Simultaneous Determination of Sulfite and Sulfide Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6 863
Simultaneous Kinetic Spectrophotometric Determination of Sulfite and Sulfide
Using Partial Least Squares (PLS) Regression
Abbas Afkhami,* Nahid Sarlak, Ali Reza Zarei,† and Tayyebeh Madrakian
Department of Chemistry, Faculty of Science, Bu- Ali Sina University, Hamadan, Iran. *E-mail: firstname.lastname@example.org
†Faculty of Materials, Malek Ashtar University of Technology, Tehran, Iran
Received January 12, 2006
The partial least squares (PLS-1) calibration model based on spectrophotometric measurement, for the
simultaneous determination of sulfite and sulfide is described. This method is based on the difference between
the rate of the reaction of sulfide and sulfite with Malachite Green in pH 7.0 buffer solution and at 25 οC. The
absorption kinetic profiles of the solutions were monitored by measuring the decrease in the absorbance of
Malachite Green at 617 nm in the time range 10-180 s after initiation of the reactions with 2 s intervals. The
experimental calibration matrix for partial least squares (PLS-1) calibration was designed with 24 samples. The
cross-validation method was used for selecting the number of factors. The results showed that simultaneous
determination could be performed in the range 0.030-1.5 and 0.030-1.2 μg mL−1 for sulfite and sulfide,
respectively. The proposed method was successfully applied to simultaneous determination of sulfite and
sulfide in water samples and whole human blood.
Key Words : Partial least squares (PLS), Sulfite, Sulfide, Spectrophotometric determination, Kinetic methods
Interest in UV-Vis spectrophotometric methods has in-
creased and been renewed through the use of signal
processing and multivariate calibration,1 partial least squares
(PLS) regression2,3 and artificial neural network (ANN).4
Multivariate calibration methods have an increased im-
portance in multicomponent analysis, specially using PLS
method with decomposition into latent variables.5 The
partial least squares (PLS) regression was successfully used
in spectrophotometry6-8 and near-infrared spectrometry.9
Multicomponent kinetic determinations when associated
with different chemometrics methods such as PLS and ANN
can resolve multicomponent kinetic systems by using dif-
ferences of behavior with respect to a common reagent10-12
without requiring prior separation.
PLS calibration of a multicomponent system can be
performed in two different ways, PLS1 and PLS2. The use
of PLS2 has a few advantages. Firstly there is one common
set of PLS factors for all analytes. This simplifies the
procedure and interpretation and enables a simultaneous
graphical inspection. Secondly, when the analyte concentrations
are strongly correlated one may expect that the PLS2 model
is more robust than separate PLS1 models. Finally, when the
number of analytes is large the development of a single
PLS2 model is done much quicker than development of
many separate PLS1 models. Practical experience, however,
indicates that PLS1 calibration usually performs equally
well or better in terms of predictive accuracy. Thus, when
the ultimate requirement of the calibration study is to enable
the best possible prediction, a separate PLS1 regression for
each analyte is advised. In the present work PLS1 models
were used for determination of analytes.13
Sulfite is widely used as additive in food and beverages
to prevent oxidation and bacterial growth and to control
enzymatic reactions during production and storage. Sulfite is
also known to present some cytotoxic, mutagenic and
antinutritional effects.14 In particularly, it interacts with some
vitamins, i.e. pyridoxal, nicotinamide, thiamine, folic acid,
reducing the nutritional quality of treated food.15
Sulfide is formed in waste water by action of anaerobic
bacteria on organic matter. Reduced sulfur compounds, such
as hydrogen sulfide are found in natural and waste waters.
From the environmental point of view, hydrogen sulfide is
one of the most important parameters to monitor in water
matrices due to its high toxicity for aquatic organisms. Also,
hydrogen sulfide controls the bioavailability of heavy metals
in anoxic environments due to the low solubility of sulfide
The determination of sulfite and sulfide in biological and
industrial samples is important. Different methods have been
reported for determination of sulfite or sulfide. These include
kinetic spectrophotometric methods,17-20 chromatography21,22
and electrochemical methods.23-25 Ghasemi and Mohammadi
applied univariate and multivariate calibration method for
the determination of sulfite26 and sulfide27 based on their
addition reaction with new fuchsin.
Several methods have been reported for simultaneous
determination of sulfite and sulfide. These include, gas phase
molecular absorption spectrometry,28 chromatographic sepa-
ration29 and fluorometric flow-injection.30 Recently ANNs
were employed for the simultaneous determination of sulfite
and sulfide.31 A simultaneous kinetic resolution of binary
mixtures of cyanide, sulfide, and sulfite by reaction with 5,5-
dithiobis(2-nitrobenzoic acid) (DTNB) in aqueous cetyltri-
methylammonium bromide (CTAB) micelles was developed.32
In this paper, we describe a rapid, simple, precise and
accurate method for the simultaneous determination of
864 Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6Abbas Afkhami et al.
sulfite and sulfide using the partial least squares (PLS-1)
regression. The method is based on the difference between
the rate of their reactions with Malachite Green in pH 7.0
buffer solution and at 25 οC. The absorption kinetic profiles
of the solutions were monitored by measuring the decrease
in the absorbance of Malachite Green at 617 nm in the time
range 10-180 s after initiation of the reaction with 2 s
Apparatus. A Pharmacia Model LKB3 UV-visible
Ultraspect(III) single beam spectrophotometer that con-
nected to a Pentium II computer with 1-cm quartz cells was
used for recording the kinetic profiles. A Jenway C15 pH-
meter was used to adjust the pH of the buffered solutions.
The computations were made with a Pentium 4 computer.
The PLS calculations were performed with the PLS_
Toolbox for MATLAB version 3.5.
Reagents. Triply distilled water and analytical reagent
grade chemicals were used. A 1000 μg mL−1 standard
solutions of sulfide and sulfite were prepared daily by
dissolving 0.7500 g of Na2S·9H2O (Merck) and 0.1574 g of
Na2SO3 (Merck) in water and diluting to the mark in a 100
mL volumetric flask. These solutions were standardized by
iodometric titration.25 Working solutions were prepared by
diluting the standard solutions with water. A 2.20 × 10−4 M
Malachite Green solution was prepared by dissolving 0.1023
g of Malachite Green (Merck) in water and diluting to 1000
mL with water. A 0.1 M phosphate buffer solution of pH 7.0
was prepared and its pH was checked by the pH meter.
Procedure. All the solutions were kept in a thermostated
water bath at 25 ± 0.1 oC before beginning the reactions.
A 30 mL of 2.20 × 10−4 M Malachite Green and 30 mL
phosphate buffer solution of pH 7.0 were added into a 100
mL volumetric flask and the solution was diluted to the mark
with water. This solution was prepared daily. Then 1 mL of
this solution was added into a 1-cm quartz cell containing 2
mL of sulfite and/or sulfide solution. The absorption kinetic
profiles of the solutions contain of sulfite and sulfide with
different concentrations were recorded at 617 nm in the time
range 10-180 s with 2 s intervals after initiation of the
Results and Discussion
Preliminary Investigations. In phosphate buffer solution
of pH 7.0 and at 25 oC, sulfite and sulfide react with
Malachite Green and decolorize it.
The reactions can be monitored spectrophotometrically by
measuring the decrease in the absorbance of the solution at
It was observed that under the similar conditions the rate
of the reaction of hydrogen sulfite and hydrogen sulfide with
Malachite Green are different (see Figure 1). Therefore, the
system seems to be appropriate for simultaneous determina-
tion of sulfite and sulfide by spectrophotometric method
using the partial least squares (PLS) calibration.
Effect of Variables. The effect of pH on the rate of the
reactions of a mixture of sulfite and sulfide was studied in
the range 4-12. The results are shown in Figure 2. As Figure
2 shows, the absorbance change increased by increasing pH
up to 7.0 and decrease at higher pHs. Therefore, pH 7.0 was
selected as the optimum pH.
(X = HSO3− or HS−)
Figure 1. Kinetic profiles for Malachite Green in the reaction with
(a) sulfite, (b) sulfide and (c) their mixture. Conditions: sulfite, 0.20
μg mL−1; sulfide, 0.20 μg mL−1; Malachite Green, 22 μM; pH = 7.0
and t = 25 οC.
Figure 2. Effect of pH on the rate of the reaction of (■) sulfite and
sulfide with Malachite Green, (▲) blank reaction and (●) their
difference. Conditions: sulfite, 0.20 μg mL−1; sulfide, 0.20 μg mL−1;
Malachite Green, 22 μM and t = 25 οC.
Simultaneous Determination of Sulfite and Sulfide Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6 865
The effect of Malachite Green concentration on the rate of
the reactions of a mixture of sulfite and sulfide was studied
in the range 2.2-40 μM. The results (Figure 3) show that the
absorbance change increased by increasing concentration of
Malachite Green up to 12 μM and remained constant at
higher concentrations. As Therefore, a concentration of 22
μM of Malachite Green was selected as the optimum
The effect of temperature on the rate of reaction of mixture
of sulfite and sulfide was studied in the range 10-40 oC. As
Figure 4 shows, the absorbance change increased by
increasing temperature up to 25 oC and decrease at higher
temperatures. Therefore, a temperature of 25
selected as the optimum temperature.
Univariate Calibration. Under the optimum conditions
calibration graphs for sulfide and sulfite were constructed by
plotting absorbance change values during 10-180 s after
initiation of the reactions as a function of the analyte con-
centration. The calibration graphs were linear in the range of
0.03-1.50 and 0.03-1.20 μg mL−1 for sulfite and sulfide,
respectively. The results are shown in Table 1.
Multivariate Calibration and Prediction Data Set.
Multivariate calibration consists of the establishment of a
relationship between matrices of chemical data. The methods
are based on a first calibration step in which a mathematical
model is built using a chemical data set (e.g. absorbance
values) and a concentration matrix data set. The calibration
is followed by a prediction step in which this model is used
to estimate unknown concentrations of a mixture from its
In particular, PCR and PLS techniques are called “factor
methods” because transform the high number of original
variables in to a smaller number of orthogonal variables
called “factors” or “principal components”, which are linear
Figure 3. Effect of Malachite Green concentration on the rate of
the reaction of (■) sulfite and sulfide with Malachite Green, (▲)
blank reaction and (●) their difference. Conditions: sulfite, 0.20
μg mL−1; sulfide, 0.20 μg mL−1; pH = 7.0 and t = 25 οC.
Figure 4. Effect of temperature on the rate of the reaction of (■)
sulfite and sulfide with Malachite Green, (▲) blank reaction and
(●) their difference. Conditions: sulfite, 0.20 μg mL−1; sulfide,
0.20 μg mL−1; Malachite Green, μM and pH = 7.0.
Table 1. Characteristics of calibration graphs for the determination of sulfite and sulfide
Intercept Correlation coefficientLinear range/μg mL−1
Limit of Detection/μg mL−1
Table 2. Concentration data for the different mixtures used in the
calibration set and prediction set for the determination of sulfite
Calibration set/μg mL−1
Prediction set/μg mL−1
866 Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6Abbas Afkhami et al.
combinations of the original variables. The first factors
contain useful information, whereas the last ones represent
the noise, which has to be discarded and not considered in
Experimental design of the calibration set multivariate
calibration methods such as PCR and PLS require a suitable
experimental design of defining the calibration set. The
calibration procedure in complete experimental design was
selected with six concentration levels for both sulfite and
sulfide. A synthetic set of 33 solutions of mixtures of sulfite
and sulfide were prepared (Table 2). From the series, 24
solutions were chosen for the calibration set and 9 solutions
were used as prediction solutions.
Procedure for Selecting the Optimal Number of
Factors. The selection of the optimal number of factors
(latent variables) used to build PLS models for represents
a decisive step to improve the prediction power of the
methods. A full cross-validation, also called leave one-out
cross-validation, was employed towards this aim. It consists
of removing one sample at a time from the calibration step
and performing the calibration with all other samples. The
concentration of the sample removed is then predicted with
the obtained model. This step is in turn repeated for each
sample considered. The procedure can be repeated after
fixing a different number of factors. The prediction error
was calculated for each ion for the prediction set, which are
the samples not participating in the construction of the
model. This error was expressed as the prediction residual
error sum of squares (PRESS). PRESS was calculated for
the first variable, which built the PLS-1 modeling in the
calibration step, then, another latent variable was added for
the model building and the PRESS was calculated again.
This process was repeated for one to 9 latent variables,
which were used in the PLS-1 modeling. This procedure was
repeated for each element. Figure 5 shows the plot of
PRESS against the number of factors for each individual
component. One reasonable choice for the optimum number
of factor would be the number which yield the minimum
PRESS. The F-statistical test was used to determine the
significance of PRESS values greater than the minimum.
The optimal number of factors for sulfide and sulfite was
obtained 3. The results obtained by applying PLS1 algorithm
to the prediction set are given in Table 3.
As Figures 2-4 show the change in absorbance strongly
depends on pH, Malachite Green concentration and tem-
perature. Therefore they are most effective factors in this
Statistical Parameters. The statistical treatment of this
study is basically the same as that of Ghasemi and
Mohammadi.27 For the optimized model four parameters
were selected, as some type of figures of merite, to evaluate
the prediction ability of the model for determination of
sulfide and sulfite in the prediction set. The root mean
square difference (RMSD), which is an indication of the
average error in the analysis, for each component:
The other parameter was relative error of prediction (REP)
that shows the predictive ability of each component:
The prediction error of a single component in the mixture
was calculated as the relative standard error (R.S.E) of the
The total prediction error of N samples is calculated as
where N is the number of samples, Cj and
concentration of the component in the jth mixture and the
RMSD = 1 n
() = 100/C 1 n
Figure 5. Plot of PRESS against the number of factors for (■)
sulfite and (▼) sulfide.
Table 3. Composition of synthetic samples, their predictions by
PLS-1 model and statistical parameters for system
Composition (μg mL−1)Prediction (μg mL−1)Recovery (%)
Simultaneous Determination of Sulfite and Sulfide Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6 867
estimated concentration, respectively, M is the number of
components, Cij is the concentration of the ith component in
the jth sample and is its estimation. The values of
statistical parameters calculated in optimum number of
factors for sulfide and sulfite in the prediction set are
summarized in Table 4.
Selectivity. To study the selectivity of the proposed
method, the effect of various ions on the determination of a
mixture of 0.50 μg mL−1 each of sulfite and sulfide was
tested under the optimum conditions. The tolerance limit
was defined as the concentration of added ion causing less
than ±3 relative error. The results are given in Table 5. As
Table 4 shows, most of the cations and anions did not
interfere on the simultaneous detrermiation of sulfite and
sulfide by the proposed method even when present in 200- to
1000-fold excess over sulfite and sulfide. Therefore the
method shows a good selectivity for the determination of
sulfide and sulfite in mixture.
Application. To evaluate the analytical applicability of the
proposed method, it was applied to the simultaneous
determination of sulfite and sulfide in water samples and in
whole human blood.
In order to separate sulfide and sulfite content of the whole
blood a gas-phase separation apparatus was used (Figure
1).35 The traps consist of two glass tubes, one for the test
sample, which has an injection port into which acid is
injected to release hydrogen sulfide and SO2. The other tube
is used to trap the hydrogen sulfide and SO2 as anionic
sulfide and SO32− in 0.1 mol L−1 of sodium hydroxide
solution. The two tubes are joined by a head unit, which
enables H2S and SO2 gases evolved to be carried into the
trapping solution. The design also allows the introduction of
inert carrier gas (N2) directly into the sample tube.
A 5 ml of 15 mol L−1 sulfuric acid solution was injected in
reaction tube that contains 15 mL of human blood. The
produced hydrogen sulfide and SO2 was carried by the
nitrogen flow from reaction tube into the trapping tube
containing 5 mL of the trapping solution (0.1 mol L−1
NaOH). The hydrogen sulfide and SO2 were quantitatively
collected in the NaOH absorber. The pH of the solution was
adjusted to about 7 by 0.2 M HNO3 its sulfite and sulfide
concentration was determined by the proposed method. The
results are given in Table 6. The results show that the PLS-1
model is able to predict the simultaneous determination of
sulfite and sulfide concentrations in such samples.
Conclusion. The above results show that PLS-1 is an
excellent calibration method to simultaneous determination
of sulfite and sulfide, based on the different in their reaction
rates with Malachite Green. The partial least squares is a
Table 4. Statistical parameters of the optimized matrix using the
aNumber of principle components.
Table 5. Tolerance limits of diverse ions on the determination of a
mixture of 0.50 μg mL−1 each of sulfite and sulfide
IonsTolerace limit/μg mL−1
Cl−, NO2−, NO3−, CO32−, HCO3− ClO3−, SO42−,
S2O82−, F−, PO43− , Br−-, BrO3−, I−, ClO4−,
SCN−, CH3COO−, Pb2+, Fe3+, As3+, V3+,
NH4+, Na+, K+, Mg2+, Co2+
Cr3+, Mn2+, Ni2+, VO2+, Ag+
Figure 6. Gas-phase separation apparatus. It consists of reaction
tube, trapping tube, gas dispersers and injection port.
Table 6. Simultaneous determination of sulfite and sulfide in water
Sulfite (μg mL−1)Sulfide (μg mL−1)
0.28 ± 0.07
0.41 ± 0.05
0.59 ± 0.03
0.81 ± 0.04
0.65 ± 0.02
0.55 ± 0.06
0.96 ± 0.07
0.75 ± 0.06
1.10 ± 0.03
0.10 ± 0.02
0.56 ± 0.06
01.1 ± 0.05
0.12 ± 0.05
0.26 ± 0.07
0.75 ± 0.03
0.84 ± 0.02
0.11 ± 0.04
0.24 ± 0.05
0.18 ± 0.06
0.48 ± 0.08
0.19 ± 0.05
0.66 ± 0.03
0.83 ± 0.02
0.48 ± 0.07
0.85 ± 0.04
1.83 ± 0.05
1.56 ± 0.02
0.90 ± 0.04
1.95 ± 0.06
1.31 ± 0.01
1.65 ± 0.05
1.04 ± 0.06
aMean ± standard deviation for three determinations. bNot detected.
868 Bull. Korean Chem. Soc. 2006, Vol. 27, No. 6Abbas Afkhami et al. Download full-text
powerful tool for the simultaneous determination of the
analytes. The results in Table 6 show that PLS-1 can
appropriately model multicomponent systems and predict
unknown analyte concentrations with satisfactory results.
1. Lobinski, R.; Marczenko, Z. Crit. Rev. Anal. Chem. 1992, 23,
2. Martens, H.; Naes, T. Multivariate Calibration; Wiley: Chichester,
3. Geladi, P.; Kowalski, B. R. Anal. Chim. Acta 1986, 185, 1.
4. Zupan, J.; Gasteiger, J. Neural Networks for Chemists; VCH:
5. Wold, H. In Sytstem under Indirect Observation: Causality-
Structure-Prediction, Vol II; Joreskoq, K. G.; Wold, H., Eds.;
North Holland Publishing: Amsterdam, 1982.
6. Otto, M.; Wegscheider, W. Anal. Chem. 1985, 57, 63.
7. Afkhami, A.; Bahram, M.; Zarei, A. R. Michrochim. Acta 2004,
8. Madrakian, T.; Afkhami, A.; Borazjani, M.; Bahram, M. Spectro-
chim Acta Part A 2005, 61, 2988.
9. Rambla, F. J.; Garrigues, S.; De La Guardia, M. Anal. Chim. Acta
1997, 344, 41.
10. Havel, J.; Jiménez, J.; Bautista, R. D.; Arias León, J. J. Analyst
1993, 118, 1355.
11. Pérez-Bendito, D.; Silva, M. Kinetic Methods in Analytical
Chemistry; Ellis Horwood: Chichester, 1988.
12. Quencer, B. M.; Crouch, S. R. Crit. Rev. Anal. Chem. 1993, 24,
13. Massart, D. L.; Vandeginste, B. G. M.; Buydens, L. M. C.; Jong,
S. D. E.; Lewi, P. J.; Smeyers-Verbeke, J. Handbook of Chemo-
metrics and Qualimetrics. Part A; Elsevier Science: 1997.
14. Stammati, A.; Zanetti, C.; Pizzoferrato, L.; Quattrucci, E.;
Tranquilli, G. B. Food Addit. Contam. 1992, 9, 511.
15. Pizzoferrato, L.; Quattrucci, E.; Di Lullo, G.; In Nutritional and
Toxicological Aspects of Food Processing; Walker, R.; Quattrucci,
E., Eds.; Taylor and Francis: London, 1988; p 93.
16. Förstner, Ü.; Salomons, W.; Stigliani W., Eds.; In Biogeodynamics
of Pollutants in Soils and Sediments; Springer: Germany, 1995; p
17. Safavi, A.; Ramezani, Z. Talanta 1997, 44, 1225.
18. Afkhami, A.; Rezaei, M. Can. J. Anal. Sci. Spect. 2000, 45, 9.
19. Mousavi, M. F.; Sarlak, N. Anal. Lett. 1997, 30, 1567.
20. Barzegar, M.; Jabbari, A.; Esmaeili, M. Bul. Kor. Chem. Soc.
2003, 24, 1261.
21. Tang, D.; Santschi, P. H. J. Chromatogr. A 2000, 883, 305.
22. Perfetti, G. A.; Diachenko, G. W. J. AOAC Int. 2003, 86, 544.
23. He, Y.; Zheng, Y.; Locke, D. C. Anal. Chim. Acta 2002, 459,
24. Hassan, S. S. M.; Marzouk, S. A. M.; Sayour, H. E. M. Anal.
Chim. Acta 2002, 466, 47.
25. Isaac, A.; Wain, A. J.; Compton, R. G.; Livingstone, C.; Davis, J.
Analyst 2005, 130, 1343.
26. Ghasemi, J.; Mohammadi, D. E. Anal. Lett. 2003, 36, 2243.
27. Ghasemi, J.; Mohammadi, D. E. Microchem. J. 2002, 71, 1.
28. Pinillos, S. C.; Vicente, I. S.; Asensio, J. S.; Bernal, J. G. Talanta
1995, 42, 937.
29. Steudel, R.; Holdt, G.; Goebel, T. J. Chromatogr. 1989, 475, 442.
30. Dasgupta, P. K.; Yang, H. C. Anal. Chem. 1986, 58, 2839.
31. Safavi, A.; Moradlou, O.; Maesum, S. Talanta 2004, 62, 51.
32. Gonzalez, V.; Moreno, B.; Sicilia, D.; Rubio, S.; Pérez-Bendito,
D. Anal. Chem. 1993, 65, 1897.
33. Williams, W. J. Handbook of Anion Determination; Butterworths:
34. Blanco, M.; Coello, J.; Ituriage, H.; Maspoch, S.; Redon, M.
Appl. Spectroc. 1994, 48, 37.
35. Richardson, C. J.; Magee, E. A. M.; Cummings, J. H. Clin. Chim.
Acta 2000, 293, 115.