Temperature effects on optical properties and chemical composition of secondary organic aerosol derived from n-dodecane

Environmental temperature plays a vital role in controlling chemical transformations that lead to the formation of secondary organic aerosol (SOA) and ultimately impact the composition and optical properties of the aerosol particles. In this study, we investigated optical properties of n-dodecane SOA under two temperature conditions: 5 and 25 C. It was shown that low-temperature conditions could enhance the real part of the refractive index (RI) of the SOA at wavelengths of 532 and 375 nm. Mass spectrometry analysis revealed that the molecular composition of n-dodecane SOA was significantly modified by temperature: a large amount of oligomers were formed under lowtemperature conditions, which led to higher RI values. These findings will help improve our understanding of the chemical composition and optical properties of SOA under different temperature conditions and will provide one possible explanation for the low visibility in suburban areas during winter.


Introduction
Organic aerosol, especially secondary organic aerosol (SOA), plays a vital role in air quality, climate change and human health (Kanakidou et al., 2005;Poschl, 2005;Mellouki et al., 2015;Poschl and Shiraiwa, 2015;von Schneidemesser et al., 2015;Shrivastava et al., 2017). SOA accounts for a high proportion of atmospheric particulate matter around the world, especially in heavily polluted areas (Liu et al., 2017;Sun et al., 2014;Huang et al., 2014). Due to the variety of precursors and oxidation pathways, the composition of SOA is very complicated and variable (Lu, 2018;von Schneidemesser et al., 2015;Poschl and Shiraiwa, 2015;Hallquist et al., 2009;George et al., 2015), and the optical properties of SOA also exhibit different characteristics (Shrivastava et al., 2017;Zhang et al., 2015;Moise et al., 2015;Laskin et al., 2015). The complex refractive index (RI), m = n + ki (n is the real part, and k is the imaginary part; they express the extent of scattering and absorbing, respectively), is the only intrinsic optical property of a particle. The RI is controlled by the chemical composition and physi-cal characteristics (e.g., morphology and shape) of a particle (Moise et al., 2015). Quantifying the RI of aerosol particles is highly important to evaluate their optical properties and to further estimate their impacts on atmospheric visibility and the Earth's radiative balance.
Long-chain alkanes, which are an important class of intermediate-volatility organic compounds (IVOCs; Zhao et al., 2014) and a large fraction of diesel fuel and its exhaust (Gentner et al., 2012(Gentner et al., , 2017, are an important potential contributor of SOA (Presto et al., 2009;Zhao et al., 2016). Several previous studies have reported the formation of SOA derived from long-chain alkanes under various conditions, including SOA compositions (Fahnestock et al., 2015;Hunter et al., 2014;Lim and Ziemann, 2005), SOA yields (Loza et al., 2014;Lim and Ziemann, 2009a) and the chemical mechanisms (Yee et al., 2012(Yee et al., , 2013Aimanant and Ziemann, 2013;Lim and Ziemann, 2009b). Recently, Lamkaddam et al. (2016) reported the temperature dependence (10, 20 and 31.5 • ) of SOA formation from high-NO x photooxidation of n-dodecane and found that temperature did not significantly influence SOA yield. They attributed this to two possible reasons: the changes in reaction rate constants lead to different SOA compositions, or the SOA formed is mainly comprised of nonvolatile compounds and is, therefore, not sensitive to temperature. Li et al. (2017a) reported the optical properties of SOA from n-dodecane, n-pentadecane and n-heptadecane under various oxidation conditions at room temperature. However, knowledge on the effect of temperature on the chemical composition and optical properties of n-dodecane SOA in the absence of NO x is still lacking, which limits our understanding of the role of SOA in visibility and the radiative balance under different temperature conditions (e.g., in winter and summer).
In the present study, we determined the temperature effects on the chemical composition and optical properties of SOA generated in a smog chamber during the photooxidation of n-dodecane under low-NO x conditions. These results will improve our understanding of the impact of temperature on SOA chemical compositions and optical properties and, in turn, on air quality and radiative forcing.

Smog chamber experiments
The experiments were performed in a dual-reactor smog chamber, the details of which are given in Wang et al. (2015). Briefly, the chamber consisted of two 5 m 3 reactors made of fluorinated ethylene propylene (FEP) Teflon film, which were housed in a thermally isolated enclosure. The temperature in the chamber was accurately controlled by a highpower air conditioner within the range from −10 to 40 • C. Multiple light sources were used in the chamber, with a center wavelength of 365 nm (GE, F40BL), 340 nm (Q-lab, UVA-340) and 254 nm (PHILIPS, G36 T8). The RH and temperature in the chamber were continuously monitored and controlled throughout the experiments. The experiments were conducted at an RH < 5 % and under two different temperature conditions: 25 • C (room temperature, R) and 5 • C (low temperature, L). The temperature fluctuation was ±0.5 • C for both conditions. As the optical properties of the SOA can be affected by many factors, other factors had to remain unchanged in order to study the impact of temperature; thus, the humidity of the experiments was constant and could not be changed. Here, the experiments were conducted under dry conditions. This was due to the fact that when the temperature changed, the saturated vapor pressure of water changed: if the RH was consistent at different temperatures, the con-centration of the water was not consistent; when the concentration of water was the same, the RH was different. Therefore, choosing different humidity (non-dry) conditions would introduce new problems, and we could only use dry conditions (RH < 5 %). The n-dodecane (≥ 99 %, Sigma-Aldrich) was photooxidized under low-NO x conditions, with hydrogen peroxide (30 % wt/wt, Beijing Chemical Works) as the OH precursor. n-Dodecane was added into the chamber first and was followed by hydrogen peroxide. The wind turbine was then turned on for 20 min to ensure that the materials in the chamber were well mixed.
When all of the substances were added to the chamber, the chamber was set to the desired temperature. The instruments were connected at room temperature, and when the temperature dropped to 5 • and stabilized, the data measured by the instruments were considered to be valid data. The lights in the chamber were then turned on, and the photooxidation reaction started. The initial conditions for these experiments are listed in Table 1. According to our experimental design, the expected concentration of n-dodecane was 50 ppb, which involved introducing 2 µL of liquid ndodecane into the 5 m 3 smog chamber. As the injection volume of n-dodecane was 2 µL, volume error during injection was inevitable, which would influence the concentration of n-dodecane in the chamber. Nevertheless, the relatively small differences in the n-dodecane concentration (43-50 ppb at low temperature and 52-58 ppb at high temperature) likely had little influence on the SOA composition and optical properties.

Measurements
All instruments were located within 1 m of the chamber, and all of the connection tubes were wrapped with insulation cotton to minimize the influence of room temperature. The concentration of NO x and the O 3 formed in the smog chamber were monitored using gas analyzers (Teledyne Advanced Pollution Instrumentation, Model T400 and Model T200UP, respectively). The concentration of n-dodecane was monitored using a proton transfer reaction quadrupole mass spectrometer (PTR-QMS 500, IONICON), and calibration of the response of the PTR-QMS to n-dodecane was achieved through permeation tubes. The NO + ion source of the PTR-QMS was used when detecting the n-dodecane (Koss et al., 2016;Shi et al., 2019;Paulsen et al., 2005).
The particle size distribution and density were detected by a scanning mobility particle sizer (SMPS, TSI) and a centrifugal particle mass analyzer (CPMA, Cambustion). A custom-made cavity ring-down spectrometer (CRDS; Wang et al., 2012) was applied to monitor the optical properties of the formed particles at 532 nm. A photoacoustic extinctiometer (PAX-375, Droplet Measurement Technologies) was used to measure the scattering, absorption and extinction coefficients of formed SOA at 375 nm.
The optical properties of the formed particles were analyzed after the mass concentration of the aerosol reached a maximum. During the following 1-2 h, the surface mean diameter and the extinction coefficients of the particles tended to be stable and did not change much. We then collected the particles on the film to analyze their chemical composition during the same period. The sample collection time was chosen to make sure that the signal of the collection filter was much higher than the background of the blank filter in mass spectrometry. Another principle is to reduce the sample volume consumed during sampling. If the membrane extraction time is too long, the chamber volume will decrease too much. The formed aerosol particles were collected on the PTFE (polytetrafluoroethylene) membrane that had a pore size of 200 nm (0.2 µm, 47 mm, Merck Millipore FELP). Each filter sample was collected for 30 min at a 10 L min −1 flow rate. The filters were then put into 5 mL of methanol (99.9 %, Fisher Chemical) and sonicated for 30 min. The dissolved solution was analyzed with a UV-Vis light spectrometer (Avantes 2048F), which was used to detect the absorbing properties (to derive the imaginary part of RI, k) at 532 nm. The solution was also analyzed using electrospray ionization time-of-flight mass spectrometry (ESI-ToF-MS, Bruker, Impact HD) to obtain the chemical composition of the formed SOA. Positive ion mode was used for the ESI-ToF-MS. The absolute mass error was below 3 ppm, and the typical mass resolving power was > 30 000 at m/z 200.

Calculation method of RI values based on the CRDS
The RI values of the particles formed in the smog chamber were estimated based on both the extinction and scattering coefficients and Mie-Lorenz theory (Bohren and Huffman, 1983). The details of the RI values' calculation method are also reported in our previous publications (Wang et al., 2012;Phillips and Smith, 2014;Li et al., 2017aLi et al., , b, 2018Peng et al., 2018). The extinction coefficients (α ext ) of the particles established using CRDS can be calculated with Eq. (1): where L is the distance between the two mirrors in the cavity, l is the length of the cavity that filled with aerosol particles, c is the speed of the light, τ 0 is the ring-down time of the CRDS when it is filled with zero air and τ is the ring-down time of the CRDS when it is filled with aerosol particles. For the homogeneous spherical particles, the α ext can be calculated using Eq. (2): a Experimental conditions: n-dodecane at room temperature (Dod-R) and n-dodecane at low temperature (Dod-L). b The mass concentration is obtained without wall loss correction. c The mass concentration is calculated with wall loss correction.
where N is the concentration of the spherical particles, σ ext is the extinction cross section, D is the particle diameter and Q ext is the extinction efficiency. Q ext is the ratio of the Beer's law extinction cross section to the geometric area of the spherical particles; it is dimensionless, and can be expressed using Eq. (3): For the polydisperse SOA particles formed in the smog chamber, which follow a log-normal distribution and have a geometric standard deviation that is always smaller than 1.5, the Q ext with a surface mean diameter of D s can be expressed as follows (with the assumption that the aerosol particles in each size bin are homogeneous spherical particles; Nakayama et al., 2010): where D p is the geometrical diameter of the particle, dD p is the size bin of the particles and N (D p ) is the number concentration of the particles in dD p with D p per unit volume. For the particles with D p in each size bin dD p , the surface area S(D p )can be expressed by Eq. (5): Thus, Eq. (4) can also be expressed as follows: where S tot is the total surface area of the particles, and the values can be obtained with an SMPS. The extinction efficiency can also be calculated with Mie-Lorenz theory: where f (D p ) is the normalized surface area-weighted size distribution function. The measured extinction efficiency (Q ext,mea ) is then compared to the calculated extinction efficiency (Q ext,cal ), and the best-fit RI value is determined by minimizing the following reduced merit function (χ r ): where N is the number of diameters to be calculated. The uncertainties of the particle concentration and the surface mean diameter measured by SMPS are ±10 % and ±1 %, respectively. The uncertainty of the retrieval method is ±0.002, and the uncertainty of the measured extinction coefficient with CRDS is ±3 %, resulting in a final uncertainty of the retrieved RI value of about 0.02-0.03. The corresponding equation for the RI value uncertainty is given in the Supplement.

Calculation method of RI values based on the quantitative structure-property relationship (QSPR)
Using the molecular formula obtained from ESI-ToF-MS, we calculated the RI values of the products in SOA with the quantitative structure-property relationship (QSPR) method (Redmond and Thompson, 2011). The QSPR can be expressed as follows: where µ is the unsaturation of the molecular formula, α is the polarizability of the molecular formula, ρ m is the mass density (g cm −3 ) and M is the molar mass (g mol −1 ). The mass density of the compound is estimated using the AIM (Aerosol Inorganics Model), the details of which are given in Girolami (1994).
µ is calculated via the conventional approach, which is used in many organic chemistry texts: where no. C, no. H and no. N are the number of the C, H and N, respectively.
α is calculated based on the molecular formula of the compound, and it can be expressed as follows: where "no." is the number of the atoms of each element in the molecular formula.
The calculated RI values of products were used to link the chemical composition and optical properties as well as to explain the observed RI differences at different temperatures in Sect. 3.4.

Calculation method of average elemental composition and ratios
Due to the complexity of the chemical composition in SOA particles, it is common to express the bulk composition as averaged elemental ratios. The average elemental composition and ratios ( where Here, x i is the peak abundance of the elemental composition.

Impact of temperature on direct radiative forcing
The simple forcing efficiency (SFE) is used to determine the relative importance of the optical properties of the aerosol to direct radiative forcing at the Earth's surface (Bond and Bergstrom, 2006): where S 0 is the solar radiation, τ atm is the transmission of the atmosphere, F c is the cloud fraction, a s is the surface albedo, Q bs and Q a are the backscattering and absorption efficiency of the aerosol particles, M is the aerosol mass, and C is the cross section of the aerosol. The SOA derived from n-dodecane has negligible absorption at the wavelengths of 532 and 375 nm under the two temperature conditions; thus, the value of Q a is zero, and the impact of temperature on the direct radiative forcing (DRF) can be expressed using DRF ratio = SFE low temperature SFE normal temperature = Q bs, low tem. Q bs, normal tem. , and this will be discussed Sect. 3.5.
3 Results and discussion

Photooxidation experiments
The profiles of the n-dodecane photooxidation experiments at different temperatures are shown in Fig. S1 in the Supplement. As similar amounts of n-dodecane and oxidant were added to the smog chamber and the reaction rate was lower under low temperature, the reaction time under lowtemperature conditions was significantly longer than that at room temperature: 4-5 h for room temperature and 8-9 h for low-temperature conditions. Under room temperature conditions, the total surface concentration and mass reached a maximum after 3 h, whereas it reached a maximum after 6 h under low-temperature conditions.

The effect of temperature on RI
The optical properties of the formed particles were analyzed after the mass concentration of the aerosol reached a maximum, and data from the last 1 h were used. During this period, the surface mean diameter and the extinction coefficients (α ext ) of the particles tended to be stable and did not change much. The extinction coefficients and the extinction efficiency of the particles as a function of the surface mean diameter are shown in Fig. S4 in the Supplement. The SOA derived from n-dodecane had no significant absorption at 532 or 375 nm, similar to our previous study (Li et al., 2017a), and the imaginary part of the RI will not be discussed here. The real part of the RI obtained in this study is shown in Fig. 1. As shown in Fig. 1, the RI values at 532 nm at room temperature were similar to our previous study (Li et al., 2017a), whereas the two temperature conditions had significantly different RI ranges: 1.472-1.486 (25 • C) and 1.502-1.526 (5 • C) at 532 nm; 1.51-1.53 (25 • C) and 1.532-1.56 (5 • C) at 375 nm. Details on the variation tendency of the RI values as a function of the surface mean diameter for SOA produced under room temperature and low-temperature conditions are given in Fig. 2. The various RI values at different temperatures indicate that lower reaction temperatures (from 25 to 5 • C) had an enhancing effect (∼ 0.03 at 532 nm and ∼ 0.02 at 375 nm) on the RI of n-dodecane SOA. The mass spectrometry analysis below (Sect. 3.3) was applied to obtain chemical composition information and explain the phenomenon above.

Temperature effect on the chemical composition and reaction mechanism
The higher RI values under low-temperature conditions indicated that the temperature might change the chemical composition of SOA by changing the reaction types or shifting the balance of different pathways. The mass spectra (MS) of n-dodecane SOA obtained by ESI-ToF-MS in positive ion mode are shown in Fig. 3, and they provide molecular insight into the chemical changes under different temperature condi- Figure 1. Summary of the averaged RI values of n-dodecane SOA at room temperature (25 • C; the light orange area) and under lowtemperature conditions (5 • C; the light blue area) at wavelengths of 532 and 375 nm. The red box is the averaged RI value for ndodecane at 532 nm, the shaded boxes are the RI values from our previous study at 532 nm (Li et al., 2017a) and the blue box is the averaged RI value for n-dodecane at 375 nm. tions. We identified about 260 individual masses for SOA under the two temperature conditions, and details are given in Fig. S2 and Table S1 in the Supplement. The spectrum under low-temperature conditions was significantly different from the spectrum at room temperature, with a large amount of ions corresponding to monomer, dimer, trimer and tetramer. This suggests that oligomerization might play a dominant role under low-temperature conditions. The molecular composition of n-dodecane SOA was significantly modified by temperature conditions, with the averaging SOA formula changing from C 14.98 H 26.47 O 5.53 at room temperature to C 21.25 H 40.44 O 7.43 under lowtemperature conditions. The average carbon number increased from 14.98 at room temperature to 21.25 under lowtemperature conditions, indicating that the SOA molecules were larger at lower temperatures. The average O/C and H/C ratios at room temperature were 0.37 and 1.72, respectively, whereas the ratios were 0.35 and 1.90 under low-temperature conditions, respectively. The products formed under lowtemperature conditions tended to have a higher H/C ratio and a lower O/C ratio compared with the products formed at room temperature. The details on the O/C and H/C ratios of the products formed under different temperature conditions can be found in Fig. 5a. The phenomenon above might be due to the presence of oligomers.
For the oligomers in the mass spectrum, one possible explanation was that low-temperature conditions promoted gas-particle partitioning and changed the particle-phase reaction. Gas-phase OH oxidation could reduce the H/C ratios and increase the O/C ratios (Li et al., , 2019Lambe et al., 2015), whereas particle-phase oligomerization basically could not change the O/C or H/C ratios (Charron et al., 2019). At room temperature, more gas-phase oxidation steps were needed to cause the less volatile products to condense into the particle phase (due to the high temperature, i.e., high saturation vapor pressure). Hence, products formed at room temperature had lower H/C and higher O/C ratios. In contrast, the more volatile (i.e., less oxidized) products were readily condensed into the particle phase under low-temperature conditions and then subse-  (Fahnestock et al., 2015;Yee et al., 2013Yee et al., , 2012. quently underwent particle-phase reactions, e.g., oligomerization, leading to the formation of products with higher H/C and lower O/C ratios. This phenomenon was consistent with Kim et al. (2014), who studied the dependence of real part of the RI on the O/C and H/C ratios of SOA derived from limonene and α-pinene and found that the higher percentage of less oxygenated semivolatile substances was responsible for the higher RI values.
For particle-phase reactions, there were two main reaction pathways (Fahnestock et al., 2015;Yee et al., 2012Yee et al., , 2013: intramolecular cyclization of multifunctional hydroperoxides (which form furan derivatives) and intermolecular oligomerization of multifunctional hydroperoxides with aldehydes (which form peroxyhemiacetal, PHA). These two pathways were competitive during the particle-phase reaction process. Based on the mass spectra analysis, we speculated that low temperature promoted the progress of the oligomerization reaction and made it the primary pathway in the particle phase, as shown in Fig. 4. As discussed above, products with a higher volatility and lower oxidation state would condense into the particle phase under low-temperature conditions and Figure 5. Panel (a) shows a van Krevelen plot of the O/C and H/C ratios for identified products from EST-ToF-MS. The orange circle is for the SOA generated at room temperature, and the blue circle is for the SOA generated under low-temperature conditions. Panels (b) and (c) show the calculated RI values of the selected identified molecular species from MS spectra at room temperature and under low-temperature conditions, respectively; the color maps refer to the relative intensity of the molecular formula.
subsequently participate in particle-phase reactions, which would further promote the oligomerization reaction.
Another possible explanation was that the pathway of the gas-phase reaction changed under low-temperature conditions compared with room temperature conditions. Combining existing mass spectrometry information, we speculated that the oligomers might also be formed by gas-phase radical oligomerization under low-temperature conditions and then rapidly deposited into the particle phase. However, these are speculations based on the existing analysis results. The specific reaction mechanism under low-temperature condition requires further investigation. Clark et al. (2016) found that the SOA formed from isoprene had similar mass spectra in the monomer range at different temperatures. However, the precursors are very different (isoprene vs. n-dodecane) and have very different oxidation pathways and partitioning processes. Hence, it is possible that we observe different results in mass spectra in the monomer range. As mentioned above, the different oxidation degree and partitioning process caused the difference in the monomer range. At room temperature, more OH oxidation steps in the gas phase can lead to the formation of some fragmentation products that may not be observed at low temperatures. For example, we observed the high intensity of m/z 195 C 10 H 20 O 2 (relative intensity 0.39) and m/z 211 C 10 H 20 O 3 (relative intensity 0.13) at room temperature but very low intensity (0.012 and 0.041, respectively) at low temperature; furthermore, m/z 195 and m/z 211 are likely products of gas-phase OH oxidation. This process may be different from isoprene (Clark et al., 2016). This is due to the fact that the carbon number is high, volatility is low and fewer oxidation steps are needed before partitioning for n-dodecane, whereas there are only five carbon atoms for isoprene so more oxidation steps are needed before partitioning regardless of temperature. Overall, the differences in the monomer range still make sense.

Relationship between RI values and the chemical composition of SOA
The RI of aerosol particles is fundamentally the results of a combination of particle chemical compositions and internal mixing. SOA particles formed in the smog chamber are treated as homogenous mixtures, and the RI values of these mixtures can be expressed as follows (Redmond and Thompson, 2011): where x i is the fraction of the ith component, and RI i is the refractive index of component i.
The RI values were calculated for the identified products (with high intensity) at room temperature and under lowtemperature conditions (details are given in Fig. 5b and c and in Table S1 in the Supplement). It can clearly be seen that the RI values of the products at room temperature were mainly in the range of 1.4 to 1.5, and the degree of oligomerization was generally in the monomer to dimer range. However, for the products formed under low-temperature conditions, the RI values were within the range of 1.4 to 1.55, and the degree of oligomerization could reach tetramerization. As the degree of oligomerization increased, the RI also gradually increased. This generally explained the higher RI of SOA under lower-temperature conditions.
To further validate our speculation and identify the relationship between the measured RI values and the chemical  composition of n-dodecane SOA, we chose a surrogate system containing 11 PHA oligomerization and 2 cyclization reactions (with different degrees of unsaturation and different functional groups) to calculate the expected RI values using the quantitative structure-property relationship (QSPR) method based on the molecular formula and structure (Redmond and Thompson, 2011). The details of the 13 reactions and the related molecular information are shown in Fig. S3 in the Supplement and in Table 2; the relationship between the predicted RI values and the degree of unsaturation as well as the relationship between the predicted RI values and the degree of oligomerization are shown in Fig. 6. Strong correlations were observed between the predicted RI values and unsaturation and between the RI values and the degree of oligomerization. With the same unsaturation, the RI values increased with an increasing degree of oligomerization; under the same degree of oligomerization, the RI values increased with an increasing degree of unsaturation. In addition, the RI values of the surrogate system were similar to the identified substances, which further confirmed the effect of oligomers on RI values. In addition to the abovementioned oligomer reasons, the difference in wall losses and gas-particle partitioning of gasphase products might partially contribute to the RI enhancement under low-temperature conditions. The particle wall loss rates under both room and low-temperature conditions were measured. The wall loss rate under low-temperature conditions (0.0025-0.0028 min −1 ) was larger than that at room temperature (0.0018-0.0020 min −1 ). After wall loss correction, the SOA mass under low-temperature conditions was higher than that at room temperature. However, this difference in the particle wall loss rate can only slightly change the SOA mass concentration, and it cannot change the particle chemical composition; therefore, it is unlikely to change the optical properties. Nevertheless, the difference in the vapor wall loss rates may change the particle composition and optical properties. Low temperature can enhance the loss rates of higher-volatility compounds (whereas the dominant fate of lower-volatility compounds is condensation, so temperature may have little impact on their losses), which may lead to their lower proportions in SOA particles. As their RI values are generally lower than lower-volatility compounds , this proportion change could enhance SOA RI at low temperature. Therefore, the difference in wall losses and gas-particle partitioning of gas-phase products might partially contribute to the RI enhancement under low-temperature conditions. Figure 7 shows the light extinction efficiency (Q ext ) ratios and direct radiative forcing (DRF) ratios of n-dodecane SOA under different temperature conditions, from which we could establish the impacts of temperature on the role of ndodecane SOA in visibility and radiative balance. As shown in Fig. 7a and c, the extinction efficiency (Q ext ) of SOA gen-erated under low-temperature conditions is higher than for SOA generated at room temperature in the size range from 50 to 200 nm, which is the most atmospherically relevant size range Guo et al., 2014). The enhancement is about 7 %-20 % at 532 nm and about 1 %-21 % at 375 nm. This suggests that the extinction efficiency (Q ext ) of SOA formed from n-dodecane (and perhaps other longchain alkanes in general) was higher in winter than in summer, which would result in lower visibility.

Atmospheric and climate implications
The enhancement in the light extinction of SOA and the oligomer compositions formed under low-temperature conditions might provide some possible insight into regional visibility issues, especially in suburban areas. It has been reported that UV-scattering particles in the boundary layer could accelerate photochemical reactions and haze production (Sun et al., 2014). The observations above showed that the scattering property of formed SOA increased under lowtemperature conditions, which might provide one possible reason for the rapid occurrence of haze in suburban areas in winter. According to field observations, haze frequently occurs in winter, especially in China (Cheng et al., 2016;Huang et al., 2014;Guo et al., 2014;Parrish et al., 2007). When haze occurs, it is often accompanied by high-NO x conditions, especially in urban areas. The temperature effects under high-NO x conditions are also important and need to be investigated in future studies.
As shown in Fig. 7b and d, the DRF values under low-temperature conditions were generally larger than at room temperature: the enhancement was about 6 %-19 % at 532 nm (50-200 nm) and about 7 %-22 % at 375 nm (50-180 nm); for the 180-200 nm size range, low-temperature conditions decreased the DRF ratio by about 3 % at 375 nm. This phenomenon implies that the SOA generated in winter (low temperatures) might have larger direct radiative forcing effects on the Earth's surface than the SOA generated in summer (room temperature). It may also imply that temperature conditions should be considered when evaluating the DRF of the aerosol particles generated in the atmosphere.

Conclusions
To the best of our knowledge, this was the first report on the optical properties of long-chain alkane SOA under lowtemperature conditions. The modification of temperature significantly changed the chemical composition of the particulate phase. From the oligomer component at low-temperature conditions, it was presumed that oligomerization was dominant at low temperatures. The presence of oligomers in the SOA particles resulted in an increase in the RI values. This study helps to improve our understanding of reduced visibility and the formation of haze in winter. Our results also show the need for further investigation of the atmospheric parameters influencing SOA formation and optical properties.
Data availability. The data used in this study are available upon request from the corresponding author.
Author contributions. WW conceived and led the studies. JL, WZ and CP performed chamber simulation and data analysis. KL, LZ, BS, YC, ML, HL and MG discussed the results and commented on the paper. JL prepared the article with contributions from all coauthors.
Competing interests. The authors declare that they have no conflict of interest.  Review statement. This paper was edited by Alex Lee and reviewed by two anonymous referees.