How we used a particle accelerator to find the hidden face in Degas's Portrait of a woman
Edgar Degas's painting Portrait of a woman is an enigmatic piece. When it was first acquired by the National Gallery of Victoria in 1937, it was unveiled to mixed reviews.
Some commented that it showed the hallmarks of the French painter's style around the 1870s. Others criticised its brown hues and the apparent discolouration across the woman's face.
Little did they know at the time that the painting held a secret: there was another portrait of a woman, inverted, lying just under the surface. Some of the discolouration was due to this other ghostly figure bleeding through.
It appears that Degas abandoned this earlier work and repurposed the canvas for the newer portrait. But when did he paint the first woman? Who was the model?
Thus began a quest to reveal the hidden woman without disturbing the portrait on top. X-ray imagery showed a little more detail, revealing the faint outline of a young woman, painted perhaps only shortly before the canvas was re-used.
Subsequent infra-red photography suggested the original figure was painted as early as 1860, while other barely visible features hinted at an earlier creation.
And this is where the Australian Synchrotron comes into the picture.
Brighter than the sun
As is often the case in cutting-edge science, a recurring challenge is to devise technology that facilitates what researchers want to achieve. For every leap forward in power or speed, supporting equipment and infrastructure is needed to make the most of the new innovation.
Such was the situation at the Australian Synchrotron. Just after beginning operations in 2008, we formed a strong collaboration with the CSIRO and Brookhaven National Laboratory in the United States.
At the time, we were receiving vast reams from the synchrotron's X-ray fluorescence microscopy beamline as it delivered light a million times brighter than the sun into a variety of scientific samples.
This collaboration helped us to develop an X-ray fluorescence detector capable of operating significantly faster than the current technology available at the time to make the most of this data.
We have also had an exciting and long-lasting collaboration with the National Gallery of Victoria (NGV).
The NGV has a fantastic collection of Australian and international art. Some items in the collection have unanswered questions that conventional analytical techniques are not able to resolve, such as Degas's Portrait of a woman.
This collaboration has proved highly successful, and the new technology developed by CSIRO and Brookhaven National Laboratory in the United States, dubbed the Maia detector, opened up innovative research that was previously not possible.
Before the Maia detector, we were restricted to acquiring small images of a sample - such as a leaf containing traces of metals, electrodes used in medical implants or the hidden layers in a painting - which contained a rather limited number of pixels, preventing us from seeing the whole picture.
With the new Maia detector, this hindrance was overcome. We are now able to routinely acquire elemental images composed of millions of pixels over large areas in only an hour or so.
With the technology available at the Australian Synchrotron, we believed there was a good chance we could reveal the hidden portrait by subjecting small areas to radiation for only a fraction of a second. And, importantly, we could do this without damaging the artwork.
After several months of planning, the painting arrived from the NGV early in the morning and we secured it in a custom mount, with the detector only 2mm above the painting's surface. We set up the coordinates of the area to scan so we could capture as much of the hidden portrait as possible.
After some fine-tuning of the X-ray beam, we launched a scan that would take approximately 33 hours to complete, giving individual images in excess of 31 megapixels, which is beyond the resolution of most of today's best digital cameras.
We were able to do an initial analysis of the elements present and load this into the computer to give a real-time image of the painting as the data was collected.
It was incredibly exciting to see the image build up on the computer monitor and reveal, pixel by pixel, the hidden portrait beneath.
Based on our analysis and comparison to other works by Degas, we suggest the hidden portrait is of the model Emma Dobigny.
We know Degas painted Emma other times, such as the famous portrait from 1869 named after her.
We hope this data can provide art historians with more information about Degas and the evolution of his work. And we hope that the fruitful collaboration with the NGV and CSIRO continues for many more years to come.