Short answer
Most pigment changes documented in paintings are not fading but conversion into a different compound, and three routes account for nearly all of them. The pigment is reduced or oxidised at its surface (chrome yellow, zinc yellow, cadmium yellow, red lead); it reacts with chloride or other salts from outside (vermilion, cinnabar, azurite); or it reacts with its own binder (smalt, emerald green, verdigris, the lead pigments). Which route a given paint takes depends on its exact composition and its surroundings, which is why one pigment name can cover both stable and unstable paint.
Several bright inorganic pigments owe their colour to a metal or a sulfur atom in one particular oxidation state, and lose it when that state changes. In chrome yellow, light reduces hexavalent chromium to trivalent chromium compounds and the yellow turns brown. Zinc yellow, the related chromate Seurat used, does the same in two steps: chromate first becomes orange dichromate, then green-brown Cr(III). Cadmium yellow goes the other way — its sulfide is oxidised to colourless sulfate, so the paint pales and grows white crystals. Red lead, reduced by light, passes through a rare lead compound called plumbonacrite on its way to white lead carbonate. In medieval window glass the browning tracks how far manganese has been oxidised. None of these is fading in the dye sense: the pigment is still there, as something else.
Vermilion was long described as a pigment that simply blackens in light. When the National Gallery examined ten of its own paintings and two from the Courtauld, chloride turned up in every discoloured sample, and the authors concluded that the vermilion was not inherently unstable but had been made so by chloride ions from the environment — dirt being one carrier. The white product they identified is calomel, a mercury chloride, and it is the mixture of white and black products that makes ruined vermilion look grey. Red wall paintings from the Pompeii area tell a parallel story with a twist: grey areas hold mercury–chlorine compounds, but the black coating is gypsum formed by sulfation of the lime ground, lying over intact cinnabar. On plaster, azurite exposed to chloride becomes green paratacamite rather than the malachite once assumed.
Oil paint is not an inert glue. As it ages it releases fatty acids, and some pigments dissolve in them. Emerald green reacts to give copper soaps and mobile arsenic compounds that spread through the film and into the varnish; in Rousseau's Descente des vaches the greens have gone brown and the picture is barely legible. Verdigris gives up its copper to resin and fatty acids so readily that browning follows how easily the copper can be extracted. Smalt, a cobalt glass, loses potassium to the oil — potassium soaps were found in every degraded sample the National Gallery studied — and without potassium the cobalt stops colouring the glass blue. Red lead and lead-tin yellow form lead soaps that gather into white lumps large enough, in a Goya portrait, to see with the naked eye.
Whether metal soaps are damage is itself debated: a review titled 'Friends or foes?' shows that painters cooked oil with lead on purpose, forming the same soaps within an hour.
A recurring finding in the synchrotron studies is how shallow the altered layer is. The reduced-chromium layer on chrome yellow is one to three micrometres deep; the oxidation front in Ensor's cadmium yellow is one to two; the altered zinc yellow on La Grande Jatte is confined to the top few; the gypsum crust on Pompeian cinnabar is about five; and the National Gallery describes the vermilion crust as extremely thin even where the visual effect is severe. Colour is decided at the surface, where light first meets the paint, so a film that is chemically intact through ninety-nine per cent of its depth can still look wrong. It also explains why these changes escaped bulk analysis for so long, and why any abrasion of a paint surface matters.
A pigment name is a poor predictor on its own. Only sulfate-rich, orthorhombic chrome yellow darkened in model paints; plain lead chromate did not, and both are present in Van Gogh's Sunflowers. Cadmium yellow failed mainly through light in Ensor's paintings and mainly through moisture and chlorine compounds in Munch's The Scream. Prussian blue's permanence depends chiefly on how much white or extender it was mixed with, and the National Gallery points to a Gainsborough and a Reynolds in which it is still of very good colour. Emerald green browned in a thickly painted, medium-rich picture. The practical consequence, drawn by the researchers themselves, is that risk has to be mapped painting by painting; the dataset behind this page records each change with the object it was found in for that reason.
| Colourant | Mechanism as published | Driver | What changed | Studied in |
|---|---|---|---|---|
| Chrome yellow (lead chromate and sulfate-rich lead sulfochromate) in oil | Hexavalent chromium is reduced to trivalent chromium compounds in a surface layer a few micrometres deep. Sulfate-rich, orthorhombic forms are the ones that darken; plain monoclinic lead chromate is lightfast. | Light and UV | Bright yellow turns brownish at the surface. | Vincent van Gogh, Sunflowers (Van Gogh Museum, Amsterdam) |
| Zinc yellow (zinc potassium chromate) in oil | Chromate converts to orange dichromate and is reduced to Cr(III) in the top few micrometres of the paint. | Light and UV; Gaseous pollutants; Humidity and water | Yellow, green-yellow and orange strokes have become ochre, olive green and reddish brown. | Georges Seurat, A Sunday on La Grande Jatte—1884 (Art Institute of Chicago) |
| Cadmium yellow (cadmium sulfide) in oil | Sulfide is oxidised to sulfate; soluble cadmium sulfate moves to the surface and recrystallises as whitish globules. | Light and UV; Oxygen | Bright yellow fades, with small white globules on the surface. | Paintings by James Ensor |
| Cadmium yellow (cadmium sulfide) in oil | Cadmium sulfide oxidises to cadmium sulfate and sulfites, with dissolution, migration and recrystallisation of water-soluble phases. | Humidity and water; Chlorides and salts | Alteration of the yellow paint to pale sulfate products. | Edvard Munch, The Scream (c. 1910, Munch Museum, Oslo) |
| Red lead (minium) in oil | Light-induced reduction of red lead leads, through the rare intermediate plumbonacrite, to white lead carbonates. | Light and UV | Orange-red turns white. | A painting by Vincent van Gogh (one microsample) |
| Vermilion (mercury sulfide), dry-process or mineral cinnabar | Chloride from the environment takes part in a light-driven reaction. One proposal is that metallic mercury nanoparticles turn the surface black and that further reaction with chloride gives white mercury chlorides; the National Gallery identified the white product as calomel. | Light and UV; Chlorides and salts | Red turns black or, where white and black products mix, grey. | Ten National Gallery paintings of the fourteenth to seventeenth centuries, including the San Pier Maggiore Altarpiece; Bernardo Daddi, Crucifixion triptych (Courtauld Gallery) |
| Cinnabar (mercury sulfide) on lime mortar | Two routes: mercury–chlorine compounds form in grey areas, and sulfation of the calcite ground produces a black gypsum coating. The long-suspected conversion to black metacinnabar was not detected. | Chlorides and salts; Gaseous pollutants | Deep red turns grey or black. | Red wall paintings from the Pompeii area |
| Azurite (basic copper carbonate) on plaster | Where chloride ions are available, azurite converts to paratacamite, a green basic copper chloride — not to malachite, as often assumed. | Chlorides and salts; Humidity and water | Blue turns green. | Wall painting of San Antonio Abate, church of San Pietro at Quaracchi, near Florence |
| Smalt (cobalt potash glass) in oil | Potassium leaches from the glass into the oil, forming potassium soaps and migrating to the surface; cobalt left in the depleted glass shifts from tetrahedral towards octahedral coordination and stops giving blue. | Adjacent materials (binder, support, varnish); Humidity and water | Blue turns grey, brownish or colourless, sometimes blanched. | National Gallery, London, and Louvre paintings containing smalt |
| Emerald green (copper acetoarsenite) in oil | The pigment reacts with free fatty acids from the oil, forming copper soaps and mobile arsenic species that spread through the paint and into the varnish. | Adjacent materials (binder, support, varnish) | Green turns brown; the picture darkens overall. | Théodore Rousseau, Descente des vaches (Mesdag Collection, The Hague) |
| Verdigris (copper acetate) and copper greens in oil–resin media | Resin acids and fatty acids extract copper(II) from the pigment; the copper diffuses as carboxylate complexes, including into varnish above. | Adjacent materials (binder, support, varnish); Low oxygen (anoxia) | Transparent green turns brown. | Laboratory or general evidence |
| Red lead and lead-tin yellow in oil | The pigment reacts with fatty acids in the oil to form lead soaps, which aggregate with lead carbonate into translucent inclusions. | Adjacent materials (binder, support, varnish) | White, translucent lumps appear in and erupt through the paint, speckling the surface. | Francisco de Goya, Portrait of Don Andrés del Peral (National Gallery, London) |
| Zinc white (zinc oxide) in oil | Zinc oxide reacts with fatty acids to form zinc soaps, which can aggregate within the paint. | Adjacent materials (binder, support, varnish); Humidity and water | Soap aggregates linked to deterioration of the paint layer. The review abstract read does not describe a colour change. | Laboratory or general evidence |
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactStrong evidence
In all ten National Gallery paintings examined for discoloured vermilion, chloride was involved in the deterioration; the authors concluded the vermilion was not inherently unstable but became so through chloride ions from the environment, and identified the white product as calomel.
Caveat: Paintings made before wet-process vermilion; later vermilion may differ.
FactStrong evidence
In Edvard Munch's The Scream (c. 1910), moisture and mobile chlorine compounds are the key factors promoting oxidation of cadmium sulfide to cadmium sulfate, while light plays a less important role.
Caveat: One painting and aged mock-ups.
FactStrong evidence
Emerald green in oil reacts with free fatty acids from the binding medium to form copper soaps and mobile arsenic species, which contributed to the change from green to brown in Rousseau's Descente des vaches.
FactStrong evidence
Zinc yellow models aged with light and sulfur dioxide turned olive green at 90% relative humidity, containing mostly Cr(III), and ochre at 50%, containing dichromate; the same species were detected in altered brushstrokes of Seurat's La Grande Jatte.
FactStrong evidence
Model oil paints darkened profoundly only when made with sulfate-rich (x ≥ 0.4), orthorhombic lead sulfochromate, with up to about 60% of chromium in the outer layer reduced to Cr(III).
Caveat: Laboratory-made paints under accelerated ageing.
Colourwise interpretationModerate evidence
The altered layer in these pigment changes is typically a few micrometres thick, far thinner than the paint film, yet enough to change the colour seen.
Based on: Colourwise's comparison of the depths reported separately by the cited studies: 1–3 µm for chrome yellow, about 1–2 µm for cadmium yellow and about 5 µm for the gypsum coating on cinnabar.
Source: Degradation process of lead chromate in paintings by Vincent van Gogh studied by means of synchrotron X-ray spectromicroscopy and related methods. 1. Artificially aged model samples (Analytical Chemistry, 2011); Characterization of a degraded cadmium yellow (CdS) pigment in an oil painting by means of synchrotron radiation based X-ray techniques (Analytical Chemistry, 2009); Blackening of Pompeian cinnabar paintings: X-ray microspectroscopy analysis (Analytical Chemistry, 2006)
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.