A small (2.0 L displacement) spark-ignited passenger car engine was used for the measurements with Russian pipeline natural gas as fuel. The methane content of the fuel was 97 %, other hydrocarbon content was 1.6 % and nitrogen content was 0.9 %. The sulfur content was below 1 ppm. The engine was run at two steady-state engine operation modes with torque of 70 Nm and speed of 2700 rpm (mode 1, M1) and torque of 35 Nm and speed of 3100 rpm (mode 2, M2). In engine mode 2, short-chain hydrocarbons were added into the exhaust to make it resemble the exhaust of a power plant NG engine. The exhaust gas composition in two operation modes simulated typical natural gas power plant exhaust gas composition. The engine, natural gas and lubricating oil properties as well as the engine operation modes have been described in more detail by , and .

Two separate after-treatment systems were applied in the measurements, both consisting of a reductive and an oxidative section. The after-treatment has been described in more detail by . The first catalyst (catalyst 1, C1) consisted of only one reactor, which targeted both oxidation of carbon compounds and NOx reduction through urea injection in the same catalyst reactor. The second catalyst system (catalyst 2, C2) consisted of a palladium- and platinum-containing methane oxidation catalyst followed by urea injection and a vanadium-SCR (selective catalyst reduction) catalyst, which were supported on metallic honeycomb substrates. Catalyst 1 was used in three different exhaust temperatures in the range of 350–450 ∘C in order to study its performance and its influence on secondary particle formation potential of the engine exhaust. The temperature of the catalyst 2 was 500 ∘C. Catalyst performance depends on the exhaust temperature e.g.,. By using the catalysts at different temperatures, effects of catalyst temperature on the formation and characteristics of primary and total aged particulate matter could be studied. The catalyst temperatures were measured upstream of the oxidation catalysts. The temperature prior to the SCR of catalyst 2 was approximately 50 ∘C lower than prior to the oxidation catalyst. The exhaust flow through the catalysts was kept constant at 80 kgh-1 by leading only a part of the exhaust gas flow through them .

The sampling system or the particle measurement instruments consisted of a porous tube diluter (PTD, ; ) with a dilution ratio (DR) of 6, followed by a residence time chamber with a residence time of 6 s. The low primary dilution ratio was used because of the very low primary particle emission concentrations see. The dilution air was heated to 30 ∘C to achieve constant dilution conditions. A second dilution stage was carried out with an ejector diluter (Dekati Ltd.) with DR 4. The dilution ratio over the PTD was adjusted using a bypass flow mass flow controller placed downstream of the residence time chamber (Fig. 1). The dilution ratios were calculated from CO2 concentrations in the raw and diluted exhaust samples and they could be used to calculate tailpipe concentrations of particle emissions. The aerosol sampling was done downstream of the exhaust after-treatment system.

A potential aerosol mass (PAM) chamber was used to simulate the aging process of an aerosol in the atmosphere. In the PAM, an oxidative environment (O3, OH and HO2, UV light) was produced using two UV lamps emitting 185 and 254 nm radiation, respectively, in a small (13 L) flow-through chamber. The PAM chamber was placed between the two dilution stages, and the flow through it was a constant 5 Lmin-1 (residence time 156 s), measured by a bubble flow meter (Gilibrator, Sensidyne Inc.) and adjusted by a pressure regulator of the compressed air flow to the ejector diluter. The PAM chamber could be either bypassed or used to measure the properties of primary and secondary aerosols, respectively.

The approximate atmospheric age, i.e., the photochemical age simulated by the PAM chamber UV lights, was modeled using the properties of the PAM chamber and the measured concentrations of gaseous components that cause external OH reactivity in the chamber. The model used for calculating the OH exposures was based on the degradation mechanism extracted from the Master Chemical Mechanism or MCM v3.3.1 via the website http://mcm.leeds.ac.uk/MCM and translated to Matlab code using the kinetic preprocessor or KPP . The model has been tested against SO2 reduction measurements in the PAM chamber.

MCM is a near-explicit chemical mechanism that describes the atmospheric degradation of volatile organic compounds in gas phase. MCM describes the degradation of a given volatile organic compound (VOC) through different generations of products until CO2 is ultimately formed. It contains about 17 000 reactions for 6700 different species . To be able to use these mechanisms with the PAM chamber, the photolysis rates have been calculated for ultraviolet light with wavelengths of 185 and 254 nm. The absorption cross section and quantum yield values needed for this are IUPAC recommendations supplemented with the Jet Propulsion Laboratory (JPL) data evaluation number 18 when necessary. Some photolysis reactions that are relevant to the PAM chamber but missing from the tropospheric MCM schemes have also been added.

KPP is a software tool for translating kinetic chemical mechanisms into Fortran 77, Fortran 90, C or MATLAB simulation code. The generated code produces concentrations of each species present as a function of time . For the model used, the KPP source code was modified to fix certain conflicts involving the MCM mechanism and the photolysis rate calculations written for the PAM chamber, as well as to allow large chemical schemes typical to MCM.

In this paper, we describe the PAM OH exposure as photochemical age, which is the equivalent time in the atmosphere in which the sample would reach the same OH exposure as in the PAM chamber. Thus, Photochemical age (days)=OH exposure1.5×106moleccm-313600h-1s×24d-1h, where 1.5×106 moleccm-3 is the average OH concentration in the atmosphere .

Relative humidity (RH) was measured downstream of the PAM chamber. The RH was high, about 80 %, due to the low primary dilution ratio that was applied during the experiments. The high RH of the sample complicated the evaluation of the PAM background mass levels – i.e., the particulate mass that was generated by only compressed air and UV lights – because the high RH could not be reproduced in the compressed air by the available instrumentation. The background levels were measured using both dry compressed air and compressed air with RH ∼30 %.

NOx concentration was measured with a chemiluminescence detector (CLD), CO and CO2 concentrations in raw exhaust were measured with a non-dispersive infrared (NDIR) analyzer and CO2 concentrations in diluted exhaust were measured with a Sick Maihak SIDOR gas analyzer. Water, methane, NH3, HNCO and the ratio of NO and NO2 were measured with a Fourier transform infrared spectroscopy (FTIR, Gasmet Cr-2000) analyzer, and methane, ethane, propane and ethylene were measured with a gas chromatograph (GC).

Aerosol instruments covered a large range of particle mobility and aerodynamic size as well as measurements of the particles' chemical composition. An engine exhaust particle sizer (EEPS, TSI Inc., ) and a high-resolution low-pressure impactor (HRLPI, ) were employed, both on 1 s time resolution, to measure the particle number, mass and size. The EEPS measures the size distribution and concentration of particles with a mobility diameter of 5.6–560 nm, and the HRLPI measures the aerodynamic size distribution of particles with diameter of ∼5–200 nm. The mass of the particles was calculated under assumptions of unit density and spherical particles. EEPS default inversion was applied.

A soot particle aerosol mass spectrometer (SP-AMS, Aerodyne Research Inc., US) – a combination of a high-resolution time-of-flight aerosol mass spectrometer and a single particle soot photometer (Droplet Measurement Technologies) – was used to measure the chemical composition and oxidative state of the aerosol sample. The SP-AMS measures both refractory and non-refractory particulate matter. It operated in V mode with a 1 min time resolution, measuring half of the time in mass spectra (MS) mode and the other half in particle size (pToF) mode. Both laser and tungsten vaporizers were used. The collection efficiency applied in the calculations was calculated using the parameterization by . The CO2 gas interference in the AMS data was corrected by using the CO2 concentrations measured with the Sick Maihak SIDOR gas analyzer. The impact of ammonium nitrate interference on CO2+, O : C and H : C ratios was evaluated to be small (less than 5 % for O : C and H : C; ). Therefore, a correction of ammonium nitrate interference was not applied for organics or O : C data.

Volatility measurements were made with a thermodenuder (TD) described in the publication by . When the remaining mass of particles was measured as a function of TD temperature, the thermodenuder was heated up to 265 ∘C and then switched off, with the sample flow still flowing through it. The decreasing temperature was recorded for at least half an hour until the temperature was below 50 ∘C.

Emission factors were calculated from fuel composition and engine performance information. Residual O2 in the exhaust was 6.2–6.3 %, the power of the engine was 12 and 20 kW and the combustion air flow into the engine was approximately 100 and 115 kgh-1 in engine modes 1 and 2, respectively. Calculated from the fuel composition information, the emission factor for CO2 EFCO2 was 2730 g kgfuel-1, and the carbon intensity was 0.74 kgC kgfuel-1.

The concept “total aged” here comprises all particle mass measured downstream of the PAM chamber, i.e., both primary and secondary particle mass. In general, primary particle mass has not been subtracted from the mass measured downstream of the PAM (total aged) to calculate the secondary particle mass separately because doing so would have created inconsistency in representation of the results, since, for example, particle size distributions or volatility behavior cannot be presented in this way. For the same reasons, the PAM background mass – i.e., the particle mass generated in the PAM chamber from clean compressed air – has not been subtracted but is instead presented separately in the Supplement of this paper. To enable a comparison to literature, an exception is made when presenting secondary particle production factors (PFs).

Figure 2 contains particulate mass measurement results derived from the three aerosol instruments. The chemical compositions from SP-AMS are also presented. The cases (engine mode, catalyst and catalyst temperature) included in this paper cover all of the tested exhaust temperatures and both engine operation modes, and they have data collected with all available instruments of both the primary and total aged aerosol measurements. In most cases in our measurements, primary exhaust particle mass concentrations from the natural gas engine were close to the detection limits of the instruments EEPS, HRLPI and SP-AMS . Exceptions were made by the high temperature catalyst cases (M2, C2, 500 ∘C and M2, C1, 450 ∘C) during which more primary particle mass was formed, especially on the size ranges of the HRLPI and EEPS: a high catalyst temperature favors the conversion of SO2 into SO3 and further into sulfuric acid, which can nucleate and condense on existing particles in the sampling process or when released into the atmosphere see, e.g.,. The primary particle formation phenomena and concentrations have been discussed in more detail in while this paper focuses on secondary aerosol formation and the total aged particle emissions.

The formed secondary particle mass concentrations were found to be significantly high in comparison with the primary particle mass emissions. In all of the investigated cases, particle mass increased when the sample was led through the PAM chamber. The increase in mass could be magnitudes larger than the primary particle mass emission (Table 1). The relative increase in mass in the PAM chamber could not be specified for all of the HRLPI measurements because of the very low primary particle mass. The total aged aerosol mass produced by natural gas engine exhaust was 0.99–2.6 mgm-3 according to SP-AMS, leading to secondary mass production of 0.96–2.5 mgm-3.

The secondary aerosol formation – i.e., the ratio of the total aged particulate mass to the primary particulate mass – was lower in the cases that already produced more primary particle mass, i.e., in the cases with a high catalyst temperature. It is possible that if the catalyst conditions are favorable, the particulate matter that would otherwise condense on particles in the PAM chamber condenses on the particle phase already in the cooling and dilution processes. In other words, if the catalyst sufficiently oxidizes the exhaust gases, thus lowering their saturation vapor pressure, they condense or nucleate already when released from the tailpipe and not later on in the atmosphere. A high catalyst temperature promoted larger total aged aerosol formation, according to the EEPS and HRLPI measurements. However, the total aged mass concentrations of the SP-AMS did not increase as catalyst temperature increased. The differences in the instruments' showings are discussed in Sect. 3.3. Also, the variation in the atmospheric ages increases uncertainty in the comparison of the catalyst temperature on secondary aerosol formation potential.

The total aged aerosol of the natural gas engine exhaust consisted of both organic and inorganic matter at the tested operating conditions (Fig. 2, Table 1). Approximately half of the total aged aerosol particle mass detected by SP-AMS consisted of organic matter. The fraction of sulfate and nitrate was measured to be 34–49 % in total, with their ratio depending on the case, and the fraction of ammonium varied between 10 and 15 %. found that even high NOx emissions can produce negligible amounts of secondary nitrate aerosol if related ammonia emissions are small. Because secondary ammonium nitrate aerosol formation is limited by ammonia, its formation is probably more related to the exhaust after-treatment than the fuel. The exact ammonia concentrations in the raw exhaust cannot be given because they were below the instrument detection limit of 2 ppm. According to these measurements, also low ammonia emissions may have atmospheric importance as secondary inorganic aerosol precursors.

The organic fraction of the total aged aerosol consisted of hydrocarbon fragments (CxHy), fragments with one oxygen atom (CxHyO) and fragments with more than one oxygen atom (CxHyOz,z>1); there was little or no CxHyN fragments (hydrocarbons with nitrogen) in the total aged particles. The main secondary organic ions detected by the SP-AMS were CO2+, CHO+ and C2H3O+. The composition of the organic aerosol was similar in all of the cases: the CxHyOz,z>1 group was the largest, followed by CxHy and CxHyO. The source of the secondary organic aerosol could be either the natural gas or the lubricating oil. However, we are not able to tell the source based on these measurements. The fuel mainly consisted of light hydrocarbons that are unable to form secondary organic aerosol e.g.,pp. 575. For example, and have suspected engine lubricating oil to be responsible for a large portion of engine-emitted particles. Therefore, we believe that also lubricating oil is a potential candidate for the source of secondary aerosol.

The O : C ratios of the total aged aerosol measured by SP-AMS were between 0.9 and 1.2. The O : C ratio of the primary aerosol in the case with the largest concentration was slightly smaller (1.1) than the O : C ratio of the total aged aerosol in the same case (M2, C2, 500 ∘C). In all of the other primary aerosol measurements, the particle mass concentrations in the sample were too low for O : C ratio analysis. In comparison with a secondary aerosol emission study on gasoline engines by , the observed O : C ratios in the total aged aerosol from the PAM chamber were rather high.

The emission factors or secondary aerosol production factors in different units can be calculated from the presented particle mass concentrations by using the following factors. If a unit factor in mg kgfuel-1 is needed, a factor of approximately 22 m3 kgfuel-1 can be applied to multiply the particle concentration calculation, e.g., in. In order to obtain emission and production factors in units of kWh-1, a factor of 2.7 m3kWh-1 (mode 1) or 4 m3kWh-1 (mode 2) can similarly be used. These factors are derived from the fuel composition and engine performance information provided in Sect. 2.1 and 2.2 and the exhaust CO2 concentration.

The production factors of secondary organic aerosol have been calculated and collected in Table 2, in units of kgfuel-1. To be able to compare the SOA production factors, here primary organic aerosol was subtracted from the total aged organic aerosol. Table 2 also contains SOA production factors of secondary organic aerosol for different diesel and gasoline vehicles obtained from the literature. Although the total aged particulate matter production of the investigated NG engine was much larger than its primary particle emissions, it was smaller than SOA production from in-use diesel and gasoline vehicles in the literature . The SOA formation potential from the NG engine, measured by SP-AMS, was similar to that of a diesel vehicle equipped with a catalytic converter or to that of a hot-start gasoline vehicle. On the other hand, the photochemical age that was simulated by a chamber in the different studies varied greatly. This is why the comparison of the SOA production factors should be done very carefully, if at all. The longest atmospheric ages in the literature collected in Table 2 were achieved in our study.

, and have seen with an oxidation flow reactor – such as the PAM chamber in our experiments – that the highest potential secondary organic aerosol formation takes place at the photochemical age of a few days and starts decreasing after that. For example, in the vehicle fleet emission study in a highway tunnel of , the peak secondary aerosol production took place after 4–10 days of equivalent atmospheric oxidation ([OH] = 1.5×106 moleccm-3), and larger OH exposures started to reduce the secondary mass. In our study, the simulated atmospheric, or photochemical, ages in the investigated cases varied between 4.6 and 10.7 days, depending on the external OH reactivity, which was affected by the concentrations of gaseous emissions (Table 1) entering the PAM chamber and by the relative humidity of the sample. The largest total aged particle concentrations were achieved with the longest atmospheric ages and the lowest particle concentrations were achieved with the shortest atmospheric ages. However, the secondary aerosol formation potential may have also been affected by the engine parameters and not only the achieved photochemical age: the total aged particle concentrations were the highest in engine operation mode 2 (M2).

The volatility of the particles was studied with a thermodenuder. Mass fraction remaining (MFR) stands for the fraction of the particle mass at a given thermodenuder temperature and the particle mass at room temperature. In Fig. 3, the particle mass fraction remaining has been calculated for two representative cases of primary emissions and four representative cases of total aged particle emissions, selected from among the cases already introduced. Figure 3 only shows data from EEPS, since the curves obtained from HRLPI were similar. Here, the curves have been smoothed by a moving average but the original 1 s resolution figure can be found in the Supplement. For the total aged emissions, the cases with both higher and lower catalyst temperature are presented. For primary particle emissions, only the case with the higher catalyst temperature is presented. This is because an accurate examination of the volatility of primary particles in low catalyst temperatures could not be done, due to the insufficient primary particle mass concentrations for high-quality analysis. In the figure, the “starting point”, i.e., the temperature where the mass fraction remaining is one, is 50 ∘C and not lower because of the decelerated cooling of the thermodenuder toward the room temperature and related time limitations.

The MFR curves for each type of particles are characteristic, i.e., each particle type can easily be distinguished by their evaporation behavior. To highlight this, the primary particle evaporation is marked with black, and the total aged particle evaporation curves are marked with cyan and blue in Fig. 3. The volatility of the particles from the natural gas engine clearly changed when the particles were aged. At high catalyst temperature, the primary particles (black triangles in Fig. 3) were more volatile than the total aged particles (cyan squares). Approximately half (46–60 % in EEPS, 43–53 % in HRLPI) of the total aged particle mass remained at a thermodenuder temperature of 250 ∘C, whereas only 5–10 % (1–4 % in HRLPI) of the primary particle mass remained at that temperature. Also, the catalyst temperature had an impact on the volatility of the total aged particles (blue vs. cyan). An easily evaporable fraction of the total aged particles was formed in the case of a low catalyst temperature, which evaporated below 110 ∘C. Because of this easily evaporable fraction, the MFR of total aged particles at 250 ∘C was 30 % in the low catalyst temperature cases, while in the high catalyst temperature cases the MFR of total aged particles at 250 ∘C was 46–60 %.

The thermodenuder used in this study has been designed to minimize nanoparticle losses by reducing the residence time . For example, in this study, the residence time in the heated zone of the thermodenuder was less than 1 s. measured the volatility of secondary organic aerosol produced during α-pinene photo-oxidation and observed that only half of the secondary particle mass evaporates in a thermodenuder (100 ∘C) if the residence times in the heated zone of the thermodenuder are less than a few seconds. With longer residence times, the remaining mass downstream of the thermodenuder decreases to less than 3 %. This means that the remaining fraction of particle mass in our study could have been smaller with longer residence times in the thermodenuder. On the other hand, a longer residence time in the thermodenuder would have increased the nanoparticle losses. In this study, with the use of a thermodenuder, we could observe the volatility differences between the different types of particle emissions emitted by a natural gas engine.

In Fig. 4, the remaining mass fractions are plotted for different chemical species of the particles drawn from the SP-AMS. In the primary emission case, approximately one-third of the particle mass – consisting mainly of organics – remained at TD temperature of 250 ∘C. The low concentrations and particle size below the detection limit of SP-AMS degrade the analysis in the case M2, C1, 450 ∘C, which was seen as a fluctuating signal. About 25 % of the total aged particle mass in the high catalyst temperature cases and less than 10 % of the total aged particle mass in low catalyst temperature cases remained at 250 ∘C, according to SP-AMS. The remaining particle matter consisted of organics, sulfate and ammonium, in this order.

The composition information reveals that the high-volatility fraction of the total aged particles in the low temperature catalyst cases consisted of nitrates, possibly of ammonium nitrate and high-volatility organics. The primary particle sulfate evaporated at thermodenuder temperatures between 100 and 170 ∘C, and the total aged particle sulfate evaporated more gradually above 120 ∘C. In all types of particles, the evaporation of organics was steady and gradual below 200 ∘C, indicating various organic compounds with different evaporation temperatures. Above 200 ∘C, the evaporation of organics decreased. This combined SP-AMS and EEPS/HRLPI derived thermodenuder temperature ramp information can be used in future measurements for particle composition analysis: the evaporation temperature of the particles can give valuable information about the composition of the particles also without access to SP-AMS.

The temperatures at which 50 % of the volatile fraction of the chemical compounds of the particles were remaining are collected in Table 3. The case “primary, M2, C1, 450 ∘C” had particle mass concentrations that were too low (see Fig. 4) for this kind of examination. The evaporation temperatures of sulfate and nitrate were the highest and the lowest, respectively, in all of the analyzed cases (all catalyst temperatures; primary and total aged particles). Similarly to , who measured ambient aerosol volatility in megacities with a thermodenuder and an SP-AMS, we found that nitrate had the highest volatility and sulfate had the lowest.

and proposed that all organic aerosol should be considered semivolatile. Our results on primary and PAM-chamber-generated organic aerosols point in that direction as well. The evaporation temperature of the volatile fraction (Tvolatile, 50 %) of organic matter remained between the Tvolatile, 50 % of nitrate and sulfate in all cases. Also, a significant fraction of the mass concentration of the organic matter did not evaporate. More exact specifications of the volatility cannot be given, but there is room left for speculation if part of the organic matter in secondary particles is semi-volatile (SV-SOA) or low volatility (LV-SOA) secondary organic aerosol .

The ammonium in total aged particles evaporated at higher thermodenuder temperatures when the catalyst temperature was high. The theory that the sulfate–nitrate trade-off phenomenon that determines the formation of nitrates is ammonium bound is supported by the evaporation temperatures of ammonium. Ammonium evaporated at approximately 20 ∘C higher thermodenuder temperatures in the high catalyst temperature cases (Table 3); thus, its evaporation temperature was closer to the evaporation temperature of sulfate when the sulfate concentration of the particles was larger. By contrast, in the low catalyst temperature cases where the nitrate concentration was higher, the evaporation temperature of ammonium was closer to that of nitrate.

The nitrate concentrations measured during the thermodenuder temperature ramp (Fig. 4) in low thermodenuder temperatures differed from the nitrate concentrations that were measured without a thermodenuder (Fig. 2a and b) in total aged particles. A possible explanation is that a long time is needed for the nitrate concentration to stabilize. In our measurement protocol, we waited 10–15 min after switching the PAM UV lights on, followed by a 10 min steady-state measurement with the aerosol instruments. After this, a thermodenuder ramp was started, which took approximately 45 min. Based on the results, the 10–15 min wait was insufficient if accurate nitrate concentrations were desired. Therefore, the chemical compound measurements performed at low thermodenuder temperature can give a truer picture of the secondary aerosol formation than the measurements presented in Fig. 2. The change in concentrations between the steady-state measurements and the thermodenuder ramp measurements was the largest for nitrate, but the concentrations of other compounds also differed slightly from each other. Because the nitrate concentrations were found to be the slowest to stabilize and the most sensitive to changes in the system, such as to changes in temperature, special attention should be given to measurements of nitrate, especially when a PAM chamber is being used. We note that because nitrate formation is limited by ammonium, the slow stabilization is probably related to ammonia.

According to our thermodenuder temperature ramp experiments, the catalyst temperature affected the total aged particle composition. With a decreasing catalyst temperature, the mass concentration and fraction of sulfate in total aged particles decreased (Fig. 4, Table 1). This was expected: at lower catalyst temperatures the oxidation of SO2 to SO3 decreases and less sulfuric acid (sulfates) can form . The mass concentration of nitrate in secondary particles increased as the catalyst temperature decreased. This could not be explained by catalyst performance improvement: gaseous NOx levels remained similar at all catalyst temperatures or rose as catalyst temperature increased see. Because ammonia concentrations after catalyst were low, below 2 ppm in all cases, the effect of catalyst temperature on ammonia emission could not be measured. However, ammonium concentrations measured by SP-AMS correlated rather well with nitrate concentrations. Therefore, we suggest that ammonium increase was related to the nitrate increase. Also, the sulfate concentrations could partly explain the behavior of the nitrate concentrations. If enough gaseous sulfuric acid is available, ammonium sulfate forms, and if not, more ammonium nitrate can form instead. A similar behavior of nitrate and sulfate trade-off has been measured by for two different marine fuels, namely heavy fuel oil (HFO) and light fuel oil (LFO).

Slightly unexpectedly, the total aged particle mass measured by SP-AMS was 2–4 times larger than the total mass measured by EEPS and 1–3 times larger than that measured by HRLPI (Fig. 2). There could be several reasons for this. In EEPS and HRLPI, unit density and spherical particles were assumed in the mass calculations. Natural gas engine primary particles have a density of 0.85 gcm-3 , but the densities of natural gas engine secondary particles can be larger than the unit density. For example, the density of ammonium nitrate, ammonium sulfate and sulfuric acid is approximately 1.5, 1.5 and 1.8 gm-3, respectively . Particle density does not completely explain the difference in instrument readings. Also, the collection efficiency (CE) estimation used in the SP-AMS calculation is probably not the reason for the differences between the instrument results in this study. Evaluation of the CE following the procedure of revealed that CE equal to 0.45 was the correct value for the studied total aged particles in the cases in Fig. 1.

However, the detection efficiency and size range varied among the aerosol instruments (EEPS 5.6–560 nm, HRLPI ∼ 5–200 nm, SP-AMS∼30–1000 nm) and can explain the differences in results; HRLPI can detect a larger fraction of the primary particles than SP-AMS because of the more suitable size range of the instrument and, correspondingly, SP-AMS can detect a larger fraction of the total aged particles formed in the PAM chamber because of its more suitable size range. Also, particle losses may play a role in the differences between instruments; particle losses in the PAM chamber were larger in the HRLPI size range than in the SP-AMS size range. Nevertheless, most probably, the largest role was played by the differences in instrument size ranges.

Mass size distributions of the total aged aerosol, measured with EEPS and HRLPI, are plotted in Fig. 5. HRLPI suggests that a part of the particle mass lies above the instrument size range, which was confirmed by SP-AMS mass size distributions in Fig. 6. According to SP-AMS, the mass size distributions of total aged particles were bimodal, with the size of the larger mode being 480–840 nm and the smaller being 150–200 nm. The mode with smaller particle size was dominated by organics. Although the mass concentration of the total aged particles was better recorded by SP-AMS, a portion of the particles on the smallest particle sizes was missed due to the lower limit of SP-AMS size range at 30–50 nm. The best overall picture is therefore gained with a combination of SP-AMS and HRLPI. See the Supplement for a comparison of the size distributions measured by different instruments in the same figure. The two instruments that measure the aerodynamic diameter of the particles (HRLPI and SP-AMS) compare quite well with each other in the size range 47–124 nm.

We can also see a difference between the EEPS and HRLPI mass size distributions. The difference is probably due to the inversion of EEPS, which forces the size distributions to follow a log-normal shape. EEPS also underestimated the mass of particles with diameter above 200 nm (see the Supplement). The different measurement principles of the instruments must also be kept in mind. EEPS measures the mobility size and HRLPI measures the aerodynamic size of the particles.

The so-called smog chambers are an established method of measuring SOA formation. An oxidation flow chamber such as PAM provides some advantages in comparison to smog chambers, such as a higher degree of oxidation, smaller physical size and a short residence time, which allows measurements with higher time resolution. Smog chamber walls may also cause large wall losses and influence the chemistry in the chamber . On the other hand, smog chambers may simulate the atmospheric oxidation of organic precursors better than oxidation flow chambers due to their more tropospheric oxidant concentrations and longer residence times .

The PAM method has been designed to produce the maximum potential aerosol mass from precursor gases . In that stage, the oxidation products of precursors have condensed into the particle phase and formed secondary aerosol. However, because the oxidant concentrations are unrealistically high in PAM, the UV-light intensity used is non-tropospheric and the residence times are much shorter than in the atmosphere e.g.,, precursor oxidation products also have other possible fates; they can be oxidized too far and form non-condensable oxidation products before condensation (accelerated chemistry) and they can exit the reactor before the condensation occurs. Also, precursor oxidation products can be lost on the PAM walls although the losses on the walls are minimized by the chamber design . The fates other than condensing on particle phase are viewed here as PAM artifacts and losses.

The losses of condensable organic oxidation products and artifact effects of the accelerated chemistry in the PAM have been evaluated following the method of for the cases in Fig. 2. HRLPI number size distributions were used to calculate the condensation sink needed in the loss calculation. A molar mass of 200 gmol-1, a diffusion coefficient of 7×10-6 m2s-1 and a rate constant for reaction with OH of 1×10-11 were applied. For sticking coefficient selection α=1 (assumed by ), the fraction of oxidation products that condensed into the particle phase was 0.94±0.03, but for α=0.1 the fraction of oxidation products that condensed into the particle phase was 0.64±0.15.

The losses of sulfuric acid were also calculated in the same way. A diffusion coefficient of 1×10-5 , α of 0.65 and molar mass of 98.079 gmol-1 were used for sulfuric acid. The fraction of sulfuric acid that condensed into the particle phase was 0.98±0.01. According to , SO2 losses in the PAM are negligible. It can be concluded that the effect of precursor losses and artifacts in the PAM was not substantial in our measurements. The measurement of ammonium nitrate and ammonium sulfate is difficult because ammonia sticks on the walls of sampling systems and instruments , which may result in wall losses or an artifact on subsequent measurements. The penetration of ammonia could not be calculated, but the measured ammonium concentrations varied clearly from one case to another, implicating that the source of the ammonia was indeed the exhaust line instead of, e.g., the walls of the PAM. However, longer times for the stabilization of the SP-AMS concentration would have been advantageous for the reliability of ammonium and, as a consequence, nitrate particle formation.

and estimated the effect of particle losses in a similar PAM chamber to be small. The particle losses measured by depend on particle size and are below 10 % at the particle sizes with most particle mass. An exact calculation of the particle losses in the PAM chamber is not possible because the particle size and number increase while the aerosol sample flows through the chamber. An estimation for the particle mass losses in the chamber can be given, calculated using the average of HRLPI particle number size distributions before and after the chamber (similarly to the precursor-loss calculation by ) and the PAM particle loss curve. The particle mass loss according to this examination was 20.3±3.7 %. Most probably, however, the actual particle mass losses in the chamber were smaller because majority of the mass actually was located at larger particle sizes that HRLPI is unable to measure, where particle losses are smaller.

No loss corrections were done based on these calculations on the results presented in this article. If corrections had been made, the presented secondary aerosol productions and production factors would be slightly higher (less than 10 %) in Figs. 2 and 4–6, and in Tables 1 and 2.