Near-Infrared Applications in Biotechnology (Practical Spectroscopy)

Citation preview

Near-Infrared Applications in Biotechnology

PRACTICAL SPECTROSCOPY A SERIES

1.InfraredandRamanSpectroscopy(inthreeparts),edited by Edward G. Brarne, Jr., and Jeanette G. Grasselli 2. X-Ray Spectrometry, editedby H. K. Herglotzand L. S. Birks 3. Mass Spectrometry (in two parts), edited by Charles Merritt, Jr., and Charles N. McEwen 4. InfraredandRamanSpectroscopyofPolymers, H. W. Siesler and K. Holland-Moritz 5. NMRSpectroscopyTechniques,edited by Cecil Dybowski and Robert L. Lichter 6. InfraredMicrospectroscopy:TheoryandApplications,edited by Robert G. Messerschrnidt and Matthew A. Harthcock 7. Flow Injection Atomic Spectroscopy, edited by Jose Luis Burguera 8. MassSpectrometryofBiologicalMaterials,edited by Charles N. McEwen and Barbara S. Larsen 9. Field Desorption Mass Spectrometry, Laszlo Prokai I O . ChromatographylFourier Transform Infrared Spectroscopy and Its Applications, Robert White 11.ModernNMRTechniquesandTheirApplication in Chemistry,edited by Alexander 1. Popov and Klaas Hallenga 12. Luminescence Techniques in Chemical and Biochemical Analysis, edited by Willy R. G. Baeyens, Denis De Keukeleire, and Katherine Korkidis 13. Handbook of Near-Infrared Analysis, editedby Donald A. Burns and €mil W. Ciurczak 14. Handbook of X-Ray Spectrometry: Methods and Techniques, edited by Rene E. Van Grieken and AndzejA. Markowicz 15. Internal Reflection Spectroscopy: Theory and Applications, edited by Francis M. Mirabella, Jr. 16.MicroscopicandSpectroscopicImaging of theChemicalState,edited by Michael D. Morris 17. Mathematical Analysis of Spectral Orthogonality, JohnH. Kalivas and Patrick M. Lang 18. Laser Spectroscopy: Techniques and Applications,E. Roland Menzel 19. Practical Guide to Infrared Microspectroscopy, editedby Howard J. Hurnecki 20. Quantitative X-ray Spectrometry: Second Edition, Ron Jenkins, R. W. Gould, and Dale Gedcke 21.NMRSpectroscopyTechniques:SecondEdition,RevisedandExpanded, edited by Martha D. Bruch 22. Spectrophotometric Reactions, lrena Nerncova, Ludrnila Cermakova, and Jiri Gasparic 23.InorganicMassSpectrometry:FundamentalsandApplications,edited by Christopher M. Barshick, DouglasC. Duckworth, and DavidH. Smith 24. Infrared and Raman Spectroscopy of Biological Materials, edited by HansUlrich Grernlichand Bing Yan

25. Near-InfraredApplications in Biotechnology,edited by RameshRaghavachari

ADDITIONAL VOLUMES INPREPARATION Handbook of Near-Infrared Analysis: Second Edition, Revised and Expanded, editedby Donald A.Bums and €mil W. Ciurczak

This Page Intentionally Left Blank

Near-Infrared Applications in Biotechnology edited by Ramesh Raghavachari Promega Corporation Madison, Wisconsin

m MARCEL

D E K K E R

MARCEL DEKKER,

INC.

NEWYORK BASEL

ISBN: 0-8247-0009-0

This book is printed on acid-free paper. Headquarters Marcel Dekker, Inc. 270 Madison Avenue, New York, NY 10016 tel: 21 2-696-9000; fax: 2 12-685-4540 Eastern Hemisphere Distribution Marcel Dekker AG Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland tel: 41-61-261-8482: fax: 41-61-261-8896 World Wide Web http://www.dekker.com For more The publisher offers discounts on this book when ordered in bulk quantities. information,write to SpecialSaleslProfessionalMarketingattheheadquartersaddress above.

Copyright 0 2001 by Marcel Dekker, Inc. All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical,includingphotocopying,microfilming,andrecording, or by any information storage and retrieval system, without permission in writing from the publisher. Current printing (last digit): I 0 9 8 7 6 5 4 3 2 1 PRINTED IN THE UNITED STATES OF AMERICA

Knowledge is the true organ of sight, not the eyes.

Panchatantra

In loving memory of my father,

who L I I W ~ I ~had S simple solutions for complex problems.

This Page Intentionally Left Blank

Foreword

I trulyenjoyedreadingaboutthevariousapplicationsdescribed

in this volume. Having worked in near-infrared (NIR) applications for pharmaceuticals for twenty years, I have looked forward to the day a reference book on NIR would be published in a field something other than food, textiles, or polymers. This collection of applications to biologicalsystemsisgroundbreaking, to saythe least. The book includes chapters focusing on fluorescence: fluorescence in biological applications, the dyes themselves, immunoassays, fluorescence lifetimes, and DNA sequencing. These constitute the best compilation of NIR-fluorescence references I have seen. Forbiologicalapplications,fluorescencespectroscopy may well prove to be of greater import than absorption methods. The chapters onmedicineandsingle-moleculedetectionareuniqueintheircompleteness. Even the polymer chapter looks at the subject in a manner I have never seen. I compliment the editor on his choice of contributors and subjects. One of the problems with NIR being applied to sophisticated samples has nothing to do withthephysics or software involved; it’s where the technique originated. It may be pass6 to imply that a person was born on the “wrong side of thetracks,”butNIR hassufferedgreatlyforitsancestry. The majorityof “reputable” or “proper” techniques-mass spectrometry, nuclear magnetic resonance (NMR or MRI), electrophoresis, etc.”used in laboratory analyses are of “good breeding.” They were discovered at distinguished universities, developed by “real” instrument companies, then introduced to industry for application. Not so for NIR. Ah, poor little NIR! It was developed (as an analytical technique) at the US. Department of Agriculture, Beltsville, Maryland (largely by Karl Norris,

V

vi

Foreword

anonacademic!)forwheat,soy,andothersuch“trivial”substances.Several smallcompanieswerestarted(mostly in Maryland)forthesolepurpose of building NIR analyzers. Since the pedigree wasn’t there, NIR wasn’t taken seriously by the scientific world, in general. At NIR meetings through the 1980s, a researcher in cereals might point to a portion of the spectrum and state, “This is the protein peak.” Real sprctrosco1,ists would never refer to the combination band resulting from the carbonyl stretch and amide N-H bend in such a manner. However,asKarland“pioneers”such as Fred McClure, PhilWilliams, “Woody” Barton, and others began explaining the science behind NIR, it began to be taken seriously. Textbooks by Williams and Norris ( I ) , Osborne and Fearn ( 2 ) , and Burns and Ciurczak (3) helped codify the science. The International CouncilforNIRnowholdsconferencesaroundtheworldandTonyDavies publishes NIR News and the Jourtlul of NIR. ConferencessuchasthePittsburgh Conference and the Eastern Analytical Symposium have presentations for achievements in the near infrared. Both conferences showcase numerous workshops, short courses, and oral and poster sessions. Yet, despitethis recentwealth of legitimacy,the“art” of NIR remains a minorforce in medicine and industry. It is still nota common partofany curriculum in any college. Indeed, even with the vastly different approaches of technique and instrumentation, it is still lumped with mid- and far infrared as “part of the infrared (IR) region of the spectrum.” Of course, it doesn’t help that practitioners of the art can’t even agree on the boundaries of the NIR spectrum. Within this very book, you will find several differing opinions as to this range. Since NIR is generallyagreed to haveits genesis in themidrange IR. I have always considered the upper boundary at 2500 nm or 4000 cm”. This is where the combination bands of the C-H, N-H, and 0 - H vibrational modes begin. It is also the upper limit of lead sulfide (PbS), the most common detector used in NIR instruments. Therefore, the question of whether this limit should 3200 nm is moot. The majority of NIR instruments be 2500 nm, 3000 nm, or don’t go above 2500 nm. The definition of the lower limitis the cause of the most fireworks. Silicon (Si) is the detector of choice for the lower end of the NIR; it is also popular for the visible and ultraviolet ranges, too. Thus, the division between NIR and visiblebecomesblurred.Thereareabsorbances as lowasthe 700 nm range that can be considered higher overtones of vibrational modes in the IR. Indeed, transuraniumelementshaveeasilyexcitedf-electronsthatcanbeexcitedby 800 nm-clearly in NIR radiation. Thus, there are electronic transitions above the “true” NIR region-blurring the distinctions at this lower boundary. of Asyouread the contributions herein, you will recognize the variety algorithms used to analyze the data. This is because the spectral data obtained, especially in complex samples such as bodily fluids or skin, is rarely clear-cut. The spectra in NIR seem to disobey all the rules of Beer’s law: the analyte is

Foreword

vii

often the smallest contributor to the spectrum, there is usuallyastrong interaction between analyte and matrix, there are no isolated peaks, and any single wavelength seldom gives a linear response to the analyte absorbance. We are forced to use chemometrics (see Chapter 1 l), that is, sophisticated mathematics, unlike typical UV or visible applications. The actual choices of sample sites, number of patients, algorithms, etc., seem almost subjective in NIR analyses. Withmanychoices (all leading to usable equations) available to thereof approaches is, to searchers and no clear paths to follow, the dizzying array an outside observer or neophyte, confusing. This is how gas chromatography began; dozens of researchers making their own coatings, packing columns, and setting their own standards. In time, we had harmonization and uniform packing practices.Indeed,clearerandclearerguidelinesareemergingforNIRanalyses. In pharmaceuticals, Ritchie (4) has been proposing guidelines that combine American Society for Testing and Materials (ASTM), International Conference on Harmonization(ICH),andcurrentGoodManufacturingPractices(cGMP) recommendations and guidelines for spectroscopic methods development. With work such as that contained in this text, Near-Infrared Spectroscopy in medicine. I foreseenonintrusive isdestined to becomeanimportanttool diagnostics becoming a reality in the very near future. I also see many current analytical methods being replaced by in-process spectrometric monitors. There is a bright future for NIR in the health services sector as this text demonstrates. Emil W. Ciurczak Research Fellow, New Technology Group Purdue Pharma L.P. Ardsley, New York

REFERENCES P Williams, K Norr~s,Near Infrared Technology in the Agricultural and Food Industries, St. Paul, Minnesota: Am. Assoc. of Cereal Chemists, 1987. 2. BC Osborne, T Fearn,NearInfraredSpectroscopyinFoodAnalysis.Englewood Cliffs,Prentice-Hall,NJ:1988. 3. DA Burns,EWCiurczak,Handbook of Near-InfraredAnalysis. NewYork:Marcel Dekker, Inc.,1992. 4. GE Ritchie,presentedatSPQ'99(Spectroscopy in ProcessandQualityControl Conference), New Brunswick, New Jersey, October, 1999. I.

This Page Intentionally Left Blank

Preface

The advent of modern technology has made the impossible possible. allowing a serious view of what the unaided human eye cannot see. I refer to the nearinfrared region, defined by Sir William Herschel as “beyond the red” in 1800. The NIRregionwas first put to use by KarlNorris of the U.S. Department of Agriculture in the mid-l960s, for the spectral analysis of many substances using absorption in this region. In a similar manner, although many dyes have been discovered over the centuries, the ones fluorescing in the NIR region have becomeimportantonlywithinthepasttwodecades. The latebloomofthis field of science can be attributed to nonavailability of economical diode lasers at these wavelengths (excitation source), silicon avalanche or lead sulfide solid state detectors, and advanced information technology to analyze the generated data. Biotechnologyisaconfluenceofbiology,chemistry,physics,engineering, and information technology. Deciphering important biological information to many advances in medicine and genetics, and using biotechnology has led in general has contributed to the welfare of humanity. The combinatorial approach to drug discovery, diagnostics of diseases in its early stages, advances in genetic identity, genemapping,andgenomicmedicine is trulymovingto the forefront of biotechnology in developing prevention and cures for complex diseases. Cloning and genetic engineering of animals and plants, when put to proper use, will have great impacton medicine and will contribute to the overall well-being of humanity. This book attempts to give a glimpse of the potential applications of this NIR applications young field of science. The book is divided into two parts: based on fluorescence and NIR applications based on absorption spectroscopy. ix

X

Preface

Bothareasarecoveredfromfundamentals to applications in severalgrowing fields. In addition, there are answers to many important fundamental questions a novice might have. The controversy regarding where the near infrared starts and ends is quite visible in this book. Each author has expressed his or her opinion without any editorial changes. As an editor, I have taken a stand to define this region in the introduction and have attempted a plausible explanation. The reader should be aware of this. The first two chapters in Part A focus mainly on the aspects of fluorescence and reference to biological processes and the dyes fluorescing in the NIR region. These two chapters are an introduction to the basics of this developing field, giving some of the finer points, which many advanced books do not cover or adequately explain. The chapter on fundamentals of fluorescence is also tuned moretowardabiologicalapplication, sincethereareinnumerablebooksthat cover the basics elaborately. The first two chapters are very relevant for many applications. Each application is thoroughly explained in the following chapters with many examples. Applications in DNA sequencing, bioanalytical immunoassays and medicine are obviously relevant to biotechnology. It may appear that the chapters onsingle-moleculedetectionandapplicationsusingfluorescence lifetimes are rather theoretical; on the contrary, one can see many biotechnology in the applications in this field thatcanevolveandbecomemoreprominent future. in polymers. However, One can see that there are not many applications relevant data has set a foundation for progress in that direction. More recently, the possibility of a polymer-based near-infrared diode laser has become imminent (1). The mostexcitingchapterforanychemistisChapter 10, “Beyond Biotechnology and into Popular Technology.” In the 21st century this will become very relevant to biotechnology as well. One can envision impregnation of near-infrared sensors into the chips that will provide vision for the blind and even expansion of their vision in the near-infrared region. This is a science in its growing stages-as one article calls it, a “teen-age science.” The resurgence in this field of science is felt in all application areas as envisaged by any other fluorescence and absorbance application. Fluorescence Spectroscopy: New A4etlmi.s urd Appliccltions, edited by Otto Wolfbeis (2) gives an idea of the possible applications that can emerge from the field ofvisible fluorescence. One day it may be a reality that all those applications would extend into the near-infrared region as well. The second part of this volume projects the possible applications based on near-infrared absorption spectrometry. The first chapter in Part B describes the fundamentals of this science and tells the reader what to expect. The next two chapters cover the applications in medicineandpharmaceuticals,thatarethe most relevant to biotechnology. The chapter on biomedical applications of NIR

Preface

xi

spectroscopydealswiththediagnosticaspectsofmedicine.Simplifyingand making diagnostics of some diseases using blood, urine, and other noninvasive techniques would make diagnostics easier as well as saving time andmoney. The chapter on pharmaceutical applications is an abridged versionof the forthcoming book edited by Ciurczak and Drennen (3). This comprehensive chapter explains applications starting from raw materials, in-process applications, and the finished productwithseveralexamples.Chapters I I , 12, and 13 wereadded to make this book a comprehensive text so rich in NIR studies that each topic by itself could have been made into a separate book. There are many textbooks that are dedicated to the science of NIR spectrometry and they are well referenced in these chapters. The aim of this book is to give the reader an overview of this young field of science and provide the possibilities of further exploration. This text should give a head start in keeping up with the current research.

ACKNOWLEDGMENTS A project like this involves many people without whom this task would have been a very difficult one. First and foremost, I would like to thank the contributors fortheircontributionsandtheirgreatenthusiasm in spite of theirverybusy schedules. I would also like to thank Dr. Vish Bhadti of Amersham-Pharmacia Biotech, Dr. Dan Simpson, and Dr. Patricia Fulmer of Promega Corporation, and Dr. Judy Schanke of Epicentre Technologies for their comments and opinions on some of the chapters in this book. Especially in an edited volume like this, it isextremely difficult to coordinate every aspect of thebookfrom all over the world. I thank the staff members of Marcel Dekker, Inc. who were involved with this project. Last but not least I thank my family for putting up with me during this time.

REFERENCES 1. http://perl.spie.org/cgi-bin/news.pl?id= I47 1. 2. Otto S. Wolfbeis,FluorescenceSpectroscopy,Springer-Verlag, New York, 1992. 3. E. Ciurzcak and J. Drennen. Pharmaceutical Applications in Near Infrared Spectrornetry, Marcel Dekker ( i n press).

This Page Intentionally Left Blank

Contents

Foreword by Ernil W. Ciurczak Prcf ’ ace Contributors

1’

ix Xv

Part A: Applications Based on Near-Infrared Fluorescence 1. Introduction Rarnesh Raghavachari 2.

3.

Some Aspects of Fluorescence of Particular Relevance for Biological Processes Lawrence J. Parkhurst Probes: Dyes Fluorescing in the NIR Region Stefan Stovanov

1

5 35

4. Analytical Applications of Near-Infrared Fluorescence in Immunoassays Anand R. Swam?, Lucjan Strekowski, and Gabor Patoncly 5.

Near-Infrared Applications in DNA Sequencing and Analysis Ronald J. Meis and Ranzesh Raghavachari

95 133

6. Near-Infrared Applications in Medicine Datluta Fragkowiak, Alfons Planner; and Krpsztof Wiktorowic::

15 1

7. Near-Infrared Applications in Single-Molecule Detection Alonso

I85 xiii

Contents

xiv

8. ApplicationsUsingFluorescenceLifetimes Jumes H. Flcrnagan, J K and Benjarnin L. Legendre, JI:

20 1

9. Applications in Polymers Olga V Przhonsku

235

IO.

BeyondBiotechnologyandinto Mmaru Matmoka

Popular Technology 263

Part B: Applications Based on Near-Infrared Absorbance 1 1.

Fundamentals of Near-InfraredSpectroscopy Howcrrd Murk

293

12. BiomedicalApplications of Near-Infrared Spectroscopy Enlil M! Ciurczak

323

13. PharmaceuticalApplications of Near-Infrared Spectroscopy Elllil M! ciuru.uk and James K. Drennerz

341

Iilcl e..r

365

Contributors

Alonso Castro, Ph.D. Staff Scientist, Physics Division, Los Alamos National Laboratory, Los Alamos, New Mexico Emil W. Ciurczak,Ph.D.

Purdue Pharma L.

P., Ardsley,New York

James K. Drennen,Ph.D.

AssociateProfessor of Pharmaceutics,School Pharmacy, Duquesne University, Pittsburgh, Pennsylvania

of

James H. Flanagan, Jr.,Ph.D.

ResearchScientist,AdvancedTechnologies and Development, Transgenomic, Inc., Omaha, Nebraska

DanutaFraqkowiak,Ph.D. (habilitation) Professor,Institute Poznan University of Technology, Poznan, Poland

of Physics,

Benjamin L. Legendre, Jr., Ph.D. Research Scientist, Advanced Technologies and Development, Transgenomic, Inc., Omaha, Nebraska

Howard Mark, Ph.D. President, Mark Electronics, Suffern, New

York

Masaru Matsuoka, Ph.D. Professor, Laboratory of Materials Science, Kyoto Women’s University, Imakumano, Higashiyama-ku, Kyoto, Japan Ronald J. Meis, Ph.D. Senior Research Scientist, Research Department, Epicentre Technologies, Madison, Wisconsin xv

xvi

Contributors

Lawrence J. Parkhurst Ph.D. Professor and Chair. Department of Chemistry, University of Nebraska, Lincoln, Nebraska

Gabor Patonay,Ph.D.

Professor,DepartmentofChemistry,GeorgiaState University, Atlanta, Georgia

AlfonsPlanner,Ph.D. (habilitation) Lecturer,InstituteofPhysics,Poznan University of Technology, Poznan, Poland Olga V. Przhonska, Ph.D. SeniorResearchScientist,Department of Photoactivity, Institute of Physics, National Academy of Sciences of Ukraine, Kiev, Ukraine

Ramesh Raghavachari, Ph.D.* Senior Scientist, Promega Corporation, Madison. Wisconsin

Stefan Stoyanov, Ph.D.,Dr.Sci. AssociateProfessor,DepartmentofChemistry, University of Sofia, Sofia, Bulgaria Lucjan Strekowski, Ph.D. Professor, Department of Chemistry, Georgia State University, Atlanta, Georgia

Anand R. Swamy,Ph.D.

PostdoctoralResearchAssociate,Department Chemistry, Georgia State University, Atlanta, Georgia

of

Krzysztof Wiktorowicz, Ph.D. (habilitation) Professor, K. Marcinkowski University of Medical Sciences, Poznan, Poland

*Cwrc.rtf c%filitrtion: Quality Systems Manager, Corning Microarray Technology, Corning.Ncw York.

Near-Infrared Applications in Biotechnology

This Page Intentionally Left Blank

Introduction Ramesh Raghavachari* Prornega Corporation, Madison, Wisconsin

1.

DEFINITION OF NEARINFRARED

The ultraviolet and visible spans of the electromagnetic spectrum have been the wavelengths most often utilized for bioanalytical techniques for the past several decades.Wavelengthsfrom190to 650 nm arethosethatmostbiomolecules absorb and fluoresce. The regionbeyond 650 nm has more recently captured the attention of many pioneers in the field of bioanalytical applications. This region, called the near infrared (NIR) which interfaces the visible and infrared portions of the electromagnetic spectrum, is gaining rapid importance in the area of biotechnology. In this book we define the “near-infrared’ as constituting wavelengths of 650-2500 nm. This region is often further divided into the deep red or far red (650-730 nm) and the near infrared (730-2500 nm). This differentiation arises [1-4], which has three types of visual from the physiology of the human eye photoreceptor cones that have different spectral responses.The response of these three types of visual cones peak at 445 nm (blue cone), 535 nm (green cone), and 570 nm (yellow cone) (Fig. I ) . The combinations of the responses of these cones represent various colors that are seen by the human eye. The yellow cone eye cannot see. In response extends up to 730 nm,beyondwhichthehuman principle, the near-infrared region starts from where the human eye has no visual response. However, with wavelengths of 650 nm and higher, the response of the humaneye is so lowthatthis tail end of thevisible spectrum is includedas part of the near-infrared region. More often the span of wavelengths between 650 and 730 nm is called the far-red or deep red region. Some experts further *Curr.c.nt @ k r t i m ;

Corning Microarray Technology. Corning.

New York.

1

2

.-P

Raghavachari

0.4

400

450

500

550

600 700650

Nanometers Figure 1 Thewavelengthresponsefactorsofthethreetypesofvisualphotoreceptor cones of the human eye.

divide the NIR into two ranges 650-1 100 nm and 1100-2500 nm based on the different detectors that solid-state technology provides for detection in several types of instrumentation. For example, silicon avalanche diode detectors have a range from 650 to 1 100 nm, and lead sulfide based detectors from 1100 to 2500 nm. Many schools of thought are represented in the literature [S-71 (see also Stoyanov, Chapter 3, this volume), suggesting various ranges starting from 650 nm where the near infrared begins and ends. For all practical purposes, this book will use this range for the near-infrared region based on both fluorescenceand absorption-related applications (Fig. 2).

Figure 2

Position o f NIR regionintheelectromagneticspectrum.

3

Introduction

Visible Fluorophorcs NIR Dyes 11111l111l11111111l1llllllllllllllllllllll

Biomolecules

600

800

Figure 3

1000 nm

Absorbance and fluorescence of biotnolecules in relation to visible and fluorophores.

NIR

II. WHY NEAR INFRARED? Mostbiologicalsubstancesandimpurities in bioprocessesabsorband fluorescebetween 190 and 6.50 nm (Fig. 3). Theirrelativesensitivity to detect biomolecules at thesewavelengthsishighlyreduceddue to highbackground caused by the molecules themselves (autofluorescence) or the impurities within the ultraviolet-visible (UV-Vis) region. The near-infrared wavelength region, beingaway fromtheseinterferences,provideshighsensitivityfordetectionof biomoleculesandhencebettersignal-to-noiseratio.Largesamplesizes(analytes) are required for a high background in order to maintain the accuracy and unambiguous detection level of signal to noise even if the absorptivity or quantum yields are high. In contrast, low background enables minimum sample sizes and high detectability even if the quantum yields are low [SI. The near infrared region provides these advantages. Time and again researchers have observed that many dyes absorbing and fluorescing in the near infrared do not have a high fluorescence quantum yield like their visible counterparts, especially the rhodamine and fluorescein family of dyes, which exhibit very high quantum yields. In spite of the low quantum yields, the sensitivity exhibited by near-infrared dyes is far superior to that of the visible dyes. The detectability in thisregion is on the order of afewmolecules(about IS atomoles,mol),whereasthevisible counterparts are 400-800-fold less sensitive despite their high quantum yields [8.9]. This consideration is based on the electronics and detector technology that are currently available for both kinds of systems.

111.

ABSORPTION VS. FLUORESCENCE

This book deals with both absorption-based and fluorescence-based applications that are being explored or that are already available as technologies. Each has its advantages and disadvantages. However, it is a well-known fact that fluorescence is far more sensitive than absorption. A minimum sample for which fluorescence spectra can be measured may not be sufficient for the measurement of its ab-

Raghavachari

4

sorptionspectra. In almost every chapter in Part A of thisbooktheauthors express that the advantage of the near infrared is that most biological substances have minimal absorption in this region and hence greater sensitivity. In simple terms, absorption in the near-infrared is observed as overtones. Exploitation of thesefunctionswiththeuse of modernalgorithmshasmadepossible all the applications mentioned in Part B of this book. The fundamental principles are embedded in the chapter by Mark in Part B. This has been possible only due to the advances in both electronics and solid-state physics along with the software algorithms that are at the center of today’s technology. In dealing with the importance of both absorption-based and fluorescencein biotechnology,thisvolume basedapplications in thenearinfraredregion gives the reader a tip of the iceberg of this young field of science. Many areas ofapplicationsare still in thedevelopmentalphase, so not enoughdataare available in the public domain for their inclusion here.

REFERENCES 1.

2. 3. 4. 5.

6. 7.

8. 9.

H Davson, ed. The Eye, Vol. 1b, Vegetative Physiology and Biochemistry. New York: Academic Press, 1984. H Davson. Physiology of the Eye. 5th ed. New York: Pergamon Press, 1990. http://ucrwcu.rwc.uc.edu/koehler/biophys/6d.html http://w3.ime.netlO/o7Ephillips/iknow~pages/humanvision/lightllight.html G Patonay. Near-infrared fluorescence: An emerging new method. In: G Patonay, ed. I . Greenwich, C T JAI Press, 1993, AdvancesinNear-InfraredMeasurements,Vol. pp 114-138. M Matsuoka, ed. Infrared Absorbing Dyes. New York: Plenum Press, 1990. JFabian, H Nakazumi, M Matsuoka.Nearinfraredabsorbingdyes.ChemRev 92:1197-1226, 1992. dyes for N Narayanan, G Little, RRaghavachari, G Patonay.Newnearinfrared 23885-IS, 1995. applications in bioanalytical methods. Proc SPIE N Narayanan, G Little, R Raghavachari, J Gibson, A Lugade, C Prescott, K Reiman, S Roemer, D Steffens, S Sutter, D Draney. New NIR dyes: Synthesis, spectral properties and applications in DNA analysis. In: S Daehne, U Resch-Genger, OS Wolf Beis, eds. Near-Infrared Dyes for High Technology Applications. Amsterdam: Kluwer Academic, 1998,pp 141-158.

Some Aspects of Fluorescence of Particular Relevance for Biological Processes Lawrence J. Parkhurst University of Nebraska, Lincoln, Nebraska

1.

BRIEF OVERVIEW AND HISTORY

Sincethereareanumber of excellenttreatments of fluorescence [I-61, the purpose of this chapter is to present a concise overview of the history and some of the fundamentals of fluorescence and then to treat several topics pertinent to an understanding of a few recent applications of fluorescence to problems of biological interest. One of the most important topics, fluorescence polarization or anisotropy, will, because of space limitations, not be covered. I base my discussion on the familiar Jablonski diagram and use simple kinetic representations of the relevant absorbance and fluorescence processes, recognizing that this simple approach may not be adequate for all descriptions. in particular for results from certain laser experiments [7]. The Jablonski diagram (Fig. 1) is drawn in the traditional, though not original, manner for ground-state singlet (So) and excited state (SI), with the lowest triplet state ( T I )drawn to the right, but where no correlation of this displacement with increased “bond length” is implied. In the diagram, transitions among various vibrational levels of a given electronic state are internal conversion (IC) processes (with an average ) connect these levels to the lowest vibrational level of that rate constant k ~ that state, the common assumption being that these rate constants are all very large with respect to the intrinsic rate constant kf for fluorescence. In the diagram, all nonradiative internal conversion processes are simply denoted kl for simplicity, with no implication they are all equal in rate. For convenience, in the absence

5

6

Parkhurst k,’

Figure 1 StandardJablonskidiagram for a three-state system (see text). A represents the allowed absorbance process; k l , k ; , and k r areinternal conversion processes; k~sc and kist denote intersystem crossing; and kf and k,, respectively. denote fluorescence and phosphorescence rate constants.

of delayedfluorescence,therateconstants for all nonradiativestepsleading , 7 - 1 forfluorescence is k~ k,. t is away from SI canbe summed as k ~ so the lifetime of S I , equal to the l/e time (the time for the intensity to decrease to 0.368 of its initial value) for the decay of the intensity of fluorescence (IF) followinga6-shapedexcitation,where IF(f) = d ( h u ~ ) / d tThe . nonradiative intersystem crossing rate constant klsc connects So and T I , and kist connects T I and SI.The rate constant k , is the intrinsic rate constant for phosphorescence. In some instances, e.g., for eosin[SI, delayed fluorescence requires the additional consideration of reverse intersystem crossing from TI to SI. The diagram can be readily augmented to allow for higher singlet excited states as well as triplet states and for ISC processes between these various S and T states. It has long been known [9] that resonance fluorescence from high vibrational levels in the excited states appears at very low pressures in the gas phase, whereas for other molecules, such as anthracene [lo], fluorescence occurs only from the lowest vibrational level evenatverylowpressures,and in solution IC within SI will in generalbeverymuchfasterthanfluorescence,resulting in aBoltzmann distribution for the population of vibrational levels of S I . In Jablonski’s paper [ 1 I], the two excited states denoted here as SI and T I were F and M , whichgiverise to fluorescenceandphosphorescence,respectively,and N wasthegroundstate.LewisandKasha [ 121 proposedthat M , fororganicmolecules,wasa triplet state, and theparamagnetismofthat state was confirmed by Lewis et al. [ 131. I n the early literature, “fluorescence” and “phosphorescence” were often used differently from the usage above. For instance, Pringsheim [ 141 termed the M + N process “slow fluorescence” and thedelayedemission M -+ F + N “phosphorescence.”Perrin 1151 proposed an operational definition, whereby fluorescence has a lifetime more or less independent of temperature,whereasthelifetime of phosphorescencetends to increase as temperature decreases (owing to decreased collisional deactivation).

+

gical

and Fluorescence

7

Leverenz [ 161 proposed a different operational definition, where fluorescence had a lifetime of about 10 nsec and phosphorescence (which would have included in this definition delayed fluorescence) had a much longer lifetime. Since that time, many examples of fluorescence lifetimes even longer than 100 nsec have been reported [ 171. Owing to the success of quantum chemical calculations of transitionmomentsandstateenergies,theLewis-Kashainterpretation is the customary interpretation for emission phenomena. Difficulties, however, in accounting for the relatively high rate of ISC (which is formally a spin-forbidden process but often competes with fluorescence and thus must occur in times on the order of 1-10 nsec) delayed acceptance of that interpretation, even with the paramagnetic evidence for assignment of TI [ 131. It has already been pointed out that ISC can compete with fluorescence of phosphorescence at processes in manymoleculessincethequantumyield low temperature or in the absence of quenchers (heavy atoms or paramagnetic species) is often within an order of magnitude the same as the quantum yield for fluorescence. In fact, under conditions where radiationless conversion of TI to the ground state So is negligible compared to phosphorescence, the ratio of quantum yields of phosphorescence to that of fluorescenceis klsc/kf [18]. In the absence of oxygen, the phosphorescence lifetime of small aromatic hydrocarbons can be expected to be on the order of 10’ sec or greater, based on studies of benzene [191. Spin quantization can be broken by paramagnetic species and by the heavy atom effect through spin-orbit coupling. In the first instance, if a doublet molecule or ion A approaches sufficiently close to a singlet ground-state molecule, the combined system will be a doublet; if A remains near when the “triplet state T I ” of the molecule is formed, the combined system will consist of a doublet and a singlet, so the perturbed triplet (what was, in the absence of A, T I )now has some singlet character and radiative and nonradiative transitions between the former TI and singlet states can occur. The result is the same for anyparamagneticion.Thiseffecthasallowedabsorptionspectrafrom So to TI to bemeasured[20],usingoxygenand NO asperturbants. The effects of paramagnetic ions on ISC and phosphorescence lifetimes have been known for over 50 years [21]. Charge transfer states of oxygen have also been implicated in the quenching mechanism [22]. In more recent times, changes in fluorescence and phosphorescence lifetimes have been used to determine oxygen concentrations [23,24] within cells and to demonstrate the ease with which oxygen can penetrate and diffuse through proteins [25]. In order to obtain the correct expression for spin-orbit coupling in an atom, a relativistic Hamiltonian is required. To the extent that spin-orbit coupling becomes important, singlet and triplet states are no longer pure states; for instance, the triplet excited state will be mixed with a nearby excited singlet. The matrix ( z ) of element that connects the two states will depend on the atomic number the atom, and if the potential is a Coulomb potential, the dependence varies as

a

Parkhurst

2'. The extent of mixing of the two states will also depend on the inverse of the energy difference between the two states( A E ) , and thus excited singlets near T I would be expected to contribute more strongly than would So to this coupling. The rate of the transition (both radiative and nonradiative) between T I and SI or So would then be expected to depend on the square of the off-diagonal matrix element, z X / ( A E ) ' [26]. McClure [27] extended the analysis to molecules (using a perturbed Hamiltonian based on a sum over the atoms in the molecule) in a study of beta-chloro-, bromo-, and iodonaphthalene, where the intensity of the singlet-triplet absorption band increased greatly in the above series. Kasha as [28] discovered the external heavy-atom effect. in which using ethyl iodide a heavy atom solvent was nearly as effective for enhancing S --f T processes as having the heavy atom covalently bonded in naphthalene. The series of dyes rhodamine,eosin(tetrabromorhodamine),anderythrosin(tetraiodorhodamine) show the expected decrease in fluorescence and increase in ISC. Both external heavy atom and paramagnetic effects were found in rare earth perturbants [21]. Spin-orbit interactions may also play a role [26] (in addition to the long radiative lifetime that derives from the small transition moment, see below) in the negligible fluorescence observed in heterocycles in which the IZ +. IT* transition is the lowest allowed electronic transition. Under certain conditions, additional rate constants must be added to Figure 1 to reflect, for instance, long-range resonance energy transfer or collisional quenching, both of which are bimolecular processes and are represented, respectively, by kt and k g . In the most useful formulation, however, kt is a first-order rate constant for energy transfer between a donor-acceptor pair separated by a distance R , and thus the concentration dependence disappears; k g is a secondorder rate constant for dynamic quenching of SI by quencher Q. In principle, if the electronic and vibrational wave functions are known for the given state, and the perturbations that result in the various transitions are also known, then the rate constants can be calculated. In practice, the transition momentsaremostoftencalculatedfromintegratedintensities [29], andthe Einstein coefficients are used to calculate intrinsic fluorescence lifetimes (st!) or the reciprocal, kr. The Einstein coefficients A and B are based on simple rate equations and on thermodynamics, and these phenomenological considerations of the relevant rate constants. do not provide a prescription for the calculation The matter of spontaneous emission deserves a comment, since, for the usual time-independentHamiltonian i n the absence of theradiation field, thewave function for the excited electronic state should represent a stationary state and decay would be impossible! This dilemma was not resolved until Dirac [30], in what is generally considered the first paper in quantum electrodynamics, showed howquantization of theradiation field andthemolecule as a singlesystem provided explicit formulas for calculating the A and B coefficients: from the A coefficient, s!, can be calculated. When the measured sf is less than this value, processes such as kl and kIsc are invoked.

ogical

and Fluorescence

9

Thus, to summarizethephysics, let Wo and Wl bevibronicwavefunctions for the ground and excited states (i.e., Wo = WoeWout, where u’ denotes a vibrationalquantumnumberforthegroundelectronicstate). In thedipole moment formulation of transition moments, ji = ,ii, ,iinUc is the sum over the coordinates of the electrons and nuclei in the molecule, multiplied by the charge of the given nucleus or electron. We can then write ,!io1 as

+

J

where the term in parentheses is an integral over the electronic wave functions and coordinates, the outer integral is a weighted overlap integral over the vibrational wave functions, and the last integral is over spin coordinates and gives rise transitions, the latter term is I , which to spin selection rules. For singlet-singlet we assume here. The quantity(M.,(R)) is a mean value averaged over the nuclear coordinatessuch that ,Go] = ( M , ( R ) ) W:,Wu,, dR. Since,as willbe shown, the intensity of an absorption transition is proportional to the square of the tranW:,WL,,, dR is the intensity attributed sition moment, the square of the term to a given vibrational component of the electronic transition, and the squared quantity is the Franck-Condon factor. It can readily be shown [31] by a completeness argument that the sum of the squaresof the second term (when summed over all vibrational levels of the upper state, starting with the lowest vibrational level of thegroundstate) is equal to 1 . Theoretically,one is thenleftwith the problem of integrating-over the electronic w a y functions for the groundstate geometry to obtain ( M e ( R ) ) The . quantity I(M,(R))I = I,iioI I is obtained from an integrated intensity, since that is equivalent to summing over all values of u” starting from u’. The Einstein coefficient for spontaneous emission from state 2 (SI)to state 1 (the ground state) is given by

s

where u is the frequency of the transition, h is Planck’s constant, c is the speed of light,and EO is thepermittivity of vacuum[8.854 x C2/(N . t n ’ ) ] . Let 141 be the length of the transition dipole in angstroms, then

and for a transition dipole length of 2 A and h = 5000 A. one obtains rlj) = 4.32 nsec.Forcomplexdyemoleculesthetransitionmomentcanbeobtained to higher precision from measured integrated intensity data than from a quantum mechanical calculation. From Beer’s law, In(lo/Z) = kc’l’ (c’ in moles per cubic

Parkhurst

10

meter, I’ in meters, k in square meters per mole). Carrying out the integration on the frequency axis (assuming randomly oriented molecules)

where N A is Avogadro’s number. Another useful relation is

s

cd In h = - 1.090 x 10“~(q1*,

in cgsunits

( h in cm, 6 in M-’ cm-I). Anapproximaterelationship is 10” x 0.917 x ~ ~ ~ ~ ~ A h = 1 /Iq(A)I’, 2 / h ~where , ~ ~Ah112 is the full width at half peak height oftheabsorptionbandhavingapeak at.,,,A Thus, if = IO’. ,,A = 500 nm, Ahll7 = 20 nm, 14 (A)[ = 1.92 A, which, from Eq. (2) and the subsequent ) nsec. discussion, gives rfo = 1.886 x 1 0 - ’ ( ~ p e a ~ A h ~ , ~ / h , n=a x4.72

II. MOLECULAR COMPLEXES AND EXCITONIC INTERACTIONS For some molecules, particularly derivatives of pyrene, it has long been known [32] that two very different emission spectra are observed depending on concentration. At low concentrations a structured spectrum is generally seen, whereas at higherconcentrationsa new spectrum,verymuchred-shiftedandwithout structure, is observed.Theinterpretation is that thesecondspectrumderives from emission of an excimer to a dissociative ground state, the excimer being an excited state of a dimerof pyrene. The excited state dimer was shown to have considerable ionic structure [33-351, in contrast to a simple excitonic dimer. In the pyrene crystal, which shows structured “monomer” absorption but “dimer” emission 1331, the pyrenes are oriented with the molecular planes parallel with the molecules separated by 3.5 1261. Such excimer spectra have been used to gain insight into interactions in membranes [36,37]. Somewhat similar, from a structural point of view, are dimers (and higher aggregates) in which excitonic interactions give rise to a band of excited states, split bytheexcitonicinteractionsandwithexcitationdelocalized[38-44]. In simpledyemolecules, thesinglet-singlettransition momentslie withinthe molecular planes. The canonical transition moment orientations and corresponding molecular stacking geometries are “side to side,” as in planar dye molecules such as porphyrins and rhodamine, and “head to tail,” as in polyenes or cyanines (Fig. 2). (These canonical as well as oblique orientations for dimers and higher aggregates have been discussed by Kasha and colleagues [41”wJ.) The very simplest interpretation is as follows. In the side-to-side dimer case, the inphase exciton state gives rise to the allowed transition, and this is blue-shifted

A

11

Fluorescence and Biological Processes

(a)

(b)

monomer

dimer

monomer

dimer

Figure 2 Canonicaltransitionmomentorientations for thedimerexcitonsplitting modelshowingthephases of thetransitionmomentsforallowedandforbiddentransitions.Solidarrowsshowallowedtransitions; wavy lines show radiationlesstransitions. (a) Side-to-sidedimerorientation.Fluorescencefromtheupperdimerexciton state, though allowed, often appears not to compete well with radiationless decay to the out-of-phase lower exciton state from which fluorescence is forbidden. (b) Head-to-tail dimer orientation. Absorption and fluorescence are allowed to the lower energy exciton state but not to the higher, resulting in red shifted spectra with respect to the monomer absorption and fluorescence spectra.

with respect to the monomer. (In a higher aggregate, a band of electronic states would result.) The lower lying exciton state, reached by internal conversion, is a state from which fluorescence to the ground state is not allowed. In accord with this picture, porphyrins are fluorescent when in the monomeric state but not at higher concentrations where aggregation occurs [45], though blue-shifted absorption appears at the higher concentrations. Similar effects are found with rhodamine and other dyes [46,47]. Such effects have been seen when clusters of lysines in proteins have been labeled with rhodamine and the spectral shifts and fluorescence quenching have been used to follow protein associations and conformational changes [48,49]. Dye aggregation effects have recently been used to detect enzyme products of gene expression [50].On the other hand, when linear molecules of the polyene type associate end to end, the excitonic interactions give rise to two excited states with the upper state forbidden and the lower state allowed for electronic transitions. The prediction is that such an aggregate will show a red-shifted absorption spectrum, and fluorescence, if allowed in the monomer, should appear in the aggregate as well [51]. An “internal” excitonic interaction model was proposed to account for the spectraof the linear polyenes [52,53], treating the molecules as an assembly of ethylenes. Polarized specular reflectionmeasurements in thecrystalshowed that thetransitionmoment of was not in accord with that theory, however, but rather with the predictions

Parkhurst

12

simple LCAO-MO theory, and axis [54].

A.

&

was shown to be aligned along the molecular

Fluorescence Resonance EnergyTransfer

Forster or fluorescence resonance energy transfer (FRET) is an exciton-like very weak coupling, generally between unlike molecules, that can result in energy transfer over distances on the order of 50-80 8,. The process has been discussed and reviewed many times [55-601, and only the key points are dealt with here. The great interest in FRET measurements derives from the fact that they can be made on very dilute solutions (even a single molecule [61]), in small volumes of afewcubicmicrometers, in timesontheorder of afewnanoseconds to minutes(allowingawiderangeofkineticprocessestobeinvestigated),and with a distance resolution on the order of 1 8,. Let kt represent the rate constant for electronic energy transfer between a donor molecule and an acceptor separated by a distance R [the distance between the centers of the transition dipoles, each of which can be on the order of 1 8, in length for a strongly allowed transition; see Eq. (4)]. The transfer requires that the emission spectrum of the donor overlap with the absorption spectrum of the acceptor. Energy transfer resultsin a decrease in donor steady-state fluorescence, which derives from a decrease in the lifetime of the donor excited state. The acceptor may be nonfluorescent (a quencher), but if it is fluorescent there will be an increase in the steady-state intensity of the acceptor and a complex time dependence of theacceptorfluorescencereflectingthetransferprocess.The expression for kt is [57]

where A = (9.79 x

for R in angstroms,

5:: = 1 / kf, tz is the refractive indexof the medium separating donor and acceptor moieties over the wavelength region of J , and K is the orientational part of the dipole-dipole operator COS(@DA) - 3 COS(@D) COS(@A). where @DA is the angle between donor and acceptor transition dipoles and @D and @A are, respectively, the angles between donor and acceptor transition dipoles with the interdipole vector of length R . J , the spectroscopic overlap integral, is given by

(6) where f ~ ( k )istheemissionintensity of the donor at wavelength h withthe emission normalized to 1 when integrated over the entire emission spectrum and

FluorescenceProcesses and Biological

13

is the molar absorptivity of the acceptor at wavelength h ( E in units of M” cm”, in cm). in terms of acritical Expression ( 5 ) is usuallymodified tobewritten transfer distance ( R o ) at which the probability of transfer equals that of decay by all other paths, i.e., k , = k f kl = 5;’ at R = Ro. Note, however, that this relation implies a specific value for kr, and thus Ro will depend on the particular reference state one chooses, one that has a corresponding lifetime T;. Thus, the correct relationship is CA

+

to thatparticularlifetime of the where R,* is thecriticaldistancereferenced donor. To state the matter somewhat differently, kl can be considered to include dynamic quenching ( ~ Q Q which ). has been used to decrease Ro [ 6 2 ] ,but such quenchers cannot affect the value of kt. Thus. in the expression for kt, 51) and Ro must be linked and not regarded as independent variables. Expression (5) shows that kt can be calculated if one knows the average value for K ~ a , quantity that can vary from 0 to 4 but has an average value of 213 for random orientations of donor and acceptor transition dipoles. The probability distribution function for K’, however, is quite unusual in that the most probable in probability and a leveling off toward value is 0, with a very rapid decrease a plateau at K’ = I followed by a slow decay toward zero, reached at K’ = 4 (see Fig. 3, curve A). Inpractice, one mustusechemicalinsightorhaveinformationonthe flexibility of the probe linkages to justify a value of 2/3 in the expression for kt. Unless the donor and acceptor are rigidly fixed to a molecular framework that is alsorigid,amodestextent of orientational flexibility willoftengive R dependsonly on avalue for K’ not far from 2/3. Owingtothefactthat theinversesixthroot of theaveragevalue of K ~ R, willnotbeparticularly , if one attaches the fluorescent probesby sensitive, in practice, to K ~ particularly flexible tethers to the macromolecule. That matter has been considered in some detail [63.64] and will be elaborated on further after the discussion of distance distributions.Theimportance ofunderstandingthe“kappa-squaredproblem” cannot be overemphasized, however, because of the a priori uncertainty in the possibility of extracting valid distances from FRET.

B. Steady-StateMeasurementsand

P(R)

Whereas having dyes coupled by flexible tethers may in practice circumvent the K~ problem, having flexible tethers necessarily introduces uncertainty in the distance R , since one must deal with a distribution of distances. Before discussing

14

Parkhurst

6

5

4 W

0

'X3

a 2

1

0

0

1

2

3

4

K2 . a: Randomorientation of fixed Figure 3 Probabilitydensityfunctionsfor K ~ Curve dipoles. Curve B: Random orientation of donor and acceptor cone axes for cones with half-apicalangle of 60'. Thecurvesweregenerated by MonteCarlosimulation with 4 x 10' points (curve A) and 2 x 10' points (curve B). Note that the x-axis intercepts for curve B at K~ = 0.41 and 1.438 correspond to structures A and C, respectively, of Figure 7.

that matter, let us examine the simple steady-state treatment for determining R . Let p equalthefluorescenceintensityofthedonoralonedividedby that of thedonor in thepresenceoftheacceptor.Thiscanreadilybeshowntobe 1 ktrD, where the lifetime is that for the particular environment of the donor in the experiment, and we assume only a single lifetime in that environment. We also assume that, aside from transfer, the environment of the donor remains unchanged in the presence of the acceptor (see Ref. 65 for a treatment of an optical cycle that links lifetimes and steady-state intensities with corrections for static quenching).

+

15

Fluorescence and Biological Processes

Rearranging the last equation and substituting for

kt gives

One usually assumes that rD = rg. A more defensible procedure is to determine the relative quantum yields (QY) for the donor in the reference state and in the particular experiment and adjust thet ’ s accordingly [i.e., QY*/QY(experiment) = rg/tD]. This, however, assumes thatno change in staticquenchingoccurs. Therefore, even in the simplest case of a fixed distance, steady-state data alone cannot reliably give the correct distance R without lifetime information, because theproperlifetime is unknown! In manyinstances,thedonoralonedoesnot decay as a single component but as at least two components or as a distribution of lifetimes. A simple example willmakethepoint. Suppose the donor actually has two components with lifetimes of 6 nsec and 0.444 nsec and in respective mole fractions of 0.36and0.64andthatthepresumedsingle-componentlifetime ( r D = tg) is 4 nsec (which gives the same quantum yield as for the biphasic decay). If p = 1.5, thenthecalculationfor R assuming only one component will give 67 A, whereas the true distance should be 71 A. This alone may be a significant error in a given application, but when coupled with the problem of distance distributions, the error may be further increased. Thus, to summarize theissue,expression (7) or anyvariantthereofused to extract R from only steady-state data suffers from several problems in addition to the well-known K~ problem:

1. The TD’S in Eq. (7) must be assumed to be equal. 2. One must assume that there is only one donor lifetime for the donor3.

only case. R yet alsoassumecomplete One must assumeasingledistancefor rotational averaging for K~ in the calculation of Ro.

The latter point is almost logically inconsistent, and this has caused several investigators to explore means for extracting distance distributions. The proposal [66] to use steady-state data from molecules having various fluorophores at one to anyextent. The proposaltouse site of attachmenthasnotbeenpursued quenching data to obtain a variety of Ro values [62] appears to work in some cases, but in my own experience, perhaps owing to multiple donor lifetimes or to perturbations to the system itself from iodide, the method led to physically unreasonablevaluesfordistances.This,togetherwithrapidimprovements in lifetime instrumentation, has led to the use of lifetime data to extract distance

Parkhurst

16

distributions from FRET measurements [66-701. Note that the correct expression, based on the discussion above, for the time domain is given in Ref. 68:

(YD; is the mole fraction of donor species having a lifetime rD;. In itsusualinterpretation,adistancedistributionisonethatrepresents distances that are fixed during the lifetime of the donor so the donor time decay results from a simple integration over a distance distribution. If, however, there is very rapid translational motion (but motion that is limited by the tethers), or rotation that accompanies a change in distance, then a single weighted average distancewillbemeasuredoverthatrange of motion[68];formotion in the intermediate time regime, some narrowingof the true distance distribution would be expected. Note that a distribution of distances between donor and acceptor will, even for a single donor-only lifetime,lead to a distribution of lifetimes (see Fig. 4). I D ( f ) . however, can generally be resolved into only two or at most three decays, which must be considered therefore only to approximate a continuous distribution. Figure 4 shows corresponding distributions in distance and lifetime. The curves are labeled to show how points at fla and f 2 a in P ( R ) lead to corresponding points in P ( s ) . At large R , the corresponding r must be that of the donor alone, here assigned the valueof 4 nsec for the calculations for Fig. 4. For two flexible tethers coupled to the ends of a rigid molecule and with individual distributions that do not overlap each other, the overall inter-dye distancedistributionwillbeverywellapproximatedbyaGaussiandistribution (Fig. 5). Assume a donor-only single lifetime of 4 nsec and a value for Ro of 60 can be most precisely determined in a distance range from about 0.67 Ro to 1.2 Ro. Figure 6 shows how distributions with three different values and varying widths (a) would lead to various calculated values of Rss, an assumed single distance based on expression (7). (The distortions would be different for other ratios of x / R o . ) These plots show how, under the best of circumstances, assuming the correct single donor lifetime, distance measurements forflexible dye tethers obtained from steady-state data Rss would differ from the most probable distance because of the presence of a distance distribution. This addresses the error in case 3 above [preceding Eq. (S)] for steady-state measurements. For commontethers of extendedlengthontheorder of 18 8, [70], (T in practice The conclusion is that for these cases, steady-state is approximately 10-12 derived distances can easily be in error by 4 Errors in cases 1-3 could add to give distance errors on the order of 10 In conclusion, lifetime data eliminate the problem of a reference lifetime agreeing with the given lifetime [Eq. (7)], deal directly with multiple donor lifetimes, and explicitly consider reasonable

where

A.x

x

(x)

A.

A.

A.

17

Fluorescence and Biological Processes

50

0

25

50

100

75

R( 4

60

40 N

0

c-

X h

P

iil

20

0

I

t

2

3

B1

4

Z (nsec)

Figure 4 Probabilitydensityfunctionsfor (a) distance P ( R ) andthecorresponding distribution (b) in lifetimes P ( 7 ) . P ( R ) corresponds to a shifted Gaussian distribution Rg = with E = 65 A, c = 10 A. P ( 7 ) is the corresponding lifetime distribution, with 60 A, 7D = 4 nsec. Points A I , B I , A2, and B2 are respectively separated from X by 1 and 2 0 . Points AI, B I , A2, and B2 are the points on P ( 7 ) that derive from A ] , B1, A?, and B2 of P ( R ) . The 7 valuescorrespondingrespectively to A2, AI, B1, and B2 are 0.604, 1.489, 3.169, and 3.56 nsec; 7 = 2.47 nsec corresponds to the peak of P ( R ) at R = 65

A.

18

Parkhurst

4 ,

20

0

60

40

R

80

100

(A)

Figure 5 The solid curve shows a probability density function for the distance between the ends of two flexible tethers of length I8 8, each connected to a straight line of length 60 8 , . The ends of each tether can occupy uniformly the region within each sphere of radius 18 8 , . The curve was from a Monte Carlo simulation with 10' points. The dashed curve shows the Gaussian distribution generated with these parameters and Eqs. 1 la,l Ib. 1.12 1.10 1.08 1.06

I

/ R=45A

& I 1.02 1.04

/

1.oo

0.98 0.96 0.94

0.92

0

2

4

6

8

10

16 12

14

( 4 Figure 6 The ratio of RSS ( R calculated from steady-state data) to 'Ti for three different as a function of cr for the P ( R ) distributions for Ro = 60 8 , .

values of

x

Fluorescence and Biological Processes

19

distancedistributions. If ~7 is on theorder of 8 p\, thereistether consistent with a distribution of K* values.

C.

Rapid but Restricted Dye Rotation and

flexibility

K~

Letusnowconsiderthe K~ distribution in somedetail. P ( K * ) (Fig. 3, curve A) hasanunusual shape,andalthough ( K * ) is 2/3, (T is 0.7. This P ( K * ) is for random orientation of donor and acceptor transition dipoles, and when it is used explicitly for FRET analysis, one assumes that there is no motion of the dipoles after excitation. Under this circumstance, in the time domain, one would observe an infinite number of decay times because of the infinite number of kt valuescalculatedovera K* range of 0 4 ; thus the kt values.thoughinfinite, would be bounded. [If, on the other hand, there is complete rotational averaging over times very short with respect to the donor lifetime, then the effectiveP ( K ~ ) collapses to the single average valueof 2/3, giving for each R only one value for kt.] There are three features of curve A in Figure 3 that can be associated with three canonical structures for the dipole orientations (see Fig. 7). In structure A, is parallel to z , is along y, and ,& is parallelto x ; all threevectordot products are zero; and K~ = 0, the most probable value for P ( K * ) . In struqure B, let again lie along z , as does (parallel orientation), and again let R be along y . Now K * = I , the value of K~ where P ( K * ) has a discontinuous first derivative. In “head to tail” orientation of structure C, let ,ii1 , k*, and R all lie along the y axis. Now all dot products equal 1 for the unit vectors, K* = 4, and P ( K ~ ”+ ) o in a small region of K* as K~ -+ 4.

ji,

,!i*

Figure 7 The three canonical structures with values of

K*

= 0 (A), 1 (B), and 4 (C).

Parkhurst

20

Consider now an ensemble of randomly oriented donor

(D) and acceptor

(A) transition dipoles in which each dipole can wobble very rapidly (with respect

to transfer)withinaconeofsemiapicalangle 8 (Fig.8). Theconeaxesare 3), but there is dynamical averaging within randomly oriented as before (Fig. the cones. The P ( K ' ) distribution changes markedly. First, under the assumption of equal cone angles for donor and acceptor, the average value of K~ will still be 2/3, but the distribution P ( K ~will ) sharpen as 8 increases from zero. The distributionfor 8 = 60" is shown in Figure 3, curve B. If 8 is 90" or 180". it can be shown that the distribution collapses to a single value of K~ = 2/3. Up to 0 = 90", the maximum width of the distribution can be determined by A and C (Fig. 7). Thusfor examiningwhathappenstocanonicalstructures 45"and60",respectively, the limiting values of K' for structure A are 0.264 and 0.417; for C the values are 2.20 and 1.448. (If 8 increases above 90", the K' valuescalculatedforstructures A andCchangeplaces, A now becoming larger, reachingamaximumvaluefor K~ = 0.75 at 120", thendecreasing to 2/3 at 180". For structure C. K' also reaches its minimum value at 120" (0.531) but then increases as 6 increases to 180". Thus, for a cone angle greater than 90", the rnuxirnunz range of values for R can be only 0.96-1.02 times the value calculated for K~ = 2/3, almost certainly an unimportant consideration.) The formulas for K~ for the three canonical structures are as follows. Let C =

1

3

Then average

D

+ cos(8) +cos(@'

K'

,

Y=-

3c - 1 2

'

S=l-C,

values for the three structures are

A:

(K?)

= D:

B:

(K')

=D

c:

(K2)

=

+ Y2;

D +4Y2

-k 3 s Y

A

Figure 8 Transition moment geometries for movement of the transition moments within a cone of half-apical angle 0 . The angles I#JD and +A are angles between the coneaxes and

the radius vector connecting the centers of the donor respectively.

and acceptor transition moments,

sogical and Fluorescence

21

Note that these cone angles are for the transition dipoles and not for the orientations of the vectors connecting the dye centers to the point of attachment on the macromolecule. If, from depolarization measurements [71] or molecular modeling, one has extra information that allows 6 to be estimated, then, with K* distributionfor 6 < 90" nootherinformation,onecansetlimitsonthe by examining the canonical structures A and C, and in turn set limits on the extrema of R withreference to whatwouldbecalculated for K ? = 2/3 (see Fig. 9). Forinstance,for 0 = 60°, the R valuescould range only from 0.86 to 1.20 times the value calculated for K~ = 2/3, regardless of the orientation of the cone axes. Note that in any actual labeled macromolecule, even one labeled at a specific site, there would be an ensemble of cone axis orientations between A and C and undoubtedly a much narrower distribution would obtain ) in Figure 3. Consider, (e.g., for 60") than that implied by the P ( K ~distribution however, the 68% confidence limits on K~ for curves such as curve B of Figure3and on the resulting distances that would be calculated with respect to that distance for K~ = 2/3 if one assumed tzo distance distribution. For various values of 6, these upper and lower distances with respect to the K~ = 2/3 value are shown in Figure 9. This gives probable ranges of uncertainty in dis-

1.4

,

,

1.4

1.2 1

. IY

I-

0.8 0.6 0.4

0.2

0 O

20

40

e

60

80

100

(degrees)

Figure 9 A plot of the upper ( 0 )and lower (+) limits of the 68% confidence region for theratio of thedistancecalculated for thecorrespondingaveragevalueof K ? for wobble within a cone and that for K~ = 2/3 as a function of the cone angle 8 . The top (A)and bottom (W) curves give the extreme valuesfor the same ratio based, respectively, on canonical structures C and A of Figure 7 and the indicated cone angle.

Parkhurst

22

tancesforrandomdistributionsoftransitionmoments(coneaxes).Thus,for 8 = 60", for the 68% confidence region, the uncertainty in would be only 4%. As an example of theuse of anisotropydatatoestimateaconeangle, when a IO-mer oligoribonucleotide was labeled at the 3' oxygen with fluorescein isothiocyanate (nine single bonds from the dye to the 0 on the phosphate) andthecomplexwasbound to 40s ribosomes,thecalculatedconeangle (8) was 55" [72]. It isthusmarkedlyadvantageous in reducingtheuncertainty in FRETderived distances to couple the dyes by tethers that allow significant rotational freedom on the subnanosecond time scale, and once 8 exceeds 60" only a small error in distances should result from using a single value of K~ = 2/3. Note that once 8 is approximately 50", the 68% confidence limits on R (Fig. 9) are nearlysymmetricalaroundthe K~ calculatedvalue,andwhen 8 > 45" one expects the error in R (assuming only a single R value) to be 2r, one can show that the average distance ( ( R ) ) between dyes is given by

( R )= R and

0

+ 2r2/5R

( I la)

for that distribution is

A,

A,

Thus, for R of about 60 r = 5-20 0 = 0.63r, and ( R ) is verynearly equal to R . One can show that P ( R ) for this problem is well approximated by a Gaussian distribution (Fig. 5). If r = 18 A, R = 60 one obtains ( R ) = 62.2 and (T = 11.2 The modelcanbeelaborated in severalways. For instance, suppose that eventhoughthetethersare flexible, oneaccountsforexcluded volume due to the tether itself by requiring that the dyes be excluded from a sphere of radius ( 2 / 3 ) r from the ends of R , or 12 in this case. One then finds that ( R ) = 62.8 and (T = 12.3 hardly changing ( R ) . Excluded volume effects from the nzncronlolecule itself can have a much greater impact. For instance, for end-labeled duplex DNA, one has an excluded volume that corresponds to the intersection of a sphere and a cylinder, a volume that depends on the ratio of the radii that characterize the geometrical structures. Suppose the dyes are attached to the sugars in a DNA duplex at positions 1 and 16 onthesamestrand,withattachmentpointsofthetethers, of maximum

A,

A. A

A,

A

A

26

Parkhurst

extended length I8 A, separated by 54.8 A in the 16-mer duplex. Excluding 12 A for the tethers, excluding the interior of the cylinder and a van der Waals distance of 3.5 A at the ends of the cylinder, one obtains ( R ) = 65.2 A and a = 9.8 (MonteCarlosimulation), in excellentaccord with experimental values of 66.5 for R and 8.8 A for a for measured values in 1 M KC1 solution [68]. Figure 10 shows the actual distribution (lo6 random pairs of points in a Monte Carlo simulation) and the Gaussian distribution that has thesame and a values. Note that there are featuresof the actual skewed distribution not modeled well by a Gaussian, suggesting that with more precise data Hermite polynomial expansions for P ( R ) might well give enhanced geometrical information on the macromolecule.Thebest fitting Gaussian(least-squaressense)totheMonte Carlodistributionhas = 66.2 A, a = 10.7 A, aneven better fit tothe experimentalthantheMonteCarlovalue.Theseresults showthat a simple Gaussian P ( R ) canyieldvaluesthatare in excellentagreement with those derived from simple Monte Carlo simulations of excluded volume. (If the DNA cylinder is not excluded, the parameter values are ( R ) = 57.7 A, a = 12.3 A,

A

A

x

5

1

0

20

40

60

80

100

R (A) Figure 10 The solid line is from a Monte Carlo simulation ( lo6 points) of the distance between two dyes at the endsof tethers each of length 18 A linked to DNA and separated by 54.8 A as described in the text. The interior of the DNA was excluded as was a shell of radius 12 8, for each tether. A 3.5 8, van der Waals excluded region was also used at the end of each cylinder. The dashed curve is a Gaussian generated with the average distance (65.2 A) and (T (9.8 A) that characterize the non-Gaussian distribution. (This is trot the best-fitting Gaussian to the distribution; see text for details.)

Fluorescence 27 Processes and Biological

showing how important excluded regions of the macromolecule, and hence local geometries, can be in affecting distances obtained from modeling.) Proceeding as above with various simple assumptions, one could extract a distance R for the rigid rod in both straight and bent forms. One can describe a bending angle for either a smooth bend or a sharp kink in the center of the If we rodasdepicted in Figure 11 (or for a morecomplexbendingmodel). return to the very simplest model of a geometric line with two spheres attached, lifetime measurements on a reference B-type DNA would provide Rstraight, and a measurement after the binding of a protein, for instance, would give a new value for R from which a bending angle could be deduced. (In the very simplest and most favorable case, such fluorescence intensity changes could be measured fromchanges in steady-statedata[see Eq. ( 7 ) ] , subjecttoall of thecaveats above, and some inferences drawnas to changes in geometry [80,81 1.) Figure I 1 shows how end-to-end distance measurements can be converted into estimates of bending angles for two models of DNA bending. In a recent study in our TATAlaboratory, we measured distance changes (from lifetime data) for various

12

1

Smooth Bend

-.

0.8 -

0.6

(Sharp Midpoint)

0.4 -

02-

0 20

40

60

80

100

120

140

I

160

0 (degrees) Figure 11 A plot of the bend angle (x axis) for two bending models as shown versus the ratio of the inter-dye distance for bent (dashed lines, where “a” is Rbent) to that for straight forms of DNA, with the bending angle as defined for the two models. For the smooth bend, the distance !?stfi,ight would be the arc length; for the single-kink model, itwouldbethesumofthe twosidesofthetriangledepicted by solidlines.Forthe smooth bend, the ratio is 2 sin (8/2)/0 ( 8 in radians); for the one-kink model, the ratio is cos(Q12). The assumption is that a single distance characterizes the distance between as extensions of geometric lines (see donor and acceptor and that the tethers function inset figures) that characterize each type of bend.

28

Parkhurst

-

R = 48.8 8,

-

R = 66.7 A

Figure 12 Two geometries for 180" bent DNA showing the different locations of the tethers and the regions availableto the dyes on each tether.The maximum extent of each of 12 8, radius was an excluded tether was 18 A. but a shell around each attachment region. The figures are scaled so the helix diameter is 20 A, and the points of attachment of the tethers are indicated by solid circles. These points of attachment are separated by 54.8 A for each structure. The shaded regions show cuts through the truncated spherical shells accessible to the dyes.

box mutations upon bindingTBP (TATA binding protein). Fitting the data to the smooth bending model of Figure 11, we found angles for the various mutations that ranged from 55" to 105" with errors of f2"for the angles. In the case of rather extreme bending (see Figs. 12a and 12b) it is clear that the location of the tether points around the circumferenceof the helix will affect the computed bend angles becauseof the excluded volume regions of the duplex. For both structures, the tether points were separated by 54.8 A, and the extended tether lengths were 18 A. The duplex was excluded, as was a 3.5 8, distance at the ends of the helix and a 12 8, radius for each tether. For case (a), i? and c were,respectively, 66.7 and 10.2 A, whereas for case (b) the corresponding values were 48.8 and 10.6 A. Clearly there are geometries for which the phasing of the excluded volume of the helix must be considered in detail. 13), for some orientations of the dyes the bend On the other hand (see Fig. angle can be calculated rather well from very simple measurements of the inter-

Figure 13 A hypothetical DNA bentin a sharp kink to an angle of 90°, a = b = 25.5 A, with the axis of the bend perpendicular to the cylinder axes.

ogical and Fluorescence

29

dye distances. Consider the 16-mer with dyes on bases 1 and 16, which i n the canonical B structure would be 180" apart when viewed down the helical axis. Consider a 90" bend in the duplex that is a sharp kink, with the two attachment points in a plane perpendicular to the axis of the kink. Let the extended tether in Figure 13. Onefinds,forthe lengths be 18 8, andtheexcludedregionsas bestGaussian fit to thedistribution, = 43.8 8,, (T = 11.4 A, andfromthe extended length of 65.5 8, (see above), one calculates (see Fig. 1 I ) an included bend angle of 83" (rather than 90"). Simple calculations show that excluded volume effects from the macromoleculemustbeconsidered if oneis to interpretthedistanceinformation correctly. Toward this end it will clearly be necessary to use multiple triangulations and incorporate studies on well-chosen reference molecules, determine optimum tether lengths, and possibly use P ( R ) distributions more complex than shifted Gaussians. FRET has been used in solution studies to follow the kinetics of bending of the TATA box of DNA upon binding of the TATA binding protein (TBP), to detect intermediates in the process, and to draw inferences as to the bending of the DNA in these intermediates [82]. Because of the unique sensitivity of FRET to small changes in distances, it can be anticipated that increased use will be made of this technique in both static and kinetic studies now that distributionmethodshaveprovidedevidencethatthe K~ problem is tractable and that reliable distances can be obtained.

IV. ACKNOWLEDGMENTS I wish to thank Professor G. A. Gallup for deriving the expressionsin Eqs. ( 1 l ) , and Robyn Powell and Jiong Wu for drawing the figures. REFERENCES I. P Pringsheim. Fluorescence and Phosphorescence. New York: Interscience, 1949. 2. SV Konev. Fluorescence and Phosphorescence of Proteins and Nucleic Acids. New York:PlenumPress,1967. 3. JR Lakowicz. Principles of Fluorescence Spectroscopy. NewYork:PlenumPress, 1967; 2nd ed, 1999. 4. JRLakowicz, ed. TopicsinFluorescenceSpectroscopy,Vol. 1, Techniques. New York:PlenumPress,1991. 5 . JRLakowicz, ed. TopicsinFluorescenceSpectroscopy,Vol. 2, Principles.New York:PlenumPress,1991. 6. JR Lakowicz, ed. Topics in Fluorescence Spectroscopy, Vol. 3, Biochemical Applications.NewYork:PlenumPress,1992. 7. J I Steinfeld. Molecules and Radiation. 2nd ed. Cambridge, MA: MIT Press. 1985, pp 285-290.

30

Parkhurst

8. TMJovin,MBartholdi,WLCVaz,RHAustin.Rotationaldiffusionofbiological

macromolecules by time-resolved delayed luminescence (phosphorescence, fluorescence) anisotropy. Ann NYAcad Sci 366:176-196, 1981. 9. GB Kistiakowski, M Nelles. The resonance fluorescence of benzene. Phys Rev 41: 595,1932. I O . BStevens,EHutton.Thefluorescenceandexcitationspectraofanthracenevapor at low pressures. Mol Phys 3:71-78, 1960. 1 1. A Jablonski. Uber den Mechanismus der Photolumineszenz von Farbstoffphosphoren. Physik 94:3846, 1935. J AmChem Soc66: 12.GNLewis,MKasha.Phosphorescenceandthetripletstate. 2100-21 16, 1944. 13. GNLewis,MCalvin.MKasha.Photomagnetism.Determinationoftheparamagnetic susceptibility of a dye in its phosphorescent state. J Chem Phys 17:804-812, 1949. 14.PPringsheim.FluorescenceandPhosphorescence. NewYork: Interscience,1949, pp 4, 290. 15. TFiirster.FluoreszenzOrganischerVerbindungen.Gottingen: Van derHoeckand Ruprecht, 195 I , p 12. 16. HW Leverenz. An Introduction to Luminescence of Solids. New York: Wiley, 1950: pp148-152. 17.WMVaughan,GWeber.Oxygenquenchingofpyrenebutyricacidfluorescence inwater. A dynamicprobeofthemicroenvironment.Biochemistry9:464-473, 1970. 18. RA Alberty, RJ Silbey.PhysicalChemistry. NewYork:Wiley,1992,pp692-693. 19. DP Craig, JM Hollas, GW King. Upper limit to the intensity of the 3400A singlettriplet absorption in benzene. J Chem Phys 29:974, 1958. 20. DF Evans. Perturbation of singlet-triplet transitions of aromatic molecules by oxygen under pressure. J Chem Soc (Lond) 1957:1351-1357, 1957. 2 1. P Yuster, SI Weissman. Effects of perturbations on phosphorescence: Luminescence of metal organic complexes. J Chem Phys 17:1182-1 188, 1949. 22. J N Murrell. The effect of paramagnetic molecules on the intensity of spin-forbidden absorption bands of aromatic molecules. Mol Phys 3:319-329, 1960. by quenchingof 23. JAKnopp, IS Longmuir.Intracellularmeasurementofoxygen fluorescence of pyrenebutyric acid. Biochem Biophys Acta 279:393-397, 1972. 24.WLRumsey,JMVanderooi, DF Wilson.Imagingofphosphorescence: A novel in perfused tissue. Science 241 :1649method for measuring oxygen distribution 1651,1988. 25.DBCalhoun. JMVdnderkooi,GVWoodrow 111, SWEnglander.Penetrationof dioxygen into proteins studied by quenching of phosphorescence and fluorescence. Biochemistry22:1526-1532,1983. 26. M Kasha.Paths of molecularexcitation.In:LGAugenstein,ed.Proceedingsof aSymposiumSponsoredbythe U.S. AtomicEnergyCommissionHeldatthe BrookhavenNationalLaboratory,Oct12-16,1959.(RadiatResSuppl2).New York: Academic Press, 1960, pp 243-275. 27.DSMcClure.Triplet-singlettransitions in organicmolecules.Lifetimemeasurements of the triplet state. J Chem Phys 17:905-913, 1949.

and Fluorescence 28.

29. 30. 31.

32.

33.

34. 35.

36. 37. 38.

39. 40. 41. 42. 43. 44. 45. 46. 47. 48.

49.

Biological 31 Processes

MKasha.Collisionalperturbationofspin-orbitalcouplingandthemechanismof J Chenl Phys fluorescence quenching. A visual demonstration of the perturbation. 20:71-74. 1952. RAAlberty,RJSilbey.PhysicalChemistry.NewYork:Wiley,1992,pp 449451, 492497. PAM Dirac. The quantum theory of the emission and absorption of radiation. Proc Roy Soc (Lond) Ser A114:243-265,1927. JI Steinfeld. Molecules and Radiation. 2nd ed. Cambridge. MA: MIT Press. 1985, pp146-147. TForster,KKasper.EinKonzentrationsumschlagderFluoreszenzdesPyrens. Z Electrochem 59976-980, 1955. J Ferguson. Absorption and fluorescence spectra of crystalline pyrene. J Chem Phys 281765-768,1958. RM Hochstrasser. Mixed dimer emission from pyrene crystals containing perylene. J Chem Phys 36:1099-1 100, 1962. JN Murrell. J Tanaka. The theory of the electronic spectra of hydrocarbon dimers. MolPhys7:363-380,1964. AC Sen, B Chakrabarti.Proximityofsulfhydrylgroupsinlensproteins. J Biol Chem 265: 14277-14284, 1990. JA Dix, AS Verkman. Pyrene eximer mapping in cultured fibroblasts by ratio imaging and time-resolved microscopy. Biochemistry 29: 1949-1 953. 1990. WT Simpson, DL Peterson. Coupling strength for resonance force transfer of electronic energy in van der Waals solids. J Chem Phys 26588-593, 1957. DL Peterson, WT Simpson. Polarized electronic absorption spectra of amides with assignments of transitions. J Am Chem Soc 79:2375-2382, 1957. ASDavydov.TheoryofMolecularExcitons.NewYork:McGraw-Hill,1962,pp 8-70. M Kasha. Energy transfer mechanisms and the molecular exciton model for molecular aggregates. Radiat Res 20:55-71, 1963. MKasha,HRRawls,MAshrafEl-Bayoumi.Excitonmodelinmolecularspectroscopy.PureApplChem11:371-392,1965. of nitrogen base-pairs M Kasha, M Ashraf El-Bayoumi, W Rhodes. Excited states and polynucleotides. J Chim Phys 58:916-926, 1961. RMHochstrasser,MKasha.Applicationoftheexcitonmodeltomonomolecular lamellar systems. Photochem Photobiol 3:3 17-33 I , 1964. QHGibson.Combinationofporphyrinswithnativehumanglobin. J BiolChem 23913282-3287,1964. E Rabinowitch, LF Epstein. Polymerization of dyestuffs in solution. Thionine and methylene blue. J Am Chem Soc 63:69-78, 1941. T Forster, E Konig. Absorptionsspektren und Fluoreszenzeigenschaften Konzentrierter Losungen organischen Farbstoffe. Z Electrochem 61:344-348, 1957. BZPackard,DDToptygin,AKomoriya. L Brand.Profluorescentproteasesubstrates: Intramolecular dimers described by the exciton model. Proc Natl Acad Sci USA93111640-1 1645, 1996. BD Hamman, AV Oleinikov, GG Jokhadze, DE Bochkariov. RR Traut, DM Jameson. Tetramethylrhodamine dimer formation as a spectroscopic probe of the con-

32

50. 51.

52. 53.

54.

55.

56. 57. 58. 59. 60. 61. 62.

63. 64. 65.

66. 67. 68.

Parkhurst

formation of Escherichia coli ribosomal protein L7/L12 dimers. J Biol Chem 271 : 7568-7573,1996. RF Service. New probes open windows on gene expression and more. Science 280: 1010-101 I , 1998. RP Haugland. Handbook of Fluorescent Probes and Research Chemicals. 6th ed. Eugene, OR: Molecular Probes, Inc., 1996. WT Simpson. Internal disperson forces. The polyenes. J Am Chem Soc 73:53635367, 195I . WT Simpson. Resonance force theory of carotenoid pigments. J Am Chem Soc 77: 6164-6168,1955. LJ Parkhurst, BG Anex.Polarizationofthelowest-energyallowedtransitionof B-ionylidenecrotonicacidandtheelectronicstructure of thepolyenes. J Chem Phys 45:862-873, 1966. JR Lakowicz. Principles of Fluorescence Spectroscopy. NewYork:PlenumPress, 1967; 2nd ed. 1999, pp 372447. CR Cantor, PR Schimmel. Biophysical Chemistry. 11. San Francisco: WA Freeman, 1980, pp 448454. HC Cheung. Resonance energy transfer.In: JR Lakowicz, ed. Topics in Fluorescence Spectroscopy, Vol 2. New York: Plenum Press, 1991, p 129. TForster.ZwischenmolekulareEnergiewanderungundFluoreszenz.AnnPhysik (Leipzig) 2%-75, 1948. T Forster. Mechanism of energy transfer. In: M Florkin, EH Statz, eds. Comprehensive Biochemistry, Vol. 22. NewYork: Elsevier, 1967, pp 61-77. ER Blout.Asystemwithrelativelyfixeddonor-acceptor SALatt,HTCheung, separation. J Am Chenl Soc 87:995-1003, 1965. M Dahan, AA Deniz, T Ha, JR Grunwell,DS Chemla, PC Schultz, S Weiss. Biophys J Abstr 43rd Annu Mtg, February 1999, p A168, MPos87. 1 Gryczynski, W Wieslaw, ML Johnson, HC Cheung, C-K Wang, JR Lakowicz. Resolution of end-to-end distance distributions of Hexible molecules using quenchinginduced variations of the Forster distance for Huorescence energy transfer. Biophys J 541577-586,1988. P Wu, LBrand.Orientationfactor in steady-stateandtime-resolvedresonance energy transfer measurements. Biochemistry 31:7939-7947, 1992. LJ Parkhurst, KM Parkhurst. Changes in the end-to-end distance distribution in an oligonucleotide following hybridization. SPIE 2137:475483, 1994. K M Parkhurst, LJ Parkhurst.Kineticstudies by fluorescenceresonanceenergy transfer employing a double-labeled oligonucleotide: Hybridization to the oligonucleotide complement and to single-stranded DNA. Biochemistry 34:285-292, 1995. CR Cantor, P Pechukas. Determination of distance distribution functionsby singletsinglet energy transfer. Proc Natl Acad Sci USA 68:2099-2101, 1971. in Fluorescence HC Cheung. Resonance energy transfer. In: JR Lakowicz, ed. Topics Spectroscopy, Vol 2. New York: Plenum Press, 199 I , pp127-1 76. i n adoubleKM Parkhurst,LJParkhurst.Donor-acceptordistancedistributions as a singlestrandandinduplexes.BiolabeledHuorescentoligonucleotideboth chemistry34:293-300.1995.

sogical and Fluorescence

33

DNAandRNAstructureanddynamics. Curr 69. DP Millar. Fluorescence studies of Opin Struct Biol 6:322-326, 1996. 70. KM Parkhurst, M Brenowitz, LJ Parkhurst. Simultaneous binding and bending of promoter DNAby the TATA bindingprotein:Realtimekineticmeasurements. Biochemistry35:7459-7465,1996. 71. E Bucci, RF Steiner. Anisotropy decay of fluorescenceas an experimental approach to protein dynamics. Biophys Chem 30:199-224, 1988. 72. R Hileman. Eukaryotic peptide chain initiation: A study using fluorescent probes. PhD Dissertation, University of Nebraska, Lincoln, NE, 1993. 73. FTanaka.Theory of time-resolvedHuorescenceundertheinteractionofenergy transfer in a bichromophoric system: Effect of internal rotations of energy donor and acceptor. J Chem Phys 109:1084-1092, 1998. 74. G Weber. Dependence of the polarization of the fluorescence on the concentration. Trans Faraday Soc 50552-555. 1954. 75. PJSims.ComplementproteinC9labeledwithfluoresceinisothiocyanatecanbe used to monitor C9 polymerization and formation of the cytolytic membrane lesion. Biochemistry23:3248-3260,1984. 76. PJ Sims, T Wiedmer. Kinetics of polymerization of a Huoresceinated derivative of complement protein C9 by the membrane-bound complex of complement protein C5b-8. Biochemistry 23:3260-3267, 1984. DNA kinking,andthe 77. MH Werner, AM Gronenborn.GMClore.Interacalation, control of transcription. Science 27 I :778-784, 1996. 78. JD Kahn, E Yun, DM Crothers. Detection of localized DNA flexibility. Nature 368: 163-1 66, 1994. 79. DA Leonard, N Rajaram, TK Kerppola. Structural basis of DNA bending and oriented heterodimer binding by the basic leucine zipper domainsof Fos and Jun. Proc Natl Acad Sci USA 94:49 I 3 4 9 18, 1997. 80. K Toth, V Sauermann, J Langowski. DNA curvature in solution measured by fluorescence resonance energy transfer. Biochemistry 37:8173-8 179, 1998. 81. MJ Jezewska. S Rajendran, W Bujalowski. Complex of Escherickicr coli primary replicative helicase DnaB protein with a replication fork: Recognition and structure. Biochemistry 3 7 3 1 16-3 136, 1998. 82. KM Parkhurst, RM Richards,MBrenowitz, LJ Parkhurst.Intermediatespecies possessing bent DNA are present along the pathway to formation of a final TBpTATA complex. J Mol Biol 289:1327-1341, 1999.

This Page Intentionally Left Blank

Probes: Dyes Fluorescing in the NIR Region Stefan Stoyanov University of Sofia, Sofia, Bulgaria

1.

INTRODUCTION

Dye research and application have always been dominated by the demands of textile chemistry. However, in recent years the fast growing high technology unconventional applications of colorants in such areas as optoelectronics-optical storage of information, liquid crystal displays, solar cells, light collectors, laser (i.e., photodyes, nonlinear optics; modern reprographics-lectrophotography copiers and laser printers, thermal and ink-jet printers); medical and biological applications (e.g., photodynamic therapy); and optical fiber sensors for diagnostic andanalyticalpurposes,fluorogeniclabeling,andsecurity-light-emitting diode (LED) and NIR bar code detection have revealed the need for new dyes [ 1-91. with specific functional properties, the so-called functional dyes There are three key terms in the title that need some additional explanationwhenusedtogether: dyes, fluorescence, and NIR region. Colorants (dyes or pigments) are characterized by their ability to absorb in the visible region of the electromagnetic spectrum (400-760 nm) and are used traditionally to impart color to textiles, leather, plastics, metals, paper, and other materials. On the other hand, when a molecule absorbs radiation, the energy taken up can be released in various ways. One possible way is fluorescence, i.e., generally the emission of light from the lowest vibrational levelof the first excited singlet state (SI) to one of the vibrational levels of the ground state (So).The near-infrared (NIR) region is usuallydefined as the650-1800nmrange,butthat is sometimesdivided intotwosubregions:650-760nm,referred to as the deep red (orfarvisible spectral region), and 760-1 800 nm, the near-infrared region. Evidently, for dyes 35

w

ORGANIC MOLECULES

Q)

TAUTOMERISM (ISOMERIZATION)

or AGGREGATION

/

n-electronic STRUCTURE

/

--

POLARIZATION

I

Color Absorbance in the UV-Vis and NIR region. microcolor filters, pleochroism, 2ptical data storage

Fluorescence in the Vis and NIR region. optical sensors. laser dyes, fluongenic probes and labels. solar energy collecton

Nonlinear properties. liquid crystal displays

Photoconductiwty. photosensitization. organic semiconductors. photowltaic cells

Chemical and biochemical senson photodynamic therapy, photochemical hole burning. color formers. thermo- and photochromism. chromoionopnores, Ruoroionophores

Figure 1 Fundamental properties of n-electronic chrornophoric system and some of their hi-tech applications. (From Ref. 15.)

Probes: Dyes Fluorescing in the NIR Region

37

fluorescing in the NIR the color is no longer of direct significance, and these dyesbelongtothegroup of new functional dye materials thatGriffiths [ 101 defined within the modern concepts of structure-spectra correlation as follows: “A functional dye is a molecule whose electronic structure permits the absorption of electromagnetic radiation by electronic excitation and whose functional property is directly related to these special electronic characteristics.” The first generation of functional dye materials were selected from the existing colorants, generally purified for the more demanding hi-tech applications. Gradually they are being replaced by new superior second generation functional dye materials, specially and specifically designed for the intended application [ 1 11. Generally an organic molecule can absorb near-ultraviolet, visible, or NIR radiation if a delocalized n-electronic system exists within the molecule. The x--x* electronictransitionsassociatedwiththelightabsorptioncaninvolve energies falling within the corresponding range of the electromagnetic spectrum, 300-400, 400-760, or 760-1 800 nm, respectively. An excited organic molecule with such a large n-electronic systemmay re-emit the absorbed light as radiative fluorescence in the visible or NIR region; such emissions are connected with numerous practical applications such as fluorescent brighteners; high visibility inks and paints; dye lasers and laser technology; solar energy collectors; liquid crystal displays; tracing in drainage systems; nonlinear applications; analytical reagents in chemical, biochemical, and medical analysis; fluorogenic labels; and probes [ I , 12-1 4). Someof these applications are discussed in detailin this book. Alternatively,thehigh degree of mobility of n-electrons, with or without irradiation of light, and their chemical (ground state) and photochemical (in excited states) reactivity can lead to specific electrical and indicator properties such as nonlinear optical properties, photoconductivity and photosensitization, and halo-, thermo-, and photochromism. The organic molecule can specifically interact with its environment and, depending on various energetic and thermodynamic factors, may existin different isomeric or tautomeric structures determinig different properties. Depending on the phase, temperature, solvent, and concentration, the molecules may exhibit solvatochromism or aggregation, which may also alter fundamental properties of the n-electronic system that are of substantial practical importance. The fundamental properties of a n-electronic system that might be exploited practically and are directly relevant to its mode of application are presented in Figure 1.

11.

HISTORYANDDRIVING OF NIR DYES

FORCES FOR DEVELOPMENT

The first far-visible polymethine dye was synthesized by Williams in 1856, i.e., in thesameyearPerkindiscoveredMauveine,andwascalledlatercyanine

38

Stoyanov

(cyanos = blue)because of itsmagnificantblue cyanine has been proved to be of structure 1.

color[8,16,17].Williams’s

The tirst NIR-absorbingdyesweresynthesized in the1930swhentheir sensitizing power for silver halide photography became important and proved usefuluptoI300nm[5.8,16].Somenaturalporphyrins [ 181; relatedsyn[ 19-22]; polytnethineand theticcyclicchromophores(e&,phthalocyanines) polymethine-likestructures;metalcomplex,quinone,andazodyes;andmiscellaneous chromophoric systems proved to be NIR-absorbing dyes 11-1 1,231. However, difficulties were associated with instrumentation and the fact that the NIR spectral region was not accessible to the available equipment, and in many cases the ability of these NIR-absorbing dyes to fluoresce was not studied. In this respect there is a current need for more information concerning this important functional property. In the case of NIR fluorescing dyes, the selective absorption and fluorescence are utilized, not their apparent color. Interest in the developmentandapplication of newNIR-absorbingand-fluorescingdyesis motivated mainly by three important achievements: Development of inexpensive NIR diode lasers, based on Ga,AI1-,rAs (where x = 0-1) and providing laser light in the range from about 850 nm for gallium arsenide down to about 750 nm as the aluminum content is increased, and the use of Q-switch dye cells to improve the energy output of solid-state Rb and Nd:YAG (I064 nm) lasers. 2. Developmentof new optoelectronicsystemssuch as laseroptical recordingandreading of information,laserprintingprocesses,and NIR laser dyes. In these systems the NlR dyes are used as effective photoreceptors or active media for laser emission. 3. Development of NIR fluorescence spectrometry based on diode lasers and inexpensive semiconductor detectors and photodiodes for the NIR spectral region, since the most widely used photomultiplier detector tubes have significantly less sensitivity in this region.

I.

Although diode lasers in the deep red or far-visible region (660-680 nm) are already commercially available [24], it was recognized that the use of NIR fluorescence spectrometry offers the following significant advantages [ 13,251 in

Fluorescing Probes: Dyes

in the NIR Region

39

spite of the fact that only a few classes of organic dye molecules exhibit NIR absorbance and fluorescence: matrix interfer1. There is no background fluorescence signal or small CH ence. The solvent isnot excited to high vibrational levels (overtone and OH vibrational bands), which prevents more sensitive detection in absorption spectrometry, providing better sensitivity. 2. The influence of background Rayleigh and Raman scattering is greatly reduced due to the 1 /A4 dependence of the intensity of scattered light, which makes NIR fluorimetry attractive for in vivo fluorescence measurements of macroscopic tissue samples or in whole blood [14]. 3. Most chemical species, including those of biological origin, have no to the absorption or emission in the NIR spectral region as opposed UV and visible regions. of the 4. Since no impurityfluorescenceinterfereswiththedetection fluorescent sample molecule, it is possible to detect some NIR fluorescing polymethine dyes down to concentrations of about lo"* M [25,26] or even lower, M [27,28], approaching the detection sensitivity at thesinglemolecule level fortheNIRtricarbocyanine fluorescing dyes 2 (IR-132) and 3 (IR-140), respectively [29,30].

IR-140

Stoyanov

40

Theseresultsstronglyimply that near-infraredfluorescencespectrometry using diode lasers and a fiber optical system is highly promising and useful for ultratrace analysis and a potential technology of the twenty-first century. It shouldbenoted that most of thecommerciallyavailable NIR dyes have simple trivial names, abbreviations, and/or numbers, which are frequently used in analytical practical and applied fields in place of their correct chemical names. A possiblereasonmightbethatspecialistsfromdifferentscientificandapplied fields suchasphysics,electronicengineering,biochemistry, medicine,andevenchemistryarenotfamiliarwithorganicchemical nomenclature. An appropriate example is the typical NIR polymethine dye 4,

I (CH2)4

80;

I S03Na

lndocyanine green (ICG) IR - 125

indocyaninegreen ( K G ) or IR-125,usedinitiallyasaphotosensitizer,as a laser dye, and for labeling protein in human serum at picomolar detection limits [24,3 1,321.Its chemical name is I , l’-bis(4-sulfob~tyl)-3,3,3’,3’-tetramethyl4,5,4’,5‘-ibenzoindotricarbocyanine or [2-17- I .3-dihydro-I, 1 -dimethyl-3-(4sulfobutyl)-2H-benz[e]indol-2-ylidene]1,3,5-heptatrienyl-[ I , I -dimethyl-3-(4sulfobuty1)- 1 H-benz[e]indolium hydroxide inner salt sodium salt. Another important application of NIR dyes connected with light emission (Fig. 1 ) is their use as active media in dye lasers. A substantial number of potentially applicable x-electron systems suitable for laser dyes are described in Refs. 3, 8, and 33-36. In these organic compounds it was especially important to elucidate the relationship between particular chemical structures and functionalities such as light absorption and emission. For example, it was shown that some polymethine laser dyes can emit efficiently up to 1800 nm [37]. Tounderstand all thefactorsthatinfluencethefluorescenceandlaser efficiency of any organic molecule it is necessary to consider all the possible radiativeandnonradiativetransitionswithinthesimplifiedJablonskidiagram presented in Figure 2 .

41

Probes: Dyes Fluorescing in the NIR Region

s2

< ?

A

VR

\

S1 IC

Absorption

Fluorescence


0.7) 2. Relativelyshortlifetime of the SI state (xfl 5 I O nsec) 3. Littleintersystemcrossing (ISC) to thetripletstates

4.

5.

High molar absorptivity of the dye molecules, needed for strong absorption of the “pumping” laser radiation Highphotochemicalstabilityandhighpurity

Themoleculardesign of dyesfordyelasers is atypicalexample of thefundamentalapproach to hightechnologyapplications. For thepurposes ofthisreview it wasinteresting to evaluatefromtheexistingandavailable literature,cited in thischapter,thenumberofNIRfluorescingdyesincludingthosefluorescing in thedeep redspectralregion (650-1800 nm)and to classify them generally according to their chemical structure, determining the specificx-electronchromophoricsystem involved. The investigatedliterature cited here is limited to scientific books, publications, and catalogs or data sheets

42

Stoyanov

[6,8,32-391, since it is practically impossible to search all the literature on NIR dyes. The two major sources are the handbook of Okawara et al. [36], which discusses about 2700 selected dyes for electro-optical applications, and the book of Maeda[33]onlaserdyes,where546dyesarelisted.The maintypes of chemicalstructuresandthecorrespondingnumber of dyesfluorescing in the deep red (650-760 nm) and NIR (760-1800 nm) spectral ranges are listed in Table 1. The number of NIR-absorbing dyes with potential use mainly for optical recording media [5,7,40-42] is rather large compared to the number of deep red and NIR fluorescing dyes (Table 1). The limited number of these fluorochromes is the major deterrent to the complete utilization of NIR fluorescence for many applications where fluorochromes with high fluorescence efficiencies are needed. Another conclusion emerging from the collected data in Table I and noted by some authors [5,13,34] is that the polymethine dyes, i.e., polymethinecyanines, merocyanines,andpolymethine-likedyessuchastriarylmethanes,constitute themajority of deep redandnear-infraredfluorescingdyes in the650-1800 to evaluate the number of cyanines nm spectral region. It was interesting also to thenumberofmethinegroupsin belonging to differentgroupsaccording the general polymethine structure Sa, where n = I , 2, 3 . . . corresponds to the

carbocyanines, dicarbocyanines, tricarbocyanines, etc. The data are collected in of the cyclic or acyclic endgroups Table 2, and more details about the nature are presented in Section 111. Evidently the two main subgroups of polymethine-cyanine dyes (Table 2) that are used in deep red and NIR fluorescence applications are the di- and tricarbocyanines, respectively. Among the members of these two subgroups are the most promising fluorescent probes for diode laser excitation in the 660-680 nm and 750-850 nm wavelengths even though the NIR dyes have smaller quantum yields, shorter fluorescence lifetimes, and in many cases poorer photochemical stability. Due to the low background fluorescence signal or lack of interfering fluorophores in most chemical species of organic and bioorganic origin, the deep to the traditional fluorored and especially NIR fluorescing dyes are superior genic labels [ 13,24,43,44].

43

Probes: Dyes Fluorescing in the NIR Region Table 1 Numbers of Deep Red and Near-IR Fluorescing Dyes in Various Chcmical Classes Spectral region

Chemical class Cyanines Merocyanines Phthalocyanines Chlorophylls Triary lmethanes Xanthenes Oxazines and thiazines Acridines Heterocyclic (pyrilium and thiapyriliuni) salts Miscellaneous Sorrrw: Refs. 6,

Deep red (650-760 nm)

Near-1R (760-1800 nm)

19

60 3 3 1 5 0 7 0 0 4

6 4 3 0 7 8 1 1

4

8. 32-39.

Table 2 Number of Polymethinecyanine Dyes Fluorcscing in the NIR Spectral Region Spectral region Deep red (6.50-760 nm)

Near-1R (760- I 800 nm)

1

3

2 4

12 4 -

5

-

2 II 35 5 7

I?

;'

3

' 1 1 = number of methine groups in thc general structure 5a.

111.

STRUCTURAL CLASSES OF DYES FLUORESCING IN THE NIR REGION

A.

Polyrnethine Dyes

Many prominent chemists, including Koenig [45], Hamer [ 161, Daehne (46,471, Brooker [48], and Kiprianov [491, contributed to our present understanding of the color and electronic structure of polyinethine dyes of general structure 5a or 5b.

Stoyanov

44

While structure Sb indicates the polymethinic structure of the conjugated chain,thevinylogyismoreadequatelydescribed by Sa [17]. In such a way di-andtricarbocyanines Sa arepenta-andheptamethinecyaninedyes Sb, respectively. The polymethine dyes can be cationic (cyanine), anionic (oxonol), orneutral(merocyanine).Thecyclicendgroupsaremainlyheterocyclic,with heteroatoms more electronegative than carbon in the majority of polymethine dyes or carbocyclic residues in the case of nonsymmetrical polymethine ionic (e.g., 6) or neutral (e.g., 7) dyes. In the latter case the number x of methine groups is even. The polymethine dye 6 [35] is a phenylogous hemicyanine or styryl dye, whereas 7 [36] is a merocyanine dye. Some examples of heterocyclic endgroups are presented in Figure 3.

-7

-6

x=,

x=4

In the case of acyclic endgroups, the cyclic ones are replaced by terminal heteroatomic groups containing nitrogen and oxygen atoms. These dyes containing the parent chromophore of the ionic or neutral polymethine dyes are called streptopolymethines and could be represented with the general structure 8.

in streptopolyrnethines

D = NR, and A = NR,

in streptooxonols D = 0- and A = 0 in streptornerocyanines D = NR, and A = 0

.s

w

0 v)

0

vi

al v)

K Z II

% X

al v)

vi

II

0

vi

K z X

al

vi

v)

v)

N

0 IY I1

0

I

0

vi

v)

al

0

vi

v)

0

X

K

I1

K z

X

II

K

z

0

vi

al

X

It

0

!5

N

-0

v)

al

X

I1

K z

0

vi

Q

cr:

Probes: Dyes Fluorescing in the NIR Region

a

d

cr:

\

i X

T

p x

\ a?

z

X

It

X

c

0

cr:

45

Stoyanov

46

In spite of the wide choiceof heterocyclic endgroups suitable as donors and acceptors, the length of the polytnethine chain, and the varietyof substituents introduced in the synthesis of polymethine dyes, the majority of synthetic methods can be generalized by relatively few principles [8,17]. The endgroups result from to form the specific endgroup synthons that react with chain-forming synthons polymethine dye. These specific reactive groups are summarized and described in a systematic way with appropriate examples for synthetic routes of various polymethine dyes in Ref. 17. The classical comprehensive report of Hamer [ 161 and further review articles and books [8,35,49-561 are useful tools for preparative organic chemists. It is not possible to cover most synthetic approaches and recent developments in the synthesis of polymethine dyes, but a few examples for the synthesis of deep red and NIR fluorescing polytnethinecyanine dyes are given below. The most convenient mode of synthesis for dicarbocyanines 5a ( n = 2) or pentamethinecyanines 5b ( x = 5) is the reaction between two equivalents of 2-methyl heterocyclic quaternary salt of indoline, benzothiazole, benzoxazole, or benzselenazole with one equivalent of phenyliminopropenylaniline derived from the condensation reaction of malonaldehyde with aniline accordingto the scheme in Figure 4. By addition of a tertiary aliphatic amine, the 2-methyl heterocyclic quaternary salt is deprotonated to the respective methylene derivative 9, which is the appropriate nucleophilic reagent, obtained in most cases in situ 181.

I

R

The tricarbocyanines 5a ( n = 3) are the main subgroup of NIR-fluorescing polymethinecyaninedyes.Theycanbeobtained by usingtheringopening of suitableheterocyclicringsactingaschain-formingsynthons in thepresence of specific endgroups. A typical example is 3,3’-diethylthiotricarbocyanine (DTTCI), presented in Figure 5 [ 171. The green tricarbocyanine dye DTTCI belongs to the strongly fluorescent dyes with absorption maximum at 768 nm, fluorescence maximum at 790 nm, quantum yield @H = 0.34 (571, and laser efficiency in the broad tuning range 790-880 nm 1361. Anotherspecificsyntheticapproachleading to stericallyrigidizedpolymethine dyes with absorbance and fluorescence mainly in the deep red and NIR spectral region consists in bridging the chain methine groups with acyclic rings [ 17,35,59-661. This structure modification makes the conformationally ‘‘loose’’

47

Probes: Dyes Fluorescing in the NIR Region

Figure 4 Convenientsynthetic mode for dicarbocyanines.

+ CH3NH2 + H A R = C,H,

and A = I

Figure 5 Representativesyntheticapproach

for 3.3”diethylthiatricarbocyaninc iodide

(D’ITCI).

polymethine chain linking the heteroaromatic fragments rigid and enhances the thermal and especially the photochemical stability due to the partial loss of flexibility and the decreased efficiency of the internal conversion, which is one of the major nonrddiative deactivation pathways (Fig. 2) in organic molecules. In the case of deep red and NIR fluorescing dyes, the restrictionof this deactivation pathway is particularly important, as the decreased excitation energy between the ground and first excited singlet states becomes closer to that of vibrational excitationenergy.Usuallythesynthesisstartswithalicyclicketones,ketals, enamines, or enol ethers. These compounds possess two activated methine or methylene groups, reacting easily with Vilsmeier’s reagent to form initial cyanine whose polymethine chain can be lengthened by nucleophilic reaction with to the scheme in Figure 6 1171, other heterocyclic methylene bases according leading to tricarbocyanine 10.

48

&

Stoyanov

+

-

2 CIANR2 A-

+ 2HCI + H A

RZN’&NRz

CI

‘3 +

I CZH5

2 H N R z + 2H A

I CZH5

10 Figure 6 Synthetic scheme leading to sterically rigidized tricarbocyanine dye 10.

A similar approach to produce stable NIR laser dyes with pyrylium and thiopyryliumterminalheterocyclicgroupswasusedbyRaueetal. [X]. The Vilsmeier reaction of cyclohexanone and cyclopentanone with POC13 and dimethylformamide leads to suitable bifunctional intermediates ll and 12. Further reactionwithappropriatebenzopyryliumandbenzothiapyryliumperchlorates produces highly stable NIR laser polymethine dyes 13-15, shown in Figure 7. of five If thephotochemicalstabilityofthedyewithamethinechain of 14, in whichthree C Catomsis put equal to I , thentherelativestability atomsarebound in asix-memberring, willbe 128, andthat of 13, where three C atoms are built into a five-member ring, will be 178. The fluorescence maximum of 15 in 1,2-dichloroethane is shifted to about 1140 nm in the NIR region. In somecasesthewholepolymethinechain is involved,makingthe acyclic skeleton rigid [56,61-651; the tetracarbocyanines 16 with fixed all-trans configurationsaregoodexamples of structureswithimprovedphotochemical stability and stronger fluorescence emission in comparison with the analogous nonrigid compounds [66]. One of the major specific requirements for NIR fluorogenic labels [ 13,141 is the existence of reactivefunctionalgroupssuchastheisothiocyanate (-N=C=S)groupthatdirectly bindthefluorophoretoreceptors or ligands containing primary NH2 groups of an analyte molecule of organic and bioorganic origin. In addition to thisspecificfunctionalproperty for covalently attached labels, a selective solubility, depending on the type of label, is needed.

49

Probes: Dyes Fluorescing in the NIR Region

CI

HoHcbcHo +

Ph

12 -

T

6

+

0

/I

HCN(CH3)z + POCI3

Figure 7 Synthetic scheme for producing stable rigidized NIR laser polymethine dyes 13-15.

Stoyanov

50

For large hydrophobic biornolecules to be labeled, the hydrophobicity of most polymethinedyes is not adisadvantage, but i n somecasessulfonategroups needed to enhance the solubility are introduced. A limited number of deep red andNIRpolymethinedyesusedasnoncovalentandcovalentfluorogenic labels were synthesized in the 1980s and 1990s [67-781. They possess functional groups that form conjugates with several classes of biomolecules such as DNA, lipids, peptides, and proteins. An interesting approach to derivatization of polymethinecyanine dyes 17, obtained in a reaction sequence analogous to the one

17 -

X = CI, R, = R2 = C,H,

or R, = (CH,),S03- and

R, = (CH,),SO;Na'

shown in Figure 6 and containing a nucleofugal group at the central position for reaction with nucleophiles, is reported by Strekowski et al. [72]. The substitutionreactions of X in 17 with MeONa,MeNH2,PhONa, PhSH, and 4-NH2PhSH yielded the corresponding methoxy, methylimino, phenoxy. andphenylthioderivatives.Thehighstabilityofthep-NH2-substituted phenylthio derivative in the latter case was usedto synthesize the isothiocyanatosubstituted derivative of 17, where the NH2 group is converted to an N=C=S group.Thisreagentwas used for NIRlabeling of proteins at aminogroups, andpreliminaryresultsshowedsuperiordetection in comparisonwithfluoresceinisothiocyanate,aclassicalreagentforproteinlabelingwithaUV-Vis

Fluorescing Probes: Dyes

in the NIR Region

51

chromophore[75]. All derivativesexhibitarelativelystrongability to fluoresce with quantum yields in the range of 0.1-0.5 in the 65G820 nm spectral range [72]. Deep redand NIR cyanineandmerocyaninedyescontainingiodoacetamide and isothiocyanate reactive groups and used as fluorogenic labels for protein sulfhydryl residues have been described [76,77]. It was found that the fluorescencepeak of dye 18 at 790 nm is sensitivetoachange of potential

across the nerve cell membranes, associated most probably with dye aggregation on the membrane, or an equilibrium of adsorption at two different sites is established where the absorption and fluorescence properties of the dye are not identical [78]. 18 with IZ = I , 2, It is interesting to note that the absorption maxima of and 3 are located at 500, 526, and 544 nm, respectively, while the corresponding fluorescence maxima are much more sensitive to the extended conjugation, being at 615, 715, and 789 nm. Obviously, in order to rationally design new deep red and NIR fluorescing dyes with favorable fluorescence properties in aqueous, nonaqueous, polymer film, or solid state, it is necessary to understand and evaluate factors that affect their photophysical and photochemical properties. This is especially important also in fundamental aspects, since polymethine dyes are no longer considered a representative class of organic dyes but rather a basic to prototypicalstructurewithintheconjugatedorganiccompounds,according Daehne’s triad theory [79].

1.

Fundamental Aspects of the Polymethine State

The existence of polymethine states is essentially bound to two properties: the nature of endgroups connected with the methine chain, and the number of x electrons ( N 1 ) distributed over N atoms. The presence of two electronegative groups X terminating the chain anda surplus of one x electron with respectto the number of atoms determine the characteristic polymethine electronic structure [17,80]. Owing to the electron-withdrawing effect of the endgroup, the carbon atomsadjacenttotheheteroatoms,aswellasanyatomconjugatedtothese carbon atoms, willhavea partial positivecharge 6+. Thus, if thenumbering starts from any of the X atoms in 19, the even-numbered carbon atoms along the chain bear a partial positive charge and the remaining odd-numbered carbon

+

52

Stoyanov

atoms bear a partial negative charge 6-, and the total electronic distribution the polymethine chain displays alternating array charges along the chain.

6’

&+

6’

of

6’

Alternating partial charges of the methine groups along with an equalization of the bond lengths are characteristic properties of polymethines, independent of the total charge of the molecule. In contrast to other classes of dyes, polymethine dyes cover the full range of colors, reaching the NIR spectral rein the deep red gion.Agreatnumberofpolymethinecyaninedyesabsorbing and NIR region are described in Refs. 5 , 16, 17, 36, and 80. Unfortunately, the number of polymethine dyes that display strong fluorescence in the 650-1800 nm spectral region is limited. Relatively high fluorescence ability or quantum yield is one of themostimportantrequirementsdeterminingtheirfunctional use as laser dyes, sunlight collectors, and fluorogenic labels and in optical fiber sensors and biosensors.

2.

Factors Affecting the Fluorescence Ability Polymethine Dyes

of

In general, the position, intensity, and shape of the fluorescence band of polymethine dyes in the deep red and near-IR spectral region depend on a number of factors. These include 1.

2. 3.

4. 5.

6.

Chemical structure, i.e., the length of the polymethine chain, nature of endgroups, basicity of the heteroatoms in the heterocyclic ring, nature and position of substituents, rigidity of the polymethine chain, and symmetry The relative rate of nonradiative relaxation pathways such as internal conversion and intersystem crossing (Fig. 2) Photochemical stability, possible Z/ E or cis/trans photoisomerization, conformational rotation of the end heteroatomic fragments during the excited state lifetime, photooxidation, and photoreduction The nature of the environment and microenvironment of the molecule (solvent, micellar systems, polymer films, etc.) Existence of a functional group conjugating or binding the fluorescent molecule to the desired object Temperature

64 638 70 744

53

Probes: Dyes Fluorescing in the NIR Region

7. Concentration, i.e., existence or lack of aggregation 8. pH or change of acid-baseproperties of photo9. Electronand/orenergytransferinspectralsensitization graphic silver halide emulsions 10. Formationofpolymethinestructuresviaphotochromicprocesses The majority of deep red and near-infrared fluorescing polymethine dyes belong to di- and tricarbocyanines (Table2 ) with general structure 20, containing indolenine, benzoxazole, and benzothiazole heterocyclic endgroups.

The spectral luminescence properties of vinylogous polymethine dyes 20 are presented in Table 3. The analysis of absorption and fluorescence spectral data presented in Table 3 confirm the established empirical rule [ 17,791 for the A -+ B symmetricalpolymethinesthatthewavelengthofthelowestenergy transition. polarized along the longitudinal molecular axis, depends linearly on the number of vinylene groups (n). The shift amounts to about 100 nm when the conjugated polymethine chain is lengthened by one double bond (100 nm

Table 3 Photophysical Data of VinylogousPolymethineDyes 20

11

0

Abs hmax

G&x

(nm)

(nm)

%I

F1

hAbs max

h u x

(nm)

(nm)

- 410 376 4340.006456 - 4220.004 0.03 482 500 0.25 0.430674585 6550.400610 0.28 710 688 855 880 0.03 -

1 568 545

2 3 4

x=s

x=o

X = C(CH3)2

Solore: Refs. 57, 66, 80. 8 I .

Abs

hmax

FI *max

@(FI)

(nm)

(nm)

QFI

0.040

558

577

0.050

0.580 0.340790 768 0.003 - 910 875

Stoyanov

54

“vinylene shift”). The energy gap at infinite chain length is commonly assumed to be zero. The problem of the band structure is not only of theoretical interest, as it is related directly to the optical properties of polymethine dyes, absorption and fluorescence in the NIR region. The question is,To what extent can the light absorption and respective fluorescence emission be further extended toward the infrared? The strong vinylene shift enabled the synthesis of dyes that absorb in the 1400-1600 nm wavelength range and emit wavelengths of up to 1800 nm [5,34,35,37,59,82]. An appropriate example is the extremely “deep colored” thiopyrylopolymethine dye 21 with reasonable photochemical stability [37], containing a long Ph

c IO,

Ph

Ph

conjugated chain of double bonds, partially incorporatedin five- and six-member rings, causing restriction of the conformation of the polymethine chain. Its absorption maximum is observed at 1420 nm in o-dichlorobenzene, and the stimulated emission is at 1800 nm (5555 cm”). The close analog of 21, containing a five-member ring in the central meso position, has its absorption maximum in the same solvent at 1510 nm [34]. Such rigidity is especially important for all organic compounds, including polymethine dyes, absorbing in the NIR spectral range. In this region, due to the small energy gap width in relation to the visible one, the process of internal conversion (Fig.2)becomeshighlyimportant,andthis is evidentfromthemuchlower ( n = 4) of thevinylogouspolymethine quantum yields for the last members dyes 20 depicted in Table 3. According to the Jablonski diagram, the vertical electronic transition results in a Franck-Condon (FC) type of excited state, Le., a higher vibrational level of the SI state (Fig. 2). The relaxation of the molecule fromthe FC excitedtotheequilibriumfluorescencestate at thezerovibrational level of the SI state proceeds through a change of two major molecular features-geometry and energy. Thus the Stokes shift, defined as the difference between the absorption and fluorescence maxima, can be explained. In the NIR spectral region, especially above 1000 nm, the electronic excitation within the absorption band overlaps partially with the high frequency overtone vibrations of thedyeandthesolvent,leadingto originatingmainlyfromC-Hbonds

Fluorescing Probes: Dyes

in the NIR Region

55

nonradiative deactivation of the fluorescent excited state [34]. As discussed in Ref. 17, Tuytuylkov predicted the energy gap of 1 eV (8068 cm”) for simple streptopolymethines 22 [83] of infinite chain lengths.

n = O , 1 , 2,....

Me

Daehne [84] presented arguments against the existence of extremely long absorption wavelengths based on polarographic half-wave potentials of cationic polymethines and concluded that such compounds are no longer stable if the energy gap between the ground and lowest excited singlet states is lower than 0.9-0.95 eV. This would limit the absorption wavelengths of these compounds to about 1400 nm. Polymethine dyes absorbing within this wavelength region should have equal oxidation and reduction potentials relative to the same reference electrode. In such a case an intramolecular redox process might be expected, resulting in a biradical species. Generally the polymethine dye molecules will gain more and more biradicaloid character as the excitation energy decreases. Atabsorptionwavelengths of about 2000-3000 nmtheelectronicexcitation becomesisoenergeticwiththevibrationalexcitation. In thiscasetheBornOppenheimerapproximation is no longer valid, andvibronicinteractionmay result in Jahn-Teller distortion of the molecular structure. Anotherconclusionrelated to theobservedlinearwavelengthvinylene shift in symmetrical polymethines like 20 and 22 and the theoretical limit ofNIR absorption and fluorescence energies comes from the analysis ofa true “vinylene shift” in inverse centimeters (cm-I) according to the classical Einstein equation A E - hv. The corresponding spectral data for vinylogous compounds 20 and 22 are collected in Table 4 and definitely show a nonlinear relationship between I I and A E in cm“, approaching a limited energy gap width of about 0.8 eV. The same is true if the absorption and fluorescence spectral data in Table 3 are converted into inverse centimeters. Another possible path for nonradiative deactivation of polymethine dyesis Z/ E or cis/trans photoisomerization within the polymethine chain. It is detected in some mono-, di-, and tricarbocyanines through observation of the transient absorption spectra originating from the photoisomer, a second component in the fluorescence decay profile and large viscosity effects in the fluorescence lifetime [ 17,431.The isomerization is a reversible process involving photochemical E --f Z isomerization and a photochemically or thermally induced reaction back to the E isomer.Thiscompetitiveroutefordeactivation of theexcitedsinglet state was studied extensively in connection with the decreased fluorescence or

Table 4 UV-Vis-NIR Absorption Spectral Characteristics of Simple Streptopolymethines 22 and Thiacarbocyanines 20 (X = S)

20

22 Abs Amax

Abs "max

n

(nm)

(cm-')

0 1 2 3 4 5 6 7

224 3 12.5 416 519 625 734.5 848

44,640

Source: Refs. 5, 80.

32,000 24,040 19,270 16,000 13,615 1 1,790

AAbs max

Ahri,+I-ri, 88.5 103.5 103.0 106.0 109.5 113.5

A ~ ~ z - ~ I - I (nm) -

12,640 7,960 4,770 3,270 2,385 1,825

422 558 655 768 875 995 1130 1250

Abs urnax

(cm-l) 23,700 17,920 15,270 13,020 1 1,430 10,050 8,850 8,000

AAn,+l-riI -

136 97 113 107 120 135 120

Auri,-ri,+,

5780 2650 2250 1590 1390 1200 850

Fluorescing Probes: Dyes in

the NIR Region

57

laser efficiency of polymethine dyes [17,43,85,86]. It is extremely difficult to discriminate sharply between the terms “configuration” and “conformation,” and normally if the isomers can be isolated the term “configuration”is used ( A H * 2 100 kJ/mol). The stereochemistry of polymethines is significantly dependent on the constitution, typeof endgroups, and kind and bulkiness of substituents within the chain. The general features of the photochemical E / Z isomerization as a potential channel of radiationless deactivation in vinylogous polymethine dyes can be summarized as follows.

20 (n = 0) fluorescefaintlyatroomtemperature if at all. The nonplanarground state of these molecules is accompanied by a torsional mobility that facilitates internal conversion. The solvent-dependent conformational rotation of the heteroatomic fragments during the excited state lifetime is mainly responsible for their low fluorescence efficiency. In the case of mono-, di-, and tricarbocyanines 20 (n = 1, 2 , 3), the 2 photoisomers absorb in general at longer wavelengths with lower intensity than the stable ground-state E isomers. The lower fluorescence efficiency of stable E isomers is partially due to the overlapping of the basic form fluorescence spectrum with the absorption spectrum of photoisomers; i.e., a reabsorption of the emitted light is possible. For example, the photoisomerization process of rather strongly fluorescing dye 20 (X = S, n = 3, and R = C2H5) is observed in ethanol and DMSO. The absorption maxima of the photoisomer at 800 and 820 nm, respectively, coincide with the fluorescence maxima in the same solvents. The fluorescenceefficiencyandlifetimedroprapidly in thevinylogous series of polymethine dyes (Table 3) on passing from the tricarbocyanines to the higher vinylogs. The reasons for this decrease are complex, including an increase in the rate of internal conversion and reduced photochemical stability. The photoisomerization proceeds through the first excited singlet state. The photoisomers formed are in general more or less unstable species with the activation energy of photoisomerization about 20 kJ/mol or less. Depending on their molecular structure, two types of stereoisoa lifetime mers are formed, a relatively long lived photoisomer with of micro- to milliseconds or a short-lived one with a lifetime in the nanosecond time scale. Radiationless deactivation via torsional motion is reduced and photoisomerization inhibited or eliminated if the dye molecule is partially or completely rigidized. A number of polymethine dyes such as 2, 3, 10, 13-17, and 21 exemplify x c h an approach in the search for efficient NIR fluorescing dyes.

1. Stericallycrowdedheterocyclicmonomethinedyes

2.

3.

4.

5.

Stoyanov

58

Other processes that weaken or quench the fluorescence of polymethine dyes are electron and energy transfer, which have been extensively studied in connection with spectral sensitization of photographic silver halide emulsions [ 17,87,88]. It is not the purpose of this chapter to discuss in detail this important aspect of their technical application, but some requirements for these dyes to act as NIR sensitizers should be pointed out. The photosensitive silver halide grain is active only with light up to500 nm. By the addition of specially designed mostly polymethine dyes, its sensitivity is extended to the deep red and NIR spectral an effective spectral sensitizerin range up to1300 nm. The main requirements for the NIR region include strong light absorption within this range, efficient transfer of the absorbed excitation energy or excited electron to the silver halide, tight adsorption at the silver halide surface, planar and compact molecular structure, and appropriate energetic positions of the ground and excited singlet states with respect to the valence and conductivity bands of the silver halide [ 17,891. The enhancement of the polymethine chain rigidity through insertion of the alicyclic bridgingunits is veryeffectiveatimprovingthephotochemicalstabilityand possible applicability [5] of NIR-sensitizing vinylogous pentacarbocyanine dyes 23 ( n = I ) .

23 n = 1 , 2 , 3,.... Thelong-wavelengthmaxima of 23 ( n = I , 2. and 3) areobservedat 1020, 1 140, and 1240 nm, with log t values of 5.0, 4.54, and 3.85. respectively. A steady decrease of absorption intensity has been observed, accompanied by a flattening out of the near-infrared absorption band. This conspicuous feature of 23 withverylongchainlengthhasbeenattributed to thepresence of an equilibrium mixture of stereoisomers [48,80]. In general, attachment of donor groups in even-numbered positions of the parent chromophore structure 19 leads to a bathochromic effect or a red shift, while donor substituents in odd-numbered positions result in a hypsochromic effect or ablueshift [ 16,17,80]. Theoppositeconsiderations on thespectral effects of acceptor substitution are effective. In a more general formulation any substitution that increases the alternating partial charges of the methine groups

Probes: Dyes Fluorescing in the NIR Region

59

alongthepolylnethinechainresults in abathochromiceffectandviceversa. The largest shift is observed by substitution in the middle of the chain (meso position), altering not only absorption but also fluorescence properties. The nature of the endgroups in Sa or 5b strongly affects the position of theabsorptionandfluorescencemaxima of polymethinedyes. It hasalready beennotedthatvarious di-andtricarbocyaninesfluoresceappreciably in the deep red and NIR regions. In the case of 4,4’-dipyrylo- or dithiopyrylocyanines (R4andRs in Fig. 3), thepolymethinechain ( n = 1-3) is extendedintothe terminal groups, leading to “deep-colored’ NIR fluorescing dyes [34,35,56]. The introduction of rigidizing tri- and tetramethylene rings not only imparts higher stability to the molecules but also results in a bathochromic and fluorochromic X = CR? in shift,reachingthespectralrangeofpentacarbocyanine(R6and Fig. 3). The properchoice of endgroups withagreatereffectivelengthlike thosementionedabove,suitablyrigidizeddi-andtricarbocyaninedyeswith generalstructure 20, is amongthemostpromisingsyntheticapproachesfor practical application of NIR-fluorescing polymethine dyes. The dyes containing amorecomplexmesoionicstructure like R12 in Figure 3 aswellasashort polymethine chain are among the NIR dyes [90]. Among the vinylogous dyes 20 the fluorescence efficiency differs slightly between oxo- and thiacarbocyanines of the same chain length, while indocyanines are less fluorescent (Table 3). Within the heterocyclic endgroups the S- and Se-containing polymethines are always more strongly bathochromic than the 0- and N-containing heteroanalogs [SO]. The substitution at the terminal benzo residues of the heterocyclic endgroups brings about moderate effects, and usually donor, as well as acceptor, substituents shift the corresponding absorption bands to longer wavelengths [17,91]. Whereas the positions of the absorption and fluorescence bands are shifted more or less regularly to longer wavelengths upon lengthening of the polymethinechain,theshape (halfbandwidth Au1p in cm-l) andintensityofthe bands are affected in a more complex way [92]. The shape of the bands, arising from electronic transitions from either the ground or first excited singlet state isdeterminedmainlybyvibronic according to theFranck-Condonprinciple, and intermolecular interactions. The valence full symmetrical vibrations within the atoms of the parent chromophore bring the main contribution to the vibronic i n nuclear equilibrium interactions in polymethine dyes, reflecting the change coordinates upon excitation or emission. At low chain length ( n = 0) the absorptionbandisbroadandsymmetrical,asthechange in geometryislarge. The higher vinylogs show a more asymmetrical and narrower absorption band, indicating a lesser change in geometry upon excitation. The Stokes shift Av,, of vinylogous polymethine dyes 20, presented in Table 3, reflects this change. In some vinylogous series such as 20 and 23, the absorption band undergoes a quite unexpected change-the band flattens, and its molar absorptivity or oscillator strength decreases at long polymethine chain lengths. This effect is mainly

60

Stoyanov

due to solvation and may also reflect the existence of photoisomers [86]; usually a long-wavelength tail of the first absorption band appears, where the photoisomer absorbs strongly. The fine structure of the absorption and fluorescence bands at low temperatures consists mainly of vibrational modes of 1375 f 25 and 1400550 cm”, respectively. The observed progressions are connected with C-C bond extension upon excitation [ 171.

3. Effects of Solvents In general, the position and shape of the fluorescence band depends on the solvent, but due to reduced nucleophilic solvation the fluorosolvatochromic shifts arelessthanthoseseen in theabsorptionspectra. The reducedspecificelectrostatic interaction (nucleophilic solvation) in the first excited singlet state is directlyconnected to thedecrease in thepositivecharge of thepolymethine chromophoreuponexcitation.Thesolventeffectsareenhanced withthe increase in the polymethine chain length and depend on the extent of delocalization of charge over the molecule. Depending on the electron-donating properties of theheterocyclicendgroupsandthelocalsitesofsolvation,includingthe polymethinechain,enhancedspectraleffectsareobserved, in particularwith nucleophilic solvent molecules [ 17,34,92,93]. The deepest hue, the narrowest absorption band, and greatest peak intensity of a given polymethine dye are realized if the solvent has the highest possible refractive index and the lowest possible nucleophilicity [93]. The pentacarbocyanine analog of 21, containing three five-member rings within the polymethine chain and relatively weak electrondonating thiopyrylium endgroups, should possess distinct solvatochromism. In accordance with this prediction, in low nucleophilic solvent o-dichlorobenzene its absorption maximum is at 1510 nm, while in strong nucleophilic CH3CN the observed hypsochromic effect is about 450 nm 1341. The fluorescence lifetimes have a tendency to decrease with increases in the nucleophilicity of the solvent. The fluorescence quantum yield of widely used NIR fluorescing dye 4 depends strongly on the solvent [43]. It increases from 0.01 in water to 0.04 in methanol and 0.07 in butanol. The fluorescence emission of many thiacarbocyanine dyes intimately depends on the solvent and has been shown to increase when the dyes are placed in less H-bonding or more nonpolar solvents [43]. The small fluorescence quantum yields reflect the difficulties associated with the use of NIR dyes and the major changes that occur in the spectroscopic properties when the dyes are placed in aqueous or more highly polar solvents. Some of these changes include extensive ground-state aggregation and solvent-dependent photophysical processes,resulting in reducedfluorescencequantumefficienciesandshorter fluorescence lifetimes [94,95]. In the case of IR-125, no apparent aggregation in water was observed, most probably due to the negatively charged sulfonic groups in the dyemoleculeprovidingabettersphereofsolvation. The short upper-state lifetime coupled with the extended length of the polymethine chain

Fluorescing Probes: Dyes

in the NIR Region

61

and additional aromatic ring on each heteroaromatic fragment may be sufficient to inhibit the conformational changes necessary for E / Z photoisomerization to occurduringthetimescale of theexcitedsingletstate.Aslightdependence of JTH on viscosity, which may arise from partial rotation of the heteroaromatic terminal groups from the molecular plane during the excited state, was observed in IR-125 [43]. The fluorescence of polymethine dyes in principle maybe enhanced at room temperature in viscous solvents such as glycerol, in micellar systems, in syntheticbilayermembranes,and in polymerfilms [17,34,43,96]. Twomajor conditions are critical-the reduced flexibility of the molecule and the lack of specificassociation or aggregation.Thesameeffect is operative if thepolyin glassymatricesatreducedtemperaturewith methinedyesareembedded concomitant increases in the fluorescence lifetime [97]. 4.

Aggregation

The phenomenon of polymethine dye aggregation was discovered a long time ago and is well documented ] 16,17,49,50], The nature of aggregates adsorbed onsilverhalideshasgreatimportance in thespectralsensitization of photoon ingraphic emulsions. The spectral shifts of dimers or aggregates depend termolecular distance and orientation and the number of molecules within the aggregate and were explained quantitatively and semiquantitatively by the application of molecular exciton theory [98]. That theory predicts an intense long( J band) in thecase of head-to-tailarrngement wavelengthabsorptionband of the chromophoric subunits or a weak one at shorter wavelengths ( H band) in head-to-head or sandwich arrangement. The slip angle a between the long molecular axes and the aggregate axis is the most important parameter that determines the direction of the spectral shift. J-Aggregates are characterized by ( a > 54"). small angles (a < 54"), and H-aggregates have greater slip angles The energy level scheme [ 151, showing the energetic distribution of two sandwich dimers (a = 90") and head-to-tail (a = 0") arrangement, is presented in Figure 8. The most favored deactivation paths are indicated within this energy level scheme. The high transition probability(Qo - Q+) of head-to-tail arranged chromophores (Fig. 8b) should favor fluorescence. Usually the intense fluorescence originatingfromJ-aggregatesappears at nearlythesamewavelength as the bathochromically shifted absorption peak. The characteristic spectral features of such J-aggregates include a shifted intensive narrow absorption band coinciding with the fluorescence maximum (resonance fluorescence). As a result, only a small or no Stokes shift is observed. The practically important indocyanine green (IR-125) also forms J-aggregates in water, and a resonance fluorescence at about 890 nm withfluorescencequantumyield Ofl = 3 x isobserved P91.

62

Stoyanov

E

Figure 8 The energy level scheme for ( a ) sandwich and (b) head-to-tail dimers.

The lowtransitionprobability (CPo - CP-) of ahead-to-headsandwich dimer(Fig.8a) mayleadtofacilitatedradiationlessprocessesofintersystem crossing, since the difference in energy between the lowest singlet and triplet levels of the H-aggregates is small. The hypsochromically shifted H-band, observed at higher concentration in monolayers and polymer matrices [ 17,34,43], indicates the existence of weakly fluorescent or nonfluorescent H-type aggregates. The fluorescence quenching of polymethinecyanine dyes in solvents with low polarity and polymer matrices is observed at relatively lower concentrations (10-s-10-6M) than in polar solvents (10-2-10-3M). A possible reason for suchconcentrationdifferences is theformation of H-type ion-pairassociates where the specific parallel arrangement of monomer molecules is achieved not only through hydrophobic interactions but also through electrostatic interactions between the dye and its counterion [34].The existence of H-type aggregation in water, leading to reduced fluorescence efficiency of NIR fluorescent dyes 2 (IR-132)and some derivatives of 17, used as labels in DNA sequencing, was of notedrecently [43,100].In general,theaggregationtendencyandthetype aggregates formed in water or on solid surfaces depend on the dye structure, compactness, low intramolecular mobility, and especially on the length of the polymethine chain. Polymethinecyanine dyes with short chains prefer to form J-aggregates, while the tendency to form H-aggregates increases with longer polymethine chains [ 171. Specially designed cyanines like24 with emission maxima at 1060 nm and containing bulky substituents in the middle of the chain and in the heterocyclic endgroups are promising in reducing the H-aggregation in the polymer matrix [34]. In thecase of tricarbocyaninedyes 17 functionalizedwithanisothiocyanatemoiety,theintroduction of sulfonatednegativelychargedderivatives significantly reduces the concentration-dependent H-type aggregation observed in the nonsulfonated analogs. The existence of bulky substituents in the meso PO-

Probes: Dyes Fluorescing in the NIR Region

63

sition of the polymethine chain and the increased water solubility of sulfonated dyesare themainreasons for thereducedH-aggregation in water.They are promising NIR fluorophores as labels in DNA sequencing [loo]. It should be noted that the principal disadvantage of almost all organic dyesfluorescing in the NIR regionabove 1000 nm istheirlowfluorescence quantum efficiency. Thequantumyieldcannotbegreatlyincreasedbecause of the higher rate of internal conversion in this spectral region. For this reason another fundamental approach, one based on excimer fluorescence, i.e., emission from an electronically excited dimer, has been described [34]. The appropriately designed S- and Se-containing tri- and tetracarbocyanines 25 and 26 exhibit no

Ph

Ph

Ph

ClO, Ph

25

Ph

Ph

CIO 4

CH=CH CH-CH

64

Stoyanov

changes in their absorption spectra within the concentration range of 10-5-10-2 M, indicating lack of aggregation in the ground state. Most probably the bulky phenyl substituents in heterocyclic endgroups and phenyl or dimethyl groups in the central ring within the polymethine chain prevent the association. In the more delocalizedfirst excited singlet state, however, the phenyl rings might be more conjugated with the main chromophoric polymethine system, and this leads to enhanced association of dye molecules. When the concentration is a new,much increasedto lo3 M, themonomerfluorescencedisappearsand moreintense,long-wavelength-shiftedexcimerfluorescencebandappears.In 25 the monomer fluorescence is observed at about 1300 nm, while the excimer emission is shifted bathochromically to about 1580 nm in o-dichlorobenzene and reaches 1700 nm in nitrobenzene. Possible advantages of excimer fluorescence include the use of a high power diode laser pumped Nd:YAG laser (1064 nm) because of its high conversion efficiency from electricity to a coherent light beam and because it lacks the typical features of J-aggregates, Le., sharp and strong absorption and resonance fluorescence, leading to a small, if any, Stokes shift.

5. Other Types of Polymethine Dyes It is well known that nonsymmetrical cyanine dyes, including ionic ones like 18 or neutral merocyanines, absorb at shorter wavelengths than would be expected from the mean absorption wavelengths of the corresponding symmetrical dyes. The loss of symmetry lead to broadening of the absorption band, and the position of the band depends more strongly on the solvent polarity. Compared with the absorption band, the fluorescence band is narrower and less sensitive to solvents. The Brooker deviation, whichis a measure of difference between the heterocyclic endgroups, is less strong in fluorescence than in absorption [loll, meaning that the difference between the basicities of the endgroups decreases in the SI state, which is more delocalized and polymethinic than the ground state. As a result a low solvatofluorochromic effect due to the reduced nucleophilic solvation in the excited state and a corresponding increase in the Stokes shift are observed. Agoodexampleforsuch a structure-spectrarelationship arethevinylogous hemicyanine dyes 18, for which the long-wavelength absorption band is slightly bathochromically shifted from 500 to 544 nm when the chain length is increased from n = 1 to n = 3. The Stokes shift, however, is found to be 115, 145, and 245 nm for I I = I , 2, and 3, respectively. The chromophore structure 18 is less delocalized in the ground state, and the color-structure relationship in terms of intramolecular charge transfer between the substituted phenyl and pyridinium moieties is more advantageous, while in the excited singlet state the electronic distribution is uniform and a polymethine vinylene shift is observed. The differences in the shapes of absorption and fluorescence bands and solvatochromic properties between the symmetrical and nonsymmetrical poly-

in the NIR Region

Fluorescing Probes: Dyes

65

methine dyes determine their specific hi-tech applications. Usually cyanine compounds withnarrowintensiveabsorptionbands aremoresuitableasmodelocking and Q-switching laser dyes, while nonsymmetrical polymethines with large half-bandwidths, high quantum yields, and large Stokes shifts are effective active media for the generation of tunable dye lasers between 650 and 1000 nm. Sensitivity to solvent effects is one of the most characteristic features of merocyanine dyes [102]. Some are positively solvatochromic like 27, and if n 2 2 they absorb in the NIR region when dissolved in polar solvents [5,103]; others, with betaine structures like 28 and 29, are negatively solvatochromic. They are

27 -

Et

28

Ph

Ph

widely used in assessment of solvent polarity [ 1021. The fluorosolvatochromic effects and large Stokes shifts of merocyanine dyes might be a useful tool in thedevelopmentanddesign of newNIR-fluorescingprobesforbiochemical application. The luminescence of a number of new red fluorescent dopants for electroluminescent display application, based on certain modifications of the mero[ 1041. The design cyanine dye structure 30, are described by Chen and Tang features of these highly fluorescent dyes with quantum yields above 0.5 in dilute solutions of 1,2-dichloroethane include chromophoric extension, structural rigidity, and steric spacing. Some representative examples, showing the effectof structural modification and leading to emission in the deep red spectral region 31-34. The existence with rather broad bandwidth, are merocyanine structures of julolidyl and pyran-fused rigid fragments in 34 leads to a red shift of about 80 nm (50 and 30 nm, respectively), which yields a significant portion of photonstoemit in thenear-IRregion since theemissionspectralbandwidth is about 100 nm. The tetramethyl steric spacer incorporated in the julolidyl frag-

66

Stoyanov

I

steric sp;

! r R1

\N, \ chromophoric extension

I rigidization I

CHI,

N

CH’

33 em

j. rnax

34 ern

= 700 nrn

j.

= 755 nrn

Probes: Dyes Fluorescing in the

NIR Region

67

ment aims to reduce the concentration quenching effects, presumably due to the aggregation of guest merocyanine molecules in the host matrix 11041. Anothertype of merocyaninedye is generatedfromspiropyranesand spirooxazines 35a by irradiation with UV light. The photochemically produced “photomerocyanines” 35b are deeply colored [l05], and compounds with ap-

& I colorless

X = C H or N

3513 colored photomerocyanine

propriate structure can absorb at relatively long wavelengths, displaying marked (X = S or Se, nature and position of substituents) solvatochromism. Irradiation of 35b with visible light gives colorless 35a. Good recycling characteristics are i n erasableopticaldisksystems[1,3-7,1051.Reneededfortheirapplication cent developments i n tailoring photoswitchable biomaterials using photochromic spiropyranes as photoisomerizable components are reviewed in Ref. 106. However, most of the colored photomerocyanines absorb in the visible region in the 500-600 nm range with no or weak fluorescence [ 1071. An attractive class of polymethine dyes for various hi-tech applications 36, which are used are squarylium and croconium dyes with general structure as xerographic photoreceptors, in organic solar cells and optical recording media [3,5,8,17,39,40], and more recently as deep red fluorescent labeling agents with (1081. The croconium dyes generally a single succinimidyl ester functionality produce a bathochromic shift of 110-1 30 nm in dichloromethane, in comparison withthatproduced by thecorrespondingcyaninedyes,whilethesquarylium dyesabsorbandfluoresce at approximatelythesamewavelengths ( f 2 0 nm) and with nearly the same efficiency as the parent cyanine chromophore [5,l08]. The croconiumdyeshavebettersolubility,greaternegativesolvatochromism, and much better light fastness than the corresponding squarylium and cyanine 36 with X = C(CH3)2 show strong dyes. The highly water soluble squaraines absorption in the red region of the spectrum [e = 20 x 1OS”30 x I o ” L/(mol . cm)].areasonablefluorescencequantumyield (Qfl = 0.10-0.15) i n aqueous mixtures, and fluorescence maxima in the appropriate deep red (640-690 nm) spectral range [ 1081. They contain a succinimidyl ester functional group attached at the heterocyclic N atom and have been used for visible diode laser-induced fluorescence detection of amines.

68

Stoyanov

X = CH=CH, C(CH,),,

0, S, Se

croconiurn dye

0'

Some diaminophenyl-substituted squarylium dyes 37, derived from squaric acid and N,N-dialkylanilines, show intense absorptionat about 620-650 nm and display multiple fluorescence emission in the deep red spectral range between 650 and 7 10 nm [ 1091. The absorption and steady-state fluorescence emission of 37, assigned as a class of donor-acceptor-donor (D-A-D) molecules, have been studied using structure-property relationships, solvent effects, and temperature.

NR2

37b

37a X = H. CH,,C,H,,OCH,.

OH

Fluorescing Probes: Dyes

in the NIR Region

69

Multipleemissionbandsareobserved in thefluorescencespectra of 37. The three bands are designated a , /?,and y according to their typical Stokes shifts. According to theproposedphotophysicalmodelthe a-band is related to the emission from the Frank-Condon excited state of the free squaraine, the /?-band is the emission from the excited state of thesolute-solventcomplex.andthe y-band is the emission of a relaxed excited state. Rotational relaxation around of squaraine the C-C bond between the phenyl ring and the four-member ring is themajorradiationlessdecayprocess of the first twoexcitedstates.As a result, a twisted relaxed excited state can undergo a rotational relaxation to the ground state or emit a photon to give y-emission [ 1091. Photochemical oxidation is one of the main reasons for the instability of polymethinedyesandespeciallyfortri-andhighercarbocyanines withbenzoxazole, benzothiazole, indolenine, and quinoline heterocyclic endgroups. Part of the reaction includes singlet oxygen, and special tri- and pentacarbocyanine iodides are used as NIR sensitizers to sensitize the production of singlet oxygen [ I IO]. In less polar solvents, tight ion pairs are formed that favor effective intersystem crossing, and the excited triplet state might be an effective sensitizer for singlet oxygen production. In water solution such ion pairs are absent, but some S- and Se-containing di- and tricarbocyanines form singlet oxygen within a solvent cage, which can oxidize the dye molecules via a self-oxidizing mechanism [ 171. In general, photochemical stability can be increased by the introduction of cyclic moieties, enhancement of the polymethine chain rigidity, as in structures 2, 3, 10, 13-17, 21, and 23-26 and other structural fragments as in 31-34 and introducing squarylium or croconium groups into 36 and 37, formally replacing the methine groups. Another interesting and successful approach to improving the photochemical stabilityoftricarbocyaninedyes is described in Ref. I 1 1 andapplied in optical recording systems [7,8]. A highly reflective, light-resistant cyanine dye film for optical disks consisting of ionic salt 38, formed between the tricarbocyanine cationic dye Df and the benzenedithiol Ni complex quencher anion Q-, is very stable against photon mode degradation, compared with the conventional cyanine dyes with CIO$ or CI- anions. Many dithiolate nickel complexes, readto be quenchers of singlet oxygen, ily soluble in organic solvents, are known strong NIR absorbers, and Q-switching dyes 11 121. It will be interesting to see of polymewhether such an approach to improving the photochemical stability thine dyes is possible in other functional applications of NIR fluorescence. Di- and triphenylmethane dyes and their higher vinylogs 39 may be regarded as polymethine-like structures if the two methine groups at each end of the simple streptopolymethines 22 are replaced by benzene rings. The resulting vinyleneshiftcomparedwiththeirbasevaluespresented in Table 4 isabout 80-100 nm [5,17.1 131.

70

Stoyanov

D

+ I

I

R

R

CI

CI

CI

CI

Q-

Some triphenylmethane dyes 39 where X is a Ph-substituted or unsubstituted naphthyl ring exhibit fluorescence properties and are used as laser dyes in the 760-820 nm very near-infrared spectral range [33]. The vinylene shift in diphenylmethane dyes 39 (X = H), starting from Michler’s hydrolblue (n = 0), is accompanied by increasedabsorptionintensityandreducedbandwidthdue to a decrease in vibronic interactions, reflecting smaller changes of the nuclear equilibrium coordinates upon excitation. The spectral characteristics of vinylogous dyes 39, presented in Table 5, confirm these conclusions.

The proper extension of the chromophoric system, the introduction of acceptor substituents at the central C atom, replacement of the central C atom with a more electronegative heteroatom, and the choice of bridging fragments and rigidizing units give rise not only to more or less significant bathochromic shifts [5,17,113] but also to appropriately designed dyes with favorable fluorescence

Probes: Dyes Fluorescing in the NIR Region

71

Table 5 SpectralCharacteristics of Vinylogous Diphenyllnethane Dyes 39 (X = H) in the Deep Red

and Near-1R Spectral Regions

5.18

0 7 10 800 880

1

2 3

4

560

5.33 5.5 1 5.53

920 880 640 5 80

5.58

efficiencies in the deep red and NIR spectral regions for different applications. A good example of such an approach, bridging the polymethine dyes with other structural classes of dyes that fluoresce in this spectral region, is the commercially available xanthene dye 40, Rhodamine 800 [24,32]. The appropriate het-

CN

Rhodamine 800

eroatomic bridging of two phenyl rings. the substituent CN at the central C atom, and the existence of bridging julolidine units in the Rhodamine 800 molecule shiftstheabsorptionmaximum at about 685 nmandinducesitsfluorescence efficiency at about 700 nm, in comparison with the parent chromophore system 39 of Michler’s hydrolblue, which absorbs at 614 nm (Table 5 ) .

6.

BasicConsiderations

In summary, the polymethine dyes are definitely the most promising and most popular NIR-fluorescing dyes due to the well-known flexibility of their chemistry and well-established structure-spectra rules. The following structural and solvent-dependent features of NIR-fluorescing polymethine dyes emerging from this review should be considered [17.34,43,114] in their hi-tech application:

Stoyanov

72

1.

2.

3.

4.

5.

6.

7.

Di- andtricarbocyaninescontainingdifferentheterocyclicgroupsare the most promising candidates. Within these dyes 0- and S-containing heterocyclic endgroups fluoresce more strongly than the corresponding indoleninederivatives. The existence of pyryliumandthiapyrylium heterocyclic rings leads to the most deeply absorbing and fluorescing NIR polymethine dyes up to I500 and 1800 nm, respectively. Partial bridging units in the polymethine chain or at the terminal endgroups impart higher fluorescence ability and photochemical stability, withthebridgeddyesabsorbingandfluorescing at slightlylonger wavelengths than the parent unbridged dyes. Anystructural or environmentalchangeleadingtoadecrease in vibronic interactions increases the fluorescence quantum yield and decreases the bandwidth of the absorption and fluorescence bands. Due toreducednucleophilicsolvation in the SI state,thefluorosolvatochromic shifts are less pronounced. If the length of the polymethine chain is increased ( n > 3), the fluorescence lifetime and quantum yield are generally decreased and the nonradiative internal conversion process becomes more efficient. The inclusion of large heteroatomic units in the terminal groups linked by thepolymethinechaininhibitconformationalreorganizationduring the excited state lifetime, reducing the amount of internal conversion and resulting in negligible rates of photoisomerization. Fluorescence efficiency increases in less polar and H-bonding solvents. The use of organized media, a rigid polymer matrix, and deuterated solvents is an effective tool for maximizing as much as possible the fluorescence efficiency of NIR-fluorescent dyes, particularly in hi-tech applications. Cyanine dyes with short chains prefer to form J-aggregates, whereas those with longer polymethine chains have a stronger tendency to form H-aggregates. Bulky substituents in the middle of the chain and in the heterocyclic endgroups prevent H-aggregation in polymer films. The incorporated charge groups within the chromophore prevent the selfaggregation of hydrophobic dyes in aqueous solution, increasing their solubility.TheH-aggregatesfluorescewith muchlowerefficiency than the monomer dyes. The formation of J-type excimers in some NIR dyesleads to muchmoreintenseandlong-wavelength-shifted excimer fluorescence, well separated from that of the monomer. The observed large Stokes shifts is an advantage compared with the resonance fluorescence of J-aggregates. The heavy atoms in the solvents, within the chromophore or as counterions, should be avoided, especially if the NIR-fluorescing dyes are used in nonpolarsolventsandpolymermatrices.Wheninserteddirectly into the molecule they reduce the quantum yield severely and

Fluorescing Probes: Dyes

Region NIR in the

73

also reduce the fluorescence lifetime owing to increased intersystem crossing. The heavy atom counterion forms ion pairs in low polarity media and enhances self-association. More tight ion pairs might be expected in the SI statesincethecharge of thecation is not uniformly distributed over the whole molecule, giving rise to local sites of solvation.

B. OxazineandThiazineDyes In diarylmethanedyes 39 (n = 0, X = H) thecentralcarbonatomcanbe replaced by anitrogen(-N=)atom. In addition,thetwoaromaticringscan be cyclized at the 2,2’-position with nitrogen, oxygen, or sulfur, producing the basic structures 41 of azines (X = NH, NR, or NAr), oxazines (X = 0),and thiazines (X = S).

A large number of oxazine and thiazine dyes were synthesized in the early stages of industrial dyestuff chemistryby a sequence of oxidation reactions of p nitrosoaniline derivatives with the corresponding phenol or thiosulfonic groups [8]. The absorption spectra of various organic colorants, including derivatives of 39 and 41, are discussed by Fabian and Hartmann [80] within the fundamental color-constitution relationships for their better conceptual understanding. Most oxazines and thiazines, because of their compact and rigid structure compared red region, 650to that of cyanines, are strongly fluorescing far into the deep 760 nm.TheyhavemodestStokesshiftsofabout 20-60 nm depending on the solvent, number of alkyl substituents on the two amine functionalities, and benzoannulation at the 1,2-position in the general structure 41. Some examples of oxazine 42-45 and thiazine 46 and 41 compounds used as laser dyes [33,361, as chromophores for labeling proteins, or as lipid and membrane noncovalent probes in visible diode laser fluorimetry [ 14,24,32,104] are discussed below. Oxazines and thiazines with compact and rigid structures generally have better photochemical stability than cyanines or rhodamines [8,14]. The relatively highmolarabsorptivities,appropriatelong-wavelengthabsorptionbands,and rather strong fluorescence make these commercially available dyes potentially useful fordiodelaser fluorimetry [24,32,108].Theapplication of asemiconductor laser oscillating at 670 nm is quite versatile because of the availability

74

Stoyanov

42

43

Oxazme 1

creG1 vlolet

lAbs= 710 nm

>.FI

44 Nllexue A

hAbs= 628 nm h FI = 690 nm

601 nm

;iAbs=

= 690 nm

; i FI

-

= 650 nm

45 Oxazine 750

hAbs = 673 nm

).Fl = 691 nm

46

47

Thlonlne

Methylene blue

kAbs = 600 nm

hAbs= 668 nm

h

FI

= 623 nm

k

FI

= 683 nm

of inany dyes in this spectral region. Some oxazines and thiazines such as 44, 47, and a derivative of 47, i.e., Azure B, in which one methyl group is substitutedwith H, havebeenused as covalent and noncovalent labels for proteins [24,1 IS]. The aspects important for the designof deep red and NIR Huorophores as covalent labeling agents are

I. 2.

The existence ofasuitablefunctionalgroup to attachthelabelto primary amine or carboxylic group of the biomolecules Arelativelyhighability to fluoresce in waterorpartiallyaqueous solvent mixtures

a

Fluorescing Probes: Dyes in

the NIR 75 Region

The labeling efficiency of NH2 or NHR groups in oxazines and thiazines is usuallyconsideredrather poor in comparison with functional groups like NCS that directly bind to NH? 1131. Another problem is the decrease in fluorescence quantum yields for the phenyl- or alkyl-substituted compounds 44 and 47, especially in aqueous solutions. For example, the fluorescence quantum yield of Azure B and methylene blueislessthan 0.05 [IO81 in water-methanol (50/50 v/v).Thewell-known aggregation ability of oxazine and thiazine dyes 44 and 47, even at low concentration in water [ 151, is most probably the reason for the decreased fluorescence efficiency. Oxazines and thiazines provide a limited perspective for the development of appropriate labels, as suggested in Ref. 108, but further improvement may be effected by development of a labeling reagent that directly binds to NH2 groups in protein,providingfertilegroundforfutureapplication.Reaction of the primary amine group with carboxylic acids in a covalent binding procedure has the same effect.

43-46 usingwater-solublecarbodiimide[24,108]

C.

Phthalocyanines

The metal-free phthalocyanines, naphthalocyanines, and certain of theirmetal derivatives are another class of deep redandNIRfluorophores. The phthalocyanine chromophore was discovered in 1925 when the chemists from Scottish Dyes Ltd., currently part of Imperial Chemical Industries, made observations on residues in a chemical reactor made of iron used in the preparation of phthalimide. A few years later its structure was identified [SI. Phthalocyanines were discovered by chance[5].Thecompanyand university chemistsrealizedthat this new compound might be of significant importance [ 1161. Coming back to our hi-tech electronic age, many applications of phthalocyaninessuchasopticaldatastorage,photodynamic therapy,guest dyesdissolved in liquid crystalline host materials in a laser-addressed system, and fiberopticprobesforNIRfluorescence in opticalsensorsrequirebasicknowledge abouttheirsynthesisandmajorphotophysicalproperties.Variousaspects of phthalocyanine chemistry have been reviewed elsewhere [5,8, I 17].The synthesis ofthemostimportantmetal-freephthalocyanine 48 is accomplishedwith relatively simple starting materials such as phthalic acid anhydride, phth aI'mide,orphthalodinitrilewithouttheisolation of intermediateproductssuchas 1,3-diiminoisoindolenine.In thepresence of Cu(I1)chloride,copperphthalocyanine is formed [SI. Substituted phthalocyanines are obtained either by direct substitution in the existing metal phthalocyanine molecule or by synthesis with substituted starting materials [21,22]. It is suggested that the parent chromophore of porphyrins and phthalocyanines is the 16-atom flat cyclic system 49. which contains 18 x-electrons within a skeleton of four pyrrole rings cyclized in their a , a'-positions by four nlethine

Stoyanov

76

groups [8,80]. However, X-ray studies are more consistent with a structure that involves more than 16 atoms as the basic chromophore [ 1 181. It is important to note that all compounds with 412 2n-electrons in the perimeter of one ring have aromatic character. As discussed earlier, polymethine dyes are built from methine chains, and if they form a ring system they lead to a highly conjugated rr-electronic structure called an annulene. Porphyrins like hemin or chlorophyll are annulenes with four nitrogens or aza[ 18lannulenes and are called pigments of life because of their central importance in living systems [SI. As in the case of polymethine dyes, methine groups in annulenes can be replaced by nitrogen (-N=) at the a- or meso-position. The absorption spectraof free-base porphyrins consist of three main bands in the 300-700 nm region, an extremely intense B or Soret band at about400 nm, and a low intense Q-band split into Q\- and Q,. components at about 620 and 530 nm. both with typical vibronic structure. The Q.l- and Q, bands are polarized parallel and perpendicular, respectively, to the H-H axis. Replacement of four methinegroups in 49 bynitrogenatomsandbenzoannulation to thepyrrole nuclei in passing from porphyrins to metal-free phthalocyanines 48 give rise to a marked change in the spectral absorption features. Phthalocyanines show strong absorption Qx and Q, bands at 698 ( E = 162,200) and 665 nm (6 = 151,400) and hypsochromically- shifted at 350 nm with a moderate intensity B-band [ 801. Replacement of the methine groups by nitrogen atomsin tetraazaporphyrin leads to an enormous change in the intensities of the Q.y and Q,. bands, while the B-band is shifted considerably toward the UV region. Further benzoannulation to the pyrrole nuclei gives rise to an additional bathochromic shift of the and Q, bands accompanied by a further increase in intensity. The Soret band is observed at nearly the same wavelength region with reduced intensity. Metal and metal-free phthalocyanines have 041, and D2/, symmetry, respectively, and, due to higher symmetry in the former case, a characteristic splitting of the Q-band

+

e.,-

the Fluorescing Probes: Dyes in

NIR Region

77

is absent. A very large number of MO calculations made on the structure and absorption spectra of porphyrins and phthalocyanines revealed the major spectrastructurerelationships [80]. The calculationsindicatedthatachargetransfer proceeds from the center toward the outsideof the phthalocyanine chromophore. Theintroduction of acentralmetalion,whichreducestheelectrondensity at theinnernitrogenatoms,leadstoahypsochromicshift that isobserved experimentally. For example, in Cu-phthalocyanine the Q-band is observed as a single peak at 678 nm with remarkable intensity ( e = 218,000). Three factors that determine the wide traditional application of phthalocyanines as pigments in various hi-tech fields are their ( 1 ) and their potential use as functional dyes bright blue to green hues with high color strength, (2) high chemical stability, and (3) exceptional lightfastness. The brightness ofphthalocyanines is one ofthemainreasonsfortheir commercial success. It can be explained by the shape of the Q-band with halfbandwidth of about300-700 c n - ' andtheredfluorescence,whichleads to anincreaseintheirbrightness.However,therearesomefeaturesthatmay The observedpolymorphism of causeproblems in theirspecificapplication. Cu-phthalocyanines leads to different physical and structural properties of the a- and p-forms. The practical insolubility of phthalocyanines in low boiling organic solvents and significant aggregation phenomena in solutions of derivatives containing water-solubilizing groups influence their color-specific applications in medicine [8]. The need for light-absorbing compounds with higher solubility in organic solvents to match the wavelengths of diode laser light determines the synthetic approaches in the search of NIR dyes within this group. Three general approaches provide means of tuning the wavelength of the Q-band absorption in the deep red and NIR regions of the spectrum: benzoannulation, substituents at the periphery of the molecule, and the nature of the central metal ion. Onpassingfromphthalocyanines 48 to1,2-and2,3-naphthalocyanines, bathochromic shifts of about 20-30 and 60-100 nm are observed, as exemplified in the latter casewheretheQ-band is at 780 nm[21,1 191.In addition, the annulated phthalocyanines are also fluorescent at room temperature. The insolubility of the phthalocyanine ring systems in organic solvents in areas such as optical data storage, photoconductivity, and photodynamic therapy of cancer [2-8,11,39,120] is overcome with the introduction of bulky or longchain substituents located CY to the point of fusion of four benzene rings to the heterocyclic rings. Such substituents should cause substantial disruption of the strong lattice forces in the parent phthalocyanine and hence help solve the solubility problem by preventing aggregation [22]. In addition, a bathochrornic effect depending on the nature of the substituent is predicted. The largest bathochromic shift is observed for the derivatives of phthalocyanine or 2,3-naphthalocyanines that bear alkoxy groups with chain lengths varying between C I and C12. Comparison of the long-wavelength band at 862 nm in alkoxy compounds with that of

78

Stoyanov

unsubstituted 2,3-naphthalocyanine at 780 nm illustrates the sensitivity of the Qband to alkoxy substituents located at a-positions. A broadly comparable shift is apparent in thephthalocyanineserieswhere(RO)s-substitutedcompounds absorb at about 760 and 740 nm, well to the red of the Q.I- and Q,. bands in phthalocyanine, located at 698 and 665 nm, respectively [22]. The presence of C1 atoms at the 2,3-positions causes a small blue shift, confirming earlier observations that the Q-band is more red-shifted by substituents at the 1,4-positions. The rhombic splitting A = Ql - Q,. for the Q-band absorption in the spectra of metal-free compounds differs from one series to another, and its magnitude is ameasure of the departure of thesystemfrom D41, symmetry [ 1211. The splitting in metal-free tetraazaporphyrins is about 2100 cm-' . Benzoannulation to this system lowers the Q-band splitting to 730 cm". The incorporation of substituentsatthe1,4-positionscontinuesthetrend,loweringthesplitting to about 390-430 cm-'. Members of the (RO)g-substituted 2,3-naphthalocyanine series show only a single Q-band. as does the unsubstituted compound itself. There is a correlation between the energy of the centroid of the Q-bands and the energy splitting of the bands; the lower the energy, the smaller the value of A . Fluorescence maxima of metal-free derivatives of phthalocyanines and 2,3naphthalocyanines in chloroform are observed at 745 and 890 nm, respectively [221. Thechoice of thecentral ion in metalphthalocyaninesandnaphthalocyanines is another tool to shift the wavelength of the Q-band absorption and emission in the deep red and NIR regions of the spectrum. The extent of this shift to shorterwavelength depends on the electronegativity of themetalion. The octabutylphthalocyanines areachemicallystablegroupwithintenseabsorption in the deep red that are intermediate in position between those of the phthalocyanines and naphthalocyanines. The synthesis and spectral properties of several new metal complexes of (CqH90)s-substituted phthalocyanines are described in Ref. 2l . Their absorption and fluorescence characteristics are collected in Table 6 along with those of the parent ligand [22]. The absorption spectraof these metal complexes in the 600-850 nm region consist of one far-red Q(0,O) band and another one (referred as the Q' band) in the 620-700 n m region, assigned to an admixture of the Q(2,O) vibronic satellite and an additional electronic transition[ 1221. The fluorescence spectrain benzene have maxima attributable to the Q(0,O) and Q' transitions. The Ql bands are in each case located at about 1300 cm-I to the red of the Q-band maxima. The that Stokes shift in almost all cases is rather small (240-390 cm"), suggesting the geometry of the SI state of the complexes is close to the geometry of the ground state. It appears that their triplet-state yields are also intermediate, being higher than those of naphthalocyanines but lower than those of phthalocyanines. 0 2 ( ' A,) These compounds [21] are being considered as photosensitizers for because of the proximity of their triplet-state energies to that of 0 2 ( ' Ax). The

in the NIR Region

Fluorescing Probes: Dyes

79

Table 6 Spectral Characteristics of Metal-Free and Metal Octabutoxyphthalocyanines

and

Q-Band Amax

E

x

10-4

(nm)

[l.mol-l.cm-l]

(nm)

None Zn

761" 737 764 767 76 1 779 724 688

13.4 19.0 19.0 20.5 21.5 18.5 20.0 9.7

738" 66 1 682

AI

Ga

Ge Sn Pd Ru

'Absorptionandfluorescencecharactenstics 2 1. 22.

x

hlll"X

Metal

10-4

[I.mol-l.cm"]

685

679 695 648 624 of Q., andbands

11.3 270 3.9 270 3.9 380 4.4 370 4.6 240 4.4 310 4.4 350 4.7 1200

AH (nm)

AVST

(an-I)

778 752 787 786 775 798 743 750

In toluene [22]

Sourc~,:Refs.

transfer of electronic energy to molecular oxygen from the triplet states of large organic molecules is an important condition in photodynamic therapy of tumors. Singlet molecular oxygen, 0 2 ( ' A S ) , is regarded as a leading candidate for the initiation of tissue damage in the presence of light, oxygen, and an absorber. As tissue allows deeper penetration with increasing wavelengths in the range 600-1200 nm, the search for effective sensitizers with high molar absorptivities in the deep red and NIR regions has been undertaken. The requirements for an effective sensitizer in photodynamic therapy were reviewed by Dolphin and Sternberg [ 1231. The photosensitizer dye should

1. Exhibithighmolarabsorptivity in the 650-850 nmregion,where tissue does not absorb considerably 2. Beusedwithanintenselight source,preferably a solid-statelaser with peak output to the 5 W range, to match the absorption peak of the drug 3 . Exhibit preferential absorption into tumor tissue, but clear readily from normal tissue 4. Fluoresce more strongly in tumor tissue than in surrounding tissue for diagnostic purposes 5 . Bereadilyavailablefromnaturalorsyntheticsources 6. Possesstheability to beagoodphototoxinwhilebeingnontoxic in the dark The first generation compound Photofrin, a derivative of hematoporphyrin, meets in part only four of the above requirements. It is an efficient sensitizer,

Stoyanov

80

demonstrates a degree of preferential accumulation into tumors, is highly fluorescent in tumors, and possesses no dark toxicity. However, it does readily clear from tumors, has a poor absorption profile for penetration into tissue, and is not a single compound 11231. Silicon naphthalocyanines 50 are among the promising photodynamic senin opticaldatastorage [ 1201 and in security sitizers (201, withpotentialuse

RO

50 R = Si(Alkyl), or Si(Alkyl),-OAlkyl

applications for invisible bar code printing [ 1241. A model for reversible energy transferbetweenthephthalocyaninesorsiliconnaphthalocyanineand 0 2 has been proposed [20,21]. The fluorescence spectrum of Si-naphthalocyanine in 2methyltetrahydrofuran consists of two peaks at 776 and 8 15 nm. The Stokes shift is only 2 nm, and these two peaks correspond to the transitions from SI( u = 0) to So(w = 0) and SO(U= I ) , respectively [ 191, with energy separation of about 620 cm" [20]. The synthesis and spectral characteristics of several tetrasubstituted aluminum2,3-naphthalocyaninesfordeterminationofmetalionshavebeendescribed 11251. Basedonthese NIR dyes, an NIR fiber-opticprobeconsisting of a semiconductor laser diode (780 nm), an NIR dye, and a detector has been reported [ 1261. A set of luminescent dyes, namely porphyrin ketones and their Zn(II), Pt(II), and Pd(I1) complexes, are reported in Ref. 127. The metal-free porphyrin ketones display strong pH-dependent fluorescence at about 640 and 620 nm due to aproteolyticequilibriumbetweenthefreebaseandthedication.The Zn(I1)complexesalsohavestrongfluorescence,whilethe Pt(I1) and Pd(J1) porphyrin ketones display room temperature phosphorescence in the

Fluorescing Probes: Dyes

in the NIR Region

81

720-900 nm rnage without any detectable fluorescence. Comparedto the existing porphyrins, the new dyes display a considerably long-wavelength-shifted luminescence along with substantially improved photochemical stability. The phosphorescent Pt(I1) and Pd(I1) complexes represent an alternative to fluorescencebasedprobingowingtosuperiorpropertiessuchaslargeStokesshifts,long decay times, and high photostability. The spectral characteristics of various proteolytic forms of the porphyrin ketone dyes dissolved in liquid polymer (PVC) membranes have been studied and appear to be promising candidates for use in optical sensors [ 1281. Evidently, a proper choice of substituents at the periphery of the phthalocyanine chromophore system, benzoannulation, and choice of the central metal ion offer good possibilities to increase their fluorescence ability, photochemical stability, and solubility.

D. Other Deep Red and NIR Fluorophores Highly fluorescent xanthene dyes like the well-known Rhodamine 6G, fluorescein, and RhodamineB are considered triaryllnethine dyesin which two aromatic rings are cyclized at the 2,2’-position by oxygen [8]. However, most xanthene dyes fluoresce below 650 nm. If the possible rotation of dialkylamino groups is restrictedbybridgingjulolidineunits as in 40, thefluorescenceability is increased along with a red shift of the fluorescence maximum. Appropriate examples are Rhodamine 1 0 1 and 640 with general structure 51, which fluoresce at 650-660 nm [33,35,36].

s-” I Y

X = SO,-, Y = SO,H X = COOH, Y = H

Rhodamine 101 Rhodamine 640

Stoyanov

82

The fluorescent lipophilic dye 52 (DZ-49), used as a pH indicator by He et al. [ 1291, also belongs to the class of xanthene dyes. The sensing scheme is

based on the selective extraction of the organoammonium ions into a PVC lipid membraneandtheconcomitantrelease of a protonfromtheprotonateddye, contained in the PVC membrane, into the sample solution. Upon deprotonation the dye undergoes a color change, which is detected optically. Perylenediimides 53 absorb in the visible range between 525 and 580 nm, fluoresce at 540-620nm,andshowaremarkablefluorescencequantumyield

R' R\

"R

R

R' R"

R = alkyl, aryl R' = H, phenoxy, 4-tert-butylphenoxy of about 0.9-1 .0 [ 1301. In addition to their application as commercial dyes and pigments they are used in modern reprographics, fluorescence light collectors, photovoltaic devices, dye lasers, and molecular switches [4,6,8].They are char-

Fluorescing Probes: Dyes

in the NIR Region

83

acterized by a brilliant color, strong fluorescence, and good thermal. chemical, and photochemical stability. Some perylene dyes such as 54 show strong fluorescence, shifted in the deep red region [ 1301.

AlkO

OAlk

hAbs= 610 nrn

= 63 000) h FI = 685 nrn (6, F I = 0.93) (E

Temylenimides 55 and 56 represent a new class of deep blue colorants that exhibit absorption maxima at 650-700 nm and fluorescence emission in the deep red to near-IR region from 670 to 750 nm [ 13I ] . These compounds have all thepropertiesexpected of excellentfluorescentdyes,such as highmolar absorptivity, high fluorescence quantum yield, and very good thermal, chemical, and photochemical stability. By varying the substituents they can be modified to serve as either soluble dyes or insoluble pigments. The absorption spectra of 55 and 56 in sulfuric acid are characterized by narrowbandsshiftedbathochromically by 160-200 nm and exhibit extremely 55. After dilution with water they high molar absorptivities, up to 508,000 for canberecoveredunchanged,indicatingtheirhighstability to acidsandoxidizingagents. Thequantumyields of 55 and 56 in methylcyclohexaneare 0.9 f 0.1 and 0.6 f 0. I , respectively. The Stokes shifts are 20-50 nm. Besides the potential applications mentioned for perylenediimides.the terryleneimides are promising for uses in photodynamic therapy and laser fluorimetry. The quaterrylene tetracaboxydiimides 57 have an extended conjugated system and exhibit absorption maxima at 764 nm for R = H and 781 nm for R = tert-butylphenoxy derivative, at much longer wavelengths than the corresponding perylene- andterrylenimides [ 13I]. The appropriate N-substitution with the solubility-increasing 1 -hexylheptyl group gives a soluble dye with remarkably strong fluorescence in the NIR region at 816 nm [ 1321.

Stoyanov

84

0

R - -N

0

+ 0 0 0

"R

R" 55 -

"R

R"

R = alkyl, aryl R'= H, 4-tert-butylphenoxy

R-

-R

Fluorescing Probes: Dyes

in the NIR Region

85

Someotherdeep redfluorescingprobes like tris(bipyridy1)osmiumand commerciallyavailableBodipyandCyderivativesarementioned in Ref. 14. A reviewdescribinganalyticalapplications of verynearinfraredfluorimetry pointedouttheneedforsynthesisof new fluorescentlabelsandprobeswith of functional groups to label good stability and solubility that carry a variety and probe numerous features of biologically important molecules [ 1331.

IV. CONCLUSIONS There are a rather limited numberof organic compounds that exhibit intensefluorescence in the deep red and near-infrared spectral region. Most of them belong to the class of polymethine dyes, which are the most popular NIR fluorescing dyes due to their flexible chemistry and well-established structure-color relationships. The introduction of bridging units into the polymethine chain, appropriate heterocyclic endgroups, and substituents are among the factors affecting their fluorescence ability. Other NIR chromophores include rigidized oxazines, thiazines, phthalocyanines, and terrylene- and quaterryleneimides. They can also be modified through annulation, substitution. and the formation of complexes. The use of covalently and noncovalently bound fluorescent labels widens their applicability to biornolecules. The introduction of diode lasers and LEDs emitting tunable light in the NIR region has strongly supported the research into and wide application of NIR dyes in various high technology fields, such as optical data storage, modern reprography, photovoltaic cells, molecular switches, photodynamic therapy, optical sensors and biosensors, and laser fluorimetry. Future developments in the NIR-fluorescing dyes are expected, especially in laser fluorimetry, due to the lack of interference and the potential for high precision in the analysis of biologically important molecules. Thefundamentalspectralpropertiesandtheirapplications may in most cases be highly specialized, and the reader of this review may not be aware of in different fields some aspects. Physicists, chemists, and biochemists working in turnmaynotbe maylackthespecialknowledge of colorchemists,who acquainted with some aspects of their hi-tech and analytical applications. The aim of this review is to partly bridge the gap between these disciplines.

V.

ACKNOWLEDGMENTS

I wish to thank Professor N. Tyutyulkov (University of Sofia, Bulgaria) and Professor F. Dietz (University of Leipzig, Germany) for their kind proposal to the Editor and encouragement to write this chapter. Thanks are due to Dr. K. Harada (Chiba University, Japan)forpresenting methevaluablebooksonvarious

Stoyanov

86

aspects of the chemistry and application of functional dyes. A highly valuable manuscript of a review [ 1341 describing the luminescent probes for near-infrared sensing applications and covering some of the aspects discussed in this chapter was kindly offered by Professor 0. Wolfbeis. The technical assistance of Dr. T. Stoyanova (University of Sofia, Bulgaria) and Dr. P. Akrivos (University of Thessaloniki, Greece) in preparing this manuscript is gratefully acknowledged.

REFERENCES I.

2. 3.

4. 5.

6. 7. 8.

9. IO.

11.

12. 13.

14.

IS. 16.

J Griffiths.Specialty dyes-New trends in moderndyechemistry. J SocDyers Colour 104:4 16424, 1988. H Nakazumi. Organic colorants for laser disc optical storage. J Soc Dyers Colour 104:121-125, 1988. F Jones. Some innovative uses of dyes. RevProgColorRelatTopics19:20-32, 1989. Z Yoshida, T Kitao, eds. Chemistry of Functional Dyes. Tokyo: Mita Press, 1989. J Fabian.HNakazumi. M Matsuoka.Near-infraredabsorbingdyes.Chem Rev 9211197-1226. 1992. Z Yoshida, Y Shirota,eds.Chemistry of FunctionalDyes, Vol 2. Tokyo:Mita Press,1993. M Matsuoka. Dyes for optical recording. Mol Cryst Liq Cryst 224%-94. 1993. H Zollinger. Color Chemistry. Synthesis, Properties and Applications of Organic Dyes and Pigments. 2nd ed. Weinheim: VCH. 199 I . P Gregory. Modern reprographics. Rev Prog Color Relat Topics 24:l-16, 1994. J Griffiths.TheMOdesignanddevelopment of functionaldyematerials.In: Z Yoshida, Y Shirota, eds. Chemistry of Functional Dyes, Vol 2. Tokyo: Mita Press. 1993, pp 1-9. P Gregory. High technology applications of functional dye materials.In: Z Yoshida, Y Shirota, eds. Chemistry of Functional Dyes. Vol 2. Tokyo: Mita Press, 1993, pp 656-664. FV Bright.Bioanalyticalapplications of fluorescencespectroscopy.AnalChem 60:1031A-l039A, 1988. to an GPatonay.MDAntoine.Near-infraredHuorogeniclabels:Newapproach old problem. Anal Chem 63321A-327A. 1991. RBThompson.Redandnear-infraredfluorimetry.In:JRLakowitz.eds. TOPICS inFluorescenceSpectroscopy. Vol 4. Probe Design and Chemical Scnsmg. New York: Plenum Press, 1994, pp IS 1- I8 1. S Stoyanov. Structure and spectral properties of potentially tautomeric systemFundamental and applied aspects. DrSci Thesis, University o f Sotia, Sofia, 1997. FM Hamer. The cyanine dyes and related compounds. In: A Weissberger, ed. The Chemistry of Heterocyclic Compounds. Vol. 18. New York: Interscience. 1964. pp

200-2 13.

17. NTyutyulkov, J Fabian, A Melhorn, F Dietz.ATadjer.Polymethine Structure and Properties. Sotia: St Kliment Ohridski Univ Press, 1991.

Dyes-

Fluorescing Probes: Dyes 18. 19.

20.

21.

22.

23.

24. 25. 26.

27. 28.

29. 30. 31.

in the NIR Region

a7

UTGJones.Chlorophyll in Photochemistry.NewYork:VanNostrandReinhold, 1973, p 75. BL Wheeler, G Nagasubramaanian, AJ Bard, LA Schechtman, DR Dininny. ME Kenney. A silicon phthalocyanine and a silicon naphthalocyanine: Synthesis, elecJ Am ChemSoc 106: trochemistry,andelectrogeneratedchemiluminescence. 7404-74 IO, 1984. PA Firey, WE Ford, JR Sounic, ME Kenncy, MAJ Rodgers. Silicon nephthalocyanine triplet state and oxygen: A reversible energy-transfer reaction. J Am Chem SocI10:762&7630.1988. BD Richter. ME Kenney, WE Ford, MAJ Rodgers. Synthesis and photoproperties of diamagnetic octabutoxyphthalocyanines with deep red opttcal absorbance. J Am Chern Soc I 12:8064-8070, 1990. JM Cook, AJ Dunn, SD Howe, AJ Thomson, KJ Harrison. Octa-alkoxy phthalocyanine and naphthalocyanine derivatives: Dyes with Q-band absorption i n far red or nearinfrared.JChemSocPerkinTrans 1 1988:2453-2458, 1988. M Matsuoka, ed. Infrared Absorbing Dyes. New York: Plenum, 1990. TImasaka,ATsuknmoto.NIshibashi.Visiblesemiconductorlaserfluorimetry. AnalChem 61:2285-2288. 1989. T Imasaka, A Yoshitake. N Ishibashi. Semiconductor laser fluorimetry in the nearinfrared region. Anal Chem 56: 1077-1079, 1984. K Sauda. T Imasaka, N Ishibashi. HPLC detector based on near-infrared semiconductor laser fluorimetry. Anal Chim Acta 187353-356, 1986. Y Kawabata. T Imasaka. N Ishibashi. Ultramicro How-cell for semiconductor laser fluorlmetry. Talanta 33281-283, 1986. PA Johnson, TE Barber, BW Smith, JD Winefordner. Ultralow detection limits for an organic dye determined by fluorescence spectroscopy with laser diode excitation. Anal Chem 61:861-863. 1989. S Soper.QMattingly, P Vegunta.Photonburstdetectionofsinglenear-infrared fluorescent molecules. Anal Chem 65:740-747, 1993. YH Lee, RG Mans, BW Smith, JD Winefordner. Laser-induced fluorescence dctection of a stngle molecule in a capillary. Anal Chem 66:41424149, 1994. K Sauda, T Imasaka, N Ishibashi. Determination of protein in humanserumby HPLC with semiconductor laser fluorimetric detection. Anal Chem 58:2649-2653. 1986.

32. TImasaka,NIshibashi.Diodelasersandpracticaltraceanalysis.AnalChem 62:363A-371A, 1990. 33. MMaeda.LaserDyes:PropertiesofOrganicCompoundsforDyeLasers.New York:AcademicPress,1984. 34. AA Ishchenko. YL Slominskii, AI Tolmachev. Polymethine dyes in quantum electronics.UkrKhimZh 55979-992. 1989. KH Drexhage.Dyestufflasersandlightcollectors-Two 35. RRaue,HHarnisch, newfieldsofapplicationforfluorescentheterocycliccompounds.Heterocycles 21:167-190,1984. 36. MOkawara,TKitao,THirashima,MMatsuoka.OrganicColorants.AHandbook of Data of Selected Dyesfor Electro-optical Applications. Tokyo: Kodansha, 1988.

88

Stoyanov

37. HJ Polland, T Elsaesser, A Seilmeier, W Kaiser, M Kussler.NJ Marx, B Sens, KH Drexhage. Picosecond dye laser emission i n the infrared between 1.4 and I .X pm. ApplPhysB32:53-57.1983. 55. 1993-1994. 38. EastmanKodak.EastmanLaboratoryChemicals.CatalogNo Rochester, NY. 39. P Gregory. Colorants for high technology. In:AT Peters, HS Freeman, eds. Colour Chemistry. The Design and Synthesis of Organic Dyes and Pigments. New York: Elsevier, 199 I , pp 193-223. 40. M Matsuoka. Molecular orbital design. synthesis and characteristics of functional dyesforelectro-opticalapplications.In: AT Peters, HS Freeman,eds.Colour Chemistry. The Design and Synthesis of Organic Dyes and Pigments. New York: Elsevier, 1991, pp 61-84. 41. G Ficken. Sensitising dyes in photography. Chem Ind 1989:672-677. 42. M Matsuoka. Absorption Spectra of Dyes for Diode Lasers. Tokyo: Bunshin, 1990. 43. SA Soper, QL Mattingly. Steady-state and picosecond laser fluorescence studies of nonradiative pathways in tricarbocyanine dyes: Implications to the design of nearIR fluorochromes with high fluorescence efficiencies. J Am Chem Soc 11637443752,1994. 44. RJ Williams,MLipowska.GPatonay,LStrekowski.Comparisonofcovalent and noncovalent labeling with near-infrared dyes f7r the HLPC determination of human serum albumin. Anal Chem 65:601-605, 1993. 45. W Koenig. Uber den Begriff der “Polymethinfarbstoffe.” J Prakt Chem I12:1-36. 1926. 46. S Daehne. Systematik und Begriffserweiterung der Polymethinfarbstoffe. Z Chem 5:44 I 4 5 1. 1965. 47. S Daehen. Der Polymethinzustand. Ber Bunsenges Phys Chem 70:618-625. 1966. in cyaninedyes.In: TS 48. LGSBrooker.Thesearchforlongerconjugatedchains Cote, BS Joshi, SV Sunthaukar, BD Tilak, eds. Recent Progress in Chemistry of Natural and Synthetic Colouring Matters and Related Fields. New York: Academic Press, 1962, pp 573-587. of PolymethineDyes.Kiev:Naukova 49. AI Kiprianov.ColourandConstitution Dumka(inRussian).1979. 50. GE Ficken. Cyanine dyes. In: K Venkataraman, ed. Chemistry of Synthetic Dyes, Vol IV. New York: Academic Press, 197 I , pp 212-324. hin51. AI Kiprianov,GGDyadyusha, FA Mikhailenko.Colourofdyesandsteric drance in the molecules. Usp Khim 35:823-852, 1966. 52. DM Stunner. In: A Weissberger, EC Taylor, eds. Chemistry of Heterocyclic Compounds, Vol 30. New York: Wiley. 1977, p 441. 53. DM Sturmer, DR Diehl. Polymethine dyes. In: M Grayson, D Eckroth. eds. KirkOthmer Encyclopedia of Chemical Technology. 3rd ed. Vol 18. New York: Wiley, 1982, pp 848-874. 54. R Raue. Cationic dyestuffs. Rev Prog Color Relat Topics 14:187-203, 1984. 5s. R Raue. Methinfarbstoffe. In: F Ullmann. ed. Ullmann’s Encyclopedia of Industrial Chemistry. 5th ed. Vol A16. Weinheim: VCH, 1990, p 487. MA Kudinova.Synthesis of polymethinedyes 56. AI Tolmachev,YLSlominskii, absorbing in the near-IR spectral region. Usp Nauchn Fotogr 22: 12-27, 1984.

Fluorescing Probes: Dyes

in the NIR Region

89

57. VA Mostovnikov, AN Ruhinov, MA Al'perovich, VA Avdeeva, I1 Levkoev. MM Loiko. Dependence of luminescent and generation properties of polymethine dye solutions on their structure. Zh Prikl Spektrosk 20:4247. 1974. 58. LStrekowski,MLipowska, G Patonay.Substitutionreactionsofanucleofugal group in heptatnethine cyanine dyes. Synthesis of an isothiocyanato derivative for labeling of proteins with near-infrared chromophore. J Org Chem 57:4578-4580, 1992. 59. MA Kudinova, YL Slominskii, VL Shokodko, AI Tolmachev. a-Thiopyrylopentacarbocyanines. Ukr KhimZh 519-96. 1985. 60. SM Makin, TN Boiko, AI Ponomarev. Synthesis and study of tetracarbocyanine dyes containing cyclic fragments in the polymethine chain. Zh Org Khim 24:410415,1988. 61. W Freyer,F Fink. Saltigbarer Absorber auf Polymethinaggregatbasis fur den nahen IR-Bereich. Z Chem 29: 105-107, 1989. Nonamethin-Cyaninfarbstoffe. Chem 62. G Hellig, W Luettke. Synthese planar fixierter Ber119:3102-3108.1986. Heptamethin-Cyaninfarbstoffc. 63. G Hellig,WLuettke.Syntheseplanartixierter Chem Ber 120:1863-1 866, 1987. 64. AA Ishchenko,MAKudinova, YL Slominskii, AI Tolmachev.Pyrylopolycarbocyanines with cyclic bridging groups. Zh Org Khim 22:170-179, 1986. 65. MA Kudinova, VV Kutdyukov. AA Ishchenko, AI Tolmachev. Sytnmetrical pyryIocyanines based on3,4-polymethylene-2,6-diphenylpyryliumsalts. Khitn GeterotsiklSoed1984:451-454. 66. G Hekg, WLuettke.UntesuchungenzurSpectroscopieundPhotostabilitat konformativtixierter Monomethin-Cyaninefarbstoffe. ChemBer I2 1 :407-410, 1988. 67. RB Mujuumdar, LA Ernst, SR Mujuumdar, CJ Lewis, AS Waggoner. Cyanine dye labeling reagents: Sulfoindocyanine succinimidyl esters. Bioconj Chem 105-1 4: 1 I . 1993. 68. RL Southnick, LA Ernst, EW Tauriello. SR Parker, RB Mujuumdar, SR Mujuundar, HA Clever, ASWaggoner. Cyaninedyelabelingreagents-Carboxymethyl indocyanine succinimidyl esters. Cytometry 11 :418430, 1990. 69. GA Casay,T Czuppon, M Lipowska.G Patonay. Near-infrared fluorescence probes. SPIEProc1885:324-336,1993. 70. G Patonay,MDAntoine, S Devanatthan,LStrekowski.Near-infraredprobefor determination of solvent hydrophobicity. Appl Spectrosc 45:457461, 1991, 71. L Strekowski. M Lipowska, G Patonay. Facile derivatization of heptamethine cyanine dyes. Synth Commun 22:2593-2598, 1992. 72. LStrekowski,MLipowska. G Patonay.Substitutionreactions of anucleofugal group in heptamethine cyanine dyes. Synthesis of an isothiocyanato derivative for labeling of proteins with a near-Infrared chromophore. J Org Chem 57:45784580, 1990. 73. AE Boyer, M Lipowska, J Zen, G Patonay. Evaluation of near-infrared dyes as labels for immunoassay utilizing laser diode detection: Developmentof near-infrared dye (NIRDIA). Anal Lett 25:415-428, 1992.

Stoyanov

90

74. MLipowska,GPatonay.LStrekowski.Newnear-infraredcyaninedyes for labelling of proteins. Synth Commun 23:3087-3094, 1993. 75. WRGBaeyens.DDeKeukeleire,KKorkidis, eds. LuminescenceTechniquesin Chcmical and Biochemical Analysis. New York: Marcel Dekker, 1991. 76. LA Ernst, RK Gupta. RB Mujumdar, AS Waggoner. Cyanine dye labeling reagents for sulfhydryl groups. Cytonletry 10:3-10, 1989. 77.RBMujumdar,LAErnst,SRMujumdar.ASWaggoner.Cyanmedyelabeling reagents containing isothiocyanatc groups. Cytometry 10: I 1-19, 1989. 78.AGrinwald,RHildeshcim, JC Farber,LAnglister.Improvedfluorescentprobes for the measurement of rapid changes in membrane potential. Biophys J 39:301308. 1982.

79. S Dachne.FMoldenhauer.Structuralprinciples of unsaturatedcompounds:Evidence from quantum chemical calculations. Prog Phys Org Chem 15: 1-130, 1985. 80. J Fabian. H Hartmann. Light Absorption of Organic Colorants. Theoretical Trcatmen1andEmpiricalRules.Berlin:Springer-Verlag. 1980, pp 162-197. 81. NJLRoth, AC Craig.Predictedobservablefluorescentlifetimes of severalcyanines. J Phys Chem 78:l 154-1 155, 1974. 82. NV Monich. AF Vompe, SM Makin, DA Shavryglns, II Levkoev. Symmetrical unsubstituted i n the chain hexa- and heptacarbocyanine dyes. Zh Org Khim 21 : 10931097,1985. I n Russian. 83. SS Malhorta, MC Whiting.Thepreparationandelectronicabsorptionspectraof homologousseriesofsimplecyanines,merocyaninesandoxonols.JChenlSoc 196013812-3822, 1960.

84. S Daehne, 0 Gurtler.Farbe.StabilitatundHalbstufenpotentalevonPolymethinfarbstoffen. J PraktChem315:786-790,1973. 85. JP Fouassier.DJ LOII~IIO~, JFaure.Transientabsorption 111 apolymcthinelaser dye. Chem Phys Lett 35: 189-194, 1975. 86. AM Kolesnikov. FA Mikhailenko.Conformationofpolymethinedyes. Usp Khim 56:466488, 1987.

87. D Doizi, JC Mialocq. Photosensitized electron-transfer reaction in the first excited singlct state of a polymethine-cyanine dye. J Phys Chem 9 1:3524-3530, 1987. 88. HKuhn.DMobius.SystemeausmonomolecularenSchichten-Zusalnmenbauund Physikalish-Chemisches Verhalten. Angew Chem 83:672-690, 197 I . 89. DM Sturmer. DW Hesseltine.TheTheoryofPhotographicProcess.4th ed. New York:Macmillan.1977. 90. KVFedotov, NN Romanov.Polymethinedyeswith 3-oxo-2.3-dihydrothiazolo [3,2-alpyrimidiniunl end group. Ukr KhimZh 52:514-519. 1986. 9 1. 11 Boiko, NA Derevyanko, AA Ishchenko, TA Markina, AI Tolmachev. Pyrylo-2carbocyanines with substituents in the heterocyclic end groups. Khim Geterotsikl Soed 1986:1607- 16 13. 92. AA Ishchenko, NA Derevyanko, VM Zubarovskii, AI Tolmachev. Influence of the polymethine chain length on the shapeof absorption bands in symmetrical cyanine dyes. Teor Eksp Khirn 20:44345 I . 1984. 93. AA Ishchenko, VA Svidro, NA Derevyanko.Solvntochromism of thecationcyaninedyes.DyesPigm10:85-96.1989.

Fluorescing Probes: Dyes

in the NIR Region

91

94. W West, S Pearce. The dimeric stateof cyanine dyes. J Phys Chem 69:18941903. 1965. 95. S Makio, N Kanamaru,J Tanaka. The J-aggregate, 5,5’,6,6’-tetrachloro- I ,l’-diethyl3,3’-bis(4-sulfobutyI)-benzimidazolocarbocyaninesodium salt in aqueous solution. Bull Chem Soc Jpn 53:312&3124. 1980. 96. F Heisel. JA Miehe, J Rachidi. The dynamics of the intramolecular and orientational motions. Chem Phys Lett 134:379-384, 1987. 97. HP Dorn, A Muller. Temperature dependence of the fluorescent lifetime and quantum yleld of pseudoisocyanine monomers. Chem Phys Lett 130:42643 I , 1986. 98. M Kasha. Energy transfer mechanisms and the molecular exciton model for molecular aggregates. Radiat Res 20:SS-70, 1963. 99. F Rotermund,RWeigand,WHolzer,MWittmann.APenzkofer.Fluorescence in water.JPhotochem spectroscopyanalysisofindocyaninegreenJ-aggregates PhotobiolA 1 10:75-78.1997. N Narayanan, S Sutter.LStrekowski, 100. DBShealy.MLipowska.JLipowski, of nearGPatonay.Synthesis,chromatographicseparation.andcharacterization infrared-labeled DNA-oligomers for use of DNA sequencing. Anal Chern 67:247251,1995. 101. TG Dyadyusha, MN Ushomirskii, VN Romanov. YP Kovtun. Topological analysis of the electronic spectraof unsymmetrical polymethine dyes. Ukr Khim Zh 54:6366, 1988. 102. C Reichardt. Solvents and Solvent Effects in Organic Chemistry. 2nd ed. Weinheim:VCH,1990. 103. LJEHofer, RJ Grebenstetter, EO Wiig. Thefluorescence of cyanlneandrelated dyes in the monomeric state. J Am Chem Soc 72:203-209. 1950. 104. OH Chen. CW Tang. Design and synthesis of red dopants for electroluminescence. I n : Z Yoshida, Y Shirota, eds. Chemistry of Functional Dyes. Vol 2. Tokyo: Mita Press, 1993, pp 536-543. 1 05. R Gugliemetti. Spiropyranes and related compounds. In: H Durr, H Bouas-Laurent, eds. Photochromism: Molecules and Systems. Amsterdam: Elsevier, 1990. pp855878. 106. I Willner.Photoswitchablebiomaterials: En routetooptobioelectronicsystems. Acc Chem Res 30:347-356. 1997. 107. R Gugliemetti. Transition from photochromic spiropyranes to spirooxazines. In: Z Yoshida, Y Shirota, eds. Chemistry of Functional Dyes. Vol 2. Tokyo: Mita Press. 1993. pp 331-338. 108. AJGMank,HTCvan derLaan,HLingeman,CGooijer, UATBrinkman.NH Velthorst. Visible diode laser-induced fluorescence detection in liquid chromatogof amines.AnalChem67:1742-1748. raphyafterprecolumnderivatization 1995. 109. K-L Law. Squarame chemistry. Effects of saturated changes on the absorption and multiple fluorescence emission of bis[4-(dimethylamino)phenyl]squaraine and its derivatives. J Phys Chem 9 I :5 184-5 193, 1987. I 10. RA Nathan. AH Adelman. Photosensitized generation of singlet molecular oxygen wlth near infrared radiation. J Chem Soc Chem Commun 1974:674-675.

92 111.

112. 113.

114.

I 15.

116.

117. 118.

119. 120. 121. 122.

123.

124. 125.

126.

127. 128. 129.

Stoyanov K Namba.Metalcomplexdyes.In:MMatsuoka. ed. InfraredAbsorblngDyes. New York: Plenum Press, 1990, pp 57-70. UT Mueller-Wersterhoff. B Vance. D1 Yoon. The synthesis of dithiolene dyes with strong near-IR absorption. Tetrahedron 47:909-932, 1991. VH Grif, VE Lavrushin. Colour and constitution of symmetrical polymethine dyes with end aromatic rings. Ukr Khim Zh 535366-870, 1987. MIDemchuk.AAIshchenko,VPMikhailov,VIAvdeeva.Theinfluenceofthe anionontheexcitedstaterelaxationtimeofcationicpolyrnethinedyes.Chem PhysLett144:99-103,1988. THigshijima,TFuchigami,TImasaka,NIshibashi.Determination of amino acids by capillary zone electrophoresis based on semiconductor laser fluorescence detection. Anal Chem 64:7 I 1-7 14, 1992. H Zollinger. Logic, psychology and serendipity in scientific discoveries in dyestuff chemistry. In: Z Yoshida, T Kitao, eds. Chemistryof Functional Dyes. Tokyo: Mita Press,1989,pp 1-8. GBooth.Phthalocyanines.In:KVenkataraman. ed. The Chemistry of Synthetic Dyes, Vol V. New York: Academic Press, 197 I , pp 241-282. KM Smith. Porphyrins and Metalloporphyrins. NewYork:Elsevier.197.5.p 10. SA Mikhalenko, EA Luk'yanets. Synthesis and some properties of 2.3-naphthalocyanines. Zh Obshch Khim 39:25.54-2558, 1969. fur dieoptischeDatenMEmmelius,GPawlowski,HWVollmann.Materialen speicherung. Angew Chem 101: 1475- 1502, 1989. LELyons.JRWalsh,JWWhite.Phthalocyaninecrystalspectra.JChemSoc 1960:167-175. TC Van Cott, JL Rose, GC Misener, BE Williamson, AE Schrimpf, ME Boyle, PN Schatz. MCD and absorption spectrum of zinc phthalocyanine in an argon matrix between 14700 and 74000 cm". J Phys Chem 93:2999-301 I , 1989. D Dolphin, E Sternberg. Medical applications of dyes: Review of photodynamic therapy. In: Z Yoshida, T Kitao, eds. Chemistry of Functional Dyes. Tokyo: Mita Press, 1989, pp 587-597. M Yousaf, M Lazzouni. Formation of an invisible infrared printing ink. Dyes Pigm 271297-303,1995. GA Casay.NNarayanan.LEvans.TCzuppon,GPatonay.Near-infraredtetrasubstitutedaluminium2,3-naphthalocyaninedyesforoptical-fiberapplications. Talanta 43:1997-2005, 1996. GACasay,FMeadows,NDaniels, H Robertson, G Patonay.Fiberopticprobe applications using near-infrared compounds-Determination of NaOH. Spectrosc Lett28:301-326,1995. OS Wolfbeis.Longwaveluminescentporphyrin DBPapkovsky,GVPonomarev. probes. Spectrochim Acta A 52: 1624-1638, 1996. DB Papkovsky. GV Ponomarev, OS Wolfbeis. Protonation of porphyrins in liquid PVCmembranes:Effect of anionicadditivesandapplicationtopH-sensing. J PhotochemPhotobiol104:1.51-158,1997. H He, G Uray, OS Wolfbeis. Enantioselective optodes. Anal Chim Acta 246:251257,1991.

Probes: Dyes Fluorescing in

the NIR Region

93

130. G Seibold, G Wagenblast, S Haremsa. New peryleneandviolanthronedyestuffs Z Yoshida, T Kitao,cds.Chemistry of Functional forfluorescentcollectors.In:

Dyes. Tokyo: Mita Press, 1989, pp 553-555. FOHoltrup,GRJMuller,HQuante, S deFeyter,FCdeSchryver, K Mullen. Terrylenimides: New NIR fluorescent dyes. Chem Eur J 3:219-225, 1997. 132. HLanghals, G Schonman,LFcilcr.A2-stepsynthesisofquaterrylenetetracarboxylic bisimides. Tetrahedron Lett 36:6423-6424, 1995. 133. JN Miller, MB Brown, NJ Seare, S Summerfield. Analytical applications of very near-IR fluorimetry. In: OS Wolfbeis, ed. Fluorescence Spectroscopy-New Methods and Applications. Berlin: Springer-Verlag. 1993, pp 189-196. 134. E Terpetschnig. OS Wolfbeis. Luminescent probes for NIR sensing applications. In: K Daehne, U Resch. OS Wolfbeis,eds.Synthesis,OpticalPropertiesand Applications of Near-Infrared Dyes in High Technology Fields. Dordrecht: Kluwer Academic.1998.pp161-182. 131.

This Page Intentionally Left Blank

Analytical Applications of Near-Infrared Fluorescence in Immunoassays Anand R. Swamy, Lucjan Strekowski, and Gabor Patonay Georgia State University, Atlanta, Georgia

1.

INTRODUCTION

Classicalbiomoleculeidentificationusuallyinvolvesseparationof a complex mixture followed by analysis of the separated fraction. No single test provides the definitive identification of an unknown biomolecule; hence a complex series of tests are required. This process is oftentime-consumingandnotavailable on thetimescaledesired in clinicallaboratories.Thesefactorsandtheneed to be able to adapt to primitive field test conditions have prompted many researcherstoexploremoderninstrumentalalternativestoclassicalprocedures. The characteristics of mostmodernanalyticaltoolsincluderapiddataacquiA sition, reproducibility, computer-aided instrument control, and data analysis. number of moderninstrumentaltechniqueshave also beenadaptedfor field applications. a significant Thelasttwodecades ofthetwentiethcenturywitnessed amount ofresearchandprogress in analyticalmethoddevelopment,resulting in several Nobel prizes being awarded in this field. The techniques developedDNA sequencing, polymerase chain reaction (PCR), X-ray crystallography, capillaryelectrophoresis (CE), circulardichroism (CD), andseveral others-also of biologicalsystems.Oneof provided a deeperinsightintothefunctioning the most interesting of these is a radioimmunoassay for thyrotropin releasing hormone developed byYalow. Guillemin, and Schelly [l]. The most valuable characteristics of this method are its very high sensitivity and specificity, which 95

Swamy et al.

96

are attributed to antibodies that constitute its principal analytical reagent. This in turn opened the doors for application of this methodology in various fields includingclinicaldiagnosisandenvironmentalanalysisofbiologicalwarfare agents, pesticides, and other hazardous substances. The generic term "immunodiagnostics" is often used to describe methods that employ antibodies as analytical reagents for various applications. This chapter describes the development of an application that utilizes the of immunoassays coupled advantages offered by the specificity and sensitivity withthelowbackgroundnoise of near-infrared(NIR)fluorescence.Tofully understand and appreciate this methodology, one needs appropriate background information in the fields of immunology and NIR spectroscopy. While this chapter is notintended to beanexhaustivereview of thesetwoareas, a general overview of related terms and definitions is presented in the following section.

II. OVERVIEW OF IMMUNOLOGY The principal function of the immune system is to protect the host animal from infectious organisms and their toxic products. The immune system hasevolved a wide range of mechanisms to locate these foreign organisms (antigens) and clear them from the body. The surveillance mechanisms involve proteins and cellsthat circulate throughout the body [ 2 , 3 ] .The two broad categories of the surveillance I ) . Nonadaptive mechanisms include adaptive and nonadaptive immunity (Fig. immunityinvolves cells that nonspecifically respond to antigens. This type of immunity constitutes phagocytosis by macrophages, secretion of lysozyme by lacrimal cells, and cell lysis by natural killers. Adaptive immunity, on the other hand, is directedagainstspecificmoleculesand is enhanced by repeatedexposure. It involves the plasma cells and the precursor B-cell lymphocytes that secrete antibodies in response to antigens. Typically, to induce antibody production the molecular weight of the invading antigen has to be about 6000 Da or higher. Molecules that can induce antibody response are also called immunogens [4-71. The antibody produced is highly specific against the antigen that evokes its production ( K D zz 1O6-IO"'). The antibody forms a complex with the antigen,and the complex is then cleared by the macrophages. The high specificity of the antibodies makes them invaluable analytical reagents in immunological research and clinical diagnostics. A description of some of the common structural features of antibodies follows.

A.

Antibodies

Antibodies are lnolecules produced by the plasma cells of the infected host in response to theinvadingantigen.Antibodiesaremembers of alargefamily

unoassaysNIR in Fluorescence

97

Antibody with high specificity

Adaptive

IIIUtlUIlt:

response

respons

1 Nonadaptive response

by

Figure 1 Immune response.

of mildly glycosylated proteins called immunoglobulins. Immunoglobulins are a diverse group of proteins that share key structural and functional features. Since several reviews on their structure have been published [8-IO], only a brief overview of immunoglobulins pertinent to this chapter is presented below. Functionally, antibodies have the ability to bind specific antigens and specializedcells or proteinsoftheimmunesystem.Structurally,antibodiesare tertiary proteins composed of a characteristic Y-shaped unit (Fig. 2). Based on their number of Y-shaped units and carbohydrate content, antibodies are further grouped into five classes: immunoglobulins G (IgG), M (IgM), A (IgA), E (IgE), D (IgD). In a typical primary immune response upon first exposure to the antigen, initially a relatively high level of IgM is produced, and after the booster dose or second exposure, IgG dominates (Fig. 3). ImmunoglobulinGisthemajorimmunoglobulinfound in mammalian serum and the focus of the applications discussed in this chapter. It serves as a model for other immunoglobulins. It contains a single Y-shaped unit. The IgG molecule has two light (L, approx 25,000 Da each) and two heavy (H, approx 50,000 Da each) chains. The light and heavy chainsare held together by disulfide ( S - S ) linkages and hydrophobic interactions [ 11-13]. There are additional disulfide linkages and hydrophobic interactions between the two heavy chains. The two arms of theY-shaped molecule are also known as the FAB region, (Fc and the stem of the Y-shaped molecule is called the crystallizable fragment region). The Fc region is composed of constant regions of the H chain and has

98

Swamy et al.

region rand

binding

Figure 2

Antibodystructurc.

10 Figure 3

Ti"I-E (days)

Kinetics of typicalantibodyproduction.

NIR Fluorescence in Immunoassays

99

little effect on the antigen binding but plays a key role in directing the biological activity of the molecule. The FAB region has two binding sites for antigen at theend of thetwoarms of the Y unit(about 100 N-terminalaminoacid sequence). This region is also called the variable region[4].Withinanintact of both the L and H chains are antibody molecule, variable region sequences structurally associated to form the antigen binding site.

B. Coupling Reactions for Labeling Antibodies Immunoassays rely on the specific reaction between antibodies (Ab’s) and antigens (Ag’s). This reaction essentially yields anAg-Ab complex that is invisible. To observe and measure this reaction, a label is usually introduced via a secondary Ab (referred to as anti-Ab) that is specific for the first (primary) antibody. The different typesof labels commonly used are radioactive isotopes, fluorescent dyes, chemiluminescent substances, and enzymes. This section describes some of the approaches that are commonly used for labeling. Untilrecently,noncovalentlabeling of proteinswasthemostcommon method used. The label is attached to the Ab by hydrophobic interactions. Two major advantages of this method are that labeling is done at physiological pH andthelabeldoesnotaffectthefunctionalactivityoftheAb.Thismethod, however, has many limitations in immunoassay applications and is rarely used. as lowbindingconstants of thelabel toAb Itsuffersfromdrawbackssuch and nonspecificity of the label. Hence, it often poses problems in actual assay conditions, as the label can associate with proteins other than the antigen, giving falsesignals. In contrast,covalentlabeling is morespecificandreliable,and labels with a variety of functional groups can be used in this approach. Williams et al. [ 141 compared the merits of covalent and noncovalent labeling of proteins. The methodologies for various covalent coupling reactions are well established [ 15-17]. A very brief overview of the most commonly used covalent coupling procedures follows. The amino groups of the Ab molecule serve as an ideal target for coupling reactions with appropriately functionalized labels. The amino groups react readily with N-hydroxysuccinimide ester (NHS ester) on the labels to form a stable amide bond between the label and Ab (Fig. 4a). The coupling reaction is efficient at concentrations greater than 2-3 mM at neutral pH and occurs at even lower concentrations at a higher pH (optimal pH M 9-10). A possible side reaction is coupling of the NHS ester with cysteine residues on the Ab molecule to form a thioester-linked label. In another method, the amino groups on Ab are allowed to undergo a condensation reaction with an aldehyde group on the label to form a Schiff base. Although Schiff bases themselves are not very stable, they can be reduced to form stable amino linkages (Fig. 4b). The amino groups can also be acylated by reaction with acid anhydride substituted labels (Fig. 4c). In

100

Swamy et ai.

N-Succlnimldylcarboxylate

Amine

on protem

0 HJLl

+

Aldehyde

R~-NH~

*

R'-N*R' H

(b)

Schiff base

Amlne on proteln

NaBH,CN 1

R'-N-R' H

R3

0

Amme on protem

Acid anhydrlde

S +

lsothlocyanate

R'-NH~

+

R'-N~N-R' H H

(dl

protem Amine on

Figure 4 Chemical reactions commonly used for labeling proteins with a label R ' .

addition to the reactive groups mentioned above, the isothiocyanate functionality on the label can react readily with primary amines on the Ab to form a stable thiourea linkage. Optimal pH for this coupling is typically around pH 9.5-10 (Fig. 4d). In determining a suitable method for coupling, factors such as the stability oftheAbandthefunctionalizedlabelatthe couplingpH,themolarratio, reaction temperature, and reaction time are critical and must be optimized. One of the most critical concerns is to ensure that the labeling of the Ab does not

NIR Fluorescence in Immunoassays

101

affecttheAgrecognitionsites. By randomlylabelingtheAbmolecule,there is always a risk of affecting the Ag recognition site, and this is especially true if a large number of labels are attached to the Ab. It is therefore necessary to optimize the number of labels per antibody so as to maintain the specificity of of the the Ab for the Ag. Thus, an ideal method would involve selective coupling label farthest from the Ag recognition site. As mentioned earlier, the FC region serves no specific purpose in Ag recognition and hence would be the best site to label. It contains glycoproteins that can be oxidized to yield aldehyde groups, which in turn can be used for selective coupling to labels by using appropriate chemistry.

C.

Immunoassays

Immunoassay (IA) is a method of analysis that relies on specific interactions between antibodies (Ab’s) and antigens (Ag’s) to detect a variety of substances, ranging from complex viruses and microorganismsto simple pesticide molecules and industrial pollutants. The IA techniques can be qualitative, semiquantitative, or quantitative. These assays have become firmly established in varied applications, especially in the biological and environmental fields. The use of antibodies as analytical reagents was first reported in 1959 when Berson and Yalow sucof picogram levels of human insulin cessfully demonstrated the measurement in samples of body fluids by radioimmunoassay (RIA) 1181. Since then, various 1As for detecting hundreds of molecules of endogenous and exogenous origin havebeen described.Thismethodology proved to bereliable,fast,andvery sensitive; many other RIAs have been developed for clinical and medical tests since then. Radiolabels were gradually replaced with enzyme labels because of the hazardsassociatedwiththeuseofradioactivematerials.Enzyme-linkedimmunosorbent assay (ELISA), which wasfirst introduced by Engvall and Perlman in 1971 [19]. has become perhaps the most popular IA format in laboratories. The modern diagnosis of many diseases, especially infectious ones, is almost completely dependent on these assays. In diseases of global importance, such as acquired immunodeficiency syndrome (AIDS) [20), cysticercosis [21], malaria [22],filariasis,andschistosomiasis 123-261, whichaffectmillions of people, immunoassays play a key role in screening and confirmatory diagnosis. The use of immunochemicaltechniques in theenvironmentalfieldwas first proposed i n 1971 by Ercegovich [27], who suggested the use of immunological screening methods for the rapid detection of pesticide residues and for confirming results of conventional analyses. An RIA for the insecticides aldrin and dieldrin was the first reported IA for environmental contaminants [28]. AIthoughafewRIAs still exist in themedical field, they areseldomused in environmental and food analysis because of the need for special handling and disposal of the radioactive materials.

Swamy et al.

102

Thegrowth in theimmunoassay field hasbeenvirtuallyexplosive,yet there is a constant quest for improvement in sensitivity, specificity, assay efficiency, ability to quantify low analyte levels, and field applicability. Specificity is a function of the immunological reaction and is independent of the assay or label type. Sensitivity is determined by the signal-to-noise ratio of theassay, which in turn is dependentontheefficiencyofthelabelanditssusceptibility to background interference. Assay efficiency is determined by speed, labor intensiveness, reagent costs, and environmental impact. Immunoassayscanbedividedintoseveralgroupsaccording to type of analysis, test sample, assay system, and assay conditions. The discussion presented below is limited to the types of assays formats frequently used, namely homogeneous or heterogeneous assay formats and competitive or noncompetitive assay formats. 1.

HomogeneousandHeterogeneousimmunoassays

In homogeneousimmunoassays,the Ag-Ab bindingcanbedetermined by changes in marker molecules without the physical separation of the bound and free components. Several methods have been reported to monitor these changes. These include detection of a conformational change of the complex, fluorescence energy transfer, inhibition of enzymatic activity upon binding and polarization, and alteration of enzymatic activity (Fig. 5). Heterogeneousimmunoassays,ontheotherhand,requireanadditional step to separate the bound and free analyte. Even though this increases the time of analysis,theseparation of boundfromfreeanalyteallowstheremovalof interfering substances and any excess Ag or Ab from the detectable complex. The method most commonly used for separating bound from free analyte is to wash off the unbound analyte with buffer containing detergent (usually Tween 20). Washingalsoreducesnonspecificbinding,whichresults in animproved detection limit and working range. Both of these factors are essential to optimization of the immunoassay design. The sensitivity of the heterogeneous assay can be improved by using a larger sample. In comparison with homogeneous assays, heterogeneous assays are more versatile and are more widely used. Homogeneous assays are limited to detecting only small molecules. Details on the various formats of heterogeneous assays are outlined in the next section.

2. CompetitiveandNoncompetitiveAssays Heterogeneous assays can be carried out in either a competitive or noncompetitive format. Competitive assays can be performed in the antigen-coated (Fig. 6) or antibody-coated format (Fig. 7). In the former analysis, labeled and unlabeled Ab molecules compete for the same epitope on a limited number of antigen sites. In this method, usually a standard curve is plotted for increasing amounts of the

103

NIR Fluorescence in Immunoassays

U Enzyme-labeled antigen

U

' +Y

i

Labeled antigen (rotates faster)

.

Fluorescence-labeled antigen + antibody (rotates slower)

A

Antibody

(b) Figure 5 Detection of Ag-Ab interaction in homogeneous immunoassays. (a) Change in enzyme activity; (b) change in fluorescence polarization.

unlabeled Ab, keeping the labeled Ab constant. The concentration of Ab in the unknown sample is determined by comparing the signal obtained from the bound labeled Ab with the standard curve. In this assay, the signal is inversely proportional to the analyte concentration. Competitive assays can also reveal other valuable information of theAg-Ab binding. For instance, in the case of specific binding, there is competition between the labeled and unlabeled antibodies. On the other hand, if the binding is nonspecific, there is no competition observed

Fix antigen to support

Add labeled antibody

(1 1

(2)

Add unlabeled antibody from sample to compete

(3)

Figure 6 Competitive immunoassayinantigen-coatedformat.

wash off unbound antibodies and quantify (4)

104

Swamy et al.

Fix antibody to support

Add labeled antigen

Add unlabeled antigen to compete

Wash off unbound antigen and quantify

(1)

Figure 7 Competitive immunoassay in antibody-coated format.

since the antibodies bind to two different epitopes on the antigen. Competitive assays have the advantage that only small amountsof analyte are required. This is particularly advantageous where the costs of producing and purifying antibodies are high. In a noncompetitive assay, the binding of limiting amounts of test Ab to an excess amount of Ag is analyzed. The assay can be performed in various formats: the director indirect method and the capture format or sandwich format (Figs. 8-1 1). It should be noted that the capture format is also commonly used for competitive assays. In the most commonly used indirect immunoassay for antibody testing, the solid support is coated with saturating amounts of Ag and then the unbound Ag is washed off. The test sample with Ab is introduced, and the excess unbound sample (Ab) is washed off. In the final step, a labeled Ab

(1) Fix antigen to solid phase

.

(2) Add labeled

antibody

-&,:A,,,3

” ,

(3) wash off unbound antibody & quantify

Figure 8 Direct immunoassay in antigen-coated format.

105

NIR Fluorescence In Immunoassays

(1) Fix antigen to

solid

(2)

phase

Add primary antibody & wash off unbound

(3) Add labeled secondary

$,z,z,-k

antibody against primary antibody

(4) Wash off unbound antibody & quantify

Figure 9 Indirectimmunoassayinantigen-coatedformat.

Fix secondary antibody on solid phase (1)

AddprimaryWashtoremove antibody unbound (2) (31

U

'I

Add labeled antigen Wash to remove unbound (4) quantify and antigen (5)

Figure 10 Indirectimmunoassayincaptureformat.

antibody

106

Swamy et al.

D

m D D

Coat wells with antibody (1)

Add antigen (2)

Add labeled antibody Wash against different antibody epitope on the antigen

Wash off unbound antigen (3)

off unbound & quantify (5)

(4) Figure 11 Heterogeneous immunoassay in sandwich format.

directed against the primary Ab is introduced and excess unbound labeled Ab is washed off, followed by quantificationof the bound labeled secondary antibody. The advantages of noncompetitive assays are short incubation times and good sensitivity. One of the disadvantages of this format is that excess reagent is used and can be expensive. The standard curve for the noncompetitive assay is the opposite to that of the competitive assay in that the signal increases with increasing amounts of analyte. The separation of the bound from the free analyte is a crucial factor in attaining good sensitivity in a heterogeneous assay [29]. This separation is enhanced when a solid-phase support is used, and the efficiency is associated with the ease with which the unbound analyte can be washed away. In solid-phase assays, however, the kinetics of Ag-Ab binding is significantly different from the liquid-phase kinetics. The kinetics of Ag-Ab interaction at solidfliquid interfaces was reviewed by Sternberg and Nygren [30]. Cell surface interactions with Ab are not normally diffusion-limited. However, reactions at solidfliquid interfaces can be diffusion-limited due to depletion of reactants close to the surface, the effects being dependent on molecular geometry, intrinsic reaction rate, and surface concentration of receptor molecules. In the solid-phase assay,

NIR Fluorescence in Immunoassays

107

the diffusion contribution is negligible, as the collision distance between Ag and AB is increased. This problem can be overcome by agitation of the solid phase, thus compensating for the difference and improving the overall reaction rate.

D. Isotopic and Nonisotopic Labeling in Immunoassays As mentioned earlier, the principal reaction in an immunoassay occurs between antibody (Ab) and antigen (Ag), yielding an Ag-Ab complex. To observe and measure this reaction, a label is introduced via a second Ab which allows the complexation to be quantified. Conventionally, this label could be a tagged radioactive isotope (followed by RIA), an enzyme (ELISA), or a fluorescent (FIA) or luminescent molecule 13I ] .

1.

Radioimmunoassay

In radioimmunoassay (RIA) the introduced label is a tagged radioactive isotope, usually 12sI. Advantages of RIAincludenegligiblebackground in biological samples and the fact that the label is not affected by surrounding environmental factors such as pH, molarity, substrate, and inhibitors, many of which can interfere with alternative labels. Despite the potential theoretical sensitivity of RIA, experience has shown that the competitive RIA assay does not permit analysis of analyte at concentrations less than 10"' mol/L (about 10" molecules/L) and the threshold is an order of magnitude lower in noncompetitive assays [32]. Although most of the immunoradiometric methods employ '"1 as a label, it is by nomeans an ideal tracer. Thesignalprovided by '"1 representsone detectable event per second per 7.5 x 10' molecules, so that only 0.000013% of thetracerisseenwithinacountingtimeof 1 sec [31]. In addition,RIA is plagued with problems such as short shelf life, cost-prohibitive instrumentation for detection of signal, labor intensiveness, and potential environmental and health hazards in the handling and disposal of radioisotopes. To overcome the problems associated with radioisotopes, nonisotopic labels drew tremendous attention. The impetus to develop nonisotopic assays was furthered by interests in the development of analytical tools that are quick, easy to use, and cost-effective and that could be adapted to perform assays in the field environment. The development of nonisotopic labels requires these labels to match RIA in terms of specificity and sensitivity. This, in turn, requires a higher number of labels per antibody without compromising Ab activity. Ease of coupling the label to Ab and stability of the conjugate are also important. The labels should be amenable of handling). While to automated application and must be nontoxic (for ease none of the nonisotopic labels currently available is ideal in all aspects, each hasitsowncharacteristicadvantages,makingthemsuitablefordeveloplnent of immunoassays. The most popular nonisotopic labels are enzyme labels and fluorescent labels.

108

2.

Swamy et al.

Enzyme-LinkedlmmunosorbentAssay

In enzyme-linked immunosorbent assay (ELISA), the most commonly used nonisotopic immunoassay, the introduced label is in the form of an enzyme. The immunological reactions are measured indirectly through the enzymatic activity of the label, which is often demonstrated by the absorbance of a chromogenic substrate. The signal obtained is highly dependent on the conditions of the substrate incubation step. Several factors such as temperature, time, and concentration of the substrate affect the overall sensitivity. High amplification of signal by the enzyme is one of the major advantages of this system. The practical detection limit of the horseradish peroxidase enzyme with 3,3',5,5'-tetramethylbenzidine (TMB) as a substrate is 10"4-10"7 m 01 [33]. One of the major factors limiting the sensitivity is therelativelylargemolecular size of the enzyme label (-60,000 Da). This in turn causesenzymeconjugates to diffuseat a slower Also, thebulk of an enzyme canafrateandincreasesnonspecificbinding. fect the antigen-antibody binding due to steric repulsions. Smaller labels would allow for more reporter molecules per antibody and help improve detection limits without affecting the antigen-antibody interactions. Though when properly stored enzymes can retain activity foryears,theyaresusceptible to environmental effects. In addition, the usefulness of ELISA for a primitive field test is limited.

3.

Fluorescence Immunoassay

Fluorescence is more sensitive and selective than absorbance as a spectroscopic tool. Theoretically, fluorescent labels have the potential for the ultimate sensitivity of single molecule detection. The properties of an ideal fluorescent label are summarized as follows: The chromophore should exhibit a high molar absorptivity and a relatively high quantum yield of fluorescence. 2. The chromophore must exhibit a large Stokes shift to help minimize excitation scatter. 3. The label should be relatively hydrophilic and highly selective for the immunological compounds of interest. It should show good solubility in reagents used in immunoassays while demonstrating a low affinity for nonspecific biomolecules and/or solid-phase surfaces. It also must possess appropriate functionalities to couple to specific antibodies or antigens. This coupling should be quick and efficient and should not significantly compromise the immunoreactivityof the labeled proteins. 4. Thefluorophore labelmustbestable. ThefluorophoreanditsAb conjugate must demonstrate long-term stability in storage and not be susceptible to the effects of photobleaching. 1.

NIR Fluorescence in Immunoassays

109

5. The labelshouldberelativelysmallwithrespecttotheAbbeing labeled,allowingfortheattachment of severalreporterlabelsper antibody for a maximum signal.The desired small size also minimizes problems associated with steric hindranceof the label by exerting little effect on the immunoreactivity of the labeled Ab. Most fluorescent compounds contain unsaturated polycyclic hydrocarbons as part of theirchromophores.Fluorescenceisenhanced byextendedconjugation,planarity,and rigidity of thestructure.Manypurepolycyclichydrocarbons are highly fluorescent; however,theirrelativelypoorsolubility in the solvents used in immunoassays and their high degree of hydrophobicity, which increasestheirnonspecificbinding,severelylimittheiruseasfluorescent labels. Derivatization of fluorescent hydrocarbons such as pyrene, perylene, naphthalene, phenanthrene, and anthrancene with hydrophilic functional groups increases their solubility in the solvents typically used in immunoassays. Derivatized fluorescent dyes including functionalized derivatives of fluorescein, rhodamine, coumarin, and porphyrins show relatively goodsolubility in aqueous solvents and are therefore preferred over hydrocarbons for use in immunoassays (Fig. 12) [34]. Fluorescein and its derivatives were among the earliest fluorescent dyes usedaseffectivelabels. It is aderivativexanthenesubstitutedwithphenolic hydroxyls at the 3’- and 6’-positions. It has a relatively strong molar absorptivity [ E = 72,000 L/(M . cm)] and a good quantum yield of fluorescence (@ = 0.85 in alkaline solutions). Its absorbance maximum at 492 nm makes it suitable for excitation with argon ion lasers.The maximum emission for fluoresceinis around 540 nm,whichlies in thedetectionrange of photomultipliertubes(PMTs). Fluorescein and its derivatives demonstrate moderate water solubility and have a net negative charge ator near physiological pH. They are also fairly photostable and does not show a significant photobleaching effect with reasonable excitation intensity. The small size of the fluorescein molecule relative to the molecular size ofan antibodymoleculesuggestsminimalstericinterferences in Ab-Ag complex formation. Different functional moieties can be incorporated into the to variousfunctionalgroupson ringsystem,whichallowsforeasycoupling immunological proteins (the most common one is isothiocyanate). Because it satisfies many of the criteria mentioned above, fluorescein is themostfrequentlyusedmarker in fluorescenceimmunoassays in thevisible region. Among the disadvantages associated with using fluorescein as a label are its particularsensitivitytoenvironmentaleffectsandasharpdecrease in quantum yielduponbinding to proteins (@ 0.3).Thequantum yield is also pH-dependent.Anotherdisadvantage is theconsiderablebackgroundinterference that occurs when the emission spectrum of fluorescein overlaps with those of biomolecules associated with protein. In particular, albumin-bound bilirubin

110

Swamy et al.

Fluorescein

Et2NmN +

\

\

/

Rhodamine

Coumarin Figure 12 Dyescommonlyused as visible fluorophores. X and Y are functional moieties for covalent labeling and solvent compatibility.

exhibits fluorescence that overlaps with fluorescein emission, generating severe background interference, which reduces the effectiveness of fluorescein labels, especially in homogeneous assay applications. The short Stokes shift of fluorescein causes scattering interference from excitation radiation. Also, the development of field applications with fluorescein-derivatized Ab is limited due to the bulky instrumentation required for excitation and detection of fluorescence. Laser-induced fluorescence provides an alternative to improving the sensitivity of fluorescence immunoassays. This can be explained in terms of the

NIR Fluorescence in Immunoassays

111

magnitude of the fluorescence signal at low dye concentrations, as given by the equation

F = 2.303Of1,~bC at the excitation where I, is the excitation power, E is the molar absorptivity wavelength, Of is thequantumyield, b isthepathlength,and C isthedye concentration. It can be seen thatthe limits of detection can be improved by using a stronger excitation source. However, one should keep in mind that the limit of detection increases only as the inverse square root of the laser power. Also, under a very high power of the excitation source, the dye can undergo photobleaching, whichcanleadtobothreversibleandirreversiblechangesinthedye,thus affectingthesignal. The useoflaser-inducedfluorescence in immunoassays was first reviewed by Hemmila [34]. The limitations of conventional lasers as excitationsourcesaretheirhighprice,size,andmaintenancecostsandthe limited choice of wavelength. As discussed in thenextsection,laser-induced fluorescence in the near-infrared (NIR) region helps overcome these limitations and aids in the development of a highly sensitive assay.

111.

NEAR-INFRARED LASER-INDUCED FLUORESCENCE IMMUNOASSAYS

The spectralregion of 650-1 100 nm corresponds to thenear-infrared(NIR) radiation of an energy range of 48-26 kcal/mol. This region of the spectrum has very low background interference in biological systems (Fig. 13). Atomic and molecular transitions in this long-wavelength region are processes that require relatively low energy photons because the ground and excited state species are close in energies. Typical NIR chromophores are polymethine, phthalocyanine, and naphthalocyanine dyes and certain elements, such as ruthenium and osmium. This section of the chapter describes the development of immunoassays that use NIR dyes as labels for observing immunological reactions.

A.

Excitation and Detection Sources in the NIR Region

Laser-inducedfluorescence in the NIR regionoffersseveraladvantages.Recent advances in semiconductor laser technology have made the use of lasers practical. This is mostly dueto the widespread application of NIR-emitting laser diodes in the telecommunications industry. This type of laseris inexpensive (typically ($150) and small (-1 cm) and has a longer operating lifetime (> 100,000 hr). A comparison of typical NIR and visible laser exitation sources is shown in Table 1.

112

Swarny et al.

Figure 13 UV-visible to NIRregionandpossibleinterferencefrombackgroundfluorescence.

TheGaAlAslaserdiodehasdrawnmuchinterestbecause its emission wavelength of 785 nm is compatible with several classes of polymethinecyanine dyes that exhibit NIR fluorescence [35-371. Nevertheless, only a small number of compounds are known to exhibit intrinsicNIR fluorescence, resulting in minimal background interference in this region [38]. Noise resulting from scatter is related to the wavelength of detection by a factor of l / A 4 . For example, detection around 820 nm versus 500 nm results in a greater than sixfold reduction in the scatter noise. The low background interference in theNIRspectralregion makes NIR fluorophores ideal probes for both biological and environmental applications.Detection in theNIRregionallowsthereplacementof commonly used photodiodes with an avalanche photodiode (APD). The APDs have excel-

Table 1 Comparison of NIR andVisibleLaser

Excitation Sources

cr Argondiode Laser Property ~

~

Wavelength Lifespan Power output Power consumption Replacement cost

785 nm 100,000+ hr 20 mW 0.150 W $5000

unoassays NIR in Fluorescence Table 2 Comparison of AvalanchePhotodiodesandPhotomultiplierTubes

Property Rcplacemcnt cost Lifetime

Approx $500

10.000 hr

80% at 820 nm

Quantum efficiency Internal amplification Size Power consumption

High Very small (mm) Verylow

Sensitive to light overexposure 0.3% at 820 nm Low Small (cm) Low

lent quantum efficiency in the NIR region. Some of the advantages of APDs are listed in Table 2. They are very inexpensive, compact, and long-lasting. In addition, they have low internal noise and very lower power consumption. All NIR fluorescence highly amenable to miniaturization, these features make the which aids in the development of a portable, compact, and rugged instrument for field application.

B.

NIR Fluorophore for Antibody Labeling

Heptamethinecyanine dyes are a class of NIR fluorophores that have been used for DNA sequencing, pH and hydrophobicity determination, metal ion detection, and antibody labeling [39]. An example of a water-soluble NIR cyanine dye, NN 382, is shown in Figure 14. The dye N N 382 and similarly functionalized

S03Na

so3

Na03S

Figure 14 Structure of NIR heptamethinecyaninedye N N 382.

114

Swamy et al.

cyanine dyes are being extensively used for antibody labeling because of their simple conjugation chemistry, namely the reaction of the isothiocyanate group aminegroups on antibodies.These (NCS) on theNIR dye withtheprimary ( > I O s M-l c n ” ) , highquantumyields dyes havehighmolarabsorptivities (20-40%), and relatively short fluorescence lifetimes (500-1000 psec), and are small molecules (“1000 Da). The small size of these dyes allows for a substantial number of labels per antibody without compromising antigen-antibody interactions,which in turn results in highersensitivity.Additionally,theuse ofasolidmatrix generates a stronger signal by concentrating the fluorescent molecules, thereby reducing the quenching effects of solvent. of NIRfluorophoresaslabels Williams [40] first demonstratedtheuse forasolid-phaseimmunoassay onanitrocellulosemembrane.Thismethod, however,suffers fromahighdegree of scattergenerated by themembrane andcancauseproblemswithnonspecificbinding of theconjugate. In addition,thelack of compatibility of theassaywithmicrotiterplates,whichare themost commonly usedsupportformodernimmunoassays,limitsitspractical use. In continuation of the development of the solid-phase near-infrared fluorescence immunoassay (NIRFIA), these issues were addressed in a study by Swamy [41], whoc developed an NIRFIA based on the heptamethinecyanine dye NN 382. The absorbance and emission characteristics of the dye are shown in Figure 15. The dye used in this study offers several advantages. It has a high molar absorptivity [ E = 180,000 L/(mol . cm)] and quantum yield (@ = 0.59 for dye conjugated to Ab). The isothiocyanate functionality undergoes a selective reaction with an amino group of the antibodies to form a stable thiourea linkage. the dye highly water-soluble. AnThe presence of the sulfonate groups makes other advantage of the presence of the negatively charged sulfonate groups on the dye is minimization of nonspecific binding to the solid matrix (polystyrene) because the matrix is often pretreated by manufacturers to have negative charges (to promote protein binding).

C. Instrumentation for NIR Fluorescence Immunoassay The instrumentation setup used in NIR fluorescence immunoassay (NIRFIA) is shown in Figure 16. This is a modified version of the LI-COR electrophoresis apparatus [42]. Thehigh voltage power supply, gel plates, and other components not required for the assay are removed, and the modified reader consists of the following components: a proprietary scanning fluorescence microscope (Model 4000X, LI-COR Inc., Lincoln, NE), an orthogonal scanner, an infrared analyzer, and a dataacquisitioninterface. The excitationsource is a 10-20mMlaser diode emitting at 785 nm mounted on the detection microscope at an angle such

NIR Fluorescence in Immunoassays

115

,

01

700

720

740

760

780

800

820

840

860

880

1wo

9W

WAVELENGTH (NM)

Figure 15 Absorptionandemissionspectra

of N N 382.

that the focused polarized radiation strikes the target at Brewster's angle (56") to minimize scattering. The focused laser beam gives a spot size of 30 p n x 50 pm. The detector in the fluorescence microscope is an avalanche photodiode (APD) cooledby a three-stage Peltier thermoelectric cooler with detection optics. The detection optics include a 20 mm focal length aspheric objective lens, two bandpass filters (820 f 10 nm) to eliminate scattered light from the excitation source, and a focusing lens. The fluorescence microscopeis mounted on a scanning platform with variable scan speeds ( 1 5 - 1 5 cm/hr) for scanning along the Y axis. The microscope is coupled with an orthogonal scanner for movement of the plate along the X axis programmed for nine separate scan speeds (4-260 cm/hr), allowing considerable overall flexibility in image resolution and acquisition time (Fig. 17). The final configured setup for the microtiter plate fluorescence scanner is shown in Figure 18. The fluorescence signal is collected by the APD in the microscope and is sent to an infrared analyzer, where the signal is amplified by a fixed-gain amplifier and a variable-gain amplifier. The variable-gain amplifier is adjustable via

116

Swamy et at.

-

I

Buffer tank

High voltage electrophoresis power supply

Laser microscope

Buffer tank

f / System power supply

-

Figure 16 Side view of LI-COR DNA sequencer.

software control to produce a system gain of 0-1000. This allows for increasing the sensitivity of assays with low conjugate concentrations by simply adjusting the gain factor. The signal is then passed on to a summing amplifier at which the direct current background is varied through subtraction of software-adjustable dc offset (1-10 V, software setting of 0-1000). This adjustment allows for differentiating between background scatter and true signal by simply adjusting the offset. Finally the signal is filtered by a lock-in amplifier and digitized by a 12 bit analog-to-digital converter. The data can be collected in either the 8 or 16 bit mode. However, the 8 bit mode severely limits the dynamic range of the actual assay. To obtain 16 bit data, the 12 bit data obtained are sampled 16 times and summed, producing a maximum 16 bit value of 65,520 (16 x 4095). The data acquisition interface is connected to an IBM computer via anIEE 488 (GPIB) cable, which in turn produces a real-time image of the scan. The scan files are then quantified by Image Pro (Image Pro, Baltimore, MD) to obtain signal values.

117

NIR Fluorescence in Immunoassays

Figure 17 Modified LI-COR scanner for two-dimensionalscanning.(Adaptedfrom Ref. 42.)

Thermoelectric cooler

Collection

Figure 18 Assembly for scanning microtiter plates in fluorescence immunoassay.

Swamy et al.

118

D.

Parameters in Evaluation of the NIRDye Labeled Antibody

Thissectiondescribes theprocedureused in definingthevariousparameters that are essential for evaluation of the labeled antibody in immunoassays. Immunoassays involve interaction between antigens and antibodies and are carried out under such conditions thatonly one of the two reactants is a limiting reagent. The first step after the preparation of a labeled Ab is the determination of conjugate excess conditions. This is usually determined by coating the microtiter with excess Ag (typical concentration of 1-2 ,ug/mL). Unbound Ag is washed off from the well, and a serial dilution of labeled Ab is added. Bound conjugate activity is quantifiedasasignal/wellratioandplottedagainsttheconcentration of conjugate. The concentration at which the activity curve first develops a plateau is considered excess. A tenfold greater concentration of conjugate excess is used in all subsequent parameter determinations. The next step is to determine the specific activity of the conjugates prepared under different conditions. Inthisstepamicrotiterplate is coatedwithaserialdilution of antigen. Excess unbound Ag is washed off, and then excess conjugate (as determined) is added and the signal/well ratio is determined. The best conjugate is the one that allows for the detection of the smallest amount of Ag under these conditions. Anotherimportantparameteristhedeterminationofthedye/Abratio of the different conjugates. In NIRFIA, the concentration of the dye is determined by absorbance of the conjugate at 780 nm. Antibody concentrations are determined [43] for protein determination. by means of the method described by Bradford The reader can obtain additional information on specific assay parameters in the work by Swamy 1411.

E. Preparation of NIR Dye-Antibody Conjugates for NlRFlA Thissectiondescribesthevariousparametersthat playacrucialrole in the preparation of conjugates with the highest specific activity. The antibody labeling procedure is adapted from that of Harlow et al. 141. Briefly, the isothiocyanate group onthe dye is allowed to reactwiththeprimary amine functionality on the antibody to give the thiourea linkage. A specified concentration of Ab [goat anti-human IgG (GAHG)] is prepared and buffered at the desired pH. A freshly prepared stock solution of the NIR dye (10 mg/mL) in 0.1 M NaCl solution is added to GAHG to achieve specified final molar concentrations, and the pH is verified again. The NIR dye and Ab are allowed to couple while being mixed end-over-end in a glass vial. After coupling for the specified time, the labeled Ab and the unreacted NIR dye are separated by size exclusion in a PD-10 column previously equilibrated with phosphate-buffered saline (PBS, pH 7.2). The final

unoassays NIR in Fluorescence

protein concentration for the conjugates is maintained at 0.2-0.4 mg/mL.The conjugate solutions prepared are then mixed with glycerol 0.77 v/v. The solutions are then split into equal volumes, with one half stored at -85°C and the working aliquot maintained at -20°C. At -2O"C, the conjugate does not freeze, due to the presence of glycerol, and hence maintains a high specific activity of the conjugates, which may otherwise decrease due to frequent freeze-thaw cycles. For more specific details, the reader is encouraged to see the procedure descried by Swamy [41]. NN382 (Fig. 14) is the dye used in the development of the NIRFIA. Its solubility exceeds 50 mg/mL, which makes it an ideal choice for labeling antibodies. The isothiocyanate group on the dye undergoes a reaction withthe primary amine group on the antibodies to form a stable thiourea linkage. CouplingpHisoneofthemostcriticalfactorsinthecouplingprocedure.The optimal pH for this reaction was determined by measuring the specific activity of the conjugates under different pH and Ag limiting conditions as described earlier.TheconjugatepreparedatapH of 10showsthegreatestspecificity andactivity(Fig.19).Undermorebasicconditionsthedye/proteinratio of the conjugate increases; however, the specific activity also decreases (Fig. 20). Additionally, at pH > 11, the stabilities of the dye and antibody are affected. In particular, the central phenoxy moiety in NN382 is hydrolytically unstable under these pH conditions. The hydrolysis resultsin removal of the label arm to give a nonfluorescent derivative. Denaturation of antibodies at high pH is also well documented. Evaluation of optimum time and temperature conditions for the coupling reaction showed that coupling at 4°C for 24 hr, 25°C for 2 hr, or 37°C for 1 hr all yield a similar dye/protein ratio(Fig. 21). Conjugates prepared

6

7

8

9

10

11

11.7

PH Figure 19 Plot of dye molecules bound to antibodies as a function of coupling pH.

Swamy et al.

120

6

8

7

9

10

11

11.7

PH Figure 20 Specific activities measured as fluorescence intensity under antigen limiting conditions for dye-Ab conjugates prepared under different coupling pH conditions.

under extreme conditions, such as 37°C for 24 hr, show a significant drop in of the dye and denaturation activity, which can be explained by the degradation of the antibody that occur under these conditions. Next, conjugates prepared with different initial NIR dye/Ab ratios were compared in terms of their binding to specific Ag and a nonspecific protein, namely bovine serum albumin (BSA). The results are shown in Figure 22. The

10

30

1440

80

120

Time (mln)

Figure 21 Plot of dye molecules bound to antibody for conjugates prepared at pH 10 under different time-temperature conditions.

NIR Fluorescence in lmmunoassays

121

. m 0

L - - Y

1.41E-008

t 3.52E-009

t

+

+

8.8OE-010

A.. 2.20E-010

m

-

-.

5.50E-011

I

L

I

1.3s-011

NHulgG Conc [MI

Figure 22 Binding of the goat anti-human NIR dye conjugate NHu/IgG prepared at 1OO:l dyelprotein molar ratio to (W) its specitic antigen (normal human IgG) and a nonspecific protein (BSA).

(e)

nonspecific binding to BSA increases for conjugates with higher initial ratios of dye to protein (>500:1), and this nonspecific binding increases even more with higher BSA concentrations. The higher initial dye/protein ratio does increase the number of labels; however, a large number of labels per antibody adversely affects the ability of the Ab to recognize specific antigen. Also, during the coupling reaction, a high initial molar concentration of the dye promotes noncovalent association of the dye molecule with the antibody. Williams et al. compared the efficiencies of the noncovalent and covalent labeling procedures [14]. Others have used the noncovalent binding to determine the three-point interaction and binding sites in BSA [44]. The noncovalent interaction is not suitable for the development of reliable immunoassays, as the noncovalently bound dye can reversibly interact with BSA or other nonspecific proteins, resulting in false signals. The hypothesis that only conjugates prepared under high initial dye/protein ratios (>500:I ) show noncovalent association in addition to covalent coupling was confirmed by diluting the prepared conjugates and passing them again through a size exclusion column. Under these conditions, only conjugates prepared under high initial dye/protein ratios (>500:1 ) separated into two bands. Conjugates with lower initial dye/protein ratios (10-200) eluted as one band (as expected) even on the second pass through the column. Conjugates prepared with an initial dye/protein ratio of 125: 1 show the best signal-to-noise ratio under Ag limiting conditions (Fig. 23).

F. Validation of NlRFlA and Bioanalytical Applications The best NIRD-Ab conjugate, as discussed in the preceding section, was evaluated for its application in a diagnostic assay, and its detection limits were com-

Swamy et al.

0

)

1-101:75

1.50

1:1w

1125

1.150

1:zw

1 '4W

lnllial dydprolein molarratlo

Figure 23 Signal-to-noiseratio of corijugates prcpared underdifferentinitial tcin/dye molar ratios for detecting limiting amounts of antigen.

pro-

pared by using ELISA. The NIRFIA developed was in an indirect assay format for detection of schistosomiasis, and the results obtained were validated with those obtained by using other common clinical methods. The analyzed samples wereobtainedfrompatientswhowereclinicallyconfirmedeitherpositive or negative or had other parasitic infections known to cross react with the schistosome antigen. The general formatof the assay was an indirect immunoassay, and the steps involved are outlined in Figure 24. A total of 66 samples-25 negative. 35 positive, and 6 cross reactors-were assayed. There was 100% agreement of

*e**

(1) Fix antigen to the wells wash and off unbound antigen

Add clinical -\~~ which has antibodies A& and washoff unbound (2)

serum

. .

secondary antibody, wash off unbound and measure signal with NIR scanner Figure 24 assay.

Indirect irnmunoassay format used for validation of NlRFlA in a diagnostic

unoassays NIR in Fluorescence

123

results obtained by NIR fluorescence assays with other methods such as FAST ELISA and western blot. Intra- and inter-run variations and detection limits between NIRFIA and in three experiments to deELISA were compared by using triplicate samples termine the ability of the two methods to detect limiting amounts of antigens. Linear regression was used to describe and compare signal values at 11 different concentrations between the two assays. The data were linearized by taking the level. Detection limits base I O logarithm of the signal value and concentration for each method were defined as the upper99% confidence limit (C.L.) of signal M. Thelowestdetectableconcentration values at aconcentration of 6.88 x was defined as thc concentration whose value is greater than the upper 99% C.L. of noise (6.88E”’). The overall geometric mean, mean square error (estimate of model variability, MSE), coefficient of variation, C.V.; root MSE/mean), r 2 (amount of variability “explained’ by the model), and detection level estimates were computed for each method. A representative plot of the results obtained is shown in Figure 25. The NIRFIA had lower variability (as percentage of mean) and lower detection limits in two out of three experiments. The detection levels in one of the experiments for the two methods were quite comparable.

1. Application of NlRFlA for Detection of Extracellular Antigen

In the validation of NIRFIA in the diagnostic assay described above. the assay wasused

to detectpurified

antigen.Thissectiondescribestheapplication

of

Signal 1.000,Ooo

100.000

level

Detection

4

r2=0,967

t

+

Geom. Mean = 2870.12 C.V. = 4.37%

.y

* * * *

10,000

NlRFlA

Geom. Mean = 30.1 C.V. = 8.25%

0.01

1

01 138

775

55

1 I

10 22

44

811

100 176

357

10

1

.ooo

111

Concentration ( ~ 1 0 ” ~ )

Figure 25 Comparison of NIRFIA with ELISA for detection limits and coefficient of

variation.

Swamy et al.

124

NIRFIA in a direct assay format for the detectionof Ag expressed on the surface of cells. The preliminary results of this application, developed by Swamy [41], were obtained in collaboration with Dr. Joseph Michl, Department of Pathology, State University of New York at Brooklyn. Inthisapplication,thehumansquamouscellcarcinomaline(HuSCC) A431waschosen[45]foravailabilityofwell-definedmonoclonalAb(Mab E48) directed against various cell surface antigens [46,47]. The Mab E48 was originally produced against cells of the HuSCC of the larynx and reacts with approximately 98% of head and neck tumors and their lymph node metastases [46]. This Ab binds specifically to a 22 kDa antigen with high affinity (1.5 x 10'" M"), of which at least 1.5 x IO6 sites are accessible on the surface of the A431 cells [46,48]. Preliminary results demonstrated the successful use of MAbE48labeled withtheNIRfluorophore (NIRF-MAbE48) in aseriesof binding studies to the HuSCC line A431. The results shown in Table 3 clearly showthatthe NIRF-mAb E48 fully retained its ability to bind to its antigen on A431 cells, providing a strong NIR signal at 820 nm in the absence of any background. Wells coated with the specific Ag showed a positive signal, and the control wells were negative. Treatment of the wells with PBS-BSA (0.003 WglrnL; 0.1 mL/well), a blocking agent, for quenching any nonspecific binding of NIR-mAb E48 to the Ag-coated wells nor did it cause a reduction in the backgroundsignalfromcontrolwells(withoutAg).Interestingly,thesignal from the BSA-treated positive wells was slightly higher than the wells without BSA treatment. This can be explained by the known increase in fluorescence of the NIR label in hydrophobic environments. Even though several recent studies have described the specificity of the labeled mAb E48 in immunodetection of SCC in humans and HuSCC-Nu/Nu mouse grafts by conventional methods such as radioisotopic single photon emission computed tomography (SPECT) [49,50], the specificity of NIR-dye-labeled Mab E48 to the E48 Ag was established in a competitive experiment. The results of the competitive experiment are shown in Table4. The results clearly

Table 3 Binding of NIR FluorophoreLabeled MAb E48 to Hs Antigen on A431 Cells

treated Wells Expt with

1

Expt 2

Expt 3

FBS Yes Yes A431 cells NIRF-MAb E48 BSA

Yes Yes Yes Yes Yes

Yes

NO

NO

Yes Yes Yes Yes

Fluorescence intensity (xu.)

59027

I037

61570

Yes No

Expt 4

Expt 5

NO

No

Yes

No

47 3

517

NIR Fluorescence in Immunoassays

125

Results of Competitive Experiments on NIR DyeLabeledMAb E48 and Unlabeled Ab to Establish Specific Binding

Table 4

1 (Noncompetitive) Expt

Wells treated Expt with A431 cells + FBS MAb E48 (unlabeled) PBS-BSA NIR-MAb E48 Fluorescence intensity

Yes Yes

Yes Yes Yes Yes

58,586

22,295

Yes No

(a.u.)

2 (Competitive )

indicate that the unlabeled MAb E48 competes with the labeled Ab, reducing the signal by more than 62%, indicating effective competition. It is clear that to maximize competition, i.e., achieve 90-100% competition, additional experiments will have to be done with different concentrations of MAb E48 (and an unrelated control Ab). The preliminary results obtained demonstrate the feasibility of using NIR labeled antibodies in different assay formats and their possible applicability in clinical imaging applications such as NIR optical tomography.

2. Application of NlRFlA in Detection of Environmental Contaminants

Near-infrared fluorescence immunoassay was evaluated as an analytical tool in environmental applications in the quantitative analysis of two pesticides, bromacil and fenvalerate. This work was done by Swamy [41] in collaboration with the Hammocks group in the Department of Entomology and Environment Toxicology, University of California, Davis. Dr. Ingrid Wengatz did the preparative work for the assay, and Dr. Ferenc Szurdoki synthesized the hapten for the assay. Bromacil, also known as Hyvar, is 5-bromo-6-methyl-3-( 1-methylpropy1)2,4( 1 H,3H)-pyrimidinedione (Fig. 26). A numberof instrumental methods have been developed for the detection of bromacil in soil and water samples [51-561. These analyses share drawbacks such as the need of laborious extraction andor cleanupprocedures,highlyqualifiedanalysts,andexpensiveinstrumentation. These required instruments could not be adapted as field-portable sensors. Hammock and coworkers developed immunoassays for the analysis of bromacil at t r x e levels that provide a low-cost alternative for screening large numbers of environmental samples [57-591. The assay system is highly sensitive (ICs0 of 0.25 ppb) and selective for bromacil [58,59]. By using this assay, water samples spiked with 0.1-160 ppb levels of bromacil can be analyzed directly, and 0.0420 ppm concentrations of bromacil can be detected in soil with only a single extraction step [%I. An alternative immunoassay using NIR dyes conjugated to a carrier protein (BSA) that is in turn coupled with bromacil hapten has been used in a tracer format assay 1601.

Swamy et al.

126

Bromacil: R = H Bromacil hapten: R = (CH2)&OOH Figure 26

Structure of bromaciland its hapten.

The NIRFIA assay was done in a capture Ab format. In this assay, microtiter plates are precoated with a secondary antibody solution (goat anti-rabbit antibody).Theplatesarewashedandblockedwithovalbuminsolution.The platesarewashedagainandcoatedwith a solution of polyclonalrabbitantiserum 2369 (primary Ab) 157,581. Excessantiserum is washed off, andthe plates are incubated with the solution of the bromacil standard and the tracer (Bromacil-BSA-dye). After excess tracer is washed off, the fluorescence signal is measured by the NIR fluorescence scanner. Separation of the bound from free

95 00

-

85.00

-

e,

u

C e,

u v)

-

3 LL e,

> .-*

2

7500-

d

6500 0 001

I 1

1

0 010

0 100

1 000

Bromacil (ppb) Figure 27 NIRFIA.

Plot of relativefluorescenceversusbromacilconcentration

as obtained by

127

NIR Fluorescence in Immunoassays

I

1

1

10

100

1000

Pyrethroid metabolite (ppb) Figure 28 Plot of relativefluorescenceversuspyrethroidmetaboliteconcentration obtained by NIRFIA.

as

labels in the coupling step can be easily carried out in a gel filtration column. The soluble tracer that did not have nonspecific binding to microtiter plates. A plot of the fluorescence signal versus bromacil concentration is shown in Figure 27. The standard curve demonstrates that the NIRFIA can detect less than 1 ppb of bromacil, and the sensitivity of the assay is comparable to that attained 10 by ELISA [57,59]. In a similar assay developed for fenvalerate, as little as ppb of the pesticide could be easily detected by the NIRFIA method (Fig. 28)

1601.

IV. CONCLUSIONS The NIR fluorescence immunoassay is fast, inexpensive, and sensitive. Systematic optimization of its various steps accomplished by minimizing contributions from potential background at every step aids in achieving the maximum signalto-noiseratioforasensitiveassay.Backgroundfluorescencefromsolutions of interest decreases the sensitivity of components other than the fluorophore detection. In biological systems, this background fluorescence is typically from the autofluorescence or scattered light contributions of certain biological components. Similar problems are encountered with environmental samples. Typical backgroundfluorescenceoccursat all wavelengths in thevisibleregionand at various intensities depending on the concentration of interfering molecules

128

Swamy et al.

present in the sample. Elimination of background fluorescence is an especially demanding problem in all immunoassay formats. The problems associated with native fluorescence or background interference are negligible in this assay because the biological components usedin it have no native fluorescence in the NIR region. In fluorescence analysis, excitation light can be scattered from the solid phase on which the assay is carried out. This scattered light effect is present in all types of fluorescence detection. Scattered light contributes a significant portion of background noise. Scattered light can be due to contributions from the Rayleigh scatteror the Raman scatter by the solvent molecules. These issues have also been addressed in NIR fluorescence assay format; the Rayleigh scatter decreases with increasing wavelength (c I

I

glass substrate

Figure 7 Schematic diagram of antibody-antigen complexes immobilized on a solid surface.

Castro

194

various antigen molecules were specifically bound to the surface and detected at the single-molecule level of sensitivity (Fig. 8).

D. Low Temperature Studies in Solids The spectroscopy of single guest molecules in solids at liquid helium temperatures is a very active area of research [14]. This method is based on the fact that the optical absorption line for a collection of guest molecules in a solid is inhomogeneously broadened. Strains, dislocations, and point defects in the solid cause a distribution of resonance frequencies for the various guest molecules. The optical absorption line width of any single molecule in the host matrix-the homogeneous width-is actually several orders of magnitude narrower than the inhomogeneous band. Therefore, the absorption of single molecules can be isolated in an inhomogeneous sample simply by using a tunable narrowband laser, provided that the average number of guest molecules present in the detection volume is less than unity. The best results are usually obtained with the fluorescence excitation method, where photons emitted as a result of the zero-phonon electronic transition are collected with high efficiency optics. The advantage of using NIR excitation has also been recognized in experiments of thistype.Orritandcoworkers [29] demonstratedthedetection ofsingledibenzoterryleneguestmolecules in naphthaleneusingNIRexcitation.Fluorescencespectra of singledibenzoterrylenemolecules(Fig. 9) were

60

I

0

1

2

3

time

4

5

6

[SI

Figure 8 Fluorescence signal from an antigen n1olecule labeled with Cy-5 and captured from a IO"' M aqueous solution. (Courtesy of S. Seeger. University of Regensburg.)

195

NIR in Single-Molecule Detection

N

m

N

ln

m

jj

c

200

400

*0

I

0

600

800

1000

1200

1400

1600

RELATIVE FREQUENCY (cm”) Figure 9 Fluorescence spectrum of a single dibenzoterrylene molecule in naphthalene at 2.0 K (upper trace). The lower trace was obtained by detuning the laser from the excitation maximum of dibenzoterrylene. The remaining lines are due to Raman scattering. (From Ref. 29.)

recorded using 758 nm excitation, a confocal collection system, and a spectrograph with a charge-coupled device (Fig. IO). This new NIR guest-host system appears very promising for low temperature single-molecule studies because it has a high fluorescence yield and narrow homogeneous lines and shows neither spectral diffusion nor significant triplet populations.

E. Surface-Enhanced Raman Scattering Using NIR Excitation Kneipp and coworkers [3O-331 recently demonstrated the detection of a single NIR dye molecule by surface-enhanced Raman scattering from molecules adsorbed on cluster particles. Although not conducted by detecting fluorescence, these studies are mentioned here because the instrumentation and possible applications of the technique bear a close resemblance to those presented throughout this chapter. Agreatenhancement of Ramanscatteringcrosssectionsoccurswhen amoleculeisattached to small(nanometersize)metallicstructures [34]. By ( it becomespossible takingadvantageoftheselargeenhancementfactors

196

Castro

c ARGON LASER

I

CORRELATOR

R I N G DYE

-I-

I

LASER

DETECTION

CUTOFF FILTER

FREQUENCY AND POWER STABILIZATION OPTICAL FIBER

1

CRYOSTAT

Figure 10 Schematic diagram of the experimental setup for detecting single molecules in host crystals at low temperatures.(Courtesy of M. Orrit. CNRS and Universityof Bordeaux 1.)

to detect a single molecule adsorbed to a colloidal silver cluster. The advantages of usingNIRlasersources, and theconsequentdecrease in fluorescenceand Raman scattering from the solvent, were exploited in these studies. Kneipp also demonstrated the detection of single-molecule vibrational spectra from dyes and nucleotides using nonresonant NIR excitation [3 11 (Fig. 1 1 ).

V.

OUTLOOK

The field of single-moleculedetection in bothbasicandappliedresearch is rapidly growing. The advantages of using NIR excitation and detection are be-

197

NIR in Single-Molecule Detection

200

j

100

-

8 .

IA

m

0-

v)

5

E1

3001

I

I

I

800

1000

I200

1400

800

1000

1200

1400

800

1000

1200

1400

'

I

~amm shift / cm" Figure 11 Surface-enhancedRamanscatteringspectrarepresentingone

(top), zero (middle), and two (bottom) adenine molecules adsorbed on colloidal silver clusters. Excitation wavelength 830 nm; laser power 80 mW: collection time 1 sec. (From Ref. 31.)

ing solidly established. New applications in chemical and biochemical analysis should appear in the next few years. This growth will greatly benetit from the designandsynthesis of morerobust NIR fluorophoresand from incremental advances in optical detection technologies, now being actively pursued.

REFERENCES WE Moerner, M Orrit. Illuminating single molecules in condensed matter. Science 28311670-1676,1999. 2. T Hirshfield.Opticalmicroscopicobservation of single small molecules. Appl Opt 15:2965-2966,1976. 1.

198

3.

4. 5.

6. 7. 8.

9.

IO. 11. 12.

13.

14. 15.

16.

17.

18. 19.

20. 21.

Castro K Peck, L Stryer, AN Glazer, RA Mathies. Single-molecule fluorescence detection: Autocorrelation criterion and experimental realization with phycoerythrin. Proc Natl AcadSciUSA 86:40874091, 1989. RP Haugland. Handbook of Fluorescent Probes and Research Chemicals. Eugene, OR: Molecular Probes, Inc., 1992. EBShera. N K Seitzinger,LMDavis,RAKeller,SASoper.Detectionofsingle fluorescent molecules. Chem Phys Lett 174:553-557. 1990. MD Barnes. KC Ng, WB Whitten, JM Ramsey. Detection of single Rhodamine 6G molecules in levitated microdroplets. Anal Chem 65:2360-2365, 1993. M Eigen, R Rigler. Sorting single molecules: Application to diagnostics and evolutionary biotechnology. Proc Natl Acad Sci USA 91:5740-5747, 1994. SA Soper, QL Mattingly, P Vegunta. Photon burst detection of single near-Infrared fluorescent molecules. Anal Chem 65:740-747, 1993. R Muller, C Zander, M Sauer, M Deimel,DS KO, S Siebert, J Ardenjacob, G Deltau, NJ Marx, KH Drexhage, J Wolfrum. Time-resolved identificationof single molecules in solution with a pulsed semiconductor diode-laser. Chem Phys Lett 262:7 16-722, 1996. T Schmidt, GJ Schultz, W Baumgartner, HJ Gruber, H Schindler. Imaging of single molecule diffusion. Proc Natl Acad Sci USA 93:2926-2929, 1996. A Castro, FR Fairfield. EB Shera. Fluorescence detection and size measurement of single DNA molecules. Anal Chem 65:849-852, 1993. PMGoodwin,MEJohnson,JCMartin.WPAmbrose,BLMarrone,JH Jett, RA Keller.RapidsizingofindividuallyfluorescentlystainedDNAfragmentsbyflow cytometry. Nucleic Acids Res 21:803-806, 1993. ACastro,JGKWilliams.Single-moleculedetectionofspecificnucleicacidsequences in unamplified genomic DNA. Anal Chem 69:3915-3920, 1997. WEMoerner.Highresolutionopticalspectroscopyofsinglemolecules in solids. Acc Chem Res 29563-57 I , 1996. YH Lee, RG Maus. BW Smith. JD Winefordner. Laser-induced fluorescence detection of a single-molecule in a capillary. Anal Chem 66:41424149. 1994. PSchwille. FJ Meyeralmes,RRigler.Dual-colorfluorescencecross-correlation J 72: 1878spectroscopy for multicomponent diffusional analysis in solution. Biophys 1886, 1997. F Loscher, S Bohme, J Martin. S Seeger. Counting of single protein molecules at interfacesandapplicationofthistechnique in early-stagediagnosis.AnalChem 70:3202-3205,1998. M Sauer, KH Drexhage, C Zander,J Wolfrum. Diode-laser based detectionof single molecules in solutions. Chem Phys Lett 254:223-228, 1996. M Sauer, KT Han, R Muller, S Nord, A Schulz, S Seeger, J Wolfrum, J Ardenjacob, G Deltau, C Zander, KH Drexhage. J Fluor 5:247, 1995. PMGoodwin,WPAmbrose,RAKeller.Single-moleculedetection in liquids by laser-induced fluorescence. Acc Chem Res 29:607-613, 1996. SA Soper, BL Legendre,JP Huang. Evaluation of thermodynamic and photophysical properties of tricarbocyanine near-IR dyesin organized media using single-molecule monitoring. Chem Phys Lett 237:339-345, 1995.

NIR in Detection Single-Molecule

199

22. SA Soper, BL Legendre. Single-molecule detcction in the near-IR using continuouswave diode-laser excitation with an avalanche photon detector. Appl Spectrosc 52: 1-6, 1998. 23. RD Guenard. LAKing.BW Smith, JD Winefordner. 2-Channel sequential singlemolecule measurement. Anal Chem 692426-2433, 1997. 24. DSKO,MSauer, S Nord,RMuller.JWolfrum.Determinationofthediffusioncoefficient of dye in solution at single-molecule level. Chem Phys Lett 26954-58, 1997. 25. M Sauer, C Zander, R Muller, B Ullrich. KH Drexhage, S Kaul, J Wolfrum. Detection and identitication of individual antigen molecules in human serum with pulsed semiconductor-lasers. Appl Phys B 65:427-43 I , 1997. 26. M Sauer,KHDrexhage, U Lieberwirth, R Muller, S Nord,CZander.Dynamics of the electron transfer reaction between oxazine dye and DNA oligonucleotides 153-163, 1998. monitored on the single-molecule level. Chem Phys Lett 284: U Lieberwirth, KMuhlegger, 27. M Sauer,JArden-Jacob,KHDrexhage,FGobel. R Muller, J Wolfrum. C Zander. Time-resolved identiticationof individual mononucleotide molecules in aqueous solution with pulsed semiconductor lasers. Bioimaging 6: 14-24, 1998. 28. AHartmann. D Bock, S Sceger.One-stepimmobilization of immUnoglObUh-g andpotentialofthemethodforapplicationinimmunosensors.SensActuatorsB 28: 143-149, 1995. 29. F Jelezko. P Tamarat. B Lounis, M Orrit. Dibenzoterrylene in naphthalene: Anew crystallinesystemforsingle-moleculespectroscopyinthenear-infrared. J Phys Chem100:13892-13894,1996. 30. K Knelpp, Y Wang, H Kneipp, LT Perelman, I Itzkan, R Dasari. MS Feld. Singlemoleculedetectionusingsurface-enhancedRamanscattering (SERS). Phys Rev Lett 78: 1667- 1670. 1997. 31. KKneipp,HKneipp, VB Kartha, R Manoharan, G Deinum. I Itzkan,RRDasari. MS Feld. Detection and identitication of a single DNA-base moleculeusing surfaceenhanced Raman scattering (SERS). Phys Rev E 57:R6281-R6284, 1998. 32. K Kneipp, H Kneipp, G Deinum, I Itzkan, RR Dasari, MS Feld. Single-molecule detection of a cyanine dye in silver colloidal solution using near-infrared surfaceenhanced Raman scattering. Appl Spectrosc 52: 175-178, 1998. 33. K Kneipp, H Kneipp,RManoharan. I Itzkan.RRDasari, MS Feld.SurfaceenhancedRamanscattering (SERS)-Anewtool forsinglemoleculedetection and identification. Bioirnaging 6:104-1 10. 1998. 34 K Kneipp, Y Wang, H Kneipp, I Itzkan, RR Dasari. MS Feld. Population pumping ofexcltedvibrational states by spontaneoussurface-enhancedRamanscattering. Phys RevLett74:2444-2447,1996.

This Page Intentionally Left Blank

Applications Using Fluorescence Lifetimes James H. Flanagan, Jr. and Benjamin L. Legendre, Jr. Transgenomic, inc., Omaha, Nebraska

1.

INTRODUCTION

The evolution of time domain spectroscopy as an analytical tool has provided a viable method for monitoring various photophysical phenomena in chemistry andbiochemistry,includingthestructureanddynamics of proteins [l], rotationaldiffusion in restrictedenvironments[2],andexcitedstateprotontransferreactions [3]. Inaddition,fluorescencelifetimedeterminationshavebeen in manyanalyticalapplicationssuchascapillary used as adetectionscheme electrophoresis [4], liquid chromatography [ 5 ] , fluorescence microscopy [6-10], determination of adsorption modes on chromatographic stationary phases [ I I], and measurements of fluorescence lifetimes for single molecular events [ 12,131. There are many advantages of time domain measurements compared to steadystate techniques, with the main advantage being that fluorescence measurements in the time domain yield information about the reaction rates of intra- and intermolecular processes. In addition, fluorescence lifetime measurements provide a method to probe the local environment of the fluorophore [ 141. Moreover, under appropriate conditions, lifetimes can be determined with higher precision than intensity-based measurements. There are two widely used methods for the determination of fluorescence lifetimes: the phase-resolved method and the time-resolved method. In phaseresolved spectroscopy, the sample is excited with sinusoidally modulated light, of the emission induced by the sample and the phase shift and demodulation relative to the excitation are used to determine the fluorescence lifetime. Timeresolved spectroscopy involves excitation of the sample with avery narrow pulse 201

202 Legendre

and

Flanagan

of light, and the subsequent time-dependent decay of the emission is determined. An advantage of time-resolvedtechniques in comparisonwithphase-resolved techniques is that time filtering canbesimultaneouslyemployed,whichcan improve the signal-to-noise ratio (SNR) during the separation by discriminating against interferences with short lifetimes or scattering photons [ 1,2, IO].

II. FLUORESCENCE LIFETIMES The fluorescence lifetime of a molecule can be defined as the average relaxation time from the excited singlet state to the ground singlet state. The expression to determine the parameters that describe an exponential decay process can be given as [ 14,151

r=l

where I I representsthenumberofcomponents in thedecay. A is thepreexponential factor, t is the time bin, and sf is the fluorescence lifetime. Because the fluorescence lifetime is proportional to the fluorescence quantum intensity, thefluorescencelifetimecanbedeterminedexperimentally by measuringthe time taken for the fluorescence intensity to fall to l / e of its initial value following the excitation of the molecule. This forms the basis of time-correlated single-photon counting (TCSPC) whereby the time-independent decay of individual fluorescence photons is analyzed. A typical decay is shown in Figure 1 along with the excitation pulse. The use of the fluorescence lifetime, sf,of a molecule depends upon the radiative and nonradiative decay processes undergone by the molecule and can be expressed through [ 14,151

where k r is the radiative rate constant (sec"), k,, is the nonradiative rate constant (sec-I), and k,,, is the total decay rate constant (sec"). From steady-state analysis, the fluorescence quantum yield, Q f , is given by [14,151 kr

QZ f"

kr

+ knr

kr

-

k,

+ kist + ki, + kd + ks,

(3)

where kist is theintersystemcrossing rate, k,, is theinternalconversionrate, kd is the photodecomposition rate, and k,, is the solvent-dependent rate. Qf is expressed as a percentage (0-100%) or as a number between 0 and 1.

203

Applications Using Fluorescence Lifetimes

1

AL

10

1

L

I

loo0

1500

2ooo

m

m

Channel Number [(2.88 psecy'] Figure 1 Decay protiles for instrumental response (solid line) and the NIR fluorophore 1R-125 inethanol (A), for which ~1 = 570 psec (f10 psec). [dye] 5 X M; = 7x5 nrn; P = 1 0 mW. Fortheinstrumentresponsefunction. full widthat hall

-

rnaxilnum (FWHM) = 165 psec.

204 Legendre

and

Flanagan

The fluorescence lifetime, q , is therefore given by sf = (kr

+ kis, + ki, + k d + k s e ) - '

(4)

As can be seen from Eqs. ( 5 ) and (6), rationalization of the photophysics and photochemistry of any singlet-state molecular species in terms of absolute rateconstantsforthevariouscompetingdecayprocessescannotbeobtained with knowledge of only the quantum yields [14]. Additionally, Eqs. ( 1 ) and ( 5 ) showthatmanyfactorsinfluencefluorescence and the fluorescence lifetime. The fluorescent molecule's microenvironment, such as the solvent propertiesof polarity and viscosity. probably plays the most important role in determining the decay kinetics.

111.

METHODS FOR MEASURING FLUORESCENCE LIFETIMES

There are two methods for determining the fluorescence lifetime of a molecule: phase-resolved spectroscopy and time-resolved spectroscopy. Phase or modulation spectroscopy incorporates a modulated excitation source such that the finite fluorescence lifetime of the sample causes the fluorescence emission waveform to be phase-shifted and of different amplitude when the signal is demodulated. If a sinusoidally modulated excitation signal, E ( t ) 114-161,

is used, then a modulated fluorescence signal,

F ( t ) = D F -tA F eXp i ( w f

F ( t ) , is produced in the form

+4)

(6)

where DE^ is the direct-current intensity component of the excitation beam, D F is the dc intensity component of the fluorescence signal, A E is~ the amplitude of the excitation signal, A F is the amplitude of the fluorescence signal, and w is the angular frequency (w = 27r f ,where f is the frequency). By substitution of Eq. (6) into Eq. ( I ) , the phase relationship is given by [ 14-16] tan 4 = wr

(7)

and the expression for the demodulation factor

AFIDF

is

1

Traditionally, phase spectroscopy has been carried out using electro-optic (Kerr or Pockelscell)oracousto-opticmodulation of continuous wave lamp or laser excitation. The upperlimit of the modulation frequency is important in determining the shortest lifetime that can be measured. Current modulation

Fluorescence Lifetimes Using Applications

205

methods operate at a maximum frequency of approximately 200 MHz, allowi% rf values as short as 1-100 psec to be measured [171. Time-resolvedspectroscopyinvolvestheexcitation of amoleculewith anarrowpulseoflightand subsequentmeasurementofthetimeinterval of the emission from the excited state. In time-correlated single-photon counting (TCSPC), the time evolution of individual photon events is processed, and upon processing many such events a histogram can be constructed that represents the decay kinetics of the excited state. The resolution and precision of the lifetime deterlnination are dictated by the width of the excitation pulse, the reproducibility of the pulse, and the instrumental response of the photodetector. Figure 2 shows a typical layout of a TCSPC instrument. The instrument consists of a pulsed light source, typically a mode-locked laser, which generates an excitation pulse train that stimulates absorption in the sample molecules. At lowlevelsofexcitationpower,eachsamplemoleculeabsorbsonephotonat most,on a timescale that is effectivelyinstantaneous.Thesubsequentrelaxation of the molecules from the excited state to the ground state via the emission of time delays as described of fluorescence photons occurs with a distribution byEq. ( 1 ). The single-photon counting technique records this distribution by measuring the time delays of the individual fluorescence photons (fphoton) with respecttothe arrival time of theexcitationpulse (tstan). The delaytimebetween the emission of the excitation pulse and the time the fluorescence photon from the sample reaches the detector, typically a photomultiplier tube (PMT) or a single-photon avalanche diode (SPAD), is measured by a time-to-amplitude converter (TAC). The excitation pulse serves as the “start” signal for charging a capacitor within the TAC. The voltage on the capacitor increases linearly until a “stop” fluorescence photon is detected. The “start”-“stop” time interval generates a proportional voltage across this capacitor. The voltage pulse is stored according to its amplitude by using an analog-to-digital converter (ADC) within a multichannel analyzer (MCA) and thereby placed into the appropriate channel number. On repeating the “start”-“stop” cycle many times, a histogram is formed that represents the fluorescence decay profile. Discriminators are used to aid in removing random noise pulses resulting from background photons and to ensure that the timing definition of the “start” and “stop” pulses is largely independent of the signal pulse height. Operation of the TAC requires the ability to register the first “stop” pulse detected after a “start” pulse. Accordingly, the “stop” pulse rate must be low enough that the probability of detecting more than one photon is negligible. In effect, the “start” pulse rate must be much greater than the “stop” pulse rate. Failure to adhere to this means that the TAC will preferentially detect photons that occur at shorter times, and the fluorescence decay time distribution will be biased, resulting in decay times that will appearto be shorter than they reallyare. This photon pileup effect makesthe single-photon counting technique inefficient

206

Flanagan and Legendre

I Emission

.“Start” : puke

Discriminator

“Stop” pulse

L

n 7

....... .......

r - 4

- tphoton ...... tstart

Figure 2 Schematic of a time-correlated single-photon counting instrument. TAC, timeto-amplitude converter; MCA. multichannel analyzer.

when the fluorescence signal is intense due to the fact that the available signal has to be wasted in order to ensure the absence of photon pileup. The time-correlatedsingle-photoncountingtechnique is adigitalrather than an analog technique as is the case for phase-resolved methods. Advantages of TCSPC include a high dynamicrangeandindependence of themeasured decay profile from fluctuations in the excitation pulse intensity. Moreover, singlephotondetectiontheory is basedonwell-documentedstatisticsforwhichthe precision, data weights, goodness of fit, etc., can be easily calculated.

Applications Using Fluorescence Lifetimes

207

IV. HISTORICAL PERSPECTIVE Earlydeterminations of fluorescencelifetimeswerecarriedoutonly by researchers who built their instruments in the laboratory. The early pulsed instruments consisted of a flashlamp excitation source and a sampling oscilloscope as the detector with corresponding timing electronics. However, many early instruments were based on the phase shift method, which used a mercury lamp output that was passed through a medium that modulated the light onto the sample cell. of Instrumentation improved in the 1960s, which made the determination fluorescence lifetimes easier and brought it into the mainstream. The introducof fluorescence tion of flashlamps made it feasible to carry out measurements lifetimes by observing the decay directly [ 181. Bennett [ 191 was one on the first (1960) to design an instrument with a nanosecond flashlamp as the excitation source. The tirst commercial instrument, based on the design of Mackey et al. [20], was the TRW Model 3 1A nanosecond spectral source system [ 181. This instrument incorporated many of the instrumental design features of Steingraber and Berlmann [21], and it was a modified version of the instrument that Chen et al. [22]used to determine the fluorescence lifetimesof 48 different compounds in1967. Alongwithinstrumentaladvances,StricklerandBerg[23]derivedthe relationship between the absorption intensity and the fluorescence lifetime for molecules. This relationship is shown in Eq. (9) where to is the fluorescence lifetime, tz is the refractive index of the medium, g~ and gu are the degeneracies of the lower and upper states, respectively, E is the molar extinction coefficient, and I is the fluorescence intensity.

where

This equation allowed researchers to estimate the fluorescence lifetime of aromatic molecules based on the molar absorptivity of the molecule and the fluorescence spectrum. They confirmed their results within experimental errorby comparing experimentally determined fluorescence lifetimes for several molecules with their calculated values. Pulsed laser, which have short pulse widths, high pulse repetition rates, in the andmorestableoutputsthanearlierpulsedlamps,begantobeused late1970s.Earlyinstrumentsusedmode-lockedAr ion lasersthatprovided 100 psec wide pulses that had a repetition rate of 76 MHz. In addition to gas ion lasers,dyelasersallowedforthedetermination of fluorescencelifetimes of molecules that could not be excited by available ion lasers. Several authors

208 Legendre

and

Flanagan

have reviewed applications, which have multiplied in number through the use of modern instrumentation, including biomedical applications 124,251, metabolic monitoring 1261, and the monitoring of membrane potentials (271 and of pH and metal ion concentrations [ 281. Inexpensive solid-state devices such as diode lasers and avalanche diode detectors can be used in the near-infrared and are attractive sources and detectors owing to their low cost, low maintenance, high output in the near infrared, simplicity of use, and small size. Diode lasers are made from semiconducting materials that are doped with either a group 111 or group V element. The semiconducting material with group 111 element doping is known as a p-type, and or one witha group V element as an n-type. />-Typematerialshave“holes” electron-deficient sites, and n-type materials are electron-rich semiconductors. When these two types of semiconductors are placed in contact with one another and a forward bias (positive potential) is placed across the material, electronhole pairs are formed in the depletion layer, and upon recombination a photon of light is emitted. A population inversion can be sustained by confining the spatial distribution of electrons by using materials with different refractive indices. One widely used diode laser consists of Gal-,AI,As. Diode lasers are attractive excitation sources in the near infrared because they exhibit stable coherent beamsof light with high output powers(- 100 mW) in the near infrared and can be modified to lase at a particular wavelength. In addition,thetypicallifetime of adiodelaser is on the orderof40,000 hr. Compared to ion lasers, diode lasers are more cost-effective over time. Diode lasers are small and can be easily used to miniaturize existing instrumentation. Semiconductor detectors can be used in the near infrared as well as for visiblewavelengths. In particular,thesingle-photonavalanchediode(SPAD) is an attractive detector. A SPAD consists of a semiconductor material and is reverse-biased above its breakdown voltage. When a photon of light strikes the diode, it creates an electron-hole pair that generates a cascade of electrons [29]. The major advantage of this device is that the detector has gain built into the system, which results fromthe cascade of electrons. This resultsin a large signal response for a single photon of light. Additionally, SPADs have high quantum efficiencies (-40%) in the near infrared. One drawback to the use of SPADS as pm in diameter), but the detectors is their small photoactive area (-150-200 proper use of focusing optics can alleviate this problem. The advent of semiconductor instrumentation has made applications based on fluorescencelifetimedeterminationinexpensiveandlessspace-consuming. Additionally,solid-stateinstrumentation is ideallysuitedforwork in thenear infrared, and many researchers are taking advantage of this.Barryetal. [30] reviewed the use of solid-state lasers for biomedical imaging applications. Soper and Mattingly 1311 determined the fluorescence lifetimesof some NIR dyes using a Ti-sapphire laser and an avalanche diode detector, and Legendre et al. [32] showed that precise lifetime measurements may be determined with a simple,

Fluorescence Lifetimes Using Applications

inexpensive solid-state diode laser as an excitation source. In 1997, Sauer et al. [33] demonstrated a techniqueforsequencingDNAthatusesapulseddiode laser and an avalanche diode detector.

V.

TCSPC INSTRUMENTATION

The two most important criteria in determining the overall system performance of a TCSPC device are the timing resolution and the sensitivity of the device. to system and are based largely on the These characteristics vary from system performanceoftheindividualcomponentsofthedevice,withthechoice of detector usually dictating the overall response of the system. A TCSPC system typicallyincludesthefollowingcomponents:apulsedlightsourcesuchas a flashlamp or mode-locked laser, a photodetector such a s a PMT or SPAD, and the counting electronics, including the CFD,TAC, and the multichannel analyzer with an ADC.

A.

Light Sources

There are basically two choices for pulsed excitation in TCSPC measurements: flashlamps or mode-locked lasers. Flashlamps operate on the premise that the to the breakdown voltlight pulses are formed when the electrodes are charged age. The typeof gas within the lamp, the pressureof the gas, and the arrangement of the electrodes determine the wavelength range. Additionally, the frequency of the pulses is determined by the type of gas and the pressure of the gas along wit the lamp capacitance and breakdown voltage. Gated lamps allow the freof the gas, pressure, and quency of the pulses to be controlled independently capacitance. A typical flashlamp has a pulse width of 2 nsec [ 14,151. Most near-infrared TCSPC applications today use a mode-locked laser for pulsed excitation. In general, a laser (which stands for light amplification by the stimulated emission of radiation)is an optical oscillator that creates a very highly directed (coherent) beam of light at a precise wavelength or frequency. There are three important components associated with all lasers: the high reflector, the gain medium, and the output coupler. gain mediumwillbe Forlasing to occur,thelightpassingthroughthe amplified. The highreflector at oneend of thelaserandtheoutputcoupler serve as the laser cavity in which the amplified light will return through the gain medium for further amplification. The output of the laser occurs when a fraction of the light is transmitted through the output coupler. In orderforlasing to occur,therearethreetypes of energyexchanges needed:absorption,spontaneousemission,andstimulatedemission. All three rely on the transitions from one energy level to another within the gain medium, with the difference between the two energy levels given as AE.

210

Legendre

and

Flanagan

Absorptionofaphoton of energy A E involvesthepromotion of the molecule of the gain medium from the ground state ( E l ) to the excited state ( E ? ) . Upon excitation, the molecule can relax back to El by releasing a photon of energy A E by either spontaneous or stimulated emission. Spontaneous emission is necessary for initiating lasing, but this process does not have the directional properties of stimulated emission and therefore represents a loss in the laser cavity. In stimulated emission, an incident photon interacts with a molecule in E2 and induces the emission of a photon with a transition to E l ; therefore the energy of the emitted photon is A E . Stimulated emission is required for lasing to occur. The photons generated by stimulated emission have two important properties; the first is that the photons have the same direction as the incident in phase with the excitation radiation, and the second is thatthephotonsare light, which is responsible for the coherence of laser light. Apopulationinversion is generated by pumpingandcanoccuronlyin systems that possess more than two levels that are involved in the lasing transition. The degree of population inversion, and therefore the efficiency of a laser, canbeenhanced by eitherincreasingthepopulation of theuppermetastable state or by decreasing the population of the lower state reached after stimulated emission. The metastable state is formed when the pump bands, or excited energy levels, of the laser rapidly relax via radiationless transitions ( x 5 0 nsec) to a longer lived metastable state ( z S msec). The initial nonradiative drop is called an idler transition. Athree-levellaserfunctions by excitation of themediumtothepump bands, relaxation to a metastable state via an idler transition, and then stimulated emission of a photon of energy h u for relaxation to the ground state. The major drawback of three-level laser systems is that the only means of depopulating the lower level is by the same relatively inefficient pumping process; thus to excite more molecules, additional pumping energy is needed. Some molecules can be pumped to the higher energy level, but a large number will remain in the ground state. A large number of atoms in the ground state diminishes the population inversion and leads to losses due to absorption of the laser beam. Most lasers, however, depend on transitions from the metastable state to another short-lived energy level that is still higher in energy than the ground state; they are called four-level lasers. Molecules in the lower short-lived energy level of the lasing transition quickly decay to the ground state by another idler transition. This greatly increases the degree of population inversion and therefore the efficiency of the laser without the expenditure of pumping energy.

B. Detectors For TCSPC measurements, a detector is needed that has a low timing dependence on wavelength, low timing jitter, low intensity after pulsing, high amplification,

Fluorescence Applications Lifetimes Using

211

lownoise,andawidespectralrange.Ultimately, it is thesecharacteristics of the detector that dictate the overall timing response of the TCSPC instrument. There are three general classes of detectors available for TCSPC experiments: photomultiplier tubes, microchannel plate photomultipliers, and avalanche photodiodes.

1.

Photomultiplier Tubes

In photon-counting experiments high gain photomultipliers can be employed owing to their great sensitivity. These PMTs are ableto detect current pulses arising from individual photons that fall on the photocathode at low light levels. Additionally, thermal dark current arising from the spontaneous release of electrons from the photocathode or dynodes can be minimized by cooling the PMT, thus improving the signal-to-noise ratio. Another important factor in photon-counting measurements is the linearity of the detector, which, for the PMT, is dictated by maintaining the dynodes at n constant voltage regardless of the incident light intensity or anode current. The timing response of the PMT is determined by the interval between the arrival of the incident light photon at the photocathode and the arrival of the amplified signal at the anode. The variations in these transit times is known as thetransittimespread,and it is mainlydictated by thegeometricpaths that photoelectrons can take through the dynode chain. Other factors that may introducetimingdisparitiesincludethewavelength of theincidentlight,the location at whichthephotoelectronsoriginate on thephotocathode,andthe of the differenttrajectoriesthatphotoelectronscantakefromthesamearea photocathode. A typical transit time for PMTs is 20 nsec Il.51. 2.

MicrochannelPlatePhotomultipliers

Microchannelplatephotomultipliers(MCPs)workonthesamepremiseas PMTs, except that instead of dynodes they use thin glass plates consisting of a series of microscopic channels. The surfaces of each plate between the chan(--I000 V ) is nelsarecoatedwithathinconductinglayer,andavoltage placed across the thickness of the plate. Each channel. which has a diameter of 12-25 p m , is lined with a secondary emitting surface and functions as an individual electron multiplier, releasing secondary electrons upon contact with the incident electron. The photoelectron generated from the photocathode travels a short distance to the first MCP, enters a channel, and strikes the wall of the channel, generating secondary electrons. These secondary electrons are accelerated further down the channel, collide with the walls, and cause the release of from more electrons. The electrons then traverse the column, and upon exiting it they spread out and enter a number of adjacent channels in the next MCP.

212

Legendre

and

Flanagan

After the cascade passes through the last plate, it is collected at the anode as a current pulse. Comparedto anordinaryPMTwithdynodes,the transit timeforthe MCP is muchless,owing to theshorterdistancethephotoelectronshaveto travel fromthephotocathodetotheanode;therefore,MCPscantimeevents [ 14, IS]. For a TCSPC device, much more accurately than dynode-based tubes this translates into a much narrower instrument response function.

3.

Single-PhotonAvalancheDiodes

Single-photon avalanche diodes (SPADs) are semiconductor-based detectors that arerununderreversebiasandabovethebreakdownvoltage 1341. Operation above the breakdown voltage is called Geiger operation. At this bias, the detector current remains zero until a carrier such as a fluorescencephoton reaches the active layer of the p n junction and triggers the cascade of electrons, in which manyelectron-holepairsareformed. The number of electron-holepairs that are formed represents the gain of the photodetector, which is determined by the energy of the incident photon and the efficiency of the detector at the incident photon’s wavelength. Photons with higher energies traverse farther into the active layer of the semiconductor; therefore, more electron-hole pairs are formed. When the first electron-hole pair is formed within the semiconductor material, arrival time. Once the leading edge of the avalanche current marks the photon the avalanche is triggered. an output pulse is generated from the detector and senttothecountingelectronics.Oncetheoutputpulseisformed,theSPAD voltage is decreasedbelowthebreakdownvoltageforelectron-holepairs to recombine. After this occurs, the SPAD bias is then restored to the operating valuc. Therefore, the dead time of the detector is determined by the recovery time and typically sets an upper limit for the dynamic range of the detector to photocurrent rates of “500 kHz. The SPAD, like the otherphotodetectors,canbetriggered by photons andalso bycarriersduetothermaleffectsinsidethesemiconductor.These processes cause a self-triggering of the cascade of electron-holepairswithin the semiconductormaterial that is calledthedarkrateofthedetector.The statistical fluctuations of these events, which compete with photons in triggering the detector, reduce the detector sensitivity. The thermal noise can be reduced by cooling the detector. With respect to conventional PMTs. SPADs have extended sensitivity in the near-infrared region, with quantum efficiencies as high as 30% at 800 nm 1341. Also, the timing response for a SPAD has a resolution on the picosecond timescale,along withsmalldeadtimesandtransittimespreads,duetothe decrease in the distance required for the photoelectron to traverse to initiate a response compared to that needed by other detectors [34].

Applications Using Fluorescence Lifetimes

213

C. TCSPC Electronics

1. Discriminators The output pulses from the detector are of varying heights due to dark noise, single-photon events, and multiphoton events. Discriminators provide the TAC with constant-amplitude pulses independent of the photodetector pulse shapes, which improves the timing capabilities on the instrument. Discriminators also improvethesignal-to-noiseratio by rejectinglowamplitudepulsesfromthe detector (i.e., thermal noise). There are two types of discriminators: leading edge discriminators (LEDs) 3 demonstrates the use of and constant-fraction discriminators (CFDs). Figure an LED [ 14,151. A broad distribution of pulse heights from the photodetector to errorswhentiming is performedviaaleadingedge pulsescangiverise discriminator. As can be seen from the figure, pulses A and B are emitted at the same time after excitation (to) but have different amplitudes and are seen tocrosseachdiscriminator level at differenttimes. At discriminatorlevels 1 and 2, it canbeseenfromthefigure that pulseAwouldappear to arrive

Time

I&

‘ h I

.................................................

214

Legendre

and

Flanagan

sooner than pulse B even though they were initiated at the same time [ 151. Also notetherelativetimeerrors At1 and A q . Differentdiscriminatorthresholds introduce more error than others depending upon the amplitudesof the generated pulses. For TCSPC measurements, constant-fraction discriminators are employed to reduce the timing error associated with the broad distribution of pulse heights comingfromthedetector. Withconstant-fractiondiscrimination,thedetector pulses are timed from a point on the leading edge that is a fixed fraction of the pulse height. For pulses of similar shape but different amplitudes, the timing point is constant. Figure 4a illustrates the way in which constant-fraction discrimination is achieved. Suppose the input pulse has amplitude V, and is to be timed at an amplitude ,f V, on the leading edge. In the discriminator, the input pulse is split into two paths. The first path delays the pulse by a factor 6 and is attenuated to a maxiinverts it, while in the other path the undelayed pulse mum f V , (Fig. 4b). The two signals are then added to form the zero-crossing signal (Fig. 4c). Only pulses with amplitudes greater than a given threshold are timed. Thediscriminator level fortheexcitationpulses is usuallyset to reject dark noise caused by signal pulses. The setting of the discriminator level for the fluorescence pulses is much more critical. Setting the discriminator level too low allows for the accumulation of background events in the fluorescence decay, whereas setting the discriminator level too high allows for multiphoton events to be counted [ 131. Fluorescence photodetector pulses have a wide amplitude distribution;therefore,there is nocleardividing line betweenpulsesarising fromdarknoiseandpulsesarisingfromfluorescencephotons.Increasingthe discriminator threshold seems to increase the signal-to-noise ratio. Consequently, thechosendiscriminator level willbea compromisebetweenacceptance of relatively more single-photon pulses and rejection of dark noise.

2. Time-to-AmplitudeConverter The TAC functions to determine the time interval between the excitation pulse and the subsequent arrival of a fluorescence photon at the detector. Upon receipt of a “start” pulse, and after a certain fixed delay, a timing capacitor is charged linearly from a constant-current source. The charge on the capacitor increases until the arrival of a“stop”pulse,and an outputpulseisgeneratedwithan amplitude proportional to the time between the “start” and “stop” pulses. If no “stop”pulse is receivedafteratimecalledthe TAC range,charging is automatically stopped. Again, a fixed time elapses before the capacitor is reset, at which time the instrument is ready to accept another “start” pulse. For TCSPC measurements, it is important that the response of the TAC be linear to minimize timing errors.

Applications Using Fluorescence Lifetimes

215

Figure 4 Schematic diagram of the function of a constant-fraction discrimmator. Thc timing is from a position on the rising edgc that is set to a fraction of the input pulse height. (a) showstheinput pulse thatisinvertedanddelayed by time 6: (b) shows the undelaycd pulsc attenuated to a nmximum amplitude (-fV;,); (c) shows the zerocrossover pulse. where thc timing is initiated.

3. MultichannelAnalyzer The multichannelanalyzer(MCA)consists of ananalog-to-digitalconverter (ADC), a memory comprising channels for storing data, and data input-output facilities. For TCSPCmeasurements, the MCAincorporateslowerandupper discriminator levels and a pulse height analysis mode for the display of fluorescence decay profiles. The data are usually displayed on a computer terminal or on an oscilloscope. For TCSPC measurements, between 200 and 600 channels are sufficient for a decay curve. However. increasing the number of channels in the decay and

216

and

Flanagan

Legendre

subsequent reduction in the time per channel improves the timing resolution the instrument, which would increase the experimental run time.

4.

of

InstrumentalResponseandConvolution

The instrument response function of a TCSPC instrument is determined by numerous factors, including the excitation pulse width, the timing jitter associated with the detector, the timing electronics (particularly the discriminator), and the optics. These factors cause the measured excitation pulse to be broader than the pure optical component. The effect of having an instrumental response functionis that the measured fluorescence decay form departs from the true fluorescence response function as described by Eq. ( I ) . The measured fluorescence decay F ( t ) can be analyzed, however, because F ( r ) can be expressed by the convolution of the instrumental response function, P ( r ) , and the theoretical fluorescence response function, i ( t )

~41, F(t)=

I’

P (t’)i(t - t’) nt‘

( 1 1)

for pulsed excitation where t’ defines the variable time delays or channel numbers of the infinitesimally small time widthsclt’ or channel widths of which P ( t ) is composed. By measuring P ( t ) experimentally over i channels, the convolved form of F ( t ) canbeobtainedfromEq. ( 1 I). The expressions explaining the convolutions are given by [ 141

r and Fy(i) =

B

+ A . F ( i + A)

(13)

for comparing with the data where i is an integer denoting the data channels, B is the background, A is a scaling factor, A is the shift parameter, and r is the fluorescence lifetime measured in channels.

D. Fluorescence Lifetime Determinations 1.

NonlinearLeastSquaresAnalysis

The fluorescence lifetime has been determined from an exponential decay process by a variety of methods, with the nonlinear least squares algorithm, accompanied by convolution or deconvolution of the instrumental response function with the goodness of fit determined by the value of x?, which is a measure of theerrorbetweentheactualdataandthe fitted function,beingonecommon

217

Fluorescence Applications Lifetimes Using

method [14,15]. Equation (13) needs to be fitted to fluorescencedecaydata (convolved decay of fluorescence response function and instrumental response function). Errors may include nonlinearity in the TAC along with the presence of high background levels due to scattered photons within the actual decay. The x 2 value determined is a function of the parameters given in Eqs. (12) and (13), namely A , B , r , and A . The least squares method aims to determine the best-fit parameters A', B', r', and A' that will yield the lowest possible value for x2. By definition [14,15],

where Y ( i ) is the fluorescence datum value, F y ( i ) is the fitting function value. a(;) isthestatisticaluncertainty of thedatumvalue Y ( i ) , and W ( i ) is the weighted residual. a(i), Forsingle-photoncountingexperiments,theexpecteddeviation, whichcharacterizestherandomnoise,canbeestimatedfromthedatafunction using Eq. (14), (15)

a(i) = I Y ( ~ ) I ' / ~

Weighted residual values are important for many reasons. for they show wherethemisfitoccurred in the fitting of the data. Also, their normalization compensatesfor thevariation in dataprecisionwithinthedatasetandfrom one data set to another. The deviations are expressed in terms of the standard deviations of the associated data noise.

2.

MonoexponentialDecayAnalysis

Forthecasewhere I Z = 1 (singleexponentialdecay),thereareseveralsimple algorithms for determining both the preexponential and exponential factors that describe a decay process. One method is the maximum likelihood estimator (MLE) [35-371. In thisalgorithm,thelifetimecanbecalculatedviathe relationship [36,37]

1=I

where t n is the total number of time channels in the decay spectrum, T is the Nt isthe total number of photocounts in timewidth in eachchannel(psec), thecalculation,and Ni representsthenumber of photocounts in theithtime channel. The left-hand side of Eq. (16) is not dependent on the data and is a function of only rf, whereas the right-hand side is determined from the experimental data. The lifetime can be abstracted from the data with the use of graphical, tabular, or iterative techniques. This algorithm has been used to calculate

Legendre

21 8

and

Flanagan

fluorescencelifetimes of singlemoleculeswithhighaccuracyandprecision [ 12,131. The relative standard deviation, ori/rr, for MLE lifetime determinations can be evaluated from the expression [38]

10.000

1.000

10

1

1000

1500 2500

2000

3000

Channel Number I(2.88 psec)"]

Figure 5 Decay protile for IR- 125 showing thc RLD method for lifetime determination. The decay profile is dividedintotwoequalwidths A t , andthecountsover At are sLlmmed. The lifctitne is then calculated v ~ Eq. a (19). The dashed vertical lines represent the boundaries for the tilne intervals.

Applications Using Fluorescence Lifetimes

When T

1 nsec) than tricarbocyanines [65]. The longer lifetime values allow researchers to probe

226

Flanagan and Legendre

COOH Figure 10 Near-infrared dye used for DNA sequencing by Sauer and coworkers.

environments that may be changing too slowly for tricarbocyanine probes. These dyes have some limitations. Naphthalocyanines are difficult to synthesize and purify [69]. Usually,synthesis of thesechromophoresoccurswith verylow yields ( < 10%).Purification is usually very difficult, requiring multiple purification steps. Additionally, the dyes are not water-soluble. The conversion of most naphthalocyaninescanbeaccomplishedthroughthesulfonationofthechromophore with sulfur trioxide, but the resulting product is usually a mixture of isomers that are difficult to purify (C. S. McWhorter, personal communication). Clearly,thesedyesexhibitphotophysicalpropertiesthatmaybemoreideal than those of tricarbocyanines, but their present synthetic limitations must be overcome before they can be more widely used as fluorescence lifetime probes. Present NIR fluorescence lifetime determinations are limitedby the lack of suitable chromophores that can be used as probes. Presently, Molecular Probes sells over 20,000 different fluorescent compounds for various applications, but only one dyeexhibitsabsorbanceandfluorescencemaxima in theNIR.LICOR (Lincoln, NE) and Amersham Pharmacia Biotech (Uppsala, Sweden) are theonlyothercompaniesthatsellprobesthathavecharacteristicabsorption and fluorescence maxima in the near infrared. To take full advantage of NIR fluorescence lifetime determinations, better probes need to be developed. New dyechemistryand new dyefamiliesneed to be explored;examplesinclude

Fluorescence Lifetimes Using Applications

227

Figure 11 Sulfonatedaluminumnaphthalocyaninedye.

lanthanidechelates,whichdemonstratelongluminescencelifetimesandNIR fluorescence. Once a proven set of probes are developed, the full potential fluorescence lifetime determination in the near infrared may be realized.

VII.

of

COMPARISON OF LIFETIME-BASED DETECTION WITH DIRECT FLUORESCENCE MEASUREMENTS

The fluorescence lifetime of a molecule represents the average amount of time a molecule resides in the excited state before it returns to the ground state. Within this time frame, several processes can occur, including excited state reactions, energy transfer, and collisionally induced quenching. Herein lies the advantage of lifetimedeterminationsoverdirectfluorescence.Determinationsinvolving the fluorescence lifetime of a molecule allow researchers to elicit information aboutthefluorophoreenvironment on atimescalethatismuchshorterthan directfluorescencewillallow.Researchershavedevelopedmethodsthatuse this advantage; the information gathered includes the measurement of rotational

Flanagan and Legendre

228

diffusion constants, energy transfer reaction constants, and excited state reaction constants. As an alternative to spectral discrimination, various groups have suggested that fluorescence lifetime discrimination can potentially serveas a viable method for base-calling in DNA sequencing applications [4,70,71]. Using this approach, primers or dideoxynucleotides are labeled with chromophores that have unique fluorescence lifetimes. After gel electrophoresis, the sequence is reconstructed based on the identification of the terminal base (Sanger sequencing) using lifetime discrimination and the mobility of the separated fragments. The principal advantages associated with lifetime discrimination for base-calling are that The calculated lifetime is immune to concentration differences, so dyelabeled terminators can potentially be usedas well as dye primers with a wide choice in polymerase enzymes to suit the particular sequencing application. 2 . Lifetime values can be determined with higher precision than fluorescence intensities under appropriate conditions, potentially improving the accuracy in base-calling. 3. Lifetimedeterminations do notsufferfrombroademissionprofiles associated with spectral discrimination. 4. Thefluorescencecanpotentiallybeprocessed on a singledetection channelwithouttheneedforspectralsorting to multipledetection channels. 1.

However, several problems do arise in considering such an approach for DNA sequencing, especially when microseparation techniques, such as capillary gel electrophoresis, are being used. The most pervasive problem is associated with the complex instrumentation required for lifetime determinations. For example, i n time domain techniques, a pulsed laser is required with a fast detector, typically a microchannel plate PM tube, and sophisticated counting electronics. In addition, poor photon statistics (low number of photocounts) produced from low loading levels and the transience of the signal can produce poor precision in the measurement. Poor precision would also be compounded by the presence of large amounts of scattering and impurity photons included in the decay profile. Finally, complex algorithms are often required for extracting the lifetime from the decay profile, making on-line determinations during electrophoresisdifficult. Many of theseconcernsassociatedwithlifetime-basedspeciesdiscrimination in DNA sequencing and other applications have been addressed using NIR fluorescence. For example, several groups have demonstrated that semiconductor diode lasers, which can be operated in a pulsed mode and lase between 680 and 800 nm in conjunction with single-photon avalanche diodes (SPADs) or photomultiplier tubes, can produce a simple time-correlated single-photon counting apparatus with performance characteristics comparable to those of visible-

Fluorescence Using Applications

Lifetimes

229

wavelength devices using mode-locked Nd:YAG lasers and microchannel plates [72-751. Soperandcoworkers [32] recentlydemonstratedthatlifetimemeasurements can be acquired in the near infrared by using solid-state components and counting electronics situated on a PC board. The also showed that simple algorithms can be used to calculate fluorescence lifetimes on-line during freesolution or capillary gel electrophoresis using NIR time-resolved fluorescence [4]. The standard deviation in the lifetime measurement of C-terminal fragments labeled with an NIR dye was found to be f 9 psec with decay profiles constructed from "20,000 photocounts. The high precision resulted primarily from the fact that in the near infrared the low scattering cross sections and the minimal number of intrinsically fluorescent components produced low numbers of interfering photocounts in the decay. FluorescencelifetimediscriminationforDNAsequencingexhibitsclear advantages over conventional direct fluorescence measurements, and coupling this technique with NIR fluorescence enhances the overall potential of the analof NIR fluorescence ysis. As this emerging technique matures, the advantages lifetime determinations will allow scientists to probe environments with unparalleled limits of detection. Advances in technology and fluorophore development willopenthe doorforthistechnique as aserioustoolforalltypesofdisciplines, including molecular biology, medicine, analytical chemistry, forensic science, organic chemistry, and inorganic chemistry.

REFERENCES I.

2.

3. 4. 5.

6.

7.

JM Beecham, L Brand. Time-resolved fluorescence of proteins. Annu Rev Biochem 54:43. 1985. JP Chauvet, M Agrawal. LK Patterson. Effects of molecular organization on photophysicalbehavior:Steady-stateandreal-timebehavior of chlorophyllAfluorescence in spreadmonolayersof dipalmitoylphosphatidylcholine.J PhysChern 92:42184222, 1988. GWRobinson.Protonchargetransferinvolvingthewatersolvent.JPhysChem 9S:lO386-10391. 1991. SA Soper, BL Legendre Jr, DC Williams. On-line fluorescence lifetime determinations in capillary electrophoresis. Anal Chem 67:4358-4365. 1995. DJ Desilets, PT Kissinger, FE Lytle. Measurement of fluorescence lifetimes during liquid chromatography. Anal Chem 59: 1830-1834, 1987. TA Louis, G Ripamonti. A Lacaita. Photoluminescence lifetime microscope spectrometer based on time-correlated single-photon counting with an avalanche diode detector. Rev Sci Instrum 61 : 1 1-22, 1990. XF Wang,TUchida,DMColeman, S Minami.Atwo-dimensionalfluorescence lifetimeimagingsystemusingagatedimageintensifier.ApplSpectrosc45:360366. 1991.

230 K Fushimi. Construction and evaluation of a frequencydomain epifluorescence microscope for lifetime and anisotropy decay measurements in subcellular domains. Biophys Chem 40:117-125, 1991. A Kusumi, A Tsuji, M Murata, Y Sako. AC Yoshizawa, S Kagiwada, T Hayakawa, S Ohnishi. Development of a streak-camera-based time-resolved microscope fluorometer and its application to studies of membrane fusion in single cells. Biochemistry30:6517-6527,1991. JR Lakowicz, KW Berndt.Lifetime-selectivefluorescenceimagingusinganrf phase-sensitive camera. Rev Sci Instrum 62: 1727-1 734, 1991. MJ Wirth, JD Burbage.Adsorbatereorientationat a water/(octadecylsilyl)silica interface. Anal Chem 63:1311-1317, 1991. SASoper,LMDavis,EBShera.Detectionandidentificationofsinglemolecules in solution. J Opt SOCAm B: Opt Phys 9:1761-1769, 1992. CW Wilkerson Jr. PM Goodwin, WP Ambrose. Detection and lifetime measurement of single molecules in flowing sample streams by laser-induced fluorescence. Appl Phys Lett 62:1804-1808, 1993. DO’Connor.DPhillips.TimeCorrelatedSinglePhotonCounting. NewYork: AcademicPress,1984. JR Lakowicz. Topics in Fluorescence Spectroscopy, Vols I and 11. New York: Plenum Press,1992. LBMcGown,FVBright.Phase-resolvedfluorescencespectroscopy.AnalChem 56:1400A-I402A,1406A,1408A.1410A,1414A.1417A,1984. I Salmeen, L Rimai. A phase-shift fluorometer using a laser and a transverse electrooptic modulator for subnanosecond lifetime measurements. Biophys J 20:335342,1977. S Udenfriend. Fluorescence Assay in Biology and Medicine. New York: Academic Press,1969. RC Bennett. Instrument to Measure Fluorescence Lifetimesin the Millimicrosecond Region. Rev Sci Instrum 3 1 : 1275, 1960. RC Mackey. SA Pollack, RS Witte. Multiple Watt Submicrosecond High Repetition Rate Light Source and Its Application. Rev Sci Instrum 36:1715, 1965. OJ Steingraber, IB Berlmann. Versatile Technique for Measuring Fluorescence Decay Times in the Nanosecond Region. Rev Sci Instrum 34:524, 1963. RF Chen, GG Vurek, N Alexander. Fluorescence decay times: Proteins, coenzymes, and other compounds in water. Science 156:949-951, 1967. SJ Strickler, RA Berg. Relationship between absorption intensity and fluorescence lifetime of molecules. J Chem Phys 37:814-822, 1962. R Richards-Kortum, E Sevick-Muraca. Quantitative optical spectroscopy for tissue diagnosis. Annu Rev Phys Chem 47555-606, 1996. JR Lakowicz, H Szmacinski, P Koen. Emerging biomedical applications of timeresolvedfluorescencespectroscopy.ProcSPIEIntSOCOptEng2136:178-192, 1994. EM Sevick-Muraca, CL Hutchinson, DY Paithankar. Optical tissue biodiagnostics using fluorescence lifetime. Opt Photonics News 7:2428, 45, 1996. J Plasek, K Sigler. Slow fluorescent indicators of membrane potential: A survey of different approaches to probe response analysis. J Photochem Photobiol B 33:lOl124.1996.

8. AS Verkman, M Armijo,

9.

10. 11.

12. 13.

14. 15.

16. 17.

18.

19. 20. 21. 22. 23. 24. 25.

26. 27.

Fluorescence Lifetimes Using Applications

231

28. H Szmacinski,JR Lakowicz. Fluorescence lifetime-based sensing and imaging. Sens Actuators B B29: 16-24, 1995. 29. TM Nordlund. Streak cameras for time-domain fluorescence. In: JR Lakowicz, ed. Topics in Fluorescence Spectroscopy. New York: Plenum Press, 199 1, p 453. 30. NP Barry, JC Dainty, K Dowhg, PMW French, SCW Hyde, R Jones, R Mellish, JM Sutherland, JR Taylor, YP Tong. BHT Chai, CJ Van Der Poel, A Valster. Novel ultrafasttuneablesolidstatelasersforreal-worldapplicationsincludingmedical imaging. Proc SPIE Int Soc Opt Eng 3 176:304-313, 1997. 31. SA Soper, QL Mattingly. Steady-state and picosecond laser fluorescence studies of nonradiative pathways in tricarbocyanine dyes: Implications to the designof near-IR fluorochromes with high fluorescence efficiencies. J Am Chem Soc 116:3744-3752. 1994. 32. BL Legendre Jr. DC Williams, SA Soper, R Erdmann, U Ortmann, J Enderlein. An all solid-statenear-infraredtime-correlatedsinglephotoncountinginstrumentfor dynamic lifetime measurements in DNA sequencing applicatlons. Rev Sci Instrum 67:3984-3989,1996. U Lieberwirth. 33. MSauer,JArden-Jacob.KHDrexhage,NJMarx,AEKargcr, R Mueller, M Neumann, SEA Nord. Online diode laser based time-resolved fluorescence detection of labeled oligonucleotides in capillary gel electrophoresis. Biomed Chrornatogr11:81-82.1997. 34. S Cova.ALucaita, F Zappa,PLovati.Avalanchephotodiodesfornear-infrared photon counting. Proc SPIE Int SOC Opt Eng 2388:9, 1995. 35. PD Waters, DH Bums. Optimized gated detection for lifetime measurement over a wide range of single exponential decays. Appl Spectrosc 47: 1 1 1-1 15, 1993. 36. P Hall. B Selinger. Better estimates of exponential decay parameters. J Phys Chem 8512941-2946,1981. 37. SA Soper,BLLegendreJr.Erroranalysisofsimplealgorithmsfordetermining fluorescence lifetimes in ultradilute solutions. Appl Spectrosc 48:400405, 1994. 38. G Ponterini. F Momicchioli. Trans-cis photoisomerization mechanism of carbocyanines: Experimental check of theoretical models. Chem Phys I5 1: 1 I 1-126, 1991. 39. RM Ballew, JN Demas. An error analysisof the rapid lifetime determination method for the evaluation o f single exponential decays. Anal Chem 6130-33, 1989. 40. J Tellinghuisen, CWJ Wilkerson. Bias and precision in the estimation of exponential decay parameters from sparse data. Anal Chem 65: 1240-1246, 1993. 41. JC Mialocq, J Jaradias, P Goujon. Picosecond spectroscopy of pinacyanol. Chem Phys Lett 47: 123-1 26, 1977. 42. LJ Hofer, RJ Grabenstetter, EO Wiig. The fluorescence of cyanine and related dyes in the monomeric state. J Am Chem SOC 72:203-209, 1950. 43. N Serpone. MRV Sahyun. Photophysics of dithiacarbocyanine dyes: Subnanosecond relaxation dynamics of a dithia-2,2’-carbocyaninedye and its 9-methyl-substituted meso analog. J Phys Chem 98:734-737, 1994. 44. VA Kuzmin,APDarmanyan.Study of stericallyhinderedshort-livedisomers of polymethine dyes by laser photolysis. Chem Phys Lett 54:159-163. 1978. 45. SP Velsko, GR Fleming. Solvent influence on photochemical isomerizations: Photophysics of DODCI 13,3’-diethyloxadicarbocyanineiodide]. Chem Phys 65:59-70, 1982.

232 Legendre

and

Flanagan

46. SK Rentsch. Modeling of the fast photoisomerization process in polymethine dyes. Chem Phys 6931-87, 1982. 47. ASTatikolov,KSDzhulibekov,ZAKrasnaya.Studyoffluorescentaggregatesof polymethine dyes. Russ Chem Bull 42:60-66, 1993. 48. W West, S Pierce. The Dimeric State of Dyes in Cyanine. J Phys Chem 69:18941903,1965. 49. SR Mujumdar, RB Mujumdar, CM Grant, AS Waggoner. Cyanine-labeling reagents: Sulfobenzindocyanine succinmidyl esters. Bioconjugate Chem 7:35&362, 1996. 50. RBMujumdar,LAErnst,SRMujumdar,CJLewis,ASWaggoner.Cyaninedye labeling reagents: Sulfoindocyanine succinimidyl esters. Bioconjugate Chem 1054: 111.1993. 51. N Narayanan, G Patonay. A new method for the synthesis of heptamethine cyanine dyes: Synthesis of new near-infrared labels. J Org Chem 60:2391-2395, 1995. 52. M Lipowska, G Patonay, L Strekowski.New near-infrared cyanine dyes for labeling of proteins. Synth Commun 233087-3094. 1993. 53. AS Tatikolov, KS Dzhulibekov, LA Shvedova, VA Kuzmin. AA Ishchenko. Infuenceof“inert”counterionsonthephotochemistryofsomecationicpolymethine dyes. J Phys Chem 99:6525-6529, 1995. 54. BSauerwein,GBSchuster.Externaliodineatomsinfluenceovertheintersystem crossing rate of a cyanine iodide ion pair in solution. J Phys Chem 95: 1903-1906, 1991. 55. MDemchuk,AIshchenko,VMikhailov,VAvdeeva.Theinfluenceoftheanion on the excited-state relaxation time of cationic polymethine dyes. Chem Phys Lett 144~99-103,1988. YL Briks. YL Slomin56. AK Chibisov, GV Zakharova, VL Shapovalov, AI Tolmachev, skii.Flashphotolysisofpolymethinedyeswithvariouscounterionsinnon-polar media. High Energ Chem 29:192-198, 1995. 57. RBMujumdar,LAEmst,SRMujumdar,ASWaggoner.Cyaninedyelabeling reagents containing isothiocyanate groups. Cytometry 10:11-1 9, 1989. 58. PL Southwick, LA Ernst, EW Tauriello, SR Parker, RB Mujumdar, SR Mujumdar, HA Clever, AS Waggoner. Cyanine dye labeling reagents: Carboxymethylindocyanine succinimidyl esters. Cytometry 1 1 :418430, 1990. 59. LA Ernst. RK Gupta, RB Mujumdar, AS Waggoner. Cyanine dye labeling reagents for sulfhydryl groups. Cytometry 10:3-10, 1989. 60. DB Shealy, M Lipowska, J Lipowski, N Narayanan, S Sutter, L Strekowski, G Patonay. Synthesis, chromatographic separation and characterization of near-infraredlabeled DNA oligomers for use in DNA sequencing. Anal Chem 67:247-25 1, 1995. 61. L Strekowski, M Lipowska, G Patonay. Facile derivatizations of heptamethine cyanine dyes. Synth Commun 22:2593-2598, 1992. 62. JHFlanagan Jr, CVOwens, SE Romero,EWaddell, S Kahn,RPHammer,SA in DNA Soper. Near-infrared heavy-atom-modified fluorescent dyes for base-calling sequencing applications using temporal discrimination. Anal Chem 70:267&2684, 1998. 63. CC Leznoff, ABP Lever. Phthalocyanines: Properties and Applications. New York: VCH,1989.

Fluorescence Lifetimes Using Applications

233

64. GA Casay, T Czuppon, L Evans 111, G Patonay. Detection of toxic metal ions with near-infrared compounds. Proc SPIE Int SOCOpt Eng 2293:42-53, 1994. 65. GA Casay, N Narayanan, L Evans 111, T Czuppon, G Patonay. Near-infrared tetrasubstituted aluminum 2,3-naphthalocyanine dyes for optical fiber applications. Talanta 43: 1997-2005, 1996. 66. M Soncin, A Busetti, R Biolo, G Jori, G Kwag, Y-S Li, ME Kenney, MAJ Rodgers. Photoinactivation of amelanotic and melanotic melanoma cells sensitizedby axially substituted Si-naphthalocyanines. J Photochem Photobiol B 42:202-210, 1998. 67. G Jori. Novel therapeutic modalities based on photosensitized processes. EPA News1 60:12-18,1997. 68. EA Luk’yanets. Phthalocyanines and their analogs in new fields of technology. Mol Cryst Liq Cryst Sci Techno1 Sect C 1:209-216, 1992. 69. M Matsuoka. Infrared Absorbing Dyes. New York: Plenum Press, 1990, pp 1-212. 70. S Seeger, G Bahteler, KH Drexhage, J Arden-Jacob, G Deltau, K Galla, KT Han, R Muller, M Kollner, A Rumphorst, M Sauer, A Schulz, J Wolfrum. Biodiagnosticsandpolymeridentificationwithmultiplexdyes.BerBunsengesPhysChem 97:1542-1548,1993. 71. LB McGown, L-C Li. On-the-fly frequency-domain fluorescence lifetime detection in capillary electrophoresis. Anal Chem 68:2737-2743, 1996. 72. T Imasaka, A Yoshitake, K Hirata, Y Kawabata, N Ishibashi. Pulsed semiconductor laser fluorometry for lifetime measurements. Anal Chem 57:947-949, 1985. 73. G Bachteler, KH Drexhage, J Arden-Jacob, KT Han, M Kollner,R Muller, M Sauer, S Seeger, J Wolfrum. Sensitive fluorescence detection in capillary electrophoresis using laser diodes and multiplex dyes. J Lumin 62:lOl-108, 1994. 74. SA Soper, YY Davidson, JH Flanagan Jr, BL Legendre Jr, C Owens, DC Williams, RP Hammer. Micro-DNA sequence analysis capillary electrophoresis and near-IR fluorescence detection. Proc SPIE Int SOC Opt Eng 2680:235-246, 1996. 75. DL Farrens, P-S Song. Subnanosecond single photon timing measurements using a pulsed diode-laser. Photochem Photobiol 54:3 13-3 17, 199 1.

This Page Intentionally Left Blank

Applications in Polymers Olga V. Przhonska National Academy of Sciences of Ukraine, Kiev, Ukraine

1.

INTRODUCTION

The idea to incorporate organic dyes into polymer materials has attracted the attention of scientists for a long time. This interest joins the efforts of researchers fromthedifferentfields of physics,chemistry,biology,andhightechnology. Amongthevariousstudiesrelatedtothe“dye-in-polymer” field, threemajor approaches may be considered. The first approach is connected with the investigation of the physical and chemical properties of polymers themselves using afluorescenceprobetechnique. In thiscasedyemolecularprobeshavebecomea very powerfultoolforstudyingvariousphenomenainsyntheticand natural polymers. This is due to the high sensitivity of spectral parameters of the dyes to their microenvironment. Different probe methods are explored, inin the quantum yield of cluding steady-state luminescence (exploring changes emission,halfwidth,andbandshapes),time-resolvedmeasurements(analysis of fluorescence decay kinetics), and fluorescence depolarization methods (especially. time-resolved anisotropy decay). Many investigations have been made to characterize molecular mobility and chain dynamics [ 11, glass transition temperature (T,) for a variety of polymers and copolymers 121, relaxations at different temperatures,whichinvolvethemotions of bothlongandshortsegments of the polymer chain [3], viscosity of the microenvironment [4], and free volume effects [ 5 ] . Itwasshownthat in amorphouspolymerstheirmicroviscosity is less then their macroviscosity by many orders of magnitude. In particular, high segmental mobility is in a highly elastic state (at temperatures above T,), which is determined by the existence of microcavities of free volume inside the matrix and by the essential increase in their size in the elastic state compared with the glassy state [5].It is now well known that the main factors that influence the rate

235

236

Przhonska

of photophysical and photochemical reactions in polymers are the free volume size at a given temperature and the size of molecular groups or the scale of their motions relative to the free volume size. The second approach in “dye-in-polymer” studies involves the investigation of matrix effects on the spectroscopic properties of organic dyes. This aspect is very important for molecular design and synthesisof dyes with desirable propof dye molecules into the erties for many practical applications. Incorporation polymer matrix can change essentially their spectral characteristics such as the Stokes shift [6], fluorescence lifetime [7], formation of aggregates 181, molecular orientation, and conformational (trans-cis) changes in the dye molecules [4]. In of the polymer matrix (glassy or elastic) the latter case, polarity and the state are the essential features that primarily determine the dye parameters. Below we give an example of matrix effects on the dye properties. The third approach involves the development of dye-doped polymer systemsasthe key elementsforhightechnologyapplications.Nonlinearoptics, quantum electronics, optical data storage, solar energy converters, fluorescence lifetime standards, optical computer elements, and laser filter systems are a few typical examples of these applications. For further progress in these fast growing fields, dyes and polymers with special characteristics are required. This chapter describes the main requirements of polymer materials and organicdyesapplied in moderntechnological fields, especially in lasertechnologies, with emphasis on elastic polymers and near-infrared (NIR) dyes and their properties in elastic media. Two examples of applications are discussed in more detail: ( 1 ) active media for polymer dye lasers and (2) dye-doped polymer elements for optical limiting devices.

II. STATE OF DEVELOPMENT OF DYE-DOPED HOSTS The first publications relating to the various phenomena in dye-doped polymeric systemsappearedabout 30 yearsagoandwerestimulated by thediscovery of the first dye laser [9,10]. Since then dye lasers have become an important tool in our technological society. This is mostly due to their tunability, which permits precise excitationof various light-absorbing molecules for the purposeof stimulating photophysical and photochemical reactions. Some examples are the analysis of environmental media for pollutants, isotope separation, photodynamic therapy in medicine, laser diagnostic medical imaging, and numerous research applications. It is desirable that the laser light be available over the entire range of the spectrum from ultraviolet to infrared. A richvariety of laser dyes are now available. Although most of the current dye lasers employ dyes in solution, liquid dye lasers have many disadvantages and limitations. They usually require large and complicated systems for liquid circulation, are subject to leakage, are

Applications in Polymers

237

notportable,andarelimitedto a narrowrange of temperatureandpressure. The procedure for changing one dye to another is difficult and hazardous due to the toxicity of many dye solutions. Such lasers are therefore limited to fixed locations such as laboratories or diagnostic centers. Polymer dye lasers were developed in an attempt to avoid the disadvantages of liquid dye lasers and to combine their advantages (wide range of tunability, high efficiency, low thresholdof pumping intensities) with the advantages of solid media (compactness, portability, simple operation, easeof changing the dye, functionality under a wide range of ambient conditions such as pressure, temperature,vibrations). The first experiments on dyelaseraction in various polymeric media were performed in 1967-1968 [ 1 1,121, but success was minimalandmostly due to the low photochemical stability and laser damage resistance of these materials. Thus, for many years only liquid dye systems were developed. During this period works on polymeric active media appeared only sporadically [ 13-20). A second wave of interest in dye-doped polymeric media arose about I O years ago, stimulated by the development ofnew applications and information gained from liquid dye systems. Intensive research in the elaboration of dye-activated solid matrices with appropriate optical and photochemical properties has been carried outby a number of teams in the former Soviet Union, the United States, France, Spain, and the United Kingdom. Below we present a brief review of recent developments in this research area and a summary of the properties of polymeric active media developed in Ukraine (Kiev). To dea number of problems must velop solid-state materials based on organic dyes, be solved: selection of solid matrices, molecular design and synthesis of dyes with the desirable properties in solid media, and improvement in the technology of the fabrication process. Application of solid matrices in highly technological fields, especially in lasertechnologies,requiresmaterials of highopticalquality,alowlevel of scattering, a wide range of transparency, a high laser damage threshold, photochemical stability, and a simple technology of doping with different classes of organic dyes. Inorganic glasses usually show very good optical properties. However, the high melting temperature (-l00OoC) necessary for production leads to the fast thermal decomposition of most organic dyes. Polymeric hosts, whose not as good as those of mechanical, thermal, and photochemical properties are inorganic glasses, attract simpler technology of dye incorporation and a broad variety of polymeric materials and polymerization processes. Highopticalqualityofelementscanbeachieved by usingamorphous The mostpopularglassypolypolymers in glassyandhighlyelasticstates. meric hosts are polymethyl methacrylate (PMMA) and its modified derivatives. Typically,thepolymerizationprocedure is a freeradicalreactioninitiated by to 40-50°C. benzoyl peroxide. To produce a solid matrix the mixture is heated Solid samples are then cut and polished for laser experiments. It was found that

238

Przhonska

in spite of the relatively low heating temperature during the polymerization procedure, many organic dyes, especially NIR dyes, were decomposed. Therefore, it wasnecessary to find a way to avoidthisdisadvantage. The majordistinguishing feature of elastic polymers is the existence at room temperature of a highly elastic state characterized by a higher mobility of polymeric chain segments. This elastic property is responsible for the high resistance to radiation damage. Unlike glassy polymers, the rubbery nature of the elastic materials has the property of self-healing radiation-damaged chemical and physical cross-link bonds.Hencethesematerialscanbeusedunderintensepumpingirradiation with minimal damage. We have developed the elastic polymeric material polyurethane acrylate (PUA), whose optical, photochemical, and technological properties satisfy most of the requirements for laser optical materials [21,22]. The band of transparency of undopedPUAissituated in thenearultraviolet to infraredregion (3501600 nm), which makes it possible to use this material with a broad variety of organic dyes. Figure 1 demonstrates an absorption spectrum of pure, undoped PUA matrix with a thickness of 1 mm. This polymer differs from other polymeric materials in itsviscoelasticandadhesionproperties.Itsglasstemperature is about -50°C; thus, at room temperature PUA exists inside the zone of high

1.5

1.o

-e

c 0 ._

0 v)

2 0.5

0.0

Wavelength (nm)

Figure 1 An absorptionspectrum of pure PUA matrix with a thickness of 1 mm.

Applications in Polymers

239

elasticity. This material is also characterized by strong adhesion to optical glass, which is useful for manufacturing elements for laser optics. The microscopic of properties of PUA are determined by its structural features, the alternation links of different degrees of rigidity and chemical makeup (such as glycol and diisocyanate), and the presence of polymer functional groups (mostly, protondonor urethane). These groups are responsible for the formation of transverse cross-linkage due to physical bonds suchas hydrogen bonds as well as a network of transverse chemical cross-links. It was shown that the laser damage threshold of the PUA matrix is about 1.5 times higher than that of glassy PMMA 1221. Also,theself-healingeffectofthePUAmatrixleadstothepartialrecovery of radiation-damaged spots of the optical elements, which contributes to their longevity. We have developed and optimized a method of polymerization known as radical photopolymerization for the formation of polymer guest-host elements. a photosensitivemixtureof Thismethodconsists of severalstages.Initially oligourethane acrylate, dye, and photoinitiator is prepared. Next this mixture is displayed between two glass or quartz plates and irradiated at the absorption band of the photoinitiator, which carries out the polymerization reaction. The polymerization of oligomers, unlike that of monomers, is characterized by small volume shrinkage-only about 2-3% of its original volume. The dye molecules are not chemically bound; instead, they are dissolved in a polymer matrix. The advantage of this method is that dye molecules do not aggregate and do not complex with the polymeric matrix and therefore their optical absorption characteristics do not change significantly from those previously determinedin liquid media. With this technology, it is possible to incorporate many different classes of organic dyes into the polymeric matrix. The elastic properties of PUA and peculiarities in theradicalphotopolymerizationproceduredeterminetheconin which polystruction of the polymer elements. They represent “triplexes,” meric film is placed between two glass or quartz plates, or “multiplexes” with two or more polymeric layers. Strong adhesion between the surface of the plates and the polymer layer is achieved by intermolecular interaction. This method of enclosure eliminates the need of polishing, which is difficult for elastomers. The glass or quartz plates are necessary (1) to carry out the photopolymerization reaction, because atmospheric oxygen inhibits the reaction; ( 2 ) to establish high optical quality of the polymer surface, which is important for optical application; (3) to minimize thermo-optical heating under intense irradiation by providing a thermal conductivity enclosure; and (4) to protect the optical element from mechanical damage and atmospheric exposure. Protection of the optical element from atmospheric oxygen reduces the photo-oxidation, whichis one of the major mechanisms for photodegradation of organic dyes. Alternative host materials and technologies are being developed [23-29]. Therearetwo maintechnologicalapproaches:sol-gelglassesandorganically

Przhonska

240

modifiedsilicate glasses (Ormosil). Sol-gel glasses allow the incorporation of organic dyes either by means ofapredopingprocedure, in whichthedyeis added to the chemical components before polymerization, or by a postdoping is diffusedintotheporesofthehostmaterial procedure, in whichthedye fromthesolution. The opticalquality of theporeglassesmaybeimproved by using index-matching materials. One of the suitable materials is PMMA. A sol gel-PMMA composite is known as Polycom glass. In the case of Ormosil glasses, the PMMA is covalently bonded to an inorganic component, producing a nonporous organic-inorganic composite. A most complete comparison of the characteristics of organic dyes in sol-gel, Polycom, and Ormosil glasses as well as in the PMMA matrix is given in Refs. 28 and 29.

111.

STATE OF DEVELOPMENT OF NIR DYES FOR POLYMERIC HOSTS

A.

NIR Dyes

Among the dyes absorbing and emitting in the NIR region the phthalocyanine and polymethine dyes are the most important and most widely used. Polymethinedyes(PDs)representalargeclass of organiccompounds thatcontain a conjugated chain of methine CH=CH groups (polymethine chain) as a basic constitutive element and have absorption maxima of up to 1600 nm [30]. ThemainpropertiesofPDsaredetermined by theexistence of thedelocalized r-electron system in the polymethine chain and two identical or different end (terminal) groups. Polymethine dyes are attractive owing to the possibility of systematically modifying their structure by using different heterocyclic endgroups,introducingspecificsubstitutesintothepolymethinechain,branching the polymethine chromophore, and cyclizing the chain with conjugated or unconjugated bridges. Many correlations between the molecular structure of PDs and their spectroscopic parameters have already been established (31,321. Therefore, there is a possibility of making predictable changes in the dye molecular in theeffective structure to obtainthedesiredspectralparameters.Achange length of the conjugated chain (number of methine groups) leads to a change in the position of the main absorption band So + SI.The simplest way to shift the absorption maximum to the red region is to increase the number of CH=CH in the chemistry groups in the polymethine chain. This has been well known a long polymethine of polymethine dyes since the 1930s. However, PDs with chromophore are chemically and photochemically unstable, which limits their applications. Another way to shift the absorption spectrum to the red and yet retain the same number of CH=CH groups is to increase the effective length of the x-electron system at the terminal chromophore groups (“heavy” terminal groups) and put the bridged groups and substitutes with different electronic

Applications in Polymers

241

structureintothepolymethinechain. In thiscasetheinteractionbetweenthe two conjugated n-electron systems in the polymethine chain and the terminal chromophore leads to a lowering of the energy level of the first excited state. This method is currently being explored. Reviews on the chemistry of PDs with typical examples of their molecular structures may be found in Refs. 30 and 32.

B. Spectroscopic Properties of Dyes in Polymeric Media It has been known that dye molecules i n an amorphous liquid and in solid media may be considered an ensemble of centers differing in configuration of the bcal on the ratio of the fluorescence environment. Their spectral behavior depends lifetime (SF) to the relaxation time(SR)within the excited State. Usually the effect of the local environment is studied in the two cases TR > SF (stationary environment). In the first case there is a dynamic averaging of localenvironments,producingmicroenvironmentalhomogeneity, forexample,dyemolecules in lowviscositysolutions at roomtemperature. The second case is typical for systems with microenvironmental heterogeneity, for example, dye molecules in frozen solutions. The spectral behavior of dye molecules in polymeric media is a less investigated problem. It reflects both the macrorigidity of the polymeric matrix and the free volume effect. Investigations of theinfluence of thepolymermatrixondyepropertiesgivesignificantinsight into studies of trans-cis isomerization processes. It is known that trans-cis of polymethinedyes in lowviscosity photoisomerization in theexcitedstate solutionsproducesone of themostimportantnonradiativepathways[33,34]. Therefore, the fluorescence kinetics usually reflects two types of excited state relaxation processes such as conformational changes (isomerization) and rearrangement of solvent-solute cages. The quantum yield of isomerization strongly In depends on the dye molecular structure, temperature, and solvent viscosity. liquid solutions it is usually difficult to distinguish between these two types of relaxations.However, in polymericmedialarge-scaleconformationalchanges such as trans-cis isomerization of the polymethine chromophore are restricted due to macrorigidity.In this case the fluorescence kinetics reflects mostly the role of microenvironmental heterogeneity and its dynamics. Quantum-chemical calculations show that only small-scale motions such as rotationsof methyl, phenyl, or dimethylamino groups are almost barrierless and are allowed in polymeric media during SF due to free volume effects [35].Decreasing the probability of conformationalmotionsgivespolymersanadvantageforstudyingtheeffects of relaxations. Below we describe some important spectral peculiarities of dyedoped polymeric media and typical experimental methods for their investigation. A typical effect of polymeric media on spectroscopic properties of the dyes can be illustrated by using a well-known polymethine dye, indopentamethinecyanine perchlorate, PD 643 (643 nm is a position of the absorption maximum in

242

Przhonska

ethanol at room temperature), as a probe 171. The absorption and fluorescence spectra of PD 643 in ethanol solution and PUA matrix at room temperature are shown in Figure 2. The absorption spectrum in PUA (Fig. 2b, curve I ) demonstrates only a slight shift, approximately 8 nm, to longer wavelengths without full width at half maximum (FWHM) significant changes in the structure and compared with ethanol solution (Fig. 2a, curve 1). The fluorescence spectrum in ethanol solution at room temperature (Fig. 2a, curve 2) does not depend on the excitation wavelength l e xOne . of the important peculiarities i n polymeric media is the dependence of the fluorescence maximum on &. At excitation longer than the absorption maximum, a shiftof the fluorescence spectra to longer wavelengths is observed (Fig. 2b, curve 3). This effect is known in the literature as the red-edge effect; it is observed in highly viscous and glassy low molecular weight solvents and presents itself as a gradual shift of the fluorescence maximum to the red on excitations longer than the absorption peak [36].In contrast to the red-edge effect, in elastic PUA the shift of l e xto longer wavelengths results in anti-Stokes fluorescence (Fig. 2b, curve 3) with the existence of a “red limit” In this case, the limiting shift is in the position of the fluorescence maximum. about 15 nm (Fig. 2b, compare curves 2 and 3). Decreasing the temperature for PD 643 in PUA leads to changes in the position of the fluorescence maximum and FWHM. These dependences are shown in Figure 2c. There is a decrease in both values from room temperature to about 200 K. Below this temperature both relationships become temperature-independent. This region (around 200 K) corresponds to the glass transition region of PUA. Thus, from regular steadystate spectroscopy it is possible to obtain information about the glass transition range of the polymer by using polymethine dye as a molecular probe. A powerful method for investigating the fast dynamics in polymeric matrices is time-resolved spectroscopy, which adds the dimension of time to the steady-state measurements. For this purposeit is usually sufficient to measure the fluorescence decay curves of the organic dye in polymer by means of a picosecond time-correlated single-photon counting technique. Analysis of the fluorescence intensity decay data as a sum of the exponentials, I ( [ ) = u ; exp(-f/ti), where a; is the relative weight of the components with lifetimes t,,may be performed by several methods. The most advancedis the maximum entropy method (MEM), which offers new possibilities in the study of molecular relaxations. Important advantages of this method are independence from any a priori models, numbers of components, and signsof their amplitudes [37,381. As an example of the possibilities of the time-resolved methods, Figure3 demonstrates the changes in theemissionspectra of PD 643 in PUAovertime,theso-calleddynamic Stokes shift. This dependence gives information about the relaxation processes on a time scale comparable with t ~The . fluorescence maximum shifts by about 4 nm during a period of 6 nsec. The most essential shift takes place within the first 1-2 nsec, which is comparable with the lifetime in PUA t~ X 2 nsec. As

243

Applications in Polymers 525

503

09

-

05

-

550

575

ea

825

0%

875

725

70)

750

C

.-

E 0 2

04

-

Wavelength (nm)

SW

1 1

_

525

,

_ 5yI ,

_575 ,

Mo

.

,

625

.

.

650

.

,

675

.

,

703

.

,

125

.

,

7501

.

1

Wavelength (nm)

-E v

5

662 5

-

6620

-

661 5

-

._ E 6610 X

rn 660 5

8 8

6600

-

2 6595 0

6590

(c)

-

Temperature (K)

Figure 2 Steady-state fluorescence measurements of the polymethine dyc indopentamethinecyanine perchlorate. (a) Absorption (curve I ) and fluorescence (curve 2) spectra i n ethanolsolutionatroomtemperature; (b) absorption (curve 1) andfluorescencespectra at hex = 600 nm (curve 2) and hex = 720 nm (curve 3 ) inPUA at room temperature; (c) position of fluorescencemaximum (2) and full widthathalf-maximum ( I ) i n PUA a s functions of temperature. hex = 573 nm.

244

660

650

640

Przhonska

630 6000

h

5000

v) 4-8

C

2

4000

0

v

Q)

2

3000

8In

p! 2000 0

3 LL

1000 0

660

650

640

_I

~

630

Wavelength (nm)

--" I 662 -

659 ,

0.0

0.5

.

I

10

.

I

15

.

,

2.0

.

I

.

I

2.5 3.53.0

.

I

.

I

.

,

4.0 5.0 4.5

.

I

.

5

Time after excitation(nsec) Figure 3 Time-resolved measurements of indopentamethinecyanine perchloratei n PUA at room temperature. (a) Fluorescence spectra measured at different times ( t ) after excitation: t = 0 (curve I ) and t = 5.3 nsec (curve 2 ) ; (b) dependence of fluorescence nlaxirnuIn on the delay time after picosecond excitation at lex = 573 nm. (Time-resolved measurements were performed by U. Stahl, 0 . Przhonska. and s. Daehne a1 Federal Institute for Materials Research and Testing, Berlin. Germany in 1995. The dye was synthesized by Yu. Slominsky at the Institute of Organic Chemistry, Kiev, Ukraine.)

Applications in Polymers

245

w:ls alreadymentioned, in polymericmedialarge-scaleconfigurationchanges are restricted by macrorigidity. Therefore, the origin of the dynamic Stokes shift of is the inhomogeneity of the polymeric medium arising from a distribution solvent-solute orientations. The relatively small value in the dynamic shift may be explained by a small change in the dipole moment of the So + SI transition, which is typical for cationicpolymethinedyeswithadistributedrr-electron system.Thus,despite themacrorigidity of thepolymericmedium,there is a dynamic rearrangement of the medium cages leading to a lowering of the excited state energy of dye molecules. This fact is important for the application of elastic dye-doped polymers in high technological fields. A further development of time-resolved spectroscopy is the time-resolved anisotropymethodarisingfromtheuseofapolarizedlightsource.Thisexloss of citationcreates an anisotropicpopulation of excitedmolecules.The anisotropy with time is connected with the reorientational processes in the excited state, which reduce the number of dye molecules aligned in the direction of the electric field. The nature of these processes is determined mainly by the overall rotation and the intramolecular configurational changes. For low viscosity solutions, polarization kinetics may reflect both the overall rotation and the configurational change processes. There are practically no data on time-resolved anisotropy in polymeric matrices. The anisotropy function A ( t ) is determined by the expression [39]

where III(t) and f i ( t ) arethefluorescencedecayintensitiesparallelandperpendicular to thepolarization of theexcitationbeam.Asanexample of the 4 demonstratesthetime-dependent time-resolvedanisotropymethod,Figure anisotropy functions of the polymethine dye ,!I-dimethylamino1.1’,3,3.3’,3’-hexamethylindodicarbocyanine perchlorate in frozen ethanol (curve 1 ) andPUA at room temperature (elastic state, curve 3 ) and at 100 K (glassy state, curve 2). These results show that the anisotropyof fluorescence emission in frozen ethanol after a short decay does not change as a function of time during 20 nsec. Therefore, wecan conclude that no significant reorientation of the emission dipole occurs on this time scale. In contrast to ethanol glass, in a glassy PUA matrix the time-dependent decay of anisotropy has been observed with a correlation time of about 40 nsec. It is essential that these dynamical processes develop on a time scale much longer than the fluorescence decay (TF = 6.5-7 nsec in glassy PUA). At room temperature (elastic state) the anisotropy decay is characterized by more complicated kinetics: the first component with a fast decay (time scale ) the second component with a much longer decay time, comparable with t ~and in the anisotropy kinetics on a time scale We did not find essential differences of 25 nsecbetweenPUA in theglassyandelasticstates.Thisdecaymaybe

Przhonska

246

0.15

0

5

10

15

20

25

Time (nsec) Figure 4 Time-resolvedanisotropymeasurements of P-dimethyIamino-l,l’.3,3.3/,3’hexamethylindodicarbocyanine perchlorate in frozen ethanol at 100 K (curve 1) and PUA at roomtemperature(elasticstate, curve 3 ) and 100 K (glassy state, curve 2 ) ; Aex = 520 nm. (Time-resolved anisotropy measurements were performed by J. Gallay, M. Vincent, and 0. Przhonska in LURE, Orsay, France in 1996. The dye was synthesized by Yu. Slominsky atthe Institute of Organic Chemistry, Kiev, Ukraine.)

explained by the slow overall rotation of dye molecules or their fragments in the polymer microcavities of free volume, The origin of the fast decay that was observed in all three cases but with different amplitudes may be connected with the possibility of an intramolecular dynamic rearrangement leading to reorientation of theemissiondipole.Thus,ourobservationsarethattheanisotropy in PUA in boththeelastic andglassystatesisafunction of time, in contrast to that of ethanol glass. These results deserve attention and require further investigation.

C.

RequirementstoLaserDyes

in Polymers

The requirements to organic dyes for laser action are well known: high fluorescence quantum yield (for low laser threshold), high absorption cross sections on pumping wavelength, low losses on excited state absorption from the first excited state and other nonradiational processes such as intersystem crossing, triplet-triplet absorption, and isomerization. One of the major requirements is photochemical stability of the dyes, which is extremely important for solid ma-

Applications in Polymers

247

trices in the absence of circulating systems. The effect of the polymeric matrix onlaserproperties is variedandhasbothpositiveandnegative aspects [40]. Incorporation of organic dyes into polymeric media often leads to an essential increase in the fluorescence quantum yield due to restriction of molecular motion and isomerization processes. This positive influence of the medium reduces the laser threshold, which is important for practical applications. The negative effect of the solid medium is connected with the inhomogeneity of the amorphous matrix, leading to a decrease in the value of the absorption cross section at the pumping wavelength and slowing down of the relaxation processes in the excited state. In many cases these effects reduce the laser efficiency and make it difficult to obtain laser action with a large red shift between the pumping and laser wavelengths. Twomajorfactorslimittheapplication of polymericelements i n laser systems:photochemicalinstability of organicdyesandlimitedshiftbetween pumping and laser wavelengths, which makes it difficult to tune the laser wavelength to theredregion. The variousprocessesleading to thedecomposition of organic dyes in polymeric media are strongly dependent on both the properties of the matrix itself (polarity. existence of some functional groups) and the parameters of light irradiation(wavelength,inputfluenceorintensity).These of the dyes at the processes may be separated into two stages: decomposition polymerization stage and decomposition of the dyes in solid polymer matrices under irradiation of different intensities in the main absorption band. We studied these processes for dye-doped PUA. The conclusions are the following: 1.

The main origin of the photodecomposition of dyes during radical photopolymerization is the interaction between the excited dye molecules and the macroradicals of the medium [41]. Selection of the proper initiator and its concentration as well as the proper choice of exposure wavelengths, which have to be different from the absorption bands of the doping dye, can reduce photodecomposition of the dyes at this stage. 2. The next stage in thestudy of thephotostability of dye-dopedsolid matrices is irradiation of the dye molecules into the main absorption band by lowpowerpumping,whichinvolvesone-photonprocesses of photoonly. In this case for the most well-known laser dyes the rate bleaching is controlled by the transient diffusion of dissolved oxygen. The excited state reaction of the dye with the dissolved triplet oxygen may result in generation of singlet oxygen, which is chemically active. Thus, this reaction leads to irreversible processes of dye transformation into the bleached molecules with absorption in the near-UV region. The photopolymerization reaction depletes the free oxygen of the medium, while oxygen diffusion into the polymer “triplex” from

248

Przhonska

outside is negligibly slow. This feature eliminates the photo-oxidation reaction and increases the photostability of the dye molecules in solid matricescomparedwithliquidsolutions.Theadditionalremoval of oxygen from the polymeric matrix results in considerable photostabilization (about IO-fold for PDs), which is important for applications [22]. We showed that PDs in PUA at lowpowerpumpingareone of the most photochemically stable classes of dyes (photobleaching quantum yield-IO”) 1221. The same trend of increased photostability in polymeric media was observed for dyes in sol-gel matrices. It was shown that the photostability of Rhodamine 6G, Rhodamine B, and Pyrromethene 567 at low power pumping was approximately twoorders ofmagnitudehigherthan in solutions[42,43].Thiseffect was explained by partial isolation of dye molecules from reactive impurities such as atmospheric oxygen. The illumination of dye molecules into the main absorption band at high powerpumpinginvolvestwo-stepprocessesandpopulation of thesecond or higher excited states. In this case the probability of photobleaching increases withtheincrease in pumping intensity, which is connectedwiththeeffective channel of irreversible transitions from the higher excited states. The strongest dependencewasobservedforPDs[44].Thus,PDs,beingthemoststable at low power pumping, decrease in photostability at high power pumping. Experiments showed that the highest photostability under strong pumping was found for phenalenone dyes [44]. For the dye-doped sol-gel matrices used as an active medium in pulsed lasers, the existence of thermodegradational processes was reported [42]. For rhodamines, DCM, and Neon Red dyes the operational lifetimes decrease with increases in the repetition rate. Under the same conditions the operational lifetime of pyrromethene dyes remains constant. Thus, for some dyes the thermodegradation processes may be very important. Successful laser performance of solid guest-host systems was achieved by usingvariousorganicdyes. In recentyears,thehighefficiencypyrromethene dyes based on substituted BF2 complexes have been synthesized. They are characterizedbyahighfluorescencequantumyieldandsmall triplet-triplet absorption [451. The highestlaserefficiency in bothliquidsolutionsandsolid matrices was obtained for Pyrromethene 567 (77% in modified acrylic plastic), pyrromethene 570 (85% in modified acrylic plastic) 1461, and Pyrromethene 580 and 567 (65% in PUA) [47]. Most of the reported solid guest-host active media showed the tuning range to be in the visible area of the spectrum, 570-620 nm. The tuning curve of the DCM-doped gel matrix covers 590-650 nm [26]. Tuning range of the sulforhodamine 640-doped silica gel laser is 610-645 nm [48]. A polymer dye laser based on PUA matrix has shown a tuning range of 550750 nm with four different dyes [49].

Applications in Polymers

249

There is strong demand for extensionof the tuning rangeto the NIR region. Important developments are expected in the field of biomedical applications, in particularforbiomedicaldiagnostics andmedicaltherapy. In thiscase, NIR lasers should be used, since they operate in the range to which human tissue is transparent. Therefore an extensive red shift between the pumping (usually is required second harmonic of the Nd:YAG laser, 532 nm) and laser wavelengths withoutsignificantdiminutioninlaserefficiency.Thisproblemhas not been solvedyet.Mostofthephysicalmechanismsleading to alargeredshift in liquid solutions such as proton transfer reactions [50,51] or solvent-dependent in effects in unsymmetricalpolymethinedyes[52] do not workwellenough solid matrices due to restrictions in the molecular motion. Thus, extension of the tuning range into the red region requires investigations into the choice of the optimal photophysical mechanism for obtaining the red shift in the emission spectrum,thechoiceandsynthesis of dyesexhibitingthisredshift in solid media, and development of technology for the preparation of active media that have desirable properties in the NIR region.

IV.

POLYMERDYELASERAPPLICATIONS

The majorproblems to besolved in polymerdyelaserdesignarethelocal heatingandphotochemicalbleaching of theactivemediumproducedbythe pumpingbeam.Tominimizetheseeffects, it isnecessary to eitherrotateor scan the active element in the laser resonator. Currently, we have several dye laser designs based on the PUA matrix. These lasers were developed for laser action in the visible and NIR regions. They are pumped by the second harmonic of the Nd:YAG solid-state laser (532 nm) with the following optimal characteristics: pulse energy 5-10 mJ, pulse duration 5-15 nsec, repetition rate up to 25 Hz, beam diameter not more than 4 mm, pumping intensity on the active element 40-80 MW/cm2(fluence0.4-0.8J/cm2),pumpbeamdivergencenot more than 3 mrad, pump polarization horizontal. In the simplest design [491 the polymeric active element is a removable plane-parallel circular disk, which is rotated in the laser resonator. The lateral movement of the disk may be manual or automated. A more complicated laser design includes a two-dimensional linear scanner device. This permits all regions of the active element to be exposed equally to the pumping beam, which avoids localized heating and retards photodegradation. Tuning curves for four active elements, which cover the spectral range 545-750 nm, are shown in Figure 5. The best laser efficiency in the dispersionresonator(two 60" prismsandtwodielectricmirrorswith reflecof 12 cm) wasobtained tioncoefficients of 60% and 99.5% andfulllength 580 show forpyrromethenedyes.BothPyrromethene567andPyrromethene a tuning range of 545-585 nm with laser efficiency of up to 65% [47]. Operational lifetimes for the polymeric elements with the scanning active area are

Przhonska

250

Wavelength (nm) Figure 5 Tuning curves for polymer dye laser based on dyc-doped PUA. Pyrromethene 567: 2. Rhodamine 6G: 3. 6-amino-phcnolenone: 4, Styril 6.

I,

1.o

E

.-

c

2 m

0.8

K

0 '=

Ea,

06

a 0

wm

0.4

._ 0.2 0

z

0.0

0.5

1.0

1.5

2.0

2.5

3.0 3.5

4.0 4.5

5.0

5.5

0

Number of laser pulses (in thousands) Figure 6 Operational lifetimes for polymer active media based on dye-doped PUA. 6-Amino-phenolenone; 2, Pyrromethene 567; 3, Rhodamme 6G.

I.

Applications in Polymers

251

in the range of several million pulses. The most photochemically stable dye is 6-aminophenalenone. Figure 6 shows the changesin normalized operational lifetimes on a number of pumping pulses (intensity % 80 MW/cm2) at one spot for 6-aminophenalenone, Pyrromethene 567, and Rhodamine 6G. As can be seen, active elements based on pyrromethene dyes are preferable to thosebased on Rhodamine 6G. The results of the development of dye-doped solid matrices for tunable lasersareveryencouraging.Theefficienciesandlifetimesarequitesuitable for practicalapplications.These highquality,compact,andinexpensivedye laserscouldfindnumerousapplications in medicine,industry,andresearch. The future of polymer dye lasers is connected with their design into microlaser structures such as thin-film organic light-emitting diodes and electrically driven laser organic materials. In this respect, the use of a zigzag resonator geometry [ S 3 ] , distributedBragg reflector, anddistributedfeedbackconfigurations [S4] may be advantageous for future technologies.

V.

DYE-DOPEDPOLYMER SYSTEMS FOR OPTICAL LIMITING APPLICATIONS

Organic molecules exhibiting strong excited state absorption are currently a subject of considerable interest for power limiting applications. These applications include the development of optical limiting devices, which protect human eyes and sensitive components of optical systems from laser-induced damage [55,56]. At high incident fluences (or intensities), organic molecules can demonstrate saturable absorption (SA) or reverse saturable absorption (RSA). Their behavior depends on the ratio between the absorption cross section from the first to the higher excited state ( ~ 1 2 )and from the ground to the first excited state ( ~ 0 1 )on pumping wavelength (see Fig. 7). Saturable absorption is observed when the absorption coefficient decreases with an increase in incident fluence, whereas RSA is observed when the absorption coefficient increases withan increase in incident fluence due to strong excited state absorption. Passive optical limiting has been demonstrated by using RSA in various organic materials. Among organic dyes the best results were obtained for nletallophthalocyanines and metallonaphthalo58 andreferencestherein).One of cyanines in liquidsolutions(Refs.57and the promising approaches for limiting applications is the development of solid guest-host systems. Atpresentthereareonlyafewpublicationsconcerning dye-doped solid matrices 159-621. The main requirements for solid optical limiters are high linear and low nonlinear transmittance levels, low power limiting threshold,largepowerdynamicrange,highdamagethreshold,photochemical stability, fast nonlinear response (picoseconds or faster), and broadband response (covering the entire visible spectrum). Typically the solid host is a transparent

252

Przhonska

Figure 7 (a) Three-levelmodelfororganicdyes. (b) Calculateddependencesofthe transmittanceversusintensity;curve 1, 001 > 0 1 2 ; curve 2, “ 0 1 = “12; curve 3, “01 < “12.

polymer such as polymethyl methacrylate, an epoxy compound, or PUA matrix and the guest is an organic dye showing optical limiting behavior. The performance is promising but still below the requirements for practical devices. The of the host medium major limitations are the relatively low damage threshold and photochemical instability of the dyes. Therefore, more work is needed on both host material improvement and dye optimization. Below we present the main results of a study of nonlinear absorption in a series of polymethine dyes. The molecular structures of the best dyes are shown in Table 1. Theroomtemperaturelinearabsorptionspectra of thesedyes in ethanol and PUA are presented in Figure 8. PD #1 with an unsubstituted polymethine chain has the shortest absorption maximum. For this dye there is almost no difference between the absorption bands in ethanol and in PUA. Inclusion of the six-link cycle with a phenyl substitute in the polymethine chain (PD #3) shifts the absorption spectrum to the red region by 15 nm compared to PD #l. The absorption spectrum of PD #6, which differs from PD #3 only by having a more complicated n-electron system in the terminal chromophore groups, is

Table 1 Molecular Structure and Parameters of Polymethine Dyes"

PD #1 Parameter kmax(abs). nm Thickness, mm TL a01 ( x lo1*cm2) q z ( x 10'8 cm2) 012/001

rsI (nsec) rs2 (psec)

PD #6

PD #3

Ethanol

PUA

Ethanol

PUA

Ethanol

PUA

755 5.0 0.67 4.5 f 1.0 5.5 1.5 120 f40 1.1 f 0 . 3 -0.5

756 2.0 0.69 5.3 f 0.7 2.6 f 0.5 5 0 f 10 2.1 zk 0.5 -4

770 2.0 0.78 1.5 z!z 0.3 3.0 f0.8 200 f 65 1 .o z!z 0.2 -2.0

78 1 2.0 0.78 2.0 zt 0.3 3.0 f 0.6 150 f 30 2.5 zt 0.6 -3.0

807 1.o 0.95 1.8 f 0.4 1.9 f 0.5 100 f 30 0.5 f0. I

818 1 .o 0.81 4.7 f 1.0 1.5 zk 0.3 32 f 7 I .o i0.2 -1.5

+

"Polymethine dyes were synthesized by Yu. Slominsky, Institute of Organic Chemistry, Kiev. Ukraine

-1.5

Prthonska

254

t

1

Wavelength (nm) Figure 8 Absorption spectra of polymethine dyes PD # I , PD #3. and PD #6 in PUA (solid lines) and ethanol solutions (dotted lines).

shifted by 37 nm compared to PD #3. The absorption spectra of PD #3 and PD #6 in PUA show a red shift of 1 1 nm compared to their absorption spectra in ethanol solutions. Nonlinear characterization of the dyes was performed at the Centerfor Researchand Education in OpticsandLasers(CREOL), University of Central Florida, Orlando. We used three well-developed techniques: Z-scan [58,63], optical limiting measurements, and picosecond pump-probe experiments. Z-Scanmeasurements were performed usingthe secondharmonic of the Nd:YAG laser (532nm) in the picosecond (30 psec)andnanosecond (10 nsec) regimes with a repetitionratethatcould be adjusted from a single shot up to 10 Hz. The Z-scan involved measuring the transmitted energy as a function of the sample position Z relative to the focal position of the pumping beam. The incident energy was fixed for the duration of the Z scan. The range of energies was 0.01-70 pJ. For all Z scans the beam was focused to a waist of radius 22 p m half width at I/e’ maximum (HWl/e2M) for picosecond pulses and 30 p m ( H W l / e 2 M )for nanosecond pulses. All transmitted energy was detected in an open-aperture configuration with care taken to collect all the energy, so our experiments were sensitive to nonlinear absorption only. Optical limiting curves were measured with 10 nsec (FWHM), 10 Hz, 532 nm laser pulses. The samples were located at the position of the pumping beam. which corresponds to the minimum transmittance and was found from Z-scan measurements. The

Applications in Polymers

255

pumping beam was focused to a waist of radius 5 ,um (HWl/e2M). Pump-probe experiments were used to study the dynamicsof photoinduced absorption and the determination of excited state lifetimes. These measurements included pumping in the materials, the samples with a strong beam, which induced nonlinearity a weak beam, which can and probing the relaxation of this nonlinearity with be delayed up to 15 nsec. Pump-probe measurements were performed for probe pulses polarized parallel and perpendicular to the pump pulse. The pump beam was focused to a waist of radius 230 p m (HWl/e’M), while the probe beam was focused to a waist of radius 34 pm (HWl/e2M). The range of pumping energieswas 10-150 pJ. Probeintensitywaskeptmuchlessthanthepump intensity, so the probe beam did not induce any nonlinearity. The two pulses were recombined at a small angle (about 5”) within the samples. The results are presented in Figures 9 and IO. Figure 9 shows the openaperture Z scans for one of the best dyes, PD #3 in PUA and ethanol. As can be seen, in the picosecond regime, there is an approximately fourfold reduction in TL (TL = 78%, transmittance at high fluence compared to linear transmittance thickness of thesampleis2mm). Atfluencesgreaterthan 0.3 J/cm2(irra0.7 J/cm2(irradiance diance > 9 GW/cm’) for PUAmatrixandgreaterthan 1 2 0 GWkm’) for ethanol solution, deformation in the curves was observed, which may be connected to laser-induced damage of the matrix and photochemical instabilityofthedyeunderstronglaserirradiation. The opticallimiting responseofPD #3 in PUA is presented in Figure 10. Ascanbeseenfrom the limiting curve, beginning from a linear transmittance of 78%, there is a decrease to approximately 15% at a fluence of 12 J/cm2. Molecular parameters of the dyes determined from linear absorption spectra, 2 scans, and pump-probe measurements are presented in Table I . The main advantage of these dyes is a large a12/a01 ratio, higher than any previously reported values. For PD #3 in PUA this ratio is up to as much as 150. This feature makes the dyes suitable for low threshold optical limiting applications. Our investigations also show the current problems in application of these materials: (1) relatively short lifetimes in the first excited state; ( 2 ) relatively long lifetimes in the second excited state, ( 3 ) photochemical which typically lead to low nonlinear saturation levels; and instability. We expect that further research and development of PDs may produce new dyes with increased rsI, reduced upper-state saturation, and improved photochemical stability.

VI.

CONCLUSIONS AND PROSPECTS

In this review we have shown that dye-doped polymer matrices and their properties have formed the basis for high technology applications. Future progress i n these fields will include several areas of research.

256

Przhonska

1.0

g

0.9

c

.-E 0.8 E

E

0.7

2

0.6

.-8 5

0.5

E

b

z

0.4

Figure 9 Z-scandata for PD #3 for picosecondregimein (a) PUAand (b) ethanol. Lineartransmittance is 78%. Thickness of thesample is 2 mm. (a) Energy E = 0.1 (curve I ) . 0.5 (curve 2). 1.0 (curve 3) and 2.0 p J (curve 4). (b) E = 0.15 (curve I), 1.2 (curve 2) and 4.5 pJ (curve 3). (Results obtained by J. H. Lim. 0. Przhonska, D. Hagan, and E. Van Stryland at CREOL, University of Central Florida. in 1997.)

Applications in Polymers

257

I

0.9 "O

Figure 10 Limiting curve for PD #3 in PUA. Linear transmittance is 78%. Thickness of the sample is 2 mm. (Results obtained by J. H. Lim, S. Yang, 0. Przonska, D. Hagan, and E. Van Stryland at CREOL, University of Central Florida, in 1997.)

First, technological progress strongly depends on the developmentof novel instrumentation that is compact, portable, convenient, simple in operation, and functional under a broad range of ambient conditions. Examples of such instrua mentation consist of tunable polymer dye lasers as useful light sources for variety of applications in science,medicine,andtechnology;polymeroptical limitingdevicesforeyeandequipmentprotection;andpolymerstandards of fluorescence quantum yield, fluorescence lifetime, and fluorescence anisotropy as an essential part of modern spectral devices. Second, dye-doped polymer matrices show promise as prospective model systems for understanding the behavior of complicated biological systems. For example, highly elastic polymers are potentially applicable as microscopic model media for the dynamic behavior of proteins [64]. Highly elastic polymers and proteins share similar properties, since they are characterized by the existence in of a broad range of microscopic conformational substates. These substates proteins may compose a unique conformation of folded polypeptide chain and in polymers produce a diversity of properties within structures with the same microscopic order. Both in polymers and in proteins this microscopic heterogeneity results in inhomogeneous broadening of the spectra. In proteins it may also result in the inhomogeneous (dispersive) kinetics of biochemical reactions [65].As we have shown, the polymers in highly elastic states exhibit a broad range of molecular relaxations, which are also observed in proteins. Stochastic dynamicsmaybeconsideredasageneralfeatureofproteins as physical

258

Przhonska

systems, while functionally important motions are a part of this dynamics that is determined by the unique construction of a particular protein and selected for performing a particular function. Some of these motions are the dynamics of localized substates. Thus relaxations in this system may be very similar to molecular relaxations in highly elastic polymers. There is a strong indication as anadequatemodelforunderstandthat highlyelasticpolymersmayserve ing the behavior in proteins. A detailed comparison should be made of protein dynamics at physiological conditions with the microscopic behavior of highly elastic polymeric matrices. Third, the ideas and methods developed for investigation of the different phenomena in biological systems enrich polymer research. The most advanced are time-resolved methods including time-resolved anisotropy measurements. It may be supposed that synthetic-natural polymeric composites will be developed in concluandinvestigatedformedicalapplications in thefuture.Therefore, sion, it is necessary to emphasize that only joint efforts of teams of physicists, chemists,biologists,andengineerscanlead to furtherprogress in this fastgrowing field.

REFERENCES 1. K Ono, Y Okada, S Yokotsuka, T Sasaki, M Yamamoto. Chain dynamics of styrene polymers studied by the fluorescence depolarization method. Macromolecules 27: 6482-6486,1994. 2. RARaja,BBRaju, TS Varadarajan. Application of fluorescence probe technique for determination of glass-transition temperature of polymers: Studies in polyacrylonitrile. J Appl Polym Sci 542327-830, 1994. 3. TDZ Atvars, E Sabadini, SM Martins-Franchetti. Study of secondary relaxations of polyethylene by photoluminescence technique. Eur Polym J 29:1259-1264, 1993. 4. S Gangopadhyay, MW Pleil, WL Borst. Fluorescence decay kinetics of “polyester yellow” in solutions and in polymers. J Lumin 4639-369, 1990. 5. J Guillet. Polymer Photophysics and Photochemistry. NewYork:CambridgeUniv Press,1985. 6. CD Eisenbach, RE Sah, G Baur. Characterlstic feature of the matrix effect on the Stokes shift of fluorescent dye molecules in pure and plasticized polymers. J Appl Polym Sci 28:1819-1827, 1983. 7. 0 Przhonska, Yu Slominsky, U Stahl, S Daehne. Excited-state relaxation processes in polymethine dye molecules in polymeric media. J Lumin 69: 105-1 13, 1996. 8. KY Low, RO Loutfy. Spectroscopy of dyes in polymer matrices: Dual fluorescence of a [p-(dialkylamino)benzylidene]malononitriledye in poly(viny1 acetate) matrices. Macromolecules 14:587-59 I , 1981. 9. PP Sorokin, JR Lankard. Stimulated emission observed from an organic dye, chloroaluminum phthalocyanine. IBM J ResDev10:162-163,1966.

Applications in Polymers IO. 11.

12. 13.

14.

IS.

16. 17. 18.

19. 20. 21.

22. 23. 24. 25. 26. 27 28.

29. 30.

259

Fp Schafer, W Schmidt, J Volze. Organic dye solution laser. Appl Phys Lett 9:306309,1966. BH Soffer, BB McFarland. Continuously tunable, narrow band organic dye lasers. ApplPhysLett10:266-267,1967. OE Peterson,BBSnavely.Stimulatedemissionfromflash-lampexcitedorganic dyes and polymethyl methacrylate. Appl Phys Lett 12:238-240. 1968. YuV Naboikin, LA Ogurtzova.APPodgornii.Spectralandenergycharacteristics of organiclasers in polymersandtoluene.OptSpectrosk28:974-985,1970 (in Russian). JM Drake, EMTam. RI Morse. The use of light converters to increase the power of flashlamp-pumped dye lasers. IEEE J Quantum Electron 8:92-94, 1972. R Ulrich, HP Weber. Solution-depositedthin films as passive and active light-guides. Appl Opt I I :428434, 1972. S Reich, G Neumann. Photobleaching of Rhodamine 6G in polyacrylonitile matrix. Appl Phys Lett 25: 1 19-1 2 1, 1974. JR Onstott. Short cavity dye laser excited by an electron beam-pumped semiconductor laser. Appl Phys Lett 31:818-820, 1977. U Itoh, M Takakusa. T Moriya. S Saito. Optical gain of coumarin dye-doped thin f i l m laser. J ApplPhys16:1059-1060,1977. S Sriram, HE Jackson, JT Boyd. Distributed-feedback dye laser integrated with a channel waveguide formed on silicon. Appl Phys Lett 36:721-723, 1980. RM O’Connell, TT Saito. Plastics for high-power laser applications: A review. Opt Eng22:393-399,1983. V Bezrodny. 0 Przhonskaya. E Tikhonov, M Bondar, M Shpak. Polymer active and passivelaserelementsutilizingorganicdyes.Sov J QuantumElectron 12: 16021608,1982. V Bezrodny, M Bondar. G Kozak, 0 Przhonskaya, E Tikhonov. Dye-activated polymeric media for frequency-tunable dye lasers. J Appl Spectrosc 50:44 “44, 1989. JL Nogues, WV Moreshead. Porous gel-silica, a matrix for opticallyactivecompounds. J Non-CrystSolids121:136-142,1990. D Levy. Sol-gel glassesfor optics and electro-optics.J Non-Cryst Solids 147/148:508517,1992. B Dunn, JI Zing. Optical propertiesof sol-gel glasses doped with organic molecules. JMaterChem1:903-913.1991. WHu,HYe,C Jiang. F Zhou. All-solid-state tunable DCM dye laser pumped by adiode-pumpedNd:YAGlaser.ApplOpt 36579-583, 1997. H Schmidt. Organic modification of glass structure: New glasses or new polymers? J Non-Cryst Solids 1 12:419423, 1989. MD Rahn, TA King. Comparison of laser performance of dye molecules in sol-gel, Polycom, Ormosil. and poly(methy1 methacrylate) host media. Appl Opt 34:8260827 1, 1995. DH Titterton. Performance of plastic dye lasers in the UK. Proceedings SPIE. San Jose, CA, 1997. AITolmachev,YuLSlominskii.AAIshchenko.Newcyaninedyesabsorbing in the NIR region. Proceedings of NATO Advanced Research Workshop, Trest, Czech Republic,1997.

260 31.

32.

33.

34.

35.

36. 37. 38.

39. 40. 41.

42. 43. 44. 45. 46. 47.

48.

49.

Przhonska 0 Przhonska.Correlationbetweenmolecularstructureandspectralpropertiesof

polymethine dyes for dye lasers. PhD Dissertation, Institute of Physics Academy of Sciences, Kiev, Ukraine, 1979 (in Russian). of Chemical ADKachkovski.Polymethinedyes.In:Kirk-OthmerEncyclopedia Technology, 4th edition, Vol 19. NewYork:Wiley,1996,pp 1004-1030. SA Soper, QL Mattingly. Steady-state and picosecond laser fluorescence studies of nonradiative pathways in tricarbocyanine dyes: Implications to the design of near-IR fluorochromes with high fluorescence efficiencies. J Am Chem SOC1 163744-3752, 1994. of fluorescence PF Aramendia, RM Negri,ESRoman.Temperaturedependence and photoisomerization in symmetric carbocyanines. Influence of medium viscosity and molecular structure. J Phys Chem 98:3165-3173, 1994. ADKachkovski.Personalcommunication,1997. AP Demchenko. Ultraviolet Spectra of Protein. Berlin: Springer-Verlag, 1986, pp 145172. AKLivesey, JC Brochon. Analyzing the distribution of decay constants in pulsefluorimetry using the maximum entropy method. Biophys J 52:693-706, 1987. M Vincent, J Gallay, AP Demchenko. Solvent relaxation around the excited state of indole: Analysis of fluorescence lifetime distributions and time-dependence spectral shifts. J Phys Chem 99:14931-14941. 1995. JR Lackowicz.TopicsinFluorescenceSpectroscopy, Vol 2. NewYork:Plenum Press, 199 1 . M Bondar, 0 Przhonskaya. E Tikhonov. N Fedotkina. Effect of polymeric matrix on dye laser efficiency. J Appl Spectrosc 52:352-357, 1990. K Gorot, G Kozak, A Marinchenko. M Bondar, 0 Przhonskaya, E Tikhonov. Characteristics of the kinetics of photoreactions of polymethine dyes with radical polymerization. J ApplSpectrosc49:1014-1020.1988. A Dubois, M Canva, A Brun, F Chaput, JP Boilot. Photostability of dye molecules trapped in solid matrices. Appl Opt 533193-3199, 1996. A Dubois, M Canva, A Brun, F Chaput. JP Boilot. Enhanced photostability of dye molecules trapped in solid xerogel matrices. Synth Metals 81:305-308, 1996. M Bondar, 0 Przhonska, E Tikhonov. Photobleaching of laser dyes in polymeric media. Exp Tech Phys 38:103-107. 1990. TG Pavlopoulos, JH Boyer. K Thangaraj, G Sathyamoorthi, MP Shah, ML Soong. Laser dye spectroscopy of some pyrromethenc-BF2 complexes. Appl Opt 3 I :70897094,1992. REHermes,THAlik, S Chandra, JA Hutchinson.High-efficiencypyrromethene doped solid-state dye lasers. Appl Phys Lett 63:877-879, 1993. MV Bondar, OV Przhonska.Spectral-luminescenceandlasingproperties of the pyrromethine dye PM-567 in ethanol and in a polymer matrix. Quantum Electron 28:753-756,1998. F Salin, G Le Saux, P Georges, A Brun, C Bagnall, J Zarzycki. Efficient tunable solid-state laser near 630 nm using sulforhodamine 640-doped silica gel. Opt Lett 1 :785-787, 1989. M Bondar, 0 Przhonska, E Tikhonov. Simple solid state polymeric dye laser for scientific research and biomedical applications. Proc SPIE 2380:330-335, 1995.

Applications in Polymers SO.

51.

52.

53.

54. 55.

56.

57.

58.

59. 60.

61. 62. 63.

64, 65

261

F Mikhailenko, 0 Morejko, 0 Przhonskaya, E Tikhonov. Fluorescence and lasing of protolytically unstable polymethine dye molecules. Kvantovaya Elektron 7572576, 1980 (in Russian). BMUzhinov.Energyandprotontransfer dyelasers.Proceedings ofNATOAdvanced Research Workshop, Trest. Czech Republic, 1997. 0 PrzhonMBondar, N Derevyanko, G Dyadyusha,MZubarovsky,AIschenko, skaya, Yu Slominsky,ASmirnova.ETikhonov,ATolmachev.Lightgeneration in thenear 1R regionwiththeuseofunsymmetricalpolymethinedyesolutions. Kvantovaya Elektron I I :46247 I , 1984 (in Russian). A Mandl. A Zavriyev, DE Klimck. Energy beam scaling and beam quality studies J Quantunl Electron 52:1723-1726. of a zigzag solid-state plastic dye laser. IEEE 1996. A Dodabalapur, EA Chandross, M Berggren, RE Slusher. Organic solid-state lasers: Past and future. Science 277:1787-1788. 1997. PA Miles. Bottleneck optical limiters: The optimal use of excited-state absorbers. ApplOpt3316965-6979,1994. I-YS Lee,X-LWu,PVBedworth, C-T Chen, DNg,SR JWPerry,KMansour, Marder, P Miles, T Wada. M Tian, H Sasabe. Organic optical limiter with a strong nonlinear absorptive response. Science 273: 1533-1536. 1996. JWPerry.Organicandmetal-containingreversesaturableabsorbersforoptical limiters. In: HS Nalwa. S Miyata, eds. Nonlinear Optics of Organic Molecules and Polymers. New York: CRC Press, 1997, Chap. 13, pp 813-840. EWVan Stryland, DJ Hagan,TXia, AA Said.Applicationofnonlinearoptics S Miyata, eds. Nonlinear optics of topassiveopticallimiting.In:HSNalwaand Organic Molecules and Polymers. New York: CRC Press, 1997, Chap. 14, pp 841860. A Kost, L Tutt, MB Klein, TK Dougherty, WE Elias. Optical limiting with 6"C in polymethyl methacrylate. Opt Lett 1834-336, 1993. GS He, JD Bhawalkar. CF Zhao. PN Prasad. Optical limiting effect in a two-photon absorption dye doped solid matrix. Appl Phys Lett 67:2433-2435, 1995. PFuqua,SRMarder,BDunn.JWPerry.Solidstateopticallimitingmaterials basedonphthalocyaninecontainingpolymersandorganically-modifiedsol-gels. Proc SPIE 2143:239-250, 1994. OV Przhonska. MV Bondar, EA Tikhonov. Nonlinear light absorption by liquid and solid solutions of organic dyes. Proc SPIE 2143:289-297, 1994. OV Przhonska, MV Bondar, YuL Slominsky. JH Lim, DJ Hagan, EW Van Stryland. in liquidandsolidmedia.JOpt Nonlinearlightabsorptionofpolymethinedyes Soc Am B 15:802-809,1998. AP Demchenko. Personal communication, 1997. AP Demchenko. Protein fluorescence dynamics and functions: Exploration of analogy between electronically-excited and biocatalytic transition state. Biochim Biophys Acta 1209: I4 1- 164, 1994.

This Page Intentionally Left Blank

10 Beyond Biotechnology and into Popular Technology Masaru Matsuoka Kyoto Women’s Universitx Kyoto, Japan

1.

INTRODUCTION

The origin of near-infrared (NIR) absorbing dyes for information recording sysof thegallium-arsenicsemiconductorlaser(diode temswasthedevelopment laser) that emits laser light at 780-840 nm. A diode laser is a very small, cheap, and convenient light source for electronic and photonic devices. It can be said that the new laser technology has developed a new dye chemistry. I reviewed these new fields of dye chemistry in 1990 [ I ] . In 1992, a full survey of nearinfrared absorbing dyes was published [ 2 ] that visualized the correlation between the absorption properties and chemical structures of NIR dyes. On the other hand, NIR dyes can be efficiently designed by using computerchemistry,particularly in conjunctionwiththesemiempiricalmolecular orbital (MO) method. The Pariser-Parr-Pople molecular orbital (PPP MO) [3,4] calculation run on a personal computer is currently used conveniently to predict the h,,,, value of dye chromophores. It can be also applied to produce a bathochromic shift of h,,, to the near-infrared region. The MO method contributed very much to the development of new NIR chromophores for various application fields. Many characteristics of NIR dyes for popular technologies were practically evaluated on spin-coated film, for which many new methodologies have been developed to characterize dye materials in the solid state. New dye media for future technologies were clearly required. These are characterized by their absorption and fluorescence, high solubility in polymer matrices, lightfastness,

263

Matsuoka

264

and durability against laser light as well as conventional light, high absorptivity, high reflection index, and so on. The NIR dye media were first applied in the optical data storage field, particularly in the direct read after write (DRAW) disk. The ideaof the DRAW disk was developed for the first time by Philips employees in 1977 (see Ref. 5). The practical use of organic NIR dyes for optical data storage was commercialized from Taiyo-Yuden as compact disk recordable (CD-R) in 1988. World demand for the CD-R disk increased quickly from 1993 and was anticipated to reach the number of 1.8 billion disks in 1999. The technology for the CD-R disk brought about many developments in the chemistry and physics of dye materials that will be applied in a wide variety of applications for NIR dyes. Some applications of NIR dyes for popular technology are summarized in Table I . On the other hand, the development of new charge generation materials (CGMs) for organic photoconductors (OPCs) in laser printing systems was also very important. NIR dyes, such as phthalocyanines and naphthalocyanines, are evaluated in terms of their crystal morphology and molecular aggregation.

Table 1 ApplicationsofNIRDyesforPopularTechnologies

ource

Light ~~~

~~

~

Diode laser, Thermal decomposition Optical disk, CD-R, optical card 780-840 nm Photoconductivity; OPC; laser printer; laser plate making charge generation Direct plate making (photoengraving) Photosensitivity Dye diffusion thermal transfer (D2T2) Thermal energy transfer Transparent bar code; forgeryReflection index preventive agent Heat-shielding absorption materialHeatSunlight (thermal light) Agricultural film, heat-retaining fiber Heat retention NIR absorption Sunglasses, goggles Photovoltaic devices Sun light absorption Halogen absorption NIR lamp, Electronic camera; automatic (800-1 100 photographic nm) exposure meter; NIR LED cutoff filter for PDP Forgery-preventive agent; dye laser; Fluorescence probe Photoinitiation Photoresist, photosensitizer Photosensitization IR photography LED = light-enuttlng diode; PDP = plasma display panel; OPC = organic photoconductor.

Beyond Biotechnology

265

Miscellaneous topics and applicationsof NIR dyes, such as bar code andor security systems, leuco-NIR dyes as color formers, color filter film for plasma display panels (PDPs), NIR light-induced heat-absorbing materials, and sensitizers for photovoltaic cells are also used to characterize the dye structures in terms of their functionalities. Almost all of thefunctionalities of dyematerialsforelectronicsand IT--71 interaction of dye chrophotonics were derived from the intermolecular mophores in aggregates, and their solid-state chemistry and physics becamevery important to knowledge of their functionality in correlation with their chemical structures. Molecular and material design of functional dyes will be discussed from the point of view of computer chemistry using molecular orbital (MO), molecularmechanics (MM), andmoleculardynamics(MD)methods.Computational chemistry has been considered a time-consuming tool for synthetic chemists, but it is a very convenient tool to interpret and simulate experimental results. Great advances in technology and the decreases in the costs of personal computers (PCs) and work stationsas well as all kinds of computational systems including software such as computer graphics and calculation packages made possible their use for the design of functional dye materials. In thischapter,themolecularandmaterialdesignofNIRdyesusing computational chemistry are discussed in Section 11. In Section 111, the effect of molecularaggregationontheabsorptionspectraofdyechromophores is correlated with their chemical structurein aggregates or crystals with reference to intermolecular IT-IT interactions. In Section IV, the practical applications of NIR dyes in popular technologies are exemplified. In the final section, applications of NIR dyes in new technologies are reviewed.

II. COMPUTER-AIDED MOLECULARANDMATERIAL DESIGN OF FUNCTIONAL DYES Functional dyes are currently used as key materials for electronic and optoelectronic devices. Various functionalities such as infrared absorption, fluorescence, pleochroism, nonlinear optical (NLO) properties, chromic properties, conductivity, photoconductivity, and electroluminescence are required for functional dye materials. In the molecular design of dye chromophores, these properties should be quantitatively correlated with their chemical structures by using molecular orbital (MO), molecular mechanics (MM), and molecular dynamics (MD) calculation methods. Functional dye materials constitute a new category of dyes, and their synthetic design should be based on the new ideas and methodologies discussed in this section. Many dyes have been traditionally used as coloring matter for polymer substrates such as textiles and plastics, whereas in the electronics and photonics

266

Matsuoka

fields, dyes are used as key materials that absorb light selectively and efficiently. Laser light such as that produced by a diode laser or light-emitting diode (LED) can be used as a powerful, highly monochromatic, collimated, and coherent light source for a variety of applications in science and technology. to apply dye The h,,,a, values of dye media must be predicted precisely materialsfor artificial lightsourcessuch as thediodelaserandhelium-neon laser (632.8 nm).Greatadvances in thequantitativeprediction of absorption spectra of dyechromophoresareattributable to thedevelopmentofthePPP MO method developed in 1953 [3,4]. The MO theory can be applied to design new dye chromophores in terms of predicting color properties such as A,,,,, E , A l p , dipole moment, and transition moment. The PPP MO calculation method analyzes chromophoric systems of dyes, and the absorption spectra can be evaluated quantitatively. The molecular structure of any dye is now accessible via absorption spectra by using the PPP MO method.* On the other hand, a more precise prediction of characteristics in molecular orbital calculation can be gained by an ab initio MO method, whichcan be used to optimize the chemical structure and reactivity of dye molecules. The molecular mechanics method predicts the optimized chemical structure, which is visualized by the three-dimensional computer graphics technique. These results are very valuable to predict the steric requirements of molecules that affect theirmolecularstacking.Ontheotherhand,moleculardynamicsapproaches to evaluating the molecular packing in an aggregate or the solid state became in a single crystal available as a practically useful method. Molecular stacking that has been analyzed by X-ray crystal analysis can now be determined with the so-called MDCP (molecular dynamics crystal packing) method [6,7]. These advancedmethodologiesforthepracticaldesign of NIR dyesarediscussed in Section 111. Molecular design ofnew dye chromophores can be performed by using semiempirical and ab initio molecular orbital (MO) calculation methods.Theoptimizedmolecularstructureandconformations canbesimulated bythe MM method.Characteristics at themolecular level suchasabsorption and fluorescence spectra, transition moment, dipole moment, and molecular hyperpolarizabilitycanbewellevaluated by computerchemistry.Ontheother thin films hand, material design of dye aggregates such as single crystals and is very difficult, becausetheevaluation of intermolecularinteractions of dye chromophores is not yet obvious. The MD method can be used to reproduce the molecular stacking of some dye chromophores in single crystals, but it is much tootime-consuming.Manyfunctionalities of dyematerialssuchasnonlinear *PPP MO software run on PC is availablefrom Dr. R. Nacf, 1M Budler 6, CH-4419Lupsingcn. Switzerland and/or T.Moschny. Hallc University, Institute of O g m c Chemistry. Geusaer Strasse. D-Oh2 17 Mcrseburg. Germany.

Beyond

267

optical (NLO) susceptibility, electroluminescence, conductivity, and photoconductivity depend mainly on their intermolecularr-x interactions. Many methods are already known to assemble or to orient dye molecules in particular forms that produce special and useful functionalities. These are the Langmuir-Blodgett (LB) membranes, bilayer membranes, single crystals, vapor-deposited thin films, and spin-coated thin film, but no one knows how to correlate their functionalities with the chemical structure of the dye chromophore. The reason is that we cannot quantitatively evaluate the intermolecular r-r interactions of molecules in thesolidstate.Theproposedcomputer-aidedmethodsformolecularand material design of functional dyes are visualized in Figure 1, and the establishment of this flowchart is the main objective in developing new functional dye materials. New functional dye chemistry includes traditional dye chemistry, computer it is necessary to establish chemistry, and material science and technology, and cooperative research among these fields. It is quite difficult for us to cover all three fields of synthetic chemistry, theoretical chemistry, and material science, andcooperativeresearch is vital to thedevelopment of newfunctionaldye of at least some chemistry.Syntheticdyechemistsshouldbeknowledgeable parts of computer chemistry and material science.

Molecular design (Molecular level) (MO, MM)

Material design (Aggregatekrystal)

Molecular stacking (MD)

a

n

n

Absorption spectra(NIR) Fluorescence Transition moment Dipole moment Molecular hyperpolarizability Solid-state absorptiodfluorescence Conductivity Photoconductivity Nonlinear susceptibility Electroluminescence LB membrane Single crystal Vapor-deposited thin film Spin-coated thin film

Establishment of functionality-structure relationship Figure 1 Molecular and material design offunctional dyes.

Matsuoka

268

111.

EFFECT OF MOLECULARAGGREGATION ON THE ABSORPTION SPECTRA OF DYE CHROMOPHORES

Dye materials for high technology fields are generally applied in the dispersed state in a polymer matrix or spin-coated or vapor-deposited thin film in a multilayered device structure. Dye molecules exist as aggregates or microcrystals in as intermolecular hydrothese conditions, and intermolecular interactions such gen bonding and intermolecular IT-IT interactions of dye molecules play a great role in the spectral and reflectance properties of dye media.In the case of optical recording media, practical light absorption and reflectance are evaluated in the A,,,, region in the solid state of the spin-coated film; the solid-state absorption in solution are no spectra then become important and the absorption properties longer important. Solid-state absorption spectra are very important with respect tothe NIR absorption of dye chromophores. Some relationships between the structure and the aggregation properties of dye chromophores are exemplified in the following. Intermolecularinteractions of dyemolecules in thesolidstatecanbe evaluated by thedifferences in h,,, fromsolution(molecularlevel)tosolid state(aggregate). The differences ( A h , nm) of somequinone-typedyesare summarized in Table 2. These dyes have strong intramolecular charge transfer chromophoric systems, but the Ah values are affected largely by the substituent, ringsize,andchainlength of thealkylgroups.For example,dye l a has a symmetricalstructure, but dye l b does not. Theyhavesimilar n-conjugation systems, and their ,A values in solutions are almost the same, but those on vapor-deposited thin film are quite different, from 641 nm for l a to 534 nm for lb. Consequently, the Ah value of l a is 100 nm, but that of l b is zero. These in theirintermolecular differences in Ah valuearecausedbythedifferences IT-n interactions. The X-ray crystal analysis of dyes l a and l b reveal that l a shows planar molecular stacking with four-centered intermolecular hydrogen bonding and has but l b shows some strong interlayer x-IT interactions at a distance of 3.3 bent structure in molecular packing caused by zigzag intermolecular hydrogen bonding as indicated in Figure 2. The interlayer distances of l b are 3.4-4.4 indicating fewer n-IT interactions in dyelb. As a result, the third-order nonlinear opticalsusceptibility of l a is 500 times that of l b [8]. In a series of tetrathiabenzoquinones (2), ring size (2a, 2b) and steric hindrance (2a, 2d) affect the Ak values [9]. Dyes 2a-2d generally have a planar structure in their n-conjugation system but have some distorted structure in the aliphatic ring system. In the case of dye 2a, the calculated distance of deviation from the planar n-system was 0.47 8, by the ab initio calculation method 1101. On the other hand, dye 2b hasaseven-memberaliphaticringthat is greatly distorted from the x-conjugation system. The structure of dye 2d is similar to

A,

A,

269

Beyond Biotechnology Table 2 Solid-state Absorptionand A h Values of DyeChromophores

’ b;X=H

la lb 2a 2b 2c 2d 3a

54 1 534 429 545

5 17 43 I 660

64 1 534 538 503 587 460 685

I00 0

I09 49 70 29 25

3b 3c 3d 3e 4a 4b 4c

672 668 670 664 560 5 10 520

716 770 760 753 682 628 644

44 102 90 89 122 118 124

lb (3.4 A 14.4 A) X%G

Figure 2 Correlationbetweeninterlayerdistance

x (3) values.

= 10” esu

of l a and l b in crystalsandtheir

270

Matsuoka

that of dye 2a, butthemethyl group occupies a relativelylargespace,which preventshighdensitymolecularpackingof 2d. Theseobservationswereobtained from the results of molecular structure optimization of dyes 2a, 2b, and 2d by using MOPAC PM3 (MM) calculations. The Ah value of 2b is half that of 2a and that of 2d is one-third that of 2a. At any rate, we could get a large bathochromic shift of the h,,, value in the solid state by intermolecular x-n interactions. The X-ray crystal analysis of 2a was performed, and strong interlayer n--71 interactions over a distance of 3.6 A were observed [9]. Each molecule aligns in the same plane and overlaps perpendicular to the quinone ring due to the n-n interactions and electrostatic dipole repulsion of the carbonyl groups (Fig. 3). In these cases, no intermolecular hydrogen bonding was attributed to their molecular stacking. The intermolecular n-x interactions of dyes 3a-3e were affected largely by the length of alkyl substituents (R) at the 4-position of the anilino groups. The Ah values increased with the length of the alkyl group from 3a to 3c,and at least a C4 alkyl group was necessary for an effective n-n interaction to give a large Ah value. The dichroic ratio indicates the orientation of dye molecules on vapor-deposited thin film, and that of dye 3d was affected largely by the surface character of the substrate. If the surface was previously covered by an oriented polymer film such aspolytetrafluoroethylene, dye 3d was stacked to give a highly orientedthinfilmthatshowedahigherdichroicratiothan that onuntreated glass. These observations to correlate large Ah values with intermolecular n-n interactions were also confirmed by similar substituent effects observed in the cases of dyes 4 a 4 c . All dyes (4a-4~)havingalkylgroupslongerthana C4 chainshowed large Ah values. The absorption spectra of 4a showed large differences from solution to the solid state. The first band ( A I ) produced a large bathochromic

Figure 3 (a) Perpendicular molecular overlap and (b) molecular stacking of dye 2a in crystal.

nergy

Beyond

shift ( A h = 122 nm), butthe second band (h2) did not shift so much. These results indicated that 4a was oriented systematically by strong intermolecular n-n interactions on the vapor-deposited thin film [l I]. As aresult,moleculardesign to producealargebathochromicshift of h,,, to the NIR region is very important from the point of view of molecular stacking, and MO design plays a great role in the development of new NIR dyes. But, at the same time, computer simulations to evaluate the optimized structure of moleculesandmolecularstackingbyusingthe MD methodarealsovery effective for the material design of dye chromophores.

IV. APPLICATIONS OF NIRDYES IN POPULAR TECHNOLOGY A.

Dyes for an information Recording System

Recent trends in the chemistry and applications of functional dye materials for an information recording system are summarized in this section. Dyes for high density optical recording media and full color hard copy systems are the most important applications in this field. There are many practical application methods for dye chromophores in combination with their special functionalities and required energy for information recording systems. These relationships among functionality, information as energy, and applications are summarizedin Tables 3 and 4.

1. NIR Dyes for Optical Recording Systems The color-structure relationship is the most important factor for the molecular design of NIR dyes. These dyes do not have any color in principle but comprise a very new category of dyes, and their synthetic design should be based on the new ideas and methodology. The absorption spectra of NIR dyes must be predicted correctly in order to apply dye materials for diode lasers, which

Table 3 Dyes for HighDensityOpticalRecordingSystem ~~

~~

Functionality NIR dye Photochromic materials Multilayered media Monolayered Changeable-wavelength media PH laser PHB: photochemical hole burning.

Diode laser Laser Laser/multiple wavelength

DRAW, CD-R Erasable recording Multiple-wavelength recording

B

272

Table 4

Matsuoka

Dyes for Full Color Hard Copy System

application Informatiodenergy Practical Functionality Pressure/MC developed Acid Lightkid

VLlpressurelMC

Acid Photopolymerizatlon/ acid Radical developer Silver halide developer

Heathherma1 head VL/MC/pressure

uv VL

Diazo developer HeaVthermal headlUV VL Photoconductivity Diode laser Photoconductivity DZT2 head HeaUthennal Charge-piezo control Charge

Pressure-sensitive color fornler (non-carbon paper) Photo-pressure-sensitive color former (color printer) Heat-sensitive recording (facsimile) Photopolymerized photography (Cycolor) Radical photography (color printer) Silver halide photography (video printer) Thermal printer (color printer) Xerography (color printer) Laser printer Heat-sensitive color printer Ink jet color printer

MC, microencapsule; VL, vlsible light; UV, ultraviolet light.

emit single-wavelength laser light at 780-840 nm. The PPP MO method could beused todesignNIRdyes; it analyzeschromophoricsystems of dyes,and then the substituent effect on the absorption spectra can be evaluated quantitatively. Some sets of PPP MO programs that can be run on a personal computer are now available for the design of dye chromophores. (See footnote in Section 11.) They are set up automatically by including structure drawing and parameter setting, and then parts of desirable results such as energy levels of the frontier orbitals, rr-electron densities, and their changes accompanying the first excitation can be printed. Several minutes are required for the calculation of a medium-size dye molecule. The applications of the PPP MO method for dye chromophoreshavebeensummarized by Griffiths [12] andFabianandHartmann [ 131. Tokita et al. published a book in Japanese [ 141, which summarizes the parameters for calculations and deals with practical examples of the design of NIR and some other dyes. In the case of the molecular design of indonaphtholtype NIR dye, for example, the x-electron density changes that accompany the first transition of the parent chromophore 5 are shown in Figure 4. The results indicate the intramolecular charge transfer (CT) character of the chromophoric system in dye 5; that is, the aniline moiety acts as a donor, and the naphthoquinoneiminemoietyactsas anacceptor.Fromtheseresults,substitution of an acceptor at the 2- and/or3-positionsorsubstitution of a carbonylgroup by a much stronger acceptor such as a dicyanomethylene group causes a large

273

Beyond Biotechnology

+ 0.05 + 0.11

+ 0.02 - 0.07

+ 0.12

.5fNONMe2 -

\ /

5. I,, 583 nm

0.01

- 0.19 + 0.08

- 0.02 + 0.03

Figure 4 X-Electron density changes accompanying the first excitation

of dye 5.

bathochromicshift of A,,;,,. The practicalexamples of alargebathochromic shift were observed in dyes 6 (776 nm) and 7 (722 nm), which have the same chromophoric system.

r

Me.

6 I,, 776 nm The other NIR dye chromophores can be designed in the same way from thepoint of view of theirabsorptionproperties.The E values of dyechromophores can bealsocalculated by the PPP MO methodasthe f' value of oscillator strength. The solubility of dyes, which is an important factor for their application by the wet-coating method, is described in connection with the length of alkyl substituents and their molecular stacking in the following section.

2.

NIR Dyes for CD-R and DRAW

An optical recording system allows the user to record and reproduce information into electrical signals and then into optically moduthat can then be converted lated signals. The user can record the information onto the optical memory disk, and it can be optically reproduced by reading the difference of the reflectance on light irradiation. The optical disk drive and its recording/reproducing system aredesigned to recordandreadtheinformationbylightirradiationontothe recording medium [ 151. Opticaldiskscanbeclassified into twotypes:theerasabletype,which allows repeated overwriting of the information, and the recordable (CD-R) or direct read after write (DRAW) type, which allows the user to record the information only once onto the recording medium. By recording principles, optical disks can be classified into two functional modes: the heat mode. which causes

Matsuoka

274

the structural alternation or decomposition of the medium by heat upon absorption of light such as in the cases of CD-R and DRAW media, and the photon mode, which causes reversible structural changes by means of light energy as in the case of photochromic media. Many types of NIR dyes have been developed and evaluated for CD-R and DRAW recording media. The general requirements for dye media are

1. Strongabsorptionataround 800 nm 2. High reflectance, over 15-30%, at 800 nm for reproducing reading 3. High recording sensitivity; large E , low thermal conductivity, and large optical difference after writing

4. Lightandthermalstabilityagainstreproducingreadinglight 5. Durability in storage,nontoxicity 6. Homogeneityofspin-coatedrecordinglayer The other practical requirements for dye media are also summarized in Ref. 15. Synthetic design and the structural modification of dyes to shift A,, into the NIR region have been studied. In general, a thin layer ofdye medium absorbs at wavelengths 30-50 nm longer than in a solution of the same dye, and it shows broader absorption spectra in the solid state. The absorption spectra of 190 NIR dye chromophores in solution and on thin film are summarized as a data book [ 161. Cyanine NIR dye with dithiol nickel complex as a counteranion (dye 8)

wasthe first example that wasusedas a practical dye medium for a DRAW system. Cyanine dyes generally have poor light stability, but the light stability was very much improved by using a singlet oxygen quencher (nickel complex) as a counteranion. The absorption spectra, reflectance curve, and transmission spectra of dye 8 on spin coating thin film are exemplified in Figure 5. Some phthalocyanine metal complexes [ 171 absorb in the NIR region and havebeenevaluatedas dyemedia.Buttheyhavepoorsolubility i n organic solvents, and their sensitivity is insufficientforpracticaluse.Introductions of phenylthiogroupsintothephthalocyaninenucleusproducedabathochromic shift of A,, into the NIR region and greatly improved their solubility i n or-

275

Beyond Biotechnology

400

500

600 800 700

900

1000

Wavelength (nm) Figure 5 Absorption,reflection,andtransmissionspectra dye 8 on thin film.

of cyanine-metalcomplex

ganic solvents due to the steric hindrance of the substituents. The intermolecular interaction to form aggregates is sterically restricted, and the solubility is improved.Otherexamples to improvesolubility by sterichindrance of substituents are well known in naphthalocyanine chromophores. Naphthalocyanine [I71 absorbs in the NIR region that is produced by the annulation in benzene ringsofphthalocyanine.Introduction of longalkylgroups orbranchedalkyl of naphthalocyanine improved its solubility groups into the naphthalene rings in organic solvents. Silicone naphthalocyanine, which has two substituents from the central silicone to the upper and lower sides from the n-plane, prevented their aggregation by steric hindrance of long alkoxyalkyl substituents as shown in dye 9. Dye 9 has enough solubility for the spin coating process to prepare

IT-IT

9 hmax 800 nm

Matsuoka

276

thin film for recording media. Pyrazinophthalocyanines with many and/or long alkyl groups also have good solubility in nonpolar solvents (see Sec. V.B). Indonaphthol metal complex NIR dyes 6, which absorb in the NIR region and have large E values corresponding to those of cyanine dyes, are also candidates for dye media [IS]. Dicyanomethylene indonaphthols 7 that have good characteristics for dye media are also known [ 181. The practical application of these NIR dyes for various optical recording systemsis being studied extensively for future technology.

Dyes for ErasableOpticalRecording

3.

Dyes that show photochromic properties due to bond alternation or tautomerism are important for use in erasable optical recording media. In these systems two or three different wavelength laser lights are used for recording, erasing, and reproducingtheinformation.Spiropyransandfulgidesarewellknownphotochromiccompounds, buttheircoloredforms do not absorb in the NIR region and their recycle durability is not enough for practical use. Spirothiopyran 10 is colorless but absorbs at 700 nm afterUVirradiation to givecyanine dye 11, whichreverts to dye 10 on exposure to visiblelight,buttherecycle durability of thissystem is poor [ 191: Diarylethenederivatives 12 have superiordurability for recycleuse ( lo4 times) and are irreversible by heat energy [20]. Butthesecompounds do nothavestrongabsorption in theNIR region,andalargebathochromicshiftofthecolored form of 12 is strongly anticipated(Scheme 1). ErasableCD-typeopticalrecordingsystemsarenow available,butorganicphotochromicdyemediaarenotusedastherecording media.

Me

Me

uv -

Me

ZY -

Me

10 -

Y

uv VL

12

Scheme 1

Y

X = 0,S,Se, NR; Y = aryl, alkyl

11

Y

277

Beyond Biotechnology

4.

Dyes for MultipleWavelengthRecording

Attemptsarecontinuallymadetoincreasethememory of opticalrecording media, and the development of a shorter wavelength laser that emits green to bluelightandtheimprovement of recordingheproducingsystems and disktrack shapes are generally studied. On the other hand, multilayered dye media composed of differentJ-aggregates of cyaninedyes havebeenproposedas multiple-wavelength recording systems [21]. The J-aggregates of cyanine dyes havevery sharpabsorptioncurveswithalargeabsorptioncoefficient. If J aggregates of LB filmwithdifferent hmax values areduplicated to produce multiple layers of dye film, we can get a total of several times as much memory for recording from each layer by using a wavelength-changeable laser as the light source. Reversible cycles of spiropyran produced by photochromism and thermal processes to give different colored forms including J-aggregates are shown in Figure 6 [21]. The h,,, value and the performance of the LB film of spiropyTans are affected by the substituents. The technology of multiwavelength optical recordingsystems is nowunderinvestigation,buttherearemanyfactorsto improve the characteristics of dye media [21]. In the process, heat, and photon modes, the recording, erasing, and reproduction of information become possible. Other highcapacityopticalrecordingsystemssuchasphotochemical hole burning (PHB) are known. Some types of tautomeric dye media such as quinizarin, naphthazarin and metal-free phthalocyanines are being investigated at very low temperatures, but there are many technological problems and it may be a long time before their practical application is possible.

6. Dyes for a Full Color Hard Copy System Informationrecordingisveryimportanttechnology,andmanytypes of full color hard copy systems are available in practical use. From the point of view of information recording systems for future technologies, two technologies are important for their chemistry.

% I

1

\

p a

AH

340 nm,

-

-

&H20Coc2’HU

/

NO-

AH

J-aggregates

Figure 6 Reversiblecyclesofspiropyranbyphotochromismandthermalprocessto give different colored forms including J-aggregates.

Matsuoka

278

1.

Dyes for Cycolor and Related Systems

Cyanine dyes have been used as photosensitizers in silver halide photography. Silver halides are sensitive to light from the UV to the blue visible region, and cyanine dyes adsorbed on silver halide microcrystals are used as photosensitizers inthevisible to NIR wavelength region. Cyanine dyes absorbs light and acts as an electron donor to reduce silver halide to silver metal to form the latent [22]. image. Some NIR cyanine dyes have been used for infrared photography A new application of NIR cyanine dyes with a singlet oxygen quencher asacounterionforopticalrecordingmediawasdescribed in Section 1V.A. Recently, photoreduction of cyanine borate as a photocatalyst was studied for the Cycolor system (Mead Co., Ohio) [23], which is well known for obtaining full color hardcopy in a one-shotcolorprocess. In thissystem,threetypes of the cyanine dye triphenylalkyl borate, which are sensitive to each of blue, green, and red light, are used as the photosensitizer to produce an alkyl radical that polymerizes acrylic monomer in a microcapsule. The microcapsule contains the cyanine photosensitizer, monomer, and leuco dye, which give three colors (yellow, magenta, cyan) when the capsule is broken after photoirradiation and contacts acid on the recording paper. The electron transfer reaction in cyanine borate ion pairs was studied by Schuster and coworkers [24] in 1988. The reaction was studied in connection with photopolymerization initiated by cyanine borate as photocatalyst. When the mixture of cyanine borate and acrylic monomer was irradiated with visible light in nonpolar solvent, polymerization proceeded depending on the strength of the induced light. The reaction mechanism is as follows. Irradiation of the cyanine borate 13 with green light leads to the formation of a locally excited singlet state of the dye. The singlet excited cyanine in the ion pair is due to capable of oxidizing the borate anion to the boranyl radical tRPh3B.1,

[Cy'] [RPh3B-] [Cy']*[RPh3B] [Cy.] [RPh3B.] [CY.] P . 1

hv

[Cy']*tRPh3B-IExcitation [Cy.]*[RPh3B]Electrontransfer [Cy.] [R.] [BPh3] C-B bond cleavage [CY-RI Alkylation

a i + - c ~ : & r ~Y - + ( ~ o Me

+

Et

Et

Ph3B-Bu-n 13 hmax 552 nm for green light

Scheme 2

[Cy'] = Cyanine chromophore [Cy']* = Singly excited state [RPh3B-]= Alkyltriphenyl borate

Beyond

279

the electron transfer from the borate to the cyanine cation in the excited state. to generate a free The boranyl radical undergoes carbon-boron bond cleavage alkyl radical. The radical can be used in the initiation of polymerization as in the case of the Cycolor system. On the other hand, recombination between the cyanine radical and the alkyl radical causes the formation of alkylated cyanine, which has no color. The mechanism of the reaction is summarized in Scheme 2. Schuster and coworkers studied the kineticsof this photoreduction of cyanine borate by time-resolved spectroscopy and by conventional chemical analysis, and they revealed the following results [24]. An intra-ion-pair charge recombinationelectrontransferreactionwasrenderedirreversible by therapid cleavage of the alkylcarbon-boron bond of the boranyl radical. The rate of bond of thealkylradical formed.When cleavagedependsdirectlyonthestability stabilized alkyl radicals are formed, carbon-boron bond cleavage is faster than the electron transfer back-reaction that regenerates the cyanine borate ion pair. Carbon-boron bond cleavage of the boranyl radical in the systems is irreversible, and the bond dissociation energy of the parent hydrocarbon is a good predictor of the yield of radicals. The Cycolor system is extremely beautiful technology for one-shot full color printing. But the system has some problems with respect to the stability of the microencapsulated leuco dyes in storage, and its use is not practical. [25,26] reported on a decolorizable toner system On the other hand, we using cyanine borate NIR dye (14). When the mixture of dye 14 and an excess oftriphenylalkylborate in resin is irradiatedwithNIRlight,thedarkgreen color of dye 14 disappears.Thedecoloredcompound wasidentifiedasthe nzeso-substituted dye 15. Whendye 14 is usedasadecolorizabletonerfor

oNEt2

Et2NQ

-

bC:CH-CH:CH-C Et2N

14 -

Q

Ph3B-Bu-n

NEt2

Et2NQ

QNEt2

dC:CH-CH-CH:C

-

\ / Et2N

" Q

15 R = B u - ~ Ph ,

NEt2

xerography, the printed paper can be decolored by photoirradiation and can be reused several times. to that shown The chemistry of this process can be explained as similar in Scheme 2. Photoirradiation of cyanine borate excites the dye into the singlet state,andsubsequentelectrontransferfromborate to theexciteddyegives dyeradical, alkyl or phenylradicals,andborane.Recombination of thedye radical with the alkyl or phenyl radicals gives the colorless dye 15. This type of decolorizable process using triphenylalkyl borate was found to be applicable to many types of dye chromophores.

Matsuoka

280

2.

Dyes for Xerography and Laser Printing

Organic photoconductors (OPCs) are widely used as a key material for xerography and laser printing. An OPC generally consists of two layers, the charge generation layer and the charge transportlayer. Many combinations of the charge generation material (CGM) and the charge transport material (CTM) are known (Table 5). The CGM includes bisazo dye (16), squarylium (17), phthalocyanine, andtrisazodye (18). TheCGMabsorbing in the NIR regionsuchasazulenium dye (19) and naphthalocyanine dye are known for their use as OPCs in laser printers that use a diode laser as a light source. Carbazolehydrazones, triarylamines, and pyrazolines are knownas CTMs,whichshouldhaveasmall ionization potential to carry the hole, have good charge transportation ability, and absorb in the UV region. The color toner, which consists of the three primary colors yellow, magenta, and cyan, can be used for full color printing. The color toner must have good transparency to duplicatethecolor to produce a full color print. Benzidine dye 20 for yellow, quinacridone or naphtholazo dye 21 for magenta, and phthalocyanine or triphenylmethane dye 22 for cyan color are generally used as color toners. The charge-controlling agent (cationic dye or metal complex azo dye) is generally added (1-3 wt%) to control the charge of the toner. Table 5 Some Combinations of CGM and CTM for Organic Photoconductors CGM

I

CTM

I I

Me

0 HO

17

18 -

I I

I

CGM, charge generation material; CTM, charge transport material.

281

Beyond Biotechnology

MI?&~;CH"-CH-NM~~

i-Pr

-

19

-

2

J

Ca2+

22

C. Leuco NIR Dyes The helium-neon gas laser, which emits laser light at 632.8 nm, has been commonly used to read the information written in point-of-sale (POS) labels. However, the diode laser has many advantages, including long life, small size, high speed, and low cost compared to the helium-neon gas laser and has become much more popular for industrial use. Leuco NIR dyes are currently used in thermal printing papers for such items as industrial labels, the so-called POP (point-ofproduction) label in automated factories. The development of new leuco NIR dyes becomes very important to replace conventional leuco dyes such as fluoran dyes. Fluoran derivatives 23 absorb in the full range of the visible region; the

+ X

Yellow Red

3-CI 243, 3-Me Orange 1,2-benzo Green 2-NHPh 2-NHPh,3-Me

23

Black

black fluoran dye called "one-dye black" is very important among them and is used on a large scale in thermal printing papers. Leuco NIR dyes generally have chromophoric systems similar to those of NIR dyes except for having a lactone ring or leaving group at which the bond cleavage occurs in the color-developing reaction. Examples of NIR dyes and a color-developingreactionareshown in Scheme 3. Leuco NIR dyes generally (27) chrohavephthalide (24, 25), fluoran (26), anddi-ortriphenylmethane mophoric systems, but they should be extended in n-conjugation to produce a bathochromic shift of h,,, in the NIR region. The tosyl group is also useful as a good leaving group to give the bis(styry1)methane-type cationic dye 28. Direct thermal printing systems using leuco NIR dyes as recording media are very important for convenient optical reading systems. In these systems, both

282

Matsuoka

Scheme 3

developing and fixing processes are unnecessary, and the image is produced by heat energy from a thermal head as an information recording device. This has many advantages, for example, it is a simple and maintenance-free system, and the equipment is cheaper than other recording systems such as ink jet printers. Direct thermal printing systems are widely used for facsimile and label printing (POS systems). The POS label systems are widely used for bar code labels to control the marketing of goods. Recently, the use of direct thermal printing was increased for factory automation by the introduction of the POP system, which has become very popular in the material distribution industry, where bar code labelsareused to controlmaterialdistribution in transportation.Development of the bar code reading system has attracted considerable interest in this field, and the development of many types of labels is anticipated.

y:;ph

24 A , ,750 - 830 nm,Black PhHN

ONH .

M e 2 N F N M e 2 \

’\

-

25 A, 800 , - 930 nm, Black

Me2N 0

26 A,,900 Black nm,

28 hm, 780 - 830 nm

27 Green

-

283

Beyond Biotechnology

Figure 7

Bar code label.

Figure 7 is an example of abarcodelabel. It consists ofalternating of severallinewidths. The information can be blackandwhitebars(stripes) stored by using various combinations of black and white stripes. Reproduction of information is carried out by optical scanningof the bar code using a bar code reader. The direct thermal paper consists of colorless leuco dyes and phenols as developer in a polymer binder that is coated onto the base paper as shown in Figure 8. Thermalpaper is colorless at thebeginning(Fig.8a)butgives an image(Fig.8b) if heatenergyisimpressedfromthethermalhead.The chemistry of color developing is the acid-induced cleavage of a lactone ring or ester bondofthe leuco dye to give a cationic NIR dye. It is calledinfrared (1R)-readable thermal paper for a bar code label. Good sensitivity in the NIR region,goodreliability,andhighstabilityagainstlightandheatarerequired for this thermal paper. Industrial thermal paper is generally used under much more severe conditions than the conventional thermal paper used for facsimile of leucodyesedited by printing. A bookonthechemistryandapplications Muthyala [27] comprehensively reviews the use of conventional leuco dyes for information recording systems and their applications in other fields. Conventional color developing processes (Scheme 3) are reversible, and the recorded information disappears when it comes in contact with hydrophobic materials. An irreversible color former is anticipated to retain recorded information forticketsandreceipts. A new irreversiblecolorformersystemwas developed by using the chemical reaction between isocyanate and phthalimide, Thermal head " .

Thermal head

Base paper Figure 8 Structure of thermalpaper (a) before and (b) after rwording.

284

Matsuoka

Scheme 4

whichgivesabrownish-blackcolorafterreaction(Scheme 4) [28]. An irreversible take off color former system is anticipated for copy-proof images and forgery-proof materials.

V.

APPLICATIONS OF NIRDYES IN FUTURE TECHNOLOGY

A.

Heat-Absorbing NIR Dyes*

The energy distributions of sunlight consists of 50% visible light (390-700 nm) and another 50% infrared light (700-1800 nm). Infrared light is generally called thermallightandisequivalenttoheatenergy. The temperature of materials that absorb IR light increases with irradiation time. NIR dyes absorb IR light efficientlyandcanthereforebeusedasheatshields.Practicalheat-shielding materialsconsist of polymermatricesandNIRdyes.NIRdyesdispersed in transparent polymer matrices such as polymethyl methacrylate, polycarbonate, as materials polyethylene terephthalate, and inorganic glass are generally used for windows, roofs, ceiling domes, telephone boxes, and so on. The shielding materials can be applied in plate, sheet, and film form and as paint. Agricultural film for greenhouses that contains fluorescence dyes or NIR dyes has become veryimportantforcontrollingplantgrowth.Somefluorescentdyesarealso used to change the energy distribution of sunlight. These dyes absorb light at around their klnax value and emit light at around their fluorescence maximum (Flnax), and consequently the energy distribution of the transmitted sunlight is in the favorable wavelength region for flowers, fruits, vegetables, and other plants. The characteristics required for heat-shielding NIR dyes are strong and wide absorption at around the NIR region, no or less absorption in the visible region, good affinity for polymer, and good durability against light and heat. Thereare many dyechromophores thatabsorb in theNIRregion,but phthalocyanine analogs are the best candidates because of their durability. The synthesis of polyarylaminofluorophthalocyanine(29) starting from tetrafluoro-

*See also Ref. 29.

285

Beyond Biotechnology RX or F

RX or F

F

F F or XR

RX

29

PC:phthalocyanine, X = 0, S, NH, R: aryl, alkyl, n = number of RX

Figure 9 Synthesis of polysubstitutedtluorophthalocyanines.

phthalonitrile has been developed as shown in Figure 9. Ocataanilinooctafluorovanadylphthalocyanine, absorbed at 884 nm on thin film, is soluble in organic solvents such as toluene (20 wt%) and 2-butanone (25 wt%) and stable above 320°C.Manyderivatives havebeensynthesized,including OPC (RX = SPh, n = 8, M = Zn, A,,, 750 and 830 nm), CD-R media (RX = 2,6-disubstituted phenoxy, I I = 4, M = VO, A,, 725 nm), and optical recording media (RX = NHPh, n = 4, M = SnC12, A,, 828 nm). There are many other candidates for NIR dyes that can be used as heat-absorbing dyes.

B. Pyrazinophthalocyanine NIR Dyes with Red Fluorescence Many chromophores for functional dye materialshave been developed and evaluated,butphthalocyanine is thebest chromophoreowing to itsstabilityand so on. durabilityforuseasOPCs,NLOs,information-recordingmedia,and We have studied the syntheses of functional dye materials based on dicyanopyrazine chromophores and evaluated their characteristics such as absorption and fluorescence properties. Newpyrazinophthalocyanineshavebeensynthesizedfromsubstituted andor annulateddicyanopyrazines,andtheircharacterizationssuchasfluorescencepropertiesandmolecularaggregationsboth in solutionand in the solid state have been correlated with their chemical structures. 2,3-Dichloro-5,6dicyanopyrazine 30 is a valuable intermediate in the synthesis of a wide variety of pyrazine derivatives (30-321. Diaminomaleonitrile 31 is a well-known raw material for the synthesis of 2,3-disubstituted 5,6-dicyanopyrazines 32 and related dyes [33,34]. Reaction of 31 with 3,4-hexanedione gave 32a, and a similar reaction of 31 with4-alkylphenylglyoxalgave 32b. Pyrazinoindoles 33 were synthesized by the reaction of 30 with enamines derived from carbonyl compound with alkylamine, followed by the ring-closure reaction of the amino group

Matsuoka

286

with the second chlorine of 30. Alkyl groups of any length can be introduced at the R', R2, and R3 positions of 33. The N-aryl analogs 34 could be synthesized from 30 with arylamine followed by further reaction with 1,3-dicarbonyl compounds. Their furano analogs 35 were obtained by similar reactions of 30 with 1,3-dicarbonyl compounds. On the other hand, the reaction of 31 with 1,4-dibromobutane-2,3-dione gave 2,3-bis(bromomethyl)-5,6-dicyanopyrazine,whichwasconverted to the Wittig reagent by treatment with triphenylphosphine and then reacted with 1,2dicarbonyl compound in the presence of sodium hydride in dimethyl formamide (DMF) to give 2,3-dicyano-6,7-disubstitutedquinoxalines 36. These annulation derivatives of dicyanopyrazines were also synthesized in the cases of pyrazinoindoles 33. Reaction of 30 with cyclohexylimine gave 37, which could be converted to 38 by aromatization with N-bromosuccinimide. Treatment of 37 with bromine in ethanol afforded ethoxy derivatives, which were similarly converted to 39. The absorption and fluorescence properties of pyrazinophthalocyanines 40

CIN NC

11

CIN NC

NH2

x

NH2 NC NC

xx

N NC NC

X N Y

32

synthesized from the corresponding dicyanopyrazines 32-39 [35,36] are summarized in Table 6. Phthalocyanines are well known to show crystal morphology induced by various intermolecular 7r-17 interactions. Some of the pyrazinophthalocyanines have many interesting properties such as molecular aggregation, red fluorescence, and good solubility even in hexane. An aluminum complex of pyrazinophthalocyanines showed red fluorescence with a small Stokes shift that indicated high efficiency of energy transformation of the absorbed light energy to fluorescence. The fluorescencequantumyields of thesehavenotyetbeen determined.Ontheotherhand,dyeswithmanylongalkylsubstituentsshow interesting molecular aggregation depending on the polarity changes of solvent

Beyond

287

Table 6 AbsorptionandFluorescenceSpectra

Product Reagent 32a 32b 32b 33 36 38 39

40a 40b 40c 40d 40e 40f 40g

of Pyrazinophthalocyanines

Ra

Metal

(nm)

2

C8 65

AI(0H) 638 AI(OH)

c16

v(0)

636 51 647 713 840 708 702

-

c

-

7

AI(0H) 720

Ph

cu

CS CS

-

cu

Fmax (nm)

656

S.S.b

(nm)

-

-

-

-

-

"rl-Alkyl group. bStokes shift. 'R1 = C I ~R2 . = CIS. R3 = C l h . Source: Refs. 35 and 36.

and temperature. Figure I O indicates the temperature dependence of the absorption spectrum of dye 40c, which exists as a monomer at high temperatures but as an aggregate at lower temperatures. Similar changes in molecular aggregation were also observed to depend on changes of solvent polarity. Window glass containing dye 40c, for example, can be used as an optical shutter. Applications of these new NIR pyrazinophthalocyanine derivatives for functional dye materials are under investigation.

C.

NIR Dye Film for a Plasma Display Panel*

Laser light-emitting diodes (LEDs) efficiently emit laser light at 780-940 nm TV setsand 3s a light andarewidely used in remotecontrolequipmentfor sourceforshort-distanceopticalcommunications. To blocknoise fromother a filter-coveredphotoreceiverareusedfor lightsources,pulseemissionand these electronic goods. But the emission of NIR light from a light source such as a plasma display should be perfectly protected to exclude the noise. On the other hand, the plasma display panel (PDP) will be used in the future to replace of a TV setbecausealarge TV display theCRT(cathoderaytube)panel such as 40 in. is difficult to achieve using a CRT and a liquid crystal display. A plasma display emits NIR light and electromagnetic waves, which must be absorbed by the front panel of PDP. Otherwise NIR light and electromagnetic waves generated by the PDP produce noise and disturb the control unit of other electronic products. Silver mesh or multilayered silver film and I T 0 glass are used to absorb the electromagnetic waves, and appropriate NIR dyes must be

*See also Ref. 37.

288

Matsuoka

55°C

400

500 600 700 Wavelength (nm)

800

Figure 10 Temperature dependence of the absorption spectra of 40c.

developed to absorb the IR light emitted by a PDP. These dyes must have the following optical properties:

I.

2.

Effective shielding against 780-940 nm laser light, which is commonly used as the light source for optical communication Little or no absorption and average transmittance in the visible region of 400-700 nm

Aminium dye 41 and bisammonium dye 42 are currently being evaluated as candidates for NIR dyes. Optical properties of a filter for PDP composed of NIR dye and vapor-deposited silver film are shown in Figure 11. These include good transparency in the region of 400-800 nm and absorption above 800 nm.

289

Beyond Biotechnology

100

20 0 400

600

800

1000

1200

Wavelength (nrn)

Figure 11 Opticalproperties of the front panelfilter for PDP.

D.

Dyes for Photovoltaic Cells

Molecular photovoltaic systems using organic dye materials for solar light harvesting and conversion to electricity have been anticipated for a long time. The inorganic counterpart consisting of a silicon-based thin film have already been used practically. Many types of organic photovoltaic cells such as the Schottky barrier cell consisting of an AVorganic dye/Ag(Au) structure have been examined, but their efficiencies are generally very low in comparison with those of theirinorganiccounterparts [38]. Graetzel [39] reported on ahighly efficient nanocrystalline photovoltaic device that was based on the spectral sensitization of a nanocrystalline semiconductor film by transition metal complexes. Carboxylated polypyridyl complexes of ruthenium and osmium gave extraordinary efficiencies of the conversion of incident photons into electric current, exceeding 90% withinthewavelength range of theirabsorptionband.Ruthenium-based sensitizerssuchasRuL2(NCS)2 43 attachedtothenanocrystallinetitanium dioxide filmshowedawiderangephotocurrentactionspectrumfrom 400 to YOOH

NCS

I

COOH

290

700 nm. New sensitizers that can cover the full range are strongly anticipated for the photovoltaic cell.

Matsuoka

of visible to NIR regions

REFERENCES 1. M Matsuoka. Infrared Absorbing Dyes. New York: Plenum Press, 1990, pp 1-220. 2. J Fabian,HNakazumi,MMatsuoka.Near-infraredabsorbingdyes.ChemRev 92:1197-1226,1992. 3. R Pariser, RG Parr. A semi-empirical theory of the electronic spectra and electronic structure of complex unsaturated molecules. J Chem Phys 21 :466-47 1, 1953. in unsaturatedhydrocarbons.TransFaradaySoc 4. JA Pople.Electroninteraction 4911375-1385,1953. 5 . R Pepperl, I Sander. Opt Acta 24:427, 1977. 6. N Tajima, T Tanaka, T Arikawa, T Sakurai, S Teramae, T Hirano. Japan Chemistry Program Exchange, Program 064. 7. NTajima,TArikawa,TSakurai, S Teramae,THirano.Aheuristicmoleculardynamics approach for the prediction of a molecular crystal structure. Bull Chem Soc Jpn 68:5 19-527, 1995. 8. M Matsuoka, A Oshida, A Mizoguchi, Y Hattori, A Nishimura. Molecular design of quinoid dyes for 3rd order NLO materials. Nonlinear Opt 10:109-1 14, 1995. 9. K Takagi, A Mizuno, A Iwamoto, M Furusyo, M Matsuoka. Spectral properties of tetrathiabenzoquinones and their self-assembly on solid state. Dyes Pigm 36:35-43, 1998. 10. M Furusyo, N Tajima, T Hirano, M Matsuoka. Self-assembling of tetrathiabenzoquinone; X-ray structure and molecular dynamics simulation. Unpublished. 1 I . JHKim,MMatsuoka,KFukunishi.Synthesesandsolidstateabsorptionspectra of aminonaphthoquinone dyes. Dyes Pigm 3 1 :263-272, 1996. 12. J Griffiths. Colour and Constitution of Organic Molecules. London: Academic Press, 1978. 13. JFabian, H Hartmann.LightAbsorption of OrganicColorants.Berlin:SpringerVerlag.1980,pp1-245. 14. S Tokita, M Matsuoka, Y Kogo, H Kihara. Molecular Design of Functional Dyes. Tokyo: Maruzen, 1989, pp 1-225 (in Japanese). 15. F Matsui. Optical recording systems. In: M Matsuoka,ed. Infrared Absorbing Dyes. New York: Plenum Press, 1990, pp I 17-140. 16. M Matsuoka. Absorption Spectra of Dyes for Diode Lasers. Tokyo: Bunshin, 1990, pp1-215. 17. M Matsuoka. Phthalocyanine and naphthalocyanine dyes. In: M Matsuoka, ed. Infrared Absorbing Dyes. New York: Plenum Press. 1990, pp 45-55. 18. K Yoshida. Syntheses and characteristics of new quinoid compounds as functional dyes. J Jpn Soc Col Material 61338-345, 1988 (in Japanese). ed. InfraredAbsorbingDyes.New 19. JSeto.Photochromicdyes.In:MMatsuoka, York: Plenum Press, 1990, pp 71-88. SOC,ed. Chemistry of Photochromic Com20. M Irie. Diarylethene. In: Japan Chem pounds. Tokyo: Japan Chem Soc, 1996, pp 89-109 (in Japanese).

Beyond

291

21. J Hibino. Multi-layered optical recording system using J-aggregates of photochromic dyes. Preprint, JOEM (Japanese Research Association for Organic Electronics Materials) Workshop '91, Atami, 1991, pp 48-56 (in Japanese). ed. Infrared Absorbing 22. TTani. Y Mihara.Infraredphotography.In:MMatsuoka, Dyes. New York: Plenum Press, 1990, pp 183-192. 23. JS Amey. Kinetic and Mechanical Descriptions of the Microencapsulated Acrylate Imaging Process. J Imaging Sci 33:l-6, 1989. X Yang, GB Schuster. 24. S Chatterjee, PD Davis, P Gottschalk, ME Kurz, B Sauerwein, Photochemistryofcarbocyaninealkyltriphenylboratesalts:Intra-ion-pairelectron transferandthechemistryofboranylradicals.JAmChemSocI12:6329-6338, 1990. 25. M Matsuoka, T Hikida. K Murobushi, Y Hosoda. Photoreduction of cyanine borate infrared absorbing dyes. J Cheln Soc Chem Comnmn 1993:299-300, 1993. 26. M Matsuoka, T Hikida. Photochemistry and applications of cyanine borate IR dyes. Mol Cryst Liq Cryst 227309-315, 1993. York: Plenum 27. R Muthyala, ed. ChemistryandApplicationsofLeucoDyes.New Press,1997. 28. K Kabashima, S Hiraishi, M Tanaka, M Matsuoka. New heat-sensitive irreversible colour former derived from aromatic isocyanates and imino compounds. Dyes Pigm 23:3342, 1993. 29. 0 Kaieda. Near-infrared heat absorbing dye and its applications. Preprint, Technological Applications of NIR Light, Kinouseishikiso Bukai Rep No. 37, Tokushima, 1997, pp 14-20 (in Japanese). 2,3-dichloro-.5,6-dicyanopyrazinewithamines. 30. DHou,MMatsuoka.Reactionof DyesPigm22:57-68,1993. 31. D Hou, A Oshida, M Matsuoka. Reaction of 2,3-dichloro-5,6-dicyanopyrazinewith enamines and some tertiary amines. J Heterocycl Chem 30:1571-1575, 1993. 32. JYJaung,MMatsuoka,KFukunishi.Synthesesandcharacteristicsof2,3,7,8tetracyano-l,4,6.9-tetraaza-S.10-dihydrophenazines. J Heterocycl Chem 34:653-657, 1997. 33. J Y Jaung, M Matsuoka. K Fukunishi. Syntheses and properties of new styryl dyes derived from 2,3-dicyano-S-methylpyrazines. Dyes Pigm 31: 141-153, 1996. 34. Y hung, M Matsuoka, K Fukunishi. Syntheses and characterization of new styryl fluorescent dyes from DAMN, Part 11. Dyes Pigm 34:255-266, 1997. 35. JY Jaung, M Matsuoka, K Fukunishi. Syntheses and characterization of push-pull tetrapyrazino[2.3-b]indoloporphyrazines. Synthesis, 1998, pp 1347-135 I . 36. JY Jaung. M Matsuoka, K Fukunishi.Synthesesandspectralpropertiesof new dicyanopyrazine-relatedheterocyclesfromdiaminoomaleonitrile. J ChemRes (S) 1998~284-285:1998: (M) 1998:1301-1323,1998. 37. KSakurai.Near-infraredlightabsorbingfilm.Preprint,TechnologicalApplicationsofNIRLight,KinouseishikisoBukalRep37.Tokushima,1997,pp21-23 (in Japanese). 38. Y Shlrota. Polymer materials for photovoltaic cell. In: K Ichimura, ed. Applications of Functional Polymers for Electronics and Photonics. Tokyo: CMC Press, 1988, pp 98-109 (in Japanese). 39. M Graelzel. Highly efficient nanocrystalline photovoltaic devices. Platinum Metals Rev38:151-159.1994.

This Page Intentionally Left Blank

11 Fundamentals of Near-Infrared Spectroscopy Howard Mark Mark Electronics, Suffern, New York

1.

INTRODUCTION

Modern near-infrared spectroscopy (NIR) is a method of performing chemical analysis via spectroscopic measurements. When we speak of “chemical analyor sis,’’ however,wemustaskthequestion:Arewetalkingaboutqualitative quantitative analysis? The flip answer is “yes,” since NIR analysis can be used for both. However, the application to quantitative analysis is far more common, and this chapter is devoted to that aspect of the technology. Thespectral region that wedefine asthe“nearinfrared”extendsfrom the red edge of the visible portion of the electromagnetic spectrum, at roughly 730-750nm, to 2500 nm. This region is sometimes further divided into two to changes in boththe subranges, 750-1 100 nmand 1100-2500nm,owing 1100 nm; these are fundamental physics and the instrumentation that occur at discussed in detail in Sections 1II.A and 1V.A. The advantages of this form of analysis is that in general it is fast (usually less than a minute from the time a sample is placed in the instrument until an answer is available), convenient, and multicomponent (the concentrations of two or more constituents can be determined from a single spectrum as easily as the concentration of one can); uses no extraneous chemicals so requires no disposal of those chemicals; requires little or no sample preparation (often samples can be used “as is”); and is nondestructive (although sample preparation, when required, may be destructive). Modern NIR spectroscopy differs from the classical use of “spectroscopy” in that it is performedbyusingfairlycomplicatedandsophisticatedmathe-

293

294

Mark

maticalmethods to relatethespectralreadingstaken in the NIR portion of the electromagnetic spectrum to the properties of the samples in a completely empirical manner. The concept is that when an instrument leaves the factory, it is a “dumb” assemblage of aluminum, plastic, glass, silicon, and other materials, and before it can do anythinguseful it needs to be “trained’ or calibrated. The calibration process is described in more detail in Section 1II.B but basically requires the user to collect a set of suitable samples, measure their spectra, determine their composition (or at least the concentrations of those constituents that are of analytical interest) by some other, reference, method, and then perform the calculations required to relate the spectroscopic measurements to the reference values of concentration. Because of the resource-intensive nature of the calibration process, this form of analysis is appropriate mainly to situations in which the routine use of the technology will involve the analysis of large numbers of similar samples over relativelylongtimeperiods, so thattheresourcesspentperformingthe calibration exercise can be justified by the time and money saved during routine analysis.

II. HISTORICALASPECTS The near-infrared spectral region was first discovered by Herschel, in his now classic experiment. Basically,Herschelwastrying to find out which colors of the spectrum contained the heat that is felt in sunlight, so he separated them with a prism and placed a black-painted thermometer in the various parts of the spectrum. He found not only that the temperature increased as the thermometer was moved toward the red end of the spectrum, but lo and behold it increased even more when the thermometer was moved beyond the red end of the range intotherangewherevisiblecolorscouldnolongerbeseen 11-31. This was the first indication that what we now call electromagnetic radiation existed that is sensitive could not be detected with the unaided eye. Since the human eye to approximately 730 nm,andwhatHerschelfoundwas“belowthered,”we now call thosewavelengthslongerthan730nm“infrared.”Aswewillsee, the infrared region is subdivided into several subregions based on the chemical, physical, and instrumental characteristics exhibited in the various regions. Almost 100 years were to pass before other scientists were able to obtain what we now recognize as spectrain this region. In the early part of the twentieth century, however, a few scientists-notably Coblentz, Ellis, and somewhat later, Willis, Kaye, and Whetsel-were active in systematically measuring the spectra of homologousseriesandotherinterestingmoleculesintheinfraredspectral

Fundamentals of NIR Spectroscopy

295

region [4-6]. A brief bibliographic review of this early “classical” work in nearinfrared spectroscopy is available 171. In general, however, this spectral region lay fallow for a long time. The vast majority of scientists who took an interest in using wavelengths longer than the visible region (we can define “visible” literally: those wavelengths to which the unaided human eye is sensitive) went directly to the spectral region we now characterize as “mid-infrared” in order to take advantage of the high absorptivities, narrow spectral bands, and highly characteristic absorption frequencies that many molecules, particularly organic molecules, exhibit there. This neglect of the near-infrared region is somewhat understandable, since it was caused to a large extent by the limitations of the equipment availableat the time. To the average scientist in the laboratory, a local digital computer, needed tocollect,manipulate,display, andanalyzedata,wasessentiallynonexistent. Therefore theyfellbackonthe tried andtrueanalogtechniquesofplotting spectraandanalyzingtheplots by eye,aidedperhaps by suchrelativelylow tech tools as rulers and planimeters. Under these conditions, every advantage that could be obtained from the useof the underlying physical characteristics of the molecules themselves was critical in allowing the scientist to extract useful information from the data. Matters remained thus until the mid to late-1960s. It was then that Karl Norris of the U.S. Department of Agriculture put together all the pieces of what we nowrecognizeas“modern”near-infraredspectroscopy-low-noisespectrometers,computerizedcontrol of thespectrometeralongwithcomputerized data acquisition, and the use of multivariate ~nathematical/statistical computer algorithms to analyze the data (what we now call “chemometrics”). Add to this list the use of diffuse reflectance measurements. This last concept, which gave the analyst the ability to obtain useful measurements from natural products and other otherwise intractable samples, was the last piece of the puzzle needed to make the technology a viable whole, in an analytical regime where the commercial need was great enough to explode the technology out of the laboratory and into widespread practical use and to provide an economic impetus for rapid and widespread development. Since this initial work was done at the USDA, the initial application of it was to those problems of interest to the USDA: measurement of the composition of agricultural materials. As Karl Norris tells the story (and this has been checked by Karl for factual accuracy), he had been investigating the utility of the near infrared for analysis when he was approached by Hayward Hunt, another USDA scientist, to see if he could measure moisture in wheat. Looking into the matter, he found that he could, but there were errors caused by an interfering material. Being a good scientist, he looked into the source of the interference and found that it was caused by the protein content of the wheat. He solved the moisture-

296

Mark

measurement problem by adding two more wavelengths to correct for the protein interference. He then tried to apply this same scheme to measuring the moisture content of soybeans, but in that case he found that the protein correction was not enough. Looking into the matter, he found that there were errors caused by of the an interfering material. Being a good scientist, he looked into the source interference and found that it was caused by the oil content of the soybeans, so he added two more wavelengths to correct for the oil. Realizing that one man’s interference is another man’s analysis, he then tried to measure the oil content of soybeans, and he succeeded. He then returned to the wheat analysis, and again realizing that since protein interfered with the moisture measurement of wheat, one man’s interference could again be another man’s analysis, he found that he could also measure the protein content of wheat this way. The rest, as they say, is history. Further research and development of the technology proceeded at such a furiousratethat,unfortunately.verylittle of it waspublished in theprimary literature, although more recent reviews summarize many of the events of those [ 101 times[8,9].Eventhepaperusuallyconsideredtheseminalpublication actually predates the explosion by several years. A later one [ 111, while closer in both chronology and topic, also misses the mark. It was the realization that the protein content of agricultural products could be measured in this way that was the spark that set off the explosion. There were actually several key concepts generated here that were critical to the development of NIR spectroscopy. One was the simple fact that quantitative chemical analysis could be done by using only spectral data; the second was the realization that it was possible to do multicomponent analysis by using one set of spectral readings and indeed even using the same set of readings for the different constituents. So the whole technology got its “kick” from this initial start by being able to measure the important components in each of two key commodities: protein, moisture, and oil in soybeans and moisture and protein in wheat. The technology was immediately commercialized by three companies that saw the value of this analytical approach. Once commercialized, new applications were generated to introduce the technology to new markets. Thus, nearinfrared analysis wasfirst used for the analysis of other raw agricultural materials (barley, corn, etc.) and then quickly spread to other raw foodstuffs (milk, fruits). From thatpoint,youwouldneedtodrawatree to describe thespread of the technology into new application areas, since in general new applications had some connection to old ones. The main branching of this tree is a twofold bifurcation: one branch, starting fromraw foodstuffs, moved the technology into processed foods of all sorts. The other branch went into the analysis of non-food agricultural products (forage, tobacco. and cotton, for example). Since cotton is used in the textile industry, that industry learned about and started using this

Fundamentals of NIR Spectroscopy

297

analytical tool; among other uses was its application to measuring the amount of lubricants, dyes, and other chemicals used during commercial spinning and weaving of the textiles. From there, it became known to the manufacturers of those chemicals. This spreading of the technology is typical of the way nearinfrared analysis has spread throughout much of modern industrial production. Currently,thenumberofdistinctapplications is seeminglyinfinite. An overview could be obtainedby perusing generalized spectroscopic journals (such as Applied Spectroscopy, Spectroscopy, and even Analytical Chemistv) as well as the one journal devoted to this field,Journal the of Near Infrared Spectroscopv [ 121. The same publisher also publishes a newsletter containing, among the other news in the field [ 131, a categorized listing of recent papers that have appeared. The newsletter also informs the interested scientist about new books, meetings, and other events of specific interest to the near-infrared community.

111.

FUNDAMENTAL PRINCIPLES

Being the result of a confluence of several disciplines, there are multiple sets of principles underlying near-infrared spectroscopy that could be labeled “fundamental.”Ofthese, the physical/spectroscopic principles and the mathematicakhemometric principles are the ones that are simultaneously of most interest so as to merit being called “funand also closest to an ab initio formulation, damental.”Forthesereasons,andbecauseofspacelimitations,welimitour discussion to those two sets of principles.

A.

Physical and Spectroscopic Principles

The physics underlying the basis of all spectroscopy is the interaction of electromagnetic (EM) radiation with a corresponding characteristic of matter. The coupling between them is due to the electric field portion of the electromagnetic wave interacting with a corresponding electric field of the molecule. To some extent,thiscanbedescribedbytheclassical(i.e.,pre-quantummechanical) of this,however, since some picture of nature. There is a limit to the extent phenomena found in nature can be described only by quantum mechanics. For others, quantum mechanics describes the actual situation more accurately. The classical approach, however, is still useful in that it provides descriptions (for those cases where it applies) that are more intuitive and therefore more easily understood. Thus we can look at the interaction between radiation and molecules from both points of view. In the classical view, an interaction can occur if the freof asystematicorcyclic quencyofthe EM radiationmatchesthefrequency change in an electrical characteristic of the molecule involved. In the quantum

Mark

298

mechanicalview,theinteractioncanoccur if theenergy of aphoton of EM radiation matches the energy difference between two states of the molecule. What is needed is an effect that causes an atom or molecule to undergo a transition between two states that have an energy difference equal to the energy of a photon of radiation. Thus the characteristic involved depends on the wavelength (or alternatively, the frequency)of the radiation, and EM radiation in different portions of the spectrum interact with matter through different mechanisms, depending on which characteristics of the molecules involved oscillate at the proper frequencies. Thus, for example, ultraviolet radiation interacts with the electron clouds surrounding each atom. Visible radiation interacts with the electron clouds surrounding molecules. Mid-infrared radiation interacts with the fundamental vibrations of moleculx nuclei, in particular those vibrations that change the dipole moment of the molecule. These interactions were all discovered and can be explained by classical physics. The near infrared, on the other hand, can be explained intuitively by classical physics, butit cannot be explained mathematically, since according to classical physics, the vibrational modes corresponding to the transitions that are active in the near infrared are “forbidden.” A rigorous mathematical explanation of these transitions had to await the advent of quantummechanics.Figure 1 illustratesthesituation.Figure l a represents the “classical” model of molecular vibrations. Essentially, this is the “ball on a spring” model; the restoring force increases linearly with the distance from the

INTERATOMIC DISTANCE Figure 1 Potential energy diagrams for (a) the ball-on-spring model and (b) the quantummechanicalmodel of molecularvibrations. (a) The parabolicenergycurveofthe simplified model leads to a simple harmonic oscillator with evenly spaced energy levels. (b) The quantum mechanical model allows for molecular dissociation at finite energies and has an energy level structure that confbrtns to that of real spectra.

Fundamentals of NIR Spectroscopy

299

equilibrium position, and therefore the energy required to displace the molecule SO thattheenergy curve from its equilibrium position increases quadratically, forms a parabola. Quantummechanicstells us thattheenergylevels ofamoleculemust be quantized. Applying quantum mechanics to the “ball-and-spring” model results in energy levels that are equally spaced, as Figure l a shows. Furthermore, thequantummechanicalanalysis of thissituationtells us thatthemolecular system can change states by only one quantum level at a time; any other transitionsare, in fact,forbidden.Theenergycorresponding to achange by one quantum level is equal to the energy of a photon of the mid-infrared spectral region, generally considered to include those wavelengths between 2.5 and SO p m (2500-50,000 nm), and therefore many molecules strongly absorb radiation in that spectral region. In reality. the energy curve of an oscillating molecule is more complicated thantheball-and-springmodelsuggestsand looks morenearly like that of Figure I b. Our knowledge of chemistry tells us that this must be so. because for a molecule to dissociate exactly means that the atoms canmove to an indefinitely real world, largedistancefromeach other. Figure IB showsthatthis,inthe happens at a finite maximum energy (when the energy curve becomes horizontal at large interatomic distances), as opposed to Figure IA, which would require infinite energy. The quantum mechanical analysis of this situation tells us that thereareanumberofsignificantdifferencesfromthesituationdepicted in Figure la. A key difference is that although the energy levels that the molecule can attain are still quantized, they are no longer equally spaced. As Figure Ib shows, the higher the energy levels are. the closer together they become. In the limit, when the moleculeis on the verge of dissociating, the energy levels become so closely spaced that they form a continuum, andso when the molecule actually dissociates it can fly apart, with the resulting pieces containing any amount of (kinetic) energy. What is not shown in the diagram is that the states that were forbidden for the ball-and-spring model are no longer absolutely forbidden. In conformance with reality, they becomepermitted in thequantummechanicalformulation, although with low probability. Thusit is possible for a moleculeto change energy by more than one quantum level. Since a change of two (or more) quantumlevels involves larger energy differences than a change of one quantum level, the higher energy photons have higher frequencies and thus shorter wavelengths than those corresponding tothe single-level (fundamental) changes. These multilevel energy changes correspond to photons in the near-infrared region, and this is the origin to a change of of near-infrared absorbance. The absorbance bands corresponding more than one energy level are sometimes called “overtone” bands. Generally, several overtones occur for each molecular vibration, with decreasing intensity for higher overtones. In fact, the “blueness” of water is due to the preferential absorption of light at the red end of the visible spectrum by high overtones of

300

Mark

the H-0-H absorbance bands-if the water is sufficiently clear to allow light to penetrate for the large distances necessary for this effect to become evident! Since in practice the absorption band of the overtones normally has become so weak by the time the wavelengths are short enough to be in the visible region, we define the “near-infrared’ spectral region as those wavelengths lying between 730 nm (the end of the visible) and 2500 nm (where the shortest wavelength fundamental absorbances are found). Also, as a practical matter, most near-infrared absorbance spectroscopy is done using the 1 100-2500 nm range. This range includes all the first overtone and some second overtone absorbance bands, which are the ones that are strong enough to be easily measurable using reasonably sized pathlengths within the samples. We discuss in Section IV those factors relating to the instruments that also contribute to this practical definition, as well as the current extension of commercial instrumentation to the 750-1 100 nm region. The fact that these changes are of low probability means that near-infrared absorbances are much weaker than mid-infrared absorbances; in general an NIR absorbance is approximately one-tenth to one-hundredth that of the corresponding mid-infrared absorbance. Furthermore, it is also possible for a molecule to change state by more than two energy levels. Changes corresponding to three, four, or more levels are possible, with continually decreasing probability, which translates into continually decreasing absorbance strengths. A side effect of the crowding together of energy levels with higher energy is that whenabsorptionofNIRradiationoccurs,theenergyneededtocause the transition between two nonadjacent states is less than twice the energy of the lowest two adjacent states. Thus, although a “zeroth-order’’ approximation tells us that NIR bands occur at one-half, one-third, etc. of the wavelength of the fundamental absorbance, the true absorbances occur at wavelengths that are displaced toward the long-wavelength side of those positions. Another phenomenon that only the quantum mechanical picture can explainabouttheway real moleculesbehave is theexistence of “combination bands.” Combination bands arise in the following manner. If a particular atom or other part of a molecule undergoes two or more vibrations (as we will see), then these vibrational modes can interact, and the molecule can absorb radiation at a wavelength corresponding to the frequency that is equal to the sum (ordifference)ofthefrequenciesoftheunderlyingfundamentalvibrations. Thenature of commonvibrationalmodes is describedbelow,butanexample of suchvibrationswouldbeastretchingvibrationandabendingvibration of the same atom. Since the frequencies corresponding to the differences of the fundamental vibration frequencies occur at lower frequencies, and thus longer wavelengths, than the fundamental themselves, they are of little interest to us here. In practice, only a limited number of molecular species are active as NIR absorbers. The basic reason for this is that only molecular vibrations involving

Fundamentals of NIR Spectroscopy

301

hydrogen atoms occur at high enough frequencies for their overtones to extend into the NIR spectral region. All other atoms, being more massive, vibrate at relatively low frequencies. Thus any overtones that might exist in the NIR are of such high order that their absorptivities are too small to be detectable. There is a small class of exceptions to this: Rare earth elements and transuranic elements have F electrons in their outer shells, with binding energies so low that changes to thenear-infraredspectral in thoseelectronenergylevelsalsocorrespond region. However, these are very much exceptions and currently of little interest to anyone not at a national laboratory. Sincehydrogen is sucha keyelement in thisregard, it is naturalthat effectively all materials that absorb in the near infrared are organic materials, because they contain large amounts of hydrogen. Thus the various absorbances we see in the NIR spectral region are due to the vibrational modes of hydrogen in the various functional groups containing this element. In organic materials, this basically limits the spectroscopically active vibrations to those of C-H, 0-H, and N-H. Since most hydrogen atoms, even in organic materials, are attached to carbonatoms, it isthe C-H absorbances that aremostprominent in NIR spectra. The vibrations that these hydrogen atoms can undergo are illustrated in Figure 2. In general, the two stretching modes are the highest frequency vibrations thatthissystemundergoes,followed by thebendingandfinallytherocking vibrations.Thusthecombinationbandsobservedarethoseof thestretching modes in conjunction with one of the bending modes.

6 . Mathematical and Chemometric Principles The rawspectrathataremeasured in thenear-infraredspectralregion (7502500 nm)areless useful in theiroriginalstate,perhaps,thanthosethatare measured in the mid-infrared. The original investigators mentioned earlier collected spectra of pure materials and found that these spectra were characteristic of the molecular structures, but the spectra also had limitations that other regions were not subject to. One key limitation is that only those functional groups that contained hydrogen were observable. Thus the vibrations involving only heavieratoms,suchascarbon-carbon,carbon-oxygen,carbon-nitrogen,peroxide, in the midand other such bonds, whose fundamental vibrations are observable infrared, are not seen in the near infrared. Add to this limitation the fact that in natural products, the original andstill of the underlying the most widespread application area, the absorbance bands components are broad, ill-defined, and completely overlapping, and you realize that simple visual observation of the spectra is not sufficient to extract useful information from them. It wastheapplicationofsophisticatedmathematicaltechniquestothe spectral data to use the spectra in ways that allow quantitative and qualitative

302

Mark

Sym Stretch

H

\/r” C Rocking inplane

Asym. Stretch

H

\J C

Scissors

C Rocking out of plane

I \/I C Figure 2

C Twist

““\

/H

C

Modes of vibrational oscillation of organic hydrogen atoms.

chemical analysis that makes NIR spectroscopy the valuable tool it has become. In this chapter we concentrate on the use of NIR spectroscopy for quantitative analysis. Although qualitative analysis through the use of NIR spectral readings can be and is performed, quantitative analysis was the original and is still the more widespread use of this technology. In one sense, some of the sophistication (and underlying complications) are hidden, because the final result (in at least some cases) can be put into the very simple-seeming form of what is sometimes considered “the near-infrared calibration equation,”

%C=bo+blAl + b 2 A z + . . . + b , , A , ,

Fundamentals of NIR Spectroscopy

303

This equation states that the concentration, C , of the analyte of interest can be calculated from the spectral data, the various A , , by multiplying these optical datavaluesbysuitable constants,thecorresponding bi. Thissimple-seeming equation hides the fact that it requires a good deal of complex mathematical theory and extensive calculations to compute the set of constants that simultaneously compensate for the interfering spectral effects of other materials in the sample, minimize the effect of noise, accommodate nonlinearities (in both the spectrochemical system and the instrumentation), and “average out” the effects of errors in the reference laboratory readings (discussed below). The mathematics used are an outgrowth of the science of statistics and are collectivelycalledchemometrics:theapplication of thesesophisticatedmathis ematical techniques to chemical analysis. The simplest of these techniques of the concept of Beer’s multiple linear regression, which is the direct application law to the problem of chemical analysis. According to this approach, the strength of the absorbance signal from the analyte of interest is proportional to its concentration at a wavelength corresponding to an absorbance peak. Therefore, a measurement at that wavelength will provide a direct measure of the concentration. The difficulty that arises is that other materials in the sample also have absorbance at that wavelength; the way to compensate for this interference is to make measurements at other wavelengths and do the necessary mathematical computations to determine how to use the measurements at these other wavelengths to correct for the effect of the interference at the primary wavelength. The generation of the necessary mathematics from this concept is described in detail in Ref. 14. For our purposes here, we simplify it, using the simulated spectral curves 3 are the two underlying shown in Figure 3. The points to note about Figure curves, representing the analyte and the interfering material, and the fact that the resultant measured spectrum is the wavelength-by-wavelength sum of these. The analyte of interest is the one represented by the narrow absorbance band on the left (absorbance peak A). The broad absorbance band (absorbance peak B) represents the interference. Since any measurement made can only measure the of the absorbance at the analyte peak alone sum of thetwo,ameasurement is insufficient, because there is no way to tell how much is due to the analyte band and how much to the interfering band. The nature of our simulated data is such, however, that the absorbance of the interfering band at the wavelength corresponding to peak A is half the peak absorbance of the interference itself (indicated as H ) . Thus, if we divide H by 2 , we arrive at the correction factor(C, which is equal to H / 2 ) , which, when subtracted from the total signal as shown, gives us the absorbance dueto the analyte band alone. Since the concentration of analyte is proportional to the absorbance, we need only multiply the net analyte absorbance by the proper scaling factor to compute the concentration.

Mark

304 1.6

-

1.4

-.

1.2

-.

1

-.

0.8 -. 0.6

-.

0.4 -.

0.2

-.

0

Figure 3 Simulated spectra showing how the effect of an interference can be subtracted from the absorbance of the analyte o f interest.

We can express this as follows. Beer's law tells us

that

%C = h / A a n a l y t e

(1)

and since the correction factor is half the absorbance of the interfering peak, Aanalyte = A t o t d -

HI2

(2)

Therefore, substituting Eq. ( 2 ) into Eq. ( l ) ,

%C = bl (Atotal - H / 2 ) and, upon multiplying this out,

%C = bl Al"t,l - ( b l / 2 )H We can express this in standard form by defining 02 as equal to b1/2:

%C = hl Atotal - bzH This short mathematical derivation is the prototype for all multiwavelength methods of spectroscopic analysis and shows how the nature of the calibration model incorporates the corrections for the interferences. When data are obtained from real-world samples, where several interferences may be present and in addition there may be noise and other sources of error, it is usually not so clear whichwavelengthsarecorrectingforwhichinterferences. In fact, it maynot actually be so clear-cut that each wavelength is correcting for only one interference or that each interference is being corrected by one particular wavelength, but underneath it all the identical processes are being carried out.

Fundamentals of NIR Spectroscopy

305

The situation becomes even more complex, and more obscure, when other methods of creating the calibration models are used. There is a whole “alphabet soup” of techniques that are used to create calibration models, with abbreviations such as PLS and PCA. These methods rely on using the whole spectrum as the basis of the calibration model rather than only a few selected wavelengths. The advantagesofthesemethodsareimprovedresistance to theeffects of noise, elimination of the possibility of missing an important wavelength’s correction capabilities, the ability to create and use orthogonal variables, and avoidance of the practical problems of finding and selecting the key wavelengths to use in the calibrationmodel.Adisadvantage of the full-spectrummethods,recentwork has shown, is that they do not compensate for nonlinearities in the spectral data as well as individual wavelengths can. The prototype basis for these full-spectrum methods is also illustrated in Figure 3 , albeit indirectly. By comparing the spectra of many samples containing varying amounts of both the analyte and the interference(s), the spectrum of the interference can be determined, and then this spectrum can be subtracted from each spectrum in the data set. The result is the spectrum of the pure analyte; this is essentiallytheinverseoperation of addingtogetherthespectra of the chemicalcomponentstoobtain the spectrum ofthemixtureshown. Thena is computed,and therelationship weightedintegral of theabsorbanceband between this integrated spectrum and the (known) concentrations of the analyte in thecalibrationsamples is determined,andthisrelationship is usedasthe calibration model. Amoredirectexample is shown in Figure 4. Here a typicalfactor is shown, compared to the spectra of the constituents it is derived from. In this simple case, the factoritself is the wavelength-by-wavelength difference between the two spectra, and this can be seen by comparing the factor with the spectra from which it is derived. The various “alphabet soup” techniques are different methods of extracting the spectra to use as the underlying factors. The technical terms for these are “basis functions” and “latent variables.” Each different method (PCA, PLS, etc.) computes a different set of basis functions; the different sets of basis functionsthatarecomputedhavedifferentmathematicalproperties.Theselatent of the chemical components of the samples. This variables can be the spectra occurs if the data are analyzed by using what is called K-matrix analysis, but otherwise it happens only rarely. It is much more common for the latent variables to be computed so as to meet more abstract mathematical criteria. Usually they are computed so as to be mutuallyorthogonal;oftentheyarecomputed SO asto bewhatareknowntechnicallyas“maximumvarianceestimators.” This latter criterion is theone that results in thePCAandPCRtechniques. Otherdefiningcriterialead to the othertechniquesforcomputingcalibration models.

0

Mark

306 1.6 14

12

w P

2

1

08

2

0.6

a

0.4

0.2 0

1

1500

1390

1100

700 2100

(a)

1900

2300

2500

2 m

2500

WAVELENGTH

1

0.9 0.8 0.7

UI

0

3 8 1700

1500

0.6 0.5 0.4

I:

1300

~

1300

(c)

,

~

1100

(b)

1500

,

~

0

00 WAVELENGTH

1100

17002500

1900 2300

2100

WAVELENGTH

Figure 4 The factor for a calibration model for a mixture is derivedfromspectraof thecomponentsofthemixture. (a) and (b) showthespectraofthetwopurecomponents, water and methanol, respectively, and (c) shows the factor, which is similar to the difference between the spectra of the two components.

Fundamentals of NIR Spectroscopy

307

IV. INSTRUMENTATION A.

General Considerations

All spectroscopic instruments, regardless of the spectral region in whichthey operate,havecertaincommonrequirements:asource of radiation. a means of determining the wavelength and the energy content of the radiation at that wavelength, and a detector. Instruments that are used for chemical analysis, such as we are concerned with here, also need to provide a means for the radiation to interact with the sample before it strikes the detector. In near-infrared spectrometers, the source is, with one exception, invariably a tungsten halogen lamp. This variation of the common incandescent lamp provides high energy and is small, rugged, and fairly stable (at least until shortly before it burnsout,butthatsourceofinstabilitycanbeavoided by routine preventive maintenance procedures). The tungsten halogen lamp, of course, has intense radiation in the visible part o f the spectrum, andthis extends well into the infrared region. These characteristics, along with their small size and ruggedness, make these lamps premier sources of radiation for near-infrared spectrometers. They are limited, however,by the decrease in their output due to the falloff of the blackbody curve at the longer wavelengths, combined with the absorbance of the quartz envelope starting at approximately 2700 nm. Since these limitations start at wavelengths that coincide with the changeover in the nature of the absorption process by organic materials (from overtones to fundamental vibrations), these form a natural limit to our defining "near infrared" to the wavelengths shorter than 2500 nm. in thedetectorsused.Historically,twotypes of Morevariationisseen detectors have been used almost exclusively (except in military and other specialized applications). The first type comprises silicon detectors, which are fast, 1100 nm and, by low-noise,andhighlysensitivefromthevisibleregionto same technology that produces virtue of their having been developed from the transistors and other semiconductors, small and inexpensive. Coupling this with instrument design that prescribes longer sample pathlengths explains the current 100 nm) introduction of instruments employing the higher overtone region (750-1 for measurements. The second common type of detector is lead sulfide (PbS). These detectors arerelativelyslowandarefinickyboth to produceandtouse,buttheyare popular because they are sensitive and provide good signal-to-noise properties from 1 100 to 2500 nm at moderatecost.Hereagain,thedetectorproperties coincide withthoseofthepopulartungstenhalogenlamps(describedabove) andtheuse of thestronger first overtoneabsorbances;these also makethe definition of 2500 nm the natural limit of "near infrared" and the use of the 1 100-2500 nm region popular.

Mark

308

More recently a third type of detector has appeared more and more commonly on the scene, one that has the speed and size characteristics of silicon and the wavelength range of PbS. This is the indium gallium arsenide (InGaAs) detector, which comes in two “flavors”: the standard InGaAs. which is sensitive to approximately 1700 n m , and the “extended range” InGaAs, which is sensitive or PbS and so to roughly 2300 nm. Both are much more expensive that the Si are used mainly in specialty applications. The interaction withthe samples is accomplished via one of two major techniques: transmission or reflection. Transmission measurements are usually made when the sample is a liquid. The liquid need not be perfectly clear. either; diffusetransmissionthroughbothliquidsandsolids is awell-acceptedmeasurement mode in this spectral region. Reflectance measurements can and are also done with highly scattering liquids, but, as betits the historical background, the vast majority of reflectance measurements are made on solid samples (most often a powder). The biggest variation in the nature of NIR instruments extant is the means used tomeasuretheradiantenergy at thedifferentwavelengths of interest, of the and so the major breakdown of instrument types deals with that aspect instrumentation.

B. Prisms We include this category of NIR instruments solely for the sake of historical completeness, since the original discovery of the near-infrared spectral region by Herschelwasdoneusingaprism,andtheearlyUV-Vis-NIRinstruments usedprisms. I, at least,am not aware of anymoderninstrument that usesa prismas its maindispersiveopticalelement. The basis of operationwasthe change in index of refraction of the prism material with wavelength (called the “dispersion”). This would cause light of different wavelengths to be refracted by different amounts and hence become spatially separated, at which point a given wavelength could be selected by placing a narrow slit in a suitable location so as to block all wavelengths except the desired one, which passed through the slit. Nowadays the spatial separation is invariably accomplished by means of a diffraction grating.

C.

DiffractionGratings

Asintimatedabove,diffractiongratingsaretheworkhorses of modernNIR spectroscopy. Although design details vary, a diffraction grating is basically a flat sheet of’ glass or other stable material, often mirrored, with a series of tine lines scratched. etched, or otherwise formed on its surface. When the lines are sufficiently fine andclosetogether,ontheorder of thewavelengthof light.

Fundamentals of NIR Spectroscopy

309

thelightleavesthesurfacewithacoherentwavefront.Lightreflectedfrom neighboring lines on the surface can then interfere constructively or destructively, depending on their phase relationships, which in turn depend also on wavelength. Thus, the net effect is similar to that of a prism, in that light (or, more important for our purposes, near-infrared radiation-we often use the word ‘‘light’’ in this way) is diffracted at an angle that depends on the wavelength. This then ‘I II ows the different wavelengths to be selected with a narrow slit, just as with a prism. The advantage of diffraction gratings over prisms is that by controlling the line spacing and width along with other parameters, the properties of the grating can be controlled in a manner not possible with prisms due to the dependency of a prism’s characteristics on the properties of the material from which it is made. Thisallowsadiffractiongrating to bemadeforvirtuallyanyspectralregion and to be optimized for the region it is to be used in, at the will of the designer. Thus, for example, a grating can be designed for a wide or narrow wavelength range and have coarse or tine spectral resolution almost independently of the range. In use, the grating is mounted so it can rotate. As it does so, the dispersed, so thattheradiation reflectedlight is sweptovera platecontainingtheslit, passing through the slit becomes related to the angle of the grating. Figure 5 illustrates how these items are related. A commonmethod of formingthelines onthegrating is to exposea light-sensitive layer (similar to that used on photographic tilm) to a holographic pattern formed by a laser; unsurprisingly, this is called a “holographic grating.” is cost: The advantage of holographicgratingsovertheconventionaldesigns It is much easier to create the required tine lines this way than by the classical method of mechanically scribing the lines onto the surface, an exquisitely delicate procedure. Diffraction gratings instruments are generally used in one of two modes, which we can characterize as “slow scan” and “rapid scan.” The difference be-

Grating

”I Mask "see Detector

Sampd

Figure 5

Instrument dcsignusing a diffraction grating.

Mark

310

tweenthesemodesiswhetherthegratingstopsmovingandstaysatagiven wavelength while the energy at that wavelength is measured or the energy measurement is made “on the fly.”

D. InterferenceFilters If two plane parallel partially reflecting surfaces are in close proximity, then the light passing through them will undergo interference effects and will be more a or less intense, depending again on wavelength. This is known formally as Fabry-Perot interferometer (hence the name “interference” filter) and is also a basis for selecting a wavelength out of the spectrum. In practice, the partially reflecting surfaces are created by depositing thin films of transparent material of alternating high and low refractive index, the index mismatch between the layers causing the reflections. The choice of materials and thicknesses of the layers determines the wavelength that such a filter will pass. Figure 6 illustrates how a filter can be interposed in the optical chain of a spectrometer. The advantageof this type of wavelength selection device is that a filter can be made relatively small and light, with a narrow passband; be stable; and have good mechanical and optical properties, all at relatively low cost. The main disadvantage is that it is a single-wavelength device; that is, a given filter can be used only at thewavelength for which it is designed and constructed. A separate filter is required for each wavelength to be used; consequently this type of device is not suited to full-spectrum data collection instruments. As an aside, we note that circular variablefilters (CVFs) and linear variable filters (LVFs), which conceivably could allow full-spectrum data collection, are used in other spectral regions but have not found application in the NIR region until recently. Another variation in the use of interference filters to scan over fairly narrow and restricted ranges of the NIR regions is the use of “tilting filters.” These depend on the fact that the wavelength of maximum transmission of an interference filter will vary with the angle of incidence of the light. Although popular

Source

Filter Sample

Figure 6 Instrumentdesignusing an interferencefilter.

Fundamentals of NIR Spectroscopy

31 1

for a while, these are no longer available, having been discontinued when true scanning instruments based on diffraction gratings became available.

E.FourierTransformSpectrometers Fourier transform spectrometers are based on the use of a Michaelson interferometer (shown in Figure 7) to create an interferngmm, a record of the intensity of radiation passing through the interferometer as one of the interferometer mirrors is moved. Sinceeach wavelengthundergoesconstructiveanddestructive interference at a different rate as the mirror is moved at constant speed, the signal produced is the Fourier transform of the spectrum. Much of the technology is similar to that used in the mid-infrared spectral region. Indeed, much of the burst in activity for the application of this technology to NIR measurement is the result of the manufacturers of Fourier transform infrared (FTIR) equipment realizing the increase in interest in the near infrared and the fact that their devices could function perfectly well in that spectral region with relatively minor changes. The advantage of the FT-NIR approach compared to, say. gratings is due primarily to the throughput (Jaquinot) advantage. This advantage, which is essentially geometrical, is almost independent of wavelength region. On the other hand, the multiplex (Fellgett) advantage does not apply in this spectral region as it does in the mid-infrared; the reason for this is that the Fellgett advantage applies mainly to instruments whose signal-to-noise performance is limited mainly by detector noise, a situation that generally does not occur in near-infrared mea-

TI

Source

/Beamsplitter

Movable Mirror Figure 7

Instrument deslgn employinga Fourier transformspectrometer.

312

Mark

surements, at least not in the common measurement situations for which instruments are normally used.

NIR

F. Hadamard TransformSpectroscopy Anothermultiplexmethod is to useadiffractiongratingalongwithamask containing not one but multiple slits. As shown in Figure 8, the detector then sees the total energy passing through all the slits. The slit patternis then changed, and another energy measurementis taken using the new slit pattern. By repeating this procedure a sufficient number of times and using a suitable set of patterns, it is possible to solve the setof measurements for the energy at each wavelength, essentially by treating the system as a set of simultaneous equations. This typeof instrument is of great theoretical interest and has the advantage that it is capable of being implemented in an instrument without any moving parts, but it is not currently in widespread commercial use.

G.

Diode Array

Another technology that is of interest because it also allows multiplexing and avoids the use of moving parts is the use of diode arrays. Basically, semiconductorfabricationtechniquesareused to create many diodedetectors(most commonly silicon detectors) on a single substrate; these detectors are arranged in a line as shown in Figure 9. This assembly is then placed at a suitable location to intercept the dispersed rays emanating from a diffraction grating. Thus, instead of passing the light through a slit through which only one wavelength at a time can be measured, all wavelengths are measured simultaneously.

Hadamard

Focussing Lens

Figure 8 Instrument based on a Hadamard spectrometer.

313

Fundamentals of NIR Spectroscopy

Grating

Sample

i

Diode Array

Figure 9 Diode array in an NIR instrument.

H. Acousto-Optical Tunable Filters The really nifty technology of acousto-optical tunable filters (AOTFs) is based on the property that some crystals have, that when excited by sound waves of sufficiently high frequency they can act as diffraction gratings. The premier type of crystal that is used for NIR instruments is tellurium dioxide, TeO?. Since the effective grating parameters depend on the frequency of the sound wave, this results in a device whose properties can be changedby changing the frequency of the acoustic wave; i.e., it is equivalent to a diffraction grating where the spacing and width of the lines can be changed. Thus the spectrum can be scanned solely by changing the electrical frequency of the exciting signal, resulting in another technologythatcanprovideaninstrumentwithnomovingparts.Figure10 shows how this can be accomplished.

1.

Light-Emitting Diodes

The purist will insist on calling light-emitting diodes (LEDs) infrared-emitting diodes (IREDs), but both types of diodes are based on the same semiconductor Mask

/

Transducer Figure 10 Instrumentemploying an acousto-opticmodulator.

Mark

314

technology. As mentioned above, LED (or IRED-based instruments) are the only type of near-infrared instrument that does not require a tungsten-halogen lamp as the source. The reason is that these diodes are used as sources and replace the incandescent type of lamps. Each diode produces a light of high intensity but relativelynarrowbandwidth compared to the incandescent type of source but still notsufficientlynarrow to beuseddirectly as the wavelength-determining device. Hence the diodes are commonly coupled to interference filters to reduce the bandwidth to ;I value considered suitably narrow. This generally implies a separate filter for each diode, and vice versa. low power The advantages of this type of' source are the small size and requirements coupled with high intensity radiation. Also, with suitable control circuits they could be cycled i n a fashion similar to the Hadarnard approach and used in a multiplex manner.

V.

MODERN USE OF NIR SPECTROSCOPY

The limitations of NIR that make it difficult to apply the classical spectrochemical methods that were historically found useful in other spectral regions make it amenable to the computerized methods that distinguish, and even define. modern NIR analysis. In particular, as we noted earlier, the near infrared is a premier spectral region in which to do quantitative analysis via spectral measurements. The general scheme is described in the flowchart of Figure I I . The basic premise of this technology is that before an instrument can be it mist first becalibrated. In NIRterminology,this usedforroutineanalysis means that the spectral readings must be empirically related to the composition of the samples that are to be analyzed. This is a multistep procedure; basically, themain steps oftheprocedureareindicated by thelabels in theboxesin Figure 11. For accurate analysis, attention must be paid to the details of each so too is an NIR analysis step; as a chain is only as strong as its weakest link, only as accurate as the care taken at each step. Although a full discussion can (and does!) occupyan entire book, we briefly discuss some of the more important salient points involved in the various steps needed to create a good calibration model per the outline given in Figure 1 1.

A.

Collecting the Samples

To calibrate the instrument, it is necessary to obtain a number of samples of the type that are to be analyzed in the future by the instrument. It is important to ensure not only that the samples are of the same type but also that they are treated in the same manneras the samples that are to be analyzed routinely in the future. For best results it is necessary to accumulate not only the correct number

315

Fundamentals of NIR Spectroscopy Analysis

Calibration

Measure with

[NIR instrument1

J/

reference laboratory

AP P lY

1

Measure with

NIR instrument

1 Ap p- Iy algorithm

)

Create calibration model

T

calibration model

ANSWER

Figure 11 Flowchart of quantitativeanalysisusing NIR spectroscopy. The twomain stcps are the calibration step and the routine analysis step.

of samples but also the correct types of samples. This is because the instrument along with the calibration model will have to do two jobs simultaneously: ( 1 ) calculate thecorrectsensitivity of thereadingstothechangingcomposition of theanalyteand (2) makethepropercorrectionfortheeffectsof all the interferences that may exist in the samples. Note that the user may not know what interferences, or even how many of them, are present. Nevertheless their effects on the spectrum must be accounted for. The goal of the sample collection procedure is to obtain a set of samples that meet two criteria: they must cover both the range of values of the analyte(s) in the samples for whichtheinstrument is tobeused in future analyses and the range of variation of interfering materials. In practice it has been found that by following certain guidelines and rules of thumb, a good selection of samples canbe obtained, eventhoughthedetails of theircompositionareunknown beforehand. Onesuchrule of thumb is: Collect 10 samples,plus 10 moresamples for each constituent for which a calibration model is desired. This results in a minimum of 20 samples, and most of the time applying this rule will result in an adequate sample set. Attention must also be paid to having the samples cover the range of values of the analyte and also the range of interferences that may be present. Since the interferences may be unknown, it is wise to make an effort to randomize the collection process so as to avoid collecting samples that are too much like ones already in hand or that are collected under conditions too similar to those at hand.

Mark

316

Sometimes circumstances make it difficult or impossible to collect a suitable set of samples at the time the instrument is first being set up.

6. Reference Laboratory Measurements

For calibration (sometimes called “training”) purposes, the concentration of the analytemustbedeterminedbysomeothermethod in orderto havevalues to relate the instrument readings to. Usually these other methods involve wet chemistry. In theearlydaysofmodernNIRtechniques,thesewetchemical laboratoryvalueswereconsidered“golden,”andanydifferencesbetweenthe instrument and the laboratory were ascribed to inaccuracies in the instrument readings. With experience, it was learned that the reference laboratory values could be in error at least as often as the instrument. Consequently, it is now common practice to send multiple aliquots of each sample to the laboratory for analysis,preferably“blinded”(i.e.,thelaboratoryshouldnotknowwhichof the samples it receives are aliquots of the sample calibration sample). Although somewhat more expensivethan sending single aliquots, sending multiple aliquots confers two major benefits: The variation in the readings of the multiple aliquots give an indication of the accuracy of the laboratory. It very quickly becomes clear that it is unreasonable to expect the instrument to agree with the laboratory any better than the laboratory can agree with itself. In addition, sometimes serious errors occur (due to blunders, misreading of equipment, etc.) in the laboratory values; having multiple readings can warn of that situation so that those values can be ignored. 2. The average of the multiple analyses of each sample is usually more accurate than a single reading. Thus, using these averages as the calibration values will improve the ultimate instrumental accuracy as well. Under these circumstances, the instrument can actually be more accurate than the reference laboratory, although that condition may be extremely difficult to demonstrate or prove. 1.

C. Obtaining Instrument Readings The main rationale for using an instrumental technique such as NIR spectroscopy is that in routine use it will be faster, easier, and less expensive and require no chemicals (with the attendant elimination of the need to dispose of those chemicals) than the laboratory procedure it is replacing. Consequently, it is common, especiallywhencollectingthedatatouseforthecalibrationcalculations,to measure each sample multiple times with the instrument as well as having the referencelaboratory do so. The manufacturer’srecommendedprocedures(for

Fundamentals of NIR Spectroscopy 31

7

wavelengthcheckingandverificationofoperationalperformance)shouldbe followed to ensure best results.

D.

Performing the Calibration Calculations

Aswesawabove,therearemanydifferentwaysthatcalibrationmodelscan be generated. Add to this the fact that instrument manufacturers and third-party software developers may implement a given algorithm in different fashions and it quickly becomes clear that very little specific advice can be given that would be valid in all cases. Thereis, however, oneprocedurethat all instrumentmanufacturersand third-partysoftwarevendorsprovideforandthatshouldbeappliedtoeach calibration model at the time it is developed: The model should be validated. This is the term applied to the concept of separating some samples for which bothreferencelaboratoryvaluesandinstrumentreadings are availablefrom the main set of calibration samples and using them to test the accuracy of the calibration model. Since those samples were not included in the calibration data, they are more nearly like the true “unknown” samples that will be analyzed in the future and those results more nearly representative of the results expected during routine analysis.

VI.

INTERPRETATION

Interpretation of NIR spectra, in the sense used in mid-infrared spectroscopy at least, is uncommon. There are several reasons for this situation. The first reason is the same as the historical reason that the near infrared was ignored for many years:Much of theinformation in anNIR spectrum is alsoavailable in the mid-IR spectrum of the compound, in a clearer and more distinct form. An NIR spectrum often lacks information about molecular structure that a mid-IR spectrum usually contains; this is due to the limitation, mentioned in Section IILA, that only molecular bonds containing hydrogen are observed in the NIR spectral region. Thus C-C, C-N, C-0, and other important non-hydrogencontinuing bonds are not seen in the NIR spectrumof a given molecule, whereas they are observed in the mid-IR spectrum. Because these modes characterize the skeletal vibrations and most of the functional groups present in molecules, this is another reason classical methods of spectral interpretation are not used in the near infrared. However, compensating for this lack is a different kind of spectral interpretation, one arising out of the very mathematical principles that give modern NIR spectroscopy its power. When the mathematical techniques are applied to NIR data, some of the results, both final and intermediate, themselves take the form

318

Mark

of spectra. These spectra reflect the underlying spectrochemical processes lying dormant in the data sets and are naturally extracted in the process of creating the calibration model. As a prototypical example of this process, we offer Figure 4, which we referred to previously, as an illustration. It shows the spectra of two pure materials, water and methanol. If a set of samples were to consist of mixtures of those two materials, then the factor that would be computed to derive the calibration model would be very similar to the difference between the spectra of the two materials.Thiseffect is shown in Figure4c,wherethefactorvalues(called “loadings”) corresponding to the water absorbance bands are pointing up in the in the factor plot, while those due to methanol, such as the absorbance bands region 2250-2310 nm, point downward. However, just as in the case of ascribing meaning to wavelength selection processes, when real spectra are used that have noise, intercorrelations, and other real-world nonideal behavior, it is difficult to ascribe meaning to the spectral displaysandrelatethem to theunderlyingchemicalandphysicalprocesses. Studies of these effects are in their infancy.

VII. ADVANTAGES,DISADVANTAGES,AND MISCELLANEOUS CONSIDERATIONS As with any analytical methodology, NIR analysis has both its strong and weak points. Tables 1-4 summarize these and suggest the situations when this technology is appropriate and when it is not appropriate. of the Tables 1 and 2 list, respectively, the advantages and disadvantages technology.Theyarebrokendownintoprimaryandderived(orsecondary) characteristics. The secondary characteristics are so labeled only insofar as they depend, in one way or another, on the primary characteristics rather than on the underlying chemical, spectroscopic, or data properties that this spectral region exhibits. The derived characteristics are deliberately not put into a one-to-one correspondence with the primary characteristics, because sometimes they depend to more than on more than one; also, a given primary property may give rise one derived characteristics. Some of these characteristics are self-explanatory; others warrant a few descriptive words. For example. it is the combination of the use of tungsten-halogen lamps, which are very high energy sources of radiation in this spectral region, with the sensitive detectors available (PbS, Si, and CdHgTe) that provides the ability to rapidly measure accurate, low noise spectra. These characteristics, in turn, allow measurements of diffuse reflectance spectra in harsh and otherwise forbidding conditions.

Fundamentals of NIR Spectroscopy

319

Table 1 Advantages of NIRAnalysis

Primary

Derived

Strong (high energy) sources Sensitive detectors Quartz/glass optics Nondestructive Low absorbance (overtoneslcombinations)

Large sample sizes Fiber optics Computer required Extensive instrumental and software infrastructure Low solvent absorbance (no “blackout” regions) Multiple-constituent measurements High scattering

High precision: high S/N Rapid analysis Measure moisture/aqueous solutions Direct sample reading (no dissolution, etc.) On-line measurement Measures inhomogeneous samples Minimal sample preparation Long pathlength; resistant to dirt and deposits Flexibility Autosampling possible Diffuse reflectance analysis

Similarly, the ability to use glass, quartz, and even sapphire as the optical materialallowsinstrumentation to be built thatcanwithstandtheunfriendly conditionsfound in factories,whichthenmakethemmoreamenabletoconsideration for in-line and on-line measurements. If conditions are too harsh for even those types of instruments, the ease of coupling the instrument to the process stream through the use of fiber optics means that the sensitive parts of the instrument can be placed in a control room or other mild environment with only the ruggedized parts of the optics exposed to the harsh environment. Another derived advantage of the use of these materials is that it makes the measurement Table 2

Disadvantagesof NIR Analysis

Primary Low absorbance (overtones/combinations)

Secondary method Poor spectral resolution High scattering Slow method development Temperature sensitivity Mainly-CH,-NH,-OH

Derived Not a micro- or trace technique Requires calibration samples, reference lab value, large sample set. calibration methodology “Classical” qualitative analysis difficult Physical variations of data

320

Mark

Table 3 When to UseNIRAnalysis I. 2. 3. 4. 5. 6. I. 8. 9. IO. 11. 12. 13.

Rapid analysis important Multiple constltuents Calibration samples available High concentrations Accurate reference lab values Method development time Natural products or organic materials Many samples to be analyzed over long period of time Unusual measurement conditions Simple or no sample preparation Automated system available Multiple sampling points Representative quality control test

of water-containing samples a relatively simple procedure; since these materials are unaffected by water, both water-containing samples and aqueous solutions can be easily measured. The generally lower absorbance of molecules in this spectral region (compared to the mid-infrared), while precluding measurements of trace quantities of material or of microsamples, has advantagesof its own. Since large sample sizes are generally de rigueur, the compensating characteristic is that the optical beam inherently samples large volumes of the material; thus sample inhomogeneities have little or no effect, since the interrogating beam sees enough of the sample to “average out” their effect optically. The requirement for an auxiliary computer is also a blessing in disguise. Since the computer is required for generating the calibration model, it is possible to createmodelsfortwoor evenmoreconstituents of asample,andwhen doing the actual analysis the concentrations of all of these components can be determined from the data of a single spectral scan. The main downside of this technology is that the user must execute the process of creating the calibration model. Taking into account the necessity to

Table 4 1. 2. 3. 4. 5.

WhenNot to UseNIRAnalysis

A single sample or onlyafewsamples are to be measured. Micro- or traceanalysisisneeded. Nogoodreferencemethodisavailable. Samplesareinorganic. Unstable chemistryhnpling situation.

Fundamentals of NIR Spectroscopy

321

collect and analyze the samples, then read them with the instrument and perform thecalibrationexercise(whichisnotalwaysstraightforward), this is afairly time- and labor-intensive procedure and generally requires a skilled operator to perform. It is thusrecommendedfor use only whenthenumberofanalyses to be performed in the future is sufficiently great to justify this expenditure of resources. The ability of the technology to deal with samples that can usually be measured “as is,” on the other hand, helps the analyst deal with the situation of having such a large sample load to measure.

REFERENCES W Herschel. Phil Trans Roy Soc Lond 90:255-283, 1800. W Herschel. Phil Trans Roy Soc Lond 90:284-293, 1800. W Herschel. Phil Trans Roy SOC Lond 90:293-329, 1800. W Kaye. Spectrochim Acta 6:257, 1954. W Kaye. Spectrochim Acta 7:181-204, 1955. K Whetsel. Appl Spectrosc 2( l ) : l , 1968. B Osborne, T Fearn. PH Hindle. Practical NIR Spectroscopy. 2nd ed. Essex, UK: Longman Scientific and Technical. 1993. 8. K Norris. NIR News 3(1):12-13- 1992. 9. KH Norris. NIR News 4(5):10-1 I , 1993. IO. KH Norris, WL Butler. IRE Trans Biomed Electron 8(3):153-157, 1961. 11. KH Norris, JR Hart. Principles Methods Measur Moisture Solids 4:19-25, 1965. 12. NIR Publications, 6 Charlton Mill, Chichester. West Sussex, PO18 OHY. UK. 13. NIK N m s . NIR Publications, 6 Charlton Mill, Chichester, West Sussex, POI8 OHY. UK. 14.HMark.PrinciplesandPractice of SpectroscopicCalibration. NewYork:Wiley, 1991. I. 2. 3. 4. 5. 6. 7.

This Page Intentionally Left Blank

Biomedical Applications of Near-Infrared Spectroscopy Emil W. Ciurczak Purdue Pharma L.i?, Ardslex New York

I. INTRODUCTION The physics of near-infrared (NIR) spectroscopy is favorable for biological applications, especially for in situ measurements. The low absorptivities inherent in NIR allow the radiation to penetrate deeper, resulting in longer pathlengths. The sources are more intense, giving more radiation to work with; detectors are more sensitive and nearly noise-free, giving a more precise and accurate spectrum. This detector and lamp combination has the advantage of being sensitive to minute differences at very low light fluxes. The physical realities allow for measurements through tissue, muscle, fat, and body fluids with great precision. Fiber-optic probes complete the picture and make NIR spectroscopy a technique adaptable to any lab or clinic. in useforroutineapplications for Although NIR spectroscopy has been several decades, it was not used for esoteric applications such as nonintrusive blood analyses until quite recently because the computing power was not available. The powerful mathematical tools, described elsewhere in this text, need the computing power and speed of modern computers to be effective. In vivo to be scanned clinical measurements cannot be made overnight; a patient has and a result generated in seconds or minutes, not hours. To this end the newest, fastest (400 MHz, 500 MHz, or faster) computer chips are required. The various applicationsin this chapter are listed by topic: blood chemistry, blood oxygen, andso forth. There may be overlaps or applications where a single application is difficult to assign. In such cases, my decision has been to place the subject in the most general application available.

323

324

Ciurczak

II. BLOODGLUCOSE One of the most publicized and pursued uses of the near infrared in the life sciences recently is for in situ glucose measurements. The number of new patients diagnosed with diabetes each year is phenomenal. With the diagnostic market estimated in themultibilliondollarrange,onecanonlyimaginethenumber of patents for blood glucose measurof workers in this field. A large number ingdeviceshavebeenissued;anexample is theonedevelopedbyHamand Cohen [ 1 I. In this device, the light is passed through the finger and is a true transmission instrument. The softwareinvolved is a neural network (NN) wherein the software recognizes the areas of greatest correlation and builds the equation itself. Because of the complexity of blood chemistry, complex algorithms are needed for the determination of any component of the blood. a t arecentAmericanDiApotentiallyimportantdevicewasdisplayed abetesAssociationmeeting.Theposter,authored by Gabrielyet al., acquired spectral data from the thumb and used it to measure clinically relevant plasma glucose[2].Scanningfrom400 to 1700nmwithafiber-opticprobe,spectra were acquired at 40 sec intervals. Although it is not a final solution, it was one of the better documented methods that use a patient’s digit for the location of the probe. Recent lawsuits on this matter resulted from undocumented clinical data on the part of the instrument manufacturer. ongoing project for Modeling of the glucose-blood system has been an numerous researchers for many years. Gary Small (Ohio University) and Mark Arnold (University of Iowa), in particular, have published widely on the subject. In 1993 131, they published a paper modeling the NIR measurement of glucose in a protein-continuing matrix. The region from 2000 to 2500 nm was used for a series of glucose solutions ranging from 1.2 to 20.0 mM in a phosphate buffer, containing such materials as bovine serum albumin (BSA) that closely resemble blood. They found that a Gaussian-shaped Fourier filter combined with a partial least squares (PLS) regression gave a reasonable standard error of calibration. In all cases, the glucose absorbance at 2270 nm gave the best correlation, with a standard error of 0.24 mM. In 1994, Arnold et al. reported work on a temperature-insensitive glucose[3], measuringmethod[4]. In anapproachsimilartotheonedetailedabove usingaFourierfilter-PLScombination.atemperaturerangeof 3 2 4 1 ° C was investigated.Thisvariationisnecessaryshouldmeasurementsbetakenfrom someone with a fever or in shock. The temperature variations caused relatively large variations in the spectra due to the water band shifts. The Fourier filtering effectively eliminated these differences. The standard error was even better than for their previous work (0.14 mM versus 0.24 mM.) The necessity of pretreatment of spectra is clearly seen in this study.

Biomedical

325

Small, Arnold, et al. [ 5 ] performed further work in 1996, this time using physiological levels of glucose in the presence of protein and triglycerides. The solutions used in the study contained 1-20 mM of glucose. The interferences were varied within each level of glucose. I t was seen that multivariate manipulation algorithms compensated for the chemical variations in the blood where the glucose level remains unchanged. The same researchers also perfonned some recently published work on using quadratic-PLS and digital-filtering techniques to account for non-glucose-related changes in the spectra [6). They concluded that in morerobust equapretreatment helped eliminate interferences and resulted tions.Theinstrumentation(eithersingle-beam or double-beam) is madeless important as interferences are accounted for in the NIR equation. in 1998 TWO interesting papers by Arnold and coworkers were published 17.81. They were devoted to the calculations used in noninvasive blood glucose measurements. The authors investigated neural networks and partial least squares and examined such things as signal-to-noise enhancements by understanding the ways in which light is attenuated as it passes through tissue. These companion articles go into detail about the scattering effectsof blood and tissue. Compensation schema are identitied and proposed to alleviate some of these interferences. It is stronglyrecommendedthatfat,water,andtissuebecompensatedfor in any algorithm or model considered. (More than a dozen salient references are listed.) Haalandetal.addedtothemodelingliteraturewitha1992paper[9]. This work used whole blood for the model. Scanning from 1500 to 2400 nm, a PLS equation was developed on glucose-spiked whole blood. The range 0.1741.3 mM yieldedan equation with a standard error of 1.8 mM.Four patients were used as models for this project. Cross-validated PLS standard errors for glucose concentration based on data obtained from all four subjects were 2.2 mM. When PLS models were developed on three patients’ blood samples and tested on the fourth, the glucose predictions were poor. The conclusion was that models must be developed for individual patients due to variability of blood chemistry. Another novel device was developed by Schrader [ l o ] in which a laser is used to illuminate the humor of the eye and the absorbance spectrum is used to measure the amount of blood glucose in the patient. The device is based on a patent developed by Backhaus et al. [ I I]. Schrader found thattheglucose levels in the anterior chamber of the eye closely follow the changes in blood glucose with a latency of approximately 20 min. The equations developed by this instrumentation allow for noninvasive monitoring of physiological glucose levels with an error of f 3 0 mg/dL. The basis of using NIR through skin and muscle is that the blood glucose level in the blood is similar, if not identical to, the glucose level in tissues. This was claimed by Fischeret al. in 1994 [ 121 and later demonstrated through a series of measurements. However, contradicting this work was a paper by Sternberg

326

Ciurczak

et al. [ 131, who claimed that tissue contained only 75% of the glucose level found in the blood at that point. Fortunately, the readings in NIR are inclusive of both blood and tissue. The calibration is based on a point of contact for each individual patient; thus, the ratio of tissue to venoudarterial blood is based on a constant. Correlation of the spectra to blood glucose readings is then acceptable. To model more correctly the in vivo realities of human body chemistry, et al. [I41 built “phantoms”were built forsimulated in vivotesting.Arnold phantoms of water, fat, and muscle tissue by reading the skin of a patient. They found that in vivo overtone spectra collected across human webbing tissue with a thickness of 6.7 mm could be simulated with a water layer thickness of 5.06.4 mm combined with a fat layer thickness of 1.4-4.2 mm. For purposes of this study, animal tissue and fat were used; there is little difference in composition between human and animal materials. They concluded that these phantom studies would help researchers develop patient-applicable methods. This “phantom” work was continued by Arnold et al. in a later publication [IS]. This is a negativelydesignedstudy,used as anobjectlesson to “warn” the inexperienced user about the pitfalls of chemometrics. They used an in vitro model of blood-simulated samples to build a model for blood glucose determination. In this case, however, they carefully omitted any glucose from the samples. The samples were randomly assigned glucose values and a PLS regression was performed. As with any PLS model, an equation was developed that gave reano glucose sonable standard errors, regression coefficients, etc. Since there was present,thisequationcouldnotpredictglucosewhensamplescontainingthe sugar were tested. This was a classic paper that showed experimenters potential wrong turns that could be taken with multivariate analyses. An interesting observation was made by Maier et al. [16], who observed that there was a correlation between blood glucose concentration and the reduced NIR scattering coefficient of tissue. Using a frequency domain NIR spectrometer, the scattering coefficient of tissue was measured with high enough precision to detectchanges in glucose. The workwasbasedonthetheorythat asthe glucose concentration increased the refractive index of the blood also increased in a predictable manner. This increased refractive index would then decrease the scattering coefficient of the blood and would be an indication of the concentration, and so forth. There is some question as to the applicability of this work to in vivo measurements, but it does demonstrate one of the novel approaches being investigated in the field. Some young researchers working at the University of Krakow have published several papers (and given numerous talks and posters) devoted to mathematical treatments of the complex spectra produced from the NIR examination of blood through skin and muscle [ 17-19]. They have been working with neural networks (NNs) in particular and have made some interesting observations. The advancement and proliferation of work in this field maybetraceddirectly to

Biomedical

327

more powerful personal computers and better partial least squares and NN programs. These complex algorithms simply couldnot be run on desktop computers a mere decade ago. One technique, recently displayed [20]. uses a fiber-optic lightpipe to meaof a finger. The device uses a portion of sure blood glucose through the skin the fiber, stripped of its cladding, as a virtual attenuated total reflectance (ATR) against the skin of the thumb. Since so much of the radiation is lost into the skin, white light is used. The monochromator is post-sample, giving better sensitivity thanwhenthewavelengths areresolved first. “Post-sample”indicates the skin and the resultant emerging light is that “white” light is impinged on collected and submitted to a monochromator. One“non-skin”applicationwaspublished by Heiseetal. 1211. In this paper,aprocedureformeasuringbloodglucosethroughthelip is described. Usingthe 1 100-1800nniwavelengthrange, partial leastsquares is themath algorithm of choice. The mean-square prediction error (PRESS’/?) is estimated as between 45 and 55 mg/dL. In this case, there is a lag time of approximately 10 minbetweenthedrawnbloodvaluesandthevaluesderivedfromthelip tissue. The authors recommend using fiber optics for further developments in this field.

111.

BLOOD OXYGEN

Themeasurement of bloodoxygenissimultaneously(relatively)simpleyet nontrivial. Early reports of the use of NIR methods for diagnostic applications camefromresearcherssuchasJobsis in 1977 [22]. HeusedNIRtechniques to monitor the degree of oxygenation of certain metabolites. Later, Ozaki et al. 1231 examined venal blood to determine the level of deoxyhemoglobin. Using aminiatureintegratingsphere.theback of thehandwasilluminatedandthe diffusely reflected light was captured by the small integrating sphere equipped withresultsfromaClBA witha PbS detector.Thespectrawerecorrelated Corning 278 blood gas analyzer. The 760 nm band in the spectrum was seen to correlate quite well with deoxyhemoglobin. A difference method or negative correlation to oxygenated hemoglobin was then built. MichaelSowaandhisgroup 1241 usedNIRimagingasanoninvasive techniquetomonitorregionalandtemporalvariations in tissueoxygenation. The purpose was to ascertain the effects of periods of restricted blood outflow (venous outflowrestriction)andinterruptedbloodinflow(ischemia). In this work, the software was the heart of the paper. Multivariate analyses of image andspectraldatatimecourseswereusedtoidentifycorrelatedspectraland regional domains. Fuzzy C-means clustering of image time courses was used to reveal finer regional heterogeneity in the response of stressed tissues.

328

Ciurczak

The wavelength region 400-1 100 nm was monitored from zero to 30 min, and a plotofthesedatawasdeveloped toproducea“topographical”representationofthephenomenon.Peaksandvalleyswereapparentwhereblood became oxygenated and deoxygenated. These standard wavelength-based values x 5 12 back-illuminated correlated well with the images developed by the 512 charge-coupled device (CCD) element. Clustering results clearly showed areas of both low and high oxygenation. These results have important implications in the assessment of the viability of transplanted tissue. Mancinietal.[25]estimatedskeletalmuscleoxygenationbyusingthe differential absorption properties of hemoglobin. Oxygenated and deoxygenated hemoglobin have identical absorptivitiesat 800 nm, whereas deoxygenated hemoglobinpredominatesat760nm.Effects of myoglobinwerealsoinvestigated, and it wasfoundthatthereadingsweredue to hemoglobin.Venousoxygen saturation and absorption between 760 and 800 nm were correlated. They researched several conclusions: ( I ) Hypoxia in KCI-arrested hearts results in only moderate activation of anaerobic glycolysis; ( 2 ) oxygenation of the epicardial and midmural LV layers is similar; and (3) a large PO? gradient exists between VS and IS in beating and arrested crystalloid-perfused hearts. An experiment wherein the influence of fat layers on the oxygenation of bloodwasexamined by Lin et al. wasreported in 1998[26]. The phantom experiments showed that fat makes a difference in patient-to-patient measurein anyindividualpatient.Yamamotoetal. ments;thesemaybecompensated [27] addressed the issue of fat interference with an oximeter that corrected for the influence of subcutaneous fat. The wavelengths, again, were the key, as was the algorithm. The effect of water on hemoglobin concentration in a tissue-like phantom was studied by Franceschini et al. in 1996 1281. Their in vitro studies consisted of aqueous suspensions containing Liposyn, bovine blood, and yeast, buffered at pH 7.2. The optical coefficients of the mixture matched thoseof biological tissue in the near infrared, and the hemoglobin concentration (23 pM)was also similar to thatfound in tissues.Theyoxygenatedanddeoxygenatedthehemoglobin by sparging the mixture with either oxygen or nitrogen. They determined that to obtainaccurateresultson waterconcentrationmustbetakenintoaccount hemoglobin concentrations. Research performed by Charles DiMarzio’s students 1291 at Northeastern University is being directed to the oxygenation problem. For thefirst application, a fiber-optic device was built to measure blood oxygen in the brain. An NIR beam is shone on the patient’s head at one point via a fiber-optic probe, and a second probe is used to collect the energy from a second point. This technique is expected to be useful in brain surgery in newborns. The other application of a noninvasive NIR probe uses it in measuring oxygen levels in the skin. This second device would be useful for patients with burns, skin ulcers, and other skin problems.

Biomedical

329

On a similar level, Jiang et al. [30] presented a device that allowed the noninvasive measurement of cerebral tissue oxygenation. Again based on fiber optics,shorterNIRwavelengthssimilartothosepresentedabovewereused. Anotherdevice to performdiffusereflectancemeasurements on theskinwas developedbyMarbachandHeise [31]. Thedevicepresentedhas anon-axis ellipsoidal collecting mirror with efficient illumination for small sampling areas of bulky body specimens. The actual schematic is too complex to describe in this chapter; anyone interestedin their work can obtaina reprint. The researchers supported the optical design with a Monte Carlo simulation study of the reflective 1600 nnipeak characteristics of skintissue.Theirworkwascenteredonthe associated with glucose (using the lip as the point of entry), butthework is applicable to other tissue research. Keiko Miyasaka presented some of his work [32] at a meeting in Toronto. As a worker in thefield of critical care for children, he introduced what he calls a Niroscope for NIR spectroscopy. His work was performed during pediatric anesthesia and intensive care. Miyasaka found that the Beer’s law relationship is not followed rigorously when the signal is passed through the cranium. Considering the massive scattering and amount of light absorbed, this is understandable. It does given, however, a semiquantitative or indicating equation. WhatMiyasakawasmeasuringwastheintercranialchromopherlevels of oxygenated hemoglobin (Hbo,), deoxygenated or reduced hemoglobin (Hb), a and cytochrome redox status. Two methods were used: photon counting and micro-type pulse laser. The photon-counting method is necessary because of the extreme attenuation of the incident radiation during traversal of the cranium. The pulse laser was used to enhance the amount of light introduced into the brain. He stated that three conclusions may be reached from the NIR data: that changes in Hboz levels reflect changes in arterial blood, that Hb changes are due to venous blood, and that total hemoglobin reflects changes in cerebral blood volume or intercranial pressure. This tool will be invaluable for emergency and operating room situations for both children and, someday, adults. Van Huffel etal. [33], in BelgiumusedNIRspectroscopy to monitor to correlate with behavioral states brain oxygenation and used the information of preterm infants and understand the development of brain hemodynamics autoregulation. The concentrations of Hbo,, Hb, and cytochrome aa3 (Cyt aaj) are used to monitortheoxygenationlevel in infantbrainblood.Some novel chemometrics were involved as well; windowed fast Fourier transform (WFFT) and wavelet analyses were employed. The purpose of the work was to find relationships between the computed chromophore concentrations and heart rate, breathing, and peripheral oxygen saturation. These researchers presented similar work in 1998 as well [34]. Cooper et al. [35] performed another study in 1998, this one aimed at the adult brain. In this work, NIR spectroscopy was used to determine the effects of

330

Ciurczak

changes in the rate of oxygen delivery on adultrat brain chemistry. Absolute levels of oxyhemoglobin, deoxyhemoglobin, and the redox state of the CUA center in mitochondrial cytochrome oxidase. An interesting finding was that although as the mean arterial blood pressure reached 100 mmHg, hemoglobin oxygenation began to fall, the oxidized CUAlevels fell only when cerebral blood volume fell autoregulationmechanismsfailed at 50 mmHg.Hemoglobinoxygenation linearly with decreases in the rate of oxygen delivery to the brain, but the oxidized CUAconcentration did not start to fall until this rate was 50% of normal. The results suggested that the brain maintained more than adequate oxygen delivery to mitochondria. Their conclusion was that NIR spectra provide a good measure of oxygen insufficiency in vivo. ArelatedstudyonhumaninfantswasperformedbyWyattetal. [36]. They used NIR to quantify the cerebral blood volume in human infants using NIR spectroscopy. Similar difficulties were encountered with the amountof light actually penetrating the cranial cavity, but useful equations were generated. Kupriyanov et al. (371 determined intracellular po? in cardiac muscle by the balance between its diffusion from vascular space (VS) to intercellular space (IS) and its uptake by mitochondria. They reasoned that cessation of mechanical work decreased 0 2 demand and should have reduced the 0 2 gradient between VS and IS. Fortheirresearch,theycomparedtheeffects of arterial / 7 0 2 o n myoglobin (Mb) oxygenation, 0 2 uptake. and lactate formation rates in beating and KCI-arrested pig hearts. Ischemia in the forearm was studied by Mansfield et al. in 1997 1381. In this study, the workers used fuzzy C-means clustering and principal component analysis of timeseriesfromthe NIR imaging of volunteers’forearms.They attempted predictions of blood depletion and increase without a priori values for calibration. For those with a mathematical bent, this paper does a very nice job describing the theory behind the PCA and fuzzy C-means algorithms. Another interesting paper was published by Wolf et al. in 1996 1391 about work i n which they used NIR and laser-Doppler flowmetry (LDF) to study the effect of systemic nitric oxide synthase (NOS) inhibition on brain oxygenation. on brainoxygenation Thestudy,performed onrats,demonstratednoeffects during cortical spreading depressions (CSD). DopplerultrasoundwascombinedwithNIRimaging in anotherstudy [40]. Leim et al. used NIR imaging and ultrasound to follow the cerebral oxygenation and hemodynamics in preterm infants treated with repeated doses of indomethacin. I n addition to the normal concentrations of oxyhemoglobin, deoxyhemoglobin, and oxidized cytochrome aa3 measured by NIR, transcutaneous po2 and pco?,arterial 0 2 saturation, and blood pressure were measured as well. Along with the cerebral blood volume, they were all used for diagnosis and research. Low oxygenation was then thought to be a possible contraindication for indomethacin treatment for preterm infants.

Biomedical

331

One interesting piece of equipment developed to measure the oxygen content of the blood was producedby the Centre for Biomedical Technologyin Australia [ 4 I]. It consists of five 1 W lasers at wavelengthsof 780, 800, 830.850, and 980 nm and uses a photodiode receiver.It uses the hemoglobin/deoxyhemoglobin absorbance differences mentioned earlier and throws in the SO2 content of the blood for good measure. The physical placement of detectors on the scalp for brain blood oxygenationwasstudied by Germon et al. in a 1998 study [42]. Detectors placed 2.7 and 5.5 cm from an NIR emitter were compared for the determination of Hhb, HbO2,oxidizedcytochromecoxidase,and total hemoglobin. The biological experimental portion of the work was to measure the chemical changes with an induced reduction of the mean decrease in middle cerebral artery blood flow. The signalchangeperunitphotonpathlengthdetected at 5.5 cm wassignificantlygreater forHhbthan for 2.7 cm. Ontheotherhand,theincrease in all chromophores detected at 5.5 cm during scalp hyperemia was significantly less than that detected at 2.7 cm. The need for more work is indicated before meaningful applications may be designed from this work. Usingsimilarinstrumentation,Henson et al. [43] determinedtheaccuracy of their cerebral oximeter under conditions of isocapnic hypoxia. Using healthy volunteers, dynamic end-tidal forcing was used to produce step changes 1 0 0 % under in PETOZ,resulting in arterialsaturationrangingfrom-70%to conditions of controlled normocapnia (resting PTETO?) or hypercapnia (resting plus 7-10 mmHg). Using standard methods, the 0 2 concentrations for each patient under each condition were determined. Excellent correlation resulted in the rsoz and[ScircIjlvbar102foreachindividualpatient:however,widevariability between patients was discovered. They concluded that under the current limitations of the equipment, the device was good for tracking trends in 0 2 but could not be used as an absolute measure for different patients. Numerous and disparate studies have been publishedor presented in recent years for the effect of various conditions on blood oxygen: Hoshi et al. 1441 investigated the neuronal activity, oxidative metabolism, and blood supply during mental tasks. 2 . Okada et a l . 14.51 presentedworkonimpairedinterhemisphericintegration in brain oxygenation and hemodynamics in schizophrenia. 3. Hoshi et al. [46] looked into the features of hemodynamic and metabolic changes in the human brain during all-night sleep. 4. Hirth et al. 1471 studied the clinical application of the near infrared in migraine patients. They assessed the transient changes of brain tissue oxygenation during the aura and headache phasesof a migraine attack. 1.

Surgeons are concerned with brain blood flow to patients undergoing cardiopulnlonary bypass surgery. Chow et al. 1481 conducted an intensive study in

332

Ciurczak

which blood flows were restricted to patients from age 2 weeks to over 20 years. NIR spectroscopy was used to correlate blood flow rate with NIR spectra of the brain. Flows of 0.6, 1.2, and 2.4 (L . m’)/min were used. Their results showed that flow was related to mean arterial pressure but did not correspond to pulsality. This was interesting in that pulse rate is often used as a diagnostic to ensure sufficient blood flow to the brain during surgery. Totaro et al. [49] published a detailed paper on the factors affecting meaNIR spectroscopy. surement of cerebrovascularreactivitywhenmeasuredby Some of the points covered were the relative transparency of the skin, skull, and brain in the 700-1 100 nm region and the oxygen-dependent tissue absorption changes of hemoglobin. Their study covered all relevantfactors such as age, sex, reproducibility (often neglected in many “academic” papers), and venous return. The test was based on a 3 min baseline, a 3 min hypercapnia (5% C 0 2 in air), and a 2 min recovery period. Changes in NIR spectra and transcranial Doppler sonography parameters weresignificantlycorrelatedwithvariations of end-tidal C 0 2 ( P < 0.005). In addition,asignificantcorrelationbetweenthereactivityindexes of NIR spectrometry parameters and flow velocity was found ( P < 0.01). Other high reproducibilitywasfoundfordeoxyhemoglobin (rl = 0.76),oxyhemoglobin (1-1 = 0.68), and flow velocity (rl = 0.60) reactivity indexes. No significant differences between the reactivity indexes of different body positions were found ( P < 0.05). The reactivity index of oxyhemoglobin decreased ( P > 0.05) and that of deoxyhemoglobinincreased ( P < 0.01) withage. TheiroverallconclusionwasthatNIRspectroscopywasaviabletechniqueforevaluationof cerebrovascular reactivity for patients with cerebrovascular disease. Some exciting work was reported by Hitachi at a recent meeting in Japan [50].The research, conducted at the Tokyo Metropolitan Police Hospital, was to use NIR to detect blood flow changes in the brain to determine sites of epileptic activity. The location of blood flow increases responded well with conventional methodssuchasintercranialelectroencephalography(EEG) or single-photon emission computed topography (SPECT). The technique was able to determine the side of the brain where the episode was taking place in all the patients on which it was tried. This technique could replace the intrusive electrodes currently in use. Hitachi plans to expand the use of this technique to other brain diseases.

IV. TISSUE Tissueisasplitsubject;theuse of NIR for suchthingsasmuscledamage is partially discussed in Section I11 with reference to oxygen determination. A particular work may have been dedicated to one diagnostic topic more strongly than the tissue portion of it. Nonetheless, a body of work exists that is purely about tissue analysis by NIR spectroscopy.

Biomedical

Dreassi and coworkers have published a series of papers on atopy of skin. In the first [51], Dreassi et al. discuss how NIR radiation penetrates complex structured matrices to at least 0.20 mm. They found that the near infrared gives valuable insights into the stratum corneum. Using principal component analyses, they decomposed the global structural information into components such as water and lipid structures. In another paper 1521, the group studied the interactions between skin and propylene glycol 400 (PEG 400), isopropyl myristate (IPM), and hydrogel. They examined spectral differences and differences in response in terms of water and lipid content between normal and atopic skin after reaction with these reagents. Theseparticularchemicalswerechosen to representaprevalentlyhydrophilic solvent (PEG 400), a prevalently lipophilic solvent (IPM), and a hydrophilic pharmaceutical gel used to promote contact in electrocardiography. By to distinusing principal component analysis (PCA), Dreassi et al. were able guish atopic from normal skin simply from their reactions to contact with these reagents (and, of course, NIR). Similar results were reportedin a later work from this group [53] (third in a series). Using a series of perfluorinated polyethers (fomblins) of differing molecular weight and viscosity, the NIR spectra of normal and atopic skin were assessed. The interaction between the chemicals and the organ (skin) consist of two stages: The skinis physically modified, and the wateris moved and redistributed. It was assumed that the chemical agents caused changes in the stratum corneum and that the changes differed between normal and atopic skin. Although different mechanisms appeared to operate in each case, each of the chemicals showed differences between atopic and normal skin. One important assessment made with NIR spectroscopy is the viability of in tissue tissue after trauma [54]. Prolonged and severe tissue hypoxia results necrosis in pedicled flaps. Dreassi’s group used NIR techniques to identify tissue regions with a low oxygen supply. The workwasperformedonreversed McFarlane rat dorsalskinflaps. It wasseenthatoxygendeliverytotheflap tissue dropped immediately. As expected, severe trauma that causes the skin to be severed from the main blood flow causes necrosis of the tissue. NIR may be used as a tool in assessing the success of reattachment of the traumatized skin. Similar work isbeingperformedbyLindinSweden [%I. Heisintroducing monitoring equipment to monitor tissue oxygenation of patients in need of hyperbaric oxygen (HBO) treatment. Tissue perfusion and oxygenation are objectiveparametersthatsupportthedecisiontoadminister HBO andwould help determine the number of treatments needed. Since hypoxia is reversible, NIR spectroscopy is believed to be the most rapid tool for saving life and limb in a trauma situation. Another tissue type, nails, was studied by Sowa et al. [56] both in vivo and ex vivo. Mid-infrared (MIR) and near-infrared (NIR) spectra were taken of viable and clipped human nails. Depth profiling was physically performed by

334

Ciurczak

MIR and performed nonintrusively by photoacoustic spectroscopy (PAS). Nearinfraredattenuated total reflectance(NIR-ATR),NIRdiffusereflectance,and PAS were compared. Assignments were made, such as an N-H stretch-amide I1 bendcombinationbeingcentered at 4868 cm" in thisbasicstudy.They concluded that for purposes of their study, the lower energy NIR-ATR gave the best results. Interesting measurements, such as of body fat in infants, are easily made with NIR images [57].Newborn body fat can be evaluated as being due to breast feeding versus non-breast feeding. This is more accurate than mere body weight that includes other tissues and bone. Another application to foetuses and newborns was published by Liu et al. in 1997 1581. In this paper,theypresentedameasureof fetal lungmaturity from the spectra of amniotic fluid. The lecithin/sphingomyelin (L/S) ratio was determined by thin layer chromatography (TLC) and used to calibrate an NIR equation using the whole amniotic fluid extracted from pregnant women. About 350 p L of fluid was required. This was scanned from 400 to 2500 nm using a commercially available instrument. The correlation between further samples of fluid and TLC results was about 0.91, considered excellent for the complexity of the solution and extremely small sample size. Of course, a PLS equation was needed because of the complexity of the samples. The temperatureoftissuewasmeasured by Barlowetal. in 1995 [59]. Absorbance changes in the water spectrum between 700 and 1600 nm (in transmission) and the spectrum between 800 and 2200 nm (reflectance) were found to correlate with the temperature of the tissue in which it is contained. The standard error of estimate (SEE = 0.02-0.12"C) and standard error of prediction (SEP = 0.04-0.12"C) were found. Since tissue, in general, is a highly dispersing medium, various attempts have been made to mitigate this scattering. Tsai et al. [60] presented a paper that was merely concerned with the absorption properties of soft-tissue constituents. They concentrated their work in the region 900-1340 nm. In the near infrared the shorter wavelength regions have lower scattering coefficients, lower absorptivities, and, consequently, deeper penetration in tissue. At the same symposium, Schmitt et al. presented a paper [611 in which they discussed the processing of NIR spectra from turbid biological tissue. Much energy hasbeen used to obviate the scattering effects of tissue. Discriminating actual absorption of light versus losses due to scattering demands the use of higher order algorithms. Since tissue both scatters and attenuates any light passing through it, many papershavebeenpublished,notaboutanyparticulardiagnosistechniquebut about the measurement approach itself. At MIT, Rava et al. [62] used an NIR Raman device to generate spectra. The group used a Nd:YAG laser to penetrate the tissue with sufficient power that a reasonable spectrum could be collected. They used a charge-coupled device (CCD) to collect sufficient light for a high signal-to-noise spectrum.

Biomedical

335

Anderson-Engels and coworkers (in Sweden) presented a series of papers on time-resolved transillumination of tissue, specifically with tumor detection in mind [63-671. In these papers, the group goes into detail about the physics involved in using a picosecond diode laser, a mode-locked argon ion/dye laser, or a mode-locked Ti sapphire laser to conduct time-resolved spectroscopy on tissue. In one case. a human (female) breast is compressed to -35 mm for the test. Light in 100 femptosecond (fs) pulses (at 792 nm) (giving aSO ps apparatus function) is dispersed to asignalthat is morethan 1 ns long. The dispersion curve obtained contains information about the optical properties of the tissue. In the case of scattering-dominated attenuation (scattering coefficient >> absorption coefficient), detection of early transmittal light will be practically insensitive to variations in the absorption coefficient. The scattering properties determine the amount of detectedearlylight.This is importantforopticalmammography, for which neovascularization surrounding a tumor causes an increase in light absorption in thetumorregion.Amodelhas beendevelopedthataccurately predicts the time-dispersion curves obtained experimentally. This breakthrough will greatly aid in mammography studies. The clinical purpose of theAnderson-Engelsgroupwastostudyintact arterial walls and compare normal and atherosclerotic patients. The technique appeared to allow monitoring a number of molecules intimately associated with atherogenisis, including collagen and elastin, cholesterol and cholesterol esters, and calcium salts. The drawbacks to their work were the 30 min collection times and the 500 mW power needed for the studies. The pioneering work in the assessment of arterial walls was performed at the University of Kentucky Medical Center. Robert Lodder and coworkers have been producing excellent results in this field for a decade [68]. In Ref. 68, the location and quantities of high density lipoprotein (HDL), low density lipoprotein (LDL), and apolipoproteins in living tissue were determined. A compound parabolic concentrator (CPC) similar to the one used for solar power concena transmitting optical fiber onto tration was used to compress the beam from a small spot on the artery surface. This CPC was molded from a polymer and contained a polished aluminum lining. Near-infrared light in the1100-2500nmrangewastransmittedthrough the concentrator onto the exterior arterial wall. The scattered light was detected to at theproximalend of theCPC by leadsulfidedetectorslocatedoff-axis the incident beam. False color maps were then produced in which the types of plaque and the amounts of each type were determined. The software used in this type of research has been steadily improved. In a recent presentation, Lodder’s group presented the latest software breakthrough [69]. In this work, the procedure was extended to the prediction of plaque at risk of breaking free and having the potential to cause a stroke. The software, named CALDATAS, was the latest incarnation of a series of multivariate programs developed by Lodder, beginning with two called BEAST and BEST. As mentioned

Ciurczak

336

previously, the nature of NIR spectra requires intensive software interpretation. Three-dimensional modeling, such as in Lodder’s work, requires intensive “data crunching.” (The body of work by Lodder is too extensive to summarize in this chapter. It is recommended that readers visit his Web page for a full accounting of the work [70].) The group has posted a minireview of their work on their Web page also [71].

V.

CHEMISTRY (IN VITRO)

“Chemistry” usually implies in vitro uses of NIR. As an example, cell culture media were analyzed by McShane and Cot6 in 1998 [72]. Samples of a 3 day fibroblast culture were analyzed by standard clinical techniques as well as by NIR.Glucose,lactate,andammoniaweredeterminedafterbuildingamodel to followthenutrifromseveral lots of cell culture media. The purpose was entlevelstodeterminenoninvasivelywhenfermentationwascomplete. The approach was deemed successful. A nice little study by Hall was performed by using NIR to analyze the major components of human breast milk [73]. This application could help nutritionists determine (quickly) whether a nursing mother needs supplements for her child. Shaw et al. [74] performed some excellent analyses of urine samples. They used NIR to quantify protein, creatinine, and urea. Since water is not as big an interference in the near infrared as it is in midrange infrared, they easily carried out the analyses. Standard errorsof prediction (SEPs) for the urea, creatinine, and protein were 16.6, 0.79, and 0.23 mmol/L, respectively. They used 127 samples for each calibration. Both MLR and PLS equations were generated, but PLS was eventually used to compensate for person-to-person variations. They concluded that the protein measurements would only be good for coarse screening whereas the other two types of measurements were comparable to current methods. The rapid nature of this test, usingno reagents, is a marked improvement over current clinically accepted methods in terms of speed and throughput. Further urine analyte analyses were performed by Jackson et al. and reported in 1997 [75]. Urine glucose, protein, urea, and creatinine concentrations were analyzed using rather simple algorithms. Urea, for instance, was calibrated by simply correlating with the absorbance at 2152 nm. The comparison with of nearly 1.00. Since standard methods gave a linear relationship with a slope in lower quantities and have lower absorpcreatinine and proteins are present tivities, a more complex algorithm, PLS, was needed to analyze the materials. The best correlation for creatinine gave a slope of 0.953, and protein produced a slope of 0.923. In critical situations, where speedis more important than absolute numbers, NIR spectroscopy would be an important tool.

Biomedical

337

Asan example of ex vivo determinations,Shawetal.[76]wereable to correlatethechemistry of synovial fluid drawnfrompatients’kneeswith a series arthritis diagnosis. Conventional chemical analyses were performed on of patientswithvarioustypes of arthritis;thenNIRscansweretaken of the fluids. A model equation was built using PLS. The prediction of further arthritis sufferers was remarkably good when the equation was tested. An interesting multitool approach was employed by Schultz et al. in which the near infrared and midrange infrared were used to examine the structure of ribonuclease A (RNase A) [77]. Because aqueous work in the MIR is difficult at best, the thermal unfolding of the protein was used as a model for the structural the N-H combinationband changesoccurring in water. In thenearinfrared, (amide bond) at 2055 nm, found in nativeRNaseA,wasshifted to 2050 nm uponthermalunfolding.Hydrogen-deuteriumexchangeexperiments,usedto validate the N-H character of this mode, were also used to estimate the number of unexchanged amide protons after exposure to D2O. The transition profiles and temperatures derived from the temperature dependence of the N-H combination mode were found to strongly correlate with those derived from the temperature dependence of the C=O amide I band in theMIRregion. The conclusionwasthattheNIRregioncanbeusedasa conformation-sensitive monitor of the thermally induced unfolding of proteins in H20 solutions. Somemodelingwasperformed onpolypeptidesbyscientistsfromthe is working on medical NationalResearchCouncil of Canada [78]. The team diagnosticmethodology. In theirstudy,near-infraredphotoacoustic(NIR-PA) spectroscopywasusedtostudy19homopolypeptides.Thebiochemicalinformation gleaned was compared with previous data from midrange infrared. Alstudy, it is one of the first though no conclusions per se were drawn from this wherein structural variation of the polypeptides were investigated at this basic level. Specific modes were assigned (e.g., the CH-stretch combination region). This is an important paper in that it does not merely use chemometrics to correlate NIR spectra with physiological activity but actually begins molecular level investigations of structural changes in proteins. Cell bioreactors are ripe targets for NIR monitoring. Mark Riley et al. of the University of Arizona reported on fermentation control at a 1999 meeting [79]. In this novel approach, the workers used a computer simulation to generate spectra of mixtures of components found in fermentation mixtures. This model allowed for reasonable assay values for, at first, a simple binary solution of glucose and glutamine. They then modeled a complex solution containing varying concentrations of ammonia,glucose,glutamate,glutamine,andlactate.They found predictions to have larger errors than standard analyses for these components, but the rapid generation of data allows for satisfactory determinations in conditions not easily assayed by conventional assays.

Ciurczak

338

Another fermentation study of note was reported by Hall et al. in 1996 [SO]. AsimultaneousNIRassayforacetate,ammonia,biomass,andglycerol was developed for an industrial Escherichicr coli (E. coli) fermentation broth. The PLS equationproducedwascapable of predictingwithstandarderrors, respectively, of 0.7 g/L, 1.4 g/L, 0.7 g/L, and 7 mmol/L for thelisted constituents. Standard wet chemical methods were used to calibrate the NIR equation.

VI.

BIOCHEMISTRY (IN VIVO)

Blood chemistry is well characterized in vitro through numerous chemical and instrumental methods. In vivo or in situ testing of blood chemistry has been perinto the patient. formed with the aid of micro (or not-so-micro) probes inserted These can now be correlated with noninvasive spectroscopic methods. One nice example of this work was performed at the University of Massachusetts Medical Center by Zhang [SI]. In this work, NIR was correlated with standard pH measurements to perform in vivo determination of the myocardial pH during regional ischemia. Some interesting work was reported on the Web at the Regensburg University Web site [82]. They had been working on developing fluorescent probes wherein the markers are covalently bound to biomolecules. The labels are activated as N-hydroxysuccinimide esters, isothiocyanates, or iodoacetamides and maleimides with reactivity toward different functional groups in biomolecules and are being tested as fluorescent markers. The penetrating power of NIR light is the driving force behind this research. The probes are used in immunology and cell biology as biotinylated fluorescent markers for enzyme amplification, membrane probes, DNA probes, and asspecific enzymesubstratesfordetermination of proteases.lipases,alkaline phosphatase, and peroxidase.

VII. CANCER Although it is in its infancy, the use of NIR for cancer research is beginning to be published. Its nonintrusive nature, as in blood chemistry work, is appealing to any number of researchers. Workers at Johns Hopkins University, under the tutelage of Chris Brown, screened PAP smears by using NIR spectroscopy [83]. Normal or healthy patients, patients with abnormal cells, and patients with cervical cancer were screened. Using discriminant analysis and principal component to examine additional samples. It analysis, the samples were grouped and used was seen that malignant and healthy tissues were distinctly different, whereas abnormal tissues carried spectral features from both sets.

Biomedical

339

Mammograms are oftcn uncomfortable and embarrassing for women. Using NIR spectroscopy, imaging [84] and an “optical biopsy” 1851 may be performed. Since NIR radiation has some unique features, it has been suggested as an alternative to both X-rays and physical, invasive biopsies. In younger women. with breasts dense to X-rays, NIR could be used to eliminate false positives. Breast cancer is one of the leading causes of death and disfigurement in women, so early detection is important [SS]. Magneticresonanceimaging (MRI) is used in caseswhereX-raysare questionable. MRI is not chemically specific and only shows masses more clearly than X-rays. Using NIR imaging simultaneously could give a better picture of the mass’s chemistry [861. A time-resolved imager capable of acquiring images simultaneously was used for this work. Short-wavelength radiation in the range 780-830 n m was found to be best. 1861. who SomeimportantworkwasperformedbyNtziachristosetal. used both magnetic resonance imaging and NIR imaging to afford precise coregistration of images and examine the potential and limitations of optical mammography. Using a time-resolved imager of their own design, the group acquired NIR images simultaneously with MR images. The intrinsic contrast at 780 and 830 nm wasused to study the relativeenhancementandkineticsdue to the administration of Infracyanamine R25, an NIR contrast agent. In 1994, Meurens et al. [87] were able to determine that cryostat sections of carcinomatous tissue were different spectrally from noncarcinomatous tissue. Four distinct wavelength regions between 1200 and 2370nm were found best for classification of the tissue samples. Samples included invasive ductal carcinoma (IDC), IDC withapredominantintraductalcomponent,mucinouscarcinoma. and invasive lobular carcinoma. Despite the varied types of canccr cells, there of cancerous versus noncancerous cells. This work is was a distinct grouping being carried over to potential in vivo measurements.

VIII. PHYSICS, PHYSICAL PARAMETERS,MATH, AND IMAGING In all the topics, the advancement of the instrumentation and software that have enabled the scientists to perform the research is often ignored. One interesting paper presented by Abbot et al. [88] used laser-Doppler perfusion imaging to followskinbloodflow. The work wasdone in theredandshort-waveNIR regions of the spectrum. These wavelengths have been used before, but Doppler imaging is a new step in instrumentation for this topic. An ingenious multichannel instrument for tissue imaging was developed at the University of Illinois (Urbana-Champaign) [89]. The thrust of the research was to develop a frequency domain instrument for noninvasive, real-time NIR

340

Ciurczak

opticaltomography of tissue in vivo. The focus of theresearchwasonconstructing a spatial map of the optical properties of a strongly scattering medium in a semi-infinite-geometry sampling configuration. Thealgorithmsareoften merelyreferred to peripherally at best. In an article by Piantadosietal. [901, thealgorithmsarethetopic of discussion. Using a half dozen research papers (by other workers) as examples, the authors discusstheapproachesusedin NIR work,citingboththehardwareandthe software.Mostofthepracticalapplicationswereonblood,andtheanalytes were hemoglobin, deoxyhemoglobin, etc. Oneinterestingpaper [911 discussedatechniquecalledfuzzyoptimal associativememory(FOAM)thatisusedforbackgroundprediction of NIR spectra. This software yields better background scans for the calculationof NIR spectra of glucose in plasma matrices (from single-beam data). FOAM is usually used in conjunction with PLS and/or some other complex algorithm. on a finite element approach for Arridge et al. [92] published some work modelingphotontransport in tissue. In thismethod,calledFEM,thephoton densityinside an objectandphotonflux at itsboundaryareintroducedinto modeling light transport through tissue. The paper covers the math with which they derive a model for one particular case. The calculation of the boundary flux is afunction of timeresultingfroma&functionpointinput to atwodimensional circle (showing as a line source in an infinite cylinder) with homogeneous scattering and absorption properties. This model may be of some use to subsequent researchers, especially in brain work, where scattering and light loss are extensive.

IX.REVIEWARTICLES Instead of re-covering the ground of several excellent review articles, their contents are merely enumerated here, and the references are given for the reader to peruse at his or her leisure. An interesting article about “bloodless testing” appeared in the October 1998 issue of Scient$c Atnericar~in the “Technology and Business” section. It listed a number of nonintrusive diagnostic tools for determining glucose in blood but mentioned only one that uses the near infrared, the Diasensor 1000 by Biocontrol Technology of Pittsburgh. Although the company has been working with the Food and Drug Administration since 1994, the device is still not approved for human use. One interesting review article, while aimed at the nonprofessional. is still worthnoting[93].Itsauthorgives a succinctpicture of thevarioustypes of noninvasivetechnologiesunderinvestigationtodayandprovides a reasonably goodbibliographyforfurtherinvestigation.HeincludesNIRmethods in the overall context of the subject.

Biomedical Applications

341

An NIR-laced article by Heise et al. 1941 discusses the technologies used in noninvasive glucose monitoring. In this article, they describe their own work of other NIR withanFT-NIRinstrumentbut alsoprovideaniceoverview applications plus luminescence, optical activity, and Raman spectroscopy. They provide 38 references on these topics. An interesting summary page was posted on the Web [95] in which the market for home blood glucose testing products was assessed. The paper covers only through 1994, but it gives an idea of why so muchworkperformed in the field. Asofearly1999, one therehasbeen source states that the diagnostic market could top $20 billion in the near future. Another rather nice review article about the state of noninvasive testing in 1997 is that of Cote 1961. In it he mentions midrange infrared, near-infrared, polarimetry, and Raman techniques. He also spends some time on the various multivariate calibration methods used by the researchers. The paper contains 41 references. Pioneer Mark Arnold critically examined a number of published claims of blood glucose measurement [97]. He concluded that many of the published claims needed work that is more fundamental before any could be considered applicable to clinicalsettings. In thesamepaper, he expresses hisoptimism about fermentation measurements using NIR technology. A very good review paper pertaining to tissue imaging up to 1992 is by Joffe [98]. He cites34 published articles on topics suchas oxygen measurements, but the strength of the article is in the detailed comparison of NIR with other methods such as X-ray, MRI, ultrasound, positron emission, thermal emission, electricalimpedance,andmore.Hediscussesthevarioustypes of equipment used, various detectors, and light sources. Healso gives the best short description of time-resolved spectroscopy I have seen in any review article.

REFERENCES I.

2.

3.

4.

5.

FMHam.GM Cohen. Non-invasivebloodglucosemonitoring.Patent5,553,616, 1997. I Gabriely. R Wozniak, M Mevorach, J Kaplan, Y Ahron, H Shamoon. Performance ofa novel near-infrared (NIR) transcutaneous glucose monitor during hypoglycemia. 59thScientificSession.AmericanDiabetesAssociationMeeting,SanDiego,CA, 1999. MA Arnold, LA Marquardt.Near-infraredspectroscopicmeasurementofglucosc in a protein matrix. Anal Chem 65:3271, 1993. MA Arnold, G Small, LAMarquart.Temperature-insensitivenear-infraredspectroscopic measurement of glucose in aqueous solutions. Appl Spectrosc 48(4):477, 1994. MA Arnold, S Pan, H Chung, G Small. Near-infrared spectroscopic measurement of physiologicalglucoselevels in variablematrices of proteinandtriglycerides. Anal Chem 68(7):1 124, 1996.

342

Ciurczak

6. GW Small et al. Evaluation of data pretreatment and model building methods for the determination of glucosefromnear-infraredsingle-beamspectra.ApplSpectrosc 53(4):402,1999. 7. JJ Burmeister, MA Arnold. GW Small. Spectroscopic considerations for noninvasive blood glucose measurements with near-infrared spectroscopy. IEEE Lasers ElectroOpt SOC12:6,1998. for 8. MR Riley, MA Arnold, DW Murhammer. Matrix-enhanced calibration procedure multivariate calibration models with near-infrared spectra. Appl Spectrosc 52( IO): 1339.1998. 9. DM Haaland, MR Robinson, RP Eaton, GW Koepp. Reagentless near-infrared determination of glucosein whole blood using multivariate calibration. Appl Spectrosc 46( IO): 1575, 1992. IO. WF Schrader. Non-invasive anterior chamber glucose monitoring by near-infrared absorption spectroscopy, an alternative to blood-glucose monitoring in diabetic patients'? Proc. 96th DOG Annual Meeting, 1998. I I . J Backhaus et al. Device for the in vivo determination of an optical property of the aqueous humor of the eye. US Patent 5,535,743, July 16, 1996. 12. U Fischer, K Rebrin, T Woedtke, P Able. Clinical usefulness of the glucose concentration in thesubcutaneoustissue-Propertiesandpitfallsofelectrochemical biosensors. Horm Metab Res 26515, 1994. EF Pfeiffer.Sub13. FSternberg,CMeyerhoff.FJMemel,HMeyer,FBischoff, cutaneousglucoseconcentration:Itsrealestimationandcontinuousmonitoring. Diabetes Care 18: 1266, 1995. 14. MA Arnold, JJ Burmeister, H Chung. Phantoms for noninvasive blood glucose sensingwithnear-infraredtransmissionspectroscopy.PhotochemPhotobiol67( 1):50, 1998. 15. MA Arnold, JJ Burmeister.GSmall.Phantomglucosecalibrationmodelsfrom simulated noninvasive human near-infrared spectra. Anal Chem 70: 1773, 1998. 16. JS Maier et al. Possible correlation between blood glucose concentration and the reduced scattering coefficient of tissues in the near-infrared. Opt Lett 19(24):2026, 1994. 17. K Jagemann et al. Applications of near-infrared spectroscopy for non-invasive determination of bloodtissue glucose using neural networks. Z Phys Chem 191:179, 1995. 18. C Fischbacher, UA Muller, B Mentes, KU Jagermann, K Danzerk. Enhancing calibration models for non-invasive near-infrared spectroscopic blood glucose determination. Fresenius J Anal Chem 359:78, 1997. 19. KDanzer,UAMuller,BMertes,CFischbacker,KUJungernann.Near-infrared diffusereflectionspectroscopyfornon-invasiveblood-glucosemonitoring.IEEE LEOS News1 4( 18). 1998. 20. H Shamoon, I Gabriely. R Wozniak, M Mevorach, J Kaplan, Y Aharm. Transcutaneous glucose monitor during hypoglycemia.Am Diabetes Assoc., San Diego, June 1999. Paper 426. 21. HM Heise, R Marboch, G Janatsch, KD Kruse-Jarres. Noninvasive blood glucose assay by near-infrared diffuse reflectance spectroscopy of the human inner lip. Appl Spectrosc 47(7):875, 1993.

Biomedical

343

22. FF Jobsis. Science 198:1264. 1977. S Maraishi, KKawauchi.Nonde23. y Ozaki, AMizuno.THayashi,KTashibu. structiveandnoninvasivemonitoringofdeoxyhemoglobinintheveinbyuseof anear-infraredreflectancespectrometerwithafiber-opticprobe.ApplSpectrosc 46(1):180.1992. 24. MG Sowa etal. Noninvasive assessment of regional and temporal variations in tissue oxygenation by near-infrared spectroscopy and imaging. Appl Spectrosc 5 l(2):143, 1997. of 25. DM Mancini. L Bolinger.HLi. K Kendrick,BChance,JRWilson.Validation near-infrarcd spectroscopy in humans. J Appl Physiol 77(6):2740, 1994. 26. L Lin, DA York. Two-layered phantom experiments for characterizing the influence of a fat layer on measurement of muscle oxygenation using NIRS. Proc SPIE 3257 (Photonics West). San Francisco, January 1998, Paper 41. 27. K Yamanloto et al. Near-infrared muscle oximeter that can correct the influence of a subcutaneous fat layer. Proc SPIE 3257 (Photonics West). San Francisco, January 1998,Paper17. 28. MA Franceschini et al. The effect of water in the quantitation of hemoglobin concentrationinatissue-likephantombynear-infraredspectroscopy.OpticalSociety of America, Washington, DC, 1996. 29. KFeldscher.Anovelway to measureoxygeninblood. The NortheasternVoice, www.voice.neu.edu/970123/oxygen.htn~l, July12,1999. 30. ZX Jiang etal. Novel NIR instrument for non-invasive monitoring and quantification of cerebral tissue oxygenation. Proc SPIE 3257 (Photonics West), San Francisco. January 1998. Paper 44. 31. R Marbach, HM Heise. Optical diffuse reflectance accessory for measurements of skin tissue by near-infrared spectroscopy. Appl Opt 34(4):610, 1995. 32. K Miyasaka. NIRS use in pediatricanesthesiaandICU. 96 PICUConference. Toronto. 33. S vanHuffel et al. Changesinoxygenationandhemodynamicsinneonatalbrain bymeansofnear-infraredspectroscopy:Asignalanalysisstudy.Dept of Pediatrics and Neonatal Medicine, University Hospital Gasthuisberg, Leuven, Belgium. www.esat.kuleuven.ac.be/sista/yearreport96/node6.html.

34. S Van Huffel et al. Modelingandquantificationofchromophoreconcentrations, basedon opticalmeasurementsinlivingtissues. eee.esat.kuleuven.ac.be/sista/ yearreport/node33.html.1998. 35. CE Cooper, J Torres, M Sharpe, MT Wilson. The relationship of oxygen delivery to absolute hemoglobin oxygenation and mitochondrial cytochrome oxidase redox state intheadultbrain: Anear-infraredspectroscopystudy.Biocheln J 332:(j27, 1998. 36 JS Wyatt, M Cope, DT Delp, CE Richardson. AD Edwards. S Wray. EO Reynolds. Quantitationofcerebralbloodvolumeinhumaninfants bynear-infraredspectroscopy. J Appl Physiol 68:1086, 1990. 37 VV Kupriyanov, RA Shaw, B Xiang, H Mantsch, R Deslauriers. Oxygen regulation of energy metabolism in isolated pig hearts: A near-IR spectroscopy study. J Mol Cell Cardiol 29243 I , 1997.

Ciurczak

344

38. JR Mansfield et al. Fuzzy C-means clustering and principal component analysis of

time series from near-infrared imaging of forearm ischemia. Comput Med Imaging Graphics 2 I(5):299, 1997. 39. TWolf,ULindauer.HObrig, J Drier, J Back,AVillringer, 0 Dirnagl. J Cereb Blood Flow Metab 16:1100-1107, 1996. 40. KD Liem, JC Hopman, LA Kollee, R Oeseburg. Effects of repeated indomethacin administrationoncerebraloxygenationandhemodynamicsinpre-terminfants: Combinednear-infraredspectrophotometryandDopplerultrasoundstudy. Eur J Pediat153(7):504,1994. 41. H Nguyen,G Murphy. Centrefor Biomedical Technology, Universityof Technology, Sydney, Australia, and P Cooper et al. CRC for Cardiac Technology, July 12, 1999. www.eng.uts.edu.auk”tnlresearch.htm1

42. TJ Germon, PD Evans. AR Manara, NJ Barnett, P Wall, RJ Nelson. Sensitivity of near infrared spectroscopy to cerebral and extra-cerebral oxygen changes is deterI . 1998. mined by emitter-detector separation. J Clin Monit 10: a cerebral oximeter in 43. LC Henson, C Calalang, JA Temp, DS Ward. Accuracy of healthy volunteers under conditions of isocapnic hypoxia. Anesthesiology 88( 1):58, 1998. 44. Y Hoshi, H Onoe, Y Watanabe, J Anderson, M Bergstram, A Lilja, B Langstrom, M Tamura. Non-synchronous behavior of neuronal activity, oxidative metabolism, and blood supply during mental tasks in man. Neurosci Lett 172:129, 1994. 4s. F Okada, Y Tokumitsu, Y Hoshi, M Tamura. Impaired interhemispheric integration in brain oxygenation and hemodynamics in schizophrenia. Eur Arch Psychiatry Clin Neurosci 244:17, 1994. 46. Y Hoshi, S Mizukami, M Tamura. Dynamic features of hemodynamic and metabolic changesinthehumanbrainduringall-nightsleepasrevealedbynear-infrared spectroscopy. Brain Res 652:257, 1994. 47. C Hirth et al. Clinical application of near-infrared spectroscopyin migraine patients. www.ukrv.de/ch/neuro/hirth.html, 1998. 48. Chow et al. The relation between pump flow rate and pulsatility on cerebral hemoJ ThoracicCardiovascSurg dynamicsduringpediatriccardiopulmonarybypass. I14(4):1123.1997. 49. R Totaro, G Barattelli, V Quaresima, A Carolei, M Ferrari. Evaluation of potential factorsaffectingthemeasurement of cerebrovascularreactivitybynear-infrared spectroscopy. Clin Sci 95:497, 1998. 50. Hitachi. “Team develops world’s first light-based procedure for examining epilepticbrainsites”(newsrelease). http://koigakubo.hitachi.co.jp/research/med/release/ br.html SI. E Dreassi, G Ceramelli, L Fabbri, F Vocioni, P Bartalini, P Corti. Application of near-infrared reflectance spectroscopy in the study of atopy. Part 1. Investigation of skinspectra.Analyst122(8):767,1997. 52. E Dreassi, G Ceramelli,PBura,PLPerruccio,FVocioni,PBartalini,PCorti. Application of near-infrared reflectance spectroscopy in the study of atopy. Part 2. Interactions between the skin and polyethylene glycol 400, isopropyl myristate, and hydrogel. Analyst 122(8):77 I , 1997.

Biomedical Applications

345

53. P Corti, G Ceramelli, E Dreassi, S Mattii. Application of near-infrared reflectance spectroscopyinthestudyofatopy.Part 3. Interactionsbetweentheskinand fomblins.Analyst 122(8):788, 1997. 54. MF Stranc, MG Sowa, B Abdulrauf. HH Mentsch. Assessment of tissue viability using near-infrared spectroscopy. Br J Plastic Surg 51:210, 1998. 55. F Lind. Tissue hypoxia monitoring and hyperbaric oxygenation. Department of SurgicalSciences,KarolinskaInstitute,Stockholm.Sweden. http://research.kib.ki.se/ e-uven/public/K3794.html

56. MG Sowa et al. Infrared spectroscopic investigation of in vivo and ex vivo human nails.VibSpectrosc10:49,1995. 57. N Kasa, KM Heinonen. Near-infrared interactance in assessing superficial body fat 82: I , 1993. in exclusively breastfed, full-term neonates. Acta Paediatr 58. KZ Liu, TC Dembinski. HH Mantsch. Prediction of fetal lung maturity from nearinfrared spectra of amniotic fluid. Int J Gynocol Obstet 57:161. 1997. 59. CH Barlow et a l . Tissue temperature by near-infrared spectroscopy. In: B Chance, RR Alfano, eds. Optical Tomography, Photon Migration, and Spectroscopy of Tissue and Model Media. SPIE 2389:818. 1995. 60. C Tsaietal.Absorptionpropertiesofsofttissueconstituentsin900-1340nm region. Proc SPIE 3257 (Photonics West), San Francisco, January 1998. Paper 14. 61. JM Schmitt et al. Interpretation and processing of NIR spectra of turbid biological tissue. Proc SPIE 3257 (Photonics West), San Francisco, January 1998. 62. RPRavaetal.Rapidnear-infraredRamanspectroscopyofhumantissuewitha spectrograph and CCD detector. Appl Spectrosc 46(2):187, 1992. of tissue for medical 63. S Anderson-Engelsetal.Time-resolvedtransillumination diagnostics. www-lmlc.fysik.lth.se/Prog939S/p43.ht1n. l . , eds. Med64. G Muller et al. Time-resolved transillumination imaging. In: R Berg aet icalOpticalTomography:FunctionalImagingandMonitoring. SPIE lnst Ser I I . SPIE, Bellingham, WA, 1993, pp. 397-424. 65. 0 Jarlman, R Berg, S Svanberg. Time-resolved transillumination of the breast. Acta Radio1 33277, 1992. 66. RBerg, S Andersson-Engels, K Rama.Medicaltransilluminationimagingusing short pulse diode lasers. Appl Opt 32:574. 1993. 67. s Anderson-Engels, R Berg, K Rama. Time-resolved transilluminationof tissue :lnd tissue-like phantoms for medical diagnostics. SPIE 2081. Budapest. Hungary, 1993, pp. I 37- 146. 68. RALodder, L Cassis.Arterialanalysiswithanovelnear-IRfiberopticprobe. Spectroscopy 5(7):12, 1990. 69. Characterization of vulnerable plaques by near-infrared spectroscopy in the atheroscleroticrabbit.AmericanCollegeofCardiology,48thAnnualScienceSession, New Orleans, March 1999. 70. RA Lodder. http://kerpuac.pharm.uky.edu/mcpr/news 71. RJDempsey.RLodder.Driventodepth:Biologicalandmedicalapplications of near-infrared spectroscopy. http://kerouac.pharm.uky.edu 72. MJMcShane. GL Cote. Near-infrared spectroscopy for determination of glucose, lactate, and ammonia in cell culture media. Appl Spectrosc 52(8):1073, 1998.

Ciurczak

346

73. J Hall. Analysis of human breast milk. Proc SPlE 3257. Photonics West, San Francisco. January 1998. 74. R Shaw. S Kotowich, HH Mantsch, M Leroux. Quantitation of protein, creatinine, and urea in urine by near-infrared spectroscopy. Clin Biochem 29( 1 ): 1 I , 1996. 75. M Jackson, MG Sowa.HHMantsch.Infraredspectroscopy:Anewfrontierin medicine. Biophys Chem 68:109, 1997. 76. RAShaw. S Kotowlch.HHEysel,MJackson. GT Thomson.Arthritisdiagnosis 15:159, baseduponthenear-infraredspectrumofsynovialfluid.RheumatolInt 1995.

77. CP Schultz. HH Mantsch. Two-dimensional mid-1R and near-1R correlation spectra of ribonuclease A: Using overtones and combination modes to monitor changes in secondary structure. Biospectroscopy 4:s19, 1998. 78. J Wang et a l . Near infrared photoacoustic spectra of model homo-polypeptides. J SPIE2089:492.1996. 79. MR Riley ct a l . Non-invasive quantification of cell cultures by near-infrared spectroscopy.AIChE 1999 AnnualMeeting.Session 280, NoninvasiveMeasurements. http://l98.6.4. I75/docs/meetapp/prog~~1~~1ni1~g/techprogra1n~abstr~1cts/173 I .html 80. JW Halletal.Near-infraredspcctroscopicdetermination of acetate,ammonia, biomass.andglycerolinanindustrial‘Escherichiacoli’fermentation.ApplSpecm s c SO( 1):26, 1996. 81. S Zhang. I n vivo determination of nlyocardial pH during regional ischemia using NIR spectroscopy. Proc SPIE 3257 (Photonics West), San Francisco. January 1998. Paper 13. 82. Redandnear-infraredlabelsandprobesforintensity-basedapplications.Regensburg University. lnst Anal Chem. 1996. http://pc.?898.ilni-regensburg.dc/Wolfbeis/e~ labels.html X3. Z Ge et al. Screcning PAP smears with near-infrared spectroscopy. Appl Spectrosc 49(4):1324. 1995. NIRmammographicexamination. 84. V Ntziachristos et al..SimultaneousMRIand Univ Pennsylvania. www. 1 rsm.upenn.edu/-vasilis/Concurrent.ht~nl 85. University of Illinois, Urbana-Champaign. Optical biopsy would be Fast, painless, andinexpensive.ScienceDaily,July 13. 1999. www.scicncedaily.com 86. V Ntziachristos et al..SimultaneousMRandNIRmanlmographicexamination. UnivPennsylvania1997. www.Irsm.upenn.edu/-vasilis/frresearch.I~t~~~l tissue bynear-infrared 87. MMeurcns et al. Identiticationofbreastcarcinomatous reflectance spectroscopy. Appl Spectrosc 48(2):190. 1994. 88. NC Abbot et al. Laser Doppler perfusion imaging of skin blood flow using red and near-infraredsources. J InvestDerrnatol107(6):2235. 1996. 89. MA Franceschini et al. Multi-channel optical instrument for near-infrared imaging of tissue. Presented at the 1995 SPlE Conference. On Web page of Univ Illinois, Urbana-Champaign. www.physics.uiuc.edu/groups/fluorescence/spie95 90. CAPiantadosi, M Hall, BJ Comfort.Algorithms for i n vivonear-infraredspectroscopy.AnalBiochem253:277,1997. for background 91. PB Harrington. BW Wabuyelc. Fuzzy optimal associative memory prediction of near-infrared spectra. Appl Spcctrosc SO( 1):34, 1996.

Biomedical

347

92. SRArridge.MSchweiger, M Hiraoka,DTDelpy.Afiniteelementapproachfor modeling photon transport in tissue. Med Phys 20(2):299, 1993. 93. DC Klonoff. Noninvasive blood glucose monitoring. Diabetes Care 20(3):433, 1997. 94. HH Heise et al. Clinical chemistry and near-infrared spectroscopy: Technology for non-invasive glucose monitoring. J Near Infrared Spectrosc 6349, 1998. 95. The U.S. marketforhomebloodglucosemonitoringproduces.Editorial. www. tindsvp.com/tocs/ML0392.htm

96. GL Cote. Noninvasive optical glucose sensing: An overview. J Clin Eng 22(4):253, 1997. 97. MA Arnold. Non-invasive glucose monitoring. Curr Opinion Biotechnol 7:46, 1996. 98.SNJoffe.Measuring andimaging in tissueusingnear-IRlight.OptPhotonNews October:27,1992.

This Page Intentionally Left Blank

13 Pharmaceutical Applications of Near-Infrared Spectroscopy Emil W. Ciurczak Purdue Pharma L.P, Ardsley, New York

James K. Drennen Duquesne University, Pittsburgh, Pennsylvania

1.

INTRODUCTION

Reported by Herschel in 1800, the near-infrared (NIR) region was ignored until the late 1950s. Publications describing pharmaceutical applications appeared approximately I O years later, with the majority appearing since 1986. Reviews of NIR spectroscopy were published in the early 1990s [ 1,2] and contain references to earlier reviews. Several texts of NIR are also available [3-61. Ciurczak published a comprehensive review of pharmaceutical applications [7], and several papers by others discuss topics in pharmaceuticals [8-1 I]. The conventional NIR region is between 700 and 2500 nm. The spectrum arises from absorption bands from overtones and combinations of fundamental mid-infrared stretching and bending modes. They have low molar absorptivities with broad, overlapping peaks. The low absorptivities, which arise from C-H, 0 - H , and N-H bonds, are a primary reason for the usefulness of the method for intact dosage forms. The earliest publications on NIR assays of pharmaceuticals appeared in the late 1960s but did not apply to intact dosage forms. Usually, the drug was extracted, then analyzed. In some cases, solid-state spectra were also collected. In 1966, Sinsheimer and Keuhnelian [ 121 investigated a number of pharmacologicallyactiveaminesaltsboth in solutionand in thesolidstate. In [I31 quantifiedtwodrugs: allylisopropylacetureide (AL) 1967,OiandInaba 349

350 Drennen

and

Ciurczak

and phenacetin (PH). Samples were dissolved in chloroform and quantified at 1983 nm for AL and 2019 nm for PH. Sinsheimer and Poswalk [ 141 determined water in several matrices. Solid samples were analyzed for hydrous and anhydrous forms of strychnine sulfate, sodium tartrate, and ammonium oxalate mixed with KC1 and compressed into disks containing 100 mg KC1 and 25 mg of drug. The water band at 1940 nm was seen in the hydrates in some samples.

II. QUALITATIVEANALYSIS A.

Raw Materials

A landmark paper presented by Rose et al. in 1982 [ 151 showed that a large number of structurally similar penicillin-type drugs could be identified and determined by using NIR techniques developed at their company. In 1984, Mark introduced Mahalanobis distances as an algorithm for discriminant analysis of raw materials. The software was described in a paper by Mark and Tunnel1 [ 161 andwas first applied to pharmaceuticals by Ciurczak [17]. With the advent of 100% testing, now in practice in Europe, this approach became popular quickly. In purematerials,liquids, or wherefewsamplesexist,thediscriminant techniquepresentsdifficulties.Ciurczak [ 181 suggestedatechnique in which artificial samples may be made either physically or electronically. Ciurczak also reported on theuseofspectralmatchingandprincipalcomponentsforraw materials [ 19,201 as wellas components ofgranulationsorblendingstudies [21,22]. There have been recent reports that the FDA may require in-process blend uniformity testing [23] for all products manufactured in the U.S.

B.

Blending Studies

Because almost all materials used in the pharmaceutical industry have NIR spectra, assuring blend homogeneity is a simple matter. Some of the first work on this subject was reported by Ciurczak in 1990 [24,25]. A fiber probe was used to collect spectra from various positions and depths throughout the mixer. Spectral matching and principal component analysis (PCA) were used to measure how close the powder mix in a particular portion of the blender was to a predetermined “good” or complete mix. The match index or PCA scores were plotted versus time to access the optimal blending time. Further work on this topic was performed by Wargo and Drennen in 1996 [26]. They used NIR to assess the homogeneity of a hydrochlorothiazide formulation. A sample thief was used to extract the samples, and diffuse reflectance NIR was used to analyze them. Single- and multiple-sample bootstrap algorithms and traditional chi-square analysis were used to determine blend homogeneities.

Pharmaceutical Applications of NIR

351

C. Verification of Supplies for Double-Blinded Clinical Studies When clinical trials are performed, the placebos are deliberately made to look like the active drug dosage form. In some cases, various levels of active ingredients are present. The most common approach has been to “sacrifice” some of the blister packs to ascertain whether the materials are in the correct order. NIR has proven useful for this task. Tablets may be read directly through the clear polymer casing of the blister packs. In two papers, Ritchie [27,28] described the approach to performing NIR qualitativelywithaneyetocGMPs(currentgoodmanufacturingpractices). Becauseclinicallotsareoftenadhocformulations, it is difficult togenerate a discriminantequationprior to theactualclinicaltrial.Ritchiedevelopeda procedurewherebyequationsarequicklygeneratedforanyparticularstudy, then discarded. Dempsteretal.[29]usedthreesamplingconfigurationstoqualifyan experimental drug present in tablets in 2%, S%, 10% and 20% concentrations, amatchingplacebo,andamarketeddrugusedasclinicalcomparators. The first required the tablets to be removed from the blister packs. In the second, tablets were scanned through the plastic packaging using a reflectance module. With the third arrangement, the tablets were analyzed through the plastic blister packaging with a fiber-optic probe. Second derivative spectra were used.In the first, all but the 2% tablets were easily classified; the 2% tablets could not be differentiated from the placebo. Using the second and third configurations, only 10% and 20%, respectively, of the tablets, placebo. and clinical comparator tablets could be properly classified. Another application of NIR in the analysis of clinical batches was published in 1994 by Aldridge et al. [30]. An NIRSystems Model 6500 spectrometer with a custom sampling configuration was used for spectral collection of the blister-packed samples; the second was used in the analysis with Spectralon used as a reference.

D. Active Ingredients Within Dosage Forms From 1982 through 1985, few NIR analyses of dosage forms were published. Since 1986, there has been a rapid growth in the number of articles. The first was a 1986 paper by Ciurczak and Maldacker [31], who used NIR for tablet formulationblendsandexaminedspectralsubtraction,spectralreconstruction, and discriminant analysis. Blends were prepared with active ingredients [aspirin (ASA), butalbital (BUT), and caffeine (CAF)] omitted from the formulation or varied over a range of 90-1 10% of label strength. For spectral subtraction, spectra of true placebos were subtracted, yielding spectra very close to that of the omitted drug. (Because this was a dry blend,

352 Drennen

and

Ciurczak

a“true”placebowaspossible. Wet blendsgivendifferentresultsdue to H bonding.) Constituents were identified by spectral reconstruction with commercially available software, based on work by Honigs [32] that was later expanded upon by Honigsetal. [33]. Usingaseries of mixtures of knownconcentrations, the spectrum of the drug was reconstructed, providing identification of active ingredients in the blend. The third experiment classified samples by discriminant analysis. In one series of blends, the CAF, BUT, and ASA concentrations varied independently between90%and 110% of label. In anotherseries,one of thethreedrugs 90% and was excluded from the mixture, and the others were varied between 110%.Mahalanobis distance was used in the classification of formulations. This techniquewasusedforsamples of completeformulations (all threedrugsat 100%of label strength), borderline formulations, and samples lacking one active component. In 1986, Whitfield [34] used discrimant analysis to ascertain that a veterinary drug, dosed in feed, was present before conducting a quantitative analysis. Because a simple MLR equation was used, Whitfield felt that a positive identification, confirming the presence of the correct drug substance, should be run prior to analysis. Aconsiderableamount of (unpublished) workhasbeenperformed by Ciurczak on counterfeit tablets. Using the same algorithms that are used to discriminate between placebos and active products, bogus products can be easily identified. The differences stem from the differences in raw materials and manufacturing processes, even though the active component may be present at the correct level.

E. PackagingMaterials As indicated elsewhere in this volume, polymers have been analyzed by NIR techniquesforsometime. In 1985,Shintani-YoungandCiurczak [35] used discriminant analysis to identify polymeric materials used in packaging: plastic bottles, blister packaging, and PVC wrap, to name a few. The replacement of the time-intensive IR technique, by using Attenuated Total Reflectance (ATR) cells, was considered quite good. Information such as density, cross-linking, and crystallinity can be measured.

F. Polymorphism When organic (drug) molecules crystallize from a solvent, the crystal structureis dependent upon the speed of crystallization, temperature, polarity of the solvent, concentration of the material, and other factors. Because the energy of the crystal affects the (physiological) rateof dissolution and thus the potency and activityof

Applications Pharmaceutical

of NIR

353

the drug, polymorphism is important [36]. The most common tool to determine crystal form is differential scanning calorimetry (DSC). Unfortunately, DSC uses small samples and may not represent the bulk of the material. X-ray diffraction is anotherexcellenttechnique but is quiteslowand is sometimes difficult to interpret. In 1985, Ciurczak [37] reported on the use of NIR to distinguish between the polymorphic forms of caffeine. This technique has been applied to proprietary drug substances, but data are unavailable for public presentation. Polymorphism was also reported upon by Gimet and Luong [38] in 1987. They found NIR a useful tool to ascertain whether the processing of a granulation led to any crystallinity changes of the active material. It has beennotedthat physical processes such as the milling, wet granulation, or compression of tablets can cause shifts in the crystal structure of a drug substance. Aldridge et al. 139 used pattern recognition to differentiate between the desired and unwanted polymorphs of an active substance. More important, the method was transferred to at least six other instruments for application. The polymorphic form of the drug substance was studied by DeBraekeleer et al. [40] in 1998. They used PCA, SIMPLISMA, and orthogonal projections to correct for temperature variation during the monitoring of polymorph conversion. This is performed in real time, on-line in a commercial process.

G.

Optical Isomers

In a presentation by Ciurczak [41], it was observed that pure d- and I-amino acids gave identical NIR spectra, whereas racemic crystals generated quite different spectra. A paper presented by Ciurczak [42] in 1986 outlines work later completed by Buchanan et al. [43]. In this work, varying percentages of d- and /-valine were mixed physically and scanned by NIR.The spectra were identical except for particle size induced baseline shifts. These mixtures were then dissolved and recrystallized as racemic crystals. The new samples were scanned by NIR;obvious qualitative and quantitative differences were observed. Mustillo and Ciurczak [44] presented a paper discussing the spectral effect of opticallyactivesolventsonenantiomers.Thisinformationwaslaterused to screenforpolarmodifiers in normalphasechromatographicsystems that included racemic mixtures [45].

H.

Structural Isomers

Structural or geometrical isomers can be distinguished by NIR. The xanthines (caffeine, theobromine, and theophylline) were discriminatedin a paper by Kradjel and Ciurczak [46]. In that same presentation, ephedrine and pseudoephedrine were shown to have different spectra. The differences in many of these cases is

Ciurczak and Drennen

354

one methyl group or the exchange of position of an H with an OH on the same carbon atom, demonstrating the power of NIR to perform identification within groups.

111.

QUANTITATIVEANALYSES

A.

Particle Size

It has long been recognized that the spectra of powdered samples are affected by their particle sizes [47,48]. The effect of particle size differences is usually seen as a sloping baseline in the spectrum. Many approaches had been suggested to circumvent this problem (screening, grinding) or to correct for it (using second derivative or multiplicative scatter correction software), but no worker used the phenomenon for quantitative purposes until 1985. In that year. Ciurczak [49] and Ciurczak et al. [50]presented work showing that there was a linear relationship between the absorbance at any wavelength and the reciprocal of the particle size. The calibration for the project was by laser low-angle scattering (LALS). Ilari et al. [51] used scatter correction in diffusely reflected NIR to determine the particle sizes of materials. Both organic and inorganic materials were determined by this technique. O’Neiletal. [52] measuredthecumulativeparticlesizedistributionof microcrystallinecellulosewithdiffusereflectanceNIR.BothMLRandPCA wereusedforthework.Theresultswereconsistentwiththoseobtainedby forward-angle laser light scattering.

B. Moisture Because the material with the greatest extinction coefficient in the NIR is water, it stands to reason that water is one of the substances most often measured by this technique. A more recent application is for (noninvasively) measuring water in freeze-dried samples. Derksen et al. [53],for instance, used NIR determine water through the moisture content of samples with varying active component contents. Warren et al. [54] described a technique for determining water in glycerides.Transmissionspectra of propyleneglycolandglycerinewereusedto calibrate and measure the water content. Correlation of total, bound, and surface water in raw materials was the subject of a paper by Torlini and Ciurczak[ S I . In that work, NIR was calibrated It wasseenthatthere by KF titration,DSC,andthermogravimetricanalysis. was a qualitative difference between “surface” and “bound’ water that could be seen by NIR but not by chemical or typical Loss on Drying (LOD) techniques. Thermal analysis methods were needed for calibration.

Pharmaceutical Applications of NIR

C.

355

Hardness

The hardness of a tablet depends onmany things; among them tableting pressure and particle size of the materials. In 1997, Morisseau and Rhodes 1561 published a paper describing their use of NIR to determine the hardness of tablets. Four formulations [two of hydrochlorothiazide (HCTZ) and two of chlorpheniramine (CTM)] and a placebo were prepared with hardness levels between 2 and 12 kg. Using MLR and PLS, equations were generated that allowed good prediction of hardness for all the products.

IV.

DETERMINATION OF ACTIVES IN TABLETS

In the earliest NIR assays, tablets and capsules were not analyzed intact. Before NIRspectralcollection,thedrugswereextractedfromthematrixintosolution. The first reported use of NIR tablets was by Sherken in 1968 [57]. In this study, meprobamate in tablet mixtures and commercially available preparations wasassayed.Twowavelengths,correspondingtothesymmetricalandasymmetricalstretchingmodesoftheprimaryaminegroup in thedrugmolecule, were used. Allen [ 5 8 ] usedNIR forthequantitativedeterminationofcarisoprodol (CAR), phenacetin (PH), and caffeine (CAF). Twenty tablets were pulverized, and an aliquot was dissolved in chloroform. Standard solutions of CA, PH, and CAF were scanned between 2750 and 3000 nm. CA and PH were determined at 2820 nm (CA) and 2910 nm (PH), with CAF determined at 3390 nm. The coefficient of variation (CV) was 1.4% or less. In 1977,ZappalaandPost[59]usedNIRformeprobamate (MEP) in fourpharmaceuticalpreparations:tablets,sustainedreleasecapsules,suspensions, and injectables. The NIR method was an improvement over that introduced by Sherken; i t took advantage of an MEP (primary amine) combination band at 1958 nm, which was not subject to the interference suffered by the peak at 2915 nm. Twenty tablets or capsules were pulverized, and an aliquot was dissolved in chloroform. Nine commercial products from four manufacturers were analyzed. The CV was 0.7% fortabletsand1.3%forcapsules (1.5% forthereference method). In 1990,Cortiet al. [60]usedanextractionprior to NIRanalysis to improvethedetectionlimit. Oral contraceptiveswereused in thestudy for ethinylestradiol (ETH) and norethisterone (NOR), two synthetic hormones. Qualitativeandquantitativeanalysesweredesired.Eighty-milligramtablets(containing 0.05 mg ETH and 0.25 mg NOR) were extracted with chloroform and scanned.

Ciurczak and Drennen

356

Quantitatively, six wavelengths were used in a Mahalanobis distance calculation. and it was possible to distinguish the ETH extracts at concentrations below 0.05%. For quantitative analysis, multiple linear regression (MLR) was employed. The correlations obtained were r 2 = 0.85 for ETH and r 2 = 0.86 for NOR. With low drug concentrations and a small range of values, the SECs were high. in 1981 AnNIR techniqueforpharmaceuticalpowderswaspublished byBecconsalletal. [61]. Near-infraredandUVphotoacousticspectroscopy wereusedforpropranolol(PR)-magnesiumcarbonatemixtures. Spectra were collected at 1300-2600nmwithcarbonblackasthereference. An aromatic C-Hcombinationband at 2200 nmand an overtoneband at 1720 nmwere used to quantify PR. The UV data were nonlinear; the NIR data were linear. In 1982, Ciurczak and Torlini [62] published on the analysis of solid and liquid dosage form. They contrasted NIR techniques for developing calibrations for natural products versus those for pharmaceuticals. Samples prepared in the laboratoryarespectrallydifferentfromproductionsamples.Usingthemfor calibration can lead to unsatisfactory results; production samples are preferred for calibration. Near-infrared spectroscopy was compared with HPLC for speed and accuracy. The effect of milling the samples prior to analysis was also investigated. Twodosageformmatriceswerestudied:acaffeine(CAF) acetaminophen (APAP)mixtureandanAPAPmixture.APAPmixtureswereanalyzedafter milling,andCAF-APAPmixtureswereanalyzedwithandwithoutmilling. Multiple linear regression (MLR) was used for the calibration. Milling of the CAF APAP mixture improved the determinationof APAP, but that of CAF wasunchanged.Thedifferencebetweenthetheoreticaland predicteddatawas -0.2596, competitivewithHPLC.NIRhasrapidanalysis times with no costs for solvent purchase and disposal. In 1987,Chasseur[63]assayedcimetidine(CIM)granules.Batches of granuleswereprepared with CIM at 70-130%oflabel.Forcalibration, first andsecondderivativespectraandoneortwowavelengthswereincluded in themodel.Atwo-wavelengthmodelusingthe first derivativegavethebest results,with SEP = 1.75%. The SEE for theNIRwas2.73%and for UV it was 2.97%. A 1987 paper by Osborne [64] used NIR to determine nicotinamide (NIC) in vitamin premixes. HPLC, the reference method for NIC, required 3 days to analyze 36 samples;the NIR method required only 30 min. Twenty-five mixtures wereusedforcalibration,withconcentrations from zero to 6%. Spectra were collected between 1200 and 2400 nm. Second derivative spectra were calculated, and the calibration obtained the ratio of the second derivative values at 2138 nm (NIC) and 2070 nm (a spectral minimum). The SEP for the validation set was 0.56% w/w. HPLC and NIR gave comparable results.

+

+

pharmaceutical Applications of

NIR

357

In 1988, Ladder andHieftje [65] usedthe quantile-BEAST(bootstrap enor-adjusted single-sample technique) to assess powder blends. In their study, four benzoic acid derivatives and mixtures were analyzed. The active content varied between 0% and 25%. The individual benzoic acid derivatives were c k sified into clusters using the nonparametric standard deviations (SDs), analogous t o SDS in par:lmetric statistics. Acetylsalicylic acid was added to the formulations at concentrations of1-20%. All uncontaminated samples were correctly identified. Simulated solid dosage forms containing ratiosof the two polymorphs were prepared. They were scanned from I 100 to 2500 nm. c v s ranged from 0.1% to 0.9%. Near-infraredspectroscopywasused in 1989 to quantifyketoprofen in gel and powder matrices for encapsulation (Corti et al. [66]). TWO ranges were used: f5% of theory and 3-30% active. The SEP was approximately 2%, with no sample having an error greater than 3.5%. Corti et al. [67] analyzed ranitidine and water in tablets. Production samples have a variation in values of active concentration, allowing for a narrower range of sample values in the calibration set-a range of 4% (to cover a 10% [90-110% of label] range). Actual drug content of the samples was determined by HPLC, and water content by the Karl Fischer moisture analysis. For prediction drug content, three NIR calibrations using MLR were developed. Lab samples provided SEE and SEP values of 8.4% with unknown samples. The second calibration of production samples provided a SEP of 1% for production samples and 6.4% for lab samples. A third calibration, using both production and modified samples, gave SEP values of -1% for both. The optimum was a calibration range of -5%. The calibration for wateremployedproductionsamples both as is and modified. Both SEE and SEP were less than 0.1%. For production samples over a 1 year period, the NIR method had the greatest error for moisture, < 1%. As a qualitative test, it erroneously rejected samples with a moisture content greater than 2%. The results showed that, for products with little variability, a small number of samples (-10-20) is sufficient. One study in which NIR was found to be unsuitable was reported by Ryan et al. in 1991 [68]. The purpose was to find a rapid method for the verification of clinical packaging. Both MIR and NIR were used to identify two cholesterollowering drugs: lovastatin and simvastatin. Both MIR and NIR had a detection to differentiate the limit of “1% (w/w) for ground samples. NIRwasunable two drugs at low concentrations. In 1992,Cortiet al. [69]analyzedantibioticcornpounds by usingNIR spectroscopy. MLR was used for the quantification and M a h a h o b i s distances for qualitative analysis. Qualitative analysis (using Mahalanobis distances) differentiated 10 antibioticpreparations,includingthreetypes of ampicillinand blends of erythromycin powder and granules. The SEE and SEP for each Calibration were less than 2%.

358 Drennen

and

Ciurczak

A 1993 paper by Blanco et al. [70] addressed concerns of laboratory manipulation of production samples prior to analysis. Two commercial preparations of ascorbic acid (vitamin C) were analyzed: one granular product and one effervescent tablet. Five batches were used. All samples were ground to a specific mesh size (either 250 or 100 pm). To expand the calibration range, samples were diluted with the primary filler to analysis,threepreprocessingmethoroverdosedwithascorbicacid.Prior ods were evaluated: multiplicative scatter correction (MSC), signal scaling, and first derivativespectra.In this study, first derivativespectraprovidedthebest calibration results. I' values SMLRcalibrations usedup to fourwavelengthsandprovided of 20.99 and SEEandSEP values of