A new method for atmospheric detection of the CH 3 O 2 radical

A new method for measurement of the methyl peroxy (CH3O2) radical has been developed using the conversion of CH3O2 into CH3O by excess NO with subsequent detection of CH3O by fluorescence assay by gas expansion (FAGE) with laser excitation at ca. 298 nm. The method can also directly detect CH3O, when no nitric oxide is added. Laboratory calibrations were performed to characterise the FAGE instrument sensitivity using the conventional radical source employed in OH calibration with conversion of a known concentration of OH into CH3O2 via reaction with CH4 in the presence of O2. Detection limits of 3.8× 108 and 3.0× 108 molecule cm−3 were determined for CH3O2 and CH3O respectively for a signal-to-noise ratio of 2 and 5 min averaging time. Averaging over 1 h reduces the detection limit for CH3O2 to 1.1× 108 molecule cm−3, which is comparable to atmospheric concentrations. The kinetics of the second-order decay of CH3O2 via its self-reaction were observed in HIRAC (Highly Instrumented Reactor for Atmospheric Chemistry) at 295 K and 1 bar and used as an alternative method of calibration to obtain a calibration constant with overlapping error limits at the 1σ level with the result of the conventional method of calibration. The overall uncertainties of the two methods of calibrations are similar – 15 % for the kinetic method and 17 % for the conventional method – and are discussed in detail. The capability to quantitatively measure CH3O in chamber experiments is demonstrated via observation in HIRAC of CH3O formed as a product of the CH3O2 self-reaction.


Introduction 20
Methyl peroxy (CH3O2) radicals are critical intermediates in the atmospheric oxidation (Orlando and Tyndall, 2012) and combustion of hydrocarbons (Zador et al., 2011). In the remote atmosphere CH3O2 is mainly formed by the reaction of methane with the OH radical via abstraction of an H atom (R1), followed by the reaction of the produced CH3 radical with O2 (R2).
OH + CH4  CH3 + H2O (R1) 25 Methyl radicals can also be formed from more complex species, e.g. the reaction of acetyl peroxy radicals with HO2 in low NOx environments or the reaction of acetyl peroxy radicals with NO in anthropogenically influenced environments. CH3O2 is predicted to be the most abundant peroxy radical in the atmosphere, yet there are no specific measurements of its concentration. 30 Daytime concentrations estimated using a box model utilizing the MCM (Master Chemical Mechanism) version 3.3.1 (Saunders et al., 2003;Jenkin et al., 2015) are ~ 6 × 10 8 molecule cm -3 in the tropical Atlantic ocean in summer (Whalley et al., 2010), ~ 2 × 10 8 molecule cm -3 in a tropical rainforest (Whalley et al., 2011), and lower in polluted environments, for example ~ 5 × 10 7 molecule cm -3 in London in summertime (Whalley et al., to be submitted).
The reaction of CH3O2 with NO (R3) usually dominates the chemistry of CH3O2, particularly in environments influenced 35 by anthropogenic NOx emissions, resulting in NO2 production and hence ozone production: The subsequent reaction of CH3O with O2 (R4) produces HO2, which in turn oxidises another NO to NO2 (R5) with further production of O3 and propagation of the HOx radical chain: CH3O + O2  CH2O + HO2 (R4) HO2 + NO  OH + NO2 (R5) 5 However, under low NOx levels (e.g. remote forested environments and the marine boundary layer) the self-reaction of CH3O2 (R6) and the reactions of CH3O2 with HO2 and other organic peroxy (RO2) species are important radical removal/termination reactions. The CH3O2 self-reaction occurs through two channels, (R6.a) and (R6.b) (Tyndall et al., 1998 Despite the importance of the reaction (R6), there are uncertainties of about a factor of two in the value of its rate coefficient at room temperature, k6, which ranges from (2.7-5.2) × 10 -13 cm 3 molecule -1 s -1 (Atkinson et al., 2006); the preferred IUPAC 15 value is k6 = 3.5 × 10 -13 cm 3 molecule -1 s -1 (Atkinson et al., 2006). The previous kinetic studies used time-resolved UVabsorption spectroscopy to detect CH3O2 radical, typically at 250 nm, Watson, 1980, 1981;McAdam et al., 1987;Kurylo and Wallington, 1987;Jenkin et al., 1988;Simon et al., 1990;Lightfoot et al., 1990). UV-absorption spectroscopy is a relatively insensitive technique and hence the detection limits of CH3O2 were quite high, for example approximately 4 × 10 12 molecule cm -3 Watson, 1980, 1981). In addition, due to the broad, featureless spectra of RO2 species, which 20 often overlap, UV-absorption is a relatively unselective technique for the study of the kinetics of individual RO2. Therefore, there is a clear need for the determination of k6 using a more selective method, which will be addressed in subsequent studies.
At present, CH3O2 is not specifically measured in the atmosphere by any direct or indirect method. Time-resolved continuous-wave cavity ringdown spectroscopy (CRDS), using the 12 transition of the A  X band at ~ 1.3 m has been used to detect CH3O2 directly in a photoreactor (Farago et al., 2013;Bossolasco et al., 2014). However, the detection limit is not 25 sufficiently sensitive to enable tropospheric detection. Typically, the sum of HO2 and all organic RO2 has been measured in the atmosphere, making no distinction between HO2 and different RO2 species, although more recently the sum of RO2 has been quantified separately to HO2. One of the methods uses Chemical Ionisation Mass Spectrometry to determine the sum [HO 2 ] + ∑ [RO 2,] or separately [HO2], depending on the control of the flows of the NO and SO2 reagents (Hanke et al., 2002;Edwards et al., 2003). The sum [HO 2 ] + ∑ [RO 2,] has also been determined for many years by the Peroxy Radical 30 Chemical Amplifier (PERCA) method, which uses NO and CO to generate NO2 amplified by a chain reaction, and subsequently measured by a variety of methods, for example luminol fluorescence, laser-induced fluorescence (LIF) or cavity absorption methods (Cantrell and Stedman, 1982;Cantrell et al., 1984;Miyazaki et al., 2010;Hernandez et al., 2001;Green et al., 2006;Chen et al., 2016). A modification of PERCA, using a denuder to remove HO2 has been used to estimate the sum of RO2 (Miyazaki et al., 2010). ROxLIF is a more recent method, which uses OH LIF detection at low pressure, known as FAGE 35 (fluorescence assay by gas expansion) (Fuchs et al., 2008;Whalley et al., 2013 Whalley 40 et al., 2013). Recently, the interference from certain types of RO2 radicals in the FAGE detection of HO2 was deliberately exploited to enable a partial RO2 speciation (Whalley et al., 2013). The method was used in the Clean Air for London campaign (ClearfLo) to distinguish between the sum of alkene, aromatic and long-chain alkane-derived RO2 radicals and the sum of short-chain alkane-derived RO2 radicals (Whalley et al., 2013).
As methoxy (CH3O) radicals can be generated by techniques such as pulsed laser photolysis and microwave discharge and detected with high sensitivity by LIF (Shannon et al., 2013;Chai et al., 2014;Albaladejo et al., 2002;Biggs et al., 1993;Biggs et al., 1997), the method has been used in kinetic studies of a range of CH3O reactions. These studies used the electronic 5 excitation of the methoxy radical from the ground state to the first electronically excited state (A 2 A1  X 2 E). The A  X excitation spectrum covers the range ~ 275-317 nm and leads to fluorescence from several vibronic bands in the near UV, and has been reported in a series of experimental and theoretical studies (Inoue et al., 1980;Kappert and Temps, 1989;Powers et al., 1997;Nagesh et al., 2014). This paper reports the development of a new method for the selective and sensitive detection of CH3O2 radicals using 10 FAGE by titrating CH3O2 to CH3O by reaction with added NO (R3) and then detecting the resultant CH3O by off-resonant LIF with laser excitation at ca. 298 nm. The method is similar to the standard method used for the detection of HO2 radicals by FAGE through conversion of HO2 to OH by reaction with added NO followed by OH on-resonance LIF at about 308 nm (Heard and Pilling, 2003). As LIF is not an absolute detection method, FAGE instruments require calibration, with the 184.9 nm photolysis of water vapour in air using a mercury (Hg) Pen-Ray lamp being a common method employed for generating 15 known concentrations of OH and HO2 (Heard and Pilling, 2003): An alternative CH3O2 calibration is also presented, consisting of the analysis of the kinetics of the CH3O2 decay by selfreaction monitored by FAGE and compared with the water photolysis method. The studies are performed within HIRAC (Highly Instrumented Reactor for Atmospheric Chemistry) which is a 2.25 m 3 , custom-built, stainless steel chamber simulating the ambient conditions (Glowacki et al., 2007). HIRAC has been used in alternative calibrations of FAGE for OH and HO2 30 using the temporal evolution of appropriate species, in validation and development of new atmospheric measurement techniques as well as in kinetic and mechanistic studies of atmospheric relevant reactions (Malkin et al., 2010;Winiberg et al., 2015;Winiberg et al., 2016).
Direct LIF detection of CH3O radicals, which is also a key intermediate in the oxidation of methane and other VOCs in the troposphere and formed by reactions such as (R3) and (R6.b), is also reported here. However, in the atmosphere CH3O is 35 exclusively consumed by reaction with O2 (R4) generating formaldehyde and recycling HO2, resulting in a very short lifetime and consequently very low concentration (~10 2 -10 3 molecule cm -3 ). For this reason no measurements in the atmosphere have previously been attempted. The photolysis of CH3OH at 184.9 nm is used to estimate the FAGE sensitivity for CH3O. The dominant photolysis channel of methanol between 165 and 200 nm generates CH3O radicals (Wen et al., 1994;Kassab et al., 1983;Marston et al., 1993) A photodissociation quantum yield of CH3O of 0.86 ± 0.10 has been found at 193.3 nm (Satyapal et al., 1989) in qualitative agreement with analysis of the end-products of the methanol photodissociation at 184.9 nm (Porter and Noyes, 1959;Buenker et al., 1984). Here we report the first measurements of CH3O concentrations in an atmospheric simulation chamber. Methoxy radicals are generated by the CH3O2 self-reaction carried out within HIRAC at 295 K and 1000 mbar of N2 containing O2 in 5 trace amounts to reduce the rate of removal of CH3O by reaction with O2. This work enhances the capability of HIRAC to measure short-lived radical species by the addition of both CH3O2 and CH3O detection, and we discuss the potential of the method for detection of CH3O2 in the atmosphere itself.

The FAGE instrument 10
Details on the HIRAC-based FAGE instrument for the detection of OH and HO2 has been presented previously (Winiberg et al., 2015). Figure 1 shows a schematic cross-section of the instrument inlet and the two fluorescence detection cells. The gas was sampled with a flow rate of 3.2 slm through a 1 mm diameter pinhole and passed down a 50 mm diameter flow tube of 280 mm length first into the OH detection axis and, after a further 300 mm, into the CH3O2 detection axis. The pressure in the detection cells was maintained at (2.65  0.05) Torr by using a high capacity rotary-backed roots blower pumping system 15 (Leybold, trivac D40B and RuVac WAU251). CH3O2 radicals were titrated to CH3O by adding high purity NO (BOC, N2.5 Nitric Oxide) with a typical 2.5 sccm flow rate (further details in Section 2.2) ~25 mm before the second detection axis into the centre of the flow. The resultant CH3O radicals were measured by LIF. fluorescence through a (308.8  5.0) nm bandpass filter (transmission > 50 %) and the second cell to detect CH3O2 after titration with added NO to form CH3O using a bandpass filter between 320-430 nm with an average transmission > 80%.
Probe laser light was generated by a Nd:YAG (JDSU Q201-HD) pumped dye laser (SIRAH Credo-Dye-N) using a DCM dye (Sirah) in ethanol and operating at 5 kHz pulse repetition frequency, with a pulse width at half maximum of 25 ns, typical 25 pulse energy of 120 J pulse -1 and a linewidth of 0.08 cm -1 at 595 nm. The frequency doubled light at either ~308 nm (OH detection) or ~298 nm (CH3O detection), was focused into fibre optic cables to be delivered to the two detection cells. OH and CH3O radicals were separately detected by LIF spectroscopy by exciting at 307.99 nm using the Q1(2) rotational line of the A 2  + (' = 0)  X 2 i (" = 0) OH transition in the first detection axis to monitor on-resonant fluorescence (308.8  5.0 nm) and excitation at 297.79 nm in the A 2 A1 ('3 = 3)  X 2 E ("3 = 0) CH3O transition in the second detection axis to monitor red-30 shifted off-resonant LIF (320-430 nm). Here 3 refers to the C-O stretching vibrational mode of CH3O which demonstrates a progression in the LIF spectrum (Inoue et al., 1980;Kappert and Temps, 1989;Powers et al., 1997;Nagesh et al., 2014). The fluorescence in the two cells was collected orthogonal to the gas flow by two microchannel plate photomultiplier tubes (MCP-PMT) (Photek PMT325/Q/BI/G) equipped with a 50 ns gate unit (Photek GM10-50) for gated photon-counting, and the signal was amplified using a pre-amplifier (Photek PA200-10). Further details on the OH detection and calibration in HIRAC have been reported previously (Winiberg et al., 2015).
The laser and photon-counting timing for CH3O detection was controlled by a delay pulse generator (9520 Quantum Composers). The relatively broad bandpass filter used for the collection of the CH3O fluorescence (average transmission > 80% between 320-430 nm) allowed some red-shifted scattered light (presumably from the walls of the chamber) generated by 5 the probe laser to be transmitted and hence detected by the MCP-PMT. In order to ameliorate this and reduce the background signal, the gate unit was opened 100 ns after the laser pulse to detect fluorescence integrated over a gate-width of 2 s. The optimum gate-width of 2 s (values in the range 1-3 s were compared) is consistent with the CH3O fluorescence lifetimes, calculated to be in the range of 0.9 -1.5 s, using the reported radiative lifetimes for CH3O of 1.5 s (Inoue et al., 1979), 2.2 s (Ebata et al., 1982) and (4  2) s (Wendt and Hunziker, 1979) and using the fluorescence quenching rate coefficients of 10 N2 and O2 (Wantuck et al., 1987) to calculate the rate of quenching at the pressure in the FAGE detection cell ((2.65  0.05) Torr). As the fluorescence lifetime of CH3O(A) in the detection cell was 0.9-1.5 s, delaying the counting of the fluorescence by 100 ns makes very little difference (~ 10%) in the fraction of fluorescence collected.
All LIF signals reported here were normalized to the probe laser power as measured with a laser power meter (Maestro, Gentec-EO) before the start of each LIF measurement. Fluctuations in the relative laser power were monitored via a photodiode 15 (UDT-555UV, Laser Components) during the measurements and were accounted for in the signal normalization. The LIF spectrum was corrected for the laser-scattered background by subtracting the normalized offline signal recorded over 60 s at the end of each LIF measurement using an offline wavelength (offline = 300.29 nm) = (online = 297.79 nm) + 2.5 nm, well away from any CH3O absorption. The signals were large enough that during conditions where CH3O2 concentrations were constant (e.g. in calibrations or during HIRAC experiments where steady-state concentrations were generated) it was 20 established that the laser-wavelength was stable over a long period once the laser wavelength had been tuned to the CH3O transition. Hence, the online wavelength position for CH3O fluorescence detection was found without using a reference cell. Figure 2 shows the laser excitation spectrum centred at ~298 nm in the 3 vibronic band recorded using an increment of  = 10 -3 nm. The spectrum agrees well with previous work (Inoue et al., 1980;Kappert and Temps, 1989;Shannon et al., 2013). Figure 3 shows typical laser excitation scans performed over a narrower range of wavelengths in order to locate (online). The 25 LIF spectra were obtained by using the CH3O or CH3O2 radicals generated in a flow tube described in Sect. 2.3.1, with the flow tube output impinged close to the FAGE sampling inlet. The radicals were generated using the 184.9 nm light output of a Hg Pen-Ray lamp by either the photolysis of methanol in nitrogen to generate CH3O or the photolysis of water vapour in synthetic air (to generate OH) in the presence of methane to form CH3O2. The CH3O radicals were directly detected, while the CH3O2 radicals were first converted to CH3O species by added NO prior to the fluorescence detection cell (Fig. 1). Similar 30 laser scans to the scans shown in Fig. 3 were recorded by using the CH3O2 radicals produced in a steady-state concentration in HIRAC using photolytic mixtures of Cl2/CH4/air as described in Sect. 2.3.2.2. There were no unexpected features in the laser scans recorded when FAGE sampled CH3O2 radicals from HIRAC, consistent with no interference being anticipated in the FAGE measurements of CH3O as there were no other species in HIRAC absorbing at 298 nm and fluorescing at the wavelengths transmitted by the bandpass filter (average transmission > 80 % over 320 -430 nm). 35 In this work the FAGE signals were large enough that during conditions where CH3O2 concentrations were constant (e.g. in calibrations or during HIRAC experiments where steady-state concentrations were generated) it was established that the laser wavelength was stable over a long period once  had been tuned to the CH3O transition. Hence, (online) was found without using a reference cell. We are in the process of developing a reference cell for field measurements in the future, when the concentrations of CH3O2 (and hence CH3O after conversion) will be both lower and more variable over short timescales. 40

Optimisation of the NO concentration for methyl peroxy radical detection
As NO was added ~ 25 mm prior the methoxy detection axis (Fig. 1), some of the methoxy radicals formed by Reaction (R3) 5 reacted further with NO before the fluorescence detection: 10 where M = N2, O2. In addition to the above reactions, CH3O reacts with O2 by Reaction (R4). Figure

FAGE calibrations
CH3O and CH3O2 calibrations were carried out using the conventional radical source employed in fieldwork OH and HO2 calibrations (Heard and Pilling, 2003) that produces radicals in a flow tube impinging just outside the FAGE inlet pinhole (Winiberg et al., 2015) and is described in Sect. 2.3.1. Two methods of calibration have been used for CH3O2: the flow tube method and the kinetics of the self-reaction of CH3O2 carried out in HIRAC. 5

Calibration for methoxy radicals
In the CH3O calibration experiments nitrogen (BOC, > 99.998 %) was used as carrier gas. Part of the N2 flow was passed through a methanol (Sigma Aldrich,  99.9 %) bubbler while the other portion bypassed the bubbler. The gas containing methanol vapour was then passed through a square cross-section flow tube of dimensions 13 × 13 (internal) × 300 mm length with a flow rate of 40 slm (ensuring turbulent flow conditions), controlled by an electronic flow controller (Brooks, 0-100 slm 10 air). The collimated light of a Hg Pen-Ray lamp (LOT-Oriel Hg-Ar) was directed across the flow tube (close to the downstream end) to photolyse methanol vapour. The flow tube output was impinged close to the FAGE inlet to sample CH3O radicals at atmospheric pressure through a 1 mm diameter pinhole (Fig. 1).
The concentration of CH3O radicals was calculated using Eq. (1): where CH3OH, 184.9 nm is the absorption cross section of methanol at 184.9 nm, (6.35  0.28 ) × 10 -19 cm 2 molecule -1 , obtained by averaging reported values (Dillon et al., 2005;Jimenez et al., 2003;Nee et al., 1985), F184.9 nm is the photon flux of 184.9 nm light and t is the irradiation time of the gas. Although it is known, based on end-product analysis, that the scission of O-H 20 bond is a major photolysis channel of methanol at 184.9 nm (Buenker et al., 1984;Porter and Noyes, 1959), the photodissociation quantum yield of CH3O at 184.9 nm, CH3O, 184.9 nm, has not been yet reported. Here it is assumed that CH3O,184.9 nm is equal to the photodissociation quantum yield at 193.3 nm, CH3O, 193.3 nm = 0.86 ± 0.10, which has been reported (Satyapal et al., 1989). In order to determine the methanol vapour concentration in the flow tube, [CH3OH], separate experiments were carried out with the same calibration system to bubble deionised water instead of methanol with the same 25 flow rate. The water vapour concentration, [H2O], was measured using a dew-point hygrometer (CR4, Buck Research Instrument) prior to the flow tube. Then [CH3OH] was calculated using the averaged [H2O] and the vapour pressures pCH3OH and pH2O at the temperatures measured for CH3OH (13 o C) and H2O (15 o C) in the bubbler: Equation 2 assumes that there were no losses of water vapour and methanol vapour by condensation in the tubing connecting the bubbler to the flow tube. This is as expected based on the small difference in temperature between the bubbler (vide supra) and the connecting tubing (typically held at ~ 20 o C) and as the gas going through the bubbler was diluted with the gas bypassing the bubbler. 35 N2O photolysis at 184.9 nm to generate NO (via reaction of the photoproduct (O 1 D) with N2O giving a known yield of NO), which was subsequently measured using a commercial analyser, was used as a chemical actinometer to obtain the product F184.9 nm × t (Winiberg et al., 2015) and hence calculate [CH3O] via Eq. (1). The photolysis time, t, was estimated to be 8.3 ms, using the volumetric flow rate and the geometric parameters of the flow tube (assuming plug flow) and was in turn used to determine F184.9 nm. Although it is the product F184.9 nm × t which is used to calculate [CH3O], any change in the volumetric 40 flow rate between the calibration and actinometry experiments will change t, and hence the product was corrected for any changes in volumetric flow rate. A range of [CH3O] at constant [CH3OH] was produced by changing the electrical current through the Hg lamp between 0 and 20 mA, and hence F184.9 nm, to generate the calibration plot presented in Fig. 5.

Flow tube method
Methyl peroxy radicals were generated by water photolysis at 184.9 nm (Reaction (R7)) to give OH followed by the reaction 15 with excess methane in air (BOC, synthetic BTCA 178) -Reactions (R1)-(R2) to give CH3O2. The calibrations were performed using the set-up described above. Methane (BOC, CP grade, 99.5 %) was flowed at 82.5 sccm to convert OH into CH3, which subsequently reacted rapidly with O2 to form CH3O2. Figure S1 where H2O, 184.9 nm is the absorption cross section of water vapour at 184.9 nm, (7.22  0.22) × 10 -20 cm 2 molecule -1 (Cantrell et al., 1997;Creasey et al., 2000) and H2O, 184.9 nm is the photodissociation quantum yield of OH, which is equal to unity. The values of F184.9 nm and t were determined as described in the Sect. 2.3.1. No loss of CH3O2 by reaction with the HO2 radicals generated by the reaction of H atoms with O2 (R8) was encountered over the residence time of the radicals in the calibration flow tube (~11 ms ) as CH3O2 reacts with HO2 on a ten second timescale as determined using a reaction rate coefficient of 5.2 5 × 10 -12 cm 3 molecule -1 s -1 (Atkinson et al., 2006) and the radical concentrations in the flow tube. The CH3O2 radicals sampled through the FAGE pinhole expansion to a pressure of 2.65 Torr reached the detection region in about 85 ms while the calculated CH3O2 + HO2 reaction half-life at this reduced pressure in the FAGE inlet was thousands of seconds and any change in the CH3O2 concentration is expected to be negligible.

CH3O2 second-order decay method
The principle behind this calibration method is that the second-order decay of CH3O2 is dependent upon its initial concentration, and hence its quantification offers an alternative way to calibrate the signal. The experiments were performed in the HIRAC chamber at 295 K and 1 bar of synthetic air obtained by mixing high purity oxygen (BOC, > 99.999 %)) and nitrogen (BOC, > 99.998 %) in the ratio of O2:N2 = 1:4. Methane (BOC, CP grade, 2-3 × 10 17 molecule cm -3 ) and molecular 5 chlorine (Sigma Aldrich,  99.5 %, 0.3-2.1 × 10 14 molecule cm -3 ) were delivered to the chamber. Eight UV black lamps (Phillips, TL-D 36W/BLB,  = 350-400 nm) housed in quartz tubes mounted radially inside the reactive volume were used to photolyse Cl2 to generate Cl atoms and initiate the chemistry: Numerical simulations using the chemical system described in Table S3 in the Supplementary Information showed that [Cl]0 = 1-6 × 10 6 molecule cm -3 (varied by changing the initial [Cl2]). The high excess of methane (2-3 × 10 17 molecule cm -3 ) relative to [Cl]0 ensured that the reactions of the Cl atoms with the self-reaction products formaldehyde and methanol were 15 negligible. In each HIRAC experiment the lamps were alternatively turned on for 2-3 min and then off over 1-2 min to generate a series of typically 3-4 CH3O2 kinetic decays.
In order to detect CH3O2 the FAGE instrument was coupled to HIRAC through a custom-made ISO-K160 flange to sample the gas with a flow rate of ~ 3 slm. For most measurements, the 1 mm pinhole of the 280 mm long FAGE inlet was sampling ~230 mm from the chamber wall as in the OH measurements reported previously (Winiberg et al., 2015). Additional 20 investigations into any CH3O2 gradient across the ~ 600 mm radius of HIRAC were conducted using measurements of CH3O2 formed by the CH4 reaction with O( 1 D) generated by the photolysis of O3 at 254 nm followed by the reaction of the produced CH3 radical with O2 at 295 K and 1 bar of synthetic air. An extended FAGE inlet (length 520 mm) was used to sample along 500 mm across the chamber starting with the inlet pinhole flush at the wall. A constant concentration of CH3O2 was found (within the 10 % overall error of the measurement) for all the sampled distance 0 -500 mm from the wall (note that 0 mm 25 here refers to the FAGE inlet being at an equivalent position to the wall away from the mounting flange). The absence of a CH3O2 gradient across the chamber provides evidence of the efficacy of the mixing in HIRAC and shows that the wall-loss of CH3O2 is negligible and hence that a shorter inlet, and hence distance from inlet to CH3O2 detection axis could be used in future CH3O2 FAGE measurements within HIRAC, improving further the sensitivity.

Methoxy radical measurements within HIRAC 30
The experiment was carried out in HIRAC at 295 K and 1 bar of N2 (BOC, > 99.998 %), but without any NO added to the FAGE cell (the cell furthest from the pinhole as shown in Fig. 1) so that [CH3O] is measured directly. Initial concentrations in HIRAC were: [CH4]0 = 4.50 × 10 17 molecule cm -3 and [Cl2]0 = 5.57 × 10 15 molecule cm -3 . After adding the reagents into the chamber the lamps (vide supra) were turned on to generate CH3O by Reaction (R6.b).

Flow tube method
The FAGE sensitivity for CH3O2 (CCH3O2) and CH3O (CCH3O), is the slope of the linear regressions in Fig. 5 and Fig. 6, which were CCH3O2 = (4.1 ± 1.4) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 and CCH3O = (5.1 ± 2.2) ×10 -10 counts cm 3 molecule -1 s -1 mW -1 . The error limits, 34 % for CCH3O2 and 43 % for CCH3O, are overall 2 uncertainties calculated using the sum in quadrature of the systematic uncertainties, 33 % for CH3O2 and 42 % for CH3O (details in Section 3.2.1), and the statistical errors from the calibration plots, ~ 8 %. The higher errors in CCH3O compared to CCH3O2 are due to the uncertainty in the methanol concentration, which is not determined directly (vide supra), 1  7 % and the error in the yield of CH3O.
The value of the CH3O photolysis yield from CH3OH reported at 193 nm was used (0.86 ± 0.10), which has an uncertainty of 5 11.63 % at the 1 level (Satyapal et al., 1989).
From the sensitivity factor, C, the limit of detection (LOD) was calculated using Eq. (4) and assuming Poisson statistics appropriate for single photon counting: where S/N is the signal-to-noise ratio, P is the laser power, BKG is the background signal and had a typical value of ~100 counts s -1 , which represents ~50 counts s -1 laser scattered light within the detection cell and ~50 counts s -1 scattered visible light which enters the pinhole from the room with a negligible contribution (1 count s -1 on average) of the detector dark counts, Although CH3O2 has not been measured specifically in the atmosphere, there have been several calculations of its 20 concentration using numerical models. In general, the concentration of CH3O2 is a function both of the loadings of volatile organic compounds (VOCs) and the levels of NOx. For the clean, remote environments at Cape Verde in the tropical Atlantic ocean and in the Borneo rainforest [CH3O2] is calculated to peak around 6 × 10 8 molecule cm -3 and about 2 × 10 8 molecule cm -3 , respectively at noon using the modeling studies reported by Whalley et al. (Whalley et al., 2010;Whalley et al., 2011). Therefore, it should be possible using the FAGE conversion method to CH3O and for an averaging time of 1 hour (vide supra) 25 to achieve a measurement of atmospheric levels of CH3O2 in such clean environments, and shorter averaging times in some cases. Further optimizations of FAGE sensitivity can be achieved by the removal of the fibre optic cables to deliver the probe laser beam directly to the CH3O detection cell to increase the laser power and, by increasing the pulse repetition frequency above the current value of 5 kHz (but without significant reduction in the pulse energy). The present investigations into the change of sensitivity with pressure in the range from 2.65-10.00 Torr found that 2.65 Torr is the optimum value in this pressure 30 interval. The result suggests that, by reducing the pressure in the above range of values, the decrease in fluorescence due to the reduction in the CH3O number density was overcome by the increase in the fluorescence quantum yield due to a lower fluorescence quenching rate. Another reason could be that the characteristics of the jet expansion and/or the ensuing flow to the LIF detection region change with pressure, leading to a more favourable transmission of radicals to the detection region, but it is difficult to test this experimentally. Hence an additional improvement in the sensitivity might be obtained by using a 35 lower detection cell pressure than the current value of 2.65 Torr using a more powerful pump. It should be also noted that the distance from the inlet pinhole to the laser-axis in the CH3O and CH3O2 fluorescence cell (Figure 1, ~ 580 mm) is considerably longer than the corresponding distance in the ground-based field fluorescence cell for OH and HO2 detection (88 mm), and improvements in sensitivity would be expected for a shorter pinhole-to-laser excitation distance for CH3O2. The further optimizations of sensitivity and the planned construction of a reference cell to find the online wavelength position could 40 potentially enable CH3O2 measurements to be made in urban environments where CH3O2 concentrations are estimated to be considerably lower, for example a few 10 7 molecule cm -3 based on modeling results (Whalley et al., to be submitted).
The calibrations using the flow tube ("wand") method have been performed under water vapour concentrations similar to the ambient [H2Ovapour] but few orders of magnitude higher than those present in the HIRAC chamber experiments. In contrast with [H2Ovapour], the methane concentrations used in the "wand" method were similar to [CH4] present in HIRAC but higher than [CH4] in the atmosphere. However, as detailed in this paragraph, the effects of methane and water on our sensitivity are minimal. Estimations using the reported fluorescence quenching rate coefficient of CH3O(A) by CH4, kquench.CH4 = 1.05 × 10 -10 5 s -1 , (Wantuck et al., 1987) and the concentrations of CH4 in the LIF detection cell for the calibrations using the flow-tube (1.7 × 10 14 molecule cm -3 and 3.4 × 10 14 molecule cm -3 , corresponding to 5.0 × 10 16 molecule cm -3 and 1.0 × 10 17 molecule cm -3 , respectively in the flow tube) resulted in only ~ 1-2% lower fluorescence quantum yield compared to the value determined in the absence of CH4. No literature value has been found for the fluorescence rate coefficient of CH3O(A) fluorescence by H2O vapour. However, even if it assumed to be as large as the above reported value for CH4 (kquench.CH4), only a few percent decrease 10 in the fluorescence quantum yield is computed (compared with a water concentration of zero) for the levels of H2O vapour which are present at the CH3O2 FAGE detection axis when using the "wand" calibration method. These levels (1-2% v/v) are similar to a typical water vapour concentration in the atmosphere. A very good agreement has been obtained between the calibration factors for CH3O2 detection with two different concentrations of water vapour in the flow tube: 7.5 × 10 16 molecule cm -3 or 3.0 × 10 17 molecule cm -3 (corresponding to 2.6 × 10 14 molecule cm -3 and 1.0 × 10 15 molecule cm -3 , respectively in the 15 FAGE cell) as shown in Figure 6 in Sect. 2.3.2.1. This very good agreement for H2O vapour and the above calculations for CH4 support the use of the flow tube method for the FAGE calibration of the CH3O2 concentrations.

Methyl peroxy calibration using kinetics of the CH3O2 second-order decay
An alternative method of calibration for CH3O2 was to generate CH3O2 radicals in HIRAC to monitor the temporal decay of 20 the CH3O2 FAGE signal once the photolysis lamps were turned off. Figure 7 shows an example of a decay in the CH3O2 signal generated by extinguishing the HIRAC lamps following the production of CH3O2 by the Cl atom initiated oxidation of CH4 in the presence of O2 (Reactions (R12) and (R2)). In the absence of other processes, the loss of CH3O2 is described by the integrated second-order rate law equation describing the CH3O2 self-reaction (Reaction (R6)): where [CH3O2]t is the methyl peroxy concentration at reaction time t, [CH3O2]0 is the initial concentration when the lights are switched off and kobs is the observed rate coefficient (which is not equal to k6, see below). Using SCH3O2 is the signal measured by FAGE and CCH3O2 is the instrument sensitivity, Eq. (6) is obtained for the temporal profile of 30 the methyl peroxy signal: In Eq. (6) (SCH3O2)t and (SCH3O2)0 are the signal at time t and t = 0 respectively. 35 Eq. (6) was fitted to the experimental decays of SCH3O2 (see Fig. 7 as an example) fixing kobs to the IUPAC recommendation, kobs = (4.8 ± 1.1) × 10 -13 cm 3 molecule -1 s -1 , in order to obtain CCH3O2. Eighteen CH3O2 decays were analysed, which yielded an average value of CCH3O2 = (5.6  1.7) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 . The error limit, 30 %, is the 2 composite error calculated as the sum in quadrature of the total systematic uncertainty, 29 % (see Section 3.2.2), and the average random error of all determinations, with 8 %, taken as two standard errors in the fit of Eq. (6) to the CH3O2 temporal decays. This value 40 agrees well with CCH3O2 = (4.1 ± 1.4) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 obtained from the flow-tube calibration method (section 3.1.1). Based on the lack of a measurable CH3O2 radical gradient across HIRAC (Section 2.3.2.2, vide supra) it is assumed that the loss of CH3O2 to the walls of HIRAC in these experiments was negligible over the timescale of 1-2 min of the temporal decay measurements. Our finding is consistent with previous results showing that the heterogeneous wall-loss rates for CH3O2 were significantly lower than the corresponding removal rates of HO2 (Miyazaki et al., 2010;Mihele et al., 1999;Fuchs et al., 15 2008). Using a 30 cm long glass tube of 2 cm diameter, Miyazaki et al. measured that the heterogeneous removal efficiency for CH3O2 was six times lower than for HO2. The HO2 wall-loss rate coefficient at room temperature and 1000 mbar in HIRAC was found to be of ~ 10 -2 s -1 (Winiberg et al., 2015). Therefore, it can be expected that the wall-loss rate coefficient of CH3O2 in HIRAC was kloss  10 -3 s -1 and so is not considered in the analysis here for CH3O2 decays which typically last for ~ 100 s.
In order to investigate the sensitivity of CCH3O2 obtained by the kinetic analysis of the CH3O2 decay to kloss higher than 10 -3 s -1 , 20 a wall-loss rate coefficient of 10 -2 s -1 was included in the analysis of the experimental decays of CH3O2 to obtain CCH3O2, but only an increase in CCH3O2 of about 6 % on average was seen. A small deviation of the experimental data from the fit was obtained at the end of the measurements whether or not kloss was included in the analysis (Fig. 7). The role of potential secondary chemistry at later times of the reaction will be investigated in future kinetic studies of the CH3O2 self-reaction.
In order to check the validity of Eq. (7) in the presence of HO2 removal by self-reaction and wall-loss, numerical simulations were performed to generate CH3O2 decays using a system incorporating the chemistry described by Reactions 20 (R4), (R6), R(13) and R(14) (vide infra) and a heterogeneous loss of HO2, kloss(HO2) (Supplementary Information). The rate coefficients were sourced from the IUPAC preferred values at 298 K (Table S3 in Supplementary Information) and kloss(HO2) was varied. The simulated decays of [CH3O2] vs. time were analysed using Eq. (5) (see Fig. S3 as an example) and gave an average observed rate coefficient of kobs = 4.7 × 10 -13 cm 3 molecule -1 s -1 , which is only 2 % lower than the IUPAC recommendation, for kloss(HO2) varied between 0.01-0.10 s -1 and, hence confirm the applicability of Eq.

Flow tube method
The 2 error associated with CCH3O2 of 34 % obtained by the flow tube method (CCH3O2 = (4.1 ± 1.4) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 ), represents the overall uncertainty calculated using the sum in quadrature of the systematic uncertainty, 33 %, and the statistical error from the calibration plots, ~ 8 %. The overall 34 % uncertainty is similar to the previous estimated total uncertainty, 36 %, in the use of the same method for calibration of OH and HO2 measurements in HIRAC, where no CH4 35 is added (Winiberg et al., 2015). The flow tube method is a proven method to generate known amounts of OH and HO2 by the photolysis of H2O at 184.9 nm in order to calibrate field instruments (Heard and Pilling, 2003).
The largest contribution to the total error of the method came from the 28 % total uncertainty in the photon flux of the calibration source, F184.9nm. The product F184.9 nm × t is determined using N2O actinometry relying on the measurement of [NO] in trace amounts (0.2-1.5 ppbv) using a commercial NO analyser (Thermo Electron Corporation 42C) followed by the 40 data analysis using four rate coefficients each with ~ 10 % uncertainty (Burkholder et al., 2015). Although it is the product F184.9 nm × t which is directly determined by the actinometric method and used to calculate the concentration of radicals to calibrate FAGE (Eq. (3)), any difference in the volumetric flow rate between the calibration and actinometry experiments will change t. Therefore, the uncertainty in t, 2 %, needs to be accounted for. The contributions from the rest of the terms in Eq.
(3) to the systematic uncertainty in the determination of [CH3O2] by this method were as follows: 6 % total error in H2O, 184.9nm (Cantrell et al., 1997), 10 % uncertainty in [H2O], taken from the instrumental uncertainty of the hygrometer and 4 % error in 5 the yield of CH3O2 produced by the OH conversion into CH3 followed by the CH3 + O2 reaction. The contribution of the uncertainties in the FAGE measurements to the 33 % overall systematic uncertainty in the calibration were estimated to consist of 12 % in the online FAGE signal and 6 % uncertainty in the laser power measured by the laser power meter and used to normalize the data. The uncertainty associated with the online signal, 12 % at 2 level, was calculated as the average deviation of the signal value due to the error limits of  5 × 10 -4 nm in the online wavelength position (see the typical laser excitation 10 scans shown in Fig. 3).

CH3O2 second-order decay calibration
The largest contribution to the calculated overall 2 uncertainty of 30 % in CCH3O2 obtained by the CH3O2 second-order decay method (CCH3O2 = (5.6  1.7) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 ), derives from the 23 % error in the IUPAC preferred value of the observed rate coefficient for the effective CH3O2 self-reaction, kobs = (4.8 ± 1.1) × 10 -13 cm 3 molecule -1 s -1 15 (Atkinson et al., 2006). It is instructive to examine the origin of the 23 % error. The studies which led to the IUPAC recommendation utilized the UV-absorption of CH3O2, typically at 250 nm, and the determined quantity was the ratio between the observed rate coefficient and the absorption cross section of CH3O2, kobs/250nm. IUPAC and the Jet Propulsion Laboratory (JPL) recommend 3.9 × 10 -18 and 3.8 × 10 -18 cm 2 molecule -1 , respectively for 250nm (Atkinson et al., 2006;Burkholder et al., 2015). The JPL recommendation (Burkholder et al., 2015) is the cross section obtained by the re-evaluation of the previous 20 reported UV-absorption spectra by Tyndall et al. in 2001(Tyndall et al., 2001, yielding 250nm = 3.78 × 10 -18 cm 2 molecule -1 . The remaining contributions to the uncertainty in the calibration using the CH3O2 second-order decay method are: 6 % error in the laser power, 12 % uncertainty in the online signal determined by how well the laser is able to find the online 30 wavelength position (vide supra) and 10 % error in (SCH3O2)0 in Eq. (6), the value of the CH3O2 signal at the moment when the HIRAC lamps were turned off to generate a second-order decay.

Comparison between the FAGE sensitivities for CH3O2 obtained by the two calibration methods
The FAGE sensitivity factor obtained using the flow tube method, CCH3O2 = (4.1 ± 1.4) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 , is 27 % lower but has overlapping error limits with the result found using the CH3O2 second-order decay method, CCH3O2 = 35 (5.6  1.7) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 (uncertainties quoted to 2). The calculated overall error in the CH3O2 second-order decay method, 30 %, is similar to the total uncertainty in the flow tube method, 34 %. The flow tube method is known to reliably generate accurate concentrations of radicals and has been used for many years in the calibration of FAGE instruments employed in field measurements of OH and HO2 (Heard and Pilling, 2003). The flow tube method has also been validated by using alternate methods of calibration, for example using the decay of a hydrocarbon in the HIRAC chamber to 40 obtain [OH] (Winiberg et al., 2015). The method of using a time-resolved kinetic quantity to derive a calibration factor was validated for HO2 in HIRAC, where CHO2 obtained from analysis of the temporal decay of HO2 agreed with CHO2 from the flow tube method (Winiberg et al., 2015). These results suggests that the sensitivity of the FAGE system, represented by the value of C, is not changed between sampling from the calibration flow tube and sampling from within HIRAC itself.
The accuracy of the CH3O2 temporal decay method is largely determined by the accuracy of kobs (see section 3.2.2. above). 5 The quantity measured in the previous kinetic studies of CH3O2 + CH3O2 is kobs/250nm and hence the accuracy of kobs is directly affected by any systematic errors in the determination of 250nm. In order to make CCH3O2 derived from the temporal decay and flow tube methods of the same, the value of kobs would need to be reduced by ~ 25 %, which in turn requires a ~ 25 % reduction in 250nm. It is noted that the UV-absorption spectrum of CH3O2 is relatively broad and hence may prevent a selective detection due to the difficulty to discriminate from the potential presence of other species also absorbing around 250 nm, such as Cl2 10 and CH3CHO used in concentrations as high as 10 16 molecule cm -3 , while [CH3O2] was ~ 10 13 molecule cm -3 (Dagaut and Kurylo, 1990;Roehl et al., 1996). As the absorption cross sections of Cl2 and CH3CHO at 250 nm lay in the range 10 -21 -10 -22 cm 2 molecule -1 (Keller-Rudek et al., 2013), the unaccounted for absorption of these species may have led to an overestimation of 250nm(CH3O2).
As noted in the 2001 review by Tyndall et al. (Tyndall et al., 2001), none of the previous laboratory studies of the CH3O2 15 recombination measured [CH3O2] by any method other than UV-spectroscopy. In addition, the traditional time-resolved measurements of CH3O2 used high CH3O2 concentrations (10 13 -10 15 molecule cm -3 ) and, as the self-reaction is fairly slow, Tyndall et al. stated that the results were potentially affected by secondary chemistry (Tyndall et al., 2001). Therefore, there is a need for the use of a complementary technique in the kinetic study of this reaction, for example by LIF as described in this paper, which may offer some advantages to probe CH3O2 selectively in the absence of interferences from other species. In 20 addition, LIF is more sensitive and hence requires significantly lower radical concentrations ([CH3O2]0 = (1-3) × 10 11 molecule cm -3 here) than for the UV-absorption studies which may help to minimize potential secondary chemistry.

Methoxy radical measurement within HIRAC
The typical concentration of [O2] = 5 × 10 18 molecule cm -3 used in the HIRAC experiments described above was lowered in 25 some experiments to decrease the consumption of CH3O by O2 via Reaction (R4). In this manner, a concentration of methoxy radicals was obtained above the FAGE limit of detection in HIRAC to enable a direct measurement over few minutes. The chamber was filled with high purity nitrogen (> 99.998 %), but the ~ 6 m long N2 delivery pipe was purposely incompletely purged before the experiment in order to deliver trace levels of oxygen to HIRAC. The initial Cl2 concentration in these experiments was 5.6 × 10 15 molecule cm -3 and hence is 1-2 orders of magnitude higher than [Cl2]0 used in the kinetic 30 experiments above in order to generate higher [Cl] and hence [CH3O]. The concentration of CH3O during the experiment was computed by using the FAGE calibration factor for methoxy radicals generated from the photolysis of methanol in N2, CCH3O = (5.1 ± 2.2) ×10 -10 counts cm 3 molecule -1 s -1 mW -1 (Sect. 3.1.1). The temporal profile of CH3O is shown in Fig. 8, together with a numerical simulation of CH3O(t) using a chemistry system described in the Supplementary Information. The best fit to the experimental CH3O concentration profile was obtained for [O2] = (5.4  0.6) × 10 15 molecule cm -3 , i.e. around 35 0.02 % relative to N2. The numerical simulations showed that Cl2 consumption was dominated by the reaction with CH3 radicals, present at a relatively high concentration, explaining the ~ 50 % decrease in [CH3O] observed during its temporal measurement shown in Fig. 8. The Supplementary Information (Fig. S5) shows the concentration profiles of Cl2, Cl, CH3 and CH3O2 obtained by numerical simulations performed over ~ 2 min. These results demonstrate the capability to measure an absolute concentration of CH3O radicals in a simulation chamber, with CH3O representing a further model target species for the validation of chemical mechanisms for the chemical oxidation of 10 VOCs. However, it is recognized that the experiments need to be performed at reduced [O2], and that [O2] needs to be known a priori in order to test robustly the accuracy of the chemical mechanism and underlying kinetic parameters.

Conclusions
Currently there is no measurement of the absolute concentration of CH3O2 radicals in the atmosphere. In this work the FAGE technique has been extended by adding the capability to detect CH3O2 and CH3O radicals to the more typical measurement of 15 OH and HO2 radicals. The method enables the speciated and sensitive detection of CH3O2 radicals by converting CH3O2 into CH3O by reaction with NO and detecting the resultant CH3O by LIF with excitation at ca. 298 nm. The limit of detection of the method obtained using the radical source commonly employed to provide accurate concentrations of OH with added CH4, is 3.8 × 10 8 molecule cm -3 for a signal-to-noise ratio of 2 and 5 min time resolution and reduces to 1.1 × 10 8 molecule cm -3 for S/N = 2 and 1 hour averaging time. Therefore, the method has the potential to be used in field measurements of the diurnal 20 profiles of CH3O2 in clean air with low NOx levels, such as remote continental environments and in the marine boundary layer.
Further improvements of the FAGE sensitivity could be achieved via the increase in the laser repetition frequency above the current value of 5 kHz, a decrease in the detection chamber pressure (currently ~ 2.65 Torr), and the use of a shorter distance between the inlet sampling pinhole and the fluorescence detection axis (presently a long distance of ~ 580 mm). The method is also demonstrated for the direct detection of CH3O, in the absence of added NO to the fluorescence cell. The limit of 25 detection for CH3O determined using the conventional radical source for S/N = 2 and 5 min averaging time is 3.0 × 10 8 molecule cm -3 .
Additional investigations into the FAGE sensitivity for CH3O2 were carried out in the HIRAC simulation chamber at Leeds, by studying the kinetics of the second-order decays of CH3O2 by its self-reaction. The second-order decays of CH3O2 were analysed by fixing the observed rate coefficient to the IUPAC recommendation, kobs = (4.8 ± 1.1) × 10 -13 cm 3 molecule -1 s -1 , (Atkinson et al., 2006) in the fitting routine to extract the FAGE sensitivity factor for CH3O2, CCH3O2. The obtained value, CCH3O2 = (5.6  0.9) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 , agrees well with the result found using the conventional radical 5 source, CCH3O2 = (4.1 ± 0.7) × 10 -10 counts cm 3 molecule -1 s -1 mW -1 (uncertainties quoted to 1). The two values have overlapping error limits at 1 level.
In addition to the quantitative detection of CH3O2, experiments were carried out to measure CH3O generated as a product by the CH3O2 self-reaction in HIRAC. Oxygen was present at a significantly lower concentration to reduce the consumption rate of CH3O by reaction with O2 in order to enable the measurement. Good agreement between the experimental data and 10 [CH3O] generated by numerical simulations using a model describing the chemical system was obtained, demonstrating the capability to quantitatively measure CH3O. As well as CH3O2, a measurement of CH3O will be useful as a further model target in future mechanistic studies of atmospherically relevant chemical systems within HIRAC.