Combined size and density fractionation (CSDF) is a method to physically separate soil into fractions differing in texture (particle size) and mineralogy (density). The purpose is to isolate fractions with different reactivities towards soil organic matter (SOM), in order to better understand organo-mineral interactions and SOM dynamics.
Combined size and density fractionation (CSDF) is a method used to physically separate soil into fractions differing in particle size and mineralogy. CSDF relies on sequential density separation and sedimentation steps to isolate (1) the free light fraction (uncomplexed organic matter), (2) the occluded light fraction (uncomplexed organic matter trapped in soil aggregates) and (3) a variable number of heavy fractions (soil minerals and their associated organic matter) differing in composition. Provided that the parameters of the CSDF (dispersion energy, density cut-offs, sedimentation time) are properly selected, the method yields heavy fractions of relatively homogeneous mineral composition. Each of these fractions is expected to have a different complexing ability towards organic matter, rendering this a useful method to isolate and study the nature of organo-mineral interactions. Combining density and particle size separation brings an improved resolution compared to simple size or density fractionation methods, allowing the separation of heavy components according to both mineralogy and size (related to surface area) criteria. As is the case for all physical fractionation methods, it may be considered as less disruptive or aggressive than chemically-based extraction methods. However, CSDF is a time-consuming method and furthermore, the quantity of material obtained in some fractions can be limiting for subsequent analysis. Following CSDF, the fractions may be analyzed for mineralogical composition, soil organic carbon concentration and organic matter chemistry. The method provides quantitative information about organic carbon distribution within a soil sample and brings light to the sorptive capacity of the different, naturally-occurring mineral phases, thus providing mechanistic information about the preferential nature of organo-mineral interactions in soils (i.e., which minerals, what type of organic matter).
Soil is a complex system which contains elements of geological and biological origin. The study of their inter-relation is a cornerstone of our understanding of ecosystem function1. In particular, organo-mineral interactions are thought to play a key role in soil organic matter (SOM) dynamics2. Unravelling SOM dynamics is presently a very active research area for several reasons. A soil with high SOM stocks will tend to show good intrinsic fertility and may also constitute an environmentally valuable carbon sequestration opportunity3,4.
Organic matter in soil is highly heterogeneous, with some components turning over in the space within a few hours while others may persist for thousands of years5. The determinants of this heterogeneity remain a controversial topic, but association with the mineral matrix is thought to be particularly important6,7, especially for subsoil horizons8. As a result, mineral phases known to closely associate with organic components are receiving increasing interest9,10,11.
Soils contain a wide range of minerals with qualitatively and quantitatively varying sorptive potential towards SOM. Minerals with large specific surface areas and/or highly reactive surfaces have been shown to have a high sorption capacity for organic compounds4,12. In soils, secondary minerals such as high-activity phyllosilicates (e.g., smectites), iron oxyhydroxides and poorly crystalline aluminosilicates have all been shown to engage significantly in the sorptive preservation of some organic compounds13,14,15,16,17. Separating soil into fractions differing in mineralogy could thus help isolate organic matter pools with relative functional homogeneity.
The aim of this paper is to present a methodology to isolate organo-mineral complexes according to composition, which then facilitates the study of their properties. The method combines size and density fractionation to physically separate bulk soil into a sequence of fractions of different composition. Combined size and density fractionation (CSDF) integrates two effective physical fractionation approaches (particle size separation and density separation). The combination of these two approaches brings improved resolution to our understanding of organo-mineral associations in soil.
There are many different approaches (chemical, physical and / or biochemical) that can be used to specify fractions in a bulk soil sample18,19. Simple density fractionation is a physical separation which has been widely used by soil scientists to study SOM dynamics (see for instance Grunwald et al., 2017 and references therein)20. In its classical form, simple density fractionation separates materials lighter than a given cutoff (generally 1.6 to 1.85 g·cm-3) – the light fraction (LF) from heavier materials – the heavy fraction (hF). The LF is sometimes further split into free light fraction (fLF) and occluded light fraction (oLF)21.
In many soils, the largest SOM pool is found in the hF22. SOM in the hF is generally thought to be more stable than that in the LF23, yet it has been shown to retain a high compositional and probably, functional heterogeneity18. This points to the need to further separate the hF into more homogeneous subfractions, with the view of isolating pools of SOM with distinct biogeochemical properties (such as residence time or functionality). Sequential density fractionation, as described by Sollins et al. (2009)24, has indeed proved to be a successful method; yet a separation done solely on the basis of density runs the risk of overlooking differences arising from variation in grain size and thus specific surface area. For instance, kaolinite has approximately the same density as quartz but may be separated on the basis of its size mode (Table 1). CSDF includes consideration of grain size and improves the resolution of the fractionation.
SOM fractionation based on physical, chemical or biochemical properties has a long history. Physical methods such as CSDF are based on physical attributes of soil components, such as size (of particles or aggregates) or density. Chemical methods include selective extractions of specific compounds or classes of compounds, as well as chemical oxidation. Biochemical methods rely on microbial oxidation under various experimental conditions. Chemical and biochemical methods are based on different principles and have different objectives compared to physical methods but are nevertheless briefly reviewed below.
The alkaline extraction (with sodium hydroxide for example) ranks among the earliest methods used to chemically isolate the organic component of soils6. Examples of more modern, chemical methods for SOM fractionation include i) alkaline extraction with Na-pyrophosphate aimed at isolating SOM bound to minerals; ii) acid hydrolysis (HCl) aimed at quantifying old, persistent SOM; and iii) selective oxidation of SOM with chemical agents aimed at attacking free or labile SOM2. While these methods can be useful to gain insight into functionally different organic matter pool, they suffer from several limitations. First, the extractions can be imperfect or incomplete. For example, the classical alkaline method extracts only 50-70% of soil organic carbon (SOC)6. Second, fractionation products may not be representative of the SOM found in situ and may be difficult to categorize5. Third, these chemical methods only offer limited insight into the organo-mineral interaction since many of them do not preserve the original association between organics and minerals.
Biochemical extraction including incubations experiments are used mainly to study labile and reactive SOM (see Strosser32 for a review of biochemical methods). Incubation experiments can be thought of as a measure of biochemical oxygen demand and is intuitively well-suited to the determination of bioavailable organic substrates. However, the need for long incubation times in conditions that differ from the field (temperature, humidity, physical disturbance, absence of new inputs) makes the extrapolation to in-situ SOM dynamics delicate.
Compared to chemical or biochemical methods which are believed to be either transformative or destructive, physical fractionation techniques can be considered as more preservative22 (with the important exception of soluble organic compounds, which are lost during the procedure). At their best, physical soil fractions can be thought of as a 'snapshot' of solid-phase soil components as present in the field and could thus relate more directly to SOM dynamics in situ33. Moreover, the non-destructive nature of the technique means that the fractions can be subsequently characterized using a variety of analyses or further fractionated according to chemical or biochemical methods.
Physical fractionation of soils is not a recent idea. Scientific literature about physical separation techniques dates back to the mid-20 century. Applications of density fractionation were reported as early as 196534, 35. During the same period and in the following decades, publications about the dynamics of SOM and its interaction with minerals were already becoming widespread amongst soil scientists36,37,38,39.
Separation based on density, aggregate size or particle size are the most common physical separation methods used currently. One of the main challenges of physical separation is the isolation of homogenous functional SOM pools, as defined by turn-over rate, size or other indicator of function. Combining separation methods or criteria, as in CSDF, may help bring functional resolution to soil fractions; indeed, these methods seem to be used more and more in combination18,40,41,42,43. By combining sequential density separation, able to yield fractions with different organic matter content and mineralogical composition, with size separation, which accounts for differences attributable to specific surface area, CSDF holds the promise of yielding insight into the diversity and function of organo-mineral associations in soil.
CSDF aims to physically fractionate bulk soil samples into fractions of relative mineralogical and textural homogeneity. The density and particle size cut-offs, as well as the dispersion energy used here have been selected based on our soil type, but these parameters may be adapted depending on the samples to be fractioned and the purpose of the study. In this example, we have chosen to use one dispersion step, two density and one size cut-offs, resulting in the separation of the bulk soil into 6 fractions (Table 2). Figure 1 gives a conceptual overview of the method. The materials being fractioned here are tropical soils, but the method can be applied to any soil type as well as sediments. CSDF is generally used as a preparatory step before further analyses, even though the distribution of materials among fractions can be very informative in and of itself. When applied to soils, CSDF yields fractions differing in (1) mineral composition (mineralogy and texture) and (2) SOM concentration and composition.
1. Sample Preparation
2. Dense Solution Preparation
3. Light Fraction (LF) Separation
NOTE: The LFs are isolated by flotation on a solution with the density of 1.62 g·cm-3 prior to ultrasonic dispersion (fLF) and following dispersion at 280 J·mL-1 (oLF).
4. Separation of the hFs According to Particle Size
NOTE: The next step is to fractionate the residue from step 3 (the hF) according to particle size. The size cut-off here is set at 8 µm and produces a clay + fine silt fraction (< 8 µm) and a coarse silt + sand fraction (> 8 µm). The grouping of clay and fine silt in the finer fraction reflects the documented affinity of both clay and fine silt for soil organic matter sorption33,48. Particle size separation here is done by sedimentation (based on Stokes’ law49). For cut-offs at 50 µm or larger, the separation can be simply effected by wet sieving without risking too much abrasion or disruption of organo-mineral complexes.
5. Separation of the Two Size Fractions According to Density
NOTE: Density fractionation is applied here to both particle size fractions. The objective was to separate silicates from oxides. We thus opted for a density cut-off of at 2.78 g·cm-3. Optional, additional separations may be performed. For instance, a 2.4 g·cm-3 solution would allow for the separation of high-activity clays from kaolinite and primary silicates. Keep in mind than high organic matter loading will decrease the theoretical density of mineral particles43.
6. Recycling of SPT
NOTE: The SPT solution can be recycled in view of re-use by passing it through a column containing activated charcoal and a cation exchange resin50. Activated charcoal retains organics while the sodium-saturated cation exchange resin removes calcium and other cations from the solution and replaces them with sodium. We digest the SPT in hydrogen peroxide prior to passing it through the column to ensure quantitative removal of dissolved organics.
Samples are tropical soils originating from the Albertine rift valley in Uganda. They consist of profiles from 3 cultivated sites receiving no external inputs such as fertilizer or phytosanitary products. These samples were chosen to represent a large spectrum of mineralogy. Preliminary analyses showed that site 1 was least weathered and richest in primary silicates (feldspars). Site 2 showed signs of more advanced weathering with a high content of secondary clays such as kaolinite and a relative enrichment in quartz. Site 3 was highly weathered with signs of desilicification and residual accumulation of iron oxides and oxyhydroxides. Site 3 contained an extremely high total iron concentration (34%, expressed as Fe2O3 oxide) due to the presence of plinthic material (iron-rich induration51,52). For each profile, two horizons were sampled: topsoil (A) and subsoil (B). CSDF was performed on these 6 samples in four replicates.
The first step in evaluating the effectiveness of the fractionation procedure is to look at recovery rates, (i.e., whether the initial material is quantitatively retrieved at the end of the experiment). We assessed recovery rates based on the whole soil and SOC contents.
Overall, whole soil recovery rates were considered to be very good, with 16 out of 20 replicates having recovery rates of more than 90% and 4 replicates showing recovery rates between 85-90% (Table 4). The cause of incomplete recovery was most likely a loss of dissolved and colloidal substances during rinsing. Two replicates showed a slight gain of mass (in the order of 1%) that could possibly be caused by SPT residues or weighing imprecisions. It should be noted that reasonable mass imbalances (<10-15%) are common and do not generally compromise the validity of the fractionation.
SOC recovery was generally within the range of reports from other studies53,54 and remarkably constant considering the large variation in initial SOC content (Table 5). Most samples showed a SOC recovery rate of 80-85%. Losses are attributable to the flushing of soluble C, which is an unavoidable feature of the method; however, soluble organic C can easily be quantified using a separate extraction in water, salt or chemical dispersant55. A small loss of dispersed organic colloids during fractionation is also likely. One site showed a slight gain of carbon which can probably be attributed to analytical error, as the absolute value for the difference was small (3 mg).
The method reproducibility may be verified by analyzing dispersion between replicates. We assessed the standard error of the mean (SEM) as well as the coefficient of variation (CV) of fraction mass between replicates.
Standard errors of the mean were small (Table 6), being generally 1 to 2 orders of magnitude smaller than mean values. This shows that working in 4 replicates allowed us to reliably estimate the central tendency for distribution of materials between fractions.
Coefficients of variation ranged from 2 to 70% (Table 7). All CVs greater than 35% occurred for fractions with small amounts of material (<0.25 g). These high values are simply due to the fact that division by a small mean yields high CVs. A few hF1 and hF3 fractions (coarse and fine silicates) showed relatively high CVs, between 20-35%, yet comprised large amounts of materials (1 – 4 g). This may reflect the high potential for human errors during several sensitive steps (i.e., (1) the separation of floating and suspended materials from the pellet in dense solutions, (2) sedimentation to isolate particle size fractions, (3) sample rinsing and recovery). This result confirms the need to work in several replicates to obtain robust results. It is also recommended that the whole process be handled by the same person, who becomes an expert in performing manipulations in a reproducible way and will acutely notice any details that might be different from previous batches.
The distribution of material mass between fractions showed strong differences between sites, as could be expected given the initial differences in mineralogy (Figure 2). At site 1, dominated by primary silicates such as quartz and feldspars, most of the material was recovered in hF1 (designed to concentrate coarse silicates). Site 2 showed a greater percentage of phyllosilicates (mostly kaolinite) during mineralogical analysis; accordingly, hF3 (designed to concentrate fine silicates) had more materials than at site 1. Finally, site 3 was the richest in oxides and also showed the greatest amount of material in hF2, designed to concentrate coarse oxides. This indicates that the method was successful in physically fractionating bulk samples into their main mineralogical components.
The amount of materials recovered in coarse (hF1 + hF2) versus fine (hF3 + hF4) fractions was compared to what was expected based on particle-size distribution determined by laser granulometry (Table 8). Agreement was good (< 10%) for three samples. The three other samples showed an excess of material in the order of 20% in the coarse fractions. The large amount of oxides in the soils (particularly at site 3) may be partially responsible for this difference. Oxide grains have a greater density compared to silicates and will sediment faster. Other factors could include incomplete dispersion or partial flocculation of samples during sedimentation, since we did not use chemical dispersion, and removal of some fine materials in the light fractions (fLF and oLF). Finally, laser granulometry is based on volume estimates under the assumption of particle sphericity, while sedimentation yields mass-based estimates. These contrasting principles of measurement are likely to give somewhat diverging results.
CSDF isolates fractions of relative mineralogical homogeneity and their associated organic matter (organo-mineral complexes). It is most useful as a preparation step prior to subsequent geochemical, biochemical and mineralogical analyses. Arguably, the most powerful experiments will aim to characterize both the organic matter and the minerals in each fraction. This will provide direct evidence for the nature of organo-mineral association in soils.
Sample analyses could include determination of SOM quantity (e.g., elemental analysis of organic C and total nitrogen) and quality (e.g., differential Fourier-transform infrared spectroscopy, pyrolysis gas chromatography mass spectrometry, or thermal analysis such as Rock-Eval pyrolysis56,57,58). When looking at the mineral partner, useful analyses could include particle size analysis, quantification of reactive aluminum and iron phases59, X-ray diffraction (XRD) on powdered samples for bulk mineralogy or on oriented slides for clay mineralogy60.
Techniques able to yield simultaneous information on both the organic and inorganic components could be of particular interest. Elemental mapping by secondary ion mass spectrometry (SIMS) or electron microscopy coupled with X-ray microanalysis (WDS or EDS, wavelength or energy dispersive X-ray spectroscopy) can allow for the co-localization of C, N and elements associated with reactive mineral phases such as Fe, Al, Mn or Ca. X-ray photoelectron spectroscopy (XPS) can reveal the chemical composition of SOM and the surface elemental composition of each fraction61.
Figure 1: Flowchart. Fractionation steps and cut-offs used in the method are presented here. Please click here to view a larger version of this figure.
Figure 2: Distribution of materials between fractions as a function of soil mineralogy for two horizons (A and B) at three sites. (A) Bar chart showing the repartition of materials between fractions. Bars represent the mean and error bars represent the standard error of the mean of four replicates. For each sample, the five bars sum to 100%. (B) Bulk sample mineralogy as assessed by powder X-ray diffraction. Please click here to view a larger version of this figure.
Soil component | Class | Density [g cm-3] | Size distribution | Source |
Organic | Organic matter | 1.00-1.50 | Variable | Multiple sources. See Rühlmann et al. (2006)25 for a review |
Imogolite | Poorly crystalline phase | 1.70-2.33 | Clay | Wada and Wada (1977)26 |
Allophane | Poorly crystalline phase | 1.84-2.35 | Clay | Wada and Wada (1977)26 ; Wilson (2013)27 |
Opal | Poorly crystalline phase | 1.90-2.30 | Variable | Hudson Institute of Mineralogy (2017)28 |
Montmorillonite | Clay mineral | 2.30-2.35 | Clay | Wada and Wada (1977)26; Wilson (2013)27 |
Vermiculite | Clay mineral | 2.30-2.50 | Clay | Wilson (2013)27 |
Gibbsite | Al oxide | 2.34-2.42 | Variable | Hudson Institute of Mineralogy (2017)28 |
K-feldspars | Primary Si-rich silicates | 2.54-2.57 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Albite | Primary Si-rich silicates | 2.60-2.62 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Kaolinite | Clay mineral | 2.60-2.68 | Clay and silt | Klein and Philpotts (2017)29; Wilson (2013)27 |
Quartz | Primary Si-rich silicates | 2.63-2.66 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Calcite | Carbonate | 2.71 | Variable | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Anorthite | Primary Si-rich silicates | 2.74-2.76 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Illite | Fine-grained mica | 2.75-2.80 | Clay | Wilson (2013)27; Hudson Institute of Mineralogy (2017)28 |
Muscovite | Mica | 2.77-2.88 | Variable | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Biotite | Mica | 2.78-3.20 | Variable | Klein and Philpotts (2017)29; Skopp (2000)30 |
Dolomite | Carbonate | 2.84-2.86 | Variable | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Amphiboles | Primary ferromagnesian silicates | 3.00-3.40 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Pyroxenes | Primary ferromagnesian silicates | 3.20-3.60 | Silt and sand | Klein and Philpotts (2017)29; Hudson Institute of Mineralogy (2017)28 |
Goethite | Fe oxide | 3.30-4.37 | Variable | Hiemstra and van Riemskijk (2009)31; Klein and Philpotts (2017)29 |
Ferrihydrite | Fe oxide | 3.50-3.90 | Clay | Hiemstra and van Riemskijk (2009)31 |
Lepidocrocite | Fe oxide | 4.00-4.13 | Variable | Hiemstra and van Riemskijk (2009)31; Hudson Institute of Mineralogy (2017)28 |
Hematite | Fe oxide | 4.80-5.30 | Variable | Klein and Philpotts (2017)29; Skopp (2000)30 |
Table 1: Main soil components in order of increasing density. Their prevalence in the main textural classes (clay fraction, 0-2 µm; silt fraction, 2-50 µm; sand fraction, 50-2000 µm) for moderately weathered soils is also indicated.
Name of fraction | Abbreviation | Cut-offs |
Free organics | fLF | < 1.62 g cm-3 (before sonication) |
Occluded organics | oLF | < 1.62 g cm-3 (after sonication) |
Coarse silicates | hF1 | > 8 µm, 1.62 g cm-3< hF1 < 2.78 g cm-3 |
Coarse oxides | hF2 | > 8 µm, > 2.78 g cm-3 |
Fine silicates | hF3 | < 8 µm, 1.62 g cm-3< hF3 < 2.78 g cm-3 |
Fine oxides | hF4 | < 8 µm, > 2.78 g cm-3 |
Table 2: List of fractions resulting from CSDF using one sonication, two density and one size separation steps.
Solution volume [mL] | Desired density [g cm-3] | Mass SPT [g] | Volume H2O [mL] |
1000 | 1.6 | 741 | 856 |
1000 | 1.8 | 990 | 810 |
1000 | 2 | 1250 | 750 |
1000 | 2.2 | 1490 | 715 |
1000 | 2.4 | 1803 | 595 |
1000 | 2.6 | 2052 | 545 |
1000 | 2.8 | 2297 | 504 |
1000 | 3 | 2552 | 450 |
Table 3: Guide to the preparation of common SPT solutions.
Site | Horizon | Replicate | Starting mass [g] |
Final mass [g] |
Difference [g] |
Difference [%] |
1 | A | 1 | 10.110 | 9.531 | 0.579 | 5.73 |
2 | 10.057 | 9.354 | 0.703 | 6.99 | ||
3 | 10.010 | 8.589 | 1.421 | 14.19 | ||
4 | 10.043 | 10.197 | -0.154 | -1.53 | ||
B | 1 | 10.054 | 9.891 | 0.163 | 1.62 | |
2 | 10.069 | 9.746 | 0.323 | 3.21 | ||
3 | 10.058 | 9.699 | 0.359 | 3.57 | ||
4 | 10.059 | 9.782 | 0.277 | 2.76 | ||
2 | A | 1 | 10.130 | 9.252 | 0.878 | 8.67 |
2 | 10.182 | 9.246 | 0.936 | 9.20 | ||
3 | 10.053 | 9.372 | 0.681 | 6.77 | ||
4 | 10.031 | 9.577 | 0.454 | 4.53 | ||
B | 1 | 10.123 | 8.824 | 1.299 | 12.83 | |
2 | 10.052 | 8.938 | 1.114 | 11.08 | ||
3 | 10.029 | 9.006 | 1.023 | 10.20 | ||
4 | 10.086 | 9.118 | 0.968 | 9.60 | ||
3 | A | 1 | 10.020 | 9.187 | 0.833 | 8.32 |
2 | 10.060 | 9.139 | 0.921 | 9.15 | ||
3 | 10.069 | 9.386 | 0.683 | 6.79 | ||
4 | 10.049 | 9.638 | 0.411 | 4.09 | ||
B | 1 | 10.071 | 9.207 | 0.864 | 8.58 | |
2 | 10.065 | 9.314 | 0.751 | 7.46 | ||
3 | 10.155 | 10.241 | -0.086 | -0.85 | ||
4 | 10.046 | 9.549 | 0.497 | 4.95 |
Table 4: Recovery rate of whole soil, showing the starting mass at the beginning of the fractionation procedure and final mass calculated as the sum of all fractions. Differences are expressed as % starting mass.
Site | Horizon | Starting SOC mass | Final SOC mass | Difference | Difference |
[g] | [g] | [g] | [%] | ||
1 | A | 0.50 | 0.41 | 0.09 | 18.07 |
B | 0.026 | 0.029 | -0.003 | -10.63 | |
2 | A | 0.34 | 0.27 | 0.07 | 20.19 |
B | 0.07 | 0.06 | 0.01 | 12.33 | |
3 | A | 1.08 | 0.94 | 0.14 | 12.56 |
B | 0.31 | 0.27 | 0.05 | 14.51 |
Table 5: Recovery rate of soil organic carbon. Initial SOC content was calculated as the product of organic C concentration measured by elemental analysis and initial sample mass. Final SOC content was calculated as the product of organic C concentration and each fraction mass, summed for all fractions. Differences are expressed as % starting mass.
Site | Horizon | fLF 1.62 g cm-3 |
oLF 1.62 g cm-3 |
hF1 > 8 µm < 2.78 g cm-3 |
hF2 > 8 µm > 2.78 g cm-3 |
hF3 < 8 µm < 2.78 g cm-3 |
hF4 < 8 µm > 2.78 g cm-3 |
1 | A | 0.0794 ± 0.0052 | 0.1093 ± 0.0076 | 5.2188 ± 0.3079 | 0.3925 ± 0.0416 | 3.3699 ± 0.1504 | 0.2478 ± 0.0689 |
B | 0.0044 ± 0.0005 | 0.0074 ± 0.0011 | 8.4351 ± 0.16 | 0.2569 ± 0.0301 | 0.9528 ± 0.1013 | 0.1226 ± 0.0124 | |
2 | A | 0.066 ± 0.011 | 0.1353 ± 0.0152 | 5.722 ± 0.1033 | 0.2575 ± 0.008 | 3.0761 ± 0.1464 | 0.1047 ± 0.0364 |
B | 0.0024 ± 0.0002 | 0.0165 ± 0.0022 | 4.5416 ± 0.0387 | 0.3082 ± 0.0072 | 4.0005 ± 0.0547 | 0.1025 ± 0.0268 | |
3 | A | 0.2107 ± 0.0099 | 0.1489 ± 0.0223 | 3.8507 ± 0.6801 | 1.762 ± 0.0923 | 3.2862 ± 0.4892 | 0.0792 ± 0.0165 |
B | 0.0305 ± 0.0035 | 0.0433 ± 0.0065 | 2.5929 ± 0.376 | 4.1277 ± 0.1025 | 2.6909 ± 0.13 | 0.0927 ± 0.0087 |
Table 6: Mean value and SEM for fraction mass (g). Each cell represents a mean of 4 replicates.
Site | Horizon | fLF < 1.62 g cm-3 |
oLF < 1.62 g cm-3 |
hF1 > 8 µm < 2.78 g cm-3 |
hF2 > 8 µm > 2.78 g cm-3 |
hF3 < 8 µm < 2.78 g cm-3 |
hF4 < 8 µm > 2.78 g cm-3 |
1 | A | 0.13 | 0.14 | 0.12 | 0.21 | 0.09 | 0.56 |
B | 0.22 | 0.29 | 0.04 | 0.23 | 0.21 | 0.20 | |
2 | A | 0.33 | 0.22 | 0.04 | 0.06 | 0.10 | 0.70 |
B | 0.17 | 0.26 | 0.02 | 0.05 | 0.03 | 0.52 | |
3 | A | 0.09 | 0.30 | 0.35 | 0.10 | 0.30 | 0.42 |
B | 0.23 | 0.30 | 0.29 | 0.05 | 0.10 | 0.19 |
Table 7: Coefficients of variation of fraction mass for 4 replicates.
Site | Horizon | Texture < 8 µm % |
Texture > 8 µm % |
Materials in fine fractions (hF3 + hF4) % |
Materials in coarse fractions (hF1 + hF2) % |
Coarse fraction 'excess' % |
1 | A | 48 | 52 | 39 | 61 | 9 |
B | 31 | 69 | 11 | 89 | 20 | |
2 | A | 43 | 57 | 35 | 65 | 8 |
B | 46 | 54 | 46 | 54 | 1 | |
3 | A | 58 | 42 | 37 | 63 | 21 |
B | 52 | 48 | 29 | 71 | 23 |
Table 8: Comparison between soil's texture determined by laser granulometry and the distribution of particles in size fractions. The last column shows the excess of materials in the coarse fractions compared to what was expected based on textural analysis.
The success of CSDF experiments hinges on the selection of appropriate parameters for the method, so that fractions of relatively homogeneous composition may be isolated. Key considerations in the selection of fractionation parameters are discussed below.
The fLF represents organic matter for which interaction with minerals is minimal. Extraction of this fraction is delicate, since the mixing of soil with the dense solution may already break-up some macroaggregates. There are, however, indications that the organic matter present in macroaggregates may be more similar to the fLF stricto sensu than to the oLF released by high-energy sonication18. Some authors have even proposed a low-energy sonication step to isolate the pool of free and weakly mineral-interacting organic matter, termed 'intra-aggregate particulate organic matter', iPOM54.
For the release of occluded organic matter, different techniques exist to disrupt soil aggregates. The most widespread are sonication, agitation with glass beads and the use of chemical dispersants33,62,63. Sonication was chosen here because the output energy can be finely controlled and is believed to distribute more or less uniformly in the sample. By precluding the need to use chemical dispersants, sonication may be considered as relatively preservative towards organo-mineral complexes22, 33. The dispersion step, however, remains one of the most delicate operations. On the one hand, a weak dispersion will leave the aggregates intact and may lead to an over-estimate of hF SOC; on the other hand, a highly vigorous dispersion step could cause re-distribution of SOC across the fractions by partial destruction of organo-mineral complexes. Weak organic-sand associations may be particularly vulnerable to this process. Since occlusion within aggregates and surface sorption are processes occurring along a continuum2, no perfect solution exists. Therefore, the energy level of sonication needs to be adjusted thoughtfully according to the soil properties. Kaiser and Berhe64 have published a very helpful review that proposes a strategy to minimize artifacts caused by ultrasound when dispersing soils.
Reported sonication energies range from 60 to 5,000 J·mL-1. Several research groups have reported that 100 J·mL-1 could be sufficient to destroy macroaggregates and effectively disperse sandy soils, while 500 J·mL-1 would destroy large microaggregates and provide a reasonable dispersion of reactive soils63,65,66,67,68. In physical fractionation schemes, complete dispersion of silt and clay-sized aggregates may not be necessary, since the protection mechanism is likely to become indistinguishable from sorptive stabilization in these size ranges. A reasonable objective of dispersion prior to size or density fractionation may be to disrupt macro- (> 250 µm) and large micro- (> 53 µm) aggregates. Energies of 100 J·mL-1 (sandy soils) to 200 J·mL-1 (loamy soils) may be appropriate choices. An energy of 200 J·mL-1 may already extract a portion of microbial metabolites (supposedly mineral-associated)69, thus the use of higher sonication energies should be subject to caution. However, mineralogically reactive soils with cemented aggregates could require up to 500 J·mL-1 to disperse. It is essential that the dispersion energy be adjusted to match each soil type as well as study objectives. Finally, it is important to remember that even after supposedly complete ultrasonic dispersion, clay-sized microaggregates are likely to persist70.
A difficulty with harmonizing physical fractionation techniques resides in the heterogeneity found in soils, in particular in their mineral composition. The choice of dense solutions should be made on the basis of known or inferred soil mineralogy, with the ultimate goal to isolate fractions which are as homogeneous as possible.
In the article, the dense solution used was SPT – pH 371, 72. The low pH minimizes losses of soluble organic compounds. However, density fractionation may be performed with different dense solutions. Historically, organic liquids were used (tetrabromoethane, tetrachloromethane), but were gradually abandoned at the profit of inorganic salts (sodium iodide, SPT) because of the toxicity of halogenated hydrocarbons and the inherent contamination of soil organics. Nowadays, SPT is the preferred solution because its density can be adjusted between 1.0 to 3.1 g·cm-3, it can be recycled and has a low toxicity (unless ingested)22, 50. Main manufacturers offer a range of SPT grades differing in the level of carbon and nitrogen contamination. For density fractionation of soils, the purest grade is recommended, particularly if the fractions are to be analyzed for isotopic composition.
A solution of density 1.6 g·cm-3 has classically been used to separate light organic from mineral-associated fractions – see for example Golchin et al.21. While some authors have suggested that a density of 1 g·cm-3 (water) could be sufficient to extract most of the light fraction73, 74, others have proposed higher density cut-offs such as 1.62 or 1.65 g·cm-3 based on the idea that some organic components could show densities up to 1.60 g·cm-3 33,75,76. Densities as high as 1.85 g·cm-3 have even been employed50. When selecting a density to separate light from heavy fractions, it should be noted that no perfect solution exists. Indeed, lower densities risk attributing some 'light' organics to the heavy fractions, while higher densities risk including some minerals into the light fractions. This last effect can be detected when observing the carbon content of the light fractions, with a % SOC lower than 40-45% indicating some degree of mineral contamination.
For heavy fractions, preliminary analysis such as XRD can provide insight into the mineralogy of the bulk sample60 and help define density cut-offs capable of distinguishing between the main mineral components of a soil, keeping in mind that high organic loadings will lower the density of a mineral compared to its theoretical value. Similarly, for particle-size separation, a textural analysis77,78 can help set appropriate limits. Particle-size separation is a particularly attractive addition to simple density fractionation whenever sequential density fractionation is difficult. This is the case for instance for soils containing large amounts of oxyhydroxides and low-activity clays, which result in sample dispersion and prevent clear separations in heavy liquids. A particle-size separation step is also indicated to separate minerals of similar densities but different sizes (e.g., quartz and illite).
Free calcium ions will react with SPT to form insoluble Ca metatungstate. The procedure is thus inappropriate for alkaline soils containing large amounts of poorly crystalline, pedogenic carbonates. Small amounts of low-reactivity carbonates do not interfere with the fractionation as long as the samples are not left in contact with SPT for too long. Ca metatungstate precipitates will lead to an over-estimate of fraction masses. If LFs are run on an elemental analyzer for C concentration, the problem will be discovered but the fractionation will be compromised.
In addition to these technical difficulties, the fundamental limitation of CSDF (or of any physical fractionation scheme) stems from the fact that reactive minerals in soils rarely occur as discrete separates, but instead as coatings and cements. The occurrence of highly sorptive but very thin coatings on otherwise unreactive minerals (such as quartz) can lead to a biased view of organo-mineral associations. Caution is thus required when interpreting results, particularly for soils whose reactivity is dominated by poorly crystalline and oxide phases. Further characterization of fractions can help alleviate such ambiguities. Nevertheless, detailed physical fractionation methods such as CSDF have an unmatched ability to gain insight into the composition of naturally-occurring organo-mineral complexes. Such insight is expected to yield new understanding of the biogeochemistry of the largest pool of organic matter in soils, the mineral-associated one.
The authors have nothing to disclose.
The development of this method was supported by the Fond d'Investissement (FINV) of the Faculty of Geosciences at the University of Lausanne. We acknowledge the Uganda National Council for Science and Technology and Uganda Wildlife Authority for granting us permission to collect research samples. The authors further wish to thank Prof. Thierry Adatte for CHN and XRD analyses. We are grateful to Prof. Erika Marin-Spiotta for providing initial training in classical density fractionation. We also thank laboratory manager Laetitia Monbaron for her assistance in securing supplies and equipment.
Fractionation | |||
Sodium polytungstate | Sometu | SPT 0 (low C and N) is recommended. Lower grade polytungstate may contaminate samples. | |
Hydrometers (1-1.5, 1.5-2, 2-2.5, 2.5-3 g.cm-3) | Allafrance | Calibrated at 20 °C, e.g. 3050FG250/20-qp | |
Vortex mixer | Fisher | Fixed speed standard vortex mixer, e.g. 02-215-410 | |
Sonifier | VWR | Qsonica LLC – Q500 system with standard probe 4220 | |
Sonifier stand | VWR | Large clamp stand | |
Sonifier enclosure | VWR | Soundproof cabinet (optional) | |
Swinging-bucket centrifuge | Beckman | Able to achieve speeds of 4000 g or more, fitted with rotor accommodating 50 mL Falcon tubes | |
High-speed centrifuge with fixed angle rotor | Beckman | Able to achieve speeds of 7500 g or more, fitted with rotor accommodating 250 mL bottles | |
50 mL centrifuge Falcon tubes | Corning | e.g. 352070 | |
250 mL centrifuge bottles | Beckman | Polycarbonate bottles (e.g. 352070) are recommended because they are clearer than other plastics. | |
Vaccum filtration units | Semadeni | Polusulfone reusable units, e.g. 3029 | |
Polypropylene hose | Semadeni | To connect the filtration unit to vaccuum source | |
Ultrafiltration disks, 0.45 µm pore size | Millipore | e.g. HAWP04700 | |
Dessicator cabinet | Fisher scientific | 3 shelves, e.g. 305317-0120 | |
Drierite absorbent indicating | Millipore | Blue drierite, e.g. 10276750 | |
Scintillation vials | Fisher scientific | HDPE – separated cap 20mL, e.g. 12341599 | |
150 mL aluminium boats (smooth sides) | Fisher scientific | Any model. | |
Laboratory oven | Fisher scientific | Any model. | |
Recycling SPT column | |||
Cation exchange resin | Sigma-Aldrich | Dowex® Marathon™ C sodium form, strongly acidic, 20-50 mesh | |
Activated charcoal | Sigma-Aldrich | Darco S-51, 4-12 mesh | |
Glass wool | Fisher scientific | Pyrex | |
Filter paper, 2.5 µm pore size | Sigma-Aldrich | Whatman grade 42, e.g. WHA1442150 | |
Hydrogen peroxide | Sigma-Aldrich | Reagent grade. | |
Ethanol | Sigma-Aldrich | Reagent grade. | |
Polycarbonate 1000mL graduated cylinder | Semadeni | Any model. | |
Stand and clamp | Sigma-Aldrich | Size L – 2-prong | |
Polypropylene hose | Semadeni | Any model. | |
Polypropylene hose clamp | Semadeni | Any model. | |
Polypropylene funnels | Semadeni | Any model. | |
Polypropylene bottle (1L, 2L) | Semadeni | Any model. | |
Heating plate | Fisher scientific | Any model. |