Infrared spectroscopy refines chronological assessment, depositional environment and pyrolysis conditions of archeological charcoals

Based on infrared spectral characteristics, six archeological sample sets of charcoals from German (5) and Brazilian (1) sites, covering the time span from the nineteenth century CE to 3950 BCE, were compared to a chronological (present to the fifteenth century BCE) series of Austrian charcoals. A typical chronological trend of several bands (stretch vibrations: O–C–O of carboxylates at 1,585–1,565 and 1,385–1,375 cm−1, C–O carboxylic acids at 1,260–1,250 cm−1) that indicate oxidation and subsequently increasing hydrophilicity (O–H stretch vibration at about 3,400 cm−1) was also contained in the archive samples. Three sample sets fit in the typical band development according to their age. For three sample sets this conformity was not observed. Despite the age of two sample sets (3950–2820 BCE), most charcoals were assigned to the Modern Period. Apart from the high degree of carbonization, anaerobic depositional conditions over a longer period of time seem to contribute to the surprising conservation. Non-removable mineral components in charcoals, as observed in a third sample set, strongly influence infrared band intensities and positions of organic compounds. The role of inorganic components in terms of charcoal aging, and the information we can obtain from spectral characteristics in an archeological context, are discussed.

www.nature.com/scientificreports/ experimental set-ups: feedstock composition, production temperature/carbonization degree and environmental conditions, e.g. soil properties 16,17 are the main influencing factors for black carbon longevity or short-term degradation. These findings were compiled in review articles 18,19 , including uncertainties regarding the stocks and stability of BC, especially in boreal regions 20 . Litterbag experiments during one year in the tropical rainforest confirmed that extensive aromatization due to higher pyrolysis temperatures led to reduced degradability. Furthermore, non-lignocellulosic materials were more susceptible to degradation 21 . Specific formation conditions result in chemical and physical differences between the pyrogenic carbon fraction in biochars and wildfires 22 . Charcoals of low carbonization degree are scarcely found in the environment after only 20 years 23 . Other effects besides degradation, e.g., additional wildfires 24 , erosion and translocation into deeper soil horizons 25 can also cause charcoal remains to disappear. Different approaches of laboratory and field incubation experiments to evaluate the stability of biochar in the environment were the basis for degradation models 26 . The quality of BC fractions and erosion should be included to improve modeling of the fate of BC in soils 27 . Degradability and quality are the main prerequisites for the function and interaction of biochar in soils 28 , such as adsorption behavior 29 , pore size and pore size distribution 30 and the effects on plant growth [31][32][33] . A variety of analytical methods and experimental approaches have been applied to investigate the chemistry of charcoals and alterations in field or lab experiments. Elemental analyses of C, O and H 34 and their ratios in the van Krevelen diagram demonstrate the relation between rising carbon content and increasing temperatures 35,36 .
The O:C ratio of fresh biochar seems to be a reliable parameter for its stability in soils 18 . The determination of benzenepolycarboxylic acids, a molecular marker for BC, is one of the established methods for quantification of BC in different environments, e.g. soils 37 and marine sediments 38 .
Analyses of isotopes can trace the fate of BC from its origin as well as elucidate its behavior and residence time 39 . Removal of contaminants (e.g. non-pyrogenic carbon) might be relevant for the characterization of pyrogenic carbon. In this context, 13 C NMR spectra were used to compare different pretreatments and the efficiency of decontamination 40,41 . For a mass balance of stable carbon, hydrogen pyrolysis was applied to the litterbag samples 21 prior to the 13 C NMR analyses in order to differentiate the stable aromatic carbon fraction from exogenous organic materials, incorporated during exposure in the environment 42 . The formation of new nitrogen heterocyclic components was observed during thermal oxidation of peat samples using 13 C-and 15 N-NMR spectroscopy 43 . 13 C NMR and FTIR spectroscopy were applied to track the progressive dichromate oxidation of charcoals, produced at different temperatures, and natural charcoals, altered in the environment. They revealed that polyaromatic structures formed at temperatures > 400 °C are highly resistant to oxidation, whereas aged samples are more susceptible 17 . Chemical alterations of charcoals along the aging process over several centuries were revealed by their infrared spectral patterns 44 .
The formation of organized polyaromatic domains in charcoals is a basis for reflectance measurements that allow the reconstruction of the production temperature of modern and archeological charcoals formed at 400 °C and above 35,45 . Aging in the environment leaves characteristic mechanical properties of charcoals that can support the identification of original wood species 46 .
Based on previous investigations that have shown a characteristic trend of infrared spectral features with age under similar environmental conditions, it was the objective of this study to corroborate this development using additional sample sets of historical and archeological charcoals from archives (from the nineteenth century CE to 3950 BCE). The classification into an existing chronological series of reference charcoals that cover the time span from the present to a period from the fifteenth to the thirteenth century BCE should verify the conformance of their spectral signature with historical, dendrochronological and/ or 14 C dating. Discrepancies between the spectral signature and the determined age are discussed in the context of literature regarding the influencing factors of aging.

Results and discussion
The relevant bands that were used for sample evaluation are compiled in Table 1. Band positions are indicated according to Smith 47 and Guo and Bustin 48 . The list is limited to visible band maxima. Aging/oxidation lead to interactions (e.g. H-bonds) and subsequently to broadening of bands 49 . 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 incubation 50 , field experiments 42 and chemical oxidation 17 . 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 www.nature.com/scientificreports/ 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 chips 42 . 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 preservation 15 . 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 14 C-dating, which confirmed the same www.nature.com/scientificreports/ 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 model 51 . It has to be emphasized that the prediction model has been calibrated with fresh charcoals from laboratory experiments 51 and applied on recent traditional kiln processes 52 . 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 samples 52 . 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. www.nature.com/scientificreports/ 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 °C 35,48 . 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 preservation 15 . 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 oxidation 17   www.nature.com/scientificreports/ 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 1 st and the 2 nd 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  www.nature.com/scientificreports/ 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 alteration 13,45,53 . For the charcoals under limestone treatment, a lower effect of alkaline conditions was observed as expected 42 .
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 components 46 , 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 wood 55 . 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 aging 15,44 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.

Conclusions
Infrared spectroscopic investigations of charcoals combined with multivariate data analysis have shown that the method provides useful information on their natural aging process, which takes place in a typical way under similar environmental conditions. The characteristic pattern can be related to their age determined by other methods ( 14 C dating and dendrochronology). The question is whether this fact can be used in the future as an additional dating tool. Deviations of this general pattern are observed if the degree of carbonization, environmental conditions and the ambient matrix disturb the steady transformation of the material. The discrepancy between the age determined by dating methods and the spectral features indicates additional influencing factors on the aging processes. Therefore, sampling of archeological charcoals should always be paralleled by a comprehensive survey of the sampling site, including a detailed environmental, especially pedological, report to support data interpretation. The infrared spectra alone already provide clues to some of these outliers, such as high carbonization degree and embedded mineral contaminants. In the future, if age and carbonization degree are well known, FTIR spectroscopy can be expected to provide profound information on the long-term soil conditions.
Due to the advantages of FTIR spectroscopy mentioned above, it is possible to characterize huge sample sets. Based on multivariate evaluation, the homogeneity of charcoal samples from a sampling site can be checked. In this case, infrared spectroscopy is a useful tool for a preliminary assessment in order to select additional, more expensive analyses. Future work will have to separate spectral characteristics, describing the degree of carbonization 51 and those indicating the aging processes 44 .
In further steps, prediction models for the age of charcoal based on spectral features might be worked out, if carbonization degree and field conditions are known.

Materials and methods
Sample sets. Details of charcoal samples, comprising reference samples from Austria 44 and samples from Germany and Brazil, are compiled in Tables 2 and 3. Charcoals were not excavated or dated specifically for this study.
Sample origin and site conditions. Table 3.
Sample preparation. The size of collected charcoal particles was 5 to 15 mm. Charcoal particles from archives were dried at 65 °C. Attached mineral components were manually removed from the surface of charcoal particles using a tweezer and scalpel with the help of a magnifying glass. The samples were stored in small sealed glass jars. For infrared spectroscopic analyses, charcoals were milled with pestle and agate mortar to a particle size < 200 µm and stored in Eppendorf vials at room temperature. www.nature.com/scientificreports/ Fourier transform infrared (FT-IR) Spectroscopy and data evaluation. Spectra were recorded in the mid wavenumber range from 4,000 to 400 cm −1 using the attenuated total reflection (ATR) mode. Instrument: Tensor 27 (diamond crystal of the BRUKER Helios FTIR micro sampler with a spatial resolution of 250 µm). Three to four replicates per sample were recorded (32 scans at a spectral resolution of 4 cm −1 ), corrected against ambient air as background and unit vector normalized with the integrated software (OPUS 7.2). The averages of normalized replicates were subjected to data evaluation using the Unscrambler X 10.1 (CAMO). The wavenumber regions 4,000-2,400 cm −1 and 1,800-400 cm −1 or 4,000-2,400 cm −1 and 1,800-1,100 cm −1 were selected for Principal Component Analyses (PCA) and SIMCA (Soft Independent Modeling of Class Analogy). SIMCA provides the classification of new samples to previously defined classes regarding spectral characteristics (or other parameters). The membership of assigned samples is based on conformity or similarity to the determined classes with a given level of significance 61 . The prediction of pyrolysis carbonization used a model published in Tintner et al. 51 Rotation of PCA was performed using the Varimax criterion 62 .

Data availability
Original data are available upon request.

German sample sets
Wittnau WI historic charcoal production site in the southern Black Forest 58 with two usage periods (fifteenth to seventeenth century CE and seventeenth to twentieth century CE); soil: acidic (parent material is Gneis-Granitic) and well drained on a slope at 645 m a.s.l

Iznang IZ
Domestic energy wood remains from fire places and remains of burnt-down houses from a cultural layer in and above permanently waterlogged conditions due to groundwater (Horgen settlement at the Untersee near Radolfzell/ Lake Constance 59 ), dendrochronologically dated settlement with its beginnings at 3,270 BCE and with a second building phase around 3,250 BCE Bodnegg BO Charcoals from fire places (settlement according to 14 C-dates at the shore of a small lake during the neolithic period 3950-3650 BCE), wetland site with permanent wet conditions; samples embedded in slightly mineralized, but permanently wet peat Olzreute OL Charcoals from excavated domestic fire places (Goldberg-III-settlement near Bad Schussenried in Upper Swabia dendrochronologically dated to 900 BCE 60 ; one or two more building phases within the period around 2850-2820 BCE; the neolithic village was built at the shore of a small lake); samples incorporated and overgrown by peat layers after silting up of the lake www.nature.com/scientificreports/