The same fitting procedure was carried out on all the dark reactions. It was found that the dark reactions measured earlier in the campaign at 300 K could not be fitted without a large[Br]0. The[Br]0 was therefore initialised to the steady state
defined by[Br2], krec, and jBr
2 (Equation 3.10). While this equilibrium may not be
established in reality, the assumption allows the optimised jBr
2 to provide a measure
of the level of un-recombined Br·, whether due to photolysis inside the chamber from the LEDs or outside from the lab lights. The fit quality achieved with this method was always good. The jBr2 that gave the best fit for the dark reactions of the whole campaign are shown in Figure 4.24.
The first weeks show a high jBr2, which was earlier determined to be caused by the
CEAS LEDs (Section 4.4.6). The blue cavity was disabled on day 25, yet an abrupt reduction in the effective jBr
with only the UV cavity on day 30, with a jBr2 of 3×10 s , was also the lowest
dark[Br2] measured, and it is probable that the Br2 was photolysed as it was being injected by some external light source. As it became clear that residual photolysis was a problem, the main lab was kept dark and the only illumination was from a small red LED and the various computer screens. The exact details of the photolytic flux incident on the Br2 syringe during the journey from the fume hood in the room adjacent to the chamber were not recorded, as well as other variables such as in the early days a white LED may have been used for illumination, or the light may have been directed into the chamber during Br2 injection instead of switching it off as soon as the needle was in position. The CEAS cavities were also re-aligned every few weeks, which may have increased or decreased the light in the chamber. After experiments resumed on day 96 after a period of laser maintenance, the protocol was more or less fixed. Unfortunately the only room temperature reactions with only Br2 carried out in the latter part of the campaign were reactions with altered fan speed, so are not suitable as a room temperature reference. If we assume that the background kthm was about 10−6s−1 throughout the latter period of the campaign,
we can calculate how changes in kreccould explain the variation.
Figure 4.25 shows the effective krec required to fit the dark reaction assuming a
constant jBr
2 of 10
−6s−1, plotted against temperature of reaction (if plotted against
date the pattern is simply the inverse of Figure 4.24). A significantly faster krec was cited as possible explanation for the results in Section 4.7.3, and Figure 4.25 shows how krecmay have changed throughout the campaign. The red and blue lines are the default values for krec at 298 and 265 K, yet almost none of the results lie within those bounds. The cold reactions are described by a much faster krec, and
the warmer reactions by a slower one, indicating that krec was substantially more
temperature dependent than the literature suggests.
However, the assumptions underlying Figure 4.25 are dubious or incorrect. The 298 K data all had much higher LED photolysis than the value assumed in this calculation, and there are likely higher wall losses of Br·at cold temperatures. The point on this graph with the coldest temperature (−9 °C) is the same experiment as the point in Figure 4.9 with the lowest [Br2] at 265 K, and it was described there as not distinguishable from Hg loss to walls. The addition of extra gases and surfaces would also provide more loss mechanisms for Br·, so it is not surprising these reactions show an increased effective krec. Clearly there are far too many
variables to measure krec in this way.
10-15 10-14 10-13 10-12 10-11 -10 -5 0 5 10 15 20 25 30 effective k rec (cm 3 molec -1 s -1 ) temperature (°C) 298 K 280 K 265 K NOx ozone FEP SSA amylene fans
FIGURE4.25: Trend of effective krecrequired to model the dark reaction with temperature, assuming constant jBr2 of 10−6s−1. The blue and red lines are the literature rates for krecat 265 and 298 K.
There were likely contributions from further protocol development, differing wall losses as the chamber conditions changed. Actual differences in krec are unlikely to
have contributed significantly to the observed changes in the dark reaction.
4.7.7
Conclusions
The oxidation of Hg in the chamber was generally slower than predicted by the most reliable literature data. The stepwise adjustment of each parameter of the model was a promising method of using all of the[Hg] decay data to separate the different reactions. However, without any measurements of the reaction intermediates or products the model was overdetermined so no unique solution exists.
If the poorer quality data at the extremes of[Br2] are discounted then altering any of kfox, krec, kcpa or kabs would fit most of the data. Of these parameters a reduction in kfox by a factor of two was the smallest adjustment of a single parameter that
could fit most of the data.
If the high and low[Br2] are included then only an extremely fast, temperature- independent kthm could fit all the points. There is no known mechanism that could produce such a fast HgBr dissociation that is consistent with the thermodynamics
of the HgBr bond. In the final section of this chapter the relevance of a fast HgBr dissociation to environmental studies is considered.