Spectrochimica Acta Part A 61 (2005) 869877

Successive ratio-derivative spectra as a new spectrophotometric method

for the analysis of ternary mixtures
Abbas Afkhami , Morteza Bahram
Department of Chemistry, Faculty of Science, Bu-Ali Sina University, Hamadan, Iran
Received 30 March 2004; accepted 9 June 2004

Abstract
A new spectrophotometric method was developed for the simultaneous determination of ternary mixtures, without prior separation steps.
This method is called the successive derivative ratio spectra. The method is based on the successive derivative of ratio spectra in two steps.
The mathematical explanation of the procedure is illustrated. In order to evaluate the applicability of the method a model data as well as an
experimental data were tested. The results from experimental data relating to the simultaneous spectrophotometric determination of Zn(II),
Co(II) and Ni(II) based on their complexes with 1-(2-pyridylazo)2-naphthol in micellar media were presented as real model for resolution of
the ternary systems. The method was compared with partial least squares (PLS) method.
2004 Elsevier B.V. All rights reserved.
Keywords: Successive ratio-derivative spectra; Spectrophotometric determination; Ternary mixtures

1. Introduction
The main problem of spectrophotometric multicomponent analysis is the simultaneous determination of
two or more compounds in the same mixtures without
preliminary separation. Several spectrophotometric determination methods have been used for resolving mixtures of
compounds with overlapping spectra. When these methods
are compared with each other, the range of application of
derivative spectrophotometry is more reliable with respect
to utility and sensitivity than normal spectrophotometry
[1]. Other spectrophotometric methods, such as partial
least squares regression (PLSR) [2], principal component
regression (PCR) [3], multi-wavelength linear regression
analysis (MLRA) [4], H-point standard addition method
(HPSAM) for binary [5] and ternary [6] mixtures have been
utilized.
Salinas et al. [7] proposed a new spectrophotometric
method, ratio-derivative spectrophotometry, for the simultaneous determination of two compounds in binary mixtures.

Corresponding author. Tel.: +98 811 8272404; fax: +98 811 8272404.
E-mail address: afkhami@basu.ac.ir (A. Afkhami).

1386-1425/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.saa.2004.06.013

Their method is based on the derivative of the ratio spectra

for a binary mixture. The absorption spectrum of the mixture
is divided by the absorption spectrum of a standard solution
of one of the compounds and the first derivative of the ratio
spectrum is obtained. The concentration of active compounds
are then determined from the calibration graphs obtained by
measuring the amplitudes at points corresponding to the minimum or maximum wavelength.
Berzas Nevado et al. [8] developed a new method for the
analysis of ternary mixtures by derivative ratio spectra zerocrossing method. In this method, the simultaneous determination of three compounds in ternary mixtures is realized by
the measuring of the amplitude at the zero-crossing points
in the derivative spectrum of the ratio spectra. These two
methods were further studied and applied for simultaneous
determination of binary or ternary mixtures [913].
Recently, Dinc et al. [1,1416] proposed a new spectrophotometric method for the simultaneous determination of
ternary mixtures. This method was called the double divisorratio spectra derivative method. The method is based on the
use of the coincident spectra of the derivative of the ratio
spectra obtained by using a double divisor (sum of two
spectra) and the measuring at either the maximum or mini-

870

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

mum wavelengths. But their method cannot be popularized,

because it can only be used for the mixtures that the ratio
of the concentrations of two interfering compounds (used as
double divisor) is known. In the other words, the ratio of the
concentrations of two interfering compounds should be the
same in calibration, prediction and unknown samples. It is
obvious that the ratio of the concentration of the analytes in
real samples is always unknown.
In this work, we proposed a new method for simultaneous
determination of the three compounds in ternary mixtures
without need to know the ratio of concentration of species.
Our method is based on the successive derivative of ratio
spectra in two successive steps. The method was evaluated
by model data and also by application to the simultaneous
spectrophotometric determination of cobalt, nicked and zinc
based on their complexes with 1-(2-pyridylazo)2-naphthol in
micellar media as experimental data.

Eq. (5) is the mathematical foundation of multicomponent analysis that permits the determination of concentration of each of the active compounds in the solution
(X in this equation) without interference from the other compounds of the ternary system (Y and Z in these equations). As
Eq. (5) shows there is a linear relation between the amount
of dD/d and the concentration of X in the solution.
A calibration curve could be constructed by plotting dD/d
against concentration of X in the standard solutions of X
or in the standard ternary mixtures. For more sensitivity the
amount of dD/d corresponding to maximum or minimum
wavelength should be measured.
Calibration graphs for Y and Z could be also constructed
as described for X.

3. Modeling
3.1. Validation of the method with model data

2. Theoretical background
Consider a mixture of three compounds X, Y and Z. If
Beers law is obeyed in the whole wavelength range used and
by considering the path length as 1 cm, the absorbance of the
ternary mixture at each wavelength can be written as:
A m = X C X + Y C Y + Z C Z

(1)

where Am is the vector of the absorbance of the mixture, X ,

Y and Z are the absorptivity vectors of X, Y and Z and CX ,
CY and CZ are the concentrations of X, Y and Z, respectively.
If Eq. (1) is divided by Z corresponding to the spectrum
of a standard solution of Z in ternary mixture, the first ratio
spectrum is obtained in the form of Eq. (2) (for possibility of
dividing operation, the zero values of Z should not be used
in the divisor):
B=

Am
X C X
Y CY
=
+
+ CZ
Z
Z
Z

(2)

If the first derivative of Eq. (2) is taken, since the derivative

of a constant (CZ ) is zero, Eq. (3) would be obtained:






dB
d X CX
d Y CY
=
+
(3)
d
d
Z
d
Z
By dividing Eq. (3) by (d/d)(Y /Z ), corresponding to
the derivative of the ratio of the spectra of the standard solutions of Y and Z, the second ratio spectrum is obtained as
Eq. (4) (for possibility of dividing operation, the zero values
of (d/d)(Y /Z ) should not be used in the divisor):
D=

dB/d
(d/d)[X CX /Z ]
=
+ CY
(d/d)(Y /Z )
(d/d)(Y /Z )

To demonstrate the analytical applicability of the proposed

method for the analysis of ternary mixtures, three spectra
were created. The three curves form model of the overlapping spectra of three compounds X, Y and Z in the range
420600 nm (Fig. 1). The absorption spectra of different concentration of X were created (Fig. 2a) and divided by the spectrum of the Y and the ratio spectra were obtained (Fig. 2b).
First derivatives of the ratio spectra were obtained with 
= 1 nm (Fig. 2c). After that, these vectors (first derivatives of
the ratio spectra) are divided by (d/d)(Z /Y ) corresponding to the derivative of the ratio of the spectra of Z and Y
and therefore, second ratio spectra according to Eq. (4) were
obtained (Fig. 2d). First derivative of these vectors were obtained with  = 1 nm (Fig. 2e). The concentration of compound X was determined by measuring the amplitude of the
resulting spectra at 435 nm corresponding to the minimum
wavelength shown in Fig. 2f.
The obtained model was validated with a number of 10
synthetic mixture set containing the considered components
of X, Y and Z in different proportions that were randomly
selected. The model spectra of these mixtures are shown
in Fig. 3a. To predict the concentration of compound X in
these synthetic mixtures the procedure that was previously
explained for calibration set was performed. Finally, sec-

(4)

If the first derivative of Eq. (4) is taken since the derivative

of a constant (CY ) is zero, Eq. (5) would be obtained:



dD
d (d/d)(X CX /Z )
=
(5)
d
d
(d/d)(Y /Z )

Fig. 1. The modeled zero-order spectra of compounds X, Y and Z.

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

871

Fig. 2. The modeled absorption spectra of different concentrations of X (a) and the ratio spectra were obtained by dividing by the spectrum of the Y (b). First
derivatives of the ratio spectra were obtained with  = 1 nm (c), second ratio spectra were obtained according to Eq. (4) (d). First derivative of these vectors
were obtained with  = 1 nm (e).

ond ratio spectra for these mixtures were obtained (Fig. 3b).
As can be seen from Fig. 3b the effect of the compounds
Y and Z was removed from mixtures spectra. The concentration of X in these synthetic mixtures was determined by
measuring the amplitude at the 435 nm corresponding to a
minimum wavelength. The results are shown in Table 1.
The amounts of Y and Z were also determined as described
for X.

of ternary mixtures was predicted. For this set of data, there

was a good agreement between synthesized and predicted
concentrations (Table 1) that shows the applicability of the
method for noisy systems.

4. Experimental
4.1. Apparatus

3.2. Effect of random noise

Random noise was added to the generated set of artificial
data in order to test the method more rigorously. In this step,
after creation of the spectra of the ternary mixtures, random
noise of 0.002 and 0.003 absorbance unit was added to
each spectrum and then the procedure was followed as explained above and the concentration of X in the noisy signal

UVvis absorbance spectra were recorded on a PerkinElmer Lambda 45 UVvis spectrometer using quartz cells
and slit width of 0.5 nm. A Metrohm model 713 pH-meter
with a combined glass electrode was used for pH measurements. All calculations in the computing process were done
in Matlab6.5 and Microsoft Excell for windows. A simple
program was written for this purpose in Matlab6.5.

872

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

4.3. Procedure

Fig. 3. The model spectra of randomly selected mixtures of the X, Y and Z

(a) and the first derivative of their second ratio spectra obtained with  =
1 nm (b) according to Eq. (5).

4.2. Reagents
All chemicals used were of analytical reagent grade and
triply distilled water was used throughout the experiments.
Stock Zn(II), Co(II) and Ni(II) solutions (1000 mg L1 )
were prepared by dissolving appropriate amounts of
Zn(NO3 )6 6H2 O, Co(NO3 )6 6H2 O and NiSO4 6H2 O (all
from Merck), respectively, in water. A 5.0 103 mol L1
PAN solution was prepared by dissolving 128 mg of PAN
(in 100 mL of hot ethanol. Buffer solution of pH 9.2 was
prepared from 0.5 mol L1 ammonia and 0.5 mol L1 ammonium chloride. Triton X-100 stock solution (14%, v/v)
was prepared by dissolving 14 mL of concentrate solution
(Merck) in hot distilled water.

A calibration graph for X is obtained by recording and

storing the spectra of solutions of different concentrations of
standard solutions of X, Y and Z and also by using the vector
of (d/d)(Y /Z ) corresponding to the derivative of the ratio
of the spectrum of the Y and Z compounds. The stored spectra of the solutions of X are divided by standard spectrum
of Z according to Eq. (2). Then the first derivative of these
vectors with respect to wavelength are obtained according to
Eq. (3). After that residual vector is divided by (d/d)(Y /Z )
according to Eq. (4). The minimum or maximum of the first
derivative of later vector with respect to wavelength is used
for the construction of calibration graph for X. For the prediction of concentration of X in synthetic mixtures and real
samples the same procedure was used except that the spectra
of the mixture were used instead of the spectra of standard
solution of X.
The construction of calibration curves for other two active
compounds and also their prediction step was performed as
described for X.
The proposed system was used to the simultaneous determination of Co(II), Ni(II) and Zn(II) based on their complexes with PAN in micellar media. We considered Co(II),
Ni(II) and Zn(II) as X, Y and Z, respectively and construction of the calibration curves and predictions were performed
as described above.

5. Results and discussion

5.1. Preliminary study of system
The chromogenic reagent 1-(2-pyridylazo)-2-naphthol
(PAN) is a very versatile reagent for the determination of
trace amounts of numerous metals. PAN form 2:1 (ML2 ) colored complexes with Zn(II), Cu(II), Fe(III), Cd(II), Co(II) and
Ni(II) [16]. The absorption spectra of Zn(II), Ni(II) and Co(II)
complexes are shown in Fig. 4. As Fig. 4 shows, the spectra
of the PAN and the complexes overlap with each other, and

Table 1
Results of analysis of modeled data of X, Y and Z mixtures by proposed method for prediction of concentration of X
Sample number

1
2
3
4
5
6
7
8
9
10

Composition of synthetic mixtures

Predict concentration of X

Noise free

0.002 noise

0.003 noise

9.0
2.0
6.0
5.0
9.0
8.0
5.0
1.0
8.0
4.0

6.0
8.0
9.0
7.0
2.0
4.0
9.0
10.0
4.0
9.0

9.0
2.0
6.0
5.0
9.0
8.0
5.0
1.0
8.0
4.0

9.5
1.7
6.0
4.3
8.6
8.1
5.0
1.0
7.7
4.2

9.4
1.8
6.0
4.7
9.1
8.0
4.8
1.2
7.8
4.2

Z
1.0
3.0
8.0
1.0
1.0
2.0
4.0
6.0
10.0
2.0

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

873

Fig. 4. The zero-order spectra of g mL1 Zn2+ (1), 1 g mL1 Ni2+ (2)
and 1 g mL1 Co2+ (3) complexes with PAN (5 104 mol L1 ) at pH 9.2
in micellar media against PAN as blank.

therefore each compound interferes in the spectrophotometric determination of the others.

The PAN and its metal complexes are insoluble in water.
But, it has been observed that they were water soluble in the
presence of the neutral surfactant Triton X-100. In order to
find the optimum conditions, the effect of pH on the spectrum
of a constant concentration of each complex was investigated.
Although there was not significant changes in the spectra of
Co(II) and Ni(II) complexes with PAN in the pH range 310,
but at lower pH values (e.g. acetate buffer of pH 4.5), nickel
complex requires about 60 min for full color development,
while it requires less than 10 min in ammonium/ammonia
buffer of pH 9.2. Zinc cannot form a complex with PAN at
lower pH values, but its complex formation reaction is rapidly
completed at pH values greater than 7. Therefore, the pH of
the solutions was adjusted at 9.2 with ammonium/ammonia
buffer.
As mentioned above, PAN and its complexes with zinc,
cobalt and nickel are insoluble in water, therefore, Triton X100, a neutral surfactant, was used to make them soluble in
water. By using surfactant for dissolving the insoluble component in water, solvent extraction or mixed solvent systems
have been avoided. In order to obtain the optimum concentration of Triton X-100, the effect of surfactant concentration
on the absorbance of the system was investigated. The results
showed that the absorbance of the solutions increased by increasing Triton X-100 concentration in the range 0.81.2%
(v/v), remained constant between 1.21.6% (v/v) and slightly
decreased at higher concentrations. Therefore, 1.4% (v/v)
Triton X-100 was applied to provide micellar media in order
to prevent extraction procedure or using the mixed solvent
systems.
A 5 104 mol L1 PAN (at least 10-fold excess over
metal ions) was applied to obtain complete complexation of
Zn(II), Co(II) and Ni(II).

Fig. 5. The zero-order spectra of different concentrations (0.1, 0.5, 1.0, 1.5
and 2.0 g mL1 ) of Co2+ complexes with PAN (5 104 mol L1 ) at pH
9.2 in micellar media (a) and the first derivative of their second ratio spectra
obtained with  = 1 nm (b) according to Eq. (5).

5.2. Proposed method

As Fig. 4 shows, the absorption spectra of Co(II), Ni(II)
and Zn(II) complexes with PAN in micellar media overlapped
in the region 550700 nm.
The absorption spectra of the solutions of the CoPAN
complex prepared at different concentrations of cobalt were
recorded in the range 550700 nm (Fig. 5a) and divided by
the spectrum of the standard solution of ZnPAN complex
that was 0.5 g mL1 with respect to Zn(II) and the ratio
spectra were obtained. First derivatives of the ratio spectra
were obtained with  = 1 nm. After that, these vectors (first
derivatives of the ratio spectra) are divided by (d/d)(Ni /Zn )
corresponding to the derivative of the ratio of the spectra
of NiPAN complex that was 0.5 g mL1 with respect to
Ni(II) and ZnPAN complex that was 0.5 g mL1 with respect to Zn(II) under the optimum conditions and therefore,
second ratio spectra (according to Eq. (4)) were obtained.
First derivative of these vectors were obtained with  =
1 nm (Fig. 5b). The concentration of cobalt was determined
by measuring the amplitude at 584 nm corresponding to a
minimum wavelength shown in Fig. 5b.
The absorption spectra of the solutions of NiPAN complex prepared at different concentrations of nickel were
recorded in the range 550700 nm (Fig. 6a) and divided by the
spectrum of the standard solution of ZnPAN complex that
was 0.5 g mL1 with respect to Zn (under the optimum conditions) and the ratio spectra were obtained. First derivatives

874

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

Fig. 6. The zero-order spectra of different concentration (0.1, 0.5, 1.0, and
1.5 g mL1 ) of Ni2+ complexes with PAN (5 104 mol L1 ) at pH 9.2
in micellar media (a) and the first derivative of their second ratio spectra
obtained with  = 1 nm (b) according to Eq. (5).

of the ratio spectra were obtained with  = 1 nm. After that

these vectors (first derivative of ratio spectra) are divided by
(d/d)(Co /Zn ) corresponding to the derivative of the ratio of
the spectra of CoPAN complex that was 0.5 g mL1 with
respect to Co(II) and ZnPAN complex that was 0.5 g mL1
with respect to Zn(II) under the optimum conditions and second ratio spectra (according to Eq. (4)) were obtained. First
derivatives of these vectors were obtained with  = 1 nm
(Fig. 6b). The amount of nickel was determined by measuring the amplitude at 552 nm corresponding to a maximum
wavelength in the first derivative of second ratio spectra as
shown in Fig. 6b.
In the same way, the absorption spectra of the solutions
prepared at different concentrations of Zn-PAN under the
optimum conditions were recorded in the range 550700 nm
(Fig. 7a) and divided by the spectrum of the standard solution
of NiPAN complex that was 0.5 g mL1 with respect to Ni
(under the optimum conditions) and ratio spectra were obtained. First derivatives of the ratio spectra were plotted with
 = 1 nm. After that these vectors (first derivative of ratio
spectra) are divided by (d/d)(Co /Ni ) corresponding to the
derivative of the ratio of the spectra of the CoPAN complex
that was 0.5 g mL1 with respect to Co(II) and NiPAN
complex that was 0.5 g mL1 with respect to Ni(II) under
the optimum conditions and second ratio spectra (according
to Eq. (4)) were obtained. First derivatives of these vectors
were obtained with  = 1 nm (Fig. 7b). The amount of zinc
was determined by measuring the amplitude at 564 nm corresponding to a minimum wavelength in the first derivative
of second ratio spectra as shown in Fig. 7b.

Fig. 7. The zero-order spectra of different concentration (0.1, 0.5, 1.0, and
1.5 g mL1 ) of Zn2+ complexes with PAN (5 104 mol L1 ) at pH 9.2
in micellar media (a) and the first derivative of their second ratio spectra
obtained with  = 1 nm (b) according to Eq. (5).

5.3. Analytical characteristics

In the proposed method, Beers law was obeyed
in the concentration range 0.12.0 g mL1 for Co(II),
0.11.5 g mL1 for nickel and 0.12.0 g mL1 for zinc.
Table 2 shows the regression coefficients and calibration
equations of the calibration graphs for active compounds
at suitable wavelengths for determination of cobalt, nickel
and zinc in their ternary mixtures. Limit of detection of the
method for determination of Co(II), Ni(II) and Zn(II) (defined as the concentration equivalent to three times the standard deviation of 5 replicate measurements of the blank) are
also shown in Table 2.
To check the reproducibility of the method five replicate
resolving of Co(II), Ni(II) and Zn(II) mixtures were performed (Table 3). The relative standard deviation (R.S.D.) for
five replicate determination of 0.5 g mL1 each of Co(II),
Ni(II) and Zn(II) was shown in Table 3. Table 3 also shows
the mean recoveries for simultaneous determination of these
species in ternary mixtures.
The effect of divisor concentration on the analytical parameters such as detection limit, slope, intercept and correlation coefficient of the calibration equations was tested. It
was observed that changing the concentration of divisors in
their linear calibration range had no significant effect on the
analytical parameters. Therefore, a 0.5 g mL1 of each of
the Co(II), Ni(II) and Zn(II) was used as divisor spectrum in
the proposed method. The amount of  had no considerable

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

875

Table 2
Analytical characteristics of the method
Analyte

Wavelength (nm)

Calibration equation

R2a

Limit of detection (g mL1 )

Zn(II)
Co(II)
Ni(II)

564
584
552

y = 129.78x + 5.59
y = 192.66x + 6.0014
y = 13.004x 0.2576

0.9921
0.9999
0.9993

0.02
0.01
0.01

Squared regression.

Table 3
Results of five replicate experiments for analysis of Zn(II), Co(II) and Ni(II) in mixtures by proposed method
Taken (g mL1 )

Found (g mL1 )

Recovery (%)

Zn(II)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

0.5
0.5
0.5
0.5
0.5

0.5
0.5
0.5
0.5
0.5

0.5
0.5
0.5
0.5
0.5

0.46
0.50
0.51
0.48
0.50

0.5
0.52
0.53
0.51
0.49

0.53
0.51
0.49
0.50
0.51

92
100
102
96
100

100
104
106
102
98

106
102
98
100
102

0.49
0.02
4.08

0.51
0.016
3.14

0.51
0.015
2.94

98

102

102

Mean
S.D.
R.S.D. (%)

effect on the derivation of ratio spectra (for both the steps).

A  = 1 nm was used.
As the derivative spectra for chemical system used in this
work was not noisy, no smoothing was necessary. But for
system with noisy derivative spectra smoothing should be
performed after derivation.
In order to obtain the accuracy and precision of the method,
several synthetic mixtures with different concentration ratios
of Co(II), Ni(II) and Zn(II) were analyzed using the proposed
method. The results are given in Table 4. The prediction error
of a single component in the mixtures was calculated as the
relative standard error (R.S.E.) of the prediction concentration [17]:

R.S.E. (%) =

1/2

2
N

j=1 (Cj Cj )
N
2
j=1 (Cj )

100

(6)

where N is the number of samples, Cj the concentration of the

j the estimated concencomponent in the jth mixture and C
tration. The total prediction error of N samples is calculated
as follows:
1/2
 
M
N
ij Cij )2
(C
j=1
j=1
100
(7)
R.S.E.t (%) = M N
2
(C
)
ij
j=1 j=1
where Cij is the concentration of the component in the jth
ij its estimation. Table 4 also shows the reasamples and C
sonable single and total relative errors for such system.
5.4. Interference study
The effects of different ions on the simultaneous determination of 0.5 mg L1 each Zn(II), Co(II) and Ni(II) by proposed method was studied. An ion was considered as inter-

Table 4
Results of several experiments for analysis of Zn(II), Co(II) and Ni(II) mixtures in different concentration ratios by proposed method
Taken (g mL1 )

Found (g mL1 )

Recovery (%)

Zn(II)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

0.1
1.4
0.2
1.0
0.7
0.9
0.9
0.5

0.5
0.4
1.0
0.2
0.3
0.5
0.4
1.2

1.5
0.3
0.5
0.4
1.2
0.4
0.3
0.6

0.09
1.37
0.19
0.92
0.69
0.92
0.92
0.46

0.50
0.41
0.99
0.22
0.32
0.55
0.43
1.18

1.50
0.31
0.51
0.41
1.20
0.42
0.29
0.63

90.00
97.86
95.00
92.00
98.57
102.22
102.22
92.00

100.00
102.50
99.00
110.00
106.66
110.00
107.50
98.33

100.00
103.33
102.00
102.50
100.00
105.00
96.67
105.00

96.52
3.74

104.24
3.70

100.62
2.71
3.47

Mean recovery
R.S.E. (%)a single
R.S.E.t (%)b total
a
b

Calculated by Eq. (6).

Calculated by Eq. (7).

876

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

Table 5
Tolerance ratio of diverse ions on the determination of a mixture of 0.5 mg L1 Co(II) Zn(II) and Ni(II) ions
Ion

Tolerance ratio

SCN , CH3 COO , S2 O3 2 , tartrate, urea, thiourea, oxalate, F ,

citrate, Cl , I , Br , ascorbic acid Mg2+ , Fe2+ , Bi3+ , Sb3+ ,
Na+ , K+ , Mo, V5+ , Al3+ , Cr3+
3+
Fe , Hg2+ , Ag+ ,
Ca2+
As3+
Mn2+
Cu2+
Cd2+
a

Tolerance ratio in the presence

of 102 mol L1 EDTA

1000

>1000a
500
100
10
1
1

>1000
1000
1000
100
100

After filtering of the produced precipitate.

ferent, when it caused a variation in the absorbance of the

sample (at constant wavelength) greater than 5%. The results (Table 5) indicate that most of the cations and anions
did not interfere. Only Cu(II) and Cd(II) interfered at levels similar to that of Ni(II) and Co(II). In the presence of
1 102 M EDTA (added after formation of Zn(II), Ni(II)
and Co(II) complexes), all complexes of common cations
with PAN except Co(II), Ni(II) and Zn(II) complexes were
decomposed. At pH 9.2, Fe3+ , Ag+ and Hg2+ at concentrations grater than 50 mg L1 would completely precipitated.
The precipitate formed was then filtered before addition of
PAN into the solution. Therefore, these cations would not
interfere at higher concentrations.

Fig. 8. Plot of PRESS against the Number of factors for Zn(II), Ni(II) and
Co(II).

model was validated with a nine synthetic mixture set containing the considered metal ions in different proportions that
were randomly selected. To select the number of factors in
the PLS algorithm, a cross-validation method, leaving out
one sample at a time, was employed [17]. The prediction error was calculated for each metal ion for the prediction set,
which are the samples not participated in the construction of
the calibration set. This error was expressed as the prediction
residual error sum of squares (PRESS). Fig. 8 shows a plot
of the PRESS vs. the number of factors for each individual
component. For finding the smallest number of factors, the

5.5. Partial least squares (PLS) method

The first step in simultaneous determination of different
metal ions by PLS methodology involves construction the calibration matrix for the ternary mixture CoNiZn. A number
of 20 ternary mixtures were selected as calibration set. Their
composition was randomly designed for obtaining more information from the calibration procedure. Under these conditions, the calibration models were obtained. The obtained

Table 6
Composition of synthetic samples, their predictions by PLS model and statistical parameters for the system
Synthetic samples (mg L1 )

Prediction (mg L1 )

Recovery (%)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

Zn(II)

Co(II)

Ni(II)

Zn(II)

0.40
0.70
1.10
0.70
0.10
0.80
1.20
1.20
1.80

1.50
0.20
1.80
0.00
0.40
1.70
1.10
2.00
0.00

0.90
1.00
2.00
1.20
1.30
0.50
0.30
0.80
0.20

0.39
0.71
1.04
0.76
0.11
0.74
1.18
1.19
1.81

1.49
0.25
1.75
0.05
0.41
1.64
1.10
1.99
0.02

0.90
0.99
0.21
1.15
1.30
0.50
0.30
0.83
0.22

97.5
101.4
94.5
109
110
92.5
98.3
99.2
100.6

99.3
125
97.2

102.5
96.5
100
99.5

100
99
105
96
100
100
100
103.7
110

3.3

2.9

2.5
2.9

R.S.E. (%)a single

R.S.E.t (%)b total
a
b

Calculated by Eq. (6).

Calculated by Eq. (7).

A. Afkhami, M. Bahram / Spectrochimica Acta Part A 61 (2005) 869877

877

Table 7
Results of analysis of Zn(II), Co(II) and Ni(II) in synthetic alloy sample
Alloy

NKK No. 920

aluminum
alloy
a

Composition (%)

Si (0.78), Fe (0.72), Mg (0.46), Cr (0.27), Zn (0.8),

Ti (0.15), Sn (0.2), Pb (0.1), Sb (0.1), Bi (0.06),
Ga (0.03), Co (0.1), Mn (0.2), Ni (0.29), V (0.15)

Found (%) PLS methoda

Found (%) proposed methoda

Zn(II)

Ni(II)

Co(II)

Zn(II)

Ni(II)

Co(II)

0.83

0.29

0.10

0.79

0.31

0.11

After addition of 2 mL of 102 mol L1 EDTA for removing interfering effect of Cu(II) and Mn(II).

F-statistic test was used to carry out to significant determination. The PLS modeling for all elements had a same number
of factors. The optimal number of factors obtained was four.
The results obtained are given in Table 6.
5.6. Application
The proposed method was successfully applied to the simultaneous determination of Zn(II), Co(II) and Ni(II) in a
synthetic sample that was prepared according to the composition of an alloy [18]. The sample was also analysed using
PLS method. The results are shown in Table 7. The good
agreement between the obtained results and known values
indicate the successful applicability of the proposed method
for simultaneous determination of Zn(II), Co(II) and Ni(II)
in complex samples.

6. Conclusion
The proposed method is simple, very sensitive and easy
to understand and apply. Accuracy, precision, reproducibility,
sensitivity and linear rang for the proposed method are nearly
the same as those for PLS method.
Standard addition can be easily used in the proposed
method and matrix effects can be removed. Therefore, this
method can be used for resolving ternary mixtures in the
complex samples with unknown matrices. In multivariate
methods such as PLS standard addition can be used but that
needs another procedure known as generalized standard addition method [19]. The linear ranges, relative standard deviations (R.S.D.), mean recoveries and relative standard errors
show the applicability, accuracy and precision of the proposed
method.

Acknowledgment
The authors thank Mr. Mohsen Faghihi-Nick for his useful
help in mathematical discussion.
References
[1] E. Dinc, F. Onur, Anal. Chim. Acta 356 (1998) 93.
[2] S. Wold, M. Sjostrom, L. Eriksson, Chemom. Intel. Lab. Syst. 58
(2001) 109.
[3] E. Dinc, C. Yucesoy, F. Onur, J. Pharm. Biomed. Anal. 58 (2002)
1091.
[4] M. Blanco, J. Gene, H. Iturriaga, S. Maspoch, J. Riba, Talanta 34
(1987) 987.
[5] F. Bosch-Reig, P. Campins-Falco, Analyst 113 (1988) 1011.
[6] J. Verdu-Andres, F. Bosch-Reig, P. Compins-Falco, Analyst 120
(1995) 299.
[7] F. Salinas, J.J. Berzas Nevado, S. Maspoch, J. Riba, Talanta 37
(1990) 347.
[8] J.J. Berzas Nevado, C.C. Guiberteau, F. Salinas, Talanta 39 (1992)
547.
[9] J.J. Berzas Nevado, C. Guiberteau Cabanillas, A.M. Contento Salcedo, Talanta 42 (1995) 2043.
[10] F. Onur, C. Yucesoy, S. Dermis, M. Katral, G. Kukdil, Talanta 51
(2000) 269.
[11] A. El-Gindy, B. El-Zeany, T. Awad, M.M. Shabana, J. Pharm.
Biomed. Anal. 26 (2001) 203.
[12] E. Dinc, I.M. Palabyik, O. ustundag, F. Yustsever, F. Onur, J. Pharm.
Biomed. Anal. 28 (2002) 591.
[13] E. Erk, Y. Ozkan, E. Bannoglu, S.A. Ozkan, Z. Senturk, J. Pharm.
Biomed. Anal. 24 (2001) 469.
[14] E. Dinc, Talanta 48 (1999) 1145.
[15] E. Dinc, E. Baydan, M. kanbur, F. Onur, Talanta 58 (2002) 579.
[16] Q. Wei, B. Du, Talanta 45 (1998) 957.
[17] H. Abdollahi, Anal. Chim. Acta 442 (2001) 327.
[18] M.A. Taher, Anal. Chim. Acta 408 (2002) 153.
[19] D.L. Massart, B.G.M. Vandeginste, L.M.C. Buydens, S.D.E. Jong,
P.J. Lewi, J. Smeyers-Verbeke, Handbook of Chemometrics and
Qualimetrics. Part A, Elsevier Science, 1997.