Particles under stress: Ultrasonication causes size and recovery rate artifacts with soil derived POM, but not with microplastics

Line 11: delete “some” [1] Done. Introduction Line 36: ultrasound is applied to a soil slurry by using a sonotrode and Lines 36-37: “light” and “heavy” needs to be explained here [2] We adjusted the Lines 36-37 „In studies on soil carbon pools, ultrasound is applied to a soil slurry to break down soil aggregates.“ [3] and added the explanation of LF and HF (Line 38): „This disaggregation allows density fractionation of the free and occluded light fractions (fLF and oLF), which largely consist of material with densities below the fractionation medium, from the heavy fraction (HF), that has higher densities.“ [4] Furthermore, „... and subsequent density fractionation of particulate organic matter ...“ is added to Line 29 to introduce the fact that density fractionation is an integral part of the method.


Introduction
The mechanical disintegration of soil aggregates by use of ultrasonication following the method of Edwards and Bremner (1967a) is widely used in the assessment of soil organic matter (SOM) stability. This includes characteristics such as aggregate composition and stability (Edwards and Bremner, 1967b), the constitution of SOM pools (Golchin et al., 1994), the stabilization of SOM in forest ecosystems (Graf-Rosenfellner et al., 2016) and the occlusive strength of particulate organic matter (POM) (Büks and Kaupenjohann, 2016). Ultrasonication is also applied to assess quantities and qualities of anthropogenic contaminants such as microplastics (Zhang and Liu, 2018;Zhang et al., 2018).
In studies on soil carbon pools, sonotrodes are applied to break down soil aggregates and separate the free and occluded light fractions (fLF and oLF) from the heavy fraction (HF). These operational fractions correspond largely to the free particulate organic matter (fPOM), the occluded particulate organic matter (oPOM) and the mineral-associated organic matter (MOM), which are assigned to the labile, intermediate and stable carbon pool, respectively, and have turnover times of <1 year (labile) to several thousands of years (stable) (Lützow et al., 2007). Furthermore, the extracted POM fractions may not only contain the natural but also anthropogenic components such as microplastic. Its quantification and characterization is a very topical task in the growing research on microplastic contamination of soils and requires a high extractive performance. In addition, when optical methods are used to determine the size and shape of the microplastic, the extraction should also cause the least possible damage to the extracted material, because both attributes provide information about the source (Zhang and Liu, 2018;Ding et al., 2020), the mobility within the soil pore space (O'Connor et al., 2019) and the ingestibility of microplastic by soil organisms (Büks et al., in review).
The common method of ultrasonication is carried out with a pieco-electric converter, that uses electric energy to generate axial vibration of a sonotrode, which is dipped into a flask containing a fluid and a submerged soil sample. The oscillating sonotrode emits acoustic pulses within the fluid. In front of the shock-waves the medium is compressed, and the increased pressure causes an increased gas solubility. Behind the wave the medium relaxes and the pressure drops below the normal level leading to an explosive outgassing (Ince et al., 2001). This so called cavitation effect produces lots of exploding micro-bubbles between particles and within cavities of the soil matrix generating very local pressure peaks of 200 to 500 atm accompanied by temperatures of 4200 to 5000 K (Ince et al., 2001). It provokes the detachment of physiochemical bondings between soil primary particles and soil aggregates and, thus, causes disaggregation. Depending on device type and settings, the vibration frequency can vary up to 10000 kHz, but low frequencies around 20 to 100 kHz are recommended for soil aggregate dispersion to avoid chemical alteration of OM, and the use of 40 kHz is very common. (Kaiser and Berhe, 2014;Graf-Rosenfellner et al., 2018) 2 27   28  29  30  31  32  33  34  35   36  37  38  39  40  41  42  43  44  45  46  47  48  49   50  51  52  53  54  55  56  57  58  59  60  61  62  63 As an artifact of the method, ultrasonication is known to provide mechanical and thermal stress strong enough to comminute mineral particles at energy levels >700 J ml -1 (Kaiser and Berhe, 2014). Also, the destructive influence on POM was tested in different studies and appears even at energy levels much lower than 700 J ml -1 . Without application of a solid mineral matrix, Balesdent et al. (1991) found >60 % of the POM in suspension comminuted after application of 300 J ml -1 . Amelung and Zech (1999) treated natural soils with 0 to 1500 J ml -1 and performed a separation into size fractions of <20 µm, 20 to 250 µm and >250 µm. At ≥100 J ml -1 POM was transferred from the >250 µm to the <20 µm fraction. In a similar manner, Yang et al. (2009) measured the mass and SOC content of sand, silt and clay sized particle fractions in natural soils using an unconventional pulse/non-pulse ultrasonication technique. The authors derived the comminution of POM at >600 J ml -1 . Oorts et al. (2005) added 13 C-enriched straw to natural soils and could show that larger amounts of POM were redistributed at 450 J ml -1 when its degree of decomposition was higher. In conclusion, those studies consistently found a comminution of POM by ultrasonic treatment, which appears, however, at very different energy levels and is likely affected by the aggregation regime (suspended without mineral matrix, added as fPOM, occluded within natural soils), direct or indirect quantification of POM and the type of POM.
The aim of this work was to test under standardized conditions how susceptible different POMs are to comminution by ultrasonic treatment. We embedded three POMs (farm oPOM, forest oPOM and pyrochar, applied as an analog for soil black carbon and biochar amendments) and also six differently weathered microplastics (fresh and weathered lowdensity polyethylene (LD-PE), polyethylene terephthalate (PET) as well as polybutylene adipate terephthalate (PBAT), a common biodegradable material) into a fine sand matrix. Then, we treated these mixtures with 0, 10, 50, 100 and 500 J ml -1 , re-extracted the organic particles with density fractionation and measured their recovery rates and particle size distributions. The sand matrix was used only to simulate the influence of pore space on cavitation and, thus, our simplified approach excluded broadly varying POM-mineral interactions resulting from aggregation processes in natural soil samples. We hypothesized the strongest comminution in case of the two oPOMs, that already started to decomposed within their former natural soil matrix, and we were curious about the effect of ultrasonication and artificial weathering on the structural stability of microplastic, which has not been studied before.

Preparation of POM
The farm and forest oPOMs were extracted from air-dried soil aggregates of 630 to 2000 µm in diameter sampled in 10 to 20 cm depth from an organic horticulture near Oranienburg/Brandenburg (N 52° 46' 54, E 13° 11' 50, texture Ss, Corg=4.9 g kg -1 , pH 5.8) and a spruce/beech mixed forest near Bad Waldsee/Banden-Württemberg (N 47° 50' 59, E 9° 41' 30, texture Sl4, Corg=7.3 g kg -1 , pH 3.4), respectively. The extraction was performed by use of a density fractionation in 1.6 g cm -3 dense sodium polytungstate (SPT) solution: In 12-fold replication, 120 ml of SPT solution were added to 30 g of aggregates in a 200 ml PE bottle. The sample was stored for 1 h to allow the SPT solution to infiltrate the aggregates and was then centrifuged at 3500 G for 26 min. The floating free particulate organic matter (fPOM) was removed by use of a water jet pump and discarded. The remaining sample was refilled to 120 ml with SPT solution and sonicated for 30 sec (≈10 J ml -1 ) by use of a sonotrode (Branson© Sonifier 250) in order to flaw the structure of macroaggregate (>250 µm). Then, centrifugation and removal of the oPOM were executed as for the fPOM. The gained oPOM was filtered off with an 0.45 µm cellulose acetate membrane filter, washed 3 to 5 times with 200 ml deionized water within the filter device until the rinse had an electrical conductivity of <50 µS cm -1 , removed from the filter by rinsing with deionized water, collected and gently dried for 48 h at 40°C. At the end, the oPOMs were sieved to 2000 µm, long-shaped residues were cut by a sharp knife, sieved again and pooled to one oPOM sample. The pyrogenic char sample (made from pine wood, pyrolysed at 850°C for 0.5 h by PYREG ® GmbH) was dried for 24 h at 105°C, ground in a mortar and sieved to <630 µm. The microplastics (LD-PE, PET and PBAT) were made from plastic films by repeated milling (Fritsch Pulverisette 14) with liquid nitrogen and sieved to <500 µm. Then, half of each sample was weathered for 96 h at 38°C, 1000 W m -2 (solar spectrum, 280 to 3000 nm) and a relative air humidity of 50 % following DIN EN ISO 4892-2/3.

Mechanical stress treatment
In order to test their stability against ultrasonication, the nine POM types (farm and forest oPOM and pyrochar as well as fresh and weathered LD-PE, PET and PBAT) were each exposed in triplicates to different mechanical stress levels (0, 10, 50, 100 and 500 J ml -1 ). Therefore, 1 % w/w POM, and 0.5 % w/w in case of the oPOMs, were embedded into an acidwashed and calcinated fine sand matrix, which simulates the soil mineral matrix. These artificial soils (each 20 g) were stored in 100 ml of 1.6 g cm -3 dense SPT solution for 1 h in 200 ml PE bottles, that did not show measurable release of plastic fragments due to sonication in preliminary tests with a pure fine sand matrix (data not shown). Mechanical stress was applied by use of a sonotrode (Branson© Sonifier 250) as described by Büks  Kaupenjohann (2016). The sonication times corresponding to 0, 10, 50, 100 and 500 J ml -1 were determined by means of the sonotrode's energy output calculated following North (1976). After the ultrasonic treatment, samples were centrifuged at 3500 G for 26 min. The floated POM was removed by use of a water-jet pump, separated and cleaned by rinsing with deionized water on a 0.45 µm cellulose acetate membrane filter until the electrical conductivity of the rinse went below 50 µS cm -1 , and then lyophilized.

Determination of recovery rates
After lyophilization, the recovery rate R=mt m0 -1 was determined by weighing as the ratio of the recovered POM mass after treatment (mt) to the initial POM mass (m0) for all POM types and energy levels. The recovery rates for each replicate were plotted over the energy levels to show initial rates at 0 J ml -1 and the influence of the mechanical stress treatment increasing to 500 J ml -1 (Fig. 1). The recovery rate at a certain energy level is assumed significantly different to the 0 J ml -1 level, if a pairwise t-test results in a p<0.05 (Table 1).

Measurement of particle sizes
All samples continued to be used for particle sizing. After pre-trials have shown that mainly the hydrophobic particles (microplastics and pyrochar) coagulated in distilled water, aggregation was avoided by suspension in 0.1 % w/v Tween© 20 detergent solution and vortexing following Katija et al. (2017). About 100 mg of POM were suspended in 500 ml 0.1 % Tween© 20 solution and size classified with a QICPIC image analysis device (Sympatec GmbH, Clausthal-Zellerfeld, Germany) using a modified method from Kayser et al. (2019). Counts were grouped into 34 size classes from <5.64 µm to 1200-1826.94 µm and plotted as cumulative histograms of each replicate and their mean values ( Fig. 2a and 2b). As the primary criterion for the reduction in particle size, the first 10 % and 50 % quantile (median) values were compared by pairwise t-test between 0 J ml -1 and each other energy level, respectively. As particle size reduction could be significant but still marginal in case of a low variance between parallels and and a low grade of comminution at the same time, the averaged comminution factor (CF) was introduced. It is defined as with i the number of parallels, x0,i the quantile value of the 0 J ml -1 energy level and xi the value of the compared energy level. A sample is then assumed significantly different to the 0 J ml -1 control and not marginal, if the p-value given by the t-test is <0.05 and the comminution factor is >1.1 for the 10 % quantile, the median or both, while its standard deviation is sd<|CF-1|. (Table 2) 6 163 164 3 Results

Resulting recovery rates
All microplastic samples (LD-PE, PET and PBAT) show a constantly high recovery rate of about 97.1±2.5 % in average over the whole range of applied energy levels. In sharp contrast, all soil derived POMs (farmland, forest) and pyrochar were decreasingly recovered along with increasing energy levels and had significant differences to the 0 J ml -1 treatment at ≥10 J ml -1 , ≥100 J ml -1 and ≥100 J ml -1 , respectively. (Fig. 1 , Table 1) 7 0 J ml -1 10 J ml -1 50 J ml -1 100 J ml -1 500 J ml -1

POM size distribution
None of the plastics shows a significant reduction of particle size due to ultrasonic treatment within the 10 % and 50 % quantile. In contrast, at ≥100 J ml -1 the particle size of farm and forest oPOM was significantly reduced compared to the 0 J ml -1 treatment in both quantiles. Ultrasonic treatment also causes a significant comminution of pyrochar, but of mainly the smaller fraction indicated by the 10 % quantile, which appeared at ≥50 J ml -1 and is only interrupted due to an outlier at 100 J ml -1 . The 50 % quantile data (median) remain insignificant. (Fig. 2a and 2b ,   Our experiments indicate that soil derived oPOM and pyrochar embedded into a fine sand matrix are prone to comminution by ultrasonic treatment at energy levels of ≥50 J ml -1 . These values are well below the 300 to 750 J ml -1 given in the literature for the complete disaggregation of various soils (Amelung and Zech, 1999;Oorts et al., 2006;Yang et al., 2009), namely in the range of values given for the destruction of macroaggregates (Amelung and Zech, 1999;Kaiser and Berhe, 2014). In consequence, particle size reduction will appear during most ultrasonic treatments aimed to extract oPOMs from soils. This underpins the former implications by some authors that ultrasonic treatment could lead to particle size artifacts. Microplastic, in contrast, shows a constant particle size distribution over all energy levels and seems to resist ultrasonication within the tested range of 0 to 500 J ml -1 . The recovery of microplastics also shows a constantly high rate of nearly 100 %, which is not affected by the applied energy. In sharp contrast, the recovery rates of soil derived POMs and pyrochar decreased with increasing energies from 95.0-78.6 % to 63.8-35.8 %, which became significant at 50 to 100 J ml -1 and therefore is quite parallel to observed size reduction.
The concurrent decrease of particle size and recovery rate of soil derived POMs and pyrochar and their absence in microplastics indicate, that there is a causal relationship between recovery rate and sensitivity against mechanical stress. We assume a mechanism that retains particles at the mineral phase after comminution. Physical disruption of large and weak particles increases the number of smaller ones, coming along with an increase of surface area and, thus, surface forces (e.g. attraction through charges or hydrophobic interaction) compared to volumetric forces (such as inertial forces). This causes an increased adsorption of small POM to mineral surfaces immediately after the ultrasonic treatment and, in consequence, a stronger retention of those particles observable as a lower recovery rate. This effect appeared in our experiment from energies around 50 J ml -1 with the beginning destruction of oPOM and might also occur with small-sized fPOM during density fractionation without application of mechanical stress.
No matter if the hypothesis on the underlying mechanism is valid, as a consequence of concurrent recovery rate and particle size reduction, farmland, forest and pyrochar POMs remain within the soil sample the more they are disrupted by stronger ultrasonic treatment. Thus, not only particle size artifacts are produced. With increasing energy level the extraction of occluded POM is increasingly hindered and, thus, parts of small POM are extracted with oPOM fractions at higher energy levels or remain within the heavy fraction -a carry-over artifact. This leads to an overestimation of the more strongly bound POM fractions or the mineral-associated organic matter (MOM), that natural part of the soil organic matter (SOM), which is adsorbed on mineral surfaces of the heavy fraction and mainly assumed to be molecular. An overestimation would have an impact e.g. on the assessment of operationally defined carbon pools within landscapes: POM is assigned to carbon pools with turnover times orders of magnitude shorter then MOM, that endures hundreds of years. Malquantifications of these pools, such as counting up to around 36.2 to 64.2 % of POM to the MOM as implied by this work, would have phenomenal influence on e.g. the estimation of SOM decomposition and CO2 emissions from land-use change. Carrying-over SOM from little to highly decomposed fractions also could alienate genuine C:N ratios, which strongly differ between the functional carbon pools (Wagai et al., 2009).
Plastic, in turn, is not prone to disruption by ultranonic treatment and its recovery rates are stable in a wide range of energy levels. We therefore assume that there will be no carry-over of particles due to comminution when extracting microplastics from soils with ultrasonication/density fractionation. In consequence, the extractive performance is higher and subsequent particle size measurements give more valid information about the original particle size spectrum compared to the measurement of farmland, forest and pyrochar POM. This is a positive sign for research on soil microplastic, however, it does not mean that microplastic will be fully extracted from soils by this method. Soil microplastics appear within a wide range of sizes between some nanometers and its upper limit of 5 mm by definition. Their smallest part, produced by physical, chemical and biological erosion within the soil, might also interact stronger with soil mineral surfaces than larger pieces causing enhanced retention onto the soil matrix. Although we have introduced billions of tons of microplastics into ecosystems since the 1950s (Thompson et al., 2009;Geyer et al., 2017), there are still problems in producing microplastic fragments <100 µm on a laboratory scale with adequate use of time and material to perform experiments within this size range.

Conclusion
Unlike weathered and fresh PE, PET and PBAT microplastic, soil derived POMs like occluded POM from farm and forest soils and pyrochar concurrently show comminution and a reduced recovery rate after ultrasonication and subsequent extraction from a sandy matrix. As comminution increases the retention, parts of the farmland, forest and pyrochar POM remain within fractions only extractable with higher energy levels or were bound to the heavy fraction, so that they are misinterpreted as MOM. An overestimation as shown in this study might lead to fundamentally different interpretations of physical protection of SOM, functional carbon pools and the expected mineralization rates in consequence of e.g. land-use change. On the contrary, the extraction of microplastics do not cause additional retention of particles at the mineral phase and do not alienate the particle size spectrum by ultrasonic-driven comminution. We conclude that density fractionation in combination with ultrasonication is an appropriate tool for analyzing occlusion of microplastics within soil aggregates and studying the size distribution of particulate microplastics. Zhang, S., Yang, X., Gertsen, H., Peters, P., Salánki, T., and Geissen, V.: A simple method for the extraction and identification of light density microplastics from soil, Science of the Total Environment, 616, 1056Environment, 616, -1065Environment, 616, , doi: 10.1016Environment, 616, /j.scitotenv.2017Environment, 616, .10.213, 2018