We have modified a commercial thermal–optical transmittance (TOT) instrument (Sunset Lab Inc.). This equipment had been originally designed (Birch and Cary, 1996) with a red laser diode (λ=635 nm) to have the possibility to monitor and correct the well-known problem of the formation of pyrolytic carbon by charring (Birch and Cary, 1996; Bond and Bergstrom, 2006; Chow et al., 2007; Cavalli et al., 2010). The assumption that OC is optically inactive at wavelengths greater than 600 nm is at the basis of the technique; therefore the laser beam attenuation is only due to the EC originally present or formed by charring in the sample under analysis. Actually, even at this wavelength, BrC can affect the reliability of the OC/EC separation and the standard methodology can be modified to quantify the BrC concentration (Massabò et al., 2016). Nevertheless, at λ=635 nm the BrC mass absorption coefficient, MAC(BrC), remains much smaller than the corresponding MAC(BC) and the modified procedure could/should be implemented at shorter wavelengths to gain in sensitivity.

We have modified our Sunset unit by making possible the alternative use of the standard laser diode at λ=635 nm or of a World Star Technologies, 100 mW, laser diode at λ=405 nm. This second laser diode can be mounted on the top of the Sunset furnace by a homemade adapter (see Fig. 1) and easily exchanged with the native red diode. With the new laser diode, the light detector placed at the bottom of the Sunset furnace has to be changed too and we selected a photodiode (PD) Thorlabs FDS1010 coupled with a bandpass filter Thorlabs FBH405-10. The responsivity of the PD FDS1010 around λ=400 nm is quite low (about 50 mA W-1) but the high power delivered by the laser diode results in signals with an amplitude comparable to the values measured with the original Sunset setup (i.e., laser diode and PD). Furthermore, the FBH405-10 filter cuts all the light background produced by the high temperature of the Sunset furnace, thus preserving the signal-to-noise ratio. Both laser and PD can be exchanged in about 10 min and no particular attention is requested other than the proper alignment to maximize the PD output signal (i.e., the transmittance value displayed by the Sunset control software). We have to note that the original configuration of the Sunset instrument adopts a lock-in amplifier to improve the signal-to-noise ratio of the PD: we did not have the possibility to manipulate the parameters of the lock-in amplifier and to tune it to the new configuration.

The new blue-light setup of the Sunset analyzer was tested using both synthetic and real-world aerosol samples, collected on quartz-fiber filters. Synthetic samples were prepared starting with a 5 % (volume) solution of Aquadag, then nebulized by a Blaustein atomizer (BLAM), and collected on quartz-fiber filters. Aquadag is the trade name of a water-based colloidal graphite coating (particle diameters between 50 and 100 nm): these samples can therefore be considered to be composed of EC–BC only. The samples were first sent to an optical characterization by the MWAA instrument (Massabò et al., 2015), which demonstrated that the optical absorption of Aquadag is independent of the wavelength. Actually, Aquadag particles tend to form conglomerates on the filter surface, with dimensions about double the longer wavelength implemented in the MWAA (i.e., 850 nm of the infrared laser diode; Massabò et al., 2015). So, the comparison between the new blue-light and original Sunset setups was made with samples having the same absorption properties. EC and TC quantifications obtained at λ=635 and λ=405 nm were in excellent agreement for both the NIOSH5040 and EUSAAR_2 protocols (Cavalli et al., 2010), as shown in Fig. 2 for the whole set of synthetic samples.

A second set of synthetic samples was prepared to mimic the behavior of real-world aerosol samples: a 3 % (weight) solution of ammonium sulfate (NH4)2SO4 in Aquadag was prepared and nebulized with the BLAM. This way, a scattering compound is mixed to the absorbing Aquadag spherules. The optical absorption measured with MWAA was independent of wavelength with this second set of samples too. The results of the Sunset analysis with both the red and blue laser setups are shown in Fig. 3. This second set of samples was analyzed through the EUSAAR_ 2 protocol only: we used two punches for each laser in each sample to have a reproducibility check. A strong correlation between the TC and EC values measured in red and blue light was obtained again with a slope close to unity.

A third and final test was performed using a set of daily PM10 samples collected by a low-volume sampler (TCR – Tecora, Italy) on quartz-fiber filters (Pall-2500 QAO-UP, 47 mm diameter) in spring 2016 in the urban area of the city of Genoa (Italy). A previous and long set of similar campaigns addressing PM10 characterization (e.g., Bove et al., 2014, and references therein) in the same urban area could not identify a sizeable contribution from biomass burning to PM composition, in particular during spring and summer. Such a situation was confirmed by the determination of the Ångström exponent in the present samples by the MWAA. Actually, in the set of 20 PM10 samples, the values of the Ångström exponent ranged between 0.9 and 1.2, confirming that black carbon is the sole or totally dominant light-absorbing component in the local PM10 (Sandradewi et al., 2008; Harrison et al., 2013). Half of the samples were then sent to Sunset analysis by the NIOSH5040 protocol while the EUSAAR_2 protocol was adopted for the remaining subset of samples. The results are shown in Fig. 4. The EC concentration values measured with the standard and modified Sunset analyzers are fully compatible when the NIOSH5040 protocol is adopted (basically, the split point position in the Sunset thermogram does not change with the two laser diodes). Instead, EC values determined by the EUSAAR_2 protocol resulted in about 30 % lower values when the blue laser diode was mounted. This corresponds to a shift of the split point position, which moves rightward and thus increases the amount of carbonaceous aerosol counted in the OC fraction. This effect is linked to the well-known issue of the formation of pyrolytic carbon during the thermal cycle in the inert atmosphere (i.e., in He). Several literature studies (e.g., Cavalli et al., 2010; Panteliadis et al., 2015) indicated that the charring is smaller at the higher temperatures reached during the NIOSH thermal protocol. In the other protocol, standard thermal–optical analyses of urban PM samples often give higher EC values (up to 50 %) when performed following the EUSAAR_2 instead of higher-temperature protocols (Subramanian et al., 2006; Zhi et al., 2008; Piazzalunga et al., 2011; Karanasiou et al., 2015; Panteliadis et al., 2015). Furthermore, as a by-product of previous PM10 studies in the urban area of Genoa by a standard Sunset unit, we could observe a systematic and very reproducible 40 % discrepancy between EC values determined in the same samples by EUSAAR_2 and NIOSH5040 protocols (with EC: EUSAAR > EC: NIOSH). Therefore, the thermal–optical analysis in blue light seems to be more sensitive to the charring formation during the EUSAAR_2 protocol.

Aerosol samples were collected in a small village (Propata, 44∘33′52.93′′ N, 9∘11′05.57′′ E, 970 m a.s.l.) situated in the Ligurian Apennines, Italy. Three different sets of PM10 aerosol samples were collected by a low-volume sampler (38.3 L min-1 by TCR Tecora): the first and the third sets had a filter change set every 24h while the second set was sampled on a 48 h basis. In total, 41 (14+13+14) PM10 samples were collected on quartz-fiber filters (Pall, 2500QAO-UP, 47 mm diameter), between 2 February and 19 April 2018. Before the sampling, the filters were baked at T=700 ∘C for 2 h to remove possible internal contamination. Field blank filters were used to monitor possible contaminations during the sampling phase. Wood burning is one of the PM sources around the sampling site, especially during the cold season, as it is used for both domestic heating and cooking purposes.

All filters were weighed before and after sampling in an air-conditioned room (T=20±1 ∘C; R.H. =50%±5 %), after 48 h conditioning. The gravimetric determination of the PM mass was performed using an analytical microbalance (precision: 1 µg), which was operated inside the conditioned room; electrostatic effects were avoided by the use of a deionizing gun.

After weighing, samples were first optically analyzed by MWAA to retrieve the absorption coefficient (babs) of PM at five different wavelengths. The EC and OC determination was performed adopting the EUSAAR_2 protocol (Cavalli et al., 2010) with both laser diodes at λ=635 and at λ=405 nm (two different punches were extracted from each filter sample).

Finally, the remaining portion of the same quartz-fiber filters underwent a chemical determination of the levoglucosan (1,6-Anhydro-beta-glucopyranose) concentration by high-performance anion exchange chromatographer coupled with pulsed amperometric detection (Piazzalunga et al., 2010). As is well-known in literature, this sugar is one of the typical markers of biomass burning (Vassura et al., 2014).

The MWAA provided the raw data to measure the spectral dependence of the aerosol absorption coefficient (babs), which can be generally described by the power-law relationship babs (λ) ∼ λ-AAE, where AAE is the Ångström absorption exponent.

The time series of the resulting AAE values is shown in Fig. 5: they range between 1.05 and 1.96 with a mean value of 1.55±0.21. This figure indicates a substantial presence of wood burning in the sampling area. In Massabò et al. (2015) and Bernardoni et al. (2017), an optical apportionment model (the “MWAA model”) based on the measurement of babs at five wavelengths had been introduced to directly obtain the BrC AAE (αBrC) and the BrC absorption coefficient (babsBrC) at each measured wavelength. It is worth noting that, at the basis of the MWAA model, there is the assumption that BrC is produced by wood combustion only (see Sect. 4 in Massabò et al., 2015; Zheng et al., 2013). In Fig. 5, we report the optical apportionment at λ=635 and at λ=405 nm, i.e., at the wavelength of the two laser diodes used in our modified Sunset instrument. At λ=635 nm, light absorption resulted mainly due to BC from both fossil fuel (FF) and wood burning (WB), and the babsBrC average value is 15 % of total babs, with the notable exception of some days on which it reached values of ∼30 %, in correspondence with AAE > 1.9. Instead, at λ=405 nm, the BrC contribution to light absorption increases up to 33 % (average percentage of total babs), with a maximum value of 51 %, again when AAEexp > 1.9. The time series of babsBrC values at both wavelengths turned out to be well correlated (R2=0.71) with the levoglucosan (levo, in the following) concentration values, as reported in Fig. 6. The slope of the correlation curve increases by a factor of 5.8 when moving from the red to the blue light.

The average αBrC value turned out to be αBrC=3.9±0.1, in very good agreement with a previous value (αBrC=3.8±0.2) obtained in the same site and with the same approach (Massabò et al., 2016). The result is also in agreement with other literature (Yang et al., 2009; Massabò et al., 2015; Chen et al., 2015).

The methodology to extract the MAC value for BrC by the coupled use of MWAA and thermal–optical analysis has been introduced in a previous work (Massabò et al., 2016). In that case, a standard (i.e., with a red laser diode only) Sunset unit was used. The entire procedure is described in detail in Massabò et al. (2016); here we briefly summarize the main steps. a.

The fraction of light attenuation due to the BrC is first calculated in each sample with the MWAA raw data.

In the present experiment, the procedure was adopted to analyze the thermograms produced with both the red and the blue laser diodes mounted in the Sunset unit: the results are summarized in Fig. 7. Despite a rather high noise in the data, the MAC(BrC) value at the two wavelengths can be determined and it turns out to be MAC(BrC) =9.8±0.4 and 23±1 m2 g-1, respectively, at λ=635 and 405 nm. This result deserves some comments.

The MAC value at λ=635 nm differs for about 3σ from the result reported in Massabò et al. (2016) and obtained for the same site and in a similar season (i.e., November 2015 to January 2016; MAC =7.0±0.4 m2 g-1). Since differences in the type of wood burned in the past and present campaign cannot be excluded, the two values can be considered to be in fair agreement.