Large contribution to secondary organic aerosol from isoprene cloud chemistry

See allHide authors and affiliations

Science Advances  24 Mar 2021:
Vol. 7, no. 13, eabe2952
DOI: 10.1126/sciadv.abe2952


Aerosols still present the largest uncertainty in estimating anthropogenic radiative forcing. Cloud processing is potentially important for secondary organic aerosol (SOA) formation, a major aerosol component: however, laboratory experiments fail to mimic this process under atmospherically relevant conditions. We developed a wetted-wall flow reactor to simulate aqueous-phase processing of isoprene oxidation products (iOP) in cloud droplets. We find that 50 to 70% (in moles) of iOP partition into the aqueous cloud phase, where they rapidly react with OH radicals, producing SOA with a molar yield of 0.45 after cloud droplet evaporation. Integrating our experimental results into a global model, we show that clouds effectively boost the amount of SOA. We conclude that, on a global scale, cloud processing of iOP produces 6.9 Tg of SOA per year or approximately 20% of the total biogenic SOA burden and is the main source of SOA in the mid-troposphere (4 to 6 km).


Aerosols and clouds play a vital role in Earth’s energy budget (1). Organic aerosol (OA), a major fraction of the total submicrometer aerosol mass (20 to 50%), is overwhelmingly secondary (SOA; up to 90%), formed through gas-to-particle conversion of condensable oxidation products from gaseous precursors (2). While our knowledge of SOA production via gas-phase chemistry has substantially improved over the past decades, SOA production in clouds has received much less attention (35). Major SOA components, such as organic acids (4), shown to form in the aqueous phase during laboratory experiments (3, 6), have been measured in the field at higher concentrations than expected from direct emissions or gas-phase oxidation (79). Current state-of-the-art models can reasonably predict OA measurements within the boundary layer close to emission sources but are subject to greater uncertainties with altitude, where most clouds exist (i.e., 1 to 6 km) (10). Despite the potential importance of cloud chemistry in modulating OA chemical composition and concentrations, the production of in-cloud aqueous SOA (aqSOA) remains virtually nonquantifiable because of major experimental limitations (e.g., simulation tool and mass spectrometer resolution and sensitivity) related to simulating aqueous chemistry at diluted conditions characteristic of atmospheric clouds.

Isoprene represents half of the flux of the biogenic volatile organic compounds emitted into the atmosphere (11). Although its SOA mass yield through gas-phase photooxidation is subject to vigorous debate, numerous publications (fig. S1) clearly show relatively low SOA mass yields (≤5% on average), with some outliers obtained under specific conditions (e.g., highly acidic seeds). Despite its low SOA yield, isoprene is presumed to be one of the largest SOA sources at the global scale because of its high emission rate (12, 13). As its oxidation products are mainly water soluble (14), their chemical reactions in cloud water could constitute a substantial aqSOA source (15), and a neglected pathway that is competitive to SOA formation from the gas phase.

Here, we present the first experimental simulation of aqSOA formation from combined photochemistry and cloud chemistry from isoprene under atmospherically relevant conditions, using a newly developed wetted-wall flow reactor (WFR), which operates in continuous flow mode. The WFR was constructed in analogy to the “rotating wetted wall flow tube” from Hanson’s group (1620). Experiments were performed at 100% relative humidity (RH) and a temperature of 295 K; isoprene and its oxidation products were introduced into the WFR along with ozone, OH, and water vapor in the presence of light (case study A) and in the dark (case study B), where they were allowed to further react in the gas phase and in an aqueous microfilm (Materials and Methods and fig. S2). We used state-of-the-art mass spectrometers to fully characterize the properties of the formed organic molecules, such as yields, volatility, and solubility. Measurements were made online in the gas phase and offline in the liquid phase through subsequent nebulization of the aqueous solutions. AqSOA yields and production rates were determined, parameterized, and implemented into the atmosphere-only configuration of the U.K. Earth System Model (UKESM1). We used the U.K. Chemistry and Aerosol Model (21, 22) of UKESM1 to assess the impact of isoprene chemistry in cloud water on the global SOA burden.


Experimental findings

We measured isoprene and its gaseous oxidation products with a proton transfer reaction time-of-flight mass spectrometer (PTR-TOF-MS) and an acetate chemical ionization mass spectrometer (acetate-CIMS). Upon the rapid consumption of 8.4 parts per billion by volume (ppbv) of isoprene by OH radicals, oxidized vapors are formed, accounting for 98 ± 14% (1σ) of the total reacted carbon (phase II of case study A; Fig. 1A). Note that as in the ambient environment, WFR experiments simulate low NOx conditions where HO2-derived products such as isoprene epoxydiol (IEPOX) and hydroxy hydroperoxide (ISOPOOH) concentrations represent the atmospheric daytime conditions well (see the Supplementary Materials for detailed calculations) (23).

Fig. 1 Time series of isoprene oxidation products measured by mass spectrometry for a case study of type A and solubility-dependent uptake ratio.

The experiment consists of six phases (all at 100% RH and 295 K). Phase I: Conditioning of the WFR flow tube in the dark to reach a steady state mixing ratio of 10 ppbv isoprene. Phase II: Initiation of the photooxidation by switching on the Xe-excimer lamps (in front of the WFR) and the ultraviolet B (UVB) lamps (around the WFR). Phase III: Injection of the water microfilm, denoted by the blue arrow. Phase IV: OH oxidation of isoprene only with the Xe-excimer lamps, bypassing the WFR. Phase V (same as phase III): Reaction mixture through WFR with both the Xe-excimer and UVB lamps on and the water microfilm present. Phase VI: UVB lights off. Phase VII: Both Xe-excimer and UVB lamps off. (A) Carbon mass balance of the gas phase derived from PTR-TOF-MS and acetate-CIMS. The red line presents the isoprene decay measured by the PTR-TOF-MS, indicating 42-ppbvC reacted isoprene carbon (corresponding to 8.4 ppbv of consumed isoprene). Gray stacked areas denote the increase of unique oxidation products measured by both the PTR-TOF-MS (dark gray) and the acetate-CIMS (light gray), with molecules encountered in both instruments only being counted via the PTR-TOF-MS. Ions with the same molecular formula from both instruments comprised 2.8 ± 0.7 ppbvC, indicating only 7 ± 2% of carbon overlap. (B) Measured (solid lines) and modeled (dashed lines; using the QEMRA model and calculated effH values) evolution of selected nonsoluble (C3H6O) and soluble (C5H10O5) oxidation products. (C) Representative uptake ratio after 5 hours of cloud processing as a function of the effective Henry’s law constant (effH) color coded with the saturation vapor concentration (C*).

Once steady state was reached, we experimentally simulated the presence of cloud water by generating an aqueous microlayer on the WFR wall (Fig. 1A, phase III). A sudden decrease of the concentrations of the oxidation products was observed, demonstrating their uptake into the microfilm. Figure 1B shows the time dependence of representative species with different solubility. Less soluble compounds such as C3H6O quickly reach their aqueous saturation concentration, and their mixing ratios in the gas phase increase back to their initial values. More soluble compounds such as C5H10O5 experience a continuous and irreversible uptake. The uptake ratio ([CClearAir − CCloud]/CClearAir) was calculated for all detected gaseous species, after, e.g., 5 hours of cloud simulation, and is displayed in Fig. 1C against their effective Henry’s law solubility constant (effH, in M atm−1). effH was estimated on the basis of the compounds’ elemental composition using structure activity relationships [i.e., GROupcontribution Method for Henry’s law Estimate (GROMHE) (24); see the Supplementary Materials]. We have developed the kinetic model QEMRA (aqueous-hydrometeor phase processing and mass transfer model; see the Supplementary Materials), which describes the partitioning of molecules of different solubility. The model captures the experimental data accurately, validating the estimated effH. According to the model, 54% of the reaction mixture diffuses to the surface of the water microlayer, of which 91 ± 13% fully dissolve therein; the remaining fraction consisting of nonsoluble and nondiffused species escapes the WFR in the gas phase. Efficient uptake (≥95% of the maximum uptake) is observed for compounds with effH > 105 M atm−1, close to atmospheric cloud conditions, where one finds notable partitioning for species with effH ≥ 107 M atm−1 (fig. S3) (25). The aqueous OH radical concentrations and the median estimates (in cases A and B) of the total dissolved organic material reached after several hours of reaction are very similar to those found in atmospheric clouds (6), ranging between 1 × 10−11 and 6 × 10−13 M OH and between 122 and 40 μM organics, respectively. Previous aqSOA experiments, based on bulk solution, were conducted using highly concentrated mixtures (around millimolar), characteristic of wet aerosols rather than cloud water (around micromolar), and they were often restricted to the processing of single oxygenated molecules. Furthermore, they unrealistically favored molecules with only slight solubility [with effH as low as 10−5 M atm−1 (3, 25)], suggesting a poor representation of the chemical composition of the dissolved organic matter in cloud droplets and thus affecting the resulting chemistry (see fig. S3). In an attempt to study the aqueous processing under more realistic conditions, Brégonzio-Rozier et al. (26) used laboratory-generated cloud droplets, where the short lifetimes of these droplets (<10 min) and losses to the chamber walls prevented a quantitative assessment of aqSOA production yields. To the best of our knowledge, our experimental setup allows the first investigation of the partitioning of organic vapors at the air-water interface under near-ambient conditions. It provides a near-atmospheric simulation of the gas-phase radical chemistry at low NOx conditions, the product distribution in the gas phase, the product/oxidant levels and distributions in clouds, the typical exposure times to clouds (~hours), and the full range of complexity found in oxidation mixtures.

After several hours of aqueous phase processing, the water mixture was collected, spiked with a solution of (NH4)2SO4 (used as internal standard), nebulized, and dried, resulting in several micrograms per cubic meter of OA detected by an aerosol mass spectrometer (AMS). Nebulization of aqueous solutions produced during control runs following the same experimental procedure (see also the Supplementary Materials), but in the absence of isoprene, showed no detectable OA mass, demonstrating that the formed aqSOA is driven by isoprene oxidation and not affected by contamination in the reaction vessel. AqSOA bulk and detailed molecular composition were characterized by an AMS and an extractive electrospray ionization time-of-flight mass spectrometer (EESI-TOF-MS) (27), respectively, and used to identify the pathway by which the observed aqSOA was produced. Since the chemical composition for case studies of type A and B was found to be very similar (see fig. S4), the following discussion will only focus on case A.

The aqSOA AMS spectrum is characteristic of an aged OA with high contribution from oxygenated fragments, such as CO2+ and C2H3O+ (Fig. 2A), and an atomic oxygen-to-carbon (O:C) ratio of 1.1 ± 0.1, higher than typical values reported (~0.8) for laboratory-generated SOA from isoprene photooxidation in the gas phase (28). While acid-catalyzed reactions of IEPOX have been shown to be important for wet aerosols (29), the absence of an AMS signal at C5H6O+ characteristic of IEPOX SOA (30) demonstrates that this is not the major aqSOA formation mechanism. These reactions require very low pH (<3) levels but are extremely slow at the pH range typical of cloud water and of our experiments (4 < pH < 7) (15). We estimate that the acid-catalyzed hydrolysis lifetime of IEPOX [τH = 1/(kH+[H+] + knuc[nuc][H+])], with the nucleophile concentration [nuc] ≡ [H2O] = 55.55 M, kH+ = 3.6 × 10−2 M−1 s−1 (31), and knuc = 2 × 10−4 M−2 s−1 (32) would still be 59 hours, when considering the lowest pH (~4) reached under our conditions.

Fig. 2 Measured gaseous and aqSOA phase composition for a case study A.

(A) Representative AMS spectrum and (B) Kendrick mass defect (KMD) plot (with CHO as unit base) of aqSOA determined from the EESI-TOF-MS. (C) Detailed chemical composition of the gas (top;, measured by the PTR-TOF-MS and the acetate-CIMS) and aqSOA (bottom; measured by the EESI-TOF-MS after nebulization) phase. Oxidation products are binned according to carbon number (x axis) and color coded by the oxygen number. (D) Relative contributions (in moles) of isoprene oxidation products in aqSOA and in the gas phase that dissolved into the WFR water microlayer, as a function of their oxygen number. (E) Relative contributions (in moles) of different compounds to the total gas (top) and aqSOA phase. The “other” group is composed of numerous compounds that each make up, at most, a few percent of the total.

The aqSOA molecular composition determined by the EESI-TOF-MS is shown in Fig. 2B as a unit-CHO–based Kendrick mass defect plot and compared with the gas-phase reaction mixture in Fig. 2 (C to E). Homologous series of monomer (C ≤ 5) compounds constitute the major aqSOA fraction (84%), indicating that oligomerization plays a minor role in the formation of aqSOA, as expected from the diluted cloud conditions (25).

The aqSOA composition retrieved from the EESI-TOF is again highly oxygenated, with an average O:C ratio of 1.1, in excellent agreement with its bulk properties determined by the AMS. The top 10 molecules, accounting for 47% of the total aqSOA, are highly oxygenated monomers having four to six oxygen atoms, whereas the major gas-phase molecules contain only two to three oxygen atoms (average O:C ratio of 0.6) (Fig. 2, D and E). The oxygen number (#O) per molecule in aqSOA species is significantly higher than #O of species dissolved into the WFR water microlayer (determined as the difference in the gas-phase composition with and without the water microfilm). This shows that the dissolution of isoprene oxidized vapors cannot explain the aqSOA production and provides strong evidence that aqSOA results from OH reactions of the dissolved species in the aqueous phase. As an example, we estimate the lifetime of IEPOX toward aqueous OH reaction [τaqOH = 1/(kaqOH[OH], with [OH] = 10−12 M and kaqOH = 1.4 × 109 M−1 s−1 (33)] to be only 12 min, substantially faster than its acid-catalyzed hydrolysis. Accordingly, 2-methylglyceric acid, recently identified as a reaction product of IEPOX with OH in the aqueous phase (33, 34), was identified in our study (fig. S5) on the basis of tandem mass spectrometry analysis. Similar to IEPOX, dissolved isoprene oxidation products rapidly react with OH to yield products with much higher oxidation state, forming aqSOA.

The increase in the oxidation state of isoprene gas-phase oxidation products through further OH oxidation in the aqueous phase is expected to translate into a significant decrease in their volatility. Figure 3A presents the effective saturation vapor concentrations (C*, in μg m−3) and the oxidation states of the molecules in the aqSOA and those that dissolved from the gas phase into the aqueous microlayer. The majority of isoprene oxidation products in the gas phase (91.1 ± 10.3%) are highly volatile (C* > 300 μg m−3), and only a minor fraction (1.7 ± 0.4% of the total carbon budget) has sufficiently low volatility (C* < 0.3 μg m−3) to form SOA under atmospherically relevant conditions. Calculated SOA yields based on the volatility distribution of the isoprene oxidation products in the gas phase (Fig. 3B) are generally consistent with the low SOA yields reported in literature (28, 3542). For comparison, we estimate the multiphasic SOA yield as the ratio between the aqSOA mass measured by the AMS and the integrated mass of the gaseous isoprene oxidation products that diffused to the WFR wetted wall. Multiphasic SOA yield values range from 0.05 to 0.28 at a load of 1 to 10 μg m−3, at least twice as high compared to reported gas-phase SOA yields and yields calculated from the volatility distribution of isoprene oxidation products in the gas phase. To the best of our knowledge, this is the first experimental demonstration of enhanced SOA production through OH oxidation in simulated cloud droplets. Figure 3C shows the strong relationship between the measured aqSOA and the total reacted products in the liquid phase estimated using QEMRA, confirming that aqSOA derives from the reaction of isoprene products with OH radicals in the aqueous phase. The fraction of the isoprene oxidation products that dissolve into the aqueous microlayer amounts to ~50% of the total gas phase, under our conditions (shown as the red line in Fig. 3A). Once in the aqueous phase, these compounds rapidly react with OH, with lifetimes of minutes under atmospherically relevant conditions: ~5 min for case A and ~70 min for case B, assuming a well-constrained generic reaction rate constant toward OH of kOH-aq (3.8 ± 1.9) × 108 M−1 s−1 (43). On the basis of the linear fit of the data in Fig. 3C, we estimate that 45 ± 11% of the species that reacted in the aqueous phase form aqSOA upon evaporation of the nebulized aqueous droplets. As aqueous concentrations of OH radicals and total organic species in our conditions are typical of values reported for atmospheric clouds (44, 45), the observed reactivity and aqSOA formation under our experimental conditions are expected to match typical conditions in clouds.

Fig. 3 Multiphase volatility distribution and SOA formation by in-cloud processing.

(A) Volatility distribution of the oxidation products in the gas phase in the absence of the water microfilm (dark line; measured by the PTR-TOF-MS and the acetate-CIMS), of the aqSOA phase (blue line; measured by the EESI-TOF), and representative dissolved phase (red line; calculated from the QEMRA model after 10 hours of experiment). The oxidation state (OSc) is presented on the right axis. (B) SOA formation yield from cloud processing of isoprene oxidation products (filled markers), compared to our own SOA from gas phase oxidation (dashed line) and literature data from isoprene photooxidation laboratory experiments at low-NOx conditions (open markers). The dashed line represents the yield determined from the measured volatility distribution in the gas phase [see (A)], where the shaded area indicates the response of the yield to a shift of the volatility bins by one order of magnitude. (C) Measured aqSOA molar concentration as a function of modeled aqueous oxidation products in the WFR microlayer. A linear fit parameterization (y = 0.45 × x), using a least-square method, is represented by the dashed line. The error (2σ) of the fit is represented by the shaded area. The OH concentrations at steady-state conditions are as follows: case A (light), 2 × 109 molecules cm−3 in the gas phase and 4 × 10−13 to 3 × 10−11 M in the aqueous phase; case B (dark), 3 × 108 molecules cm−3 in the gas phase and 3 × 10−13 to 5 × 10−12 M in the aqueous phase. The aqueous OH concentration compares well to typical ambient conditions in cloud droplets [10−14 to 10−12 M (6)]. Note that the high OH dose in the gas phase is needed to initiate a sufficient concentration of oxidation products but does not result in atmospherically irrelevant conditions (see the Supplementary Materials for details).

Global model simulation and atmospheric implications

To assess the atmospheric implication of aqSOA formation from isoprene on a global scale, we implement the parameterizations of experimental data in the global climate model UKESM1. The uptake fraction of isoprene oxidation products in cloud droplets was parameterized as a function of liquid water content using our measured solubility distribution; it ranges between 50 and 70% under typical atmospheric conditions (see the Supplementary Materials and fig. S6). We also used a kOH-aq of 3.8 × 108 M−1 s−1 and the experimentally derived yield of 0.45 in the model. Figure 4 shows the spatial variation of aqSOA and its contribution to the total SOA budget at the global and continental level. AqSOA exerts a substantial impact on the vertical aerosol distribution (Fig. 4A). The most important SOA precursors in our model, monoterpenes, are relatively short-lived, so they produce SOA close to the ground, while the contribution of aqSOA is more important at higher altitudes where clouds are ubiquitous (4 to 6 km). The aqueous chemistry process accounts for 6.9 Tg of SOA annually or ~20% of the total biogenic SOA budget (Fig. 4B). Including aqSOA in the model changes the direct radiative effect of aerosols at the top of the atmosphere by up to around −0.5 W m−2 (Fig. 4C) in Central Africa and the Amazon over a 1-year simulation. In our model, SOA is produced in the gas phase mainly close to sources (see the Supplementary Materials). Underestimating gas-phase production in remote regions and overestimating it near sources may affect the local radiative effects, and our calculations should be refined in the future using a model with a more sophisticated SOA production mechanism. However, it is unlikely that these effects significantly influence global averages. Hence, the substantial effects that we simulate demonstrate the importance of representing isoprene cloud chemistry to help improve the ability of models to capture the magnitude and variability of aerosol in the free troposphere to quantify aerosol climate effects. The large source of aerosol mass from aqSOA is located mostly in the Southern Hemisphere (80% of total aqSOA), whereas most previous observations of SOA were made in North America, Asia, and Europe. Therefore, our study calls for more dedicated field measurements in the Southern Hemisphere to elucidate the role of aqSOA formation in past, present, and future climate systems.

Fig. 4 aqSOA mass concentration due to isoprene cloud chemistry.

(A) Simulated longitudinal mean vertical profile of the SOA formation rate from aqueous phase isoprene cloud chemistry. (B) Relative change in global SOA mass formation rate after implementing isoprene cloud chemistry, where the pie charts represent the total mass of SOA formed (in Tg year−1) via three different pathways: terpene (tgasSOA) or isoprene (igasSOA) gas-phase SOA formation and isoprene cloud processing (aqSOA) over the entire globe and over each continental area that is marked on the map with a box. Where the total SOA formation rate is below 1 Tg year−1, the change is masked (white on the map). (C) Change in the direct radiative effect of aerosols at the top of the atmosphere when aqSOA is implemented in the model.

Overall, our results represent the first step toward a better representation of in-cloud aqSOA formation under relevant conditions to be integrated in models. Obviously, further studies beyond the limited range of conditions presented here (e.g., room temperature and low NOx regime) are needed. Hence, our novel method opens up new avenues of in-cloud chemistry investigation under various environmental conditions (e.g., high NOx, low temperatures representative of tropospheric clouds, and low pH) and for other volatile organic compound precursors.


Wetted-wall flow reactor

Aqueous phase processing experiments of isoprene OH-oxidation products were performed in a rotating WFR. Although similar devices have been described in the literature (17, 20, 46), only occasional studies on the uptake of individual compounds in a solution were performed, while, to our knowledge, such wetted-wall reactors were never used so far to investigate the uptake of a complex mixture of gas phase oxidation products including all reactants, presumably because of technological limitations. The WFR consists of a quartz glass cylinder (length, 125 cm; internal diameter, 6 cm) into which between 8 and 58 ml of water were injected and rotated (15 rotations/min) to maintain a 35- to 250-μm water microlayer on the wall. While a stirring bar (length, 123 cm; diameter, 3 mm) at the bottom of the cylinder helps mix the solution, it mainly facilitates the wetting of the wall and maintains the microfilm. Note that the inner surface was sandblasted to increase the wettability of the glass. Around the WFR, nine ultraviolet B (UVB) lamps (Philips lamp TL 40W/12 RS) were used in certain experiments to start the photochemistry. When all lamps were switched on, the NO2 and O3 photolysis rates were JNO2 = 3.64 × 10−3 s−1 and JO1D = 5.30 × 10−3 s−1. A ventilation-cooling device on top of the UVB lamps’ frame allowed us to control the temperature at ±1.5°C during the experiments. In addition, the laboratory room temperature was regulated to 22 ± 2°C with an air conditioner.

WFR conditioning and experimental protocol

Between each experiment, the WFR was cleaned three times with ultrapure water (18.2 megohm·cm; Millipore) by rotating it with 500 ml of water for 30 min. The WFR walls were then heated with a heat gun to 200°C, while a pure dry air stream of 20 liter min−1 was maintained for 1 hour.

All experiments were performed with a stream of synthetic clean air (10 liter min−1) by mixing nitrogen (8 liter min−1) produced from the evaporation of pressurized liquid nitrogen (purity, >99.999%; Messer) and oxygen (2 liter min−1; purity, >99.995%; Messer) from gas cylinders. This dry air stream was passed through a porous PTFE (polytetrafluoroethylene) GORE-TEX tube (Gore) immersed in ultrapure water to keep a constant RH of 100%.

OH radicals were produced by irradiating the humidified air stream with a Xe-excimer laser (7.2 eV, 172 nm), where photolysis of H2O and O2 leads to the formation of not only OH radicals but also O3 and HO2 molecules (47). Isoprene was then injected from a gas cylinder [100 part per million by volume (ppmv) in N2; purity, >99%; Carbagas AG] and combined with the humidified air stream containing the OH radicals. In this way, OH oxidation was initiated 34 ms before entering the WFR. In certain experiments, UVB lights were also used as a source of additional OH radicals within the flow tube. The residence time in the WFR was 21 s. A description of the experimental setup is shown in fig. S2 (in the Supplementary Materials) for different case studies and is explained in the following. A summary of the experimental conditions is provided in table S1.

Case study A: Aqueous phase processing of water-soluble organic compounds in the presence of light

Aqueous phase uptake and photochemical processing was performed simultaneously, as shown in fig. S2A (in the Supplementary Materials). A flow of 10 ppbv isoprene was reacted with OH radicals generated with the Xe-excimer laser, followed by reactive uptake in the water microfilm in the presence of UVB light. Here, aqueous phase uptake and photochemical processing occurred simultaneously, i.e., in the presence of UVB light.

Case study B: Aqueous phase uptake of water-soluble organic compounds in the absence of light

Isoprene was reacted in the dark with OH radicals generated with the Xe-excimer laser, as shown in fig. S2B (in the Supplementary Materials). This step allowed water-soluble compounds to be taken up by the water microfilm but with a reduced photochemical processing compared to case A.


Temperature was measured with two thermocouples of type K, one positioned at the inlet and the second one at the outlet. RH was monitored with a hydroclip (Rotronic HygroClip) and a dew point mirror (EdgeTech DewPrime II). Mixing ratios of the parent hydrocarbon and its oxidation products were monitored at the exit of the WFR with a PTR-TOF-MS (Ionicon, Series 8000) and an acetate chemical ionization atmospheric pressure interface time-of-flight mass spectrometer (acetate-CIMS; Airmodus/Aerodyne, Tofwerk). In certain experiments, a Vocus PTR-TOF-MS (Tofwerk/Aerodyne) was also used to determine the mixing ratios of gaseous products. Ozone was measured with a commercial monitor (Monitor Labs Inc., 8810).

At the end of the experiment, water samples were collected and stored at 4°C until chemical analysis, which was performed within 24 hours. Offline aqueous-phase chemical composition analysis was explored with three different instruments. Liquid samples were atomized in N2 using an Apex Q nebulizer (Elemental Scientific Inc., Omaha, NE 68131, USA) operating at 60°C. The resulting aerosol was dried by passing it through a Nafion drier (resulting in an RH of <30%; Perma Pure, Toms River, NJ 08755, USA) and subsequently analyzed by a high-resolution time-of-flight AMS (Aerodyne) and an EESI-TOF-MS (27). In addition, collected samples were analyzed with liquid chromatography with a column (150 mm by 2.1 mm; particle size, 2.6 μm; Accucore RP-MS; Thermo Fisher Scientific) and a precolumn (Accucore RP-MS Defender Guards included; 10 mm by 2.1 mm; particle size, 2.6 μm), coupled with an LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). Analyses were performed at isocratic conditions, with two different mobile phases consisting of the following: (i) 0.05% ethylenediamine (v/v), 39.98% water (v/v), and 59.97% methanol (v/v) and (ii) 0.05% formic acid (v/v), 19.99% water (v/v), and 79.96% methanol (v/v). Spectra were detected in positive and negative mode. Collision-induced dissociation multistage mass spectrometry (MSn; n = 2) was performed on the most intense precursor ions. Details on operation and calibration can be found in the Supplementary Materials.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: This research has received funding from the Swiss National Science Foundation (grant no. 200020_172602), the European Union’s Horizon 2020 Research and Innovation programme through the EUROCHAMP-2020 Infrastructure Activity under grant agreement no. 730997, the Swiss Innovation Agency Innosuisse, the Swiss Competence Center for Energy Research SCCER BIOSWEET, and the Energy System Integration (ESI) platform at the Paul Scherrer Institute. Author contributions: J.D. and U.B. conceived the original idea of the project. U.B. provided funds, instruments, and supercomputing service. H.L., G.W., and J.D. designed, constructed, and prepared the WFR facility. H.L. and I.E.H. designed the research. H.L., D.S., S.B., J.K., and F.M. performed data collection. H.L. performed data analysis and interpretation. A.R. and H.G. performed the global simulation. H.L. drafted the manuscript. All coauthors commented on the results and participated in the writing and critical revision of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The experimental data are available at Zenodo (DOI: 10.5281/zenodo.4498915). Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article