Non-Invasive Evaluation of Polymeric Protective Coatings for Metal Surfaces of Cultural Heritage Objects: Comparison of Optical and Electromagnetic Methods

Featured Application: Non-invasive measurement of protective coatings applied to metallic surfaces for monitoring their efficacy in the conservation of cultural heritage. Abstract: The application of protective coatings is an effective preventive strategy to avoid metal corrosion. Constant monitoring of the coating’s quality is fundamental for the successful preservation of the metallic objects by reducing their interaction with corroding agents. Their evaluation over time helps to identify failure at early stages and promote their removal and substitution. Several methods have been employed for coating evaluation (i.e., chemical analysis, thickness and homogeneity investigation). In this paper, we compare three methods—Optical Coherence Tomography (OCT), Confocal Raman Microspectroscopy (CRM), and Eddy Currents (ECs)—to evaluate thickness values and coating integrity. The results from the two optical techniques (CRM and OCT) agree, being able to detect the inhomogeneity of the layer on a micron scale but requiring correction to account for the refraction phenomenon. The Eddy Current is a fast and efﬁcient method for thickness estimation, providing data with millimetric lateral resolution.


Introduction
Metal corrosion has an economic and safety impact in several fields, such as automotive and oil industries and civil engineering [1]. In cultural heritage artifacts, corrosion products developed over centuries represent part of their history; consequently, they are generally preserved or partially removed. However, active corrosion phenomena can damage the objects, particularly those made of iron and copper alloys [2][3][4]. In this scenario, conservators implement protection and inhibition treatments and strategies to avoid further active degradation. For centuries, the application of protective coatings (e.g., resins, natural and synthetic waxes, cellulose derivatives, or acrylic-based polymers) has been an effective method for this purpose [2,3]. While not completely isolating the metallic surface [5], they can slow down the corrosion phenomena by reducing access to oxidizing agents and other pollutants [6]. A limited number of materials are suitable for application on metal-based cultural heritage objects because of the reversibility and stability characteristics required by conservation standards [2,4,7]. Evaluation of the electrochemical properties, which generally requires direct contact between the probe and the object surface, is a relatively common approach in the metal conservation field because of its ease of use and portability. Eddy Currents (ECs) [10] allow for thickness estimation through sample conductivity (σ) measurement. This is performed by detecting the impedance variation induced by the secondary magnetic field produced by the sample as a result of the Eddy Currents [11,12]. Recently, innovative EC set-ups have been developed with more accurate thickness measurements, sometimes achieved through a contactless modality [13]. Electrochemical Impedance Spectroscopy (EIS), on the other hand, aids in inspecting protective coatings applied to different metallic objects by evaluating their barrier properties [2,4]. Despite the valuable complementary information provided, it might operate with limited lateral resolution, long acquisition times, and considerable uncertainty regarding the quantification of coating thickness [14].
In the last few decades, alternative systems to evaluate and monitor protective coatings have been tested. X-ray fluorescence spectroscopy (XRF) has been applied to measure and map the thickness of coating layers on silver artifacts [15,16]. The most recent Evaluation of the electrochemical properties, which generally requires direct contact between the probe and the object surface, is a relatively common approach in the metal conservation field because of its ease of use and portability. Eddy Currents (ECs) [10] allow for thickness estimation through sample conductivity (σ) measurement. This is performed by detecting the impedance variation induced by the secondary magnetic field produced by the sample as a result of the Eddy Currents [11,12]. Recently, innovative EC set-ups have been developed with more accurate thickness measurements, sometimes achieved through a contactless modality [13]. Electrochemical Impedance Spectroscopy (EIS), on the other hand, aids in inspecting protective coatings applied to different metallic objects by evaluating their barrier properties [2,4]. Despite the valuable complementary information provided, it might operate with limited lateral resolution, long acquisition times, and considerable uncertainty regarding the quantification of coating thickness [14].
In the last few decades, alternative systems to evaluate and monitor protective coatings have been tested. X-ray fluorescence spectroscopy (XRF) has been applied to measure and map the thickness of coating layers on silver artifacts [15,16]. The most recent advances in molecular spectroscopy, such as reflectance Fourier-transform infrared (rFTIR) and Near-infrared (NIR), have shown promising results for studying the chemical composition and evaluating the homogeneity of protectives on bronze surfaces [17]. Other optical methods, such as Fiber Optic Reflectance Spectroscopy (FORS) [18], UV-induced visible fluorescence imaging [19], Fluorescence Lifetime Imaging (FLI) [20], ellipsometry [21] and Optical Coherence Tomography (OCT) [14,22], are suitable tools for measuring protective coatings, yet their employment in metal conservation is still limited. Confocal Raman Microspectroscopy (CRM) is another non-destructive tool suitable for coating evaluation (i.e., thickness, homogeneity and distribution of chemical species) [23,24]; however, it is refraction-limited due to spherical aberration and laser refraction, particularly when testing with dry objectives, producing depth resolution broadening [23][24][25]. This inconvenience can be overcome by employing oil-immersion objectives or mathematical corrections [23]. CRM spatial resolution can be improved by coupling it to an Atomic Force Microscope (AFM) [26].
In this work, we compare the potential of three techniques: two portable techniques-Eddy Current (EC), widely used by conservators, and Optical Coherence Tomography (OCT)-and benchtop Confocal Raman Microspectroscopy (CRM), to evaluate film thickness of three of the most commonly used polymers in metal preservation-two acrylic copolymers (Paraloid ® B-72 and Paraloid ® B-44) and one cellulose nitrate lacquer (Zapon ® ). We aim to identify the advantages and disadvantages of each technique along with their complementarity, as well as the critical analytical points that should be considered. Our goal is to offer a methodological comparison that would allow professionals in charge of the preservation of metallic cultural heritage to make informed decisions. The data evidenced each technique's advantages (e.g., portability, precision, and ease of use) and drawbacks (e.g., optical aberrations or poor lateral resolution), providing an overview of critical characteristics. These data are presented with the aim of guiding the formulation of analytical strategies.
We studied nine transparent polymer disks (Table 1) of 13 mm (±1 mm) in diameter obtained by depositing 35 µL of the polymer, dissolved in butyl acetate at different concentrations, over a KBr pellet. After one week of curing, the KBr substrate was dissolved in distilled water, and the polymer disks were left to dry for 24 h under laboratory conditions.

Spectral-Domain Optical Coherence Tomography (SD-OCT)
For each disk, we performed 12 cross-sectional analyses (B-scans), spaced 1 mm apart, with a commercial SD-OCT device (Thorlabs Telesto-II). The device source is a superluminescent diode with a central wavelength of 1300 nm and a bandwidth of about 100 nm. The axial resolution in the air is 5.5 µm, while the lateral resolution is 13 µm. The detector consists of a spectrograph made of a diffraction grating and a fast camera. The system is controlled via 64-bit software preinstalled on a high-performance computer. The 3D scanning path probe with an integrated video camera performs high-speed imaging (76 kHz) for rapid volume acquisition and live display. The XZ acquisition field of view (FOV) was 10 × 1.50 mm 2 with a pixel size of 8.94 × 3.55 µm 2 . The data were processed using ThorImage 5.0 and ImageJ software, taking into account the refractive index of each polymer (n PB72 = 1.49, n PB44 = 1.48, n Zapon = 1.54) to calculate the real thickness values.

Confocal Raman Microspectroscopy (CRM)
Confocal Raman Microspectroscopy (CRM) spectra were measured in depth-profiling mode using a Renishaw inVia Raman confocal microscope equipped with a Leica DM2700 optical microscope and a 532 nm excitation source. Three depth profiles along the sample Z-axis for samples were measured by displacing the motorized microscope stage vertically relative to the laser focus with a step of 1-2 µm. The measurements were performed in an extended spectral range of 100-3200 cm −1 using a grating of 1800 L/mm and a thermoelectrically cooled CCD detector (spectral range 400-1060 nm) with a spectral resolution of 1 cm −1 per CCD pixel (functional resolution of 3 cm −1 ). The laser power on the sample was 16 mW, with typical 10 s integration times and 1 accumulation. The data spectra were collected with a 100x long-distance objective (HC PL FLUOTAR NA = 0.75; theoretical spot size 532 = 0.43 µm, FWD = 4.7 mm) and processed with Wire5.5 and OriginPro8.5 software. The measurements were performed following the OCT profiles.
The depth-profile data were obtained by plotting the high intensity of selected bands from the polymers studied as a function of depth and calculating the full width at halfmaximum (FWHM)-nominal depth (∆)-after Gaussian fitting using OriginPro 2018 (Origin-Lab, Northampton, MA, USA) software. We calculated the corrections of nominal depths with Matlab R2021a (MathWorks, Natick, MA, USA) software.

Eddy Current (EC)
The Eddy Current (EC) measurements were performed with a portable coating thickness gauge, Leptoskop 2042 (Karl Deutsch-D), with a dual probe (Fe/NFe 0 • ) of 5 mm in diameter (0-3000 µm Fe, 0-1250 µm NFe). To assess the accuracy of Eddy Current analyses, we used certified calibration films provided with the device. The disks were placed over a non-magnetic and conductive substrate, and 10 measurements were recorded for each sample.
The coating thickness was estimated automatically by the instrument software; it is determined following the DIN EN ISO 2360 standard (for non-conductive layers on nonmagnetic, conductive based materials). For the coating thickness range (coating < 100 µm) studied in this work, the measuring uncertainty after calibration is 1% (±1 µm).

Spectral-Domain Optical Coherence Tomography (SD-OCT)
Three profiles in the x-z axis were measured, and we later quantified the thicknesses at 12 different points for each profile. By way of example, visible images of PB72 30 and Zapon 20 polymer disks placed on graph paper and the relevant OCT B-scans are shown in Figure 2. As Figure 2b shows, the OCT data are capable of illustrating inhomogeneities and defects in the films. The polymers at higher concentrations produced typical lamella and rim dispersion after their deposition. As shown in Figure 2b, the OCT profile recorded in PB-72 30 samples exhibits thicker layers on the edges of the disk (rim) that become thinner in the center. OCT images also allowed for identifying defects in different areas of the film. The samples with lower polymer concentration, e.g., Zapon 20 (Figure 2c,d), show thinner and more homogeneous layers. As was expected, the thickness of the layers studied is proportional to the volume of polymer dissolved in 35 µL of solution, while its homogeneity may be related to the lower viscosity of solutions with low polymer concentration and the longer time required for complete solvent evaporation.
Three profiles in the x-z axis were measured, and we later quantified the thicknesses at 12 different points for each profile. By way of example, visible images of PB7230 and Zapon20 polymer disks placed on graph paper and the relevant OCT B-scans are shown in Figure 2. As Figure 2b shows, the OCT data are capable of illustrating inhomogeneities and defects in the films. The polymers at higher concentrations produced typical lamella and rim dispersion after their deposition. As shown in Figure 2b, the OCT profile recorded in PB-7230 samples exhibits thicker layers on the edges of the disk (rim) that become thinner in the center. OCT images also allowed for identifying defects in different areas of the film. The samples with lower polymer concentration, e.g., Zapon20 (Figure 2c,d), show thinner and more homogeneous layers. As was expected, the thickness of the layers studied is proportional to the volume of polymer dissolved in 35 μL of solution, while its homogeneity may be related to the lower viscosity of solutions with low polymer concentration and the longer time required for complete solvent evaporation.  Table 2 reports the results obtained with the OCT system. By plotting the thickness values as a function of sampling points (Figure 3), it is possible to observe the layer thickness distribution. Such 3D representation enables evaluation of the coating's integrity and, possibly, the quality of application over the metal surface to be conserved.   Table 2 reports the results obtained with the OCT system. By plotting the thickness values as a function of sampling points (Figure 3), it is possible to observe the layer thickness distribution. Such 3D representation enables evaluation of the coating's integrity and, possibly, the quality of application over the metal surface to be conserved.

Benchtop Instrumentation: Confocal Raman Microspectroscopy (CRM)
Before recording the dataset for the depth profiling analysis, the full-range spectra of each polymer were acquired (Figure 4a). Two characteristic and intense bands were selected for each polymer to evaluate thickness measurement reproducibility.   Table 2.

Benchtop Instrumentation: Confocal Raman Microspectroscopy (CRM)
Before recording the dataset for the depth profiling analysis, the full-range spectra of each polymer were acquired (Figure 4a). Two characteristic and intense bands were selected for each polymer to evaluate thickness measurement reproducibility.
For PB-72, the bands at 1725 cm −1 and 1451 cm −1 , assigned respectively to ν(C=O) and δ(CH2) from methylene groups, were selected. To measure PB-44 films, the bands at 1451 cm −1 and 812 cm −1 , δ(CH2) from methylene groups and νsym(C-C) from the main chain, were employed [36]. For Zapon ® samples, thickness was calculated considering the relative intensities of the bands at 1284 cm −1 and 853 cm −1 from νsym(NO2) and ν(NO) [37]. After setting up the experimental conditions, a sequence of spectra was recorded by focusing the laser at different depths inside the coating until the disk had been completely crossed. An example of PB-4425 depth profiling is shown in Figure 4b.
The comparison between the apparent thickness data obtained with the two different selected Raman bands shows a negligible difference (>0.6 μm) (Appendix A). This confirms the possibility of choosing a distinct band in case of signal overlapping in multilayer systems (e.g., Paraloid ® B-72 and Paraloid ® B-44 both having an intense band at 1451 cm −1 ). In the following discussion, we report only the results obtained with the most intense band, since they showed a lower standard deviation.  Table 2.  The CRM results agree with the OCT findings, showing thickness variability in the different points studied, due to the samples' inhomogeneity, and correlation between thickness and polymer concentration ( Figure 5). Some differences in thickness measurement were detected due to optical aberrations produced by the refraction of the laser inside the polymeric film. For PB-72, the bands at 1725 cm −1 and 1451 cm −1 , assigned respectively to ν(C=O) and δ(CH 2 ) from methylene groups, were selected. To measure PB-44 films, the bands at 1451 cm −1 and 812 cm −1 , δ(CH 2 ) from methylene groups and ν sym (C-C) from the main chain, were employed [36]. For Zapon ® samples, thickness was calculated considering the relative intensities of the bands at 1284 cm −1 and 853 cm −1 from ν sym (NO 2 ) and ν(NO) [37]. After setting up the experimental conditions, a sequence of spectra was recorded by focusing the laser at different depths inside the coating until the disk had been completely crossed. An example of PB-44 25 depth profiling is shown in Figure 4b.
The comparison between the apparent thickness data obtained with the two different selected Raman bands shows a negligible difference (>0.6 µm) (Appendix A). This confirms the possibility of choosing a distinct band in case of signal overlapping in multilayer systems (e.g., Paraloid ® B-72 and Paraloid ® B-44 both having an intense band at 1451 cm −1 ). In the following discussion, we report only the results obtained with the most intense band, since they showed a lower standard deviation.
The CRM results agree with the OCT findings, showing thickness variability in the different points studied, due to the samples' inhomogeneity, and correlation between thickness and polymer concentration ( Figure 5). Some differences in thickness measurement were detected due to optical aberrations produced by the refraction of the laser inside the polymeric film. The CRM results agree with the OCT findings, showing thickness variability in the different points studied, due to the samples' inhomogeneity, and correlation between thickness and polymer concentration ( Figure 5). Some differences in thickness measurement were detected due to optical aberrations produced by the refraction of the laser inside the polymeric film. To account for the influence of the refraction phenomenon in the thickness measurements, we calculated the corrected thickness ( ̅ ) using Equation (1) [24]: (1) To account for the influence of the refraction phenomenon in the thickness measurements, we calculated the corrected thickness (z) using Equation (1) [24]: (1) the equation considers the true point of focus of the laser (z(m)) and the weighting factor ((NA eff ) 2 ) that accounts for the volume from which emitted photons can be captured by the objective [38]. As proposed by Everall [39], z(m) can be calculated as: where ∆ is the nominal depth (thickness) as obtained from the FWHM measurement from the profile, NA the objective numerical aperture, n the polymer refractive index, and m the normalized radius. On the other hand, (NA eff ) 2 is obtained with Equation (3): Table 3 summarizes the nominal depth (∆) measured for each sample and the corrected thickness (z). As expected, laser refraction has a higher influence when studying thicker layers producing a higher error value (h z ), as calculated with Equation (4) [24]. Yet, CRM provides a good estimate of the thickness of the protective layers, being sensible even to slight variations. Additionally, CRM is capable of obtaining chemical information relevant for the assessment of degradation processes, possible mixtures (e.g., pigmented coatings), or inclusions in the layer [20]. Table 4 summarizes the results obtained with the three techniques. Besides the mean thickness values, for OCT and CRM the minimum and maximum thicknesses are reported to indicate their variability. In the first place, it is necessary to consider the different sizes of the areas analyzed. The EC system employed in this work offers average information from an area of 19.63 mm 2 ( 5 mm), while the optical methods provide information with a microscopic resolution, thus registering variability throughout the film. Despite the variability of the thickness values measured with the three techniques, the results are comparable. It is worth noticing that, in addition to the mean values, OCT and CRM data show the minimum and maximum thickness of the layers at a microscopic scale. Spot CRM results depend on the area analyzed; however, performing a depth-profile mapping could solve this issue, offering a profile comparable to the OCT results.

Discussion: Optical vs. Electromagnetic Methods
The results from the thinner samples (i.e., the ones obtained with lower polymer concentration solutions) showed lower variability for the three techniques due to the homogeneity of the layers. This points to the main disadvantage of the EC: irregularities in the coating are not registered due to the averaging area. Moreover, OCT and CRM offer morphological information about the layer, thus registering the topography of the coating induced by possible irregularities present in the metallic surface; this feature is not registered by EC.
Regarding the presence of corrosion products in the metallic surface, such as passive patina layers, the optical methods are not affected by them. OCT registers changes in the refractive index and, consequently, the corrosion patina does not interfere in the estimation of the polymeric layers. Accordingly, CRM is able to differentiate between the polymeric coatings and the metallic patina/surface. The main advantage of CRM is the possibility to simultaneously characterize the corrosion products present and, in some cases, the possible inclusions (i.e., colorants in artificial patinas) in the coating layers [20]. On the contrary, EC results may be influenced by the presence of corrosion products under the coating layer and even by superficial deposits (i.e., dust). It either includes their thickness in the estimation of the polymeric coating or potentially detects a multilayer system without a specific identification, since it exclusively registers conductivity differences.
In contrast, CRM, as used in this study, is non-portable and only samples that can fit under the microscope can be studied. Additionally, fluorescence emission by the probed material can interfere with the profile measurements. Ultimately, the measurements are highly time consuming. Both OCT and CRM require corrections for the material refraction index. Table 5 summarizes the main features of each technique. The choice of one or the other must be evaluated according to the object's characteristics. The optical methods offer more detailed results by registering micrometric coating features; however, ready-to-use (i.e., not requiring mathematical corrections), fast, and in situ measurements can be performed with ECs.

Conclusions
Protective coatings have a significant role in metallic cultural heritage preservation. The evaluation of layer features, such as thickness estimation, fractures or interstice identification, and the detection of chemical changes, contributes to the monitoring of their performance. The three techniques tested in this work are not interchangeable and each one showed specific advantages and disadvantages, such as the possibility to identify detailed layer features or the identification of chemical composition. The results of the two optical techniques (OCT and CRM) are in reasonable agreement and revealed the layers' inhomogeneity; their image and graphical outputs are easy and fast to interpret but require corrections considering the refractive index of the polymer. In contrast, CRM, as used in this study, is not portable and the object's size and shape may therefore limit its employment. However, the recent advances in portable Raman instrumentation may allow the in situ analysis of coatings using the confocal modality.
The EC measurements agree with the other two techniques when homogeneous layers are tested; however, the main issue with ECs is that the results are averaged over a relatively big area, thus hampering the detectability of microscale inhomogeneities and defects in the polymer coatings. Additionally, corrosion products and deposits over the surface can interfere with thickness estimation.
All in all, compared to electrochemical methods, the optical techniques allow for more features from the polymeric layers to be registered yet require specialized instrumentation. In the future, this issue might be solved by more user-friendly systems being made available to conservators.

Data Availability Statement:
The data presented in this study are available on reasonable request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.