Assessing the laser cleaning of paintings by means of OCT and reflection FT-IR
OCT and FT-IR measurements carried out on the red background before the laser cleaning tests provided the local painting stratigraphy and the chemical composition of the surface layers, respectively. In the OCT tomograms (Figs. 1d, 2a,d and 4a) the three surface transparent layers over the overpaint layer were recognized. The red ochre overpaint layer (No. 6) is semi-transparent to OCT probing light and, thus, varnish layers below (Nos. 3–5) are visible as well (Figs. 1d, 2a, c, d, f, 4a, c). It is worthwhile to note that fading “tails” visible below the surface of the opaque paint (No. 2) are the result of multiscattering phenomena in the paint and do not reflect its real structure.
The uppermost layer was attributed to a wax coating identified by FT-IR. In the infrared spectrum collected before the treatment (Fig. 2b—black line), indeed, absorption features of beeswax at 2954 cm−1 (asymmetric stretching CH3), 2922 and 2853 cm−1 (asymmetric and symmetric stretching CH2, respectively) and the doublet at 732 and 716 cm−1 due to the CH2 rocking mode were observed [27]. Additionally, the band at 1320 cm−1 revealed the presence of oxalates [28], probably of calcium, while the OH stretching signals at 3398 and 3555 cm−1 and the characteristic SO42− vibration modes (ν3 at 1120 cm−1 and ν4 at 676–608 cm−1) indicated gypsum [29] from deposition or migration.
The main goal of laser cleaning on this multifaceted painting was to gradually remove the upper varnish layers until a safe level was reached, with the red ochre overpaint left intact; this safe level was established to consist in maintenance of a “buffer” residual varnish layer over the red ochre overpaint so as to protect it from photochemical oxidation due to laser illumination. Furthermore, additional tests aimed at investigating the removal of the red ochre overpaint layer itself have also been performed. Towards these goals a series of cleaning tests employing different fluence values (F: 0.7–1.1 J/cm2) were executed, resulting into a different ablation depth and thus cleaning level. The cleaning tests were fine-tuned using variable number of pulses (N: 10–50) according to the indications provided by the OCT and reflection FT-IR survey carried out before and after each laser application (treated areas shown in Fig. 1a, e).
Table 1 summarises results of one of these series of cleaning tests (C series) providing the relationship between various combinations of F and N and the different cleaning levels, namely reduction or removal of the outermost wax layer and the two upper varnish layers.
From Table 1 it can be seen that the cleaning depth can be regulated by using different laser cleaning parameters. In particular, by adopting the lowest fluence of 0.7 J/cm2 with 25 pulses (C4 area) a mild cleaning result was obtained, which mainly refers to the removal of the wax layer and partially of the upper varnish layers, corresponding to a total thickness of 19 ± 5 μm. On the other hand, roughly doubling the number of pulses at the same fluence (46 pulses, C5 area) the surface appears over-cleaned, showing a patchy distribution of the residual varnish. The most acceptable outcome for the needs of conservation, intended at removing the varnish layers but leaving the overpaint intact, was reached at laser irradiation fluence of 0.9 J/cm2 with 10 pulses (C3b area); in this case, removal of wax and partially of the upper varnishes is observed from the comparison between the OCT tomograms acquired before and after the laser treatment (varnish removal is estimated to be 30 ± 6 μm, Table 1 and Fig. 2a, c). These findings were confirmed by the FT-IR profiles collected on the same area before and after the cleaning step. In detail, in the spectrum related to this latter step, red line in Fig. 2b, the infrared features of beeswax disappear, while the varnish is still visible. This latter is clearly distinguished from the underlying oil binder thanks to the identification of the CH2 asymmetric bending mode at 1383 cm−1, which usually appears much more pronounced in the natural varnish spectral pattern than in the oil one and can be used as a marker band in reflection mode FT-IR spectroscopy [30]. In addition, the infrared analysis allowed us to detect the removal of calcium oxalates (νs(CO) band at 1320 cm−1), and of surface gypsum (from deposition or migration).
For higher number of pulses (C3a with 20, C2b with 40 and C2a with 50), maintaining the same value of F = 0.9 J/cm2, a fine modulation could be reached: the thickness of the removed material was increased with increasing N values. From the surface maps generated by OCT data (Fig. 3a) a significant difference in ablation efficiency can be already noticed by doubling the delivered number of pulses (C3a area, N = 20, vs C3b area, N = 10); also FT-IR measurements before and after laser cleaning of the C3a area (Fig. 2e) exhibit a significant reduction of the varnish signal, while spectral contributions ascribable to the underneath paint layer start to be visible, as better clarified from the analysis of the C1a area (see discussion below). The same conclusion can be drawn from the map of the removed varnish (Fig. 3b). Close inspection of both images (Fig. 3a, b) reveals that despite the ablation of the varnish, the craquelure structure remains intact as it is seen by direct comparison of details in the surface maps (Fig. 3a) and as yellow lines (indicating significantly lower local ablation) in the map of Fig. 3b. This effect is especially visible for the C3b area where the amount of removed material is also more homogenous over the whole area with respect to the C3a area; this should be evaluated as a desirable result from the conservation point of view and confirms the conclusion about the most acceptable outcome for area C3b, stated above, based on the analysis of the OCT cross-sections and FT-IR data. Areas C2b and C2a (N = 40 and 50, respectively), on the other hand, appear over-cleaned; the two upper varnish layers were removed while the overpaint was partially ablated, as seen by OCT and FT-IR results (Table 1).
At last, in the attempt of revealing the original red vermilion paint layer, for the highest F value tested (1.1 J/cm2) at increasing number of pulses (C1b = 20, C1a = 30) the red ochre overpaint layer as well as the varnish coatings were removed (Table 1). In Fig. 4a, c OCT tomograms acquired before and after the laser treatment with the most intense operative parameters of this study (area C1a, F = 1.1 J/cm2 and N = 30) clearly show the uncovering of the original red vermilion paint layer, which appears with increased surface roughness. Reflection FT-IR measurements confirmed OCT results: after the cleaning step (red line in Fig. 4b) the diagnostic signals of beeswax and of the surface varnish layers are not visible anymore, while the lipid component of the oil binder is detected (diagnostic bands: C=O stretching and δ + ν CH modes) along with signals in the spectral region of the amide I and amide II bands (1550–1650 cm−1); these latter possibly come from the underneath ground layer, as also suggested by the permanence of a sulphate band at about 1100 cm−1, despite the fact that the surface sulphate deposits were removed by laser cleaning. Interestingly, the infrared spectrum acquired on the original paint layer uncovered by laser cleaning is particularly noisy, especially at low wavenumbers (red line in Fig. 4b); a similar result was obtained for all the areas where either the red ochre overpaint or the original red vermilion paint were partially uncovered (Table 1). This fact might be explained by a possible irregular action of cleaning with the specific laser parameters on the treated paint surfaces, which, notably, has been readily revealed by both OCT and FT-IR.
Assessing the laser cleaning of paintings by means of LIF
While the combined use of OCT and reflection FT-IR spectroscopy has proved to be successful in assessing the treated surface at different laser parameters, the possibility of equipping the laser cleaning workstation with a diagnostic modulus for on line monitoring and control of cleaning process is still to be explored. In this perspective, laser induced fluorescence spectroscopy has been evaluated having the definite advantage of being compatible with the optical set up of a laser cleaning workstation (e.g. sharing the same laser source used at different fluence values). The use of UV–Vis–NIR fluorescence spectroscopy (exploited with different type of excitation sources, namely, filtered Xenon lamp, laser, and more recently LED) to analyse and study cultural heritage objects and materials is well established [31, 32]. Its use as a diagnostic tool is recommended in combination with UV–Vis reflection spectroscopy and when organic dyes and/or pigments are present with rather good emission quantum yield [33]. Conversely, fluorescence spectroscopy of organic materials, like varnishes and binding media, does not provide satisfying specificity for their differentiation [32]. Nevertheless, the use of LIF as a tool to assess and monitor the cleaning processes (using conventional [34] as well as laser-based [9, 12, 35, 36] methodologies) can be done on the basis of direct comparison of luminescence spectra before and after cleaning.
The fluorescence intensity (IF) depends on the amount of energy absorbed at the excitation wavelength and thus, according to the Beer–Lambert law, on the thickness of the remaining film after cleaning (\(I_{F} = Z \cdot I_{0} \cdot \varPhi_{F} \cdot \left( {1 - 10^{ - \varepsilon c d} } \right)\), where ΦF is the fluorescence quantum yield, Z is the geometrical (collection angle) factor, I0 is the intensity of the incident probe laser beam, ε is the molar extinction coefficient, c is the molar concentration and d is the optical path length). Therefore, in our study, the reduction of the fluorescence intensity recorded on the laser treated surfaces may be indicative of the thinning of the varnish layer. In parallel, changes to the shape and position of the fluorescence profile can be indicative of a photo-ageing process. Specifically, broadened and red-shifted bands are associated with the presence and formation of new fluorophores (i.e. new C=C bonds show increased fluorescence at longer wavelengths) and can be linked with oxidized and cross-linked products that characterize aged polymers [37]. Finally, the appearance of new bands may indicate signal from underlying materials (e.g. specific emitting pigments in paint layers).
In the case study presented herein, the LIF spectra recorded after each laser cleaning test (red lines in Fig. 5a, c) are compared with the ones recorded before laser irradiation (reference spectra, black lines in Fig. 5a, c) on the basis of the intensity (Imax) and wavelength (λmax) of their main fluorescence band (Table 1 and Fig. 5). All spectra are background corrected in order to compare the absolute maximum fluorescence intensity reflecting the chemistry of the exposed surface. Moreover, the spectra are also normalized to their maxima in order to clearly visualize any changes in λmax (lower part in Fig. 5a, c).
As it can be seen in Fig. 5, all the recorded fluorescence spectra are relatively broad with maxima at ca. 473 nm. In most of the cases Imax appears reduced indicating elimination of the varnish thickness in agreement with OCT and FT-IR analysis. A characteristic example is demonstrated in Fig. 5a (upper part); Imax of the LIF spectra recorded for the area C1a is significantly decreased after cleaning, confirming the elimination of the varnish layer. This is also illustrated in the digital microscope UV-induced fluorescence image recorded (Fig. 5b, lower part), where the area C1a after the laser treatment appears darker due to absence of emitting materials. Comparison of the normalized spectra in Fig. 5a (lower part) leads to the observation of a notable red-shift of the fluorescence maxima (Δλmax ~ 10 nm) and a new band (shoulder) at ca. 610 nm. This observation was recorded for most of the laser-treated areas except for the C3b (Fig. 5) and C4 ones (Table 1). As discussed previously, the red-shift can be correlated to photo-oxidation of the probed layer while the shoulder at ca. 610 nm (Fig. 5a) is tentatively assigned to vermilion (HgS) [14, 32, 38, 39], thus indicating that cleaning may have partially reached the original paint layer (layer No. 2 in Fig. 1c).
In contrast in test areas C3b (Fig. 5c, d) and C4, where FT-IR and OCT analyses have shown partial removal of the outermost varnish layers, the fluorescence spectrum, although maintaining the same profile, presents higher intensity values after the cleaning. No other changes (e.g. λmax shifting) are observed after laser treatment (Fig. 5c, lower part). This result is interpreted as partial varnish removal up to a layer where photo-oxidation may be present and is also illustrated in Fig. 5d, through the digital microscope images, which both confirm that a varnish layer remains and no paint layer is reached.