Supporting information for: CCN Spectra, Hygroscopicity...
Transcript of Supporting information for: CCN Spectra, Hygroscopicity...
Supporting information for: CCN Spectra,
Hygroscopicity, and Droplet Activation Kinetics of
Secondary Organic Aerosol Resulting from the 2010
Deepwater Horizon Oil Spill
Richard H. Moore,† Tomi Raatikainen,‡ Justin M. Langridge,¶,§ Roya Bahreini,¶,§
Charles A. Brock,¶ John S. Holloway,¶,§ Daniel A. Lack,¶,§ Ann M. Middlebrook,¶
Anne E. Perring,¶,§ Joshua P. Schwarz,¶,§ J. Ryan Spackman,¶,‖ and Athanasios
Nenes∗,†,‡
School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta,
Georgia, USA., School of Earth & Atmospheric Sciences, Georgia Institute of Technology,
Atlanta, Georgia, USA., Earth System Research Laboratory, National Oceanic and Atmospheric
Administration, Boulder, Colorado, USA., Cooperative Institute for Research in Environmental
Sciences, University of Colorado, Boulder, Colorado, USA., and Science and Technology
Corporation, Boulder, Colorado, USA.
E-mail: [email protected]
1
∗To whom correspondence should be addressed†School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA.‡School of Earth & Atmospheric Sciences, Georgia Institute of Technology, Atlanta, Georgia, USA.¶Earth System Research Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA.§Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA.‖Science and Technology Corporation, Boulder, Colorado, USA.
S1
Flight Overview Maps2
S2
30 30
29 29
28 28
-90 -88 -86 -84 -82
DWH Site
Spill Extent
10 June 20100.6
0.5
0.4
0.3
0.2
0.1
0.0
CC
N-D
eriv
ed
Hygro
scopic
ity (
κ)
30 30
29 29
28 28
-90 -88 -86 -84 -82
8 June 2010
Figure 1: Complete aircraft trajectories for the survey flights on 8 June (top) and 10 June (bottom).Both flights originate and end in Tampa, Florida. Markers are colored by the CCN-derived hygro-scopicity. The gray shaded area represents the extent of surface oil (1). The ordinate and abscissadenote degrees latitude and longitude, respectively.
S3
CCN-Derived Hygroscopicity3
The CCN-derived hygroscopicity parameter, κCCN is obtained from the supersaturation-dependent4
CCN concentration and aerosol size distribution following Moore et al. (2):5
1. The normalized cumulative size distribution was calculated for each particle size bin as6
ncum(Dp) =1
NCN
∫ Dp
4nm
dNCN
d logDpd logDp (1)
where Dp is the particle diameter, NCN is the fine particle concentration larger than 4 nm,7
dNCN is the fine particle concentration in the bin, and d logDp is the bin width in logarithm8
space.9
2. The CCN-active aerosol fraction, Ra = NCCN/NCN was calculated, where NCCN is the mea-10
sured CCN concentration at a given supersaturation. For an internally-mixed aerosol whose11
composition does not vary with size, all particles larger than some critical supersaturation,12
Dp,c act as CCN. Thus, Dp,c can be found directly by interpolating the normalized cumula-13
tive size distribution as Ra = 1−ncum(Dp,c).14
3. The interpolated value of Dp,c is used in Köhler theory (Eq. 1 in the main text) to find κCCN .15
Since κ is strongly dependent on the critical diameter through its cubic dependence in Köhler16
theory, determining Dp,c is likely to be the largest source of uncertainty in deriving κCCN . Conse-17
quently, we also calculate the sensitivity of the derived κCCN to uncertainties associated with NCCN18
and NCN . The relative CN concentration uncertainty, εCN is reported as 11%, while εCCN depends19
on the concentration-dependent Poisson statistical uncertainty as well as that due to the CCNC20
flow rate uncertainty (εQCCN ∼ 7%). Thus, the overall uncertainty of Ra is21
εRa =
√ε2
CN +
(τCCN
NCCNQCCN
)+ ε2
QCCN(2)
S4
where τCCN is the OPC integration time (1 second), QCCN is the CCNC sample flow rate (45.4522
cm3 min−1). In addition to accounting for the uncertainty of the activated ratio, Ra, an additional23
10% uncertainty was applied to Dcrit after interpolation in Step 2 above to account for the UHSAS24
sizing uncertainty (3). The derived uncertainty in κCCN was found to vary significantly between25
50-200% during the flight on 8 June (see Figure 3 and Figure 4 in the main text).26
CRD-Derived Hygroscopicity27
The data from the cavity ringdown (CRD) spectrometer were used to infer the hygroscopicity28
parameter, κ , as follows:29
1. The extinction-derived humidification factor, γext , was used to calculate the change in ex-30
tinction between a dry particle at 10%RH and a humidified particle at 85%RH, f (85%RH),31
as32
f (85%RH) =σ85%RH
σ10%RH=
[1−0.851−0.10
]γext
(3)
where σRH is the is the aerosol extinction at relative humidity, RH.33
2. Mie theory equations (4) were iteratively solved to find the hygroscopic growth factor,34
g(85%RH)=Dp(85%RH)Dp(10%RH) , that reproduces the measured humidified light extinction at 85%RH,35
where Dp(RH) is the RH-dependent particle diameter. The Mie theory calculations were per-36
formed at the CRD laser wavelength of 532 nm and used the fine particle size distribution37
(Dp < 2 µm) with a prescribed complex refractive index (RI) of 1.45 - 0i that is characteristic38
of alkane and aromatic species (5). At each iteration step, the humidified particle refractive39
index is calculated by volume-weighting the dry particle refractive index with that of water40
(1.33 - 0i).41
3. The hygroscopicity parameter, κ , is then calculated for each aerosol size distribution bin as42
in Petters and Kreidenweis (6)43
S5
κ =(g3−1
)(exp( ADpg)
RH100
−1
)(4)
where A = (4Mwσw)/(RT ρw), R is the ideal gas constant, T is the absolute temperature of44
the measurement, and σw, Mw and ρw are the surface tension, molar mass, and density of45
water, respectively. The overall κ is calculated as a volume-weighted average of the κ for46
all size distribution bins.47
Table 1: Pure component densities and refractive indices used to compute κCRD.
Density RefractiveComponent (kg m−3) IndexAmmonium Sulfate 1769a 1.53−0.00ia,b
Ammonium Nitrate 1725a 1.61−0.00ia
Organic Carbon 1400c 1.45−0.00id
Elemental Carbon 1800e 1.95−0.79ie
Water 996a 1.33−0.00ia
aGreen and Perry, (7)bToon et al., (8)cAlfarra et al., (9)dRiazi and Al-Sahhaf, (5)eBond and Bergstrom, (10)
Implicit in the above method is the assumption that the aerosol are internally-mixed and can be48
described well by a single, complex refractive index. To test the sensitivity of the derived κCRD to49
the RI assumption, the measured aerosol composition from a compact time of flight aerosol mass50
spectrometer (C-ToF-AMS) and a single particle soot photometer (SP2) was used to compute an51
RI based on a volume-weighted average of the pure component refractive indices (Table 1). As52
shown in Figure 2a, the derived κCRD is insensitive to the assumed RI over the range of observed53
values (Figure 2b,c) with the uncertainty in κCRD from this assumption estimated to be less than54
0.02, absolute. While elemental carbon has the potential to significantly impact aerosol extinction55
because of its large RI, the low mass loadings observed during the the 8 June flight do not appear56
to have a large influence (Figure 2b,c). The observed insensitivity of κCRD to the refractive in-57
dex assumption may not be true for some heterogeneous environments with large local sources of58
S6
elemental carbon, however, the finding is consistent with at least one previous study (11). Uncer-59
tainties arising from the measurement of f (RH) were found to vary between 5-8%, on average,60
which were propagated in the calculation procedure above to find the κCRD uncertainty over the61
range of observed values (shown in Figure 3 and Figure 4 in the main text). This uncertainty62
analysis neglects uncertainties associated with the aerosol size distribution measurement, such as63
non-idealities arising from particle non-sphericity, inhomogeneity, or core-coating structure. Thus,64
the derived κCRD uncertainty should be considered a lower limit.65
Figure 3 shows a direct comparison between κCCN and κCRD for the June 8 flight, with the66
corresponding timeseries shown in Figure 4 of the main article. Overall, the CRD-derived hygro-67
scopicity is approximately two-fold lower than the CCN-derived hygrosopicity, with κCCN ranging68
from 0.05-0.6, while κCRD varies from 0.01-0.2. A linear fit between the two quantitities gives69
a slope of 0.47 and a coefficient of determination, R2 = 0.75. An interesting feature of Figure 370
is that a clear minimum value of κCCN is apparent between 0.05-0.10, despite wider variation in71
κCRD. Similarly, κCRD reaches a maximum value of 0.15-0.20, despite much wider variation in72
κCCN . These asymptotic limits coincide with high and low organic volume fractions, and seem to73
imply that organic species reduce subsaturated hygroscopic growth more so than CCN activation.74
Size-Dependent CCN Composition75
Figure 4 shows the size-resolved organic aerosol volume fraction (εSR,org) plotted versus the bulk76
(i.e., size-averaged) organic volume fraction (εorg). The Aitken mode aerosol contains mostly77
organics (εSR,org∼0.85-1), but its influence is eclipsed by the more-massive accumulation mode78
aerosol resulting in lower values of bulk εorg. A 100-nm cutsize was used to differentiate the79
two modes, and the average composition in each mode was used to determine the modal κ =80
εSR,orgκorg + εSR,inorgκinorg, assuming κorg=0.05 and κinorg=0.6. As shown the lower part of Fig-81
ure 4, the accumulation mode κ is roughly two-fold greater than the Aitken mode κ . The modal κ82
were linearly regressed against the bulk εorg, and were combined in an aerosol number-weighted83
S7
average to find the overall aerosol hygroscopicity, shown as filled circles in Figure 4. Using a84
number-weighted average versus a volume-weighted average is analogous to assuming the two85
modes are externally mixed rather than internally mixed.86
S8
0.01
2
3
4
5
6
789
0.1
2
0.012 3 4 5 6 7 8 9
0.12
500400300200100
0Fre
qu
en
cy
-0.05 0.00 0.05
1.0
0.9
0.8
0.7
0.6
0.5
0.4Org
anic
Volu
me
Fra
ction
CRD-Derived Hygroscopicity κ
(Composition-Dependent Refractive Index)
CR
D-D
erived H
ygro
scopic
ity κ
(Consta
nt R
efr
active Index)
300
200
100
0
Fre
quency
1.561.541.521.501.481.461.441.42
Real Refractive Index
500
400
300
200
100
0
0.100.080.060.040.020.00
Imaginary Refractive Index
Fre
quency
B
C
A
Absolute Error
Figure 2: (a) Comparison plot of the CRD-derived hygroscopicity for the 8 June flight calcu-lated assuming constant and composition-dependent refractive indices. Data from the high altitudetransit legs are excluded. (b,c) Frequency of occurrence of the real and imaginary parts of thecomposition-dependent refractive index during the 8 June flight.
S9
0.01
2
3
4
5
6
78
0.1
2
3
4
5
6
0.012 3 4 5 6 7 8 9
0.12 3 4 5 6
CCN-Derived Hygroscopicity (κ )CCN
1:1
1:2
1:4
1.00.90.80.70.60.50.4
Organic Volume Fraction
250
200
150
100
50
0
Fre
qu
en
cy
-1.0 -0.5 0.0 0.5
C
RD
-De
rive
d H
yg
rosco
pic
ity (κ )
CR
D
Relative Error
Figure 3: Comparison plot of the CRD-derived and CCN-derived hygroscopicity parameters ob-served during the 8 June flight.
S10
10
2
4
6
100
2
4
6
1000
0.900.800.700.600.500.400.30
6
8
2
4
Number Geometric Mean Diameter
Total Mass Geometric Mean Diameter1.0
0.1
Hygro
scopic
ity
Para
mete
r (κ
)
Part
icle
Dry
Dia
mete
r (n
m)
0.8
0.6
0.4
0.2
0.0
Siz
e-R
eso
lve
d O
rga
nic
Vo
lum
e F
ractio
n (
ε )S
R,o
rg
Bulk Organic Volume Fraction (ε )org
AccumulationMode κ
Aitken Mode κ
2-Mode κ
Linear Fit
κ = 0.45 - 0.35 εorg
κ = 0.21 - 0.15 εorg
Figure 4: Average size-resolved organic volume fraction (εSR,org) from the C-ToF-AMS (top)and calculated modal hygroscopicity parameters (κ) (bottom) plotted versus the bulk (i.e., size-averaged) C-ToF-AMS organic volume fraction εorg for both flights. The 2-Mode κ is an aerosolnumber-concentration-weighted average of the linear fits to the Aitken and accumulation mode κ .
S11
Coupled CCNC and Droplet Growth Model87
A detailed description of the coupled CCNC instrument and droplet growth model is given by Raatikainen88
et al. (12), so only a brief description is included here. A simplified form of the coupled model89
was employed here, which assumes parabolic velocity fields while solving the water vapor and en-90
ergy conservation equations. Neglecting explicit calculation of the velocity fields has a negligible91
effect on the simulation results, but substantially decreases the necessary computational time (12).92
The model has been successfully applied in the past to simulate the instrument behavior for both93
steady-state and transient operation (13–15). Recent work by Lathem and Nenes (16) has shown94
that moderately high CCN concentrations in the growth chamber can slightly decrease the cen-95
terline supersaturation profile. This effect appears to be unimportant for measurements of CCN96
concentrations below 104 cm−3 (16), which is the typical mode of operation. However, supersatu-97
ration depletion can have a detectable effect on the OPC-measured droplet size distribution when98
the CCN concentrations exceed several hundred per cm3. The model can be used to simulate the99
CCN droplet size distributions, both with and without such supersaturation depletion effects as a100
function of the effective water uptake coefficient, γcond . The optimal value of γcond is then found101
by iteratively matching the simulated and measured droplet size distributions.102
As described in detail by Raatikainen et al. (12), a number of uncertainties exist that challenge103
the simulations. For example, aerosol mixing state and size-dependent composition give rise to104
a polydisperse hygroscopicity distribution, which is poorly-constrained. Additionally, there are105
various instrument non-idealities (e.g., column thermal resistance, possible OPC sizing biases,106
incomplete wetting of the column wall and overall mass transfer coefficient changes, non-linearity107
of the wall temperature profile). To quantify the impact of these effects, CCN droplet distributions108
for ammonium sulfate calibration aerosol were compared to those predicted by the model, and it109
was found that the model simulations overpredict the measured mean droplet size by approximately110
2.1 µm, which agrees well with the correction bias of 2.3 µm uncovered for γcond=0.2 (Table 2)111
and applied to the June 8th simulation timeseries in Figure 4 of the main article.112
Figure 5 shows 1:1 comparison plots for the uncorrected simulated mean droplet size (D̄p,sim)113
S12
versus the OPC measured mean droplet size (D̄p), and the linear regression coefficients are listed114
in Table 2. It can be seen that forcing the regression slope to unity has a negligible effect on the115
coefficient of determination for all simulations except those where supersaturation depletion effects116
are neglected. Simulations assuming γcond=0.05-1 show the best correlation with observations117
(R2∼0.51), and the overprediction bias for the γcond=0.1-0.2 simulations is most similar to the118
2.1 µm bias identified from calibrations. The correlation worsens significantly for γcond <0.05 or119
when depletion effects are neglected. Thus, the regression analysis indicates that CCN near the120
Deepwater Horizon oil spill do not exhibit slow activation kinetics and derived values of γcond are121
consistent with those for ammonium sulfate calibration aerosol and previously reported values for122
pure water droplets (17, 18).123
Table 2: Linear regression coefficients between modeled and observed mean droplet sizes formodel simulations with varying water uptake coefficients and with supersaturation depletion ef-fects considered or turned off. Both 1-parameter (D̄p,sim = D̄p +Bias) and 2-parameter (D̄p,sim =Slope× D̄p +Bias) are listed.
Model Simulation 1-Parameter Fit 2-Parameter Fitγcond Depletion Effects? Bias R2 Slope Bias R2
1.00 Yes 2.46 0.49 0.76 3.20 0.530.20 Yes 2.29 0.50 0.76 3.03 0.530.10 Yes 2.09 0.49 0.76 2.83 0.540.05 Yes 1.70 0.49 0.76 2.45 0.530.04 Yes 1.53 0.47 0.76 2.28 0.520.03 Yes 1.25 0.44 0.75 2.03 0.500.02 Yes 0.78 0.39 0.73 1.63 0.460.01 Yes -0.21 0.26 0.65 0.88 0.370.008 Yes -0.53 0.22 0.62 0.67 0.350.005 Yes -1.11 0.04 0.52 0.41 0.291.00 No 2.69 0.15 0.56 4.06 0.360.20 No 2.50 0.18 0.58 3.82 0.360.10 No 2.27 0.20 0.60 3.53 0.360.05 No 1.85 0.24 0.63 3.02 0.360.04 No 1.66 0.25 0.64 2.79 0.360.03 No 1.36 0.25 0.65 2.46 0.350.02 No 0.85 0.26 0.66 1.91 0.350.01 No -0.19 0.22 0.63 0.97 0.340.008 No -0.51 0.19 0.61 0.72 0.330.005 No -1.10 0.04 0.52 0.42 0.29
S13
6.2
6.0
5.8
5.6
5.4
5.2
5.0
3.83.63.43.23.02.82.6
6.4
6.2
6.0
5.8
5.6
5.4
5.2
3.83.63.43.23.02.82.6
With Depletion Effects Without Depletion Effects
6.0
5.8
5.6
5.4
5.2
5.0
4.8
3.83.63.43.23.02.82.6
6.2
6.0
5.8
5.6
5.4
5.2
5.0
3.83.63.43.23.02.82.6
5.8
5.6
5.4
5.2
5.0
4.8
4.6
3.83.63.43.23.02.82.6
6.0
5.8
5.6
5.4
5.2
5.0
4.8
3.83.63.43.23.02.82.6
5.6
5.4
5.2
5.0
4.8
4.6
4.4
3.83.63.43.23.02.82.6
5.6
5.4
5.2
5.0
4.8
4.6
4.4
3.83.63.43.23.02.82.6
3.6
3.4
3.2
3.0
2.8
2.6
2.4
3.83.63.43.23.02.82.6
3.8
3.6
3.4
3.2
3.0
2.8
2.6
2.4
3.83.63.43.23.02.82.6
γ = 1.00cond
γ = 0.20cond
γ = 0.10cond
γ = 0.05cond
γ = 0.01cond
γ = 1.00cond
γ = 0.20cond
γ = 0.10cond
γ = 0.05cond
γ = 0.01cond
Sim
ula
ted M
ean D
rople
t S
ize (
µm
)
Measured Mean Droplet Size (µm)
Figure 5: Comparison plots of the simulated CCN droplet sizes obtained from the instrumentmodel versus the measured mean droplet sizes during the 8 June flight. Solid traces are a 1-parameter linear fit (constant bias), and dashed traces are a 2-parameter linear fit (slope and bias).Regression coefficients are listed in Table 2.
S14
References124
(1) NOAA-NESDIS, Satellite Derived Surface Oil Analysis Products – Deepwater Horizon125
(www.ssd.noaa.gov/PS/MPS/deepwater.html). 2010.126
(2) Moore, R. H.; Bahreini, R.; Brock, C. A.; Froyd, K. D.; Cozic, J.; Holloway, J. S.; Middle-127
brook, A. M.; Murphy, D. M.; Nenes, A. Hygroscopicity and Composition of Alaskan Arctic128
CCN During April 2008. Atmospheric Chemistry and Physics 2011, 11, 11807–11825.129
(3) Brock, C. A.; Cozic, J.; Bahreini, R.; Froyd, K. D.; Middlebrook, A. M.; McComiskey, A.;130
Brioude, J.; Cooper, O. R.; Stohl, A.; Aikin, K. C.; de Gouw, J. A.; Fahey, D. W.; Fer-131
rare, R. A.; Gao, R.-S.; Gore, W.; Holloway, J.; Hübler, G.; Jefferson, A.; Lack, D. A.;132
Lance, S.; Moore, R. H.; Murphy, D. M.; Nenes, A.; Novelli, P. C.; Nowak, J. B.; Ogren, J. A.;133
Peischl, J.; Pierce, R. B.; Pilewskie, P.; Quinn, P. K.; Ryerson, T. B.; Schmidt, K. S.;134
Schwarz, J. P.; Sodemann, H.; Spackman, J. R.; Stark, H.; Thomson, D. S.; Thornberry, T.;135
Veres, P.; Watts, L. A.; Warneke, C.; Wollny, A. G. Characteristics, sources, and transport136
of aerosols measured in Spring 2008 during the Aerosol, Radiation, and Cloud Processes137
Affecting Arctic Climate (ARCPAC) project. Atmospheric Chemistry and Physics 2011, 11,138
2423–2453.139
(4) Bohren, C. F.; Huffmann, D. R. Absorption and Scattering by Small Particles; Wiley, 1983.140
(5) Riazi, M. R.; Al-Sahhaf, T. A. Physical Properties of n-alkanes and n-alkylhydrocarbons:141
Application to petroleum mixtures. Industrial and Engineering Chemistry Research 1995,142
34, 4145–4148.143
(6) Petters, M. D.; Kreidenweis, S. M. A single parameter representation of hygroscopic growth144
and cloud condensation nucleus activity. Atmospheric Chemistry and Physics 2007, 7, 1961–145
1971.146
(7) Green, D. W.; Perry, R. H. Perry’s Chemical Engineers’ Handbook, 8th Edition; McGraw-147
Hill, 2008.148
S15
(8) Toon, O. B.; Pollack, J. B.; Khare, B. N. The optical constants of several atmospheric aerosol149
species: Ammonium sulfate, ammonium oxide, and sodium chloride. Journal of Geophysical150
Research 1976, 81, 5733–5748.151
(9) Alfarra, M. R.; Paulsen, D.; Gysel, M.; Garforth, A. A.; Dommen, J.; Prévôt, A. S. H.;152
Worsnop, D. R.; Baltensperger, U.; Coe, H. A mass spectrometric study of secondary or-153
ganic aerosols formed from the photooxidation of anthropgoenic and biogenic precursors in154
a reaction chamber. Atmospheric Chemistry and Physics 2006, 6, 5279–5293.155
(10) Bond, T. C.; Bergstrom, R. W. Light absorption by carbonaceous particles: An investigative156
review. Aerosol Science and Technology 2006, 40, 27–67.157
(11) Ervens, B.; Cubison, M.; Andrews, E.; Feingold, G.; Ogren, J. A.; Jimenez, J. L.; DeCarlo, P.;158
Nenes, A. Prediction of cloud condensation nucleus number concentration using measure-159
ments of aerosol size distributions and composition and light scattering enhancement due to160
humidity. Journal of Geophysical Research 2007, 112.161
(12) Raatikainen, T.; Moore, R. H.; Lathem, T. L.; Nenes, A. A coupled observation – modeling162
approach for studying activation kinetics from measurements of CCN activity. Atmospheric163
Chemistry and Physics Discussions 2012, 12, 1821–1865.164
(13) Moore, R. H.; Nenes, A. Scanning flow CCN analysis – A method for fast measurements of165
CCN spectra. Aerosol Science and Technology 2009, 43, 1192–1207.166
(14) Rose, D.; Gunthe, S. S.; Mikhailov, E.; Frank, G. P.; Dusek, U.; Andreae, M. O.; Pöschl, U.167
Calibration and measurement uncertainties of a continuous-flow cloud condensation nuclei168
counter (DMT-CCNC): CCN activation of ammonium sulfate and sodium chloride aerosol169
particles in theory and experiment. Atmospheric Chemistry and Physics 2008, 8, 1153–1179.170
(15) Lance, S.; Medina, J.; Smith, J. N.; Nenes, A. Mapping the operation of the DMT continuous-171
flow CCN counter. Aerosol Science and Technology 2006, 40, 242–254.172
S16
(16) Lathem, T.; Nenes, A. Water vapor depletion in the DMT continuous flow CCN chamber:173
Effects on supersaturation and droplet growth. Aerosol Science and Technology 2011, 45,174
604–615.175
(17) Li, Y. Q.; Davidovits, P.; Shi, Q.; Jayne, J. T.; Kolb, C. E.; Worsnop, D. R. Mass and thermal176
accommodation coefficients of H2O(g) on liquid water as a function of temperature. Journal177
of Physical Chemistry A 2001, 105, 10627–10634.178
(18) Shaw, R. A.; Lamb, D. Experimental determination of the thermal accomodation and con-179
densation coefficients of water. Journal of Chemical Physics 1999, 111, 10659–106663.180
S17