Introduction
The uptake and reaction of water-soluble volatile organic compounds (VOCs) in
cloud droplets and aerosol liquid water is likely a significant source of
secondary organic aerosol (SOA) material (Carlton et al., 2008; Fu et al.,
2008, 2009; McNeill, 2015; McNeill et al., 2012). These processes may be
referred to, collectively, as aqueous SOA (or aqSOA) formation.
Glyoxal (CHOCHO, GLYX) and methylglyoxal (CH3C(O)CHO, MGLY) are both
atmospherically abundant gas-phase oxidation products of multiple VOC
precursors, including isoprene and toluene. Both GLYX and MGLY are
water-soluble, GLYX more so than MGLY (Betterton and Hoffmann, 1988; Zhou and
Mopper, 1990). Taking into account salting effects, the effective Henry's law
constant for GLYX in aerosol water can be several orders of magnitude higher
than that of MGLY, depending on the aerosol ionic content (Kampf et al.,
2013; Waxman et al., 2015). As α-dicarbonyl species, GLYX and MGLY
exhibit similar aqueous-phase chemistry: they undergo reversible hydration
and self-oligomerization (Ervens and Volkamer, 2010; Hastings et al., 2005;
Sareen et al., 2010; Shapiro et al., 2009), they can be oxidized by
aqueous-phase radicals to form organic acids or organosulfates (Carlton et
al., 2007; Lim et al., 2013; Perri et al., 2010; Schaefer et al., 2012,
2015), and they can react with nitrogen-containing species to form brown
carbon (De Haan et al., 2018; Lee et al., 2013; Maxut et al., 2015;
Nozière et al., 2009; Powelson et al., 2014; Sareen et al., 2010; Schwier
et al., 2010; Shapiro et al., 2009; Yu et al., 2011).
GLYX and MGLY received significant attention in the atmospheric chemistry
modeling community (Carlton et al., 2008; Fu et al., 2008, 2009) following
early experimental demonstrations of their potential significance as aqSOA
precursors (Carlton et al., 2007; Hastings et al., 2005; Kroll et al., 2005;
Loeffler et al., 2006). Fu and co-workers predicted that uptake of GLYX and
MGLY to low-level clouds was a significant source of organic aerosol over
North America, with MGLY producing more than 3 times more SOA than GLYX
(Fu et al., 2009). Carlton et al. (2008) found that including in-cloud aqSOA
production by GLYX in CMAQ improved agreement with aircraft observations.
Since these initial studies, more information has become available regarding
the gas–particle partitioning of glyoxal and methylglyoxal (Ip et al., 2009;
Kampf et al., 2013; Waxman et al., 2015; Yu et al., 2011) and their chemical
processing in the aqueous phase, allowing a refinement of their
representation in models.
Here, we calculate reactive uptake coefficients for glyoxal and methylglyoxal
for several cloud and aerosol types for application in large-scale
atmospheric chemistry modeling. We take into account salting effects,
aerosol thermodynamics, mass transfer considerations, and aqueous-phase
chemical kinetics. We base our calculations on the cloud and aerosol types
used in GEOS-Chem v11, so these recommendations can be applied directly to
that model, but the approach is general.
Methods and data
Following Hanson et al. (1994), the reactive uptake coefficient, γ, is
calculated according to
1γ=1α+ω4H*RTkIDaq1cothq-1/q,
where α is the mass accommodation coefficient, ω is the
gas-phase thermal velocity of the organic species, H* is the effective
Henry's law constant (Schwartz, 1986), R is the universal gas constant, T
is temperature in Kelvin, kI is the first order aqueous loss rate, and
Daq is the aqueous-phase diffusion coefficient for the organic species.
The particle radius, Rp, and in-particle diffusion limitations are
taken into account through the parameter q=Rp/l, where l is
the diffuso-reactive length:
l=DaqkI1/2.
The aqueous-phase diffusion coefficient used for both GLYX and MGLY was
Daq=10-9 m2 s-1. Daq does not vary
much for small species, and this value is typical for small organics (Bird et
al., 2006). The mass accommodation coefficient used was α=0.02. This
value of α is an estimate based on the assumption that α values
for GLYX and MGLY are similar to that of formaldehyde uptake to water (Jayne
et al., 1992). Following Eq. (1), the calculation is insensitive to within
10 % for a 50 % variation in α for values of
γ < 10-3.
Particle types and composition
Although the approach described here is general, we applied it to the liquid
cloud and aerosol particle types in GEOS-Chem v11. A complete listing can be
found in the Supplement. Briefly, we considered marine and remote continental
liquid cloud droplets, coarse and fine sea salt aerosol particles as a
function of relative humidity (RH), and sulfate/nitrate/ammonium (SNA)
aerosols as a function of RH and composition. Calculated results for
γGLY and γMGLY as a function of S : A,
S : N and calculated pH are available in the Supplement. Sea salt
aerosols are assumed to be composed of 100 % NaCl. The Windows
stand-alone executable for ISORROPIA-II (Fountoukis and Nenes, 2007) was used
in forward mode to calculate the equilibrium inorganic ion composition of the
aerosols, in order to calculate the Henry's constant. The temperature was
held constant at 280 K, and calculations were performed for each of the
desired relative humidities (99, 95, 90, 80, 70, and 50 %). Solid
formation was suppressed (metastable mode). For the SNA aerosols, the amount
of NO3- was held constant at 2 µmol m-3 air, while
the SO4-2 and NH3 amounts were each varied from 1 to
8 µmol m-3.
Mean values and range of in-particle hydroxyl radical
concentrations, as reported by Herrmann et al. (2010).
Cloud/aerosol type
Mean [OH] (M)
Max [OH] (M)
Min [OH] (M)
Maritime aerosols
10-13
3.3 × 10-12
4.6 × 10-15
Remote aerosols
3.0 × 10-12
8 × 10-12
5.5 × 10-14
Maritime clouds
2.0 × 10-12
5.3 × 10-12
3.8 × 10-14
Remote clouds
2.2 × 10-14
6.9 × 10-14
4.8 × 10-15
Reaction and mass transfer parameters.
Species
kOH (M-1 s-1)
KH,w (M atm-1)
Ks,NaCl (m-1)
Ks,(NH4)2SO4 (m-1)
Ks,NH4NO3 (m-1)
Glyoxal
1.1 × 109a
3.5 × 105c
-0.10e
-0.24d
-0.07e
Methylglyoxal
7 × 108b
3.71 × 103c
0.06e
0.16e
0.075e
a Schaefer et al. (2015). b Schaefer et
al. (2012). c Betterton and Hoffmann (1988). d Kampf
et al. (2013). e Waxman et al. (2015).
Aqueous-phase reaction
The formulation in Eqs. (1) and (2) describes uptake due to irreversible
aqueous-phase loss processes only. Based on our previous analysis of the
system using the multiphase photochemical box model GAMMA, the dominant
irreversible atmospheric aqueous-phase reactive process for GLYX and MGLY is
the reaction with OH (McNeill, 2015; McNeill et al., 2012). This reaction is the
initiation step for most radical-based chemistry of GLYX and MGLY in the
atmospheric aqueous phase, including organic acid formation and organosulfate
formation (McNeill et al., 2012; Perri et al., 2010). Other irreversible loss
processes, such as imidazole formation, occur on much longer timescales
(Teich et al., 2016; Yu et al., 2011). Therefore, the aqueous loss is
represented by the pseudo-first-order rate constant for the reaction between
the organic species of choice and OH (i.e., kI=kOH[OH]).
Reversible reactive processes, e.g., spontaneous hydration and
self-oligomerization of glyoxal and methylglyoxal, which substantially
promote uptake of GLYX and MGLY to the aqueous phase, may be taken into
account by the use of an effective Henry's law constant (McNeill et al.,
2012). However, we note that the form of Eq. (1) implies no uptake
(reversible or irreversible) in the absence of OH.
Considerable uncertainty exists in the aqueous concentration of OH in
cloud water and especially aerosol particles. In order to calculate kI,
we use OH concentrations for maritime and remote continental clouds and
aerosols following Herrmann et al. (2010) (Table 1). They reported a range
of [OH] for each scenario calculated using the CAPRAM 3.0 model. This [OH]
range was used in the calculation of the uncertainty in γ.
Calculating the Henry's constant
The solubility of glyoxal and methylglyoxal in aqueous solutions depends on
the salt content (Ip et al., 2009; Kampf et al., 2013; Waxman et al., 2015;
Yu et al., 2011). Glyoxal becomes more soluble with increasing salt
concentration (i.e., it exhibits “salting in”), whereas the opposite is
true for methylglyoxal (it “salts out”). Therefore, the Henry's constants
for glyoxal and methylglyoxal are a function of particle type and liquid
water content (and therefore RH).
The Henry's constants are calculated for sea salt aerosols following Waxman
et al. (2015) using the following equation:
logKH,wKH,NaCl=Ks,NaClcNaCl,
where KH,w is the Henry's constant for pure water,
KH,NaCl is the Henry's constant for the salt-containing aerosol,
cNaCl is the NaCl concentration in molality as calculated using
ISORROPIA-II, and Ks,NaCl is the salting constant (Table 2).
Note that the KH values are effective Henry's constants, which
account for hydration of the carbonyl species upon uptake. Waxman and
co-workers showed that salting constants were additive for a mixed
(NH4)2SO4 / NH4NO3 system, following
logKH,wKH,salt=Ks,NH42SO4cNH42SO4+Ks,NH4NO3cNH4NO3,
where KH,salt is the Henry's constant for the salt mixture,
cNH42SO4 and
cNH4NO3 are the concentrations in molality, and
Ks,NH42SO4 and
Ks,NH4NO3 are the salting constants. The sum of
sulfate and bisulfate was used to calculate cNH42SO4.
For cloud droplets, KH,w is used due to the low ion
concentrations in cloud water (Ervens, 2015; McNeill, 2015).
Statistical analysis and parameter estimation
The calculated reactive uptake coefficient, γ, for MGLY and GLYX and
each particle type was parameterized as a function of RH via
weighted least squares regression. Assuming that the errors in the reactive
uptake coefficients are log-normally distributed, a covariance matrix for the
model parameters was calculated based on the mean square errors of the data
and the design matrix of the linear regression. The standard deviations of
the model parameters were then determined from the diagonal of the covariance
matrix (Aster et al., 2005). Student's t tests were then performed on each
model parameter for the hypothesis that the model parameter in question is
equivalent to zero in order to assess the necessity of each parameter. The
nonzero model parameter was kept for t tests in which there was at least
98 % confidence that the hypothesis of the model parameter being zero
could be rejected.
Results and discussion
The reactive uptake coefficient, γ, was calculated for MGLY and GLYX
as a function of [OH], RH, particle size, and in the case of SNA aerosol,
particle composition. Calculated values of γ varied over several
orders of magnitude. In most cases these values are lower than those
previously used to model reactive uptake of these species in large-scale
models.
Liquid cloud droplets
The results for marine and remote continental cloud droplets are shown in
Table 3, with the mean value and error bars given. The uncertainty reflects
the uncertainty in [OH]. γMGLY is lower than γGLYX
by a factor of roughly 100 in each case, consistent with its lower KH,w
and kOH.
Aerosols
For aerosols, the reactive uptake coefficients were found to vary
significantly with RH due to salting effects. Figure 1 shows calculated
values of γGLYX for the three particle types as a function of RH.
The range of uncertainty in the calculated values, indicated by the red
shading, is due to the uncertainty in [OH] (Table 1), and in the case of SNA
aerosols, variations due to the different aerosol compositions tested. The
black lines indicate the weighted least squares fit to the data, and the grey
lines indicate the confidence interval for the fit. The average values and
the results of the least squares fits are summarized in Table 4.
Calculated reactive uptake coefficients for uptake of glyoxal to
sulfate/nitrate/ammonium (SNA) aerosols, and fine and coarse sea salt (SS)
aerosols, as defined in GEOS-Chem v11. Red shading indicates the uncertainty
in γGLYX. The black lines show the results of weighted least
squares regression, with the confidence intervals in grey.
Recommended γ for liquid cloud droplets. Cloud types and
size as defined in GEOS-Chem v11.
Cloud type
Reff (µm)
γGLYX
γMGLY
Marine
10
7.5 × 10-4 (+0.001, -7.4 × 10-4)
5.7 × 10-6 (+9.4 × 10-6, -5.6 × 10-6)
Remote continental
6
4.3 × 10-6 (+9.2 × 10-6, -3.4 × 10-6)
3.2 × 10-8 (+6.7 × 10-8, -2.5 × 10-8)
Summary of γGLYX
recommendations for aerosols. Aerosol types and specifications as defined in
GEOS-Chem v11.
Aerosol type
γGLYX average value
γGLYX parameterization
x: RH as fraction
SNA
1.0(±0.1) × 10-2 (RH = 50 %)
γ=exp(a+bx+cx2)
4.9(±1.0) × 10-4 (RH = 70 %)
a=12.1 (±0.6)
1.0(±0.3) × 10-4 (RH = 80 %)
b=-44.5 (±1.7)
2.6(±1.3) × 10-5 (RH = 90 %)
c=22.3 (±1.1)
2.8(±1.4) × 10-5 (RH = 95 %)
conf = 0.9997
8.5(±4.6) × 10-5 (RH = 99 %)
Sea salt (fine)
2.6 × 10-6 (+0.04, -2.6 × 10-6)
γ=exp(a+bx+cx2)
a=-7.5 (±0.1)
b=-10.0 (±0.3)
c=4.4 (±0.2)
conf = 0.9998
Sea salt (coarse mode)
4.8 × 10-7 (+0.013, -4.8 × 10-7)
Average value recommended
Summary of γMGLY recommendations for aerosols.
Aerosol type
γMGLY average value
γMGLY parameterization x: RH as fraction
SNA
1.6(±0.4) × 10-10 (RH = 50 %)
γ=exp(a+bx+cx2)
4.3(±1.0) × 10-9 (RH = 70 %)
a=-25.7 (±0.4)
1.3(±0.3) × 10-8 (RH = 80 %)
b=2.5 (±1.0)
3.5(±1.5) × 10-8 (RH = 90 %)
c=8.3 (±0.7)
7.7(±4.1) × 10-8 (RH = 95 %)
conf = 0.9990
5.3(±2.9) × 10-7 (RH = 99 %)
Sea salt (fine)
6.5 (±1.3) × 10-9
γ=exp(a+bx+cx2)
a=-17.9 (±0.9)
b=-10.4 (±2.6)
c=10.3 (±1.7)
conf = 0.9909
Sea salt (coarse mode)
5.5 × 10-10 (+0.016, -5.5 × 10-10)
Average value recommended
γGLYX decreases with increasing RH, due to salting in. Therefore,
the maximum γGLYX (10-2 for SNA, 50 % RH) exceeds the
dilute (cloud water) case. It also exceeds the general case used by Fu et
al. (2008) (γGLYX,Fu = 2.9 × 10-3), which was
based on the experimental observations of Liggio et
al. (2005) for (NH4)2SO4 aerosols at 55 % RH. The
parameterization presented in Table 4 yields
γGLY = 3.6 × 10-3 at 55 % RH.
In the case of coarse sea salt, in-particle diffusion limitations led to
smaller γGLYX at the mean [OH] than at the minimum [OH] for some RH
values. The scatter in the calculated γGLYX led to a low-confidence
result from the weighted least squares regression. For this reason, we
recommend use of the error-weighted average γGLYX value in lieu of
a parameterization (Table 4).
Calculated reactive uptake coefficients for uptake of methylglyoxal
to sulfate/nitrate/ammonium (SNA) and sea salt (SS) aerosols. See text for
details.
Figure 2 shows calculated values of γMGLY as a function of
RH. The average values and the results of the least squares fits are
summarized in Table 5. In contrast to glyoxal, methylglyoxal salts out, so
γMGLY increases with increasing RH. All calculated values
(10-10 < γMGLY < 10-6) are
much smaller than the general case used by Fu et al. (2008)
(γMGLY,Fu = 2.9 × 10-3). Those
investigators had assumed that the reactive uptake coefficient for
methylglyoxal would be equal to that for glyoxal as measured by Liggio et
al. (2005).
Similar to the glyoxal case, the variability in γMGLY for
the coarse-mode sea salt aerosols due to in-particle diffusion limitations
led to a low-confidence weighted least squares fit. The error-weighted
average value is recommended.
For a given RH, γGLYX and γMGLY
show weak, nonmonotonic dependence on S : A and S : N, with
significant scatter (see the Supplement). Plotting γGLYX
and γMGLY as a function of calculated aerosol pH shows a
general positive trend, with γ increasing with increasing pH at a
given RH. These plots can be found in the Supplement along with a
parameterization of the trend. We note that aerosol pH is not an independent
parameter in our calculations, but rather a calculated output of ISORROPIA II
as a function of input aerosol composition and RH. Furthermore, the mass
transfer and reactive loss processes considered here are not explicitly pH-dependent. Therefore, we interpret this apparent dependence on pH to be
reflective of pH being a proxy for variations in solute concentration in the
aerosol with varying aerosol liquid water content. All variations in
γGLYX and γMGLY as a result of varying SNA aerosol
composition or pH are incorporated in the error bars shown in the top panels
of Figs. 1 and 2, and therefore accounted for in the weighted least squares
fits and the uncertainty ranges provided in the parameterizations in Tables 4 and 5.
Atmospheric implications
We present revised recommendations for the reactive uptake coefficients for
glyoxal and methylglyoxal for several cloud and aerosol types. The values we
calculated under many conditions are lower than those currently used in
large-scale models such as GEOS-Chem, although we note that the
parameterization presented in Table 4 under the experimental conditions of
Liggio et al. (2005), 55 % RH, yields γGLY = 3.6 × 10-3,
which is within 24 % of their experimental value
(2.9 × 10-3). We expect application of these parameterizations
will result in a decrease in the calculated contribution of MGLY uptake to
aqueous SOA formation and better representation of spatial variability in
aqSOA formed from glyoxal. The reduced contribution of MGLY to aqueous SOA
formation due to salting out is consistent with the calculations of Sareen et
al. (2017).
Reactive uptake of glyoxal and methylglyoxal to other hygroscopic aerosols
such as organic aerosols is possible, although given the importance of salting
effects on this chemistry, and the low expected [OH] concentration in organic
aerosols (McNeill, 2015), we expect the contribution of these processes to
aqSOA formation to be minor.
This representation of aqueous SOA formation by GLYX and MGLY, with the
treatment of Henry's constants described here, does not take into account the
contribution of reversible uptake of GLYX, which could be a significant,
although transient, source of aerosol mass under some conditions (McNeill et
al., 2012; Woo and McNeill, 2015). The use of this parameterization together
with simpleGAMMA (Woo and McNeill, 2015) would give representation of both
aqSOA formation types by GLYX.