Pigment Dates and the Anachronism Test

The asymmetry: an anachronism convicts, a clean result acquits nobody

Pigment dating rests on a single logical asymmetry. If a paint layer contains a substance that was not manufactured until a given year, that layer was applied after that year. The inference runs one way only, and it is close to unanswerable, which is why it has retired more pictures than any argument from style. The reverse inference does not exist. A painting in which no anachronistic material is found has not been authenticated; it has merely failed to be convicted.

The Courtauld Gallery's version of Dirck van Baburen's The Procuress makes the point cleanly. It was given to the Courtauld in 1960 by Geoffrey Webb, one of the Monuments Men, who believed Han van Meegeren had painted it. Scientific analysis in 2009 identified no modern pigments, which was treated at the time as a surprise. More recent chemical analysis found Bakelite, and because van Meegeren is the only forger known to have used it, the picture is now positively attributed to him. Two rounds of analysis on the same object, opposite conclusions, and only the second was capable of a verdict.

Two further cautions belong in every report. Restoration contaminates the sample: titanium white in a retouching is evidence about the restorer, so the position of the sample in the layer structure matters more than the pigment name. And manufacture is not adoption. Titanium white was industrially available from 1918, yet a survey of 179 twentieth-century paintings in Dutch collections found the pigment convincingly in use by Dutch artists only from the 1970s onward. The manufacturing date is the exclusion boundary. The adoption date is a probability, and probabilities do not convict.

The dates that carry weight, and how they move

Only a handful of introduction dates are firm enough to build an argument on, and each has a documented history that a specialist should be able to cite rather than assert.

Prussian blue is the instructive case because its boundary has moved. The pigment was first synthesised in Berlin by Johann Jacob Diesbach at about 1706, and was available at the Royal Academy of Arts in Berlin no later than 1709. The received view, stated by Harley, was that painters rarely used it before the 1720s and that it became well established only by the middle of the century. Jens Bartoll's study of more than fifty French eighteenth-century paintings from the collection of Frederick II of Prussia, examined with mobile energy-dispersive X-ray fluorescence at 45 kV and 600 microamps, optical microscopy at up to fifty times magnification, and reflectance spectroscopy from 360 to 740 nanometres, found Prussian blue in two Watteau paintings conventionally dated to about 1710: La mariee de village at Sanssouci and the Embarkation to Cythera in the Staedel Museum, Frankfurt. The earliest dated painting known to contain it remains Pieter van der Werff's Entombment of Christ of 1709, also at Sanssouci. The exclusion boundary is the earliest evidenced use, and it can move against you.

Bartoll's method also shows why a pigment identification is harder than it sounds. Once copper and cobalt were ruled out by fluorescence, the remaining candidates were Prussian blue, ultramarine and indigo, and X-ray fluorescence cannot separate them: indigo is organic and invisible to it, while silicon and aluminium from the ochre in the ground appear in almost every measurement. The identification came from particle size under the microscope, where ultramarine appears in large particles while Prussian blue and indigo are powders below one micron, combined with reflectance spectra in which ultramarine and indigo reflect strongly above 650 nanometres and Prussian blue does not.

At the other end of the timeline, lead-tin yellow first appears in paintings around 1300 and drops out of use around 1750, which makes its presence an argument for an early picture and its absence in a supposed seventeenth-century work worth a question. It exists in two forms, the common type I, a mixed oxide of lead and tin, and a type II that also carries tin oxide and traces of silica. Artificial ultramarine follows a successful synthesis in 1828 and rapidly displaced the natural pigment. Zinc white was marketed by Winsor and Newton in 1834 but cost over four times as much as lead white and became popular only after 1850. Lead white itself was banned or restricted for interior painting in most countries between 1905 and 1930 under the Geneva White Lead Convention, and its reported presence in paintings falls from effectively universal in the early twentieth century to about a tenth by 1945.

Titanium white carries the modern boundary. Industrial production of composite titanium white pigments began in November 1918 at Fredrikstad in Norway; pure anatase pigments are reported after 1926; surface treatments and rutile grades entered the market in the 1940s, and the composites were discontinued around 1970. There is a finer instrument inside it. Niobium is an impurity of the sulfate process only, so its absence from an X-ray fluorescence spectrum indicates a rutile pigment and therefore a date after 1959.

Beltracchi: every test anticipated except the white

Wolfgang Beltracchi's Rotes Bild mit Pferden, presented as a Heinrich Campendonk of 1914, collapsed in 2011 when scientific analysis at the Doerner Institut in Munich and by Nicolas Eastaugh in London established that titanium white was present in the paint layer. A picture dated 1914 cannot contain a pigment first manufactured in November 1918. The Cologne Regional Court gave judgment in the associated civil case on 28 September 2012.

What makes the affair worth study is everything the pigment did not do. A comparative technical study at the Cologne Institute of Conservation Sciences, which examined eighteen genuine Campendonks alongside eight Beltracchi forgeries that had passed through a Cologne auction house, assembled a set of material observations that convicted nothing individually and converged completely. After 1913 Campendonk preferred fabric that had been primed before it was stretched; on the Rotes Bild the ground was plainly applied after stretching, as the tacking edges show. Under magnification the grounds of the forgeries are frequently sanded before painting, with canvas threads left exposed beneath the paint, consistent with an earlier image having been scraped from an old fabric. Tacking margins were lightly patinated to look dirty. The stretcher was of a type uncommon in Germany, whereas every examined Campendonk used a keyed stretcher with two keys in each corner; the authors are careful to add that because stretchers are routinely replaced, this observation can only support a suspicion.

Beltracchi had planned for the instruments he knew about. In his own account he could not remove the original painting entirely from the old canvas, so he built the surviving diagonal, a path with buildings and a large tree in the middle ground, into his composition so that those elements would not appear as anomalies under X-ray examination. The radiograph does indeed show parts of an old, discarded composition. He anticipated radiography and defeated it. He did not anticipate a tube of white.

Van Meegeren, Bakelite, and the measurement that finally closed the case

Han van Meegeren's method was designed against the tests of his day. In the first half of the twentieth century the practical age test was to swab a little alcohol onto the surface and try to insert a heated needle: paint residue on the swab, or a needle that penetrated easily, indicated a recent picture. Van Meegeren mixed phenol-formaldehyde resins into his paint so that the film hardened, and the swab and the needle then told the examiner what he wanted to hear. Paul Coremans's team identified the resins after the war.

Even so, the argument did not close. As late as the 1960s some experts held that Christ and His Disciples at Emmaus, the best of the group, was a genuine Vermeer, and that van Meegeren had confessed to it only to escape a charge of treason. What closed it was a measurement of natural radioactivity in the lead white rather than any pigment identification. The method was developed with the support of the National Gallery of Art in Washington and applied by Bernard Keisch. In freshly smelted lead white, lead-210, with a half-life of 22 years, is present far in excess of its long-lived parent radium-226, whose half-life is 1,600 years; in a pigment several centuries old the two have returned to balance. Every van Meegeren tested returned paint that could not be more than a few decades old. The cover of Keisch's own booklet for the United States Atomic Energy Commission shows a picture once believed to be by Frans Hals, proved by polonium-210 and radium-226 measurements to be no more than fifty years old.

Two details from that episode belong to any modern reading of a forgery file. The radiograph of the Emmaus shows traces of the original painting van Meegeren scraped off in order to obtain an old canvas, the same manoeuvre Beltracchi would attempt seventy years later. And the sale figure usually attached to the picture, about 300,000 US dollars, is a conversion: the painting was sold in the Netherlands in guilders, so the dollar sum is derived, is dependent on the date of conversion, and should be replaced by the primary Dutch record.

Knoedler: materials convicted a group more surely than a picture

Between 1994 and 2011 the Long Island dealer Glafira Rosales brought thirty-one Abstract Expressionist paintings to Knoedler and to another New York gallery; Knoedler sold twenty-four of them. All were painted by Pei-Shen Qian, an artist then living in Queens. These were dealer transactions rather than auctions, so there is no hammer price and no buyer's premium to separate; the figures below are taken from a list prepared by Knoedler and filed in federal court, reported by The Art Newspaper in 2013.

The spreads are instructive on their own. Knoedler paid Rosales 950,000 US dollars in 2001 for a Pollock, Untitled, 1950; Pierre Lagrange bought it in November 2007 for 15.3 million dollars, about 17 million with commission fees. Knoedler paid 950,000 dollars for a Rothko, Untitled, 1956, which Domenico and Eleanore De Sole bought in December 2004 for 8.3 million. Michael Hilti paid 5.5 million in November 2002 for a Rothko for which Knoedler had paid 775,000.

The materials evidence, developed largely by James Martin of Orion Analytical, was decisive across the group. Testimony reported from the 2016 De Sole trial recorded that works contained paint not available until well after their supposed date of creation, and that one surface had been sanded down with an electric sander, a procedure incompatible with Motherwell's technique.

But the group is the unit of proof, and the exhibition record makes that plain. On the very first painting Rosales brought to the gallery, scientific analysis identified no anachronistic pigment at all. What it found was that the signature had been toned with a yellow colour, probably in an attempt to make it look older than it was. One picture yielded an ambiguous material result. Thirty-one pictures sharing one provenance story, one hand and recurring anachronisms did not.