Total luminescence spectroscopy for selectivity enhancement in multicomponent analysis

MISSING IMAGE

Material Information

Title:
Total luminescence spectroscopy for selectivity enhancement in multicomponent analysis
Physical Description:
viii, 307 leaves : ill. ; 28 cm.
Language:
English
Creator:
Inman, Eugene Lee, 1956-
Publication Date:

Subjects

Subjects / Keywords:
Luminescence spectroscopy   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1982.
Bibliography:
Includes bibliographical references (leaves 298-306).
Statement of Responsibility:
by Eugene Lee Inman, Jr.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000318239
notis - ABU5072
oclc - 08939832
System ID:
AA00003857:00001

Full Text












TOTAL LUMINESCENCE SPECTROSCOPY FOR SELECTIVITY ENHANCEMENT
IN MULTICOMPONENT ANALYSIS










BY

EUGENE LEE INMAN, JR.


A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA
1982














This dissertation is dedicated to the two most important people

in my life. My wife and companion, Wanda, has been a continuous

source of encouragement, love, and understanding throughout my

educational career. My son, Joshua, has added challenge to my

life, and has taught me numerous principles over the past year.

His continuous curiosity and effort to explore and understand

epitomizes the need for individual intellectual improvement and

personal growth.

Wanda and Joshua have helped to place my scientific efforts

into perspective both in this life, and the life to come.


















The mind of the intelligent seeks knowledge, but the mouth

of fools feeds on folly.

Proverbs 15:14













ACKNOWLEDGEMENTS

The author would like to thank Dr. James D. Winefordner for his

guidance, encouragement, support, and friendship throughout this

research effort. His example of scientific and individual stature

has been a model for emulation. The JDW research group, both as a

whole and as individual members, has been a continuous source of

friendship and daily challenge. The author would like to especially

thank Dr. Edward Voigtman for scientific exchange and thought-provoking

discussions, many of which resulted in total disagreement. He would

also like to thank his wife, Wanda, for her assistance in putting

this work in its present form. Finally, the author would like to

thank the faculty members who have served on his advisory committee

for their help and interest throughout this educational process.














TABLE OF CONTENTS

CHAPTER PAGE

ACKNOWLEDGEMENTS ...................................... iv

ABSTRACT ........................................... vii

1 INTRODUCTION TO MOLECULAR LUMINESCENCE SPECTROSCOPY .... 1

Energy Model ........................................ 1
Stokes Shift ........................................ 6
Advantages of Molecular Luminescence Spectroscopy .... 7
Spectral Overlap ..................................... 11
Note .............................................. 22

2 TOTAL LUMINESCENCE SPECTROSCOPY ........................ 23

Theory .............................................. 23
Instrumentation ................................. ...... 25
Fluorimeter .................................. ...... 28
Computer Hardware .................................. 37
Computer Software .................................. 37
Spectral Features .................................... 46
Conclusions .......................................... 62

3 SYNCHRONOUS LUMINESCENCE SPECTROSCOPY .................. 67

Theory ......................... .......... ............ 67
Instrumentation ..................................... 75
Constant Energy Synchronous Fluorescence
Spectroscopy ...................................... 76
Background ........................................ 76
Raman Scatter Reduction ............................ 80
Introduction ..................................... 80
Instrumentation .................................. 89
Results and Discussion ........................... 89
Conclusions ...................................... 106
PAH Analysis ........................................ 106
Introduction .................................... 106
Ternary Mixture ................................. 121
Complex Mixture ................................. 134
Synchronous Profiles ............................ 155
Conclusions ......................................... 160







4 LOW TEMPERATURE LUMINESCENCE SPECTROSCOPY .............. 165

Introduction .. ... .. ... ..... ............. ........ 165
Instrumentation .,,............ .................... 170
Results and Discussion .............................. 175
Phosphorescence Spectroscopy ........................ 211
Synchronous Profiles ................................ 215
Conclusions ....................................... .. 220

5 MATHEMATICAL AND STATISTICAL METHODS ................... 221

Introduction ........................................ 221
Basic Matrix Principles ............................. 222
Data Smoothing ....................................... 225
Multiple Linear Regression .......................... 226
Factor Analysis .................................... 232
Spectral Ratioing ................................... 235
Correlation Methods ................................. 238
Introduction ....................................... 238
Experimental ...................................... 239
Background ........................................ 239
Results and Discussion ............................ 240
Conclusions ....................................... 257

6 APPLICATIONS IN OIL ANALYSIS ........................... 258

Introduction ........................................ 258
Total Luminescence Spectroscopy ...................... 259
Synchronous Luminescence Spectroscopy ................ 260
Low Temperature Fluorescence Spectroscopy ............ 274
Phosphorescence Spectroscopy ......................... 274
Conclusions ......................................... 275

7 CONCLUSIONS AND FUTURE WORK ........................... 276

APPENDICES

1 CHEMICALS AND THEIR SOURCES USED IN THIS RESEARCH ...... 281

2 BANDWIDTH REDUCTION AND PEAK LOCATION IN CESL .......... 283

Geometric Model for Bandwidth Reduction .............. 283
Algebraic Model for Bandwidth Reduction .............. 286
Peak Location for the General CESL Scan .............. 288
Bandwidth Reduction for the General CESL Scan ........ 290

3 SOFTWARE FOR LTCESL ................................... 292

Memory Allocation ................................... 292
BASIC Program --ENERGY .............................. 292
Assembly Routines ................................... 294

REFERENCES ..................... ........................... 298

BIOGRAPHICAL SKETCH ................................... 307













Abstract of Dissertation Presented to the Graduate
Council of the University of Florida in Partial
Fulfillment of the Requirements for the Degree
of Doctor of Philosophy

TOTAL LUMINESCENCE SPECTROSCOPY FOR SELECTIVITY ENHANCEMENT
IN MULTICOMPONENT ANALYSIS

By

Eugene Lee Inman, Jr.

May, 1982

Chairman: J. D. Winefordner
Major Department: Chemistry

Total Luminescence Spectroscopy (TLS) involves the measurement of

the luminescence emission spectrum over a range of excitation wave-

lengths, providing fingerprinting capabilities for many organic

compounds. A TLS spectrometer was built to collect data under computer

control, sequentially scanning through the wavelength regions of

interest. TLS spectra contain large quantities of information, useful

for analyte characterization. Instrumental and mathematical methods

were developed to optimize the interpretation of the information

contained in these spectra.

Synchronous Luminescence Spectroscopy (SLS) is a subset of TLS,

where a single scan is made through the TLS spectrum. By studying TLS

spectral relationships, SLS improvements are described, resulting in the

development of Constant Energy Synchronous Luminescence Spectroscopy

(CESL). By maintaining a constant energy difference between the

monochromators, Raman interference is reduced and vibrational







relationships in the spectra of polynuclear aromatic hydrocarbons are

used for analyte identification.

Low Temperature Constant Energy Synchronous Luminescence

Spectroscopy (LTCESL) supports the results of room temperature analyses.

Selectivity is enhanced by narrowing luminescence bands, making multi-

component analysis practical. By relating these results back to

luminescence theory, parameter optimization is achieved by evaluating

characteristic spectral relationships. A model was developed to explain

the resultant spectra.

Several mathematical and statistical methods are described for TLS

data evaluation. Using basic linear algebra concepts, spectra

evaluation is described for cases with extensive overlap. Cross-

correlation methods have direct application to TLS data, confirming

the conclusions drawn for CESL.

Finally, an evaluation of the applicability of these techniques to

oil samples is described briefly. Historically, these techniques have

been extensively used in oil analysis, based on empirical, rather than

theoretical, considerations. By building on luminescence theory,

several improvements are described for these techniques. These

improvements result in selectivity enhancement in multicomponent

analysis.


viii













CHAPTER 1
INTRODUCTION TO MOLECULAR LUMINESCENCE SPECTROSCOPY

Energy Model

Molecular luminescence spectroscopy concerns the measurement of

emission accompanying the deexcitation of a molecule from an excited

electronic state, where the molecule had been previously excited by the

absorption of a photon of light. The intensity, frequency distribution,

and temporal characteristics of the emitted radiation have been used

extensively for the identification and quantitation of a wide variety

of molecular species. Numerous analytical techniques and instruments

have been developed utilizing these molecular luminescence character-

istics. Effective application of luminescence techniques requires an

understanding of luminescence theory.

Many molecular spectroscopic techniques are envisioned utilizing

a molecular potential energy diagram, as shown in Figure 1. The

absorption of light by molecules at room temperature or below

generally occurs from the lowest vibrational levels of the ground

singlet state to a number of vibrational levels in one or more excited

singlet states. Whether this transition is labeled as nr-* or n-i*,

electronic absorption processes generally occur in the ultraviolet

to visible region of the spectrum. Once in an excited state, various

deexcitation mechanisms are available with characteristic rates and

therefore, probabilities for occurrence. Luminescence spectroscopy

encompasses those processes resulting in radiative deexcitation.



























Potential energy diagram of the luminescence
processes for an organic molecule.


A = absorption
F = fluorescence
P = phosphorescence
IC = internal conversion
ISC = intersystem crossing
VR = vibrational relaxation
S = singlet
T = triplet


Figure 1.


















2

1

oT2


3

2


oTI


S2








SI




A


SO







Luminescence generally accompanies a transition from the lowest

vibrational state within the excited electronic energy manifold. Once

excited, rapid radiationless deexcitation places the molecule in the

lowest vibrational state within the first excited singlet state, S1.

If a molecule then returns to the ground electronic state, SO, emitting

a photon, this radiation is referred to as fluorescence. If, however,
intersystem crossing occurs and emitted radiation accompanies a

transition originating in the lowest vibrational state of the first

excited triplet, the radiation is referred to as phosphorescence. Both

fluorescence and phosphorescence processes are included in subsequent

discussions of luminescence.

Assuming monochromatic excitation, the measured fluorescence from

a sample containing a single fluorescing compound can be mathematically
described as
Mij = Io (i)f(l-e-2.3(x i)bc)yf(xj)K(j) (1)

where M.. is the measured relative intensity at excitation wavelength

.i and emission wavelength .j, efficiency (a fraction), 1-e-2'3e(Xi)bc is the optical density with
e(xi) the molar absorptivity of the compound (L mol- m- ), b the path
length (m), and c the sample concentration (mol L-1), Io(Xi) is the

relative intensity of the incident radiation on the sample, yf(Ax) is

the fraction of the total fluorescence emitted at wavelength Xj, and

i(c() is the detection system relative response function. Several
assumptions are necessary for this relationship to be strictly valid.
First, no pre-filter, post-filter, or self-absorption effects are
present. Second, the excitation source is monochromatic or e(xi) is

constant over the bandpass of the excitation radiation. Also, exact








background correction is possible. These conditions are met for dilute

solutions of simple mixtures. Using conventional instrumentation,

fluorescence emission is not measured at a single wavelength, but is

integrated over the spectral bandpass of the emission monochromator.

This instrumental limitation is included in yf(xj). For low

concentrations of luminescing species, Equation (1) reduces to
Mij = 2.31q(X i)fE(xi)bcyf(Xj)K(Xj) (2)

An analogous equation can be written for phosphorescence,

Mij = 2.31o(X )0ISCp (Ai)bcYp(A j)i(Aj) (3)
In this case, 0ISC is the intersystem crossing quantum efficiency, p

is the phosphorescence quantum efficiency, and y p(X) is the fraction
of the total phosphorescence emitted at j. For simplicity, Equation

(2) can be written in a shortened form, with the phosphorescence

equation simplification following similarly. The terms a, xi, and yj
are defined as

a = 2.3 fbc (4)

xi = (X.)E(X.i) (5)

j = Yf(Xj)(~j) (6)
In this way, a is the product of wavelength independent terms, xi is the

product of excitation wavelength dependent terms, and yj is the product

of emission wavelength dependent terms. Equation (2) is then written as
Mij = axi (7)

for a single fluorescing compound.
Many of the spectral features observed in absorption and

luminescence spectra can be easily explained by the corresponding
processes depicted in the energy level diagram. For example,
fluorescence excitation and emission spectra are approximately mirror







images of each other. The distribution of vibrational levels within the

first excited singlet and ground singlet energy manifolds determines

the structure with each respective spectrum. Often these manifolds

are similarly spaced, resulting in mirror image spectra. This

symmetry is centered around the resonant 0-0 excitation/emission

transitions with excitation occurring at shorter wavelengths (higher

energy) and fluorescence occurring at longer wavelengths (lower energy).

This characteristic was observed as early as 1949 with respect to the

1400 K vibrational spacing observed in the fluorescence spectra of

many polynuclear aromatic hydrocarbons (1) and has been of considerable

interest in the literature (2,3). Changes in spectral features due

to temperature reduction or fluorescence quenching can also be explained

by processes within the energy diagram model.

Stokes Shift

One spectral feature often observed in fluorescence spectra

obtained at room temperature is the Stokes Shift. The Stokes Shift

corresponds to the difference between the wavelength of excitation

corresponding to a transition from the lowest vibrational state of the

ground singlet state to the lowest vibrational state of the first

excited singlet state, and the wavelength of emission corresponding to

a transition between these same two energy levels. This difference

occurs because of an energy change that occurs between the time a photon

is absorbed and the time a photon is emitted, approximately 10- s.

During this time, the solvent cage around the newly excited molecule

has time to rearrange itself to a state of lower potential energy. The

magnitude of the Stokes Shift can be calculated from








x-Um = (2/hcco)((mo-mi)2/a3 )((D-1)/(D+2)-

(n -l)/(n +1))+(2/hcEo)((ao-i)*

00 ii 00 ii
(3(muo)225(m u u) ))/a6

((D-1)/(D+2)-(n )/(n+1))2 (8)
o o
where ux is the frequency of the absorbed radiation (K), m is the

frequency of the emitted radiation (K), h is Planck's constant (Js), c

is the speed of light (ms- ), mu is the dipole moment of the solute in
O0
the ground state (Cm), m. is the dipole moment of the solute in the

excited state (Cm), a is the effective cavity radius of the solvent (m),

D is the static dielectric constant of the solvent, n is the solvent

refractive index at zero frequency, a is the polarizability of the
solute in the ground state (m3), ao is the polarizability of the solute
1
in the excited state (m 3), and E is 8.857 X 1012 CN- m2, the

permittivity of vacuum. This solvent reorientation occurs more

slowly than the 10-15 s absorption process. Therefore the energy levels

of excitation and emission do not exactly coincide. The Stokes Shift

commonly observed is on the order of a few hundred K.

Advantages of Molecular Luminescence Spectroscopy

Molecular luminescence spectroscopy has been used extensively in
a wide variety of applications and has proven to be a powerful

analytical tool. Its widespread importance is evidenced by the numerous

luminescence books describing its theory and practice (2,4-9).

Luminescence techniques are routinely used in clinical chemistry,

pharmacology, forensic chemistry, pesticide analysis, environmental

monitoring, and hydrology utilizing fluorescent tracers.

Many of the characteristics of luminescence measurement make it

an excellent analytical tool for the analysis of molecules in solution.








The major advantages are often listed as 1. high sensitivity and

detection power, 2. high selectivity, and 3. wide linear dynamic

range. High sensitivity and detection power are obtained due to the

basic measurement process relative to absorption measurement. Instead

of measuring differences between two large signals, luminescence signals

are essentially measured against zero background. In this way,

luminescence detection limits are commonly three orders of magnitude

better than absorption detection limits. High selectivity results

from two sources. First, more compounds absorb ultraviolet-visible

radiation than fluoresce or phosphoresce. The second selectivity

advantage is the dependence upon two independent wavelengths, the

excitation and emission wavelengths. The luminescence measurements

yield three spectra for identification: excitation, fluorescence, and

phosphorescence spectra. The wide linear dynamic range is due to

the low detection limits and makes luminescence techniques effective

over a concentration range up to five orders of magnitude.

The sensitivity of luminescence techniques has attracted a great

amount of research interest (10). The past decade has seen a

continuous push toward lower and lower detection limits (11,12). Recent

work has shown that using conventional fluorescence instrumentation,

detection limits in the pg/mL range are achieved for many polynuclear

aromatic hydrocarbons (PAHs) (13). These detection limits extend

below those needed for routine analysis in many areas. Scientific

speculation has been used to determine the theoretical lower limit

to experimentally attainable detection limits (14).

Equation (2) indicates the direct relationship between the measured

fluorescence intensity and the intensity of radiation incident on the









sample. With the introduction of lasers as excitation sources, incident

radiation can be increased by several orders of magnitude (15-18).

While some work has demonstrated modest detection limit improvements

with laser excitation (19-22), more powerful lasers generally do not

yield substantially lower detection limits. This is due to the limiting

noise or background in fluorescence measurements. Luminescence

detection limits are generally limited by spectral interference due to

the blank. Blank spectral interference result from two broad classes

of effects. First are those features not associated with fluorescence

and phosphorescence. Included in this group are Rayleigh and Raman

scatter. The second group includes the luminescence properties of

compounds found in the blank. Often this luminescence overlaps the

luminescence of the compound of interest. Because of this, the

implementation of laser excitation sources must take advantage of

laser temporal, spectral, and spatial characteristics rather than high

intensity alone.

Variation of the excitation wavelength in luminescence analysis

improves selectivity by utilizing the method of selective excitation.

Selective excitation optimizes the selected excitation wavelength by

matching this wavelength with a spectral region with an absorption band

of the compound of interest. The excitation wavelength is selected

to excite preferentially only one species independently of all others

that may be present. In this way, each compound is excited indepen-

dently and its fluorescence measured. For a mixture of six components,

six excitation wavelengths are selected. This technique requires one

to locate spectral regions characteristic of each analyte.

Selective excitation requires familiarity with the compounds of

interest, their luminescence characteristics, and potential spectral








interference. However, this may be difficult or even unnecessary in

cases where a large number of compounds are present, and complete

identification and quantitation of each compound are unnecessary. For

example, the analysis of polynuclear aromatic compounds in oil samples

represents this situation. Oil identification has made extensive use

of luminescence techniques to characterize these samples by their

temporal and spectral characteristics. In cases such as this, one is

generally more interested in locating spectral characteristics common

to broad categories of compounds. This is especially true for samples

where the chemical composition is unknown. For elemental analysis, it

may be possible to document all potential elemental spectral inter-

ferences. This task for molecular interference becomes formidable.

Because of this, selective excitation often becomes "hit and miss"

analysis.

Development of luminescence spectroscopy has focused on the

analysis of a single component in a pure solvent. This represents the

simplest case for an analytical technique. Attempts to apply many

luminescence techniques to mixtures, as in real samples, have

proven that many of these techniques have limited potential. Thus the

present research emphasis is on the development of techniques

capable of analyzing compounds in real sample matrices. In

applications where luminescence is commonly used, sensitivity is

often not the limiting factor, but rather, selectivity.

Theoretically, only a single measurement value is needed for

quantitation of a single component if that point is unique. But for

identification, a spectrum is required due to the vast number of

potentially luminescing compounds. Spectroscopists are accustomed








to using spectral maxima only, not fully using the vast amount of

information available. In many cases, the maximum may be a poor

choice and inadequate for multicomponent analysis. With computers

readily available, there is an increasing ability to more effectively

utilize the available data, taking full advantage of its information

content. This is the primary objective of this research.

Spectral Overlap

If the luminescence spectra of several components in a mixture

overlap extensively, identification and quantitation become difficult.

There are two broad means of resolving the spectrum for each compound.

The first group of methods are mathematical and statistical manipula-

tions. Multiple linear regression and factor analysis are important

techniques in this area. The alternate group falls under the category

of instrumental developments, including band narrowing methods and

temporal resolution. The principle underlying this group is that if

the spectral features are narrowed and the spectral range spanned by

each component is reduced, the probability of overlap is reduced and

chances are greater that there exists a region of the spectrum unique

to each component in the mixture. Examples of band narrowing

techniques are the use of low temperature and of Shpol'skii solvent

spectroscopy. Both of these broad groups of techniques will be

evaluated with respect to their utility in Total Luminescence

Spectroscopy and Synchronous Luminescence Spectroscopy.

To introduce the extent of spectral overlap routinely observed

in conventional fluorimetry at room temperature, a set of 24 poly-

nuclear aromatic hydrocarbons (PAHs) were selected as a model set.

PAHS are an analytically important class of compounds because many







compounds in this class demonstrate varied degrees of mutagenicity

and carcinogenicity and are seemingly ubiquitous. A brief perusal of

any major analytical journal demonstrates the challenge they present

to analytical chemists. In luminescence spectroscopy, PAHs are

especially important because many of these compounds demonstrate ideal

luminescence qualities. Technique development often begins with PAH

analysis before further applications to other classes of compounds.

The 24 PAHs used throughout this research are listed in Table 1.

Compounds selected are the more important PAHs from a biological

position, include a number of similar compounds and isomers, represent

a wide range of luminescence characteristics, and include enough

variety to allow one to draw conclusions based on trends found in the

data. For the spectral overlap analysis, a database was generated for

the fluorescence excitation and emission spectra for the 24 PAHs in

Table 1. The instrumentation required for this is described in Chapter

2. Using spectral bandpasses of 2 nm for both the excitation and

emission monochromators, excitation spectra were collected and digitally

stored over the range of 220 to 420 nm in 1 nm increments, and emission

spectra were collected and stored over the range of 250 to 600 nm in

1 nm increments. Because all spectra were uncorrected for instrumental

response (see Chapter 2), exact numerical significance cannot be

established, but rather, general characteristics can be studied.

Concentration selection was an important consideration for this

analysis to prevent concentration effects, such as non-linear

absorption, dimer formation, and intermolecular energy transfer, from

appearing in the spectra. PAHs, as a class of compounds, have

exceptionally large molar absorptivities (2) and large quantum









Table 1. Polynuclear Aromatic Hydrocarbons Selected as a Model Set for
Database Generation With Corresponding RTF Concentrations


Compound Concentration (pg/mL)


1. 1,2-Benzopyrene 4.0

2. Anthracene 0.10

3. 1,12-Benzoperylene 1.0

4. Pyrene 3.0

5. 1,2-Benzofluorene 1.0

6. 2,3-Benzofluorene 0.21

7. Perylene 0.022

8. 1,2:5,6-Dibenzanthracene 0.23

9. 3,4-Benzopyrene 0.11

10. Chrysene 0.84

11. Fluoranthene 0.90

12. Phenanthrene 5.0

13. 1,2:3,4-Dibenzanthracene 0.51

14. 2-Methylanthracene 0.12

15. 9-Methylanthracene 0.10

16. 1,2-Benzanthracene 0.32

17. 1-Methylpyrene 1.1

18. Naphthacene 0.12

19. 9,10-Diphenylanthracene 0.090

20. Fluorene 0.30

21. 9,10-Dimethylanthracene 0.082

22. Naphthalene 12.

23. Acenaphthene 1.0

24. Anthanthrene 0.014








efficiencies, resulting in intense fluorescence with detection limits

often in the 10-12 M range. It has been shown that fairly low concen-

trations must be selected to remain in the linear region of the

calibration curve. The concentrations selected for the 24 PAHs are

also listed in Table 1. The following criteria were used in their

selection. The typical dark current for the system used was 0.6 nA

with peak-to-peak noise about 0.1 nA. To provide sufficient signal-

to-noise, yet remaining on the linear region of the calibration curve,

a concentration was selected yielding a fluorescence maximum signal

between 20 and 100 nA. As a rough estimation, this corresponds to a

concentration about 200 to 1000 times the detection limit of the

selected compound. This proved to be quite effective. A time constant

of 0.5 s was used for all measurements, with a scan rate of 40 nm/minute.

The solvent for each solution was n-heptane.

The method for spectral overlap calculation used is similar to that

described elsewhere (23). Each spectrum was normalized so that for an

nX1 vector V and its transpose VT,

VTV= 1 (9)

The overlap was calculated by multiplying the spectrum of A by the

spectrum of B. The results have been tabulated in Tables 2 and 3.

From the excitation spectra data in Table 2, 68.1% of the pairs have

an overlap of at least 25% and 35.5% have an overlap of at least 50%.

From the emission spectra data in Table 3, 54.3% of the pairs have an

overlap of at least 25% and 36.6% have an overlap of at least 50%.

Thus spectral overlap is considerable in the fluorescence spectra

of PAHs.

The values obtained from this study cannot be directly compared

to those previously presented but are similar in magnitude (23). The










Table 2. Conventional Fluorescence Excitation Relative Spectral
Overlap

COMPOUND Aa


1 2 3 4

1.000 0.219 0.192 0.832

--- 1.000 0.248 0.267

--- 1.000 0.146

-- --- 1.000


5 6 7 8

0.747 0.762 0.043 0.770

0.297 0.290 0.130 0.173

0.222 0.191 0.924 0.265

0.539 0.611 0.047 0.592

1.000 0.953 0.061 0.745

--- 1.000 0.056 0.653

--- 1.000 0.043

-- --- 1.000


1

2

3

4

5

6

7

C 8

0 9

M 10

P 11

0 12

U 13

N 14

D 15

16

Ba 17

18

19

20

21

22

23

24


-"

"'











"'


"'

"'









-"

"'


"-




"'






"-

"'


"'




-"




"'






-"









Table 2-continued


COMPOUND Aa


9 10 11 12


1

2

3

4

5

6

7

C 8

0 9

M 10

P 11

0 12

U 13

N 14

D 15

16

Ba 17

18

19

20

21

22

23

24


0.332

0.427

0.581

0.232

0.382

0.326

0.252

0.507

1.000

---


0.724

0.250

0.204

0.567

0.872

0.897

0.061

0.678

0.378

1.000

---


0.618

0.511

0.319

0.581

0.476

0.493

0.094

0.615

0.611

0.497

1.000


13 14 15 16


0.532

0.131

0.212

0.276

0.629

0.546

0.060

0.683

0.453

0.701

0.493

1.000


0.700

0.203

0.219

0.478

0.664

0.628

0.048

0.749

0.441

0.715

0.706

0.861

1.000


0.250

0.948

0.263

0.275

0.373

0.363

0.141

0.195

0.450

0.334

0.491

0.199

0.244

1.000


0.715

0.355

0.278

0.538

0.636

0.611

0.070

0.763


0.201

0.734

0.395

0.197

0.324

0.306

0.214

0.165

0.637

0.298

0.456

0.188

0.206

0.854

1.000


0.547

0.673

0.866

0.813

0.947

0.378

0.333

1.000


"'

"'

"'

"'

"'



"'


"'

"'

'"'

"'

"'





"'


"'

"'

"'

"'

"'



"'

"'


"'

"'

"'

"'

"'





"'







Table 2-continued


COMPOUND Aa


17 18 19 20


21 22 23 24


0.627

0.340

0.163

0.727

0.406

0.497

0.047

0.500


1

2

3

4

5

6

7

C 8

0 9

M 10

P 11

0 12

U 13

N 14

D 15

16

Ba 17

18

19

20

21

22

23

24


0.279

0.036

0.286

0.184

0.311

0.320

0.229

0.309


0.264

0.504

0.298

0.650

0.590

0.077

0.081

0.521


0.170

0.565

0.485

0.185

0.302

0.306

0.314

0.148


0.665

0.321

0.441

0.233

0.218

0.607

0.729

0.321


0.432

0.225

0.209

0.231

0.709

0.653

0.055

0.588


0.390

0.775

0.348

0.797

0.687

0.324

0.300

0.632


1.000 0.197 0.224 0.199

--- 1.000 0.182 0.614

--- 1.000 0.332

-- --- 1.000


0.144

0.522

0.405

0.149

0.378

0.363

0.288

0.105


0.528

0.341

0.342

0.250

0.196

0.591

0.631

0.271


0.505

0.161

0.189

0.233

0.623

0.558

0.049

0.600


0.393

0.703

0.510

0.928

0.924

0.230

0.209

0.839


0.619

0.071

0.226

0.325

0.755

0.640

0.032

0.842


0.434

0.727

0.472

0.848

0.832

0.114

0.093

0.762


0.267

0.133

0.763

0.199

0.360

0.312

0.757

0.315


0.314

0.305

0.189

0.181

0.200

0.149

0.194

0.198


0.174 0.217 0.233 0.136

0.146 0.671 0.478 0.180

0.878 0.235 0.100 0.298

0.394 0.817 0.740 0.192

1.000 0.255 0.096 0.297

--- 1.000 0.829 0.167

-- 1.000 0.305

-- --- 1.000


aCompound


numbers match those in Table 1.


0.294

0.420

0.705

0.266

0.445

0.340

0.270

0.604












Table 3. Conventional Fluorescence Emission Relative Spectral

Overlap


COMPOUND Aa


1 2 3 4


5 6 7 8


1.000 0.789 0.489 0.915


--- 1.000 0.486 0.861


--- 1.000 0.422


--- --- 1.000


0.332 0.116 0.167 0.732


0.352 0.148 0.150 0.735


0.081 0.014 0.787 0.673


0.442 0.183 0.131 0.751


1.000 0.836 0.020 0.205


--- 1.000 0.003 0.068


-- 1.000 0.282


-- --- 1.000


I.


1


2


3


4


5


6


7


C 8


0 9


M 10


P 11


0 12


U 13


N 14


D 15


16


Ba 17


18


19


20


21


22


23


24


___ ___


___ ___


___


___


___ ___


___


___ ___


___ ___


___


___


___


___ ___


___ ___


___


___


___


___ ___ ___


___ ___ ___


___ ___ ___ ___


___ ___ ___


___ ___


___ ___ ___ ___


___ ___ ___ ___


___ ___ ___


"'


"' "'


'"' "'


"' "'


"-


"' "'


"' "'


"'


"' '"'


"' "'


"' "'


"' "'


"--


"'


"'


"'


"' "'






"'


"' "'


"" "'


"'


"" "'


"'


--- --- --- ---


--- --- --- ---




--- --- --- ---






--- --- --- ---


--- --- --- ---
- --- --- --


--- -- --


---" -- -








Table 3-continued


COMPOUND Aa


9 10 11 12


0.634

0.638

0.611

0.562

0.139

0.036

0.298

0.636


0.689

0.707

0.248

0.826

0.692

0.378

0.068

0.473


0.253

0.246

0.781

0.208

0.037

0.007

0.832

0.378


0.451

0.444

0.122

0.567

0.938

0.639

0.034

0.273


13 14 15 16


0.910

0.905

0.422

0.960

0.431

0.180

0.132

0.714


0.895

0.856

0.596

0.889

0.323

0.120

0.270

0.739


0.914

0.637

0.670

0.796

0.236

0.067

0.352

0.788


0.921

0.714

0.560

0.882

0.291

0.093

0.248

0.741


1.000 0.443 0.411

--- 1.000 0.122

--- 1.000


0.201

0.789

0.057

1.000


0.565

0.785

0.206

0.556

1.000

m---


0.821

0.739

0.375

0.430

0.871

1.000


0.677

0.546

0.396

0.340

0.780

0.853

1.000


0.713

0.664

0.322

0.412

0.849

0.929

0.933

1.000


M 10

P 11

0 12

U 13

N 14

D 15

16

Ba 17

18

19

20

21

22

23

24


___



___

___

___

___

___


"'

"'

"'



"'



"'

"'


"'

"'

"'



"'

"'

"'


"'

"'

"'

"'

"'

"'








Table 3-continued


COMPOUND Aa


17 18 19 20


21 22 23 24


0.800

0.924

0.351

0.886

0.405

0.177

0.089

0.688


1

2

3

4

5

6

7

C 8

0 9

M 10

P 11

0 12

U 13

N 14

0 15

16

Ba 17

18

19

20

21

22

23

24


1.


0.021

0.014

0.275

0.029

0.007

0.005

0.443

0.073


0.034

0.006

0.625

0.009

0.015

0.060

0.052

0.043


0.677

0.691

0.872

0.586

0.125

0.026

0.514

0.786


0.867

0.391

0.598

0.183

0.596

0.814

0.791

0.739


0.000

0.000

0.013

0.002

0.025

0.039

0.009

0.013


0.000

0.001

0.014

0.073

0.001

0.012

0.001

0.005


000 0.009 0.510 0.000

--- 1.000 0.101 0.034

--- 1.000 0.000

-- --- 1.000


0.673

0.758

0.802

0.595

0.139

0.036

0.431

0.754


0.856

0.411

0.560

0.195

0.611

0.812

0.704

0.697


0.

0.

0.

0.

1.


0.044

0.051

0.042

0.062

0.354

0.400

0.031

0.050


0.027

0.121

0.036

0.309

0.060

0.054

0.035

0.045


541 0.

100 0.

968 0.

007 0.

000 0.

--- 1.


0.049

0.060

0.016

0.076

0.436

0.497

0.004

0.047


0.015

0.151

0.005

0.364

0.077

0.052

0.031

0.045


0.231

0.232

0.619

0.160

0.023

0.000

0.571

0.296


0.511

0.112

0.586

0.042

0.182

0.379

0.363

0.310


054 0.071 0.121

029 0.009 0.077

032 0.015 0.607

402 0.309 0.000

036 0.021 0.567

000 0.926 0.024

--- 1.000 0.001

--- 1.000


aCompound


numbers match those in Table 1.


0.459

0.720

0.155

0.514

0.952

0.768

0.645

0.706


.








selection of monochromator bandpasses of 2 nm is significant for this

and later studies. A major objective in selectivity enhancement is

band narrowing of the fluorescence spectra. In attempting to achieve

this objective in fluorescence spectroscopy, the monochromator band-

passes must not introduce significant band broadening beyond the

narrowest bandwidths inherent in the fluorescence spectra. Because

PAH fluorescence bands are as narrow as 4 to 10 nm at room temperature,

a bandpass less than this should be selected. Therefore, monochromator

bandpasses were selected to reduce instrumental broadening effects on

the luminescence spectra. This significance will be important through-

out this work.

A major advantage of luminescence techniques is the capability of

varying two wavelengths during an analysis, as mentioned. The full

potential of this advantage is usually not realized in conventional

luminescence spectroscopy. Conventionally, a spectrum is obtained by

maintaining one monochromator (excitation or emission) at a constant

wavelength and scanning the other monochromator (emission or excitation)

over a range of wavelengths. Two techniques that take advantage of this

selectivity capability are Total Luminescence Spectroscopy and

Synchronous Luminescence Spectroscopy. Exploitation of this capability

to enhance selectivity has been the underlying determination in this

research.

Chapter 1 has described the basic luminescence concepts of

considerable importance in Total Luminescence Spectroscopy and

Synchronous Luminescence Spectroscopy. Total Luminescence Spectroscopy

is described in detail in Chapter 2, while the concepts of Synchronous

Luminescence Spectroscopy are set forth in Chapter 3. Chapter 4 extends








the concepts of Synchronous Luminescence Spectroscopy to low temperature

analysis. Chapter 5 includes several mathematical and statistical

methods available for data evaluation from the techniques described

in Chapters 2-4. The objective of the techniques in Chapters 2-5 is to

enhance selectivity in multicomponent analysis using the concepts

established in Chapter 1. Chapter 6 summarizes the application

development of these techniques, with Chapter 7 suggesting areas for

further investigation.

Note

1Throughout this work, energy units are expressed as kaisers,
defined as cm-l, and abbreviated K. The kaiser is the accepted primary
unit while the wavenumber is common throughout the literature and is
the accepted secondary unit. The symbol K is also used throughout
this work for degrees kelvin. While both are used extensively in
Chapter 4, the desired units are clear from the context.













CHAPTER 2
TOTAL LUMINESCENCE SPECTROSCOPY

Theory

Total Luminescence Spectroscopy (TLS) measures the emission

spectrum of a sample over a range of excitation wavelengths. Data

obtained in this manner represent variation in both the excitation

and emission wavelengths, making more spectral information available

for sample characterization. While an abundance of information is

potentially present within the data, full understanding of data

relationships over analyzed spectral regions is necessary to exploit

the data to their maximal analytical potential. As the analytical

chemist approaches this goal, improved selectivity in multicomponent

analysis should be the natural result.

As will become obvious, TLS data encompass a broad set of closely

related luminescence techniques. Therefore a thorough investigation

was completed in attempting to understand TLS from an instrumental

and mathematical perspective. Thus TLS bridges the gap between

fundamental luminescence principles as described in Chapter 1, and

the more specific techniques described in later chapters.

The mathematical description of TLS data is a natural extension of

Equation (7). Now instead of measuring a single data point, groups of

measurements are made. Equation (7) then becomes

M = aX Y (9)

where M is an nXm matrix defining the TLS data plane, X is an nXl matrix

(array of length n) representing a conventional fluorescence excitation








spectrum, and Y is a 1Xm matrix representing a conventional

fluorescence emission spectrum. Thus the measured quantity Mi. is a

single point within the matrix M. Each point within M represents a

unique pair of experimental wavelengths. Selection of an i,j pair

defines an excitation, emission wavelength pair and a corresponding

fluorescence measurement.

For a single fluorescent component, the measured emission specter


is independent of the excitation wavelength, and the

spectrum is independent of the emission wavelength.

analysis, this is no longer true. For a solution of

compounds, Equation (9) becomes


measured excitation

In multicomponent

r fluorescing


r
M= E akXk (10)
k=l
Analytical applications of TLS directly depend on this relationship.

There are several advantages of TLS over conventional luminescence

spectroscopy. Selectivity is improved by varying both the excitation

and emission wavelengths, as described earlier. TLS requires minimal

a priori knowledge of an unknown's luminescence characteristics. TLS

has excellent potential resolvability of complex mixtures with little or

no sample pretreatment. Also, TLS spectra are unique for each

luminescing compound, offering fingerprinting capabilities, ideal for

analyte identification (24).

The last advantage of TLS spectra suggests that because there are

unique spectral characteristics for each compound, one need only identify

those spectral characteristics that are specific for the analyte of

interest and recognize these characteristics within the spectrum of an

unknown solution. Because the TLS spectrum extends over such a large

range of wavelengths, there is an excellent probability that if the


um








compound fluoresces, its characteristics are defined by the TLS spectrum.

Thus, one need not examine several data sets, corresponding to various

experimental parameters, but rather, a single data set, capable of

identifying numerous compounds with minimal experimental variation of

parameters.

TLS,as an analytical tool,was established in the middle 1970's with

several instrumental developments, making data collection practical (25).

For TLS to be a routinely used technique, data collection must be rapid

and automated. To speed the data collection process, one must either

scan more rapidly or make many measurements simultaneously. While the

matrix format of TLS data was pointed out earlier (26), experimental

realization had not taken a foothold until the late 1970's. This can be

traced to two technological advances necessary for TLS instrumentation:

the computer and rapid data collection devices.

Instrumentation

Instrumentation for TLS data collection has been described by

several authors by interfacing the monochromator scan drives of a

conventional fluorimeter to a computer (27). What would be manually

difficult and time-consuming then becomes practical for routine analysis.

The monochromators are sequentially scanned over the desired excitation

and emission ranges. This requires excellent reproducibility in the

stepping motors of the monochromators. The scan speed is determined by

the mechanical limitations of the monochromators and the signal-to-noise

ratio (S/N) required for the analysis. Mechanical improvements have been

made with high-speed stepping motors (28,29) and galvanometrically

scanned mirrors (30).








An alternate data collection system for TLS utilizes a multichannel

detection system. The Videofluorometer is based on this system. The

Videofluorometer was presented in 1976 (31) as a result of TLS research

at the University of Washington. This instrument and its applications

have been described in detail (32-35). The unique features of this

instrument are the polychromatic irradiation of the sample cuvette and

an SIT vidicon detector. Using a xenon arc lamp excitation source, the

wavelength of excitation is a function of the physical height within

the cuvette. The multichannel detector allows an entire TLS spectrum

to be collected in less than 20 ms, although integration times of

approximately 1 s are commonly used. As improvements are made in the

sensitivity of SIT vidicon detectors, the TLS data collection efficiency

using instrumentation similar to the Videofluorometer will increase

(36-38).

Talmi has reviewed the status of multichannel image detectors and

their utility in spectroscopy (39-41). A major direction of instru-

mental development has been in the application of these systems to

HPLC, most commonly for spectrophotometric measurements. For this

use, an entire spectrum can be obtained in approximately 1 s. With this

speed, image detectors make fluorescence detectors practical for

HPLC analysis (42).

The Videofluorometer suffers from several major problems. For

adequate sensitivity, spectral resolution is generally around 10 nm,

with additional degradation of resolution by pixel summing. With

present-day image detectors, this system is less sensitive to the

emission region below 360 nm. Also, its cost, on the order of $50,000,

is prohibitive for routine analysis. The unique sample illumination








design makes adaptation to other luminescence techniques other than

room temperature fluorescence difficult. Thus instrumental improvements

are needed to overcome many of these limitations.

Baird (Baird Corp., Bedford, MA 01730) has designed a TLS system

that is presently on the market. It uses photomultiplier detection

with computer-controlled monochromator scanning. A PDP 11/04 computer

is used for instrument control, data manipulation, and contour plotting.

Information on this system was obtained after most of the instrumental

development for this research was completed although many similarities

can be found.

For further study, a database of TLS spectra was necessary. While

the research efforts of others have emphasized instrumentation, this

research is more concerned with the spectra obtained and spectral

relationships. Because of this, adequate instrumentation was developed

with the main effort made towards software development, establishment

of a database, and evaluation of data. The following criteria were

established to direct instrument development (43):

-adequate resolution (varies with process measured)

-broad spectral range (220-650 nm)

-on-line computer for

-data acquisition

-data display

-data analysis

-plotting

-complete instrument control

-spectra library organization








Implementation of these requirements suggests three areas of instrument

development: the fluorimeter, computer hardware, and computer software.

Fluorimeter

Previous research has proven that room temperature fluorescence

spectroscopy (RTF) is the best starting point due to its relative

simplicity in instrumentation and data interpretation. A fluorimeter

was constructed as shown in Figure 2, with individual components and

their sources listed in Table 4. Each component was selected to

provide sufficient versatility to expand as research needs dictated.

The components of the fluorimeter correspond to what is commonly

found in commercial instruments. The secondary goal of this development

was to keep instrumental costs low for practical reasons.

The xenon arc lamp is the most commonly used source of excitation

in luminescence spectroscopy. It provides sufficient intensity over a

broad spectral region, and its output is more continuous than the

sharp-spiked output characterisitc of medium-pressure mercury lamps.

Various optical arrangements and reflector descriptions have been

described to improve upon the standard arrangement of xenon arc lamp use.

The monochromators were selected because of their versatility,

compactness, and ease of automation. Both monochromators and scan

controls are identical except for the grating blaze wavelength. These

monochromators are f/3.5 and have a reciprocal linear dispersion of

8 nm/mm. With 2 nm bandpasses, sufficient throughput is achieved for

the RTF studies. The excitation monochromator has a blaze wavelength

of 250 nm and the emission monochromator has a blaze wavelength of

450 nm. These are optimal for PAH analysis. Wavelength calibration of

these monochromators using a low-pressure mercury lamp (Oriel Corp.,








































4-
4-)

E













4->
0





ca
(U








a)
C .


(0-



.,
<(










*-J
0
I--


0
E




*10



0




CM


ra




30












uZ

U,0
o


mu
I-


0





















S0 9
4u S








ZZ
mu








CL








u, 0
IL_




5 o





a.








Table 4. Experimental Equipment Used in Total Luminescence Spectroscopy


Item Model Source


Eimac Xenon Arc Lamp
150 W

Eimac Illuminator Power Supply
operated at 12 A

Excitation Monochromator
f/3.5, holographic grating
1200 grooves/nm

Sample Housing



Emission Monochromator
f/3.5 holographic grating
1200 grooves/nm

Photomultiplier Tube


High Voltage Power Supply
operated at -1000 V


VL-150-2


P250S-2


H-O1UV


H-1OV



1P928


EU-42A


Nanoammeter

DC Amplifier


Monochromator Scan Controls


Minicomputer


1020-SS


PDP 11/34


Eimac, Division of Varian
San Carlos, CA 94070

Eimac, Division of Varian
San Carlos, CA 94070

American ISA, Inc.
Metuchen, NJ 08840


American Instrument Co.
Silver Spring, MD
20910

American ISA, Inc.
Metuchen, NJ 08840


Hammamatsu
Waltham, MA 02154

Heath Co.
Benton Harbor, MI
49022

Laboratory Constructeda

Laboratory Constructed

American ISA, Inc.
Metuchen, NJ 08840

Digital Equipment Corp.
Maynard, MA 01754


Analog-to-Digital Converter


Laboratory
Peripheral
System (LPS-11)


Maynard, MA 01754


Digital Plotter


HIPLOT DMP-4 Houston Instrument
Division of Bausch &


Lomb
Austin, TX


78753


aReference 44









Stamford, CT 06902) demonstrated accuracy of +0.2 nm over the

entire spectral region.

The 1P928 photomultiplier tube was specifically selected for its

wide spectral response range, extending well into the near infrared

region. Other work with this PMT has shown excellent sensitivity out

to 1050 nm. This is especially important for PAH analysis as the

phosphorescence of many PAHs extends from 600 to 800 nm.

In all these studies, the spectra are uncorrected for instrumental

response and excitation source output. Ideally, all spectra should be

corrected to provide information that is specific-instrument independent.

While this instrument is capable of correcting spectra digitally, this

feature was not used in this work. Even without corrected spectra, one

should be aware of the spectral limitations of the instrument for proper

data interpretation. As a result, the instrument response functions

were obtained for the excitation and emission arms of the fluorimeter.

Several procedures have been described for the correction of

luminescence spectra (45). To obtain the spectral response functions

for this instrument, a 1000 W quartz-halogen tungsten coiled-coil

filament lamp was used for reference (Optronic Laboratories, Inc.,

Silver Spring, MD 20910). Although absolute irradiance data are

available for this lamp, only relative measurements were made. This

lamp is recommended by the National Bureau of Standards as a spectral

irradiance calibration standard (46). The lamp was operated at 8.00

+ 0.008 A dc using a Model 83 Precision DC Current Source (Optronic

Laboratories, Inc., Silver Spring, MD 20910).

The procedure used for calibration required two sets of measurements.

The sample cell was modified to allow direct illumination by the lamp








through the sample cell into the measurement arm of the fluorimeter. A

bandpass of 2 nm was used on the emission monochromator. The emission

spectral response function was obtained by dividing the measured spectrum

by the published spectral irradiance of the lamp. The lamp was then

removed and a quartz cuvette wall was placed in the sample cell to

scatter the xenon arc lamp output directly into the measurement system.

A bandpass of 2 nm was also used on the excitation monochromator. The

total instrument response function was obtained by scanning the emission

monochromator to the selected wavelength, scanning the excitation

monochromator through the same spectral region, and recording the

maximum signal obtained. This compensated for imperfect placement of the

quartz plate in the sample cell. The excitation spectral response

function was then obtained by dividing the total response function by the

emission response function. Each of these is displayed in Figure 3. The

general shapes of these functions agree with what has been reported (47).

Each function defines the relative instrumental response per nm, measured

at each wavelength interval. The xenon arc lamp suffers from low

relative intensity below 300 nm, with a peak intensity around 800 nm.

A decision was made to leave spectra uncorrected to eliminate the

artificial distortion imposed by digital correction. Although the

emission response functions shows several minor spectral features,

correction for emission response causes few changes in the spectra

obtained from this system. The excitation output of this system,

however, involves substantial change over the commonly used wavelength

range. Thus when correcting spectra, the region below 300 nm is

multiplied by fairly large factors. With noise present, the S/N becomes

grossly distorted with noise peaks almost as large as strong signals





























Figure 3. Relative spectral response functions for TLS
instrument.

A. Excitation

B. Emission












, X10

0
a_
ul -



CI
LU





w




250 300 350 400 450 500

EXCITRTION WAVELENGTH (NM)


B


f-I





I--
.J





250 350 450 550 650 750


EMISSION WAVELENGTH


(NM)







towards the red region of the spectrum. An excellent example of this

distortion is given in Reference 48.

Because of the common use of xenon arc lamps in luminescence

spectroscopy, most instruments have similar response functions. While

corrected spectra are important in absolute measurements and

quantitative data, the conclusions drawn and principles set forth in

this work are in no way altered by the lack of instrumental independence.

Of more concern in the present research is the temporal output

fluctuation of the excitation source. The radiant flux from a xenon arc

lamp is generally noisy, drifts over time, and is interrupted by

occasional spikes. To stabilize the lamp output, a closed loop feedback

system has been described in the literature (49).

Deoxygenation of solutions has been suggested as a method to

increase fluorescence sensitivity. By bubbling nitrogen through the

solution, oxygen quenching is decreased with a resultant increase in

fluorescence signal. An apparatus was designed to provide deoxygenation

of solutions for the TLS system. A teflon cap is fitted over the

cuvette with a stainless steel needle extending to the bottom of the

cuvette along the back corner. House nitrogen is bubbled through the

solution for 5 minutes and then blown over the solution during the

analysis using a three-way stopcock to direct the nitrogen flow to

another stainless steel needle extending to just above the surface of

the solution. To prevent solvent evaporation, the nitrogen is saturated

with solvent. While some researchers suggest that deoxygenation is

always necessary (50), it proved to be of limited utility in routine

analysis. Although analyte fluorescence signals increase by a modest

factor of two or three, oxygen-free nitrogen is not practical for








routine use. Also spectra are generally unchanged by deoxygenation,

making it unnecessary for the study at hand.

Computer Hardware

The PDP 11/34 with 32 K words RAM was selected for instrumental

control and proved to be more than sufficient for all computing

requirements. A parallel digital interface to the PDP 11/34 was

connected to the monochromator scan controls requiring three lines for

each scan control: ground, direction selection, and scan speed control.

Data collection is through the analog-to-digital converter of the LPS-11.

With this available, little hardware development was necessary.

Computer Software

The RSX-11M operating system was used for all software development

and corresponding instrumental control throughout this work. Software

development comprised a major effort in successful completion of this

research. Rather than a collection of independent programs, a rather

extensive network of files, programs, and directories was necessary to

meet all the requirements placed on the computer for the TLS system. The

chart in Figure 4 illustrates the overall philosophy in the software

development. Two analogous sets of programs are available for Total

Luminescence Spectroscopy and Synchronous Luminescence Spectroscopy

experimentation, with several peripheral programs. Each of the six main

programs shares a set of about 25 library subroutines, performing a wide

variety of operations. The network of programs and subroutines and

their functions are listed in Tables 5 and 6.

Data are organized into four types of data files, identified by the

three letter extension after the file name. DAT identifies a TLS

spectrum of raw data as collected, BKD files have been background

















































4-


0)






C)


S-
C
4-
0



0,





01
I-

C
C




S.-
0


1-








LL




















/ I\ I I

I-----
3 I






I I I I
I I
*0 0u 0
-
mu








m m
Ml








Table 5. FORTRAN Programs Developed for TLS System


Program Name Function



BANDPS to convert a spectrum with bandpass 1 to spectrum
with bandpass 2

ENERGY to convert a spectrum as a function of wavelength
to one that is a function of energy

MAINFL to maintain TLS data files and PARAMS.DAT

MIXTUR to simulate TLS mixtures

NRNBR to perform a nearest neighbor smooth on a TLS
data file

OPT to calculate synchronous optimization parameters

OVRLAP to calculate extensive synchronous optimization
parameters

PEAKS to determine wavelength position of spectral
features in vectors files

SCATT to remove Rayleigh scatter from vector files in
LTF

START to initialize PARAMS.DAT

UNIQUE to calculate spectral overlap between luminescence
spectra

VECFRM to create vector files directly

VECPLT to plot vector files








Table 6. Library Subroutines Developed for TLS System


Subroutine Name Function


AXES2D to draw axes in two-dimensional plots

AXES3D to draw axes in three-dimensional plots

AXLB2D to label axes in two-dimensional plots

AXLB3D to label axes in three-dimensional plots

DISPLA to display TLS data file

FIND to locate file parameters in PARAMS.DAT

FLIST to list file parameters in PARAMS.DAT

INIT to initialize the digital plotter

LOCMAX to locate the local maximum in a TLS data file

MAXFND to locate the maximum in a TLS data file

MONO to scan monochromators to their desired locations

NEWFIL to initialize parameters for a new data file

OLDFIL to locate file parameters for an existing data
file

SCAN to scan monochromators

SMOOTH to perform a nearest neighbor smooth on a TLS
data file

SMTH to smooth a TLS data file using basis vectors

SPECTR to plot an array of data

STAND to generate a standard synchronous spectrum

SUBBLK to blank subtract a TLS data file

SYNLUM to control synchronous instrumentation

SYNMAT to perform mathematical operations on synchronous
data files

TOTLUM to control TLS instrumentation








Table 6-continued


Subroutine Name


TOTMAT


TOTPLT

VECTOR


Function


- to perform mathematical operations on TLS data
file

- to plot TLS data files

- to create a vector file from a TLS data file








subtracted with the appropriate TLS blank spectrum, while SMO files have

been digitally smoothed. The VEC file contains two records; the first

is the luminescence excitation spectrum and the second is the

luminescence emission spectrum. VEC files are obtained by one of two

possible routes. Both spectra can be obtained by directly scanning the

desired spectrum with the other monochromator at the peak maximum, or

the spectra can be obtained by selecting the most intense column and row

of the TLS data matrix after proper background subtraction. Because of

the mathematical relationship described by Equation (9), a TLS data file

can be generated from the VEC file. Therefore, in the generation of a

database, the data are more compactly stored using a VEC file with a

resulting storage savings of a factor of 30 to 50 for each file. For

TLS reference spectra of single component solutions, the data file is

generated from a VEC file as needed.

Access to any data file is through the file PARAMS.DAT. This file

contains a directory defining the parameters under which each data file

was collected, along with identifying information. Each data file is

named as X##LLL.EXT, where ## corresponds to an identification number

for each compound (01-24 were used for the compounds in Table 1). LLL

is a group of three letters identifying the luminescence method used to

collect the data, such as, RTF for room temperature fluorescence and LTP

for low temperature phosphorescence, and EXT corresponds to one of the

four file extensions. Thus,each compound-method pair has four extension

possibilities and four corresponding data files. PARAMS.DAT records

which files have been created and are available for evaluation and which

files have been deleted.

TOTLUM and SYNLUM are used for data collection and instrumental

control. The TLS data are collected using a rapid scan procedure.








From its initial wavelength settings, the emission monochromator is

scanned forward to the terminal wavelength, the excitation wavelength is

advanced, and the emission monochromator is scanned in reverse. Each

emission scan is stored as a separate data record upon completion of the

scan. Scan rates, wavelength intervals and ranges, and amplifier gain

are variable and selected by the user. All software was written in

FORTRAN with A/D control and signals sent via the parallel interface for

monochromator scan control using a set of FORTRAN-callable assembly

language routines obtained from DEC. With this software, a 1700 points/s

data collection rate is achievable.

Due to the vast amount of data that is collected for a TLS

experiment, effective visualization becomes a challenge. If one cannot

visualize the data and understand their relationships, full utility of

their informational content is generally not realized. Attempts were

made to locate a plotting package at a moderate cost that was compatible

with the PDP 11/34 system. After several unsuccessful attempts, a

plotting package was written for the specific needs of this work. Three

plotting modes were needed: two-dimensional plots, three-dimensional

isometric projections, and contour plots. This work is certainly not

unique and comprehensive, but it fulfills the current needs and will be

described briefly.

Several textbooks have described considerations in plotting

software development (51,52). The steps involved in the plotting

routines for the TLS system include initializing the plotter,

establishing a point of reference, defining a data window, drawing axes,

labeling axes, plotting the spectrum, and terminating plotting

operations. All plots are produced on a digital x-y recorder. Two-








dimensional plotting is fairly routine with the spectrum scaled to its

maximum or to any selected size. The three-dimensional plotting routines

required substantially more effort to develop.

Isometric projection plots depend on hidden line calculations, rapid

enough to be practical. Because TLS spectra include such a large number

of data points, the calculations must be simple, not having large

memory requirements. Isometric projection plots display the emission

wavelength along the x axis, the excitation wavelength along the z axis,

and the relative intensity along the y axis. The x,y,z position is first

determined for each point and its projection onto the x-y plane defined

by the paper is calculated. If this projection is below a previously

plotted surface, the line from the previous point to the current point

is hidden.

Implementation of this hidden line concept is straight forward. The

data file is plotted horizontally an emission row at a time at whatever

selected excitation resolution. For any scanned row, only two records

are needed in memory at one time, the current data row and the

previously plotted row. For each new row, an array is created from the

data record, linearly interpolated to fit the plot size. This new array

fits each point to a single position on the plotter, providing maximum

resolution. The projection of this array onto the x-y surface is then

calculated and plotted. As the array is plotted, each point is compared

to the corresponding point of the previously plotted array. A pen up or

pendown command is generated as needed for hidden lines. When completed,

the plotted array is then stored as the previously plotted array. This

cycle is then repeated for the entire plot. A typical plot requires

about 10 to 15 minutes to complete. The viewpoint for the plot is








determined by the definition of the z axis location. That is, an x,y

point is selected by the user as the terminal point for the z axis.

This point then defines the angles of projection for an x,y,z point onto

the x-y plane. In this way, one has 90 of freedom in viewing the TLS

data.

Contour plotting is generally implemented using triangularization

(53). For this work, a simple contour concept was used, similar to that

described for isometric projection plotting. Again, only the previously

plotted record and the current record are kept in memory. A data array

is generated that is a linear interpolation between the two data records

to the desired resolution. This current array is plotted with a pen up

or pen down command sent to the plotter each time a contour boundary is

crossed. When completed, contour regions are represented by alternating

regions of black and white.

To illustrate the spectra obtained using these three plotting

routines, a dilute solution of quinine sulfate in 0.01 N sulfuric acid

was analyzed. Figure 5 shows the conventional fluorescence excitation

and emission spectra for this solution. Figure 6 is the isometric

projection of the TLS data file, while Figure 7 is the contour plot for

the same solution. Both the isometric projection and contour plots

have advantages and disadvantages, so each is used as necessary. For

complete visualization, both plots are needed.

Spectral Features

Once TLS data are obtained, interpretation of these data is

necessary to identify the compound of interest. To do so, the unique

spectral features for that compound must be determined. Several methods

have been tried to highlight these features. One method that has been


























0


0.
S-1


0





0-
4-)
























(o
-r--












4-'
C-
Ea















-r-



*r-




0.
o-
0










4- C


C:
a)
-I'















> c--
r_
v3




0r-





Oa














U-
C*r-
>
> C3




























D 5


ID





0*
-o
La.






O




O
0)


L1ISN31NI 3AI1U713














































0i





C,
*1c







C-,
a)

0I






a)
Ec
0






I-,


U)
S.













15Z



1nK 8






















































C,

*w
4-



U,
-o





0
r-
I-















f..


r-
0-
r--
D-










oI-,











E)





L)
)





UU
--
Q --I






w
w








(0)



ID







U in UI U
tr)N


(WN) Hi9N313AUM NOIJ.UI3X3








successful in simple mixture analysis and in oil sample screening is the

generation of difference spectra (48). The TLS spectrum of a spectrally

interfering compound can be subtracted from the TLS spectrum of the

analyte to reveal characteristic spectral regions. To illustrate,

Figure 8A is the TLS spectrum of 1-methylpyrene and Figure 8B is the TLS

spectrum of pyrene. When the difference spectrum is plotted, shown in

Figure 9, unique regions are readily identifiable for 1-methylpyrene.

The excitation region between 340 and 350 nm appears to be the region of

choice in this situation.

Another suggestion to enhance key spectral features is the use of

log plots. That is, log relative intensity is plotted on the y axis.

This tends to enhance regions of low intensity and suppress regions of

high intensity. While perhaps useful in limited cases, multicomponent

analysis still remains difficult. Both of these methods are empirical

in nature and do not rely on theoretical considerations.

To fully exploit TLS data, spectral features must be interpreted

relative to luminescence theory. The TLS spectrum of a ternary mixture

of PAHs, anthracene, perylene, and naphthacene, reveals that the spectra

of PAH mixtures become complicated rapidly, as in Figure 10. Yet this

spectrum suggests an inherent symmetry, perhaps useful in further

analysis.

When solutions of low fluorescence intensity are analyzed, their TLS

spectra contain several readily identifiable spectral features. Three

such features are Rayleigh (elastic) scatter, Raman scatter, and second

order Rayleigh scatter, as shown in Figure 11. Rayleigh scatter, the

redirection of the incident photons by atoms and molecules within the

solution, appears along a 45* line through the plot where the excitation




























Figure 8. TLS contour plot of:

A. 1-methylpyrene

B. pyrene













A



355





I-
300


c-








350 400 450 50

EMISSION WAVELENGTH (NM)





B


350 -







z
0 3





X
S 250




350 400 450 501


EMISSION WAVELENGTH (NM)















Ln










D aD
Lr)






z
*-j

0
Un

LJ


(WN) H19N3713AUM NOIIUII3X3


_ I ~


S4b









































C











,..

r..
5-

Cf



4-







E
(J





Q..
0)
C,
S-
X












4-I





L) C
r-

r-c:













CD,
*f-










'-4W
Z U-Z
0 0 X-

M a L.u

I, x Z
"W-- ..J 2
1^ U >n *
ll X I"


hC


z



U1
..J

to
L)
Z
0
S-4
in
LnL

































(A

0




4-)






tri
C








Lc,
*r-








4-
-Nd




C *


S -
r-




4-
.0-




L
U- O
s-



E0
0 U

4--





ur-





I- C
-4-










5-
U-



I-
Icn

LL.




61















CD
qqr


g UJ
x
w u


c LtD

-.------ ouJMP-
....--- =n a
CD -ic
u .
w r

E))


0
Ln











cr.


to



cr ED

in CD~
z -4
cr.i
cr ID
w 0 U








wavelength and the emission wavelength are equivalent. Raman scatter

also appears along a curved path through the TLS spectrum-that is

characteristic of the solvent used. This is generally of low intensity.

Second order Rayleigh scatter occurs where the emission wavelength is

twice the excitation wavelength. This appears because of the dispersive

(interference) characteristics of the grating within the monochromator.

The upper triangle in the TLS data matrix corresponding to the

region where the excitation wavelength is greater than the emission

wavelength is of little analytical utility. Stokes Law states that the

emission wavelength must be greater than or equal to the excitation

wavelength. While luminescence has been observed in the region where

.i is greater than .j, it is generally of low intensity and is not

useful in trace analysis. Two instrumental modifications were made to

exploit this feature. In obtaining TLS spectra, this region is scanned

around to save time, and the matrix is zero-filled in this region. The

time savings varies but can extend up to 50%. Secondly, this region is

placed in the rear of the TLS isometric projection plot, providing as

broad a view as possible of spectral regions with analyte luminescence.

Conclusions

Many of the assumptions made for TLS analysis must correspond to

those that are made in conventional luminescence spectroscopy. These

assumptions will be basic to all luminescence theory and discussions

described for these studies. They include: 1. The excitation and

emission processes are completely independent. That is, the shape of

the emission spectrum is independent of the wavelength selected to

excite the sample. 2. No appreciable electronic energy transfer

occurs between species. 3. Absorbance is low and the optical density








is less than 0.01. Other luminescence principles of conventional

luminescence apply in the areas of filter and quenching effects. These

assumptions are usually met when working with dilute solutions. For

PAHs, concentrations less than 10 ug/mL are sufficient to meet these

requirements.

TLS as an analytical technique suggests several basic disadvantages.

Generally, large amounts of data are generated. For example, in the

PAH TLS spectra used throughout this study, there are more than 35,000

data points in each data file. This number can be reduced substantially

by reducing the spectral range or by increasing the sampling interval,

and therefore, reducing spectral resolution. Also a much more elaborate

experimental and data handling system is required over conventional

luminescence instrumentation, especially if an image detector is used.

Often this makes its cost prohibitive for routine use. Within the TLS

data matrix, there is extensive data redundancy. Later, this redundancy

will be utilized for statistical analysis of the data set. Finally,

data collection and evaluation are more time-consuming. Some of these

difficulties are instrumental in nature and improvements will

inevitably come. However, some of these difficulties are inherent in

TLS.

Thus, it appears that while extensive work has been done to allow

for more data to be generated at increasingly faster rates, that is,

instrumental improvements have been made, one is still faced with the

need to interpret these data, whether manually or by computer assistance.

The problem of data redundancy and the generation of voluminous amounts

of data is shared by other analytical techniques.








For example, during the course of a capillary GC-MS experiment,
it is easily possible to obtain over 7000 mass spectra per hour.
There is every indication that the development of liquid
chromatographic-mass spectrometer (LC-MS) systems will follow
the same course. While LC-MS may not require the fast scan times
of GC-MS, each LC-MS experiment on a complex mixture will
generate a tremendous amount of mass spectral data. The analyst
is often only interested in some small subset of these data;
and it is becoming increasingly necessary to utilize the speed
of computers to search through the spectra obtained and screen
for compounds of interest in the particular analysis. (54)

As improvements are made in multichannel detectors, TLS spectra can be

collected more rapidly, more reliably, and with greater sensitivity.

These developments are extending the utility of TLS as detection

systems in HPLC and TLC.

To the analytical chemist, TLS is more than the generation of

large amounts of data. TLS data are a superset of the data from other

luminescence techniques. Figure 12 depicts the relationship between

TLS and luminescence techniques such as Synchronous Luminescence

Spectroscopy (SLS). Perhaps by studying TLS spectra, SLS will be better

understood, making optimization more systematic. Instrumental and

mathematical similarities make the study of SLS a natural extension of

the principles set forth in TLS.
































0
4-3

0)





S-









4-)
aI


a)




E3

r-
C



a-




SE
U






a1






4.2
C-)
(UI








-,
a)
U







cr


















U-
*w
LL
S-
0)
rrr











I-
CL











D
Sn




z
0





N -
LD
z
z
u-i
-J
.t. z



g '



u -







0 z
Sz 0










ED
u, ED


(WN) Hi9N313AyM NOIIU1I3X3













CHAPTER 3
SYNCHRONOUS LUMINESCENCE SPECTROSCOPY

Theory

In the analysis of samples with completely unknown luminescence

characteristics, TLS appears to have many desirable qualities.

However, in many analyses, the luminescence characteristics of the

sample are approximately known, as in the search for an impurity in

a product, or where slight changes in spectral characteristics provide

the desired information. In these cases, TLS generates more data than

is necessary. If one has catalogued the spectral regions characterizing

the compound of interest, the focus of the analysis should be on this

spectral region. That is, previous information about the sample's

characteristics should be used to limit the analysis to those spectral

regions providing the greatest amount of information for analyte

identification and quantitation. Synchronous Luminescence Spectroscopy

(SLS) offers an alternate solution to this problem.

SLS requires the simultaneous scan of both the excitation and

emission monochromators, synchronized so that a well-defined

relationship is maintained between the wavelengths. Conventionally,

this relationship is a constant wavelength difference. Mathematically,

SLS data closely resemble TLS data. Equation (7) also defines each

SLS data point in a single component sample with one added restriction.

That is,

Xj-Ai = AA (11)
where Ax is the wavelength difference between the excitation and








emission monochromators. With this added condition, SLS becomes a

subset of TLS data. The synchronous scan path can be envisioned as a

line through the TLS plane. In this way, one can again take advantage

of the selectivity capability of molecular luminescence spectroscopy

by varying both the excitation and emission wavelengths during analysis.

SLS,as an analytical technique,depends not only on the compound's

luminescence excitation characteristics or the luminescence emission

characteristics, but on both,and especially the relationship between

the two. From the perspective of the energy level model from Chapter

1, the analyst is not interested in a single process, but rather, the

relationship between two processes. Proper application of SLS

techniques to multicomponent samples requires an understanding of the

luminescence principles in Chapter 1 and the relationship they hold to

TLS theory.

There are several advantages of SLS over conventional luminescence

techniques that make it a technique for consideration in multicomponent

analysis (55). The spectra obtained in SLS are reduced in complexity.

Another selectivity advantage is the reduction of spectral bandwidths

because of the scan process employed. A practical improvement is a

reduction in the effect of Rayleigh scatter on the data. This

technique can be combined with several conventional luminescence

measurement systems. The instrumentation of SLS, as well as data

handling requirements,.is significantly simpler than in TLS. The

analysis focuses on the key spectral regions of known analytes,

making SLS useful for qualitative and quantitative mixture analysis.

Perhaps a more thorough discussion of each of these characteristics

will reveal some of the analytical potential of this technique.








The reduction in spectral complexity is best demonstrated with an

example. Figure 13A shows the conventional fluorescence spectra for

a solution of anthracene in n-heptane. The SLS spectrum for this

same solution using a AA of 3 nm is shown in Figure 13B. As will

become clear, the selection of AX determines the complexity of the

spectrum obtained. The objective is to provide sufficient complexity

for analyte identification, yet maintaining enough simplicity to

make multicomponent analysis practical.

All SLS spectra are plotted as a function of emission wavelength,

as shown in Figure 13B. Given the position along the emission wave-

length axis, the excitation wavelength can be easily calculated using

the conditions described to obtain the spectrum, such as a constant

wavelength difference. Thus,the SLS spectrum is plotted as a

projection of the data array onto the emission wavelength axis.

An improvement in selectivity is difficult to quantitate. A clear

demonstration of this improvement is the band narrowing capabilities

of SLS. This advantage has been stated by proponents of this technique

but has not been quantitatively evaluated. One published quantitation

of this improvement by Andre et al. (56) states that

s = +6 -(2 +2)1/2 (12)
s x m x m
where 6As is the effective synchronous bandwidth (nm), 6Xx is the

luminescence excitation bandwidth (nm), and 6xm is the luminescence

emission bandwidth (nm). This, once again, demonstrates the dependence

of SLS characteristics on the characteristics of both the excitation

and emission spectra. While this equation seems to yield approximate

results for limiting cases, no derivation of this equation is

available in the literature.





























Figure 13. A.


Conventional RTF excitation and emission
spectra of anthracene in n-heptane.


B. SLS spectrum of anthracene in n-heptane
using a Ai of 3 nm.

































250 300 350 400 450 500 550 600

WAVELENGTH (NM)




B




















250 300 350 400 450 500 550 600

EMISSION WHVELENGTH (NMI








To gain an understanding of the quantitation of this effect, a

geometric model and an algebraic model were designed. A basic

assumption was made that the luminescence band can be described as a

Gaussian function on both the excitation and emission axes and that the

synchronous scan path passes directly over the center of the band. If

the bandwidth is defined as the full-width-half-maximum, the bandwidth

of the entire band defines an ellipse. Using the simple relationships

shown in Appendix 2, the relationship
s = / 2 2)1/2 (13)
s x m x m
is obtained. Three cases are of special interest. If the excitation

and emission bandwidths are approximately equal, the bandwidth defines

a circle and the band is completely symmetrical around the band

maximum. For this case, Equation (13) reduces to

6As= 6x // = 6&m/ (14)
This is envisioned by realizing that although the scan path passes over

the same band size, the SLS spectrum is plotted as a projection onto

the emission wavelength axis. In room temperature fluorescence

spectroscopy, this case has significance for many compounds. While a

2- reduction seems modest at best, it is substantial under these

conditions. For example, a spectrum with a bandwidth of 30 nm is

reduced to 21 nm, important for a technique that is limited by

extensive band overlap. Two other limiting cases are also of special

interest. If the excitation bandpass is much greater than the

emission bandpass so that sax >> m Equation (13) reduces to

6Xs = Dm (15)

Analogously, if the reverse is true, SAm >> Xx and

xs = 6x (16)








This confirms the characteristic of SLS spectra that its features follow

most closely the features of the excitation or emission spectrum with

the narrowest features. For some compounds, the emission spectrum is

more structured than the excitation spectrum, while for other compounds,

this condition is reversed. In SLS, only one of the two conventional

spectra needs to contain sharp spectral features for the SLS spectrum to

also be well-structured, a natural consequence of the multiplication

effect describing this technique.

To model luminescence data more closely, luminescence bands are

more effectively described in energy-related terms. Wavelength units

are approximations and are limited to special cases. Appendix 2 gives

the conversion of these equations to energy units. For the following

discussion, energy terms are used.

Algebraically, similar results are obtained. For the previous

assumption of Gaussian intensity functions, xi and yj are defined as
= x exp(-(ui- )2 /(2 2)) (17)
0 0
and

y = yexp(-(u-u. )2/(2a)) (18)
Y Yo J -o J
where xi and yj are intensities at wavelengths xi and x., xi and y

are intensities at the peak wavelengths x. and o, u', ui U, and

Uj are the energy values corresponding to the respective wavelength

terms, and oi and a. are the standard deviations of the Gaussian bands,

given in K. If 6di and 6sj are the energy equivalents of 6Xi and 6xj,

and 6ux /Ui and um /u.o are less than 0.05, then ai and ja are

proportional to 6ux and S m. The exact relationship is given by

ai = ((l+(6x/i )2) 1/2-1)((1076u x/u )(21n2)1/2)-1 (19)
Soo o o o
with an analogous equation for ja. (57). With the details given in
J








Appendix 2, the synchronous bandwidth again reduces to Equation (13).

Using energy units for these calculations provides a more realistic

model than the wavelength approximation. In SLS, the scan paths need

not pass exactly over the excitation and emission maxima. This presents

a more difficult situation. A more general equation can be given for

6 s. Equation (13) is a limited case of the expression
-2 -2 1/2 -2 -2 2)1/2
S= ( x /(Sux+2 )/2)( +6m-(u -u-u )2)1/2 (20)
-0 0
where Au is the energy equivalent of AA.

The reduction in Rayleigh scatter is a major consideration in SLS.

This characteristic is explained by examination of the relationship

between TLS and SLS, pictured in Figure 12. Rayleigh scatter is

observed when the bandpasses of the excitation monochromator and the

emission monochromator overlap, reaching a maximum along the line where

Xi and Aj are equal. The Rayleigh scatter line can be thought of as an
SLS scan with a AX of 0 nm. That is, the SLS scan path runs parallel

to the Rayleigh scatter ridge. Proper selection of AX and the

bandpasses eliminates this scatter.

TLS and SLS can be used under a variety of sample conditions.

Fluorescence or phosphorescence can be measured at room temperature or

at low temperatures. Implementation of SLS requires only the

instrumentation required for the conventional luminescence technique.

SLS data requirements are much less than those for TLS, reducing the

computational and data storage needs.

SLS has had limited success since its introduction by Lloyd in

1971 (58-61). This group was the only one to publish in this area until

the work of Andre and Vo-Dinh. Lloyd's work has focused on oil sample

screening, with empirical rather than theoretical developments. The








work of Andre extended the application of SLS to drug samples (56) and

fluorescence tracers in hydrology (62), again with little theoretical

backing. The work of Vo-Dinh, and especially his paper in 1978 (63),

was the first to attempt to force development along a path backed by

luminescence theory. The same year saw a question raised as to the

practical potential of SLS (64). Recently, Vo-Dinh has presented a

review of SLS principles and applications (55). Significant improve-

ments are necessary to make SLS more than a curiosity in the analytical

literature.

Instrumentation

Thus, the stage was set for the current investigation of SLS and its

potential in multicomponent analysis. The same experimental system that

was described in Chapter 2 was used for this work. Additional software

was written as needed to fit the overall software philosophy. Several

software developments are worthy of mention. With a database of TLS

spectra on file, standard SLS spectra can be obtained from either the

TLS data files, using the appropriate extension, or generated from

the excitation and emission spectra in the VEC files. A thorough

investigation demonstrated that the generated spectra and instrumentally

obtained SLS spectra agree very closely. Thus, to obtain an SLS spectrum

using a different AA, the generation parameters need only be changed

accordingly. Thus a new experiment is not necessary for each desired

spectrum. Seemingly trivial, without this ability, this entire project

would be prohibitively time-consuming. The database requirements for

TLS data files complement those of SLS, again providing considerable

data storage savings.

Bandpass selection influences experimentally obtained SLS spectra

and should also be reflected in the standard spectrum generation program.








That is, a separate database for each bandpass is undesirable.

Fortunately, a similar problem has been studied in conventional lumi-

nescence spectroscopy. For this system, a triangular function is used

as the transfer function of the monochromator and the spectrum is

corrected accordingly. The new spectrum is determined in the following
manner. If 6 1 is the desired effective bandpass and no0 is the band-

pass used to generate the database spectra then
12
F = z kSk+w (21)
k=l
1
where Fw is the corrected relative intensity at wavelength w Sk+w is

the intensity in the original spectrum at wavelength xk+w, sk is the

scaling factor, 11 is the lower wavelength limit, and 12 is the upper

wavelength limit. Generally 11 and 12 are set equal to -SAl and a6l'

respectively, and sk is defined as

sk = l-lkl/Sx1 (22)
Several examples using this triangular approximation resulted in an

average 3.2% error for assumed Gaussian peaks. Thus, the bandpass

effects can be quantitatively predicted.

One limitation is imposed on the above bandpass correction; 6SO
must be smaller than a l. Because of this, database spectra were

obtained with 2 nm bandpasses, the smallest practical value for RTF

using this system. This correction procedure is more properly labeled

a bandpass degradation calculation.

Constant Energy Synchronous Fluorescence Spectroscopy

Background

Once the instrumentation development was completed, a comprehensive

evaluation of SLS was begun. Several major disadvantages became obvious,








perhaps limiting its practical application. Problems included the

requirement for more a priori knowledge about the sample of interest

before analysis, the observation that improper instrumental parameter

selection may not detect all the luminescing compounds present, that

is, "special blindness," and the key spectral regions must be well

characterized prior to analysis. Essentially, these are all directly

related to parameter selection. More specifically, the selection

of AX is critical.

Stated by one author,

the value selected for the step (AX) is arbitrary, but
maximum sensitivity is obviously obtained when the value
corresponds to the separation between the maxima of the
excitation and emission spectra.(56)

In a previous evaluation of SLS by the Winefordner group, AX

selection seemed to be fortuitous (64-66). The recommended AX is one

that results in the best resolved and simplest spectral features. If

the maxima separation rule is used, n values for AX must be selected

for the n components present. This is similar to selective excitation.

In previous theoretical discussions of SLS, the simplest case was

selected to demonstrate the technique's capabilities. That is, a AX of

3 nm is usually used, approximately equal to the Stokes Shift for many

PAHs in n-alkane solvents (63,67). Then by selecting carefully chosen

compounds, synthetic mixtures are resolvable (68). However, theory

breaks down for the remaining non-ideal compounds. A major group of

compounds causing difficulty are those molecules with a weak S1 0-0

absorption. Examples of this are pyrene, 1,2-benzopyrene,

1-methylpyrene, and 3,4-benzopyrene, which are analytically significant

compounds.








What has been obviously lacking in SLS is a strong theoretical

foundation, one that can be traced to invariant photophysical properties

of the analyte. Luminescence theory has been building over the past 30

years, so a vast storehouse of information is available for

consideration. If, as I. M. Kolthoff said, "Theory guides, experiment

decides," proper guidance is needed.

In 1978, Weiner (69) suggested the use of Total Luminescence

Spectroscopy to optimize the selection of AX in SLS. He noted that the

main difficulty in SLS is in finding the best value for AX. In fact,

he described the selection of a range of AX's with the SLS spectra

combined to reconstruct the TLS spectrum. This short note appeared

fairly trivial, but contained a concept that has been consistently

overlooked by SLS users. In the past, the development of SLS and TLS

has been completely independent, with the suggestion that the techniques

are competitive rather than complementary.

By using SLS, the data generated are substantially less than those in

TLS. But to be effective in multicomponent analysis, the data selected

must retain as much information as possible. The analytical value of a

subset of TLS data, such as an SLS scan, varies considerably, depending

on the parameters of subset selection. Based on TLS theory as outlined

in Chapter 2, a natural suggestion is to use a different subset of the

TLS spectrum, one that more clearly identifies characteristic spectral

regions. Experimental implementation of this change translates to

selecting a different scan route. Mathematically, this corresponds to

changing the restriction imposed by Equation (11).

Due to the nature of the scan process in conventional SLS, it can

be best described as Constant Wavelength Synchronous Luminescence








Spectroscopy (CWSL), to contrast it with other scan modes. Identified

as Variable Wavelength Synchronous Fluorescence Spectroscopy, the first

suggestion was to scan the two monochromators at different rates. By

scanning both monochromators at constant rates, with the emission mono-

chromator scanned faster than the excitation monochromator, a scan path

is traced through the TLS plane defining a line similar to CWSL, not on

a 45* line relative to the emission wavelength axis, but rather less

than 45.

The instrumentation described for SLS was designed for a Variable

Wavelength Synchronous Luminescence Spectroscopy (VWSL) scan, with CWSL

representing a special case. The scan rates, dxi/dt and dXA/dt, are

defined as

dA./dt-di./dt = c (23)

where c is a constant greater than zero. In CWSL, c is set equal to

zero. Again, reference spectra generation is possible using the same

database.

Evaluation of VWSL indicated immediately that it is hampered by most

of the same problems of CWSL. The lack of a solid theoretical foundation

again makes multicomponent analysis difficult. Although there have been

reports of research in this area, no publications are available. This

evaluation was soon abandoned after a more favorable alternative was

discovered.

The evaluation of TLS data and relationships that exist between

spectral regions of PAH and pharmaceutical solutions has led to the

development of the new method, Constant Energy Synchronous Luminescence

(CESL), with fluorescence, Constant Energy Synchronous Fluorescence

(CESF), and phosphorescence, Constant Energy Synchronous Phosphorescence

(CESP), applications. This technique is performed by maintaining a









constant energy difference between the excitation and emission wave-

lengths. The restriction placed on the TLS data matrix is mathematically

defined as

G = (1/X.-1/x.)*107 (24)

where A- is the energy difference between wavelengths, in K. Evaluation

of this technique was directed towards Raman scatter interference reduc-

tion in pharmaceutical analysis and information retention in the analysis

of PAHs. The relative merits of this technique against conventional SLS

and conventional fluorescence techniques will be presented.

Raman Scatter Reduction

Introduction. The fluorescence emission spectra of most common

solvents used in RTF exhibit a strong Rayleigh scatter signal when the

excitation and emission wavelengths are identical, and a weak Raman

scatter signal, shifted in energy so that

AR = (1/xx-1/m)*107 (25)

where A R is the energy of the active Raman transition, in K, \x is the

excitation wavelength, in nm, and xm is the emission wavelength, in nm.

The contribution of the solvent can become significant if the

fluorescence signal is small and falls within the spectral region of

the solvent Raman scatter.

The intensity and position of the Raman band is dependent upon the

solvent used. However, similarities can be found among commonly used

solvents. Figure 14 shows the Raman scatter for cyclohexane, n-heptane,

ethanol, and water excited at 330 nm. All four were measured under the

same instrumental conditions, including amplifier gain. Spectral band-

passes of 8 nm were used with each monochromator. The Raman band































C
0 "-C L
E CU 1 W


EJ r- U

> I>
4- 3'a
E
S- *C
,0-, 0 Q
4- r-
,I II

(0 C 4-+ E
Ci0 '*- t
A X u

C C S-L
00 W

cn U m- 4-)
.r- 0 "












4--
L )C


U L 4C



S> II
3 r- *0
e- 0 C >,


a *
C4Ur
E C 0 E

*r- 4fl C: 0

>0
0 i- *
C.) 9- S3




















ED





E o




Sz
Ci








oN
C
C






C


IUSN3iNI 3AIlUI3








positions agree with the Raman shifts reported by Parker (2) for the

solvents, shown in Table 7. All solvents with a CH or OH group show

a weak Raman band with an approximate 3000 K shift.

The total emission of water clearly demonstrates the spectral

relationship of the Rayleigh and Raman scatter (Figure 15). The line

through the Raman ridge corresponds toCWSF scans with a AX of 32, 37,

and 42 nm. Upon examination of the spectra obtained from these three

scans, Figure 16, two effects are clearly noticeable. First, the

Raman band becomes broadened compared to the Raman band in conventional

fluorescence spectroscopy. Second, the position of the Raman band is

determined by the selection of AX. The lines through the total emission

spectrum run parallel to the Rayleigh scatter ridge using CWSF.

Plotted on a wavelength axis, Figure 15 shows that the Raman scatter

ridge and Rayleigh scatter ridge diverge, maintaining constant energy

difference between the two. It is this basic natural relationship that

forms the foundation for Constant Energy Synchronous Fluorescence (CESF).

In this technique, the curve defined through the TLS spectrum runs

parallel to both the Rayleigh scatter ridge and the Raman scatter

ridge when plotted in energy units.

The spectral region around the Raman scatter ridge is an analytically

important region. Many pharmaceuticals fluoresce in this region, as

well as other aromatic compounds, such as phenol (70). Therefore, a

method capable of Raman scatter reduction or elimination could provide

widespread utility.

Raman scatter can generally be subtracted out with a blank measure-

ment. Rayleigh scatter is almost impossible to subtract out and

theoretically should not be the same for the sample and blank solutions.















Table 7. Raman Bands Observed in Common RTF Solvents


Solvent


Water

Ethanol


Cyclohexane


Chloroform


Energy Shift (K)a


3380

2920
1400

2880
1360

3020
700


aReference 2


II

































n-




a) c
,o a


I-> I
C *
01--



*0
S II
(L) 4I-
4-) C: e<
3 a)


S- -

'4- a) -=
S- 4I-)

0 Q.) 0-






,-- 4-3
e C







cr- S.-
Lo 0 0






0- =1 -





c 0m S
S00 C



4-) S- C
o L
4-)
0 .C 3











LOL
Ie -1~


*~E~
LI..cc











IS







\ E








z
0
th















O0 C
ln 0 m


(WN) HI9N313AUM NOI01IJX3

































o c



4-
0






C ;



*r-




cc


cLa
S-


U
0)

















LL
cn -


C0

LE





r-








Q.
c*



















0,
r-
Q..




88







II)









z


LD
--Z
z





0
O
/-4
(U























IISN31NI 3AI873N









Because Raman scatter is due to the solvent, its intensity is about the

same from solution to solution with the same solvent. However, many

situations make a blank spectrum unavailable or a truly representative

blank unachievable.

Instrumentation. All CESF measurements were made on the system

described in Chapter 2. In order to scan the monochromators and to

maintain a constant energy difference between the two, the emission

monochromator is stepped at a constant speed while the excitation

monochromator speed is varied. A linear approximation is made for this,

dividing the spectral range to be scanned into 1000 segments. The

emission monochromator is scanned by a fixed increment, while the new

position for the excitation monochromator is calculated and moved

accordingly. The PMT signal is integrated over approximately 0.3 s,

using an RC constant of 0.1 s. The speed of the FORTRAN calculations

makes the scan essentially continuous, with the calculation time

negligible. Thus a single scan requires approximately 5 minutes.

Results and Discussion. To illustrate the relationships among

conventional fluorimetry, CWSF, and CESF, tripelennamine hydrochloride

was selected as a model compound. Figure 17 shows the conventional

fluorescence emission spectra for solutions of 0.0, 0.1, and 1.0 ug/mL

tripelennamine hydrochoride in water. Note the presence of the Rayleigh

scatter band and the position of the Raman scatter relative to the

fluorescence. The total luminescence isometric projection of the

1.0 pg/mL solution illustrates the relative positions of the spectral

features (Figure 18). The bandpass selected for both monochromators


























E
S -
O




.-a



S-
0 e.-




0 e-
O










0
s- -












0 .
U) *


0
"C4


r- *


L)


0)C0

* .- 4..
) *



0 C:
4- 0
O *"--



C) t o
>-




r-e
0 3
C-



,.0
'r
ccnC
01~-
|Zl~
























z


L-
(I

-j



mm
0
1-4

U,























IISN31NI 3AIIB13S































4-
O

O
C
0



rl-


E







4r-
0

CL u
O



S-
O



eo










S-
Sr-
CO

Q- C
0-

au









5.-
I-I-





1]/



cr-