It’s a Friday morning in the painting conservation offices at the Art Institute of Chicago. I’m ushered through security and brought to a staging room. It’s cavernous. I tilt my head up: Towering black matte walls meet a black ceiling somewhere out of sight, and I can just make out the stage lights overhead. I shift my line of sight to the walls around me. Rolling wooden easels of every size stand sentry among folded tripods. Stacks of calibration targets, poster-sized checkerboard patterns and bright white sheets of paper, line the shelves. My eyes are just starting to adjust to the dark when a conservator wheels in a cart, and suddenly Pablo Picasso’s jaunty asymmetric painting A Faun Musician lies before me.
It’s smaller than I expected, about the size of a page from a hardcover book. Actually it is from a hardcover book, painted opposite the title page in a 1947 edition of Petrarch’s sonnets. The painting depicts a faun—a mythical goat-like creature—surrounded by leaves and playing an aulos, which resembles a two-tailed clarinet. The painting has a muted palette of gray-blues and dark olive green but maintains a sense of dynamic movement through the continuous black line that defines the faun’s features. Flashes of bright blue around the edge of the painting contrast with the gray tones of the portrait and hint at what makes this Picasso stand out and why I came to study it. These unexpected pops of color alerted conservators to a hidden Picasso painting believed to be a vase of flowers under Faun. It’s my job as a cultural heritage scientist to identify and map the pigments, or colorants, that Picasso used in each layer of this painting.
My work on Faun comes on the heels of other discoveries by my colleagues at the Center for Scientific Studies in the Arts of hidden masterpieces under many of Picasso’s paintings. These findings show that Picasso incorporated pre-existing shapes on the canvas into his new creations and made alterations frequently, such as painting over the hand of the main figure in the Blue Period (1901-1904) painting La Miséreuse accroupie (1902). Information about each of the layers in Picasso’s paintings reveals much about his creative process. My work on Faun continues this effort to understand Picasso’s techniques and the underlying composition of his paintings.
In its simplest form, paint has two components: a binder and a pigment. The binder is what holds the paint together and prevents it from spreading too much on the canvas or support. The opacity and color of the paint come from the pigment component, which is what I study. Pigments come in every imaginable color and demonstrate a wide array of chemical properties, and this chemical diversity makes them ideal materials to study using different types of imaging.
Imaging is the bread and butter of cultural heritage scientists because it’s non-invasive to the object—we can study the object closely without damaging or altering it. Before I came to the Art Institute, scientists in the conservation lab collected X-ray fluorescence (XRF) maps of Faun. When X-rays ping off an atom in a pigment, they force the atom to re-shuffle its electrons, giving off distinctive energy signals in the process. With a detector tuned to these signals, we can identify which kind of atom the X-rays excited. XRF is most useful for inorganic pigments because they contain metals—copper, iron, tin, lead and chromium, for example—with lots of electrons to shuffle around. XRF isn’t as useful for the elements in organic dyes—carbon, oxygen, and nitrogen—because these atoms have fewer electrons to rearrange and produce signatures that overlap. After scanning the painting, the scientists generated grayscale maps showing the distribution of elements in each region of the painting.
XRF doesn’t give us the whole picture, however, because it only tells us which elements are in the pigments—that is, it doesn’t tell us about the molecules that are present. This is an important distinction in the world of paint. Sister pigments with similar elemental makeups may have distinct molecular arrangements and appear as different colors. Red lead and lead white, as their names suggest, both contain lead and would give identical XRF signals but are different colors due to their different compositions.
My job is to probe these molecular configurations to precisely identify the pigments present in the painting. To do this, I use hyperspectral imaging. Atoms in pigments are bonded together, and when these bonds absorb just the right color of light, they vibrate like microscopic springs. The bonds are choosy about which colors of light they absorb and which they reflect. By shining light containing every color (white light) on the pigment, I collect a fingerprint-like spectrum that shows which colors the pigment absorbed or reflected. Spectral fingerprints are unique to each material and can distinguish between sister colorants, like lead white and red lead. Hyperspectral imaging earns its “hyper” prefix by collecting these spectral fingerprints at every pixel in an image of the painting. The result of this technique is a map of the painting like XRF, but this time the molecular fingerprint defines each region instead of the type of metal atom present.
Back in the dark staging room, collecting a hyperspectral image doesn’t take long. The camera is about the size of a box of tissues and sits on a track that allows it to slide horizontally. Once we have the camera secure and calibrated, we shine a white LED lamp on the painting and turn out the overhead lights to prevent any stray light from contaminating the results. The camera moves along its track almost imperceptibly in front of the painting. Fifteen minutes pass quietly in the dark, and the scan is over. This relatively short scan time contrasts with XRF, which can take several hours to collect a full scan. The speed of hyperspectral imaging is part of what makes this technique attractive to conservators. Shorter imaging times mean less exposure to light for paintings, and hyperspectral imaging has the added benefit that it works well with low light levels, which pose minimal threat to the object.
The real challenge of hyperspectral imaging is the sheer volume of data that a single 15-minute scan can produce. We measure reflection across 240 wavelengths, or colors, at each pixel. A high-definition image usually has a resolution of 1280 x 720 pixels. Doing the math, that means there are 998,400 pixels for us to acquire. We end up with millions of datapoints, so we turn to computer algorithms to sort the data into manageable groups. We borrow one of our key algorithms directly from developers at Facebook, who use similar codes to sort through pictures and streamline search results. When applied to our data, the algorithm sorts pixels into islands on a color map. It clusters pixels that are mostly green together in a region of color space separate from the blue pixels, for example. From there, we take the spectral fingerprints of these simplified color groups and match them using a library of known pigments to make an identification of the colorants Picasso used.
Scanning the painting at the Art Institute is just a blip in the months-long process that goes into studying Faun; the real work of data processing begins back in the lab on campus. The bulk of my time is spent with my laptop trying to decipher data representations almost as abstract as the art I am studying. This is where my story with Picasso ends, for the time being. Although we are closer than we’ve ever been to understanding the paints Picasso used in the many alterations of A Faun Musician, we are still a few months away from a positive identification.
The Faun Musician, 1947, (327 x 502 mm; Gift of Dorothy Braude Edinburg to the Harry B. and Bessie K. Braude Memorial Collection; The Art Institute of Chicago 1998.720) © 2018 Estate of Pablo Picasso / Artists Rights Society (ARS), New York
La Miséreuse accroupie, 1902, Pablo Picasso, Oil on canvas (101.3 x 66 cm), Art Gallery of Ontario, Toronto ©Picasso Estate, SODRAC (2017)