Photochemical reactions of contaminants in snow and ice can be important sinks for organic and inorganic compounds deposited onto snow from the atmosphere and sources for photoproducts released from snowpacks into the atmosphere. Snow contaminants can be found in the bulk ice matrix, in internal liquid-like regions (LLRs), or in quasi-liquid layers (QLLs) at the air–ice interface, where they can readily exchange with the firn air. Some studies have reported that direct photochemical reactions occur faster in LLRs and QLLs than in aqueous solution, while others have found similar rates. Here, we measure the photodegradation rate constants for loss of the three dimethoxybenzene isomers under varying experimental conditions, including in aqueous solution, in LLRs, and at the air–ice interface of nature-identical snow. Relative to aqueous solution, we find modest photodegradation enhancements (3- and 6-fold) in LLRs for two of the isomers and larger enhancements (15- to 30-fold) at the air–ice interface for all three isomers. We use computational modeling to assess the impact of light absorbance changes on photodegradation rate enhancements at the interface. We find small (2–5 nm) bathochromic (red) absorbance shifts at the interface relative to in solution, which increases light absorption, but this factor only accounts for less than 50 % of the measured rate constant enhancements. The major factor responsible for photodegradation rate enhancements at the air–ice interface appears to be more efficient photodecay: estimated dimethoxybenzene quantum yields are 6- to 24-fold larger at the interface compared to in aqueous solution and account for the majority (51 %–96 %) of the observed enhancements. Using a hypothetical model compound with an assumed Gaussian-shaped absorbance peak, we find that a shift in the peak to higher or lower wavelengths can have a minor to substantial impact on photodecay rate constants, depending on the original location of the peak and the magnitude of the shift. Changes in other peak properties at the air–ice interface, such as peak width and height (i.e., molar absorption coefficient), can also impact rates of light absorption and direct photodecay. Our results suggest our current understanding of photodegradation processes underestimates the rate at which some compounds are broken down, as well as the release of photoproducts into the atmosphere.
Snow and ice contain a wide variety of chemical compounds (Grannas et al., 2006), which can be transformed via photochemical reactions (Bartels-Rausch et al., 2014; Domine and Shepson, 2002; Grannas et al., 2007). While snow and ice are comprised primarily of crystalline water ice, under environmental conditions there are also small areas of disordered water molecules that contain most of the solutes present in a snowpack (Barret et al., 2011; Bartels-Rausch et al., 2014, 2017; Grannas et al., 2007; Jacobi et al., 2004). Although the terminology used in the literature can vary, at the air–ice interface these regions are commonly called quasi-liquid layers (QLLs), while those located at ice grain boundaries and other locations within the ice matrix are referred to as liquid-like regions (LLRs). Photochemistry can be important in snowpacks (Grannas et al., 2007), as light can penetrate tens of centimeters below the snow surface (France et al., 2011; Galbavy et al., 2007; Phillips and Simpson, 2005), and chemical species can exchange with the firn air in the snowpack. Photochemical reactions are classified as either direct – where a compound absorbs sunlight and is transformed – or indirect – where a reactive species (e.g., hydroxyl radical) formed from a direct photoreaction reacts with the compound of interest.
Despite their importance, only a small number of direct photochemical reactions have been studied in/on ice, with variable and occasionally conflicting findings. Measurements of direct photodegradation rates for a number of inorganic solutes (e.g., nitrate, nitrite, and hydrogen peroxide) found the same temperature dependence in aqueous solution and LLRs, suggesting both compartments provide similar environments for chemical reactions (Chu and Anastasio, 2003, 2005, 2007). The picture is more complicated for PAHs (polycyclic aromatic hydrocarbons). Two studies found little difference in PAH photochemistry in/on ice compared to solution: phenanthrene, pyrene, and fluoranthene had similar photodegradation rates in aqueous solution and in LLRs (Ram and Anastasio, 2009), while anthracene and pyrene had similar rates in aqueous solution, in LLRs, and at the air–ice interface (QLLs) (Hullar et al., 2018). However, two other studies reported that the photodecay of anthracene and naphthalene was faster in LLRs and at the air–ice interface compared to in solution (Kahan and Donaldson, 2007; Kahan et al., 2010b). Harmine has also been reported to photodegrade faster at the air–ice interface (Kahan et al., 2010a). Most recently, we found that guaiacol photodegradation was somewhat faster in LLRs and considerably faster at the air–ice interface than in aqueous solution (Hullar et al., 2020).
To evaluate the possible causes of photodegradation enhancements in/on ice
compared to solution, consider the variables that control the direct
photodecay rate (M s
Many previous studies did not measure photon fluxes at the point of the reaction, so it is difficult to accurately determine the significance of local flux differences in accounting for photodecay enhancements in, or on, ice. However, measurements in different solute locations, e.g., in solution, in LLRs, and at the air–ice interface, found that photon fluxes varied by less than a factor of 1.5 (McFall and Anastasio, 2016). In addition, in our recent work with guaiacol we normalized photodecay rate constants for loss by photon flux but still saw large differences in rate constants between solution, in ice, and at the air–ice interface (Hullar et al., 2020). Thus local photon flux differences do not appear to a major factor in observed reaction rate enhancements in/on ice.
Because natural solar photon fluxes increase by several orders of magnitude between 295 and 400 nm (Madronich and Flocke, 1999), even a small shift in compound absorbance towards longer wavelengths (i.e., a red, or bathochromic, shift) could substantially increase the amount of sunlight absorbed by a compound, increasing its reaction rate. Several studies have measured absorbance shifts for compounds in LLRs and at the air–ice interface relative to solution (Corrochano et al., 2017; Heger et al., 2005; Heger and Klán, 2007; Kahan and Donaldson, 2010; Kania et al., 2014; Krausko et al., 2015; Malongwe et al., 2016; Matykiewiczová et al., 2007). The absorbance of some compounds was the same as in aqueous solution, with others showing shifts of up to 15 nm either to the red or blue (i.e., a hypsochromic shift); for several compounds concentrated in LLRs, shifts of up to 100 nm were reported (Heger and Klán, 2007). However, these large shifts were attributed to aggregated test compounds and resulting intermolecular interactions, rather than individual molecules. Unfortunately, measuring a compound's absorbance at the air–ice interface can be challenging, particularly when using low concentrations representative of environmental conditions. Accurate absorbance measurements typically require relatively high concentrations, which can lead to aggregation on the ice surface, potentially influencing the absorption characteristics. To avoid this problem, we recently used molecular modeling to estimate the absorbance shift for guaiacol at the air–ice interface (Bononi et al., 2020; Hullar et al., 2020). While we did find a slight bathochromic shift (5 nm), this shift explained less than 10 % of the enhanced reaction rates experimentally measured at the interface.
Finally, an increased quantum yield at the air–ice interface could explain a faster reaction rate, due to a greater fraction of absorbed photons resulting in photochemical reaction of the chemical. Some studies suggest LLRs and solution represent similar reaction environments (Chu and Anastasio, 2003, 2005, 2007; Ram and Anastasio, 2009), while others have found higher quantum yields at the air–ice interface (Hullar et al., 2020, 2018; Zhu et al., 2010). Our recent work with guaiacol (Hullar et al., 2020) found that changes in the quantum yield were the dominant contributor to reaction rate differences between aqueous solution, LLRs, and QLLs, with values up to 40-fold higher at the air–ice interface compared to solution.
Taken together, previous studies show the importance of determining various
factors to understand the reasons for enhanced chemical reaction rates in
snow and ice. In particular, our recent results (Hullar et al., 2020)
indicate that the direct photodecay of guaiacol is different in aqueous
solution, LLRs, and QLLs and demonstrate how molecular modeling can be
used to assess the relative contributions of changes in light absorbance and
quantum yield. Here, we extend those results to three additional organic
compounds chosen as model aromatics in the environment: 1,2-,1,3-, and
1,4-dimethoxybenzene (abbreviated 1,2-DMOB, 1,3-DMOB, and 1,4-DMOB,
respectively; chemical structures given in Fig. 1). DMOBs can be emitted
into the atmosphere by biomass burning (Smith et al.,
2020). Several studies have examined the direct photodegradation of DMOBs
and methoxybenzene (anisole), but few have used wavelengths relevant to
tropospheric sunlight. At wavelengths greater than 290 nm, 1,2- and 1,3-DMOB
have been reported to photodegrade slowly, with 1,4-DMOB loss being somewhat
faster (Amalric et al., 1993). 1,2-DMOB in acetonitrile forms a
triplet excited state when illuminated with 418 or 514 nm radiation
(Schurmann and Lehnig, 2000). Aqueous 1,4-DMOB excited at 266 nm forms
a triplet excited state, which decays to a solvated electron and a
relatively long-lived organic radical cation (Grabner et al., 1996, 1980). Another study (Tajima et al., 1999) with
266 nm excitation under acidic conditions (pH
Here, we measure the direct photochemical reaction rate constants for loss
of the three DMOB isomers in aqueous solution, LLRs, and QLLs, normalizing
each to the measured photon flux for a given sample type. To assess the
contribution of absorbance shifts, we model DMOB absorbance in aqueous
solution and on an ice surface. As with guaiacol, the DMOBs are all
doubly substituted aromatic rings; however, the hydroxyl group of guaiacol
is replaced by a methoxy group, eliminating the possible unwanted reaction
with triplet excited states (
1,2-, 1,3-, and 1,4-DMOB (99 %,
We placed samples in 10 mL glass beakers (Pyrex) and covered them with nylon
film (McMaster-Carr, approximately 25
Sample illumination followed the method described previously (Hullar et
al., 2020). We set sample beakers upright in a drilled aluminum block set
within a temperature-controlled chamber; dark samples were completely
covered with aluminum foil and placed in the aluminum block next to the
illuminated samples. The samples were held at 5
After illumination, we melted the frozen samples and measured DMOB
concentration using a Shimadzu HPLC (Hullar et al., 2018) with an eluent
of 60 : 40 acetonitrile : MQ water, a flow rate of 0.70 mL min
To account for differing photon fluxes across samples types and experiment
days, we used 2-nitrobenzaldehyde (2NB) as a chemical actinometer
(Galbavy et al., 2010; Hullar et al., 2020, 2018). Except for
snow samples, we prepared 10
We used the Tropospheric Ultraviolet and Visible (TUV) model (Madronich and Flocke, 1999) to model spectral actinic
fluxes for Summit, Greenland, at noon on the summer solstice (subsequently
referred to as “Summit conditions”). We used default settings (option 1)
except for wavelength interval
We determined DMOB photodegradation rate constants for loss using the same
approach as for guaiacol and PAHs (Hullar et al., 2020, 2018). We illuminated samples with simulated polar sunlight, periodically
removing a beaker (and corresponding dark beaker) for analysis. To determine
the photodegradation rate constant for loss, we first calculated the natural
logarithm of the ratio of the DMOB concentration at time
We calculated quantum yields for each DMOB using methods described
previously (Hullar et al., 2020). In short, the quantum yield was
estimated for each DMOB by dividing the dark-corrected experimental
photodegradation rate constant (
To investigate possible shifts in light absorbance at the air–ice interface for the three dimethoxybenzene isomers, we used a multimodel approach that combines classical and first-principles molecular dynamics (FPMD) simulations, excited state calculations using time-dependent density functional theory (TDDFT), and machine learning (ML) (Bononi et al., 2020; Tibshirani, 2011).
As in our recent work on phenol and guaiacol, models of DMOB in aqueous
solutions and at the ice surface were equilibrated in classical MD
simulations using the LAMMPS code (Plimpton, 1995), the OPLS force field, and the TIP4P/ice water model
(Abascal et al., 2005). To model the air–ice interface we
utilized an ice slab model, which captures a well-equilibrated surface
structure and reproduces recent measurements for QLLs (Kling et al., 2018;
Sanchez et al., 2017). We then performed FPMD simulations of the DMOB isomers
in solution at 27
As a refinement to our former approach, we developed a universal ML model to
predict the absorption spectra for all three DMOB isomers. To accomplish the
transferability, we adopted a more sophisticated atomic descriptor – the
bispectrum component (BC) (Bartok et al., 2013; Thompson et al., 2015). The BC
describes each molecule's atomic environment by projecting the weighted
atomic densities to four-dimensional hyperspherical harmonics, and it has
been previously applied to ML interatomic potential development and material
property predictions (Cusentino et al., 2020; Legrain et al., 2017). By
using the BC with the least absolute shrinkage and selection operator (LASSO)
regression model (Tibshirani, 2011), we
attain a more precise estimate of the low-energy, long-wavelength tails of
the spectra, which are important for calculating rates of photon absorption
since the photon flux is increasing in this region. To assess the relative
contributions of the phenyl ring and methoxy groups to the light absorbance
of each DMOB isomer, we decomposed the predicted peak wavelength from over
We prepared samples using one of several methods designed to place the DMOB
isomer into aqueous solution, LLRs, or at the air–ice interface (Sect. 2.2 and 2.3). Then, we illuminated the samples, periodically removing them
for analysis. Figures S1 through S12 in the Supplement show the results for every
illumination experiment, with each data point representing one sample
beaker. Generally, dark controls show slight loss of DMOB, probably
attributable to volatilization; illuminated samples often show considerably
greater loss due to photodegradation, but the extent of loss depends on DMOB
isomer and sample preparation method. Figure 1 summarizes the experimental
results for each of the three DMOBs in aqueous solution and the various
frozen sample preparations. As described above, we divided each
dark-corrected, measured rate constant for DMOB loss by the corresponding
measured
Photon-flux-normalized photodegradation rate constants for loss
for each dimethoxybenzene isomer (
As shown in Fig. 1a, the 1,2-DMOB photodegradation rate in aqueous
solution is slow, and the normalized rate constant for loss is statistically
indistinguishable from zero. For frozen solution experiments, the average
rate constant was negative, and the data were quite noisy. Samples frozen
with liquid nitrogen (“LN2”) should, like freezer-frozen samples, place
solutes primarily in internal LLRs. However, the variability in 1,2-DMOB LN2
experiments is considerably less than for freezer-frozen experiments, and
the rate constant is roughly equivalent to that determined for aqueous
solution. Previous work (Hullar and Anastasio, 2016) suggests more
homogeneous solute distribution in LN2 samples compared to frozen solution
samples, which may explain the reduced experimental variability in LN2
samples. The reduced variability might also be due to the fact that freezing
in LN2 is fast (less than 90 s), which reduces the time available for
the DMOB to react as solutes concentrate during freezing; in contrast, the
freezer requires much more time (typically several hours) to make ice, which
can lead to more, and more variable, DMOB loss. For both frozen solution and
LN2 treatments, the rate constants are indistinguishable from zero. The two
treatment methods which put 1,2-DMOB at the air–ice interface, VD to ice and
VD to snow, both show normalized rate constants for loss approximately 15 times faster than in aqueous solution or in LLRs. However, while
experimental results for the VD-to-ice treatment are highly variable (with
an average rate constant indistinguishable from zero), VD-to-snow
experiments are more reproducible and give a normalized rate constant
statistically greater than zero, showing the advantage of using
nature-identical snow to study photodegradation at the air–ice interface. As
discussed previously (Hullar et al., 2020), the specific surface area
(SSA) for our VD-to-snow samples (approximately 600 cm
1,3-DMOB results are summarized in Fig. 1b. Because the frozen solution and VD-to-ice experiments were very noisy for 1,2-DMOB, we did not run experiments with these sample treatments for 1,3-DMOB. For aqueous solution, the 1,3-DMOB average rate constant for loss is slightly negative and indistinguishable from zero. In LLRs (LN2 sample treatment), 1,3-DMOB photodegrades at a moderate rate, statistically greater than zero. Finally, at the air–ice interface (VD-to-snow samples), the photodegradation rate constant is approximately 4 times faster than in LLRs although statistically indistinguishable from zero because of very high variability.
For 1,4-DMOB in aqueous solution (Fig. 1c), the average photodegradation rate is slow but statistically greater than zero. As with 1,3-DMOB, we did not run experiments in frozen solution for 1,4-DMOB; however, LN2 experiments, which should also place solutes primarily in LLRs, showed photodecay rates both statistically greater than zero and approximately 3-fold faster than in aqueous solution. Measured VD-to-ice rates were variable, and although the average normalized rate constant for loss was similar to LN2, it was not statistically different than zero. As with 1,2-DMOB, the average 1,4-DMOB photodegradation rate constant at the air–ice interface (VD-to-snow experiments) is considerably faster than in either aqueous or LLR compartments, with a 26-fold enhancement relative to aqueous solution, and is statistically greater than zero.
To determine if the various sample treatment rate constants are
statistically different from each other, we used the Tukey–Kramer test for
multiple comparisons (
Table 1 presents the rate constant enhancements for each frozen sample type
relative to aqueous solution; Table S3 provides details for the
various measured and computed experimental parameters. For 1,2-DMOB,
photodegradation proceeds at approximately the same rate in LLRs and aqueous
solution but roughly 15-fold (
Summary statistics for each experimental preparation method
Figure 2 presents the wavelength-dependent molar absorption coefficients for 1,2-, 1,3-, and 1,4-DMOB, as well as guaiacol (which was studied in our previous work; Hullar et al., 2020). 1,2- and 1,3-DMOB in solution have nearly identical absorbance curves, with maximum absorbance at 274 and 273 nm, respectively. While guaiacol absorbs less strongly, its curve shape and peak location are similar to 1,2- and 1,3-DMOB. In contrast, 1,4-DMOB absorbs at longer wavelengths, with a peak absorbance at 287 nm. For comparison, the two black lines in Fig. 2 show the photon flux of our experimental system (dashed line) and the modeled actinic flux for Summit conditions (solid line); a more detailed graph is shown in Fig. S13. While the actinic flux at Summit starts at approximately 297 nm and increases quickly with increasing wavelength, the experimental flux begins earlier (roughly 280 nm) and increases more gradually. 1,2- and 1,3-DMOB in solution absorb small amounts of light under our illumination conditions and virtually none in the Arctic environment. In contrast, the 1,4-DMOB absorbance curve has substantial overlap with both photon flux curves and therefore absorbs light under both experimental and natural conditions.
Light absorption spectra for the dimethoxybenzene (DMOB) isomers and guaiacol, along with photon fluxes in our experiments and for Arctic summer conditions. Solid colored lines are the measured molar absorption coefficients for each DMOB isomer, while dashed colored lines are predicted absorbance spectra at the air–ice interface, estimated using the results of our molecular modeling. The solution guaiacol spectrum (dotted grey line) is provided for comparison to previous work (Hullar et al., 2020). Black lines (right axis) represent the modeled actinic flux for Summit conditions (solid line) and the photon flux measured in our laboratory illumination system (dashed line).
While we can measure light absorption by the DMOB isomers in solution, we also want to understand their absorption at the air–ice interface. To estimate this, we use molecular modeling combined with machine learning for each compound in aqueous solution and at the air–ice interface; these modeled curves are shown in Fig. 3. As shown in Fig. S14, modeled absorbance bands for aqueous DMOBs peak at longer wavelengths (7 to 21 nm) compared to measurements, equal to or greater than the 7 nm difference we observed for guaiacol (Bononi et al., 2020; Hullar et al., 2020). These differences are caused by systematic underestimation in our simulations, which is a known limitation of TDDFT calculations; the peak wavelength offset relative to measured spectra tends to increase with larger molecules (Leang et al., 2012; Miura et al., 2007), consistent with the greater difference here for the DMOB isomers compared to our previous work with guaiacol. These differences can be corrected by applying the same shifts to both solution and ice spectra (Ge et al., 2015).
Modeled absorbance spectra in aqueous solution (solid lines) and
at the air–ice interface (dashed lines) for each DMOB isomer. Absolute
absorbance values are arbitrary but accurately reflect the relative
absorbance differences between isomers and conditions. Temperatures were 27
While the modeling does not accurately reproduce the absolute wavelengths of
absorbance, it provides useful insights into the differences between
absorbance in aqueous solution and at the air–ice interface. We note that a
similar modeling approach comparing phenol absorbance in gas and aqueous
phases successfully predicted the experimentally observed
We also used the molecular model results to assess the relative contributions of the phenyl ring and methoxy groups to the light absorbance of each DMOB. As indicated in Fig. S15, small geometrical changes in the phenyl ring are primarily responsible for the shifts in the absorption spectra for all three DMOB isomers, while the methoxy groups make a minor contribution. Changes in the geometry of the phenyl ring are responsible for 95 %–98 % of the light absorbance shifts in aqueous solution and 96 %–98 % at the air–ice interface. These findings are consistent with our previous work on guaiacol (Bononi et al., 2020; Hullar et al., 2020). Overall, these results suggest that differences in the atomic environments around the aromatic ring modify its geometry and determine their vertical excitation and are the primary factor controlling light absorption changes between aqueous solution and the air–ice interface.
As seen in Fig. 2, the predicted spectrum for each isomer at the air–ice interface (dashed colored line) is noticeably different than the measured aqueous spectrum (solid colored line), with bathochromic peak shifts and changes in absorbance spectrum shape. To assess the impact of these changes on light absorbance, for each isomer we multiplied the aqueous and air–ice interface wavelength-specific molar absorption coefficients by the experimental or Summit photon fluxes to determine the rate constant for light absorbance at each wavelength (Fig. S16). For each DMOB isomer, the rate constant for light absorbance is a wavelength-specific value giving the rate at which photons are absorbed per molecule of test compound for a given light condition. We then summed the wavelength-specific values to obtain the overall rate constant for light absorbance in aqueous solution and at the air–ice interface for each isomer, for laboratory and Summit light conditions (Table S4). Because all three isomers show bathochromic absorbance shifts at the air–ice interface relative to aqueous solution, the overall rate constants for light absorption are generally higher at the air–ice interface. 1,3-DMOB, which has the largest absorbance spectrum bathochromic shift (5.2 nm), shows the largest change in overall light absorption, with a 5.3-fold increase relative to aqueous solution for experimental light conditions; for Summit actinic flux, the rate constant of light absorption increases by a factor of 170 from solution to air–ice interface. Conversely, the light absorption peak for 1,4-DMOB shifts only slightly from solution to ice and has a greater overlap with the photon flux curves in solution, so the rate of light absorption increases only slightly (10 % or less) from solution to the air–ice interface. These results show that the amount of light absorbed can be dramatically affected by absorbance changes and that this effect depends strongly on the position of the absorbance spectrum relative to photon fluxes and on the magnitude of the absorbance shift on ice. Comparing the overall light absorbed under laboratory versus Summit light conditions, 1,2-DMOB in either aqueous solution or at the air–ice interface absorbs around 200 times as much light in our lab system compared to Summit, while for 1,4-DMOB the light absorption is approximately equal in both systems. 1,3-DMOB presents a more complex picture: in aqueous solution, the rate constant of light absorption is about 400-fold greater under laboratory illumination compared to Summit light conditions, but at the air–ice interface, light absorption is only 12-fold greater in the lab relative to Summit conditions due to the absorbance shift on ice. For 1,2- and 1,3-DMOB, wavelengths from 275 to 295 and 295 to 315 nm are most photochemically important for lab and Summit light conditions, respectively; for 1,4-DMOB, these ranges are 280–315 and 300–320 nm.
Our observed increases in photochemical degradation rates at the air–ice interface can be caused by increases in light absorbance or quantum yield or a combination of both. As shown previously (Hullar et al., 2020), by solving Eq. (1) for quantum yield, we can use the calculated enhancements in the rate constant of light absorbance from our modeling results to estimate how quantum yields change from solution to the air–ice interface. Using the measured aqueous and predicted ice spectra for each compound, we calculated the quantum yields for each isomer under various conditions (Table 1). Our experimental results suggest LLRs may represent an environment different from either aqueous solution or QLLs. However, we did not model light absorbance changes in LLRs, and the available literature is inconclusive on the likelihood of absorbance shifts in LLRs, so for the quantum yield calculations we assumed our test compounds have the same molar absorption coefficients in LLRs as in aqueous solution.
For 1,2-DMOB, our quantum yield in aqueous solution calculated from our
experimental results is 0.015 (
Next, we evaluated the relative contributions of increased light absorbance and larger quantum yields to the photodegradation rate enhancements at the air–ice interface relative to solution. For 1,2- and 1,4-dimethoxybenzene, the faster photodegradation on ice is primarily due to an increase in quantum yield. In contrast, for 1,3-DMOB, the enhanced photodegradation at the air–ice interface is roughly equally due to increases in quantum yield and light absorbance. As in our earlier work with guaiacol (Hullar et al., 2020), light absorbance changes are never the dominant factor controlling rate constant enhancements. Increased light absorption accounts for 16 %, 49 % or less, and 4 % of the reactivity enhancement on snow relative to aqueous solution for 1,2-,1,3-, and 1,4-DMOB, respectively. Thus, higher quantum yields account for the bulk of the enhancement seen at the air–ice interface, accounting for 84 %, at least 51 %, and 96 % of the observed enhancements, respectively. These results are roughly consistent with our previous observations for guaiacol, where the quantum yield increased at the air–ice interface by a factor of 41, accounting for 95 % of the overall 77-fold increase in reactivity compared to aqueous solution (Hullar et al., 2020).
To assess the environmental significance of our findings, we calculated
dimethoxybenzene photodegradation rate constants for loss and photochemical
lifetimes in each compartment for Summit, Greenland, conditions (Table 1).
For these calculations, we used modeled actinic fluxes at Summit (Sect. 2.3) and our estimated quantum yields (Sect. 3.2); because our
computational modeling did not include LLRs, we used measured aqueous
spectra to represent absorbance in both aqueous and LLR compartments and
our predicted ice spectra (Fig. 2) for the air–ice interface. 1,2-DMOB has
slow photodegradation rate constants and very long photochemical lifetimes
(
As discussed above, enhanced reactivity at the air–ice interface is
primarily due to increases in the quantum yield, ranging from at least
5-fold (1,3-DMOB) to 41-fold (guaiacol) (Hullar et al., 2020). However,
although we can predict absorbance shifts at the air–ice interface using
molecular modeling techniques, we cannot currently predict quantum yield
changes using either computational or experimental methods. While changes in
quantum yields affect photodegradation rate constants linearly – a doubling
of quantum yield will double the rate constant for loss – absorbance shifts
cause nonlinear effects. To evaluate the impact of absorbance shifts on
compound photodegradation, Fig. 4a shows the calculated ratios of
absorbance-shifted rate constants to the unshifted rate constant. We
estimated environmental (
Predicted changes to photodegradation rate constants for loss and
lifetimes resulting from absorbance shifts for DMOB isomers. Rate constants
for loss were determined using calculated aqueous quantum yields, aqueous
absorbance spectra shifted either hypsochromically (towards shorter
wavelengths) or bathochromically (towards longer wavelengths), and either
experimental photon fluxes (dashed lines) or the modeled actinic flux for
Summit conditions (solid lines).
While the impact of a red shift in absorbance can be dramatic, this does not necessarily translate to a short lifetime. For example, a 10 nm red shift for 1,2-DMOB increases the rate constant for photodegradation by a factor of 90 (Fig. 4a), but this only reduces the lifetime from 23 000 to 260 d (Fig. 4b). 1,3-DMOB, which has essentially the same absorbance spectrum, behaves similarly (Figs. 4, 5, and S17). The behavior of 1,4-DMOB is different, however, since it overlaps the most with the solar spectrum (Fig. 2): while its rate constant for loss is less sensitive to a shift in absorbance (e.g., increasing by a factor of 16 for a 10 nm red shift), this changes the lifetime from 71 to 4.5 d (Fig. 4b), which is short enough to be significant for its environmental fate.
To generalize our experimental findings to other chemicals, we calculated
photodegradation rate constants and lifetimes for a hypothetical model
compound with an assumed Gaussian absorbance spectrum under Summit
conditions and with a quantum yield of 1. We first made a single absorbance
curve for a hypothetical model compound and then evaluated the impact of three
variables: peak position, peak width, and peak height. We represented the
model compound absorbance spectrum as a Gaussian curve with its peak at 280 nm, peak height (molar absorption coefficient) of 3000 M
We then evaluated the impacts of shifting the peak position widely, by
Predicted changes to photodegradation rate constants for loss
(
Next, we examined the impact of peak width, as illustrated in Fig. S22.
From our modeling, the largest peak width change was approximately 2 nm (for
1,3-DMOB). As seen in Fig. S23, narrowing the hypothetical
peak from 7 to 5 nm reduces
Finally, we evaluated the impact of changing the peak height (hyper- and hypochromic shifts). Figure S24 shows the spectra tested and Fig. S25 the results; for comparison, our largest modeled peak height change was 17 %, for 1,2-DMOB. Because the area of a Gaussian curve is proportional to its peak height, doubling the height doubles the area, and therefore the light absorbed would double as well. However, compared to the impact of peak location and width, even a peak height doubling exerts a relatively small influence on peak area and therefore light absorbed. To evaluate the relative impact of absorbance shifts, broadenings, and peak height (molar absorption coefficient) changes on photodegradation, we assumed the largest modeled absorbance changes between aqueous solution and at the air–ice interface for the three DMOB isomers are typical for chemicals in the environment. Based on this assumption and applying these changes to our hypothetical peak, peak location and width changes at the air–ice interface probably control overall differences in light absorption, while changes in peak height likely make a minor contribution.
Our results, together with previous studies (Hullar et al., 2020; Kahan and Donaldson, 2007; Kahan et al., 2010a, b), suggest that for some organic compounds, QLLs and LLRs represent different photochemical reaction environments that are distinct from aqueous solution. While molecular modeling and laboratory measurements have both found evidence of absorbance shifts (Corrochano et al., 2017; Heger et al., 2005; Hullar et al., 2020; Malongwe et al., 2016), our results indicate that increases in quantum yield are the major reason for enhanced photochemical reactivity at the air–ice interface. For compounds absorbing appreciable amounts of sunlight in aqueous solution, QLL and LLR reactivity increases may cause environmentally significant changes in direct photoreaction rates and lifetimes, but for chemicals that absorb very little or no sunlight, these changes do not appear to make direct photochemistry a significant sink.
Our ability to make statistically significant conclusions depended on the choice of the experimental treatment; samples frozen in liquid nitrogen or vapor-deposited to nature-identical snow provided useful insights into LLR and QLL compartments, respectively. In contrast, samples frozen in a laboratory freezer or vapor-deposited to a water ice surface gave results that were noisier and less valuable. In addition, computational methods allowed us to determine absorbance spectra at the air–ice interface, where experimental observations would have been difficult.
While we find that quantum yields at the air–ice interface can be much higher than in aqueous solution, the reason for this remains unclear. Our modeling suggests small geometric changes in the configuration of the phenyl ring shifts molecular absorbance, and it is possible the change in the carbon atom positions could also increase the quantum yield. Despite our use of small amounts of DMOBs deposited to the air–ice interface, which should reduce aggregation and areas of high local concentration, the observed photodegradation rate constant enhancements might be caused by higher local concentrations at the air–ice interface, increasing secondary chemistry. An additional possibility for the higher quantum yields is a weakening of the cage effect at the air ice interface. In solution (including in LLRs), the chromophore is surrounded by a cage of water molecules, which can inhibit dissociation of the excited state into products. At the air–ice interface, however, this cage will be incomplete since the molecule is exposed to air on one side. This reduced cage should increase the efficiency of the excited state decaying into products, leading to a higher quantum yield. A reduced cage effect was proposed by Meusinger et al. (2014) to possibly explain enhanced photodecay of nitrate ion in natural snow studied in the lab, although later work in Antarctica (Barbero et al., 2021) found no enhancement in nitrate quantum yield in the field.
We used publicly available software to conduct the research and write the paper, including two commercial products, Microsoft Word and Microsoft Excel. For the modeling work, we used open-source tools: Quantum Espresso (QE) (
Data for the experimental portion of this work are included in the Supplement. For our computational efforts, data are available at
The supplement related to this article is available online at:
Overall, TH, TT, OP, and CA were responsible for the experimental portion of the research, while ZC, FB, and DD did the modeling work. Research oversight was provided by DD and CA. TT, OP, and FB performed most of the experimental and modeling efforts, with assistance and guidance from TH. ZC developed the machine learning model. TH wrote most of the manuscript, with additional text from ZC. TT, ZC, FB, DD, and CA commented on the paper drafts, with significant input from DD and CA.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Calculations were performed using the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by the National Science Foundation (grant no. ACI-154856).
This research has been supported by the National Science Foundation (grant nos. CHE 1806210, AGS-PRF 1524857, and ACI-154856).
This paper was edited by Sergey A. Nizkorodov and reviewed by Dominik Heger and one anonymous referee.