Hidden for centuries, Rembrandt’s secrets are finally being revealed

Rembrandt van Rijn’s The Night Watch is one of the world’s best-known paintings. Dating from 1642, the strikingly large canvas – 4.6m across and 3.8m high – depicts a group of civic militia guards led by Captain Frans Banninck Cocq, the central figure with a red sash and lacy collar (the man to his left, resplendent in cream doublet and feathered hat, is his lieutenant, Willem van Ruytenburch).

The scene, which contains 31 figures in total, is unusually chaotic. Rather than solemnly lining up his subjects for a portrait, as was customary at the time, Rembrandt captures them in full, messy action. Guardsmen brandish muskets, pikes and swords; a dog barks at a drummer mid-beat; a young girl peers through the crowd, a chicken hanging from her waistband. Some details glint in Rembrandt’s trademark dramatic lighting while others are cast in shadow, forcing your eyes to dance around the image, picking out new features with each glance.

For more than 200 years, The Night Watch has been in Amsterdam’s Rijksmuseum, where it hangs at the end of the Gallery of Honour, the altarpiece in a cathedral to Dutch painting. On a Wednesday in August 2019, it looks somewhat different to usual. It has been removed from its frame to reveal a few inches of grubby, naked canvas around the edges, and only half of the painting is visible. The rest is obscured by a pair of scissor lifts, on top of which sits a large machine consisting of a grey box with a metal frame positioned parallel to the painting. This is a macro X-ray fluorescence (macro XRF) scanner, an imaging tool used to detect the presence of different chemical elements in an object. The painting and equipment are all encased in a large glass box.

The Rijksmuseum is currently undertaking an unprecedented research and conservation project on The Night Watch, which sees it using new imaging tools and data analysis techniques to investigate the painting down to a molecular level. Dubbed Operation Night Watch, the plan combines the expertise of chemists, conservators, computer scientists and historians in an attempt to gain new insight into how Rembrandt approached his masterpiece, what it would have looked like when he first painted it, and how it has changed in the centuries since. The methods developed during the process consider the work from a scientific perspective as much as an artistic one, and could help transform the way we study and restore art.

Vincent Fournier

At 8:30am, half an hour before the Rijksmuseum opens to the public, paintings conservators Susan Smelt and Anna Krekeler stand on top of the scissor lift in front of The Night Watch, their feet level with Captain Banninck Cocq’s face. They position the macro XRF scanner in front of a section of the painting about 78cm wide – a patch of muddy-looking background – and carefully manoeuvre the scanner head so that it is exactly one centimetre from the painting’s surface.

“A painting is not like a flat leaf or a flat board; it has a surface, so we need to be sure that we are parallel to the painting,” Smelt says. She and Krekeler work methodically, measuring the distance between scanner head and painting at regular intervals with a piece of soft foam. Shortly after 9am, the first tourists trickle in to see its star exhibit.

It takes until 9:45 for Smelt and Krekeler to complete the setup. As the scanner head starts its slow movement from left to right, orange and red lights appear on the machine to indicate that X-rays are being emitted. Smelt and Krekeler descend on the scissor lift with a clunk and exit the glass box – given the painting’s renown, the museum has made the decision to conduct every stage of Operation Night Watch in sight of its visitors. On the glass door hangs a yellow sign: “Warning: X-ray radiation”.

Macro XRF scanning works differently from the kind of X-ray you might get in hospital (although this has also been applied to The Night Watch). Whereas a regular X-ray image simply shows up areas where X-rays have been absorbed before reaching a detector on the other side (be it by bone or by pigments containing heavy metals), macro XRF scans give much more detailed information about the materials the X-ray beam encounters.

As the scanner head moves across the painting’s surface, the X-rays are scattered back to a detector. By measuring the energy of these returning X-rays, known as “fluorescent” X-rays, the Operation Night Watch team can tell which chemical elements are present at any point, and infer the pigment used. The presence of lead, for example, could indicate lead white or lead-tin-yellow pigment, depending on which other elements are found alongside. One pigment used a lot by Rembrandt is smalt, a type of ground blue glass. “There is cobalt in the blue glass, but also nickel and arsenic from the glass, so if you see these three elements together you know you have smalt present,” Smelt says.

The scanner produces “maps” of the painting for every element studied – all 33 of them. These are recognisable as black-and-white images of The Night Watch, but each one looks completely different, with the element in question shining bright. Some look almost like photographic negatives. In the calcium map, details in the dark background pop out much more clearly (calcium is found in a pigment known as bone black, made from charred animal bones). In the cobalt map, the boot of the figure to the far left of the painting, usually murky brown, shines bright white.

Vincent Fournier

While it may seem relatively obvious which parts of the painting contain white or blue paint, the macro XRF scans reveal more than can be seen by the eye. By showing exactly which pigments are used where, they give insight into Rembrandt’s approach. We can see, for instance, that he used smalt not only to create a blue colour but also in other mixtures, such as the brown of the boot – perhaps to enrich the colour, but also because smalt affects the texture of paint, making it dry quicker. Katrien Keune, a chemist and the Rijksmuseum’s head of science, says Rembrandt’s inventiveness in combining pigments is part of his artistic legacy. He is particularly known for his impasto – the use of thick paint to create a 3D effect. “His hunger for creating certain effects, I think that’s what makes him special, and it’s what we try to understand better,” she says.

X-rays, of course, also penetrate beyond the surface of the painting and so can reveal information from below the top layer. The Rijksmuseum researchers hope their scans may provide clues as to how Rembrandt composed the Night Watch scene. Did he change his mind during the commission, perhaps repositioning the figures part-way through, or adding extra elements at the last moment? Smelt says they don’t expect to find anything extreme. “Rembrandt didn’t first use this painting to start something else and then scrape it off,” she says. Rather, they are looking for small changes that may shed some light on the artist’s creative process.

Studying the painting on a molecular level will also help researchers understand how it has changed and degraded over time, and assess the impact of previous conservation attempts. One of the triggers for the current project is a whitish haze that has settled around the dog in the bottom right corner, and that Keune says was certainly not intended by Rembrandt. The scans will help to determine why this discoloration has occurred.

The Rijksmuseum’s decision to work on the painting now, however, is as much down to technological advances as it is to any urgent need for repair. The last major restoration was in the 1970s; researchers today have imaging tools at their disposal that may not have been available even five years ago, and modern computers are able to process much more data.

Today’s scan will take 21 hours to complete and is one of a total of 56, covering the entire canvas. After imaging a few additional details, the macro XRF scanner will be replaced with a high-resolution camera system, which will be suspended in front of the painting from a metal beam and will take up to 12,500 photographs. The researchers will also use other imaging techniques, such as imaging spectroscopy, which takes images at hundreds of different wavelengths; optical coherence tomography (OCT), which is useful for imaging the translucent layers of varnish; and structured-light 3D scanning, which measures the topography of the painting’s surface, like a contour map.

Only after they have analysed all this data will the project move on to the conservation phase – a “full body scan” approach that Keune thinks should become standard practice. “It’s very important,” she says. “You’re not going to operate if you don’t know where the cancer is.”

Part of the iron “map” of The Night Watch

Vincent Fournier

Across the road from the main entrance of the Rijksmuseum, with its postcard-friendly red brickwork and steep grey roofs, is a building known as the Ateliergebouw. It is here that researchers and conservators work to unravel the secrets of historical paintings and art objects, and to restore them to their full glory. Robert Erdmann, a senior scientist at the Rijksmuseum and a professor at the University of Amsterdam, has an office on the second floor. The only paintings here are in the screensavers on his computer monitors – four of them, positioned so that he can easily twist his chair to look at each one. His workspace looks more like a set from CSI than something you’d expect to find in an art museum.

Erdmann is responsible for processing and visualising the imaging data collected in Operation Night Watch. Previously a materials scientist at the University of Arizona, he moved to Amsterdam full-time in 2014 to combine his interests in science and art. “I saw that there was such a huge need for someone who could process images and data and help tell stories with pictures that I just jumped and made a complete left turn,” he says.

One of the main challenges he faces with Operation Night Watch is the sheer amount of data collected – a relatively new problem for the art world. “We had a long time where we had terrible images, because we didn’t have high resolution digital cameras,” he says. “We didn’t have computing infrastructure. We didn’t have fast internet. We didn’t have things like neural networks or these various tools that help us process this. And so now we’ve gone from that world to one in which, if we’re not careful, we drown in data.”

He estimates that Operation Night Watch will require dealing with around 600 terabytes of data. Take the high-res photographs: captured with a Hasselblad H6D-400c MS camera, these will produce the most detailed image yet of The Night Watch, with a resolution down to 4.75 micrometres. This means that one pixel in the photographs covers an area smaller than a single red blood cell; the resulting image of the painting as a whole will be almost 1 million pixels wide. “It’s not just a cult of resolution,” he says. “It’s not just feeling cool because we have the highest-resolution photos in the world. We can see paint pigment particles at that resolution, and you can’t see them at lower resolutions. We can see details of the brush strokes, and streaks of individual hairs in the paint brush.”

To Erdmann, being able to see the painting in such close detail does nothing to dispel his wonder at Rembrandt’s work. He talks about the “mystery of mastery”: zoom in really close and all you can see are abstract blobs of colour, but zoom out and somehow they create the perfect illusion of an object. He demonstrates on a digital image of two other Rembrandt paintings, Portrait of Marten Soolmans and Portrait of Oopjen Coppit, painted as a matching pair in 1634 to commemorate the subjects’ marriage. The only full-length portraits Rembrandt made, they were jointly purchased by the Netherlands and France from the Rothschild family in 2015, and were recently restored by the Rijksmuseum, trialling many of the imaging techniques that will now be used on The Night Watch.

On one of his monitors, Erdmann zooms in closely on Marten’s lace collar, then on Oopjen’s pearl necklace. Both become unidentifiable, filling the screen with smudges of grey, cream, black, tan. But when he zooms out, the accessories are immediately recognisable, almost photo-realistically so. “If you asked me to paint a pearl, I wouldn’t think that I should do that,” he says, gesturing at the swirls of pigment before zooming out again. “But then out here it’s like, oh yeah. Those are totally pearls … Somehow he knows how to translate what’s happening with the paint into what you’re going to perceive it to be.”

Vincent Fournier

To make one complete high-res image of The Night Watch, Erdmann has to combine the photographs so that there are no detectable flaws. If you’ve ever tried to stitch multiple photos together to make a panorama, you may have some idea of the difficulties this presents – the light may not be perfectly uniform from photo to photo; there may be some vignetting (where the image darkens at the edges); there may be some lens distortion; there could be some variation in focus; the camera may be tilted at a slightly different angle. Imagine doing this, but with thousands of images and much less scope for error.

Then there are all the other types of imaging. Much of Erdmann’s work involves finding new ways to visualize the data gathered by the different tools so art conservators and historians, often more visually inclined, can actually make use of it. One technique he has created, which he calls the “curtain viewer”, is a browser-based tool that stacks up to four different images on top of each other so that they are perfectly aligned. You can then see an interactive split-screen of all four image types, with your cursor at the intersection. Move the cursor, and you can immediately contrast any point on one image with the same point on another.

This makes it easier to compare image types – rather than moving your eyes back and forth to spot the difference, you can keep focused on the area of interest. Erdmann demonstrates the tool with a work by Hieronymus Bosch depicting John the Baptist next to a strange-looking plant. Move the cursor so that the infrared reflectogram image is over the plant, however, and you can see that there was once another figure there – possibly a donor who refused to pay up and so was painted out.

With The Night Watch, this tool could be used to compare the different element maps produced by the macro XRF scans, in order to see where multiple elements are present in the same spot and infer which pigment was used there. Overlap between tin and lead? You may well have lead-tin-yellow.

But there is a limit to how much data you can see in a single image, even if it is interactive. Erdmann shows me another tool he is working on, which he currently calls “pixel swarm”, and lets you interrogate all 33 element maps at once.

On screen, he brings up a painting called The Man in a Red Cap, attributed to the school of Rembrandt. First, he makes a scatter plot comparing two elements: cobalt and arsenic. This shows each pixel of the image that contains arsenic, cobalt or both; cobalt and arsenic together are indicative of the smalt pigment. Erdmann selects these pixels in the scatter graph and turns them blue. Arsenic alone is indicative of orpiment, a pigment that was highly toxic and degraded very quickly. There was a reason it was commonly used, however: it glitters gold. Erdmann tags these pixels yellow.

Going back to the image of the painting, the selected pixels now shine blue and yellow. The blue ones are largely present in the brown background but, curiously, also in the edges of the red hat. From this, we can deduce that the hat was originally smaller; the outer edges were added later, on top of the background. The yellow pixels form a patch on the man’s red coat: this would have originally sparkled.

Vincent Fournier

Beyond just comparing two elements, Erdmann can create a pixel swarm in 33-dimensional space, showing where every pixel of the painting falls in terms of all of the elements it contains. Exploring the graph, there are islands of pixels similar in composition, with bridges between them; the closer one pixel is to another, the more chemically similar they are.

Select a group of pixels and you can see where they correspond to on the painting. One pixel island highlights the edges of both the man’s hat and coat. To understand the full relevance of this, a bit of art history knowledge is required. In the 17th century, painters would have to mix their own paints, which would dry quickly, meaning they would never work with a full palette of colours but stick to one or two at a time. “What that means is whatever you’re painting with on that day is going to be chemically homogenous,” Erdmann says. As a result, we can infer that Rembrandt decided to extend the hat at the same time as he adapted the coat, changing the figure’s pose. “So what this is suggesting is that this is a giornata – this is what he did in one day.”

As well as interrogating a single painting in close-up, Erdmann develops tools to explore works in the context of an artist’s broader oeuvre. Training neural networks on all of Rembrandt’s paintings, for instance, can help researchers to identify common features or spot anomalies. He demonstrates with The Jewish Bride, another famous Rembrandt painting in the Rijksmuseum. Based on our knowledge of how Rembrandt usually combines different elements, he trained a neural network to predict what the painting should look like based only on its macro XRF data. He shows the results on screen side by side: on the left, a photograph of The Jewish Bride; on the right, the neural network’s prediction.

They are impressively similar, but with a few obvious flaws in the neural network’s attempt. On the bride’s face, for instance, are black speckles – a skin condition that isn’t present in the original. Looking at the element maps, it’s clear that calcium is present in these spots; the neural network has assumed, therefore, that they are bone black. Calcium, however, is present in something else: gypsum, often used in restoration. The speckles the neural network misidentified are not part of the original paint; they have been restored. “So in other words, to take a big step back, this is giving us the wrong answer because the assumption that those areas are painted by Rembrandt is wrong,” Erdmann says. “So I now have a neural network that can give me clues about whether something is authentic.”

Vincent Fournier

The Night Watch has a storied past when it comes to conservation. Officially titled Militia Company of District II Under the Command of Captain Frans Banninck Cocq, the group portrait was commissioned for the Kloveniersdoelen, the headquarters of the civic guard, where it was likely not particularly well looked after. “It hung there in the big hall, which meant that they held target practices there and feasts, so I’m sure some of the damages on the painting may have already occurred then,” says conservator Esther van Duijn, who is researching the conservation history of the painting.

Van Duijn’s work consists of digging through old archives to find any documentation relating to the state of the painting and previous restoration efforts. Her research helps make sense of some of the results of the scientific imaging, and tell the story behind The Night Watch’s present condition.

The painting was moved from the militia headquarters to Amsterdam town hall around 1715. It’s at this point it suffered probably its greatest and most irreversible damage, in a decision that today would be unthinkable: as the painting was too large for its new location, whole sections were chopped off all four sides.

We have an idea of what The Night Watch must have looked like before this butchery thanks to smaller copies made by some of Rembrandt’s contemporaries. Currently hanging next to the original in the Rijksmuseum is one version, only 86cm wide, attributed to Dutch painter Gerrit Lundens. The most obvious difference is the presence of two extra figures on the left hand side, presumably sliced off the original when it was re-sized. Before the Operation Night Watch researchers started imaging The Night Watch, Smelt says, they tested some of the tools on this smaller version. One intriguing thing they found was a transfer system beneath the paint – a grid of squares similar to the kind you might find in a child’s colouring book to help accurate copying.

Computer analyses of the Night Watch canvas also reveal that there is material missing. The painting is made from three strips of canvas sewn together (if you look at it in the gallery, you can see the horizontal seams about a third and two-thirds of the way down). Erdmann has built computer models that compare the thread pattern in the different pieces and show that they are all cut from the same roll of fabric. Crucially, there is no “cusping” at the edges of the painting. When a canvas is attached to a stretcher, it distorts the threads, creating a sort of scalloped pattern around the tacks. This is not the case with The Night Watch, suggesting those edges were subsequently cut off.

While in the town hall and then the Rijksmuseum, the painting underwent many restoration treatments. It was cleaned, had various types of lining – additional canvases adhered to the back of the painting – applied and removed, and was coated in layers of varnish. The history of The Night Watch is in a sense the history of art conservation, reflecting the changing techniques and fashions through the ages. In the 19th century, thick yellow varnishes were popular. “They prefer what they call the ‘golden tone’ or ‘golden glow’, especially on 17th century old masters paintings,” says Van Duijn.

This caused a problem, however, as microscopic cracks would appear in the layers of varnish, creating a cloudy effect. At the same time, restorers started getting more hesitant to do much to old paintings, for fear of damaging them. In 1870, German chemist Max von Pettenkofer came up with what seemed to be a miracle treatment: a way to make the varnish clear without touching it. His “regeneration” method involved using alcohol vapour to soften the varnish and smooth over the cracks, making it translucent once more. This was applied to The Night Watch for the first time in 1889, to great success. “The only thing was, it didn’t last,” says Van Duijn. “So within ten years it had to be done again.”

After a few decades, the treatment had stopped working. Before the museum could decide what to do, the Second World War broke out and The Night Watch was evacuated for safekeeping, at one point rolled up – paint side out, to minimise cracking. When the painting returned to the Rijksmuseum in 1945, a new lining was added and the layers of thick yellow varnish were removed. The difference was, quite literally, night and day – despite its nickname, The Night Watch doesn’t actually depict a night scene; the darkening layers of varnish, on top of Rembrandt’s painting style, made it look gloomier than intended. Every time the painting is treated, van Duijn says, people say the same thing: “It’s not a Night Watch any more, it’s a Day Watch!”

By the 1970s, the museum was considering another restoration effort, but its hand was forced in September 1975 following a brutal attack. A museum visitor slashed the painting 12 times with a bread knife, slicing straight through the canvas and leaving vertical gashes up to half a metre tall. A piece of canvas in the middle of the painting was left dangling off between Captain Banninck Cocq’s legs. If you know where to look, you can still see a slight scar where the wounds were repaired. They show up even clearer on the macro XRF element maps, where the materials used to fix the tears contrast with the chemistry of the original paintwork.

The new technological analyses are already helping to fill in some of the remaining puzzle pieces when it comes to the painting’s conservation history. Van Duijn will compare data from the OCT imaging, for example, with archival information about when and how different varnish layers were added. To test the technique, the team tried out the OCT imaging on an old insert – a small piece of painted canvas – that was once added to The Night Watch to fill a hole. It was removed during the 1975 restoration, but they did not know when it had been attached.

The OCT results showed that the small patch had quite the history of its own. “Now we think it’s actually quite old – the insert was maybe even added in the 19th century, maybe even earlier,” Van Duijn says. “So it has gone through all these cycles – it has gone through the regeneration phases, it has the varnish that was put on in 1945-1947 – and now it’s this bit of evidence from the past, and it’s much more interesting than I ever thought.”

By revealing information about the painting’s current state, the imaging research will also inform any attempt at restoration. The cloudiness around the dog, for instance, may be due to a reaction in the paint or in the varnish layer. Only by interrogating the chemistry will the conservators be able to pin down the causes of degradation and decide how best to proceed.

Vincent Fournier

Today, the general ethos of art conservation is to remove anything that is not original, then retouch where necessary, taking care not to touch the original paint. The Operation Night Watch team will not decide what to do to the painting – or whether to do anything at all – until they have analysed the data collected and have a clear picture of its condition. Obvious candidates for restoration include removing and reapplying varnish, and relining the canvas.

This restoration phase of the project will begin around June 2020, with all of the work being completed under the gaze of the public. The glass box is large enough that the painting can be removed from the wall and laid flat. “It is still a painting, it’s an object – you can lift it,” Smelt says. “We are going to touch it, because that’s our work. That’s normal for us, but I understand it’s not that normal for visitors.”

If questions remain unanswered after the imaging, the researchers may even take new cross-sections of the paint layers – tiny, sub-millimetre samples that are removed by scalpel to undergo further microscopical analyses. “It’s really tiny, but it gives us such valuable information,” Smelt says. “But you need to be sure to do it because, yeah, you put a knife into your painting.”

The biggest transformation, says Van Duijn, would be if the team decides to change the varnish, which currently makes The Night Watch look noticeably yellower than some of the other militia paintings hanging nearby. It would look most spectacular halfway through the varnish removal process, when the contrast would be most striking. Individual retouchings of paint should not be noticeable, she says, but there may be an accumulative effect: “It creates a more even continuous layer; it’s a bit less restless for the eye to look at.”

The new imaging technology means any work the team does will now be documented on a micro level – a prospect at once exciting and scary. Future conservators will be able to compare individual pixels before and after the treatment, making it easier to track changes but leaving no space for mistakes. “You’re really under a magnifying glass yourself as a conservator in that way,” Van Duijn says. Perhaps in 50 or 100 years, today’s gold-standard restoration methods will seem as outmoded as the 19th century’s penchant for thick yellow varnish.

Beyond The Night Watch, the team aims to leave a legacy for the entire museum community. The Rijksmuseum will apply the technologies developed along the way to more artworks in its collection, and will make them available for use by other museums too. “It’s a very special project, but it’s also a great excuse to be able to work on making our technology advance,” Erdmann says. People from the aerospace and medical communities have already been asking about some of the imaging technology and data visualisation tools, he says. “Historically, the museum world has been 20 years behind the technology curve, but now we’re really sitting at the forefront.”

More great stories from WIRED

🐘 The illegal trade of Siberian mammoth tusks revealed

🙈 I ditched Google for DuckDuckGo. Here’s why you should too

📧 How to use psychology to get people to answer your emails

Like this article?

Share on facebook
Share on Facebook
Share on twitter
Share on Twitter
Share on linkedin
Share on Linkdin
Share on pinterest
Share on Pinterest

Leave a comment