Synchrotron Radiation-Based XRF Reveals the Secrets of Famous Paintings

Koen Janssens, professor of chemistry at the University of Antwerp (Belgium), uses synchrotron radiation–based X-ray fluorescence (XRF) to analyze historic works of art. In a recent study, he used various spectroscopic techniques, including several X-ray methods, to reveal the process by which the yellow paint in some of Vincent Van Gogh’s paintings darkened over time. Here, Janssens explains his work, including how synchrotrons enabled him to do more detailed analyses with XRF than is possible with laboratory instruments.

Spectroscopy: In a recent study, you used several X-ray methods to reveal the process by which the yellow paint in some of Van Gogh’s paintings darkened over time. Can you explain briefly what you found?

Janssens: We were intrigued by the fact that in some quite famous works of Vincent Van Gogh, the yellow he uses gradually becomes darker. The most famous cases in point are the sunflowers. There are several sunflower paintings made by Van Gogh, and some of them have this tendency that the originally fairly bright yellow becomes darker and darker, becomes a kind of brown. And so we tackled this problem, which was known about even while Van Gogh was still living, by a combination of X-ray–based methods, X-ray fluorescence (XRF) analysis, and X-ray absorption near edge spectroscopy (XANES).

XRF allows you to get information on which chemical elements are present in a material, while the other method, XANES, goes one step further. It allows you to determine, for specific elements you’re interested in—in our case, chromium—in which chemical or valence states the chromium is in.

The yellow pigment we investigated and which suffers from this tendency to get more brown, more dark, is called chrome yellow or lead chromate. The chromium in there is in the chromate form, which means it’s present as chromium(VI)+, the most oxidized form of chromium, and that gives it a yellow color. While what we found, essentially, by doing this combination of XRF and XANES, is that the browning or darkening of the yellow is caused by a reduction of the chromium. So we found, in a number of the painting samples we examined, that the chromium is partially reduced, so that part of the chromium — let’s say two thirds of the upper parts of the yellow paint — has transformed from a chromium(VI) into a chromium(III) compound. That’s the essence of our discovery.

And in addition to being able to see that in real samples from paintings of Van Gogh, we also could reproduce this process in the lab, in circumstances which are very similar to what the paintings were going through. And because we can reproduce this, we can turn more knobs and change more parameters than what you normally can do when you deal with real paintings or real works of art.

We also have a pretty good idea now why some of the paintings of Van Gogh which contained the chrome yellow show this behavior and why some others do not. This is also something which has puzzled a lot of people. There are paintings by Vincent Van Gogh and by some of his contemporaries who also use a lot of yellow where there is actually no change visible, while with others there is this quite serious problem.

It turns out that what is needed is an extra component which induces the browning. Not only do you need lead chromate, which is a yellow powder, there is also another component required; in the case of Van Gogh, this is frequently a white powder or pigment called lead sulfate. So in fact it is Van Gogh’s intention in a number of paintings to create a light yellow effect that more or less is the origin of the problem. If you use the orange-yellow lead chromate powder on its own, then there is not really a problem after a number of years. The warm yellow color remains the same, and in a number of paintings of Van Gogh, and also by Gauguin and Cezanne, the original paint is as it is. It has retained its original yellow-orange color. It’s only in those paintings by Van Gogh where he was trying to paint a lot of sunshine, a lot of light, yellow textures on ripe corn in the fields, where he really wanted to create a light yellow color, that those lighter yellows are now becoming darker. And that’s because there is this excess sulfate in the paint. That is the key ingredient that causes a reduction of the chromate.

Spectroscopy: As you mentioned, you used several different X-ray techniques in this study. Can you explain the role of each in reaching your conclusions?

Janssens: X-ray fluorescence analysis is a fairly well established method of elemental analysis. But normal tabletop machines today are made for analyzing fairly large samples, so they give you information on the elemental contents and this is averaged out over a square centimeter or several square centimeters.

Here, of course, we are looking at paint samples—very small bits of paint which were taken from the paintings. Usually this is done as a part of a larger restoration effort where there is already damage to the painting from the past, and little bits of paint may come off or can be sampled there. These little bits are about a cubic millimeter in size, or a fraction of that. And usually these paint samples consist of a series of layers. Each layer is typically only 20 or 30 μm in thickness. And therefore you need a method that allows you to analyze each layer separately.

So we used microscopic X-ray fluorescence for that. If we use beams of 1 μm in size, that allows us to make scans over these samples which are cross sections. And that allows us to see which combinations of elements are present. We record these elemental maps to figure out the composition of each layer and to see which elements are in which locations.

Then we record the XANES spectra on specific locations. So, with X-ray fluorescence you get information on all kinds of different elements. With XANES, what you do is you focus on one specific element, such as chromium or sulfur, in this case.

Then you vary the energy of the X-rays with which you irradiate the sample. Inside chromium you have the electrons which are the closest to the nucleus, which are in the 1s orbital (or in the K shell for atomic physicists). If you give these electrons a certain amount of energy, which is their binding energy, you can eject these electrons completely from the atom.

There is a very small difference in the energy of these 1s or K shell electrons depending on whether the chromium atom is in its metallic state, when it has all its electrons, in the chromium(III)+ state, or in the chromium(VI)+ state. For chromium(VI)+, where there are six electrons missing, all of the other electrons that are still in the atom are more tightly bound to the nucleus than when, for instance, the chromium is metallic and has all its electrons.

By determining more or less the energy required to remove these K shell or 1s orbital electrons from the atom, we can derive information on its oxidation state.In the locations which have turned a little bit brown or very brown on the surface, we see very thin layers in which the chromium has been reduced.

So the yellow paint, which was originally homogenously yellow, became covered by an extremely thin 1- to 2-μm layer of brown material. And this brown material is caused by the presence of chromium(III)+. In the world of pigments and paints, everyone is very aware that you can make yellow with chromate, with chromium(VI), and that with chromium(III)+ it’s possible to make a green pigment called viridian, which is chromium oxide. It’s Cr2O3 with some hydration water attached.

So this is what is really happening: Inside the yellow part of the chromium, atoms are reducing, and instead of being yellow, they become green. And it’s this combination of yellow and green, together with some oxidation of the oil in the oil paint that causes this chocolate brown color which is present at the surface.

Spectroscopy: In some of your recent work, μXRF was also used to map the distribution of the different chromium species, but I thought XRF was not capable of distinguishing species. Can you explain a bit more about how the μXRF was involved in that mapping?

Janssens: Well, we use a special version of XRF. As I was saying, in chromium(VI) and in chromium(III) there is a difference in the extent to which the 1s electrons are bound to the nucleus of the atom. We can capitalize on that small difference, which is just a few electron volts difference, by carefully choosing the energy we use to record the XRF maps. So if you have chromium in the VI+ state, rather than do what we normally do with XRF, which is to give enough energy to eject the K-shell or 1s orbital electrons from the atom, what we do is we give them just enough energy to promote the 1s electrons to the 3d orbitals of chromium.

In chromium(VI)+, the 3d orbitals are all empty. So for chromium(VI)+ the transfer of an electron from the 1s to the 3d orbital works fairly efficiently, but it happens at only one energy. So you could call this a resonance energy, for chromium(VI)+.

If you shoot at material which contains a mixture of chromium(III) and chromium(VI) and use a particular energy, only the chromium(VI) will absorb the radiation and will emit chromium fluorescence X-rays. The chromium(III)+ will not do that. So we are able to make chromium(VI)-specific maps in this way, by carefully selecting the energy with which we irradiate materials. The same to some extent also is possible for chromium(III). If we increase the energy a little bit, we come into a regime where both the chromium(VI) and the chromium(III) will become ionized and will emit X-ray fluorescence. And from that map, we can then subtract the chromium(VI) contribution. And in this way it’s possible to obtain what we call chemical state maps, distributions of where chromium(III) and chromium(VI) are present. In these maps, you then clearly see the difference between the surface of the yellow paint and what is below—what is still yellow.

Spectroscopy: Some of these X-ray analyses were carried out at synchrotrons in Grenoble, France, and Hamburg, Germany. Why was a synchrotron necessary for this work?

Janssens: When you want to do very laterally resolved analysis, you need very small beams to excite the material only at a specific location on the sample. Because these layers are so small, and because the alteration, the conversion from yellow to brown, only happens in a very thin layer on the surface, we needed a very small X-ray beam. X-ray beams produced in tabletop devices, using X-ray tubes, can be focused down with various kinds of optics, but these optics generally produce beams which are about 10 to 20 μm in size. And this is too large for our purpose. In this case, we needed a beam which was around 1 μm or smaller. And to obtain those beams you need a synchrotron. A synchrotron is a more powerful X-ray source and makes it possible to focus X-ray beams down to the μm- or sub-μm range.

Another reason for using a synchrotron is that to do the XANES measurement and to be able to choose the primary energies very carefully, to get the resonance effects, we need highly monochromatic primary X-rays, and again, this is something you normally can only do at a synchrotron facility. Most instruments at a synchrotron are equipped with monochromators, such as double crystal monochromators, and those are able to select a very narrow energy band from what is produced in the synchrotron.

It’s not only the very narrow width of the selected energy band that is useful. It’s also possible to change the settings of the monochromator so that you can, in a continuous way, move or “tune” this energy. And that’s another requirement perform XANES measurements.

Spectroscopy: Are there other new types of studies that synchrotron X-ray analysis enables you to do that you could not do without it?

Janssens: Another thing we are doing with X-rays is to see what other changes to paintings, including those done by the original artist.

In many paintings, what you see now is what the artist intended, obviously, but you could say it’s what the artist intended in the last phase of finishing the painting. In some cases, paintings are made very quickly. The artist has an idea, executes it, and the work of art is finished. But in a number of cases, it’s not so straightforward. For instance, an artist can change his mind several times about what is shown in the picture. Either details are changed—you can imagine somebody who wears a hat and waves at another person in the original idea of the artist, but after a while the person doesn’t wear a hat anymore and doesn’t wave or moves his hands in another way. These kinds of changes, which are part of the creative process of a painter, are more or less recorded in a painting. They are recorded not at the surface, but they are recorded in the form of buried paint layers.

So an artist paints something, then changes his mind, so he paints over it. These thin layers, which may be 20, 30, 100 μm thick, are no longer visible at the surface. But if you have methods to show the chemical elements which are part of these buried paint layers you can reconstruct, more or less, earlier versions of the painting.

Of course it’s also possible that it’s not really the original artist that made the changes. It can be that later on another artist, or the same artist, recycled a canvas or a panel, and completely covered the original painting with a layer of white paint and reused the new surface for a completely different painting. When the artist that came second is the most famous one, and what he used as a substrate was done by an unknown artist, nobody really pays attention. But if the reverse is the case, if, let’s say a painting from an artist who is now famous was covered up by a nobody, or by himself in a latter period of his life, art historians are very interested to uncover these original layers.

By using scanning X-ray fluorescence analysis with high-energy primary beams, beams that go very deeply into or through these multiple layers of paint, and by scanning this beam over large areas, it’s possible to build up large-scale images that in a number of cases show you how the original paintings, which were later covered up, looked. We have been able to do that with paintings by Van Gogh, Rembrandt, and Memling.

We have also done this with a painting by the 18th century Spanish painter Goya. Goya lived in a period when Spain was occupied by the Emperor Napoleon. Napoleon put his brother, Joseph Bonaparte, on the throne of Spain. Joseph Bonaparte had a court in Madrid, and commanded a number of paintings from Goya, who also lived in Madrid at that time.

Nevertheless, the reign of Napoleon in France and of his brother in Spain was short lived. So Goya later ended up with a number of paintings of officers of Joseph Bonaparte which he could not sell anymore. So one of these paintings he covered up and painted something else on top of it — another portrait, which he could sell.

Well, by using this scanning XRF technique, we were able to more or less reconstruct the original painting. The painting as you see it now, and which is now in the Rijksmuseum in Amsterdam, shows a civilian dressed in black, a fairly ordinary man. What we were able to uncover below is an officer, a classical 18th century high-level military officer wearing a uniform with a lot of gold brocade, some medals, and epaulettes. So a quite different, much more impressive portrait is below the layer that you see now.

So with this technique, microscopic XRF, it’s possible to have quite a different outlook on paintings. Normally to inspect paintings, people use X-ray radiography, where you just shine with a broad beam of X-rays through the painting, like with medical radiography, and record what has been absorbed. Although that technique is useful, it doesn’t show you all the information. With our technique we get information which is separate for the different elements, and then we can combine all these large-scale elemental maps. In the case of this Goya portrait, for instance, we could combine the lead image with the antimony image and the mercury image. You can then create a kind of color radiography. And that usually is more informative than the traditional black-and-white radiographs which people have been using since Roentgen invented X-rays at the end of the 19th century.

Spectroscopy: Was a synchrotron necessary for that work?

Janssens: Well, the first time we did this kind of investigation in 2008, we did it at a synchrotron.

The only reason why a synchrotron was necessary was to speed up the analysis. Because the intensity you can use at the synchrotron is very high, you can scan more quickly. If you use X-rays from an X-ray tube, it’s still possible to do the same thing, but you have to work a bit slower.

So in the meantime, we have constructed a kind of mobile microscopic X-ray scanner. The big advantage of using an instrument that uses X-rays generated in X-ray tube is that you can move the instrument to the painting, rather than moving the painting to the synchrotron. It’s not impossible to convince people in a museum to allow their painting to be moved to a synchrotron, but it’s not really possible for all paintings. Very large paintings are very difficult to move, and, in general, all paintings, especially the famous ones, are very expensive to insure. And when you do these experiments at the synchrotron, you need to guard the painting 24 hours a day.

So there are all kinds of reasons to prefer to do these kinds of analyses inside a museum. The security is already in place, the painting usually can stay in the gallery. We have already analyzed paintings while they were being seen by the public, so that the people that visit the gallery are not disappointed because their favorite painting is not there for a week because it’s at the synchrotron.

And it seems to work quite well. We have compared our analyses of a number of paintings that we analyzed both at the synchrotron and using this microscopic X-ray scanner. And even though the latter takes longer, more or less the same results come out. That’s very promising. All kinds of museums are starting to show an interest because they realize that now you can get a new view of their paintings. This is really starting to generate a lot of interest from various museums, big and small, around the world.

Spectroscopy: Is the instrument that you use for those in situ analyses a commercial instrument or something that you developed in the lab?

Janssens: It’s something we built here in my department. My research group has been building their own equipment for many years. In this case, the entire instrument is being moved, in small steps, in front of a painting, without disturbing the painting. This is very comfortable for museum people. Then we can just let the instrument do its job. And it’s both nondestructive and noninvasive.

Editor’s note: This interview has been edited for length and clarity.