Sorption to soil, biochar and compost: is prediction to multicomponent mixtures possible based on single sorbent measurements?

Sorbents

Biochars produced from soft wood and miscanthus grass straw at pyrolysis temperatures of 550 and 700 °C were purchased from the UK Biochar Research Centre (SWP550, SWP700, MSP550 and MSP700, respectively). A compost containing 11.4% organic carbon was kindly provided by fk Agrar Umweltservice (Pixendorf, Austria). Two top soils were sampled from agricultural fields in Austria: a sandy loam from Eschenau (Lower Austria) and a clay loam from Kaindorf (Styria). After sampling, the two soils were air dried, sieved at <2 mm and stored in the dark. Soil samples were amended with compost and/or biochar (all air dried, see the details in the sorption section), well mixed and conditioned for about three days at 20 °C. The mixtures were then dried overnight at 80 °C, crushed with a pestle and mortar, and sieved at <250 µm to ensure homogenisation before the sorption experiments. All materials were thoroughly characterized in previous studies, including elemental analysis, organic carbon, pH, specific surface area and pore volume (Sigmund et al., 2017b; Bielská et al., 2017). All properties are presented in Tables S1–S3 of the supporting information. Scanning electron microscope images were obtained for the four biochars with an Inspect™ S50 scanning electron microscope operating under high vacuum, 10 kV acceleration voltage equipped with a secondary electron Everhart–Thornley detector (Figs. S1 and S2) following sputtering of the samples with carbon.

Chemicals and analysis

Hexane and methanol were of residue analysis grade (Labscan, Dublin, Ireland and Acros Organics, Geel, Belgium). Pyrene (99.5%) and pyrene-d10 (99.5%) were purchased from Dr. Ehrenstorfer GmbH (Augsburg, Germany). Pyrene was analyzed by gas chromatography-mass spectrometry (GC-MS): Agilent 7890A coupled to Agilent 5975C; HP-5MS fused silica column: 60 m × 250 μm × 0.25 μm, J&W Scientific; pulsed splitless mode; oven temperature of 55 °C for 1 min, then 10 °C/min up to 300 °C. Quantification was based on deuterated internal standards added to the samples before extraction. The limit of detection was 0.0054 μg/L for the batch sorption experiment and 0.0002 μg/L for the passive sampling method (the concentration refers to that remaining in the aqueous phase of the sorbent suspension after sorption).

Sorption measurements

Sorption of pyrene was first measured on the four biochars (SWP550, SWP700, MSP550 and MSP700) and the compost. The biochar exhibiting the highest sorption potential was further used in the sorbent mixtures: compost + biochar (2:1) and for each soil: soil + 10% compost, soil + 5% biochar, and soil + 10% compost + 5% biochar (all in dry weight based). The rates were selected to be consistent with previous studies considering combined amendments (Bielská et al., 2017; Sigmund et al., 2018). Sorption was measured using two techniques: a classical batch and a passive sampling method using POM. All sorption experiments were conducted at 20 °C and using aqueous background solutions containing 0.01 M CaCl₂ and 0.385 mM NaN₃ to inhibit biological activity. Pyrene stock solutions were prepared in methanol, and the amount spiked was kept below 0.1% in volume to minimize cosolvent effects on sorption. Controls and blanks were included in all experiments and the sorbed concentrations were calculated by mass balance.

For the batch method, sorbent suspensions were prepared in 50 mL glass centrifuge tubes (10 mg of biochar, 20 mg of compost, 200 mg of soil or soil mixture in 40 mL of background solution) and equilibrated for 48 h by horizontal shaking at 125 rpm. Pyrene was then spiked (the initial concentration ranged 1–50 µg/L), before samples were return to shaking for 48 h, which was previously demonstrated to be sufficient to reach sorption pseudo-equilibrium (data not shown). Samples were then centrifuged at 1000 g for 40 min, 30 mL of supernatant were collected, and extracted three times with hexane after addition of the internal standard (Pyr-d₁₀). Extracts were combined, concentrated down to 1 mL under N₂ flow and analyzed by GC-MS (see the details above). Sorption was measured by batch at six different concentration levels for each sorbent and in triplicates.

The POM method allows sampling the truly dissolved portion of pyrene and studying the possible effects that dissolved organic matter may have on sorption measurements. The POM method also allows measuring sorption down to lower concentrations than with the classical batch. The method was previously described and validated for pyrene (Kah et al., 2011, 2016) and it is only briefly described here. Clean POM strips of about 100 mg were preconditioned in the background solution for 72 h, before being added to sorbent suspensions prepared as described for the batch method. The samples were then spiked with pyrene (the initial concentration ranged 1–75 µg/L), and shaken horizontally in the dark at 125 rpm for 31 days in order to achieve equilibrium (Kah et al., 2011). The POM strips were then removed from the vials and wiped with a wet tissue. The internal standard (pyrene-d10) was added and the POM strips were extracted with 20 mL of hexane by shaking for three days at 125 rpm (Cornelissen & Gustafsson, 2004). Extracts were concentrated under N₂ flow and analyzed by GC-MS. Sorption was measured by the POM method at 5–6 concentration levels and in duplicates.

Data analysis

Sorption isotherms were fitted with the Freundlich model: (1) $C_{\text{s}}=K_{\text{f}}·C_{\text{w}}{}^n$ where C_s (μg/kg) is the sorbed concentration, C_w (μg/L) is the concentration remaining in solution, K_f ((μg/kg)/(μg/L)ⁿ) is the Freundlich affinity and n (-) is the linearity parameter. The model fitted all isotherms adequately, and there was no need to consider more complex models (potentially bringing a risk of over parametrisation, Kah et al., 2011). Sorption coefficients (K_d, L/kg) were calculated for C_w 0.02 and 2 μg/L with the aim to compare the different sorbents at the low and high concentration levels included in the isotherms, respectively. Normalized coefficients, K_OC values were also calculated based on the organic carbon content of each sorbent and sorbent mixtures (Table S3).

The dual-mode model was also fitted to the sorption isotherms in order to quantify the contributions of adsorption Q_ad (µg/kg) and partitioning Q_p (µg/kg): (2) $C_{\text{s}}=Q_{\text{ad}}+Q_{\text{p}}=\left(Q_{\text{max}}*C_{\text{w}}\right)/\left(b+C_{\text{w}}\right)+K_{\text{p}}*C_{\text{w}}$ where Q_max (µg/kg) is the adsorption maximum capacity, b is the affinity coefficient (µg/L) and K_p is the partitioning coefficient (L/kg).

All statistical analyzes, curve fits and graphs were produced using GraphPad Prism 6 (2016; GraphPad Software Inc., San Diego, CA, USA), with a significance level set to α = 0.05.

Sorption of pyrene to four biochars

Sorption isotherms of pyrene to the four biochars could be well fitted by the Freundlich model (0.92 < r² < 0.97, Fig. 1, all parameters are presented in Table S4). The linearity parameter n ranged between 0.45 and 0.67, which is characteristic of the highly nonlinear sorption of pyrene to carbonaceous sorbents (Kah et al., 2016). Nonlinearity increased with the pyrolysis temperature and was greater for the biochars derived from soft wood than from miscanthus (p < 0.001).

The biochars produced from miscanthus sorbed significantly more pyrene than biochars produced from soft wood, which was unexpected in view of the greater OC%, surface area and pore volume of the latter (Table S2). Sorption of organic sorbates generally increases with the biochar production temperature (Kah et al., 2016), and this was observed for soft wood, but not for miscanthus in our study. The discrepancy cannot be interpreted in terms of surface chemistry (e.g., aromaticity and polarity indices shown in Table S2) and neither in terms of physical parameters (surface area or porosity). Behazin et al. (2015) have previously shown that wood chip biochars exhibit atypical characteristics relative to miscanthus biochars, possibly related to differences in their structure, as shown on the electron microscope images (Figs. S1 and S2). Identifying the exact cause(s) for discrepancies in behavior between the two feedstock materials will require further investigations. Biochar MS550, which exhibited the highest sorption for pyrene, was considered for further investigations and it is designated as “biochar” in the following discussion.

Comparison of the batch and POM methods

We had hypothesized that the compost would generate dissolved organic matter that can sorb significant amounts of pyrene. A previous study carried out with the same materials showed that concentrations in dissolved organic carbon increased by 40% for the clay loam and 100% for the sandy loam upon compost addition, reaching concentrations of about 850 mg/kg (details are available in Bielská et al., 2017). Dissolved organic matter may not be fully separated from the aqueous phase during the centrifugation step applied in the batch method, which can lead to K_d batch < K_d POM (the POM method only considers the freely dissolved portion of pyrene). Fig. S3 shows that the data generated by the batch and POM methods were consistent, and the isotherms generated in the low (with POM) and higher concentration range (by batch) constituted a continuous isotherm. When considering all sorbents, there was no significant difference in the K_d values derived by the two methods (paired t-tests on K_d values calculated based on isotherms fits at 0.05, 0.08, 0.2 and 0.5 µg/L, representing the range of concentration where both methods could be applied). When differences were noticed, K_d values generated by batch tended to be greater than K_d generated by POM, which goes against the hypothesis that dissolved organic matter played a significant role on the sorption of pyrene. It was previously demonstrated that batch and POM methods can generate complementary results (Kah et al., 2011). In view of the general continuity of the isotherm generated by batch and POM (Fig. S3), the data were combined. The following discussion is thus based on isotherm analysis across several orders of magnitude.

Sorption isotherms

All sorption isotherms were very well fitted with the Freundlich model (r² ≥ 0.95, Figs. 2 and 3 all presented in the Table S5, and isotherms are presented in logarithmic scale in Figs. S4 and S5). The sorption affinity of pyrene to the clay loam was greater than to the sandy loam, which is consistent with the higher organic carbon content of the former.

The sorption isotherm to the pure biochar was the most nonlinear (n = 0.67), while sorption to the compost was one of the most linear (n = 0.89). The addition of compost to the biochar increased the n value from 0.67 to 0.73 and brought the isotherm close to that of the compost (Fig. 2), indicating that significant fouling of the biochar sorption sites by the compost must have occurred (further discussed below). Similar observations were made with the soil mixture presented in Fig. 2: the linearity of the isotherms decreased upon the addition of biochar, and increased upon addition of compost (either with or without biochar). Changes in the isotherm linearity are consistent with previous literature, and reflect the contribution of partitioning to labile organic matter from soil and compost (close to linear), relative to the contribution of adsorption (nonlinear) to the carbonized surface of biochar (Schwarzenbach, Gschwend & Imboden, 2017; Sigmund et al., 2017b).

All sorption isotherms were also fitted with the dual mode model with the goal of distinguishing and quantifying the respective contributions of absorption (mainly driven by the soil and compost) and adsorption (mainly driven by the biochar). In many cases, the fits were ambiguous, meaning that several set of values for the three parameters could equally well fit the curve. This is likely because the dual mode model is too complex (over-parameterized) for the isotherms, leading to highly unreliable values that should not be interpreted. The following discussion is thus solely based on isotherm fits with the Freundlich model. Differences in sorption affinity cannot be discussed based on K_f values due to great differences in isotherm nonlinearity. Using the isotherm parameters presented in Table S5, sorption coefficients were calculated at two different concentrations and are discussed in the next section.

Impact of amendment mixtures on sorption: K_d values

Sorption coefficients calculated at C_w = 0.02 and 2 mg/L (K_d0.02 and K_d2) are presented in Fig. 4 (grey crosses, all values are available in Table S5) to allow comparisons of sorption among sorbents and their mixtures. Predictions of K_d values for the mixtures were also computed based on the K_d value of pyrene to each individual sorbent, and their respective proportion in a given mixture. The theoretical contributions of the soil, compost and biochar are represented by the brown, green and black bars in Fig. 4 respectively. The top of the bar indicates the K_d value predicted for the mixture by the additive model. For single sorbents, the prediction matches the measurement (100% contribution of one sorbent). For the mixtures, differences between the predicted (bars) and measured K_d values (crosses) indicate the extent to which sorption was reduced through the interactions between sorbents.