You are on page 1of 9

Analytica Chimica Acta 387 (1999) 165173

Automatic potentiometric titration in monosegmented


ow system exploiting binary search
Patrcia B. Martelli, Boaventura F. Reis*, Mauro Korn1, Jose L.F. Costa Lima2
Centro de Energia Nuclear na Agricultura Universidade de Sao Paulo, PO Box 96, Piracicaba, SP 13400-970, Brazil
Received 27 August 1998; received in revised form 5 January 1999; accepted 9 January 1999

Abstract
An automatic monosegmented ow system (MSFA) based on the binary search concept to perform potentiometric titration is
proposed. A tubular hydrogen ion-selective electrode without inner reference solution, consisting of a conducting epoxy
cylinder machined with an axial hole coated with tridodecylamine (TDDA) was employed as a sensor. The titration procedure
was implemented by exploiting the binary search approach, after each analytical cycle data were evaluated to decide the
variation of the titrant volumetric fraction to be inserted for the next cycle. The ow network comprised three-way solenoid
valves controlled by a microcomputer running software in QUICKBASIC 4.5. Combination of binary search and MSFA resulted in
an automatic system providing possibilities to perform ow titrations without any operator assistance. The main features of the
proposed system were veried by titrating hydrochloric and acetic acid solutions and the feasibility of the approach was
ascertained by analyzing vinegar, coke, lemon soda, isotonic, industrial and natural orange juice samples. Results were in
agreement with those obtained with a conventional potentiometric titration, and no signicant difference at 95% condence
level was observed. A 1% standard deviation (n9) in results was also observed. # 1999 Elsevier Science B.V. All rights
reserved.
Keywords: Potentiometric titration; Tubular pH electrode; Monosegmented ow; Binary search; Multicommutation

1. Introduction
Titration is a widely used technique to monitor
industrial process, presenting as disadvantage the long
time required when a manual titrimetric procedure is
employed. To overcome this drawback, automatic
batch titrators and continuous ow techniques have
*Corresponding author. Fax: +55-19-4294610; e-mail:
reis@cena.usp.br
1
Universidade do Estado da Bahia, Salvador BA, Brazil.
2
CEQUP/Faculdade de Farmacia Universidade do Porto,
Portugal.

been proposed. The rst automated titration set up was


introduced by Zeigel in 1914 [1]; addition of the titrant
was done by controlling a burette by means of an
electromagnetic device. The requirement of mechanical devices to deliver the titrant solution and to
homogenize the mixture in general could make the
analytical procedure expensive. This disadvantage
could be circumvented by employing continuous
ow techniques, that present as advantages high
throughput, good precision, and that can be implemented with low cost apparatus. In this sense, the
classical work introduced by Blaedel and Laessig [2]
can be mentioned; the sample stream was pumped

0003-2670/99/$ see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S0003-2670(99)00092-6

166

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

at a constant ow rate and the reagent stream was


varied linearly up to attain the end point. Following
this work, several procedures have been described
to carry out titrations in continuous ow systems
[3,4]. The manifolds were designed to perform
mixing between analyte and titrant solutions into a
reaction coil followed by detection. In general, the
analyte solution ows at constant ow rate while
the titrant solution ow rate is varied up to reach
the end point.
Another ow system exploiting a gradient titration
technique was developed by Fleet and Ho [5] based on
the variation of the titrant concentration, while the
other parameters were constant. The feasibility of the
procedure was veried by titrating sulde ion with
mercuric nitrate solution monitoring with a sulde
selective electrode. Also, a titration gradient based on
the triangle-programmed titration technique was
developed [6]. The titrant was coulometrically generated in a triangular way by employing a controlled
current source. After mixing with the sample stream at
constant ow rate, two end points were obtained. The
time elapsed between them was directly proportional
to the analyte concentration.
Since its introduction in 1975 [7], ow injection
analysis (FIA) has received considerable interest, and
a number of methods have been adapted for routine
analysis in different elds. The FIA titration technique
was introduced by Ruzicka et al. [8] and afterwards
several systems have been proposed for different
applications [911]. A basic difference between ow
injection titration and other FIA procedures is the use
of peak width rather than peak height as the analytical
signal. In the original work [8], the titrant solution
concentration was used as a carrier stream. After
sample injection, it was mixed with titrant in a gradient chamber, where chemical reactions occurred
before detection. In this sense, no reaction stoichiometry was reached as in conventional titrations. The
time interval elapsed between signal appearance
and its return to the baseline was considered as the
analytical parameter. This work showed some advantages such as high sampling rate and good precision.
Nevertheless, it requires analytical curves to determine the analyte concentration and the titration curves
presented a non-linear behavior. Since this procedure
is based on analytical curves, it is considered a
pseudo-titration according to the IUPAC denition

[12] and also considering the arguments by Pardue


and Fields [13].
Ingenious continuous ow titration approaches
based on changing the ow rate to generate concentration gradients have been proposed [14,15]. The
system comprised two pumps, one running with constant ow rate and the other ow rate was changed in
the range from 0 to 5 ml min1. A linear concentration
gradient was obtained by varying the ratio of titrant
and sample ow rates. Other work employing two
peristaltic pumps and exponentially changing the ow
rates of titrant and titrand solutions was presented
[16]. The authors mentioned as advantage a decrease
in the time interval to attain the end point when
compared with methods based on linear ow rate
variation [14,15].
All the above mentioned titration techniques require
calibration curves to obtain the analyte concentration.
Flow titrations without calibration runs would be
possible if one knows the volumes of sample and
titrant which should be constant under ow conditions
[17]. In this sense, the work presented by Korn et al.
[18] based on the binary search concept fullls this
requirement. The stoichiometry condition was
attained by stepwise varying both sample and titrant
volumetric fractions. The sample concentration was
obtained directly and no calibration curves were
necessary.
Dispersion is an inherent feature of the ow system,
and as a consequence sample distribution inside the
carrier stream presents a gradient pattern, the MSFA
can be used to minimize this effect [19]. In a ow
system based on this approach, sample aliquots are
introduced in the analytical path between two air
bubbles in order to minimize the contact of sample
solution with carrier solution, resulting in a low dispersion effect.
In the present work, an automatic potentiometric
titration ow procedure based on the binary search
concept [18] employing a monosegmented ow system [19] was developed. A tubular ion-selective electrode for hydrogen ion based on the ionophore
tridodecylamine without inner reference solution
[20] was employed. The software developed is able
to control all steps of the titration procedure with a
feedback structure in order to allow loading of sample
and titrant solutions into analytical path following
application of the binary search. The feasibility of

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

the procedure was demonstrated by the determination


of acidity in vinegar, coke, lemon soda, isotonic,
industrial and natural orange juice.
2. Theory
The potentiometric titration employing binary
search as proposed here can be considered as a
successive approximations strategy to locate the end

167

point by varying the volumetric fraction of the titrant


solution.
The signal generated by the electrode when only the
carrier solution is owing through the analytical path
is registered prior to the analytical cycle and stored as
a pre-set potential value (Eb, baseline measurement).
This value is adopted as a reference to decide about the
titration course to be followed after each run and also
to nd the titration end point. If the signal generated by
running an analytical cycle higher than Eb indicates an

Fig. 1. Schematic representation of the binary search titration: (a) sampling pattern, (b) hypothetical results. Eb average readings of the
baseline. Eb1Ebks and Eb2Ebks.

168

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

excess of sample solution, then the titrant volumetric


fraction must be increased. On the other hand, if a
signal lower than Eb indicates an excess of titrant
solution, then the titrant volumetric fraction must be
decreased.
The strategy followed in performing a hypothetical
titration employing binary search is represented in
Fig. 1. The aliquot volumes of sample and titrant
solutions (V1 and V2) are introduced into the system
as a time function, so that the volumetric fraction can
be related with the previously adjusted time intervals
T1 and T2. After sample and titrant aliquots introduction into the analytical path, the microcomputer
reads the difference of potential generated by the
potentiometer. The collected signal is processed in
order to decide either to increase or to decrease the
titrant volumetric fraction. As indicated in Fig. 1, the
titrant solution volume is varied up to the measured
signal was converged to the baseline value according
to Eq. (1):
Es Eb  ks;

(1)

where Es is the signal reading when performing a


titration run, Eb the baseline signal, k the arbitrary
constant, and s is the baseline standard deviation.
The titration procedure ends when the read signal is
within the range dened by Eq. (1), emphasizing that
the stoichiometric condition was attained.
As indicated in Fig. 1, equal volumes of the sample
and titrant solutions are inserted to perform the rst
run. If the stoichiometric condition is not attained, the
volume of titrant solution is halved and added or
subtracted from the last titrant aliquot according to
the following equations:
V0 V1 =2;

(2)

V2 V1 V0 ;

(3)

V20

(4)

V1 V0 ;

where V0 is the titrant volume variation, V1 is the


titrant volume of the rst run, V2 and V20 are the titrant
volume for the next run.
If the value is higher than the range dened by
Eq. (1), the concentration of the sample solution is
higher than the titrant, then the titrant volumetric
fraction must be increased as indicated by Eq. (3);
if lower, the titrant solution must be diluted as indicated by Eq. (4). For the next run up to attaining the

stoichiometry condition, the variation of titrant volumetric fraction is calculated as indicated in the following equations:
Vn Vn1 =2;

(5)

Vn Vn1 Vn ;

(6)

Vn0

(7)

Vn1 Vn :

This strategy is followed until the condition established in Eq. (1) is attained. The volumetric fraction of
the sample solution is constant to avoid variations of
the sample matrix, therefore, when the titrant volumetric fraction was decreased (Eq. (4)), a diluent
solution aliquot with a equal volume to that of the
variation of titrant was inserted in order to maintain
the concentration and ionic strength of the sample
zone.
All paths followed by the hypothetical titration
based on the binary search process up to the fth trial
is shown in Fig. 1. The end point is between the third
and fourth trials and the end point was attained in the
fth trial.
3. Experimental
3.1. Apparatus
The ow manifold was built up by assembling a set
of three-way solenoid valves (161T031 NResearch),
mixing coils of PTFE tubing (0.8 and 1.6 mm i.d.) and
transmission lines of polyethylene tubing (0.8 mm
i.d.). Signal detection was carried out with a 2002
Crison potentiometer equipped with a double junction
electrode (Orion 900029 Ag/AgCl) as a reference
electrode. A home-made ow through selective electrode for hydrogen ion [20] was employed as working
electrode. The sensor unit was adapted to the ow
system by using a cell constructed as described elsewhere [21]. A 486 microcomputer equipped with an
electronic interface (Advantech PCL-711S), running
software written in QUICKBASIC 4.5, was used to control
all steps of the titration procedures. An Ismatec IPC-4
peristaltic pump was employed and variation of pumping rates was performed by using its serial port (RS232) coupled to the microcomputer. An electronic
interface [18] was used to match the current intensity
required to switch the solenoid valves.

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

169

Fig. 2. Flow diagram of the system for automatic titration. (a) Sampling pattern. V1V5 three-way solenoid valves, B coiled reactor
(30 cm, i.d. 1.6 mm), C carrier stream, S sample, T titrant, D diluent, X confluent point, ISE ion-selective electrode, REF
reference electrode, W waste, mV potentiometer, P peristaltic pump, flow rate2.0 ml min1. Dashed lines represent the flow paths
when the valves are switched on.

3.2. Reagents and solutions


All solutions were prepared with analytical grade
chemicals and distilled deionized water. A buffer
solution containing 5.7105 mol l1 sodium tetraborate plus 20105 mol l1 disodium hydrogen phosphate plus 13105 mol l1 sodium citrate was used
as carrier stream (C). Hydrochloric acid or sodium
hydroxide solution was added in order to obtain pH
values between 7 and 9. The ionic strength of this
solution was adjusted to about 0.1 mol l1 with NaCl.
Hydrochloric and acetic acid solutions (S) were
prepared from stock solutions (1.0 mol l1) by appropriate dilution. Sodium hydroxide solutions (T) free of
CO2 were prepared daily from a saturated sodium
hydroxide solution using water boiled for 5 min.
Before dilution, an aliquot of the sodium hydroxide
stock solution was ltered to remove the sodium
carbonate precipitated and the ionic strength was
adjusted to 0.1 mol l1 with NaCl. While experiments
were performed, argon was bubbled through the
sodium hydroxide solution to maintain it free of

CO2. As diluent (D) a 0.1 mol l1 NaCl solution


was used to maintain the ionic strength in the sample
bulk.
Vinegar, coke, lemon soda, isotonic, industrial and
natural orange juice samples were analyzed without
any previous preparation.
3.3. Flow diagram and procedure
The ow diagram of the system is shown in Fig. 2,
where the solution loading step is performed by
synchronizing its start with pumping pulsation. Thus,
when the software was run, the microcomputer reads
through the PCL-771S analog input the tachometer
signal from the peristaltic pump. The ow system was
controlled by means of the microcomputer running the
software shown in the owchart of Fig. 3. The software also was able to decide about the best course
followed in the titration close to the end point. The
acquired data were treated in real time to decide
straightforwardly about the pathway of the next run
to be executed.

170

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

Fig. 3. Software flowchart.

In the conguration of Fig. 2, all valves are off and


only the carrier stream is owing through the analytical path. Afterwards, by switching of the valves V1
and V5 at the same time, an argon volume
(VVzt1, Vz, ow rate) is introduced in the analytical path. The aliquots of sample and titrant solutions were inserted in tandem by simultaneously
switching on valves V1 and V2 followed by V1 and
V3. This cycle should be repeated several times in
order to attain the sample volume appropriate to per-

form the titration. Once the solution loading steps are


nished, valves V1 and V5 are switched on again
during a time interval t1 to insert another argon
bubble. Under this condition, the analytical path is
loaded with a sample string presenting a pattern
analogous to that in Fig. 2(a). By switching all valves
off, the carrier stream ows again to transport the
sample string towards the detector. Solutions mixing
and reaction take place in the reaction coil (B). The
signals generated by the sensor are read by the micro-

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

computer through the analog input of the PCL-711S


interface, which was coupled to the potentiometer
analog output.
Before inserting the sample string to begin the next
run, the microcomputer reads the signal corresponding
to the carrier stream (baseline signal) that is used as a
reference to nd the titration end point.
Length of the sample string and sample volume
were constant in the analytical path. Then, during the
searching procedure to nd the end point, the slug of
the titrant solution was decreased and a slug of diluent
solution was inserted between slugs of the sample and
titrant solutions to maintain the string length. The
diluent solution slug was introduced by switching
valves V1 and V4 on.
The analytical path length, sampling string length,
sample aliquot volume and gas bubble volume were
studied in order to establish better conditions to carry
out the titration. The volume of the sampling string
was varied from 32 to 640 ml (50% sample solution)
by repeating the loading step from 1 to 20. Gas bubble
volumes of 10, 20, and 30 ml were inserted by sandwiching the sampling string solution. In this sense, the
analytical path length of 30 and 120 cm with inner
diameter of 1.6 and 0.8 mm, respectively, were
employed.
The feasibility of the procedure was veried by
titration of HCl and H3COOH with NaOH, and by
acidity determination in vinegar, coke, lemon soda,
isotonic, industrial and natural orange juice.
4. Results and discussion
Selection of the volumetric fraction was based on
the molar ratio method, which has been used for
determination complex stoichiometry [22]. In this
work, it was implemented by binary search as depicted
in Fig. 1. The analytical signal should attain a steady
state condition in order to improve the decision about
the titration course. In this sense, the ow network was
designed by exploiting the monosegmented approach
owing to its provided facilities to achieved this condition more easier than that based on a usual ow
system.
Titrant and titrand solutions aliquots were inserted
into the analytical path by employing the binary
sampling process, thus the reaction coil was loaded

171

Fig. 4. Typical recorded tracings of a binary search titration.


Dashed lines (a) indicate the end signal of the first gas bubble and
dashed lines (b) indicates the initial signal of the second gas
bubble. Curves 1, 2 and 3 correspond to titrant solution slugs of 32,
16 and 24 ml, respectively. Eb baseline signal. Eb1Ebks and
Eb2Ebks. Sample slugs32 ml. Titrant concentration8.74102
mol l1 NaOH. Flow rate2.0 ml min1.

with a string formed by slugs of titrant in tandem with


slugs of titrand. Under this condition, the mixing
between solutions could be affected by the reaction
coil geometry. Thus, experiments were carried out
utilizing two reaction coils with equal volume (600 ml)
presenting inner diameters of 0.8 and 1.6 mm. In this
way, by maintaining the string volume at 512 ml and
the slugs volume at 32 ml, only the inner diameter
could affect the mixing condition. The plateau pattern
without undulation shown in Fig. 4 indicates that
steady state situation and good mixing conditions
were attained. Nevertheless, utilizing the 0.8 mm
i.d. tubing a recorded signal with undulation was
observed, and to obtain similar results the solution
slug volumes should be decreased to 4 ml. A low initial
volume could affect the nal volume variation, thus
making the end point location more difcult. Moreover, gas bubble segmentation occurred frequently
affecting the precision of the measurements. With
the 1.6 mm i.d reaction coil, this effect was not
observed.
The volumes of the two gas bubbles were xed at
33 ml and the sampling string was inserted between
them. If the volume was smaller than this value,
bubbles segmentation occurred affecting the signal
stability. The signals were read without bubble
removal from the analytical path, and as a conse-

172

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

quence, the generated signals exhibited the uctuations pointed out in Fig. 4 caused by the gas bubbles
when they passed through the electrode.
The above experiments were carried out by inserting a sample string with a volume of 512 ml. Further
experiments showed that a string with a volume of
256 ml yielded a signal of similar magnitude (95%),
hence, this volume was employed further.
The software was designed to read the signal as a
function of time and to display it on the video screen to
allow a real time view of the titration. At the same
time, a data string of 120 measurements (average of 10
analog/digital converter readings) were generated.
Afterwards, the measurements related to the gas bubble were eliminated, and the data were processed to
decide about the course of the titration. Also, these
data were saved in a le to allow further analysis.
A baseline value was used as a reference to nd the
end point of titration. Therefore, the carrier stream
solution was passed through the analytical path and
200 measurements of the potentiometer analog output
were collected before the beginning of each trial.
Afterwards, the mean and standard deviations were
calculated. If the r.s.d. was lower than 2%, baseline
and r.s.d. values were stored, otherwise the measurements were repeated.
As indicated in the model of the binary search
titration (Fig. 1), on increasing the number of trials,
the measurements tend to the baseline value. The end
point titration was found when Eq. (1) was satised.
The coefcient k dened as an arbitrary constant was
set as 5. The pattern in Fig. 4 indicates a behavior
similar to that proposed in the model, so that for

additional trials, the measurements tend to the baseline value. In this gure, baseline value and standard
deviation were 86.7 and 0.9 mV, respectively. In this
sense, the titration end point was reached when the
measurement (Es) was comprised in the 82.291.2 mV
range. In Fig. 4, only three trials of the titration are
shown corresponding to titrant solution slugs volume
32, 16 and 24 ml. The potential difference of curve 3
was larger than 91.2 mV, thus for the next trial the
titrant slug volume should be 28 ml. In this situation,
the volume variation was 4 ml and for the next trial it
should be 2 ml. This value must be added if the
measurement was higher than 91.2 mV or subtracted
if the measurement was lower than 82.2 mV, as indicated by Eqs. (6) and (7). This strategy of slug variation was continued to locate the end point or till it
attained the smallest variation that was pre-set as
0.3 ml (loading time variation of 0.01 s).
After the operational conditions were established,
the feasibility of the approach was assessed by analyzing samples of vinegar, coke, lemon soda, isotonic,
industrial and natural orange juice, yielding the results
in Table 1. Accuracy was checked by applying the
paired t-test with those data obtained by a conventional titrimetric procedure [22], and no signicant
differences at the 95% condence level were
observed. A relative standard deviation of 1%
(n9) was achieved.
The time elapsed for performing one trial was 30 s,
therefore, the overall time to carry out the entire
titration depends on the required number of trials.
For the analyzed sample (Table 1), the minimum
number of trials was 5 and the maximum 8.

Table 1
Mean values and uncertainties (n3) for acidity determination in vinegar, coke, lemon soda, isotonic, industrial and natural orange juice
samples
Sample

FIA (102 mol l1)

Reference methoda (102 mol l1)

White vinegar
White vinegar
Color vinegar
Color vinegar
Coke
Lemon soda
Isotonic
Orange juice (natural)
Orange juice (industrial)

7.250.03
7.210.02
7.580.01
7.290.02
0.4630.03
2.640.04
3.550.01
9.080.01
8.880.01

7.430.03
7.220.01
7.580.02
7.280.05
0.4690.07
2.580.07
4.090.01
9.080.03
8.760.04

Conventional potentiometric titration with glass electrode.

P.B. Martelli et al. / Analytica Chimica Acta 387 (1999) 165173

5. Conclusions
The proposed system was able to perform all steps
required for the titration without any operator assistance. The combination of the binary search concept
and the monosegmented ow approach provided possibilities to locate the titration end point with good
precision. The acidity of the samples was determined
without using analytical curves. The system can work
without removing the gas bubbles. Their passage
through the electrode cause signal uctuations which
do no affect the analytical signal.
Acknowledgements
The authors are grateful to E.A.G. Zagatto for
critical comments. FAPESP, FINEP/PRONEX, CAPES
and CNPq are thanked for nancial support.
References
[1] H. Zeigel, Z. Anal. Chem. 53 (1914) 755.
[2] W.J. Blaedel, R.H. Laessig, Anal. Chem. 36 (1964) 1617.
[3] W.J. Blaedel, R.H. Laessig, Anal. Chem. 37 (1965) 332.

173

[4] J.M. Calatayud, P.C. Falco, M. Albert, Analyst 112 (1987)


1063.
[5] B. Fleet, A.Y.W. Ho, Anal. Chem. 46 (1974) 9.
[6] G. Nagy, Z. Feher, K. Toth, E. Pungor, Anal. Chim. Acta 91
(1977) 97.
[7] J. Ruzicka, E.H. Hansen, Anal. Chim. Acta 78 (1975) 145.
[8] J. Ruzicka, E.H. Hansen, H. Mosbaek, Anal. Chim. Acta 98
(1977) 235.
[9] J.F. Tyson, Analyst 112 (1987) 523.
[10] J.F. van Staden, H. Plessis, Anal. Commun. 34 (1997) 147.
[11] G.D. Clark, J. Zable, J. Ruzicka, G.D. Christian, Talanta 38
(1991) 119.
[12] H.M.N. Irving, H. Freiser, T.S. West, Pure Appl. Chem. 18
(1969) 427.
[13] H.L. Pardue, B. Fields, Anal. Chim. Acta 124 (1981) 39.
[14] J. Marcos, A. Rios, M. Valcarcel, Anal. Chim. Acta 261
(1992) 489.
[15] J. Marcos, A. Rios, M. Valcarcel, Anal. Chim. Acta 261
(1992) 495.
[16] C.H. Yarnitzky, N. Klein, O. Cohen, Talanta 40 (1993) 1937.
[17] J. Bartroli, L. Alerm, Anal. Lett. 28 (1995) 1483.
[18] M. Korn, L.F.B.P. Gouveia, E. Oliveira, B.F. Reis, Anal.
Chim. Acta 313 (1995) 177.
[19] C. Pasquini, W.A. Oliveira, Anal. Chem. 57 (1985) 2575.
[20] P.B. Martelli, B.F. Reis, E.A.G. Zagatto, J.L.F.C. Lima, R.A.
Lapa, Quim. Nova 21 (1998) 133.
[21] S. Alegret, J. Alonso, J. Bartroli, A.A.S.C. Machado, J.L.F.C.
Lima, J.M. Paulis, Quim. Anal. 6 (1987) 278.
[22] D.A. Skoog, D.M. West, F.J. Holler, Fundamentals of
Analytical Chemistry, 7th ed., Orlando, 1996.

You might also like