Uptake of acetone, 2-butanone, 2,3-butanedione and 2-oxopropanal on a water surfaceThe uptake of acetone, 2-butanone, 2,3-butanedione and 2-oxopropanal on pure water was investigated by means of a single drop experiment. This method allows to follow the concentrations in an aqueous drop in situ by UV-VIS-spectroscopy. Due to Henry's law saturation, the uptake coefficients γmeas decreased with time. Maximum values (T = 293 K) of 1.8 × 10−4, 1.5 × 10−4, 8 × 10−5 and 1.2 × 10−4 were observed for acetone, 2-butanone, 2,3-butanedione and 2-oxopropanal, respectively. A fit of the experimental data to the resistance model and correction for the effect of gas phase diffusion yield values for the mass accommodation coefficients αacetone = (5.4−2.6+4.5) × 10−3, α2-butanone = (2.1−0.8+0.9) × 10−3, α2,3-butanedione = (1.0 ± 0.3) × 10−4 and α2-oxopropanal = (1.5 ± 0.5) × 10−4. A separate model approach is used to discuss the impact of turbulent mixing of the drop.Oxygenated hydrocarbons, such as the ketones being subject of the present study, are either emitted from biogenic and anthropogenic sources[1] or formed as intermediates in the oxidation of atmospheric pollutants. The photolysis of acetone, the most abundant ketone in the atmosphere, in the upper troposphere (UT) may quite efficiently produce HOx. This channel might even be competitive to HOx formation from O(1D) + H2O under UT conditions.[2]
Because of their high solubility, the mentioned gas phase species can be taken up at the surfaces of liquid tropospheric particles, such as cloud droplets or deliquescent aerosol particles. It is necessary to know the parameters that govern this process in order to evaluate the rate of transfer and thereby the compounds' budgets in either phase. In the present work, the uptake of acetone, 2-butanone (methyl ethyl ketone), 2,3-butanedione (biacetyl) and 2-oxopropanal (methylglyoxal), respectively, on water is investigated. Key parameters of the uptake process are extracted allowing subsequent modeling of the processes studied here.
The uptake process is a convolution of different processes. For species that do not show reactions in the liquid phase these are (i) the gas phase diffusion described by the diffusion coefficient Dg, (ii) the actual transfer across the interface which is governed by the mass accommodation coefficient α, and (iii) the Henry's law saturation described in terms of HD0.5l.[3] Here H is the Henry's law coefficient and Dl is the diffusion coefficient in the liquid phase. For convenience H is given here in the frequently used units of mol l−1 atm−1. This can be converted to the SI units mol N−1 m−1 by mol N−1 m−1 = 9.87 × 10−3 mol l−1 atm−1. γmeas, which is the net probability for the uptake of a molecule of the trace gas species that hits the surface, can be expressed according to the so-called resistance model as follows:[3] Here v is the average thermal velocity of the gas phase molecules, R is the gas constant, T is the temperature, and t represents the time the liquid phase is exposed to the gas phase; Γg accounts for the effect of gas phase diffusion.
A number of determinations of the Henry's law coefficient of acetone and 2-butanone are reported in literature. The H-values at 293 K range from 35 to 45 mol l−1 atm−1 (= 0.34–0.45 mol N−1 m−1) for acetone[4–8] and from 25 to 27 mol l−1 atm−1 (= 0.25–0.27 mol N−1 m−1) for 2-butanone.[4,7,8] For 2,3-butanone only one H-value for 293 K is reported. Betterton states 105 mol l−1 atm−1 (= 1.0 mol N−1 m−1).[6] His result for 298 K is about 30% higher than H = 58 mol l−1 atm−1 (= 0.57 mol N−1 m−1) given by Snider and Dawson.[8] The only determination for the Henry's law coefficient of 2-oxopropanal yields H = 5.42 × 103 mol l−1 atm−1 (= 53 mol N−1 m−1) at 293 K which is very high compared to the other species.[9]
There is only one single determination of the mass accommodation coefficient for acetone with αacetone = 5.4 × 10−3 at T = 293 K.[10] For 2-butanone, 2,3-butanedione and 2-oxopropanal no experimental data for α are available. Possibly, because of this lack of data, these species are not included in current chemical multiphase mechanisms despite of their atmospheric importance, see ref. [11] and references therein.
In a recent publication by Ervens et al[11]. a method for the estimation of unknown mass accommodation coefficients was proposed. According to this, αestacetone = 3.9 × 10−3, αest2-butanone = 2.1 × 10−3, αest2,3-butanedione = 3.5 × 10−3 and αest2-oxopropanal = 2.9 × 10−3 are estimated for T = 293 K.
The uptake experiments were performed in a flow tube reactor (i.d. = 2.13 cm) which includes a section for generation and analysis of a single drop on a pipette’s tip (see Fig. 1). Both phases can be analyzed by UV-VIS spectroscopy. Details of the setup are described elsewhere.[12]
The experiments were done at 293 K in humidified He (>99,999%, Linde) at 10.1 kPa total pressure. The total gas flow in the reactor was set to 1.00 × 104 cm3 min−1 by calibrated mass flow controllers (MKS 1259). Gas mixtures of the carrier and the trace gas were prepared in a special mixture apparatus and stored in large flasks. Before entering the flow tube via the main inlet a small flow containing the trace gas species was added to the main flow with partial pressures of pHe = 8.0 kPa and pH2O = 2.1 kPa.
The pipette for drop generation had an outer diameter of 1.55 mm. The radius of the drops was rdrop = 2.2 mm resulting in a gas droplet contact time of tcont = 2.34 ms. The surface area of the drop Sdrop is calculated to be 1.67 × 10−6 m2 and the liquid volume Vdrop was 6.9 × 10−9 m3. These values were kept constant for all experiments. All drops were made of pure water (Millipore MilliQ, 18 MΩ cm).
Acetone (99.8%, Fluka), 2-butanone (99.5%, Riedel-de Haën) and 2,3-butanedione (99%, Fluka) were used without further purification. The absorption spectra of 1 × 10−2 mol l−1 solutions measured with a commercial spectrometer (Hewlett Packard, HP 8453) are shown in Fig. 2. One band at 265 nm with εmax = 17.5 l mol−1 cm−1 is featured in the spectrum of acetone. 2-Butanone has an absorption maximum of εmax = 20.5 l mol−1 cm−1 at 267 nm. These values compare well with literature data.[13] The spectrum of 2,3-butanedione has two bands in the 200–500 nm region. The strongest one centered at 284 nm has a maximum absorption coefficient of 26.5 l mol−1 cm−1. No comparable literature data were found.
In order to obtain water-free and unpolymerized 2-oxopropanal, the commercially available aqueous solution (40%, Fluka) was processed similarly to the method Schweitzer et al. used to produce pure glyoxal.[14]
The normalized spectrum of diluted 40% aqueous solution of 2-oxopropanal is shown in Fig. 3. A spectrum of a drop exposed to 2-oxopropanal in the gas phase is also shown. Both are qualitatively different. Neither the positions of the band and the minima nor the relative absorptions for different wavelengths match. This is supposed to be due to oligomerisation in the undiluted stock solution.
An attempt was made to prepare a solution of the monomer carbonyl by putting the pure substance into water. Although the resulting solution showed a spectrum that was qualitatively the same as that observed in the single drop experiment, it was not possible to quantify the amount of 2-oxopropanal dissolved. In all experiments a white precipitation was forming if the pure 2-oxopropanal was about to dissolve in the water indicating no complete dissolution of the substance. Therefore, 2-oxopropanal was dissolved from the gas phase into water using the setup described below.
Helium was flowed over the water-free substance and subsequently bubbled through a bottle containing water. After this the gas phase was analyzed by UV-VIS spectroscopy. There was a bypass for the bubbler in order to check the gas phase concentration upstream the water bottle. Gas phase concentrations were calculated from the measured absorption difference between 451 and 470 nm using Δσ (451, 470 nm) = 6.2 × 10−20 cm2.[15] This was done alternating with the bypass of the bubbler shut and open. The concentration downstream the bubbler (bypass shut) was less than 1% of the concentration upstream (bypass open) indicating that almost all 2-oxopropanal is taken up in the bubbler. The amount of 2-oxopropanal taken up after 180 s of bubbling through the water bottle was calculated to be 4.26 × 10−3 mol. Knowing the volume of the water in the bottle from weighing, a liquid phase concentration of 4.7 × 10−2 mol l−1 resulted. The water bottle and the 2-oxopropanal saturator were weighed before and after the measurement. The mass balance was closed within the error limits.
The spectrum of the resulting solution matches the drop's spectrum very well (see Fig. 3). The wavelength dependent absorption coefficient is shown in Fig. 2. One band at 284 nm shows a maximum absorption coefficient of εmax = 16.0 mol−1 l cm−1. No previous measurements of this spectrum are reported in literature.
The gas phase concentrations were determined spectroscopically to be (0.67–11.1) × 1016 molecules cm−3 of acetone, (0.81–3.8) × 1016 molecules cm−3 of 2-butanone, (0.49–2.49) × 1016 molecules cm−3 of 2,3-butanedione and (0.60–2.14) × 1016 molecules cm−3 of 2-oxopropanal, respectively. For this σacetone(280 nm) = 5.05 × 10−20 cm2 molecule−1,[16]σ2-butanone(280 nm) = 5.74 × 10−20 cm2 molecule−1,[16]σ2,3-butanedione(280 nm) = 4.71 × 10−20 cm2 molecule−1,[17] and σ2-oxopropanal(430 nm) = 1.04 × 10−19 cm2 molecule−1,[15] were used.
The primary result of each measurement is the absorbance A(λ) = caqdε(λ) with caq: liquid phase concentration, d: optical path length in the drop, ε: absorption coefficient. A(λ) is the decadic logarithm of the quotient of the intensities of the light at wavelength λ entering and leaving the drop, respectively. As shown in a previous publication,[12] the measured uptake coefficient γmeas in the single drop experiment is related to caq by: where cg is the gas phase concentration of the trace gas species. The factor F is unity for all substances in this study; a is an abbreviation used in the following.
Because of the limited solubility and diffusivity in the liquid phase, evaporation back to the gas phase will occur as the species taken up is accumulating in the liquid. In the resistance model, this is accounted for by the time dependence of the right hand term in eqn. (1) which increases with increasing exposure time t. Merging eqns. (1) and (2) and subsequent integration gives the time dependence of the concentration in the liquid: with and
A fit of the experimental data to eqn. (3) with γ0 and b as parameters gives values for HD0.5l and the uptake coefficient γmeas for t = 0 which equals γ0. The latter includes only the mass accommodation coefficient α and the effect of the gas phase diffusion.
A previously developed computer model was employed in order to correct for this effect and obtain values for α. Shortly, the model solves the diffusion equation in the gas phase numerically for the period of time the gas phase takes to pass the drop tcont. Details are given elsewhere.[12] The result of the model calculation is the dependence of the observed γ0 on the mass accommodation coefficient α. The plots for the four species investigated in this study are shown in Fig. 4. The gas phase diffusion coefficients Dg were calculated from the binary diffusion coefficients DX–He and DX–H2O according to:[18]
Experimental data for the binary diffusion coefficients of the species being the subject of this study are not available. They were calculated using the method of Fuller et al.[19] However, a change of 10% in Dg would only cause a change in the calculated α of less than 5% even for the highest γ0-values observed in this study.
Two limiting cases are visible in Fig. 4. For very low α-values, the uptake is only limited by α itself resulting in γ0 = α. On the other hand, for very high values of α, the gas phase diffusion limits the rate of uptake and γ0 is independent of α. Note that the four curves shown in Fig. 4 are nearly identical showing almost no visible difference.
The time course of the liquid phase concentration of acetone which is shown in Fig. 5 is calculated from the observed differences of the absorptions at 265 and 320 nm, respectively. The difference of the absorptions was applied in order to correct for base line shifts.
The measurements were done at four different gas phase concentrations of acetone. Between 9 and 14 individual experiments were used for averaging. The qualitative behavior was the same for the different gas phase concentrations and Fig. 6 shows the calculated γmeas according to eqn. (2) for all four of them. As expected, no dependence of the uptake coefficient on the gas phase concentration can be seen. The only difference is the decreasing signal to noise ratio with increasing concentration. All curves shown in Fig. 6 show a maximum after about 4 s. This behavior was already found in a previous study[12] and is attributed to the establishment of a steady state mixing of the drop which is driven by the gas phase flowing around. The initial region is excluded in the following fitting procedure because the actual features of the uptake process are only visible after steady state mixing has been reached. The maximum value of the uptake coefficient that was observed was 1.8 × 10−4.
An assumption made in the resistance model described by eqn. (1) is that the liquid phase is infinitely deep.[20] This results in an infinite capacity of the liquid for uptake of the trace species. This, however, can never be the case in a real experiment. Therefore, this model can only fit experimental data for conditions far away from saturation. The calculated saturation concentration for a gas phase acetone concentration of 1.11 × 1017 molecules cm−3 is about 0.17 mol l−1 using H = 40 mol l−1 atm−1.[4] Only data below a threshold of 40% of this value were used for fitting the measured uptake coefficient to eqn. (3).
The parameters of the best fit shown in Fig. 5 are γ0 = 3.0 × 10−3 and b = 4.0 × 103 s−0.5. As expected, the fit line shows strong deviations from the experimental data for longer times. The uptake process becomes slower than predicted as the drop's capacity for the solute gradually saturates. But, nevertheless, the initial behavior of the concentration is represented very well.
After correction for the effect of gas phase diffusion (see Fig. 4), α = (5.4−2.6+4.5) × 10−3 is obtained estimating an overall error for γ0 of 30%. This is in excellent agreement with the data of Duan et al[10]. and the estimation method of Ervens et al.[11]
HD0.5l = (1.5 ± 0.5) × 10−5 mol N−1 s−0.5 is obtained from bvia eqn. (5). Assuming Dl = (1 ± 0.2) × 10−9 m2 s−1 and an overall error for b of 30%, H is estimated to be in the range 29–71 mol l−1 atm−1 which is in agreement with the measurements.
The differences between the absorptions at 267 and 320 nm give the time dependent concentration of 2-butanone in the drop. As shown in Fig. 7, the uptake of 2-butanone is quite similar to that of acetone.
Between 12 and 14 experiments were averaged for gas phase concentrations of 0.81, 2.44, 3.12 and 3.75 in units of 1016 molecules cm−3 of 2-butanone. The maximum value of the time dependent uptake coefficient calculated according to eqn. (2) was 1.5 × 10−4, independent of the gas phase concentration. Using H = 27 mol l−1 atm−1,[4] the saturation liquid phase concentration for cg = 3.8 × 1016 molecules cm−3 is about 4.0 × 10−2 mol l−1. Using a 40% of this value threshold, the same fitting procedure was employed as for acetone. The parameters for the fit shown in Fig. 7 are γ0 = 1.6 × 10−3 and b = 4.9 × 103 s−0.5 with an estimated overall error of 30%. α = (2.1−0.8+0.9) × 10−3 is obtained after correction for the effect of the gas phase diffusion using Fig. 4 which equals the estimation according to Ervens et al.[11]
HD0.5l = (1.1 ± 0.3) × 10−5 mol N−1 s−0.5 can be extracted from bvia eqn. (5). With Dl = (1 ± 0.2) × 10−9 m2 s−1, H is estimated to be between 23 and 50 mol l−1 atm−1 which is in agreement with the measurements.
Uptake of 2,3-butanedioneUsing the differences of the absorptions at 284 and 334 nm the time dependent concentration in the drop, see Fig. 8, was calculated. The measured time dependent uptake coefficient γmeas is shown in Fig. 9 for the different gas phase concentrations of 2,3-butanedione that were used. The mean of at least 12 individual measurements for each gas phase concentration was taken for calculation according to eqn. (2). For times larger than 7 s there is no significant difference between the four graphs. The variations before this time are attributed to the onset of the mixing of the drop. The highest uptake coefficient that was observed is around 8 × 10−5 neglecting the first 7 s.
Fig. 8 shows that the saturation of the drop after 100 s is very small unlike the acetone and 2-butanone experiments. This is at least partially due to the much larger Henry's law coefficient. Using H = 105 mol l−1 atm−1,[6] all experimental data are within 30% of the saturation concentration which is around 0.1 mol l−1 for cg = 2.49 × 1016 molecules cm−3. Consequently, all data were used for fitting to eqn. (3) except the first 5 s. The best fit shown in Fig. 8 as a solid line is obtained using γ0 = 1.0 × 10−4 and b = 1.9 × 103 s−0.5. Employing the computer model for correcting for the gas phase diffusion effect, α = (1.0 ± 0.3) × 10−4 and HD0.5l = (2.6 ± 0.8) × 10−5 mol N−1 s−0.5 are obtained. The correction is smaller than 5% in this case indicating that the uptake is hardly limited by gas phase diffusion. This value for the mass accommodation coefficient is more than one order of magnitude lower than the estimation according to Ervens et al.[11]
The estimation for H assuming Dl = (1 ± 0.2) × 10−9 m2 s−1 is 53 < H/mol l−1 atm−1 < 122 covering literature data.
The time dependent concentration of 2-oxopropanal in the drop is calculated from the difference of the absorptions at 284 nm and 334 nm, respectively. The picture shown in Fig. 10 is quite similar to the one for 2,3-butanedione because H is high for this species, too. A saturation concentration of caq = 4.6 mol l−1 is calculated for cg = 2.14 × 1016 molecules cm−3 with H = 5.4 × 103 mol l−1 atm−1.[9] All experimental data are well below 2% of this. Therefore virtually no effect of saturation is expected. This is verified by the plot which nearly is a straight line with constant slope.
Measurements were done for gas phase concentrations of 0.6, 0.88, 1.35 and 2.14 × 1016 molecules cm−3 of 2-oxopropanal. 14 Individual experimental runs were averaged, respectively. For all four concentrations the calculated uptake coefficients γmeas showed the same time course similar to that of 2,3-butanedione with a maximum of about 1.2 × 10−4.
The fit to eqn. (3) was performed using all data except the first 5 s yielding γ0 = 1.5 × 10−4 and b = 7.5 × 102 s−0.5 with estimated overall errors of 30%. The correction for the gas phase was less than 5% leading to α = (1.5 ± 0.5) × 10−4. This value is failed by about a factor of 20 by the estimation method of Ervens et al[11].HD0.5l = (7.2 ± 2.2) × 10−5 was calculated via eqn. (5). From this, H is estimated to be in the range (1.4–3.4) × 102 mol l−1 atm−1. This is more than one order of magnitude lower than the value given by Betterton and Hoffmann.[9] This discrepancy might be due to different influences of the oligomerisation of the carbonyl in both experiments. However, the estimate is still much higher than the values for the other species.
A different model approachThe resistance model shows strong deviations from the experimental data for larger times as can be seen in Figs. 5 and 7. This is due to the assumptions made for deriving eqn. (1) which are not fully met by the present experiment.
In order to check which are the effects of these assumptions, the uptake process is treated numerically in an explicit form. The liquid was treated as a sphere with concentrical layers of thickness Δr. The fluxes of the trace species in and out each sphere where calculated for discrete time steps according to the diffusion equation in spherical coordinates: The flux into the outermost layer which results from the uptake of the trace species from the gas phase is: Here p is the partial pressure of the trace species in the gas phase.
For a test of the explicit model, the radius of the drop was set to infinity. This situation is comparable to the resistance model where an infinitely deep liquid with a plane surface is assumed for the derivation of the resistance of the bulk liquid (right hand term in eqn. (1))[20].H and γ0 were taken from the best fit in Fig. 5.
Both models are compared for the uptake of acetone in Fig. 11. The principal behavior of the explicit model (dashed line) is the same as that of the resistance model (straight line). There is only a little shift which seems to be built up at early times. The shift might be due to another assumption implied in the resistance model, i.e. the decoupling of the individual processes convoluted in the phase transfer. This is expressed by an instantaneous saturation of the liquid layer adjacent to the surface. The outermost layer becomes only saturated step by step in the explicit model which results in a steeper concentration gradient. Therefore, the flux is larger in the first phase of the uptake process.
Setting rdrop to the experimental value of 1.1 × 10−3 m and keeping all other parameters constant gives the dash-dot plot in Fig. 11. While the initial behavior remains unchanged, the modeled concentrations for larger times become smaller than in the resistance model. This slower increase is in good agreement with the experiment. Therefore, the deviation of the resistance model from the experiment seems to be mainly resulting in the radial symmetry of the system.
Another effect that is not included in the resistance model is possible turbulent mixing of the bulk liquid. Danckwerts who derived the resistance term for liquid phase saturation in eqn. (1) assumed a ‘quiescent liquid’ in which mass transport is only performed by diffusion.[20] On the other hand, the presence of a mixing motion was observed in the single drop experiment.[12] This is supposed to be driven by the gas phase flowing around the drop. It takes along the surface layer of the liquid by means of friction from the most up stream area of the drop to the most downstream one. In these regions the liquid is supposed to come from and to go into bulk liquid, respectively. This, in turn, might induce turbulent mixing in the core of the drop.
In order to approach this situation in the explicit model, the liquid sphere was divided into two parts, (i) the inner core of radius rmix which is totally mixed resulting in a uniform concentration profile and (ii) the outer shell which follows the gas phase laminarly and where radial transport is performed by diffusion only. This is called the ‘Film model’ in the treatment of Danckwerts.[20] The results for different radiuses of the core are compared with the experimental data for acetone in Fig. 12. It can be seen that, although the initial behavior is unaffected, the deviation from the experiment gets larger with increasing core radius.
It is concluded that in a drop under the present experimental conditions the effect of turbulent mixing on radial transport is negligible. It can be proposed that the observed mixing process proceeds as follows, see Fig. 13 for illustration. The surface liquid is flushed to the downstream part of the drop laminarly. Subsequently it is passing the drop in a narrow tube around its symmetry axis that is parallel to the gas flow in the converse direction. At the end of this tube the liquid is again flushed onto the surface and dragged along by the gas phase again. All other parts of the drop follow this circulation laminarly with tracks of a similar shape but smaller size. This flow pattern was also the result of study by Chen who solved the governing differential equations for the diffusion simultaneously with those for convective transport.[21]
The application of the resistance model is verified only for conditions far away from liquid phase saturation as it was done in this study.
Summary and implications for the atmosphereThe uptake of four ketones was investigated using a single drop experiment that allows in situ analysis of the liquid phase by absorption spectroscopy. After fitting to a model values for the mass accommodation coefficients α are obtained. For acetone and 2-butanone similar values of some 10−3 are obtained. For mass accommodation coefficients of this order of magnitude the rate of gas uptake under tropospheric conditions is affected by both Γdiff and α.[22]
However, under the experimental conditions of this study, i.e. large drops with r ≈ 10−3 m, saturation is reached after roughly 102 s. The capacity of tropospheric cloud droplets with r ≈ 10−5 m is about six orders of magnitude lower. Therefore, saturation is expected to occur on a very short time scale if no reactive sink in the liquid phase would be present.
For 2,3-butanedione and 2-oxopropanal mass accommodation coefficients somewhat larger than 10−4 are obtained. The uptake rate under tropospheric conditions is therefore predominantly limited by α. The Henry's law coefficients for these bicarbonyls are much larger than for the monocarbonyls. Therefore, Henry's law saturation is not reached on the timescale of the present experiment. However, regarding atmospheric aerosol particles having a much lower volumes, it is expected to be reached on a still very short timescale even if the very high H for 2-oxopropanal determined by Betterton and Hoffmann is considered which is about two orders of magnitude larger than that for acetone.[9]
According to the present knowledge, tropospheric aqueous phase radical chemistry could establish a sink driven by OH during the day and NO3 during night-time. If Henry's law equilibrium is established for the trace species, the relative importance of the reactions in the gas and aqueous phase, respectively, can be estimated. In Table 1 this is done considering only the reactions with OH in either phase of a continental cloud at 25 °C. The partitioning of the trace species between the two phases Naq/Ng is calculated as the product of HRT and the liquid water content (LWC). In terms of molecule number fractions, less than one per thousand molecules of acetone, 2-butanone and 2-3-butanedione are in the aqueous phase. For 2-oxopropanal having a much higher Henry's law coefficient, only a few percent of the molecules are dissolved.
Nevertheless, the picture changes if the ratio of the rates of the reactions between the ketones and OH in either phase are regarded. The rate of the aqueous phase reaction is at least four orders of magnitude larger than that for the gas phase reaction for all substances being subject to this study. This is due to the higher concentrations of the reactants in the aqueous phase.
Multiplying the partitioning Naq/Ng with the ratio of the rates Raq/Rg gives the ratio of the conversions due to the reaction with OH in the aqueous and gas phase, respectively. This value is given in the last column of Table 1. Obviously, the conversion by the aqueous phase reaction is much higher for all substances.
The uptake parameters obtained in this study will be used for the extension of atmospheric multiphase chemistry models such as CAPRAM in its different versions, see ref. [11] and references therein.