Cumene hydroperoxide (80 %), tert-butyl hydroperoxide (35 %), 2-butanone peroxide (35 %), peracetic acid (32 %), di-tert-butyl peroxide (98 %), benzoyl peroxide (≥98 %), di(dodecanoyl) peroxide (97 %), 2-nonenal (97 %), meso-erythritol (≥99 %), cis-pinonic acid (98 %), formaldehyde (37 %), acetaldehyde (>99 %), hexanal (98 %), methyl glyoxal (40 %), glyoxal (40 %), hydrogen peroxide (H2O2, 30 %), α-pinene (≥99 %) and ammonium acetate (≥99.99 %) are all purchased from Sigma-Aldrich (Canada). Isoprene-4-hydrox-3-hydroperoxide (ISOPOOH) is synthesized according to the literature (Rivera-Rios et al., 2014). Methanol (MeOH, LC-MS grade) is purchased from VWR, Canada. All the chemicals and reagents are used as received.

Stock solutions for ROOH standards (i.e. cumene hydroperoxide, ISOPOOH, tert-butyl hydroperoxide, 2-butanone peroxide and peracetic acid), other peroxides (di-tert-butyl peroxide, benzoyl peroxide and di(dodecanoyl) peroxide), 2-nonenal, meso-erythritol and cis-pinonic acid are prepared by dissolution of the substances in MeOH (2.0–10 mM). One hundred microliters of stock solutions are further diluted in MeOH to a final volume of 1 mL for mass spectrometry analysis. In some cases, ammonium acetate (AA) is added (∼5 mM) to enhance the signal for the ammonium adducts of the molecular ions in the mass spectra.

Reactions of the selected aldehydes (i.e. formaldehyde, acetaldehyde, hexanal, glyoxal and methyl glyoxal) with H2O2 are performed by mixing the reactants in MeOH (∼3 mM and ∼13 mM for aldehydes and H2O2, respectively) at room temperature (295±3 K) for ∼10 min. AA is added (∼5 mM) before the samples are analyzed. Sample blanks (i.e. aldehydes + AA and H2O2 + AA) are prepared and analyzed in the same manner as the reaction mixtures.

SOA is generated in a steady-state manner in a 1 m3 Teflon (FEP) chamber from gas-phase reaction of α-pinene with ozone (Aljawhary et al., 2013). α-Pinene is introduced into the chamber by passing a ∼2 mLmin-1 nitrogen through a bubbler containing α-pinene that is chilled to -20 ∘C and mixed with 5 Lmin-1 purified air. Ozone is generated by passing ∼10 Lmin-1 purified air through a mercury lamp. The final concentration of ozone is monitored to be ∼7.4× 1012 moleculescm-3 by a UV photometric ozone analyzer (Thermo Model 49i), and the α-pinene concentration is estimated to be ∼3.7× 1012 moleculescm-3.

SOA generation is confirmed by a Scanning Mobility Particle Sizer (SMPS, Model 3034) and collected on quartz fiber filters (47 mm diameter) for 72 h at a flow rate of ∼15 Lmin-1. The filters are preheated at 500 ∘C for 24 h to remove organic impurities before SOA collection. After SOA collection, the filter is extracted with 10 mL MeOH, and the extract is immediately analyzed by the mass spectrometer.

A unit resolution APCI-MS/MS instrument (Thermo TSQ Endura) is operated in positive ion mode with direct infusion of the samples into the mass spectrometer. The sample is injected at a flow rate of 10 µLmin-1 using a syringe pump (Chemyx, Inc., USA; model: Fusion 101) to the APCI source through polyether ether ketone (PEEK) tubing. The spray voltage is set at +2500 V; vaporizer temperature and ion transfer tubing temperature are set at 200 ∘C. Sheath gas, auxiliary gas and sweep gas flows are set (arbitrary units) at 5, 2 and 0, respectively. The mass spectrometer is a triple quadrupole that is calibrated with polytyrosine. Mass spectra are obtained under either full-scan mode or selected ion monitoring (SIM) mode.

In full-scan mode, RF lens voltages are the default from the optimization in the mass calibration. The SIM scan is achieved by isolating and monitoring a range of masses with maximum mass range of 50 Da. The RF lens voltages in SIM mode can be manually varied and are optimized to maximize the intensities of the ammonium adducts of the molecular ions.

Ion fragmentation in the MS/MS is accomplished by application of electrical potentials (2–10 V) and collision gas (Argon, 0.5 mTorr) in the collision-induced dissociation (CID) cell. Mass spectra from product scans are obtained by transmitting ions of a specific m/z through the first quadrupole (precursor ions), fragmenting in the CID cell, and monitoring the resulting fragment ions (product ions) by the third quadrupole.

In the NL scan mode, the first and third quadrupoles are scanned at the same rate over mass ranges of the same width; i.e. the third quadrupole transmits ions at a fixed mass-to-charge ratio lower than the first quadrupole. In the NL scan, the CID voltage and argon pressure are set at 2–10 V and 0.5 mTorr, respectively. The limit of detection (LOD) of the analytical method is established by direct injection of the ROOH standard solution in MeOH with 2–10 mM ammonium acetate into the APCI source with the tandem mass spectrometer being operated under NL scan mode. Three ROOH standards – namely 2-butanone peroxide, tert-butyl hydroperoxide and cumene hydroperoxide – are analyzed. Since a neutral loss of 51 Da from the ammoniated molecular ion ([M+NH4]+) of the ROOH is characteristic for ROOH molecules, as will be seen in the next section, the calibration is performed by operating with a loss of 51 Da in NL scan. The ROOH concentrations are selected when the [M+NH4]+ ions are well above the noise and clearly seen in the average mass spectra. The LOD is reported as 3 times the S/N where noise is estimated at mass-to-charge ratios different from the mass-to-charge ratio for [M+NH4]+.

Figure 1 presents examples of the direct infusion (+)-APCI mass spectra for (a) cumene hydroperoxide (cumene HP) and (b) 4,3-ISOPOOH, with and without addition of AA. The chemical structures and molecular weights (MWs) of the ROOH analyzed are given in Fig. S1 in the Supplement. Note that we do not attempt to interpret the mass spectra of ROOH and other standards obtained under full-scan and SIM modes due to the presence of stabilizers and other impurities in the standard samples that make the mass spectra complex. Instead, we focus on the protonated and ammoniated molecular ions of the ROOH molecules.

It can be seen from Fig. 1 that, although cumene HP demonstrates the proton adduct of the molecular ion at m/z=153 ([M+H]+ (top panel of Fig. 1a), addition of AA results in the ammonium adduct at m/z=170 ([M+NH4]+) (bottom panel of Fig. 1a). For ISOPOOH, the molecular ion is not seen in the full-scan mass spectrum (top panel of Fig. 1b), and the dehydrated molecular ion at m/z=101 ([M+H-H2O]+) is clearly observed instead (top panel of Fig. 1b). This is consistent with previous studies claiming that hydroperoxy group (-OOH) is not a favorable protonation or deprotonation site with ESI or APCI (Reinnig et al., 2008; Rondeau et al., 2003; Nilsson et al., 2008). The addition of AA again leads to significant production of the ammonium adduct of the molecular ion at m/z 136 (bottom panel of Fig. 1b). Similar effects of AA on the APCI mass spectra are observed for other ROOH species.

In comparison, a number of other common molecules in atmospheric samples – e.g. ROOR, carbonyls, alcohols and carboxylic acids – were also analyzed. The chemical structures of the other oxygenated organics are given in Fig. S2. Similar to the ROOH samples, the ammonium adducts of the molecular ions ([M+NH4]+) are obtained for all the substances. The only difference is that in some cases, e.g. benzoyl peroxide, the ammonium adduct of the molecular ions can be clearly seen without addition of AA, via trace levels of ammonia present in the water or air.

Figure 2 gives the CID fragment patterns (i.e. product spectra) for ammonium adducts of the molecular ions ([M+NH4]+) of (a) cumene HP and (b) ISOPOOH. Two types of fragmentation, i.e. loss of 35 and 51 Da from the respective [M+NH4]+ ions, are observed in both cumene HP and ISOPOOH. The loss of 35 Da, corresponding to [-H2O-NH3] (Fig. S3), is also observed in meso-erythritol and pinonic acid (Fig. S4). Neutral loss of 51 Da, corresponding to [-H2O2-NH3] (Fig. S3), is only observed from fragmentation of the [M+NH4]+ ions of ROOH (Figs. 2 and S4). It is known that the loss of the H2O molecule from protonated molecular ions ([M+H]+) is typical for epoxides, alcohols and carboxylic acids (Holcapek et al., 2010). But the loss of 35 Da from [M+NH4]+ is not characteristic for ROOH. Instead, the loss of 51 Da from [M+NH4]+ is only observed in ROOH standards, including peracetic acid (Fig. S5). Hence we propose that the neutral loss of 51 Da from [M+NH4]+ is characteristic for ROOH molecules. This is consistent with previous work that suggested that neutral loss of H2O2 from [M+H]+ is also characteristic for ROOH species (Reinnig et al., 2009).

Using 1H NMR spectroscopy, it has been shown that ROOH species form from the reactions of aldehydes with H2O2 (Zhao et al., 2013). As shown in Fig. 3, the reaction proceeds via reversible nucleophilic addition of H2O2 to the carbonyl group in aldehydes, leading to α-hydroxyhydroperoxides (HHP). The addition of H2O2 to methylglyoxal (MGL; MW =72) gives rise to MGL HHP (MW =106) (Fig. 3a), whose ammonium adduct is seen at m/z 124 in Fig. 4a. The mass spectrum of the reaction mixture of glyoxal and H2O2 suggests that, rather than a direct addition to glyoxal, H2O2 is instead added to the glyoxal geminal diol formed by the hydrolysis of glyoxal (Fig. 3b), producing [M+NH4]+ at m/z 128 (Fig. S6a). The [M+NH4]+ ROOH peaks are also observed in the reactions of other aldehydes with H2O2 (data not shown).

The fragmentation spectra of the [M+NH4]+ of the ROOH from methylglyoxal and glyoxal are given in Figs. 4b and S6b, respectively. It is clear that the neutral loss of 51 Da is again observed in the fragments of the ROOH molecules. The ROOH products from all the other aldehyde reactions with H2O2 also show loss of 51 Da in MS/MS mode. Overall, we conclude that the neutral loss of 51 Da from [M+NH4]+ fragmentation can be used to identify ROOH molecules.

Figure 5 presents mass spectra of the methanol extract of SOA generated from ozonolysis of α-pinene under dry conditions (relative humidity (RH) <5 %). There are three features of note. First, the well-characterized products from this reaction – such as nopinone aldehyde (MW =154), terpenylic acid (MW =172) and cis-pinic acid (MW =186) (Jenkin et al., 2000; Larsen et al., 2001; Claeys et al., 2009) – are present as protonated molecular ions ([M+H]+) in the full-scan mode (Fig. 5a). Second, the full-scan mass spectrum (Fig. 5a) shows the presence of monomer (m/z=80–250), dimer (m/z=250–450) and trimer (m/z=450–600) species, as has been reported in previous work (Venkatachari and Hopke, 2008). Third, and perhaps most importantly, the 51 Da neutral loss scan (Fig. 5b) indicates that all the ROOH species have masses <300 Da (Fig. 5b).

To confirm the observation of ROOH obtained with the neutral loss scan, CID fragmentation spectra of a few intense peaks at m/z 190, 202, 206, 218 and 220 in Fig. 5b are analyzed. All the fragments of these products show loss of 51 Da. An example of the fragment pattern for m/z 206 is given in Fig. S7.

The ROOH products in SOA are tentatively identified and listed in Table 1. It should be noted that the chemical structures of these products are only obtained from their molecular ions and, therefore, remain speculative.

Several mechanisms for oligomer product formation in SOA arising from VOC oxidation have been proposed: (i) self- and cross-reactions of the peroxy radicals (RO2) (Zhang et al., 2015); (ii) reaction of ozonolysis products in the condensed-phase, such as aldol condensation, esterification, hemiacetal and peroxyhemiacetal formation (Ziemann, 2003; Tolocka et al., 2004; Kristensen et al., 2014; Docherty et al., 2005; Muller et al., 2009; Yasmeen et al., 2010; Hall and Johnston, 2012; Witkowski and Gierczak, 2014; DePalma et al., 2013; Lim and Turpin, 2015); (iii) dimer cluster formation from carboxylic acids (Hoffmann et al., 1998; Tobias and Ziemann, 2000; Claeys et al., 2009; Camredon et al., 2010; DePalma et al., 2013); (iv) reactions of Criegee intermediates (CIs) with VOCs oxidation products (Bonn et al., 2002; Lee and Kamens, 2005; Tolocka et al., 2006; Heaton et al., 2007; Witkowski and Gierczak, 2012; Kristensen et al., 2016; Wang et al., 2016); and (v) reactions of RO2 radicals with Cis (Sadezky et al., 2008; Zhao et al., 2015). Among them, the reactions of CIs with protic substances (water, alcohols, acids and hydroperoxides) can form ROOH. However, the nature of the ROOH products observed in SOA material suggests that these reactions do not take place to a significant extent given that ROOH do not contribute significantly to the dimer and trimer SOA signals (Fig. 5b). The similar mass patterns for the SOA and ROOH obtained under dry (RH <5 %) and humid (RH =50 %) conditions support this conclusion (Fig. S9). Of course, we can not rule out that the oligomeric ROOH may not be sensitive with the analytical method used. Additionally, while the ROOH formation mechanisms are proposed via gas-phase ozonolysis of α-pinene (Fig. S8), the ROOH observed in the SOA could also potentially be arising from the decomposition of peroxyhemiacetals during methanol extraction.

In the neutral loss scan mode, the LOD were measured to be 0.2, 0.3 and 20 mM for 2-butanone peroxide, cumene hydroperoxide and tert-butyl hydroperoxide, respectively. The LODs obtained by this method are a rough estimation and will vary dependent on a number of parameters: ionization voltage, sample injection flow, gas flow, CID gas pressure, CID voltages etc. More importantly, if a specific ROOH is to be analyzed, then multiple reaction monitoring mode would be applied and the LOD of the ROOH would be substantially reduced.