Details of the experimental apparatus and protocol used in the current study have been previously described (Shang et al., 2017). Briefly, the main body of the environmental chamber is a 400 L polyvinyl fluoride (PVF) bag filled with synthetic air (high purity N2 (99.999 %) mixed with high purity O2 (99.999 %) with a ratio of 79:21 by volume; Beijing Huatong Jingke Gas Chemical Co.). The chamber is capable of temperature (∼293 K) and relative humidity (0.8 %–70 %) control using a water bubbler and air conditioners, respectively. The chamber is equipped with two light sources, both with a central wavelength of 365 nm. One is a set of 36 W tube lamps with a main spectrum of 320–400 nm and a small amount of 480–600 nm visible light (Fig. S1a). The other is a set of 12 W LED lamps with a narrow main spectrum of 350–390 nm (Fig. S1b). The light intensities for the tube and LED lamp at 365 nm were 300 and 200 µW cm-2, respectively, as measured in the middle of the chamber. NOx concentrations at the outlet of the chamber were monitored by a chemiluminescence NOx analyzer (ECOTECH, EC9841B). HCHO was generated by thermolysis of paraformaldehyde at 70 ∘C and detected via an acetyl acetone spectrophotometric method using a UV–Vis spectrophotometer (PERSEE, T6) or a fluorescence spectrophotometer (THERMO, Lumina), depending on different initial HCHO concentrations. The particle size distribution was measured by a scanning nanoparticle spectrometer (HCT, SNPS-20). Electron spin resonance (Nuohai Life Science, MiniScope MS 5000) was used to measure ⚫OH on the surface of the particles. A 5,5-dimethl-1-pyrroline-N-oxide (DPMO, Enzo) was used as the capture agent. Fifty µL particle-containing suspension mixed with 50 µL DMPO (concentration of 200 µM) was loaded in a 1 mm capillary. Four 365 nm LED lamps were placed side by side, vertically, at a distance of about 1 cm from the capillary, and the measurement was carried out after 1 min of irradiation. The modulation frequency was 100 kHz; the modulation amplitude was 0.2 mT; the microwave power was 10 mW; and the sweep time was 60 s.

In our experiments, two nitrate salts – potassium nitrate (AR, Beijing Chemical Works Co., Ltd) or ammonium nitrate (AR, Beijing Chemical Works Co., Ltd) – were composited with pure TiO2 (≥ 99.5 %, Degussa AG) powder or TiO2 (1 wt %) / SiO2 mixed powder to prepare NO3-–TiO2 or NO3-–TiO2 (1 wt %) / SiO2 samples; 250 mg TiO2 was simply mixed in the nitrate solutions at the desired mass mixing ratio (with nitrate content of 4 wt %) to obtain a mash. The mash was dried at 90 ∘C and then ground carefully for 30 min. A series of samples with different amounts of nitrate were prepared, and diffuse reflectance Fourier transform infrared spectroscopy (DRIFTS) measurements were made to test their homogeneity. Figure S2 shows the DRIFTS spectra of these KNO3–TiO2 composites, of which the 1760 cm-1 peak is one of the typical vibrating peaks of nitrate (Aghazadeh, 2016; Maeda et al., 2011). The ratio value of the peak area from 1730–1790 cm-1 for 1 wt %, 4 wt %, 32 wt %, 80 wt % composited samples is 1:4.1:29.8:81.6, which is very close to that of theoretical value, proving that the samples were uniformly mixed. SiO2 (AR, Xilong Scientific Co., Ltd.) with no optical activity was also chosen for comparison, and samples of KNO3-SiO2 and KNO3–TiO2 (1 wt %) / SiO2 samples with a potassium nitrate content of 4 wt % were prepared. The blank 250 mg TiO2 sample was solved in pure water, using the same procedure as mentioned above. Four wt % HNO3–TiO2 composite particles were prepared for comparison. Concentrated nitric acid (AR, Beijing Chemical Works Co., Ltd) was diluted to 1 M, and 250 mg TiO2 was added to the nitric acid solution and stirred evenly. A layer of aluminum foil was used to cover the surface of the HNO3–TiO2 homogenate, which was then dried naturally in the room and then ground for use. We also selected Arizona Test Dust (ATD, Powder Technology Inc.), whose chemical composition and weight percentage were shown in Table S1, as a substitute of NO3- / TiO2 to investigate the “photocatalytic renoxification” process of nitrate and the positive effect of HCHO.

For the chamber operation, we completely evacuated the chamber after every experiment; we then cleaned the chamber walls with deionized water and dried them by flushing the chamber with ultra-zero air to remove any particles or gases that had collected on the chamber walls. The experiments carried out in the environmental chamber can be divided into two categories according to whether HCHO was involved or not. (1) No HCHO involvement in the reaction. The PVF bag was inflated by 260 L synthetic air, and then 75 mg particles were instantly sprayed into the chamber by a transient high-pressure airflow. As shown in Fig. S3, the particle number concentration of the KNO3–TiO2 or TiO2 samples decreased rapidly owing to the wall effect, including the possible electrostatic adsorption of the particles by the environmental chamber. The size distributions of the KNO3–TiO2 and TiO2 samples were similar, with both becoming stable after about 60 min. The peak number concentration was averaged between 3991 and 3886 particle cm-3 during the illumination period for the KNO3–TiO2 and TiO2 samples, respectively, indicating that the repeatability of the introduction of particles into the chamber is good. This can be attributed to the strict cleaning of the chamber and the same operation of each batch experiment. (2) With the participation of HCHO. The PVF bag was inflated by 125 L synthetic air followed by the introduction of HCHO, and then the chamber was filled up with zero air to about 250 L. In order to know the HCHO adsorption before and after the particles' introduction, we conducted a conditional experiment in the dark. It can be seen from Fig. S4 that it took about 90 min for the concentration of HCHO to reach stability and be sustained. Then, 75 mg TiO2 or NO3- / TiO2 powders were introduced instantly, and the concentration of HCHO decreased upon the introduction. It took about 60 min for HCHO to reach its second adsorption equilibrium and for the concentration of HCHO to remain stable for several hours in the dark. Therefore, for the irradiation experiments, the particles were injected at 90 min after HCHO's introduction, and the lamps were turned on at 60 min after the particle's introduction.

To determine the background value of NOx in the reaction system, four blank experiments were carried out under illumination without nitrate: “synthetic air”, “synthetic air + TiO2”, “synthetic air + HCHO”, and “synthetic air + HCHO + TiO2”. In the blank experiments of “synthetic air” and “synthetic air + TiO2”, the NOx concentration remained stable during 180 min illumination, and the concentration change was no more than 0.5 ppb (Fig. S5a). Therefore, the environmental chamber, the synthetic air, and the surface of TiO2 particles were thought to be relatively clean, and there was no generation and accumulation of NOx under illumination. When HCHO was introduced into the environmental chamber, NOx accumulated ∼2 ppb in 120 min with or without TiO2 particles (Fig. S5b). Compared with the blank experiment results when there was no HCHO, NOx might come from the generation process of HCHO (impurities in paraformaldehyde). However, considering the high concentration level of NOx produced in the NO3-–TiO2 system containing HCHO under the same conditions in this study (see later in Fig. 2), the NOx generated in this blank experiment can be negligible.

We investigated the photocatalytic role of TiO2 on renoxification. The light source was two 365 nm tube lamps containing small amounts of 400–600 nm visible light; this setup was suitable for exciting TiO2 and the photolysis of available nitrate radicals. Raw NOx data measured in the chamber under dark and illuminated conditions for 4 wt % KNO3-SiO2 and 4 wt % KNO3–TiO2 (1 wt %) / SiO2 are shown in Fig. 1. The ratio of 1 wt % TiO2 to SiO2 corresponds to their ratio in sand and dust particles. We observed no NOx in the KNO3-SiO2 sample under darkness or illumination, indicating very weak direct photolysis of nitrate under our 365 nm tube-lamp illumination conditions. However, when the sample containing TiO2 / SiO2 was illuminated, NOx continually accumulated in the chamber. This finding confirms that NOx production arising from photodissociation of NO3- on TiO2 /,SiO2 was caused by the photocatalytic property of TiO2 (i.e., photocatalytic renoxification) and was not due to the direct photolysis of NO3- (photolytic renoxification).

TiO2 can be excited by UV illumination to generate electron-hole pairs, and the h+ can react with adsorbed NO3- to produce NO3⚫ (Ndour et al., 2009). Thus, in the present study, NO3⚫ mainly absorbed visible light emitted from the tube lamps, which was subsequently photolyzed to NOxthrough Eqs. (3) and (4) (Wayne et al., 1991), which explains why NOx was observed in this study. Thus, we demonstrated that TiO2 can be excited at illumination wavelengths of ∼365 nm, even when then content was very low, and that NOx accumulated due to the production and further photolysis of NO3⚫. However, the production rate of NOx was very slow, reaching only 1.3 ppb during 90 min of illumination. This result may have been caused by the blocking effect of K+ on NO3-. K+ forms ion pairs with NO3-, and electrostatic repulsion between K+ and h+ prevents NO3- from combining with h+ to generate NO3⚫ to a certain extent, thereby weakening the positive effect of TiO2 on the renoxification of KNO3 (Rosseler et al., 2013). 1TiO2+hυ(λ<390nm)→e-+h+.2NO3-+h+→NO3⚫.3NO3⚫+hυ(λ<640nm)→NO2+O⚫.4NO3⚫+hυ(585nm<λ<640nm)→NO+O2.

LED lamps with a wavelength range of 350–390 nm and no visible light were used to irradiate 4 wt % KNO3–TiO2 without generating NOx (NO2 and NO concentrations fluctuate within the error range of the instrument) (Fig. S5). TiO2 can be excited under this range of irradiation, producing NO3 radicals as discussed above. The lack of NOx generation indicates that neither nitrate photolysis nor NO3⚫ photolysis occurred under 365 nm LED lamp illumination conditions. In addition, it has been shown that NO3⚫ photolysis only occurs in visible light (Aldener et al., 2006). Therefore, the LED lamp setup was used in subsequent experiments to exclude the direct photolysis of both KNO3 and NO3⚫ but to allow the excitation of TiO2. This approach allowed us to investigate the process of photocatalytic renoxification caused by HCHO in the presence of photogenerated NO3⚫.

Atmospheric trace gases can undergo photocatalytic reactions on the surface of TiO2 (Chen et al., 2012). As the illumination time increased, the concentration of HCHO showed a linear downward trend, which was found to fit zero-order reaction kinetics (Fig. S7). The zero-order reaction rate constants of HCHO on TiO2 and 4 wt % KNO3–TiO2 particles were 9.1×10-3 and 1.4×10-2 ppm min-1, respectively, which were much higher than that for gaseous HCHO photolysis (Shang et al., 2017). We suggested that the produced NO3⚫ contributed to the enhanced uptake of HCHO. In the following study, the effect of HCHO on the photocatalytic renoxification of NO3-–TiO2 was explored.

The variation in NOx concentrations within the chamber containing nitrate–TiO2 particles with or without HCHO is shown in Fig. 2. For 4 wt % KNO3–TiO2 particles, the NOx concentration began to increase upon irradiation in the presence of HCHO, reaching ∼ 3861 mmol normalized ppb (equivalent to 110 ppb) within 120 min. This result indicates that HCHO greatly promoted photocatalytic renoxification of KNO3 on the surfaces of TiO2 particles. This reaction process can be divided into two stages: a rapid increase within the first 60 min and a slower increase within the following 60 min, each consistent with zero-orde reaction kinetics. The slow stage is due to the photodegradation of HCHO on KNO3–TiO2 aerosols, which led to a decrease in its concentration, gradually weakening the positive effect. NOx is the sum of NO2 and NO, both of which showed a two-stage concentration increase (Fig. S8). The NO2 generation rate was nearly 6 times that of NO as compared to using the zero-order rate constant within 60 min (1.18 ppb min-1 NO2, R2= 0.96; 0.19 ppb min-1 NO, R2= 0.91). This burst-like generation of NOx can be ascribed to the reaction between generated NO3⚫ and HCHO via hydrogen abstraction to form adsorbed nitric acid (HNO3(ads)) on TiO2 particles. We measured the pH of water extracts in NO3-–TiO2 systems with and without HCHO. It was found that the pH decreased by 1.7 % for KNO3–TiO2, suggesting the formation of acidic species such as HNO3(ads) in this study. Based on the analysis of the absorption cross section of HNO3 adsorbed on the fused silica surface, the HNO3(ads) absorption spectrum has been reported to be red-shifted compared to HNO3(g), extending from 350 to 365 nm, with a simultaneous cross-sectional increase (Du and Zhu, 2011). Therefore, HNO3(ads) was subjected to photolysis to produce NO2 and HONO (Eqs. 6–8) under the LED lamp used in this study. A previous study of HNO3 photolysis on the surface of Pyrex glass showed that the ratio of the formation rates of photolysis products (JNOx/J(NOx+HONO)) was > 97 % at RH = 0 % (Zhou et al., 2003), suggesting that NOx is the main gaseous product under dry conditions. Thus, the effect of HONO on product distribution and NOx concentration was negligible in this study. Together, these results suggest that NO3⚫ and HCHO generate HNO3(ads) on particle surfaces through hydrogen abstraction, which contributes to the substantial release of NOx via photolysis. This photocatalytic renoxification via the NO3-–NO3⚫–HNO3–NOx pathway is important considering the high abundance of hydrogen donor organics in the atmosphere. 5NO3⚫+HCHO→CHO⚫+HNO3(ads)6HNO3(ads)+hυ→[HNO3]*(ads)7[HNO3]*(ads)→HNO2(ads)+O(3P)(ads)8[HNO3]*(ads)→NO2(ads)+⚫OH(ads)

To demonstrate the proposed HCHO mechanism and the photolysis contribution of HNO3 to NOx, we prepared an HNO3–TiO2 sample by directly dissolving TiO2 into dilute nitric acid. The formation of NOx on HNO3–TiO2 without HCHO under illumination was obvious and at a rate comparable with that on KNO3–TiO2 with HCHO (Fig. 2). The renoxification of HNO3–TiO2 particles was further enhanced following the introduction of HCHO. This is because HNO3 dissociates on particle surfaces to generate NO3-, such that HNO3 exists on TiO2 as both HNO3(ads) and NO3-(ads). Similarly, NO3-(ads) completed the NO3-–NO3⚫–HNO3–NOx pathway, as described above, through the reaction process shown in Eqs. (2) to (8). The rates of NOx production from HNO3–TiO2 particles with and without HCHO were similar for the first 60 min (Fig. 2), mainly due to the direct photolysis of partial HNO3(ads). However, after 60 min, NOx was generated rapidly in the presence of HCHO, perhaps due to the dominant photocatalytic renoxification of NO3-(ads). These findings indicate that HCHO converts NO3- on particle surfaces into HNO3(ads) by reacting with NO3⚫, and then HNO3(ads) photolyzes at a faster rate to generate NOx, allowing HCHO to enhance the formation of NOx. Overall, the photocatalytic renoxification of NO3-–TiO2 particles affects atmospheric oxidation and the nitrogen cycle, and the presence of HCHO further enhances this impact.

Photocatalytic renoxification reactions occur on the surfaces of mineral dust due to the presence of semiconductor oxides with photocatalytic activity such as TiO2 (Ndour et al., 2009). In order to confirm this, we synthesized nitrate with inert SiO2 as a comparison. It can be seen from Fig. S9 that no NO2 formation was observed regardless of whether or not HCHO was present, indicating that photocatalytically active particle TiO2 is critical to the photocatalytic renoxification process. Furthermore, a kind of commercial mineral dust ATD was selected to study the effects of HCHO on this process. We detected ⚫OH in irradiated pure TiO2 and ATD samples using the electron spin resonance (ESR) technique and found that, for ATD samples, the peak intensity of ⚫OH generation was 40 % that of TiO2 samples (Fig. S10). ⚫OH originates in the reaction of h+ with surface-adsorbed water (Ahmed et al., 2014). ATD contains semiconductor oxides such as TiO2 and Fe2O3 and is thought to exhibit photocatalytic properties affecting the renoxification of nitrate. The NO3- content of ATD is 4×1017 molecules m-2, which is ∼ 0.25 wt % of the total mass (Huang et al., 2015; Park et al., 2017). The NOx concentration changes observed in the environmental chamber demonstrated that HCHO promoted the renoxification of ATD particles (Fig. S11). This result suggests that mineral dust containing photocatalytic semiconductor oxides such as TiO2, Fe2O3, and ZnO can greatly promote the conversion of granular nitrate to NOx in the presence of HCHO.