Imaging 17th century art

Materials World magazine
1 Jan 2018

A non-destructive imaging technique may offer a greater insight into 17th century art. Kathryn Allen reports. 

The methods by which artists prepared canvases and applied layers of paint have been observed by researchers at the Georgia Institute of Technology, USA, and detailed in Global mapping of stratigraphy of an old-master painting using sparsity-based terahertz reflectometry, published in Scientific Reports

Using terahertz scanners and advanced signal processing techniques, developed for petroleum exploration, an unprecedented insight into the layers of 17th century artwork has been gained. 

The observations could be useful in authenticating artworks, highlighting previous restoration efforts and showing damage. Outside of the art world, they could be used to detect skin cancer, the thickness of automotive paints and the adhesion of turbine blade coatings. 

17th century secrets

Using a commercial terahertz scanner, the team studied Madonna in Preghiera by the workshop of Giovanni Battista Salvi da Sassoferrato, loaned from the Musée de la Cour d’Or, Metz Métropole, France. Described in the paper as typical of the period and having previously resisted attempts to analyse its stratigraphy (the order and distribution of layers of paint) – due to the layers measuring only tens of micrometres (μm) – it is an oil painting on canvas, mounted on a wooden stretcher. 

The scanner’s electromagnetic wave generator emits signals that penetrate through the layers of paint. Different pigments and physical structures, such as imperfections in the layers, are revealed by wavelengths as they reflect some of the beam back to the scanner. Oil paint and newer varnish do not fluoresce under UV – meaning recent retouchings appear as darker patches on the surface. 

The data was processed by a computer-based signal processing technique – sparsity-based time-domain deconvolution – and a three-dimensional map of the image was constructed using the signals reflected from each layer. The team was able to identify layers including the canvas support, ground (the background surface), imprimatura (the initial colour stain painted on the ground), underpainting, pictorial and varnish layers. A formerly unknown layer of varnish restoration was also detected. 

Sparse deconvolution

This new technique allowed layers 20μm thick to be detected – while previous methods have only identified layers of 100–150μm. This level of detail can be achieved as the scanner operates at very high frequencies – one terahertz (THz) is equal to 1012 hertz – easily penetrating layers of paint, but blocked by carbon black and other conductive pigments. 

David Citrin, a professor in the Georgia Tech School of Electrical and Computer Engineering, USA, explained the process. ‘The system produces roughly single-cycle terahertz pulses with a spectrum extending from about 100GHz to 3THz.  The pulses are produced at a repetition rate of around 100MHz. Terahertz pulses are directed onto the painting and the temporal profile of the reflected terahertz pulses are measured – both their amplitude and phase. The painting is layered, so we obtain a reflected terahertz pulse from the interface between each pair of layers due to the small refractive-index difference between the successive materials.’ The time delay of these received signals can then be mapped to represent depth. 

Signals received from thick layers of paint (around 100μm) don’t overlap much in time. However, as the layers become thinner, the signals overlap more frequently. According to Citrin, received signals from layers of a few tens of micrometres thick become impossible to visually separate as they overlap too much. 

Citrin said, ‘Our received signal is made up of two sources of information – one is determined by the temporal terahertz pulses produced by our apparatus. These we can measure quite accurately. The other part is determined by the unknown intrinsic structure we have, in other words, the impulse-response function, which contains the information about the time delays between the various [received signals] as well as the relative amplitudes of [these signals]. Our received terahertz signal is thus the convolution of the signal produced by the apparatus and the impulse-response function of the structure.’ 

Sparse deconvolution allows the team to separate these signals. Junliang Dong, Georgia Institute of Technology, carried out the experiments applying sparse deconvolution. As Citrin explains, it ‘enables us to back out the impulse-response function from our received signal. It involves considerable offline computer numerical processing after the data is acquired. The end result is the impulse-response function we are after – a sequence of time-delayed peaks. These various time delays, together with reasonable assumptions about the indices of refraction of the layers [...], enable us to reconstruct the local stratigraphy of the painting.’

A new insight 

Claire Meunier, Curator of Fine Arts at the Musée de La Cour d’Or, Metz Métropole, told Materials World, ‘The most important difference and advantage of the terahertz technique compared with other non-invasive techniques is that it does not show only one layer of the painting but all the layers. The infrared rays only show the preparatory layer and is usually used to reveal if there is any preparatory drawing beneath the painting. Using fluorescence ultraviolet is useful only on the varnish layer – the upper layer of the painting – to show the past retouchings and/or restorations.’

According to Meunier, no other analysis techniques were used on the Madonna in Preghiera, but to get similar results – identifying all the layers – the only other way would be to use a destructive technique, such as microscope analysis. This involves taking a micro-sample of the painting, and in this case several samples would have been needed to identify the various layers in the dark sections, such as the Virgin figure. 

Of the layers identified, Meunier explains how terahertz scanning confirmed the presence of a previously unknown layer of varnish restoration. A difference in texture of the black colour around the head of the figure had been detected by experts, however they could not identify whether it was a restoration or degradation. ‘The Terahertz technique shows that the varnish is thicker at this place, which is coherent with a restoration. A degradation of the painting would have lead to a thinner varnish layer. It helped us in understanding what was happening to this precise part of the painting,’ said Meunier. 

The team believe this technique improves on current art analysis methods. According to Citrin, it offers a less restricted insight compared to optical and infrared imaging, which are limited to near-surface analysis. Infrared imaging – using near and short-wave infrared (up to 2.5μm) – works in a similar way, by observing the absorption and reflectance of infrared light by an object. X-radiation methods are also used, however, as Citrin explains, ‘X-rays penetrate though paintings but they exploit a different contrast mechanism [to terahertz imaging]. X-ray cross section is connected to the elemental atomic numbers of the paint constituents – that often does not allow them to resolve the sequence of thin paint layers applied by the artist.’ He also points out concerns raised over the safety of X-ray operators due to ionising radiation and the potential for X-ray-induced chemical changes to the painting. 

Other conventional approaches to art analysis include raking light – whereby the painting is illuminated with bright light, usually angled, to reveal features on the top layer of paint – and UV fluorescence, where the painting is illuminated using incident radiation. The varnish on paintings exposed to UV radiation display different glows, because the UV photons are absorbed by the varnish layers and emit lower energy photons, which are observable to the human eye, highlighting restoration efforts.

Moving forward

Having already imaged a Byzantine coin through a layer of oxidation, the researchers plan to use the technique to study a small part of a 12th century wood panel painting. Damage to the painting and an uneven surface pose challenges, as the terahertz pulses may be scattered considerably. The paint is also very thin, however, the team have so far been able to detect the pictorial composition using the scanner. 

It could also be used to detect damage, such as delamination, in composites on the scale of tens of micrometres and larger. However, with carbon-fibre reinforced polymers in particular, it can be difficult to detect damage, as the conductive material is highly reflective to the incident terahertz pulses.

To read the paper visit