“ Welcome to the jungle ” : TLC-SERS to wade through real complex mixtures of synthetic dyes

The finding of numerous synthetic strategies between the 19th and 20th century fostered the production of an extensive spectrum of organic dyes for artistic and industrial applications. Shortly after, challenges emerged related to the rapid deterioration of these products. Nowadays, the characterization of coloured pieces of art produced in the first half of the XXth century is further complicated by the absence of reference materials with an unambiguous composition, which are frequently a mixture between the starting dyes and their degradations products. Here we demonstrate as Surface Enhanced Raman Spectroscopy coupled with the simplest separation technique, i.e. Thin Layer Chromatography, is able beyond to recognize several synthetic dyes also to identify their isomers and degradation products. The occasion was the study of a catalogue of dyed textile swatches, produced in 1951 by the well-known Italian industry ACNA (Azienda Coloranti Nazionali e Affini). An apparently easy task turned out to be, indeed, not at all obvious as we found out that the actual composition of the reference samples seldomly corresponded to the declared one. TLC-SERS investigation of historical textile swatches was complemented by the integration of liquid chromatography coupled with a diode array detector and high-resolution mass spectrometry (HPLC-DAD-HRMS). This approach allowed us to validate the findings of TLC-SERS and to distinctively characterise synthetic by-products and degradation markers resulting from the deterioration of the dyes.


Introduction
Since the late 19th century, the colour palette available to artists and industry was confined to natural pigments and dyes extractable from minerals, and plants or animals.At the end of 1800, we witnessed a transformative surge in the chemical industry, ushering in a new era for colours.Initially, chemical reagents, such as sulphuric and acetic acid, were incorporated to selectively modulate the extraction process of natural dyes from the raw materials (e.g.modified madder-based products [1]), or to modify the physical properties of the resulting natural dye-based materials (e.g.unoxidised iron gall ink [2]).Chemically modified natural dyes were subsequently supplanted by synthetic dyes, which offered superior performances in terms of ease of use and reproducibility of the obtained hues, and economic advantages.This transition promoted the production of novel colouring molecules, allowing the development of synthetic organic pigments (SOPs), but also of synthetic analogues of natural dyes [3].Given that synthetic dyes and pigments often result from mixing multiple dyes without transparently declaring their compositions, possibly due to patent controversies, the role of analytical chemistry in studying their real nature is pivotal.The ability to unveil the individual components through advanced analytical methods is indeed the cornerstone of a rational conservation strategy.
Typically, the chemical analysis of dyes or pigments entails the application of advanced chromatographic and spectroscopic techniques, given the complexity and the small size of the samples.
The characterisation of synthetic dyes in historic sample is a particularly challenging task, considering the extension of synthesis byproducts and dyes decomposition (due to photo-ageing and/or oxidation, occurring over time or even starting at the moment of their synthesis or application [4]) as well as the lack of information by producers.A reliable database could be built starting from the catalogues released in the past by dye industries, however it has been demonstrated by Tamburini that frequently a dye's label does not correspond to its actual composition, as documented for organic pigment-lakes [5][6][7].
In a recent work Tamburini determined by HPLC-MS 2 the composition of numerous reference dyes from a catalogue published in Germany in 1896 [5].The same analytical technique was exploited by Serafini et al. to investigate 15 samples of azo dyes produced by the Italian industry "Azienda Coloranti Nazionali e Affini" (ACNA) [8].Successively the same research group extended their analytical toolset for the characterization of these dyes to Raman spectroscopy [9].However, in this context Raman comes with certain limitations, such as scarce sensitivity, potential interferences from multiple overlapped signals, and issues related to fluorescence of both dyes and the textile substrates.Surface-Enhanced Raman Spectroscopy (SERS) has emerged as a notable alternative, and already proved its efficacy in investigating natural and synthetic dyes [10,11].Here, we coupled SERS to Thin Layer Chromatography (TLC-SERS) to separate and identify the multiple components of selected reference samples from a catalogue of basic dyes on textile fibres by ACNA (Fig. 1).TLC-SERS emerged in 2009 to separate and characterize a mixture of several artists' dyes [12][13][14].In our work, samples were selected based on the reported label on the catalogue, focusing on some of the most used and some of the less studied dyes.TLC-SERS proved its efficiency in separating complex mixture of dyes, thus confirming that the label used never referred to a single molecule, but mostly to a poorly purified product.On the one hand, this highlights the importance of being cautious in considering historical textile swatches as reliable reference materials; on the other hand, it confirms how the composition of each dye is highly specific for its manufacturer, and thus useful for dating and provenance assessment.Most importantly, the study of the catalogue gave us the chance to identify for the first time by TLC-SERS several dyes decomposition products.The obtained results were further checked by HPLC-DAD-HRMS analysis, confirming the usefulness of the technique for a fast and affordable screening of complex samples.

ACNA reference samples
The analysed samples were selected from the ACNA catalogue "Coloranti Basici (Basic Dyestuffs)", printed in Milano, Italy in 1951 and purchased in 2019 in an antique shop by one of the authors of the manuscript.Chosen samples are listed in Table 1, with the associated ACNA code, name, photos, and results of the chemical investigation.The dyes were applied on silk fibers.

Raman, surface enhanced Raman scattering (SERS) and thin layer chromatography-surface enhanced Raman scattering (TLC-SERS)
A citrate-reduced silver colloid was obtained according to Lee and Meisel [15].Prior to SERS analysis, the dyes were extracted from approx.1-2 mg of textile sample suspended in 50 μL of 0.5 M oxalic acid/methanol/acetone/water (1:30:40:40 v/v/v/v) solution and ultrasonicated for 30 min at 60 • C. For SERS analysis, 5 μL of the colloidal Ag solution (10×) was mixed with 1 μL of sample extract and 1 μL of 2 M KNO 3 (aggregating agent), a drop was then deposited on a microscope glass slide.Spectra were acquired immediately by focusing on the top of the drop.
SERS measurements were made on a Renishaw InVia instrument coupled with an optical Leica DLML microscope, equipped with a NPLAN objective 20× and 50×.The laser source was a NdYAG laser at 532 nm or a HeNe laser at 633 nm, and a laser power output was between 1 and 5 mW.The spectrometer consists of a single grating monochromator (1200 lines mm − 1 ) coupled with a CCD detector, a RenCam 578 × 400 pixels (22 μm × 22 μm) cooled by a Peltier element.
The spectral calibration of the instrument was performed on the 520.7 cm − 1 band of a pure silicon crystal.All SERS spectra were recorded between 350 and 2000 cm − 1 , where the main Raman and SERS signals of dyes are located, with 3 accumulations and exposure times of 10 s.The spectra preprocessing was realized using Spectragryph software, version 1.2.15 [16].Spectra were baseline-corrected, smoothed by Savitzky-Golay filters with five points and a second-order polynomial function, and the spectral intensities were normalised between 0 and 1.
For TLC-SERS, the separation took place on Silica gel 60 plates (Supelco) as stationary phase and a mixture ethyl acetate/methanol/ ammonia/water (7:2.2:1:1) as eluent.Two μL of each extract were deposited onto the TLC plate by means of a glass capillary and eluted in a glass developing chamber.The separated spots were visualized under a UV lamp at 254 and 365 nm.The separated components were analysed directly on the TLC plate by placing 1 μL of the colloidal Ag solution (10×) on top of each spot.

High performance liquid chromatography (HPLC-DAD-MS 2 )
The HPLC-DAD system consists of a PU-2089 quaternary pump equipped with a degasser, an AS-950 autosampler and an MD-2010 spectrophotometric diode array detector (DAD) (all modules Jasco International Co., Japan).The DAD detection parameters were: wavelength range 200-650 nm, resolution 4 nm, acquisition rate 0.2 s.The HPLC-ESI-Q-ToF consists of an HPLC 1200 Infinity, coupled to a Jet Stream ESI-Q-ToF 6530 Infinity detector and equipped with an Agilent Infinity autosampler (Agilent Technologies, Palo Alto, CA, USA).High resolution MS and MS/MS spectra were obtained in positive ionisation mode in the acquisition range of 100-1700 m/z, with a scan rate of 1.04 spectra/sec (CID voltage 30 V, collision gas N 2 , purity 99.999%).Further details on MS working conditions can be found in [17].Chromatographic separation was performed in both instrumentation on an analytical reversed-phase column Poroshell 120 EC-C18 (3.0 × 75 mm, particle size 2.7 μm), equipped with a Poroshell 120 EC-C18 (3.0 x 5 mm, particle size 2.7 μm) guard-column.The eluents used as mobile phase were: A = water with 1% (v/v) formic acid (FA); B = acetonitrile with 0.3% (v/v) formic acid (FA).The flow rate was 0.6 mL/min and the program was 5% B for 2.6 min, then to 50% B in 13.0 min, to 70% B in 5.2 min, to 100% B in 6.

Software
ChemDraw Ultra version 12.0 (ChemOffice 2010, CambridgeSoft Corporation) software was used for drawing molecular structures and to calculate their topological polar surface area (tPSA, measured in square angstroms).Polar surface area is defined as a sum of surfaces of polar atoms (usually oxygens, nitrogens and bonded hydrogens) in a molecule.The methodology for the calculation of TPSA is described in details in [18].

Results and discussion
The study of a catalogue produced in 1951 by the well-known industry ACNA to promote their dyes showing their final appearance on textiles (here silk) was the occasion to deeply investigate the composition of some of the most important and some of the less known synthetic materials produced in the 1900s.In the following paragraphs the results of SERS and TLC-SERS analysis, as well as HPLC-DAD and HPLC-HRMS data, are presented for each class of compounds.The intent is to show, beyond the characterization of the historical samples, how TLC-SERS can be exploited to characterize mixtures of different dyes as well as compounds with similar structures (e.g.compounds with various degree of N-methylation).The results achieved are summarised in Table 1, while in the Supporting Information file the list of compounds detected by HPLC-DAD-HRMS (Table S.1) as well as those detected by each analytical technique (Table S.2), and HPLC-DAD (Figs.S1-S3) and HPLC-HRMS (Figs.S4-S5) chromatograms not reported in the main text are provided.Fig. S6 shows the whole picture of the TLC plates analyzed

Table 1
Photo, ACNA code, label (in Italian as catalogued, and translated in brackets), and detected chemical composition of the selected samples.A non-aged silk sample was also extracted and analyzed according to the methods described in Section 2. No silk specific markers were identified through either SERS or HPLC-DAD/MS analysis.

Auramine-based dyes
Auramine (bis(4-dimethylaminophenyl)methyleneiminium chloride, Basic Yellow 2, C.I. 41000) was first patented in 1884 by Caro and Kern [19], and since then it was extensively used for the dyeing of paper, cotton and silk.In the two selected yellow samples (ACNA 61 and 62), catalogued as "Auramina II" and "Auramina O", SERS analysis confirmed the presence of an arylmethane dye (υ(C ring N)/δ s (CH 3 ) at 1543 cm − 1 ; δ as (CH 3 ) at 1478 and 1434 cm − 1 ; υ s (CC center C)/υ(CN) at 773 cm − 1 ; C=C-C out of plane bending at 731 and 646 cm − 1 ) [20].The spectra acquired at 532 nm, shown in Fig. 2, were compared with that of the available analytical standard of Auramine O (CAS 2465-27-2, C.I. 41000), finding similar values of Raman wavenumbers and thus confirming the presence of the basic dyes in both samples [21].In addition to auramine vibrational signature, in the spectra of ACNA 61 and 62 are  also detectable a doublet around 580 cm − 1 , a doublet/triplet peak at 800 cm − 1 not present with the same intensity in the standard, and a very intense doublet around 1200 cm − 1 , which could be ascribed to demethylated forms of auramine.
Analogue results were obtained by HPLC-DAD (Fig. S1), while from TLC-SERS analysis it was not possible to detect any coloured spot (neither under visible, 254 or 365 nm light).HPLC-HRMS in positive ionisation mode (Fig. S4) enabled us to outline results consistent with Tamburini previous findings [5].Auramine O ([M] + = 268.181)was detected along with Michler's ketone ([M] + = 269.165),and their mono-and bis-demethylated derivatives were identified.Michler's ketone (4,4′-bis (dimethylamino)benzophenone) is an intermediate in diand triarylmethane dyes synthesis (employing Michler's ketone, zinc chloride and ammonium chloride) and its presence might be due to an insufficient purification of the product, but its formation in the samples due to hydrolysis or oxidation of the original dyes cannot be excluded [22,23].
While Auramine O can be easily traced back to the known basic dye having C.I. 41000, we were not able to find any commercial reference for Auramine II.The SERS spectra of the two sample are apparently identical, except for a peak at 645 cm − 1 of ACNA-Auramina II (respect to the peak at 648 cm − 1 for both ACNA-Auramina O and standard Auramine O) and an inverted ratio between the peaks at 1543 and 1595 cm − 1 .

Safranine-based dyes
Safranines were first discovered by Perkin as by-products of ).All chromatograms are presented in the same scale and are stacked for purpose of clarity.mauveine production, but the correct nature in terms of chemical structure of these products was not revealed during the 19th century [24].Synthesis of phenazines was first introduced by Felix Duprey in 1865, in France, based on heating commercial aniline dissolved in acetic acid with lead nitrate.A different route was reported by Wohl and Aue in 1901, by heating aniline and nitrobenzene at 140 • C in presence of base, and several other strategies were proposed throughout the years.[25].
The results of HPLC and SERS analysis of ACNA 71 sample, catalogued as "Safranina T extra", were at one in pointing to Safranine O (3,7-diamino-2,8-dimethyl-5-phenylphenazinium chloride, Basic Red 2, An interesting feature of historical samples of Safranine O is the presence of multiple positional isomers, although commercial sources of this dye are usually assumed to consist of a single compound [26].The work of Andersen and colleagues, who isolated and characterized the three major components of commercial Safranine [26], helped us to identify the same isomers in the analysed ACNA sample based on elution order, labelled as Safr I, Safr II, Safr III and Safr IV in the following.Remarkably, besides HPLC analysis, isomers separation was obtained also by TLC (see Fig. 3), where three pink spots where clearly visible (Safr III is present in much lower amount respect to the other three isomers, and is thus probably below the technique detection limit).From each spot it was possible to acquire a distinct SERS spectrum, and their superimposition returns the SERS spectrum of Safranine O (reported in Fig. S7).In this work the most visible spot, corresponding to spectrum b in Fig. 3, lies in between the other two forms of Safranine as well as in [26], and this was also confirmed by HPLC techniques where the most abundant isomer (Safr II) elutes between the other two main Safranine isomers (Safr I and Safr IV).Some peaks from Safranine O Raman spectrum were assigned by DFT calculations to specific vibrational modes in previous studies [27].Among these, the strong signal at 371 cm − 1 characteristic of the spot with the lowest retention factor (Rf 0.58, spectrum c in Fig. 3), was proposed to be the results of combined phenazine ring CC stretching, CCC bending, and CCN bending.
The characteristic dyes of for ACNA 63 and 64 were expected, based on the labels, to be Chrysoidine G (4-Phenylazo-m-phenylenediamine monohydrochloride, Basic Orange 2, C.I. 11270) and Chrysoidine R (2-Methyl-4-phenylazo-m-phenylenediamine monohydrochloride, Basic orange 1, C.I. 11320), respectively, which differ only for a -CH 3 substituent on the m-phenylenediamine group.However, an in-depth characterisation of their chemical composition enabled us to determine Safranine O, Auramine O and Michler's ketone in both samples 63 and 64, while Chrysoidine related compounds were not detected.Thus, as reported in Fig. 4, our research reveals that ACNA industry exploited different ratios of Safranine O and Auramine O to achieve the specific shades of orange desired in their products, or that an error occurred during the synthesis of Chrysoidine leading to Safranine O by a condensation mechanism.
A reference SERS spectrum for standard Chrysoidine G is reported in Supplementary Information (Fig. S8) for completeness.Basic Blue 6, C.I. 51175) was initially identified by SERS analysis in sample ACNA 85 ("Blu Meldola BB concentrato") and 86 ("Blu Meldola R") by comparison with the spectrum reported in the literature [28].Moreover, the presence of this dye was further confirmed by TLC-SERS analysis (Rf 0.92).A closer look at the SERS spectra (Fig. S4), however, revealed a difference respect to the literature one in the region between 1340-1370 cm − 1 .Specifically, the spectra recorded in this work had a main signal at 1344 cm − 1 while in [24] two signals are located at 1352 and 1377 cm − 1 .Meldola's Blue, among the first dyes synthetized by Raphael Meldola since 1879, can be regarded as the prototype of angular phenoxazines.Similarly to quaternary ammonium cations with alkyl substituents, it can undergo dealkylation processes due to photooxidation [29], consistently with the detection in the HPLC-HRMS chromatograms of its mono-demethylated and bis-demethylated forms.Moreover, HPLC-DAD-HRMS enabled us to determine three red-violet compounds both in ACNA 85 and 86 (Fig. 5 and Fig. S2).

Meldola's blue-based dyes
Based on MS and MS 2 results, these markers can be tracked to oxidative products of Meldola's blue and its demethylated products.Their detection, never reported before in the literature, suggests an exposure to oxidative conditions, such as those granted by HClO 4 used as strong mineral acid during the naphthoxazinium synthesis [30], or those occurring during ageing.On the basis of HPLC results, we could assign the spectra of Fig. S9 to oxidized Meldola's Blue or to monodemethylated oxidized Meldola's Blue.
Nonetheless, HPLC-HRMS analysis revealed Meldola's Blue (oxidized or oxidized and demethylated) as the predominant component in sample 86, whereas it was only detected at trace levels in sample 85.This could suggest the presence of another blue dye in sample 85, likely contributing to the chromatic disparities observed between the two samples.
Indeed, together with Meldola's blue, HPLC-HRMS, HPLC-DAD and TLC-SERS detected also Methylene Blue (3,7-bis(Dimethylamino)phenazathionium chloride, Basic Blue 9, C.I. 52015) in sample 85.In this work we were able to identify by TLC-SERS three spots traceable to Methylene Blue, distinguishable among them by a limited number of Raman peaks (see Fig. 6).The most characteristic band, which shifts from 1515 (in the spectrum of the compound having Rf 0.66) to 1504 cm − 1 (in the spectrum of the compound having Rf 0.40), in the literature is referred to ring stretch + H in-plane bend (S side) + CH 3 bend [31].
In the light of HPLC analysis, these forms could correspond to Methylene Blue (Rf 0.66) and its mono and bis-demethylated forms (Rf 0.63 and 0.40, respectively).The calculated topological polar surface area (tPSA) for these compounds is 18 (Methylene Blue), 27 (monodemethylated Methylene Blue), 36 (bis-demethylated Methylene Blue on different N atoms), 41 (bis-demethylated Methylene Blue on the same N atom).Considering the high difference in terms of Rf between the first and second spot (0.40 and 0.63) we could suppose that the spot having Rf = 0.40 corresponds to Methylene Blue bis-demethylated on the same N atom.

Vesuvine-based dyes
Bismarck Brown or Vesuvine is the common name used to identify a group of azo dyes, which are found commercially as a mixture of closely related compounds (structures reported in Fig. 7A).Among the earliest, Bismark Brown Y (Basic Brown 1, C.I. 21000) was first described in 1863 by Carl Alexander von Martius.Other azo-dyes commonly used at the beginning of the 20th century for textile dyeing were Bismarck Brown R (Basic Brown 4, C.I. 21010), obtained by azo coupling using 2,4-diaminotoluene as both diazo and coupling component, and Bismarck Brown G, a dye mixture that contains a monoazo dye formed when diazotized m-phenylenediamine couples with itself.
Fig. 7B shows the SERS spectra obtained from sample 67 ("Vesuvina R"), compared to the reference spectrum of standard Bismarck Brown Y.It is possible to observe some correspondence between the two spectra, but the unambiguous identification of the main dyes characterizing the sample is not feasible.HPLC-DAD-HRMS analysis (Fig. 8) confirmed the absence of Bismark Brown Y in the sample, while enabled us to detect specific products due to the formaldehyde or acetaldehyde-induced degradation of Bismarck Brown R in the textile swatch samples [32].These findings are consistent with previous literature reports [5], confirming that Vesuvine-based reference textiles are chemically unstable and suggesting that an in-depth study of their degradation mechanism is necessary.

Triarylmethane-based dyes
Triarylmethane dyes, such as Methyl and Crystal Violets, Diamond green and Magentas, were produced since the late 19th century and consist of mixtures of homologous compounds, differing only for the nature, presence, or position of the substituents on the aromatic rings.Different substituents allowed the producers to obtain a wide range of colours, ranging from red to blue and purple; the composition of the dye mixtures was fundamental in achieving the desired colour [3].The name  frequently is followed by a capital letter, e.g.R and B, to differentiate the redder from the bluer shades.A number can precede the letter, to provide information on the lightness of the colour: for instance, in their dye form, Methyl Violet 6B is darker than 2B and lighter than 10B.
Methyl violets have been extensively studied by SERS [20,33,34] finding that this technique, alone or coupled to chemometric analysis, is not able to distinguish between hexa-N-methylpararosaniline (hexa-MP in the following) and penta-N-methylpararosaniline (penta-MP).In the case of penta-MP, SERS analysis followed by principal component analysis is indeed able to discriminate just di-amino from tri-amino derivatives.Cesaratto and coworkers [33] followed the photodecomposition of hexa-MP and penta-MP by SERS, finding that the spectra of the standard compounds exposed to UV light gradually changed up to resemble that of PR, consistently with the results described by chromatographic analysis by Confortin et al. [35].
Here the complexity of the formulation of ACNA samples 87-88-89-90, marked as "Nero Basico" (i.e.basic black), is highlighted by the several coloured spots visible on the TLC plate.At least 4 spots were detected for each sample, with Rf from 0.78 to 0.89 and SERS spectra consistent with triarylmethane-based dyes.Fig. 9A shows the SERS spectra obtained from the analysis of four adjacent TLC spots from sample 87.Analogous spectra were obtained from TLC-SERS analysis of standard hexa-MP and penta-MP, which as expected turned out to be a mixture of N-methylated pararosanilines.The spectral pattern of the four spots appears almost superimposable, as shown in Fig. 9A, except for few bands around 559-567 (symmetric 6a benzene mode, bending of C-N-C, out-of-plane deformation of C-C-C), 725-732 (C-N stretching), 1172-1178 (asymmetric 9a benzene mode, asymmetric stretching of the central C-C-C), and 1373-1377 (bending of C-H and of the central C-C-C) cm − 1 .In particular, a gradual shift of SERS signal from 1178 to 1172 cm − 1 was observed to correlate with compound's Rf, i.e. to the polarity of the compound (see Fig. 9B).HPLC-HRMS and HPLC-DAD analysis confirmed the presence of compounds at different N-methylation degrees, identified as tri-MP, tetra-MP, penta-MP and hexa-MP (see Fig. 9D).
Based on the elution order in liquid chromatography, TLC stationary phase polarity (direct phase) and calculated tPSA for each compound, structural hypotheses can be outlined (Fig. 9C).Specifically, we propose an elution order on TLC plate from tri-MP (lowest Rf) to hexa-MP (higher Rf).
In addition to methyl violet-based dyes, we detected by TLC-SERS also Malachite green (C.I. 42000; Basic green 4, which is a triarylmethane dye as well, reference SERS spectra in Fig, S10) in samples 87-88-90, and Safranine in sample 87.Malachite green was also a component of ACNA sample 84 ("Verde Basico 2G"), along with Auramine O and Michler's ketone.All these results were confirmed by chromatographic analysis.The Rf of malachite green (higher respect to the Rf values found for triarylmethane-based dyes) confirmed the correlation with its calculated tPSA (6.5) and the elution order proposed for the different pararosaniline forms.In addition, HPLC-DAD-HRMS analysis (Figs. S1 and S3, HPLC-DAD; Figs.S4 and S5, HPLC-HRMS) enabled us to identify three yellow products with a molecular ion of [M] + = 226.122(t R = 17.7 min), [M] + = 212.106(t R = 15.6 min),and [M] + = 198.090(t R = 13.0 min) in ACNA 84, 87, 88, 90, which have already been reported in the literature by Tamburini [5] as degradation product of Brilliant green (C.I. 42040, Basic Green 1) or Malachite Green.However, the molecular structure of the specie proposed by Tamburini is also consistent with a degradation product of Michler's ketone, which was detected in the sample, thus its origin cannot be ascertained.
ACNA samples 87, 88, 89, and 90 were also found to contain Methylene Blue by HPLC-DAD-HRMS (traces levels for sample 87 and 90).For sample 89 three TLC spots were visible, with the same Rf of sample 85, and showing SERS spectra attributable to methylene blue with different number of N-methyl substituents (see Fig. 9A).
Table S2 summarizes the results of the analyses performed by the different techniques.

Conclusion
Summarising, the research carried out on ACNA catalogue enabled us to outline the complex, and in most cases unexpected composition of a wide array of synthetic organic dyes used in the 20th century textile industry.Hence, the molecular profile of Auramine O, Safranine T, Malachite green, Methylene blue, Meldola's Blue, Bismarck Brown R and Methyl violet formulations were fully characterised by SERS, TLC-SERS, and HPLC-DAD-HRMS analysis.Our findings enabled us to prove significative differences between the label reported in the catalogue and the actual chemical composition of ACNA silk textiles.Fabrics labelled with different names turned out to contain qualitatively identical synthetic dye profiles (e.g."Auramina II" and "Auramina O", "Crisoidina G" and "Crisoidina R", and "Nero basico C" and "Nero basico SC"), proving that variations in chromatic shades in the swatches were only related to adjustments in the relative proportions of dye components.In addition, our investigations revealed disparities within similarly catalogued names (e.g.Blu Meldola BB and Blu Meldola R, Nero basico B and Nero basico BT), where differences were identified concerning the presence of additional dyes compared to those reported by the producer.
For the first time TLC-SERS was exploited in the investigation of historical textile dyes, not only to unravel the composition of multiple dyes mixtures, but also to distinguish structurally similar compounds (i.e. having a variable methylation degree, such as triarylmethane dye compounds).The results obtained from TLC-SERS analysis of liquid extracts from reference dyes applied on silk fibres in 1951, for the distribution of a catalogue, were confirmed by the more consolidated HPLC-DAD-HRMS approach.Elution order on TLC plates and consideration regarding polar surface area enabled us to assign the SERS spectra from each TLC spot to a specific form of each dye.The next step will be a joint experimental and theoretical study to attribute the SERS signals distinctive of each compound to the corresponding vibration.
More generally, this study contributes to further increase the existing database of reference dyes and of their ageing markers or synthetic byproducts, including their SERS spectra, and confirms the unreliability of the association between name and composition in historical catalogues.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Cover, first page and example of a page from the catalogue "Coloranti Basici" released by ACNA.The publication date (1951) is visible on the first page.
PhotoCode Name as catalogued

Fig. 2 .
Fig. 2. SERS spectra acquired at 532 nm of ACNA sample 61 ("Auramina II") and 62 ("Auramina O").The spectrum of Auramine O analytical standard is added for comparison; in the inset, the chemical structure of Auramine O is provided.All spectra are presented in the same scale and are stacked for purpose of clarity.

Fig. 3 .
Fig. 3. SERS spectra acquired at 633 nm from TLC plate of ACNA sample 71 ("Safranina T"), along with a photo of the corresponding TLC plate and calculate Rf.The chemical structure of Safranine O is shown.The arrow shows the characteristic peak at 371 cm − 1 (see main text).All spectra are presented in the same scale and are stacked for purpose of clarity.

Fig. 6 .
Fig. 6.SERS spectra acquired at 532 nm from TLC plate of ACNA sample 85 ("Blu Meldola BB concentrato"), along with a photo of the corresponding TLC plate and calculate Rf.All spectra are presented in the same scale and are stacked for purpose of clarity.

Fig. 7 .
Fig. 7. A) Chemical structures of Bismarck Brown-based dyes.B) SERS spectra acquired at 532 nm from ACNA sample 67 ("Vesuvina R").The spectrum of Bismarck brown Y analytical standard is added for comparison.All spectra are presented in the same scale and are stacked for purpose of clarity.