Winter wheat (Triticum aestivum, cultivar JB Asano) samples were taken from an ongoing long-term field experiment in NE Germany. A comprehensive site description and details on this long-term field experiment can be found in Puppe et al. (2021). We took leaf samples from harvest-fresh plants at the end of July 2018, which were used for heating experiments (subsection Heating of Winter Wheat Leaves) and phytolith extraction (subsection Phytolith Extraction and Treatment).

Heating of winter wheat leaves was performed following a protocol modified from Dabney et al. (2016): i) leaf samples were carefully cut (approx. 0.5 × 1.5 cm): with a razor blade, transferred to a microscope slide, and covered with a cover slip, ii) the samples were heated in a muffle furnace, whereby the temperature was increased gradually until the intended temperature (120 , 170, 220, 270, 320, 370, or 420°C) was reached and then kept for 6 h, iii) the samples were allowed to cool slowly down in the muffle furnace (off-state, furnace door closed) over night before microscopic analyses (subsection Microscopic Analyses).

Phytoliths were extracted from wheat leaves as follows: i) 10 g of dry plant material was weighed, ii) organic matter was oxidized using H₂O₂ (30%) and HNO₃ (65%) at 80°C until the reaction subsided, iii) carbonates were dissolved by the use of hot HCl (10%, 30 min, 80°C), and iv) the obtained phytoliths were washed with distilled water and dried at 105°C. Phytoliths were heated in the muffle furnace for 6 h at different temperatures, i.e., 120 , 170, 220, 270, 320, 450, and 600°C.

Additionally, we analyzed potential auto-fluorescence of untreated phytoliths in soil samples randomly selected from retained samples of a previous study (Kaczorek et al., 2019). These soil samples with (relatively) high phytolith contents were screened under the light microscope until recognizable soil phytoliths were detected. Thus, these phytoliths were neither extracted nor heated, but directly observed with light as well as fluorescence microscopy (subsection Microscopic Analyses).

Light and fluorescence microscopy was performed with an inverted epifluorescence microscope (Nikon ECLIPSE Ts2R-FL) equipped with a fluorescence kit (Nikon C-LEDFL470, excitation at 470 nm). Micrographs were taken at magnifications 100 × and 200 × with a high-resolution digital camera (Nikon DS-Fi3) using the software Nikon Nis-Elements (version 4.60). Micrograph processing (sharpness, contrast) was performed with the software ImageJ (version 1.52, http://imagej.nih.gov/ij).

For elemental analyses (Si and carbon) of phytoliths from selected treatments (control and 450°C) a SEM (ZEISS EVO MA10) equipped with an element detector for EDX (Bruker QUANTAX EDS) was used. All SEM-EDX scans were performed using samples sputter-coated with gold (coating thickness approx. 5 nm) and quantitative results were displayed in percent. SEM-EDX analyses were run for all occurring elongate phytoliths, short cells, and silicified tissue fragments (cf. Kaczorek et al., 2019). We performed SEM-EDX analyses without measuring reference materials for calibration as we intended to use carbon contents of phytoliths solely for a statistical grouping of phytoliths via multivariate cluster analysis (subsection Statistical Analyses).

For transmission FTIR 0.5 mg of extracted phytoliths of selected treatments (control-phytoliths and phytoliths heated at 220, 270, 450, or 600°C) were diluted in 99.5 mg potassium bromide (KBr, Merck). Subsequently, these 100 mg phytolith-KBr mixtures were finely ground in an agate mortar and pressed to pellets (5 repetitions each), which were used for transmission FTIR measurements. Transmission FTIR spectra were recorded as 16 co-added scans between wave numbers (WN) 4,000 and 400 cm⁻¹ at a resolution of 4 cm⁻¹ and related to background spectra of ambient air using a Biorad FTS 135 spectrometer. All spectra were smoothed (boxcar moving average algorithm, factor 11), and corrected for baseline shifts using the software WIN-IR Pro 3.4 (Digilab, United States). Mean spectra were calculated from five repetitions using the Average tool of Win-IR Pro. Interpretations of IR spectra were performed by means of standard literature (e.g., Van der Marel and Beutelspacher, 1976) and publications specifically focusing on BSi analyses (e.g., Sandford, 2003; Fraysse et al., 2006; Loucaides et al., 2010; Cabanes et al., 2011; Yan et al., 2014; Puppe and Leue, 2018).

For statistical analyses specific peaks of eight functional groups were used: i) O-H groups around WN 3430 cm⁻¹, ii) the sum of aliphatic C-H groups between WN 3000 and 2,800 cm⁻¹, iii) the sum of C=O groups at WN 1709, 1,645, 1,632, and 1,609 cm⁻¹, iv) the sum of C-O and N-H groups around 1,390 cm⁻¹, v) Si-O groups at WN 1092 cm⁻¹ (asymmetric Si-O-Si stretching), vi) Si-O groups at WN 957 cm⁻¹ (Si-OH bending), vii) Si-O groups at WN 797 cm⁻¹ (symmetric Si-O-Si stretching), and viii) Si-O groups at WN 469 cm⁻¹ (Si-O-Si bending).

Monotonic relationships were analyzed via Spearman’s rank correlation. For grouping of phytoliths based on carbon contents obtained from SEM-EDX analyses, a multivariate cluster analysis (Ward’s minimum variance method with squared Euclidean distances) was applied. Differences between means were tested with the Kruskal–Wallis analysis of variance (ANOVA) followed by pairwise multiple comparisons (Dunn’s post-hoc test). Statistical analyses were performed using the software package SPSS Statistics (version 22.0.0.0, IBM Corp.).

Fluorescence microscopy revealed strong auto-fluorescence of organic matter in the analyzed leaves, which decreased with increasing temperature and the accompanied gradual combustion of organic matter (Figure 1). At temperatures up to 220°C organic matter auto-fluorescence masked the detection of auto-fluorescence in phytoliths. At higher temperatures (270–420°C) auto-fluorescent phytoliths were clearly visible.

While auto-fluorescence of organic matter in untreated (no extraction, no heating) soil samples was clearly visible again, phytoliths in these samples showed no auto-fluorescence at all (Figure 2).

Analyses of phytoliths extracted from plant samples indicated ambiguous results at temperatures ≤320°C with some phytoliths showing strong auto-fluorescence and others showing only weak or no auto-fluorescence (Figure 3). At higher temperatures, i.e., 450 and 600°C, all phytoliths showed auto-fluorescence.

Elemental analyses showed that control-phytoliths were characterized by Si and carbon contents ranging between 19.7–39.4% and 14.7–53.2%, respectively (Figure 4). The grouping of control-phytoliths based on carbon contents resulted in four clusters in the first step. Cluster 1 comprised phytoliths with carbon contents between 23.9–29.4%, cluster 2 was characterized by phytoliths with carbon contents between 14.7–20.2%, cluster 3 included phytoliths with carbon contents of 37.7–39.3%, and cluster 4 comprised phytoliths with carbon contents of 42–53.2%. In the second and third steps clusters 3 and 4 and clusters 1 and 2 were joined, respectively. In the fourth (last) step these clusters were finally joined. The grouping of phytoliths heated at 450°C showed less distinct results as heating strongly affected carbon-based functional groups (subsection FTIR Spectroscopy). The grouping of control-phytoliths indicated that most phytoliths were classifiable as lumen phytoliths (62%) with mean carbon contents of about 23%, while cell wall phytoliths represented 38% of the control-phytoliths. Cell wall phytoliths were further divisible into recognizable (in terms of phytolith morphology) cell wall phytoliths (19%, carbon content of about 40%) and silicified tissue fragments (19%, carbon content of about 47%). Carbon contents of lumen and cell wall phytoliths differed statistically significant.

Transmission spectra of control-phytoliths differed from those of heated phytoliths (Figure 5). The strongest Si-O band of control-phytoliths occurred at WN 1102 cm⁻¹ (asymmetric Si-O-Si stretching) and was shifted towards WN 1092 cm⁻¹ with higher temperatures. In this context, also the band of silanol groups at WN 957 cm⁻¹ (Si-OH bending) became more prominent in the spectra. The signal intensity at WN 795 cm⁻¹ (symmetric Si-O-Si stretching) remained constant, while the band at WN 465 cm⁻¹ (Si-O-Si bending) slightly increased after thermal treatment. In the spectra of control-phytoliths absorption bands assigned to organic matter were prominent in the regions WN 3000–2,800 cm⁻¹ (aliphatic C-H), WN 1709–1,609 cm⁻¹ (C=O, C-O, N-H, C=C, O-H groups of adsorbed water), and around WN 1395–1,382 cm⁻¹ (C-O and N-H). These bands were quite smaller, but did not completely disappear in the spectra of the samples treated with higher temperatures. Signal intensities of bands assigned to free O-H groups and O-H groups of minerals (around WN 3800–3,400 cm⁻¹) decreased with increasing temperature indicating dehydration during heating. While signal intensities of aliphatic C-H groups strongly decreased at temperatures ≤220°C, the intensities of all other bands decreased more or less gradually with increasing temperatures.

The results described above were also reflected in our statistical analyses (Figure 6). With increasing temperature we found a decrease of the summed peaks of i) the O-H groups (r _s = -0.400, p = 0.505), ii) the C-H groups (r _s = -0.700, p = 0.188), iii) the C=O groups (r _s = -0.800, p = 0.104), and iv) the C-O and N-H groups (r _s = -0.900, p = 0.037). Contrary, with increasing temperature we found an increase of (v) the Si-O groups at WN 1092 cm⁻¹ (r _s = 0.900, p = 0.037), vi) the Si-O groups at WN 957 cm⁻¹ (r _s = 0.900, p = 0.037), vii) the Si-O groups at WN 797 cm⁻¹ (r _s = 0.700, p = 0.188), and viii) the Si-O groups at WN 469 cm⁻¹ (r _s = 0.600, p = 0.285).

Our findings indicate that organic compounds associated with phytoliths induce phytolith auto-fluorescence in phytoliths treated at temperatures below approx. 400°C. In phytoliths treated at higher temperatures, i.e., 450 and 600°C, phytolith auto-fluorescence is then mainly caused by molecular changes of phytolith silica, because organic compounds are largely, but not entirely, combusted at these temperatures. In general, our observations are in line with the results of previous studies. Li and Delvaux (2019), for example, examined the Si release from rice straw biochar at different temperatures of pyrolysis. They found an increase of amorphous silica (phytoliths) in biochar with increasing temperatures between 150 and 500°C, but at higher temperatures (500–700°C) the content of amorphous silica decreased. While hydrated amorphous silica (SiO₂·nH₂O) is relatively stable at temperatures below 150°C, SiO₂·nH₂O is converted to amorphous silica by dehydration at higher temperatures (150–500°C). At temperatures higher than 500°C amorphous silica starts to convert to mineral polymorphs of silica, i.e., cristobalite and tridymite (see Li and Delvaux, 2019 and references therein). These temperature-dependent molecular changes in the Si-O groups of phytoliths are also reflected in our FTIR results. However, such heating effects are not restricted to biochar, but have also been observed in mineral soils, for example, with consequences for Si cycling (Schaller and Puppe, 2021).

The association of phytoliths with organic compounds has been shown in several previous studies. In fact, organic compounds like proteins or glycoproteins have been found occluded in phytoliths (e.g., Elbaum et al., 2009; Alexandre et al., 2015; Kumar et al., 2020), while other studies have shown that phytolith surfaces are also associated with cell wall components like cellulose, cutin, hemicellulose, or lignin (e.g., Law and Exley, 2011; Soukup et al., 2017; Zancajo et al., 2019). These differences in phytolith composition are directly linked to the origin of phytoliths: While cell wall phytoliths are associated with a carbohydrate matrix, lumen phytoliths seem to contain more proteins and glycoproteins than cell wall phytoliths with consequences for phytolith dissolution kinetics and carbon sequestration (see the authoritative reviews of Hodson, 2016; Hodson, 2019, and references therein).

Nguyen et al. (2021) found the amount of organic carbon associated with phytoliths to decrease with increasing pyrolysis temperature. It is noteworthy that they still detected 3.4 and 1.6% of organic carbon in phytoliths heated at temperatures of 800 and 900°C, respectively. These findings are underlined by our FTIR results regarding functional groups assigned to organic matter. While aliphatic substances (C-H) showed a weak thermal stability, other functional groups representing organic matter (C-O, N-H, C=O, C=C) were relatively stable and detectable even at a temperature of 600 °C. In this context, associations between C-H enriched organic matter and phytogenic silica might be less stable than associations between C=O enriched organic matter and phytoliths. Such binding strength differences have also been found for associations between mineral matter and C-H enriched organic matter showing relatively weak associations and mineral matter and C=O enriched organic matter showing relatively strong associations (Leue et al., 2016). The association with organic matter has been proposed to stabilize phytoliths, whereby heating at specific temperatures might even stabilize this organic matter by aromatization enhancing phytolith resistance against dissolution (Nguyen et al., 2021).

In general, the chemistry of amorphous silica is mainly determined by its surface properties (Zhuravlev, 2000), but what actually causes fluorescence in phytoliths? Traditionally, fluorescence has been attributed to materials with (poly)-aromatic or (poly)-heterocyclic compounds representing conventional chromophores. These conventional chromophores have delocalizable electrons that are needed for photon absorption, energy-induced electron excitation, and electron relaxation leading to fluorescence as illustrated in the Jabłoński diagram (Jabłoński, 1933). However, this viewpoint has been challenged by numerous descriptions of fluorescing materials lacking conventional chromophores. This phenomenon can be well-explained by so-called clusterization-triggered emissions, which are caused by overlapping electron clouds forming non-conventional chromophores responsible for fluorescence (a comprehensive overview of the underlying mechanisms can be found in the review of Liao et al., 2021). Based on our results and the discussion above we propose that auto-fluorescence in phytoliths is caused by clusterization-triggered emissions, which can be induced by i) a strong increase in pH or ii) heating (Figure 7). While increasing pH leads to an increase in deprotonated silanol groups (≡Si−O⁻), increasing temperatures lead to increased dehydroxylation, i.e., the release of hydroxyl groups (−OH⁻) by forming water molecules, which results in an increased formation of siloxanes. Deprotonation results in overlapping lone pair electrons, which form electron clouds representing non-conventional chromophores that cause fluorescence events (pH effect). Dehydroxylation and the increasing formation of siloxanes result in the formation of oxygen clusters representing non-conventional chromophores (heating effect). At temperatures <400°C dehydroxylation is readily reversible by the introduction of an excess of water (Zhuravlev, 2000). The extent to which this is influencing phytolith auto-fluorescence in archaeological contexts has to be analyzed in future studies.

Our SEM-EDX results corroborate the hypothesis of Hodson (2019) that there are two distinct classes of phytoliths, i.e., cell wall and lumen phytoliths, with relatively high and low carbon concentrations, respectively. At this point it should be noted that the carbon contents measured by us are not directly comparable to the ones compiled by Hodson (2019). This is because we did not measure reference materials for SEM-EDX calibration, which would have been necessary for high-precision quantitative carbon analyses needed, e.g., for the determination of carbon sequestration in phytoliths. However, we ensured the compilation of reliable data for multivariate cluster analysis by SEM-EDX measurements under standardized conditions. This implies that the resulting proportions of different phytoliths (i.e., cell wall vs lumen phytoliths) will be directly comparable to corresponding results of future studies. Regarding the auto-fluorescence of phytoliths our results indicate that cell wall phytoliths with strong silica-organic-compounds-associations and corresponding high carbon concentrations might show strong auto-fluorescence without being burnt at all (auto-fluorescence of R groups, see Figure 7). Interestingly, cell wall phytoliths seem to be the ones that can persist in soils and sediments for centuries or even millennia making them ideal candidates for long-term carbon sequestration (cf. Hodson, 2019). In this context, the differences in carbon concentrations of cell wall and lumen phytoliths might explain the different viewpoints regarding the carbon sequestration potential of phytoliths (Song et al., 2016; Santos and Alexandre, 2017; Song et al., 2017). Moreover, this would explain the relatively strong FTIR peaks assigned to organic matter that was previously detected in phytolith samples by two of us (Puppe and Leue, 2018).

In future studies it has to be examined how much time is needed for the decomposition of organic matter associated with phytoliths in soils of different ecosystems. In fact, there are big differences in the Si uptake of different kinds of vegetation and the corresponding storage of phytogenic Si in soils (reviewed by, e.g., Katz et al., 2021 and Schaller et al., 2021). Our SEM-EDX analyses indicated that 62 and 38% of extractable phytoliths were classifiable as lumen and cell wall phytoliths, respectively. Cell wall phytoliths in turn were dividable equally in recognizable (in terms of phytolith morphology) cell wall phytoliths (short cells, 19%) and silicified tissue fragments (19%). Thus it seems that only a minority (19%) of extractable phytoliths is represented by recognizable cell wall phytoliths. This seems to be underlined by our fluorescence analyses of un-extracted soil phytoliths, which showed no auto-fluorescence, and thus might be classified as lumen phytoliths. However, detailed research on this aspect is necessary to corroborate our assumption. Nevertheless, altogether our findings might be a hint that more lumen phytoliths than cell wall phytoliths are formed in winter wheat in general. In fact, lumen phytoliths have been found to be much more common in grasses and cereals compared to other plant groups (Hodson and Guppy, 2022). In this context, it will be of great interest to examine if the proportion of cell wall and lumen phytoliths determined by extraction techniques is comparable to the cell-wall-to-lumen-phytolith-ratio determined by microscopical in situ observations. In fact, fragile phytogenic silica (e.g., thin epidermal silica) in plants are not covered by standard phytolith extraction procedures (Puppe et al., 2017). Considering that only about 16% of the total Si content of grasses is represented by extractable phytoliths (Puppe et al., 2017), the cell-wall-to-lumen-phytolith-ratios determined by extraction techniques and microscopical in situ observations might differ considerably.

What we need now is systematic research on the proportion of cell wall and lumen phytoliths in different plant species determined by a combination of phytolith extraction methods and microscopical analyses. This will allow us to assess the ‘flux’ or translocation of different phytoliths from these species into soils. The corresponding findings will both i) complement the existing information on morphology and taxonomy of phytoliths (Neumann et al., 2019) and ii) allow us to evaluate the potential of species-specific phytoliths for long-term carbon sequestration (e.g., Song et al., 2016). Additionally, the proportion of cell wall and lumen phytoliths in different plant organs like leaves, stems, and roots should be analyzed. This knowledge is crucial to evaluate the potential of plant material recycling for long-term carbon sequestration via phytoliths as well as for the Si supply of crops by the application of biochar (Li and Delvaux, 2019) or crop straw (Puppe et al., 2021). While the relatively stable cell wall phytoliths seem to be well-suited for carbon sequestration in the long term, the less stable lumen phytoliths seem to be important drivers of Si cycling in shorter terms. Our preliminary results on hand indicate that the ratio of cell wall to lumen phytoliths in winter wheat is close to 1:1, i.e., about 1:1.6, suggesting this plant species as both a well-suited candidate for long-term carbon sequestration and Si supply via phytoliths. Indeed, grass species are known for their relatively high Si contents and their potential for carbon sequestration as well as Si recycling has been demonstrated in previous studies (e.g., Puppe et al., 2021; Song et al., 2022).