THESIS

 

Madam

‘ LIBRARY
Worthy

 

This is to certify that the

dissertation entitled
A 7/’ 7/914 7701-! - can meaning y I; if VdLVAG-t

S’f’flflk ATOM/c smug/o»; pé'lr'éc’rot lie/t
(xv/444:7 6-295 CflKOMflTOG-KA/kr

presented by
k6 Ira xé/w/e-‘r/I 77/5:le

has been accepted towards fulfillment
of the requirements for

fl/i 'é- degree in CH!“ “VT/z

W/ZM’

Major professor

Date ////?/§/§/

Msui.n.4mn...m..‘ - r“ m ~ I - ~ 0.12771

 

 

 

MSU

LIBRARIES
n

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

 

 

A THYRATRON-CONTROLLED, HIGH VOLTAGE
SPARK ATOMIC EMISSION DETECTOR FOR

CAPILLARY GAS CHROMATOGRAPHY

By

Keith Kenneth Trischan

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Chemistry

1984

ABSTRACT

A THYRATRON-CONTROLLED, HIGH VOLTAGE
SPARK ATOMIC EMISSION DETECTOR FOR
CAPILLARY GAS CHROMATOGRAPHY

By
Keith Kenneth Trischan

The construction of a high voltage, dc, nanosecond spark source for
atomic emission analysis is described. The source is used for
element-selective detection of capillary gas chromatographic effluents.
In operation, a high power spark discharge is formed in an argon atmosphere
between two thoriated tungsten electrodes by the discharge of a coaxial
capacitor. A ceramic, hydrogen thyratron is used to. trigger the discharge.
This establishes a highly reproducible firing rate. Effluents from a capillary
gas chromatograph are transferred through a heated glass interface to the
spark where atomization and excitation of the eluted compounds occurs.
A microcomputer-controlled data acquisition system allows the acquisition
of time-resolved emission intensities. With this method, intense continuum
radiation, generated when the discharge is formed, is discriminated against
in favor of relevant atomic emissions. Improved signal-to-noise ratios are
obtained by acquisition of computer—averaged intensities for several
consecutive sparks. A menu-driven software package is described which
simplifies the data acquisition process. Characterization studies of the
thyratron-controlled spark source are described. The entire detection system
is applied to universal and element-selective detection of capillary gas

chromatographic effluents.

To my wife, Anna,
parents, and
brother

ii

ACKNOWLEDGEMENTS

Many people deserve more thanks than I could ever give. Most of all,
I wish to thank my wife for all the support, moral and financial, which she
has given me. Without her typing assistance this work would be as hard
to read as my handwriting.

I also thank my parents and brother for their assistance in this endeavor.
They have always impressed me with the idea that nothing is impossible
if you try your hardest and if you persevere.

Dr. Stan Crouch deserves a fair share of the credit for this as well.
He has allowed me to develop along paths that I found interesting and which
otherwise would have remained unexplored while he still provided direction
when needed.

Many thanks to the members of my research group who have always
insured that work was full of unexpected surprises. They have made my
stay here much more enjoyable than it otherwise would have been. I would
like to especially thank Marquerite who has proven that the size of the heart
is definitely not proportional to the size of the person.

Finally, I would like to thank the excellent technicians, machinists,
and specialist in the department who have helped solve problems which I

could not overcome on my own.

iii

TABLE OF CONTENTS

29.92
LIST OF TABLES ........................................................................ vii
LIST OF FIGURES ....................................................................... viii
CHAPTER I - INTRODUCTION ..................................................... 1
CHAPTER I] - ELEMENT—SELECTIVE GAS .................................... 4

CHROMATOGRAPHIC DETECTORS

A. Introduction ................................................................... 4
B. Thermionic Detector ...................................................... 5
C. Flame Photometric Detector ........................................... 6
D. Electron Capture Detector ......... ~ ..................................... 8
E. Atomic Absorption Detector ........................................... 9
F. Atomic Emission Detectors ............................................. 10
1. Microwave Induced Plasma Detector ......................... 10
2. Direct Current Plasma Detector ............................... 12
3. Inductively Coupled Plasma Detector ........................ 13
4. Spark Discharge Detector................................ ......... 13
CHAPTER III - INSTRUMENTATION ............................................. 15
A. Overview of the Spark Atomic Emission System ................ 16
B. Spark Source Design ....................................................... 18
C. Thyratron Controlled Spark Source .................................. 23

iv

PAGE

D. Capillary Gas Chromatograph - Spark Interface ............... ‘24
E. Solution Introduction System .......................................... 27
F. Gas Chromatographic Components.....-. ............................ 32
G. Alternate Spark Imaging Scheme ................. > .................... 35
CHAPTER IV — ELECTRONICS AND OPERATION ....................... A ‘ 37
A. The Intel 8085 Based Microcomputer ............................... 37
B. Minicomputer Facflltles 42
C. Data Acquisition Electronics ................................. 42
1. Device Selection Circuitry ........................... - ............ 42

2. Data Acquisition Timing Circuitry .......... ‘ .................. 45

3. Analog Electronics. .................................................. 53

a. Gated Integrator ................................................ 53

b. Baseline Subtraction Circuitry ............................. 55

c. Programmable Gain Amplifier .............................. 55

d. Analog-to—Digital Converter Circuitry .................. 56

4. Photomultiplier Gate Circuitry ................................. 57

D. Monochromator Controller/Microcomputer Interface. ........ 59
E. Spark Detection Circuitry ......... - ....................................... 65
CHAPTER V - SOFTWARE .......................................................... 67
A. Data Acquisition Software .............................................. 69
B Interrupt Service Routine ................................................ 73
C Buffer Management Routine ............................................ 74
D. Data Storage Routine ...................................................... 77
B Data Disk Organization .................................................... 77
F Minicomputer Programming Language Selection ................ 79
1. CVT - Data Conversion Software ................................ 81

PAGE

CHAPTER VI - SPARK SOURCE CHARACTERIZATION ............... 83
A. Capacitor Charging Characteristics ................................. 85

B. Effect of Gap Voltage on Current Waveform .................... 90

C. Background Emission Studies ........................................... 93

D. Effect of Oxygen on Carbon Emission ............................. 99

E. Excitation Temperature Studies ....................................... 99

F. Effect of Time on Excitation Temperature ....................... 102

G. Effect of Charging Voltage on Excitation Temperature ..... 102

H. Electron Density Measurements ...................................... 108

1. Determination of Ionization Temperature ........................ 110
CHAPTER VII — CHROMATOGRAPHIC EVALUATION ................... 116
A . Introduction .................................................................. l 1 6

B. Evaluation of the GC — Spark Interface ........................... 117

C. Universal GC Detection Capabilties ............................... 121

D. Element-Selective Detection of Lead in Gasoline ............. 127

1. Lead Calibration Curve ........................................... 128

2. Lead Determination in Gasoline ............................... 131
CHAPTER VIII - CONCLUSIONS AND PERSPECTIVES ................. 134
REFERENCES ........................................................................... 1 3 9

vi

LIST OF TABLES

TABLE M
4-1 Monochromator Controller Commands ........................ 63
5-1 Data Acquisition Options ............................................ 71
6-1 Typical Acquisition Parameters ................................... 94
6-2 Argon Line Emission Parameters ................................. 104
7-1 Jet Fuel Gas Chromatographic Conditions ................... 120
7—2 Carbon Standards Gas Chromatographic Conditions ...... 123
7-3 Lead Standards Gas Chromatographic Conditions .......... 129

vii

LIST OF FIGURES

 

FIGURE PAGE
3-1 Block Diagram of the NSS—GC System .......................... 17
3-2 The N88 Housing ........................................................ 19
3—3 The Analytical Chamber, expanded view ....................... 21
3-4 Circuit Diagram of Thyratron-Controlled Spark Source.. 25
3-5 GC - Spark Interface ............ . ...................................... 26
3-6 Ultrasonic Nebulizer ................................................... 29
3-7 Cross-Flow Nebulizer ...... .. .......................................... 30
3-8 Miniaturized Desolvation Chamber ............................... 31
3-9 Capillary GC Injector .................................................. 33
3-10 Alternate Lensing Scheme ........................................... 36
4-1 Block Diagram of Data Acquisition Electronics ............. 38
4-2 Block Diagram of Intel 8085 Microcomputer System ...... 39
4-3 Interface Device-Selection Circuit ............................... 44
4—4 Block Diagram of Advanced Micro Devices AM9513 ....... 46

Timing Controller

4-5 Block Diagram of AM9513 Counter Group ..................... 48
4—6 Timer Circuit. .............................................................. 51
4-7 Analog Electronics ...................................................... 54

viii

FIGURE PAGE

4-8 Photomultiplier Gate Circuit ................................... 58
4-9 Microcomputer/Monochromator Interface .................. 60
4-10 Photodiode Trigger Circuit ....................................... 66
5-1 Data Acquisition Software Organization .................... 70
5-2 Flow Chart for Interrupt Service Routine .................. 75
5-3 Flow Chart for Buffer Management Routine .............. 76
5-4 Flow Chart for Buffer Flushing Routine ..................... 78
6-1 Gap Voltage as a Function of Repetition Rate ............ 87
6-2 Relative Intensity as a Function of

Repetition Rate ...................................................... 88
6-3 Relative Intensity as a Function of Charging Voltage.. 89
6-4 Discharge Current as a Function of Charging Voltage. 91
6-5 Maximum Discharge Current as a Function

of Charging Voltage ................................................ 92
6-6 Background Emission as a Function of

Wavelength (0.3 - 1.0 us delay time) .......................... 95
6-7 Background Emission as a Function of

Wavelength (2.0 - 4.0 us delay time) .......................... 96
6-8 Relative Carbon Emission Intensity as a

Function of Oxygen Flow Rate ................................. 100
6-9 Spectrum of Neutral Argon Emission Lines ............... 103
6-10 Excitation Temperature Slope Plot ........................... 105
6—11 Excitation Temperature as a Function of

Delay Time ............................................................ 106
6-12 Excitation Temperature as a Function of

Charging Voltage ................................................... 107
6—13 Stark Broadened Hg Line Profile .............................. 109

ix

FIGURE PAGE

6-14 Electron Density as a Function of Delay Time .............. 111
6-15 Calcium Emission as a Function of Delay Time ............. 112
6-16 Ionization Temperature as a Function of

Delay Time ................................................................ 114
7—1 Comparison of FID and N88 Detection ........................ 119
7-2 n-alkane Chromatogram ............................................. 1 24
7-3 Carbon Calibration Curve ........................................... 125
7-4 Chromatographic Peak Digitization Errors .................... 126
7-5 Lead Standard Chromatogram ...................................... 130
7-6 Lead Calibration Curve ............................................... 132

CHAPTER I

INTRODUCTION

Chemists are being required to perform analyses of samples in extremely
complex matrices. The main forces behind this demand are the determination
of trace level components which impart either a beneficial or a hazardous
critical characteristic to the sample. Often such a critical component will
contain a functionality or element such as bromine, lead, or other heavy metals
which is contained in only a small fraction of the compounds in the sample.
When this is true, a detection method which responds only to the critical
feature will greatly simplify the identification and quantitation of that
compound.

Even when the critical component contains an identifiable, unique species,
the matrix effects associated with the sample may interfere with the
determination, producing inaccurate results. Both gas and liquid
chromatography are quite powerful for separating complex mixtures. Because
of the separation, matrix effects are eliminated except for the case of
coelution or incomplete separation. The degree of separation is so high, that
for complex common samples, in excess of one hundred individual peaks can
be readily obtained (1). Identification of each peak is tedious and generally

unnecessary since only a few critical components are of interest. Most

detectors in general use have little or no species-specificity; hence,
identification of the critical compounds is by comparison of the retention
times of standard mixtures to those of the sample. This requires the existence
of such a standard. In addition, the chromatographic system must be stable
over time, otherwise, repeated injections of the standard would be required.
This could quickly become the rate limiting step in the analysis. When a
selective detector is used, variations in the retention time are not critical
since only the species of interest and a few similar compounds are detected;
thus, a simpler chromatogram is produced.

Element-selective detectors are one group of selective detectors.
Detectors based on atomic absorption or atomic emission spectroscopy offer
high detectability and linearity for several metals. Detection of both metals
and nonmetals can be performed with plasma sources. Since atomic carbon
emission can be observed in plasmas, these detectors can be used in a universal
detection mode. By tuning the observation wavelength, signals for all
constituent atoms can be obtained. This allows detection of all species
containing the critical element. In addition, when element-selective detection
is used for gas chromatography, the empirical formulae of the individual
peaks can be determined. Thus, the powerful separation capabilities of
chromatography coupled with element-selective detection provides the analyst
with a valuable technique.

Plasma sources are, in general, quite expensive to purchase and operate
with initial costs an order of magnitude greater than that of the
chromatographic system. This is prohibitive and has limited the application
of plasma detectors for chromatography. A relatively inexpensive alternate
atomic emission source is the nanosecond spark source.

This work describes the construction and characterization of an

electronically-triggered, high-voltage spark discharge. A description of the
hardware and software used to monitor the spark and acquire emission intensity
data is presented. The spark is also applied to the element-selective detection

of capillary gas chromatographic effluents.

CHAPTER II

ELEMENT-SELECTIVE GAS CHROMATOGRAPHIC DETECTORS

A. Introduction

Gas chromatography has become a powerful analytical tool. Since
extremely complex mixtures can be separated, identification of all detected
components is extremely difficult. Choice of a detector which provides
chemical information about the detected species simplifies the identification
process. In addition, detection of a specific chemical property or functionality
simplifies the chromatogram obtained since only compounds containing the
specific characteristic are detected. A final advantage to selective detection
methods is a reduction in the chromatographic quality required for a separation,
since resolution of only selected components is necessary.

Element-selective chromatographic detection is one area which has
received much attention recently. Several elements such as S, N, P, heavy
metals, and the halogens are often associated with an environmental hazard;
hence, techniques for the analysis of these elements in complex samples are
required.

Element-selective detection can be achieved by atomic spectroscopic
techniques or by one of several molecular methods which are drastically

enhanced by the presence of one or more specific elements in the eluted

4

component. Atomic absorption and atomic emission techniques are popular
atomic spectroscopic methods applied to GO detection. This chapter briefly
reviews those element-selective detection methods which have been applied
to gas chromatography; a more thorough review of this tapic can be found
in works of Lantz (2), Koepling (3) and Calkin (4).

Several universal detectors commonly in use demonstrate enhanced
response for compounds which contain certain elements. Enhancements

between 10 and 1000 are obtained for various techniques.

B. Thermionic Detector

The earliest thermionic detector was developed by Cremer (5,6). The
thermionic detector is similar to the flame ionization detector except that
an alkali salt crystal is mounted in the flame. In operation, the
chromatographic effluent is burned in an HZ-air diffusion flame. The
conductivity of the flame is measured; hence, changes in the flame environ-
ment cause an increase or decrease in the conductivity of the flame. Sevcik
(7) has proposed that thermal ionization of the alkali salt establishes a constant
current in the flame. Current fluctuations are caused by electron capture
processes, the formation of thermally stable compounds, or formation of
species with high ionization potentials. Response enhancement is observed
for species containing phosphorus (8,9), nitrogen (10,11,12) sulfur (l3), and
the halogens (14). In early TID designs, the collector electrode was at a
negative potential with respect to the alkali salt. As a result, loss of alkali
material was observed and the detector sensitivity decreased with use. Kolb
and Bischoff (15) reversed the polarity of the collector. In this configuration,
no deterioration in performance was observed with time. To account for

this, a charge transfer mechanism based on the formation of radicals was

developed. In this mechanism, radicals such as CN-, PO-, P023 or X- leach
electrons from the alkali to form stable species. In the original design the
cation was detected; hence, alkali erosion occurred. However, with polarity
reversal, anion detection occurs and the cation is electrically attracted back
to the salt. Thus, no erosion occurs. Recent advances indicate improved
stability is attained when independent heating of the alkali material is
performed. Patterson _e_t _a_l_. has successfully designed a ceramic core with
a surface coating of alkali (16). The surface coating determines the sensitivity
of the detector such that a low degree of coverage and operation in an HZ/air
environment demonstrates a preferential response to P and N containing
compounds, while a higher coverage enhances the response to compounds
containing electronegative functional groups such as nitro groups. In general,
the thermionic detector has become well accepted and is not undergoing
rapid technological advances as a gas chromatographic detector. Bombick
gt _a_l. (17) have suggested that a thermionic emission process can be used
as a tunable, selective chemical ionization media for the detection of gas

chromatographic effluents by mass spectrometry.

C. Flame Photometric Detectors

Flame photometric detectors (FPD) monitor molecular emissions produced
by burning the GC effluent in an Hg/Oz/air flame. Molecular emissions are
monitored through an optical filter with a photomultiplier. The FPD is
commonly used to detect phosphorus and sulfur; however, selective detection
of boron (18), nitrogen (18), chlorine (19), iodine (20), and iron (21,22) has
been demonstrated. The FPD can also be operated in a universal detection
mode by operation in an N2 atmosphere (23).

One major drawback of the FPD is a nonlinear dependence of the response

on the concentration of the detected species. The most common case is for
the detection of S. A rough dependence on the square of the S concentration
is observed since an 82 emission band is monitored (20). When the GC effluent
is combusted in the presence of ozone, an $02 emission band can be observed;
however, the signal obtained is not linear with S concentration (25). Nelson,
e_t_ 11:, have obtained a fluorine-induced chemiluminescence response which
is linear with the S concentration (26). Detection limits of between 24 and
260 pg/s were found for several compounds. The fluorine—induced
chemiluminescence detector had no response to $02, COS, H28, and CSZ;
however, the otherwise good detection limits make this an excellent tool
with which to monitor other atmospheric sulfur compounds.

A similar approach has been used for the detection of iodine in iodinated
hydrocarbons (20). An alternate analysis for the halogens requires observa-
tion of chemiluminescence radiation generated when polyhalogenated
hydrocarbons are passed through sodium vapor (27). A number of problems
were noted with the containment of the sodium vapor.

A second drawback of the FPD is that the flame occasionally extinguishes
when the solvent front elutes. This has been eliminated with a two flame
system (28,29). In these designs, the lower flame is used to atomize the
effluent and the upper excites the residual species. Improvements in the
detection limits for P, N, and S with the inclusion of a quartz sleeve around
the upper burner were referenced by Karasek g at. (29).

As with the thermionic detector, the FPD has reached a level of maturity
and acceptance as a reliable, well-established chromatographic detector

and is used in routine analysis and applications.

D. Electron Capture Detector

The electron capture detector (ECD) is one of the most widely used,
extremely sensitive detectors for routine analysis. With optimum operating
conditions, a detection limit of 90 ag was found for
N,N-dipentafluorobenzoylpentafluoranaline (30). In operation, a beta particle
emitter, typically 63Ni or tritium, ionizes nitrogen molecules in the support
gas and low energy electrons are formed. Application of a potential across
the detector causes electron migration to the anode and the generation of
a current. When an electronegative species enters the detector, some of
the electrons are captured and the current decreases; thus, the detector
response is a function of the mass flow of the analyte and its electronegativity.
This gives a selective response to the halogen- and oxygen-containing species.
Early ECD designs established a constant potential across the detection cell.
Poor stability and linear range of two orders of magnitude or less were reported
(31). Improvement in the stability and linear range of six orders of magnitude
for the ECD have been achieved by pulsing the applied potential (32). Two
mechanisms for the response of the pulsed ECD have been proposed by Aue
gt 31. based on his experiments (33). The first mechanism is the accepted
neutralization process (34) where one analyte molecule reacts with a maximum
of one electron. The second mechanism is a space-charge model where a
molecule sequentially combines with several electrons (35). This mechanism

is summarized with the reactions:
k1
RX + e" ——-> R + X“

X‘ + P+ ‘9 PX + neutral

9

k
Px+e-—2—9X-+P

where R is an alkyl group, X is the electronegative species, and P” is a positive
ion, most likely hydrogen. Reaction A is the conventional electron capture
reaction. In the pulsed operating mode, several additional electrons can be
neutralized by undergoing reaction C. Hence, when reaction C is kinetically

and energetically favorable, an enhanced response is observed for RX.

E. Atomic Absorpbtion Detectors

Several benefits exist for selective detection by atomic spectroscopic
techniques. These detectors offer high sensitivity and selectivity. In addition,
selectivity for several elements is easily obtained by adjustment of the observed
wavelength. Also, atomic spectroscopy has been applied to many types of
analyses; hence, the application of these methods to chromatographic analysis
does not require the development of an entirely new technique, but rather
the extension of an existing one.

Atomic absorption is one spectroscopic technique which has been
successfully adapted to chromatographic detection of metal-containing species
(36-40,49). Connection of the chromatographic column to both flame and
electrothermal AA detectors is trivial. Two approaches have been attempted
with equal success. Wolf (39) simply drilled a hole into the side of a burner
and connected the column directly to the burner. Other workers (36,37) have
used a heated transfer tube as the interface. To minimize absorption of
chromatographic effluents, a glass lined burner head (41) and quartz transfer
tubes (42) have been developed. Detection limits at the nanogram level for
Pb (39,48), Cr (39,42), Hg (40), and Se, Sn, and At (36,47) have been reported.

A severe limitation to the application of AA to chromatographic detection

10

is the inability to determine more than one metal at a time. In addition,
the need for a hollow cathode lamp for each element further limits this
approach. To circumvent this limitation, Harnly §_t_ a1. (43) have developed
a continuum emission, wavelength—modulated AA method that can be applied
to simultaneous, multielement determinations. Recently, a comparison of
this method with wavelength—modulated atomic emission detection (44) presents

detection limits of 0.09 ng/ml and 0.02 ng/ml, respectively for Cr in urine.

F. Atomic Emission Detectors

Atomic emission detectors are extremely powerful and are rapidly
becoming the most popular element-selective detector available. When
operated with a suitable light detection device, the atomic emission detector
should respond to any element which the source can excite. This includes
nonmetallic as well as metallic elements. Hence, atomic emission detectors
can be operated in a universal mode or an element-selective mode. Several
atomic emission sources have been developed and have been shown to be
extremely valuable methods for chromatographic detection.

Use of polychromators foremission detection (45) allows several elements
to be simultaneously monitored. This further increases the power of the
technique. Atomic spectrometric detectors, particularly those using microwave
induced plasma and inductively coupled plasma emission spectroscopy, have

been thoroughly reviewed by Risby (46) and Qing-Yu (51).

1. Microwave Induced Plasma

Microwave induced plasma (MIP) sources have been among the most
popular atomic emission detectors due to the relatively low cost, ease of
operation, and high quality of response. Early MIP designs required the use

of helium plasma support gas at reduced pressure (50,52,53). This made the

11

MIP difficult to use with the gas chromatograph; however, detection limits
in the subnanogram range and a linear dynamic range of 3 to 4 orders of
magnitude were reported for several elements (54,55). An atmospheric pressure
argon or helium MIP was made possible with a major redesign of the resonance
cavity by Beenaker (56-58). Using the atmospheric pressure MIP, the
chromatographic effluent can be eluted directly into the resonance cavity
(59,60). Carnahan and coworkers (61) have presented a review on
chromatographic applications of plasma emission techniques and have included
details of GC interfaces which have been used. Several workers have
demonstrated the usefulness of sequential element-selective detection
(57,62-66); however, this technique requires one injection for each element
of interest. Hence, sequential detection is slow and subject to errors caused
by unstable chromatographic conditions and random variations in the injection
technique. Three approaches to simultaneous multielement detection of
chromatographic effluents are possible. The rapid scanning detector of
Mulligan 53 al. (62,63) requires adequate chromatographic separation to allow
multielement detection. The use of a polychromator to detect MIP emissions
has received much attention (67-71). Some of these systems have the capability
of detecting 20 to 30 elements simultaneously (51,62). Alternately, an array
detector can be used for detection of a wavelength region containing several
emission lines. One drawback of the MIP, as with other spectroscopic
techniques, is the presence of molecular emissions which occur when carbon
and nitrogen compounds elute. Hence, the background level changes as the
peak is eluted and accurate emission intensities are difficult to obtain. This
interference is particularly acute for detection of Br and Cl. Eckhoff and
coworkers (71) have developed a wavelength modulated polychromator as

the GC-MIP detector. Packed column detection limits of 86 pg/s and 76

12

pg/s were obtained for Br and C1 with this system. These compare quite
well with values of 67 pg/s and 86 pg/s obtained by Estes gt gt. (72) with a
capillary GC with monochromatic detection of the MIP emissions. A maximum
acquisition rate of 3 background corrected intensities per second was obtained
for each of the four available channels. While this sample rate is adequate
for packed column studies, a faster acquisition rate is prefered for capillary
chromatography. Holland gt gl_. (73) suggest that a minimum of 2 and preferably
5 measurements per second be performed to ensure accurate representation
of the peak. Finally, stoichiometric ratios of several elements (C, N, F, H,
Cl, Br, S, and I) of the chromatographic effluents were determined with
polychromatic detection of the MIP emissions (58,74,7 7—7 9).

Two disadvantages exist for the MIP. First, the plasma is sensitive to
the support gas flow rate and, second, the elution of the solvent front
occasionally extinguishes the plasma (80). Both problems arise from the small
size of the plasma. Although these problems exist, the MIP has become popular

because of its high power and relatively low cost.

2. Direct Current Plasma

The direct current plasma (DCP) source has a higher energy than the
MIP; hence, the DCP is less sensitive to the carrier gas flow rate and solvent
front elution. The higher energy also makes this a more sensitive atomic
emission source. The DCP has been applied as a GC detector with high
sensitivity for several elements (64,65,7 5). The development of three-electrode
designs has improved the sensitivity of the source in two ways. Increased
source stability has lowered random fluctuations in the location of the plasma
and improved signal-to—noise ratios. Also, this design has shifted the location
of the analyte emission maximum away from the intense continuum region

of the plasma, further improving the sensitivity of the technique (76). Although

13

the DCP is a more stable and energetic plasma than the MIP, the relatively

high cost of the source has limited its use as a chromatographic detector.

3. Inductively Coupled Plasma

Inductively coupled plasmas (ICP) are one of the most powerful plasma
sources used for emission spectroscopy. Because of this, the ICP demonstrates
superior detection limits and linearity of response. In addition, the high energy
available in these sources makes them ideal for analysis of liquid samples
where solvent vaporization and atomization processes consume large amounts
of energy.

These plasma sources have been applied to many simultaneous,
multielement detection schemes for both solution and liquid chromatographic
analyses (81,82). Reports of the application of the ICP to gas chromatographic
detection are sparse. Brown has successfully determined oxygen (83) and
nitrogen (84) containing effluents. Lead as tetraethyl lead in fuel and air
has been determined by Sommer (85) with the ICP. Detection limits at the
ng Pb/s level were obtained. Detection limits of 0.05 ug with a linear range
of greater than 5 x 103 have been found for C1 and Br while monitoring
non—resonant emissions (86). The application of ICP detectors to gas
chromatography has been limited due to the high cost of the source relative

to the cost of the rest of the chromatographic system.

4. Spark Discharge Detector

Single (2) and double (3,4) gap spark discharges have been used as atomic
emission sources for the detection of GC effluents. Many elements, both
metals and nonmetals, have been successfully detected. Detection limits
between 10'8 g/s and 10‘12 g/s were achieved. Response curve linearity
over two to three orders of magnitude were typical. Lantz (2) has demonstrated

the capacity to determine C/H/N/O ratios for several compounds in an effluent

14

stream.

Band broadening caused by excessive detector dead volume has been
eliminated with the incorporation of a low-volume flow cell. Although the
double—gap spark source was allegedly free from affects of variations in the
analytical gap environment, this was not the case. Firing rates and emission
intensities were dependent on the gap environment. This degraded the
analytical usefulness and made the spark source difficult to use. Although
these problems existed, the spark source provided several of the desirable
characteristics of more expensive atomic emission sources. Hence, the spark
source remained an economical alternate plasma source which could excite
many elements in an effluent stream. This work describes the construction
and characterization of a high power, thyratron-triggered spark source. In
addition, a microcomputer-controlled, data acquisition system for the spark
source is described. Finally, the spark source is applied to the

element-selective detection of capillary gas chromatographic effluents.

CHAPTER III

INSTRUMENTATION

The nanosecond spark source has evolved through several generations
as a gas chromatographic element-selective detector (2-4). Spark system
designs incorporating three major triggering mechanisms have been attempted.
Originally, a single gap system was developed. The electrical breakdown
potential and, hence, the energy dissipated in the spark were dependent on
the spark environment. Since the observed signal intensity is dependent on
the dissipated energy, any variation in the gap environment caused a variation
in the signal intensity. The problem was particularly acute during the elution
of the solvent front in gas chromatography, often completely extinguishing
the spark discharge and causing difficulty in immediate reignition of the
plasma. This compromised the value of the single gap system as a gas
chromatographic detector.

In an attempt to alleviate the environmental dependency, a double gap
system was developed. A high breakdown potential spark gap was used to
control the ignition of the lower breakdown potential analytical gap. By
maintaining a pure argon atmosphere in the control gap, a reproducible firing
rate was to be achieved. This, in turn, was to establish a constant total

discharge energy. Although the dependence of the detector on the

15

16

interelectrode enviornment was diminished in this design, a constant firing
rate and the associated energetic consistency were unobtainable. In spite
of this difficulty, the spark discharge system proved to be a useful
element-selective gas chromatographic detector.

Most recently, a single gap design incorporating a thyratron tube to control
the spark firing rate has been developed. Control over the firing rate is
achieved by placing the thyratron in series with the spark gap. Since the
thyratron is electronically triggered, a firing rate independent of the gap
environment is attained. Preliminary .work by Calkin (4) with the
thyraton-triggered spark gap indicated that the firing rate was completely
independent of the spark environment; however, the ability to hold the
electrodes at a voltage greater than the breakdown potential caused
uncontrollable arcing to the spark chamber walls.

Additional mechanical and electrical problems arose due to the
characteristics and requirements of the thyratron-controlled spark source.
This necessitated a redesign of the spark chamber.

This chapter presents an overview of the spark system and a description
of the gas chromatographic components. The spark chamber design and

spark/GO interface are discussed.

A. Overview of the Spark Atomic Emission System .

The spark system is illustrated in Figure 3-1. In operation, gaseous sample
is continuously introduced into the spark chamber perpendicular to the
interelectrode axis. Spark ignition occurs shortly after the thyratron is
triggered into conduction. Once this occurs, the sample is atomized and
excited, and radiation is emitted. This radiation consists of an intense

continuum emission and characteristic atomic emission from the sample and

17

swim ooAmz 2: do Seaman v.85

 

mutation—z.

nu\PF.Ema

 

moon
Amhz.

 

mupaazooa

Tm Rama

 

~30th

 

 

I‘ll

 

fill:

 

 

muzzomhounfi

zo..:m.:oo<
(bio

 

 

 

._<z_2mmh
hmo

 

 

 

do

 

 

muhua..

noxhouaw

 

 

 

zomk<¢>1b

 

 

 

 

zomh<m>1h

 

 

 

Kwamaoo

 

 

 

H mmAJOmeOO.

 

mop<zo¢zoozoa

 

 

 

uo<suo>
:o:.

 

 

 

 

00

 

 

 

18

the plasma support gas atoms. The monochromator (GOA-McPherson EU—700)
isolates the desired wavelength band. Discrimination against the relatively
short-lived continuum emission is accomplished by gating the photomultiplier
tube in the spectrometer. Consistent timing relative to the spark ignition
is accomplished with a synchronization pulse obtained from the thyratron
driver circuitry. After a short delay relative to the spark ignition, the
photomultiplier is rapidly gated from a low to a high gain state. The PMT
signal is passed into a gated integrator and programmable gain amplifier;
it is then digitized by a twelve bit analog—to-digital converter. The digitized
signal is then read by the microprocessor and temporarily stored. To improve
the signal—to-noise ratio, emission intensities from multiple sparks are averaged
by the microprocessor, and stored as a single data point. This data point
can be permanently saved on flexible disk media. Alternatively, the averaged
signal can be displayed on a strip chart recorder after digital-to—analog
conversion; this permits real time observation of the acquired data. Subsequent
to the data acquisition run, the data are transferred through a serial line
to an LSI—11/23 minicomputer where higher level data manipulation is
performed. A CRT terminal is used for operator interaction with the data
acquisition software running on the microprocessor and for communicating
with the minicomputer. Hard copy plots and printouts are obtained from

a dot matrix printer attached to the minicomputer.

B. Spark Source Design

A diagram of the spark chamber is shown in Figure 3-2. The spark chamber
consists of three sections: a coaxial capacitor and mount, the analytical
chamber, and a cathode mounting stage/high voltage connection. A description

of these sections follows.

19

+HIGH
VOLTAGE

COOLING WATER

COOLING WATER

IN

BRASS
TEFLON,

V‘|I|--.III‘II.I|-II"I'l'l‘-‘ IIIIIIIIIIIII

 

 

 

 

 

‘
‘
.

 

'I-‘ ~‘i---'~-~- --'--|"-----'-"-|'-

\\\\\\\S SS. /%

”mm ”m /////,\
. q
[I]. //,/ /////\::

3/

 

.////1//
////u.../

 

  

A
W

[I'II/n

.\\\\\\\\>”r

PURGE GAS

_, n
m1,
2
/n..

f
I.
.

Wém
/:\./

IV“.

_ ,4

IN

SAMPLE

—-) OUT

w”

. Vl/u /

 

/

W
m
,

:0",

“"‘r

V/é

SAMPLE
IN

EXHAUST

\\:\\

I’ll/’4

,, W \
, . \HA/AAAIM; It
__Tmm"m—”” Lsss

//m.////// /////// //\\\\\.

<\\\\\c \\\\\\\\\

 

THYRATRON

The NSS Housing

Figure 3-2

20

The coaxial capacitor section is similar to that described by Calkin (4).
A few minor modifications have been made. First, the set screw holding
the anode into the capacitor has been relocated. Orginally, the set screw
was mounted perpendicularly to the electrode axis, causing the sharp edge
of the set screw access hole to be located within two millimeters of the
dielectric material. Under extreme operating conditions, this generated
an electric field of sufficient strength to cause the electrical breakdown
of the dielectric material in the area of the set screw access hole. This led
to the destruction of the capacitor and to a tedious repair procedure. To
eliminate this problem, the set screw was angled into the electrode from
the bottom of the central capacitor core. This provides sufficient clearance
to avoid capacitor destruction from this cause under all conditions studied.

Secondly, a screw contact to the bottom of the capacitor was incorporated
to allow connection of an anode» access wire to the capacitor. The wire is
passed out of the spark chamber, to provide one contact for a bridging resistor.
Incorporation of an external bridging resistor greatly simplified resistor
replacement and spark disassembly procedures.

The final modification fo the capacitor section is the incorporation of
a Teflon wall between the capacitor section and the spark chamber. This
provides a barrier to prevent the sample from contaminating the capacitor
section and to improve electrical and thermal isolation of the capacitor section
from the remainder of the spark source.

An expanded view of the analytical chamber is shown in Figure 3—3. With
thyratron control over the spark breakdown, it is possible to overvolt the
spark gap. To eliminate arcing caused by the overvolt condition, the entire
analytical chamber interior was insulated with approximately 1/4 inch thick

Teflon. Complete electrical insulation was obtained by machining all parts

   
 

Ear?

II

.7//. S
'0
co
'0

  
  

I: ////'

//////Vfl/
NZ 4' /

/I"I¢
”a % #

 
    

  

22

to provide a tight, press-fit with adjacent components.

The upper portion of the analytical chamber is fitted with the purge
gas inlet. Connection to an external gas line are made with Teflon Swagelok
fittings. Gas exits from the purge gas inlet through a series of holes
concentrically arranged around the interelectrode axis. This sets up a regular
flow pattern from the top to the bottom of the analytical chamber, and purges
any extraneous sample. Purge gas exits the chamber through a 3/8—inch opening
in the lower chamber wall.

The center section of the analytical chamber houses an interchangeable
Teflon flow cell, the spark gap, and an observation window. The flow cell
decreases the effective detector dead volume and allows optimization of
the detector geometry for a particular sampling environment. A sample
stream and heated argon support gas stream enter one end of the flow cell
through a Teflon Swagelok fitting. At the center of the cell, the flow path
intersects with the spark gap. This region is monitored through 3/8-inch
observation port in the flow cell. A one-inch focal length lens, mounted in
the chamber wall, focuses the discharge radiation emitted through the
observation port.

The two electrodes that form the spark gap are 0.04—inch diameter, two
percent, thoriated tungsten electrode (Union Carbide, Linde Division) and
are located in the center of the flow cell. The anode is connected directly
to the coaxial capacitor. The cathode is electrically connected to the anode
of the thyratron. To decrease the severity of spark wander between the two
electrodes, the final half inch of the electrodes are electro-polished to a
very fine point. Sharpening is performed by placing the end of the electrode
into a saturated solution of sodium nitrite and applying approximately 40

volts a.c. between the spark electrode and a counter electrode placed in the

23

solution. Approximately ten minutes are required for complete sharpening.

The lower third of the analytical chamber contains the purge gas exhause
port and a Teflon wall to separate the analytical chamber from the lower
electrode mounting stage.

The last section of the spark chamber contains the lower electrode
mounting stage. The lower electrode is fastened into the Teflon-covered
brass stage with a set screw. A cathode access wire is attached to the stage
with a screw connection. The access wire is passed through the chamber
wall and soldered to the cathode end of the bridging resistor- A high voltage
connector (Amphenol 82-856) is screwed into the base of the stage. The outer
threading of the connector is screwed through the base of the spark chamber.
Rotation of the connector moves the cathode relative to the anode, to allow
adjustment of the gap length. This compensates for erosion of the lower
electrode during normal operation. The entire spark chamber is fastened
together with four 4 3/8-inch bolts.

The anode and cathode access wires pass through the chamber wall and
into the bridging resistor housing. A resistor is soldered between the access
wires and an insulated brass cover is placed over the resistor. Since the resistor
is externally mounted, it is effectively isolated from the high temperature
environment often present in the spark chamber. Because of this isolation,
resistor burnout associated with the Calkin design has been eliminated. In
addition, spark chamber disassembly procedures have been simplified because

of the increased resistor accessibility.

C. Thyratron Controlled Spark Source
Thyratrons have been used successfully in applications which require

rapid, reproducible high voltage switching (87-89). these include triggering

24

a spark discharge (89). The schematic diagram of the thyratron-controlled
spark source is shown in Figure 3-4. In operation, a high voltage power supply
(Spellman UHR10P100) charges the capacitor through the current-limiting
resistor RL. When the thyratron (EGécG model HY-6 hydrogen thyratron)
is non-conducting, both sides of the spark gap remain at the same potential
and no spark is formed. When a positive voltage trigger pulse is imposed
on the grid of the thyratron, the spark gap- cathode rapidly switches to ground.
This generates a potential difference between the spark electrodes, causing
a spark to form. After the energy stored in the capacitor is dissipated, the
thyratron is returned to the non-conductive state, and the charging cycle
resumes.

A constant hydrogen atmosphere inside the thyratron is generated by
electrically heating a titanium hydride reservoir (90). When an 800 volt trigger
pulse is imparted on the grid, hydrogen ionizes, and the thyratron becomes
conductive. After the trigger pulse is removed, the hydrogen deionizes, and
conduction ceases.

A six-volt, twelve-ampere, DC power supply provides the current for
the hydride heater and the thyratron cathode. A DC supply is used to minimize
trigger jitter caused by line-frequency modulations (91). Trigger jitter of
under ten nanoseconds duration can be obtained by use of a regulated power
supply. Thyratron ignition is accomplished with a EGGLG TR-27 trigger module.
A pulse from the trigger module synchronizes data acquisition timing with

spark ignition.

D. Capillary Gas Chromatograph - Spark Interface
The chromatographic interface used with the spark is pictured in Figure

3-5A. The interface design is based on Calkin's (4) packed column interface.

25

woe-Low xeaam co:obcoO::o.53>.:E mo Ewawwa 3:050 Tm mazwrm

ENmem xm<mm QquomHZOO Zomh<m>IH

 

 

 

 

 

 

 

 

 

 

5:536 3025 o». _S -
C
zome<m>zc I
uoemmm 5.236
mm.
Manama .5125 e
MW .6 W354?
+ :2:
aompzoooa
.. :23”.

EMKmxm xmcim Q<0 womDOQ

26

 

 

 

 

 

 

é—to In x 1/4 o.d. i;
GC OVEN :
I
i
CAPILLARY : x Ar on In
COLUMN 5 w : 9
i n
I I”
E 5 In. I p
i I
I
I
I

 

Side View End View

Figure 3-5 GC - Spark Interface

27

The interface is constructed from l/4-inch O.D. glass tubing. Since
capillary GC flow rates are inadequate to provide sufficient support for the
spark plasma, an auxiliary argon support flow is supplied through the sidearm
on the interface. The support gas flow rate is controlled with a Lab Crest
flow meter. Glass is used in preference to a sturdier metal interface to allow
observation of the column as it is threaded through the interface. This
minimizes possible column breakage during this process. In addition, the
glass interface insulates the equipment and the operator from the high voltage
present in the spark gap.

The entire interface is heavily wrapped with a flexible heating cord to
prevent condensation of the chromatographic effluent. Measurements with
a chromel—constantan thermocouple indicate no difficulty maintaining
interfacial temperatures at 280°C, the maximum recommended column
temperature. To minimize the detector dead volume, the capillary column
is extended past the end of the interface to within two to three mm of the
spark gap. Attempts to position the column closer to the gap caused
intermittent arcing to the end of the column, and resulted in an excessively
noisy signal and destruction of the column coating. The effective detector
volume under these conditions is approximately 50 n1.

The Teflon collar pictured in Figure 3-5B is used to connect the interface
to the spark chamber. Adjustment of the three radial screws allows‘ accurate

positioning of the column relative to the discharge axis.

E. Solution Introduction System
The current spark chamber has been designed to accommodate many
sampling environments, including operation in solvent mists as produced by

a nebulizer. Two types of nebulizers have been used for solution introduction.

28

A Plasma-Therm UNPS-l ultrasonic nebulizer system, pictured in Figure
3-6, was tested for use with the spark system. This setup is commonly used
for sample introduction into inductively coupled plasma torches (92). In
operation, a ceramic piezoelectric transducer is powered by a radio-frequency
(RF) generator tuned to the resonant frequency of the transducer,
approximately 1.4 MHz. Frequency adjustment of 10 kHz on either side of
this central frequency was possible (93). Cooling water was supplied at a
rate of 3 gallons per hour to dissipate heat generated in the transducer.
Solution was pumped to the transducer surface with a single channel peristaltic
pump (Ismatek AMY8.1-ABZ.SSR) at approximately one ml/min. A heated
argon gas flow flushed the solution mist from the nebulization chamber to
the spark chamber via a five-inch long piece of l/4-inch i.d. glass tubing.
The nebulization chamber and the transfer tube were wrapped with heating
tape and heated to approximately 120°C to assist in vaporization of the sample.

The crossflow nebulizer of Donohue and Carter (94) pictured in Figure
3-7 was also used for sample introduction. Two four-inch stainless steel
tubes, 0.095-in o.d., are used for solution uptake and argon inlet lines. A
l/4-in piece of 0.029-in o.d. stainless steel capillary column is silver-soldered
into the tip of each inlet line. To avoid excessive backpressure, the length
of fine bore capillary tubing is kept as short as possible. A high velocity
stream of argon across the uptake line aspirates solution into the nebulizer
chamber. Proper capillary alignment is required for optimum nebulizer
performance. Alignment is obtained by tighening the appropriate nylon
mounting screws for each capillary. Earlier nebulizer designs incorporated
metal-jacketed glass capillaries or all-glass capillaries which were frequently
chipped or broken; this caused erratic, non-uniform nebulization. The

incorporation of stainless steel capilaries has eliminated this problem.

29

couzsnoz 2:03.55 min oeswmm

 

 

_t u . y.
7 EE®M_ . (—

ZERO

 

It. _ Z. mud—2.4m

 

Z_ m<.0 memdU/

z. owxlv

z. elm

. ; . . 4 9m. ,
.So ONII \ .. _ . / a
. ‘ no 5.59» , ,

_1

 

 

 

3&5 oz_x_s_

 

 

 

IDS—MEMO

 

\\ \\\\\\\\\\s\\ T
\\\\\\\\\\\\\\\\\\

 

 

 

 

_ mums—<10
Dukdwl .0...

. EEbm..

NY LON SCREW

ARGON GAS
INLET

 

TEFLON BODY—u

 

 

30

SAMPLE UPTAKE

I/4 IN. NYLON SWAGELOK FITTING
TEFLON INSERT
RIGID PLASTIC FLANGE

O-RING

 

 

 

 

Figure 3-7

Cross-Flow Nebulizer

31

$5820 .8533me UoNEEEEZ mum ocsmmm

Ego

 

d c. _onoLu<
)0301 pogo:

J

My
fluu
. 0U
flnu
”nu
\ %
mu.
«no 0350“. 30 Loan; uczooo

:— Luau? 05.000

5 .303 05.000

30 L303 05300

 

 

 

 

 

 

32

Adequate nebulizer performance was observed above an argon flow rate
of 0.5 l/min. Care must be taken to provide a snug fit of all components
to insure consistent positioning control during use.

The low-volume mixing/vaporization/desolvation chamber pictured in
Figure 3- 8 was used to decrease the amount of solvent reaching the spark
chamber. The front four inches of the chamber are wrapped with heating
tape to vaporize the nebulized solution. The desolvation portion of the chamber
consists of a four-inch long condenser with an inner, water-cooled, helical

glass coil; this provides a relatively large surface area volume.

F. Gas Chromatographic Components

All gas chromatography was performed on a Varian 1400 series GC. Two
capillary columns have been used with this instrument; a 30 meter long, wide
bore (250 micron) column with an 0.25 um thick DB—l, equivalent to SE-30
or OV-l, (Anspec DB-1-30W-.25) and a 25 meter 320 mm bore column with
a 5.0 pm thick methyl silicone coating (Quadrex 007-1-25W—5.0F). Both of
these columns are similar to an OV-l column (95). The greater film thickness
available in the Quadrex column allows for heavier column loading, up to
1.0 ul, without serious degradation of chromatographic separation. The injector
was constructed from Scientific Glass Engineering Inc. on-column injector
kit OCI-3 shown in Figure 3- 9. On-column injection permits quantitative
transfer of the sample onto the front of the column, and is most appropriate
for very dilute and wide boiling range samples. High purity helium (99.9999
96 pure Linde) carrier gas is passed through a 0.5 ml/min flow controller to
provide temperature independent flow rate through the column. Compressed
air was supplied at 40 psi to provide injector coolant gas and to drive the

pneumatic head seal. Injector coolant gas was required to prevent sample

33

.8392: 00 Ewaaao min meswE

 

 

  

cEDBU
oZo> :<
:02, co>o III III 35:00 commocano
Luce. .
moo
a: 0x02
coaoouoo
moo
Lotcoo

 

mczooo

 

 

 

 

 

 

 

 

 

 

 

 

 

mcwzoLucoo
30-; 1LT LOuMMuo< o>_o>
0.30Eo3cm Hun mchfizm
cacao “co>

ocammoca nl_oo

34

fractionation in the syringe and to minimize deterioration of the injector
seals when the switching valve is opened. Pneumatic pressure is applied to
the head seal and the head seal opens. The head seal is closed when the
switching valve is closed and pneumatic pressure is released. Proper injection
technique was required to provide analytically useful information and to keep
the injector in good working order for a long time. To perform an injection,
the head was inserted until it touched the silicone head seal. After opening
the head seal with the switching valve, the syringe was completely inserted
and the head seal was closed. After a three second delay to smooth an inherent
pressure surge in the system, the sample was injected. One of two injection
techniques were used depending on the sample size. For a sample size of
less than 0.4 ul, a rapid injection is preferred. Alternately, for a large sample,
a slow injection rate allows the sample to be carried down the column by
the carrier gas, preventing sample loss off the start of the column (96). After
injection, the syringe was removed without reopening the head seal. When
connecting the capillary column to the injector, the end of the column must
be cut smoothly and inserted into the countersunk end of the injector and
fastened. If the syringe cannot be smoothly inserted into the column, the
column should be reconnected. Failure to do this leads to syringe breakage
and injector damage. During initial setup of the injector, the final 1/4 inch
of the injector was extended into the GC oven to permit heating of the entire
column.

A 5.0 ul syringe (SGE 5A-500-1005) with a 100 mm long stainless steel
sheathed fused silica needle (SGE VSl70/100) was used for all injections.
For adequate injector performance, the final 20 mm of the needle must be
unsheathed to ensure proper insertion into the column. Needle breakage

is commonly caused by improper allignment of the end of the column in the

35

injector. When this occurs, the needle should be replaced and the column

realligned to insure proper injector performance.

G. Alternate Spark Imaging Scheme

It was found that imaging the spark with a single'lens mounted in the
spark chamber wall provided less than optimal light throughout. To improve
this aspect of the system, an auxiliary adjustable multi-element lens mount
shown in Figure 3-10 was developed. The design consists of three concentric
tubes and two variable position, one inch diameter lends holders. A third.
lens can be positioned in the spark chamber wall. One end of the mount
replaces the lens retaining ring in the observation port, providing a stable
reference point for the lens system. The two lens holders incorporated in
the imaging system are each capable of approximately one inch of independent
axial movement. Each lens can also be independently rotated to provide
for specific angular orientation of radially nonsymmetrical cylindrical lenses.
By use of lenses with various focal properties and exploitation of the
adjustability of this lens mount, a wide range of spark imaging characteristics

is attainable.

 

    

 

 

k

\

 

 

 

 

 

 

 

CHAPTER IV

ELECTRONICS AND OPERATION

The data acquisition process is under complete microcomputer control.
A block diagram of the acquisition system is presented in Figure 4-1. The
Intel 8085 microcomputer designed by Newcombe (97) performs all
communication, control, and process coordination between the instrument,
mass storage capabilities, the operator, and the multiuser minicomputer.
The basic data acquisition scheme consists of microcomputer control of several
timing functions, acquisition of digitized emission intensities, and storage
of averaged data onto flexible disk media. Subsequent to the acquisition
of data from an entire experiment, data are transferred via a serial line to
an LSI-11/23 minicomputer where higher level smoothing, plotting, and
statistical analysis are performed.

This chapter presents a brief description of the microcomputer system.
The circuitry unique to the spark detector is discussed in depth. A brief
discussion of the higher level computing capabilities available on the

minicomputer is presented.

A. The Intel 8085 Microcomputer

The block diagram shown in Figure 4-2 illustrates the components of

37

38

 

 

oo<
._._m NF

 

 

 

 

 

mam <._.<n_ mmom

828.58% :o_:m_=co< 3am Co £2.35 :85 Te oesma

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n2< 220
u4m<§2<mooma

 

 

 

 

 

 

zoCommmoo
mz_._um<m

 

 

 

 

 

 

 

 

 

 

 

man

 

 

 

 

 

 

 

 

 

 

 

 

m0h<mouh2_
auh<o

 

xooao
m2: .2“;
IL
.6007.»
m<a<om m<ueom
m+<mou»z_ ><auo
coLoEoEoocoE
E0:
2:.
omp<o

\2

 

 

39

 

INTEL 8085
MICROPROCESSOR

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

12K BYTES 28K BYTES
ROM RAM
DUAL L—%PROGRAMMABLE
p
USART: .U,__7IN.UT/OUTPUT
3
(D
t
(D
7 D
INTERRUPT , CHIP
CONTROLLER SELECT
INSK .
CONTROLLER D A C ’

 

 

 

 

 

 

 

 

 

 

USER SPECIFIC
COMPONENTS

 

 

 

Figure 4-2 Block Diagram of Intel 8085 Microcomputer System

40

the general microcomputer system used in the spark data acquisition system.

The 8085 microcomputerhas an eight-bit data bus and a sixteen-bit address
bus which allows for a maximum of 64K bytes of address space. The
microcomputer contains 12K bytes of read only memory (ROM) and 28K bytes
of random access memory (RAM). The 12K bytes Of ROM are used for storage
of the system routines such as: communication software, an editor, and the
FORTH compiler used onithe system. The 28K bytes of RAM are used to
store system variables, the screen buffers used in the FORTH system, and
user variables, programs, and buffers. Communication .to a terminal and
the minicomputer is performed with the two universal
synchronous/asynchronous receiver transmitter (USART) and the interrupt
controller chips provided in the system. The minicomputer USART is hardwired
to Operate at 4800 BAUD; however, the terminal USART is wired to allow
software selection of a 4800 or 9600 BAUD transmission rate. The higher
BAUD rate allows for a more rapid printout of help frames during the data
acquisition process.

The disk controller circuit is based on the National Semiconductor
INSl77l—l floppy disk controller chip. Two Siemens FDD-100-8 single density,
8 inch floppy disk drives are used as the mass storage devices for the system.
TO accommodate slight differences in the design of the less expensive Siemens
drives versus the standard Schugart drives, a 20 ms interval (98) is required
between stepping pulses to the drives instead Of the standard 6 ms. These
changes are accomplished through software configuration Of the controller.
In addition, these disks require the read/write head to be disengaged when
the disk is not active and a 10 ms delay after the head is loaded. Each drive
can access a total of 250K bytes or 250 FORTH blocks of storage per disk,

which gives a total Of 500K bytes of mass storage for the system. In this

41

application, one of the disks is used for software storage, while the other
is used primarily for data storage during an acquisition run. This configuration
provides between fifteen and twenty hours of continuous data acquisition
per disk.

The final board required for microcomputer operation is the chip select
board, which decodes the signals present on the address bus. When an address
in the appropriate range is present, an active low pulse is generated. The
active range for the board is determined by the hardwired base address on
the board. The subsequent 512 addresses are decoded to give eight independent
regions Of sixty-four unique addresses. In addition, the highest active region
is further decoded to provide eight read/write qualified segments of eight
bytes each. These segments are selected only if a read or write operation
is performed; this provides immunity from erroneous selection due to bus
noise. If any of the addresses in a region are present on the bus, the chip
select for that region will become active. In addition to providing a centralized
way to access system devices, such as the USARTs and floppy disks, this
board provides a simple method to address the data acquisition electronics,
digital-tO-analog converters (DAC's), and other devices.

Two, eight-bit DAC's with selectable output range (Analog Devices AD558)
are used in the computer system. One of these is configured for 0 to 10 V
full scale and drives a Heath model SR255B chart recorder. The second DAC
is configured to output between 0 and 2.56 V and is part Of the automated
baseline subtraction circuit discussed later.

The final system board is the programmable input/output (PIO) board
based on the Intel 8255A programmable peripheral interface chip. This device
allows for configuration of three, eight-bit ports for parallel input or output

to external devices. For two Of the ports, port A and port B, all eight bits

42

must be dedicated to input or output for a given Operation. However, the
third port, port C, can be configured as eight bits of input or output, or, four
bits of input and four bits of output. The configuration of each port is under
software control which makes this a highly versatile interface device. The
PIO board is an integral component in the monochromator controller interface
which is discussed later.

The above components are mounted on one of two mother boards. The
mother boards, in turn, plug into a backplane, thereby connecting the bus
between the mother boards. Enough space is available on the backplane to
incorporate four mother boards. The data acquisition electronics board

occupies one Of these slots.

B. Minicomputer Facilities

All high level data reduction routines were performed on a Digital
Equipment Corporation LSI-11/23 minicomputer running the RSX-llM multiuser
Operating system. The minicomputer performs graphical data display and
generates high quality plots and figures. File based archival data storage
on flexible disk media is available at this level. An Integral Data Systems
Paper Tiger IDS-560 printer was used to printout of run parameters and source

code listings.

C. Data Acquisition Electronics (DAE)

The data acquisition process for a spark discharge detector system requires
the incorporation of several powerful circuits to perform time-resolved atomic
emission measurements. These components and the associated interface
to the controlling microcomputer are presented in this section.

1. Device Selection Circuitry

Communication to the various devices in the DAE is accomplished by

43

assigning each device a unique address for access and reading from or writing
to that address. To simplify interconnections between the microcomputer
and the acquisition electronics, only one chip select line is required in addition
to the microcomputer bus available on the backplane. All additional address
decoding is performed on the DAB board; hence, the chip select line performs
as a board select signal, mi.

The circuitry required for successful address decoding is shown in Figure
4-3. To enable address interpretation, a read or write operation must occur
to an address betweeen BFC016 and BFFF16. This causes two events to occur.
First, the output of the XOR gate goes to a high logic level and second, the
BDSEL line form the microcomputer is asserted low. These events enable
the 3-to-8 decoding of address lines A3, A4, and A5. When these three lines
are asserted low, i.e. addresses between BFCO and BFC7 are accessed; the
chip select for the AMD9513 timer chip and its associated bus driver are
selected over the 58-95-173- line. For addresses between BFC8 and BFCF, the
1' output of the 3-tO-8 decoder A is asserted low. This enables the second
decoder to demultiplex address lines A2, A1, and A0. In addition, the T output
is, after inversion, combined in a NAND gate with the least significant address
bit (A0) to enable the gain select latch. The resulting signal is inverted and
used to latch the data present on the input of the 7475 four bit latch. The
net result Of this circuit is that data latching occurs only if one of the four
Odd addresses between BFC8 and BFCF is accessed. To complete the gain
selection logic, the _1', 3, 5, and 7 outputs of the 3-to-8 decoder B are tied
to the data inputs Of the latch. The outputs Of the latch are used to activate
the gain selection relays. The latch outputs are connected to the five volt
power supply through 0.01 uF capacitors to minimize switching transients

and improve switching reliability.

44

329:0 :ofioflomuoggn doctor:

”.4 SEE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.552.» m m «min—V
oo<omal m H mm m+
Jr Mn no m w u< I ~<
no“ mun. m we m ._ ..< A P<
«or .6. u. «o .n. a. o< I o<
J» H mm A on w A U
L: To Fur..." mu 5
H - .
fl i
m+ m+
o «mllw 2:31. {m3
w n Milan“ E.
n+ m. U I Ebb
com: in ._ ~< I n<
npmmmoh m .4 I I. .2
h o< IIAn<

 

 

 

45

The 9 output Of the second 3-to-8 decoder is used to access the
analog-to-digital converter by asserting the RITA-DC line. Finally, the 2 output
asserts the WP] line, which initializes the integrator reset pulse. These
two circuits are discussed later.

2. Data acquisition timing

Several timing functions are required for proper Operation of the data
acquisition process. After the spark fires, a short delay period is required
to allow the initial continuum radiation to diminish. Following this, an
integration period elapses during which the photomultiplier output is integrated.
These periods are typically a few microseconds in duration and are controlled
by the microcomputer. In addition to the rapid timing required, an elapsed
time clock capable of Operation over the maximum duration of a typical gas
chromatographic run is necessary for subsequent reconstruction of the
chromatogram. These timing criteria were met and exceeded by a single
integrated circuit, the Advanced Micro Devices AM9513. A brief description
of this device is required for full appreciation of its versatility and flexibility.

The AM9513 system timing controller shown in Figure 4-4 consists of
system control registers, a frequency source, and five independent, hardware
and/or software gateable, sixteen-bit counters/timers. The system control
registers consist Of a data register, a command register, and a master mode
register. The command register and the data register send and receive all
communications with the microcomputer. The command register receives
instructions concerning the loading of data into the various registers,
initialization, activation and deactivation of individual counters, and
modification Of specific system modes. The data register acts as a buffer
between the microcomputer data bus and the source or destination register

on the AM9513. By design, data are only transferred directly from the data

46

  
    

#30,

#30

@310.

8:05:00 wEEE.

”32>; moot/On 98:2 coocw>n< Ho Ewowwa :83 To 959m

 

 

maomu 0.00.—

 

 

 

 

 

 

 

 

 

 

@—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. muhzooo , 4 ooopzoo
II .6528 moo m...ooo
Amotooo
) ,
A:
ooomo o_oo._ - L4
u mmpzooo 556mm «5......on
_ L_~ moo: 55.3 33.23
. \Xo _
moooo o_oo._ ,
n «32:3 5.6.3:. r «526..
H mop<hm _ <p<o
moomo 063 28.. 5335 Sou
4 «32:8 Budm Saz. So... to:
L_A R 5.2% .
.. omo
ooomo 063 No com... comm: m
00
m mMHZD \Nm mwh<0
_I...L \ oom

47

register or the command register. When information concerning another
register is transferred, the command register points to the appropriate internal
register, a read/write operation is performed to the data register, and the
AM9513 transfers the information between. the data register and the
appropriate internal register.

The master mode register controls the configuration Of the frequency
source, the data bus width, and the time-of-day counting mode. The frequency
Of the internal frequency source is derived from the system clock or any
of the source or gate inputs for any of the five counters. A wide range Of
output frequencies is available by the incorporation of an internal frequency
divider.

Each of the five counters diagrammed in Figure 4-5 consist of independent,
software-accessible mode, load, and hold registers and hardware-accessible
source, gate, and output ports. In addition, software gating of the timers
is available. Each counter can be programmed to Operate in any one Of nineteen
different modes. This capability gives this device much Of its versatility.

To Operate a given counter, the specific mode command must be loaded
into the mode register, and starting values for the counter should be written
into the load and hold registers as required. A counter-specific load command
is then issued to initialize the counter to the appropriate value. If the
application requires a specific initial output state, the output state should
be set or cleared at this time. When the counter is to be activated, a
counter-specific ARM command is issued. Counting/timing then commences
as allowed by the source and gate inputs.

The AM9513 chip is used to control the critical timing requirements
of the data acquisition electronics. These include generation Of delay and

integrate times, generation of a pulse stream to drive the monochromator

48

95.5 3:500 2352 no Eoomwa OBOE one moswmm

 

 

556mm.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_
_ >._zo 2.5? ton?
_ a oz< .
_ mmhzo he
_
_ mop<m<ozoo
_ Lento.
_ It?
I .L I l IIIIIIIII _
556mm
ooo: tone.
I . u.
z oomhzoo fl mupzaoo
5.30 ban—#30 _ .Emlw-
- on
$5.3m

0(04 .EmIm-

 

 

mwhwaum
.waoz twin —_

 

o

 

 

 

 

Azooo
ooohzoo
mopzooo
.203 7.23
Louoom coma
Fonz. op<o
oomoom

 

 

49

scan motors, initiation of the analog-to—digital converter, and operation of
the elapsed time clock for time correlation of chromatographic data.

TO perform these functions, three Of the nineteen operation modes are
selected for the various counters/timers. A brief description of these three
modes will be useful in understanding the timing circuitry operation.

The first important mode is designated as "Mode A". A counter configured
in this mode acts as a simple, one-shot, up or down counter. The counter
will count edges occurring at the source input following an ARM command.
On reaching the terminal count, i.e. the minimum or maximum values allowed
for the counter, the counter disarms itself and inhibits additional counting
until another ARM command is supplied. The internal count register is also
reloaded with the contents of the load register on the terminal count. One
Of the two output functions can be obtained in this mode; either a pulse
occurring during the terminal count or inversion of the output state following
the terminal count. The second of these output modes is called "terminal
count toggled output".

Square wave generation is accomplished using "Mode D". In this mode,
the counter repetitively counts to the terminal count once it is armed. On
the terminal count, the counter reloads itself from the load register. Selection
of the terminal count toggled output mode will generate a square wave with
a frequency equal to the count source frequency divided by two timesthe
value stored in the load register.

The third mode of importance in this application is "Mode L". After
loading the contents Of the load register and arming the counters, a rising
or failing edge at the gate input triggers the counter. The counter commences
counting source edges. On the first terminal count, the counter automatically

reloads itself from the hold register and continues counting. On the second

50

terminal count, the counter reloads from the load register. At this point
counting stops until another triggering edge occurs at the gate input. Any
gate edges which occur before the counter is armed or while the counter
is counting are ignored. Use Of the terminal count toggled output option
in this mode generates a hardware-triggered, delayed pulse in .which the length
of delay and the pulse length are proportional to the values in the load and
hold registers respectively.

The schematic diagram of the circuitry associated with the AM9513
timer is presented in Figure 4-6. The output of a 20 MHz crystal oscillator
(Motorola K1091A) is halved using one stage of a 748112 JK flip-flop. The
resulting 10 MHz signal is supplied to the X2 input of the AM9513 to act
as a master clock for the timer chip. Connection of the timer to the
microcomputer data bus is made through a 7 4LSZ45 line transceiver.

The direction of operation of the transceiver is controlled by the S/R
input, which is connected to the RT) bus signal. Both the transceiver and
the timer are enabled with a low pulse on them signal, which is generated
in the device select logic circuitry.

Access to the various registers on the timer is controlled by the three
inputs WR, RD, and C/D-. These inputs are connected to the microcomputer
bus lines WR, RT), and A0 respectively. In this configuration, a write
instruction sends data to the timer and a read obtains data from the timer.
The state of the C/D input determines whether the command register or
the data register is accessed during an operation. Since this line is connected
to the least significant address line, an input/output Operation to the even
address 589—51-3- accesses the data register and an I/O Operation to the odd
address (Egg-51.3 + 1) accesses the command register.

All five available timers are used in the data acquisition scheme. Timer

51

2:26 85:. o4 Sana

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

nAnoa< . unumons
fiLnocm so . so :1 so
Sou no . no n<...lllA no
one no no «(Iii no
23m». :3 Po . .o I .o
ea a.c. » so . on Elli so
53;:Y «:3 mm mm «\m
H «So on. It. o<
E on. . It
IE1 ".29 .. A g
o¢.mvs A . u t
E Q. So x l.
:.
F2. A.
+ .
.L a. ,. J. L
, . wfi
00> 020 mum L
. L 2:. on. o A A H n...
ppm . n+ h: fl. xo .n -
onpmvs <.oo.x . ~.Amcs «\L
ox A n\. «Jacobo:

m+

52

2 is configured in "Mode L" and controls the delay and integrate times. Use
of the 10 MHz clock frequency for the source frequency of the timer allows
0.1 us resolution of measured intervals. As such, the delay and integrate
times, specified in tenths Of microseconds, are stored in the load and hold
registers. Synchronization of timer triggering with spark ignition is
accomplished by sensing the rising edge Of the thyratron trigger module .
synchronization pulse output. Once the delay time elapses, the output of
this timer toggles to a high logic level. This signal is passed through a 74Sl40
NAND line driver. The resultant active low signal enables the gated integrator
and switches the photomultiplier tube to a high-gain state. Termination Of
the integrate time returns the output to a low state, which disables the
integrator and returns the photomultiplier tube to a low—gain state. The
falling edge of this output pulse is also used to trigger timer 1.

Timer 1 is also configured in "Mode L" and Operates with a 10 MHz source
clock. This timer provides a 12.8 us setting time before a 500 ns
start—of—convert pulse .is sent to the analog-to-digital converter. Following
inversion, the low logic level pulse is asserted on the CVT line to initiate
conversion.

Timer 3 is configured in "Mode D" to act as a square wave generator.
The pulse stream is used as the external scan rate input to the monochromator
controller. A software-selected source frequency of either 1 kHz or 10 kHz
allows generation of scan rates between 0.001 nm/sec and 2.55 nm/sec with
a wavelength increment of 0.01 nm.

Counters 4 and 5 perform as the elapsed time clock for the data acquisition
system. Counter 4 is configured in "Mode D". The FOUT divider supplies
a 200 Hz source frequency for counter 4. A 4 Hz square wave is generated

at the output by repetitively counting to 25 and selecting the terminal count

53

toggled output option. The output for counter 4 is used as the source for
counter 5. Counter 5 is configured as a simple 16 bit counter in "Mode .A".
Since the input frequency 4 Hz, the accumulated count is simply the number
of quarter seconds since timer initialization. At this rate, a total elapsed
time is retrieved by issuing a SAVE command for counter 5. The SAVE
command transfers the accumulated count from an internal register to the
associated hold register without interruption of the counting process. A
subsequent read from the hold register transfers the count to the
microcomputer.

3. Analog Electronics

The analog circuitry presented in Figure 4-7 performs signal integration,
amplification, and conversion to the digital domain. The three major sections
of this cirucit are a gated integrator, a programmable gain amplifier, and
an analog-to-digital converter.

a. Gated Integrator

An Evans Associates model 4130 gated integrator card performs integration
Of the photomultiplier signal. Two additional analog inputs are combined
with the photomultiplier signal at the summing point Of the integrator. A
static Offset control to set the zero point of the integrator is constructed
from a variable voltage source present on the integrator card. The output
of the voltage source is available at the Vx output of the card and is connected
to the summing point through a 10kI2 resistor. The second Offset voltage
is generated by one of the system digital-to-analog converters (DACs). The
DAC output is connected to the summing point through a 33kn resistor. This
input is used as part of the background correction circuitry discussed in the
next section.

In addition to the analog inputs, two TTL inputs are required to control

54

 

 

 

 

 

 

 

I

 

l I

 

 

 

 

EVANS ASSOCIATES

4130 GATED INTEGRATOR

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

”>—
"0%
RITE-5% E-—‘
79-3245 - 9 new:
REF—fix???
DO)— .6
D, >———A1 'q
oak—A2 .2
03>...“ ’3 1/2 74123
DSHA4 8‘ 0. 100 +5
D'>'— A. a. D, a 100 kg
D7) .LPJ. ”11 sis—«MPH:

 

Figure 4-7 Analog Electronics

55

gating and to reset the integrator. Gating of the integrator is controlled
by the active low TNT signal generated in the timing circuitry. Integration
continues for the duration fo the low pulse. This period is the integrate time.
The integrator reset signal is an active low, 2.8 us pulse derived from the
Q1 output of a 74123 monostable multivibrator. The monostable is triggered
by a low pulse on the MST-1. line. Since the m pulse is generated
in the device select logic, a software command is required to reset the
integrator.

b. Baseline Subtraction Circuitry

In many circumstances, the background radiation signal was found to
account for up to 75% of the usable dynamic range of the gated integrator.
This was true particularly at high gain settings. To circumvent this problem,
a baseline correction circuit was developed. Since the photomultiplier supplies
electrons to the integrator, an electron drain or positive voltage source is
required to counteract excessive background signal. In addition, for maximum
usefulness, a wide range of voltages should be readily and rapidly available
with the selected voltage based on the observed background signal. These
features are obtained by using. a digital-to-analog converter under
microcomputer control. Since the microcomputer has access to the intensity
level and the current gain setting, an appropriate offset value can be calculated
and sent to the DAC. A 33kg resistor is chosen to connect the DAC output
to the integrator summing point; this provides offset capabilities over the-
entire integrator output range at all gain settings and integration times
typically used.

c. Programmable Gain Amplifier

The programmable gain amplifier consists of four normally-open five

volt relays (Magnecraft W107DIP—1), five precision resistors used for gain

56

determination, and a Teledyne Philbrick 1026 FET operational amplifier.
The operational amplifier is wired as a voltage follower with gain. The
noninverting input is connected directly to the output of the gated integrator
and the inverting input is connected to ground through a 1kQ precision resistor.
A feedback resistor connected between the inverting input and the output
determines the gain associated with the circuit. Assertion of a high logic
level on either the G1, G2, G3, or G4 inputs activates the corresponding relay
and switches the appropriate resistor into the feedback loop. This effectively
selects a specific gain. The device select logic assures that only one of the
relays is activated at any one time. Reed relays are used in preference to
other electrically controlled switching devices because there is no leakage
current through the relay. A leakage current would divert some of the signal
through each of the feedback resistors, and undesirably alters the gain
characteristics of the circuit.

d. Analog-to—digital Converter Circuitry

Signal conversion to the digital domain is performed by an Analog Devices
AD574JD analog-to—digital converter (ACD). Power supplies and offsets
are wired according to manufacturer's specifications for operation in a unipolar,
0-10 V input mode (99). The analog input is connected directly to the output
of the programmable gain amplifier.

Data conversion begins when both the ES and the R/E inputs are asserted
low. This occurs following the settling delay at the end of the integration
period. Connection of the CS— input to the INT line prevents conversion or
data access from occurring during the integration period. Control over the
conversion precision is governed by the state of the A0 input. A lZ-bit
conversion occurs when the A0 input is low. Since this input is connected

to the Q output of the integrate reset monostable and the Q output is low

57

at this point in the acquisition process, a 12-bit conversion is performed.
After the conversion is complete, the STS output line is asserted low. The
high-to low transition of the STS output generates an end-of—convert interrupt
on the RST5.5 microprocessor input. The interrupt is programmed to initiate
a routine to access the digitized signal.

The microcomputer used here is an eight-bit machine; hence, two read
operations are necessary to input all twelve bits of information from the
ADC. The A0 line controls what information is available on the ADC data
bus. When A0 is held low, the most significant eight bits are accessible;
however, when a high‘ logic level is asserted on A0, the four least significant
bits are placed on the bus.

The ADC bus and the system bus are connected with a 74LSZ45 8-bit
bus transceiver. Access of the entire 12-bit data word is accomplished in
three steps. First, a one-byte read from the ”1513766 address inputs the high
eight bits. Following this read, a read or write operation must be performed
to the TIN—TESTI— address. This triggers the associated monostable to reinitialize
the integrator and to assert a high logic level on the A0 input. Finally, a
second read fromm transfers the low four bits to the microcomputer.

4. Photomultiplier Gate Circuit

Switching the photomultiplier from a low-to a high-gain state is required
in the data acquisition scheme to protect the photomultiplier from saturation
during the initial intense continuum emission generated by the spark discharge.
The high-speed photomultiplier gate circuit used is identical to that described
by Calkin (4) and is shown in Figure 4-8. In normal operation, a small potential
difference, approximately 20 V, is maintained across the third and fourth
dynodes. This keeps the photomultiplier in a low-gain state. A low pulse

on the PMT line switches the photomultiplier to the high-gain state by applying

58

mg 346

2:26 83 amasseosfi M: 2%:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

rllll

 

_
I|1 IIIIIL
.2388 8.3 4426.5

 

 

 

 

 

59

80 to 100 V across the dynode pair. The rising edge of the WT pulse returns
the photomultiplier to the low gain state. Switching times of approximately
30 ns are reported with this gate circuit (4).

D. Monochromator controller/microcomputer interface

In the data acquisition process for any spectroscopic technique, information
concerning the identity of the sample is encoded by optical signals at specific
wavelengths; therefore, the capability for accurate mechanical selection
of the appropriate wavelength is an important criterion in the design of a
spectroscopic instrument. Wavelength selection with the original
GOA-McPherson EU-700 series monochromator and controller was manually
performed by slewing to within a few nm of the desired location and then
scanning slowly to the final wavelength. Unfortunately, due to electrical
and mechanical shortcomings in the original design, final wavelength selection
often became a matter of trial and error. To alleviate this deficiency, the
microcomputer interface to the monochromator controller, shown in Figure
4-9, was designed.

The interface provides microcomputer access to the current wavelength
setting and control of slewing and scanning operations. .Communication
between the microcomputer and the interface is performed with the INTEL
8255 programmable input/output circuit. This provides three, eight-bit ports.
Port A is not used in this application. Port B is used as a bidirectional data
bus between the microcomputer and the interface. The low three bits of
port C select the interface function to be performed, either a wavelength
read, or, initiation or termination of a scan or slew operation. External control
of the scan rate is accomplished with one output of the AM9513 timer.

In normal operation, a zero is written to port C, which causes the? output

of the 74LSl38 3-to-8 decoder to go low. The signal is inverted and connected

60

aha:
z<0m
#2.

28h:—

zmhxu

8:“

z<om

hxu

m.npa 20 am

«.nwa a: am

833:: ecquoEoocoZEousa58232 muv ogswmm

 

    

      

on.-
Ill

bvam—nn
'
-¢«<«<

N

   
 

mm; cuhmaum Ihozu4u><3 mm:

«9.; pp 4.x a o_: h.m o _o t n.u m N n o

    

3.. D
_I

  

          
  

       

   

    

so .aua.asa

n nmavh
o a .

anoonn .a
MW *5

oawa
p

”a

   

n

 

a
n magi.
p o a

 

fl

 

B'JMN'I‘I
h
‘

 

V’
(

  

I
O
<

 

.

         

     

   

O .0-
o bros
or.

 

    

          
  

         

                       

hack
0:.

61

to the .58- input of the 74LSZ45 bus transceiver and the E inputs of the three
74LS373 wavelength latches. This disables the line transceiver and enables
the latches to track the current wavelength. The data inputs of these latches
are connected to the binary-coded-decimal wavelength signals available in
the controller at J14 (100). To accommodate the eight-bit architecture,
the five—digit wavelength must be read in three operations. Writing a nonzero
value to port C latches the wavelength latches and selects the line transceiver.
The state of bit C2 determines the direction of data transfer, i.e. a low level
on CZ causes data to be transferred from the interface to port B, and a high
level reverses the data flow. Writing the value 1 to port C causes the Toutput
to attain a low level. This enables the output of the latch which holds the
two most significant wavelength digits. The information is then transferred
from the latch through the transceiver to the port B inputs, from which a
read operation is performed. The next two digits and the final digit are read
with the same process, except the values 2 and 3 are written to port C,
respectively. Following the complete wavelength transfer, the value 0 is
written to port C to return the interface to the monitor level.

To initiate scan or slew commands, data must be transferred from port
B to the 74LS373 command latch. This requires port B to be reconfigured
via software as an output port. The appropriate command is then written
to port B. Writing the value 4 to port C generates a low signal on theToutput
of the 74LSI38 three-to—eight decoder. This changes the direction of the
bus transceiver and disables the output of the control latch. The 71' output
is inverted to enable data tracking on the control latch. Writing the value
5 to port C maintains valid data on the interface bus and latches the control
latch. Writing a zero to port C enables the control latch outputs and reinstates

wavelength tracking.

62

The valid monochromator. functions controllable with the interface are
scanning or slewing in either direction. As purchased, the controller included
the logic circuitry required for external control of these functions; however,
to permit microcomputer selectionof the function, three connections had
to be broken. These were the INTRN and EXTRN lines from the hardwired
internal/external switch and a direction sense line originating on the front
panel and accessible at J13 pin 5 (97). All modifications of the original design
were performed on the wavelength circuit board. Manual monochromator
control with the front panel switches on the controller is possible, provided
in INTRN line is not asserted low.

Microcomputer control of the INTRN and EXTRN signals is provided
by the Q7 output of the control latch. A high Q7 output asserts the EXTRN
line high and the INTRN line low, which selects external monochromator
control and inhibits manual operation. Scan direction is determined by the.
state of the Q6 output, such that a high level selects a scan to longer
wavelengths. Assertion of the Q0 or Q1 control bits to a high level initiates
a slew command to longer or shorter wavelengths, respectively. Assertion
of both bits inhibits all slewing. Since the command is latched, microcomputer
intervention is required to terminate a scanning or slewing Operation. A
summary of inputs required for interface operation is presented in Table
4-1.

Minor modifications were made to the monochromator circuitry to
minimize the detrimental effects of radio. frequency interference and random
voltage spikes associated with spark ignition. These effects included loss
of correlation between the actual wavelength and that determined by the
monochromator controller and unsolicited activation of the scan stepper

motors. Four 0.1 uF ceramic disk capacitors were placed between ground

63

ccmEEoo €qu
Hobcoo 3:522
* a: :wom

._. Esoo :wom
Esov 33w

9. 33m

33:8 .63an0

«nu—mom

m>oaw ma 05mm

zzxxxxxfi
zzxxxxfio
,zzxxxxco
ofixxxxco
chxxxoc
>>xxxxcc
555355559
atom

.555
0 tom

8:25.30 3332— 90:05—30 3352:0982

.mlv 036E

com
cam
com
com
cam
cam

cam

" mcomuocam 3.5.50

2303 tom
0.35500 OE

64

2qu “833
3&8 N 3.522
33% N 55

Loumcwb o: . 32

Hun—mum

-cum-
cow- -cam-
-ccm- -omfi-

GEO;

engnunnnvnmncnsn
m E...

82 .38 @239» 33536 gm: ”_sz Ann _..

in. .850 we 32.: omwofi 25% on $2: nu a
~33 033 .858 Ann x

H a a mom

a H c mam

H o a New

o c. a «cm

H «:9; face ~ 263

.55”: :38. to.—
0 tom c5568 OE

.28 Tu 03$.

65

and the four stepper motor drive lines on the wavelength circuit board. This
largely eliminated activation of the stepper motors due to electical noise.
To minimize problems with wavelength synchronization, two 0.03 uF capacitors
were placed between ground and the ¢A and ch outputs of the wavelength
encoder. Installation of capacitors larger than 0.03 uF caused valid encoder
pulses to fail to trigger the wavelength counters; hence, synchronization
was again lost. In addition, a 1.0 uF tantalum capacitor was installed between
ground and the five volt supply used by the encoder to further minimize false
triggering of the wavelength monitoring circuitry.

These steps significantly improved the noise immunity; however, during
periods of extremely high interference, wavelength synchronization was still
lost.

E. Spark Detection Circuit

The synchronization pulse output by the thyratron trigger initiates the
data acquisition process; therefore, it is necessary to know the timing
relationship between the trigger pulse and actual spark ignition. A photodiode
trigger circuit similar to one designed by Seng (107) was used to detect spark
ignition. The photodiode circuit is shown in Figure 4-10. The only modification
to Seng's design is the omission of a 4.7 k9 resistor (bold face in the figure)
connected between the photodiode anode and ground. Removal of the resistor
increases circuit sensitivity and improves the fall time when light strikes
the photodiode. An increase in the rise time is also observed; however, since
the spark ignition time is of interest, the rise time is not important in this

measurement. In operation, fall times of less than 50 ns were observed.

€86 amaze meoGSofi S-.. 95me

m...

 

/\
l

F9

cxor

axow

 

m.+

 

mmnNZN

mmnNZN

 

A
N

 

 

cxm.F

 

m+

thé.

com am:

2

CHAPTER V

Software

Microcomputer control of instrumentation is a rapidly growing, beneficial
process applicable in many areas. Benefits include the rapid synchronization
of the many small tasks required for data acquisition, transferral of data
between different computer facilities to exploit various capabilities of the
different machines, and immediate archival data storage. In addition,
microcomputer control permits the incorporation of increased versatility
into the acquisition process as the need arises, without necessitating a redesign
of the entire system.

Selection of a programming language is based on several considerations
associated with the particular application such as, the available memory,
the required computation speed and power, and the overall complexity of
the control process. Often, the basic instrumental control process requires
the repeated manipulation of several parameters and fundamental operations
in an identical manner. More complex control structures incorporate several
of the fundamental routines in a precise sequence. The function of the
software used for a given application is to provide for the development of
fundamental acquisition and control routines, and allow development of

higher level control structures by linking the fundamental routines in a logical,

67

68

useful order. To perform these operations, mechanisms for editing and
compiling are required. Instrument control operations are most conveniently
performed from the lowest programming level, assembly language; however,
data manipulation and process coordination are more conveniently and
understandably implemented with a higher level programming language.
Thus, the ideal interface control language should provide mechanisms for
combining assembly language routines into the higher level programming
language programs.

Two languages were available for the microcomputer used here, SLOPS
and FORTH. SLOPS or Structured Library Oriented Programming System
as developed by Gregg (101) is an assembly language based package. It consists
of a set of user-callable subroutines to perform general functions such as
multiplication, division, character input, and conversion. Although this
vastly simplified coding assembly language acquisition routines, no compilation
or editing capabilities were available on the microcomputer. Therefore,
program development and compilation were performed on the LSI-11/23
minicomputer and downloaded to the microcomputer. When this procedure
was used, program development was a time consuming, error prone, and
often futile process. In addition, since all routines had to be written in
assembly language, highly versatile programs required long development
times. These factors made SLOPS a less than ideal language for this
application.

The second language available was .FORTH, which offered many
advantages over SLOPS. The FORTH installation used is similar to that
described by Brodie (102). The base level -of FORTH includes a line editor
and compiler. This allows program development to be performed on the

microcomputer and eliminates the time consuming compilation and

69

downloading steps required in SLOPS. As a consequence of FORTH's structure,
assembly language routines or "words" are easily incorporated into high
level FORTH words. This ability makes FORTH an excellent choice for
instrumental control operations. Additional benefits derived from FORTH
include: double-precision integer math, a multiple-buffer block input/output
structure for data transfer, multitasking. capabilities, and a threaded
programming structure. These features are all exploited in the data
acquisition software.

This chapter describes the data acquisition software used on the
microcomputer. The programs used to massage the data and convert it

to plottable form are also discussed.

A. Data Acquisition Software

The software controlling the acquisition process is hierarchically arranged
as shown in Figure 5-1. In this organizational scheme, a menu-drive, monitor
routine running as a foreground task, is the general user interface to all
levels of the system. All hardware and software acquisition parameters,
the monochromator interface commands, and file handling routines are
accessed from this level.

The menu organization implements two important features. First, any
table entry can be accessed with a single keystroke, and second, a prompt
screen that describes all options available on the current page is listed.
These features make option selection a quick and easy procedure, minimize
the amount of memorization of word names required, and decreases the
time required to optimize the system parameters during a particular
experiment.

The four, twenty-six entry menu tables shown in Table 5-1 allow selection

70

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cotwficmmeo ohwstom 5362630 3me Tm. charm
55.53.51 uzzaox uzzaom mzfiaom mezoEoud
333m AAAAAAA c.5332; AAAAAA “Baum AAAAA zoEmSao<
nux . 7.6.. (:3 «wine r3532. £3
\1 ) ) \1
0 \l \/

MW 5 5 D D

) \I ) x/

o w » n n

M 5 m

000000 mzfizom motzo: zoEmSoo< 000000000000

 

 

 

 

 

 

 

 

Table

Key

EQ'TJITJUOUD'JP

D—i

wowozzt‘xm

U)

H

DTBL

CHOOSE

DIR

GTBL

FILSAV

GTBL

VHD

DDISK-INIT

VDEL

?LINE

MONO

NEWWAVE

23LINE

LTBL

GTBL

$DAC
STBL

TALK

71

Table 5-1.

SPKPTS

BSSET

CHOOSE

*DLY

GTBL

FINDPEAK

GAIN

OBSL

*ISET

FSTA RT

FSTOP

?LINE

MASK

CVT

GWATCH

PARAMS

ABRTRN

RPAVG

ETBL

TAKE

Available Acquisition Options

?AVER

!BKGND

CHOOSE

GTBL

FINDPEAK

!PEAKVALUE

-BKGND

PARAMS

ABRTRN

RPT

ETBL

TAKE

AVING
BSSET
CHOOSE
DSCAN

GTBL

?LINE
SCANINC
CVT
PARA MS
ABRTRN
RPT
ETBL

$DAC

Table 5—1 cont .

Table

Key

N N N 2 < C

HCOPY

>LAMDA

>3TBL

>2TBL

>2TBL

72

UNMASK

>3TBL

>LAMDA

TPTS

?>1TBL

>3TBL

>2TBL

>LAMDA

SVLINE

>2TBL

?>4TBL

>LAMDA

CNCN

?>3TBL

73

of options from three major categories. These are microcomputer support
functions, general acquisition parameter control functions, and specialized
acquisition routines. The first menu consists of routines for disk drive control
and communication between the microcomputer and the minicomputer.
In the second and third menus, a generalized acquisition routine can be
selected. While the experiment is in progress, the user, by appropriate option
selection, can modify any experimental parameter, such as delay or integrate
times, and can observe the effect of that modification immediately. This
greatly simplifies optimization of experimental conditions. The specialized
acquisition routines available in the fourth menu are designed to record
the effect of a single variable on an observed emission intensity. This is
accomplished by synchronizing the change of the parameter with the
acquisition of a number of data points. The value of the variable, the average
emission intensity, and the standard deviation of the emission are then saved
as an X, Y, Z triplet for later analysis. Synchronous delay time, integrate
time, or wavelength scans are examples of routines available in this category.

The three control routines which perform the acquisition and data storage
process for all acquisition are the interrupt service, buffer management,
and data storage routines. The net effect of the three routines is the
acquisition of individual emission intensities, followed by the storage of
average intensity data to flexible disk media. Since these are interrupt
driven routings and background tasks, the monitor routine can be executed

concurrently to allow user interaction at any point in the acquisition process.

B. Interrupt Service Routine
The RST5.5 microprocessor interrupt input is triggered by the

end-of-convert pulse. When enabled and unmasked, the RST5.5 signal initiates

74

the execution of the interrupt service routine diagrammed in Figure 5-2.
Upon execution, the 12-bit digital value representing the intensity is read
from the analog-to-digital converter and summed into a grand sum. the
interrupt routine then terminates, and returns control to the monitor task.
After a user-selected number of spark intensities are acquired in this manner,
the interrupt routine resumes the buffer management background task. The
buffer full flag is then checked. If the buffer is marked as full, the data
storage background routine is resumed. Following this, control returns to

the monitor routine.

C. Buffer Management Routine

The buffer management routine controls the data buffer filling process
and normal run termination procedures. The flowchart shown in Figure
5—3 depicts the buffer management background task. Upon initialization,
the average intensity of the grand sum for the selected number of points
is calculated. If appropriate, this value is sent to the DAC controlling the
strip chart recorder. The 12-bit average is then stored in one of the two
disk buffers available in FORTH. When only six bytes remain in the current
buffer, the elapsed time and current wavelength are read from the elapsed
time clock and monochromator interface. These values are then saved as
the last six bytes in the data buffer, and the buffer full flag is set. Run
termination conditions are checked. If no termination condition exists,
the buffer management routine exits. When a termination condition is met,
the run-in—progress flag is cleared, all active data buffers and current run
parameters are saved to disk, and the disk directory block is updated to
reflect the presence of another data file. Finally, the flags indicating an

active background task are cleared to prevent inadvertent corruption of

75

ocmusom 833m 3:23.: .50 :95 32m mum 25mg

zmahum

Pasmmuhz
w4m<zm

a.

mmuhmamm
muchmwm

 

 

 

 

 

 

 

1_

 

 

 

33m
IMO...

 

00¢
a<um

 

 

 

mmuhmgum
Imam

a

hmammuhz
u..m<m.o

 

 

 

 

 

 

 

h¢<hm

76

25:30 EoEmwwcaE Hotsm .50 «.820 32m

zmahum

 

  

 

a

 

 

 

 

>m0hommj—

meson:

 

man—m
<H<o

u><m

 

 

   

    

0.00.5
mqm<i<><
hm<4

  

 

 

 

 

 

E 2&3

 

 

 

 

 

uo<mm><
:mou

 

 

77

the disk buffers, and the buffer management routine exists.

D. Data Storage Routine

The data storage background task diagrammed in Figure 5-4 correctly
identifies a data buffer for disk storage and executes the FORTH buffer
storage routine. To avoid corruption of the disk as a result of unintentional
execution of this task, a task-active flag, which is set at the beginning of
an experiment, is used. A second flag' is set only while the disk save routine
is active. After this task exits, control is returned to the monitor routine.

The disk routine has complete control of the microprocessor while
information is transferred to disk; therefore, the RST5.5 interrupt must
be disabled at this point in the acquisition scheme. This results in the loss
of one to five data points whenever the disk routine is executed; the exact
number lost depends on the particular disk. interleave used. Interleave values
of four, five, or six give the fastest block access times on this system and
result in the fewest number of lost data points. Since each block contains
approximately 500 data points and an acquisition rate of two points per

second is obtained, one data point is lost every four minutes.

E. Data Disk Organization

On run termination, all emission data are saved to disk. In addition,
a parameter block containing all microprocessor-controlled variables, an
eighty character run descriptor, and a time stamp corresponding to the start
of the experiment are saved. Information which selects default data
conversion parameters, for use on the minicomputer, is also contained in
this block.

One of the major drawbacks to using the FORTH programming language

78

ocssom mcEmEm Sham .80 «2:5 32m

zmahum

 

<
l _
045.... .
022.53.:
m<w._o

 

    
 
   

 

rankzaoo-
xooam .
uk<aan

 

 

xooqm
rapzuo_

 

E 25$

 

I;

 

houqmm
Imaqk
z<u._o

 

 

 

 

 

 

 

0(4...
02.7534...
hum

 

79

for data acquisition is that by design, all data and software are saved in
blocks with no file based organization system; this requires the user to keep
track of where information is stored manually. The lack of file structure
is a particularly serious problem with data storage, since very littletext
is present with which to identify related blocks of data and/or experiments.
To alleviate this deficiency, a simple file structure for the data storage
disk was developed. In this protocol, the second block on a data disk is
reserved for directory information, which consists of the total number of
blocks used, the number of experiments currently saved on the disk, and
a directory entry for each experiment saved. The directory entry includes
a nine character file name and the first and last block numbers of blocks
which contain data from the corresponding experiment. A maximum of
sixty directory entries per data disk are permitted. In addition, theability
to interpret and display the parameter block for an experiment is available.
The incorporation of these and similar disk-based routines makes the current
system very amenable for use as an immediate archival storage and retrieval

method.

F. Minicomputer Programming Language Selection

Since the microcomputer system used does not support any hard copy
devices and high level, floating-point operations are difficult to perform
using the integer math functions available in the 'FORTH programming
language, data transfer to the LSI-11/23 minicomputer, where these
capabilities exist, was required. At this level, hardcopy printouts, high-level
computation procedures, and a versatile data plotting program are available.

The ideal programming language for the conversion process should have
several critical capabilities, including: floating-point calculation, disk file

access, high computation speed, and the ability to execute without repeated

80

user interaction. The choice of a suitable programming language on the
minicomputer was limited to BASIC, FORTH, FORTRAN-77, and MACRO-11.
BASIC, although versatile and easily programmed, was judged to be too
slow for the high volume of data associated with a typical experiment. The
version of FORTH available on the minicomputer did not offer --the
floating-point math functions required for statistical treatment of the data;
hence, FORTH was unsuitable at this level. MACRO-11, the LSI-11/23
assembly language, offers access to all system routines. This includes
floating-point capabilities; however, user access to the routines from
MACRO-11 is a tedious, error-prone programming exercise. FORTRAN-77
offers all of the features required of the conversion programming language
and was chosen to perform all data manipulations.

One weak point of FORTRAN-77 was that user interaction was typically
required for entry of several data manipulationparameters. One approach
to overcoming this difficulty is to access the command line used to initiate
the conversion software. The FORTRAN routines available to access this
information are awkward and difficult to use. MACRO-11 supports relatively
easy-to—use routines which access and parse the command line; hence, a
MACRO-11 subroutine was written to perform this function. Since the
MACRO subroutine uses the system routines for parsing the command line,
input of filenames and options conforms to the command line standards
established for the operating system. This feature allows specification of
all data conversion options on a single line. Furthermore, this line can be
specified in a command file to allow unsupervised conversion to occur.
Because command lines to initiate all programs associated withthe data
conversion and plotting routines are included, the user simply starts the

process and is free from the tedium of waiting for the conversion to terminate.

81

The user has an option to specify that the entire process be run when computer
usage is normally low so that the computer is free for userintensive tasks

such as editing and program development.

1. CVT - Data Conversion Software a

After experimental data are transferred to the minicomputer, they
are converted into a form readable by the plotting package available. In
addition, several options .exist which select specific, nondefault conversion
conditions, display experimental parameters and statistics, and enhance
the resultant plot. All options are accessible via the initial command line
and are selected by the inclusion of a two character mnemonic. Input and
output file specifiers are also entered in the command line; hence, all data
manipulation parameters .are selected on the single line used to initiate
the conversion program.

Data files present. on the minicomputer are of two major types. The
simpler of the two contains an experimentalparameter block followed by
data blocks. Each data block contains 509 intensity values. followed by the
wavelength setting and elapsed time associated with the last intensity value.
Enough data blocks are present to contain the entire experiment. This type
of data file is generated when the general acquisition routine is used and
does not allow for synchronization between data acquisition and variation
of a specific experimental parameter. The more complex data structure
also consists of the parameter block and data blocks; however, a specific
value for the abscissa and the standard deviation associated with each
intensity value are present. A synchronized acquisition run, such as a delay
or monochromator scan experiment, generates this type of data file.

The parameter block contains data descriptor information used by the

82

conversion routine to select the default conversion modes most appropriate
for the particular experiment stored in the data file; this allows for relatively
simple user interaction with the data manipulation software. The data
descriptor section indicates whether the file consists of intensity data only
or synchronously acquired data, the total number of data sets present, and
the length in bytes of each data set. In addition, flags indicating the
appropriateness of standard deviation or pooled standard deviation calculations
exist in the data descriptor. The parameter block also contains the specific
experimental parameters existent during the experiment; these include,
delay and integrate times, the initial wavelength, the number of sparks
averaged per point, and the gain used for the experiment. This information
is displayed on the user's terminal and optionally printed on the line printer
to permit complete, rapid experimental identification.

Manual override of the default conversion parameters is accomplished
by specifying the desired option in the command line. Other options available
include: abscissa selection, description of an appropriate output window,
baseline substraction, and inclusion of standard deviation calculations. After
the data conversion process was completed, the data sets were displayed
by the MULPLT plotting program written by Atkinson (103). Hardcopy plots

of the data were obtained at this level.

CHAPTER VI

SPARK SOURCE CHARACTERIZATION

The transient nature of the spark discharge requires the use of
time-resolved spectroscopy for discharge characterization. Previous work
on a similar spark discharge in argon has shown that the plasma is initiated
by the ionization of metastable argon atoms (104). Earlier studies of similar
spark discharge designs indicated that the spectroscopically useful time
window existed on the microsecond time scale after the cessation of current
through the plasma. Therefore, optical and electrical characterization of
the discharge were performed on the. microsecond time scale.-

Since the thyratron trigger synchronization pulse was used to initiate
the data acquisition process, the timing relationship between the leading
edge of the trigger pulse and actual spark ignition was determined. To
determine spark ignition, the output of the photodiode circuit (see Chapter
III) was monitored with a Tektronics 5403 single beam oscilloscope equipped
with a 5A48 dual trace module and 5842 time base.

A minimum lag time of approximately 0.3 us exists for all gap lengths
studied. In addition, the charging voltage required to attain the minimum
lag time increases slightly with increased gap length. These characteristics

indicate the lag time consists of two components. The more significant

83

84

component determines the minimum observed lag time and arises from the
thyratron breakdown time, the impedance characteristics of the high voltage
connection between the spark cathode and the thyratron anode, and
nonsimultaneous generation of the synchronization pulse and the thyratron
grid pulse. Walters (104) has observed a similar, although shorter, lag time
with his equipment and has found the lag time to be independent of anode
voltage, but influenced by specific thyratron characteristics. One attribute
affecting the lag time is the thyratron hydrogen pressure. Since the pressure
decreases with extended use, the lag time is expected to increase with
thyratron age. Other characteristics which affect the lag time include
thyratron geometry, cathode voltage, and high voltage transmission line
impedences.

The second factor determining the timing relationship between the
synchronization pulse and the spark formation is the actual breakdown voltage
of the spark gap. When a longer gap is used, a higher breakdown potential
is obtained. Nasser (105) reported that the rate of gap breakdown is dependent
on the amount of overvoltage across the gap. Hence, the longer gap length
requires a greater amount of overvoltage to produce the same spark breakdown
times. Two mechanisms are responsible for streamer formation and
subsequent breakdown. Near the breakdown potential, a Townsend mechanism
(105) is active. This requires the formation of a space charge region, followed
by electron transfer from the cathode to the anode with subsequent streamer
formation. This is a relatively slow process occurring on the microsecond
time scale. At higher overvolt potentials, a "streamer mechanism" is present.
In this mechanism, support gas ionization occurs near the anode. The
ionization process continues to self-propagate to subsequent layers of the

support gas until a streamer extends to the cathode. At this point, the spark

85

gap breaks down and the plasma is formed. At high overvoltage conditions,
this mechanisms occurs in 0.01 microseconds (105), which is negligible in
comparison to the constant lag time arising from the external electrical
characteristics of the system.

In all experiments, the quoted delay time presented is relative to the

trigger pulse.

A. Capacitor Charging Characteristics

In the early, free-running, double gap spark source design, the voltage
across the analytical gap and, consequently, the energy available during
the discharge, was determined by the breakdown potential of the control
gap. Therefore, the repetition rate was a function of the charging time
of the storage capacitor, which was determined by external factors such
as charging voltage, the current limiting resistance, capacitance, and the
longevity of conductive species in the spark gap. With the incorporation
of the thyratron switch into the discharge circuit, a constant repetition
rate is electronically generated and the charging time became independent
of these factors. Under these constraints, the effective spark gap voltage
is dependent on the time constant of the charging circuit and the repetition
rate.

To obtain optimum performance, the energy dissipated in the spark
should be as high as possible. Since the energy available for the discharge
is proportional to the gap voltage, knowledge of the gap voltage is useful.
The theoretical maximum for the gap voltage is derived from the calculated
value of the charging time constant and the experimental repetition rate.

The equation for the capacitance of a cylindrical capacitor is

86

C = Elf/2d

where E, the dielectric constant, equals 2.26 for polyethylene, l is the length,
r is the average radius, and d is the thickness of the dielectric. With this
equation, a capacitance of the storage capacitor of 300 pF is found. This
value, coupled with the current—limiting resistance, gives a time constant
of 300 us. Hence, a charging time of 690 us, or a repetition rate of 1.5 kHz,
is required to charge the capacitor to 9096 of the available voltage. Although
this constraint was recognized, a repetition rate of 2.2 kHz was generally
used, since this frequency is more conveniently and consistently available
from the thyratron trigger circuitry. Operation at this frequency corresponds
to a maximum gap voltage which is 78% of the charging voltage. The
calculated maximum and the observed gap voltage as a function of repetition
rates are shown in Figure 6—1. The slight negative deviation from the
theoretical maximum is probably caused by inaccuracies in the calculated
values of the electrical components.

Figure 6-2 shows the relation between the repetition rate and the
observed signal intensity for a 6000 V spark. The decreased signal intensity
at increased repetition rates is caused by the decreased gap voltage associated
with incomplete charging of the storage capacitor. The associated decrease
in the overvoltage across the gap causes an increase in the trigger jitter
of the spark. Since the slope of the charging curve is steeper at shorter
charging times, trigger jitter causes a greater variation in the gap voltage
at higher repetition rates and an increased noise in the observed signal
intensity.

The effect of operation at higher charging voltages is an increase in

the energy available for each spark. As shown in Figure 6-3, the signal

Gap Voltagc (x vcharg.)

100.

95.

87

  
 
  

{—— THEORETICAL

 

 

33, -L
' EXPERIMENTAL -—->
00, db
1 A' 1
75. I f U f ' I f I U fl U I I j I
1.0 1.5 2.0 2.5

Repetition Rate (kHz)

Figure 6-1 Gap Voltage as a Function of Repetition Rate

Relative Intensity (Z)

88

 

 

100.1-
90.--
; .
J
.1
BO.-#
fi
7O.db
I
1 l J, L j
60. T V i 1 I V V I 1 I r V U ' j V I T j
05 10 1.5 21) ‘ 25

Repltltion Rate ' (kHz)

Figure 6—2

Relative Intensity as a Function of Repetition Rate

Int. (0.11.)

Rel.

89

 

2500. q I-
J

2000. «I I-
.4

1500. -I i-

1000. q I-
1

500. - -
I

o. - MW

4000. 5000. 0000. 7000. - 8000.

Charging 'Voltage (V)

Figure 6-3 Relative Intensity as a Function of Charging Voltage

90

intensity is proportional to the charging voltage. In addition, a slight decrease
in the noise. level at 415 nm is observed at higher charging voltages. With
the more energetic conditions, a higher overvoltage is present across the
spark gap. Under these conditions, the spark fires more consistently relative
to the trigger pulse; hence, the same time window is monitored by the

acquisition electronics and a less noisy signal is obtained.

B. Effect of Gap Voltage on Current Waveform

Calkin (4) has monitored the discharge current ~ waveform for the
thyratron-controlled spark source; however, due to insufficient electrical
insulation in the chamber, determination of these waveforms under high
overvolt conditions was impossible. The current waveform is dependent
on the gap voltage; therefore, it can be affected, with the same result, by
modification of either the charging voltage or the repetition rate.

A series of current waveforms obtained at different charging voltages
is presented in Figure 6-4. The maximum discharge current as a function
of charging voltage is shown in Figure 6-5. The current waveform was
monitored with an IPC CM-lO-M current monitor which generates a 0.1
V/A response. The monitor output was observed on a Tektronics 5403 single
beam oscilloscope with 5A4B dual trace module and 5B42 time base modules.
The output was digitized and plotted.

These figures show that as the charging voltage is raised, a corresponding
increase in the discharge current is observed. In addition, the duration of
the discharge is increased slightly. The main discharge also occurs earlier
in time and is temporally more stable for the higher energy discharges. The
current waveform of the 8000V spark was reproducible within the limits

of oscilloscopic resolution; however, the initiation of the main discharge

Current (A)

Current (A)

91

 

 

 

 

 

 

 

 

 

 

- 5000 V
‘50.‘P Vc - 4000 V ‘50.}. Vc
zoo.{. ,‘ IOO.€-
i 3 4
i 1
' .E 50 d"
50.1. o ‘
4 t .
1 3 :
0.; " 0.1
3 1
-so.‘---------:..- ----.: ~50. -- -- --:-- - ---i1
0 00 0.25 0.50 0.00 0.25 0.50
Time (us) Time (us)
- 0 V I- 7000 V
150.1! Vc 600 V 150.1. C
I 1
100.1. .~
1 <
V
' "E
50.1- o
‘ t
1 3
0.‘
-so.' 4. ....... : ......... ‘
000 0.25 0.50

 

Time (ue) Time. (us)

vc - 5000 v

 

 

 

150.1-
4
100.eb
F‘ 4
< .
v 1
fl 1
c 50.1-
” 1
L 1
I...
3 i
O 0.-
l
-50 ‘-- - v .4
000 0.25 050

Figure 6—4 Discharge Current as a Function of Charging Voltage

Maxirnum Diecharge Current (A)

92

Discharge Current as a Function of Charging Voltage
140.--

 

 

 

‘0' I I I U j
3000. 4000. 5000. 0000. 7000. 0000.

Charging Voltage (V)

Figure 6-5

Maximun Discharge Current as a Function of Charging Voltage

93

pulse of the 4000V spark varied by as much a 0.1 us relative to the initial
thyratron conduction. This behavior is understood by the spark formation
mechanisms previously presented. The oscillatory behavior of the discharge
current is caused by capacitance and inductance characteristics of the high
voltage conductor between the spark chamber and the thyratron. Since
the thyratron acts as a unidirectional switch, no current is allowed in the
reverse bias direction. Walters (104) has performed much work in the area
of characterization of oscillatory spark discharges through modification
of circuit impedance components. He has found that by manipulation of
the discharge circuit impedance characteristics, discharge current waveforms

can range from unipolar to oscillatory in nature.

C. Background Emission Studies

To determine regions of possible spectral interference, background
emission spectra from 200 to 600 nm were observed at several delay times
for an argon discharge. In all studies, except where noted, the parameters
listed in Table 6-1 were used. The spectra obtained are shown in Figures
6-6 and 6-7.

The first significant feature observed in the spectra is the emission
from neutral carbon atoms at 247.8 nm. This emission was not reported
by Zynger (106) or Lantz (2) in the original spark chamber design; however,
Koeplin (3) and Seng (107) did report this emission and attributed it to tank-
impurities. Four possible sources of the carbon background signal are
outgassing from the Teflon components, contamination of the support gas
from residual oil or contaminants in the argon tank, aspiration of particulate
carbon from the gas purifying network, or leakage of atmospheric carbon

into the chamber.

94

Table 6-1

Typical Acquisition Parameters

Charging Voltage

Gap Length

Flow cell argon flow rate

Purge gas flow rate

Repetition rate

Slit width

Photomultipl ier voltage

Delay time

Integrate time

Sparks per point

Gain

6000 V

3 mm

250 ml/min
50 ml/min
2.2 kHz

70 u

900 V

1.0 us

4.0 us
1024

various

95

385 >200 m: 9798 59.2ng «o cofiocsm a ma :ommflEm 959.9.on mum oaswmm

int (ADC unite)

Int (ADC units)

int (ADC units)

000*

000¢

Chev face—go;

 

 

 

000 com 00¢ com com
P i P n b r b r b p b .
>200 n1 0;
com com oo.‘ com cow
3 P p D b . p P . y F p L
>2 . a: 0.0
000 com 00* com oou
- L . b n h h n b p p b »
>0. . n1 n.0

0
ON
0*
00
00
00v

ON
0*
00
00
00'

ON
0*
00
on
00..

(z) '1“! 'l98

(x) 'M was

(7.) 'iUi 'Iau

96

3E: >28 3 c.0798 598353 00 53055 a ma :oRflEm cczoamxowm

Int (ADC units)

int (ADC units)

Int (ADC units)

000

000+
000

 

 

 

 

 

Te 2&5
C55 595.263
oom oat com com
. _ p i . 1 c
on
oe
on
on
>300 «1 OJ 00.
com 00¢ com com
-- . . . . . .. i i Ii 0
on
0*
00
on
>030 n1 o.n 00"
com 00¢ con cow
r p . r . h L . L L p . . 0
0w
ov
on
on
>200 n1 0..N 00—

(x) '30: vs

(3) 'lul '13::

(x) '3": ’l‘ti

97

Removal of the purifying network had no effect on the peak intensity;
consequently, contamination must occur after the purifying network or the
contaminants are unaffected by the network. Since only 99.99% pure argon
was used for they support gas; moderate levels of hydrocarbons (maximum
4 ppm (108)) were present. Repeated baking of the Teflon chamber
components at approximately 230°C was performed to rid the Teflon of
any volatile contamination. During the first hour of baking, a brown substance
was formed which was partially volatilized after heat treating for
approximately 8 hours. This indicated the presence of unstable volatiles
on or in the Teflon, which either originated during the manufacture of these
components or was caused by adsorption of organic compounds during normal
use. After reassembly, a diminished yet significant carbon background was
obtained. Attempts to seal the spark chamber thoroughly from the atmosphere
had no effect on the observed background. signal; therefore, the carbon
background is probably due to outgassing from the Teflon and tank impurities.

Since the spark is most suitable as a gas chromatographic detector,
the carbon line is of interest when a universal detection mode is chosen.
Therefore, for optimum performance, this interference should be eliminated.
A background substraction scheme was used in this work to minimize the
carbon interference. Replacement of the Teflon insulation with a glass
or ceramic material should completely eliminate one source of this problem.

The second major feature observed in the background spectra is an
OH radical emission band centered at 310 nm. This molecular band arises
from three sources. The first source is any residual water vapor in the spark
chamber from assembly or from solution studies. Since the background
spectra presented were obtained approximately one week after some solution

work, this is the most likely source.

98

A second source of water is the possible diffusion of atmospheric water
vapor into the spark chamber. Since the spark chamber was well sealed
when the background scans were performed, this is. probably not the primary
source of the background water. '

Finally, Ascarite is used as a carbon dioxide scrubbing agent. Due to
the hygroscopic nature of NaOH, which is the active agent on the Ascarite,
adsorbed water vapor is most likely present. As the dry argon stream is
passed over the material, the adsorbed water slowly evaporates into the
support gas and is observed by the detector. This process appears to be
responsible for the increased hydroxyl emission intensity which was observed
when fresh Ascarite was used. Thus, the Ascarite appears to be the major
source of water when the detector is operated as a gas phase detector. This
problem could be minimized by drying the Ascarite immediately prior to
use.

The third and most obvious feature present in the background spectra
is the continuum radiation observed between 200 nm and 500 nm. There
are two factors which contribute to this continuum: bremsstrahlung radiation
and radiative recombination. Zynger has demonstrated, through electron
density calculations, that the high energy requirements of bremsstrahlung
radiation occur only within the first 0.1 to 0.2 us relative to spark ignition.
The radiative recombination reactions which occur later in time in the spark
environment diminish to acceptable levels within approximately 1.0 us. Prior
to this time a high-intensity, noisy spectrum is obtained.

The final feature observed in the spark background is the series of neutral
argon lines which occur primarily between 390 nm and 460 nm. These lines

have relatively long lifetimes of up to 40.0 us and are generally quite intense.

99

D. Effect of Oxygen on Carbon Emission

In an effort to counteract the troublesome background carbon emission
and clear the electrode of any carbon buildup, a small amount of oxygen
was mixed with the argon support gas. Figure 6—8 shows the effect of added
oxygen for several oxygen flow rates at three argon flow rates. In all cases,
the emission intensity decreases rapdily to a minimum between oxygen flow
rates of 45 and 80 ml/min, after which the intensity increases to levels higher
than the original. Oxygen strongly quenches atomic emissions; hence, the
initial drop in the signal occurs. Following the decrease, the emission intensity
rises again. Since the C emission was due to low level impurities, the emission
signal should become constant when a large excess of oxygen is introduced
into the instrument. This effect was not observed in this experiment. Three
possible molecular species have emission bands in the near vicinity of the
247.86 nm C emission. These are CO, CO+, and 02 at 248.38 nm, 247.42
nm, and 248.0 nm (109) respectively. Since the bandpass of the
monochromator was 0.14 nm, the detection of the CO and CO+ emissions
should be minimized. The 02 emission; however, falls within the bandpass
used. In addition, since oxygen is being added to the spark, the intensity

increase observed is due to the increased oxygen levels in the support gas.

E. Excitation Temperature Studies

One useful plasma diagnostic which was performed on the spark was
the determination of the excitation or spectroscopic temperature. This
calculation requires the derivation of a relationship between the excitation
temperature and the observed relative intensities. Line reversal and
absorption-emission methods are not used since absorption measurements

are much more difficult to perform on the spark. Therefore, a slope method

Relative Intensity (X Full Scale)

80

I

100

O - 1.5 l/rnin Argon
A - 2.5 l/rnin Argon
O I: 3.5 l/rnin Argon

71/

l _J

l I l
2°1lT1U'TTTU'ffiIU‘1IITIUUUUIIUV'1

O

100 200 300

Oxygen Flow Rate (mls/min)

Figure 6-8 Relative Carbon Emission Intensity

as a Function of Oxygen Flow Rate

101

relating the observed relative intensity and the excitation temperature is

used. Zynger (106) among others has derived the equation

log (1 A/g2A2?1) = log 0' - E2/2.303kTex

where I is the observed intensity, A is the central wavelength for an emission
originating in an excited state of energy E2 with a degeneracy of g2 and
a transition probability of A291, 0' is a proportionality constant, k is
Boltzmann's constant, and T is the excitation temperature. This equation
states that a plot of the lefthand side of the equation _v_s_. E2 results in a
straight line with a slope which is inversely proportional to the temperature.

When E2 is expressed in cm“1 the excitation temperature is given by

T = —hc/(2.303k (slope))

where h is Planck's constant and c is the speed of light.

Reif (110) has shown the validity of the slope method for obtaining
the excitation temperature. The requirements of the technique are that
no self-absorption occurs and that the spectral response of the optical
components are known. A constant spectral response can be assumed over
a narrow wavelength range; therefore, selection of a series of spectral lines
over a narrow bandwidth circumvents the second requirement. Knopps,
gt a_l_. (111) and Tourin (112) have found the argon plasma transparent below
650 nm; therefore, the self-absorption criterion is met in the spark.

Two major advantages of the slope technique are that no light source
other than the spark is required and there is no theoretical limit on the

applicable temperature range.

102

A series of seven neutral argon atomic emission lines located between
415 nm and 430 nm were chosen for this study. This spectral range is shown
in Figure 6-9. The transition characteristics, obtained from tables (113)

associated with these lines are presented in Table 6-2.

F. Effect on Time on Excitation Temperature

Spectra from 415 nm to 430 nm were obtained for delay times of 1.0,
1.5, 2.0, 3.0, 4.0, 10.0, and 20.0 us. A monochromator slit width of 25 um
and a scan rate of 0.05 nm/s were used. After baseline substraction, the
peak intensities were measured and plots similar to Figure 6-10 were
generated. The excitation temperatures were calculated from the slope
of the curve. Estimates of the error in the measurement were derived from
the baseline noise observed for the individual runs and the uncertainty in
the listed transition probabilities. The results are shown in Figure 6-11.
The 0.3 us measurement appears inordinately high and may be inaccurate
due to the random errors associated with the high noise level present in
this spectrum. A temperature of 5600 K was found at a delay of 1.5 us;
the temperature steadily decreased over the period studied. The values
obtained were approximately 800 degrees higher than those observed by
Zynger (106). This is due to the increased energy available in the spark

since overvolting of the gap is prevalent in this thyratron-controlled spark.

G. Effect of Charging Voltage on Excitation Temperature

For a second set of data, the charging voltage was varied while a constant
delay time of 2.0 us was maintained. Data were treated in the same manner
as outlined in the previous section. The results are presented in Figure 6-12.

Temperatures between 4800 and 5700 K were obtained. The highest

103

int (ADC units)

mos: commflEm come< 3.:st mo €0.50on

AEcV 50:20:;

mm¢

n D )‘l

0N¢

p

Te 9590

 

000¢ .-

 

 

 

I

 

 

 

 

 

 

 

>200 mi 0.0

 

 

 

 

 

 

 

 

 

ON

0*

00

cow

(x) '1‘” '193

104

Table 6-2

Argon Line Emission Parameters (113)

1(nm) E2(cm"1) g2 A291 (108 s‘l)
415.9 117184 5 0.0145
416.4 117151 3 0.00295
418.2 118460 3 0.0058
425.9 118871 1 0.0415
426.6 117184 5 0.00333
427.2 117151 3 0.0084

>430.0 116999 5 0.00394

LOG (I 0/92 A241)

105

 

 

 

6.8 w-
.1
6.7 m-
J
6.6 --
.1
6.54.:....:.- .4,
117000. 118000. 119000.

Energy (cm—1)

Figure 6—10 Excitation Temperature Slope Plot

‘Excitation Temperature (K)

8000. -r-

7000. --

 

6000. J-
4

5000. --

 

4000. r 1

Figure 6-11

106

11” _

= - -.J

T I I 1 r 1 I I I

5.0 10.0
Delay Time (us)

Excitation Temperature as a Function of Delay Time

Excitation Temperature (K)

 

107

 

 

6000. --
,
5000. ‘-
4ooo. .:4...:....:....:....:j.
4000. 5000. 6000. 7000. 8000.
Charging Voltage (V)
Figure 6-12 Excitation Temperature as a Function of Charging Voltage

108

excitation temperature was obtained for the highest charging voltage used.
Initially, the temperature appears to rise; however, little difference between
the 7000 and 8000 V conditions is observed. A higher level of noise in the
7000 and 8000 V spark indicates a more energetic environment; however,
this severely compromises the accuracy of the temperature determinations.
One explanation for the temperature rise is that the power dissipated in
the spark is proportional to the product of the current and the voltage. Since
both parameters increase when the charging voltage increases, a temperature
rise should be observed.

The excitation temperature obtained from the spark are somewhat
higher than the 4850 and 4920 K reported by Taylor, e_t a]: (114) and Fallgatter
(115) respectively for an atmospheric pressure microwave induced argon
plasma. This indicates that the spark should perform better than the

microwave induced plasma as an excitation source for gas phase samples.

H. Electron Density Measurements

Stark broadening is a major line broadening mechanism present in a
plasma. This broadening is caused by the electric fields of charged species
which surround the excited atom. Hydrogen atomic emission is subject to
a linear Stark broadening effect (116). In addition, pronounced broadening
on the order of a few nanometers is observed. Since the broadening is caused
by charged particles, the electron density in the spark can be calculated
from the width of a hydrogen emission line.

Figure 6-13 shows the Stark broadened H8 line at 486.1 nm. A delay
time of 0.3 us, a spark charging voltage of 6000V and a slit width of 50 um
were used for this spectrum. A total of 8384 sparks were averaged into

each data point. The argon support gas was bubbled through distilled water

Relative intensity (a.u.)

109

 

 

 

1500 -,

1000 --

500 --
0 , j 1 0 A, T . . . %
480.0 485.0 490.0

Wavelength (nm)

Figure 6-13 Stark Broadened H8 Line Profile

110

to supply hydrogen to the spark.

Through the Kolb and Griem (117) theory of Stark broadening, the Stark
width of a hydrogen line, when raised to the 3/2 power, is proportional to
the electron density. The proportionality coefficient is slightly dependent
on the electron temperature. Since this parameter is unknown, coefficients
tabulated by Griem (116) for several electron temperatures were averaged.

The equation

N6 = 1.12 x 1016 W15

expresses the relation between the line width, W, expressed in nm and the
electron density, Ne, expressed in electrons/cm3. Figure 6-14 shows the

relation between the calculated electron density and the delay time.

1. Determination of Ionization Temperature

To measure the ionization temperature of the spark, the intensity ratio
of the 393.3 nm Ca(II) line and the 422.7 nm Ca(I) line was obtained while
a 50 ppm solution of Ca was aspirated into the nebulization/desolvation
system. Thirty data points were collected at each delay time at both
wavelengths and averaged. The standard deviation of the averaged intensity
was used as a measure of the error. Figure 6-15 shows the decay of the
Ca atomic and ionic emission lines as a function of time.

The Saha equation describes the relation between the degree of ionization,
the ionization temperature (T), and the electron pressure (Pe). Using Boumans'

(119) notation, the intensity ratio of an ion-atom pair is

111

 

 

 

5.0 -P
6‘ 4.0 «l
E .
o .
\
,, .
C -1
3 3.0 q-
3 9
_g .
3’ a
.1
2‘3 2.0-'1'
o I
\ , ‘
"-1
z” 1.0-:-
1 1 I n 1 n
0.0 ........-1f...,.-..,..-.1
0.0 1.0 2.0 3.0 4.0 5.0

Delay time (us)

Figure 6—14 Electron Density as a Function of Delay Time

intensity (a.u.)

10.00

8.00

6.00

4.00

. 2.00

0.00

112

50 ppm Co. 4.0 [1: integration

Ca(ll)

Ca(l)

 

T I I I I I I I I I I j I W TTI # T I ' r I!“
0.0 1.0 2.0 3.0 4.0 5.0

Delay time (p3)

Figure 6-15 Calcium Emission as a Function of Delay Time

113

log (I+/l) = -log Pe + log g+A+2_)1 A /gA2_)1 +

—5040(vij + v"q - Vq)/T + 5/2 log T - 6.18

where the superscript '+' refers to the ions, Vq is the excitation level from
which the excitation occurs, and Vij is the apparent ionization potential
for the atom. According to Boumans, the apparent ionization potential
for Ca is 5.83 eV. Values of 1.0 and 0.91 were reported by Corliss and Bozman
(120) for the gA values for the atOmic and ionic emission lines. Excited
state energy levels of 2.93 eV and 3.15 eV for the atom and ion were obtained

from spectral tables (121). The electron pressure is
re = NeT/7.34 x 1021

With the electron densities calculated in the previous section, equations
6—5 and 6-6 were solved recursively on the minicomputer for T at each delay
time. The results of these calculations are presented in Figure 6-16.

In comparison to Zynger's characterization of the single gap system
(106), the time averaged ionization temperature for the present design is
approximately 600 degrees lower. This is probably due to more complete
desolvation with the desolvation chamber Zynger used. The increase in
the quantity of water vapor would require more energy from the spark for
desolvation and excitation. This, in turn, lowers the temperature of the
spark. I

Since the excitation temperature was determined in a pure argon stream

and the ionization temperature measurement was performed in an argon

Ionization Temperature (K)

 

114

 

 

5000 .
0.0

Figure 6-16

1.0 2.0 3.0 4.0 5.0

Delay Time (us)

Ionization Temperature as a Function of Delay Time

115

stream which contained some water vapor, a comparison of the two
temperatures is difficult; however, some conclusions can still be made from
this information. Even under the energetically less favorable conditions,
the ionization temperature was found to be approximately 1000 K higher
than the excitation temperature. This indicates that thermodynamic
equilibrium is not achieved in the spark. Similar characteristics were found

in the original single gap system of Zynger.

CHAPTER VII

CHROMATOGRAPHIC EVALUATION

A. Introduction

Lantz, Koeplin, and Calkin (2-4) have demonstrated the element-selective
detection capabilities of the spark source for use with packed column gas
chromatography. Their work has been concerned with the detection of
metallic and non-metallic elements in CC effluent and has demonstrated
the visability of spark detection for several elements. Detection limits
in the ng/s range were determined for many elements. In recent years,
capillary gas chromatography has been applied to a wide range of sample
analyses. Because the capillary media demonstrates extremely high
theoretical plate counts, the resolution of chromatographic peaks is much
higher than with traditional packed column technology. In conjunction with
the increased separation power of capillary GC, an increase in the quantity
of discernable peaks and, hence, an increase in the complexity of peak
identification was inescapable. One approach to simplifying the chromatogram
is by use of a detection method which detects only the component(s) of
interest. Ideally, the chromatographer could specify the peak of interest
and the detector would respond quantitatively to that component and have

no response to any other constituent. One method which approaches this

116

117

ideal is element-selective detection. In this method, the detector responds
preferentially to those components which contain the element of interest
and remains nonresponsive to other species. By using atomic emission
spectroscopy, the element of interest is detected by adjustment of the
observation wavelength. Considering the broad range of applicability of
emission spectroscopy, the extension of the spark source to the detection
of capillary gas chromatographic effluents would simplify the interpretation
of complex chromatograms.

This chapter consists of an evaluation of this capillary GC-spark
interface, a discussion of the universal detection capabilities of the spark,

and an application of element—selection to the analysis of leaded gasoline.

B. Evaluation of the GC-Spark Interface

One of the major advantages of capillary GC is that band broadening
is kept to a minimum, since column volumes are extremely small. Hence,
one requirement of the detector is that the high resolution not be compromised
by the detector characteristics. To this end, the effective volume of the
detector must be kept to a minimum. In the detector interface designed
for the spark (Figure 3-5), the capillary column is threaded through a 15
in. x 1/4 in. o.d. glass tubing which is wrapped with heating tape. The heating
tape was used to maintain temperatures above that required for the particular
separation study. One end of the interface protrudes approximately 1/2
in. into the GC oven. The other end of the interface slides through a Teflon
positioning collar and protrudes approximately 3/4 in. into the flow cell
in the spark chamber. At this end, the interface consists of two concentric
tubes. The column passes through the inner, 1/8 in. o.d. tube, and argon

plasma support gas is supplied through the outer tube. A low argon flow

118

rate of 0.25 1/min assures a non-turbulent gas flow out of the end; of the
interface. This acts to restrict the diffusion of the effluent stream throughout
the flow cell. Manual positioning of the end of the column to within 2 mm
of the spark gap minimizes the effective detector volume to approximately
1.6 n1 for an 0.32 mm i.d. capillary column. Arcing to the end of the column
at shorter separations prevented further reduction in the detector volume.
An alternate aluminum interface was constructed to take advantage of its
increased thermal conductivity and less fragile nature; however, occasional
arcing to the interface was observed. This created a potential electrical
hazard to the operator and the equipment; hence, the glass interface was
preferred.

To determine the effect of the interface on column performance, the
resolution of the spark system was compared to that of the flame ionization
detector (FID) supplied with the GC. The resolution, R3, of two adjacent

chromatographic peaks is given by the equation:

RS = “R2 - tR1)/0.5(tw1 + th)

where the subscripts l and 2 refer to the earlier and the later eluted
components respectively, tR is the retention time of the peak, and tw is
the width at the base of the peak (122). For the determination of the
resolution, a chromatogram of a jet fuel sample was obtained with both
the FID and spark detectors, using the conditions shown in Table 7-1. Figure
7-1 shows the chromatogram obtained with the FID and the spark detector.
The two peaks labeled in the figure were used to determine the resolution
for each case. Values of 1.4 and 2.2 were obtained for the resolution for

the FID and spark chromatograms respectively. This indicates that band

119

FID
JET FUEL A

 

 

 

 

 

 

 

 

 

 

 

MSS
JET FUEL A

 

 

 

 

 

 

 

 

Figure 7-1 Comparison of FID and NSS Detection

Table 7-1.

Jet Fuel Gas Chromatographic Conditions

Column

Sample

Sample Size
Helium Flow Rate
Initial Temperature
Initial Hold Time
Program Rate
Final Temperature
Final Hold Time

FID Parameters
02 Flow
H2 Flow
Helium Makeup Flow
Attenuation
Range

Spark Parameters
Ar Makeup flow
Ar Sheath Flow
Interface Temperature

. Gap Voltage

Gap Length
Observed Wavelength
Slit Width
PMT Voltage
Delay Time
Integrate Time
# Sparks per Point
Gain

30 m x 0.32 mm
0.25 11 DB-l

1.0 96 Jet Fuel
1.0 pl

50 cm/sec
30°C

0 min

4 deg/min
150°C

10 min

300 ml/min
30 ml/min
30 ml/min
16

10-11 A

1.0 l/min
250 l/min
240°C
6000 V

3 mm
247.8 nm
125 pm
1000 V
0.9 us

4.0 us
1024

250

121

broadening was lower with the spark than with the FID. This is possible
since the FID associated with the GC was not originally designed for use
with a capillary column. As such, a commercially available detector insert
(SCE DC-8) was required for capillary compatibility.

Since the sample inlet for the FID was oriented at right angles and
approximately 2 cm below the base of the FID, a relatively large detector
dead volume of approximately 18.0 n1 was unavoidable. Considering this
aspect of FID, the higher resolution of the spark is understandable.

The second important criterion for the interface is the capability to
attain acceptably high temperatures to prevent condensation of those
components which elute at high temperatures. If condensation were a
problem, the detector response would decrease and peak widths would increase
for high temperature components. In practice, interface temperatures of
250 :I: 10°C were maintained throughout the chromatographic run. Again
a comparison between the two chromatograms shown in Figure 7-1 indicates

no loss of chromatographic quality with the spark detector.

C. Universal GC Detection Capabilities

Most GC detectors in general use are classified as universal detectors,
since they respond to all compounds with few exceptions. The FID and,
in one mode of operation, the electron capture detector belongs in this
category. The spark detector can be operated in this mode as well by
monitoring the neutral carbon emission at 247.8 nm. Lantz (2) has found
detection capabilities for several compounds are independent of the functional
groups present in the compound. This includes carbonylic and nitrilic carbon
which demonstrate poor detectability in the FID. Detection limits of 2

x 10‘8 gs"1 with linearity over two orders of magnitude were reported by

122

Koeplin and Calkin (3,4) for packed column chromatography.

To determine the detection limit for carbon for the capillary system,
a series of n-alkanes in hexane was chromatographed with the conditions
shown in Table 7—2. A representative chromatogram is shown in Figure
7—2. The undecane peak was digitally integrated. The resultant integrated
peak intensities were used to generate the calibration curve shown in Figure
7—3. As in the packed column studies cited, a linearity of two orders of
magnitude was observed; however, a detection limit of 4.0 i 0.5 x 10'8 gs”1
was obtained.

The presence of the background carbon signal made carbon determination
less accurate and sensitive than would otherwise be expected. The high
background and the accompanying baseline noise made observation of small
analyte signals difficult even though electronic baseline subtraction was
enabled.

An alternate explanation for the slight decrease in sensitivity concerns
the time scale of capillary GC. Typically, peak widths in packed column
chromatography are approximately 20 to 30 s for major components. However,
with capillary technology, peak widths of under 5 s are common. Since the
spark is a discontinuous plasma, sharp peaks can be undersampled or aliased.
Figure 7-4 shows typical peaks obtained with capillary and packed column
gas chromatography. The vertical dashed lines indicate possible times when
the peak is sampled. For the packed column, this sampling has no effect
on the observed peak height since the top of the peak is broad with respect
to the sampling rate. However, if the top of the capillary peak is narrow
in comparison with the sampling period, loss of accurate peak reproduction
occurs. Expressed in terms of the Nyquist constraint for sampling, the Nyquist

frequency of the packed-column peak is much lower than that of the capillary

Table 7-2.

Carbon Standards Gas Chromatographic Conditions

Column

Sample

Sample Size
Helium Flow Rate
Initial Temperature
Initial Hold Time
Program Rate
Final Temperature
Final Hold Time

Spark Parameters
Ar Makeup Flow
Ar Sheath Flow
Interface Temperature
Gap Voltage
Gap Length
Observed Wavelength
Slit Width
PMT Voltage
Delay Time
Integrate Time
# Sparks Per Point
Gain

30 m x 0.32 mm
0.25 p DB-l

Neat CnH2n+2 Mix
1.0 111

50 cm/sec

25°C

5 min

4 deg/min

200°C

0 min

1.0 1/min
250 l/min
250°C
6000 V

3 mm
247.8 nm
70 um
900 V

0.4 us
10.0 us
1024

250

124

 

EwewoquoEO 25:01: wingswmm

AEEV oEr— cozcoaom

 

 

 

 

 

 

y-CD

 

[UUTI'IIjI'IIUT'IIII'WUUI'

000

000w

000w

000w

000w

coon

(°n'o) Kigsuaiul engiolaa

Integrated Intensity (a.u.)

125

3000.

1000.

IllLILIAIJIIIJJILJJILL14l141IJ

 

 

 

Carbon present (pg/s)

Figure 7-3 Carbon Calibration Curve

126

295m com—33005 xwmm 25020308200

     

cEJBo

vmxoom
_occhm>coo

 

I. 2:90

 

 

:EJBU
>Lo:_aoo

—_-——_-————————A_——_—————_-——

-___————_—————_-——-———_———

 

127

peak, hence, a higher minimum sampling rate is required for capillary
chromatography. Since the sampling rate for the spark was 2.2 kHz, Figure
7-4 exagerates the severity of the problem for data acquisition. However,
the data display update frequency is only 2 Hz; hence, some biasing of the
chart recorder output does exist. Because of this, the digitized peak areas
are accurate representations since they are a time averaged signal over
the entire peak even though peak height information exhibits some biasing.
A similar evaluation by Holland _e_t_ a]: (61) concerning capillary GC-MS data
acquisition concludes that a sampling rate of two and preferably five data

points per second is sufficient for regeneration of a chromatographic peak.

D. Element-Selective Detection of Lead in Gasoline

The major benefit of using an atomic emission technique for
chromatographic detection is the associated element-selectivity. Previous
work has demonstrated the detectability of several elements which include
Ca, Si, B, Al, P, Cu, Cr, Br, I, N, and O. The current work applies the capacity
for element-selective detection to the analysis of Pb in gasoline.

The currently approved standard method for the determination of lead
in fuels consists of a sample preparation step followed by lead detection
by atomic absorption. The preparation step consists complexation of the
lead with iodine, stabilization of the alkyl lead iodide with
tricaprylmethylammonium chloride, and a 10-fold dilution with methyl isobutyl
ketone (123). The preparation step is time-consuming since the reaction
between the iodine and lead is slow. In addition, it has been reported that
the preparation step is not highly reproducible (124). In an attempt to avoid
the slow preparation step, gas chromatographic analysis of a neat gasoline

sample was performed.

128

1. Lead Calibration Curve

Gasoline additives, the most common being tetraethyl lead, are used
to increase the octane rating of the fuel. An increase in the octane rating
decreases the tendency of the fuel to knock or ignite before a spark is fired.
One common property among the antiknock agents is that they possess a
high boiling point in comparison to the remainder of the fuel components.
This information allows chromatographic conditions which simplify and
accelerate the determination to be selected. Specifically, a high initial
temperature is selected to rapidly elute most fuel components. The
chromatographic conditions used for all lead standards and gasoline samples
are shown in Table 7-3.

A series of standard lead samples were prepared by dilution of the 0.537
jig/111 (1.994 g/gal) lead reference mixture (NBS standard reference material
1636) with n-hexane. These standards are a mixture of 91% by volume
2,2,4-trimethylpentane and 9% by volume n-heptane (125). Tetraethyl lead
supplies the analyte concentration.

A lead standard chromatogram acquired in both the universal carbon
mode detection and the element-selective mode for lead is shown in Figure
7-5A and B respectively. The carbon mode shows two distinct features.
First, a large signal centered at 2.2 min corresponds to the coelution of
the 2,2,4-trimethylpentane and heptane. Second, a smaller peak occurring
at 6.1 min is observed; this is the tetraethyl lead. The chromatogram obtained
while monitoring the Pb(I) emission at 405.7 nm also shows features
corresponding to the solvent front and the lead compound; however, the
solvent front in this mode causes a negative deviation in the baseline. This
response is caused by cooling of the spark by the entire plug of solvent and

has been observed in all previous versions of the nanosecond spark source

Table 7 -3.

Lead Standards Gas Chromatographic Conditions

Column

Sample

Sample Size
Helium Flow Rate
Initial Temperature
Initial Hold Time
Program Rate
Final Temperature
Final Hold Time

Spark Parameters
Ar Makeup Flow
Ar Sheath Flow
Interface Temperature
Gap Voltage
Gap Length
Observed Wavelength
Slit Width
PMT Voltage
Delay Time
Integrate Time
# Sparks Per Point
Gain

30 m x 0.32 mm
0.25 u DB-l

Neat Pb standards
1.0 111

50 cm/sec

90°C

5 min

4 deg/min

140°C

0 min

1.0 1/min
250 m1/min
250°C
6000 V

3 mm
247.8 nm
70 um
900 V

0.4 vs
10.0 us
1024

250

130

EwLwaEoEO 030:me 00mg mus opsmrm

AEEV 0E2. cozcoaom

 

as 3... 9m 0.... ad .od 9. o.o
.7 b r bL - n b P n - p L P P -y b P P h — b b h . - h LP 1r P b h b E b _ no
.oom
.83
.89
E: 02: S 8.5.5 San... .ooou

0.20 me: 80:23.

OK. 0.0 0.0 0.¢ 0.0 0.N 0; 0.0

P10 P Flint-11h Phrr-(er r—L b hr?- p'lblP- Pb F b P» h'P PI— 00

 

.oom
.
1 door
.
n
n1 .000—
E: 6.5% so 5.3.5 So u. .ooou

(°n°o) 8”] .138

('fl'D) .1,” .138

131

(2-4). The tetraethyl lead peak in the lead mode is shifted approximately
0.1 m relative to the peak observed in the carbon mode. The shift is caused
by a lack of reproducibility in the carrier gas flow rate and variation in
the temperature program initiation timing.

A series of standard lead mixtures were separated and the lead peak
was digitally integrated. The calibration curve shown in Figure 7—6
demonstrates the results obtained. The highest reference concentration
available was 0.537 ug/ul lead; hence, the upper limit of the calibration
curve was not determined. Under the chromatographic conditions studied,
a detection limit of 7.5 i 0.8 ngs‘1 was found. At this mass flow value,
a signal equal to two times the background noise was obtainable. For the
above determination, the detection limit corresponds to a mass of 0.15 :1:
0.016 ug or a concentration of 0.55 :l: 0.06 g/gal.

Figure 7-6 also shows that the calibration curve does not pass through
the origin but has a negative y-intercept. An explanation for this
characteristic is that the spark is cooled by the presence of the analyte.
This lowers the background level as the peak elutes and causes the acquisition

of a dimished intensity value.

2. Lead Determination in Gasoline

Carbon mode chromatograms of leaded gasoline are similar to the lead
standard chromatogram. However, because gasoline is a complex mixture
of several compounds, the early portion of the chromatogram demonstrates
a response to a series of components. These components are not well resolved
for two reasons. First, the chromatographic conditions were optimized
for the determination of high-boiling components; hence, the majority of
the gasoline (components are rapidly eluted. Second, a large injection volume

is required to adequately determine the lead-containing species. This resulted

132

02:0 0030,5300 0004 0A. 9503

3.0 638?. an. 30:25
so so to 3 «d to ad

.PPDI-bbnD-b-DI-DDPP-DPDI-BDPb-DPbL-DDIL-Dbbb-Db-b-DDDP-DPDP CO

I .mN

l .00

.1 .mn

 

 

 

‘dddd‘dddddddddfidI-J‘fidequd11A-d04¢.ddld-dfidd 00°F
0.N mé . 04 0.0 0.0
Cam\3 0o$abcoocoo an.

(mm) :3er noed BAHDISH

133

in a column overload for the low—boiling, major components. In addition
to the response for low-boiling components, a single, small peak is observed
at the retention time for tetraethyl lead. The chromatogram for leaded
gasoline observed while monitoring the lead emission line is similar to that
obtained for the lead standards except that the initial negative response
to the solvent front is slightly longer-lived for the gasoline studies. This
is because a longer time is required to elute all of the low-boiling components
of the gasoline. As was true of the lead standards, a single positive response
is obtained at the retention time for the tetraethyl lead standard. Digital
integration of the peak and correlation with the previously presented
calibration curve indicated a lead concentration of 290 t 35 ug/g, or 0.76
1: 0.09 g/gal. Other researchers (126,127) have reported values ranging from
0.01 g/gal to 1.9 g/gal for gasoline from various sources. In comparison
to these values, those determined by spark detection are reasonable.

Although the detectability of lead has been demonstrated in this work,
the detection limit obtained with these conditions is not comparable to the
0.002 jig/gal obtained with the standard method. However, much higher
capacity capillary columns have been developed since these experiments
were performed. Preliminary experiments with such a column indicate that
injection volumes of up to 4.0 ul do not seriously degrade the chromatographic
performance. This would improve the detection limits by a factor of four

or more since higher injection volumes could be used.

CHAPTER VIII

CONCLUSIONS AND PERSPECTIVES

The current work has shown that the thyratron triggered spark is
temporally stable and reproducible. The arcing problems noted earlier have
been eliminated with the present, insulated chamber design. A lower level
of radio frequency interference is attained with this design since more
complete shielding of the high voltage path was incorporated. This permitted
increased microcomputer reliability. These modifications have made spark
operation at higher gap voltages possible, and a more energetic plasma is
generated. Associated with the higher gap voltages are increased spark
currents and improved temporal stability of the discharge, this combination
produces larger signals and decreased noise. The coaxial capacitor design
allows a rapid discharge to occur; hence, the spark can be classified as a
high power discharge. The incorporation of a complete, menu-driven,
microcomputer software package for data acquisition, coupled with
improvements in the reliability of the spark, have made the spark system
significantly easier to operate than previous installations. The implementation
of the FORTH operating system for data acquisition and control functions
has allowed comparatively rapid development of a complex, versatile software

package. In addition, the use of FORTH should allow future software to

134

135

build on the current software package. Several complex, high speed timing
operations were performed by a single, large scale integration, integrated
circuit. This device should be seriously considered in any future
microprocessor-controlled timing requirements.

Future work on the spark system itself should center on several areas.
The spark is a high power discharge; however, due to the short discharge
duration, the energy dissipated is only 60 to 100 W. Development of a high
energy spark should give more immunity to solvent effects and permit more
reliable background level determinations to be made. With the incorporation
of thyratron control over the spark production rate, higher gap voltages
have been attained; however, the gap voltage is now, in part, determined
by the available charging time derived from the firing rate. Efforts to
decrease the time constant for the charging cycle should be made to allow
capacitor charging to proceed to at least two time constants. This should
improve the signal-to—noise ratio. Further work on improving the positional
stability of the spark is required, as this appears to be the major source
of noise in the signal. Two possible approaches are increasing the firing
rate as described by Coleman and Walters (91) or providing an axial argon
flow along the electrodes (128). Both of these methods have met with success.
The former again requires modification of the charging characteristics to
permit adequate performance. The latter may be accomplished through
a redesign of the purge gas inlet and flow cell. The current waveform studies
have indicated that triggering of the data acquisition electronics from the
trigger synchronizationpulse is not particularly precise on a spark-to-spark
basis at low overpotentials and demonstrates small time lag until spark
ignition. This time lag is expected to increase with the age of the thyratron.

As such, a return to the use of the photoiode trigger is appropriate.

136

The spark was found to perform adequately as a capillary GC detector.
The interface which was developed worked well and demonstrated no adverse
band broadening characteristics. The ability to position the capillary to
within 2 mm of the spark was necessary for minimizing the effective dead
volume. The capillary GC detection limit for carbon is comparable to the
packed column value obtained with previous spark source designs. To improve
this, the carbon background signal must be minimized. The use of ultrahigh
purity argon would eliminate most of the background; however, a drastic
increase in spark operating expenses would be incurred. Replacement of
the Teflon flow cell with one of a non-carbonaceous structure would also
improve results.

Element-selective chromatographic detection was shown to be a viable
means of simplifying component identification. In addition, coupling of
chromatographic techniques with element-selective detection can be used
to minimize or eliminate time-consuming sample preparation procedures
in some cases.

Improvements in the capillary GC/spark detector should focus on ways
to increase the quantity of analyte sampled in each spark. Preliminary work
with a currently available high-capacity. capillary column indicates that
sample volumes can be 4 to 10 times greater than those used with the original
column. This will cause a corresponding increase in the detectability of
individual components. Axial introduction of the effluent instead of the
current transverse introduction mode may further increase detection
capabilities (107) since the entire gap could be filled with effluent.

Studies with partially desolvated aerosol streams indicate that sufficient
energy to desolvate, atomize, and excite many species at low levels is present.

The major problem with this is that too much energy is required to atomize

137

the large excess of solvent molecules. This effectively cools the spark to
an analytically marginal level. More complete desolvation than the
miniaturized nebulization/desolvation chamber was capable of producing
is necessary for a solution interface to the spark. To attain the goal of
an HPLC interface, a very low effective dead volume must be obtained.
One promising technique involves an introduction scheme based on isolated
droplet generation developed by Hieftje (129). In this technique, a liquid
stream is forced through a fine bore capillary. The capillary is physically
attached to a bimorph transducer which is electrically oscillated. The
oscillation imparts a longitudinal wave onto the emerging stream. The
mechanical perturbation of the stream causes it to break into a series of
equally spaced droplets. Once the droplets are formed, no mixing with
adjacent drops can occur. Thus, the dead volume is limited to the volume
of a single droplet, or a few nanoliters. By electrically charging some of
the droplets and applying a voltage across a pair of deflection plates it is
possible to remove selected droplets from the stream and reduce the
introduction rate.

In this scheme, a single drop could be selected to pass into the gap,
and nonselected drops could be deflected to a trap. Synchronization of spark
firing to the presence of a droplet would then vaporize, atomize, and excite
only the molecules in the droplet. Since a droplet pulsing rate of
approximately 20 kHz is common, 90% of the solvent and analyte are
immediately eliminated. Additional desolvation steps could further minimize
the amount of solvent reaching the spark. In preliminary work by Wiegand
(130), individual droplets were successfully passed through the spark gap.
This droplet generation may prove to be the long-sought HPLC interface.

The present detector is limited in practicality since only a single emission

138

at a time can be acquired. Implementation of an optical multichannel analyser
(OMA) on the spark system would vastly improve the throughput of
multielement detection. Rapid gating of the detector is necessary to reduce
the background continuum. This criterion is met by the Princeton Applied

Research OMA III (131).

681

'ZOI-L6

 

‘oz ‘zsst ‘IOS ‘womo 'r '0 ‘soaanuo f'V'H ‘uanas 9T3 ‘uosaaned

 

ms ‘81 ‘7161 '10s 'woulo 'r 'P‘JJOHOSIH Pa ‘9on

 

'8661 ‘81? ‘ILGI 'UJallO 'IBUV "I'EI ‘KIIGX 9V ‘UGUJBX

 

'191 ‘1. ‘6961 ‘IDS 'woauo T '1' ‘xeuer 2’w ‘JaIssa-Ia
'18‘991

‘LLGI °u101q3 '1‘ '3 ‘SJBBOH é'g ‘ueiuowes £°p ‘XOIBIQ é-f ‘ziimoanq

 

'188 ‘9 ‘L961 “HO-NO 599 '1’

'00 ‘818011 5'10 ‘Bumms f'oa ‘aipum. f'M'O ‘axaueo 5'V'M ‘env

 

'991 ‘I. ‘6961 "PS 'woauo °r ’H'O ‘UBWIJBH
’661‘981‘LL61 means '1' 'a ‘alqwif'w ‘1“93J385'V'H‘SSGIPOOH

‘8 ‘617‘9961 oxovr 'O'G ‘xOEMlsoa 5'.~.I'N ‘saAI 5"1 ‘89101109

 

“SH-681 ‘9 ‘SLGI Biudmfiomwoauo T 500438

 

'688 ‘9 ‘L961' 'wOJUO 399 'P '21 ‘JamaJO

 

'699 ‘88 ‘1961 Meal 'BUI walls '3 ‘91010098 2'1 ‘sumx 5'3 ‘Jewaao
’1861 m

‘BUISUF'I 18921- =KuSJeAiun 9191s uvfiiuoiw ‘sysaui am "1'0 ‘umvo
82.61 M

‘Bugsuaq 1863- :KusaeAiun 91813 uefiiqoiw ‘sisqu, °q'qd °w°s ‘undaox

LL6I 1W

‘Bugsuaq 1363 :KusaaAgun 91mg uangogw ‘SlSOlLL 'CI’lld 31°}; ‘ziueq

 

170173-8697. ‘99 ‘8861 'weuo 'IBuv 'x ‘HOUIUBN "r ‘qooar ‘9 ‘Jawwi-IE)

SHONHHHJHH

'91

'91

'VI

'81

'Z1

'11

'01

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

140

Bombick, D.; Pinkston, J.D.; Allison, J. Anal. Chem. 1984, _5_§, 396—402.-

 

Aue, W. J. Chrom. Sci. 1975, 13, 329.

 

Versino, B.; Vissers, H. Chromatographia 1975, 8, 5.

 

Getty, R.H.; Birks, J.W. Anal. Lett. 1979, 12, 469-476.
Aue, W.; Flinn, C.G. J. Chrom. 1972, 142, 145.

Jackson, J.A.; Blair, W.R.; Brinckman, F.E.; Iverson, W. P. Envir.

Sci. Techn. 1932, 16(2), 110-159.

Sutton, D.L.; Westberg, K.R.; Melger, J.E. Anal. Chem. 1979, 51,

 

1399-1401.

Burnett, C.H.; Adams, C.F.; Farwell, S.O. J. Chrom. Sci. 1978, 16,

 

68-73.

Kelly, T.J.; Gaffney, J.S.; Phillips, M.F.; Janner, R.L. Anal. Chem.

 

1983, 55, 135-138.

Nelson, J.K.; Getty, R.H.; Birks, J.W. Anal. Chem. 1983, 55, 1767-1770.

 

Yamada, M.; Ishiwada, A.; Hobo, T.; Suzuki, 8.; Araki, S. J. Chrom.
1982, 238, 347-356.

Patterson, P.; Howe, R.; Abu-Shumays, A Anal. Chem. 1978, 50,

 

339.

Karasek, F.W.; Onaska, F.I.; Yang, F.J.; Clement, R.E. Anal. Chem.

 

1984, 56, 174R-199R.

Corkill, J.A.; Joppich, M.; Kuttalo, S.H.; Giese, R.W. Anal. Chem.

 

1982, 54, 481-485.

McNair, H.M.; Bonelli, E.J. "Basic Gas Chromatography", Varian
Instruments: Berkeley, CA, 1969; p. 118.

Karasek, F.W.; Field, L.R. Res/Dev 1977, 42.

Aue, W.A.; Siu, K.W.M. J. Chrom. 1982, 239, 127-144.

Sullivan, J.J.; Burgett, L.A. Chromatographia 1975, 8, 176.

 

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

141

Grimsrud, E.P.; Knighton, W.B. Anal. Chem. 1982, 54, 565—570.

 

Parris, G.; Blair, W.; Brinckman, F. Anal. Chem. 1977, 49, 378.

 

Seger, D. Anal. Lett. 1974, 7, 89.

Andree, M.O. Anal. Chem. 1977, 49, 820.

 

Wolf, W. J. Chrom. 1977, 134, 159.

Dumarey, R.; Dens, R.; Sandra, P. J. High Resolut. Chrom. 1982,

 

5, 687-689.

Chau, Y.K.; Wong, P.T.S.; Saitoh, H. J. Chrom. Sci. 1977, 14.

 

Wolf, W.R. Anal. Chem. 1976, 48, 1717.

 

Harnly, J.M.; O'Haver, T.C.; Golden, 8.; Wolf, W. Anal. Chem. 1979,

 

51, 2007.

Harnly, J.M.; Patterson, K.Y.; Wolf, W.R.; O'Haver, T.C. Anal. Chem.

 

1983, 55, 1417.

Eckhoff, M.A.; McCarthy, J.P.; Caruso, J.A. Anal. Chem. 1982, 54,

 

165-168.

Risby, T.H.; Talmi, Y. CRC Crit. Rev. Anal. Chem. 1983, 14(3),

 

231-265.

Chau,Y.K.; Wong, P.T.S.; Bengert, G.A. Anal. Chem. 1982, 54, 246.

 

Robinson, J.W.; Kiesel, E.L.; Bliss, R. Anal. Chim. Acta 1977, 92,

 

321.

Yamagisawa, M.; Suzuki, H.; Kitigawa, K.; Tange, S. Spectrochim.

 

Acta Part B 1983, 38, 1143-1149.

 

McCormack, A.J.; Tong, S.C.; Cooke, W.D. Anal. Chem. 1965, 37,
1470.

Qing-Yu, O.; Guo—Chuen, W.; Ke-Wei, Z.; Wei-Lu, Y. Spectrochim.

 

Acta Part B 1983, 38, 419—425.

 

Fehsenfeld, F.C.; Evenson, H.M.; Broida, H.P. Rev. Sci. Instr. 1965,

 

36, 294.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

142

Greenfield, S.; McGeachin, H.; Smith, P.B. Talanta 1975, 22, 553.

Van Dalen, J.; Sezenne-Coulancki, A.; Galan, L. Anal. Chim. Acta

 

1977, 94, 1.

Serravallo, F.; Risby, T. Anal. Chem. 1975, 47, 2141.

 

Beenaker, C.I.M. Spectrochim. Acta Part B 1976, 31, 483.

 

Beenaker, C.I.M. Spectrochim. Acta Part B 1977, 32, 173.

 

Beenaker, C.I.M. Spectrochim. Acta Part B 1978, 33, 53.

 

Reamer, D.C.; Zoeller, W.H.; O'Haver, T.C. Anal. Chem. 1978, 50,

 

1449.

Quimby, B.D.; Delaney, M.F.; Uden, P.C.; Barnes, R.M. Anal. Chem.,

 

1979, 51, 875.

Carnahan, J.A.; Mulligan, K.J.; Caruso, J.A. Anal. Chim. Acta 1981,

 

130, 227.

Mulligan, K.J.; Zerezhgi, M.; Caruso, J.A. Spectrochim. Acta Part

 

g 1983, 38, 369-375.

Tanabe, K.; Haraguchi, H.; Fuwa, K. Spectrochim. Acta Part B 1981,

 

36, 633-639.

Uden, P.C.; Barnes, R.M.; Des Sanzo, F. Anal. Chem. 1978, 50, 852—855.

 

Lloyd, R.J.; Barnes, R.M.; Uden, P.C.; Elliot, W.G. Anal. Chem. 1978,

 

50, 2025-2029.

Quimby, B.D.; Uden, P.C.; Barnes, R.M. Anal. Chem. 1978, 50,

 

2112-2118.

Windsor, D.L.; Denton, M.P. Appl. Spectros. 1978, 32, 366-371.

 

Zerezghi, M.; Mulligan, K.J.; Caruso, J.A. J. Chrom. Sci. 1984, 22,

 

348-352.

Morita, M.; Uehiro, T.; Fuwa, K. Anal. Chem. 1980, 52, 349-351.

 

Hausler, D.W.; Taylor, L.T. Anal. Chem. 1981, 53, 1223-1227.

 

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

143

Eckhoff, M.A.; Ridgway, T.H.; Caruso, J.A. Anal. Chem. 1983, 55,

 

1004-1009.

Estes, S.A.; Uden, P.C.; Barnes, R.M. Anal. Chem. 1981, 53, 1829.

 

Holland, J.F.; Enke, C.G.; Allison, J.A.; Stults, J.T.; Pinkston, J.D.;

Newcome, 8.; Watson, J.T. Anal. Chem. 1983, 55, 997A.

 

 

Steiglitz, L.; Zwick, G. Anal. Org. Micropollut. Water 1982, 105,
112.

Uden, P.C.; Bigley, LE. Anal. Chim. Acta 1977, 94, 29.

 

Decker, R.J. Spectrochim. Acta Part B 1980, 35, 19.

 

McLean, W.R.; Stanton, D.L.; Penketh, G.G. Analyst 1973, 98, 432.
Brenner, K.S. J. Chrom. 1978, 167, 365.

Dingham, H.P.; DeJong, H.J. Spectrochim. Acta Part B 1981, 36,

 

325.

Van Loon, J.C. Can. J. Spectros. 1981, 26(4), 22A-32A.

 

Fraley, D.M.: Yates, D.A.; Manahan, S.E. App. Spectros. 1981, 35,

 

. 525.

Irgolic, K.J.; Stockton, R.A.; Chakraborti, D.; Beyer, W. Spectrochim.

 

Acta Part B 1983, 437-445.

 

Brown, R.M.; Fry, R.C. Anal. Chem. 1981, 53, 532-538.

 

Brown, R.M.; Northway, S.J.; Fry, R.C. Anal. Chem. 1981, 53, 934-936.

 

Sommer, D.; Ohls, K. Z. Anal. Chem. .1979, 295, 337.

 

Hughes, S.K.; Fry, R.C. Anal. Chem. 1983, 53, 1111.

 

Sinclair, D.A.; Whitten, R.N. Siectrochim. Acta 1960, 16, 760.

 

Nakajima, T.; Kawaguchi, H. Spectrochim. Acta 1962, 18, 201.

 

Walters, J.P. Anal. Chem. .1968, 40, 1540.

 

"Ceramic Metal Grounded Grid Thyratrons Data Sheet H5003C-1",

EGch Electro—Optics: Salem, MA 1977.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

144

Coleman, D.M.; Walters, -J.P. Spectrochim. Acta Part B 1976, 31,

 

547-587.

Black, M.S.; Browner, R.F. Anal. Chem. 1981, 53, 249.

 

"Ultrasonic Nebulizing System Model. UNS-l", Plasmatherm: Kresson,
NJ 1979.

Donahue, D.L.; Carter, J.A. Anal. Chem. 1978, 50, 686.

 

"Gas Chromatography Catalog #6", Ansepc Co.: Ann Arbor, MI 1983,
pp. 26,44.

"Insulation and Operation Instructions for On-Column Injector", SGE
Co.: Austin, TX 1980.

Newcome, B.H. "Ph.D. Thesis", Michigan State University: East
Lansing, MI 1983‘.

"Floppy Disk Drive FDD 100-8 Technical Manual", Siemens Inc.:
Anaheim, CA 1980; pp. 1—1, 1-16.

"Technical Data for AD574 A/D Converter", Analog Devices Inc.:
Norwood, MA 1981.

"700 Systems Instruction Manual Model EU-700-32 Controller",
GCA/McPherson Inc.: Acton, MA 1975, pp. 7-10, 7-24.

Gregg, H.L. Ph.D. research in progress, East Lansing, MI 1984.

Brodie, L. "Starting FORTH", Prentice-Hall Inc.: Englewood Cliffs,
NJ 1981.

Atkinson, T.V.; Gregg, H.L. "MULPLT: A Multiple Data Set, File-Based
Data Plotting Program", East Lansing, MI 1982..

Walters, J.P.; Coleman, D.M. Spectrochim. Acta Part B 1976, 31,

 

547.
Nasser, E. "Fundamentals of Gaseous Ionization and Plasma

Electronics", Wiley 6c Sons., Inc.: New York, NY 1971, pp. 219-297.

106.1

107.

108.

109.

110.

111.

112.

113.

114.

115.

116.

117.

118.

119.

145

Zynger, J. "Ph.D. Thesis", Michigan State University: East Lansing,
MI 1973.

Seng, G.T. "Ph.D. Thesis", Michigan State University, East Lansing,
MI 1978.

"General Catalog 92", Scientific Gas Products, Inc.: South Plainsfield,
NJ 1983, p. 7.

Pearse, R.W.B.; Gaydon, A.G. "The Identification of Molecular Spectra
4th Ed.", Wiley 6( Sons: New York, NY 1976.

Rief, I.; Fassel, V.A.; Kniseley, R.N. Spectrochim. Acta Part B .1973,

 

28, 105.

Knopp, C.F.; Gottschlich, C.F.; Cambel, A.B. J. Quant. SpectrL

 

Radiative Transfer 1962, 2, 297.

 

Tourin, R.H. J. Quant. Spectry. Radiative Transfer 1963, 3, 89.

 

Wiese, W.L.; Smith, M.W.; Miles, B.M. "Atomic Transition Probabilities",
U.S. Government Printing Office: Washington, DC 1969.

Taylor, H.E.; Gibson, J.H.; Skogerboe, R.K. Anal. Chem. 1970, 42,

 

1569.

Fallgatter, K.; Svoboda, V.; Winefordner, J.D. Appl. Spectrosc. 1971,

 

25,347.

Griem, H.R. "Plasma Spectroscopy", McGraw—Hill: New York, NY
1964.

Kolb, A.G.; Griem, H.R. Phys. Rev. 1958, 111, 514.

Malone, B.S.; Corcoran, W.H. J. Quant. Spectry. Radiative Transfer

 

1966, 6, 443.
Boumans, P.W.J.M. "Theory of Spectrochemical Excitation", Plenum

Press: New York, NY 1966.

120.

121.

122.

123.

124.

125.

126.

127.

128.

129.
130.

131.

146

Corliss, C.H.; Bozman, W.R. "Experimental“ Transition Probabilities
for Spectral Lines of Seventy Elements, NBS. Monograph 53", U.S.
Government Printing Office: Washington, DC 1962.

Zaidel', A.N.; Prokof'ev, V.K.; Raiskii, S.M.; Slavnyi, V.A.; Shreider,
E.Y. "Tables of Spectral. Lines", Plenum Press: New York, NY 1970.
Kirkland, J.J.; Snyder, L.R. "Introduction to Modern Liquid
Chromatography, 2nd Ed.", Wiley 6: Sons: New York, NY 1979, pp.
34-36.

"Annual Book of ASTM Standards", Am. Soc. For Testing Materials:
Philadelphia, PA 1973, p. D3237.

Scott, D.R.; Holboke, L.E.; Hadeishi, T. Anal. Chem. 1983,55, 2006.

 

"Certificate of Analysis for Standard Reference Material Lead in
Reference Fuel", U.S. Dept. of Commerce, NBS: Washington, DC
1980.

Koizumi, H.; McLaughlin, R.D.; Hadeishi, T. Anal.~Chem. 1979, 51,

 

387.

Ng, K.C.; Caurso, J.A. Anal. Chem. 1983, 55, 2032.

 

Walters, J.P.; Goldstein, S.A. "ASTM Special Technical Publication
# 540", Am. Soc. For Testing Material: Philadelphia, PA 1973.

Hieftje, G.M.; Malmstadt, H.V. Anal. Chem. 1968, 40, 1860-1867.

 

Wiegand, P.M. Ph.D. research in progress, East Lansing, MI 1984.
"Optical Multichannel Averager 111 Technical Bulletin", Princeton Applied

Research: Princeton, NJ 1984.