FTIR spectral characteristics of functional groups in the bindersFrom the molecular structure of the epoxy resin adhesive shown in Fig. 2, the structure contains the amide bond, ether oxygen bond, benzene ring, hydroxyl group, imine bond, etc. The main characteristic peaks related to these bonds of all samples can be found in Fig. 3, indicating that the IR spectral characteristic peaks of the adhesive here measured match the bisphenol A epoxy resin cured with polyamide.Figure 3FTIR spectra of the epoxy resin binders served on the restored terracotta warriors. The spectra obtained using KBr plates (a, b, c and d) and ATR (e).Although the infrared spectra of all samples shown in Fig. 3 are extremely similar, some differences still exist. For the spectra obtained by using the KBr pressed-plate method as shown in Fig. 3a,b,c,d, the following results were obtained. Firstly, the FTIR characteristic peaks from PA650 and E-44 are found in Fig. 3. The intensive and relative sharp absorption band centered at 3420 cm−1 ranging from 3685 to 3000 cm−1 corresponds to the overlapping stretching modes of the primary amine group form PA650 and the hydroxyl groups formed by the reaction of PA650 with epoxy groups of E-4425. Additionally, the stretching modes of the C–H bonds at near 2920 cm−1 and 2850 cm−1 bands, and the bending mode of the H–C–H angle at 1462 cm−1, originated from the CH2 groups in PA650 and E-44 also appeared. For PA650, the most prominent FTIR characteristic peak is the carbonyl of a primary amide vibration at 1650 cm−1 19,26. For E-44, the characteristic peaks of the bisphenol A epoxy resin, 1510 cm−1 elongation vibration associated with aromatic ring (–C=C), 1362 cm−1 (the CH3-bending vibration)27, the band at 1184 cm−1 to the stretching mode of Ar–C–Ar (Ar represents the aromatic ring), the band at 1038 cm−1 attributed to the stretching C–O–C of ether, the peak at 830 cm−1 of the benzene ring distribution, appear in the spectra of all samples28. Above results indicated that the adhesive used in the restoration of terracotta warriors belongs to a mixture of polyamide and bisphenol A epoxy resin.Although the PA650 polyamide and E-44 epoxy resin respectively have the above typical characteristic peaks and the FTIR spectra of all samples are similar to some extent, there are overlap of some characteristic absorption peaks between PA650 and E-44, resulting in the difficulty to precisely assign these peaks. For example, the peaks at 1250 cm−1 (asymmetric C–C–H bending) and 1164 cm−1 (symmetric deformation C–C–H)29 from PA650 respectively overlap the peaks at 1250 cm−1 (the anti-symmetric stretching of aromatic ether –C–O–C) and 1164 cm−1 (symmetric stretching of aromatic ether) from E-4430,31,32. However, the relative content of E44 in the binders can be determined by using the relative intensity of E44 characteristic peaks. These peaks include the elongation vibration at peak 1510 cm−1 associated with the aromatic ring (–C=C) and the peak at 830 cm−1 of the benzene ring distribution28. In the infrared spectra of all samples, the samples (ES-2018-2, IC-1990-3, CT-1989-3, BF-1989-3) showed strong characteristic absorption at 1510 cm−1 and 830 cm−1, indicating that these samples contained higher E44, also reflecting that there was some degree of randomness in the relative amount of E44 and PA650 used in the restoration practice of the Qin terracotta warriors. In fact, the difference in the relative amount of PA650 and E-44 in the binders has some universality in practice33.Additionally, there are some distinct differences in details of the FTIR spectra ranging from 1250 to 1000 cm−1 for the samples shown in Fig. 3. These differences mainly reflect changes of various vibration modes involving aromatic ether, fatty ether, and alcohol hydroxyl dominated by the reaction of PA650 with E-44 in a different relative amount34.For MT-2015-3 (Fig. 3d), compared with all samples, the intensities of absorption mainly attributed to E-44 at 1510 cm−1, 1462 cm−1, 1184 cm−1 and 830 cm−1 are lower, indicating that the ratio of E-44 to PA650 used in MT-2015-3 is the lower among samples measured. Accordingly, the peaks related to the alkyl ether groups and the secondary alcohol groups at 1085 cm−1 (alkyl ether C–O symmetric stretching), 1038 cm−1 (C–O bending vibration), at 3420 cm−1 (O–H stretching vibration, N–H stretching vibration), formed by the reaction between E-44 and PA650 appear very obvious. For BF-1989-3 (Fig. 3d), compared with MT-2015-3, the absorption intensities at 1530 cm−1, 1462 cm−1, 1184 cm−1 and 830 cm−1 originated from E-44 are higher, indicating that the ratio of E-44 to PA650 used in MT-2015-3 is the higher among samples measured. Additionally, due to the presence of two epoxy groups in bisphenol A epoxy resin E-44, more ether and alcohol hydroxyl groups are present in the reaction products between E-44 and PA650. As a result, the peaks related to the alkyl ether groups and the secondary alcohol groups at 1085 cm−1 (alkyl ether C–O symmetric stretching), 1038 cm−1 (C–O bending vibration), the at 3420 cm−1 (O–H stretching vibration), exhibit more obvious.Based on the infrared spectrum analysis mentioned above regarding the two typical ratios of E-44 and PA650 for preparation of binders, it can be inferred that samples ES-2018-2 and ET-1989-3 binders were prepared in a relatively high ratio of E-44 to PA650, while the binders (SF-2018-2, IC-1990-3, MT-2020-3, UC-1996-3, CF-1989-3) were prepared in a relatively lower ratio of E-44 to PA650. For the two-components epoxy resin binders that require on-site mixing of two components with strong viscosity, it is difficult to strictly control the amount of components in practical operation. In fact, the above deduction mentioned is quite similar to the relative change of aromatic ether and aliphatic ether in the bisphenol A epoxy resin system reported in the literature35. Therefore, it is not surprising that the binders used at different times have different ratios of E-44 and PA650. However, in view of practice requirements, it is necessary to develop a convenient method for on-site quantitative mixing of epoxy resins and cross-linking agents.Because the amount of samples (MA-2018-2, ES-1989-3, IC-1996-3) obtained was not enough to be determined using the KBr method, these samples were measured by the ATR mode. As for the spectra obtained using ATR mode (Fig. 1S), due to the inherent feature of this method, the measured result obtained from this method is not exactly same as the KBr method. Firstly, a lower intensity at higher wave-numbers is one of the innate effects associated with ATR due to the decreased penetration depth of the evanescent wave36, this results in greater absorption on the longer wavelength side of an absorption band, contributing to band distortion and band broadening. However, the FTIR spectra obtained by using the KBr method are qualitatively similar to the ATR spectra in the fingerprinting region37. Secondly, ATR mode mainly responds to the surface information of a sample. For the surface composition of the sample may being different from that of the interior, the ATR mode has more advantages over the KBr method in detecting the FTIR spectrum. In any case, under the same detecting conditions, the differences in the spectra of monitored samples with different aging times can still provide some useful information. The ATR FTIR spectra of the samples (MA-2018-2, ES-1989-3, IC-1996-3) show in Fig. 4. For the samples of ES-1989-3 and MA-2018-2, their main characteristic peaks were basically similar to that found in Fig. 3. According to the similar analysis of the spectra shown in Fig. 3, there is no significant advantage in the relative amounts of E44 and PA650 used in samples (MA-2018-2, ES-1989-3, IC-1996-3) shown in Fig. 4. Unexpectedly, for ES-1989-3 and MA-2018-2, the strong peaks at 2920 cm−1 and 2850 cm−1 corresponding to the stretching modes of the C–H bonds of –CH2– groups appeared, unlike the spectrum determined by ATR mode to be commonly lower intensity at higher wave-numbers. This situation may be related to the relatively more abundant distribution of PA650 chain segments on the surface. PA650 has the hydrophobic chain segment38. As a result, such hydrophobic chains tend to move towards a surface, spontaneously, the groups of –CH2– groups as well as the corresponding functional groups in PA650 can be specially monitored by ATR mode (Fig. 4). It has been reported that the chemical composition of an epoxy resin is not uniform along the direction normal to the solid interface39. The primary amide vibration characteristic peak at 1650 cm−1 and N–H stretching vibration at 3300 cm−1 from PA650 also indicate the distribution advantage of PA650 on the surface of the binder. These results indicate that ATR FITR spectroscopy has a significant advantage in characterizing surface functional groups of materials. In general, the aging of materials first occurs on the surface. Therefore, ATR FITR has obvious advantages in characterizing material aging. By polishing the surface of the sample and examining the FTIR spectra of the bulk resin materials, Miller SG et al. confirmed that the chemical aging processes, including cure and oxidation, were limited to the surface of the resin40. Therefore, to some extent, the infrared spectroscopy detected by the ATR mode is more effective for evaluating the epoxy resin aging.Figure 4ATR FTIR spectra of the epoxy resin binders served on the restored terracotta warriors. The spectra obtained using ATR mode.In particular, the intensity in IR spectrum of the sample IC-1996-3 is significantly weaker even under the same test conditions. Obviously, the result is related to the sample itself rather than others. It has been found that rigid materials can be problematic to measure with the ATR technique as it is difficult to create close optical contact with the diamond crystal. As contact is confined to small areas, weak spectra are produced, the effects of which are greatest at shorter wavelengths where the depth of penetration is lowest41. It is clear that aging causes an increase in relative rigidity, and it is explained by the reduction in free volume during the physical aging process42. Meanwhile, within the range of 1250–880 cm−1, the strong absorption peak indicates that the sample IC-1996-3 may have a high cross-linking density. This is also another reason why the sample has strong rigidity. Therefore, the sample IC-1996-3 may have too strong rigidity due to physical aging and high cross-link density so it is difficult to create close optical contact with the diamond crystal in ATR mode, resulting in decreased optical absorbance.Although there are some differences in the specific details of the infrared spectrum caused by the difference in the relative content of E-44 and PA650 or the detection mode, all the samples have similar normal spectral features. Unfortunately, the lack of FTIR spectra of the corresponding original samples made it impossible to directly compare with present spectra. However, many investigated results regarding epoxy resin aging reported in the literature provided an important reference to estimate whether the measured samples have been aged. It has been reported that the carbonyl group was an important marker of epoxy resin aging25,43,44,45, and the formation of different carbonyl units might originate from the photo-oxidation of different alkyl and phenyl units. In terms of hygrothermal aging, most of the literature showed that hygrothermal aging can also cause the oxidation of α–CH2 or amines29,39,46,47,48 and isomerization of oxirane ring49 to form carbonyl groups. Fortunately, the characteristic peak at 1730 cm−1 assigned to the carbonyl functional groups was not present in the FTIR spectra of all samples shown in Fig. 3, indicating that all samples did not undergo visible chemical aging25,40,41,42,43,44. In fact, the results are not unexpected to some extent. All exhibited unearthed Qin terracotta warriors and horses are set in the exhibition hall of the Mausoleum Museum of Emperor Qin Shihuang, and this museum is located in the Guanzhong Plain, China, with a semi-arid and semi-humid climate. As described in the 2.3 section, the humidity and the temperature of the museum don’t exceed 90% and 43 °C, respectively. Generally, in a year, the humidity and the temperature in two extreme climatic conditions are respectively in summer 2.8–13.6 °C, 20.0–57.6%, in winter 23.2–37.5 °C, 4 3.0–78.5%24. It has been reported that epoxy resins were not prone to chemical aging in such environmental conditions. Many studies have shown that the UV photo radiation degraded materials much more compared to hygrothermal expose25. Additionally, the hygrothermal aging of epoxy resins at below 60 °C did not cause any chemical modifications, the chemical aging at this temperature was very slow even at 90% relative humidity36,47. For instance, the literature reported that the epoxy resins did not significantly change in colour upon in-room storage after 28 years50. Therefore, in a mild indoor temperature and humidity environment for 40 years, no significant chemical aging of the epoxy resins used for the restoration of the Qin terracotta warriors is not surprising.XPS spectral characteristics of functional groups in the bindersX-ray photoelectron spectroscopy (XPS) is an indispensable technique in materials science for the determination of surface chemical bonding. To further investigate the possible chemical aging of the epoxy resin served in the Qin terracotta warriors, the XPS spectra of all samples involved in this investigation were detected. It can be found that there are three elements, C, N and O, in all samples, and N is related to PA650, and C and O are related to both E-44 and PA650. Taking into consideration of the molecular structure of the samples and the reports in reference51,52,53,54,55,56,57, here we tentatively performed curve-fitted high-resolution XPS spectra of C1s, O1s and N1s, and the components that correspond to the different functional groups. The corresponding results of the samples (SF-1989-3, IC-1990-3, MT-2020-2), (UC-1996–3, CT-1989-3), (ES-2018-2, ES-1989-3), (MT-2015-3, BF-1989-3), (MA-2018-2, ES-1989-3, IC-1996-3), were shown in Fig. 5a,b,c,d,e), respectively. For the C1s spectra, five different carbon environments were found, including 284.8 eV (C–H, C–C, C=C), 285.3 eV (C–N), 286.1 eV (C–OH, C–O–C), 288.2 eV (NC=O)58,59. For the O1s spectra, three different oxygen environments were found, including 533.0 eV (C–OH and C–O–C), 531.8 eV (NC=O, adsorbed water or oxygen). It can not be divided an unequivocal signal related to C=O groups at ~ 534 eV in the O1s spectra and to C=O groups at 289–290.1 eV in the C1s spectra60,61,62, further indicating no chemical aging occurred visibly in all samples. For the N1s signal in all samples, the peak centered at ~ 400 eV appeared, which can be assigned to C–N CONH species57. It is worth noting that the N1s spectra of the samples (IC-1990-3 Fig. 5a; CT-1989-3, Fig. 5b; T-1989-3, Fig. 5c; BF-1989-3, Fig. 5d; MA-2018-2, ES-1989-3, IC-1996–3, Fig. 5e) emerged a small shoulder peak at 401.8 eV attributed to the amine groups (N–H) species63, indicating that these samples have a certain number of uncross-linked amine groups.Figure 5(a) XPS of C1s, N1s and O1s spectra of the epoxy resin binders (SF-1989-3, IC-1990-3, MT-2020-2). (b) XPS of C1s, N1s and O1s spectra of the epoxy resin binders (UC-1996-3, CT-1989-3). (c) XPS of C1s, N1s and O1s spectra of the epoxy resin binders (ET-1989-3, ES-2018-2). (d) XPS of C1s, N1s and O1s spectra of the epoxy resin binders (MT-2015-3, BF-1989-3). (e) XPS of C1s, N1s and O1s spectra of the epoxy resin binders (MA-20187-3, ES-1989-3, IC-1996-3).The XPS spectra shown in Fig. 5a,b,c,d,e effectively reflect the main chemical bonds of all samples detected in this study, which consistent with the conclusions obtained from the corresponding infrared spectra shown in Figs. 3 and 4. Figure 5a,b,c,d,e indicates that the peak attributed to C–O–H and C–O–C species in ES-2018-2, ET-1989-3, SF-1989-3, MT-2015-3 and BF-1989-3 is more obvious than that in IC-1990-3 and ES-1989-3, which is related to the fact that the ring-open reaction of epoxy group occurred in the former samples is more significant than in the latter ones. These results are consistent with the characteristic peaks ranging from 1200 to 1132 cm−1 and the centered 3400 cm−1 shown in Figs. 3 and 4. It is worth mentioning that there is an evident difference in the consistency of the relative intensity between the species CONH and the species (C–O–C, C–OH) in C1s and in O1s in Fig. 5 for all samples. For samples, the species CONH compared with the species (C–O–C, C–OH) in O1s are always dominant but not in C1s. The unexpectedly high portion of CONH in the deconvolution of O1s spectra may be associated with the oxygen species unconsidered. It has been found that the O2 binding energy (BEs) corresponds to various oxygenated species, as for example, C=O, having similar BEs (531.8 eV) and thus hardly discernable by curve-fitting64. This report implies that the variability mentioned above may be related to the presence of oxygen in the samples. In practice, air inclusion can occur when mixing two-part epoxy liquid systems by hand65. Only vigorous stirring can mix the sticky E-44 and PA650 due to both E44 and PA650 being highly viscous. These results imply that the binder served in Qin terracotta warriors has oxygen that had been randomly introduced during the restoration process.Thermal decomposition characteristics of the bindersGenerally, a initial decomposition temperature is defined as the temperature at which the material loses 5% of its weight66. Therefore, thermogravimetric analysis (TG) is one of the important methods to assess material aging67,68.Figure 6 showed the thermogravimetric curves (TG(A1, B1), DTG(A2, B2)) of the binders used restoration of Qin terracotta warriors and horses. As shown in Fig. 6, for all samples, their TG curves have the similar profile, and the initial temperatures at which the samples significantly lost weight were above 300 °C, being very similar to the results of unaged epoxy resins reported in the literature69,70,71. Generally, the loss weight ranging from 310 to 480 °C is mainly caused by the elimination reaction of the end groups (hydroxyl, amino) in the cured structure and the degradation of the epoxy resin backbone72. To make it easier to identify the thermal decomposition temperatures of all samples, the derivative results (DTG) of the TG curves are shown in Fig. 6A2,B2, evidently indicating that there are certain differences in the thermal decomposition temperature of the measured samples. The temperature difference of the sample during the significant weight loss stage may be related to the difference in the contamination of inorganic substances in the sample. Comparing the DTG curves (Fig. 6A2,B2) and the residual weight in the thermogravimetric curves (Fig. 6A1,B1), it can be observed that the turning point of the thermal decomposition curve at the high temperature end of the sample with higher residual weight shifts to the left, resulting in a decrease in the weight loss temperature. In any event, the thermal decomposition behaviors of all samples shown in Fig. 6 are quite consistent with the literature report, suggesting that there has been no noticeable change over time in the structural and compositional features of all detected samples. It can therefore be concluded that the binder served in Qin terracotta warriors and horses did not undergo obvious chemical aging.Figure 6TG/DTG curves of the epoxy resin binders.It is worth noting here that there were still some subtle differences in the thermal decomposition behavior of these binders. Firstly, for the samples, including UC-1996-3, IC-1996-3, MA-2018-3, IC-1990-3, and MT-2020-3, exhibited the slight loss weight (approximately, 4%) ranging from 50 to 250 °C shown in Fig. 6A, and the corresponding curves were slightly different. These phenomena may be related to the volatilization of residual oligomers and residual moisture in epoxy resins72,73,74,75. In other words, it is normal that the slight loss weight at below 250 °C for the determined binder served in Qin terracotta warriors and horses. Secondly, a stabilization of mass percentage appeared after 500 °C for all samples, however, there were a significant difference in the final residue proportion between the determined samples. This difference is primarily attributed to the contamination of ceramics. Epoxy resin, as a masonry binder, has a so strong adhesion that the ceramic particles can be also peeled off in the sampling process of peeling epoxy resin16. Polarization micro-graphs and EDX spectrum of the corresponding samples respectively confirm this hypothesis. Using SF-1989-3A as a typical sample, the corresponding results are shown in Fig. 7. As expected, prominent characteristic peaks of the typical elements, such as Si, Al, K, Mg and Na related to ceramic materials, appeared in the EDX spectrum of the sample SF-1989-3. The bright regions originated from the bi-refraction of ceramic grains appeared in the polarization micrograph of the sample SF-1989-3 shown in Fig. 7, further demonstrating the presence of inorganic minerals in the sample SF-1989-3. The above results not only verify the rationality of the higher residual ratio observed in the thermogravimetric curves but also demonstrate that the binder still maintains a strong bonding ability. In other words, the performance degradation of the binder served in the Qin terracotta warriors is not significantly.Figure 7SEM image, EDX spectrum and polarization micro-graph of the sample SF-1989-3.
