The results are based on the comparison of the reaction of the samples during the preparatory procedure and on the counting of coccolith in the smear slides. Even though several extraction procedures already exist (54,56,79), we considered that they are not conservative enough for the low amount of material of our sample and would increase the loss and contamination with a probability too high to be risked with such specific samples. The counting of coccoliths was done under light microscopy even though it may lead to taxonomical problems (80,81). The verification of the species was not the focus of this study and microscopy stays the easiest method available (81). It follows the rule used by Backmann et al. (82) which chose a specific number of frames to be counted at a specific magnification. The difference of coccoliths between our 3 model samples was high. For RSG, the average value counted was 307 coccoliths, for LO, it was 112 and finally, for the historical model sample, it was 101. Compared to the reference value of 300 counted coccoliths on 10 frames. Commonly accepted, the counting should be between 300 and 400 coccoliths per slide (83). The counted coccoliths difference is relatively high and can be due to the distribution of the coccoliths onto the smear slide. All slides were first observed, and the best-distributed site was taken for the counting (low number of aggregates, separation of the coccoliths). We suggest that the smear slides could also be prepared according to Gallagher (56) for a combined determination through scanning electron and optical microscope as also developed in other works (84–87). This will assure proper identification of the different species present.
Disaggregation methods
During the M1 disaggregation procedure, one model sample MS01 (Fig. 4) reacted with water, peroxide and both micro-emulsions. The largest whole coccoliths occurrence was with peroxide at 70◦C and the largest occurrence of coccoliths including fragments was found in the peroxide at room temperature. Compared to the peroxide with pure chalk, the presence of whole coccoliths increases by 2.5x and of broken coccoliths decreases from 1/4x. The reaction between RSG and H2O2 allow fewer molecules to react primarily with the chalk particles. This preserves the chalk particles as long as the binder is around. The disaggregation was very low with model samples MS04 (Fig. 5) and MS16 (Fig. 6), which shows a high variation in results according to the material. The peroxide was discarded because its result could not be predicted if the samples were of unknown composition. Water also worked on MS01 (Fig. 4), but not on MS04 (Fig. 5), and neither on MS16 (Fig. 6). It can still be used as a binder pre-dissolution for a surface re-hydration of the binder that will ease the later disaggregation with one of the other procedures. Acetone showed a reduced impact only on MS04 (Fig. 5).
In the M2 procedure, micro-emulsions ME1 and ME2 dissolved effectively MS01. It is also efficient to lose the aggregates and it gives a clear gravity separation between the components compared to the other ME. ME2 has a very good dissolution rate (no aggregates to very small ones) and seems more effective on MS04 than ME1 even though ME1 is more concentrated. The linseed oil specimen turned orange combined with ME1 or ME2 (Fig. 7). The change in colour is probably due to the formation of degradation products of proteins owing to the action of the strongly alkaline quaternary base. The separation of the material is very good, the solution does not turn into a gel and there is no apparent dissolution of the inorganic particles. Only micro-emulsion ME1, ME2 and ME3 dissolved rabbit skin glue. Emulsion ME3 has a lower dissolution rate than ME1 and ME2 for the glue samples. The solution gels with the glued sample and the separation of the component are unclear. ME4 does not dissolve very well the glue that stays at the bottom of the assay tube. ME5 dissolves partially the glue that gels the solution as ME3, the rest stay at the bottom of the essay tube as aggregates. ME4 and ME5 can therefore be discarded as they are not dissolving proteins that are in many cases the main component of paintings. During the interpretations of the smear slides, some showed a kind of layer on top of the particles which could be related to the ME. This lowered in some cases the determination rate of nannofossils.
M3 gives similar results on MS04 (Fig. 5) specimens than M1 and M2 but gives better results on MS01 (Fig. 4). MS16 (Fig. 6) was quicker and better disaggregated with M3 than with M1 or M2. The smear slides showed no glue traces nor the layer visible in M2 and generally fewer aggregates than M1 and M2 as well. We had therefore a higher rate of whole nannofossils.
Binder analyses
Gas chromatography coupled to mass-spectrometer (GC-MS) offers precise analysing possibilities but is bound to the destruction of the sample (47). GC-MS was chosen for this analysis as it is a fine-tuned way for organic matter determination (48). FAME and lipids were analysed. The full process of fatty acid methyl ester (FAME) analysis consists of esterification of lipids and of injection, separation, identification, and quantitation of the FAME (47). We analysed each component separately (oil, glue. chalk, micro-emulsion) at different ratios to get a molecular mapping of each as a reference. We analysed MS3 alone and disaggregated it with our methods. The samples had dry masses between 0.1 and 0.5 mg or were already dissolved in 1.5 mL Eppendorf. We also analysed the real samples to see if there was pigments or dye interference (49).
The concentration of fatty acid (FA) in a sample will indicate more or less precisely the type of organic components (plant or animal-based, species) present but also depending on some FA markers and their ratio information about the drying process (age), the presence of other chemicals and reactions occurring can be interpreted (96). For example, palmitic (16:00) and stearic (18:00) acid ratio is used to distinguish animal fats from plant oils (97) but this ratio is not always clear as for Mills (98) who differentiated linseed, walnut and poppy seed oils or Mateo-Castro who combined several fatty acids ratios and gas chromatography with flame ionization detection (GC-FID) (99). Several Fatty acids markers are used for the determination of specific binders such as myristic acid (C(14:0)), palmitic acid (C(16:0)), stearic acid (C(18:0)), oleic acid (C(18:1)), linoleic acid (C(18:2)), suberic acid (2C(8)), azelaic acid, (2C(9)) and sebacic acid (2C(10)) for drying oils (100) or azelaic, palmitic and stearic fatty acids which tend to give the indication of lipoid binder (99). Degradation products such as pelargonic acid as in the Mateo-Castro study will give information on the degradation state and process occurring (99). Their interpretation is therefore rich in information about the technique, the school, and the period of production for conservator and restorer.
In our study, we analysed pure components (RSG100, LO100, ME1, ME2) and mixed components (RSG50/30/10, LO50/30/10 - example: 50% RSG for 50% chalk (CC) = RSG50).
We observed that the pure rabbit skin glue contains 9 different fatty acids (FA; arachidic (14:00), beheric (22:00), Palmitoleic/Hypogeic (16:1w7), 7-octadecenoic (18:1w7), stearic (18:00), palmitic (16:0), linoleic/octadienoic (18:2w6,9), ambiguous structure/oleic (18:1w9), alpha/gamma-linoleic/octadecatrienoic (18:3w3,6,9)) (Fig. 9) while RSG + CC samples have similar FA but in other concentrations and also contains other FA (penthadecyclic (15:00), 15-methylpalmitic (i-17:0), isopalmitic (i-16:0), Palmitoleic/5-hexadecenoic (16:1w5), palmitic (a17:00), margaric (17:00), myristic (14:00), 12-methylmystiric (a15:0)). These additional FA can be due to contamination while the preparation (brush, fingers’ fat, medium on which it was deposited), from the chalk or because of the medium-chalk interaction (Fig. 9).
The difference between the LO samples is not as clear as in the RSG samples. The FA of LO100 (stearic (18:00), Palmitic (16:00), Linoleic/octadienoic (18:2w6,9), Ambiguous structure/oleic (18:1w9), alpha/gamma-linoleic/octadecatrienoic (18:3w3,6,9), pentadecyclic (15:00), margaric (17:00)) and LO30 (LO100 FA + arachidic (20:00), palmitoleic/hypogeic (16:1w7), 7-octadecenoic (18:1w7), myristic (14:00)) matches mostly in type and concentration. LO30 has the higher concentrations of FA with
The difference between RSG100 (Fig. 9) and LO100 (Fig. 10) is the absence of 7-Octadecenoic (18:1w7), Hypogeic (16:1w7), arachidic (20:0) and beheric acid (22:0) in LO100 compared to RSG100 and the absence of Pentadecyclic (15:0) and margaric acid (17:0) in RSG100 compared to LO100. Compared to the lower concentration samples, we found higher values of arachidic (20:00) and 7-Octadecenoic acid (18:1w7) and the presence of pentadecyclic acid (15:00). LO10. Arachidic acid (20:00) appears in LO + CC samples and consistently in all RSG samples. This FA is probably in CC and related to RSG. 12-methylmystiric acid (a15:0) only appears in RSG30 and LO50 which is probably due to external contamination. The same goes for beheric acid (22:0) which is only detected in LO10, RSG + LO25 and RSG100.
RSG100 and RSG+LO25 are the closest in terms of FA presence but we observe that we have an additional FA as mystiric acid (14:00) which is also higher than in the RSG+CC samples (Fig.10). Mystiric acid (14:00) concentration in RSG+LO25 is similar to LO30 when almost absent from LO50 (Fig.10)
Linseed oil fatty acids (FA) concentration in parts per million (ppm) on a logarithmic scale. LO100 has lower concentrations of FA which weren’t expected when LO30 and LO10 have higher concentrations.
The influence of the micro-emulsions on the samples result is observed in Fig. 11. We observed that ME1 which is the strongest micro-emulsion in terms of active principles contains some of the same FA as LO and RSG and in terms of concentration is closer to RSG10. ME1 contains 6 FA (stearic (18:00), palmitic (16:00), linoleic/octadecenoic (18:2w6,9), Ambiguous structure/oleic (18:1w9), alpha/gamma-linoleic/octadecatrienoic (18:3w63,6,9), margaric (17:00)) when ME2 only contains 2 FA (stearic (18:00), palmitic (16:00)). This is probably due to the detection limit of the method.
The ratio between stearic and palmitic acid can be used for the determination of the presence of drying oil along with other FA (98,100,101).
As ME1 and ME2 contain both acids, there is a risk that it could change the ratio depending on the primary amount of FA detected by GC-MS in the painting samples. According to this, we recommend using ME2 which will have a lower impact on the result than ME1.