Radiocarbon and Pigment Dating

What a Date on a Painting Measures

A painting is a stack of materials assembled over time. The panel may have been seasoned for years before a ground went on. The canvas may have sat in a colourman's shop. The oil may have been pressed from seed harvested a season before the tube was filled. A later restorer may have lined the reverse, varnished the front, or retouched a loss with a twentieth-century white. Each of those carbons and minerals has its own clock. The laboratory dates the clock you sample, not the biography you hope for.

The result that matters in court and in a catalogue is often a negative. Canvas fibres with a bomb-peak radiocarbon signature cannot belong to a painting finished before atmospheric nuclear tests. Titanium white in an original layer cannot belong to a picture dated before that pigment reached artists' paints. A panel whose outermost measured ring formed in 1513 cannot have been painted in 1500. Those statements are strong because they rest on physics and industrial history. The matching positive statement is weaker. A canvas grown in the 1880s can carry a later picture. A palette that matches van Gogh's Arles years can still be a copyist who bought the same tubes. A felling date after 1613 does not prove Rembrandt held the brush.

Attribution is a separate argument: style, documents, workshop practice, and the technical image of how the paint was put down. A date can rule a claim out. It cannot, on its own, put a specific artist in.

Radiocarbon Chemistry: Decay, Calibration and the Bomb Peak

Cosmic rays strike nitrogen in the upper atmosphere and produce carbon-14, a radioactive isotope. Plants take that carbon in as carbon dioxide. Animals eat the plants. While the organism lives, its 14C/12C ratio tracks the atmosphere. After death, 14C decays with a half-life of 5,730 years (the figure the Royal Institute for Cultural Heritage in Brussels uses in its public account of the method). Measure the remaining 14C, compare it with a calibration curve built from tree rings and other archives, and you obtain a calendar range. Conventional "before present" (BP) ages still use Libby's older half-life and treat 1950 as the zero year, because after that date fossil-fuel carbon and nuclear-test carbon both scramble the atmospheric record.

Early counting needed grams of carbon, enough to destroy a tacking margin. Accelerator mass spectrometry (AMS), demonstrated in 1977 by groups publishing in Science (Bennett and colleagues; Nelson, Korteling and Stott), counts 14C atoms rather than waiting for them to decay. Sample sizes dropped to milligrams, then, with gas ion sources on compact AMS instruments such as MICADAS, to tens of micrograms of carbon. A paintings laboratory can now date a canvas thread or a binder scraped from an existing crack.

Atmospheric 14C has never been a flat line. The international IntCal curves convert a radiocarbon age into calendar years for material older than 1950. For younger material, laboratories use post-bomb curves (Northern or Southern Hemisphere). Atmospheric nuclear tests from the mid-1950s drove 14C in the northern troposphere to about double its pre-test level by 1963–64. The Partial Test Ban Treaty of 1963 stopped most atmospheric detonations. The excess then declined as the carbon mixed through oceans and biosphere and as fossil-fuel carbon dioxide diluted it. The resulting "bomb peak" is a steep, well-measured spike. A sample with more 14C than the 1950 atmosphere (reported as percent modern carbon, pMC, above 100) formed after the tests began. Because the curve rises and then falls, one pMC value can match two calendar windows: one on the way up, one on the way down. Hendriks and colleagues, dating the oil binder of a known fake in 2019, obtained 1958–61 or 1983–89. Both windows post-date 1950. That was enough.

The bomb peak is why radiocarbon now reaches modern pictures and forged "old masters." A linen plant harvested in 1910 and a linen plant harvested in 1960 do not look different under a microscope. Their 14C content does.

How a Sample Is Taken and Interpreted

Sampling a painting is a conservation decision before it is a physics experiment. Every fibre or chip is a loss. Codes that govern treatment (AIC, Icon, ICOM-CC) apply here: document the location, take the smallest amount that can answer the question, and refuse a sample that would destroy original design when a tacking edge, a reverse thread, or an existing craquelure already offers carbon. A private owner who posts a flake to a commercial "authentication" site without a conservator in the room is not following that logic. Museum practice puts a paintings conservator and a scientist on the same object: one to choose a location that does not collapse the image, the other to say whether the carbon in that location can be cleaned of varnish, lining adhesive and later retouch.

The sequence below is professional laboratory practice, not a kit. Solvent recipes and acid strengths belong in the lab notebook, not in a public how-to.

Choosing the material that answers the question

The question decides the sample. If the claim is "this canvas was woven in Léger's lifetime," you date a canvas fibre, after checking that the cloth is original and not a lining. If the claim is "this image was painted in 1866," canvas alone is the wrong clock: a forger can buy an old support. Hendriks, Hajdas, Ferreira and colleagues showed that in 2019 on a picture signed "Sarah Honn" and dated 1866. The canvas age was compatible with a nineteenth-century support. The oil binder, measured on a paint chip under 200 micrograms, carried bomb-peak 14C. The seeds that produced the oil were harvested in 1958–61 or 1983–89. Robert Trotter had already admitted at trial that he painted it in 1985.

Beck and colleagues at LMC14 (Université Paris-Saclay), working with the French Central Office for the Fight against Illicit Trafficking in Cultural Property (OCBC) in 2022, met the inverse problem. White paint and stretcher wood from an alleged Impressionist picture gave pre-bomb ages on the calibration plateau from the late seventeenth century to about 1950, too wide to exclude the artist's lifetime. Canvas bast fibres measured 108.59 ± 0.27 pMC: after 1950, calibrating to 1957 or 2000–2003. The alleged artist had died in the 1940s. The support was not a lining. The image could not be that artist's work.

Pigment screening comes first when the target is binder. Inorganic whites (lead white, zinc white, titanium white) contribute little or no contemporary organic carbon. Hendriks's group chose a white passage that Raman and related methods showed was inorganic pigment plus oil, then removed varnish with ethanol and carbonates with hydrochloric acid so that the AMS saw the binder. A sample mixed with chalk, later varnish, or a synthetic resin will date the mixture.

Pretreatment, combustion and the AMS measurement

Laboratories treat canvas and wood with solvent extraction (often Soxhlet) and an acid–base–acid wash to strip carbonates, humic acids and many conservation adhesives. Paint chips need a protocol matched to the layer: varnish off, geological carbonate off, binder left. ORAU's published pretreatment methods (Brock, Higham, Ditchfield and Bronk Ramsey, Radiocarbon, 2010) are the reference many art papers cite. KIK-IRPA in Brussels runs the same logic on its MICADAS accelerator, in house since 2013.

The cleaned sample is combusted to carbon dioxide. Larger samples are graphitised and sputtered as a solid target. Microgram samples can go as gas into a gas ion source. The accelerator strips molecular ions, counts 14C against 12C and 13C, and reports an age or a pMC value with an uncertainty. Constant-contamination corrections matter at microgram scale: a few micrograms of modern dust shift a small sample more than they shift a milligram. Hendriks's paint chip started at 160 micrograms; after cleaning, 58 micrograms remained; 19 micrograms of carbon reached the AMS.

Turnaround is days to months depending on the lab's queue.

Calibration, time lags and the sentence you can print

Raw pMC is not a year. The lab calibrates with OxCal (Christopher Bronk Ramsey, Oxford) or equivalent software, using IntCal or a post-bomb curve. The output is a probability distribution, often with two peaks on the bomb curve. The scientist then applies art-historical constraints: the artist died in 1955, so a 1959 harvest date for the cotton is decisive; a 1983–89 oil date on a picture claiming 1866 is decisive in the other direction.

A further lag sits between harvest and the finished picture. Fiona Brock, Nicholas Eastaugh, Tom Ford and Joyce Townsend dated canvas from eighteen mid-twentieth-century Scandinavian paintings (Radiocarbon, 2019). Most post-bomb samples implied two to five years between plant harvest and completion; some implied up to ten. Beck's group used that window to place their Impressionist fake after 2000, and more likely 2002–08, once the 2000–2003 harvest interval was the probable match. The printed sentence should name the material ("the flax in this canvas was harvested in …") and the lag, not "the painting was made on 4 June 2003."

Pigment Terminus Post Quem

A pigment with a known first date of manufacture gives a terminus post quem for any original layer that contains it. The chemist identifies the colourant (Raman spectroscopy, scanning electron microscopy with energy-dispersive X-ray spectroscopy, X-ray diffraction, X-ray fluorescence). The historian supplies the industrial date. Presence of a modern pigment in a layer the conservator judges original is a hard stop. Absence of that pigment proves nothing: an artist in 1925 could still squeeze lead white.

National Gallery research by Jo Kirby and David Saunders treats Prussian blue as the first modern synthetic pigment, prepared around 1704 by the Berlin colourmaker Diesbach (an accidental product while he aimed at a cochineal lake), announced in 1710, and found in pictures from a few years later. A "Rembrandt" whose original blue is Prussian blue is not a Rembrandt. Cobalt blue (Thénard, 1802) and artificial ultramarine (Jean-Baptiste Guimet's synthesis in 1826, prize in 1828) are the next well-dated blues. Titanium white (titanium dioxide) entered commercial production around 1916 and reached artists' oil paints in the years that followed. Nicholas Eastaugh, lecturing for Icon, described how titanium white in a painted sketch on the reverse of a purported Heinrich Campendonk of 1914 broke Wolfgang Beltracchi's Rotes Bild mit Pferden. The pigment was not in use at the claimed date. That is terminus post quem as a forensic sentence, not a biography of the forger; the detection narrative sits in the sibling article on how forgeries are detected (pending).

Lead white can now be dated by radiocarbon as well as identified. Historical stack-process recipes used vinegar and fermenting organic matter to corrode metallic lead to hydrocerussite and cerussite. Organic carbon dioxide is locked into the carbonate. Lucile Beck and colleagues have shown that this 14C can date the manufacture of the pigment, and that heating protocols can separate that carbon from geological chalk (Meudon white) which contains no 14C. The method is young, sample-hungry by paintings standards, and sensitive to later restoration whites.

Dendrochronology for Panel Paintings

A wooden panel is a slice of a tree. Each year the tree adds a ring whose width records climate. Dendrochronologists measure those widths along the end grain, build a curve, and match it to a master chronology for the species and region. Oak from the eastern Baltic, shipped to the Low Countries, is the classic case for Netherlandish panels. Beech from German forests is the classic case for Cranach. The match dates the rings that are present. It does not date the day a workshop planed the board, and it does not date the paint. Peter Klein, working from the University of Hamburg, measured thousands of painted panels; the RKD in The Hague and the National Gallery of Denmark (SMK) have been entering that archive into Dendro4Art. The National Gallery of Victoria's published note states the limit: the method yields a terminus post quem, the earliest date at which the painting could have been executed, by establishing a plausible felling date for the tree.

Measuring rings and matching a chronology

The scientist needs a readable sequence: enough rings, a species with a regional master curve, and an edge that still shows the outer wood. Oak and beech in northern Europe are the cooperative cases. Tropical hardwoods, walnut, and poplar often lack the published chronologies or the ring clarity that oak provides. Measurement used to mean a travelling microscope on the unframed panel. Current work also uses high-resolution imaging of the end grain. The curve is then cross-dated against masters (Baltic oak, German oak, German beech, and others). A match is a statistical claim with a date for the outermost measured ring.

Klein's 21 January 1994 report on Cranach the Elder's Madonna with Child at the National Gallery of Art, Washington (inv. 1953.3.1), is a compact example. Two beech boards: 48 rings dated 1406–1453, and 32 rings dated 1483–1513. The youngest ring formed in 1513. For beech panels, Klein noted, makers as a rule removed the bark and used the tree to the outside, so 1513 is an earliest felling date. With a minimum of two years' storage, he judged creation plausible from 1515 upward. That is a dated board, a seasoning estimate, and a cautious sentence. It is not "Cranach painted this in 1515."

Sapwood, seasoning and the terminus post quem

Oak panels are harder. Heartwood is what you see on many trimmed edges. Sapwood, the living outer wood, was often cut away because it is vulnerable to insects. If sapwood is missing, the felling date is after the last heartwood ring plus a statistical minimum of sapwood rings for that region. A 2025 replication study on two Rembrandt portraits (Humanities and Social Sciences Communications) restated Klein's 1990s numbers with current masters. For the panel that lacks sapwood, Polish oak statistics give a minimum of nine sapwood rings, so a last heartwood ring in 1602 becomes a felling date after 1611, and with two years' seasoning an earliest production after 1613. The study also shows why replication matters: new reference datasets can shift the "more likely" window even when the outermost ring date holds.

Seasoning is a workshop guess informed by period practice, not a stopwatch. Klein's two-year minimum for beech and oak is a floor. A board can sit longer. A painter can use a panel cut from furniture or from an older picture. The printed result should keep the words "earliest possible" in view.

Limits of a felling date

A felling date after 1613 does not prove Rembrandt. It proves the tree was still growing until at least the last measured ring, and that the panel is unlikely to predate the seasoning estimate. A workshop assistant, a later copyist, or a forger with an old board can all paint after that year.

Dendrochronology has grouped panels from the same tree, which is strong evidence that two pictures passed through the same supply, sometimes the same shop. That is still not a signature. An old panel with a convincing ring sequence is a purchased alibi, the wooden equivalent of Trotter's old canvas. Date the paint or the binder as the counter.

How These Methods Developed

Libby's counting laboratories in the late 1940s could not date a painting without destroying it. Willard Libby proposed the method in 1946 at the University of Chicago; in 1949 he and James Arnold published the "Curve of Knowns" in Science; the Nobel Prize in Chemistry followed in 1960. Art applications waited for AMS. Oxford decided in 1980 to build a dedicated accelerator facility with the university's nuclear physics department, hoping for trial dates by summer 1981. ORAU became one of the first AMS laboratories in the world. Robert Hedges ran it in the early years. Christopher Bronk Ramsey later directed it and wrote OxCal, the calibration package most art papers now cite.

KIK-IRPA in Brussels began radiocarbon work in the 1960s with a proportional gas counter, sent prepared samples abroad for AMS from 1989, and installed its own MICADAS in 2013, the only AMS for carbon-14 in Belgium. Mathieu Boudin heads the lab. Private individuals, the institute says, call in particular for the dating of artworks.

The last generation added Raman and SEM-EDS on microscopic samples, and published industrial chronologies that make titanium white in 1914 a one-line disproof. Dendrochronology of panels matured in the 1970s–90s in Hamburg, Berlin and Copenhagen (Bauch, Eckstein, Klein, Wadum). Lead-isotope work on lead white goes back to Keisch and Callahan in 1976; Fortunato, Ritter and Fabian's 2005 Analyst paper on Rubens and van Dyck is the high-precision cluster study later papers still cite.

Labs and Published Studies

A date stands or falls on pretreatment, on whether the sample is the layer you think it is, and on the sentence written on the report. The groups below are the ones the research literature keeps returning to for paintings. Commercial pages that advertise carbon dating without naming an AMS facility, a pretreatment protocol, or a calibration curve are selling a number, not a method. Museum conservation departments, listed in the museums directory, commission these tests through in-house scientists or through the labs named here. ETH Zurich's Laboratory of Ion Beam Physics (MICADAS, Hendriks and Hajdas), INFN-Labec in Florence (the Léger canvas), and LMC14 at Université Paris-Saclay (Beck's police case) sit in the same published circuit. You go to them with a question a conservator has already framed, not with a jpeg and a hope.

Oxford Radiocarbon Accelerator Unit

ORAU sits in Oxford's School of Archaeology, descended from the Research Laboratory for Archaeology and the History of Art (founded 1955). It offers project design, sampling of delicate objects, contaminant removal, AMS measurement, and OxCal calibration. Commercial work is listed at £500 per sample excluding VAT, including sampling at Oxford, preparation, stable carbon isotopes and AMS; failed samples through no fault of the lab are £100. Academic archaeology samples are cheaper in groups (£380 for one to four). Those figures are the instrument line, not a paintings-authentication fee. Conservator time, pigment identification and a written interpretation sit on top. ORAU reserves the right to refuse material of uncertain provenance.

Brock's 2019 canvas study is the ORAU-linked paper that paintings specialists cite for the harvest-to-picture lag and for the warning to screen canvas for synthetic fibres before dating. A polyester thread in a "linen" sample will not behave as plant carbon.

KIK-IRPA, Brussels

The Royal Institute for Cultural Heritage combines a painting laboratory, a radiocarbon lab, and dendrochronology. MICADAS, in house since 2013, measures small samples in hours once they are prepared. Two online databases hold sixty years of dates.

For a painting, the useful neighbour is the painting lab: original layer versus later addition, then a fibre date if the question needs one. The institute invites a quote by email, with photographs, and will visit on site for complex studies. That is how a private owner reaches a heritage AMS without inventing a chain of custody.

Van Gogh palettes and forgery cases at method level

Van Gogh's materials show both uses of a date: ruling a claim out, and opening a period window without finishing the attribution. At Otto Wacker's 1932 trial in Berlin, the Dutch restorer A. M. de Wild reported that pigments in the disputed pictures did not match those van Gogh used. That was identification against a documented palette. Experts at the trial still disagreed on which pictures were genuine; pigment evidence sat beside connoisseurship. X-rays shown in court belong with the imaging guide, not here.

In 2013 the Van Gogh Museum reversed a 1991 rejection of Sunset at Montmajour. Teio Meedendorp, writing in The Burlington Magazine, noted that almost all the pigments were ones van Gogh had on the palette in 1888, including cobalt blue from the summer of 1887. The pigments made an 1888 date possible. Letters, the Montmajour site, and canvas comparison with The Rocks in Houston carried the attribution. Pigment dating did not, by itself, name van Gogh.

The AMS cases already walked through above follow the same rule. They name a material and a calendar. They do not name a forger. The pending how forgeries are detected article is the place for the human plot.

Failure Modes: Old Wood, Reused Canvas, Contamination

The old-wood problem is the oldest warning in radiocarbon. Wood from the heart of a long-lived tree can be decades or centuries older than the year the tree was cut. A panel, a stretcher, a frame, and a carved element can all be reused. Picasso and van Gogh reused canvases; Beck's paper cites that studio habit as a reason a "too old" support does not prove a fake. A forger who understands the test buys the old support on purpose. Dating the support then dates the victim, not the crime.

Contamination moves the date in both directions. A petroleum-based varnish or consolidant is 14C-dead and makes a sample look older. A modern glue, a lining, or bomb-peak retouch makes it look younger. Beck's Impressionist white, taken at the edge, was zinc white in oil: in principle a clean binder date. Dust and possible varnish drips still sat on the surface. The authors called the binder date inconclusive and trusted the canvas.

Synthetic fibres in canvas, incomplete ABA pretreatment, and chalk (geological carbonate) in grounds are the other regular traps. Brock's group flagged pre-screening for synthetics. Lead-white radiocarbon protocols heat at a chosen temperature (Beck's thermogravimetric work points to about 400 °C in mixed lead white / Meudon white / linseed oil systems) so that organic and lead-carbonate carbon come off while geological calcite stays. Get the temperature wrong and you date the chalk.

The calibration plateau from the late 1600s to 1950 is a failure mode of interpretation. Many calendar ranges overlap. A stretcher that calibrates to 1697–1724 or 1812–36 or 1880–1911 has not dated an Impressionist picture. It has failed to exclude one. Bomb-peak samples escape that plateau, which is why they dominate forensic papers.

Lead Isotopes and Other Brief Methods

Lead white carries the isotopic signature of the ore from which the metal was smelted. Fortunato, Ritter and Fabian (EMPA, 2005) measured high-precision lead-isotope ratios on samples from Rubens, van Dyck and other northern pictures, and compared them with Italian pictures of the same centuries. The Flemish group formed a tight cluster; the Italian group sat elsewhere. The authors inferred distinct ore sources and little mixing of European leads in the pigment trade they sampled.

Later work has narrowed what that cluster can mean. Dutch seventeenth-century lead white was in large part made from English lead, Derbyshire in the account of D'Imporzano and Davies's Vermeer study (2025). Vermeer's securely attributed pictures share that English-lead signature with many contemporaries. A match to Vermeer's lead white therefore matches a national supply, not a studio. The same authors argue that isotope data cannot support or refute the attribution of Saint Praxedis: too many painters used the same commercial white. Heterogeneity inside a single picture (different whites in ground, paint and retouch) can flag a later addition. It still does not name the painter.

Costs, Access and When Dating Is Worth Doing

ORAU's £500 commercial AMS fee (excluding VAT) is a floor for one prepared sample, not a quote for "date my painting." KIK-IRPA asks for photographs and a description, then prices the study; complex objects get a site visit. ETH Zurich, INFN-Labec, LMC14 and other AMS centres work through research collaborations or police instructions more often than through a public web form. Pigment identification in a museum laboratory is staff time. In a private conservation studio it is an analysis invoice: Raman, SEM-EDS and a report can exceed the AMS fee before anyone burns a fibre.

A serious campaign therefore runs to thousands of pounds or euros once you add examination, imaging (the pending infrared and X-ray article), sampling by a paintings conservator, two or three AMS targets (canvas, binder, perhaps stretcher), and interpretation. Museum objects absorb that cost as research. Private owners should expect to pay for a written question ("Can this canvas predate 1955?") rather than for a certificate of authorship. No reputable lab sells a certificate of authorship. ORAU's refusal right on uncertain provenance is the ethical version of that limit.

Access tracks the question. If you need to know whether a family landscape is a print or an oil, you do not start with MICADAS. If a catalogue raisonné committee, a museum acquisition, or a criminal investigation has a specific anachronism to test, dating is in play. The conservator who already knows the layer structure should choose the sample. Chain of custody matters if the result may reach court; Hajdas's 2022 intercomparison exists because lawyers ask whether Lab A and Lab B would have agreed.

Studio restoration, lining and varnish history belong with how paintings are restored. Those treatments are also the contamination sources this page has already named. Date first, or date from a location the restorer has not flooded with new carbon.

FAQ

Search results for "dating a painting" mix living artists asking whether to write a year on the front, collectors hoping a signature will reveal a fortune, and a thin scientific literature behind paywalls. Google's related questions add "how can I tell the age" and, off this fence, "is it worth anything." The answers below stay with laboratory methods: what you sample, what carbon-14 can say about a hundred-year-old canvas, why a date is not an attribution, the old-wood problem, cost, and which parts of a picture have no 14C clock. Value, selling, and identifying an artist from a photo are different jobs. Forgery as a narrative is the pending detection article. Imaging is the pending infrared and X-ray article.

How do scientists date a painting?

They date a material. Radiocarbon on canvas, paper, wood or oil gives a calendar window for when that organic carbon left the atmosphere. Pigment identification gives a first-use date for a colourant in an original layer. Dendrochronology gives a terminus post quem for a panel from the last growth ring plus sapwood and seasoning estimates. The report should name the material, the method, the calibration, and the lag between harvest or felling and the finished picture.

Can you carbon-date a painting that is only a hundred years old?

Yes. The bomb peak makes the last seventy years more precise than the century before. Pre-1950 samples from about 1700 to 1950 sit on a calibration plateau with wide ranges. Post-1950 plant carbon carries excess 14C from nuclear tests. A 1920 linen without bomb carbon will not give a one-year answer.

Does a radiocarbon date prove who painted the work?

No. It proves when the sampled carbon was alive, within calibration error and after pretreatment. An old canvas can carry a new picture. A new canvas cannot carry a picture by an artist who died before the plant was harvested, unless the cloth is a lining or a later restretch. Authorship still needs documents, style, and how the paint was handled.

What is the old-wood problem?

The laboratory dates the growth of the wood, not the painting of the panel. Heartwood can be a century older than the bark. Stretchers are replaced. Forgers buy old boards. The same logic applies to reused canvas. Pair the support date with a binder date, or treat the support date as a terminus post quem only.

How much does it cost to radiocarbon-date a painting?

ORAU lists £500 excluding VAT for a commercial sample, and £380 for a single academic archaeology sample. That is the AMS line. Conservator sampling, pigment analysis, extra targets, and a written interpretation are separate. A full technical study sits in the thousands. Anyone offering a cheap "carbon date certificate" of authorship is not describing these laboratories.

How can I tell the age of a painting without taking a sample?

You can read labels, stretchers, nails, and the history of the frame. Craquelure and varnish are ageing, not a calendar. Non-invasive pigment screening (XRF, Raman) can sometimes flag titanium white or Prussian blue without a chip. A firm terminus post quem or a bomb-peak window still needs a laboratory and, for 14C, carbon. If the object cannot be sampled, the honest report stops at "materials visible so far are compatible with …"