The Technical Examination of Nineteenth-Century Artists' Pigments from John Opie’s Paintbox

ABSTRACT

St Agnes Museum, Cornwall, currently holds in its collections a paint box with a metal plaque reading ‘John Opie 1806’, containing pigments and other artists’ materials. John Opie produced hundreds of oil paintings until his untimely death in 1807 and was a well-known and successful painter of portraits and historical scenes. However, his work has been little studied, in comparison to his contemporaries such as Joshua Reynolds. This paper discusses the identification of the pigments in the paint box using infrared spectroscopy (FTIR-ATR), optical microscopy, scanning electron microscopy with energy dispersive spectroscopy (SEM-EDX), Raman spectroscopy (SERS) and X-ray fluorescence (XRF). Seven pigments have been identified as ultramarine, lead white, fustic and carmine lakes, carbon black (likely ivory), hematite, and burnt sienna, consistent with the bottle labels where present. The identification has been further confirmed using historical literature, which has shown that the pigments are likely to be contemporary with the stated date of 1806. This information adds to the body of knowledge about the artist specifically, and the artists pigments generally available at the time.

KEYWORDS:

• Nineteenth century pigments
• artist materials
• heritage science
• Raman spectroscopy
• FTIR

Introduction

Studies of historic pigment composition can be used to enhance our understanding of artists’ materials from different time periods. Although many of these studies look specifically at paintings, other artists’ materials, such as painters’ palettes and paint boxes, can provide valuable evidence. This paper discusses the analysis of the composition of seven pigments from a paint box in St Agnes Museum, Cornwall. The box belonged to the English artist John Opie, and a metal plaque on the box states both his name and the date of 1806. Opie (1761-1807) was an English painter, who painted for members of the Royal Court including King George III (Hendra 2007, 37). However, since his death, his work has often been overshadowed by that of his contemporaries, such as Joshua Reynolds (1723–1792), and research into Opie’s paintings and techniques is very limited. This study aims to increase the knowledge around Opie’s use of pigments and also to add to knowledge about the use of artists pigments in the early nineteenth century.

John Opie was born in May 1761, near Truro in Cornwall. He showed a love of painting from an early age but was not encouraged in this by his father, a carpenter, who took him on as his apprentice. In 1775, Opie’s work came to the attention of a local doctor, John Wolcot (Rogers 1878, 10). From this point onwards, Wolcot encouraged Opie’s art and bought him out of his apprenticeship, to enable him to concentrate on painting (Earland 1911, 8). Despite receiving an extensive education during his time in Cornwall, on his move to London in 1781 Opie was presented by Wolcot as ‘The Cornish Wonder’, an untaught genius; though it is suggested this was a marketing ploy (Earland 1911, 28). Throughout his career, he used a strong chiaroscuro effect in his paintings to create bold contrasts between the figures and surroundings (MacLeod 2009). Opie’s career quickly took off, and he painted portraits for members of the court, including commissions for King George III. He was a prolific painter and produced at least 760 oil paintings, which are listed in Rogers catalogue of 1878, though Earland (1911) lists 1147 works. As well as portraits, which provided much of Opie’s initial work and income, Opie expanded his range of subjects to include historical scenes and ‘Fancy’ paintings, using everyday life but with added elements of storytelling, such as The Peasant’s Family (1783-5) currently held by the Tate Gallery, London (inv.no. N05834).

Opie’s opinions on pigments used for his artworks are difficult to ascertain from existing archival records, but a letter to one of his amateur pupils may give an indication of his opinions: ‘ … I believe one kind of yellow is as good as another, a little more or a little less white will make it what you want … ’ (John Opie to Rev. John Owen, 24 Dec 1789, quoted in Rogers 1878, xv). The tone of the letter overall reads as though it is intended to be light-hearted, and is perhaps not to be taken seriously, but it does suggest that he may not have been overly concerned with buying the most expensive pigments himself and was willing to mix pigments to create the desired result. Opie’s 1807 Royal Academy lecture on colour shows a change in his approach to the use of colour, where he claims in his fourth lecture, on colour, that ‘the less colours are mingled, the greater their purity and vivacity’ (Opie 1809, 164). This suggests a move away from his earlier assertation of just adding white, and pigments may be found in a relatively un-mixed form on his later paintings if his lecture is an accurate reflection of his painting style of the time.

John Opie was clearly willing to experiment and try unfamiliar pigments, as his letter in 1790 to Reverend Thomas Kerrich demonstrates:

I have had a quantity of a very fine pigment made me a present of today of which I shall send you a specimen soon it comes from the East Indies and I am told is the urine of some animal. Mr. Hodges the landscaper painter [likely William Hodges RA (1744-1797)] of whom I had it tells me he has us’d it these ten years and that it stands perfectly well if so it must be a great acquisition. (Opie 1790)

This letter is perhaps referring to an Indian Yellow, which was in use in England from the early 1700s, but grew in popularity in particular during the eighteenth century, due to its radiant colour (Bailkin 2005).

In 1782, Opie first exhibited his work at the Royal Academy (Rogers 1878, 25), and became an associate of the Royal Academy in 1787 (Earland 1911, 68). It was not until 1805 when a Professorship position became available, into which Opie was duly elected (Rogers 1878, 41). As a Professor, it was expected that he would deliver a series of six lectures. However, the last two lectures remained unwritten, as Opie died in the prime of life, 9 April 1807 at age 45. This was exactly one month after the completion of his first series of lectures at the Royal Academy covering design, invention, chiaroscuro, and colouring (Opie 1809, 81; Rogers 1878, 45). The cause of his death is uncertain, but it has been suggested that it was due to a chill caught on a visit to a friend, shortly after the completion of his last lecture (Earland 1911, 224). Modern medical knowledge suggests an infection, in combination with sepsis, as the potential cause of his death (Hendra 2007, 167). He was buried in St Pauls Cathedral next to Joshua Reynolds, an indication of the high regard in which he was held by his contemporaries (Rogers 1878, 47).

After Opie’s death, most of the contents of his house in Berners Street, London were sold at auction by his widow, Amelia (Hendra 2007, 175). However, it seems that the paintbox may have been retained, as it was subsequently donated to the museum in St Agnes by descendants of the Opie family. It is thought that this paintbox was presented by the Royal Academy of Arts after Opie was granted the position of Professor of Painting in 1805. The paintbox is made from mahogany and contains 14 pigments in glass vials, with cork stoppers, plus a glass pestle and mortar and measuring flask. All of the pigments appeared to have been at least partly used. Three larger glass bottles, with glass stoppers, are labelled as containing linseed, turpentine, and varnish. These have only minimal traces of their contents remaining. In order to examine the contents of the paint box, samples were removed from seven of the pigment bottles for analysis (Figure 1). The samples were limited to only those where the cork could be removed without causing damage. While four of the sampled pigments were named, with what appeared to be original labels on the outside, it is not unusual for there to be some variations between the expected contents, and the actual chemical composition, because of different manufacturing and supply methods (Carlyle 1995). Three further pigments were unlabelled, and the contents were unknown. Additional labelled bottles, which were not sampled, are labelled as Vermillion, Vandyke Brown, Light Red, Pn [Prussian] Blue, Raw Sienna, and Yw [Yellow] Ochre. All are loose powder pigments, apart from one bottle containing yellow drops. Field (1835, 83) states that yellow lakes ‘are usually in the form of drops’ i.e. a compacted and shaped dry pigment, though this pigment could not be analysed as the cork was too fragile to remove.

Figure 1. Pigments sampled in the study: (a) Unlabelled pink; (b) unlabelled black; (c) unlabelled blue; (d) bottle labelled as Indian Red; (e) Bt. Sienna; (f) F. White; and (g) Yw. Lake. Pigment images by Rhiannon Sinha.

There are a number of ways to study the materials used by artists. The paintings themselves are of course a vital source of information about techniques and materials. Another important source of information is the materials that originally belonged to artists, which were used in their work, such as Turner’s palette and pigments in the Tate collections (Townsend 1993; 2004). The study of documentary sources and the use of analytical techniques offer methods for gaining a more comprehensive understanding of the materials and techniques used by artists. Surviving artists’ papers may provide valuable documentary evidence about their working practices. Where specific archives are not present, publications around historic recipes and working techniques may provide additional information. In this study, the authors used a combination of analytical techniques combined with the use of historic recipes, in order to determine the chemical composition of seven of the pigment samples from Opie’s paintbox. This multiple methodology, combining results from a range of appropriate analytical techniques with historical literature, allows more robust conclusions to be drawn by combining findings from a range of sources.

Opie was a highly regarded painter in the late eighteenth to early nineteenth century, and painted portraits of many of the most significant figures in society. However, more recently his work has been overlooked and is very little studied. Studying the pigments from this box adds significant new information about the artist and his materials. Furthermore, increasing the knowledge of pigments used in this era is of great value to researchers working in this area.

Materials and methods

Sampling: Samples were taken from seven pigment bottles, four of which were labelled: F. White, Yw. Lake, Bt. Sienna, Indian Red, and three unlabelled bottles containing pink, blue, and black pigments. The cork was removed from the bottle, and a small amount of pigment was removed using a cocktail stick. The pigment was placed into a glass vial and sealed with a plastic stopper.

SEM-EDX: Studies were carried out with an Impact S SEM-EDX from FEI using the manufacturer’s software. Samples of loose pigment were mounted onto an SEM pin stub mount and conductive carbon adhesive tab and examined under vacuum.

FTIR-ATR: Samples were studied using a Perkin Elmer Spectrum 100 FTIR Spectrometer with a Specac Golden Gate diamond lens attenuated total reflectance (ATR) attachment, single bounce. The scan range was 4000–400 cm⁻¹ with a scan number of 32, a resolution of 4 cm⁻¹, and a scan speed of 0.5 cm s⁻¹ The resulting infrared spectra were compared to in-house reference spectra, obtained on the same machine, using the Euclidean algorithm (processing software: Perkin Elmer Spectrum, version 6.1.0.0038). Correlations were assessed using the processing software. The FTIR data presented in Figure 2 was obtained from the Opie sample and Flake White (Cornelissen) with an Agilent 4300 handheld FTIR with diamond ATR attachment and compared using their Microlab software. Scan range and number were used as above.

Figure 2. (a) Opie’s white lead pigment viewed using polarising light microscopy, showing birefringence, image width 160 µm; (b) FTIR-ATR spectra of Opie’s F. White (blue) and a comparison Flake White pigment (orange).

Microscopy: Optical and polarising light microscopy was carried out using a Leica DM2700M microscope and Motic BA310Pol microscope respectively. Images were produced using Moticam X camera in Motic Live Imaging Module software. Loose pigment was mounted onto glass slides with a coverslip, using Meltmount 1.662 (Cargille). A series of identification tests were performed using the polarising light microscope, following the procedures described by McCrone (1982) to characterise the refractive index (Becke test), and visual assessment of the pigment size and shape, relief, pleochroism, and birefringence.

XRF: A Handheld XRF Tracer IV System was loaned by the Bruker Corporation for this study. Readings were taken at 40 keV for 60 s, then peaks were manually assigned to elements using the manufacturers S1PXRF software.

Surface Enhanced Raman Spectroscopy (SERS): Due to the fluorescence of the pigment obscuring the spectra peaks, Raman spectroscopy tests were unsuccessful. Therefore, to reduce fluorescence, SERS spectra were obtained for a range of control samples and the pink and yellow Opie pigments. SERS spectra were obtained using a Horia Jovin Yvon LabRam instrument, fitted with a green laser at 532 nm. Spectra processing used the suppliers associated LabSPEC software (version 5.64.15, 2008). Silver colloid was prepared as described in White and Hjortkjaer (2014). Comparisons were made with both the potential active ingredients of the yellow and pink pigments, and with modern pigments.

Comparison spectra for the yellow pigment were obtained from modern yellow pigments; Stil de Grain, Fustic, and Reseda (Kremer Pigments), and their respective chromophore components, Emodin and Quercetrin, Morin, Luteolin, and Apigenin (Sigma). The pink pigment was compared to Lac Dye, a red Indian Lake, Carmine Lake pigment, Rose Madder, and Red Madder (Kremer Pigments), and their components are carminic acid and alizarin (Sigma).

For comparison samples with a known composition, a 10⁻⁴ molar solution was prepared in distilled water. A range of concentrations of pigment and aggregating agent were tested in order to optimise the spectra. For most samples, 2 µl of the 10⁻⁴ M sample was combined with 60 µl 0.1 M nitric acid (Fisher Scientific) and left for 6 min before adding 500 μl silver colloid. After pipetting 2 µl onto a glass slide, it was left to dry prior to Raman analysis. With Carmine Lake, 10 µl 0.1 M nitric acid was sufficient, and 20 µl 0.1 M nitric acid was used with Emodin.

Opie’s pink pigment was prepared with 0.002 g pigment in 100 µl distilled water. Optimisation showed that 2 µl pigment solution combined with 10 µl nitric acid prior to adding silver colloid gave the clearest Raman signal. As the yellow pigment was only partially water soluble, it was first solubilised by adding 0.002 g pigment to 30 µl MeOH (Fisher Scientific HPLC grade), plus 10 µl distilled water. After vortexting for 2 min, the solution became an even yellow colour. Ten microlitre of this solution was diluted into 40 µl distilled water. Six µl of the pigment solution was combined with 20 µl nitric acid before proceeding as above.

Figures 2–8 contain graphs prepared using Spectragryph version 1.2.16.1 (Menges 2022).

Figure 3. (a) Yellow lake pigment viewed under transmitted light, showing low relief yellow flakes, image width 160 µm; (b) comparison of Opie’s yellow lake pigment (blue) with Raman SERS spectra for morin (red) and fustic extract (orange) with baseline correction applied.

Figure 4. (a) Burnt Sienna pigment viewed under transmitted light microscopy, showing a heterogeneous sample. Image size 200 µm; (b) Labelled XRF peaks from Bruker’s S1PXRF analysis programme, showing the peaks indicating the presence of iron (left) and arsenic (right). Readings are in counts.

Figure 5. (a) Transmitted light microscopy of pink pigment, image width 160 µm; (b) Raman SERS spectra of Opie pink pigment (orange), with a comparison sample of carmine lake (blue), and the silver colloid spectra (red).

Figure 6. (Left) Indian red pigment under transmitted light (top) and plane polarised light (below), image width 160 µm; (Right) SEM image of Indian Red (top) and EDX results showing iron and oxygen (below).

Figure 7. (a) Blue pigment viewed in transmitted light, image width 160 µm; (b) FTIR-ATR spectra of the unlabelled blue pigment (blue) and light ultramarine pigment from Cornelissen (orange).

Figure 8. (a) Opie black pigment viewed in transmitted light, x200 magnification; (b) FTIR-ATR spectra of Opie’s black pigment.

Results

Seven of the pigment samples were analysed in this study. All of these pigment bottles had previously been opened, and some pigment was missing. It was decided to only sample those pigments which could be accessed without damaging the pigment containers. This meant that a further seven bottles were not sampled, as it could have caused extensive damage to the cork stopper had it been removed.

As four of the sampled bottles had labels, this provided some indication as to what could be expected of the contents and enabled a more targeted analysis of the pigments. However, it was not known if the original pigments were indeed still present, or whether they had been refilled at a later date. A further three bottles analysed were unlabelled.

Flake white

The bottle labelled F. White contained a white pigment, and the name implied that this would be a flake white. Flake white is a form of lead white. This pigment was first manufactured by the ancient Greeks and remained the most commonly used white pigment by artists until the introduction of zinc and then titanium white. It is still regarded as the whitest of the white pigments, but is, of course, toxic, hence its decrease in usage from the late nineteenth century, when other whites became available (Barnett et al. 2006).

Lead white would have been manufactured by the stack method in the nineteenth century, using lead plates suspended over vinegar to create the pigment through the corrosion process (Harley 1982, 167). The addition of horse manure over the jars of lead and vinegar caused carbon dioxide and heat to be released, as the manure decomposed. This decomposition process led to the formation of hydrocerussite (Pb₃(CO₃)₂(OH)₂) and cerussite (PbCO₃). These products formed by the stack process were generally produced in an 80:20 ratio respectively, through post-production treatments can alter that ratio. Gonzalez et al. (2016) found significant variations between the ratio in their study of pigments used in a range of paintings from the Renaissance era to the late nineteenth century. Flake white was thought to be the purest form of lead white pigment, as it was less likely to be adulterated with other pigments or extenders (Harley 1982, 170).

When this pigment was examined using transmitted light microscopy, it appeared to confirm the expectation that this was indeed a lead white. At x400 magnification, the particles were quite fine, with evidence of hexagonal fragments and was highly birefringent under polarised light (Figure 2(a)). There is certainly no evidence of additional crystal types visible, which shows that this was indeed unadulterated and purely a lead white.

This was reinforced through the use of SEM-EDX and XRF, which both confirmed the presence of lead (data not shown), with EDX additionally detecting oxygen, and no additional impurities. XRF is unable to detect elements lighter than magnesium, so the absence of oxygen and other light elements from the XRF reading was to be expected.

Further evidence to support this being a lead white was obtained from FTIR-ATR, where it produced a 97% match with a modern flake white pigment (Cornelissen); a significant match (Figure 2(b)). The bands shown on FTIR can be attributed to lead white features based on prior FTIR data characterising these minerals (Siidra et al. 2018). The peak at 3633 - 3535 cm⁻¹ links to O-H stretching. A number of bands can be attributed to C–O vibrations, due to CO₃^2- anions. The broad peak around 1360-1400 cm⁻¹ relates to asymmetric stretching of CO₃^2-, obscuring the 1405 cm⁻¹ region peak which would be expected to correspond to symmetric stretching. Bending vibrations for those anions cause peaks at 852 and 837 cm⁻¹ (out of plane) and 677 cm⁻¹ (in plane). The small peak at 874 cm⁻¹ in Opie’s pigment was not found in the flake white comparison sample and relates to the presence of calcite within the pigment. Siidra et al. (2018) tentatively assign a broad band around 1090-1108 cm⁻¹, which is labelled as 1104-1108 cm⁻¹ in Figure 2(b), to the presence of SO₄^2- anions. The broad band shown at 767-768 cm⁻¹ is specific to hydrocerussite and is due to the presence of Pb … O-H bonding. The peak at 837 cm⁻¹ is specific to the presence of cerussite, demonstrating that the mix of minerals is present, as expected from a lead white. The difference in peak size for cerussite and hydrocerussite between the Opie pigment and the reference sample indicates a potential difference in the ratio of the two minerals, however further analysis through X-ray diffraction would be required to produce a precise ratio between the cerussite and hydrocerussite products.

Yw. lake

Lake pigments are defined as soluble organic colours, which are co-precipitated with metal salts and alkali. These form insoluble compounds consisting of a metal oxide and a colourant (Bersch 1901, 1). While the label of yellow lake gives some idea of what the pigment bottle may contain, there are a number of different dyes and bases which can be found in yellow lake pigments. Metal substrates are commonly lead, tin, or alumina, while other substrates may include limestone or starch, though Bersch (1901, 90) states alum was the most frequently used, due to its greater resistance to discolouration. Dyes can come from a variety of organic sources, with weld and buckthorn berries being some of the most common in the preparation of yellows (Bersch 1901, 89–92; Eastaugh 1999, 361).

Examination of the pigment using FTIR-ATR produced a 95% match with rose madder when compared to the in-house pigment library (Sancho, Sinha, and Skipper 2014). Rose madder is generally a pink colour, with the organic component obtained from the root of the plant (Harley 1982, 140). However, the organic component in a rose madder lake pigment is a relatively low concentration, and it may not be detectable using FTIR. Therefore, this is more likely to be picking up the inorganic base material only, implying that would have been the same base material used for both this yellow lake and the rose madder lake pigment used for the in-house pigment library.

Comparing the FTIR spectra to that obtained by Kirby, Spring, and Higgitt (2005) for hydrated alumina lake, the peak pattern strongly resembled that of alumina prepared with potassium aluminium sulphate, with AL–O bonds below 950 cm⁻¹, broad water bands around 3400 and 1650 cm⁻¹, and sulphate bands around 1080 and 980 cm⁻¹. This indicates the presence of a significant amount of phosphate, which causes the S-O peak (normally found at 1125 cm⁻¹) to shift to the 1080 cm⁻¹ position.

Microscopy showed that there was a low relief, relatively large yellow coloured particle, which did not show birefringence under polarising light. This is a very good fit for the characteristics of an alumina base for the lake (Figure 3(a)). There were also tiny, birefringent particles on the surface of the flake, which could not be conclusively identified just using microscopy.

SEM-EDX indicated the presence of aluminium, sulphur, oxygen, and potassium, which would further fit with the expected components of an alum lake, and reinforces the indication from the FTIR spectra (data not shown). We are therefore able to conclude that this lake pigment was indeed formed from an alumina precipitate. Additional SEM-EDX data from other points within the pigment showed the presence of lead within the sample. This was further reinforced by the XRF data, which also detected lead (data not shown).

Evidence of the use of lead to make a variety of yellow lakes can be found in literature both pre- and post-dating the Opie pigment sample. The Paduan Manuscript, which may have been composed during the middle or later part of the seventeenth century, contains a recipe for a yellow pigment composed of a mixture of white lead and gialdo santo (Merrifield 1999, 650–651), with gialdo santo being described as a yellow lake made from a variety of flowers. Osborn (1845) describes Stil de Grain (buckthorn berry lake) recipes with lead oxide, to make the pigment less fugitive. Eastaugh (2004, 162) states that the London colourman’s firm Berger prepared yellow lakes in this manner in the early nineteenth century, a date that fits well with the date of the Opie pigments. The pigment is combined with white lead in the manufacturing process. In Bersch’s 1901 book The Manufacture of Mineral and Lake Pigments, it states that ‘since the least trace of iron damages the shade of the colour, Fustic Lake is more frequently made by means of lead oxide’. This produces what Bersch describes as a handsome yellow, and well suited for paintings (351). It is likely that the method described by Bersch in 1901 would have been in use around 100 years previously. It would therefore seem likely that the small, birefringent particles detected are a lead oxide, added to an alum base with yellow fustic providing the colour. This would account for the presence of lead detected through XRF and SEM-EDX.

In order to investigate the organic component of the pigment further, Raman spectroscopy was used. The high fluorescence of the pigment masked the Raman scattering, which meant that a surface enhanced Raman spectroscopy (SERS) technique was employed, using a silver colloid technique to reduce the fluorescence.

Historically, the most common pigments used in yellow lakes were buckthorn berry (Stil de Grain), Reseda (weld), and Fustic (dyer’s mulberry, Maclura tinctoria) (Mayhew et al. 2013). Spectra of these pigments, and their active ingredients, were used as a comparison to the spectra of the unidentified yellow lake pigment. The chromophores are carbon-based metabolites found in plants and fungi, and in this instance is the aspect of the plant extract that provides the colourant to the dyestuff. SERS spectra were obtained for Stil de Grain, Fustic, and Reseda pigments (Kremer Pigments), and their respective chromophore components. These are Luteolin and Apigenin for Reseda lake; Emodin and Quercetrin for Stil de Grain; and Morin for Fustic lake (Sigma). The SERS spectra for Opie’s yellow lake pigment were still a relatively weak signal with low peak intensity, despite optimisation through adjusting the concentrations and Raman settings. However, when comparing the spectra Opie’s pigment appeared most closely matched to that of the extract of Fustic and to Morin, the flavonoid found in Fustic, with SERS peaks around the 1200–1600 cm⁻¹ region in particular being seen only on the Opie pigment, Morin, and Fustic (Figure 3(b)). We can therefore conclude that this is most likely an aluminium-based Dyer’s Mulberry lake pigment, which was also commonly known as Old Fustic. Given the mix of other components identified within the pigment, it is unsurprising that it was not possible to produce an exact spectral match.

Bt. Sienna

The bottle labelled Bt. Sienna contains a dark brown pigment. Microscopically, the pigment contains a mix of grain sizes and coloured particles varying though transparent to yellow, orange, and brown (Figure 4(a)), which is consistent with the description of burnt sienna in Eastaugh (1999, 367).

A comparison of FTIR-ATR spectra of the Bt. Sienna pigment with those of modern burnt sienna did not show a significant (over 95%) match (Sancho, Sinha, and Skipper 2014) although there were a number of common features present in the sienna’s examined. This however is not surprising due to the impurities found within the pigment. Burnt sienna pigments are obtained by heating brown earth pigments, composed of iron oxide with various additional impurities. The impurities vary depending on the region in which the earth was obtained. Although the name Burnt Sienna originally referred specifically to earth from the Siena region of Italy, over time this changed to refer to any burnt earth pigment of a dark brown colour (Eastaugh 2004, 339).

Studies of the pigment using XRF and SEM-EDX confirmed the presence of iron within the pigment, along with oxygen and silica. It is likely that the silica content comes from the quartz impurities present in the sample, while the iron content is due to the presence of iron oxide as would be expected from a sienna earth pigment. This therefore confirms the bottle label is still representative of the contents, and the pigment is indeed a burnt sienna. Interestingly, a further finding from both XRF and SEM-EDX was the presence of arsenic in the sample (Figure 4(b)). Arsenic has been detected in a few technical examinations of Burnt Sienna samples from Italy, such as in the studies by Nel et al. (2010), and Manasse and Mellini (2006), and Montagner et al. (2013). This suggests that the Opie pigment comes from a similar geographical region, and so may in fact be an Italian imported pigment, rather than a more generic example of a burnt earth pigment. This may prove an interesting marker for future studies of Opie’s paintings, to examine his use of this specific pigment.

Unlabelled pink

This unlabelled bottle contained a very vivid coloured pink pigment. As was found with the yellow lake, the FTIR-ATR spectrum for the Opie sample of unlabelled pink shows an 88% match with rose madder when compared to the in-house libraries. This is not a significant result but suggests a relatively similar composition for the pink pigment, meaning that it is likely to also be a lake pigment. It does not however mean that the pigment is rose madder, for, as previously discussed for the yellow lake, the FTIR-ATR similarities are only to the inorganic component (i.e. an alumina base) rather than the organic lake dye. The pink pigment was almost identical to the yellow pigment spectra, indicating a similar composition and manufacturing method.

Under x400 magnification, there were clearly a number of different components present within the pigment sample. The bulk of the sample was composed of pale pink, rounded, isotropic pigments on a low relief, colourless base. This is the pink lake pigment, and the base has the appearance of alumina. A far darker, anisotropic pigment with bright red/orange birefringence was also observed, often clustered onto the alum base (Figure 5(a)).

The likelihood of this being a lake pigment is reinforced by the SEM-EDX and XRF data, which confirmed the presence of elements indicating an alumina compound. This, therefore, would be the base onto which a dye was precipitated. Further elemental analysis using these techniques showed the presence of mercury within parts of the pigment sample, and further grains composed of calcium, carbon, and oxygen (data not shown). The presence of these elements may be accounted for by examination of historic recipes and referring to the microscopy data. Polarising light microscopy suggests that the darker, anisotropic pigment could be vermillion, which would account for the presence of mercury found from SEM-EDX and XRF analysis. Vermilion is a widely used historical term for synthetic forms of the mineral cinnabar (mercury (II) sulphide) (Eastaugh 2004, 386).

Red lake pigments tend to be formed from either madder (plant extract) or carmine (insect-based) dyes. The presence of carmine was confirmed by SERS spectra of the unknown pigment. The spectra of Opie’s pink pigment were compared to SERS spectra for a modern carmine lake pigment, red madder, rose madder, and lac, all of which have been used historically to produce a red or pink colour in lake pigments. Overlaying the spectra showed a match to carmine (Figure 5(b)), with peaks at 1301–1302 cm⁻¹ for both the Opie pink and the carmine lake pigment clearly visible in comparison to the silver colloid alone. No matches were found to lac or to madder.

The microscopic appearance of the pink particles is a close fit to that shown in Eastaugh (1999, 358) of a carmine lake adulterated with vermillion, and this is reinforced by the analytical data. Field states that a carmine may be ‘sophisticated’ with vermillion, especially the scarlet lakes (1835, 79–80). Genuine carmine faded rapidly and was relatively expensive; the addition of vermillion was claimed to make the colour more durable (Harley 1982, 138). Vermillion may have had an added benefit to the manufacturer, as Bersch (1901, 94) adds that carmines may be adulterated with vermillion, implying that this also was used to reduce the quantity of carmine needed to make the pigment.

A transparent pigment was also visible under high magnification. This was birefringent under polarising light and showed undulose extinction. This may indicate the presence of chalk white, which is consistent with the SEM-EDX data. Characteristic standing cross coccoliths were visible under polarising light, indicating the presence of a natural chalk. The chalk may have been added as an extender to the pigment, or used to modify the colour of the pigment to produce pink instead of red. This again seems usual for this type of pigment, as other studies of nineteenth-century artists’ pigments, such as Kirby, Spring, and Higgitt (2007) and Pozzi et al. (2014) found that many contained chalk grains, which they state were added as an extender.

Indian red

As the pigment bottle was labelled with the term Indian Red, this provided an indication of the potential composition of this pigment. Indian Red is known to be a red earth pigment composed of iron oxide from the Persian Gulf, though it gradually decreased in availability. By the late eighteenth century, the pigment named Indian Red was generally available as a manufactured iron oxide, rather than a natural earth (Harley 1982, 119). With the expected date of manufacture of the Opie pigment in the early nineteenth century, it appears likely that this pigment would be of a synthetic origin.

A sample of modern Indian Red produced an FTIR spectrum with a 45% match to the Opie sample. Although these spectra are similar, the match is clearly not significant, and visually the spectra are not identical (Sancho, Sinha, and Skipper 2014). As with the Opie Bt. Sienna sample, this is likely to be due to the variations in chemical composition between the historic and the modern pigments. The FTIR spectra showed iron peaks at 420 and 515 cm−¹, and Si-O-Si stretch at 1026 cm⁻¹, but in contrast to the haematite sample examined by Čiuladienė et al. (2018), did not show CaCO₃ vibration peaks. This provides additional support to the microscopic identification of this being a relatively pure pigment sample.

Microscopically, it is apparent that the Indian Red has regular grain sizes with no obvious impurities and showed a strong red birefringence under polarising light. This indicates a synthetic origin, as natural red earth pigments would have some impurities present and tend to be more variable in size (Eastaugh 1999, 363). Elemental analysis through SEM-EDX showed the elements oxygen and iron are present, as expected of an iron oxide pigment (Figure 6), with some additional lighter elements that are not identifiable through this technique.

Taken together, evidence suggests that this red is a synthetic form of hematite, Fe₂O₃. This iron oxide pigment may have been composed of caput mortuum, an iron oxide by-product of the sulphuric acid industry (Harley 1982, 122). This would be consistent with the dates for pigment availability of Indian Red, as by this point in time we would expect to find an artificial iron oxide pigment sold under this name.

Unlabelled blue

Microscopically, the unlabelled blue pigment consisted of a mix of pigment sizes of bright blue colour, with fracture lines, and some colourless birefringent impurities visible under polarised light (Figure 7(a)). SEM-EDX analysis detected the presence of sodium (Na), aluminium (Al), silicon (Si), oxygen (O), and sulphur (S) (data not shown). All these elements are what would be expected within an ultramarine blue.

Ultramarine blue was originally obtained from the lapis lazuli mineral, initially imported from Afghanistan, but a new supplier in Siberia increased the availability of this pigment in the nineteenth century, and in 1835 Field describes ultramarine as being brought from China, Tibet, and Lake Baikal (Siberia) (Field 1835, 122; Harley 1982, 45). Although there is some disagreement with the published dates, it is generally accepted that the artificial form of ultramarine was not commercially available until the 1820s (Eastaugh 2004; Plesters 1993, 55). Therefore, this pigment, if natural, would help to indicate whether the pigments either postdated the labelling on the box or had been used and refilled at a later date.

In both natural and synthetic forms, ultramarine is essentially a sodium aluminium silicate combined with a small amount of sulphur. Ultramarine does not have a fixed formula and the ratios of the various constituents can change within limits. Generally, the chemical formula for ultramarine is given as 3Na₂O.3Al₂O₃.6SiO₂.2Na₂S (Harley 1982, 43). It is the additional sodium and sulphur atoms that determine the colour of the ultramarine (Gettens and Stout 1966, 164). Because of the very similar chemical structures between natural and artificial ultramarine, it can be difficult to ascertain the difference between the pigment types. Under magnification, synthetic ultramarine generally has fewer impurities and is more regular sized. This pigment examined under microscopy showed a range of different grain sizes, which would indicate a natural pigment, although it was low in impurities. A small proportion of clear crystals were visible, which resembled calcite on visual inspection. The visual appearance of the Opie sample under microscopy did not conclusively indicate its source, since although the pigment appeared relatively low in impurities (potentially suggesting a synthetic pigment), this may also indicate a high-grade natural source, as suggested by Eastaugh (1999, 45).

FTIR-ATR of the pigment demonstrated a main peak in the region of 960 cm⁻¹. This is a slightly lower wavenumber than the Si,Al–O band observed in ultramarine by Milani at 1010 cm⁻¹, but depending on the source there may be some variation due to the overall chemical composition and source of the pigment. For example, Bruni et al. (1999) determined a range of different peaks in that region by looking at standard samples of lazurite, sodalite, and haüyne, all of which can be found within the lapis lazuli mineral. These peaks did not prove a good match for the FTIR spectrum generated by the Opie sample, but Bruni used a NaCl pellet preparation technique, which may have led to variations in the readings. The spectrum did produce a clearer visual match to lapis lazuli in the ATR-FTIR database produced by Vahur and colleagues (2016), which showed a large peak at 966 cm⁻¹, and smaller peaks in the 698-637 cm⁻¹ region.

Experiments carried out by Miliani et al. (2012) comparing synthetic and natural ultramarine conclude that a small difference is visible in FTIR-ATR analysis. They demonstrate a difference due to the presence of reagent residues such as kaolinite in the synthetic pigment (O–H stretchings at 3624 and 3700 cm⁻¹), and the presence of calcite in the natural pigment (2512, 2340, and 1800cm⁻¹). As shown in Figure 7(b), FTIR-ATR of the unlabelled blue pigment from Opie did not show significant peaks at any of these points. However modern synthetic light ultramarine sample (Cornelissen) run as a comparison did demonstrate the typical kaolinite O-H stretching peaks. As only a small proportion of the sample contained calcite or other impurities on visual inspection using microscopy, it is possible that the calcite peak was too small to visualise on FTIR-ATR, because of its low concentration within the sample.

On balance, based on the appearance of the pigment and FTIR-ATR data, it is likely that the unlabelled blue pigment from Opie is a natural ultramarine.

Unlabelled black

The black pigment showed a range of sizes and shapes of pigment when viewed under magnification, suggesting a natural origin due to its heterogeneity (Figure 8(a)). Blacks in use at the time of the paintbox were mostly based on burnt organic material, such as lamp black, from soot, and ivory or bone black, from charred ivory or bone respectively (Harley 1982, 156–7). Analysis of the Opie sample using SEM-EDX showed that the elements contained within this pigment were calcium, phosphorus, magnesium, and oxygen, all of which were consistent with the chemical formula for bone and ivory black. This fits with the data from the FTIR-ATR spectrum, which gave a 95% match with ivory black from the in-house library database (Ca₃(PO₄)₂) (Sancho, Sinha, and Skipper 2014). Tomasini, Siracusano, and Maier (2012) characterised a range of black pigments using FTIR, including lamp black, ivory black, and bitumen. The FTIR bands seen on the spectra for Opie’s pigment (Figure 8(b)) closely match those found in their ivory black sample. Bands at 556, 600, and 1015 cm⁻¹ relate to the presence of phosphates (apatite). Carbonated hydroxyapatite is responsible for bands in the regions of 874, 1413, and 1452 cm⁻¹. The broad band around 1600 cm⁻¹ region relates to C = O amide stretching. Interestingly, a peak at 2013cm⁻¹ is not present in the Opie sample. Daveri, Malagodi, and Vagnini (2018) state that this band is due to the formation of cyanamideapatite during pigment manufacture, where some combustion of the source material has taken place in the absence of air and indicate this could be a suitable marker for the presence of an animal-based pigment. However, the lack of marker at this point with the Opie pigment, when all other peaks indicate the presence of an animal black, suggests that this marker may not be effective as an identification method alone.

Although evidence indicates an animal black, this does not differentiate between bone and ivory as a source, since chemically the components are very similar. Ivory black or bone blacks generally contain about 10% carbon, 84% calcium phosphate, and 6% calcium carbonate (Gettens and Stout 1966, 99). Magnesium is found in both bone and ivory blacks, though Freund et al. (2002) determined that the atomic mass ratio of magnesium to calcium is lower in bone (1:18) compared to ivory, where the ratio is 1:8. Our SEM-EDX readings indicate the P:Ca ratio in this pigment is 1:11. Although this is not quite the expected ratio for ivory, it is higher than expected for bone, and suggests an ivory source is more likely.

Ivory black is produced by charring ivory in closed vessels and then grinding, washing, and drying the black residue. Gettens and Stout state that the term is now commonly used for the black from animal bones known as bone black, and in fact, the names have historically been used interchangeably, with the term ivory black adapting over time to mean a higher quality black, and not necessarily implying an ivory source (Gettens and Stout 1966, 122). Although we cannot conclusively assign a source for the animal component, the pigment in question is a black pigment formed primarily from bone or ivory and is more probably an ivory source. This pigment was the darkest and considered to be the best type of black available (Harley 1982, 158).

Discussion

Through a combination of scientific analysis and referring to historical literature and recipes, it has been possible to determine the composition of all seven of the pigments sampled from Opie’s paintbox. It contains a lead white, known as flake white, with no extenders, indicating a high-quality source. The yellow lake is composed of an alum base, with yellow from Dyer’s Mulberry (Old Fustic), most likely prepared using a lead oxide production method. Kirby, Spring, and Higgitt (2005) state that the presence of sulphur within an alumina lake pigment indicates a nineteenth-century production method for the lake, where the alum and dye were mixed first, prior to adding the alkali. The other lake pigment also has an alum base and is formed from carmine and vermillion. Although Burnt Sienna is indeed a natural, earth pigment, Indian Red is a synthetic iron oxide. The blue is a high-grade natural ultramarine. The black is a calcium and phosphate-based pigment and is likely to be an ivory black, again a high-quality source material. All of the seven pigments identified would have been available at the purported date of the paintbox, 1806. There is no evidence that they have been re-filled at a later date, as the labels on the bottle matched the pigment found within. It does therefore seem likely that they are contemporary with the label on the box, so could indeed have belonged to John Opie, and been used in his paintings in the year before his death.

Conclusion

Opie was a strong advocate for the importance of ‘truth’, which was a recurring theme throughout his lectures to the Royal Academy in 1806. So, what is the truth of these pigments? The pigments within the box were of good quality. The wooden box containing the pigments is well-made, using mahogany, an expensive wood, and lined with red felt. The ultramarine blue is of a high-grade natural lapis lazuli, with a low number of impurities. Flake white was known to be the best quality of the lead whites (Harley 1982, 170). Bone or ivory black was thought to be superior to the other available blacks, with ivory being the higher quality (Harley 1982, 158). The pink lake is carmine-based, with the addition of vermillion, both relatively expensive pigments at the time. Overall, this was an expensive gift, so the giver certainly held Opie in high regard. Could they indeed be a gift from the Royal Academy, related to his professorship? Or could they be a present from a friend to mark the occasion? There is currently no historical information available to confirm or deny either of these possibilities. It is interesting that there are no indications of a maker, as one would perhaps expect from a colourman, a paint supplier.

Cawse produced a manual entitled Introduction to the art of painting in oil colours in 1822, and stated that he was a ‘pupil to the late John Opie R.A.’ It is therefore likely that ideas in the book would have been influenced by Opie’s teachings. He provides a list of colours needed by artists which are summarised as follows: White (lead-based), Black (made from vine and ivory), Blue (ultramarine and Prussian blue), Yellow (light and brown ochre), Red (light red, a calcined light ochre), India red, Carmine Lake, Umber, and Vermillion (Cawse 1822, 5–7). He suggests that many then add ‘sienas’ [sic] and Vandyke Brown. As a vehicle (a medium used to carry the pigment) he recommends a mix of oil and turpentine (Cawse 1822, 8). Although the varnish bottle in the paintbox is labelled simply as varnish, Cawse (1822, 9) states that copal or mastic is available, but mastic is preferred. Similarly, the 1795 updated reprint of Bardwell’s manual of oil painting lists flake white ‘the very best white we have’ (Bardwell 1795, 77), Ivory black as the ‘best black’, ultramarine as the ‘best blue’, Prussian Blue, Light Ochre (a yellow), light Red, Vermillion, Carmine, Lake (described as a deep red), Indian Red, Brown Pink, and Burnt Umber as being desirable colours for an oil painter (Bardwell 1795, 77–81). Brown-pink may have been a buckthorn lake pigment (Clarke 2013).

The colours in the paintbox therefore seem likely to represent a slightly extended version of the ideal colour palette needed for painting almost any portrait or landscape in oils, along with the necessary equipment to prepare the pigments for use.

Characterisation of these pigments provides valuable information to others studying pigments from this period, demonstrating the compositions used in relation to bottle labels, and also giving an enhanced understanding of the materials used. In addition, the identification of these pigments may prove helpful in future studies of Opie’s paintings. Although it is clear that the pigments have been used, it would be necessary to examine the chemical composition of pigments on Opie’s paintings in order to investigate whether they were indeed used by Opie himself, before his untimely death. The arsenic component of the burnt sienna could be a particularly useful indicator of a genuine Opie painting, if it can be shown that he did indeed employ these pigments from the paintbox in his works. Measurement of the hydrocerussite:cerussite ratio of the white in the paint box could be compared to that found in his paintings. These pigments from 1806, though, will only be found in paintings produced in the last year of his life, and so further studies of Opie’s paintings would be necessary to build up a fuller picture of the colour palette that he used throughout his career. We also cannot exclude the possibility that the pigments were used by a descendent; the artist Edward Opie (1810-1894) was the great-nephew of John Opie, and there is a possibility that he may have used the paint palette also in the collection of the St Agnes Museum (BBC News 2012). Further research into paintings by both authors, and the palette itself, may help to clarify this.

Scientific analysis combined with the use of historic recipes has therefore been able to reveal at least some of the truth surrounding this paintbox. We can show that it may indeed date from 1806 and, in the words of Opie himself, ‘I am confident a knowledge of the truth will in the end equally benefit the art and the artist’ (Opie 1809, 21). In the modern day, we hope this knowledge will also benefit conservators, art historians, and heritage scientists as well.

Acknowledgments

Thanks to the Bruker Corporation for the use of a handheld Tracer IV system XRF. We are grateful to St Agnes Museum for allowing us to carry out this study, and to the University of Lincoln Life Sciences department, who provided the Raman, SEM-EDX, and FTIR used in the study. Sampling of the pigments was carried out by Rhiannon Sinha. Further thanks to Tom Hendra, an undergraduate chemistry student from the University of Birmingham, who assisted with the polarising light microscopy and some of the historical research, and to Dr Philip Skipper for his comments on the manuscript. This work was part-funded by a research grant for returners to work after maternity leave, from the University of Lincoln.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by University of Lincoln, UK.