Summary of Organic and Elemental Carbon/Black Carbon Analysis Methods and Intercomparisons

Many different thermal, optical, and thermal/optical carbon analysis methods for organic carbon (OC), elemental carbon (EC) or black carbon (BC) have been applied throughout the world to evaluate visibility and the Earth’s radiation balance. Dozens of intermethod and interlaboratory comparison studies have been conducted. Several of these studies are catalogued and summarized here. BC or EC concentrations are found to differ by up to a factor of 7 among different methods; factor of 2 differences are common. Differences between methods are not consistent among comparison studies, with some methods showing higher BC for one set of samples and lower BC for other sets relative to a common benchmark. The absorption efficiency relating light absorption (babs) to EC that is derived from collocated optical and chemical measurements can vary by a factor of 10, depending on the collocated babs and EC measurement methods. Future intermethod and interlaboratory comparisons must include components that seek to understand the causes of these differences.


Introduction
Elemental carbon (EC) and black carbon (BC) are operationally defined by the measurement method applied, although EC and BC are often used interchangeably.The objective of this paper is to: 1) summarize filter methods used to measure organic carbon (OC) and EC; 2) assemble interlaboratory and intermethod comparisons; and 3) identify knowledge gaps and research needs.
EC occurs as the mineral graphite or as diamond in its purest forms, but these structures of more than 0.1 micrograms (µg) are seldom found in ambient particulate matter (PM).Freshly emitted diesel soot consists of agglomerates of small spherical graphitic particles consistent in size from 20 to 30 nanometers (nm).Even the soot from incomplete combustion contains non-carbon components and has a non-crystalline structure (Akhter et al., 1984(Akhter et al., , 1985)).It has a large surface-to-volume ratio and reactive surfaces, so it attracts condensable materials (e.g., polycyclic aromatic hydrocarbon [PAH] gases) soon after emission into ambient air.
Of the major components of PM 2.5 and PM 10 , OC and EC are the most uncertain with respect to sampling and analysis (Huebert and Charlson, 2000;Jacobson et al., 2000;Turpin et al., 1994).Most EC and BC characterization involves collecting PM on filters and measuring either the carbon content on the filter or the attenuation of light reflected from or transmitted through the filter.Filterbased optical techniques include the British Smoke Shade method (Hill, 1936), the coefficient of haze (COH) (Hemeon et al., 1953), the integrating plate method (IPM) (Lin et al., 1973), the aethalometer (Hansen et al., 1984), and the particle soot absorption photometer (PSAP) (Bond et al., 1999).The scattering and absorption properties of particles distributed on top of and throughout a filter are not the same as they are in the atmosphere.Light absorption coefficients (b abs ) determined from these methods are often biased (Horvath, 1993).
It is generally agreed that EC is the major contributor to b abs (e.g., Horvath, 1993;Watson, 2002).
EC absorbs light due to conduction electrons associated with the graphitic structure.EC is, therefore, often referred to as BC.However, the specific mass absorption efficiency of EC has been estimated to range from 2 to 20 m 2 /g (Liousse et al., 1993).Particle light absorption depends on the wavelength (λ) of the incident light.Moosmüller (1998) reported that EC absorption efficiency varied as λ -2.7 near Denver, CO, while Horvath et al. (1997) reported that absorption efficiency for aerosols in Santiago, Chile, varied as λ -0.92 .Kirchstetter et al. (2004) found that b abs from engine exhaust varied as λ -1 whereas b abs from biomass burning varied as λ -2 .Carbonaceous material from different sources (e.g., diesel versus wood burning) has different structures and compositions.A small quantity of carbonates (e.g., CaCO 3 ) is found in some fine particulate samples, but this is rarely comparable to the EC content (Chow and Watson, 2002;Cao et al., 2005).The remainder of the carbonaceous material is organic matter that is a complex mixture of hundreds of organic compounds covering a wide range of molecular forms and volatilities (Jacobson et al., 2000).Organic matter can be emitted directly from combustion sources with EC, or it can be formed in the atmosphere through condensation of low-volatility oxidation products of hydrocarbons (i.e., secondary organic aerosol [SOA]).Some components of OC may be weakly light-absorbing in the visible spectrum, but OC mainly influences direct radiative forcing through light-scattering or through mixing with EC to enhance the EC absorption efficiency (e.g., Fuller et al., 1999).
Particles change when they are extracted from the air onto a filter, on which most EC or BC measurements are made.The rate at which the material evaporates depends on how the filter is handled and stored between sampling and analysis.Quartz-fiber filters used for thermal carbon analysis absorb some organic vapors throughout their thickness; these vapors are often interpreted as OC by thermal methods and possibly as a portion of EC if the OC is charred.Fuller et al. (1999) hypothesized that differences among estimates for soot extinction efficiencies are due to: 1) different wavelength dependencies; 2) deviations from spherical particles; 3) mischaracterization of the soot refractive index; 4) inaccurate densities; and 5) mixtures of graphitic material with other compounds.By applying radiative transfer models for non-spherical particles of non-homogeneous composition, Fuller et al. (1999) found that for the same quantity of EC, b abs decreased rapidly for particles >0.1 µm, the particle size of most of the aged EC in the atmosphere.
Up to 60% higher efficiencies were calculated for long chain aggregates relative to the same amount of EC in a sphere.Absorption efficiencies >10 m 2 /g-exceeding 25 m 2 /g in some situations-were estimated for soot imbedded in a sulfate particle.Efficiency decreased as the EC fraction in the particle increased, implying that a lower EC concentration may yield a higher contribution per EC mass to light absorption under some circumstances.Efficiencies decreased by nearly 1 order of magnitude as an EC core at the center of a concentric sphere migrated through the sphere to its surface.Martins et al. (1998) found similar results, with maximum soot absorption efficiencies approaching 30 m 2 /g when the carbon constituted 0.5% of the volume of a 0.5 µm diameter particle.
Empirically derived soot extinction efficiencies are usually greater at non-urban than urban monitors (Horvath, 1993).This is consistent with an aged aerosol in which more of the fresh emissions become coated with condensed, absorbed, and adsorbed material.Horvath (1993Horvath ( , 1997) ) shows how light transmission through a filter varies depending on the filter loading, the presence or absence of light-scattering particles, and the location of particles within a filter.These results imply that constant conversion factors used to infer EC from b abs , such as the 10 m 2 /g often used in visibility studies (Watson, 2002), are subject to large uncertainties.
OC and EC are measured directly by thermal evolution methods that quantify the amount of carbon that leaves the filter at different temperatures (Currie et al., 2002;Schmid et al., 2001).These methods use different combinations of temperature and analysis atmospheres to evaporate, pyrolyze, and combust the carbon-containing compounds on a filter sample, then detect the evolved carbon gases.The separation of OC from EC is ambiguous because some of the EC combusts in the presence of oxygen, and some of the OC chars (turns to EC) in an oxygen-deficient atmosphere.
Light reflected from (Johnson et al., 1981;Huntzicker et al., 1982;Chow et al., 1993) or transmitted through (Turpin et al., 1990;Birch andCary, 1996a, 1996b;Chow et al., 2001) the filter during the analysis is used to monitor and correct for this charring.Interlaboratory and intermethod comparisons (e.g., Chow et al., 2001;Countess, 1990;Currie et al., 2002;Schmid et al., 2001) show EC differences of a factor of 2 or more among thermal methods, depending on the protocol and type of sample.Analysis methods alone can account for the large differences in EC emission rates among inventories.In addition to OC and EC, carbon that evolves at several different temperatures has been found useful for source apportionment studies (Watson et al., 1994;Kim and Hopke, 2004).

Thermal Organic and Elemental Carbon Analysis Methods
Table 1 summarizes several of the thermal methods that have been applied to estimating total carbon (TC), OC, and EC.All of these produce OC and EC concentrations that are defined by the method rather than by an absolute standard.Many of these methods use a two-step temperature in which the carbon evolving below the temperature (~350 to 550 °C) is termed OC and the remaining carbon evolving at a higher temperature (~650 to 1100 °C) is termed EC.Between different methods, lower OC temperatures are used in an oxidizing atmosphere for which the combustion rate of EC is assumed to be low; higher OC temperatures are usually applied in a non-oxidizing atmosphere, with an oxidizer added at a time after which most of the OC is assumed to have left the sample.The evolved carbon is converted to carbon dioxide (CO 2 ), which can be detected directly or converted to methane (CH 4 ) for more sensitive detection.
As Table 1 shows, many of the methods employ thermal/optical reflectance (TOR) and/or thermal/optical transmission (TOT) to monitor the conversion of OC to EC as part of the analysis.
Since EC is not volatile, it is released only by oxidation (typically in an oxygen [O 2 ] atmosphere at a temperature below 800°C) and is thereby separated from OC. Heating in an O 2 -free environment, however, causes certain OC components to pyrolyze and form non-volatile, light-absorbing, "charred," material that can be mistaken for atmospheric EC.Light reflected from or transmitted through the sample monitors the darkening of the particle deposit on the filter due to OC charring.
When O 2 is added to the analysis atmosphere (i.e., carrier gas) at a sufficiently high temperature (e.g., >350 °C), this black char combusts along with the original EC on the filter, and the filter becomes whiter.When the reflected or transmitted light attains its original intensity, the charred, or pyrolyzed, OC (POC) is considered to have been removed.All of the remaining carbon is associated with the EC that was originally on the filter.Therefore, a partitioning can be made by assigning carbon evolved before this split point to OC and after this point to EC. Johnson et al. (1981) and Yang and Yu (2002) pointed out that such partitioning assumes that: 1) charred OC evolves before the original EC in the thermal analysis, and 2) charred OC and original OC equally attenuate reflectance (R) and transmittance (T).Several of the thermal methods appear to be similar, but they contain subtle differences with respect to: 1) analysis atmospheres, 2) temperature ramping rates, 3) temperature plateaus, 4) residence time at each plateau, 5) optical pyrolysis monitoring configuration and wavelength, 6) standardization, 7) oxidation and reduction catalysts, 8) sample aliquot and size, 9) evolved carbon detection method, 10) carrier gas flow through or across the sample, 11) location of the temperature monitor relative to the sample, and 12) oven flushing conditions.These differences are not always well characterized or reported along with analysis results, but this information is critical to interpret variability in the results.For example, Chow et al. (2005a) demonstrate the influence of temperature calibration on TOR analysis for temperature-resolved carbon fractions.Chow et al. (2004) and Chen et al. (2004) discuss how TOR and TOT might determine different OC/EC splits for various types of samples.Owing to differences in the form of EC and the methods that operationally separate OC from EC, there is no reason to expect a consistent relationship among samples measured in different laboratories.This is borne out by the summary of 40 different intermethod and interlaboratory comparisons studies in Table 2.These comparisons date from the early 1980s to the present and cover a wide range of samples, the thermal evolution methods described in Table 1, and types of sampled particles.
In some studies, several simulated and ambient samples were sent to different laboratories.These comparisons showed that TC was the same for well-calibrated instruments, but the OC/EC splits were different.Since EC usually constitutes the smallest fraction of TC, it shows the largest variation among laboratories (up to a factor of 7 as reported by Currie et al., 2002).Many of the methods agreed on EC from diesel exhaust or powdered graphite samples, but they often disagreed for ambient and biomass burning samples.Schmid et al. (2001) showed that biases were not consistent from sample to sample for paired measurements from several thermal/optical protocols.
This demonstrated that differences depend on the samples analyzed as well as on the instruments and methods used for the analyses.Potential sample biases include: 1) non-uniform particle deposits on the filter; 2) particle deposits that are too light or too dark, which make R and T charring corrections uncertain; 3) organic vapor filter adsorption and its charring during heating; 4) catalytic and oxidation interactions between OC, EC, and non-carbonaceous material in the sampled particles; and 5) changes in optical properties of the particles during thermal evolution.
It is doubtful that future carbon comparisons will add much information to those already completed unless they include components that systematically quantify the effects of sample properties and the analysis variables cited above.This degree of systematization is lacking in most of the studies.While it is possible to observe differences, it is not possible to explain why.
As an example, Chow et al. (2004) examined charring of the filter backside and microscopic crosssections of a filter punch at different parts of the temperature cycle during thermal analysis.It appeared that R was dominated by charring of OC that co-existed with EC in particles on the surface of the filter, while T was dominated by the charring of organic vapors distributed throughout the filter.When oxygen was added to the analysis atmosphere, the surface EC (original EC and charred OC) evolved before the charred OC that was distributed throughout the filter.Hence, T achieved its initial value later than R. Using a radiative transfer model, Chen et al. (2004) showed that this explanation is plausible, and that simultaneous R and T measurements can be used to estimate contributions to charring from the surface particulate OC and the charred vapors adsorbed throughout the filter.a Advance from one temperature to the next when a well-defined carbon peak has evolved.
b Sample is acidified with 30 µl of 0.02N hydrochloric acid (HCl) and preheated at 120 ºC in contact with a bed of granulated manganese dioxide (MnO 2 ) in helium (He) for 180 s to remove volatile OC and carbonate.The sample remains in contact with MnO 2 throughout analysis.c Temperature change is accomplished by moving the sample from a lower-temperature oven to a higher-temperature oven.
d Sample is preheated at 350ºC in 98% He/2% oxygen (O 2 ) environment until all volatile OC is removed.
e Cool to ~ 350 ºC before the introduction of O 2 (Turpin et al., 1990).
f The third of four carbon dioxide (CO 2 ) peaks evolving during thermal analysis is assigned to EC; this peak is usually accompanied by an increasing filter transmittance.g Residence times for loadings < 50 µgC and > 50 µgC per cm 2 are 300 s and 900 s each, respectively, for both 400 °C and 700 °C temperature stages.h Sample is preheated at 350 ºC in ambient air for ~ 450 s to remove volatile OC.
i OC on half a filter is extracted with 10 ml of a 50:50 volume %-mixture of toluene and 2-propanol for 24 hr, and the halffilter is dried in pure nitrogen (N 2 ) before thermal analysis.OC is estimated from the difference between TC and EC acquired from two separated filter halves.j The sample is cooled to 80 ºC from 700 ºC.Carbon evolved during cooling is also counted as EC.
k Temperature changes are accomplished by moving the sample through an increasing temperature gradient in the oven.
l The overall analysis time is adjustable but should not be > 30 min.
Figure 1 shows how EC determined by TOT compared with EC by TOR for a high and low temperature analysis protocol.EC by TOT was lower than that by TOR in both cases, with the high temperature protocol yielding much lower EC values.Chow et al. (2004) hypothesized that much of the organic vapors adsorbed throughout the filter is desorbed at the 120 °C and 250 °C plateaus of the low temperature protocol, leaving less to char during the remainder of the analysis.The high initial temperature protocol begins analysis at 310 °C, and charring is immediately observed before volatilization is possible.
Figure 2 compares EC from the high temperature and low temperature protocols using the TOR correction.The values are in good agreement for this case.The correction is less sensitive to the temperature protocol than is the transmittance correction.
Carbonate carbon present with the OC and EC may interfere with the analysis if it decomposes during the analysis.For the high temperature protocol used in Figures 1 and 2, temperatures exceed 850 °C during the OC step, while for the low temperature protocol the temperature does not reach 800 °C until the EC step.Chow and Watson (2002) demonstrated that calcium carbonate is uncommon in many atmospheric samples and is not measured by thermal methods with temperatures <800 °C.Chow et al. (2001) showed that high temperatures (e.g., 800 °C) in the OC step may also oxidize EC via reactions with O 2 -containing minerals in the sample (Fung, 1990;Fung et al., 2002).
While most air quality studies in the U.S. have measured OC and EC by thermal/optical methods, the two-temperature method of Cachier et al. (1989aCachier et al. ( , 1989b) is most widely used in global inventories and ambient studies.Owing to their widespread use, similarities and differences between the IMPROVE, STN, and CNRS-CEA protocols in Table 1 need special attention with respect to their comparability.

Optical Light Absorption Measurements
Also summarized in Table 2 are comparisons of thermal EC measurements with in-situ or filterbased optical BC measurements.A fundamental measurement of in-situ aerosol b abs can be achieved by the photoacoustic instrument (Arnott et al., 1999;Moosmüller et al., 1997), which quantifies minute changes in the speed of sound in response to heating and cooling of PM by a modulated laser beam.It may be possible to relate EC or BC measurements to their absorption properties by collocating filter-based samplers with photoacoustic measurements.
Photoacoustic instruments have been used to measure BC in engine exhaust for more than two decades (Faxvog and Roessler, 1979;Killinger et al., 1980;Japar et al., 1982Japar et al., , 1984;;Roessler, 1984), though it has only been recently that lasers have become adequately compact and powerful so that practical portable instruments can be fielded.The large dynamic range of the photoacoustic photometer allows it to measure b abs over a wide range of BC concentrations in source and ambient samples (Moosmüller et al., 2001a(Moosmüller et al., , 2001b)).For BC size distributions with mass median diameter < 0.3 µm, the BC mass absorption efficiency (b abs /BC) is stable (Killinger et al., 1980).As explained in the introduction, b abs can be translated into a BC or EC concentration with an appropriate mass absorption efficiency.Consistent with other studies cited above, Table 2 shows that mass absorption efficiencies differ from study to study.These differences result from variability in the shape, density, material refractive index, and internal mixing of particles as well as inconsistencies in quantification EC by other methods.
A more common method of b abs and BC (or EC) measurement is filter-based absorption using an aethalometer.The aethalometer measures optical aerosol absorption by quantifying the attenuation of light transmitted through a filter tape on which aerosol particles are collected (Hansen et al., 1984).
Once the spot monitored on the filter tape exceeds a certain optical density, the aethalometer automatically advances the tape to a new spot.In this manner, the aethalometer can perform months of ambient monitoring without operator attention.The measured attenuation is reported as BC concentration after conversion with an empirically determined factor.The aethalometer also obtains its calibration by comparison with thermal EC measurements (Hansen et al., 1984;Hansen and Novakov, 1990).Arnott et al. (2005) suggested that a filter-loading correction is needed to account for the multiple scattering effects of filter material in the aethalometer.A 50% reduction occurs between when the filter is pristine and white after a filter change, and when it is dark right before a filter change.Time averaged aethalometer data tends to average out this variation if the source of BC is sufficiently steady.Otherwise, the aethalometer b abs would not be linearly related to BC.This correction was not made in the earlier studies summarized in Table 2, and a wide range of b abs /BC ratios is observed.
Similar to the aethalometer, the integrating plate method (IPM, Lin et al., 1973), measures the transmission of diffused light through a polycarbonate-membrane, Teflon-membrane, or quartz-fiber filter.Quartz-fiber filters have more internal scattering than the thin polycarbonate-or Teflonmembrane filters, and the transmittance is usually double that for a given deposit on the membrane filters.This method can be applied to 47 millimeter (mm) Teflon-membrane filters with a photographer's densitometer for filter transmission that is calibrated with photographers' neutral density filters (Wratten Kodak, Rochester, NY).Filter transmission is highly correlated with the aethalometer measurement and has been used as a surrogate for BC in several studies (e.g., Chow et al., 1997;Watson and Chow, 2002).Bond et al. (1999) note the importance of correcting for particle scattering when a mixed aerosol is sampled and measured by filter transmission as with the IPM.

Conclusions and Knowledge Gaps
This summary documents and compares many OC and EC measurement methods and the efforts made to determine the equivalence among them.OC and EC are operationally defined by these methods, but they are often not sufficiently documented to allow the methods to be repeated outside of the laboratory origin.In seeking to determine equivalence, samples have been prepared from organic chemicals, graphite, diesel exhaust, and wood smoke.Ambient samples have also been circulated among laboratories or taken in parallel in a variety of environments.Measurements of particle absorption have been compared with thermal evolution and photoacoustic methods.The general conclusion is that different studies give different results for method comparisons, and that citation of a single comparison study is insufficient to establish comparability.More systematic comparisons are needed that hold most variables constant while varying only a few.These variables apply to the type of sample analyzed as well as to the analysis method.
Future comparison studies need to: Evaluate the sensitivity of OC and EC concentrations to variations in thermal evolution temperatures, pyrolysis corrections, analysis atmosphere compositions, presence or absence of oxidizing minerals and catalysts, vapor adsorption, and optical pyrolysis correction methods.
Create reproducible and well characterized samples with homogeneous deposits of light, medium and heavy particle loadings that represent simple (e.g., graphite powder, organic compounds, carbon arc emission) and complex (diesel exhaust, wood smoke, mixtures with inorganic minerals and salts) situations that might be found in the environment.
Implement the methods in Table 1 on a single instrument so that variables can be systematically changed for analysis of the prepared samples and selected ambient samples.
Develop and apply methods to calibrate and audit temperatures, analysis atmospheres, and optical monitoring that can be applied to a variety of hardware to determine that analysis assumptions are met in practice.
Interpreting the data from comparisons that incorporate these features will provide the basis for determining why different methods give different results.

Table 1 .
Summary of organic and elemental carbon thermal analysis protocols.