Temperature-dependent kinetics study of the reactions of O(1D2) with N2 and O2

A laser flash photolysis–resonance fluorescence technique has been employed to investigate the kinetics of the reactions of electronically excited oxygen atoms, O(1D2), with N2 (k1) and O2 (k2) as a function of temperature (197–427 K) in helium buffer gas at pressures of 11–40 Torr.

The results are well-described by the following Arrhenius expressions (units are 10−11 cm3 molecule−1 s−1): k1(T) = (1.99 ± 0.06) exp{(145 ± 9)/T} and k2(T) = (3.39 ± 0.03) exp{(63 ± 3)/T}.

Uncertainties in the Arrhenius parameters are 2σ and represent precision only; estimated accuracies of reported k1(T) and k2(T) values at the 95% confidence level are ±10% around room temperature and ±15% at the temperature extremes of the study.

The O(1D2) + O2 kinetic data reported in this study are in very good agreement with available literature values.

However, the kinetic data reported in this study (and two other new studies reported in this issue) suggest that the O(1D2) + N2 reaction is significantly faster than previously thought, a finding that has important implications regarding production rates of tropospheric HOx radicals as well as stratospheric HOx and NOx radicals calculated in atmospheric models.


Atomic oxygen in its lowest-energy excited electronic state, O(1D2), is an important reactive intermediate in atmospheric chemistry.1

Production of O(1D2) in the atmosphere occurs primarily via ultraviolet photodissociation of ozone,2 and O(1D2) is rapidly destroyed via reaction with N2 and O2, the two most abundant atmospheric constituents: O(1D2) + N2 → O(3PJ) + N2O(1D2) + O2 → O(3PJ) + O2 In competition with reactions (1) and (2), a small fraction of atmospheric O(1D2) atoms are destroyed via interaction with atmospheric trace gases.

Of particular importance are the following chemical reactions: O(1D2) + H2O → 2 OHO(1D2) + N2O → 2 NOO(1D2) + N2O → N2 + O2Reaction (3) is the single most important primary source of HOx radicals in the troposphere while reaction (4a) is the dominant primary source of NOx radicals in the stratosphere.1

Assuming that atmospheric destruction of O(1D2) is completely dominated by reactions (1) and (2) (reaction (3) can actually make a small but significant contribution under humid lower tropospheric conditions), eqns. (I) and (II) can be readily derived to represent the production rates of tropospheric HOx and stratospheric NOx, respectively: d[HOx]/dt = 2J[O3]k3[H2O]/(k1[N2] + k2[O2])d[NOx]/dt = 2J[O3]k4a[N2O]/(k1[N2] + k2[O2])In the above equations, J represents the first order rate coefficient for production of atmospheric O(1D2) via O3 photodissociation, i.e., d[O(1D2)]/dt = J[O3].

Clearly, accurate evaluation of the above production rates requires accurate rate coefficients for reactions (1) and (2) over the range of temperatures encountered in the troposphere and stratosphere.

The Arrhenius parameters for k1(T) and k2(T) used in models of atmospheric chemistry are those recommended by the NASA3 and IUPAC4 panels for chemical kinetic and photochemical data evaluation.

In units of 10−11 cm3 molecule−1 s−1, the Arrhenius expressions currently recommended by the NASA panel are k1(T) = 1.8 exp(110/T) and k2(T) = 3.2 exp(70/T) while the expressions recommended by the IUPAC panel are k1(T) = 1.8 exp(107/T) and k2(T) = 3.2 exp(67/T).

The recommended room temperature rate coefficients are based almost entirely on experimental work carried out more than twenty years ago,5–8 and the recommended activation energies are based on a single temperature dependent study.5

While the earlier experimental work was at the state-of-the-art for its time, the accuracy of reported results was somewhat limited by available technology and, as a result, uncertainties in the recommended rate coefficients are undesirably high, i.e., about 20% at 298 K and over 40% at 200 K at the 67% confidence level.3,4

Over the years, the ability of the atmospheric field measurement community to accurately determine the concentrations of reactive free radicals has evolved to the point where detailed comparisons of field data with photochemical models is limited by the accuracy of rate coefficients for key reactions.9

For this reason, new kinetics studies of important atmospheric processes (like reactions (1) and (2)) that focus on high accuracy are needed.

While studying a variety of O(1D2) reactions of atmospheric interest, three respected gas kinetics research groups have independently reinvestigated the temperature dependence of k1.

The results of the three studies are in quite good agreement with each other, but suggest that k1(T) is significantly faster than suggested by previously available kinetic data.3–8

To disseminate this important new result to the atmospheric chemistry community as rapidly as possible, a joint communication was recently published10 where the results of all three groups are summarized and recommendations concerning values for k1(T) that should be used in atmospheric models are presented.

In this paper and two companion papers,11,12 each of the three individual studies is presented in detail.

Also, presented in this paper are the results of a temperature dependent kinetics study of reaction (2).

Experimental technique

The experimental approach is similar to one we have employed in several previous studies of O(1D2) reactions of atmospheric interest,7,13–18 although the data acquisition and analysis methods are improved over those employed in all but the most recent studies.17,18

The experiments involve time-resolved detection of ground state atomic oxygen, O(3PJ), by atomic resonance fluorescence spectroscopy following laser flash photolysis of O3/N2/O2/He mixtures.

The O(3PJ) atoms are produced with a small quantum yield via O3 photodissociation,13,20 and with a large yield as a product of O(1D2) deactivation by the gases in the reaction mixture.3,4

An apparatus diagram is published elsewhere.19

Some experimental details that are particularly relevant to this study are given below.

A jacketed, Pyrex reaction cell with an internal volume of approximately 160 cm3 was used in all experiments.

The cell was maintained at a constant temperature (±2 K at the temperature extremes of the study) by circulating either ethylene glycol (for T > 298 K) or 1∶1 ethanol-methanol (for T < 298 K) through the outer jacket.

O(1D2) atoms were produced by 248 nm or 266 nm laser flash photolysis of O3.

The source of 248 nm photons was a Lambda Physik model Compex 102 excimer laser operating with a KrF gas fill; the laser pulsewidth was ∼25 ns and the range of fluences employed was 24–32 mJ cm−2 pulse−1.

The source of 266 nm photons was fourth harmonic radiation from a Quanta Ray Model DCR-2A Nd:YAG laser; the pulsewidth of this laser was ∼6 ns and the range of fluences employed was 12–32 mJ cm−2 pulse−1.

An atomic resonance lamp, situated perpendicular to the photolysis laser, excited ground state O(3PJ) atoms in the reaction cell.

Radiation was coupled out of the lamp through a MgF2 window and into the reaction cell through a MgF2 lens.

Fluorescence was collected by a CaF2 lens on the axis normal to both the photolysis laser beam and the resonance lamp beam and imaged onto the photocathode of a solar blind photomultiplier; the CaF2 lens prevented impurity lamp emission of Lyman-α radiation from reaching the detector.

The regions between the resonance lamp and reactor and between the reactor and photomultiplier were purged with N2 to prevent absorption of the atomic oxygen resonance radiation (130–131 nm) by atmospheric compounds such as O2 and H2O. Signals were processed using photon-counting techniques in conjunction with multichannel scaling.

The multichannel scaler sweep was pre-triggered in order to allow a pre-flash baseline to be obtained.

All experiments were carried out under “slow flow” conditions with a linear flow rate through the reactor of about 25 cm s−1.

Since the laser repetition rate was 10 Hz, the gas mixture within the detection volume was replenished between laser pulses.

Ozone, O2, N2 (in most experiments), and zero air (low hydrocarbon impurity levels) were allowed to flow into the reaction cell from 12-L Pyrex bulbs containing dilute mixtures in helium buffer gas while helium was allowed to flow directly from its high-pressure storage tank.

In a few of the experiments designed to measure k1, the N2 source was a high pressure cylinder containing a certified 3.15% UHP N2 in UHP He mixture.

The concentrations of each component in the reaction mixture were determined from measurements of the appropriate mass flow rates and the total pressure.

The mole fraction of ozone in its storage bulbs was checked frequently by UV photometry.

The monitoring wavelength for the photometric measurements was 253.7 nm (Hg line) and the room temperature O3 absorption cross section that is needed to convert the measured absorbance to an O3 concentration was taken to be 1.144 × 10−17 cm2.20–25

Research grade helium (>99.9999% purity) was used as the buffer gas (the total pressure was 20 Torr in most experiments) and in the preparation of all gas mixtures.

The stated minimum purities of the other gases were: N2, 99.999% and O2, 99.997%; these gases were used as supplied, although in some experiments the O2/He or N2/He mixture was passed through a glass-bead-filled liquid nitrogen trap during transit from the storage bulb to the reactor.

The zero air had a stated minimum purity of 99.997% and stated upper limit total hydrocarbon mole fraction of 0.001% it was used as supplied.

Ozone was prepared by passing O2 through a commercial ozonizer; the O3 was trapped on silica gel at 196 K and excess O2 was removed by pumping.

Results and discussion

All experiments were carried out under pseudo-first-order conditions with O3, N2, O2, zero air, and of course, helium buffer gas in excess over oxygen atoms.

In reaction mixtures containing, for example, O3, N2, and He, the processes that control the temporal evolution of O(3PJ) are the following: O3 +  → O(3PJ) + O2O3 +  → O(1D2) + O2O(1D2) + O3 → O(3PJ) + O3O(1D2) + O3 → 2 O(3PJ) + O2O(1D2) + O3 → 2 O2O(1D2) + N2 → O(3PJ) + N2 In the presence of O2, production of O(3PJ) can also occur via the following (exothermic) processes:O(1D2) + O2 → O(3PJ) + O2(X 3Σg)O(1D2) + O2 → O(3PJ) + O2(a 1Δg)O(1D2) + O2 → O(3PJ) + O2(b 1Σ+g)

It is well-established that the dominant pathway for reaction (1) is channel (1a).

At a pressure of one atmosphere the yield of N2O from reaction (1) is only ∼10−6 while under the conditions of a majority of our experiments (20 Torr He buffer gas) the branching ratio for channel (1b) is less than 10−7.26

Published studies of the mechanism of reaction (2) suggest that the branching ratio for channel (2c) is high (∼0.8)27–30 and the branching ratio for channel (2b) is low (≤0.05).31

The temporal profile of the O(3PJ) fluorescence following the laser flash is described by the following expression:St = {ka/(kd − ka)}A{exp(−kat) − exp(−kdt)} + Bexp(−kdt)

In eqn. (I), St is the fluorescence signal at time t after the laser flash, ka is the pseudo-first order rate coefficient for O(3PJ) appearance, kd is the rate coefficient for O(3PJ) decay (assumed to be first order), and the parameters A and B are related to the concentrations of O(1D2) and O(3PJ) produced by the laser flash: A = [O(1D2)]0B = C[O(3PJ)]0In eqns. (II) and (III), C is a proportionality constant that relates the resonance fluorescence signal to the O(3PJ) concentration and Φ is the number of O(3PJ) produced per O(1D2) destroyed.

If (for example) the reaction mixture contains O3, He, and N2, then in eqn. (IV), k0a is the pseudo-first-order rate coefficient for O(3PJ) appearance in the absence of N2.ka = k1[N2] + k0aGround state oxygen atoms are unreactive toward He and N2, and react very slowly with O3.

Hence, under the experimental conditions employed in this study, ka ≫ kd.

As a result, the assumption that kd is a first order rate coefficient, while not strictly correct, does not compromise the accuracy with which ka, the parameter of interest, can be determined.

Some typical O(3PJ) temporal profiles observed following laser flash photolysis of O3/N2/He mixtures are shown in Fig. 1.

Bimolecular rate coefficients for the collisional removal of O(1D2) by N2, k1(T), are obtained from the variation of ka with [N2] at constant concentrations of O3 and He.

Typical data are shown in Fig. 2.

Similarly, values for k2(T) are obtained from the variation of ka with [O2] in experiments with [N2] = 0 and constant concentrations of O3 and He.

The results of all experiments carried out to evaluate k1(T) are summarized in Table 1 and the results of all experiments carried out to evaluate k2(T) are summarized in Table 2.

Arrhenius plots for reactions (1) and (2) are shown in Fig. 3.

The following Arrhenius expressions are derived from the data (units are cm3 molecule−1 s−1): k1(T) = (1.99 ± 0.06) × 10−11 exp{(145 ± 9)/T}k2(T) = (3.39 ± 0.03) × 10−11 exp{(63 ± 3)/T}Uncertainties in the above expressions are 2σ and represent precision only.

A discussion of possible systematic errors along with an estimate of the accuracy of reported rate coefficients is given below.

One potentially important source of systematic error in elementary reaction kinetics studies is the occurrence of secondary reactions that are not accounted for properly in the data analysis.

The photochemical system employed to study reaction (1) seems to be devoid of possible secondary reactions that could destroy or regenerate oxygen atoms on the time scale for O(3PJ) production via reaction (1).

However, in the case of reaction (2), oxygen atoms can be regenerated via the reaction of the major molecular product, O2(b1Σ+g), with O3: O2(b1Σ+g) + O3 → O(3PJ) + 2 O2O2(b1Σ+g) + O3 → other products

It is well-established that k7 ∼ 2.2 × 10−11 cm3 molecule−1 s−17,8,29,32–37 with little or no temperature dependence,36 and that k6a/k6 ≈ 0..733,34

Because very low O3 concentrations were employed in our study of reaction (2) (Table 2), the pseudo-first-order rate coefficient for O(3PJ) formation (via reaction (2)) exceeded the pseudo-first-order rate coefficient for O(3PJ) regeneration (via reaction (7)) by factors of 50–700 in all experiments where O2 was present in the reaction mixture.

As a result, the occurrence of reaction (7) had a negligible effect on observed O(3PJ) appearance rates.

Side reactions of oxygen atoms with background impurities in the reactant samples is usually not a problem when the rate coefficients of interest are more than 10% of the gas kinetic collision rate coefficient.

However, given the extreme level of reactivity of O(1D2) and the fact that the values for k1(T) measured in this study are somewhat faster than nearly all previously reported values, a number of experimental checks were carried out to verify that impurity reactions were unimportant.

In most experiments, the N2 source was a certified 3.15% N2 in He mixture obtained from Matheson.

However, in other experiments, the N2/He mixtures were prepared manometrically by the investigators using two different He cylinders and two different N2 cylinders.

As seen from the results reported in Table 1, excellent precision in room temperature rate coefficients was obtained even though multiple sources of N2/He reactant mixtures were employed.

In some of the k1(T) experiments, the N2/He mixture was allowed to flow through a glass-bead-filled liquid nitrogen trap.

Back-to-back experiments where the N2/He mixture (a) was allowed to flow through the trap and (b) by-passed the trap gave essentially identical results.

In most experiments, the N2/He reaction mixture was transported to the reaction cell through Teflon tubing.

However, in some room temperature experiments, the Teflon tubing was replaced with stainless steel tubing that was heated and purged with helium for several hours before being used in kinetics experiments.

As in the case of the other variations described above, the measured rate coefficient was found to be independent of whether the reactant line tubing was Teflon or stainless steel.

Finally, to test for possible mixing problems, about half of the room temperature experiments were done using a 1/4″ stainless steel cross to mix the O3/He and reactant/He mixtures with additional He, while about half of the experiments employed a 200 cm3 Pyrex “mixing chamber” in place of the stainless steel cross; the measured rate coefficient was independent of this variation in apparatus configuration.

As a final test of the internal consistency of the reported values of k1(T) and k2(T), the deactivation of O(1D2) by zero air was investigated.

The data are summarized in Table 3 and an Arrhenius plot of the ln kvs.

T−1 data is shown in Fig. 4.

The solid line in Fig. 4 is the best linear fit of the ln kvs.

T−1 data; the best fit Arrhenius expression is k(T) = (2.11 ± 0.17) × 10−11 exp{(140 ± 21)/T} cm3 molecule−1 s−1.

The dashed line in Fig. 4 is obtained from the expression k(T) = 0.79 k1(T) + 0.21 k2(T) where k1(T) and k2(T) are evaluated using the Arrhenius expressions obtained in this study.

Clearly, the O(1D2) + zero air results are consistent with the O(1D2) + N2 and O(1D2) + O2 results.

Since secondary chemistry and impurity reactions appear to be very minor sources of error in this study (see above), the largest source of systematic error is probably the measurement of the reactant concentration.

Flow meter calibrations were checked frequently over the course of the study, and are believed to be accurate to within a few percent.

We conservatively estimate the overall uncertainty in the reported room temperature rate coefficients to be ±10% at the 95% confidence level.

At the temperature extremes of the study, the overall uncertainty increases a little because (a) precision is not as good as at room temperature and (b) uncertainty of ±2 deg in the temperature leads to additional uncertainty in the reactant concentration.

We estimate the overall uncertainty in the rate coefficients measured at the temperature extremes of the study to be ±15% at the 95% confidence level.

The NASA panel for chemical kinetics and photochemical data evaluation parameterizes temperature-dependent uncertainty factors in second order rate coefficients using the following relationship:3f(T) = f(298)exp{|(ΔE/RT)(T−1 − 298−1)|} At the 67% confidence level (used by the NASA panel), the values of the parameters f(298) and (ΔE/R) that are consistent with the above error estimates are f1(298) = f2(298) = 1.05 and (ΔE/R)1 ∼ Δ(E/R)2 ∼ 15 K, where the subscripts refer to the reaction number.

The kinetic data obtained in this study are compared with other published rate coefficients in Tables 4 (k1) and 5 (k2).

For the sake of clarity, only results published since 1973 are included in the Tables.

The results summarized in the Tables include all studies where absolute rate coefficients are reported5–7,11,12,38,39 as well as one recent relative rate study,8 but do not include a number of relative rate studies published during the 1960s and early 1970s.27,28,40–49

None of the early relative rate data are used to derive current recommendations of k1(T) and k2(T) for use in atmospheric models.3,4

Uncertainties in room temperature rate coefficients are not quoted in Tables 4 and 5 because it is very difficult to obtain a consistent set of uncertainty estimates from the published information.

It appears that reasonable uncertainty estimates for previously published absolute room temperature values of k1 and k2 are typically around ±20% at the 95% confidence level.

The rate coefficients reported by Husain and coworkers38,39 are significantly faster than all others summarized in Tables 4 and 5; this trend also holds for a number of other O(1D2) reactions,3,4 and is widely believed to result from a problem with the experimental determination of the modified Beer–Lambert law parameter γ that is needed to relate the O(1D2) absorbance to its concentration.

A theoretical treatment50 suggests that γ should be near unity under the experimental conditions employed by Husain and coworkers, while the experimental value used to obtain the rate coefficients given in Tables 4 and 5 is γ = 0.41.

Assuming γ = 1 would decrease the measured rate coefficients by more than a factor of two.39

The results of Husain and coworkers are not included in the data set used by the NASA3 and IUPAC4 panels to arrive at recommendations of O(1D2) rate coefficients for use in atmospheric models.

The Arrhenius parameters for the O(1D2) + O2 reaction that are reported in this study are in very good agreement with current panel recommendations3,4 (which are based on the room temperature study of Amimoto et al6. and the temperature-dependent study of Streit et al5.), and also agree well with results reported in the two very recent studies of Dunlea and Ravishankara11 and Blitz et al.12

On the other hand, the values of k1(T) reported in this study are among the fastest of any listed in Table 4, although our results agree quite well with those reported in the two other very recent studies.11,12

The room temperature rate coefficient reported in this paper is 25% faster than the currently recommended value,3,4 which is based on the absolute studies of Streit et al.,5 Amimoto et al.,6 and Wine and Ravishankara,7 as well as the competitive kinetics study of Shi and Barker.8

The 1981 study of Wine and Ravishankara7 employed the same experimental approach as this study and another recent study by Dunlea and Ravishankara.11

However, the precision and accuracy of the recent studies is believed to be superior to the 1981 study because of improved instrumentation for data collection and analysis, which allows higher quality data to be obtained and more experimental tests for systematic errors to be performed in a workable time frame.

It is well established that photodissociation of O3 at 248–266 nm generates translationally hot O(1D2).51–56

Matsumi and coworkers have investigated the competition between translational relaxation and electronic quenching in collisions of O(1D2) with N255 and O2,54 and find that, in pure N2 or pure O2, electronic quenching occurs before translational deactivation is complete, i.e., the translational energy distribution of the O(1D2) atoms undergoing electronic quenching is superthermal.

The cross section for reaction (2) appears to depend only weakly on the relative translational energy of the reactants,54,57 but both experimental55 and theoretical58–60 results suggest that the cross section for reaction (1) decreases sharply with increasing reactant translational energy.

All absolute kinetics studies of reactions (1) and (2) have employed reaction mixtures where the reactant (N2 or O2) is diluted in helium, thus ensuring thermalization of the O(1D2) translational degrees of freedom.5–7,11,12,38,39

However, the results discussed above55,58–60 suggest that if O(1D2) is generated in the atmosphere with significant translational excitation, its effective rate coefficient for deactivation by N2 could differ from the thermal rate coefficient measured in this study and in all other absolute kinetics studies summarized in Table 4.

In the troposphere and lower stratosphere, O(1D2) is generated via O3 photolysis at wavelengths longer than 300 nm, i.e., near the thermodynamic threshold for the spin-allowed O(1D2) + O2(a 1Δg) channel which accounts for well over 90% of O(1D2) production at the relevant wavelengths.2

Hence, it appears that the thermal rate coefficient k1(T) can be employed in simulations of atmospheric chemistry without introducing significant error.

However, in the upper stratosphere and above, where significant O(1D2) production occurs via short wavelength photodissociation of O2 and O3, use of thermal rate coefficients in simulations could result in systematic errors in evaluation of HOx and NOx production rates.

Arguing against this idea is a study by Greenblatt and Ravishankara,61 who have measured the yield of NO resulting from reaction of O(1D2) with gas mixtures containing N2O, N2, O2, and He under a variety of experimental conditions.

Within the combined uncertainties, the NO yields measured by Greenblatt and Ravishankara agreed quite well with yields calculated from values for k1, k2, and k4a that were accepted at the time of their study (1990).

Furthermore, NO yields were found to be independent of (a) O(1D2) source, i.e., 193 nm photolysis of N2O versus 248 nm photolysis of O3, and (b) whether or not helium was present in the reaction mixture, suggesting that the stratospheric N2O production rate can be evaluated with good accuracy using thermal rate coefficient data.

Reanalysis of the Greenblatt and Ravishankara data using the new faster value for k1 reported in this study and two other new studies11,12 improves the agreement between NO yields measured in their “integrated study” and yields calculated from the individual thermal rate coefficients.11

For reasons discussed in the introduction, the higher rate coefficient for the O(1D2) + N2 reaction that is reported in this study has important implications for model calculations of the production rates of tropospheric HOx as well as stratospheric HOx and NOx radicals.

For example, in the marine lower troposphere, where water vapor concentrations are relatively high but concentrations of other potential OH precursors are relatively low, using the Arrhenius expression for k1(T) reported in this study would lead to a calculated OH production rate that is ∼18% smaller than the production rate that is calculated using the currently recommended3,4 Arrhenius expression.

In the upper troposphere, however, where H2O levels are very low and, as a result, the O(1D2) + H2O reaction is a less important OH source, use of the expression for k1(T) reported in this study would reduce the calculated OH production rate by only ∼5%62.