Ways of Seeing: Fourier Transform Raman Spectroscopy in the examination of artist’s materials

15. Ways of Seeing: Fourier Transform Raman Spectroscopy in the examination of artist’s materials.

Alex Drake and Kristine Moore*

‡Department of Pharmacy, King’s College London, Manresa Road, London SW3 6LX

*To whom correspondence should be addressed: e-mail:kristine.moore@kcl.ac.uk

Abstract

A brief survey of the challenges presented by, and some of the techniques that have been used in, the examination of artist’s materials is presented. Pigmented egg yolk tempera and oil films were examined by both visible & near infrared Raman spectroscopy. The potential of Fourier Transform Raman Spectroscopy (FTR) as a non-destructive technique for the examination of artist’s materials is discussed.

 Keywords:

FTR, EGG YOLK TEMPERA, OIL PAINT, PIGMENTS,

Introduction

Paints made using the egg yolk of the domestic fowl (Gallus gallus) have been documented since at least AD 1140 [1] and are well known for their durability [2] and relative resistance to treatments. The so called “drying oil” paints gradually superseded tempera in Europe during the fifteenth century. However, as it became known, pictures in oil were less resistant to change than those of egg tempera but the oils had offered advantages in composition. Despite the hardy nature of tempera pictures, some undesirable changes may occur.

A requirement exists to distinguish between works executed in egg yolk tempera, oil, egg and oil, and more complex intermediate mixtures for some, or all, of the following reasons: to assist in the completion of the documentation of a picture, to allow informed selection of the appropriate preventative conservation measures that will allow the art work to be seen by future generations as the artist intended, and to contribute to the discussion on the question of attribution. An understanding of the mechanisms underlying the progressive ageing of films will assist in the identification of egg yolk and egg yolk mixtures in paint media.

Identification of media by Fourier Transform Infrared Spectroscopy (FTIR) alone can be unreliable as absorbance from the pigment can obscure bands due the media. In addition, infrared spectra of aged films are inevitably much more complex than those of the initial constituents. Conservators are understandably cautious in releasing anything but the smallest, inconspicuous, samples from works of art and so a further constraint is placed on the technique used for identification. A combination of methods, for example, FTIR/Differential Scanning Calorimetry, may be used to give a greater degree of certainty in identification. The sequential application of techniques to a sample usually follows a non-destructive to “more-destructive” technique rationale. Identification of proteins or peptides by staining, or by a separative technique, is undesirable as the former method is unreliable and both methods are inevitably destructive. Many techniques have been applied to pictures to find a combination of non-destructive techniques or, preferably, a non invasive technique that allows conclusive identification of both the media and pigment present.

Fluorescence has limited the application of Raman spectroscopy, using conventional lasers, in this field but the technique has proved useful in the identification of extremely small pigment particles on, or very near, the surface of manuscripts and ceramics [3, 4] A recent survey [5] showed 56 out of a total of 64 pigments in use before ~AD 1850 to produced high quality spectra by Raman microscopy. Ostensibly, the application of FTR to the study of both media and pigments is very inviting. The technique using near infrared excitation is, on the whole, free of the fluorescence problems associated with the blue-green lasers. It is, usually, non-destructive and a real possibility exists for in situ, non destructive study of the materials with which pictures are made. The use of FTR in tandem with FTIR will yield the “whole vibrational picture” [6].

FTR using a remote probe sampling system has recently been used to determine the source of an ivory inlay [7]. The introduction of a non-invasive, non destructive, technique, such as FTR, yielding molecular information in situ on the chemical species, both organic and inorganic, within a picture could be a major advance in this field. To speculate wildly – if FTR, or perhaps Raman Optical Activity Spectroscopy, could provide the ability to ascertain the absolute age of a representative sample of media within a picture there would be major implications for the art world

Professor Hendra at Southampton University kindly helped me to make exploratory investigations into the application of FTR to the identification of media.

A consideration of some of the limitations in the application of FTIR to the study of ageing of paint films.

In practice, when an infrared spectroscopist is asked by a conservationist to address a particular query, the question is very rarely “what is it?” but rather “is it oil or is it egg? glue or gum? is this resin the same as the other resin?” Contextual information concerning location of the picture, likely provenance, condition, pigments present (by elemental analysis, scanning electron microscopy) and information from optical and polarised light microscopy provide data to assist in the interpretation of the spectrum.

Fundamental research starts with known materials and simple situations and common practice is to study the processes of ageing that may occur in simple media mixtures with single effects, for example, light or thermal ageing, and predict and match the behaviour of authentic films when possible. FTIR is a mature technique, readily available in these situations, at relatively low cost, for the examination of even extremely small quantities of paint samples.

The mechanisms underlying the ageing of oil are still the subject of debate [8] and Raman spectroscopy has contributed to a revival of this discussion. The proposed mechanisms underlying the drying of oil have been used as a point of departure for the discussion of those underlying the drying and ageing of tempera as the triacylglycerol (TAG) fractions are common. Tempera films differ from those of oil in possessing three different, significant, components: protein, water and phospholipid. We are aware that tempera and oil films have different physical properties, there must be a physico-chemical basis for this.

Egg tempera and oil, when unpigmented, are easily distinguished by FTIR – the simultaneous presence of four characteristic bands Amide A, B, I & II and the common ester carbonyl at ~1740 cm-1 identifying the protein/oil mixture. However the addition of pigment often masks a significant and diagnostic portion of the spectrum and leads, in turn, to difficulty in interpreting age and pigment dependent changes. Unpigmented media have been studied to provide a baseline in uncovering age dependent changes in media mixtures but they are rarely, if ever, used in pictures.

Workers in this field have used sodium chloride or potassium bromide (KBr) plates as supports for fresh films and KBr disks as method for investigating historic, and therefore scarce and valuable, samples [9]. Hygroscopic materials of this type may be suitable for oil media where, we presume, there is little or no water involvement with the stabilisation of the film. However, we have shown that it is possible to remove water from some tempera paints and there is probably a reversible exchange of water with the ambient air. We have also demonstrated that it appears possible to reproduce some of the effects of ageing by the desiccation of tempera samples and this may confirm the role of water in the inhibition of lipid oxidation. The application of a technique that does not have the effect of abstracting water from films is desirable.

The use of FTIR micro-sampling (beam condenser with diamond cell) provides a method whereby water is allowed to remain in contact with the sample. However, there are concerns that intensity information obtained when using this technique can be misleading as the light throughput is compromised both by beam condensing and the crystal lattice absorption of the diamond cell Sections of appropriate thickness can be prepared for FTIR transmission microscopy. However, specimens are often “gritty” proving difficult to microtome and embedding resins can infiltrate the sample to a varying extent. Although spectral subtraction of the embedding media can be applied, media peaks can be swamped by those of the resin. Workers at the National Gallery in London have successfully used silver chloride (AgCl) as an embedding material for thin sections from pictures [10]. The AgCl is transparent in the mid IR and so spectra of the media/pigment only may be obtained but, difficulties exist in supporting sections of this type and they lack long term stability as the AgCl eventually discolours. The preparation of cryosections is theoretically interesting but the same difficulties of supporting the section exist.

The difficulty of sampling a homogenous section of a sample within even an extremely small section of pigmented media must not be ignored. We are inevitably sampling a molecular array containing areas of media:media, media:pigment and media air:interaction We see part of, but, a mixed picture. Raman spectroscopy will encounter the inadequate spatial resolution problem and so the sampling of an inhomogenous molecular array, is common to both techniques. FTR offers the advantages of new and complementary information without a requirement for excessive manipulation, adulteration or destruction of sample.

FTR Spectroscopy at Southampton

Experimental

Semi quantitative media rich films were prepared by the traditional tempera method. Organic eggs (Best Before 24/12/97, used on 11/12/97) were obtained from Waitrose Supermarket, King’s Road, London. Pigments were obtained from L. Cornelissen & Son Ltd, Great Russel St, London. London tap water supplied by Thames Water, calcium concentration ~ 4mmol/l. The oil paints were obtained from Daler Rowney (Bracknell, Berks) – Titanium white (009) and alizarin crimson (515).

The pigment was initially triturated, mixed with water and further grinding took place to thoroughly wet the pigment. The wetted pigment is then briefly mixed with yolk before application, in this case, onto glass microscope slides. The films were been allowed to dry for ~ 7 days in ambient conditions.

The commercial oil paints were prepared by painting directly painting onto microscope slides and allowed to dry for ~ 12 days.

Instrumental

Perkin Elmer System 2000. Excitation wavelength 1064nm, Backscattering configuration, 8 cm-1 resolution.

Results and Discussion
FTR spectrum of egg yolk tempera

Although lacking suitable backing material, to enhance the collection of the scattered light, a spectrum was obtained and tentative and alternative assignments have been made.

Figure 1. FTR spectrum – Egg yolk tempera

Raman Shift Intensity Attributed to Comments
3012 medium n=CH cis unsaturation TAG
2900+/- 100 strong n -CH protein & TAG
2721 weak n CH(O) aldehyde ?
1741 weak n C=O ester carbonyl TAG
1658 medium n C=C /Amide I unsaturated TAG or protein amide groups
1522 weak ? unassigned
1441 strong d CH CH2 deformation – protein & TAG / dNH2 (7)
1303 weak d CH CH2 twist rock – protein & TAG
1270 medium Amide III / d C=H n CN + d NH / cis unsaturation / dCH2 (7)
1160 weak n C-O ? alcohol TAG
1082 medium n C-O O-CH2– TAG

Table 1. Tentative assignments of bands present in egg yolk tempera (TAG = Triacylglycerol)

FTR spectra of pigmented media

All the samples examined gave Raman spectra, except for vermilion (HgS) tempera. Sample decomposition appeared to be occurring when attempts were made to collect the spectrum.

Figure 2. FTR spectra of tempera & drying oil films

Material Pigment ord span Comment
1. Egg yolk tempera None 6 Spectrum of medium alone.
2. Lead white tempera 2PbCO3.2Pb(OH)2 30 Medium + 3290/3057/1364cm-1 + 1051cm-1 pigment feature sharp
3. Yellow ochre tempera Fe2O3.H2O 20 ? broad band fluorescence pigment feature 388cm-1
4. Terre verte tempera FeMgKALOSi 6 Medium + pigment feature 551cm-1
5. Ultramarine tempera Na8[Al6Si6O24]Sn 4 Medium + pigment features 547, 354cm-1 noise
6. Titanium dioxide oil paint TiO2 30 Medium + features 990, 610, 452cm-1 broad band 3100-2500cm-1
7. Alizarin oil paint C14H8O4 3 medium + many sharp bands some possibly attributable to aromatic pigment,   noise. See below

Strong features, where present in agreement with (5). Table 2. Spectra of pigmented media – general comments

Lead white tempera

Here the pigment scattering enhancement is realised with a spectrum five times more intense than that of the unpigmented material. Three additional bands that probably arise from the media or media pigment products are also present and possible assignments are: 3290cm-1 Amide A (n NH) and 3057cm-1 aromatic n CH arising from the tyrosine and phenylalanine content of yolk. These bands may be lost in the noise of the corresponding spectrum of the egg yolk film. The band at 1354 cm-1 is unassigned. The media features show an anticipated similarity with those of egg yolk tempera as no significant ageing has occurred in the sample.

The infrared spectrum of this material would show a broad n asCO3 2- feature effectively obscuring the region between 1500 and 1350 cm-1. Four features are infrared active in lead white and the n sCO3 2- occurring at ~ 1046cm-1 is a distinctive but small, sharp, band. The band shown in the Raman spectrum at 1051 cm-1 is probably the stronger, corresponding, Raman active vibration. A small difference between the infrared and Raman shifted frequencies is expected although I have not been able to ascertain why this difference arises.

Yellow ochre tempera

The spectrum of yellow ochre tempera is overwhelmed by a broad band that extends over most of the Raman shifted region. The media features are weak. We can consider possible reasons that may give rise to a broad band of this nature.

  • Water. At a relative humidity of 30-90% 0.3-0.5g of water is associated with 1g of protein[11] In Raman spectroscopy the d OH vibration of water occurs at 1600 ± 1000cm-1 and is broader than its infrared counterpart. The very weak n OH of water occurs between 3700-3200cm-1 [12]. A weak band centered at ~3300cm-1 can be seen in the spectrum of the ochre paint but it is also visible in those spectra that do not exhibit the broad feature. Inspection of the FTR spectra of proteins [13] would tend to favour the assignment of n N-H (Amide A) of the protein. The pigment here is hydrated iron oxide (Fe2O3.H2O) and it is noteworthy that this is the only pigment of those samples chosen that is hydrated and that exhibits this broad band. FTR of the pigment alone would confirm the pigment as the source of the broad band and deuteration may identify dOH as its cause.
  • Fluorescence. The band is very broad, extending from near the excitation line at 1064nm to ~ 1560nm. Without absorbance data on this pigment it is venturesome to speculate on fluorescence as origin of the band. although perhaps it could be attribute the “non-classical” fluorescence that can occur in FTR.

Terre Verte

The spectrum here is as weak as the egg yolk but appears much noisier. On examination under the optical microscope the paint has many small air bubbles present. Features associated with the media are clearly visible.

Ultramarine

The signal to noise ratio for this spectrum is very low but bands attributable to the media are observable

Commercial titanium white oil paint

The spectrum of this material shows two distinct bands attributable to TiO2 rutile form, titanium octahedrally co-ordinated to oxygen, at 610 & 452cm-1. A sharp band at 991cm-1 probably arises from the addition of barium sulphate [5] as an “extender” used to dilute the pigment. A weak broad band underlies the n CH region which, if this were an infrared spectrum, would be attributed to the n OH and combination bands of carboxylic acids. It would be of interest to prepare a further sample to verify the existence of this band or attribute it to an artefact. The bands attributable to the media are discussed below when the media are compared. Sharp bands that appear to be non vibrational in origin are discussed along with the spectrum of alizarin.

Alizarin oil paint

The film examined here was almost transparent and so the low signal to noise was anticipated. Further sample preparation may improve this. Sharp lines at ~547, 2083 & 2852cm-1 (1129, 1367& 1529nm respectively) may arise from the reflection of non lasing lines from within the laser [13]. Similar peaks occur, but to a lesser extent, in the spectrum of TiO2/oil. If the lines did arise from the laser then we would expect them to be reflected more efficiently in the TiO2 sample. An alternative explanation is the collection of light from an external source and the bands do show coincidence with the emission lines of mercury although the relative intensities are not retained (Table 2a). A strong possibility exists that mercury vapour in the fluorescent tubes illuminating the laboratory was the source of the lines.

Observed Raman shift (cm-1) Observed lines “corrected” for laser frequency (cm-1) Detected wavelengths (nm) Known wavelengths of Hg emission lines (14)
2863 6535 1530 1529
2229 7169 1395 1395
2083 7315 1367 1367
2028 7370 1357 1357
1483 7915 1263
543 8855 1129 1129
534 8864 1128 1129
Excitation wavelength 1064nm                        9398cm-1

Table 2a.

Sharp peaks which appear more vibrational in origin may be attributable to the alizarin (dihydroxyanthroquinone), for example 1483cm-1 . If examination of alizarin by FTR could confirm these bands as originating from the pigment then FTR may prove diagnostically useful for this natural organic pigment. Of interest is the fact that no detectable Raman signal for alizarin has been obtained using radiation at 514.5 or 632.8nm [5] although its more fugitive partner in the madder root colour, purpurin (trihydroxyanthroquinone) does produce a good Raman spectrum using visible excitation wavelengths.

Comparison of FTR spectra of egg & oil based media

Figure 3. FTR spectra – Egg yolk tempera/Lead white tempera

Figure 4. FTR spectra – Egg yolk tempera/Titanium white oil paint

 Figure 5. FTR spectra – Lead white tempera/Titanium white oil paint

Band (cm-1) Egg Yolk tempera Lead white tempera Oil/ TiO2 Comments
~1654 Y Y Y Possibly mixed band – See below
~1526 Y Y x unassigned
~1365 x Y x Possibly arising from phospholipid ~ see below
~1266 Y Y x n CN + d NH / protein dCH2(7)
~1162 Y Y x unassigned
~1005 x Y x Possibly arising from phospholipid ~ see below

Table 3. Comparison of bands observed in egg and oil based samples Y observed x not observed

The presence of the band at 1656 cm-1 in both types of material could lead to an assignment of n C=C and not Amide I. However, when ratioed against the n CH peak height in each spectrum the relative peak height of this band is highest in egg tempera (0.27), lower in lead white tempera (0.23) and lowest in the titanium white/oil (0.18). The measurement of the n CH band in the oil paint was made from a manually corrected baseline as a broad band broad band underlies the region. The relative intensity findings are supported intuitively. Egg yolk alone would show the Amide I & the unsaturated element, the lead tempera would have dried to a certain extent (lead white is noted as a fast drier of oils) and shows Amide I & a reduced unsaturated element and titanium white paint shows simply its remaining unsaturated element. If the band at 1656 cm-1 is assigned as arising from both vibrations then it may be utilised to follow changes in unsaturation as the presumption that the Amide I band should remain constant during ageing time scales proposed is not unreasonable.

Bands at 1365 & 1005 cm-1 present in lead white tempera may arise from phospholipids present in the egg yolk that account for ca 20% dry weight of the yolk. Infrared active P-O vibrations can occur in this region. The bands are perhaps not visible in the yolk tempera spectra because this particular sample, despite being a thick opaque film is not scattering efficiently Preparation of a TiO2/egg tempera film and observation of these bands would confirm egg yolk and possibly phospholipids as their source. Alternatively they may be products of media/pigment interaction in the lead white tempera film.

Raman spectra of pigmented media.

Figure. 6 shows the fluorescence problems associated with the use of an excitation wavelength in the visible region of the spectrum. A contribution to the broad band, at ~ 950cm-1, is present in a number of the spectra and may arise from the silicates of the microscope slides supporting the specimens.

Figure 6. Raman spectra of tempera & drying oil films

Conclusions We have shown that FTR spectra of unpigmented and pigmented paint media can be obtained using coarsely prepared samples. In most cases the spectra produced were free of “fluorescence”. It is possible to distinguish between oil and egg based paints in unaged films by observation of bands at 1526, 1266 & 1162 cm-1 . Optimisation of samples may improve the differential diagnosis. Although the distinction between unaged oil & egg in infrared spectroscopy is facile, in aged films the situation becomes much less clear.

The unsuitabilty of Raman spectroscopy using excitation in the visible region for the examination of media is confirmed.

FTR may offer a method whereby the aging of paint films of this type can be further investigated by the loss of species or the formation of products.

References:

  1. THEOPHILUS, P., On Divers Arts, Trans. Hawthorne, J.G., & Smith, C.S. Dover Publications, New York (1979).
  2. THOMPSON, D.V., The materials and techniques of medieval painting. Dover publications, NewYork, (1956).
  3. BEST, S.P. et al., Identification by Raman Microscopy and visible reflectance spectroscopy of pigments on an Icelandic manuscript. Studies in Conservation 40 (1995) 31-40
  4. CLARK, R.J.H et al, The identification of Lapis Lazuli on Medieval Pottery fragments from the South of Italy. Spectrochimica Acta A , 53, 597-603 (1997)
  5. BELL, I.M. et al.,  Raman spectroscopic library of natural and synthetic pigments (pre ~ 1850AD) Spectrochimica Acta A, 53A, 2159-2179 (1997) Internet URL: http://calcium.chem.ucl.ac.uk/webstuff/resources/raman/speclib.html
  6. HENDRA, P.J., What is Raman Spectroscopy? Internet J. Vib. Spect., 1(1998) 16.
  7. HOWELL, G.M. et al.,  FTR of ivory: a non-destructive diagnostic technique. Studies in Conservation, 43, (1998) 9-16
  8. HIGUCHI, S.,et al., Examination of the photochemical curing and degradation of oil paints by laser Raman spectroscopy. J. Appl. Spectr, (1997) 51, 8, 1218-1223
  9. MEILUNAS, R.J.et al, Analysis of aged paint binders by FTIR spectroscopy, Studies in Conservation 35 (1990) 33-51
  10. PILC, J. WHITE, R., The application of FTIR-microscopy to the analysis of paint binders in easel painting, National Gallery Technical Bulletin, 17 , 73-84
  11. KARPOWICZ, A. Ageing and deterioration of proteinaceous media. Studies in Conservation, 26 (1981), 153-160
  12. NAKANISHI. K., SOLOMAN, P. Infrared absorption spectroscopy 2nd Edit., Holden-Day, San Francisco, (1957)
  13. HENDRA, P.J. et al. Fourier Transform Raman Spectroscopy; Instrumentation & Chemical Applications. Ellis Horwood, Chichester &New York (1991)
  14. 14. The Handbook of Chemistry & Physics, 38th Edition, Chemical Rubber Publishing, Ohio, (1956).
  15. CENNINI, C., The craftsman’s handbook “Il libro dell’arte”. Trans. Thompson. D.V.,Dover Publications, (1960).
  16. LASERNA, J.J.(Ed.), Modern Techniques in Raman Spectroscopy. John Wiley & Sons, Chichester & New York (1996)

Received 1st April 1998, received in revised format 24th April 1998, accepted 25th April 1998.