1. Introduction

Surface-enhanced Raman scattering (SERS) has been firmly established as a powerful tool for identifying organic colorants in works of art [1–3]. Indeed, because of its molecular specificity, high spatial resolution, and exquisite sensitivity afforded by the electromagnetic and resonance Raman enhancement mechanisms [4–6], SERS is capable of identifying chromophores even down to the single-molecule limit [7–9]. Accordingly, many studies have demonstrated the SERS-based identification of different classes of dyestuffs, dyes, and pigments in actual cultural heritage objects including textiles, paper, and paintings [10–15]. Despite these advances, however, the SERS spectrum of a colorant can be exceedingly sensitive to enhancing substrate as well as other chemical and instrumental variables (e.g., solubility, matrix, pH, excitation wavelength, baseline correction). These variations limit the reproducibility and reliability of SERS as an identification technique and ultimately, its widespread application to cultural heritage research.

To address these challenges, several studies have demonstrated reproducible microanalytical approaches and searchable SERS databases that provide analysts with the ability to identify dye class, and in some cases, the chromophore itself [16–24]. However, the majority of these studies employ water-soluble dyes as the test analyte (e.g., rhodamine 6G, crystal violet, alizarin, carminic acid), which are known to exhibit exceptional SERS signals. Furthermore, significant spectral variability with SERS-active substrate is still observed, both with respect to different synthetic approaches as well as batch-to-batch variations within one protocol [20,22,25]. How, then, can analysts assess the activity of SERS substrates before applying them to a precious art sample? The focus of this study is to develop a quality assurance (QA) protocol to test whether SERS substrates are capable of producing high-quality spectra and ultimately, provide for the accurate and reproducible identification of unknown colorants in art.

In this study, we examine the most widely used SERS substrate in cultural heritage research: silver colloids prepared according the Lee and Meisel method [26]. Silver colloids are relatively easy to prepare from low-cost reagents and produce highly-enhancing nanostructures that are capable of elucidating the SERS spectrum of a single molecule [27]. However, the use of silver colloids also introduces several potential drawbacks such as nanoparticle polydispersity, batch-to-batch variability in colloid production, spurious signal from the citrate capping agent, as well as susceptibility of the nanoparticles to oxidation, dissolution, aggregation, and agglomeration [25,28–30]. Here, we examine the spectral variations produced by different batches of silver colloids that are prepared under the same experimental conditions. Although different batches of colloids exhibit significant variations in SERS intensity, alizarin is successfully identified in all trials. However, by increasing the complexity of the analyte and its matrix (i.e., by moving from dye to pigments and paints), we find that some batches of silver colloids are capable of identifying red organic unknowns and others are not. These results are integrated into an overall QA protocol that enables an analyst in the museum setting to effectively test substrate activity prior to applying the colloids to art.

2. Experimental

Alizarin (Acros Organics, 97%), HCl, HNO₃, and MeOH (Fisher) were used as received. Madder lake, carmine naccarat, charcoal black, flake white, and linseed oil were obtained from Kremer Pigments and used to prepare reference paints [11,18]. Glassware was cleaned with aqua regia prior to use. Silver nitrate (Strem, 99.9995%) and sodium citrate (Alfa Aesar) were used for colloid synthesis [26]. The colloids were centrifuged (Eppendorf, 1 mL aliquots with ~0.93 mL of supernatant removed) for two cycles at 12,000 g for 15 min per cycle. Microscopic samples were obtained using surgical blades (Feather Safety Co.), placed on glass coverslips, and coated in 0.75-μL colloids either before or after sample treatment with 0.75 μL of a 1:2 mixture of 1 M HCl in MeOH [16,18]. SERS studies were performed on an inverted microscope equipped with a 632.8-nm laser described in detail elsewhere [11,16–18,31]. We apply two criteria for the positive identification of pigments: (1) the detection of at least two peaks associated with carmine (i.e., ~1300 and 459 cm⁻¹) and seven peaks for madder (i.e., 1610, 1543, 1418, 1326, 1296, 1162, 343 cm⁻¹), with intensities greater than 2 times the standard deviation of the background [19] and (2) relative SERS intensities ($I_{SERS}$) are consistent with reference samples of madder (i.e., $I_{SERS}$ = $4\pm 2$:$4\pm 2$:1 for peaks at 1418, 1296, and 1610 cm⁻¹, respectively) and carmine (i.e., $I_{SERS}\ge$ 3:1 for peaks at 1300 and 459 cm-1). To account for intensity fluctuations due to random variations in pigment/paint samples and nanoparticle concentration (i.e., as a result of nonstoichiometric synthesis, supernatant removal after centrifugation, solvent evaporation of the colloid spot, etc.), SERS measurements were performed in triplicate and by different analysts.

3. Results and Discussion

To examine batch-to-batch variability in colloid production and corresponding SERS signal, we synthesized 149 batches of silver colloids using the same Lee and Meisel experimental procedure. Batches were prepared by 10 different researchers in the same laboratory conditions and stored in the dark. The resulting colloids exhibit significant variations in terms of color (i.e., brown, white, green-yellow) and corresponding UV-vis spectra, with absorption maxima (${\lambda }_{max})$ that range from 398 nm to 425 nm. No correlation between UV-vis spectra and activity was detected. The average SERS intensity of adsorbed citrate on the colloids (i.e., $I_{SERS}$ at ~1400 cm⁻¹) [32,33] is 9000 $\pm$ 5000 ADU mW⁻¹ s⁻¹, with individual values that vary by over two orders of magnitude among different batches. These variations highlight the need for a reliable method to ensure colloid quality and SERS activity before their application to a precious cultural heritage object. As a first step toward QA, we performed SERS measurements of alizarin, one of the most important dyes in cultural heritage research [13,19,34,35]. Consistent with previous studies, the SERS spectra of 10⁻⁴ M alizarin in ethanol exhibit major peaks at 1599, 1413, 1327, 1295, 1160, and 342 cm⁻¹ [19,21,31]. Although batch-to-batch variations in SERS intensity are observed, alizarin is successfully identified in all trials, and in some cases, even after the colloids have been stored for over a year (Fig. 1). Figure 2 summarizes the performance of ten representative batches of colloids prepared by the same researcher. This observation demonstrates the activity and stability of silver colloids for a highly-enhancing dye like alizarin, but it is unclear whether this activity translates to unknown colorants in art.

 

Fig. 1 (left) SERS spectra of alizarin obtained using the same colloids (a) freshly synthesized and (b) ~1 year after synthesis. (right) SERS spectra of reference red lake pigments obtained using various colloid batches. (c) carmine lake with batch J, (d) carmine lake with batch A, (e) madder lake with batch G, (f) madder lake with batch D, and (g) blank colloids.

Download Full Size | PDF

 

Fig. 2 Performance of various colloid batches (A - J) to QA tests.

Download Full Size | PDF

To address this concern, we tested the colloids using more complex samples that are commonly encountered in painting conservation (i.e., pigments and paints). Previous studies have demonstrated that silver colloids can be directly applied to paint samples in order to successfully identify carmine, the red lake pigment of carminic acid, using extractionless non-hydrolysis SERS [11,14,31,34,35]. Therefore, testing the colloids with carmine represents a sensible next step in terms of increased sample complexity and reported reproducibility [11]. Figure 1 presents SERS spectra of carmine pigment obtained using two different colloid batches. In one case (batch J) the SERS spectrum of carmine demonstrates significant peaks at 1483 (m), 1298 (s), 459 (m), and 431 (m) cm⁻¹ [11,19,35,36]. However, the SERS spectrum of carmine from batch A is relatively low intensity (e.g., $8.2\times {10}^3$ and 90 ADU mW⁻¹ s⁻¹ at ~460 cm⁻¹ for batches J and A, respectively) and obscured by peaks due to adsorbed citrate (e.g., at 1484 (m), 1397 (s), 1026 (m), and 942 (m) cm⁻¹) [29,37]. Therefore, carmine is not unambiguously identified using batch A colloids and the direct application of these colloids onto an art sample is not recommended. Ultimately, two trials (batches A and B) do not pass the criteria for carmine pigment and corresponding paint, suggesting that the other batches may be effective for unknown identification in art. To test this hypothesis, we examined the efficacy of the colloids for identifying analytes of increased difficulty: madder lake pigment and paint.

Although some studies have demonstrated extractionless non-hydrolysis SERS of madder lake pigment and paint using silver colloids [31,38], others have made a strong case for pretreating madder-containing samples [19,23]. Therefore, relative to carmine, madder lake represents a more challenging analyte for colloid testing. Figure 1 presents SERS spectra of madder lake obtained using two different colloid batches that passed the alizarin, carmine pigment, and carmine paint tests. The colloids from batch G produce a SERS spectrum of madder with major peaks at 1610 (m), 1543 (m), 1418 (s), 1326 (s), 1296 (s), 1162 (m), and 343 (m) cm⁻¹ [19,21,31]. However, the SERS spectrum of madder from batch D colloids is dominated by signal from adsorbed citrate (i.e., at 1484, 1398, 1026, and 946, and 797 cm⁻¹) and does not pass the QA test. Therefore, although testing with carmine suggests that colloids C through J are effective for direct analysis of art, further testing with madder reveals this is not the case. Figure 2 summarizes the results of testing colloid batches A through J with alizarin, carmine pigment, carmine paint, madder pigment, and madder paint. Implementing these QA tests demonstrates that just 20% of colloid batches are viable for direct application to unknown samples from art.

Fortunately, several SERS studies have shown that sample pretreatment using extraction and/or hydrolysis can improve reproducibility [16–18,21–24]. Therefore, as a final step after QA testing, we subjected reference pigment and paint samples to an established pretreatment approach based on HCl and MeOH [16,18]. Colloid batches that passed the QA tests for alizarin, carmine pigment, and carmine paint, but had failed for madder lake pigment and/or paint, could be successfully identified following treatment. However, if colloid batches failed the QA tests for carmine, sample treatment did not provide for colorant identification. Thus, QA testing with carmine represents a critical first step to establish intrinsic colloid activity before analysts attempt to identify unknowns. These results also suggest that this QA protocol (Fig. 3), when coupled to an effective sample pretreatment method, will provide for the reproducible identification of a wide variety of unknown colorants. Indeed, we examined disperse samples from the floral dress details of Portrait of Queen Elizabeth I, which demonstrate the physical characteristics typical of organic lake pigments. First, we synthesized a new batch of silver colloids, which passed the QA test for carmine pigment (Step 1), but failed for reference madder paint. Therefore, following the QA protocol, we pretreated the paint and retested the colloids. Finally, disperse art samples were treated with HCl/MeOH before applying the colloids for SERS analysis. Figure 3 presents the resulting SERS spectra of three disperse art samples following QA testing and sample pretreatment. In all cases, major SERS peaks at ~1483, 1300, and 462 cm⁻¹ are observed, with corresponding intensities that meet the criteria for identifying carmine. These results demonstrate that our QA protocol is effective for testing substrate activity before applying the colloids to art.

 

Fig. 3 (left) Schematic of QA protocol. (right) SERS spectra of samples from Portrait of Queen Elizabeth I (1533-1603), British, 1590-1600, oil on canvas transferred from wood, accession #1945-20. The Colonial Williamsburg Foundation, Gift of Mr. Preston Davie.

Download Full Size | PDF

4. Conclusion

This study demonstrates that significant batch-to-batch variations (i.e., in UV-vis spectra and $I_{SERS}$ of adsorbed citrate) are commonly observed for silver colloids prepared using the same protocol. Furthermore, SERS measurements of alizarin do not provide a sufficient means to establish the utility of silver colloids for art analysis. However, a simple QA protocol using carmine and madder can establish colloid activity before a precious art sample is consumed. These observations motivate further studies aimed at understanding why some colloid batches do not produce high-quality SERS spectra of unknown pigments and paints (even after sample treatment) and perhaps more serious consideration of alternative nanostructures that exhibit higher synthetic fidelity. Current studies are underway to elucidate these fundamental structure-activity relationships. Ultimately, we believe that the combination of this QA procedure to test intrinsic SERS activity with sample treatment strategies to enhance dye-nanoparticle interactions will significantly enhance the reproducibility and applicability of SERS in cultural heritage research.