The electron donating capacity of biochar is dramatically underestimated

Biochars have gathered considerable interest for agronomic and engineering applications. In addition to their high sorption ability, biochars have been shown to accept or donate considerable amounts of electrons to/from their environment via abiotic or microbial processes. Here, we measured the electron accepting (EAC) and electron donating (EDC) capacities of wood-based biochars pyrolyzed at three different highest treatment temperatures (HTTs: 400, 500, 600 °C) via hydrodynamic electrochemical techniques using a rotating disc electrode. EACs and EDCs varied with HTT in accordance with a previous report with a maximal EAC at 500 °C (0.4 mmol(e−).gchar−1) and a large decrease of EDC with HTT. However, while we monitored similar EAC values than in the preceding study, we show that the EDCs have been underestimated by at least 1 order of magnitude, up to 7 mmol(e−).gchar−1 for a HTT of 400 °C. We attribute this existing underestimation to unnoticed slow kinetics of electron transfer from biochars to the dissolved redox mediators used in the monitoring. The EDC of other soil organic constituents such as humic substances may also have been underestimated. These results imply that the redox properties of biochars may have a much bigger impact on soil biogeochemical processes than previously conjectured.

0.83 mmol(e − ).g char −1 . This is much higher than the EDC previously measured for wood (0.03 mmol.g −1 ) or grass-based biochars (0.10 mmol.g −1 ) with identical HTT 20 . Unfortunately, the ratio of Ag + either reduced or only adsorbed under cationic form could not be quantified. Another report showed that the bacterium Geobacter metallireducens was able to perform either denitrification (i.e. a reduction) or acetate oxidation with wood-based biochar (HTT: 550 °C) as the only electron donor or acceptor, respectively 22 . The bacteria were not expected to oxidize (or reduce) all the redox moieties "chemically available" within the char because of their restricted access in char pores as well as the thermodynamic limits of their metabolism. The authors were therefore surprised to find that their "bioavailable" EEC (0.86 ± 0.01 mmol(e − ).g char −1 ) was even higher than the chemically determined EECs measured by Klüpfel et al. for similar wood-based biochars produced at similar HTTs (0.59 ± 0.09 mmol.g −1 and 0.18 ± 0.2 mmol.g −1 for 500 °C and 600 °C HTT, respectively) 20 .
The redox properties of biochars could be substantially involved in numerous biogeochemical phenomena such as, e.g., N 2 O mitigation 21 . Regarding the limited amount of data related to EEC values of biochars as well as some apparent discrepancies, we aimed here to quantify the amount of electrons that can be accepted or donated by wood-based chars pyrolyzed at different HTTs. To do so, we took advantages of highly accurate hydrodynamic electrochemical techniques with rotating disc electrodes (RDE) in order to monitor the redox state of oxidative or reductive solutions reacting with the biochars 29 . After validation of the method, we compared our values with previously published data obtained with similar biochars (feedstock from identical genus -Pinus -and identical HTTs). We showed that the EDC of biochars has so far largely been underestimated, very likely because a very slow rate of electron donation from biochars to redox solutions is quickly reached and could have been unnoticed. We observed a similar phenomenon for grass-based biochars as well as synthetic lignin powder and synthetic humic acid. We review the putative reasons for this sluggish electron transfer. Finally, we discuss the need for further investigations of these redox mechanisms and their possible implications for biochar-mediated soil biogeochemical processes.

Results and Discussion
Electrochemical analysis for EDC determination. Three sets of char were prepared at different HTTs (400, 500 or 600 °C) and designated in this study as 'char-400' , 'char-500' and 'char-600' accordingly. The ground chars reacted for different times in abiotic, oxidative aqueous solution (buffered 50 mM ferricyanide, pH 6.5) under mixing before electrochemical analysis. All procedures are detailed in the Method section.
Hydrodynamic cyclic voltammograms (CVs) show that the 50 mM ferricyanide solution oxidized redox moieties from the char-400 in suspension for 20 days (Fig. 1). While the control devoid of char (red, dotted curve) showed an almost zero anodic plateau current (j la,ctrl , 20 μ A.cm −2 ), the presence of 35.0 mg char-400 increased j la to 1080 μ A.cm −2 (black curve). The amount of electrons donated by the char (ED m ) is proportional to (j la − j la,ctrl ) and was 4.8 mmol(e − ).g char −1 (see Methods and equations (9) and (10)). The presence of char did not modify the sigmoidal shape of the CV nor the difference between the anodic and cathodic plateaus currents. This proves that the total amount of dissolved [ferricyanide + ferrocyanide] stayed constant over the experiment 29 . In particular, it demonstrates the absence of notable adsorption of these redox species on the char surface. When the char suspension was devoid of ferricyanide (grey curve), the CV was flat with non-significant current over the full range of potentials (− 0.5 V to + 0.7 V vs. Ag/AgCl). Typical redox moieties from chars have formal potentials (pH 7) ranging from − 0.5 V to 0 V for electron accepting quinones 30 , and + 0.2 V to above + 0.6 V for electron donating phenolic compounds 31 . The absence of peaks or plateaus on the CV proves that no significant amounts of these redox moieties from char-400 were dissolved, and that they were tightly bound to the char. This is in good agreement with studies showing that soluble organic carbon content from biochar drastically decreases with increasing HTTs and becomes quasi-nil for HTT ≥ 400 °C 14,18,32,33 .
If the CV measurement allows the aforementioned controls, the ED m of chars was generally determined faster and even more accurately via chronoamperometry (C.A., see Supplementary Fig. S1). NaCl ("char + ferri", black curve); in controls in the absence of char ("ferri", red dotted curve) or in the absence of ferricyanoide ("char", grey curve). Recorded at a scan rate of 50 mV.s −1 and an electrode rotation speed of 1000 rpm. (B) Zoom on the anodic plateaus corresponding to the mass transfer limiting current density j la for ferrocyanide oxidation. The amount of electrons donated by the char is proportional to (j la − j la,ctrl ), difference between the anodic plateau currents recorded in the presence of char and in the control without char. Amount of electrons donated over time. The kinetics of electron donation for the three chars are presented in Fig. 2. Electron transfer rates from chars were initially high (average of ~0.1 mmol(e − ).g char −1 .h −1 for the three first hours of reaction) and decreased very sharply over time ( Supplementary Fig. S3). The ED m of chars showed a continuous and drastic decrease with the HTT from 400 °C to 600 °C, as previously observed in Klüpfel et al. 20 . However, no redox equilibria were reached after 66 days of reaction under our conditions, whereas Klüpfel et al. assumed that apparent equilibria were achieved for all chars after only ~1 h. In addition, the differences in EDCs between both studies are particularly large for similar pinewood-derived biochars. We observed between one and two orders of magnitude higher EDCs (t = 66 d) for char-400 (7.0 mmol(e − ).g char −1 vs. 0.20 mmol.g −1 in Klüpfel et al.), char-500 (3.7 mmol.g −1 vs. 0.03 mmol.g −1 ) and char-600 (1.4 mmol.g −1 vs. 0.03 mmol.g −1 ). Furthermore, the fact that the equilibria were not completely reached (only pseudo-plateaus reached on Fig. 2) indicates a certain underestimation of our own values of EDCs (which is also demonstrated by the substantial impact of initial ferricyanide concentration on the kinetics and pseudo-plateau values, see Supplementary  Fig. S4). The minor difference in the pinewood-char preparation (30 min charring at HTT vs. 60 min for Klüpfel et al.) would not explain such differences in EDCs considering that char particles sizes (≤ 50 μ m) and elemental compositions of the chars produced at identical HTT were similar in both studies (see Table 1 in Methods section for our values). Respective EDC values are compared on Fig. 3. Most redox moieties in the char are believed to contain oxygen 20 . The EDC increased almost linearly with the elemental oxygen content of the chars ( Supplementary Fig. S5). The average ratio of electron donated per O slightly decrease with HTT from 0.71 (char-400) to 0.43 (char-600), while it was calculated to be between 0.003 and 0.015 for the similar biochars of Klüpfel et al. 20 . A probable reason for these discrepancies and another comparative experiment will be presented further. The decrease of EDC with HTT is very likely associated with the decrease in the abundance of oxygen functional groups (e.g. hydroxyl or ether) with charring temperature (see Table 1) 14,20 . In particular, the very large decrease in EDC between char-400 and char-600 could be related to a transition point at HTT between 450 °C-550 °C with the onset of aromatization coupled with a drastic decrease in hydroxyl functions due to the dehydration of lignin-derived phenols and alcohols 34 . The increasing aromatization of the chars with increasing HTTs was also confirmed by the continuous rise of their aromatization index (AI) from 0.63 for char-400 to 0.83 for char-600 (see Table 1). Wood-based biochars from Klüpfel et al. presented similar increasing trend and values of AI from 0.52 (char-400) to 0.80 (char-600).   Controls. Figure 4A shows the CVs recorded in redox suspensions of different mass of char-600 (t = 30 d).
Conservation of the difference between anodic and cathodic current plateaus of CVs for each mass of char suspended (and control) confirmed the absence of noticeable adsorption of ferri-or ferrocyanide on the char surface (vide supra). Even for the maximum mass of char (96.4 mg), the final fraction of ferricyanide which has been reduced to ferrocyanide was only 3%, which indicates only a limited decrease in the oxidative power of the redox solution. Corresponding CAs recorded at + 0.7 V (Fig. 4B) confirmed and refined the values of the anodic plateau currents j la to access the amount of electrons donated by the char (ED). Figure 4C shows that the amount of electrons donated after 30 days was quasi-proportional to the mass of char-600 suspended. The final redox potentials (E h ) of the char-600 suspensions are plotted in Fig. 4D. E h value is determined by the ratio [ferri-]/[ferrocyanide] and follows the Nernst equation. The decrease of E h with the mass of char initially suspended reflects the increasing amount of ferricyanide reduced and therefore a decrease of the oxidative power of the solution over time (and for larger mass of char). The good linearity of ED with the mass of char strongly suggests that the redox solution was sufficiently oxidative (i.e. concentrated in ferricyanide) to oxidize most of the oxidable redox moieties of the char. Still, a slight decrease in the EDC with the mass of char introduced indicates that minor thermodynamic and/or kinetics limitations of the redox process may have led to a slight underestimation of our EDC values ( Supplementary Fig. 6). For all char suspensions from Fig. 2  The proportional relationship between the mass of char and ED (Fig. 4C) is expected if only the redox moieties of the chars are responsible for the reduction of ferricyanide. Still, all controls were performed to discard any (improbable) ferricyanide reduction by other chemical entities present. The chloride ion (used as supporting electrolyte) is a particularly weak reductant and did not reduce ferricyanide, even in the presence of char (absence of any putative catalytic effect of the latter), as shown on Supplementary Fig. S7. Water was not significantly reducing ferricyanide since its oxidation would necessarily produce some O 2 , which was not the case ( Supplementary  Fig. S8). Even the nature of the container surface itself (polypropylene) was tested and not involved in any electron transfer ( Supplementary Fig. S9).
Finally, we showed that even for the less stable of the three chars (lowest HTT char-400) 35 , the high EDC measured was not due to an hypothetical accelerated rate of biochar carbon mineralization to CO 2 because of ferricyanide ( Supplementary Fig. S8 and Method S1). All these controls confirms that the high EDCs measured were indeed related to the electron donating moieties from the chars, such as e.g. phenolic compounds 20 .
Electron accepting capacities. EAC initially increased with HTT and was maximal for char-500 (0.40 ± 0.02 mmol(e − ).g char −1 ) before sharply decreasing for char-600 (0.10 ± 0.01 mmol.g −1 ). The linearity of the amount of electrons accepted with the mass of char-500 and corresponding raw data (CAs) are presented Error bars represent 2 standard errors (n = 3 for all results except our EDC where n = 2). Note the 10-times difference in scale between the EDC and EAC charts.
in Supplementary Fig. S10. Contrary to EDC, EAC values as well as the trend of EAC variation with HTT are in excellent agreement with the data from Klüpfel et al. 20 , as illustrated on Fig. 3, and despite the use of different dissolved electron donors (NR in the present study while they used a zwitterionic viologen). This tends to prove that both methods were adequate for EAC monitoring. A very large difference with our slow EDC kinetics is that the redox equilibria for EAC measurements were reached in less than 4 days (no measurement was made at a shorter time). The putative reasons for this difference in kinetics will be discussed in the section "Possible reasons for sluggish electron donation". Contrary to EDC, the EAC did not increased monotonically with the elemental O content in chars but showed a maximum for char-500 ( Supplementary Fig. S5). The trend of EAC with HTT is probably due to the generation of electron-accepting quinone moieties up to 400-500 °C followed by their consumption for higher HTTs 20 . A fraction of EAC is also believed to be due to polycondensed aromatic structures newly formed for HTT ≥ 600 °C 20 .
Discrepancy between EDC values. We determined EDCs via a different method than the one conducted  20 . In that way the electrons supplied by a certain amount of char were transferred to the electrode and the EDC was obtained by integrating a baseline subtracted current of a CA over ~1 h (chronocoulometry). A good redox mediator must be stable in aqueous solution for the time scale of the experiment 37 and ABTS is a well-established colorimetric mediator for fast processes (s to min) such as enzyme assays 38 . However the radical anion ABTS •− can lack long term stability and is slowly reduced by water 39,40 . We attempted to perform EDC measurements with 1 mM ABTS •− rather than 50 mM ferricyanide and observed that 89% of ABTS •− had been reduced to ABTS 2− in the control devoid of char after 6 days ( Supplementary Fig. S11). In comparison, only 0.065% of the stable ferricyanide had been reduced for the same delay in the absence of char. For periods longer than 2 days, the remaining ABTS •− was reduced faster by water only (control) than in the char suspensions and the "apparent" ED m of chars were therefore decreasing over time. EDCs of chars were consequently underestimated with our method when using ABTS •− . Still, the "apparent" EDC values we obtained with ABTS •− (i.e. the maximum ED m reached for t ≥ 20 h) were between 3-times and 14-times higher than the values from Klüpfel et al. Interestingly, our ED m values after only 1 h of reaction with ABTS •− were very similar to their EDC values, recorded for ~ 1 h of reaction at which they assumed to have reached at redox equilibrium ( Supplementary  Fig. S12) 20 .
In a similar method to Klüpfel et al. to monitor the EDC of humic substances with ABTS •− , no measurements were performed at working electrode potential above + 0.52 V vs. Ag/AgCl because this led to a baseline current above 100 μ A, at least partially because of ABTS •− instability (i.e. homogeneous reduction by water followed by heterogeneous oxidation on the working electrode) 41 . Typical baseline current values were not provided in Klüpfel et al. 20 , but it can be speculated that the low CA potential (+ 0.405 V vs. Ag/AgCl, where only ~10% of ABTS is in its radical form ABTS •− and able to oxidize the chars) was chosen to avoid this excessive baseline current. With their direct measurement and after the initial fast rate of electron donation from the chars (> 1 h), it could be challenging to distinguish between a low, almost constant rate of electron donation and a potentially higher, constant baseline current of ABTS re-oxidation. Furthermore, integration of the unsubstantial difference would very likely lack accuracy. This is also strongly suggested on Supplementary Fig. S13 where we present our own continuous measurement of initial electron donation from char-500 to ferricyanide as if recorded in a system as used by Klüpfel et al. 20 . We therefore believe that the EDCs values provided in ref. [20] may have been underestimated because of an inadequacy of the method and mediator to notice and monitor the slow reaction of electron donation by the biochars.
The EDCs of compounds other than biochars have previously been assessed via identical or very similar ABTS •− mediated measurements over short times (~1 h) 41 . Humic substances show numerous chemical and structural similarities with biochar 42 . The EDC of a synthetic humic acid (Aldrich) was previously measured at 1.6 mmol.g −1 41 . We assessed a 3-time higher EDC (5.0 ± 0.2 mmol.g −1 , n = 3) for the same humic acid with our method after 100 days of reaction ( Supplementary Fig. S15). Similarly, our EDC for synthetic lignin powder was much higher (15.2 mmol.g −1 after 15 days, Supplementary Fig. S16) than the one previously monitored in ~1 h with ABTS •− (4.3 mmol.g −1 ) 20 . Batch-to-batch variation from the same supplier could explain a relative difference in EDC values of these commercial compounds. However, the 3-times differences in EDCs monitored with both techniques as well as the much longer redox reaction strongly suggest that a slow release of electrons from these compounds has also been unnoticed after 1 h of reaction with the continuous, ABTS •− -based method.
One may claim that comparison of results from different methods could be biased, however, we strongly believe that EDC and EAC values should be intrinsic properties of specific compounds and therefore not be substantially method-dependent, similarly than for e.g. specific surface area, cation exchange capacity, elemental composition, etc. A standardized method of determination should be developed to record accurate values and allow relevant comparisons between compounds. Before all, a clear definition of "EDC" itself should be discussed. Should the choice of a certain EDC value be determined by a certain time of reaction in a specific solution (concentration of the redox compound and redox potential)? Should the EDC include a putative fraction of electrons donated by mineralization or via irreversible oxidative polymerization (vide infra)? Should it correspond to the amount of electrons that the char can only donate in environmentally relevant conditions (and then which one)?
Possible reasons for sluggish electron donation. Wood-based biochars are relatively hydrophobic and a fraction of the char particles can occupy the gas/liquid interface of a static solution. The addition of 1 mM cetyltrimethylammonium bromide (CTAB, surfactant enhancing porous carbon wettability) 43 did not impact the electron donation kinetics of the most hydrophobic biochar (char-600, Supplementary Fig. S17). This proves that under our mixing conditions, a hypothetical wettability issue was not involved in the slow kinetics. Since CTAB is an effective antiseptic 44 , it also confirms that no (unexpected) microbial process was involved in the redox process.
The surface of biochars is negatively charged at circumneutral pH 18 . Pore sizes of chars range over several orders of magnitude but smaller ones are typically at the nanometer level 1,13 . This is also the typical scale of the electrical double layer thickness where mostly cations can reach the close vicinity of the negative surface 45 . An electrostatic repulsion could therefore restrain the penetration of dissolved anions in small pores, and in particular for the ferricyanide trianion. Porosity and pore accessibility quickly increase with HTT 18,19 , while the surface charge density decreases with HTT 1 . Despite having the lowest EDC, the char-600 showed a faster initial electron donation than the chars prepared at lower HTT (see Supplementary Fig. S18). The ferricyanide molecules very likely encounter an easier/faster access to the oxidizable moieties present in the pores of the char-600 than in the less accessible, more negatively-charged pores of lower HTT chars. We also showed that the redox equilibrium was reached much faster with the uncharged 46 NR (for EAC) than with the ferricyanide trianion (for EDCs). All these results would initially favor the assumption of electrostatic repulsion of ferrocyanide limiting the kinetics within the pores of the chars. However, Klüpfel et al. recorded similar EAC values with a negatively charged zwitterionic viologen in only 1 h 20 , which rather thwart the hypothesis of an electrostatic impact. Finally, the relatively slow kinetics observed for the non or less porous lignin and humic acid powders (Supplementary Figs S15 and S16) also favor another hypothesis. A possible electrostatic impact could still be assessed by testing neutral or positively charged oxidants to monitor EDCs.
A slow mineralization (i.e. oxidation) of biochar carbon to CO 2 could account for the slow kinetics of electron donation observed, but only partially (see Supplementary Fig. S8 and associated Method S1: a maximum of 7% of the electrons donated by char-400 after 10 days could be the result of mineralization).
Last but not least, the sluggish reaction could be due to intrinsically slow kinetics of oxidation for some recalcitrant compounds or to sequential irreversible oxidations of some phenolic species. The reversible 1-electron oxidation of phenolic moieties produces phenoxy radicals. The latter can further perform slower complex, irreversible oxidative polymerizations [47][48][49] . This has been observed abiotically either with ferricyanide 50 or ABTS •− 51 , or via enzymatic processes 52 . The first, reversible oxidation would allow a redox cycling of the char, conserving its "redox buffer" properties for its environment, as shown by Klüpfel et al. 20 . On the contrary, the irreversible oxidative polymerization would not allow further redox cycling and would decrease the EEC. Further investigations are needed to understand the slow kinetics of electron donation from the chars as well as their redox cyclability. It is also essential to assess to which extent these electrons can be donated in environmentally relevant conditions.

Conclusion
Understanding and evaluating redox reactions in soil are of crucial importance. This work provides strong evidences that biochars may donate much more electrons to their environment than previously considered. It also shows that wood-based biochar can typically give one order of magnitude more electrons than they can accept, potentially favoring reductive reactions in the subsurface. Our results also suggest that the EDC of other organic or mineral compounds from soil may have been underestimated if their kinetics of electron donation is slow. In this case, punctual measurements with stable redox probes may be preferred to continuous ones if their detection limit in term of reaction rate is too high.
Given their much larger mineral content, the redox properties of biochars from other feedstock sources (e.g. manures or sewage sludge) should be carefully assessed. For example, a high content in Fe and/or Mn in the char may significantly impact the EECs and the kinetics at which electrons are exchanged with their environment. Due to the high chemical complexity of chars or humic substances, it may also be useful to assess the EDC of redox model molecules bound within the biochar structure after primary degradation of lignin during pyrolysis (e.g. phenolic and aromatic compounds). Valuable information could derive from their (chemical and biological) oxidation mechanisms, kinetics and products determination, as well as the possible occurrence and extent of these redox processes in environmental conditions. A key question is indeed how much electrons a biochar can actually donate to a specific biogeochemical environment with respect to the chemical EDC measured in laboratory studies where the oxidation rate and extent are attempted to be maximized. Furthermore, ageing of biochars very likely impacts their redox properties before and after their addition to soil. In particular, ageing of chars can modify their surface via precipitation or sorption of organic and mineral compounds to their surface 53 .
The EDC and EAC should also be accurately measured and taken into account when studying any redox process occurring in the presence of biochars. For example, several recent reports showed that biochars 27,54 can promote microbial activity by acting as electron shuttle between two delocalized microbial species 54 or between a bacterial strain and Fe(III) mineral 27 . On the other hand, Saquing et al. recently proved that biochars can substantially support microbial activity only via their ability to provide or harvest electrons within the microbial metabolic pathways (i.e. irrespectively to biochars conductivity) 22 . This stresses the need to assess which of these nonexclusive mechanisms (conductivity or EECs) is mostly involved in supporting microbial metabolism, especially for biochars pyrolyzed at relatively low HTT (< 600 °C) where EECs are high but electrical conductivity unsubstantial 55 .
More generally, our results further emphasize the potentially large impact of biochars on soil redox processes, either microbial or abiotic. The nature and extent of these redox processes will depend on the char characteristics (EEC, porosity, surface chemistry, charge and area, etc., which mostly rely on feedstock choice and charring conditions) and its surrounding soil and climate environment (soil composition and morphology, pH, plant and microbial ecology, temperature, hygrometry, etc.). A better understanding of the numerous and complex redox interactions between well-defined biochars and specific biogeochemical environments is therefore crucial in order to design adequate biochars for targeted beneficial purposes.

Methods
Chemicals. Na 2 HPO 4 .2 H 2 O, NaH 2 PO 4 .2 H 2 O, NaCl and potassium ferricyanide were purchased from Carl Roth; potassium ferrocyanide, neutral red (NR) from Sigma. All chemicals had higher purity than 98% and were used without further purification. Humic acid (20 wt% ash) and lignin (alkali) were from Aldrich.
Solutions. All solutions were made with deionized water (18.2 MΩ.cm), passed through a Milli-Q purification system (Millipore) and made anaerobic by N 2 gas bubbling (~30 min) with subsequent storage (> 2 h) before use in an anaerobic workstation (GP-Campus, Jacomex, TCPS NV, Rotselaar, Belgium) under a N 2 :CO 2 (90:10, v/v) atmosphere. Solution for EDC determination was 50 mM potassium ferricyanide (E 0′ = + 0.29 V vs. Ag/AgCl), 3 M NaCl and 0.1 M sodium phosphate buffer (PB) at pH 6.5. At the end of the experiments, the char suspensions were all at pH 6.5 ± 0.1. Highly concentrated NaCl acts as the supporting electrolyte and avoids migration impact on ferricyanide mass transfer, which could bias electroanalytical measurements based on diffusion/convection (Levich equation). EAC was determined in a solution made of ~5 mM of reduced NR RED (E 0′ = − 0.49 V vs. Ag/AgCl) in an identical electrolyte. The electrolysis step to reduce commercial NR before its use is described in Supplementary Method S3.
Chars. Commercially available bark-free pinewood (Bemap Houtmeel BV, The Netherlands) was used as feedstock containing a mixture of different pine species, of which Pinus sylvestris was the predominant one. Further compositional data are available elsewhere 56 . The production of char was carried out under N 2 in a slow pyrolysis retort, which consisted of a vertical, tubular, stainless steel reactor (d × L = 3.8 cm × 30 cm) heated by an electric furnace, full details on the setup and procedure are available elsewhere 57 . From 70 g to 100 g of ~6 mm wood chips were first loosely packed in the reactor. The reactor was heated at a rate of ~17 °C min −1 until the HTT was reached and held constant for 30 min, after which the reactor was cooled at ambient temperature. Pyrolysis gases were continuously evacuated by continuous N 2 flow. Elemental analysis (C, H, N) was performed on biochars in triplicate in an elemental analyser (FLASH 2000 Organic Elemental Analyser, Thermo Scientific) according to standard procedure 58 . The oxygen content was calculated by mass balance on the ash-free basis. The samples (3 mg) were dried overnight (105 °C) before analysis and 2,5-bis(5-tert-butyl-benzoxazol-2-yl)thiophene (BBOT) was used as a standard reference. The ash content in the chars was determined in triplicate according to standard procedure 59 in a muffle furnace (AAF 11/3, Carbolite). BET specific surface area of biochars was determined via Scientific RepoRts | 6:32870 | DOI: 10.1038/srep32870 N 2 adsorption at 77.4 K with a TriStar II surface area analyzer (Micromeritics). The samples were pretreated with a VacPrep Degasser (Micromeritics). Selected physicochemical properties of the chars are listed in Table 1.
Redox reactions in char suspensions. All dry chars were manually ground with a ceramic mortar and pestle to increase accessibility of the redox solution to the pores and accelerate the electron exchange rate. Particles of ground chars smaller than 50 μ m were selected via sieving (Retsch, Germany). Unless otherwise specified, ~30 mg of chars were weighed (0.1 mg accuracy) in 50 mL polypropylene tubes wrapped under aluminum foil to avoid light exposure (the latter increases the very slow natural reduction of ferricyanide, see Supplementary  Fig. S20). The chars were exposed to vacuum for at least 1 h in the airlock of the anaerobic workstation to remove residual O 2 . Note that a longer vacuum exposure (or its absence) did not substantially modified the EDC obtained ( Supplementary Fig. S21). All the following steps were performed inside the anaerobic workstation. The tubes were filled with 55 mL of anaerobic oxidative or reductive solutions (~0.55 g char .L −1 ), without substantial headspace to favor wetting of relatively hydrophobic chars. The tubes were sealed and immediately subject to continuous shaking at 100 rpm (orbital shaker, Phoenix Instrument), initiating the redox reactions between the redox moieties of the chars and dissolved oxidant or reducer (t = 0): Electrochemical measurements for EDC determination. The measurements were performed in the char suspensions (or controls in the absence of char) using a potentiostat (VSP, Biologic, France) inside the anaerobic workstation at 30 °C. Glassy carbon RDEs (3 mm diameter (0.0707 cm 2 ), A-011169, ALS, Japan) were used as working electrodes. They were successively polished on microcloth pads with 1 μ m and 0.05 μ m diameter alumina slurries (Buehler, USA), and a particle free pad, and thoroughly rinsed with distilled water after each step. No decrease of electroactivity toward ferri/ferrocyanide couple was detected over 20 successive measurements, whereas the RDE was re-polished before every measurement in NR solutions due to slow but significant decrease of electroactivity. A platinum spiral wire (10 cm) was used as counter electrode and all potentials in this manuscript are referred to a Ag/AgCl (3 M KCl) reference electrode (ALS, Japan, + 0.205 V vs. SHE at 30 °C). The RDEs were rotated using an RRDE-3A rotator (ALS, Japan) in a 50 mL glass beaker. Cyclic voltammetries (CVs) were performed at 50 mV.s −1 at a RDE rotation speed of 1000 rpm. Constant potential chronoamperometries (CAs) were periodically performed at 1000 rpm in the supernatants of the suspensions which were reintroduced in their respective tube after measurement. The potential was sufficiently high (+ 0.7 V vs. Ag/AgCl) to reach the anodic limiting current density j la [A.cm −2 ] which is proportional to the total ferrocyanide concentration [ferro] tot according to the Levich equation 45 : Finally, the EDC of biochars corresponds to the final ED m once a (quasi-)equilibrium between the char and the sufficiently oxidative solution is reached.
Electrochemical measurements for EAC determination. The method was similar than for EDC determination. EA m is proportional to the absolute value of the background-subtracted cathodic limiting current density (j lc − j lc,ctrl ) for NR reduction recorded at − 0.65 V vs. Ag/AgCl (CVs available in Supplementary Fig. S19): With D NR the diffusion coefficient of NR at 30 °C in 3 M KCl. The latter was not retrieved from a database and was assumed to be 3.57 × 10 −6 cm 2 .s −1 (recalculated with the Stokes-Einstein equation from the diffusion coefficient at 25 °C of fluorescein in water (4.25 × 10 −6 cm 2 .s −1 ) 62 , which has a similar structure and molecular weight than neutral red). The EAC corresponds to the EA m once an equilibrium between the char and the sufficiently reductive solution is reached. Experimental values in text are provided as the mean + /− standard deviation for n samples.