The relevant bands that were used for sample evaluation are compiled in Table 1. Band positions are indicated according to Smith47 and Guo and Bustin48. The list is limited to visible band maxima. Aging/oxidation lead to interactions (e.g. H-bonds) and subsequently to broadening of bands49. Strong bands of inorganic components overlap smaller organic bands. Nevertheless, underlying features are included by multivariate data analysis of the spectral pattern. Data analyses were performed with selected wavenumber regions.
Stepwise procedure in the interpretation of sample sets from additional sites
Principal Component Analysis (PCA) of infrared spectra (wavenumber regions of 4,000–2,400 cm−1 and 1,800–400 cm−1) is performed for all samples (reference samples and dated sample sets). If their scores correspond to the age determined by the reference method, it is indicative that alteration proceeds according to the spectral pattern of the reference samples. Otherwise, further investigations are necessary as described below.
Trend of the spectral pattern of natural charcoal aging
Due to the determined age, it can be assumed that all samples were pyrolyzed at temperatures > 400 °C. Charcoals produced at lower temperatures are more affected by microbial degradation as shown by incubation50, field experiments42 and chemical oxidation17. During the aging process a common succession of the spectral changes is noticeable. Figure 1a and b display the PCA (4,000–2,400 cm−1 and 1,800–400 cm−1) of reference samples from Austria (recent, A, B, C) and sample sets of charcoals from Brazil (Rio) and Germany (Wittnau WI and Iznang IZ). Despite the heterogeneity within the group, their scores in PC1 are proportional to age. The loadings plot (Fig. 1b) clearly reveals the relevance of the organic bands with regard to the aging process.
The average infrared spectra of each sample set are shown in Fig. 1c to support the loadings interpretation. They feature the characteristic development of relevant bands in the region from 1,800 cm−1 to 1,100 cm−1: the increase in intensity of the carboxylate bands (1,585–1,565 cm−1 and 1,385–1,375 cm−1) and the concomitant increase, followed by a relative decrease in the oldest samples, of the carboxylic acid bands at about 1705 cm−1 and 1,260–1,250 cm−1. This decline is observed starting from the fifteenth to the thirteenth century BCE (reference samples “C”). The samples from Brazil (Rio), Wittnau (WI) and Iznang (IZ) fit in the series according to their determined age. Emerging functional groups and changes of band intensities in the carboxylic/carboxylate region are paralleled by a corresponding increase of the O–H stretch band in the spectral region 3,500–2,500 cm−1, which is in accordance with the increasing hydrophilicity due to oxidation (as shown in Fig. 3). Despite different sampling sites (Germany, Brazil) and therefore different climatic conditions, the oxidation process follows a useful regularity. The degree of carbonization seems more important than some differences in the environment, as confirmed by litterbag experiments, where degradation was generally highest in 500 °C chars and lowest in 300 °C chars, independent of storage conditions such as soil surface, litter, or layer of limestone chips42. Changes regarding drought, humidity and temperatures might be counterbalanced over the long-lasting residence time in the environment. Nevertheless, some environmental conditions have a strong impact on alteration or preservation15. Heterogeneity within groups is pronounced from the nineteenth to the thirteenth century CE, whereas older groups become more uniform. Over a period of millennia, the relevance of individual degrees of carbonization or environmental exposure abates and only samples with high degrees of carbonization and appropriate environmental conditions remain.
Archeological sites Bodnegg and Olzreute
Two sample sets (Bodnegg and Olzreute) do not fit in the series of reference samples in terms of their spectral features. Despite the age of several thousand years (3950–3650 BCE and 2900–2820 BCE, respectively), sample positions in the scores plot of the PCA (not shown), based on all reference samples and the wavenumber regions 4,000–2,400 cm−1 and 1,800–400 cm−1, are close to recent samples (rec) or overlap samples from reference set “A” (about 1800 CE). This first PCA was the basis for the second PCA (Fig. 2), which emphasizes these two relevant periods. This second PCA used the wavenumber regions 4,000–2,400 cm−1 and 1,800–1,100 cm−1. The wavenumber region from 1,100 to 400 cm−1 was excluded, as mineral compounds were not visible in the spectra.
According to their position in the scores plot, close to either recent samples (rec) or reference samples “A”, single spectra of the marked samples in Fig. 2 from both groups (BO1 and BO2, OL1 and OL2) are displayed in Fig. 3. The wide variability of intensities in the carboxyl/carboxylate region indicates different partial oxidation degrees among different BO and OL samples (within-group variation).
As the spectra of BO2 and OL2 feature properties of reference samples C (increase of the bands at 1,585–1,565 cm−1 and 1,385–1,375 cm−1 and the concomitant decrease of the bands at 1,705 cm−1 and 1,260 cm−1), a SIMCA (Soft Independent Modeling of Class Analogy) was calculated to find out the membership of the samples.
Most of the samples are located in the area “neither–nor”, but closer to the class “rec” than “C”. Only 7 out of 125 samples from Bodnegg, and none from Olzreute, are assigned to the class “C”, as would be expected from their determined age.
The similarity to recent samples or reference samples “A” (i.e., low degree of partial oxidation) indicates some protective mechanism from aging or high charcoal recalcitrance, which could be provided by a high degree of carbonization. The minor evidence of aging and the discrepancy between the indicated age and the spectral signature required additional investigation.
The scores plot of the PCA (Fig. 2) and the Coomans plot (Fig. 4) reveal the heterogeneity within these groups (BO and OL), which raises the question of whether charcoals with a wider range of age coexist in the same sites. In such cases FT-IR spectroscopy provides advantageous information as it allows analyses of huge sample sets due to low costs and a rapid procedure. The spectral feature can be used for sample screening to confirm previous dating results or to initiate additional investigations. Two samples from each site, Olzreute and Bodnegg, with the highest distance in the scores plot (see squares in Fig. 2) were subjected to 14C-dating, which confirmed the same age for both contrasting BO- and OL-samples. Therefore, we can conclude that a high degree of carbonization and/or special environmental conditions are responsible for the preservation.
In the next step, the carbonization temperature that the wood was exposed to, was determined based on spectral characteristics using an established prediction model51. It has to be emphasized that the prediction model has been calibrated with fresh charcoals from laboratory experiments51 and applied on recent traditional kiln processes52. The aging effect is not considered in the model. Prediction results for the current sample sets are presented in Fig. 5a and indicate that many charcoal samples from Olzreute (OL) and Bodnegg (BO) were subjected to similar temperatures as kiln samples52. Comparison of standard deviations of all sample sets (Fig. 5b) confirms that the application of the temperature prediction model is limited to charcoal samples that are similar to recent charcoals. High standard deviations indicate that the material departed from material characteristics of the calibrated recent charcoals.
It has to be emphasized that the sample set “rec” comprises samples with a high carbonization degree. It is not evident that such carbonization conditions can be presupposed for all pyrolysis processes to which charcoals were subjected at archeological sites. Reflectance measurements might provide more information about the production temperature, at least for > 400 °C35,48. The boxplots show that most samples BO and OL had been exposed to temperatures > 400 °C, half of them even > 580 °C corresponding to high thermal alteration.
Apart from the high degree of carbonization in the BO and OL samples, environmental conditions have to be considered. According to the archeological information about the sampling sites, charcoals from Bodnegg and Olzreute were embedded in a permanently wet peat. As oxygen availability and accessibility are essential factors for degradation, anoxic environments such as waterlogged ecosystems, peats and river sediments foster preservation15. Anaerobic environmental conditions over longer periods of time, together with the high degree of carbonization, seem to be responsible for the good state of preservation. Lab experiments with a strong chemical oxidative reagent revealed a considerable resistance of charcoals produced at 600 °C against oxidation17.
Sample set with partially high content of mineral compounds (site Speckhau)
The spectral pattern of the sample set from Speckhau (SP) conspicuously indicates the environment where the charcoals were buried. These charcoals originated from a tumulus and were embedded in a mineral matrix. Mineral components (silicates, clay) had permeated the pores and could not be removed without additional chemical methods. According to Huisman et al.53, who investigated remains of a Neolithic settlement, alkalinity is a main factor for charcoal alteration and pedological processes that lead to clay coatings. Figure 6a displays the average spectra of both samples containing high mineral contents (SP-H) and samples with a low content (SP-L).
The high mineral content with intense infrared bands (e.g. Si–O at about 1,030 cm−1)54 obliterates other bands, with organic indicator bands disappearing almost completely. In 21 (SP-L) out of 40 samples the characteristic indicator bands of aging are at least visible. The PCA (Fig. 6b) based on infrared spectra of the whole sample set in the wavenumber range 4,000–2,400 cm−1 and 1,800–400 cm−1 illustrates the conspicuous heterogeneity due to different portions of mineral components. The corresponding loadings plot (Fig. 6c) of the 1st and the 2nd Principal Component (PC) reveals the spectral regions that are responsible for the main variance (explained variance by PC1 and PC2: 74% and 18%, respectively). Besides the characteristic spectral regions that represent the aging process (> 1,200 cm−1), the dominant contribution of mineral components with atom pairs with large reduced mass, such as Si–O, Al–O, Fe–O etc., resulting in stretching bands at low wavenumbers (region < 1,200 cm−1) is obvious. Charcoals with low contents (SP-L) overlap reference samples “C” and pretend an earlier period.
Although the classification into the chronological series is not possible due to the strong influence of mineral components, we can obtain some relevant information about these charcoals by means of spectral data and multivariate analysis: knowledge of the mineral environment, the heterogeneity within the group and the characteristic development of indicator bands that reveal the oxidation process.
Figure 6 displays the PCA with the same settings as in Fig. 1. Rotating the PCA (Fig. 6d, e), it is possible to isolate both sources of variation: the aging in the component 1 and the mineral contaminants in the component 2. However, the estimated age for SP samples, based on their scores in component 1, is overestimated (appearing as “C” or even older).
Ex post we can only hypothesize which factors were responsible for the accelerated aging process. Alkaline environments promote the fragmentation of charcoals and facilitate further alteration13,45,53. For the charcoals under limestone treatment, a lower effect of alkaline conditions was observed as expected42.
The question of why a part of the sample set SP incorporated more mineral components can only be answered by some assumptions. The anatomical structure of wood species might be a reason for the different ingress of mineral components46, charcoal burying and incorporation in the soil, and clay migration during pedogenesis. However, a close exposure to mineral components, especially to clayey materials, influences the aging process. This fact has also been observed for wood55. Rechberger et al.56 documented an accelerated aging process for biochar in acidic soils. Furthermore, a lower degree of carbonization might have led to the accelerated aging15,44 The simple previous inspection of the infrared spectra and identification of a large amount of minerals preclude the use of these spectra in the current models. New models for this kind of samples could overcome this limitation.