“Millet” is an umbrella term used to describe different grass species that share a set of structural and functional characteristics. They are small-seeded cereals that produce a large number of grains in their inflorescences, belonging to the C₄ photosynthetic pathway, which is typical of warm climatic areas. Millets had different centers of domestication in Asia and Africa, and then progressively expanded to other continents. In particular, Panicum and Setaria were first cultivated in northern China in the early Neolithic (Lu et al., 2009a; Stevens et al., 2021), and during the 2nd millennium BC they spread across Eurasia (Miller et al., 2016; Martin et al., 2021) to Europe (Motuzaite-Matuzeviciute et al., 2013; Filipović et al., 2020). The dispersal of both Chinese millets is associated with social, economic, and human and animal food changes (Lightfoot et al., 2013; Wang et al., 2019; Kirleis et al., 2022; Sun et al., 2024). Millets have recently attracted keen interest in archaeological science with increasing taxonomic identification of preserved macroscopic remains (archaeobotany) but also thanks to recent methodological advances and multi-proxy approaches: phytoliths (Lu et al., 2009b), new analytical biomolecular techniques applied to sediments and artifacts to detect miliacin, a distinctive broomcorn millet compound (Jacob et al., 2008), and stable isotope analysis of human and animal bones (Tafuri et al., 2009).

Grains of Panicum and Setaria have essentially the same structure (Lorenz, 1977). They are formed by a caryopsis encapsulated by two membranous and robust husks: a palea and an upper lemma. The caryopsis is divided into the pericarp, the germ, and the endosperm (Figure 1). The pericarp is the outermost protective layer, made up of one-layered aleurone and endosperm (Serna-Saldivar and Espinosa-Ramírez, 2018; Akanbi et al., 2019). The germ, which includes the embryo and scutellum, is comparatively larger in relation to grain size than in other cereals and also varies between different species of millets (Akanbi et al., 2019). The endosperm contains two different types of cells and starch shapes and sizes (McDonough and Rooney, 1989; Taylor and Belton, 2002; Bean et al., 2018), namely vitreous/corneous/glassy cells and floury/opaque cells. The outer layers of the endosperm are made of vitreous cells which are relatively compact, with large polygonal starch granules pressed together and frequent indentations owing to this dense mesh, and a high concentration of kafirin protein bodies (Watterson et al., 1993; Krishna Kumari and Thayumanavan, 1998; Bean et al., 2018). Proteins become less frequent toward the grain's inner core, which has a less dense and more scattered matrix. Starch-rich cells contain small spherical starch grains and there is a less dense protein mesh in the innermost layers of the grains (Krishna Kumari and Thayumanavan, 1998; Bean et al., 2018). Both Panicum and Setaria contain different proportions of starch granules, the polygonal type being more common in Setaria than in Panicum and vice versa for the spherical type (Krishna Kumari and Thayumanavan, 1998). The major dietary fibers are cellulose, arabinoxylans (AX), b-glucans, other hemicelluloses, pectin, and small amounts of lignin (Serna-Saldivar and Espinosa-Ramírez, 2018). Compounds inside the grains are distributed unequally: husks are particularly rich in cellulose, while the germ contains more lipids than the endosperm.

Understanding how seeds react to charring has been a frequent concern of the archaeobotanical discipline. Several studies have addressed experimental charring of common millet and foxtail millet grains in controlled conditions to understand how charring modifies their morphology and structure (Märkle and Rösch, 2008; Yang et al., 2011b; Motuzaite-Matuzeviciute et al., 2012; Walsh, 2017; Dong et al., 2022; Liu et al., 2023) and also their carbon and nitrogen isotope ratios (Yang et al., 2011a; Fraser et al., 2013a; Dong et al., 2022). Although in all of these cases different temperatures, times and firing atmospheres have been used to answer the specific questions raised in each paper, the results converge and complement each other. When heated, grain size of both cereals increases from before reaching charring threshold temperatures (Yang et al., 2011b), simultaneously with weight loss (Märkle and Rösch, 2008; Yang et al., 2011b). At temperatures up to 200°C, no structural changes occur in the starch granules (Yang et al., 2011b). The change from a fresh to charred state occurs in both species around 220–235°C in short exposures of 1 h to 4 h (Märkle and Rösch, 2008; Motuzaite-Matuzeviciute et al., 2012), and at lower temperatures after longer exposures: 225°C at an exposure time of 4 h (Märkle and Rösch, 2008) or 24 h at 215°C (Dong et al., 2022). Although distorted, the grains remain identifiable to species up to around 250°C (Walsh, 2017). At higher temperatures grains tend to burst (Märkle and Rösch, 2008; Walsh, 2017; Dong et al., 2022), leading to the disintegration of the grains into ash from 320°C after 4 h of heating (Märkle and Rösch, 2008).

The grain analyzed in the experiments was obtained from traditional varieties cultivated by Celia in the village of Busto (Rois, A Coruña, Spain) and by Luis in the village of Rebollar (Degaña, Oviedo, Spain) in 2012 as part of an ethnobotanical research project (Moreno-Larrazabal et al., 2015; Teira Brión, 2022), and sown annually in the following years. The experimental fields where the traditional varieties were grown were located in Artes, Galicia, Spain (Mercator: 42°34′59^′′N 9°00′32^′′W), within the biogeographic and climatic Atlantic region. The field is a terraced plot with a long tradition of intense organic manuring (cow, humus, and farmyard manures) and the addition of some chemical fertilizers and rabbit manure from the 1980s until 2013, when it remained uncultivated till the date of the experiment. Millets were sown separately in different fields but sharing very similar growing conditions in terms of soil, drainage, insolation, and humidity. Mechanical means were used to plow the soil, sowing was done by hand and the field was not manured. The plants grew between the 13th of June and September 2020. Common millet and foxtail millet were harvested 85 days and 94 days after sowing, respectively, once the grains were fully ripe. The seeds were naturally sun dried for 5 days and then stored in a traditional granary.

The experimental charring methodology was based on previously established protocols for other crop seed types (Fraser et al., 2013a; Charles et al., 2015; Nitsch et al., 2015; Styring et al., 2019; Stroud et al., 2023). A total of 2520 grains were charred in bulk samples of 10 grains each. The different charring conditions were repeated on four sets of samples including a sample of husked and huskless grains of common millet and foxtail millet. The first stage of the experiment comprised grains charred in an oven at four different temperatures, namely at 30-degree intervals between 200 and 290°C. Subsequently, a second round of samples were charred at different temperatures, namely ten-degree intervals between 220°C and 260°C to obtain more detailed data on the processes occurring within this range. As a result, the final experiment comprised seven different charring temperatures: 200, 220, 230, 240, 250, 260, and 290°C. At each temperature three different heating atmospheres were tested: high oxygen availability in uncovered glass beakers; medium oxygen availability with grains covered in 10 cl of sand; and low oxygen availability with grains wrapped in aluminum foil and covered in 25 cl of sand. Each set of grains was charred for 2, 4 or 6 h under each temperature and atmosphere combination.

The grains were charred in a Gallenkamp Plus II oven and temperature fluctuations were monitored using a Datalogger Thermometer with three thermocouples, one located inside the recipients and the other two at the top and at the bottom of the oven respectively (graphs are available as Supplementary material Data Sheet 1). After charring, the grains were cut crosswise with a scalpel to obtain a view of their internal structure and were photographed with a Leica Z6 APO stereo microscope x5.6-x36 magnification with a Leica DFC495 microscopy camera using LAS v4.12 software. Scanning electron microscopes belonging to Research Laboratory for Archaeology and the History of Art (RLAHA) at University of Oxford (JEOL JSM-5510) and the Research Support Infrastructure Network -RIAIDT at Universidade de Santiago de Compostela (ZEISS EVO LS15) were used to image the internal structures of the grains. To record the degree of charring of the outer and inner surface of the grains as a function of time and temperature, a color code system similar to that developed in previous work (Charles et al., 2015; Holguin et al., 2022; Stroud et al., 2023) was created.

To assess how charring affects isotopic values, only grains for the same ear of each crop were analyzed to avoid possible disparity of results between different plants, and to obtain an average representation of the single ear. A total of 300 grains of Panicum and the same amount of Setaria were selected randomly using a riffle box. Bulk samples of 10 grains, husked—caryopses with paleae and upper lemmas still attached—or huskless, were analyzed. Batches of uncharred fresh grains, and grains charred at optimal conditions for preservation (230 and 250°C after 2 and 4 h respectively and under reducing condition) were analyzed (see Section 4.1). This included batches of husked and huskless grains, to understand the effect of charring on the isotopic composition of grains and grains with husks. The same apparatus for temperature measurement and charring specified in the Section 3.2 was used. All bulk samples were weighed before and after charring to determine their mass loss (see Supplementary material Table 1).

Analysis of the stable carbon and nitrogen isotopic compositions of the samples were conducted at the Research Laboratory for Archaeology and the History of Art at the University of Oxford. Each sample was homogenized within Eppendorf tubes with two steal balls inside using a QIAGEN TissueLyser LT mill for 1 h at an oscillation of 50 pulses per second. The powder obtaining was weighted into tins containers using a Sartorious R200D balance. Carbon and nitrogen isotopic compositions of the common and foxtail millet samples were determined in separate runs on a SerCon Europa EA-GSL sample converter connected to a Sercon 20–22 stable isotope gas-ratio mass spectrometer. Isotopic compositions were calibrated relative to VPDB and AIR scales using the in-house DL glutamic acid monohydrate (δ¹³C = −10.67 ± 0.07‰, δ¹⁵N = −1.73 ± 0.21‰) and Leucine (δ¹³C = −28.31 ± 0.03‰, δ¹⁵N = 6.33 ± 0.07‰) standards for carbon, whereas for nitrogen the glutamic acid and a seal collagen standard (δ¹³C = −12.54 ± 0.13‰, δ¹⁵N = 16.14 ± 0.09‰) were used. To understand precision, accuracy and overall uncertainty check standards were run as well as a duplicate of each tenth sample as minimum. Alanine (δ¹³C = −27.18 ± 0.16‰, δ¹⁵N = −1.50 ± 0.07‰), EMA–P2 (δ¹³C = −28.19 ± 0.14‰, δ¹⁵N = −1.57 ± 0.19‰) Pea protein (δ¹³C = −26.47 ± 0.08‰, δ¹⁵N = 1.97 ± 0.10‰) and seal collagen were used as check standards during the carbon runs, while for the nitrogen run leucine was used instead of seal. Precision, accuracy and overall uncertainly were calculated following Szpak et al. (2017). For the δ¹³C values precision [u(Rw)] was ± 0.09‰ and accuracy or systematic error [u(bias)] was ± 0.18‰. For δ¹⁵N precision was ± 0.14‰ and systematic error [u(bias)] was ± 0.18‰. Overall uncertainty (u_c) was ± 0.20‰ for carbon and ± 0.23‰ for nitrogen. All uncorrected raw data from sample and standard measurements are included in Supplementary material Table 2. All statistical tests, calculations, and graphing was conducted in RStudio 2023.12.1, and R version 4.3.3.

At 200°C grains of Panicum showed no internal or external evidence of starch transformation into solid carbon material resultant from the thermal decomposition of organic matter and release of volatile compounds (Figures 2–4). At 220°C signs of transformation began in the outermost part of the endosperm. Although changes in color took place and a vitrification process started, the cell structure of the core of the Panicum endosperm was very similar to that of the uncharred grains after 2 h. At 220°C the appearance of the grains is vitreous and internal indentations appear. Most relevant changes started at around 230°C from short 2-h exposures. Carbon material and inner voids formed in the outer layers of the endosperm, whereas some areas inside remained dark-brown color, preserving the cells walls (Figure 3). After 4 or 6 h the whole grains were charred. However, germ cells did not show clear evidence of black material. At about 240°C degrees the internal structure and external appearance were very similar to those of well-preserved archaeobotanical grains. Cracks formed regularly in some cases, and surface deformation was evident in a large number of seeds. At 250°C, deep cracks appeared, the caryopsis were heavily distorted, and grains began to form aggregates. The caryopsis compressed by the husk showed visible loss of tissues in irregular shapes. At 260°C (and at higher temperatures, 290°C) the grains burst and were hardly identifiable, resembling popcorn-shapes or amorphous masses. All of these changes occurred in oxidizing, semi-reducing and reducing atmospheres but with some differences –see Figure 2. Differences in grain structure between husked and huskless grains at lower temperatures were observed. At higher temperatures, the compression caused by paleae and upper lemmas resulted in the explosion of the external surface from 240°C onwards and the formation of cracks and more stubby grains. From 260°C pyrogenic material release produced eccentric forms in husked grains.

The same pattern was observed in Setaria, but at slightly higher temperatures (Figures 2, 5). Black material was not detected at 200°C or at 220°C in short exposures. The internal structure of the Setaria endosperm cells underwent changes at around 230°C. At this temperature the floury cells still preserved their appearance, even with some indentations caused by stress due to chemical changes—e.g., water or lipid evaporation, whereas vitreous cells became char. After 2 h of heating, only the external layers of the endosperm showed black color, while internal cells remained still uncharred. Between the charred layers of floury cells voids and indentations formed, which led to the beginning of the deformation of the caryopses. The whole endosperm after 4 h or 6 h, depending on the atmosphere, was black in color (Figure 2). The grains were completely made up of pyrogenic material at 240°C. The deformation intensified at 250°C, but grain distortion was more tangible at temperatures above 260°C and higher temperatures, with visible cracks and loss of internal material, but on a smaller scale than observed in Panicum. Low/medium/high oxygen availability atmospheres drove similar grain modification, although grain color darkened quicker under the lower oxygen/higher reducing conditions (Figure 2). As also observed in Panicum, germ cells were more resistant to heating than endosperm cells.

This study shows that the most relevant changes due to charring in Panicum and Setaria grains occur between 230 and 250°C over short time exposures of 2 to 6 h. These charring conditions produced good, low distortion grains which resembled well preserved archaeological grains. Changes in both millets happen in a narrower window of temperatures than observed in other cereals (Charles et al., 2015; Stroud et al., 2023). This could also relate to the tendency of millet to form aggregates (Teira-Brión et al., 2024), possibly due to the layers that make up the caryopsis, the two types of cells and mesh of proteins that constitute their endosperms, their shape and molecular composition, and how they react distinctively to charring. In other cereals such as wheat, barley, rye, rice or oat, the endosperm contains a single type of floury cells, which could result in a similar response to charring and seed deformation. Conversely, millets have distinctive endosperm cells.

The conversion of starch present in the endosperm and cellulose in the pericarp initiates around 220°C (Braadbaart et al., 2004; Braadbaart, 2008). This transformation affects each layer differently. When charring takes place in the floury cells, there is an expansion and bursting of the grains, with cracks and loss of material to the outside. By contrast, the glassy cells layer when subjected to 220°C heating show some signs of vitrification but still preserve tissue structures. When the grains are heated at 230°C for 2 h, voids begin to form between the glassy and floury cells, which still preserve starch granules. However, tissue charring only affects the outermost layers and the cells in the core of the endosperm still contain uncharred tissues; this also occurs in the germ, which is more resistant to heating. The grains heated at 230°C still contain a high proportion of carbohydrates, which is indicative of a starch-rich content (Teira-Brión et al., 2024).

The formation of indentations on the grain surface appeared from 230°C, being a constant feature of grains exposed to 240°C and above. From 250°C the scars are deep and at 260°C, the bursting of the grain is the most characteristic external feature. This is the temperature at which “popcorn” shapes are produced. Some studies of coffee beans suggest that these cracks start to form after 8 min of charring (Charles et al., 2015). Other work indicates that mass loss increases before 40 min at the same temperature (Braadbaart et al., 2004; Braadbaart, 2008). Hence, it is possible that the number of indentations in the experimental millets is correlated to the longer duration of this experimental charring (≥2 h).

As for the temperature window in which we can find identifiable grains, it has been set between 230 and 250°C for the conditions and variables in this experiment. The lower limit obtained for charring differs from that determined by Dong et al. (2022), which suggests that charring would occur from 200°C. However, in this and a related study (Teira-Brión et al., 2024), although changes occurred in the cellular tissues, we were not able to detect charring of the grains at 200°C when heated for 2 to 24 h. The upper limit of 250°C for good grain preservation has also been considered in experiments on Panicum miliaceum and Setaria italica, albeit with larger and less precise temperature gaps of 50°C (Yang et al., 2011b; Walsh, 2017).

The germ was found to be more resistant to charring than the endosperm. Only at temperatures of 260°C was it fully bonded with the endosperm. This could be the origin of a scarce presence of embryos in the millet grains come from the archaeological assemblages, which would have been disarticulated or fragmented (Martín-Seijo et al., 2020). This aspect may also be significant in the isotopic values since the elemental composition in embryos and endosperm differs (see Section 2.2).

Huskless grains are usually interpreted as a sign that they were fully processed and stored clean before consumption (Tereso et al., 2013; Filipović et al., 2019). However, the absence of paleae and lemmas does not necessarily indicate that they were stored without husks. It has been observed that glumes are typically more susceptible to destruction at lower temperatures compared to grains (Boardman and Jones, 1990). In our experimental design, the paleae and upper lemmas demonstrated better resistance to changes in morphology, particularly under oxidizing conditions, but they became extremely fragile and prone to breakage. Differences in the preservation of husks in some archaeological assemblages (e.g., Martín-Seijo et al., 2020) may be explained by the conditions of charring or the recovery methods.

Minor deviations in δ¹³C values of the charred samples compared to the fresh samples are similar to results obtained in other work on isotopes in millets (Yang et al., 2011a; Dong et al., 2022; Wang et al., 2022). In this study, the mean offset has been established as an average value of ±0.16‰ relative to the bulk samples of fresh grains taken from the same inflorescence for Panicum and Setaria.

Since different compounds deplete into specific isotopes at different temperatures, the timing of this reaction in the grains and which compounds are most affected could explain the pattern observed in the charring offset. Grain composition and chemical changes can make a difference in that carbohydrates comprise approximately 74% of the grain, whereas lipids make up 5% (Ravindran, 1991). Starches and sugars have ¹³C:¹²C values similar to that of the carbon initially fixed in photosynthesis, whereas lipids are up to 15‰ higher and cellulose tend to be approximately 2‰ lower (Ehleringer, 1981). The ¹³C depletion observed during charring may be partially attributed to the loss of cellulose, a carbohydrate that is more prevalent in seeds than other compounds with lighter isotopic values, such as lipids. Lipids are known to volatilise at temperatures >150°C, implying that their residues will be higher when charring initiates (Czimczik et al., 2002). The negative δ¹³C offset of the Panicum husked grains charred at 230°C from the uncharred material has also been observed by other studies among C₃ cereals (bread wheat, einkorn, rye) and millets (common millet) (e.g. Nitsch et al., 2015; Dong et al., 2022; Stroud et al., 2023). It is plausible that these values may be linked with a greater prevalence of uncharred tissues, when the temperature is still insufficient to transform all organic compounds.

The δ¹⁵N values of the charred samples tend to a higher offset from the uncharred samples than do the δ¹³C values, with δ¹⁵N values tending to be more positive in charred material. This correlation has been widely reported in different studies on other crops, but not in all species and experiments (Fraser et al., 2013a; Nitsch et al., 2015; Styring et al., 2019; Dong et al., 2022; Wang et al., 2022; among others). Variation in δ¹⁵N values can have an impact when considering fractionation processes in body tissues, which has an effect between diet and bone collagen in the order of approximately +3‰ (Reitsema, 2013). Therefore, when comparing bones with charred plant remains, the isotope effect must be considered to determine the trophic effect more accurately (Fraser et al., 2013b). Nitrogen is also key for understanding soil nitrogen composition and potential manure supply to the soil (Bogaard et al., 2013). The δ¹⁵N offset must be taken into account in the interpretation of manuring levels in the archaeological samples.

A relationship between nitrogen isotopic effect in millets and their different biochemical subtypes within the C₄ photosynthetic pathway has been proposed (i.e., Varalli et al., 2023). These authors posit that charred millets of the NAD-ME subtype, such as Panicum miliaceum and Cenchrus americanus, exhibit a higher isotopic offset in δ¹⁵N values compared to those from NADP-ME subtype, such as Setaria italica and Sorghum bicolor (Varalli et al., 2023). Conversely, the results obtained within the parameters of the present experimental design differ from those of the previous study. Specifically, the offset for charred Panicum huskless has been shown to increase less than that for the equivalent Setaria batch. Consequently, this contrast suggests the need for further investigation to determine the impact that the biochemical NAD-ME and NADP-ME subtypes have on the isotopic effect. Additional factors must be considered, such as variations in the compound composition of each millet species—even within the same species—which can lead to different isotopic values in the grains. Furthermore, differences in temperature and heating exposure between experiments may reflect different stages of compound depletion, thereby affecting isotopic values.

Although no specific values have been measured for paleae and lemmas, experimental results indicate that grains with intact bracts have lower isotopic values than grains without. Previous works have reported distinct δ¹³C and δ¹⁵N values among leaves compared to grains in Setaria (Lightfoot et al., 2016, 2020). As paleae and lemmas develop before grain formation (Motuzaite-Matuzeviciute et al., 2012), their values are also associated with different growth stages, along with the different components that make up each constituent. Just as each part of a plant has distinct isotopic values, the isotopic signatures of humans and livestock would differ depending on which parts of the plant they consume.