Application of the ECT9 protocol for radiocarbon-based source apportionment of carbonaceous aerosols

Carbonaceous aerosol is mainly composed of organic carbon (OC) and elemental carbon (EC). Both OC and EC originate from a variety of emission sources. Radiocarbon (14C) analysis can be used to apportion bulk aerosol, OC, and EC into their sources. However, such analyses require the physical separation of OC and EC. Here, we apply of ECT9 protocol to physically isolate OC and EC for 14C analysis and evaluate its effectiveness. Several reference materials are selected, including two pure OC (fossil “adipic acid” and contemporary “sucrose”), two pure EC (fossil “regal black” and “C1150”), and three complex materials containing contemporary and/or fossil OC and EC (“rice char”, NIST urban dust standards “SRM1649a” and “SRM8785”, i.e., fine fraction of resuspended SRM1649a on filters). The pure materials were measured for their OC, EC, and total carbon (TC) mass fractions and corresponding carbon isotopes to evaluate the uncertainty of the procedure. The average accuracy of TC mass, determined via volumetric injection of a sucrose solution, was approximately 5 %. Ratios of EC/TC and OC/TC were highly reproducible, with analytical precisions better than 2 % for all reference materials, ranging in size from 20 to 100 μg C. Consensus values were reached for all pure reference materials for both δ13C and fraction modern (F14C), with an uncertainty of < 0.3 ‰ and approximately 5 %, respectively. The procedure introduced 1.3± 0.6 μg of extraneous carbon, an amount compatible to that of the Swiss_4S protocol. In addition, OC and EC were isolated from mixtures of pure contemporary OC (sucrose) with pure fossil EC (regal black) and fossil OC (adipic acid) with contemporary EC (rice char EC) to evaluate the effectiveness of OC and EC separation. Consensus F14C values were reached for all OC (∼ 5–30 μg) and EC (∼ 10–60 μg) fractions with an uncertainty of∼ 5 % on average. We found that the ECT9 protocol efficiently isolates OC or EC from complex mixtures. Based on δ13C measurements, the average contribution of charred OC to EC is likely less than 3 % when the OC loading amount is less than 30 μg C. Charring was further assessed by evaluating thermograms of various materials, including aerosol samples collected in the Arctic and from tailpipes of gasoline or diesel engines. These data demonstrate that the ECT9 method effectively removes pyrolyzed OC. Thus, the ECT9 protocol, initially developed for concentration and stable isotope measurements of OC and EC, is suitable for 14C-based apportionment studies, including μgC-sized samples from arctic environments. Copyright statement. The works published in this journal are distributed under the Creative Commons Attribution 4.0 License. This licence does not affect the Crown copyright work, which is reusable under the Open Government Licence (OGL). The Creative Commons Attribution 4.0 License and the OGL are interoperable and do not conflict with, reduce, or limit each other. © Crown copyright 2021

Abstract. Carbonaceous aerosol is mainly composed of organic carbon (OC) and elemental carbon (EC). Both OC and EC originate from a variety of emission sources. Radiocarbon ( 14 C) analysis can be used to apportion bulk aerosol, OC, and EC into their sources. However, such analyses require the physical separation of OC and EC.
Here, we apply of ECT9 protocol to physically isolate OC and EC for 14 C analysis and evaluate its effectiveness. Several reference materials are selected, including two pure OC (fossil "adipic acid" and contemporary "sucrose"), two pure EC (fossil "regal black" and "C1150"), and three complex materials containing contemporary and/or fossil OC and EC ("rice char", NIST urban dust standards "SRM1649a" and "SRM8785", i.e., fine fraction of resuspended SRM1649a on filters). The pure materials were measured for their OC, EC, and total carbon (TC) mass fractions and corresponding carbon isotopes to evaluate the uncertainty of the procedure. The average accuracy of TC mass, determined via volumetric injection of a sucrose solution, was approximately 5 %. Ratios of EC/TC and OC/TC were highly reproducible, with analytical precisions better than 2 % for all reference materials, ranging in size from 20 to 100 µg C. Consensus values were reached for all pure reference materials for both δ 13 C and fraction modern (F 14 C), with an uncertainty of < 0.3 ‰ and approximately 5 %, respectively. The procedure introduced 1.3 ± 0.6 µg of extraneous carbon, an amount compatible to that of the Swiss_4S protocol.
In addition, OC and EC were isolated from mixtures of pure contemporary OC (sucrose) with pure fossil EC (regal black) and fossil OC (adipic acid) with contemporary EC (rice char EC) to evaluate the effectiveness of OC and EC separation. Consensus F 14 C values were reached for all OC (∼ 5-30 µg) and EC (∼ 10-60 µg) fractions with an uncertainty of ∼ 5 % on average. We found that the ECT9 protocol efficiently isolates OC or EC from complex mixtures. Based on δ 13 C measurements, the average contribution of charred OC to EC is likely less than 3 % when the OC loading amount is less than 30 µg C.
Charring was further assessed by evaluating thermograms of various materials, including aerosol samples collected in the Arctic and from tailpipes of gasoline or diesel engines. These data demonstrate that the ECT9 method effectively removes pyrolyzed OC. Thus, the ECT9 protocol, initially developed for concentration and stable isotope measurements of OC and EC, is suitable for 14 C-based apportionment studies, including µg C-sized samples from arctic environments. Putaud et al., 2010;Yang et al., 2011a, b;Hand et al., 2013;Ridley et al., 2018) and is a complex mixture composed of mainly light-scattering organic carbon (OC) and highlyrefractory, light-absorbing elemental carbon (EC, also referred to as black carbon) (Pöschl, 2005). The OC and EC fractions play important and often distinct roles in climate (Bond et al., 2013;Hallquist et al., 2009;Kanakidou et al., 2005;Laskin et al., 2015), air pollution, and human health (Cohen et al., 2017;Grahame et al., 2014;Janssen et al., 2012). Moreover, both OC and EC were identified as shortlived climate forcers (SLCFs) by the IPCC expert meeting (https://www.ipcc-nggip.iges.or.jp/public/mtdocs/1805_ Geneva.html, 15 April 2021) in 2018. To develop and monitor the efficiency of mitigation strategies for both climate change and air pollution, it is required to have a better understanding of the temporal and spatial dynamics of OC and EC emission sources.
The majority (> 50 %) of carbonaceous aerosol is OC, which has a wide size range. Coarse OC (in PM 10 ) consists of plant debris, microorganisms, fungal spores, and pollen. Fine OC (in PM 2.5 ) is formed predominantly via the oxidation or nucleation and coagulation of volatile organic compounds, such as mono-and sesquiterpenes, from both biogenic and anthropogenic sources (Shrivastava et al., 2017) but can also be directly emitted from combustion sources (Hallquist et al., 2009;Fuzzi et al., 2015;Liggio et al., 2016). In contrast, EC is found primarily in fine particles, e.g., PM 1.0 or smaller (Chan et al., 2013;Bond et al., 2013). It is emitted through incomplete combustion of fossil fuels and biomass and biofuels (Bond et al., 2013;Huang et al., 2010;Evangeliou et al., 2016;Winiger et al., 2016Winiger et al., , 2017Winiger et al., , 2019. Measuring the isotopic signature and composition, i.e., radiocarbon ( 14 C) content and stable isotope ratio ( 13 C/ 12 C) of aerosol, offers a powerful tool for quantifying the sources of bulk aerosol and its OC and EC fractions. Aerosol 14 C content can be used to quantify the relative contributions from contemporary biomass and fossil sources (Heal, 2014). 14 C is a naturally occurring radioisotope (5730 year half-life) produced in the atmosphere. After its oxidation to carbon dioxide ( 14 CO 2 ), 14 C enters the food chain through photosynthesis so that all living organisms are labeled with a characteristic 14 C/ 12 C ratio and described as "modern" carbon. Materials containing carbon older than about 50 000 years ( 14 C 12 C) are described as "fossil" carbon. Over the past centuries, the 14 C content of the atmosphere has undergone distinct changes (Graven, 2015;Graven et al., 2020;Levin et al., 2010): anthropogenic combustion of fossil fuels emits 14 C-depleted carbon into the atmosphere (i.e., dilutes the proportion of 14 C relative to 12 C). In contrast, nuclear weapons testing doubled the 14 C content of CO 2 in the Northern Hemisphere in the mid-20th century, followed by mixing of this bomb-derived 14 C-enriched carbon into the ocean and biosphere. Similarly, aerosol stable isotope ratios provide insight to different types of anthropogenic sources (e.g., combustion of solid and liquid vs. gaseous fossil fuels).
However, 13 C data cannot distinguish emissions from mixed fossil fuel combustion and live C 3 plant biomass (Huang et al., 2006;Winiger et al., 2016). Thus, isotope-based source apportionment studies become particularly insightful when both 14 C and stable carbon isotopes are considered (Andersson et al., 2015;Winiger et al., 2016Winiger et al., , 2017 or when combined with analyses of specific source tracers, such as levoglucosan or potassium for wood burning emissions (Szidat et al., 2006;Zhang et al., 2008) and/or remote sensing data and modeling analysis (Barrett et al., 2015;Mouteva et al., 2015b;Wiggins et al., 2018).
The objective of this study is to evaluate the effectiveness of separating OC and EC via the ECT9 (EnCan-Total-900) protocol (Huang et al., 2006;Chan et al. 2010Chan et al. , 2019 for 14 C-based source apportionment studies of carbonaceous aerosols. The ECT9 technique was originally developed to physically separate OC and EC mass fractions for concentration quantification and stable carbon isotope analysis. This protocol has been used since 2006 to monitor carbonaceous aerosol mass concentrations and stable isotope composition over Canada, including in the Arctic at Alert, as part of the Canadian Aerosol Baseline Measurements (CABM) Network by Environment and Climate Change Canada (Chan et al., 2010(Chan et al., , 2019Eckhardt et al., 2015;Sharma et al., 2017;Xu et al., 2017;Leaitch et al., 2013Leaitch et al., , 2018Huang, 2018). It has also been used to monitor carbonaceous aerosol over China (Yang et al., 2011a, b). Furthermore, EC concentration measurements made with the ECT9 protocol correlate well with those derived from light absorption by an aethalometer and refractory black carbon (rBC) using a Single Particle Soot Photometer (SP2) Chan et al., 2019). It was demonstrated that the ECT9 protocol can be used to quantify OC/EC concentrations and provide source information at the same time.
The ECT9 protocol is a thermal evolution analysis (TEA) protocol that is different from commonly used thermal optical analysis (TOA) methods for monitoring air quality, such as the Interagency Monitoring of Protected Visual Environments (IMPROVE) protocol (Chow et al., 2001;Watson et al., 2007), the National Institute for Occupational Safety and Health protocol (NIOSH method 5040, Birch, 2002), and the European Supersites for Atmospheric Aerosol Research (EUSAAR) protocol (Cavalli et al., 2010). In those protocols, the OC fraction is thermally desorbed from filter samples in an inert helium (He) atmosphere at relatively low temperatures, and the EC fraction is combusted at higher temperatures by introducing oxygen (O 2 ) into the He stream while the filter reflectance or transmittance for a laser signal is continuously monitored. During the analysis, a fraction of the OC may char (forming pyrolyzed OC or PyOC), causing the transmittance or reflectance to decrease. While TOA methods use the changes in laser signal to mathematically correct for PyOC within the measured EC fraction, the ECT9 protocol aims to minimize or remove PyOC, together with carbonate carbon (CC), during an intermediate temperature step of 870 • C in pure He via high-temperature evaporation (Chan et al., 2019). With much longer retention times at each temperature step (see Sect. 2) and without either reflectance or transmittance used, the ECT9 protocol effectively isolates OC, PyOC + CC, and EC.
It should be noted that other methods have been also developed mainly for 14 C analysis of OC and EC, such as the CTO-375 (Zencak et al., 2007), the Swiss_4S protocol (Mouteva et al., 2015a;Zhang et al., 2012;Szidat et al., 2004), or hydropyrolysis (Meredith et al., 2012;Zhang et al., 2019), which use distinct temperature protocols and gas mixtures and/or remove water-soluble OC or inorganic carbon prior to EC analysis. In contrast to the ECT9 protocol, however, these approaches differ substantially from the protocols that are widely used for monitoring OC/EC mass concentrations in the field, which limits the relevance of this data for improving the representation of carbonaceous aerosols in chemical transport models.
Here we analyzed the 14 C content of OC and EC fractions (< 100 µg C) isolated with the ECT9 protocol from four pure fossil and contemporary reference materials. These materials were analyzed on their own to quantify the amount and source (modern or fossil) of extraneous carbon introduced by the procedure, as well as its reproducibility. Mixtures of two reference materials were measured to elucidate how efficiently the ECT9 protocol isolates OC from EC. In addition, we investigated the laser signals of three reference materials and three aerosol samples (tailpipe emissions, ambient aerosol from Alert, and SRM8785) to assess how efficiently the ECT9 protocol removes PyOC. Our evaluation of the ECT9 protocol on its ability to physically separate OC from EC for 14 C-based source apportionment studies significantly expands the existing opportunities for characterizing and monitoring sources of carbonaceous aerosol at regional or global scales while at the same time providing a solid base for EC and OC concentration measurements.

The ECT9 protocol for the physical separation of OC and EC
The ECT9 protocol was developed at the carbonaceous aerosol and isotope research (CAIR) lab of Environment and Climate Change Canada (ECCC) to quantify the amount of OC and EC in carbonaceous aerosol and their δ 13 C values (Huang et al., 2006;Chan et al., 2010Chan et al., , 2019. Carbon fractions are isolated with an OC/EC analyzer (Sunset Laboratory Inc.) coupled to a custom-made gas handling and cryogenic trapping system for CO 2 collection from OC and EC fractions (Fig. 1a). The fractions are separated from each other, according to their degree of refractoriness. Specifically, carbon fractions are released by the ECT9 protocol in three steps ( Fig. 1b): (1) OC at 550 • C for 600 s in pure He (99.9999 % purity), (2) PyOC and CC at 870 • C for 600 s in pure He, and (3) EC at 900 • C for 420 s in a mixture of 2 % O 2 with 98 % He. All fractions are fully oxidized to CO 2 by passing through a furnace containing MnO 2 maintained at 870 • C. For concentration determination, the CO 2 passes through a methanator at 500 • C, is converted to CH 4 , and is quantified with a flame ionization detector. For isotope analysis, the CO 2 is cryo-trapped with liquid N 2 (−196 • C) in a U-shaped glass trap, purified on a vacuum system (to remove He), sealed into a Pyrex ampoule, and analyzed for its δ 13 C ratio with an Isotopic Ratio Mass Spectrometer (IRMS), i.e., MAT253 or F 14 C with an Accelerated Mass Spectrometer (AMS).

Reference materials and their composition
To evaluate the ECT9 method for separating OC and EC for 14 C analysis, we isolated and measured the 13 C and 14 C content of the OC or EC fraction or total carbon (TC) from five to six modern or fossil reference materials (Table 1), including two pure OC (adipic acid, sucrose), two EC (C1150, regal black), and two natural OC/EC mixtures (rice char and urban dust SRM1649a). Some of the reference materials have previously been utilized to compare different protocols that quantify OC/EC fractions (Hammes et al., 2007;Willis et al., 2016) and to determine the mass of extraneous carbon introduced during OC/EC isolation from carbonaceous aerosol (Mouteva et al., 2015a). Table 1 provides an overview of their chemical compositions, i.e., total carbon contents and relative fraction of OC and EC, respectively (for individual measurements, see Table S1 in the Supplement). Primary methods (i.e., gravimetric or volumetric) are used for mass loading of the materials, whereas the mass of TC, OC, and EC are quantified via the ECT9 thermal protocol. Based on repeat injections of sucrose results (20-80 µg sucrose, n = 117), the accuracy of the TC mass is about 5 %. The reproducibilities of both OC/TC and EC/TC percentages are 2 % or better. Although uncertainties of weighing pure EC mass (i.e., regal black and C1150) via microbalances are relatively large (due to static electricity and variable relative humidity), the EC/TC and OC/TC ratios for all reference materials are highly reproducible (one SD < 2 %). The results show that the two EC materials (i.e., regal black and C1150) contain 97 % and 98 % EC, with only 3 % and 2 % OC, respectively. The two OC materials (i.e., sucrose and adipic acid) are 99 % and 100 % OC and less than 1 % EC (likely due to charred OC contribution), respectively. Thus, the materials are suitable for the purpose of this study.
We also analyzed the 13 C and 14 C isotopic composition of each reference material, using offline combustions and ECT9 coupled with cryo-purification to convert them into CO 2 . The results are summarized in Table 2 (for individual results see  Tables S2 and S3 in the Supplement). The 14 C analysis of µg C-sized carbonaceous aerosol samples requires the as- Figure 1. Overview of the carbonaceous aerosol measurements (concentration and isotopes) via EnCan-Totol-900 (ECT9). (a) Schematic procedures for 13 C and 14 C measurements of OC/EC via ECT9, including (1) OC/EC isolation and CO 2 collection via cryo-trapping, (2) CO 2 purification, and (3) isotope analysis with IRMS ( 13 C/ 12 C of CO 2 ) or AMS ( 13 C/ 12 C and 14 C/ 12 C of graphite targets). (b) Thermogram of the ECT9 protocol on a Sunset OC/EC analyzer. First, organic carbon (OC) is thermally desorbed at 550 • C for 600 s in 100 % He, and then any pyrolyzed OC (PyOC), refractory OC, and carbonate carbon (CC) is released at 870 • C in 100 % He for 600 s. Finally, elemental carbon (EC) is combusted at 900 • C for 420 s by introducing 2 % O 2 in He. All carbon fractions are oxidized to CO 2 followed by reduction to CH 4 and quantification via flame ionization detection (FID) for carbon content or purified and cryo-trapped in Pyrex ampoules for isotope analysis. Example FID signals are shown for a pure OC reference material (sucrose) mixed with a pure EC material (regal black) and the internal standard (CH 4 ). The total of 101 % is obtained from the ratio of TC measured to TC calculated from the injected solution of sucrose; b A total of 49 % of TC to bulk material in adipic acid based on its molecular mass. c A total of 58.6 % of TC to bulk material in rice char obtained from Hammes et al. (2006). d A total of 17 % of TC to bulk material in SRM 1649a obtained from a critical evaluation of inter-laboratory data by Currie et al. (2002). sessment of extraneous carbon . This is achieved by measuring multiple smaller-sized materials with known 14 C content. Consequently, the results in Table 2 are critical, as those 14 C values provide the reference for quantifying the extraneous carbon introduced during the isotope analysis procedures.

Isolation of OC, EC, or TC with the ECT9 protocol and purification of CO 2
The isotopic analysis of carbonaceous aerosol via the ECT9 system involves three steps ( Fig. 1a): (1) OC and EC isolation and CO 2 collection, (2) CO 2 purification, and (3) isotope analysis for either 13 C/ 12 C by IRMS or 14 C by AMS (i.e., coupled measurements of 13 C/ 12 C and 14 C/ 12 C of µg Csized graphite targets), as desired. The initial masses of the pure reference materials ranged from 5 to 47 µg C (n = 3-13; Table S6 in the Supplement), whereas those for the mixed materials ranged from 5-30 µg C for OC and 5-60 µg C for EC (n = 5-6; Table S7 in the Supplement). The loaded mass of each material was determined via a microbalance (MX5, Mettler Toledo or CCE6, Sartorius) with the lowest reading being 1 or 0.1 µg C, respectively. Filters before mass loading were pre-combusted at 900 • C in a muffle furnace overnight and wrapped into pre-fired aluminum foil before cooling below 200 • C. Usually, OC materials were first dissolved in Milli-Q water with known volume to obtain its concentration, and then a known amount (5-10 µL) of OC solution was very carefully applied onto a pre-cleaned quartz filter surface (1.5 cm 2 , Pall Canada Limited) via a syringe injection. After the injection, the quartz boat holding the punch is pushed to the right position inside of the analyzer. The volume of OC solution used does not saturate the filter but merely moistens the surface. After purging the filter for about 20 min ensuring the water is gone, the filter is ready for analysis. EC (i.e., regal black and C1150) and mixed materials (rice char or SRM 1649a), which cannot be completely dissolved in water, were directly weighed onto pre-cleaned quartz filter punches in the form of solids (powders). Adipic acid was also loaded as powder. The final power mass was determined by the difference weighted before and after analysis. A filter punch with the loaded mass was carefully carried to the Sunset analyzer by a Pyrex glass Petri dish with a cover for analysis with the ECT9 protocol.
OC and EC were separated and the combusted OC or EC fractions (as CO 2 ) were cryo-collected in a U-shaped flask submerged in liquid N 2 (Fig. 1a, step 1). Following this, the flask containing CO 2 and He was connected to a vacuum line with four cryo-traps and several open ports (Fig. 1a, step 2), where the CO 2 is purified by sequential distillation when passing cryo-traps 1 through 3. Finally, the pure CO 2 is transferred and sealed into a 6 mm glass ampoule for 13 C or 14 C analysis. Pressure is read by a Pirani gauge before sealing the ampoule for an estimation of the amount of gas, and consequently sample size could be determined as µg C. Table 2. Overview of the isotopic composition of the reference materials used in this study. Radiocarbon ( 14 C/ 12 C, reported as fraction modern, FM 14 C) was measured at the KCCAMS facility, and δ 13 C was measured at the CAIR lab. CO 2 extraction and 13 C/ 12 C analysis Sample CO 2 is cryogenically purified on a vacuum line and sealed into an ampoule for analysis with a MAT253 Isotopic Ratio Mass Spectrometer .
n/a * Sucrose was loaded as a solution (µg C), and regal black, C1150, adipic acid, rice char, and SRM-1649a were loaded as a fine powder (µg dry mass); n/a stands for not applicable

14 C measurements
At the KCCAMS facility, the OC and EC fractions or TC (in form of CO 2 ) were reduced to graphite on iron powder via hydrogen (H 2 ) reduction using equipment and protocols specifically developed for smaller-sized (≤15 µg C) samples (Santos et al., 2007a, b). Briefly, sample-CO 2 was introduced into a vacuum line, cryogenically isolated from any water vapor, monometrically quantified, and then transferred to a heated reaction chamber, where it was mixed with H 2 and reduced to filamentous graphite. To characterize the graphitization, handling, and AMS analysis, two relevant standards (oxalic acid II as modern carbon and adipic acid as fossil carbon), with similar size ranges to the samples prepared via ECT9, were also processed into graphite. The graphite was then pressed into aluminum holders and loaded into the AMS unit alongside measurement standards (Table S6) and blanks for 14 C measurement (Beverly et al., 2010). The data are reported as a fraction of modern carbon (F 14 C), following the conventions established by Stuiver and Polach (1977) and also described elsewhere (Reimer et al., 2004;Trumbore et al., 2016).
To establish consensus values (Table 2), we also analyzed the 14 C content of the bulk reference materials ranging in size from 0.06 to 1 mg C, using our standard combustion and graphitization methods. Larger aliquots of material were weighed into a pre-combusted quartz tube with 80 mg CuO, evacuated, and combusted at 900 • C for 3 h. The resulting CO 2 was cryogenically purified on a vacuum line, reduced to graphite using a closed-tube zinc-reduction method (Xu et al., 2007), and analyzed as described above.

Quantification of extraneous carbon
Any type of sample processing and analysis introduces extraneous carbon (C ex ). Therefore, the measured mass of any sample will include the mass of this sample and of any C ex incorporated throughout the analysis (Eq. 1): where m spl_meas , m spl , and m ex are the measured and theoretical mass of the sample and of C ex , respectively. For small samples (with a mass of a few µg C), the mass of C ex can compete with or overwhelm the sample mass and cause the measured F 14 C value of a sample to deviate from its consensus value.
Here, we estimated the mass of C ex introduced during the ECT9 protocol and the 14 C analysis following Santos et al. (2010), where C ex is understood to consist of a modern and of fossil component (Eq. 2): where m mex and m fex is the mass of the modern and fossil C ex , respectively.
Following an isotope mass balance approach, the measured isotopic ratio ( 14 C/ 12 C) of a sample (R spl_meas ) can be expressed as Eq. (3).
where R spl is the theoretical isotopic ratio of the sample, and R m and R f are the consensus isotopic ratios of a modern and fossil standard, respectively. This equation can be further simplified because R f is 0. R m is determined by measuring regular-sized aliquots of this reference material. In addition, all 14 C/ 12 C ratios are corrected for isotope fractionation using their δ 13 C measured alongside 14 C on the AMS (Beverly et al., 2010). The mass of modern C ex can be quantified by analyzing fossil reference materials, which are highly sensitive to modern pollutants and insensitive to fossil pollutants. Based on Eq.
(3), the measured isotopic ratio of the fossil reference (R f_meas ) can be expressed as Eq. (4): The smaller the mass of the fossil reference material, the greater the effect of the constant mass of modern C ex on the isotope ratio of the fossil reference material, i.e., R f_meas deviates toward R m .
Similarly, the mass of fossil C ex can be quantified by analyzing modern reference materials. With decreasing mass, the measured isotopic ratio of the modern reference (R m_meas ) will deviate more strongly from R m (toward R f ). Based on Eqs. (1)-(3) and assuming m spl m mex , the R m_meas can be expressed as Eq. (5): Finally, we can calculate the C ex -corrected isotope ratio of an unknown sample (F spl_cor ). This value is reported as the ratio between the theoretical isotopic ratio of this sample and the accepted value of a modern standard (R/R m ), also known as "fraction modern" (F ; with all R corrected for stable isotope fractionation). This data are reported as Eq. (6): where F m * is determined from the direct measurement of the modern primary reference material (OX1) used to produce six time-bracketed graphite targets measured in a single batch after isotopic fractionation correction and normalization (Santos et al., 2007a, b). The individual uncertainty of F spl_cor is determined from counting statistics and by propagating the quantified blanks using a mass balance approach. Long-term and continuous measurements of various types of blanks indicate that the mass of C ex within one analytical method or system can vary as much as 50 % (see Santos et al., 2010; Fig. 1). Therefore, we applied a 50 % error in m fex and m mex from long-term measurements of variance in m ex of small samples (Santos et al., 2007a).
In this study, we used a multi-step approach to quantify m ex introduced by the ECT9 protocol and 14 C analysis (i.e., graphite target preparation for CO 2 sample plus AMS analysis). First, we quantified m ex introduced during 14 C sample preparation and analysis by analyzing different masses of our bulk reference materials without involving ECT9 protocol. Extraneous carbon is introduced during sealed tube combustion and graphitization followed by graphite target handling and AMS measurement at the KCCAMS facility. Typically, 14 C sample preparation and AMS measurement contributes a small portion to m ex (Mouteva et al., 2015a;Santos et al., 2010). Second, we quantified the portion of m ex added during the isolation of OC and EC with the ECT9 protocol. This portion of m ex allows us to determine the practical minimum sample size limit for the entire method, including m ex contributions from filter handling before OC/EC analysis, instrument separation, and transfer to the cryo-collection system and Pyrex ampoules. To isolate this portion, we quantified m ex of the entire procedure (ECT9 protocol plus 14 C analysis) by analyzing the 14 C signature of OC and EC from different masses of a large set of reference materials and then subtracted the portion of m ex introduced during 14 C analysis.

Recovery estimation
The reference materials used in this study, including the modern and fossil end-members (i.e., the major carbon sources) found in carbonaceous aerosol and their TC, OC, and EC concentrations are shown in Table 1. Reference materials were separated into OC, EC, or TC using the ECT9 method at ECCC's CAIR lab (Fig. 1) and analyzed for their 14 C content at UC Irvine's KCCAMS facility, including graphitization and AMS analysis. Figure 2 shows the cross-validation of carbon mass between the mass determined at ECCC's CAIR lab and the mass quantified at UC Irvine's KCCAMS lab indicating a very good positive correlation (R 2 = 0.93 for pure materials and R 2 = 0.95 for two-material-mixtures in Fig. 2a and b, respectively). Reassessment of sample masses by manometric measurements at UC Irvine (UCI) show good agreement with initial mass loaded at ECCC's CAIR lab via gravimetric or volumetric methods ( Fig. 2a and b and Tables S6 and S7). It is suggested that no major losses or gains of carbon occurred Figure 2. Cross-validation of carbon mass processed by the ECT9 protocol and collected via cryo-trapping at ECCC and the carbon mass retrieved during the purification and graphitization on a KCCAMS vacuum line. Carbon fractions (OC, EC, or TC) were isolated from two reference materials for OC (sucrose, adipic acid) and for EC (regal black, C1150) and one OC and EC mixture (rice char). Most of the points deviating from the 1 : 1 line are carbon-rich reference materials, e.g., regal black and C1150 (> 90 % TC), wherein there are usually greater uncertainties in initial mass determination via weighing using microbalance. and (e) rice char is a mixture of modern OC and EC. Open and solid symbols represent 14 C data before and after correction for extraneous carbon introduced during OC/EC isolation and subsequent 14 C analysis, respectively. The dashed line indicates the consensus value determined from regular-sized bulk samples of these materials undergoing offline combustions (see Table 2).  Mouteva et al., 2015a).
during the entire analytical process and the overall recovery was close to 100 %, with a 5 % uncertainty for samples ranging in size from about 5 to 60 µg C.

Quantification of extraneous carbon and its sources
All types of samples, regardless of size, show deviations in their measured F 14 C value from their consensus values to certain degree due to C ex introduced during sample analysis. In µg C-sized samples (mass < 15 µg C), significant bias from any C ex can be observed because C ex constitutes a large fraction of the total sample. Previous work (using solventfree analytical protocols) has shown that modern C ex is introduced mostly through instrumentation and sample handling techniques, while fossil C ex originates from iron oxide used as a catalyst for the reduction of CO 2 to graphite prior to AMS analysis (Santos et al., 2007a, b).
The F 14 C values of the pure modern or fossil reference materials generally agreed with their accepted F 14 C values for both OC and EC fractions (within approximately 5 % uncertainty on average; see Fig. 3 and Table 2, S6, and S7) after applying a constant amount C ex correction in F 14 C determination. Specifically, the overall agreements for all individual pure (Table S6) and mixed reference materials (Table S7, excluding the OC data from adipic acid + bulk rice char) are within 2 ± 3 % of their corresponding values (Table 2). On average, for samples containing > 10 µg C the agreements are within 1 ± 1 %, whereas samples containing between > 5 µg C and < 10 µg C are around 7 ± 5 %. This constant C ex is a critical prerequisite for accurately correcting the F 14 C value of unknown samples. Hence, our data demonstrated that the ECT9 method (and subsequent 14 C analysis) introduces a small, reproducible amount of C ex .
According to Eqs. (4) and (5) in Sect. 2.5, C ex can be quantified by measuring F 14 C of pure modern or fossil materi-als with different sizes. Figure 3 demonstrates that regardless what 14 C content is in carbon fractions isolated from the reference materials and what size it is, the corrected F 14 C values match with the consensus value within propagated uncertainty.
To evaluate the suitability of ECT9 for 14 C analysis of aerosol samples, a comparison is made between the results of a published method (i.e., Swiss_4S) and those of ECT9. The two protocols are listed in Table 3, and their C ex distribution is shown in Table 4. The total amount of C ex introduced by the complete procedure through ECT9 and determined based on all reference materials was 1.3 ± 0.6 µg C, with 70 % originating from contamination with modern carbon (Table 4). The isolation of OC and EC with the ECT9 protocol introduced 65 % of total C ex (0.85 out of 1.35 µg C), with 65 % derived from modern carbon. Overall, the total amount of C ex introduced during OC/EC isolation with the ECT9 protocol is comparable to that for the Swiss_4S protocol established at UCI within uncertainties (Table 3, Mouteva et al., 2015a). Thus, it is demonstrated that the ECT9 protocol serves as a suitable alternative for the 14 C analysis of aerosol samples with masses > 5 µg C.

Effectiveness of OC/EC separation
To investigate the effectiveness of the ECT9 for separating OC from EC in more complex mixtures while minimizing OC into the EC fraction via pyrolysis, mixtures of the modern and fossil reference materials (Table 2) were used for measuring δ 13 C (Tables S4 and S5 in the Supplement) and F 14 C (Table S7).
First, it was found that the F 14 C values of OC and EC fractions isolated from mixtures of pure sucrose (modern OC) and pure regal black (fossil EC) were within the measurement uncertainty of their accepted F 14 C values after correc- . Radiocarbon ( 14 C) composition, expressed as fraction of modern carbon, of (a) organic (OC, triangles) or (b) elemental (EC, squares) carbon fractions isolated with the ECT9 protocol from mixtures of pure modern OC (sucrose) with fossil EC (regal black). Open and solid symbols represent 14 C data before and after correction for extraneous carbon introduced during OC/EC isolation via ECT9 and subsequent 14 C analysis via AMS, respectively (see Table S7). The dashed line indicates the consensus value (see Table 2). Table 4. Comparison of the procedural contamination with extraneous carbon for aerosol reference materials partitioned into organic carbon (OC) and elemental carbon (EC) with the ECT9 or Swiss_4S protocols based on their 14 C contents. We assume a measurement uncertainty of 50 % (see Sect. 2). tion for a constant amount of C ex (Fig. 4) for samples with 5-34 µg OC carbon and 10-60 µg EC carbon, showing a good separation of OC from EC. This amount of C ex was identical to that applied to the pure reference materials above, further corroborating the constant background introduced by the ECT9 protocol and 14 C analysis. Next, the mixtures of fossil adipic acid (pure OC) and modern rice char (mixture of OC and EC) were isolated and analyzed. It was found that after correction for C ex , the F 14 C values of the OC (from the mixture) were systematically greater than the consensus value of the pure adipic acid, i.e., a F 14 C of zero (Fig. 5a), indicating that there was certain level of modern fraction contributed to the measured OC from the modern rice char. Based on an elevated mean value of 0.1081 ± 0.0259 (n = 6) after blank corrections, a mass balance calculation indicates that 10 ± 3 % of rice char OC is present. The high end of this estimation is close to ∼ 14 %, within a validity range of what one would expect.
To confirm that ECT9 could remove OC contained in rice char, an additional step was taken before mixing modern rice char's EC with the fossil OC (adipic acid). Specifically, we stripped the OC fraction of rice char by running rice char (on a filter) through the ECT9 protocol. Adipic acid (fossil OC) was then injected onto the filter with the remaining rice char EC. The results show that the F 14 C of OC values of this mixture lie well within the expected range of the consensus value (Fig. 5b) after a C ex correction as described above, demonstrating an excellent removal of rice char OC.
In both mixtures (fossil adipic acid with modern bulk rice char or rice char EC), the corrected F 14 C values of the isolated EC fractions were within the expected range for the rice char reference material ( Fig. 5c and d). This provides further evidence that the ECT9 protocol isolates modern EC from fossil OC with no obvious evidence of transferring fossil OC into the EC fraction. Together, the three sets of mixing experiments (Figs. 4 and 5) provide strong evidence for the effectiveness of separating OC from EC via ECT9 protocol.
In addition to F 14 C measurements, δ 13 C measurements in mixtures of OC and EC can also provide quantitative information on the effectiveness of OC and EC separation via ECT9. Various amounts of sucrose (pure OC, 10-30 µg C) were first mixed with varying amounts of regal black (pure Open and solid symbols represent data before and after correction for extraneous carbon introduced during OC/EC isolation via ECT9 and subsequent 14 C analysis via AMS, respectively (Table S7). The dashed line indicates the consensus value (see Table 2). EC, 20-66 µg C). The mixtures were then physically separated into OC and EC fractions by ECT9 for δ 13 C measurements. The measured δ 13 C values of OC and EC from these mixing experiments are listed in Table S4. Based on the δ 13 C values of individual pure reference materials (Table S3) and a two end-member mixing mass balance, it is estimated that the average fraction contributed into each other in the mixtures (i.e., sucrose fraction into regal black or vice versa) was likely less than 3 % (Table S5).

Charring evaluation and PyOC removal using the ECT9 protocol
It is known that some of OC (e.g., oxygenated OC or watersoluble OC) would char to form pyrolyzed organic carbon (PyOC) when heated in an inert He atmosphere, darkening the filter (Chow et al., 2004;Watson et al., 2007) and causing decreased laser signals due to light absorption of charred OC. In most TOA protocols, this PyOC would combust and contribute to EC when O 2 is added. However, PyOC can be also be gasified and released as CO at high temperatures (> 700 • C) with limited O 2 supply, e.g., oxygenated OC at 870 • C (Huang et al., 2006;Chan et al., 2010Chan et al., , 2019. Most TOA protocols estimate PyOC by quantifying the mass associated with reflectance and transmittance changes, i.e., the mass released between the time when O 2 is introduced and the OC/EC split point (where the reflectance and transmittance returns to the initial value). In contrast to other TOA protocols, ECT9 defines PyOC as the mass released at the temperature step of 870 • C (during a period of 600 s). This includes charred OC, calcium carbonate (CaCO 3 ) that decomposes at 830 • C, and any refractory OC not thermally released at 550 • C (Huang et al., 2006;Chan et al., 2010Chan et al., , 2019. Although ECT9 does not use laser signals to quantify PyOC, it is expected that the changes of laser signals during the stage of 870 • C would provide useful information about PyOC. Thus, four sets of samples were selected, including Figure 6. Thermograms of pure or bulk references: (a) regal black, (b) sucrose, and (c) rice char. Temperature (solid blue line) and FID signals (integrated yellow area with green line) on the left axes and laser (solid red line) on the right axis. It is observed that on the three thermograms during the temperature stage of 870 • C, the laser transmittance signals decrease first and increase again before the next temperature stage, minimizing PyOC fraction, i.e., possible charred OC contribution to EC. those of pure reference materials and ambient aerosol samples from different sources with heavy or light mass loading (e.g., those arctic sample filters from different seasons) to evaluate the possible charring via ECT9. Their thermograms are shown in Figs. 6 to 9. Figure 6a-c show thermograms of pure or bulk references for regal black, sucrose, and rice char, respectively. It is observed in all three panels that the laser transmittance signals first decrease and then increase again during the 870 • C step and that they return to their initial values just before EC is released at the next step of 900 • C. This demonstrates that the ECT9 method minimizes PyOC contributions to the EC fraction.
The thermograms of aerosol (on filters) collected directly from tailpipe exhaust of a diesel engine vehicle and a gasoline engine passage car, respectively are shown in Fig. 7. These data suggest that the amount of PyOC generated during analysis are sample/matrix dependent. Specifically, the Figure 7. Thermograms of the filters directly collected from tailpipe exhaust of a diesel engine vehicle in (a) and a gasoline engine passage car in (b). The legends are the same as Fig. 6. Note that the mass fraction from the temperature stage of 870 • C in (b) is obviously larger than that in (a). The latter is negligible, indicating that the amount of PyOC fraction is sample matrix dependent. The amount of PyOC from gasoline vehicle emissions is likely larger than that from diesel vehicle emissions. Note that the laser signal reaches the initial value before the 900 • C stage for EC releasing, demonstrating that the charring contribution to EC is minimized. mass fraction during the 870 • C temperature is larger for the gasoline than the diesel engine. This finding supports previous work showing that PyOC is proportional to the amount of oxygenated OC (Chan et al., 2010). It is noticed that the laser signal reaches the initial value before the EC step, further demonstrating that the charring contribution to EC is minimized.
Another set of thermograms of two total suspended particle filter samples collected during the summer (August) and winter (December) of 2015 at an arctic site (i.e., Alert) are shown in Fig. 8. More details about these samples can be found in Wex et al. (2019). The laser signal patterns are similar to those shown in Figs. 6 and 7, yet more pronounced. During the 550 • C step, the laser signals decrease. During the 870 • C step, the signals further decrease, then increase, and finally increase to their initial point before EC is released at 900 • C. These thermograms further demonstrate ECT9 is able to minimize PyOC by gasification.
Finally, the thermographs of NIST urban dust reference material SRM 8785 (the resuspended SRM 1649a urban dust with a fine fraction < 2.5 µm collected on quartz filter) analyzed with ECT9 and Swiss_4S are shown in Fig. 9. Both thermograms obtained with the ECT9 method ( Fig. 9a and b) show the similar patterns as those in Figs. 6-8, i.e., the laser signals reaching the initial value just before the EC release at 900 • C, suggesting that the charring contribution to EC is minimized during the stage of 870 • C even though some PyOC might remain.
In the thermogram obtained with the Swiss-4S protocol (Fig. 9c), the laser signal increases from the beginning of the run while the first two stages (375 and 475 • C) are under the conditions of pure O 2 stream, inferring that light absorbing carbon is released during the first two OC stages. The laser signal continues to increase while the temperature increases up to 650 • C (the third stage) under the pure He gas stream, indicating that no charred OC is formed. However, when the temperature starts decreasing from 650 • C, the laser signal decreases, indicating PyOC formation below that temperature. This signal decrease continues until the beginning of the next pure O 2 stage. It is important to note that to ob-  Fig. 6. It is clearly shown on both thermograms that during 550 • C stage, the laser signal starts decreasing (implying charred OC formation) and begins increasing at 870 • C and reaches the initial value before the EC stage (indicating the contribution to EC by charred OC is minimized or removed). tain EC fraction, the Swiss-4 (Table 3) method calls for filter sample pre-treatment, i.e., extraction with water before the thermal separation of OC/EC to minimize the contribution of charred OC from the 3rd stage to EC at the 4th stage . However, for a method comparison, the thermogram shown in Fig. 9c was from a filter without pretreatment. While it is difficult to make direct comparisons between OC and EC from Fig. 9a and c, the laser profiles from those thermograms in Fig. 9a and b indicate that in both cases charred OC is negligible or minimum via ECT9.
Together, the thermograms (Figs. 6-9) elucidate that the ECT9 protocol can effectively remove or minimize charred OC (PyOC) to achieve good physical separation of OC and EC. Another great advantage of using ECT9 to separate OC from EC for isotope analysis (both 13 C and 14 C) is its consistency with the protocol used for OC and EC concentration measurements. Moreover, the ECT9 method does not require filter samples to be pre-extracted with water before EC analysis (to reduce PyOC).

Conclusions
We demonstrate the effectiveness of the ECT9 protocol to physically isolate OC and EC from aerosol samples for 14 C and 13 C analysis by using OC and EC reference materials on their own and as mixtures. It was found that the ECT9 protocol successfully separates OC and EC fractions with a low (but largely modern) total carbon blank of 1.3 ± 0.6 µg C. The majority (65 %) of this extraneous carbon originates from the isolation with the ECT9 protocol, with 35 % contributed from graphitization and 14 C measurement of the samples at the KCCAMS facility. After mass balance background corrections, the F 14 C results from both bulk pure materials and mixtures (with sample size as small as 5 µg C) can Figure 9. Thermograms of the SRM 8785 filters (the fine fraction (PM 2.5 ) of resuspended urban dust particles from SRM 1649a and collected on quartz filters) with various amount of materials ranging from 614 to 1723 mg via two different thermal protocols. Panels (a) and (b) were obtained by ECT9. The legends are the same as Fig. 6. Thermograms in (a) and (b) show the similar patterns as in Figs. 6-8 that the laser signals reaching the initial value are just before the temperature stage of EC, suggesting that the charred OC contribution to EC is minimized. The thermogram in (c) is obtained from the same filter in (b) but by Swiss-4 protocol for comparison. The legends are similar except for the integrated area with green line, which stands for CO 2 in parts per million (by non-dispersive infrared) instead of FID signals. reach the consensus values (Table 2) with an average uncertainty of about 5 %.
In addition, we evaluated potential PyOC formation during ECT9 by investigating thermograms of a variety of reference materials and ambient filter samples. It is demonstrated that ECT9 provides a good alternative for carbonaceous aerosol source apportionment studies, including ultra small sized (5-15 µg C) samples obtained from arctic regions. To increase the application of isotope data ( 14 C or 13 C) in atmospheric research, future efforts should be focused on the comparison on OC/EC separation via different methods/protocols using the same sets of reference materials. At the same time, the isolation results should be also compared among those meth-ods/protocols widely used in long-term national monitoring network for OC/EC contents, ensuring a consistency in measurements between OC/EC concentrations and their corresponding isotopic compositions. Author contributions. Conceptualizing and designing the study: LH, CIC, and GMS. Developing analytical methods and ensuring data quality: LH, GMS, WZ, CIC, BTR. Performing the experiments and data acquisition: WZ, GMS, SRH, VV, BTR. Data organizing/analysis and interpretation: LH, CIC, BTR, GMS, WZ. Writing the paper, including editing and preparing figures and tables: LH, CIC, BTR, GMS, WZ.
Competing interests. The authors declare that they have no conflict of interest.