Amanda L.
Gomez
,
Kevin D.
Easterbrook
,
Nicole M.
Johnson
,
Shanu
Johnson
and
Hans D.
Osthoff
*
Department of Chemistry, University of Calgary, 2500 University Drive N.W., Calgary, T2N 1N4, Alberta, Canada. E-mail: hosthoff@ucalgary.ca
First published on 30th April 2025
The peroxycarboxylic nitric anhydrides (PANs; RC(O)O2NO2 with R ≠ H) are important trace gas constituents of the troposphere. One of the lesser studied molecules of the PAN family is peroxyacrylic nitric anhydride (APAN; CH2CHC(O)O2NO2) which is found in elevated concentration in biomass burning (BB) plumes and downwind from petrochemical plants. In this work, we conducted laboratory and field experiments to constrain the thermal decomposition (TD) rates of APAN in the atmosphere. The TD of APAN was studied in laboratory experiments using a Pyrex reaction coil at temperatures between 295.2 K and 320.7 K as a function of flow rate (i.e., residence time). Gas streams containing APAN were generated from a diffusion source containing a synthetic sample stored in tridecane at water-ice temperature. Nitric oxide (NO) was added to this gas stream to prevent recombination of the TD products. Concentrations of APAN were monitored by gas chromatography with electron capture detection (PAN-GC). The TD rate constant is best described by 10(17.88±0.80)
e−(121.2±4.8) kJ mol−1/(RT) s−1, where R is the universal gas constant, and T is the temperature in kelvin. We report ambient air mixing ratios of peroxyacetic nitric anhydride (PAN), peroxypropionic nitric anhydride (PPN), and APAN measured by PAN-GC at the Calgary Central (Inglewood) air quality station from April 17 to May 31, 2023. From May 16 to May 21, the measurement location was blanketed by a BB plume as judged from co-located observations of fine particulate matter (PM2.5) and carbon monoxide (CO). During this time, mixing ratios as high as 3.4 ppbv (PAN), 455 pptv (PPN), and 220 pptv (APAN) were observed. After sunset, mixing ratios of the PANs decreased with pseudo-first order kinetics, rationalized by a combination of dry deposition and loss by TD.
Environmental significanceThe peroxyacyl nitrates (PANs) are important tropospheric trace gases acting as NOx reservoir species that are removed mainly by thermal decomposition (TD). Peroxyacrylic nitric anhydride (APAN; also called peroxyacryloyl nitrate or vinyl-PAN, CH2=CHC(O)O2NO2) is found in elevated concentration in motor vehicle exhaust, petrochemical, and biomass burning (BB) plumes. Grosjean et al. (1994) studied the TD of APAN but reported Arrhenius parameters at odds with other PAN-type compounds. Here, we report rate constants for APAN's TD measured in laboratory experiments and revised Arrhenius parameters which will allow for more accurate modeling of nitrogen oxides in BB plumes. We also report ambient air measurements of PANs (including APAN) during a BB episode in Calgary and investigated the removal of PANs after sunset. |
Reference | Campaign | Platform | Major source type | Max. APAN (pptv) | APAN:PAN (%) |
---|---|---|---|---|---|
8 | RISOTTO 1999–2000 | Ground | n/d | 65 | 4 |
9 | TexAQS 2000 | Ground | O&G | 502 | 30 |
10 | SENEX 2013 | Aircraft | BB | 26 | 5.3 ± 0.7 |
6 | OPECE 2018 | Ground | O&G | >400 | 1–4 |
7 | SUNVEx 2021 | Ground | BB | 80 | 3.7 ± 0.1 |
11 | KORUS-AQ 2016 | Aircraft | O&G | 850 | 14.2 |
This work | FOOBAR 2023 | Ground | BB | 220 | 4–8 |
The PANs are primarily formed as by-products of photochemical ozone (O3) and smog production, i.e., from the photooxidation of volatile organic compounds (VOCs) in the presence of NOx (= NO + NO2), and are often associated with biomass burning (BB) plumes.7,12 Aldehydes are usually the most common VOC precursors of PANs. For example, PAN is mainly derived from the hydroxyl radical (OH) initiated oxidation of acetaldehyde, and the unsaturated APAN and MPAN are primarily derived from oxidation of acrolein and methacrolein, respectively, e.g.,
![]() | (R1) |
Subsequent reaction of the acrolein peroxyacyl (APA) radical with NO2 yields APAN:
CH2![]() ![]() | (R2) |
Acrolein stems mainly from the photooxidation of 1,3-butadiene, though also has primary sources such as automobile exhaust;13 acrolein and 1,3-butadiene have also been associated with emissions from petrochemical industries.11 In contrast, methacrolein is mainly derived from isoprene, which is predominantly biogenic in origin. Consequently, the relative abundances of MPAN and APAN provide insight into whether anthropogenic or biogenic hydrocarbons dominated the O3 production upwind.4
The PANs are important molecules for two main reasons: for one, because their production consumes and their decomposition releases nitrogen dioxide (NO2), the PANs can act as reservoir species of NOx and odd oxygen (Ox = O3 + NO2) and can transport NOx and Ox from polluted urban air to remote regions where nitrogen oxides are scarce. In this way, the PANs contribute to the redistribution of NOx and, by extension, affect the rates of tropospheric O3 and secondary organic aerosol (SOA) production throughout the troposphere.14,15 The second reason why PANs are important molecules is that the PANs are noxious in high concentration. For instance, PAN has been shown to be phytotoxic,16 and PAN and PPN are known lachrymators.17 The toxicity of the other PANs has not been studied though may be inferred from the relative toxicity of their precursors in the atmosphere. For example, given acrolein's higher toxicity relative to that of acetaldehyde,18 APAN likely surpasses the potency of PAN (at least on a per-molecule basis).
In the lower troposphere, the removal of PANs is usually driven by thermal decomposition (TD; (R-2)), for example:
![]() | (R-2) |
To correctly model the abundances of PANs and their impact in the troposphere, accurate knowledge of their TD kinetics is needed. Table 2 summarizes measured Arrhenius parameters for the TD of PAN, PPN, MPAN, and APAN. While the TD of PAN has been the subject of numerous studies, there have only been two studies of MPAN and only a single study of APAN (prior to this work). Grosjean et al.31,36 investigated the TD of both compounds at T between 285 K and 300 K using samples generated in situ in a Teflon chamber. They reported activation energies that are quite different from the other PANs (in part due to a relatively narrow T range) and appear specious.9 Furthermore, the MPAN data are inconsistent with measurements at T > 302 K by Roberts and Bertman.27 Hence, corroboration of the TD parameters by Grosjean et al.31,36 is warranted, especially for APAN for which no data at T > 300 K exist.
Molecule | A (s−1) | E a (kJ mol−1) | k −2 (298 K) (10−4 s−1) | Reference |
---|---|---|---|---|
PAN | — | — | 2.8 ± 0.8 | 23 |
PAN | (0.8 ± 1.1) × 1015 | 104 ± 3 | 4.7 | 24 |
PAN | (1.9 ± 2.7) × 1016 | 113 ± 4 | 3.7 ± 0.4 | 25 |
PAN | 1018 | 121 ± 8 | 5.4 | 26 |
PAN | 3.16 × 1016 | 113 | 4.9 | 27 |
PAN | 2.52 × 1016 | 112.85 | 4.17 | 28 |
PAN | 1.21 × 1017 | 101 ± 4 | 4.12 ± 0.25 | 29 |
PAN | — | — | 3.1 | 30 |
PAN | (2.5 ± 2.3) × 1017 | 119 ± 2 | 3.1 | 27 |
PAN | (1.6 ± 5.8) × 1016 | 113 ± 9 | 3.0 | 31 |
PAN | (0.5 ± 1.3) × 1016 | 109 ± 6 | 4.2 | 32 |
PAN | — | — | 4.40 ± 0.13 | 33 |
PPN | 2 × 1015 | 106 | 4.4 | 34 |
PPN | (1.6 ± 4.7) × 1025 | 164 ± 7 | 3.4 | 31 |
PPN | 7.2 × 1016 | 116 ± 2 | 3.5 | 35 |
PPN | — | — | 3.67 ± 0.10 | 33 |
Fur-PAN | (3.7 ± 0.2) × 1016 | 113.6 ± 4.2 | 4.6 | 7 |
MPAN | (1.6 ± 2.6) × 1016 | 112 ± 4 | 3.5 | 27 |
MPAN | (1.6 ± 6.7) × 1023 | 153 ± 10 | 2.3 | 36 |
APAN | (0.3 ± 4.1) × 1020 | 131 ± 31 | 3.0 | 31 |
APAN | (0.75 ± 1.39) × 1018 | 121.2 ± 4.8 | 4.3 | This work |
In this work, we studied the TD of APAN using a T-controlled flow reactor coupled to a gas chromatograph with an electron capture detector (PAN-GC) to monitor APAN concentrations. We report TD rate constants of APAN at T between 295.2 K and 320.7 K and at a pressure of (662 ± 11) torr and calculate Arrhenius parameters. To corroborate our results, we measured the TD of PAN at 313.2 K for which literature data are available.27
We also present ambient air measurements of PAN, PPN and APAN mixing ratios which were made at the Calgary Central (Inglewood) air quality station from April 17 to May 31, 2023, as part of the “Focus on nitrOgen diOxide at the Bird sAnctuaRy” (FOOBAR) campaign. From May 16 to May 21, elevated concentrations of carbon monoxide (CO), fine particulate matter (PM2.5), and PANs were observed, associated with BB emissions that originated from several out-of-control wildfires to the north-northwest (NNW) of the measurement location. On several occasions, mixing ratios of PAN, PPN, and APAN decreased exponentially after sunset, allowing for an assessment of APAN loss rates in ambient air with the revised TD rates obtained in this work.
To corroborate the identity of APAN using relative elution times, gas streams containing PAN, PPN or PiBN were generated from photolysis at 285 nm of acetone, diethyl ketone, or diisopropyl ketone in the presence of NOx as described previously.39
CH2![]() ![]() | (R3) |
The value of k3 has not been experimentally determined for the reaction of NO with the APA radical but is likely of similar magnitude as that for NO + PA. Using the rate coefficient reported by Villalta and Howard43 for reaction of the PA radical and NO, we calculate a PA and APA radical lifetime of <1 ms with respect to (R3) under our experimental conditions, i.e., the re-formation of PANs via(R2) is calculated to be negligible.
The radical generated in (R3) dissociates to carbon dioxide (CO2) and vinyl radical (CH2CH), which will react with O2 to form formaldehyde (HCHO) and formyl radical (HCO),44 which in turn reacts with O2 to form carbon monoxide (CO) and hydroperoxyl radical (HO2).45 Under our experimental conditions, HO2 is expected to be titrated by NO to form OH, which in turn is expected to react with NO or the inner walls of reaction vessel (rather than adding to the double bond of APAN and speeding up its decomposition).
Kinetic measurements were conducted using a reaction vessel constructed from ∼1 cm outer diameter (o.d.) and ∼0.8 cm inner diameter (i.d.) Pyrex glass that was double-coiled (12 revolutions downwards in an inner coil and another 12 revolutions on the outside back to the top) and submerged in a T-controlled water bath (Lauda Proline RP 1290, ±0.01 °C). The internal volume of this reaction vessel was (234 ± 2) mL, determined by measuring the volume of water needed to fill it. When not in use (e.g., overnight), the reaction vessel and connecting tubing were continuously purged by a flow of ∼5 sccm of zero air. Thermal decomposition in the connecting tubing outside and downstream of the heated vessel (internal volume ∼5 cm3) was estimated at ∼0.1% and hence negligible.
The vessel effluent was combined with an always-on flow of 50 to 55 sccm of dry N2 and analyzed by PAN-GC. The line pressure was monitored using a transducer (MKS Baratron 722A) mounted on the overflow vent line (not shown) and ranged from 650 torr to 671 torr (855 to 883 hPa).
Concentrations of PANs were monitored using a customized Hewlett-Packard (HP) 5890 PAN-GC described previously.42 This GC was equipped with a 50 μL stainless steel sample loop and a 15 m long megabore analytical column (Restek RTX-200) with a film thickness of 1 μm. The GC was operated with N2 carrier and make-up gas delivered from the “blow-off” of a liquid N2 dewar. Injections were automated (and usually every 5 min). At a column flow rate of 15 mL min−1, PAN, PPN and PiBN eluted at 110 s, 181 s, and 296 s, respectively, whereas APAN eluted at 163 s (Fig. S1A†). These relative elution order and times are consistent with those reported by Roberts et al.9 who also used an RTX-200 column and with values by Tokarek et al.40 who utilized a marginally more polar RTX-1701 column. Chromatographic peak areas were determined by fitting the parameters of a Gaussian expression to the observed peaks as described by Tokarek et al.40
Concentrations of APAN (or PAN) entering (C0) and exiting (Ct) the reaction vessel were measured by manually bypassing the gas flow with the aid of 2-way valves (Swagelok PFA-43S4). The residence time (tres), calculated by dividing the reactor volume by the total volumetric flow rate, was then systemically varied (Table S1†). In a typical experiment (Fig. 2), the reaction vessel was initially bypassed, and the PAN under investigation was eluted from the diffusion source. Once a stable (or only slowly changing) peak area (i.e., concentration, C0) was observed by the PAN-GC, the reaction coil was switched in-line to measure Ct and bypassed again after consecutive GC injections showed stable peak areas. Values of Ct/C0 were calculated by dividing the peak areas observed with the reactor in-line (shown as red squares in Fig. 2) by a linear interpolation of the peak areas when the reactor was bypassed (shown as a blue line in Fig. 2). In the example shown in Fig. 2, Ct/C0 equaled 0.544 ± 0.014, where the error is 1σ precision. The apparent loss rate constant (k′) in the reaction coil was then obtained by linear regression from plots of versus tres.
Pyrex surfaces can destroy PANs, particularly at low humidity,27,46 which adds to k′:
k′ = k−2 + kPyrex. | (1) |
For experiments at T < 300 K, TD rates were slow, necessitating long run times that made it challenging to maintain stable diffusion source outputs. In those cases, k′ was calculated by dividing by tres for individual experiments and averaging the results at each T. For experiments at T > 300 K, k′ values were calculated from the slopes of plots of
versus tres. At higher T, both approaches gave equivalent results (Table 3), with the second method being more precise.
Reference | T (K) | n | k′a (10−5 s−1) | k′b (10−5 s−1) | r | k −2 (10−5 s−1) |
---|---|---|---|---|---|---|
a Average and standard deviation of k−2 values calculated in individual experiments (Table S1).
b Derived from linear fits of ![]() |
||||||
This work | 320.7 | 6 | 1336 ± 20 | 1334 ± 9 | 99.98% | 1306 ± 9 |
This work | 318.2 | 6 | 968 ± 53 | 958 ± 17 | 99.8% | 930 ± 17 |
This work | 315.2 | 6 | 674 ± 67 | 680 ± 20 | 99.6% | 652 ± 20 |
This work | 313.2 | 7 | 534 ± 50 | 527 ± 14 | 99.6% | 499 ± 14 |
This work | 310.7 | 8 | 364 ± 58 | 352 ± 20 | 97.7% | 324 ± 20 |
This work | 308.2 | 8 | 230 ± 37 | 228 ± 13 | 97.7% | 200 ± 13 |
This work | 303.2 | 6 | 137 ± 30 | 130 ± 15 | 94.8% | 102 ± 15 |
This work | 295.2 | 2 | 63 ± 2 | n/d | n/d | n/d |
This workd | 295.2 | 4 | 51 ± 4 | 51 ± 2 | 99.6% | n/d |
Grosjean et al.31 | 299.1 | 1 | 67.0 ± 12.9 | |||
Grosjean et al.31 | 297.8 | 1 | 29.9 ± 2.5 | |||
Grosjean et al.31 | 293.2 | 1 | 8.3 ± 0.2 | |||
Grosjean et al.31 | 293.2 | 1 | 8.3 ± 0.2 | |||
Grosjean et al.31 | 290.2 | 1 | 6.5 ± 0.5 | |||
Grosjean et al.31 | 290.2 | 1 | 7.7 ± 0.4 |
To corroborate the results with APAN, we measured the TD rate of PAN at T = 313.2 K and compared to literature values. Complete lists of experiments conducted for APAN and PAN are shown in Tables S1 and S2,† respectively.
Rates for TD reactions are commonly shown in the form of an Arrhenius plot, i.e., of ln(k−2) versus 1/T. In this work, the k′ data were fit to the parameters of a rearranged Arrhenius expression:
![]() | (2) |
Auxiliary measurements included mixing ratios of NO2 and ΣPAN by TD-CRDS,47 NO, total odd nitrogen (NOy) and O3 using commercial NO–O3 chemiluminescence and absorption analyzers (Thermo Scientific 42i-Y and 49i), and meteorological data collected using a commercial sensor (Vaisala WXT520). All instruments sampled from the same height, ∼4.5 m above ground.
The instruments were housed in a mobile laboratory parked adjacent to the Calgary Central (Inglewood) air quality station located at 51° 1′ 47.7840′′ North Latitude and 114° 0′ 29.1996′′ West Longitude and operated under the guidance of the Calgary Region Airshed Zone (CRAZ), a non-profit association with members from government agencies (federal, provincial, and municipal), non-government organizations, industry, and the public. The Inglewood station houses several continuously operated instruments, including a commercial carbon monoxide (CO) infrared analyzer (Thermo Scientific 48i) and a real-time particulate monitor (Thermo Scientific 5030 SHARP) quantifying sub-2.5 μm aerosol mass (PM2.5).
![]() | ||
Fig. 3 Plots of ln(Ct/C0) for APAN versus tres (i.e., reaction time) as a function of reactor temperature. Error bars are at the 1σ precision level. |
To probe whether the Pyrex walls contributed to the apparent APAN loss in our experiments, experiments were conducted at 295.2 K using humidified flows and with an internally Teflon-coated flow reactor. These experiments yielded lower k′ values than those obtained with the reaction coil and the dry zero air, (51 ± 4) × 10−5 s−1versus (63 ± 2) × 10−5 s−1 (Table 3), confirming that APAN indeed is partially scrubbed on the inner walls of the Pyrex reactor. A lower limit for kPyrex of ≥1.2 × 10−4 s−1 was calculated from the difference of the value with and without Teflon. This value is a lower limit as humidification and the Teflon coating may not fully eliminate wall losses.
The reactor wall loss contributes to all experiments but will have a lesser impact at the higher T where it is not as fast as TD and hence TD dominates. With the assumption that kPyrex is not T-dependent, we estimated its magnitude from r values of plots of ln(k′ − kPyrex) versus in the 303.2–320.7 K T range as a function of kPyrex (not shown). A maximum r value of 0.996 was observed at kPyrex = 2.8 × 10−4 s−1. This value was subtracted from all k′ values to calculate k−2; the results are summarized in Table 3 under the column heading k−2c. These values are plotted in Fig. 4 and were fitted to eqn (1), with the results of the fit shown in Table 2. By extrapolation, a loss rate constant of k−2 = 4.3 × 10−4 s−1 at T = 298 K was calculated, in excellent agreement with k−2 (298 K) of other PANs (Table 2). In contrast, the same analysis with assumed kPyrex = 1.2 × 10−4 s−1 gave k−2 (298 K) of 5.1 × 10−4 s−1, which would be inconsistent with other PANs.
![]() | ||
Fig. 4 Arrhenius plot of k−2 (blue circles) versus 1/T for APAN. Error bars for the laboratory data shown are ±1.4 × 10−4 s−1 (½ of the magnitude of kPyrex). Data from Grosjean et al.31 and pseudo-first order loss rate constants derived from ambient air data are shown as red squares and green triangles, respectively. The grey bar indicates the range of dry deposition loss rates for PAN reported by Roberts.2 |
For PAN, we measured k′ = (3.3 ± 0.1) × 10−3 s−1 at T = 313.2 K (Table S2 and Fig. S2†). Subtracting kPyrex = 2.8 × 10−4 s−1 yields k−2 = (3.0 ± 0.1) × 10−3 s−1, in excellent agreement with the value of (2.9 ± 0.5) × 10−3 s−1 reported by Roberts27 at that T.
Mixing ratios of PAN, PPN, and APAN were elevated during this event, peaking at 3.4 ppbv, 455 pptv, and 220 pptv, respectively. These data are amongst the highest mixing ratios of BB-generated APAN reported at a surface site to date (Table 1) and are the largest mixing ratios measured by this research group.40 Prior to (and after) the BB event, APAN mixing ratios were <10 pptv. In the plume, the APAN:PAN ratio ranged from 4% to 8% (Fig. S4†), which is a larger ratio than reported in other BB studies though smaller than APAN:PAN ratios associated with industrial sources (Table 1). Average ratios of PPN to PAN and APAN to PAN mixing ratios were 12.5% and 6.0%, respectively (Fig. S4†).
In addition to the period shown in Fig. 5, BB smoke was observed on the morning of May 22 as well as from May 24 to May 28, 2023 (data not shown). During those times, PAN, PPN, and APAN mixing ratios peaked at 1.15 ppbv, 200 pptv, and 40 pptv, respectively, and CO and PM2.5 peaked at 602 ppbv and 47 μg m−3, respectively.
NO2 + O3 → NO3 + O2, k4 | (R4) |
Quantity\Period | May 17 22:00–midnight | May 18 01:40–02:40 | May 21 01:00–04:00 | May 25–May 26 23:30–02:50 |
---|---|---|---|---|
a Calculated based on values by Kabir et al.33 b Based on laboratory experiments in this work. | ||||
PAN0 (pptv) | 815 | 511 | 1222 | 762 |
PPN0 (pptv) | 106 | 69 | 174 | 113 |
APAN0 (pptv) | 47 | 24 | 89 | 22 |
NO (pptv) | <LOQ | <LOQ | <LOQ | <LOQ |
NO2 (ppbv) | 14.1 ± 3.5 | 7.0 ± 0.5 | 10.4 ± 1.4 | 16.8 ± 3.8 |
NOy (ppbv) | 16.5 ± 3.5 | 9.1 ± 0.5 | 16.1 ± 1.5 | 18.9 ± 3.8 |
O3 (ppbv) | 31.3 ± 8.0 | 29.5 ± 4.5 | 20.4 ± 5.6 | 18.3 ± 5.9 |
P(NO3) (ppb h−1) | 0.77 ± 0.10 | 0.35 ± 0.06 | 0.41 ± 0.13 | 0.47 ± 0.15 |
RH (%) | 44 ± 10 | 52 ± 5 | 64 ± 8 | 75 ± 6 |
CO (ppbv) | 436 ± 36 | 256 ± 5 | 943 ± 26 | 409 ± 18 |
PM2.5 (μg m−3) | 33.4 ± 2.3 | 21.0 ± 2.6 | 164.3 ± 3.7 | 23.2 ± 1.3 |
T (°C) | 12.0 ± 2.0 | 8.0 ± 1.1 | 13.2 ± 1.3 | 9.1 ± 1.2 |
k −2 (PAN)a | 5.6 × 10−5 s−1 | 2.9 × 10−5 s−1 | 6.9 × 10−5 s−1 | 3.5 × 10−5 s−1 |
k −2 (PPN)a | 4.7 × 10−5 s−1 | 2.4 × 10−5 s−1 | 5.7 × 10−5 s−1 | 2.9 × 10−5 s−1 |
k −2 (APAN)b | 4.7 × 10−5 s−1 | 2.3 × 10−5 s−1 | 5.9 × 10−5 s−1 | 2.8 × 10−5 s−1 |
![]() |
(3.5 ± 0.4) × 10−5 s−1 | (6.7 ± 1.0) × 10−5 s−1 | (4.6 ± 0.9) × 10−5 s−1 | (3.5 ± 0.5) × 10−5 s−1 |
r | 0.948 | 0.968 | 0.785 | 0.853 |
![]() |
(3.6 ± 0.4) × 10−5 s−1 | (7.0 ± 0.2) × 10−5 s−1 | (3.8 ± 0.8) × 10−5 s−1 | (3.2 ± 0.5) × 10−5 s−1 |
r | 0.939 | 0.9991 | 0.780 | 0.862 |
![]() |
(3.7 ± 0.7) × 10−5 s−1 | (6.2 ± 0.7) × 10−5 s−1 | (4.1 ± 0.9) × 10−5 s−1 | (3.3 ± 1.7) × 10−5 s−1 |
r | 0.862 | 0.981 | 0.767 | 0.485 |
![]() |
(4.8 ± 0.2) × 10−5 s−1 | (9.8 ± 0.6) × 10−5 s−1 | (4.3 ± 0.3) × 10−5 s−1 | (4.2 ± 0.2) × 10−5 s−1 |
For the selected cases, plots of ln(C0/Ct) versus time were linear as judged from their r values (Fig. S5† and Table 4). Further, PAN, PPN, and APAN decayed in unison (Fig. 5) with statistically identical values of their apparent loss rate coefficients during each episode (Table 4).
The values for PAN and PPN were compared to k−2 values calculated using the parameterization by Kabir et al.33 The
values for APAN are superimposed in Fig. 4 for comparison with k−2 predicted from the laboratory parameterization determined in this work. In all cases, the
values were of the same order of magnitude as k−2. The apparent loss rates were ∼1/3 slower than the predicted k−2 on May 17 and May 21 and equal to the k−2 on May 26. For the May 18 data, however, the rate constants derived from the scatter plot shown in Fig. S5† were three times as large as the predicted k−2 value.
Mixing ratios of odd oxygen (Ox = O3 + NO2) also decreased with first-order exponential kinetics and rate constants between 4.2 × 10−5 and 9.8 × 10−5 (Table 4), factors between 0.9 and 1.5 larger than those determined for the PANs. Mixing ratios of Ox were chosen (as opposed to O3) to account for titration of O3 by NO to NO2.
However, there is a fundamental difference between the two studies: the experimental approach. Grosjean et al.31 utilized a Teflon smog chamber and generated APAN in situ using sunlight, acrolein, and NO, whereas we utilized a glass reaction coil and an APAN diffusion source. Following APAN generation, their chamber would have contained a considerable amount of NO2, which increases the effective lifetime of APAN (eqn (3)) as the TD fragments would have had a reasonable chance to recombine rather than react with NO. The apparent PAN loss rate constant under these conditions is
![]() | (3) |
In contrast to the smog chamber studies by Grosjean et al.,31,36 the study by Roberts and Bertman27 and this work utilized a Pyrex reactor to dissociate PANs. These experimental setups suffer from a potential bias: We observed that APAN decomposes faster on dry Pyrex surfaces than on either humidified or Teflon-coated Pyrex, necessitating a correction by kPyrex = 2.8 × 10−4 s−1, assumed to be T-independent. An outstanding question is why the decomposition of APAN is faster on dry than on humidified Pyrex as has been reported for other PANs.27,46 Theoretical studies, e.g., by density functional theory,52 are needed to probe and rationalize the role of water in the TD of PANs on Pyrex surfaces.
The choice of kPyrex limits the accuracy of our study: For example, if we assume a value of kPyrex = 1.2 × 10−4 s−1 (the minimum value at T = 295.2 K), we would have obtained k−2 = 10(16.9±0.8)e−(115.1±4.6) kJ mol−1/(RT) s−1 and k−2 (298 K) = 5.1 × 10−4 s−1 which are quite different from the values calculated with kPyrex = 2.8 × 10−4 s−1 that are shown in Table 2. Even so, our kinetic data are corroborated by observations made with heated instrument inlets. For example, Veres and Roberts37 reported APAN to have TD kinetics of similar magnitude as those of the other PANs. Overall, a consistent picture emerges: The TD parameters for all PANs fall in the ranges of 106 kJ mol−1 < Ea < 122 kJ mol−1 and 2.8 × 10−4 s−1 < k−2 (298 K) < 4.6 × 10−4 s−1 (if studies published before the year 1980 are omitted). This makes sense chemically, as the TD rate is mainly determined by the strength of the –O–NO2 bond, and the nature of the side chain would be of lesser importance. Only the values reported by Grosjean et al.31,36 fall outside those ranges (Table 2).
As such, the ambient air observations provided a fortuitous opportunity to evaluate the revised parameterization for TD of APAN from this paper's laboratory experiments under “real-world” conditions. During the periods selected, the mixing ratios of PAN, PPN and APAN decreased with identical pseudo-first order rate constants, . Further, the observed
values appear to agree (within error limits) with k−2 values predicted using laboratory parameterizations (Table 4 and Fig. 4). For APAN, this parameterization was extrapolated from the T range of our lab experiments (295.2–320.7 K) to that of the ambient air data (280–287 K).
At first glance, this seems like an excellent result. However, because of these relatively cool temperatures, the lifetimes of the PANs with respect to TD was relatively long, making it possible for other loss pathways to dominate. Further, for TD to dominate the atmospheric removal loss of PANs, the generated PA radicals need to be removed, and quickly. For the four ambient air periods selected, the NO mixing ratios were below our ability to quantify (i.e., <50 pptv), in contrast to the laboratory experiments (Section 3.1), where kloss (PA) was fast because NO was deliberately added in high concentration. It is unclear which reaction(s), if any, other than (R3) would have removed PA radicals and to what extent. Potential candidates for PA radical removal include aerosol uptake,53 and reaction with NO3 (ref. 54) or with organic peroxy radicals, XO2.22,55 Villalta et al.53 reported a relatively large uptake coefficient (γ) of 0.0043 for uptake of CH3C(O)O2 radicals on deionized water at 274 K, and the aerosol surfaces in the BB plume, while not quantified, were likely large. Thus, aerosol uptake of PA radicals was likely their largest sink. Still, one would expect that reaction of PA radicals with NO2(R2) would have increased the apparent lifetime of the PANs (eqn (3)), i.e., the apparent loss rate constitutes a lower limit to k−2. Perhaps not surprisingly then, the field data (shown as green triangles in Fig. 4) tend to fall below the linear fit of the TD rate extrapolated from higher T data.
Another assumption of this analysis is that nocturnal sources of PANs such as reactions of NO3 with aldehydes54 were negligible. Neither mixing ratios of NO3 nor those of VOCs (from which NO3 abundances could be constrained) were quantified during FOOBAR. Nevertheless, the instantaneous NO3 production rate, P(NO3), was large during the selected periods, 0.35 to 0.77 ppbv h−1 (Table 4), which is within the range of 0.1 to 1.5 ppbv h−1 reported by Decker et al.56 who observed nocturnal BB plumes from an aircraft platform. Thus, all of the selected periods were conducive to NO3 formation. However, Decker et al.56 also noted exceptionally large total NO3 reactivity toward BB-VOCs of between 17 and 70 s−1 that only gradually decreased during the several hours transport time between the emission source and observation point. Large NO3 sinks were likely also present during the BB events in this data set, implying that NO3 mixing ratios would have been small during those times. How much chemistry went through the NO3 + aldehyde channel (which can yield PANs) during FOOBAR is unclear.
Finally, the analysis assumes that direct sinks of PANs (other than TD followed by removal of the PA radicals) such as aerosol uptake57 and dry deposition20–22 were also negligibly small. Lifetimes of PAN with respect to dry deposition have been estimated to be in the 5–10 h range,2 which translate to a loss rate constant of between 5.6 × 10−5 s−1 to 2.8 × 10−5 s−1, which is on par with the values determined in this work (Table 4). In this context, the Ox loss constants,
, are informative since deposition velocities (vd) for O3 are of similar magnitude (with a range of 0.1 to 0.7 cm s−1)58 as those of PAN, for which vd values between 0.13 and 0.54 cm s−1 have been reported.2 The values of
are slightly larger than
of the PANs, which supports the notion that both O3 and PANs are removed mainly by dry deposition. However, Decker et al.56 reported a total O3 reactivity toward BB-VOCs of up to ∼6 × 10−4 s−1 which rapidly decreased with plume age to ∼2 × 10−4 s−1, about a factor of two to four larger than
in this work. In other words, the observed Ox loss could have entirely been due to chemical reactions as opposed to dry deposition; in the absence of speciated VOC, it is not possible to tell. For APAN, the implication is that the nocturnal losses are explained through a combination of dry deposition and TD.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ea00032g |
This journal is © The Royal Society of Chemistry 2025 |