Absolute rate constant and O ( 3 P ) yield for the O ( 1 D ) + N 2 O reaction in the temperature range 227 K to 719

Absolute rate constant and O(P) yield for the O(D)+N2O reaction in the temperature range 227 K to 719 K S. Vranckx, J. Peeters, and S. A. Carl University of Leuven, Department of Chemistry, 200F Celestijnenlaan, 3001 Leuven, Belgium Received: 4 March 2008 – Accepted: 15 April 2008 – Published: 19 May 2008 Correspondence to: S. A. Carl (shaun.carl@chem.kuleuven.be) Published by Copernicus Publications on behalf of the European Geosciences Union.

received attention due to their occurrence in gas-phase plasmas used in the oxidation of silicon surfaces (Kaspar et al., 2003) and in photochemical processes in other planetary atmospheres and around comets (Bhardwaj and Haider, 2002;Nair et al., 2005).
In this work we focus on the reaction of O( 1 D) with N 2 O, which has seven exothermic 5 product channels for which both direct and indirect quantification of the products, N 2 , NO, and O( 3 P) have been employed to determine product branching fractions. For the important NO channel, k R1d /k R1 appears to be reasonably well established at room temperature; a value of 0.61±0.06 (95% confidence) was recommended by Cantrell et al. (1994) following k R1c /k R1 =0.04±0.02. The rest of the reaction flux (ca. 35% to 40%) passes through either of, or a combination of, the last three channels (R1e)-(R1g) yielding O 2 +N 2 . The electronic state in which O 2 is preferentially produced is not established. As well as the branching ratios, the overall rate constant, k R1 , has been determined in several studies, using a variety of methods (Carl, 2005, and references therein) 5 for following the time profile of O( 1 D). The current NASA/JPL panel recommendation (Sander et al., 2006) for k R1 (298 K) is (1.17±0.40)×10 −10 cm 3 s −1 , with the uncertainty representing approximately 95% confidence. This value is based on early k R1 determinations by Davidson et al. (1979), Amimoto et al. (1979), Wine and Ravishankara (1981), and the very recent studies of Blitz et al. (2004), and Dunlea and Ravishankara 10 (2004). Since the reported k R1 values of the latter two studies differ by some 18% at room temperature and by almost 30% at lower stratospheric temperatures, no large improvement in the uncertainty of the recommended value over the previous recommendation was forthcoming. In fact the latest four determinations of k R1 show a fair spread in values at room temperature. For the two studies mentioned above, Dun-15 lea and Ravishankara (2004) determined k R1 to be (1.21±0.04)×10 −10 cm 3 s −1 , and Blitz et al. (2004) determined a value of (1.07±0.08)×10 −10 cm 3 s −1 , whereas the latest two room-temperature determinations by Takahashi et al. (2005) and by Carl (2005) reported values of (1.35±0.08)×10 −10 and (1.43±0.08)×10 −10 cm 3 s −1 , respectively, where all values are given with their reported 95% confidence limits.

20
The most extensive set of measurements to date are those above by Dunlea and Ravishankara (2004) for which twenty four determinations of k R1 were carried out over the temperature range 220 K-370 K. Actually, the high stated precision of those determinations reveals a statistically significant difference between the weighted average of all nine 295 K data, (1.21±0.04)×10 −10 cm 3 s −1 , and the k R1 (295 K) value of 25 (1.34±0.04)×10 −10 cm 3 s −1 predicted from an Arrhenius fit to all other k R1 (T ) determinations of that study (fifteen in all). This, together with the most recent data of Takahashi et al. (2005) and Carl (2005), suggests that k R1 is significantly greater than the current NASA/JPL recommendation, though still within its broad uncertainty limits.

8884
The aim of the present study is three-fold. Firstly, to further reduce the uncertainty in k R1 for this very important reaction by employing a recently-developed, highly-precise method (Carl, 2005) for the determination of O( 1 D) rate constants; secondly, to clearly establish its temperature dependence by extending k R1 determinations to higher temperatures; and thirdly, to determine the branching fraction k R1c /k R1 over a wide temper-5 ature range to establish its value at stratospheric temperatures. High-temperature rate constant determinations could also aid predictions of stratospheric NO formation rates under conditions where the population of photolytically-produced O( 1 D) is not entirely thermalized before reaction with N 2 O (Kharchenko and Dalgarno, 2004).
2 Experimental section 10 We use a method to monitor O( 1 D), described recently by Carl (2005), based on the 431 nm CH(A→X) chemiluminescence resulting from the reaction, Pulsed laser (10 ns) photolysis of the precursors C 2 H 2 and N 2 O, at 193 nm, generates the reactants of the above chemiluminescence reaction. Aside from production of 15 O( 1 D), N 2 O photolysis at 193 nm results in a small fraction of O( 3 P) (Φ(O( 1 D))=0.995, Φ(O( 3 P))=0.005)) (Nishida et al., 2004). Additional O( 3 P) can result from O( 1 D) quenching by the precursor molecules C 2 H 2 and N 2 O, and by the bath gas He. The presence of O( 3 P) in the reaction volume also leads to CH(A→X) chemiluminescence by the analogous, but less efficient, reaction: Rather than being a hindrance to the study of O( 1 D) reactions, the occurrence of Reaction (R3) yields precise information on the fraction of O( 1 D) quenched during its reactive lifetime. 8885 The radiative lifetime of CH(A) due to spontaneous emission is sufficiently short, at ca. 540 ns, (Luque and Crosley, 1996;Tamura et al., 1998)  components: the first arising from Reaction (R2) and the second from Reaction (R3). The chemiluminescence intensity, I chem2 (t), due to Reaction (R2) is given by  and k C 2 H could be determined by fitting the decay rates of I chem (t)≡I chem2 (t) + I chem3 (t) at short times and at long times, respectively. This analysis is valid provided initial O( 3 P) formation is minor, such that the decay profile at short times is representative of 5 O( 1 D) and C 2 H decay only. The full I chem expression is given below in Eq. (2).
The ratio of the chemiluminescence channel rate coefficients k R2 and k R3 in Eqs. (1a), (1b) and (2a), (2b) is equal to 3.0±0.2 at room temperature. This (T -dependent) value was determined in this study by simply taking two chemiluminescence intensity profiles: one when photolysing a mixture of N 2 O and C 2 H 2 in helium buffer gas,  and 2.6×10 5 s −1 respectively. Thus, extrapolation of the emission intensities to t=0 and 20 taking the ratio in N 2 (or Ar) over that in He gives k R2 /k R3 , provided that CH(A) is not significantly quenched by the buffer gas (Tamura et al., 1998). The concentration of each gas in the reaction chamber was calculated from the measured gas flows using the gas law together with the known fractional composition in the high-pressure cylinders. Though not a critical parameter we used the value of 25 the fractional composition for C 2 H 2 in He as that stated by the manufacturers of 0.0096. High-purity (99.9997%) helium was used as the bath gas for all kinetic experiments. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
The concentration of N 2 O is, of course, an important parameter. The suppliers specification is 5.0% N 2 O in high-purity He. However, the fractional concentration of the cylinder was checked by single-pass uv optical absorption in a 1.2 m absorption cell using a D 2 lamp as the light source. The resulting spectra together with fits using the known absorption cross sections (Sander et al., 2006;Selwyn et al., 1977) are dis-5 played in Fig. 1. Our fits give the percentage of N 2 O in the cylinder as (5.10±0.30)% (95% confidence).
The possible impurities of the N 2 O/He cylinder were also checked by electron-impact ionization mass-spectrometry, in which several mass spectra were taken as a function of electron energy to eliminate ions resulting from N 2 O fragmentation from the analy-10 sis. Here only trace amounts of NO and N 2 were detected and their estimated mole fractions of less than 1×10 −6 were too small to significantly influence the kinetic measurements.
A  Fig. 2. To enable a large range of temperatures to be covered, two reaction vessels of entirely different construction were connected in series, such that the photolysis laser beam was able to pass through both of 20 them at the same time.
The reactor on the left in Fig. 2 is made of a single tube of chemically-inert PFA (perfluoroalkoxy) of internal diameter 10 mm and length 30 cm with a gas inlet and outlet. As connections to the single Spectrosil quartz window of the PFA reactor and for the gas inlet and pressure meter, PFA Swagelock "Tee" connectors were used (not shown).

25
This reactor was designed to cover the ranges from ca. 500 K to ca. 220 K. It is cooled or heated by fluid flowing in a closed circuit through a metallic jacket surrounding the PFA reactor tube. There is a quartz entrance window for the laser beam; the exit window is placed after passage through the second reactor volume. Interestingly, no 8888 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion window for detection of the chemiluminescence is required for the PFA reactor: the PFA tubing is partially transparent to the 431 nm chemiluminescence and according to our test does not significantly fluoresce at this wavelength on passage of the uv photolysis pulse. The chemiluminescence detection system comprises a lens, an interference filter (430±10 nm) and a photomultiplier tube that are placed close to the PFA reactor, 5 and arranged such that the chemiluminescence is detected at right-angles to the passage of the laser beam. For this, the metallic cooling jacket exposes the PFA tube on one side for 3 cm at about 3/4 along its length. The cooling fluid was maintained at the correct temperature by a commercial temperature controller. Pressure in the reaction cell was determined using a calibrated capacitance manometer that was cross-checked 10 regularly with other calibrated pressure gauges. The pressure measurement point was located about 12 cm downstream of the chemiluminescence observation zone, using a second PFA Swagelock "Tee" connector (not shown) placed between the two reactors. At the flow rates used in this experiment there was negligible pressure drop between the observation point and the pressure-measurement point. Upstream of the reactor, 15 the separate flows of He, C 2 H 2 /He, and N 2 O/He, were combined in a small volume to ensure complete mixing before entering the reactor. The reactor on the right is a larger heatable stainless steel cell that has been used for many previous studies for C 2 H, OH, and CF 2 reactions (Elsamra et al., 2005;Khamaganov et al., 2006;Dils et al., 2004). It can cover a temperature range of 290 K to 20 850 K. A glass window is used as exit window for the chemiluminescence, detected perpendicular to the axis of the laser beam. The interference filter, collection lens and PMT are mounted together on a translation stage such that they are easily moved between the observation points of the two reaction cells, thus providing a very convenient way to directly compare rate constants at two different temperatures if need be.

25
The temperature of the gas mixture in the reaction cells was measured using retractable calibrated thermocouples. We estimate the uncertainties in temperature to be ±10 K at 719 K decreasing to ±1 K at room temperature and then to increase again to 220 K to ±5 K, all at ca. 95% confidence. 8889 3 Results and discussion

Determination of k R1 (T )
Preliminary results revealed a very small O( 3 P) yield for the title reaction, in qualitative agreement with the previous studies mentioned above. Thus, under the conditions used for the rate constant determinations, the chemiluminescence decay profiles are 5 effectively single exponential at short times (t≤10 µs) and represent the sum of the decay rates of O( 1 D) and C 2 H only, with interference from any growth of O( 3 P) therefore negligible. In fact, even for reactions with substantial quenching to O( 3 P), the decay rate of O( 1 D) alone is in principle relatively easily extracted as described previously (Carl, 2005). Here though, the determination of k R1 is more transparent as it involves 10 fitting to a single-exponential decay only. A typical chemiluminescence time profile generated using our new method is displayed in the log-linear plot of Fig. 3 The solid line is a single-exponential fit to the data, neglecting the longer-time portion that includes chemiluminescence arising from a small fraction of O( 3 P) produced by quenching of O( 1 D).

20
Our recent detailed experimental and theoretical investigation of the C 2 H+N 2 O reaction (manuscript in preparation) shows it to have a significant barrier and a low rate constant ranging from ( Figure 4 shows examples of such k R1 determinations at 227 K, 298 K and 422 K. The ordinate intercepts correspond to the sums of the known removal rates of O( 1 D) and C 2 H by C 2 H 2 , with rate constants at 295 K of 3.08×10 −10 cm 3 s −1 and 1.3×10 −10 cm 3 s −1 , respectively (Carl, 2005;Van Look and Peeters, 1995;Vakhtin, 2001). That our k C 2 H+N 2 O was found to be very small at room temperature provides ad-5 ditional support of the high purity of the N 2 O mixture used for the O( 1 D) experiments, given the general high reactivity of C 2 H, especially toward hydrocarbons. The high precision of the data in the plots of Fig. 4 is typical of our chemiluminescence method, which allows O( 1 D) removal rates of up to 450 000 s −1 to be measured.
Note that the data presented in this study, including those taken at room temperature,  Table 1 along with three columns of uncertainties. The first of these gives the 95% confidence limits in the fitted slopes for plots such as those given in Fig. 4 that are derived statistically from the weighted least-squares fitting procedure. For fitting a suitable function to the 20 data, the relative weights of the individual data points need to be estimated. The random uncertainties on individual determinations just mentioned (column 3 of Table 1) should not, in this case, be used as relative weighting factors, as this would imply, for example, that the confidence in the value at 227 K is a factor four greater than that at 447 K, whereas it is noted that the scatter of the data is greater than the random 25 uncertainty associated with many of the individual determinations, including the one at 227 K. In fact the scatter of the data around some mean value -given by the standard deviation -has two random error contributions. One of these is a pooled average of the individual uncertainties; the other represents the random error introduced by repeating 8891 an experiment, which is likely due mainly to uncertainties in flow controller calibrations. To determine the average scatter of the data points we consider the data in the range 227 K to 447 K and assume k(T ) it to be constant (this will tend to overestimate the scatter). Thus, the standard deviation, SD, is 0.045 cm 3 s 1 . An estimate of the likely spread in the data at 95% confidence is ±2×SD=±0.090×10 −10 cm 3 s −1 . The 5 average contribution to this value of the uncertainties of individual determinations is (N/Σ i (1/s 2 i )) 0.5 =0.029, where s I are the 95% confidence on individual determinations and N is the number of data points considered. Thus the contribution, at 95% confidence, of experiment repeatability to 2×SD is (0.090 2 −0.029 2 ) 0.5 =0.085. This last value is now propagated with the uncertainty of each data point (column 3, Table 1) 10 to give an estimate of the relative weights of the data. These are given in column 4 of Table 1 and also plotted as error bars in Fig. 5. Using these values as error bars is somewhat artificial since they have been derived partly from the data itself, however they do give a visual representation of the relative weights of the data points. Additional to those random errors, is the systematic uncertainty of ca. 6% in the fractional concentration of N 2 O of our cylinder that was based on analysis of the results of our absorption measurements. This uncertainty, which applies equally to all points with the same sign, affects only the absolute value of the whole set of rate constant data and not their temperature dependence. It is statistically added to the overall errors at 95% confidence in the final column. These overall uncertainties cannot be used in a least-20 squares fitting procedure of the data. We estimate our confidence in temperature of the monitored reaction zone (at ca. 95% limits) to be ±10 K at 719 K decreasing to ±1 K at room temperature and then to increase again to 220 K to ±5 K. Given the relatively flat temperature profile of the rate constant data over 446 K to 227 K region, such uncertainties in temperature will make no significant contribution to the final uncertainties 25 of the results. All of our k R1 values below 450 K are significantly greater than the current recommendation. Between 227 K and 446 K the determined values -twenty one in all -range from 1.28×10 −10 cm 3 s −1 to 1.43×10 −10 cm 3 s −1 (standard deviation of 0.05×10 −10 cm 3 s −1 ) (1.37±0.09)×10 −10 cm 3 s −1 at 95%, which includes the 6% uncertainty in our N 2 O fraction. These full confidence limits are plotted as dashed lines around the mean value in Fig. 5. This value is in excellent agreement with the room temperature study of Takahashi et al. (2005) of (1.35±0.06)×10 −10 cm 3 s −1 as well as the average of all determinations (beside those at room temperature, as discussed in the Introduction) by Dunlea and Ravishankara (2004)  This effect is not unusual if the so-called "bottle-neck structure" is located on a purely attractive entrance part of the potential energy surface. Conservation of the rotational quantum number J during the reaction means that the amount of energy available for random distribution amongst all other modes changes as the reactants approach one another. The bottle-neck structure is the structure (or point on the potential energy hy-20 persurface) that has a minimum number of states, counted from the zero-point energy to total available randomizable energy for that structure. As the reactants approach and the overall moment of inertia decreases, the energy associated with J, E J , increases leading to proportionally less energy available for randomization. The total available randomizable energy (and therefore the number of states) depends therefore 25 both on the shape of the potential energy surface and on how E J changes over the surface. The first is independent of temperature whereas the latter is temperature dependent. At higher temperatures, the differences in E J over the surface become more pronounced with the result that the bottle-neck structure -which will also change with 8893 temperature -has proportionally fewer available randomizable states. This leads to a decrease in rate constant with increasing temperature. A similar effect can be caused by partial re-dissociation of an initially-formed reaction complex. In this case there is an increased propensity to re-dissociate to reactants over the entrance-channel transition state rather than undergo transformation via a second 5 transition state (that can lie lower in energy than the reactants) leading to reactants.

Determination of the O( 3 P) yield
For the accurate interpretation of O( 3 P) yields from Reaction (R1), sources of potential interferences need to be considered. The first of these is direct production of O( 3 P) from the 193 nm photo-dissociation of N 2 O. The quantum yield for this process had would be ±0.002. Note that O 2 or N 2 impurities, e.g. from air leaks, could affect the results by chemiluminescence via C 2 H+O 2 →CH(A)+CO 2 (Elsamra et al., 2005) or by O( 1 D→ 3 P) quench- 15 ing. It was duly ascertained in this work that the O 2 and N 2 traces were so small as to have an entirely negligible effect. We must now conclude that the anomalously high quenching fraction value of 0.056 reported in the earlier investigation of this laboratory by Carl (2005) was most likely due to a very small air leak in the reactor, whose influence may have been amplified due to its proximity to the chemiluminescence ob-20 servation zone.
The general equation for the chemiluminescence decay profile is (Carl, 2005) with Q x the fractional yield of O( 3 P) from the reaction O( 1 D)+X; Q x =1 for X =N 2 , and He.

10
The ratio k R2 (T )/k R3 (T ) was experimentally determined using our new detection method for O( 1 D) and O( 3 P). An absolute determination of either k R2 or k R3 as a function of temperature was not possible here because the absorption cross-section of the oxygen atom precursor, N 2 O, is not accurately known as a function of the temperature. However, it was possible to determine their ratio, as described in the experimental section. 15 Thus, the ratio of initial chemiluminescence intensities, one taken with buffer gas He that does not contribute to any O( 1 D→ 3 P) quenching, and the other with buffer gas N 2 that rapidly quenches all initial O( 1 D) to O( 3 P), is given by In fact the temperature dependence given by the denominator, which is the sum of the T dependence of k 3 and the T dependence of σ(N 2 O) at 193 nm (assuming Φ(O 1 D)=1) is quite similar to that found by Devriendt et al. (1996) for the T dependence of k 3 alone.

20
This indicates that σ(N 2 O) changes by a factor 1.5 at most between 300 K and 800 K. For the O( 3 P) yield determinations, a typical decay profile is given in Fig. 7 The results of all determinations of the O( 3 P) yield from the title reaction, such as the one given in Fig. 7, are summarized in Fig. 8 together with the results of earlier studies of Wine and Ravishankara (1982) and Nishida et al. (2004). The results clearly show the yield of O( 3 P) to be less than 0.01 over the temperature range 550 K to 250 K, slightly decreasing with decreasing temperature. At the lowest 15 temperature of ca. 250 K an upper limit for the O( 3 P) yield of 0.002 could be established. Channel (1c) therefore need not be taken into consideration in stratospheric chemistry models. The overall impact of the study related here has been firstly to provide a set of k R1 (T ) data that taken with that of Dunlea and Ravishakara (excepting their anoma- taken from this work, and that for O( 1 D)+C 2 H 2 was taken from Carl (2005) and assumed to be independent of temperature. The rate constant for C 2 H+N 2 O is taken from recent, unpublished, experiment determinations of this laboratory that shows it to be less than 2×10 −14 cm 2 s −1 below 600 K. The rate constant for C 2 H+C 2 H 2 was taken to be 1.3×10 −10 cm 3 s −1 , independent of temperature. 8897 all rate constant for this reaction, and to significantly increase its best estimate, to be used in atmospheric chemistry models. The impact of our recommended values on modeling calculations naturally depends on values currently adopted for a particular model. At one extreme, models that rely on the JPL/NASA recommendations prior to 2006 with an overall T -independent rate constant k R1 =1.16×10 −10 cm 3 s −1 , that in-5 terpret the branching of 0.60 as k R1d /(k R1e +k R1f + k R1g ), and not k R1d /k R1 , and then consider an additional reduction of k 1d by 4% due to channel (1c), as discussed by Nishida et al. (2004), effectively use a value for the NO channel k R1d of 1.16×0.60 (1-0.04)=6.7×10 −11 cm 3 s −1 . On the other hand the 2006 recommendation (Sander et al., 2006) of k R1d =6.7×10 −11 exp(20/T ) gives a rate constant of 7.3×10 −11 cm 3 s −1 at 10 220 K. The IUPAC recommendation (Atkinson et al., 2004) for k R1d is 7.2×10 −11 cm 3 s −1 , independent of temperature. If we also adopt a branching ratio k R1d /k R1 =0.60, the present study results in a value for k R1d of 8.3×10 −11 cm 3 s −1 at 220 K, which represents significant increases of ca. 15% over the last two values and of 24% over the first.

Conclusions
We have determined the rate coefficient k R1 of the reaction O( 1 D)+N 2 O over the wide temperature range 227 K-719 K using a new and highly sensitive technique for monitoring O( 1 D), that provides a high k R1 (T ) measurement precision. We have firmly established that the rate constant has negligible temperature dependence over atmo-20 spheric temperature ranges, but shows a pronounced negative temperature dependence for T >450 K. Our k R1 (T ) data are significantly higher than the current JPL/NASA recommendations. At stratospheric temperatures, at which the title reaction is most important, our rate constant is about 15% above the current recommendation. We have also determined that the minor channel leading to O( 1 D→ 3 P) quenching is entirely 25 negligible at all atmospheric temperatures.
8898 8902 Fig. 1. Spectra of mixtures of 5% N 2 O in He taken in a single-pass absorption cell of 1.2 m length at three different total pressures. The fit to the data is based on the total cell pressure, the room temperature absorption cross-section of N 2 O and the fractional concentration of N 2 O, which is the variable in the fit.  (2005). The dotted line through the middle of the graph and the outer dotted lines represent the current JPL/NASA recommendation and its ca. 95% uncertainty limits, respectively. The inner lines represent the best fit to our data between 227 K and 446 K assuming a T -independent k R1 and 95% confidence limits that includes the 6% uncertainty in the N 2 O fraction. The plot on the right better shows individual room-temperature determinations of this work and those from the most recent studies by other groups. was recorded under exactly the same conditions as for profile (a) except that a small fraction of the He flow was replaced by N 2 flow. In this case, O( 1 D) is rapidly quenched by N 2 to O( 3 P). Here, except at short times, the decay rate is that of [C 2 H] only and the intensity proportional to [C 2 H][O( 3 P)]k(C 2 H+O( 3 P)→CH(A)+CO). The dotted lines are expected time profiles of the chemiluminescence signals in each case. For profile (b), at 0<t<2 µs, a deviation from the expected profile is seen due to the time constant of the collection electronics. The ratio of the extrapolated intensities of the two exponential profiles to t=0 gives the ratio of the rate constants for the two chemiluminescence reactions considered in the paper.   In this case the adjustable parameter Q N 2 O (=k R1c /k R1 )=0.02±0.02. Also plotted are simulated curves for Q N 2 O =0.01, 0.02, and 0.04. Note that Q N 2 O =0.04 was found by Nishida et al. (2004) and was suggested as an upper limit by Wine and Ravishankara (1982). Fig. 9. Summary of the results of the determinations of k R1c /k R1 . All error bars indicate 95% confidence.