September 1998 Volume 21 Number 1

Evaporation of Fatty Acids and the Formation of Ghost Images by Framed Oil Paintings

by Michael R. Shilling, David M. Carson, and Herant P. Khanjian


On occasion, oil paintings are framed with a protective glass layer to prevent damage from vandalism and accidents. Over time hazy films, termed "ghost images", may appear on the inside surface of the protective glass which, ultimately, impair proper viewing of the painting. To enhance visibility, these so-called "ghost images" are periodically removed by museum preparators using glass cleaners.

Studies by Williams at the Canadian Conservation Institute, using infrared spectroscopy, determined that ghost images consist almost entirely of palmitic acid (Williams 1989). To explain his findings, Williams speculated that volatile ketones are released by the paint, some of which undergo oxidation on the glass surface to form palmitic acid.

Somewhat later, Michalski postulated another explanation for Williams's findings, that free fatty acids simply may evaporate from the painting and condense on the glass. He reasoned that enrichment of the image in palmitic acid occurs because the boiling point of palmitic acid is much lower than that of stearic acid (Michalski 1990). However, these ideas were never subjected to experimental verification.

Thus, a study was undertaken at the Getty Conservation Institute to further elucidate the composition of ghost images, and to determine whether evaporation may, indeed, play a role in the loss of fatty acids from oil paints. Thermogravimetry (TG) was used to determine the evaporation rates of palmitic, stearic, and azelaic acids in an inert gas atmosphere. Additionally, gas chromatography-mass spectrometry (GC-MS) was used to analyze ghost images that evolved from numerous test paints, to identify the materials that were present in the images and to determine which paints were most likely to produce ghost images.


Thermogravimetric analyses were conducted on samples of free fatty acid. The samples were heated from 25° C to 300° C using various heating rates. Sample weight was recorded as a function of temperature for each test, and subsequently plotted as percent of original weight versus furnace temperature. A reaction that leads to weight loss (such as evaporation) will give rise to a step on the TG curve.

Each set of TG test results were processed using a thermal analysis kinetic evaluation program to obtain routine Arrhenius equation kinetic parameters (activation energy and reaction rate). From the kinetic data, calculations were performed in order to estimate the half-time of the evaporation rate (i.e., the amount of time that would be required for one half of the fatty acid sample mass to evaporate) for a number of ambient temperatures.

Reaction rate and half-time data for the evaporation of the three fatty acids, calculated at four reference temperatures, are listed in Table 1. Based on kinetic evaluation of the TG data, the estimated half-time for the rate of evaporation at 25° C is approximately 40 years for palmitic acid, 140 years for stearic acid, and 80 years for azelaic acid.

Table 1. Reaction rate and half-time data calculated at four reference temperatures, for the evaporation of palmitic, stearic, and azelaic acids.
palmitic acid stearic acid azelaic acid
T,(C) rate, 1/sec t1/2 rate, 1/sec t1/2 rate, 1/sec t1/2
25 6.1E-10 36 years 1.6E-10 136 years 2.7E-10 82 years
50 9.7E-09 2.3 years 3.0E-09 7.3 years 4.5E-09 4.9 years
80 1.6E-07 50 days 5.8E-8 137 days 7.9E-8 102 days
100 8.1E-07 9.9 days 3.2E-07 25 days 4.1E-07 20 days

In addition to TG analysis, ghost images were generated artificially for GC-MS testing purposes. Oil paint samples were placed onto single-depression microscope slides, a second slide placed onto the first with the depressions facing together, and the two slides clamped together. Duplicate samples of each paint formulation tested were placed in separate slide assemblies, with one sample kept on a sand bath set to 80° C for a six week period, and the second exposed to light in a Heraeus Suntest chamber for four weeks, followed by sand bath heating for one week. The test paints used in this study had been stored in the laboratory for up to five years prior to testing (Schilling et al. 1997).

Ghost images which developed on the upper slide well were rinsed with diethyl ether into sample vials, derivatized with Methyl-8 reagent, (from Pierce Chemical Company), and the fatty acid methyl esters were analyzed by gas chromatography-mass spectrometry (GC-MS).

Ghost images were formed by a majority of the paints, but the intensity varied with the type of exposure and the paint formulation. For the sand bath samples, no visible images were produced by cold-pressed linseed and stand linseed oil paints, and the others tended to produce images with variable intensity. On the other hand, much less intense images were formed by the Suntest exposure, but nearly all of the paints formed some visible residue.

Table 2 lists the fatty acid molar ratios for the ghost images formed by the test paints. Although the GC-MS results revealed that high levels of palmitic acid were indeed present in the images, other fatty acids were also detected, such as stearic acid, azelaic acid, and short-chain dicarboxylic fatty acids.

Table 2. Fatty acid molar ratios for ghost images generated by test paints, determined by gas chromatography-mass spectrometry. For entries with a dash, one or both of the fatty acids in the appropriate ratio could not be detected.
          Sandbath     Suntest        
Medium     Pigment A/P P/S A/D   A/P P/S A/D    
Blown Linseed   Lead White     0 11.9 -   1.3 2.3 0.5  
(P/s=1.8)     Vine Black     0.2 6.4 2   1.5 1.9 0.4
      Yellow Ochre   0.4 4.5 1.3   0.4 2.5 0.7  
Cold-Pressed Linseed   Lead White     0 2.8 -   0 0.9 0  
(P/s=1.7)     Vine Black     2.1 - 1.7   3.1 1.8 0.9
      Verdigris   - -     0.2 0.9    
      Yellow Ochre   - 2.4     5.1 1.9 1  
Egg Yolk     Lead White   0 5.2 -   0 4.1 1  
(P/s=3.3)     Yellow Ochre   0 5.6 -   0 1.8 -  
Stand Linseed   Lead White     -       0.2 1.1 0.7  
(P/s=1.8)     Vine Black     0 - -   0.9 2.1 0.8
      Verdigris   - - -   0 1.3 -  
      Yellow Ochre   - - -   0.3 2.3 2  
Walnut   Lead White     0.1 4 -   0.1 2.2 2.4  
(P/s=3.3)     Vine Black     0.2 - 0.9   0.4 1.9 1.6
      Yellow Ochre   6.4 2.3 1.8   0.9 3.5 1.1  

In general, the chromatograms for the sand bath samples were dominated by high levels of saturated fatty acids. The fatty acid compositions in the sand bath ghosts were proportional to the levels of free fatty acids present in the paint samples, but in every instance the ghosts were enriched in the more volatile, lower-boiling species. Accordingly, the P/S ratios of the sand bath ghost images were higher than that of the corresponding paints. On the contrary, paints exposed in the Suntest chamber often formed images that were enriched in dicarboxylic fatty acids, thus providing evidence for the role of light in oxidative breakdown of oil paints.

An additional test was performed to determine what effect molecular weight would have on the extent of fatty acid evaporation. A standard mixture of free monocarboxylic fatty acids (C12, C14, C16, C17, C18, and C20) and dicarboxylic fatty acids (C7, C8, C9, C10) was placed in a glass slide assembly and heated on the sand bath in order to form a ghost image. Upon examining the slide assembly after it was heated on the sand bath, a pronounced white ghost image was visible on the upper slide well, and a much smaller amount of residue remained in the lower slide well. The ghost image that developed on the upper slide well, and the residual fatty acids that remained in the lower slide well, were derivatized individually using Tri Sil reagent, (from Pierce Chemical Company), and analyzed by GC-MS.

Figure 1 is an overlay of GC-MS chromatograms of the silyated extracts from the upper slide well (a), the lower slide well (b), and an untreated fatty acid standard (c). It is clear each of the fatty acids present in the standard, both monocarboxylic or dicarboxylic in nature, were detected in the ghost image.

figure 1 GC-MS chromatograms

Furthermore, the ghost image is significantly enriched in the lower molecular weight fatty acids, whereas the residue in the lower slide well is depleted in the lower molecular weight fatty acids. Area ratios of the peaks in the ghost image versus the residue in the lower slide reveal a 5:1 enrichment for both azelaic acid and stearic acid in the ghost image, and a 16:1 enrichment for palmitic acid in the ghost image.


In analytical studies of aged oil paints conducted a number of years ago at the Getty Conservation Institute, it was discovered that the saturated fatty acid content of drying oil paints may be reduced by exposure to heat and light (Schilling et al. 1997). The largest overall reductions in saturated fatty acid content were observed for heat-aged paints made from yellow ochre mixed with walnut or poppy oil, but many of the other slow-drying paint formulations exhibited moderate reductions.

At the time, these findings were somewhat surprising because saturated fatty acids were believed to be relatively stable marker compounds in drying oil media. In fact, the relative amounts of palmitic and stearic acids are the basis for a method of identification of drying oil media that has been in practice for more than thirty years (Mills 1966).

However, in light of the present study, it is clear that evaporation does play a role in the reduction of saturated fatty acids from the test paints. The TG results demonstrate that saturated and dicarboxylic fatty acids evaporate readily, even when extremely low heating rates were employed. Interestingly, on close examination of glass sample vials that contain oil paint samples we have also observed hazy films on the inner wall of the vials, primarily near the samples.

The TG data are in broad agreement with the observed high levels of palmitic acid in ghost images, although strict comparison between half-time estimates and the composition of ghost images from the test paints is not possible due to limitations in the design of the experimental apparatus.

Williams observed that ghost images were most intense over dark colors and less intense over light colors, which gave rise, more or less, to a negative ghost image of the painted subject. Based on the results from the present study, we may offer an explanation for this phenomenon. In general, slow-drying paint formulations tend to produce intense ghost images enriched in saturated fatty acids; this is consistent with the mechanism of partial hydrolytic decomposition of the triglyceride oil matrix (Boon JJ, Peulve SL, Van der Brink OE, Duursma MC and Rain-ford D, 1996). Paints that were slow to dry include those pigmented with ochre or vine black, or made from walnut or poppy oil media. In addition, ghost image formation was greatly reduced over paints made with lead white and verdigris, presumably due to formation of pigment soaps (Koller and Burmester 1990). Thus, variation in ghost image intensity may be due strictly to the quantity of free fatty acids in paints, which in turn is affected by the specific combination of pigment and medium.

In published studies of crystalline bloom on contemporary art pieces (Koller and Burmester 1990), it has been observed that oil-rich paints made with pigments that do not promote cross-linking (such as alizarin, titanium dioxide, and ochres) are particularly susceptible to the formation of whitish surface deposits of fatty acids. Presumably, the fatty acids were released either from the oil matrix through hydrolysis of the glyceride ester backbone or else from decomposition of pigment extenders (such as stearates), and once liberated were able to migrate through the paint and deposit on the surface. Therefore, these works should be extremely likely to lose fatty acids through evaporation and, if framed behind glass, form ghost images.


Using thermogravimetry and gas chromatography-mass spectrometry, free fatty acids were shown to evaporate readily to form ghost images over oil paints. By expressing evaporation rates in units of half-times, palmitic acid was found to evaporate approximately four times as rapidly as either stearic acid or azelaic acid at room temperature.

Ghost images were found to consist of mixtures of free fatty acids, with higher levels of the more volatile species predominating in both homologous series of fatty acids. Images formed by the heating of test paint samples tended to consist largely of saturated fatty acids, whereas the composition of images formed by exposure of paint to light generally were more enriched in dicarboxylic fatty acids.

Free fatty acids are liberated by hydrolysis of oil triglycerides, migrate through the paint, and subsequently evaporate to form ghost images. In general, slow-drying paints yielded more intense ghost images than did fast-drying paints, presumably because of variation in the degree of hydrolysis of oil triglycerides.


Boon JJ, Peulve SL, Van der Brink OE, Duursma MC and Rainford D, 1996. Molecular aspects of mobile and stationary phases in ageing tempera and oil paint films. Early Italian Paintings Techniques and Analysis, Limburg Conservation Institute, Maastricht. 35-56.

Koller J and Burmester A, 1990. Blanching of unvarnished modern paintings: a case study on a painting by Serge Poliakoff. In: Mills JS and Smith P, eds. Cleaning, Retouching and Coatings: Technology and Practice for Easel Paintings and Polychrome Sculpture: Preprints of the Contributions to the Brussels Congress. London: 138-143.

Michalski S, 1990. A physical model for the cleaning of oil paint. In: Mills JS and Smith P, eds. Cleaning, Retouching and Coatings: Technology and Practice for Easel Paintings and Polychrome Sculpture: Preprints of the Contributions to the Brussels Congress. London: 85-92.

Mills JS, 1966. The gas chromatographic examination of paint media. Part I. Fatty acid composition and identification of dried oil films. Studies in Conservation 11: 92-107.

Schilling MR, Khanjian HP, Carson DM, 1997. Fatty acid and glycerol content of lipids; effects of ageing and solvent extraction on the composition of oil paints. Techne 5: 71-78.

Schilling MR, Carson DM, Khanjian HP, 1998. The composition of solvent extracts from oil paints. Art et Chimie - La Couleur conference, the Louvre.

Wendlandt WW, 1974. Thermal methods of analysis. Second edition. In: Elving PJ and Kolthoff IM, eds. Chemical analysis, volume 19. New York:Wiley and Sons:45-61.

Widmann G, 1982. Kinetic measurements on polymers-applications and limits. Journal of Thermal Analysis 25:45-55.

Williams SR, 1989. Blooms, blushes, transferred images and mouldy surfaces: what are these distracting accretions on art works? Proc. of the fourteenth annual IIC-CG Conference, IIC-Canadian Group, Ottawa. 65-84.

 [WAAC]  [WAAC Newsletter]  [WAAC Newsletter Contents]  [Search WAAC Newsletter]  [Disclaimer]

[Search all CoOL documents]