Development of portable microfading spectrometers for measurement of light sensitivity of materials

Full text

SCIENTIFIC RESEARCH
1
Keywords: preventative conservation,
accelerated aging, reciprocity, colour
change, microfadometry
ABSTRACT
Microfading was originally designed for ef-
ficiently detecting extremely light-sensitive
materials on objects in situ to determine the
appropriate exhibition lighting conditions. By
focusing an intense beam of light to a tiny sub-
millimetre sized spot and simultaneously mon-
itoring the colour change over time, the fading
rate of the material can be measured without
producing noticeable damage. The increased
intensity of light allows rapid determination of
light-fastness of materials. This paper examines
an improved design of microfading spectrom-
eter that is easy to assemble, compact, robust,
capable of fully automatic acquisition of data
with precision control of the fading time to
produce higher precision measurements and
to allow simultaneous monitoring of colour,
spectral reflectance and other changes in real
time. The effects of various parameters such as
thickness and concentration of paint layer, the
binding medium and substrate on the fading
rates are examined for selected pigments and
found that in certain cases substrates, bind-
ing media and thickness can affect the fading
rate. Reciprocity in the context of microfading
compared with realistic exhibition conditions
is examined and found that it breaks down for
some pigments.
RÉSUMÉ
La microdécoloration a été conçue à l’origine
pour détecter de manière efficace les maté-
riaux extrêmement sensibles à la lumière sur
des objets in situ afin de déterminer les condi-
tions d’éclairage appropriées pour leur expo-
sition. En dirigeant un faisceau de lumière in-
tense sur un minuscule point de moins d’un
millimètre et en surveillant simultanément le
changement de couleur au fil du temps, il est
possible de mesurer la vitesse de décolora-
HAIDA LIANG*
Nottingham Trent University
School of Science & Technology
Nottingham, UK
haida.liang@ntu.ac.uk
REBECCA LANGE
Nottingham Trent University
School of Science & Technology
Nottingham, UK
rebecca.lange@ntu.ac.uk
ANDREI LUCIAN
Nottingham Trent University
School of Science & Technology
Nottingham, UK
andrei.lucian@ntu.ac.uk
PAUL HYNDES
Nottingham Trent University
School of Science & Technology
Nottingham, UK
JOYCE H. TOWNSEND
Tate Britain
London, UK
joyce.townsend@tate.org.uk
STEPHEN HACKNEY
Tate Britain
London, UK
stephen.hackney@tate.org.uk
*Author for correspondence
DEVELOPMENT
OF PORTABLE
MICROFADING
SPECTROMETERS
FOR MEASUREMENT
OF LIGHT SENSITIVITY
OF MATERIALS
INTRODUCTION
Microfading is a micro-destructive technique that makes it possible to
examine the light sensitivity of materials in situ, which is particularly
useful when the exact composition of the material is not known. It has
been used in the conservation field to compare rapidly the rate of fading
of known pigment samples or other coloured materials and also to measure
directly unidentified pigments on works of art that might fade if exposed
to excessive light on display. Since the method causes hardly discernable
damage on a micro level to only the most sensitive pigments, it is potentially
a very useful way of pre-empting more extensive damage on display.
Ten years since the invention of the first microfadometer (Whitmore et al.
1999), a new microfading spectrometer with improved portability (a few
kilograms) and accuracy was built in a collaboration between the Tate
and Nottingham Trent University, taking advantage of the availability of
compact light sources and portable fibre optic spectrometers (Lerwill et al.
2008). The instrument consists of a probe head with a 0/45° geometry
using identical focusing optics in both illumination and collection. The
illumination spot is slightly smaller than the measuring spot since the
collection probe is at 45° to normal. The probe head needs to be aligned
such that the illumination and collection spots are coincident when in
focus. The probe head is placed on a motorized X-Y-Z stage such that
the Z focusing is controlled by the motorized micrometer stage ensuring
accurate focusing and hence high precision estimates of incident power
per unit area. The motorized X-Y stage allows automated measurements
of various pre-determined positions on an object or a well plate. Detailed
design of the instrument was published by Lerwill et al. (2008). The
repeatability of the instrument is better than ∆E00 = 0.1 for repeat colour
measurements of the same spot (Lerwill 2011). The accuracy of fading
measurements is limited by intensity fluctuations of the light source which
is less than 1% in 7 hours. Heating by the focused light increases the
temperature by only a couple of degrees centigrade.
While the instrument has improved portability, the proper alignment of
the probe head when re-assembled each time can be a non-trivial task.
In addition, the motorized stages and the spectrometer are not computer
controlled by the same program, the shutter is controlled manually and
the focussing is not automated. This paper describes a new upgraded

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
2
tion du matériau sans produire de dommage
visible. L’intensité accrue de la lumière permet
de déterminer rapidement la résistance à la
lumière des matériaux. Cet article examine la
version améliorée d’un spectromètre de mi-
crodécoloration facile à assembler, compact,
robuste, qui permet l’acquisition entièrement
automatique des données avec un contrôle
précis de la durée de la décoloration, afin de
fournir des mesures de précision supérieure
et de permettre un contrôle simultané des va-
riations de couleur, de la réflectance spectrale
et d’autres changements avec le temps. Les
effets de plusieurs paramètres comme l’épais-
seur et la concentration de la couche picturale,
le liant et le substrat sur les vitesses de déco-
loration ont été examinés pour une sélection
de pigments. Les résultats indiquent que dans
certains cas, les substrats, les liants et l’épais-
seur de la couche picturale peuvent influer sur
la vitesse de décoloration. La réciprocité dans
le contexte de la microdécoloration comparée
avec des conditions d’exposition réalistes a été
examinée, montrant que cela ne fonctionne
pas pour certains pigments.
RESUMEN
La microdecoloración se diseñó originalmente
para detectar de manera eficiente materiales
extremadamente sensibles a la luz en objetos
in situ, y poder determinar las condiciones de
iluminación adecuadas para su exposición.
Enfocando un rayo de luz intensa en un pe-
queño punto de menos de un milímetro, y
monitoreando simultáneamente el cambio
de color a lo largo del tiempo, se puede medir
la tasa de decoloración del material sin causar
daños visibles. La intensidad aumentada de
luz permite determinar rápidamente la resis-
tencia de los materiales ante la luz. Este artí-
culo analiza un diseño mejorado de espectró-
metro de microdecoloración, fácil de montar,
compacto, robusto y capaz de adquirir datos
de manera totalmente automática con control
de precisión del tiempo de decoloración, que
permite obtener medidas de mayor precisión
y hacer un monitoreo simultáneo del color,
la reflectancia espectral y otros cambios en
tiempo real. Se examinaron los efectos que
varios parámetros, como el grosor y la concen-
tración de la capa de pintura, el aglutinante y
el sustrato en algunos pigmentos selecciona-
dos, y se descubrió que en ciertos casos, los
sustratos, los aglutinantes y el grosor pueden
afectar a la tasa de decoloración. Se estudió
la reciprocidad en un contexto de microde-
coloración comparado con condiciones de
exposición realistas y se descubrió que en el
caso de algunos pigmentos se rompe.
microfading spectrometer with improved portability and simplicity in the
probe design as well as being fully computer controlled.
THE UPGRADED MICROFADING SPECTROMETER
Figure 1 shows the latest upgrade to the microfading spectrometer.
The probe head is re-designed so that it operates in retro-reflection
mode hence avoiding the need for alignment between the illumination
and collection probes. As a result, the probe head is significantly more
robust and compact; measuring only 12 cm by 4 cm. The motorized
focusing stage, the X-Y stage, the shutter for the light source and the
spectrometer are all computer-controlled by the same program ensuring
synchronization of the measurements with the light source shutter and
hence improved accuracy in the fading measurements.
Similar to the instrument described by Lerwill et al. (2008), a high-powered
continuous-wave xenon light source (Ocean Optics HPX2000) is used
with a filter that cuts off the ultraviolet and near infrared radiation.
An Ocean Optics HR2000+ portable fibre optics spectrometer is used
instead of an Avantes spectrometer. The two brands of spectrometers have
similar designs and the choice of the Ocean Optics spectrometer is for
compatibility with the Ocean Optics light source in order to synchronize
control of the shutter and the spectrometer.
Auto-focussing is achieved by attaching the probe to a computer-controlled
motorized linear micrometre stage and finding the position when the
counts detected from the reflected light is a maximum. The light intensity
is always reduced by an attenuator during focusing. The user is able to
monitor the colour and spectral changes online.
Fading rate and degradation rate
It is the convention to measure the light-sensitivity of a material through
monitoring the colour change ∆E over the time of exposure, which is
useful as an indication of how noticeable the degradation is. However,
colour change does not correspond to the rate of degradation linearly and
nor does it correspond linearly with the measured spectral reflectance.
This paper presents not only the colour change but also the change in
spectral reflectance ∆R=R(t)-R(0) averaged over the wavelength range
where the change is occurring. The most perceptually uniform colour
space is CIE2000, with colour difference expressed as ∆E00 (Luo et al.
2001, Sharma et al. 2005). However, for convenience of calculation, ∆Eab
(also called ∆E76) (CIE 1986) and ∆E94 (CIE 1995) are still commonly
used within the conservation community (Druzik 2010), although they
are perceptually less uniform, which means that a ∆E76 value might be
perceptible for some colour, but not for others. Here the authors will
use ∆E00 throughout the paper but give comparisons in ∆Eab and ∆E94
for fading of common standards like the ISO blue wool series, of which
standards BW1, BW2 and BW3 have been used as comparators by earlier
(Whitmore et al. 1999) and current microfadometer researchers.

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
3
System accuracy and repeatability
The stability of the system was examined in detail and found to have a
systematic drift of ~2% over 60 hours of continuous run. The drift had
the greatest rate of change of 0.5% in the first hour of the spectrometer
taking measurements. This indicates that the initial drift was partly
due to the spectrometer warming up as the steeper slope at the start
of taking measurements was observed even after the lamp had been
on for a number of hours. It is important to note that the relationship
between the observed drift in counts (or reflectance) and colour change
∆E is not linear. In the following sections, the drift in ∆E00 is simulated
for each material based on their initial spectral reflectance and the
observed system drift measured from the light reflected off a stable
ceramic white tile.
Accuracy of auto-focus was tested on a standard matt ceramic white tile.
The peak intensity has a plateau over a distance of 40 microns around
the focus position. The accuracy at finding the focus position is better
than 20 microns. Re-focussing on the same spot (i.e. not moving the
X-Y stage) using auto-focus gave accuracies of 0.2% in reflectance
for the standard matt white tile and 0.06% for a standard matt ceramic
black tile.
As with any microfading technique, if the typical roughness scale of the
material is of the order of the spot size, then the surface texture would
cause variations in spectral reflectance and colour across the surface.
Conventional spectrometers and colorimeters have measurement spot
diameters of 3 mm to 8 mm, whereas typical spot sizes in microfading
are 0.25 mm to 0.5 mm full-width half-maximum (FWHM).
Spot size and efficiency of the system
In order to measure the diffuse reflection from a material, the probe is
set up at 45° to the surface of the sample in a retro-reflection geometry.
Figure 2 shows a profile of the incident spot across the major and
minor axes as measured by a CCD camera. The minor axis of the spot
is ~0.46 mm FWHM similar to the original Whitmore et al. design
(~0.4mm) but larger than the ~0.25mm in our previous design given in
Lerwill et al. (2008). The total power over the focused spot is ~2 mW
and the average intensity is ~7 kW m-2 over the elliptical spot of 0.46 mm
by 0.76 mm. It is ~7 times less in intensity than our previous design
(Lerwill et al. 2008), i.e. ~2x106 lux, and ~3 times less than the original
Whitmore et al. instrument (1999).
Figure 3 shows the change in the mean spectral reflectance ∆R, the rate
of change dR/dt of BW2, as well as the colour change corresponding to
the evolution of the spectrum averaged over three fading measurements.
BW2 was found to reach ∆E00 ~0.7 after 20 minutes. The same colour
change was reached in 2.5 minutes (Whitmore et al. 1999), 1.5 minutes
(Lerwill et al. 2008) and 4 minutes (Druzik 2010) in different instruments.
Figure 1
A picture of the upgraded microfading
spectrometer showing the light source
(blue box), spectrometer (black box above
the light source), the probe on the right
attached to a motorised linear stage, the
input and output fibre optics attached
to the probe and a sample placed on a
motorised stage
Figure 2
Profiles of the focussed spot along the
minor axis (green line) and the major axis
(black line)

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
4
Being slower means it is less efficient, but by reducing the intensity by
almost an order of magnitude means it is more likely to yield realistic
results closer to exhibition conditions. Using ∆R or dR/dt plot as a
guide, the lightfastness of BW2 can be determined within the first
minute when dR/dt is greatest. It is easier to use ∆R than ∆E for the
determination of lightfastness of materials, because of the simplicity in
the error estimates associated with ∆R. ∆R measures the initial rate of
degradation as well as providing a higher signal to noise measure than
dR/dt by calculating the accumulated change over time.
The measurement integration time was typically 7ms and the number of
spectra averaged was 10. Increasing the number of averages beyond 10
has little effect on the signal-to-noise ratio. For comparison, the rate of
change in spectral reflectance due to the initial hour of drift corresponds
to 1.4x10-4 percent per second and the effect on the colour change due
to the drift of the system was found to produce a change of ∆E00 = 0.06
for BW2 over the hour.
The large spread in BW2 measurements is due to the surface texture of
the wool which was found to be ~ 200 microns in height and 800 microns
in the lateral direction (same order as the size of the focused spot)
using optical coherence tomography (OCT) (Liang et al. 2005, Spring
et al. 2008). The sample was placed between two glass microscope
slides in order to reduce the surface texture. The fading was reduced
to ∆E00 ~0.4 after 20 minutes when measured through the 1 mm thick
glass microscope slide.
PARAMETERS AFFECTING MICROFADING
The effects of different substrates, thickness of the paint, shade (or
concentration of the pigment) and binding medium are examined to
understand to what extent these parameters affect the fading rates. Samples
were painted out on glass microscope slides, waterleaf paper and filter
paper. In the following experiments, samples on paper were clamped
between two glass microscope slides to keep them flat. The experiments
were conducted in a temperature controlled lab at ~22°C.
Effect of substrate
The stability of the substrates, waterleaf paper and filter paper, were tested
first. The spectral reflectance of the two types of paper is fairly similar
with average reflectance of ~67% within the visible range (400-700nm).
Figure 4 shows that filter paper is more stable than waterleaf paper.
To examine the effect of substrate on the fading of a pigment, a sample
(Tate Gallery Archive 7315.7 TTB6) of Prussian blue from the studio
pigments of J.M.W. Turner (1775–1851) was mixed in gum Arabic and
painted on waterleaf paper, filter paper and a glass microscope slide. The
average spectral reflectance of the paint on filter paper and waterleaf
paper were similar at ~14%, but the paint on glass placed over a white
Figure 3
Direct fading of BW2 (green symbols) and
BW2 clamped between glass microscope
slides (blue symbols). Top: evolution of
∆R in the wavelength range of 450nm to
490nm in units of percentage reflectance;
the error bars are plus and minus one
standard deviation and the dotted line is
the system drift. Middle: evolution of dR/dt
in the same wavelength range as above (in
units of percentage reflectance per second).
Bottom: the corresponding evolution of
colour difference (circles indicate ∆E00,
crosses indicate ∆E94, squares indicate ∆Eab);
the black line close to zero is the expected
colour change from the system drift in ∆E00
units
Figure 4
Fading of waterleaf paper (magenta) and
filter paper (black). Top: evolution of ∆R for
waterleaf paper averaged over 400-500nm
and filter paper averaged over 400-700nm;
the dotted line corresponds to the system
drift. Bottom: the corresponding colour
change; the dotted line show the simulated
colour changes due to the system drift

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
5
background was fainter at ~6%. Figure 5 shows that it degrades faster
on waterleaf and filter paper than on glass. In all cases, rate of change
is greatest at the beginning of the exposure.
Effect of thickness and concentration of paint
The initial average spectral reflectance over the 400 to 700nm range for
the two shades of Prussian blue painted in gum Arabic on filter paper
range between 9 to 14% and between 35 to 41% for the darker and
lighter shades respectively. Figure 6 shows that in this case the fading
rate is independent of the concentration of the pigment.
Orpiment mixed in linseed oil and painted on glass microscope slides
in varying layer thickness was tested. The OCT measured thicknesses
were 400 microns, 320 microns and 100 microns. Figure 7 shows that
the fading rate is the same for the two thicker samples but slower for the
thinnest sample. The samples were painted out three years ago and kept
in dark storage. However, the thinner sample appears to have started to
degrade over the years. It was noticed that the same orpiment pigment
kept in a glass bottle had started to degrade, since those pigments next
to the glass have started to turn orange. The pigment powder pressed
between two glass microscope slides was also tested and found to fade
differently than those mixed with linseed oil and painted out on glass
microscope slides. The thickness of the pigment powder was found to
be 400 microns from OCT images. The difference between the final
and initial spectra also showed that the thinnest sample and the powder
sample responded differently compared to the two thicker samples
after the same amount of exposure. Note that the turning point in the
reflectance spectrum of the orpiment sample is at ~550nm.
Reciprocity
The validity of accelerated aging methods depends on the reciprocity
principle to a large extent. The reciprocity principle states that the
amount of degradation only depends on the total energy that the sample is
exposed to. Microfading spectrometers usually operate at light intensity
levels that are at least 4-5 orders of magnitude greater than exhibition
lighting. The intensity of the current instrument is about 4 orders of
magnitude more intense than exhibition lighting. Reciprocity principle
over 3 orders of magnitude was tested on a 400 micron thick paint of
orpiment in linseed oil. Orpiment was chosen because it fades fast
and has been tested for reciprocity in conventional accelerated aging
experiments using light boxes and found to obey the reciprocity principle
for light intensities between 80 and 8000 lux judging by ∆Εab (Saunders
and Kirby 1996). Figure 8 shows that reciprocity principle breaks down
for the orpiment sample where the reaction pathway is different for the
different light intensities. The degradation appears to be slowed down
for intensity levels between 2x104 and 2x105 lux. It should be noted
that the orpiment sample used here and the one used in Saunders and
Figure 5
Top: evolution of ∆R (averaged over the
wavelength range of 440nm to 480nm) of
TTB6 (Prussian blue) in gum Arabic on glass
(red), filter paper (blue) and waterleaf paper
(green). Bottom: the corresponding colour
change; the expected effect due to the
system drift is shown as a black line
Figure 6
A sample of TTB6 in gum Arabic painted
in two shades on filter paper (the blue line
shows the darker shade and the green
line shows the lighter shade). Top: ∆R
(averaged between 440nm and 480nm) in
units of percentage reflectance. Bottom:
corresponding colour change; the expected
effect on colour change due to the system
drift is shown in solid black

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
6
Kirby (1996) are from different manufacturers. Tests of Prussian blue
also showed that reciprocity breaks down.
CONCLUSIONS
The advantage of the latest upgrade to the microfading spectrometer is the
automation such that all parts of the instrument are centrally controlled by
a laptop and that the probe is more robust, smaller in size and easy to use
with no need for alignment. These improvements increase the portability
and user-friendliness which can potentially increase the use of microfading
tests to assist conservation management decisions.
The evolution of the change in average spectral reflectance ∆R over a spectral
region where most of the change occurs can be used as an alternative to
monitoring degradation rate through colour change. It is easier to understand
the statistical characteristics of ∆R than ∆E. Since degradation of material
is independent of human vision, there is no real advantage in monitoring
∆E other than noting the visibility of the damage.
The reciprocity principle was tested on an orpiment sample and found
to break down. Since reciprocity is most likely to break down at the
highest intensities typical for microfading, it is both efficient and important
to examine reciprocity at the top one or two orders of magnitudes in
intensity for a significant sample of common pigments in the future. It is
likely that many pigments do not follow the reciprocity principle at these
high intensities, but microfading is still likely to provide a prediction for
light induced degradation that results in conservative decisions for light
exposure.
ACKNOWLEDGEMENTS
The authors would like to thank Simon Godber for technical assistance,
Jo Kirby and Jim Druzik for valuable discussions, Marika Spring for the
orpiment paint sample, Nottingham Trent University Stimulating Innovation
for Success award and UK Department of Innovation and Skills Public
Sector Research Exploitation (PSRE) for funding to Tate 2006-09.
REFERENCES
CIE. 1986. Colorimetry. 2nd Ed. CIE Publ. No. 15.2. Vienna: Central Bureau of the CIE.
CIE. 1995. Technical report: industrial colour-difference evaluation. CIE Pub. No. 116.
Vienna: Central Bureau of the CIE.
DRuzIk, J. 2010. Evaluating the light sensitivity of paints in selected wall paintings at
the Mogao Grottoes: Cave 217, 98 and 85. In Conservation of Ancient Sites on the Silk
Road, Dunhuang, June 2004, ed. Neville Agnew, 457–463. Getty Publications.
LERwILL, A. 2011. Micro-fading spectrometry: an investigation into the display of traditional
watercolour pigments in anoxia. Ph.D. Dissertation, Nottingham Trent University, UK.
LERwILL, A., J. TowNSEND, H. LIANg, J. THomAS, and S. HACkNEy.
2008. Portable microfading spectrometer for versatile lightfastness testing. e-Preservation
Science 5: 17–28.
Figure 7
Fading of orpiment in linseed oil at various
thickness (400µm in black, 320µm in
green and 100µm in blue) and a 400µm
layer orpiment powder packed between
two glass microscope slides (in red). Top:
evolution of ∆R averaged between 520 and
545nm. Middle: colour change. Bottom:
difference spectra between the final
spectrum and the initial spectrum
Figure 8
Reciprocity test on a 400µm thick layer
of orpiment in linseed oil paint on a glass
substrate using 100% (black line), 10%
(green line), 1% (blue line) and 0.1% (red
line) of the total intensity (~2x106 lux).
Top: ∆R between 520 and 545nm in units
of percentage reflectance as a function
of incident energy; note that 80mJ is
equivalent to ~2x104lux·h. Middle: colour
change as a function of total energy of
exposure; the dotted lines correspond
to simulated effects due to system drift
for fading periods associated with each
intensity levels. Bottom: difference spectra
after a dose of 55mJ and the initial spectrum

SCIENTIFIC RESEARCH
DEVELOPMENT
OF PORTABLE MICROFADING
SPECTROMETERS
FOR MEASUREMENT OF LIGHT
SENSITIVITY OF MATERIALS
7
LIANg, H., m. gomEz CID, R.g. CuCu, g.m. DobRE, A. g. PoDoLEANu, J.
PEDRo, and D. SAuNDERS. 2005. En-face optical coherence tomography – a novel
application of non-invasive imaging to art conservation. Optics Express 13: 6133–6144.
Luo, m., g. CuI, and b. RIgg. 2001. The development of the CIE 2000 colour-
difference formula: CIEDE2000. Color Research & Application 26: 340–350.
SAuNDERS, D., and J. kIRby. 1996. Light-induced damage: investigating the
reciprocity principle. In ICOM-CC 11th Triennial Meeting Preprints, Edinburgh, 1–6
September 1996, ed. J. Bridgland, 87–90. London: James and James.
SHARmA, g., w. wu, and E. DALAL. 2005. The CIEDE2000 color-difference
formula: implementation notes, supplementary test data, and mathematical observations.
Color Research & Applications 30(1): 21–30.
SPRINg, m., H. LIANg, b. PERIC, D. SAuNDERS, and A. PoDoLEANu.
2008. Optical coherence tomography – a tool for high resolution non-invasive 3D-imaging
of the subsurface structure of paintings. In ICOM-CC 15th Triennial Meeting Preprints,
New Delhi, 22–26 September 2008, ed. J. Bridgland, 633–640. Paris: International Council
of Museums.
wHITmoRE, P., X. PAN, and C. bAILLIE. 1999. Predicting the fading of objects:
identification of fugitive colorants through direct non-destructive lightfastness measurements.
Journal of the American Institute of Conservation 38: 395–409.
Post publications notes: The orpiment from Beijing pigment factory referred to above was later
analysed by XRD and Raman and found to be a mixture of realgar and its degradation product
pararealgar.