COMPUTER ASSISTED CHARACTERIZATION

OF A FILAMENT ELECTROTHERMAL ATOMIZER

BY

David Norman Baxter

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY
Department of Chemistry

1977

ABSTRACT

COMPUTER ASSISTED CHARACTERIZATION

OF A FILAMENT ELECTROTHERMAL ATOMIZER

BY

David Norman Baxter

The influence of modern solid state electronic instru-
mentation and computers upon analytical chemistry is pro-
found. The continuing decrease in cost and increase in
capabilities of laboratory minicomputers, particularly
with the advent of the microprocessor, have combined to
make sophisticated computer supervision of scientific
instrumentation highly beneficial. Electrothermal atomizers
are also becoming increasingly important in atomic spectro-
metric research. Many devices have been proposed in the
literature, but considerable research remains to be done
in further characterizing them.

Three specific investigations, all of which employ
versatile, computer-controlled instrumentation, have been
performed with one such atomizer, the graphite braid.
Scanning electron microsc0py with x-ray microprobe fluores-
cence has been used to examine the physical appearance
and evolution of graphite braids as they receive and

atmmize analytical samples. Changes which occur in

 

‘ “AGO!
h... ..

 

.'..;.Ho
I
u... .h. .

”‘v-n .
O

(I!

"b...

. .0
C. " ‘
”'c o.._.
....._.
.w h
.‘--‘A\'

 

 

' n

‘ ~
.. bu.
.""... ..
bco. .'

f.

x
.
—.‘_.“
. u ‘
‘.-“ .
s
" ~-.“
u .-

~..-.

 

-
o
n

 

.
a...
“ “~.‘
n“:
i
‘ c
¢-.‘
."I 3 I.
" h
h
‘ O
‘ r
v “
‘ ’
‘¥‘ .
‘-
.
:"~;
‘4: Q
.‘~ f
\
‘.
O
Q "~
" \ .
"v "-
.
.
n. .
n‘: H'-
. N-—
'1‘:

David Norman Baxter

braids as they are in service, and some of the behavior
exhibited by deposited samples are observed and discussed.
The control of temperature in electrothermal atomic spec-
trometry is an extremely important parameter. Of the
various feedback signals which can be used for this con-
trol, a method known as radiation programming, in which
the intensity of the blackbody emission from the incan-
descent atomizer is the control signal, is a superior
method of temperature regulation. Radiation programming
is shown to provide lengthened atomizer lifetimes, insensi-
tivity to changes in certain system parameters, and im-
proved atomization characteristics with implications for
lower detection limits and freedom from sample matrices.
Temperature calibration of the atomizer is also illustrated.
In conclusion, to learn more about the mechanisms of
atomization and the behavior of free atoms, two dimension-
al spatial profiles of atomic concentrations above the
atomizer have been acquired. This work was carried out
with the assistance of two computer-controlled instruments
of original design: a "positioner" which translates the
atomization cell with respect to a fixed optical axis,

and a "sampler" which automatically deposits aliquots of
solution on the atomizer. The general shapes of the free
atomic clouds of various elements are shown, and observa-
tions which concern the effects of some sample matrices

are described.

DEDICATION

To my parents and family:

for so much support through so many years

ii

r 4
I"

V
h,’ -0
.so.r. u”
o

":0; ...

O
.915..-

 

~.' . '
.¢. 5-
...\

 

tn
4 I
U.
I?’

.‘. AI

 

\
I
Q
N
‘L‘
.
‘
“‘-~
‘-
s
u

ACKNOWLEDGMENTS

I am indebted to many people for making much of this
work possible, and I wish to express to them my deepest
gratitude.

I wish at first to thank my research preceptor, Pro-
fessor Stanley R. Crouch, for his support and guidance
for the past four and a half years. Thanks are also due
Professor Chris Enke, who served as my second reader,
and Professors James Dye and Richard Schwendeman, who
also were on my guidance committee.

I owe much to our postdoctoral fellow of 1974-1976,
Dr. Eric Johnson. He personally increased my knowledge
of computers and electronic instrumentation by at least
an order of magnitude. Similar credit should be given
to my coworker, Gene Pals, without whose comprehensive
software and instrumental contributions this work would
have gone nowhere, and who always was a valuable second
opinion and source of ideas.

To Professor Andrew Timnick goes my thanks for being
such a good friend and coworker through many terms of
teaching. To Dr. Tom Atkinson I extend my gratitude for
three pleasant terms of employment, and for formidable
talents and assistance in all facets of computer hard-
ware and software.

I have drawn heavily upon the services of our depart-

mental support facilities in constructing the instrumentation,

iii

and I owe much to the men who work in them for putting up
with my unreasonable demands and periodic raids on their
stock. Included in this category are Ron Haas and the
staff of the electronics shop, Russ Geyer and the staff

of the machine shop, and Marty Rabb of the instrumentation
consulting service.

I have had much material assistance from sources out-
side the university as well. Thanks are due Dr. Paul
Beckwith and the BASF Corporation of Wyandotte, Michigan
for their generous donation of both time and supplies on
their scanning electron microscopy facility. I acknow-
ledge also five terms of research support in the form of
both a summer and a full year ACS Analytical Fellowship
under the kind sponsorship of the Procter and Gamble
Company of Cincinnati, Ohio. My thanks also go to Peri-
Anne Warstler for typing the original manuscript.

Finally, I wish to thank all of my fellow graduate
students in the analytical division who either contributed
in some way to this work, no matter how small, or who
simply were good friends throughout my stay here. To
name everyone would take up another page in this already
oversized mound of ink, paper, and finely divided silver.
May I simply say that losing your daily companionship is

the most sincere regret I have in leaving MSU.

iv

A!
.
I...c'
vn-u ._.
b

"‘n .—
.

.‘h. ..
.O-- A.
‘ O
.‘b. ..

 

I “g
u...__-.
\

n1 :‘J

D.’

a
r

(H

Chapter

LIST OF

LIST OF

TABLE OF CONTENTS

TABLES O O O I O l O O I O O O O O O O O

FIGURES . . . . . . . . . . . . . . . .

INTRODUCTION 0 O C O O O O O O 0 O I O O O O O O

HISTORICAL. . . . . . . . . . . . . . . . . . .

A.

w

OWMUO

Early Observations and Developments
of Atomic Spectrometric Techniques. . .

Comparison of Flame and Electrothermal
Me thOdS O O 0 O O O I O O O O O O O O O

Furnace Blectrothermal Atomizers. . . .
Filament Electrothermal Atomizers . . .
Sampling Boat and Probe Techniques. . .
Chemical Atomization Techniques . . . .

Specific Literature Background
of the Present Work . . . . . . . . . .

COMPUTER CONTROLLED INSTRUMENTATION . . . . . .

A.
B.

Introduction. . . . . . . . . . . . . .
The Positioner. . . . . . . . . . . . .

1. General Introduction and
Overview. . . . . . . . . . . . . .

2. Sequence Structure and
the Flow of States. . . . . . . . .

3. Detailed Logic. . . . . . . . . . .

a. Dual Synchronous Clock
Source 0 O O O O O O O O O O O O

b. Sequence Initiation
Circuitry . . . . . . . . . . .

c. Destination Preparation
Circuitry O I O O O O O O O O 0

d. Destination Seeking
Circuitry . . . . . . . . . . .

e. Overcount Circuitry . . . . . .

Page

ix

xii

13
18
22
24

25
28

28
29

29

36
43

43

45

47

52
57

.-
OK-

b;’

n... .5.
o

1

('7

Chapter

Page

f. Motor Driving Circuitry . . . . . . 61
4. Interface . . . . . . . . . . . . . . . 67

5. Power Supplies and Physical
Structure 0 I O O O O O O O O O O O O O 67

6. Testing and Performance . . . . . . . . 69
7. Operating Software. . . . . . . . . . . 74
The Autosampler . . . . . . . . . . . . . . 75
1. History and Introduction. . . . . . . . 75
2. Device Description. . . . . . . . . . . 80
3. Detailed Logic. . . . . . . . . . . . . 87
a. Registers . . . . . . . . . . . . . 87
b. Vertical Motor. . . . . . . . . . . 98
c. Turret Motor. . . . . . . . . . . . 99
d. Rotation Motor. . . . . . . . . . . 102
e. Syringe Motor . . . . . . . . . . . 103
4. Interface . . . . . . . . . . . . . . . 110

5. Power Supplies and Physical
Structure . . . . . . . . . . . . . . . 114

6. Testing and Performance . . . . . . . . 115
7. Operating Software. . . . . . . . . . . 116
Data Acquisition System . . . . . . . . . . 118

1. Multifrequency Real Time
CIOCk O O O O O O O O O I O O O I O O O 118

2. A/D Converter . . . . . . . . . . . . . 126

3. Power Supplies and Physical
Structure . . . . . . . . . . . . . . . 127

4. Testing and Performance . . . . . . . . 130
5. Operating Software. . . . . . . . . . . 130
Perspectives. . . . . . . . . . . . . . . . 131
l. Positioner. . . . . . . . . . . . . . . 132
2. Autosampler . . . . . . . . . . . . . . 136
3. Data Acquisition System . . . . . . . . 138
Miscellaneous Instrumentation . . . . . . . 139

1. Braid Heating Level
Converter . . . . . . . . . . . . . . . 139

vi

Chapter Page

2. General Purpose F1ags...................142

3. Computer Conveniences. ....... .......... 143

G. Remote Terminal Facility . . . . . . . . . .145
MECHANICAL AND OPTICAL INSTRUMENTATION AND

BASIC SYSTEM PERFORMANCE . . . . . . . . . . . . . 147

A. The Atomization Cell . . . . . . . . . . . 147

B. The Optical System . . . . . . . . . . . . 152

C. Atomic Absorption Instrumentation. . . . . 155

D. Braid Power Supply and Controller. . . . . 156

E. Operating Software . . . . . . . . . . . . 157

F. Summary of Events for a Single
Sample . . . . . . . . . . . . . . . . . . 159

G. Calculation of Absorbance. . . . . . . . . 161

H. Perpendicular Versus Parallel
Braid Orientation. . . . . . . . . . . . . 165

I. Effect of Entrance Slit Geometry . . . . . 167
J. Nature of Instrumental Noise . . . . . . . 169
K. Importance of Background Correction. . . . 176
L. Alternate Methods of Data Reduction. . . . 179

SCANNING ELECTRON MICROSCOPY STUDIES OF
GRAPHITE BRAIDS O O O I O O O O O O C O I O O O O O 183

AN EVALUATION OF THE RADIATION METHOD FOR
ATOMIZER TEMPERATURE CONTROL . . . . . . . . . . . 203
A. Introduction . . . . . . . . . . . . . . . 203

B. Qualitative Comparison of the
TeChniqueS O O O O O O O O O O O O O O O O 205

C. Temperature Calibration of the
Graphite Braid . . . . . . . . . . . . . . 207

D. Long Term Temperature Drifts of
Graphite Braid . . . . . . . . . . . . . . 217

E. Sheath Gas Flow and Contact
Resistance Effects . . . . . . . . . . . . 227

F. Performance of Methods During
Atomizations O O O O O O O O O 0 O O O O O 230

G. Summary and Conclusions. . . . . . . . . . 247

vii

Chapter

ATOMIC CONCENTRATION PROFILES OVER

GRAPHITE BRAIDS. . . . . . . . . . . . . . .
A. Introduction . . . . . . . . . . . .
B. Vertical Appearance of Concentration
Profiles . . . . . . . . . . . . . .
C. Horizontal Appearance of Concen-
tration Profiles . . . . . . . . .
D. Effect of Temperature Upon Silver
Integrated Absorbance. . . . . . .
E. A Discussion of Some Problems of
Interpretation . . . . . . . . . . .
F. Effect of Oxygen Upon Free Atom
Populations. . . . . . . . . . . . .
G. Effects of Matrices Upon Concen-
tration Profiles . . . . . . . . . .
SUMMARY AND CONCLUSIONS. . . . . . . . . . .

APPENDIX A: SIMPLE HARDWARE CONTROL APPROACH
FOR SEQUENCING CHEMICAL INSTRUMENTATION. . .

APPENDIX B: SOURCES OF COMMERCIAL
INSTRUMENTATION. . . . . . . . . . . . . .

REFERENCES 0 O C O I O 0 O O O O O O O O O 0

viii

Page

251
251

254

261

271

279

281

286
304

308

332

335

Table

10

11

LIST OF TABLES

Page

Command set for positioner

interface. . . . . . . . . . . . . . . . 66
Precision of location of the

reset point: vertical. . . . . . . . . . 71
Precision of location of the reset

point: horizontal. . . . . . . . . . . . 73
Bit assignments of the autosampler
registers. . . . . . . . . . . . . . . . 89
Ballast resistor values for the

autosampler motors . . . . . . . . . . . 109
Command set for autosampler

interface. . . . . . . . . . . . . . . . 113
Frequencies available from the

data acquisition system. . . . . . . . . 124
Command set for the data acquisi-

tion system interface. . . . . . . . . . 129
Command set for the final instru-

mental sections. . . . . . . . . . . . . 144
Improvement of signal and precision

for parallel versus perpendicular

braid orientation. . . . . . . . . . . . 166
Effect of monochromator orientation

upon results . . . . . . . . . . . . . . 168

ix

 

Table Page

12 Significance of the need for
background correction. . . . . . . . . . 177
13 Comparison of algorithms for

the calculation of absorbance. . . . . . 181
14 Comparison of pyrometrically

observed temperatures with the

present method . . . . . . . . . . . . . 216
15 Instrumental parameters for the

comparison of radiation and power

programming. . . . . . . . . . . . . . . 232
16 Analysis of the cadmium-copper

mixture. . . . . . . . . . . . . . . . 248
17 Common instrumental parameters

for atomic concentration profile

work . . . . . . . . . . . . . . . . . . 255
18 Change in integrated absorbance

of silver samples with temperature . . . 273
19 Transit times for silver samples

at different temperatures. . . . . . . . 276
20 Effect of sheath flow rate upon

cadmium spatial data . . . . . . . . . . 278
21 Effect of sulfate ion from ‘

sulfuric acid upon integrated

absorbance of cadmium. . . . . . . . . . 287
22 Effect of chloride ion from

potassium chloride upon integrated

X

Table

23

24

Page

absorbance of cadmium. . . . . . . . . 292
Matrix effects of zinc and

cadmium at lowered total salt

concentrations . . . . . . . . . . . . . 296
Effects of matrices upon inte—

grated absorbance for copper

and magnesium. . . . . . . . . . . . . . 297

xi

Figure

10

11

12

13

LIST OF FIGURES

Block diagram of computer-controlled
atomic spectrometric instrumentation.
Assignment of atomizer cell locations
Block diagram of positioner control
logic . . . . . . . . . . . . . . . .
Flow of states and functions in a

GO cycle. . . . . . . . . . . . . . .
Flow of states and functions in a
RESET cycle . . . . . . . . . . . . .
Schematic diagram of a sequencer. . .
Schematic diagram of the dual,
synchronous clock source. . . . . . .
Schematic diagram of the sequence
initiation circuitry. . . . . . . . .
Schematic diagram of the decimal
registers . . . . . . . . . . . . . .
Schematic diagram of the principal
positioner registers. . . . . . . . .
Schematic diagram of the gates

and direction control circuitry . . .
Schematic diagram of the condition

4 detect logic. . . . . . . . . . . .
Schematic diagram of the overcount
circuitry . . . . . . . . . . . . . .

xii

Page

32

34

37

4O
42

44

46

48

49

53

54

59

Figure Page

14 Schematic diagram of the stepper

motor waveform generator. . . . . . . . . . 62
15 Schematic diagram of the stepper

motor drivers . . . . . . . . . . . . . . . 63
16 Normal rest position of the auto—

sampler . . . . . . . . . . . . . . . . . . 83
17 Sample depositing position of the

autosampler . . . . . . . . . . . . . . . . 83
18 Turret addressing position of the

autosampler . . . . . . . . . . . . . . . . 84
19 Sample filling or emptying position

of the autosampler. . . . . . . . . . . . . 84
20 Schematic diagram of the gate,

direction control, and error detect

circuitry for the vertical motor. . . . . . 91
21 Schematic diagram of the fine

positioning circuitry for the

vertical motor. . . . . . . . . . . . . . . 94
22 Schematic diagram of the stop

circuitry for the vertical motor. . . . . . 97

23 Schematic diagram of the circuitry

for the turret motor. . . . . . . . . . . . 100
24 Schematic diagram of the circuitry

for the rotation motor. . . . . . . . . . . 104

xiii

Figure Page

25 Schematic diagram of the gate,

error detect, and stop circuitry

for the syringe motor . . . . . . . . . . . 106
26 Schematic diagram of the sample

size control circuitry for the

syringe motor . . . . . . . . . . . . . . . 107
27 Schematic diagram of the autosampler

interface flag and clock sources. . . . . . 111
28 Schematic diagram of the clock

source and frequency select

circuitry for the data acquisition

system. . . . . . . . . . . . . . . . . . . 120
29 Schematic diagram of the frequency

dividing and gating circuitry for

the data acquisition system . . . . . . . . 121
30 Schematic diagram of the data ac-

quisition system interface flag and

data reading circuitry. . . . . . . . . . . 128
31 Improved flow of states and func-

tions for a RESET cycle . . . . . . . . . . 135
32 Schematic diagram of an improved

gate circuit for the autosampler

motors. . . . . . . . . . . . . . . . . . . 137
33 Schematic diagram of the braid

heating level control circuitry . . . . . . 140

xiv

 

’0

‘1

A

Figure Page

34 Illustration of the atomization

cell. . . . . . . . . . . . . . . . . . . . 148
35 Optical system for the atomic

absorption experiments. . . . . . . . . . . 153
36 Illustration of the effect of

readout time constants upon

integrated absorbance . . . . . . . . . . . 164
37 Noise characteristics of the

instrumentation over various

periods of observation: silver

hollow cathode. . . . . . . . . . . . . . . 172
38 Noise characteristics of the

instrumentation over various periods

of observation: lead hollow cathode . . . . 173
39 Virgin graphite braid at 50

diameters . . . . . . . . . . . . . . . . . 186
40 Degassed virgin braid at 50

diameters . . . . . . . . . . . . . . . . . 186
41 Virgin braid at 500 diameters . . . . . . . 188
42 Individual fiber of braid at

20,000 diameters. . . . . . . . . . . . . . 188
43 Individual fiber of a used braid

at 20,000 diameters: electrode

holder region . . . . . . . . . . . . . . . 190

XV

.

wr-y

...‘.
.

4‘ ?.4

1‘.

‘

—

~l

:

.-
'N

3: .

n

~

‘

C
'3
Q.

N
:p
v.
t‘
'\
:-
~(

I).
«J

"I

1"

 

Figure Page

44 Individual fiber of a used braid

at 20,000 diameters, with apparent

extensive damage. . . . . . . . . . . . . . 190
45 Individual fiber of a used braid

at 10,000 diameters, with deposited

particulates. . . . . . . . . . . . . . . . 192
46 Individual fiber of an old braid at

20,000 diameters: central region. . . . . . 192

47 Braid intentionally damaged through
atmospheric exposure, 500 diameters . . . . 193
48 Braid used in high temperature work,

100 diameters . . . . . . . . . . . . . . . 193
49 Copper nitrate salt particle on braid,

10,000 diameters. . . . . . . . . . . . . . 195
50 Nickel nitrate salt particle on

braid, 10,000 diameters . . . . . . . . . . 195
51 Glass bead on braid, 10,000

diameters . . . . . . . . . . . . . . . . . 198
52 Combined cadmium and zinc nitrate

salt particle on braid, 10,000

diameters . . . . . . . . . . . . . . . . . 198
53 Exploded silver nitrate salt

particle on braid, 10,000 diameters . . . . 200
54 Copper nitrate salt particle on

braid, 10,000 diameters . . . . . . . . . . 200

xvi

e

I‘N'HV
06“.
l

'1! olll ‘

C s
\. st.
and D...
IN: rev
AI. 3..
Q I nnc

In. on.
pQJ and
nil

Figure Page

55 Oscilloscope photographs of power

supply responses under radiation

and power programming . . . . . . . . . . . 206
56 Temperature calibration of graphite

braid with respect to reference

voltage: low temperature region . ._. . . . 212
57 Temperature calibration of graphite

braid with respect to reference

voltage: high temperature region. . . . . . 214
58 Temperature calibration of graphite

braid with respect to phototransistor

output: low temperature region. . . . . . . 218
59 Temperature calibration of graphite

braid with respect to phototransistor

output: high temperature region . . . . . . 219
60 Temperature decay of graphite braid

under radiation and power programming:

low temperature . . . . . . . . . . . . . . 220
61 Temperature decay of graphite braid

under radiation and power programming:

moderate temperature. . . . . . . . . . . . 224
62 Temperature decay of graphite braid

under radiation and power programming:

high temperature. . . . . . . . . . . . . . 226

xvii

Figure Page

63 Cadmium transients under ratiation
programming . . . . . . . . . . . . . . . . 234
64 Cadmium transients under power

programming . . . . . . . . . . . . . . . . 23S

65 Lead transients under radiation
programming . . . . . . . . . . . . . . . . 236
66 Lead transients under power

programming . . . . . . . . . . . . . . . . 237

67 Silver transients under radiation
programming . . . . . . . . . . . . . . . . 238
68 Silver transients under power

programming . . . . . . . . . . . . . . . . 239
69 Copper transients under radiation

programming . . . . . . . . . . . . . . . . 240
70 Copper transients under power

programming . . . . . . . . . . . . . . . . 241
71 Nickel transients under radiation

programming . . . . . . . . . . . . . . . . 242
72 Nickel transients under power

programming . . . . . . . . . . . . . . . . 243
73 Distortion of transients under

radiation programming by readout

time constants. . . . . . . . . . . . . . . 245

xviii

Figure Page

74 Normalized concentration profiles with

vertical height for silver, cadmium,

and lead . . . . . . . . . . . . . . . . . 257
75 Normalized concentration profiles

with vertical height for zinc, mercury,

and thallium . . . . . . . . . . . . . . . 258
76 Normalized concentration profiles

with vertical height for copper

and magnesium. . . . . . . . . . . . . . . 259
77 Normalized concentration profiles with

horizontal displacement at a vertical

height of 0 mm . . . . . . . . . . . . . . 262
78 Normalized concentration profiles

with horizontal displacement at a

vertical height of 6 mm. . . . . . . . . . 263
79 Normalized concentration profiles

with horizontal displacement for

silver at various vertical heights . . . . 266
80 Normalized concentration profiles

with horizontal displacement at a

vertical height of 0 mm and a per-

pendicular braid . . . . . . . . . . . . . 268
81 Normalized concentration profiles

with horizontal displacement at a

vertical height of 6 mm and a per-

pendicular braid.. . . . . . . . . . . . . 269

xix

5.3-1
.

I].

Figure

82

83

84

85

86

87

88

89

Page

Normalized concentration profiles

with horizontal displacement at a

vertical height of 12 mm and a

perpendicular braid. . . . . . . . . . . . 270
Normalized concentration profiles

with horizontal displacement for

silver at various atomizer tempera-

tures. . . . . . . . . . . . . . . . . . . 274
Effect of oxygen upon the vertical
concentration profile of cadmium . . . . . 283
Effect of oxygen upon the vertical
concentration profile of zinc. . . . . . . 284
Effect of oxygen upon the vertical
concentration profile of lead. . . . . . . 285
Effect of added chloride ion as

ammonium chloride upon the vertical
concentration profile of cadmium . . . . . 289
Effect of added chloride ion as

potassium chloride upon the high

resolution vertical concentration

profile of cadmium . . . . . . . . . . . . 293
Effect of sulfate ion as potassium

sulfate upon the vertical concentra-

tion profile of copper . . . . . . . . . . 298

XX

t

1'!

Figure Page

90 Concentration dependence of the

effect of added fluoride ion upon

the vertical concentration profile

of magnesium . . . . . . . . . . . . . . . 301
91 Effect of added fluoride ion upon

the high resolution vertical concen-

tration profile of magnesium . . . . . . . 302
A1 Basic flowchart units for transfer

conditions and transfer functions. . . . . 314
A2 Sequencer flowchart for the example

of a bath thermostat . . . . . . . . . . . 316

A3 General schematic diagram of a
controller . . . . . . . . . . . . . . . . 317
A4 Sequencer flowchart for the example

of an automatic fraction collector . . . . 320
A5 Schematic diagram of the controller
for the example of the automatic

fraction collector . . . . . . . . . . . . 322

A6 Sequencer flowchart for the filament
positioning instrument . . . . . . . . . . 326
A7 Schematic diagram of the controller

for the filament positioning instrument. . 328

xxi

 

Ci.

he.

‘ 0

on

 

 

 

 

 

 

INTRODUCTION

The practice of analytical chemistry has been revolu-
tionized in recent years by the tremendous advances made
in scientific instrumentation. The variety of physical
phenomena of analytical importance for which precision
instrumentation is available has expanded dramatically,
and the sophistication of both new and familiar instru-
ments increases annually. Much of this progress has been
made possible by a parallel and equally burgeoning growth
in the technology of computers and solid state electronics.
The size, cost, and power consumption of computers and
electronic instruments have steadily dropped, whereas their
capabilities, speed, and performance have continually risen.
Because of these trends, modern analytical instrumentation
can not only perform determinations or observe phenomena
previously unobtainable, but can also guide its own
activities, monitor and correct drifts in its perfor-
mance, and relieve the user of many necessary but routine
duties. When coupled to a computer, an analytical instru—
ment becomes a particularly powerful tool for scientific
investigations, in which the complex capabilities of the
computer can serve to provide detailed and flexible con-
"trol over the instrumental functions, as well as compre-
luensive treatment and presentation of the acquired data.

Ida recent years, the attainment of such goals has been

 

'u .‘

' ‘ .
set

6.,

U...“
h-
"“5

t v
u.

 

made particularly possible through the introduction and
growth of microprocessors, in which the entire central
processing unit of a modern minicomputer can be placed

on a single solid state device. Their impact has already
been felt in existing instrumentation, and the promise
they hold for future instruments is virtually limitless.
The research described in this thesis has as its central
goal the use of such modern, custom-designed instrumenta-
tion in conjunction with computer-controlled operation to
investigate various facets of an analytical technique
that would be prohibitively difficult to study without
the benefits such control can provide. The technique under
study is filament-type electrothermal atomic absorption
spectrometry, with an atomizer known as the graphite
braid.

Atomic spectrometry is an analytical technique rich
in both instrumental requirements and chemical complexity.
Studies of variables in these categories have been per-
formed for some time by many workers. Most of these have
been with the traditional flame atomizer. The nonflame,
or electrothermal atomizers are more recent developments;
the oldest of them has existed less than twenty years.
iIn the past decade, a great many such devices have been
ciescribed in the literature. Much work remains to be

done, however, in better characterizing atomizers that

already exist.

If a device is to serve effectively as an electrothermal

atomizer, it follows that studies of the device itself
can be of great potential benefit in learning what physi-
cal and chemical processes occur as it is used. In the
present work, scanning electron microscopy and x-ray
microprobe fluorescence have been employed to observe the
appearance of graphite braid and the changes which it
undergoes while in service as an atomizer.

The control of temperature in electrothermal atomizers
is an extremely important parameter. Because an atomizer
will age with continued use, it is a serious matter to
assure that the power supply which heats the device can
be effectively and sensitively regulated to preserve as
nearly consistent temperatures from atomization to atom-
ization as possible. If an electrical parameter such as
the voltage develOped across the atomizer is the control
signal, the drift of atomizer temperature with continued
use can be considerable, because control is based upon
an input to the atomizer rather than an output from it.
The radiation method of atomizer control, which follows
the blackbody emission from the device, is such an output
parameter. It will be shown in the present work that this
method is an excellent control signal for electrothermal
atomic spectrometry, and that it provides superior con-
sistency in atomizer temperatures, increased atomizer
lifetime, and the promise of improved detection limits and
ability to deal with complex sample matrices.

Although computer control and powerful instrumentation

plays a significant role in the above studies, their full
capabilities are realized in the final field of investiga-
tion, which is a study of the spatial profiles of free
atom concentrations above the atomizer. To perform such
studies, repeated atomizations of a constant sample are
taken while the atomization cell is systematically trans-
lated in two dimensions with respect to a fixed optical
axis. This work is extremely time consuming because of
the discrete nature of electrothermal atomic spectrometric
sampling. It is also very tedious to perform manually.

In the present work, two sephisticated, computer-controlled
instruments are described which make these studies practi-
cal. One of these is a positioning device which is capable
of performing the orderly movement of the atomization cell.
The second is a versatile and precise autosampling unit.
With these two instruments, and additional instrumentation
which includes a dedicated data collection system, a com-
plete, computer-controlled atomic absorption spectrometer
is described which is able to run itself virtually un-
assisted by the user. As an aid and convenient reference
for the descriptions to follow, Figure 1 shows a block

diagram of the entire operating system.

.cofiuwu
Icmssuumcw oauumEouuoomm vasoum pmaaouuCOOIumusmEoo mo EcumMAU xoon .a musmwm

¢0P<xo¢10 Jawu
62°! zo~h<NnxOh<

¢u_u—3ax< Nam—3;.psax
.> or — apex;

 

 

 

 

8°~h~¢~300<
<h¢O

cmqaoxpzou

Ema 3
CwN~80h< 38<00h < cmzonhnmom

 

 

 

 

4<2—xcuh cowauuomn Dunn
cu“: 4<¢k2wu rttoau

 

b.

6»

b».

1.

by

LL

Q».

HISTORICAL

Atomic spectrometry is today a well established
laboratory technique. It is capable of performing a wide
variety of both qualitative and quantitative determina-
tions, and finds a place in laboratories which do every-
thing from the most current and speculative research to
the most routine and common repetitive sample analysis.
Atomic Spectrometric methods are applicable to nearly
every metallic and metalloidic element of the periodic
table, and are frequently the techniques of choice for
the determination of trace quantities of sample consti-
tuents, or for multielement determinations of several

species within the sample.l"4

A. EARLY OBSERVATIONS AND DEVELOPMENTS OF

ATOMIC SPECTROMETRIC TECHNIQUES

Initial observations of atomic spectrometric phenomena
probably date from pre-Lavoisierian chemistry in the form of
the familiar colored flame tests for qualitative analysis.
These early discoveries were given a more systematic and
scientific chemical treatment in the pioneering work of

5 at

Bunsen and Kirchhoff over one hundred years ago,
which time the observation of prism sorted emission spectra
represented one of the prime experimental techniques for

the discovery and confirmation of additional known elements.

With the growth of technological skills in optics and
electrical power generation, the study of emission spectra
and their permanent recording on photographic plates came
to be established as one of the earliest instrumental
techniques. In the years since, electrical arc and spark
atomic emission spectrography have developed a literature
rich in both theory and practice for a great many chemical

6'8 In fact, emission methods remain today as

systems.
the workhorse techniques for the analysis of materials
having metallurgical or geological significance, or for
studies in specialized fields such as astronomy. Applica-
tions and advances in flame emission and radiofrequency
emission techniques have enjoyed a parallel growth no less
impressive.9-11
A satisfactory theoretical and mathematical explana-
tion of atomic spectra was not possible until the develOp-
ment of quantum mechanics in the early part of this century,
after which a full rationalization of atomic spectra was
rapidly developed. By the early thirties, a very complete
body of knowledge had been acquired, which could effectively
and consistently explain all of the basic atomic spectro-
metric phenomena, as well as a considerable number of the

12 In

more common deviations from ideal model behavior.
return, atomic spectrometry served as one of the principal
techniques for the experimental confirmation of early
quantum mechanical predictions, the Bohr atom being a

particularly prominent and well known example. Inherent

in the theoretical treatment, and in some experimental
observations at the time, was the realization of the exis-
tence of the other two atomic spectrometric techniques,
atomic absorption and atomic fluorescence.

Although atomic absorption had been known since 1802
with Wollaston's discovery of the Fraunhofer lines in the
sun's spectrum, true chemical applications of the method
did not occur until the early 1950's, partly because of
instrumental limitations imposed by the lack of a practical,

13 With the adoption of the

stable, narrow line source.
hollow cathode discharge tube, and the groundbreaking work
of Alan Walsh in Australia, atomic absorption moved into
the realm of viable analytical chemical techniques.14'15
After a somewhat slow start during the remainder of the
fifties, atomic absorption grew rapidly during the mid-
sixties. Several lines of competitive instrumentation
appeared in the marketplace, and an intensive though at
times overly idealistic sales program got underway. By
the end of the sixties, an atomic absorption instrument
of some sort was likely to be found in nearly every major
academic and industrial lab, and the growth of the method
has been prodigious ever since.16'17
Atomic fluorescence is an even more recent develop-
ment than atomic absorption. Although it too had been
recognized theoretically and experimentally for some

years,18'19 it was not until 1962 that Alkemade proposed

20

its application to chemical analysis, and not until

1964 that the first modern paper on the subject appeared

21

by Winefordner and Vickers. In the short time since, it

too has enjoyed a significant growth,22'23

although it
remains at this writing primarily a research method only,
chiefly because of the lack of commercial instrumentation,
and some reluctance by instrument builders to accept it as

a companion to atomic absorption rather than a duplication

or a competitor.

B. COMPARISON OF FLAME AND ELECTROTHERMAL METHODS

With the exception of the electrical arc and spark
techniques, most experimentation in atomic spectrometry
has been performed using a flame as the medium with which
to produce free atomic vapor. This is historically the
natural development of the discipline, and a flame support-
ing instrument is still the standard configuration for
commercial atomic spectrometric systems. This is not to
imply, however, that the flame is an outmoded or funda-
mentally unsuitable atomization device. Flames do possess
a number of characteristics that recommend their use.24
Burners and auxiliary apparatus to ignite and sustain a
flame are common and readily available in a variety of
modifications. They are capable of producing a flame of
high stability and reproducibility. The number of different
fuel and oxidant combinations available, as well as the

ability to adjust fuel-to-oxidant ratios, allows flames

10

to be used over a wide range of temperatures. Flames offer
very good levels of detectability and sensitivity for
many elements with resonance lines in the common region of
200 to 400 nm. Finally, the sample in a flame system may
be continually aspirated, which gives rise to a DC signal
that may be detected as such or modulated with an AC com-
ponent for phase-locked detection, and improvement of
signal-to-noise qualities.

Yet, in spite of their popularity and capabilities,
flame atomizers do suffer from numerous disadvantages
that detract from their usefulness. The flame itself is
an extraordinarily complex chemical environment, rich in
molecular, atomic, and radical species which can seriously
plague an analysis. This is true both from the chemical
standpoint of depOpulating the supply of free atomic species
through formation of stable analyte compounds such as
hydroxides or oxides, and from the physical standpoint
through such effects as molecular quenching of fluores-
cence. The high rates of flow and the expansion effects
of the combusting gases act to dilute the atomic concentra-
tion of analyte in the flame. Flames can often contribute
significantly to the background emission or absorption
character of the system, especially for those elements
with analysis wavelengths near the vacuum ultraviolet.
The overall efficiencies of aspiration, desolvation, and
vaporization are rather low for flame systems. Finally,

the amount of sample solution required may be excessive,

11

if not impossibly large, for critical samples such as are
found in clinical work. It may certainly be counter-
argued that much modern research has been directed towards
alleviating these problems. The use of ultrasonic nebu-
lization to improve sample transport efficiencies is a case

25 Nonetheless, because of effects such as these,

in point.
alternative methods of atomization in atomic spectrometry
have found increasing use in recent years.

The alternative practical energy source for atomic
spectrometric atomization is electrical power. Over the
past fifteen to twenty years, many researchers have pro-
posed and characterized such "flameless", or more properly
stated, electrothermal atomizers, which are generally
fashioned out of some substance which is sufficiently
refractory to withstand sudden and repetitive heating
to temperatures as high as 3000° K.26 Perhaps the most
fundamental contrast between these methods and the more
classical flame approaches is that in electrothermal
techniques, the sample is not usually continuously as-
pirated by some sort of nebulizer, but is rather deposited
as a discrete sample of typically only a few microliters
total volume directly onto the atomizer. In a similar
manner, the atomizer itself is not usually continuously
at the atomization temperature, but is instead stepped
through a heating sequence. First, gentle heating is
employed to remove the solvent, followed by strong heating

to affect atomization, with an optional intermediate

12

heating level for the ashing of organic matrices. From
these differences arise many of the comparative advantages
and disadvantages of the electrothermal techniques. Elec-
trothermal atomizers lack the complex chemical environ-
ment of a burning flame, as well as much of the expanding
gas dilution effect. They also possess considerably less
spectral background interference. These effects all com-
bine favorably to give electrothermal atomizers a general
advantage over flames in terms of detectability and sensi-
tivity for many elements. The very small sample sizes are
beneficial to those analyses where sample conservation is
essential. Furthermore, the fact that the entire sample
may be discretely placed directly upon the atomizer sur-
face substantially improves the overall efficiency in making
the transitions from a dilute solution to an atomic vapor.
At the same time the electrothermal methods suffer
difficulties as well. One of the most vexing, at least
for those devices employing discrete sampling, is the
transitory nature of the analytical signal. In place of
the DC level or AC waveform that can be observed and
processed over tens of seconds, many electrothermal
atomizers produce an absorbance spike of duration as
short as one quarter second. In addition, sampling cannot
be repeated much faster than one sample per minute. The
instrumental readout must consequently be prepared to
follow transients, and integration of the transient is

advisable for quantitation of results over a reasonable

13

dynamic range. Furthermore, the same atomizer must be

used over and over for each successive sample, which in-
evitably leads to drifts due to aging, and often requires
the eventual discarding of a spent device. A final problem
is the difficulty in reproducibly drawing and depositing
sample quantities on the order of only a few microliters.
Electrothermal atomizers can be divided into either the
"furnace" or "filament" categories. They are discussed

here in that order.

C. FURNACE ELECTROTHERMAL ATOMIZERS

The concept of using a furnace—like device for atomic
spectrometric studies dates back to the early part of this
century, when King proposed the use of a furnace manu-
factured from graphite for studies in atomic spectrophys-

27-29 The first such device intended specifically for

ics.
analytical chemical work was that proposed by L'vov about
1961.30 His atomizer consisted of a graphite tube 30 to
50 mm long and 2.5 to 5 mm in inner diameter. The sample
was placed onto a treated graphite electrode which was
inserted into an opening in the side of the furnace. To
atomize the sample, an arc was struck from this electrode,
while simultaneously another power source heated the main
furnace body to temperatures as high as 3000° K. To pro-

tect the device from atmospheric attack, it was enclosed

in an argon-filled container fitted with quartz windows

Of

bk

Tu

O»

Q:

uni

\H‘

14

to permit passage of the source beam down the furnace axis.
L'vov and Lebedev improved upon the original design with
such additions as two channel operation with an internal
standard, a simpler means of heating samples, use of
pyrolytic graphite in the furnace to retard sample dif-
fusion into the walls, and the ability to function at

31’32 Some forty ele-

greater than atmospheric pressure.
ments were examined and detection limits on the order of
picograms of material were realized.

In the late sixties, Woodriff and coworkers construct-
ed a furnace atomizer 150 mm long and 7 mm in diameter.33-39
It was continuously heated by a high wattage supply, and
was enclosed in an insulated, steel container provided
with a continuous flow of argon. Samples were either
discretely deposited, or nebulized into the furnace through
a side arm. A number of elements were examined with this
device, as were some matrix effects.

About the same time as Woodriff's work, Massmann
devised another furnace similar to L'vov's, but simpler

40’41 The tube measured 55 mm long and

in construction.
6.5 mm in diameter. It was resistively heated up to 2600°
K by a high wattage supply. Samples were deposited through
a small hole bored in the side of the furnace. Both solu-
tion and solid samples were examined, and the device was
also redesigned in an alternate shape specifically intended

for atomic fluorescence. A further modification of the

Massmann device was marketed in 1970 by the Perkin-Elmer

‘1:

(I

(I)

I).

(J

J

'1'1

(I)

'1

HI

I"

H’

15

Corporation, and named the HGA-2000 ("Heated Graphite

42-44 It was the first of several commercial

Atomizer").
electrothermal devices, and is in increasingly widespread
use.

A furnace atomizer for use in atomic fluorescence was

45 It featured observation slits

designed by Wineforder.
cut in the graphite so as to prevent observation of black-
body radiation by the detector, as well as a septum and
syringe method of sample injection. Robinson and coworkers
described an induction heated atomizer for the determination

46-48

of air pollutants. Air is drawn through a tube filled

with carbon chunks and heated to about 1400° C. The carbon
monoxide produced acts with additional hot carbon to re-
duce metallic ion vapors to the elemental state, after
which they are analyzed in a second heated tube made of

quartz. The system was capable of detecting cadmium at

levels as low as 0.005 microgram per cubic meter of air.49

Other inductively heated graphite tube furnaces were pro—

posed by Headridge and Smith,50

8611.51

and by Langmyhr and Thomas-

The advantage of continuous sample introduction first
realized by Woodriff has also been retained in several
additional devices. Veillon and coworkers have described
a vertically oriented graphite tube furnace, useful in
atomic fluorescence, which receives sample continuously

52

in the form of an argon carried aerosol. Wineforder

and his students have explored techniques for continuous

16

sample introduction into furnaces through the use of a
pneumatic nebulizer and steel desolvation chamber.53’54
Papers have also appeared in which already existing
atomizers have been modified to improve their performance.

Chapman, Dale, and Kelly reduced the diameter of a fur-
nace on either side of the sample deposition region to
produce regions of greater heat that would retard the dif-
fusion of atomic vapor out of the observation window.55
Robinson and coworkers have described techniques in which
a sample in a complex matrix is placed into the side arm
of Woodriff-type furnaces for extensive pretreatment de-
signed to destroy and flush out the matrix. The analyte
of interest can then be atomized relatively matrix

56'57 Successful direct analyses for trace metals

free.
in such difficult matrices as blood, urine, and sea water
were reported. Church et 31. have designed a low tem-
perature Woodriff-type furnace for mercury determinations
in which oxygen is actually drawn through the furnace to
assist in sample matrix combustion.58
Several authors have performed investigations designed
to improve and extend furnace methodology. Veillon and
his students have experimented with techniques to deposit
and maintain pyrolytic graphite coatings upon atomizers,
through use of a constant supply of methane gas at trickle

59,60

flow rates. Similar studies have been reported by

61 62

Molnar and Winefordner, and by Sturgeon and Chakrabarti.

Runnels and his associates have proposed a related process

17

in which the coating is a refractory metal carbide. Im-

proved performance of the furnace in analyses of metals

(1.53

prone to carbide formation was observe Fuller and

Thompson have evaluated some concepts of solid sampling

64 Thomassen gt 31. have proposed analyses in

techniques.
which the trace metals of interest are electrodeposited on
graphite sticks, which are then scraped, and the resulting

65 Extensive explora-

powder analyzed in a cup-like device.
tory work has been performed by Ottaway and coworkers to
extend the applications of furnaces to atomic emission
analysis of elements such as the alkali metals and alkaline
earths.66 DeGalan has published a critical study of some
of the practical pitfalls and implications of furnace
atomic spectrometric work.67
Applications of furnace type atomizers to specific
analyses are extensive; dozens of references could be
cited.1'2'26 Some representative papers from only the

past few years may be given here as examples. Among the

clinical and biological methods may be listed analyses

68 69 70 71 d72

and lea
75

chromium,
74

cadmium, gold,

73

for aluminum,

in blood, mercury, and lead in urine, or chromium

76 in tissues. Many applications may be found

77

and selenium

for geological samples, including cadmium, silver, lead,

78

thallium, and zinc in rocks. Environmental samples con-

tinue to be of interest, with methods reported for silver,

79

beryllium, cadmium, and lead in air, cadmium in seawa-

80

ter, or selenium in industrial effluents.81 Among the food

‘m

n:

'1‘

I").

0‘.-

(1

(ll

n;

K).

()1

7

18

analyses may be included aluminum, chromium, copper, lead,

82 83

and zinc in mussels, and cobalt in feed grains.

Metal and alloy analyses are well represented also, with

84 silver,

d'86

techniques available for aluminum in steels,

85

bismuth, and cobalt in tin, iron in silver and gol

and lead, bismuth, selenium, tellurium, thallium, and tin

in metal chunks.87

D. FILAMENT ELECTROTHERMAL ATOMIZERS

The furnace devices were the first to be developed
and are the most numerous among commercial instrumentation
in the field. Aside from the comparative advantages and
disadvantages with respect to flames previously mentioned,
furnaces as electrothermal atomizers have the advantages
of solid sample capabilities, and a more flamelike environ-
ment for the sample in that it is surrounded by the hot
atomizer. Samples are thus more likely to be efficiently
atomized and kept in the atomic state. Problems associ-
ated with their use include their bulkiness, a necessity
for cooling facilities, and the requirement of a power
supply of arc welder proportions to heat them rapidly
to the requisite high temperatures. Because of these
detractions, the alternative type of electrothermal
device, known as the filament atomizer, was developed.

The first filament atomizer was that reported by

T. S. West and his colleagues in l969-l970.88"91 It

 

(7'

19

consisted of a filament originally 25 mm long by 2 mm
in diameter with a notch grooved in the center to take
the sample. It was secured to stainless steel electrodes
and enclosed in an argon purged chamber with quartz
windows. Performance of this device in both atomic
absorption and atomic fluorescence was reported for
about twenty elements, with detection limits roughly
comparable to the furnaces.

Belyaev and coworkers developed a solid sample
graphite atomizer shaped like a miniature cup, and
used to examine samples prepared in graphite powder.
Representative determinations included cadmium and silver

92-95 A commercial device based on the fila-

96,97

in rocks.
ment design was marketed by Varian Techtron. It is
actually a miniaturized Massmann furnace consisting of
a rod 9 mm in diameter with a 3.5 mm transverse hole,
supported by water cooled graphite electrodes. This
atomizer has the provision for supplying hydrogen gas
around the filament during analysis, allowing the user
to suppress many matrix effects.

Several other authors have also reported filament
type devices, most of which are not too dissimilar from
the original West concept. Such atomizers would include

98

the designs of Dipierro and Tessari, as well as the

atomizer used in this work which was reported by Crouch

99-101 Still another device was reported

102

and Montaser.

by Molnar and Winefordner, which was surrounded

9h

(n

\

'71

It“

5"

20

during use by a laminar hydrogen-argon-entrained air
flame to assist in preserving the atomic population.

Among the filament devices, there are also a variety
of atomizers which are made of materials other than graph-
ite. Donega and Burgess developed an atomizer prepared
from 25 micron thick tantalum or tungsten foil, and
shaped into a boat structure 50 mm long by 6 mm wide,
with a depression for containing the sample solution.103
The device was provided with a quartz envelope, and oper-
ated at subatmospheric pressures. A simplified version
of the Donega and Burgess device was developed by Hwang
and coworkers, and eventually became the Flameless Sampler

104

produced by Instrumentation Laboratory. Other devices

in this category include the tantalum strip of Takeuchi
1.105

the tantalum filament of Maruta and Takeuchi,106

107

93
and the tungsten filament of Cantle and West.
A closely related class of nongraphite atomizers

are the wire loop devices. Tungsten and platinum loops

were investigated by Winefordner and coworkers.108-110

Their loop had a diameter of 1/32 inch, and receives
sample solution through a dipping process. Crouch gt 31.

investigated automated analysis in atomic fluorescence

111

using platinum wire loops. A tungsten filament atom-

izer was proposed by Williams and Piepmeier using the

112

tungsten element of an ordinary light bulb. Chauvin

and colleagues used loops prepared from an exceptionally

113

high melting tungsten-rhenium alloy. Other similar

21

atomizers would include the tungsten filament of McCul-

114

lough and Vickers, the molybdenum ribbon of Ohta and

Suzuki,115 the molybdenum filament of McIntyre, Cook,

and Boase,116

117

and the tungsten wire 100p of Newton and
Davis.

Applications of the filament and loop type atomizers
to real analysis problems are competitively numerous with
the furnace methods. The nature of the matrices en—

countered are similar as well. Thus, reports have been

offered on the determination of zinc,118 copper,119 and

120 121

in blood or serum, zinc in saliva, and

h.122

lithium
Arsenic, antimony, and tin have been
124

cadmium in fis
123 iridium in
126

examined in steel, selenium in c0pper,

125

noble metals, and selenium in copper and iron.

Filament techniques have been applied to the analysis of

128 129 in water.

130

arsenic,127 beryllium, and manganese

Cadmium has been examined in air particulates, as has

131 Rock samples have been analyzed

132

zinc in sediments.

and the cobalt
133

for gold, cobalt, lead, and vanadium,

levels of soils have received attention. Additional

successful analyses would include those of metals in jet

134 gold in photographic film,135 barium and

136

engine oil,

chromium, c0pper, and iron
k.138

antimony in gunshot residue,

137 and lead in canned mil

in polymers,
Compared to the furnaces, a filament atomizer enjoys
the prime advantage of not needing nearly as much elec-

trical power to reach the highest analysis temperatures,

22

which makes a corresponding reduction in the size of both
the atomizer and its supporting facilities. The filament
devices can also be more beneficial in atomic fluores-
cence because their smaller size cuts down on the amount
of background radiation observed by the detector. On

the negative side, the filament devices do not sustain
the sample after atomization. Unlike flames and furnaces
which surround the sample with a hot medium, the filament
expells the sample into the sheath gas, where it is
rapidly cooled and subject to condensation. As a result,
it is often observed that analyses in difficult matrices
can at times be better done in a furnace. Several
workers have observed, however, that use of hydrogen in
the sheath gas to ignite and sustain a hydrogen dif-
fusion flame during atomization is greatly beneficial to

96,102,140

preserving the atomic population. Confining

the optical field of view to the region immediately above

the filament is also of merit.139

E. SAMPLING BOAT AND PROBE TECHNIQUES

Aside from the purely electrothermal techniques,
there are also several other classes of atomizers which
exploit chemical peculiarities of specific systems, or
attempt to combine virtues of both the flame and electro-
thermal methods. Prudnikov has explored the enhance-

ment of flame method sensitivities through introduction

23

of liquid or solid samples directly into the flame on
the tip of a graphite micrOprobe, which may or may not
be externally heated.141-143 A technique employing a
tantalum sampling boat was developed by Kahn and cowork-

ers,144 and by Ringhardtz and Welz,145

in which a sample
is pipetted into a small tantalum trough, dried by

gentle heating, and then pushed reproducibly into a
conventional flame. A variety of clinical determinations

146 147 have

such as cadmium in blood or thallium in urine
been performed with this technique, and the boats them-
selves are available commercially. Perhaps the best

known of these combination approaches is the Delves cup,
which combines the concepts of mechanical sample introduc—
tion with the use of an absorption tube resident in the
flame. This tube serves as a containment chamber for

the gases of the atomized sample during analysis.148
Delves intended his method to address the problem of lead
poisoning, through analysis of lead in blood. An air-
acetylene flame is used, and the absorption tube is manu-
factured from nickel. A 10 ufi sample of blood is placed
in a nickel cup, dried, and treated with peroxide. It

is then inserted into the flame for analysis. For
determinations of lead in blood, it is extremely suc-
cessful, and has been developed into a rapid, mass screen-

149-152

ing clinical procedure. Other researchers have

extended its capabilities to elements other than lead,

and to matrices other than blood.153'156

24
F. CHEMICAL ATOMIZATION TECHNIQUES

A final field of atomic spectrometric techniques which
do not employ flames for production of atomic vapor are
the chemical vaporization processes for mercury, and for
elements forming volatile hydrides. The latter class
of elements includes arsenic, antimony, selenium, bismuth,
germanium, tin, and tellurium. The concept of determin-
ing these elements by volatilizing them out of solution
as an easily decomposed hydride was first proposed by

157 The hydride

Holak in a flame determination of arsenic.
was generated by the reaction which takes place in acidic
solution in the presence of potassium iodide, stannous

chloride, and granular zinc. At a later date, the method

was modified by several workers to eliminate the flame

and substitute an absorption tube heated to about 700°

158-160

C Another improvement was the use of sodium

borohydride to generate the hydride species,161-163
which allows a significant reduction in the time required
for hydride generation.

The cold vapor technique for mercury takes advantage
of mercury's unusually high vapor pressure and ease of
reduction to the atomic state. The cold vapor method

164

was first proposed by Poluektov and coworkers, and

has since developed rapidly into a method of choice for
determining mercury in just about every matrix encountered.

The technique generally consists of reducing mercury

25

to the atomic state with stannous chloride solution,

and sweeping the resulting atomic vapor out of solution
and on to an absorption tube for analysis. Representa-
tive recent determinations would include mercury in sedi-

165 167 168

ments, grain,166 fish, and oil. The method has

also received attention from a critical standpoint, and

has been the subject of optimization studies.169
The popularity of the volatile hydride and mercury

cold vapor techniques arises from their convenience,

and from the fact that they provide an unusual opportunity

to combine the actual analysis with an effective and highly

quantitative separation of the analyte from even the most

complex of matrices. Also appealing is the simplicity

of the extra apparatus which must be assembled to convert

existing instruments to such methods.

G. SPECIFIC LITERATURE BACKGROUND OF

THE PRESENT WORK

There are some direct literature precedents for
portions of the present work. The concept of using var-
ious electrical and optical parameters as a means by
which to control the heating of an electrothermal atomizer
was explored by Montaser and Crouch using the graphite
braid.101 A total of four possible parameters were ex-
plored as candidates for serving as the feedback control

signal in atomizer temperature regulation. They are

26

the voltage across the atomizer, the current passing through
it, the power dissipated by it, or the blackbody radiation
emitted from it. The first two methods, although better
than an unregulated system, were not very acceptable as
control signals, due to their inability to correct for
changes in atomizer characteristics with age. Combining
both voltage and current into a signal representing
electrical power improved the situation somewhat, but
the greatest benefits were definitely seen using the
blackbody emission technique for atomizer control.
Similar observations were noted as well by Lundgren,
Lundmark, and Johansson, who used a photodiode in combina-
tion with a triac-controlled switching regulator for
governing the heating of a Massmann-type furnace.170
These authors noted the superior precision of tempera-
ture control under the radiation method, and demonstrated
how the improved rate of atomizer heating can be put to
good advantage in real sample determinations, such as
cadmium in sea water.

The exploration of atomic populations at various
positions with respect to an electrothermal atomizer
has been attempted by several workers. Winefordner and
his students examined the decay of atomic populations up
to 18 mm above their filament device through use of both
atomic absorption and atomic fluorescence techniques.171'172

A number of elements were investigated, including silver,

cadmium, copper, mercury, lead, tin, thallium, and zinc.

27

The decay was generally observed to be exponential in
shape. Dramatic improvements in both the initial concen-
tration and the later preservation of free atoms were
observed for most elements by including hydrogen in the
sheath gas flow. This was attributed to the efficacy
of hydrogen in breaking up metal oxide species and in
serving as a general scavenger of oxygen.

West and coworkers have briefly examined the decay
of atomic populations with vertical height for several

139’173-176 The intent

elements above their filament.
here was principally to locate an Optimum height above

the filament for analyses of the element in question.
However, a few specific matrix studies were also performed.
Finally, Torsi and Tessari have studied vertical concen-
tration profiles in the experimental testing of their
theoretical models for the production and dissipation

177 In all of

of atomic vapors from filament atomizers.
these systems, the ability to examine populations was
limited to the vertical dimension only. In addition,

the instrumentation was not automated under computer
control.

The use of electron microscopy to examine electro-
thermal atomizer appearances appears to be rather un-
precedented, except for an occasional reference in which
the technique was applied to the examination of a Specific

point of interest.62

COMPUTER CONTROLLED INSTRUMENTATION
A. INTRODUCTION

As has been previously mentioned, the nature of much
of this investigation is such that extensive computer
control of many aspects of the necessary instrumentation,
including mechanical devices as well as traditional real
time data acquisition, is all but indispensable. The
monotonous, time consuming tasks of depositing samples,
locating the atomization cell, and cycling the atomizer
itself constitute a great potential source of procedural
errors if done manually. In addition, manual execution
is subject to inconsistencies in the timing of various
portions of a complete instrumental cycle, even if the
cycle itself is performed error free. A computer, how-
ever, is well suited to the repetitive performance of
a defined task, and furthermore will do so from cycle
to cycle with a consistency as rigid as the executing
software and instrumentation under control will permit.

For achieving such computer control in the present
work, three sophisticated devices were designed and con-
structed. The first is what is commonly referred to as
the "positioner", which controls the movement of the
atomization cell in two dimensions with respect to a
fixed optical axis, for the purpose of scanning the array

of space to be included in an atomic concentration profile.

28

29

The second is the "sampler" which, through computer con-
trol, automatically deposits samples of a specified size
and source on the atomizer. The third is a dedicated,
multifrequency real time clock and data acquisition system
for on—line observation of signals produced during atomiza-
tions. These instruments will be discussed in the follow-

ing sections in the order just listed.

B. THE POSITIONER

1. General Introduction and Overview

To develop two dimensional profiles of atomic popu-
lations above an atomizer, it is necessary to have some
means by which a systematic scan of the two dimensional
space may be performed as samples of the solution under
investigation are repetitively atomized. This require-
ment is met in the present work by locating the optical
axis of an atomic absorption instrument at a fixed posi-
tion in space. The atomization cell is then mounted on
two stacked, Optical translation stages driven by lead-
screws, which permits the cell to be independently dis-
placed vertically and horizontally with respect to the
optical beam. The motive force for the leadscrews is
supplied by two stepper motors, chosen because of the
ability of a stepper motor to rotate through a precisely
defined angle in either direction, as well as the inherently

digital nature of such motion.178

30

To control the orderly movement of this mounted cell,
an instrument known as the positioner was designed and
constructed. The positioner is a self-contained device
which independently moves the atomization cell in the
vertical and horizontal directions under the guidance
of internal logic sequencers. The instrument in general
is built under the philosophy of providing computer trig-
gered, but not software controlled functions. In other
words, the computer has only to give the appropriate com-
mand and supply the proper data, and the positioner will
then internally assume all responsibility for executing
that command properly. No further interaction with the
computer is required until the done flag is signalled.
Thus, the device's own logic governs such functions as
the distance and the direction of the cell movement.
Operation is either computer controlled or manual, as
selected by a front panel switch.

The positioner performs two basic cycles of Opera-
tion, known as GO and RESET. When a G0 command is received,
the positioner moves the atomization cell to the vertical
and horizontal destinations specified. When requested to
RESET, the cell is moved steadily in a fixed direction
until the moving stages encounter mechanical boundaries
defined by miniature optointerrupters. The optointer—
rupter and a piece of metal are mounted on appropriate
structural surfaces of the assembly so that as movement

in that respective direction occurs, the metal piece

31

and optointerrupter either approach or recede from each
other. When a RESET is requested, the cell is moved until
the metal piece intercepts the beam of the interrupter,
and when this occurs, that location of the cell in that
dimension is defined as location 0. In this way, the
positioner is capable of calibrating the absolute location
of the cell at any time. (This feature is chiefly of
importance when the positioner is first powered up, and
has no knowledge of where the cell might be.)

Because the computer used with these studies is a
twelve-bit machine, and is thus capable of directly
representing decimal integers between 0 and 4095, it was
decided that the complete segment of distance over which
the cell could be moved in either dimension would be
divided up into 4096 equally spaced locations, numbered
sequentially from 0 to 4095, with location 0 being the
position of the reset boundary. The convention selected
for assigning these locations is shown in Figure 2. When
the cell is so positioned that the optical beam is directly
above and centered over the atomizer, (the condition
known as "grazing incidence") the vertical dimension is
at location 0 and the horizontal is at the midpoint, or
location 2048. Movement away from the interrupter (towards
locations of higher number) is defined as being positive
motion.

The translation stages upon which the cell is mounted

have a total displacement of approximately one inch, and

32

    

ATOMIZER

OPTICAL AXIS

Figure 2. Assignment of atomizer cell locations.

33

are driven by forty-thread-per-inch leadscrews. The
stepper motors turning the leadscrews take two hundred
steps per revolution. Therefore, eight thousand motor
steps are required to translate the cell one inch in the
given dimension. With the highest location being numbered
4095, it was decided that adjacent locations would be two
motor steps apart. Thus, the movement from one end of

a dimension to the other requires 8192 motor steps, which
results in a net actual displacement of 1.024 inch, and
which permits a scan of an area slightly in excess of one
square inch above and horizontally centered with respect
to the atomizer itself.

Before discussing the detailed logic of the positioner,
it would be helpful to gain some general insight into its
operation with the aid of Figure 3, which illustrates in
block form the basic interrelationships among the instru-
ment's major register arrays, of which there are two
identical such units, one for each motor.

When operated manually, the user enters the desired
horizontal and vertical locations of the cell as numbers
which range from O to 4095. The numerical locations are
entered from separate, front panel, four decade thumb-
wheel switch assemblies. These switches are coded in
a BCD format. When a toggle switch is momentarily de-
pressed to issue the GO command, these destinations are

loaded into the sixteen—bit decimal register, after which

34

.owmoH Houucou umGOwuwmom mo Sandman xoon

 

4

 

 

a

 

 

 

 

 

nourh

 

 

 

 

 

 

 

 

 

 

 

C

 

 

 

 

 

 

 

 

N: can

 

no:

 

 

 

 

 

 

 

 

 

NIX onllv

 

«math

3

 

 

 

 

 

"0:1.

 

 

 

 

 

3

HDQHDO

:>

 

 

moe<qaznuu<

 

ocuNAIIH

 

a

 

 

N13 cull-J “month
2809

 

 

 

 

L7

 

 

Nauru.

 

 

3

“ Nauru

 

 

3

.m «Human

 

 

capo—om:
.oncn zo_p<uoa
noes ¢o2<¢<axoo
cupoaouc
.uars zo_»<z-muo
cupmuouc
IIIIII, nodes 4<z_uun

 

 

3

wmIQPHZw 4mmIzszIk

35

a BCD-to-binary conversion occurs by clearing the twelve-
bit destination register, and then simultaneously decre-
menting the decimal and incrementing the destination
registers until the former reaches zero. Should the

user enter a value larger than 4095 from the thumbwheel
switches, it is detected by the code-converting logic

and causes the sequence to abort prior to motor movement,
a condition known as "overflow". When operated from the
computer, the destinations are loaded in binary format
directly into the destination register.

The destination register is compared at all times
through the use of cascaded four bit digital comparators
with a third twelve-bit register, known as the location
register, which contains at all times the actual location
of the atomization cell. Once a valid destination has
been either loaded or clocked into the destination regis-
ter, the comparators detect whether or not this destina-
tion is different from the current location, and, if so,
in what direction the new destination lies. The motors
are then stepped toward this new destination, and, simul-
taneously, the location register is incremented or decrem-
ented at one half the motor stepping rate. This con-
tinues until the comparators once more detect a state of
equality between the two registers. A final procedure
designed to eliminate "lost" steps due to mechanical
looseness in the translation stage leadscrews may then
take place; it is described in detail in a subsequent

section.

36

2. Sequence Structure and the Flow of States

The flow of logic in the positioner that executes
all of the individual details of the two basic cycles in
proper order is supervised by two identical, sophisticated,
yet simple decision-making sequencers based on a design

179,180 A basic understanding Of

by C. L. Richards.
the terminology and function of this sequencer, given in
Appendix A, is essential in comprehending the details of
the circuitry to follow.

The flow of states and functions involved in a G0
cycle are illustrated in Figure 4. While idle, the se-
quencer holds in a wait loop formed by state 0, a condi-
tion indicated by a front panel "READY" LED. When a G0
command is received, the overflow flag is cleared and the
sequencer determines whether operation is under manual
or computer control. If manual, a BCD-to-binary conver-
sion of the destination is in order. The system checks
for overflow, and if overflow is not present, it checks
to see if the BCD-to-binary conversion is complete. If
conversion is not complete, the two registers are clocked
as previously described, and the checks for overflow and
end of conversion are repeated. This cycle continues
until conversion is complete, or until overflow occurs.
Should overflow happen, the cycle is aborted with the
overflow flag set, as shown by a front panel "OVERFLOW"

LED. For legal destinations the end of the BCD-to-binary

37

.maowo ow m sH chfluocau can mmuwum Mo 30am .w wusmwh

 

mcupmnomc
znm 4000
20040

 

 

 

 

 

u ZOMPUZDu

 

 

 

     
      
    

 

up<o
wwoqu fimzoo
AMWVA zo_ocu>zoo xommnzav wnwuumuwn
zo.hwwc_o > 22m-oum >
»

 

 

 

 

 

 

 

 

w zoahuzau m ZO—huzau

 

 

 

 

 

o :o—Pozzu

 

 

 

 

up<o
macho cupm~muc o<sa
AHHVAIIII >¢<z_n oAanwwu zesacu>o
zo_pww¢.a ¢<uou ¢<uoo
»

 

 

 

 

 

 

 

 

 

O 20~h023u U ZO—huzau < zonputzu

 

 

 

38

 

o mr<hmATI|l

 

 

 

4 zo—ruzau

.cmscwucoo

 

 

 

 

 

x zonhuzau

 

 

 

 

Cupm—omc

 

 

 

3 ZO—rozau

 

 

 

 

   

  

uh<o
zumo IIILVO mr<hw
— zo—Puzau
bomzu<mc
Anwwazav zcnp<004

 

  
  

 

 

 

1 zonhuzau

 

.v madman

39

conversion permits the sequence to advance beyond state
three and start moving the motors in the direction indi-
cated by the comparators. As previously mentioned, if

the system is under computer control, no BCD-to-binary
conversion is needed (and no overflow is possible). Thus,
the sequence jumps directly from state one to state four
and motor movement.

The motors move, and the location register is clocked
until the new destination is reached. At that point, the
sequencer reminds itself of whether the movement just
performed was positive or negative. If negative, the GO
cycle is complete, and the sequencer returns immediately
to the state 0 wait loop with the READY LED lit once more.
If motion was positive, however, the process of moving
and counting locations continues upward ten extra locations
beyond the nominal destination, at which point the motor
direction and location register clocking are reversed,
the ten extra locations are counted back, and the cycle
concludes. In this manner, no matter whether the net
movement was positive or negative, the cell is moved
last in the negative direction, leaving the leadscrews
tightly meshed in that direction of movement.

The flow of states and functions for a RESET cycle
is basically that of an abbreviated GO cycle, as illus-
trated in Figure 5. The cycle begins in the same manner

as a GO operation to a destination of zero would, except

40

.maomo Bmmmm m :e maceuocsu

 

o up<prTIII up<o

zuuo

 

 

 

— zonrozau

 

 

@fi

 

 

uh<o

umogu

zo—houxnn
u>nh<ouz

 

 

 

O ZO—Pozau

GAIL

zone mane
u>~h<ouz

 

 

 

O zo—huzau

 

curmnomc
>c<znm
c<m40

 

 

 

u zonrozau

 

 

  
    

kaochzoo
4<004

 

can mmumum mo 30Hm

.m wusowm

 

Anumazav

 

   

  

Gouda—Ch
COFADCCth—
uohao

 

 

I Zo—huzau

 

 

namwazav

 

  

03°4u¢u>o

 

 

u zo—huzau

 

 

 

 

 

< zo—huzzu

41

that the location register is cleared at the outset and
held in the clear state throughout the cycle. If under
manual operation, no BCD-to-binary conversion is needed
nor is overflow possible, since by definition of RESET
the destination is inherently zero. Thus, the branches
of the cycle associated with these steps are missing as
they will never be followed. Motor movement commences,
and continues not until location zero is reached, but
until the respective optointerrupter is tripped. At

this point the cycle concludes with the location register
defined equal to zero through its having been held clear.
There is no possibility of positive motion overcounting,
as again, the definition of a RESET automatically implies
negative movement, so once more, this section of the flow
is missing.

The specific implementation of a sequencer based on
these cycles is shown in Figure 6. The structure of the
sequencer is based squarely upon the suggested Richards
design with the exception of the three gates which pass
the state address lines to the function decoders. Addi-
tion of these gates was found necessary to assure that
the total propagation delay involved in responding to a
state change would be shorter for the multiplexer than
for the decoders. Should the opposite be true, the pos-
sibility exists that false function pulses may be produced

by the decoders until the multiplexer catches up with them.

42

REE t

EEEEEBKE

 

llllllll

OHNOCDOB

«.129

N12 0. lb
840

 

molllll

 

 

 

 

 

 

.uwocwsomm a mo Bouquet oeumfimnum .m
PE... h.
a J- t
c .— Hg
a
ham 0 Afilenh uh
ounce
azu o An*Ylhu
Bu<o<

 

 

 

 

 

 

 

 

 

 

 

 

£41441

 

 

 

no

Rflr
Qt

 

 

 

 

 

 

L

 

 

 

Inrla

 

.anoc-noh

HlllHl

.nNN'noh
00000000

 

 

whomem

uh.
PM

43

(At the time, the preferred technique of solving this
problem through use of, say, a Schottky multiplexer chip
was not possible due to lack of the prOper IC. The ex-
pedient solution shown here has thus survived in the
present instrument.) Two identical such sequencers are
present in the positioner, one for each motor. In fact,
with the exception of the clock circuit discussed next,

all circuitry of the following subheading, "Detailed Logic",

is similarly duplicated in the instrument for each motor.

3. Detailed Logic
a. Dual Synchronous Clock Source

The instrument logic and motors are driven by a dual,
synchronous clock source. This clock is shown in Figure
7 along with a representation of its output waveform.
A base frequency of 10 kHz is supplied by a 555 timer
configured for astable operation. This is divided to
10 KHz by two flipflops which are wired to toggle on
opposite edges of the source. The output of one of these
flops is used as the clock source for driving the logic
sequencers. The output of the other flop is further
divided by a factor of 20 to produce a frequency of 500
Hz, used for stepping the motors and clocking the location
registers. (The remaining circuitry which acts as a
gate upon the output of this 500 Hz clock is discussed

later.)

44

.wou50m xooao msocounocwm .stp may no EMHOMHU owumfiwnom .5 gunmen

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N: n.~w

 

 

 

 

 

 

 

 

 

 

 

 

 

N: n.~¢

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

45

The use of two flipflOps to perform the first divi-
sion of the source frequency, and the interwiring of these
flops with each other's outputs and J-K inputs assures
that the 10 KHz and 500 Hz clocks always bear the timing
relationship to each other shown in the waveforms. When
the instrument was first completed, the two clock fre-
quencies were derived from separate, asynchronous 555
timers. This led to occasional anomolous instrumental
behavior due, it is believed, to periodic, random syn-
chronizing or near synchronizing of the clocks such that
certain portions of circuitry required to respond at times
to both clocks could not do so reliably. With the present
clock, the relationship of the two frequencies is always
rigid, and ample time between an edge of one and an edge
of the other always exists for the satisfaction of propa-

gation delays in the logic.

b. Sequence Initiation Circuitry

Circuitry involved in initiating the instrumental
cycles is shown in Figure 8. The front panel switch,
which selects manual or computer operation, controls a
set of multiplexers which pass GO or RESET commands
originating either from the computer or from debounced
front panel switches. When given, either command from
either source will cause the clearing of the READY flip-

flop, which extinguishes the corresponding front panel

46

.xuuesoueo cowumHuflce mocwsomm on» no asymmep oeumEosom .m mesmem

 

 

 

w
cl
‘ I

:hmwmma

 

 

 

 

 

 

 

 

 

uh

 

 

mF----

 

 

l
y.

 

 

 

 

 

 

 

 

 

 

so .L
[1» m H u
<
i n,

tow:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

47

LED and satisfies condition 0 of the logic sequence to
begin the cycle. This flop may be reset to the ready
state, indicating a completed or aborted cycle, by any
of the sequence concluding function pulses or by the
INIT pulse. A second flipflop of the RS variety is used
to distinguish the GO from the RESET command, being set
by a RESET command and cleared by the same sources as
the READY flop. The multiplexers' select line also serves
as condition 1 to the sequencer.

The remaining three sections of logic to be discussed
systematically follows the actions taken by the instrumen-

tal logic in executing the sequences.

c. Destination Preparation Circuitry

The first major task of the sequencer is to perform
the BCD-to-binary conversion of the destination, should
the positioner be under manual control. The circuitry
associated with these steps of the sequence is illustrated
in Figures 9 and 10, which show the decimal register,
the OVERFLOW flipflop, and the actual wiring of the major
registers shown previously in block form.

A manual GO command causes the loading of the desti-
nation from the thumbwheel switches by signal LD'BCD.

An example of one of the sixteen input lines to the deci-
mal register is shown. The thumbwheel switches produce

complementary BCD code. This is inverted and passed to

48

.muwumemwu HmEflomp on» mo Enumwwp owumfiwnom

 

Ow.— ~304u¢u>0~

NU >¢¢<u

 

 

.m musmwh

com ._0

49

.mumumflmmu umcowuwmom Homeoceum mop mo Emummep owumaonom .oH gunmen

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

¢ 4 c3 8..
oun<n§ coo \ no o3
STIR... 13.1.1312. n .1 an E
1o..s so no; use
a u a < a. -

 

 

a? E

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2:29 IIIIEM u a pzullu 51% o n :5 k 5% u n pzuUIn.
.: 1 1111 1111a t
_ _ _ :1 _ at
h «a a b.14 O m n..P < ._.; I “a o “u <

50

the register. The sequencer passes from state 0 to state
1, which generates pulse FA and sets the OVERFLOW flip-
flop. The test for manual operation is performed and,
since at present manual Operation is assumed, the test
will be true, which causes pulse F5 to clear the destina-
tion register in preparation for the conversion of data.
The test for absence of overflow takes place, and it will
be initially true. Pulse FE is generated, but not used
anywhere. The test for complete conversion takes place
next. The condition for complete conversion is derived
from the NOR gate which monitors the BORROW lines of the
decimal register IC's. If the destination entered happen-
ed to be zero, the conversion would be "complete" since
none is needed, and the logic would proceed to state 4.
We will assume a nonzero destination however, so that
conversion is not complete and condition 3 is false.
Pulse FF is then generated, thereby decrementing the
decimal and incrementing the destination registers by
one, and backing up the sequencer to state 2. The over-
flow test is performed again, followed by the conversion
complete check, etc., until conversion is complete or
overflow occurs.

An overflow is indicated by the destination register
having reached 4095 with conversion still not complete.
When the destination register reaches 4095, signal CARRY
will be asserted, and if an additional FF pulse occurs

in an attempt to continue the conversion, these two

51

signals will together clear the OVERFLOW flipflop. The
next test Of condition 2 will then be false, and the cycle
will abort through generation of pulse FD, leaving the
front panel OVERFLOW LED lit.

If conversion does complete itself at or before the
assertion Of CARRY however, the zero state of the decimal
register will be sensed by the NOR gate already mentioned.
Condition 3 is thus satisfied and the sequence is allowed
to proceed.

The discussion has thus far assumed manual operation.
If computer operation is in effect, the test for condi-
tion 1 will be false, and almost all elements of the se-
quence that deal with the above circuitry are skipped.
Instead, prior to starting the sequence, the destination
is loaded directly into the destination register with the
signals EB DR and CDATA (which are both required tO
defeat the synchronous loading characteristics Of a 74161)
after which the cycle is initiated by issuing CEO. The
actions of loading the destination and starting the sequence
are separate under computer control, since the computer
can supply destination data for only one dimension at a
time. With manual operation, the loading and starting
processes are simultaneous because of the separate thumb-
wheel switch assemblies.

Most of the above circuitry is also unused if the
given command is a RESET. A manual reset command asserts

the signal CL BCD which puts a destination of zero in

52

the decimal register, satisfies the tests for no overflow
and complete conversion in rapid succession as previously
shown, and allows the sequence to proceed on to state 4.
Similarly, a computer originated reset clears the destina-
tion register with signal CRST (which unlike loading is

an asynchronous event in a 74161) and proceeds directly
from state 1 to state 4. Note that in the event of a reset
sequence, the RESET RS flipflop asserts the signal RESET
throughout the entire sequence, and, thus, holds the loca-

tion register in the clear state.

d. Destination Seeking Circuitry

The next major section of circuitry is that which
controls the movement of the motors in the proper direc-
tion to the desired destination and simultaneously clocks
the location register. This circuitry is shown in Figures
11 and 12 and acts identically for either manual or com-
puter controlled Operation.

The gating Of 500 Hz pulses into the motor driver
and location register is governed by the RS flipflop
formed by gates A and B. This flop is set by either FE
or F5, the two pulses which advance the sequence into state

4 depending on the source Of control. It is cleared by

 

INITlor either of the two pulses FT and FL which conclude
a sequence after state 4. The JK flipflop accomplishes

the task of clocking the location register at one half

53

.muuenoueo Houucoo oceuomuwc paw mwumm wnu mo scammeo Oeumemnom .HH muamem

 

c8 3..
2m Reopen. allusions
3H1]. ..

«19 coho:

EB

 

 

 

no

Fug-om .Pn.
25 Solar 17%| 5. so» 8.731.
COPO! MAO - o
ooa soullo x . rnnun
.A‘ .i

P...
uh

 

 

 

 

 

 

E1...

 

54

 

.OflmoH uomumc e :Oeuepcoo on» no Emummep OeumEmnom .NH mnsmam

 

 

 

 

 

 

 

 

 

TT

 

 

 

 

Ni! O—

55

the motor stepping rate.

The direction of movement is indicated by the state of
the signal DES > LOC produced by the cascaded comparators.
If high or low, the same pulses which close the gate will
set or clear, respectively, the direction flipflop, the
complementary output Of which controls the direction line
on the location register.

The final section Of circuitry shown in Figure 12
is a multiplexer for controlling the assertion Of condi-
tion 4, which depends on whether the cycle in progress is
a G0 or a RESET.

TO understand this circuitry, consider first a GO
command. As state 4 is entered and motor movement is to
begin, the pulse FE or F5 closes the gate and properly
prepares the state of the direction flipflop in light of
signal DES > LOC. The action Of the gate is very care-
fully designed to assure that the stepping of the motors
and the clocking of the location register at half that
rate is done without ambiguity. (The success Of this
precise Operation depends in large measure on the rigid
relationship between the two clocks, as previously dis-
cussed.) Until the gate closes, the location register
is disabled by signal LOC EN, and the JK flipflop is in
the set state by virtue Of the low level at its K input.
The fact that this flop is set assures that the clock

line of the location register is high when idle, a condition

56

required by the 74191 IC if its clock direction line is

to be unambiguously assigned, as is the case here. When
the gate closes, the register is enabled, and the next

500 Hz falling edge toggles the flop and steps the motor.
The next falling edge does the same, and in addition,
through the flop, clocks the location register. Thus,

the stepping Of the motor and clocking of the register

are in phase, and the divide by two process on the register
clock line is conducted properly.

When eventually the gate Opens again, the divide by
two flipflop will have just toggled into the set state,
and clocked the location register the final time, which
causes the comparators to once more assert DES=LOC. This
signal, in combination with RESET (which in effect means
GO), then satisfies condition 4 Of the sequence. Pulse
_H is generated but not used, and the test for positive
or negative motion is performed, with the output Of the
direction flipflop tested for condition 5. If the motion
just performed was negative, condition 5 will be false,
and pulse RI will Open the gate and conclude the sequence.

A special case occurs if the destination called for
happens to be the location at which the cell already is.
In such an instance, as state 4 is entered, the DES=LOC
line will still be asserted, and the combination of that
signal with RESET will prevent FR or F5 from closing the
gate. Simultaneously, DES > LOC will be low, allowing

FB'or FG'tO put the direction flipflop into the clear

57

state, and causing the sequence to conclude with pulse

FT, thereby "arriving" at the new destination without having
to move. This special case occurs quite Often in actual

use Of the positioner, as one dimension is scanned while

the other is held at a constant location.

In the event of a RESET command, the action of the
gate is similar to that of a G0, except that the pulses
which attempt to decrement the location register are
disregarded, since that register is held clear from the
outset during a reset cycle. The direction flipflop is
inherently cleared, and the sequence always concludes
with condition 5 false and pulse FT generated. The one
distinctive aspect of the reset cycle is in the circuitry
used to control the assertion of condition 4. During a
reset, satisfying condition 4 means tripping the opto-
interrupter. When this occurs, the tripping Of the inter-
rupter is squared up by a Schmitt trigger. Then, because
this tripping is independent of the sequencer, it is syn-
chronized with the sequencer clock by a flipflop, and
used in combination with signal RESET to assert condi-

tion 4.

e. Overcount Circuitry

The final circuitry Of the sequencer is that which
governs the correction for mechanical hysteresis by count-

ing upward ten extra locations past the proper destination

58

and then back again whenever a positive motion GO command
is executed. This circuitry performs identically under
either manual or computer control, and is never used at
all during RESET commands because Of the inherently nega-
tive motion of a reset Operation.

Figure 13 shows the logic involved. The ten register
is a single decade counter clocked from the same flipflop
that clocks the location register. While the motor is
moving to its desired destination, this register counts
redundantly and without effect. When the destination is
reached however, and the test of condition 5 shows posi-
tive movement, the sequence does not conclude as before
with pulse FT. Instead, pulse F3 is produced, which clears
the ten register to zero and sets the D flipflop. Move-
ment continues, and ten extra locations are counted. As
the tenth extra location is reached, the ten register
rolls over from a nine to a zero, causing the flop to
load the state Of the signal C6,7 EN. This signal, derived
from the direction flipflOp, will inherently be low at
this time because of the positive motion, so the net
effect is the clearing of the flop, and the satisfaction
of condition 6. In response to this, the sequencer pro-
duces PK, which clears the direction flipflop, and causes
the motor movement and location counting to reverse.
Another ten pulses are then counted as the cell returns
to the proper destination, and at the end of the tenth

pulse, the flop again loads C6,7 EN, which will now be

CLK TEN

59

 

 

 

 

 

  

 

 

 

 

 

B, can EN —10 a; c7
TE A9
., ~ 953:8": c t—-— cs
3 a a a ° 5

 

 

 

Figure 13.

Schematic diagram of the overcount circuitry.

60

high. This causes condition 7 to be satisfied, and the
sequence to conclude by assertion of pulse FL.

There is one complication to this action. It was
discovered when the positioner was first tested that the
sudden reversal Of motor movement during this correction
sequence would not work at motor stepping speeds much in
excess Of 100 Hz, due to the inability of the motors to
overcome their mechanical inertia of movement with such
a sudden change. The net result was that the motors
stuttered, and lost steps. However, because a stepping
speed Of 500 Hz was desired, it was decided to correct
the problem by inserting a pause into the sequence before
reversing direction, to allow the motors to truly stop
before reversing. This pause is accomplished with the
circuitry previously postponed from the discussion of
the clock circuit Of Figure 7.

Normally, flipflop A of Figure 7 is clear, which
holds flipflop B and counter C in the clear state as
well. Under these conditions, 500 Hz pulses are passed
without interruption to the two sequencers. If, however,
either sequencer should generate a PR, which indicates that
it is about to reverse direction in an overcount correc-
tion, that FK will set flop A, and gate Off the 500 Hz
pulses to the motors. At the same time, flop B and counter
C, acting together as a divide by 32 counter, will begin
counting pulses of frequency 62.5 Hz. As the thirty

second such pulse is counted, the falling edge at the

61

output of C returns flop A to the clear state, and gates
on the 500 Hz pulses once more. The net effect is that
assertion of a F? pulse from either sequencer suspends
the motor stepping pulses for about one half second;
adequate time for the motor to dissipate its inertia and
reverse without difficulty. Note that with the same
clock line running both motors, it often happens that one
sequencer, in calling for this half-second pause, stOps
the other motor from moving as well for no reason. The
effect is harmless however, as the sequencers have no con-
cern for how long it takes the system to reach a given
destination; they will wait patiently in the proper state

until the destination is indeed finally reached.

f. Motor Driving Circuitry

Figure 14 shows the circuitry used to control the
stepper motor drivers. It is a full-wave, four-phase wave-
form generator Of commercial origin.181

The schematic Of the drivers themselves is illustrat-
ed in Figure 15, the set for each motor being composed Of
four identical units of the type illustrated. If the
input signal to the diode is low, the current passed by
the 1 k0 resistor is sunk to ground through the diode by
the low TTL output. The Specific use of a germanium diode

assures that the base of T1 is at a low enough voltage to

hold Tl fully off. If the input signal is high, the diode

62

.uoumumcmm EhOmm>m3 HouOE Hmmmmum one mo Emummwo owumfimnom

mm am N< —<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.41 messes

coho: XJU

 

 

 

 

 

c—O more:

63

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.mum>wuc Houos ummmoun on» no Enummwo Ofluwaonom .mH mesmeh
H ._. ._. ._.
¢u>1¢o ¢u>1¢o ¢u>1¢o cu>1¢o
fl 1 1 1
31ou 11oo 1.00 11ou
echo: coho: capo: coho:
_ _ T 1
3 cu x ow
=1 1 :1
nu.
u» x.
1»
32:: MN can...
nooazu
a.
a.
tween... . h. .. ..

64

is reverse biased and T1 is switched on. T1 forms the
first stage Of a Darlington pair. Its emitter current
switches on the high power transistor T2, which passes
current from the +25 volt supply through the motor coil
and ballast resistor to ground. The value of the ballast
resistor is that recommended by the manufacturer of the
motors to provide the proper coil current of l A. (The
other components of the circuit are such that T2 is driven
into near saturation, able to conduct as much as 4 A if
needed.) The zener diode on the collector of T2 serves

to clip the peaks of inductive voltage spikes generated

by the motor coils as they turn off. Without such clipping,
these spikes could potentially damage T2 by exceeding its

maximum collector-to-emitter voltage rating.

4. Interface

The computer interface to the positioner need not be
illustrated, as it is a rather simple circuit of the
programmed I/O type very common for DEC 8/e computers

using the KL8A Positive I/O Bus Interface.182

It
consists basically Of the familiar circuitry used to
interpret device codes and IOP pulses, coupled with an
instrument flag which is set by the ANDed combination Of
the READY signals from both sequencers, and cleared by

the start Of a new sequencer cycle or by a separate com-

mand. This flag is usually checked by computer software

65

on a skip test basis, but provision is present for an
interrupt driven facility as well. The complete command
set Of the interface is given in Table 1, along with the
names Of the digital signals they represent. In addition
to these signals, a signal representing device code 34

is also decoded and passed to the sequencers to serve as
signal CDATA, previously mentioned in the discussion of
the sequencer logic as necessary to defeat the synchronous
loading of a 74161 IC. The device codes selected for the
positioner are arbitrary, and were chosen at the time of
construction so as not to interfere with any other known
peripherals. Alteration Of the codes is easily accomplish-
ed if desired by the relocating Of jumper wires on the
interface circuit board, with the restriction that all
three codes must share the same more significant octal
digit.

The interface is also the source Of the signal INIT,
frequently seen in the schematics of the sequencer logic.
This signal is composed of a ORed combination of the com-
puter's INIT line with a local initialize pulse which is
generated by the positioner interface logic at power-up

time.

66

Table 1. Command Set for Positioner Interface

6341 Load AC into destination_£egister of horizontal
dimension sequencer (LD DES)

6342 Load AC into destination_£egister Of vertical
dimension sequencer (LD DES)

6344 Reserved for future use

6351 RESET (ORE-'17)

6352 GO (666)

6354 Skip on Flag

6361 Clear Flag

6362 Enable interrupt facility from positioner

6364 Disable interrupt facility from positioner

67

5. Power Supplies and Physical Structure

The entire instrument is housed within a commercially
produced enclosure. Power for the device is provided by
separate +5 and +25 V supplies. The +5 V regulated sup-
ply used to power the digital logic is Of original, but
highly conventional design, and need not be illustrated
in detail. It is based upon the familiar 723 IC voltage
regulator, employing two parallel, external pass transistors,
and is capable of delivering in excess of 6 A of current.
The +25 V supply, used for the stepper motors, is a very
basic, filtered but unregulated DC supply capable of
delivering 12 A. As presently constructed, the existing
circuitry consumes approximately 67 and 33 percent, respec-
tively of the ratings Of these supplies. Adequate cooling
of such parts as the +5 V supply pass transistors or the
motor ballast resistors is assured by the presence of
a blower fan.

The bulk of the instrumental circuitry is assembled
on six printed circuit boards. Two Of these contain the
digital logic Of the two sequencers. These boards are Of
commercial origin, with provisions for thirty six chips,
and accompanying power and ground buses. The chips are
socket mounted and interconnected with wire wrap tech-
nology. Two additional boards Of original design and
layout contain the components of the associated stepper

motor drivers. The final two boards contain the

68

circuitry of the clock sources and the computer interface.
They are also of original layout and design and employ
wire wrapping for signal connections. These six boards
stand vertically in a horizontal row within the instrument
and are inserted into standard edge connectors which are
mounted in a printed circuit backplane board which lies
along the bottom of the chassis and serves as a signal
bus for all necessary interconnections among the six
boards. A final printed circuit board is located at the
front panel, and contains the inverters associated with
the thumbWheel switches as well as TTL drivers for the
front panel LED's. Connections Of this board with the
bus board are made with simple cable bundles, with the
exception Of the data from the thumbwheel switches, which
is passed to the appropriate sequencer logic board directly
through use of bundled wires terminated by an IC header
cap which inserts into one Of the IC sockets on the logic
board. Connection Of the computer interface terminal to
the positioner is made via a ribbon cable patch which
inserts into an edge connector in the rear Of the chassis.
Similarly, three multiple conductor cables leaving the
rear Of the instrument serve to make electrical connection
to the horizontal motor, vertical motor, and optointer-
rupters, respectively.

There is sufficient physical room on the front panel
and signal bus for the addition of a complete, third se-

quencer and driver should future applications Of the

69

instrument require a third motor. Similarly, the computer
interface is already configured to recognize a third motor,
and the power supplies have ample reserve capacity to

handle a third load.

6. Testing and Performance

Most of the initial work involved in testing the
positioner consisted of the inevitable task of debugging
the digital logic. (Indeed, the schematic diagrams of
the sequencer logic given in the figures are, in some
cases quite different from their original design concepts.)
This work was done with oscillosc0pes and test probes and
with one very advantageous technique involving the sequen-
cer clocks. Many individual problems in the logic, par-
ticularly the verification of correct action of the gate
and overcount circuitry, were solved by using an external
clock source running at only a few Hz. This slow source
enabled the individual states of the sequencer to be
followed with test probes, as well as permitting a direct
count of pulses received by the stepper motors at critical
stages of a sequence by simply holding the motor shaft
and feeling the pulses one by one. SO useful was this
external clock in proving out the instrument that pro-
visions for using it have been kept in the final device:
the output of the 555 timer that provides the basic clock
frequency is passed to the remaining circuitry via a

long, loOping jumper wire which stands Off the clock

7O

circuit board. By removing the timer chip and connect-
ing an external TTL source to the loop, the sequencer and
motors can be run at one half and one fortieth, respec-
tively, of most any reasonable frequency desired.

Of more direct concern to the intended use of the
positioner was an examination of the optointerrupter reset.
If the optointerrupter is to serve as the absolute cali-
brator of location zero, then it must act in a highly
reproducible manner. That is, the exact instant at which
the interrupt occurs must always happen with interrupter
and metal occluder at precisely the same consistent
orientation with respect to each other.

To test the quality of this action, the following
experiment was conducted. An events counter was tempor-
arily connected to the sequencer signal CLK MOTOR, so that
the counter could register the total number of pulses
sent to the motor during a given instrument cycle. The
cell was then displaced to a known location. Finally,

a reset was commanded, and during the reset Operation,
the number of pulses sent to the motor was counted. If
the optointerrupter defines location zero consistently,
then the number of pulses required to reach the reset
point should always be simply twice the value of the
location from which the reset was performed. Repetitive
counts were taken for both motors from locations 1000,
2000, 3000, and 4000, thus covering the total range of

displacement. The results are shown in Tables 2 and 3.

71

Table 2. Precision of Location of the Reset Point -

Vertical Number of Counts Accumulated

 

 

From Location 1000 From Location 2000

 

2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001

 

From Location 3000

From Location 4000

 

6001 8001
6001 8001
6001 8001
6001 8001
6001 8001
6001 8001
6001 8001
6001 8001
6001 8001
6001 8001

 

 

72

Table 3. Precision of Location of the Reset Point -

Horizontal Number of Counts Accumulated

 

 

From Location 1000 From Location 2000

 

2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2001 4001
2002 4001
2001 4001

 

From Location 3000

From Location 4000

 

6001 8001
6001 8001
6001 8001
6001 8001
5999 8001
6001 7999
6001 8001
6001 8001
6001 8001
6001 8001

 

 

73

The results are amazingly precise. No deviation
is seen at all for the vertical motor at any location,
any slight errors being rounded off by the inherently
quantized motion of the cell. (The consistent extra count
arises from an inherent gating artifact in the events
counter.) Only a few scattered, minor deviations are
seen for the horizontal motor, (due perhaps to the fact
that the horizontal motor must move not only the cell but
the vertical motor and its stage and support as well)
but even here the deviations are not significant with
respect to this work. Eight motor steps are required to
displace the cell one milliinch. Conservatively rated,
the location of the reset points is equally as good. But
one milliinch is far more precise than the atomic absorp-
tion work performed here. The resolution of the optics
alone is at least an order of magnitude larger, and even
if the resolution of the optics matched that of the cell
location, it is doubtful whether transients of atomic vapor
violently ejected from a hot atomizer into a gently flow-
ing stream of sheath gas would be equally as precise.
It may be concluded therefore, that the calibration of
the cell by the reset operation is as good as the mechani-
cal rigidity with which the translation stages and Opto-

interrupters are assembled.

74

7. Operating Software

The basic package of three assembly language sub-
routines are employed to operate the positioner. These
routines were written and are maintained by Mr. Eugene
Pals, to whose thesis the reader is directed for details
of their Operation. A short commentary on the use of these
routines will be given here for completeness of the instru-
mental discussion.

The first routine receives two arguments from the
calling routine which represent the desired horizontal
and vertical destinations of the cell. These destinations
are entered in units of millimeters to the calling routine
by the user, with 0.0 mm vertical being at grazing inci-
dence, and 0.0 mm horizontal being centered. In these
units the vertical locations may range between 0.0 and
26.0 mm above the braid, and the horizontal locations
may vary out to :12.7 mm on either side of the braid.

When passed to the assembler subroutines, the necessary
conversion from millimeters to twelve bit format is per-
formed. The positioner flag is cleared, after which the
arguments are converted into machine code and sent out
successively to the horizontal and vertical destination
registers. In conclusion, the cycle command G0 is issued.
The second routine clears the flag and issues the cycle
command RESET. The final routine performs a skip test

of the positioner flag. With these three routines the

75

full sc0pe of positioner operations can be accessed, with
the exception of the interrupt facility, which is not
presently in use. The divorcing of the skip test from
the cycle commands is particularly convenient in that it
permits the computer to start the positioner on a task
and then direct its attention elsewhere, only returning

to check the positioner flag when truly necessary.

C. THE AUTOSAMPLER

1. History and Introduction

One of the principal reasons for constructing the
positioner was to provide a sophisticated instrument that
would relieve the user of all responsibility for correctly
placing and translating the atomization cell. Indeed, the
general goal was to achieve this degree of independence
in the entire AA system. Ideally, once the user had
specified to the computer the tasks to be performed, the
instrumentation would automatically take care of executing
them without the need for further user intervention; the
system would be able to run itself for hours on end. In
order to truly reach this goal, it was necessary to have
along with the positioner a companion instrument to per-
form a mechanical task even more mundane and common than
locating the atomization cell: namely, the deposition

(of samples. There is considerable precedent in these

76

laboratories for the construction and use of automatic
samplers. Two such devices have in fact been employed
with the graphite braid.183'184
The first autosampler was a simple device designed
to deliver repetitively a fixed amount of a single sample.
This was accomplished by forcing the sample in question
out of a delivery tip manufactured from a metallic syringe
needle. This needle was mounted in an aluminum arm, which
in turn was mounted in a pivot on a main chassis. At the
time of sample deposition, the arm and needle were pivoted
forward through an angle of 90 degrees by a small pneu-
matic cylinder, thereby bringing the tip of the needle
to the top edge of the braid. As the tip reached the braid,
a microswitch would be tripped by the moving arm that would
engage a motor and cams combination, which would then dis-
pense the sample by momentarily compressing a short length
of the tygon tubing through which sample was drawn from
the source cup. The cams were used not only to compress
the tubing, but also to block reverse flow, so that the
sample would be forced out Of the needle onto the braid.
Upon concluding this action, the pneumatic cylinder would
release, and the arm would return to its rest position.
The precision of this sampler was quite acceptable,
about 2-3%. It had the advantages of a reasonably small
physical size and rapid sampling rate; the entire sampling

Operation took about five seconds. For repetitive sampling,

77

it was in fact fairly servicable. However, it lacked
many features of a truly versatile automatic sampler.
The sample size was variable from about one-half to five
microliters (by controlling how much tubing was compressed
by the cam), but it was not computer controllable, being
governed instead by a manual screw assembly. The source
of the sample was fixed by what the network of tubing
happened to contain at the moment, and changing the sample
was very tedious, as it involved continuously pumping Old
sample through the tip five microliters at a time until
the system was drained, and then following this up with
more pumping of first rinse water and then new sample
solution. Because of the fairly large dead volume of
tubing involved, this process could take as much as a
half hour. The need for both electrical and pneumatic
power was inconvenient, and the metal delivery tip was
prone to often severe scavenging of solutions through elec-
trochemical deposition of the trace metals present.

Many of the limitations of the first autosampler
were recognized soon after its completion and a second
device was constructed to attempt to overcome them. The
new autosampler retained the concept of pivoting an arm
and delivery system through 90 degrees with a pneumatic
cylinder at the time of sample deposition, but the sampling
device itself was changed to a stepper motor driven
micrometer syringe mounted in a housing which could be

finely translated and rotated with screw adjustments to

78

allow precise positioning of the delivery tip over the
graphite braid. A rotating sample turret driven by a DC
motor was provided which contained room for as many as
six sample cups. The pneumatic cylinder which pivoted

the arm and syringe forward was accompanied by an addi-
tional cylinder which could rotate the arm about 45 de-
grees horizontally so that the syringe, when thus rotated,
could address either the atomizer or the sample turret.
Operation of the device was controlled by an ingeniously
constructed series of cams turned by slow speed appliance
motors. Upon receiving the command to deposit a sample,
one such series of cams would be turned through a com-
plete revolution, bearing upon a series of microswitches
that would automatically sequence the pneumatic cylinders
properly to pivot the syringe to the braid, wait for sample
deposition, and then return. Similar cam systems could,
upon receipt of a single command, move the syringe over
to the sample turret and refill it, or even perform a
discarding of old sample and double rinsing of the syringe
assuming that one sample cup would be reserved for waste
and another for deionized water.

This sampler was a great improvement over the pre-
vious one in that it provided computer-selectable sample
size and sample source, and employed a delivery system
of greater potential resolution which could be cleaned
or changed more readily and had far less dead volume.

The precision of delivery was again on the order of a

79

few percent. Unfortunately, these advantages were gained
at some cost. The size of the sampler was tremendously
increased. It literally enveloped the old atomization

cell used with it, and was inherently physically incom-
patible with the new positionable cell. The requirement

of pneumatic power was still present, as was the use of a
metal needle on the syringe. The new device was ponder-
ously slow. The intricate cams which could automatically
execute the complicated mechanical sequences had no ex-
tensive means of detecting the completion of various stages
of the cycles, and thus had to operate on a basis of allow-
ing generous times for the completion of individual steps.
The time for sample deposition was on the order of thirty
seconds, and the more complicated refilling or purging
operations required times on the order of minutes. Finally,
certain desirable automatic features were missing; for
example, the computer had to provide the pulse train of
proper length and frequency to step the syringe motor
instead of having this task done externally.

Thus, neither sampler was truly acceptable as a
general purpose device. It was therefore decided to con-
struct a third system which would combine as best as pos-
sible the small size and rapidity of the first sampler
with the versatility of the second, and place it all under

the control of a sophisticated external controller.

80

2. Device Description

The basic principles of the new sampler's construc-
tion may be most easily conveyed with the assistance of
the photographs on the following pages. The main frame-
work of the device consists of a small baseplate bolted
to the spectrometer optical rail behind the atomization
cell which, with the aid of triangular braces, supports
two bar shaped vertical members. These members resemble
a block letter U in cross section, and are oriented so
that the "U's" face each other, thereby forming a vertical
channel. Located in this channel and able to move up
and down in it is a metal plate, to the rear of which is
bolted a large stepper motor. The shaft of this motor
protrudes through the plate and is fitted with a 8" gear.
This gear bears against the larger member of a 2%"-8"
combination gear, the smaller member of which in turn
bears against a length of rack bolted to the right hand
vertical bar. Thus, by rotating the stepper motor, the
motor, plate, and anything secured to the plate will ride
up and down in the vertical track. The gears provide a
fivefold mechanical torque advantage for the lifting of
considerable weight. The plate is lined with strips Of
teflon along those edges which contact the vertical bars
to minimize friction when moving.

On tOp of the vertically moving plate is secured

another, horizontally oriented plate. Under the rear

81

half of this plate is mounted a second, smaller stepper
motor, having a protruding shaft and %" gear similar to
the first. This gear bears against a single 2%" gear which
is mounted on a teflon disc centered directly over the
vertical plate, and fitted with a shaft and bearing so that
rotation of the small stepper motor will cause the large
gear and anything secured to it to also rotate in the
horizontal plane. To this gear is affixed an open, rec-
tangular, box-shaped structure, upon one end of which is
bolted the stepper motor driven syringe assembly. This
syringe was salvaged virtually intact from the Older
autosampler previously described. It consists of a 200
ul micrometer syringe driven by a 200 step per revolution
motor. One revolution of the micrometer will dispense
or draw 10.0 ul of solution, thus giving a theoretical
resolution of 0.05 ul per step. The structures in which
this syringe is mounted include screw driven fine position-
ing adjustments which allow the precise locating of the
delivery tip in both the X and Y dimensions, defining the
Z dimension as vertical.

Associated with the main unit is a second, physically
separate piece which serves as the source Of samples.
It consists basically of a flat metal disc with sixteen
equally spaced holes around its outer perimeter. These
holes are of proper size to accommodate 2 ml disposable

Autoanalyzer sample cups. This disc is mounted with a

82

shaft and bearings to a simple framework chassis. Also
mounted on the shaft is a third 2%" gear, which again is
turned by a stepper motor and 5" gear combination. The
motor used here is small, owing to the minimal torque
requirements.

The four photographs illustrate the basic movements
performed by the sampler when in operation. The usual
rest position of the device is shown in Figure 16, with
the moving assemblies located at the top of the vertical
track. Figure 17 illustrates the position taken during
sampling. When the commands to deposit a sample on the
braid are received, the large motor (vertical motor) lowers
the entire assembly to the braid. Upon arriving, the syr-
inge dispenses the sample, after which the vertical motor
returns the apparatus to the top of the track, out of the
way of the atomization cell and optics.

If replenishing or changing the contents of the syringe
is necessary, the syringe must be moved over to address
the sample turret. The first step in this process is
shown in Figure 18, in which the small motor (rotation
motor) has rotated the box structure and syringe assembly
90 degrees counter-clockwise so that the delivery tip
is now over the sample turret instead of the braid.
Finally, Figure 19 shows the assembly lowered by the
vertical motor to the turret. Notice how the open struc-
ture of the rectangular box permits the principal vertical

members to pass actually through the box as the assembly

83

.umamemmousm
on» mo noduemoo
mcfluflmoooo wamfiom .FH musmfim

.Hoamfimmousm
on“ no sewuflmom umwu Hmeuoz

 

.11 wnsa1a

 

84

.noaoEmmouow mop mo seeuemOd
mnemuQEw no mcflaaflw mHmEom

.mH wusmflm

.uwamsmmousn
on» no cofluwmoa
mcemmmupom uwuusa

.m1 mnsm1a

 

85

is lowered.

The proper places at which each motor is to stop
moving are indicated by optointerrupters identical to
those used in constructing the translatable atomization
cell. Three such devices define the locations of "top",
"braid", and "turret" to the vertical motor; two of them
are used to indicate "left" (over the braid) and "right"
(over the turret) to the rotation motor; and a single
device is used to indicate sample cup "zero" to the sample
turret. These interrupters are tripped by apprOpriately
mounted small pieces of metal. Similarly, two micro-
switches activated by a structure on the syringe plunger
indicate whether the syringe is full or empty.

Unlike the positioner, the sampler is designed to
operate only under computer control; manual sampling is
inherently more efficient using a simple hand-held syringe.
Though operable by computer only, the sampler control
logic still retains the philosophy employed with the posi-
tioner of requiring only minimum attention from the com-
puter. However, considered as computer-controlled devices,
there are some differences between the two instruments.
Each Of the nine separate requests responded to by the
positioner was given its own computer command. It was
thought, however, that this was a rather wasteful procedure,
as only two (potentially three) of the commands have data
transfer associated with them. The other six are bare

instructions. A similar situation exists for the sampler.

86

Only movements of the sample turret and syringe require
data transfer of any appreciable sort. The vertical

and rotation movements are essentially data free, as are
again such commands as flag checks and interrupt statuses.
It was thus concluded that two commands could serve all
possible mechanical movements of the sampler. One of
these would direct the actions of the syringe, using eleven
bits of the accumulator for data, and the remaining bit
as a sign bit to signal filling or emptying. The other
command would control all motions of the other three
motors, using four bits for data to the sample turret,

and leaving the remaining eight free for assignment as

a sort of enable register whose exact contents would
indicate a particular set Of motions.

Another basic difference between the two instruments
is also well served by this combination-of-commands con-
cept. Unlike the positioner, whose two motors and stages
are completely independent physically and able to move
back and forth at will oblivious to each other, there are
several constraints imposed in the sampler with respect
to simultaneous movement by more than one motor. For ex-
ample, if the sampler is in the positions of Figures 17
or 19, simultaneous movement by both the vertical and
rotation motors will cause the rotating assembly to strike
and labor against the vertical members long before they
are cleared in ascending. Thus, precautions against this

and other forbidden movement combinations must be taken

87

to avoid damage to the unit. This too is facilitated by
use of a single command and enable register concept, in
that illegal requests can be detected as forbidden combina-
tions of register bits by the control logic, which can

then act to block the disallowed command from actually
starting the motors moving.

A final difference between the two is that in the
sampler, there is, along with the enable register, a
status register which can be used to input information
to the computer regarding the sampler's physical location.
By reading this register, the computer can determine,
for example, that the sampler might be vertically at the
top and rotated over turret sample cup zero with the
syringe empty. Such ability to check orientations faci-

litates initializing the device upOn power up.

3. Detailed Logic
a. Registers

None of the four motors of the sampler executes a
function which is excessively complicated. Thus, the
logic of the device consists of four independent groups
custom designed to the needs of each motor, all serviced
by a common interface and signalling a combined flag.

A sequencer such as the Richards unit of the positioner
is not needed either, as the logic sequencing is direct

enough to be Of the self guiding, "fall through" type.

88

The logic of the four motors will be considered one by
one, after which the flag and interface will be treated.
First, it will be helpful to list the bit assign-
ments of the enable and status registers to make the
logic easier to coordinate. These assignments are given
in Table 4. The enable register, as previously mentioned,
is an output register containing four bits of data for the
turret and potentially eight bits to determine which
motors are to move and to what destinations. The status
register is an input register which uses four bits to
signal attempts at illegal commands for each motor, and
eight bits for informing the computer of the physical
orientation of the instrument. With these assignments
available for reference, an examination of the four logic

networks of the motors can begin.

b. Vertical Motor

The logic of the vertical motor is easily the most
complicated of the four, due to the existence of some
specialized manual controls associated with the precise
positioning of the syringe tip over the graphite braid.
There are three basic positions of the vertical motor.
The first is at the top of the track, which is the normal
rest position andtfluaposition at which rotation takes
place. The others are the braid and turret positions,

being respectively the proper heights at which to address

89

Table 4

Autosampler Registers

 

 

 

Enable Register Status Register
EN 0 : Unassigned ST 0 : Vertical Error
EN 1 : Unassigned ST 1 : Rotation Error
EN 2 : Vertical Up/Down ST 2 : Syringe Error
EN 3 : Vertical Enable ST 3 : Turret Error
EN 4 : Rotation Left/Right ST 4 : Syringe Full
EN 5 : Rotation Enable ST 5 : Syringe Empty
EN 6 : Turret MSB ST 6 : Rotation at Left
EN 7 : Turret Bit ST 7 : Rotation at Right
EN 8 : Turret Bit ST 8 : Vertical at Top
EN 9 : Turret LSB ST 9 : Vertical in Turret
EN 10 : Turret Enable ST 10 : Turret at Zero
EN 11 : Unassigned ST 11 : All Motors Idle

 

 

90

either the sample turret cups or the braid itself. These
latter two positions are easily set up as necessary by
simply moving their optical interrupters up or down on

the mounting bracket. The logic also contains the ability
to recognize yet a fourth position, which is not currently
in use, but which will be explained anyway.

Three basic sections Of the vertical logic are shown
in Figure 20. Certain aspects of the first two of these
are used repeatedly with the other three motors. The
first section is the gate used to start or stop motor
movement. It is closed by setting flipflop A. This takes
place when the computer gives the combined motor command
(CMD 1), provided that movement of this motor is requested
by bit 3 of the enable register (EN 3), and that the com-
mand is not blocked by an error condition preventing it
(VERR is not low). The following AND gate, inverter, and
flipflOp B serve to synchronize the computer request with
the 600 Hz clock which actually steps the motor. With
flipflop A set, the next falling clock edge sets flipflop
B, which in turn gates the clock to the motor via a final
OR gate whose use will be explained later. In essence,
this circuitry is thus a synchronous gate which provides
a "clean" waveform to the vertical stepper motor driver
whenever a legal movement of this motor is requested by
the computer.

The second logic section, shown at the lower right,

is used to detect illegal movements of the vertical motor.

91

uouum pow .Houucoo aceuoouwo .mumm on» no Bouquet owumEmsom

w3h<hm

E

 

> KEN

 

 

 

> u>~¢o core!

.HouoE Hmoeuum> may now mavenouwo uomuop

N: 000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

> wzoo

 

 

 

 

.om messes
zxooxa:
> :3 .33.. ELI
b c
:33 u. a]... an
o
.51.. or... 28
w
Eh.
% c1a uzap
c
b olnzu
< E
o ulIAAHHHH1 axe

92

An illegal motion is requested, and must be prevented,
through any mechanism which asserts VERR. From the gates
preceeding the origin of this signal, it can be seen that
illegal vertical motion is requested if simultaneous
rotation is also requested (EN 5), if rotation from a
previous legal command is still in progress (DONE—R),

or if the syringe is over the turret and turret motion is
requested (EN 10 ' LEFT), or in progress (DONE—T - LEFT).
Any one or more of these conditions will assert VERR

and block the setting Of flipflop A by CMD 1. But, in
addition, flipflop C will be set by the same CMD 1, pro-
ducing signal ERR V, which is passed forward to the sampler
flag to signal the computer that the requested action

was aborted. The computer can then respond by detecting
ERR V as bit 0 of the status register with the command
STATUS. Similar error detect networks will be seen later
in the other motors.

The final section Of logic merely determines the direc-
tion of motion of the motor by preparing flipflop D in
light of the state Of EN 2 at CMD 1 time. The outputs
of this flop control not only the direction of the vertical
motor waveform generator, but also provide the signals UP
and DOWN, used in a future section of logic to detect when
motor movement is to cease through generation of the signal
§T_R. The gates associated with the R and S inputs of
this flop are concerned with the previously mentioned

manual controls, which will now be explained.

93

Although the framework upon which the syringe is
mounted contains provisions for fine screw adjustments
of the exact positioning of the delivery tip, one aspect
of this positioning available to the control logic is
the precise point at which the vertical motor stops in
descending to the braid. The optical interrupter nominally
known as the Braid 01 is in fact actually located a quarter
inch or so above the level of the braid, with the remain-
ing distance covered through guidance of the circuitry
shown in Figure 21. The circuit is centered around two
eight bit registers, known as the reference and working
registers. Clock pulses of one tenth the normal stepping
frequency can be gated into the reference register from
the debounced front panel switch "TUNE". These pulses
cause this register to either increment or decrement ac-
cording to the setting of the panel switch "UP-DOWN".
These same pulses also cause the vertical motor to move
synchronously up or down through use of the signals
TUNE, which is passed to the motor through the OR gate
postponed from the discussion of the gate circuit, and
UP/DOWN, which forces the state of flipflop D through the
circuitry on its R and 8 inputs. Gate A detects overflow
or underflow of the reference register and cuts Off the
60 Hz pulses at a count of 255 or 0, respectively, should
either event threaten to occur. Gate A also stops the
motor motion. The contents of the reference register may

be cleared by either the signal INIT, or front panel

94

on» How havesoufio

.HouoE HMOfluuu>
oceoofluwmom mafia on» we Emummflw oeumfionom

 

 

 

 

 

 

 

 

.HN musm1n
Nz one u h
z 1 1 E11
n¢474>4> I
(a _ _
ox: x16 ox: x19 ~<1<
cum“ 1
cure—mu: on uz1xmox 2. on
~0—rh ‘—
= u

 

 

 

mzmxmoz zu
_o_cs 1
m < _ 1. u m

A

 

 

 

uzzoobuau I

< gl .1. m o
zxonxaal
thh Mg. *
no

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m: 1
~z coo
_ < x.
h oxanum<xx oxaoum<xxvluw
H. or muzmxuuum zmo Qua mozmzmuuc 2m AWN
.¢<uao- 1 1m1cs you IA.1 1n1ck x1u
n u m < o o m <
ms. b
no

 

 

 

 

 

x1
:19 ”2:1
“kph

09

95

switch "CLEAR".

When the vertical motor is sent down to the braid,
it continues down until the Braid OI is reached. When
this happens, the event is latched and monostable A is
triggered which loads the contents of the reference regis-
ter into the working register. Motor movement then con-
tinues synchronously with decrementation of the working
register until this register reaches zero, at which point
the signal REGISTER ZERO stops the motor, as will be seen
in the next figure.

Typically then, the user, upon turning the sampler
on or pushing the "CLEAR" switch, executes a computer
routine which causes the syringe to descend to the braid.
The reference register will be initially clear at this
point, so the motor stops at the Braid OI, since the work-
ing register will load a zero and assert REGISTER ZERO
immediately when the OI is struck. Using the "TUNE"
and "UP-DOWN" switches, the user then positions the tip
of the syringe precisely to his satisfaction, and signals
the computer to return the syringe to the top rest posi-
tion. Whenever told to descend after that, the net number
of pulses entered into the reference register are used to
bring the syringe past the Braid OI to exactly the tuned
position in the manner just described. In this way, slight
differences in positioning of successive braids can be
accommodated without having to move interrupters or adjust

syringe mounts.

96

When the gate circuit was discussed, only the means
for closing the gate were outlined. Figure 22 shows the
circuitry which detects, for each destination of the
vertical motor, when to open the gate again to stop mo-
tion. All of this circuitry is dedicated to generating,
by one means or another, the signal STOE, which causes the
clearing of both flipflops A and B of the gate circuit (A
directly, and B as a result of A). The A01 gate A divides
this circuitry into two basic branches, namely that which
controls stopping motion for descending movements and that
which controls it for ascending movements. The signals
UP and DOWN at two inputs Of A activate one branch or the
other as necessary.

Let us consider first the branch which stops ascend-
ing motions. As of this writing, the only possible des-
tination when ascending is the top. Thus, when the TOp
OI is reached, its signal in combination with UP asserts
STOR through all of the gating. (Note that to conserve
space in the diagrams, the interrupter producing signal
TOE—OI is not shown, nor will be those of any further
interrupter signals. They are all identical to the one
illustrated in full for the Braid OI in the previous
figure.) FlipflOp E and gate B are presently not used,
and have to do with the possibility of a fourth vertical
destination previously mentioned. When the sampler was
designed, it was foreseen that when rinsing between sample

types, the syringe would be quite busy at the turret

97

.uouos
Hmoeuum> may now muuflsouwo moan on» mo Bouquet Oflumemsom .NN muzmem

 

 

 

 

 

 

 

 

E

 

 

 

OCUN
Cuhwnomc

 

 

 

 

 

 

 

 

 

 

 

98

drawing and dispensing rinse water and solutions. It was
thought that in such cases, forcing the syringe to return
all the way to the top while rotating the turret would

be wasteful of time. It would be better to have an alter-
nate ascending destination which would allow the syringe
tip to just clear the turret, and only go all the way up
to the tOp when ready to rotate over to the braid. It
turns out however that the best place to put the sample
turret at present is on top of the monochromator, which
already puts it very close to the top anyway. Thus, the
extra destination was left out as being only marginally
useful now. Should future applications require, however,
it can be installed, and its OI connected to the avail-
able Schmitt trigger shown. The D input of flipflop E
could then be connected to one of the remaining available
enable register bits instead of to +5 volts, and then each
occurrance of CMD l for ascending motions would prepare
the flop to recognize either the Top or the new OI as
necessary for generating STOP through AOI gate B. Note
that the vertical position "top" is available to the
status register as bit 8.

The other branch of gate A controls stopping in the
downward direction. Such stopping is to take place at
either the Turret or vicinity of the Braid OI's. The stop-
ing point depends on whether the syringe is rotated right

or left. Flipflop F keeps track of the current position

99

of rotation. If at the left, F is clear, and §Tr'6p‘ will

be generated during descending movements by the signal

from the Turret OI in combination with DOWN. If at the
right, F is set, and ETOR is derived from the network
around flipflOp G, which prOperly responds to the manual
tuning circuitry previously described. When descending
over the braid, as the Braid OI is passed, the same
monostable which loads the working register sets flip-

flOp F with the signal BRAID‘NEAR. When, in addition,

the signal REGISTER ZERO is asserted (either immediately

or after decrementing a nonzero working register), these
two combine with DOWN to assert STOR. Flipflop G is cleared
for the next such occurrence by the absence of signal

DOWN when the inevitable ascention after sampling takes
place. Note that the vertical position "turret" is present
in the status register as bit 9, and that the signal TNTT
assures that the vertical motor is halted when the sampler

is first powered up.

c. Turret Motor

The circuitry of the turret motor logic is shown
in Figure 23. The lower section is the illegal command
detect section, which is structured in a manner almost
identical to that of the vertical motor. The error condi-
tions possible are either turret movement attempted simul-

taneous with or during movement of the vertical motor,

100

.nouos umuusu may you huueauuwo map mo Emumneo oeumsmnom

w3r<pw

E

r 81w

 

 

 

 

 

z”

 

Oun<

Ox: 8.3

a
.O—rh

 

 

 

oun<

 

,2;

 

 

Man.

J node «a
00:.
r< 0< N< ~<

 

 

 

 

 

 

 

w3h<ru

 

 

 

 

 

 

 

.mm manage

 

101

should the syringe be located at the left over the turret.
If an error occurs, it is detectable as status register
bit 3.

The upper section is the gate for motor movement and
the counting circuit for seeking the proper sample cup.
The method used is suggested by the techniques employed
in the positioner. If CMD l is given along with EN 10,
and the signal is not blocked by an error, the identity
of the sample cup to be rotated into position under the
syringe is loaded into a four bit latch from enable
register bits 6 through 9. The contents of this latch
is compared by a four bit comparator with a counter
which contains the actual identity of the current sample
cup. Removing the condition of equality between the latch
and counter by specifying a different cup asserts DUNE‘T,
which allows clock pulses of 10 Hz to pass to the turret
motor driver. For every twenty five pulses passed to the
driver (as counted by two successive divide by five sec-
tions of decade counters), the cup counter is increased
in value by one unit, twenty five being presently the
proper number of pulses at the motor shaft between adjacent
sample cups. When the condition of equality is reestab-
lished between latch and counter, the assertion of DONE‘T
is lost, and pulses are cut off. (Note that the turret
motor moves in one direction only.) At present, the
stepper motor used with the turret is a rather large

angle device (9° per step) that was conveniently available.

102

Because of the large step size, only eight of the poten-
tial sixteen cup positions are currently in use, spaced
two positions apart, and numbered 0 through 7. It is for
this reason that EN 6 is not actually connected where

it belongs, which causes the latch and counter to func-
tion as three-bit devices.

The final few gates assure the proper initialization
of the turret. An optical interrupter and piece of trip
metal are situated so as to define sample cup zero. Every
time the OI is struck by the revolving metal piece, the
monostable is triggered, which resets both the cup counter
and divide by twenty five network to zero. This reset is
also accomplished by the TNTT pulse, so that when first
powered up, the turret will think it is at position zero
no matter where it might really be. Two successive com-
mands to send it first to some nonzero, and then to the
zero location will assure that the turret makes a complete
revolution. Sooner or later during that process, the
interrupter will be struck, and the true position Of zero
will thus be established. Predictably, this important
turret interrupter is available to the computer as a

status register bit, bit 10.

d. Rotation Motor

The circuitry associated with the rotation motor is
even simpler, as this motor has only the humble task of

moving back and forth through (at present) 90 degrees

103

between the interrupter defined limits of left or right.
Figure 24 illustrates the logic. The gate for starting
and stopping the motor is all but a direct OOpy of that
used with the vertical motor, and the illegal movement
detect circuit is once more very similar. Here the
illegal movements consist of rotation attempted simul-
taneous with or during vertical movement, or rotation
under any circumstances whatsoever if the syringe is not
vertically at the top. The Opening of the motor gate is
also equally direct. Simply put, the motor is stopped
when the Left OI is struck if moving left, or when the
Right OI is found when moving right. The assertion of
the proper signal LEFT or RIGHT is governed directly by
a flipflop prepared according to the state of EN 4 at CMD
1 time. Again, the TNTT signal assures that motion is
halted when power is first applied. Both the left and
right interrupter signals are Open to computer query as

status register bits 6 and 7 respectively.

e. Syringe Motor

The three motors discussed thus far are all serviced
by the same command to begin movement, to sort out the
proper duties through use of the enable register, and to
respond to the computer with the status register. The
final motor, namely that Of the syringe itself, has its

own separate command since it needs the full twelve

104

.uouoE
coeumuou may now mhuwsoueo may no Eoummflo oeumEmnom .vm whomem

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a can
war<pw
E
E
c E—O Echo:
..I o: n _m :82.
3:3. E
him—c .0. cl: 2m
Ohr E hum.—
E 53.10 01.1 98
my)
a c
a n. b n. m an
o O o u E
a nice coroxlAl—Ifi) 1 98
N: 00—

 

105

accumulator bits for data. It does, however, still use
the same status register and device flag.

Figure 25 shows the gate, error detect, and stop

circuitry for the syringe. Once more, the gate is very
similar to those of the vertical and rotation motors,
except for lack of an enable bit at the D input Of the
first flipflop. The syringe motor is halted as usual
by TNTT, by two mechanical limit microswitches similar
in action to the two interrupters of the rotation motor
should the syringe reach the end of its micrometer travel
when filling (UP) or emptying (DOWN), or by the signal
ZERO which will be explained with the sample quantity
computing circuitry of the next figure. The full and
empty limit conditions are available respectively as bits
4 and 5 to the status register. The error conditions are
attempted movement of the syringe if the vertical motor
is moving or if the vertical motor is at the top. Techni-
cally, these errors are mechanically harmless, as the
syringe can theoretically be moved whenever desired.
The restrictions were incorporated though as the syringe
has no business being used at any position other than at
the braid or in a sample cup. A syringe error is detect-
able as status register bit 2.

The logic drawn in Figure 26 controls the all important
functions of filling or of emptying, and Of determining
how much in either case. The twelve bits of the accumu-

lator are structured as a twelve bit two's complement

106

.uOuOE omceuhm on» you muueaouwo
moon can .uoouwo nouum .oumm on» «o Emummwo owumEOnOm .mm musmem

0+ nu. BK» Run
325 P

 

 

OTI§(I)n+
an

 

 

 

 

 

 

 

> uzoo

 

O
0

 

 

 

 

 

 

 

o u
m nice 332$ u 28
N... com Db.

107

.Houoe wmcwuhm
may you wuuwsuuwo Houucoo muflm meEmm map No EMHmeU oaumfimnom

        

N1 000

 

2300 b. o

HDQHDO mOh<JDZDQQ<

.mm ousmflm

u<

108

number with bit 0 being the sign bit used to signify fill
or empty. The remaining eleven bits are used for data to
a synchronous, down counting register. Three NOR gates
and a NAND gate provide a zero detect for this register,
the resulting signal being 5555, seen previously as one
of the motor gate opening signals of Figure 25.

Both the sign and data bits are loaded into a flip-
flop and the register, respectively, by the signal 55,
which is merely the combination of CMD 2 and the consent
of the error detect logic. Upon loading, the motor direc—
tion is set, and movement begins, with the data register
down counting synchronously with motor movement. When
the register reaches zero, the requested amount of solu-
tion has been either drawn up or dispensed, and the result-

ing signal ZERO stops the motor.

f. Motor Drivers, and Clock Sources

The pulse trains directed from the gates to the motors,
as well as the motor direction control lines, are used
to run four full-wave stepper motor sequencers of the
exact same construction as that presented with the posi-
tioner as Figure 14. Similarly, the outputs of these
sequencers in turn are used to control four sets (sixteen
total units) of motor coil drivers identical to those
of positioner Figure 15, except that the ballast resistors

are different for each motor as listed in Table 5. (The

109

Table 5

Ballast Resistor Values for Sampler Motors

 

 

Vertical Motor 79
Syringe Motor 160
Rotation Motor 200

Turret Motor 420

 

 

110

use of identical drivers for each motor simplified con-
struction of the electronics of the sampler, even though
the large vertical motor is the only motor truly requiring

the full current switching capability they possess.)

4. Interface

The flag circuit used for the entire sampler is il-
lustrated in Figure 27. As can be seen, the flag may
signal the computer through either the skip test or pro-
gram interrupt modes, as was the case with the positioner.
The interrupt facility is enabled or disabled with the
command SAM INT and associated accumulator output bit 0.
Operation of the flag itself is centered around flipflop
A and gate B. When first powered up, or after the com-
puter responds to a previous raising of the flag with the
CER'SAE'FEE'command, flipflop A is clear, which keeps
gate B low and the SE? and FT lines inactive. If either
CMD l or CMD 2 is given to start an operation, flipflop
A will be set. The action of gate B then depends on
what happened when the CMD instruction was given. If
the attempted command was legal, all of the ERR i sig-
nals will be low, as will be at least one of the DONE i
signals. This causes the other input of gate B to be low
as well. Once all motors are finished moving and all DONE
i signals are again asserted, gate B will assert the flag.

If the command given was illegal, it will have been

111.

N: Du

 

 

 

an cc
9. ones;
n< OD

l

 

 

ri—

 

 

 

 

 

 

 

 

 

 

 

 

 

.mwouaom xooHo
can moan mommumusw umHmEmmousm on» mo Emummww oaumsmnom

.sm whamfim

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

u: so. \Arlux coo
Icl HI N: 00 cal I|I
o..H 3.an .v W o..H .o.% u_.u
o.~ .u n . a: ox o.~ n“ a _
x.o.o non o<oacs_uu. a. on»
a c o oo _<oo a c o
x. x. ,
L ikln+ E LL} no
N: no»
c .thh
ale 3 o E
p:. x<o
o u
m uzoo
c uzoo
» uzoo
m >mzoo
( . 2m
cunux<m paw» a.xu .
F; .l A. :“w

 

112

blocked by some error detect logic network. As a result,
at least one of the ERR i lines will be high, which causes
gate B to signal the flag immediately to notify the com-
puter that something was wrong. Finally, if the command
given was legal, but redundant in that what it requested
was already true anyway, none of the DONE i signals will
leave the high state, and again the flag will be signalled
immediately. The fact that all four motors are done with
their tasks is available to the computer as status register
bit 11. This facility is provided so that the computer
can test for a stable state of the sampler without having
to wait for the skip test or risk pulling an interrupt
which may not be opportune. The complete command set

is shown in Table 6.

Also shown in Figure 27 are the two 555 astable multi-
vibrators and their associated frequency dividers which
are used as the sources of the clock frequencies for the
various motors.

The remainder of the sampler interface need not be
illustrated. As was the case with the positioner, it
consists of the usual circuitry for decoding device
identities and commands from the.computer, circuitry for
transferring data of the various registers into or out
of the accumulator, and circuitry for managing the skip
and interrupt lines. The INIT signal is again derived
from the interface as a combination of the computer
initialize pulse or a local pulse generated at power

up time.

113

Table 6

Computer Command Set for Autosampler

 

 

6301:CMD l - Combined motor command for vertical, rota-

tion, and turret motors

6302:5TATUS - Input status word of sampler to computer

6304:CMD 2 - Motor command for syringe motor

63ll:CLR SAM FLG - Clear sampler flag

6312:SKIP TEST SAMPLER - Skip test of sampler flag

6314:SAM INT - Enable or disable sampler interrupt

facility

 

 

114

5. Power Supplies and Physical Structure

The electronics for the autosampler are assembled
on printed circuit cards of original design and construc-
tion mounted in a commercially produced card cage. The
logic of the vertical and syringe motors are contained
on one board, and that of the rotation and turret motors
and the sampler flag on a second. These boards are general
purpose integrated circuit cards having room to accommodate
as many as 42 fourteen or sixteen pin devices, with power
and ground buses. Signals are brought to the boards via
double sided edge connectors of 58 contacts per side on
tenth inch spacing. All chips on these boards are socket
mounted and interconnected with wire wrapping. Decoupling
capacitors are provided at each package. The three front
panel controls associated with the vertical motor are
mounted on a simple aluminum bracket secured to the cir-
cuit board.

The sixteen motor coil drivers for the sampler motors
are mounted on a third board. A fourth board similar to
the logic boards is used to contain the computer inter-
face, with connections to the computer terminal being
made via the usual ribbon cable and paddle board system.
The four boards communicate with each other through a
backplane bus wire wrapped among the terminals of the
various edge connectors. Signals taken out to the sampler

for powering the interrupters and microswitches are routed

115

with bundled groups of ordinary hookup wire, which are
demountable at the wall of the card cage through use of a
nylon connector.

The power supply used for the sampler logic is a
commercial device capable of supplying as much as 12 A of
regulated +5 V. The large supply is used because the card
cage contains the circuitry of the data acquisition system
as well as that of the sampler, and even then is only half
full. The supply used for the stepper motors is a +25 V
filtered, but unregulated, device of original, but ordinary
design. It contains the elements of the power supply, a
cooling fan, and the motor ballast resistors. Connections
from the supply to the motors are made with a socket
demountable multiple conductor cable. Return lines from
the motors to the motor driver board are similarly joined,
and a single ground wire of heavy gage completes the cir-

cuit back to the motor supply.

6. Testing and Performance

All facets of operation of the autosampler were
tested with apprOpriate computer routines, and changes
were made in the original logic design until proper per-
formance was realized. This process was simpler than that
encountered with the positioner because of the separable
nature of the sampler motor movements and the use of self-

guiding logic rather than control sequencers in executing

116

the operations. It was at this time that the proper fre-
quencies of motor stepping were selected and permanently
hardwired.

A very extensive characterization of the sampler as
an instrument for use in electrothermal atomic spectroscopy
was undertaken by Mr. Pals, to whose thesis the reader
is referred for evaluations of performance under a variety

of conditions.

7. Operating Software

Two assembly language routines written by Mr. Pals
are used to control the sampler, both of which operate
on the principle of storing in memory a series of words
whose values represent the prOper combinations of bits
to direct various motor motions. The simple act of samp-
ling, for example, involves a descent to the braid, a
deposition of the sample, and a return to the top. The
bit patterns for a vertical descent, delivery of an amount
of sample specified by a passed argument, and vertical
ascent are stored in sequential memory locations. When
the controlling routine calls for sample deposition there—
fore, it will call the sampler subroutine in such a manner
as to summon these three words from memory one by one and
output them to the sampler along with the proper motor
command. Skip tests occur between each such word to allow

completion of the task. Status register checks occur

117

after each step for an illegal command or malfunction.
Should an error be detected, the computer will halt to
prevent possible damage to the sampler through further
commands.

One of the two subroutines operating in this manner
is used to initialize the sampler. When called it directs
the sampler, via the self indexing word system described
above, to execute the following series of motions: ascend
to top, rotate over turret, send turret to cup one, send
turret to cup zero, descend to turret, empty syringe,
ascend to tOp. This sequence of operations assures that
the sampler will legally "find" all of the optical inter-
rupters properly, and leave itself in a well defined
position: over turret cup zero (reserved for waste) with
an empty syringe. (The double turret command initializa-
tion was previously explained in the turret logic, and
total emptying of the syringe to strike the lower limit
microswitch is assured by two successive deposit sample
commands of maximum quantity.) This routine may be called
whenever desired, and contains a built in check to assure
that it is called before regular use of the sampler is
attempted.

The second subroutine manages routine use of the
sampler. When first called to load the initial sample,
or whenever called afterwards to change to a different
sample, it directs the sampler to the turret to empty its

contents into cup 0, twice fill (cup 1) and empty (cup 0)

118

itself completely in a rinsing process with deionized water,
fill completely with the proper sample, and finally deposit
the same in an argument specified amount on the braid.
The routine then keeps track of what the syringe contains,
as well as a running tally of how much so that in succes-
sive sampling operations, it is able to choose among three
possible sequences of code words either to deposit another
sample, return to the turret to refill with more of the
same sample solution before depositing it, or change
samples entirely by executing the full rinsing Operation
described above. It is obvious that creation of new or
modification of existing code word sequences can be used
to alter the sampler's activities as desired.

At present the sampler is not interrupt driven, and

so no interrupt facility is in use in the software.

D. DATA ACQUISITION SYSTEM

1. Multifrequency Real Time Clock

Data acquisition for the graphite braid system con-
sists of performing repetitive analog-to-digital conver-
sions at Specified rates upon the appropriately amplified
spectrometer photomultiplier signal. The final major
piece of instrumentation to be discussed is a computer-
controlled, dedicated, multifrequency real-time clock for

providing the necessary timing and signalling for such

119

conversions. It consists basically of a crystal clock
source whose frequency is divided in orders of magnitude
by a LSI device, the Mostek 5009 divider, and then further
divided by smaller amounts with familiar logic to produce
an output frequency which is passed to the convert input
of an analog to digital converter. The logic associated
with the device is illustrated in Figures 28 and 29.

Figure 28 shows the clock source and the latch cir-
cuitry used to receive the desired frequency of operation
from the computer. The clock source is a 1.000 MHz quartz
crystal which is activated and maintained by a simple
series resonant circuit. The output of this circuit is
squared up by an RS latch and symmetrically divided to 100
kHz by a decade counter. As shown in Figure 29, the 100
kHz line is passed into the external input of the 5009
divider, where it is divided by orders of magnitude accord-
ing to the data present at the control inputs of the chip.
The output from the 5009 is in turn passed into a second
decade counter and a flipflop configured to toggle. The
counter A0, counter Do' and flipflop Q outputs represent
divisions, respectively, by 10, 5, and 2 upon the output
of the 5009. These three lines are fed into three inputs
of a four-to-one multiplexer, whose fourth input is ground-
ed, being unused. The output of this carries the selected
frequency clock train to the A/D converter via some con-
cluding circuitry which will be dealt with later.

At this time, it will be best to discuss how the

120

.EwumMm GOwuwmwsvom upon onu Hem hupflaouao voodoo
mucosvoum can monaOm xooHo on» no Emuomwo oaumfimsom .mn muswfim

buoj
we an 2. oc
N2: 0.. o< one» no
.< co

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   

 

9*HWIIIII1
x. 2.... sail—o o o 3

0
mm: x3110 O h o<

 

 

 

 

 

 

m40<zu

121

.Emumhm cowufimwsvom mama way how huufisouwo
mcwumm can mcflow>wv mocmsvoum ogu mo Bouquet oaumemnom

.mm magmas

 

 

 

 

 

 

— 13%
N 22%

m

ON in N .N

ow: x3! rum
_ allows x::
w m
0 ~ SL150 oooo talus. 2:

a r _ I on on we 5:... .3
high—:—

wan<zm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

     

 

 

 

o old/”xino o m
p¢u>zou

no
2.
< 2 u
h>zou z<8 \) n b u<
(a. u a AT. <VIQH§¢<2~

w: on

 

 

 

 

 

 

 

122

control lines of the 5009 and MUX are used to determine
the A/D conversion frequency. Refer back to the remaining
section of Figure 28. Bits 7-11 of the accumulator are
used to control the frequency of the clock. These bits
are latched out into the clock circuit when the computer
command ENABLE is given. Bits 9-11 are used to address
the 5009, and thus control the exponent of the output
frequency, whereas bits 7—8 address the MUX and govern the
mantissa. The complement of the latched bits 9-11 is sent
to a four-bit full adder, where a constant binary five is
added to them. The outputs of this adder are then used

to address the 5009. (This seemingly strange series of
events provides a convenient code for remembering what
frequency the clock is producing, as will be shortly seen
by example.) Bits 7-8 are latched and sent directly to the
MUX as its control lines.

To give an example of how the system functions, con-
sider what frequency would be produced if bits 7-11 of the
accumulator contained the binary pattern 10010 at the time
ENABLE was asserted. Let us first deal with bits 9-11.
They will be latched out as binary 010. The complement
of this, or 101 will be sent to the four bit adder. There,
a constant of five, or 101 will be added to them, producing
the result 010, with the carry into the fourth adder bit

being discarded. Thus, the 22, 21, and 20

lines of the 5009
will be addressed by 0, l, and 0 respectively. From the

manufacturer's truth table, this combination of address

123

bits will cause the 5009 to divide by 100, producing a
frequency at the 5009 output of 1 kHz. Bits 7-8 control
the MUX, and with the pattern 10 in these hits, the select-
ed channel will be that of the 2 input, which is the divide
by five output of the second decade counter. Thus, the
frequency at the output of the MUX will be one-fifth of 1
kHz or 200 Hz, and it is this frequency at which the A/D
converter will function. The use of the circuitry employed
to achieve this is now more reasonably seen. The original
binary word used to produce this frequency was 10010.

If a space is put between bits 8 and 9 in this word, it
reads 10 010. In this manner, bits 7 and 8 are in binary
the mantissa of the frequency, or two, and bits 9-ll give
in binary the exponent, also two: 2 x 102 Hz. With this
arrangement, it is easier to remember how to construct
computer software to decode and request frequencies of the
desired value. The complete list of frequencies available
from the clock is shown in Table 7. Addressing both lines
of the MUX with 0 selects the grounded channel, making the
device inactive. Also, addressing both lines with l
selects a mantissa of five, not three. It can also be

seen that the three hits used with the 5009 actually func-
tion as a three bit two's complement word for selecting

the exponent, with the one notable exception that bit
pattern 100, which in strict two's complement should be
negative four, is in fact positive four in this applica-

tion. The total range of frequencies selectable runs from

124

Table 7

Frequencies Available from the A/D Clock

 

 

 

Bits 7-8: Mantissa Bits 9-11: Exponent
00 - inactive 011 - 103 Hz
01 - 1 010 102
10 - 2 001 101
11 - 5 000 100
111 10"1
110 10’2
101 10’3
100 104

 

 

125

50 kHz to l mHz, values which are respectively too fast
for the computer to handle, and orders of magnitude slower
than typical atomization times from the braid.

The remaining circuitry of Figure 29 may now be cover-
ed. It consists of a system used to start the clock at any
desired frequency with minimum time jitter. A final ac-
cumulator bit, bit 6, is used as an enable bit to start
the clock. If bit 6 is present when ENABLE is given, a
monostable is fired, which sets all portions of the fre-
quency division chain to their maximum count values. This
assures that when the monostable pulse concludes, there
will be a maximum time jitter of only 1 us between the
conclusion of the pulse and the first rising (converting)
edge out of the MUX no matter what the selected clock
frequency may be. This circuit in effect assures a nearly
exact synchronization between the crystal clock and the
computer to within a margin of error insignificant with
respect to the operating clock frequencies.

Whenever the ENABLE command is given, flipflOp A
is cleared. If bit 6 is present also at the time, A will
be reset again after 10 us, and its outputs will permit
the clock frequency coming from the MUX to be passed on
to the A/D converter as the signal CONVERT. If bit 6
is not present when ENABLE is given, however, A will remain
clear as the monostable is never fired, and clock fre-
quencies coming from the MUX will not be passed to the

converter. Instead, the line MAN CONVT, or "manual convert"

126

will be active, and allow conversions to be requested from
a source other than this clock if so desired. (The main
reason this seemingly cumbersome flipflop arrangement is
used is that it prevents spurious edges from passing down
the CONVERT line as the clock frequency is being set up,

a potentially important service if data acquisition should
ever be incorporated into an interrupt driven facility.)
Typically then, when this clock is used, it is activated
at the start of a data acquisition interval by issuing
ENABLE to latch the frequency data along with bit 6 to
start the clock. At the conclusion of acquisition, ENABLE
is again given with at least bit 6 low to turn the clock

off.

2. A/D Converter

The signal CONVERT is passed directly to the start
convert input of the A/D converter, which is a 12 bit, 8
us device. This converter is wired for unipolar, 0 to
+10 volt operation, with trimpots provided for the adjust-
ment of gain and offset as recommended by the manufacturer's
data sheet. The digital output is complementary straight
binary.

Figure 30 shows the flag and data reading circuitry
associated with the converter. The flag is a straightfor-
ward device signalled by the conclusion of the converter's

BUSY line. Upon being set, the flag may be queried by the

127

computer as usual either by the command SKIP TEST A/D or
by pulling an interrupt if this facility has been enabled

with the computer command A/D INT and bit 11 of the accumu-

 

lator. The computer may also assert CLR A/D FLAG to pre-
pare for the next conversion.

The digital data from the converter is inverted back
to straight binary and strobed onto the accumulator input
lines with the command READ A/D. Table 8 gives a summary

of all computer commands for the data acquisition system.

3. Power Supplies and Physical Structure

The circuitry of the data acquisition system is housed
in the same card cage used to contain that of the auto-
sampler. The logic of the clock and flag is contained
on a general purpose logic board of the type previously
described for the sampler, and is physically assembled
with an identical philosophy as well. The A/D converter
itself as well as the twenty four gates immediately as-
sociated with its digital outputs are contained on a second
board. All unused space on this board is left covered
with copper laminate to serve as a ground plane for the
converter. Manufacturer's recommendations for suppression
of noise and routing of grounds are observed. The analog
input is a length of coaxial cable terminated with a BNC
connector.

The data acquisition system is serviced by the same

128

.wnuwsouflo onwvmmu mumn can moan 00mm
unmucfl Emummm coHuwmwsvom camp on» no Emummav oaumfimnom .om musmflm

Ame—8 Nnv

IIEHHLFFF

ol—u u<

 

in

 

 

Ilit...

O\< D<w¢ h h—O D\<

hlrfirkbk

u: D\< hum» A!“

bk”

0 ul»:— O\<

 

 

 

Zn
4... D\< 140

pa
TA?

 

 

01:8. 0...

 

 

o u loAlzao

 

129

Table 8

Computer Command Set for A/D Converter and Clock

 

 

6521:ENABLE - Start or stop clock and prepare desired

frequency of same

6522zA/D INT - Enable or disable A/D interrupt facility

6524:CLR A/D FLG — Clear A/D Flag

6531:MAN CONVT - Do an A/D conversion without using the

clock

6532:SKIP TEST A/D — Skip test A/D flag

6534:READ A/D — Input digital output from A/D converter

to computer

 

 

130

computer interface used for the autosampler, as well as
by the same wire-wrapped backplane for distribution of
signals.

The +5 V power requirements of the circuitry are
met by the same supply which provides +5 V to the auto-
sampler logic. The 115 V demands of the A/D converter
are met by a separate supply of commercial origin. (The
-15 V side of this supply is also used to provide the -12
V required by the Mostek 5009 chip through the expedient

of a simple two resistor voltage divider.

4. Testing and Performance

The logic of the data acquisition system was debugged
as usual with computer test routines. In addition, the
A/D converter was calibrated to zero offset and exact 0
to +10 volt (really 0 to +9.9976 volt) operation by adjust-
ing the gain and offset potentiometers with the assistance

of a voltage reference source.

5. Operating Software

The data acquisition system and its dedicated clock
are controlled by a sophisticated series of subroutines
written by Mr. Pals which control the frequency of the
clock as specified by the user and arrange to acquire

and store in a buffer region of memory the results of

131

analog to digital conversions. These routines process

the converted data in real time directly into OS/8 Fortran
IV compatible floating point numbers through machine code
techniques too complicated to warrant discussion here.
Because of this real-time processing, it is possible to
acquire data at rates as high as 1 kHz or more, which is
far beyond the frequency attainable by acquiring and
processing the data through the traditional method of

subroutine calls managed by the Fortran IV run time system.

E. PERSPECTIVES

A distinct disadvantage to most academic instrument
building is that the complex instruments built as one of
a kind devices often suffer a lack of the benefits of
prototyping. A system too complicated to check out by
bread-boarding must simply be built with as much fore-
sight as possible and then corrected or modified as neces-
sary within the confines available. So it is with these
instruments. Having built them, I have several thoughts
on how they could have been better built. I mean by this
various conceptual matters that if observed would have
produced or could now improve an instrument having more
capable, convenient, or reliable function than that which
it now possesses.

For all three devices it may be immediately noted

that if their design were begun from scratch tomorrow,

132

they would almost certainly be configured around micro-
processors dedicated to managing each individual instru-
ment's tasks, probably from a basis of local subroutines
held in read only memory. All requirements of the instru-
ment would be handled by its micro, which in turn would
receive its assignments and data from the central mini-
computer. Microprocessor control would provide a reduc-
tion in the amount of hardware logic required by the instru-
ments. It would also give them the flexibility to customize
their sequences of operations to accommodate virtually any
reasonable task without the necessity to reconfigure elec-
tronic hardware. Finally, the software burden borne at
present by the minicomputer could be substantially lowered.
Even if micrOprocessor-based instruments were not possible,
however, there are still a number of improvements which

could be made in the units as they are currently built.

1. Positioner

A modest but useful addition to the positioner would
be the use of seven segment decoders and displays to pre-
sent visually the contents of the location register. When
under computer control, the user can only estimate the
location of the atomization cell by observing its physical
position, and even under local control, the thumbwheel
switches only illustrate where the cell might be, not where

it necessarily is. A straightforward LED display of this

133

information would be an appreciable convenience.

A more significant change would be the substitution
of a static for a dynamic method of performing the front
panel BCD-to-binary conversion of destination data. The
present method, while logically sound, is cumbersome.
Fully three states of the GO cycle, numbers 1-3, are
dedicated to managing this system. Alternative methods
of performing the conversion, such as using the 74184
BCD-to-binary converter or the 7483 four bit adder185
would help to simplify the instrument. Because these
methods would perform the conversion directly from the
state of the thumbwheel switches on a constant basis, the
three states of the GO cycle mentioned above could be
eliminated altogether, which would compress the cycle to
only five states, and would simplify the GO sequence from
a branching to a nonbranching one. The distinction between
local or computer-controlled operation would be removed
from the sequence as well, and a possible saving in chips
might result, although this is questionable because of
the greater—than-unity relation between the number of
converter packages required and the number of BCD decades
being converted.

A third change would be a major restructuring of the
RESET cycle. At present, when a RESET is performed, the
positioner assumes that the action was not redundant;
that is, the cell was not already reset when the command

was given to do so. This is not necessarily the case in

134

practice. Consecutive RESET commands can be given, and
each such extra command will move the cell one motor pulse
farther past the true reset point into regions of hypo-
thetical negative position. In addition, if the cell is
physically moved so that the metal piece which signals
the reset is screwed far down into the interrupter slot,
the positioner cannot recognize the difference between
this "reset" and the true reset achieved when the photo-
electric beam is just barely cut off. To combat this
problem, I would reconfigure the instrument to perform
resets via the sequence illustrated in Figure 31. The
sequence begins by moving the cell in the positive direc-
tion, away from the reset point, until the interrupter is
definitely seen to be not tripped. Then an extra series
of steps in the positive direction is taken, perhaps ten
locations worth, to eliminate gear hysteresis, and then
finally motion in the negative direction is begun and
continued as is done presently until the true reset point
is found. This new sequence would guarantee the validity
of resets with respect to the above problems, and would
yield a sequence shorter than the old one, nonbranching,
and transparent to manual or computer-originated requests.

Mechanically the positioner has always performed well.
As presently constructed, it lacks very little that is

either essential or convenient to its intended purpose.

135

.maomo Bummm m HON mcowuocam

 

 

O uh<hmATlll

 

  
 

ecumm-Ch

 

 

 

 

 

n zonruzau

 

u zonhozau

 

 

971

turn—Owe

 

 

zo—poucno
u>~h~woa

Ur<o
mango

8045¢w>o
¢<uau

 

 

 

 

n zonhuzau

 

 

< zo—huzau

can mwumum mo 30Hm 00>OHQEH

 

 

.Hm magmas

 

136
2. Autosampler

The electronics of the sampler have served well.
Generally speaking, they control the movements of the
device as thoroughly and properly as desired. Two small
points might be mentioned with respect to them though.

It would be convenient to have the turret capable of moving
in both directions, so that it could back up from say, cup

1 to cup 0, instead of having to rotate all the way around
as it now does. A direction control, perhaps best handled
in the operating software, could productively extend the
sophistication of the turret. In addition, the gate circuit
used to start or stop three of the motors could be replaced
with the edge triggered Jk gate circuit of Figure 32, which
accomplishes the same task of producing a synchronized,

clean waveform with elegant simplicity.185

(Figure 32
illustrates how this gate would be connected with the
rotation motor signals, for example.)

Mechanically the sampler could stand some improvement.
The movement of the turret by a stepper motor, particularly
a large angle stepper, is rather jerky, and causes solu-
tion splashing at times. The turret might well be better
turned by a small DC motor or DC servo and position indi-
cator system in order to provide smoother motion. Another
problem is the drive of the rotation system. At present

it is also rather jerky and somewhat under torqued. A

fivefold gear reduction added to the drive train should

137

DONE B-— —-100 HZ

CMD 1
O J O MOTOR DRIVE 8
mj >‘c T
T

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 32. Schematic diagram of an improved gate circuit
for the autosampler motors.

138

be of great benefit in correcting these problems.

There is also the question of having the sampler
delivery system be only a multi-sample capacity device.
Many commercial analogs of this instrument employ a simple
tube delivery system that picks up and deposits samples
one by one from the sample tray instead of filling a
reservoir for multiple sampling. In much of the work
done here, multiple sampling is desirable, but in most
real sample situations, frequent sample changing is far
more common. The possibility of another autosampler de-
vice, either commercial or self designed, to provide the
capability for handling such situations, is one worthy
of consideration. Some problems also exist in the sampler
due to air space in the syringe barrel which acts like a
cushion when the plunger is moved, and can prevent the
deposition of reproducible sample volumes. Designs of
new syringes that minimize this problem are of potential

merit.

3. Data Acquisition System

The analog-to-digital conversion system and dedicated
clock are rather simple devices. Any appreciable changes
in them would result only from significant changes in
either the manner or type of data collected, such as
synchronization of the clock to an alternating pulsed

hollow cathode source lamp system for purposes of

139

correcting nonspecific background absorption. It may be
pointed out, however, that future extensions in the sophis—
tication of this atomic absorption system might well extend
into just such cases Of multichannel data acquisition, or
entirely alternate data acquisition systems such as multi-

channel detectors.

F. MISCELLANEOUS INSTRUMENTATION

A few small pieces of instrumentation remain to be
mentioned which are not part of the three major instru-
ments, but yet are of essential importance to the entire

Operating system.

1. Braid Heating Level Converter

The most significant of these small pieces is the
digital-to-analog converter used to set the heating levels
of the graphite braid. It is a straightforward device
illustrated in Figure 33. Digital data representing the
desired heating level of the braid are latched out to the
D/A converter with the computer command BURN. This data
is supplied from a subroutine written by Mr. Pals which
receives the desired level as an argument between 0 and
10 V from the calling routine and scales it accordingly
as a binary number between 0 and 4095. Like the A/D con-

verter of the data acquisition system, the D/A is a

140

.muufisoufio Houusoo Hm>ma mcfiusms wanna man no Emnmmfip owumsmsom

on<¢m or

OONO

klélo a:

 

f
1
.1“...

i D In <31

To \ngri
x~.Nt

{\leéftlhaapao <\o
xn.No

 

 

 

.mm «Human

llfi run 04
Ilzcan

 

1.:

141

complementary device, and so it is addressed by the 5
outputs of the latches. These latches may be cleared,
turning off the braid, either by latching out all zeros
with the BRAID command, or by simply giving the command
KIEE—Efififi, which is an ORed combination of a computer
command and a front panel push button switch.

The D/A converter is wired and calibrated for uni-
polar, 0 to +10 volt output from complementary straight
binary input as recommended by the manufacturer. The out-
put Of this converter is fed to a low noise, low drift
operational amplifier, where it is amplified by factors
of anywhere from 0.1 to 1.5 in steps Of approximately 0.1,
as selected by a binary combination of switch closures
from a miniature switch array. This facility is used to
scale the output of the converter to levels appropriate
with respect to the atomization temperature of the element
under study. In addition, current may be supplied directly
to the summing point of the CA from either of two push-
button front panel switches. Their heating levels of
the braid correspond roughly to a few hundred degrees and
to about 1500 degrees, for purposes of conveniently de-
gassing newly mounted braids.

These facilities are contained on the same circuit
board used to hold the A/D converter of the data acquisi-
tion system, and are powered from the same supplies.

The final voltage output to the graphite braid power '

supply controller is provided by a length of coaxial cable

142

terminated with a BNC connector. This circuit board should
be redone to replace the miniature mechanical switches

with latches and PET switches so as to put the range
scaling facility under the scope of the computer's con-

trol.

2. General Purpose Flags

The most common technique for regulating the heating of
the braid is radiation programming, in which the blackbody
radiation of the braid is monitored. Radiation program-
ing is only possible if the braid is incandescent within
the spectral response curve of the sensing phototransistor,
which is not the case at the low levels of desolvation.
When desolvating, the braid heating is regulated by moni-
toring the electrical power dissipated in it, and radiation
programming is only activated at the onset Of atomization.
A flag is necessary to signal this transition. Similarly,
there are other events which may conceivably be triggered
at the onset of atomization, such as scope sweeps or chart
paper tracings, which also require controlling flags.

The interface board of the sampler-acquisition card cage
contains provisions for three such flags. They are merely
individual D flipflops set by a computer command FLAG if
an accumulator bit at the D input is high, and cleared
with FLAG if that bit is low. The outputs of these flags

are available to accomplish whatever task is desired.

143

A subroutine written by Mr. Pals governs their activities.
One such flag is presently in service to control the switch-
ing from power to radiation programming at the onset of

atomizations.

3. Computer Conveniences

Finally, because the entire atomic absorption system
is Operated through use of a remote interface facility
which communicates with the minicomputer over a distance
of some 100 feet, it is valuable to have certain conven-
iences available to minimize trips back and forth from the
instrument to the computer. Three such conveniences are
present on the card cage interface board. One of them is
a simple LED indicator which informs the user of whether
or not the computer is running. A second is a debounced
momentary action switch which when depressed fires a several
us monostable which forces the computer to perform a skip.
This is convenient if the computer has become stalled in
skip testing a flag which cannot be accessed due to, say,
oversight in hooking up the instrument in question to the
interface terminal. The final device is a set of six
miniature switches, which can be read with a computer com-
mand to simulate options called for by settings in the
computer's front panel switch register.

Table 9 gives the computer command summary for these

final instrumental sections.

144

Table 9

Command Summary for Final Instrumental Sections

 

 

6111:

6112:

6114:

6121:

Latch digital data out to heating level D/A

converter (BURN)

Address control flags (FLAG)

Clear braid D/A converter to zero (KILL BURN)

Read miniature switch array

 

 

145

G. REMOTE TERMINAL FACILITY

The complete atomic absorption instrument in use
here occupies a goodly portion of an entire lab bench,
and cannot physically fit around the laboratory mini-
computer. To allow for computerization of the instrument
without sacrifice of physical room or permanence Of loca-
tion, a remote terminal facility was designed and con-
structed which transports all fifty three computer signals
available to the outside world through the DEC Positive
I/O Bus Interface Option back and forth between the com-
puter and the instrument, a distance of some 100 feet.

The Bus Interface Option is traditionally accessed
by the user through the Heath Computer Interface Buffer,
commonly known as the "buffer box". For the remote facil-
ity,a1terminal is provided which sits adjacent to the buffer
box and connects to it with three short patches of paddle
board terminated ribbon cable. Thirty seven of the sig-
nals from the computer travel from it to the instrumenta-
tion being interfaced. Each such signal is sent to an
integrated circuit differential line driver which, with
proper impedance matching and termination, places it onto
the leads of a shielded, twisted pair wire group. These
differential, paired signals are carried through two
large cable bundles across the laboratory false ceiling
to a central location in the atomic spectrOSOOpy lab,

where they are brought down to a second terminal which

146

contains identically matched and terminated differential
line receivers for each of the signals. The restored TTL
output of these receivers is presented to the user on edge
connectors physically identical in number and pattern to
those of the commercial buffer box. The remaining sixteen
signals travelling from the instrumentation to the com-
puter simply duplicate the process along the reverse route.

With this facility, the full interfacing capabilities
of the computer are available in the alternate laboratory.
Tests of the facility reveal that it can routinely trans-
port the various signals with excellent noise immunity and
at frequencies even exceeding those at which they are
normally produced by the computer.

Two additional cables are also laid across the false
ceiling which provide for remote connection of a CRT
terminal and line printer for user-computer communica-
tions. The nature of these devices' interfaces is that,
unlike the interface signals, they do not require special
driving and receiving techniques for reliable data trans-

mission.

MECHANICAL AND OPTICAL INSTRUMENTATION

AND BASIC SYSTEM PERFORMANCE

A. THE ATOMIZATION CELL

A considerable amount has already been described about
the graphite braid atomization cell in the discussion of
the positioner. There remain, however, several points of
significance which may now be dealt with.

The atomization cell is pictured in Figure 34. It
is constructed around an aluminum base and support which
secures to the side of the optical rail. Upon this struc-
ture is mounted the horizontally moving translation stage
and its motor. To that stage is bolted the vertically
moving stage, and to that in turn, the vertical motor and
atomization cell proper. In addition, both of the lenses
of the optical system are supported from the baseplate
as well.

The cell itself is structured around a hollow, rec-
tangular aluminum chamber which holds the components of
the sheath gas flow system and serves to prevent air from
getting into the cell through the bottom. A shelf is cut
around the entire upper perimeter of the inside of this
chamber, upon which rests the atomizer block. This block
consists of two fixed, aircooled brass electrodes for

holding the graphite braid, and machined plexiglas for

147

148

.HHTO cofiumNHEOOO Tau mo :oflumuumOHHH

. A.

.vm munmfla

 

149

securing the electrodes and supplying necessary mass where
needed to complete the structure. The interior of this
block, slightly over one square inch in area, is open,

from which a continuous flow of sheath gas passes the
braid at all times during use. The electrodes grip graph-
ite braids by simple pinching. By loosening a screw, a
section of the electrode may be slid back, exposing an
Open space into which an end of a braid may be placed.

By pushing the section back again and tightening the screw,
the mount is completed. Two simple knurled screws also
serve to make connection of the electrodes to the power
supply. The design of the electrodes is such that the
braid itself is the highest object in the entire cell.

This makes possible the passage of a source lamp beam
directly at grazing incidence over the braid even if the
braid is parallel with and just under the optical axis.

The braid can be used in this position, or it may be turned
perpendicular to the optical axis by a simple changeover
operation amounting in essence to a ninety degree rotation
of the atomizer block.

OnLone side of the block, aimed at 45° with respect
to the braid, is a phototransistor housed in a short
plastic tube which monitors the braid blackbody emission.
During burns, the signal produced by the transistor may
be used as a sensitive and precise means of regulating
the heating of a braid to constant temperature.

The atomizer is protected from damage due to

150

atmospheric attack with a rectangular chimney fashioned
from glued pieces of eighth-inch thick supracil quartz
plates. This chimney sits in recesses cut into the plexi-
glas of the atomizer block, and extends upwards from the
atomizer to a height of about two inches. The interior of
this enclosed space defines the region available to the
positioner for the scanning of atomic concentrations.
Electrothermal atomizers require a flow of sheath
gas to carry atomized products through the optical beam
for analysis and to protect the atomizer from damage by
the atmosphere. In many devices, a simple flow of gas
such as argon is sufficient to accomplish these purposes.
In the present system however, a laminar sheath gas flow
of high uniformity is desired so that the profiles of
atomic concentrations reflect as little as possible the
eddy and swirl patterns of the sheath. It is not possible
to treat this problem too rigorously, as even the best
laminar flow of gas designable cannot be expected to main-
tain its integrity when an object in its path is suddenly
heated to as high as 2500°K; nevertheless, the establish-
ment of a good laminar flow is of high importance in
obtaining meaningful profiles. In the present system this
was accomplished by supplying the sheath gas through
capillaries. Ordinary glass melting point tubes about
three quarters of an inch long were held together with
household glue in a honeycomb-like array about one inch

square in cross section. This array is located in the

151

interior, Open area of the atomizer block, directly under-
neath the braid. It is held by a closed, rectangular
chamber made of plexiglas, which is mounted inside the
principal aluminum chamber of the cell. Sheath gas whose
flow rate is regulated by a rotometer is brought in through
the bottom of the plexiglas holder using tubing. It dis-
perses in the holder, and rises uniformly through the
capillaries past the braid. Once beyond the capillaries,
there is nothing to disturb the laminar flow except the
braid itself, and the short sections of the electrodes
which must be included within the chimney. A second,

slow trickle of argon is allowed to flow into the aluminum
body Of the cell during use to keep the cell interior
flushed free of oxygen.

The quality of flow provided by the capillaries may
be appreciated from the behavior of freshly mounted braids.
When a braid is first mounted, it must be heated Once or
twice to drive off contaminants. They are given off as
smoke, which can be seen to rise in a steady and smooth
pattern straight up from the braid all the way to the
chimney tOp, where it scatters in the atmosphere. Some
further observations along these lines were also done using
a smoke generator. In a humid atmosphere, titanium tetra-
chloride hydrolyzes to the oxide, which fumes in a dense,
white smoke. Observations were made in which argon gas
was passed over water and then through the cell, in which

was mounted a braid upon which drops of titanium

152

tetrachloride were introduced with a stirring rod. A
streamer of smoke was Observed to rise directly upwards
from the braid without dispersion over the entire height
of the chimney, although a slight wavering in the stream,
amounting to about one or two millimeters, was frequently
in evidence. It was also observed that unless the flow
rate of gas was on the order of 2.5 to 3.0 liters per
minute, a steady flow up the entire chimney height could
not be sustained; rather, the gas would stagnate and lose

its flowing qualities somewhere below the chimney top.

B. THE OPTICAL SYSTEM

The optics for a typical atomic absorption system often
consist of a simple lens to collimate light from the source
lamp and direct it through the atomization cell to the
entrance slit of the monochromator. In the present system
however, because of the desire to be able to isolate small
regions of space above the atomizer for analysis, and to
permit having the braid parallel to the optical axis, a
somewhat more elaborate arrangement was required. The
system used is illustrated in Figure 35. A large quartz
lens, having both diameter and focal length equal to two
inches, is used to collect light from a hollow cathode
source. This lens is located six inches from the cathode.
At its image focal point, three inches away, is located

the focal point of a second, smaller quartz lens of

153

.musmfifinmmxm cowumHOmnm owfioum on» How Emumam Hmowumo .mm Tasman

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

all
_
W I? Vol
mmahmwa< th<30 NPC<DG
z:0.m «\u :m.& «\u =O.N ._.DUI
:8.“ ‘ saw-m — :Quw '1—

154

diameter and focal length equal to one-half inch, which
acts as a collimator. The collimated light beam passes
through the chimney and cell to the entrance slit of the
monochromator. There it is limited by an aperture which
defines the vertical size of the region of space examined.
The horizontal size is defined by the width of the slit.
This combination of lenses gives an Optical beam which
has reasonably good intensity and which allows the use
of manageable amplifier gain settings and photomultiplier
voltages in spite of the limited field of view. It also
permits the braid to be parallel to the optical axis without
blocking the source light.

To be certain that the size of the space examined
is accurately reflected by the aperture size and slit
width, measurements were taken of the divergence of the
beam after it passes the collimating lens. The Optical
system was prepared as shown, and a hollow cathode was
used as a source. The collimated beam from the second
lens was directed into the body of a single lens reflex
35 mm camera, and photographed at two different lens-to-
camera distances. After film development and selection
Of representative exposures, the divergence of the beam
was calculated from the difference in size of the spots
at the two distances. The results showed about 1°30'
divergence from the Optical axis to the beam edge, or about
3° across the beam. These values are somewhat arbitrary,

as they will change somewhat with wavelength due to effects

155

of chromatic abberation, but as a general estimate, they
do confirm the fact that the beam collimation is of good
quality, and that aperture and slit sizes may be regarded
as fair estimates of the size of space examined in the cell.
Alignment or complete assembly of these optics is
greatly facilitated through use of a low power continuous
laser. The process is as follows. First, the laser is
started and used to define the optical axis. Care is
taken to assure that the laser beam is truly perpendicular
to the plane of the entrance slit, and centered squarely
on the slit itself. Next, the small lens is installed and
so aligned to keep the diffused laser spot centered on
the slit. Then the large lens is placed at its prOper
distance and aligned. Finally, the turret which holds the

hollow cathodes is secured at its proper location.

C. ATOMIC ABSORPTION INSTRUMENTATION

The remainder of the spectroscopic instrumentation
is all of commercial origin and essentially unmodified.
Hollow cathode lamps are powered by the Heath EU-703-70
Hollow Cathode Supply, and Operated at the maximum recom-
mended currents. The monochromator is the Heath EU-700
with accompanying wavelength controller. The photomulti-
plier is powered by the Heath EU-701-30 Photomultiplier
Supply. Generally, for work done at wavelengths above

300 nm, the photomultiplier used is an RCA 1P28. Below

156

300 nm, a solar blind Hamamatsu R166 is used instead, be-
cause of its significantly greater Spectral cathodic
sensitivity. The output current from the photomultiplier
undergoes current-to-voltage conversion with a Keithley
427 Current Amplifier, which provides adjustable gain,
offset, and signal rise time. The output voltage from
this amplifier is sent directly to the analog-to-digital
converter of the data acquisition system, and is usually
also monitored on an oscilloscope to facilitate the loca-
tion of analytical lines and adjustment of the gain to

provide a full 0 to +10 V output.

D. BRAID POWER SUPPLY AND CONTROLLER

The power supply used to heat the graphite braid is
of original design and construction, but was in existence

183 The raw supply is a

before the present work began.
high current, 50 V, full-wave rectified DC supply. Power
from it is switched into the braid on a regulated basis
from a bank Of parallel-wired, high current transistors
mounted on water cooled heatsinks and driven by a multi-
stage Darlington circuit. An instrument for regulating
the heating was constructed by Mr. Pals; the reader is
referred to his thesis for details of its operation.
Simply put, it consists of a control operational amplifier

which regulates the heating of the braid by comparing

at its noninverting input the reference voltage supplied

157

from the digital-to-analog convertor and amplifier already
described, with a feedback signal at its inverting input
derived from the braid. The feedback signal may be derived
from any one of four different sources, each of which has
appropriate sensing and forming circuitry in the controller.
First, the voltage dropped across the braid during heating
may be used. Second, the current passed through the braid
may be Observed. Third, the power dissipated by the braid
may be monitored by combining the voltage and current
signals in an analog multiplier. Finally, the intensity

of the blackbody emission from the braid as provided by

the phototransistor mounted on the atomizer block may be
followed. This final method is conceptually different from
the first three techniques and has several attractive vir-
tues which recommend it as a method of atomizer regulation;

they are discussed in a later chapter.

E. OPERATING SOFTWARE

Certain sections of the controlling software written
by Mr. Pals have already been described in the discussions
of the instrumentation. These routines are part of a very
extensive composite of software which supervises all aspects
of available control over the system. A brief summary may
be given here to clarify the manner in which the equipment
is employed.

The software is written in Fortran IV, with assembly

158

language routines in Digital Equipment Corporation's as-
sembler, RALF. The ability of Fortran IV to segment a
complete package of software into overlaying sections has
been of critical importance to this work, permitting the
execution in a 16K computer environment of a software
package that would require on the order of 28K of virtual
storage. The central point of this software is a routine
known as INPUT, which serves as a command interpreter for
receiving instructions. All parameters of operation are
under control of this routine. These include not only
obvious quantities such as atomizer powers or positioner
destinations, but also control variables such as whether
or not to plot results as they are acquired, record them
on a mass storage device, or wait for operator evaluation
between successive runs. When the software is first
started, INPUT "comes up" with a default set of values
for all variables. The Operator then modifies selectively
those he desires in order to achieve the configuration
pertinent to the work at hand. An extensive series of
error checks is built into the software. For example,
the computer checks to see that a reading of the dark
current has been taken before allowing work to begin.

The computer also flags improbable or unworkable values
for variables, such as a call for a larger number of data
points than there is room allotted in core for their
storage.

Once all parameters are set and checked for legality,

159

execution of the task begins as ordered. Several major
routines supervise the proper coordination of procedures,
while others arrange for the sequencing of repetitive tasks,
such as the systematic scanning of space by the positioner.
A routine is also available if desired to plot the results
of each sample on a high resolution CRT terminal. Perhaps
most ingenious of all is the fact that the software contains
provisions for including optional, user written routines
that may be used to suspend the normal execution of the
standard software to achieve unconventional functions from
the instrumentation. For example, a subroutine may be
included which suspends the normal method for the reduc-
tion of data in favor of the user's alternate approach.

The software can even be configured in a batch mode to
execute a predefined series of ordinary or non-standard

tasks without need of operator intervention.

F. SUMMARY OF EVENTS FOR A SINGLE SAMPLE

To appreciate the discussions to follow, a summary
may now be given of the general sequence of events occurring
in the process of collecting the data from one single sample.
The user is assumed to have prepared the instrumentation
for use (such as selecting the proper hollow cathode),
entered the desired values of all parameters to the input
routine, and performed initializations consisting of

resetting the positioner, initializing the sampler, and

160

taking a reading Of the dark current.

When control is relinquished by the user to the
system, the sampler begins to fill itself with the desired
solution while the positioner sends the cell to location
Horiz = 0.0 mm, Vert = 0.0 mm. A sample is deposited,
and the braid gently heated to drive off the solvent.

Next the cell is moved to the desired position of analysis,
and upon arriving, lOO readings of 100%T are acquired and
averaged. The braid is then successively heated to its
ash and atomize levels. During atomization, data are
acquired at the specified rate. After a brief delay to
allow for cooling, the three step cycle of desolvate,

ash, and atomize is repeated while data acquisition is
taking place for the purpose of background correction.
Upon conclusion of this second cycle, the cell is returned
to the zero location to receive the next sample, while
simultaneously the data from the sample just completed

are processed and disbursed as requested.

This basic series of events continues in this manner
for each successive sample (assuming the user has not
modified the normal sequence) until the entire assigned
task is completed, all without need for any intervention
on the part of the user, after which the system returns
control of itself to the user and awaits further instruc-

tions.

161

G. CALCULATION OF ABSORBANCE

There are four types of readings acquired from the
system. Nominally, they are the dark current, 100% T,
sample data, and background data. If these are numbered
one through four respectively, they can be written as

four equations:

where Id is the dark current, I0 is the 100% T level, IS
is the sample signal, and Ib is the background signal,
these last two being variable with time during a given
atomization. If these readings are combined algebraically
according to the following relationship, they yield the

correct definition for absorbance:

H

O _ (2-1)
I; ' 109 (3-4) + (2-1)

 

A = log

In the data reduction software, this calculation is
performed on a point by point basis. The routine uses

the average value for dark current, the average but

162

reacquired value for 100% T, and each associated pair of
sample and background atomization data points. The point
by point technique is better than the use of an averaged
background reading, as the background level changes during
the earlier portion of atomization while the braid is
brought up to final temperature. The correction is how-
ever Only useful for samples that possess a matrix which

is either nonabsorbing or can be completely driven off
prior to atomization. If such is not the case, and non-
specific absorption is present, other correction techniques
such as the use of a continuum lamp or analysis by standard
additions would have to be employed.

Because the signal from the atomizer is a transient,
the integrated absorbance of the sample is the ultimate
readout value. In the present application, this is simply
the sum of all the point by point absorbance readings.

As a result, integrated absorbance is in itself not a
meaningful parameter, as its value depends upon the rate

of data acquisition. If the rate is known, it is trivial
to scale values properly from different data sets. However,
a more disturbing phenomenon exists which makes integrated
absorbances even more nebulous, namely the fact that inte-
grated absorbance depends on peak shape.

While working with the system it was noticed that
increasing the rise time of the amplifier caused a de-

crease in integrated absorbance with no change in the

163

sample. For example, a series of atomizations of 2 02 Of

1 ppm silver were performed with rise times of 30 and 300
ms. For the former rise time, the integrated absorbance
of five samples was 10.84 i 0.33, whereas five samples
with the latter time yielded a value of 8.13 t 0.14.

These data would seem at Odds with the fact that a filter
does not change the total area of a peak, though it does
alter the shape. Indeed, if a signal generator and OA
integrator are used to simulate transients, the rise time
setting of the amplifier will, as expected, not affect

the areas under the transient, given enough time for them
to completely pass the filter. The explanation is that,
in the present system, the area of the peaks is taken
after the point by point conversion of the raw data to
absorbance, which involves the nonlinear operation of
taking logarithms. To illustrate, consider the idealized
pair of peaks shown in Figure 36. Peak 1, with small rise
time, penetrates to 10% T for one second. Its "integrated
transmittance" or true area may be taken as 9.0, whereas
with a sampling rate of 10 Hz, its integrated absorbance
is 10.0. Peak 2, low and broad because of the effects

of a filter, penetrates to 50% T for 1.8 seconds. The
true area is still 9.0, but the integrated absorbance is
only (18).(0.3) or 6.0. Thus peak shape is of great sig-
nificance to integrated absorbance, which indicates that
integrated absorbances from different systems or even dif-

ferent data sets may not be meaningfully comparable in

164

100— ~— ~—-—-—- ~—

90-—

80—— {-189 <—a.es—->

78-

68—-

 

 

 

'/,T50~ ————————————

Tau: 13:1 - 9.0
1+3 -- nu us - 0.0

38-

29 —- mu: AREA - 9.0
INT ABS - 10.0

 

 

 

10— -----

(ASSUME A IO HZ SAHPLINO BATE)

 

 

0

Figure 36. Illustration of the effect of readout time
constants upon integrated absorbance.

165

the absence of peak shape knowledge, even though the

samples are chemically identical.
H. PERPENDICULAR VERSUS PARALLEL BRAID ORIENTATION

The braid may be mounted either parallel with or
perpendicular to the Optical axis. When samples are
placed onto a braid and soak into it, it might be expected
that the parallel orientation would give greater sensi-
tivity because of longer pathlength, and less contribution
to background because the braid itself is not viewed as
directly by the monochromator.

Some simple observations of the braid blackbody
radiation at 650 nm indicate that turning the braid parallel
to the axis reduces the Observed braid emission intensity
by about an order of magnitude. This fact can be signi-
ficant for analyses of high boiling elements with reason-
ance lines at longer wavelengths, such as calcium at 422.7
nm. Table 10 Shows the improvement in integrated absor-
bance and precision realized for several elements when
analyzed with the parallel as compared to the perpendicular
geometry. The increase in integrated absorbance is rather
small, generally only a factor of one to four, which indi-
cates that longitudinal soaking of samples on the braid
is not a very significant phenomenon. Precision is better
in the parallel orientation as well, and again by roughly

similar factors. This improvement may well be due to the

166

Table 10
Improvement of Signal and Precision for

Parallel versus Perpendicular Braid Orientation

 

 

 

Ag Cd Pb Cu

IA * 7
pa/IApn 1.5 3. 2.8 1.8
Parallel 2.0% 3.1% 6.2% 8.4%
Perpendicular 7.3% 4.8% 20.0% 9.7%

 

 

*
Ratio of integrated absorbance in the parallel geometry
to that in the perpendicular.

167

fact that with a perpendicular geometry, small incon-
sistencies in the exact point of sample deposition are far
more critical than with a parallel orientation. The im-
provements do not show any clear cut trends; however, the
consistent improvement as well as the reduced background
level of emission do indicate that routine use of the

parallel orientation is to be preferred.

I. EFFECT OF ENTRANCE SLIT GEOMETRY

Because of the high collimation of the source beam,
the horizontal size of the space examined at any given
cell location is determined by the monochromator slit
width, which is typically only 200 um, fully an order of
magnitude smaller than the diameter of the braid into which
samples soak. An experiment was performed in which the
monochromator was rotated through ninety degrees with the
aid of a specially constructed platform. The purpose of
this experiment was to see what changes might be observed
if the 200 pm dimension was used to limit the vertical size
of the observation window rather than the horizontal.
The results are given in Table 11, with several different
apertures to define the other window dimension.

The two monochromator orientations appear to be
rather similar. For all three apertures, the normal
orientation gives slightly superior results, both in terms

of integrated absorbance and precision, although the

168

Table 11

Effect of Monochromator Orientation Upon Results

 

 

 

Traditional Rotated 9O
1 mm aperture 10.84 i 0.29 2.7% 10.14 i 0.29 2.9%
3 mm aperture 10.99 t 0.20 1.8% 9.43 t 0.31 3.3%
no aperture 10.29 i 0.22 2.1% 9.04 i 0.19 2.1%

 

 

Each data set consisted Of 20 samples of 5 ppm cadmium.

169

precisions are inherently so close to the 2-3% limit of the
sampling error that rigorous comparisons of them are
probably not very meaningful. The integrated absorbance
generally decreases consistently with aperture size; this
is consistent with the fact that the larger apertures

view a steadily increasing region of space which atomized
samples, through diffusive and convective spreading, are
not obliged to fill consistently. The behavior of these
two orientations may well change with other elements re-
quiring different atomization conditions than the humble
ones necessary for cadmium, but at least it would appear
that using the monochromator in the traditional orientation
does not ignore an Obvious means by which to improve the

quality Of results.

J. NATURE OF INSTRUMENTAL NOISE

Some brief investigations were performed to identify
roughly the predominant sources of noise in the system,
so that their effects might be recognized for possible
future instrumental improvements.

The single most persistent source of noise in all
of electrothermal atomic spectroscopy is what is generally
referred to as the sampling error, which reflects the
imprecision inherent in the delivery of sample volumes of
only a few microliters, not only in the variation of the

exact sample size, but also the variation in precisely

170

where it is deposited on the atomizer and where it might
migrate to or soak into on the atomizer as the heating
process begins. This source of noise is often the prime
contributor to the total system noise of 1-2% at best,
and can render a systematic study of other noise sources
difficult or meaningless.

For the present work a computer program was written
which would operate the data acquisition system at user
specified rates and output the raw twelve-bit data con-
versions to the terminal or line printer along with the
standard deviation of the set. With this program, data
sets of one hundred conversions each could be taken over
variable time periods.

The first item investigated was the amplifier and
data acquisition system itself. With no input to the
amplifier, and a large offset intentionally introduced to
place the output at nearly the full-scale limit of the
converter, one hundred data point sets were taken over
courses of time ranging from one second to about one
quarter hour. Amplifier gains ranging over the settings
commonly used in spectroscopic work, 106 to 1010 volts per
ampere were used. This portion of the system was found
to be essentially noise free. Regardless of the gain
or time of measurement, the standard deviations of all
data sets were significantly less than one LSB of the ADC.

Thus, the noise level of the readout is below the resolution

of the data acquisition system at least for single readings.

171

Virtually the same behavior was observed if the photo-
multiplier was connected as an input with its shutter
closed. Only at the highest gain settings, such as 109
V/A and 800 V would the dark current noise seem to reach
one converter LSB in magnitude. It was observed though
that at such high amplifier gains, shaking the cable con-
necting photomultiplier and amplifier would introduce a
very noticable determinate noise in the output level.

The use of special low noise cable for this connection,
and perhaps a critical evaluation of the wiring of the

photomultiplier base itself, might be of merit to guard
against noise introduced from mechanical vibrations.

Of more interest were some Observations taken using
hollow cathode lamps as the sources of signals. Two
lamps were studied. The first was a multielement lamp
from which the silver reasonance line at 328.1 nm was
isolated. The second was a lead lamp examined at 283.3
nm. The lead lamp was more worn than the silver, and
could not be used at as high a current rating, and thus
required significantly more gain from the readout to get
a solid output voltage. The settings used were: silver,
108 V/A and 700 v, and lead: 109 V/A and 610 v, both
with a 200 pm slit width. One hundred point data sets
were taken of the output from either lamp over times
ranging from one second to about one quarter hour.
Figures 37 and 38 show plots of the data sets over the

various time intervals. The ordinate values of the sets

172

.mponumo zoaaon um>afim ”:Oflus>HmmnO
mo mOOHumm msoflsw> Hm>o coflumucoeduumcw on» mo mOflumflumuomumno mmfloz .hm musofim

N- S;
mumzzz hzmom <h<o
900.1 ooo.o

Egg/EL...
HMN/HQUCH “coomm H

Hm>uwusw pcooom OH 1

Hm>HODCH psoomm ooa i

8"
WVNSIS

Hm>uwu2H psoomm oooa

l_.l

 

-Emm.~

173

.Oposumo 30Haos puma "sowum>nmmno
mo mpoflumm msoflum> Hm>o soflumucmfisuumcfi on» mo moaumwuouomumno wmaoz .mm whomflm

m- o;
mmmznz hzuom <h<o
ooo.l ooo.s

too” . m
H
L-

Hm>umucH pcoomm H UH
H x
1... I S

Hm>umuca pcoomm oa .. 0 mo.
-- “N
-1 . v
1.. 8 .II

Hs>umucfl pcoomm ooa

m>umusfl pcoomm coca

 

174

have been uniformly shifted where necessary to separate
the sets.

In the case of the silver lamp, a long term drift
effect is quite evident, in spite of the fact that both
lamps were allowed to warm up for more than a half hour
before any of this data was taken. The magnitude of the
noise is greater for the lead lamp, reflecting the higher
gain necessary to achieve good on-scale signals from it.
The quickest data set, taken over one second of time, is
comparable to the rate of data acquisition employed during
normal spectroscopic analysis. For this set, the silver
lamp shows a relative standard deviation of 0.17%, and the
lead lamp 0.32%. To compare with these results, a similar
set of observations was made with an ordinary tungsten
lamp as the source, the same gain and photomultiplier
settings as were used with the silver and lead lamps,
and a change in the monochromator wavelength setting to
get comparable signal intensities. The tungsten lamp
yielded results so indistinguishable from the hollow
cathodes that they need not be plotted. The one-second
data set under the silver lamp conditions showed a standard
deviation of 0.19% and that for lead conditions was 0.36%.
The tungsten lamp also showed the same general suscepti-
bility to long term drifts, and the same apparent random-
ness of Short term fluctuations as did the hollow cathodes.

One would expect a tungsten lamp to have inherently

superior short term stability compared to a hollow cathode.

175

Yet the very similar performance of the two sources in

the above studies would seem to indicate that shot noise,
and not source flicker is the principal contributor to

the noise magnitude. This is not an unreasonable finding
for a system like the present one in which light through-
put is severely restricted to a small region of space.

The drift of the hollow cathodes is not surprising. They
are regulated by monitoring the current passed through the
lamp. A superior method of regulation would be to do it
Optically, based on the intensity of the analytical line.
This would not be trivial to implement, as unfortunately
in atomic spectroscopy the sample cell preceeds the mono-
chromator. Optical regulation or compensation of the
source would require a double beam geometry or a synchron—
ously detected modulated source. Some sort of source
stabilization might well be desirable however because of
the time separated nature of the instrumental data.
Readings of 100% T, sample atomization, and background
correction all take place sequentially in time, with several
seconds elapsing between each. Uncompensated drift of the
source during such time could cause a significant contribu-
tion to the total imprecision of an analysis. For example,
a decrease in source intensity by 0.2% T between the read-
ing of 100% T and the sample atomization would make a
contribution of 0.087 to the integrated absorbance ob—

served during a one second atomization at a data acquisition

176

rate of 100 Hz. At this data rate, the transient signal
seen from, say copper at 1 ppm, yields an integrated
absorbance on the order of 16 or so. Consequently, the
source drift would make a contribution of roughly 0.5%

to the total imprecision of the analysis. This is not as
much as that expected from the sampling error, but it is
still a significant amount. It might also be pointed

out that this analysis would not at all be close to the

detection limit.

K. IMPORTANCE OF BACKGROUND CORRECTION

For several elements that can be analyzed with a
parallel braid at relatively low temperatures, the con—
tribution of the braid blackbody to the total photocurrent
is quite tiny. For example, a direct measurement at the
silver line of 328.1 nm of the blackbody intensity of a
braid heated to 1700°K (very adequate to atomize silver)
indicates that the blackbody contributes less than 0.01%
to the total photocurrent, a negligible amount. An experi-
ment was performed in which an alternate data reduction
subroutine was used simultaneously with the usual one to
eliminate taking into account the second, background
burn of the braid. Four data sets of 20 samples each were
atomized, the results of which are presented in Table 12.

The first set compares the integrated absorbance seen

177

Table 12

Significance of Background Correction

 

 

 

Data Set With Back. Cor. Without Back. Cor.
A 5.88 i 0.55 9.3% 6.73 i 0.57 8.4%
B -0.048 0.56
C -0.047 0.54
D -0.003 0.02

 

 

A = 2 pi 1 ppm silver at vertical height 1.5 mm.
B = Baseline at vertical height 1.5 mm.
C = Baseline at vertical height 5.0 mm.

D = Baseline at vertical height 5.0 mm, braid not burned.

178

from samples of 2 pi Of 1 ppm silver. Without background
correction, there is a slight improvement in precision,
but the integrated absorbance is distinctly higher. Al-
though the blackbody intensity is known to be negligible,
it is interesting to note that the difference in integrated
absorbance is Opposite to that expected if the blackbody
were significant. Set B compares baseline data between
the two methods, burning the braid without sample. A
noticeable absorbance without the correction is still
present. It was suspected that perhaps this was due to
the fact that observations were taking place very close
to the braid (vertical height of 1.5 mm), and that during
atomization, the braid was bowing upwards through thermal
expansion into the source beam. Set C repeats this base-
line data, but at a vertical height of 5 mm, sufficient
to avoid the bowing effect. Although the uncorrected
absorbance is reduced, it is still present. Finally in
set D, comparisons are taken with the braid not burned

at all. Here the integrated absorbances finally agree.
It is evident that background in this system consists

not merely of blackbody emission. Possible contributiOns
from such sources as carbon specks lost from the braid

or Schlieren effects from the heated sheath gas will also
contribute to the total integrated absorbance, as well as
to the precision of analysis. These effects would no
doubt be quite difficult to quantify, as there is no

simple means available to the present system by which to

179

detect them separately. Certainly a great deal of irrepro-
ducibility in these effects is to be expected. Worst of
all is the fact that by turning the braid parallel to the
optical axis to cut down on "background intensity",
Schlieren effects and particulate losses are actually
intensified, as they occur along the entire length of

the braid, and would be Observed by the system as such.

L. ALTERNATE METHODS OF DATA REDUCTION

One final test of performance was to investigate
whether or not the precision of analysis could be improved
by a data collection routine which would look for a peak,
and confine the integration of absorbance to it only
instead of the usual total summing of all points taken
during the atomization time. The routine employed was
configured to not begin accumulating absorbances for
integration until at least five consecutive points were
monotonically increasing and above a threshhold absorbance
arbitrarily set at 0.005. Integration would then continue
until five points monotonically decreased or the absorbance
fell below 0.01. Along with this modification, the routine
also contained provisions to present the data in the usual
mode of integrated absorbance, as well as in the form of
what might be called the "integrated nontransmittance”,

namely:

180

 

with the latter expression referring back to the numbered
equations of the previous section. Again, these data
were presented both with and without the peak seeking
algorithm.

The results are shown in Table 13 for the analysis
of three different concentrations of 2 01 samples of silver.
The average peak maximum absorbance is given with each
data set to give a rough idea of the magnitude of the
transient. It can first of all be seen that essentially
no difference in precision is provided by using the peak
seeking algorithm. Regardless of the peak height, the
precision Obtained in either the absorbance or transmittance
mode of data representation is comparable with or without
peak seeking. It is interesting however to compare the
trend in precision between the absorbance and transmit-
tance modes. For the concentrated solution of 1 ppm silver,
the precision of the data set is about 1.58 times worse
calculated as absorbance rather than transmittance. The
intermediate solution value is 1.15, and the dilute solu-
tion gives a ratio of 1.04, or nearly comparable results.
This effect is due again to the nonlinear nature of
logarithms. The lower the percent transmittance, the
greater the change in absorbance per unit change in trans-

mittance. In the case of the concentrated solution then,

 

 

181

 

wa.¢a Hm.o H mm.m mm.w mm.o H vm.m wv.m an.o u mm.mm wane xmmm
mm.ma Hm.o H mm.m wm.m Hm.o H mm.m wv.m «v.0 w mm.mm Hmfiuoz
.mcmnucoz .ucH
wm.va ma.o w mm.o w~.> mm.o H oo.v mm.m mm.o u mm.mm xaco xsmm
mm.va mH.o H «o.a wo.> m~.o H ma.v wm.m mm.o H nv.mm HsEuOZ
mad .UGH
mg Ema «0.0 mm Egg «.0 m< Ema a

 

 

manufluoma< coflumasoamu mo cOmwummEou

ma manna

182

the variations in data points from sample to sample near
the peak maximum make abnormally large contributions to the
imprecision in the total integrated absorbance. With more
dilute solutions, in which the transmittance and absorbance
functions have more similar slopes, the effect is gradually
minimized. It is thus unfortunate that even if integrated
absorbance is used as the readout parameter rather than

the peak height, the peak height still can exert an in-
fluence with the type of imprecision that peak integration

is hopefully intended to overcome.

SCANNING ELECTRON MICROSCOPY

STUDIES OF GRAPHITE BRAIDS

Graphite braids undergo some rather notable changes
in physical appearance with time which depend on how
they are used. In addition, the rather porous nature of
the braid causes deposited samples to soak in during
desolvation: their ultimate fate in undergoing desolva—
tion to form dry salt particles is not well known. For
these reasons, and for the sake of being able simply to
take a closer look at the physical substance of the
atomizer itself, scanning electron microscopic and x-ray
fluorescence micrOprobic investigations were undertaken
of typical graphite braids, both with and without samples.
These studies were performed using instrumentation and
photographic supplies provided by the BASF Corporation
at their analytical research facilities in wyandotte,
Michigan. In this chapter, a discussion of the observa-
tions noted from selected photographs will be presented.

Scanning electron microscopy is an extremely power-
ful technique for the examination of samples requiring
resolutions of as little as one hundred Angstroms or

.186'187 The SEM can cover a very broad range of

so
magnification, from the nearly molecular figure just
quoted up to modest enlargements similar to those ob-

tained with hand held magnifiers. The interpreted

183

184

electron image can be displayed on a phosphor or televi-
sion screen with image contrasts and lighting effects
which greatly imitate the appearance of illumination
by natural light. Perhaps most significant of all is
the fact that the SEM has a tremendous depth of field,
which permits the examination of samples having con-
siderable length along the electron axis with very little
loss of focus. Accompanied by an x-ray microprobe, the
SEM becomes a powerful analytical tool. The x-ray probe
can be configured to scan the entire viewed area of
sample, and yield an averaged readout of elemental composi-
tion, or it can.be narrowed to a tiny, fixed spot only a
few Angstroms wide, for a highly specific observation of
a feature of particular interest. In this mode it can
also be scanned across a specific location of the sample
to produce a readout of elemental analysis versus sample
location. This may be used to typify the chemical con-
tent of sample features. Analysis may be either of all
recognizable elements simultaneously, or of just one
specific element. Signal integration may be performed
over long time periods as well for signal-to-noise ratio
improvement.

Graphite braid is something of an ideal sample for
SEM study. The carbon itself is of too small an atomic
number to be detected by the x-ray microprobe (which
was capable of elements of atomic number as low as sodium),

and is also inherently electrically conducting, which

185

frees it of the accumulation of static charge that can
white out an SEM image. This eliminates the pretreat-
ment step, often needed for nonconducting samples, of
the thermal vacuum deposition of a surface layer of
platinum or gold. During transport to Wyandotte, sample
braids were protected by enclosure in a piece of saran
wrap. For analysis, a short section of braid (about one
quarter inch long) was simply cut from the desired region
of analysis, and affixed to a small aluminum stage with
a conducting adhesive. The stage was then inserted into
the sample port, and could be fully examined a few
minutes later after establishment of the requisite high
vacuum in the instrument.

Figure 39 illustrates virgin graphite braid at 50
diameters. This particular piece was taken straight off
the supply roll and examined as such. Even physical
handling of the braid was minimized. The braided nature
of the material is well illustrated, as is the composi-
tion of braid bundles by individual graphite strands.
The strands are quite small; each bundle may well con-
tain a thousand or more of them. Also visible are small,
randomly scattered specks in the braid, which may be
the principal source of the initial impurities that are
visibly ejected as smoke from a fresh braid when it is
first heated. This idea is supported by the braid Shown
in Figure 40, which is a braid that has been subjected

to one initial heating to remove the impurity smoke,

 
  
 
      
 

5:;

Q"

\

\\\\\. \

\ \N ‘
‘N‘

J

'\

\ .
\ .
\‘ ’

\

    

\y_ \\“\.n1 Od
\\\\\
\ \\\\\\\

\

\

’x

.
\v

.
‘\
‘\

\\\‘

Figure 39. Virgin graphite braid at 50 diameters.

 

Figure 40. Degassed virgin braid at 50 diameters.

187

but which has been otherwise unused. The frequency of
occurence of the small particles seen in Figure 39 is
quite markedly reduced.

Figure 41 shows virgin braid again, but at a magni-
fication of 500 diameters. The open, porous nature of
graphite braid is rather strikingly revealed in this
image. Individual fibers are well collimated and uni-
form, but they are by no means very densely packed. It
is Obvious how easily deposited samples may soak into the
fiber spaces during desolvation. Equally apparent is the
fact that the potential total surface area of contact
between atomizer and sample is quite large; this excellent
contact undoubtedly plays a major role in promoting two
of graphite braid's attractive features as an atomization
device, namely its general freedom from memory effects,
and its ability to perform complete atomizations on the
subsecond time scale. Also notable from this photograph
is the high consistency of interfiber dimensions. The
diameter of any given fiber seems to be quite representa-
tivecfifthem all, and individual fibers are apparently
able to extend unbroken for considerable distances through
the braid, if not damaged. Figure 42 shows one indivi-
dual virgin fiber at 20,000 diameters. Again, the par-
ticulate matter visible on the fiber surface is suspected
Of being impurity deposits, as other photographs of
fibers from braids that have been burned do not show them.

From measurements of this photograph, the diameter of

 

.\

Figure 41. Virgin braid at 500 diameters.

 

Figure 42. Individual fiber of braid at 20,000 diameters.

189

a fiber is about ten um. Interestingly, there appear
to be crystalline layers in the fibers which run along
their entire length, as evidenced by the transitions

in image brightness that occur in crossing the width of
the fiber.

As braids are used in spectrometric work, the regions
close to the electrode holders gradually become dull in
appearance, and eventually develop hot spots of unusually
high electrical resistance that ultimately become the sites
of braid burnout. An examination of these dull areas
in the next few figures is quite revealing with respect
to the cause of this appearance.

Figure 43 shows a 20,000 diameter magnification of
one individual fiber from the electrode holder region of
a braid that had been in service sufficiently long to
develop the dull appearance effect. It is plainly seen
that the fiber has become pitted with numerous submicron
size cavities, with perhaps slight evidence that these
cavities tend to form near the boundaries of the crystal-
line layers previously noted. Their presence in the
fibers accounts well for the visual observations listed
above. As the fibers become pitted and scarred, light
reflection from the fiber surface is cut down markedly,
which leads to the dull appearance. In addition, the
cavities, once formed, probably serve as sites of still
further fiber destruction, which increases the fiber's

electrical resistance and deterioration progressively

190

 

Figure 43. Individual fiber of a used braid at 20,000
diameters: electrode holder region.

 

Figure 44. Individual fiber of a used braid at 20,000
diameters, with apparent extensive damage.

191

until the braid ruptures. A particularly battered fiber
is illustrated in Figure 44. Here the pitting and scarring
has progressed extensively, and the fiber is rapidly
losing its mass and integrity. A final illustration is
given in Figure 45 at 10,000 diameters. Here a fiber
which does not appear to be badly damaged by pitting
seems instead to be covered with deposited particulate
matter. This is also reasonable in light of the fact
that some of the dull appearance of a braid can be re-
moved, and the shiny appearance restored, by simply
rubbing a dull braid gently with one's finger. Finally,
in Figure 46, a fiber is illustrated from the central,
sample receiving portion of a braid that is very near to
burnout. The dull appearance is beginning to extend into
the center of the braid by this time, and expectedly,
the fibers there are now beginning to Show the pitted
characteristics as well.

Two more types of braids without samples are shown
in Figures 47 and 48. Figure 47 shows a braid that has
been intentionally damaged by direct exposure to the
atmosphere during heating. The magnification is 500
diameters. In contrast to the orderly arrangement of
fibers seen in Figure 41, the fibers in Figure 47 are
in great disarray, much like a frayed hemp rope, and large
numbers of them have been ruptured and oxidized. Figure
48 shows a braid that has been in service at high tem-

peratures, about 2200-2300° K, at a magnification of 100

192

 

Figure 45. Individual fiber of a used braid at 10,000
diameters, with deposited particulates.

 

Figure 46. Individual fiber of an old braid at 20,000
diameters: central region.

 

193

 

Figure 47. Braid intentionally damaged through atmos-
pheric exposure, 500 diameters.

 

Figure 48. Braid used in high temperature work, 100
diameters.

194

diameters. Braids employed at the highest temperatures
develop a shiny gray appearance in place of the usual
black, and also seem to be perceptibly stiffer when
handled. It was suspected that this was due to a marked
change in fiber quality, such as the fusion of adjacent
fibers together. However, Figure 48 shows that the braid
is not greatly different in appearance from braids used
at more conventional temperatures. Examinations at higher
magnification show a similar lack of recognizable dif-
ferences. Apparently, the alteration observed is one of
the crystallinity of the graphite itself rather than a
change in the surface structure.188
For the observation of desolvated samples on graphite
braids, it was felt necessary to use sample solutions that
were abnormally concentrated in order to assure that
desolvated particles could be reliably found. The general
procedure followed was to deposit a five microliter sample
of a 1000 ppm solution of the element of interest, desol-
vate the sample, and then carefully package the braid
at that point for transportion to wyandotte. When these
samples were subjected to analysis, it was found that
even at such unrealistically high concentrations, authentic
particles of desolvated sample were few and far between,
and at times could not be located at all. This is some-
what surprising, as the volumes and concentrations of
samples used should result in something like five to ten

micrograms of salt crystals being desolvated. Some

195

possible explanations may be offered. First, the SEM can
only observe the outermost fibers of the braid bundles.
Considering that samples soak into the braid, it may

well be that the great bulk of desolvation takes place

in the braid interior, and that the population of salt
particles is far denser within the depths of the fibers.
Second, there is always the possibility of loss of the
sample because of inevitable physical disturbances that
occur when the loaded braid is stored for transport, or
prepared for the microscope. With respect to the first
problem, it might be suggested that an intentional spread-
ing of the braid fibers before analysis to expose some of
the interior regions might reveal an increased population
of salt crystals. However, it is then equally obvious
that any gains which might be realized through this pro-
cedure may be seriously offset by further disturbance of
the braid, and consequent aggravation of the second
mechanism. This is particularly so because unravelling
of braid is not all that easy to do gently.

In Spite of these problems, some authentic salt par-
ticles were located and identified with the x-ray micro-
probe, which was Operated in the focussed spot mode.
Figures 49 and 50 show, respectively, a copper nitrate
and a nickel nitrate particle on braids examined at 10,000
diameters. (This is the magnification of all remaining
pictures.) The copper particle measures about four um

long by three wide, whereas the nickel particle is

196

 

Figure 49. Copper nitrate salt particle on braid, 10,000
diameters.

 

Figure 50. Nickel nitrate salt particle on braid, 10,000
diameters.

197

roughly spherical with a diameter of only 0.8 um. Such
dimensions are rather typical of all particles located

on the braids, which indicates that samples are desolvated
as quite a few microcrystals of salt. This also attests
to the braid's efficiency as an atomizer.

Figure 51 shows a phenomenon observed rather fre-
quently with the braids, namely the presence of glass
beads on the fibers. X-ray analysis shows a principal
elemental composition of potassium, silicon, and calcium,
with traces of iron, chromium, and titanium. Most of
these elements are reported by Union Carbide to be present

183 The glass beads may

as trace impurities in braid.
well be formed from them upon heating and their common
presence may stem from the fact that these braids were
burned for only a few seconds to drive off the usual
smoke before the sample was deposited. It may well be
that if the braids had been used more extensively before
being loaded with samples, many of the glass beads would
have been volatilized away. In testimony to the sensi-
tivity of the x-ray microprobe, it may also be mentioned
that a fair number of particles found on these braids
gave elemental compositions which strongly hinted that the
particles were airborne chalk dust. (Calcium and sulfur
peaks were quite prominent.)

Figure 52 shows a particle about two microns in
diameter that was desolvated from a sample solution 500

ppm each in cadmium and zinc. X-ray micrOprObe analysis

198

 

Figure 51. Glass bead on braid, 10,000 diameters.

 

Figure 52. Combined cadmium and zinc nitrate salt particle
on braid, 10,000 diameters.

199

revealed that this particle contained a mixture of salts
of both elements. That they Should have desolvated to-
gether is perhaps not too unusual, as both elements were
in solution as the nitrates, and both occupy the same
group in the periodic table. In contrast to this are the
particles shown in Figures 53 and 54, which are, respec-
tively, nitrates of silver and COpper. Here the two
were again present together in solution at a level of

500 ppm each, but desolvation seems to have occurred
separately. (Another particle was located in which both
were combined.) Here it may be that a separation occurred
because, although both elements are again from the same
periodic group, the COpper ion has twice the charge of the
silver ion. This may lead to sufficiently different
crystallizing tendencies to promote separate growth.

Most interesting is the observation that the silver
particle appears to have exploded. It is quite likely
that it originally desolvated as a crust of salt which
surrounded a still liquid center. When the enclosed
solution boiled to final dryness, it caused a violent

and widespread rupturing of the outer salt Shell. Such
occurances could constitute a source of imprecision in
electrothermal atomic spectrometry if they are a common
desolvation phenomenon because of a loss of analyte from
the atomizer. The effect could be most severe in devices
such as the more traditional filaments, which do not have

the braid's soaking-in capabilities.

200

 

Figure 53. Exploded silver nitrate salt particle on
braid, 10,000 diameters.

 

Figure 54. Copper nitrate salt particle on braid, 10,000
diameters.

201

In future applications of these studies it might be
of interest to compare the physical appearance and per-
formance of graphite braid to commercially produced graph-
ite devices, or to some of the other types of filament
atomizers such as metal foils and wires. Such devices
should be equally easy to examine, as the electrical con-
ductivity of the medium would still exist, and the lack
of sample soaking would keep all of a desolvated sample
in plain view on tOp of the atomizer for observation.
In the cases of the metallic atomizers, there might be
interesting trends in the progression of oxidation Of
the atomizer surface, with different chemical compositions
at different depths. Different chemical deterioration
with sample variety and sample matrix may occur as well.
Electron microscopy would also permit a firsthand look
at atomizer changes which occur as a result of lifetime
enhancement techniques, such as the use of traces of
methane in the sheath gas flow.

The greatest disappointment in the present work is
the scarcity of salt particles seen even with enormously
concentrated sample solutions. Reasons for this and the
problems likely to be encountered in doing anything about
them have already been touched upon. Perhaps one tech-
nique which might offer some assistance would be to unravel
a braid prior to sample deposition, and then mount and
deposit the sample upon only one bundle of fibers. This

would in effect act like a miniature braid. With a

202

thin device such as this, and fewer fibers to contend
with, it may be possible to locate and observe a dense
population of salt particles, yet still have them be
reasonably representative of the behavior likely to be
experienced within the interior of an atomizer of full

diameter.

AN EVALUATION OF THE RADIATION METHOD

FOR ATOMIZER TEMPERATURE CONTROL

A. INTRODUCTION

One of the most important variables in atomic spec-
troscopy is the regulation of the temperature of the atom-
ization device. In flame methods the atomization tem-
perature is adjusted roughly by selection of the particu-
lar fuel and oxidant combination and finally through
adjustment of the fuel-to-oxidant ratio. The hottest
flame occurs at the stoichiometric mixture. Moreover,
the regulation of temperature in the flame methods is large-
ly self controlling. The flame is a dynamic medium, con-
tinually refreshed by new volumes of fuel and oxidant.

A natural steady state condition is established between
heat gained from further combustion and heat lost from
convective and radiative dispersions. Given a high
quality, reproducible gas flow controller and a well
designed burner head, an extremely stable and well-be-
haved flame can be routinely produced. In electrochemical
methods, the temperature is adjusted by the amount of
power delivered to the atomizer from its supply, which
gives the advantage of nearly continuous electrical
selection of temperature up to the upper limit of the
device in question. But the regulation of this tempera-

ture is a significant problem. The same atomizer must

203

204

be used repeatedly throughout a long lifetime. This can
give rise to temperature drifts as the device ages, par-
ticularly since it must be driven from room temperature

up to atomization temperature in the process of vaporizing
samples. A flame burns steadily at the desired temperature,
whether sample is present or not.

It was previously mentioned that there are (at present)
four possible ways in which to regulate the heating of the
graphite braid. These are the voltage developed across
the braid, the current passing through it, the power
dissipated by it, or the blackbody radiation emitted from
it. The last method, regulation by monitoring of the black-
body emission intensity, is rather different from the
others. Philosophically speaking, this is so because,
of the four techniques, the first three are Observing in
one way or another a physical input to the braid, whereas
the blackbody emission method, hereafter known as radia-
tion programming, functions on the basis of a physical
output from the braid. Radiation programming can, there-
fore, be free from the effects of some system parameters
or errors which can greatly affect regulation performed

with the other three methods.

205

B. QUALITATIVE COMPARISON OF THE TECHNIQUES

The differences between radiation programming and the
other methods are to an extent visible to the user when
graphite braids are heated. A braid for which the elec-
trical power dissipated, hereafter known as power program-
ming, is the means of regulation can be observed to heat
up to its final temperature with a sort of ramp function.
A braid heated under radiation programming seems to flash
up to the final temperature very quickly. This effect
can be seen in the plots Of Figure 55, which are oscillo-
sc0pe photographs of signals from two different observa-
tion points in the braid power supply. The top two trac-
ings show the voltage developed across the supply current
sensing resistor for the two programming methods. In
power programming (on the left), the current starts and
remains at a more or less constant level, being changed
only Slightly as required by the power regulation. In
radiation programming, however, the current begins with
a near square wave rise to a very high value, follows
this with a continued high level in which some fluctua-
tion is observed as the supply recovers from the heavy
drain, and then suddenly settles down to a low, quite
constant level. The explanation is relatively Simple.

In radiation programming, blackbody emission must be
developed from the braid to establish the regulated

heating level. However, when power is first applied,

206

.mcfiEEmuqoum NOBOQ
pom OOHDMAOmH pops: wwaOQmON wammom umsom mo mcmmumouozm wooomOHHflowo

 

m H u wEflu mmwa m m n wEfiu ammzm

 

 

207

the braid is not incandescent, so the supply turns on

very hard in an attempt to establish emission and stays

on hard until emission is seen. AS a consequence, the
heating rate is very rapid compared to the power method.
The bottom traces in Figure 55 Show the signal developed
from the braid-observing phototransistor under the same
conditions. With the power method, the emission, and hence
the braid temperature, rises smoothly but slowly to an
asymptotic final value. In radiation programming, the
temperature rises very quickly to the final value, reaches
it with no appreciable overshoot, and then holds that level

precisely for the remainder of the burn.

C. TEMPERATURE CALIBRATION OF THE GRAPHITE BRAID

Blackbody emission from an Object is described by
the well known Stefan-Boltzmann expression, in which the
total electromagnetic radiation emitted from a source
is proportional to the fourth power of the absolute tem—
perature of that source. As a consequence, radiation
programming of an electrothermal atomizer carries with
it the inherent possibility of accurate temperature cali-
bration of the atomizer and a convenient way to monitor
temperature in the output signal from the phototransistor.
Such a temperature calibration was performed for the
graphite braid.

At first it was proposed that the calibration be

208

done by burning the braid under radiation programming

at a series of reference voltage and output operational
amplifier switch settings and observing the blackbody
emission at each with an optical pyrometer. A computer
program was written which would burn the braid at user
specified levels and times, allow delays for cooling of
the braid between burns and sighting of the pyrometer.
The pyrometrically determined temperatures would be
entered by the user for statistical evaluation. After
some preliminary work with the pyrometer however, it was
decided that a complete pyrometric calibration would not
be practical for two reasons. First, the graphite braid
is not nearly as uniform in temperature as the pyrometer
is intended to expect. The bundled, braided nature of
the fibers gives rise to a fair amount of thermal texture
in the atomizer, which is easily observable through a dark
red filter. As a result, a consistently objective assess-
ment of an average temperature against which to match the
pyrometer filament is extremely difficult to make. The
pyrometer setting may be easily shifted by fifty degrees
and still give a filament color which could conceivably
be taken as a reasonable average of the entire atomizer.
Second, to get a reliable reading with the pyrometer,

it is necessary to repeat the observation five or six
times at each temperature, burning the braid about six
seconds per observation. Under such conditions, an

individual braid will not survive long enough to

209

complete an extensive calibration before burning out or
aging beyond the point of reliability.

To overcome such difficulties, an alternate means
of observing the blackbody was develOped. A graphite
braid was mounted, and one particular temperature was
accurately determined with the pyrometer. The braid
was then turned perpendicular to the Optical axis and
again heated to the same level. The blackbody emission
was recorded by the spectroscopic instrumentation on a
chart recorder with the monochromator set to 650 nm; the
wavelength of the pyrometer. The braid was then burned
at a regular series of additional levels. The intensity
of each burn was recorded on the chart. Simultaneously
with the acquisition of this data, a second chart recorder
was used to record the corresponding phototransistor
output signals.

Interpretation of the data was performed with the
Planck equation for the spectral emission from a blackbody.
According to this equation, the emission E of a blackbody

at a wavelength A and temperature T is given by:

-5
coll

E =
(eczr'IT-1)

 

where e is the emissivity (l for a blackbody) and c1
and c2 are composite fundamental constants. This equa-

tion was solved for E at the braid temperature which

210

was determined with the optical pyrometer. The deflection
Obtained on the chart recorder for this temperature then
served as a reference for all other temperatures examined.
To determine any other temperature, it was merely neces-
sary to multiply the reference emission value by the

ratio of the chart deflections observed for the new and
reference temperatures, and then reverse the Planck equa-
tion with the new emission value to yield the new tempera-
ture. For example, let us suppose that a temperature of
1500°K was determined for a braid with the pyrometer and
that at this temperature, the chart recorder Shows an
arbitrary deflection of 200 mV. To determine the tempera-
ture which causes a 500 mV deflection, one first solves
for the blackbody emission at 1500°K, with a result of

0.4 x 10'"4 wsr-lcm-znm-l. This value is then multiplied by

the ratio of the two deflections, which yields 1.0 x 10-4

2nm-l. Finally, by reversing the Planck equation

war-1cm-
for this emission value, a temperature of 1599°K is Obtain-
ed.

Fortunately, the assumption inherent in this process,
that the braid is a perfect blackbody, does not influence
the results. Strictly speaking, if the emission of a
blackbody at a temperature T is B, then that of graphite
at the same temperature will be eE, with e typically about
0.85 to 0.9 (89). After multiplication by the ratio of the

chart deflections one obtains a new value eE' for the emis-

sion of graphite at the new temperature T'. To correct

211

this value back to that of a blackbody at T', one must
divide by the value of e for graphite. Thus, the emissiv-
ity cancels out, and the braid may be assumed to be a
blackbody for purposes of the calculations.

With this method of observation, the braid needed to
be burned only once at each setting to get an accurate
reading, and the burn itself could be shortened to only
a few seconds. This permitted not only completion of a
calibration run on one braid, but even two complete repeti-
tions of the entire run on the same braid for statistical
evaluation. In addition, the chart deflections Obtained
are somewhat representative of the average temperature
of the braid. The work was done with the monochromator
slit open fully to 2 mm. Even with such a wide slit, the
limited acceptance angle of the monochromator probably
confines the region of the braid viewed to the central
few millimeters. However, as this is the point at which
samples are deposited, it may be argued that this region
is the one for which a temperature calibration has most
meaning.

Shown in Figure 56 is a family of curves giving the
absolute temperature of the braid as a function of the
reference voltage supplied from the DAC for a series of
OA switch settings. The pyrometric reference temperature
was obtained with a 5.00 volt DAC output on OA switch
setting 1, for which the pyrometer gave a temperature

of 1571 t 7°K from six averaged readings. (This particular

212

2.1887—

TEMPERATURE (K)
-3
X18
1 1 a 1
l T I l

 

1.288

b

I. ' : I. : 1 .
8.888 9.888
DAC VOLTAGE (V)

__

Figure 56. Temperature calibration of graphite braid with
respect to reference voltage: low temperature
region.

213

level was selected as being convenient to observe on the
pyrometer.) The data points in Figure 56 are the average
of three complete calibrations, which agree with each other
at all points to within five or ten degrees. Agreement
among different braids is on the order of ten to twenty
degrees, which is only a factor of two or so above the
statistical validity of the pyrometer. Some slight varia-
tion is to be expected among braids, because the exact
orientation of given braids with respect to the photo-
transistor can vary and cause local minor inconsistencies
in the distribution of temperature.

Calibration at temperatures higher than those shown
here was not at first possible, as the phototransistor
becomes saturated by directly observing temperatures
above 2000°K. This causes the braid power supply to go
out of regulation. To calibrate the upper temperature
range, it was necessary to attenuate the light reaching
the phototransistor. This was accomplished by inserting
a small neutral density filter made from several glued
layers of fogged spectrographic glass plate film in front
of the transistor. Figure 57 shows a similar family of
curves obtained with this filter for the high temperature
range. With the filter, it is possible to drive the braid
to the highest temperature attainable, about 2600°K. To
get beyond this point, a different power supply would be
required. At a level of 2600°K the present supply is

dumping current continuously at 22-24 A, which closely

2. 788'“—

TEMPERATURE (K)

1.688

Figure 57.

x18
1

 

214

1 1 l L, I 1 1 1

T I l l l I T 1

8.888 9.888
DAC VOLTAGE (V)

Temperature calibration of graphite braid with

respect to reference voltage: phigh temperature
region.

215

matches the maximum predicted current for a 50 V supply
presented with approximately a 2 0 load. Higher tempera-
tures are probably not practical, however, as the braid
suffers rapid and severe damage at such levels. Indeed,
although the data of Figure 57 is again the average of
three complete runs on the same braid, it was necessary
to correct these runs for the damage suffered at the
highest temperatures by reobserving the reference tempera-
ture level frequently throughout the data collection
process. Because of this damage, the range previously
quoted of ten to twenty degree accuracy for the drawn
loci probably degrades to perhaps twice as much for the
upper temperature end of Figure 57.

As a check on the accuracy of this method Of calibra-
tion, the pyrometer was used to examine some of the cali—
brated temperatures. The results are Shown in Table 14.
Agreement between the two techniques is rather good.

The maximum temperature attainable is dictated by the
limitations of the power supply. The minimum attainable
is the lowest temperature at which the braid has sufficient
incandescence to provide light to the phototransistor.

For the present system, that limit occurs at about 1100°K.
This limit could be extended downward if desired by use
of a photosensor with better spectral response in the
near infrared than the present device.

As previously mentioned, along with the blackbody

emission data, a second chart recorder was simultaneously

216

Table 14. Comparison of Pyrometrically Observed Tempera-
tures with the Present Method.

 

 

 

Present Pyrometric
1345°K 1353 1 6°K
1571 1563 i 8
1608 1595 i 6
1777 1779 i 7
1861 1851 i 12
1923 1919 i 11
1968 1966 t 5

 

 

217

run which observed the phototransistor output for each
temperature level. These readings were highly reproducible,
as they ultimately are compared directly to the reference
voltage at the power supply control OA. Figures 58 and 59
show plots of the phototransistor output versus tempera-
ture for the low and high temperature ranges respectively.
On these plots, of course, the data from all seven (six)
curves overlay each other into one smooth locus. With the
data of Figures 56-59 available, it is possible to select
and monitor braid temperatures reliably if the braid in

question is not excessively aged or otherwise damaged.

D. LONG TERM TEMPERATURE DRIFTS

Aging is inevitable in electrothermal atomizers.
Regardless of the device employed, with constant use it
will eventually deteriorate to the point where its per-
formance is no longer acceptable. The principal variable
to preserve during an atomizer's lifetime is again the
temperature. The radiation programming method, because it
monitors a quantity sensitively dependent on temperature,
is a superior means of sustaining constant temperatures
throughout the useful life of an atomizer.

The upper locus of Figure 60 Shows a plot of the
decay in intensity of the blackbody radiation typically
observed throughout the lifetime of a graphite braid that
is heated repeatedly under radiation programming to about

1500°K. The ordinate is scaled in arbitrary units of

218

.COAmOH musumummEmu 30H "usmuso noumwmsmuu

 

louonm ou uowmmmu nuflz wanna muflnmsnm mo sowusunfiamo wusumnogsma .mm Tasman
A>v bombao mahmmmz<mh0hozm
rooo.m _ _ l _ 800.9

1 _ N _ — Sow-H

l.

11 3

N

.d

m s

1|.AU v.

I.

.n

11 “H

no.1

)

11 “X

 

1100a .N

219

.cofiwwu musumumafiwu so“: ”usmuso Houmfimsmuu

 

louonm on powmmmu nuwz Osman wuflnmmum mo :Owusunwamo onsumnwmfiwe .mm musmfim
A>v hamhoo mosm~mz<mh0hoza
_000.m _ _ _ _ 000.0
_ _ _ J _ 00m.“
1..
1:...
l.
11 3
w
u .1
11.58 V.
l.
.n
11 ”a
Ab_d
)
11 “X

 

1100K. .N

220

. musumummsau 30H

"mswaemumoum

 

umzom pom coaumflpmu Hops: pawns muwnmmum mo hmomo OHSOMHOQEOB .00 Tasman
N - o a x
mszm no mwmznz
000 . m _ _ 000 . 0
F . b _ P _ p b p p p p _ n - P b P — p 0
fl 4 . a . — a a _ 4 a . _ _ a q q q a . . . 008 o
MOHmMHIII II.
.u
M m 11. -. .1
o mma v.
I as
MOHSHII 1 X
x m
MOHMVHIIII 1.. I a
no 1A
xOmvaII 11 ..1
.. N
M0331 - 3
+ N
V6331. .838 1.. m
.- .1
003:1 A
MOmmwdll 11
oceanspmm .
MOnomHII -.

 

L000 .0

221

blackbody intensity at 650 nm, as observed on a chart re-
corder, whereas the abcissa is scaled in number of burns,
with each burn having a duration of three seconds. Along
the right hand side of the plot are some temperatures
appropriate to the decay observed. Throughout the total
lifetime of the braid, the decay in temperature with radia-
tion programming amounts to about 50°, or roughly three
percent. Such a decay can be accounted for by observing
visibly what happens to a braid as it ages at this tem-
perature. With continued heating, damage begins to set in
at the vicinity of the brass electrodes. This leads to

a catastrophic mechanism of failure in which the damaged
areas begin to dissipate abnormally large fractions of the
total electrical power, which causes them to become hotter,
and to experience yet greater damage. Several possible
explanations might be offered for why the Spots develop
specifically at the electrodes. Because of the proximity
to the chimney wall, this is the region of the least ef-
fective flow of sheath gas. In addition, the quality of
the contact between individual fibers may be lessened by
the mashing of the braid in the electrode grip. Finally,
the bending of the braid at the electrodes in order to
secure them may cause strains that rupture fibers under
heating. Whatever the cause, the phototransistor which
regulates the heating observes the braid at about a 45°
angle, whereas the intensity data in Figure 60 were observ-

ed by the monochromator with the field of view confined

222

to the central few millimeters of the braid. Thus, as the
hot spots develop, the phototransistor observes at least
one of them directly, but the monochromator does not.

The transistor is thus unduly influenced by the hot spots,
and steadily cools the braid down.

The second locus shown in Figure 60 gives the anal-
ogous performance of a braid heated under power programming.
The level of heating is that which eventually achieves the
same final temperature as the corresponding level under
radiation programming. The two loci have been normalized
to the same initial value. Before discussing the second
locus, a brief comment may be given on the comparison of
the methods in this manner. It is not possible to compare
the radiation and power methods fairly by simply having
the times of heating in each burn be identical. As the
photo tracings at the start of this chapter illustrated,
radiation programming reaches the final temperature much
faster than power programming. A fair comparison must
account for this difference. When the data for the power
locus of Figure 60 were taken, the burns were increased
to six seconds each as an estimate of the required cor-
rection. At a later time, a more exact comparison was
made in which an electronic integrator was used to sum
the total area of the blackbody emission from single burns
by the two methods. From these experiments it was found

that one burn for three seconds at 1500°K under radiation

223

programming was equivalent to 0.98 burn for six seconds
which approaches 1500 K under power programming. The orig-
inal data for the power locus has thus been adjusted in
Figure 60 to account for this slight correction, by length-
ening the abcissa in the amount of 1.02. The superior per-
formance of the radiation method for preserving constant
temperature is obvious. The power locus decays about

115°K in temperature over a range in which radiation de-
cays only about 150K. This is a direct consequence of
regulating with a temperature sensitive probe.

The same comparison of methods at the significantly
higher temperature of about 1920°K is illustrated in
Figure 61. The lengths of the original burns here were
three seconds for radiation and four seconds for power.
Integration of the blackbody as before yielded an equiva-
lent of one radiation burn for every 1.1 power burns. The
same general effect is seen, with the braid lifetimes
generally reduced becausexof the higher temperature. The
decay of the power locus is, however, even more severe
than the radiation, with a loss of fully 300°K over a range
in which the radiation method loses at most only 30°K.

Such a large loss under power programming could easily
upset analyses through several mechanisms. As the cooling
progresses, the braid might lose its ability to atomize
samples completely, or free them from stubborn matrices.
Also, the atomic plume from a cool braid would be lowered

and broadened with respect to that from a hot one, and

224

.mnspmummfimu Tusuwpoe umcfififisnmoum

 

 

 

uw3om pom soflumwomu Hops: pwmup muanmmum mo amomo musumumaema .Hm Tasman
m- o;
mzmnm no mwmznz
. 000.0
.QQrN _N _ _ _ _ r E 1 p . _ r _ _ _ _ _ . _ r .
— 1 q q — . a 4 — u — J — a — a — . — q a 000 o
.u
.1
V.
as
N”
.8
VA .U
I 0
AU .1
.I
.llN
II.
a.
”N
c.
1.
IA
sodoma casumflomm

 

000.0

225

differences in peak shape have already been shown to have
great effect upon integrated absorbances.

One additional set of such temperature decay compari-
sons was also performed for a high temperature of about
2300°K. Some results of the study are presented in Figure
62. However, the examples shown are not totally represen-
tative. In contrast to Figures 60 and 61, in which a pro-
gressive loss in temperature is seen for both methods,
the observations at the highest temperature were rather
erratic. Under both radiation and power programming, the
fluctuation in temperature throughout the braid lifetime
was rather small, only a few tens of degrees, but it does
not progress in a smooth decay. Instead, a region of decay
might be Observed at first, but later this might well be
followed by a region of temperature gain. Such rather
random alteration between gain and decay would continue
down to the final burnout of the atomizer. This would
generally occur after approximately 100-120 three second
radiation burns or their power equivalent. A clue to an
explanation for this altered behavior might be the change
in appearance of braids with repeated heating at high
levels. Hot spots do not seem to form as readily near the
electrodes. In fact, the braid will almost always burn
out somewhere in the middle section. The lack of the
electrode hot spots may account considerably for the more
nearly constant temperature-time curve under radiation

programming. Perhaps, at high temperatures, the electrodes

.wusumquEmu amen “mswEEmumoum
um3om pom sofiumwpmu “moss Osman ouflsmmum mo mmowp wusumnwmfime .00 Tasman

0Hx

226

 

 

N.u
mzmnm no mwmznz
000 N 000.0
r _ _ .1 l r _ . . . . u . _ r .
1.....1 1... _... mass
.8
I .1
v.
as
“X
1 8
VA .u
I 0
AU .A
sovzm 11 1 .I
.N
MCQOHNI .T II.
.3
3
0.03.31... 0038 I
l.
mogmml I A
0102.21
sooomml.
sesflmm ..I coflmSmm

000.0

227

function to conduct potentially damaging heat away from

the braid ends.
E. SHEATH GAS FLOW AND CONTACT RESISTANCE EFFECTS

Two additional variables to which radiation programming
of temperature should be immune are the rate of sheath gas
flow and the possible presence of atomizer-electrode con-
tact resistance. Comparisons of both radiation and power
programming were performed for these variables. The ef-
fects of the sheath gas flow upon temperature were largely
negligible. Theoretically, under power programming, the
slower the flow of sheath gas, the less effective its
contribution to the removal of heat from the atomizer,
and thus the hotter the atomizer itself. It was found,
however, that in varying the sheath flow from one to five
liters per minute, no appreciable effect upon the braid
emission intensity could be Observed. Instead, the suc-
cessive burns of the braid merely continued to reflect
the usual slow decay of temperature with time Similar to
Figures 60 and 61, regardless of what the sheath flow did.
The same behavior was of course observed for radiation
programming. One curious and unexpected phenomenon was
observed though. If the flow rate of the sheath gas were
allowed to fall below about 1.25 liters per minute for either
method of programming, the regulation of braid temperature

seemed to be lost. In place of the usual smoothly rising

228

curves of emission intensity with time, the emission from
the braid fluctuated wildly and randomly under power
programming, and similarly although less severely under
radiation programming. Furthermore, after restoration

of more traditional flow rates, the emission intensity,

and hence the braid temperature, was seen to have jumped
significantly higher under either method of regulation.

The explanation was revealed by a physical inspection of
the braid. After being heated even only once or twice at
the very low sheath gas flow rate, the braid became notice-
ably thinner and more worn in appearance, evidence of
atmospheric damage. (Exposure of the braid directly to

the atmosphere without any sheath will cause its destruc-
tion with only a few burns at even low temperatures.)
Apparently with excessively low sheath flow rates, the
feeble movement of sheath gas can be very easily upset by
the convective currents produced by the hot braid. AS

a result, atmospheric oxygen from the top of the chimney
can be caught in these currents and swept down to the braid
itself. The fact that the temperature of the damaged braid
is higher after restoration of normal sheath gas flow can
be explained in light of this damage. In power programming,
the braid will have lost some of its original mass, and
obviously goes to higher temperatures if still controlled
to dissipate the same amount of power. In radiation pro-
gramming, the regulating phototransistor still demands the

same amount of light flux to establish the proper output

229

signal. Thus, a smaller braid of decreased size, and
consequent decreased apparent light-emitting surface area,
must go to higher temperatures to supply the required light.
This illustrates well the fact that even radiation program-
ming is not the ultimate temperature regulating technique
for electrothermal atomic spectroscopy. Even though black-
body intensity depends only upon temperature, any real
system which uses radiation programming will still have
Optical and physical variables associated with it which
can affect both the means and amount of Observation of
atomizer emission.

The presence of contact resistance between the atom-
izer and electrodes was simulated by introducing a 0.27 0
resistor in series with the braid. The results were as
expected. With radiation programming, the phototransistor
simply calls for more power from the supply to preserve
the required level of emission intensity. With power pro-
gramming, a large reduction in temperature is observed as
the resistor dissipates a significant fraction of the
allotted power. The practical meaning of these observations
was brought to light with some experiments on how the tem-
perature of the braid varies with the technique employed
to mount braids in the electrodes. Comparisons were
taken of the emission from four mountings. The first was
normal. The second was somewhat loose, in the sense that
the braid was slightly long and curved upward in the middle.

A third was tightly stretched between the electrodes, and

230

the fourth was normal, but with the grip of one electrode
poorly secured. For both regulation methods, the first
three types of mounts gave very comparable emission levels.
Some slight differences were seen, but they did not follow
any regular trend, and amounted to temperature changes of
only a few degrees. This is reassuring in the sense that
it is not essential to have an exceptionally rigorous or
reproducible technique to follow when mounting braids.

If reasonable care is taken, any temperature differences
caused by variations in Operator mounting technique are
quite negligible. The last variant mentioned above (a
poorly secured holder) did, however, show an adverse effect
on results. Essentially, it created a contact resistance,
which under power programming caused a severe drop in
temperature. Even though radiation programming acts to
correct this problem, it is to be avoided, because the
offending electrode is subjected to severe overheating

as it dissipates the excess power.

F. PERFORMANCE OF METHODS DURING ATOMIZATIONS

After investigating the physical parameters of per-
formance for the two programming methods, they were com-
pared chemically as well by performing atomizations of
selected elements at different temperatures. For each
element in question, a recommended analysis wavelength

was selected with the monochromator, and the photomultiplier

231

and amplifier gains were adjusted to yield a full-scale,
100%T level of nearly ten volts at the A/D converter, with
the appropriate hollow cathode lamp Operated at its maximum
recommended current. The solutions analyzed were prepared
from deionized water stocks of the pure salts and were
intentionally selected to be of significant concentration so
as to Obtain strong, relatively noise-free signals. Five
elements were investigated. They were cadmium, lead, sil-
ver, copper and nickel, listed here in order of increasing
temperature required for atomization. The various system
parameters for each of these elements are given in Table
15. For each element, a number of equivalent atomizations
was done at each of a series of temperatures under radia-
tion programming. The normal output of peak absorbance,
time of peak absorbance, and integrated absorbance was
logged by the line printer. A second set of data was then
taken in which five atomizations were done at each tempera-
ture, and the point by point absorbance readings along

each transient were stored on floppy disc. At a later
time, another computer routine read and averaged these

five stored peaks into one composite, representative peak,
and pnfixmed the averaged readings on the line printer.

An identical procedure was followed as well for
atomizations performed under power programming. In each
case, the heating level chosen for a group of power pro-
grammed atomizations was carefully selected so as to yield

ultimately the same final temperature of the braid as the

232

.mE om umEau Oman umamaamad “muscaE\uouaa m "30am mow numwnm .Efio.m 1 amoauum>
.EEo.o 1 amazoNauom "aamo mo cOauamom “Ed oom "suma3 uaam .a9 0 ”Oman mamfism “mucmumsoo

.mnsu ecaan umaom
n L

 

omvN

ommm
a\> oa .>oos mama aco.~m~ sooomm ~.ao.az soon az

maam
mama
mmma m m
«\> OH >omm mama ecs.¢~m somsna . 02.50 200a so

moma
mmma
mova
¢\> moa >omn mama Esa.mmm xoomma 0204 Emma ma
omna
ooma
ovoa m m
«\> oa >osm mam Eam.mmm soomma A 02.00 soda no
omna
ooma
ovoa m
¢\> N.oa >omn man Eca.mmm xoomma aOOO Edam mo

 

CHOU mad uao> 2m Baum Sumcmam>wz .QEOB xflhamz .OCOU unmfiwam

 

 

.maaEEmumoum umsom pom coaumammm mo cowaumoeoo may now muwumEmumm amucmfisuumca .ma manna

233

corresponding data under radiation programming.

Shown in Figure 63 is a plot of absorbance versus time
peak shapes for cadmium samples atomized under radiation
programming at four temperatures. (The data for the
three highest temperatures are essentially superimposable.)
All of the peaks are well shaped Gaussians, which rise to
a peak absorbance about 0.14 second after the onset of
atomization, and return to baseline in 0.50 second. In
contrast to this are the corresponding plots done under
power programming illustrated in Figure 64. Here the
four peaks are well separated. The times required for
atomization are significantly longer, even for the highest
temperatures. Furthermore, the peak absorbances Obtained
are much lower than those observed with radiation program-
ming. These facts are well brought out by the superim-
posed radiation programmed peak repeated from Figure 63
with the dotted line.

Figures 65-72 illustrate the analogous data obtained
for the other four elements. Consistently, the radiation
method of programming gives rise to atomizations that are
accomplished in less time and with higher peak absorbances
than the power method. In fact, at the highest radiation
programmed temperature for all elements studied, the tran-
sients assume a rather good Gaussian shape. With all
elements available for comparison, some additional conclu-
sions may be drawn from details of the plots.

The cadmium and lead data are extremely similar to

234

 

 

 

 

 

 

8. 388"
1* 1440,1600,l760 0K
1280 0K
LU
U
z:
1<
a1 1.
m:
c:
(D
a:
<
8.888 1
8.888 8.588
TIME (8)

Figure 63. Cadmium transients under radiation programming.

0.1201“

ABSORBANCE

8..8EB8

Figure 64.

 

 

235

1 Radiation

l
l
l
I 1760 0K
I
I
1

fl 1
1
1600 OK

 

1440 oK 1280 oK

 

 

 

{dfllhms , 1 %
1..5E30
T114E '(S)

Cadmium transients under power programming.

236

0.500”.

(All temps.)

 

ABSORBANCE

 

 

 

 

0.000

 

0.000 ' 0.500
1TIME: (5)

Figure 65. Lead transients under radiation programming.

237

 

 

 

 

 

0.200‘"
w
‘t
* 1760 OK
Lu
(J
z
< £
m sh ‘*
S " 1600 0K
(D
m
< r
P
.. O
1280 0K
0.000 ~A». . v} ' o
0.000 5.000

TIME (3)

Figure 66. Lead transients under power programming.

238

 

0..8E30"
__ 1695 0K
1535 0K
n: 1405 0K
L.)
z:
<
(D ._..
a: .
O 4
U)
m
<
0.000

 

 

 

TIME (5)

Figure 67. Silver transients under radiation programming.

239

0.300T'

1695 0K

ABSORBANCE

 

1280 0K

 

 

0.000

 

 

TIME (8)

Figure 68. Silver transients under power programming.

240

 

 

 

 

 

 

 

 

 

0.800T'
” 2115 OK
'4!—
1975 0K
m P
U
z:
<
a: 0..
m:
c:
0‘)
a:
<
1865 OK
- 1775 0K
0.000 nA I
2.000
TIME (S)

Figure 69. Copper transients under radiation programming.

241

 

 

0. 400‘“-
2115 0K
.1;
u: ‘ >
U
z
<
m ._.-
a:
0
(.0 1975 0K
m
<
0. 000

 

 

 

TIME (5)

Figure 70. Copper transients under power programming.

242

0. S000"

ABSORBANCE

 

 

 

0.000

 

 

3.000

TIME (3)

Figure 71. Nickel transients under radiation programming.

0. see——

ABSORBANCE

0.000

Figure 72.

 

243

 

 

 

 

 

 

2450 0K
+ 2330 0K
7 - . 1 1| __ .
l f '
0.000 3.500
TIME (8)
Nickel transients under power programming.

244

each other, as is expected from their similar volatility.
It can now be seen that the plots for these elements over-
lay each other under radiation programming because even
the lowest temperature investigated for them is signifi-
cantly in excess of the minimum required to atomize ef-
fectively. The Gaussian peak shape merely reflects the
kinetic process of getting all of the sample out of the
braid. In fact, the minimum attainable temperature with
radiation programming is probably sufficient to atomize
cadmium and lead. In the case of silver it is possible
to see the effect of temperature in the radiation mode.
The peaks begin low, broad, and asymmetric, and gradually
lose these qualities as they become Gaussian. The power
mode peaks for silver do not show the asymmetry that radia-
tion does, which reflects the slower rise time of tempera-
ture under power programming. These effects are even more
pronounced for copper. In fact, the copper peaks under
radiation programming resemble somewhat the appearance of
a Gaussian peak affected by a low pass filter of varying
time constants. Finally, the nickel results begin to de-
emphasize the differences between radiation and power
programming. The radiation method is still at an advantage,
but because the power supply for the braid is now approach-
ing the upper limit of its capabilities, the advantage of
radiation programming is beginning to level off.

One consequence of the use of radiation programming

is seen in Figure 73, namely, the effect of readout time

245

 

 

 

 

 

 

 

 

  
    

 

 

0.500'F
10 ms
30 ms
u: ‘—
L)
2
'<
m
a:
O
(.0
m
< “‘1'-
0.000 ---« ~. ._
0.000 1.000
TIME (8)

Figure 73. Distortion of transients under radiation program-
ing by readout time constants.

246

constants upon the data. From the peak shapes for lead
illustrated, a small but noticeable improvement in the true
peak shape still occurs in going from an amplifier rise
time of 30 ms, as was used for Figures 63-72, to one of
only 10 ms. This would seem to indicate that use of radia-
tion programming almost necessitates an all electronic
readout; mechanical systems, even with fast response
chart recorders, might not be quick enough to present
properly the fast transients obtained for volatile elements
on a radiation programmed instrument.

An area of possible future interest with these studies
is the opportunity of "dematrixing" a complex sample.
With the rapid heating and precise temperature control
afforded by radiation programming, a multistep atomization
sequence might often be beneficial in providing sequential
volatilization of several components in a single sample.
Multielement analysis is not presently possible at all
with the current system of instrumentation; however, a
crude simulation of this concept was attempted. Three
solutions were prepared in deionized water: 5 ppm cadmium,
1 ppm c0pper, and a mixture of both 5 ppm cadmium and 1 ppm
copper. The cadmium solution was analyzed under conditions
appropriate to a good, radiation—programmed cadmium atomiza-
tion, and the mixture was analyzed in the same manner.
The nonatomized copper was driven off by forced heating
between samples. Next, the copper solution was analyzed

appropriately, and then the mixture, with both cadmium

247

and copper driven off by the single burn. Table 16 gives
a comparison of the integrated absorbances observed. Al-
though the match is not perfect, the basic concept of se-
quential volatilization seems workable. The cadmium in
the mixture is easily driven off without interference

from copper, and with no change in parameters from those
used for pure cadmium solution. A similar disregard of
cadmium by the c0pper atomizations is observed. Admittedly
these elements differ greatly in volatility, and had
roughly comparable concentrations in the mixture. However,
in certain favorable cases, it may be possible to put this
technique to use in suppressing some of the interelement

effects often seen in real samples.

G. SUMMARY AND CONCLUSIONS

To summarize, perhaps the most significant conclusion
to draw from this comparison is that the maintenance of
strict and reliable control over the atomization tempera-
ture in electrothermal atomic spectroscopy is a concern
of paramount importance. Compared to electrical means
of atomizer temperature control such as the power dissi-
pated, the method of radiation programming can offer
significant advantages in achieving this goal. This
method is far superior in preserving constant temperature
throughout the useful life of an atomizer, and can actually

extend an atomizer's life, because the rapid rise of

248

Table 16. Analysis of the Cadmium-COpper Mixture.

 

 

 

Solution Analyte Integrated Absorbance
Sppm Cd Cd 8.06 i 0.20
lppm Cu Cu 22.94 i 1.50
Mixture Cd 8.00 t 0.19

H-

Mixture Cu 24.97 0.75

 

 

249

temperature to the final level permits significant reduc-
tions in the total atomization time. Atomic transients
come off the atomizer with greater peak height, which
makes them easier to detect above baseline. In addition,
the ability to calibrate atomizer temperature, and monitor
it as well, would be of great.importance in thermodynamic
or kinetic studies of atomization mechanisms. At the same
time, the use of radiation programming obligates the user
to make companion improvements in his or her atomic spec-
troscopic system. These consist principally of a fast
response readout that can accurately represent the quick
transients that radiation programming produces, as well as
an atomizer power supply with sufficient voltage and cur-
rent reserve to take full advantage of the benefits provided
by the rapid, near step function atomizer heating. Also,
although the monitoring of temperature is excellent and
free from interference by several variables, radiation
programming is still not the ultimate means of regulation,
as it is subject to optical variations in the observation
of the atomizer.

Perhaps the best means of programming would be the
direct sensing of temperature in the atomizer by a thermo-
couple. Such a facility might be rather difficult to
implement with the graphite braid, not only because of
its small size, but because most attempts in this writer's
experience to undo or modify the braided bundles (as would

be necessary to install a thermocouple) usually damages

250

the braid beyond use. In a more massive and rigid device
such as a furnace, however, a thermocouple implant might
be very beneficial indeed, and offer an excellent means

of holding a constant atomizer temperature even beyond the

precision with which radiation programming functions.

ATOMIC CONCENTRATION PROFILES

OVER GRAPHITE BRAIDS

A. INTRODUCTION

Although the full array of computer-controlled instru-
mentation has played essential parts in the atomic spec-
trometric experiments that have been discussed thus far,
the benefits and powers of the broad computer control
are best appreciated and most fully exercised in the ex-
amination of atomic concentration profiles. To get some
understanding of why this is true, consider briefly what
is involved in obtaining concentration profile data with
both flame and filament electrothermal instruments.

For concentration studies in a flame, the flame
would be ignited, and sample would be aspirated continu-
ously from a suitable reservoir. By analogy to the present
instrumentation, the burner would then be systematically
moved with respect to the optical axis as data is gathered.
Because of the continuous sample aspiration, the distribu-
tion of free atoms in the flame would be in the steady
state. This means that data could be acquired as fast
as the burner is translated, or perhaps even continuously
as it is translated. A high quality log amplifier could
be used to present the readout in an absorbance mode,

and extensive advantage could be taken of phase locked

251

252

detection and signal averaging techniques for improved
signal-to-noise ratios. A thorough examination in two
dimensions could be concluded within minutes.

In a filament electrothermal system, the situation
is somewhat altered. The sample is not continuously
atomized, but is rather deposited on the atomizer as a
discrete aliquot. The atomization cell is then displaced
to the proper position of analysis, and the regular se-
quence of heating intervals must be performed. The
analytical signal will be a transient, and must be treated
as such although the point by point data acquisition pro-
cess does provide valuable information about each sample,
such as time relationships of the signal shape and signal
peak, or the ability to integrate the total signal derived
for each sample. Finally, the cell must be returned to
its initial position so that it may receive the next sample,
after allowing a few seconds for cooling of the atomizer.
All of this work results in one solitary data point for
one single position in space. Several more samples must
undoubtedly be repeated at the same position to get a
reasonable statistical picture of their validity, and
then the entire process must be repeated for the next
atomizer cell position. This continues until the full
region of space has been examined, with the entire pro-
cedure consuming some hours of time, and with no change
in instrumental or sample parameters save the position

of the cell.

253

Such is the case for the present system. Typically,
it takes about 75 seconds for the instrumentation to ex-
ecute an entire cycle of sampling, atomizing, and data
processing for one sample. Four samples are usually
done at each cell location. A typical scan would be to
examine atomic populations at intervals of 3 mm from 0 to
24 mm vertical height above the braid. This is a total
of thirty six samples, and requires about 45 minutes to
perform. Additional such vertical columns at other fixed
horizontal locations would also take 45 minutes each.
Hours of (H1 line computer time can easily be consumed,
and yet the sample is constant throughout. Changing a
parameter to see what effect it might have upon results
implies further hours of repetitive work. During all of
this time, the atomizer is aging and undergoing subtle
but progressive changes. Were it not for computer control
and the use of automated instruments, these studies would
be so tedious and prone to error that it would be nearly
impractical to do them. With automated computer control,
a fixed though lengthy task can be defined, and the com-
puter and instruments can see to its execution without
the need for operator intervention. Such is the manner
in which these studies have been done.

To do concentration profile work, a constant sample
of the analyte of interest is repetitively delivered in
a fixed amount to the atomizer. The concentration of the

analyte is intentionally selected to give transients of

254

significant peak absorbance, in order to assure that they
may be easily detected above baseline noise as far away
from the atomizer as possible. Radiation programming is
used for all the work, and the other conditions of atomiza-
tion such as heating times and levels are usually adjusted
to produce the analyte transient as a quick, sharp signal,
as was seen in the previous chapter. This also enhances
the detectability of signals, and extends the lifetime

of an atomizer as much as possible. Table 17 lists most
of the common instrumental parameters for the various
elements involved in these studies.

As data points are acquired, the records of cell
positions, integrated absorbances, peak absorbances, and
times of the peak absorbances are simultaneously printed
by the line printer and stored on a floppy disc. At a
later time, a separate computer routine is used to pick
up the magnetically stored data, average the results of
each cell location, and print out these averages as well

as their standard deviations.

B. VERTICAL APPEARANCE OF CONCENTRATION PROFILES

The shapes of concentration profiles can be intui-
tively predicted. A plot of integrated absorbance versus
vertical height above the atomizer should resemble some
sort of decay curve. The concentration of free atoms will

be greatest at the atomizer, and progressively less above

25:

 

 

 

 

 

Table 17. Common Instrumental Parameters for Atomic
Concentration Profile Work.
Element Conc. Matrix Temp. Wavlnth. HCDT
Cd 5 ppm, Cl , 1440°K 326.1 nm 10 mA
100 ppb N0;
503'
Pba 1 ppm N0; 1440°K 283.3 nmb 8 mA
Ag 1 ppm NOS 1695°K 328.1 nm 15 mA
Cu 1 ppm NO} 2115°K 324.7 nm 15 mA
Zn 100 ppb N0; 1440°K 213.8 nmb 15 mA
Mg 1 ppm N0; 2115°K 285.2 nmb 15 mA
Hga 10 ppm NO} l440°K 253.6 nmb 15 mA
T1 1 ppm N03 1440°K 276.8 nmb 10 mA
Constants:
Sample Size: 2 ul Amplifier Gain: 108 V/AC
Slit Width: 200 um Amp Rise Time: 30 ms
Sheath Flow: 3 l/min

aSolution 0.5% in H 0

b

Solar blind PM

C1o7 V/A for Mg

2

256

it as diffusive, convective, and chemical mechanisms all
combine to diminish the available atomic population. A
plot versus horizontal displacement will be akin to a sym-
metrical distribution curve, with the greatest atomic
concentration directly over the atomizer, and lesser con-
centrations to either side. The higher the vertical
height for such a horizontal examination, the lower and
broader the concentration curve should become. Generally,
this is what was observed in the present work.

Shown in Figures 74, 75, and 76 are some concentration
profiles with vertical height for a series of elements.
The concentration axis is in integrated absorbance, but
the readings have all been normalized so that all plots
begin with an arbitrary value of 10.0 at grazing incidence,
to allow for intercomparisons on a fair basis. The various
elements display a wide variation in free atom lifetimes.
The best ability to survive in the free atomic state is
shown by zinc, which retains fully 50% of its original
population as far as almost one inch away from the atom-
izer. One inch above the atomizer is quite far removed
from the atomizer heat, particularly at the low tempera-
ture sufficient to atomize zinc efficiently. The tem-
perature of gases at that point is probably very nearly
back to room temperature again. The worst survival is
shown by thallium, which is lost completely by a height of
12 mm.

An interesting effect is seen for those elements

257

.wmma can .Edwfipmo

 

  

.u0>aflm mom unmflws Hmowuum> cums moafimoum coflumuucmocoo omeHmEuoz .vh musmflm
T. o;
Azzv hIonI 4<u~kmw>
ammw$_.N _. h: _ _ _ _l __wmw®..®
1 _ _ 4 _ _ some
I
1! “N
l.
q:
[I .U
H
. 11 VA W“
t. .1
n. nu
[I .V
T.au
II a:
nu
”a
.8
II .V
N.
no
It .3
um>aflm

 

 

omwa .fi

258

.Eswaawnu can .ausonms

 

.ocfiu now uzmflmn Hmowuum> sues mwaflmouo cowumupcmocoo omnwamsuoz .mn musmflm
T o;
Azxv kronI 4<u~k¢m>
. ooo.®
.08.. N _ _ _ _ _ F P .
_ q _ _ _ _ _ Goo o
Esflaamce
I
II. N
I.
.3
1! .u
8
VA “V
annoumz II T. M
a. nu
11. . v.
Y:8
II .5
0
”H
.u
11 ”V
“N
.J
11 .3
ocfls
I

 

000."

259

.55ammcmms can

 

Hoodoo How unmflwn Hmofiuum> nufl3 mmawmoum coflumuucmocoo omuflamfiuoz .mn musmwm
7. 3x
Azzv hoHwI 4<onhmw>
. ooo.o
_o®¢ N_ _ _ _ _ p .
7 _ H — — 808 Q
I
II N
I.
Edwmmcmmz .d
1... 9
U
V
I.” m
0 0
11 .V
1.8
11 .5
0
U
ad
II V
”N
no
Hoodoo 1| .1
goes.”

 

260

which require the highest levels of heating, namely silver,
copper, and magnesium. In each of these cases, the decay
curve begins with a region of convex rather than concave
curvature, with an inflection point occurring a few
millimeters above the atomizer. This may possibly be

an illustration of the benefits provided by hot sheath
gas in preserving free atoms. With the atomizer heated
strongly to drive off the analyte, the sheath gas im-
mediately above its surface is also hotter than usual,
and may help to retard chemical condensation mechanisms
which begin to scavenge free atoms almost immediately for
the elements that atomize well with less heat.

The rates at which the various elements' pOpulations
decay do not correlate well with any single physical
prOperty which is likely to explain them. If plots are
made, for example, of estimates of the rate of decay of
the atomic populations against such likely thermodynamic
quantities as the heat of vaporization of the element, or
the heat of formation of the common elemental oxide, there
is no clearly defined locus. The plots are instead quite
scattered and random. This is unfortunate, because if
such a correlation could be established, it would provide
an excellent theoretical basis from which to design and
interpret further experimentation. At the same time,
however, it is perhaps not unexpected that no such cor-
relation exists. Thermodynamic or kinetic data depend

critically upon temperature, and any experiments which

261

hope to receive a rigorous thermodynamic or kinetic ex-
planation must either have a constant temperature or be
able to predict and account for a changing temperature

in an orderly fashion. In filament atomic spectrometry,
however, the desolvated samples are subjected to tempera-
ture rises of hundreds of degrees as they are atomized,
after which the free atomic vapor experiences an equally
rapid temperature drop of comparable amount as it is
carried away for analysis. There is little chance for
even an approximation to equilibrium. The decay of a
population is thus most likely a rather nondescript
combination of both physical and chemical mechanisms
which do not readily admit to a systematic interpretation,
and which probably shift in relative importance from
element to element, or even within the course of events

173'174 Not surprisingly, studies

for a single sample.
done to date on thermodynamic or kinetic processes in
electrothermal atomizers are most often performed with
furnace devices, in which the sample continues to be
exposed to high temperatures for some time after atomiza-

tion.190

C. HORIZONTAL APPEARANCE OF

CONCENTRATION PROFILES

The horizontal distribution of atomic populations

for three typical elements is shown in Figures 77 and 78

262

um ucmEmomammHo HmucoNHuon cufl3 moaflmoum COAumuucmocoo pwuwamauoz

®®®.w

DU

.1»-

.EE 0 mo usmfloz Havauum> m

Azzv Hzmzwu<4omHo 4<HZONHCOI

9a

U0

.4.—

GD

we

50

'11“

.mn whoops

01X

BONVBHOSSV 031V8931NI

 

LI®®® .fi

263

080 .

w

_
~1-

Azzv Pzwzwu<4mmHo 4<PZONHmOI

.EE 0 mo unofimn HMOfluum> m
up useEmomHQmHo HmucoNfiuon spas mwaflmoum :oflumnucmocoo owufiameuoz .mh wusmflm

 

q

_ 1%

/
/

-H-

U0

 

9m

DU

®®®.ma
r C
. WOQQ 8
IL]
I
11 N
I.-
3
1.. 9
H
X V
1.. I l
3
0 n.
1... .V
I8
I] S
0
H
8
1.. V
N
3
11 3

 

1r®®® .«

263

.68 o no usmflmn Hmofluum> m

 

 

um ucmEmomHomHo HmucoNfluon suwB mwaflmoum :oHumuucwocoo pmNHHmEuoz .mn musmfim
Azzv Hzmzwu<4¢mHo 4<PZONHmOI
®®®.m ®®®.w1
T . _ . _ . _ . _ r F . .
. . 5 . 1 . _ . meme 0
L

I

11. N

I.

3

.11 .5

1. H

V.

ll VA .1

I 3

0 0

me 11 . V

II I8

11 .5

0

“H

9. co po . so 11 8

V

. . _N

.3

1! .3

 

Ir®®® .fi

264

for vertical heights of 0 and 6 mm. The elements selected,
cadmium, silver, and copper, cover a reasonably represen-
tative range of atomizer temperatures. The distance over
which the atomic populations are scattered is not very
great. (The braid itself is about two millimeters in
diameter.) A significant amount of the sample is present
a short distance away from the atomizer, however, even
at grazing incidence. This is undoubtedly a direct conse-
quence of the fact that samples soak into the braid when
they are deposited, and distribute themselves more or less
throughout the entire braid cross section. At the time
of atomization, portions of the sample will be located at
the sides of the braid, or perhaps even the bottom, and
will thus find that exit from the sides or bottom is
most probable. Such exit would be further assisted by
convective forces created by the hot atomizer itself.
The width of the plots increases steadily in the order
cadmium, silver, copper. This coincides directly with
the temperatures required for atomization, and indicates,
reasonably so, that the violence and strength with which
samples are atomized depends on how strongly the atomizer
is heated.

At 6 mm height, the plots retain their same general
shape. The precision of the data points is, however,
noticably worse, particularly along the sides of the

curves. These regions represent the points at which

265

the streamer of warm sheath gas which is carrying the
sample is coming into conflict with cooler gases that are
rising through the remaining bulk of the atomization cell.
This could lead to some local turbulence in the sheath
flow. In addition, if some wavering in the sheath flow
patterns were present, individual samples examined at
these cell locations might by chance happen to intersect
the optical axis well in some instances, and poorly in
others. These effects would combine to cause the impre-
cision seen.

At still greater heights, these trends simply con-
tinue. Eventually, the magnitude of the imprecision
becomes so comparable to the size of the signal itself
that the plots lose much of their meaning. Throughout the
entire 24 mm of examinable height however, the horizontal
distributions of free atoms stay remarkably compact.

The width of the plots does not change much at all from
the initial width created at grazing incidence. This

can be seen somewhat in the illustration of Figure 79,
which shows the distribution of silver at three heights
without the clutter of the error bars. This lack of
spreading into the wings is probably due somewhat to the
fact that the transit time for free atoms from grazing
incidence to full cell height is usually only a few tens
of milliseconds. Diffusion simply does not have much time
to occur. Nevertheless, it is also an additional testi-

mony to the quality of the sheath gas flow. A turbulent

266

.munmflmn HMOfluum> msowum> um H0>Hflm

 

How usmEoomHome HmucoNHuo: nuw3 mwaflmoum coauwuucmocoo omuwameuoz .mh musmwm
Azzv hzmzwo<4mmuo 4<hZONHCOI
.ooo.m _ _ _ u _ ooo.wa
_ 1 _ 1 _ I 3 1 _ 1 . 1 080.9
1.... N
I.
3
11 9
8
x V
. 1-.. .....
0 0
11 .V
I8
11 .5
0
U
8
.1] V
N
3
as... 1.. 3
SEQ \ [1.
a: - _ . 1-25..

 

267

flow would certainly widen the free atom distribution as
the gas formed spirals and vortices. A final curiosity
observed for all elements, and just starting to be visible
in Figure 79, is the invariable shifting of the atomic
cloud towards regions of negative displacement. By the
time an atomized sample reaches the full 24 mm height,

the peak concentration is usually found in the range of

-l to -2 mm horizontally. It is conceivable that this
effect could be imaginary, and caused by a slight lack of
parallel alignment in the two faces of the chimney through
which the Optical beam passes. However, if a laser is
targeted on a distant surface, and the chimney is then
placed in the beam path, no appreciable displacement of
the target spot is seen at all. This shifting effect is
therefore real, and perhaps results from the influence
exerted on the sheath gas flow pattern by the prevailing
laboratory air currents in the region of the atomization
cell.

Another interesting insight into the nature of hori-
zontal concentration profiles is provided by an examination
of their behavior on braids that have been turned perpen-
dicular to the optical axis. Shown in Figures 80, 81,
and 82 are plots of the three test elements at respective
vertical heights of 0, 6, and 12 mm.

Although the profiles are very similar in general shape
to those observed with a normal, parallel braid, they are

somewhat wider because of the ease of soaking of the

268

.pflmun umasowwsmmnwm m can 58 o no uzmwws H00fluum>

 

m an ucoamomammfio Hansenwuon nufi3 moawmoum cowumuucwocoo omuflamauoz .om onsmwm
Azzv kzwzmo<3dmuo 3<P20Numox
®®®.m ®®®.m1
r p — _ — _ — n h p — h F b P p I
fl q — . — . % q — q — . 4 14 — q 808 o
11 N
I.
3
11 9
8
v
1. m ..
3
0 0
so 9. po 11 .v
I8
.1 S
0
9. 6o . 8
\ 1. m
/ N
3
11 3
so /// \\ 4
, 11®®o..

 

269

.pflmun HMHDOHpchHmQ 0 ppm as o no unmfiws Hwowuum>

 

0 pm useEmomadep HmuconHon saws mwawmoum coaumuucmocoo wmufiameuoz .Hm musmwm
.zzv hzwzwofzmuo ._.;ZONHmoz
000.0 000.01
r r. _ . 0 . p _ _ P _ . 1P1 _ _ p
— a u . — . — 4 d 4 — A — q — . QQOOQ
I
11 “N
I.
.1
11 .U
, H
V.
1 m a
.A 0 0
\ Ir.
\ . V.
I8
11 .5
0
.5
00 1 .U
1 ”V
5 N
mm 00 o co .3
so 11 .3

 

11000 ..

270

.owmun uma30flocmmumm m was 880... m0 ”imam... amounuumtw

 

m on ucoemomammflp HmucoNfluon nufl3 moaflmoum cofluwuucwocoo cmuwflmeuoz .Nm onsmflm
Azzv hzwzmo<4mmmo 4<hZONH¢OI

~®®®.m_ L .1 . . _ _®®®.m1
_ + . . _ x _ w . x . x q . . x 000.0

L1
I
11 “N
I.
.3
11 .5
U
1. x u
I 3
0 8
1.. . V
71.8
I] 3
so on. so .00..
.8
.11. V
”N
so 1.. n

so m...
1 In

-'1.AA--m-_WWIY. 1F®mw® .m

 

271

sample along the length of the braid. The increase in
width is not excessive, however, which indicates that
desolvation to dry salt takes place quickly, before the
sample can spread too far. In contrast to the parallel
braid data, there is a small tendency for the profiles

to broaden with height, (more easily visible if the plots
can be overlaid) and there is not as severe a loss in
precision in the profile wings. It is felt that this is
due to the lack of severe thermal gradients along the
width of these profiles. Horizontal spreading above
parallel braids must cross a large thermal and convective
gradient, whereas comparable spreading above a perpendicu-
lar braid more or less fans out into regions of comparable
thermal and convective influence. Similar reasoning may
well account for the lack of a clear relationship between
profile width and atomizer temperature, as was seen with

the parallel braid data.

D. EFFECT OF TEMPERATURE UPON SILVER

INTEGRATED ABSORBANCE

It was illustrated in a previous chapter that the
shape of an atomic transient has great influence upon the
integrated absorbance, because of the nonlinearity intro-
duced through the point by point computation of logarithms.
The general effect observed was that for two otherwise

identical transients, the transient which is narrow and

272

tall will yield a larger integrated absorbance than the

one which is broad and short. In addition, as the atomizer
temperature is increased, it is observed that atomic
transients become narrower and taller. One might logically
expect, therefore, that the integrated absorbance of
transients will increase as temperature increases.

In practice, some elements do not show much change
in integrated absorbance with temperature at all. Cadmium
and lead fall in this category. But this is not unexpect-
ed, as cadmium and lead are quite volatile, and under
radiation programming produce transients which have es-
sentially the same shape no matter what temperature is
used. A most significant change is seen for silver,
however, and it is opposite to the trend predicted above.
This is illustrated by the data of Table 18, which shows
how the integrated absorbance of silver transients falls
more than 50% over the range of temperature covered, in
spite of the fact that the transient shape is becoming
narrow and tall.

There is a twofold explanation for the effect. Part
of the reason is the general observation, previously dis-
cussed, that as the atomizer temperature increases, the
width of the free atomic plume increases as well. This
is seen for silver in the plots of Figure 83, which show
integrated absorbance as a function of horizontal dis-
placement for the family of temperatures. Only one side

of the cell is plotted, and the error bars are left out

273

Table 18. Change in Integrated Absorbance of Silver

Samples with Temperature.

 

J

 

Vertical Height 1280°K 1405°K 1565°K 1695°K
0 mm 35.05 29.69 15.72 12.91
33 mm 28.11 25.33 15.85 13.43
6 mm 21.86 20.09 13.71 11.96
9 mm 17.85 16.98 11.86 10.50
12 mm 14.99 13.34 10.13 9.02
15 mm 12.92 10.70 7.25 6.95

 

 

274

.mousuwnmmEmu kuwsouw msoflum> um H0>HHm
you useEmomHQmflo HmucoNfiuoz nufl3 moafimoua cofluwuucmocoo vmuwamfiuoz .mm musoflm

Azzv szzmu<gmmuo 4<hZONHmOI

 

7000 .H _ _ _ _ 000 . r ..
. a . . . 000.0
11
I
11 N
I.
3
1.. 9
8
V
1- m 1
3
0 8
LI
. V
8
so 83 1.. .1 s
0
so mom. 8
8
11 V
N
11 3
so was. 1..
11000 ..

 

275

for clarity. Clearly, as the temperature is increased,
there is a net reduction in the absolute quantity of
absorbing atoms within the observation window of the
spectrometer at any given time. This will tend to lower
integrated absorbance with atomizer temperature.

The second contributor to the phenomenon is the ef-
fect that the hot atomizer has upon the flow rate of the
sheath gas. Table 19 lists the time elapsed from the
start of atomization until the occurrence of the peak
maximum absorbance for silver transients examined at
vertical heights of 0 and 12 mm. The differences between
these sets of times represent the time required for the
transient to cover the 12 mm distance up the chimney.

It can be seen that this time drops steadily, with the
exception of the highest temperature. This effect almost
certainly results from the expansion of sheath gas by

the presence of thelmfi:atomizer. The laminar flow of the
sheath may further assist this process by permitting the
hot gas to "stream" rapidly up from the atomizer without
much hindrance. As a result, as atomizer temperature
increases, the cloud of free atoms is carried through

the observation window at any point at a growing velocity.
For a fixed rate of data acquisition, this leads to
steadily less and less observation of the sample as it
passes, and again, a lowered integrated absorbance.

Why the transit time should rise again for the final

temperature is not conclusively known. Perhaps at that

276

Table 19. Transit Times for Silver Samples at Different
Temperatures.

 

 

 

Peak A Time: Peak A Time
0 mm 12 mm Difference
1280°K 222 ms 265 ms 43 ms
1405°K 240 ms 277 ms 37 ms
1565°K 270 ms 300 ms 30 ms

1965°K 272 ms 312 ms 40 ms

 

 

277

temperature some slight turbulence is being set up which
interrupts a smooth gas flow. The integrated absorbance
still goes down in spite of this fact. Apparently the
continued widening of the atomic cloud is enough to keep
the net atomic concentration declining. The peak shape
effect is no longer of much importance at that tempera-
ture, as is apparent from the very slight difference in
shapes between the 1565 and 1695°K plots of Figure 83.

Some brief tests were performed with cadmium as the
analyte to investigate the effect thatthe flow rate of the
sheath gas in general has upon results. Vertical profiles
of concentration were performed for cadmium, with rotometer
flow rate settings of 2, 3, and 4 liters per minute of
argon. The results are given in Table 20. There is no
appreciable drop in the transit time for the peak absor-
bance maxima with increase in flow rate, but a small drOp
in the integrated absorbance. The effect is very minor,
however. In light of these results, it was concluded that,
given a reasonable bulk flow rate of sheath gas to carry
the sample, any alterations in the instrumental performance
which arise from perturbations upon the sheath gas flow
can be ascribed to the presence of the hot atomizer.

The basic flow of the sheath itself is a parameter of

only slight influence.

278

Table 20. Effect of Sheath Flow Rate upon Cadmium Spatial

 

 

 

Data.
Integrated A Time of Peak A

2 l/min:
0 mm 12.68 i .18 113 ms
12 mm 4.89 i .22 165 ms
24 mm 3.05 t .60 201 ms

3 l/min:
0 mm 12.56 t .10 114 ms
12 mm 4.55 1 .37 168 ms
24 mm 2.85 i .38 203 ms

4 l/min:
0 mm 12.27 t .14 112 ms
12 mm 4.48 i .16 166 ms

24 mm 2.82 .16 201 ms

1+

 

 

279

E. A DISCUSSION OF SOME PROBLEMS

OF INTERPRETATION

At this point it might be well to pause briefly and
consider how complex explanations of atomic spectrometric
phenomena can be. The susceptibility of results to changes
in the flow of sheath gas caused by the atomizer has just
been illustrated. With so many different atomizers in
use, each of which as often as not has its own particular
sheath gas delivery system, it can be appreciated how ef-
‘fects observed for one given system are not of importance
in another. If the sheath gas flow in this instrument
were turbulent, for example, the changes in integrated
absorbance with temperature for silver might not be serious,
because the sheath would more effectively resist the altera-
tions to which it is actually susceptible at present.
Contradictory behavior can be seen even within the limits
of the present system. Copper, for example, also shows
an effect of decreased integrated absorbance with tempera-
ture, though not as severe as that of silver. Yet in the
case of copper, the transit times (between 0 and 12 mm)
for the atomic cloud rise steadily with temperature from
a minimum of 10 to a maximum of 25 ms. Thus both this
effect and the peak shape effect should resist the trend
in integrated absorbance observed. If plots are compared
of the horizontal displacement of integrated absorbance

for these temperatures, the changes in width vary somewhat

280

from temperature to temperature, but in not nearly as
regular a manner as was seen for silver. This raises the
possibility of chemical differences in the mechanism of
atomization which influence the population of free atoms.
Another illustration of the complexities involved in
this work is given by the vertical height decay behavior
shown by zinc, cadmium, and mercury in Figures 74 and 75.
The elements are listed in the order of their longevity.
Seemingly contradictory to this is the fact that they are
also listed in the order of decreasing reactivity for other
chemical systems, such as the standard electrochemical
potentials. Some insight into the problem is possibly
suggested by an examination of the times at which the peak
absorbance occurs at a given vertical height. At grazing
incidence, for example, the times of peak absorbance are
as follows: zinc, 160 ms into atomization, cadmium, 120
ms, and mercury, 70 ms. These distinct separations in time
might explain the longevity trend. Zinc does not leave
the atomizer until well into the burn. Decomposition
products from its salt or any other trace sources have
already been expelled into the sheath. When the zinc
finally comes off, the atomizer is somewhat hotter, and
the hot gas helps to preserve the free zinc atoms. Mercury,
on the other hand, is very volatile, and leaves the atomizer
extremely early. There is not much benefit to be gained
from hot sheath gas, and the mercury atoms are expelled

along with their matrix. For mercury, this may be

281

unusually severe because of the necessity of using peroxide
to prevent mercury losses prior to atomization. Interest-
ingly enough, once the mercury atoms are a few millimeters
away from the atomizer, the severe loss in population is
strongly arrested, and further decay occurs at a rate no
more severe than that of zinc. Cadmium, being intermediate
in atomization time, is also intermediate in these ef-
fects. This is all speculative, but it does illustrate

how very complex the activities within an atomic spectro-
metric cell are, and how many explanations can often arise

to decipher unusual behavior.

F. EFFECT OF OXYGEN UPON FREE ATOM POPULATIONS

In an attempt to learn somewhat more about the ability
of free atoms to survive as they leave the hot atomizer,
experiments were done in which oxygen was intentionally
included in the sheath gas. The normal flow of 3 liters
per minute of argon was retained, but it was supplemented
by a flow of 0.2 liters per minute of oxygen. The ef-
fect upon the graphite braid was devastating. After only
five or six burns in such an atmosphere, the braid was
severely damaged, and an equal amount of additional burns
would bring the braid close to burnout, in spite of the
use of very low temperatures. Because of this fact, only
elements of high volatility were examined, and even for

them it was possible to get only one single data point

282

per cell location. Shown in Figures 84, 85, and 86 are
the results obtained for cadmium, zinc, and lead. The
lack of repetitive data makes it tenuous to draw conclu-
sions, but it is apparent that the presence of the oxygen
does not automatically mean the total and rapid scaveng-
ing of all free atoms produced. The plots for the oxygen
data all begin at a disadvantage with respect to the normal
data. This is particularly true for zinc, which perhaps
reflects zinc's greater tendency to form an oxide than
cadmium or lead. Most of the damage is done near the
braid, however. Once away from the braid into cooler
regions of the sheath, the data taken in oxygen do not
decay rapidly. In all three plots, in fact, the loci

tend to converge. It seems that if a troublesome matrix
or scavenger is present in a sample, it will do its damage
in the vicinity of the atomizer, where the generally
higher prevailing temperatures will better promote the
energetics and kinetics of such reactions as the formation
of oxides. At greater distances from the atomizer, the
prevailing temperatures around a sample are cool, and

free atom-scavenging mechanisms may lack the necessary
energy to remain functioning. Some additional support

for this concept is provided by the distinct lessening

in slope that many of the vertical decay curves of Fig-

ures 74-76 show in the region of 4-6 mm vertical height.

283

 

.Eswspmo mo oawmoud :oflumuucwocoo Hmoauuo> map com: cmohxo mo uoommm .vm musmwm
. .. o . x
Asz hIGHwI 4<0Hhmw>
_oo~.. _ t _ soo.o
_ 1 _ 1 _ 1 _ 1 cos . o
111:
I
11 N
I.
3
1 9
1 a
V
/ [.1 m an“
/ a 0
11 . V
c0903 3.33 I N
11 O
.80wa usonufis . 8
// 11 V
N
3
11 3

 

.000.“

284

.ocflu mo mammoum cofiumuucmocoo Hmofiuum> 0:» sons cmm>xo mo pommmm .mm madman

.- 0.x
Atty kzoawx 4<Q~hmu>
Fsow; _ _ .ooo.s
. _ . _ . 03.0

q
41-
.4?—
_

0IX

C‘.
(D
0"
>1
X
0
.C:
o
'H
3
//
' I
BONVGHOSGV 031V8931NI

 

comwxo usonufls

 

W000..

285

 

.ommH mo mammouo cofiumnucoocoo H00fiuuo> mnu com: comaxo mo uomuwm .00 musmflm
.. 5.x
Azzv #1000: 3<onhmw>
_ . . . 1 . . 000.0

cmmmxo nuH3

cmmmxo usonufl3

1

 

0IX

BONVBHOSBV OBIVUSEINI

000..

286

G. EFFECTS OF MATRICES UPON CONCENTRATION PROFILES

A final field of investigation of concentration pro-
files was that of some matrix effects. A large percentage
of current analytical problems involves the performing of
analyses for substances that are familiar in themselves,
but which are contained in an unusual or stubborn matrix.
Atomic spectrometry is no exception, and the ability to
do a spatial analysis of an electrothermal atomization
cell can provide a means for gaining further knowledge
about matrix effects and using this knowledge to eliminate
or minimize such effects.

Some preliminary work was done using cadmium as the
analyte and solutions of the common mineral acids as the
matrix. The concept was to begin these studies with a
matrix that should be relatively gentle and easily vola-
tilized. Solutions of 5 ppm cadmium made from stock
cadmium chloride, nitrate, and sulfate were supplemented
by additional amounts of standardized hydrochloric, nitric,
and sulfuric acids to produce solutions of different con-
centrations of the respective anion, up to a maximum of
250 ppm. It was found that the presence of these acids,
even up to the highest concentration level, had very
little effect upon the integrated absorbance of cadmium
solutions compared to that obtained from solutions which
lacked the acids. An example of some of this data is

shown in Table 21, for additions of sulfuric acid. With

287

Table 21. Effect of Sulfate Ion from Sulfuric Acid Upon
Integrated Absorbance of Cadmium

 

 

 

Height 0 ppm 803- 50 ppm 50:. 250 ppm 502
0 mm 10.00 t .78 9.23 t .24 8.84 i .21

3 mm 6 48 t 20 6.44 t 14 6.37 i .04

6 mm 5 03 i 15 4.90 i 13 4.94 i .06

9 mm 4 13 t .13 4.09 t .14 4.10 i .43
12 mm 3.45 i .11 3.72 i .24 3.43 i .17
15 mm 3.00 i .17 3.01 i .32 3.05 i .19
18 mm 2.58 i .37 2.59 i .43 2.72 i .31
21 mm 2.25 i .32 2.35 i .39 2.28 i .44
24 mm 2.09 i .45 1.91 i .41 2.25 i .23

 

 

289

the possible exception of grazing incidence, there is
virtually no significant difference between the data
sets. The same can be said for the data of hydrochloric
and nitric acids as well.

This is not very unusual. All three acids are rather
volatile. Hydrochloric acid would be completely evapor-
ated at desolvation temperatures. Nitric acid would be
effectively decomposed. Only sulfuric acid could survive
desolvation, but it too would be removed from the braid
before cadmium and, being a liquid, would not effectively
remove cadmium salts prematurely because of entrainment.

As a next step in worsening the matrix, the studies
on cadmium with added chloride, nitrate, and sulfate were
repeated, but with the ammonium salts as the anion sources.
The reasoning was that ammonium salts should retain the
matrix as a solid, and survive desolvation conditions
better than the mineral acids, but still present a matrix
which can be quite readily decomposed by heat.

Shown in Figure 87 are the results for cadmium in
the presence of ammonium chloride. Here a consistent
matrix effect is seen, beginning with even small concen-
trations of the added salt. Ammonium nitrate and sulfate
did not show any such consistent trends, but left the
cadmium absorbance largely unaffected. The effects of
increased temperature upon these matrices can explain the
behavior seen. Ammonium nitrate decomposes at low tem-

peratures into nitrous oxide and water, and ammonium

289

.Esfiaomo mo mammoum coflumnucmocoo

 

Hmowunm> onu coma mpfiuoHno EsacoEEm we saw wpwuoano poops mo uowmmm .hm madman
.. 0.x
Azzv hIonI 4<0Hhmm>
788.7 N. _ _ p _ I _®®® O .
_ d . s1 3 q . . 0mw®_.0

 

 

33NV880$8V 031V8931NI

000..

290

sulfate decomposes at 235°C. Ammonium chloride, however,
remains intact at higher temperatures, though it does
sublime at 340°C.191 It is possible that in the present
instance, the sulfate and nitrate are destroyed prior to
atomization of cadmium, but that the chloride, or at least
some of it, survives long enough to be atomized with the
cadmium where it can exert a depopulating influence. It
may be noted that the slopes of the decay curves are quite
similar, with or without the matrix present. This indi-
cates that the matrix does its damage directly at the site
of atomization and ceases to have any influence upon the
population once it is in the sheath gas.

As a last step in the progression of difficulty of
the matrices, the anion studies of cadmium were repeated
a third time, with potassium salts as the source of the
matrix. In these cases, the survival of the matrix through
the desolvation step is assured. All three salts are suf-
ficiently involatile to be present during cadmium atomiza-
tion. Indeed, to assure that excess matrix salt did not
build up on the braid, atomization was performed at 1760°K
rather than the usual l440°K.

As was the case with the ammonium salts, only potas-
sium chloride showed a significant matrix effect. Both
the nitrate and sulfate did not yield any clear cut trends.
In the case of potassium chloride, for the first time, the
matrix contributed significantly to the total absorbance

observed. Indeed, at a concentration of 250 ppm added Cl',

291

the ejection of the matrix from the braid was faintly,

but perceptibly visible as a brief white smoke. Some
results are listed in Table 22. A small, but noticable
effect is seen upon addition of only a slight amount of
KCl. At high KCl levels, however, there seems to be an
increase in the absorbance again, even after correcting

the raw data for the absorbance by the matrix itself.

An exact reason for this apparent reversal of events is

not known, but it does illustrate one of the instrumental
limitations of the present system. The atomic absorption
spectrometer of these studies is only a single beam instru-
ment. To correct for nonspecific absorbance by matrices,
it is necessary to run the matrix alone as a separate sample
at some later time, and apply corrections. Aside from

the excessive waste of time involved in running such

sample matrices without analyte, the possibility remains
that this is not truly an exact means for achieving the
necessary correction.

To conclude the cadmium studies, an investigation was
done in a 50 ppm Cl' matrix of KCl, in which a small
aperture of diameter only 0.5 mm was used at the mono-
chromator in place of the traditional 3.0 mm aperture.

The decay of a cadmium population in 0.5 mm increments
away from the braid, both with and without added KCl, is
shown in Figure 88. In spite of the greatly improved
vertical resolution, it can still be seen that the cadmium

populations in the presence of the matrix are reduced

292

Table 22. Effect of Chloride Ion from Potassium Chloride
upon Integrated Absorbance of Cadmium.

 

 

 

Height 0 ppm Cl- 10 ppm Cl- 250 ppm Cl-
0 mm 10.00 i .32 8.62 i .10 9.09 i .52

3 mm 6.49 i .16 6.19 i .33 6.66 i .08

6 mm 4.94 i .11 4.52 i .11 5.35 i .13

9 mm 4.02 i .10 3.86 i .10 4.63 i .20
12 mm 3.47 i .25 2.97 i .56 3.65 i .35
15 mm 3.05 i .30 2.59 i .34 3.16 i .18
18 mm 2.73 i .21 2.23 i .15 2.86 i .43
21 mm 2.53 i .19 2.22 i .20 2.03 i .17
24 mm 2.25 i .43 1.89 i .29 1.85 i .62

 

 

293

.Esflanmo mo mammoum coflumnucmocoo Honduum> cowu5H0mmu

 

 

:0“: man coma mcfluoHno Esfimmmuom mm cow mcfiuoHSU omvvm mo pummmm .00 musmfih
nitv hIQHwI 4<0Hhmm>

Goo.m .

—1 n w w .— 1F _ b b! h b owl’s o O

— a _ u a a _ u q . OQQ 0

I
.1] N
I.
3
11. a.
H
v
1.. m m...
0 .u
111
.v
I8
[.1 s
0
H
8
V
N
3
3

 

 

 

000..

294

right from the edge of the braid; the matrix curve is
separate from the nonmatrix curve at grazing incidence
instead of meeting it. Again, the matrix effect would
appear to be one of interference in the primary atomiza-
tion process. The exact cause can arise from several
mechanisms. During atomization, potassium chloride may
provide a ready population of effective free cadmium
scavengers. chemical affinities during desolvation may
cause the cadmium to be well occluded in KCl crystals,
so that it is ejected in the KCl salt smoke, or other-
wise prevented from atomizing effectively. Once away
from the braid, the population decays both with and with-
out a matrix are quite similar, as was observed for am-
monium chloride.

If other elements are investigated for matrix effects,
it is found that each individual element seems to have
its own particular behavior. A few examples may be given
here in order to illustrate some of the other tendencies
and effects that may play a role in matrix interferences.

One of the most simple is that the severity of a
matrix effect may depend on the absolute quantities of
sample and matrix. This is illustrated by some observa-
tions with zinc and cadmium. Most of the work done in
these studies with cadmium was at the triplet resonance
line at 326.1 nm. The reason for this was that the sensi-
tivity of cadmium at the singlet line, 228.8 nm, is so

high that severe and vexing problems exist with automated

295

delivery of trace cadmium levels. At the triplet line,
sensitivity is sufficiently lowered so that more concen-
trated solutions could be used. Zinc is equally as sensi-
tive at its singlet line as cadmium, but a similar pro-
cedure was not possible for zinc because its triplet line
at 307.6 nm is a very weak absorber. Zinc studies thus
had to be done at the singlet line, 213.8 nm, with zinc
concentrations of only 100 ppb, and matrix concentrations
of up to 5 ppm. This gives the same maximum 50:1 matrix
to analyte ratio as was used for 5 ppm cadmium and up to
250 ppm matrix. Zinc matrices, however, were not all

that serious. Table 23 shows the integrated absorbances
obtained for zinc solutions, and for zinc solutions made
with ammonium and potassium chloride matrices, such as
were effective with cadmium. The differences are rather
slight. For a comparison with this data, cadmium solu-
tions were also analyzed at 100 ppb levels with 5 ppm

of matrix, using the cadmium singlet reasonance at 228.8
nm. The results are also shown in the same table, and
they too show an insensitivity to the presence of the
matrix. This indicates that the analyte to matrix ratio

is not always of sole importance in predicting the pos-
sible existence of a matrix effect. The absolute quantity
of material can play a role as well. With more salt on
the atomizer, it is possible that analyte may be entrapped
within matrix salt crystals of large size, and prevented

from atomizing as effectively as it would with less total

296

Table 23. Matrix Effects of Zinc and Cadmium at Lowered
Total Salt Concentrations.

 

 

 

No Matrix 5 ppm Cl- (KCl) 5 ppm Cl-(NH4cl)
Zinc 12.09 .14 12.21 .29 12.36 .20
cadmium 13.72 .28 13.94 .11 13.74 .48

 

Zinc, cadmium concentration = 100 ppb.

material present.

Results for copper are of interest to illustrate how
the effectiveness of a given substance as a matrix inter-
ferent can change for different elements and atomization
conditions. Shown in Table 24 are some results observed
for matrices of various salts. In contrast to cadmium,
copper is not severely interfered with by potassium
chloride, but is significantly affected by potassium
nitrate and sulfate. Figure 89 shows the vertical profile
of the sulfate interference. The initial loss in inte-
grated absorbance is still seen, as with cadmium, but so
too is an increase in the decay rate of free copper popu-
lation in the presence of the matrix. These results give
further clues to the importance of analyte and matrix
volatilities. The ammonium salts have little effect on
capper, which reinforces the concept that these matrices
are decomposed and ejected prior to the atomization of

the analyte. Similar reasoning can be proposed for the

297

Table 24. Effects of Matrices upon Integrated Absorbance
for COpper and Magnesium.

 

 

 

Matrix Copper Magnesium
None 15.40 t .29 21.03 t .53
KF 10.27 ii.33 0.42 i .05
KCl 13.79 i .20 7.65 i .84
KBr 12.96 t .24 13.28 i 1.3
KI 10.49 i .38 8.78 i .65

9.71 t .15

9.56 i .24
KNO3 8.01 1 .29 ' 1.96 2 .11
1(st4 6.59 i .15 2.09 i .15
NH4NO3 15.73 i .09 20.81 i .67
NH4C1 15.67 i .60 18.91 i 1.0
(NH4)ZSO4 15.47 i .26 19.57 i 2.3

 

 

Copper, Magnesium concentration = 1 ppm.

Matrix concentration = 250 ppm in the anion.

298

88:.
P

N

.Hmmmoo mo mawuoum coflumuu
ucmocoo Hmofluum> mcu com: mummasm Edfimmmuoa mm cOH ovumasm mo uommmm .mm musmwm

T o;
AZZV hxwuwI 4<u~hmu>

.888.8

 

1

888.8

    
   

umvom and omm

BONVBHOSSV 031V8931NI

888 o [I

 

299

matrix effectiveness of potassium nitrate and sulfate.
Copper requires much more heat to atomize than cadmium
does. In the cases of c0pper and the potassium nitrate
or sulfate matrices, the analyte and matrix may be vola-
tilized essentially simultaneously, which provides the
opportunity for the matrix to exert a depOpulating in-
fluence. For cadmium and these matrices, the two may
volatilize sufficiently separate in time to avoid much
interaction. Unfortunately, neither potassium nitrate

nor sulfate show any appreciable nonspecific absorption
at the copper wavelength and the 250 ppm concentration
level, so it is not possible to compare directly the times
of maximum absorbance for samples of copper with and
without these matrices, to see if the simultaneous volatil-
ity concept is true.

Another interesting effect is shown by the matrix of
potassium iodide upon copper. In the presence of iodide,
copper II ion undergoes the reaction:

Cu2+ + 31' + CuI+ + 12
Notice how the integrated absorbance averages for the
copper-potassium iodide mixture decrease with time. No
visible precipitation was present in solution at this
trace copper level, and even if precipitate were present,
it should still be delivered from the autosampler to the

atomizer. The drOp in integrated absorbance suggests

300

that insoluble CuI is adsorbed or otherwise retained upon
the surfaces of the autosampler syringe parts, thereby
reducing the total copper concentration delivered to the
atomizer in each successive sample. Similar processes
might also explain matrix effects that are observed in,
for example, supernatants drawn off from suspensions

such as effluent samples. Many of the investigations
performed for lead in the present work were extremely
confusing until the use of hydrogen peroxide in lead solu-
tions was adopted to prevent the take up of lead by glass
surfaces such as the autosampler syringe barrel.

A final illustration which emphasizes the contribu-
tion of chemical affinities to matrix effects is afforded
by a comparison of effects between magnesium and copper.
Both elements were studied at the same levels of analyte
and matrix concentration, and under the same conditions
of atomization as well. Yet magnesium suffers far worse
in the presence of the matrices than does copper. Only
for the ammonium salts does magnesium seem to be relatively
matrix immune. This is again consistent with the volatil-
ity concepts of these salts. The increased susceptibility
of magnesium to other matrices in general can be ascribed
to its greater inherent chemical reactivity than copper.

The concentration dependence of the severe fluoride
matrix with magnesium is illustrated in Figure 90, and
a high resolution examination close to the braid is pre-

sented in Figure 91, with the assistance of the small

301

.Edflmmcmme mo mawmoum cowumuucwosoo Hmowuuo>
on» coma cofl mofluosHm poops mo vommmw on» no oocmocmmmp cofiumuuomocou .om musmwm

 

 

 

Ha 8fix
Azxv hIOHmI 4<o~bmm>
P88N.~ 888.8
_ . I oooé
LT ..N.
I.
3
11 9
w
X
.1].- I m...—
0 0
II .V
Ema om 1%
11 0
8
8
11 v
N
3
1... 3

 

zoos;

302

.Esfimmcmmfi mo wawmoum coflumuucmocoo

 

 

 

Havauum> GOwUSHommH no“: mzu coma Goa mowuosam poops mo pommmm .Hm musmwm
szv pIonI 4<unhmu>
.888.m _ _ 888.8
7 . . . _ . . 1 _ w 4 1 888.8
I
E. N
l.
3
!. 9
8
v
-..ma
0 0
E. .V
.l8
5. S
0
8
L a
. V
. N
w .J
I 3
F888.”

 

 

303

aperture. The former figure is drawn with all data ori-
gins normalized to 10.00. The raw data shows some vari-
ability in the absorbance at grazing incidence regardless
of fluoride content. It is a problem at times to get
reproducible atomizations at given locations in space
with elements that decay as rapidly as magnesium does.

The flexibility of the braid, and the expansions it under-
goes in heating give rise to small perturbations in the
exact location of grazing incidence with time that can
cause noticable alterations in the integrated absorbances
observed for such elements. The latter figure shows the
familiar pattern that, in the presence of the matrix, the
free atom population is reduced from the outset. The
noticable rise in integrated absorbance, with or without
the matrix, for the first short distance of vertical
travel, is something new, and indicates that for magnesium
at least, mechanisms favorable to free atom production are

still in effect away from the atomizer surface.

SUMMARY AND CONCLUSIONS

The goal of this work was to characterize aspects
of the performance of a filament electrothermal atomizer
through the use of powerful instrumentation and automated
computer control. Insofar as these studies were carried,
this goal has been achieved. Experiments that require
a great deal of time, as well as considerable attention to
the control of a variety of instrumental links have been
executed with thoroughness and consistency.

The two major instruments constructed, the positioner
and autosampler, have many additional potential applica-
tions. The positioner can continue to provide the spatial
control over the instrumentation that concentration pro-
file studies require. But, there is little about the
positioner that ties it specifically to its present
application alone. If viewed as a separate instrument,
it is really a dual channel, digital positioning device.
The two motors can be connected to any sort of mechanical
instrumentation that requires spatial variation under
computer control. The autosampler is more directly tied
to filament atomic spectrometric work, but it should
have a long future of service in that area. Regardless
of the exact mechanical configuration of the device, it
has been the experience in this laboratory that an auto-
sampler can consistently outperform a manual, hand-held

syringe in terms of the precision of sample delivery.

304

305

This holds true for both the size of the sample itself,
and its consistent placement on the atomizer. The present
device is not perfect, and changes such as those outlined
previously should be made. Its continued active use is,
however, of certain benefit to further atomic spectrometric
work.

The largest single shortcoming of the present instru-
mentation is that it is single channel. Additional work
should be accompanied by refinements to correct this
problem. At the least, a simple background corrector
formed from a continuum lamp and beamsplitter could be
coupled with a lamp pulsing circuit and phase-locked
detector to accomplish this goal. The necessary lamp and
optics are already available for such an addition. Alter-
nate techniques are equally possible, such as the wave-

192 or the use of a

length modulation method of O'Haver,
true, multichannel device such as a diode array. The
incorporation of such a device would imply extensive
changes in the data collection electronics and interface.
Microprocessor applications to this problem would be
numerous.

The advantages of radiation programming for atomizer
heating have been well illustrated. The use of a power
supply of higher base voltage would permit further im-
provements in the characteristics of radiation programming.

Investigations should also be done with photosensors which

have improved infrared response to allow a lowering of

306

the regulation threshhold. Comparative studies on other
atomizers would assist in assessing the method for electro-
thermal atomic spectrometric work in general. With some

of the more massive and rigid atomizers, the opportunity

to do direct temperature regulation with a thermocouple
also exists.

The microscopy studies should be carried on with an
emphasis on methods by which to locate the principal salt
crystal population. The scarcity of authentic salt par-
ticles located, even with inordinately concentrated samples,
tends to remove these studies somewhat from truly reflect-
ing atomizer conditions during real analyses. If undoing
of the braid fibers should prove impractical, the examina-
tion of alternate atomizers which do not allow sample soak-
ing could be substituted. The ability to examine a dried
sample directly, and take its x-ray fluorescence spectrum,
would be of great benefit in evaluating hypotheses about
matrix effects.

The spatial examination of a filament atomization
cell is a field of inquiry that is almost without bound.

If time permits, much additional work could be done, not
only for a wide range of matrix effects, but also for
other phenomena which were not addressed in this work,
such as interelement effects, the behavior of the sheath
gas in the presence of the hot atomizer, or the nature
of the hydrogen diffusion flame.

Finally, beyond the extensions of these topics listed

307

above, the graphite braid in general is deserving of further
work to investigate its potentials as a general purpose
atomizer, particularly for elements of low to moderate
volatility. The soon anticipated construction of a tunable
dye laser in this research group will offer a powerful
instrumental tool for such work, particularly in the

field of electrothermal atomic fluorescence, which has

not as yet been the subject of much study on either the

braid or many other devices.

APPENDICES

APPENDIX A

SIMPLE HARDWARE CONTROL APPROACH

FOR SEQUENCING CHEMICAL INSTRUMENTATION

by

S.R. Crouch*, D.N. Baxter, E.H. Pals and E.R. JohnsonJr

Department of Chemistry
Michigan State University
East Lansing, Michigan 48824

* Author to whom reprint requests should be sent
+ Present address: Varian MAT

25 Hanover Road
Florham Park, NJ 07932

308

309

In modern analytical instrumentation, digital controllers
have become more and more complex and flexible in recent years
as instruments have become more highly automated. The
controller directs the sequencing of various instrument
functions. For example, in a modern, automated gas chromato-
graphy system, the controller might be required to send
signals to an automatic sampler for introduction of the
sample into the column, to control the oven temperature and
its temperature program, to control the sensitivity of the
detection system, and to control the detection and integration
of the various chromatographic bands as they elute from the
column. Such controllers may be simple fixed order sequencers,
or may require branching to skip certain functions if certain
instrument conditions are set by the user (i.e., unless the
temperature program button and rate have been set, the
controller would skip this part of its sequence).

In modern instruments, controllers take the form of
hardwired logic systems, minicomputers or microprocessors.

The microprocessor is certain to become the controller of
choice in the future for many control operations, because
the program (sequence) can be altered without changes in
hardware. Also, even the most complex control operations,
with multiple and nested branches can be readily carried out
under the supervision of a microprocessor. However, the
microprocessor requires the instrument designer to be re-
educated in terms of software and processor capabilities.

This education process may require months to years in order

 

310

to get working systems into the laboratory with microproces-
sors as "intelligent” instrument controllers.

A few years ago our research group became heavily
involved in controllers when we decided to undertake a
project which involves the spatial profiling of atomic
concentrations above a filament-type electrothermal atomizer
using atomic absorption spectrometry. We desired to automate
fully the operations of sample introduction to the filament,
movement of the filament in both x and y directions away
from the AA optical axis to predetermined destinations,
temperature programming of the filament, data acquisition of
background corrected peak area, and automatic return to
position "zero" for pickup of a new sample. The filament
position controller was intended to move the atomization cell
freely over a total displacement of one inch in either of two
dimensions,through the use of stepper motor driven micrometer
translation stages. The controller was required to function
in either a "local" mode, in which the desired filament
position could be set by the user via front panel BCD
switches, or in an "on line" mode in which the controller
received binary information from a minicomputer and functioned
as a hardware slave to the processor. In addition, we
required the controller to convert BCD position code to
binary code in the local mode, to use position instead of
displacement as the input (i.e., to remember where it is and
to determine the direction to move and how far to go), and

to be capable of a hardware reset to an optically defined

311

reference position. Finally, we requirtd the controller

to correct for play in the leadscrews of the translation
stages in that if a requested movement was in the positive
direction (i.e., away from the optical reference), the
controller would intentionally move the cell to a known
distance beyond the desired destination, and then reverse
direction and return to the true destination, in contrast to
movement in the negative direction, which would proceed
directly to the destination without such overshoot. In this
manner, no matter whether the net movement was positive or
negative, the cell would be moved last in the negative
direction, thus leaving the leadscrews tightly meshed in
that direction of movement.

Because our group had no microprocessor experience at
that time, we decided on the simple hardware approach
described in this paper. Although the complexity of our
application is such that we might at this time have chosen
to use a microprocessor, there are many more simple control
functions for which the approach described is highly suit-
able. Even for our complex application, this systems-level
approach to controller design enabled us to design in about
3 hours a unit requiring over 50 IC's per motor controller.
Construction and debugging of the position controller required
additional time of course. The speed of our design was made
possible by the use of a method originally described by
Richardsl79, which allows all the complex sequential

operations to be treated by using only a simple flow chart

312

at the beginning, and by implementing the flow chart with
2-3 MSI IC's.

To understand the approach, we present in the first
section of this paper the general concepts and principles
of the "state" of a controller and the simple implementation
which results from the Richards179 approach. We then
present a chemical example of a "fall through" sequencer
which controls an automatic fraction collector used in
elution chromatography. In the last section we describe a
branching sequencer and illustrate its use as an electro-

thermal atomizer position controller.
A. THE CONCEPT OF THE STATE

In all but the most trivial control applications,
a series or sequence of functions must be controlled. For
such applications there are many approaches, but as the
number of functions and the complexity increases, solutions
can become exceedingly complicated. The systematic approach
advocated in this paper allows even the most complex control
applications to be handled with ease.

To help understand this approach several terms must
be defined. This can be done most conveniently with the aid
of an example. Consider the controller for the thermostat on

a temperature bath. Using the method described by

Richardsl79, it can be considered to have three states: State

0 = thermostat off - heater off; State 1 = heater on; and

313

State 2 = heater off. In State zero the thermostat con-
troller waits for the condition of the thermostat power
switch to be on. The on condition of the switch then is

referred to as a transfer condition. When the transfer

 

condition for state zero is true, i.e., the switch is on, the
controller advances to state one. The process of this
advance causes a transfer function to occur. The heater

is turned on. The second state (state one) tests to see if
the bath temperature exceeds the limit set on the thermostat.
When the bath temperature does exceed the limit, the con-
troller steps to state two, causing another transfer function
to occur. The heater is turned off. The transfer condition
for state two is a bath temperature below the thermostat
limit. When this condition is met, no transfer function is
required except to return to state zero and repeat the cycle.

Viewing such a trivial example in terms of states,
transfer conditions, and transfer functions may seem overly
complicated, but it is useful for two reasons. First,
the method is also applicable to much more complex control
applications. Second, by using this approach, the circuit
design is greatly simplified by use of a standard circuit for
the heart of the controller.

The application of this approach involves first convert-
ing the states, transfer conditions and transfer functions
into a standard flowchart format. The two basic flowchart
units for each state are shown in Figure A1. The unit which

is used depends on what is to occur at that particular state.

314

 

TRANSFER ’
Fougg‘ON STATE I

 

 

#:Nuggga
FU‘fIION _—€.STATE “OI

 

 

 

 

 

 
 

PERFORM

conoztxou '13 Tnnusrcn
n51? ruufifigou L'*""'E "”
v

   

 

 

 

Figure Al. Basic flowchart units for transfer
conditions and transfer functions.

315

Either unit contains a diamond shaped box which represents

the state and contains the transfer condition. Depending on
the application, the basic unit will also contain one or two
rectangles which represent the transfer functions which occur
depending on whether the transfer conditions are met or not.
In this method the transfer function on the "no" leg is
available only if the transfer condition not being met

causes a move to a state I other than the present state. This

transfer function, if it is used, is called a secondary

 

transfer function as opposed to the primaryptransfer function

 

on the "yes" leg. Figure A2 shows the flow chart for the
bath thermostat example. No secondary transfer functions
are required for this application and the controller is
therefore an example of a fall through sequencer (no condi-
tional branches).

Once the flowchart for the particular application has
been constructed, the circuitry required for the "heart" of
the controller can be chosen. The heart of a controller
constructed using this method consists of a k-bit counter,
an n-bit multiplexer and one or two n-bit decoders (the
number of decoders depends on whether the application requires
secondary transfer functions), where the number of states is
n and n E 2k.

Schematically, the heart of the controller is shown in
Figure A3. The output of the counter indicates the present

state of the controller. Using these outputs the multiplexer

supplies the transfer condition appropriate to that state

316

.umumOEumcu Lama m 80 mHQmeo on»

Hem upmao3oflm Hoocwoomm

.m< wusmflm

 

 

O mh<h¢
Oh Icahmt

 

 

 

o Ibnhuzau

Dm>

 

, cuu<ux
duo agar

 

 

 

 

n Zouhuzau

 

 

 

4

 

 

Imh<u1
80 883p

 

 

 

 

cowzuu
4<8¢m3h

 

 

<

lanruzah

DH>

62°

tornxo

 

 

 

Iowa!“

 

 

OATA
INPUTS

CLOCK

OF

OR AL
JUMP FUNCTT

INPUT
TRANSFER
CONDITIONS

Figure A3.

317

 

._g

 

 

 

L
0N8

 

 

 

 

 

 

 

 

 

 

——u
-—y
-—’

a \
-i

 

 

 

 

 

 

. a
u 311 ' ’
cgufi%gn oscoozn 2-—4>sscoqunv
secouoanv ' TRANSFER
- FUNCTIONS
ENABLE Ham u-n H
LO OUTPUTS SELECT " ""
1 1:
. SELECT SELECT . l
a OUTPUT —o-a ENABLE : -->
2 2~—4> PRIMARY
u 311 - tnnusren
”5°00!" ruucrlous
u 311 PRIMARY .
aux
N-I u~1 —-—>
u uu—-+»

 

 

General schematic diagram of a controller,

 

318

to the state counter and to the decoder(s). If the
transfer condition is true, then the clock on the state
counter is allowed to increment to the next state. The
presence of a true transfer condition also causes the
associated primary transfer function from the decoder to go
true, but this is only a pulse since the next clock pulse
will step the state counter to the next state. If the
transfer condition is false either a secondary transfer
function may be generated if the controller is to go to a
new state or the state counter may wait for the transfer
function to become true.

Implementing this circuit for an eight state sequencer
requires a 74163 counter, a 74151 multiplexer, and one or
two 7442 decoders. At current single unit quantity prices

they can be obtained for just over three dollars.

B. THE FALL-THROUGH SEQUENCER

The fall-through sequencer is the simplest type of
sequencer. In this sequencer the only type of decision to
be made is whether to remain in the present state or proceed
to the next state. In a closed-100p operation this decision
is based on receiving an "all done" signal (flag) from a
previously actuated circuit. In an open-loop Operation the
change of state may be immediate or may occur upon receiving

a signal that a time delay has elapsed.

 

319

A chemical example of a fall-through sequencer is a
controller for an automatic fraction collector used in
elution chromatography. Typically, the fraction collector
has a rotatable tray which contains vials that are sequen-
tially rotated beneath the column outlet valve. When a
vial is directly beneath the outlet valve, the valve is
opened for a preset time, then closed, and the next vial
rotated into the receiving position. These motions are then
repeated until the desired number of fractions have been
collected.

In Figure A4 the decisions and actions involved in
controlling this apparatus are represented in flow diagram
form, along with the inputs used by the decisions in
functioning. Assume that the operator has initialized the
apparatus by rotating the first vial into position beneath
the valve and loaded the number of fractions desired into a
counter register. Also assume that the controller is
initially in State 0. Condition A is satisfied since the
counter is non zero, so the controller generates Function A,
which is to open the delivery valve and to start the delay
timer. Now the controller moves to state 1 and repeatedly
queries the delay timer until it signals that enough solution
has been delivered (Condition B). When Condition B has been
satisfied the controller closes the valve and starts the vial
tray moving (Function B). Now the controller moves to State

2 and waits for a position indicator microswitch beneath the

.Houomaaoo
cofluomum caumeousm no mo mamamxm may now pumno3on umocmzomm .vm muomflm

 

 

><¢h
tOrw

Garzaou
hzuxmcouo

 

 

 

320

u Icaruzau

 

ozn>ox

 

    
   

 

 

 

 

 

 

O zonhulau

 

 

bount<4u
><4ub
uxnp

Cut—h
><JUD
P¢¢hu

U>J<>
Into

 

 
   

   

9¢uh8300
OCUanoz

 

 

< zo~h0293

 

321

outlet valve to be tripped (Condition C). When Condition

C is met the controller must decrement the fraction counter
register, stop the tray, and effect a jump back to controller
state a (Function C). This sequence is then repeated until
the fraction counter register is decremented to zero. When
this happens condition A can never be met, so the apparatus
halts since it becomes hung up in State 0.

This controller can be quite easily implemented using a

 

sequencer composed of three IC's: a 74163 counter, a 74151
multiplexer, and a 7442 decoder. Of course additional
packages may be necessary for such things as delay timing

but these three IC's handle the generation of all the

signals necessary to initiate the actions at the appropriate
times. Figure A5 shows how these IC's are interconnected

and also shows the input and output signals that are

required in this particular application. When power is
turned on, an initialize pulse is required to clear the
fraction counter register and to assert the clear input of
the state counter. This forces the counter to zero, which is
State 0. The QA, QB, and QC outputs of the state counter

are all LO, which sets the multiplexer's Y output L0. This
LO signal is fed to the state counter's ENABLE P input, which
inhibits the counter from counting clock pulses at its CLOCK
input. Thus it remains in State 0. The multiplexer's
complementary output W is directed to the decoder's D

input. Output W is HI, and examination of the truth table

322

95

 

COUNTER

f,"

O

 

 

CNT

L

FTTOPEN VALVE, START Tin: DELAV

 

 

 

an
'1' DJ‘ ENP 9A 03 Oc
m l

 

 

 

r '.—:

 

 

 

CO Y A O C
C1—'
C2 MUX

“CIR-D.N-.

 

 

 

 

riicLosE VALVE, NOV! 3AuPLE TRAY
731310? TRAY, DECRENENT COUNTER
oD-——FT
0EconEn ‘ P“.
A zp———EE
8 D
«D
ab
ED
P—-1° 7 D
CIICOUNTER 18 NONZERO
cactan DELAY 18 DONE

D-j1. czsnxcnosultcu stnucx‘

 

 

Figure A5. Schematic diagram of the controller for the
example of the automatic fraction collector.

323

for the decoder indicates that the 0 through 7 decoder
outputs must be HI when D is HI (Negative logic is used at
these outputs so that a function is generated only when the
output goes L0). This means that initially none of the
functions are being asserted. When the operator gives the
"GO" command, the number of fractions desired is loaded into
the fraction counter register, and Condition A is satisfied.
Now the multiplexer's w output goes L0, and the decoder now
points to a decimal 0, which asserts the Function A signal.

Function A triggers external circuitry to open the
delivery valve and start a time delay circuit. In addition,
the multiplexer Y output is HI, so that the state counter's
ENABLE P input is asserted, and on the first clock pulse the
counter is incremented once (State 1). Now the multiplexer
shifts from selecting data channel 0 (Condition A) to data
channel 1 (Condition B). Assuming that the required time
delay has not yet elapsed, Condition B is not yet
satisfied so that the multiplexer Y output falls, locking the
state counter in State 1, and the W output goes high,
concluding function A and inhibiting the decoder from
generating any functions.

When the required time delay has elapsed, Condition B
is satisfied and a similar sequence occurs in the state
counter, multiplexer and decoder. The multiplexer's W output
enables the decoder, and its A, B, C and D inputs point to
decimal 1. Thus, output 1 (Function B) is asserted. This

triggers the valve to close and starts the tray moving. The

 

324

multiplexer Y output goes high allowing the counter to count
one pulse and move to State 2. Now the state counter
requests the multiplexer to address the status of Condition
C.

When the new sample vial reaches the position beneath
the outlet valve, a microswitch is tripped and Condition C
is satisfied. Again this activates the decoder, and Function

C is asserted. Function C stops the tray and decrements the

 

fraction counter register. Also, since this is the last
function in the sequence, it is used to assert the CLEAR
input of the state counter and return the controller to
State 9.

Unless this was the last fraction to be collected, the
fraction counter register will be non-zero when the state
counter enters State a so that Condition A is immediately
satisfied. Thus, as soon as the Function C pulse has cleared
the state counter to State U, Function A will be generated
and on the next clock pulse State 1 will be achieved. This
entire delivery sequence will continue until the fraction
counter register reaches zero, which prevents Condition A
from being satisfied. Then the apparatus will halt because
the controller becomes hung up in State 8 until the operator

intervenes to start a new collection sequence.

325

C. THE BRANCHING SEQUENCER

In many cases it is desirable to have a sequencer
capable of branching; that is, one in which sections of
the basic state sequence may be skipped, repeated, or other-
wise altered as requested by the instrumentation under
control. An example of such a sequencer is the electro-
thermal atomizer position controller mentioned previously.

The flow diagram of its sequence is shown in Figure A6.
The system waits in State U until data representing a new
location are received, and a command to move is given. Upon
receipt of this command, the overflow flag is cleared, and
the system determines if it is being operated under local or
computer control. If control is local, the desired
destination has been entered in BCD format from front panel
thumb-wheel switches, which requires a BCD to binary conver-
sion. A check is made to see if the system is free of over-
flow, and if so, another check is made to see if BCD to
binary conversion is complete. If conversion is not complete,
the registers are clocked, and the checks for overflow and
complete conversion are repeated. If an overflow should
occur, the sequence will abort to State a with the overflow
flag set. (Overflow results from entering a destination
outside the defined limits of the system.) If no overflow
occurs, then upon completion of the conversion, the stepper

motor direction is determined, and the stage begins to move.

 

 

326

 

.ucmEsnumofi mcflcOMuHmoo ucmEmHHm may now uumnosoam “mocmovmm .84 musmwm

 

57
a p

.kV NV, . AV,
08hr. :83
:2. :58 .
Eu»!
. 2 3.8.8.. 8 .3293...

O

 

 

 

 

Capo:
pear.
lo— poucn. b
pun

 

 

 

 

to»:
3r.
as to u r0885
0‘...-
ran-(u
bu.
no i n r0835

no tags
I I

. k7 NV
E O E .9 E O
.- Enho‘n ad gag

 

 

 

 

3%..

«a 328...».

 

 

 

”I 'wgu

327

If control is from the computer, the BCD to binary
conversion is unnecessary. Thus, this section of the
sequence is omitted.

The stage moves until the destination is reached. At
this point the system checks to see if the direction of
movement was positive. If so, the stage continues to move
ten additional increments positive, and when finished,
reverses direction and comes back ten increments negative.
If movement was originally negative, the sequence is
concluded immediately. In this manner gear play is removed
from the mechanism by assuring that movement always occurs
last in the negative direction.

To implement a branching sequencer such as this, it
is necessary to add only one additional decoder IC to the
basic controller circuit. The actual wiring is shown in
Figure A7. The actions of this controller are identical to
the simpler three chip system previously discussed as long
as the basic, non-branching sequence is followed. If a
branch is called for, however, (if the selected data
channel of the multiplexer is low when a potentially branch-
able state is reached), the following events occur. The
L0 state of output Y from the multiplexer causes the
function pulse to be generated by the second decoder instead
of the first decoder. This function pulse is used by the
instrument to perform tasks in the same manner as any function
pulse from the first decoder. In addition, this pulse is

also fed back to the state counter where it serves

328

 

 

 

 

 

 

 

 

 

 

 

 

 

 

95
.L
7 A on
'8, ,. cauuun DECODEB ‘ a——rl'!’
ab—-rt§
F“ ’C [NT . h,“
0 ‘IP
.rLer——ch sp——rf§
WIT—4 an "" "9— ° ' D
L 05 Or. Re A . c 7 D
It! [ l
V” [
FF! .
'" 1 1 '1 J
cu-—-o A T ‘: A T c OD—-$TT
Cl—l Y— ogcoosn I D—F'IT
c2—— 2 NW 2 p——r'c1’
ca-—-o up———n a>——EUT
ca-—wa a>——FFT
cs-—us s>——FFT
cs-—Jo s o>——¢FT
c7——Jz D:L 7>——rfir

 

 

 

 

 

 

Figure A7. Schematic diagram of the controller for
the filament positioning instrument.

329

simultaneously to provide binary data representing the state
to which to branch as well as to load those data into the
state counter. For example, in the positioner sequence, if
the system is under computer control when state 1 is entered,
the LO level at the data channel 1 of the multiplexer disables
the first decoder, but causes the second decoder to generate
function 82. Function 82 is used in the circuitry to update
the direction of movement and start the motor, just as
function 01 of the first decoder does under local control.
But, it is also present at the C and Load inputs of the state
counter, so that as it occurs, the state counter is forced
to skip directly to state 4 through parallel loading. All
additional functions which are part of other such branches
are similarly directed back to the Load and appropriate data
inputs of the state counter, using simple gates as shown to
OR them together as required.

By adding additional multiplexer and decoder chips to
the controller, it is possible to produce sequencers of
even greater complexity in which the branches consist not
only of single function pulses, but also of additional states
or even complete subsequences reminiscent of nested computer
subroutinesl79. Similarly, it is also possible to
substitute basic chips different from those employed here
to produce sequences of more than eight basic states.
However, such systems can quickly become unwieldy in their

complexity and might very well be better performed using

330

microprocessor-managed controllers even if the user has no

previous experience with such devicesng—lgs.

D. CONCLUSIONS

We have described in this paper a hardware approach
for designing instrument sequencers, which we feel greatly
reduces design time, yet allows complex sequences to be
readily implemented. The increasing education of chemists
as to the potential, the software, and the interfacing of
microprocessors will certainly lead to a level of controller
complexity at which a microprocessor is the sequencer of
choice. Even so, the concepts presented here should enable
simple controllers to be designed and implemented readily
by chemists unfamiliar with microprocessors or for applica-
tions in which it is undesirable to "tie up" the processor
in simple, unchanging control tasks. In our own laboratories,
the cross-over point between using hardwired-logic and micro-
processor controllers has continuously declined as our know-
ledge of microprocessor technology has improved. At
present, in our case, a controller requiring >30 IC's
would probably be implemented with microprocessors. For
electronics designers, the crossover point is undoubtedly
fewer IC's than our figure. Nonetheless, the simplicity

of the approach presented here, should prove valuable to

 

331

other chemists involved in the automation of chemical

instruments.

ACKNOWLEDGMENTS

We are grateful to Dr. T.V. Atkinson for his assistance
in the use of a computerized graphics facility for the
production of our figures. D.N.B. acknowledges both a
summer and a full year ACS Analytical Fellowship under the
sponsorship of the Procter and Gamble Company of Cincinnati,

Ohio.

APPENDIX B

SOURCES OF COMMERCIAL INSTRUMENTATION

Computer:
Model 8/e - Digital Equipment, Maynard, MA
Internal Bus Options:
8K core memory
8K random access memory
Extended arithmetic element
Real-time clock
Positive I/O bus interface
External Bus Options:
RK8E hard disc
Dual floppy disc (Model 7200, Sykes Datatronics,
Rochester, NY)
Graphics terminal (Consul-980, Applied Digital
Data Systems, Hauppauge, NY)
LA-30 Decwriter
Computer Interface Buffer (Heath, Benton Harbor, MI)
Spectrometer Components: GCA McPherson, Acton, MA
HCDT Supply: Model EU-703-70
Monochromator: Model EU-700
Photomultiplier: Model EU-701-30

Spectrometer Accessories:

332

 

333

Hollow Cathode Lamps:
Jarrell-Ash, Fisher Scientific, Livonia, MI
Westinghouse, Elmira, NY
Flowmeters: Type 13 Roger Gilmont Instruments, Great
Neck, NY
Lenses, chimney plates: Grade Sl-UV quartz, Esco
Optics, Oak Ridge, NY
Photocurrent Amplifier: Model 427, Keithley Instruments,
Cleveland, OH
Laser: Model LS-32, Electro-Nuclear Labs, Waltham, MA
Positioner:
Translation Stages: Model 420-51, Newport Research,
Fountain Valley, CA
Circuit Boards: Type ll-DE-S, Douglas Electronics,
San Leandro, CA
Motors: Model 23D6102A, Computer Devices, Santa Fe
Springs, CA
Autosampler:
Syringe: Model 51100, Roger Gilmont Instruments,
Great Neck, NY
Motors: Models 23D6102A (rotor) and 23D6306A (vertical),
Computer Devices, Santa Fe Springs, CA
Models 202215D200-Fl.6 (syringe) and 10-2013D40-F75
(turret), Sigma Instruments, Braintree, MA
Data Acquisition and Other:

Converters:

D/A: Type DACHY12BC, Datel, Canton, MA

 

334

A/D: Type ADCHYlZBC, Datel, Canton, MA (no longer
available)
Power Supplies:
HD5-12/OVP (+5 volts) and HBAA-40W (£15 volts),
Power One, Camarillo, CA
Card Rack:~ Model SR-20105, Bud Radio,

Willoughby, OH

 

REFERENCES

 

11.

12.

13.
14.

15.

16.

REFERENCES

Hieftje, G. M., Copeland, T. R., and deOlivares,
D. R., Anal. Chem., 48, 142R (1976).

Winefordner, J. D., and Vickers, T. J., Anal. Chem.,
'46, 192R (1974).

Woodward, C., "Annual Reports on Analytical Atomic
Spectroscopy, 1973", Vol. 3, Society for Analytical
Chemistry, London, 1974.

WOodward, C., "Annual Reports on Analytical Atomic
Spectroscopy: 1974", Vol. 4, Society for Analytical
Chemistry, London, 1974.

Kirchhoff, G., and Bunsen, R., Fresenius Z. Anal.
Chem., 1, l (1862).

Mika, J., and Torok, T., "Analytical Emission Spec-
trosc0py", Crane and Russek, New York, 1974.

Ahrens, L. H., and Taylor, S. R., "Spectrochemical
Analysis", Addison-Wesley, Reading, Massachusetts,
1961.

Slavin, M., "Emission Spectrochemical Analysis",
Wiley-Interscience, New York, 1971.

Barnes, R. M., Syst. Mater. Anal., 3, 23 (1974).

Butler, L. R. P., Human, H. G. C., and Scott, R. H.,
Handb. Spectrosc., l, 816 (1974).

Skogerboe, R. K., ibid., 1, 845 (1974).

Mitchell, A. C. G., and Zemansky, M. W., "Reasonance
Radiation and Excited Atoms", Cambridge University
Press, New York, 1961.

Walsh, A., Anal. Chem., 46, 698A (1974).

Walsh, A., Spectrochim. Acta, 1, 108 (1955).

Russell, E. J., Shelton, J. P., and Walsh, A.,
Spectrochim. Acta, 8, 317 (1958).

Gaydon, A. G., "The Spectroscopy of Flames", 2nd.
ed., Halsted, New York, 1974.

335

 

17.

18.

19.

20.

21.

22.
23.
24.

25.

26.
27.
28.

29.

30.

31.

32.
33.

34.

35.

36.

336

Kirkbright,G. F., and Sargent, M., "Atomic Absorption
and Fluorescence Spectroscopy", Academic Press,

New York, 1974.

Wood, R. W., Phil. Mag., 22, 513 (1905).

Nichols, E. L., and Howes, H. L., Phys. Rev., 22,
472 (1924).

Alkemade, C. Th. J., Proc. Colloq. Spectrs. Internat.,
Spartan Books, Washington, D.C., 1963.

Winefordner, J. D., and Vickers, T. J., Anal. Chem.,
22, 161 (1964).

Browner, R. F., Analyst, 22, 617 (1974).
West, T. 8., Analyst, 22, 886 (1974).
Kirkbright, G. F., Analyst, 22, 609 (1971).

Denton, M. B., and Malmstadt, H. V., Anal. Chem.,
32, 241 (1972).

Syty, A., Crit. Rev. Anal. Chem., 3, 155 (1974).
King, A. S., Astrophys. J., 22, 318 (1922).
King, A. S., and King, R. B., ibid., 22, 377 (1935).

King, R. B., and Stockbarger, D. C., ibid., 22,
488 (1940).

Lvov, B. V., Spectrochim. Acta, ll, 761 (1961).

Lvov, B. V., and Lebedev, G. G., Zh. Prikl. Spektrosk.,
Z. 264 (1967).

Lvov, B. V., Spectrochim. Acta, 248, 53 (1969).

Woodriff, R., and Ramelow, G., Spectrochim. Acta,
238, 665 (1968).

Woodriff, R., Stone, R. W., and Held, A. M., Appl.
Spec., 22, 408 (1968).

Woodriff, R., and Stone, R., Appl. Optics, 1, 1337
(1968).

Woodriff, R., Culver, B. R., and Olsen, K. W., Appl.
Spec., 21, 530 (1970).

 

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

337

Woodriff, R., and Shrader, D., Anal. Chem., 22,
1918 (1971).

Pagenkopf, G. K., Neuman, D. R., and Woodriff, R.,
Anal. Chem., 12, 2248 (1972).

Woodriff, R., Culver, B. R., Shrader, D., and Super,
A. 8., Anal. Chem., 22, 230 (1973).

Massmann, H., Spectrochim. Acta, 23B, 215 (1968).
Massmann, H., in "Flame Emission and Atomic Absorption
Spectrometry", Vol. 2, Dean, J. A., and Rains, T. C.,
Eds., Marcel Dekker, New York, 1971.

Perkin-Elmer Corporation, Instrument News, 22, 4
(1970).

Kahn, H. L., and Slavin, S., A. A. Newsletter, 22,
125 (1971).

Perkin-Elmer Corporation, "Analytical Methods for
Atomic Absorption Spectroscopy using the HGA Graphite
Furnace", Norwalk, Conn., March, 1973.

Winefordner, J. D., Pure Appl. Chem., 22, 35 (1970).

Christian, C. M., and Robinson, J. W., Anal. Chim.
Acta, 29, 466 (1971).

Robinson, J. W., and Slevin, P. J., Amer. Lab., August,
1972, pp. 10.

Robinson, J. W., Slevin, P. J. Hindman, G. C., and
Wolcott, D. C., Anal. Chim. Acta, 22, 431 (1972).

Robinson, J. W., Wolcott, D. K., Slevin, P. J., and
Hindman, G. D., Anal. Chim. Acta, 22, 13 (1973).

Headridge, J. B., and Smith, D. R., Talanta, 22,
247 (1971).

Langmyhr, F. J., and Thomassen, Y., Z. Anal. Chem.,
264, 122 (1973).

Murphy, M. K., Clyburn, S. A., and Veillon, C., Anal.
Chem., 32, 1468 (1973).

Black, M. S., Glenn, T. H., Bratzel, M. P., and
Winefordner, J. D., Anal. Chem., 22, 1769 (1971).

Molnar, C. J., and Winefordner, J. D., Anal. Chem.,
.12, 1419 (1974).

 

 

 

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.
68.

69.

70.

71.

72.

73.

338

Chapman, J. F., Dale, L. S., and Kelly, J. W., Anal.
Chim. Acta, 22, 207 (1974).

Robinson, J. W., and WOlcott, D. K., Anal. Chim.
Acta, 12, 43 (1975).

Robinson, J. W., Wolcott, D. K., and Rhodes, L.,
Anal. Chim. Acta, 12, 285 (1975).

Church, D. A., Hadeishi, T., Leong, L. McLaughlin,
R. D., and Aak, B. D., Anal. Chem., 22, 1352 (1974).

Kantor, T., Clyburn, S., and Veillon, C., Anal. Chem.,
22, 2205 (1974).

 

Clyburn, S. A., Kantor, T., and Veillon, C., Anal.
Chem., 22, 2213 (1974).

Molnar, C. J., and Winefordner, J. D., Anal. Chem.,
22, 1807 (1974).

Sturgeon, R. E., and Chakrabarti, C. L., Anal. Chem.,
22, 90 (1977).

Runnels, J. H., Merryfield, R., and Fisher, H. B.,
Anal. Chem., 21, 1258 (1975).

Fuller, C. W., and Thompson, 1., 102, 141 (1977).

Thomassen, Y., Larsen, B., Langmyhr, F., and Lund,
W., Anal. Chim. Acta, 22, 103 (1976).

Ottaway, J. M., and Shaw, F., Analyst, 100, 438
(1975).

DeGalan, L., Anal. Chim. Acta, 21, 259 (1976).

Fuchs, C., Brasche, M., Paschen, K., Nordbech, H.,
and Quellhorst, E., Clin. Chem. Acta, 22, 71 (1974).

Perry, E. F., Koirtyohann, S. R., and Perry, H. M.
Jr., Clin. Chem., 22, 626 (1975).

Pekarek, R. S., Hauer, E. C., Wannemacher, R. W.,
Jr., and Beisel, W. R., Anal. Biochem., 22, 283 (1974).

Kamel, H., Brown, D. H., Ottaway, J. M., and Smith,
W. E., Analyst, 101, 790 (1976).

Fernandez, F. J., Clin. Chem., 22, 558 (1975).

Toffaletti, J., and Savory, J., Anal. Chem., 21,
2091 (1975).

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

339

Ebert, J., and Jungmann, H. Z. Anal. Chem., 272,
287 (1974).

Chao, S. S., Kanabrocki, E. L., Moore, C. E., Oester,
Y. T., Greco, J., and vonSmolinski, A., Appl. Spec.,
22, 155 (1976).

Ihnat, M., Anal. Chim. Acta, 22, 293 (1976).

Gong, H., and Suhr, N. H., 22, 297 (1976).

Langmyhr, F. J., Stubergh, J. R., Thomassen, Y.,
Hanssen, J. E., and Dolezal, J., Anal. Chim. Acta,
12, 35 (1974).

Siemer, D., and Woodriff, R., Spectrochim. Acta,
293, 269 (1974).

Frech, W., and Cedergren, A., Anal. Chim. Acta. 22,
57 (1977).

Henn, E. L., Anal. Chem., 21, 428 (1975).

Lord, D. A., McLaren, J. W., and Wheeler, R. C.,
Anal. Chem., 22, 257 (1977).

Hageman, L., Torma, L., Ginther, B. E., J. Assoc.
Off. Anal. Chem., 22, 990 (1975).

Persson, J. A., Frech, W., and Cedergren, A., Anal.
Chim. Acta, 22, 119 (1977).

Medina, R., Z. Anal. Chem., 271, 346 (1974).

Kragsten, J., and Reynaert, A. P., Talanta, 22,
618 (1974).

Marks, J. Y., Welcher, G. G., and Spellman, R. J.,
Appl. Spec., 22, 9 (1977).

West, T. S., and Williams, X. K., Anal. Chim. Acta.
21, 281 (1969).

Anderson, R. G., Maines, I. S., and West, T. S.,
Anal. Chim. Acta, 22, 355 (1970).

Alder, J. F., and West, T. S., Anal. Chim. Acta, 22,
365 (1970).

Jackson, K. W., West, T. S., and Balchin, L., Anal.
22, 249 (1973).

 

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

106.

107.

108.

340

Belyaev, Uy. I., Pchelintsev, A. M., Zvereva, N. F.,
and Kostin, B. I., Zh. Anal. Khim., 22, 492 (1971).

Belyaev, Uy. I., Karyakin, A. V., and Pchelintsev,
A. M., Zh. Anal. Khim., 22, 852 (1970).

Belyaev, Uy. I., Koveshnikova, T. A., and Kostin,
B. I., Zh. Anal. Khim., 22, 2111 (1973).

Belyaev, Uy. I., Pchelintsev, A. M., and Zvereva,
N. F., Zh. Anal. Khim., 22, 1295 (1971).

Amos, M. D., Bennett, P. A., Brodie, K. G., Lung,
P. W. L., and Matousek, J. P., Anal. Chem., 22,
211 (1971).

Amos, M. D. and Matousek, J. P., 23rd Pittsburgh
Conf. Anal. Chem. Appl. Spectrosc., Cleveland, Ohio,
March 1972.

Dipierro, S., and Tessari, G., Talanta, 22, 707
(1971).

Montaser, A., Goode, S. R., and Crouch, S. R., Anal.
Chem., 22, 599 (1974).

Montaser, A., and Crouch, S. R., Anal. Chem., 22,
1817 (1974).

Montaser, A., and Crouch, S. R., Anal. Chem., 21,
38 (1975).

Molnar, C. J., Reeves, R. D., Winefordner, J. D.,
Glenn, M. T., Ahlstrom, J. R., and Savory, J., Appl.
Spec., 22, 606 (1972).

Donega, H. M., and Burgess, T. E., Anal. Chem., 22,
1521 (1970).

Hwang, J. Y. Ullucci, P. A., and Smith, S. B., Amer.
Lab. August 1971, pp. 41.

Takeuchi, T., Yanagisawa, M., and Suzuki, M., Talanta,
22, 465 (1972).

Maruta, T., and Takeuchi, T., Anal. Chim. Acta, 22,
253 (1973).

Cantle, J. E., and West, T. S., Talanta, 22, 459
(1973).

Winefordner, J. D., Pure Appl. Chem., 23, 35 (1970).

109.

110.

111.

112.

113.

114.

115.

116.

117.

118.

119.

120.

121.

122.

123.

124.

125.

126.

341

Bratzel, M. P., Dagnall, R. M., and Winefordner, J. D.,

Appl. Spec., 22, 518 (1970).

Bratzel, M. P., Dagnall, R. M., and Winefordner,
J. D., Anal. Chim. Acta, 22, 197 (1969).

Goode, S. R., Montaser,
Spec., 21, 355 (1973).

A., and Crouch, S. R., Appl.

Williams. M. and Piepmeier, E. H., Anal. Chem., 22,
1342 (1972).

Chauvin, J. V., Newton, M. P., and Davis, D. G.,
Anal. Chim. Acta, 22, 291 (1973).

McCullough, M. R.,
22, 1006 (1976).

and Vickers, J. J., Anal. Chem.,

Ohta, K., and Suzuki, M., Talanta, 22, 465 (1975).

McIntyre, N. S., Cook, M. G.,
Chem., 22, 1983 (1974).

and Boase, D. G., Anal.

Newton, M. P.,
2003 (1975).

and Davis, P. G., Anal. Chem., 21,

Kolihova, D., Sychra, V., Chem. Listy, 22, 1091
(1974).

Evenson, M. A., and Warren, E. L., Clin. Chem., 22,
537 (1975).

Stafford, D. T., and Saharovici, F.,
Acta, 298, 277 (1974).

Spectrochim.

Henkin, R. I., Mueller, C. W., and Wolf, R. O., J.
Lab. Clin. Med., 22, 175 (1975).

Blood, E. R.,
1438 (1975).

and Grant, G. C., Anal. Chem., 21,

Ratcliffe, D. G., Byford, C. S., and Osman, P. B.,
Anal. Chim. Acta, 12, 457 (1975).

Mullen, J. D., Talanta, 3;, 846 (1976).

Nikolaev, G. I., Podgornaya, V. I., Kalinin, S. K.,
and Zakharenko, V. M., J. Anal. Chem. USSR, 22,
124 (1975).

Ohta, K., and Suzuki, M., Anal. Chim. Acta,

llr
288 (1975).

127.

128.

129.

130.

131.

132.

133.

134.

135.

136.

137.

138.

139.

140.

141.

142.

143.

144.

145.

342

Tam, K. C., Environ. Sci. Technol. 2, 734 (1974).

Janouskova, J., Sulcek, Z., and Sychra, V., Chem.
Listy, 22, 969 (1974).

Shigematsu, T., Matsui, M., Fujino, O., and Kinoshita,
K., Anal. Chim. Acta, 12, 329 (1975).

Janssens, M. and Dams, R., Anal. Chim. Acta, 12,
25 (1974).

Brady, D. V., Nantalov, J. G., Jr., Glowacki, G.,
and Pisciotta, A., Anal. Chim. Acta, 12, 448 (1974).

Riandey, C., Linhares, P., and Pinta, M., Analusis,
2, 202 (1975).

Oborne, A. C., and West, T. S., Proc. Soc. Anal.
Chem., 2, 198 (1972).

Chuang, F. S., and Winefordner, J. D., Appl. Spec.,
22, 215 (1974).

Dittrich, K., and Mothes, W., Talanta, 22, 318 (1975).

Goleb, J. A., and Midkiff, C. R., Appl. Spec., 22,
44 (1975).

Henn, E. L., Anal. Chim. Acta, 12, 273 (1974).

Huffman, H. L., Jr., and Caruso, J. A., J. Agric.
Food Chem., 22, 824 (1974).

Alger, D., Anderson, R. G., Maines, I. S., and West,
T. S., Anal. Chim. Acta, 21, 271 (1971).

Reeves, R. D., Patel, B. M., Molnar, C. J., and Wine-
fordner, J. D., Anal. Chem., 22, 246 (1973).

Prudnikov, E. D., Zh. Prikl, Spektrosk., 12, 145
(1971).

Prudnikov, E. D., Zh. Anal. Khim., 21, 2327 (1972).

Prudnikov, E. D., and Shapkina, U. C., Zh. Anal.
Khim., 22, 1055 (1973).

Kahn, H. L., Peterson, G. E., and Schallis, J. E.,
A. A. Newsletter, 1, 35 (1968).

Ringhardtz, I., and Welz, B., Z. Anal. Chem., 243,
190 (1968).

146.

147.

148.
149 O

150.

151.

152.

153.

154.
155.

156.
157.

158.

159.

160.

161.

162.

163.
164.

343

Hauser, T. R., Hinners, T. A., and Kent, J. L., Anal.

Chem., 51, 1819 (1972).

Curry, A. J., Read, J. F., and Knott, A.
22, 744 (1969).

R., Analyst,

Delves, H. T., Analyst, 22, 431 (1970).

Ediger, R. D., and Coleman, R. L., A. A. Newsletter,
11, 33 (1972).
Joselow, M. M., and Bogden, J. D., A. A. Newsletter,

11, 99 (1972).

Rose, G. A., and Willden, E. G., Analyst, 22, 243
(1973).

Barthel, W. F., Smrek, A. L., Angel, G. P., Liddle,
J. A., Landrigan, P. J., Gehlbach, S. H., and Chisolm,
J. J., J. Assoc. Offic. Anal. Chem. 22, 1252 (1973).

Clark, D., Dagnall, R. M., and West, T. S., Anal.
Chim. Acta, 22, 219 (1972).

Lieberman, K. W., Clin. Chem. Acta, 22, 219 (1973).

Heinemann, G., Z. Klin. Chem. Klin. Biochem., 11,
197 (1973).

Henn, E. L., A. A. Newsletter, 12, 109 (1973).
Holak, W. Anal. Chem., 21, 1712 (1969).

Lichte, F. E., and Skogerboe, R. K., Anal. Chem.,
22, 1480 (1972).

Chu, R. C., Barron, G. P., and Baumgartner, P. A. W.,
Anal. Chem. 22, 1476 (1972).

Vijan, P. N., and WOod, G. R., A. A. Newsletter, 12,
33 (1974).

Schmidt, F. J., and Royer, J. L., Anal. Lett., 2,
17 (1973).

Pollock, E. N., and West,
11, 104 (1972).

S. J., A. A. Newsletter,

Fernandez, F. J., A. A. Newsletter, 12, 93 (1973).

Poluektov, N. S., Vitkun, R. A., and Zelyukova, Uy.
V., Ah. Anal. Khim., 12, 937 (1964).

165.

166.

167.

168.

169.

170.

171.

172.

173.

174.

175.

176.

177.

178.

179.

180.

181.

344

Agemian, H., and Chan, A. S. Y., Anal. Chim. Acta,
12, 297 (1975).

Dessani, S. D., McClellan, B. E., and Gordon, M.,
J. Agric. Food Chem., 22, 671 (1975).

Teeny, F. M., J. Agric. Food Chem., 22, 668 (1975).
Knauer, H. E., and Milliman, G. E., Anal. Chem., 21,
1263 (1975).

Hawley, J. E., and Ingle, J. D., Jr., Anal. Chem.,
21, 719 (1975).

Lundgren, G., Lundmark, L., and Johansson, G., Anal.
Chem., 22, 1028 (1974).

Patel, B. M., Reeves, R. D., Browner, R. F., Molnar,
C. J., and Winefordner, J. D., Appl. Spec., 21,
171 (1973).

Reeves, R. D., Patel, B. M., Molnar, C. J., and Wine-
fordner, J. D., Anal. Chem., 22, 246 (1973).

Aggett, J., and West, T. 8., Anal. Chim. Acta, 22,
359 (1971).

Aggett, J., and West, T. S., Anal. Chim. Acta, 22,
349 (1971).

Ebdon, L., Kirkbright, G. F., and West, T. S., Anal.
Chim. Acta, 22, 187 (1972).

Johnson, D. J., West, T. S., and Dagnall, R. M.,
Anal. Chim. Acta, 22, 171 (1973).

Torsi, G., and Tessari, G., Anal. Chem., 22, 1812
(1973).

Denton, M. B., Routh, M. W., Mack, J. D., and Swartz,
D. B., Amer. Lab., 2 (2), 69 (1976).

Richards, C. L., in "Circuit Designer's Casebook",
(Electronics Magazine, eds.), McGraw-Hill, New York,
pp. 88-940

Crouch,
E. R.,

S. R., Baxter, D. N., Pals, and Johnson,

submitted to Anal. Chem.

B. H.,

"Sigma Stepping Motor Handbook",
eds.), Braintree, Mass, 1972, pp.

(Sigma Instruments,
23.

 

182.

183.

184.

185.

186.
187.
188.

189.

190.

191.

192.

193.

194.

195.

196.

345

Malmstadt, H. V., Enke, C. G., and Crouch, S. R.,
"Digital and Analog Data Conversions", Benjamin,
New York, 1973.

Montaser, A., Ph.D. Thesis, Michigan State University,
East Lansing, Michigan, 1974.

Crouch, S. R., Montaser, A., Goode, S. R., in "In-
formation Chemistry: Computer Assisted Chemical
Research Design", (8. Fujiwara and H. B. Mark, Jr.,
eds.),University of Tokyo Press, Tokyo, 1975, pp.
107-124.

"Designing with TTL Integrated Circuits", (R. L.
Morris and J. R. Miller, Eds.), McGraw—Hill, New York,
1971, pp. 177, 241.

Beer, M., Anal. Chem., 48, 93R (1976).

Beer, M., Anal. Chem., 22, 428R (1974).
Ubbelohde, A. R., and Lewis, F. A., "Graphite and
Its Crystal Compounds", Clarendon Press, Oxford,
1970.

"C. R. C. Handbook of Chemistry and Physics", (R. C.
Weast, ed.), The Chemical Rubber Company, Cleveland,
1969-1970: ppo E-2370

R. E. Sturgeon, C. L. Chakrabarti, and C. H. Lang-
ford, Anal. Chem. 22, 1792 (1976).

"C. R. C. Handbook of Chemistry and Physics", (R. C.
Weast, ed.), The Chemical Rubber Company, Cleveland,
1969-1970, pp. B-174 to B-l77.

A. T. Zander, T. C. O'Haver, and P. N. Keliher, Anal.
Chem. 22, 1166 (1976).

T. R. Blakeslee, "Digital Design with Standard MSI
and LSI", Wiley-Interscience, NY, 1975.

D. R. McGlynn, "Microprocessors", Wiley-Interscience,
NY, 1976.

B. Soucek, "Microprocessors and Microcomputers",
Wiley-Interscience, NY, 1976.

A Barna and D. I. Porat, "Introduction to Micro-
computers and Microprocessors", Wiley-Interscience,
NY, 1976.