Sequential injection analysis for automation of the Winkler methodology,

cavalcadehorehoundMechanics

Nov 5, 2013 (3 years and 9 months ago)

73 views

1

Sequential injection analysis for automation of the Winkler methodology,
with real
-
time SIMPLEX optimiza
tion and shipboard application

Burkhard Horstkotte
1
,
*
,
Antonio Tovar
1
,
Carlos M.
Duarte
1
, Víctor
Cerdà
2

1

Department of Global Change Research.
IMEDEA (
CSIC
-
UIB) Institut Mediterráni d'Estudis
Avançats, Miquel Marques 21, 07190 Esporles, Spain

2

University of the Balearic Islands, Department of Chemistry

Carreterra de Valldemossa km 7,5, 07011 Palma de Mallorca, Spain


Abstract:

A multipurpose analyzer sy
stem based on sequential injection analysis
(SIA)
for the determination of
dissolved oxygen (DO) in seawater is presented. Three operation modes were established and
successfully applied onboard during a research cruise in the
Southern
ocean: 1
st
, in
-
line
execution of
the entire Winkler method including precipitation of manganese(II) hydroxide, fixation of DO,
precipitate dissolution by confluent acidification, and spectrophotometric quantification of the
generated iodine/tri
-
iodide (I
2
/I
3
-
), 2
nd
, spectroph
otometric quantification of I
2
/I
3
-

in samples prepared
according the classical Winkler protocol, and 3
rd
, accurate batch
-
wise titration of I
2
/I
3
-

with thiosulfate
using
one
syringe pump of the analyzer as automatic burette.

In the first mode, the zone sta
cking principle was applied to achieve high dispersion of the
reagent
solutions
in the sample zone. Spectrophotometric detection was done at the isobestic wavelength
466

nm of I
2
/I
3
-
.
Highly reduced c
onsumption of reagents and sample compared to the classi
cal
Winkler protocol
, l
inear response up to 16

mg

L
-
1

DO
,

and an injection frequency of 30

per hour were
achieved.
It is noteworthy that for the offline protocol, sample metering and quantification with a
potentiometric titrator lasts in general over 5

min

without counting sample fixation
,
incubation
, and
glassware cleaning
.
The modified SIMPLEX methodology was used for the simultaneous optimization
of four volumetric and two chemical variables. Vertex calculation and consequent application including
in
-
lin
e preparation of one reagent was carried out in real
-
time using the software AutoAnalysis.
The
2

analytical
s
ystem featured high

signal stability, robustness, and a repeatability of 3

% RSD
(1
st

mode)
and 0.8

% (2
nd

mode)
during shipboard application.


Keyw
ords:


Sequential injection analysis, Dissolved Oxygen, Winkler protocol, Seawater Monitoring, modified
SIMPLEX method, AutoAnalysis

Fax: + 34 971 173 426, email:
Burkhard.Horstkotte
@uib.es

3

1 Introduction

The concentration of dissolved oxygen (DO) in seaw
ater is a
n

essential property used to characterize
water masses and derive ocean circulation
[
1]

and ecosystem metabolism
[
2]
, and assess the risk of
hypoxia of the various water masses

[
3]
.

To this aim, the Winkler method
[
4]

is routinely used in oceanog
rap
hic research, specially where
high
analytical precision and independence from ambient pressure, temperature, or ion strength of the
sample are required
[
5
-
7]
.

Although Clark
-
type electrodes and fluorescence
-
based optochemical oxygen sensors (oxygen
opt
r
odes) are available to measure oxygen concentration in seawater, their performance is
only
comparable to that of high
-
precision Winkler measurements
.
By example, t
he
both approaches are
generally
affected by membrane
/ sensor
bio
-
fouling
,

the requirement
of a constant approach flow to
the membrane,
temperature and pressure dependency,
and
considerable
delay time
s

to reach steady
state signal
.

N
on
-
linearity of the calibration function, costs,
photo
-
bleaching of the sensor layer,
and
limited lifetime are typ
ical shortcuts of oxygen optrodes
[
8,9]

but can be overcome
by adequate
protective layers on the sensor
[
10]
.

Use of cyclic voltametry for DO quantification renders high
robustness but lengthens also the measurement procedure

[
11].

Both sensor types are ad
aptable to
in
-
situ measurements and DO depth profiling.

However, amperiometric sensors are inadequate for incubation experiments due to the consumption
of dissolved oxygen by the sensor itself
. On the other hand,

invasive
optodes
have to be used during
th
e entire incubation
experiment

although only the integral consumed DO is of interest mostly. In
consequence the
number of proces
sed
samples
is limited by the
number of
purchased optodes
.

The Winkler method is based on the quantitative oxidation of manganes
e (II) hydroxide precipitate by
DO in alkaline media followed by the reduction in presence of iodide in acidic media. The coupled
reduction of iodide to iodine can be determined by titration with thiosulfate either using starch as visual
end
-
point indicato
r or assisted by potentiometry. Alternatively, iodine can be quantified directly by
spectrophotometry with improvements reported in respect of manual labor, time of analysis, and
reliability
[
1
2,13
]
.

With a surplus of iodide, iodine (I
2
) forms the tri
-
iodi
de ion (I
3
-
)
[
14
]
, which has led to
the consideration to perform spectrophotometric quantification of I
2
/I
3
-

at their isosbestic wavelength
[
15
]

for an improved reliability. Main disadvantages of the Winkler method are lengthy times for
4

reaction and determ
ination, considerable reagent consumption, and a large and precisely metered
sample volume.

Analytical flow
techniques present an elegant way for the application of wet
-
chemistry procedures for
monitoring, i.e. semi
-
continuous, automated proceeding by sim
ultaneous reduction of reagent
consumption and sample handling. Therefore, t
hey are widely used where fast, economic, reliable,
and automated analysis of a large number of similar samples is demanded.


In spite of these potential advantages, mainly the qua
ntification of I
2
/I
3
-

originated from the classical
Winkler method has been automated applying by example spectrophotometry
[
16
]
, amperometry
[
17
]
,
or indirect spectrofluorimetry
[
18
]
. Only in two works, fully
[
1
9
]

and semi
-
automated
[
20
]

implementation of

the entire Winkler chemistry in complex, multichannel flow injection analyzers (FIA)
[
21
]

has been described. In both analyzers, peristaltic pumps were used as liquid drivers implying the
well
-
known shortcomings of flow rate drift, high solution throughpu
t, and laborious optimization by
manual manipulation of the manifold.

Among the distinct flow techniques extensively described in recent reviews
[
22
-
25
]
,
Sequential
Injection Analysis (SIA) proposed by Ruzicka and Marshall in 1990
[
2
6
]

ranks as one of the

most
simple, economic, and versatile flow techniques, given by the multitude of analytical applications
[
27
-
28
]
. It uses only one single, pressure
-
robust syringe pump
module
for the precise handling of minute
volumes of the required solutions, aspirated f
rom the lateral ports of a selection valve and introduced
into a flow detection cell. Any required procedures including mixing of sample and reagents and
incubation are carried out following a software
-
based instruction protocol, which can easily be adapte
d
using distinct
operation schemes for each analytical parameter.

In this work, we
describe a SIA automation of the entire Winkler reaction (precipitation of manganese
(II) hydroxide, reaction with dissolved oxygen in the sample, precipitate dissolution,
and generation
and spectrophotometric detection of I
2
/I
3
-
). Determination of I
2
/I
3
-

in samples prepared according the
classical Winkler method was further enabled by either batch titration with thiosulfate or by
spectrophotometry obtaining an efficient, mu
ltipurpose analyzer. Due to the linear operation of SIA,
zone stacking of reagents and sample is imperative, which is in contrast to FIA, where merging flows
of sample and reagents is a usual concept. To enable acid addition after the

oxidation of the
prec
ipitate, the SIA was extended by an additional syringe using a multisyringe device

as liquid
5

propelling device
described in detail elsewhere
[
2
9
-
31
]
. To achieve the maximal reaction volume of
sample with the precipitate formed in
-
line by the mixture of the

used reagents manganese sulfate and
alkaline iodide solution, thorough optimization of especially the volumetric variables affecting the
precipitate formation and sample penetration was required.

Computer assisted optimization such as by the modified SIMP
LEX method
[
3
2
,3
3
]

provides a highly
useful tool to find an optimum of the variables effecting the performance of an analytical instrument.
This potential has been demonstrated for flow technique
-
based analyzers before
[
34
-
36
]
, however,
considerable user's

work has still been required for SIMPLEX optimization such as preparation of
reagents and manual modification of the tubing manifold. In contrast, Gine et al. presented
for the first
time
in 1998
[
37
]

real
-
time SIMPLEX
optimization of
a multicommutated FI
A
system
[
3
8
,3
9
]
. Here, the
calculated variables of each vertex were directly applied to the operation scheme of the solenoid
valves implemented in the analyzer manifold.

There is nevertheless a paucity of real
-
time application of SIMPLEX to flow techniqu
e systems, one
reason being the availability of versatile control programs able to control instruments typically used in
flow techniques such as pumps, valves, and detectors, of on
-
line data evaluation, and perform the
required calculations. A second bottl
eneck is that the variables of interest need to show
interdependence. Finally, the information deficit about the robustness of the optimum and the
presence of other optima has to be acknowledged as a main drawback of SIMPLEX optimization,
which requires a
posterior univariant study of each variable for the validation of the found optimum.

In this work, we applied the modified SIMPLEX methodology for the real
-
time optimization of four
volumetric variables and two concentrations of one in
-
line prepared reage
nt of a SIA system using the
software AutoAnalysis. The optimized and characterized analyzer was used for the determination of
DO in seawater samples during a research cruise in the Southern Ocean


2 Materials and methods

2.1 Reagents and standards

All che
micals were of analytical
-
reagent grade (Scharlab, Barcelona, Spain) and
for the preparation of
the overall solutions
distilled water was used
for all laboratory experiments and MilliQ water during
6

shipboard application
. A solution of 0.3

mol

L
-
1

MnSO
4

was

used as a reagent 1 (R1). A solution of
0.6

mol

L
-
1

NaI and 0.2

mol

L
-
1

NaOH was used as a reagent 2 (R2). Reagent 3 (R3) being 0.3

mol

L
-
1

H
2
SO
4

was propelled by syringe 2 and a carrier being 0.05

mol

L
-
1

H
2
SO
4

was propelled by syringe 1.
The reagents we
re not sparged with nitrogen prior to use for practical reasons.

For the classical Winkler protocol the following reagents were prepared and used according the
standard protocol

[6]
: 3

mol

L
-
1

MnSO
4

(R1c), 4

mol

L
-
1

NaI and 8

mol

L
-
1

NaOH (R2c), and 5

mol

L
-
1

H
2
SO
4

(R3c).
KIO
3

was dried at 105°C overnight and used to prepare solutions of 0.023

mol

L
-
1

and
0.0023

mol

L
-
1
. A 0.2

mol

L
-
1

Na
2
S
2
O
3

solution was prepared as a titrant of I
2
/I
3
-

and standardized
against 10

mL of 0.0023

mol

L
-
1

KIO
3
, previously mixed

with 1

mL of R3c and R2c. A 30

w/v

% starch
solution in glycerol was used as visible I
2
/I
3
-

end
-
point indicator for titration. For modified SIMPLEX
optimization, stock solutions of NaI and NaOH of 2

mol

L
-
1

each were used.

Artificial seawater (ASW
) as de
tailed elsewhere
[5]
by dissolving the following compounds in
distilled water in the given order 3

mg

L
-
1

NaF, 20

mg

L
-
1

SrCl
2
∙6 H
2
O, 30

mg

L
-
1

H
3
BO
3
, 100

mg

L
-
1

KBr, 700

mg

L
-
1

KCl, 1,470

mg

L
-
1

CaC
l
2
∙2H
2
O, 4,000

mg

L
-
1

Na
2
SO
4
, 10,780

mg

L
-
1

MgCl
2
∙6 H
2
O,
23,500

mg

L
-
1

NaCl, 20

mg

L
-
1

Na
2
SiO
3
∙9 H
2
O, and 200

mg

L
-
1

NaHCO
3
.

For calibration of the in
-
line Winkler method, about 280

mL
ASW

solutions of different DO contents
were measured on the proposed analyzer. Standards were prepared by sparging of
ASW

with
n
itrogen, air, or oxygen for at least 15

min, which were found to yield steady
-
state gas saturation.
Intermediate contents of DO were obtained by mixing of saturated
ASW

in a Winkler bottle, which was
closed and shaken vigorously
hereafter
.
The objective wa
s not to prepare DO standard of known but
different concentrations to be quantified by the classical method.
Oxygen was

generated in the
laboratory by the catalytic decomposition of
H
2
O
2

(about 15

%v/v) continuously added to an initial
volume of
KMnO
4
. Aft
er in
-
line analysis, the unknown contents of DO of each standard were
determined by application of the classical Winkler titration protocol
[
6]
.

For this, Winkler bottles of
approximately 120

mL were filled carefully with the standards and 0.5

mL of R1c an
d R2c were added.
The bottles were sealed, shaken vigorously, and incubated for at least 1

h in the darkness. Afterwards,
0.5

mL of R3c was added and the formed I
2
/I
3
-

was quantified by titration.

For the spectrophotometric determination of I
2
/I
3
-
, calibr
ation standards of 100
-
800

µmo
l

L
-
1

DO were
prepared by appropriate addition of 0.023 mol

L
-
1

KIO
3

to about 280

mL
ASW
, where 1

µmol of KIO
3

7

corresponds to 1.5 µmol DO. Afterwards, 1

mL of R2c and 1

mL of R3c were added, by which iodate
is reduced quantita
tively to I
2
/I
3
-

[
6,16
]
.


2.2 Flow analyzer instrumentation

The implemented SIA system is depicted in Fig. 1
-
A wherein tubing dimensions are indicated. A valve
module VA2 equipped with two rotary 8
-
port selection valves and a syringe module Bu4S
[
2
9
-
31
]

pu
rchased from Crison Instruments S.A. (Allela, Barcelona, Spain) were used. The syringe module
was equipped with two glass syringes of 5

mL (S1) and 1

mL (S2) total dispense volume (Hamilton
Bonaduz AG, Bonaduz, Switzerland), driven simultaneously by the si
ngle step motor of the instrument
(16000 steps, 24 s


1024 s for

total dispense). Solenoid head
-
valves allow

the connection of the
syringes either to the manifold (ON) or to the respective carrier reservoir (OFF) for re
-
filling. Both
modules were connecte
d in series via a RS232C serial interface to a PC for remote software control.

All liquid contacted parts were made of the chemical resistant polymers ETFE, PMMA, PEEK, and
PTFE. The central port of the selection valve was connected via the holding coil HC
1 to the position
ON of the syringe head valve. About 6

cm from the selection valve, a 3
-
way connector was integrated
into HC1 used as a confluence of R3 (in
-
line addition of acid).

A thermostatization coil (TC) was used to connect the selection valve and

the detection flow cuvette. It
was inserted into a homemade reactor flushed continuously with water provided from a miniature
precision thermostat PT31 (Krüss Optoelectronic GmbH, Hamburg, Germany). The PT31 was further
used for the thermostatization of t
he flow cuvette holder using a processor water
-
cooler copper shoe
(Conrad, Hamburg, Germany).

An USB
-
2000 miniature spectrophotometer was used for detection and was directly connected to a
cuvette support CUV
-
UV (both Ocean Optics Inc., Dunedin, USA) with
a flow cuvette, type 75.15 SOG
from Starna (Essex, UK). Dual wavelength detection was accomplished throughout for compensation
of the schlieren effect
[
40
]
using the isosbestic wavelength 466

nm of I
2
/I
3
-

[41]
as detection
wavelength and processed by subtr
acting the absorbance at a reference wavelength of 580

nm. A
homemade light source was used for all measurements consisting of a halogen bulb and a 383UBC
LED (390

nm
-

510

nm, range < 20

% emission intensity) from Roithner Lasertechnik, Vienna, Austria.
8

B
oth were arranged perpendicular such that the light emitted of the LED passed the bulb and both
spectra were superpos
ed achieving in approximation
uniform

emission spectra.


2.3 Software AutoAnalysis and Integration of SIMPLEX methodology

The software pac
kage AutoAnalysis 5.0 (http://www.sciware
-
sl.com)
[
42
]

was used for instrumental
control, data acquisition, and evaluation. The basic program is written in Delphi and provides tools for
in
-
line data evaluation, use of variables, basic calculations, loops,
conditional inquiries, and enable a
modular setup of the instruction method by definition of pr
ocedures. Instrumental control wa
s done via
specific dynamic link libraries establishing the communication to the individual instrumental assembly.

The modified

SIMPLEX method programmed on the AutoAnalysis 5.0 platform enabled the
optimization of up to seven variables, six used in the present work. It enabled the calculation of the
initial SIMPLEX from a given center point and SIMPLEX diameter. Three stop criter
ia were
established: maximal optimization cycles, maximal repetition of the best vertex, and minimum
threshold of the achieved improvement. It automatically calculated a new vertex either by normal
reflection, expansion, external contraction, or internal c
ontraction including a correction if one or more
parameters of the new vertex passed prior defined working range limitations. It finally enabled the
permutation of the second worst vertex if permutation of the worst vertex led to a
n

even worse result.


2.4

Analytical protocols

The analytical protocol for the in
-
line execution of the Winkler method is given
in the
supplement
material
s,
No
1
. To achieve a high dispersion of the reagents in the sample zone and already in the
holding coil, the zone stacking pri
nciple was applied by aspiration of three volumes of the same
sample intercalating small volumes of R1 and R2 (VR1, VR2). The aspiration order was 1750

µL
(VS1) of sample, 125

µL of R1, 100

µL of sample (VS2), 125 µL of R2, and 400

µL of sample (VS3).
By t
his procedure, the penetration of VR1 and VR2 and consequently the formation of the
manganese(II) hydroxide precipitate did mainly proceed during the aspiration of VS3 leading to
enhanced dispersion. After 40

s, during which the precipitate could react wit
h DO leading to the
manganese oxyhydroxide

and

the syringe
s were

refilled from the reservoir
s in head
-
valve position
9

OFF
. F
inally
,

the composite zone was dispensed towards the detector. At
the confluence, the
precipitate

dissolve
s

by the merging flow of R3

from syringe 2
under simultaneous formation o
f

I
2
/I
3
-
.

Using the acidic carrier, an additional protocol enabled the quantification of I
2
/I
3
-

in samples processed
off
-
line by the classical Winkler method. For this, a volume of 800

µL of sample was aspirat
ed,
propelled through the thermostatization coil into the detection flow cell, and the absorbance value was
measured in stopped flow over an averaging time of 5

s

in order to compensate fluctuations of the light
source intensity. D
uring
this time,
the rema
ining sample in HC1 was discharged to waste. Finally, the
detection cell was flushed with carrier for cleaning.

Syringe 1 could further be used as an automatic burette for the quantification of I
2
/I
3
-

of off
-
line
processed samples by titration with sodium
thiosulfate. A software protocol was established for the
dispense of user
-
defined volumes between 4 mL and 1

µL including the automated refilling of the
syringe whenever required, counting the total consumed volume of the titration agent, and, if required,

cleaning the syringe with the titrant and at the beginning and end of the procedure, respectively.


2.5 Real
-
time modified SIMPLEX optimization

For the in
-
line execution of the Winkler protocol, the five concentrations of reagents and carrier, five
aspir
ation volumes, and the reaction time had to be optimized. Leaving apart the reaction time, these
variables were interdependent. For example
, a higher concentration of NaOH in R2 requires a higher
concentration of R3 for dissolution of the precipitate; by a

larger VS2, VS3 has to increase in order to
enhance the dispersion and achieve penetration of R1 and R2. Due to the complexity of this
interdependence, the modified SIMPLEX method was used for the optimization of VS2, VS3, VR1 and
VR2 and the concentratio
ns of NaOH and NaI in R2. For this, the volumes of interest were defined by
variables, which values were calculated and directly applied by the used software.

In order to enable in
-
line preparation of R2, the analyzer system was expanded by a thir
d glass
syringe of 10

mL, using

water
as carrier (see figure 1
-
B). S
yringe
3
was connected via a holding coil
HC2 to the central port of the second selection valve with an open and continuously stirred mixing
chamber of 4

mL volume as described elsewhere
[
4
3
,4
4
]

a
nd stock solutions of NaOH and NaI of 2
mol

L
-
1

each positioned on its lateral ports. The mixing chamber was further connected to a lateral port
of the first selection valve. For each optimization experiment, 1

mL of R2 was prepared automatically
10

using the

calculated volumes of the stock solutions (VOH, VI) and a replenishing volume of water.
After application of the so
-
prepared reagent, the mixing chamber was emptied
and cleaned. For this,
the remaining reagent was aspirated by syringe 3

and

dispensed to w
aste
.
Then the mixing chamber
was filled with
water
from
syringe 3

(carrier), which was again
aspirated and discharged

to waste
.

Since simultaneous optimization of all influencing variables was not possible, the following preliminary
considerations were m
ade. To mimic the addition of manganese in the classical Winkler protocol
[
6]

with a final concentration of about 30

mmol

L
-
1
, the concentration of R1 was affixed to 0.3 mol

L
-
1

estimating a dispersion factor of at least 10 for R1 in HC1. Considering a sim
ilar dispersion factor for
R2, a stochiometric ratio of Mn
2+
:OH
-
:I
-
:H
+

of 1:2:3:4, and a merging flow ratio of sample and acid of
5:1, a concentration of 0.5

mol

L
-
1

for R3 guaranteed a surplus of hydronium ion downstream the
confluence. In order to comple
te the optimization within one day and evaluate the influence of the
variables of interest on the reaction kinetic, a reaction time of only 5

s was applied. The sample
volume was fixed by applying VS1 = 2.5

mL


(VS2 + VS3 + VR1 + VR2).
The variation of VS
1 in
function of VR1 and VR2 was done to avoid the
overfilling of the holding coil and to favor SIMPLEX
progress to small reagent volumes.
In order
to suppress
the formation of manganese
precipitate in the
carrier and by this to avoid the alteration of the

signal height by the DO content of the carrier
,
0.1

mol

L
-
1

H
2
SO
4

was used as sample carrier
.

The SIMPLEX optimiz
ation
started from the initial point 50, 200, 150, 250, 150, 300
[
µL] (VOH
-
, VI
-
,
VR2, VR1, VS2, VS3) and a SIMPLEX diameter of 80

%. For opt
imization, N
2

saturated, continuously
sparged
ASW

and air saturated
ASW

where measured in duplicate obtaining the average signal
heights Abs(N
2
) and Abs(Air). To avoid sole optimization of
the methods sensitivity

and evolution to
high blank values and reag
ent consumption
, 20

% of the blank
were

subtracted

from the sensitivity.
Consequently, a lower blank value at constant sensitivity would be evaluated still as a better result.
The
used response function
was
“Abs(Air)


1.2 x Abs(N
2
)”.

Finally, in order to
avoid, that errors during the execution of the optimization experiments such as false
peaks caused by air bubbles could affect the SIMPLEX evolution, the user had to confirm the use of
each data set (four peaks) or else, order the repetition of the last ex
periment by binary input (Yes/No).


2.6 Shipboard application

11

The analyzer system was tested, under true operating conditions, on board of the Spanish
oceanographic research vessel
Hespérides

during a research cruise in the Antarctic Peninsula sector
(Bel
inghausen Sea, Bransfield Strait and Weddell Sea)of the Southern Ocean) in February 2009.
Seawater was sampled, down to 200 to 1,000

m depth, depending on the studied

area, from a SBE 32
carousel water sampler from Sea
-
Bird Electronics Inc. (Washington DC,

USA) equipped with 24 Niskin
bottles of 12

L each from OceanTest Equipment Inc. (Fort Lauderale, FL, USA) combined with a
multiparameter SeaBird 9 CTD sensor registering, among others, temperature, pressure, salinity and
dissolved oxygen. Surface water wa
s sampled using a 30

L Niskin bottle from the same company.
Glass Winkler bottles of about 280

mL and 110

mL were used for sampling for both the in
-
line and
classical Winkler method. The bottles were filled from the Niskin bottles via a silicon tube (50

cm
, 1

cm
i.d.), were let overflow by about twice their volume, and were closed bubble
-
free. For the classical
Winkler protocol, oxygen was fixed following the standard method (see section 2.1).


3 Results and Discussion

3.1 SIMPLEX optimization of in
-
line W
inkler method

Optimization with the modified SIMPLEX method proceeded efficiently using the response function
Abs(Air)


1.2 x Abs(N
2
)
.

O
ptimization of the sole sensitivity corresponding to the
response function
Abs(Air)


Abs(N
2
)
showed to
le
a
d to increas
ing sensitivity but
also
to
an
unaccept
ably high blank
value

of > 1

AU
.
On the other side, increasing the weight of the blank value to 1.3
caused SIMPLEX
evolution to decreasing sensitivity. Only twice
,

experiments had to
be
repeated
by user demand due to
the
presence of air bubbles
,

which had caused
false
peak height evaluation
.
Initial and final vertex
data

including
start and stop conditions are given
in

table 1
.

The o
ptimization stopped automatically after 7 initial and following 32 cycles due to repeti
tion of the
best vertex 10 times. During automation, all SIMPLEX progress modes had been successfully applied.

The parameter set of the vertex in cycle 26 was chosen as the
highest sensitivity was
achieved
. The
optima of the volumetric variables were verif
ied by univariant study of each variab
le with results given
in table 2
. In 3 of 4 cases, the optima found with the modified SIMPLEX method could be confirmed,
thus proving the efficiency of the procedure. However, a larger VS3 showed to improve the sensiti
vity
due to enhanced dispersion of the manganese (II) hydroxide precipitate in the sample zone.

12

It is noteworthy that SIMPLEXS optimization allowed approach
ing

an ideal starting parameter set for
univariant study with a minimum of experiments. Performing t
he optimization in real
-
time was of high
advantage since the labor
and time
for preparation
of reagent 2 as well as for modification of the
volumes of interest in the software protocol fall upon

and t
he SIMPLEX optimization
could be
terminated

in even
less

than
7 hours

(about 10 min for reagent preparation, performance of four
measurements, and cleaning of the mixing chamber afterwards)
.


3.2. Selection of chemical and physical variables for in
-
line Winkler method

Since higher sensitivity was achieved by
increasing VS3, a further study of the remaining variables
was done including the reaction time and the sulfuric acid concentrations of R3 and the carrier. The
experiments were started from the former found optima 125

µL of R1, 190

µL of R2, 100 µL of VS2,

300

µL of VS3,
and 0.2

mol

L
-
1

NaOH and 0.4

mol

L
-
1

NaI of R2
.

The influence of the volume VR2 on the analytical response was studied over the range of 100 to
275

µL for concentrations of NaOH and NaI of 0.2

mol

L
-
1

and 0.4

mol

L
-
1
, respectively. Experim
ental
results are depicted in figure 2
-
A. Increasing VR2 led to higher peaks for both N
2

and air saturated
ASW

but decreasing sensitivity.
To use the entire absorbance range (up to 1 AU) for the DO
concentration range of interest (0


400 µmol L
-
1
),
a volu
me of 125

µL R2 was chosen.

Testing the linearity of the method up to 9

mg

L
-
1
, it turned out, that iodide was in limiting
concentration. Therefore, the iodide concentration was heightened to 0.6

mol

L
-
1
, without any
significant alteration of the method’s

sensitivity but gaining linearity up to 16

mg

L
-
1
. Though the
isosbestic wavelength of I
2
/I
3
-

was chosen for detection, a higher iodide concentration further
warranted the presence of iodine as tri
-
iodide and by this, offered the possibility of reliable d
etection at
shorter, more sensitive wavelengths.

The influence of the sulfuric acid concentration of R3 on the analytical response was studied over the
range of 0.2 to 1

mol

L
-
1
. Experimental results are depicted in figure 2
-
B. While no significant influe
nce
on the peak height of N
2

saturated
ASW

was observed, the peak height of air saturated
ASW

decreased nearly linear with increasing acid concentration. To ensure the complete dissolution of the
precipitates at a minimal consumption of acid, a concentrati
on of 0.3

mol

L
-
1

H
2
SO
4

was chosen.

13

The influence of the reaction time on peak height was studied for 5, 10, 20, 40, and 80 s. Experimental
results are depicted in figure 2
-
C. The peak heights of N
2

saturated
ASW

increased nearly linear with
time, whereas
for air and oxygen saturated
ASW
, considerable decelerations of the increments were
observed with increasing time. Since the observed improvement passing from 40

s to 80

s was low, a
reaction time of 40

s was chosen as a compromise between time of analysis

and sensitivity. The
increase of the blank value was likely due to diffusion of oxygen through the PTFE tubing walls into
the sample. So, a longer reaction time would mainly lead to a lower affection of the ambient
temperature due to a more progressed rea
ction but not to a significant increase of sensitivity.

The influence of the sulfuric acid concentration of the carrier on the analytical signal was studied over
the range of 0.025 to 0.3

mol

L
-
1
. Experimental results are depicted in figure 2
-
D. While no s
ignificant
influence on the peak height of N
2

saturated
ASW

was observed, the peak height of air saturated
ASW

clearly decreased slightly with higher acid concentration. As a compromise between sensitivity and the
requirement to suppress the formation of m
anganese precipitate in the carrier, a concentration of
0.05

mol

L
-
1

was chosen.

The influences of the volumes VS2 and VS3 on the analytical readout were studied over the range of
0 to 150

µL and 100 to 400

µL, respectively. Experimental results are depic
ted in figure 3. Increasing
VS2 led to lower peak heights for N
2

saturated
ASW

while for air saturated
ASW

a maximum was
found for 50

µL. The highest sensitivity was achieved at a volume of 100

µL, which was therefore
applied further on. A higher VS3 led t
o increasing sensitivity due to a higher dispersion of the
manganese (II) hydroxide precipitate in the sample zone and consequently a larger reaction volume of
sample. Therefore, a volume of 400

µL was chosen; larger volumes were not tested since the
achie
ved sensitivity allowed to cope with the desired working range of 13

mg

L
-
1
.

As former reported
[
19
]
, a brown coating of the inner tubing walls of the holding coil by manganese
oxyhydroxide
was observed. This for one proved the fully separation of the acid
ic carrier from the
reaction composite zone but affected both reproducibility and the method's sensitivity. In order to
proceed a cleaning of the holding coil prior to sample analysis, a small volume of R2 (50

µL) was
aspirated prior to VS1. Applying this
step, the precipitate dissolved immediately at the penetration
zone of R2 and the acidic carrier. An aspiration step


former
performed with the syringe head
-
valve in
position OFF
-

was required nevertheless in order to overcome the backlash of the syringe

module
and improve the precision of the following aspiration step, so did require additional time.

14


3.3 Analytical performance
of the in
-
line Winkler method

The entire analytical method required 120

s, where 40

s were for the oxidation time of the mangan
ese
precipitate. A sample frequency of 30 full analyses per hour is therefore achievable during monitoring
but is slightly reduced by a cleaning protocol for the sample channel when discrete samples are
measured. The high sample frequency was enabled by ca
rrying out mixture of sample and reagents
and precipitate oxidation already in the holding coil.

Linearity was proven
up to
16

mg

L
-
1

(
480 µmol

L
-
1
) following a calibration function of 0.020

L

mg
-
1


+
0.15 AU

(see supplementary materials, No 3)
. The relati
ve standard deviation (RSD) of the blank was
generally <

1.5

%. The limit of detection, calculated as the DO content yielding a signal equal to the
blank plus three times the blank's standard deviation
, was 0.34

mg

L
-
1

applying 466

nm as detection
waveleng
th
.

An increment by factor 2.2 of the sensitivity and blank value can be achieved by using a
detection wavelength of 440

nm, yielding a higher sensitivity than in former reported reversed FIA
analyzer for automation of the Winkler methodology
[
19
]
.

A typic
al calibration example is given in
Figure 4
-
A.

In comparison with the classical Winkler method, the consumption of reagents and sample was highly
reduced. Only 2.25

mL of sample, 0.125

mL of R1 and R2, 0.7

mL of R3, and 3.5

mL of carrier were
required for
one analysis, which represents a noteworthy advantage for monitoring applications and,
particularly, for experiments that allow limited sample sizes
[
4
5
]
.

The system showed a notable memory effect affecting generally the first peak of repeated sample
measu
rements. This effect was considered to be attributed to diffusion of gases into pores of the
PTFE tubing material.


3.4 Determination of I
2
/I
3
-

by
titration and

spectrop
hotometry

For titration, syringe 1 was used as automated burette using 0.2

mol

L
-
1

sod
ium thiosulfate as titration
agent. Three
-
fold repetitions of titration of two Southern Ocean seawater samples (10

m depth:
7.5

mg

L
-
1

DO; 41

m: 7.4

mg

L
-
1

DO) prepared by the classical Winkler protocol gave RSD <

0.2

% or
a standard deviation of <

3

µL ti
tration agent applying 250

mL
of sample, respectively. Lowering the
15

concentration of the titration agent would either allow increasing the precision or lowering the required
volume of sample. For the entire procedure including syringe refilling, generally
less than 2

min were
required. The procedure was unaffected by shipboard application with dispense volumes as low as
1

µL.

Without variation of the manifold, the analyzer could be further applied to the spectrophotometric
determination I
2
/I
3
-

in samples p
repared by the classical Winkler protocol. The volumes of sample
aspiration, dispense to the detection cell, and flushing after measurement were properly adjusted (see
section 2.4) to perform measurement of the steady state sample concentration of I
2
/I
3
-
.
Relative large
volumes had to be applied as former reported
[
16]

due to adsorption of I
2
/I
3
-

at the tubing inner walls,
leading otherwise to an increasing baseline and memory effect.

In comparison with a former described continuous flow analyzer
[
1
6
]
, a t
hreefold higher sample
frequency of 90

h
-
1

was achieved. For cleaning of the sample channel, additional 15

s were required
after sample change. The methods featured a high

reproducibility with
an
average RSD of 0.8

%
(n

=

5)
corres
ponding to LOD
and LOQ of

<

0.01

mg

L
-
1

and
0.
06

mg

L
-
1
,

respectively, being
comparable with commercial
titration automats
.
The
analytical response
was
0.043



0.002)

L

mg
-
1

+

0.01



0.02)

AU
with mean correlation coefficient of 0.999
averaging 21
daily
shipboard calibrations. C
omparing the content of DO of 50 seawater samples determined by titration
and spectrophotometry, the average deviation was 2.8

% or 0.
3

mg

L
-
1
, respectively
. The variation of
the laboratory temperature of T
Max



T
Min

= 15

K affecting the density of the KIO
3

standard stock and
titrant ha
s

to be considered as main error sources.
A typical calibration example is given in Figure 4
-
B
.

A comparison of analytical results obtained by applying the classical Winkler procedure but applying
iodine quantification by tit
ration with thiosulfate or spectrophotometric quantification, respectively, is
given in
F
igure 5
-
A

proving the good data agreement between both methods
.


3.5 Shipboard application of the in
-
line Winkler method

In comparison with laboratory operation,
a low
er reproducibility of the in
-
line Winkler method during
shipboard application was found.
In the first, this was reduced to the considerable
variation of the
ambient temperature affecting the dispersion of the sample
in the holding coil and the
reaction
kin
etic
.

In the second, g
as bubble evolving
in
the sample at warming up in the tubing manifold
can further
16

affect the dispersion and mixing of the sample with the reagents
.
Nevertheless, a good reproducibility
with an average RSD of 3.1

% (n = 5) was obtained

leading to a LOD of 0.68

mg

L
-
1
.

The sensitivity and blank value were increased during shipboard application with an analytical
response of 0.02
7



0.00
3
)

L

mg
-
1

+

0.18



0.05)

AU averaging 6 calibrations for likely the same
reasons. This observation w
as explained by the affection of the ship
-
movement on the
notably
denser
precipitate
leading to a constant motion in the sample volume, and higher
turbation

and contact
volume.

Comparing the content of DO of
over 30
seawater samples processed with both wit
h in
-
line Winkler
method and with the classical Winkler protocol following spectrophotometric determination, a mean
deviation of 1.4

% ± 6.7 % was obtained, thus proving
good
agreement and accuracy of the proposed
in
-
line Winkler method.

A comparison of an
alytical results obtained by applying the in
-
line Winkler
method and the spectrophotometric iodine quantification for the classical Winkler protocol (operation

mode 2) is given in
F
igure 5
-
B

proving the usability of the in
-
line method for DO quantification
.

The in
-
line Winkler method can be of especial interest for monitoring applications or for small sample
volumes, whereas the both further operation modes present alternatives for potentiometric titrations
especially with advantages in respect of labor and

time. The spectrophotometric detection presents
further an alternative to titration procedures since it did not
any
require further reagents.


4

Conclusion

A robust and reliable sequential injection analyzer system for in
-
line automation of the entire Wi
nkler
method was presented. To this end, volumetric and concentration parameters were optimized by real
-
time modified SIMPLEX method. The optimized and characterized system featured a
high
sample
frequency and sensitivity
appropriate for oceanographic appl
ications
over a wide working range with a
highly reduced consumption of reagents and sample compared to the classical Winkler method. The
system enabled further operation modes for the spectrophotometric determination as well as the
batch
-
wise titration of

I
2
/I
3
-

in samples prepared according the classical Winkler protocol. Applying the
analyzer to real seawater samples, conformity with determinations
,

proce
ss
ed
according
the classical
Winkler protocol
,

was found.
The system
was successfully used during a r
esearch cruise in the
Southern Ocean, which


with
high
temperature differences between in situ (
-
1 º C to 2 ºC) and the
17

laboratory, which experienced broad, 15 K, temperature oscillations
-

provides demanding conditions.
The results were satisfying in res
pect of reproducibility and precision for all operation modes, thereby
providing an avenue to increase the analytics throughout and lower the analytical cost, while achieving
satisfactory precision, for dissolved oxygen concentration, a key parameter in oc
eanographic
research.


Acknowledgements

This work was funded by the projects "Design and application of analytical systems for the
quantification of trace elements in the ocean" supported from the Government of the Balearic Islands
(CAIB Progecib
-
5C), and

of the ATOS project, funded by the Spanish Ministry of Science and
Innovation (POL2006
-
00550), a Spanish contribution to the International Polar Year. BH was funded
by a JAE Postdoctoral fellowship from CSIC.

We

further
thank
Marta Alvarez Rodriguez for u
seful
information.

18

References

[1]

A. Ganachaud
,
C.
Wunsch, Improved estimates of global ocean circulation, heat transport and
mixing from hydrographic data,
Nature 408

(2000)

453
-
457

[2]

C.
R
obinson
,
P.J. Le B.
W
illiams
,

Respiration and its measurement in surfac
e

waters, i
n:
P.A.
del Giorgio
,

P.J. Le B.
Williams (
E
ds.), Respiration in Aquatic Ecosystems. Oxford Univ. Press.

2005

[3]

R.

Vaquer
-
Sunyer,

C.M.
Duarte,
Threshold for hypoxia for marine biodiversity,
PNAS 105
(2008) 15452
-
15457

[4]

L.W.
Winkler, Chem. Ber. 27
Die

Bestimmung des in Wasser gelősten Sauerstoffes
(1888)
2843
-
2855

[5]

APHA, AWWA, WPCF, Standard methods for the examination of water and wastewater, L.S.
Clesceri, A.E. Green
berg, R. Rhodes Trussell
(
Ed
s
.
)

20th Edition, Washington DC.

1998

[6]

A.G.
Dickson, Determ
ination of dissolved oxygen in sea water by Winkler titration
,

WOCE
Operations Manual. P
art 3.1.3 Operations & Methods,

WHP Office Report WHPO 91
-
1.
1995

[7]

K.
Grasshoff, K.

Kremling,

M.

Ehrhardt,
Methods of seawater analysis
,

third ed.
, Wiley
-
VCH,
Weinheim

19
76

[8]

J.F.
Gouin,
F.
Baros,
D.
Birot,
J.C.
André
A fibre
-
optic oxygen sensor for oceanography,
Sensors
and Actuators B 38
-
39

(1997)

401
-
406.

[9]

D.
Xiao,
Y.
Mo,
M.M.F.
Choi
,

A hand
-
held optical sensor for dissolved oxygen measurement,
Meas. Sci.
Technol. 14

(2003)

8
62
-
867

[10]

A. Tengberg, J. Hovdenes, J.H. Andersson, O. Brocandel, R. Diaz, D. Hebert, T. Arnerich, C.
Huber, A. Körtzinger, A. Khripounoff, F. Rey, C. Rönning, J. Schimanski, S. Sommer, A.
Stangelmayer, Evaluation of a lifetime
-
based optode to measure oxygen
in aquatic systems,
Limnol. Oceanogr.: Methods 4 (2006) 7
-
17

[11]

M. Sosna, G. Denuault, R.W. Pascal, R.D. Prien, M. Mowlem, Field assessment of a new
membrane
-
free microelectrode dissolved oxygen sensor for water column profiling. Limnol.
Oceanogr.: Methods 6
(2008) 180
-
189

19

[12]

F.
Roland
, N.F.Caraco, J.J.
Cole
, Rapid and precise determination of dissolved oxygen by
spectrophotometry: evaluation of interference from color and turbidity,
Limnol.
Oceanogr. 44
(1999) 1148
-
1154

[13]

S.
-
C. Pai, G.
-
C.
Gong,
K.
-
K.
Liu,
Determina
tion of dissolved oxygen in seawater by direct
spectrophotometry of total iodine,
Mar. Chem. 41 (1993) 343
-
351

[14]

J.D.
Burger
,

H.A.
Liebhafsky,
Thermodynamic data for aqueous iodine solutions at various
temperatures. Exercise in analytical chemistry
,
Anal. Ch
em. 45

(1973)

600
-
602

[15]

T.
Labasque,
C.
Chaumery,
A.
Aminot,
G.
Kergoat,
Spectrophotometric Winkler determination
of dissolved oxygen: re
-
examination of critical factors and reliability,
Marine Chemistry 88

(2004)

53
-
60


[16]

T.

Reinthaler,

K.

Bakker,

R.

Manuels,

J.

van Ooijen,

G.J.

Herndl,

Fully automated
spectrophotometric approach to determine oxygen concentrations in seawater via continuous
-
flow analysis,
Limnol. and Oceanogr.

Meth
.

4

(2006)

358
-
366


[17]

M.
Novic, B

Pihlar,

M

Dular,
.
Use of flow injection analysis

based on iodometry for automation of
dissolved oxygen (Winkler method) and chemical oxygen demand (dichromate method)
determinations,
Fresenius’ Z. Anal. Chem. 332 (1988) 750
-
755

[18]

T.
Sakai,
H.
Takio,
N.
Teshima,
H.
Nishikawa,
Extraction
-
flow injection spectr
ofluorimetric
measurement of dissolved oxygen in environmental waters using 2
-
thionaphthol,
Anal. Chim.
Acta 438 (2001) 117
-
121

[19]

S.
Muangkaew, I.D.

McKelvie,

M.R.

Grace,

M.

Rayanakorn,

K.

Grudpan,

J.

Jakmunee,

D.

Nacapricha,

A reverse
-
flow injection analysi
s method for the determination of dissolved oxygen
in fresh and marine waters
,
Talanta 58
(2002) 1285
-
1291

[20]

T.
Sakai,
S.
Piao,
N.
Teshima,
T.
Kuroishi,
K.
Grudpan, Flow injection system with in
-
line Winkler’s
procedure using 16
-
way valve and spectrofluorimetric

determination of dissolved oxygen in
environmental waters
,

Talanta 63 (2004) 893

898

[21]

J.
Ruzicka, E.H.
Hansen, Flow Injection Analysis. Part 1. A New concept of fast continuous flow
analysis,
Anal.
Chim. Acta 78

(1975)

145
-
157.

[22]

Trojanowicz, M. (
Ed.
)

Flow i
njection analysis. Instrumentation and applications
,

World Scientific
Publishing, Singapore

2000

20

[23]

M.Trojanowicz

(
Ed.
)

Advances in Flow Methods of Analysis
,
Wiley
-
VCH, Weinheim

2008

[24]

S.D.
Kolev,
I.D.McKelvie, (Eds)

Advances in Flow Injection Analysis and rela
ted techniques.
Compreh
ensive Analytical Chemistry 54,
Elsevier, Netherlands

2008

[25]

A. Cerdà, V.
Cerdà, An introduction to flow analysis (2009) Sciware, Palma de Mallorca

[26]

J.
Ruzicka,
G.
Marshall,
Sequential injection: a new concept for chemical sensros, proces
s
analysis and laboratory assays,
Anal.
Chim.
Acta
237

(1990)

329
-
343.

[27]

A.
Economou,
Sequential
-
injection analysis (SIA): A useful tool for on
-
line sample handling and
pre
-
treatment,
T
R
A
C
24

(2005)

416
-
425

[28]

C.E. Lenehan, N.W.
Barnett,
S.W.
Lewis,
Sequential i
njection analysis,
Analyst 127

(2002)

997
-
1020

[29]

V. Cerda, R.
Forteza,
J.M.
Estela,
Potential of multisyringe flow
-
based multicommutated
systems,
Anal.
Chim. Acta 600

(2007)

35
-
45

[30]

B.
Horstkotte, O.

Elsholz,

V.
Cerdá, Review on Automation using Multisyringe F
low Injection
Analysis,
J. Flow Injection Anal 22

(2005)

99
-
109

[31]

V. Cerda, J.M. Estela, R. Forteza, A. Cladera, E. Becerra, P. Altimira, P. Sitjar, Flow techniques
in water analysis, Talanta 50 (1999) 695
-
705

[32]

J.
Nelder,
R.
Mead,
A Simplex Method for Functio
n Minimization,
Computer Journal 7

(1965)

308
-
313

[33]

M.

Blanco,
V.
Cerda,
Quimiometria
,

Publications de la Universitat Autonoma de Barcelona,
Bellaterra, Spain 1988, ISBN 84
-
7488
-
395
-
4

[34]

D.
Betteridge,
T.J.
Sly,
A.P.
Wade,
J.E.W.
Tillman, Computer
-
Assisted Optimiz
ation for Flow
Inj
ection Analysis of Isoprenaline,

Anal. Chem. 55

(1983)

1292
-
1299

[35]

A.
Pasamontes, P.
Callao, Fractional factorial design and simplex algorithm for optimizing
sequential injection analysis (SIA) and second order calibration
,
Chemometrics and
Intelligent
Laboratory Systems
83

(2006)

127
-
132

[36]

H.A.D.F.O. Silva, L.M.B.C.
Alvarez
-
Ribero,
Optimization of a flow injection analysis system for
tartaric acid determination in wines,
Talanta 58

(2002)

1311
-
1318

[37]

M.F.
Gine,
R.L.
Tuon,
A.A.
Cesta,
A.P.
Packer, B.
F.
Reis, Real
-
time simplex optimization of flow
-
injection systems for chemical análisis,
Anal
.

Chim
.

Acta 3668

(1998)

313
-
318

21

[38]

B.F. Reis, M.F.
Giné,
E.A.G.Zagatto, J.L.F.C.Lima,
R.A.S.
Lapa, Multicommutation in flow
analysis, Part 1, Binary sampling: Concept
s, instrumentation and spectrophotometric
determination of iron in plant digests,
Anal
.

Chim
.
Acta 293

(1994)

129
-
138

[39]

F.R.P. Rocha, B.F.Reis, E.A.G.Zagatto, J.L.F.C.Lima, R.A.S.Lapa, J.L.M. Santos,
Multicommutation in flow analysis: concepts, applications
and trends, Anal.
Chim. Acta 468
(2002) 119
-
131

[40]

A.C.B. Dias, E.P.Borges, E.A.G.Zagatto, P.J.Worsfold, A critical examination of the components
of the Schlieren effect in flow analysis, Talanta 68 (2006) 1076
-
1082

[41]

W.W. Broenkow, J.D. Cline, Colorimetric det
erminations of dissolved oxygen at low
concentrations, Limnol.
Oceanogra. 14 (1969) 450
-
454

[42]

E. Becerra, A.Cladera, V.
Cerda,
Design of a Very Versatile Software Program for Automating
Analytical Methods,
Lab.
Ro
botics & Automat. 58 (1999) 131
-
140

[43]

F.

Albert
ús, B.

Horstkotte
, A.

Cladera,

V.
Cerdà
,
A robust multisyringe system for process flow
analysis, Part I. On
-
line dilution and single point titration of protolytes
,

Analyst 124

(1999)

1373
-
1381

[44]

N.
Ornelas Soto,
B.
Horstkotte, J.G.

March,

P.L.

Lopez de Alba,

L.

Lopez Martiınez,

V.

Cerdà
,

An environmental friendly method for the automatic determination of hypochlorite in commercial
products using multisyringe flow injection analysis,
Anal
.

Chim
.

Acta
611 (
2008
)

182
-
186

[45]

C.M.
Duarte,
J. Dachs
,
M.
Llabrés,
P.
Alonso
-
La
ita
,
J.M. Gasol
,
A.
Tovar
-
Sánchez,
S. Sañudo
-
Wilhemy
,
S.
Agustí
,

Aerosol inputs enhance new production in the Subtropical NE Atlantic
,
J.
Geophys. Res.

111

(2006)

G04006, doi:10.1029/2005JG00


22


Figure 1:
Scheme of the SIA analyzer system in normal con
figuration (Part A) and during
SIMPLEX optimizati
on (including Part B). HC1: Holding coil 1 (350

cm, 1.5

mm

i.d.), HC2:
Holding coil 2 (1

m, 1.5

mm

i.d.), TC: Thermostatization coil (120

cm, 0.8

mm

i.d.), D:
Spectrophotometer USB2000 (λ
Det
: 466

nm, λ
Ref
: 580

nm), light source, and flow cuvette,
additiona
l tubes a: 6

cm, 1.5

mm

i.d., b, c: 10

cm, 0.8

mm

i.d., S: Sample. V1,V2,V3: 3
-
way
multicommutation solenoid valves with normally closed positions (ON, activated) dotted and
normally open positions (OFF, deactivated) straight.

23


Figure 2: Evaluation of th
e effect of A: Volume of R2, B: concentration of sulfuric acid of R3,

C: reaction time, and D: concentration of carrier sulfuric acid of carrier on the analytical
readouts.

24


Figure 3: Evaluation of the effect of the volumes of sample VS2 and VS3 on the
analytical
readouts.


25


Figure 4
: Calibration using the on
-
line Winkler mode (A) and off
-
line validation mode (B).

Conditions A:
125

µL of
0.3

mol

L
-
1

MnSO
4

,
125

µL of
0.2

mol

L
-
1

NaOH and 0.6

mol

L
-
1

NaI,
confluent acidification with 0.3 mol

L
-
1

H
2
SO
4

,

reaction time 40 s
,
s
ample volumes 1
-
3

were
1850 µL, 100

µL, and 300

µL, respectively.


26


Figure
5
:
Comparison analytical results for southern ocean seawater samples

obtained by
different operation modes.
A: titration (operation mode 3) versus spectropho
tometric iodine
quantification (operation mode 2) both for of samples processed according the classical
Winkler protocol. B
:
in
-
line Winkler (operation mode 1)
versus
spectrophotometric
iodine
quantification
(operation mode 2) of samples processed accordin
g the classical
Winkler
protocol.


27

Table 1: SIMPLEX

optimization
of the in
-
line Winkler method*.

Cycle

c(OH
-
)

[
mol L
-
1
]

c(I
-
)

[
mol L
-
1
]

VR1

[
µL]

VR2

[
µL]

VS2

[
µL]

VS3

[
µL]

N
2

**

[
AU]

Air
*
*

[
AU]

Air


N
2

[
AU]

Air


1.2 ∙ N
2

[
AU]

01

0.140

0.492

174.5

281
.5

165.5

326.0

151.3

179.7

28.4

-
1.9

02

0.060

0.492

174.5

281.5

165.5

326.0

80.3

59.4

-
20.9

-
37.0

03

0.116

0.216

174.5

281.5

165.5

326.0

100.2

106.1

5.9

-
14.1

04

0.100

0.400

76.5

281.5

165.5

326.0

50.8

48.6

-
2.2

-
12.4

05

0.100

0.400

150.0

123.5

165.5

3
26.0

91.1

90.4

-
0.7

-
18.9

06

0.100

0.400

150.0

250.0

72.5

326.0

130.4

153.0

22.6

-
3.5

07

0.100

0.400

150.0

250.0

150.0

143.0

85.0

99.2

14.2

-
2.8























11

0.190

0.440

127.5

190.0

119.0

150.0

112.4

167.1

54.7

32.2























17

0.206

0.296

113.5

166.0

91.0

203.5

119.5

176.0

56.5

32.6























26

0.202

0.472

131.0

194.0

100.5

225.0

135.7

192.4

56.7

29.6

* Initial SIMPLEX diameter 80

%, thresholds: 40
-
600, 40
-
600, 25
-
250, 25
-
250, 50
-
300, 150
-
1250, stop
criteria:
maximal 40 cycles, minimum 0.1

% improvement, and maximal 10 repetitions of best point.

*
*

N
2
, Air: Peak heights
obt
ained from N
2

and air
-
saturated
ASW

corresponding to blank and standard
solutions
.

28

Table
2
: Verification of the SIMPLEX optimization with v
olumes yielding maximal sensitivity in
bold (SIMPLEX optima estimation were VR1: 130

µL, VR2: 195

µL, VS2: 100

µL, and VS3:
225

µL).

VR1
[
µL] *

50

75

100

125

150

175

200

225


N
2

sat.
ASW

[
AU]

0.132

0.143

0.151

0.171

0.172

0.183

0.189

0.191


Air sat.
ASW

[
AU]

0.190

0.201

0.214

0.232

0.233

0.238

0.243

0.247


Difference

[
AU]

0.058

0.057

0.063

0.061

0.061

0.056

0.055

0.056


Regressio
n of difference and maximum:
-
5.
429E
-
07∙VR1
2

+ 1
.
279E
-
04∙VR1 + 5
.
284E
-
02 /
118 µL

VR2
[
µL] **

100

125

150

175

200

225

250

275


N
2

sat.
ASW

[
AU]

0.103

0.112

0.127

0.141

0.152

0.163

0.169

0.178


Air sat.
ASW

[
AU]

0.162

0.179

0.201

0.223

0.225

0.237

0.234

0.241


Difference

[
AU]

0.059

0.067

0.074

0.082

0.073

0.074

0.065

0.063


Regression of difference and maximum: 2
.
219E
-
06∙VR2
2

+ 8
.
364E
-
04∙VR2


1
.
911E
-
03 /
189 µL

VS2
[
µL]

0

25

50

75

100

125

175

200

225

N
2

sat.
ASW

[
AU]

0.192

0.179

0.177

0.171

0.168

0.152

0.144

0.130

0.125

Air sat.
ASW

[
AU]

0.233

0.231

0.227

0.219

0.221

0.204

0.194

0.184

0.172

Difference

[
AU]

0.041

0.051

0.0
50

0.048

0.053

0.052

0.050

0.053

0.048

Regression of difference and maximum:
-
4
.
812E
-
07∙VS2
2

+ 1
.
299E
-
04∙VS2 + 4
.
380E
-
02 /
135 µL

VS3
[
µL]

100

150

175

200

225

250

275

300

325

N
2

sat.
ASW

[
AU]

0.121

0.141

0.140

0.144

0.148

0.147

0.151

0.153

0.149

Air sa
t.
ASW

[
AU]

0.164

0.197

0.200

0.205

0.210

0.215

0.216

0.220

0.218

Difference

[
AU]

0.043

0.056

0.060

0.061

0.062

0.068

0.065

0.067

0.069

Regressio
n of difference and maximum:
-
5.
557E
-
07∙VS2
2

+ 3.
393E
-
04∙VS3 + 1
.
592E
-
02 /
305 µL

* 0.3

mol

L
-
1

MnSO
4

**
0.
2

mol

L
-
1

NaOH and 0.4

mol

L
-
1

NaI

29

Supplementary materials

Supplement
1
: Automated procedure of in
-
line Winkler protocol for
the quantification of
dissolved oxygen exploiting sequential injection analysis.*

Instrument

Instruction

Description

Valve

Valve
A move to position 6

Position of R2

Syringe
s

Pickup 50 µL at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of R2 for holding coil
cleaning and simultaneously
overcome of syringe backlash

Valve

Valve A move to position 2

Position of sample

Syringe
s

Pickup 1.75 mL
at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of first sample volume

Wait

Wait 1 seconds


Valve

Valve A move to position 5

Position of R1

Syringe
s

Pickup 125 µL at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of R1

Wait

Wait 1 seconds


Valve

Valve A move to position 2

Position of sample

Syringe
s

Pickup 100 µL at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of second sample volume

Wait

Wait 1 seconds


Valve

Valve A move to position 6

Position of R2

Syringe
s

Pickup 125 µL at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of R2

Wait

Wait 1 seconds


Valve

Valve A move to position 2

Position of sample

Syringe
s

Pickup 400 µL at 6.6 mL/min
[
1
-
Off 2
-
On]

Aspiration of third sample volume

Wait

Wait 40 seconds

Incubation and refilling of syringes

Syringe
s

Priming in pickup at 10 mL/min
[
1
-
Off 2
-
Off]

Syringe refilling

Valve

Valve A move to position 8

Position of detector

Spectrometer

λ
Det
: 466 nm /
λ
Ref
: 620 nm at 3.3 Hz

Measurement start

Syringe
s

Dispense 2.5 mL at 6.6 mL/min
[
1
-
On 2
-
On]

Acid merging and detection of I
2
/I
3
-

Spectrometer

Stop measure

Measurement stop

Syringe
s

Dispense 1 mL at 10 mL/min
[
1
-
On 2
-
On]

Discharge of sampl
e

* All volumes and flow rates refer to a syringe size of 5

mL (Syringe 1)
, size of syringe 2 was 1

mL
.

30


Supplement
2
: Linear calibration of in
-
line Winkler
system with artificial seawater, conditions
as given in figure 4.
Aliquots of the samples measu
red with the in
-
line Winkler method were
fixated according
the
classical
off
-
line Winkler method

and their DO content was quantified by
operation mode 3 (titration with
0.2 mol

L
-
1

Na
2
S
2
O
3
).