Improvement of signal-to-noise and selectivity in fluorescence spectrometry

MISSING IMAGE

Material Information

Title:
Improvement of signal-to-noise and selectivity in fluorescence spectrometry
Physical Description:
x, 166 leaves : ill. ; 28 cm.
Language:
English
Creator:
Walden, Gary Lyle, 1951-
Publication Date:

Subjects

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

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1979.
Bibliography:
Includes bibliographical references (leaves 160-166).
Statement of Responsibility:
by Gary Lyle Walden.
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 - 000100369
notis - AAL5830
oclc - 07390744
System ID:
AA00003488:00001

Full Text










IMPROVEMENT OF SIGNAL-TO-NOISE AND
SELECTIVITY IN FLUORESCENCE SPECTROMETRY





BY

Gary Lyle Walden


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


1979














ACKNOWLEDGI21' N:TS



I wish to express my graditude to my research director,

Dr. James D. Winefordner, for his constant help, encourage-

ment, and belief. Special thanks to Jim Bower and Jimmie

Ward for their aid and advice. Thanks to Dave Bolton for

help with the assembly language programming and to Jeanne

Burton for the hours spent on this document. Also, my

graditude is extended to the JDW group, past and present.

I would like to thank my parents, Raymond and Elma

Walden, for their support, and finally to my wife, Marsha,

for her sacrifice, patience, and devotion.














TABLE OF CONTENTS


Page

ACKNOWLEDGEMENTS ii

LIST OF TABLES vi

ABSTRACT vii

CHAPTER

I BACKGROUND INTRODUCTION 1

Noise Power Spectra 2

Fluorescence Theoretical Background 4

Experimental Approach 9

Fluorescence Temporal Measurements 10

PM-Boxcar 11

Transient Digitizer 12

Streak Camera 12

Time-Correlated Single-Photon Counting 14

Optical Gating 15

Cross Correlation 16

Phase Shift 16

Power Spectra-CW Laser 17

II NOISE POWER SPECTRA OF THE ICP 19

Introduction 19

Power Spectrum 20









Page

Fast Fourier Transform (FFT) 20

Noise Components 24

Experimental 24

Results and Discussion 31

Spectra 58

Conclusions 61

III FLUORESCENCE TEMPORAL MEASUREMENTS 66

Introduction 66

Time Resolution 67

External Heavy Atom 74

Per Pulse Fluorescence 76

TEA N2 Laser 76

PM-Boxcar 82

Streak Camera 86

Computer Link 87

Triggering 87

Computer Routines 89

SIT Camera 90

Experimental 92

Results and Discussions 95

PM-Boxcar 95

Streak k Camnra 101

IFluorescence Lifetimes 105

Limits of Detection 108

Mixtures 111








Page

External Heavy Atom 116

Methods Comparison 117

IV SUBNANOSECOND PULSED DYE LASER 122

Introduction 122

Dye Laser Design 123

Results and Discussions 126

V SUMMARY AND FUTURE WORK 129

APPENDICES

A FORTRAN PROGRAM USED FOR ICP DATA REDUCTION 134

B ASSEMBLY LANGUAGE PROGRAM FOP THE FAST
SAMPLING OF ONE CHANNEL OF THE LPS-11 141

C FORTRAN PROGRAM USED FOR DATA TRANSFER FROM
THE TEMPORAL ANALYZER AND FOR FLUORESCENCE
LIFETIME DETERMINATION 145

D ASSEMBLY LANGUAGE PROGRAM FOR DATA TRANSFER
FROM THE TEMPORAL ANALYZER TO THE DEC PDP
11/34 MINICOMPUTER 156

LIST OF REFERENCES 160

BIOGRAPHICAL SKETCH 167














LIST OF TABLES


Table Page

I. Experimental Equipment Used for Noise Power
Spectra of ICP 25

II. Noise Components of the ICP With Varying
Operating Parameters 32

III. Summary of the Effects of Changing ICP Operating
Parameters Relative to the Base Conditions 62

IV. Experimental Equipment Used for Temporal
Measurements 93

V. Fluorescence Lifetimes with the PM-Boxcar
System 100

VI. Fluorescence Lifetimes Measured with the
Streak Camera System 102

VII. External Heavy Atom Effect on Fluorescence
Lifetimes 103

VIII. Streak Camera Limits of Detection and Linear
Dynamic Range 110

IX. Comparison of Fluorescence Lifetime Measure-
ment Systems 118

X. Dye Laser Relative Outputs 127
















LIST OF FIGURES


Figures

1. Energy Level Diagram


2. Block Diagram of Experimental Setup for ICP


Page

6


Power Study


Powe r

Power

Power

Power

Power

Power

Powe r

Power

Power

Power


Spectrum,

Spectrum,

Spectrum,

Spectrum,

Spectrum,

Soectrum,

Spectrum,

Spectrum,

Spectrum,

Spectrum,


13. Noise Power Spectrum,


14. Theoretical


PM Dark Current

Water Background

40 ppm Zn

P = 1.0 kW

Q = 20 L/min

2000 ppm Zn

N = 1.5 L/min

100 ppm Ba

20 ppm Li

Commercial Type Torch

100 opm Ba-Aliasing


Approximation of Fluorescence


Temporal Response to an Excitation Pulse

15. Natural Logarithm of Summation of Two Exponen-
tial Decays

16. Block Diagram of Experimental Setup Using the
PM-Boxcar System

17. Block Diagram of Exporimental SoLup Using the
Streak Camera System

18. Wiring Diagram of Photomultiplier Dynode Chain


vii


Noise

Noise

Noise

Noise

Noise

Noise

Noise

Noise

Noise

Noise

Noise


3.

4.

5.

6.

7.

8.

9.

10.

11.

12.









Figures Page

19. Photomultiplier Saturation with Anthracene
Fluorescence at Various Wavelengths 99

20. Quinine Sulfate Fluorescence Temporal Response
to the TEA N2 Laser and Natural Logarithm of
Decay 107

21. Cephradine-Quinine Sulfate Mixture Fluores-
cence 115

22. Block Diagram of TEA N Laser Pumped Dye Laser
for Subnanosecond Excitation Pulses 125

23. Flow Chart for FORTRAN Program Used for ICP
Noise Power Study 136

24. Flow Chart for Assembly Language Program
Used for A/D Sampling 143

25. Flow Chart for FORTRAN Program for Data
Transfer and Fluorescence Lifetime Determina-
tions Using Streak Camera System 147

26. Flow Chart for Assembly Language Pror r.iri Used to
Transfer Data from the Temporal Analzyer to the
DEC PDP 11/34 158


viii














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


IMPROVFIIl'T OF SIGNAL-TO-NOISE AND
SELECTIVITY IN FLUORESCENCE SPECTROMETRY

By

Gary Lyle Walden

December 1979

Chairman: James D. Winefordner
Major Department: Chemistry

The aim of this work was to improve molecular fluori-

metric analysis. Since the noise attributed to an analysis

system is the limiting property of the determination, the

noise spectral analysis of the system is important. The

noise power spectra of an inductively coupled plasma (ICP)

were determined as an example of a typical radiative source.

The system used for the fluorimetric analysis was a pulsed

source-gated temporal detector.

Noise power spectra for the ICP were determined under

various conditions by Fast Fourier Transform (FFT) digital

techniques. The quantitative results for the three major

noise components are given. These noise components are

white noise, low frequency noise, and proportional noise.

The proportional noise observed increased with concentration

of analyte, radio frequency (RF) power, nebulizer flow rate,

ix








and coolant gas flow rate. Increasing the observation wave-

length increased the frequency at which the low frequency

component reached 10% of its maximum amplitude. Changing the

observation height in the plasma determined which noise com-

ponents were visible in the power spectrum. At observation

heights near the power coils, all three noise components were

present, whereas, at a significantly greater height, primar-

ily the low frequency noise dominated. Changing the torch

design changed the relative amplitudes of the different pro-

portional noise components, but did not greatly change the

central frequencies. Changing the drain tube length or

cavity volume had little effect on the noise power spectra.

Also reported is the temporal measurement of fluores-

cence signals using two different pulse measurement techni-

ques. The first measurement technique discussed is a photo-

multiplier tube-boxcar (PM-boxcar) approach. The second is

a streak camera system. Although the PM-boxcar system was

more sensitive, it suffered from poor overall temporal resolu-

tion due to trigger jitter and PM temporal response. The

streak camera approach is by far the best in terms of tem-

poral resolution, however, the signal levels needed for its

operattion :are qui ite I e. Th( streak camera was tised to

determine external heavy atom effects on fluorescence li fe-

times and an attempt was made to recover the fluorescence

lifetimes of two spectrally overlapping compounds in a

mixture. A comparison of the most often used temporal

metasuroment techniques for fluorescence is given.

x















CHAPTER I
BACKGROUND INTRODUCTION


When molecular luminescence is used for analysis, the

analyst must be aware of the overall signal-to-noise ratio

(S/N) of the method. Whenever he attempts to improve a

method, he is trying to increase the signal and/or lower the

effective noise in order to increase S/N. Through the use

of theoretical S/N consideration, described by Alkemade

et al. (1), Boutilier et al. (2,3), and Boutilier (4), it

is possible to determine certain experimental approaches

which could lead to an increase in the S/N. However, this

must be done only with the inclusion of previous experimental

knowledge and must be tested under actual experimental condi-

tions. To quote I.M. Kolthoff, "Theory guides, experiment

decides."

The effect of the S/N on analytical figures of merit are

important to visualize. The limit of detection (LOD) is

the most obvious case in which an improvement in S/N results

in a lower LOD. Using the TUPAC definition (5), the limit

of detection is the analyte concentrate ion, or amount, that

corresponds to a measure (signal) equal to three times the

standard deviation ("noise") of the measurement of the blank

for a sufficiently large number of blank determinations

(16 suggested). More simply, the concentration that results

in a S/N of three. 1








Another figure of merit affected is the linear dynamic

range (LDR). An increase in the S/N generally results in

an increase of the LDR, barring any other change in the ana-

lytical calibration curve, and hence, increases the useful

working range of the method. Also by increasing the S/N,

the percent relative standard deviation (%RSD) decreases,

which is an improvement in the precision of the method.



Noise Power Spectra


As implied from the definition of limit of detection,

if a signal is present, from a given system, the S/N can be

improved simply by decreasing the noise component to its

absolute minimum (assuming that point has not already been

reached). The determination of the noise components in a

given system and their contribution to the overall noise of

the method is of paramount importance in improving any

method. The determination of the noise components in a sys-

tem is experimentally accomplished by obtaining the noise

power spectra for that system. Classically, this has been

done in a rather laborious fashion (6-8). With the advent

of modern computers and the development of the Fast Fourier

Transform (FFT) (9-11), this process has been simplified (12).

At the present, there are several commercially available FFT

instruments. With either method, the information obtained

is a power measurement per frequency interval (units pro-
portional to W Hz-) over a given frequency range. As will be
portional to W Hz ) over a given frequency range. As will be








discussed later, this information can be used to define

analytical working conditions or even to localize and elimi-

nate or minimize noise sources. It is also possible to use

these measurements to decide upon experimental setups where

signal power might be greater at certain frequencies and/or

where noise power is lower.

As an application of the determination of noise power

spectra, the present work is applied to a popular atomic

emission source, an inductively coupled argon plasma (ICP).

The noise power spectra obtained indicate the possible appli-

cation of an AC detection system. The use of noise power as

a diagnostic tool for torch design and characterization is

also indicated.

Previous workers (6-13) have described the power spec-

tra of several sources of analytical importance. In most

cases, the noise power spectra are composed of a low fre-

quency, assumed to be of the 1/f variety, component of con-

siderable amplitude, proportional noise, spikes due to the

60 cycle line frequency and its harmonics, and a white noise

background that extends throughout the spectra. Also report-

ed in certain cases are higher frequency proportional noise

components, so called whistle noise (1). In the present

work, all of these noise types were observed in the ICP

under various conditions. Different experimental parameters

were changed and the effects noted.








Fluoirescence Theo rtical Background


In molecular luminescence spectrometry, the noise

sources are a subset of the type of measurement approach

chosen. In the present, work, molecular fluorimetry is the

technique used.

Fluorescence of an organic molecule corresponds to the

radiative transition of an excited molecule from its excited

state singlet, S1, to the singlet ground state, SO (refer

to Figure 1). If the molecular population of a single

species is instantaneously displaced from equilibrium, e.g.,

by a burst of excitation light, the population will decay

back to equilibrium by emitting fluorescence radiation that

will follow an exponential function given by

_t/T F
I = oe F (1)


where t is time, I is the observed fluorescence intensity,

10 is the fluorescence intensity at time t = 0 (where t = 0

is some time after the burst of excitation light has ended),

and TF is the observed lifetime of the fluorescence. The

observed fluorescence lifetime is determined to be the time

it takes the intensity of the fluorescence to decay to 1/e of

its value at t = 0. Hence, the observed lifetime is the

rec'iplrocal of the rate at which the excited population decays

back to equilibrium. This rate is determined by a combina-

tion of rate constants of the different processes that are

occurring in the excited state molecular population (see





























Figure 1. Energy Level Diagram

k = rate constant

A = Absorption
F = Fluorescence
P = Phosphorescence
IC = Internal conversion (radiationless transi-
tion between states of equal spin)
ISC = Intersystem crossing (radiationless transi-
tion between spin forbidden states)
ISC = T -+ SO (radiationless transition)
SO = Singlet ground state
S1 = Excited singlet state
T1 = Excited triplet state



























!



11111
III|I

I lrii

llIII
II Il I
11111
II(I
ti;i
(Hi!
k ',',' k *
IIIll


11,11
S(II I
lilli
Itall



'lilt
II II
IIII
I I II
IIII

IIIII
iIII
1II(
"11


_ .... ___ -| _t vt 3l ......
_ ... ____ T11 ....--
IIII JY,'T


k
'SC


Si












k


k
F


I ll
liii'
fi ll











II I
11111
I l l
Illit

11111
11II1








II III
gl ilt
II Ill

'lilti
II Ill

11111
11111
II III



11111

11111
tilli
II III


IIIll'
II III
litil t


ATI


~"T~'


---


I








Figure 1). There are three main processes contributing to

the decay of the excited state (14-17): (i) fluorescence

(spontaneous emission), (ii) internal conversion, (radiation-

less transition between S1 and SO), and (iii) intersystem

crossing (radiationless transition between S1 and T1). Their

respective rate constants (s-1) are kF, kic, and kISC. The

observed fluorescence lifetime is given by


1
T= (2)
F = k + kc + ks (2)
F IC ISC


Another important quantity that can be measured is

the fluorescence quantum efficiency, YF dimensionlesss).

This is the fraction of molecules that decay by fluorescence

from the excited singlet. It can be expressed by

kF
Y = (3)
F k + ki + k(3)
F IC ISC

From the observed fluorescence lifetime (s) and the quantum

efficiency dimensionlesss), the spontaneous radiative life-

time, TR, is obtained,


1 F
T F (4)
R kF Y
F F

The radiative lifetime is the lifetime that would be measured

if no other processes were involved.

Also of importance is fluorescence quenching. This

effect is attributable to many causes, the majority of them

being environmental effects. Examples of quenching (14-17)







are: temperature quenching, oxygen quenching, concentration

quenching, and impurity quenching. Other environmental

effects that are of importance are: solvent effects, heavy

atom effects, pH effects, hydrogen bonding effects, and the

effects of other solutes. This last effect can be further

delineated as the prefilter (inner filter) effect and

collisional energy transfer.

In the most analytically useful form, ie., assuming no

prefilter, no postfilter, no self-absorption, monochromatic

excitation, the relationship between fluorescence and con-

centration is


I = YFIEX(1-e-2.303bc) (5)


where I is the total isotropic fluorescence intensity inte-

grated over all fluorescence wavelengths (relative intensity

units), IEX is the intensity of the excitation radiation

(relative intensity units), e is the molar absorptivity
-1 -1
coefficient (L mol m ), b is the sample thickness (m),

and c is the concentration (mol L-1). Other than concentra-

tion, there are three major factors that affect the fluores-

cence intensity: the quantum efficiency, excitation intensity,

and absorptivity. If any of these is increased, then the

fluorescence increases.

In very dilute solutions, equation (5) reduces to


I = 2.303YFIEX bc (6)


Equation (6) indicates that, in dilute solutions, fluorescence








intensity is linear with concentration. At higher concen-

trations, the fluorescence becomes non-linear, mainly from

absorption increasing with concentration. It has been

estimated that up to 5% of the exciting radiation can be

absorbed by the sample and a linear response obtained (16).



Experimental Approach


To improve the signal-to-noise ratio over that obtained

with conventional fluorimetry, a pulsed source-temporal

measurement approach was used. The reason for this choice

was indicated by previous works reporting the use of temporal

measurements to improve analytical determinations (4,18-26),

particularly in phosphorimetry.

The pulsed excitation source chosen was a transverse

excitation atmospheric (TEA) nitrogen laser. This source

has a high peak power and a relatively short duration pulse,

as compared to conventional sources. The main drawback of

this laser was that the excitation wavelength was fixed at

337 nm, whereas conventional sources, gas filled flashlamps,

offered a greater excitation wavelength selection. As a

development from the N2 laser system used, and to be used

with the system as an excitation source, an N2 laser pumped

dye laser producing subnanosecond pulses was developed.

This dye laser provided a very inexpensive, simple means of

(btaininl! excitation wavelength selection with even shorter

pulse durations than that provided by the N2 laser.








Two separate systems were used for detection and

measurement of the temporal fluorescence signals. The first

system consisted of a photomultiplier tube (PM) detector

and a boxcar signal average, or sampling oscilloscope, for

signal measurement. The second system was a streak canera-

SIT vidicon combination. The PM-boxcar system showed great-

er sensitivity, and hence, lower LOD's (27), but suffered

from poor temporal resolution and jitter. The streak camera

system offered superb temporal resolution, speed, and ease

of use, but suffered from poorer sensitivity and higher

cost. The streak camera system was used to observe heavy

atom effects on fluorescence lifetimes (4,14-17,28-31) and

an attempt was made to demonstrate the utility of the instru-

ment for the time resolution of a mixture of two spectrally

overlapping compounds.

Previous workers (32) have demonstrated solvent effects

of fluorescence lifetimes, though the analytical use of this

type of information is yet to be implemented. Others (18-20,

33-36) have used time resolution as an aid in analysis or

to deconvolute two spectrally overlapping compounds.



Fluorescence Temporal Meastu iements


At the present. "state of the art," there are several

techniques for the measurement of fluorescence lifetimes. A

brief list is presented here, and later in this work. For

a more detailed description, refer to Ware (14). The tech-

niques used can be divided into two basic groups, pulsed

methods and phase shift methods.








The basis of the pulsed methods is to generate a short,

intense, optical excitation pulse and observe the intensity

(signal) of the fluorescence decay with time as described

in equation (1). The typical excitation source used is a

flashlamp giving optical pulses of approximately 1 ns dura-

tion (14,16). However, a tailing effect is noted for flash-

lamps that is not seen for N2 lasers(37). Recently, mode

locked lasers have been used as excitation sources for

fluorescence with pulse widths from 1 to 200 picoseconds

(38-43).



PM-Boxcar


The most popular, and easiest to use, measurement tech-

nique is the sampling oscilloscope used in conjunction with

a fast rise time photomultiplier tube as the detector.

Suitable PM tubes are commercially available (44-46); how-

ever, numerous designs are available to modify less suitable

PMs (47-49). Unfortunately, the shorter rise time PM

tubes have a lower gain and a significantly increased cost

(up to %$10,000). Sampling oscilloscopes are available with

rise times as short as 25 ps. The use of this combination

provides a small portion of the total decay curve with each

pulse of the excitation source. To obtain an entire decay

curve, the source is pulsed many times. To improve signal-

to-noise ratio, many pulses are averaged for each portion of

the decay curve. In this averaging mode, the sampling

oscilloscope is known by another name, the boxcar signal








average. Like the sampling oscilloscope, boxcar signal

averagers are commercially available.



Transient Digitizer


Also available are transient digitizers, or high speed

signal averagers. These operate on the same basic principle

as the sampling oscilloscope, except that they sample several

different portions of the decay curve per pulse and retain

each sampling separately. Hence, an entire decay curve can

be approximated with a single excitation pulse, and through

averaging, improve the signal-to-noise of the entire decay

signal on each pulse. The drawback is that the sampling

interval is not as short as that obtained with the sampling

oscilloscope, the bandwidth not as high, and the cost is

increased considerably.

It should be added that recently a 1 GHz real time

oscilloscope has been introduced commercially. Hence, for

many fluorescent compounds, an entire decay curve can be

obtained real time. The improvement of electronics will

continually improve the simplest of temporal measurement

techniques and keep it in constant competition with more

sophisticated techniques.



Streak Camora


Streak camera techniques are an extension of the tran-

sient digitizer methods, except the time resolution is








greatly improved. The principle is relatively simple and

very similar to a normal oscilloscope in operation. Under

uroDer timing conditions, the fluorescence is focused on a

semitransparent photocathode. Ejected electrons are accel-

erated and pass between two deflecting electrodes. The

relative potential between these electrodes is rapidly

sept with time. The electrons ejected at different points

in time (corresponding to the decaying fluorescence signal)

are deflected further and further. The electrons then strike

a phosphor screen and produce a spatially displayed time

scan, the fluorescence intensity decay. In most cases, the

electrons are amplified just prior to the phosphor screen,

increasing the sensitivity of the instrument. A photograph

can be taken of this intensity distribution and a densito-

meter used to obtain lifetimes. In the more sophisticated

instruments, a SIT vidicon camera is used and this infor-

mation passed on to a computer for processing.

The advantages of this technique are that with a single

pulse from the excitation source, a lifetime can be approx-

imated, and the time resolution can be as low as 2 ps. The

maximum time range is 100 ns, with a temporal resolution of

better than 1% of the time range. The disadvantages of this

system are lower sensitivity than the sampling oscilloscope -

low cost PM method and higher cost. Also, the temporal

resolution of a given time range is a difficulty when apply-

ing the streak camera to a time resolution separation of a

mixture of compounds. Improvements are forthcoming, e.g., a

three dimensional temporal scan will be discussed later.









Time-Corr elated Single-Photon Counting


Currently, the most popular technique is time-correlated

single-photon counting (TCSP) (14,32,33). In this method,

fluorescence intensity is reduced to a level such that, at

most, only a single photon is detected per pulse. The

detection electronics then measure the delay between the

excitation pulse and the detected photon corresponding to

that pulse. This process is repeated for many excitation

pulses, with the number of times that a specific delay is

obtained stored in a multichannel analyzer (MCA) preceded

by a time-to-amplitude converter (TAC). Each incremental

channel of the MCA corresponds to an incremental time delay.

After a sufficient number of pulses to insure a good S/N,

a histogram showing the number of counts versus delay time

is produced. The number of counts in any given time interval

is directly proportional to the probability of emission at

that time delay. Hence, the histogram corresponds to the

fluorescence decay curve.

The important features of this technique are that it

is a counting technique, and follows digital statistics, and

that it relies on a time difference measurement. This last

feature means that. this technique depends on the stability

and reprodiuc ii i t y of the detect ion lcOct.ron i cs, and not on

the analog time response or the ability to reproduce a time-

dependent wave form. The temporal resolution of this method

depends on the dispersion in the detector and the accuracy








of the deconvolution process, (50,51). Timing uncertainties

can be as low as 30 ps (50), and the time range applicable

is from 100 ps to 10 lis. One of the problems with this tech-

nique is the relatively long data acquisition times; however,

improvements are being made. The improvements include high-

er repetition rate excitation sources, especially mode lock-

ed lasers, and better data acquisition systems to operate

at the higher repetition rates. Also, this is strictly a

lifetime measurement technique; its quantitative and temporal

sensitivity is no better than the sampling oscilloscope

techniques.



Optical Gating


Optical gating methods, such as the optical Kerr cell,

are not widely used for fluorescence lifetime measurements,

especially since the methods discussed previously are avail-

able. The Kerr cell technique employs a very high power

laser to send a pulse through the Kerr cell. The Kerr cell

is optically transparent only when the laser pulse is present.

Hence, the optical gate can be approximately a picosecond

wide, if the proper laser is selected. A portion of the

high power laser beam is split off and used to excite the

sample in question. The delay between the two pulses is

varied by optical delays and the fluorescence decay scanned

optically by scanning the optical delay, very much like the

electronic scanning of the sampling oscilloscope. The








advantage of this is the time resolution of about 1 ps. The

disadvantages are poor S/N, because of amplitude instability

of the laser, and a background or scatter also due to the

laser, and slow data acquisition due to the low repetition

rate of the laser.



Cross-Correlation


An interesting new pulse method has recently been

published, implementing a cross-correlation detection system

(52,53). The system consists of a mode-locked laser as

excitation source and a microwave double-balanced mixer and

optical delay as the cross-correlator for a reference laser

signal and the fluorescence decay signal. Fluorescence life-

times as short as 80 ps were measured with + 10 ps uncer-

tainty. The results obtained, corresponded to those obtain-

ed via the sampling oscilloscope techniques.



Phase Shift


Phase shift methods are all based on the relationship

that a fluorescent sample excited by a sinusoidally modu-

lated source produce fluorescence signals modulated at the

same frequency but with a phase shift relative to the source.

This phase shift can be related to the lifetime by


TF = !tanO (7)


for an exponential decay. In equation (7), w is the








angular modulation frequency (w = 2nf) and 0 is the phase

shift. The instrumentation needed is some means to modulate

the source and a phase sensitive detector,

Light source modulators that have been used are Kerr

cells, Pockel cells, ultrasonic modulators, and RF discharges

(14). Various phase sensitive detectors have been developed,

but generally radio frequency techniques are employed. For

example, an instrument has been designed employing a phase-

shift null point approach. An optical delay line is used

to null the phase while the IF circuitry of an AM radio is

used for null point detection (54,55).

Modulation frequencies are limited to approximately

20 MHz, allowing measurement of fluorescence lifetimes from

100 ps to 10 us. The main disadvantage is that, in the

past, only one modulation frequency has been used, corre-

sponding to a single point determination on the fluorescence

decay curve. Hence, the reliability is questionable. How-

ever, improvements are being made (56,57), such as using

more than one modulation frequency for the excitation source.



Power Spectra-CW Laser


A relatively new approach to fluorescence lifetime

determinations has been introduced by Ramsey et al. (52,58).

This approach uses the beat noise of a CW laser as a

multifrequency modulated source to obtain the power spectrum

of this signal. Hence, it is a frequency determination type

of experiment to get temporal information, similar to the





18


phase shift methods. The disadvantages are low sensitivity,

limited lifetime measurement range of from 200 ps to 10 ns,

and a large DC background due to the laser, caused by

scatter.















CHAPTER II
NOISE POWER SPECTRA OF THE ICP



Introduction


Some of the factors which set limits on precision,

accuracy, and limits of detection in chemical analysis are

directly relatable to fluctuations recorded in the measure-

ment of the signal. These fluctuations are termed noise (59)

and can arise from any point in the total analytical proce-

dure. The fluctuations appear to be random about the mean

value of the signal. The smaller the fluctuations (the lower

the noise), the higher the reproducibility of the measure-

ment (the better the precision), and the smaller the signal

that can be determined (the lower the limit of detection).

This also enables a better estimate of the "true" value of

the quantity of the analyte corresponding to the measured

signal. Hence, determining the noise sources in an analyti-

cal procedure, with the hope of minimizing their contribution

to the signal is one of the best approaches for improving

the analytical procedure (13). In the present work, we are

concerned with determining the noise components associated

with a popular analytical method, namely the optical system,

detection device, and electronic signal processing system of

an inductively coupled argon plasma (ICP).

19








Power Spectrum


To characterize the noise associated with the system

chosen, the spectral noise power spectrum (called power spec-

trum) was determined. The power spectrum indicates the

intensity (power) of each frequency component, over a given

frequency interval, in the original waveform. It is present-

ed as a noise power per unit bandwidth as a function of fre-

quency (W Hz-1). It can also be presented as the root-mean-

square (RMS) of the noise current per square root bandwidth

(A Hz ), as a function of frequency. Since the noise spec-

trum displays noise components at their respective frequen-

cies, it can be used in identifying noise sources and to pre-

dict their origins. It has been used in the past to find

regions of the noise spectrum containing low amplitude noise

components into which the information of interest can be

transferred, e.g., by modulation (7,60-62), to improve the

S/N of the measurement.



Fast Fourier Transform (FFT)


There are both analog and digital methods available for

obtaining power spectra. The analog methods used in the past

have proved to be tedious and time consuming (6-8), whereas,

if a digital computer is available, the digital methods can

be implemented rather easily (as attested to in a recent

article in a microcomputing journal (63)). There are, also,

numerous dedicated instruments available to perform the








computation necessary. These digital computation techniques

exploit the Fourier transform properties of the power spec-

tral density function (10). The power spectrum is the

Fourier transform of the autocorrelation function, and this

was the digital approach used until the advent of the "Fast

Fourier Transform" algorithm in 1965 (9). Since that time,

the "FFT" has been the method of choice (12) and was the

method used here, as implemented by a digital computer.

In using the FFT to determine power spectra, several

choices and criteria must be determined. The choices are

frequency range to be covered, bandwidth or resolution to

be used, and number of data points to be sampled per power

spectrum determined. Any two of these parameters specifies

the remaining choice. The relation is a simple one


Af f)(N) (8)
MAX 2

where Af is the frequency range, fMAX is the maximum fre-

quency (all measurements always start at 0, or "DC"), 6f is

the frequency resolution and N the number of data points used.

A criterion that must be met to prevent aliasing is the

Nyquist limit. This limit specifies that the sampling rate

must be at least twice the rate of the highest frequency of

interest'. If this criterion is not met, aliasing occurs and

information at a higher frequency than the Nyquist limit is

"folded over" back onto the power spectrum. In an effort to

minimize high frequency information from folding back into

a certain frequency range of interest, a low pass electronic








filter is used to attenuate all signals above the highest

frequency allowed by the Nyquist limit for a given data sam-

pling. It is also strongly suggested that the sampling rate

be chosen such that the highest frequency of interest passed

by the filter be near the 3 dB point of the filter, and that

both of these be significantly below the Nyquist limit. How-

ever, if the sampling frequency is too far below the Nyquist

limit then resolution is sacrificed.

At the other end of the frequency spectrum, a high pass

filter is suggested to attenuate the very low frequency com-

ponents. This filter attenuates the signals that appear in

the first few resolution elements of the power spectra. The

first resolution element, or zeroth harmonic, contains all

the power of frequencies from the lowest frequency resolu-

tion element to "DC" (10). In some cases, this component

could be the dominating feature of the spectrum. If the

algorithm used for the FFT requires a scaling procedure, a

large DC component could be of extreme importance. The

scaling procedure is needed because only a certain number of

binary bits are used by the computer in the computation of

the FFT, usually one computer word or less. To keep the com-

put er from overflowing its limited word size during the cal-

culation for the FFT, the algorithm constantly checks for

overflow errors. If an overflow occurs, usually after an

addition or subtraction, the data are scaled by dividing the

four (or shifting right twice) and the process during which

the overflow occurred, repeated. (It- should be noted that








this is only one way to overcome the overflow problem.) In

this case, small data values rapidly lose significance.

Therefore, it is important to attempt to limit the "DC" (or

first resolution element) amplitude by the use of an elec-

tronic or digital filter.

It is also possible to do various types of digital pro-

cessing, or smoothing, either before or after the FFT. Prob-

ably, the most important and commonly used digital processing

is the selection of a "data window" (10,64-66). In sampling

a given waveform, the data are collected for a specified amount

of time, giving a representative data set of the waveform.

Unmodified, the data were obtained through a rectangular data

window. In the FFT, the data window is convoluted with the

data, resulting in broadening of the frequency bands and

side lobes appearing on either side of a given high intensity

frequency component. In an effort to minimize the side lobes,

especially where important information is located at a fre-

quency next to a very intense frequency component, the

retangular data window is mathematically modified by modi-

fying the data, particularly the data at the beginning and

end of the data set. The net effect is to reduce the side

lobes relative to the main frequency component. Also, the

amplitude of the main frequency component is reduced and the

frequency band is broadened, as compared to that obtained

with the rectangular data window. (For an excellent dis-

cussion of the above, see references (10) and (64,66).)







Noise Components


In previous works (6-8,12,13), the observed noise was

broken down into three major groups: (i) white noise, (ii)

low frequency (or excess) noise (usually referred to as 1/f

noise), and (iii) proportional noise (or whistle noise).

White noise is a noise that has a constant amplitude at all

frequencies and, as such, forms a baseline for measurements

of other noise components above this level. White noise is

due to completely random variations in the analytical system.

Low frequency noises are noise components that occur at the

low frequency end of the power spectrum. In some cases, they

follow a 1/f function; hence 1/f-noise has become the more

popular terminology. The principal cause for such noise is

the slow drift in the instrumental components. This can come

about because of temperature changes, changing line voltage,

aging of the instrumental components, etc. Proportional

noises occur at given frequencies throughout the range of the

spectrum. Their cause can be varied, although when noted in

flames (12,13), the possible cause was speculated as an

"organ pipe" effect; hence the term whistle noise.



Experimental


In Table I, the instrumentation used in this section

is given. In Figure (2), a block diagram of the experimental

setup is shown. In Appendix A, a flow chart and a descrip-

tion of the FORTRAN program used to manipulate the data is















Compu


A/D,


X-Y p
analo

Oscil


AC am


High
power

Elect


Pho t o


Monoc


ICP


Signa


Table I. Experimental Equipment Used for Noise Power
Spectra of ICP.


Item Model Source

iter PDP 11/34 Digital Equipment Corp.
Maynard, MA

D/A LPS-11 Digital Equipment Corp.
Maynard, MA

letter, Plotmatic Bolt, Bernard, and Newma
9g Cambridge, MA

loscope 122 AR Hewlett-Packard
Palo Alto, CA

iplifier 103 A Keithley Instruments, In
Cleveland, OH

voltage 244 Keithley Instruments, In
supply Cleveland, OH

rometer 601 Keithley Instruments, In
Cleveland, OH

multiplier R 928 Hamamatsu Corp.
Middlesex, NJ

.hromator EU-700 Heath-Mc Pherson
Benton Harbor, MI

1500 Plasma-Therm, Inc.
Kresson, NJ

.1 generator 180 Wavetek


San Diego, CA


(DEC)





n, Inc.






c.


c.


c.


II





















Figure 2. Block Diagram of Experimental Setup for ICP Noise
Power Spectra

(A) ICP
(B) Monochromator
(C) Capacitor (1 yF, mylar)
(D) PM
(E) High Voltage Power Supply
(F) AC Amplifier
(G) A/D (LPS-11)
(H) D/A (LPS-11)
(I) Minicomputer
(J) Terminal
(K) Analog Plotter
(L) Oscilloscope




27













B
D ----




g








F -- }- 6

Hr
1- 1-:^-

^^ -- /








given. In Appendix 13 a flow chart and a description of the

fast A/D sampling routine, written specifically for the high

frequency range but used for all ranges, is given.

All chemicals used were reagent grade. The chemicals

and manufacturers are: zinc metal (Mallinckrodt Chemical

Co., St. Louis, MO), barium carbonate (J.T. Baker Chemical

Co., Phillipsburg, NJ), lithium nitrate (Fisher Chemical Co.,

Fairlawn, NJ), and yttrium oxide (American Potash and Chem-

ical Corp., Chicago, IL). Doubly dionized water was obtained

from a Barnstead Nanopure water system (Barnstcad, Boston,

MA).

For each noise power spectrum determined, three differ-

ent sampling rates were used, 100 IIz, 1 kHz, and 25 kHz.

From the Nyquist limit, the frequency r:i'r,,.-; covered were,

respectively, 50 Hz, 500 Kz, and 12.5 kHz. For each sam-

pling, 2048 data points were collected giving a frequency

spectrum of 1024 unique points. Hence, the frequency reso-

lution for the three sampling rates were, respectively,

0.049 IHz, 0.49 Hz, and 12.2 Hz. For the resulting noise

power spectra, 100 separate noise spectra were determined

and averaged in order to improve the signal-to-noise ratio

of t he spect ra.

The photornultiplier was operated at -1 kV. The Keithley

AC amplifier was set to 0.1 Hz for the 3 d13 point of the

high pass filter in all cases, and the low pass filter was

changed as the sampling rate was changed. The roll-off for

these filters was -6 dB per octave. The 3 dB point of the








low pass filter was set to 30 Hz for the 100 Hz sampling rate,

300 Hz for the 1 klz sampling rate, and 10 kHz for the 25 kHz

sampling rate. A constant 1000 Q load resistor was used

across the output of the photomultiplier and this voltage sig-

nal amplified by the AC amplifier. The output of the ampli-

fier was AC coupled to the DEC LSP-11 A/D converters through

a mylar, 1 pF capacitor. This was necessary to reduce the DC

offset and drift inherent in the amplifier. The signal gain

was adjusted by the gain control on the amplifier in conjunc-

tion with changing the input channel of the A/D. The A/D

allowed a choice of either + 1 V full scale input or + 5 V

input, depending on the channel chosen. In all cases, the

signal was monitored, prior to sampling, on an oscilloscope

to determine which channel to use, such that the great major-

ity of the signal waveform fell within the input voltage

limits of the appropriate channel. The channel and gain in-

formation was used by the FORTRAN program to scale the data.

The FORTRAN program returned the power spectrum values

as RMS current per square root of the frequency bandpass. A

calibration curve was obtained using a Wavetek signal gener-

ator and the power spectra were found to be linear in ampli-

tude as a function of frequency range and sampling rate to a

precision of + 5 %.

The ICP was operated under normal manufacturer's condi-

tions. A set of parameters was established as a base and

various changes were made to note their effect on the noise

power spectrum. The base conditions were P = 1.5 kW RF








power, Q = 15 L/min coolant argon gas flow through the torch,

N = 1.5 L/min nebulizer flow rate, an observation height (H)

of 32 mm above the coil, and 40 ppm Zn sample solution being

aspirated (A = 213.9 nm, slit height (SH) = 1 mm, slit width

(SW) = 100 um). The torch design was similar to that pro-

posed by Genna et al. (67). The DC signal level was mea-

sured with a Keithley electrometer. Other values of the P,

Q,N parameters were P = 1.0 kW, Q = 20 L/min, and N =

1.0 L/min. Water and 2000 ppm Zn were aspirated also. An-

other experiment was performed with the same base conditions,

but for 100 ppm Ba (X = 455.4 nm, ion line, SH = 1 mm, SW =

50 jm) and 20 ppm Li (X = 670.7 nm, SII = 1 mm, SW = 50 pm)

aspirated; the corresponding noise power spectra were

determined.

A third experiment involved yttrium, 500 ppm, being

aspirated (A = 597.2 nm, SH = 1 mm, SW = 100 pm) and the

observation height being varied from 15 mm above the coil to

30 mm, to 45 mm; the noise power spectra were determined at

each height.

The torch was changed to the type commercially available

and base conditions employed to determine the power spectra

from this torch design.

Finally in an attempt to localize the noise sources

observed in the power spectra, the nebuli zer drain tube was

shortened and the base conditions repeated. Also, graphite

blocks were placed in the torch cavity such that approxi-

mately 50. of the cavity volume was occupied, and the noise

power spectra determined using the base conditions.








Results and Discussions


In Table II, the results obtained for the noise power

spectra under varying conditions are listed. The table lists

white noise amplitude, the frequency and amplitude where the

low noise components reach 10% of their peak values (omitting

the zeroth harmonic), and any proportional noises, by peak

amplitude, frequency, and bandwidth. In Figures 3-12, exam-

ples of the noise power spectra obtained are given.

The noise power spectra of only the 500 Hz range is pre-

sented in all cases, except where the observation height was

varied and yttrium was measured. In that case, the 12.5 kHz

range is presented because the low frequency noise appears to

extend to significantly higher frequencies at the largest

observation height. For all other experimental conditions,

the high frequency range offered little information, except

to indicate that no other proportional noises existed at high-

er frequencies. In some cases, the 50 Hz range was not of

use because aliasing had occurred resulting in fold over of

information from frequencies higher than 50 Hz. This is

particularly evident when very strong components are present

at higher frequencies.

Figure 13 is an example of the results of this fold over

effect. The 50 Hz spectrum of barium corresponds to the

500 Hz spectrum in Figure 10. The broad peak at %10 Hz in

Figure 13 is due to the peak at %210 Hz in Figure 10. Hence,

the low frequency information is completely hidden by this

high frequency component.
























002


Ct: L l S C JN N
in)~ O --> c c -


I I I I I C


a-
;=c:
-YC

V]



CC-



3=

=Ls;
Jir
C
51



Z-

-C
Oi





33.
If
-e

c


C
5C~'
=S

C'u
L
Lr





C

O

`rr 1
'i i

n
c-~
a=E


oc
x
ILeF:
3
3~3~1

cr=

L







YO
V;'L)


Z-


E

r:
E


03
00
000


1 1 0


E'J "3


I i I 0


N N













1 I I -
0 '- vT v.
p)3 I- N






M CJ -















0CCO00ON 00


C -- 0 1

- :-V -~ -













- 0 ~ c C 0

00 0 000000N C


0 0 0 0 0 0 0 0

m = 1-0 1) 0


C

-E 7


00










zC'


- ~ I.. 0 0 0 -


0 0











0 0















. 0 N C rN

000 CO O


- :11 M :. 0







33






N C











I I I















S I 0 0
00 ? 0 0 I I I I I


01- -
--


I 0


to o w '
-- I -












c cm e- n











03 0 03 =i o'.-
S o -o o c
-0 N


0 C0o0 000
o 0n o 0 n o
L7 U3n ,n n .-


N











I












0













00 0000 -


.' -e .


0 NO N oNO
0 in -,O ,


0

E 0
C N
in



*i ~ r >* ~=





c,3 cl c I
INN
- S -^ _C"c-

- C- 'I i' -


C- e .- 0 0 C-
Nl N N N N -l N


FI I
.. :
ZI.0I jo
0e



SC)


0
1 ~"




s- X =^C

N N N n '-

















.7 -




I In













I


.I PN 0

C -C C


c

C E





N C
= .7*


-1 -^ *T
N N- cN


r


rJ
cl-


r:
"
~
L
9_-

f=L
N3
j=-r


~3


- N N f

OC C -







C I




0 C C -



> p




: -2


I' i i










C- N I


34








1 LfC I
N -








0I 0



















0- -



















' -~








C


N


N J
. -


N


















2: -



:1 -













-- I

- '

N 1


I I






35



































































E E





J[ y









































E


4-1
U


Cl) Q)








cti








.r-
44a














0
0
n

0
-0
N


-0>
--0 -O

U
Z
o W

cru


o L
-0



-0
I i ID

LO LO n
SN ru



S/ ZH6f7"O]/ OT K V)

H IIMONVE/N-Nt nD S k






































C
E


LO
1-1



Zrc
rC II E

d C



O ,C II

II r- Q
Ic 0 ,U

'-c O.C
.0 P -H A-
0 II
a) N cd HC
O C rH ;H
0 D u ,-l -H
11 0 (1) 1 r-
MC) *OTO,


Od
0






00 c
S0
0 r













0
0
O


0
-D


I
0
N


0

-0

U



o LL
-C



0-


n 0 0O



P( /; [ZH6-'0O]/60T V)


tI-i I"CN /v../.N j n 3
;__. / "1





































E



C) 0
0



S -*H 0 -

ON 0 O
0 C ( 0
E CO

wQ. 0 0U)
00 0) 3 M





tO
1.o
E a















0

-0
I


-0
N
I


"0
O
o w
o D
cu 0







wo
Son o LL
-0







( /C [ZH6t'O]/60T; a V)

SHIOIMOaNVa9/IN33JIn3 SIWd
















































2
C

C-)


C/)

c)

0


C)TH

CD
"-1 f


LO












0'-
4I




4-)C


cX
Q) Q
E




cd x
ulQ
















O
_0



0
0-
N

0
-0>
i W
ow
-o







o Li-


-0



0
1N 1
00



0 000


(2/ [CZH61'O]/ 60T V)


/-T HLOIMONV/8/IN3dl0n 3


SWI,












































E

U

4-)



E
c







Do.


Ci2

4)

r- L
CQ)





0
(D 00
2r b1 fl




U) 4, CH
U2~0














0
-


0





-0
U
z
-o0

lu0

It

-0



-0



0 0

(2,; EZH6trO3/ 6 OT V)


/- HLOQIMONVS/ IN3

SWCI






































Ca)



E F

C
> d
O

W 0
00 1




OE 0
P I0 0
0
oo o


00 c0
Z CI 0





0b















0


0

0
-0
N


O
0



0
m U


0 W





-0



-0



cu ru W4
./o










( I [ZH6tr'O/6 OT V)
. HOI O N IM O /IN3c\ JIn3 S W











































2

4-J
0








Q')
II
0


UI)

Cd


0( C) -H
*Hl U -H



0 0





U2 PLAz











0
0


0
0
N



o
O
I






0
o u




I I






i v rcu o0


( Z/ [ZH6t7'O3/ t0T a V)
i3 / Q






< /;ZHl0/^TV












































U))C
Ci2




-H 0




0 L00
O Q)LOO

ha II ?I
2 bn II1












0
0



-0
T0
N
I
Jo
0
-o 0



oz
-0 w



-0



0


S0 0 0
o d d o

2/E g ZH6"P03/9OT V)

23/, HIOIMONVa/IN3 I fn3 SW !










































E


4-)
U
a)







E
a)





00C



Cd



0 ,
0) .





0 00
Q) L- LO

Sh II
Sd -H i
In PT l< U













0I
0-





N

0
-0
SU


ow


0 Z



-0



m cu 0o
0 CD z0 0 0


( g/[ZH603/6OT a V)

3/,T HJLCIMONV/LN3)nO3 SWk









































-'0
E-:
EU





ic~i
o

oa~

'HE




0 0
zRu


Cd
bfl


w) a)




0 UQ

oa),
.H a
10 4-j
[0

o 4-1
tU 4-

d- a)-
a) PL4








































I 1


-0
o
- o
^r


u u
U

0L


w
o W




o LL
-0

-o

-0


( %/E [ZH6tr'0]/ 60; N tV)


SWcI


2/N;H L0MiCNVa /L N3Bln3
















0
I)


0
N
I
0
Li

z
03




I ---- I ---- -- I ---
CU Cy
w
O




0


O) 0 0 0


( / CZH6trO3]/6 OT V)
/ HIOIMONVS/IN3ynfO SWNi








The reason for this fold over is that the low pass

filter used only attenuated at -6 dB/octave, and for large

amplitude signals, this attenuation was not enough. In the

500 Hz range, fold over was probably also occurring, but

since there were no proportional noises at higher frequen-

cies, i.e., only white noise, this effect was not observed

and not a problem. Whenever information was acquired from

the 50 Hz range, the 500 Hz range was checked to determine

that there were no large amplitude proportional noises to

interfere with this low frequency range.



Spectra


In Figure 3, the noise spectrum of the photomultiplier

dark current is shown. This spectrum shows a low frequency

component and a contribution from 60 Hz and its harmonics,

due to line frequency pickup. It is appropriate to note here

that the 60 Hz components were observed in every spectrum and

have been recorded previously (13). The change of the rela-

tive amplitude of the even and odd harmonics from spectrum

to spectrum is also noted. Although the reason for this is

unclear, it is believed to be related with the normal opera-

tions in the laboratory, i.e., room lights being turned on

and off at different times which change the contributions to

the different harmonics. However, since these components

were constant in frequency, of narrow bandwidth, and deter-

minant, whatever their contribution, they were viewed more







as a frequency calibration than as a hindrance. Also,

because of their large intensity, they too were folded over

into the 50 Hz spectrum. Again, they were not viewed as

entirely detrimental, but as a constant reminder of the

aliasing that was occurring.

Another point of interest are the spikes that occur in

certain spectra. Even though the spectra presented are the

average of 100 separate spectra in each case, miscellaneous

spikes, at apparently random frequencies, are observable

with good amplitude. The cause of these spikes is not known

in all cases; however, in one instance, a major contribution

was localized and eliminated as being caused by a poor cable

connection from the AC amplifier to the A/D converter.

Hieftje and Bystroff (13) also noted spikes in their work

which they attributed to sudden aeration of the flame. In

Figure 4, a water background of the ICP for the base condi-

tions is given. The only observable noise components, other

than white noise, are the low frequency components, which are

similar in shape to the dark current components, but they are

generally more intense, as is the white noise.

For the base conditions chosen with Zn (Figure 5) as

compared to the spectrum obtained with water, white noise

increases, the low frequency noise increases and its 10%

point (i.e., the frequency at which the low frequency compo-

nent reached 10% of its maximum amplitude) moves to higher

frequency, and, of course, the DC signal increases. Also, a

proportional noise is observed at %210 Hz, with a less








intense band at %120 Hz. Decreasing the RF power decreases

the amplitude of the noises as well as the DC signal (Figure

6), as does increasing the coolant gas flow rate (Figure 7),

though not as much. Increasing the concentration of the Zn

solution being aspirated increased all the noise components

as well as considerably increasing the DC signal (Figure 8).

Decreasing the nebulizer flow rate also increased the noise

and DC amplitudes, though not to the same extent (Figure 9).

For barium,at 455.4 nm,the proportional noises, the white

noise and the DC signal increased compared to the Zn case

while the low frequency noise components became indistin-

guishable relative to the proportional noise (Figure 10).

For lithium, at 670.7 nm, the low frequency noises over the

proportional noise components increased and their 10% points

occurred at higher frequencies than for the other elements

(Figure 11).

Interestingly, very few noise components, other than

white noise, were distinguishable with changing observation

height at 590.7 nm. However, increasing the observation

height, with yttrium being aspirated, extended the low

frequency components to a 10". point of approximately 1000 Hz.

Decreasing the observation height, with yttrium being aspi-

rated, accentuated the proportional noises.

Changing torches (Figure 12) produced a small shift in

the frequency of the proportional noises and greatly in-

creased their amplitude, although the DC signal remained

about the same. Changes in drain tube length and cavity

volume had little effect on any of the observed noises.








Conclusions


In Table III, the effects of changing the operating

parameters of the ICP relative to the base conditions chosen

are given. From the change in concentration of the zinc

solutions used, it is apparent that the noise components in-

crease in amplitude with increasing DC signal. This is also

the case when the RF power was lowered and less DC signal

observed. Decreasing the coolant flow rate also increased

the signal, and hence, the noise. The same was true for

the nebulizer flow rate; decreasing the flow rate increased

the signal. This effect can be related to zinc and its

residence time in the plasma. Aspirating barium and changing

the wavelength to 455.4 nm (near the peak of the argon back-

ground of the ICP) seemed to enhance the proportional noises.

However, the barium signal was very large. The water back-

ground spectrum at this wavelength, compared to that obtained

for water at the zinc wavelength, showed a smaller amplitude

low frequency component that extended to slightly higher

frequencies. This indicates that the low frequency noise was

the major component affected by the increase in wavelength.

This conclusion was born out when lithium was used and the

wavelength changed to 670.7 nm. Here, the major feature is

the low frequency noise, even when the water spectrum was

determined.






4 62


0 ft I I + + I I I
.*--H





Q)
7 0




) ,--H + O O + 0 + I
.,0


4-




C OZ
S *HCI2






S O-H + I 0 I I + +
00 O




0 0 u a)







z na)





0




0 0
E *Hca












t. + I +







.. ) u d U
4 -H





0 -', 4 )
4- )U
00 0



*- 0 Co 0 U H i
S0 H








a)-' D > 0
0 a> -A












r.0 4 0 m0 0
cO M 4 0 w 0

0 *0i 0 0 11 0 0 0i H
U 9 0 0 0 a aP
0 0














SC C 0 C Q a00
Ci pi bii + + c I1 + 0 a)
+ a) a-) a) a a)



>a



















2;








For yttrium, some interesting effects are noted. The

oxide was monitored with the hope of observing changes in the

noise power spectrum as the observation height above the coil

was changed. The major point noted was, as the height in-

creased, the power in the low frequency components increased.

Their 10% point was %1000 Hz. The increasing height corre-

sponded to an increase in signal-to-background ratio. When

water was aspirated at each height, these low frequency com-

ponents were not observed. Hence, they were related to the

yttrium oxide. This could be an indication of the turbulence

in these regions of the torch and the formation of the oxide

itself.

When the torch was changed, a slight change in frequency

of the proportional noise was noted, but the magnitude of

this change was within the variations observed for the torch

used in the base conditions. However, the amplitude of the

proportional noise was increased substantially which is most

likely related to the design change. This "new" torch is the

type commercially available; the torch used for the bas

conditions was of an improved design. The design improvement

consisted of constricting the inlet side of the coolant gas

tubes giving a greater pressure drop at this point. The

results are a more turbulent flow of argon and a reduction in

the amount of gas used (67). From the noise power spectra,

it appears that the commercial type torch had more noise

power in the proportional noises, and the constricted inlet

torch had more noise power in the low frequency components.








Thus, the commercial type torch probably has more turbulence

at the higher frequencies than does the constricted inlet

torch.

An attempt to localize the noise sources by changing

the drain tube length and the cavity volume gave no defini-

tive results. The idea of an organ pipe effect (12,13)

causing the %210 Hz proportional noise indicated a pipe

length of approximately 75 cm for a tube open at both ends,

and 38 cm for a tube open at one end only, while the torch

itself is about 15 cm long. Hence, the attempt at changing

the larger volumes associated with the torch.

The large proportional noises shown in the power spectra

are readily observable on an oscilloscope. At one point in

the trial and error stages of this experiment, a longer torch

was tried and a noise component of considerably higher fre-

quency was observed with the oscilloscope. However, that

torch was destroyed in one of the many laboratory errors that

occur and a replacement was not obtained prior to this

writing. It is suggested that the torch design, particularly

the length, is responsible for, at least, the frequency of

the proportional noises, and contributes greatly to the

amplitude of the noise.

The main purpose for these experiments with the ICP was

to report noise components at various frequencies in the ICP.

In relation to signal-to-noise improvement, the majority of

the ICP systems used today are equipped with DC detection

electronics employing an integrator or a very low, low pass








filter. Hence, the higher frequency noises will be filtered

out. However, at low signal levels, as indicated by the

water background spectra, the dominant noise is the low

frequency components, the largest of which are passed by the

integrator or low pass filter. The extent of their contri-

bution to the "DC" signal measured by the integrator or low

pass filter is not clear from the present work and indicates

an area of future interest, to characterize the extremely low

frequency noises (3,13).

Another region of interest is that of the proportional

noise. This "noise" could probably be used as a signal if

an appropriate AC amplifier were used. Because of the broad-

ness of the peak, indicating drift or jitter, a lock-in

amplifier would not be the system of choice. However, a

broad band AC amplifier would certainly be a starting point.















CHAPTER III
FLUORESCENCE TEMPORAL MEASUREMENTS



Introduction


For conventional molecular fluorescence spectrometry,

the excitation scatter and luminescence from the blank have

been shown to be the major factors controlling the limits of

detection. This has also been shown to be the case for laser

excited molecular fluorescence (27,68). However, because of

their high intensity, an increase in sensitivity (signal/

concentration) is often noted with laser excitation. From

theoretical calculations (3,69), as well as experimental re-

sults, pulsed source excitation with gated detection, using

time resolution, could be used to improve S/N in certain

instances. Pulsed lasers are very well suited for these

measurements because of their high intensities and short

pulse widths.

The operating conditions for which the application of

time resolution could improve the S/N ar e important. First,

the emission from the analyte and the interfering component,

scatter, blank luminescence, etc. should spectrally overlap;

otherwise the problem could be spectrally resolved. (This

assumes that the interfering component is excited simulta-

neously with the analyte.) Second, the analyte should have

66








a significantly longer lifetime than the interfering compo-

nent, indicating the use of time resolution. If this second

point is not true, then the use of time resolution becomes

complicated and of little quantitative value. More informa-

tion is needed about the entire temporal signal and a decon-

volution is used to separate the analyte signal from the

interfering component. This increase in experimental com-

plexity usually results in a poorer S/N than that obtained

with conventional fluorimetry. Other problems result from

energy transfer, photodecomposition, etc., and these problems

have to be considered on an individual basis to determine

if time resolved fluorimetry would offer a more accurate

result than conventional fluorimetry.



Time Resolution


The basis of time resolution is shown in Figure 14. The

excitation pulse is assumed to be a square pulse, for sim-

plicity. The dashed line is the response of a fluorophor

that has a relatively short lifetime. The dotted line repre-

sents the response of a fluorophor with a lifetime ten times

longer. The remaining dot-dashed line is what would be

observed if the two were combined. The principle of time

resolution is to allow the short lifetime component to "die

out," such that it is contributing very little to the overall

signal, and the long lifetime component to dominate. The long

lifetime component can then be quantitated with little






















Cd

O



0
a)



0






E
a)
0E

a)

0,
0










0
O0
rH










O

0
O




a)c














0-H
aEC
o





Or-i
*H























E- M


hD

0



E






H+-1 Q

0) 0
z G r. I(1
0- 0 ) >
E EH 0
+.. 0 4 O

-4- )0
P'M O


w 0



S0 C
Pt, O

OOad

00



00 c0



0) 0 c





69











I
-*/
*


'I






../ tw
H
.'/ /- /




**




N










ALISN3.LNI

3ON3OS3-OfldO- 3AIIV73W








interference from the short lifetime component. This has

recently been applied analytically to fluorescence (36).

If information is desired about the short lifetime com-

ponent, more information can be collected from the entire

decay of the 2 component mixtures. Once the lifetime of the

long lived component is calculated, the intensity (signal)

contribution to the overall signal can be estimated and decon-

voluted from the total signal, giving the contribution due

strictly to the short lifetime component. It is theoretically

possible to extend this to more than two contributing species

(70-79). However, at the present, the practical limit

appears to be the deconvolution of only two spectrally over-

lapping components (under ideal conditions 3 components

could be measured).

If lifetime information is desired, and the relative

lifetimes of the two components are significantly different,

it is possible to plot the natural logarithm of the signal

and obtain a graph that contains two linear regions. The

linear region of the graph that has the least negative slope

corresponds to the lifetime of the long lived component; the

short lived component can then be calculated by deconvolution/

subtraction of the long lived component from the total signal

as a funct ion of time. An example of this is shown in Figure

15. This is a plot of the natural logarithm of the summation

of the two exponential decays shown in Figure 14. The two

linear re'.ions are clearly visible. The slope of the lower

linear r.:-ion corresponds to the lifetime of the component































co U
Ui)
* 0




Srl 4-

Qi) 0
z2 () H
oe 5o



4E E
02 4--)
o d




E-O
0
*HHc
4-3 4)


O 8 rH


-Pl-
o 7




0

0 4-) -1 b
E~
4a-,





u /2 4-3-i
0 c-r-4 zi c
Or-4












br-r
P r4 r-



4-)


cj








Fr-



































0
S
S
S
S
S
S
S


AIJISN3LNI


3ON3oS3dOcfld


3AIIV~13.4









with the longer decay time. The upper linear region corre-

sponds to a summation of lifetimes of the two components.

As mentioned previously, there are inherent problems

with time resolution. First, the closer the two lifetimes

are to each other and/or the greater the signal level of the

short lived component compared to the long lived component,

the more difficult it is to evaluate the long lifetime com-

ponent without interference from the shorter lived species.

Hence, the information about the short lifetime species is

even harder to obtain. Secondly, the further from the laser

pulse the decay signal is measured to determine the long

lived component, the smaller the signal being measured.

Therefore the S/N decreases, even though the relative signal

of the long lived component to the short lived component is

increasing. Thirdly, what is being observed is fluorescence

intensity levels. A large intensity from a short lifetime

component makes it extremely difficult to determine a low

intensity, long lifetime component. In other words, the

relative intensity contributions of the two species can be

significantly different only if the relative lifetimes are

also significantly different. The reverse case--low inten-

sity, short lifetime component, large intensity, long lived

component-is not as susceptible to this problem and time

resolution should work better.

An interesting problem that develops from this third

point is that, if the quantum efficiency is small for one

component and large for the other, then to obtain approximately








equal fluorescence intensities, the concentration of the com-

ponent with the poorer quantum efficiency must be relatively

large. Higher concentrations could lead to quenching and

other effects not immediately apparent from the time resolved

experiment.

Other problems with time resolution are more subtle,

but the experimenter is quickly made aware of some of these.

The more important of these subtle problems are generally

true of all temporal measurements, especially in the fluores-

cence decay time range. These are time jitter, drift, pulse

to pulse variations, and signal reproducibility. Also, more

information is available from the temporal signal than just

its application to time resolution. This information is more

in the form of a qualitative nature, but with this informa-

tion it could be possible to improve the quantitative results.

Changing the environmental conditions of the sample changes

the observed fluorescence lifetime (14-16,32,36,69,80-82).

Other effects, such as quenching, photodecomposition, and

excited state reactions, could be monitored temporally, pro-

viding more information about the sample.



External Heavy Atom


The use of the external heavy atom effect for fluori-

metric and phosphorimetric analysis has been demonstrated

in the past (4,28-31,83-87). The external heavy atom effect

is presumably due to an increase in singlet-triplet








probability from an increase in spin-orbit coupling. The

resulting effects reported are, decreased fluorescence inten-

sity, decrease in fluorescence quantum efficiency, decrease

in phosphorescence observed lifetime, and an increase in the

observed phosphorescence intensity.

The present work shows an attendant decrease in fluores-

cence lifetimes with increasing concentration of silver

nitrate or potassium iodide. The external heavy atom effect

on the phosphorescence of two compounds used here has been

reported previously (4,31). The external heavy atom effect

could be used as a selective means of fluorimetric analysis,

as suggested by Zander (28-30). This selectivity could be

enhanced with the use of temporal information.

There has been an increase in the interest in the tempo-

ral information obtainable via luminescence, as attested to

by the reviews by O'Donnell and Solie in 1976 and 1978

(88,89). This interest has grown with the advent of short

pulse width lasers and improved detection of events on the

same time scale as fluorescence lifetimes or faster. Time

resolution as applied to Raman (34,35) or atomic fluores-

cence (18,19) has been shown to be of considerable use. For

phosphorescent species, temporal measurements have been of

analytical use for some time (reference 4 and references

therein).

With the commercial availability of mode locked lasers

and streak cameras, even more interest will be generated in

obtaining temporal information (90-92). The analytical use

of this information is to be decided in the next decade.








Per Pulse Fluorescence


By being able to determine an entire fluorescence decay

on a per pulse basis, as with the use of a streak camera, the

temporal information mentioned above is made available rapid-

ly. With greater temporal resolution, nulse to pulse varia-

tions are visible and signal reproducability is observable

with each pulse. Also, jitter in the time between when the

excitation pulse is fired and when the signal is measured

can be easily determined, as can the instrumental drift. For

most temporal measurement, the only way around such varia-

tions is by signal averaging, which tends to blur out the

temporal information. If the information is made available

on a per shot basis, decisions can be made as to the signi-

ficance of a particular signal and corrections made to

improve the overall signal or averaged signal.

Hence, obtaining the entire fluorescence decay curve

per shot of the excitation source, could provide important

information about the sample and allow instrumental correc-

tions to give a better measure of the signal.



TEA N2 Laser


The instrumentation used to obtain the fluorescence tem-

poral information can be quite complex. For the systems used

in this work (Figures 16 and 17), the excitation source






























































p
p 0
C) r-4 4--)
U) -A m~
0C) E
'-U 0

C\l Q)

0c 0

li 0
p c-H I


C)

0






U) 0H

0 f

Q)0
CO





C) ;-4


0- (D~
o -C


ko)


Q)



U)


U)





4-)










0 )
c


PC

EU






01


~~~C) ~ r-F i'S






78

































L=
(--r















































O
4rl

C),
0

bi

f-.

S0

M,--l O

U)r 0
cd E E



WH) C 0
~rcd
O~ CO
Hcid'O


a)
0

0 U 0
* H C)


;Q E-0 r-
000
H *H >



EI-
- >H
E I [-
(GQM Q >
HET< -iM


Q,


-)
Q)







-,- ,
E-0





rO


dok
C) 0 0
F -P


O) 0


E-1- 0
ri ^



0 4>


*^CUQMMOai-'-i-) ;7
11- 1- 11 -1 1-,1e -11-1-
p 4 PL,0 M t4 -











employed is common to both. This source was a transverse

excitation atmospheric (TEA) nitrogen laser. It had a pulse

width (measured at full width half maximum, FWHM) of 1.0 ns,

a maximum repetition rate of 100 !z, and an energy per pulse

of %20 IJ. The TEA N2 laser is distinctive from the low

pressure N2 laser in that it produces two lines spectrally

at 337 nm (93-95). The low pressure N2 laser produces numer-

ous spectral features (4,95,96). Hence, the TEA N2 laser is

spectrally cleaner offering less problems from scatter due

to unwanted spectral features. The TEA N2 laser also has a

shorter pulse width than the low pressure N2 laser, and it

can be made even shorter by using air as the fill gas,

instead of nitrogen; however, there is an attendant drop in

output power. This means that time resolution can be applied

to more systems than if a longer pulse width excitation

source was used.

The TEA N2 laser pulse width is still considerably

longer than that produced by mode locked lasers. Also, the

excitation wavelength is set at 337 nm. In an effort to

reduce the pulse width and increase the variability of exci-

tation wavelength selection, a simple dye laser was con-

structed which produced pulses as short as 225 ps, FWIIM. A

more detailed description of this laser will be given in

Chapter IV.

The sample cell used was a common fluorescence cuvette,

chosen because of the manufacturers claim of being "fluores-

cence free." In practice, a smaller volume sample cell could









have been used (97). The laser beam was focused into the

cell to provide a higher power density and allowing only a

small portion of the sample to emit fluorescence. The diam-

eter of the beam entering the sample was approximately 1 mm,

giving an illuminated sample volume of about 8 uL. The

emission of the sample was focused onto the slit of a mono-

chromator with a simple lens giving a one-to-one image. The

maximum slit width was 2 mm, therefore the observed sample

volume was about 2 pL. In the optical arrangement used, the

slit height of the monochromator was only partially filled

with the available sample fluorescence, %1 mm. However, in

the case of the streak camera, the largest entrance slit to

the camera available was 100 pm high and 1 mm wide, so only

a small portion of the total fluorescence available was ob-

served. For the PM-sampling oscilloscope/boxcar case, an

improvement in the signal level observed could be made if

the monochromator slit height and excitation pulse into the

sample cell were aligned in parallel.



PH-Boxcar


In Figure 16, the P'l-sampling oscilloscope/boxcar system

employed is shown. In reality, the majority of the work

performed for this section was done using a boxcar signal

average in place of the sampling oscilloscope because it

offered a direct method to signal averaging. However, the

sampling oscilloscope showed less jitter in the overall








signal measured, for either system, when it was triggered

internally (part of the signal used to trigger the sampling

head).

The photomultiplier used was wired for fast rise time

response (Figure 18). Although other wiring designs are

available, as mentioned previously, this design was chosen

as a compromise between rise time, versatility, ease of PM

replacement, and ease of construction to achieve the desired

speeds. It is noted that proper shielding and grounding of

the PM was necessary to minimize Radio Frequency Interference

(RFI) and ringing of the signal. As such, the PM had a rise

time of about 2 ns. Its response to the 1 ns TEA N2 laser

pulse was %4 ns, FWHM.

The gate widths, or sampling times, used were approxi-

mately 350 ps for the sampling oscilloscope and approximately

75 ps for the boxcar. In either case, the system response

was limited by the response of the PM. In making the tem-

poral measurements with the boxcar, the laser and the boxcar

were both triggered by an external pulse generator. When

the sampling oscilloscope was used, and when the signal was

large enough, the laser was triggered from the pulse gener-

ator and the sampling head was triggered internally. When

the signal level was small, the sampling oscilloscope and

laser were triggered externally by the pulse generator. In

either case, to produce an entire fluorescence decay curve,

the sampling gate was internally swept, in time, across the

signal. When amplitude measurements were made, the delay

































Figure 18. Wiring Diagram of Photomultiplier Dynode Chain

R = 50 kQ
C = 2 nF

All resistance values are in ohms. All resistors
used were carbon film. All capacitance values
are in nanofarads. All capacitors are disc
ceramic.











-HV








time from the trigger pulse was fixed. The only variation

in the gate position, in that case, was due to jitter or

drift.



Streak Camera


In Figure 17, the streak camera system is shown. The

temporal disperser (TD) is the actual streak camera and its

function is to provide an output exactly described by its

name, a spatial dispersion of a temporal event. This output

was three dimensional. First, the spatial image of the light

passed by the entrance slit is imaged across the output

screen in the x direction. This corresponds to the 1 mm slit

width. Second, the temporal scan is displayed in the y

direction, with the progression of time going from top to

bottom of the output screen. Finally, intensity is displayed

as the third axis and this corresponds to the fluorescence

signal that is being streaked.

To monitor this output from the temporal disperser, a

SIT-vidicon camera was used. The vidicon took the three

dimensional optical output, converted it to an electronic

signal, and passed it to the temporal analyzer (TA), a micro-

computer, for processing. The TA was hooked up to a video

monitor which provided a real time display of the "streaks"

and also a means of outputting the intensity versus time

interpretations of this information by the microcomputer.

Although the temporal disperser provided three dimensional








output, the temporal analyzer used here only interpreted

this information in a limited fashion; the spatial image

information was not used. However, the microcomputer added

a great deal of versatility to the system. It provided a

means of setting up the initial operating parameters of how

to handle the information obtained from TD, whether to use

a dark current subtraction from the data, how to trigger the

system, and how to present the information in the various

forms of output (monitor, recorder, teletype, or digitally

pass it to another computer).



Computer Link


Finally, in the mode used most frequently for this work,

the TA was connected to a minicomputer. This connection was

only a one way connection for the transfer of data (with

minimal "handshaking" controls) from the TA to the minicom-

puter. When the system was operated with the minicomputer

receiving the data, one streak was performed per data trans-

fer. This allowed the data to be processed by the minicom-

puter before the next streak occurred. In this way, even

more versatility was obtained, and it was possible to make

corrections for drift and jitter on a per pulse basis.



Triggering


The triggering of the system was important and varied

depending on .just how much of the total configuration was








used and what the operating parameters were. Both the N2

laser and the temporal disperser were triggered by an exter-

nal pulse generator. After the TD was fired, it, in turn,

triggered the temporal analyzer to process the information

from the vidicon camera.

The temporal analyzer required about 0.4 s to interpret

and display the data on the monitor for a single integration

and a persistence of ten (discussed below). If a dark current

subtraction was performed, the time required doubled. If

the streaks are integrated, the data is acquired at %0.1 s

per shot and the analysis time of 0.4 s, or 0.8 s, is added

on. The data transfer through the TA to the minicomputer

is done only on a per shot basis and requires the 0.4 s, or

0.8 s, analysis time per shot, depending on whether or not

a dark current subtraction is performed before the transfer.

With the laser capable of a 100 Hz repetition rate and the

temporal disperser capable of following a 1000 Hz repetition

rate, it is easy to see that the temporal analyzer is the

rate determining step in the total data acquisition process.

When the pulse generator is triggered internally, the

laser and temporal disperser can produce information at a

rate up to 100 Hz. However, the TA can not handle such a

large throughput. (It should be noted that the vidicon

camera provides one complete frame every 17 ms, so it could

not follow this rate either. But up to a repetition rate of

60 Hz, the monitor displays the streaks as the vidicon

collects the information. Hence, the time consuming process









is the interpretation of the temporal information by the

microcomputer.) This is the reason for driving the pulse

generator externally.

With the pulse generator triggered externally, it is

possible to "single shot" the system using a momentary switch

or to have the TA trigger the pulse generator after each data

analysis. The last technique is the most efficient in terms

of analysis time. It also keeps photodecomposition to a

minimum and reduces the aging of the laser. For most of the

work presented here, this is the triggering technique used.

During the initial setup, the system was free run (pulse gen-

erator on internal) or "single shot."

Another modification is to have the laser free running

and "single shot" the temporal disperser or have the TA gate

the temporal disperser such that it only obtains a streak

after the microcomputer was ready. Problems here include an

increase in the possible photodecomposition per analysis,

triggering of the data acquisition is not optimized, and an

overlap of streaks, before analysis, if the triggering rate

of the laser is too high.



Computer Routines


For the analysis of lifetimes and pulse widths of lasers,

averaging was important to improve the S/N, hence the mode

of data transfer to the minicomputer was used. Several types

of data processing routines were tried to overcome the








problems of jitter and drift. The two that were the most

useful were a "step seek" routine and a "peek seek" routine.

The step seek routine was used mainly for the fluorescence

temporal measurements. With the pulse width of the excita-

tion source being 1 ns, or less, the rise time of the pulse

was considerably faster. In most cases, the growth of the

fluorescence signal was faster than the signal decay, because

of the excitation pulse. Hence, the fluorescence signal had

a relatively sharp leading edge. The step routine seeks out

this leading edge as it crosses a specified threshold value,

and adjusts each data set such that they are all added

together from this point.

The peak routine was used mainly for the pulse width

measurements of the lasers tested, in conjunction with a

single shot measurement. This routine scanned the data sets

for their maximum values and added each data set together

such that the points in time where maxima occurred overlap-

ped.



SIT Camera


With the use of the vidicon camera, one has to be care-

ful to work with certain inherent characteristics (98-101).

For the work reported here, the most important characteristic

to be aware of is the persistence of the camera itself. It

has been found that for the best S/N, several frames of video

signal should be summed together after a streak to deplete

completely the charges on the silicon target (102). As far