Environ. Sci. Technol. 2001, 35, 3519-3525
An Improved Thermal Oxidation Method for the Quantification of Soot/Graphitic Black Carbon in Sediments and Soils YVES GE ´ L I N A S , * ,† K E N N E T H M . PRENTICE,† JEFFREY A. BALDOCK,‡ AND JOHN I. HEDGES† School of Oceanography, Box 357940, University of Washington, Seattle, Washington 98195-7940, USA, and CSIRO Land and Water, PMB #2, Glen Osmond, SA 5064, Australia
Recent findings have confirmed the importance of black carbon (BC) in the global biogeochemical cycles of carbon and oxygen through its important contribution to the slowly cycling organic carbon (OC) pool. Yet, most BC determination methods published to date measure operationally defined BC fractions, oftentimes with a high potential for artifacts and a lack of specificity for one of the two major forms of the BC continuum, soot/graphitic BC (GBC) and char/charcoal BC (CBC). This paper describes a method that reduces the potential for artifacts to accurately and selectively measure the concentration of GBC in complex mineral and organic matrixes. Marine and lacustrine sediments, river sediments, suspended particles, and a marine plankton sample were first demineralized with a mixture of hydrochloric (HCl) and hydrofluoric (HF) acids to expose any biochemical entrapped in a mineral matrix. The hydrolyzable organic matter fraction (mostly proteins and carbohydrates) was then removed with O2free trifluoroacetic acid and HCl, after which the non-GBC, non-hydrolyzable OC fraction was finally removed by thermal oxidation at 375 °C for 24 h. The specificity of the method for GBC was assessed with pure CBC and GBC samples. Detection limit and GBC recovery in spiked samples were 10 mg kg-1 and ∼85%, respectively. Typical GBC concentrations measured in a series of natural samples ranged from 93% at this temperature. The fact that the recovery for bituminous coal was essentially zero at a temperature of 375 °C suggests that there was little formation of GBC during the thermal decomposition process. Leaves, wood, and charred wood residues were also completely oxidized at 375 °C (leaves and wood data not shown in Figure 2 for clarity). The diesel soot and activated carbon samples lost intermediate amounts of carbon over the 350-400 °C temperature interval. Because of the wide range of carbon loss from activated carbon on heating between 350 and 400 °C, activated carbon can be used as a standard to compare the homogeneity of heating conditions between different groups of samples. The ∼40% carbon mass loss detected for diesel soot upon thermal treatment at 350 °C was suspected to arise from the volatilization of a fraction of non-GBC carbon contained in the sample. The cross-polarization 13C NMR spectrum acquired for diesel soot (Figure 3a) revealed a large contribution from alkyl carbon (0-45 ppm) in addition to the unsaturated carbon signal (110-145 ppm). The contriVOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3521
TABLE 1. Soot-BC Recovery sample Mex 306 spiked with diesel soot diesel Soot particulate
initial OC (%)
initial soot-BC (%)
measured soot-BCa (%)
recovery
8.93
0.33
0.29 ( 0.01
87.9
78.2
45.0
38.4 ( 0.6
85.3
a
The standard deviation only applies to the thermal oxidation step.
FIGURE 3. Cross-polarization (a) and Bloch Decay (b) 13C NMR spectra of the diesel soot sample used in this study. Resonance peaks centered at 30 and 130 ppm correspond to alkyl and unsaturated carbon, respectively. Integration of the Bloch Decay spectrum gives a relative signal intensity of 37% for alkyl carbon and 63% for unsaturated carbon. bution from N-alkyl (46-60 ppm), O-alkyl and di-O-alkyl (60-110 ppm), phenol (145-165 ppm), and carbonyl (165210 ppm) carbon was negligible. Given that all unsaturated C may not be observable in cross-polarization 13C NMR, a Bloch decay analysis was completed. Bloch decay does not selectively under-represent carbon atoms that are not bound to hydrogen atoms, such as internal aromatic carbons in polyaromatic layers. Relative intensities of 37 and 63% were obtained for alkyl and unsaturated carbon, respectively, by integrating the signal intensity acquired in the Bloch decay analysis (10) (Figure 3b). The alkyl carbon fraction, most likely present in alkyl groups attached to aromatic rings or in straight-chain hydrocarbons, is not stable at high temperature and is therefore lost during the thermal oxidation treatment. Thus, in agreement with Gustafsson et al. (33), it appears that GBC can be selectively measured by thermal oxidation at 375 °C with no interference from CBC. Although graphite-BC cannot be distinguished from soot-BC at that temperature, graphite is rare in nature and most likely represents a negligible fraction of the GBC concentration in environmental samples. GBC Recovery. One of the main challenges faced in BC analysis when using wet chemical pretreatment is to quantitatively recover the GBC fraction. Because individual GBC particles are highly hydrophobic and generally in the low micrometer size range, there is a significant potential for losses either by sorption to the walls of the containers or when separating the supernatants from the solid residues. Extreme caution should thus be taken when handling samples during pretreatment. To evaluate the importance of this source of error, the Mexican coast (Mex306) marine sediment with a very low GBC content was spiked with diesel soot to obtain a natural organic-rich sediment sample with a GBC concentration of 0.33% (or a 3.7% GBC/TOC concentration). The spiked Mex306 sediment and an aliquot of the pure diesel soot particulate sample were run through the entire method and the GBC concentration was measured in the combusted residue. The results shown in Table 1 illustrate typical recoveries obtained with this method. About 85% of the initial GBC mass was recovered from both samples. GBC losses were most likely physical rather than due to oxidation since the total mass losses during the chemical pretreatment were 3522
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001
also ∼15% and the GBC/TOC concentration in the demineralized and HOM-extracted diesel soot sample was identical to the GBC/TOC concentration in the natural sample. The losses are unlikely to be due to oxidation during pretreatment because (i) GBC is fairly inert toward oxidation, and (ii) the demineralization and HOM removal steps were carried out with non-oxidizing acids in solutions purged with N2 to minimize oxidative losses. However, because about 15 to 20% of GBC was lost through thermal oxidation and sample manipulation, the results obtained may slightly underestimate actual GBC concentrations in the samples. Importance of Demineralization and HOM Removal. Dispersion of GBC in the global environment mainly follows the atmospheric transport and deposition pathway, a route that entails fairly uniform concentrations over wide geographical areas (3). Since total OC concentrations in coastal marine and abyssal sediments vary widely, it is surprising to find literature data suggesting remarkably uniform GBC/ TOC concentrations in marine and estuarine sediments collected in widely varying depositional settings in Europe or America. This uniformity is best shown by plotting GBC concentrations against TOC concentrations for sediments listed in recently published studies (Figure 4a,c; refs 16, 38). The slope of the GBC vs TOC concentrations varied between sample types and investigators (ranging between 0.03 in Figure 4c and 0.17 in Figure 4a). Similar covariation was found in our laboratory for soils and sediments when using the same thermal oxidation method with no chemical pretreatment (Figure 4b,d). While a global relationship between GBC and TOC cannot be ruled out based on this small data set (the only published GBC data available), this correlation might also be due to the incomplete removal of non-GBC organic materials or to the formation of condensation products during the combustion step. To probe these possibilities, a set of samples varying in nature (suspended particles, riverine, lacustrine and marine sediments, and marine plankton), as well as in OC concentration (from 0.46 to 8.32 wt % OC) were demineralized and their HOM fraction was removed as described in Experimental Section. Aliquots of the natural, demineralized, HOM-extracted, and demineralized + HOM-extracted samples were then simultaneously combusted at 375 °C for 24 h. The results from this experiment are shown in Figure 5. The highest GBC concentrations were measured following thermal oxidation of the natural samples, with a strong GBC to OC covariation (r2 ) 0.66, squares in Figure 5). Particularly surprising is the 0.54 wt % GBC concentration found for Dabob Bay plankton (OC% ) 8.32), a sample of fresh organic material that should contain very little, if any, GBC. Demineralizing (r2 ) 0.35, diamonds) or removing HOM (r2 ) 0.10, triangles) from the sediments before the combustion step resulted in lower GBC concentrations in most samples, particularly for the samples having higher OC contents. The lowest concentrations of GBC were obtained when minerals and HOM were removed from the samples before the thermal oxidation step (r2 ) 0.02, circles). The results suggest the presence of an artifact caused by the presence of organic materials that should have been removed during thermal oxidation, or by the formation of condensation products when
FIGURE 4. Correlation between GBC concentration and total OC in (a) marine sediments from Molenplaat, Schelde estuary (Netherlands), the Atlantic Iberian margin, the northwestern Black Sea, the North Sea, the Madeira abyssal plain, and eastern Mediterranean (38); (b) Australian soils and lacustrine sediments (our data and 29); (c) marine sediments from the Gulf of Maine and the Palos Verdes shelf off Los Angeles (16); (d) marine, riverine, and lacustrine sediments as well as riverine suspended particles (this work). Three data points (marked **) were excluded from graph (c) and two from graph (d) in the calculation of the correlation coefficients (p > 0.95). The thermal oxidation methods used for these measurements are described in refs 33 (circles and solid line) and 15 (squares and dotted line).
FIGURE 5. GBC concentration measured by thermal oxidation for a wide range of natural and chemically treated samples. Each sample was thermally oxidized in the natural form (thermal oxidation), after demineralization (demineralization + thermal oxidation), after removal of hydrolyzable OM (HOM removal + thermal oxidation), and after demineralization and removal of HOM (demineralization + HOM removal + thermal oxidation). Results from the Buffalo River sediment, a sample with an unusually high GBC concentration, were not included in the calculation of the correlation coefficients. Correlation coefficients >0.30 are significant at the 90% confidence level. Samples and chemical treatments are described in Experimental Section. hydrolyzable biochemicals are heated at 375 °C, leading to an overestimation of the actual GBC concentration. While demineralization removes the more labile hydrolyzable biomolecules and most of the minerals (34), a fraction of macromolecular proteins, carbohydrates, and lipids are not
hydrolyzed and might fuel condensation reactions at high temperatures. Indeed, a simple 1:1 mixture of proteins (bovine serum albumin) and glucose heated at 375 °C for 24 h in an open crucible produced a 2.6 wt % concentration of thermally stable organic materials, equivalent to an apparent VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3523
TABLE 2. Soot-BC Measurement soot-BC/total weight (%)
soot-BC/total OC (%)
sample
sample type
initial OC (%)
no treatment
treated
no treatment
treated
MEX 313 MEX 305 MEX 306 WEC 204 WEC 216 WEC 219 Saanich LWSM Buffalo R. estuarine Dabob Bay Madeira BUCK ACU SS6 SS7 SS8
marine sediment marine sediment marine sediment marine sediment marine sediment marine sediment marine sediment lacustrine sediment riverine sediment estuarine sediment marine plankton riverine susp. part. soil soil soil soil soil
1.76 3.30 7.99 1.42 1.27 1.41 2.69 4.69 2.43 0.51 8.32 0.46 1.56 2.55 7.16 12.33 2.63
0.1688 0.2469 0.2647 0.0853 0.1243 0.1260 0.1812 0.0856 0.3720 0.0473 0.5438 0.0432 0.0731 0.1498 0.3783 0.5155 0.1520
0.0024 0.0053 0.0037 0.0146 0.0237 0.0194 0.0245 0.0438 0.1928 0.0109
