Reactive quenching of electronically excited NO ∗ 2 and NO ∗ 3 by H 2 O as potential sources of atmospheric HO x radicals

. Pulsed laser excitation of NO 2 (532–647 nm) or NO 3 (623–662 nm) in the presence of H 2 O was used to ini-tiate the gas-phase reaction NO ∗ 2 + H 2 O → products (Reac-tion R5) and NO ∗ 3 + H 2 O → products (Reaction R12). No evidence for OH production in Reactions (R5) or (R12) was observed and upper limits for OH production of k 5 b /k 5 < 1 × 10 − 5 and k 12 b /k 12 < 0 . 03 were assigned. The upper limit for k 5 b /k 5 renders this reaction insigniﬁcant as a source of OH in the atmosphere and extends the studies (Crow-ley and Carl, 1997; Carr et al., 2009; Amedro et al., 2011) which demonstrate that the previously reported large OH yield by Li et al. (2008) was erroneous. The upper limit obtained for k 12 b /k 12 indicates that non-reactive energy transfer is the dominant mechanism for Reaction (R12), though generation of small but signiﬁcant amounts of atmospheric HO x and HONO cannot be ruled out. In the course of this work, rate coefﬁcients for overall removal of NO


Introduction
The capacity of the atmosphere to oxidise trace gases released at the Earth's surface is sensitively dependent on the concentration of the hydroxyl radical OH (Lelieveld et al., 2008).Most atmospheric OH is believed to be generated via a combination of primary photolytic processes involving O 3 (λ ≤ 370 nm; IUPAC, 2018) (Reactions R1, R2) and HONO (λ: 280-370 nm; IUPAC, 2018), for example, as well as in the reaction of NO with HO 2 , the latter being formed in the troposphere via the oxidative degradation of organic trace gases.
As a large fraction of the oxidation of organic trace gases is initiated by reaction with OH, the conversion of HO 2 back to OH (e.g. via reaction with NO) is often referred to as recycling; the relative importance of direct OH formation and recycling depend on the concentrations of organics and NO.Together, OH and HO 2 are referred to as HO x .Any reaction that can generate OH or HO 2 directly or indirectly (e.g. via generation of a short-lived OH precursor such as HONO) thus contributes to atmospheric oxidation capacity.Processes that form HONO (both gas phase and heterogeneous) are therefore of great interest to atmospheric science and have been the subject of many studies (see, e.g.Stemmler et al., 2007;Li et al., 2014;Meusel et al., 2016).Two processes that may potentially generate HO x and HONO are the gas-phase reactions of H 2 O with electronically excited nitrogen dioxide (NO 2 A 2 B 2 henceforth NO * 2 ) and the electronically excited nitrate radical (NO 3 A 2 E and B 2 E , henceforth NO * 3 ).
3 troposphere.It was argued that non-dissociative absorption by NO 2 (Reaction R4) could lead to formation of OH and HONO in a process (channel R5b of the overall reaction with H 2 O, R5) that is exothermic for excitation wavelengths across the visible absorption spectrum of NO 2 , which extends to ≈ 650 nm.
NO 2 + hν(≤ 650 nm) → NO * 2 (R4) The rate of OH formation following NO 2 excitation in the atmosphere depends on the OH yield (k 5b /k 5 ) and on the relative rates of NO * 2 deactivation by H 2 O (Reaction R5) and by N 2 and O 2 (Reaction R6).For details of the NO 2 cross sections, quantum yields, quenching rate constants, and associated photophysics for these processes, we refer to our previous publication (Crowley and Carl, 1997).
Crowley and Carl (1997) used 532 nm pulsed-laser excitation of NO 2 to determine an upper limit to the OH yield of (k 5b /k 5 ) ≤ 7 × 10 −5 .Crowley and Carl (1997) also identified routes to O( 1 D) at shorter wavelengths that involved two-photon excitation of NO 2 , and which lead indirectly to OH formation via reaction of O( 1 D) with H 2 O. Whilst of some utility in the laboratory, such processes that require multiphoton excitation are generally of no consequence for the atmosphere.
More than 10 years later, Li et al. (2008) carried out similar experiments but at longer wavelengths (560-640 nm) and reached very different conclusions, deriving a yield of OH (and thus also HONO) close to 1×10 −3 , a factor of 14 larger than the upper limit of Crowley and Carl (1997).Calculations of the impact of Reactions (R4)-(R5) using the large yield reported by Li et al. (2008) led to the conclusion that Reaction (R5) is important for air quality under highly polluted conditions; use of the lower yield from Crowley and Carl (1997) resulted in minimal impact (Wennberg and Dabdub, 2008;Ensberg et al., 2010).Subsequent to the work of Li et al. (2008), two further experimental studies (Carr et al., 2009;Amedro et al., 2011) appeared to confirm the conclusions of Crowley and Carl (1997) and suggested that the high yield reported by Li et al. (2008) was an experimental artefact, resulting from multiphoton laser excitation of NO 2 in their focussed laser beam (Amedro et al., 2011).However, the experiments of Amedro et al. (2011) at 565 nm andCarr et al. (2009) at 563.5 and 567.5 nm used NO * 2 prepared at wavelengths that covered only a small portion of the 560-630 nm range from Li et al. (2008).The single wavelength (532 nm) used by Crowley and Carl (1997), whilst interrogating the same excited states of NO 2 , was outside of the range of wavelengths covered by Li et al. (2008).The principal goal of the experiments on Reaction (R5) described in this work was therefore to measure OH yields (k 5b / k 5 ) using a range of photoexcitation wavelengths similar to those employed by Li et al. (2008) but avoiding potential complications related to multiphoton excitation.

NO
The NO 3 radical is generated throughout the atmospheric diel cycle via the oxidation of NO 2 by O 3 : At night, NO 3 can acquire mixing ratios of hundreds of parts per trillion by volume.The high reactivity of NO 3 towards unsaturated, organic trace gases (especially biogenically emitted ones in forested regions; Liebmann et al., 2018a, b) makes it an important nocturnal oxidant.NO 3 is generally considered to be unimportant during the daytime due to rapid photolysis.Rapid photodissociation (Reactions R8a and R8b) following absorption of visible light reduces the daytime NO 3 lifetime to only a few seconds and usually limits mixing ratios to less than 1 pptv.
NO 3 photophysics has been the subject of many studies, up to 1991 reviewed by Wayne et al. (1991).Briefly, the NO 3 absorption spectrum (≈ 400-665 nm) is broad and diffuse with an extended excited-state lifetime of several hundred microseconds (Nelson et al., 1983) for excitation beyond the photodissociation limit.The extended lifetime results from coupling between ro-vibrational levels of the ground (X 2 A 2 ) state and the excited (A 2 E and B 2 E ) electronic states, so that excitation into the strongest feature (centred at ≈ 662 nm) can be considered to populate a manifold of mixed ground and excited electronic states (Carter et al., 1996).For simplicity, we refer to excited state NO 3 as NO * 3 .NO * 3 can dissociate (Reactions R8a, R8b, dominant at excitation wavelengths < 630 nm), fluoresce (Reaction R9), and return to the ground state or be quenched in collisions with the main atmospheric bath gases N 2 , O 2 , and H 2 O (Reactions R10-R12).Fluorescence and collisional quenching are important only at wavelengths longer than ≈ 630 nm.
The net result of NO 3 formation in Reaction (R7) and photolysis via the main channel, Reaction (R8a), is no change in NO x (NO x = NO + NO 2 ) or O 3 .The net effect of formation in Reaction (R7) and photolysis via the minor (20 %) channel (Reaction R8b) is conversion of NO 2 to NO (i.e.no net loss of NO x ) and conversion of O 3 to O 2 (loss of odd oxygen).Reaction of NO * 3 with H 2 O to form OH + HNO 3 (Reaction R12b) changes this picture dramatically.As illustrated in Fig. 2, if NO * 3 reacts with H 2 O to form OH + HNO 3 (Reaction R12b), the net effect is conversion of NO 2 to HNO 3 (i.e.loss of NO x ) and conversion of O 3 and H 2 O to OH.This process (Reactions R7, R12b) therefore allows formation of atmospheric OH from O 3 in the absence of actinic UV radiation normally required to generate O( 1 D) from O 3 (Reaction R1).If NO * 3 reacts with H 2 O to form OH + HONO + O 2 , as in Reaction (R12c), the net effect is conversion of NO 2 to NO (no loss of NO x ) and formation of two HO x molecules, again bypassing the need for the actinic radiation in the UV wavelength.Using literature values for the wavelength-dependent NO 3 absorption cross sections (Yokelson et al., 1994) and photolysis quantum yields (Orlando et al., 1993) as well as actinic flux (calculated using the TUV program (http://cprm.acom.ucar.edu/Models/TUV/Interactive_TUV/, last access: June 2018) for 50 • N at a zenith angle of 27 • , an overhead O 3 column of 300 Du, a surface albedo of 0.1, and an aerosol optical depth of 0.235), we calculate that, averaged over the NO 3 absorption spectrum, 60 % of actinic photons absorbed result in dissociation of NO 3 .The residual 40 % results in formation of NO * 3 , which can then undergo chemical and photophysical transformation.Figure 1 gives an example of the relative rates of photodissociation and (non-dissociative) photoexcitation across the NO 3 absorption spectrum.
The relative importance of fluorescence and the collisional deactivation processes depends on the fluorescence lifetime and the rate constants for quenching.Nelson et al. (1983) report two components to the NO 3 fluorescence decay they observed following excitation at 661.9 nm, with collision-free fluorescence lifetimes of 27 and 340 µs.
The longer-lived component (accounting for > 85 % of the total fluorescence) was quenched by N 2 and O 2 with rate coefficients of Nelson et al. (1983) did not report a quenching rate coefficient for H 2 O but determined large quenching coefficients for propane (1.09 × 10 −10 cm 3 molecule −1 s −1 ) and nitric acid (3.07 × 10 −10 cm 3 molecule −1 s −1 ), presumably resulting from more efficient energy transfer due to higher densities of states in these polyatomic molecules.A substantially larger rate coefficient for quenching of NO * 3 by H 2 O of k 12 = (6.9± 0.5) × 10 −10 cm 3 molecule −1 s −1 ) was reported by Fenter and Rossi (1997).The quenching rate constants are sufficiently large that, at the pressures of N 2 , O 2 , and H 2 O available in the troposphere, relaxation of NO * 3 via fluorescence can be neglected.
The fraction, f H 2 O , of tropospheric NO * 3 that will be quenched by collision with H 2 O rather than N 2 or O 2 is given by expression (1): (1) Using this expression we calculate that, at the Earth's surface (1 bar of pressure) and a temperature of 25 • C, f H 2 O can vary between 0.2 and 0.5 for relative humidities between 20 % and 80 %.As mentioned above, daytime concentrations of NO 3 are generally low due to rapid photolysis (and reaction with NO), though measurements in polluted environments indicate maximum daytime concentrations of Geyer et al., 2003).The atmospheric production rate of OH via NO 3 excitation may be written as Using an NO 3 concentration of 1 × 10 8 molecule cm −3 and J exci = 0.15 s −1 (Fig. 1) enables us to calculate an OH production rate (at 80 % relative humidity) of 7.5 × 10 6 molecule cm −3 s −1 if all quenching of NO * 3 by H 2 O is reactive and forms OH.To put this value in context, we note that typical OH production rates from photolysis of O 3 are around 2 × 10 5 molecule cm −3 s −1 , a factor of ≈ 40 lower.The principal objective of this work was therefore to determine the OH production rate via NO 3 photoexcitation and subsequent reaction of NO * 3 with H 2 O (Reaction R12b).To best constrain these measurements, rate coefficients for total removal (quenching and chemical reaction) of NO * 3 by H 2 O (k 12 ) and N 2 (k 10 ) were determined.

Experimental
All experiments were conducted in a 500 cm 3 jacketed photolysis cell as described previously (Wollenhaupt et al., 2000;Dillon et al., 2006).Laser light entered and exited the reaction vessel via Brewster-angle quartz windows; laser fluence at each wavelength was recorded using a Joule meter located behind the exit window.An excimer laser was used to generate ≈ 20 ns pulses of light at 193 nm (ArF) or 248 nm (KrF).Dye lasers pumped by Nd:YAG lasers were used to generate pulsed (≈ 6 ns) tuneable radiation at visible wavelengths.
The pressure and the gas flow rate (300-2000 cm 3 (STP) min −1 ) were regulated to ensure that a fresh gas sample was available for each laser pulse for operation at 10 Hz.The pulsed laser-based schemes for generation of excited NO 2 and NO 3 are described below, as are the schemes for calibration of the OH signal.
Concentrations of the key reactants and precursors (NO 2 , HNO 3 , and H 2 O) were monitored by UV-visible absorption spectroscopy, reducing potential uncertainties in each of these parameters to ≤ 10 %.NO 2 was measured in situ using a multipass absorption cell positioned upstream of the reactor.Light from a halogen lamp passing through the cell was focused onto the entrance slit of a 0.5 m monochromator.A diode-array detector was used to record NO 2 absorption in the visible range of light in the range 398 ≤ λ ≤ 480 nm at an instrumental resolution of 0.32 nm, determined from the full width at half maximum (FWHM) of the 436.8 nm Hg emission line.Optical absorption by HNO 3 and H 2 O was determined using a "dual-beam" absorption cell (Hg line at 184.95 nm, l = 43.8nm) located downstream of the photolysis reactor.In experiments in which both HNO 3 and H 2 O were present, they were added sequentially so first the optical density due to a single component was measured before the second was added and the resultant total optical density was monitored.
The output from a Nd:YAG pumped dye laser operating with Rhodamine 6G dye was frequency doubled to 282 nm and used to detect OH via excitation the A 2 (v = 1) ← X 2 (v = 0) transition close to 282 nm.Laser-induced fluorescence (LIF) was detected by a photomultiplier tube shielded by a combination of a 309 nm (±5 nm) interference filter and BG 26 (glass) filter.Directly following experiments to measure formation of OH in the title reactions, known amounts of OH were generated via pulsed laser photolysis of HNO 3 (at 248 nm) or H 2 O (at 193 nm).
For the experiments on NO * 2 , a small flow of HNO 3 diluted in N 2 was added to the N 2 bath and a series of experiments were conducted that covered a range of laser fluences at 248 nm.The additional flow was compensated for by reducing the main N 2 flow so that different concentrations of OH were generated in essentially unchanged conditions of pressure, temperature, [NO 2 ], and [H 2 O].When using Reaction (R14) to calibrate the OH signal, the NO 2 supply to the experiment was replaced with N 2 , and 193 nm light was used to dissociate OH from the H 2 O already present (in unchanged conditions of pressure, temperature, and [H 2 O]).
The uncertainty associated with conversion of LIF signals into OH concentrations stemmed partially from uncertainties in (measured) [HNO 3 ] and [H 2 O] but was dominated by uncertainty in the measurement of the laser fluence at the centre of the reactor.Such measurements depended on both the accuracy of the Joule meter and corrections for beam divergence and the assumption of a homogeneous light intensity over the cross section of the laser beam.An overall uncertainty of 40 % was estimated for the conversion of LIF signals to absolute [OH] required for (k 5b /k 5 ) determinations.For determination of k 12b /k 12 , the self-calibrating chemistry (Reactions R13, R15) results in a smaller contribution of 2 and NO * 3 14009  laser fluence uncertainty to the overall uncertainty, which is dominated by assumptions regarding the NO 3 profile (see later).Chemicals used are as follows.NO 2 (ABCR 99.99 %) was subject to repeated freeze-pump-thaw cycles at 77 K prior to dilution in N 2 and storage in blackened glass bulbs.H 2 O (Milli-Q de-ionised water) and HNO 3 (prepared in house from H 2 SO 4 + KNO 3 ) were added to the reactor via bubblers.O 3 was generated via electric discharge through O 2 in a commercial ozoniser (Anseros).N 2 O 5 was prepared by mixing O 3 with NO 2 and trapping the resulting N 2 O 5 at 195 K (Wagner et al., 2008).N 2 and O 2 (Westfalen, 99.999 %) were used as supplied.NO 2 was excited at a number of different wavelengths: 532 and 567-647 nm; reagent concentrations and conditions for these experiments are given in Table 1.In general, large concentrations of H 2 O were used to promote reaction of NO * 2 over deactivation by other colliders, notably N 2 , and to ensure that changes in other reagent concentrations (e.g. for calibration; see above) had a minimal effect on fluorescence quenching or other processes that impact OH-LIF detection sensitivity.
Figure 3 displays the results of an experiment in which NO 2 was excited at 532 nm (at t = 280 µs) to generate 10 13 to 10 14 molecule cm −3 of NO * 2 .The delay of 280 µs is the time between the triggering of the flash lamps (at t = 0) and the Q-switch of the YAG laser.The solid black triangles were obtained with the OH-excitation laser tuned to 282 nm (on resonance) and indicate a change in signal of ≈ 200 to 350 µs.This signal does not display the kinetic behaviour of OH in this chemical environment and remains when the OHexcitation laser is tuned off resonance (red triangles).It is also present when the 532 nm light is blocked and we conclude that this weak signal, having neither kinetics nor spectroscopy characteristic of OH, is an artefact with electronic origin, possibly related to the output of the pulse generator used to trigger the laser Q-switch.
The data represented by open circles (roughly independent of reaction time) are the results of OH-calibration experiments using the 193 nm photolysis of H 2 O (1.5 × 10 17 molecule cm −3 ) at four laser fluences between 0.3 and 6.8 mJ cm −2 in the absence of NO 2 .The OH concentration was calculated using a 193 nm cross section for H 2 O of 2.1 × 10 −21 cm 2 molecule −1 (Sander et al., 2011).
The roughly constant OH level over 1000 µs is consistent with the fact that OH does not react with any components of the gas mixture.An experiment at 193 nm using the same OH-generation scheme but in the presence of NO 2 is dis-played as solid stars.OH now decays exponentially at a rate which is consistent with its loss via reaction with NO 2 .In this experiment, some OH was also generated by the reaction of O( 1 D) (formed by the 193 nm photolysis of NO 2 ) with H 2 O and it was not used for calibration purposes.The signals obtained in the absence of NO 2 were converted to OH concentrations (right y axis) using Joule meter readings as described in Sect.2.1.
The solid black line in Fig. 3 represents the OH signal and concentration expected from our experimental conditions (NO 2 concentration, H 2 O concentration, total pressure, and 532 nm laser fluence) and literature data for NO 2 absorption cross sections, NO * 2 deactivation rate constants, and the yield of OH from NO * 2 + H 2 O reported by Li et al. (2008).Within experimental uncertainty (see below) our data are clearly not consistent with the large yield of OH reported by Li et al. (2008).In order to rule out the possibility that this is a result of using different excitation wavelengths, similar experiments were carried out in which we explored different regions of the NO 2 absorption spectrum.OH signals were not observed at any wavelength, enabling us to set upper limits to k 5b /k 5 .The upper limits were calculated from the minimum observable OH signal (assumed to be twice the RMS noise levels on the OH signal) and accounting for uncertainty in parameters such as laser fluence (30 %), NO 2 concentration (10 %), and concentration of H 2 O (10 %).
The results are summarised in Table 1, which lists the experimental conditions in detail and in Fig. 4, in which we also compare to literature determinations of k 5b /k 5 .The present dataset and those reported by Crowley and Carl (1997), Amedro et al. (2011), andCarr et al. (2009) found OH formation in the reaction between NO * 2 and H 2 O to be inefficient, with upper limits to k 5b /k 5 of between 6 × 10 −6 and 1.4 × 10 −4 at all wavelengths investigated.Together, these datasets contradict the yield of 1 × 10 −3 reported by Li et al. (2008) for excitation across the wavelength range of 560 to 630 nm.Our dataset, covering three absorption features of the NO 2 absorption spectrum within the range reported by Li et al. (2008), also rules out that the poor agreement is due to use of different excitation wavelengths.As discussed by Amedro et al. (2011) the use of focussed laser beams and the resulting multiphoton processes are the most likely explanation for OH formation in the work of Li et al. (2008).The results from this work reduce the maximum yield of OH from the reaction of NO * 2 with H 2 O to 6 × 10 −6 at 532 nm as opposed to 7×10 −5 measured by Crowley and Carl (1997).The assumption that this value is valid across the non-dissociative part of the absorption spectrum of NO 2 enables us to conclude that formation of atmospheric OH (and HONO) via Reaction (R5b) is insignificant.

Generation of NO 3
For the experiments to investigate the reaction of NO * 3 with H 2 O (Reaction R12), NO 3 was generated via the reaction of OH with known amounts of HNO 3 (Reaction R15).
The rate constant and the yield of NO 3 (unity) for Reaction (R15) are well known (Brown et al., 1999(Brown et al., , 2001;;Carl et al., 2001;Dulitz et al., 2018), enabling the timedependent NO 3 concentration profile to be calculated if the initial amount of OH is known.This initial concentration of OH depends on the 248 nm laser fluence (measured by a Joule meter, uncertainty 30 %) and the HNO 3 concentration (measured by optical absorption at 185 nm, uncertainty 10 %).As OH was formed from HNO 3 photolysis (Reaction R13) and the OH decay was monitored, these experiments were self-calibrating as long as a sufficient excess of In the conditions employed in this work (see Table 2), radical losses via unwanted self-reactions and cross reactions of OH and NO 3 were < 5 % of the total OH loss rate, which was dominated by Reaction (R15).In experiments to measure the rate constant for NO * 3 quenching by N 2 (Reaction R10) and H 2 O (Reaction R12), NO 3 was generated via the 248 nm photolysis of N 2 O 5 : In this scheme of NO 3 generation, NO 3 is formed instantaneously (in contrast to Reactions R13-R15).(Fenter and Rossi, 1997) and the value derived (6.9 × 10 −10 cm 3 molecule −1 s −1 ) is unexpectedly large, we chose to remeasure k 12 .In these experiments, NO 3 was generated in Reaction (R16) and He was used as the main bath gas, with traces of N 2 and H 2 O added.

Atmos
An excitation laser pulse at 662 nm was triggered when the NO 3 concentration was close to its maximum value (i.e. when > 95 % of the primary OH had been consumed by reaction with HNO 3 ) to generate NO * 3 .Time-dependent fluorescence from NO * 3 (λ > 690 nm) was detected using a redsensitive photomultiplier and recorded on a 100 MHz digital oscilloscope.Fluorescence decay constants in the presence of various concentrations of H 2 O were then used to derive k 12 .We also conducted a set of experiments using N 2 as a quenching molecule to test our experimental methodology by comparison with literature measurements of k 10 .
NO 3 fluorescence profiles from these experiments are displayed in Fig. 5, in which datasets are depicted in which various amounts of N 2 (Fig. 5a) and H 2 O (Fig. 5b) were added to the He bath gas.The fluorescence decay rate constant (k f ) derives from the sum of processes that depopulate the excited state and includes fluorescence, inter-system crossing, and quenching by N 2 , H 2 O, and N 2 O 5 with rate constants In line with previous studies (Nelson et al., 1983), the slow component of the NO 3 fluorescence was found to decay mono-exponentially (black and red lines in Fig. 5a and b) and depended on the pressure of N 2 or H 2 O.
The decay constant (k f ) was derived from exponential fits to the data and plotted against the concentration of N 2 or H 2 O (Fig. 6) to obtain (from the slopes) the rate constants k 10 and k 12 for quenching by N 2 and H 2 O, respectively.Assuming negligible contribution from OH, NO 2 , and NO 3 , due to their low concentrations, the y-axis intercepts in Fig. 6 (≈ 0.5-0.8× 10 6 s −1 ) are the combined terms where k q (N 2 O 5 ) is the unknown rate constant for quenching of NO * 3 by N 2 O 5 .As the collision-free lifetime of excited NO 3 is several hundred microseconds, the terms k f and k ISC contribute insignificantly to the fluorescence decay.The intercept (≈ 5-8 × 10 5 s −1 ) is consistent with N 2 O 5 concentrations in the range of 10 15 molecule cm −3 and a value of k q (N 2 O 5 ) of the order of 10 −10 cm 3 molecule −1 s −1 .
Our result obtained for N 2 , k 10 = (2.1 ± 0.2) × 10 −11 cm 3 molecule −1 s −1 , is in reasonable agreement with the value of (1.7 ± 0.2) × 10 −11 cm 3 molecule −1 s −1 reported by Nelson et al. (1983).In contrast, our result for quenching by water vapour, k 12 = (1.6 ± 0.3) × 10 −10 cm 3 molecule −1 s −1 , is more than a factor of 4 lower than that reported by Fenter and Rossi (1997).As both studies used 662 nm excitation of NO 3 and similar methods to derive k 12 , the differences are likely to be related to the measurement of the H 2 O concentration.As we measured the H 2 O concentration in situ (optical absorption at 185 nm), the uncertainty of our result is expected to be determined by uncertainty in the absorption cross section of H 2 O at this wavelength, which, based on good agreement across several measurements (Cantrell et al., 1997;Hofzumahaus et al., 1997;Creasey et al., 2000), we estimate to be < 10 %.Fenter and Rossi (1997) relied on flow measurements to derive the concentration of H 2 O in their experiments.Because of this, we consider our measurement of k 10 to be the more accurate one and use this value for further evaluation of our experiments to derive k 12b /k 12 .

Yield of OH from NO
Figure 7 displays the results of an experiment using three pulsed lasers.The first (excimer laser at time zero) generated OH from the 248 nm photolysis of HNO 3 .In this particular experiment the HNO 3 concentration (monitored at 185 nm) was 6.3 × 10 15 molecule cm −3 and a laser fluence of 13 mJ cm −2 was used to generate 2.0 × 10 12 OH cm −3 .This OH monitored by the 282 nm LIF laser out to a reaction time of 10 ms (open circles in Fig. 7) was observed to decay at a rate consistent with its well-characterised reaction with HNO 3 k 15 (298 K, 30 hPa) = 1.3 × 10 −13 cm 3 molecule −1 s −1 (Dulitz et al., 2018).NO 3 is the unique product of Reaction (R15) (Brown et al., 2001;Carl et al., 2001).The NO 3 profile (dashed line), calculated from initial OH and HNO 3 concentrations, is also displayed in Fig. 7.Here we calculate that 97 % of the initial OH will react with HNO 3 , the balance resulting from diffusion and reaction with NO 2 .The decay of NO 3 at long reaction times is due to NO 3 diffusion from the reaction volume so that its concentration at 8.28 ms (when the 662 nm laser is triggered) was reduced by ≈ a factor of 2 compared to the stoichiometric yield of 2 × 10 12 molecule cm −3 (i.e. when all OH is converted to NO 3 ).The decay of NO 3 was calculated from the known diffusion loss constant for OH at this pressure and the relative reduced masses of OH/N 2 and NO 3 /N 2 .A delay of 8.28 ms allowed the primary OH to decay to very low values (i.e.≈ 10 9 molecule cm −3 ) before triggering the 662 nm excitation laser.The measured laser fluence at 662 nm was then combined with the NO 3 concentration at 8.28 ms to calculate the fractional excitation of NO 3 (generally about 10 %) and thus the concentration of NO * 3 formed.When using very large laser fluences at 662 nm we calculate that the transition was saturated and then assume equal concentrations of ground and excited-state NO 3 directly after the excitation pulse.
The solid lines starting at t = 8.28 ms represent the expected OH signal if the value of k 12b /k 12 were 0.0, 0.01, 0.05, and 0.1 and were calculated using the rate constants for quenching of NO * 3 by N 2 and H 2 O as derived in this study as well as the concentrations of N 2 and H 2 O. 2 and NO * 3 14013 Clearly, the data from the experiment illustrated in Fig. 7 are consistent with a value of k 12b /k 12 that lies between 0 % and 1 %.Similar experiments were repeated for different starting conditions and photoexcitation wavelengths (623, 629, and 662 nm) corresponding to strong absorption features of NO 3 .No evidence for OH production in Reaction (R12) was observed in any experiment and an upper limit to the yield of OH was obtained from the random noise on the experimental OH-trace data and the expected OH signal.These values are tabulated in Table 2.The major sources of uncertainty in the calculated OH yield are uncertainty in the measurement of laser fluences (30 %) required to calculate the initial OH and NO * 3 concentrations and assumptions related to the (unmeasured) NO 3 time profile.NO 3 is relatively unreactive in this system as it does not react with HNO 3 and only slowly with NO 2 (formed in Reaction R13) at these pressures.We calculate that ≈ 5 % of the NO 3 formed is lost via reaction with OH (k(OH + NO 3 ) = 2 × 10 −11 cm 3 molecule −1 s −1 ) (Atkinson et al., 2004), its major removal processes being diffusive transport.The diffusive loss rate constant for NO 3 in this system was calculated from the known diffusive loss rate constant of OH under the same conditions of pressure and temperature.In the absence of corroborative measurements of the NO 3 profile in these experiments, we conservatively assume a factor of 2 uncertainty in the NO 3 concentration at the time of the excitation pulse.We thus derive an upper limit of k 12b /k 12 < 0.03 following photoexcitation at 623, 629, and 662 nm.This indicates either that the rapid quenching of NO * 3 by H 2 O predominantly involves energy transfer rather than reaction or that the products formed in reactive quenching do not include OH.

Atmospheric implications and conclusions
The results obtained in this work and elsewhere (Crowley and Carl, 1997;Carr et al., 2009;Amedro et al., 2011) clearly demonstrate that the large values of k 5b /k 5 reported by Li et al. (2008) were erroneous.In this work we were able to reproduce, extend, and improve upon previous results (i.e.obtain smaller upper limits for k 5b /k 5 ).The extension of the database to a wider range of photoexcitation wavelengths was important since the majority of the data from Li et al. (2008) were obtained at wavelengths red-shifted from those of the other groups.In the modelling study by Wennberg and Dabdub (2008) the largest impacts of Reaction (R5b) on air quality (enhancements in O 3 of ≈ 40 %) were found when using k 5b /k 5 = 10 −3 from Li et al. (2008).Small but still significant impact changes in O 3 and particle mixing ratios were calculated when using the upper limit of k 5b /k 5 = 7×10 −5 provided by Crowley and Carl (1997).Results from this work, with upper limits to k 5b /k 5 an order of magnitude smaller than those available previously, enable us to conclude that the formation of OH in NO * 2 +H 2 O is not an important atmospheric process.
Our upper limit of 3 % to OH formation from the reactive quenching of NO * 3 by H 2 O can be put in context using Eqs.( 1) and (2).We combine our measurements of k 10 = 2.1 × 10 −11 cm 3 molecule −1 s −1 and k 12 = 1.6 × 10 −10 cm 3 molecule −1 s 1 with the literature value for k 11 (2.1 × 10 −11 cm 3 molecule −1 s −1 ; Nelson et al., 1983) to derive f H 2 O = 0.16 at 25 • C and a relative humidity of 80 %.Using the same excitation rates and concentrations of NO 3 described in Sect.1.1 and our upper limit of k 12b /k 12 = 0.03, we derive an OH production rate of ≈ 7 × 10 4 OH cm −3 s −1 .Whilst this value is ≈ 2 orders of magnitude lower than that calculated in Sect.1.1 in which we assumed that all NO * 3 + H 2 O interactions form OH and used the high value of k 12 from the literature (Fenter and Rossi, 1997), it is nonnegligible compared to OH production rates from photolysis of O 3 (see Sect. 1.2) and may still represent an important contribution to OH formation in environments in which OH generation via traditional routes involving absorption of UV radiation is suppressed, for example, at high latitudes in winter.
The low yield of OH most likely results from the dominance of collisional energy transfer over reactive quenching of NO 3 by H 2 O (k 12b k 12 ).However, we also consider the possibility that the non-observation of OH in our experiments reflects the fact that the preferred products are HONO + HO 2 (i.e.k 12c > k 12b ) even though the molecular rearrangements required to form these products are less straightforward than for formation of OH and HNO 3 if excited-state NO 3 has the same (approximate) D 3h symmetry as the ground state and formally contains no O-O bonds.The conversion of HO 2 to (detectable) OH via addition of NO was not feasible owing to the rapid reaction of NO with NO 3 (k(HO 2 + NO) ≈ 8 × 10 −12 cm 3 molecule −1 s −1 , k(HO 2 + NO) ≈ 3 × 10 −11 cm 3 molecule −1 s −1 (Atkinson et al., 2004;IUPAC, 2018).
Given that our experiments were blind to formation of HO 2 or HONO, a detailed discussion of the atmospheric role of Reaction (R12c) is not warranted.However, the potential importance of Reaction (R12c) can be illustrated by assuming a 10 % yield of HONO and HO 2 (k 12c /k 12 = 0.1) and the same temperature, NO 3 concentration, and relative humidity outlined above.With this scenario, we calculate production rates of HO 2 and HONO of ≈ 2 × 10 5 molecule cm −3 s −1 .For HONO, this production rate is comparable to its formation in the gas-phase reaction between OH and NO under low-NO x conditions but lower than the missing production rate of ≈ 1-5×10 6 molecule cm −3 s −1 that has been observed in several environments as summarised by Meusel et al. (2016).In terms of HO 2 formation, a rate of 2 × 10 5 molecule cm −3 s −1 would be comparable to that obtained by the photolysis of ≈ 0.5 ppbv of HCHO (assuming a J value for HCHO of ≈ 2 × 10 −5 s −1 ).In conclusion, whilst our experiments indicate that the reactive quenching of ex-

14007Figure 1 .
Figure 1.Rate constants for dissociative (black line, J diss ) and non-dissociative (blue line, J exci ) excitation of NO 3 .The data use solar radiation actinic flux at the surface at 50 • N and a solar zenith angle (SZA) of 27 • (red line) as well as the NO 3 absorption cross sections and quantum yields.J values (and fraction of NO 3 dissociated) were obtained by integration of the excitation rate (quanta s −1 nm −1 ) over the wavelength range of absorption.

Figure 2 .
Figure 2. Net effects of reactive removal of NO * 2 and NO * 3 by H 2 O.

Figure 3 .
Figure 3. Photoexcitation of NO 2 at 532 nm.The open circles are OH calibrations obtained by the 193 nm photolysis of 1.5 × 10 17 molecule cm −3 H 2 O (in the absence of NO 2 ) at different laser fluences (mJ cm −2 ).The solid stars are data points from an OH calibration in the presence of NO 2 .The black triangles are data obtained by photoexcitation of [NO 2 ] = 4.0 × 10 15 molecule cm −3 using 532 nm (50 mJ cm −2 ) in the presence of [H 2 O] = 1.5 × 10 17 molecule cm −3 .The red triangles are the results of an identical experiment, but with the OH-excitation laser tuned away from the OH feature at 282 nm.The solid black line represents the OH signal and concentration expected from the yield of OH from NO * 2 + H 2 O reported by Li et al. (2008).

Figure 4 .
Figure 4. Summary of data obtained following photoexcitation of NO 2 at various wavelengths.The data from this study, Crowley and Carl (1997), Carr et al. (2009), and Amedro et al. (2011) are all upper limits, indicated by the downward arrows.The NO 2 absorption cross sections were taken from Vandaele et al. (1998).

Figure 5 .
Figure 5. Exponential decay of fluorescence from NO 3 following photoexcitation at 623 nm in the presence of N 2 (a) and H 2 O (b).An approximate NO 3 concentration of 3×10 12 molecule cm −3 was generated via the 248 nm photolysis (Reaction R16) of N 2 O 5 (≈ 10 15 molecule cm −3 ) in all quenching experiments.

Figure 7 .
Figure 7. Plot of primary OH and expected OH (solid lines after 8.28 ms) from NO * 3 + H 2 O at various values (0 % to 10 %) of k 12b /k 12 .The initial OH concentration (right y axis) was 2.0 × 10 12 molecule cm −3 .The dashed red line displays the calculated NO 3 concentration, which at 8.28 ms (time of 662 nm excitation pulse) was 9 × 10 11 molecule cm −3 .In these conditions 50 % of available NO 3 was excited to NO * 3 by absorption at 662 nm; 35 % of this NO * 3 proceeded to react with H 2 O in Reaction (R12), with the balance quenched by N 2 or HNO 3 .The solid lines (t > 8.28 ms) represent expected OH signals for values of k 12b / k 12 between 0 % and 10 %.
. Chem.Phys., 18, 14005-14015, 2018 www.atmos-chem-phys.net/18/14005/2018/ 2 and NO * 3 14011 2 O (k 10 and k 12 ) The fate of electronically excited NO 3 radicals in the atmosphere is controlled by the relative rate of quenching by H 2 O and the predominant bath gases N 2 and O 2 , which depends on both the concentration of H 2 O and the quenching rate coefficients k 10 , k 11 , and k 12 .As the rate constant for quenching of NO * 3 by H 2 O (k 12 ) has been addressed only briefly in a single study