GEOPHYSICAL RESEARCH LETTERS, VOL. 29, NO. 0, XXXX, doi:10.1029/2002GL015078, 2002

Adsorption isotherms of acetone on ice between 193 and 213 K Florent Domine´ and Laurence Rey-Hanot CNRS, Laboratoire de Glaciologie et Ge´ophysique de l’Environnement, St Martin d’He`res cedex, France Received 8 March 2002; revised 4 April 2002; accepted 30 April 2002; published XX Month 2002.

[1] The adsorption isotherms of acetone on ice were measured at 193, 203 and 213 K using a volumetric method with mass spectrometric detection. Henry’s law applies for values of the acetone partial pressure, Pacetone, lower than 10
3 Pa. Where Henry’s law applies, the number of acetone molecules adsorbed per cm2 of ice, is: nads = 90.53  Pacetone  exp (6610.2/T), with Pacetone in Pa and T in K. The measured enthalpy of adsorption of acetone on ice is
Hads =
55±7 kJ/mol. Acetone values previously measured in Arctic snow are too high to be due to adsorbed acetone. Acetone was then probably dissolved in ice or in organic aerosols contained in snow. Adsorption of acetone in the snowpack or on ice crystals in cirrus clouds is insufficient to affect Pacetone above the snow or in the INDEX TERMS: 1863 Hydrology: Snow and ice clouds. (1827); 3947 Mineral Physics: Surfaces and interfaces; 0320 Atmospheric Composition and Structure: Cloud physics and chemistry. Citation: Domine´, F., and L. Rey-Hanot, Adsorption isotherms of acetone on ice between 193 and 213 K, Geophys. Res. Lett., 29(0), XXXX, doi:10.1029/2002GL015078, 2002.

1. Introduction [2] Recent observations have shown that acetone could be emitted by the snowpack [Couch et al., 2000] and physical processes such as desorption from snow crystal surfaces or release during snow metamorphism have been proposed to explain the observations. Subsequent measurements at Alert (82.5
N, 62.3
W) during the ALERT 2000 field campaign [Guimbaud et al., 2002] also showed that acetone was exchanged between the snowpack and the atmosphere, and indicated that photochemical processes could contribute to acetone emission by the snowpack in the springtime. Also during ALERT 2000, Houdier et al. [2002] measured snow-phase acetone concentrations in the range 1.5 – 3.1 ppbw and noted that night-time values appeared more elevated than daytime ones, which is compatible with temperature-driven adsorption-desorption. [3] Understanding and modeling exchanges of acetone between the snowpack and the lower troposphere requires the knowledge of numerous parameters, among which the adsorption isotherms of acetone on ice, i.e. the relationship between acetone partial pressure, Pacetone, and the number of molecules adsorbed, nads. These isotherms may also help improve our understanding of acetone chemistry in the upper troposphere, where ice is often present, as cirrus clouds cover 25 –30% of the earth’s surface [Winkler and Trepte, 1998]. Since acetone is a major HOx source in the upper troposphere [Arnold et al., 1997], and because it strongly impacts O3 and NOx chemistry [Mu¨ller and BrasCopyright 2002 by the American Geophysical Union. 0094-8276/02/2002GL015078$05.00

seur, 1999; Folkins et al., 1998], a detailed understanding of its chemistry is clearly of interest. Investigating its interactions with ice may reveal new processes that would need to be included in upper tropospheric chemistry models. We have therefore measured the adsorption isotherms of acetone on ice using a volumetric method, and use the data obtained to evaluate the impact of adsorption on Pacetone above snow surfaces and in cirrus clouds.

2. Experimental Methods [4] Figure 1 shows the experimental system used. A glass container called the expansion volume Ve is coated with ice about 1 mm thick. Thinner ice results in pieces of ice falling off during the experiments. The 1 mm ice thickness is obtained by successive deposition of submillimetric layers, formed by rotating the expansion volume filled with a few ml of water in a cold room at
15
C. The expansion volume is then immersed in a thermostated bath. To measure the adsorption isotherms, purified and degassed acetone is leaked into the glass introduction volume Vi and Pacetone is measured by a high precision capacitance manometer, 1 Torr full scale (120A type baratron, MKS instruments). Knowing the ambient temperature and the volume of Vi, the number of acetone molecules in the gas phase, n1, is calculated with the ideal gas equation. [5] The valve between Vi and Ve is then opened. Acetone mixes with the water vapor in equilibrium over the ice and adsorbs on the ice. Because the main gas is now H2O, the new and lower Pacetone value is measured with a quadrupole mass spectrometer with electron impact ionization, using the mass 43 peak. The new number of acetone molecules in the gas phase, n2, is again determined with the ideal gas equation, and the number of acetone molecules adsorbed is nads = n1
n2. Small thermal transpiration corrections were performed using the equations of Takaishi and Sensui [1963]. The parameters used in these equations were determined using a molecular diameter of 0.255 nm for water, obtained from the kinetic theory of gases that links viscosity to molecular diameter. More details on the volumetric method can be found in Chaix et al. [1996] and Legagneux et al. [2002]. Incremental additions of acetone yields the adsorption isotherm nads = f(Pacetone) at the temperature of the bath. Because the ice used is made by freezing water, its surface is assumed to be flat. The ice surface area used to analyze the data is the geometric surface area of Ve that is immersed in methanol: 250 cm2. [6] The response of the mass spectrometer was calibrated by injecting Pacetone values measured with the capacitance manometer. Values down to 4  10
4 Pa (3  10
6 Torr) were used for calibration. The Knudsen number of the effusive beam entering the mass spectrometer chamber through a 200 mm gold-plated orifice varied with PH2O,

X-1

X-2

DOMINE´ AND REY-HANOT: ADSORPTION OF ACETONE ON ICE

adsorption is at least partially reversible. At lower surface coverages, isotherms are linear (Figure 3), and thus follow a Henry’s law: nads = H(T)  Pacetone, where H(T) is the Henry’s law constant. H(T) values were determined from least square fits of the data of Figure 3: 6.88  1016, 1.22  1016 and 2.77  1015 molecule cm
2 Pa
1 at 193, 203 and 213 K, respectively. The correlation coefficients of the least square fits are all >0.99. [9] The enthalpy of adsorption of acetone on ice at low surface coverage was deduced from these data. Figure 4 shows an Arrhenius plot of LnH vs. 1000/T. We are aware that 3 data points is not a large number to calculate an enthalpy, however the correlation coefficient of the plot of Figure 4 is 0.9997. The slope yielded
Hads =
55.0 kJ/ mol, with an estimated uncertainty of 7 kJ/mol. Figure 4 yields equation (1): 6610:2 T

nads ¼ 90:53 Pacetone e

ð1Þ

with nads in molecule cm
2, Pacetone in Pa and T in K. Equation (1) is valid where Henry’s law applies. Considering the data of Figure 3, and given that atmospheric Pacetone values are almost always less than 2  10
4 Pa [Singh et al., 1994; Arnold et al., 1997; Guimbaud et al., 2002], Henry’s law probably always applies under atmospheric conditions.

4. Discussion Figure 1. Experimental system used to measure the adsorption isotherms of acetone on ice. i.e., with the ice temperature. This resulted in an increase in the signal with increasing PH2O. Calibrations were thus performed with each PH2O value used in the experiments. The detection limit of our system is 1.3  10
4 Pa for pure acetone, and 6  10
5 Pa for acetone diluted in PH2O = 1.1 Pa (saturating pressure of ice at 213 K). [7] The accuracy of the capacitance manometer is critical for the accuracy of the data. Its nominal accuracy is 0.8% of reading, but this value can be degraded by ambient temperature variations, that affect gain and offset. The manometer, thermostated at 45
C, was placed in an insulated box to further reduce the effects of temperature variations. The measurements accuracy depended on temperature and Pacetone values. We estimate that most Pacetone and nads values were determined with an accuracy of 10% and 15%, respectively, while precision was better than 5%.

[10] Studies of acetone adsorption on ice are few. Schaff and Roberts [1998] studied acetone adsorbed on amorphous and crystalline ice by thermal desorption spectroscopy and observed 2 states that desorbed at 140 and 157 K, called aand b- acetone, with respective desorption energies of 35 and 39 kJ/mol. The a-state was readily observed on annealed ice, while the b-state on annealed ice was only observed at very low surface coverage. [11] Picaud et al. [2000] used molecular dynamics to simulate the adsorption of acetone on proton-ordered ice at 0 K. Under those conditions, acetone formed an ordered layer on the ice surface, with 2 molecules per surface unit cell of ice (i.e., one monolayer of acetone on ice consisted of 2.45  1014 molecule cm
2), adsorbed with an energy
Uads =
49 kJ/mol. Since their calculations are at constant pressure and volume, their definition of energy is equivalent

3. Results [8] Adsorption isotherms of acetone on ice (Figure 2) were obtained at 193, 203, and 213 K. At 223 K, n2 was almost as large as n1, resulting in large errors on nads, and values obtained were not reliable. The shape of the 193 K isotherm appears compatible with a Langmuir isotherm or with a Type II isotherm in the BET classification [Gregg and Sing, 1982]. However, the number of data points at high surface coverage is insufficient to determine which one of these 2 types best describes our data. Three desorption data points were obtained at 213 K, indicating that acetone

Figure 2. Adsorption isotherms of acetone on ice at 193, 203, and 193 K. Three desorption data points were also obtained at 213 K.

DOMINE´ AND REY-HANOT: ADSORPTION OF ACETONE ON ICE

-1

Pa )

38

-2

to our definition of enthalpy. With the same approach, Picaud and Hoang [2000] studied the adsorption of acetone on proton-ordered ice between 50 and 150 K. For a surface coverage of 2.45  1014 molecule cm
2, they calculated
Uads =
46.3 kJ/mol at 50 K, that decreased regularly to
41.8 kJ/mol at 150 K. They also performed calculations at 175 K for a coverage of 1.22  1014 molecule cm
2, and obtained
Uads = – 38.9 kJ/mol. At coverages above 1 monolayer, they observed the formation of a second layer over the first one, with
Uads in the range
30 to
34 kJ/ mol, depending on temperature. From these values, they suggested that the b-state identified by Schaff and Roberts [1998] was acetone adsorbed directly on ice, while the astate was overlayers of acetone. [12] Our enthalpy value (
55 ± 7 kJ/mol) was obtained from data at very low surface coverage, and should therefore be compared to the value of the b-state measured by Shaff and Roberts (
39 ± 2 kJ/mol) and to that calculated at 175 K at half layer coverage by Picaud and Hoang (
38.9 kJ/mol). Our value was obtained at lower coverage and higher temperature than both these previous studies, and it can be suggested that both these parameters affect
Hads. Calculations performed by Picaud (personal communication) using a single acetone molecule adsorbing on an ice surface comprising 160 water molecules at 250 K yielded
Uads =
38kJ/mol. High temperature (that would induce more surface disorder) and low coverage then probably cannot explain the difference. Another possibility is that the ice used here, contrary to that used in simulations, also contained defects that may contribute to stronger interactions. In support of this suggestion, Marinelli and Allouche [2001] found from ab initio calculations that adding a defect to the ice structure increased
Uads of acetone on protonordered ice from
31.2 to
53.4 kJ/mol. This latter value is in excellent agreement with ours. [13] The data presented here can be used to determine whether acetone measured in Arctic snow at Alert [Houdier et al., 2002] was adsorbed on the snow crystal surfaces or dissolved in their volume. Their night-time measurements were done while T = 243 K, and the snow had a specific surface area around 400 cm2/g [Domine´ et al., 2002]. Their concentration of 3.1 ppbw then translates into a surface concentration of 8.1  1010 molecule cm
2, i.e., less than 0.001 monolayer, if all the acetone is assumed to be

adsorbed. Pacetone was around 500 pptv [5  10
5 Pa, Guimbaud et al., 2002], and from equation (1), the surface concentration was 3  109 molecule cm
2, which accounts for less than 4% of the value of Houdier et al. This result is valid only if our experimental data can be extrapolated from 213 to 243 K. The ice surface is known to become more disordered at higher temperature [Girardet and Toubin, 2001, and references therein] and changes in surface structure may invalidate this extrapolation. However, Kuroda and Lacmann [1982] report that ice surface properties remain constant in given temperature ranges, and the physical studies and structural models that they report do not show any changes between 243 and 213 K. Acetone adsorption may also enhances surface disorder, especially at 243 K. No detailed studies are available on this point. However, the surface coverages involved at 243 K are