**4. Discussion**

The main results obtained from data analysis are recapped in Table 1. From the FORS and SERS results, different typologies of synthetic dyes were undoubtedly used for the puppets' manufacturing. In some cases, the identification of the present dyes is straightforward: for the D817-2 sample, the apparent absorption shoulder at 490 nm and the maximum at 530 nm are attributable to a xanthene pink dye [11,40]. Among the several colorants belonging to this family, which includes Eosin, Rhodamine, and Phloxine, the SERS spectrum (Figure 5, bottom) allows a clear identification of Rhodamine-based dyes [42,49,50]. Moreover, among the two main dyes of this typology (Rhodamine B and Rhodamine 6G), the acquired spectra present higher similarities with the Rhodamine B spectra reported in the literature [50]. Analogously, the maximum at 640 nm observable in the apparent absorption spectra of the D816-7 sample suggests the presence of a triphenylmethane dye, such as Malachite Green and Patent Blue V [11]. The SERS spectra obtained for this sample confirm these preliminary data, because both their general patterns and the Raman bands are characteristic of this family of dyes. However, it is difficult to identify the specific dye: the experimental spectra, indeed, show great similarities with both Crystal Violet and Malachite Green dyes, both belonging to the same class [44]. In addition, Crystal Violet's SERS spectrum is very similar to the Methyl Violet one; these three dyes should be thus considered for the attribution [54,55]. With reference to the acquired spectra (Figure 6, top), some bands (732, 760 cm−1) suggest the presence of Crystal Violet, while others (1176, 1221, 1364, 1399 cm<sup>−</sup>1) suggest the presence of Malachite Green. Taking into account the FORS spectra, a possible hypothesis consists in the identification of Malachite Green as the principal dye, with minor amounts of Crystal Violet (which, although in a minor quantity, is reported to be more easily detectable through SERS spectroscopy in comparison to Malachite Green) [44].

**Table 1.** Assignment of the dyes for the different samples, with reference to their appearance and their spectral features.


The presence of a triarylmethane dye is also suggested for the D817-4 sample, but, in this case, the identification of the molecule is clear: the SERS spectrum (Figure 5, bottom) presents a fulfilling matching with the SERS spectrum of Crystal Violet (or Methyl Violet), and there is no ambiguity with other compounds of the same class (e.g., fuchsine based dyes) [42,54], whose characteristic peaks are not evident in the spectra. Nevertheless, there is an actual contradiction with the FORS results: the main apparent absorption band of Crystal Violet should be at 594 nm, but it is not visible in the experimental spectrum. According to the literature, the main matching of the apparent absorption spectrum bands at 506 and 540 nm occurs with the spectra of orange mono-azo dyes, for instance, Lithol Fast Scarlet RNP or Orange II [11]. A possible hypothesis to explain this unusual behavior could be formulated with reference to the sampling area. The stubbon was used to sample the

dye from a part of the puppet with an orange-yellow background, and this is in agreement with the bands observed in the FORS spectra. However, this area is decorated with a dark motif. Considering the high SERS detectability of Crystal and Methyl Violet, the minor amount of this dye present in the decoration represents the molecules effectively detected with this technique, while the orange dye (likely a mono-azo) could be not visible for a lower affinity to the silver nanoparticles or for the lower Raman cross-section.

For the remaining samples, the identification of the present dyes is less easy, but some hypotheses can be formulated. The FORS spectra obtained for the D816-4 sample suggest the presence of a xanthene dye: in particular, the apparent absorption bands at 501 and 550 nm present a good matching with the Visible Light absorption of Rhodamine B. Moreover, one of the SERS spectra (Figure 4, top) presents some peaks at 622, 735, 1051, 1364, 1510, and 1645 cm−<sup>1</sup> (observable also in the spectrum of D817-2, with eventual reduced shifts in wavenumbers), which can confirm the presence of this colorant [50]. Other bands at 637, 714, 1329, 1452, 1577, and 1623 cm−<sup>1</sup> could be indicative of Eosin Y [14,43,47,48]. For the D816-9 sample, even if some signals confirm the presence of Eosin Y (479, 1184, 1623 cm<sup>−</sup>1), the lower quality of the SERS spectra, along with the change in intensity in the apparent absorption spectra, does not allow for inferring any plausible hypothesis on the composition of the area. However, two samples derived from close areas of the same color and, thus, a similar composition is expected.

The two last samples, D816-5 and D816-15, represent the most complex ones for the interpretation. The FORS spectra are not really informative because of the low and noisy reflectivity spectra and consequent artifacts in the apparent absorption spectra. However, the SERS spectra (Figure 6) present several affinities, thus allowing to hypothesize a similar composition. With reference to the literature, it is hard to individuate a dye whose reported spectrum completely matches the experimental ones. Moreover, the variation in intensity of some peaks in different spectra suggests that several dyes could be present in mixture. For instance, peaks at 658, 1003, 1030, 1130, 1183, 1209, 1585, and 1627 cm−<sup>1</sup> could be indicative of the presence of Sudan Black B [51,52], even if the relative intensity of bands does not match the literature spectrum of this dye [51,52]. On the other side, peaks observable in some spectra at 570, 733, 1323, 1443, and 1585 cm−<sup>1</sup> could be indicative of the use of carminic acid dye [15,53], even if some characteristic signals are missing. Moreover, if Sudan Black B could represent the main colorant of the sampled black area, the carminic acid could not explain its color. In order to verify and confirm these hypotheses, a separative analytical technique should be used to discriminate the different colorants (for instance, High Performance Liquid Chromatography coupled to Photo-Diode-Array HPLC or High Resolution Mass Spectrometry).

From the analytical point of view, the combination of two laboratory spectroscopic techniques resulted useful for the identification of the present dyes. However, even if the micro-FORS provided preliminary hints, we remark that the acquisition of spectra on sampling stubbons resulted critical. If the acquired spectra resulted highly significant for the identification in some samples (e.g., D816-7, D817-2), the individuation of marker bands was not easy for other ones, especially if the concentration of the sampled dye was not high and the interference from the background—likely due to the fibrous stubbon matrix—was remarkable. For instance, in the case of the D816-5 and D816-15 samples, no main spectral feature is clearly observable. Nonetheless, the general performances obtained for the micro-FORS approach applied to the sampling stubbons suggest that an on-site FORS analysis would have provided remarkable results: the higher concentration of dyes on the objects and the absence of background interference from the stubbon would represent improvement factors for the acquisition of informative Vis-Light reflectance spectra. In this study, it was not possible to acquire FORS spectra in situ, however, its application presents a high potential not only for its non-invasiveness but also for the quality of obtained data. On the other side, the stubbon sampling was fundamental for the vibrational spectra collection. Indeed, the acquisition of on-site Raman spectra would likely not be useful for similar matrices. The portable Raman spectrometers are, in general,

less performative than benchtop instruments, especially when highly fluorescent analytes are present, as in this case. Moreover, as highlighted by the preliminary measurement, the conventional Raman spectra resulted not useful for the identification of the synthetic dyes, and the micro-invasive addition of SERS colloid was fundamental in obtaining high quality vibrational spectra. Even in the case of samples containing Malachite Green and Crystal Violet, whose standard Raman spectra are reported in the literature, it was not possible to detect any clear spectral feature attributable to the dye. Probably, the high fluorescence of these dyes could be increased by the presence of degradation products and impurities from the art object. Moreover, focusing directly on the dye particles on the stubbons was more difficult in comparison to focusing on Ag nanoclusters close to them, because the fibers are more susceptible to move under the lens, with consequent loss of focus. Actually, for similar historical objects where the presence of synthetic dyes is highly probable due to the manufacturing period, a sampling could be necessary, but the high sensitivity of SERS spectroscopy allows for minimizing the number of sampled analytes (or the amount of analyzed sample). In order to maximize the results achievable through an integrated spectroscopic approach, the utilized protocol, enriched by a preliminary on-site FORS analysis, is suggested for similar matrices with synthetic dyes to address the following chromatographic analyses and when it is not possible to sample fragments for the analysis with low sensitivity methodologies.
