@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Graduate and Postdoctoral Studies"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Schulze, Georg"@en ; dcterms:issued "2009-03-17T18:58:52Z"@en, "1996"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """The measurement of neurotransmitter secretions by living cells, both in living organisms or in preparations, constitutes an enduring and vexing problem for neuroscientists due to the large number of substances involved at very low concentrations. An ability to correlate neurotransmitter secretions with various factors including organismic behavior would greatly advance our understanding of the organization and functioning of central nervous systems. This, in turn, has many important implications for the diagnosis and treatment of disorders of central nervous systems (mainly in humans) as well as for the design and implementation of information processing and control systems. The work presented here was undertaken in order to explore a novel approach to this demanding problem. The objective was to develop a probe capable of measuring neurotransmitter secretions in real time, at physiologically relevant concentrations, and non-invasively in situ. Data were obtained using an ultraviolet resonance Raman spectroscopic analytical technique performed via optical fibers, and were analyzed primarily with artificial neural networks. To this end, a prototype tunable ultraviolet resonance Raman system was designed, assembled, comissioned and employed. A general introduction to the problem and a discussion of existing techniques for neurotranmitter measurement are given in Part I. In Part n, the analytical method was shown to allow discrimination between several different neurotransmitters and some of their precursors, both on the basis of their spectra and the selective resonance enhancement of their spectra. Optical fibers were characterized with regard to their suitability for use with pulsed ultraviolet radiation in Part III and on the basis thereof selected for the construction of optical fiber probes. It was found that the performance of optical fibers varied greatly when subjected to pulsed ultraviolet radiation, making the selection of fibers a crucial factor in probe construction. Various design features influencing the efficiency of optical fiber probes were investigated using both theoretical and empirical techniques. A right-angle geometry using a small diameter excitation fiber and several larger collection fibers in close proximity produced the most efficient probe. In Part IV the use of cell secretions as samples modelling in vivo conditions were investigated. It was also shown that these probes could be inserted via surgically implanted cannulae into and operated in the crania of experimental male rats without producing discernable behavioral artifacts. In Part V some signal recovery methods were investigated and it was shown that artificial neural networks could be used to identify and quantify neurotransmitters based on their Raman spectra. Part VI contains an assessment of the neuroprobe using neurotransmitter secreting cultured cells as a model system. The thesis is concluded with a discussion of the charateristics of an ideal biosensor, reviews the work done, and highlights some future directions. This thesis represents my contributions toward the development of a tunable ultraviolet resonance Raman neurotransmitter probe. Within the scope of this work, limitations of the available equipment and other resources precluded the complete development of a high-performance neuroprobe, however, the data presented here demonstrate proof-of-concept and feasibility. In particular, what has hitherto been considered impossible - the use of optical fibers for pulsed ultraviolet remote resonance Raman spectroscopy - has been shown to be distinctly feasible. It has further been shown that ultraviolet resonance Raman spectroscopy is well-suited to the problem of resolving a mixture of neurotransmitters in a biological matrix. With the appropriate state-of-the-art equipment, there is now a very real possibilty of obtaining detection limits of lx10~9 M for the catecholamine neurotransmitters and 1x10"0" M for the aliphatic neurotransmitters with 30 s exposure time, thus providing a novel and general solution to the problem of neurotransmitter measurement."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/6171?expand=metadata"@en ; dcterms:extent "20198340 bytes"@en ; dc:format "application/pdf"@en ; skos:note "THE D E V E L O P M E N T OF A FIBER-OPTIC PROBE FOR THE IN VIVO R E S O N A N C E R A M A N SPECTROSCOPY OF NEUROTRANSMITTERS by Georg Schulze B. Sc. Eng. (Chemical), The University of Pretoria, 1980. M . A . (Psychology), The University of British Columbia, 1991. A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES I N T E R D I S C I P L I N A R Y S T U D I E S Biotechnology Laboratory Chemistry Psychology We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A October 1996 © H . Georg Schulze, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of pS-j C^tApgpj , CIA&IASU^) / V ^ K ^ * ^ The University of British Columbia Vancouver, Canada Date IS gcvV^r (?fQ> DE-6 (2/88) Abstract The measurement of neurotransmitter secretions by living cells, both in living organisms or in preparations, constitutes an enduring and vexing problem for neuroscientists due to the large number of substances involved at very low concentrations. An ability to correlate neurotransmitter secretions with various factors including organismic behavior would greatly advance our understanding of the organization and functioning of central nervous systems. This, in turn, has many important implications for the diagnosis and treatment of disorders of central nervous systems (mainly in humans) as well as for the design and implementation of information processing and control systems. The work presented here was undertaken in order to explore a novel approach to this demanding problem. The objective was to develop a probe capable of measuring neurotransmitter secretions in real time, at physiologically relevant concentrations, and non-invasively in situ. Data were obtained using an ultraviolet resonance Raman spectroscopic analytical technique performed via optical fibers, and were analyzed primarily with artificial neural networks. To this end, a prototype tunable ultraviolet resonance Raman system was designed, assembled, comissioned and employed. A general introduction to the problem and a discussion of existing techniques for neurotranmitter measurement are given in Part I. In Part n, the analytical method was shown to allow discrimination between several different neurotransmitters and some of their precursors, both on the basis of their spectra and the selective resonance enhancement of their spectra. Optical fibers were characterized with regard to their suitability for use with pulsed ultraviolet radiation in Part III and on the basis thereof selected for the construction of optical fiber probes. It was found that the performance of optical fibers varied greatly when subjected to pulsed ultraviolet radiation, making the selection of fibers a crucial factor in probe construction. Various design features influencing the efficiency of optical fiber probes were investigated using both theoretical and empirical techniques. A right-angle geometry using a small diameter excitation fiber and several larger collection fibers in close proximity produced the most efficient probe. In Part IV the use of cell secretions as samples modelling in vivo conditions were investigated. It was also shown that these probes could be inserted via surgically implanted cannulae into and operated in the crania of experimental male rats without producing discernable behavioral artifacts. In Part V some signal recovery methods were investigated and it was shown that artificial neural networks could be used to identify and quantify neurotransmitters based on their Raman spectra. Part VI contains an assessment of the neuroprobe using neurotransmitter secreting cultured cells as a model system. The thesis is concluded with a discussion of the charateristics of an ideal biosensor, reviews the work done, and highlights some future directions. This thesis represents my contributions toward the development of a tunable ultraviolet resonance Raman neurotransmitter probe. Within the scope of this work, limitations of the available equipment and other resources precluded the complete development of a high-performance neuroprobe, however, the data presented here demonstrate proof-of-concept and feasibility. In particular, what has hitherto been considered impossible - the use of optical fibers for pulsed ultraviolet remote resonance Raman spectroscopy - has been shown to be distinctly feasible. It has further been shown that ultraviolet resonance Raman spectroscopy is well-suited to the problem of resolving a mixture of neurotransmitters in a biological matrix. With the appropriate state-of-the-art equipment, there is now a very real possibilty of obtaining detection limits of lxl0~9 M for the catecholamine neurotransmitters and lxlO\" 0\" M for the aliphatic neurotransmitters with 30 s exposure time, thus providing a novel and general solution to the problem of neurotransmitter measurement. iv T A B L E OF CONTENTS Page Abstract i i List of tables xii List of figures xvi Acknowledgement xxvi Dedication xxvii P A R T I General Introduction C H A P T E R 1 l . l I N T R O D U C T I O N 1 1.2. O V E R V I E W O F B R A I N F U N C T I O N 2 1.3. O V E R V I E W O F B I O S E N S O R S 4 1.4. O V E R V I E W O F N E U R O T R A N S M I T T E R S E N S O R S 4 1.4.1. Non-selective sensors 5 1.4.1.1. Collection techniques 5 1.4.1.2. Spectrometric techniques 6 1.4.2. Selective sensors 7 1.4.2.1. Electrodes 8 1.4.2.2. Optrodes 8 1.5. T H E N E E D F O R A N E W N E U R O S E N S O R 9 1.5.1. The need for a fast technique 10 1.5.2. The need for a general technique 11 1.6. P R O P O S A L F O R A N E U R O S E N S O R 11 1.6.1. The problem 11 1.6.2. The conceptual solution 13 1.6.3. The approach 16 1.6.4. General principles and axioms 17 1.7. T H E S I S O U T L I N E 18 1.8. R E F E R E N C E S 19 PART II Spectroscopy Chapter 2 2.1. I N T R O D U C T I O N 23 2.2. N O R M A L R A M A N S P E C T R O S C O P Y 23 2.3. R E S O N A N C E A N D S U R F A C E E N H A N C E D R A M A N S P E C T R O S C O P Y 25 2.4. I N S T R U M E N T A T I O N 27 2.5. R E F E R E N C E S 29 V C H A P T E R 3 3 . 1 . I N T R O D U C T I O N 3 0 3 . 2 . I N S T R U M E N T A T I O N 3 1 3 . 2 . 1 . System requirements and characterization 3 1 3 . 2 . 2 . The detector 3 2 3 . 2 . 3 . The spectrograph 3 3 3 . 2 . 3 . 1 . The single monochromator 3 3 3 . 2 . 3 . 2 . The double monochromator 3 3 3 . 2 . 3 . 3 . The fiber-optic adaptor 3 4 3 . 2 . 4 . The fiber-optic probe and coupling optics 3 4 3 . 2 . 5 . The light source 3 5 3 . 2 . 5 . 1 . The second harmonic generator 3 5 3 . 2 . 5 . 2 . The dye laser 3 5 3 . 2 . 5 . 3 . The pump laser 3 7 3 . 2 . 5 . 4 . The Ar+ laser 3 8 3 . 2 . 6 . System throughput 3 8 3 . 2 . 7 . Synchronization 3 9 3 . 3 . M E T H O D O L O G Y 4 0 3 . 4 . D A T A A N A L Y S I S 4 1 3 . 5 . D I S C U S S I O N 4 2 3 . 6 . R E F E R E N C E S 4 3 C H A P T E R 4 4 . 1 . I N T R O D U C T I O N 4 4 4 . 2 . N O R M A L R A M A N S P E C T R A 4 5 4 . 2 . 1 . Neurotransmitter spectra and band assignments 4 5 4 . 2 . 1 . 1 . Physiological saline 4 6 4 . 2 . 1 . 2 . Acetylcholine 4 7 4 . 2 . 1 . 3 . Dopamine 4 8 4 . 2 . 1 . 4 . Epinephrine 4 9 4 . 2 . 1 . 5 . Norepinephrine 5 0 4 . 2 . 1 . 6 . Serotonin 5 1 4 . 2 . 1 . 7 . Histamine 5 2 4 . 2 . 1 . 8 . Aspartate 5 3 4 . 2 . 1 . 9 . y-Amino butyric acid 5 4 4 . 2 . 1 . 1 0 . Glutamate 5 5 4 . 2 . 1 . 1 1 . Glycine 5 6 4 . 2 . 2 . Other spectra 5 7 4 . 2 . 2 . 1 . Tryptophan 5 7 4 . 2 . 2 . 2 . Cerebrospinal fluid 5 7 4 . 2 . 2 . 3 . Anesthetic 5 8 4 . 2 . 3 . Difference spectra 5 9 4 . 2 . 3 . 1 . Acetylcholine in D M E M 5 9 4 . 2 . 3 . 2 . Acetylcholine in cerebrospinal fluid 6 0 4 . 3 . F I B E R - O P T I C P R O B E R A M A N S P E C T R A 6 1 4 . 3 . 1 . Neurotransmitter spectra 6 2 4 . 3 . 1 . 1 . Acetylcholine 6 2 4 . 3 . 1 . 2 . Dopamine 6 3 4 . 3 . 1 . 3 . Serotonin 6 3 vi 4 . 3 . 1 . 4 . y-Amino butyric acid 6 4 4 . 3 . 1 . 5 . Mixture spectra 6 5 4 . 4 . D I S C U S S I O N 6 6 4 . 5 . R E F E R E N C E S 6 8 C H A P T E R 5 5 . 1 . I N T R O D U C T I O N 6 9 5 . 2 . R E S O N A N C E R A M A N S P E C T R A 7 0 5 . 2 . 1 . Different concentrations of methyl orange 7 0 5 . 2 . 2 . Methyl orange at different visible frequencies 7 1 5 . 2 . 3 . Differentiation between methyl orange and chromate 7 2 5 . 3 . U V A B S O R P T I O N S P E C T R A 7 3 5 . 3 . 1 . Methyl orange 7 3 5 . 3 . 2 . Some aromatic neurotransmitters and amino acids 7 4 5 . 3 . 3 . Other neurotransmitters 7 5 5 . 3 . 4 . Cerebrospinal fluid 7 5 5 . 4 . U V R E S O N A N C E R A M A N S P E C T R A 7 6 5 . 3 . 2 . Neurotransmitters 7 7 5 . 4 . 2 . Other 8 0 5 . 5 . D I S C U S S I O N 8 1 5 . 6 . R E F E R E N C E S 8 4 PART I I I Fiber-optics C H A P T E R 6 6 . 1 . I N T R O D U C T I O N 8 5 6 . 2 . F I B E R - O P T I C S 8 6 6 . 2 . 1 . Light propagation 8 6 6 . 2 . 2 . Light coupling 8 7 6 . 2 . 3 . Light delivery efficiency 8 8 6 . 3 . R E M O T E R A M A N S P E C T R O S C O P Y 9 0 6 . 3 . 1 . In vivo Raman spectroscopy 9 0 6 . 4 . R E F E R E N C E S 9 1 C H A P T E R 7 7 . 1 . I N T R O D U C T I O N 9 3 7 . 2 . O P T I C A L F I B E R C H A R A C T E R I Z A T I O N 9 4 7 . 2 . 2 . Input coupling 9 6 7 . 2 . 1 . I N P U T C O U P L I N G 9 4 7 . 2 . 1 . 1 . Continuous wave excitation 9 4 7 . 2 . 1 . 2 . Pulsed excitation 9 4 7 . 2 . 2 . Optical damage 9 6 7 . 2 . 2 . 1 . Continuous wave light 9 7 7 . 2 . 2 . 2 . Pulsed light 9 7 7.2.3. Light transmission 100 7.2.3.1. Continuous wave excitation 100 7.2.3.2. Pulsed excitation 100 7.2.4. Beam profiles 107 7.3. D I S C U S S I O N 108 7.4. R E F E R E N C E S 111 C H A P T E R 8 8.1. I N T R O D U C T I O N 113 8.2. F I B E R - O P T I C P R O B E D E S I G N 114 8.2.1. Design considerations and simulations 114 8.2.1.1. Model parameters 115 8.2.1.2. Excitation and collection fiber sizes 117 8.2.1.3. Fiber tip modifications 118 8.2.1.4. Bundles 120 8.3. F I B E R - O P T I C P R O B E F A B R I C A T I O N 121 8.3.1. End-face preparation 121 8.3.2. Probe assembly 123 8.3.2.1. Dual fiber probes 123 8.3.2.2. Multi-fiber probes 124 8.3.3. Glue 128 8.4. F I B E R - O P T I C P R O B E E V A L U A T I O N 129 8.4.1. Working curves 129 8.4.1.2. Excitation and collection fiber sizes 129 8.4.1.3. Fiber tip modifications 132 8.4.1.4. Bundles 132 8.4.1.5. A side-casting bundle 135 8.5. D I S C U S S I O N 135 8.6. R E F E R E N C E S 139 PART I V Biopsychology C H A P T E R 9 9.1. I N T R O D U C T I O N 141 9.2. C U L T U R I N G P C 1 2 C E L L S 142 9.3. M E A S U R E M E N T O F N E U R O T R A N S M I T T E R S E C R E T I O N S 143 9.3.1. Cell depolarization 143 9.3.2. High performance liquid chromatography 143 9.3.3. Resonance Raman spectroscopy 144 9.4. D I S C U S S I O N 149 9.5. R E F E R E N C E S 150 C H A P T E R 10 10.1. I N T R O D U C T I O N 151 10.2. T H E B R A I N - P R O B E I N T E R F A C E 151 10.2.1. Interface assembly design 151 10.2.2. Placement coordinates 153 10.2.3. Cannulations 154 10.3. IN VIVO P R O B E E V A L U A T I O N 155 10.3.1. Rayleigh scattering in the lateral ventricle 155 10.3.2. Probe fouling 157 10.4. P O S S I B L E B I O L O G I C A L E F F E C T S O F IN VIVO S P E C T R O S C O P Y 159 10.4.1. In vivo pulsed U V probe operation 159 10.4.2. Motor deficits 160 10.4.3. Brain lesions 162 10.5. D I S C U S S I O N 162 10.6. R E F E R E N C E S 164 PART V Signal Processing C H A P T E R 11 11.1. I N T R O D U C T I O N 165 11.2. S I G N A L R E C O V E R Y 166 11.2.1. Noise removal 166 11.2.2. Deconvolution 168 11.2.3. Baseline removal 169 11.3. S I G N A L I D E N T I F I C A T I O N A N D Q U A N T I F I C A T I O N 170 11.3.1. Classical least squares 171 11.3.2. Artificial neural networks 171 11.3.2.1. Architecture 172 11.3.2.2. Layer computation 175 11.3.2.3. Node computation 178 11.3.2.4. Weight optimization 180 11.3.2.5. Biological neural networks 181 11.3.2.6. Modifications 183 11.3.3. Water as an internal standard 185 11.4. R E F E R E N C E S 186 C H A P T E R 12 12.1. I N T R O D U C T I O N 191 12.2. N O I S E R E M O V A L 191 12.2.1. Filtering 192 12.2.2. Autoaccumulation 193 12.2.3. Moving product processing 196 12.3. D E C O N V O L U T I O N 198 12.3.1. Maximum entropy 198 ix 12.4. B A S E L I N E R E M O V A L 199 12.5.1. Curve fitting 199 12.5. D I S C U S S I O N 201 12.6. R E F E R E N C E S 203 C H A P T E R 13 13.1. I N T R O D U C T I O N 205 13.2. S I G N A L C L A S S I F I C A T I O N W I T H N E U R A L N E T W O R K S 205 13.2.1. Layer sizes 208 13.2.2. Transfer functions 210 13.2.3. Generalization and discrimination 211 13.2.3. Input features 217 13.3. D I S C U S S I O N 218 13.4. R E F E R E N C E S 222 C H A P T E R 14 14.1. I N T R O D U C T I O N 224 14.2. M I X T U R E R E S O L U T I O N 225 14.2.1. Artificial neural networks 225 14.2.1.1. Training 226 14.2.1.2. Layer number and size 228 14.2.1.3. Transfer functions 230 14.2.1.4. Split layers 231 14.2.1.5. Local connections 232 14.2.2. Classical least squares 233 14.3. D I S C U S S I O N 236 14.4. R E F E R E N C E S 238 C H A P T E R 15 15.1. I N T R O D U C T I O N 240 15.2. Q U A N T I F I C A T I O N W I T H N E U R A L N E T W O R K S 241 15.2.1. Training 241 15.2.2. Spectral removal procedure 242 15.2.3. Binary mixtures 244 15.2.4. Tertiary mixtures 245 15.2.5. Measured mixture spectra 246 15.2.6. Removal fraction coefficient 248 15.2.7. Multi-component mixtures 250 15.2.8. Spectral shifts 251 15.3. D I S C U S S I O N 253 15.4. R E F E R E N C E S 255 P A R T V I Evaluation, Summary and Future Directions C H A P T E R 16 16.1. I N T R O D U C T I O N 257 16.2. A N A P P L I C A T I O N : T H E M E A S U R E M E N T O F C U L T U R E D C E L L S E C R E T I O N S 257 16.2.1. The samples 257 16.2.2. The fiber-optic probe 258 16.2.3. The neural network 259 16.2.4. The resonance Raman measurements 260 16.2.5. The results 261 16.2.5.1. H P L C results 261 16.2.5.2. Neuroprobe results 264 16.3. D I S C U S S I O N 265 16.4. R E F E R E N C E S 268 C H A P T E R 17 17.1. I N T R O D U C T I O N 269 17.2. I D E A L B I O S E N S O R C H A R A C T E R I S T I C S 270 17.2.1. Dynamic range 270 17.2.2. Linearity 271 17.2.3. Sensitivity 272 17.2.4. Dynamic response 272 17.2.5. Hysteresis 272 17.2.6. Stability 272 17.2.7. Calibration 272 17.2.8. Selectivity 274 17.2.9. Background signal 274 17.2.10. Noise characteristics 275 17.2.11. Limit of detection 276 17.2.12. Dependencies 276 17.2.13. Biocompatibility 277 17.2.14. Lifetime 277 17.3. F U T U R E W O R K 277 17.3.1. Efficiency improvements 277 17.3.1.1. Double monochromator filter 27 8 17.3.1.2. Frequency doubler 279 17.3.2. Filtering improvements 279 17.3.2.1. Liquid filters 280 17.3.2.2. Optical amplification 280 17.3.3. Digital signal processing 281 17.3.3.1. Neural networks 282 17.4. C O N C L U S I O N S 285 17.5. R E F E R E N C E S 286 xi A P P E N D I X A 2 8 8 Probe simulation program in Microsoft Quick Basic 4 . 0 A P P E N D I X B 2 9 3 Program to perform randomly displced autoaccumulations of a spectrum, written in Microsoft Quick Basic 4 . 0 . A P P E N D I X C 2 9 9 Program to amplify signals and reduce noise in a spectrum using a moving product method. Written in MicroSoft Quick Basic 4 . 0 . LIST OF T A B L E S Table 3.1 36 The average dye laser output energy (10 pulses) (a) as a function of the oscillator dye concentration and (b) as a function of time given a dye concentration of 276 mg/L. Table 4.1. 48 Table 4.1 shows some of the major peaks in the Raman spectrum of acetylcholine (0.5 M), their vibrational band assignments, and relative intensities. Table 4.2. 49 Some of the major peaks in the Raman spectrum of dopamine (0.5 M), their vibrational band assignments, and relative intensities. Table 4.3. 50 Table 4.3 shows some of the major peaks in the Raman spectrum of epinephrine (0.5 M), their vibrational band assignments, and relative intensities. Table 4.4. 51 Table 4.4 shows some of the major peaks in the Raman spectrum of norepinephrine (0.5 M), their vibrational band assignments, and relative intensities. Table 4.5. 52 Table 4.5 shows some of the major peaks in the Raman spectrum of serotonin (0.1 M), their vibrational band assignments, and relative intensities. Table 4.6. 53 Table 4.6 shows some of the major peaks in the Raman spectrum of histamine (0.5 M), their vibrational band assignments, and relative intensities. Table 4.7 54 Table 4.7 shows some of the major peaks in the Raman spectrum of aspartate (0.5 M), their vibrational band assignments, and relative intensities. Table 4.8. 55 Table 4.8 shows some of the major peaks in the Raman spectrum of y-amino butyric acid (0.5 M), their vibrational band assignments, and relative intensities. Table 4.9. 56 Table 4.9 shows some of the major peaks in the Raman spectrum of glutamate (0.1 M), their vibrational band assignments, and relative intensities. Table 4.10. 56 Table 4.10 shows some of the major peaks in the Raman spectrum of glycine (0.5 M), their vibrational band assignments, and relative intensities. Table 7.1. 96 The effects of placing the beam waist outside (A) or inside (B) the fiber. Excitation at 225 nm and 10 Hz. Table 7.2. 99 The effects of increasing pulse energies on the transmission through a 4x300 pm input bundle of 6 cm length. Excitation at 266 nm and 10 Hz. Table 7.3. 102 The effects of increasing pulse energies on the transmission efficiency (average of 10 pulses) through a 15 cm section of 300 pm diameter Fiberguide Industries fiber (Superguide G). Excitation at 266 nm and 10 Hz. Table 8.1. 117 The effects of varying excitation and collection fiber diameters (pm) on the relative collection efficiencies (%) of dual fiber probes for maximum probe depths of 500 pm. The separation distances between fibers were a function of their diameters. Table 8.2. 119 The effects of probe tip geometry on the relative collection efficiencies of dual fiber probes for maximum probe depths of 500 pm and in solutions with varying absorbances. The separation distances between fibers were 20 pm. Table 8.3. 120 The effects of probe geometry on the relative collection efficiencies of dual fiber probes for maximum probe depths of 500 pm. The separation distances between fibers were 20 pm and that between collection fiber endface and excitation fiber apex varied as indicated. Table 9.1. 145 The number of viable and non-viable PC 12 cells per hemocytometer counting area before depolarization. Table 9.2. 146 The concentration of neurotransmitters secreted by approximately 8.30x10^ cells into 1 ml depolarizing solution. Table 10.1. 161 The means, standard errors of the means, and probabilities (a = 0.05) of 3 different behavioral measures obtained on 4 consecutive days before and after in vivo operation of the U V resonance Raman neuroprobe. Table 12.1. 194 The characteristics of signal, noise, and approximate SNR from an acetylcholine spectrum autoaccumulated with varying bandwidths and using uniform or normal distributions. Abbreviations: N = normal distribution; U = uniform distribution; n = number of samples; sd = standard deviation. Table 12.2. 195 The characteristics of signal, noise, and approximate SNR from an acetylcholine spectrum processed with autoaccumulation and varying bandwidths of a zero-order Savitsky-Golay filter. Abbreviations: SG = Savitsky-Golay filter; others as in Table 12.1. Table 14.1. 227 This table shows the desired output node values for the neurotransmitter Raman spectra used in the training data set. Abbreviations: A C H , A = acetylcholine; DOP = dopamine; EPI, EP = epinephrine; NOR = norepinephrine; SER = serotonin; HIS = histamine; ASP = aspartate; G A B = y-amino butyric acid (GABA); G L U = glutamate; G L Y = glycine; and combinations of these. xiv Table 14.2. 228 This table shows the desired output node values for the neurotransmitter Raman spectra in the BASIC, NOISE, SH5, difference spectra, and INDEP testing data sets. See text for further explanation. Abbreviations as in the text and in Table 1; M I X = equimolar mixture of A C H , DOP, and EPI (the spectrum was not normalized and the table values reflect the relative Raman signal intensities for these three neurotransmitters). Table 14.3. 229 The test results of three artificial neural networks when given the BASIC testing set as input. The networks had different configurations but used only hyperbolic tangent transfer functions. Consult the text for further details. Tfunction = transfer function; RMS = root mean square error value. Table 14.4. 231 The test results of three artificial neural networks when given the BASIC testing set as input. The networks used different hidden layer transfer functions: sigmoidal (SGSG); hyperbolic tangent (TNSG); and sine (SNSG) transfer functions. Consult the text for further details. Tfunction = transfer function; RMS = root mean square error value. Table 14.5. 232 The test results of artificial neural networks when given the BASIC testing set as input. (a) The networks had configurations and used transfer functions as indicated in the table. (b) The table shows a detailed breakdown of the network performance (RMS error values) on the binary and ternary mixtures in the BASIC set plus two additional binary mixtures. Also shown are RMS error values based only on the average values of the training set (AVG). Consult the text for further details. Tfunction = transfer function (or network name as it appears in the text); RMS = root mean square error value. Table 14.6. 234 The mixture composition estimates of data in the BASIC testing set obtained with the classical least squares method (CLS) and a neural network (LCI) using local processing. Table 14.7. 235 The RMS values and composition estimates of the acetylcholine difference spectra and data in the INDEP testing set obtained with the classical least squares method (CLS) and a neural network (LC2) using local processing. Also shown are the RMS values for the NOISE (5 runs) and SH5 testing sets. Table 15.1. 245 Consecutive network estimates of the components present in the mixture and the highest estimate for components not present in the mixture (Other) for the A E P I G L Y spectrum and residual spectra after component removal. Also shown are spectral average values. Value of a used: 80%. Abbreviations: ach = acetylcholine; epi = epinephrine; gly = glycine; nor = norepinephrine; ser = serotonin; Resid. = Residual. Table 15.2. 246 Consecutive network estimates of the components present in the mixture and the highest estimate for components not present in the mixture (Other) for the A E P I G L Y spectrum and residual spectra after component removal. Also shown are spectral average values, a used: 100%. Abbreviations: ach = acetylcholine; epi = epinephrine; gly = glycine; nor = norepinephrine; ser = serotonin. XV Table 15.3. 247 Consecutive network estimates of largest component in mixture (Estimate), the fraction of that component removed from the mixture spectrum RF (RF = a x Estimate), the scaling factor SF used to renormalize the remaining mixture spectrum, and CFs obtained for every component recovered from the MIX2 spectrum. Value of a used: 50%. Abbreviations: ach = acetylcholine; dop = dopamine; epi = epinephrine. Table 15.4. 251 The table shows the network estimates for the three components present in the mixture of the MIX2 spectrum when the spectrum was shifted towards the excitation frequency by 0, 1, 2, 3, and 4 cm\" 1 . Table 16.1. 262 The concentration of neurotransmitters secreted by approximately 4.27x10\"\" cells into 0.5 ml depolarizing solution as determined by HPLC. Table 17.1. 276 The estimated improvement in detection limit that could be obtained with the use of different equipment or modifications to the existing system. * Table 17.2. 284 The results (after 2000 trials) of two networks with sine transfer functions trained by initial random assignment of either the weights (W) or the hidden layer values (H) followed by optimization with the simplex method. xv i LIST OF FIGURES Figure 1.1. 3 A schematic representation of a neuron. Figure 1.2. 14 A schematic representation of the in vivo Raman sensing process. It is for conceptual purposes only and not drawn to scale. In particular, the probe diameter is about 500 pm and the synaptic cleft about 50 nm. Figure 1.3. 18 Schematic diagram showing the organization of the thesis and the major research areas involved. Figure 2.1. 25 Energy level diagram showing the relationships between (a) IR absorption; (b) Rayleigh (R), Raman, and resonance Raman scattering; (c) fluorescence; and (d) resonance fluorescence. Figure 2.2. 26 Absorption and Raman spectra of a compound with two chromophores (A and B) showing how selective resonance enhancement of the chromophoric Raman bands can be obtained by the appropriate choice of the excitation frequency. Figure 3.1. 31 A schematic representation of the Raman spectroscopic system. Figure 3.2. 36 The output of the dye laser as a function of the oscillator voltage of the pump laser after optimization of the dye laser using coumarin 460. Figure 3.3. 37 The output energy from the pump laser at 532 nm as a function of the oscillator voltage setting: 500-700 V (dotted line) and 500-590 V Q-switched (solid line). Figure 3.4. 39 A schematic representation showing the throughput of the Raman spectroscopic system at various points. Figure 3.5. 40 A schematic representation of the optical path and electronic circuit time lags used to synchronize the optical and gating pulses. Figure 4.1. 46 The Raman spectrum of physiological saline solution. The prominent peak is due to water deformation. Figure 4.2. 47 The Raman spectrum of acetylcholine (0.5 M) in physiological saline. Inset shows structural formula. XVII Figure 4.3. 48 The Raman spectrum of dopamine (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.4. 49 The Raman spectrum of epinephrine (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.5. 50 The Raman spectrum of norepinephrine (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.6. 51 The Raman spectrum of serotonin (0.1 M) in physiological saline. Inset shows structural formula. Figure 4.7. 52 The Raman spectrum of histamine (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.8. 53 The Raman spectrum of aspartate (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.9. 54 The Raman spectrum of y-amino butyric acid (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.10. 55 The Raman spectrum of glutamate (0.1 M) in physiological saline. Inset shows structural formula. Figure 4.11. 56 The Raman spectrum of glycine (0.5 M) in physiological saline. Inset shows structural formula. Figure 4.12. 57 The Raman spectrum of tryptophan (0.1 M) in physiological saline. Figure 4.13. 58 The Raman spectrum of cerebrospinal fluid taken during surgery (solid line), 24 h later (dashed line), and 72 h later (dotted line) Traces offset for clarity. Figure 4.14. 59 The Raman spectrum of the anesthetic used during surgery. Figure 4.15. 60 The Raman spectrum of D M E M (solid top line), D M E M spiked with acetylcholine to 0.5 M (dotted line), and the difference spectrum (bottom line). Figure 4.16. 61 The Raman spectrum of cerebrospinal fluid taken during surgery (see Figure 4.13) spiked with acetylcholine to 0.2 M (top trace), and the difference spectrum (bottom trace). XVll l Figure 4.17. 62 The Raman spectrum of acetylcholine dissolved in physiological saline (0.5 M) obtained with a fiber-optic probe (baseline adjusted). Figure 4.18. 63 The Raman spectrum of dopamine dissolved in physiological saline (0.5 M) obtained with a fiber-optic probe (baseline adjusted). Figure 4.19. 64 The Raman spectrum of serotonin dissolved in physiological saline (0.1 M) obtained with a fiber-optic probe (baseline adjusted). Figure 4.20. 65 The Raman spectrum of y-amino butyric acid dissolved in physiological saline (0.5 M) obtained with a fiber-optic probe (baseline adjusted). Figure 4.21. 66 The Raman spectrum of an equimolar mixture of serotonin and acetylcholine dissolved in physiological saline (0.5 M) obtained with a fiber-optic probe (baseline adjusted). Figure 5.1. 71 Spectra of different methyl orange concentrations obtained at 514.7 nm: (a) 1x10\"^; (b) l x l 0 \" 4 ; (c) l x l O - 5 , and (d) l x l O \" 6 M . Figure 5.2. 72 Spectra of methyl orange (lxl0\"5 M) excited at different visible frequencies: (a) 488 nm; (b) 472.7 nm; and (c) 457.9 nm. Resonance enhancement occurred with excitation at 472.7 nm. The traces are offset for clarity. Figure 5.3. 73 Spectra of 20xl0\"6 M methyl orange (a) and 200x10\"\" M chromate (b) excited at 364 nm and 472.7 nm showing differential resonance enhancement. Figure 5.4. 74 The U V absorption spectra of 3.3xl0\"5 M dopamine (solid line), 3.8x10\"^ M epinephrine (dotted line), 108x10\" 5 M tryptophan (dashed line), and 110x10\" 5 M tyrosine (dash/dot). A l l analytes were dissolved in water. Figure 5.5. 75 A compilation of the U V absorption spectra of some of the 10 small-molecule neurotransmitters: (a) dopamine, epinephrine, norepinephrine, serotonin; (b) histamine; (c) glutamate; (d) aspartate; (e) acetylcholine; (f) glycine. Figure 5.6. 76 The U V absorption spectrum of cerebrospinal fluid diluted approximately 20 times with deionized water. Figure 5.7. 77 The resonance Raman spectrum of 1x10\" ^ M dopamine (dotted line) dissolved in distilled water obtained at 227 nm, 20 Hz, 90 s exposure time, a mononchromator setting of 235.3 nm and with = 17 mJ/pulse. The spectrum of the solvent only (solid line) and obtained under the same conditions. xix Figure 5.8. 79 The U V resonance Raman spectrum of 1x10\" 3 M epinephrine (dotted line) dissolved in distilled water measured at 227 nm, 20 Hz, 40 s exposure time, a mononchromator setting of 234.8 nm and with - 1 7 mJ/pulse. The spectrum of the solvent only taken under the same conditions (solid line). Figure 5.9. 79 The U V resonance Raman spectrum of 1x10\"^ M serotonin dissolved in distilled water obtained at 227 nm, 20 Hz, 90 s exposure time, a mononchromator setting of 235.3 nm and with ==17 mJ/pulse. Figure 5.10 80 The U V resonance Raman spectrum of 1x10\" 3 M histamine dissolved in distilled water obtained at 227 nm, 20 Hz, 90 s exposure time, a mononchromator setting of 234.3 nm and with ~ 20 mJ/pulse. Figure 5.11. 81 The U V resonance Raman spectrum of 50x10\"\" M melatonin (top), 820x10\"\" M tryptophan (center), 100x10\"^ M tyrosine dissolved in distilled water (solid line) obtained at 227 nm, 20 Hz, 40 s exposure time, a mononchromator setting of 234.8 nm and with - 1 8 mJ/pulse. The spectrum of water obtained under the same conditions (dotted line). Figure 5.13. 83 Lorentzian curves fitted to the measured spectra of dopamine, epinephrine and norepinephrine indicating peak positions. Figure 6.1. 88 A diagram showing the coupling critical half-angle a and the transmission critical angle (3 of an optical fiber. Figure 6.2. 89 Transmission curves of fused silica (a) and glass (b), redrawn from the Newport Catalog (1993). Figure 7.1. 95 The pulse energy (mean ± S.E.M.; n = 10) as a function of oscillator setting when measured at 266 nm and 10 Hz before (solid line) and after (dotted line) a 150 mm focal length lens. Figure 7.2. 101 The attenuation for a 15 cm section of 300 pm diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 100 pulses per point, as a function of time. The recovery exhibited by the fiber after a 5 minute rest period is also shown (last measurement). Figure 7.3. 102 The attenuation for a 1 m section of 600 pm diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 10 pulses per point as a function of the distance between the input face and the beam waist. XX Figure 7.4. 103 The attenuation for a 15 cm section of 300 pm diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 100 pulses per point as a function of time. The fiber tip extended freely either 2 mm (top) or 6 mm (bottom) from its support. In neither case was the dynamic attenuation affected. Figure 7.5. 104 The attenuation (mean ± S.E.M.; n = 10) for 15 cm sections of 300 pm fused silica fibers from Polymicro Technologies (dotted line), 3 M (dashed line), Fiberguide Industries (dash/dot), and the new experimental fiber from Polymicro Technologies (200 pm, solid line) at 230 nm and 10 Hz as a function of time. The last point in each of the two top traces were obtained after a 5-minute recovery period. Figure 7.6. 105 The transmission efficiency (mean ± S.E.M.; n = 10) of an experimental fiber (20 cm section, 200 pm) as a function of recovery time at 230 nm and 10 Hz. Figure 7.7. 106 The transmission efficiency (dotted line) and output pulse energies (solid line) of an experimental fiber (20 cm section, 200 pm) as a function of input energy at 230 nm and 10 Hz (mean ± S.E.M.; n = 10) Figure 7.8. 107 The output energy as a function of time for fibers tested under 3 conditions: increasing the input energy with small increments (e.g. oscillator voltage from 640-646; solid line); with a moderate increment (590-615 V; dashed line), and with a large increment (590-639 V ; dotted line) to the final operating level. Data were scaled to the same initial values. Figure 7.9. 108 The central section emission beam profile of a 300 pm fused silica fiber with a Gaussian fit superimposed for comparison. Figure 8.1. 116 The simulated intensity distribution of light emerging from the excitation fiber of a front-casting probe. Figure 8.2. 118 The simulated intensity distribution of light emerging from the excitation fiber of a side-casting probe. Figure 8.3. 122 Two 300 pm fibers with polished enfaces before being assembled into a probe. Figure 8.4. 123 A silastic collar being stretched with fine wire to accommodate two optical fibers. Figure 8.5. 124 The aligned endfaces of two optical fibers of 300 pm diameter being held in place by the silastic collar. Figure 8.6. 125 End view of an optical fiber probe consisting of a central collection fiber (300 pm) surrounded by 8 angled excitation fibers (200 pm each). x x i Figure 8.7. 126 Side view of the optical fiber probe consisting of a central collection fiber surrounded by 8 angled excitation fibers shown in Figure 8.6. Figure 8.8. 127 End view of the input bundle of the optical fiber probe consisting of a central collection fiber surrounded by 8 angled excitation fibers (200 pm each) shown in Figure 8.6. The single lighter fiber functions as a spacer to improve alignment of the fibers inside the glass collar. Figure 8.9. 128 A dual front-casting probe glued beyond the silastic collar with a suspension of silicone rubber, cyclohexane, and tetradecane, to produce a thin film of glue. Figure 8.10. 130 The working curves of 3 probes with 300 pm excitation fiber diameters and 300 (dotted line), 600 (dashed line), and 1000 pm (solid line), collection fiber diameters, respectively. The working curves were normalized with respect to output power. Figure 8.11. 131 The working curves of two probes with the same collection fiber diameter (1000 pm) and varying excitation fiber diameters, 300 pm (dotted line) and 1000 pm (solid line), respectively, normalized with respect to output power. Figure 8.12. 132 The working curves of a 200x200 front-casting probe (dotted line) and a 200x200 side-casting probe (solid line) with aluminum pigmented acrylic paint applied to the angled facet of the excitation fiber to provide a reflective surface. The working curves were normalized with respect to output power. Figure 8.13. 133 A diagrammatic cross-section of a probe consisting of an input bundle of 10 excitation fibers of 100 (im diameter each and a 400 p.m collection fiber. Some excitation fibers were misaligned. Figure 8.14. 133 A probe consisting of a 300 pm to 100 pm tapered excitation fiber surrounded by 4 collection fibers of 200 pm diameter each. Figure 8.15. 134 The working curves of the 2 probes shown in Figures 8.13 (input bundle, dotted line) and 8.14 (tapered input, solid line), respectively. The working curves were normalized with respect to output power. Figure 8.16. 136 The working curve (closed symbols) and detailed simulated working curve (open symbols) of a probe with 300 pm excitation fiber diameter and 600 pm collection fiber diameter. The working curves were normalized with respect to output power. Figure 8.17. 137 The working curve for a side-collection probe using pulsed ultraviolet resonance Raman spectra of tryptophan dissolved in water excited at 227 nm, 20 HZ, and 300 pW average power. The tryptophan signal (solid line) shows a maximum near 80 p M while the water signal (dotted line) declines monotonically. Figure 9.1. 143 Secretions of neurotransmitters (dopamine, top; serotonin, center solid line; epinephrine, dotted line; norpeinephrine, dashed line) from rat PC 12 cells, measured with an HPLC, as a function of time. Figure 9.2. 147 H P L C trace of neurotransmitter secretions from rat PC 12 cells taken at 30 minutes after depolarization. Sample A (1.5 pi, solid line) was filtered with a 0.2 pm filter and sample B (3 pi, dotted line) with an ultrafilter. Figure 9.3. 147 H P L C trace of neurotransmitter secretions from rat PCI2 cells taken at 30 minutes after depolarization, as in Figure 9.2, but enlarged 4 times. Figure 9.4. 148 Resonance Raman spectra of neurotransmitter secretions from rat PC12 cells taken after 30 minutes of depolarization (dotted line) and depolarizing agent (solid line, scaled down by a factor of 2 for comparison). Spectra were obtained with 227 nm excitation. Figure 9.5. 149 Resonance Raman spectra of neurotransmitter secretions from rat PC12 cells taken after 30 minutes of depolarization: 8 spectra obtained with 227 nm excitation added together (top line); 3 spectra obtained with 218 nm excitation added together (bottom line). Backrounds were removed. Figure 10.1. 152 A schematic diagram of the protective sleeve designed for inserting the fiber-optic probe into the centrally implanted guide-cannula. Figure 10.2. 154 A digitized histological slide from a trial implantation into the lateral ventricle of a male rat showing the cannula tract entering the left lateral ventricle. The figure was digitally enhanced to demonstrate the relative scales of the cannula tract and the ventricles. Figure 10.3. 156 An anesthetized rat with fiber-optic probe inserted into the guide-cannula. Figure 10.4. 157 Ray lei gh scattering obtained in vivo from an anesthetized rat with 473 nm excitation and different monochromator settings: 896 (top), 1080 (middle), and 1187 cm~l (bottom). Figure 10.5. 158 Two spectra obtained in vivo with 473 nm excitation and the same monochromator setting: (top) initial spectrum and (bottom) 25 minutes later. Figure 10.6. 160 Spectra obtained in vivo from an anesthetized rat with pulsed 230 nm excitation and 3 different monochromator settings after processing: 1000 cm~l (top), 2000 cm\"! (middle), and 3000 c m - l (bottom). Spectra are vertically displaced for ease of viewing. Figure 11.1. 173 This figure shows an artificial neural network with 250 input nodes, 30 hidden nodes, and 10 output nodes. Each output node corresponds to one of 10 neurotransmitters. Also shown are data from a dopamine spectrum being given as input to the network and a network output classifying the input as dopamine. Figure 11.2. 179 This figure shows (from left to right) the computations performed at hidden layer node 1. The data transferred from each of the 4 input nodes, multiplied by the respective connection weights (in small rectangles), are combined into a single value at this hidden node. This value is passed through the transfer function F(x) to determine the hidden node's output. As before, this output is multiplied with the connection weights (in small rectangles) and transferred to the 3 output nodes. Figure 11.3. 182 A simplified rendition of the primate retina showing a 3-layered structure similar to the backpropagation artificial neural network discussed in the text. The input layer consists of the rod receptor cells, the hidden layer consists of the bipolar cells, and the output layer of the ganglion cells. The axons of the ganglion cells form the optic nerve. Figure 12.1. - 192 The effects of filtering with a zero-order Savitsky-Golay filter with 10 (dotted line) and 50 (dashed line) point spread on a section of dopamine Raman spectrum. Figure 12.2. 195 The effects of 50-point normally (dotted line) and uniformly (dashed line) displaced autoaccumulations on an acetylcholine spectrum. Figure 12.3. 197 The effects of using a moving product to smooth a spectrum with SNR of 1. The Gaussian signal after Gaussian noise addition (top trace). The effects of smoothing with a 6 (dotted line), 20 (dashed line) and 25-point (solid line) moving product on the noisy spectrum. Original signal (before noise addition) shown at the bottom. Spectra were normalized (using factors of 4, 300, and 10,000 respectively) and offset for comparison. Figure 12.4. 200 First stage in the baseline correction of a dopamine spectrum showing a polynomial fit (dashed line) and the removed outliers (Raman signals, dotted lines). The baseline-corrected spectrum is shown at the bottom. Figure 12.5. 201 The effects of polynomial order on baseline fitting, displaced vertically for ease of viewing: 3rd order (top), 5th order (middle), and 6th order (bottom). Figurel2.6. 202 The recovered signal (solid line) from a 0.01 M acetylcholine in physiological saline Raman spectrum (dotted line) after a 3rd order polynomial background removal and 1000 autoaccumulations. Figure 13.1. 207 Typical learning curves for neural networks using sine transfer functions (solid line) and sigmoid transfer functions (dotted line). xxiv Figure 13.2. 209 Performance of neural networks with the same ratio of hidden to input layer nodes but different numbers of input layer nodes (IN136: 136 input nodes, IN150: 150 input nodes, IN250: 250 input nodes). Testing sets consisted of the baseline-corrected Raman spectra of neurotransmitters manipulated as follows: smoothed with a 10-point moving average (SM10); smoothed with a 5-point moving average and also used for training (SM5); not processed (RAW); shifted towards the exciting frequency by 5 cm\"l (SH5); and shifted towards the exciting frequency by 10 cnr* (SH10). Figure 13.3. 210 Performance of neural networks with the same number of input nodes but different ratios of hidden to input nodes: ratio = 0.10 (RATIO10); ratio = 0.16 (RATI016); and ratio = 0.25 (RATI025). Testing sets as for Figure 13.2. Figure 13.4. 212 This figure shows the consequences of a relaxation of the training criterion on SGSG networks from a low root mean square (RMS) error value (SGSGL) to a higher RMS error value (SGSGH). The performance of a SNSG network with the more relaxed criterion (SNSGH) is shown for comparison. The discrimination index is defined in the text and the testing sets in the caption of Figure 13.2. SNSG networks use sine and sigmoid transfer functions for the hidden and output layers, respectively, and SGSG networks use only sigmoid transfer functions. Figure 13.5. 213 Performance of neural networks using different transfer functions: a sigmoid function (SGSG) and a sine function (SNSG). Testing sets are as given in Figure 13.2. Figure 13.6. 215 Abilities of two neural networks (SNSG, SGSG) with different transfer functions (sine, sigmoid) to discriminate between various spectra: 5-hydroxytryptophan (TRP, not in the training set) and serotonin (in the training set) as well as between epinephrine (ENP) and norepinephrine (NNP) and vice versa (both in the training set). Figure 13.7. 217 Discrimination index (see text) of two neural networks (SNSG, SGSG) with different transfer functions (sine, sigmoid) when presented with acetylcholine difference spectra (not included in the training set) obtained in cerebrospinal fluid (CSF) and Dulbecco's modified eagle medium (DMEM). Performance of neural networks using different transfer functions: sine (SNSG) or sigmoid (SGSG); and different input features: interval means (IM) or uniform sampling (US). Testing sets as for Figure 13.2. A schematic illustration of the discriminating abilities of ideal networks on hypothetical data sets. The solid line represents the performance of a discriminating network on noiseless data while the dotted line represents the performance of the same network on data with an arbitrary noise content. The dashed line represents the performance of a network with a better ability to generalize between data sets while the dotted line represents a network with a better ability to discriminate between data sets. The negative numbers on the abscissa represent an arbitrary increase in the loss of spectral detail (e.g. increasing the degree of smoothing) and the positive numbers an arbitrary increase in the loss of spectral peak positions (e.g. increasing the extent of spectral shifts). The '0' point Figure 13.8. 218 Figure 13.9. 221 XXV on the abscissa represents the spectral characteristics of the training data set. The discrimination index is given in arbitrary units and defined in the text. The figure is intended to show general trends only. Figure 15.1. 247 The Raman spectra of acetylcholine (a) and two experimentally measured mixtures, MIX1 (b) and MIX2 (c), both containing acetylcholine. In the mixture spectra, considerable overlap of the major component peaks occur. Figure 15.2. 249 The sum of all the recovered component fractions are plotted against the coefficient of the removal fraction used during spectral removal iterations. This graph illustrates that component fraction sums tend to converge to a minimum which is indicative of correct spectral recovery. The results shown are for the A E P I G L Y (broken line, left scale) and MIX2 (solid line, right scale) spectra. Figure 15.3. 250 The number of iterations required for mixture decomposition as a function of the a value used. The results shown are for the A E P I G L Y (broken line, right scale) and MIX2 (solid line, left scale) spectra. This figure indicates that the number of iterations required to resolve a mixture into its individual components is inversely proportional to the value of a used. Figure 16.1. 260 The U V resonance Raman spectra of cell secretion samples F (5 minute depolarization, top trace) and G (45 minute depolarization, bottom trace) obtained at 227 nm, 20 Hz, and approximately 400 pW employing a fiber-optic probe. Figure 16.2. 261 The U V resonance Raman difference spectrum of cell secretion samples F (5 minute depolarization) and G (45 minute depolarization) obtained at 227 nm, 20 Hz, and approximately 400 p.W employing a fiber-optic probe. Figure 16.3. 263 The full-scale chromatograms of cell secretion samples F (5 minute depolarization, dotted line) and G (45 minute depolarization, solid line). Figure 16.4. 263 The scale-expanded chromatograms of cell secretion samples F (5 minute depolarization, dotted line) and G (45 minute depolarization, solid line). Figure 16.5. 265 The average output values produced by the neural network for the difference spectrum of cell secretion samples F (5 minute depolarization) and G (45 minute depolarization). Negative values (epinephrine and serotonin) were set to zero. Figure 17.1. 271 The working curves for front-(dotted lines) and side-casting probes (solid lines) plotted on a linear scale to show the linear ranges of the two probes for resonant (positive slope) and non-resonant (negative slope) analytes in an absorbing medium. xxvi A C K N O W L E D G E M E N T S I wish to thank my comittee members - Drs. Michael Blades, Alan Bree, Boris Gorzalka, and Robin Turner - for having given me this opportunity to pursue an interdisciplinary project and for their encouragement and support. Special thanks to my colleague, Shane Greek, for his collaboration, support, and patience, in spite of the many hours spent with severely attenuated visibility below 550 nm. I have also received courteous and expert help from technical professionals in the mechanical, electrical and glassblower's workshops of the Department of Chemistry, from technician Chris Sherwood in the Biotechnology Laboratory, and Dr. Alina Kulpa, a scientific engineer in the Department of Electrical Engineering. I wish to thank Dr. Ariane Coury in the Department of Psychology for analyzing some of my samples with her H P L C and Dr. Karl Klein from Friedberg, Germany, for his shared insights into optical fiber transmission processes. Finally I acknowledge with appreciation the many graduate, and sometimes undergraduate students, in the various departmental laboratories where I spent research time, for their patience with a novice and their willingness to show me the ropes. Dedicated with love to the memory of Neil Lewis Andrews: 2 November 1958 - 21 November 1994. 1 P A R T I General Introduction C H A P T E R 1 1.1 I N T R O D U C T I O N 1.2. O V E R V I E W O F B R A I N F U N C T I O N 1.3. O V E R V I E W O F B I O S E N S O R S 1.4. N E U R O T R A N S M I T T E R S E N S O R S 1.4.1. Non- selective sensors 1.4.1.1. Collection techniques 1.4.1.2. Spectrometric techniques 1.4.2. Selective sensors 1.4.2.1. Electrodes 1.4.2.2. Optrodes 1.5. T H E N E E D F O R A N E W N E U R O S E N S O R 1.5.1. The need for a fast technique 1.5.2. The need for a general technique 1.6. P R O P O S A L F O R A N E U R O S E N S O R 1.6.1. The problem 1.6.2. The conceptual solution 1.6.3. The approach 1.6.4. General principles and axioms 1.7. T H E S I S O U T L I N E 1.8. R E F E R E N C E S 1.1 I N T R O D U C T I O N The fourfold requirement of performing rapid simultaneous measurements of many low concentration analytes in situ makes the development of an ideal sensor for investigating brain function an extremely difficult task. When to this requirement are added those of identification, quantification, biocompatibility, dynamic range, stability, ease of use, etc., the design of an ideal brain sensor becomes truly daunting. In spite of these considerable demands, the study of brain function is of such enormous importance to many disciplines that several techniques have been developed to investigate brain activity. Although these techniques have often lead to great advances in our understanding of the brain and its operation, their many limitations have denied researchers the opportunity to investigate numerous important aspects of brain activity effectively. The functioning of the normal and diseased brain therefore remains largely uncharted territory. 2 The aim of the interdisciplinary work undertaken and reported in this thesis was to conceptualize, design, and lay the foundations for the development of a brain sensor capable of meeting the general requirements of sensitivity, speed, and versatility while operating non-invasively in an aqueous in situ environment. In this Chapter, brief overviews of general brain function, biosensors, and brain sensors are given. In addition, the need for and advantages of a new brain sensor are described, the central problem formulated, and the approach taken to solve this problem sketched. The chapter concludes with an outline of the thesis organization indicating the important areas of research involved in this work. 1.2. O V E R V I E W O F B R A I N F U N C T I O N Brain function generally depends on two factors: the organization of brain cells and the operation of these cells. One of the most important classes of brain cells are nerve cells or neurons which operate by the propagation of intercellular electrical and chemical signals. Due to the electrochemical character of nerve signals, investigative techniques aimed at the elucidation of brain function typically detect electromagnetic (e.g. electroencephalograph) or chemical (e.g. in vivo dialysis) events associated with neuronal operation. In order to limit the scope of inquiry, only chemical sensing will be considered here. A neuron, shown in Figure 1.1, consists of three parts: the cell body or soma, an input region of dendritic protrusions from the soma, and a single output conduit, the axon. The axon terminals abut on the somata and dendrites of other cells but for a minute gap (about 50 nm), the synaptic cleft. 3 Dendrites Soma Axon Terminal buttons Figure 1.1. A schematic representation of a neuron. Neurons are bathed in extracellular fluid rich in ions, primarily N a + and CI\". The presence of a selectively permeable cell membrane, intracellular protein anions, and an ion pump, result in the generation of a resting membrane potential of about -70 mV due to osmotic and electrostatic gradients and the activity of a transmembrane ion pump. Ion channels situated in the membrane can be gated electrically or molecularly to allow the free movement of N a + into the cell followed by the movement of K + out of the cell across the membrane. These ion fluxes locally change the membrane potential and signals are transmitted within a cell by a sequence of such changes. When potential changes are transmitted down the axon they are called action potentials and cause the release of gating molecules (neurotransmitters) from the axon terminals which then diffuse across the synaptic cleft and act on the ion channels of neighboring cells. Such neurotransmission causes fluxes of cations and anions in the neighboring cell which produce excitatory and inhibitory postsynaptic potentials. If a summation of 4 these potentials exceeds a cellular threshold, an action potential is triggered at the axon near the soma and propagated toward the axon terminals. Neurons are often highly organized into structures in many brain areas. The cortex, for instance, clearly consists of six layers (Kelly, 1985). The cerebellum, which plays an important role in motor control, shows exquisite organization in the arrangement of Purkinje cells and lateral fibers (e.g. Pellionisz and Llinas, 1979). In the tectum, input neurons terminate in a highly organized manner to form topographical maps of various sensory modalities (see Camhi, 1984). 1.3. O V E R V I E W O F B I O S E N S O R S Hall (1992) defines a biosensor as consisting of an analyte selective interface that relays the interaction between the surface and the analyte via some transducer to a detection system. This definition makes no reference to biology. In contrast, Buerk (1993) defines a biosensor as any measuring device that contains a biological (sensing) element. A third option is to define a biosensor as any measuring device that has been specifically designed or adapted to perform measurements in biological systems and this definition will be employed here. Sensors are often categorized on the basis of their transducers into electrochemical (amperometric, conductimetric, and potentiometric) and electromagnetic (optical and mass) sensors, but can also be categorized on the basis of their application (glucose sensors, pH sensors, etc.). Brain sensors, hence termed neurotransmitter sensors or neurosensors, can be seen as an application-specific subcategory of the set of biosensors. 1.4. O V E R V I E W O F N E U R O T R A N S M I T T E R S E N S O R S Chemical neurosensors can be divided into two broad groups: non-selective and selective sensors. 5 1.4.1. Non-selective sensors 1.4.1.1. Collection techniques Ventricular perfusion uses artificial cerebrospinal fluid to perfuse the ventricular system of an experimental animal. This can be done while the animal is unanesthetized and freely moving (Pappenheimer et al., 1962). Samples can be taken at specific intervals and analyzed. It is impossible to be sure of the precise site of release of the neurotransmitters identified. Using recording electrodes in conjunction with this technique may allow for the identification of some of the sites where neurotransmitters were released (Obata and Takeda, 1969). In vivo dialysis can overcome this problem to some extent, but suffers from another. This method involves the intracranial implantation of a probe into a specific site. A selectively permeable membrane in this probe allows certain neurotransmitters to diffuse upon release into a minute carrier stream which is collected at regular intervals (e.g. 12 minutes) before analysis with a high performance liquid chromatograph (Ungerstedt, 1986; Ungerstedt, 1984). This long time interval severely complicates the correlation of behavior with neurotransmitter release. The time spent in transport may also allow the breakdown of neurotransmitters and render the temporal resolution of transmitters released in close temporal contiguity to one another difficult. Another problem is that the selectivity of the membrane limits the number of neurotransmitters that can be detected. This technique is also rather invasive. Suck-blow micropipettes are closed systems where material is collected during the period when a cell actively secretes a substance. This is then later released at the same position to determine the effects on the cell. Electrodes incorporated into the pipette are used for this purpose. The minute amounts of neurotransmitter involved can be collected and pooled for analysis. Push-pull cannulae are implanted cannulae allowing the withdrawal or application of substances centrally. A disadvantage is the fact that tissue damage is bound to occur and that collected material could be drawn from large 6 numbers of perhaps heterogeneous types of neurons (see Mitchell, 1975 for a discussion of these methods). A definite disadvantage to all the above techniques is the time required for analysis. During the period between collection and analysis, the compounds in the sample may naturally degrade or may be degraded by other substances in the same sample. 1.4.1.2. Spectrometric techniques. Nuclear magnetic resonance (NMR) spectroscopy is based on the absorption of radio frequency radiation by analytes subjected to an intense magnetic field (e.g. Skoog, 1985). N M R , employing surface coils, can be used to detect the formation of metabolites from radiolabeled precursor molecules introduced into living organisms (den Hollander et al., 1984; Tungsal et al., 1990). Although there is a fairly wide range of compounds that can be investigated with this method, it has several disadvantages. These include the lack of fine spatial localization and potential heating effects of the sample being investigated (Hammer et al., 1990), the difficulty in differentiating among compounds with a common precursor (Tungsal et al., 1990), the low signal-to-noise ratio (Sekihara and Ohyama, 1990), and the need to inject radio-active substances. Surface-enhanced Raman spectroscopy (see Chapter 2) has been applied to the analysis of neurotransmitters (McGlashen et al., 1990; Morris et al., 1990) and peptides (Garrell et al., 1990). It is capable of detecting the biological amines at stimulated release concentrations (Morris et al., 1990). At present, some problems exist with the electrode coating to prevent protein fouling: thick coatings are more robust but increase response times, while thinner coatings have faster response times but are more fragile (Morris et al., 1990). Detection is dependent on the adsorption of neurotransmitters on metal surfaces and the adsorption kinetics (whether fast or slow adsorption occurs) as well as the degree of adsorption and the intensity of spectra generated by adsorbates would have 7 a bearing on the usefulness of this approach. This method, however, may have the potential for the development of a fast, general in vivo neurotransmitter probe. Infrared absorption (IR) spectroscopy is based on the absorption of infrared radiation. Organic molecules produce complex spectra and IR spectroscopy is primarily used for their identification (e.g. Skoog, 1985). In vivo near-IR spectroscopy may also have the potential for a general method but appears to be currently used mainly to determine tissue and blood oxygenation through the detection of hemoglobin (Dickensheets and Cheung, 1989; Thorniley et al., 1988) and cytochrome c oxidase (Thorniley et al., 1988). In this method, near-infrared radiation is relayed to the sample (e.g. the patient's head) by optic fiber, passed through the sample (e.g. dermis, cranium, meninges etc.) and, upon emergence from the sample, collected and relayed to the sensors by optic fiber (Rea et al., 1985). Since tissue is radiated, the effective optical pathlength has to be estimated (Delpy et al., 1988). This technique, as currently used, has poor spatial resolution and does not appear to have been extended to investigate a wider range of substances. In addition, IR methods have poor detection limits. 1.4.2. Selective sensors. Selective techniques are highly sensitive to a particular molecular species. This quality is generally gained by designing the measuring device so as to interact selectively with a particular type of molecule. Such an interaction then provides, either directly or indirectly, a signal indicating the presence of the analyte. Selectivity can be accomplished through a variety of ways: specific molecules can be 'selected' on the basis of their physical properties (e.g. size - permeable membranes and molecular sieves, natural fluorescence, light scattering, etc.) or on the basis of their chemical properties (e.g. interactions with antibodies, specific redox reactions, specific enzymatic reactions, etc.) or a combination of these methods. Many of the selective techniques combine fiber optics and spectrometry. 8 1.4.2.1. Electrodes. In vivo voltammetry is a technique selectively sensitive to several of the monoamines (Blaha and Lane, 1983). This technique uses a chemically modified graphite paste electrode. Currents recorded are those arising as a result of monoamine oxidation subsequent to their release. The use of electrode arrays can produce very low detection limits (Monita and Niwa, 1996). However, should a cell release other substances, those remain unidentified. Furthermore, it is difficult to differentiate between the monoamines and/or their metabolites, e.g. dopamine can only be discriminated from norepinephrine after pharmacological pretreatment to suppress interfering metabolites (see Gonzalez-Mora etal., 1991). There are two types of ion-selective microelectrodes: ion-selective glass microelectrodes and ion-specific liquid ion-exchanger electrodes. They measure ionic activity and not concentration. The ionic activity coefficient is also not known in and around living cells and has to be estimated. Other ions, intracellular diffusion barriers and compartmentalization could further complicate such measurements. Although these electrodes allow for more specific measurements, potentials recorded are still a summation of all potentials occurring between the microelectrode and the reference electrode and transmitter identification is problematic (for a discussion of these recording methods see Kelly et al., 1975). Iontophoresis is related to the above techniques and is essentially used to record potentials resulting from the intracranial administration of neuroactive substances (Kelly et a l , 1975). Such administration can be controlled with great accuracy using small electrical currents, but a major problem remains the identification of the actual neurotransmitters released and the temporal aspects of their release. 9 1.4.2.2. Optrodes. Biological molecules with natural fluorescence can be detected with fluorescence spectroscopy which is much more sensitive and more selective than absorption spectroscopy (Bright, 1988). Optical fiber is used to relay exciting incident radiation and to collect emitted radiation (Kulp et al., 1988). Non-fluorescent species can frequently be fluorescently labeled (Bright, 1989) or transformed into a fluorescent molecule by chemical reaction which would render these substances detectable. Such modification of non-fluorescent molecules has to occur in vivo and is accomplished through the use of an immobilized indicator phase on the sensing surface of the optic fiber. The analyte is made fluorescent when it combines with a reagent adsorbed on the indicator phase. The same general principle is used to construct optrodes that employ immune (Tromberg et al., 1988) and enzymatic (Kulp et al., 1988; Pantano and Kuhr, 1995) reactions to indicate the presence of analyte. In the latter cases, chemical reaction is not aimed at generating a fluorescent molecule, but to modify incident radiation to indicate the presence of the analyte (e.g.evanescent wave attenuation). The reader is referred to several reviews of this technology (Schultz, 1985; Seitz, 1989; Seitz, 1984; Sepaniak et al., 1988). The very selectivity of these methods also constitutes a limitation. Although some of these fiber optic probes can be very small, it is perhaps not feasible to use more than a few of them, each sensitive to a different analyte, in close proximity in vivo. Furthermore, some of these methods cannot yield real time results due to the duration of diffusion and reaction processes (Sepaniak et al., 1988). . 1.5. T H E N E E D F O R A N E W N E U R O S E N S O R . Ideally, an investigative tool would be able to indicate which events are occurring and when they are occurring. That is, real time identification of neuronal events is desirable. At present, no technique can accomplish this. Existing techniques can generally be divided into two groups. In the first group are those techniques capable of 10 detecting and identifying a wide variety of neurotransmitters (e.g. in vivo dialysis). These techniques are generally slow, which is one of their major drawbacks. In the second group are techniques that are generally faster, but limited to the detection of single compounds or families of compounds (in vivo voltammetry). Their inherent specificity thus also constitutes an important deficit. A brief discussion of these two groups of techniques has been given above. There does not seem to exist an in vivo technique that can be used with unanesthetized, freely moving animals to allow the rapid identification and quantification of all organic substances released in the process of neuronal signaling occurring at a specific central site. This view is in agreement with that of Morris et al. (1990). These authors concluded that there was a need for a general (amine) neurotransmitter probe providing resting-level sensitivity and real-time response; usable in vivo, with brain slices, extracts, and dialysates; and that was easy to fabricate and operate. 1.5.1. The need for a fast technique. Many behaviors occur at timescales in the millisecond range. For instance, the lordosis reflex of the female rat (an index of sexual receptivity) occurs approximately 160 milliseconds (ms) after mounting by the male (Pfaff and Lewis, 1974) and lasts from about 200 ms to several seconds. Our understanding of this behavior would be greatly improved if we could determine the chemical signaling events associated with its onset, maintenance and termination (e.g. Schulze and Gorzalka, 1995). Another example is provided by Parkinson's disease, which is neurologic ally primarily characterized by a deterioration of the dopaminergic nigrostriatal pathway and behaviorally characterized by, among other symptoms, a rhythmic tremor of 3 - 6 Hz (e.g. Cote and Crutcher, 1985). Detection of the chemical signaling events responsible for these tremors will enhance our understanding of this disease and so facilitate treatment efforts. The vast majority of interneural signaling events are very rapid and only two examples from a 11 possible multitude have been presented. Clearly, to gain insight into the temporal characteristics of these processes, these events have to be detected as they occur. This will necessitate a technique with an appropriately fine temporal resolution. 1.5.2. The need for a general technique. To date dozens of neurotransmitters have been identified. These fall into two broad classes: small-molecule neurotransmitters (about 10 different compounds) and neuropeptides (more than 30 different compounds). Several of these compounds coexist and are released concomitantly (e.g. Schwartz, 1985). A method that can detect a wide variety of unrelated neurotransmitters (e.g. peptides and small-molecule transmitters) is necessary to investigate this type of concomitant signaling. A consequence of concomitant signaling is that at least some postsynaptic neurons must be sensitive to two or more different types of neurotransmitters. Furthermore, many neurons receive inputs from a variety of sources perhaps using a variety of transmitters. In addition, many functional central neural circuits involve neurons that use different transmitters. For instance, an imbalance in the dopaminergic-cholinergic-GABA (y-aminobutyric acid)-ergic loop, which is biochemically distinct but functionally integrated, can cause movement disorders. This happens in Huntington's disease where acetylcholine synthesizing and gamma-aminobutyric acid synthesizing neurons of the striatum are lost (e.g. Cote and Crutcher, 1985). Understanding the interactions between different transmitters will have undeniable advantages and demands a method capable of detecting them. Here too, a clear need exists. 1.6. P R O P O S A L F O R A N E U R O S E N S O R 1.6.1. The problem The eventual aim of this work is to furnish researchers with a sophisticated technique for the effective correlation of behavior and cognition with neurotransmitter 12 release in health and disease. In order to formulate the requirements for such a technique, some important issues will be considered in this section. Chemical neurotransmission at the synapse takes about 10 ms (see Kandel and Siegelbaum, 1985) while behaviors become manifest within about 200 ms (e.g. Hyman, 1953; Pfaff and Lewis, 1974). For the adequate real time correlation of brain function with behavior and cognition, a temporal resolution of at least 10 ms is required. Baseline levels of dopamine in the rat nucleus accumbens are about 1 nM as determined with in vivo microdialysis (Fiorino et al., 1993) while release levels around 1 p M can be obtained in the striatum under some conditions (Wood et al., 1992). Other neurotransmitters can be expected to have similar resting and release concentrations, thus establishing the required limit of detection at approximately 1 nM. Up to now, approximately 40 chemicals have been identified as neurotransmitters (e.g. Schwartz, 1985). If one supposes at least one precursor and one metabolite for each of these, in excess of a 120 substances are directly or indirectly involved in neural signaling. As mentioned above, these different compounds are often jointly involved in some brain function. Furthermore, many neurotransmitters and their metabolites are often simultaneously present in some brain areas (e.g. Fiorino et al., 1993; Wood et al., 1992). Therefore, an adequate understanding of brain function cannot be obtained without the ability to measure several neurotransmitters, their precursors, and their metabolites concurrently. In addition, measurements should take place in the extracellular fluid of the living brain under natural conditions (Crespi, 1990; Ungerstedt, 1986). This requirement implies remote measurement in an aqueous environment with minimal invasiveness. Taken together, the following 4 stringent performance requirements apply to the ideal neurosensor: (i) real time measurements (10 ms resolution); (ii) physiological detection limits (1 nM); 13 (iii) concurrent multi-component detection (120 plus); and (iv) in situ operation (remote aqueous samples). To these are added the need to identify and quantify the various concurrently measured substances and the requirements of the ideal biosensor (e.g. linearity, dynamic range, biocompatibility, etc. - see Buerk 1993). 1.6.2. The conceptual solution The best candidates for a fast, general technique appear to be the various analytical spectrometric methods. These methods involve the detection of radiation, can be extremely fast, and can be used with optical fibers, thus rendering remote data collection possible. Moreover, if the data thus collected can be appropriately processed and interpreted, the need to limit or select the data can be eliminated. Theoretically at least, spectrometry offers the possibility of fast and complete data collection in vivo. There are several potential spectrometric techniques, e.g. fluorescence, absorption, reflectance, infrared absorption, Raman, nuclear magnetic resonance, to name but a few. Of these methods, Raman and especially resonance Raman spectroscopy appear to be the most feasible for a fast, general in vivo neurotransmitter analytical method for the following reasons: (i) they permit fast data collection due to the inherent rapidity of the scattering process; (ii) organic molecules have unique and relatively narrow Raman band profiles in the fingerprint 1500 cm\" 1 to 500 cm~l region (unlike the broad bands produced by fluorescence); (iii) a low detection limit can be realized with resonance Raman spectroscopy (unlike infrared absorption spectroscopy); (iv) interfering fluorescence is absent below 260 nm; (v) significant interference by water, glass, and silica Raman scattering does not occur; (vi) selective resonance enhancement can be obtained with wavelength tuning and (vii) moderately uncomplicated instrumentation is required (unlike nuclear magnetic resonance spectroscopy). 1 4 In vivo Raman spectroscopy could be effected by the use of optical fibers to connect a living animal to a Raman spectrometer. An optical fiber probe could be constructed and inserted through an implanted intracranial cannula. When the probe is implanted and all the equipment operational, the release of a neurotransmitter or neurotransmitters from their nerve terminals would lead to a local increase in the concentration of such neurotransmitter(s). Diffusion would then occur away from the site of release. This process would bring the neurotransmitter molecules into the sensing area of the probe as shown in Figure 1.2. Neurotransmitter-Synaptio=cleft Probe Raman scattering Excitation Figure 1.2. A schematic representation of the in vivo Raman sensing process. It is for conceptual purposes only and not drawn to scale. In particular, the probe diameter is about 500 \\im and the synaptic cleft about 50 nm. The sensing area of the probe could either continuously or periodically be subjected to electromagnetic radiation of the required frequencies delivered by the excitation fiber. Interaction between the molecules and the radiation would change the frequency characteristics of some of the scattered radiation which can be collected and 15 relayed to a spectrometer by the collection fiber. The collected light can be decoded into its constituent frequencies by the spectrograph before delivery to a detector. The radiation so measured can then be mathematically processed to enable the identification of the different neurotransmitter species and their concentrations. The dynamic nature of in vivo neurotransmitter levels makes it possible in principle to employ difference spectroscopy and matrix algebra to identify and quantify individual neurotransmitters. If neurotransmitters fluctuate independently in vivo (given a certain time scale), then a number of spectra taken at suitable intervals can be used to identify those neurotransmitters fluctuating most rapidly with the standard methods of matrix algebra (e.g. Gaussian elimination). For instance, if correlated with a certain behavior, 5 neurotransmitters show independent fluctuations in their levels (i.e. released at different times and/or rates), then 5 spectra taken at different intervals can be used to identify and quantify these 5 components. The fact that the tissue/extracellular fluid ratio for neurotransmitters can increase by up to 3 orders of magnitude upon pharmacological stimulation while those of precursors and metabolites show little change (see Westerink et al., 1987), provides added support for using difference spectroscopy. Methods such as the blind identification of independent source signals in sensor arrays (Cichocki and Unbehauen, 1993) and principal component analysis (e.g. Erickson et al., 1992) could also be applied to this problem. In addition, methods exist (e.g. classical least squares regression and artificial neural network processing - Schulze et al., 1995) for the resolution of the individual components of static mixtures which could be applied to individual spectra (a dynamic mixture could be considered static for a very short period of time during which a spectrum is obtained). Finally, for quantification purposes, water could serve as a convenient and practical internal standard. Using fiber optic based resonance Raman spectroscopy (Chapter 2 contains a brief discussion of Raman spectroscopy and its instrumentation) has been considered difficult due to the problems detailed in Part III. These mostly concern the ability to 16 transmit pulsed ultraviolet radiation with optical fibers (see Chapter 7) and designing an appropriate probe for such work (see Chapter 8). 1.6.3. The approach In order to establish practical goals, it was proposed to develop a Raman neuroprobe capable of identifying and quantifying the 1 0 small-molecule neurotransmitters (acetylcholine, dopamine, epinephrine, norepinephrine, serotonin, histamine, aspartate, y-amino butyric acid, glutamate, and glycine) and mixtures of these in vivo. This consisted of 1 1 basic goals: (i) the characterization of the individual neurotransmitters with normal and resonance Raman spectroscopy; (ii) the determination of the viability of resonance Raman spectroscopy for obtaining selectivity and sensitivity; (iii) the spectroscopic characterization of the intended environment; (iv) the design and development of a suitable fiber-optic probe for in vivo operation; (v) the determination of the conditions for optimum probe use; (vi) the testing of the probe in vitro with neurotransmitters: individually, in mixtures, and using cell culture secretions to simulate in vivo conditions; (vii) the testing and development of the brain/probe interface; (viii) the testing for possible behavioral effects due to probe operation in vivo; (ix) the selection/development of post collection processing methods to improve signal-to-noise ratios in order to help achieve stringent detection limits; (x) neurotransmitter identification and quantification from complex spectra obtained under difficult conditions; and, if possible, 17 (xi) the fiber-optic based detection of these neurotransmitters in vivo including a comparison with in vivo voltammetry and in vivo dialysis to determine the viability of in vivo spectroscopy. These goals were advanced in parallel, as much as circumstances permitted, rather than serially. This was done to avoid solving earlier problems in ways inconsistent with later requirements. Of these 11 basic goals, all but the last one have been completed. Contrary to the situation when this work was commenced, it is now for the first time possible, and as a direct consequence of this work, to perform in vivo resonance Raman spectroscopy with state-of-the-art equipment. 1.6.4. General principles and axioms The development of this neuroprobe is based on the following general principles: That all data available in situ be collected and mathematically post processed for species identification and quantification before chemical and/or mechanical selection is attempted. In other words, it is preferable that selectivity be obtained mathematically (which retains all collected information) and not physically (which rejects much useful information before collection). The following axioms were postulated: (i) neurotransmitters will not all be released at the same time; (ii) neurotransmitters will not all be released at the same rate; (iii) neurotransmitters will not all diffuse away from the site of release at the same rate; (iv) neurotransmitters can all be differentiated by spectroscopic means, and (v) failing (iv), neurotransmitters can be mechanically separated (e.g. with permeable membranes) until spectroscopic differentiation is possible. Assumptions (i)-(iii) collectively assert the independence of neurotransmitter release (which may only be true by first approximation), while assumption (iv) is 18 justified by the uniqueness of Raman spectra (Chapter 2) and (v) by existing techniques (Chapter 1). 1.7. T H E S I S O U T L I N E The present thesis describes the progress made in the development of a resonance Raman neuroprobe for in vivo use. A tree diagram outlining the structure of the thesis and the important areas of research involved is given in Figure 1.3. T h e s i s s t ructure Introductbn Area Theory Devebpment and sub-applications Application Conclusion Chapter 1 Spectroscopy Fiber optics Biopsychology Signal processing Chapter 2 Chapter 6 Chapter 1 Chapter 11 Chapter 3 Chanter 4 Chapter 5 Chapter 16 Chapter 17 Chapter 12 Chapter 13 Chapter 14 Chapter 15 Figure 1.3. Schematic diagram showing the organization of the thesis and the major research areas involved. The thesis is divided into 6 parts. Part I, consisting of Chapter 1, serves as a general introduction to neurotransmission and the problem of neurotransmitter measurement. Each of the next 4 parts has an introductory chapter discussing the basic theory applicable to that part, hence serving as a reference chapter. Part II describes 19 Raman and resonance Raman spectroscopy and instrumentation; Part III describes the characterization of optical fibers and the development of optical fiber probes; Part IV describes the measurement of neurotransmitter secretions from cell cultures, the development of the brain/probe interface, and the behavioral effects of in vivo probe operation; and Part V deals with signal analysis. Part VI contains the penultimate chapter which attempts to draw all 4 primary areas of investigation together in a single application and the last chapter, Chapter 17, contains a summary and discussion of the progress made and addresses issues for future research. 1.8. R E F E R E N C E S . Blahs, C. D. and Lane, R. F. 1983. Chemically modified electrode for in vivo monitoring of brain catecholamines. Brain Research Bulletin 4: 861-864. Bright, F. V . 1989. Multifrequency phase fluorescence study of hapten-antibody complexation. Analytical Chemistry 61: 309-313. Bright, F. V . 1988. Bioanalytical applications of fluorescence spectroscopy. Analytical Chemistry 60: 1031A-1039A. Buerk, D. G. 1993. Biosensors. Technomic Publishing: Lancaster, 1-19. Camhi, J. M . 1984. Neuroethology. Sunderland: Sinauer, 240-244. Cichocki, A. ; Unbehauen, R. 1992. Neural networks for optimization and signal processing. New York: Wiley, 461-471. Crespi, F. 1990. In vivo voltammetry with micro-biosensors for analysis of neurotransmitter release and metabolism. Journal of Neuroscience Methods 34: 53-65. Cote, L . ; Crutcher, M . D. 1985. Motor functions of the basal ganglia and diseases of transmitter metabolism. In E. R. Kandel and J. H . Schwartz (eds.) Principles of neural science (2nd ed.). Amsterdam: Elsevier, 523-535. Delpy, D.T.; Cope, M . ; van der Zee, P.; Arridge, S.; Wray, S.; Wyatt. J. 1988. Estimation of optical pathlength through tissue from direct time of flight measurements. Physics in Medicine and Biology 33: 1433-1442. den Hollander, J. A. ; Behar, K. L . ; Shulman, R. G. 1984. Use of double-tuned surface coils for the application of ^ C - N M R to brain metabolism. Journal of Magnetic Resonance 57: 311-313. 20 Dickensheets, D. L . ; Cheung, P. W. 1989. Pathlength independent spectrophotometric measurement of hemoglobin in solution. IEEE - Engineering in Medicine and Biology, 11th Annual Conference 4: 1090-1091. Erickson, C. L . ; Lysaght, M . J.; Callis, J. B. 1992. Relationship between digital filtering and multivariate regression in quantitative analysis. Analytical Chemistry 64: 1155-1163. Fiorino, D. F.; Coury, A. ; Fibiger, H . C ; Phillips, A . G. 1993. Electrical stimulation of reward sites in the ventral tegmental area increases dopamine transmission in the nucleus accumbens of the rat. Behavioral Brain Research 55: 131-141. Garrell, R. L . ; Heme, T. M . ; Ahern, A. M . ; Sullenberger, E. 1990. Surface-enhanced Raman spectroscopy of peptides. Proceedings of SPLE - The International Society for Optical Engineering, vol. 1201, Optical Fibers in Medicine V : 451-460. Gonzalez-Mora, J. L . ; Guadalupe, T.; Fumero, B.; Mas, M . 1991. Mathematical resolution of mixed in vivo voltammetry signals. Journal of Neuroscience Methods 39: 231-244. Hall, E. A . H . 1992. Overview of biosensors. In P. G. Edelman and J. Wang (eds.) Biosensors and chemical sensors: optimizing performance through polymeric materials. Washington, DC: American Chemical Society, 1-14. Hammer, B. E.; Sacks, W.; Bigler, R. E.; Hennesey, M . J.; Sacks, S.; Fleischer, A. ; Zanzonico, P. B. 1990. Design of a ^C{ ^H} RF probe for monitoring the in vivo metabolism of [1-^C] glucose in primate brain. Magnetic Resonance in Medicine 13: 1-5. Hyman, R. 1953. Stimulus information as a determinant of reaction time. Journal of Experimental Psychology 45: 188-196. Kandel, E . R.; Siegelbaum, S. 1985. Principles underlying electrical and chemical synaptic transmission.. In E. R. Kandel and J. H . Schwartz (eds.), Principles of neural science. Elsevier: Amsterdam, 89-107. Kelly, J. P. 1985. Anatomical basis of sensory perception and motor coordination. In E. R. Kandel and J. H . Schwartz (eds.), Principles of neural science. Elsevier: Amsterdam, 222-243. Kelly, J. S.; Simmonds, M . A. ; Straughan, D. W. 1975. Microelectrode techniques. In: P. B. Bradley (ed.), Methods in brain research. London: Wiley, 295-332. Kulp, T. J.; Camins, I.; Angel, S. M . 1988. Enzyme-based fiber optic sensors. Proceedings of SPLE - The International Society for Optical Engineering, vol. 906, Optical Fibers in Medicine III: 134-138. McGlashen, M . L . ; Davis, K . L. ; Morris, M . D. 1990. Surface-enhanced Raman scattering of dopamine at polymer coated silver electrodes. Analytical Chemistry 62: 846-849. 21 Mitchell, J. F. 1975. Collection techniques. In: P. B. Bradley (ed.), Methods in brain research. London: Wiley, 333-377. Monita, M . ; Niwa, O. 1996. Electrochemical detection using interdigitized array carbon microelectrodes - detection limits of 5/1,000 trillionths grams achieved. NTT Review 8:77-80. Morris, M . D.; McGlashen, M . L. ; Davis, K. L . 1990. Surface-enhanced Raman (SERS) probes of neurotransmitters. Optical Fibers in Medicine V SPLE 1201: 447-450. Obata, K. and Takeda, K . 1969. Release of gamma-amino butyric acid into the fourth ventricle induced by stimulation of the cat's cerebellum. Journal of Neurochemistry 16: 1043-1048. Pantano, P.; Kuhr, W. G. 1995. Enzyme-modified microelectrodes for in vivo neurochemical measurements. Electroanalysis 7:405-416. Pappenheimer, J. R.; Heisey, S. R.; Jordan, E. F. and Downer, J de C. 1962. Perfusion of the cerebral ventricular system in unanesthetized goats. American Journal of Physiology 203: 763-774. Pellionisz, A . and Llinas, R. (1979) Brain modeling by tensor network theory and computer simulation. The cerebellum: distributed processor for predictive coordination. Neuroscience 4: 323-348. Pfaff, D. W.; Lewis, C. 1974. Film analyses of lordosis in female rats. Hormones and Behavior 5: 317-335. Rea, P. A. ; Crowe, J.; Wickramasinghe, Y . ; Rolfe, P. 1985. Non-invasive optical methods for the study of cerebral metabolism in the human newborn: a technique for the future? Journal of Medical Engineering and Technology, 9: 160-166. Schultz, J. S. 1985. Medical applications of fiber optic sensors. Medical Instrumentation 19: 158-163. Schulze, H . G.; Gorzalka, B. B. 1995. Behavioral modeling: the use of chemical reaction kinetics to investigate lordosis behavior in female rats. Journal of Theoretical Biology 174: 355-369. Schulze, H . G.; Greek, L . S.; Gorzalka, B. B.; Bree, A . V. ; Blades, M . W.; Turner, R. F. B. 1995. Neural network and classical least squares methods for neurotransmitter mixture analysis. Journal of Neuroscience Methods, 56: 155-167. Schwartz, J. H . 1985. Chemical messengers: small molecules and peptides. In E. R. Kandel and J. H . Schwartz (eds.) Principles of neural science (2nd ed.). Amsterdam: Elsevier, 148-158. Seitz, W. R. 1989. Transducer mechanisms for optical biosensors. Part 1: The chemistry of transduction. Computer Methods and Programs in Biomedicine 30: 9-19. 22 Seitz, W. R. 1984. Chemical sensors based on fiber optics. Analytical Chemistry 56: 16A-34A. Sekihara, K.; Ohyama, N . 1990. Parameter estimation for in vivo magnetic resonance spectroscopy (MRS) using simulated annealing. Magnetic Resonance in Medicine 13: 332-339. Sepaniak, M . J.; Tromberg, B. J.; Vo-Dinh, T. 1988. Fiber optic affinity sensors in chemical analysis. Progress in Analytical Spectroscopy 11: 481-509. Skoog, D. A . 1985. Principles of Instrumental Analysis. Philadelphia, Saunders College Publishing. Thorniley, M . S.; Wickramasinghe, Y . A . B. D.; Rolfe, P. 1988. Near infra-red spectroscopy: a new technique for the non-invasive monitoring of tissue and blood oxygenation in vivo. Biochemistry Society Transactions 16: 978-979. Tromberg, B. J.; Sepaniak, M . J.; Vo-Dinh, T. 1988. Development of antibody-based fiber-optic sensors. SPIE - The Society for Photooptical Instrumentation Engineers, vol. 906, Optical Fibers in Medicine III: 30-38. Tungsal, B. ; Hofmann, K. ; Stoffel, W. 1990. In vivo A ^ C nuclear magnetic resonance investigations of choline metabolism in rabbit brain. Magnetic Resonance in Medicine 13: 90-102. Ungerstedt, U . 1986. Microdialysis - a new bioanalytical sampling technique. Current Separations 7: 43-46. Ungerstedt, U . 1984. Measurement of neurotransmitter release by intracranial dialysis. In C. A . Marsden (ed.) Measurement of neurotransmitter release in vivo. New York: Wiley, 81-105. Westerink, B. H . C ; Damsma, G.; Rollema, H. ; de Vries, J. B.; Horn, A . S. 1987. Scope and limitations of in vivo brain dialysis: a comparison of its application to various neurotransmitter systems. Life Sciences 41: 1763-1776. Wood, E. R.; Coury, A. ; Blaha, C. D.; Phillips, A . G. 1992. Extracellular dopamine in the rat striatum during ischemia and reperfusion as measured by in vivo electrochemistry and in vivo microdialysis. Brain Research 591: 151-159. 23 PART II Spectroscopy Chapter 2 2.1. I N T R O D U C T I O N 2.2. N O R M A L R A M A N S P E C T R O S C O P Y 2.3. R E S O N A N C E A N D S U R F A C E E N H A N C E D R A M A N S P E C T R O S C O P Y 2.4. I N S T R U M E N T A T I O N 2.5. R E F E R E N C E S 2.1. I N T R O D U C T I O N In Part II the focus is on the spectroscopic work done for this multidisciplinary thesis. This work served to establish the uniqueness of Raman and resonance Raman spectra of neurotransmitter for identification purposes, the feasibility of using resonance Raman spectroscopy to obtain signal enhancement and selectivity, and to briefly characterize the intended measuring environment. In addition spectra were obtained for use in the signal processing part of the thesis (e.g. difference spectra). Part II also contains a description of the Raman system designed for this work. The present chapter establishes the theoretical basis for Part II. It provides brief introductions to Raman, surface enhanced Raman, and resonance Raman spectroscopy and the instrumentation used for Raman measurements. 2.2. N O R M A L R A M A N S P E C T R O S C O P Y Rayleigh scattering of light by matter consists of the elastic scattering of incident radiation. In contrast, the Raman effect discovered in 1928 by the Indian physicist C V . Raman, consists of the species-characteristic inelastic fractional (about 1/10^) scattering of incident radiation. In the particle theory view of Raman scattering, molecules are seen as being excited to non-quantized virtual energy states from where they immediately (10\" 15 s) relax to vibrational ground electronic states. The scattered radiation is Stokes shifted if the molecule was originally in a ground vibrational state and relaxation 24 occurred to an excited vibrational state and is anti-Stokes shifted vice versa, hence the Boltzmann distributions of thermally occupied states account for the relative intensities of these differently shifted frequencies. In the wave theory view, a proportional electric dipole moment, m, with proportionality constant, a, is induced in a molecule subjected to an electric field E. The size of the proportionality constant depends on the polarizability of the molecular bond involved in the vibration, which is in general dependent on the bond length. The intensity of scattering due to the Raman-active vibrational mode (I s, erg) is proportional to (proportionality constant K) the squared derivative of the polarizability tensor elements (cc'2) induced by the molecular vibration (of frequency vi , Hz) caused by the excitation light, the concentration of the analyte ((proportional to n/e\" E/kT) where E is energy (erg), k is Boltzmann's constant (1.3805x10-6 erg/degree), T is absolute temperature (degree), n^ are the number of molecules in state i), the intensity of the incident radiation (IQ, erg), and the fourth power of the frequency of the incident radiation (VQ, H Z ) as shown in Equation 2.1 (e.g. Asher, 1988): I s = K I o ( v o ± v i ) 4 a ' 2 n / e - E / £ T (2.1) Energy exchanges of vibrational quanta between radiation and matter involve infrared frequencies, thus accounting for the close relationship between infrared and Raman spectroscopy. However, Raman spectra (dependent on the change in molecular polarizability) often provide information not available from IR spectra (dependent on the induced molecular dipole moment), rendering the two methods mostly complementary. In particular, water and glass exhibit weak Raman scattering while nonpolar bonds (e.g. organics) show strong Raman scattering (Gerrard, 1991), thus making Raman spectroscopy well suited for the fiber-optic based analysis of biological samples. The relationships between IR absorption, Raman and resonance Raman scattering, and fluorescence and resonance fluorescence, are shown in Figure 2.1. 25 F 2 F V2 v o R R Ra R (a) (b) (c) (d) Figure 2.1. Energy level diagram showing the relationships between (a) IR absorption; (b) Rayleigh (Ra), Raman (R), and resonance Raman (RR) scattering; (c) fluorescence; and (d) resonance fluorescence. (Ei = electronic energy level i; v; = vibrational energy level i.) 2.3. R E S O N A N C E A N D S U R F A C E E N H A N C E D R A M A N S P E C T R O S C O P Y The normally weak Raman signals can be enhanced in a number of ways, the two most important ones being resonant Raman scattering (RRS) and surface-enhanced Raman scattering (SERS). When the energy of the incident radiation approaches that of an electronic transition of the molecule, increases in the intensity of Raman scattering, up to 6 orders of magnitude, are observed (e.g. Vickers et al., 1991). This occurs because extra intensity is \"mixed\" into vibrational modes that are otherwise weak (e.g. Stencel, 1990). Consider again the change in polarizability term (a') from Equation 2.1 which can be rewritten as: a' = l//*Z((M m e M e „)/( vem - v0 + iTe) + (MmeMen)/(vem + v0 + iTe)) (2.2) where h is Planck's constant (6.62x10\"27 erg s), the M's are electric transition moments (erg), the vem is the frequency (Hz) corresponding to the energy difference between 26 states e and m, m and n index ground electronic vibrational states, e indexes an excited electronic vibrational state, iTe is a damping constant (Hz), and VQ has been defined in Equation 2.1 (Ferraro and Nakamoto, 1994). In normal Raman scattering, the excitation frequency is much lower than that of a molecular electronic transition and Raman bands are proportional to (VQ - V j ) 4 . However, when the excitation frequency is close to the absorption frequency of a particular chromophore in a molecule, the Raman band associated with that chromophore would get selectively enhanced (Asher et al., 1986; Ferraro and Nakamoto, 1994; Johnson et al., 1984). This occurs because V Q approaches v e m and the denominator of the first or \"resonance\" term of Equation 2.2 becomes very small, leading to a large increase in the value of the resonance term. This is illustrated in Figure 2.2. 1 02 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 h 0.0 -0.1 ******** A B 200 220 240 260 280 300 320 340 360 380 400 Wavelength (nm) Figure 2.2. Absorption (dotted line) and Raman spectra (solid lines) of a compound with two chromophores (A and B) showing how selective relative resonance enhancement of the chromophoric Raman bands can be obtained by the appropriate choice of the excitation frequency. 27 Resonance Raman scattering (10 - 1 ^ s) is slower than Raman scattering because relaxation now occurs, not from a virtual state, but from an excited electronic state. It differs from fluorescence (10~9 s) in the respect that no prior relaxation to the lowest vibrational level of the excited electronic state occurs and is therefore a faster process (Skoog, 1985). Fluorescence is also absent in Raman scattering excited by wavelengths below about 260 nm (Johnson et al., 1984; Vickers et al. 1991). Extraordinary enhancements of Raman signals, up to 6 orders of magnitude, can be obtained from analytes adsorbed on some rough metal surfaces such as Cu, Ag, Au, Ni , Pt, Pd, Ti , Hg, and others (Chang and Furtak, 1982; Kerker, 1990; Stencel, 1990). Initially, two general models have been advanced to account for this phenomenon: the electromagnetic model and the chemical model (Kerker, 1987; Pettinger, 1986). Enhancement according to the electromagnetic model occurs as a result of intensified local electromagnetic fields at or near metal rough spots. Chemical models attribute the enhancement to a modification of the molecular polarizability due to an interaction with the metal surface thus producing molecular resonances. Currently, it is widely accepted that there are at least two contributions to SERS: electromagnetic enhancement and an enhancement due to changes in the electro-optical parameters of the scattering process. 2.4. I N S T R U M E N T A T I O N The typical Raman spectroscopic system consists of a light source, excitation optics, the sample and sample holder, collection optics, a frequency decoder, detector, and data processing equipment. The advent of the laser, providing intense monochromatic light, has facilitated spectroscopy based on the Raman effect. Continuous wave (cw) visible lasers are frequently used and the excitation wavelength should be chosen to avoid analyte absorption (except for resonance Raman spectroscopy), solvent absorption, sample 28 fluorescence and photodecomposition. Although less common due to the difficulties encountered with high peak energies and low duty cycles, pulsed lasers can also be used for Raman spectroscopy. These are presently the only means of providing continuously tunable U V radiation for U V resonance Raman work. The excitation optics often consist of a single lens to focus the light on the sample but could also include an interference filter to remove unwanted laser radiation and/or optical fiber to deliver the light to the sample. In general, small samples with little preparation can be used for Raman spectroscopy. Glass and silica can be used as sample holders and capillary tubes are often employed for this purpose. The collection optics generally consist of a high quality lens to collect and focus the scattered light onto the entrance slit of a monochromator and often include a notch filter to eliminate Rayleigh scattered light. Optical fibers can also be used for collection. The frequencies of the scattered light are spatially separated with a spectrograph. The use of a triple spectrograph significantly reduces stray light and is common. Photomultiplier tubes, intensified photodiode arrays, and intensified charge coupled devices are detectors in frequent use for Raman spectroscopy. Photomultiplier tubes are highly sensitive but bulky devices and cannot easily be used for multichannel spectroscopy. Both photodiode and charge coupled device arrays permit multiplex spectroscopy with the general difference that photodiodes are more sensitive in the ultraviolet and charge coupled devices in the visible range of the electromagnetic spectrum. A controller is associated with the detector and allows the experimenter to vary a number of detector parameters such as temperature, exposure time, synchrony, etc. Microcomputers with appropriate software are widely used for experimental control, as well as data collection and manipulation. General descriptions of Raman and resonance Raman spectroscopy and its instrumentation can be found in Skoog (1985), Asher (1988), and Ingle and Crouch (1988), amongst others. 29 2.5. R E F E R E N C E S Asher, S. A . 1988. U V resonance Raman studies of molecular structure and dynamics: applications in physical and biophysical chemistry. Annual Review of Physical Chemistry 39: 537-588. Asher, S. A. ; Ludwig, M . ; Johnson, C. R. 1986. U V resonance Raman excitation profiles of the aromatic amino acids. Journal of the American Chemical Society 108: 3186-3197. Chang, R. K.; Furtak, T. E. (eds.) 1982. Surface enhanced Raman scattering. New York: Plenum. Ferraro, J. R.; Nakamoto, K. 1994. Introductory Raman spectroscopy. Boston: Academic Press, 1-94. Gerrard, D. L . 1991. Organic and petrochemical applications of Raman spectroscopy. In: J.G. Graselli and B. J. Bulkin (eds.) Analytical Raman spectroscopy. Wiley and Sons: New York, 275-323. Ingle, J. D.; Crouch, S. R. 1988. Spectrochemical analysis. Prentice Hall: Englewood Cliffs, 494-513. Johnson, C. R.; Ludwig, M . ; O'Donnell, S.; Asher, S. A . 1984. U V resonance Raman spectroscopy of the aromatic amino acids and myoglobin. Journal of the American Chemical Society 106: 5008-5010. Kerker, M . 1990 (ed.) 1990. Selected papers on surface-enhanced Raman scattering. Bellingham: SPLE (Milestone Series). Kerker, M . 1987. Estimation of surface-enhanced Raman scattering from surface-averaged electromagnetic intensities. Journal of Colloid and Interface Science 118: 417-421. Pettinger, B. 1986. Light scattering by adsorbates at Ag particles: Quantum mechanical approach for energy transfer induced interfacial optical processes involving surface plasmons, multipoles, and electron-hole pairs. Journal of Chemical Physics 85: 7442-7451. Skoog, D. A . 1985. Principles of instrumental analysis. Saunders, Philadelphia, PA. Stencel, J. M . 1990. Raman spectroscopy for catalysis. New York: Van Nostrand Reinhold, 141-174. Vickers, T. J.; Mann, C. K.; Zhu, J.; Chong, C. K . 1991. Quantitative resonance Raman spectroscopy. Applied Spectroscopy Reviews 26: 341-375. 30 30 C H A P T E R 3 3.1. I N T R O D U C T I O N 3.2. I N S T R U M E N T A T I O N 3.2.1. System requirements and characterization 3.2.2. The detector 3.2.3. The spectrograph 3.2.3.1. The single monochromator 3.2.3.2. The double monochromator 3.2.3.3. The fiber-optic adapter 3.2.4. The fiber-optic probe and coupling optics 3.2.5. The light source 3.2.5.1. The second harmonic generator 3.2.5.2. The dye laser 3.2.5.3. The pump laser 3.2.5.4. The Ar+ laser 3.2.6. System throughput 3.2.7. Synchronization 3.3. M E T H O D O L O G Y 3.5. D I S C U S S I O N 3.6. R E F E R E N C E S 3.1. I N T R O D U C T I O N The aim of the interdisciplinary work undertaken and reported in this thesis was to conceptualize, design, and lay the foundations for the development of a brain sensor capable of meeting the general requirements of sensitivity, speed, and versatility while operating non-invasively in an aqueous in situ environment as stated in Chapter 1. As part of this effort, it was necessary to design, procure, and commission equipment, individually and in concert, to create an analytical system suitable for in vivo continuously tunable UV resonance Raman investigations. The following chapter contains the specifications established for this system including the appropriate justification. They also contain a description of the selected equipment, their characterization, and integrated operation. This chapter does not include the design and testing of fiber-optic probes, which are presented in Part H I . For an introduction to Raman spectroscopy and instrumentation the reader may refer to Chapter 2. 31 3.1. I N S T R U M E N T A T I O N . Technically speaking, the design of a spectrometric system should proceed from the detector backwards to the light source to ensure that the entire system matches the performance requirements (e.g. Heiman et al., 1989). Given the fact that funds were not available to purchase all the constituent items simultaneously and that a number of important items were shared with or borrowed from other users, this was not feasible. However, the system, shown in Figure 3.1, will be discussed this way for illustrative purposes. M i r r o r Spectrograph M i r r o r Detector Figure 3.1. A schematic representation of the Raman spectroscopic system. 3.2.1. System requirements. In order to identify the various neurotransmitters from their Raman spectra, it was necessary for these spectra to cover the infrared \"fingerprint\" region (500 cm~l to 2000 cm\"l) at a sufficiently high resolution (4 cm\"1) (e.g. Tanabe et al. 1992). For quantification purposes, water, which has a peak near 1630 cm\"l, could be used as an 32 32 internal standard. An ability to collect a full spectrum every 10 ms would provide the required temporal resolution (see Chapter 1.6) and make it equal to or better than that of in vivo voltammetry (Stamford, 1989). Taken together, a bandwidth of 1000 cm\"l to 2000 cm\"' , a spectral resolution of 2 cm\"l to 5 cnr 1, and a temporal resolution of 10 ms were desired. In addition, a tunable light source to generate light in the 200 - 300 nm region was required for U V resonance Raman spectroscopy (Asher, 1988). Finally, multi-mode, small-diameter optical fiber with high U V transmission and a high damage threshold to withstand the high pulse energies from the light source was needed. 3.2.2. The detector. An intensified diode array detector (EG&G model 1455B-700-G, Princeton, NJ) was selected and purchased for the Raman neuroprobe system. This model consisted of a linear array of 700 intensified diodes, was blue enhanced to give a quantum efficiency of about 12% between 200 and 300 nm, gateable to 5 ns, had a 14 bit dynamic range, was thermoelectrically cooled to below -25 °C to reduce dark current, had a spectral range of 180 - 910 nm, and a 500:1 variable gain. The detector was controlled with a ST120 controller from the same supplier, enabling up to 1000 s integration time, 33 ms array readout time, and external triggering and synchronization. The detector was gated with an FG100 pulse generator, also from E G & G , with variable time delay and variable pulse width for use with pulsed light sources. Gating allows the experimenter to collect a spectrum for only that period of time during which the sample is being excited by a pulse of light. This reduces the number of dark counts and hence gives better signal-to-noise ratios. The detector was thermoelectrically cooled and with the circulation of water coolant temperatures of -25 °C or less could be reached. In order to obtain noise levels, the baseline standard deviations of spectra of 1 s integration measured at 5 minute intervals were calculated. Spectra were collected starting immediately after commencing 33 33 detector cooling. The noise values decayed exponentially and stabilized to ~ 5 counts/s about 30 minutes after the start of detector cooling with 1 L/minute water flow. 3.2.3. The spectrograph. The spectrograph (capable of multichannel spectroscopy as opposed to the spectrometer which is a scanning instrument) could be operated as a single monochromator. With the addition of a double monochromator in subtractive mode to function as an optical filter for Rayleigh rejection, a triple monochromator was formed. 3.2.3.1. The single monochromator. A Model 207 spectrograph from McPherson (Acton, MA) , was acquired for the Raman system. This instrument was of the Czerny-Turner arrangement, had a 0.67 m focal length, accommodated a 120 mm by 140 mm 3600 G/mm holographic grating to give a f/4.7 throughput, had 2 entrance slits with folding mirror, an exit port that could accommodate the detector, a linear dispersion of 0.415 nm/mm in first order to give a bandwidth of 1203 cnr ^ and resolution of 1.7 cm~l (with 3600 G/mm grating operating at 250 nm). The incoming light was dispersed in the single monochromator by a grating and then focussed onto the detector. The calibration of the instrument was checked with visible Hg lines and a 1200 G/mm grating. For a given setting, the center pixel was found to be 530 and the dispersion 0.031 nm/pixel (equivalent to 0.402 nm/mm with a 3600 G/mm grating). 3.2.3.2. The double monochromator. The Model 275D double monochromator with 2 holographic gratings (1200 G/mm) from the same supplier was purchased to function as a stray light rejecter (specified better than 10\"^ at 1.5 bandpasses from a given line). This unit was attached to 34 the Model 207 such that the exit port of the Model 275D mated with the entrance port of the Model 207, had an aperture ratio of f/4.2, a linear dispersion of 4 nm/mm to give a maximum bandwidth of 2576 c i r r i a t 250 nm, and covered the spectral range 185 - 1000 nm. The double monochromator functioned by dispersing the incoming light onto a central slit that could be adjusted to give a certain bandpass. The light passing the central slit was then recombined and imaged onto the entrance slit of the single monochromator. When the instrument was in operation, it was found that the exit slit affected the bandpass of the unit. This was anomalous. Upon questioning, the manufacturer modified the instrument by inserting inversion optics at the position of the central slit which corrected the unit's operation. 3.2.3.3. The fiber-optic adapter. A fiber-optic adapter, Model 132, with a S M A connector, x-y micropositioners for aligning the fiber with the entrance slit of the spectrograph, U V grade optics to aperture-match the fiber (f/4.5) and spectrometer (f/4.7 or f/4.2), adjustable focus, and filter receptacle was also obtained from McPherson. This unit made possible the use of a single optical fiber input to either the single or double monochromator entrance slit assemblies. 3.2.4. The fiber-optic probe and coupling optics. Fiber-optic probes were mostly constructed from glass (F-MLD-10) from Newport (Fountain Valley, CA) , U V grade fused silica (Superguide G) from Fiberguide Industries (Stirling, NJ), and high U V transmission fused silica (Polymicro Technologies, Phoenix, AZ) , but occasionally from other suppliers (see Chapter 8). These were step-index fibers with numerical apertures of 0.22 (f/4.5), core diameters from 100 - 1000 pm, cladding thicknesses typically 10% of the core diameter, and mostly acrylate 35 35 jacketing material. The input ends of the excitation fibers were held in a chuck and positioned with a fiber-optic positioner (Newport Model F915). Light was coupled into the fiber with a microscope objective (Newport M-10X, visible; U-27X, UV) or a 150 mm focal length plano-convex silica lens. 3.2.5. The light source. The tunable ultraviolet light source consisted of the frequency doubled output from a dye laser pumped by a N d : Y A G laser. For visible light excitation, the lines from an A r + laser was used. 3.2.5.1. The second harmonic generator. Second harmonics were generated with a CSK Optronics Super Doubler (Culver City, CA) using {3-barium borate crystals (205-220 nm/ 220-250 nm), cut at phase-matching angles, and separated from the fundamental with a fused silica prism. The efficiency of the frequency doubler was specified as being larger than 5% in the 205 -250 nm range. When measured at 225 nm, the efficiency was found to be 6.4%. 3.2.5.2. The dye laser. Tunable pulsed visible light was obtained with a Quanta Ray PDL-1 narrow-band (less than 0.25 cm~l at 560 nm) dye laser (Mountain View, CA), pumping coumarin 460, Exciton 417, or Exciton 418 dye (Exciton, Dayton, OH), and cited efficiencies from 5 to 20%. It was difficult to establish the efficiency of the dye laser due to the lack of a suitable energy meter head to measure the Q-switched output of the pump laser at 355 nm and high oscillator energy settings. However, the output of the dye laser was measured as a function of the oscillator energy setting of the pump laser after optimization of the dye laser using coumarin 460 and the results are shown in Figure 3.2. 36 j • i i i i i i i i i i i i i i i _ 560 580 600 620 640 660 680 700 720 Oscillates- setting (V) Figure 3.2. The output of the dye laser as a function of the oscillator voltage of the pump laser after optimization of the dye laser using coumarin 460. By placing the energy meter head a suitable distance beyond the focal point of a 50 mm lens, the active area of the energy meter head could be filled with radiation, thus avoiding possible damage to the head. However, the formation of an air plasma at high pulse energies (plateau in Figure 3.2) clearly reduced the throughput thus limiting this approach. The dye laser was further optimized with regard to the oscillator dye concentration (initially 300 mg/L oscillator and 60 mg/L amplifier) and the initial concentration found to be near optimal. Subsequently, the output (shown in Table 3.1) was measured over a 30 minute period at high pump energies (pump laser oscillator setting at 710, 5 Hz) and found to be stable. Table 3.1 The average dye laser output energy (10 pulses) (a) as a function of the oscillator dye concentration and (b) as a function of time given a dye concentration of276 mg/L. (a) Concentration (mg/L) Output (mJ) 324 4.40 314 4.50 303 4.60 300 4.60 293 4.65 276 4.60 273 3.90 37 (b) Time (minutes) Output ±S.E.M. CmD 0 4.65 ±0.15 5 4.60 ±0.11 10 4.72 ±0.10 15 • 4.76 ± 0.08 20 4.67 ±0 .10 25 4.66 ± 0.07 30 4.65 ±0.10 3.2.5.3. The pump laser. The 355 nm output (shown in Figure 3.3) from a Lumonics HY-400 N d : Y A G laser (Rugby, England) operating at 10 Hz was used to pump the dye laser. A half wave plate in a rotary mount (CVI, Albuquerque, NM) was interposed between the pump and dye lasers to adjust the vertically polarized pump light to the horizontal orientation required by the dye laser. Figure 3.3. The output energy from the pump laser at 532 nm as a function of the oscillator voltage setting: 500-700 V (dotted line) and 500-590 V Q-switched (solid line). 38 38 The doubled (532 nm) and quadrupled (266 nm) output from the pump laser was sometimes used directly for Raman excitation. The output of the pump laser was re-optimized whenever a different wavelength was used. 3.2.5.4. The A r + light source. A Spectra Physics Stabilite 2017 A r + laser (Mountain View, CA) was used to provide continuous wave visible and U V excitation. The 488 nm, 473 nm, and 364 nm, lines were mostly used. The output from this laser was measured with a power meter and the readings found to correspond with those of the instrument's built-in power meter. Furthermore, plasma lines from ion lasers can interfere with Raman measurements (see Figure 10.4, top trace, about pixel 730) and interference filters are often used to avoid this problem. Since an interference filter was only available for the 488 nm line, and since several of the other lines were periodically used for excitation, the laser was occasionally detuned to verify the existence of plasma lines. When the laser was detuned, lasing ceased, but the plasma lines persisted to the extent that they could be easily and safely detected with the spectroscopic equipment. 3.3.1.8. System throughput. Using 200 mW at 514 nm, the throughput of the system as shown in Figure 3.4 was measured. With the single monochromator the throughput to the detector was about 12%, but with a triple monochromator arrangement the throughput to the detector was too low to measure and estimated at 1%, an additional 93% (of 12%) loss. Values reported in the literature range from 1% - 5% (e.g. Chase, 1994). The most severe losses occurred at the gratings. These results suggested that a prism monochromator system may be more advantageous for Raman spectroscopy, notably in the ultraviolet region where the dispersion of quartz would be more pronounced. 39 Detector Figure 3.4. A schematic representation showing the throughput of the Raman spectroscopic system at various points. 3.3.1.9. Synchronization. In order to test the synchronization of the gating of the detector with the arrival of the optical pulse from the laser, the system was set up as shown in Figure 3.5. The Pockels cell synchronization output started 100 ns before the optical pulse from the laser. From the moment of triggering by the Pockels cell to the arrival of the light at the detector, about 120 ns elapsed, while the electrical signal took about 64 ns to reach the detector. The electrical path consisted of two sections of coaxial cable (30 foot and 4 foot, respectively) and the fixed internal delay of the pulse generator (30 ns). The variable delay feature of the pulse generator (part of the electrical pathway) was used to compensate for the difference. Furthermore, the delay required proved to vary with the oscillator energy setting. In general, the gate width was set at a maximum to bracket the signal whereafter it was reduced to optimize the signal-to-noise ratio. The gating could perhaps be improved by splitting a small fraction of the beam to allow triggering from a 40 separate detector. Due to small pulse-to-pulse variations in the Q-switch operation, using the Q-switch for triggering demands a wider gate width allowing for the build-up of unnecessary dark charges. However, the gate pulse generator has an internal delay of 30 ns while the optical path is traversed in 19 ns thus requiring a delay of the main optical beam. Laser 1 0 0 ns 30 ns 19ns Adapter Spectrograph Pulser 30 ns 4 ns Detector Figure 3.5. A schematic representation of the optical path and electronic circuit time lags used to synchronize the optical and gating pulses. 3.3. M E T H O D O L O G Y . The general manner in which the equipment was operated and data collected is described in this section. The construction of the optical fiber probe is described in III. More detailed information will be provided in subsequent chapters describing spectroscopic measurements. Prior to data collection, the input fiber endface was rinsed with deionized water, and subsequently visually aligned with the laser beam and optimized with the aid of a 4 1 piece of white paper (which reflected visible radiation and produced violet fluorescence from U V radiation). Optimizing the coupling visually was generally satisfactory as verified with a power/energy meter. Optimizing the coupling of light from the collection fiber(s) into the spectrometer was done with the 1050 cm~l peak of KNO3 (visible excitation) and the Rayleigh line (UV excitation). The slit width on the spectrometer was generally on the order of the collection fiber/bundle diameter and the intermediate slit on the double monochromator fully open for maximum bandpass. A 1200 G/mm holographic grating was used for frequency decoding visible wavelengths and a 3600 G/mm holographic grating for U V wavelengths. A variety of sample holder shapes and sizes were employed for Raman measurements. With some probe designs (ends aligned flush) and visible excitation, a measuring cylinder was used. It was modified by covering the bottom with a piece of black plastic to attenuate the amount of stray light collected by the probe (and hence the generation of large spectral backgrounds). When using U V excitation, a background was collected after the detector had been cooled sufficiently to minimize dark current noise and this background was subtracted from the measured spectra to remove the deterministic readout noise of the detector. This was rarely done with visible excitation because signals were mostly clearly detectable. The components comprising the light source were optimized only initially or after some change in set-up or configuration had been implemented (e.g. changing to U V optics on the A r + laser). 3.4. D A T A A N A L Y S I S . For the identification of neurotransmitters from their Raman spectra, artificial neural networks were used. Difference spectroscopy was investigated with the intention 4 2 4 2 to remove static and slowly changing spectral components and thus simplify complex spectra before identification. For quantification, the water peak at 1630 cm\"' served as an internal standard and working curves for some individual neurotransmitters were established. The data analysis is discussed in detail in Part V. 3. D I S C U S S I O N . This discussion serves to briefly outline some of the operational implications inherent in a system as described above. In general, one wishes to collect as much scattered light as possible, this would give spectra in the shortest possible time. Collecting spectra for a longer time generally reduces the noise level and increases the signal strength (see Chapter 11) so that even very weak signals could be detected if collection continues for an adequate length of time. Due to the reduction in noise level, the resolution of the collected spectra are better. If a higher resolution of the spectra is required, the entrance slit to the monochromator could be narrowed (e.g. Skoog, 1985), thus requiring more collection time. The resolution could also be increased by using a dispersive element (grating or prism) with greater dispersion. This would also lead to a reduction of throughput to the detector and require more collection time. Hence improved resolution and better signal-to-noise ratios are generally gained at the expense of collection speed. Higher intensity signals could be obtained by increasing the laser power thus producing more intense scattering from the sample. Under some conditions, samples and/or fiber optic probes (see Part UI, Chapter 7) may be sensitive to high powers, thus establishing effective upper limits to the laser power that could be used. Higher intensity signals could also be obtained by using more sensitive detectors, or a better cooling of a detector. Furthermore, a better throughput of the system (i.e. using more efficient dye lasers, doubling optics, beam steering optics and optical fiber probe, as well as spectrometer - see Chapters 8 and 17) would also shorten the collection time required to 43 43 obtain high resolution spectra. Finally, the application should be kept in mind in order to determine the appropriate balance to maintain between these variables. In the present case, a resolution sufficient for spectral identification is required which is perhaps not as severe a requirement as that necessary to detect spectral shifts caused by changes in physical environment (e.g. change of solvent, protein denaturation, isotope substitution, temperature braodening, etc.). 3.6. R E F E R E N C E S . Asher, S. A . 1988. U V resonance Raman studies of molecular structure and dynamics. Annual Review of Physical Chemistry 39: 537-588. Chase, B. 1994. A new generation of Raman instrumentation. Applied Spectroscopy 48: 14A-19A. Heiman, D.; Zheng, X . L . ; Sprunt, S.; Goldberg, B. B.; Isaacs, E. D. 1989. Fiber-optics for spectroscopy. Raman Scattering, Luminescence, and Spectroscopic Instrumentation in Technology, SPIE 1055:, 96-104. Tanabe, K. ; Tamura, T.; and Uesaka, H . 1992. Neural network system for the identification of infrared spectra. Applied Spectroscopy 46: 807-810. Stamford, J. A . 1989. In vivo voltammetry - prospects for the next decade. Trends in Neuroscience 12: 407-412. 44 C H A P T E R 4 4 . 1 . I N T R O D U C T I O N 4 . 2 . N O R M A L R A M A N S P E C T R A 4 . 2 . 1 . Neurotransmitter spectra and band assignments 4 . 2 . 1 . 1 . Physiological saline 4 . 2 . 1 . 2 . Acetylcholine 4 . 2 . 1 . 3 . Dopamine 4 . 2 . 1 . 4 . Epinephrine 4 . 2 . 1 . 5 . Norepinephrine 4 . 2 . 1 . 6 . Serotonin 4 . 2 . 1 . 7 . Histamine 4 . 2 . 1 . 8 . Aspartate 4 . 2 . 1 . 9 . y-Amino butyric acid 4 . 2 . 1 . 1 0 . Glutamate 4 . 2 . 1 . 1 1 . Glycine 4 . 2 . 2 . Other spectra 4 . 2 . 2 . 1 . Tryptophan 4 . 2 . 2 . 2 . Cerebrospinal fluid 4 . 2 . 2 . 3 . Anesthetic 4 . 2 . 3 . Difference spectra 4 . 2 . 3 . 1 . Acetylcholine in D M E M 4 . 2 . 3 . 2 . Acetylcholine in cerebrospinal fluid 4 . 3 . F I B E R - O P T I C P R O B E R A M A N S P E C T R A 4 . 3 . 1 . Neurotransmitter spectra 4 . 3 . 1 . 1 . Acetylcholine 4 . 3 . 1 . 2 . Dopamine 4 . 3 . 1 . 3 . Serotonin 4 . 3 . 1 . 4 . y-Amino butyric acid 4 . 3 . 1 . 5 . Mixture spectra 4 . 4 . D I S C U S S I O N 4 . 5 . R E F E R E N C E S 4 . 1 . I N T R O D U C T I O N The spectroscopic characterization of the individual small-molecule neurotransmitters, the intended environment, and the testing of the probe in vitro with individual neurotransmitters and mixtures with normal Raman spectroscopy, some of the 1 1 basic goals established in Chapter 1, is the purpose of the research described in Chapter 4 . Measuring the normal Raman spectra of the 1 0 small-molecule neurotransmitters would determine their degree of uniqueness and hence their identifiability. Considerable difficulty in differentiating the individual neurotransmitter spectra would render the 45 method unfeasible. The 10 small-molecule neurotransmitters can structurally be subdivided into 3 groups: (i) acetylcholine; (ii) the biogenic amines consisting of dopamine, epinephrine, norepinephrine, serotonin, and histamine; and (iii) the amino acid neurotransmitters aspartate, glutamate, glycine, and y-amino butyric acid. It was therefore expected that the neurotransmitters belonging to the same group would exhibit similar spectra. These spectra were also required as bench marks for later measurements with optical fiber probes in the event that fiber-based measurements produced distortions of the spectra. They could therefore aid in probe development. Both regular collection and fiber-based collection normal Raman spectra are reported in this chapter. Band assignments were made for the reference spectra based on tables of characteristic Raman frequencies by Dollish et al. (1974) and Lin-Vien et al. (1991). As part of the characterization of the intended measuring environment, the normal Raman spectra of cerebrospinal fluid samples were obtained as well. This was necessary in order to ensure that a probe capable of functioning under these conditions was developed. Furthermore, one of these spectra, along with that of a cell culture medium, were used to obtain neurotransmitter difference spectra in biological matrices for use in the signal processing investigations. The matrix, as well as difference spectra, are also reported in this chapter. 4.2. N O R M A L R A M A N S P E C T R A . 4.2.1. Neurotransmitter spectra and band assignments. At the outset, while a Raman/resonance Raman system was not available in-house, access to another Raman instrument was obtained to measure the normal Raman spectra of the small-molecule neurotransmitters. They were measured between 2000 cm\" 1 and 500 cm\"l from the excitation frequency. This range is considered to be effective for the identification of infrared spectra, and, due to the similar information probed by Raman spectroscopy, thus also for Raman spectra. The neurotransmitters were dissolved 46 in physiological saline (0.9% NaCl) to concentrations between 0.1 M and 0.5 M . Raman scattering was excited with the 488 nm line from an Ar +-laser operating at 200 mW and measured with a i m focal length spectrometer (JASCO Model NR-1100, Tokyo, Japan) scanning at 120 cm\" Vmin and with slits set at 500 pm. The spectra were digitized at 1 cm~ 1 intervals resulting in 1501 points per spectrum with approximate signal-to-noise ratios (maximum peak height/noise standard deviation) between 20 and 50. Between 3 and 8 scans of each spectrum were accumulated. The backgrounds of the spectra were removed with a polynomial fitting procedure. More detail about data analysis is given in Part V . 4.2.1.1. Physiological saline. The spectrum of physiological saline (0.9 % NaCl in water) is shown in Figure 4.1. 1.0 r _Q 4 I i i i i i i i i i i i i i i i i i i i 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.1. The Raman spectrum of physiological saline solution. The prominent peak is due to water deformation. 47 The spectrum shows only a single prominent feature near 1630 cm\" 1 assigned to H2O deformation. Physiological saline is routinely used as a vehicle for the administration of drugs to living animals and was often used as solvent for neurotransmitters. 4.2.1.2. Acetylcholine. The Raman spectrum of 0.5 M acetylcholine dissolved in physiological saline is shown in Figure 4.2 and the band assignments of some of the major features of this spectrum are made in Table 4.1. Note the relative intensity of the H2O deformation peak near 1630 cm\"l. •a 3 a <4 Pi CH 3 O CH^-CH.CH^O-C'-CH^ CH, J 1 L 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman Shift (cm\") Figure 4.2. The Raman spectrum of acetylcholine (0.5 M) in physiological saline. Inset shows structural formula. 48 Table 4.1. Table 4.1 shows some of the major peaks in the Raman spectrum of acetylcholine (0.5 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensitv(%,) 646 0-C=0 in plane def. 38 722 C-N symmetric stretch 100 840 C-C stretch 16 877 C-O-C stretch 22 950 -- 23 1451 CH-3 deformation 25 1742 C=0 stretch 13 4.2.1.3. Dopamine. The Raman spectrum of 0.5 M dopamine dissolved in physiological saline is shown in Figure 4.3 and the band assignments of some of the major features of this spectrum are made in Table 4.2. A ring stretch vibrational mode at 1617 cm~l is superimposed on the H2O deformation peak near 1630 cm\"l. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm\"1) Figure 4.3. The Raman spectrum of dopamine (0.5 M) in physiological saline. Inset shows structural formula. 49 Table 4.2. Some of the major peaks in the Raman spectrum of dopamine (05 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensitv(%) 599 ring vibration 22 717 C-N stretch 33 756 ring vibration 36 788 ring vibration 100 953 C-C stretch 24 1293 ring vibration 42 1617 ring stretch 25 4.2.1.4. Epinephrine. The Raman spectrum of 0.5 M epinephrine dissolved in physiological saline is shown in Figure 4.4. The ring stretch vibrational mode at 1617 cm _ l is again superimposed on the H2O deformation peak near 1630 cnr 1. The band assignments of some of the major features of this spectrum are made in Table 4.3. H O '3 3 03 3 3 .0 -2.5 -2.0 •e 1.5 CO 1.0 0.5 I 0.0 -0.5 H a / ^ V C H C H 2 - N H - C H 3 H O ' 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm\") Figure 4.4. The Raman spectrum of epinephrine (0.5 M) in physiological saline. Inset shows structural formula. 50 Table 4.3. Table 4.3 shows some of the major peaks in the Raman spectrum of epinephrine (7J.5 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment IntensitvC^) 762 ring vibration 78 788 ring vibration 100 944 C-C stretch 28 1294 ring vibration 74 1613 ring stretch 46 4.2.1.5. Norepinephrine. The Raman spectrum of 0.5 M norepinephrine dissolved in physiological saline is shown in Figure 4.5. The ring stretch vibrational mode at 1613 cm~l is superimposed on the H2O deformation peak near 1630 cm\"l. Note the similarity to the spectrum of epinephrine in Figure 4.4. The band assignments of some of the major features of this spectrum are made in Table 4.4. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm'1) Figure 4.5. The Raman spectrum of norepinephrine (0.5 M) in physiological saline. Inset shows structural formula. 51 Table 4.4. Table 4.4 shows some of the major peaks in the Raman spectrum of norepinephrine (0.5 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensitv(%l 763 ring vibration 60 781 ring vibration 100 1294 ring vibration 58 1613 ring stretch 46 4.2.1.6. Serotonin. The Raman spectrum of 0.1 M serotonin dissolved in physiological saline is shown in Figure 4.6 and the band assignments of some of the major features of this spectrum are made in Table 4.5. The peak near 1630 cm~l is due to H2O deformation. H _0 5 I 1 1 1 • 1 1 1 • 1 1 1 1 1 1 1 • 1 • 1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.6. The Raman spectrum of serotonin (0.1 M) in physiological saline. Inset shows structural formula. 52 Table 4.5. Table 4.5 shows some of the major peaks in the Raman spectrum of serotonin (0.1 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensitv(%) 762 phenyl ring vibration 79 829 C-N symm. stretch 46 937 C-C stretch 86 1240 -- 47 1348 indole ring vibration 100 1435 N H bending of ring 51 1550 indole ring vibration 72 4.2.1.7. Histamine. The Raman spectrum of 0.5 M histamine dissolved in physiological saline is shown in Figure 4.7 and the band assignments of some of the major features of this spectrum are made in Table 4.6. The shoulder near 1630 cnr 1 is due to H2O deformation. _0 2 l 1 1 1 > 1 < ' • ' 1 < 1 > 1 1 1 1 1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.7. The Raman spectrum of histamine (0.5 M) in physiological saline. Inset shows structural formula. 53 Table 4.6. Table 4.6 shows some of the major peaks in the Raman spectrum of histamine (0.5 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensitv(%) 650 -- 42 860 C-N stretch symm. 26 990 C-C stretch 27 1010 C-N stretch antisym. 29 1108 — 26 1160 ring breathing 46 1236 — 41 1272 ring vibration 100 1310 -- 65 1452 C=N stretch 53 1495 -- 42 1575 C=C stretch 79 4.2.1.8. Aspartate. The Raman spectrum of 0.5 M aspartate dissolved in physiological saline is shown in Figure 4.8 and the band assignments of some of the major features of this spectrum are made in Table 4.7. The peak near 1630 cnr 1 is due to H2O deformation. _0 2 1 1 • 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.8. The Raman spectrum of aspartate (0.5 M) in physiological saline. Inset shows structural formula. 54 Table 4.7. Table 4.7 shows some of the major peaks in the Raman spectrum of aspartate (0.5 M), Raman shift Band assignment Intensitv(%l 825 C-C stretch 100 1738 C=0 stretch 85 4.2.1.9. y-Amino butyric acid. The Raman spectrum of 0.5 M y-amino butyric acid dissolved in physiological saline is shown in Figure 4.9 and the band assignments of some of the major features of this spectrum are made in Table 4.8. The peak near 1630 cm~l is due to H2O deformation. 1.8 r 1.6 --t-» 1.4 -'c 1.2 -(arbitrar 1.0 -(arbitrar 0.8 -• -% r—< 0.6 -i—I u • 0.4 -i 0.2 -P6 0.0 --0.2 L o I. HO / _-CH 2 CH 2 CH 2 -NH 2 _l 1 L 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.9. The Raman spectrum of y-amino butyric acid (0.5 M) in physiological saline. Inset shows structural formula. 55 Table 4.8. Table 4.8 shows some of the major peaks in the Raman spectrum ofy-amino butyric acid Raman shift Band assignment Intensitv(%) 867 C-C stretch 58 897 C-C stretch 49 944 — 32 976 — 42 1052 C-N stretch 26 1314 -(CH 2 ) twist 58 1408 CO2\" symm. stretch 100 4.2.1.10. Glutamate. The Raman spectrum of 0.1 M glutamate dissolved in physiological saline is shown in Figure 4.10 and the band assignments of some of the major features of this spectrum are made in Table 4.9. The peak near 1630 cnr 1 is due to H2O deformation; note the relative intensity. i - H •4—» 1 B .S Pi 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm\") Figure 4.10. The Raman spectrum of glutamate (0.1 M) in physiological saline. Inset shows structural formula. 56 Table 4.9. Table 4.9 shows some of the major peaks in the Raman spectrum of glutamate (0.1 M), Raman shift Band assignment Intensitv(%) 938 C-C stretch 100 1351 CH2 twist 64 1412 CC»2\" symm. stretch 100 4.2.1.11. Glycine. The Raman spectrum of 0.5 M glycine dissolved in physiological saline is shown in Figure 4.11 and the band assignments of some of the major features of this spectrum are made in Table 4.10. The peak near 1630 c m - ' is due to H2O deformation. .C-CHj-NH^ 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm\"1) Figure 4.11. The Raman spectrum of glycine (05 M) in physiological saline. Inset shows structural formula. Table 4.10. Table 4.10 shows some of the major peaks in the Raman spectrum of glycine (0.5 M), their tentative vibrational band assignments, and relative intensities. Raman shift Band assignment Intensity(%) 873 1734 C-C stretch C=0 stretch 100 32 57 4.2.2. Other spectra. 4.2.2.1. Tryptophan. The spectrum of tryptophan, a precursor of serotonin, was also measured to investigate the correspondence between neurotransmitters and some of their precursors and/or metabolites. This spectrum is given in Figure 4.12 (note the similarity to the serotonin spectrum of Figure 4.6). I 3 3 tt 2 •tt 1 c £ 0 --1 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm\"1) Figure 4.12. The Raman spectrum of tryptophan (0.1 M) in physiological saline. 4.2.2.2. Cerebrospinal fluid. CSF was obtained from the lateral ventricle of a male rat under anesthesia (sodium pentobarbitol and ketamine, see Chapter 10) during the implantation of an indwelling cannula as well as removed via the cannula 24 h and 72 h after surgery from the unanesthetized animal. The Raman spectra from these samples were measured and are shown in Figure 4.13. The spectrum taken during surgery differs from those taken after surgery, possibly due to sample contamination with traces of blood hence changing 58 the background fluorescence (e.g. compare to the spectrum of the cell culture medium containing 5% serum as shown in Figure 4.15). 2 5 r 5 I i i i i i i i i i i i i i i i i i i i 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm4) Figure 4.13. The Raman spectrum of cerebrospinal fluid taken during surgery (solid line), 24 h later (dashed line), and 72 h later (bottom line, dotted) Traces displaced for clarity. 4.2.2.3. Anesthetic. A Raman spectrum of the anesthetic (see previous paragraph) employed in the cannula implantation procedure mentioned above was also taken for comparison with the spectrum of CSF taken during surgery to determine whether the presence of the anesthetic could be detected in the CSF sample. No sign of the anesthetic was evident in the spectrum of CSF taken during surgery. No attempt was made to dilute the anesthetic to physiological concentrations. The spectrum of the anesthetic is shown in Figure 4.14. 59 0.4 w -o . i -400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.14. The Raman spectrum of the anesthetic (see text) used during surgery. 4.2.3. Difference spectra. The Raman spectra of two biological matrices were measured before and after having been spiked with acetylcholine. The spectrum taken before spiking was then subtracted from that taken after spiking to obtain an acetylcholine difference spectrum. The collection parameters were as described above. 4.2.3.1. Acetylcholine in D M E M . A Raman spectrum was taken of Dulbecco's modified Eagle medium (DMEM) * with 5% calf serum and 0.2% sodium azide (as preservative). D M E M is a complex cell culture medium consisting of several inorganic salts, amino acids, vitamins, and other nutrients. The Raman spectrum of D M E M , shown in Figure 4.15, exhibits a strong fluorescence band. The sample was subsequently spiked with acetylcholine to a concentration of approximately 0.5 M and the Raman spectrum measured again (also 60 shown in Figure 4.15). The difference spectrum of acetylcholine in D M E M was obtained by subtracting the two spectra and it is shown in Figure 4.15. Figure 4.15 (bottom line) clearly resembles the acetylcholine spectrum shown in Figure 4.2, especially the 722 cm\" 1 and 646 cm\" 1 peaks. i . i . i . i . i . i . i . i . i . i 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm 1 ) Figure 4.15. The Raman spectrum of DMEM (solid top line), DMEM spiked with acetylcholine to 0.5 M (dotted line), and the difference spectrum (bottom line). 4.2.3.2. Acetylcholine in CSF. The Raman spectrum of CSF taken during surgery is shown in Figure 4.13. This sample was subsequently spiked with acetylcholine to an approximate concentration of 0.2 M and then remeasured. The spectrum is shown in Figure 4.16 and the difference spectrum of acetylcholine in CSF is shown in the same figure. The difference spectrum can be compared to the spectrum of 0.5 M acetylcholine dissolved in physiological saline shown in Figure 4.2. 61 -1 -I i i i i i • i . i i i i i i i i i i i 400 600 800 1000 1200 1400 1600 1800 2000 2200 Raman shift (cm1) Figure 4.16. The Raman spectrum of cerebrospinal fluid taken during surgery (shown in Figure 4.13) spiked with acetylcholine to 0.2 M (top trace), and the difference spectrum (bottom trace). 4.3. F I B E R - O P T I C P R O B E R A M A N S P E C T R A . The spectra of four neurotransmitters were obtained with a fiber-optic probe. The spectra were excited with the argon ion laser producing 200 mW at 514.5 nm of which 60% was coupled into the probe, thus giving an effective power of 120 mW at the sample, and measured with the single monochromator's entrance slit set at 250 pm. The Raman shift was calculated from the pixel (or diode) number with the following formula: S = C W N - R(P - CP) (4.1) where S is the Raman shift (cnr 1), C W N is the monochromator center wavenumber setting (cnr 1), P is the pixel number, CP is the number of the center pixel (530, see Chapter 3), and R is the wavenumber to pixel ratio (cm-Vpixel) at the center wavenumber based on the instrument's dispersion (0.031 nm/pixel, see Chapter 3). The probe used consisted of a small excitation fiber (core 50 W), to continuous wave sources. 7.2.2.1. Continuous wave light. Because high average powers were not used (200 mW maximum), coupling the beam from the A r + laser into optical fibers never produced any damage. 7.2.2.2. Pulsed light. Coupling pulsed light into optical fibers was more difficult and generated several modes of failure. To prevent damage due to focusing inside the fiber, the beam was focused in front of the fiber. At high fluences an air plasma was generated which absorbed part of the laser beam as mentioned above. Under these conditions, the fiber was observed to fracture approximately 1 cm from the input face which was reminiscent 9 8 of a mode of failure described by Allison et al. (1987). Pitting of the input end was also observed. Therefore, the external focus should be some distance away from the fiber endface yet have expanded to no more than about 80% of the core diameter (De Hart, 1992). The use of a longer focal length lens (150 mm) improved the coupling efficiency, prevented the formation of an air plasma, and allowed positioning of the fiber further from the focal point, thus avoiding damage to the fiber. However, focusing the beam in front of the fiber does not produce as efficient coupling as focusing inside the fiber as described above. Therefore, focusing inside the fiber to improve the coupling efficiency while using low pulse energies and a long focal length lens appears desirable. The use of a fiber-optic bundle to avoid high fluences was also investigated. The input face area of a single excitation fiber could be multiplied by using several excitation fibers of the same diameter at the cost of increasing the probe diameter by the diameter of a single excitation fiber. For instance, using 6 excitation fibers of 300 pm diameter surrounding a single collection fiber of the same diameter increases the input surface area 6 times but the total diameter only 1.5 times. The optical damage threshold of a 4x300 pm input bundle of 6 cm length was tested under increasing pulse energies and the results are shown in Table 7.2. No visible surface or transmission damage was noted up to the maximum pulse energies,tested (12 mJ/pulse). In contrast, damage occurred to a o 600 pm diameter fiber at pulse energies of about 5 mJ/pulse. Disadvantages of using bundles are that it is impossible to avoid coupling light into the cladding thus reducing throughput and risking absorption-induced damage to the probe, that the open spaces between the fibers will cause a further reduction in throughput and may make the bundle more susceptible to damage, especially with small diameter fibers. It should also be noted that the power distribution of a Gaussian beam would cause fibers situated centrally in the bundle to receive a disproportionate amount of the pulse energy, that is, the input energy is not evenly distributed among the fibers in a 99 bundle. The beam diameter of a Gaussian beam is defined by 1/e2 or 2 standard deviations from the mean. The beam intensity has dropped to about 60% of the maximum value at 1 standard deviation from the mean and about 70% of the throughput in the area defined by 1 standard deviation. This suggests that coupling the center 25% of a Gaussian beam cross-sectional area into a bundle would yield a more even distribution of light amongst the fibers with about 30% light loss. Table 7.2. The effects of increasing pulse energies on the transmission through a 4x300 [im input bundle of 6 cm length. Excitation at 266 nm and 10 Hz. Oscillator (V) Bundle output (ml) Std. errorfmJ) 590 0.131 0.005 600 0.207 0.005 610 0.329 0.009 620 0.445 0.019 600 0.213 0.007 630 0.651 640 0.816 0.016 600 0.214 0.007 640 0.835 0.015 650 1.083 0.024 600 0.188 0.004 660 1.337 0.038 670 1.420 0.044 600 0.206 Another way to increase the input area to facilitate light coupling and minimize the risk of damage associated with high fluences, is to use a tapered fiber. These fibers gradually taper from one diameter to another over a certain fiber length. This allows one to use a large input face area without sacrificing the probe diameter. A probe employing a tapered excitation fiber is described and evaluated inChapter 8. 100 7.2.3. Light transmission The quoted attenuation for U V transmission for fibers from different manufacturers vary considerably. For example, at 200 nm fiber from Polymicro Technologies has an attenuation of about 1 dB/m while fiber from Fiberguide Industries has an attenuation of about 2 dB/m. These values represent an attribute which could be termed 'static' attenuation. Static attenuation results from Rayleigh scattering in the fiber and absorption by dopants and impurities in the fiber. For use in the deep U V , an equally or more important 'dynamic' attenuation exists due to the formation of reversible and irreversible color centers in the fiber (e.g. Karlitschek et al., 1996; Klein et al., 1995; Toriyaetal , 1995). This type of attenuation, however, was not mentioned by the fiber manufacturers and its discovery was somewhat unexpected. There is some evidence that manufacturers are aware of this issue and that they have started to address it, driven by the need to find optical fibers suitable for the delivery of excimer wavelengths (e.g. Klein et al., 1995). Recently, Polymicro Technologies have developed an experimental fiber aimed at reducing dynamic attenuation and made some of it available for testing (see Figure 7.5). Another fiber manufacturer, 3M, is also actively engaged in developing optical fiber suitable for pulsed U V transmission. 7.2.3.1. Continuous wave excitation. Continuous wave light transmission at 364 nm exhibited more static attenuation than at visible wavelengths but dynamic attenuation was not noticed. 7.2.3.2. Pulsed excitation. Pulsed light transmission was severely hampered in the U V by both static and dynamic attenuation. The static attenuation was characterized by the manufacturer's specifications, but the dynamic attenuation was unknown until a transmission time-101 course study revealed unexpected attenuation. Figure 7.2. shows the attenuation with time for a 15 cm section of 300 pm diameter Fiberguide Industries fiber excited with a pulsed N d : Y A G at 266 nm, 10 Hz repetition rate, and averaging 100 pulses per point. The recovery exhibited by the fiber after a 5 minute rest period is also shown. It is evident that complete recovery occurred for this fiber and under these conditions. 0.41 r 0.40 -0.39 0.38 0.37 53 ft 0.36 0.35 0.34 0.33 -100 0 100 200 300 400 500 600 700 800 3700 3800 Pulses Figure 7.2. The attenuation for a 15 cm section of300 [im diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 100 pulses per point, as a function of time. The recovery exhibited by the fiber after a 5 minute rest period is also shown (last measurement). The attenuation was found to be a function of the pulse energies as can be seen from Table 7.3. Higher pulse energies excited a greater attenuation resulting in a reduced transmission efficiency. This was further confirmed by investigating the degree of attenuation as a function of the distance between the input face and the beam waist. The results (given in Figure 7.3) showed that the attenuation was less severe when the fiber 102 was a greater distance from the beam waist (thus subjected to decreased energy densities). Table 7.3. The effects of increasing pulse energies on the transmission efficiency (average of 10 pulses) through a 15 cm section of 300 Jim diameter Fiberguide Industries Superguide G fiber. Excitation at 266 nm and 10 Hz. Oscillator (V) Input (mJ) Output (mJ) Efficiency (%) 570 0.248 0.126 50.8 580 0.571 0.255 44.7 600 1.821 0.495 27.2 0.35 h 0.30 -X 0.25 h I & 6 0.201-0.15 -0.10 11 12 13 14 15 Distanoc from beam waist (mm) 16 17 Figure 7.3. The attenuation for aim section of 600 \\im diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 10 pulses per point as a function of the distance between the input face and the beam waist. The cause of this dynamic behavior was not known and hypothesized to be due to temperature effects. To eliminate the possibility that misalignments due to mechanical distortions caused by temperature changes contributed to this phenomenon, the degree of dynamic attenuation was tested with either 2 mm or 6 mm of the bare fiber protruding beyond the supporting point of the fiber-holding chuck. As Figure 7.4 shows, the 103 dynamic attenuation was not caused by mechanical instability and was due to other causes. A literature search suggested that color center formation in the fibers (e.g. Klein et al., 1995; Kubo et al., 1994; Toriya et al., 1995) may have been the cause of dynamic attenuation. Figure 7.4. The attenuation for a 15 cm section of 300 pm diameter fused silica fiber excited at 266 nm, 10 Hz, and averaging 100 pulses per point as a function of time. The fiber tip extended freely either 2 mm (top) or 6 mm (bottom) from its support. In neither case was the dynamic attenuation affected. Subsequent testing at 225-235 nm and -0.20 mJ/pulse also produced strong dynamic attenuation in low U V attenuation fibers from Polymicro Technologies, 3M, Oriel Corporation, and Fiberguide Industries, some of which are superimposed in Figure 7.2 for comparison, and appeared to vary based on the manufacturer. A direct comparison also revealed a dependence of dynamic attenuation on wavelength. Whether dynamic attenuation varied with respect to placing the beam inside or outside the fiber was not investigated. 104 ' i i i i i i i i i i I i I i i i i i -200 0 200 400 600 800 1000 1200 1400 1600 Time(s) Figure 7.5. The attenuation (mean ± S.E.M.; n = 10) for 15 cm sections of300 [im fused silica fibers from Polymicro Technologies (dotted line), 3M (dashed line), Fiberguide Industries (dash/dot), and the new experimental fiber from Polymicro Technologies (200 [im, solid line) at 230 nm and 10 Hz as a function of time. The last point in each of the two top traces were obtained after a 5-minute recovery period. It can be seen from Figure 7.5 that the experimental fiber, despite its smaller diameter, outperformed the other fibers both in terms of decay and recovery characteristics. The new fiber was further characterized with regard to recovery time, throughput and efficiency, as well as attenuation. A 20 cm segment of 200 pm diameter experimental fiber was tested with approximately 75 pj input energy at 230 nm and 10 Hz to determine its recovery behavior. The results are shown in Figure 7.6 and indicated that a resting period of 1 minute produced nearly full transmission recovery. 105 Recovery time (minutes) Figure 7.6. The transmission efficiency (mean + S.E.M.; n - 10) of an experimental fiber (20 cm section, 200 \\x.m) as a function of recovery time at 230 nm and 10 Hz. The efficiency of the same section of fiber was subsequently tested as a function of input energy. The input energy was incremented from about 80 pj/pulse to about 250 mJ/pulse in 10 steps without any intervening recovery periods. The results, shown in Figure 7.7, indicated that the throughput efficiency of the fiber declined with increasing pulse energies (consistent with data given in Table 7.3 for a different fiber). Output energies gradually increased and eventually exceeded 50 pi/pulse. Although output energies in excess of 50 pj/pulse could be achieved and sustained via a gradual increases in input energy, it could not be done with a single very large increment. Less large increments produced initial sharp declines in fiber throughput as shown in Figure 7.2 and, for comparison, again in Figure 7.8. Small increments did not seem to produce these declines at all. 106 55 50 h 45 40 \\t 3 5 EP «S 3 0 25 20 15 - i 55 - 50 - 45 40 _ i _ _ i _ _ i _ _ i _ 35 §• 30 ^ 25 20 60 80 100 120 140 160 180 200 220 240 260 280 Pulses 15 300 Figure 7.7. The transmission efficiency (dotted line) and output pulse energies (solid line) of an experimental fiber (20 cm section, 200 \\xm) as a function of input energy at 230 nm and 10 Hz (mean ± S.E.M.; n = 10). For example, when coupling about 203.43 ± 1.57 pJ/pulse (representing a 200 pj/pulse increase) into a fresh 20 cm section of experimental fiber, the fiber output dropped to below the detection threshold after the first reading of 40.40 ± 0.35 mJ/pulse. When the experiment was terminated 5 minutes later, the output energy was still below the detection threshold. In contrast, an input of 217.50 ± 1.81 mJ/pulse (representing a 28 pj/pulse increase), produced an output of 48.86 ± 0.16 mJ/pulse for at least 2 minutes when the previous input energy levels were attained via gradual increases. The moderately large pulse increase consisted of an input of 77.17 ± 1.15 mJ/pulse and produced a maximum decline of about 13%. These results indicated that a gradual step-wise increase in input energy was required to achieve high energy throughputs. 1 0 7 42 -40 -38 36 34 h 32 30 -28 --I- -P • 0 20 40 60 80 100 120 Time (s) Figure 7.8. The output energy as a function of time for fibers tested under 3 conditions: increasing the input energy with small increments (e.g. oscillator voltage from 640-646; solid line); with a moderate increment (590-615 V; dashed line), and with a large increment (590-639 V; dotted line) to the final operating level. Data were scaled to the same initial values. 7.2.4. Beam profiles. The emission profiles from optical fibers of varying diameter were measured as part of the optical fiber characterization. Light from a HeNe laser (5 mW) was coupled into an optical fiber of given diameter and the emitted light was projected onto a Reticon 4096 element diode array detector (EG&G, Princeton, NJ). The fiber, 133 mm from the array, was mounted onto a translation stage that could be adjusted to move the beam incrementally across a mask containing a 1 mm slit covering the array. The readout from the array was moved to a microcomputer using Computerscope software via a Reticon 1020 interface and a ISC-16 data acquisition board. A C program (by L.S. Greek) was used to convert the binary data to ASCII format. The results indicated that the emission 108 beam profiles from smaller diameter fibers approached Gaussian distributions, but larger diameter fibers showed increasingly bimodal beam profiles. The central section beam profile of a 300 pm fiber is shown in Figure 7.9. Figure 7.9 reveals that although the beam profile of a 300 pm fiber can still be approximated by a Gaussian distribution, some bimodality is already evident. Although it was known that the different lasers used exhibited radically different beam profiles, time did not permit investigating the dependencies of the emission profile on the profile of the input beam and on other coupling parameters. Such dependencies could reasonably be expected to exist. j i i i i i i i i i i i i 0 200 400 600 800 1000 1200 Pixel Figure 7.9. The central section emission beam profile of a 300 \\lm fused silica fiber with a Gaussian fit (dashed line) superimposed for comparison. 7.3. D I S C U S S I O N . The central issues regarding the use of optical fibers for U V resonance Raman spectroscopy are those of physical damage to fibers, light coupling into the fibers, and light propagation by the fibers. 109 The findings reported here concerning fiber damage, light coupling, and dynamic and static attenuation, were in general agreement with those found in the literature (Allison et al., 1987; Klein et al., 1995; Kubo et a l , 1994; Toriya et al., 1995; Taylor et al., 1988). However, due to variations in fiber attributes based on the manufacturer and due to the absence of critical information on the supplier's specification sheets and in the literature, individual fiber characteristics had to be determined. From the reports mentioned above, it was learned that the dynamic attenuation was due to color center formation in the fiber. Therefore, fibers with a weak tendency to form color centers should be used as excitation fibers when fabricating an optical fiber probe. Physical damage was shown to occur at high pulse energies and be aggravated by short focal length lenses, consequently, these are best avoided. Lower pulse energies will allow the beam to be focused inside the fiber to improve light coupling into the fiber as shown in Table 7.1. The use of a long focal length lens would avoid damage at first bounce and first internal focus due to modal spreading. The best method for coupling light into an optical fiber may be to use a Galilean telescope to reduce the input beam diameter to somewhat less than that of the fiber core. With this arrangement, no focusing of the beam would occur anywhere in the optical path. The transmission of light through a fiber was given by the Equation 6.2 where the linear attenuation has previously been termed 'static' and the UV-induced attenuation 'dynamic'. The dynamic attenuation was found to be a function of the pulse energies as can be seen from Table 7.3, number of transmitted pulses, light frequency, and laser repetition rate. Equation 6.2 could be restated as: - dl/dz = (a0 + Aad(p, v, R))I + p(t))I2 (7.1) where the symbols are as defined before and p is the number of pulses transmitted, v the light frequency (Hz), and R the laser repetition frequency (Hz). The transmission coefficients are now being determined for the new experimental fibers from Polymicro Technologies. 110 Using lower pulse energies would therefore result in less dynamic attenuation and greater average throughput. Taken together, low pulse energies will avoid fiber damage, enable efficient light coupling into the fiber, and ensure more efficient light propagation by the fiber. Energy delivery could be maximized by using the new fibers and by gradually increasing the input energy to avoid catastrophic failure due to the sudden transmission of high energy pulses. Additional methods to increase the average power delivered to the sample should be investigated or other means found to deliver high energy pulses to the sample. Increasing the pulse repetition rate or extending high energy pulses could be used to increase the average power. Since the laser repetition rate of the laser used here cannot be increased beyond 20 Hz, the need to find a suitable pulse extender becomes all the more important given the many problems it would ameliorate. Two recently developed resonance Raman instruments boost average power along these very lines: one employs a continuous wave (infinitely extended pulse) intracavity doubled argon laser (FRED) as a U V source (Russell et al., 1995) and the other a tripled/quadrupled high repetition rate Thsapphire laser (Manoharan et al., 1994). Some researchers have also had success in decreasing dynamic attenuation thus boosting fiber throughput by annealing the fiber before use (Toriya et al., 1995). Concerning alternative light guides, hollow, small-bore light pipes with aluminized reflecting surfaces could potentially be used. Short sections (3-5 mm) of fused silica or sapphire optical fiber could be made into windows for the hollow light guides for use in solutions. These windows could be shaped (e.g. lens or angle), anti-reflection coated, and provided with reflecting surfaces if required. Silver coated small-bore glass waveguides have recently been developed for light delivery from CO2 and E r : Y A G lasers (Matsuura et al., 1995). I l l The work reported in this chapter provided the information regarding fiber damage, light transmission characteristics, and input coupling necessary to proceed with the design and fabrication of actual optical fiber probes. 7.4. R E F E R E N C E S . Allison, S. W.; Cates, M . R.; Gillies, G. T.; Noel, B. W. 1987. Fiber optic pulsed laser delivery for remote measurements. Optical Engineering: 26, 538-546. Allison, S. W.; Gillies, G. T.; Magnuson, D. W.; Pagano, T. S. 1985. Pulsed laser damage to optical fibers. Applied Optics, 24: 3140-3144. Beck, T.; Reng, N . ; Richter, K . 1993. Fiber type and quality dictate beam delivery characteristics. Laser Focus World: October, 111-115. Chang, R. S. F.; Ge, Z.; Djeu, N . 1995. UV-visible characteristics of sapphire fibers grown by laser-heated pedestal growth technique. In J. A . Harrington; D. M . Harris; A . Katzir (eds.) Proceedings of biomedical optoelectronic instrumentation SPIE2396: 145-150. De Hart, T. 1992. Where and why optical fibers fail... and how to prevent it. Photonics Spectra, November: 107-110. Karlitschek, P.; Klein, K.-F. ; Hillrichs, G.; Grzesik, U . 1996. Improved UV-fiber for 193 nm excimer laser applications. SPLE 2677: paper 16, in press. Klein, K.-F. ; Hillrichs, G.; Karlitschek, P.; Grzesik, U . 1995. Improved optical fibers for excimer laser applications. LASERmed 95: paper 107b, in press. Krohn, D. A. ; McCann, B. P. 1995. Silica optical fibers: Technology update. In J. A . Harrington; D. M . Harris; A . Katzir (eds.) Proceedings of biomedical optoelectronic instrumentation SPLE 2396: 15-24. Kubo, U . ; Hashishin, Y . ; Nakano, H. ; Nakayama, T. 1994. ArF excimer laser beam delivery systems for medical applications. In J. A . Harrington; D. M . Harris; A . Katzir (eds.) Proceedings of biomedical fiber optic instrumentation SPLE 2131: 28-34. Manoharan, R.; Wang, Y . ; Boustany, N . ; Brennan, J. F.; Baraga, J. J.; Dasari, R. R.; Van Dam, J.; Singer, S.; Feld, M . S. 1994. Raman spectroscopy for cancer detection: Instrument development and tissue diagnosis. SPLE 2328: 128-132. Matsuura, Y . ; Abel, T.; Harrington, J. A . 1995. Optical properties of small-bore hollow glass waveguides. Applied Optics 34: 6842-6847. 112 Russell, M . P.; Vohnik, S.; Thomas, G. J. 1995. Design and performance of an ultraviolet resonance Raman spectrometer for proteins and nucleic acids. Biophysical Journal 68: 1607-1612. Taylor, R. S; Leopold, K . E.; Brimacombe, R. K.; Mihailov, S. 1988. Dependence of the damage and transmission properties of fused silica fibers on the excimer laser wavelength. Applied Optics 27: 3124-3133. Toriya, T.; Kaneda, K.; Tsumanuma, T.; Sanada, K. 1995. Characteristics of optical fiber for high power excimer laser. In J. A . Harrington; D. M . Harris; A . Katzir (eds.) Proceedings of biomedical optoelectronic instrumentation SPIE 2396: 138-144. 113 C H A P T E R 8 8.1. I N T R O D U C T I O N 8.2. F I B E R - O P T I C P R O B E D E S I G N 8.2.1. Design considerations and simulations 8.2.1.1. Model parameters 8.2.1.2. Excitation and collection fiber sizes 8.2.1.3. Fiber tip modifications 8.2.1.4. Bundles 8.3. F I B E R - O P T I C P R O B E F A B R I C A T I O N 8.3.1. End-face preparation 8.3.2. Probe assembly 8.3.2.1. Dual fiber probes 8.3.2.2. Multi-fiber probes 8.3.3. Glue 8.4. F I B E R - O P T I C P R O B E E V A L U A T I O N 8.4.1. Working curves 8.4.1.2. Excitation and collection fiber sizes 8.4.1.3. Fiber tip modifications 8.4.1.4. Bundles 8.4.1.5. A side-casting bundle 8.5. D I S C U S S I O N 8.6. R E F E R E N C E S 8.1. I N T R O D U C T I O N The aim of the interdisciplinary work undertaken and reported in this thesis was to conceptualize, design, and lay the foundations for the development of a brain sensor capable of meeting the general requirements of sensitivity, speed, and versatility while operating non-invasively in an aqueous in situ environment. Therefore, the basic goals consisting of the design and development of a suitable fiber-optic probe for in vivo operation and the determination of the conditions for its optimum use (see Chapter 1) were of pivotal importance to this project. The present chapter discusses the investigations concerning the optimal geometry of a fiber-optic probe for resonance Raman work in vivo. These investigations consisted of computer simulations of probe performance, researching different fabrication procedures, and evaluations of the probe in vitro. For the theoretical background, the reader may wish to refer to Chapter 6. 114 8.2. F I B E R - O P T I C P R O B E D E S I G N . Criteria important in the design of optical fiber probes can sometimes be investigated with the aid of simulations and has been done for single (e.g. Zhu and Yappert, 1992a) and double fiber sensors (e.g. Zhu and Yappert, 1992b; Schwab and McCreery, 1984). Such simulations can then be experimentally verified and used to optimize the probe design. Some design considerations and two simulations are discussed in the next sections: an intensity profile and collection efficiency simulation and a more fine-grained probe simulation (the latter generated by L.S. Greek and evaluated by us, see Greek et al., 1996). 8.2.1. Design considerations and simulations. When using optical fibers for Raman spectroscopy, Raman and luminescence signals generated in the optical fiber core or cladding should be considered. Prominent Stokes shifted silica Raman features which can interfere with analyte signals occur at 1535, 817, and 516 c m - ' and are pronounced in long fibers (Kercel et al., 1990), while the anti-Stokes band near 500 cm\"l has also been observed (Gambling and Poole, 1988). In addition, strong Rayleigh scattering by the fiber may completely obscure the much weaker Raman signals generated by analytes (Heiman et al., 1989). Although the use of a single fiber for both excitation and collection is efficient because of a complete overlap of excitation and collection volumes the interference caused by Raman and Rayleigh scattering from the fiber itself requires a dual or multiple fiber arrangement (Myrick and Angel, 1990). In a dual fiber geometry, one fiber is used for delivery of the excitation radiation, while another fiber is used to collect the scattered radiation from the sample. Myrick et al. (1990) found the collection efficiency in a very dilute rhodamine 6G solution (-30 pM) to be optimal for small angles between collection and excitation fibers and poor for right 115 angle collection. The same advantage also benefits single excitation, multiple collection fiber arrangements. Such dual (Heiman et al., 1989) and multiple fiber (Schwab and McCreery, 1984; Wang et al., 1992) probes have found applications in remote Raman spectroscopy. A forward scattering geometry is also efficient, but requires filters to remove the laser line (Myrick et al., 1990) and Raman lines produced by the fiber (Ma and L i , 1994) which could be impractical where very small probes are needed. For these reasons, it was decided to use separate excitation and collection fibers. As part of the characterization of the intended measuring environment, the normal Raman (Chapter 4) and U V absorption (Chapter 5) spectra of cerebrospinal fluid samples were obtained. Although normally a clear fluid, the U V absorption spectrum of CSF (Figure 5.6) indicated that it was highly absorbing in the region of interest. It was therefore necessary to allow for the effects of a highly absorbing environment in the probe design. 8.2.1.1. Model parameters. A knowledge of the excitation intensity profile was required in order to aid in the determination of the optimum probe geometry. In order to obtain this profile, a coarse-grained simulation was generated. The detailed simulation available (Greek et al., 1996) did not accommodate differences in probe tip geometry and did not allow for a visualization of the excitation light intensity distribution in the analysis volume. Parameters that could be varied in the simpler model included excitation and collection fiber diameters, separation distance between these fibers, numerical aperture, sample absorptivity, intensity gradations of the excitation light, and fiber tip geometry. The source code for this program (in Basic 4.0, Microsoft Corporation, Redmond, WA) is given in Appendix A. A gaussian beam profile with centred mean and 3 standard deviations across the fiber radius was assumed for the exiting light, but could be changed if required. This assumption was justified by the results obtained from measuring the 116 emission profiles from optical fibers as discussed in Chapter 7. Figure 8.1 shows the simulated intensity distribution of light emerging from the excitation fiber of a front-casting probe. Figure 8.1. The simulated intensity distribution of light emerging from the excitation fiber of a front-casting probe. The profile simulation program was used to calculate the efficiencies of fiber optic probes with flush or 45° angled excitation tips in samples with different absorbances. This calculation was not integrated over the entire analysis volume, but consisted only of the central section through the analysis volume. In addition, the maximum sample depth on which the calculations were based, was only 533 pm. It was assumed that the collection fiber collected photons as long as they were scattered from the overlapping region between the excitation and collection fiber cones. This was an oversimplification which tended to overestimate the collection efficiency of probes where the fibers were aligned flush and overestimated to a lesser degree the efficiency of probes 117 with angled excitation fibers. Furthermore, the likelihood of a photon being collected by the collection fiber depended on the intensity of the light in the volume element where it was scattered from and on the distance to the collection fiber endface. Excitation and scattered photons were in a 1:1 ratio. However, scattering, though in direct proportion to the excitation light intensity, has a very much smaller probability. For these reasons, the results obtained with the coarse-grained simulation were mostly of qualitative value. 8.2.1.2. Excitation and collection fiber sizes. The effects of varying sizes of excitation and collection fibers were investigated with the model. The results are given in Table 8.1. Table 8.1. The effects of varying excitation and collection fiber diameters (fJm) on the relative collection efficiencies (%) of dual fiber probes for maximum probe depths of500 pm. The separation distances between fibers were a function of their diameters. Excitation (j) (pm) Collection <|) (pm) Separation (pm) Efficiencvf%) 100 100 20 17.2 100 200 30 13.4 100 300 40 10.5 100 100 20 17.2 200 100 30 6.4 300 100 40 3.4 The results show that for a given diameter collection fiber, smaller excitation fibers are more effective. This is consistent with the fact that with smaller excitation fibers all the light energy gets \"compressed\" into a volume closer to the collection fiber which improves collection efficiency. Using larger diameter collection fibers given a certain excitation fiber diameter leads to small decreases in collection efficiency. This makes little conceptual sense because with larger diameter collection fibers, scattered light ought to be collected more easily. The latter result therefore reflects only the effects of increased separation distances between the fibers. 118 8.2.1.3. Fiber tip modifications. Due to internal filtering in samples with pronounced molar absorptivity, it is important for the excitation volume to be in close confinement to the collection fiber endface. This can be achieved by side-casting the excitation light across the collection surface. Figure 8.2 shows the intensity distribution of light emerging from the excitation fiber of a side-casting probe. Figure 8.2. The simulated intensity distribution of light emerging from the excitation fiber of a side-casting probe. Although the total analysis volume of the side-casting probe is smaller than that of the front-casting probe, a comparison of Figures 8.1 and 8.2 reveals that those areas of the analysis volume of the side-casting probe excited by the most intense light fall within the acceptance cone of the collection fiber in contrast to the other arrangement where it falls mostly outside of the collection cone. A comparison of Figures 8.1 and 8.2 further reveals 119 that the average distance from a volume element in the analysis volume of the side-casting probe to the collection face is shorter than the corresponding distance in the front-casting probe. Both considerations favor the side-casting probe in absorbing solutions and possibly in arrangements with restricted sample volumes. Table 8.2 shows the relative collection efficiencies of front and side-casting probes in samples with different absorbances. Table 8.2. The effects of probe tip geometry on the relative collection efficiencies of dual fiber probes for maximum probe depths of500 pm and in solutions with varying absorbances. The separation distances between fibers were 20 pm. Excitation (j) (pm) Collection § (\\xm) Absorbancef%) Efficiencv (%) Front-casting probe 100 100 0 1.55 100 100 50 0.84 100 100 90 0.00 Side-casting probe 100 100 0 38.86 100 100 50 41.34 100 100 90 33.77 An inspection of Table 8.2 confirms that the side-casting probe is much more efficient than the front-casting probe in samples with limited probe depth, especially absorbing samples. The model was also used to investigate the positioning of the collection fiber endface relative to the excitation fiber tip. These results are shown in Table 8.3 and indicated that the collection efficiency of the probe increased as the collection fiber endface was moved away from the excitation fiber tip. Moving the endface of the collection fiber away caused more overlap between excitation and collection volumes. An optimum could not be attained with the current model, but is expected to occur at a distance where the entire output 'endface' (the fiber surface area over which the excitation fiber emits light) of the excitation fiber falls within the collection cone of the collection fiber. 120 Table 8.3. The effects of probe geometry on the relative collection efficiencies of dual side-casting fiber probes for maximum probe depths of500 pm. The separation distances between fibers were 20 pm and that between collection fiber endface and excitation fiber apex was varied as indicated. Excitation