Measurements of Indoor and Outdoor Fine Particulate Matter during the Heating Period in Jinan, in North China: Chemical Composition, Health Risk, and Source Apportionment

Fine particulate matter (PM2.5) was simultaneously collected from the indoor and outdoor environments in urban area of Jinan in North China from November to December 2018 to evaluate the characteristics and sources of indoor PM2.5 pollution. The concentrations of indoor and outdoor PM2.5 were 69.0 ± 50.5 µg m−3 and 128.7 ± 67.9 µg m−3, respectively, much higher than the WHO-established 24-h standards for PM2.5, indicating serious PM2.5 pollution of indoor and outdoor environments in urban Jinan. SO42−, NO3−, NH4+, and organic carbon (OC) were the predominant components, which accounted for more than 60% of the PM2.5 concentration. The total elemental risk values in urban Jinan for the three highly vulnerable groups of population (children (aged 2–6 years and 6–12 years) and older adults (≥70 years)) were nearly 1, indicating that exposure to all of the elements in PM2.5 had potential non-carcinogenic risks to human health. Further analyses of the indoor/outdoor concentration ratios, infiltration rates (FINF), and indoor-generated concentration (Cig) indicated that indoor PM2.5 and its major chemical components (SO42−, NO3−, NH4+, OC, and elemental carbon) were primarily determined by outdoor pollution. The lower indoor NO3−/SO42− ratio and FINF of NO3− relative to the outdoor values were due to the volatility of NO3−. Positive matrix factorization (PMF) was performed to estimate the sources of PM2.5 using the combined datasets of indoor and outdoor environments and revealed that secondary aerosols, dust, cement production, and coal combustion/metal smelting were the major sources during the sampling period.


Introduction
Epidemiological and toxicological studies have repeatedly shown that fine particulate matter (PM 2.5 ) is linked with increased health risk, such as respiratory illness, cardiovascular diseases, Alzheimer's disease, and increased mortality [1][2][3]. The toxicity of PM 2.5 is closely associated with its chemical composition (e.g., water-soluble ions, organic carbon (OC) to elemental carbon (EC) ratio, trace elements, and polycyclic aromatic hydrocarbons (PAHs)) [4,5]. Previous studies of PM 2.5 and its chemical characteristics have mostly focused on the outdoor atmosphere rather than indoor environments [6][7][8][9][10][11]. Some studies have shown that urban population spends~80% of time indoors [12,13]; therefore, it is of particular importance to make clear the characteristics of indoor PM 2.5 pollution and to evaluate the effect on human health. NO 3 − , and NH 4 + had no indoor sources in a typical office in Milan. To date, although several studies have investigated the indoor PM 2.5 components and their sources (e.g., OC and EC [22,42], OC, EC, S, Ca, K, and V [43,44], and PAHs [29,45]), the contribution of outdoor sources to the indoor PM 2.5 pollution in polluted environments in North China remains unclear. Comprehensive and systematic evaluation is required to fill this knowledge gap. As the capital of Shandong Province in the North China Plain, Jinan has experienced heavy PM 2.5 pollution in both indoor and outdoor environments in the last few years [24,26]. In recent years, the PM 2.5 and SO 4 2− concentrations have rapidly decreased due to the stringent control of SO 2 emissions [46,47]; however, NO 3 − concentrations have exhibited an increasing trend in the outdoor environment in Jinan [46][47][48], which could have an influence on indoor PM 2.5 pollution. In this study, sample collection and chemical analysis of aerosol in an office and outdoor environment were synchronously performed in Jinan to investigate the characteristics and main sources of indoor PM 2.5 , and assess its human health risk using the data of constituent elements. The results of this study were also compared with those of the previous studies conducted in Jinan to identify the variations of indoor and outdoor PM 2.5 pollution.

Sample Site and Chemical Analysis
The measurements were synchronously conducted in a typical office room and the outdoor atmosphere in an urban area in Jinan (36 • 69 N, 117 • 06 E) from 12 November to 5 December 2018 ( Figure 1). The outdoor PM 2.5 samples were collected on the rooftop (approximately 20 m above ground level) of a six-story office building on the west campus in the University of Jinan, whereas an office room on the second floor of the same office building was selected for indoor sampling. The building is more than 20 years old and has not been decorated or furnished in the last 10 years. The room has two sliding windows facing the street and one door opening into the corridor. During the sampling period, the office was empty and the windows were closed without any natural and mechanical ventilation. The door was always closed throughout the sampling process, although students entered and exited the room daily to collect samples. The sampling site is surrounded by residential and commercial areas. There are two major roads nearby (~500 m away) that frequently have heavy traffic, viz., ErHuan South Road in the south and Nanxinzhuang West Road in the west. The inlets for PM 2.5 collection were set 1.5 m above the floor.
OC and EC were determined with a thermal-optical carbon analyzer (Dual-oven model, Sunset Laboratory, OR, USA) with a non-dispersive infrared detector, which used the thermal-optical transmittance (TOT) method and complied with the NIOSH 5040 protocol. The principles and operation of this analyzer were described by Wang et al. [51]. The trace elements (K, Cl, Al, Fe, Ca, Ti, Mn, Cu, Cr, Se, Ni, Zn, As, Pb, S, V, and Co) in the PM2.5 samples were quantified by an energy dispersive X-ray fluorescence spectrometer (XRF, Epsilon 4, Malvern Panalytical, Malvern, UK). The estimated detection limits were 0.010-0.084 µg m −3 for all of the ions, 0.04 µg m −3 for OC/EC, and 0.3-1.0 ng cm −2 for elements, with measurement uncertainties of approximately 10%. The temperature was obtained from the PM2.5 intelligent sampler (Model TH-150A, Wuhan Tianhong Corporation, Wuhan, China).  PM 2.5 samples were collected manually on quartz fiber filters (90 mm in diameter with a 2 µm pore size; Pall, NY, USA) using an intelligent sampler (Model TH-150A, Wuhan Tianhong Corporation, Wuhan, China) at a flow rate of 100 L/min. Most samples were collected in two 12-h phases in a day (around 7:30-19:00 and 19:30-7:30), and in a single 24-h phase (from 7:30 to 7:00 the next day). Finally, 34 sets of PM 2.5 samples (7 sets for 24 h, 26 sets for 12 h, and 1 blank sample) were collected from both indoor and outdoor environments. The filter was mounted on the sampler for 24-h to collect the blank sample when the sampler was not operated, and the blank sample was processed simultaneously with the field samples. Before sampling, the quartz filters were baked in a muffle oven at 600 • C for 4 h, and after sampling, all of the filters were kept in plastic Petri dishes and then stored in a freezer at −4 • C.
To obtain the mass concentrations of PM 2.5 , the filters were weighed before and after sampling with a Sartorius ME 5-F electronic microbalance (sensitivity: ±1 µg, Sartorius, Göttingen, Germany) after equilibration for 24 h at 20 • C to 23 • C and 35% to 45% relative humidity. The filter was cut into three sections for analysis of water-soluble ions, OC/EC and trace elements. The concentrations of water-soluble ions (including F − , Cl − , NO 2 − , NO 3 − , SO 4 2− , Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) were determined by ion chromatography (Dionex IC 90, Dionex Corporation, Sunnyvale, CA, USA) after the filters were extracted with ultra-pure water and the water extracts were filtered. The details of this process can be found in reports by Nie et al. [49] and Zhou et al. [50]. The contents of OC and EC were determined with a thermal-optical carbon analyzer (Dual-oven model, Sunset Laboratory, OR, USA) with a non-dispersive infrared detector, which used the thermal-optical transmittance (TOT) method and complied with the NIOSH 5040 protocol. The principles and operation of this analyzer were described by Wang et al. [51]. The trace elements (K, Cl, Al, Fe, Ca, Ti, Mn, Cu, Cr, Se, Ni, Zn, As, Pb, S, V, and Co) in the PM 2.

Positive Matrix Factorization (PMF) Analysis
A receptor model from the U.S. Environmental Protection Agency (EPA), PMF 5.0, was used to identify the contributing sources of the major components in PM 2.5 in Jinan. The PMF model is a multivariate factor analysis tool that is widely used for atmospheric samples [52,53]. Detailed information on the principles and use of the PMF model can be found in the user guide and related literature [52,53]. The concentrations of water-soluble ions, OC, EC, and elements were input into the PMF model to estimate the sources of PM 2.5 . This study made use of the same uncertainty estimation method used by Wang et al. [53]. (1) where MDL ij is the method detection limit of species j in sample i and C ij is the concentration of species j in sample i. After that, three to nine factors and plenty of combinations of chemical species were trialed to obtain the best solution of PMF. One hundred runs were performed for each calculation to ensure the robustness of the statistics. The number factors were selected mainly based on the physical interpretability of PMF solutions and the factor matching rate calculated by bootstrap error estimation. Solutions with five to seven factors show the high physical interpretability. However, each factor profile of six-factor PMF result has the most physicochemical meaning. In five-factor PMF solution, road dust factor has seldom SO 4 2− and NO 3 − and there is no Ti in it. For seven-factor PMF results, there is a factor hardly to be interpreted, which has abundant secondary inorganic ions, OC, EC and some mineral species but no Ti, in which SO 4 2− and Ni has the highest contribution. Furthermore, the factor matching rates are approximately or less than 80% for five-and seven-factor solutions, while more than 90% for the six-factor solution.
Hence, six-factor solution was selected as the optimal PMF scheme.  [54], which implies that PM 2.5 pollution in both indoor and outdoor environments in Jinan is severe. Table 1 summarizes the statistics (including daily, daytime, and nighttime mean values) of indoor and outdoor PM 2.5 and its major chemical components. In general, the concentrations of indoor PM 2.5 and its major chemical components were always lower than those of the outdoor counterparts ( Figure 2 and Table 1). The mean outdoor and indoor PM 2.5 concentrations were 69.0 ± 50.5 µg m −3 and 128.7 ± 67.9 µg m −3 , respectively. In a highly polluted urban area such as Jinan, indoor PM 2.5 was approximately 50% of the outdoor PM 2.5 and therefore indoor had slightly better air quality. This holds for an unoccupied, unused office that has closed the door and windows and is probably determined by outdoor air leaking into the building. Compared with the findings of previous studies in urban Jinan [24,26], both indoor and outdoor concentrations in this study were lower, suggesting alleviation of PM 2.5 pollution. Water-soluble ions were the most abundant species. The total concentrations of water-soluble ions (F − , Cl − , NO 2 − , NO 3 − , SO 4 2− , Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) in outdoor and   The indoor and outdoor PM2.5 composition were similar (see Figure 3), and SO4 2− , NO3 − , NH4 + and OC were the predominant chemical components, together contributing 61.9% and 63.7% to the indoor and outdoor PM2.5 concentrations, respectively. The main difference observed between the indoor and outdoor PM2.5 was in the NO3 − /SO4 2− mass ratio. The outdoor NO3 − /SO4 2− mass ratio was 1.94, approximately twice the indoor ratio of 1.00. This difference can be explained by the loss of semi-volatile NO3 − during the transport of PM2.5 from the outside atmosphere to indoor air due to  from 2005 to 2015 in Jinan [48]. We further compared the concentrations of indoor PM2.5 and its chemical components in this study with those in previous studies [24,27]. The indoor PM2.5 and water-soluble ion concentrations (except NO3 − ) in 2018 decreased, following the decrease in outdoor concentrations. However, the indoor NO3 − concentration in 2018 increased by almost 25%, which was consistent with the increase in the outdoor NO3 − concentration. These phenomena may be due to the following two causes: (1) the strict control of SO2 emissions has led to a remarkable decrease in SO4 2− concentration in China [35,46]; and (2) the sampling site is different from those in the previous studies; and specifically, it is close to ErHuan South Road and Nanxinzhuang West Road, which frequently have heavy traffic.

Assessment of Health Risks for Metals and Sulfur in PM2.5
PM2.5-bound heavy metals can enter deep into the lungs and cause risks to human health due to their toxicity. Thus, in this study, the possible non-carcinogenic risk to human health from both indoor and outdoor environments was assessed based on representative elemental components of PM2.5 for three highly vulnerable groups: children (aged 2-6 years and 6-12 years) and older adults (≥70 years). According to the method described by the U.S. EPA [56], the elemental risk (R) was calculated using the formula where DE: dose of exposure (mg/kg-day); C: mean concentrations (mg/m 3 ); I: inhalation rate (m 3 /day); F: exposure frequency (days/year); D: exposure duration (years); t: average time (days); W: body weight (kg); R: elemental risk; RD: reference dose [56]. The values of W, I, F, and D were 16, 6.00, 180 and 4 for young children, 29, 11.04, 180, and 6 for children aged of 6-12 years, 70, 19.92, 180, and 30 for adults, respectively. Table 2 shows the calculated R values for the indoor and outdoor environments. concentration [24,26,27,55]. In this study, the NO 3 − /SO 4 2− mass ratio (1.94) was higher than the results obtained from 2005 to 2015 in Jinan [48]. We further compared the concentrations of indoor PM 2.5 and its chemical components in this study with those in previous studies [24,27]. The indoor PM 2.5 and water-soluble ion concentrations (except NO 3 − ) in 2018 decreased, following the decrease in outdoor concentrations. However, the indoor NO 3 − concentration in 2018 increased by almost 25%, which was consistent with the increase in the outdoor NO 3 − concentration. These phenomena may be due to the following two causes: (1) the strict control of SO 2 emissions has led to a remarkable decrease in SO 4 2− concentration in China [35,46]; and (2) the sampling site is different from those in the previous studies; and specifically, it is close to ErHuan South Road and Nanxinzhuang West Road, which frequently have heavy traffic.

Assessment of Health Risks for Metals and Sulfur in PM 2.5
PM 2.5 -bound heavy metals can enter deep into the lungs and cause risks to human health due to their toxicity. Thus, in this study, the possible non-carcinogenic risk to human health from both indoor and outdoor environments was assessed based on representative elemental components of PM 2.5 for three highly vulnerable groups: children (aged 2-6 years and 6-12 years) and older adults (≥70 years). According to the method described by the U.S. EPA [56], the elemental risk (R) was calculated using the formula where DE: dose of exposure (mg/kg-day); C: mean concentrations (mg/m 3 ); I: inhalation rate (m 3 /day); F: exposure frequency (days/year); D: exposure duration (years); t: average time (days); W: body weight (kg); R: elemental risk; RD: reference dose [56]. The values of W, I, F, and D were 16, 6.00, 180  Table 2 shows the calculated R values for the indoor and outdoor environments. Overall, the risk values of various elements were slightly higher in the outdoor environment than the indoor environment. Individual elemental risk values greater than 0.1 were considered to denote adverse health effects for children [57]. As shown in Table 2, the elements in PM 2.5 with individual risk values greater than 0.1 for the study populations were Mn, Cr, Co, and S in both indoor and outdoor environments. To consider the cumulative effect of the non-carcinogenic risks of elements, the individual risk values of the 10 elements were summed to give the total risk values. The total risk values were 0.78 for children aged 2-6 years, 0.80 for children aged 6-12 years and 0.59 for elderly adults in the indoor environment. These values were slightly higher in the outdoor environment. In addition, a comparison between the risk values for children and older adults indicated that children were the most susceptible group to non-carcinogenic effects. Compared with the data reported by Yang et al. [57], the risk values of individual elements in this study were lower, which is consistent with the reduction of PM 2.5 concentration as described above. These results indicated that the exposure to elements found in PM 2.5 may pose a reduced public health risk in this study area. However, the results of the risk assessment are affected by some uncertainties, associated mainly with the estimates of toxicity values and exposure parameters. In this study, we calculated the risks of only some heavy metals in PM 2.5 (e.g., Cd was not assessed), and PM 2.5 contains other harmful substances as well, such as PAHs [27]. Therefore, although the total risk value of elements calculated in this study was less than 1, the harmful effects of PM 2.5 to human health cannot be ignored because the actual risks could be greater than our calculated results.

Indoor/Outdoor Ratio (I/O Ratio)
The indoor/outdoor ratio (I/O ratio) is calculated as C in /C out [37], where C in and C out are the indoor and outdoor species concentrations, respectively. Figure 4 shows I/O ratios of PM 2.5 and its chemical components. The average I/O ratios of PM 2.5 and its chemical components were all lower than 1.0, indicating that the indoor PM 2.5 concentration was strongly affected by outdoor pollution. The I/O ratios of SO 4 2− and OC were the highest, both of which were 0.81 (SO 4 2− range: 0.65-1.05, OC range: 0.55-1. 16), whereas that of NO 3 − (0.41 ± 0.10) was lower, as also observed by previous studies [24,26]. The decomposition of ammonium nitrate and other volatile components due to higher indoor temperatures led to a reduction in the PM 2.5 I/O ratio.

Infiltration Factor (FINF) and Indoor Source Contribution (Cig)
The linear regressions between indoor and outdoor PM2.5 and its major chemical components can be expressed as Cin = FINF Cout + Cig [37], where Cig is the indoor particle concentration contributed by indoor sources, and FINF (called the infiltration factor) is the fraction of outdoor PM2.5 and its chemical compounds that enter the indoor environment and remain suspended. The linear regressions of PM2.5 and its major chemical components are depicted in Figure 5. The FINF of PM2.5 was approximately 0.68, indicating that more than half of the outdoor PM2.5 had entered indoors. It should be noted that the office windows were closed during most of the sampling periods, which limited the effect of natural ventilation on the indoor-outdoor air circulation. The Cig of PM2.5 was negative, indicating the very limited contribution of indoor sources to indoor PM2.5 concentration. A similar finding was reported by Sangiorgi et al. [5] in cold conditions.
As with the I/O ratio, the FINF of major chemical components also followed the order of SO4 2− > OC > EC (>PM2.5) > NH4 + > NO3 − , which could be attributable to the semi-volatile characteristic of NH4NO3 and the non-volatile characteristic of (NH4)2SO4 and EC. Temperature is the main factor affecting the gas-particle conversion of semi-volatile compounds [5,14,40]. During the transfer of NH4NO3 from outdoor to indoor, the higher indoor temperature (21 °C on average) compared with the outdoor temperature (10 °C on average) led to the partitioning from the particle phase to the gas phase [5,26]. However, (NH4)2SO4 and EC were not affected by the temperature. The Cig values for major chemical components were positive, except for NH4 + . The Cig values of SO4 2− , OC, EC, NH4 + , and NO3 − only accounted for less than 20% of their indoor concentrations, implying that the indoor sources of PM2.5 were negligible, which was consistent with the characteristics of the selected office.

Infiltration Factor (F INF ) and Indoor Source Contribution (C ig )
The linear regressions between indoor and outdoor PM 2.5 and its major chemical components can be expressed as C in = F INF C out + C ig [37], where C ig is the indoor particle concentration contributed by indoor sources, and F INF (called the infiltration factor) is the fraction of outdoor PM 2.5 and its chemical compounds that enter the indoor environment and remain suspended. The linear regressions of PM 2.5 and its major chemical components are depicted in Figure 5. The F INF of PM 2.5 was approximately 0.68, indicating that more than half of the outdoor PM 2.5 had entered indoors. It should be noted that the office windows were closed during most of the sampling periods, which limited the effect of natural ventilation on the indoor-outdoor air circulation. The C ig of PM 2.5 was negative, indicating the very limited contribution of indoor sources to indoor PM 2.5 concentration. A similar finding was reported by Sangiorgi et al. [5] in cold conditions.
As with the I/O ratio, the F INF of major chemical components also followed the order of SO 4 2− > OC > EC (>PM 2.5 ) > NH 4 + > NO 3 − , which could be attributable to the semi-volatile characteristic of NH 4 NO 3 and the non-volatile characteristic of (NH 4 ) 2 SO 4 and EC. Temperature is the main factor affecting the gas-particle conversion of semi-volatile compounds [5,14,40]. During the transfer of NH 4 NO 3 from outdoor to indoor, the higher indoor temperature (21 • C on average) compared with the outdoor temperature (10 • C on average) led to the partitioning from the particle phase to the gas phase [5,26]. However, (NH 4 ) 2 SO 4 and EC were not affected by the temperature. The C ig values for major chemical components were positive, except for NH 4 + . The C ig values of SO 4 2− , OC, EC, NH 4 + , and NO 3 − only accounted for less than 20% of their indoor concentrations, implying that the indoor sources of PM 2.5 were negligible, which was consistent with the characteristics of the selected office.

Source Identification by PMF
According to the relationships of the indoor and outdoor PM 2.5 , I/O ratios, and the F INF mentioned above, indoor PM 2.5 pollution was primarily determined by the outdoor pollution, which suggests that indoor and outdoor PM 2.5 in Jinan had the same or similar sources. Thus, PMF was performed using the combined datasets of the PM 2.5 chemical components (SO 4 2− , NO 3 − , NH 4 + , OC, EC, Cl, K, Al, Fe, Ca, Ti, Mn, Cu, Cr, Se, Ni, As, Zn, and Pb) in both indoor and outdoor environments to obtain the possible sources of PM 2.5 . Six sources were identified, and the source profiles and contribution of each source to PM 2.5 are described in Figures 6 and 7. These major sources include secondary aerosols, local road dust, long-range dust storm, cement production, and coal combustion/metal smelting.  performed using the combined datasets of the PM2.5 chemical components (SO4 2− , NO3 − , NH4 + , OC, EC, Cl, K, Al, Fe, Ca, Ti, Mn, Cu, Cr, Se, Ni, As, Zn, and Pb) in both indoor and outdoor environments to obtain the possible sources of PM2.5. Six sources were identified, and the source profiles and contribution of each source to PM2.5 are described in Figures 6 and 7. These major sources include secondary aerosols, local road dust, long-range dust storm, cement production, and coal combustion/metal smelting.  EC, Cl, K, Al, Fe, Ca, Ti, Mn, Cu, Cr, Se, Ni, As, Zn, and Pb) in both indoor and outdoor environments to obtain the possible sources of PM2.5. Six sources were identified, and the source profiles and contribution of each source to PM2.5 are described in Figures 6 and 7. These major sources include secondary aerosols, local road dust, long-range dust storm, cement production, and coal combustion/metal smelting.  Factor 1 showed a high loading of Cl (56% of the total Cl mass concentration) and low loadings of other metals such as Zn, Pb, and Fe, which could be associated with waste incineration [58,59]. The contribution of this factor was minor (4% and 1% to the outdoor and indoor PM 2.5 concentrations, respectively; Figure 7). Factor 2 showed a high loading of Ca (contributing 59% to the total Ca concentration) and moderate loadings of Fe, Ti, NO 3 − , and EC, potentially highlighting links with local road dust resuspended by moving traffic [60,61]. The contributions of factor 2 were 17% and 13% to the outdoor and indoor PM 2.5 concentrations, respectively (Figure 7). Factor 3 was primarily associated with long-range dust storm due to high loadings of Ti, Fe, Al, and Ca. As shown in Figure 8, a dust storm from Mongolia affected Jinan from 27-30 November and from 3-5 December 2018, when the observed hourly PM 10 concentrations reached 581 µg m −3 and the ratio of PM 2.5 /PM 10 declined to 0.20-0.40 (as compared with >0.40 outside the dust storm period). This dust storm caused serious air pollution in Jinan and increased the hourly air quality index to 481 (shown in Figure 8, the hourly data of PM 2.5 , PM 10 and AQI from http://fb.sdem.org.cn: 8801/AirDeploy.Web/AirQuality/MapMain.aspx). In addition, high loadings of NO 3 − and NH 4 + (22% and 11% of factor mass, respectively) indicated that the dust plumes had aged during the transport. The time series of PMF results further showed that the contribution of long-range dust storm to PM 2.5 was reasonable (Figure 8). This factor accounted for 6% and 10% of the outdoor and indoor PM 2.5 concentrations, respectively (Figure 7). and the ratio of PM2.5/PM10 declined to 0.20-0.40 (as compared with >0.40 outside the dust storm period). This dust storm caused serious air pollution in Jinan and increased the hourly air quality index to 481 (shown in Figure 8, the hourly data of PM2.5, PM10 and AQI from http://fb.sdem.org.cn:8801/AirDeploy.Web/AirQuality/MapMain.aspx). In addition, high loadings of NO3 -and NH4 + (22% and 11% of factor mass, respectively) indicated that the dust plumes had aged during the transport. The time series of PMF results further showed that the contribution of long-range dust storm to PM2.5 was reasonable ( Figure 8). This factor accounted for 6% and 10% of the outdoor and indoor PM2.5 concentrations, respectively (Figure 7). Factor 4 had the highest loadings of NO3 -, SO4 2− , and NH4 + . NO3 -, SO4 2− , and NH4 + in this factor were major contributors to the PM2.5 concentrations of factor 4 (51% and 20%, respectively) and also contributed 63%, 30%, and 51%, respectively, to the total NO3 − and NH4 + concentrations. This factor could thus be identified as secondary aerosols. In addition, many other components (e.g., Se, As, Pb, OC, and EC) were observed with moderate loadings and contributions in factor 4, potentially highlighting links with coal combustion and vehicle emissions [62,63]. The secondary aerosols contributed an average of 49% and 25% to the outdoor and indoor PM2.5 concentrations, respectively (Figure 7). Factor 5 included Ni, Al, Ca, Cr, and Cu, which may be associated with cement production [64]. This factor contributed 11% and 26% to the outdoor and indoor PM2.5 concentrations, respectively ( Figure 7). factor were major contributors to the PM 2.5 concentrations of factor 4 (51% and 20%, respectively) and also contributed 63%, 30%, and 51%, respectively, to the total NO 3 − and NH 4 + concentrations.
This factor could thus be identified as secondary aerosols. In addition, many other components (e.g., Se, As, Pb, OC, and EC) were observed with moderate loadings and contributions in factor 4, potentially highlighting links with coal combustion and vehicle emissions [62,63]. The secondary aerosols contributed an average of 49% and 25% to the outdoor and indoor PM 2.5 concentrations, respectively ( Figure 7). Factor 5 included Ni, Al, Ca, Cr, and Cu, which may be associated with cement production [64]. This factor contributed 11% and 26% to the outdoor and indoor PM 2.5 concentrations, respectively ( Figure 7).
Factor 6 was suggested to be associated mainly with coal combustion [60,61], because it was characterized by the high loadings of As, Se, Zn, OC, and EC and contributed 59%, 43%, 41%, 42%, and 39% to the total As, Se, Zn, OC, and EC concentrations, respectively. Coal is still a primary energy source in Shandong Province, particularly during the heating season, with As, Se, Pb, Cr, Mn, Cu, and Zn as representative element tracers [62]. This factor also contained large contributions from Mn, Cu, and Pb and was associated with metal smelting [65]. This factor contributed 13% and 25% to the outdoor and indoor PM 2.5 concentrations, respectively (Figure 7).

Conclusions
To understand the pollution characteristics and sources of indoor PM 2.5 , PM 2.5 samples were collected simultaneously from inside a typical office and its outdoor atmosphere at an urban site in Jinan from 12 November to 5 December 2018. Subsequent chemical analysis revealed high PM 2.5 concentrations in both indoor and outdoor environments.
The chemical compositions of the indoor and outdoor PM 2.5 were similar, with SO 4 2− , NO 3 − , NH 4 + , and OC being the predominant components. A comparison with the results of previous studies in urban Jinan revealed significant decreases in PM 2.5 and SO 4 2− concentrations but an increase in NO 3 − concentration in recent years. The results of elemental risk assessment indicated the potential risk of PM 2.5 to health. The obtained I/O ratios and F INF indicated that indoor PM 2.5 pollution was largely affected by the outdoor environment. No significant indoor sources were found for the major chemical components of PM 2.5 based on the low C ig values, as the contributions of all sources to the indoor PM 2.5 concentrations were less than 20%. The main sources of PM 2.5 in Jinan were found to be secondary aerosols, local road dust, long-range dust storm, cement production, and coal combustion/metal smelting.