Examination of Organic Vapor Adsorption onto Alkali Metal and Halide Atomic Ions by using Ion Mobility Mass Spectrometry

Abstract We utilize ion mobility mass spectrometry with an atmospheric pressure differential mobility analyzer coupled to a time‐of‐flight mass spectrometer (DMA‐MS) to examine the formation of ion‐vapor molecule complexes with seed ions of K+, Rb+, Cs+, Br−, and I− exposed to n‐butanol and n‐nonane vapor under subsaturated conditions. Ion‐vapor molecule complex formation is indicated by a shift in the apparent mobility of each ion. Measurement results are compared to predicted mobility shifts based upon the Kelvin–Thomson equation, which is commonly used in predicting rates of ion‐induced nucleation. We find that n‐butanol at saturation ratios as low as 0.03 readily binds to all seed ions, leading to mobility shifts in excess of 35 %. Conversely, the binding of n‐nonane is not detectable for any ion for saturation ratios in the 0–0.27 range. An inverse correlation between the ionic radius of the initial seed and the extent of n‐butanol uptake is observed, such that at elevated n‐butanol concentrations, the smallest ion (K+) has the smallest apparent mobility and the largest (I−) has the largest apparent mobility. Though the differences in behavior of the two vapor molecules types examined and the observed effect of ionic seed radius are not accounted for by the Kelvin–Thomson equation, its predictions are in good agreement with measured mobility shifts for Rb+, Cs+, and Br− in the presence of n‐butanol (typically within 10 % of measurements).


Introduction
Ion-induced nucleation [1,2] occurs when it is energeticallyf avorable for vapor molecules to adsorb repeatedly ontoions, growing them substantially in size (into droplets). The study of ioninduced nucleation is of fundamentali mportance in understanding condensed phase speciesf ormation from vapor [3][4][5][6] and also finds applicationi nt he design of condensation based detection systems (i.e. for analytes in the vapor phase [7][8][9] ). Classical modelso fi on-induced nucleation, which incorporate the Kelvin [10,11] and Thomson effects [12] to evaluatet he vapor pressure of as mall droplet,c an be used to predict both ion induced nucleationr ates and activation efficiencies for vapor molecule-ion complexes; however,s uch predictions are not in agreement with all experimental measurements. [6,[13][14][15][16] Most notably,c lassical Kelvin-Thomson-basedm odelsc an explain neither observed dependencies on the sign of the ion, [3] nor observed dependencies on the ion chemical composition. [4,[17][18][19][20][21] Model predictions in ion-induced nucleation are heavily dependentu pon the properties of the so-called criticalc luster, [22] that is, the ion-vapor molecule complex of maximum free energy,w hich is typicallyi nt he nanometer to subnanometer size range and is composed of al imited number of vapor molecules. To better understandw hy discrepancies arise between classical predictions and measurements, it is also desirable to probe the properties of ion-vaporm olecule complexes at the size scale of critical clusters. [16,23,24] However,t he majority of experimental approaches to examine ion-induced nucleationr ely upon detection of nucleated droplets significantly larger than the criticals ize, [22,[25][26][27] with nucleation theorem based extrapolation applied to infer properties of critical clusters. Distinct from these techniques is ion mobility mass spectrometry, [28][29][30] which, via doping driftg ases with organic vapor molecules, has recently been employed to examine ion-vapor molecule complexes. [19,20,[31][32][33][34] Thoughv apor molecules typically desorb from seed ions in mass spectrometer inlets, during ionmobility measurement, which takes place at controlled pressure and temperature, ions and the surroundingv apor molecule are in equilibrium with one another.M easurement of shifts in an ion's mobility with changes in vapor saturation ratio can then be used to infer the extent of vapor molecule adsorption. [19,31] As vapor dopant concentrationsa re below saturation during ion mobility measurements, such experiments are the converse to the traditional manner in whichi on-induced nucleationi s examined;t raditionally micrometer sized droplets (supercritical sizes) formed under supersaturated vapor conditions are probed, while in ion mobility-mass spectrometry nanometer scale complexes (subcriticals izes) are studied. At the same We utilize ion mobility mass spectrometry with an atmospheric pressure differential mobilitya nalyzer coupled to at ime-offlight mass spectrometer (DMA-MS) to examine the formation of ion-vapor molecule complexes with seed ions of K + ,R b + , Cs + ,B r À ,a nd I À exposed to n-butanola nd n-nonanev apor under subsaturatedc onditions. Ion-vapor molecule complex formationi si ndicatedb yashift in the apparent mobility of each ion. Measurement resultsa re compared to predicted mobility shifts based upon the Kelvin-Thomson equation, which is commonly used in predicting rates of ion-induced nucleation. We find that n-butanola ts aturation ratios as low as 0.03 readily binds to all seed ions, leading to mobility shifts in excess of 35 %. Conversely,t he binding of n-nonane is not detectable for any ion for saturation ratios in the 0-0.27 range. An inverse correlation between the ionic radius of the initial seed and the extent of n-butanol uptake is observed, such that at elevated n-butanol concentrations, the smallest ion( K + )h as the smallest apparent mobility and the largest (I À )h as the largest apparent mobility.T hough the differences in behavior of the two vapor molecules types examined and the observed effect of ionic seed radius are not accounted for by the Kelvin-Thomsone quation, its predictions are in good agreement with measured mobility shifts for Rb + ,C s + ,a nd Br À in the presence of n-butanol (typically within 10 %ofm easurements). time, thoughv apor concentrations are below saturation, they can be higher than are achievable in high pressure-mass spectrometry,atechnique which has been used previously in examined ion-vapor molecule complexes formed in subsaturated conditions. [35][36][37] Therefore, ion mobility-mass spectrometry measurements are well suited to providei nformation on the earliest stages of ion-vaporm olecule complex formation.
To date, studies utilizing ion mobility-mass spectrometry to examinei on-vapor molecule complexes have been focused on proof-of-concept measurements, [32,33] the development methods to analyze and interpret results, [31] examination of how complexf ormation influencesi nstrument calibration, [38] and the examination of water and alcoholu ptakeb ys alt cluster ions. [19,20] Though the latter are of interest in understanding ion-induced nucleation,c omparison to theoretical predictions is complicated by the possibility that salt cluster ions mayp artially or wholly dissociate upon vapor molecule adsorption (as is suggested by computational predictions [19] ). This has an influence on the ion-vapor molecule complex free energy (solvation energy), and is difficultt oq uantify without the use of computational approaches specific to the cluster ion and vapor molecule under examination. As impler examination of vapor uptake would involve the use of atomic ions as seeds for vapor adsorption, for which dissolution or changes in ion conformation upon vapor adsorption need not be considered. The purpose of this study is to perform measurements along these lines. Specifically,w eutilize an atmospheric pressure differentialm obility analyzer coupled to am ass spectrometer (DMA-MS)t oe xamine the formation of alkali metal cation complexes with n-butanola nd n-nonane( which have been utilized prevalently in ion induced nucleation/condensation experiments [39,40] ), as wella sh alide anion complexesw ith the noted organic species. Results are comparedt omodified classical predictionsu sing the analysisf ramework described by Oberreite tal. [18,19] and Rawat et al., [31] linking the shift in mobility/collision cross section( inferred from mobility measurements)b rought about by vapor molecule adsorption to the equilibriums orptionc oefficients for successive adsorption events.

Experimental Section
The DMA-MS system is described in detail in prior studies. [19,20,[41][42][43] Briefly,i tc onsists of ap arallel-plate DMA (P5, SEADM, Boecillo, Spain, with ar esolving power in excess of 50) coupled with a QSTAR XL quadrupole-time-of-flight mass spectrometer (MDS Sciex). Atomic ions were generated via electrospray ionization of 10 mm methanol solutions of potassium, rubidium, and cesium iodide salts, as well as tetraheptylammonium bromide (purchased from Sigma-Aldrich, St. Louis, MO, USA). Positive mode was employed to generate cations, and negative mode was employed for anions. Electrospray ionization of salt solutions generates primarily singly and multiply charged cluster ions. [44] Here, we focus only on measurement of the atomic cations/anions produced. Ions were drawn into the DMA electrostatically against a0 .2 Lmin À1 counterflow of ultrahigh purity air (Airgas). The DMA sheath flow was also ultrahigh purity air,a nd was maintained at at emperature in the 303-305 Kr ange via application of aw ater based heat exchanger. For mobility measurements, the potential difference across the DMA was scanned from 500 to 2500 Vi n1 0V increments. Mass spectra were recorded at each voltage step using the time-of-flight section of the mass spectrometer.C ontrolled amounts of n-butanol and n-nonane vapor were introduced into the DMA sheath using a constant output nebulizer described previously. [18,19] Prior to all measurements, the entire system was allowed to operate for more than two hours, to ensure thermal equilibration of the DMA sheath flow and that vapor concentration profiles within the DMA were uniform. Between measurements, the DMA-MS system was not used for any other experiments, in order to minimize the potential for contamination from other chemicals. The compounds n-butanol and n-nonane were chosen for several reasons. First, they have been examined in prior ion-induced nucleation experiments [39,40] with clusters/particles in the nanometer size range. Second, n-butanol is prevalently used in condensation particle counters, [9,17] which are commercially available devices used to detect ions/nanoparticles in the gas phase via condensation of n-butanol onto analytes (growing them to sizes detectable via light scattering). Third, these solvents, of clearly disparate molecular structure, have similar saturation vapor pressures at 304 K( 1.3 kPa for n-butanol and 0.6 kPa for n-nonane) and similar surface energy densities (0.024 Jm À2 for n-butanol [45] and 0.023 Jm À2 for n-nonane). Shown subsequently, classical theory predictions of the extent of uptake are dependent upon the saturation vapor pressure (defining the saturation ratio) and the surface energy density,h ence it is of interest to examine solvents with similar bulk properties yet distinct molecular structures.
To quantify vapor uptake by ions, the potential difference in the DMA required to maximally transmit each examined cation and anion was monitored as af unction of saturation ratio. In differential mobility analysis, the potential difference is linearly proportional to the inverse mobility of the ions transmitted. [46] DMA calibration was performed both in the absence and in the presence of organic vapor by determining the voltage required to transmit the tetraheptylammonium ion, whose inverse mobility (1.03 Vscm À2 ) was measured in air at atmospheric pressure by Ude and Fernandez de la Mora. [47] As noted in several studies [31,38] and also observed here, this ion's mobility appears insensitive to saturation ratio (the voltage required to maximally transmit it does not vary substantially) and it does not appear to form complexes with either of the vapor molecule types examined in the test saturation ratio range.

Ion-Vapor Molecule ComplexM obilities
In total, we made measurements of the inverse mobilities of K + ,R b + ,C s + ,B r À ,a nd I À at 304 Ka nd atmospheric pressure in air,w ith butanol saturation ratios in the 0-0.17 range and nonanes aturation ratios in the 0-0.27 range (similars aturation ratio rangesw ere accessible because of the similar saturation vapor pressures of theset wo solvents). Inverse mobility is proportionalt ot he apparent collision cross section of the ion under measurement conditions, hence larger inverse mobilities correspond to larger ions (i.e. larger ions have smaller mobilities). The inverse mobilities of the formed ion-vapor molecule complexes are plotted in Figure 1a  DMA, it is important to note that the number of vapor molecules bound within an ion-vapor molecule complex is not a constant; each complex is in equilibrium with its surroundings and probest he equilibrium distribution of vapor molecules bound (which is af unction of the vapor molecule sorption and desorption rates). [18] Therefore, the measured inverse mobilities do not correspond directly to complexesw ith as pecific number of bound vapor molecules.M odeling in the subsequent sectioni su sed to comparem easuredi nverse mobilities with theoretical predictions.E ven without such modeling, it is evident that ion-butanolc omplexesf orm readily,a si on inverse mobilities increase with increasings aturation ratio. Meanwhile n-nonaned oes not adsorb onto any of the examined ions (at the examined saturation ratios). Qualitatively, this is in agreement with the dropleta ctivationm easurementso fW inkler et al., [39] who found that smaller sized tungsten oxide seed ions could be used to initial droplet growth of n-propanol vapors than could be used for n-nonane. The increasei ni nverse mobility for butanol is most pronouncedf or the cations, and is inverselyc orrelated with ion mass/size;t hough potassium is the smallest ion examined, upon introduction of butanol to the DMA it has the largesti nverse mobility.D ata hence reveal ac lear sign dependency for butanol uptake, as well as a size dependency.The magnitude of increaseininverse mobility (more than af actor of 2f or the cationsa ts aturation ratios greater than 0.10) is larger than what has been observed in prior studies where the vapor dopants were water [19,48] and isopropanol. [31,32] In Li and Hogan, [20] ion-vapor molecule complex formation was examined for( NaCl) n Na + and (NaCl) n Cl À ions with n-butanol, ethanol, methyl ethyl ketone (1-butanone), and toluene vapor molecules. Though such ions potentially dissolve/change structure during complex formation, similar findings were observed in this study. n-Butanol, for ag iven solvent vapor concentration, led to the largests hifts in mobility for all sodium chloride clusteri ons;i nverse mobility shifts of more than af actor of 2w ere observed. Adsorption of ethanola nd methyl ethyl ketone led to increases in inverse mobility above 1.5 (in as imilarv apor concentration range), whilet oluene, which, like n-nonane, has ad ipole momentb elow 0.5 D, led to minimal inverse mobility shifts.
Prior to more detailed model comparison, we remark that the initial inverse mobilities of the atomic ions in dry air are also within expectations.F or the five ions examined, Ta ble 1 lists the measured inverse mobility,a sw ell as the predicted inverse mobility based upon the gas molecule scattering calculation approachd escribed by Larriba and co-workers. [49][50][51][52][53] Calculations were performed modeling ions as spheres with radii equivalent to their ionic radii (noted in the table) and gas molecules as spheresw ith effective radii of 0.15 nm (based on prior measurements [54,55] ). The ion-induced dipole potential between ions and gas moleculesw as also considered (with ag as molecule polarizability 1.7 10 À30 m 3 ), and has al arge impact on the predicted inverse mobilitieso fa tomic ions. Calculations were performed modeling ion-gasm olecule collisions as completely elastic and specular( elastic hard spheres cattering, EHSS), as well as with the diffuse-inelastic scattering model (diffuse hard sphere scattering, DHSS)o fL arribaa nd Hogan. [50] While prior studies [41,42,48] reveal that gas molecule-ion collisions in diatomic gases are neither whollys pecularn or wholly dif-  With the exceptiono f the potassium cation, we find this to be true;m easurements are bounded by the DHSS prediction as an upper limit the EHSS calculation as al ower limit. We suggestt hat the anomalously high inverse mobility of the potassium cation maybea ttributable to either the transient adsorption of contaminant vapor species duringm obility analysis (this could shift mobilities by several percentf or all examined ions);a lthough efforts were made to minimize contamination of the system, completely removing all potential condensable species in ion mobility measurements has been shown to be difficult. [38]

Comparison to Classical Model Predictions of Vapor Uptake
Because we find non-negligible mobility shifts in the presence of n-butanol only,w ec ompareamodel of the mobility shift of ions in the presence of this vapor to measurements. Following the procedure developed in Oberreite tal., [18,19] the mobility of an ion (K S )e xposed to vapor at saturation ratio S relative to its mobility in the absence of vapor (K 0 )c an be computed using Equation (1): where P g is the probability an ion-vaporm olecule complex has g vapor molecules adsorbed to it at equilibrium (at the prescribed saturation ratio), m 0,b is the reduced mass of the bare ion and the bath gas, mg ,b is the reduced mass of the ion-vapor molecule complex containing g vapor molecules, and W g is is the collision cross section of ion-vapor molecule complex consideringc ollisions with the bath gas (with W 0 the bare ion collision crosss ection). Equation (1) is developed accounting for the fact that if an ion equilibrates with the surrounding vapor during mobility measurement, the number of vapor molecules bound is not ac onstant, ratherv apor molecules continually sorb and desorb from each complex,w ith the probability of finding an ion-vapor molecule complex containing precisely g vapor molecules determined by the equilibrium binding coefficients for individual vapor molecules. Equation (1) neglects the influenceo fc ollisions between the dopant vapor and ionvapor molecule complex on drag/mobility.F or the vapor pressures examined in this study we find this influence negligible, thoughn ote it has been shown in prior work to lead to a small, linear change (with vapor concentration)i nt he mobility of an ion in the absence of binding. [31] Implementation of Equation ( In Equation (2), k a,g-1 is the association rate coefficient for the noted reaction, k d,g is the dissociation rate coefficient, kT is the thermale nergy,a nd DE g is the enthalpyd ifference between the ion-vapor molecule complexes [IV] g and [IV] g-1 at saturation. P g can be expressed in terms of such equilibrium coefficients [Eq. (3a) and Eq. (3b)]: [19] As in prior studies, [18] the association and dissociation rate coefficientsc an be approximated as [Eq. (4a) and Eq. (4b)]: where mg ,v is the reduced mass for av apor molecule and an ion-vapor molecule complex (composed of g molecules), r g is the effective radius of an ion-vapor molecule complex, r v is the effective radius of av apor molecule, kT is the thermale nergy, and h D is ad imensionless enhancementf actor accounting for the influence of the ion-dipole potentialo ni on-vapora ssociation (considered for butanol only). We approximate this factor using the equation [Eq. (4c)]: where m D is the permanent dipole moment of the vapor molecule (1.66 Df or n-butanol), ze is product of the ion absolute charges tate and the unit charge, and C 1 is aconstant quantifying the fraction of time the dipole is aligned in the direction of the ion-vapor molecule complex (taken to be 0.6 here). Nadytko and Yu [5] have presenteda na lternative equationt oE q. (4c), which can be expressed as [Eq. (5)]: with decreasing ion size, [56] which is physically unreasonable, and predicts rates in excesso ft he C 1 = 1 in Equation (4c), which is the fully alignedd ipole collision rate derived via the approacho fV asil'ev and Reiss. [57] We therefore utilize Equation (4c) in all calculations presented here.
In Equations (4 a-c), the ion-vapor molecule complex and vapor molecule are modeleda ss pheres. While prior work shows that ac ollision radius cannot be universally defined for an on-spherical entity withoutc onsidering the size and shape of its collision partner,( e.g. r v should depend upon g) [58,59] we find that the spherical approximation does not strongly affect model predictions here. The ionicr adiif or each atomic ion, providedi nT able 1, are used for r 0 of ions with no vapor molecules bound, and ion-vapor molecule complex radii calculated using the equation [Eq. (6)]: The radius of ab utanol monomer was approximated from its molecular weight and bulk density,with avalue of 3.3 .
Following the classical ion induced nucleation approach, DE g can be writtenast he sum of two terms [Eq. (7a)]: where the subscript K and T denote the Kelvin and Thomson contributions to the free energy, respectively.The Kelvin contribution can be writtenas[ Eq. (7b)]: where s is the surface tension/surface energyd ensity of the ion-vaporm olecule complex (assumed to be the surface tension of the condensed vapor species, 0.024 Jm À2 ). The Thomson contribution is [Eq. (7c)]: where e 0 is the permittivity of free space,a nd k is the complex dielectric constant (i.e. the condensed vapor dielectric constant, 17.8). These two terms combined serve as the basis classical ion-induced nucleationt heory predictions. [1,6] DE g;K is a positive term and quantifies the enthalpy barrier to growth, while DE g;T is an egative term and quantities the barrier reduction brought about by the presence of charge. Sample Equation (3a, b) calculations of P g for n-butanol with the Rb + ion are providedi nFigure 2. For each saturation ratio, an on-negligible probability is found for multiple ion-vapor molecule complexes, with the number of ion-vapor molecule complexf or which P g ! 0.01 increasing with increasing saturation ratio. This highlightst he importance of accountingt he sorption and desorptiono fv apor molecules from complexesd uring measurement;i ti si mprobablet hat an ion would traverse the mobility analyzer with ac onstantn umber of vapor molecules bound to it.
For (b), following prior studies of the collision crosss ections of nanometer scale ions, [49,50] we approximate the ratio where r b is the effective radius of the bath gas (1.55 )a nd L Y p;g ÀÁ describes the influencet he induced-dipole potential between bath gas molecules and the ion-vapor molecule complex have on momentum transfer upon close approach. We calculate L Y p;g ÀÁ using the equation from Larribaa nd Hogan [Eqs. (9a), (9b) and (9c)]: [50] L Y p;g ÀÁ ¼ 1 þ 0:329Y p;g þ 0:089Y 2 p;g Y p;g 1 ð9aÞ L Y p;g ÀÁ ¼ 1 þ 0:368Y p;g À 0:005Y 2 p;g Y p;g > 1 ð9bÞ Equation (9b) is only required for calculations involving the bare ion Y p;0 ÀÁ ,asi ti so nly these ions for which Y p;g > 1. Ac omparison of Equation (1) predictionst om easurements in terms of the ratio K 0 /K S (whichi ncreases with increasing saturation ratio) is shown in Figure 3. Experimental measurements and model predictions are in qualitativea greement;b oth show ar apid increase in K 0 /K S at low saturation ratios,f ollowed by am ore graduali ncrease. This is in contrast to Kelvin-Thomson predictions for larger ions; [18,19] in these instances as mall increase( below 10 %) in inverse mobility is predicted saturation ratios below 0.10, and then ad rastic increase at higher saturation ratio (dependent upon the modeled activity coefficient).A dditionally,f or largerc luster ions, Kelvin-Thomsonp redictions have been found to be in poorer agreement with ion mobility-mass spectrometry measurements than have simpler, Langmuir adsorption based models. [19,20,31] However, for the Rb + , Cs + ,a nd Br À ions examined here, model predictions are within 10 %o fm easured K 0 /K S values, suggesting that deviations observed in prior studies are at least partially attributable to the influence of cluster ion dissolution upon vapor sorption, as well as the influences of ion structure on sorption. Poorer agreement is observed for K + ,a nd I À ,t he smallest and largest ions examined, with underprediction in the extent of mobility shift forK + and overprediction for I À .T hough the comparison is not shown, poorer agreement is also found betweenp redictions and measurements of n-nonane facilitated K 0 /K S shifts. The n-nonanea nd n-butanol model predictions primarily differ in that n-nonane has an egligible dipole moment, and while this reduces the predicted extent of mobility ratio shift, it does not lead to model predictions of zero shift (as is observed for n-nonane). Therefore, while Kelvin-Thomson predictions fit some results well, we caution against universal application of this modelt od escribe vapor uptake, even by atomic ions. Without utilizing the bulk surface tension of n-butanol in modeling and instead fitting the surface tension to measurements (minimizing the square error),f or K + ,R b + ,C s + ,B r À ,a nd I À ,w ef ind effective surface tensions of 0.010, 0.022, 0.025, 0.027, and 0.044 Jm À2 ,r espectively, suggesting that the effective surfacee nergy scales with ionic radius. The finding that smaller atomici ons uptake organic vapor to ag reater extent (reflected in the smalleri nferred surfacet ension)i si ng ood agreement with the high pressure mass spectrometric measurements of Dzidic and Kebarle, [37] who found that smaller alkali metal ions form larger complexesw ith water vapor than do larger ions, with Li + exhibiting the largest (negative) enthalpy and Gibbs free energy change upon water vapor binding, and Cs + the smallest enthalpy and Gibbs free energyc hange. However,t he correlation betweenc ore ion size and extent of uptake does not appear to be universal for all vapor-atomic ion combinations;h igh pressure mass spectrometry also reveals that the monovalent Sr + ion would adsorb more water vapor than the Li + ion, [36] and Castleman et al. [35] observed a weaker link between the extent of ammonia uptake and ion size. Using the effective surface energies noteda bove,t he "prenucleation cluster" size can also be extrapolation asafunction of saturationr atio, and is plottedi nF igure 4. The prenucleationc luster size corresponds to the largest size at which Equation (2) predicted equilibrium coefficients are greater than unity.F ollowing directly from the fit surfacee nergies, the largest prenucleation clusters are predicted for K + and the small-  est forI À .U nder subsaturatedc onditions, for Rb + ,C s + ,a nd Br À ,prenucleation ion-vapor molecule complexes are anticipated to have between 3a nd 30 vapor molecules bound;t he largesto ft hese clusters would be expected to have effective diameters near 1.4 nm, below the size detectable in condensation based particle detectors, [17] but potentially with masses in excess of 1000 Da, which is larger than commonly encountered in ambient environments. [41]

Conclusions
We apply ion mobility-mass spectrometry to examine the formation of ion-vapor molecule complexes with seed ions of K + , Rb + ,C s + , Br À ,a nd I À with n-butanola nd n-nonane as the vapors,i na ir at atmospheric pressure near 304 K. Mobility shifts can be directly compared to model predictions based upon the Kelvin-Thomson equation, which is commonly invoked to predicti on induced nucleation rates. Based on these studies, we draw the following conclusions: 1. As was recently observed for sodium chloride cluster ions, [20] the extent of mobility shift observedf or atomici ons exposed to butanol is substantial;m obilities of all test ions are reduced by more than 40 %a tb utanol saturation ratios of 0.17. Conversely, n-nonaned oes not appear to bind to atomici ons at similarly low saturation ratios, as even the transient binding of as ingle n-nonane molecule would be lead to ad etectable mobility shift. Ion mobility-mass spectrometry experiments hence confirmastrong chemical dependency in the earliest stages of ion-molecule complex formation in the vapor phase. 2. While the datad os uggest that n-butanolb inds more strongly to cations than to anions (more uptake is observed for positivelyc harged species), ac learer correlation is observed between the extent of mobility shift and ion size, with greater sorption observed for smaller atomici ons. The difference in the extent of sorption is large enoughs uch that smaller atomic ions have lower mobilities at elevated butanol concentrations, that is, smaller ions actually form largeri on-vapor complexes. This is consistent with high pressure mass spectrometry experiments with alkali metal seed ions and water vapor. [37] However, it is not accounted for in Kelvin-Thomson equation predictions and suggests the binding energieso fs olvent molecules are not accurately predicted by this simple model alone. 3. Though the Kelvin-Thomson model does not accurately capture the influence of vapor molecule structure on complex formation (andp rior work has shown it is difficult to modifyt his model to account for detailed chemical interactions), [4] predicted mobility shifts based upon it are in reasonable agreement with observed shifts for Rb + ,C s + ,a nd Br À seed ions in the presence of n-butanol. Therefore, despite what is noted in concluding remark 2, measurements here do suggest that for atomic seed ions, the Kelvin-Thomson model at least qualitativelyc aptures features of ion-vapor molecule complex formation. Complimentarye xperiments examining vapor sorptiona nd nucleationu pon theses eed ions under supersaturated conditions [22] will be useful to fully describev apor uptake.