Potential interferences in photolytic nitrogen dioxide converters for ambient air monitoring: Evaluation of a prototype

ABSTRACT

Mixing ratios of the criteria air contaminant nitrogen dioxide (NO₂) are commonly quantified by reduction to nitric oxide (NO) using a photolytic converter followed by NO-O₃ chemiluminescence (CL). In this work, the performance of a photolytic NO₂ converter prototype originally designed for continuous emission monitoring and emitting light at 395 nm was evaluated. Mixing ratios of NO₂ and NO_x (= NO + NO₂) entering and exiting the converter were monitored by blue diode laser cavity ring-down spectroscopy (CRDS). The NO₂ photolysis frequency was determined by measuring the rate of conversion to NO as a function of converter residence time and found to be 4.2 s⁻¹. A maximum 96% conversion of NO₂ to NO over a large dynamic range was achieved at a residence time of (1.5 ± 0.3) s, independent of relative humidity. Interferences from odd nitrogen (NO_y) species such as peroxyacyl nitrates (PAN; RC(O)O₂NO₂), alkyl nitrates (AN; RONO₂), nitrous acid (HONO), and nitric acid (HNO₃) were evaluated by operating the prototype converter outside its optimum operating range (i.e., at higher pressure and longer residence time) for easier quantification of interferences. Four mechanisms that generate artifacts and interferences were identified as follows: direct photolysis, foremost of HONO at a rate constant of 6% that of NO₂; thermal decomposition, primarily of PAN; surface promoted photochemistry; and secondary chemistry in the connecting tubing. These interferences are likely present to a certain degree in all photolytic converters currently in use but are rarely evaluated or reported. Recommendations for improved performance of photolytic converters include operating at lower cell pressure and higher flow rates, thermal management that ideally results in a match of photolysis cell temperature with ambient conditions, and minimization of connecting tubing length. When properly implemented, these interferences can be made negligibly small when measuring NO₂ in ambient air.

Implications: A new near-UV photolytic converter for measurement of the criteria pollutant nitrogen dioxide (NO₂) in ambient air by CL was characterized. Four mechanisms that generate interferences were identified and investigated experimentally: direct photolysis of HONO which occurred at a rate constant 6% that of NO₂, thermal decomposition of PAN and N₂O₅, surface promoted chemistry involving HNO₃, and secondary chemistry involving NO in the tubing connecting the converter and CL analyzer. These interferences are predicted to occur in all NO₂ P-CL systems but can be avoided by appropriate thermal management and operating at high flow rates.

Introduction

Nitrogen oxides (NO_x), comprised of nitric oxide (NO) and nitrogen dioxide (NO₂), are generated as by-products when air is used as an oxidant in combustion reactions. Once released to the atmosphere, NO_x plays a central role in photochemical ozone (O₃) production and aerosol formation (Frost et al. 2006; Meng, Dabdub and Seinfeld 1997), impacting air quality. Hence, continuous monitoring of NO_x concentrations at the points of emission at mixing ratios exceeding 1 part per million by volume (10⁻⁶, ppmv) and of NO_x in ambient air at trace level, usually below 1 part per billion by volume (10⁻⁹, ppbv) are needed. For both applications, NO-O₃ chemiluminescence (CL) is commonly chosen (or mandated by law) for NO_x analysis (Jahnke 2000; Penkett et al. 2011). In this method, NO is oxidized in excess O₃ to produce excited state NO₂*, which relaxes to the ground state by emitting a photon (Clough and Thrush 1966); the intensity of the generated light is proportional to the NO concentration sampled (Fontijn, Sabadell, and Ronco 1970):

(1) ${\mathrm{O}}_3+\mathrm{N}\mathrm{O}\to {\mathrm{O}}_2+\mathrm{N}{{\mathrm{O}}_2}^{\ast }$(1)

(2a) $\mathrm{N}{{\mathrm{O}}_2}^{\ast }\to \mathrm{N}{{\mathrm{O}}_2}^{\hspace{0.17em}}+\mathrm{h}\nu \left(600\mathrm{n}\mathrm{m}<\lambda <2800\mathrm{n}\mathrm{m}\right)$(2a)

(2b) $\mathrm{N}{{\mathrm{O}}_2}^{\ast }\to \mathrm{N}{\mathrm{O}}_2\left(\mathrm{n}\mathrm{o}\mathrm{n}-\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\right)$(2b)

Mixing ratios of NO_x (and, by difference, NO₂) are quantified following reduction of NO₂ to NO. One example method by which NO₂ is converted to NO is the heated molybdenum (Mo) converter (Fehsenfeld et al. 1987; Gluck et al. 2003; Williams et al. 1998):

(3) $\mathrm{M}\mathrm{o}+3\mathrm{N}{\mathrm{O}}_2\to \mathrm{M}\mathrm{o}{{\mathrm{O}}_3}^{\hspace{0.17em}}+3\mathrm{N}\mathrm{O}$(3)

Heated Mo converters also reduce many other components of total odd nitrogen (NO_y) to NO, such as peroxyacyl nitrates (PAN; RC(O)O₂NO₂), alkyl nitrates (AN; RONO₂), and nitric acid (HNO₃) which are formed as by-products of photochemical O₃ production (Dunlea et al. 2007; Steinbacher et al. 2007; Winer et al. 1974). In spite of this, heated Mo converters remain in use for NO₂ measurement in some ambient air quality monitoring networks. For more accurate measurements of NO₂, photolytic conversion (Kley and McFarland 1980; Pollack, Lerner and Ryerson, 2010; Ryerson, Williams and Fehsenfeld 2000; Reed et al. 2016b; Sadanaga et al. 2010) of NO₂ to NO followed by CL detection (P-CL) has been recommended (Dunlea et al. 2007; Penkett et al. 2011) over CL monitors equipped with heated Mo catalysts since the former does not suffer from this systematic interference:

(4) $\mathrm{N}{\mathrm{O}}_2+\mathrm{h}\nu \left(\lambda <420\mathrm{n}\mathrm{m}\right)\to \mathrm{N}\mathrm{O}+\mathrm{O}$(4)

Since air contains oxygen (O₂; ~21% by volume), the atomic oxygen generated in reaction (4) reacts with O₂ to form O₃:

(5) ${\mathrm{O}}_2+\mathrm{O}\to {\mathrm{O}}_3$(5)

Table 1 gives an overview of photolytic NO₂ converters that have been reported in the literature.

Table 1. Photolytic NO₂ converters reported in the literature. BLC = blue light converter. n/d = not determined or stated.

Reference	Light source	λ (nm)	Volume (cm^3)	j(NO2) (s^−1)	k[Ox] (s^−1)	Zero artifact (pptv)	CF (1 s)
Kley and McFarland (1980)	High- pressure Xe arc lamp, 300 W	>300	220	0.37	n/d	<500	0.29
Ryerson, Williams, and Fehsenfeld (2000)	Hg arc lamp (50 or 100 W)	350 – 400	36	0.54 or 0.73	n/d	10 – 20	0.40 or 0.53
Pollack, Lerner, and Ryerson (2010)	Hg arc lamp	350 – 400	36	0.927	0.068	15	0.59
(Pollack, Lerner, and Ryerson 2010)	Hamamatsu LEDs	365 ± 10	15	0.993	0.105	8	0.60
Pollack, Lerner, and Ryerson (2010)	Nichia LEDs	365 ± 10	5	8.393	0.883	20	0.90
Sadanaga et al. (2010)	Nichia array of 32 LEDs	385 ± 10	166	n/d	n/d	n/d	0.90
Pollack, Lerner, and Ryerson (2010)	BLC	395 ± 20	115	0.515	0.273	40	0.36
Reed et al. (2016b)	BLC	395	16	0.2 – 0.6	n/d	n/d	0.22 – 0.42 (0.96 s)
Reed et al. (2016b)	BLC (high-power)	395	10	6.5	n/d	n/d	0.93 (0.6 s)
This work	LED arrays	395 ± 18	96	4.2	0.16	300	0.96

An important parameter for these devices is their conversion factor or fraction (CF), given by the ratio of the amount of NO generated, Δ[NO], over the amount of NO₂ entering the converter, [NO₂]₀ (Kley and McFarland 1980):

(6) $CF=\frac{\mathrm{\Delta }\left[\mathrm{N}\mathrm{O}\right]}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}\approx \frac{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0-{\left[\mathrm{N}{\mathrm{O}}_2\right]}_{\mathrm{f}}}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}$(6)

If NO_x is conserved, Δ[NO] equals [NO₂]₀ – [NO₂]_f, where [NO₂]_f is the amount of NO₂ exiting the converter. In practice, CF is a function of the residence time t_res of the sampled gas in the photolysis cell (Ryerson, Williams, and Fehsenfeld 2000):

(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7)

Here, j(NO₂) is the NO₂ photolysis frequency, and k[Ox] denotes a pseudo-first order rate coefficient multiplied by the effective concentration of oxidants which convert some of the photochemically generated NO back to NO₂:

(8) $\mathrm{N}\mathrm{O}+\left[\mathrm{O}\mathrm{x}\right]\to \mathrm{N}{\mathrm{O}}_2$(8)

These oxidants include O₃, the hydro and alkyl peroxyl radicals (HO₂ and RO₂), and unknown species. The magnitudes of j(NO₂) and k[Ox] can be determined from a fit of Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) to a scatter plot of CF against t_res (Kley and McFarland 1980; Pollack, Lerner, and Ryerson 2010).

The quantification of NO₂ in ambient air by P-CL has been extensively validated by inter-comparison with other NO₂ measurements (Fehsenfeld et al. 1990; Fuchs et al. 2010). Though recent converters utilizing light centered in the 360 nm to 395 nm wavelength range provide highly selective NO₂ photolysis, they are still prone to undesired photolysis of species such as HONO, bromine nitrate (BrONO₂), or nitryl chloride (ClNO₂) (Pollack, Lerner, and Ryerson 2010). Furthermore, these devices usually exhibit a small zero offset or artifact (Table 1). In addition, recent studies have reported positive interference in certain photolysis systems from PAN thermal decomposition as a result of poor thermal management (Reed et al. 2016b) and negative interference from photolysis of α,β-dicarbonyls when present at mixing ratios near the parts-per-million (ppmv, 10⁻⁶) mark (Villena et al. 2012).

In this article, we characterize a photolytic NO₂ converter prototype, originally designed for continuous emission monitoring systems (CEMS) (Odame-Ankrah and Rosentreter 2017), to determine its optimum operating conditions for ambient air NO₂ monitoring. The converter uses a proprietary set of chip arrays containing forty ~395 nm light emitting diodes (LEDs) to photodissociate NO₂ to NO, and was developed primarily for selective reduction of NO₂ at operating pressures of typically <200 Torr (Odame-Ankrah and Rosentreter 2017). Here, the performance of the converter was evaluated outside its normal operating range (i.e., at higher pressure deviating the converter from its optimum performance) to intentionally accentuate interferences that, when operated under manufacturer specified conditions, are much reduced. The amounts of NO₂, NO_x, and other components of NO_y entering and exiting the converter were quantified by blue diode laser cavity ring-down spectroscopy (CRDS) (Taha, Odame-Ankrah, and Osthoff 2013). The measurement of nitrogen oxides by CRDS has been extensively validated (including by comparison with CL---see, for example, Figure 4 of Fuchs et al. 2009); the key advantages of CRDS is that mixing ratios are measured by optical absorption such that the instrument response does not need to rely on external calibration. Parameters such as the NO₂ to NO conversion fraction (CF), the NO₂ photolysis frequency (j(NO₂)), the NO oxidation rate (k[Ox]), zero offset, relative humidity (RH) dependence, and the extent of potential interferences from HONO, PAN, isopropyl nitrate, and HNO₃ were determined. The purpose of this manuscript is not to present a device that surpasses existing converters in interference suppression, but to highlight interfering processes that likely exist in all such devices but are rarely quantified. Considerations to improve the performance of the photolytic converter prototype are discussed.

Methods

Photolytic NO₂ converter

The design of the photolytic NO₂ converter has been described elsewhere (Odame-Ankrah and Rosentreter 2017). Briefly, Figure 1 shows a sketch of the photolysis cell which has an internal volume of ~88 cm³ and is comprised of a cylindrical body made of ultraviolet (UV) grade quartz and two windows on opposite ends of the cell. The body of the chamber is wrapped with a highly reflective UV grade material (i.e., aluminum foil, Fabritech) to increase the photolysis of NO₂. The photolysis cell features an outlet tube extending axially along the wall of the chamber and perpendicular to the inlet (Figure 1). This design promotes a cyclonic flow pattern inside the reaction chamber which tapers toward the entrance of the outlet within the reaction chamber. The relatively small inner diameter of the outlet tube (~3.8 mm) and its proximity to one of the light source modules (the distance between the end of the outlet tube and the surface of the distal quartz window is ~5 mm) reduces the time available for secondary chemistry between the points of photodissociation and observation of NO downstream and outside the converter. At its manufacturer-recommended operating condition, i.e., a pressure of ~160 Torr and a flow rate of 600 standard cubic centimeters per minute (sccm), the residence times (calculated assuming plug flow) are ~1.7 s and ~0.05 s within the reaction chamber and outlet tube, respectively. The particular model used in these experiments contained three internal cooling fans: two to cool the LED chips, and one to cool the photolysis cell.

Figure 1. Schematic of the photolytic NO₂ converter. (a) Light source modules comprised of LED chip arrays and cooling elements. (b) Quartz window proximal to inlet. (c) Quartz window distal to inlet. (d) Inlet tubing (tangential to the plane of view). (e) Exit tubing for transport of the photolytic products to the reaction chamber of the chemiluminescence analyzer (not shown).

Gas-phase measurements

The CRDS used in this work was a compact, two-channel instrument with 55 cm long optical cavities called General Nitrogen Oxide Measurement (GNOM; Figure S1e) (Taha, Odame-Ankrah, and Osthoff 2013). Briefly, mixing ratios of NO₂ were quantified by optical absorption at 405 nm at a time resolution of 1 s. Mixing ratios of NO_x (and, by difference, NO) were quantified in parallel by converting NO to NO₂ through the addition of excess O₃ to the inlet (reaction 1) (Fuchs et al. 2009; Odame-Ankrah 2015). Alternatively, the sampling channels were equipped with heated quartz inlets to convert other components of NO_y to NO₂, such as PAN, alkyl nitrates, or HNO₃ (Paul, Furgeson, and Osthoff 2009). Concentrations of PAN, for example, were determined with one CRDS channel operated at room temperature (to measure background NO₂), while the other sampled via a quartz tube heated to 250°C to dissociate PAN to NO₂; PAN was then quantified by difference (Paul and Osthoff 2010). Mixing ratios of NO_y were quantified by thermal dissociation of HONO (to HO and NO) and HNO₃ (to HO and NO₂) at 600°C followed by addition of excess O₃ which oxidizes NO to NO₂ (Wild et al. 2014). In the absence of added O₃, this channel did not yield quantitative data since NO partially oxidized to NO₂.

In addition, mixing ratios of NO, NO₂, or NO_x were monitored using a commercial NO/NO_x CL-analyzer equipped with an internally mounted heated Mo converter (Thermo 42i-TL; Figure S1c), mixing ratios of O₃ using a commercial UV-absorption O₃ monitor (Thermo 49i), and mixing ratios of HONO by a cavity-enhanced absorption spectrometer (CEAS) (Jordan and Osthoff 2020).

The instruments sampled from a common inlet manifold (Figure S1c) with 1/4” (0.635 cm) outer diameter (o.d.) and 3/16” (0.476 cm) inner diameter (i.d.) fluorinated ethylene propylene (FEP) Teflon™ tubing (VWR) and through membrane filters (2 μm pore size, Pall) housed in Teflon™ filter holders (Cole Parmer).

Generation and delivery of gas mixtures

The common inlet was flooded with synthetic, NO_x-free “zero air” delivered from a custom zero air generation system (Figure S1b). Part of this flow was passed through a water bubbler to control the RH, which was measured using a commercial RH probe (VWR) installed inline. For low RH experiments, dry air was delivered from a compressed gas cylinder containing NO_x-free air (Air Liquide, Alphagaz grade). Excess gas was vented into an exhaust line.

Calibration gases were added to the common inlet via a pair of Teflon™ tee fittings (Entegris Fluid Handling) in a similar fashion as shown in Figure 5 of Odame-Ankrah and Osthoff (Odame-Ankrah and Osthoff 2011). One tee was placed inside the main inlet line, and its third arm was connected to the second tee (Figure S1c). One arm of this second tee was connected to the calibration gas stream (generated as described below), while the other was connected to a normally open 2-way valve (Cole-Parmer) whose other end was connected to an exhaust pump to enable rapid on/off switching of the calibration gas stream (Odame-Ankrah and Osthoff 2011).

Mixtures of NO and NO₂ in air were generated using the output of a gas cylinder (100.2 ppmv NO in O₂-free N₂, Scott-Marrin; Figure S1a). This gas cylinder’s 2-stage regulator was connected via 1/8” (0.318 cm) o.d. and 0.060” (1.52 mm) i.d. polyether ether ketone (PEEK™) tubing (VICI Valco) to an all-metal mass flow controller (MFC; MKS Instruments; 20 or 100 sccm capacity). The tubing was constructed from PEEK™ to minimize O₂ contamination. The output of the MFC was combined with a flow of ~50 sccm of O₃ in O₂, delivered via a 50 μm critical orifice and a cylinder back pressure of 20 pounds per square inch (psi; ~1.4 bar). The O₃ was generated by irradiating part of the O₂ flow with a Hg pen-ray lamp in a commercial “O₃ generator” (VWR). The combined flow was passed through a ~ 7.2 m long coil of 1/4” (0.635 cm) o.d. and 3/16” (0.476 cm) i.d. FEP Teflon™ tubing (Figure S1a) to allow reaction (1) to go to completion.

Synthesis and delivery of PAN, HONO, isopropyl nitrate (IPN) and HNO₃ is described in the S.I.

Evaluation of converter performance by CRDS

The CF was determined over a range of cell pressures and residence times, t_res, by choosing appropriate inlet flow restrictions and varying the total flow rate. For most experiments shown in this article, a flow restriction was placed at the tip of each CRDS inlet, such that CRDS cell pressures were in the range of ~140 Torr to ~680 Torr at total sample flow rates of between 0.2 to 2.0 standard liters per minutes (slpm). The converter was placed between the CRDS and the CRDS inlet flow restriction (i.e., in the reduced pressure region) and was connected via 1/4” o.d. Teflon™ tubing. The internal volume of the connecting tubing was ~50 cm³ (length ≈ 2.8 m) for experiments conducted prior to April 2018 (results shown in Figure 3 and S3) and reduced to ~12 cm³ (length ≈ 0.7 m) in all other experiments. The pressure within the converter was monitored using an external pressure transducer (MKS). Interferences were evaluated at a pressure of 420 Torr and flow rate of 0.90 slpm, i.e., a 3.5 s residence time, unless noted otherwise.

The photolytic converter could be manually bypassed (typically once every other minute) using a pair of 2-way Teflon™ valves (Entegris Fluid Handling). When it was inline, GNOM quantified the mixing ratios of NO₂ and NO_x exiting the converter, from which the amount of NO was calculated by difference. When the converter was bypassed, GNOM quantified [NO₂]₀ and [NO_x]₀, i.e., the amounts entering the converter.

For some experiments, the flow restriction was moved between the photolytic converter and the CRDS, and a bypass flow of up to 12 slpm was added to facilitate larger flow rates and shorter photolysis cell residence times.

Box model simulations

Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) neglects potential recombination of the photodissociation products in the tubing connecting the photolysis cell and NO_x analyzer. Most importantly, the O₃ generated by reaction (5) can titrate NO to NO₂ (reaction (1)). This is neglected in the instruments described by Kley and McFarland (Kley and McFarland 1980) and Pollack et al. (Pollack, Lerner, and Ryerson 2010) because their respective photolysis cells and NO analyzers were tightly integrated. However, with the external converter setup, the residence time and chemistry in the connecting tubing are significant. A box model was constructed using the software package Igor PRO (Wavemetrics) to simulate the chemistry in the converter and in the connecting tubing. Details are given in the S.I.

LED emission spectrum and j constants

The LED emission spectrum was recorded using an Ocean-Optics USB-2000 (wavelength range 200 nm–500 nm, 200 μm entrance slit). Photolysis frequencies (referred to as j constants) of NO₂, NO₃, PAN, and HONO were calculated by multiplying the observed emission spectrum, F(λ), with room-temperature absorption cross-sections, σ(λ,T), and quantum yields, ϕ(λ,T), taken from the NASA-JPL evaluation (Burkholder et al. 2015), and integrating over all wavelengths:

(9)

Photolysis frequencies were scaled to j(NO₂), which was determined from a fit to Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) .

Results

Determination of NO₂ conversion fraction

Sample time series and CF as a function of RH

A sample time series of a typical experiment is shown in Figure 2. The CRDS sampled NO_x-free air delivered from a cylinder using an MFC at a flow rate of 1.8 slpm before 18:38, from 18:48 to 18:52, and after 19:02. At 18:49, the make-up gas was humidified with the aid of a bubbler. The top panel shows the RH of the make-up gas, which was continuously monitored. A mixture of NO and NO₂ was added to the gas stream at a flow rate of 100 sccm from 18:18 to 18:48 and from 18:52 to 19:02. After dilution, the NO mixing ratio was in the range of 120 to 130 ppbv, and the NO₂ mixing ratio in the range of 70 to 74 ppbv. The converter was then switched in- and offline approximately every 120 s (indicated by light yellow and gray backgrounds, respectively). The green dashed line indicates the interpolated mixing ratio of NO₂ delivered from the NO_x source during this time, [NO₂]₀.

Figure 2. Sample time series demonstrating the determination of the conversion fraction (CF) and its dependence on relative humidity (RH) and a pressure of 415 ± 1 Torr. The CRDS sampled NO₂-free air delivered from a cylinder using a mass flow controller at a flow rate of 1.8 slpm before 18:38, from 18:48 to 18:52, and after 19:02. From 18:18 to 18:48 and 18:52 to 19:02, a mixture of NO and NO₂ was added to the gas stream at a flow rate of 100 sccm. The converter was switched in- and offline approximately every 120 s (indicated by light yellow and gray backgrounds, respectively). The green dashed line indicates the interpolated mixing ratio of NO₂ delivered from the NO_x source. CF was calculated using Equation (6(6) $CF=\frac{\mathrm{\Delta }\left[\mathrm{N}\mathrm{O}\right]}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}\approx \frac{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0-{\left[\mathrm{N}{\mathrm{O}}_2\right]}_{\mathrm{f}}}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}$(6) ) and is shown on the right axis. At 18:49, the make-up gas was humidified with the aid of a bubbler. The top panel shows the RH of the make-up gas that was continuously monitored using an RH probe.

Since the concentration of NO_x is unaffected by the photolytic converter, the CF can be calculated from the right-hand side of Equation (6)(6) $CF=\frac{\mathrm{\Delta }\left[\mathrm{N}\mathrm{O}\right]}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}\approx \frac{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0-{\left[\mathrm{N}{\mathrm{O}}_2\right]}_{\mathrm{f}}}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}$(6) , i.e., by dividing the difference between [NO₂]₀ and [NO₂]_f by [NO₂]₀. The CF is shown on the right axis. In the example shown, an average CF of 89.7% (±0.2% standard deviation) was calculated at 0% RH, while the CF was (89.1 ± 0.2)% at 93% RH. These RH values cover the extremes found in ambient air; we conclude that CF is not a function of RH.

CF as a function of residence time

The experiment shown in Figure 2 was repeated over a range of pressures and sample flows and with NO_x samples that contained >90% NO₂. A subset of the results is shown in Figure 3a. At low flow rates (and long photolysis cell and connecting tubing residence time), the observed CF was ~70%. As the flow rate is increased and/or the pressure is decreased, the CF increases and approached unity. At a flow rate of 2 slpm and <300 Torr pressure (Figure 3a), the residence time in the converter is too short to fully dissociate NO₂, such that the CF is lower. Optimum conditions (i.e., maximum CF) could be reached at a variety of flow rates and pressures; this optimum region is shown with a gray underlay, and encompasses flows and pressures of commonly used NO analyzers. The CF in this region was evaluated prior to this study and found to be >0.95.

Figure 3. (a) NO₂ conversion efficiency (CF) as a function of photolysis cell pressure and flow rate. The symbols denote different flow restrictions used in the experiments. The area shaded in gray indicates the optimum operating region. The open black circles show typical pressures and flows of NO_x analyzers. T42 = Thermo Scientific 42, AT200 = Advanced Pollution Instrumentation 200, S6040 = Sabio Environmental 6040, ECLD8 = Ecophysics CLD88sp. (b) CF as a function of photolysis cell residence time (data from Figure 3a are shown in red color). The black dashed line is a fit to the data <0.7 s residence time assuming that k[Ox] equals zero. The solid blue line is a fit of Equation (7(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) ) to the data shown in blue color.

A plot of CF against photolysis cell residence time is shown in Figure 3b. Data points corresponding to those shown in Figure 3a are shown as red circles. There are two distinct regimes: At short residence times (data points shown in blue), conversion of NO₂ is photon limited, and CF increases. The solid blue line shows a fit of Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) to these data, which gave j(NO₂) = 4.2 s⁻¹ and k[Ox] = 0.16 s^{^−1}. The dashed black line shows a fit of the data with residence < 0.7 s to Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) , with j(NO₂) = 3.7 s⁻¹ and k[Ox] constrained to 0.0 s⁻¹.

At long residence times (data points shown in red), CF decreased because of secondary chemistry in the tubing connecting the converter and CRDS, namely the relatively slow reaction of the photolysis products NO and O₃ (reactions 1 and 2), which is investigated in more detail below in the section titled “Oxidation of NO to NO₂ in the connecting tubing.”

The intermediate, optimum region with highest CF values is highlighted in gray and reaches from ~1.3 s to ~2.0 s residence time.

Photolysis frequencies

Figure 4 shows the LED emission spectrum, whose peak intensity is at 394.9 nm. The spectrum is slightly asymmetric with higher intensity at longer wavelengths. The parameters of an exponentially modified Gaussian function (Jeansonne and Foley 1991) were fitted to the observed spectrum using the “fExpGauss” function of Igor Pro, which gave the following parameters: Gaussian location = 389.7 ± 0.5 nm, full-width-at-half-maximum (FWHM) = (12.1 ± 1.1) nm (σ = (5.15 ± 0.45) nm), and τ = 10.16 ± 1.04.

Figure 4. Overlap of NO₂, HONO, and NO₃ absorption cross-sections (Burkholder et al., 2015; Voigt et al., 2002) (left axis) and NO₂ quantum yield (ϕ; right axis) with the LED emission spectrum (right-axis).

Table 2 summarizes photolysis frequencies calculated using Equation (9)(9) from the observed LED emission spectrum and absorption cross-sections and quantum yields, scaled to the photolysis frequency of NO₂. Seven molecules have a photolysis frequency >0.01 s⁻¹; beside NO₂, these are Cl₂, NO₃, HONO, biacetyl, glyoxal, and ClNO₂. In ambient air, abundances of Cl₂, NO₃, and ClNO₂ are generally much lower than those of NO₂ and too small to matter in its quantification; the only molecule from which a significant interference can be expected is HONO: at a residence time of 3.5 s, ~40% of this molecule is expected to photodissociate to NO. Table 2 also lists a few molecules that are present in relatively high abundance such as formaldehyde (HCHO), hydrogen peroxide (H₂O₂), PAN, and HNO₃, but whose photolysis frequencies are negligibly small.

Table 2. Photolysis frequencies calculated from the overlap of the LED emission spectrum with absorption cross-sections and quantum yields, scaled to the photolysis frequency of NO_2.

Species	j (s^−1)
NO2	4.2
Cl2	0.21
NO3	0.18
HONO	0.14
biacetyl	0.080
glyoxal	0.030
ClNO2	0.013
HCHO	2.4 × 10^−7
H2O2	9.0 × 10^−14
PAN	3.6 × 10^−15
HNO3	9.5 × 10^−16

Experimental characterization of artifacts and interferences

Interference from nitric oxide

The NO₂ conversion fraction decreases when high concentrations of NO are sampled alongside NO₂, because (a) the high background will speed up conversion of NO to NO₂ by reaction with O₃ in the connecting tubing after photolysis, and (b) some NO will be oxidized to NO₂ in the photolysis cell due to the k[Ox] term. Both effects are investigated in the following sections.

Oxidation of NO to NO₂ in the connecting tubing

Figure 5a shows an experiment in which the CF of ~21 ppbv NO₂ was examined as a function of added NO. The time series demonstrates that the CF (calculated using the right-hand side of Equation (6))(6) $CF=\frac{\mathrm{\Delta }\left[\mathrm{N}\mathrm{O}\right]}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}\approx \frac{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0-{\left[\mathrm{N}{\mathrm{O}}_2\right]}_{\mathrm{f}}}{{\left[\mathrm{N}{\mathrm{O}}_2\right]}_0}$(6) decreased with the amount of NO added. Figure 5b shows the results of box-model simulations (with k[Ox] = 0 s⁻¹) under the conditions of Figure 5a. The simulations predict that CF is a function of NO at the time of observation (gray-shaded area). Figure 5c shows CF as a function of added NO (black circles) and the CF expected from the box model simulations (blue shade). The simulations overestimate the NO yield by 4%, or a factor of ~(j(NO₂)+k[Ox])/j(NO₂). Once this factor is taken into account (red shade), the simulated conversion fractions agree well with the experimentally determined CF values (Figure 5c). The simulations further predict that this effect scales with the residence time in the connecting tubing, i.e., the inverse of the volumetric flow rate.

Figure 5. (a) Determination of the CF for ~21 ppbv NO₂ as a function of NO at 420 Torr and a flow rate of 0.90 slpm. The yellow shaded region indicates times when the converter's LEDs were turned on. (b) CF expected from box model simulations of NO₂ photolysis (yellow shaded region) and subsequent oxidation of NO by the O₃ generated by NO₂ photolysis. In these simulations, k[Ox] was set to 0 s^-1. The area shown in gray brackets the time of observation by CRDS. (c) Comparison of experimental (black circles) and expected (shaded) CF. The blue shaded area was derived from the data shown in Figure 5B. The red shaded area was obtained by multiplying the blue shaded area with 0.96 = j(NO₂)/(j(NO₂)+k[Ox]).

Photo-oxidation of NO to NO₂ in the photolysis cell

The rate of NO oxidation by the unknown oxidant (k[Ox]) was measured directly by sampling a high mixing ratio of NO at low flow rates and long residence times (> 10 s) while the converter was turned on. A subset of the results is shown in Figure 6. The amount of NO₂ produced scales linearly (r² > 0.99) with residence time with a slope of 7 × 10⁻⁴ s⁻¹ (scatter plot not shown). This value is three orders of magnitude below that required to justify the CF shown in Figure 3b and the conversion of NO₂ to NO (Figure 5). This suggests that the unknown oxidant is a by-product of NO₂ photolysis.

Figure 6. Oxidation of NO to NO₂ at very long residence times (shown on top). The converter’s LEDs were on for the entire experiment, while the flow rate was reduced in a step wise fashion to increase the residence time. A constant amount of NO was periodically added to the sampled air and analyzed for its NO₂ content. The observed NO₂:NO ratio (right-hand side) is rationalized using k[Ox] = 7 ± 10⁻⁴ s⁻¹, which is inconsistent with the k[Ox] value derived from eq.

UV-induced artifact

Most photolysis systems exhibit a UV-induced artifact, where a small amount of NO is photochemically produced when sampling NO_x-free air. This converter is no exception, and an example is shown in Figure S2. Here, the CRDS sampled dry zero air from 19:41 to 19:49 through the converter, which was turned on from 19:43 to 19:46. The NO_x signal increased to ~0.32 ppbv during this time and returned to its baseline as soon as the converter’s LEDs were turned off. The magnitude of the UV-induced artifact was variable and depended on sample history.

Interference from NO_z species

There was no interference from IPN (Figure S2). At a converter residence time of ~3.3 s, 60% of HONO and 66% of PAN converted to NO (Figures S2 and S3). We also observed partial conversion (6%) of HNO₃ to NO (Figure S5), though this conversion likely occurred via photo-dissociation of surface-adsorbed nitrate.

Discussion

In this work, potential interferences in a new photolytic NO₂ converter have been evaluated. The converter is operated near the upper end of the wavelength region with unity NO₂ quantum yield. This region is preferred for NO₂ conversion since the conversion efficiency (per photon) is high, most of the LED light intensity falls between the absorption spectra of HONO and of the nitrate radical (NO₃), and unwanted photolysis of, for example, HNO₃, PAN, N₂O₅, HO₂NO₂, methyl nitrate (CH₃ONO₂), and chlorine nitrate (ClONO₂) is avoided (Ryerson, Williams, and Fehsenfeld 2000).

The key distinguishing feature of this converter is its high CF of up to 96% (Table 1). The high CF and large j(NO₂) are advantageous as they enable a large linear dynamic range, which is useful for on-stack CEMS, and allow the unit to be operated at low pressures and high flow rates, which is useful for suppression of artifacts and interferences. The artifacts and interferences elucidated and quantified during the photochemical conversion of NO₂ to NO in this work likely also occur in other photolytic converters operated in the same wavelength region (Table 1) to some degree, such as the widely used “blue light converter” (BLC), but are rarely investigated and disclosed, though there are notable exceptions (e.g., the work by Pollack, Lerner, and Ryerson 2010). Four distinct mechanisms that generate artifacts and interferences were identified (Table 3): direct photolysis, thermal decomposition, surface promoted photochemistry, and secondary chemistry in the connecting tubing. All of these can be remedied to a certain extent by lowering t_res.

Table 3. Measurement artifacts. The extent of thermal dissociation was evaluated at temperatures of 75°C (estimated from the extent of PAN dissociation – see S.I.) and 45°C. obs. = observed. pred. = predicted.

Cause	Species	j or k (s^−1)	Conversion at tres = 3.5 s	Conversion at tres = 0.35 s
Direct photolysis	HONO	0.25 (0.14)	60% (obs.) – 39% (pred.)	8% – 5% (pred.)
Direct photolysis	NO3	0.18	47% (pred.)	6% (pred.)
Thermal dissociation (T = 75°C)	PAN	0.31	66% (obs.; tres = 3.3 s)	10% (pred.)
Thermal dissociation (T = 45°C)	PAN	0.0077	3% (pred.)	0.3% (pred.)
Thermal dissociation (T = 75°C)	N2O5	6.0	>99% (pred.)	88% (pred.)
Thermal dissociation (T = 45°C)	N2O5	0.338	69% (pred.)	11% (pred.)

Direct photolysis

Perhaps the least surprising of these is the direct interference from the photolysis of HONO to NO. From the overlap of the LED emission spectrum with the absorption spectrum of HONO, we calculated a photolysis frequency of 0.14 s⁻¹; experimentally, a somewhat larger j(HONO) of ~0.25 s⁻¹ was determined. The reason for the larger than predicted j(HONO) is unclear, though it may indicate a slightly higher than measured light intensity at lower wavelengths within the photolysis chamber.

With t_res = 3.5 s as tested, a j(HONO) of ~0.25 s⁻¹ implies a HONO to NO conversion rate of ~60%. We calculate that the conversion of HONO to NO is lower at the optimum flow rate (~22%) and could potentially be suppressed to <10% if a 10-fold larger flow rate were used (e.g., by adding a by-pass flow) and if the chamber is operated at a lower pressure yielding a shorter residence time (Table 3).

In ambient air, HONO/NO₂ ratios are usually largest at night, driven by heterogeneous conversion of NO₂ to HONO on surfaces. Actual ratios vary as a complex function of terrain, location, season, and relative humidity, amongst others. For example, Stutz et al. reported HONO/NO₂ ratios of ~0.04:1 over grass in Milan, whereas Wojtal et al. reported average ratios of ~0.30:1 with peaks >1:1 in the marine boundary layer off Saturna Island, BC (Stutz, Alicke, and Neftel 2002; Wojtal, Halla, and McLaren 2011). Such abundances would constitute a significant interference, even at the optimal flow rates and operating pressures.

Interference from HONO has been investigated for other photolytic converters. Inherently, the HONO conversion is less with the 395 nm LEDs than with those at 365 nm (due to the reduced overlap of the HONO absorption cross-section with the emission spectrum); this wavelength is used in the National Oceanic and Atmospheric Administraiton (NOAA) instrument for which the HONO to NO conversion was reported at 37% (Pollack, Lerner, and Ryerson 2010). Sadanaga et al. (2010) reported a 28% conversion efficiency in their converter which is of similar magnitude as what we observed at 1 s residence time. More recently, Reed et al. (2016a) described a technique to quantify HONO by differential photolysis using 385 nm and 395 nm LEDs. The latter converted 18% of HONO at a chamber residence time of 0.96 s, a rate that is consistent with what was observed in this work.

Though not experimentally verified, NO₃ is also expected to be partially converted to NO₂ and NO. The largest NO₃ mixing ratios reported are on the order of several 100 pptv, under the rare condition of large NO₂ and O₃ mixing ratios and small NO₃ sinks (Brown and Stutz 2012). Furthermore, there are well-documented inlet line transmission issues with NO₃ (Fuchs et al. 2008), such that it is unlikely that NO₃ would constitute a significant interference during the measurement of NO₂ in ambient air when measured at a ground site.

Another potential interference is the photolysis of carbonyls such as glyoxal or biacetyl. Photolysis of these molecules produces HO_x (= HO + HO₂) radicals that can feed into secondary chemistry cycles and reduce the CF (Villena et al. 2012). However, the mixing ratios required to achieve measurable effects are usually large (~ ppmv). Though not unrealistic for smog chamber experiments and certain environments such as road tunnels or street canyons, these mixing ratios are well above maximum abundances of glyoxal in ambient air which range from 200 pptv in forests to 1.85 ppbv in Mexico City (Huisman et al. 2008; Volkamer et al. 2005). Tropospheric abundances of biacetyl are less certain, though an abundance of 90 pptv was recently assumed to rationalize PAN mixing ratios in summertime Beijing, China (Zhang et al. 2015). The study by Villena et al. (2012) is the only one which has investigated this chemistry, and more work will be needed to constrain this artifact when quantifying NO₂ mixing ratios in ambient air.

Thermal decomposition

We observed positive interference from TD of PAN (Figure S3). This is qualitatively similar to what was observed by Reed et al. (2016b) who reported that their BLC thermally dissociated PAN to NO₂ caused by heat given off by their LEDs. There are many environments where PAN concentrations can approach or even exceed those of NO₂ (an example is the higher Arctic, Ford et al. 2002), such that interference from TD of PAN can be significant. We observed that decomposition of PAN was effectively accelerated by NO, which is partially generated in the converter, titrating the PA radical and preventing its recombination with NO₂. At 75°C, the lifetime of PAN with respect to thermal dissociation is ~3.2 s. If a 10× faster flow rate is used (i.e., under optimized conditions when coupled to a CL analyzer), ~10% of PAN is expected to dissociate (Table 3), which would constitute a negligible interference under most conditions.

There are other thermally labile components of NO_y that can interfere. The most important one is the nocturnal species N₂O₅, which has been observed in mixing ratios up to 8 ppbv approximately equal to those of NO₂ (Brown et al. 2016), although its abundance is usually insignificant at surface measurement stations (due to faster loss of NO₃ and N₂O₅ at the surface) and during daytime (Osthoff et al. 2017). Its lifetime with respect to TD to NO₂ and NO₃ at 75°C is ~0.16 s, such that it is expected to efficiently decompose. Clearly, improved internal thermal management of the converter unit is desirable. To demonstrate the benefits of a lower chamber temperature, we calculated expected TD conversion rates for PAN and N₂O₅ at 45°C (Table 3). At this temperature and with a short residence time of 0.35 s, 11% of N₂O₅ and <1% of PAN are expected to dissociate.

One important consideration is that the photolysis cell in the prototype evaluated was cooled only by air-circulating fans. A relatively straightforward improvement would be the implementation of a thermoelectric cooling, i.e., active cooling to maintain a constant, near-ambient cell temperature.

Role of connecting tubing

A third type of interference results from secondary chemistry in the connecting tubing between the photolytic converter and NO detector. High mixing ratios of NO and, by inference, O₃, were shown to negatively interfere (Figure 5). Possible solutions are either to eliminate the connecting tubing altogether and integrate the photolytic converter as tightly as possible with the detector or to reduce the residence time by increasing the volumetric flow rate, i.e., lowering the operating pressure.

Surface photochemistry

Lastly, positive interference arises from surface photochemistry on the inner walls of the photolysis chamber. A so-called UV-induced offset has been reported for most photolytic converters when sampling NO_x-free air (Table 1). Our results indicate that wall-adsorbed HNO₃ is a likely culprit contributing to this artifact since it was enhanced after sampling HNO₃ (Figure S5). Ryerson et al. (Ryerson, Williams, and Fehsenfeld 2000) reported that the interference could be reduced by washing the inner walls with liquid water, which is consistent with wall-adsorbed nitrate. In addition, HONO may also play a role in surface-promoted chemistry (Syomin and Finlayson-Pitts 2003). When the HONO source was placed inline (or taken offline), we observed fast rise and fall times (e.g., Figure S3) indicating that the adsorption and desorption kinetics of HONO on the inner walls of the photolysis cell were fast, which in turn suggests that wall-adsorbed HONO did not play a significant role.

The HNO₃ interference test was particularly intriguing, as it hints at complex secondary chemistry. The temporary loss of NO_y in these experiments (Figure S5, 23:01) implies that a component of NO_y condensed on the inner surfaces of either the photolysis chamber or connecting tubing. It then appears that this surface passivated or saturated, as the NO_y signal recovered. Potential molecules that could fit this behavior are NO₃ and N₂O₅, for which wall losses are well documented and which readily convert to nitrate (Fuchs et al. 2008). It is unclear how NO₃ would form in the cell (it is not predicted by kinetic model simulations). One potential pathway is reduction of surface-bound nitrate anion by an electron transfer reaction, though this is speculative, and more work would be needed to confirm this.

The ZA offset reported was larger than those reported in other recent converters (Table 1), though likely because of the sampling history of the particular converter investigated. A lower offset could have been achieved by washing the cell with deionized water (which was not done) and perhaps by lowering the surface to volume ratio of the photolysis cell.

Mystery oxidant

While Equation (7)(7) $CF=\left[\frac{j\left(\mathrm{N}{\mathrm{O}}_2\right){\mathrm{t}}_{\mathrm{r}\mathrm{e}\mathrm{s}}}{j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}+k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]\left[1-e^{-j\left(\mathrm{N}{\mathrm{O}}_2\right)t_{res}-k\left[\mathrm{O}\mathrm{x}\right]t_{res}}\right]$(7) has been used to phenomenologically describe the conversion of NO₂ to NO in a photolytic converter, the “unknown” oxidant term, k[Ox], remains an enigma. The data presented in Figure 6 shows that an oxidant of the required magnitude does not exist in the converter per se but is a direct byproduct of NO₂ photolysis. A plausible candidate for this oxidant is the hydroperoxyl radical (HO₂), which oxidizes NO to NO₂,

(10) $\mathrm{H}{\mathrm{O}}_2+\mathrm{N}\mathrm{O}\to \mathrm{N}{\mathrm{O}}_2+\mathrm{H}\mathrm{O},$(10)

though it is not known how this molecule would be generated or what other compounds (perhaps excited state oxygen species) are involved.

There has been some speculation that electronically excited NO₂ can act as a source of the hydroxyl (HO) and/or the hydroperoxyl radical (HO₂), though this topic has remained controversial (Dillon and Crowley 2018; Li, Matthews, and Sinha 2008). Reactions that have not been fully explored are those between excited state NO₂* and H₂O:

(11) $\mathrm{N}{{\mathrm{O}}_2}^{\ast }+{\mathrm{H}}_2\mathrm{O}\to \mathrm{H}\mathrm{O}+\mathrm{H}\mathrm{O}\mathrm{N}\mathrm{O}$(11)

Our data are not consistent with reaction (11) since we did not observe an RH dependence and the NO_x budget is closed (Figure 2), leaving no room for HONO production.

Any photolytic converter that is operated with ambient air would likely quickly build up an internal surface covered with some water and nitrate. A surface effect would be consistent with the variability of reported k[Ox] between photolysis cells (Table 1). On the other hand, a wall-driven process would have a finite oxidation capacity and likely not scale with NO_x concentration as observed in Figure 6. Clearly, more work is needed to identify the process or processes behind the k[Ox] phenomenon.

Outlook

This work has improved the understanding of interferences in photolytic converters. The high CF exhibited by this prototype, which is higher than other photolytic converters described to date (Table 1), allows the converter to be operated at low pressure and high flow rate to minimize interferences as well as to function when sampling high NO₂ concentrations such as in stack emissions. Current efforts in our laboratories are focused on the construction of a device with improved thermal management and reduced photolytic conversion of HONO.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

This work was supported by the Natural Sciences and Engineering Research Council of Canada [NSERC Engage grant EGP515261-17].