Colour Matching of Dyed Wool by Vibrational Spectroscopy Mandana Mozaffari-Medley BSc. A thesis submitted in partial fulfilment of the requirement of the Degree of Master of Applied Science School of Physical and Chemical Sciences Queensland University of Technology July 2003
258
Embed
Colour Matching of Dyed Wool by Vibrational Spectroscopyeprints.qut.edu.au/15867/1/Mandana_Mozaffari-Medley_Thesis.pdf · known as keratins (from the Greek word 'κερας' meaning
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Colour Matching of Dyed Wool by Vibrational
Spectroscopy
Mandana Mozaffari-Medley BSc.
A thesis submitted in partial fulfilment of the requirement of the Degree of Master of Applied Science
School of Physical and Chemical Sciences Queensland University of Technology
July 2003
ii
Statement of Original Authorship
The work contained in this thesis has not been previously submitted for a
degree or diploma at any other higher education institution. To the best of my
knowledge and belief, the thesis contains no material previously published or
written by another person except where due reference is made.
iii
"w¬Ã�f ͤ‡ ä… ,w¬� ÑŸ£•"
dZ¯‹®• pz®•
“I conquered, I got published”
Forough Farokhzaad, Pioneer female poet I would like to dedicate this thesis to my family, friends, and colleagues who believed in me and encourage me to be a true scientist. Especially Dr. B. Frost, the dean of faculty of science at the University of Queensland.
iv
Acknowledgement
I would like to acknowledge my supervisor, Dr. Serge Kokot for providing me with a very interesting project. I would like to thank him for being there whenever I was facing a challenge in the course of my study. I would like to thank Professor George, my co-supervisor. For that although as the dean of the faculty he was very busy, he always found time to listen and provide guidance and encouragement. I also wish to thank Dr. Jeff Church, from CSIRO, for providing me with samples to study and for his help at the beginning of my study. Many many thanks go to Dr. Llew Rintoul for teaching me how to use the instruments and for discussing various aspects of my work and providing me with excellent ideas. Not to mention that he spent a few Saturday and Sunday afternoons reading my thesis. Thank you Llew, you are an invaluable friend. I would like to thank Associate Professors Brian Thomas and Peter Fredericks and Ms. Elizabeth Stein. Without your support this thesis would not have been possible. I would like to thank Drs. Dalius Sagatys, Bob Johnson and Graham Smith for listening to me when I was down and for inviting Greg and I to their circle of friends. All the postgraduate students are not forgotten, specially Dr. Shona Stewart, Sandra Dütt and Thanh Van den Elst, for making my life at QUT that much more pleasant. Also I would like to thank a few people at work: my boss Cathrine Neuendorf for supporting me in taking time off work to finish, Sandra for listening to all my stories and buying me lots of gifts to cheer me up and Helen Woods for making sure that I was on the track and that I had a plan to follow. I’d like to thank my parents. To whom I owe my achievements: To my father, the greatest textile engineer, whose passion for textiles inspired me to do this project. Dad, many times when I read about wool processing, I remembered you, teaching me enthusiastically about the textile science of fabrics whenever the possibility came up. Does summer holidays in Europe ring a bell!!! To my mother, a teacher by profession and in life. You are my role model. You have taught me never to give up and always think positive. Thank you for being there for me.
v
I also wish to thank my brother, Maziar, who is also finishing his PhD thesis. Thank you for all your support and encouragement. I am proud to be your sister. Last but by no means least; I’d like to thank my husband, Dr. Greg Medley. For all that time that I discussed my scientific ideas with him, for teaching me how to write a thesis and for correcting all my grammatical errors and funny English!!! For many times that he worked 12-14 hours at uni so that I would keep going, for listening to my grumbles and always saying: “Sweetie, you are going to be fine, keep going”.
vi
Abstract
The matching of colours on dyed fabric is an important task in the textile industry.
The current method is based on the matching the visible reflectance spectrum to
standard spectral libraries. In this study, the amount of dye on various wool and
wool-blend fabric was measured using vibrational-spectroscopic techniques.
FT-IR PAS and FT-Raman spectroscopy was used to analyse the following set of
samples: woollen fabrics (supplied by CSIRO- Geelong, Australia), dyed with
Lanasol dyes (Red 6G, Blue 3G and Yellow 4G) and wool/polyester fabrics (supplied
by Ceiba-Geigy, Switzerland), dyed with Forosyn dyes (grey, yellow, green, brown,
orange, red). A minimum of six spectra was recorded for each sample. The spectra
recorded were consistent with those reported previously. FT-IR PA spectral data were
block normalised with Y-mean centring and examined using Principle Component
Analysis (PCA) and Partial Least Squares (PLS).
Although PCA separates the woollen fabrics dyed with a combination of two colours,
it does not do equally well for samples dyed with three colours. The dyed wool/
polyester blend samples appeared in a totally random fashion on the PCA plot.
The PLS analysis of PA spectra of various ratios of dyes on woollen fabrics as well as
wool/polyester blend was found to be a viable procedure and should be investigated
further, perhaps with a broader set of data.
vii
FT-Raman spectra were examined using PCA and PLS. The best pre-treatment for
FT-Raman spectral data was found to be normalising followed by Y-mean centring.
The PCA plots demonstrate that woollen samples are separated according to the dye
ratios and that the presence or absence of some of the peaks is influenced by
individual dyes. For example, the presence of the peak at 1430cm-1 is inversely related
to the presence of blue dye on the fabric. The PLS resulted in SEE and SEP values of
around 1 and 2 respectively indicating that the prediction of the dye ratios have not
been very successful and suggesting that there was some problem with the measured
values of the calibration set.
PCA plots of wool/polyester fabrics dyed with a single colour indicate that PC1
separates the samples according to how close the shades are together, while PC2 and
PC3 separate samples according to their individual colours. PC4, although explaining
only a small percentage of variance, suggests that the samples are not homogeneously
dyed. PCA plots of the samples dyed with various combinations of the three main
dyes display each cluster of samples in their right position on the colour card.
Calculated SEE and SEP values (Yellow: ~0.30, ~0.55, Brown: ~0.30, ~0.79, Red:
0.16, 0.49 and Grey: ~0.2, ~0.40, respectively) indicate that FT-Raman spectroscopy
and chemometrics may offer promising methods for measuring the ratio of various
dyes on wool/polyester fabrics.
FT-Raman spectroscopy and chemometrics were also used to investigate the change
in the ratio of dyes on UV-treated dyed woollen samples. Samples were weathered for
7 and 21 days, using accelerated weathering instrument. The substrate subtracted
viii
spectral data were normalised to 100% substrate of the first derivative (9 points and 7
degrees) followed by double centring of the matrix in the spectral region of 1500-
500cm-1. PCA effectively separated non-irradiated from the irradiated sample but did
not separate the irradiated samples further according to the number of days of
irradiation. The pre-treatment used for PLS was first derivative of substrate subtracted
spectral data normalised to 100% substrate, and then Y-mean centred. PLS failed to
predict the ratio of the irradiated dyes very well. This may be because degradation
products are not modelled by PLS or because the total amount of dye has reduced
Res. Conf.’, Pretoria, Vol. 1, 1980. 10 J.H. Bradbury, K. F. Ley, Australian J. Biol. Sci., 25, 1972, 1235. 11 J. L. Bradbury, G.V. Chapman, N.L.R. King, Australian J. Biol. Sci., 18, 1965, 353. 12 J.A. MacLaren, B. Milligan, ‘Wool Science, The Chemical Reactivity of the Wool Fibre’,
Science Press, Ch.1. 13 J.D. Leeder, Wool Sci. Rev., 63, 1986, 3 14 P.R. Blakey, R.Guy, F. Happey and P. Lockwood, ‘The Effect of Chemical Modifications on
the Morphological Structure of Keratin’, Applied Polymer Symposium No. 18, 1971, pp. 193-200
Conf., Tokyo, 5 (1985) p. 99. 21 J.P. Human, J.B. Speakman, J. Text. Inst., 45, 1954, T497. 22 A.Korner, Proc. 1st Intern. Symp. On Speciality Animal Fibre, Aachen, 1987, p.104. 23 D.E. Rivett, Wool Sci. Rev., 67, 1991, 1. 24 R D Fraser, T.P. Macrae and G.E. Rogers; ‘Keratins-Their Composition, Structure and
Biosynthesis’ (Springfield, USA: C.C Thomas, 1972) 25 H. Baumann in ‘Fibrous Proteins: Scientific, Industrial and Medical Aspects’; D.A.D. Parry
and L.K. Creamer (ed.); vol 1; (London: Academic Press, 1979) p. 299. 26 R.D. Fraser and T.P. Macrae, Milton Harris; ‘Chemist, Innovator and Entrepreneur’,
M.M. Breuer (ed.), (Washington DC: Amer. Chem. Soc., 1982), p.109. 27 R.S. Asquith, N.H. Leon in ‘Wool Dyeing’, D.M. Lewis (ed.) (1992), Ch.5. 28 J.A. Rippon in ‘Wool Dyeing’, D.M. Lewis (ed.) (1992), p.11. 29 A.C. Welham in ‘Wool Dyeing’, D.M. Lewis (ed.) (1992), p.116. 30 K.R. Makison, ‘Shrinkproofing of Wool’, (Marcel Dekker Inc., New York, 1979). 31 D.M. Lewis in ‘ Wool Dyeing’, D.M. Lewis (ed) (1992), Ch.8. 32 D. Mäusezahl; Textilveredlung, 5, 1970, 839. 33 A. Büher, R. Hurter, D. Mäusezahl, J.C. Petitpierre; ‘Proc. Int. Wool Text. Res. Conf.’,
Aachen, Z63 (1975). 34 J. Shore; J. Soc. Dyers Col., 84, 1968, 408. 35 J.A. Maclaren, B. Milligan; ‘Wool Science, The Chemical Reactivity of the Wool Fibre’,
(1981), p. 168. 36 P.G. Cookson & F.J. Harrigan;’ Wool Dyeing’, D.M. Lewis (ed.) (1992), p. 257. 37 B. Angliss, Textile Progress, 12 (3), 1982, 9. 38 A.M. Duthie, J.S. Church, W-H. Leong, ‘A Study of Solution-Dye Interactions for Rhodamine
6G’, Proceedings of 1st Australian Conference on Vibrational Spectroscopy, Sydney (1995), p.137.
39 L.A. Holt, L.N. Jones, I.W. Stapleton, ‘Interactions Between Wool Weathering and Dyeing’, Proceedings of the 8th International Wool Textile Research Conference, N.Z., Vol. IV (1990), p. 117.
40 D.M. Lewis, J. Soc. Dyer Color. , 98 (5-6), 1982, 165-176. 41 I.Rusznàk, J. Frankl, J. Gombkötő, J. Soc. Dyer Color., 101, 1985, 130-136.
Chapter 1 Introduction 39
42 A.S. Davie, W.H. Leong, D.J. Tucker, J.S. Church, ‘FT-Raman and FT-Infrared Studies of
Lanasol Type Dyes and Their Reactions with Selected Amino Acids’, Proceedings of 1st Australian Conference on Vibrational Spectroscopy, Sydney (1995), p.p. 35.
43 H. Baumann, J. Soc. Dyer Color., 90, 1974, 125-129. 44 H. Baumann, ‘The Effect of Reactive Dyes on the Main-chain Scission of Wool on Exposure
to Light’, J. Soc. Dyer Color., 90, 1974, 326-328. 45 F.W. Billmeyer, Jr., M. Saltzman; ‘Principles of Colour Technology’, 2nd ed. (John Wiley &
Sons, 1981), p.17. 46 National Geographic, Vol. 196, No. 1, 1999. 47 D. Spitzer, R. Gottenbos, P. van Hensbergen, M. Lucassen, 29, 1996, 235-238. 48 P. Pugh, American Ink Maker, 75(6), 1997, p.p. 65-66. 49 J.C. Crowther, J. of the Am. Leather Chemists Association, 84(6), 1989, 184-199. 50 R. Zbinden, ‘Principles of Colorimetry’, Colorimetry, (1962,Basle). 51 C.J. Sherman, Polymer Paint Colour Journal, 177, 4190, 1987, 296. 52 J.S. Church, W-H. Leong; ‘ The Analysis of Wool Textile Blends by FT-IR PAS and FT-
Raman Spectroscopies’, ‘9th Intn. Wool Text. Res. Conf.’, Biella, Italy, IV (1995), pp. 114-122.
53 A. Rosencwaig, E. Pines, Biochimica Biophysica Acta, 493, 1977, 10-23. 54 A. Rosencwaig, ‘Photoacoustics and Photoacoustic Spectroscopy’, (John Wiley and Sons,
New York, 1980), pp. 94-124. 55 J.M. Chalmers, M. W. Mackenzie in ‘Advances in Applied Fourier Transform Infrared
Spectroscopy’, M.W. Mackenzie (ed.), (John Wiley & Sons Ltd., 1988), Ch. 4. 56 L.H. Lee, “Photoacoustic Spectroscopy For the Study of Adhesion and Adsorption of Dyes
and Polymers”, Polymer Science and Technology Series, 12(A), 1980, pp. 87-102. 57 L.E. Jurdana, K.P. Ghiggino, I.H. Leaver, P. Coleclarke; Applied Spectroscopy, 49 (3), 1995,
199. 71 J.S. Church, K.R. Millington, Biospectroscopy, 2, 1996, 249-258. 72 G. Lee-Son, R.E. Hester, J. Soc. Dyer Color., 106, 1990, 59-63. 73 I. Keen, ‘Forensic Application of Raman Microprobe’, Masters Thesis, Queensland University
of Technology, (1998), Chapter 4. 74 J. Cheng, ‘Characterisation of Wool Treated with Metal Ions’, Master Thesis, Queensland
University of Technology, (1993), Chapter 6.
Chapter 2 Experimental 40
Chapter 2
Experimental
2.1 Materials
Fabrics
Two types of fabric were used in these studies. The CSIRO Textile and Fibre Technology
located in Geelong, Victoria provided a set of twenty-four samples. These were plain weave
Merino woollen fabrics dyed with three types of Lanasol dyes: red, blue and yellow. Each
sample was dyed with various ratios of two or three of these colours, so that the ratios added to
10. For example, a sample referred to from here on as XYZ means that there are X parts of red,
Y ratios of blue and Z ratios of yellow dyes on the fabric, and where X+Y+Z = 10 ratios.
The other set of fabrics studied consisted of a 55:45 wool/polyester blend. They were
plain-weave fabrics dyed with Forosyn dyes (Sandoz Ltd., Switzerland) pre-set for 30 seconds at
190°C.
Three colour squares (or diamonds) were studied here. Each diamond has samples with various
combinations of three out of four colours present in that particular diamond. Appendix 1 shows
Diamonds 1, 2 and 3. As seen, each sample has a reference, which consist of three numbers. The
first one indicates the number of parts of one colour (yellow or green), while the next two give
the number of part of the other two colours (brown and grey, respectively). For example, a
reference number 600 represents Yellow colour, which is equal to 1.8% Forosyn Yellow 2RL.
Chapter 2 Experimental 41
The same rule may be applied to the other two diamonds as well. The colours available in each
diamond are as follows:
I) Diamond 1:
Forosyn Yellow 2RL, Forosyn Brown RL, Forosyn Green 2GL, Forosyn Grey 2BL
II) Diamond 2:
Forosyn Brown RL, Forosyn Yellow 2RL, Forosyn Grey 2BL, Forosyn Red BL
III) Diamond 3:
Forosyn Brilliant Orange R, Forosyn Green 2GL, Forosyn Red BL, Forosyn Brown RL
Fading of Colour on Woollen Fabrics by UVA Irradiation
Swatches of the pure wool samples described in section 1.1, were used to study the fading of the
colours on these samples. UVA irradiation was applied to nine samples selected out of the
twenty-four. These samples were selected so that they had various ratios of red dye. The colour
chosen was however picked out of the three at random and not for a particular reason. The
samples were irradiated for a period of seven and twenty-one days respectively. They were
observed at 24 hour intervals to monitor the fading rate of the colours. After the completion of
each irradiation period, the samples were stored in the dark until they were studied by
vibrational spectroscopy.
Chapter 2 Experimental 42
2.2 Instrumentation
UV Accelerated Weathering Instrument
The equipment used for this purpose was called SUNTEST CPS+ (W.C. Heraeus GmbH,
Germany).
The test chamber, which consists of a parabolic reflector, sample table, xenon lamp and photo
diode to measure the radiation of the lamp. The lamp and test chamber are air-cooled. The
specifications of the instrument are shown in Table 2.1
Table 2.1- Specifications for the Accelerated Weathering Instrument
Dose 453600kJ/m2
Distance from lamp axis to sample level Approx. 230mm
Irradiance for the wavelength range below 800nm for filter system "max. UV"
765W/m2
Sample swatches of the woollen fabric approximately 2×9cm were stapled onto a cardboard
mount and then placed in the sample holder of the instrument such that the fabric was 23cm
away from the illumination tubes.
2.3 Vibrational Spectroscopy Analysis
Both sets of samples were analysed using FT-IR Photoacoustic Spectroscopy (FT-IR PAS) and
FT-Raman spectrometry.
Chapter 2 Experimental 43
Sample Preparation
Circular pieces of dyed woollen fabric of approximately one centimetre in diameter were
prepared with a wad punch. These were then arranged in various manners with careful
attention to the direction of the warp and weft.
Three layers of samples were put on the top of each other such that their respective warp and the
weft were running in the same direction. Meyer1 has previously shown that samples with 90°
and 180° warp and weft orientation exhibit the smallest difference in energy in PAS, and give
identical spectra. While other orientations, such as when a sample is positioned with its warp at
45° angle to the next one, spectra with slightly different intensities were obtained.
Wool/polyester blend fabrics were also studied using FT-IR PAS. Three small swatches (approx.
5×5mm) were positioned in the sample holder. Again, the samples were positioned so that warp
and weft of each fabric was aligned with respect to those of the sample underneath. These
sample stacks were put in the middle of the phto-acoustic cell. The alignment of the sample with
the IR beam was aided by the use of a liquid crystal thermal imaging sheet supplied by the
manufacturer of the instrument.
After the acquisition of each spectrum, the top swatch was moved to the bottom of the stack
hence exposing the next sample for measurement.
Chapter 2 Experimental 44
Vibrational Spectroscopy Instrumentation
FT-IR Photo-Acoustic Spectroscopy
PA spectra were acquired with an MTEC Model 200 standard sample gas-microphone accessory
with an accompanying MTEC pre-amplifier power supply. The spectral signal was then
transferred to the FT-IR series 2000 signal processor.
A background spectrum of finely powdered pressed carbon black was collected at the beginning
of each session. Both the background material and the samples were purged for 10 minutes at
20cc.min-1 with UHP grade Helium. The chamber was then sealed and the spectrum recorded.
The correct purging time for the samples was determined by first testing different times until an
acceptable spectrum could be recorded. That was done on the basis of the water vapour content
in the spectrum.
The conditions at which the spectra were recorded are tabulated below and unless indicated
otherwise, these conditions were kept constant throughout the studies, for both types of fabric.
Chapter 2 Experimental 45
Table 2.2 Specification and operating parameters for the PE-2000 FT-IR spectrometer Detector MTEC
Wave number range 4000-450cm-1
Number of Scans 128
Apodisation Strong
Resolution 8.0cm-1
OPD velocity 0.2cm.s-1
Interferogram Bi-directional, double sided
Acquisition time 10 min
MTEC Pre-AMP Gain 2.0
Phase correction Self, 256 pts
Three swatches were taken from each fabric to ensure that there was no variation across the area
of the fabric. Spectra were recorded for each of these swatches on two different days to ensure
that there was no variation over time.
FT- Raman Spectroscopy
The samples prepared for FT-IR PAS method were also examined by FT-Raman spectroscopy.
Samples were studied using Perkin Elmer System 2000 NIR/ FT-Raman spectrometer. This
instrument is equipped with Nd:YAG laser emitting at 1064nm.
Various conditions such as different laser power and mirror velocities were tried as well but it
was found that the below conditions were optimal. The spectra were then recorded six times for
three pieces of the same sample. The conditions of the recording spectra are tabulated below:
Chapter 2 Experimental 46
Table 2.3 Specification and operating parameters for the PE-2000 NIR/ FT-Raman
spectrometer
Laser Type Nd-YAG (1064 nm)
Laser Power 300mW
Phase correction Self
Interferogram type bi-directional, double sided
Detector InGaAs
Beam splitter Quartz
Gain 1.0
Wavenumber Range 3800-200cm-1
Resolution 8cm-1
Number of Scans 300
Mirror Velocity 0.2cm.s-1
Beer Norton Apodisation Strong
The optimum sample alignment was determined using BaSO4 (Hopkin and William, London).
2.3 Data Processing
Data Manipulation
All data manipulations in these studies were performed using the Grams 32 software package2,
Excel 97 and SIRIUS3 6.0. Data collected were transferred to Grams 32 and changed to ASCII
file type. The baseline was flattened and zeroed. (ie. a linear baseline was subtracted to bring
both ends of the spectrum to zero intensity). The data interval was increased by averaging over 4
cm-1. Since the region 1800-800cm-1 contained most of the peaks for wool, the spectra were
truncated to this region. Further studies of the spectra have revealed that the changes due to
Chapter 2 Experimental 47
various colours occur at ca.1500-925cm-1 for FT-IR PAS. Therefore, the data were studied in
this region.
Both the photoacoustic and the Raman spectra were also analysed as their first derivatives. First
derivative (9 Points, 7 degrees) spectra were obtained using the MacroWizard package in
Grams 32.
Data Pre-treatment
Before more detailed chemometric analysis is performed, it is often useful to pre-treat the raw
spectral data to make all of the spectra compatible with the chemometric process. The type of
pre-treatment that is required depends on the origin of the data and the context of the problem4.
There are two basic types of pre-processing: transformation of variables and transformation of
samples. In variable transformation, systematic changes across the columns of the data matrix
are removed. The most common form of variable transformation is column centring where the
mean of each column in the data matrix (X) is subtracted from every value in that column. ie.
m
XXX
m
1kki
ijcentred]-[column
ij
∑=−=
Where, m is the number of rows in the data matrix.
Depending on the nature of the problem5, it may be helpful to column-standardise the data by
dividing the column-centred values by the standard deviation of the column. i.e.
Chapter 2 Experimental 48
m
m
XX
XX
m
1k
m
1lil
ik
centred][columnij[standard]
ij
∑∑
=
=
−
−
=
Sample transformation follows similar procedures to variable transformation with the
transformation happening across the rows of the data matrix rather than the columns.
Data may be row-centred. Since almost all data is column-centred, this is known as doubled-
centring.
n
XXX
n
1k
centred][columnin
centred][columnij
centred][doubleij
∑=
−
−− −=
Double-centred data sets have no features in common throughout the data; they only show the
data variation.
The sample data may be normalised. This may be by an internal standard in which case the
value for the standard in subtracted from the corresponding data values. Alternatively, it may be
normalised to a constant sum in which case the normalised value ijX ′′ is given by:
∑=
=′′ n
1jij
ijij
X
XX
Chapter 2 Experimental 49
In this study, the data were imported into the Excel 5.0 spreadsheet for pre-treatment. Once in
the spreadsheet, the data matrix was double-centred, normalised and standardised as appropriate
before being transferred to the SIRIUS 6 software package for further chemometric analysis.
2.4- Chemometrics
There are two general objectives in the chemometric analysis of spectral data: qualitative
classification and quantitative measurement of component concentrations.
Qualitative analysis involves the classification of a spectroscopic sample into one or more
classes on the basis of its spectrum. The class is defined by a set of samples of known class (the
training set). The calculation of the quality of fit of the spectrum to the class may also be
possible. In this study, principal component analysis (PCA) and subsequent use of the SIMCA
technique were used for classification of spectral data.
Quantitative measurement requires first a calibration with a training set of multiple samples with
a known concentration of multiple analyte components and then the calculation of the
concentrations in an unknown sample in such as way as to minimise prediction error. In this
study, partial least squares analysis was used for quantitative measurement.
Chapter 2 Experimental 50
Principal Component Analysis
Principal Component Analysis (PCA) is a multivariate reduction technique in which the original
multivariate data is reprojected onto a set of orthogonal axes. The new projections are known as
the principal components (PCs).
When the raw data (Xij) consists of n objects described by m independent variables, we may
describe the data by a set of principal components (PCij) which are linear combinations of Xij.
∑=
=m
1kjkikij XaPC 3.1
where aik is the loading of the variable k on PCi.
When there are m independent variables, there will also be m orthogonal PCs in any given set
and all m PCs will be needed to fully describe the data.
In PCA the loadings are chosen such that the maximum amount of variance is described in the
first PC (PC1). The next largest variance is described by PC2 and so forth with progressively
decreasing variance down to PCm. In this way it is often possible to give a good description of
the data objects with only the first few PCs rather than a large number of independent variables.
Because most of the variance is described by the first few PCs, it is possible to display the data
in two-dimensional plots of one PC versus another. In this way subtle relationships between the
objects, which would otherwise be concealed by the mass of data, may be revealed.
Chapter 2 Experimental 51
Useful information may also be obtained by inspection of the loadings. For example, if two
classes of objects are effectively distinguished by a particular PC (PCi) then an inspection of the
loadings ai1…aim may reveal which of the original variables distinguish the two classes of object.
In the raw data this relationship may be obscured by all the other variables.
When PCA is applied to spectral data the objects are the individual spectra, the independent
variables are the wavelengths or frequencies and the raw values (Xij) are the spectral intensities.
The raw data matrix is therefore:
ν1 ν2 … νm
spectrum 1 I1
1
I1
2
… I1m
spectrum 2 I2
1
I2
2
… I2m
spectrum 3 I3
1
I3
2
… I3m
: : : … :
spectrum n In
1
In
2
… Inm
In a typical experiment, the number of spectra (n) is generally less than twenty while the number
of wavenumbers (m) is generally several hundred.
Chapter 2 Experimental 52
When PCA is performed, this is transformed into the PC data matrix:
PC1 PC2 … PCm
spectrum 1 PC11 PC12 … PC1m
spectrum 2 PC21 PC22 … PC2m
spectrum 3 PC31 PC32 … PC3m
: : : … :
Spectrum n PCn1 PCn2 … PCnm
Where the PCs are described by the loading factors aij :
PCij = ai1.Ij1 + ai2.Ij2 + … + aim.Ijm
Theoretically, the data has not been reduced — it is still an n×m matrix. If the data is amenable
to PCA however, it will be adequately described by only the first few PCs.
If, for example, the data is adequately described by three PCs , the data matrix is effectively
reduced to n×3 in size.
PC1 PC2 PC3
spectrum 1 PC11 PC12 PC13
spectrum 2 PC21 PC22 PC23
spectrum 3 PC31 PC32 PC33
: : : :
spectrum n PCn1 PCn2 PCn3
Chapter 2 Experimental 53
A two dimensional PCA plot may be made of one PC versus another in which each spectrum is
represented by a single point. Information about the relationships between spectra may be seen
in the groupings on the PCA plot. When a particular PC is found to separate two classes of
spectra, the loading spectra for that PC may be examined to see in which wavenumber region
these classes differ.
SIMCA
Soft Independent Modelling of Class Analogies (SIMCA) is a technique for the supervised
classification of PC data. In SIMCA, a number of object classes are defined and the PC scores
for each object are tested to determine how well they fit each class.
The degree to which a class is defined is given by the Residual Standard Deviation (RSD)
which, is defined as:
∑= −−
=cN
1n c
2n[c]
[c] 1PNε
RSD 3.2
where: Nc is the number of classes
P is the number of PCs in the model
and εn[c] is the error for object n in class c
A small value of RSD indicates a tightly clustered class. Depending on the level of confidence,
a critical RSD value (RSDcrit) may be set to define the boundary of the model:
SDF.RSDcrit = 3.3
Chapter 2 Experimental 54
where: F is the critical F value
and SD is the mean standard deviation.
An object may be said to belong to a class if, when it is fitted into that class, the RSD value of
the class is less than RSDcrit.
One class (C) can be fitted into another class (D) and the RSD[C|D] calculated as:
∑ == CN
1nC
2n[D]
D]|[C Nε
RSD 3.4
This allows the distance between two classes (dCD) to be calculated:
1RSDRSDRSDRSD
d 2[D]
2[C]
2[D|C]
2D]|[C
CD −+
+= 3.5
Fuzzy Clustering
The fuzzy clustering technique is an unsupervised analysis method. In this technique the user
does not assign the objects to any class but rather the fuzzy clustering technique designates each
object with a degree of membership to a particular class. The number of classes available is
determined by the user. A membership value between zero and one is then assigned to each
object for any given cluster with a higher value indicating a high degree of similarity between
the object and the class. The sum of the membership values for each object across the whole all
of classes designated equals one. An object with close to equal membership values in each class
is referred to as a fuzzy object.
Chapter 2 Experimental 55
Quantitative Analysis
The simplest (univariate) way to calculate concentration spectroscopically is to make a Beer’s
law plot of absorbance versus concentration for a calibration set and then fit a regression line to
the data using a linear least squares method.
This technique can be extended to multiple analyte components and multiple frequencies and is
known as the multivariate Classical Least Squares (CLS) method.
When there are n samples containing l components and n frequencies in the spectra, the
absorbance may be written as an m×n matrix (A) and the concentrations as an m×l matrix (c).
Beer’s law is therefore:
A = cK + EA 3.6
where K is an l×n matrix of pure-component spectra
and EA is an m×n matrix of spectral noise.
The estimated component spectra ( K ) may be found from the calibration set by:
ACCCK ′′= −1)(ˆ 3.7
and the estimated concentration ( c ) may then be calculated for the unknown spectra (a) as:
aKKKc ˆ)ˆˆ(ˆ 1−′= 3.8
While CLS represents a great improvement over univariate methods7,8,9,10, it can fail to take
account of nonlinearities and baseline effects.
Chapter 2 Experimental 56
Many of these shortcomings can be solved by the techniques of Principal Component Regression
(PCR) and Partial Least Squares regression (PLS) in which the raw data is first converted to
principal components and the PC scores (T) are fitted to the concentrations. ie:
veTvc += 3.9
where v is a coefficient relating concentration and principal component scores. In PLS the
coefficient v is analogous to the pure component spectra k in CLS but v is an “abstract spectra”
which may have no obvious chemical significance. The v spectra in PLS may reflect baseline
variations, instrumental anomalies or non-linear chemical effects as well as analyte absorbances.
Although PLS yields generally poorer qualitative information than CLS, it produces superior
quantitative measurements.
Because the PC coefficients (v) have no obvious physical interpretation — unlike CLS k factors
which are component spectra — the problem arises of how may factors should be included in a
PLS analysis. Too few factors would fail to account for the systematic variation; too many
would overfit the data and produce spurious effects by fitting random error.
The optimal number of factors for any calibration set can be determined by the process of cross-
validation11,12,13 in which the PLS analysis is conducted with different numbers of factors to find
the model with minimal prediction error.
A model is cross validated by systematically removing one or more of the objects from the
calibration set and then measuring how well the removed values are predicted using the
remaining calibration set. This process is repeated until all of the elements of the calibration set
have been removed at least once.
Chapter 2 Experimental 57
Both for validation and for estimation of the error of the final concentration measurement, the
Standard Error of Estimation (SEE) is required for the calibration set and the Standard Error of
Prediction (SEP) is required for the validation set. These may be calculated for centred data
by14:
∑= −−
−=
n
1i
2ii
1n)cc(SEE
df 3.9
and
∑= −
−=
n
1i
2ii
1n)cc(SEP 3.10
where ic is the predicted concentration of sample i ci is the actual concentration of component i n is the number of samples and df is the number of factors used in the analysis (ie. the degree of freedom)
Chapter 2 Experimental 58
2.5 References:
1 U. Meyer, Private notes, Queensland University of Technology, 1990. 2 Grams 32, Version 4.01, level 1,1996, Galactic Industries Corp., New Hampshire, USA 3 Pattern Recognition Systems AS, 1998, Bergen, Norway. 4 O.V. Kvaleim, “Pretreatment of Multivariant Data” in “SIRIUS: A Program for Multivariant Calibration
and Classification”, T.V. Karstang, O.M. Kvalheim (ed.), (Pattern Regognition Systems, Norway, 1990) 5 A. Thilemans, D. L. Massart, Chimica, 39, 1985, 236-242. 6 O.M. Kvalheim, SIRIUS (version 2.3); Department of Chemistry, University of Bergen, Norway 7 D.M. Haarland, Easterling, R.G.; Appl. Spectrosc., 34, 1980, 539 8 D.M. Haarland, Easterling, R.G., Vopicka, D.A.; Appl. Spectrosc. 39, 1985, 73 9 M.A. Sharaf, D.L. Illman, B.R. Kowalski; “Chemometrics”, (1986), Wiley, N.Y. 10 S.N. Deeming, S.L. Morgan; “Experimental Design: a Chemometric Approach” (Elsevier, New York,
Figure 3.1 FT-IR PA spectra of undyed Merino wool fabric purged with ultra-pure Helium gas for: A) 5minutes (top spectrum), B) 10 minutes (middle spectrum) and C) 15 minutes (bottom spectrum). (Resolution= 8.0 cm-1, OPD velocity= 0.2 cm S-1, No. of Scans= 128)
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.2 FT-IR PA spectrum of undyed wool fabric.
Amide I Amide II CH deformation
Amide III ν(S-O)
Chapter 3 FT-IR Spectroscopy of Dyed Wool 63
Considering the advantages of PAS over the other IR techniques, in particular its
ability in depth profiling allows studying the dye-fibre system in situ; i.e. in the cortex,
where the dye molecules migrate upon dyeing the fibre. Therefore, it was decided to
use PAS to examine the dyed fabrics in this project.
Therefore, a mirror velocity of 0.2cm.s-1 was chosen. This, according to the
Rosencwaig equation23, assuming that the thermal conductivity of wool24 is 9.0531 X
10-5 cal °C-1 s-1 cm-1, corresponds to a thermal depth of 2 and 4µm at 1800 and
800cm-1 respectively. As the thickness of the cuticle layer in Merino wool is
estimated24 to be between 0.5 and 1.0 µm, it is concluded that the information
obtained in these studies was from the cortical area of the fibre.
3.2 FT-IR PA spectroscopy of undyed Wool
In this project, the samples were purged with helium for different time intervals in
order to find the optimum purging time of the sample. Figure 3.1 shows the spectra
for a simple-weave, undyed Merino wool fabric. The sample was purged for different
time intervals, namely five, ten and fifteen minutes respectively. As can be seen, the
spectrum obtained after five minutes purging is very noisy and of lower quality. In
contrast, the spectra purged for ten and fifteen minutes show minimal difference in
quality. Therefore, it is more economical and time effective to purge the sample for
only ten minutes.
Figure 3.2 shows a typical spectrum of 21µm Merino wool fabric. Visual comparison
of this spectrum with previously reported spectra24, 25 shows that they are very similar.
The major peaks for this spectrum are listed with their assignments in Table 3.1. Also
Chapter 3 FT-IR Spectroscopy of Dyed Wool 64
in this table, are the peak positions found in previous studies24,25,26 for comparison
purposes. All the major peaks in this spectrum have been reported previously.
However there are a few differences found between the spectra obtained in these
studies and one of the published spectra24; for example, in the spectrum recorded by
Carter, the peaks near 1400 and 1300cm-1 are not found. These peaks have been
assigned to ν(coo-) and νs-o bands of cystine dioxide respectively. Other published
spectra 25 however, do show these peaks. The peak at 1720 cm-1, is assigned by
Carter24 to ν (coo-). In the spectrum acquired here, this peak may be obscured by the
dominant Amide I band, which is centred at 1667 cm-1 and has a shoulder at around
1720 cm-1. The peaks for cysteine-S-sulfonate present in Carter’s spectrum are
missing from Figure 3.2.
The lack of these peaks and the presence of the peaks for cystine monoxide and
cystine dioxide indicates that probably the surface of the fabrics studied here have
been slightly oxidised.
Chemometrics
The data were then studied qualitatively using chemometrics method of analysis in
order to investigate the regions of the spectrum where the dyed and undyed samples
differ from each other. This was done with the aid of PCA. The pre-treatment of all of
the data discussed for woollen samples in this chapter, unless stated otherwise,
consisted of normalising in MS-Excel before being submitted to SIRIUS 6.0 for
chemometric analysis.
A matrix was built consisting of 19 objects and 251 variables. The objects considered
were repeat spectra of undyed wool along with the spectra of wool dyed with a
mixture of two colours, i.e. 073, 370 and 703.
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.3 PCA and loading plot of spectra of undyed and dyed wool samples
dyed with Lanasol dyes
Comp. 1 (62.0%)
Comp. 2 (16.6%)
-2.8 -1.1 0.6 2.3 4.0 *10-3
-4.1
-2.4
-0.7
1.0
2.8 *10 -3
**** *
*
* ** ** **
*
** ***
Undyed wool
073
703
370
-0.3-0.25-0.2
-0.15-0.1
-0.050
0.050.1
0.150.2
1800
1756
1712
1668
1624
1580
1536
1492
1448
1404
1360
1316
1272
1228
1184
1140
1096
1052
1008 964
920
876
832
Loadings comp. 2
Loadings comp. 1
Loadings comp. 2
Chapter 3 FT-IR Spectroscopy of Dyed Wool 65
Three components were significant to explain 83% of the variance, with PC1
separating undyed wool from the dyed ones. See Figure 3.3. The loading plot of PC1
indicates that bands in the region of 1720-1380 cm-1 have greater contribution to the
undyed samples; whereas the spectral region of 1308-828 cm-1 has a greater
contribution to the dyed samples, with the peak at around 1020 cm-1 having the
greatest weight (probably corresponding to νs (S-O) at 1024 cm-1).
This indicates that the S-O bond in the oxidized cysteine is more strongly affected by
the dyeing process than amide I and II.
3.3 FT-IR Studies of dyed wool
Introduction
Dyed wool has been studied extensively using various analytical techniques,
especially UV-VIS spectroscopy. In this method the dye is normally extracted into a
solution that can then be studied spectroscopically. The fabric may also be Soxhlet-
extracted to remove any unbound lipids from the surface of the fibre. Papini27 has
studied the monochromatic reflectance, transmittance and absorbance for the solar
spectra for undyed as well as dyed wool and cotton fabrics; and found that the dye
does not influence the NIR spectra, while it does influence the UV-VIS spectrum. It
was shown that the influence of the dye on the UV-VIS spectrum is greater for the
cotton than for the wool fabric.
Church et al.28 have studied Lanasol dyes, the general formulas of which are shown in
Figure 3.4. They have concluded that lysine, cysteine and histidine are the reaction
FT-IR Spectroscopy of Dyed Wool Chapter 3
Table 3.2 Table of peak positions and relative intensities in IR vibrational mode
for undyed wool and wool dyed with Lanasol dyes (Note: str. = strong, vw = very weak, sh = shoulder, shp = sharp, br = broad, m = medium)
Figure 3.4 The Lanasol dye containing chromophore attached to α,β-
dibromopropionylamido or α-monobromoacrylamido groups respectively.
NH
N
amino-acid
O
Chromophore
aziridine ring
Figure 3.5 The Lanasol dye product with lysine.
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.6 FT-IR PAS spectra of wool samples undyed and dyed with
different ratios of Lanasol dyes.
10
20
30
40
50
60
70
1800 1600 1400 1200 1000
Wavenumber (cm-1)
730
307
073
000 (Undyed)
343
800
Regions where spectra show differences.
Chapter 3 FT-IR Spectroscopy of Dyed Wool 66
sites and observed that the other potential sites of reaction, namely tryptophan,
tyrosine or serine, were not involved at all. They also confirmed previous studies
indicating that the dye forms an aziridine ring with lysine side chain groups as
illustrated in Figure 3.5.
In the studies conducted here, PAS was used to study dyed woollen fabrics with the
aid of chemometrics. This was done with a view to examine the dyes on the samples,
both qualitatively and quantitatively.
PAS Analysis of Woollen Fabrics Dyed with Lanasol Dyes
Figure 3.6 shows PAS spectra of undyed wool fabric (i.e. 000) along with the 343,
073, 307 and 730 dyed samples. Table 3.2 lists the main peaks found. Unlike the
FT-Raman spectra of the same samples (Cf. Chapter 4) no peaks were observed that
could be assigned exclusively to the dye molecules. This might be due to the low total
percentage of dye on the fibre (~2-3%) or to the greater IR activity of the wool
groups. Visual comparison of these spectra indicates that there are some minor
differences between the undyed wool spectrum and those of the dyed ones. Apart
from these minor differences however, keratin peaks dominated the spectrum. The
changes are particularly obvious in the region of 1200-900 cm-1. This area is marked
in Figure 3.6 with arrows. The main wool keratin peaks are present and can be
assigned. It must be noted that due to saturation, it was anticipated that the peaks for
Amides I and II, at around 1657 and 1547cm-1, would not be very informative.
Chapter 3 FT-IR Spectroscopy of Dyed Wool 67
Therefore, apart from acknowledging their correct position on the spectrum they were
disregarded. Changes were, however, observed in the 1400-800 cm-1 region.
The spectra were next submitted to chemometrics for further analysis, both
qualitatively and quantitatively.
3.4 Chemometrics studies of dyed wool
Chemometrics has been used to interpret the spectral data obtained using various
analytical methods. It is especially useful when there are no obvious differences
between the spectra obtained for quite similar samples. Chemometrics has been
applied to DRIFT spectra of dye mixtures extracted from a polyester/cotton shirt29.
Raw data and pre-treated spectral data matrices were studied using PCA, SIMCA and
Fuzzy Clustering. It was concluded that in PCA the objects cluster according to their
position on the garment, therefore offering more information than previously used
methods such as TLC.
Kokot and co-workers30 have demonstrated that DRIFT combined with chemometrics
can classify samples not only according to the quality of the cotton fabrics but also by
its stages in a processing sequence. In 1994, the same authors31 reported the
possibility of the application of chemometrics to DRIFT spectra of the dye mixtures
of worn clothing. In these studies the dyes were extracted from the sample. It was
shown that, by interpreting the spectral data using PCA and SIMCA, it is possible to
discriminate and match unknown dye mixtures extracted from microscopic samples of
worn fabrics as well as unknown dye mixture extracted from the same fabric in a
wash and wear situation.
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.7- PCA plot of samples with a mixture of two and three dyes
Note: Brown = 046, Light Orange = 073, Sea Green = 208, Lime = 307, Pink = 550, Rose = 703, Grey 40% = 343, plum = 127, Indigo = 154, Aqua = 181, Sea green = 208, Violet = 213, Blue-Grey = 217, Red = 235, Dark Red = 244, Dark Green = 361, Olive Green = 424, Dark Yellow = 433, Sky Blue = 532, yellow = 613, Pale blue = 811.
Comp. 1 (49.9%)
Comp. 2 (19.9%)
-3.4 -1.7 -0.0 1.7 3.3*10-3-3.1
-1.4
0.3
2.0
3.7 *10
-3
**
** **
**
*
*
**370a
*
** ** * *
*
**
*
**
**
**
235b**235e*
*
244b
** *
***
**
*
*
*
**433d
*
** *424c
* *
*
**
*
***
* *** *
*
*
***
*** *
**
*
*
***
*
*
**
*
**
**
**
*
*
*
****
*
***
***** *
*
**
**
*
*
**
*
*
**
**
*
**
*
*
****
***
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.8 PCA plot of samples with a mixture of two dyes
Note: Brown = 046, Light Orange = 073, Sea Green = 208, Lime = 307, Pink = 550, Rose = 703
Comp. 1 (50.6%)
Comp. 2 (19.1%)
-2.0 -1.0 0.0 1.0 2.0*10
-3-2.0
-1.0
0.0
1.0
2.0 *10 -3
**
*
*
**
**
*
**
**
*
*
*
*
*
*
* **
*
* **
*
***
*
**
*
073
046
208
307
550
703
Figure 3.9- PCA plot of samples with a selection of samples Note: Light Orange = 073, Lime = 307, Rose = 703, Grey 40% = 343, Aqua = 181, Pale Blue =
811, Blue- Grey = 217.
Comp. 1 (62.0%)
Comp. 2 (16.6%)
-2.2 -1.0 0.2 1.4 2.6*10
-3-2.1
-0.9
0.3
1.5
2.7 *10
-3
*
****
**
*
*
** * * *
*
*
**
*
**
*
*
**
*
*
*
**
*
*
* **
*
* **
*
*
073307
811
181
703
217 343
Chapter 3 FT-IR Spectroscopy of Dyed Wool 68
In view of these results the spectral data were submitted to PCA for pattern
recognition followed by PLS for the prediction of the ratio of dyes. The advantage of
this procedure to the previous studies mentioned above is that the dyes were examined
in situ. It is anticipated that by doing so more information about the dye-wool fibre
may be obtained, and that it would eliminate the lengthy procedure of dye extraction.
PCA
A data matrix consisting of repeated PA spectra of all the samples studied here was
submitted to PCA. The matrix consisted of 141 objects and 151 variables. Three
components accounted for 76% of the total variance. Figure 3.7 shows PC1 versus
PC2 plot for these data, indicating that there are too many data points to be
conclusive.
Therefore, smaller groups of samples were considered in PCA analysis.
Next, a matrix consisting of 34 objects and 151 variables was built using samples
containing only two colours. The data matrix was then analysed with PCA. 75% of
the total variance was explained using three components. Figure 3.8 shows the plot of
PC1 versus PC2. As it can be seen in this figure, even though the repeated spectra of
the same sample have not clustered together tightly, a general pattern is observed with
respect to the amount of red dye. Starting with zero red dye on the right-bottom of the
plot and moving anti-clock wise.
Another group considered was with a mixture of two or three dyes present on the
sample. The samples chosen for this purpose were 073, 181, 217, 307, 343, 730, 811.
FT-IR Spectroscopy of Dyed Wool Chapter 3
Table 3.3: Predicted and Measured Values for the Validation Sets
(For Dependent Variables Red, Blue and Yellow respectively)
Some loss of information is observed as peak saturation is observed which may be
due to the OPD velocity used.
Visual comparison of the spectra reveals minor differences between them. See Figure
3.13. Thus they were submitted to chemometrics for further analysis.
Various spectral data pre-treatments were tried on this reduced data set in order to
find the one that offered the best separation of the samples in PCA. For example,
Figure 3.14 shows PCA plots obtained using first derivative spectra (i) and non-
derivative spectra (ii). Both of these sets have then been block normalised and
Y-mean centred. As it can be seen non-derivative spectra offer a much clearer PCA
plot.
Hence, the best pre-treatment for this set of data was decided to be non-derivative
spectra block normalised followed by Y-mean centring.
Subsequently, all objects and 251 variables were submitted to PCA for analysis. Three
components were significant to explain 90.5% of the variance in the data. Figure
3.14(ii) shows PC1 versus PC2. PC1 explains 65.8% of the total variance and
separates yellow and green colours (positive) from brown and grey colours (negative).
PC2 and PC3 do not separate the samples any further.
Spectral data (pre-treated in the same manner) for colours red and orange were then
introduced to the same set of data. More data for yellow, brown, grey and green were
added as well. These data were pre-treated in the same manner as the existing data.
The PCA plot is shown in Figure 3.15. 93 objects and 251 variables were introduced
FT-IR Spectroscopy of Dyed Wool Chapter 3
Figure 3.15: PCA plot of individual dyes red, yellow, orange, green, grey and brown on wool/polyester fabric samples.
(Note: The colour of the stars in the plot corresponds directly to the colour on the fabrics)
Com p. 1 (51.9%)
Com p. 2 (35.0%)
-0.84 -0.38 0.08 0.55 1.01*10-2-1.33
-0.87
-0.41
0.05
0.51*10 -2
****
*****
* *
*
** * ** **** **
*
*
* *
*
* * *
*
**
* *
** ****
***
***
*
*
*
*
**
* ** ****
*
**
*
*
*** *
***
***
** *
***
** *
* * ** *** *
Chapter 3 FT-IR Spectroscopy of Dyed Wool 74
to PCA. Three components were significant to explain 93% of the total variances in
amongst the data. As it may be seen in this figure, there is a poor separation of colours
in this plot.
In summary, PA spectra of woollen fabrics dyed with Lanasol dyes and wool/
polyester fabrics dyed with Forosyn dyes were recorded and studied qualitatively
(PCA) and quantitatively (PLS).
The results show that PCA separated the woollen samples dyed with two colours
according to their position on the colour card. Samples dyed with three dyes however,
do not separate that well and although spectral repeats still clustered together, they do
not do so in any order. On the other hand, wool blend fabrics spectra were displayed
on the PCA plot totally in a random fashion.
PLS was performed on the spectral data of dyed woollen fabrics. The calculations
indicate that the results obtained are promising. They do however need to be studied
further with a much larger calibration set.
These samples were then examined further using FT-Raman spectroscopy and
chemometrics, in order to investigate if this combination would offer better solution
to the problem of colour matching.
Chapter 3 FT-IR Spectroscopy of Dyed Wool 75
3.6 References
1 N.M. Morris, R.A. Pittman and R.J. Berni; ‘ Fourier Transform Infrared Analysis of Textiles’, 16 (2), 1984, 30-44. 2 M.P. Fuller and P.R. Griffiths; Applied Spectroscopy, 34 (5), 1980, 533-538. 3 R.S. Davidson and D. King; J. Text. Inst., 74 (6), 1983, 382-384. 4 A. Rosencwaig, ‘Photo-acoustics and Photo-acoustic Spectroscopy’, (Interscience, New York, 1980). 5 J.S. Church and K.R. Millington; Biospectroscopy, 2, 1996, 249-258. 6 R.S. Davidson, D. King, P.A. Duffield and D.M. Lewis; J. Soc. Dyer Color., 99, 1983, 123-
126. 7 E.A. Carter, P.M. Fredericks, J.S. Church; Textile Res. J., 66, 1996, 787. 8 J.S. Church, W-H. Leong; ‘ The Analysis of Wool Textile Blends by FT-IR PAS and FT-
Raman Spectroscopies’, 9th International Wool Textile Research Conference, Biella, Italy, IV (1995), p.p. 114-122.
9 R.S. Davidson and G.V. Fraser; J. Soc. Dyer Color., 100, 1984, 167-170. 10 L.E. Jurdana, K.P. Ghiggino, I.H. Leaver, C.G. Barraclough, P. Coleclake; Applied
Spectroscopy, 48 (1), 1994, 44-49. 11 J.F. McClelland &R.W. Jones, S. Luo & L.M. Seaveron; ‘A practical guide to FTIR
photoacoustic spectroscopy’, March 1992. Reprint of a chapter to appear in Proper sample handling with today’s IR instruments. Edited by P. Coleman and published by CRC press.
12 J.S. Church and A.L. Woodhead; ‘The Vibratinal Spectroscopic Analysis of Wool/ Polyester Textile Blends’, XXX CSI (1997), Melbourne, p.p. PS54. 13 J.S. Church and D.J. Evans; J. of Applied Polymer Science, 57 (13), 1995, 1585-1594. 14 D.C. Jones, C.M. Carr, W.D. Cooke and D.M. Lewis; Textile Res. J., 68 (10), 1998, 739-748. 15 K. Krishnan, ‘Application of the FT-IR Microsampling Techniques to Some Polymer Systems’, Proceedings fro SPIE the international Society for Optical Techniques for
Industrial Inspection, 665, 1986, p.p.252-257. 16 J.M. Chalmers and N.J. Everall; ‘FT-IR Microscopy Advances in Techniques for Characterising and Structure-Property Elucidations of Industrial Materials and Chemicals’, XXX CSI (1997), Melbourne, p.p. I25. 17 D.J. Gerson and C.A. Chess, Practical Spectroscopy, 6, 1988, 73-83. 18 D. Wielbo and I.R. Tebbett, Journal of Forensic Science, JFSCA, 37 (4), 1992, 1134-1148. 19 H.A. Harris and T. Kane, Journal of Forensic Science, JFSCA, 36 (4), 1991, 1186-1191. 20 P.L. Lang, J.E. Katon, J.F. O’Keefe and D.W.Schiering, Microchemical Journal, 34, 1986,
361-366. 23 A. Rosencwaig,; ‘Photo Acoustic and P.A.S.’ (John Wiley & Sons, New York; 1980), pp. 94-
124 24 E. Carter, ‘Vibrational spectroscopic Studies of Wool’, PhD thesis, Queensland University of
Technology, 1998. 25 M.A. Moharram, T.Z. Abdel-Rehim, S.M. Rabie, J. App. Polym. Sci., 26, 1981, 921-932 26 D.M. Lewis in ‘Wool Dyeing’, D.M. Lewis (ed) 27 M. Papini; Infrared Phys., 29 (1), 1989, 133-137. 28 A.S. Davie, J.S. Church, P.J. Scammells and D.J. Tucker; ‘A Spectroscopic Analysis of the Dibromopropionyl/ Bromoacryl Dyes With Wool’, XXX Colloquium Spectroscopium Internationale (1997), Melbourne, p.p. C11. 29 S. Kokot, S. Carswell and D.L. Massart; Applied Spectroscopy, 46 (9), 1992, 1393-1399. 30 C.Gilbert, S. Kokot and U. Meyer; Applied Spectroscopy, 47 (6), 1993, 741-747. 31 S. Kokot and C. Gilbert; Analyst, 119, 1994, 671-675.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 76
Chapter 4 FT-Raman Spectroscopy
of Dyed Wool
4.1 Introduction
Vibrational spectroscopy has played a significant role in both structural determination
and compositional analysis of various materials. Fourier Transform Raman
Spectroscopy has been used in many studies of biological tissues. It has been shown1
that good quality spectra of Human skin, nail, hair and calluses can be obtained with
this method. Human keratotic biopolymers have been studied using FT-Raman
spectroscopy. It has been concluded that despite their functional differences, these
samples show molecular similarities.
FT-Raman spectroscopy has also been used in studying natural fibres such as cotton.
Kokot et al.2 have shown that various Australian cotton fibres may be identified using
FT-Raman and chemometrics. FT-Raman microscopy has been used3 to study the
fixed dye on the thread of a jean garment. It has been shown that the spectrum of the
dye may be observed with no chemical pre-treatment of the sample. In a related study
of reactive dyes on cotton fabric4 it was concluded that PCA could discriminate
between samples with four different dye states, as well as the individual dye
concentration subgroups. PLS analysis was also applied to predict the concentration
of the unfixed dye on the fabric.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 77
Although much useful information may be obtained from wool fibres using
vibrational spectroscopy, traditional infrared techniques have been limited by several
practical problems5 which may be overcome by using Raman spectroscopy in the
visible or near-infrared region. In many samples visible laser excitation produces a
fluorescence signal that mask the Raman spectrum. Near-infrared excitation also
reduces the problem of fluorescence.
4.2 Dyed Wool For the past three decades keratins such as wool have been studied using vibrational
spectroscopy methods. In 1972, Raman spectroscopy was used to study the
interaction of mercuric chloride with wool6,7. In 1976, the Amide I and Amide III
band regions in unordered polypeptide chains present in feather keratin were studied8
with Raman spectroscopy, as were structural changes in the wool fibre after
annealing9. The effect of stretching on the conformation of the peptide backbone of
the wool fibre, pre-treated with sulphite, has also been investigated10 with FT-Raman
spectroscopy.
In 1994, the structure of untreated and hydrogen peroxide bleached wool samples
were studied11. The authors in this paper claim it was the first FT-Raman study on the
structure of wool. Rapid, good quality FT-Raman spectra of bundled and woven
scoured merino wool were recorded. The spectrum of a single wool fibre was
measured using an FT-Raman microscope. The authors concluded that the highest
quality spectra were obtained from the wool fabric, but there were no major
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 78
spectroscopic differences between this and the spectra obtained from the bundle of the
wool fibres.
The conditions of the instrument used in the above study are summarised in Table 4.1.
These instrumental conditions are important in order to be able to compare various
studies of the same nature.
Table 4.1 Instrumental conditions used in Hoggs11, Carter12 and this study
In the same year, untreated wool and wool in different stages of shrink-proofing
treatment was studied using FT-Raman spectroscopy12. The instrumental conditions
used in this study are also noted in Table 4.1. Two objectives were considered here:
to find the optimum conditions for FT-Raman spectral acquisition of wool in various
presentations and to study the stability of the wool sample under laser irradiation. This
was achieved by assessing the wool discolouration and by observing any spectral
changes by varying the laser power. The following conclusions were reached:
1. No damage to the wool samples was observed up to the maximum laser power
Phe and Trp 1002 1006 1003 Aromatic ring breathing mode
955 (m) CH2 rock 952 959 N.R. N.R. 931 (m)
Skeletal C-C stretch (α)
934 935 N.R. N.R.
N.O. Skeletal C-C stretch (α)
905 N.O. N.R. N.R.
897 (m)
Skeletal C-C stretch (α)
897 N.O. N.R. N.R.
876 (w, sh) Trp 881 883 N.R. N.R. 851 (s) Tyr 851 852 N.R. N.R. 826 (m, v br, broadened between 835-826cm-1)
Tyr 828 835 N.R. N.R.
801 (w) 801 811 N.R. N.R.
Notes:
N.O. (Not Observed): the reference does not acknowledge the peak at all N.M.S. (Not Mentioned Specifically): there is a reference to the range where the peak is found but it has not been mentioned and assigned individually. N.R. (Not Reported): the range in which that peak was to be found has not been discussed at all by the authors.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 79
2. Optimum quality spectra for various sample holders were obtained using different
conditions. For example, the best spectrum for a dense plug of dry wool fibres in a
glass cuvette was obtained using the laser power of 200mW with 500 scans while
frame held fibres require 300mW and 1000 scans.
Their band frequencies and assignments are also shown in Table 4.2.
Reactive dyes form a covalent bond with the wool fibre substrate, giving rise to the
high wet fastness of these dyes13,14,15. The reactive dyes available for dyeing wool can
be divided into two types: those reacting via nucleophilic substitution reactions and
those by reacting via Michael additions. Lanasol dyes, the most successful reactive
dyes, belong to the second type. The chemistry of reactive dyes has been explained
in chapter one, hence it suffices here to point out their general chemical structure:
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 80
NHCH2
Br
OD NH2peptide
NHCH2
NH
OD
peptide
NH NH
BrO
peptide
D
NH
NpeptideO
D
substitution
reaction
addition
reaction
NHpeptide
NHD
O NHpeptide
NH2 peptide
NHBr
OD
Br
H2O
HBr
+
+
Using model compounds consisting of simple chromophores attached to the reactive
sites of Lanasol dyes, it has been further shown16 that lysine, cysteine and histidine
are the amino acid providing reaction sites. Tryptophan, tyrosine and serine on the
other hand were not involved in the reaction of the wool fibre with the Lanasol dyes.
In this paper, the authors claim that they have observed the presence of an aziridine
ring in the product of the dye and the lysine amino acid, while they found no evidence
of two amino acids covalently binding to one dye molecule. The observations
challenge the previous assumption17 that the Lanasol dyes are bifunctional (ie. they
Chapter 4 FT-Raman Spectroscopy of Dyed Wool
Table 4.4 FT-Raman dye peaks found in each fabric dyed with two dyes (1800-400cm-1) Lee-Son and Hester have assigned all these bands as Dye On Fibre (D.O.F.)
Table 4.5 FT-Raman dye peaks found in selected fabrics dyed with three
Table 4.11: Dye peaks found in the FT-Raman spectra of dyed (45:55) wool/polyester Using a mixture of various colours after subtraction of the substrate spectrum
Figure 4.21: PCA Plot for Side 1, 4 and 5 of the Top Triangle in Diamond 1
D ataS et: tt2kn , S u b se t: a5 , S co res 1 vs 2
C o m p . 1 (94 .4% )
C o m p . 2 (3 .4% )
-2 .8 -1 .3 0 .2 1 .8 3 .3*10-2-3 .2
-1 .7
-0 .1
1 .4
3 .0 *10 -2
*********** ************** ***
*** *
** ******
************* ***
* * * * ** * * ** * * * * **
6 0 0 Y ellow
5 0 1
5 1 0 4 2 0
2 4 0
3 3 0 /4 0 2 3 0 3
0 6 0 B row n
1 5 0
0 5 1
2 0 4
0 4 2
0 3 3
0 2 4
1 0 5
0 1 5
0 0 6 G rey
Figure 4.22: PCA Plot For The Top Triangle In Diamond 1 The position of some of the samples (in Pink) located in the middle of the triangle are shown in relation to the ones seen in Figure 4.21.
DataSet: tt2kn, Subset: a7, Scores 1 vs 2
Comp. 1 (93.2%)
Comp. 2 (4.6%)
-2.8 -1.3 0.2 1.7 3.2 *10-2-3.0
-1.5
-0.0
1.5
2.9 *10 -2
* **********
*****************
*******
****
********* *******
****** **************
************
006 Grey 015
105
033051
303411
600 Yellow
231
321
312 501
510
330
240
150
420 060 Brown
****
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 107
Top Triangles of D1 and D2
PCA for Three Sides of Top Triangle of D2
A matrix consisting of 174 objects (sides 1,4 and 5) and 97 variables (1476-1092cm-1)
was analysed using PCA. The objects in this set include samples with both single dyes
and with mixtures of two dyes. Figure 4.20 shows PC1 vs. PC2 plot. A total of 99.0%
variance was explained by three components. PC1 explained 92.4% and PC2 6.3% of
the total variance.
The PCA analysis conducted here has been used as an indication of the location of the
samples only. As it is seen in Figure 4.20, there are too many samples and too closely
spaced for a complete separation of the samples into positive and negative
components.
The PCA plot however, clearly shows that the repeated spectra of each sample are
clustered together and that they are positioned regularly according to the ratio of each
of the three dyes.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 108
PCA for Top Triangle of D1
A matrix of 72 objects and 126 variables representing three sides was submitted to
PCA for analysis. The results are shown in Figure 4.21. PC1 explained 94.4% of the
variance while PC2 explained 3.4%.
Overall, the samples are separated and positioned according to their location on the
diamonds, with the three main colours well separated. The samples with a
combination of two dyes are clustered on the PCA plot, in the appropriate locations
between their component colour.
In Figure 4.22 a few of the samples with a combination of three dyes have been
included in the matrix composed and analysed in Figure 4.21. Again the plot is an aid
to view the position of these samples in the context of the triangle studied before.
From Figure 4.21 it can be concluded that the samples are located in the right position
according to their dye ratios. For example, sample 231 is situated before sample 240.
Samples 312 and 411 are also correctly positioned before samples 330 and 420
respectively.
Therefore, the overall conclusion is that PCA plots can separate the samples in
accordance to their dye ratios of one, two or three dyes on the substrate thus allowing
the data to be studied qualitatively.
The data were then analysed quantitatively by Partial Least Squares.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool
Table 4.12: Results of PLS Model for the prediction of brown dependant
variable in the regions 1476-1092cm-1 and 1800-800cm-1 Name Predicted
(1476-1092cm-1)
Predicted
(1800-800cm-1)
Measured
303B212 2.7 2.9 3.00
303B213 3.21 3.2 3.00
303B215 3.2 3.1 3.00
303B211 2.8 3.1 3.00
033B221 -0.10 0.17 0.00
033B223 -0.29 -0.17 0.00
033B224 0.13 0.13 0.00
033B225 0.03 -0.00 0.00
033B226 -0.28 -0.09 0.00
330Y11 2.3 2.9 3.00
330Y12 2.6 3.0 3.00
330Y13 2.2 2.5 3.00
330Y14 2.2 2.6 3.00
330Y15 2.2 2.7 3.00
330Y16 2.0 2.7 3.00
Chapter 4 FT-Raman Spectroscopy of Dyed Wool
Figure 4.23: Predicted vs. Measured values for (i) calibration and (ii) validation sets for Sides 1, 4 and 5 of Diamond 2 respectively, dependent variable: Brown.
Two PLS1 models were built for diamonds 1 and 2 (see Appendix 1), using the sides
of the top triangle of diamonds 1 and 2 respectively. Another PLS1 model was built
using all of the points in the top triangle of diamond 1. Two ranges of independent
variables were studied for each: the whole spectrum (1800-800cm-1) and the dye
region (1500-1000cm-1).
Sides 1, 4 and 5 of Diamond 2
Validation Set: 033, 303, and 330
Dependent Variable: Brown, Yellow, Red
Independent Variable: 1800-800cm-1
PLS Analysis for the Brown Dependent Variable
A PLS model consisting of 154 objects in the calibration set was built.
These samples were then analysed with the model and a cross-validation was
performed again. Eight factors were found to be significant and explained 99.46% of
the independent variables and 98.18% of the dependant variables. The last three
factors however, explained only 4.02% of the total variance of the dependent
variables. Thus, the model consisted of 5 factors explaining ~94% variance.
A validation set consisting of 15 objects was introduced to this model. The plots of
measured versus predicted values for (i) the calibration and (ii) the validation sets are
shown in Figure 4.23. The values for the slope, intercept and correlation coefficient
show that there is a good correlation between the objects on the calibration set and
that this set may be used with confidence. The results of the prediction for the
validation set are tabulated in Table 4.12.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 110
SEE and SEP values were also calculated to be 0.35 and 0.37 respectively.
Sides 1, 4 and 5 of Diamond 2
Validation Set: 033, 303, and 330
Dependent Variable: Brown, Yellow, Red
Independent Variable: 1476-1092cm-1 (dye region)
PLS Analysis for Dependent Variable Brown
Another PLS1 model was built using the same calibration and validation sets. The
independent variable region chosen however was 1476-1092cm-1, which is the region
of the spectrum containing the dye peaks. Five factors were found to be significant,
explaining 96.92% of the variance for the dependent variables. The results for the
prediction of the validation set are presented in Table 4.12. These show that the first
model has a better predictive ability and that the whole 1800-800cm-1 region should
be used as independent variables. This may be the result of interactions between the
dye molecule and the substrate. The SEE and SEP values were calculated to be 0.79
and 1.2 respectively.
Based on these results the rest of the models were constructed using the whole region
of the spectrum as independent variable.
Hence, the PLS model with independent variables 1800-800cm-1 for yellow and red as
dependent variables were also studied:
PLS Analysis for the Yellow Dependent Variable
Chapter 4 FT-Raman Spectroscopy of Dyed Wool
Table 4.13 Results of PLS model for Sides 1, 4 and 5 of Diamond 2 for the prediction of yellow and red dependant variable in the region 1800-800cm-1 for smaller validation set.
Name Predicted Measured Name Predicted Measured
303B212 0.01 (0) 0.00 303B212 3.31 (3) 3.00
303B213 -0.06 (0) 0.00 303B213 3.11 (3) 3.00
303B215 -0.04 (0) 0.00 303B215 2.96 (3) 3.00
303B211 0.037 (0) 0.00 303B211 3.45 (4) 3.00
033B221 3.43 (3) 3.00 033B221 3.16 (3) 3.00
033B223 3.30 (3) 3.00 033B223 3.28 (3) 3.00
033B224 2.98 (3) 3.00 033B224 2.99 (3) 3.00
033B225 3.42 (3) 3.00 033B225 3.01 (3) 3.00
033B226 3.37 (3) 3.00 033B226 3.18 (3) 3.00
330Y11 2.78 (3) 3.00 330Y11 0.72 (1) 0.00
330Y12 2.69 (3) 3.00 330Y12 0.65 (1) 0.00
330Y13 2.88 (3) 3.00 330Y13 0.76 (1) 0.00
330Y14 2.88 (3) 3.00 330Y14 0.63 (1) 0.00
330Y15 2.85 (3) 3.00 330Y15 0.85 (1) 0.00
330Y16 2.89 (3) 3.00 330Y16 0.97 (1) 0.00
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 111
Five components were significant for this model, explaining 98.01% of the total
variance of the dependent variable, with the first two explaining 94.24%. The plot of
predicted versus measured values gives a slope of 1.02, intercept of (-0.015) and a
correlation coefficient of 0.986. Thus indicating that the predicted values are
reasonably close to the measured ones. The results of the prediction are shown in
Table 4.13. SEE and SEP values for this model are 0.35 and 0.37 respectively.
PLS Analysis for the Red Dependent Variable
Six factors were significant here, with the first three explaining 94.83% of the total
variance of the dependent variable. The objects in the calibration set show a good
correlation indicating that this set may be used with confidence. The slope of the line
of best fit is found to be 0.989. This line intercepts the axis at 0.017.
When the validation set is introduced to this model however, the slope (0.799) and the
intercept (0.768) for the fitted line are not as favourable as for the calibration set. This
is reflected in the predicted values for the validation. SEE and SEP values for this
model were calculated to be 0.24 and 1.1 respectively. The high value for SEP
confirms that the predictive ability of this model for red dye is not good.
Since overall this model shows a reasonable ability to predict the values for the
dependent variables it was decided to expand the number of objects in the validation
The next model studied consisted of 72 objects in the calibration set and 102 in the
validation set. Six factors were significant explaining a total of 99.05% variance of the
dependent variable.
A cross-validation was performed on the calibration set. The values for the slope
(0.990), intercept (0.019) of the predicted versus measured plot and the correlation
coefficient (0.995) indicate that the set chosen is a suitable one. The validation set was
then introduced to this model. Here, the values for the slope (0.784), intercept (0.161)
and the correlation coefficient (0.973) indicate that there is a reasonable correlation
between these objects. See Appendix 2 (Figure A2-4). This is also confirmed by the
actual values obtained for the prediction of the dependent variable brown, shown in
Table 4.14. SEE and SEP values for this model are 0.19 and 0.83 respectively. The
value for SEP is considerably high. The calibration and validation sets studied here
were used for the dependent variables yellow and red.
PLS Analysis for the Yellow Dependent Variable
A model consisting of 72 objects in the calibration set and 102 in the validation set
was built. Five factors explained 99.24% of the total variance of this dependent
variable, with the first three explaining 98.13%. The calibration set showed a good
correlation between its objects. The correlation between the validation set, although
not as good as for the calibration set, is quite favourable, as is shown by the predicted
values for its objects. See Appendix 2, Figure A2-5 and Table 4.14. The SEE and SEP
values were calculated to be 0.17 and 0.65 respectively. The value for SEP also shows
that the prediction for this dependent variable yellow is better than the one for brown.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool
Table 4.14: Results of PLS Model for Sides 1, 4 and 5 of Diamond 2 for the prediction of dependant variables brown, yellow and red in the region 1800-800cm-1 with a larger validation set
these objects is not as good as the one built for samples with one and two dyes.
However, these results show that it is still a reasonable correlation. A validation set
consisting of samples 222, 321 and 411 was then submitted to this model for
prediction. Figure 4.25(ii) shows that the data for sample 222 in the validation set are
quite spread. The SEE and SEP values were calculated to be 0.44 and 0.51
respectively for this model.
The same calibration and validation sets were then used for dependent variables
brown and grey.
PLS Analysis for the Brown Dependent Variable
A calibration set containing the same number of objects for the calibration set and the
independent variables was built. The cross validation of the calibration set indicated
that there is a reasonable correlation between the objects in this set. The validation set
was then introduced into the model. Six factors were found to be significant
explaining 93.96% of the variance in the dependent variable. The last two factors
explained only 3.87% of the total variance. Again it was found that the correlation
between the objects in the model with brown dependent variable is not as good as the
one for the other variables. See Appendix 2, Figure A2-9 (I, II).
The values for the brown independent variable were then predicted using SIRIUS 6.0.
These values are displayed in Appendix 2, Table A2-6. SEE and SEP values
calculated for this model are 0.55 and 0.82 respectively. These values are certainly
very much higher than what has been calculated before.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 118
On introduction of the validation set, it was found that the correlation between these
objects is not very favourable. The values for the slope (1.156) and the intercept
(-0.094) for the plot of predicted versus measured values for the validation set
indicated that three of the spectral repeats for sample 222 might be outliers. Therefore,
they were taken out. This caused an improvement in the values for the slope (0.784)
and the intercept (0.278).
The predicted values for the objects in the validation set (after the outliers have been
taken out) are still not as accurate as the one for yellow however confirming that the
brown dye interferes with the predictive ability of the PLS1 model. See Appendix 2,
Figure A2-10 (III) and Table A2-6. In this model SEP was found to be 0.28. This is a
great improvement on SEP value calculated previously. This also confirms that the
three samples should be treated as outliers.
PLS Analysis for the Grey Dependent Variable
A calibration set consisting of the same number of objects was built. Cross validation
calculations on this set showed that there is a good correlation (0.989) between its
objects. Six factors were significant explaining a total of 97.74% of the variance for
the latent variable. The first three factors explained 95.33% of the total variance of the
grey dependent variable. The validation set consisting of 15 objects was then
introduced into this model. The values for the slope (0.517) and the intercept (0.655)
of the plot of predicted verses measured values indicated that there is not a good
correlation between these values. See Appendix 2, Figure A2-11. The actual predicted
values are shown in Table A2-7. The values for SEE and SEP are 0.32 and 0.45
respectively.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 119
Overall, the results indicate that the PLS models built using FT-Raman spectra of
these textile samples are a promising method of quantitatively measuring the ratio of
various dyes on each fabric. It is however necessary to study a larger population in
order to find out how robust this model is. Some dyes offer a better predictive ability
in these systems than others. This might be due to the fact that the dye peaks are less
intense because of the presence of fluorescence in those spectra. Therefore, it is
proposed to study a large population as well as to investigate a wider range of dyes
as dependent variables.
It is interesting to note that the PLS has worked quite well for the polyester blend, but
failed to do equally well for the dyed wool even though the dye peaks are easily
visible in the spectra of dyed wool without the aid of chemometrics. It may be that the
failure of the chemometrics to correlate the dye peaks to the amount of dye is due to
some, unknown, interaction between the different dyes. It is perhaps more likely that
the given dye values do not reflect the amount of dye actually on the fibres.
Chapter 4 FT-Raman Spectroscopy of Dyed Wool 120
4. 6 References
1 A.C. Williams, H.G.M. Edwards, W. Barry, J. Raman Spectroscopy, 25, 1994, 95-98. 2 Y. Liu, S. Kokot, T.J. Sambi, Analyst, 123, 1998, 633-636. 3 C. Coupry, G. Sagon, P. Gorguet-Ballesteros, J. Raman Spectroscopy, 28, 1997, 85-89. 4 S. Kokot, N.A. Tuan, L. Rintoul, Applied Spectroscopy, 51 (3), 1997, 387-395. 5 G. Lee-Son, R.E. Hester, J. Soc. Dyer Color., 106, 1990, 59. 6 R.H. Fish, J.R. Scherer, E.C. Marshall, S. Kint, Chemosphere, 6, 1972, 267-272. 7 V.J.C. Lin, J.L. Koenig in ‘ Advances in Infrared and Raman Spectroscopy’, R.J.H. Clark &
R.E. Hester (ed), 1, (Chichester, West Sussex, John Wiley, 1975), p. 35. 8 S.L. Hsu, W.H. Moore, S. Krimm, H. Randall, Biopolymers, 15, 1976, 1513-1528. 9 R. Shishoo, M. Lundell, J. Polymer Science (Polymer Chemistry Ed.), 14, 1976, 2535-2544. 10 J.S. Church, G.L. Corino, A.L. Woodhead, J. Molecular Structure, 440, 1998, 15-23. 11 L.J. Hogg, H.G.M. Edwards, D.W. Farwell, A.T. Peters, J. Soc. Dyer Color, 110, 1994, 196-
1927-1936. 13 D.M. Lewis, S.M. Smith, Proc. 8th Internat. Wool Tex. Res. Conf. (Christchurch) 1990 14 L.N. Jones, D.E. Rivett, D.J. Tucker in ‘Handbook of Fibre Chemistry’, Edited by M. Lewin,
E.M. Pearce, (Marcel Dekker Inc., New York), 2nd ed. 1998, p. 366. 15 G. Blankenburg, K. Laugs, A. Thiessen, Textilveredlung, 24, 1989, 10. 16 A.S. Davie, J.S. Church, P.J. Scammells, D.J. Tucker, ‘A Spectroscopic Analysis of the
Reactions of Dibromopropionyl/ Bromoacryl Dyes with Wool’, Proc. XXX Colloquium Spectroscopium Internationale (1997).
17 H. Zollinger, ‘Color Chemistry: Syntheses, Properties and Applications of Organic Dyes and Pigments’, (Weinheim New York), (1987), p. 140.
18 J.H. Bradbury, D.E. Peters, Text. Research J., 42, 1972, 248 19 S. Kokot, N.A. Tuan, L. Rintoul, U. Meyer, Appl. Spec., 51 (3), 1997, 387 20 H. Martens, T. Næs, ‘Multivariate Calibration’, (John Wiley & Sons Ltd., 1991), p.166. 21 PRS, Pattern Recognition System, SIRIUS version 6.0, User Guide (1993), p. 143. 22 B.G.M. Vandeginste, D.L. Massarat, et al., “Data Handling in Science and Technology-
Handbook of Chemometrics and Qualimetrics: Part B”, p.366. 23 G.E. McGraw, ‘Polyester Structure by Laser Raman Specroscopy’, C.D. Craver (ed), Polym.
Character., Interdisciplinary Approaches, Proc. Symp. (1971). 24 J.V. Miller, E.G. Bartick, Appl. Spec., 55, 2001, 1729-1732. 25 J.S. Church, J. O’Neil, A.L. Woodhead, Textile Research J., 69 (9), 1999, 676-684 26 G.E. McGraw, Amer. Chem. Soc., Div. Org. Coatings Plast. Chem., 30, 1970, 20-26. 27 J. Purvis, D.I. Bower, J. Polym. Phys. Ed., 14, 1976, 1461-1484. 28 G.m. Venkatesh, P.J. Bose, A.H. Khan, J.P. Sibilia, S.L. Hsu, J. Appl. Polym. Sci., 26, 1981,
223-230. 29 J.S. Church, W-H. Leong, ‘ The Analysis of Wool Textile Blends Using FT-IR Photo-
Acoustic and FT-Raman Spectroscopies’, Proc. 9th int. Wool Text. Res. Conf., Biella, 4 (1995), p.p. 114-122.
30 J.L. Koenig, “Spectroscopy of Polymers”, (American Chemical Society, Washington D.C. 1992), p.120.
31 E.D. Lipp, M.A. Leugers in ‘Analytical Applications of Raman Spectroscopy’, M.J. Pelletier (ed), 1st ed. (1999), Ch.3.
32 D. Lin-Vien, N. B. Colthup, W.G. Fateley, J.G. Grasselli, “The Handbook of Infrared and Raman Charactristic Frequencies of Organic Molecules”, (Academic Press, 1991), pp.485-490.
33 S. Carswell, “Microanalysis of Dyes From textiles”, Master Thesis, Queensland University of Tecnology, 1991.
34 J. Cheng, ‘Characterisation of Wool Treated with Metal Ions’, Master Thesis,Queensland University of Tecnology, 1993.
35 I. Keen, ‘Forensic Application of Raman Microprobe’, Masters Thesis, Queensland University of Tecnology, 1998.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 121
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool
5.1 Introduction
Sunlight is an important cause of physical and chemical damage to polymeric
materials such as plastics, textiles and paints. The damage is caused mainly by short
wavelength ultraviolet light1. When exposed to light, natural fibres such as wool and
cellulose undergo photo-bleaching (caused by blue light, i.e. above 400 nm), followed
by yellowing and tendering of the fibres caused mainly by the UV radiation between
290-310 nm. The degree of fibre yellowing has been shown to increase linearly with
its exposure up to 104 hours of sunlight2.
The extent of photo-degradation on the fibre may be assessed by the measurement of
the tensile strength of the fibre. Tensile strength decreases dramatically after 52 hours
of exposure of the wool fibre to accelerated light2. Simpson et al.3 have found that
both peptide and cystine cross-linkages of wool are destroyed upon photo-degradation
and a new range of ionic groups (mainly acidic) is formed. They have also reported
some success in reducing the rate of photo-degradation on wool by the use of UV
absorbers and dyes. On the other hand, chemical bleaching of the fibre and the use of
fluorescent whitening agent (FWA) on fibres accelerates photo-degradation 4.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 122
The most likely reason5 for yellowing in the wool fabric is environmental effects on
the lipids situated on the surface of the fibre. Miller and Smith6 found that, although
some dyes accelerate the loss of strength in the fibre on exposure to sunlight, treating
the fibre with aluminium salts followed by dyeing with mordant dyes has the opposite
effect.
The chemistry of the photo-degradation of wool is not fully understood. Amino acid
analysis of UV irradiated wool fibre has indicated that, following irradiation of the
fibre, there is considerable degradation of the amino acids tryptophan, cystine,
tyrosine, phenylalanine and histidine, with tryptophan being the precursor in the
photo-yellowing of the wool fibre 2.
It has been suggested that, when wool is exposed to sunlight, a large number of
disulfide bonded cysteine residues are oxidised to cysteic acid7. This causes a
reduction in the tenacity of the fibre. The change in the elasticity of the fibre is
attributed to changes in intra- and inter-molecular forces in the α-helix with the
possible formation of new covalent cross-linkages.
Photo yellowing due to the degradation of tryptophan upon exposure to light is also
greatly influenced by oxygen. It is suggested8 that the photo-degradation of
tryptophan occurs via a cleavage of the indole ring resulting in the formation of
kynurenine derivatives some of which are yellow. The photo yellowing of wool is not
due to one particular yellow chromophore, but to melanin-like structures or
ommochrome dyestuffs.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 123
Several significant changes occur in the molecular structure of wool when it is
exposed to stimulated sunlight, viz.:
1. Disulfide bonded cystine amino acid residues are oxidised with cysteic acid
being the major product.
2. There is a reduction in the α- helical content, while random coil or β-sheet
protein chain conformations are favoured.
3. There are structural changes in the fibre as a consequence of cystine
photo-oxidation.
4. Tryptophan amino acid residues in the wool fibre are degraded upon
exposure to light, causing photo yellowing in the fibre.
It has been suggested9 that Raman spectroscopy may be used to obtain detailed
information about UV absorbers in wool and that this technique is particularly
useful since structural information about the UV absorbers may be obtained
without any special sample preparations and since there is minimal interference
because of the relatively weak scattering by the substrate.
Various studies have also been conducted on fading of colours on different
substrates. A study on the colour fading in dehydrated flowers using photo-
acoustic spectroscopy10 concluded that the fading of the vibrant colours is due to
the humidity in the air rather than temperature. The same observation has been
made for photo-oxidation of wool, where it is suggested that atmospheric oxygen
interacts with light and humidity in the air11. Choi et al.12 have studied photo-
catalytic degradation of three Lanasol dyes and suggested that this technique may
be used for treating textile wastewater.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 124
Irradiated Undyed Wool Fabric Church and Millington7 studied the photo-degradation of wool using Raman and
FT-IR. Wool samples were irradiated in air using various sources — UVC
(254nm), UVA (360nm) and blue light (420nm). Two photolytic reaction
mechanisms involving the cystine residues were proposed with the pathways
dependent on the wavelength of the applied radiation. The conditions used for the
Raman spectrometer are as indicated in Table 5.1.
Table 5.1 Conditions chosen in FT-Raman by Church and Millington7
Six spectra of each sample were collected, averaged and then normalised on the
strong band assigned to CH2 and CH3 bending modes at 1450 cm-1. The spectral
changes caused by various exposing wavelengths have been studied; in particular the
changes in the S-S and S-H stretching lines due to the changes in the concentration of
cysteic acid as well as tryptophan-phenylalanine band were considered.
UV-C radiation was found to cause oxidative cleavage of the S-S bond in cystine to
produce cysteic acid and the partially oxidised derivatives cystine S- monoxide and S,
S-dioxide residues. UVA and blue light produce cysteic acid and cysteine-S-sulfonate
residues. Thiol groups are also produced after UVC and UVA irradiation. Significant
halla
This table is not available online. Please consult the hardcopy thesis available from the QUT Library.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 125
degradation of the amino acid tryptophan was also found after irradiation with UVA
and the blue light.
Lewis and co-workers2 studied the photo-oxidation of wool using FT-Raman
spectroscopy, where the samples were exposed to accelerated light exposure
conditions. The conditions for the instrument are tabulated in Table 5.2.
Table 5.2- Conditions used in FT-Raman spectrometer by Lewis and co-workers2
In order to allow comparison of changes in band intensities after photo-oxidation, they
normalised the spectra against the peak intensity of the CH2 anti-symmetric stretch at
2933cm-1 rather than the band at 1450cm-1 as done by Church because of the a strong
band associated with Cibafast W (a UV absorber) at 1454cm-1. They concluded that
differences between the irradiated and non-irradiated spectra were mainly found in the
1800-500 cm-1 region. Lewis et al. argued that changes in the intensities in the S-S
and C-S peaks confirmed that there is oxidation of cystine residues, which in turn was
supported by the appearance of a peak assigned to the S-O vibration of cysteic acid.
They have also shown that the relative intensity of the Tyrosine Fermi doublet at 830
and 854 cm-1 was affected by exposure. They have offered two explanations for that.
One is that tyrosine was photo-oxidised; another is that the intensity ratio of these two
peaks was sensitive to the strength of the H-bonding to the phenolic hydroxyl group
of tyrosine and that the increase in the intensity of the band at 854cm-1, showed that
the tyrosine residues were strongly H-bonded or buried within the hydrophobic
halla
This table is not available online. Please consult the hardcopy thesis available from the QUT Library.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 126
region, being protected from further photo-oxidation. Lewis et al.2 have also
concluded that the bands belonging to tryptophan do not show any obvious changes,
despite the fact that it does degrade upon the exposure of wool to sunlight. This is in
agreement with earlier studies8, 13 indicating that tryptophan is affected by its exposure
to sunlight. In those studies they examined tryptophan amino acid in aqueous solution,
as well as by isolating it from the wool fibre using hydrolysis.
Lewis2 argued that changes in the intensities of the peaks at 1245, 1665 and 935cm-1
were indicative of an increase in the disordered content of the fibre by irradiation. The
results of Lewis’s studies are displayed in Table 5.3.
Table 5.3 Changes reported by Lewis et al. 2 in the vibrational spectra of
wool due to UV-irradiation
halla
This table is not available online. Please consult the hardcopy thesis available from the QUT Library.
Figure 5.1 FT-Raman spectra of undyed wool (texp= 0, 168, 504 hrs.) Note: The arrows show some of the peaks affected by UV-radiation
-0.004
-0.002
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
1500
1448
1396
1344
1292
1240
1188
1136
1084
1032 98
092
887
682
477
272
066
861
656
451
2
Unexposed wool Texp.= 7 days
(168 hrs) Texp. = 21 days (504 hrs)
Wavelength (cm-1)
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 127
5.2 FT-Raman Spectroscopy of Irradiated Wool
In this chapter, irradiated undyed and dyed wool were studied using Raman
spectroscopy with the aid of chemometrics.
Briefly, in this project, samples of undyed woollen fabric were weathered using a
Sun-Test instrument. One set of samples was irradiated for seven days, while the other
set was weathered for twenty-one days. A set of swatches was kept as a control set
(texp = 0 days). These samples were kept in dark, under constant temperature and
humidity. FT-Raman spectra were then recorded for each set. The spectra for samples
at texp = 0, 7 and 21days is shown in Figure 5.1.
Figure 5.1 shows the spectra, in the region of 1500-500 cm-1, of samples with
different exposure times. These spectra are very similar to those reported by Lewis et
al. 2. The peaks marked by an arrow are the ones found both in this study and by
Lewis that were affected by exposure to accelerated weathering.
The frequencies and assignments of these peaks are given in Table 5.4. The peak at
977cm-1 reported by Lewis et al.2 was also found to be present in these spectra. They
have not assigned this peak and indicated that they do not know its origin. It may be
that this band is due to the products of photo-oxidation of the wool cystine residues;
dithiocarbamates have a strong Raman band at ca. 970cm-1 and both monothiolic
acids and ionic dithiolates have strong Raman bands in this region14. This is also
evident from the changes in the intensity of S-O vibrational band at 1040cm-1. It may
be that, upon increasing the exposure time, the S-O bond is further oxidised.
Figure 5.2(b) Loading plot for the component 1 of PCA plot in Figure 5.2 (a1)
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
1524
1460
1396
1332
1268
1204
1140
1076
1012 94
8
884
820
756
692
628
564
Loadings Comp. 1
Wavenumber (cm-1)
Figure 5.2 (a) PCA Plot for Undyed Wool Irradiated (texp= 7, 21 days or 168,
504 hrs.)
I) Spectral region: 1524-524 cm-1
Comp. 1 (81.1%)
Comp. 2 (6.1%)
-2.0 -1.0 0.0 1.0 2.0*10
-1 -1.8
-0.8
0.3
1.3
2.3 *10
-1
** * ***
Undyed wool (texp=504hrs)
Undyed wool (texp=168hrs)
II) Spectral region: 800-524 cm-1
Comp. 1 (60.7%)
Comp. 2 (13.1%)
-4.0 -2.0 0.0 2.0 4.0 *10-2-4.0
-2.0
0.0
2.0
4.0 *10 -2
*
** **
*
texp.=504 hrs. texp.=168 hrs.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 128
As seen in Table 5.4, the position of the peaks as well as their intensity changes
recorded in this thesis and those reported in the literature2 are in good agreement.
5.3 Chemometrics
PCA
Irradiated Undyed Wool Fabric
Three FT-Raman spectra of undyed wool samples irradiated for 168hours (7days) and
504hours (21 days) respectively were recorded. The first derivative of each spectrum
was calculated in Grams32 (9 points, 7 degrees). Those spectra were then block
normalised (i.e. the variables for each spectrum is divided by their sum so that the
block normalised variables add to 100 for each of the spectra) in Excel 97.
A matrix consisting of 55 objects consisting of undyed samples (irradiated for 7days
and 21 days) and 251 variables was built for the 1524-524cm-1 region to be analysed
by Principle Component Analysis (PCA). Although the number of objects in the data
matrix was small, the PCA plot (Figure 5.2(a)) clearly shows the differences between
the samples irradiated for different times. PC1 has separated the samples irradiated for
168hours (negative) from the ones irradiated for 504hours (positive). The loading plot
for PC1, Figure 5.2(b) shows the peaks that cause this separation.
The loading plot indicates that peaks in the regions 976-930, 1020-1004, 1380-1344
and 1500-1452cm-1 exhibit the greatest variation with exposure time. These regions
correspond with the previously reported spectral changes on irradiation as shown in
Table 5.4. While Lewis et al.2 reports that the peaks belonging to tryptophan are too
weak to observe any changes in them, the loading plot demonstrates that the peaks
Table 5.4- FT-Raman spectral frequencies and assignments of the bands
(1500-500 cm-1) affected by exposure to stimulated sunlight in this
thesis and by Lewis 2
Note: * = the gradual increase of the peak at 977cm-1 w.r.t. exposure time was more obvious than the increase or decrease observed for the other peaks.
Frequencies (cm-1) found
in literature2
Changes Reported
With Exposure
Frequencies (cm-1) found
in this work
Changes Observed
With Exposure
Peak Assignments
516 (7 days) 513 Decrease 522 (21 days)
Decrease Cystine S-S band
623 621 (s)
Not changed 623
Not changed Phenylalanine
644 643 (s)
Decrease 644
Decrease Tyrosine
664 665 (br, wk) Decrease 664
Decrease Cystine C-S stretching
830: 854 830: 854 (Fermi Doublet)
Decrease in 854 peak intensity w.r.t. 830 peak
830: 854
Decrease in 854 peak intensity w.r.t. 830 peak
Tyrosine residues
936 935 Decrease
934
Decrease C-C skeletal stretching in α-helix
967 977 Increase gradually* 972
Increase gradually
Not assigned
1004 1004 (s) No change
1004
No change Ring vibration of Tryptophan & Phenylalanine
1040 1040 Increase 1040
Increase S-O vibration of cysteic acid
1244 1245 Increase 1244
Minor increase
Random coil structure
1336, 896 1341, 882 Has shifted in position 1336, 896
Too weak in FT-Raman
Tryptophan
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 129
belonging to this amino acid residue, namely 1341 and 882cm-1, are influenced by the
exposure time. This agrees with Asquith and Rivett13 who have concluded that photo-
degradation causes a decrease in the tryptophan content of wool. The loading plot also
indicates that there is a change in the C-C skeletal structure of α-helical backbone.
This is demonstrated in the region of 1250-1220cm-1, where there is an indication of a
conformational change.
In previous studies on photo-oxidation of wool fabric15, PCA analysis of Raman
spectra was performed on the 800-400cm-1 region. However, in this study it was
found that the 1524-524cm-1 region offers a better separation and grouping of the
data. See Figure 5.2(a, II).
Irradiated Dyed Wool Fabric
In these studies, wool samples dyed with various ratios of three dyes were exposed to
accelerated sunlight. The fabric swatches were left in the weathering instrument and
then checked for changes in colour at 24 hour intervals. Since the first obvious colour
change was observed after seven days, one set was exposed for seven days. After
twenty-one days the colour-change was visually near to extreme.
The samples were then kept in the dark until their Raman spectra were recorded.
Three spectra were recorded for each sample. Three layers of the same sample were
positioned on the top of each other in the sample holder, with the weathered side
exposed to the laser. The spectral data were transformed to their first derivative and
block-normalised as for the undyed wool spectra.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 130
A comparison of the spectra for the irradiated and non-irradiated samples shows that
the dye peaks are still at approximately the same frequencies but exhibit changes in
their intensity. Figure 5.3 demonstrates this point by showing the spectra for irradiated
sample 073 at the two different exposure times. The arrows identify peaks that change
in intensity with exposure.
It is known16 that, since the perceived colour of an object is associated with the
absorbance of some wavelengths of light and reflectance of the rest by the
chromophore, the alteration — or disappearance — of the colour may be related to
chemical changes in that chromophore.
The exact structures of the Lanasol dyes are not available to the public and hence it is
not easy to readily identify the chemical changes that might have occurred during
photo-oxidation. This was not a significant disadvantage because chemometrics relies
on statistical analysis of the spectra and requires no previous chemical knowledge of
the samples. Even without knowledge of the original chemical structure of the dye,
analysis of the PCA loadings may provide some information on what chemical
changes accompany the weathering process.
Group (i)
Data matrix: Woollen samples (texp = 0, 168, 504 hrs.)
Figure 5.8 (a): Predicted vs. Measured plot for the calibration set for RED (non irradiated samples) (Pre-treatment: 1st derivative, block normalised, Y-mean centred)
Figure 5.10 Predicted vs. Measured blue values for calibration set (Pre-treatment: 1st der., subtracted spectra normalised to 100% wool, Y-mean centred)
A calibration set consisting of 142 objects was built. The set was then cross validated.
The plot of predicted versus measured (Figure 5.12(I)) shows that there is a
reasonable correlation between the data in this set. The irradiated samples were then
introduced to this set as the unknown set. The results are shown in Table 5.8.
Dependent Variable: blue
The same calibration set was used here as well. The data are again spread across the
best fitted line. See Figure 5.12 (II). The unknown set was then introduced to this
model. Predicted values for this set are shown in Table 5.9.
Dependent Variable: Yellow
The same calibration set as described previously was chosen. See Figure 5.12(III).
125 objects were introduced as the unknown set to this model. Samples in the
unknown set with zero ratios of yellow dye were not considered since the
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 142
corresponding samples in the calibration set were not predicted very well. The results
of the prediction are displayed in Table 5.10.
Therefore, overall it is concluded that:
• The pre-treatment of the spectral data used here, does not enhance the ability of
the PLS model to predict the ratio of dyes.
• Considering all the results obtained so far, it may be concluded that the best
pre-treatment to be used for PLS1 model here is normalising the spectral data to
100% substrate of the first derivative FT-Raman spectra of the samples, followed
by Y-mean centring of the matrix.
• The PLS model used to predict the dye ratios of the irradiated samples indicates
that the model shows a small ability of the system for prediction, but the
calibration set for the model has to be reviewed.
It has been suggested20,21 that, the correlation and error parameters may be improved
by including data from the same statistical population as the predicting set. This
strategy defines how “similar” a sample is to the rest of samples contained in the
calibration set. A sample is considered to be similar to the one in the calibration set if
the model is able to predict the properties of a sample. If the sample is found to be
dissimilar to the rest of the samples, then it is assumed that there is new information
in that sample unknown to the calibration set and the new sample is then added to the
calibration set automatically in order to improve the model.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 143
It is anticipated that this procedure would indeed improve the predictive ability of the
PLS model. It was, however, not possible here to do so since the ratio of the dyes
remaining on the irradiated samples are not known so it is suggested that a further
study should be conducted in which a larger matrix could be built and the ratios of
the irradiated dyes would be estimated by other means. In this way a calibration set
containing some of the new properties introduced to the model by the irradiation
could be constructed.
Another approach may be to look at the faded dyes in colour-space — that is the shift
colour from the original colour — or in colour-space coordinates. In this way, all of
the data for the original and faded fabric may be used in the calculations rather than
the amount of dye exhausted onto the fabric.
Chapter 5 FT-Raman Spectroscopy of Irradiated Dyed Wool 144
5.4 References 1 P. Brennan, C. Fedor, “Sunlight, UV, & Accelerated Weathering”, (The Q Panel Company,
26200 First St., Cleveland, Ohio 44145), p.1 2 D.C. Jones, C.M. Carr, W.D. Cooke, D.M. Lewis, Textile Res. J., 68, 1998, 739-748. 3 W.S. Simpson, C.T. Page, ‘The Effect of Light on Wool and the Inhibition of Light
Tendering’, Wool Research Organisation of New Zealand Inc., Report No. 60 (1979). 4 S. Collins, R.S. Davidson, J. Photochem. Photobiol. A; 77, 1994, 277-282. 5 E. Wojciechowska, A. Pielesz, A. Wlochowicz, Textile Res. J., 62, 1992, 580-585. 6 I.J. Miller, G.J. Smith, J. Soc. Dyer Color., 111, 1995, 103-106 7 J.S. Church, K.R. Millington, Biospectroscopy, 2, 1996, 249-258. 8 K. Schäfer, D. Goddinger, H. Höcker, J. Soc. Dyer Color., 113, 1998, 350-355. 9 I.H. Leaver, R.E. Hester, R.B. Girling, Textile Research Institute, 1998, 182-184 10 T.J. Racey, P.L. Rochon, Canadian Journal of Applied Spectroscopy, 39, 1994, 38-42. 11 I.Rusznàk, J. Frankl, J. Gombkötő, JSDC, 101 (1985), p.p. 130-136. 12 B. Neppolian, H.C. Choi, S. Sakthivel, B. Arabindoo, V. Murugesan, Journal of Hazardous
Materials, B89, 2002, 303-317. 13 R.S. Asquith, D.E. Rivett, Appl. Polym. Symp., 18, 1971, 333. 14 G. Socrates, “Infrared and Raman Characteristic Group Frequencies, Tables and Charts” 3rd
ed., (John Wiley & Sons, 2001) 15 V. Fredline, S. Kokot, C. Gilbert, , Mikrochim. Acta [Suppl.] 14, 1997, 183-184. 16 Private communication with Dr. S. Wallis (8 Feb. 2002); Dept. of Medicine, University of
Queensland. 17 S. Kokot, S. Carswell, D.L. Massart, , Appl. Spec., 46, 1992, 1393-1399. 18 C. Gilbert, S. Kokot, ‘An Analysis of Oxidised Wool Fabric Using FT-Raman Spectroscopy
and Chemometrics’, Proceeding of 13th Australian Symposium on Analytical Chemistry, Darwin, Australia (1995), AS 40-1.
19 W.S. Simpson, ‘Photo-protection of Wool Fabrics by Dyestuffs’, Wool Research Organisation of New Zealand Inc., Communication No. C77 (1982).
20 D.R. Tallant, R.L. Simpson, ‘The Thermal History of Charred Materials by Raman Spectroscopy’, Sandia Report, SAND2001-0131, Unlimited Release, Printed February 2001, p.p.3-19.
“We yearn for the day when we can totally rely on colour
instrumentation to decide when two colours match, and remove the
responsibility from that colour matcher working third shift.” Bruce Mulholland, Hoeechst-Celanese Engineering Plastics Div., Florence, Ky.1
6.1 Conclusion
To date, the procedure used by the textile industry referred to as “Colour Matching”
involves studying the colour using spectrophotometry of the dye and examining the
reflectance properties of the dye on the substrate. The studies conducted in this thesis
were concerned with analysing dyed wool and dyed wool blend fabrics using
vibrational spectroscopy, viz. FT-IR PAS and FT-Raman carrying on from the work
of Lee-Son et al.2. In this thesis, three dyes on woollen fabrics were examined with
two of the dyes being the same as those studied by Lee-Son et al. FT-Raman spectra
for the dyes were found to be identical with those recorded by Lee-Son et al. The
peaks due to wool keratin are also easily identified. However, unlike the FT-Raman
spectra, the PA spectra do not display well-resolved peaks that are solely due to the
dye molecule. This may be because the backbone vibrations of the dyes are
intrinsically active Raman scatterers and hence weak IR absorbers while the
functional groups in the wool fibre are weak Raman scatterers and hence strong IR
absorbers.
Chapter 6 Conclusion 146
FT- Raman and FT-IR PA spectra of undyed wool and undyed wool/ polyester
samples were also recorded and compared to the ones in the literature3,4. It was found
that they were very similar apart from some minor differences. For example, it was
found that the surface of the undyed wool sample studied in this project was oxidised.
Some variations in the position of the peaks reported in different literature sited were
also observed. Examining the FT-Raman spectrum of the wool sample studied in this
project, it was then possible to draw conclusions to which peak positions reported in
various literatures could be more correct. An undyed wool/ polyester blend sample
was also examined using FT-Raman spectroscopy. All the peaks could be assigned
according to the literature.
One of the main objectives of this project was to investigate the possibility of colour
matching and prediction of dye mixture ratio in dyed woollen fabrics using vibrational
spectroscopy and chemometrics, both qualitatively and quantitatively. For this
purpose PCA and PLS methods of analysis were used respectively.
PCA plots of the FT-IR PA spectra for the woollen fabrics dyed with Lanasol dyes
indicate that the samples of the same colour do cluster loosely together. They do not
position themselves in the same manner as they are positioned on the colour card.
PCA plots for the PA spectra of wool/ polyester blend fabrics displayed the same
conclusion as above. In these plots the spectral data are too scattered to even display
clusters for the repeat spectra of the same sample.
The FT-IR PA spectral data for the woollen fabrics were then submitted to PLS. The
results show that the prediction ability of the PLS1 model studied is quite reasonable
Chapter 6 Conclusion 147
(within ±10% of the correct value) as long as the calibration set chosen demonstrates
satisfactory prediction ability as well.
FT-Raman spectra of woollen fabrics and wool/ polyester blends were also submitted
to PCA and PLS for further analysis. PCA plots of both types of samples indicate that
not only spectral repeats of each sample cluster together but also the arrangement of
the clusters is in accordance with that of the colour card. This is particularly obvious
in the case of wool/ polyester blend samples.
Therefore, overall it may be concluded that the combination of FT-Raman
spectroscopy and PCA offers a better colour recognition than FT-IR PA spectroscopy
and PCA method of analysis for both of these types of samples.
The spectral data were also studied quantitatively by PLS. In these studies it was
found that PLS1 model was the preferred model and that PLS2 model did not work
well at all for this set of data. Unlike the results obtained for the PLS1 model used to
study PA spectra of woollen fabrics, it was found that this method of analysis did not
work as well for the FT-Raman spectral data. On the other hand, when PLS was
applied to the data for the wool/ polyester blend samples, the prediction ability was
satisfactory. Again it was found that the prediction ability of the model depends
heavily on the calibration set to begin with.
The effect of the dye on the Raman spectra was observable without the aid of
chemometric and yet the PLS analysis of the Raman spectra failed to calculate the
Chapter 6 Conclusion 148
correct amount of dye. This strongly indicates that the measured values do not reflect
the true amount of dye on the fabric.
Therefore, overall it was concluded that PLS method used along with FT-Raman data
works well for wool blend, while for pure wool samples it is advisable to use FT-IR
spectra in combination with PLS.
The last objective of the project was to study the fading characteristics of the wool
fabrics with mixed dyes.
A selection out of the woollen samples were UV-irradiated for 7 and 21 days.
FT-Raman spectra of the undyed irradiated wool fabric agreed well with those found
in literature5. In this thesis, the peak at 997cm-1 mentioned by Lewis et al.5 was also
observed. It was suggested that this peak appears and gradually increase due to further
oxidation of –S-O-S- bond in the wool fibre.
A number of dyed woollen samples were also UV-A irradiated and studied by FT-
Raman spectroscopy. Minor changes were observed in the intensities of some of the
peaks with respect to the exposure time. The data were then submitted to PCA and
PLS for pattern recognition and prediction of the ratio of the dye remaining intact,
respectively. The plot for PCA for samples at exposure times zero, 7 and 21 days
indicated that PCA separates non-irradiated samples from the rest. The spectral data
for the exposed samples were also studied in order to find out if these samples would
position themselves according to the colour card, i.e. the way they did before being
irradiated. The results showed that although spectral repeats of the same sample still
Chapter 6 Conclusion 149
clustered together they did not follow the pattern shown by the same ones at
texp= 0 hrs. One possible explanation for this is that the ratio of dyes on the exposed
samples differed from those of the original set due to differential degradation of the
three dyes.
The data were then studied using PLS. It was found that the ability of this method for
predicting the ratio of all the three colours was poor. It was however found that the
ability of prediction for dependent variable yellow was better than for the colours red
and blue. According to the literature6,7 the poor results obtained may be justified since
the irradiated samples may posses properties that are not accounted for in the
calibration. For example, the degradation of the underlying wool in addition to the
dye molecule has not been considered in modelling of the calibration set. Also, the
ratio of the dyes might have not changed but the absolute amount of dye could have
become different. Any combination of these reasons could have in turn reduced the
ability of the model in correctly predicting the ratio of the colours.
This project has therefore shown that:
FT-Raman spectroscopy of dyed wool along with application of
chemometrics (namely, PCA and PLS) is a feasible method of studying
woollen samples both qualitatively and quantitatively; and that the ratio of
colours on the samples may be predicted quite successfully.
The above procedure when applied to UVA exposed samples of the same
origin, does not offer the same success rate.
Chapter 6 Conclusion 150
6.2 Future Work
It is therefore suggested that further studies of the same nature containing a larger data
matrix and for more colours should be conducted in order to make it applicable for the
dyeing industry. As it stands, vibrational spectroscopy techniques are not well suited
to industrial analysis of dyed fabrics since they involve complicated procedures using
expensive equipment and produce results of limited accuracy. With further
development , however, these technique may find useful application in forensics or
other specialized analysis.
The same suggestion applies to the irradiated samples. There, a much larger data
matrix should be considered. The ratio of the dye still intact should also be measured
by some other means. A correlation between the irradiated and non-irradiated samples
should be realised. Followed by building an appropriate calibration set containing
some information about the irradiated samples. Only then the unknown samples
should be introduced to the PLS model and predicted.
Chapter 6 Conclusion 151
6.3 References
1 B. Mulholland, Plastics Compounding, 12, 1989, 33. 2 G. Lee-Son, R.E. Hester, J. Soc. Dyer Color., 106, 1990, 59-63. 3 E. Carter, ‘Vibrational spectroscopic Studies of Wool’, PhD thesis, Queensland University of
Technology, 1998. 4 M.A. Moharram, T.Z. Abdel-Rehim, S.M. Rabie, J. App. Polym. Sci., 26, 1981, 921-932 5 D.C. Jones, C.M. Carr, W.D. Cooke, D.M. Lewis, Textile Res. J., 68, 1998, 739-748. 6 D.R. Tallant, R.L. Simpson, ‘The thermal History of Charred Materials by Raman
Spectroscopy’, Sandia Report, SAND2001-0131, Unlimited Release, Printed February 2001, pp.3-19.
This figure is not available online. Please consult the hardcopy thesis available from the QUT Library.
Figure A1-2: Diamond 2
halla
This figure is not available online. Please consult the hardcopy thesis available from the QUT Library.
Figure A1-3: Diamond 3
halla
This figure is not available online. Please consult the hardcopy thesis available from the QUT Library.
Appendix 2 page A2/1
Appendix 2
Appendix 2 page A2/2
Table A2-0 FT-Raman Peak Positions and their Assignments for Undyed Wool and
Wool Dyed with Two Lanasol Dyes Note: The three-digit dye code numbers refer to the ratio of red, blue and yellow dyes on the wool (e.g. sample 073 has red:blue:yellow in the ratio 0:7:3).
Table A2-0 FT-Raman Peak Positions and their Assignments for Undyed Wool and Wool Dyed with Two Lanasol Dyes
Note: The three-digit dye code numbers refer to the ratio of red, blue and yellow dyes on the wool (e.g. sample 073 has red:blue:yellow in the ratio 0:7:3).
1. G. Lee-Son, R.E. Hester, J. Soc. Dyer Color., 106, 1990, 59. 2. E. Carter, P.E. Fredericks, J.S. Church, R.J. Denning, Spectrochimica Acta, 50A (11), 1994,
1927-1936
Appendix 2 page A2/4
Group: (i) Model:2 Cal. Set:84 Val. Set:39 Dep. Var.: Red Figure A2-1 Plot of predicted vs. measured values for calibration (I) and validation (II) sets, group (i), model 2, dep. Var. Red I)Calibration Set:
Pred. vs Measured, Var %R,(5 Comp), Avg. Pred. Err.= 0.950, Date 5/1/2002
Table A2-1-Results of PLS model 2 applied to Group (i), dependent variable red Note: The number in the parenthesis indicates the predicted value rounded to the nearest digit for comparison with the measured value Name Predicted Measured 073Ra -0.89 (-1) 0073Rb -0.72 (-1) 0073Rc -1.8 (-2) 0073Rd -2.4 (-2) 0073Re -2.8 (-3) 0073Rf -19 0181Ra 1.3 (1) 1181Rb 1.8 (2) 1181Rc 1.7 (2) 1181Rd -0.60 (-1) 1181Re 1.4 (1) 1181Rf 1.6 (2) 1217Ra 3.1 (3) 2217Rb 1.9 (2) 2217Rc 2.7 (3) 2217Rf 3.0 2307Ra 2.5 (3) 3307Rb 1.9 (2) 3307Rc 2.2 (2) 3307Rd 1.3 (1) 3307Re 0.50 (1) 3307Rf 2.0 3343Ra 5.0 3343Rb 5.4 (5) 3343Rc 5.8 (6) 3343Rd 5.9 (6) 3343Re 6.7 (7) 3343Rf 5.0 3811Ra 13 8811Rb 10 8811Rc 11 8811Rd 11 8811Re 13 8811Rf 12 8730Ra 4.4 (4) 7730Rb 7.6 (8) 7730Rc 4.3 (4) 7
Appendix 2 page A2/6
Group: (ii) Cal. Set: 86 Val. Set: 47 Dep. Var.: Blue Figure A2-2 (a)- Plot of predicted vs. measured values for calibration (I) and the validation (II) sets, group (ii) (dep. Var. Blue) I)Calibration Set:
Pred. vs Measured, Var %BL,(6 Comp), Avg. Pred. Err.= 1.264, Date 2/7/2002
Group: (ii) Cal. Set:86 Val. Set: 47 Dep. Var.: Yellow Figure A2-3- Plot of predicted vs. measured values for calibration (I) and validation (II) sets, group (ii), (dep. Var. Yellow)
I)Calibration Set:
Pred. vs Measured, Var %Y,(8 Comp), Avg. Pred. Err.= 1.069, Date 2/7/2002
Figure A2-4 Plot of predicted vs. measured for sides 1,4 and 5 in Diamond 2, Calibration set (I) and Validation set (larger set) (II), dep. Var.= brown, indep. Var. 1800-800 cm-1
I) Calibration Set:
Pred. vs Measured, Var Brown,(8 Comp), Avg. Pred. Err.= 0.212, Date 10/7/2002
Figure A2-5 Plot of predicted vs. measured for sides 1,4 and 5 in Diamond 2, Calibration set (I) and Validation set (larger set) (II), dep. Var.= yellow, indep. Var. 1800-800 cm-1
I) Calibration Set:
Pred. vs Measured, Var Yellow,(5 Comp), Avg. Pred. Err.= 0.180, Date 10/7/2002
Figure A2-6 Plot of predicted vs. measured for sides 1,4 and 5 in Diamond 2, Calibration set (I) and Validation set (larger set) (II), dep. Var.= red, indep. Var. 1800-800 cm-1
I) Calibration Set:
Pred. vs Measured, Var Red,(7 Comp), Avg. Pred. Err.= 0.180, Date 10/7/2002
Figure A2-7 Plot of predicted vs. measured for sides 1,4 and 5 in Diamond 1, Calibration set (I) and Validation set (II), dep. Var.= brown, indep. Var. 1800-800 cm-1
I) Calibration Set:
Pred. vs Measured, Var Brown,(6 Comp), Avg. Pred. Err.= 0.262, Date 9/1/2002
Figure A2-8 Plot of predicted vs. measured for sides 1, 4 and 5 in Diamond 1, Calibration set (I) and Validation set (II), dep. Var.= grey, indep. Var. 1800-800 cm-1
I) Cal. Set:
Pred. vs Measured, Var Grey,(6 Comp), Avg. Pred. Err.= 0.178, Date 4/7/2002
Figure A2-9 Plot of predicted vs. measured for top triangle of Diamond 1 for Calibration set (I) and Validation (II,III) sets, dep. Var.= Brown, indep. Var. 1800-800 cm-1
I) Cal. Set:
Pred. vs Measured, Var Brown,(6 Comp), Avg. Pred. Err.= 0.569, Date 5/7/2002
Figure A2-10 Plot of predicted vs. measured for top triangle of Diamond 1 for Calibration set (I) and Validation (II) set, dep. Var.= Grey, indep. Var. 1800-800 cm-1
I) Cal. Set:
Pred. vs Measured, Var Grey,(6 Comp), Avg. Pred. Err.= 0.330, Date 5/7/2002