Uptake of Hydrogen Peroxide from the Gas Phase to Grain Boundaries: A Source in Snow and Ice

Hydrogen peroxide is a primary atmospheric oxidant significant in terminating gas-phase chemistry and sulfate formation in the condensed phase. Laboratory experiments have shown an unexpected oxidation acceleration by hydrogen peroxide in grain boundaries. While grain boundaries are frequent in natural snow and ice and are known to host impurities, it remains unclear how and to which extent hydrogen peroxide enters this reservoir. We present the first experimental evidence for the diffusive uptake of hydrogen peroxide into grain boundaries directly from the gas phase. We have machined a novel flow reactor system featuring a drilled ice flow tube that allows us to discern the effect of the ice grain boundary content on the uptake. Further, adsorption to the ice surface for temperatures from 235 to 258 K was quantified. Disentangling the contribution of these two uptake processes shows that the transfer of hydrogen peroxide from the atmosphere to snow at temperatures relevant to polar environments is considerably more pronounced than previously thought. Further, diffusive uptake to grain boundaries appears to be a novel mechanism for non-acidic trace gases to fill the highly reactive impurity reservoirs in snow’s grain boundaries.

For an uptake experiment, the flow of H 2 O 2 in humidified N 2 was diverted from the bypass line through a pre-cooled (253 K) PFA tubing housed within an ethanol-chilled cooling jacket and then entered the DIFT, which was placed within the same cooling jacket. The concentration of H 2 O 2 exiting the flow tube was monitored over one hour, much longer than the approximately 2 s residence time of the carrier gas, using an online analyzer situated downstream of the DIFT. At the end of each uptake experiment, the flow is directed back to the bypass line. Subsequently, a desorption experiment is conducted by introducing precooled and humidified N 2 gas into the bypass line and then into the DIFT while monitoring the peroxide signal. At the end of the desorption experiment, the flow is directed back to the bypass line. All experiments were done with one of the two DIFT tubes. Between experiments, the tubes were kept within the set-up with no gas flow, and at 253 K. Coated wall flow tube experiments were operated analogously.

Experimental, data processing
The direct observable in these experiments is changes to the gas-phase concentration of H 2 O 2 exiting either the DIFT or the by-pass line. The uptake to the ice is quantified as the difference between the by-pass signal to the uptake signal and that of the desorption as the difference between the desorption signal and that of the baseline. The by-pass signal showed drifts upwards during the day due to conditioning. This trend in the concentration of H 2 O 2 during the by-pass periods was accounted for by fitting and interpolating the by-pass periods directly before and after each uptake experiment ( Fig. SI-2). There might further be small desorption of H 2 O 2 from the transport wall lines after the experiments that only slowly ceases. A linear baseline fit when both the H 2 O 2 source and the DIFT tube were bypassed accounted for this. A control experiment revealed 0.1 × 10 12 molecules of H 2 O 2 cm -3 released from the MC-DIFT into H 2 O 2 -free carrier gas. The PC-DIFT did not show such a release when H 2 O 2 -free carrier gas flowed through the DIFT. We assign this background to traces of peroxides being released from the MC-DIFT tube that might originate from earlier experiments performed with the same flow tube. This background was subtracted during data processing. We assign the higher scatter in the MC-DIFT recoveries to day-to-day variability in this background H 2 O 2 release.

Results, drilled ice tubes
The total uptake of H 2 O 2 to the ice for the 60 min experiments resulted in apparent surface coverages of 3 × 10 14 molecules cm -2 to 6 × 10 14 molecules cm -2 for the MC-DIFT and 8 × 10 14 molecules cm -2 to 9 × 10 14 molecules S2 cm -2 for the PC-DIFT. For this analysis, the area between the total uptake curves and an interpolation of the bypass signal was integrated between t= 0 minutes and t= 60 minutes and divided by the geometric surface area S4 of the DIFT. The apparent surface coverage quantifies the uptake without differentiating between adsorption to the surface and uptake to other reservoirs, formally expressing the total uptake as adsorption. This approach has no physical meaning. It serves to compare the observed uptake with results from other flow tube studies that lack a detailed analysis of the uptake mechanism and often report such apparent surface coverages.
The four repetitions of the experiment in the same MC-or PC-DIFT, respectively, are color-coded in Figure 2 of the manuscript. Of these four experiments in the MC-and PC-DIFT each, two had the same concentration of H 2 O 2 in the gas phase, and the concentration was slightly varied for the other runs (see table SI-T2). The curves differ somewhat in their shape during the first 10 minutes of each experiment. We judge the reproducibility towards the end of each experiment at 60 min., the period from which we discuss the recovery level, as excellent. Uncertainty in these results comes mainly from instrumental variability and data processing. To assess the uncertainty from data processing, including the fits to account for shifts in the baseline and by-pass signal with time, each line represents the average of two data recordings and independent processing of the same experiment using both available channels of the H 2 O 2 analyzer. The shaded area illustrates the deviations between these two data recordings, which we judge as minor.   Fig SI-4; the area between the uptake curve and this fit gives the number of molecules adsorbed. This approach is similar to those previously used by 1, 2 . This analysis yields surface coverages of 4 10 13 molecules cm -2 at 235 K to 3 10 12 molecules cm -2 at 258 K for these experiments. Reynolds numbers were calculated for both flow regimes, resulting in Re = 440 for 2000 ml min -1 STP and Re = 110 for 500 ml min -1 STP, indicating laminar flow throughout all experiments. The dew point of the gas flow was compared to the dew point of the ice with a commercial dew point sensor (EdgeTech Dewmaster) before each experiment. For this, the dew point of the gas flow was measured; subsequently, the dew point after the CWFT was measured and compared to the temperature inside the CWFT and the dew point of the gas flow. If necessary, the temperature of the CWFT was adjusted to the dew point of the water vapor in the gas flow. The temperature inside the coated wall flow tube was measured at experimental conditions with a type K thermocouple.

The resistor model
The detailed derivation of the kinetic model can be found in 3 . Briefly, we assume that the quasi-stationary condition is satisfied and that there is a uniform velocity profile through the flow tube (plug flow assumption), leading to a linear flow rate ( ). Under this assumption, one expects an initial rapid drop in the trace gas number density due to adsorption. A first-order decay of the uptake towards zero follows this initial phase as the adsorption sites become filled. In Eq. 1, we model the H 2 O 2 number density at the time ( ) observed in the DIFT ( ), normalized to its initial concentration ( ). The uptake coefficient ( ) describes the net probability ( ) 0 ( ) that a molecule from the gas phase will enter the condensed phase as it collides with the surface. Note that, saturation effects and desorption from the surface result in a pronounced time dependency for the uptake coefficient 3 . (1) In addition to , the residence time of H 2 O 2 in the DIFT is also governed by known experimental variables, ( ) including the length of DIFT , the surface area of the DIFT borehole available for adsorption , the volume ( ) ( ) of DIFT borehole , and the mean thermal velocity of the gas molecules . We assume that the net uptake ( ) ( ) coefficient is a composite of two independent loss pathways to the surface through Langmuir adsorption, given in Equation 3, and loss to the bulk via diffusion, given in Equation 4: (2) To model the time-dependent surface uptake coefficient for H 2 O 2 ( (t)), we assume a Langmuir adsorption isotherm, which describes gas uptake to a finite number of independently available, non-interacting surface sites (n sites ). Then is given by the effective mass accommodation coefficient ( ), modulated by a first-order ( ) decay of the number of available sites due to the previous adsorption. The mass accommodation coefficient describes the fraction of molecules entering the condensed phase relative to those colliding with interface from the gas phase. It relates to the uptake coefficient by . At the time limit t→∞, =0 indicating the ( →0) = surface is at equilibrium, and there is no net surface uptake. This time dependency of is captured in the time constant ( ) for Langmuir adsorption, which describes the characteristic time to reach surface equilibrium. The bulk uptake coefficient ( that in principle captures diffusive loss into any bulk reservoir such as the ( )) ice crystal as well as grain boundaries is formulated using the "resistor" approach 4-6 . Diffusive loss from the gas phase is related to the mean thermal velocity of the gas , dimensionless Henry's law coefficient ( ) parameterizing the solubility of H 2 O 2 in bulk ice, and diffusivity of H 2 O 2 in bulk ice 3 . Since the surface ( ) accommodation and diffusion processes are contingent, their resistances are added in series. Substituting Equation (3) and (4) into (1), gives the overall uptake coefficient for our DIFT system, Equation 5. Similar equations describing uptake processes that couple the surface and bulk reservoirs have been derived elsewhere [6][7][8][9] .
To constrain the fits to Eq.5, we first fit the last 500 s of the uptake curves to the diffusive limit for the flow tube equation, where in the limit t → ∞, The parameters derived from these fits are presented in Table 1. A coupled adsorption-diffusion flow tube expression is used to fit the uptake curves over the entire period of observation (250 -3500 s): The Hsqrt(D) was constrained to the 95% confidence interval derived from fits using Eq. 7. The resulting extracted fit parameters are displayed in Table 1. Interestingly, the predicted effective mass accommodation coefficient and lambda are similar for the MC-and PC-DIFT. This indicates that adsorptive processes are identical in adsorption probability and energetics, irrespective of the number of grain boundaries in the ice sample. Compared to the reported value of the mass accommodation coefficient of 0.2 for aqueous surfaces at 273 K 10 , the results here reflect the strong influence of gas phase diffusion on the interface and also the fact that the gas undergoes multiple collisions with the flow tube in these experiments at atmospheric pressure.
Figure SI-5 presents an asymptotic fit to the last 500 s of the uptake curves. The equations and parameters derived from these fits are presented in Table 1. As expected, the parameterization for the solid-state diffusive limit (Eq. 7) fails to reproduce the uptake curves over the entire experiment for either the MC-or PC-DIFT results because the adsorptive uptake that dominates in the initial period of the uptake is neglected. The fits allowed deriving Hsqrt(D) that dominates the uptake at longer time scales for both types of DIFT tubes. The 95% confidence interval derived from the asymptotic fits was then used to constrain fits to the entire uptake curves (250 -3500 s) using a coupled adsorption-diffusion flow tube expression (Eq. 8) in Fig. SI-5b and as discussed in the manuscript.  Results, data table   Table SI