Freshwater Suspended Particulate Matter – Key Components and Processes in Floc Formation and Dynamics

Freshwater suspended particulate matter (SPM) plays an important role in many biogeochemical cycles and serves multiple ecosystem functions. Most SPM is present as complex floc-like aggregate structures composed of various minerals and organic matter from the molecular to the organism level. Flocs provide habitat for microbes and feed for larger organisms. They constitute microbial bioreactors, with prominent roles in carbon and inorganic nutrient cycles, and transport nutrients as well as pollutants, affecting sediments, inundation zones, and the ocean. Composition, structure, size, and concentration of SPM flocs are subject to high spatiotemporal variability. Floc formation processes and compositional or morphological dynamics can be established around three functional components: phyllosilicates, iron oxides/(oxy)hydroxides (FeOx), and microbial extracellular polymeric substances (EPS). These components and their interactions increase heterogeneity in surface properties, enhancing flocculation. Phyllosilicates exhibit intrinsic heterogeneities in surface charge and hydrophobicity. They are preferential substrates for precipitation or attachment of reactive FeOx. FeOx form patchy coatings on minerals, especially on phyllosilicates, which increase surface charge heterogeneities. Both, phyllosilicates and FeOx strongly adsorb natural organic matter (NOM), preferentially certain EPS. EPS comprise various substances with heterogeneous properties that make them a sticky mixture, enhancing flocculation. Microbial metabolism, and thus EPS release, is supported by the high adsorption capacity and favorable nutrient composition of phyllosilicates, and FeOx supply essential Fe.


INTRODUCTION
Freshwater suspended particulate matter (SPM) is an integral part of global geological, geochemical and biological cycles in the aquatic environment, due to its ubiquitous presence, mobility, and physicochemical properties (Eisma, 1993). About 10 9 -10 13 tons of SPM are discharged into the oceans every year (Beusen et al., 2005;Meybeck, 1982;Viers et al., 2009;Walling and Webb, 1987), supplying loads of inorganic (Martin and Meybeck, 1979) and organic (Meybeck, 1982) matter to the marine ecosystem. For most elements, particulate loads by far exceed dissolved ones (Jeandel and Oelkers, 2015;Meybeck, 1976), making riverine SPM the major nutrient-supply to the oceans (Beusen et al., 2005;Eisma, 1993;Jeandel and Oelkers, 2015). Substantial particulate carbon loads (Meybeck, 1982) indicate the importance of SPM in global carbon-cycling. About 44 % of the annual terrestrial C flux into rivers are returned to the atmosphere, ~22 % are buried in lake and reservoir sediments, and about one third reaches the ocean (Aufdenkampe et al., 2011), where salinity boosts aggregation and sedimentation (Eisma, 1993). Both, freshwater and marine sediments constitute important carbon sinks, and burial may subject SPM to diagenesis or hydrocarbon formation processes, closing the geological cycle (Eisma, 1993;Leithold et al., 2016).
Freshwater CO2 release is partly attributed to colonized SPM flocs functioning as bioreactors ( Figure 1). SPM pools elemental nutrients and organic substrates which fuel aquatic microbial production and attract microorganisms to attach (Grimm, 2017;Grossart and Ploug, 2000;Servais and Garnier, 2006). To access nutrients, microbes modify pH and redox potentials inside SPM and create a microenvironment with increased enzyme activity. During the degradation of organic matter biomass is built up, inorganic nutrients are recycled and partly released to the water together with labile organic substances, feeding free-living organisms (Grossart and Simon, 1998;Simon et al., 2002). Microbial biomass improves the nutritional value of SPM (Crump et al., 2017;Rosi-Marshall and Meyer, 2004) for grazing filter feeders (e.g. crustaceans, fish, whales), making SPM an essential part of the pelagic food web ( Figure 1) (Zimmermann-Timm, 2002). SPM is mobile and connects the pelagic to the benthic zone by sedimentation and resuspension, rivers to lakes and oceans by advective transport, and terrestrial to aquatic environments via runoff/erosion and flooding ( Figure 1) (Aufdenkampe et al., 2011;Zimmermann-Timm, 2002).
Sedimentation is favored at slow flow and in zones of reduced turbulence (Eisma, 1993) and supplies nutrients to benthic organisms. Floc formation continues in sediments (Wotton, 2007;Zimmermann-Timm, 2002) until increased shear through high discharge in rivers, tidal impact in estuaries, or temperature/wind-driven mixing in lakes, (re)mobilizes SPM (Eisma, 1993). High concentrations of SPM can physically impair aquatic ecosystems. Increased turbidity hampers bioproduction due to reduced light and mechanical stress, and mineralization of organic-rich SPM upon deposition can induce oxygen shortage in poorly mixed waters (Bilotta and Brazier, 2008).
Different conceptions of SPM are reflected in the diversity of terms across disciplines. The most common ones are suspended sediments (Droppo et al., 1998;Hillier, 2001;McConnachie and Petticrew, 2006;Pont et al., 2002), suspended (particulate) matter (Baalousha et al., 2019;Breitenbach et al., 1999;Le Meur et al., 2016;Ran et al., 2000), (total) suspended solids (Berger et al., 1996;Bilotta et al., 2012;Droppo and Ongley, 1994), which share the idea of a material suspended under given conditions. Suspended sediments stresses sedimentation and resuspension as a major SPM control but falls short of implying material inputs by runoff, soil-erosion and pelagic bioproduction. Suspended solids underlines the distinction from the dissolved phase, but dissolved substances are building blocks incorporated by adsorption or diffusion into SPM pores.
So far, SPM in river systems has received much less attention as compared to estuarine and marine systems. In the light of recent research addressing particulate pollutants, a fundamental understanding of freshwater SPM is a prerequisite to assess the fate of these emerging contaminants in river systems. Knowledge about freshwater SPM is found in manuscripts from various disciplines, ranging from geosciences through microbiology and ecology to environmental sciences and modelling. There is no comprehensive review available on the formation of SPM in river systems which analyzed the plethora of information from these different disciplines. Many studies have considered SPM as substrate for contaminant interactions, controlling behaviour, fate and bioavailability of pollutants. In most cases SPM was treated as a black-box, constituting the particle-bound fraction of pollutants. Here we abandon this contaminant-centred view and focus on the characteristics and dynamics of SPM as pivotal micro-reactors in freshwater chemistry and ecology. The aim of this review is to analyze the complexity of freshwater SPM composition and identify the key components and mechanisms governing physicochemical floc formation processes with a focus on river systems. We elucidate the quantitative and qualitative mineralogical and organic composition of riverine SPM and identify components that play crucial structuralfunctional roles in SPM formation. The origins, flocculation-relevant properties and dynamic interactions of these SPM components are discussed in detail, and spatiotemporal dynamics affecting floc formation, morphology, and composition are explained ( Figure 1).

COMPOSITION AND STRUCTURAL FEATURES OF RIVERINE SPM
With about 1 -30 %wt organic matter content, riverine SPM mass, not necessarily volume, is overall dominated by minerals, mainly phyllosilicates, carbonates, quartz, and feldspars. The organic matter content and its composition underlie seasonal and discharge-related dynamics.
Generally, less bioavailable fulvic and humic-like substances dominate, but during periods of substantial bioproduction, easily biodegradable substances can constitute up to 70 %wt of the organic matter in SPM.
The mineral composition is of course coined by the catchment lithology and soil-types (Håkanson et al., 2005;Le Meur et al., 2016;Lučić et al., 2019;Pont et al., 2002), but was found to be quite constant throughout seasons and different flow regimes (Nachtnebel et al., 1998;Pont et al., 2002;Slomberg et al., 2016;Wall and Wilding, 1976), except for the mentioned grain-size relation (higher quartz, feldspar, carbonate concentrations in larger size fractions). a) original data included organic fraction; numbers were recalculated so that the total minerals represent 100 % b) only data for 0.5 -1µm size fraction were included c) Averages should only provide orientation, as data are of different quality (regarding sampling frequencies and seasons).

Particulate organic matter (POM) fraction mostly below one third
Particulate organic matter (POM) denotes the organic mass fraction in SPM, distinguished from dissolved organic matter (DOM) by filtration (POM > 0.2 or > 0.45 µm). Sometimes colloidal organic matter (< 0.45 µm and > 1 kDa) is separated as well. This distinction is solely operational, since natural organic matter (NOM) constitutes a continuum, from molecules via detrital matter up to microorganisms, and the POM fraction integrates parts of all size-fractions.
POM is typically determined as mass loss on ignition or applying conversion factors to particulate organic carbon (POC). The reported POM fractions in SPM of 9 rivers range from a few percent to above 30 %wt, with a median of 15 %wt (Table 2). Employing the reported conversion factors 1.7 -2.5 (Guo and Macdonald, 2006;Hillier, 2001;Wall and Wilding, 1976) to POC data (SI- Figure 1), 95 % of the reviewed 54 rivers fall within the range of 1.2 -40 %wt POM, with a median of 6.5 -9.5 %wt. The POM content was found to be inversely correlated with SPM size and reached its maximum in the smallest (1 -0.2 µm) SPM subfraction (Ran et al., 2000). This is confirmed by 55 % of the total organic carbon load being colloidal (0.45 µm -1 kDa) and 6 -20 % particulate (> 0.45 µm), leaving 15 -28 % in the truly dissolved fraction (< 1 kDa) (Guo et al., 2003;Guo and Macdonald, 2006).

15.3
(1.3…33.0) a) LOI = loss on ignition; POC = particulate organic carbon b) same factor as for Chena River located in the same area c) mass-weighted average for size fractions > 0.2 µm calculated from size fractionated data Quantitative data on the composition of POM is scarce. Typically, easily biodegradable labile and hardly degradable refractory organic matter are distinguished. The labile fraction mainly comprises carbohydrates and protein-like substances, the refractory fraction represents mostly pedogenic, but also aquagenic fulvic or humic-like substances. The POC composition depends on the total SPM load (Ittekkot, 1988) (Table 3). Overall, refractory substances dominate with about 29 -96 %. Average labile POC fractions of 13 major world rivers amount to 4.3 -46.6 %, but local or seasonal maxima can be higher, e.g 71 % in river Lütschine (Chanudet and Filella, 2007). In the labile fraction protein-like substances mostly prevail over carbohydrates. Table 3. Refractory (R-) and labile (L-) POC fractions in riverine SPM. L-POC includes carbohydrates (CH-C) and protein-like substances (Prot.-C). Data represent min…max (average).

POM content and composition are dynamic.
Whether the quality of POM is rather refractory or labile is related to the source of NOM.
The maximum POC values are frequently measured from spring to early summer (Chanudet and Filella, 2007;Henning et al., 2001;Servais and Garnier, 2006) at intermediate discharge (low runoff-to-precipitation ratios) and increased temperatures (Besemer et al., 2009;Onstad et al., 2000). In such conditions, inputs of soil-derived carbon and inorganic nutrients meet the onset of increasing bioproduction, while the shear suffices to resuspend (and avoid sedimentation of) light organic-rich fractions but not to mobilize minerals from the riverbed or banks.

Floc-like aggregates are the dominant SPM structure.
The primary mineral and organic constituents of SPM are hardly found isolated, but assembled in complex floc or aggregate-like structures (Bibby and Webster-Brown, 2005;Droppo, 2001;Henning et al., 2001;Kravchishina et al., 2010;Lučić et al., 2019;Nomizu et al., 1987;Wilkinson et al., 1997;Zimmermann-Timm, 2002), which constitute more than 90 % of the total SPM volume, 10 -60 % by number (Droppo and Ongley, 1994;Droppo and Stone, 1994) and about 60 % by mass (Woodward and Walling, 2007). Droppo and Ongley (1994) summarized three possible origins of floc or aggregate-like structures (1) soil-erosion supplying some water-stable aggregates, (2) flocculation within the river by physical, chemical and biological means, and (3) hybrids combining both. Traditionally, flocs referred to a riverine origin, whereas aggregates were considered soil-borne. Also structural differences were inferred: aggregates show higher density and compactness, lower porosity and rounder shape (Droppo et al., 2005;Woodward et al., 2002). A distinction within a river-water sample was, however, not possible (Droppo et al., 2005). Most soil aggregates (partially) disaggregate in rivers (Grangeon et al., 2014); heavier factions are incorporated into sediments or organic-rich sediment surface layers, from where they can be resuspended as flocs , whereas light fractions stay suspended and form flocs or aggregate-floc hybrids in the water column (Droppo et al., 2005). Biofilm fragments that detach from their substratum also contribute to SPM (Wotton, 2011(Wotton, , 2007 and may capture minerals within their organic matrix while suspended (Figure 1). In this work, the term floc refers to any complex SPM structures.
The formation and stability of flocs depends on the constituents' physicochemical surface properties, which give rise to short-range interaction forces. Based on particle aggregation theory (Elimelech et al., 1995;Gregory, 2005), electrostatic forces arising from the constituents' structural charge or ionized functional groups, as well as van der Waals and hydrophobic attractions are most relevant. Organic molecules may also favor flocculation, acting as bridges between particles or inducing surface charge heterogeneities or neutralization upon adsorption (Buffle et al., 1998).

CRUCIAL STRUCTURAL-FUNCTIONAL SPM COMPONENTS
The major physicochemical processes driving SPM floc formation can be established around three crucial structural-functional components due to their physicochemical surface properties: EPS, FeOx, and phyllosilicates ( Figure 1).

Microbial EPS-type NOM is the "glue" in SPM flocs.
NOM is omnipresent in aquatic systems and no solely mineral associations are found in riverine SPM samples (Droppo, 2001;Droppo et al., 1997). Different types of NOM exhibit different flocculation potentials (Buffle et al., 1998;Wilkinson et al., 1997) depending on their molecular characteristics and the degree of particle surface coverage (Buffle et al., 1998;Philippe and Schaumann, 2014;X. Wang et al., 2021). Adsorbed rigid, high-MW molecules may form bridges to other particles, while low-MW and flexible molecules flatten on the particle surface but may induce flocculation via charge neutralization, or patchy particle surface coverage (electrostatic patches). Particles are stabilized against flocculation if fully coated by highly charged, low-MW molecules, or by large, rigid molecules, inducing steric hindrance. An early conceptual model of complex SPM flocs (Greiser 1988, cited in Henning et al., 2001 already stressed the importance of organic gels, binding other particles. Evidence of network-like and fibrillar organic structures was found in numerous electron microscopy images of natural SPM (Buffle and Leppard, 1995;Chanudet and Filella, 2007;Droppo et al., 1997;Filella et al., 1993;Henning et al., 2001Henning et al., , 2000Lartiges et al., 2001;Liss et al., 1996;Neu, 2000;Perret et al., 1994) and a relation between seasonal increase of microbial activity and increased network densityinvolving thicker, longer, more branched organic fibrils -was observed (Chanudet and Filella, 2007;Henning et al., 2001). These fibrils were concluded to consist of macromolecular microbial EPS (Filella et al., 1993;Leppard, 1997) and pores in SPM were found mostly filled with EPS (Liss et al., 1996;Neu, 2000;Wheatland et al., 2017). At times of high microbial activity, a matrix of macromolecular EPS integrates mineral particles, yielding large fluffy flocs; while in winter, when pedogenic HS is the major NOM source, SPM is dominated by small, compact, mineral-rich flocs and single mineral grains (Böckelmann, 2000;Droppo and Ongley, 1994;Henning et al., 2001;Sigleo and Helz, 1981;Walling and Phillips, 2004) (Figure 1).
Heterogeneities may even persist on a particle level: On goethite, EPS formed nano-scaled patches enriched in different subfractions (proteins, lipids, polysaccharides), suggesting that patches exhibit distinct physicochemical properties (Xinran .

Microbial EPS are a diverse mix of substances.
The origin of EPS is mostly linked to microbes, such as bacteria and microalgae ( Figure 1), but also fungi and archaea (Droppo et al., 2004;Flemming and Wingender, 2010;Zimmermann-Timm, 2002) or larger organisms contribute -from insect silk up to fish mucus (Wotton, 2011).
EPS are dominated by polysaccharides and a previously underestimated share of proteins (Flemming and Wingender, 2010;Laspidou, 2002), complemented by some extracellular DNA, lipids and fragments of all these (peptides, amino-and nucleic acids, sugars) (Flemming and Wingender, 2010). Here we consider EPS the sum of biopolymers released by microorganisms, (a) actively as part of their metabolic activities (exudates) or (b) passively upon cell lysis (intracellular substances) or during cell decomposition (structural biopolymers) (Laspidou, 2002).
Active exudation serves purposes of attachment/detachment to/from surfaces or each other; protection from toxic components, UV-radiation, oxidation, desiccation or mechanical stress; nutrition through exoenzymes, capture of nutrients and particles in the EPS matrix or use of EPS as substrate; and establishment of microenvironments with physicochemical gradients (Flemming and Wingender, 2010). Most exudates are either matrix-forming polysaccharides (e.g. xanthan, pullulan, dextran, alginate (Buffle et al., 1998)) or enzymatic proteins Wingender, 2010, 2001). Matrix polysaccharides are long molecular chains that may associate to form fine strands attached to cells or complex networks surrounding cells (Sutherland, 2001) and are comparatively resistant to degradation (Flemming and Wingender, 2001). Enzymatic proteins cleave organic matter, yielding bioavailable low-MW products, or degrade the EPS matrix, allowing detachment (Flemming and Wingender, 2010). Some lipopeptides are exuded as well (Flemming and Wingender, 2010), they render surfaces more hydrophobic (e.g. serrawettin) or hydrophilic (e.g. surfactin, viscosin), aiding attachment or detachment (Neu, 1996). More recently, the active exudation of DNA, which forms grid-or network-like filament structures, was attributed an attachment function (Flemming and Wingender, 2010).

Heterogeneous molecular characteristics confer stickiness to EPS
There are neutral as well as charged EPS. Anionic character mostly results from carboxyl-groups (in uronic acid and ketal-linked pyruvate), and partly from inorganic groups such as phosphate (e.g. in DNA or phospholipids) or sulphate (e.g. in cyanobacterial EPS), while cationic character arises mostly from amino groups. Most polysaccharides are polyanionic heteropolysaccharides (e.g. alginate, xanthan, colonic acid), while homopolysaccharides such as fructans and glucans are usually neutral (e.g. cellulose, pullulan, dextran, levan). Proteins are ampholytic, as they consist of basic and acidic amino acids carrying carboxyl-and amino groups (Flemming and Wingender, 2010;Sutherland, 2001). Polycationic EPS, exuded by some gram-positive bacteria (Vu et al., 2009), are rare (an example being adhesin) and the hydrophilic moieties of lipid-type amphophiles (phospholipids, glycolipids and lipopolysaccharides) can exhibit either charge condition (Neu, 1996). Charged EPS interact with particle surfaces through electrostatic interactions, H-bonding or ion-bridging (Philippe and Schaumann, 2014). Cationic amino groups for example sorb to negatively charged basal planes of montmorillonite, illite or kaolinite, while anionic carboxylgroups and phosphate groups (on phospholipids, phosphorylated proteins or DNA) prefer mostly positive Fe-or Al-(hydr)oxide surfaces and phyllosilicate edges, or require ion-bridges (Cao et al., 2011;Gu, 1992;Hedges and Hare, 1987;Lin et al., 2016;Xinran Liu et al., 2013;Yeasmin et al., 2014). Compared to carboxyl-rich pedogenic HS, negative charge densities of most aquagenic EPS are low, which eases interactions with mostly negative mineral surfaces (Buffle et al., 1998), and charge heterogeneities within the EPS mix or even on single EPS molecules (e.g proteins) favor flocculation.
Hydrophobic interactions let hydrophobic parts of EPS molecules flatten on particle surfaces, as modelled for negative montmorillonite basal planes (Henao and Mazeau, 2008), while hydrophilic side chains (e.g. carrying charged carboxyl-or hydroxyl-groups) extend and may induce flocculation (Henao and Mazeau, 2008;Labille et al., 2005). This reveals the importance of structural EPS features like molecular weight, structural flexibility/rigidity and conformation.
Rigid, high-MW EPS have a higher chance to induce bridging flocculation, as they extend from the particle surface into the water where electrostatic forces become negligible. A full coating by rigid biopolymers, however, prevents flocculation through steric hindrance. (Buffle et al., 1998) Rigid, high-MW EPS mainly comprise structural biopolymers (e.g. cell-wall components) and polysaccharidic exudates, but also DNA (Buffle et al., 1998). Polysaccharide primary structures are long molecular chains, either branched for most heteropolysaccharides or linear for many homopolysaccharides (Sutherland, 1990). Their rigidity depends on the types of linkages between their monomers: 1,3-β-or 1,4-β-linkages confer rigidity (e.g. the cellulosic backbone in xanthan), while other linkages may yield more flexible structures (e.g. 1,2-α-or 1,6-α-linkages in dextran) (Sletmoen et al., 2003;Sutherland, 2001). Highly charged EPS behave rigid because they strongly bind water (sometimes up to 80 %) (Buffle et al., 1998). Some EPS can also self-assemble in rigid supramolecular structures via (1) H-bonding, as between hydroxyl-or methyl-groups, (2) electrostatic or (cation-induced) ionic interactions, as on carboxyl-or sulfate groups, (3) molecular The most common secondary conformation of polysaccharidic exudates (and DNA) in environmental conditions are rigid double or triple helices, yielding long, thin, fibrous structures in single strands or networks (Sutherland, 2001).
With a MW of 10 4 -10 5 g mol -1 , fibrils reach a thickness of a few nm and a length up to the µmrange. Nano-sized minerals can attach to fibril segments with favorable characteristics and act as connectors between fibrils, yielding large, loose flocs. (Buffle et al., 1998) Polysaccharides can also form gels, which can bind particles. Gelling is favored by carboxyl-groups that allow bridging by multivalent cations (e.g. in gellan or alginate). Increasing acylation, in turn, renders polysaccharides more hydrophobic, which inhibits interactions with ions, weakening gels and promoting flexible random coil structures. Weak gels will not persist in rivers as they dissolve in excess solvent. (Sutherland, 2001) Flexible polymers often exhibit globular appearance and easily change conformation, which lets them flatten on the particle surface and excludes bridging, but they may induce aggregation via electrostatic patch formation or charge neutralization (Buffle et al., 1998;Gregory, 2005). Flexible EPS comprise proteins, intracellular reserve polysaccharides and aquagenic refractory organic matter (i.e. recalcitrant fragments and recombinations of degradation products of microbial cells) (Buffle et al., 1998).
Associations between bacteria and algae are mutualistic: bacteria use labile algal EPS as substrate, and algae benefit from inorganic N-fixation and reductive dissolution of Fe by bacteria (Flemming and Wingender, 2001;Srivastava et al., 2018 and refs. therein). Since free-living and attached bacteria are equally correlated to SPM floc size (Zimmermann, 1997), attachment is not a prerequisite for microbially induced flocculation (Figure 1).
Preferential adsorption of dissolved EPS also yields enrichment within SPM (Aufdenkampe et al., 2001;Cao et al., 2011;Lin et al., 2016;Xinran Liu et al., 2013). With increasing microbial activity in spring or during algal blooms, EPS release induces a shift in DOC composition, with an increase of larger biopolymers (> 20 kDa) (Lee et al., 2019); increased MW of proteinaceous substances, and lower MW of aromatic compounds   (Lee et al., 2019).

Pedogenic HS counteract EPS.
Flocculation by dissolved EPS is in competition with stabilization by soil-derived HS (Guo et al., 2003;Guo and Macdonald, 2006) (Figure 1). Pedogenic HS typically constitute the largest NOM fraction in fresh waters, with 40 -80 % of NOM being fulvic acids and < 5 -10 % humic acids. Fulvic compounds exhibit a low MW (0.5 -5 kg mol -1 ) and high maximum charge densities (-6 to -11 meq g -1 ). Humics are larger, with lower charge density, higher hydrophobicity, C:O and C:H ratios, but else similar. (Buffle et al., 1998) The highly negative charge of HS results from many carboxyl-groups, which are deprotonated at environmental pH values. HS are generally thought to electrostatically stabilize colloids against aggregation, as adsorption increases negative charge of mineral surfaces or induces charge reversal of positively charged surfaces (Buffle et al., 1998;Reinhardt and Wilkinson, 2004;Srivastava et al., 2018). Adsorption of low-MW HS is kinetically favorable, but their hydrophilicity eases desorption and HS will be replaced by higher-MW, aromatic, hydrophobic NOM, the adsorption of which is thermodynamically favored (Philippe and Schaumann, 2014). Therefore, the flocculation potential of river water decreases after rain events, when soil-derived HS increase their share compared to EPS (Lee et al., 2017(Lee et al., , 2019. The addition of EPS-like substances to HS-stabilized montmorillonite, in turn, triggers slow aggregation of HS-coated montmorillonite (Wilkinson et al., 1997).

Iron oxides and (oxy)hydroxides (FeOx) are the "cement" in SPM flocs.
Due to their physicochemical properties, FeOx play a crucial role in SPM floc formation (Praetorius et al., 2020), acting as aggregating agent or cementing agent, analogously to their role in soil aggregates (Filella, 2007). FeOx can be of soil origin, reaching rivers with soil aggregates or as crusts on mineral surfaces, or form and transform in-situ through redox-driven dissolution and precipitation reactions (Davison and De Vitre, 2019) (Figure 1). Precipitation initially yields amorphous and poorly crystalline FeOx, like ferrihydrite or lepidocrocite (Guo and Barnard, 2013;Vindedahl et al., 2016). This so-called labile FeOx fraction was found positively correlated to SPM floc size (Markussen et al., 2016), as it exhibits a high specific surface area and reactivity (Davison and De Vitre, 2019;Hirst et al., 2017;Jilbert et al., 2018). Dissolution and reprecipitation, however, can transform labile FeOx into more cryctalline goethite, while hematite requires dry and warm conditions to form and indicates soil origin (Davison and De Vitre, 2019;Guo and Barnard, 2013). With points of zero charge mostly above pH 7 up to 9 (Kosmulski, 2014), bare FeOx surfaces are expected to be positively charged in many rivers (Salminen et al., 2005), favoring association with negatively charged minerals and NOM. Indeed, only 3-8 % of the total Fe in four river water samples were discrete positively charged FeOx colloids (filtration < 1 µm) and most Fe was associated with SPM (1-10 µm) and with colloidal (< 1 µm) HS or aluminosilicates (Hiraide et al., 1988b).

Reactivity of FeOx surfaces favors aggregation and flocculation processes.
Due to (1) the dominance of labile FeOx phases and (2) a high density in (positively charged) surface hydroxyl-groups, most FeOx surfaces are highly reactive. The transformation of amorphous and poorly crystalline FeOx into crystalline phases is limited by stabilization through ligation, and by relatively short residence times in rivers (as compared to soils) (Davison and De Vitre, 2019;Guo and Barnard, 2013). Consequently, labile, reactive FeOx phases with high specific surface area predominate (Guo and Barnard, 2013;Hirst et al., 2017;Vindedahl et al., 2016 and refs. therein). Labile FeOx can be stabilized (Figure 1) by NOM and inorganic ligands, like silicates or phosphates (Baalousha et al., 2019;Cameron and Liss, 1984), or by precipitation or attachment onto phyllosilicates (Arias, 1995;Schwertmann, 1988;Wei et al., 2011). The surface chemistry of FeOx is governed by hydroxyl-groups, exhibiting pH-dependent variable charge (Vindedahl et al., 2016 and refs. therein). Due to their high points of zero charge, FeOx surfaces are a major contributor of positive surface charge in SPM.
Aggregation by HS proceeds slowly and yields small, compact aggregate structures (Baalousha et al., 2008;Buffle et al., 1998;Mylon et al., 2004;Tipping et al., 1981). Slow aggregation driven by FeOx can, thus, explain smaller and more compact SPM aggregate structures observed in winter when HS are the dominant NOM source (Figure 1).

FeOx precipitation and morphology are governed by heterogeneous oxidation
Autochthonous FeOx form when waters rich in dissolved Fe get supersaturated upon hydrochemical transitions (Guo and Barnard, 2013). Generally, abiotic oxidation dominates over biotic or biotically mediated oxidation (Baken et al., 2013;Bullen et al., 2001;Davison and De Vitre, 2019). The latter proceeds in floc microenvironments or sediment redox transitions, where at low pH, biota may outcompete abiotic oxidation (Davison and De Vitre, 2019).
High amounts of FeOx form when (1) anoxic Fe(II)-rich groundwaters or soil porewaters discharge into oxic rivers Baken et al., 2013;Bullen et al., 2001), or (2) waters The dominance of nano-sized FeOx and patchy coatings results from heterogeneous oxidation (Teutsch et al., 2005), which involves adsorption of Fe(II); Subsequent oxidation creates an unstable Fe(III) surface complex which hydrolyzes and precipitates. When anoxic groundwater enters a river, anoxia extends into the hyporheic zone (Baken et al., 2013;Ingri et al., 2006;Teutsch et al., 2005;van der Grift et al., 2014), supplying Fe(II) to the sediments, where it sorbs to minerals and NOM (Figure 3). Similarely, microbial reduction in organic-rich, anoxic sediment layers generates Fe(II), which diffuses and adsorbs to surface sediments. In most rivers and oligotrophic lakes reducing conditions start below the sediment surface, in eutrophic systems or stratified lakes anoxia may extend into the water (Davison and De Vitre, 2019), where suspended minerals and NOM offer further adsorption sites. When the groundwater table drops or surface-sediments get resuspended, adsorbed Fe(II) is oxidized and precipitates (Gerhardt and Schink, 2005). Some oxidation even proceeds under anoxic conditions, as catalyzed by the sorbate, e.g. NOM (Daugherty et al., 2017) or phyllosilicates (Van Groeningen et al., 2020), or by microorganisms (Bryce et al., 2018).
Nucleation of FeOx on clay-attached NOM provides an alternative explanation for FeOx granules (10 -20 nm) seaming clay edges attached via NOM (Poulton and Raiswell, 2005), which the authors attributed to aggregation of NOM-coated FeOx with positively charged clay edges.

Phyllosilicates are a major substrate for floc formation processes.
Being one of the most abundant SPM constituents, phyllosilicates constitute the scaffold for floc formation processes (Figure 1). Five interaction mechanisms can be discerned (Figure 4) in which phyllosilicates play a key role through intrinsic surface heterogeneities, a high affinity for NOM, as a preferential substrate for reactive FeOx coatings, and as substrate for microbial growth. iron oxides/(oxy)hydroxides).

Phyllosilicate surface heterogeneities favor aggregation and EPS-flocculation.
At the edges or surface defects, phyllosilicates carry reactive surface hydroxyl-groups, mostly Si-OH and Al-OH, but also some Mg-OH. Those exhibit pH-dependent, variable surface charge, while unimpaired basal planes exhibit permanent negative charge due to isomorphic substitution of Si 4+ with Al 3+ and Fe 3+ , or Al 3+ with Mg 2+ and Fe 2+ in the crystal lattice (Preocanin et al., 2016).
Zero charge conditions for phyllosilicate edges were reported at pH ≥ 7.5 (for chlorite, kaolinite, talc, muscovite) (Kosmulski, 2001;Yin et al., 2013). A higher density of Mg-OH and Al-OH groups at the edges favors positive edge charge as they are more easily protonated than Si-OH (Xiandong Tournassat et al., 2016;Yin et al., 2013). The intrinsic surface charge heterogeneities of phyllosilicates favor (edge-to-face) aggregation ( Figure 4) especially at lower pH when edge-group protonation is more pronounced.
A close correlation of the organic C content in SPM with Al contents and specific surface area (Poulton and Raiswell, 2005) indicates interactions of NOM with phyllosilicates. Phyllosilicate hydroxyl-groups are prone to adsorb NOM. Negatively charged, low-MW HS coat positive edge groups and stabilize phyllosilicates (Kretzschmar et al., 1998;Tarchitzky et al., 1993), while EPS typically induce flocculation (Labille et al., 2005;Lee et al., 2017;Zhang et al., 2013). Due to their heterogeneous properties, phyllosilicates preferentially interact with similarly diverse EPS.
Surface charge heterogeneity allows electrostatic interactions of anionic NOM groups (e.g. carboxyl-or phosphate) with phyllosilicate edges, and of cationic (amino) groups with phyllosilicate basal planes (Cao et al., 2011;Hedges and Hare, 1987;Lin et al., 2016;Yeasmin et al., 2014). Since the basal planes exhibit increasing hydrophobicity with lower degrees of isomorphic substitution (e.g. montmorillonite < illite < kaolinite < talc), while the edges are usually hydrophilic (Yin, 2012), hydrophobic interactions with NOM occur on basal planes (Henao and Mazeau, 2008). The high specific surface area and cation exchange capacity of phyllosilicates additionally supports weaker NOM interactions, like H-bonding and cation bridging at lower pH and in the presence of divalent cations (mainly Ca 2+ ) (Cao et al., 2011;Philippe and Schaumann, 2014).

Phyllosilicates are a preferential substrate for reactive FeOx coatings.
The visual evidence of frequent FeOx coatings on aluminosilicates mentioned earlier, is supported by positive correlations of reactive Fe with Al contents, with decreasing SPM constituent particle size and with increasing specific surface area (Markussen et al., 2016;Poulton and Raiswell, 2005). Aluminosilicates constitute a preferential substrate for precipitation or attachment of FeOx in patches, and support formation and persistence of labile reactive FeOx.
FeOx patches yield surface charge heterogeneities, and reactivity enhances NOM interactions, favoring incorporation of FeOx-carrying aluminosilicates into SPM.
Ligation by reactive mineral surface groups can catalyze oxidation of adsorbed Fe(II) by water (Géhin et al., 2007;Van Groeningen et al., 2020), allowing for oxidative precipitation of adsorbed Fe(II) even below saturation and under strictly anoxic conditions (Géhin et al., 2007;Mikutta et al., 2009;Van Groeningen et al., 2020). Al-OH or Mg-OH groups at edges or surface defects of phyllosilicates (e.g. montmorillonite) constitute such catalysts (Géhin et al., 2007). Due to a limited availability of these sites, low Fe(II) surface loadings yielded higher oxidized Fe(III) fractions than high loadings (Van Groeningen et al., 2020). Aluminosilicate surface pits filled by FeOx provide visual evidence (Poulton and Raiswell, 2005). Below-saturation precipitation of FeOx in anoxic conditions also occurred on Al-and Fe-oxides (goethite), but not on quartz, and a stronger adsorption of Fe(II) suggests a higher affinity for Al-OH and Fe-OH than Si-OH (Mikutta et al., 2009). As phyllosilicates are way more abundant than metal(hydr)oxides, they will be the relevant substrates for surface-catalyzed FeOx precipitation, but FeOx patches themselves may facilitate further precipitation.
Attachment of FeOx to phyllosilicates behaves similarly, as indicated by ferrihydrite granules sitting in surface defects of montmorillonite after mixing (Yan et al., 2021). Electrostatic interactions favor attachment of FeOx (ferrihydrite) to phyllosilicate basal planes at low to circumneutral pH (Yong and Ohtsubo, 1987). Only high loadings yielded a full coating, else ferrihydrite distributed in patches of particles or aggregates centered around certain kaolinite surface sites (Jones and Saleh, 1987). Again, specific interactions with high affinity sites can be concluded.
As shown for FeOx surface precipitates on kaolinite (Arias, 1995;Wei et al., 2011) as well as for ferrihydrite particles being mixed with different clay minerals (Schwertmann, 1988), phyllosilicates (as opposed to quartz) increase lability, reactivity and specific surface area of FeOx, and inhibit or slow their crystallization. The authors hypothesized that (1) dissolved Si or Al released from phyllosilicates inhibits or retards FeOx crystallization, (2) reactive clay surfaces hamper FeOx crystallization more than silicates (in line with the above), (3) the cementing action between phyllosilicates and amorphous FeOx inhibits dissolution-precipitation reactions and structural rearrangement of FeOx into crystalline forms (Arias, 1995;Schwertmann, 1988;.

Phyllosilicates support microbial nutrition.
Due to their high specific surface areas and ion exchange capacities, phyllosilicates adsorb nutrients and organic substrates, supporting microbial growth (Cuadros, 2017;Henning et al., 2001;Liss et al., 1996). Phyllosilicates offer hydrophilic and hydrophobic surfaces, easing microbe attachment, and microbes exude EPS to attach and access nutrients, fostering typical organomineral aggregates (Cuadros, 2017). 3D-nanotomography imaging of riverine SPM shows complex porous structures of clay minerals and few larger grains being colonized by microorganisms (Wheatland et al., 2017). So-called house-of-cards structures are actively constructed from phyllosilicates, EPS, and some FeOx, constituting the microbial "minimal nutritional sphere" (Lunsdorf et al., 2000). In this context, phyllosilicates are also thought to eliminate harmful waste products by adsorption. Microbes prefer nutrient-rich (e.g. smectite) and easily leachable clays: montmorillonite benefitted microbial activity more than kaolinite, as probably related to their adsorption capacity (Cuadros, 2017 and refs. therein). Moreover, microbes can mediate precipitation of low-crystallinity, Si-Al-Fe-containing clay minerals.
Functional groups on cell walls and EPS (negative carboxyl-, hydroxyl-, phosphoryl-and positive amino groups) thereby provide sorption and nucleation sites. A detailed review of these and other microbe-mineral interactions (e.g. transformation, weathering, dissolution, leaching of clays) is given in Cuadros (2017).

CONCLUSION
In this review we elucidated how the most important physicochemical SPM floc formation processes are governed by three crucial structural-functional SPM components, i.e., EPS, FeOx, and phyllosilicates. These components contribute to an overall heterogeneity in physicochemical surface properties, which favors their association in complex flocs or aggregates. Other materials are rather inert (e.g. quartz), or play similar roles but are less abundant, like Mn-or Aloxides/(oxy)hydroxides.

Flocculating EPS counteract stabilizing HS.
EPS, as opposed to HS, promote heterogeneity in physicochemical surface properties upon adsorption. They comprise substances spanning a wide range of molecular weights, various degrees of rigidity and hydrophobicity, and diverse functional groups and charge densities.

Heterogeneity favors EPS-induced flocculation through
• (multivalent-ion assisted) bridging by large rigid EPS, • charge heterogeneities within coatings through fractionated sorption or patchy coatings, • charge neutralization allowing predominance of weak interaction forces such as van der Waals and hydrophobic forces.
Highly degraded HS are more homogeneous regarding flocculation-relevant properties. They mainly comprise low-MW fulvic and to a lesser extent high-MW humic acids, exhibiting high densities in negatively charged carboxyl-groups, which yield homogeneous, highly negative coatings that electrostatically stabilize minerals against flocculation. Due to their antagonistic roles flocculation or stabilization depend on the relative concentrations of HS and EPS.

Phyllosilicates constitute the base for SPM floc formation.
Phyllosilicates dominate the mineral mass fraction in SPM. Due to their heterogeneous surface properties, they provide the base for FeOx-and EPS-enhanced floc formation processes. Both, Fe, and organic C contents in SPM are closely correlated to phyllosilicates. Phyllosilicates exhibit intrinsic surface charge heterogeneities, with structural permanent negative charge on the basal planes and pH-dependent variable charge at the edges and surface defects. Edges are hydrophilic, while basal planes are more hydrophobic, and edges as well as surface pits exhibit increased densities of reactive hydroxyl groups. These properties allow phyllosilicates • to undergo electrostatic edge-to-face aggregation, • to interact with NOM, preferentially EPS, and • to act as preferential substrate for FeOx attachment or surface-precipitation.

FeOx increase surface charge heterogeneities and reactivity of phyllosilicates.
Interactions between FeOx and phyllosilicates increase phyllosilicate surface charge heterogeneity and FeOx reactivity, favoring flocculation. Attachment or precipitation of FeOx yields patchy coatings on phyllosilicates, since dissolved Fe as well as FeOx particles are preferentially ligated by surface-pit or edge hydroxyl-groups. Additionally, bare FeOx can electrostatically interact with phyllosilicate basal planes, due to mostly positive surface charge at circumneutral pH. Association with phyllosilicates stabilizes labile FeOx phases against crystallization, keeping them reactive and their specific surface area high. Consequently, FeOx-carrying phyllosilicates enhance flocculation • through surface charge heterogeneities arising from patchy distributions of FeOx precipitates or granules; • through a phyllosilicate-mediated increased reactivity of FeOx, favoring interactions with NOM and mineral surfaces.
• Negatively charged and hydrophobic phyllosilicate basal planes preferentially adsorb cationic amino groups (on basic amino acids) and hydrophobic parts (e.g. acyl-groups).
• FeOx and clay edges strongly interact with anionic carboxyl-groups (on polysaccharides, acidic amino acids, or HS) and phosphate groups (on phospholipids, phosphorylated proteins, or nucleic acids).
Preferential adsorption of EPS on FeOx and phyllosilicates supports flocculation mechanisms involving EPS. At high concentrations of HS, however, the affinity of HS carboxyl-groups for hydroxyl-groups on FeOx and phyllosilicate edges yields full coating and electrostatic stabilization. Due to their high hydroxyl-group density, FeOx can still allow for some degree of flocculation at elevated HS concentrations.
The high affinity between NOM and Fe can induce co-precipitation, whereby NOM-type and concentration impact morphology and reactivity of FeOx precipitates. Fully Fe-coated EPS yield elongated fibrous FeOx, while at lower Fe:EPS ratios nano-sized FeOx granules nucleate on EPS.
HS promote globular co-precipitates with negative surface charge due to the high density of carboxyl-groups in HS.
Phyllosilicate-FeOx-NOM associations are a favorable substrate for microbial colonization. They pool essential inorganic nutrients and organic substrates, as their high specific surface area and cation exchange capacity favor adsorption. Microbial colonization triggers flocculation through EPS release at different stages. Adhesive EPS aid attachment, attached microbes exude ligands or enzymatic EPS to access nutrients, and when the substrate becomes refractory, starvation induces further EPS release. Additionally, dissolved EPS preferentially interact with FeOx and phyllosilicates and flocculate new phyllosilicate-FeOx-NOM associations that provide fresh substrate for microbial colonization.

Flocculation dynamics are governed by EPS and FeOx concentrations.
Large, fluffy flocs form at times of high bioproduction when EPS-release is high. Their loose structure is supported by high contents of fibril-or network-like organic EPS, and hosts bacteria and microalgae. At low temperatures or after rain events high fractions of soil-leached HS decrease the flocculation potential of river waters. Small, compact, aggregate-like SPM, inhabited by few bacteria, and single mineral grains are dominant. FeOx may induce slow formation of compact aggregates despite elevated HS concentrations. The content of labile, reactive FeOx is positively related to SPM floc size, as they constitute aggregating and cementing agents as well as an essential Fe source to microbes. Authigenic FeOx formation also exhibits seasonal fluctuations depending on the dominant Fe source. At (winter) baseflow higher shares of Fe(II)-rich anoxic groundwater may increase precipitation, while soil-derived Fe-NOM complexes increase after rain events.
SPM is an important part of many biogeochemical cycles and provides multiple ecosystem services. It functions as habitat and feed, as bioreactor involved in element cycling impacting aquatic food chains up to climate change, and as transport vehicle for nutrients and pollutants affecting freshwaters, sediments, inundation zones, and the ocean. Analyzing the scientific wealth of more than 220 publications from several decades of research and various scientific disciplines it could be shown that freshwater floc formation processes and structural dynamics involve three