Report

Rate of Gas Phase Association of Hydroxyl Radical and Nitrogen Dioxide

See allHide authors and affiliations

Science  29 Oct 2010:
Vol. 330, Issue 6004, pp. 646-649
DOI: 10.1126/science.1193030

Honing in on HONO2

Modeling air pollution requires knowledge of all the interrelated reactions occurring in the atmosphere. Among the most significant is the formation of nitric acid (HONO2) from OH and NO2 radicals. One sticking point in the study of this reaction has been the uncertainty in how often radicals link through an O-O rather than an O-N bond. Mollner et al. (p. 646) measured the partitioning coefficient, as well as the overall consumption rate of the radicals, with an array of highly sensitive spectroscopic techniques in the laboratory. The measurements yielded a well-defined rate constant for nitric acid formation, which was applied to the prediction of ozone levels in atmospheric simulations of the Los Angeles basin.

Abstract

The reaction of OH and NO2 to form gaseous nitric acid (HONO2) is among the most influential in atmospheric chemistry. Despite its importance, the rate coefficient remains poorly determined under tropospheric conditions because of difficulties in making laboratory rate measurements in air at 760 torr and uncertainties about a secondary channel producing peroxynitrous acid (HOONO). We combined two sensitive laser spectroscopy techniques to measure the overall rate of both channels and the partitioning between them at 25°C and 760 torr. The result is a significantly more precise value of the rate constant for the HONO2 formation channel, 9.2 (±0.4) × 10−12 cm3 molecule−1 s−1 (1 SD) at 760 torr of air, which lies toward the lower end of the previously established range. We demonstrate the impact of the revised value on photochemical model predictions of ozone concentrations in the Los Angeles airshed.

The gas phase three-body association reaction of hydroxyl radical (OH) with nitrogen dioxide (NO2) to form nitric acid (HONO2) (Eq. 1a)OH + NO2 + M → HONO2 + M (1a)(where M is N2 or O2 in air) is important throughout the lower atmosphere. It sequesters reactive hydrogen (HOx, the sum of OH and peroxides) and reactive nitrogen (NOx, the sum of NO2 and NO) as long-lived nitric acid, shortening their atmospheric lifetimes and slowing catalytic cycles responsible for photochemically driven formation of air pollution in the troposphere and ozone depletion in the stratosphere.

The impact of the reaction in Eq. 1a is especially notable in polluted air masses, where its rate affects the formation of pollutants including ozone (O3), nitric acid, and fine particulate nitrate. This reaction leads to lower ozone production at elevated NOx concentrations, because excess NO2 removes OH. Previous sensitivity analyses (1) of photochemical models used to predict O3 from inputs of precursor emissions have shown that the rate constant for this reaction, k1a, ranks among the top normalized sensitivities for all chemical rate constants. Because these models are used to assess the future impact of emission reduction strategies for improving air quality, it is essential that k1a be known to a high accuracy.

Determination of k1a has been a long-standing problem in gas phase kinetics. It is an effective bimolecular rate constant that should depend nonlinearly on pressure according to the Lindemann-Hinshelwood mechanism, but Robertshaw and Smith found the rate constant to be anomalously high at higher pressures (2). They explained this behavior by proposing a second channel to form peroxynitrous acid (HOONO, in Fig. 1), a weakly bound isomer of nitric acid (Eq. 1b): OH + NO2 + M → HOONO + M (1b)HOONO formation is now believed to be a minor channel at low pressures (3), but the branching ratio, α(T, [M]) = k1b/k1a, is expected to increase with pressure, with the two channels predicted to be comparable in the high pressure limit (47). Although HOONO can form initially as one of several isomers, all HOONO molecules rapidly isomerize to the lowest energy cis-cis conformer, which is stabilized by an intramolecular hydrogen bond (8).

Fig. 1

Energy diagram for the reaction OH + NO2 + M. Energies are at 0 K and include zero point energy (3, 29, 30).

Under conditions pertaining in the boundary layer, HOONO will redissociate rapidly to OH and NO2 (9); only nitric acid is stable enough to act as a sink for HOx and NOx. Atmospheric models thus need k1a, but experiments generally measure time-resolved loss of OH in the presence of NO2 and hence determine the total rate constant k1 = k1a + k1b. Deriving the rate constant needed for models, k1a, therefore requires measuring α.

The difficulties in determining k1a arise from two key issues. First, there are technical problems in making laboratory measurements of k1 in air. Most previous measurements detected OH by fluorescence-based methods and suffered from poor signal-to-noise ratios at pressures above a few hundred torr of air, because collisional quenching reduced the OH fluorescence quantum yield (1012). In addition, systematic errors in the measurements of NO2 concentration propagate directly into measurements of k1. This is because the experiments measure the first-order loss of OH in excess NO2. Previous studies in nitrogen near atmospheric pressure differ by almost a factor of two (1012). The first study by Anastasi and Smith (10) at N2 pressures up to 500 torr agrees with the most recent results of D’Ottone et al., who have measured the rate in both air and nitrogen at pressures up to 700 torr (11), but Donahue et al. found substantially lower rates (12). Second, despite a large body of evidence to support the existence of the HOONO channel (3, 4, 1316), there are no quantitative measurements of α at atmospheric temperature and pressure. There have been many efforts to compute k1 and α, and the reactions in Eqs. 1a and 1b have become a test of theoretical kinetics, but reliable first principles predictions are not yet feasible (4, 6).

These uncertainties make it hard for data evaluation panels to recommend a value for k1a for use in atmospheric models. Recommendations for k1a at 760 torr of air and 298 K by the International Union of Pure and Applied Chemistry (IUPAC) (17) and Jet Propulsion Laboratory (JPL) (9) evaluations (Table 1) differ by 14% and have large uncertainties due to the discrepancies among the laboratory results. Because predicted ozone levels in air quality models have high sensitivities to this rate constant, a re-examination of this reaction is necessary.

Table 1

Recommended effective bimolecular rate constants at 760 torr of air and 298 K, except the IUPAC recommendation, which is for 750 torr N2. Reported uncertainties are 1 SD and include both statistical and systematic errors.

View this table:

Here, we report the results of room temperature measurements of k1 at pressures of N2, O2, and air up to 900 torr, and α at pressures up to 750 torr in N2. A fit to both data sets yields parameters that describe k1a over the range 100 to 760 torr. We then use an atmospheric chemistry and transport model to estimate the effect of the new results on simulated ozone levels in the Los Angeles Basin.

The total rate constant k1 was measured with a Pulsed Laser Photolysis–Laser Induced Fluorescence (PLP-LIF) apparatus designed to overcome the experimental difficulties described above (fig. S1) (18). Pseudo first-order ([M] » [NO2] » [OH]0) rate constants were measured in air, N2, and O2 buffer gases over the pressure range 20 to 900 torr at 298 K. We compensated for the reduction in OH sensitivity at higher pressures by probing the OH with a high-repetition-rate (10 kHz) laser. The high repetition rate was coupled with the use of photon counting detection and enhanced photon collection efficiency to further improve the precision of the OH kinetic decay profiles. High-sensitivity OH detection allowed data to be collected above 760 torr while keeping the [OH]0 low (≤1011 cm−3). To better quantify NO2 concentrations, the NO2 was measured directly in the reaction zone by absorption spectroscopy over the wavelength range 410 to 440 nm. NO2 concentrations were obtained by fitting the entire spectral region using cross sections measured recently by Nizkorodov et al. (19) over a wide range of pressures and temperatures relevant to the conditions of the kinetic measurements. The combination of improved OH detection sensitivity and direct NO2 concentration measurements led to kinetics data with high precision and accuracy.

At each pressure, k1 was determined by fitting 10 to 15 measurements of the first-order loss of OH as a function of NO2 concentration. The observed fall-off curves of k1 versus pressure in nitrogen, oxygen, and air from 100 to 900 torr (600 torr for O2) are given in fig. S6. Our results support the higher range of k1 values of D’Ottone et al. (11) for N2 and air near atmospheric pressure, as well as those of Anastasi and Smith (10) for N2, rather than the lower values of Donahue et al. (12). Our measurements in air differ slightly from a weighted linear combination of measurements in O2 and N2 by 3 to 4%, which we attribute to systematic errors. Our best estimate is an average of these two data sets. Our rates at 700 torr are ~10% lower than those of D’Ottone et al. (11) in air, which we attribute to improved sensitivity and to our more accurate NO2 concentration measurements.

Experiments to determine α were performed in a second apparatus using PLP for the generation of radicals and infrared cavity ringdown spectroscopy (CRDS) for the detection of products (fig. S3). The high sensitivity of CRDS enabled detection of the ν1(OH stretch) bands of HOONO and HONO2 products in absorption (3). We measured relative yields of the primary products from the reaction of photolytically generated OH with excess NO2 in the presence of predominantly (>90%) N2 buffer gas at pressures of 100 to 750 torr. Spectra were recorded after a delay of 100 μs, sufficient to allow for complete reaction with NO2. The ratio of product concentrations was determined from the integrated band intensities (after correcting for a small deviation in the HONO2 absorption due to nonlinearities in the CRDS signal) using the ratio of ν1 cross sections obtained from high-level ab initio calculations. These calculations include electrical and mechanical anharmonicities; furthermore, the ratio of intensities is largely invariant with respect to level of theory (3), because systematic errors cancel. This ratio should be accurate to 5% based on the agreement between calculated absolute intensities and experimental results for nitric acid and agreement of calculated frequencies with experimental results for both nitric acid and HOONO (20).

Because HOONO has an internal hydrogen bond, the OH absorption is blue-shifted when torsional vibrations, which break the hydrogen bond, are excited; the thermal population in these hot states is not fully accounted for by the measured HOONO band (21). To determine the fraction of the population that did not contribute to the observed HOONO band, we developed a three-dimensional (3D) potential surface as a function of the OH bond length and the HOON and OONO torsion coordinates (18). We extended our earlier work to include both torsions, as they were found to be strongly coupled near the potential minimum. Using the calculated energy levels, we corrected for contributions to the HOONO intensity absent from the observed band (multiplying our observed intensities by 1.2).

Experiments were performed in N2, because the presence of O2 would have led to the rapid formation of HO2 and spectral interference in the region of interest. We expect the relative collisional efficiency of O2 in the two product channels Eqs. 1a and 1b, to be similar; the presence of O2 would then not significantly alter the ratio α. The N2 pressures for our α data were converted to effective air pressures by assuming the collision efficiency of air to be 94% that of N2 (Fig. 2B).

Fig. 2

Pressure dependence of rate parameters for the reaction (Eqs. 1a and 1b): OH + NO2 + M, M = air at 298 K. Error bars represent 2 SD uncertainty. The curves are derived from fitting the data to four fall-off rate parameters; these are given in table S7. (A) Total effective bimolecular rate coefficient k1 = k1a + k1ba. Measurements (black circles) and fit (solid black line) from this work; previous measurements by D’Ottone et al. (blue triangles) (11); recommendations by IUPAC (dotted red line) (17), and NASA/JPL 2006 (dashed cyan line) (9). (B) Branching ratio α = k1b/k1a. Measurements (black squares) and fit (red line). Buffer gas was predominantly N2; nitrogen pressures were scaled by 1.063 to account for the collisional efficiency of N2 relative to air. (C) Effective bimolecular rate constant k1a for the nitric acid product channel. Fit to the data from this work in (A) and (B) (solid black line); recommendations by IUPAC (dotted red line) (17), NASA/JPL 2006 (dashed cyan line) (9), and NASA/JPL 2000 (dot-dashed purple line) (22). Pressures for the IUPAC recommendation (in N2) were scaled by 1.063 to account for the collisional efficiency of N2 relative to air.

Our measurements of k1 and α were then fit simultaneously to obtain the four rate parameters for the two channels of the reaction Eqs. 1a and 1b (k01a, k1a, k01b, and k1b), using the JPL formulation of the fall-off expression equation S1 (9) (see table S7 for new recommended values). Fits to both data sets were obtained, as shown in Fig. 2, A and B, and figs. S7 and S9. The pressure dependence of k1a at 298 K (Fig. 2C) falls 13 to 23% below the values derived from the current recommendations over the range 100 to 760 torr.

At 760 torr of air, we find a branching ratio of α = 0.142 (±0.012) (1 SD) and a rate coefficient for the formation of nitric acid of k1a (298 K) = 9.2 (±0.4) × 10−12 cm3 molecule−1 s−1 (1 SD), as shown in Table 1. Our value at 1 atm air is 13% lower than the current JPL recommendation and 23% lower than the current IUPAC recommendation (which assumes α = 0); the uncertainty in our measurement is substantially smaller than stated in the evaluations, which had to account for the large variation in k1 among existing data sets. We anticipate that the agreement between our k1 measurements and the higher previously reported values (10, 11), in combination with quantifying the HOONO yield, will lead to a reduction in the k1a uncertainty in future data evaluations. Our measurement agrees with the earlier JPL 2000 Evaluation, which has been used in many photochemical models (22), but the agreement is fortuitous, as this evaluation was strongly influenced by k1 values we now believe to be too low (12) and did not account for reaction 1b.

Do the differences in k1a shown in Table 1 imply changes in air quality predictions that are important at the policy level? To estimate the impact of our measured k1a, we applied a 3D Eulerian photochemical model (18) to predict ozone formation in southern California under summertime conditions. Descriptions of the model, input data, and chemical mechanism are available elsewhere (1, 23, 24). Figure 3 shows the spatial distribution of ozone concentrations and changes relative to the base case (NASA/JPL 2006) when lower (this study) and higher (IUPAC) values of k1a are used. The photochemical model predicts 5 to 7 parts per billion (ppb) higher O3 concentrations when using our value of k1a relative to the base case and 10 to 12 ppb higher O3 concentrations relative to the IUPAC recommendation. In these regions of high O3 concentrations, a 10% decrease in k1a leads to a roughly 10% increase in modeled O3 concentration. The direct effect of reducing k1a is a widespread OH increase, which in turn leads to higher ozone levels. The relative changes in NO2 and HNO3 concentrations are smaller than for OH and vary in sign depending on location. Other measures of photochemical air pollution besides midday O3 concentrations have large sensitivities to k1a and will likewise be affected by any revision to the value of k1a. In particular, the U.S. Environmental Protection Agency’s health-based air quality standard for O3 now considers daily maxima of 8-hour average rather than 1-hour peak concentrations. In the supporting online material, we show that the 8-hour average O3 concentrations are similarly sensitive to k1a (18).

Fig. 3

Spatial distribution of predicted midday (1200 to 1300 hours) summertime ozone concentrations in parts per billion in southern California using emission inventory estimates for 2010. (A) defines k1a as recommended by NASA/JPL 2006 (see Table 1). (B and C) show ozone changes relative to the base case for (B) lower k1a measured in this work and (C) higher k1a recommended by IUPAC. Apart from differences in k1a, all three panels rely on the same chemical mechanism (26) without other adjustments to describe relevant atmospheric chemistry.

A consideration in applying this kind of analysis is that most air pollution models use chemical mechanisms where volatile organic compound (VOC) oxidation steps that are not constrained by laboratory data are sometimes adjusted during mechanism development to improve consistency of the model predictions with O3 measurements and other results of smog chamber experiments [see, for example, Wang et al. (25) for the case of aromatics]. To assess the extent to which a revision of k1a might necessitate changes to other portions of the mechanism, we reran the simulations of all chamber experiments used in the evaluation of the mechanism used to produce Fig. 3 with the new, lower value of k1a indicated by this work (18). Briefly, the reduction in k1a required a corresponding reduction in the magnitude of the chamber radical source incorporated in the model to represent chamber effects when simulating the data (24, 26), but otherwise the modified mechanism simulated the chamber data sufficiently closely to the original mechanism that no revisions are indicated, at least for the VOCs important in the ambient simulations discussed here. As a result, the mechanism we used to produce Fig. 3B is consistent with the chamber data.

We have shown that simulations using our revised value of k1a lead to higher predicted ozone concentrations by as much as 5 to 10 ppb. We can assess the relevance of a change of this magnitude by putting it in the context of human health and emissions control strategies. From a health standpoint, it was recently found that a 10-ppb increase in ozone concentrations leads to a 4% increase in risk of death from respiratory causes (27). For emissions controls, the 2010 sensitivity of peak ozone in southern California to anthropogenic VOC has been estimated to be about 16 metric tons per day per ppb O3 (28). Comparing this to the total estimated anthropogenic VOC emissions in 2010 of 740 metric tons per day, (1) VOC emissions would need to be roughly 10% lower to offset a 5-ppb higher predicted O3 concentration. These two factors alone show that the magnitude of the changes in simulated ozone, caused by a revision of k1a, can lead to small but noticeable differences in the predictions of photochemical models used in assessing the impact of air pollution.

Revisions to the rate of reaction 1 are likely to have an impact beyond that of air pollution. Chemical mechanisms are also used extensively in the interpretation of laboratory kinetics data, atmospheric field data, and satellite measurements for a host of issues, including derivation of rate parameters, the global HOx budget, and stratospheric ozone depletion. For these applications, k1a needs to be known precisely over a wide range of temperatures and pressures (29).

Supporting Online Material

www.sciencemag.org/cgi/content/full/330/6004/646/DC1

Materials and Methods

SOM Text

Figs. S1 to S16

Tables S1 to S7

References

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. This work was supported by National Aeronautics and Space Administration (NASA) grants NAG5-11657, NNG06GD88G, and NNX09AE21G; California Air Resources Board contracts 03-333 and 07-730; National Science Foundation grant CHE-0515627/0848242 (A.B.M.); a NASA Earth Systems Science Fellowship (A.K.M.); and a Department of Defense National Defense Science and Engineering Graduate Fellowship (M.K.S.). Research at JPL was supported by the NASA Upper Atmosphere Research and Tropospheric Chemistry Programs. This work was carried out in part at JPL, California Institute of Technology, under contract with NASA.
View Abstract

Navigate This Article