The importance of aromaticity to describe the interactions of organic matter with carbonaceous materials depends on molecular weight and sorbent geometry

Dissolved organic matter (DOM) is ubiquitous in aquatic environments where it interacts with a variety of particles including carbonaceous materials (CMs). The complexity of both DOM and the CMs makes DOM–CM interactions difficult to predict. In this study we have identified the preferential sorption of specific DOM fractions as being dependent on their aromaticity and molecular weight, as well as on the surface properties of the CMs. This was achieved by conducting sorption batch experiments with three types of DOM (humic acid, Suwannee River natural organic matter, and a compost extract) and three types of CMs (graphite, carbon nanotubes, and biochar) with different geometries and surface complexities. The non-adsorbed DOM fraction was analyzed by size exclusion chromatography and preferentially sorbed molecular weight fractions were analyzed by UV/vis and fluorescence spectroscopy. All three sorbent types were found to preferentially sorb aromatic DOM fractions, but DOM fractionation depended on the particular combination of sorbent and sorbate characteristics. Single-walled carbon nanotubes only sorbed the smaller molecular weight fractions (<1 kDa). The sorption of smaller DOM fractions was not accompanied by a preference for less aromatic compounds, contrary to what was suggested in previous studies. While graphite preferentially sorbed the most aromatic DOM fraction (1–3 kDa), the structural heterogeneity of biochar resulted in reduced selectivity, sorbing all DOM > 1 kDa. The results explain the lack of correlation found in previous studies between the amount of aromatic carbon in a bulk DOM and its sorption coefficient. DOM sorption by CMs was generally controlled by DOM aromaticity but complex sorbent surfaces with high porosity, curvatures and functional groups strongly reduced the importance of aromaticity.


Introduction
Natural organic matter (NOM) derived from decaying plant and microbial biomass is ubiquitous in the natural environment, playing a major role in the global carbon cycle. 1 In general, for a water sample, a lter cutoff between 0.2 and 0.7 mm is used operationally to dene the dissolved organic matter (DOM) fraction of NOM. 2 It comprises a complex mixture of humic substances, carbohydrates, lipids, proteins, and other biomacromolecules, that cover a continuum of molecular weights ranging from a few hundred Dalton (Da) to >100 000 Da. 3 DOM aromaticity and the presence of functional groups depends on its origin. 4 Of this variety of compounds, operationally dened humic acid (HA) and fulvic acid (FA) have oen been used as surrogate compounds to represent DOM in experiments. 5 Because of its ubiquitous distribution, DOM can inuence the surface charge, aggregation, and transport of particles, including carbonaceous material (CM) particles, that are present in natural environments. 6 CMs can be produced as a result of incomplete biomass combustion or pyrolysis, as well as petrogenic processes, and are released as particles into the environment in large quantities, amounting to several Tg of carbon per year. 7 CMs such as biochar or activated carbon can be deliberately introduced into the environment for contaminant remediation, soil improvement, or carbon mitigation. On the other hand, carbon nanotubes, fullerenes, and graphite used for technological applications can enter the environment unintentionally during any stage of their life cycles. 7,8 Naturally produced CMs such as soot or wildre chars constitute the largest pool of CMs in the natural environment and can play a considerable role in environmental systems. Once introduced into surface waters, the high sorption capacity of CMs make them effective sorbents for DOM in aqueous systems, affecting the fate of both the DOM and the CMs. 5 Understanding the interactions between DOM and CMs is rendered difficult by the complexity and heterogeneity of DOM. Sorption is oen driven by p-p electron-donor-acceptor (EDA) interactions. Other sorption mechanisms include H-bonds, electrostatic interactions, and hydrophobic effects, with their relative contributions to overall sorption depending on the properties of both sorbent and sorbate. 9,10 Sorbent properties, such as the diameters of carbon nanotubes or the presence of surface functional groups, can affect sorption. 9,[11][12][13] Furthermore, the different chemical and size fractions of DOM are not sorbed equally. Smaller DOM molecules can sorb in mesopores of CMs, while bigger molecules are excluded through steric hindrance, 14-16 but generally preferential sorption of aromatic and high molecular weight DOM fractions by CMs was observed. 9,10,14,17,18 Hence, the percentage of aromatic carbon in a bulk DOM sample would be expected to relate to the extent of sorption to CMs. However, a recently published review on the sorption of a large variety of DOM to carbon nanotubes found no such correlation. 5 Further investigation of the factors other than aromaticity that affect sorption was therefore required to explain this discrepancy.
Since each molecular weight fraction of DOM exhibits different chemical properties, adsorptive fractionation should be regarded as a combined effect of both chemical and physical characteristics. In this study we have therefore analyzed the effect of DOM aromaticity, molecular weight, and CM surface properties on DOM sorptive fractionation. For this purpose, different DOM types of increasing complexity regarding their chemical composition and size distribution were selected as sorbates. HA, representing a more aromatic homogeneous fraction of DOM, was selected to represent interactions with (mostly aromatic) high molecular weight compounds and DOM from Suwannee River NOM (SRNOM) was employed as a natural, more aliphatic, lower molecular weight DOM. A compost extract DOM (CDOM) was chosen as a complex, environmentally representative DOM that contains a variety of undened substances, polar macromolecules, and proteins. Graphite, carbon nanotubes and biochar, having distinct structural properties, were used to compare the effects of different CM surface properties on sorption. These three different types of CMs represent (i) non-porous, mostly aromatic surfaces, (ii) strongly curved aromatic surfaces, and (iii) structurally heterogeneous carbonaceous surfaces with high porosity and a high number of functional groups, respectively. This study aimed to explain the discrepancies found between aromatic carbon content of a bulk DOM and its sorption affinity to CMs by separately analyzing the inuence of DOM aromaticity and molecular weight as well as the CM surface properties on sorptive fractionation of structurally and compositionally complex different DOM.

Materials
Commercial HA (technical grade, CAS number 1415-93-6) extracted from leonardite was purchased from Sigma Aldrich and SRNOM (RO isolate, 2R101N) from the International Humic Substances Society. CDOM was extracted from composted biosolids according to a protocol by Bhaduri et al. 19 DOM stock solutions were prepared by dissolving the dry material in ultrapure water (MQ, 18.2 MU cm, PURELAB Ultra, ELGA LabWater Global Operations), adjusted to pH 8.3 with 0.1 M NaOH to dissolve the material and re-adjusted regularly for 3 days while being agitated on a reciprocal shaker at 125 rpm. Aer 3 days the solution was adjusted to pH 7 with 0.1 M HCl, centrifuged (rcf ¼ 1000 Â g) for 30 minutes, and the supernatant then ltered through a sterile 0.22 mm PES lter (Thermo Scientic, Nalgene, Rapid-Flow). HA was dialyzed (molecular weight cutoff 1 kDa) against MQ for 7 days and HA and CDOM was then puried with Chelex resin (complexing cation exchanger, Sigma-Aldrich, Chelex® 100 sodium form, dry mesh 50-100) to minimize possible effects of inorganic ions on the observed interactions.
Graphite (Sigma Aldrich, 99.99% carbon, CAS: 7782-42-5, <20 mm powder) and single-walled carbon nanotubes (Sigma Aldrich, $95% carbon, CAS: 308068-56-6) were purchased from Sigma Aldrich and used as received. Single-walled carbon nanotubes were chosen as a well-dened and material with the minimum amount of surface functional groups that distinguishes itself from graphite mostly through its tubular geometry. This would allow a more direct comparison of effects that can be attributed to the sorbent geometry. Biochar (SWP700, UK Biochar Research Center, produced from so wood pellets pyrolyzed at 700 C) was crushed and sieved to obtain a size fraction of 64-250 mm. To avoid measurement interference the leachable organic carbon fraction was removed by pre-leaching the biochar in MQ water (2 g L À1 ) for 24 hours, followed by wet sieving (64 mm). 20 The retained biochar was then oven-dried at 40 C.

Sorbent characterization
The specic surface area and pore size distribution of the sorbents were derived from sorption isotherms of N 2 at 77 K aer degassing overnight under a vacuum at 378.15 K, using a Quantachrome Nova 2000 Surface Area and Pore Size Analyzer, following a protocol by Sigmund et al. 21 The total C, H, N and S content was measured using an elemental analyzer (Vario MACRO, Elementar), the ash content was determined by loss on ignition (750 C, 4 hours), and the oxygen content was estimated by mass balance: O [wt%] ¼ 100 À (C + H + N + S + ash content). The results are listed in Table S1. † Surface functional groups for all sorbents were determined by Fourier transform infrared (FT-IR) measurements (Fig. S1) as detailed in the ESI. †

Sorption experiments
Between 0.2 and 20 g L À1 of sorbent was suspended in DOM solutions with 30 mg C L À1 to achieve approximately 20% sorption of DOM. Samples were prepared in 50 mL vials with Teon screw caps and placed on a reciprocating platform shaker at 125 rpm under exclusion of light. Aer an equilibration time of 7 days (chosen based on previously conducted kinetic studies 10,22-24 ) the samples were centrifuged (rcf ¼ 1000 Â g) for 30 minutes and ltered (0.22 mm (ref. 17, 23 and 25-27)) for analysis of the DOM remaining in solution. Sample batches included duplicate samples and sorbent blanks, to avoid pH and temperature disturbance in our observations, all experiments were conducted at constant conditions (pH 7 at 274 K). The pH was monitored before and aer the experiments: it remained constant at 7.1 AE 0.3. The amount of DOM in the supernatant was quantied aer sorption to the CMs using a TOC analyzer (Shimadzu, TOC-L series), measuring non-purgeable organic carbon (NPOC). Samples containing biochar were corrected for blank values. Surface area normalized sorption coefficients (K SA ) were calculated as (C sorbed /specic surface area)/C aqueous .
The UV absorbance was measured using a UV/vis spectrophotometer (PerkinElmer, Lambda 35) with a 10 mm quartz cuvette. SUVA 254 values were calculated by dividing the specic absorbance at 254 nm by the dissolved organic carbon (DOC) concentration (SUVA 254 ¼ 100 Â A 254 /DOC), which can be related to the aromaticity of the sample. 4 The ratios of specic UV absorbance at 250 nm to that at 365 nm (E 2 /E 3 ratios) being negatively correlated with aromaticity were calculated as an indicator of aromaticity. 28,29 Fluorescence measurements were obtained using a spectrophotometer (Horiba FluoroMax 4) in a mirrored glass cuvette (1 cm path length) with a slit width of 3 nm and an integration time of 0.2 seconds. The uorescence index (FI) was calculated as the ratio of the emission intensity at 470 nm to that at 520 nm, obtained at an excitation wavelength of 370 nm. 30 Measurements included peak C (ex. 340/em. 440 nm) and peak B (ex. 275/em. 305 nm), corresponding to the maximum uorescence intensities of humic-like substances and tyrosine/ protein-like substances, respectively. 30 The FI has an inverse correlation with both the degree of humication and the aromatic carbon content, and the C : B ratio describes the ratio of humic-like uorophores to tyrosine-like uorophores. 29,31 The C : B ratio was only used as a supporting indicator to predict aromaticity.
The molecular weight distribution of DOM before and aer sorption was analyzed by size exclusion chromatography (SEC, XBridge® Protein BEH SEC 125 A, 3.5 mm, Waters). The column was used in a liquid chromatography system (Agilent Technologies 8800 Series) equipped with a UV/vis spectrophotometer (Agilent 1200 Series Diode Array and Multiple Wavelength Detector, detection at l ¼ 254 nm and 365 nm) and a uorescence detector (FLD, Agilent 1200 Series, excitation l ¼ 350 nm, emission l ¼ 450 nm). A ow rate of 0.4 mL min À1 was used with a 25 mM sodium phosphate buffer (pH 6.7) as eluent and a sample injection volume of 30 mL. The samples were diluted in the eluent to avoid water peak interference. The column was calibrated with polystyrene sulfonate molecular weight standards (PSS, Polymer Standards Service GmbH) of 0.2, 1.1, 1.9, 3.6, 6.8 and 10.6 kDa. The column void volume was tested with polystyrene sulfonate (molecular weight 282 kDa) and the exclusion limit with acetone. Following elution, specic fractions ( Fig. S2 †) were collected using a fraction collector (Agilent 1200 Series). Signal data from both UV and uorescence detectors were exported in milli absorbance units (mAU) and luminescence units (LU), respectively, and were analyzed using Origin® Graphing and Analysis soware.

The importance of DOM aromaticity for sorption depends on the sorbent properties
Graphite, carbon nanotubes and biochar all showed an overall preference for aromatic DOM fractions, as indicated by the spectroscopic measurements of CDOM, SRNOM and HA before and aer sorption to these three sorbents (Fig. 1a-d). The SUVA 254 values in the non-sorbed DOM fraction decreased slightly aer sorption, indicating preferential sorption of the aromatic DOM fraction regardless of DOM type. While changes in the SUVA 254 value were not always signicant (p < 0.05), the increase in E 2 /E 3 ratio following sorption to all CMs conrmed the preferential sorption of aromatic fractions, which is in line with previous observations. 10,12,16,18,22 The slight increase in the FI of DOM following sorption, accompanied by a decrease in the C : B ratio provided further conrmation that CMs had a general preference for aromatic DOM fractions. An exception was the FI of HA, which remained constant following sorption to carbon nanotubes. This may partially be explained as being a result of the high molecular weight of HA (Fig. 3) and the curved geometry of carbon nanotubes, as will be discussed in Section 3.2. In any case, HA is only one of the many constituents of heterogeneous natural DOM. The fact that the FI of both SRNOM and CDOM indicated a fractionation aer sorption emphasizes the shortcomings of using HA as a DOM surrogate when investigating interactions in natural environments. Since HA represents a rather dened DOM fraction and only occupies specic types of sorption sites, it cannot reect the full range of interactions that occur between DOM and CMs. Thus, HA may not respond in the same way as more complex DOM to changes in the factors that affect sorption.
DOM-CM interactions are mostly governed by hydrophobic effects and p-p EDA interactions between aromatic carbon structures of the sorbent and the sorbate. 32 Interactions between graphite and the most aromatic DOM fractions are therefore strongly favored. The formation of H-bonds plays a minor role due to the absence of functional groups on the surface of graphite and carbon nanotubes (carbon purity is >99.99% and >95%, respectively; Table S1 †), suggesting that DOM sorption to CMs occur mainly through p-p EDA interactions. 32 This may explain why graphite in particular showed a strong preference for the aromatic fraction, as shown by the markedly reduced SUVA 254 values (Fig. 1a).
DOM with higher aromaticity was oen found to result in a higher distribution coefficient (K D ) for sorption to carbon nanotubes, graphite, or activated carbon. 9,10,14,17 However, Zhang et al. (2015) found that the K D for sorption of HA to carbon nanotubes decreased systematically with increasing molecular weight and aromaticity. 33 Recently Engel and Chefetz (2019) presented a linear correlation between the percentage of aromatic carbon and the extent of HA or FA sorption to carbon nanotubes. 5 However, this correlation could not be conrmed when extending the comparison across multiple studies and a range of different types of DOM, rather than just FA and HA. The lack of any linear correlation between aromatic carbon content and the adsorption coefficient indicates that there are additional factors that need to be considered. Our ndings show that the properties of the sorbent can affect the importance of aromaticity for sorption (Fig. 2).
K SA values for graphite and carbon nanotubes were considerably greater than that for biochar. This can be explained by the negative charges on the biochar surface that can cause electrostatic repulsion of the negatively charged DOM, reducing the K SA considerably. 34 Furthermore, CMs show a preference for aromatic DOM fractions, which can form p-p EDA interactions with the aromatic surfaces of the CMs. 9,14 Graphite and carbon nanotube surfaces consist of aromatic sheets of carbon, whereas biochar surfaces can contain plenty of oxygen-containing functional groups among the aromatic surface patches. 35 These highly condensed aromatic patches are usually smaller than 5 nm as demonstrated by highresolution transmission electron microscopy of high temperature biochar surfaces. 36 The highest K SA values would therefore be expected for the most aromatic bulk DOM sorbing to graphite or carbon nanotubes. Fig. 1 shows that HA is the most aromatic bulk DOM in our set, but it was nevertheless not the most strongly sorbed by either graphite or carbon nanotubes. Although biochar mostly showed a less marked preference than graphite for the aromatic DOM fraction (Fig. 1) it sorbed HA, the overall most aromatic bulk DOM, more strongly than it sorbed SRNOM or CDOM. This supports the hypothesis that DOM aromaticity is an important factor affecting DOM sorption to all types of CMs (Fig. 1), but that on its own it is unable to account for all differences in DOM sorption, as discussed in the following sections.

DOM fractionation is controlled by the interplay between sorbate molecular weight and sorbent geometry
Although DOM aromaticity generally correlates with molecular weight, this does not necessarily apply for all DOM types. 16 Since both factors can exert different inuences on sorption, in a rst step they have been considered separately in this study. Firstly, to investigate the inuence of molecular weight on sorption, size exclusion chromatograms of the three DOMs were measured before and aer sorption to CMs (Fig. 3a and S3 †). Our data show that graphite strongly sorbed the fraction with molecular weights between 1 kDa and 3 kDa, biochar was less selective and sorbed all fractions with molecular weights >1 kDa, and carbon nanotubes mainly sorbed those fractions with molecular weights <1 kDa. The increase in high molecular weight DOM during HA sorption by biochar is most likely due to aggregation of smaller HA molecules due to the release of calcium from biochar samples. 37 Although each sorbent preferentially sorbed a different molecular weight fraction, the aromatic DOM fractions were preferred by all sorbents, as shown in Section 3.1. To understand the combined inuence that both factors have on sorption, we investigated the aromaticity of each of the preferentially sorbed fractions separately ( Fig. S2 and S4 and Table S2 †).
The SUVA 254 values decreased together with molecular weight, thus linking aromaticity with molecular weightas has been previously reported for other types of DOM. 28,38 This appears to contradict the observation that carbon nanotubes preferentially sorbed the aromatic fraction (as supported by the spectroscopic aromaticity indicators shown in Fig. 1) and yet showed preferential sorption of the smaller and relatively less aromatic fractions with molecular weights of <1 kDa (Fig. 3a). A possible explanation can be derived from SEC measurements with simultaneous detection at 250 nm and 365 nm, which yielded the E 2 /E 3 ratio of each molecular weight fraction before and aer sorption (Fig. 3b). As the E 2 /E 3 ratio is inversely correlated with aromaticity the graphs show the aromaticity of the original DOM and the DOM aer sorption for each molecular weight.
The E 2 /E 3 ratio for the smaller molecular weight DOM fraction that was preferentially sorbed by carbon nanotubes was considerably greater than that for the non-sorbed DOM ( Fig. 3b), indicating preferential sorption of aromatic compounds within each molecular weight fraction. Although the spectroscopic measurements (Fig. S4 †) indicate that the smaller molecular weight fractions were less aromatic overall, Fig. 3b illustrates that there were aromatic compounds present within the smaller molecular weight fractions that were preferentially sorbed. This is therefore consistent with an overall preference for the aromatic fraction in bulk measurements (Fig. 1). All three sorbents showed a stronger fractionation in the lower molecular weight range (Fig. 3b). Single-walled carbon nanotubes only sorbed these lower molecular weight compounds and accordingly only fractionated these. Biochar and graphite, however, especially sorbed the higher molecular weight compounds but showed a stronger fractionation in the lower molecular weight range. This indicates that the lower molecular weight compounds are chemically more heterogeneous; while the high molecular weight compounds (2.5-1 kDa) are mostly aromatic, the smaller fractions contain both aliphatic and aromatic compounds resulting in a stronger chemical fractionation. The highest molecular weight fraction (>2.5 kDa) contains high amounts of tyrosine-and protein-like compounds which is shown by the uorescence measurements (Fig. S5 †) but possibly not detected by the E 2 /E 3 ratio measurements (Fig. 3b). Thus, the highest molecular weight fraction could be equally fractionated as the smaller one. Sorbent surface properties and geometry can explain why higher molecular weight DOM fractions are sorbed less by carbon nanotubes (Fig. 2) despite being more aromatic. The carbon nanotube pore size distribution (Fig. S6 †) extends over a large range of mesopores with pore widths between 3 and 15 nm. However, the use of pore size distributions for carbon nanotubes has to be interpreted with caution as the dry aggregated powder may exhibit different pores than the actual dispersed carbon nanotubes in water. Sorbed DOM disperses and stabilizes carbon nanotubes in solution and may limit DOM sorption in the interstitial spaces between carbon nanotube bundles. 39,40 The inner pores of carbon nanotubes are too small for the diffusion of DOM molecules and it has been argued that possible sorption mostly occurs on the cylindrical surface of the outermost carbon nanotubes. 17,18,41 Pore exclusion effects can hardly explain the selectivity of carbon nanotubes for certain molecular weight fractions; most studies to date found preferential adsorption of higher molecular weight fractions of DOM by carbon nanotubes since that this is the most aromatic fraction. 10,14,17,18 Although the sorption of some of the DOM fractions may occur in the grooves of carbon nanotube aggregates, 42 it does not explain the exclusion of the higher molecular weight fractions that are overall more aromatic than the lower molecular weight fractions. Carbon nanotubes preferentially sorbed aromatic compounds even within the small MW fractions (Fig. 3b), which demonstrates the inuence of the sorbent geometry. Aromatic structures are relatively rigid, with low rotational degrees of freedom; they are consequently not able to adjust their conformation to the highly curved surfaces of carbon nanotube with small diameters and thereby increase the contact area. 30,33 This restricts sorption to only the smaller aromatic compounds, which maintain the face-to-face orientation allowing p-p EDA interactions to take place 32 and is supported by the previous nding that HA sorption to carbon nanotubes decreases as their outer diameter decreases 13 and as the molecular weight of the HA increases. 33 It was previously suggested that aliphatic compounds have a high affinity to carbon nanotubes due to their rotational degrees of freedom resulting in an improved structural compatibility than rigid aromatic compounds. 33 However, this study showed that the sorption of smaller molecular weight DOM by carbon nanotubes does not correlate with a preference for aliphatic and less aromatic compounds. Aromatic compounds are, in fact, preferentially sorbed despite being less abundant in the small molecular weight fractions. This shows that aromaticity is a key factor determining adsorption to carbonaceous materials when considering pure and dened sorbent structures. For more heterogeneous sorbent surfaces as found in the natural environment polar and aliphatic moieties can play a considerable role for sorption. 43

Sorbent heterogeneity decreases selectivity for aromatic fractions
Of the sorbents investigated, graphite showed the strongest preference for the aromatic DOM fraction (Fig. 1). Accordingly, the molecular weight fraction between 1 kDa and 3 kDa, which was the most aromatic fraction according to the SUVA 254 value and the C : B ratio (Fig. S4 †), was the fraction most efficiently sorbed by graphite (Fig. 3a). Aromatic molecules can maximize their contact areas on graphene surfaces because of their planar nature, in contrast to the strongly curved surfaces of carbon nanotubes. 44 Interactions between aromatic high molecular weight fractions and the planar aromatic graphite surfaces are therefore favored.
Graphite sorbed the highest molecular weight fraction (>3 kDa) considerably less effectively than the fraction with molecular weights between 1 and 3 kDa (Fig. 3a) due to differences in composition between the two fractions. Although earlier observations that aromaticity generally decreases with molecular weight were conrmed, the C : B ratio of the highest molecular weight fraction was lower than that of the secondhighest fraction (Fig. S4 †). The comparably more pronounced peak B indicates the presence of protein-like or tyrosine-like compounds. 30 Furthermore, SEC chromatograms indicated the presence of protein-like and tryptophan-like compounds in the highest molecular weight fraction of CDOM (Fig. S5 †). 30 This structurally more heterogeneous fraction sorbs to CM surfaces through different mechanisms than the more aromatic hydrophobic fractions do. 12 Graphite has hardly any surface functional groups and preferably sorbs the aromatic DOM fraction through hydrophobic and p-p EDA interactions on its planar aromatic surface. Biochar, on the other hand, contains functional groups on its otherwise carbonized and aromatic surface. Previous research on SWP700 biochar has identied surface functional groups by FT-IR, 45,46 Boehm titration, 47 and X-ray photoelectron spectroscopy. 48 FT-IR measurements (Fig. S1 †) indicate the presence of functional groups such as hydroxyl groups that can exert a different sorption behavior than the graphite or carbon nanotube surfaces: next to p-p EDA interactions and hydrophobic effects, which contribute strongly to sorption to the aromatic surface patches of biochar, electrostatic interactions and the formation of H-bonds between oxygen-containing functional groups of biochar and polar groups of DOM can occur. Biochar therefore sorbed the highest molecular weight fraction just as strongly as the second-highest molecular weight fraction, despite their different chemical characteristics. In contrast, graphite preferentially sorbed the second-highest molecular weight fraction, which did not contain as much polar less aromatic carbon as the highest molecular weight fraction. Furthermore, polar interactions were shown to strongly contribute to the sorption of HA to graphene oxides. 49 The relative contribution of each type of interaction is dependent on the abundance of surface functional groups. 10,25 Low temperature produced biochars have a lower aromaticity and higher number of functional groups and better adsorb polar DOM compared to the high temperature SWP700 used in this study. 25 Previous studies explained sorption of distinct molecular weight fractions by pore sieving, i.e., large organic molecules and high molecular weight DOM molecular weight are excluded from micro-and mesopores. 14,15 The biochar pore size distribution (Fig. S6 †) shows that biochar has a large mesopore volume between 1.8 and 2 nm. Measurements of the same biochar material by Sigmund et al. (2017) show that a large proportion of pores are in the micropore range 21 which are likely inaccessible for most DOM molecules. 24,50 Although sorption of the smallest DOM fractions in the range of the measured lower mesopores and reported micropores may have occurred through pore lling, pore sieving effects do not explain the sorption of the fraction >1 kDa.

Conclusion
The ubiquitous input of highly aromatic CMs to surface waters can inuence the composition of the DOM, which can in turn affect the fate of both DOM and CMs. By individually investigating sorptive fractionation of three compositionally different DOM types we could demonstrate how DOM-CM interactions are inuenced by a variety of factors including aromaticity, molecular weight, sorbent functionality, and sorbent geometry. Most importantly, these factors were interdependent and could not be considered independently. It is therefore possible for a rather aromatic DOM fraction to not be sorbed because of other factors such as the sorbate molecular weight or sorbent geometry not being favorable for sorption. This may have led to discrepancies between the results of previous investigations and needs to be considered in any future research.
HA is oen used as a surrogate for natural DOM and its inuence on processes occurring under natural conditions, even though it only represents a discrete fraction of the highly heterogeneous DOM. Our investigations have shown that this inuences the way it interacts with CMs, so that the choice of DOM can have a strong inuence on the outcome of any experiment, especially when ternary interactions are analyzed. It is therefore critically important in any future research to consider carefully the choice of DOM to be used as environmentally representative DOM.

Conflicts of interest
There are no conicts to declare.