Creagh D., Bradley D. (Eds.) Physical Techniques in the Study of Art, Archaeology and Cultural Heritage. Volume 2

Подождите немного. Документ загружается.

4.1.3. Application 1: blue paint

Identifying blue paints is important to the paintings conservator when choosing the right

colour for retouching (Morgans, 1982; Berns and Imai, 2002; Day, 2003a; van der Weerd

et al., 2003; Clarke, 2004). Blue, in particular, is difficult to match, and choosing the incor-

rect pigment can result in severe metamerism. Pigment identification was tested using

indigo, Prussian blue, light and deep cobalt blue, cobalt turquoise, cerulean blue, ultrama-

rine, phthalocyanine blue, and phthalocyanine turquoise in different binding media.

∑ Indigo blue is dark blue and fade prone. Indigo is found as glucoside indican on leaves

of Isatis tinctoria (woad) and Indigofera tinctoria from India (McLaren, 1983). It has

been exported to Europe from the twelfth century (McLaren, 1983). Synthetic indigo

from o-nitrophenyl acetic acid has been available since 1878.

∑ Prussian blue is a synthetic pigment developed in 1704 by Dippel and Diesbach. Also

called iron blue, it consists of ferric ferrocyanide (Fe

[Fe(CN)

]

) (Barnes, 1939b).

∑ Cobalt blue deep and light were isolated in 1735, but not developed as a pigment until

1802 to replace the similarly coloured smalt (Bergen, 1986). This blue contains oxides

of cobalt and aluminium

∑ Cobalt turquoise is a variety of cobalt blue additionally containing chromium.

∑ Cerulean blue is paler and greener than cobalt blue, and is an artificial pigment that

consists of oxides of cobalt and tin (Johnston-Feller, 2001).

∑ Ultramarine (French) is an artificial substitute for the expensive natural ultramarine

derived from lapis lazuli (Tate, Paint and Painting, 1982). It was discovered in 1826, but

not commercialised until 1830 by Guimet and Koetting (Bergen, 1986).

∑ Phthalocyanine blue was prepared by Von Diesbach and Von der Weid in 1927

(Loebbert, 1992). Its exceptional stability, lightfastness, and resistance to acids, alkalis,

and heat led to copper phthalocyanine being commercialised as a pigment by ICI in

1935, with IG Farben and DuPont starting soon after. Winsor and Newton produces

phthalocyanine blue under the name Winsor blue (Tate, Paint and painting, 1982).

∑ Phthalocyanine turquoise is a mixture of phthalocyanine blue and phthalocyanine

green (C

CuN

Reference spectra were collected directly from these common blue paints seen in Fig. 25.

As seen with the yellow pigments, the nine blue pigments separate easily, based on both

different value and shape reflectance peaks in the blue region (400–450 nm), and different

levels of reflection in the infrared (Fig. 25). Most pigments reflect light in the blue region

of the spectrum and absorb well into the red region. With the exception of phthalocyanine

blue, phthalocyanine turquoise, and Prussian blue, all blues show an upwards slope at 780 nm.

Cobalt blue deep and cobalt blue light may only be separated by percentage reflectance

peaks at 450 nm. Some of the deep blue/black pigments, Indigo in particular, extinguishes

the incoming beam in the visual region to a large extent, but the intensity of the long-

wavelength reflection of the spectrum (700–750 nm) is high. The deep colour of Prussian blue

is explained by its low reflectance curve, which reflects little light between 400–1000 nm.

This reflection spectrum is almost featureless. Some of the spectra were quite close in

shape and were therefore difficult to separate by spectrum (e.g. cobalt blue light and cobalt

blue deep). These results are confirmed by van der Weerd et al. (2003) and Berns and

)

⋅n

(CoO Al O ).

⋅n

232 M. Kubik

Imai (2002), who also found blues easy to separate. Berns and Imai incorrectly identified

manganese blue as phthalocyanine blue, which may have been due to similar incorrect

labelling of pigments.

Blue pigments also provided the opportunity to measure CIE L*a*b* colours using

Thermo Grams AI©. This program converted reflectance spectra using 2∞ observer angle

and standard illuminant CIE A, and calculated CIE L*a*b* values between 400 and

830 nm (Table 2). This demonstrates how easily CIE L*a*b* values are determined, useful

for recording colour change in an objective manner.

Hyperspectral Imaging: A New Technique for the Non-Invasive Study of Artworks 233

Fig. 25. Reflectance spectra of representative blue paints.

Table 2. CIE L*a*b* values derived from reflectance spectra of blue paints

Paint L* a* b*

Ultramarine blue oil 24.005 5.356 -43.318

Ultramarine blue tempera 29.673 3.576 -72.191

Ultramarine gamblin 25.130 7.061 -47.966

Cobalt blue deep tempera 29.530 -0.123 -54.090

Cobalt blue gamblin 31.121 1.450 -60.498

Cobalt blue light tempera 45.237 -6.571 -61.429

Prussian blue tempera 12.402 3.004 -14.658

Prussian blue gamblin 19.770 0.072 -11.424

Indigo oil 25.690 2.008 0.395

Manganese blue gamblin 40.015 -27.707 -54.071

Phthalo blue oil 23.443 -1.582 -12.553

4.1.4. Application 2: gouache painting in blue

To test the identification capacity for blue paints, the system was applied to a watercolour

and gouache painting made by the author (Fig. 26). This also validated methodology before

field testing. A number of blue Artist Spectrum (Australia) gouache paints and Windsor and

Newton (UK) watercolours were combined into a single portrait of a lady. Schmincke refer-

ence plates were used to attempt matching and identification of the “unknowns”. Figure 27

presents the location of these gouaches and watercolours that were tested.

All blues present in the reference set were identified (Figs. 28–32). Where overlap was pres-

ent between reference “known” paints and the “unknowns” of the painting, successful match-

ing was achieved, e.g. phthalocyanine blue (Fig. 31) and ultramarine (Fig. 28). Locations

for these identified paints mapped well across the image. While this is normally seen as

red highlights against a black and white image, reproductions in this chapter do not permit

colour and instead locations are mapped as bright white areas. Cerulean blue (Fig. 29) had

234 M. Kubik

Fig. 26. Blue painting with Schmincke reference plates.

Ultramarine

Cobalt Blue

Prussian Blue

Cerulean Blue

Zinc White

Turquoise

Marine Blue

Fig. 27. Location of Gouache and watercolour “unknowns”.

Hyperspectral Imaging: A New Technique for the Non-Invasive Study of Artworks 235

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Fig. 28. Ultramarine locations and spectrum.

Fig. 29. Cerulean blue locations and spectrum.

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Fig. 30. Indigo blue locations and spectrum.

236 M. Kubik

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

100

Wavelength (nm)

% Reflectance

Fig. 31. Phthalocyanine blue locations and spectrum.

Fig. 32. Prussian blue locations and spectrum.

been used in a diluted manner across the face and hands; however, the diluted spectrum of

these areas was not strong enough for identification. Limits were further presented where

reference spectra were not available for particular paints present, e.g. turquoise blue

(Fig. 33), marine blue, and cobalt blue. This highlights the need for either an extensive

reference set to be included in the analysed image, or a searchable library database of

reflectance spectra to be accessible for comparison.

4.1.5. Application 3: small war memorial paintings

With the assistance of David Keany, senior paintings conservator of the Australian War

Memorial, two studies in oil measuring 7 cm ¥ 12 cm were provided on loan for further

investigation (Fig. 34). These were deemed suitable due to their small size, and the appar-

ently large variety of blues employed.

The locations of pure Prussian blue could easily be identified by comparison of a refer-

ence plate in the same image (Figs. 35 and 36). Mixtures using lead white to obtain lighter

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Fig. 33. Turquoise blue locations and spectrum.

Fig. 34. Small war memorial painting (Wings) with selected blue reference plates.

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Fig. 35. Wings Prussian blue locations and reflectance spectrum.

shades of blue are further investigated. The blue pigment composition was confirmed

using Raman spectroscopy (Lee et al., 2005). Results of Raman analysis using the Kremer

database are presented in Table 3, while sample locations are given in Fig. 37. It appears

the artist has used a very simple palette: the lead white is almost ubiquitous, while all blue

areas consisted of various mixtures of Prussian blue and lead white.

238 M. Kubik

100

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Table 3. Location and identification of AWM plane pigments using Raman spectroscopy

Sample Description Identification

P1 Blue sea Prussian blue

P2 White cloud Lead white, small amount of Prussian blue

P3 Green earth Prussian blue + lead white + possibly kibeni

orange or priderite yellow

P4 Light brown ground Lead white, other inclusions seen but no spectra

generated.

P5 Medium blue Prussian blue

P6 Medium blue sky Prussian blue, lead white

P7 Green foreground Prussian blue + lead white + possibly kibeni

orange or priderite yellow

P8 Purple pants Prussian blue

P9 Pink man on wing Lead white

P10 Red hair on LH man Possibly kibeni orange or priderite yellow

P11 Red hair on LH man Possibly kibeni orange or priderite yellow

P12 Dark blue wing Prussian blue, small carbon peaks indicating

carbon black

P13 Light brown fabric coating Lead white

P14 Light blue Lead white, low amount of Prussian blue

Fig. 36. Plane Prussian blue locations and reflectance spectrum.

These dark blue regions have also been identified as Prussian blue using imaging spec-

troscopy. This shows that identification by imaging spectroscopy has been partially

successful in identifying the blue paint used, although lighter mixtures using lead white

have made this less than straightforward.

4.1.6. Application 4: Sidney Nolan

In order to test the portability and effectiveness of the instrument setup, the imaging

system was taken to the Australian War Memorial (Fig. 45). Here, it could be tried with

ambient lighting conditions and actual paintings with unknown pigment composition. It

also allowed for field testing, seeing how portable the instrumentation was and how easy

it was to carry and set up in a museum location with uncontrolled lighting. This was an

important consideration for the system’s design. Application proved to be very straightfor-

ward, with the entire system being able to be dismantled, carried by hand, and fitted in the

back of a car. The War Memorial venue consisted of an active conservation studio, with

diffuse natural light, fluorescent light, and incandescent lights, simultaneously. As in the

lab, the camera was positioned at 150 cm from the painting, which was sufficient to

capture the entire image. Projectors could be placed at slightly longer distances as the

allowed space was wider, in this case 120 cm compared to 90 cm. This affected maximum

exposure readings for the Spectralon reference, so that calibration had to be performed

before images were captured.

The Angry Penguins group was of particular interest, with little materials research

performed to date on some of Australia’s best known artists, e.g. Sydney Nolan (1917–1992).

This has led to the formulation of incorrect hypotheses about his materials, and incorrect

labelling in catalogues (Kubik, 2007). Nolan was unusually experimental in his approach

to paints, in part due to influence from overseas artists, but also caused by rationing and cost

of materials during the early stage of his career. Returning to interviews (Nolan, 1962; 1980;

1987) and letters written by Nolan (Nolan, 1943b; Nolan, 1943c; Nolan, 1943d; Nolan,

1943e; Nolan, 1947; Nolan, 1948) reveals what may be expected in materials analysis.

Hyperspectral Imaging: A New Technique for the Non-Invasive Study of Artworks 239

P6 P14 P2 P1

P11

P10

P9P4P3P7

P13

Fig. 37. AWM plane locations for Raman sampling.

In his letters, there are several requests by Nolan for a particular blue dye (Nolan, 1943f),

which has never been identified. By 1954, Nolan had switched to mixing his own pigments

into polyvinyl acetate binder (Llyn, 1967). Advantage was taken of the War Memorial’s

collection of Nolan paintings to shed light on these details. The following three paintings

containing blue paint were tested:

(1) Gallipoli Landscape, half-lit (1958): Synthetic polymer paint on coated card, ART

91243

(2) Cove at Hydra (1956): Synthetic polymer paint on coated card, ART 91261

(3) Gallipoli Landscape with gun (ca. 1963): Acrylic with yellow oil crayon on gloss art

board, ART 90217

All three contained unknown blue paint in either sky or sea, so that these could be tested

against known references and the library database created earlier. The blues also appeared

in different hues, so that several types of pigments were expected.

The first painting tested was Gallipoli Landscape, half lit. This showed dappled blue

paint, almost ultramarine in colour. Nolan’s technique at this stage involved wiping paints

using a squeegee and sponge (Moore, 1998). The painting was placed on an easel, and

projector lights placed at 1.0 m at 45∞ incident to the surface plane of the painting. As had

been established at ANU, the CCD camera was placed at 150 cm from the painting, which

was sufficient to capture the entire image, as well as six blue reference plates and the

Spectralon reflectance standard in the middle of these (Fig. 38). The positions of the blue

references remain consistent throughout the three Nolan paintings tested. The white

balance was calculated at 4000 maximum and the exposure times adjusted accordingly.

Images were captured sequentially between 400 and 1000 nm, and stacked into a data cube

240 M. Kubik

Cerulean

Prussian

Indigo

Ultramarine

Phthalocyanine

Cobalt

Fig. 38. Nolan’s Gallipoli Landscape, half lit with six reference plates of blue pigments

and the Spectralon standard.

using the Hypercube software. From this, identification could either be achieved by select-

ing the six standards and finding similarities within the image, or reversely by selecting

key points in the sky and allowing the search to find similarities within the reference plates.

After collating the images in Hyperspec and searching both the distribution of unknown

blue areas and the references, both indigo and phthalocyanine blue were found to be pres-

ent as a mixture (Figs. 39 and 40). The two different colours appeared thoroughly mixed

on the canvas, and were only discernable using imaging spectroscopy. This was therefore

a positive outcome, as it indicates the possibility of identifying individual components

within a mixture.

Hyperspectral Imaging: A New Technique for the Non-Invasive Study of Artworks 241

100

400

450

500

550

600

650

700

750

780

800

830

850

880

900

930

950

980

1000

Wavelength (nm)

% Reflectance

Fig. 39. Gallipoli Landscape, half lit locations for indigo (left) and phthalocyanine blue

(right).

Fig. 40. Reflectance spectra of indigo (dotted line) and phthalocyanine blue (solid line).