Microsoft Word - 144-569-2-PRA

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

A NEW METHOD FOR THE DETERMINATION OF

TOXIC DYE USING FTIR SPECTROSCOPY

MOHAMED E. S. MIRGHANI
1
*, NASSERELDEEN A. KABBASHI

1
, ISAM Y. QUDSIEH

AND FAIZ A. ELFAKI
2

1
Bioenvironmental Engineering Research Unit (BERU), Department of Biotechnology

Engineering,
2
Department of Science in Engineering, Kulliyyah of Engineering, International

Islamic University Malaysia (IIUM), P. O. Box 10, 50728 Kuala Lumpur, Malaysia

*E-mail: elwathig@iiu.edu.my

ABSTRACT: A new method was developed to determine toxic dyes content in textile

and other products using Fourier Transform Infrared (FTIR) spectroscopy with

Attenuated Total Reflectance (ATR) element and KBr transmission cell. The

wavelengths used were selected using pure dyes and dye mixtures. Transmittance values

from the wavelengths regions 3500 – 2650 and 1675 – 1500 cm
-1
and partial least square

(PLS) regression method were used to derive FTIR spectroscopic calibration model for

dyes containing –N=N– in their structure. The coefficient of determinations (R
2
) for the

models were computed by comparing the results obtained from FTIR spectroscopy

against the actual values of the dyes concentrations. R
2
were 0.9321 and 0.9819 for two

samples of toxic dyes respectively. The standard errors (SE) of calibrations were 1.84

and 1.36 respectively. The calibration model was cross validated within the same set of

samples and the standard deviation (SD) of the difference for repeatability and accuracy

of the FTIR method were determined. With its speed and ease of data manipulation,

FTIR spectroscopy is a useful alternative method to wet chemical methods for rapid and

routine detection of azo dyes as toxic dyes in such products for quality control.

KEYWORDS: FTIR spectroscopy, PLS, Toxic dye.

1. INTRODUCTION

Dyeing is an ancient art which predates written records. It was practiced during the

Bronze Age in Europe [1]. Primitive dyeing techniques include sticking plants to fabric or

rubbing crushed pigments into cloth. Nowadays, most of the colors used commercially for

dyeing are synthetic, which mostly derived from non-renewable coal tar and petroleum.

They are synthesized by various means, from by-products of fossil fuels, e.g., aniline and

other aromatic derivatives [2]. Because clothing comes into prolonged contact with skin,

toxic chemicals are absorbed through the skin, especially when human body is warm and

skin pores have opened to permit perspiration. Once absorbed by humans, heavy metals

tend to accumulate in the liver, kidney, bones, heart and brain. The effects on health can

be significant when high levels of accumulation are reached. The effect is particularly

serious in children because toxic dye and/or heavy metals accumulation may negatively

affect their growth and may be their life as well.

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

It is often the dye fixative which is used to bond the dye color to the fabric that causes

the most problems. Unfortunately, heavy metals have often been used in dye fixatives and

also in dyes. Toxic chemicals sometimes found in the dyeing process include:

• Dioxin – a carcinogen and possible hormone disrupter.

• Toxic heavy metals such as chrome, copper, and zinc – known carcinogens.

• Formaldehyde – a suspected carcinogen.

• Azo dyes group – which give off carcinogenic amines.

In general, azo dyes can occur in two tautomeric forms; azo (–N=N–) or hydrazone

(=N–NH–). The latter is said to be more prone to the oxidative fading, which is the most

common photodegradation mechanism in the presence of light, moisture, and oxygen. In

oxidation of an azo dye, singlet oxygen attacks on the hydrazone tautomer to form

unstable peroxide, which then undergoes decomposition. Under anaerobic conditions, azo

dye can be reduced to its corresponding amines by abstracting a hydrogen atom from a

hydrogen donor [3].

Amendments and regulations in some countries stated that azo-dyestuffs, which can

release carcinogenic amines, should no longer be used in dyeing consumer goods.
13
C and

15
N-NMR studies of the azo-hydrazone tautomerism of some azo dyes were carried out by

Antonin and Valdimir [4]. Many acid and direct dyes which liberate harmful amines such

as benzidine, o-tolidine and o-dianisidine after reduction are, however, still used [5]. Toxic

chemicals from dyes also create widespread environmental havoc. Large amounts of water

are used to flush conventional synthetic dyes from garments and then this waste water

must be treated to remove the heavy metals and other toxic chemicals before it can be

returned to water systems, sewers and rivers.

Most garments are produced in developing countries where pollution controls are often

lax or nonexistent. Discharges from huge numbers of the textile producers go straight into

rivers where the river water might be bright green one day and yellow the next.

Developing countries are also lacking in standards and enforcement concerning the use of

toxic chemicals in dyes and garment finishes. Figures 1 and 2 showed the chemical

structure of some of the first commercial cellulosic fiber-reaction dyes used in fabric

industries that make use of the mono-chlorotriazinyl group. Figure 3 shows an orange dye

that is covalently bonded with the cotton fiber.

Low-impact, fiber-reactive dyes have become the dye of choice for many organic

clothing manufacturers who want a diverse palette of vibrant colors. Depending upon the

nature and degree of their chemical sensitivities, people with mild chemical sensitivities

can often wear organic clothing with fiber-reactive dyes. Un-dyed, natural color or color-

grown fabrics are the best choice for people who react to fiber-reactive dyes or who want

only pure fabrics on their skin.

As the awareness of the need to preserve our natural resources and environmental

factors increases, interest is growing in finding renewable resources, which can be used as

alternatives. Natural dyes are seen as more eco-friendly as, unlike their synthetic

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

counterparts, they are all derived from nature. Production of natural indigo from plant

material (Polygonum tinctorium Ait) applying low-technology methods was done

successfully by Bechtold et al. [6]. A better quality indigo dyestuff was extracted from the

indigo plant (Indigofera tinctoria L.) [7].

Earlier all colors were derived from natural sources, including plants. Looking for

better quality colors for industry, researcher went to the synthesis of dyes from by-

products of the petrochemical. Recent research is looking for the possibility to use plants

as commercially viable source of dyes [8]. This means that in order to satisfy the demand

for high quality and choice, the plants in question must be studied more closely to allow

breeding of useful lines and improved economic returns.

In this study, an FTIR spectroscopy was used to develop qualitative and quantitative

analytical environmentally friendly technique for different types of dyestuffs or dyed

substrates.

a
1

a
2

SO
2
Na

Fig. 1: Anthroquinonoid reactive dye (Procion Blue).

N
H

SONa

Fig. 2: Procion Brilliant Red.

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

Dye

Cotton
Celullose

Fig. 3: Reactive orange dye covalently bonded with the cotton fiber.

2. MATERIALS AND METHODS

2.1 Samples and Chemicals

All chemicals were of analytical grade. Different types of dyes samples were purchased

from local market. Fabric and hair dyes samples were used in this study. Solid samples

were used to make stock solution followed by preparing dilutions of known concentrations

ranged from 0 to 80 mg/mL (w/v) to be used as standard solutions. The samples were

tightly covered and shaken vigorously to homogeneity on an Autovortex SA1 mixer

(Stuart Scientific, Redhill, UK).

2.2 Chemical Analysis

Hair dyes and some fabric dye samples were determined by photometry using alkaline

solutions of Fe (II) triethanolamine as reducing agent following the methods suggested

and used by Bechtold et al. [9] and Merritt et al. [10] respectively, after some

modifications on handling samples.

2.3 Instrumental Analysis

The infrared spectra were recorded at room temperature with a Perkin-Elmer Fourier

Transform Infrared Spectrometer, Model spectra 100 series (Perkin-Elmer Corporation,

Norwalk, CT, USA), equipped with a deuterated triglycine sulfate (DTGS) detector and

controlled by a Perkin-Elmer PC. The software used for the FTIR data collection was the

Infrared Data Management (IRDM) system. The instruments were maintained in constant

humidity to minimize water vapor interference.

Drops from each standard were placed on ATR element and scanned. After each scan,

the ATR diamond were rinsed three times with acetone and dried with soft tissue before

adding the next sample. Calibration spectra were obtained from 64 scans at a resolution of

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

2 cm
-1
with strong apodization through 4000 - 400 cm

-1
frequency region. The spectra

were ratioed against the background air spectrum. All the scans were done in triplicate

with the spectra recorded as absorbance and stored on a disk for subsequent chemometric

analysis.

2.4 Statistical Analysis

All the experiments and measurements were done in triplicate. The relationships

between each of the FTIR spectrum parameter and the original data from the standard

solutions were determined using the software Nicolet Turbo Quant IR-Calibration and

Prediction Package, Version 1.1 (Nicolet Instrument Co., Madison, WI, USA).

PLS regression was used to derive the dye contents in the standard solutions. The actual

and spectral data were correlated and the correlation coefficients (r) taken as estimates of

the factor scores, which were then used as regressors to model both the spectral and actual

data. Selected factors were used in a multiple linear regression (MLR) to predict the

spectral values. The optimum number of factors employed in the calibrated models was

indicated by the predicted residual error sum of squares (PRESS) values. An example of

the PRESS plot is shown in Fig. 11, which also shows the F-test significance result for the

method (<0.05) and number of factors included in the calibration at the lowest value. The

F statistic for each PRESS value was calculated with all the factors prior to the number

with the smallest PRESS value. The optimum number of factors was empirically chosen as

that which gave the smallest PRESS value such that the F ratio probability drops below

1.0 or below 0.75.

Microsoft Excel spreadsheet software was used to organize the correlation of the FTIR

predicted and actual or chemical data. The good correlation obtained for the 30 calibration

samples indicated the adequacy of the FTIR calibration. Accuracy was assessed based on

the smallest standard error (SE) and the highest coefficient of determination (R
2
) [11].

2.5 Validation

The ‘leave-one-out’ cross-validation technique was used to verify the calibration

model. The PRESS was computed from the error in prediction from the standards and

plotted as a function of the number of factors employed in the calibration. The accuracy

was assessed by the standard error of cross validation (SECV) and R
2
. Further verification

was by the mean difference (MD) and standard deviation of difference (SDD) for

repeatability and closeness of fit (accuracy) between the chemical data and FTIR predicted

values.

3. RESULTS AND DISCUSSION

3.1 Chemical and FTIR Predicted Results

Table 1 shows the FTIR predicted values by the PLS statistical method as means and

standard deviations (SD) of the actual manually prepared data (Standard solutions) for dye

contents in the samples. The means and SDs for fabric dye were 46.53, 4.05 and 44.17,

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

3.75 mg/kg for calibration and cross-validation, respectively. For hair dye, they were

45.20, 3.69 and 45.82, 3.85 mg/kg, respectively.

Table 1: Calibration and cross-validation for dye content in fabric and hair dye samples

by FTIR methods in comparison with actual values of dye concentration.
a

Fabric dye Hair dye

Data Set Mean SD Mean SD

Calibration 46.53 4.05 45.20 3.69

Validation 44.17 3.75 45.82 3.85

a
FTIR, Fourier transform infrared spectroscopy; SD, standard deviation; All dataset obtained from three

replications.

Figures 4, 5 and 6 illustrate the spectra for dying stuff samples of hair dye, green dye,

and henna, respectively. The spectra are in the frequency range 4000 – 400 cm
-1
. Each

spectrum showed the characteristic absorption bands of dyeing stuff according to its

chemical structure [12, 13]. Figure 4 shows bands at 2020, 1480 and 1419 cm
–1
which

indicates the presence of alkyl groups. The band at 1510 cm
–1
could be assigned for the

aromatic compound possibly nitrile type compound alkyne ( C C, C N, and –N=N–)

or isocyanate (–N=C=O). The band at 1033 cm
–1
could be assigned for carboxylic acid

ester (R–CO–O–C–) or (R–CO–N–). The bands at 856 and 691 cm
–1
could be assigned for

aromatic ring and –C=S=C–, respectively. The spectrum of green dye (Fig. 5) shows

bands at 3308 cm
–1
and 2970 cm

–1
which could be assigned for overlaps of the C–H

stretching region and methyl C–H asymmetric stretch, respectively. The band at 1738 cm
–1

indicated the presence of carbonyl group (–C=O), 1635 cm
–1
may be assigned for quinine

or conjugated ketone, 1365 cm
–1
absorption band could be assigned for the presence of

nitro compounds NO2 stretch. The bands at 1229 and 1217 cm
–1
may suggest the presence

of aromatic ring. The weak band at 1092 cm
–1
could be assigned for C–O stretch, most

likely for secondary alcohol [14]. For the spectrum of dry henna in Fig. 6, the band at

3281 cm
–1
indicates the presence of OH stretch or H– bonded OH stretch. The bands at

2919 and 2851 cm
–1
could be assigned for methylene C–H asym./sym. stretch. The band at

1630 cm
–1
is for alkenyl C=C stretch and may be aryl- substituted C=C. The band at 1365

cm
–1
-(which is weak band compared to that one at same wavenumber on Fig. 5)- could be

assigned for the presence of trimethyl or “tert-butyl” (multiplet). A big band at 1027 cm
–1

could be assigned for –OH group and/or –NH2 group or both in the same structure.

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

Fig. 4: Spectrum of black hair dye sample (powder).

Fig. 5: Spectrum of green dye (liquid sample).

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

Fig. 6: Spectrum of dry henna (powder).

3.2 Selecting the Optimal Frequency Region for Prediction

The software Spectrum Quant Plus was used to obtain the optimal frequency region for

prediction, which is calculated by multiplying the difference between each standard

spectrum and the mean spectrum at each wavelength by the difference between the

corresponding property concentration and the mean property concentration, and summing

over all the standards. Peaks that do not correlate with the change in concentration are

summed to zero, producing a spectrum that highlights the peaks that change with change

in concentration, i.e. the peaks that relate to dye content. Thus, the correlation spectrum

was used to choose the 3500 – 2650 cm
-1
and 1400 – 800 cm

-1
regions use for calculating

dye contents of the samples as described by Fuller et al. [15]. High variances were seen in

the same regions (3000 – 2700 cm
-1
and 1600 – 1000 cm

-1
) for C-H, -OH, -NH2, –N=N-

and C=O [16]. Hence, the correlation and variance spectra were used to select the best

region for prediction using the PLS statistical technique to develop calibration from the

FTIR spectra and chemical data. R
2
and SEC from the PLS calibration were used to

choose the best region for determining dye content in dyestuff samples, 3500 – 2650 and

1675 – 1500 cm
-1
region was found to be the best for determination of these types of

dyestuffs.

3.3 Statistical Analysis

Figure 7 plots the data of actual values against the PLS FTIR-predicted values for dye

content in the fabric dye samples. The best correlation obtained at the highest R
2
(0.9321)

and lowest SEC (1.84). Figure 8 plots the cross-validation for the determination of dye

content in fabric dye samples (R
2
= 0.9618 and SECV = 1.45). Figure 9 plots the actually

prepared dye samples against the PLS FTIR spectroscopy predicted dye contents in hair

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

dye samples (R
2
= 0.9819 and SEC = 1.36) and Fig. 10 the cross-validation plot for dye

content in hair dye samples (R
2
= 0.9711 and SECV = 1.64).

Fig. 7: Calibration of the actual precisely prepared values of fabric

dye from 35 samples versus the PLS FTIR predicted values.

R
2
= 0.9618

-5

0 20 40 60 80

Actual value - fabric dye content (mg/kg)

F
T
IR
-p
re
d
ic
te
d
v
a
lu
e
-

fa
b
ri
c
d
y
e
c
o
n
te
n
t
(m

g
/k
g
)

Fig. 8: Validation plot of PLS FTIR predicted values versus actual

values of fabric dye content in 35 samples.

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

y = 0.9466x - 0.1677

R
2
= 0.9819

-5

0 20 40 60 80

Chemical value - hair dye content (mg/kg)

F
T
IR

-p
re
d
ic
te
v
a
lu
e
-

h
a
ir
d
y
e
c
o
n
te
n
t
(m

g
/k
g
)

Fig. 9: Calibration plot of PLS FTIR predicted values versus chemically

determined hair dye content in 35 samples.

R
2
= 0.9711

-5

0 10 20 30 40 50 60 70 80

Chemical value - hair dye content (mg/kg)

F
T
IR

-p
re
d
ic
te
d
v
a
lu
e
-

h
a
ir
d
y
e
c
o
n
te
n
t
(m

g
/k
g
)

Fig. 10: Validation plot of PLS FTIR predicted values versus chemically

determined hair dye content in the 35 samples.

The optimum number of factors employed in the calibrated models was indicated by

the predicted residual error sum of squares (PRESS) values and were 6 and 8 for the fabric

and hair dyes samples, respectively. PRESS plot is shown in Fig. 11, which also shows the

F-test significance result for the method (<0.05) and number of factors included in the

calibration of fabric dye samples was 6. The F statistic for each PRESS value was

calculated with all the factors prior to the number with the smallest PRESS value. The

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

optimum number of factors was empirically chosen as that which gave the smallest

PRESS value such that the F ratio probability drops below 0.75.

Fig. 11: Press plot obtained from cross-validation of PLS calibration for determination of

fabric dye sample.

4. CONCLUSION

FTIR spectroscopy gives very accurate frequencies in the spectrum - this enables

processing techniques such as spectral subtraction as well as it has a much shorter

sampling time compared to wet chemical methods and even shorter than conventional

spectroscopic techniques. FTIR spectroscopy is a useful environmental friendly method

for rapid and routine detection of azo dyes as toxic dyes in such products for quality

control because no chemicals are used in the detection. The full possibilities of FTIR

spectroscopy are still far from being exploited.

REFERENCES

[1] P. J. W. Stone, A. L. Rhoden, L. S. Miller, D. J. Walton, A. M. McRoberts and J. Heptinstall,
Polysiloxanes for Langmuir-Blodgett Applications, Thin Solid Films, Vol. 244, no. (1-2)

1012-1017, 1994.

[2] Hélène Rudyk, Michael H. Knaggs, Snezana Vasiljevic, James Hope, Chris Birkett and Ian
H. Gilbert. Synthesis and Evaluation of Analogues of Congo red as Potential Compounds

against Transmissible Spongiform Encephalopathies, European Journal of Medicinal

Chemistry, Vol. 38, no. 6, 567-579, 2003.

[3] Katri Vikman, Studies on Fastness Properties of Ink Jet Prints Using Vibrational
Spectroscopic Methods, Thesis for the degree of Licentiate of Science in Technology

submitted for inspection, Espoo, November 18
th
, 2002.

IIUM Engineering Journal, Vol. 9, No. 2, 2008 Mirghani et al.

[4] L. Antonin, and M. Valdimir,
13
C and

15
N-NMR Studies of the Azo-hydrazone Tautomerism

of Some Azo Dyes. Dyes and Pigments, Vol. 7, no.3, 171-185, 1986.

[5] S. W. Oh, M. N. Kang, C. W. Cho and M. W. Lee. Detection of Carcinogenic Amines from
Dyestuffs or Dyed Substrates, Dyes and Pigments, Vol. 33, no. 2, 119 – 135, 1997.

[6] T. Bechtold, A. Turcanu, S. Geissler, and E. Ganglberger. Process Balance and Product in the
Production of Natural Indigo from Polygonum tinctorium Ait, Appling Low-technology

Methods, Bioresource Technology, Vol. 81, 171-177, 2002.

[7] H. Schmidt, Indigo – 100 Jahre industrielle Synthese, Chem. Unserer Zeit, Vol. 31, 121-128,
1997.

[8] G. Stoker Kerry, and T. Cooke David. Dyes from plants: Past usage, present understanding
and potential, Plant growth regulation, Vol. 34, 57-69, 2001.

[9] T. Bechtold, E. Burtscher, and O. Bobleter. Konzentrationsbestimmung von Indigo in Ansatz-
und Prozessnadern, Text.-Prax. Int., Vol. 47, 44 – 49, 1992.

[10] J. T. Merritt III, K. R. Bock, C. B. Smith, P. J. Hauser, and W. J. Jasper, Determination of
Indigo in Dye baths by Flow Injection Analysis and Redox Titrations, AATCC Rev. ¼: 41 –

45, 2001.

[11] J. Dubois, F. R. van de Voort, J. Sedman, A. A. Ismail and H. R. Rawaswamy, Quantitative
Fourier Transform Infrared Analysis for Anisidine Value and Aldehyde in Thermally

Stressed Oils, J. Am. Oil Chem. Soc., Vol. 73, 787 – 794, 1996.

[12] K. Nakanishi, and P. H. Solomon. Infrared Absorption Spectroscopy, 2
nd
Edn. Holden-Day,

Boca Raton, Fl, pp. 100 – 235, 1977.

[13] N. P. G. Roeges, A Guide to the Complete Interpretation of Infrared Spectra of Organic
Structures, John Wiley & Sons Ltd, Baffins Lane, Chichester, England, p. 339, 1994.

[14] J. Coates, Interpretation of Infrared spectra, A Practical Approach, in Encyclopedia of
analytical Chemistry, R. A. Meyers (ed.). John Wiley & Sons Ltd, Chichester. pp. 10815 –

10837, 2000.

[15] M. P. Fuller, G. L. Ritter, and C. S. Draper, Partial Least-Square Quantitative Analysis of
Infrared spectroscopic Data, Part I: Algorithm Implementation, Applied Spectroscopy, Vol.

42, no 2, 217-227, 1988.

[16] M. P. Fuller, G. L. Ritter, and C. S. Draper, Partial Least-Square Quantitative Analysis of
Infrared spectroscopic Data, Part II Application to Detergent Analysis, Applied Spectroscopy,

Vol. 42, no. 2, 228-235, 1988.