Nonlinear optical microscopies (NLOMs) are primarily employed in the biological and biomedical fields for the non-invasive, label-free imaging and diagnostics of both cells and tissues, providing morphological information in deep tissue with sub-cellular spatial resolution [1–6]. Biological samples from different medical fields, such as regenerative medicine, tissue engineering, neuroscience and in vitro 3D models, can be easily imaged with these techniques because of their intrinsic properties, e.g. the presence of autofluorescent species or the hyperpolarizability nature of some molecules, such as collagen or myosin, but also cellulose and starch [7]. NLOMs application in the Cultural Heritage (CH) field dates back some 15 years [8–10] and their ability to perform optical sectioning in turbid media [2, 10], overcoming the limits of well-established optical techniques such as OCT, brought them to the fore. In these techniques, the signal intensity is a nonlinear function of the laser power: the excitation is, therefore, spatially confined to the focal volume, reducing any out-of-focus photochemical phenomena such as photobleaching and, hence, the risks of damage [11]. The possibility of simultaneously detecting multi-photon excited fluorescence (MPEF) and second and third-harmonic generation (SHG and THG) signals allows for the acquisition of morphological, structural and chemical information at the same time. In specific, two and three-photon excited fluorescence (TPF and 3PF, respectively) permit the detection of fluorophores located deeper into the specimen; second harmonic generation (SHG) enables the investigation of crystalline and non-centrosymmetric structures, and third harmonic generation (THG) detects the interfaces through the local differences in refractive indexes [12–18]. TPF, 3PF and THG, first applied to measure the thickness of varnish layers to support the cleaning process [19], were then used to perform cross-sectional imaging of varnish layers to monitor the effects of laser ablation and ageing [20, 21]. TPF and SHG have been usefully applied to investigate the microstructures of different wood species composing artworks [13, 22, 23]; to characterize cellulose-based fabrics treated with a flame-retardant for tapestry preservation [24] and collagen degradation in parchments [14, 25]. TPF has been used to observe and quantify the corrosion layers in silver-based artefacts [26] and measure acrylic and egg tempera paint layer thicknesses [10, 12, 27]. As NLOMs are progressively triggering interest in the CH field, the scientific community has to face the safety of these techniques, establishing this evaluation on previous experiences in biological and biomedical fields [28]. Fluorescence emission can experience fading effects when high intensity and high repetition sources are employed, leading to photochemical phenomena in the molecular structure of fluorophores [29]; fluorescence intensity can also be enhanced under specific conditions of laser irradiation (pulse duration and energy) and different damaging mechanisms within the sample [30]. The potential laser-induced damage for NLOMs on paints has been examined for the first time by Liang et al [31], but extensive analysis of artistic materials has not been carried out so far. Such extensive analysis is particularly important when considering nonlinear optical interactions, since the amount of energy deposited in the material does not have a simple linear dependence on the fluence, as in linear optical interactions, but it has a more complex dependence on parameters describing pulsed laser emission, such as repetition rate, pulse duration, and average power. For ease of comparison with respect to other studies using different experimental setups and laser sources, we always referred to the peak power, rather than to the fluence, as the energy deposited in the sample has a defined dependence (i.e. quadratic for two-photon processes) on the peak power.

Recently, we have defined the laser power threshold for a set of modern acrylic-painted samples as a function of their chemical composition and optical properties. Repetitive laser irradiations were performed and the photodamage induced by TPF was monitored to set the laser power range for safe operation [32]. Starting from our experience, this work aims to determine the optimal experimental conditions for TPF imaging without causing photodamage on a representative set of six acrylic paints. In particular, we performed long-term continuous imaging of the different paints with increasing laser power to set safe exposure time intervals. We defined the safe working conditions on acrylic-painted artworks using parameters such as peak power and irradiation time to provide a reference potentially useful for the researchers employing different NLOM setups.

The samples analysed in this work, as described in our previous paper [32], are made of six commercial extra-fine acrylic paints (Maimeri Brera, IT) laid with a thickness of around 40–50 μm on glass coverslips. The acronyms used throughout the text are the following: primary red magenta (PRM), lemon yellow (LY), cadmium yellow (CY), permanent green light (PGL), permanent blue light (PBL) and zinc white (ZW). The six samples and their corresponding microscopy images are shown in figure 1, whereas the colour indexes and chemical composition, as declared by the manufacturer, are reported elsewhere [32].

Figure 1. Six acrylic paints on glass coverslips (top) and corresponding optical microscope (OM) images (bottom). Scale bar: 300 μm. PRM—primary red magenta, CY—cadmium yellow, LY—lemon yellow, PGL—permanent green light, PBL—permanent blue light, ZW—zinc white.

Download figure:

The choice of a glass substrate for the acrylic application, although differing from the material traditionally used in paintings as support (canvas or wood), does not affect the analysis, being the excitation volume in TPF imaging spatially confined within the paint layer.

The NLOM, a custom-made laser-scanning nonlinear microscope developed at the Istituto Nazionale di Ottica—Consiglio Nazionale delle Ricerche (CNR—INO), has already been described in a previous article [33]. It employs several contrast mechanisms, including TPF, the only technique used for this study.

The light source is an OPO (optical parametric oscillator) pumped by a doubled Yb-based pulsed laser with 80 MHz repetition rate (Chameleon Discovery, Coherent, Santa Clara, CA, US), emitting ultrashort pulses (100 fs at the output) in the 680–1300 nm range; in this experiment, the excitation wavelength was set at 800 nm. A Plan-Apochromat 10× objective lens, with a numerical aperture of 0.45 and a working distance of 2.1 mm (Carl Zeiss Microscopy, Jena, Germany), focuses the laser beam on the sample placed on a motorised translator (M687-XY, Physik Instrumente, Karlsruhe, Germany). In this configuration, the laser spot on the sample has a FWHM of 1.0 µm on the radial plane, and 6.2 µm on the axial direction. Fast laser beam scanning is guaranteed by two galvanometric mirrors (Cambridge Technology, Bedford, MA, USA). A piezoelectric system (P-725KHDS PIFOC, Physik Instrumente, Karlsruhe, Germany), placed behind the objective lens in tandem with a manual actuator, allows both gross axial movements for focusing and fine axial scanning. The signal emitted by the sample is collected by the objective lens, filtered by a long-pass dichroic filter (FF665-Di02-25 × 36, Semrock Inc., New York, NY, USA), and detected by a photomultiplier tube H7422-40 (Hamamatsu, Hamamatsu City, Japan). Data acquisition is controlled via a custom software developed in LabVIEW 2015 environment (National Instruments, Austin, TX, US).

For a single laser pulse, the higher the instantaneous peak power (PP), the greater the probability of 2 (or more) photon absorption. Given the average laser power (PL) at the output of the microscope objective, the laser repetition rate (rL) and pulse duration (δ), this parameter is defined as:

In particular, the intensity of two-photon emission increases proportionally to the inverse of laser pulse duration, 1/δ [34]. When the laser beam passes through optical elements—e.g. lenses—the pulse duration increases according to the following equation [35]:

where δ0 and δ are the initial (100 fs) and final pulse durations, respectively, and ϕ2 is the group-delay dispersion (GDD, i.e. the product of the group-velocity dispersion—GVD—and the length of the dielectric material). In our experimental setup, we estimated a final pulse duration of ∼460 fs by considering all elements and their known GVD/GDD: 4 N-BK7 lenses (overall GDD = 1390 fs2), 14 mirrors (12210 fs2), 2 half wave-plates and 1 quarter wave-plate (240 fs2), 1 Glan-Taylor polarizer (560 fs2), 1 beamsplitter (50 fs2), 1 dichroic (80 fs2) and the objective (1550 fs2), resulting in a total GDD of 16 080 fs2. The obtained δ value was used to compute peak power values during laser irradiation.

Hereafter, both the average and peak laser powers are reported for a more comprehensive description of our experimental conditions and for ease of comparison with other experimental setups and conditions.

Optical microscopy (OM) images have been collected, before and after laser irradiation, using a Leica DMRXE upright wide-field microscope (objective Leica HCX PL FLUOTAR L 20×, NA 0.45), enabling fluorescence, bright-field and dark-field imaging, equipped with a Nikon D3200 (CMOS—APS-C sensor, 24.2 Megapixel resolution, Nikkor AF-S DX 18–55 mm f/3.5–5.6 lens).

This study aims to define the laser time-exposure damaging thresholds for the paint materials. The excitation source was tuned to 800 nm (Ti:Sapphire peak emission wavelength), as this wavelength can effectively excite fluorescence in all samples. The irradiated areas on each sample were marked by laser ablation, allowing for easy localization with both the optical and nonlinear microscopes. For each measurement, 550 images were consecutively acquired on the same area (FOV = 100 × 100 μm2 and resolution = 256 × 256 pixels) for an overall acquisition time of ∼10 min corresponding to 440 s of effective laser exposure, due to the shutter closing/opening dead time. This acquisition time is suited for measurement on CH materials based on our previous experience. Several laser power values (with a minimum increasing step of 0.25 mW) were tested on each sample ranging from 4 mW to 58 mW.

Finally, a wider image (FOV of 500 × 500 μm2) centred on the same area was acquired to evaluate better the damage in the central portion of the FOV. All irradiated areas were equally distanced from each other by about 1 mm to avoid unwanted multiple laser exposures.

To quantify the maximum exposure time before damaging the sample, for each laser power, we measured the time at which the fluorescent properties of the sample started being altered. To this aim, the mean fluorescence intensity (Ii) and its standard deviation (σi) were computed over each ith image, where index i ranges from 1 to 550. We defined an area as damaged when Ii differed from I0 (intensity value before irradiation) more than the standard deviation of Ii. We computed the relative variation between I0 and Ii, i.e. the quantity (I0—Ii)/I0. Despite slight differences from sample to sample, the average variation resulted in ∼ 7%. Its minimum value (6.8% for PBL) was the threshold for safe measurement conditions.

Therefore, for each laser power, the time threshold is defined as the product of the number of images sequentially acquired before reaching the 6.8% intensity variation and the effective exposure time of a single image.

To define safe exposure time intervals on paint materials, we performed long-term continuous imaging of the acrylic samples with increasing laser power. We defined the safe working conditions through the peak power and irradiation time to provide parameters that can be specified with different NLOM setups.

Figure 2 shows the results obtained on the CY acrylic sample, irradiated with laser power values in the range 6–17 mW, with 1 mW step for 10 min (440 s effective exposure). We report a selection of the TPF and OM large FOVs (figure 2, columns). The upper and middle rows display the TPF images acquired before and after exposure, respectively, whilst the corresponding OM images after irradiation are shown in the bottom row. Laser power of 6 and 8 mW did not cause any alteration. When increasing power to 10 and 12 mW, a slight decrease in the fluorescence intensity appears. This effect becomes more evident for laser powers of 14 and 17 mW. The damage is also visible in the corresponding OM images, where a darker yellow square fades in. The effect of damage is more evident in the TPF images, making it the best-suited technique for monitoring any superficial alteration, consistent with our previous work [32].

Figure 2. CY acrylic sample. TPF images before (top row) and after irradiation (middle row) with laser power values of 6, 8, 10, 12, 14 and 17 mW; OM images (bottom row) acquired after laser irradiation. FOV = 500 × 500 μm2 and scale-bar = 100 μm.

Download figure:

All the acrylics showed a fading in TPF intensity at increasing laser powers with the only exception of PBL and PRM, which are characterized by an initial enhancement of the fluorescence intensity, followed by a progressive darkening. According to the literature, the enhancement could be explained by the occurrence of redox reactions, leading to the formation of new fluorescent products with a subsequent signal enhancement [30]. In general, this finding indicates that the observed damages could be related to the chemical composition of the pigments rather than the acrylic binder, or more likely to an altered interaction of the pigments with the acrylic binder.

The results of the maximum safe irradiation time per laser power are shown in figure 3 for all samples and all values are reported in the Supplementary Material (table SM1-6).

Figure 3. Plots of the laser power (left) and peak power (right) as a function of the maximum safe exposure time for the six acrylic paints, PGL (a), PBL (b), CY (c), LY (d), PRM (e) and ZW (f). The coloured areas correspond to the safe conditions for TPF measurements.

Download figure:

The coloured areas define the safe measurement conditions. PGL, PBL, and CY show a similar trend (figures 3(a)–(c): for laser power below 6.5 mW, 8.25 mW, and 9 mW, respectively, the maximum time exposure (440 s) is safe; with a power increase of 0.25 mW for PGL and PBL, and 1 mW for CY, the time threshold nearly halves, showing slight signs of alteration. With further power increases of 1 mW (PGL and PBL) and 2 mW (CY), the time threshold is reduced by one order of magnitude, corresponding to an exposure time of around 50 s. For LY and PRM (figures 3(d) and (e)), power increases of 0.5 mW from 7.5 mW and 18 mW, respectively, cause a small reduction in safe exposure time (415 s and 435 s, respectively). Then, an additional 0.5 mW increase determines a halving of the exposure time for LY and a sharp time reduction by a factor of 20 for PRM. On the left side of each panel, the decrease in exposure time progressively becomes smaller and smaller with higher powers. Finally, ZW (figure 3(f)) shows a stronger resistance to damage: the exposure time decreases more slowly compared to the other paints and therefore can be irradiated with higher power (up to 58 mW for ∼ 40 s).

In figure 4, the laser power and peak power (left and right y-axis, respectively) are reported as a function of the exposure time in a logarithmic scale for all the acrylic paints. The plot displays the safe working conditions: for PGL and PBL (green and blue dots), the exposure time can be arbitrarily increased (up to 440 s), provided that the laser power is lower than 10 mW (200 W peak power). For LY (yellow squares), below 20 s exposure time, the laser power can be increased up to 40 mW for an exposure time of 4 s; above 20 s exposure time, a laser power lower than 10 mW is safe. CY (orange squares) has a smoothest trend compared to LY: below 20 s exposure time, the laser power can be increased up to 15 mW for an exposure time of 10 s; above 20 s exposure time, a laser power lower than 17.5 mW is safe. PRM (red squares) has a similar trend with an exposure time cut-off of 10 s. ZW (white squares) has a completely different behaviour since the exposure time is significantly higher than for all other samples. In particular, a power greater than 30 mW (about 850 W peak power) can be used for 440 s without inducing any damage. For shorter exposure time, the average power can be increased up to 57 mW at 40 s exposure time, meaning that zinc white has a significantly reduced absorption compared to other pigments, consistently with its reflective properties.

Figure 4. Plots of the laser power (left y axis) and peak power (right y axis) as a function of the maximum safe exposure time in a logarithmic scale for the six acrylic paints.

Download figure:

A double-slope trend can be observed from the examination of the six graphs in figure 3, suggesting the occurrence of two distinct effects: one dominating at high laser power levels and short exposure times and the other at lower levels and longer exposure times. According to the existing literature [36, 37], these effects are ascribable, respectively, to thermal and photochemical phenomena. Specifically, Suzuki et al [36] report that these two mechanisms are responsible for the progressive darkening of cinnabar (HgS), a semiconductor pigment like CY sample (CdS). Other studies suggest that these effects could be distinguished as laser-induced photooxidation [38] and laser-induced heating [39]. The chemical composition of the pigments plays an essential role in both photochemical and thermal processes. For instance, lower purity of the pigment may increase laser-induced heating effects [40]. In our case, the relative dominance of one phenomenon over the other at each power level could be influenced by the pigment's chemical properties [36]. The nature of the pigments, whether inorganic (CY and ZW), organic (PRM and LY) or mixed (PGL and PBL), may also affect their behaviour under laser exposure. Further investigation with thermal- and molecular-sensitive imaging techniques would be necessary to fully understand the chemical and physical changes induced to the paints by pulsed laser irradiation; however, such goal is beyond the scope of the present work.

In this study, we introduce a new methodology for determining exposure time thresholds for TPF imaging of painted artworks. We monitored the alterations in fluorescence intensity of six different acrylic paints while increasing the acquisition time for several laser power values. Specifically, we measured the laser-induced fluctuations in fluorescence intensity, namely fading or enhancement of the emitted fluorescence, with a maximum exposure time of 440 s, based on our previous experience with paint materials. For each specimen, we determined the maximum safe exposure time as a function of both the average and the peak laser power. Our main outcome is that the time threshold decreases following a specific trend for each acrylic paint, entailing that the nonlinear order of the optical interaction causing damage is peculiar for each sample. This proof-of-concept study provides valuable insights for optimizing the signal-to-noise ratio (SNR) and the imaging depth without inducing any damage when using TPF microscopy to image acrylic paintings. Although these results are limited to a specific excitation wavelength (800 nm), this methodology offers the potential to be replicated in future studies with different excitation wavelengths and/or contrast mechanisms. The data obtained will contribute to establishing safe protocols for the application of NLOMs on real artworks, thereby enabling the full potential of these promising techniques to be effectively harnessed in non-invasive analytical studies.

This work has received funding from European Union—Next Generation EU in the framework of the research programs: PRIN 2022, Project 2022Y9YP9C (ALIAS) and PNRR H2IOSC (Humanities and Cultural Heritage Italian Open Science Cloud) Project (IR0000029), CUP_B63C22000730005 and from Regione Toscana (PORFSE2014–2020, 'Giovanisì', Intervention Program 'CNR4C', CUP B15J19001040004). The contents reflect only the authors' view, and the European Commission is not responsible for any use that may be made of the information it contains.

The data cannot be made publicly available upon publication because no suitable repository exists for hosting data in this field of study. The data that support the findings of this study are available upon reasonable request from the authors.

The authors declare no conflicts of interest.