In the atmosphere, HO2 and OH radicals (OH+HO2=HOx) are closely coupled via several reactions as shown in Scheme 1. The short lifetimes of HOx radicals mean that concentrations are determined by chemical production and removal and not by transport processes, making them ideal candidates as test species for our understanding of atmospheric chemical mechanisms (Stone et al., 2012; Monks, 2005; Stockwell et al., 2012).

In Scheme 1, the reaction of alkoxy radicals with molecular oxygen is a major route to HO2 formation; however, this is not the only significant HO2 formation process; for example, in the atmospheric oxidation of n-butanol, HO2 can be formed via two different mechanisms. Abstraction by OH at the α position leads to a radical which reacts with oxygen to directly produce HO2 (Reactions R1a, R2) whereas abstraction at other sites leads to alkylperoxy radical (C4H9O2) formation with varying fractions of the RO2 forming alkoxy radicals, and subsequently HO2 (McGillen et al., 2013) on a longer timescale; see Scheme 2. R1aOH+CH3CH2CH2CH2OH→CH3CH2CH2CHOH+H2OR2CH3CH2CH2CHOH+O2→HO2+CH3CH2CH2CHO The fraction of alkoxy radicals formed depends on the mechanism of RO2 removal (reaction with NO or self- or cross-reactions) and the yield of HO2 from the alkoxy radical depends on the competition between decomposition, isomerization, and reaction with O2, which in turn will depend on the structure of the alkoxy radical, temperature, pressure, and concentration of oxygen (Orlando et al., 2003). Therefore, in order to determine the HO2 yield from the OH-initiated oxidation of compounds such as n-butanol, it is important to have a selective, sensitive, and time-resolved method of HO2 detection.

The importance of HO2 chemistry is not limited to atmospheric processes; HO2 is a key intermediate in low-temperature (500–1000 K) combustion processes, particularly those involving oxygenated fuels (Zador et al., 2011). The mechanisms of low-temperature combustion are of particular interest in the development of new engine technologies such as reactively controlled compression ignition (RCCI) (Reitz and Duraisamy, 2015) and are closely linked to atmospheric oxidation mechanisms. Monitoring HO2 concentrations under the elevated temperatures and high pressures of combustion processes is therefore also of interest. In low-temperature combustion, HO2 formation is a chain-inhibition process, with OH reformation a chain-propagating or chain-branching process. The ratio of chain-branching to chain-inhibition processes is often the controlling factor in modelling ignition delay times (Agbro et al., 2017). High temperatures and concentrations of oxygen may be required to convert atmospheric processes, which take several tens of seconds at ambient temperatures (and hence may be influenced by surface chemistry or secondary reactions) to the milli- or microsecond timescale where they can be studied by flash photolysis techniques without such interferences (Medeiros et al., 2018).

Direct measurements of HO2 rely on absorption techniques, and kinetic information on HO2 reactions has been determined mainly using absorption spectroscopy. This can be achieved with either conventional absorption techniques, often in the UV (including multipass optics to enhance the pathlength), or in the IR with cavity ring-down spectroscopy (CRDS) (Assaf et al., 2018; Onel et al., 2017). However, the HO2 UV absorption spectrum (200–260 nm) is broad and featureless (Crowley et al., 1991), and as such overlaps with the UV absorptions of many other species present in atmospheric degradation pathways or combustion systems (particularly H2O2 and RO2). To utilize the selectivity of the structured IR spectra, absorption methods have been developed in both the mid- and near-IR (NIR) (Taatjes and Oh, 1997). Mid-IR absorption features for HO2 provide sufficient absorption cross sections for study (Jemialade and Thrush, 1990) but suffer from severe pressure broadening, reducing sensitivity under the conditions relevant to atmospheric and combustion systems (Thiebaud and Fittschen, 2006). Detection in the NIR has similar advantages in terms of a structured spectrum providing greater selectivity; the weaker absorption cross sections are compensated for by the higher powers and ease of use of NIR laser sources (Gianella et al., 2016). However, pressure broadening and interference from H2O absorptions can make these measurements difficult at even low concentrations of water (1014 molec. cm-3).

In the atmosphere (Stone et al., 2012) and in chamber studies (Glowacki et al., 2007), HO2 is detected using a sensitive but indirect method via conversion to OH, with detection of OH via laser-induced fluorescence (LIF) (Hard et al., 1984; Brune et al., 1995; Fuchs et al., 2011) or conversion to H234SO4 with subsequent detection of the acid via mass spectrometry (Edwards et al., 2003; Hanke et al., 2002). In the LIF method, also known as fluorescence assay by gaseous expansion (FAGE; Hard et al., 1984), which is the technique used in this study, OH is sampled into a low-pressure region through a pinhole. Low pressures allow for the temporal separation of resonant 308 nm fluorescence from the excitation pulse. Following the first detection axis for OH, a flow of NO is introduced which reacts with HO2 (Reaction R3): R3HO2+NO→OH+NO2. The resulting OH is monitored at a second detector. The high sensitivity with which OH can be detected gives HO2 detection limits in the 108 molec. cm-3 range for 5–10 s averaging; however, to extract concentrations, both OH detection methods require calibration (Winiberg et al., 2015). For chamber measurements of HO2, comparisons with direct CRDS measurements have verified the reliability of the calibration process (Onel et al., 2017).

HO2 detection by LIF can be potentially sensitive to interferences from certain RO2 species, which may also be converted to OH on short timescales. Interferences can be minimized by short conversion times between NO injection and OH monitoring, utilizing low pressures, high flow rates of the sample gas, and low NO concentrations to separate OH generation from HO2 and RO2; reduced conversion of HO2 reduces the sensitivity of this technique and as such in practice a compromise between selectivity and sensitivity is used (Fuchs et al., 2011; Hard et al., 1984; Whalley et al., 2013).

The current paper describes a significant development on our earlier FAGE-based instrument for time-resolved OH detection (Stone et al., 2016). In this improved system, laser flash photolysis in a high-pressure (up to 5 bar), temperature-controllable (300–800 K) reactor (shown in Fig. 1), generated radicals which were then sampled through a pinhole forming a jet within the low-pressure detection region (shown in more detail in Fig. 2). OH radicals were monitored by LIF close to the pinhole. The jet breaks down after ∼20 mm and NO was injected after this point to convert some HO2 into OH, which was then detected by a second monitoring system. In general, LIF becomes less sensitive at higher temperatures (due to distribution of population over more rotational levels) and O2 concentrations (due to quenching). Sampling into the low-pressure region reduces both the effect of collisional quenching and temperature on the sensitivity of LIF detection, although there is a reduction in the number density of the HOx species in the expansion. We report the adaptation of our time-resolved OH FAGE instrument to allow HO2 detection, the characterization of the instrument (including development of a calibration method for HO2 yields of OH initiated reactions), and the investigation of the influence of RO2 species. Finally, we discuss the application of the technique to determine the yield of HO2 from the reaction of OH with n-butanol. The instrument has some similarities to that presented by Nehr et al. (2011) where a conventional OH lifetime instrument was altered to allow for chemical conversion of HO2 to OH and hence the sequential determination of OH and HO2.

Reactions were carried out in a high-pressure (0.5–5 bar) reaction cell, which is described in greater detail in Stone et al. (2016) and schematics of which are shown in Figs. 1 and 2. The high-pressure reactor was a 0.5 m stainless-steel tube with a 22 mm internal diameter. Gas flows were delivered to the high-pressure cell from a mixing manifold where calibrated mass flow controllers (MFCs) allowed for accurate control of flow rates. Low-vapour-pressure compounds, OH precursors (H2O2), and substrates methanol and butanol, were delivered to the mixing manifold from thermostatted bubblers in pressure-regulated backing flows of nitrogen (N2). Ethane and oxygen were delivered directly from cylinders into the mixing manifold through MFCs. The gas flow rate through the cell was kept under laminar conditions with typical Reynolds values (Re) of 480 (corresponding to a flow rate for an experiment of 10 SLM (standard litres per minute) at 2 bar); in general conditions were maintained between 400 and 800 Re (Re <2400= laminar flow), with some experiments carried out with higher flow rates, up to 1800 Re.

Temperature control of the reactor between room temperature and 800 K was achieved by altering the voltage applied to a coil heater (WATROD tubular heater, Watlow) over the last 30 cm of the stainless-steel tube. The heated region was fitted with a quartz liner (inner diameter 18 mm) to reduce wall-initiated chemistry. A temperature readout, from a type-K thermocouple in the gas flow, close to the pinhole, was calibrated for given flow rates, pressures, and voltage settings by measuring the highly temperature-sensitive OH and methane rate coefficient, using the temperature dependence reported by Dunlop and Tully (1993). A more detailed description of this method is described within instrument characterization (Sect. 3.4).

The photolysis of the OH precursor, H2O2, at 248 nm (Lambda Physik, COMPex 200 operated using KrF at 1 or 5 Hz) or 266 nm (frequency quadrupled Nd:YAG output, Quantel, Q-smart 850 at 1 or 10 Hz) initiated the chemistry. No significant difference was noted in the kinetics or yields as a function of laser repetition rate. R4H2O2+hν→2OH Hydrogen peroxide was used as the OH precursor for all experiments where HO2 detection was performed because it also acts as an internal calibrator to relate OH and HO2 (highlighted in Reaction R5) in a 1:1 relationship, via Reaction (R5): R5OH+H2O2→H2O+HO2. However, in general, other OH precursors can be used. The OH precursor was maintained at low concentrations (1×1014–1×1015 molec. cm-3) to minimize errors associated with assigning pseudo-1st-order kinetics for the loss of OH and to reduce radical–radical reactions. Maintaining a low radical precursor concentration had the additional advantage of minimizing attenuation of the photolysis beam, ensuring consistency in the initial radical concentrations generated along the length of the high-pressure cell. Initial OH concentrations were in the range 2×1011–5×1013 molec. cm-3.

A pinhole (diameter <0.15 mm) at the end of the high-pressure reactor couples the reactor to the low-pressure (0.3–5 Torr) detection cell. Details on OH detection can be found in Stone et al. (2016). The accuracy of the instrument for OH measurement has recently been verified by measurements of the rate coefficient of the reaction of OH with isoprene (Medeiros et al., 2018), which are in excellent agreement with the literature. A more detailed schematic for the low-pressure detection cell is shown in Fig. 2.

In the first low-pressure detection cell, the OH was probed within the jet-expanded gas, close to pinhole (<5 mm), perpendicular to the gas flow. The OH was detected by off-resonance laser-induced fluorescence (LIF) at 308 nm following excitation with 282 nm light (A2Σ(v′=1)←X2Π(v′′=0),Q11). The 282 nm light was the frequency-doubled output of a dye laser ((Rhodamine 6 G, Spectron) pumped at 532 nm by a Nd:YAG laser (Spectron), or (Rhodamine 6 G, Continuum) pumped by a Nd:YAG laser (Quantel, Q-smart 850)). Measuring the off-resonance fluorescence allowed the use of a filter (308±5 nm, Barr Associates) before the photomultiplier (PerkinElmer C1943P) to remove scattered light and improved the signal-to-noise ratio.

A delay generator (BNC DG535) was used to vary the delay (time resolution ∼10 ns) between the photolysis and probe laser, facilitating generation of time profiles of the OH concentration. The traces, typically 200–300 data points and ranging in time from ∼50 µs to 20 ms, were scanned through multiple times (5–20) and the signal at each time point was averaged, giving high-precision OH loss traces. An example OH trace from the first detection cell for Reaction (R5) is presented in Fig. 3. As reactions were carried out under pseudo-1st-order conditions ([OH]≪[substrate]), the time dependence of the OH LIF signal, If (proportional to the [OH]), was given by If,t=If,0e-kOHt, where kOH=k5[H2O2]. In Fig. 3 two traces are presented from the first, OH, detection axis, these two traces were taken in consecutive experiments with a constant [H2O2] where the first trace (kOH,1st=(2351±22) s-1) was taken where N2 was flowed into the low-pressure region; the second trace (kOH,1st=(2389±18) s-1) was taken when this flow had been switched to NO to allow HO2 detection in the second detection cell, and errors are given as 2σ. The similarity of the OH decay traces when either N2 or NO was injected shows that there was no back-streaming of NO in the low-pressure cell and hence no HO2 conversion at the first detection axis.

HO2 radicals were monitored by the chemical transformation of HO2 to OH via reaction with NO (Reaction R3) in the low-pressure cell. Following the breakdown of the jet, after the Mach disk (>2 cm beyond the pinhole), a small flow (5 sccm) of NO or N2 was introduced into the low-pressure cell via a 1.5 mm i.d. stainless-steel pipe (for a typical 0.5 Torr pressure in the FAGE cell and the NO concentration was 5.5×1013 molec. cm-3). After passing through the first detection cell, the probe beam was redirected through the second low-pressure detection cell downstream of the NO pipe, allowing for the measurement of the OH concentration by LIF in the same manner as in the first cell.

By switching between a flow of N2 and NO, through the pipe, traces for OH loss and HO2 formation could be elucidated, examples of which are shown in Fig. 4. Subtraction of the two OH traces in Fig. 4 (upper red trace is with NO injection and the signal corresponds to reactant OH and OH produced from the titration of HO2 to OH; lower blue trace with N2 injection is reactant OH only) gave a resultant signal associated with HO2 production in the high-pressure reactor, shown as the pink trace in Fig. 4. The signal from the first PMT allowed for correction of the signal heights at the second PMT for changes in the probe laser power or wavelength, as any variations in laser power or wavelength affect the absolute signal retrieved from both PMTs; however, the relative signals retrieved from the PMTs remain consistent.

Fits to the HO2 formation traces and OH loss traces from the second cell generated kinetic parameters which differed from the accurate parameters collected at the first detection axis, kOH,2nd=(1390±44) and kHO2,2nd=(1080±150) s-1, where the accurate loss parameters from the first cell were kOH,1st=(2389±18) s-1, 2σ errors. This difference was the result of transport effects. By comparison of the loss and formation parameters derived for OH+H2O2, for the first and second detection cells, HO2 formation rates could be assigned from a calibration plot (Fig. 7).

Neither of the OH determinations in the two detection axes provide absolute measurements of radical concentrations. Each detection axis could be calibrated as for chamber measurements, but for our purposes a calibration reaction linking photolytically produced OH and HO2 removes many sources of error compared to an absolute calibration. The reaction of OH with the radical precursor H2O2 which directly forms HO2 with a 100 % yield was used for calibration. R5OH+H2O2→H2O+HO2 For reactions carried out where a reagent was added in addition to the H2O2, the resulting ratios can be compared with those from the calibration reaction to allow assignment of an observed HO2 yield. To assign the HO2 yield from the test reaction required accounting for secondary HO2 production in the high-pressure reactor, from OH+H2O2 and photolysis processes. From the known rate coefficients, it was possible to calculate the fraction of OH reacting with the H2O2 (typically 5 %–10 %) and hence the expected contribution to the HO2 signal. Photolytic production of HO2 was accounted for by measuring the observed HO2 signal in the absence of any H2O2.

In a typical experiment, the reaction of OH and H2O2 would be carried out four times, twice in the absence of NO and twice with the addition of NO to calibrate the instrument. Exponential fits to the OH decay as monitored in the first cell determine the peak OH signal. The OH signals at the second detector were recorded with only N2 addition (reagent OH reaching the second detector) and subtracted from the signal with NO added (reagent OH and HO2) to give the net HO2 signal. This profile was fitted to a bi-exponential growth and decay function to extract the peak HO2 signal for that set of conditions. Combinations of traces were then used to obtain an averaged value (and uncertainty) of the signal on the first PMT (OH) to the net HO2 signal at the second PMT for this calibration reaction where OH reactant and HO2 product have a 1:1 relationship. The same process was then performed in the presence of the compound of study. The removal pseudo-1st-order rate coefficient with H2O2 and the reagent of study (k′OH,1st=kOH+H2O2[H2O2]+kOH+TEST[TEST]) were compared to the removal pseudo-1st-order rate coefficient with only H2O2 (k′OH,1st=kOH+H2O2[H2O2]) to assign what fraction of the OH reacted with the H2O2 precursor and hence the resulting contribution to the observed HO2. Comparison of the remaining peak ratio to the ratio from the H2O2 and OH calibration experiment provided the experimentally derived HO2 yield for reaction of OH and the reagent of study.

Branching ratios to direct HO2 formation could be assigned with an accuracy of ∼10 %; the limitations to this were signal-to-noise effects, where improved signal-to-noise ratio could be achieved by increasing the precursor concentration and photolysis energy. However, this was limited by the need to ensure pseudo-1st-order conditions were maintained and to minimize radical–radical processes. Ensuring the dominant reaction was between OH and the reagent of study, whilst still being able to accurately measure the initial OH conditions provided a limit to the maximum removal rates achievable (kOH,1st<30000 s-1).

Many reactions of atmospheric and combustion interest are studied in the presence of oxygen leading to the generation of peroxy radicals (RO2). For certain RO2 there is a potential to generate OH and HO2 on a fast timescale and therefore three well-known reactions were chosen to characterize the instrument, OH and ethane, OH and ethylene, and OH and methanol. R6OH+C2H6→H2O+C2H5R7OH+C2H4→HOC2H4R8a, R9OH+CH3OH→H2O+CH2OH,CH2OH+O2→HO2+HCHO(fast)R8b, R10OH+CH3OH→H2O+CH3O,CH3O+O2→HO2+HCHO(slow) OH and ethane (Reaction R6) give an assessment of any false yields generated from RO2 and NO from prototypical alkyl RO2 species that will be formed from many atmospherically relevant reactions. Ethylene and OH (Reaction R7) form a hydroxy alkyl peroxy radical, the typical RO2 species known to create interferences in FAGE HO2 detection systems (Fuchs et al., 2011; Hard et al., 1984; Whalley et al., 2013). Minimizing and understanding the HO2 yield from this reaction allowed for limits to the selectivity of the instrument to be known. The reaction of OH with methanol is a well-understood reaction; the two isomeric radical products react with oxygen on differing timescales to generate HO2. Complete conversion of both isomers should yield 100 % HO2.

As discussed in the experimental section, transport effects after the breakup of the sampling jet mean that rate coefficients measured in the second cell kX,2nd (X=OH or HO2) differ from each other (transport effects scale with mass) and from those made in the first detection axis (kOH,1st). Pseudo-1st-order rate coefficients from the two detection axes were compared to ascertain whether measurements in the second detection axes can be used to make quantitative kinetic measurements.

Finally, the layout of the apparatus makes it hard to accurately measure the temperature at which the reaction occurs; for reactions occurring on a millisecond timescale, the relevant reaction distance from the sampling pinhole is approximately 0.05–0.5 mm. Compared to a conventional slow flow laser flash photolysis/laser-induced fluorescence apparatus, where the reaction volume is the overlap of the perpendicular photolysis and probe laser beams, it is hard to accurately position the thermocouple, and additionally any thermocouple located close to the sampling pinhole may affect the flow into the low-pressure system. In addition to the difficulties in correctly siting a thermocouple, there are additional errors (flow, conduction, and radiative) that derive from measuring the temperature of a flowing gas with a thermocouple. We have therefore performed additional experiments to determine the temperature based on the well-characterized and temperature-sensitive reaction of OH and methane.

The branching ratios for the sites of OH attack on n-butanol, as presented in Scheme 2, are of significance to the modelling of the ignition delay times for n-butanol (Agbro et al., 2017). Abstractions at the α and OH positions are chain-terminating reactions at low temperatures due to the formation of the relatively inert HO2 radical, and abstraction at the β site leads to chain propagation, through OH recycling. The new instrument permitted determination of the attack at the α and β sites; attack at the α site leads to prompt HO2 formation in the presence of O2; at elevated temperatures bi-exponential fits to non-single exponential OH loss traces in the absence of O2 (due to decomposition of the β hydroxy radical to OH and iso-butene) allowed for attack at the β site to be measured.

An instrument based on the FAGE technique for monitoring OH radicals and validated via reproduction of the literature kinetics of several OH reactions over a wide range of temperatures and pressures has been extended to allow for simultaneous HO2 and OH detection via the chemical titration of HO2 to OH with NO.

As mentioned in the introduction, the instrument has similarities to that presented by Nehr et al. (2011), where a FAGE system for sequential OH and HO2 is coupled to a lifetime instrument and yields of HO2 from OH-initiated reactions are reported. Although the principles of HOx detection used in both systems are similar, there are some significant differences between the two instruments. Some differences relate to the reaction cell in which the kinetics take place: 1 atm of air and 298 K for Nehr et al. (2011) and 0.5–5 atm of any gas and 298–800 K for this work. However, in principle, the Nehr et al. (2011) FAGE cell could be coupled to a different reaction cell to probe a wider range of conditions. A more substantial difference is the timescale of the chemistry taking place. Typical temporal profiles from Nehr et al. (2011) are of the order of a second compared to <10 ms in this work. The enhanced sensitivity of the Nehr et al. (2011) instrument means that radical–radical reactions should not interfere, but the technique may be subject to interferences from 1st-order (or pseudo-1st-order) reactions including heterogeneous processes. Detection of radicals in kinetics or yield experiments is difficult and studying reactions under a range of conditions is important to identify systematic errors; hence both instruments have a role to play.

The use of H2O2 as an OH precursor has been shown to provide a reliable method of internally characterizing our system for HO2 yield detection. Interferences that could arise from using this precursor for HO2 detection have been accounted for, and the presence of water that the H2O2 precursor introduces has no effect on the sensitivity of the LIF method, unlike IR absorption methods.

It has been demonstrated that this instrument can reliably assign HO2 yields and simultaneously measure OH kinetics, even under conditions of high temperatures and high oxygen concentrations, which could be challenging for other detection systems. Such conditions are important for exploring key combustion chemistry reactions and for converting slow atmospherically relevant processes to the microsecond timescales required to minimize secondary or heterogeneous chemistry.

By performing reactions under low NOx and low radical densities (<1×1013 molec. cm-3), HO2 yields formed on fast timescales (>300 s-1) can be assigned to direct HO2 channels or reactions of alkyl (R) radicals with oxygen. Whilst some time resolution is lost at the HO2 detector, sufficient time resolution is retained in order to separate varying sources of HO2, for example the two channels leading to HO2 production in the OH/CH3OH/O2 system (Sect. 3.2) or from unwanted secondary chemistry.

For particular reactions, illustrated in this paper by the study of OH with methanol and butanol in the presence of oxygen, the simultaneous measurement of OH kinetics and HO2 yields can provide important site-specific information. In other systems, the onset of HO2 formation could allow for the assignment of new channels becoming important within complex mechanisms, potentially allowing for verifying the onset of OOQOOH chemistry within OH regeneration processes.