Non-destructive testing of silk: problems and possibilities

Abstract

Silk deterioration can be catastrophic, with splits and tears leading to powdery and fri-able surfaces which are difficult to conserve. The deterioration is reported to be accelerated for silks which have been treated with metallic salts; however, these materials are difficult to identify. Methods of non-destructive testing to identify the types of silk and their condition have been explored. The use of these techniques in situ to reduce sampling or object movement is discussed along with the problems and possibilities of each technique. résumé La dégradation de la soie peut être catastro-phique, puisque les déchirures et les accrocs rendent les surfaces poudreuses et friables, et donc difficiles à conserver. La dégradation est décrite comme accélérée pour les soies ayant subi des traitements à base de sels mé-talliques ; toutefois, ces matériaux sont diffi-cilement identifiables. Des méthodes d'essais non destructifs destinés à identifier les types de soie et leur état de conservation ont été explorées. L'utilisation de ces techniques in situ afin de réduire l'échantillonnage et les déplacements des objets est discutée, ainsi que les problèmes et les possibilités relatifs à chaque technique. resumen El deterioro de la seda puede ser catastrófico, con rajaduras y desgarres que hacen que las superficies se vuelvan pulverulentas y fria-bles, y sean difíciles de restaurar. Se sabe que el deterioro se acelera en sedas que han sido tratadas con sales metálicas. Sin embargo, estos materiales son difíciles de identificar.

Full text

TEXTILES
1
Keywords: silk, non-destructive test-
ing, condition assessment, elemental
analysis
ABSTRACT
Silk deterioration can be catastrophic, with
splits and tears leading to powdery and fri-
able surfaces which are difficult to conserve.
The deterioration is reported to be acceler-
ated for silks which have been treated with
metallic salts; however, these materials are
difficult to identify. Methods of non-destruc-
tive testing to identify the types of silk and
their condition have been explored. The use
of these techniques in situ to reduce sam-
pling or object movement is discussed along
with the problems and possibilities of each
technique.
RÉSUMÉ
La dégradation de la soie peut être catastro-
phique, puisque les déchirures et les accrocs
rendent les surfaces poudreuses et friables,
et donc difficiles à conserver. La dégradation
est décrite comme accélérée pour les soies
ayant subi des traitements à base de sels mé-
talliques ; toutefois, ces matériaux sont diffi-
cilement identifiables. Des méthodes d’essais
non destructifs destinés à identifier les types
de soie et leur état de conser vation ont été
explorées. L’utilisation de ces techniques in
situ afin de réduire l’échantillonnage et les
déplacements des objets est discutée, ainsi
que les problèmes et les possibilités relatifs à
chaque technique.
RESUMEN
El deterioro de la seda puede ser catastrófico,
con rajaduras y desgarres que hacen que las
superficies se vuelvan pulverulentas y fria-
bles, y sean difíciles de restaurar. Se sabe que
el deterioro se acelera en sedas que han sido
tratadas con sales metálicas. Sin embargo,
estos materiales son difíciles de identificar.
NAOMI LUXFORD*
Centre for Sustainable Heritage, Bartlett
School of Graduate Studies
University College London
London, UK
n.luxford@ucl.ac.uk
DAVID THICKETT
English Heritage
London, UK
PAUL WYETH
c/o Centre for Textile Conservation and
Technical Art History
University of Glasgow
Glasgow, UK
*Author for correspondence
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND
POSSIBILITIES
INTRODUCTION
Object and conservation records, along with archives, can provide a wide
variety of information on collections. Despite this, elucidating the composition
and condition of textiles requires scientific analysis. However, a balance
has to be achieved between the care, conservation and display of the objects
and the removal of material for analysis to inform these areas. Generally,
destructive analytical techniques require large samples and are limited in
application to surrogate materials for laboratory experiments. In some cases,
micro-sampling of objects for destructive analysis is acceptable due to the
value of information obtained. For example, silk fibroin molecular weight,
an indicator of the state of deterioration, is revealed by high-performance
size exclusion chromatography (HPSEC), which requires just a millimetre
of thread.
Ideally, analysis would be performed without removing samples and
even while an object is on display. The techniques would need to be both
non-invasive (samples do not need to be removed) and non-destructive
(no damage caused), whilst still providing useful results. Such analysis
can help inform curators and conservators both of the technical art history
as well as the condition of the objects. Research has been undertaken to
determine the potential of two non-destructive techniques for revealing
the condition and composition of historic silks on open display.
ANALYSIS METHODS
An initial review identified a number of techniques, most notably
spectroscopic techniques such as Raman spectroscopy and polarised
attenuated total reflectance infrared spectroscopy (pol-ATR), which could
be used to provide detailed information on the condition and composition
of silk (Luxford 2009). However, almost all of the techniques identified
required samples, although often very small (less than 1 mg), and very
few could be used on site or truly non-destructively. Two techniques,
X-ray fluorescence (XRF) and near-infrared (NIR) spectroscopy, had the
potential to provide information on the condition and composition of the
silk. Both techniques are non-invasive and non-destructive making them
suitable for in situ analysis of historic collections. Furthermore, both are
becoming increasingly portable and more widely used within heritage
science enabling access to this equipment.

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
2
Se han estudiado métodos no destructivos
para identificar los tipos de seda y su estado
de conservación. El artículo discute el uso de
estas técnicas in situ para reducir la toma de
muestras y el movimiento de los objetos, así
como los problemas y las posibilidades de
cada técnica.
XRF spectroscopy
XRF has been applied to textiles for the identification of mordants
(Masschelein-Kleiner and Maes 1978) and weighting agents (Becker et al.
1987), as well as pesticides (Odegaard et al. 2006), painted details (Skelton
1995) and mineral dyes (Gardiner et al. 2000). However, previous studies
have usually relied on objects being taken to larger-scale laboratory-based
equipment. More recently, portable, hand-held devices have become
available and the potential of the Bruker TRACeR III-V handheld XRF was
assessed for use with historic textiles. The instrument is battery powered
with an associated PDA for data analysis, although a laptop and power
cables can also be used. When used with the KeyMaster-Bruker vacuum
pump attachment, elements with atomic mass down to magnesium can be
identified. The sampled area is an ellipse approximately 5 by 6 mm and,
in this study, a real count time of 40 seconds was used.
NIR spectroscopy
NIR spectroscopy has already found application for paper and plastics
within conservation science and some studies on textiles have also been
reported (Richardson et al. 2008, and references therein). A variety of
natural and synthetic fabrics can be identified, as long as there is a suitable
reference dataset for comparison. NIR has also been shown to be of value
for the diagnosis of areas of strain within silk (Richardson and Garside
2009). Previously, NIR has indicated potential for condition monitoring
of silk (Zhang and Wyeth 2007, Kim and Wyeth 2009).
To characterise the condition of the historic silk collection, a NIR spectrometer
was taken to Brodsworth Hall, an English Heritage property. NIR spectra
have been collected using a Perkin Elmer Spectrum One Fourier-transform
near infrared spectrometer (FT-NIR) with an Axiom fibre optic probe
scanning between 4000-12000 cm-1 with a resolution of 8 cm-1 and a scan
accumulation of 64. The background reference was Spectralon®. All
data were collected in absorbance at 22 ± 2°C and 52 ± 5% RH. Baseline
corrections and spectral averaging were carried out in Thermo Galactic
Grams AI version 8. Additional spectral pre-processing and multivariate
analysis (MVA) was performed with The Unscrambler® version 9.7 software
by Camo Technologies Inc.
EXPERIMENTAL RESULTS AND DISCUSSION
XRF results
XRF was used at two English Heritage historic houses, Audley End House
and Brodsworth Hall, to provide elemental analysis of textiles on display.
One of the immediate benefits of on-site analysis was the greater number
of areas that could be analysed. For example, in Figure 1 individual colours
could be selected to determine differences in mordants and weighting
agents used. By using a non-invasive and non-destructive technique, it was
possible to select areas to sample for analysis by other techniques, such as
Figure 1
Detail from the festoon curtain in the
Little Drawing Room at Audley End House.
Numbers indicate representative areas of
the different coloured threads analysed

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
3
HPSEC. The use of the portable XRF also allowed ready determination
of the composition of a number of metal threads, including gilt silver,
and brass (see Figure 2, for example), found within the silk artefacts. In
most cases, these would have been impossible to sample without causing
obvious aesthetic damage to the textiles design.
A variety of elements were identified within the silk, including some that
were expected such as tin (probably weighted silk), iron (probably used
as a mordant on black silk) and chromium (used as a mordant on modern
reproduction silks). For example, the prize ribbon in Figure 2 has high
levels of tin, indicating the ribbon has been weighted. Treatment of silk
with metallic salts, especially tin, was common, particularly for costume
and fashion textiles (Hacke 2008). These materials are seen as inherently
deteriorating due to the presence of the metallic salts, but are particularly
difficult to identify as there are no obvious visible indicators. Hence,
finding a non-destructive method of identifying these materials is of
crucial importance for curators and collections managers.
Whether the element identified is a result of weighting or mordanting (or
an alternative source such as pollution) is less certain, thus limiting to
some degree the value of the information for conservators and curators.
This arises as the equipment standards are based on metal alloys and
without access to the XRF software it was not possible to establish an
internal XRF calibration for silk textiles (further details can be found
in Luxford 2009). Therefore, an element can be qualitatively identified
as present, but the quantitative concentration could not be determined,
making assignment of its source difficult. In most cases, the source can
be assumed, e.g. chromium in modern textiles is most likely to arise from
the mordant used. However, historically tin was used both as a mordant
for light coloured textiles and a weighting agent, as was iron for dark
coloured dyes, primarily black.
Identified elements are present at more than twice the baseline level,
whereas trace elements were present above the baseline but below this
point. Baselines for each element were determined by comparison of the
amount found in all analysed objects with the line drawn above the low
concentration present in all samples (Luxford 2009). Results from the in
situ XRF analysis at Brodsworth Hall are presented in Table 1. In some
cases, other materials also seem to have been sampled. For example, on
the sofa in the south hall the silk upholstery is light coloured and therefore
unlikely to contain iron as a mordant. However, iron has been identified,
which may arise from something like an upholstery tack beneath the
upholstered layer. Lead was identified in situ during analysis at both
Brodsworth Hall and Audley End House. Although lead was sometimes
used as a weighting agent, it is rare for upholstery fabrics to be weighted.
In most cases, when lead was identified it was on wall silk, and a possible
explanation is the presence of lead paint beneath the silk.
Due to perceived inherent deterioration of weighted silk, objects in a poor
condition are often assumed to be weighted. However, in some cases silk
Figure 2
Prize ribbon [90015011] from Brodsworth
Hall with brass metal threads

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
4
in a very poor condition showed no presence of any metallic elements. For
example, only sulfur was identified on the cabinet in the north corridor,
which is in a particularly perilous condition. Other authors have previously
noted that much degraded silks, labelled as weighted due to their condition,
contained no metallic elements, but high levels of sulfur (Ballard et al.
1990). In these cases, the damage is thought to be the result of sulfur
bleaching processes rather than any weighting agents.
Table 1
XRF results from in situ analysis at Brodsworth Hall
room location identified elements trace elements
south hall sofa good condition area S, Sn, Ca K, Fe
wall gold silk S, Pb
drawing room wall behind exit door S Ca
sofa behind door S, Sn, Ca
sofa behind exit door shattered area Sn S, K, Ca, Pb
drawing room sofa behind exit door good condition area S, Sn, Ca, Pb
wall silk (by entrance door) gold area S
wall exit door gwwood condition S Pb
north corridor cabinet grille Cu, Zn S
cabinet RHS flat area silk S
store narrow cummerbund metal thread Cu, Zn S
wide cummerbund metal thread Cu, Zn S
prize ribbon tassel Cu, Zn P, S
prize ribbon Si, K, Sn, Ca P, S
Practical problems included the weight of the XRF when held for long
periods and the relatively short battery-life, although on more recent
models the battery life has been significantly improved and now lasts for
a full day of analysis. The portable XRF allows elements to be identified
within the silk. However, due to the problem of establishing an internal
XRF calibration for silk textiles, it was often necessary to check the
identification after the analysis, which increases the interpretation time.
There were significant advantages in using the portable equipment on site;
for example, the inorganic composition of many of the objects could not
otherwise have been determined, as sampling would have been overly
damaging. It also allowed a much wider range of objects to be analysed,
for example wall silk could never have been taken to a laboratory for
analysis. Although the amounts of elements were not quantified in this
work, it was possible to qualitatively identify the presence of elements
and make assignments based on known processing methods.
NIR identification results
In some cases, identification of materials can help curators establish the
date or the status of an object, for example the brass (rather than gold)
metal threads above in Figure 2. NIR can be used for identification when a
suitable database of spectra is available for comparison. During the research,
identifying the material used to make the pile for velvets was problematic
and for a number of objects fibre identification by NIR proved especially

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
5
useful. The majority of objects were wool, but in some cases silk velvet was
found. Silk upholstery and cellulosic padding layers were also successfully
separated and identified using this technique (Figure 3). This could be done
by separately analysing areas of remaining silk and the underlying padding
visible in areas of damage, such as split areas of the silk. In some cases, the
padding contribution was subtracted from the spectrum to give a clearer
indication of the remaining upholstery layer composition.
Figure 3
NIR spectra of silk upholstery (red) and padding (purple)
Identification by NIR spectroscopy has a number of advantages, including
the fast analysis and the fact that samples are not required. However, in
order to identify a material, a large reference database is required (Kawano
2002). This can be difficult as known and well-characterised materials are
required as reference materials. The broad overlapping peaks that typify
NIR spectra can also make it difficult to identify materials and some
spectral processing can be required for materials with similar chemical
groups such as wool, silk and nylon. This processing can significantly
add to the interpretation time, meaning that although the data collection
is rapid, the actual identification is time-consuming. Although the NIR
spectrometer used can be moved easily, it is relatively large and so not ideal
as a portable instrument; furthermore, it requires a power supply during
operation. This can limit its use, although more portable models are now
available. The lack of portability is compensated, to some degree, by the
use of a fibre-optic probe limiting the need to move the spectrometer.
NIR condition assessment results
In order to assess small changes within NIR spectra multivariate analysis
(MVA) has been used. This allowed the development of a model which
can be used to predict the tensile strength of silk from the NIR spectrum.
The model was built using NIR spectra and the tensile properties of a
reference set of artificially aged surrogate silks (Luxford 2009). Tests of the
model with other samples of the same silk gave results within the model’s

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
6
deviation (around 30 N). This indicated condition predictions based on the
NIR spectra were possible. The deviation is calculated by Unscrambler
using the model error, sample leverage and residual X-variance. Large
deviations indicate the predicted sample is different from the calibration
samples.
Subsequent prediction of condition for the historic silks suggested better
performance than expected, but with relatively large prediction errors
(Figure 4). For example, unaged silk in the accelerated ageing experiments
had a tensile strength of 160 N. For most of the objects analysed in situ the
results were equal to, or greater than 160 N. This indicates that the historic
silks are not well described by the current model. As the model used here
has been built solely using artificially aged samples of unweighted modern
silk, it is unsurprising there are differences between the real objects and the
model. A more robust model would require a great number and variety of
historic silks with suitable data on the condition, either tensile properties or
molecular weight from size exclusion chromatography (HPSEC) analysis
of micro-samples.
Figure 4
Predicted tensile strengths (in N) of historic silks (white line) of in situ NIR spectra with the deviation shown
as a blue bar (sample number from 1 to 39 is shown on the x-axis)
As the model was not as successful in situ as expected from the laboratory
trial, the NIR spectra from a number of objects were investigated to
determine some of the factors which may influence the model. Coloured
samples were first investigated, since additional, complicating dye-related
absorptions would be expected. However, the current model limits the
analysis to a small spectral region (4100-5100 cm-1) and colour had little
significant effect.
As the XRF analysis had demonstrated some of the objects were made
of silk that contained elements such as tin, these were also studied to
determine if there was any impact. In Figure 5, a clear change can be
seen between the unaged (unweighted) silk and the two tin weighted silk
spectra, which have an increased peak intensity at 5180 and 6970 cm-1.
Although outside the limited spectral region used in the current model,
weighting appeared to have some influence. A future model could include

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
7
weighted samples which may improve both the prediction of condition
and help identify weighted silk within collections.
Figure 5
Effect of inorganic elements on NIR spectra (unaged silk: red; weighted samples: blue and purple)
Although the predicted values of the tensile strength are greater than would
be expected, the samples were successfully ranked in order of condition. For
example, those samples in the poorest visible conditions also had the lowest
predicted tensile strength values. The current model cannot be used to give
exact values of condition, but it can provide comparative ranking, which
could help in prioritising future interventive treatment. NIR spectroscopy
offers rapid, non-destructive and non-invasive silk characterisation, although
its application may be constrained by the requirement for a comprehensive
reference spectral dataset and the generation of a robust condition model.
CONCLUSION
The research has demonstrated the feasibility and illustrated the value of
non-destructive analysis of textiles on open display. XRF has been found
to be suitable for in situ analysis of textiles and provides a qualitative
identification of silks treated with metallic salts. It can also be used to
determine the composition of metal threads whilst on display. The types
of textile can be identified using NIR spectroscopy with comparison
against a suitable database, without removing samples for analysis.
Furthermore, the relative condition of the silks can be gauged by NIR/
MVA. Appropriate extension of the NIR reference dataset should enable
prediction of absolute values for condition related parameters such as
tenacity and fibroin molecular weight, offering more detail and better
informing curatorial and conservation strategies.
ACKNOWLEDGEMENTS
The authors would like to thank the curators, staff and volunteers at
Brodsworth Hall and Audley End House for facilitating the in situ visit.

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
8
We are grateful to the AHRC for the Doctoral Award funding of N. Luxford
(2006–2009).
REFERENCES
BALLARD, M.W., R.J. KOESTLER, AND N. INDICATOR. 1990. Recent results
concerning the degradation of historic silk flags. In ICOM-CC 9th Triennial Meeting
Preprints, Dresden, German Democratic Republic, 26–31 August 1990, 277–282. Paris:
International Council of Museums.
BECKER, M.A., S.P. HERSH, and P.A. TUCKER. 1987. The influence of tin weighting
agents on silk degradation. Part 1: inorganic weighting and mordanting agents in some
historic silk fabrics of the 18th and 19th centuries. In ICOM-CC 8th Triennial Meeting
Preprints, Sydney, Australia, 6–11 September 1987, ed. K. Grimstad, 339–344. Paris:
International Council of Museums.
GARDINER, J., J. CARLSON, L. EATON, and K. DUFFY. 2000. “That fabric
seems extremely bright”: non-destructive characterization of nineteenth-century mineral
dyes via XRF analysis. In Conservation Combinations: Preprints North American Textile
Conservation Conference 2000, Asheville, North Carolina, March 29–31 2000, 100–115.
Asheville, N.C.: Biltmore House.
HACKE, M. 2008. Weighted silk: history, analysis and conservation. Reviews in
Conservation 9: 3–15.
KAWANO, S. 2002. Sampling and sample presentation. In Near-infrared spectroscopy
principles, instruments, applications, ed. H.W. Siesler, Y. Ozaki, S. Kawata, and H.M.
Heise, 115–124. Weinheim: Wiley-VCH.
KIM, J.-J., and P. WYETH. 2009. Towards a routine methodology for assessing the
condition of historic silk. e-Preservation Science 6: 60–67.
LUXFORD, N. 2009. Reducing the risk of open display: optimising the preventive conservation
of historic silks. Ph.D. thesis, University of Southampton, United Kingdom.
MASSCHELEIN-KLEINER, L., and L.R.J. MAES. 1978. Ancient dyeing techniques
in Eastern Mediterranean regions. In ICOM-CC 5th Triennial Meeting Preprints, Zagreb,
1–8 October 1978, 78/9/83. Paris: International Council of Museums.
ODEGAARD, N., D.R. SMITH, L.V. BOYER, and J. ANDERSON. 2006. Use of
handheld XRF for the study of pesticide residues on museum objects. Collection Forum
20: 42–48.
RICHARDSON, E., G. MARTIN, P. WYETH, and X. ZHANG. 2008. State of
the art: non-invasive interrogation of textiles in museum collections. Microchimica Acta
162: 303–312.
RICHARDSON, E., and P. GARSIDE. 2009. The use of near-infrared spectroscopy
as a diagnostic tool for historic silk artefacts. e-Preservation Science 6: 68–74.
SKELTON, M. 1995. Use of qualitative non-destructive X-ray fluorescence analysis in the
characterization of Chinese or Western provenance for 18th-century painted/printed silk.
In The conservation of 18th-century painted silk dress, ed. C. Paulozik, and S. Flakerly,
6–15. New York: Metropolitan Museum of Art.
ZHANG, X., and P. WYETH. 2007. Moisture sorption as a potential condition marker
for historic silks: non-invasive determination by near-infrared spectroscopy. Applied
Spectroscopy 61(2): 218–222.

TEXTILES
NON-DESTRUCTIVE
TESTING OF SILK:
PROBLEMS AND POSSIBILITIES
9
MATERIALS LIST
KeyMaster-Bruker TRACeR III-V handheld XRF and KeyMaster-Bruker vacuum pump:
Mike Dobby, Portable XRF Applications Manager (Europe), Bruker Advanced X-Ray
Solutions Banner Lane, Coventry, CV4 9GH, UK
PerkinElmer Spectrum One NTS spectrometer with Axiom probe.
Perkin Elmer Life and Analytical Sciences
Chalfont Road, Seer Green, HP9 2FX, UK
Unscrambler version 9.7
CAMO Software AS.
Nedre Vollgate 8, N-0158, Oslo, Norway