1111mmnlmm‘mmmmn lL

32193 00671 4707

This is to certify that the

dissertation entitled
Design, Characterization, and Applications of

a Miniature Continuous Flow Analysis System
presented by
Charles J. Patton

has been accepted towards fulfillment
of the requirements for

Ph . D . Chemistry

degree in

 

 

WKW

 

Major professor

Date December 1, 1982

 

MS U L! an Affirmative Action/Equal Opportunity Institution 0-12771

PLACE IN RETURN BOX to roman m- ohockout from your record.
TO AVOID FINES rota-n on or More on. we.

DATE DUE DATE DUE DATE DUE ||

9K W ’5‘? I

If”

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmdlvo Action/Equal Opportunity Initiation
emu»:

DESIGN. CHARACTERIZATION. AND APPLICATIONS OF A

MINIATURE CONTINUOUS FLOW ANALYSIS SYSTEM

by

Chas. J. Patton

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemistry

1982

Gzaoei’j

ABSTRACT
DESIGN, CHARACTERIZATION, AND APPLICATIONS OF A
MINIATURE CONTINUOUS FLOW ANALYSIS SYSTEM
by

Chas. J. Patton

A.modu1ar, single channel. continuous flow analysis instrument
that can be configured for either air-segmented continuous flow analy-
sis (CFA) or flow injection analysis (FIA) has been developed. System
components include a modified, commercially available. peristaltic
pump, a dual-beam, fiber optic. filter photometer, and an electronic
bubble gate that removes the air segment artifact from the detector
signal when.bubble-through flow cells are used for colorimetric CFA
determinations. Instrumentation and hardware designed and built for
use with this system are described in detail.

Performance of the miniature CFA system (0.1 cm ID manirold com-
ponents and bubble-through flow cells) is characterized for a number of
equilibrium based colorimetric determinations. Novel open tubular cad-
mium reactors (OTCRs) were develOped and characterized that were used
to reduce nitrate to nitrite in conjunction with automated colorimetric
determinations of nitrate in water and seawater. Peaks with flats and
less than 1% interaction were achieved in conjunction with OTCRs at a

l

sampling rate of 120 hr- . Comparable performance in conjunction with

packed bed cadmium reactors generally used for automated nitrate

determinations could only be achieved at half this sampling rate
(60 hr'l).

Results from an experimental comparison of the performance
(sampling rates, sample and reagent consumption, and precision of
analytical results) of CFA and FIA are presented. Reagent consumption
and sample dispersion in the miniature CFA system were less than in FIA
systems equipped with either coiled Open tubular reactors “L05 cm ID)
or single head string reactors. 'The advantages of using single bead
string reactors for merging zones FIA determinations are demonstrated.
CFA and FIA are shown to be complementary techniques, and the relative

merits of each for various applications are discussed.

Fedor Mikhailovich Dostoevski. the Russian novelist. said one time
that. 'One sacred memory from childhood is perhaps the best education.’
I can think of another quickie education for a child. which, in its
way. is almost as salutary: Meeting a human being who is tremendously
respected by the adult world. and realizing that that person is

actually a malicious lunatic.

Slapstick -- Kurt Vonnegut

ii

ACKNOWLEDGMENTS

I am extremely indebted to Professor Stanley Crouch for his
guidance, enthusiasm, and friendship throughout the course of my
research effort. Many thanks also to Professor Chris Enke who served
as my second reader. His advice was consistently helpful and incisive.
Who else would have known that BALD. stands for Bradlee, Vorhees. and
Day? This information is now pinned discretely to my B.V.D.'s. The
efforts of the other members of my Guidance Committee. Professors Tom
Pinnavaia and George Leroi. are also appreciated. It was also a
pleasure to teach for Professor Andrew Timnick. who allowed me to
incorporate a continuous flow analysis experiment into his Chemistry
333 course.

A great deal of assistance (32; necessarily constructive) was pro-
vided by my buddy and fellow sufferer. Eugene Ratzlaff - AKA: AZURE
FLEETFANG. Together we managed to advance the 'Boy's Shop' approach to
analytical instrumentation to new levels of perversity. Thanks. too,
to Robert Thompson who shock-tested early versions of my instrument.
You will be happy to know, Rob. that the leering grin of 'Bow tie Man'
still haunts my troubled dreams. Sincere thanks are also extended to
Keith.Trischan.who wrote several computer programs that were extremely
useful in my work. The friendship and help of other Crouch Group

members will not be forgotten.

iii

The help. good humor (ice cream?), programming expertise. and down
home philosophy of Dr.‘Tom Atkinson were invaluable. Thanks. too, to
Marty Babb for help with the design of electronic components of the
miniature continuous flow analysis system. This work would have been
impossible without the craftmanship of Russ Geyer, Ben Stutsman, Dick
lenke. and Deak Watters in the Chemistry Department Machine ShOp. Andy
Seer and Manfred Langer in the Chemistry Department Glass Shop. and Ron
Haas and Scott Sanderson in the Chemistry Department Electronics Shop.

Special thanks to JO Iotarski who prepared the figures and

drawings in this dissertation and to Margy Lynch for technical typing.

iv

FOREWORD

This dissertation is more like a collection of short stories with
a common theme, than a fully deveIOped novel. With the exception of
Chapter 1 where the origins and principles of air-segmented continuous
flow analysis (CFA) and flow injection analysis (FIA) are discussed.
the chapters are arranged in the approximate order that research
reported was performed. As work progressed new lines of research began
to suggest themselves. I have thus included more FIA experiments than
I had planned at the onset of this project.

In Chapter 2 the miniature continuous flow analyzer that evolved
during the course of this research is described in detail. Here I have
attempted to provide sufficient information to allow others to use the
system as it now exists and to modify it for new applications. Also
included are procedures for data reduction and analysis.

A description of the electronic bubble gate that removes the air
segment artifact from the detector signal when.bubble-through flow
cells are used with air segmented continuous flow analyzers can be
found in Chapter 3. Details of individual circuit elements in the
bubble gate are included. ‘The results of experiments in which this
bubble gate was used in conjunction with a commercial CFA system and
the miniature CFA system described in Chapter 2 are also presented. and
the performance of the latter is compared with performance reported for

flow injection analysis.

The developmental work on open tubular cadmium reactors for rou-
tine nitrate determinations in water and seawater reported in Chapter 4
is the culmination of an idea that occurred to me in 1976 after I had
returned from an ACS summer symposium on immobilized enzymes. The
major obstacle to beginning this work was locating a source of small
diameter cadmium tubing. Apparently it is used in the nuclear reactor
industry. I had worked on this project sporadically for several years,
but until about six months ago, the successful outcome of this project
was still very much in doubt.

The experimental comparison of PIA and CPA presented in Chapter 5
resulted from some lively, late-night discussions with several FIA
enthusiasts in attendance at the ACS summer symposium on flow injection
analysis held in New York in 1981. Upon returning from this symposium.
I realized that the miniature continuous flow analysis system which can
be easily configured for either FIA or CFA determintions was ideally
suited for an unbiased comparison of these two techniques. Please note
that the results of this comparison pertain only to equilibrium based
colorimetric determinations. and they should not be extrapolated to
include determinations in which the formation of concentration
gradients can be exploited (titrations. viscosity measurements, refrac-
tive index measurements). for which I believe FIA has some clear advan-
tages over CPA.

Finally, in Chapter 6. some future applications for the miniature

continuous flow analysis system are discussed.

vi

CHAPTER

TABLE OF CONTENTS

FOREWORDO O O O O O O O O O O O O O O O O O O O O O O O O O O v
TBLE 0F FIGURESO O O O O O O O O O O O O O O O O O O O O O O x
LIST OF Tux-IE8 O O O O O O O O O O O O O O O O O O O O O O O O xv

HISTORICAL AND THEORETICAL INTRODUCTION TO CONTINUOUS

FLOW ANALYSIS.. . .. . .. . .. . .. . .. . .. . ..

A. Overview. . . . . . . . . . . . . . . . . . . . . . . .

B. Historical. . . . . . . . . . . . . . . . . . . . . .
l. CFA. . . . . . . .
2. FIA. . . . . . . . . . . . . . . . . . . . . . . .

C. Principles. . . . . . . . . . . . . . . . . . . . . . .
l. CFA. . . . . . . . . . . . . . . . . . . . . . . .
2. FIA. . . . . . . . . . . . . . . . . . . . . . . .

O O
He 0
‘OQQUINNHH

INSTRUMENTATION, DATA ACQUISITION AND PROCESSING AND
OPERATIONAL PROCEDURES. . . . . . . . . . . . . . . . . . . .25
. General Considerations - CFA. . . . . . . . . . . . . . .25
. Pump. .. . .. . .. . .. . .. . .. . .. . .. .29
. Air Segment Phasing. . . . . . . . . . . . . . . . . . . 31
. Detector. . . . . . . . . . . . . . . . . . . . . . . . .34
Bubble Gate. . . . . . . . . . . . . . . . . . . . . . . 37
Data Acquisition. . . . . . . . . . . . . . . . . . . . .41
Data Processing Techniques. . . . . . . . . . . . . . . .41
1. Data Reduction. . . . . . . . . . . . . . . . . . . .41
2. Graphical Procedures for Estimation of b and ot. . . 42
H. Manifold Components. . . . . . . . . . . . . . . . . . . 49
1. CFA. . . . . . . . . . . . . . . . . . . . . . . . . 49
2. FIA Manifold Components. . . . . . . . . . . . . . . 52
I. Plumbing. . . . . . . . . . . . . . . . . . . . . . . . .53

PHMUOG>

DESIGN AND CHARACTERIZATION OF AN ELECTRONIC BUBBLE
GATE FOR AIRrSEGMENTED CONTINUOUS FLOW ANALYSIS. . . . . . . 60
A. Overview. . . . . . . . . . . . . . . . . . . . . . . . .60
B. Experimental. . . . . . . . . . . . . . . . . . . . . . .62
1. CF Systems. . . . . . . . . . . . . . . . . . . . . .62
2. Determination of Nitrite and Silicate. . . . . . . . 63
a. Nitrite Reagents. . . . . . . . . . . . . . . . .63
b. Nitrite Standards. . . . . . . . . . . . . . . . 63

vii

CHAPTER

D.

E.

PAGE

c. Silicate Reagents. . . . . . . . . . . . . . . . 63

d. Silicate Standards. . . . . . . . . . . . . . . .64
3. Bubble Gate. . . . . . . . . . . . . . . . . . . . . 65
Results and Discussion. . . . . . . . . . . . . . . . . .70
1. Introduction. . . . . . . . . . . . . . . . . . . . .70
2. Steady State Experiments. . . . . . . . . . . . . . .71
3. Sample Interaction. . . . . . . . . . . . . . . . . .77
4. Mixing Effects. . . . . . . . . . . . . . . . . . . .79
Analytical Results. . . . . . . . . . . . . . . . . . . .79
1. Nitrite Determinations. . . . . . . . . . . . . . . .79
2. Silicate Determinations. . . . . . . . . . . . . . . 80
Conclusions. . . . . . . . . . . . . . . . . . . . . . . 84

NOVEL CADMIUM REACTORS FOR DETERMINATION OF NITRATE IN
WATER AND SEAWATER BY SEGMENTED. CONTINUOUS FLOW
COLORIflmY O O O O O O O O O O O O O O O O O O O O O O O O O 88

A.
B.
C.

A.
B

C.

Overview. . . . . . . . . . . . . . . . . . . . . . . . .88
General Considerations. . . . . . . . . . . . . . . . . .89
Summary of Nydahl's Work. . . . . . . . . . . . . . . . .92
1. pH Effects. . . . . . . . . . . . . . . . . . . . . .92
2. Dissolved Oxygen Effects. . . . . . . . . . . . . . .93
3. Effect of Chloride. . . . . . . . . . . . . . . . . .94
Experimental. . . . . . . . . . . . . . . . . . . . . . .94
1. Reagents. . . . . . . . . . . . . . . . . . . . . . .94
2. Standards. . . . . . . . . . . . . . . . . . . . . . 95
3. Cadmium Reactors. . . . . . . . . . . . . . . . . . .96
4. Manifolds. . . . . . . . . . . . . . . . . . . . . . 98
5. pH Measurements. . . . . . . . . . . . . . . . . . . 98
Some Definitions. . . . . . . . . . . . . . . . . . . . .98
Preliminary Experiments. . . . . . . . . . . . . . . . .101
1. Open Tubular Cadmium Reactors. . . . . . . . . . . .101
2. Cadmium Foil Reactors. . . . . . . . . . . . . . . .110
Final Experiments. . . . . . . . . . . . . . . . . . . .112
Discussion. . . . . . . . . . . . . . . . . . . . . . . 120

EXPERIMENTAL COMPARISON OF CFA.AND FIA.. .. .. .. ..122
Introduction. . . . . . . . . . . . . . . . . . . . . . 122
Experimental. . . . . . . . . . . . . . . . . . . . . 124
1. Dye Dispersion Experiments. .. . .. . .. . .. . 124
2. Chloride Determinations. . . . . . . . . . . . 131
Discussion. . . . . . . . . . . . . . . . . . . . . . . 138

FUTURE APPLICATIONS O O O O O O O O O O O O O O O O O O O O O 14 1

A.
B.

C.

D
E.

Overview. . . . . . . . . . . . . . . . . . . . . . . . 141
Automated Sample Pretreatment and Pre- and Post-Column

Reactors for Liquid Chromatography. . . . . . . . . . . 141
Continuous Flow Kinetics Determinations. . . . . . . . .143
Active Open Tubular Reactors. . . . . . . . . . . . . . 144
Extended Data Acquisition and Processing Systems. . . . 145

viii

CHAPTER

LIST OF REFERENCES.

ix

FIGURE

1-1

1-2

1-5

TABLE OF FIGURES

PAGE

Diagrams of generalized CFA and FIA instruments.

8 I sampler, SV I sampling valve. L I sample loop.

DB I debubbler. D I detector. PC I flow cell.

36 I bubble gate, CR I chart recorder. A I air,

RIreagent. A)CFA. B)FIA................8

Signals recorded for the same flow stream with bubble
gated and debubbled flow cells. . . . . . . . . . . . . . . 10

Expanded scale plot of smoothed and regenerated
interaction test patterns shown in Figure 1-2. . . . . . .14

Variation of a as a function of n. d .
andeort=300 2s. A)FI0.0083 mLs
1: dt:0.2cm. 2: ddtIOJ cm.
3: dltI 0.05cm. _?)d I01 cm. 1: Fl
I0.0167mLs . 2: FIO.OO83 mLs .
3: F I 0.004 mL s 1. Minor variables:

8. 9 x 10- poise, 15 I c33 dyne
m . D' 25 I 5 x 101.. . .. . .. . .. . .. 17

0.3

Dispersion in FIA as a function of injected sample

volume (8 ) and reactor length.(L). Carrier stream I

pH 9.5 borate buffer. sample I 10 a! phenol red in

pH 9.5 borate buffer. flow rate I 2.0 mL min-1. coiled

Teflon reactor diameter -=.O5 cm. A) L I 50 cm.

8': 1 I 25 uL. 2 I 50 uL. 3 I 100 uL. 4 I 200 uL.

5I 400uL. 6- 800 uL. B)Sv I50uL.

L: 1 I 10cm. 2 I 50 cm. 3 I 100 cm. 4 I 200 cm. .. . . 21

Methods for phasing air injection with pump pulsations.

A) Dual pump tube method. B) Solinoid valve method.

C) Schematic diagram of solinoid valve controller.

ICl I LM311 comparator. IC2 I 96LS02 dual monostable
multivibrator, D1 I germanium diode. 01 I 2N3904 NPN
transistor. All resistances in ohms. . . . . . . . . . . . 33

FIGURE

2-2

2-4

2-6

2-7

2-8

PAGE

Schematic representation Of the dual-beam, fiber Optic,

filter photometer designed for the mCFA system.

S I source. F0 I randomized. bifurcated fiber optic

bundle. H I fiber optic/flow cell holder. D I dove-

tailed slide, PC I flow cell. M I flow cell/interference
filter mount. F I interference filter, PD I photodiode.

I/V I operational amplifier current-to-voltage

converter. A I second stage amplitier.. .. .. .. .. ..35

Schematic diagrams of detector circuits. A) Photodiode

(PD) and current-to-voltage converter. B) Second stage
variable-gain inverting amplifier. ICl and IC2 I LF351

FET input operational amplifiers. A11 resistances in

ohms. Power supply bypass capacitors omitted for

clarity. .. . .. . .. . .. . .. . .. . .. .. .. ..38

Detailed schematic diagram of the bubble gate designed
for the mCFA system. ....................39

Schematic diagram Of a modified circuit to trigger and

clear the monostable multivibrator of the bubble gate.

ICl I LF351 Operational amplifier. IC2,IC3 I LM311

voltage comparators. IC4.IC5 I SN74OO quad 2-input

NAND gates. .All resistances in ohms.. . .. . .. . .. . 4O

Extraction of b from rise curves of CFA peaks. Bubble

gated flow cell: VI smoothed At points. VI smoothed
(Ass-At) points. Debubbled flow cell: 4;: smoothed

At points. AI smoothed (Au-At) points. . . . . . . . . . .46

Graphical estimation of “t by means of a cumulative
probability plot.. . .. . .. . .. . .. . .. . .. . . 48

Manifold components designed for the mCFA system.

A I air inlet, R I reagent inlet. S I sample inlet.

E I epoxy cement. G I glass. A) sample inlet blocks:

left. used for plastic coils. Right. used for glass

coils. B) Reagent addition tees: left. used for

plastic coils. Middle and right used for glass coils.

C) Flow stream debubblers. .. . . .. . . .. . .. . .. .51

Manifold components for FIA. C I carrier stream inlet.

M I to manifold. W I to waste, S I sample inlet.

L I sample loop. A) Six-port rotary valve. B) Dual

four-way slider valve. C) Single four-way slider valve.

D) Reagent addition tee.. . . .. . . .. . . .. . . .. .54

xi

FIGURE

2-10

2-11

3-1

3-3

PAGE

Methods for interconnecting various manifold components

used for continuous flow analysis. T I teflon tubing.

P I pump (not shown). LS. I inner plastic sleeve.

0.8. I outer plastic sleeve. MC I mixing coil.

R I reagent inlet. G I gripper fitting. C I standard

HPLC connector. 8 I plastic sleeve.. . .. . .. . .. .. 56

A) Preferred method of sample introduction for FIA

systems. C I carrier stream. R I reagent. S I sample.

W I to waste. B) Pecked sample introduction used for

CFA systems. S I sample. W I wash solution.

IAS I intersample air segment. (3 Effect of intersample

air segments on dispersion that Occurs in the sample

line of CFA systems. . . . . . . . . . . . . . . . . . . . .57

Fluctuations in the detector signal caused by the

successive passage of liquid and air segments through

the flow cell: Position 1. air segment completely

within cell. Postion 2. air segment exiting cell.

Position 3. cell completely filled by liquid segment.

Position 4, air segment entering cell. . . . . . . . . . . .66

Generalized schematic diagrams of bubble gate circuitry.
All resistances in ohms. all capacitances in microfa-
rads. A) Basic bubble gate. 1C1 I LF311 FET input
comparator. IC2. 74LS123 dual monostable multivibrator,
IC3 I LF398 sample-and-hold amplifier. IC4 I AD755
logarithmic amplifier. IC5 I TL084 quad FET input
Operational amplifier. B) Differential edge sensor.

IC5 I TL084 quad FET input operational amplifier.

1C6 I LF311 FET input comparator. C) Sample-and-hold
amplifier update logic. 1C7 I 7410 triple 3-input NAND
gate. D) Automatic threshold adjustment for comparator.
ICl. IC8 I TL084 quad FET input operational amplifier. . . 67

Composite diagram of various CF manifolds used for

nitrite determinations. ‘Values outside or within

parentheses refer to AAII or mCF manifolds.

respectively. . . . . . . . . . . . . . . . . . . . . . . . 72

Results of percent steady state 0588) experiments:

1. standard AAII CF system. 2. bubble gated AAII CF

system. 3. bubble gated mCF system without peeked

sampling. 4-6. bubble gated mCF system with peeked

sampling and n equal to 1.5. 2.0. and 3.0 s-l.

respectively. . . . . . . . . . . . . . . . . . . . . . . . 74

xii

FIGURE

3-5

4-5

PAGE

Effect of segmentation frequency and liquid flow rate on

SSS with the residence time of the sample slug in the

mCF system held constant at 110 s: l. n.I=1.5 s-l.

F .. 0.0067 mL .‘1. 2, n - 3.0 .‘1. F .. 0.013 mL .‘1. . . ..76

Recording of 2. 6. 10. 14 and 18 “M nitrite standards
run in ascending and descending order. followed by an
interaction test pattern and replicate determinations of
thel6 a! standard. A) 120 samples hr-l. B) 360 samples

hr O O O O O O O O O O O O O O O O O O O O O O O O O O O O 82

Diagram of the manifold used for silicate determinations
'ith the “CF 'y‘temO O O O O O O O O O O O O O O O O O O O O 83

Recording of 5. 15. 25. 35 and 45 UM silicate standards

run in ascending and descending order followed by an
interaction test pattern and replicate determinations of

the 15 u! :fandard. A) 120 samples hr-l. B) 180

samples hr . . . . . . . . . . . . . . . . . . . . . . . . 86

Diagrams of manifolds used to determine nitrate and

nitrite with the mCFA instrument. A) System for Open

tubular and foil cadmium reactors. B) System for packed

bed cadmium reactors. . . . . . . . . . . . . . . . . . . . 99

Percent recoveries of nitrate and nitrite. A) Open
tubular cadmium reactor. B) Packed bed cadmium reactor. . 106

Effect of chloride concentration on percent recoveries
of nitrate and nitrite with Open tubular cadmium
re‘ctor‘O O O O O O O O O O O O O O O O O O O O O O O O O .109

Peaks recorded for seawater spikes with a 60 cm open

tubular cadmium reactor. A) Nitrite spikes. Nominal
concentrations from left to right: blank. 0.5. 1.5. and

2.5 9!. B) Nitrate plus nitrite spikes. Nominal
concentrations of N02- + N03- from left to right:

0.5 + 5.0. 1.5 + 15.0. 2.5 + 25 ufl. Sample and wash

times were 23 s and 7 s. respectively. . . . . . . . . . . 117

Peaks recorded for seawater spikes with a packed bed

cadmium reactor. A) Nitrite spikes. Nominal

concentrations from left to right as in Figure 4-4.

B) Nitrate spikes. ‘Nominal concentrations from left to

right: 5.0. 15. and 25 pg. Sample and wash times were

40 s and 20 s. respectively. . . . . . . . . . . . . . . . 118

xiii

FIGURE

4-6

5-2

PAGE

Correlation of data obtained with open tubular and
packed bed cadmium reactors. O I seawater spikes.
II== distilled water standards. . . . . . . . . . . . . . .119

Composite diagrams of manifolds used to compare FIA and

CFA. A I air. S I sample. C I carrier. R I reagent.

A) Manifold used for dye dispersion experiments. B) FIA
manifold for chloride determinations. C) mCFA manifold

for chloride determinations. . . . . . . . . . . . . . . . 125

Curve tracings of peaks from dye dispersion experiments
with Open tubular reactors. single head string reactors,
and air segmented reactors. . . . . . . . . . . . . . . . .129

Peaks recorded for chloride determinations with the
single line FIA manifold. . . . . . . . . . . . . . . . . .133

Peaks recorded for chloride determinations with a
merging zones FIA manifold. .A) 50 cm open tubular
reactor. B) 50 on single head string reactor. . . . . . . 134

Peaks recorded_ior chloride determinations at a sampling

rate of 360 hr . A) FIA with merging zones and a 25 cm
single bead string reactor. B) mCFA. . . . . . . . . . . .136

xiv

TABLE

1-1

2-2

2-3

3-1

3-2

3-3

3-4

LIST OF TABLES
PAGE

Evolution of Techniconfs air-segmented continuous flow
‘n‘lyzersO O O O O O O O O O O O O O O O O O O O O O O O O O 4

Theoretical values of o as a function of F and (1t
calculated with Snyder's model for t I 300 s. . . . . . . . 26

Minimum values for at as a function of F and d ‘when
limitations in n imposed by flow cell volume are taken
into ‘ccountO O O O O O O O O O O O O O O O O O O O O O O O 28

Segmentation frequencies and delivery factors as a
function of speed control settings on the modified
Brinkmann IP-12 peristaltic pump. . . . . . . . . . . . . . 30

Flow rates for standard pump tubes at various speed
control settings of the IP-12 pump. . . . . . . . . . . . . 32

Data for the first 10 s of the rise curve for the steady

state peak shown in Figure 1-2. This data set pertains

to the bubble gated flow cell. . . . . . . . . . . . . . . .44
Data for the first 10 s of the rise curve for the steady

state peak shown in Figure 2-1. This data set pertains

to the debubbled flow cell. All other notation is shown

in Table 2-5 O O O O O O O O O O O O O O O O O O O O O O O O 45
Major variables for SSS experiments. . . . . . . . . . . . .75

Percent intOIOCtione O O O O O O O O O O O O O O O O O O O O78

Comparison'between mCFA and FIA for colorimetric nitrite
detemin.tion‘O O O O O O O O O O O O O O O O O O O O O O O 81

Performance of mCFA system for colorimetric silicate
detemin.tion‘O O O O O O O O O O O O O O O O O O O O O O O 85

XV

TABLE

5-3

5-4

PAGE

Percent recovery of nitrite and nitrate standards for
copperized OTCRs of various lengths. . . . . . . . . . . . 102

Linear regression parameters and percent recoveries

calculated for nitrate and nitrite standards as a

function of pH. . . . . . . . . . . . . . . . . . . . . . .104
Linear regression parameters and percent recoveries

calculated for nitrate and nitrite standards with

imidazole buffers and a 30 cm copperized OTCR. . . . . . . 108

Effect of segmentation gas on the pH measured for the
effluent from a 60 cm OTCR. . . . . . . . . . . . . . . . .111

Data from determinations of nitrate and nitrite in
distilled water and seawater with OTCRs and PBCRs. . . . . 113

Linear Regression parameters for data listed in Table

4-5O O O O O O O O O O O O O O O O O O O O O O O O O O O O 114

Sample volume as a function of sample time for dye
dispersion experiments. . . . . . . . . . . . . . . . . . .126

Reactor lengths used for dye dispersion experiments. . . . 126
Precision of dye dispersion experiments. . . . . . . . . . 127

Comparison of mCFA and FIA: Colorimetric chloride
deterMInatiOn-SO O O O O O O O O O O O O O O O O O O O O O O13?

xvi

CHAPTERl

HISTORICAL AND THEORETICAL INTRODUCTIW TO

CONTINUOUS FLOW ANALYSIS

AL Overview

Until about 1975 the term continuous flow analysis was used almost
exclusively to describe a technique now known as air-segmented con-
tinuous flow analysis (CFA) that was invented by Leonard Skeggs [1] and
developed commercially by the Technicon Corporation under the trade
name. AutoAnalyzer. For a number of years an air segmented analytical
stream was considered a prerequisite for continuous flow analysis. but
in the mid-1970's two promising experimental approaches toward prac-
tical nonsegmented continuous flow analysis were reported almost simul-
taneously by Kent Stewart and coworkers in the United States [2] and
E10 Hansen and Jaromir Ruzicka in Denmark [3]. Despite pronounced
differences in hardware. the basic concept for nonsegmented continuous
flow analysis was essentially the same for both groups. Ruzicka and
Hansen named this technique flow injection analysis (FIA) by which it
is commonly known today.

There have been a number of fairly thorOUgh reviews of CFA [4-6]
and FIA [7-10] which I shall not attempt to duplicate here. Instead I
shall focus on the strengths and weaknesses of the two techniques in
relation to each other. To do this. it is necessary to place the de-

veIOpment and theoretical foundations of CFA and FIA into historical

perspective. since there is little question that both these elements
have had a profound effect on the manner in which these two techniques

are currently perceived.

g; Historical

1... 93

Details concerning Skeggs invention of CFA and its ensuing early
history are somewhat aprocryphal and inexorably linked to the Technicon
Corporation. It appears that both the increased demand for clinical
laboratory services and the technological boom that followed World War
II provided Skeggs with both motivation and inspiration to automate
routine colorimetric determinations. He dismissed the possibility of
building a machine to manipulate the test tubes and pipets used for
conventional batch analysis as being overly complex and impractical
[1]. Instead. he began to explore the feasibility of performing assays
sequentially in a continuously flowing stream contained within a net-
work of small diameter open tubes (now called a manifold). At some
point in the late 1940s or early 1950s. Skeggs began to build prototype
CFA systems and devised novel solutions to problems posed by the CFA
approach. He found. for example. that dialysis was particularly adap-
table to CFA. and he used it to separate low molecular weight analytes
from blood and serum proteins that interfere in many assays. His most
significant innovation. however. was unquestionably 'the bubble'. In-
termixing of samples as they passed through the manifold in boxcar
fashion severely limited the performance of Skeggs' early systems. He

soon discovered. however. that insertion of air bubbles into the

analytical stream at regular intervals reduced this mixing dramatically.
Thus the bubble became and remains the 3125 pg; 323 of CFA.

In the mid 1950s. the Technicon Corporation bought and patented
Skeggs' CFA concept and prototype instrumentation. and in 1957 they
marketed a single channel CFA system. the AutoAnalyzer. which was the
first in a series of highly successful and profitable CFA instruments.
It must be appreciated that virtually none of the hardware that CFA
practicioners presently take for granted was available when Skeggs
began his developmental work. Reliable multichannel peristaltic pumps.
and more importantly. durable and inert pump tubes that could deliver
relatively small volumes with good precision had to be developed and
manufactured in quantity. The same is true for the multitude of
specialized manifold components such as dialysis blocks. mixing coils.
and fittings that now can be obtained from a number of suppliers.
Technicon must be credited with developing and constantly improving the
hardware and technology that transformed CFA from a laboratory curiosi-
ty into a highly practical and sOphisticated technique.

While a typical present-day CFA instrument is much the same as the
first AutoAnalyzer with respect to the basic components and their
general arrangement in the system. the hardware itself and the way in
which data are acquired and processed has now evolved through three
major design changes as summarized in Table 1-1. Several comments are
in order here. First. CFA was develOped by Technicon primarily for
clinical determinations and in general. new technology has been aimed
at this market. While first generation CFA instruments were general

purpose systems that were readily adapted for non-clinical

.ema-m_m .mfl .eeon Luz .Eo_em-:oumcmzv .eOcu .=.Lu .: .Lommmepmzuo: A.e.n .mmmoxme

.mcwmmououn mam cemuwmwzcom
mums pounceou .mucoconaoo myomficme
mam .mssg .pofiasmm mo cufimomou ucmmz

.mcomum0wgmmm ”newcwfluuco: you xfimumewum
.Epom Hoccmco OHmCMm cw zuomoccuou <2m

.m-ou 30am .mucocomsoo
naomwcme .massa mo :uwmomop Home:

.ucoespumcw oaxuouopm m.pOmmmuum;oo:
mam mmuoxm mo ~cvoe newuoncOLQ

.upmcu camcwm a co

umpuouou mcofiumcfiau0uom fifim pom mufismo:
.muo>:o o~nm>oe mam Houoouop o~mcmm
.xuuoewpofiou >9 evapoficu mam ommxoMm
:ospmu MILuOE0uoca osmfim xp ESwOOm vcm
ssfimmmuoa mo m:0wum:flsuouom mzoocmunsewm

.cowouuw: mug: voofin mce
mmoozfiu mo mCOMHmCMEuouop msoocmufinswm

.mofiasmm _m0mcw~u cw coucpuw: mop: coo—n
no omoonfiw mo :ofiumcmsuouom pom mocuflmon

 

mucceeou

i III l:.9.l(ll..1 I'.‘ .1.-- (l1.l(.‘.( ..1.! -.1 A O 'I'.'I 'y' I In- lll'i'lfli

omH DNA u<2m .I mum“
Eoumxm :oMumpocoo thank

 

 

 

 

ce-o~ H CH emsxnm=<ous< .u Damn

oo N2 oe\NH <2m .u “can

mEm mem COM HGHOCOD vcooow

cm NH om\NH <2m .u comm
Neconxfimc< Hoccmco

ON w -mpmsz ogxuouoee .: eemfl
punxmmc<

om e muxmoeuuonm .: seam

om N eoxsnmc<oos< seesaw .. mean

cm L H Lessnme<oe=< .. ammn

use; no; mfioccmcu mEoumxm newumuocoo amp“;
mcowumcmeuouoo mo 902532

(1...!

.mpcmxfimcm zofiw msoscwucoo voucoewom-uflm m.:ooM::ooH mo mewu=~o>m .H-~ ofinmh

applications. a trend toward specialized multichannel systems that
perform standard determinations at a single work station is apparent in
the second generation SMA 12/60 (sequential. multichannel analyzer. 12
channels. 60 samples per hr) and in the third generation SMAC (sequen-
tial multichannel analyzer with computer) system. Second. Technicon
has tended to reserve its best technology for its large. clinically-
oriented customers. Although SMA technology was made available to non-
clinical customers in the form of the AAII single channel instrument.
state-of-the-art SMAC technology in single channel form is still
unavailable.

.2... EIA

The origins and early history of FIA have recently been reviewed
by Kent Stewart [12] and Horacio Mottola [13]. Stewart traces the FIA
concept-sample injection into a continuously flowing stream with con-
tinuous downstream detection-~back to James and Martinis original work
on gas chromatography in 1952 [14]. Mottola sets the clock somewhat
later. at Blaedel and Hicks' [15] develOpment of a continuous flow
kinetics instrument in 1962. FIA in its present form. however. did not
appear until the mid 1970s and evolved through two distinct pathways
which have now merged for the most part into a single technique con-
taining elements of both original approaches. FIA as developed by
Stewart. Beecher. and Hare [2] has very obvious connections to the non-
segmented. continuous flow. post column LC reaction colorimeter
designed by Spackman. Stein. and Moore [16] for amino acid analysis.
The carrier (reagent) stream was propelled through 0.025 cm ID teflon

tubing at relatively high pressure. and samples were introduced through

a multiport sampling valve. It is unfortunate that much of this
group's early work is buried in rather obscure literature sources.
Ruzicka and Hansenfls first apparatus on the other hand seems much more
related to CFA systems. A peristaltic pump was used to propel the
carrier stream through much larger (0.15 cm ID) polyethylene tubing.
Samples were literally injected. at first directly through the walls of
the tubing and later through a septum. with a hypodermic needle and
syringe. Clearly FIA. was not invented by a single group. but Ruzicka
and Hansen deserve a tremendous amount of credit for their developmen-
tal efforts. They coined the generic term for nonsegmented continouous
flow analysis. recognized the great potential of the technique.
developed many novel applications for its use. and continue to promote
it with unflagging enthusiasm.

At the time FIA was being develOped. the hardware required was
readily available from a number of suppliers. By the early 1970s.
peristaltic and reciprocating LC pumps were for the most part perfected
and a wide variety Of small diameter polymeric tubings from which FIA
reactors are usually fibricated were readily available. SO too were
low dead volume connectors and fittings designed primarily for use with
LC systems. In addition. the absence of air segments simplified inter-
facing the flow stream to a variety of detectors. Part of the great
appeal of FIA. especially in small industrial and academic labora-
tories. is the ready availability of hardware and the ease with which
the FIA technique can be adaptd to existing equipment. In addition.
FIAfs lack of association with commercial concerns (until quite
recently) has led to freer and faster dissemination of information and

technological advances.

The sampling rates achieved with relatively simple single channel
FIA.systems often surpass those of third generation CFA systems. In
general. high (>200 hr-l) sampling rates are possible only when the
flow rate is high and the residence time of samples in the reactor is
short. ‘This was particularly true for early systems that were often
wasteful of reagents and limited to very fast chemical reactions. As
sample dispersion in FIA systems became better understood. however.
these shortcomings have been minimized. The trend in FIA has been
toward lower flow rates. smaller diameter reactors. and smaller sample
volumes. Multiport valves have gained almost universal acceptance as
the sample introduction method. Innovations in FIA hardware and

methodology continue to be reported at a rapid pace.

9; Principles

LEE;

The spatial arrangement of interconnectd modules for a generalized
CFA instrument is shown in Fig. 1-1A. It is common practice to
aspirate samples sequentially from containers positioned on the sampler
into the manifold where chemical separations and reactions are ef-
fected. Usually an analyte-free wash solution contained in a reservoir
on the sampler is aspirated between each sample. The sampler probeis
residence time in a sample or the wash solution. t‘ or t'. respective-
ly. and the flow rate of the pump tube to which it is connected deter-
mine the volume of sample drawn into the manifold. Failure to control
these variables rigidly. therefore. can adversely affect the precision
of analytical results. Sample and wash slugs are uniformly segmented

with air bubbles as they enter the manifold and are thus divided into a

A) pump WASTE

2:: <3— .m\ IQ)”
i \..\' —~ :

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l
I
l __.
.. B
B)
PUMP
R WASTE L
l} é [
_ WASTE ——3—
FC D

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 1-1. Diagrams of generalized CFA and FIA instruments.
sampler, SV I sampling valve, L = sample loop, DB I debubbler,
detector, PC = flow cell, 86 = bubble gate, CR = chart recorder,
air, R I reagent. A) CFA. B) FIA.

>60:
uu

number of nominally identical subunits. The segmented stream is pro-
portioned with reagents at various points along the manifold as it is
propelled toward the flow- through cuvet of a recording photometer.
The highly reflective air segments pose special problems for photo-
metric detection in CFA. In first and second generation systems air
segments are removed from the analytical stream just prior to detec-
tion; in third generation systems the air bubble artifact is removed
electronically. a technique known as bubble gating.

In the absence of dispersion. the trace observed at the recording
photometer of a CFA instrument would resemble a train of square waves.
Transitions to or from some time invariant signal level proportional to
the analyte concentration in the wash solution or a sample (base line
or sample gggggy ;;;;53) would be instantaneous. In the trace actually
observed. however. these transitions are more gradual and steady states

are approached along skewed rise and {all gurves shown in Fig.1-2.

 

Data for this figure were obtained experimentally with the miniature CF
system that was developed during the course of this research and they
will be discussed in more detail in paragraphs that follow. and in
Chapter 2. Deviation of these recorded signals from the idealized
square wave output function is quite obvious. Data sets plotted to the
right and left Of the dashed vertical line in this figure were obtained
with bubble gated and debubbled flow cells. respectively. Note that
steady state was fully attained in the broad center peak (t‘I30 s). but
that steady states were not achieved at shorter sample and wash inter-
vals (t'I8s. t'I2s) which were used to generate the peaks on either
side of the center peak. The sequence of peaks to the right of the

center peak is known as an interaction test pattern. Here samples were

10

.mHHoo 30am wofinnzsom was coumu o~£nzn saw: Ecouum zofiw 05mm ozu pom OOUHOOOH mfimcwmm .NIH whzwwm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

$va?
:9% 03m Oxm me c—m cm? om? cm? ca 00 on AV
_.- _ _ m _ _ _ _ _ _ _ _ 0.0
_ _:ll+¥ _ fi/J QJ _ k _ _ d
_ 1.. No
_ 11 Go V
_ ”.350 m.
: .jE I: to o
_ J
an
_ NEDU J] 0.0 w
_ mm... o
_ 11 ed a
_ J] 50
E g _ 94pm .i we
zamCE 5m: >9qu I ..
20.54%? _
i no
.38 38... _ .38 33...

owkdu MJmmDm OBmQmeo

11

determined in the concentration sequence of low. high. low and it is
apparent that the tail of the high-concentration peak contributes
significantly to the height of the second low-concentration peak.
Percent interaction (‘11) is defined as the difference in absorbance
between the second and first low-concentration peaks divided by the
absorbance of the high-concentration peak. This quantity multiplied by
100 is the percent interaction. The extent of sample interaction
depends on the magnitude and nature of dispersion experienced by
samples as they pass through the flow system.

Dispersion in CFA arises from two distinct processes. The first
is large scale convective mixing (longitudinal dispersion) that occurs
in unsegmented zones of the system--e.g.. sample lines. flow cell
debubblers. flow cells. Here flow is essentially laminar and. as a
result. recorded signals are deformed exponentially [17] relative to
the idealized square wave output function. An empirical linear rela-
tionship between concentration and time for longitudinal dispersion is

given by Equation l-l.
ln<Ass-At) I 1n Ass-t/b (l-l)

where Ass is the steady state absorbance. At is the absorbance of the
peak profile at any time t. and b. the exponential factor. is the lepe
constant. Equation l-l. due to Walker» 21,31; [18] is a formalized
version of an equation first reported by Thiers. gt _a_}_,, [19]. It is
customary to express the magnitude of longitudinal dispersion in terms
of b. Walker [17] showed that because of the exponential relationship
between sbsorbance and time. peak.heights withinl90. 95. and 99% of

steady state result when ts I 2.3b. 3.0b. and 4.6b. respectively. and

12

that percent interaction of 7. l. or 0.5 7. will occur when the time
between samples (t‘+t') is 2.65b. 4.6b. or 5.3b. respectively. This of
course assumes that longitudinal dispersion is much greater than axial
dispersion as is the case for first and second generation CFA systems.

The second effect. axial dispersion. occurs in the air segmented
analytical stream. Here large scale mixing between liquid segments is

prevented by air bubbles. while mixing within segments is greatly en-

 

hanced by so-called bolus flow [20]. The only medium for intersegment
mass transfer in the segmented stream is the stagnant liquid film that
wets the walls of the flow system. By this mechanism. analyte mole-
cules initially contained in a single segment are incorporated into
succeeding liquid segments that were initially analyte free. After an
arbitrary period of flow. the concentration of analyte molecules in any

segment is approximated by Equation 1-2.

Ck/c = e-qqk/kl (1-2)

where C is the analyte concentration in the initial undispersed seg-
ment. Ck is the analyte concentration in segment k (kI0.l.2...). and q
is a dimensionless parameter which in simplest terms is the ratio of
the liquid volume that wets the walls of the entire flow system (Vf)
and the volume of a M liquid segment (V‘). Note that the expres-
sion on the right hand side of Equation 1-2 is the Poisson distribu-

tion. When the sample is a series of analyte containing segments. as

 

is the usual case in CFA. the concentration distribution is given by
the summation of Equation 1-2 and for a large number of segments the

cumulative Poisson distribution is approximated by the cumulative

13

Gaussian distribution. Therefore axial dispersion is generally expres-
sed in terms of the standard deviation. o. of the cumulative Gaussian
distribution. In this case q is equal to the displacement of the 50%
maximum concentration point in the dispersed sample profile from the
leading edge of the undispersed sample slug (kIO). and c : (q)1/2.
Note that both a and q are dimensionless in the sense that they are
measured in terms of segment number. Equation 1-2 was the first model
of axial dispersion in CFA and is due to Hrdina [21]. The trouble with
this and several other models that followed it. most notably those of
Begg [22]. Thiers [23]. and Walker [24]. is that they did not relate
axial dispersion to relevant experimental variables--e.g.. residence
time. flow rate. segmentation frequency. tube diameter. For a more
detailed discussion of axial dispersion in CFA see reference 25 and the
figures therein.

Longitudinal dispersion can be minimized by improved design of
system components--e.gu.'pecking' samplers [26]. bubble gated flow
cells [27-29] - or its effects can be removed after the fact by means
of’a data processing technique known as curve regeneration that was
first reported by Walker [30]. Here the final steady state absorbance
is calculated before it is physically attained by means of Equation
1-3.

dA
A88 = At-l-b —£ (1-3)

dt
Details of curve regeneration are discussed in Chapter 2. The
effectiveness of bubble gating and curve regeneration is illustrated in

Figures 1-2 and 1-3. respectively. In both figures curve tracings to

14

.N-H assume an

czozm mcaoupm; umOu sewuumpoucm emumpocouou ace vocuoosm mo uofia ommum moccmmxm .muu ouzwau

Amvorceh

 

 

 

eouoqlosqv

 

 

 

 

 

 

 

 

5.40 4.50 5.40 5.40
ombumwzwowm owIBOEm owkdmwzwwwm othOOZm

 

1.1.8 3017.. _ JJwU 301......
owhdw mgmmam . _ 03mm3mmo

15

the right of the dashed lines pertain to the bubble gated flow cell.
Dispersion is very obviously less in the bubble gated flow cell. Peak
heights more nearly approach steady state and interaction is reduced.
The value of b estimated for the bubble gated and debubbled system were
1.0 s and 2.3 s. respectively. This information was used to regenerate
the bubble gated and debubbled data sets and the curves that resulted
are shown in Figure 1-3. Smoothed and regenerated data are plotted
side by side for each data set to facilitate comparison. Note that the
effects of longitudinal dispersion are greatly reduced for data sets
obtained with either flow cell and that the peak profiles closely
resemble the integrated Gaussians predicted by Hrdina's model.

Axial dispersion is fundamental to CFA as it is currently prac-
ticed because flow systems fabricated from wettable materials are
required for hydraulic stability in segmented streams. In 1976 Snyder
and Adler [25] experimentally verified Hrdina's model and presented a
semiempirical derivation of it. The unique feature of their derivation
was that Vf was expressed in terms of the length and diameter of the
flow system. This allowed q (and therefore a) to be related to experi-
mental variables. They then extended the model to account for slow
mixing between segments and the liquid film [20]. Previous models had
assumed instantaneous mixing. Later Snyder [31] recast the extended
model into a form that made the effects of major experimental variables

on axial dispersion more obvious as shown in Equation 1-4.

:1/3

 

 

r538d’{’(F + 0.92d3n)"3n’” + 1/ ] [2.3507 + din)"3 77
= n

2

7’ FD,” 7 Fd.3

16

Here °t is the standard deviation of the peak expressed in seconds. dt
is the internal diameter of the flow system in cm. F is the liquid flow

rate in mL s-1

. n is the segmentation frequency in Hz. t is the resi-
dence time of the sample in the flow system in s. n is the viscosity Of
the liquid in poise. 7 is the surface tension of the liquid in dyne
cm”. and Dw.25 is an empirical diffusion coefficient (cm2 s-l) that
pertains only to diffusion in coiled tubes [31]. The basic assumptions
of the model follow: 1) air segment volumes are the minimum required
to totally occlude a tube of a given diameter (711/24 (it3 3 0.92 :13);
2) the flow system is perfectly wetted; 3) longitudinal mixing in the
film is negligible: and 4) the flow system is free of mixing effects
(longitudinal dispersion). Generally the experimenter has control over
only dt' n. F. and t. although it should be noted that a surfactant is
generally added to reagents to avoid changes in 7 that would otherwise
occur as the concentration of the analyte varied [24].

It is clear from Equation 1-4 that at is proportional to the
square root of t. The effect of the other major variables (F. dt’ n)
is not so apparent. but they can be readily visualized with the aid of
Figure 1-4A and 1-4B. IPart A of this figure shows the relationship
between at and n for several values of dt when F and t are fixed at
0.5 mL min"1 and 300 s. respectively. and values for minor variables
are in keeping with a similar set of calculations by Snyder [32]. Here
we see that for each value of dt' at passes through a minimum. As (It
decreases. °t decreases while the value for n where at is minimum
increases. Part B of this figure shows the relationship between 6t and

n for several values of F when (It is fixed at 0.1 cm and all the other

variables are the same as for Figure 1-4A. Note that the minimum value

17

a
.7m «50 m1o~xm .- 3 3o .73 0:3 an n r
638 AZ: .2: r. : $023.5, 8&2 .7... a... 285 1 n. 2.. .7m .2... 38... n I ~

 

 

.7... .E 385 a a z .3 1c 1 he S .3 8.: u we 3 .8 1c 1 an .N .5 «a a he 3
.H m a... 386 u m 2 .m can u a new "— 28 .um .: mo 530:3 a me an we .8339; .7” 95m:
AFImV xucmacmfi cozoucmccomm Clmv Iocmncmcu co_LoLcmEomm
060— 06' 0.. —.O 0.8, 06' O.— pd
Twit.“ x "xix“ I "xIII “ 9o TAXI“ I “31:“ “ “xflxn I 90
4| 0; ll 0—

(S) ‘o

 

 

D
‘.'
)
s
1i o.n ( 1! on
m
N
— lo. 113
N
— L
11 on Io...

Am 3.

18

for ot decreases as F decreases and also that these minima occur at
lower values of n.

These trends explain. in part. the relative performance of Tech-
nicon's second and third generation CFA systems. The values of F. n.
and :1t for second generation systems are about 1.5 mL min-1. 0.5 {-1.
and 0.2 cm. while for third generation systems they are about
0.6 mL min-1. 1.5 s-l. and 0.1 cm. If we assume t I 600 s and retain
the same values for minor variables as before. values of at calculated
for second and third generation systems are 3.82 s. and 2.57 s.
respectively. Snyder [32] has suggested.that for peaks with apprec-
iable steady state intervals (flat tops) the time between samples
should be about 8 at. Therefore the predicted sampling frequencies for
second and third generation systems are about 120 hr”1 and 175 hr-1
respectively. Second generation systems generally are limited to a
maximum sampling frequency of about half the theoretical limit because
of longitudinal dispersion in the sample line and in flow cell debub-
blers. In third generation systems the sampling frequency (150 hr.1)
approaches the theoretical limit more closely because longitudinal
dispersion is minimized with peeked sampling and bubble-through flow
cells. Although there is no theoretical limit to the extent that axial
dispersion can be decreased by reduction of dt and F. there are several
practical constraints imposed by flow cell volumes. precision and
delivery rates of pumps. and other factors. Snyder takes these into

account in references 31 and 32. and I shall discuss them in more

detail in Chapter 2.

19

2.. _IA

The arrangement of components for a generalized FIA instrument is
shown schematically in Fig. 1-1B. A nonsegmented carrier stream (the
reagent for the simplest case) is pumped continuously through a poly-
meric tube that terminates at a1detector which.for the sake of sim-
plicity we shall assume to be a recording photometer. Samples are
abruptly introduced into the carrier stream either by means of a
syringe or (usually) a multiport valve. The sample mixes with the
carrier stream enroute to the detector. For systems with a sampling
valve the amount of sample introduced into the reaction network is
determined solely by the sample loop volume. Provided that the 1o0p is
completely filled prior to injection. the residence time of the sample
line in a sample does not affect the precision of analytical results.
Furthermore. the low dead volume of the sample line. valve. and sample
loop. and the nonwettable materials from which they are made generally
preclude the need for a wash solution. A slug of air aspirated between
successive samples is usually sufficient to purge remnants of the
previous sample and the carrier stream from the sample line and sample
lOOp. respectively.

At the moment of injection. a sample with initial concentration Co
is entrained by the carrier stream which propels it through the reactor
toward the detector. To a first approximation. sample molecules in the
center of the reactor flow toward the detector with twice the mean
velocity of the carrier while those at the wall are stationary (i.e..
laminar flow). Convective mixing of the sample with.the carrier occurs
along the stream lines and dilutes the sample so that the1maximum

sample concentration. C sensed at the detector is some fraction of

20

C The ratio of Co to cmax provides a useful empirical measure of

o'
dispersion. D [33] in FIA systems. Ruzicka and Hansen classify disper-
sion in FIA as limited (D I l to 3). medium (D I 3 to 10). or large
(D 2 10). If the flow rate and reactor diameter are held constant. the
injected sample volume. Sv. and the reactor length. L. have a major
influence on D. which approaches unity either as Sv inreases or L
decreases. as illustrated in Figure l-5A and 1-5 B. Further details
can be found in the figure captions. So far in our discussion we have
only considered dispersion of the sample by convection in a straight
tube. Under these conditions. skewed peaks with infinitely long tails
would be predicted. Tailing of FIA peaks is much less severe than
this. however. due to radial diffusion and secondary flow that we shall
now briefly discuss.

Sample molecules in stagnant zones near the reactor wall are
transported into more rapidly flowing stream lines by radial diffusion.
and thus tailing is reduced. At very low flow rates. in fact. radial
diffusion becomes the dominant contributor to dispersion. and under
these conditions a Gaussian peak profile results. Peak shapes actually
Observed in the normal range of operating conditions for FIA. however.
generally reflect the combined effects of convection. radial diffusion
and. if the reactor is coiled (the usual case for FIA). secondary flow.
Vandersl ice 5; 5;. [34] presented numerical solutions to convection-
diffusion equations that allow calculations of dispersion in straight
Open tubes under a variety of experimental conditions. The problem
with such models. however. is that they fail to account for secondary
flow--f1ow perpendicular to the axis of the reactor resulting from

centrifugal forces-that develops in a coiled tubes. All other factors

21

on
meoe
n%J
/
«J
m oN
n.
o.

 

 

 

 

 

   

 

 

 

 

.Eu com u v .Eo oo~ u m .50 cm H N
:5 S "H 5 :3 em 1 >m E :5 cow .1. e :3 8e .1. m :3 SN 1 e :3 2: u m :3 cm 1 N
.4: mm H H ">m .50 cm a A< .Eelmc. u mouoemwm pouumou commoh cofiwou .~1:«E as o.~ n
ouch zofiw .hommsn Campos m.m :9 CM mop fiocozm 2: o“ u ofiasmm .uomwzn Campos m.m :2 u seekum uofluumu
.nmv caucofi heuueoa mam n>mv oss~o> o~c5mm m0u00wcfi mo comuocsm m we <~u cm :OMmuommMa .m1H opzwwu
Amv NEE. 35.2....
mm Omvm m. N. w 0 Nvmm0m ¢N m. N. w 0 ,
u u q )r I i q d d d - 0.0
1 Qmu .
v .._.o . 1_.o
. m ow- N
.. V O
8 3 l V
m“ 1 . MW nwu . on
NO% / .NOMW
«J
l 1 . nu
wH ‘m.uN1 m” 1 8
m w
— 1M.O 10.0%
1 0.1
Am -vo v .Io
o . m 0
[[1111 U [[Ill'llmo o.—l[[OU[[[l [[[[[[ 0.0

 

 

 

 

22

being equal. dispersion in a tightly coiled tube is about 2 to 3 times
less than in a straight tube [5.35]. The extent of secondary flow
increases as the coil diameter approaches the inside diameter of the
tube and as the mean velocity of the carrier system increases. Note
that secondary flow actually stabilizes the laminar flow regime [36] so
that there is no physical basis for the notion of‘incipient turbu-
lance' in coiled Open tubes.

It is also possible to minimize dispersion in FIA systems by using
packed rather than open tubular reactors. More care must be taken in
the preparation of packed reactors. especially if small diameter parti-
cles are used. and the pressure drop may be too great to allow a peris-
taltic pumping system to be used. Recently. however. FIA reactors have
been reported in which a small diameter tube is packed with solid glass
spheres with diameters only slightly less than that of the tube [37].
The pressure drop across these so-called single head string reactors
(SBSR) or pearl string reactors are only slightly greater than that of
an open coiled tube of comparable diameter. and dispersion in an SBSR
is 2 to 3 times less than that of tightly coiled Open tube. Also the
extent of dispersion in SBSRs is less affected by changes in the flow
rate then are coiled Open tubular reactors [38]. This is an obvious
advantage for FIA.because reagent consumption decreases with decreasing
flow rates. Furthermore. the lower dispersion obtained with SBSRs
should allow longer reaction times with less dispersion.

SO far we have only considered dispersive effects in the reactor.
The total dispersion of the system. however. is also dependent on
dispersion that occurs in the injector and detector. which is seldom

negligible. Chemical kinetics. too. greatly influence FIA peak heights

23

and widths. If the FIA system is used only to transport the sample to
a detector. limited dispersion is desirable so that even though the two
ends of the sample plug experience mixing with the carrier. there
remains a central undispersed zone. This is clearly not acceptable
when it is desired to measure the end product of some reaction between
the sample and the carrier stream. Here the entire sample volume must
mix with the reagent and remain in the reactor for a sufficient period
of time for the reaction to proceed to an appreciable extent prior to
detection. Medium dispersion systems are generally used in these cases
and a compromise must be reached between adequate mixing and reaction
time on one hand and excessive dispersion of the sample zone on the
other. If the sampling rates generally associated with FIA (90-150
hr’l) are to be maintained. maximum residence times are on the order of
30 s. With the advent of SBSRs. however. the maximum residence time
might be extended to 60-90 s. if several samples can reside in the
reactor at a given time with minimal interaction.

At present there is no general model for FIA that allows disper-
sion to be calculated 5 2512;; from experimental variables such as tube
ID. flow rate. and residence time. Betteridge provides a good summary
of dispersion in FIA based on the tanks in series model [39] in Table
II of reference 7. It must be noted. however. that at present only
trends can be predicted. This is because subtle differences in the
geometries of sample injectors and flow cells. as well as differences
in wall roughness in reactors and connectors can profoundly influence
the total observed dispersion [40]. Also. because there is a gradient
both axially and radially within the dispersed sample zone. measurement

of some property within this zone will always result in a composite

24

average of concentration distributed across or along the reactor. This
is entirely different from CFA where each segment is essentially homo-
geneous. Nonetheless. for a given FIA system. dispersion generally is
highly reproducible and therefore the precision of analytical results
is good. Furthermore for a number of applications such as kinetic
assays. titrations. viscosity measurements. and other techniques where
a gradient can be exploited. FIA offers possibilities that are not

easily achieved by CFA.

CHAPTER 2
INSTRUMENTATION. DATA ACQUISITION AND PROCESSING.

AND OPERATIONAL PROCEDURES

A. General Considerations - CFA

In Chapter 1 the origins of dispersion in CFA systems were
discussed largely from a theoretical point of view. It is now
worthwhile to examine limitations to this theoretical treatment that
are imposed by hardware used to implement real CFA systems. Returning
to Snyder's model and assuming that the dwell time. t. is fixed at 300
s and that the values of minor variables [32] are assigned as follows:
7 I 32 dynes cm-1;n I 8.9 x 10"3 poise; Dw.25 I 5 x 10-5 cm2 s'l. it is
a simple matter to calculate the magnitude of axial despersion. °t' as
a function of liquid flow rate. F. segmentation frequency. n. and
manifold inside diameter. dt' Results of such calculations. shown in
Table 2-1. reveal that for each value of dt and F. there exists a value
of n (nopt) where at is minimum. Two trends are observed in this

table. First. increases as dt decreases. and second for any fixed

nopt
value of dt' as F decreases. nopt decreases continuously. Note.
however. that “t at first decreases as F decreases. but at some point
further reduction in F causes at to increase. The minimum Ot value
(otmin) and the associated value of nopt for each value of dt are

underlined in Table 2-1. Note the steady decrease in “t that occurs as

dt and F decrease. This trend is without theoretical limit. In

25

26

Table 2-1. Theoretical values of (It as a function of F and dt

calculated with Snyder's model‘2 for t = 300 s.

 

d = 0.2 cm d = 0.1 cm d = 0.05 cm d = 0.025 cm

 

 

 

 

 

 

 

 

 

 

 

 

 

t t t t
I:1 (“L 5-1) nopt O1: r_‘_op_t_ at nopt Ct _n_op_§_ or
50.0 1.31 2.94 5.0 2.84 19.0 3.08 70 3.60
25.0 1.00 2.46 4.2 2.13 16.4 2.14 61 2.38
16.7 0.90 2.29 3.7 1.87 14.9 1.77 56 1.90
8.3 0.60 2.16 2.9 1.58 12.4 1.35 48 1.33
4.2 0.45 2.15 2.2 1.44 9.80 1.10 40 0.98
1.7 0.25 2.31 1.4 1.42 6.8 0.96 30 0.74
0.8 0.15 2.53 0.9 1.48 5.0 0.93 24 0.65
0 4 0 08 2 80 0 6 1.60 3 3 0 96 17 0 62

 

 

aValues for minor variables: Y = 32 dyne cm-1

2 ; n = 8.9><10-3 poise;
Dw 25 = S><10'5cm1 5-1. Underlined values indicate minimum n and

at values at each value for dt'

27

practice. however. the range over which n. dt' and F can be varied is
quite limited due to restrictions imposed by the pumping system and
flow cell volumes.

In CFA the multichannel peristaltic pumps used to aspirate samples
into the manifold. and to proportion samples with air and reagents. are
far from pulseless. Pump pulsations can cause nonuniform proportioning
that is manifested as noise on the steady state portions of recorded
signals. This problem is minimized if air segments are added in phase
with pump pulsations. that is. at the time each roller leaves the
platten surface. Under these conditions, however. n is restricted to a
fairly narrow range of values that is determined by the pump speed and
the number of pump rollers. The range of allowed segmentation frequen-
cies is further restricted if bubble-through flow cells are used.
because in order to measure the absorbance of individual liquid seg-
ments. the flow cell volume (Vc) must be considerably less than the
liquid segment volume (V‘). Flowcells with volumes of about 2 pl. (path
length 8 1.0 cm. ID - 0.05 cm) are commercially available and so the
minimum value of V‘ can be conservatively set at about 4 pl. (twice Vc).

At any flow rate. the maximum segmentation frequency (11 ) can be

max
determined by dividing F by V‘. If the values for at are now calcu-
lated assuming these restrictions, results shown in Table 2-2 are
obtained. In this table values in parentheses are for u when nopt<nmax'
The minimum axial dispersion is achieved when at 8 0.1 cm,with values
for F in the range of 8.33 to 4.17 pl. (1 (0.50 to 0.25 ml. min-1).
Choice of the higher flow rate is advantageous because. as shown in

Chapter 1. dispersion due to mixing effects decreases as the flow rate

increases. other things being equal. The experimental CFA system that

28

Table 2-2. Minimum values for at as a function of F and d1: when

limitations in n imposed by flow cell volume are taken

into account.

 

 

 

 

 

 

 

 

 

 

dt = 0.2 cm dt = 0.1 cm dt = 0.05 cm
-1 -l b

F (UL s ) nmaX (5 l 0t (S) Gt (5) 0t (S)
50.0 (3.333“ 12.5 2.94 (1.3)0 2.84 (5.0) 3.12
25.0 (1.67) 6.2 2.46 (1.0) 2.13 (4.2) 2.34
16.7 (1.11) 4.2 2.29 (0.9) 1.82 (3.7) 2.12
8.3 (0.55) 2.1 2.16 (0.6) 1.62 2.02
4.2 (0.28) 1.0 2.15 (0.45) 1.61 2.13

1.7 (0.11) 0.4 2.31 (0.25) 1.78 2.41

0.8 (0.05) 0.2 2.53 (0.15) 1.98 2.69
aValues in parentheses indicate F in units of mL min-1.

Minimum segment volume = 4 uL = 2 VC.
cValues in parentheses are for n when n <11 .

opt opt max

29

I shall now describe was designed to operate within this Optimum range

of experimental variables.

L. L092

The lodel IP-12 variable speed. 12 channel peristaltic pump
(Brinkmann Instruments. Westbury. NY) used for all work reported here
was modified locally by replacing the standard roller assembly (8
rollers) with one containing 16 rollers. This pump is equipped with a
two decade (00-99) digital control that allows linear adjustment of the
motor speed in the range of 0 to 13.2 RPM in 0.13 RPM increments. At a
speed control setting of 42, nominal and experimentally determined flow
rates for Technicon SMA Flow-Rated pump tubes agreed to within about
10%. Table 2-3 lists segmentation frequencies and delivery factors
associated with commonly used speed control settings.

The pump was further modified as follows. An optical chopper with
16 blades (one in line with each roller) was attached to one end of the
roller assembly and an opto-interrupter (positioned so that the emitter
and detector were on opposite sides of the chopper) was mounted on the
pump frame. The signal from the 0pto-interrupter provided timing
pulses that were used to synchronize injection of air segments with
roller lift-off. Also. notched metal end blocks were attached to the
pump housing on Opposite sides of the roller assembly. These allowed
pump tubes to be stretched (slightly) across the roller assembly. When
pump tubes were held under constant tension in this way. flow rate
stability was improved.

Technicon SMA Flow-Rated pump tubes were used in all experiments.

A light spray of silicone lubricant was applied to the tubes on a

30

Table 2-3. Segmentation frequencies and delivery factors as a
function of speed control settings on the modified
Brinkmann IP-12 peristaltic pump.

 

 

 

 

Segmentation
Frequency
Speed Control Setting (5'1) Delivery Factora
14 0.5 0.33
28 1.0 0.67
42 1.5 1.00
56 2.0 1.33
70 2.5 1.67
84 3.0 2.00
98 3.5 2.33

 

aFactor by which nominal pump tube flow rate must be multiplied to
estimate actual flow rate.

31

weekly basis. Previous experience has shown that this practice
approximately doubled tube life. A list of nominal flow rates for pump
tubes at several different pump speeds can be found in Table 2-4. The
'COLOR CODE' heading in this table refers to the color of two plastic

shoulders bonded to each pump tube that identify the nominal flow rate.

£34 A}; Segment Phasing

Two methods were used to add air segments in phase with pump pul-
sations. In the first method that was originally reported by Habig. g;
_a_l. [27]. two pump tubes (A and B) are connected as shown in Fig. 2-1A.
Tube A acts as a compressor for tube 3 which is connected to the air
inlet port of the manifold. Inside tube B. uniform segments of com-
pressed air are trapped between rollers and when the forward roller
leaves the platten. this compressed air rapidly expands into the mani-
fold. For 0.1 cm ID manifolds. best results were obtained when the
nominal flow rates of tubes A and B were 0.05 and 0.10 mL min—1.
respectively. I developed an alternate method for air phasing that
used information from the optical chopper (described in the previous
section) to actuate a normally closed miniature solenoid valve (LIF
valve, Lee Company. Westbrook. CH). The inlet of the valve was con-
nected to a source of low pressure gas (a pump tube or a cylinder of
compressed gas); the outlet was connected to the air inlet port on the
manifold. as shown in Figure 2-1B. The T11. pulse train from the opto-
interrupter triggered a monostable-multivibrator which generated a
delay pulse that was adjusted to terminate at the moment of roller
lift-off. The falling edge of the delay pulse triggered a second

monostable that generated the pulse used to actuate the valve. The

32

Table 2-4. Flow rates for standard pump tubes at various speed
control settings of the IP-12 pump.

 

Nominal Flow Rate (mL min-1)

 

Pump Speed Control Settings

 

 

 

 

Technicon
Number Color Code _14_ 2_8 42‘z _s_o__ _7_o_ _8_4_
116-0549P01 orn-blk .005 .010 .015 .020 .025 .030
02 orn-red .010 .020 .030 .040 .050 .060
03 orn-blu .017 .033 .050 .070 .083 .100
04 orn-grn .033 .067 .100 .130 .167 .200
05 orn-yel .053 .107 .160 .210 .267 .320
06 orn-wht .077 .153 .230 .310 .383 .460
07 blk-blk .107 .213 .320 .430 .533 .640
08 orn-orn .140 .280 .420 .560 .700 .840
09 wht-wht .200 .400 .600 .800 1.00 1.20
10 red-red .267 .533 .800 1.07 1.33 1.60
11 gry-gry .333 .667 1.00 1.33 1.67 2.00
12 yel-yel .400 0.80 1.20 1.60 2.00 2.40
13 blu-blu .533 1.07 1.60 2.13 2.67 3.20
14 grn-grn .667 1.33 2.00 2.67 3.33 4.00

 

aStandard speed.

33

.meco cw moocmumwmoh
~n< .ooomnmcooo zoz vommzm n he .oooflo sancosnom u no .oooooon>nonss onooomocoe “use Nommom u Non

.poumumaeoo -m24 n fium .po-opu:oo o>~m> wfiocwfiom mo Empumwn odomeocom flu .vonuoe o>~m> wfiocwfiom mm

.vozuoe oasuxrsxhfimso A< .m:0wuwmfisa can; cu“: cofluooncw haw ucwmmcm pow mwocuoz .Hum opswwu

 

 

 

     

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o 1
.o 8 o. 0.0 mutawwpwwu
.. n I I I.
r 1 1 #N w1 1 _
e . _
> .. _ _
_ _
r I II I
.66 com
>~$w
8
123a Q35.”—
m m
10%|.— m 1PT.. m
> m 7.4
5.4 4
Am 2

 

 

 

34

circuit diagram for the valve controller is shown in Fig. 2-1C. This
method of air injection has two advantages. First. a source of
compressed gas external to the pump can be used to segment the analyti-
cal stream. Thus the number of pump channels required for a given
manifold can be reduced. Second. if a variable-modulus counter is
added to the circuit. air segments can be added in phase with with
every second. third. etc.. roller lift-off. The latter feature would
be particularly useful for teaching excercises. For routine work.
however. I used the dual pump tube method of air segment phasing

because it was simple and reliable.

Q; Detector

A schematic diagram of the dual beam. fiber optic photometer
designed for the miniature continuous flow analysis system is shown in
Figure 2-2. Models 03000 and 03100 miniature tungsten-halogen lamps
(Welch-Allyn. Skanetelles Falls. NY) were used interchangably as the
light source. Both lamps have identical voltage and current require-
ments (3.5 V at 0.75 Amp). but the model 03100 has a flame formed lens
that approximately doubles the radiant energy impinged on the fiber
optic bundle. These lamps are rated for only 20 hours of operation at
3.5 Volts. but lifetimes in excess of 150 hours were obtained when the
lamps were operated at about 3a3‘volts. as was customary.

An EK-15 randomised. bifurcated fiber Optic bundle (Dolan-Jenner.
'oburn. NA) made from either glass or quartz fibers was used to
transmit light from the source to the front windows of the flowcells.
The fiber optic-flowcell interfaces are rectangular. black delrin

blocks that slide along dovetail grooves machined in the base plate Of

35

03:35
oumum vacuum u < .pouco>:oo ommuHo>lounu:OHn=o nowmflfiasa Hacoflunnomo n >\H .owOMvOuosm n on .nouamw
oocopomnoucw n u .ucsos noufifim ooconomnoucfi\fifiou 30am u 2 .fifioo zofiu n Um .Ouflfim vofiwmuo>ov n a .hovH0:

HHoo zofim\omuao gondw n : .oficcsn owuao honfim umpmupswfln .vonwfioucmh n Om .OUHSOm n m .Eoumxm <uue
osu pow vocuwmov houoeouozm houfifim .ofiuao Honwm .Ewon-~m:w ecu mo :ofiuwucomOHQOH ofiumsozow .NnN casuam
\\VV\ .N\\\\\ \\

 

  

wwflmmrnflflwwnmldLJhnW%Ww

2 on
wgwmmumm
B&§HHM.H h.H.uM
@5330
.1 s. o...
m4a2<w

 

 

 

 

 

RRRR\\\\\

 

36

the photometer. Flowcells with saphire windows. a 1.0 cm pathlength.
and an internal diameter of 0.05 cm were Obtained commercially (PN 178-
B724-02. Technicon Instruments. Tarrytown. NY). Generally the
reference channel was used only to compensate for source drift and in
these cases a 1.5 cm x 0.2 cm diameter delrin rod. center bored with a
0.05 cm hole. replaced the reference flowcell. Wavelength discrimina-
tion was accomplished with 1.27 cm diameter. narrow bandwidth (:8 nm).
3-cavity interference filters (Ditric Optics. Hudson. NA). Threaded.
circular mounts machined from black delrin hold the interference fil-
ters in place centered directly in front of the photodiodes. The front
faces of these mounts were center drilled to allow insertion of the
rear flowcell end cap to a depth of about 0.2 cm. This arrangement
allows for easy installment and removal of flowcells and interference
filters. and insures prOper optical alignment.

Silicon photodiodes with an operational range of 200-1200 nm (30V-
0408. E.G.5 6.. Electro-optics Division. Princeton. NJ) were used as
detectors. These were operated without external bias to minimize dark
current. Photocurrent was converted to voltage with simple Operational
amplifier current-to-voltage converters that had fixed transfer
functions of about 2 x 108 VIA. The photodiodes and current-to—voltage
converters were shielded from other circuit elements by a metal cover.
It should be noted that these photodiodes have a maximum response at
about 900 nm. and interference filters with good near IR blocking
properties must be used to avoid stray light problems. Variable gain
(2-100) second stage amplifiers invert and amplify signals from the
current-to-voltage converters. The gain Of each amplifier is adjusted

with a lO-turn potentiometer. These were mounted on top of the photo-

37

meter housing and permitted the sample and reference signals to be
balanced. usually at a level of about 5 volts. This type of electronic
balancing provided an effective and much simpler alternative to
mechanical slits. Sample and reference signals are connected to the
inputs of a log-ratio amplifier (AD755. Analog Devices. Norwood. NA).
This device provides an output of 1.0 volt per absorbance unit. Base
line drift of the sample and reference channels are approximately 10%
per hour for the first hour and about l-2$ per hour thereafter. but as
expected. drift of the ratioed signal is negligible over the same time
intervals. Sample. reference. and log-ratio outputs are made acces-
sible through BNC connectors mounted on the rear panel of the photo-
meter housing. Refer to Figure 2-2 for further details. Schematic
diagrams for the photometer"s first and second stage amplifiers are

shown in Figure 2-3.

g. Bubble Gate

 

 

The electronic bubble gate that removes the air segment artifact
from the detector signal when bubble-through flow cells are used with
the photometer has been described previously [29]. A detailed sche-
matic diagram of this bubble gate is shown in Figure 2-4. Note that a
second sample-and-hold amplifier was added to this circuit so that both
transmittance and absorbance (log-ratio) signals could be acquired
simultaneously. A full description of this device Operating over a
wide range of experimental conditions can be found in Chapter 3. A
different circuit that can be used to trigger and clear the buble
gate's monostable multivibrator is shown in Figure 2-5. This circuit
appears to eliminate the need to autorange the comparator threshold

level (see Chapter 3. 3.3. and Figure 3-2). but is has not been

38

A) 5”:

 

 

 

 

200M
e
5k
3 ) 1T)
lOk 5 IOOO F “6' 2-38
I
V ' “-ISV
IOOpF
a) ,L
I 20k 50k
I' +|5v 1k -
, e
> - 6 I“\ SIGNAL
3 £52 5 ~ .1 OUTPUT
4 g,
t
v O’BV
Figure 2-3. Schematic diagrams of detector circuits. A) Photodiode

(PD) and current-to-voltage converter. B) Second stage variable-gain
inverting amplifier. ICl and IC2 = LF351 FET input operational
amplifiers. All resistances in ohms. Power supply bypass capacitors
omitted for clarity.

39

HTP ‘7

 

 

. lOk
OFFSET

 

 

 

 

 

 

 

 

 

 

LOW CLIP
LEVEL

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LO - Chip 0
2. Circuits on us indicated:

@ 1108‘:

® LF398N

@(7) LFSIIN

© SN74L5123N

© snmon
3. All PS. Bypass Cups on .0471”:

Figure 2-4. Detailed schematic diagram of the bubble gate designed
for the mCFA system.

 

40

.mszo :« moozmumfimou HH< .moumm az<z usmcwum vmsv oovnzm n
mu~.vuw .muoumcmgeoo oumu~o> Hfimzq u mu~.muH .howmflfimem fimCOWHNHQQO Hmmmg u ~UH .Oumw ofinnsn ecu mo

houmhnfi>wumsa o_£mumo:OE ozu nwofio vcm youuflhu ou umsohmo voflmwpoe m we Eupumwu Umumsozom .mnm ohzwmm
do 00.
>n.o+
..oo. .
.:

e
+
‘ non

o a m.)
.3 .w
. m n
3350202 0» c x: .m
> —___‘-Ju- e >mr+
& >m+
O
m 1.

 

:89”: .1de0

 

m
0
.—
0
.lm

m

C

E a only

ml. Go 1
v. . >9-

m. c.
m.-. ole
fl .1. E
v. . .
o

 

 

 

 

 

 

>m ._ o F 1
+ >m+ :
. ) e NB
—1 _ u. m
>38 50ng . n
ie m
>o_+

41

characterized fully at present.

II”

Data Acquisition

Data were recorded with a Model SRrZSS 10 in. strip chart recorder
(Heath Company. Benton Harbor. MI). In many experiments. data were
logged simultaneously with a microcomputer. FORTH based software rou-
tines written by Eugene Ratzlaff [41]. in this laboratory. acquired
data points (voltage and time of acquisition) on command from the
update pulse of the bubble gate. These data were temporarily stored in
RAM and later shipped to an LSI 11/23 minicomputer (Digital Equipment

Corp“, Maynard. NA) for further processing.

I."

Data Processing Technigues

 

L. M Reduction

When data are recorded on a strip chart recorder. data reduction
is a simple matter of reading and recording the steady state absorbance
value Of each peak. Before analytical results can be calculated. how-
ever. the baseline absorbance . which is equivalent to the reagent
blank in batch colorimetry. must be subtracted from the peak absor-
bance. Under usual sampling conditions (t‘)t') the signal level does
not return fully tO baseline between samples. and the baseline absor-
bance for each peak is estimated by linear interpolation between the
initial and final baseline absorbance values. Baseline drift due to
instability of electronic circuits and the light source is less than 1%
per hour when the mCF system is Operated in the dual beam mode. Larger
drifts. however. can result from other factors such as build up of
particulate matter on flow cell windows. instability of reagents. or

partial blockage of pump tubes. If the refractive index of samples is

42

significantly different from that of the wash solution. a small frac-
tion of the total absorbance measured for a sample is due to refractive
index effects. The contribution Of refractive index to the absorbance
measured for a sample can be quantitated if the normal reagent solution
is substituted with one from which the reagent necessary for chromo-
phore formation has been omitted. When samples are redetermined under
these conditions. the baseline corrected absorbance of the resulting
peaks is due only to refractive index changes. Corrections are made by
subtracting these values from the peak heights measured under standard
conditions.

2; Graphical Procedures £25 Estimation 23 b 52d at-

As discussed in Chapter 1. two constants. b and at. provide mea-
sures of longitudinal and axial dispersion. respectively. and therefore
describe for the performance of a given CFA system. Both these
constants can be extracted from rise or fall curves by simple. graphi-
cal procedures. The data plotted in Figure 1-2 will be used to provide
working examples of these procedures. Data for this figure were
Obtained experimentally with a manifold similar to the one shown sche-
matically in Figure l-lA. In this system values for n. dt' F. and t
were 2 s-l. 0.1 cm. .014 ml. s-l. and 85 s. respectively. Samples were
aqueous phenol red solutions and the reagent was borate buffer (pHle)
containing Brij-35 surfactant. The analytical stream flowed sequen-
tially through two identical flow cells (1.0 cm x 0.05 on ID). The
first flow cell was bubble gated‘while the second was debubbled, and
data (time and voltage) from both flow cells were acquired simul-
taneously with a microprocessor; Data were then shipped to the L31

11/23 minicomputer where they were converted to absorbance and then

43

smoothed and derivatixed by means Of a moving. 7-point polynomial
Savitxky-Golay algorithm [42]. The smoothed data points were also
regenerated according to Equation 1-3. Processed data acquired for the
rise curve of the high concentration sample (the broad center peaks in
Figure 1-2) with the bubble gated and debubbled flow cells are listed
in Tables 2-5 and 2-6. respectively. Values of b used to regenerate
these data sets were determined as described below.

Inspection of Equation l-l reveals that b is simply the inverse
negative slope of the linear portion of a plot of ln(A‘s-At) versus
time. In Figure 2-6 smoothed absorbance values and smoothed values for
the difference between A,8 and At are plotted as a function of time for
both data sets. The right y-axis in this figure is logarithmic and
pertains to the (A‘s-At)‘values. while the left y-axis is linear and
relates to the At values. See the figure caption for further details.
The lepes of the linear portions of the logarithmic plots for data
Obtained with the bubble gated and debubbled flow cells were ~0Jl36 and
~0.37l. respectively. which correspond to b values of 1.36 s and 2.70
s. Note that the flow cell debubbler approximately doubled the magni~
tude of longitudinal dispersion for this CF system.

It is also possible to determine b emperically by regenerating the
data sets with different values for b and plotting the results on a
graphics terminal. If the value chosen for b is tOO small. regenerated
peaks will still appear to be exponentially deformed. Larger than
Optimunlvalues for'b on the other hand result in regenerated peaks
that are excessively noisy and have large spikes at the leading edge of
rise and fall curves that extend above and below the sample and base

line steady state absorbance levels. Optimum b values determined in

44

 

 

 

 

 

 

Table 2-5. Data for the first 10 s of the rise curve for the steady
state peak shown in Figure 1-2. This data set pertains
to the bubble gated flow cell.

. a .*b * .fc * +
time (5) At At dAt/dt At lr1(ASS-At) At norm
63.49 .0096 .0043 .0429 .0472 -.2000 .9524
64.01 .0180 .0225 .0837 .1061 -.2225 .8801
64.54 .0590 0732 .1312 .2044 -.2879 .7594
65.05 .1600 .1593 .1737 .3330 -.4099 .6015
65.53 .2771 .2726 .1999 .4725 -.5971 .4303
66.01 .3938 .3931 .2045 .5977 -.8442 .2766
66.52 .5031 .5013 .1888 .6902 -1.1341 .1630
*67.03ci .5945 .5921 .1595 .7516 -1.4658 .0877
67.54 .6640 .6629 .1253 .7882 -1.8320 .0427
68.04 .7136 .7140 .0930 .8071 -2.2164 .0195
68.55 .7490 .7490 .0655 .8146 -2.6037 .0103
69.03 .7722 .7720 .0443 .8163 -2.9757 .0082
69.50 .7863 .7869 .0291 .8160 -3.3215 -—
*70.00 .7955 .7960 .0190 .8150 -3.6119 -
70.50 .8021 .8022 .0126 .8149 -3.8728 ——
70.98 .8062 .8066 .0088 .8154 -4.1105 -—
71.46 .8103 .8097 .0067 .8164 -4.3200 -—
71.99 .8116 .8123 .0054 .8176 -4.5375 -—
72.51 .8144 .8147 .0043 .8190 -4.7915 -
73.01 .8172 .8169 .0033 .8202 -5.0995 -
73.50 .8186 .8186 .0022 .8208 -5.4262 -—

aAt = raw absorbance.

bA: = smoothed absorbance.

2A: = regenerated absorbance.

*

..* inclusive data points used to calculate regeneration factor.
See text for details.

4S

 

 

 

 

 

 

Table 2-6. Data for the first 10 s of the rise curve for the steady
state peak shown in Figure 2-1. This data set pertains
to the debubbled flow cell. All other notation is shown
in Table 2-5.

. * * + * +
time (5) At At dAt/dt At 111(ASS-At) At norm
68.55 .0118 .0100 .0248 .0670 -.2070 .9243
69.03 .0184 .0210 .0480 .1315 -.2206 .8392
69.50 .0449 .0492 .0765 .2250 -.2564 .7170
70.00 .0968 .0976 .1050 .3390 -.3210 .5695
70.50 .1643 .1629 .1280 .4574 -.4154 4179
70.98_ .2391 .2385 .1418 .5647 -.5370 .2824
71.46 .3157 .3169 .1451 .6506 -.6810 1757
*71.99 .3944 .3931 .1391 .7129 -.8442 1003
72.51 .4644 .4627 .1270 .7548 -1.0208 .0517
73.01 .5239 .5232 .1118 .7804 -l.2046 .0240
73.50 .5733 .5735 .0960 .7942 -1.3883 0108
73.99 .6158 .6167 .0810 .8030 -1.5784 .0037
74.50 .6524 .6526 .0675 .8080 -1.7696 .0009
75.01 .6850 .6832 .0559 .8117 -1.9675 -
75.53 .7060 .7077 .0459 .8133 -2.l602 -—
76.03 .7281 .7268 .0374 .8129 -2.3413 ——
76.54 .7420 .7426 .0304 .8125 -2.5195 —-
77.02 .7550 .7564 .0250 .8139 -2.7091 ——
77.50 .7685 .7669 .0207 .8145 -2.8806 —-
77.99 .7760 .7758 .0171 .8151 -3.0534 -—
*78.47 .7824 .7836 .0140 .8159 -3.2340 ——

 

46

.mucmom nu<umm<v
62300.5 04 .mucfioa u< wofoog n 4 ”:oo 30$ pognsnoo .mucwoa au<-mm<v voguOOEm "D .mucmom u<
wozuoosm "D ”:06 30C woven oznam .mxmon— «Eu mo moiso om: thm a mo 5:28.536 .ouN ops»:

A368;

or m o
0.4.. _ _ o.

 

0N1 ll

Q a

eouoqlosqv

 

 

.0
<0

 

CO ®.O

47

this way for the bubble gated and‘debubbled data sets were 1J)s and
2.3 s. respectively; These values are somewhat smaller than those
determined graphically. This is because the derivative of absorbance
with respect to time that is required to regenerate the data amplifies
both the hydraulic and electronic noise associated with the CF system.
In this regard. the choice Of smoothing parameters can have a pro-
nounced effect on the quality of regenerated data. I found that 5 to
7-point Savitsky-Golay smooths and derivatives gave the best results
with this CF system.

As shown in Figure 1-3. regeneration of both data sets resulted in
peaks that closely resembled the integrated Gaussian curves predicted
by Snyder's model. The value of at can be estimated by plotting the
values for regenerated data points on a cumulative probability scale as
a function Of time [17]. Because the limits of the probablity scale
are 0 to 1. regenerated data must be normal ized by dividing the dif-
ference between the sample steady state absorbance and the regenerated
absorbance at each time. by the difference between the absorbance Of
the sample and baseline steady states [17]. Values for normalised.
regenerated data can be found in the last column of Tables 2-5 and 2-6.
As shown in Figure 2-7. normalized points of both data sets fall on the
same straight line. Thus the assumption that regenerated peaks
approximate cumulative Gaussian curves is good. (h1the cumulative
probability scale. the interval between 0.16 and 0.50 corresponds to
one standard deviation. This interval is indicated by dashed horizon-
tal lines in Figure 2-7. Vertical lines dropped from the points where
the horizontal lines intersect the plot cross the time axis at 2.4 s

and 3.7 s (At-1.3 s). Thus at measured for this CF system (n-2 s-l.

0.0000l

0.0005
0.00l

0.002
0.005

0.0l
0.02

0.05
0. l

0.2

0.3
0.4
0.5
0.6
0.7
0.8

0.9
0.95

NORMALIZED REGENERATED ABSORBANCE

0.98
0.99

0.998
0.999

0.9999

Figure 2-7.
probability

48

I

 

  

 

 

 

I
I 0" -I.3s
| 0 Bubble gated
I o Debubbled
|
l
P l
I- I I
| I
l 1 a 1 I l 14 1 l l l
0.5 L0 l.5 2.0 3.0 4.0 5.0 6.0 7.0 8.0
TIME (s)

Graphical estimation of o by means of a cumulative

plot. t

49

dt-0.1 cm. F-.014 mL s-l. t-85 s) is approximately 1.3 s. This com-
pares favorably with the value calculated with Snyder's model
(Ot'lJ) s) assuming these values for major variables. and values for

minor variables as in the first section of this chapter.

L Manifold Components

1.. LA

For the most part. manifold components (mixing coils. connectors)
were fabricated locally. Glass tubing (1.0 cm ID x 2.3 cm 0D) in
various lengths was wrapped around 0.8 cm or 3.0 cm mandrels in the
Chemistry Department Glass Shop. The resulting coiled glass tubes
served as mixing and time delay coils. respectively. The latter were
mounted in water tight plexiglass jackets so that they could be ther-
mostated by means of a circulating water bath. For many applications
mixing coils fabricated from 0.10 cm ID x 0.18 on 0D Micro-line tubing
(#18160. Thermoplastics Scientifics. Inc.. Warren. NJ') were highly
suitable. To form coils the tubing was wrapped around glass rods.
Tube ends were secured to the rod with masking tape. The rod was
immersed in boiling water for about one minute and then removed and
placed in cold water for about five minutes. A similar method was used
by Amador [43] to fabricate mixing coils from polyethylene tubing.
These plastic coils have a number of advantages. They are easy to
make. inexpensive. flexible. unbreakable. and compatible with Omnifit
flangeless gripper fittings (ThermOplastics Scientifics. Inc.. Warren.
NJ) that allow xero dead volume connections to be made between coil

ends and reagent addition tees. The major disadvantage Of these plas-

tic coils is that they are considerably less wettable than glass coils.

50

Therefore hydraulic instability (surging due to back pressure. and
bubble break-up) is likely to occur. Usually these problems can be
eliminated if the surfactant concentrations in reagents are approxima-
tely doubled relative to the<concentrations that would be used for a
glass manifold. IFor determinations that are adversely affected by
surfactants or that require heating to acceleratre chemical reactions.
glass mixing coils are best.

Specially designed fittings are required to connect sample. air.
and reagent pump tubes to various points along the flow system of the
CFA instrument. It is particularly important to minimise the dead
volume of these fittings to prevent increased dispersion due to mixing
effects. In general fittings should have central bores that match the
inside diameters of mixing coils and side arm bores that are no larger
than those of the pump tubes to which they are connected. A large
assortment of manifold fittings are commercially available from a
number of suppliers. but they are relatively expensive. For this
reason I designed the fittings illustrated in Fig. 2-8 that were fabri-
cated in the Chemistry Department Machine and Glass Shops. For most
applications only two basic fitting types are required: 1) sample
inlet blocks. and 2) reagent addition tees. These fittings were
machined from plexiglass rods. The central and side arm bores were
0.10 and 0.05 cm. respectively. Pump tubes are connected to the fit-
tings by means Of stainless steel side arms (0.05 on ID). These press-
fit into holes drilled at right angles to the center bore of the fit-
ting. Side arms were secured with epoxy cement. Further construction
details for inlet blocks and reagent tees can be found in Figures 2-8A

and 2-8B. respectively. Flow stream debubblers were constructed from

51

.muo_nnsnou Emopum 36am Au .mawou mmmfim pom pom:
“moon :owuwvvw acoueom Am .mfiwoo mmmfim pom pom: .uzwwm
.mmwfiu u u .ucoEou xxono u m .uoficfl oHnEmm n m
.w-~ osswna

ucuwh can omvvmz .m~moo omummHL pom pom: .uuom
.meou owummfia you pom: .umo~ ”mxoofin uo~c~ o~mewm n<
.uoficfi ucomaoh u z .uoficfl Ana 0 < .Eoumxm ¢uue 0:» you vocwflmov mucocomsou vHomMcmz

102?. o»

 

 

E

 

 

0

 

 

 

 

 

 

 

 

 

 

 

 

52

0.10 cm ID glass tubing. Debubblers can be used in two ways. Either
the bubbles can be pumped away. usually at about twice the rate at
which they were introduced. or the analytical stream can be pumped away
at about 75% of the total (air + liquid) flow rate so that bubbles (and
a small portion of each liquid segment) are forced to waste. Both
configurations are illustrated in Figure 2-8C. The former is generally
used to debubble flow streams prior to passage through packed bed
reactors. while the latter is generally used to debubble flow cells.
Readers with an interest in more exotic fitting types are refered to
William Furman's. Continuous EM Analysis: £5951 _an_d Pragtice [6].

.2_,_ __IA Manifold Components

Reaction coils for FIA manifolds are generally constructed from
flexible. nonwettable. polymeric tubing. In this work open tubular

reactors were made from lengths of 0.05 cm ID Teflon TM

tubing. Tubes
were terminated with Omnifit flangeless gripper fittings and then
wrapped in tight overlapping coils around a 1.0 cm diameter glass rod.
The coiled tubing was slipped off the rod and secured at two points
with plastic cable ties. Packed single bead string reactors (SBSRs)
were fabricated from 0.08 cm ID Teflon tubing and 0.05 cm diameter
glass beads (Propper Manufactoring Company. Long Island City. NY). A
gripper fitting was attached to one end Of the tube which was then
crimped (directly above the gripper fitting) with a pair of surgical
forceps. A disposable pipet tip (the tip must be trimmed slightly to
allow free passage of the beads) was attached to the other end of the
tube by means of a plastic sleeve. This assembly was then mounted in a

vertical position and the pipet tip was partially filled with the glass

beads. When the pipet tip was tapped gently. the beads began to enter

53

and fill the tube. I found that the packing procedure worked best when
the pipet tip and tube were completely filled with water before the
beads were added. The packed tube was crimped just above the topmost
bead and terminated with a gripper fitting. The SBSR was then coiled
and secured with cable ties.

Three different modes of sample injection (A-C) were used for FIA
experiments as illustrated in Figure 2-9. In mode A. a Type 50 six-
port rotary valve (Rheodyne Inc.. Cotati. CA) was used. A shaft
attached to the valve core was turned manually to switch between the
'LOAD' and 'INJECI" positions. Dual and single four-port slider valves
(Alltech Associates. Deerfield. IL) were used for modes B and C.
respectively. The sliders on these valves were pneumatically actuated.
External sample loops determined the injected sample volume in modes A
and B. In mode C sample volume was determined by the time interval
that the slider remained in the 'INJECT' position. Reproducible timing
was achieved by means Of a microprocessor controlled valve actuator
[44]. The parts Of both valve types in contact with solutions were
made from Teflon and had 0.8 cm bores. The tee connectors used for
reagent addition are illustrated in Figure 2-9D. Details Of more spe-

cialized manifold components can be found in Ruzicka and Hansen's Flow

Injection Analysis [10].

L Plumbing

Short lengths of Teflon tubing (.04 cm ID. .05 cm 0D) were used
for sample and reagent withdrawal lines. These were attached to pump
tubes either by direct insertion. or by means Of plastic sleeves if

there was a poor match between the ID of the pump tube and the CD of

54

.O>~m> povwam xmzuuaom ofiwcwm mu

.noo~ cumsmm u A .uoch Ouaswm n m
.m-~ oosmna

.oou :onofiooo ucomooz no

.o>~w> xpmuop whonuxmm m<

.o>~w> howwum xmzuusom “one am
.<Hm pom mucocoaaou ufiomwcwz

2 .w_ommcme 0» u z .uoficfl smouum pumnhmu n u

 

 

 

 

 

 

 

 

 

 

 

 

 

.oummz Cu n
”V
- A
fluhhhH—lll I I I I 00
<4Il!4“\v brill>>
[0.4L
8
mw
Fhuw54‘_n:un.
MW >> mw >>
1 i i 1
ya 0 D;
We... ..I .... I. WILL
0 ob “v
ooa\ /~o call-36

 

 

 

55

the Teflon tube. The same techniques were used to connect the delivery
end of the pump tubes to manifold fittings (see Figures 2-10A and 2-
108). A stock of worn pump tubes were kept on hand from which sleeves
were made. Plastic sleeves were also used to connect glass mixing
coils to fittings on CFA manifolds. Note that the coil and fitting
ends must be butted securely within the sleeve. Otherwise the integri-
ty of the air segmented stream was degraded (e.g. -- air bubbles were
trapped in the void volume between the coil and fitting ends) and
increased dispersion due to mixing effects was noticeable. The LC
fittings used in conjunction with plastic mixing coils (for both CFA
and FIA) insured zero dead volume connections (see Figures 2-10C and2-
10D).

A few words about sample introduction methods used for FIA and CFA
are in order here. In FIA it is best to withdraw samples from their
containers into the sample loop as shown in Figure 2-llA. Note that in
this configuration. the sample does not pass through the pump before it
reaches the sample loop. Therefore the time required to fill the
sample loop is short and cross contamination of successive samples is
minimized. In CFA on the other hand. samples generally pass through
the pump before they enter the manifold a shown in Figure 2-11B.
Because the sample moves through the pump tube under essentially
laminar flow conditions. considerable mixing occurs either between
successive samples. or between a sample and the wash solution. The
intersample air segment (IAS) that enters the sample line when it is
cycled between a sample and the wash solution (or the reverse) provides
a partial barrier to intersample mixing. If additional IASs are intro-

duced at the start of each sample and wash interval by rapid.

S6

     
  

[I’ll/1111111111110,”
I

\\\\\\\

D
f

 

os
\§\ammmmmm
T_____JEE@

\\\\\\\\\\\\

   
  
     
 

 

 

 

 

 

 

 

 

 

 

 

MC

s ll...""'o"fi.a.I.-.A.l
. l

  

oilife'o'ohffl.‘
O

 

Figure 2-10. Methods for interconnecting various manifold components
used for continuous flow analysis. T = teflon tubing, P = pump

(not shown), I.S. inner plastic sleeve, 0.8. = outer plastic sleeve,

MC = mixing coil, = reagent inlet, G = gripper fitting, C = standard
HPLC connector, 8

plastic sleeve.

"50"

57

A) PUMP
C or R 2,—‘__>

 

 

 

 

 

 

 

 

 

 

 

 

:7—

 

 

 

C)
‘c’
O
.D
3
8 o
q I
0 so 67)
Time (5)

Preferred method of sample introduction for FIA

Figure 2-11. A)
systems. C = carrier stream,R = reagent, S = sample, W = to waste.
= sample,

B) Pecked sample introduction used for CFA systems. S
W = wash solution, IAS = intersample air segment. C) Effect of
IAS on dispersion that occurs in the sample lines of CFA systems.

58

repetitive withdrawal and insertion of the sample line ('pecking').
mixing within the pump tube can be greatly reduced as shown in Figure
2-11C.

This figure is a generalized composite of data collected from a
number of individual experiments in which no manifold was used.
Instead. one end of a pump tube was connected directly to the flow cell
and the other end was fitted with a 20 cm length of Teflon tubing (.025
cm ID) that served as the sample withdrawal line as shown in Figure 2-
118. The sample and wash solutions used in these experiments were
alkaline phenol red and distilled water. respectively. The absorbance

at 540 nm (11“ for alkaline phenol red) was recorded while the sample

x
line was manually cycled between the two solutions at 60 s intervals.
As previously mentioned. one IAS forms naturally when the sample line
is transferred from one solution to the other. The IAS can be e1 imi-
nated. however. if the pump is turned off during this Operation. Ad-
ditional IASs can be introduced by pecking. In this way the amount of
dispersion that occurred in the sample line and pump tube was monitored
as a function of the number of IASs introduced at the leading edge of
each sample and wash slug at several different flow rates.

Results for one set Of experiments where the flow rate was 0.32 mL

min"1

are shown in Figure 2-llC. With no IAS. dispersion in the sample
line and pump tube was large. With one IAS. dispersion was reduced.
but still significant. With three IASs. however. dispersion was almost
eliminated. As expected. in the absence of an IAS. dispersion
decreased as the flow rate increased. When three IASs were introduced

at the leading edge of each sample and wash slug. however. dependence

of dispersion on flow rate was minor. These experiments underscore the

59

importance of pecked sampling for minimum dispersion in CFA experi-

ments .

CHAPTER 3

DESIGN AND CHARACTERIZATION OF AN ELECTRONIC BUBBLE GATE

FOR AIRrSEGMENTED CONTINUOUS FLOW ANALYSES

A; Overview

As already mentioned. photometric detection with CFA systems is
complicated by the presence of air segments in the analytical stream.
In commercially available first and second generation systems. the
analytical stream is debubbled just prior to detection. Unfortunately.
this expedient allows previously segregated liquid segments to mix in
the debubbler and flow cell. and therefore the rate at which analyses
can be performed is reduced. As shown in Chapter 2. mixing effects
that occur in flow cell debubblers and flow cells can be removed (after
the fact) by analog [45.46] or digital [47] curve regeneration tech-
niques. or they can be eliminated by using bubble-through flow cells
and a gated detector (bubble gating) as is done with the Technicon
third generation SMAC clinical analyzer. This chapter deals exclusive-
1y with the latter approach.

Habig and co-workers [27] appear to have develOped the first
bubble gate. which was activated by conductance changes within a
specially designed flow cell. They concluded. however. that their
design was preliminary and was not suitable for routine use. Sub-

sequently a more robust electronic bubble gate was described by Neeley.

6O

61

et a1. [28].'The same group later reported a slightly modified version
of this circuit [48] which they incorporated into a high performance
colorimeter for a miniaturized CF analyzer. This bubble gate repeti-
tively sampled and stored the detector signal at a frequency determined
by an adjustable internal time base. Then. for a fixed time interval.
the stored signal level was compared with the real-time signal level by
means of a window comparator circuit. When the stored and real-time
signal levels were within the limits Of the window comparator. the
real-time signal level was stored in a second sample-and-hold circuit
connected to the readout device.

The bubble gate reported here is much simpler. It uses the
periodic fluctuations of the detector signal. caused by the successive
passage of air and liquid segments through the flow cell. to syn-
chronize the time Of data acquisition and temporary storage with the
brief interval during which each liquid segment completely fills the
flow cell. Data are presented to demonstrate that improved performance
of both commercially available and custom built CF analyzers can be
achieved using the bubble gate described. The analytical performance
Of a miniature CF (mCF) system with 0.1 cm ID manifold components and a
bubble gated detector used for the colorimetric determination of
nitrite and silicate in water is presented. Results of nitrite deter-
minations Obtained with mCF system are compared with those reported for

flow injection analysis.

62

13... W

L _C_F Systems

The filter photometer used with the mCF system is described fully
in Chapter 2. The 0.2 cm ID manifold CF (AAII CF) system consisted Of
a Pump III (with air bar). and an Industrial S.C. colorimeter equipped
with 1.5 cm x .15 cm ID debubbling flow cells and 540 nm interference
filters (Technicon Instruments Inc.. Tarrytown. NY). In bubble gating
experiments with the AAII CF system. the Technicon colorimeter was
modified by removing capacitors C-201. C-203. and C-204 from the
'control module' circuit board. This modification reduced the rise time
of the detector signal from about 0.3 s to about 10 ms. Also. the
sample side debubbling flow cell was removed and replaced with a 1.5
cm x .1 cm ID bubble-through flow cell (Gamma Enterprises. Inc.. NF.
Vernon. NY). The cell was mounted directly onto the phototube housing
with an adaptor to occlude the phototube entrance and exclude stray
light. The colorimeter was then adjusted in the usual manner. The
'telemetry output' (0-5 V) Of the colorimeter was connected to the bubble
gate described below through a unity gain inverting amplifier with
offset. This made the logarithmic output of the calorimeter compatible
with the logic of the bubble gate which is described in detail below.

Manifold components for the AAII CF system were obtained com-
mercially from standard sources. while those for the mCF system were
for the most part custom made as described in Chapter 2. Technicon SMA
Flow-Rated Pump tubes were used for both systems.

Data were recorded with a Model SR-255 10 inch strip chart recor-
der (Heath CO.. Benton Harbor. MI). In some experiments data were

simultaneously logged with a microcomputer (see Chapter 2).

2‘ Determination pf Nigrite ppg Silipate

a. Nitrite Reagents

The sulfanilamide (SAN) reagent was prepared by dissolving 10 g
SAN in 500 mL of deionized. distilled water (DDW) and 100 mL of concen-
trated HCl contained in a l L volumetric flask. This solution was
diluted to the mark with DDW. mixed. and transferred tO an amber
bottle. Then 0.5 mL Brij-35 wetting agent was added.

The N-(l-Naphthyl) ethylenediamine dihydrochloride (NED) reagent
was prepared by dissolving 1.0 g NED in l L of DDW. This solution was
transferred to an amber bottle. and 0.5 ml Brij-35 was added.

b. Nitrite Standards

The primary nitrite standard was prepared by dissolving 0.345 g (5
mmod) of dried. analytical reagent grade sodium nitrite in.l L DDW.
Working standards in two ranges were prepared by dilution Of the pri-
mary standard using digital microliter pipets and volumetric flasks.

c. Silicate Reagents

The molybdate reagent was prepared by dissolving 10.8 g of am-
monium molybdate and.2.8 mL of concentrated sulfuric acid in about 500
mL of DDW contained in a l L volumetric flask. This solution was
diluted to the mark with DDW. mixed. and transferred to a plastic
bottle. Then 2 ml.of an aqueous sodium lauryl sulfate solution was
added.

The tartaric acid reagent was prepared by dissolving 100 g of
tartaric acid in 950 mL of DDW contained in a l L plastic bottle.

The stock stannous chloride solution was prepared by dissolving 10

g of stannous chloride in 20 mL of 50% (V/V) hydrochloric acid.

64

Heating is sometimes required to effect complete dissolution. This
solution was stored in a plastic bottle.

The working stannous chloride reagent was prepared by adding 0.5
mL Of the stock reagent to 50 mL of 1.2 N HCl solution contained in a
plastic bottle. This solution is not stable for more than about 8
hours.

The wetting agent was prepared by dissolving 10 g Of sodium lauryl
sulfate in 100 mL of DDW.

d. Silicate Standards

The primary silicate standard was prepared by fusing 0.3007 g (5
mmol) of dried silicon dioxide (Alfa-Ventron. 99.9%) with 0.7 g of
sodium carbonate in a platinum crucible. Heating was effected with a
compressed air-natural gas flame. After the melt formed a glass (:5
min). heating was discontinued and the crucible was allowed to cool.
When cooling was complete. DDW was added to the crucible which was then
covered and allowed to stand overnight. The next day the contents of
the crucible were transferred quantitatively to a l L volumetric flask.
This solution was diluted to the mark with DDW and transferred to a
plastic bottle.

Working standards in the range of 5 to 45 pH were prepared by
dilution of the primary standard by means of digital microliter pipets
and 100 mL volumetric flasks. Working standards were transferred to
plastic bottles immediately following their preparation to avoid the
possibility of silica in the glass volumetric flasks being leached into

the standards.

65

3; Bubble Gate

 

 

When bubble-through flow cells are used with colorimetric. air~
segmented CF analyzers. the detector signal closely resembles a square
wave. With an air-segmented blank solution the detector output varies
between the 100% T level (5.0 V in our case) and the 074 T level (0.0
V). as shown in Figure 3-1. As an air segment traverses the flow cell.
it reflects most of the light away from the photodetector. and the
signal approaches the 0* T level (0.1 to 054 V in our case). The
actual magitude of the signal observed in the presence of an air
segment is somewhat erratic. and it decreases as the transmittance of
the two adjoining liquid segments decreases. When the air segment
exits the flow cell. the detector rapidly rises to a level that cor-
responds tO the transmittance of the liquid segment (see Figure 3-1).
The bubble gate described here uses the information encoded in the
periodic fluctuations of the detector signal to synchronize the time of
data acquisition and temporary storage with the brief interval during
which each liquid segment completely fills the flow cell.

The schematic diagrams presented in Figure 3-2 depict the func-
tional units of the bubble gate circuit. In the basic bubble gate
circuit (Figure 3-2A) comparator ICl converts the detector signal to a
TTL logic level signal that indicates the presence or absence Of an air
segment in the flow cell. This is accomplished by setting the
threshold level of the comparator at about 0.5 V. A HI logic level at
the output of the comparator indicates the absence of an air segment in
the flow cell. Note. however. that the logic level transitions of the
comparator lag the entrance and exit Of air segments into and out of

the flow cell by several ms because of the time required for the

66

 

 

V0 LTS

 

 

 

 

 

 

() 4_:--

TIME

Figure 3-1. Fluctuations in the detector signal caused by the
successive passage of liquid and air segments through the flow cell:
Position 1, air segment completely within cell. Position 2, air
segment exiting cell. Position 3, cell completely filled by liquid
segment. Position 4, air segment entering cell.

I0

67

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

' k I m ‘( I000 ' LEDZ 1:471: ——I_.
00 00 .. . .4 > =m- @09E900
. k
,. >
IOk ”122.40 rcz-e ©
an +mv I C)
1’59 0
..— Okuvsarro 11’" v@
LEDI To
.I-
ggocn: »_. ': PW. 2 C
H Log
new“; "I S/H Our S/H Out
4
39k
‘ 1C3 R24
1c: ’ S/H LOG
A 3 From
b© L A h S/H 00'
Thread! LeveI Fnure 20 Frggc© FM 20
A)
Detector
Snnol OQI
In K
19' IM .
p v
TO '1'
- Figure 2A
0)

 

Figure 3-2. Generalized schematic diagrams of bubble gate circuitry.
All resistances in ohms, all capacitances in microfarads. A) Basic
bubble gate, ICl = LF311 FET input comparator, IC2, 74L8123 dual
monostable multivibrator, IC3 = LF398 sample-and-hold amplifier,
IC4 = AD755 logarithmic amplifier, IC5 = TL084 quad FET input
Operational amplifier. B) Differential edge sensor, IC5 = TL084
quad FET input operational amplifier, 1C6 = LF311 FET input
comparator. C) Sample-and-hold amplifier update logic, 1C7 = 7410
triple 3-input NAND gate. D) Automatic threshold adjustment for
comparator, ICl. IC8 = TL084 quad FET input operational amplifier.

68

detector signal level to cross the threshold level of the comparator.
The rising edge of the comparator output triggers monostable multi-
vibrator IC2-A. which generates a time delay pulse. The duration Of
this pulse is adjusted with potentiometer P1 to terminate at the
approximate midpoint of the interval during which a liquid segment
completely fills the flow cell. Two light emitting diodes (LEDl and
LED2) facilitate this adjustment by giving‘visual indication of the
logic states of the comparator and monostable. respectively. The
falling edge of the time delay pulse triggers monostable IC2-B which
generates a pulse with a fixed width Of approximately 100 us. A gated
version of this 100 pa pulse is used to update sample-and-hold ampli-
fier IC3 which samples on a TTL HI and holds on a TTL LO.
Irregularities in the segmentation pattern of the analytical
stream. generated when the sample probe is cycled between samples and
the wash solution. can produce two conditions which cause the basic
bubble gate to malfunction. IFirst. irregular liquid segments that
completely fill the flow cell for a time interval much shorter than the
selected time delay interval cause the comparator and the monostable to
get out of phase; several liquid segments Of normal length must pass
through the flow cell before these two circuit elements are again
synchronous. Erroneous sample-and-hold amplifier updates could occur
in the interim. 'This problem was corrected by clearing monostable IC2-
A with the falling edge of the comparator output so that the time delay
pulse is aborted in the event that it is still in progress when an air
segment enters the flow cell. Second. irregular liquid segments that
completely fill the flow cell for a time interval approximately equal

to the duration of the delay pulse can cause incorrect sample-and-hold

69

amplifier updates because of the lag between the initial entry of an
air segment into the flow cell and the logic level transition of the
comparator output as discussed previously. This problem was eliminated
with the differential edge sensor circuit shown in Figure 3-2B. which
generates a TTL HI at its output whenever the detector signal has a
non-zero derivative. Because the output of this circuit responds to
changes in the detector signal level almost instantaneously. it can be
used to eliminate the possibility of sample-and—hold amplifier updates
at the time of initial entrance or exit of an air segment into or out
of the flow cell. Three conditions must be sastisfied in order to
update the sample-and-hold amplifier:

1) The output of the monostable IC2-B must be HI.

2) The output of the differential edge sensor must be LO.

3) The ouput of comparator 101 must be HI.
Figure 3-2C shows the logic that prevents sample-and-hold amplifier up-
dates when the above conditions are not satisfied.

This bubble-gate circuit will not operate when the signal level of
a liquid segment is less than the threshold level of comparator 101.
If the threshold level Of the comparator were to be fixed at 0.5 V.
liquid segments with transmittance less than about 0.1 would be
indistinguishable from air segments. and the bubble gate would be
restricted to an operational range Of 100% T to 10% T. However.
because an air segment's effective transmittance is not constant but
decreases as the transmittance of the two liquid segments adjoining it
decreases. it is possible to extend the lower Operational limit of the
bubble gate considerably by using the detector signal level stored in

the sample-and-hold amplifier for automatic adjustment of the

70

comparator threshold to an appropriate level. This function is per-
formed by the circuit shown in Figure 3-2D. A buffered voltage divider
applies one-half of the sample-and-hold amplifier output to the
threshold level input of comparator 101 through a unity gain buffer
(IC8-D). Two active diode clippers limit the maximum and minimum
threshold levels to about 2 V (fixed) and 0.1 V to 0.5 V (adjustable).
respectively. This circuit extends the bubble gate's lower limit of
operation to about 3% T (1.5 A). and in addition improves performance
in the high transmittance range because it provides greater separation
between the comparator threshold level and the somewhat erratic
detector signal produced by the passage of air segments through the
flow cell.

A logarithmic amplifier (Model AD755. Analog Devices. Norwood. MA)
is also included in the bubble gate circuitry so that peak heights are
linear functions of concentration.

A detailed schematic diagram of the bubble gate can be found in
Chapter 2.

It should also be noted that bubble-gating can be accomplished
with a microcomputer and apprOpriate software. Apparently this
approach was used by the designers of the Technicon SMAC clinical CF

system.

Q; Resulgs ppd Dispuggion
L Introduptipn

The bubble gate was tested with the Technicon AutoAnalyzerII CF
System (AAII CF System) and the mCF system described in Chapter 2. The

colorimetric determinaion of nitrite via a modified Griess reaction

71

procedure [49] was chosen as the reference assay with which the
performance of the two CF systems could be evaluated and compared. A
composite manifold diagram for both CF systems is shown in Figure 3-3.
Pecked sampling did not improve the performance of either the standard
or the bubble-gated AAII CF system noticeably. As described in Chapter
2. however. it did improve the performance Of the mCF system. There-
fore. pecked sampling was used routinely with the mCF system. For both
systems the time required to reach steady state and the percent inter-
action were determined as described below.

g_,_ Steggy State ggperimgngs

A set of five aqueous nitrite standards with nominal concentra-

 

tions Of 5 11!. 15 (1!. 25 11!. 35 31!. and 45 II! were prepared. and the
detector gain was adjusted such that these standards produced a
response of approximately 10%. 30%. 50%. 70%. and 90% full scale.
respectively. on the strip-chart recorder. With the AAII system each
standard was sampled in triplicate for 5 s. 10 s. 20 s. 30 s. 40 s. 50
s. and 60 s with a 60 a wash between each sample to minimize carryover.
With the mCF system each standard was sampled in triplicate for 5 s. 10
s. 15 s. 20 s. 25 s. 30 s. and 60 s with a 30 s wash between each
sample. The average peak height for each sampling interval was divided
by the steady-state peak height (60 s sampling interval) and multiplied
by one hundred to obtain the per cent of steady state (%SS) as a
function of sample time. .A grand average of %SS for all five standards
at each sampling interval was then calculated. The results of this
calculation showed that %SS was independent of concentration to within
about 1% for the 5 s and 10 a sampling intervals and to within about

(L5% for sampling intevals greater than 10 s. The increased scatter in

72

NOMINAL FLOW RATE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[mL/min]

AIR [Fl—(JO) 1 , WASTE
54C)
32(05) 4.? a _-_ flit.
E31,? I5mmxlmmID
SAMPLE .3206) 1 .4” a _0 0000 _»00mm05mmm)
edgngfiiflfl} iflfiflfiw.. . a 9:41
:DILUEN_ l..80(.23 .4182
SAN 10(05)
NED ,|o(,05) I5 mm x l.5mmlD
"°° 9 WASTE
PUMP
Figure 3-3. Composite diagram of various CF manifolds used for

nitrite determinations. Values outside or within parentheses refer to
AAII or mCF manifolds, respectively.

73

the peak heights at shorter sampling intervals is probably a result Of
the manual sampling employed.

Results of the %SS experiments are presented graphically in Figure
3-44 ‘Values for major experimental variables can be found in Table 3-1
as well as the estimated dispersion. at. of the air segmented sample
slug during its passage through the CF systean The latter quantity was
calculated according to the model of Snyder and Adler [25.20]. Values
of minor variables shown in Table 3-1 were in keeping with a similar
set of calculations by Snyder [32].

As shown in Fig. 3-4. bubble gating reduced the sampling time
required for the AAII CF system to reach 98 %SS by about 8 s. A
similar reduction in the sampling time required for the mCF system to
reach 98 %SS was achieved by using the peeked sampling techique. An
apparent trend for the mCF system also shown in this figure is a
decrease in the sampling time required to attain a given.%SS as the
segmentation frequency. n. increases. Note. however. that with the mCF
system. n was increased by increasing the pump speed (and therefore the
liqiud flow rate. F) in order to keep air segmentation in phase with
the roller lift off of the peristaltic pump. as discussed in Chapter 2.
Thus the residence time. t. for the sample slug in the manifold
decreased as n and F increased (see Table 3-1). To differentiate
between the effects of n and F. the experiment was repeated for
n . 1.5 .‘1. a - 0.0067 .1. .‘1 and n - 3.0 s". F - 0.014 mL .‘1 with t
held constant at 110 s by altering the length and number of mixing
coils in the manifold. As shown.in Figure 3-5. the same trend was
observed. These results were were contrary to the predictions of

Snyder's model [10] (see Table 3-1). This was not surprising. because

74

 

 

 

1C)O '-F' __ ._____
___,_.__. - ‘.— .1-
.__ 6
5
m 9‘) —-_t 4
o .__
on
>\ BI) --» 3
'0
o _—
0
w 7 0 --
}; _._
o
2 5(3 ..L. 2
O)
0.
5() -u-' 1
I l l I
I I I I 1 fl]
0 5 1 0 1'5 2(3 255 3()
Sonprng Intervol (s)
Figure 3-4. Results of percent steady state (%SS) experiments:

1, standard AAII CF system. 2, bubble gated AAII CF system.

3, bubble gated mCF system without pecked sampling. 4-6, bubble
gated mCF system with pecked sampling and n equal to 1.5, 2.0, and
3.0 5'1, respectively.

7S

.xHo>wuuommou .No.~

mam mm.~ one: mucwuuom Honucou conumpnwfimu wumvcmam .mEoumzm mu HH<< voumwumnnnan mam vumvcmum ecu mom
.Hum «EU mIoHoAm n mmN.3oU acumowmmooo scamsmmflm o>wuoommo

H IEU ocxv HoHoA~.m u nxv ”coflmcou oommHSm .Omfiom

a

IoHoAm.w u may xufimoomw> "moHnmflum> momma

 

 

 

 

 

 

I. H m u
no.H 96.6 Nw.o em.o wo.H wo.H mH.~ I. . Hog oHao oo
oHH cHH mo ooH omH omH com com H83 0
mHo.o aooo.o mHo.o mwoo.o aooo.o aooo.o mmo.o -o.o HH-m Hey an
H.o H.o H.o H.o H.o H.o ~.o N.o H563 no
o.m m.H c.m o.N m.H m.H m.o m.o HH-mU =
wcwuuom Houumoo
4w me am om N4 NV .I I. 6666. mesa
. . . . . . . .. . .. ooeoonomoo
mo mo we we mo mo oon mom. ooHaomHHE
. . . . . . . . H563 HomeoH
o H o H o H o H o H o H m H m H good HHoo 36H“
wcHHooa gon wcHHooa oooao oHoosm onaocoom oHoaHno>
unoguwz Hmucoawuomxm
zmem>m mos zmem>m mu HH<<

.mucosfinonxo mww HomNomofinmwpm> Hana: .HIM canes

76

 

 

1C)O -——-
0>
_._J -I-
C)
U)
955 -*-
x ~—
13
C3 __
.3 __ 2
(f)
4, -_
C: 9(3 --
0>
0 HI.
k.
0) _-
O.
_ 1
I l l .gj
855 I I I I I I
O 5 10 15 20 25 30
Sampling Intervol (s)
Figure 3-5. Effect of segmentation frequency and liquid flow rate on

%SS with the residence time of the sample slug in the mCF system held
constant at 110 s: l, n = 1.5 s’ , F = 0.0067 mL 5’1. 2, n = 3.0 5‘1
F = 0.015 mL s-l.

77

Snyder’s model assumes that mixing effects in unsegmented zones of the
CF system are negligible. As discussed in Chapter 2. this assumption
is weak even for CF systems with bubble-through flow cells. Other
factors being equal. mixing effects decreae as the flow rate increases.
At standard pump speed. approximately 20 seconds elapse between.the
time of sampling and the time at which the leading edge of the sample
slug is segmented with air. Doubling the pump speed halves this inter-
val. As demonstrated earlier. pecked sampling and bubble gating reduce
mixing effects. but clearly do not eliminate them. 'This set of experi-
ments suggests that the decreased dispersion of the sample slug
observed when the pump speed was doubled resulted primarily from a
decrease in mixing effects and that the concomitant doubling Of the
segmentation frequency had little effect.

3‘ Sppplp Interaction

Percent interaction (%I) was determined by the method of Thiers.
_L _l. [50]. Nitrite standards in the concentration sequence Of 5 pg.
45 11!. 5 a! were used throughout the experiments. Sampling intervals
for the AAII and mCF systems were fixed at 30 s and 25 s. respectively.
and %I was determined as a function of wash interval. When the second
low standard produced a shoulder peak. the peak height just prior to
the fall was recorded. Results Of these experiments can be found in
Table 3-2. On the AAII CF system with and without bubble gating. 1.0
%I was achieved using 10 s and 15 a wash times. respectively. On the
mCF system a 5 a wash time resulted in 1.0 %I and 0.5 %I at standard
and twice standard pump speeds. respectively. The decrease in %I at
higher flow rates is again presumed to result primarily from decreased

mixing effects as discussed previously.

78

 

 

 

an
em
ma
OH
m

 

 

I I I H.o Nd
H.ov H80 H8 :3 m5
:3 NH. NS 3. we
Nd 4.6 33 as H..H
.2. no oH .2 5H

3 gum H-m mHuc H838 6:55 a

5.5»... "as 55$ no :5

.cofiuomnoucw vacuums

nmuHm>Poucm awe:

.Nnm O~DMH

79

L Iixipg Efgegts

It is apparent from the foregoing experiments that mixing effects
contribute significantly to the Observed loss of wash in both of the CF
systems used for this study. I was able to minimize mixing effects in
the mCF system by using standard pump tubes with nominal flow rates
approximately half that Of the desired flow rate. and then Operating
the pump at about twice standard speed. Under these conditions analy-
sis rates which approximate theoretical limits assuming Snyder's 8 0t
criteria [10]. were achieved as is demonstrated in the final section of
this chapter. A better approach would be to use a miniaturized pump
fitted with shorter pump tubes. It is also possible to remove the
effects of longitudinal dispersion by means of curve regeneration as

described in Chapter 2.

1L Analytical Resulgs
1,, Nitrige Determinations

Aqueous nitrite standards were determined with the mCF analyzer
with the manifold shown in Figure 3-3. Note that the dilution pump
tube and first mixing coil (shaded area of Figure 3-3) were omitted to
increase sensitivity. A nitrite concentration range of 2 n! to 18 u!
was chosen to allow direct comparison of the mCF procedure with a
recent flow injection analysis (FIA) procedure for the determination of
nitrite [51]. The FIA procedure used the merging zones technique to
minimize reagent consumption and a novel 'intermittant flow' technique to
increase the sampling rate. Performance of the mCF system at several
sampling rates in terms Of linearity of standard curves. %I. and the

precision Of five replicate determinations of the 6 .1! standard can be

80

found in Table 3-3. Recordings of data Obtained at analysis rates of
360 hr.1 and 120 hr'1 are shown in Figure 3-6. Also listed in Table 3-
3 are experimental conditions and performance criteria reported for the
FIA.procedure.

Inspection of Table 3-3 reveals that even at a sampling rate Of
360 hr’l. where precision was limited because of manual sampling. the
performance of the mCF system is comparable to that Of the FIA system
which had a maximum sampling rate Of about 70 hr’l. Note also that on
a per assay basis. the sample and reagent requirements of the mCF
system were less than those of the FIA system. These results conflict
with the usual claims made about the performance of FIA relative to
CFA. ‘The sampling rate of about 70 hr"1 reported.by ZagattO. et a1.
[12] is somewhat superior to sampling rates generally achieved with
second generation CFA systems. but is decidedly inferior to sampling
rates achieved with the mCFA system. Furthermore. the FIA system
required an elaborate system of valves to inject both samples and the
reagent into the carrier stream. and to vary the flow rate of the
carrier stream during the course Of each determination. In this
respect the mCFA system is much simpler than the FIA system with the
exception of the gated detector.

‘2‘ Silicate Determination;

Aqueous silicate standards were determined with the manifold shown
schematically in Figure 3-7. All mixing coils*were made from 041 cm ID
Micro-line tubing. The lengths Of tubing used to make the first.
second. and third mixing coils were approximately 250 cm. 100 cm. and
50 cm. respectively. These provided about 230 s. 70 s. and 30 a

reaction times after the addition of the molybdate. tartaric acid. and

81

.Hm>nou:w 5mm:

mam oHnEmm one now confiscop oEHu on» wcfinsm vomo>wHov mucoumou omz mam z<m mo oas~o> vocwneoua
.mucoswhomxo :m pom m 2: up can AaIm m.Nu5 on uvoonm 9556

some oz oo.H oo.H oo.H oo.H oo.H oo.H H.Hmooo .Honuoov zxo
some oz wooo.o omoo.o Hooo.o wmoo.o mmoo.o «moo.o oaooooocH >
some oz mmo.o mmo.o mmo.o mmo.o mmo.o mmo.o ooon
H m~.o 4H.o am.o o~.o om.o mm.o H4
m.o m~.o 4N.o HH.o mH.o wm.o 44.o Hamzau :onHoooo
omH mm mm om om N4 mm Han HoesHo> snowmom
com wNH om om 4o 46 mm Han oasHo> oHosam

on oNH ONH owH owH cam com HHInzv oooz mHmzHoa<

II oH mH m CH m m HmVoeHP can:

II om mH mH oH 0H m HmVoEHP oHosam

 

5... 6 <5...

II] 111]] 1- 15

.m:0wpmcwsnouov Omaha“: OHPuoEwHoHoo pom <Hm mam <mus accrues :OmwhmnEou .mIm canny

82

 

0...-III III-I II .II
muunnwan;mumM ---IH-I
.I.III.I.IIII-.HI-u-IIIIh.I-.I ”III- I IHIHII IIIIIIhIIHI. _I.\s
.rIIIIIIIIII
II IIIIIII IIIIrIIII II I I II.“
K! II

 

IIII
FLJ I? U¥ LT—‘l
500

 

 

0.6 -—

0.5 —~

04-—
.3 -I-
02 ‘—

OOCOQCOmn<

Time (s)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LI
180 210 240

150

Time (s)

III III

 

IHIWIIIHIIIIIIIIINII IIIIIIIIHuIi I
fl L
1III‘I“
N 1
— — _ _ b —
— — _ _ _ _
6 a. A 13 2 .I 0
o O o o o O o

moconc6mo<

30 60 90 120

0

Recording of 2, 6, 10, 14 and 18 0M nitrate standards
B) 360 samples hr‘l.

run in ascending and descending order, followed by an interaction
test pattern and replicate determinations of the 6 uM standard.

A) 120 samples hr'l.

Figure 3-6.

83

NOMINAL

 

 

 

 

 

 

REAGENT FLOW RATE

(lemin) A

"' 7““ PUMP SPEED-56

: 91.0.: L n-2

0.05
A'R L——l 2.5m I.Om 0.5m WASTE
SAMPLE 0-23 ,
MOLYBDATE W
TARTARIc ACID °-'° '
STANNOUS 905 820nm
CHLORIDE

Figure 3-7. Diagram of the manifold used for silicate determinations

with the mCF system.

84

stannous chloride reagents. Silicate was determined by a reduced. 9"
lZ-silicomolybdic acid procedure [52.53] that has an absorption maxima
at about 820 nm. This procedure was adapted from an AAII CFA procedure
used to determine silicate in seawater [7].

Aqueous silicate standards with nominal concentrations Of 5. 15.
25. 35. and 45 n! were determined at several values Of t' and t'.
Data from these experiments are summarized in Table 3-4. Precision was
estimated from five replicate determinations of the 15 a! standard at
each sampling rate. Recordings of data Obtained at analysis rates of
180 and 120 samples h"1 are shown in Figure 3-8. Inspection of Table
3-4 reveals that precision and interaction were both about 1% or less
at all sampling rates. Linearity Of standard curves were also very
good. The maximum sampling rates that were achieved with acceptable
precision and interaction were less than for nitrite detrerminations.
but this was expected because the dwell time of samples in the silicate

manifold was three times longer than in the nitrite manifold.

1; Conclusions

At present. third generation CFA systems with bubble gated detec-
tors are not commercially available in single channel form. and the
high cost. complexity. and specializd nature of the Technicon SMAC CFA
system restricts its use to major clinical laboratories. This explains
in part why many academic and small industrial laboratories have
embraced FIA which is currently perceived in these circles as a versa-
tile. less costly. and more efficient alternative to CFA. The data
presented here. however. demonstrate that in conjunction with the

bubble gate circuit. high performance. miniature CFA systems can be

85

 

Table 3-4. Performance of mCFAczsystem for colorimetric silicate
determinations
Sample Time(s) 10 15 20 23
Wash Time(s) 10 5 5 7
Analysis Rate (hr-1) 180 180 144 120
Sample Volume (uL) 51 76 102 117
Reagent Volume (uL)Z> 137 137 172 206
Precision (%RSD) 1.1 0.8 0.4 0.4
Percent Interaction 0.07 0.90 0.97 0.16
$10pe .010 .0105 .011 .011
Y Intercept 0.000 0.000 0.000 0.000
rxy (correl. coeff.) 1.00 1.00 1.00 1.00

 

aPump speed = 56 (n = 2.0 5.1) and t = 300 s for all experiments.

Combined volume of molybdate, tartaric acid, and stannous chloride
reagents delivered during the time required for one sample and
wash interval.

86

 

Absorbonce
j
:I

 

 

 

 

 

 

 

04.: (1 .1 '1U[1 {1
..LCIIII. .. 1,077qu

I 1 1 I . I l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 90 ISO 270 360 450 540 630
Time (s)
0.5 -P
.._- II I
q) ..
U
co.-- I I
C
.D .I.
L
8 02-- I I
..D
<( ..
0.1 4-
" IIIII III I
0.0 , 4 4 E 4 I j 4 4 I 0: . i 4 4 , . 9 ; * s~1
0 60 120 180 240 300 360 420
Thne (s)

Figure 3- 8. Recording of 5,15,25,35 and 45 uM silicate standards
run in ascending and descending order followed by an interaction test
pattern and replicate determinations of the 15 uM standard. A) 120
samples hr 1. B) 180 samples hr 1

87

assembled and Operated with moderate ease, and that such CFA systems
need not be more elaborate or expensive than comparable FIA systems.
In fact. the combined cost of electronic components for the bubble gate
and a 2 9L bubble-through flow cell (:3500) required for the mCFA
system is comparable to the cost of sampling valves and valve control-
lers required.for FIA systems. vhile the cost of other major system
components (samplers. pumps. detectors) is about the same for either
technique. Hopefully the results reported here will encourage
researchers in academic and small industrial laboratories to think
seriously about custom designed mCFA systems. Such systems should be
particularly advantageous for the automation of routine colorimetric
determinations in vhich the analytical reaction requires more than
about 30 s to reach an appreciable degree of completion, especially
mhen lov dispersion of samples is required to achieve maximum sensiti-

vity. For an experimental comparison of FIA and mCFA. see Chapter 5.

W4

NOVEL CADNIUH REACTORS FOR DETERMINATIW OF NITRATE IN 'ATER

AND SEAWATER BY SEGMENTED. CONTINUOUS FLOW COLORIMETRY

A; Overview

Nitrite in water and seawater is routinely determined colorimetri-
cally with high specificity. sensitivity. and precision by diasotixa-
tion with sulfanilamide and coupling with N—(l-Napthhyl)-ethylene-
diamine as described in Chapter 3. An equally simple and sensitive
colorimetric assay for nitrate has yet to be devised. and for this
reason nitrate is generally determined as nitrite after it is reduced
to that species with a suitable reagent. Although procedures for
reduction of nitrate to nitrite with zinc [54]. hydrazine/Cu2+ [$5].
and immobilized nitrate reductase [56] have been reported. they have
not been widely applied. This is because the reduction with zinc is too
vigorous at ambient temperatures. reduction withhydraxine/Cu2+ is
sluggish and difficult to control. and enzymatic reduction. although
highly specific. requires reagents that are relatively expensive and
difficult to prepare. Cadmium metal has proved to be a much more
suitable reagent for reducing nitrate to nitrite. and is generally used
in the form of packed bed reactors. although reactors fabricated from

cadmium wire inserted into narrow plastic tubes [57.58] have also been

reported.

88

89

unfortunately, packed bed and tubular wire cadmium rmactors are
incompatible with segmented.Icontinuous flow analyzers because the
analytical stream must be debubbled prior to entry into these types of
reactors. Air segments cause channeling in the packed bed reactors and
as a result, reduction efficiency decreases and sample dispersion
increases. Tubular wire reactors will not support a stable segmenta-
tion pattern and so generally the analytical stream is segmented on the
downstream side of such reactors. By now it should be obvious that
other factors being equal, dispersion will be greater for a system in
which air segments must be removed and reintroduced downstream, than
for a system where segmentation is uniform and continuous. This means
that in determinations incorporating packed bed or wire reactors. the
sample and wash intervals required to achieve acceptable levels of
precision and interaction will increase. which of course decreases the
rate at which determinations can be performed. Furthermore. the flow
characteristics of packed bed reactors can change during the course of
a run due to accumulation of particulate matter. which will change the
back pressure. and therefore the flow rate through the reactor.

Reactors fabricated from cadmium tubing should eliminate these
problems. They would support a stable segmentation pattern in a
similar fashion to other mixing coils that make up the manifold. the
difference of course being that the inner walls of the cadmium tube are

active rather than passive.

5‘ General Consideratiggg

Reduction of nitrate to nitrite is generally 90 to 98% complete in

packed bed cadmium reactors (PBCRs) under neutral'to mildly alkaline

90

conditions. The dimensions of the reactor, as veil as the size distri-
bution of the cadmium particles with which it is packed. will have a
major influence on reduction efficiency. Larger reactors and smaller
cadmium particles lead to higher reducing power. It must be stressed.
however, that the active surface area of cadmium particles with similar
size distributions can be quite variable and may change during the
course of the reduction. For example Davidson and 'oof [59] reported
that different forms of cadmium -- e.g.. spongy. electrolytically
precipitated. filings, powder -- had different reducing powers on a
weight-to-weight basis. These conclusions are somewhat suspect because
they did not consider differences in the surface areas of the different
forms of cadmium. nor did they account for changes in pl! that are
likely to occur during the course of the reduction due to reaction of
dissolved oxygen with cadmium. Certainly one of the major difficulties
in achieving good precision with cadmium reactors is the maintenance of
a reproducibly active surface during the course of experiments.

There are several ways to increase the active surface area of cad-
mium particles. For example, if cadmium filings are reacted with
nitric acid prior to packing. their reducing power is increased [60].
This is because nitric acid pits the cadmium and thus the surface area
is increased. It is also well established [59,61] that amalgamated.
silverised. or capperised cadmium is more reactive than pure cadmium.
Amalgamated cadmium tends to form lumps [61.62] and is therefore not
well suited for the preparation of PBCRs. Copperized cadmium on the
other hand. does not lump and is highly reactive. For this reason it
has been widely used for the preparation of PBCRs. Several workers

2+

[62,63] have suggested that Cu is a product of nitrate reduction in

91

such reactors. Recently. however. Sherwood and Johnson [64] presented
convincing evidence that copper is not directly involved in the
reduction process; instead it functions primarily as a surface activa-
tor.

Other variables such as pH and the amount of dissolved oxygen and
chloride in samples and reagents also affect the reduction efficiency
of a given cadmium reactor. These effects may go unnoticed in rela-
tively large (25 cm x 0.8 cm ID) PBCRs used for batch reduction of
nitrate to nitrite in natural water samples [62]. because of the great
excess of cadmium relative to the amount of nitrate present in samples.
As PBCRs are miniaturized to make them compatible with continuous flow
analyzers [65.49]. however. control of reaction parameters becomes much
more critical. and changes in the active surface of the cadmium that
can occur during the course of the reduction become more apparent.
Nydahl [61] reported a detailed study on the cptimum conditions for the
reduction of nitrate to nitrite with PBCRs. This study was highly
useful in my work with open tubular cadmium reactors (OTCRs). I will
therefore summarize his results in the paragraphs that follow to
provide some background information necessary to understand the results

with OTCRs presented later.

92

_C_,_ Summary; mghahl's !Qrk
L pl_i Effects
Equations 4-1. 4-2. and 4-3 [61] suggest that

no; + 211* + 2e'—>N02’ + 320 (4—1)
we; + 611* + 4.’—>33Non* + 1120 (4-2)
no; + 311* + 5.'—->Nn4+ + 2n20 (4-3)

the pi! at which the reduction occurs has a major influence on reduction
products. Nydahl [61] found that in the pH range of l to 3. hydroxyl-
amine is the major product. Small amounts of ammonia are also formed.
Reduction rates decrease sharply in the pH range of 3 to 5. and nitrite
is the major reduction product. The rate of nitrate reduction con-
tinues to decrease in the pH range of 7 to 9. and 95 to 994% of nitrate
originally present in samples is recovered as nitrite. The fraction
depends on the flow rate through the PBCR and its size. As the pi!
increases in this range. either the size of the PBCR must be increased.
or the flow rate of sample through it must be decreased to achieve the
same degree of reduction. The rate of reduction of nitrite to
hydroxylamine also decreases as the pl! increases. but even at pl! 9.5.
slight reduction of nitrite still occurs. Nydahl reported the same
trends for copperized cadmium but found that the rate of reduction
observed for both nitrate and nitrite was a factor of 10 to 20 times
greater for copperized cadmium than for pure cadmium. These results
emphasize the importance of pH control during the reduction process for

all types of cadmium reactors. The presence of dissolved oxygen in

93

samples and reagents makes pH control more difficult than might be
expected.

2‘ Qissoived ngggg Effects

Cadmium(II) ions are a product of the reduction of nitrate to

nitrite as shown in Equation 4-4.
- + 2+ -
Cd + N03 + 2!! ->Cd + No2 + £20 (4—4)

Because hydrogen ions are consumed in this reaction. the pH of unbuf—
fered samples increases slightly during reduction. For samples con-
taining micromolar concentrations of nitrate. however. this increase is
negligibly small. Nydahl [66] pointed out. however. that a much more
significant increase in pH occurred within reactors because dissolved
oxygen in samples and reagents reacts quantitatively with cadmium as

shown in Equation 4-5.
ca + 1/202 + 2n*—>Cd2* + 1120 (4—5)

Under mildly alkaline conditions this reaction proceeds at a rate about
30 times faster than the rate at which nitrate is reduced to nitrite.
The concentration of oxygen in dilute aqueous solutions in equilibrium
with air is on the order of'0.5 mg-At L"1 [67] which. in accordance
with the reaction stoichiometry. is equivalent to 1 mmol of base per
liter of solution passed through a PBCR. For this reason. the buffer
capacity of the sample must be much greater than would be predicted on
the basis of nitrate reduction. if constant pfl'is to be maintained
throughout the reduction process. Note also that oxygen reduction

42*

simultaneously increases the concentration of C to a value of

94

approximately 0.5 ml_l. Nydahl estimated that at this concentration.
cadmium hydroxide would begin to precipitate at a pH of about 8.5. and
was likely to adhere to the surface of the cadmium particles. As a
result. the reducing power of the reactor would decrease. This can be
prevented if compounds such as ammonium chloride or EDTA that form

soluble complexes with Cd2+

are added to the sample prior to reduction.
Nydahl [61] suggested the use of imidazole (pg-7.09) which can be used
to buffer samples and complex Cd2+ (84:17.5) simultaneously.

L M 21 Chlorigg

Nydahl [61] found that chloride ions greatly retard the rate at
which nitrate is reduced to nitrite. Thus the reduction of nitrate in
seawater ”Cl-120.3 )1) proceeds more slowly than in fresh water. This
is generally of little consequence when PBCRs are used because of their
relatively large reducing power. In fact. because chloride ions also
decrease the rate at which nitrite is reduced to hydroxylamine. yields
of nitrite may actually be increased in the presence of chloride.

especially when highly reactive capperized cadmium is used as the

re duc ing agent.

Q. Experimental

L Reagents
Sulfanilamide (SAN) and N-(l-Naphthyl)ethyenediamine (NED)

reagents were prepared exactly as described in Chapter 3.

The ammonium chloride reagent was prepared by dissolving 10 g
(0.19 mol) of ammonium chloride in l l. of 00'.

Buffer solutions (0.1 )1) with nominal pH values of 5.0. 6.0. 7.0.

and 8.0 were prepared from acetic acid and sodium acetate.

95

2-(N-morpholino)ethanesulfonic acid (HES) and sodium hydroxide.
imidazole and hydrochloric acid. and tris [hydroxymethyl] aminomethane
(TRIS) and hydrochloric acid. respectively. These buffers were chosen
because in each case. their pK.s or plbs are within 0.25 units of the
desired pH. Brij-35 surfactant (1 ml. L-l) was added to each buffer.

Copper sulfate (0.01 _l_) was prepared by dissolving 2.5 g of
CuSO4'5H20 in l L of DDW.

2... W

Primary (5 m!) and working nitrite standards were prepared exactly
as described in Chapter 3.

Primary nitrate standards (5.0 m!) were prepared by dissolving
0.5056 g of dried. reagent grade KN03 in 1 L of DDW. 'orking standards
were prepared by dilution of the primary standard by means of digital
microliter pipets and volumetric flasks.

Seawater standards (spikes) were prepared as follows. Volumetric
flasks (100 mL) were filled to the mark with low-nutrient filtered sea-
water. Then l.00 mL of the seawater was withdrawn from each flask with
a digital pipet and discarded. Next the required volume of primary
nitrate or nitrite standard was added to the flasks by means of digital
microliter pipets. The flasks were then diluted to the mark with DDW
and mixed. This procedure ensured that the dilution of the seawater
was small (150 and that the dilution was uniform regardless of the
concentration of the standard prepared.

The filtered. low nutrient seawater. collected in midsummer from
Long Island Sound. was a gift of the Oceanographic Sciences Division of

Brookhaven National Laboratory.

96

1,, ngmium Reactor;

Packed bed cadmium reactors were prepared as follows. Cadmium
filings (40-60 mesh) were stirred with a magnet to remove ferrous metal
and then washed with methylene chloride to remove grease. The filings
(:30 g) were air dried on filter paper in a hood. and then transferred
to a 100 m1. beaker. The filings were etched with 25 mL of 8 g nitric
acid for about 1 minute and then rinsed thoroughly with DDW. 1.2 !
hydrochloric acid. and again with DDW. Next the filings were treated
with several portions of the copper sulfate solution until the blue
color no longer faded from the solution. The capperized filings were
then rinsed with ammonium chloride solution to remove colloidal copper
and cadmium hydroxide. COpperized cadmium filings prepared in this way
can be stored indefinitely under a solution of ammonium chloride.

PBCRs were prepared in small glass tubes (7 cm x 0.2 cm ID) with a
constriction at one end. A small plug of glass wool was pushed to the
constricted end of the tube with a stiff wire. and a small funnel was
attached to the unconstricted end of the tube by means of a plastic
sleeve. The funnel and tube were filled with ammonium chloride solu-
tion and copperized cadmium filings were added to the funnel. one
spatula tipful at a time. Between additions. the walls of the tube
were tapped with the spatula to ensure uniform packing. 'hen the
filings were within 0.2 cm of the top of the tube. a second plug of
glass wool was added to secure the filings. The funnel was then
removed and the PBCR was ready to install in the manifold.

Open tubular cadmium reactors were prepared from 0.127 cm
ID x .229 cm 0D cadmium tubing (Reactor Experiments. San Carlos. CA).

Tubing was cut into desired lengths and wrapped around a 1 cm diameter

97

glass rod to form a closely spaced helix. Next the ends of the coiled
tube were connected by a short length of plastic tubing. and the
exposed surfaces of the cadmium coil were sprayed with several coats of
acrylic lacquer. This allowed the reactors to be handled freely
without fear of cadmium poisoning.

A syringe was used to wash the inner walls of the OTCRs with
methylene chloride. which was subsequently removed by flushing the 0TCR
with nitrogen for several minutes. The OTCRs were copperized as
follows. A 50 mL beaker was filled with equal volumes of imidazole
buffer (pH-7.0) and copper sulfate solution (0.1 l). A syringe was
attached to one end of the OTCR with a plastic sleeve and a short piece
of plastic tubing was attached to the other end of the 0TCR. The
syringe was then used to draw the buffered cepper solution into the
OTCR. The solution was forced back and forth through the reactor for
several minutes. after which the reactor was ready to be installed in
the manifold. 'hen the OTCR was not in use. it was removed from the
manifold. filled with imidazole buffer. and sealed by joining the ends
with a short length of plastic tubing.

Cadmium foil reactors were prepared from a 0.0025 cm cadmium foil
(Alfa Ventron. Danvers. NA). The foil was polished with steel wool.
rinsed with methylene chloride. and dried between tissues. The foil
was then mounted between standard 6 or 12 inch dialysis blocks (#178-
2538-07 or 178-2540-01. Gamma Enterprises. It. Vernon. NY). designed
for use with the Technicon SMAC clincial CFA system. The foil was then
capperized within the dialysis blocks by the same procedure used for

OTCRs .

98

1‘ Nanifolgs

Nanifolds used for nitrate and nitrite determinations with OTCRs
and PBCRs during these experiments are shown schematically in Figures
4-lA and 4-1B. respectively. Note that the flow rates of the air
delivery pump tubes are twice the normal values in the manifold used
with OTCRs. Recall from Chapter 2 that the minimum air segment volume
necessary to sufficiently occlude a tube of diameter dt is equal to
0.92 dt3. The only cadmium tubing commercially available had an inside
diameter of 0.13 cm. so an air segment volume of about 2 9L was
required. This is about twice the volume required when dt . 0.1 cm. as
is the case for other components that make up the manifold.

5‘_ pg leasurements

A11 pH measurements were made with an Orion 611 digital pH meter
and a Ross combination pE/reference electrode (Orion Research. Inc..

Cambridge. MA.)

_E_,_ Some Definition;

 

As discussed in the previous section. both nitrate and nitrite are
reduced to lower oxidation state species in cadmium reactors. and the
extent of reduction depends primarily on the size of the reactor. the
flow rate of solution through the reactor. and the pH and chloride con-
centration of the solution in contact with the reactor. Since only
nitrite reacts with the Griess reagents. ideal experimental conditions
would be those where reduction of nitrate to nitrite is quantitative.
while reduction of nitrite to lower oxidation state species is negli-
gible. In this case determination of nitrite in the cadmium reactor

effluent would yield the sum of nitrite and nitrate originally present

99

.muouomop sawsvmo won moxomn you soumxw Am .muouomou ezwsvmo awom mam magmas» mono uom.eoumxmmmm<
.ucoempumcfi <mue ecu saw: ouwpuwc mam mummy“: ocwshouov ou vow: mpfiomficms mo memummwo Hue on an

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3.0 a!
Sol: 25.
.ImmTWI“ m:«
o. .o
r. l l. I
5.58 54883 xmnazou
1 I J _ 0.6 \mmmhsm
. . f 1 . . “Jasa
e o . En o mood . «.4
Bass
- mum:
on . omwam axon.
Am
E 9.... am:
33%;
mpg; 52 «z
I u .
on. ommam 25¢ Hum 30.:

 

Adi=sxxg AnN

100

in the sample. An independent determination of nitrite would then
allow the nitrate concentration to be calculated by difference. For
automated nitrate determination procedures that use small PBCRs for
reduction. these conditions are generally closely approximated. The
surface-to-volume ratio of such reactors is large. which favors rapid
reduction of nitrate to nitrite. while the mean residence time is
short. which minimizes reduction of nitrite to lower oxidation state
species. The surface-to-volume ratio is much less favorable for OTCRs.
and a large number of experiments had to be performed before results
for nitrate determinations in water and seawater obtained with OTCRs
were comparable to those obtained with PBCRs. Details of these experi-
ments are summarized in the next section.

In future discussions a quantity called percent recoveg is used
to provide a quantitative measure of the reduction that occurs in a
cadmium reactor. Percent recovery is defined as follows. First con-
sider solutions that contain only nitrite. In this case percent
recovery is the ratio (x 100) of absorbances measured for the end
product of the Griess reaction when the same nitrite standard is deter-
mined with and without a cadmium reactor installed in the manifold.
For example if absorbances of 0.587 and 0.603 were measured for the
same nitrite standard with and without a cadmium reactor on line. the
percent recovery would be 97.3 [(0.587/0.603) x 100)]. This implies
that with the cadmium reactor on line 2.75 of the nitrite originally
present in the standard was reduced to lower oxidation state species
that do not form colored products with the Griess reagents. For solu-
tions that contain nitrate. percent recovery was calculated by dividing

the absorbance measured for a nitrate standard with a cadmium reactor

101

on line. by the absorbance measuredIfor a nitrite standard with the
same nominal concentration after the cadmium reactor had been removed
from the manifold. In this case. however. some of the nitrite produced
from nitrate reduction may be further reduced to lower oxidation
states. and so percent recovery will reflect the net yield of nitrite

resulting from these two competitive reduction reactions.

E; Preliminary Experiments

1‘ Open Tupular Cadmium Reacpors

Results obtained in preliminary experiments with OTCRs conformed
to trends reported.by Nydahl [61] for nitrate reduction with PBCRs.
Recoveries of nitrate were in the range of only 5 to 10% for untreated
OTCRs. These low nitrate recoveries could not be attributed to nitrite
reduction because recoveries calculated for nitrite standards were in
the range of 98 to 99%. When OTCRs were activated by perfusion with a
solution of 0.01 5 copper sulfate. however. nitrate recoveries
increased dramatically. Percent recoveries calculated for 25 n!
nitrite and nitrate standards as a function of copperized OTCR length
are given in Table 4-1. As expected percent recovery calculated for
the nitrate standard increases as the length of the OTCR increases. but
there is no clear trend in percent recovery calculated for the nitrite
soltution. In these experiments. the buffer/complex reagent (see
Figure 4-1) was an unbuffered solution of ammonium chloride containing
copper sulfate (100 uL of 0.01 ! CuSO4 per 100 ml. of 0.19 ! NH4C1).
The pH of this solution after mixing with the sample in the first coil
was 4w58. After passage through a 15 cm copperized OTCR. the pH

measured was 8.08. an increase of 3.5 pH units. Variation of the

102

Table 4-1. Percent recovery of nitrite and nitrate standards for
copperized open tubular reactors of various lengths.

 

a
Percent Recovery

 

 

 

OTCR Length (cm) N02 N03
3 93.0 42.2

6 96.6 61.8

9 96.9 82.1

12 98.3 87.8

15 98.2 91.8

18 98.3 92.4

21 95.3 91.3

30 96.1 93.7

 

aBased on absorbance values recorded for 25 pH nitrite and nitrate
standards as described in text.

103

nitrate concentration in the sample (0 to 35 pl) had no measurable
effect on these results. which indicated that the pH change was due to
oxygen reduction as reported.by Nydahl. Thus the somewhat erratic
percent recoveries listed in Table 4-1 are attributable to changes in
pH that occurred during the course of reduction.

When this experiment was repeated sequentially with a 0.1 !
acetate buffer (pH-=4.75). a 0.1 Ll NES buffer (pH-6.0). a 0.1 !
imidazole buffer (pm-7.0). and a 0.1 g Tris buffer (pH-8.2) in place of
the ammonium chloride/copper sulfate reagent. the pH of solutions
entering and exiting the OTCR remained constant to within 041 p8 unit.
Slepes. intercepts. and correlation coefficients for calibration curves
calculated from the absorbance of 5. l5. and 25 pl nitrite and nitrate
standards with and without the OTCR installed in the manifold at each
pH can‘be found in‘Table 4-2. Also listed in this table are percent
recoveries calculated for nitrite and nitrate standards on the basis of
the ratio of the slopes of the calibration curves for nitrite and
nitrite at each pH; Since the slopes of the calibration curves for
nitrite at each pH were linear and constant to within about 45 without
an OTCR on-line. the low percent recoveries calculated for nitrite and
nitrate at the two lower pH levels cannot be attributed to changes in
the sensitivity of the Griess reaction when different buffers were
used. Instead low recoveries with these buffers are due to reduction
of nitrite to lower oxidation state species (notice the large. negative
y-intercept values) as observed by Nydahyl in his studies of PBCRs in
this pH range. Percent recovery of nitrite is maximum at pE-8.2. while

percent recovery for nitrate is maximum at pR-7JL

104

.ofiomficme can Eoem vo>oEou
«who one new: mcnmvcmum cameo“: you woummmos moocmanmnm Eonm voumfimofimo one: monogamoumm cw mo=Hm>o

 

 

 

 

 

o.we w.om w.Hm o.ma m.~m e.wm H.om ~.oe sao>oooa w
“mam. ”mamm.voooo.m mama. fleamm.voooo.m Rama. hemmm.vmmmm. emma. hemmm.vwmmm. axe
ooo.- fiNoo.-Veoo.- moo.- ”moo.-vooo.- emc.- Avoo.vmmo.- ”Ho.- Amos.-V¢~c.- eaooaoeem-»
mama. ”memo.voe~o. moms. AwVNo.UemNo. ono. Awmwo.vmmmo. «boo. NeaoVNo.Ummoo. odomm
-moz -Noz -moz -moz -moz -moz -moz -Noz

c~.w II oc.e cc.c me.v ":8

 

.mn mo cowuocmw a ma mvumvcmum
momma“: can mummy“: pom woumfisofimo mowno>ooou ucooeom mam muouoemnma :oflmmouwom neon“; .Nuv oanmh

105

Percent recoveries of nitrite and nitrate were further investi-
gated over the pH range of 6.5 to 8.5 using imidazole (pH-6.5. 7.0.
7.5) and Tris (pH-8.0. 8.25. and 8.5) buffers. A short (10 em) OTCR
was used in these experiments so that changes in percent recoveries of
nitrate would be more apparent. This experiment was also performed
with a cepperized PBCR using the manifold shown in Figure 4-lB. Per-
cent recoveries calculated from experimental results obtained with the
OTCR and the PBCR are plotted as a function of pH in Figure 4-2A and 4-
28. respectively. As before. percent recoveries were calculated on the
basis of the ratio of the slopes of nitrite and nitrate calibration
curves (5. l5. and 25 pl concentrations) with and without a cadmium
reactor on line. Nitrite recoveries increase slightly as the pH
increases in experiments with both reactor types. as expected. Nitrate
recoveries calculated from data obtained with the PBCR also increases
as the pH increases in agreement with Nydahl's study. Nitrate re-
coveries calculated from data collected with the OTCR. however. first
increases slightly as the pH increases. and then decreases sharply with
further increases in pH. Although this trend can be rationalized in
terms of the decrease in the rate of nitrate (and nitrite) reduction
that occurs as pR increases. it seems likely that the ability of

d2+ is also involved. There was no complexing

imidazole to complex C
agent in the Tris buffers and so formation of Cd(OR)2 on the surface of
the OTCR was a distinct possibility. This seems even more probable in
light of some later experiments that are discussed shortly. Throughout

these experiments nitrate recoveries were always highest when imidazole

buffers were used.

106

A)
'00 F W

(O
O
l

m
0

fl
0
I

O)
O
U

 

Percent Recovery
T

50£

(51’ I 41 l l J
6.0 6.5 7.0 7.5 8.0 8.5

 

B) pH

>~

§l00-

E; 535).

o: 98-

C 97 b 0N0;

£95, 0N0;
95-

0% L 4 1 1 1

6.0 6.5 20 75 8.0 8.5
pH

Figure 4-2. Percent recoveries of nitrate and nitrite. A) Open
tubular cadmium reactor. 8) Packed bed cadmium reactor.

107

In the next set of experiments percent recoveries were determined
for nitrite and nitrate standards in the concentration range of 5 to
25 '1! using the three imidazole buffers from the previous experiment
and a 30 cm copperized OTCR. Slopes. Y-intercepts. correlation coef-
ficients. and percent recoveries calculated for data from these
experiments can be found in Table 4-3. These results were encouraging.
Recoveries for both nitrite and nitrate were in the range of 95 to 97‘.
When low nutrient seawater that had been spiked with nitrate was deter-
mined with this OTCR. however. nitrate recoveries were less than 80%.
although the calibration curves remained linear. This observation was
not unexpected because Nydahl [61] reported a sharp decrease in the
rate of nitrate reduction when the chloride concentration of the solu-
tion in contact with PBCRs increased. A plot of nitrite and nitrate
recoveries as a function of chloride concentration is shown in Figure
4-3. Here calculations of percent recoveries are based on the absor-
bance measured for 25 ll! nitrite and nitrate standards to which sodium
chloride was added. Imidazole buffer (pH-7.5) and a 15 em copperized
OTCR were used for this experiment. At this point it became apparent
that longer OTCRs would be necessary to approach quantitative reduction
of nitrate in a seawater “Cl-1:03!) matrix. When a 60 cm copperized
OTCR was installed in the manifold. however. the segmentation pattern
of the analytical stream emerging from the reactor was irregular.
recorded baseline signals were noisy. and peak shapes were severely
distorted.

These problems were initially attributed to irregularities or par-
tial blockages within the reactor. When several shorter OTCRs that

individually supported stable segmentation patterns were connected

108

.vfiomwnme can seem uo>oaou
2090 on» saw: mvpmvcmum comma“: mom poemmmos moozmpuomnm scum voumfimomwo one: monogamonmm a“ momam>o

 

 

 

 

 

 

0.0m 2.5m 0.8m ~.cm o.ma o.mm muo>ouo¢ 3
oooc.~ Rocco.mvoooo.m oooo.~ floooo.mvmmmm. oooo.~ .flcooo.mvcoco.m axe
Hoo.- flooo.-vmco.- voo.- flees.-vmoo.- moo.- ”woo.-vwoc.- udooaoeem-s
OMNO. ameD.VHm~o. ammo. ”ammo.vm-c. ammo. ”VHGMNo.Ue~No. odomm
Inez -moz -moz . -Noz -moz -Noz
m.e . o.e m.o. ”mg

 

.mupo vouwnonmoo Eu cm a can muommmb oHonvflaw new: mvuevcmum
comma“: was museum: mom woumfisu~mo mofiuo>ooon acoouom new muouoammmm cowmmouwou among; .nue oHnmh

109

 

80

Percent Recovery
V
o

O)
O

 

50¢

°o.o 6| 0.2 0.3 6.4 6.5 6.6 6.7 6.8 69
Chloride Concentration (M)

Figure 4-3. Effect of chloride concentration on percent recoveries
of nitrate and nitrite with open tubular cadmium reactors.

 

 

110

together. however. the same symptoms were observed. ‘This was puzzling
at first. but it finally became apparent that oxygen in the air
segments was being reduced at the reactor walls. which caused the air
segments"volume to decrease. In longer OTCRs. the air segment volume
became too small to occlude the tube and so the integrity of the
analytical stream was lost. When nitrogen rather than air was used as
the segmentation gas. stable segmentation patterns were achieved for
OTCRs as long as 150 cm. At this point. the p11 of the buffered analy-
tical stream was measured before and after it had passed through the 60
cm OTCR. first with air. then with nitrogen as the segmentation gas.
Results are shown in Table 4-4. The buffer capacity was clearly
exceeded when air was used as the segmentation gas and a 60-cm OTCR was
on-line. which confirmed the hypothesis that oxygen in the air segments
was being reduced at the OTCR walls (see Equation 4-5) and also
explained the distorted peak shapes and noisy base lines observed when
the seawater spikes were passed through the 60 cm OTCR with air as the
segmentation gas. As is demonstrated in the final section of this
chapter. nitrate recoveries comparable to those measured with PBCRs can
be achieved with a 60 cm OTCR when nitrogen is used to segment the
analytical stream.

2‘ Cadmiumqui; Reactors

As mentioned earlier. the ID of commercially available cadmium
tubing is somewhat larger than the ID of other manifold components in
the mCF system. Recall that. other factors being equal. dispersion in
gas segmented reactors increases as the inside diameter of the reactor
increases. It occurred to me that a cadmium reactor with improved wash

characteristics could be fabricated by inserting a piece of cadmium

111

Table 4-4. Effect of segmentation gas on the pH measured for the
effluent from a 60 cm OTCR.

 

Measured pH

 

 

 

 

 

Segmentation
Gas: Air Nitrogen
Nominal Distilled Distilled
pH Water Seawater Water Seawater
a
6.5 6.80(6.54) 6.96(6.63) 6.52 6.64
7.0 7.35(7.01) 6.38(7.11) 7.01 . 7.12
7.5 8.59(7.49) 8.38(7.64) 7.50 7.64

 

apH values in parentheses were measured before the buffered analytical
stream entered the OTCR.

112

foil between the blocks of a dialyzer designed for use with the SNAC CF
system. This approach was tried. but unfortunately it proved very
difficult to produce a good seal between the dialysis blocks and the
cadmium foil. Even when a seal was achieved temporarily. leaks began
to develop within a few hours of operation time. This was not
considered acceptable for routine use. In the few successful experi-
ments with 12 inch dialyzer blocks. recoveries calculated for nitrite
and nitrate standards were on the order of 977- and 75%. respectively.
It did appear that sample interaction was lower in this reactor than in
OTCRs. although insufficient data were collected to make a definitive
comparison. In these experiments the analytical stream was buffered
with imidazole (pH-7.0) and segmented with nitrogen. These results
were encouraging. and if a solution to the problem of forming a good
seal between the dialysis blocks and the cadmium foil could be solved.

this system would warrant further study.

Qp Einal Experiments

The results of the final experiment in which the performance of a
60 cm cepperized OTCR and a 7 cm x .2 cm ID cOpperized PBCR were
compared for both distilled water standards and seawater spikes are
summarized in Tables 4-5 and 4-6. The concentration range of standards
and spikes for both nitrate (0 to 25 p!) and nitrite (0 to 2.5 u!) was
chosen to be representative of concentrations normally encountered in
unpolluted seawater. Data listed in Table 4-5 resulted from five
replicate determinations of each standard and spike. Three sets of
standards and spikes were prepared for this experiment. One set con-

tained only nitrate. another set contained only nitrite. and a third

113

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4-5. Data from determinations of nitrate and nitrite in
distilled water and seawater with OTCRs and PBCRs.
Baseline Corrected Absorbancea
Nominal
Concentration
(uM) OTCR PBCR
- Distilled Distilled
N02 + NO3 Seawater Water Seawater Water
0.5 + 5.0 .1491.001 .1401.001 .l35t.001 .136i.002
1.5 + 15.0 .4121.001 .409i.001 .4071.001 .4071.001
2.5 + 25.0 .660r.001 .669:.001 .6681.001 .6681.003
NO3
5.0 .127.+..002 .127i.001 .124i.001 .1211.001
15.0 .373 :t .001 .374 i .003 .366 r .001 .368 i .002
25.0 .6111.001 .607i.001 .6111.001 .6011.002
NO2
0.5 .Ollt.001 .009:.001 .0111.001 .0121.001
1.5 .035:.001 .0321.001 .03Si.002 .037i.002
2.5 .065:.001 .059:.001 .0621.001 .0641.001
(Cadmium Reactors Off-Line)
0.5 + 5.0 .013i.001 .011:.001 .0101.001 .0121.001
1.5 + 15.0 .037i.001 .0381.001 .0371.001 .040i.001
2.5 + 25.0 .0621.001 .063:.001 .0601.001 .0651.001
N02
0.5 .013i.001 .011¢.001 .0111.001 .0121.001
1.5 .036r.001 .0381.001 .037:.001 .039:.001
2.5 .067i.001 .0631.001 .0641.001 .0671.001

 

aAverage of five replicate determinations.
.0041 .002.

Refractive index correction

114

 

 

 

 

.momceucou ore ofimfiw

xx

 

 

 

 

 

 

 

 

mama. mmmm.e mama. mama. e

nmm. amm. mmm. mmm. .eea mo eoeeo .mem

Sm. e 8m. 8m. a 8m. mmm. e mmm. mmm. e mmm. .m.m e mama—See;

mmmmm. e mamam. mmmmm. e e28. mmmmm. e Seam. mmmmm. e. Seam. .m.m e emcee

8:358

mamm. mama. amam. mmmm. axe

_mm. emm. mmm. mmm. .eaa ea eoeea .eem

Nam. e So: So. e Nam: 8m. e «mm: So. an mm: 5.3 emaoeoemeIa

ammm. e 88. ammm. e 3.8. ammm. e mmem. ammm. e maNm. .m.m e emcee

-Noz

mmmm. mmmm.e mmmm.e mama. axe

«mm. _mm. mmm. mmm. .eae mo eoeeo .mem

8m. e 8m. 8m. e Nmm. 8m. e amm. Nmm. e m8. emooeoeeeIa

emmm. e emam. ammmm.nammm~m. ammmm.nemme~m. emmm.ne emem. emcee

Inez

mama. mama. mmmm.e mama. axe

mmm. mmm. mmm. mmm. .eea ea eoeeo .mem

mmm. e mmm. 8m. x «mm. 8m. e mmm. 8m. e mam. .m.me emooeoemeIa

ammm. e 2.8. mmmmm. a 23m. mmmmm. e mmxam. mmmmm. e mammm. .m.m e emcee
Hope: ecumzmom noun: ecumzmom Inez + -Noz

Imoemweaem maeemeemm

mama ammo emoeaeoewem

.mav ofinmh cm woummm meme eom meoaoememn cowmmouucm emcee; .ose smack

115

 

 

 

axe

.emo mo eoeeo .mum
uncoeouceua

oaoem

 

accessou
axe
.umo mo eoeeo .mum
«acoeouceua
omoem

 

 

mama. Nmaa. mama. moaa.
moo. moo. goo. moo.

8o. H So... So. H woof So. H go? So. H So...
mooo. H 3.8. mooo. H ammo. mooo. H memo. mooo. H ammo.
mamm. aaaa. mama. coma.
com. com. moo. mac.

80..” woof 89H Non..- 39H Nocf 89H moo...
aooo. H mama. Boo. H memo. Boo. H memo. aooo. H came.
mama. Haaa. mama. amaa.
Ham. moo. moo. coo.
So. H Sci 8o. H woof 8o. H woof 8o. H 8o.
aooo. H memo. Boo. H ammo. aooo. H memo. Boo. H memo.

eoewz eoemzmom ecumz ecumzmom
vofifiwumem wofifieemam

 

mzmqaumo mm0ku<mm 23H2o<u

N

02
axe

.Hmm MO HOHHO UHNVflMHW

.m.m H emoueaemeIa
.mmeeeza

 

aoz + Noz

 

mmonmoewoz

 

.moameemom me memee

116

set contained both nitrate and nitrite in the concentrations indicated
in Table 4-5. The latter two standard and spike sets were also deter-
mined with the OTCR and PBCR removed from the manifolds shown in
Figures 4-lA and 4-lB. respectively. Imidazole buffer (pH-7.0) con-
taining 100 uL of 0.01 ! Cu804 solution per 100 mL was used for both
systems. Nitrogen was used to segment the analytical stream for deter-
minations with the OTCR. Sample and wash times were 23 s and 7 s for
experiments with the OTCR. and 40 s and 20 s for experiments with the
PBCR. These values resulted in percent interaction of 1.0% or less for
both systems.

Strip chart recordings of peaks obtained for nitrate and nitrite
determinations in seawater with the OTCR and PBCR are reproduced in
Figure 4-4 and Figure 4-5. respectively. Part A of both these figures
shows peaks recorded for seawater spikes that contained only nitrite.
The noise on these peaks resulted from refractive index differences
between the distilled water used as the wash solution and the low
nutrient seawater in which the spikes were prepared as discussed in
Chapter 2. Peaks shown in part B of Figure 4-4 (OTCR system) are for
seawater spikes that contained both nitrate and nitrite. while those
shown in part B of Figure 4-5 (PBCR system) pertain to seawater spikes
that contained only nitrate.

Figure 4-6 shows the correlation between absorbances measured with
the OTCR system and absorbances measured with the PBCR system. In this
figure the baseline (and blank. in the case of seawater spikes) cor-
rected absorbance values determined for all standards and spikes with
the OTCR system are plotted against those obtained with the PBCR system

(see Table 4-5). The slope (0.999 + 0.005). standard error of the

117

I
—. --o-b
-I
'1
' l
I
a-—. -
- — .— a.
-0
“-

 

 

 

 

 

 

I T..-” - --A '-- I
.3 ‘ I .
'I ; '
.. .'...--- _‘E . I I I I I '
.TL 4 ___ I - ~“‘—_.‘ .m' - -- e... &—-_
I . I . I I ' ‘ I
l . . . . . , ‘ . .
7---m J _ -'---—:II—-——1-.-70— I I JI——
' , ' ' II fl '5: 'I ' I I
3.--- - I-i-'—-'—III:IIII7II I .. --.----..;...II—I- '

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

d -- .— -—-—-—“r—- -—--- - lo -- <- -I<~ >4 v-U- I— I «o - o—— ._ .......Y —I .4 f -
I I I : ' ' ‘ I I
IL‘ -qt- _4I—-——T—e—. e—A‘L—--.-— -- ..PI) «>144» “—4-. .0. ..xI—I J -.l.._... -.. _é- . . - --
I h I I . I . . , I
I 4 :
qp—- — -T- I dy‘ a- - ,—-—.J ..I 1.---..
. I z
- I g _ . . I I
-—J .-.---'-—- «---J«-«I *“I‘ I- ~ J-I-
I 9 - . :
! I ' F F I . '
5“”‘I ”r “I“ ‘II ‘II‘rv- -~1~I Ir“ -
I ‘ ' I I 5
I ' . I
l I I e I
. n
'I I _, -.. .' --ab. «I I-I ..n 1 [I1 .— vq I .1 —.
i I I a ‘ I I
. l I I . I
”1* - -‘ r 1 5 :1 - l x
I I .
\- I I '} TI 7
I -- 1' 1
I
r C
-i. , -_‘ _ r I - ,
_ E I ' . I _
' ' D ' . ‘ 4
, _. _.._1-_...-l_-__. .. . ..-- _ -_ _ o , «'4- fl .- _ . -. -

 

 

Figure 4-4. Peaks recorded for seawater spikes with a 60 cm open
tubular cadmium reactor. A) Nitrite spikes. Nominal concentrations
from left to right: blank, 0.5, 1.5, and 2.5 uM. B) Nitrate plus
nitrite Spikes. Nominal concentrations of NOéVTNoj'from left to
right: 0.5-+5.0, 1.54-15.0, 2.54-25 uM. Sample and wash times

were 23 s and 7 s, respectively. '-

118

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I
s
I
-i-.
I

 

Figure 4-5. Peaks recorded for seawater spikes with a packed bed
cadmium reactor. A) Nitrite spikes. Nominal concentrations from
left to right as in Figure 4-4. B) Nitrate spikes. Nominal concen-
trations from left to right: 5.0, 15, and 25 pg, Sample and wash
times were 40 s and 20 s, respectively.

119

0.7

.0
m

Q
0|

Absorbonce (PBCR)
Q o
o: 45

Q
N

9

 

 

1

OJ 0'2 d3 074 6.5 6.6 57
Absorbonce (OTCR)

Figure 4-6. Correlation of data obtained with open tubular and
packed bed cadmium reactors. O = seawater spikes; - = distilled
water standards.

120

estimate (0.005). Y-intercept (0.002 t, 0.002). and correlation coef-
ficient (0.9998) of this plot indicate that results obtained with both

systems are equivalent within the limits of experimental error.

L Dis u n

The goal of this research was to develop a new type of cadmium
reactor that would be more compatible with air segmented. continuous
flow analysers than the PBCRs that are most frequently used for routine
nitrate determinations in seawater samples. The OTCRs described here
fulfill this objective. Inspection of Tables 4-5 and 4-6. and Figure
4-6 reveals that equivalent analytical results can be obtained with
either reactor type. both for standards prepared in distilled water and
for seawater spikes. Furthermore. dispersion that occurred in OTCRs
where the analytical stream remained segmented during reduction was
much less than that which occurred in PBCRs where the analytical stream
was not segmented during reduction. This is shown clearly in Figures
4-4 and 4-5. Peaks with flats and less than 1% interaction were
achieved in conjunction with the OTCR at a sampling rate of 120 hr-l.
Comparable performance in conjunction with the PBCR could only be
achieved at half this sampling rate (60 hr'l).

OTCRs are much more convenient to use than PBCRs that require
elaborate procedures both for preparing cadmium filings and for packing
prepared filings into small glass columns. OTCRs are also much less
prone to clogging and channelling than PBCRs, and their use eliminates
the need to debubble and rebubble the analytical stream before and
after reduction. Thus when an OTCR is used in place of a PBCR for

nitrate determinations. fewer pump channels and manifold components are

121

required and therefore less can go wrong. The fact that the analytical
stream must be segmented with nitrogen rather than air when OTCRs are
used is somewhat inconvenient, and might be particularly cumbersome for
shipboard work where compressed gas cylinders must be secured to the
outside bulkheads for reasons of safety.

Some work still remains to characterize OTCRs completely. No ex-
periments were performed to determine the long-term stability of OTCRs.
The optimum concentration of Cu2+ in the imidazole buffer should also
be investigated more thoroughly. Addition of small amounts of copper
to this reagent kept the activity of the OTCR high and uniform, but in
larger amounts, reduction of nitrite to lower oxidation state species
increased. Nonetheless. results presented here suggest that OTCRs are
a practical alternative to PBCRs for routine nitrate determinations.
Some field studies are now needed to determine how well OTCRs will per-

form in extended. routine operation.

CHAPTER 5

AN EXPERIMENTAL COMPARISON OF CFA AND FIA

A; Introduction

From the time of its inception. FIA has been promoted as a
simpler, less costly. and more efficient alternative to CFA, although
this assertion has been hotly contested in the recent literature [68-
73]. Snyder [5] recently compared CFA and FIA and demonstrated that
from a theoretical standpoint. dispersion is generally much lower in
air segmented flow reactors than in nonsegmented flow reactors. This
study predated the first report of single head string reactors for FIA.
Very recently Rocks and Riley [11] compared FIA and CPA from the
standpoint of clinical applications and concluded that FIA would
eventually replace CFA in the clinical laboratory; This review appears
to have compared second, rather than third. generation CFA.systems with
FIA and therefore the conclusions drawn seem overly optimistic. To
date. however. there has been no direct experimental comparison of CFA
and FIA, perhaps because ground rules to prevent bias are difficult to
establish. and also because a single channel. third generation CFA
instrument is not commercially available. The miniature continuous
flow analysis system that I developed during the course of my research
can.be configured for either CFA or FIA operation. and is therefore
ideally suited to an experimental comparison of these two continuous

flow analysis techniques.
122

123

The experimental comparison of CFA and FIA presented in this
chapter is unique in several respects. lost importantly. instrumen-
tstion bias is minimal because the same pump. flow cell. and detector
were used for both FIA and CFA experiments. Furthermore. for the first
time. dispersion in open tubular reactors (OTR). single bead string
reactors (SBSR). and air segmented reactors (AS!) is compared under
identical experimental conditions. In the first set of experiments.
dye dispersion in the three reactor types was monitored as a function
of the volume of dye injected into the reactors and the liquid flow
rate through the reactors. Here chemical reaction was not required for
detection. and therefore recorded peak profiles provided an unambiguous
indication of the extent of dispersion that occurred in each reactor at
each flow rate. The situation is more complex when chemical reaction
is required for detection. as is often the case for continuous flow
determinations. Here recorded peak profiles will depend not only on
the dispersion characteristics of the reactor. but also on the kinetics
of the chemical reactions involved. Analytical rections with rela-
tively slow kinetics were expected to bias the comparison of CFA and
FIA in favor of CFA. To avoid this type of bias in the second set of
experiments. I chose the colorimetric determination of chloride by
reaction with mercuric thiocyanate and ferric nitrate [74] as a
reference assay by which the performance of CFA and FIA could be com-
pared. This reaction reaches equilibrium in a few seconds of reaction

time.

124

_B_,_ er men al

L m Dispersion Experiments

The simple flow system used for dye dispersion experiments is
shown schematically in Figure S-lA. Here a single four-way slider
valve was used to direct either the dye solution [phenol red (1.0
uncle) in 100 ml. of pi! 9.5 borate buffer] or the carrier solution (pH
9.5 borate buffer) into the reactors. Valve switching was controlled
by a microprocessor as described in Chapter 2. The volume of dye
injected as a function of pump speed and valve actuation time can be
found in Table 5-1. Pump tubes with nominal flow rates of 1.0 ml. min"1
were used for all experiments with OTRs and SBSRs. For experiments
with ASRs at the two lower flow rates. however. pump tubes with nominal
flow rates of 0.42 ml. min"1 were used. This was necessary to achieve

1 at all flow rates (see

segmentation frequencies of at least 1.5 s-
Table S-l). Plastic OTRs, SBRs. and ASRs were fabricated in the
lengths listed in Table 5-2. Fabrication details can be found in
Chapter 2. Letters in parentheses in Table 5-2 are used to identify
each reactor in the remainder of this discussion. Reactor lengths were
chosen such that the dwell time of the dye slug in each reactor type
would be about the same in each set of experiments.

Signals that resulted when 25. 50. 100. 200. 400. and 800 pl. of
dye were injected sequentially into each of the twelve reactors at four
different flow rates were recorded with a strip chart recorder. In
addition. five replicate determinations for one sample volume (usually
50 pl.) were obtained at each flow rate for each reactor. The average

and the standard deviation for each set of replicate determinations can

be found in Table 5-3. Relative standard deviations of 1% or less were

125

NOMINAL

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A) FLOW RATE
Jul-mm)-
: _ .019. 4. .1
1A"; I..._ SL952 #— —————— .1
SAMPLE I. O. 2) WASTE
“EV '-
O
BUFFER "00.42” .—I
[_J 540nm
B)
REAGENT __ 9.8_._ __ _ _ .. ._, WASTE
REAGENT -8 H
or DDW .' O.
A |.O : 480nm
[ w WASTE
SAMPLE
C) twp SPEED ~56
r‘ ""’ '" -' W n-2
I T_' a *
I (31C) J VMAErrE:
AIR L-._Q£§ 0.3m
REAGENT 0 32 1
SAMPLE ° '5
-———J 480nm
Figure 5-1. Composite diagrams of manifolds used to compare FIA and

CPA. A = air, S = sample, C = carrier, R = reagent.
B) FIA manifold for chloride determina-
C) mCFA manifold for chloride determinations.

for dye dispersion experients.
tions.

A) Manifold used

126

Table 5-1. Sample volume as a function of sample time for dye
dispersion experiments.

 

Sample Volume (uL)

 

 

 

 

 

 

25 50 100 200 400 800
Flow Rate
Pump Speed (mL min'l) Samplipg Time
21“, 45b 0.5 3.0 6.0 12.0 24.0 48.0 96.0
42“, 90b 1.0 1.5 3.0 6.0 12.0 24.0 48.0
63a 1.5 1.0 2.0 4.0 8.0 16.0 32.0
84a 2.0 .75 1.5 3.0 6.0 12.0 24.0

 

aPump tubes with nominal flow rate of 1.0 mL/min at standard speed.

Pump tubes with nominal flow rate of 0.42 mL/min at standard speed
used with ASRs only.

 

 

 

Table 5-2. Reactor lengths used for dye dispersion
experiments.
Reactor Type Lepgth (cm)
OTR 10 (A) 50 (B) 100 (C) 200 (D)
SBSR 10 (E) 25 (F) 50 (G) 100 (H)

ASR 10 (I) 30 (J) 50 (K) 75 (L)

 

127
a
Table 5-3. Precision of dye dispersion experiments.

 

------------------ Flow Rate (mL min’l) -—---——--------

 

 

 

 

 

 

Reactor 0.5 1.0 1.5 2.0
PA 0.217 1.003 .241 1.002 .258 1.007 .274 1.004
B_OTR 0.232b1.001 .280b1.000 .170 1.002 .164 1.002
c 0.153 1.001 .150 1.001 .160 1.001 .204 1.001
_0 0.124b1.001 .227c1.002 .257 1.001 .305’01.000
_ E 0.323 1.002 .334 1.002 .350 1.003 .364 1.003
1=__SBSR 0.247 1.002 .254 1.002 .273 1.003 .285 1.003
G 0.173 1.002 .183 1.002 .194 1.002 .203 1.001
_H 0.128 1.001 .138 1.001 .148 1.002 .146 1.002
' 0.436 1.004 .430 1.000 .408 1.004 .392 1.015
J_ASR 0.424 1.002 .422 1.003 .371 1.007 .364 1.009
0.440 1.002 .435 1.001 .368 1.008 .352 1.014
LL 0.433 1.002 .431 1.002 .354 1.009 .309 1.017

 

 

“Sample volume = 50 uL unless otherwise indicated.
b100 uL sample volume.

0200 uL sample volume.

128

calculated for all data sets except those obtained with ASRs at the two
higher flow rates. The higher relative standard deviations calculated
for these data (3-5fi) resulted from unfavorable experimental conditions
that could not be avoided. At the two lower flow rates the dye slug
was divided into about 10 segments (e.g.. 6 s/slug x 1.6
segments/s - 9.6 segments/slug; 3 s/slug x 3.2 segments/s - 9.6
segments/slug). while at the two higher flow rates. the dye slug was
divided into only about 5 segments (e.g.. 2 s/slug x 2.25
segments/s - 4.5 segments/slug; 1.5 s/slug x 3.00 segments/s - 4.5
segments/slug). For this reason proportioning errors were much more
pronounced at the higher flow rates. Under normal operating conditions
used for CFA. this situation would not arise. In any case. the pre-
cision of these two data sets is sufficient for qualitative compari-
sons.

Tracings of peaks recorded when 50 pl. of dye were injected into
each reactor are shown in Figure 5-2. This figure provides a great
deal of qualitative insight into the dispersion characteristics of the
three reactor types. Examination of the peak profiles obtained with
the four OTRs (A.B.C.D). reveals that peak heights decrease and peak
widths broaden as the reactor length increases. Of course. reactor
length and dwell time of the sample slug in the reactor are two sides
of the same coin. It is generally accepted [75] that dispersion in
OTRs increases with the square root of either the reactor length or the
mean dwell time of the sample slug in the reactor. This relationship
was found to hold to a fair approximation under all experimental condi-
tions. Notice also that as reactor lengths were (increased. the peaks

became more symmetrical as predicted by the 'tanks-in-series'model of

129

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FLOW RATE
0. SmL Imin. l.0 mL/m‘n. l.5 mL/min. 2 0 mL lmtn.
<13- FUXWDpentube
g 0.24 A
g a
a:
‘1 0.l- F. I
C , ‘ D
r—fi f——T_I—fi T T T fl" r T T T T l T
0 IE) 0 90 0 45 0 36
0.4- FIAISBSR
0.3- E
F
8
3 0.21 G
a N
“4° 0.:-

OH H J
I—_I—-1 7—1—4 fi—T—fi r T I T fifi—I
0 l20 0 60 0 45 0 30

(ls- rnCEA
I .I

0.44 K
t; L
ggcis-
a:
é
2_O.2«

0.l-

OJ L Msk J \L .1 LL}; L
f—‘l—fi [_T_—I r—T'fi l T I T 1—1
060l20 03060 0|5304506l2l82430

TIME“)
Figure 5-2. Curve tracings of peaks from dye dispersion experiments

with open tubular reactors, single bead string reactors, and air
segmented reactors.

130

dispersion in (Ills [76]. Another clearly visible trend is the increase
in peak heights and decrease in peak widths that occurred as the flow
rate increased. This trend is due primarily to the decreased residence
time of the dye slug in the reactor. and to a lesser extent. an
increase in secondary flow effects. These same trends hold for peaks
recorded for the four SBSRs (E.F.G.H). Note. however. that at com-
parable residence times. peak widths for SBSRs were about half those
recorded for OTRs. Also tailing. which was particularly noticeable in
peaks recorded for the shortest OT! (A). was much less pronounced in
the SBSR (E) of comparable length. especially at lower flow rates. In
general. peaks recorded for SBSRs were more symmetrical than those
recorded for OTRS. This suggests that the mixing efficiency of SBSRs
is better than in OTRs as reported by Reijn. et al. [37] who stressed
that dispersion is not synonymous with adequate mixing. especially in
OTRs. The peaks recorded for the four ASRs (I.J.K.L) exhibited very
different characteristics. Notice first that reactor length had very
little effect on either peak heights or peak widths. Note that subtle
differences in peak heights and widths that might be expected were
masked by mixing effects that occurred in the sampling valve before the
dye slug was segmented with air. Under normal operating conditions for
CFA where peeked sampling is employed. sharper peaks would have been
recorded. Nonetheless the ability of air segments to minimize disper-
sion is readily apparent. This study shows very clearly that under
comparable experimental conditions the extent of dispersion in the

three reactor types can be ranked as follows: ASR<SBSR<0TL

131

2‘ Chloride Determinatigns

In the second set of experiments. the three reactor types were
evaluated in terms of their performance in the simple colorimetric

determination of chloride. The reaction sequence is shown below.

a; (SCN)2+2c1’—> RgC12+28CN— (5-1)
SCN'+Fo3*—> Fe (SCN) 2" (5—2)
Max : 480 mm

Although it has been suggested [77] that the major product of Reaction
5-2 is Fe(SCN)2+. and that even higher order complexes might also be
formed. the 1:1 complex indicated is the mostly likely product [74].
The reagent solution was prepared by dissolving 0.626 g of mercury (II)
thiocyanate. 30.3 g of iron (III) nitrate. 3.3 ml. of concentrated
nitric acid. and 150 mL of methanol in about 500 mL of dionixed
distilled water (DDW) contained in a 1 L volumetric flask. The solu-
tion was diluted to the mark with DDW. mixed. and transferred to an
amber colored bottle. Then 1 ml. Brij-35 surfactant was added. This
reagent formulation is identical to the one recommended by Hansen and
Ruzicka [77]. The primary chloride standard (1000 ppm) was prepared by
dissolving 1.648 g of sodium chloride in l L of DDW. Working standards
in the range of 5 to 45 ppm were prepared by dilution of the primary
standard with digital microliter pipets and volumetric flasks. The FIA
and CFA manifolds used for these experiments are shown schematically in
Figures 5-lB and 5-1C. respectively. Samples were introduced into the

FIA system by means of a six-port rotary valve equipped with a 60 pl.

132

sample loop. Pecked sampling was used to introduce samples into the
CFA manifold.

'hen chloride standards in the concentration range of 5 to 35 ppm
were injected into the single line FIA manifold (Figure 5-lB) at 30 s
intervals. the peaks shown in Figure 5-3 were recorded. Note that
there is a large blank signal (i.e.. injection of DD! resulted in a
peak) and also that each positive peak was followed by a negative peak
before the signal returned to base line. These effects are frequently
observed when a colorless sample is injected into a colored reagent
stream [78] or when there are large differences between the viscositics
of the sample and the reagent stream [79]. The effects are due to
temporal changes in refractive index and concentration along the
dispersed sample slug. It is interesting that this behavior was not
observed in an almost identical experiment described by Ruzicka and
Hansen [77]. probably because the internal volume (18 pL) of the flow
cell they used was nine times greater than that (2 p1.) of the flow cell
used in these experiments.

The above problem was eliminated by adding a second pump tube to
the FIA system (dashed line in Figure 5-1B). Here samples were
injected into a distilled water carrier stream that merged with the
reagent stream in a tee mixer. Thus the extent to which the reagent
stream was diluted remained relatively constant‘when samples were
injected. Peaks obtained with the two line FIA system (50 em CTR) are
shown in Fig. 5-4A. As expected. no response was observed when
distilled water was injected into the system. but peaks and the base
line signal were excessively noisy. This was due to poor mixing

between the reagent and sample carrier streams. When the 50 cm OTR was

55 PPM

25””!

 

 

 

 

 

5P?“ " LI M

 

 

 

«I u u

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

134

{15'

. __ '35

25

I5

 

 

 

 

 

 

 

 

 

 

NM

Figure 5-4.
merging zones FIA manifold.

 

km

8) 50 cm single bead string reactor.

15

 

 

3S

 

 

K.

 

 

k

45

 

 

K

L

Peaks recorded for chloride determinations with a
A) 50 cm open tubular reactor.

135

replaced with a 50 cm SBSR. the peaks shown in Figures 5-4B were
recorded. Note that base line noise was reduced to negligible levels
and that peak heights increased and returned to baseline more rapidly.
Here we see clearly the advantages of using SBSRs for routine FIA
determinations. Dispersion is decreased so the sampling frequency can
be increased. and the mixing efficiency is much improved relative to a
comparable OTR. Further experiments showed that good results were
still obtained when the length of the SBSR was reduced to 25 cm. At a
flow rate of 1.6 ml. min—1 (carrier. 0.8 ml. min-1; reagent. 0.8 ml.
miufl). sampling rates of 180 hr"1 were achieved using the 25 cm SBSR
without significant interaction. 'hen the pump speed was doubled. it
was possible to increase the sampling rate to 360 hr’l. Peaks recorded
at this sampling rate shown in Figure 5-5A.

Next the same chloride standards were determined with the CFA man-
ifold shown schematically in Figure 5-lC. 'The pump speed and segments-

1

tion frequency were 56 and 2 s- . respectively. Peaks recorded with

the CFA system at a sampling rate of 360 hr.1 (t' . 5 s. t - 5 s) are

w
shown in Figure 5-53. Data from chloride determination experiments for
both FIA and CFA systems are compiled in Table 5-4. Note that in all
cases. less reagent and sample were required for CFA determinations.
Furthermore the linearity of the standard curves and the precision of

analytical results is somewhat better for the CFA system than the FIA

system.

136

......

    

! l
0

  

i
1
I
Figure 5-5.
rate of 360 hr‘l.
string reactor. 8) mCFA.

Peaks recorded for chloride determinations at a sampling
A) FIA with merging zones and a 25 cm single bead

137

MO OESHO> OFF ”<Hm

.ppmvcmum okuoficu Ema mm oz» mo m:0wum:flsnouov oumoflamou m scum woummsofimu

w.

.m:0wuoon:m oHaEmm coozuon Hm>uou=w oswu on» wcwusu veno>wfiov psowmoa
.Hm>uou:« cm“: can ofiasmm 6:0 wcwnsm v6u6>fifiom ucowmmn mo osmao> cab "<muo

.mmmm so mu .61 1 66666 61:66
.50 H.o u up .m on u u .Nncqom u macaw manna

 

 

al.!‘al'l' I ‘ II“! '.!5

566. 566. R66. 666. 666. 166.6 A.66666 .m66666v axe
amo. one. one. mmo. «no. mmo.o 66666606“ >
6H6. one. 156. Ame. “No. omo.o 66656
..A HA N-Hnu 6.6 6.6 6.6 1603666661H 1
56.6 66.6 Na._ 6m.o m~.o m6.o “enema *1 662606666
mmm 666 new N62 Rom Ha Aefigav 6s=Ho> 6166661
:6 66 co 6m 6m 12 56:6 6e=~o> 6~asmm
-11. .1. .111 m m m Amy 6206 :66:
.111 .11 .111 mm as m Amy 61“» 6Haeam
owm oem 66m owm cum 66m ”5-;V 6666 6261~61<
651 egos

 

.m:0mumcfishouov opwpofizo ofipuoeflhofiou

"<Hm cam <mUE mo

:omwummaou .v-m omens

138

g. Digcugsion

The foregoing experiments underscore the fundamental differences
between FIA and CFA. In FIA the sample enters the reactor as a slug
and begins to disperse into the carrier stream at the moment of injec-
tion. Note that the sample loop is itself a mixing stage. as is the
flow cell. Except in cases where samples are not required to undergo
chemical reaction prior to detection. such as when FIA is used to
transport samples to an ion selective electrode. some dispersion of the
sample slug is necessary. In single line FIA systems. in fact. disper-
sion is the only mechanism by which the sample and reagent can
interact. In dual line FIA systems where the sample and reagent are
merged in a tee connector. mixing is improved somewhat but is still far
from complete. For colorimetric determinations. this is a major weak-
ness in the FIA approach to continuous flow analysis. because it is
generally necessary to compromise between adequate mixing of the sample
and reagent on the one hand. and excessive broadening of the sample
slug on the other. There are a number of 'little tricks'. however.
that can be used to enhance mixing and minimise dispersion in FIA that
help to make this compromise more favorable. As shown in the experi-
mental section. these include inducing secondary flow by using tightly
coiled rather than straight OTRs. and using SBSRs rather than OTRs.
Both these approaches are successful because they disrupt the parabolic
velocity profile associated with laminar flow. Another effective way
to increase residence time without excessive dispersion is simply to
inject the sample into the carrier stream and then stop the pump [80].
To a close approximation when the flow stops. dispersion stops. After

sufficient reaction time has elapsed. the pump is restarted and the

139

reacted sample is propelled into the detector. Of course this approach
is somewhat cumbersome and has some rather obvious limitations.

The situation is quite different for CFA.because the process of
mixing the sample with reagent is not dependent on sample dispersion.
Recall that in CFA the sample slug is divided into a number of nominal-
ly identical segments before it enters the reactor. Then each segment
is proportioned with reagent. and mixing within each segment is
enhanced by the mierocirculation pattern (bolus flow) that develOps
naturally in a segmented. flowing stream. Furthermore. dispersion in
CFA is minimized by one 'big trick'. the bubble. Air segments provide
barriers to the establishment of’parabolic velocity profiles in.all
zones of a CFA system: pecked sampling greatly reduces dispersion that
would otherwise arise during the sampling process; segmented flow
greatly reduces dispersion in the reactor; and bubble-through flow
cells reduce dispersion associated with detection. In addition. other
factors being equal. dispersion decreases as the flow rate decreases
which leads to conservation of samples and reagents.

From the foregoing experiments. it is clear that for equilibrium
based to colorimetric determinations. CFA retains a slight edge over
FIA even for relatively simple chemistries with fast reaction rates.
It cannot be denied. however. that in the limit FIA is more simple.
operationally. than CFA. Consider the case where a sample is reacted
with a single reagent and detected. The minimum requirements for an
FIA system are a single channel pump. an injection valve. and a
recording detector. The same determination.with CFA would require
three pump channels (one each for sample. air. and reagent). and the

detector signal would have to be gated to eliminate the air bubble

140

artifact. Otherwise a fourth pump channel would be required to debub-
ble the analytical stream prior to detection. Also. when an automatic
sampler is not available. as was the case for my work. introduction of
samples with a valve in FIA is much less taxing than manual. peeked
sampling used for CFA. because no timing is required. Thus for single
reagent colorimetric determinations where the differences in per-
formance between CFA and FIA are small. FIA may be preferred due to the
relative ease and simplicity with which determinations can be perfor-
med. On the other hand. for colorimetric determinations requiring
multiple reagent additions. dialysis. or reaction times exceeding about
30 s. the increased complexity of CFA is more than offset by its higher
performance. The point at which this trade-off is reached. however. is
not clear-cut. and the choice between these two continuous flow tech-
niques will be strongly influenced by the analyst's ingenuity and

predisposition to CFA and FIA.

CHAPTER 6
FUTURE APPLICATIWS

_A_,_ Overview

The miniature continuous flow analysis system described and
characterized in the preceding chapters has proved to be a versatile
instrument. In conjunction with the electronic bubble gate. state-of-
the-art performance for air-segmented continuous flow analysis can be
achieved with relative ease at low cost. Flow injection analysis can
be performed with the same basic system except that a multiport valve
is used for sample introduction. and the bubble gate is removed from
the detection circuitry. The modular design of this system. and the
ease with which it can be configured for either CFA or FIA are unique
features that offer'a number of intriguing possibilities for future
analytical applications. A few of these are outlined in the paragraphs

that follow.

g, .Automated,§;gplg Pretrgatment ggg £52; gag Post-Column Reactors £2;
ngpig_Chroma10graphy

A survey of the recent literature concerning detection schemes for

liquid chromatography (LC) reveals a progressive shift in interest from

universal to selective detectors. This trend reflects the growing

awareness that because of the complex matrices associated with samples

of biological or environmental origin. a truly universal LC detector

141

142

would create more problems than it would solve. Of the many
conceivable selective LC detectors. those based on UV‘VIS absorption.
fluorescence. or electroactivity are most widely used. Often both the
selectivity and the sensitivity achieved with these detectors can be
increased by several orders of magnitude if sample constituents of
interest are derivatized with appropriate reagents prior to detection.
either before or after chromatographic separation. Generally sample
pretreatment and derivatixation steps are labor intensive. and it is
therefore not surprising that continuous flow analysers. which allow
these procedures to be automated. are beginning to gain acceptance as
efficient and reliable adjuncts to LC systems. 'To date continuous flow
post-column reactors have received the most attention. This subject
has been extensively reviewed by Frei and co-workers [81-85].

There has been a great deal of discussion about the optimum.design
for LC post-column reactors. Obviously. the extra-column band
broadening that occurs in post-column reactors must be held to a mini-
mum to prevent loss of chromatographic resolution. There are three
basic types of post-column reactors. In two of these. coiled open
tubular reactors and packed bed reactors. the column effluent is not
segmented. In the third type. segmented open tubular reactors. the
column effluent is segmented either with gas or an immiscible liquid.
Very recently Scholten. 23. 1.1- [86] presented results of experiments
with all three post-column reactor types and concluded that segmented
reactors were to be preferred even when the kinetics of the post-column
reaction were relatively fast. This is not surprising in light of the
experimental comparison of FIA and CFA presented in Chapter 5.

Scholten used second generation CFA equipment for this study and found.

143

predictably. that most of the extra-column band broadening that
occurred could be attributed to mixing in debubblers and phase separa-
tors. As Snyder [31] pointed out. extra-column band broadening can be
reduced still further with third generation CF systems that are
equipped with gated detectors which eliminate the need for phase
separation. At present. single channel third generation CF systems are
not available commercially. The single channel mCF system is ideally
suited for post column reactor applications and is sufficiently simple
to be fabricated in-house at low cost. In addition the mCF system
should be well suited for sample preparation (dialysis. solvent extrac-
tion. derivatisation) prior to the chromatographic step. In this case
of course. the effluent from the mCF system would be interfaced to the
sample injection valve of the liquid chromatograph rather than a detec-
tor; For either application. the mCF system should produce better

results than any presently available commercial systems.

9. Continuopg Blow ginetigg Qeterminatiggg

There are presently two approaches to kinetics determinations with
continuous flow analysers. In the case of third generation CFA systems
[4]. the analytical stream flows sequentially through several gated
detectors which are interconnected by means of thermostated delay
coils. Because the integrity of the analytical stream is maintained
during detection. dispersion between detectors is minimal. Thus the
steady state signal recorded at each detector provides a measure of the
extent of the chemical reaction at fixed time intervals. ‘This approach
is somewhat cumbersome unless a computer is available for data acquisi-

tion and processing. In the case of FIA systems the so-called

144

'stopped-flow' [87] technique (not to be confused with conventional
rapid-mixing stopped-flow analyzers) is generally used. Here some
portion of the dispersed sample some is stopped within the flow cell.
after sufficient time has elapsed for mixing. by turning off the pump.
After the flow stops the progress of a chemical reaction can be
monitored continuously over any desired time interval.

The FIA approach is attractive because of its simplicity. Only
one detector is required and data can be acquired with a strip chart
recorder. For determinations that require preliminary reactions prior
to the final analytical reaction. or that require sample pretreatment
such as dialysis. CFA may still be preferred despite its complexity.
Neither technique is applicable for kinetics determinations where the
reaction half-life is on the order of the mixing time (:5-10 s) of the
continuous flow analyzer. This constraint is not prohibitive for a
number of kinetics based determinations of clinical interest. however.
and it would be worthwhile to compare the performance of continuous
flow analyzers with that of conventional stOpped-flow analyzers for

these applications.

Q; Active Qpen Tubulgr Reactgrg

Robert Thompson. a recent graduate of the Crouch research group.
used the mCF system to do some nice work with open tubular immobilized
enzyme reactors [44.88]. There are several opportunities for continued
research in this area. These include characterization of reactors to
which more than one enzyme has been bonded. the use of several immobi-
lized enzymes reactors for a single determination. and characterization

of immobilized enzyme single bead string reactors. Readers with an

145

interest in this line of research are directed to Thompson's Ph.D.
dissertation [44]. Nonenzymatic active tubular reactors also warrant
further investigation. As demonstrated in Chapter 4. open tubular
cadmium reactors can be used for nitrate determinations and they are
more compatible with CFA then packed bed cadmium reactors. Other
metals such as iron. copper. and zinc in the form of narrow open tubes
may also have useful applications for routine determinations with CFA

and FIA.

g, Extended Data Acguisitign 5p; Procesging Systems

A dedicated microcomputer based data acquisition and processing
system would greatly extend the usefulness. of the mCF system. This.
along with the automatic sampler recently acquired by the Crouch group.
should go a long way toward eliminating the drudgery and tedium that
characterized the manual sampling and data processing required for many
of the experiments performed during the course of my research. The
general purpose FORTH based data acquisition routines developed by
Eugene Ratzlaff [41] are already available to members of the Crouch
research group. These routines could be stored in read only memory and
should be extended to allow for curve regeneration. peak finding.
baseline drift corrections and sample interaction corrections.
Software routines could also be written to control the pump and
sampler. These routines would be particularly useful for FIA experi-
ments. Although a software bubble gating routine could also be
written. my own feeling is that analog bubble gating is just as
effective and it has the advantage of freeing the microcomputer to

perform higher level tasks.

LIST OF REFERENCES

10.

11.

12.
13.
14.
15.

16.

17.

18.

LIST OF REFERENCES

L.T. Skeggs. Apex, J, Clin, Path, 28. 311-322 (1957).

LR. Stewart. G.R. Beecher and P.B. Rare. Fed, Prdt, 13,. 1439-
(1974).

J. Ruzicka and ER. Hansen. Anal, Chim, A21; 1d. 145-157 (1975).

L.R. Snyder. J. Levine. R. Stoy and A. Conetta. a Ch 5;, dd.
942A-956A (1976).

L.R. Snyder. dnat, Chtm, Adta 111. 3-18 (1980).

LR. Furman. Continuous Flow Analytit: Theory ddd Pradtige.
Nareel Dekker. New York (1976).

D. Betteridge. Anal Chem L0. 832A-846A (1978).

 

J. Ruzicka and E.R. Hansen. Anal, Chim Agta L. 19-44 (1980).
C. Ranger. Anal, Cheat fl. 20A-32A (1981).

J. Ruzicka and E. Hansen. Flow Ldjeetion Analyttt. John Wiley and
Sons. New York (1981).

B. Rocks and C. Riley. Clin, Chgm, (Uigston-Salem, NC) 3d. 409-
421 (1982).

(.1. Stewart. Talantd 21. 789-797 (1981).

R. Nottola. Anal, Chgm, fl. 1312A-1316A (1981).

A.T. James and A.T. lartin. Analyst 11. 915-932 (1952).

7.1. Blaedel and G.P. Ricks. Antt, Che_m,,_ .31. 388-394 (1962).

D.R. Spackman. LR. Stein and S. Nore. Antl, Chggg 19. 1190-1206
(1958).

I.R.C. lalker. in Continuout Flow Adalysis: Theoty £13 Pragtigd
(LB. Furman. ed.) Narcel Dekker. New York (1976) PP. 207-255.

'.R.C. Walker. C.A. Pennock and OJ. NcGowan. Clin Chim, Agtd
2.1. 421-435 (1970).

146

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.
36.

37.

38.

147

R.B. Thiers. R.R. Cole and '.J. Kirsch. i C im I n-
Satem, NC) 11. 451-467 (1967).

L.R. Snyder and R.J. Adler. 9111, Chgm, dd. 1022-1027 (1976).

J. erina. 6th Colloq. Amino Acid Analysis. Technicon Corp”
Nonograph No. 3. 1967.

R.D. Begg. Anal, Chtm, 11. 631-632 (1972).

LE. Thiers. A.R. Reed and I. Delander. Cttn, Chem, (Vinttdn-
Stlem, NC) 11. 42-48 (1971).

I.R.C. 'alker and LR. Andrew. £119, Chim, Act; 11. 181-185
(1974).

L.R. Snyder and R.J. Adler. dds), Cdgg, L8. 1017-1022 (1976).

R. Diebler. N. Pelavin. in ddymu g Automatgd Analystt,
Technicon Sygposta 1912, . Nediad Inc.. Tarrytown. NY. Vol. I
(1972) pp. 7-13.

R.L. Rabig. 3.7. Schlein. L. Ialters and R.R. Thiers. Cl tn, Chem,
11. 1045-1055 (1969).

LR. Neeley. S. 'ardlaw and I.R.T. Swinnen. C n Ch m I ns n-
Satem, NC) 10. 78-80 (1974).

C.J. Patton. N. Rabb and SJ. Crouch. A1151: Chgm, fl. 1113-1118
(1982).

'.R.C. Walker. Clin, Chim, A213 11. 305-306 (1971).
L.R. Snyder. ,1, Chrogatogr, 1.3.5.: 287-306 (1976).

L.R. Snyder. in m .11.! Autdmdted Adalytit: Teghnigon Inter-
national Congrgtt 121d. Nediad Inc.. Tarrytown. NY. Vol. I.
(1977) pp. 76-81.

J. Ruzicka and R.R. Ransen. Ploy Idjegtion Anatys 1. John Riley
and Sons. New York. (1981) p. 16.

J.T. Vanderslice. LI. Stewart. A.O. Rosenfeld and DJ. Riggs.
Ta1anta 2;. 11-18 (1981).

R. Tijssen. gal, Chim, Act; 111. 71-89 (1980).
R. Rofman and I. Ralasz. 1, Chrdmatggt, 113,. 211-228 (1979).

J.N. Reijn. LR. Van Der Linden and R. Poppe. Ana), Chtg, dgtg
123.. 229-237 (1981).

J.N. Reijn. '.E. Van Der Linden and R. Poppe. Anal, Chtm, Adtd

39.

40.
41.
42.
43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.
56.
57.
58.

59.

148

O. Levenspiel. Chemigal Readtidn Engineering. 2nd ed.. John Riley
and Sons. New York (1972).

I. Rofman and I. Ralasz. mgatogr, 122. 3-22 (1980).
E. Ratzlaff. Ph.D. Dissertation. Nichigan State University. 1982.
A. Savitzky and N.J’.E. Golay. Anal, Cteg, 16. 1627-1639 (1964).

E. Amador. Clin, Chem, (Vintton-Salem, NC) L8. 164 (1972).

R.Q. Thompson. PR.D. Dissertation. Nichigan State University.
1982. pp. 154-155.

'.R.C. Walker. J. Townsend and P.N. Keane. Clin, Chim, Agta 1d.
119-125 (1972).

E.B.N. DeJong and A.R. Ieyden. Anal, Chtm, Atta 11 . 311-317
(1980). '

ILA. Nacaulay. LN. Nathers. C.L. Jacklyn and C.A. Nunro.
Ctin, Biochem, 2. 111-116 (1976).

LE. Neeley. S.C. 'ardlaw. T. Yates. v.0. Roll ingsworth and LG.
Swinnen. Clin Chem 2;. 227-231 (1976).

 

T.E. 'hitledge. S.C. Nalloy. C.J. Patton and C.D. Virick.

Autogated Nutrient Analyses d_n Seayatgr, Technical Rgport.
Brookhaven National Laboratory. Upton. NY (1981) pp. 50-53.

 

R.E. Thiers. J. Ileyn and R.F. 'ilderman. Clin, Che;,_1§. 832-839
(1970).

E.A.G. Zagatto. A.O. Jacintho. J. lortatti and F. Bergamin.
Anal, Chim, Agta 120. 399-403 (1980).

VJ. Truedale and C.J. Smith. Analyst 191d ,10_1. 19-31 (1976).

F.J.A. Armstrong. C.R. Stearns and J.D.R. Strickland. Deep-
Sea Res, d1. 381-389 (1967).

T.J. Chow. and 11.8. Johnstone. Anal, Chim, Acta _21. 441-446
(1962).

LR. lull in and J.P. Riley. Ant), Chim, Agtt 11. 464-480 (1955).

D.R. Senn and PJ. Carr. Anal, Chegy 51. 954-958 (1976).

 

N.P. Stainton. Anal Chcgfl. 1616 (1974).

LR. Iillis. Anal Chem 51. 1377-1379 (1980).

 

I. Davison and C. 'oof. M 191. 403-406 (1978).

60.

61.

62.

63.

64.

65.
66.

67.

68.

69.

70.
71.
72.
73.

74.

75.

76.

77.

78.

79.

80.

81.

149

J.D.R. Strickland and T.R. Parsons. A 3.3.1.1231 gt Sayater
galysis, _2_n_d 91» Queens Printer. Ottawa (1965). p. 203.

 

F. Nydahl. Talanta 21. 349-357 (1976).

E.D. Wood. F.A.J. Armstrong and F.A. Richards. L_!ar, Biol,
All: 0": fl. 23-31 (1967)e

A. Otsuki. gal, Chim, Agta a. 375-377 (1978).

C.A. Sherwood and D.C. Johnson. Ana Chim, Acta 129. 87-100
(1981).

P.G. Brewer and J.P. Riley. DeetSea Researgh _1_2_. 765-772 (1965).
F. Nydahl. Talantt 2_1. 1259-1267 (1974).

J.P. Riley and G. Skirrow. Chemital Oceanograp y. Academic Press.
New York. (1965) pp. 204-205.

If. Nargoshes. Anal Chem 52, 17-19 (1977).

 

J. Ruzicka. E.R. Ransen. R. Nasbaek and F.J. Krug. Anal, Cheg,
fl. 1858-1861 (1977).

N. Nargoshes. Ana), Chem, 12. 1861-1862 (1977).
I. Nargoshes. Anal, Chgm, §_4_. 678A-679A (1982).

C.B. Granger. Anal Che,m_,_ L4,. 1106A (1982).

 

N. Nargoshes. Anal. Chem 25.: 1106A (1982).

 

D.R. Zall. D. Fisher and N.Q. Garner. Anal Cheat 2 . 1665-1668
(1956).

 

J. Ruzicka and R. Ransen. Flow Injection Analysis. John Wiley and
Sons. New York. (1981) p. 20.

J. Ruzicka. and E. Ransen. Floy Injedtion Analysis. John Wiley
and Sons. New York. (1981) p. 41.

E.R. Hansen and J. Ruzicka. Chem, Ed, dd. 677-680 (1979).

J. Ruzicka and F. Ransen. Blow Injection Adalysit. John Wiley and
Sons. New York. (1981) p. 144.

D. Betteridge. Talantd _2_3_. 409-410 (1976).

J. Ruzicka and E. Ransen. Flow Injgction Analyt t. John Wiley and
Sons. New York. (1981) p. 21.

R.W. Frei and W. Santi. 2, Anal, Chg!_,_211. 303-310 (1975).

82.

83.

84.
85.

86.

87.

88.

150

R.W. Frei. L. lichel. and W. Santi. J, Chromatdgr, 142. 261-270
(1977). '

R.W. Frei and A.R.N.T. Scholten. d, Chrogatogr,.S§i, A1. 152-160
(1979).

R.W. Frei. J, Chromatdgr, 1 5. 75-86 (1979).
R.W. Frei. Chromatographia 11. 161-166 (1982).

A.R.N.T. Scholten. U.A.Th. Brinkman. and R.W. Frei. Anal C e,a_1,_
14.. 1932-1938 (1982).

J. Ruzicka and E.R. Ransen. loy Injection Analysit. John Wiley
and Sons. New York. (1981) pp. 61-65.

R.Q. Thompson. C.J. Patton and S.R. Crouch. Clin Chem Wins on-
Salem, 1491 2.1;. 556-557 (1982).