Measurement of interferences associated with the detection of the hydroperoxy radical in the atmosphere using laser-induced fluorescence

One technique used to measure concentrations of the hydroperoxy radical (HO2) in the atmosphere 1 involves chemically converting it to OH by addition of NO and subsequent detection of OH. However, some 2 organic peroxy radicals (RO2) can also be rapidly converted to HO2 (and subsequently OH) in the presence of 3 NO, interfering with measurements of ambient HO2 radical concentrations. This interference must be 4 characterized for each instrument to determine to what extent various RO2 radicals interfere with measurements 5 of HO2 and to assess the impact of this interference on past measurements. The efficiency of RO2 to HO2 6 conversion for the Indiana University Laser-Induced Fluorescence – Fluorescence Assay by Gas Expansion (IU7 FAGE) instrument was measured for a variety of RO2 radicals. Known quantities of OH and HO2 radicals were 8 produced from the photolysis of water vapor at 184.9 nm, and RO2 radicals were produced by the reaction of 9 several volatile organic compounds with OH. The conversion efficiency of RO2 radicals to HO2 was measured 10 when NO was added to the sampling cell for conditions employed during several previous field campaigns. For 11 these conditions, approximately 80% of alkene derived RO2 radicals and 20% of alkane derived RO2 radicals were 12 converted to HO2. Based on these measurements, interferences from various RO2 radicals contributed to 13 approximately 35% of the measured HO2 signal during the Mexico City Metropolitan Area (MCMA) 2006 14 campaign, where the measured VOCs consisted of a mixture of saturated and unsaturated species. However, this 15 interference can contribute more significantly to the measured HO2 signal in forested environments dominated by 16 unsaturated biogenic emissions such as isoprene. 17


Introduction
The hydroxyl radical (OH) is one of the primary oxidants in the atmosphere (Levy, 1972).The reaction of OH radicals with volatile organic compounds (VOCs) leads to the production of peroxy radicals, both the hydroperoxy radical (HO 2 ) and organic peroxy radicals (RO 2 ), which in the presence of nitrogen oxides (NO x = NO + NO 2 ) can lead to the production of ozone and secondary organic aerosols in the atmosphere.As a consequence, the development of effective control strategies for the formation of these pollutants requires an accurate understanding of the OH, HO 2 , and RO 2 radical chemistry in the atmosphere.Measurements of OH and HO 2 (together HO x ) can provide a robust test of our understanding of this complex oxidation chemistry.
These results question our understanding of HO x radical chemistry and the ability of models to simulate future changes in the chemical composition of the atmosphere.However, a recent intercomparison of several instruments measuring HO 2 found that the agreement between the different instruments was variable, although the measurements were highly correlated (Fuchs et al., 2010).While the differences were within the combined uncertainties of the measurements, there were several measurement periods when the differences could not be explained by instrumental uncertainties.These results suggested the possibility of potential interferences in the HO 2 measurement technique.
Laser-induced fluorescence using the fluorescence assay by gas expansion technique (LIF-FAGE) is a common method for measuring HO 2 radicals in the atmosphere.In this technique HO 2 radicals are measured indirectly after sampling ambient air at low pressure through chemical conversion to OH by addition of NO as shown in Reaction (R1) and subsequent detection of OH by LIF: It was previously believed that the detection of HO 2 radicals using the LIF-FAGE technique was free from interferences from the reaction of RO 2 radicals with NO, as model simulations and measurements suggested that the rate of conversion of RO 2 radicals to HO 2 by Reactions (R2) and (R3) and subsequent conversion to OH through Reaction (R1) was negligible due to the slow rate of Reaction (R3) under the reduced oxygen concentration in the low-pressure LIF-FAGE cell and the short reaction time between injection of NO and detection of OH (Heard and Pilling, 2003).
For example, RO 2 radicals produced from the OH-initiated oxidation of small alkanes were found to produce a negligible yield of HO 2 (Stevens et al., 1994;Kanaya et al., 2001;Tan et al., 2001;Creasey et al., 2002;Holland et al., 2003).However, recent laboratory studies have shown that there are interferences associated with measurements of HO 2 us-ing this technique from the conversion of RO 2 radicals derived from the OH-initiated oxidation of alkenes and aromatics to HO 2 (and subsequently OH) by reaction with NO.Measured RO 2 -to-HO 2 conversion efficiencies of 95 % for the peroxy radicals derived from the OH-initiated oxidation of propene and 86 % for the peroxy radicals derived from the OH-initiated oxidation of benzene have been reported (Fuchs et al., 2011).The high conversion efficiency of alkene-based peroxy radicals to HO 2 is due to the ability of the β-hydroxyalkoxy radicals produced from Reaction (R2) to rapidly decompose, forming a hydroxyalkyl radical which then reacts rapidly with O 2 and leads to the production of a carbonyl compound and HO 2 (Fuchs et al., 2011;Whalley et al., 2013).The conversion efficiency depends on the instrumental characteristics and the configurations employed (Fuchs et al., 2011;Whalley et al., 2013).As a result, this interference must be characterized for all LIF-FAGE instruments for the accurate analysis of ambient HO 2 .This paper will describe the characterization of the RO 2 interferences associated with the Indiana University LIF-FAGE instrument under several past campaign configurations.These include the Mexico City Metropolitan Area (MCMA) campaign in 2006 (MCMA-2006;Dusanter et al., 2009a, b), the Community Atmosphere-Biosphere INteractions EXperiments (CABINEX) in 2009 (Griffith et al., 2013), and the California Research at the Nexus of Air Quality and Climate Change campaign in Los Angeles (CalNex) in 2010 (Griffith et al., 2016).The impact of this interference on the previously published results from the MCMA 2006 campaign and a reanalysis of these HO 2 measurements will be discussed.

Instrument description
The Indiana University LIF-FAGE instrument (IU-FAGE) has been described in detail previously (Dusanter et al., 2008(Dusanter et al., , 2009a;;Griffith et al., 2013Griffith et al., , 2016)).In the LIF-FAGE technique, OH radicals are detected by laser-induced fluorescence after expansion of ambient air to low pressure.This enhances the OH fluorescence lifetime, allowing temporal filtering of the fluorescence from laser scatter (Heard and Pilling, 2003).A diagram of the IU-FAGE detection cell is illustrated in Fig. 1.Ambient air is expanded through an orifice between 0.635 and 1.016 mm diameter located at the top of a cylindrical nozzle (5 cm in diameter and 20 cm long).The size of the orifice was kept unchanged during each campaign but was varied between the different campaigns reported here.Two scroll pumps (Edwards XDS 35i) connected in parallel maintain a pressure inside the cell between 4 and 7.5 Torr depending on the sampling size of the orifice and the pumping speed, resulting in a flow rate between 3 and 10 SLPM through the sampling nozzle.The original IU-FAGE laser system used in this study and in the MCMA 2006 campaign consisted of a Spectra-Physics Navigator II YHP40-532Q diode-pumped Nd:YAG laser that produced approximately 5.5 W of radiation at 532 nm at a repetition rate of 5 kHz.This laser pumped a Lambda Physik Scanmate 1 dye laser (Rhodamine 640, 0.25 g L −1 in isopropanol) that produced tunable radiation around 616 nm, which was frequency doubled to generate 2 to 20 mW of radiation at 308 nm (∼ 20 ns pulse width).This laser system was recently replaced with a Spectra-Physics Navigator II YHP40-532Q that produces approximately 8 W of radiation at 532 nm at a repetition rate of 10 kHz that pumps a Sirah Credo Dye laser (255 mg L −1 of Rhodamine 610 and 80 mg L −1 of Rhodamine 101 in ethanol), resulting in 40 to 100 mW of radiation at 308 nm.Measurements of the conversion efficiencies were similar for the two laser systems.After exiting the dye laser, the beam was focused onto the entrance of a 12 m optical fiber to transmit the radiation to the sampling cell.In the detection cell, the laser crosses the expanded air perpendicular to the flow in a White cell configuration with approximately 24 passes.
OH radicals are excited using the A 2 + υ = 0 ← X 2 υ = 0 transition near 308 nm (Stevens et al., 1994).The net signal is measured by turning the wavelength on and off resonance in successive modulation cycles.A reference cell where OH is produced by thermal dissociation of water vapor is used to ensure that the laser is tuned on and off the OH transition.The OH fluorescence is detected using a microchannel plate photomultiplier tube (MCP-PMT) detector (Hamamatsu R5946U-50), a preamplifier (Stanford Research System SR445), and a gated photon counter (Stanford Research Systems SR400).The MCP-PMT is switched off during the laser pulse through the use of electronic gating allowing the OH fluorescence to be temporally filtered from laser scattered light.
A Teflon injector located approximately 2.5 cm below the inlet and 17.5 cm above the detection axis (Fig. 1) allowed for the addition of NO (Matheson, 99.8 %) to convert ambient HO 2 to OH through Reaction (R1).The fraction of HO 2 (C HO 2 ) converted into OH was measured during calibration experiments (Dusanter et al., 2008).The NO flow (approximately 1-3 × 10 13 cm −3 ) maximized the conversion of HO 2 into OH while minimizing the removal of OH by the OH + NO reaction.

Instrument calibration for OH and HO 2
The IU-FAGE instrument is calibrated by producing known quantities of OH and HO 2 radicals from the photolysis of water vapor in air (Reactions R4 and R5;Dusanter et al., 2008): The calibration source consists of an aluminum flow reactor (1.27 × 1.27 × 30 cm) equipped with quartz windows on two sides (Fig. 2).The light source consists of a low-pressure mercury pen lamp (UVP, LLC) housed in an aluminum cartridge that is continuously purged with dry nitrogen to prevent light absorption by gases in addition to helping to stabilize the lamp temperature.The radiation from the lamp passes through a band-pass filter centered at 185 nm (Acton Research) prior to entering the reactor and is detected by a photodiode.The lamp housing can be adjusted along the length of the calibrator to measure the loss of radicals between the source region and the exit of the calibrator.The concentration of OH and HO 2 radicals produced by the calibration source can be determined from the following equation: (1) In this equation, φ OH+H is the quantum yield of OH from water photolysis and σ H 2 O is the absorption cross section of water (7.14 × 10 −20 cm −2 molecule −1 ; Cantrell et al., 1997;Hofzumahaus et al., 1997;Creasey et al., 2000).The product of the photon flux (F ) and the photolysis time (t) can be determined from oxygen actinometry, as the photolysis of oxygen at 185 nm leads to the production of ozone (Reactions R6 and R7; Okabe, 1978): The concentration of HO x radicals can thus be calculated from measured concentrations of water and ozone using Eq.(2) (Heard and Pilling, 2003;Holland et al., 2003): Here φ O 3 is the quantum yield of O 3 from oxygen photolysis and σ O 2 is the absorption cross sections of O 2 , which must be experimentally determined for each pen lamp due to the overlap of the highly structured absorption spectrum of O 2 and the lineshape of the emission at 184.9 nm.The lineshape depends on the operating conditions of the lamp as a result of line reversal and potential fluorescence of the fused silica envelope (Cantrell et al., 1997;Hofzumahaus et al., 1997;Lanzendorf et al., 1997).

Measurement of the RO 2 conversion efficiency to HO 2
Various alkenes (isoprene; methyl vinyl ketone, MVK; methacrolein, MACR; methyl ethyl ketone, MEK; ethene; trans-2-butene; tetramethylethylene, TME), alkanes (propane, butane, octane), and aromatic compounds (toluene) were used to measure the conversion efficiency of RO 2 radicals to HO 2 .These VOCs were added to the main calibrator flow, either by direct addition of a gas mixture or by bubbling air through the liquid compound (Tables S1  and S2 in the Supplement), approximately 190 ms prior to the radical source to ensure that the added VOC was well mixed into the humid air flow before photolysis within the calibration source.The concentration of each VOC added to the calibrator was increased to react and remove the majority of the OH produced in the calibrator, resulting in RO 2 concentrations that were approximately equal to the concentration of OH reacted away.These RO 2 radicals are then sampled into the IU-FAGE instrument.Addition of NO inside the detection axis converts a fraction of the RO 2 radicals to HO 2 through Reactions (R2) and (R3).Since RO 2 is produced together with HO 2 in the calibrator, there is a subsequent conversion of both RO 2 and HO 2 into OH in the IU-FAGE cell, which is then detected by LIF.
Figure 3 illustrates two typical experiments designed to measure the conversion efficiency of RO 2 radicals to HO 2 in the IU-FAGE instrument.The total HO x signal is defined as the sum of the total OH (S OH ) and HO 2 (S HO 2 ) produced by the mercury pen lamp in the absence of the added VOC (Eq.3): (3) The OH concentration produced by the pen lamp is measured at the beginning, middle, and at the end of each experiment to ensure that the concentrations remained stable (experimental mode 1 in Fig. 3).Once the OH signal (S OH ) stabilizes, NO is added internally to the detection cell to convert HO 2 into OH and measure the total HO x signal (S HO x ; mode 2 in Fig. 3).The conversion efficiency of HO 2 to OH is defined by Eq. ( 4): S OH 0 is the OH signal after accounting for the loss of OH on the walls of the calibrator (approximately 20 %).The wall loss for HO 2 is negligible in the calibrator (Dusanter et al., 2008).Next, internal NO addition is stopped and the OH signal is measured again to ensure the stability of radical production during the experiment.The VOC is then added to the calibration system, resulting in a decrease in the observed OH signal (mode 3 in Fig. 3).The remaining OH signal in the presence of the VOC is denoted as S OH+VOC .For alkenes such as isoprene, the fast reaction with OH results in an almost total removal of OH radicals from the calibration source and S OH+VOC is close to zero.However, for less reactive alkanes such as butane, the added VOC concentration was often not sufficient to completely remove OH radicals due to the short reaction time in the calibrator, resulting in a nonzero S OH+VOC signal.The conversion efficiency in which OH radicals are converted to RO 2 radicals (C OH→RO 2 ) is defined by Eq. ( 5), which is derived from integrating the expressions for the rate of OH loss and the rate of RO 2 production from the OH + VOC reaction: where k w t is the product of the rate constant for the reaction of OH radicals on the wall of the calibration source with the reaction time t, reflecting the measured loss of OH on the walls of the calibrator (Dusanter et al., 2008).
The subsequent addition of NO to the detection cell will convert a fraction of RO 2 radicals and HO 2 radicals to OH (mode 4 in Fig. 3).The conversion efficiency of RO 2 to OH (C RO 2 →OH ) is determined by multiplying the fraction of RO 2 radicals converted to HO 2 (f RO 2 →HO 2 ) with the conversion efficiency of HO 2 to OH (C HO 2 →OH ): The signal due to RO 2 radicals (S RO 2 ) is defined as the original OH signal (S OH 0 ) multiplied by the conversion efficiency of OH radicals to RO 2 radicals (C OH→RO 2 ) and multiplied by the conversion efficiency of RO 2 to OH (C RO 2 →OH ; Eq. 7): For OH + VOC reactions that lead to the production of HO 2 with a yield of y (OH + benzene and toluene for example; Klotz et al., 1998;Nehr et al., 2011), the OH to RO 2 conversion efficiency (C OH+VOC ) must be multiplied by the overall yield (1 − y) of RO 2 radicals produced from the OH + VOC reaction.Taking this yield into account, the signals due to RO 2 and HO 2 radicals become The measured OH signal under these conditions (S RO x ) reflects the contribution of RO 2 , HO 2 , and unreacted OH radicals (experimental mode 4): Combining Eqs. ( 3), ( 6), and (9a) results in an expression for the fraction of RO 2 radicals converted to HO 2 (f RO 2 →HO 2 ) that can be expressed as the measured signals for each experimental mode (S OH , S HO x , S OH+VOC , S RO x ) as seen in Eq. ( 10): When the yield of HO 2 from the OH + VOC reaction is zero (y = 0), and under conditions where all the OH radicals are converted to RO 2 (S OH+VOC = 0), the above equation (with Eqs. 3 and 4) simplifies to the following: Because this method cannot distinguish between the different peroxy radicals that could be produced from each OH + VOC reaction, the measured conversion efficiency reflects the average conversion efficiency of all peroxy radicals for a given VOC.

Results
The pressure and flow conditions for the three campaigns conducted with the IU LIF-FAGE instrument are summarized in  santer et al., 2008, 2009a).The reason for this discrepancy is unclear and may indicate problems in precisely recreating the flow conditions during this campaign in these laboratory experiments.In addition, the NO flow rate was varied during MCMA-2006 in order to maximize the HO 2 -to-OH conversion efficiency and to quantify the photolytic interference associated with high NO concentrations in the detection cell.Thus, it is possible that the actual flow rate used to maximize the conversion efficiency was slightly greater than the 1 sccm reported.Since it is not known whether the flow was greater than the 1 sccm that was measured, or whether the flow conditions led to more efficient mixing, we chose to conduct the experiments using the measured 1 sccm flow rate.It is difficult to quantify how the higher HO 2 -to-OH conversion efficiency measured during MCMA-2006 would translate into the various RO 2 -to-HO 2 conversion efficiencies, although it is likely that many of them would be larger.As a result, the conversion efficiencies measured in this study for the MCMA-2006 configuration may represent a lower limit to the actual conversion efficiencies during the campaign.
The RO 2 conversion efficiency into HO 2 (f RO 2 →HO 2 ) measured for the inlet conditions for the MCMA 2006, CAB-INEX, and CalNex campaigns are summarized in Table 2 and represent the results of several experiments similar to those illustrated in Fig. 3, with the uncertainty representing 1 standard error of the mean of the measurements.The largest RO 2 interference was observed for the CalNex inlet conditions where alkenes produced interferences ranging from 83 ± 7 % for isoprene-based peroxy radicals to 96 ± 6 % for tetramethylethylene (TME)-based peroxy radicals while the conversion efficiency of aromatic, aldehydes, and ketone compounds ranged from 54 ± 4 % for methacrolein (MACR) to 91 ± 4 % for methyl vinyl ketone (MVK).The RO 2 -to-HO 2 conversion efficiency of a number of alkanes ranged from an average measured value of 15 ± 3 % for propane-based peroxy radicals to 62 ± 4 % for octane-based peroxy radicals, with the RO 2 -to-HO 2 conversion efficiency increasing with the carbon number.The inlet configuration and conditions used during the MCMA 2006 campaign generally resulted in lower RO 2 interferences likely due to the higher flow rate (and shorter reaction time) in the detection cell and the lower NO concentration (Table 1), although the measured conversion efficiency was found to be somewhat greater for some VOCs.Under these inlet conditions the RO 2 -to-HO 2 conversion efficiency for propane-based peroxy radicals was measured to be 22 ± 11 % while the conversion efficiency for octane-based peroxy radicals was 30 ± 5 %.Because the CABINEX campaign occurred in a remote forested environment, measurements of the RO 2 -to-HO 2 conversion efficiency focused on characterizing interferences from peroxy radicals produced from isoprene and its oxidation products (MVK and MACR), as isoprene peroxy radicals were predicted to contribute to more than 80 % of the total RO 2 concentration during the campaign (Griffith et al., 2013).The inlet and instrumental configuration during CABINEX resulted in a higher pressure and slower sampling rate compared to the MCMA-2006 configuration.For this instrumental configuration, the RO 2 -to-HO 2 conversion efficiency was found to be 91 ± 5 % for isoprene-based peroxy radicals, while the conversion efficiencies for MVK and MACR were found to be 62 ± 5 and 30 ± 7 %, respectively.
These observations are consistent with results reported for other FAGE instruments (Fuchs et al., 2011;Whalley et al., 2011) and assume that the photolysis of each VOC does not contribute to the production of radicals in these experiment.However, tests to determine whether photolysis of the various VOCs resulted in the formation of HO x radicals in the absence of water vapor revealed that the photolysis of methyl vinyl ketone (MVK), methacrolein (MACR), methyl ethyl ketone (MEK), and toluene can lead to the production of HO x radicals.The radical signals from the photolysis of methacrolein and toluene were small and negligible relative to the total HO x signal produced from the photolysis of water.However, the signal from the photolysis of MVK and MEK during these tests was significant and could interfere with the measurements of the RO 2 -to-HO 2 conversion efficiency.These results are in contrast to that reported by Fuchs et al. (2011), who found that the photolysis of VOCs during similar tests in dry air did not produce any radicals.The reason for this discrepancy is unclear but may be related to differences in the UV flux produced by the different mercury lamps or impurities associated with the VOC samples (Tables S1 and S2).Addition of water vapor may reduce the HO x radical production from photolysis of these VOCs due to quenching of the excited VOC, and as a result it is difficult to quantify the interference in these experiments.However, any interference from HO x radicals produced from the photolysis of MVK and MEK would result in higher apparent conversion efficiencies, as they could represent an additional source of HO x radicals when the VOCs are added, and could contribute to the higher RO 2 -to-HO 2 conversion efficiency reported here for MVK compared to that reported by Fuchs et al. (2011).
As previously observed, the RO 2 -to-HO 2 conversion efficiency of alkene-based β-hydroxyalkyl peroxy radicals was found to be greater than the conversion efficiency of alkanebased alkyl peroxy radicals (Fuchs et al., 2011).As discussed above, this is due to due to the ability of the β-hydroxyalkoxy radicals produced from the RO 2 + NO reaction to rapidly decompose to form a hydroxyalkyl radical.The hydroxyalkyl radical reacts rapidly with O 2 in the FAGE detection cell, leading to the production of a carbonyl compound and HO 2 .However, the ability of large alkoxy radicals to rapidly isomerize and decompose (Atkinson, 1997;Finlayson-Pitts and Pitts Jr., 2000) also results in a rapid production of HO 2 radicals and a larger conversion efficiency.
In general, reducing the reaction time in the IU-FAGE instrument reduces the conversion of these peroxy radicals to HO 2 , as illustrated by the reduced conversion efficiencies between the CalNex and MCMA operating conditions for the majority of the VOCs tested.However, the measured conversion efficiencies of some of the tested VOCs did not always display this behavior, and the reasons for the discrep-ancies are unclear.For example, the conversion efficiency for ethene peroxy radicals was lower for the CalNex configuration compared to the CABINEX and MCMA configurations even though the overall flow rate was slower for the Cal-Nex configuration.However, the HO 2 -to-OH conversion efficiency was also lower for this inlet configuration, suggesting that reaction time may not be the only factor limiting the conversion efficiency under these instrument conditions.Similarly, the conversion efficiency of MVK and MACR measured for the CABINEX instrument configuration was lower than that measured for the MCMA inlet configuration, even though the overall slower flow rate in the CABINEX configuration leads to a longer reaction time in the IU-FAGE detection cell.This may suggest that the chemistry of peroxy radicals produced from the OH-initiated oxidation of MVK and MACR is different than that of the peroxy radicals produced from the OH-initiated oxidation of alkenes and alkanes, with competing channels that result in a more complex dependence on reaction time.As discussed in Fuchs et al. (2011), the fates of the peroxy radicals produced by the OH-initiated oxidation of MVK and MACR are not well known.For the peroxy radicals produced from the OH-initiated oxidation of MVK, three decomposition channels are possible, with two channels potentially leading to fast HO 2 production and one likely leading to slower HO 2 production (Fuchs et al., 2011).For the peroxy radicals produced from the OH-initiated oxidation of MACR, the channel that leads to the formation of a hydroxyalkyl peroxy radical likely leads to fast HO 2 production, while the channel that leads to the formation of an acyl peroxy radical would lead to much slower HO 2 production (Fuchs et al., 2011).

RO 2 radical concentrations during MCMA-2006
The previous analysis of the HO 2 radical concentrations during the Mexico City Metropolitan Area 2006 campaign did not take into account interferences from RO 2 radicals (Dusanter et al., 2009b).As discussed above, the instrumental conditions during MCMA-2006 resulted in the conversion of a fraction of RO 2 radicals to HO 2 , resulting in the measurements reflecting HO 2 * = HO 2 + αRO 2 and overestimating the actual HO 2 concentrations.To determine the fraction (α) of RO 2 radicals likely detected during the HO 2 measurements, the RO 2 radical concentrations during MCMA-2006 that were previously modeled using the Regional Atmospheric Chemistry Mechanism (RACM) were used to calculate the modeled HO 2 * concentrations (Dusanter et al., 2009b).
As discussed in Dusanter et al. (2009b), the RACM model is a condensed chemical mechanism that describes the gasphase oxidation of 17 inorganic and 32 organic species.Kinetic parameters for the reactions of OH, O 3 , and NO 3 with inorganic species and for reactions involving organic species treated explicitly in RACM (methane, ethane, ethene, formaldehyde, glyoxal, methyl peroxide, and isoprene) were updated using the NASA Jet Propulsion Laboratory database (Sanders et al., 2006).Rate constants and branching ratios for OH, O 3 , and NO 3 reactions with surrogate species were used as described in the RACM model (Stockwell et al., 1997).Heterogeneous chemistry, such as the incorporation of trace gases into aerosols, was not included.
The total average modeled RO 2 concentration from 09:00 to 18:00 consisted of 54 % alkane-based, 27 % alkene-based, and 14 % aromatic-based peroxy radicals (Fig. 4).On average, the modeled composition of peroxy radicals was relatively constant throughout the day during the MCMA campaign.The modeled relative contribution of aromatic-based peroxy radicals was greater in the morning, which is consistent with the observed elevated concentrations of benzene and toluene during the morning hours (Dusanter et al., 2009b).

Implications of RO 2 interferences for HO 2 measurements during MCMA-2006
The modeled diurnal average concentrations of total RO 2 radicals during MCMA is shown in Fig. 5, along with the modeled HO 2 concentrations and the measured HO 2 * concentrations.As discussed in Dusanter et al. (2009b), the modeled HO 2 concentrations were in good agreement with the measurements during the afternoon but the model underestimated the measured HO 2 concentrations during the morning hours by a factor of approximately 2 to 5.However, these conclusions were based on the assumption that the measured HO 2 concentrations were free from interferences and could be compared to the modeled HO 2 concentrations.Based on the conversion efficiencies reported for RO 2 radicals in the present study, it is clear that the MCMA measurements represent an upper limit to the actual HO 2 concentrations and should be compared to the modeled HO 2 * = HO 2 + αRO 2 concentrations.
The RACM modeled HO 2 * concentrations were calculated by applying the measured RO 2 -to-HO 2 conversion efficiencies for the instrumental conditions reported in Table 2 for MCMA-2006 using Eq. ( 12): The contribution for isoprene peroxy radicals (ISOP), ethene peroxy radicals (ETEP), and toluene peroxy radicals (TOLP) were taken directly from Table 2.The average RO 2 -to-HO 2 conversion efficiency for trans-2-butene and tetramethylethylene-based peroxy radicals was used for the conversion efficiency of peroxy radicals from internal olefins (OLIP) and external olefins (OLTP), while the conversion efficiency for trans-2-butene was used for the conversion efficiency for α-pinene and other cyclic terpene peroxy radicals (APIP).The measured conversion efficiency for toluene-based peroxy radicals was used to represent the conversion efficiency for xylene (XYLP) and cresol (CSLP) peroxy radicals.The conversion efficiency of methacroleinbased peroxy radicals was used to represent the conversion  efficiency of acetyl peroxy and higher saturated acyl peroxy radicals (ACO3) as well as unsaturated acyl peroxy radicals (TCO3), while the conversion efficiency of methyl vinyl ketone-based peroxy radicals was used to represent the efficiency of ketone-based peroxy radicals (KETP).
The overall average contribution of peroxy radicals to the modeled HO 2 * and the relative contribution of each RACM peroxy radical category to the RO 2 interference are shown in Fig. 6.Because the HO 2 -to-OH conversion efficiency of 80 % in these experiments was approximately 20 % lower than the conversion efficiency measured during the campaign (Dusanter et al., 2008(Dusanter et al., , 2009a)), the relative peroxy radical contributions illustrated in this figure are likely lower limits to the actual contribution during the campaign.
On average, RO 2 radicals contributed to approximately 35 % of the total modeled HO * 2 (Fig. 6).While alkanes compose the majority of the modeled peroxy radicals (Fig. 4), they only contributed to about 29 % of the RO 2 interference, whereas alkenes contributed to approximately 51 % of the interference (Fig. 6).While isoprene peroxy radicals contributed to only 5 % of the total RO 2 concentration, they contributed approximately 11 % to the interference.The overall contribution of RO 2 radicals to the measured HO 2 * concentrations in this environment is similar to that observed during the CalNex campaign, where RO 2 radicals were modeled to contribute to approximately 30 % of the measured HO 2 * concentrations, although during CalNex peroxy radicals from isoprene (ISOP) and its oxidation products accounted for approximately 40-50 % of the modeled interference and olefins (OLTP, OLIP) contributed approximately 20-30 % (Griffith et al., 2016).Unlike these urban environments, in forested environments where the OH reactivity is dominated by isoprene and other unsaturated biogenic emissions, isoprene and other biogenic hydroxyl alkyl peroxy radicals can be the dominant peroxy radicals and can make a significant contribution to the measured HO 2 * concentrations due to their high conversion efficiency to HO 2 in the FAGE detection cell (Table 2).For example, during the CABINEX campaign in a northern Michigan forest, isoprene peroxy radicals were modeled to be the dominant peroxy radical in this environment and the main contributor to the interference, contributing to approximately 50 % of the modeled HO 2 * concentrations during the daytime (Griffith et al., 2013).As a result, previous measurements of HO 2 in these environments by LIF-FAGE or other chemical conversion techniques are likely influenced by an interference from β-hydroxyalkyl peroxy radicals such as those produced by the OH-initiated oxidation of isoprene and other biogenic emissions.While it is possible to subtract the modeled speciated RO 2 concentrations from the measured HO 2 * and compare the results to the modeled HO 2 , this method increases the uncertainty associated with the measured concentrations due to the additional uncertainty associated with the modeled RO 2 concentrations as well as the uncertainties associated with the measured RO 2 -to-HO 2 conversion efficiencies.We estimate the uncertainty associated with the RACM mod-eled RO 2 to be approximately ± 70 % (2σ ), which is similar to that for the modeled HO 2 (Dusanter et al., 2009b).As a result, we prefer to compare the modeled HO 2 * to the measured HO 2 *.
The diurnal average modeled HO 2 * concentrations for the MCMA 2006 campaign are also shown in Fig. 5.As can be seen in this figure, the model overestimates the measured HO 2 * by approximately 35 % between 12:00 and 17:00 CST, although the modeled results are generally close to the upper bound of the calibration accuracy (36 %, 2σ ) (Dusanter et al., 2009b).As discussed above, the modeled HO 2 * is likely a lower limit given that the RO 2 -to-HO 2 conversion efficiencies during the campaign may be greater than shown in Table 2 due to the higher HO 2 -to-OH conversion efficiency measured during the campaign.Although the measured HO 2 * is still likely to be within the overall uncertainty of the model, which was estimated to be approximately a factor of 1.7 (Dusanter et al., 2009b), these results suggest that the model likely overestimates the measured concentrations during the afternoon.
These results are in contrast to the results from the CalNex campaign, where the simulations using the RACM2 model tended to underestimate the measured HO 2 * concentrations during the week, when NO mixing ratios were greater than 4 ppb (Griffith et al., 2016).The reason for this difference between the campaigns is unclear but may be related to the relative concentrations of dicarbonyl species and their treatment in the RACM and RACM2 models.Dusanter et al. (2009b) demonstrated that the RACM model results for MCMA-2006 were highly sensitive to the concentrations of dicarbonyl species in the model, with the model significantly overpredicting the concentration of HO x radicals when unmeasured concentrations of these species were not constrained.Daytime average measured glyoxal mixing ratios during MCMA- 2006 were approximately 0.4 ppb (Dusanter et al., 2009b), which were greater than the maximum daytime mixing ratios of 0.16 ppb during CalNex (Washenfelder et al., 2011), suggesting that the MCMA-2006 results may be more sensitive to the treatment of dicarbonyl chemistry compared to CalNex.Additional analysis and modeling will be needed to resolve this issue.
While the model tends to overestimate the measured HO 2 * concentrations during the afternoon, it underestimates the HO 2 * concentrations in the morning by a factor of 3 between 09:00 and 11:00.As discussed in Dusanter et al. (2009b), this may suggest that a significant radical source may be missing from current atmospheric models under polluted conditions.Similar results were observed in Wangdu, China, by Tan et al. (2017).In this study, total peroxy radical concentrations were measured by chemical conversion to HO 2 in an external reactor with subsequent detection of HO 2 (after chemical conversion to OH) in an LIF-FAGE instrument.They also measured HO 2 radicals using a second LIF-FAGE detection axis that minimized interferences from RO 2 radicals.They found that a model using the updated RACM2 mechanism was able to reproduce the observed HO 2 concentrations during the day but underestimated the observed total RO 2 concentration by a factor of 3 to 5 in the morning when NO concentrations were higher than 1 ppbv.The observed RO 2 concentrations could be explained by a missing RO 2 source of 2 ppbv h −1 (Tan et al., 2017).Dusanter et al. (2009b) also compared the measured HO 2 * / OH ratio to the RACM modeled HO 2 / OH ratio and found that the model underpredicted the observed ratio, especially under conditions where the mixing ratio of NO was greater than 5 ppb.At NO mixing ratios of 10 ppb, the model underestimated the measured ratio by a factor of 2 (Dusanter et al., 2009b).However, comparing the measured HO 2 * / OH ratio to the modeled HO 2 * / OH ratio improves the agreement even though the model tends to overpredict both OH and HO 2 * in the afternoon (Fig. 7).This may indicate that there is either a missing sink of HO x radicals in the model or a miscalculation of the relative rates of initiation and/or termination.At an NO mixing ratio of 10 ppb the modeled HO 2 * / OH ratio is in good agreement with the measurements, although it still underestimates the measured HO 2 * / OH ratio at higher NO mixing ratios by as much as a factor of 4, and may also overestimate the HO 2 * / OH ratio for mixing ratios of NO less than 5 ppb by as much as a factor of 2 (Fig. 7).It is interesting to note that a model underestimation of the total OH reactivity at high NO mixing ratios may contribute to this discrepancy.Unfortunately, total OH reactivity was not measured during MCMA-2006 and the reliability of the model to simulate it could not be assessed.Similar results were observed for the CalNex campaign (Griffith et al., 2016), which included direct measurements of the total OH reactivity.Although accounting for the missing reactivity in the analysis of the CalNex data improved the agreement between the measured and modeled HO 2 * / OH ratio, the model still underestimates the measured ratio at high mixing ratios of NO (Griffith et al., 2016).These results suggest that our understanding of the radical propagation chemistry under high NO conditions may be incomplete.

Summary and conclusions
The RO 2 interference associated with measurements of HO 2 by the IU-FAGE instrument was characterized for three different instrument configurations that were used in previous field campaigns (MCMA 2006, CABINEX 2009, and Cal-Nex 2010).Similar to that reported for other LIF-FAGE instruments, the RO 2 -to-HO 2 conversion efficiency was highest for alkene-and aromatic-based RO 2 radicals, producing higher levels of interference, while the conversion efficiency of alkane-based RO 2 radicals was less but increased with increasing carbon number.In general, the conversion efficiency was higher for instrument configurations that involved slower sampling flow rates and longer reaction times between the peroxy radicals and NO in the detection cell.
The similarities in the measured RO 2 conversion efficiencies reported here with those reported for other LIF-FAGE instruments suggest that the main factor controlling the conversion efficiency is the rate of reaction of RO 2 radicals with NO and that increasing the efficiency of the conversion of HO 2 to OH will also increase the RO 2 -to-HO 2 conversion efficiency.Although the impact of differences in the characteristics of the low-pressure expansion in LIF-FAGE instruments cannot be ruled out, these results suggest that the interferences reported here associated with measurements of HO 2 are likely similar for all instruments that measure HO 2 by chemical conversion through reaction with NO.Previwww.atmos-meas-tech.net/11/95/2018/Atmos.Meas.Tech., 11, 95-109, 2018 ous measurements of HO 2 radicals by instruments using this method were likely influenced by the conversion of RO 2 radicals, with measurements of HO 2 in forested environments likely influenced by interferences from peroxy radicals derived from biogenic alkenes such as isoprene due to the high RO 2 -to-HO 2 conversion efficiencies of these radicals.Because of the lower conversion efficiencies of alkane-based peroxy radicals, the impact on previous measurements in urban areas will depend on the relative concentrations of alkanes versus alkenes and aromatics contributing to the overall pool of peroxy radicals in these environments.While this interference was taken into account to investigate the radical chemistry during CABINEX (Griffith et al., 2013) and CalNex (Griffith et al., 2016), this issue was not known when the radical measurements from the MCMA 2006 field campaign were published (Dusanter et al., 2009b).An analysis of the impact of this interference on the results for the MCMA 2006 campaign suggests that the RO 2 radical contribution to the measured HO 2 * concentration was approximately 35 % based on the RACM modeled RO 2 concentrations.Taking this interference into account, the resulting modeled HO 2 * concentrations were generally greater than the measured concentrations by 35 % during the afternoon, although the model results were within the calibration uncertainty of the measurements (36 % at 2σ ).Given that the modeled HO 2 * concentrations likely reflect a lower limit to the interference during the campaign, these results suggest that the model likely overestimates the measured concentrations during the afternoon.However, the model still underestimates the HO 2 * concentration by a factor of 3 in the morning, suggesting that the model may be missing an important radical source in the morning.Although the measured HO 2 * / OH ratio was in better agreement with the modeled HO 2 * / OH ratio compared to the modeled HO 2 / OH ratio, the model still significantly underestimates the HO 2 * / OH ratio by up to a factor of 4 for NO mixing ratios greater than 10 ppb, suggesting that our understanding of radical propagation under these conditions is still incomplete.
Future measurements of peroxy radicals by the IU-FAGE instrument will involve measurements at lower NO concentrations to minimize the RO 2 -to-HO 2 conversion efficiency.Recent experiments have demonstrated that the addition of an NO concentration of approximately 9 × 10 11 cm −3 results in an HO 2 -to-OH conversion efficiency of approximately 17 % and a conversion efficiency of isoprene-based peroxy radicals to HO 2 of approximately 10 %.Even at this low HO 2 -to-OH conversion efficiency, the resulting HO 2 signals are still significantly greater than the limit of detection of the instrument, but at this low NO concentration the subsequent conversion of isoprene peroxy radicals to OH is negligible, allowing for measurements of ambient HO 2 concentrations without interferences from RO 2 radicals.

Figure 1 .
Figure 1.Indiana University LIF-FAGE cross section (a) and a schematic of the excitation axis and the detection axis (b; Dusanter et al., 2008).

Figure 2 .
Figure 2. Cross section of Indiana University calibration source for the IU-FAGE instrument.

Figure 3 .
Figure 3. RO 2 interference measurement experiment for isoprene (a; with an OH reactivity of approximately 290 s −1 ) and butane (b; with an OH reactivity of approximately 30 s −1 ).The boxed numbers within the figure represent the various experimental modes: (1) S OH , (2) S HO x with internal NO addition, (3) S OH + VOC with VOC added, and (4) S RO x with VOC added and internal NO addition.

Figure 5 .
Figure 5.Diurnal average HO 2 * measurements from MCMA-2006.The grey solid circles are 30 s averages and solid blue square symbols are binned 1 h averages.The solid black line represents the RACM modeled HO 2 , the solid red line represents the modeled HO 2 *, and the dotted black line represents the total modeled RO 2 + HO 2 .The error bars reflect the calibration accuracy of the measurements (±36 %, 2σ ).

Figure 6 .
Figure 6.Modeled speciation of the RO 2 interference for MCMA-2006.The pie chart on the left is the modeled HO 2 * composition after adding the fraction of RO 2 interference to the modeled HO 2 .The pie chart on the right is the composition of the RO 2 interference.

Figure 7 .
Figure 7. Correlation plot for HO 2 * / OH vs. NO.Small grey circles are individual measurements recorded for the whole campaign.Large blue circles are average values calculated on binned NO data and the blue line is a fit to the average measurements.The modelcalculated HO 2 / OH ratio is displayed by the green line for the campaign averaged measurements, while the red line represents the modeled HO 2 * / OH ratio.Dashed lines are the 95 % confidence interval from the nonlinear power regressions.

Table 1 .
Configuration of the IU-FAGE instrument during various previous field campaigns.

Table 2 .
Average measured f RO 2 →HO 2 for various alkenes and alkanes under different inlet conditions.Uncertainties represents the standard error of the mean from all individual experiments, with the number of experiments shown in parentheses.