Development of a Mechanism for Nitrate Photochemistry in Snow

Development of a Mechanism for Nitrate Photochemistry in Snow

(Parte 1 de 3)

Development of a Mechanism for Nitrate Photochemistry in Snow

Josue Bock* and Hans-Werner Jacobi*

UniVersite Joseph Fourier s Grenoble 1/CNRS, Laboratoire de Glaciologie et Geophysique de l’EnVironnement, Grenoble, 54 Rue Moliere, 38402 St. Martin d’Heres, France

ReceiVed: September 24, 2009; ReVised Manuscript ReceiVed: NoVember 17, 2009

A reaction mechanism to reproduce photochemical processes in the snow is reported. We developed a box model to represent snow chemistry. Constrained by laboratory experiments carried out with artificial snow, we deduced first a reaction mechanism for N-containing species including 13 reactions. An optimization tool was developed to adjust systematically unknown photolysis rates of nitrate and nitrite (NO2-) and transfer rates of nitrogen oxides from the snow to the gas phase resulting in an optimum fit with respect to the experimental data. Further experiments with natural snow samples are presented, indicating that NO2 - concentrations were much lower than in the artificial snow experiments. These observations were used to extend the reaction mechanism into a more general scheme including hydrogen peroxide (H2O2) and formaldehyde (HCHO) chemistry leading to a set of 18 reactions. The simulations indicate the importance of

H2O2 and HCHO as either a source or sink of hydroxyl radicals in the snow photochemistry mechanism. The additionof H2O2 and HCHO in the mechanismallowsthe reproductionof the observedlow NO2- concentration.


Various studies have shown that a variety of chemical and physical processes occur in the snow.1,2 A range of chemical reactions takes place in the snow and the interstitial air of the snow leading to modifications in the concentration of impurities present in the snow.1-3 As snow is connected to the overlying atmosphere through the open porosity network,4,5 compounds present in the snowpack or in the air can interact and exchange from one place to the other through various processes such as dry deposition, sublimation, adsorption, or condensation.2 Therefore, a modification of the composition of the snowpack can also impact the composition of the atmospheric boundary layer just above the snow. As snow-covered areas represent as much as about 14% of the Earth’s surface, i.e., up to 40% of the land masses during the boreal winter,6,7 these interactions may have an impact in snow-covered regions on the regional budgets of atmospheric nitrogen oxides (NOx ) NO + NO2) and oxidants since they are commonly produced in the snow- pack.2

Photochemical reactions are of a special importance in snow chemistry. They involve species present both in the air and in the snow itself.2 Especially, the observation of the production of NOx following the nitrate (NO3-) photolysis in the snow8 initiated several field, laboratory, and modeling studies2,3,9-14 to characterize and quantify the processes and reactions that participatein the NO3- and subsequentnitrite(NO2-) photolysis. At present,reliablechemicalmodelingof snow does not exist, even if some studies have been performed in this direction.

Because NO3- is ubiquitously present in the snow,13,14 the determination of a comprehensive mechanism for the NO3 - photolysis is a first step for snow chemistry modeling. Jacobi and Hilker14 proposed such a snow chemistry mechanism including 14 reactions based on laboratory experiments with artificially produced snow and using reaction rate coefficients determined in the aqueous phase. They further proposed a simplified mechanism by comparing the rates and determining which of the 14 reactions are significant. A similar reaction mechanism for the NO3- photolysis in snow including

18 reactions was presented by Boxe and Saiz-Lopez.13 Their model also includes an atmospheric chemistry part to account for emissions from the snow pack. More recently, Anastasio and Chu15 proposed a mechanism with eight reactions for nitrogen chemistry in the snowpack at Summit, Greenland, to investigate the effect of pH on the release of HONO to the gas phase. Here, we report the further development of a generalized mechanismfor NO3- photochemistryin the snow.We developed a specific box model to reproduce snow chemistry. We present the results of a range of simulations concerning previous laboratory experiments using artificial snow, starting with a full mechanismcombiningreactionsproposedby Jacobiand Hilker14 and Boxe and Saiz-Lopez.13 Subsequently, we propose a simplified mechanism by identifying reactions that are negligible. To test this reaction mechanism optimized with artificial snow, we performed an additional laboratory experiment using natural snow samples and applied the model to this specific experimentto furthervalidateand refinethe reactionmechanism. Because of predictabledifferencesbetween the experimentsand the results modeled with the mechanism taking into account only nitrogen containing compounds, we added reactions of further compounds that are always present in natural snow

(especiallyhydrogenperoxide(H2O2) andformaldehyde(HCHO)) to get closer to a typical natural snow composition.


ExperimentalMethods.Experimentswithnaturalsnowwere performed using snow samples collected on March 2, 2005, on the roof of building C of the Alfred Wegener Institute located in Bremerhaven, Germany. The samples consisted of the top 5-10 cm of fresh snow deposited in the last 12 h before sampling. In early March the air temperature was always below zero with averaged temperatures between -2 and -7 °C. The samples were collected in 1 L Schott bottles and stored in a refrigerator at ∼-20 °C until used for the experiments. The experimental setup of the photolysis experiments was identical to previously performed experiments using artificial snow

J. Phys. Chem. A X, x, 0 A

10.1021/jp909205e X American Chemical Society

samples.14 In short,snow sampleswere filledfrom the 1 L Schott bottles into cylindrical 1 cm long Teflon cells, equipped with quartz windows. The samples were radiated using a mercuryarc lamp (Oriel Instrument, Stratford, CT) with a power input of 500 W. A sequence of a 10 cm long liquid filter filled with Milli-Q water and a 10 cm long cylindrical extension was directly coupled to the output of the lamp housing condenser. The reaction cell was located inside a freezer, which was regulated to a temperature of -31 to -30 °C. Before each experiment,the filled cell was stored several hours in the freezer to ensure thermalequilibrium.Before and after each experiment,

NO3- and NO2- concentrationsin the meltedsnowsampleswere determined using an ion chromatography system.16 The system was always calibrated with a range of standard solutions and Milli-Q water before and after the analysis of the samples. The for NO2- or (10%, whichever is larger. Model Description. We based our model on the assumption that ions and dissolved molecules of the snow are present in a disordered surface layer. At present, the location of impurities within the snow crystals is subject to debate.2 However, in the case of the artificial snow used in the experiments, it was concluded that at least a large fraction of the NO3- is always present in the surface layer.16 As a consequence, all chemical reactions should occur in this layer. At temperatures close to the melting point, this is a thin layer on the surface of the ice crystals, which is commonly called quasi-liquid layer (QLL), because its properties are intermediate between a liquid and a solid.5,17-19 It has further been demonstrated that such a layer exists down to very low temperatures of 228-200 K.20,21 As its structure is poorly understood, we decided to represent the QLL as a homogeneous layer, testing if its complex properties can reasonably well be described using this simplification. We elaboratedour snow chemistrymodel as a conceptualbox model consideringthe QLL of the snow and the gas phase representing the snowpack interstitial air. This model was programmed in Fortran. We used aqueous phase data for the reaction rate coefficients after extrapolating the values to the temperature of the experiments if activation energies are available. Of course, applying aqueous phase data to subfreezing temperatures is possibly not appropriate. However, recent studies15,2-25 show continuous properties (same temperature dependence, for example) for reactions investigated in the liquid (aqueous) and solid (ice) phases, which confirm the assumption that most of the chemicalreactionstake place in the surfacelayer. We further do not attempt to calculate the pH in the QLL since this issue is still unresolved.24,26,27

Initial concentrations taken as input in our model have to be concentrations in the QLL. Since concentration measurements were made on bulk samples of the snow after melting, it is necessary to calculate the volume of the QLL relative to the bulk volume to derive the QLL concentrations. As given by

Cho et al.,21 the fraction of the QLL ( H O) compared to the bulk solid is a function of temperature and total solute

with mH O the molecular weight of water, R the gas constant, Tf the water freezingtemperaturein K, and Hf0 the meltingenthalpy of water. Since we assume that all impurities are located in the

QLL,14 the QLL concentration CT0 can be calculated from the bulk concentration (Cbulk) using this fraction of water

Equation 3 can be derived after substitution of CT0 in eq 1 by eq 2

With a given mechanism, the evolution of the concentration of each compoundcan be expressedusinga differentialequation. Depending on the mechanism used, we implemented different versions of our model with specific sets of coupled differential equations. The different sets of reactions and used rate coefficients are summarized in Table 1.

The concentrations of the considered species are simultaneously calculated using a unique time step of 10-2 s adjusted to give converging results in any case. Among the considered species, NOx can be present in both the QLL and the gas phase. We assume that the air surroundingthe snow during the artificial snow experiments is a “clean” NOx-free atmosphere, so that

NOx produced in the snow will only be transferred to the gas phase and not back to the condensed phase. This assumption is also based on the fact that the solubility of NO and NO2 in the QLL and their interaction with ice is very weak as suggested by previous studies.2

Error Calculation. For all simulations, we used an error quantification criteria based on the calculation of relative errors over each experimentalpoint,ratherthan classicalsum-of-square errors.28 We further averaged the errors using all points of the experiments as given by eq 4 with n the number of experimental measurements and x the

NO3- and NO2- concentrations.When severalexperimentaldata sets are used simultaneously, we further averaged the relative errors for all data sets.

Optimization Tool. For the optimization of unknown reaction rate coefficients or concentrations, we implemented a specific tool for our model to test all combinations of reaction rate coefficients (k) or initial concentrations ([C](t)0)) in a given range and computed the errors between simulated concentrations and experimental data. We first set the unknown experimental reaction rate coefficients to centered values k based on a previous study.14 Then, we performed calculations for a wide interval of values ranging from kcent

× 0.01 to kcent × 100 with a grid of nine different values (one centered value plus four smaller and four larger values) for each initial rate coefficient. The grid was regularly spaced on a logarithmic scale. Thus, the ratio between consecutive values corresponds to a factor of (100)1/4 ≈ 3.16. As a result, in the case of five unknown rate coefficients, 95 simulations need to be performed to consider all possible combinations. Next, we computed the errors between simulated concentrations and experimental data for all combinations. We repeated this procedure in a second step after changing the centered values kcent to the set of optimum rate coefficients kopt giving the lowest errors in the first step. We refined the grid with a n |xi,modeled - xi,measured| xi,measured (4)

B J. Phys. Chem. A, Vol. x, No. x, X Bock and Jacobi

reduced range from kopt × 0.1 to kopt × 10. In total, five steps like this with gradually reduced ranges were necessary to get converging results with a precision in accordance with the uncertainties inherent to the experiments.

Results and Discussion

Optimized Mechanism for the Photolysis of NO3- in the QLL of Snow. We used as a starting point the mechanism with

14 reactions for the NO3- photolysis (Table 1, reactions 1-14) proposed by Jacobi and Hilker14 that we implemented in our own code. This mechanism includes the transfer of NO and

NO2 from the condensed phase (in our case, the QLL) to the gas phase (Table 1, reactions 10 and 13) and rate coefficients at -31 °C. Jacobi and Hilker14 implementedthis set of reactions using a commercial software and compared the evolution of the concentrations of NO3-,N O2-, and NOx predicted by the model with experimental results. These experiments (see ref

14 for details) were carried out with four batches of artificial snow with differing initial NO3- and NO2- concentrations. We confirmedthat our newly programmedmodel deliveredidentical results for the four experiments compared to the commercial software used by Jacobi and Hilker14 (Figure 1). The averaged relative error using the 14 reactions for both NO3- and NO2 - and for the four batches was 48%, with much larger errors for

Another set of reactions for the photolysis of NO3- in snow was recently proposed by Boxe and Saiz-Lopez.13 This set of reactions is similar to the one proposed by Jacobi and Hilker14 even if some reactions are expressed differently: e.g., paths involving nitrogen oxide dimer formation (N2O3 and N2O4) and hydrolysis are expressed as single reactions each (Table 1, reactions 7 and 8) in the Jacobi and Hilker mechanism14 while they are described by two reactions by Boxe and Saiz-Lopez.13 However, Boxe and Saiz-Lopez13 used two additional reactions

(Table 1, reactions15 and 16) both transformingNO2- to NO3- .

We added these two reactions to the previous set of 14 reactions of Jacobi and Hilker14 to complement the reaction mechanism. In their work, Jacobi and Hilker14 adjusted and optimized the experimental photolysis rates for NO3- and NO2- (Table 1:

reactions 1, 2, and 4) and the transfer rates of NO and NO2 from the QLL to the gas phase (Table 1: reactions 10 and 13) by minimizing an error function comparing the four studied experimentsof artificialsnow and their modeledresults.Because of the two additional reactions 15 and 16 converting NO2- into

NO3-, we had to search again for the best reaction rate coefficientsfor the photolysisand transfer reactions.We applied the above-described optimization tool to find the best values for the five unknown reaction rate coefficients 1, 2, 4, 10, and 13. Initial values for the rate coefficientswere taken from Jacobi and Hilker.14

The procedure delivers an optimum set of rate coefficients from a purely mathematical point of view. It can be considered as the minimumvalue of the error functionin a five-dimensional space related to the five determined rate coefficients. However,

TABLE 1: Reactions and Rate Coefficients Used in the Mechanism for the Photochemical Transformations in Snow no. reaction reaction rate coefficienta (T )- 31 °C) reference a Reaction rate coefficients are extrapolated to -31 °C using available activation energies obtained in the aqueous phase. This concerns reactions 7, 1, and 17-21. b Not included in the reduced mechanism. c In this reaction, we explicitly used the hydrated form of HCHO, i.e.,

Nitrate Photochemistry in Snow J. Phys. Chem. A, Vol. x, No. x, X C

this absolute minimum may be related (and in fact it is) to a set of reaction rate coefficients that are unrealistic from a chemical point of view. Therefore, we added further constraints for the rate coefficients to exclude certain parts of the five-dimensional space in order to remain consistent with the results of previous studies. First, the NO2- photolysis rate (Table 1, reaction 4) needs to be faster than NO3- photolysis (Table 1, sum of photolysis rates 1 and 2), because the absorption of NO2- is stronger than NO3- over the entire wavelength range of the emission of the lamp used in the artificial snow experiments.16

Yet, both absorption spectra have similar magnitudes, so we used an upper limit to reject solutions with a NO2- photolysis rate larger than the NO3- photolysis rate by more than 2 orders

of magnitude. Second, the photolysis of NO3- to NO2 (Table 1, reaction 1) was limited to less than 10 times faster than the photolysis of NO3- to NO2- (Table 1, reaction 2). Indeed, several experiments carried out both in the aqueous phase29,30 and in ice (frozen aqueous solution)2-24 suggest such a maximum ratio between the two photolysis pathways. The experimental ratios cover a range of 0.25-8.4.2-24,31 Due to the experimentaluncertainties,we used a somewhat higher ratio of 10 for the calculations. Third, we rejected potential solutions with a discrepancy between the two transfers rates of NO and

NO2 (Table 1, reactions 10 and 13) higher than a factor of 15 as suggested by Jacobi and Hilker.14

The new optimized reaction rate coefficients are reported in

(Parte 1 de 3)