Varnish removal

Why can stripping a painting's yellowed varnish take the artist's own final glazes with it?

A varnish is a separate coat, brushed on top of a finished picture, often by someone other than the painter. It sits above the paint, not in it. So removing it ought to be the simplest job in conservation — find something that dissolves the resin, leave the paint alone underneath.

And yet cleaning is the operation that generates the loudest quarrels in the field, and the losses it causes are always described the same way: not gouges, not scraped-off passages, but the last things the painter did — the thin darkening glaze over a shadow, the final scumble across a sky. Why should the topmost paint be exactly what a solvent aimed at the layer above it takes?

Start with what the varnish becomes. A traditional picture varnish is a natural resin — mastic or dammar — laid down dissolved in a solvent that then evaporates. Fresh, it is a soft, largely non-polar film that a mild hydrocarbon will lift off cleanly. Then it ages: oxygen attacks it, and the oxidation does two things at once. It introduces polar groups — acids, ketones — so the film's chemical character shifts, and it forms cross-links between the resin molecules, so the film stops being a collection of dissolvable units and becomes one network.

Follow what that implies for the solvent. A more polar film needs a more polar solvent. A cross-linked film cannot really be dissolved at all; it can only be swollen until it breaks up. So the requirement on the solvent rises steeply with age — and here is the first half of the problem, stated as a rule rather than an anecdote: a varnish becomes hardest to remove at exactly the moment it becomes most necessary to remove. The yellowing and the insolubility are not two separate misfortunes; they are the same oxidation, seen through the eye and through the chemistry.

Now turn to the paint. The intuition that the paint is inert is the thing to question. Aged oil paint is also a cross-linked network — and it is not pure. Inside it sit low-molecular-weight fragments: free fatty acids, breakdown products, metal soaps formed between the oil and reactive pigments. Those are not bound into the network; they are held in it. Ask what a polar solvent does when it swells such a film, and the answer follows without any special knowledge: it takes them into solution and carries them out. The paint is not being dissolved, but it is being extracted.

So there are two curves, and they move toward each other. The polarity needed to shift the varnish climbs with age; the polarity at which the paint starts giving up material stays roughly where it was. The window between them narrows. That, and not clumsy hands, is why cleaning gets more dangerous the older the picture is.

Why the glazes specifically, though? Because of what a glaze is. It is a thin, transparent, medium-rich, pigment-poor layer — and many painters thinned that medium with resin, sometimes the very same resins later used for varnishing. A final glaze can therefore be chemically closer to the varnish than to the opaque paint beneath it. Add that it is measured in a few micrometres, so there is no thickness to spare; that it is the outermost paint, in direct contact with the varnish and often partly interpenetrated by it; and that varnish applied to a still-porous surface sinks into it. The boundary a conservator is trying to find is, in the worst cases, not a boundary.

Is any of this testable, or is it just a plausible story? Here is the test. If glaze loss were purely mechanical — swab pressure, rubbing — then applying solvent without abrasion, as a gel or a vapour, would be safe. It is not: aged oil films exposed to polar solvents with no rubbing at all still swell and still give up soluble fractions, which is measurable by weighing the extract and analysing it. The observation that would refute the model: cross-sections taken before and after a wholly non-abrasive clean showing the glaze unchanged in thickness, medium content, and composition. Then the risk would collapse back to technique, and better swabs would solve it.

An honest qualification: how much the extraction matters is genuinely argued. That soluble material leaves is not in dispute. Whether losing it embrittles the film, or whether a well-cross-linked paint is essentially unharmed by it, is still contested, and different laboratories weigh it differently.

The common picture — a clean layer above a clean layer, needing only the right solvent — fails on three counts at once, and it is worth keeping all three separate. Chemically, the two layers are not distinct kinds of material. Physically, the glaze is thin enough that a small loss is a total one. Historically, the interface has usually been disturbed already, by earlier varnishes, earlier cleanings, and retouchings that are themselves resinous.

Notice also what the diagram below cannot show: reversibility. Solvent action on paint is not a switch that trips at some threshold; it is a rate, so time and how many passes matter as much as which liquid is in the swab. And unlike almost everything else a conservator does, this one cannot be undone. That is why the discipline's answer is procedural rather than chemical — test in an inconspicuous area, document what was used, stop while material remains — and why "partial cleaning", leaving a thin residue rather than going to bare paint, is a defensible position and not merely a timid one.

• How solubility parameters were adapted from industrial chemistry to map safe cleaning windows, and where that map stops being trustworthy.
• Why synthetic varnishes were adopted specifically to keep the removal window open, and what was given up in appearance to get it.

Every term the collection defines is gathered in the glossary.