Discussion
Here we used model soils to characterize both clay minerals and SOM for understanding the mechanisms controlling the formation efficiency of mineral-associated SOM during litter decomposition. We demonstrated interactive effects of clay mineral and litter types, rather than litter type alone, on the chemical composition (Fig. 1) and formation efficiency (Fig. 4) of SOM. Different clay minerals discriminatively protected litter and microbially-derived residues due to differences in mineral-organic association mechanism (Fig. 5) and mineral-protection strength (Fig. 6). Since the mineral-protection strength was higher for surface adsorption by vermiculite than for pore entrapment within domains of kaolinite or illite, vermiculite discriminatively protected more litter-derived labile compounds and fungal residues and had a higher SOM formation efficiency than kaolinite and illite. The significance of mineral-discriminative protection and its control over diversity of organic compounds in SOM has been recently reported for long-term preservation/stabilization of SOM12. The discriminative protection oflitter and microbial residues by different clay mineral types explains why SOM structures often look similar for soils with similar soil mineralogy, but different for soils with contrasting soil mineralogy in different climate zones32-35.
We found that clay minerals were associated with both labile (i.e. O–alkyls and anomerics) and recalcitrant (aromatics and aromatic C–O) litter residues and that litter-residues were dominant over microbial residues in mineral-associated SOM (Figs. 2 and 3). These findings suggest that litter residues, regardless of their recalcitrance, could be associated with clay minerals as very fine SOM observed using transmission electron microscopy36. However, these findings do not support the hypothesis that mineral-associated SOM was derived only from labile litter compounds10-11 or microbially processed products26. The discrepancy arises likely as the previous studies allowed only labile substrates into mineral phases through leaching during litter decomposition above the ground10, preferentially labelled labile compounds in litter to trace SOM formation from litter decomposition in soil as indicated by a low 13C abundance (4%)11 or used only labile substrates26.
We demonstrated that more labile litter residues and fungal residues (Figs. 2 and 3) were better protected through surface adsorption by vermiculite than through pore entrapment within domains of kaolinite and illite, irrespective of litter types. Several studies have also demonstrated a shift toward retention of more fungal than bacterial residues in model soils consisting of vermiculite when compared to illite25 or in natural soils dominant with vermiculite37. We attributed this phenomenon to higher relative abundance of fungi in the vermiculitic soil than in other soils (Fig. 3) and higher recalcitrance of fungal residues compared to bacterial residues, as suggested by previous study38. Bacterial residues can be decomposed by fungi for growth39. So bacterial residues were not protected when they were exposed on vermiculitic surfaces but were better protected when they were isolated in pores within domains of illite and kaolinite.
We provided the first model to describe the feedback effects of mineral-organic association on litter decomposition within mineral matrices and to quantify mineral-protection strength. This novel model consisted of two separate and interactive pools, which is fundamentally different from the conventional SOM model, often consisting of two discrete pools40-41. We provide a simple and reliable approach to quantify mineral-protection strength for specific mineral types or soils and to understand some physico-chemical and physical protection processes of SOM. The modeled mineral-protection strength explained well the variance of the measured SOM formation efficiency of the model soils mixed with either litter types (Fig. 6). Although several cutting-edge models provide a framework to describe the role of mineral protection in controlling SOM dynamics and stabilization13,42, no models are available to describe the control of mineral composition over SOM formation efficiency. In addition, those SOM models have not yet incorporated parameters that consider mineral-protection strength.
We were able to predict the mineral-protection strength of the carbon-free natural soil material based on the mineral-protection strengths of the pure clay minerals and their relative abundances in the soil. The vermiculite, regardless of its origin, had a much larger specific surface area than the illite (Supplementary Table 1). The natural soil material was taken at a depth below 2 m and is not highly weathered, so its illite would have a relatively large particle size and then a small specific surface area compared to the pure illite (Supplementary Table 1). Contrasting X-ray diffractogram changes in illite from the same subsoil as ours were observed in a previous study43, suggesting that SOM was adsorbed on the surfaces of < 2 μm illite, but entrapped within pores of 2-5 μm illite domains. However, natural soil minerals, particularly in surface soils, may differ notably from pure minerals in their particle size and surface properties, which will inevitably impact mineral-organic association mechanism and strength, as shown for illite in our study. In addition, our model soils were initially C-free. However, SOM may be protected through adsorption to existing SOM, rather than through mineral associations in soils with high organic carbon contents44. Therefore, further studies are needed using more mineral and litter types or soils with different initial C contents and running for longer time scales to better understand mineral protection of SOM. With better knowledge about soil mineralogy and mineral-organic association, our novel model can likely be incorporated into next generation soil and terrestrial C cycling models to reliably predict and compare SOM dynamics among different soil types.