Charge Dependence of Solvent-Mediated Intermolecular Coster

Dec 3, 2010 - Advanced Search .... Since intraionic Coster−Kronig decay channels from the (2s)−1 state are closed for free Na+, Mg2+, ... reminisc...
0 downloads 0 Views 403KB Size
J. Phys. Chem. B 2010, 114, 17057–17061

17057

Charge Dependence of Solvent-Mediated Intermolecular Coster-Kronig Decay Dynamics of Aqueous Ions ¨ hrwall,*,† N. Ottosson,‡ W. Pokapanich,‡ S. Legendre,‡ S. Svensson,‡ and O. Bjo¨rneholm‡ G. O MAX-lab, Lund UniVersity, P.O. Box 118, SE-221 00 Lund, Sweden, and Department of Physics and Astronomy, Uppsala UniVersity, P.O. Box 516, SE-751 20 Uppsala, Sweden ReceiVed: September 19, 2010; ReVised Manuscript ReceiVed: NoVember 16, 2010

The 2s and 2p photoelectron spectra have been measured for Na+, Mg2+, and Al3+ ions in aqueous solution. In all cases, the 2s lines are significantly broader than the 2p features, which is attributed to a shorter lifetime of the respective 2s hole. Since intraionic Coster-Kronig decay channels from the (2s)-1 state are closed for free Na+, Mg2+, and Al3+ ions, this is evidence for an intermolecular Coster-Kronig-like process, reminiscent of intermolecular Coulombic decay (ICD), involving neighboring water solvent molecules. The observed 2s Lorentzian line widths correspond to lifetimes of the (2s)-1 state of 3.1, 1.5, and 0.98 fs for the solvated Na, Mg, and Al ions, respectively. Introduction The degree of association between cation and solvent molecules in a solution can be probed in various ways. One characteristic that conceivably could be affected by chemical interaction between the ion and the solvent, and provide an experimental observable, is the lifetime of a vacancy in the outermost core orbital, which decays primarily via radiationless Auger decay, in which the core hole is filled by one outer electron while another outer electron is emitted. The influence of the surrounding on the valence electrons can affect the rate of core-valence-valence Auger decays, which indeed has been observed in free small molecules.1 An extreme situation occurs when all of the valence electrons of the cation have been donated to the counterion, which in principle would prevent an Auger decay from the outermost core level. Auger-like decays involving orbitals from the surrounding solvent molecules will still be possible, however. An example of such a decay has been found in aqueous solutions of KCl, where Auger signal from final states involving orbitals on water molecules surrounding the K+ ion has been observed after ionization of the L2,3 edges.2 In that case, the core orbital involved was not the outermost one, and normal K LMM Auger still dominated the spectrum. A similar process also occurs after resonant oxygen 1s excitation of OH- in water, leading to very efficient delocalized decays involving surrounding water molecules.3 For pure water, holes in the inner valence O 2s level have recently been shown to undergo Auger-like intermolecular Coulombic decay (ICD).4 In the case of the aqueous ions studied in this paper, Na+, Mg2+, and Al3+, the 3s and 3p orbitals are usually considered empty, and the 2s and 2p levels are the outermost ones. In the corresponding metals, which can be described as consisting of such ions surrounded by delocalized valence electrons in s- and p-bands, a 2p vacancy will predominantly relax by Auger decay. A 2s hole predominantly relaxes by Coster-Kronig decay, which is faster than normal Auger decay, since for Coster-Kronig decay one of the final-state holes is in the same shell as the * To whom correspondence should be addressed. E-mail: gunnar.ohrwall@ maxlab.lu.se. † Lund University. ‡ Uppsala University.

intermediate-state core hole. For the corresponding free ions, these decays are forbidden, because of the lack of available valence electrons to fill the 2p hole and for energy reasons for the 2s vacancy. In the related case of solvated Li+ ions, where the valence 2s electron is missing, a theoretical prediction has been made that 1s holes can decay through a process where one electron from a neighboring water molecule will fill the vacancy, and another will be ejected, originating either from the same or a different water molecule. This process has been termed electron transfer mediated decay (ETMD).5 One can also compare to the situation in solid sodium halides, where it is known that interatomic Auger and Coster-Kronig processes occur after Na 2p and 2s ionization, respectively, and the electrons that take part in the relaxation originate from the neighboring halide ions.6 This could also be expected for crystalline magnesium and aluminum salts, but in a solution, where the interaction between the cation and the solvent molecules is weaker than in an ionic compound, the situation is by no means self-evident. We have therefore made an investigation of the widths of the cation 2s and 2p photoelectron lines from solutions of NaCl, MgCl2, and AlCl3, using a liquid microjet setup, to see whether a difference in lifetime related to the different ion-solvent interaction of the three differently charged ions can be observed. We have also recorded spectra in the kinetic energy region where electrons taking part in intermolecular Auger and Coster-Kronig decays could be expected to appear. Experimental Section The measurements were performed at the undulator beamline I411 at the Swedish National Synchrotron Facility MAX-lab.7,8 Aqueous solutions of sodium chloride, magnesium chloride, and aluminum trichloride were prepared by mixing commercially obtained salts (purities of g99.5%, Sigma Aldrich, g 98%, Fluka, and g99.99%, Sigma Aldrich, respectively) with deionized water. The concentrations were 3.0, 3.0, and 0.5 m (mol/ kg solvent) for NaCl, MgCl2, and AlCl3, respectively. The lower concentration for aluminum chloride was motivated by the corrosive character of the solution, which could harm the equipment used. One should note that, at this concentration, the hydrolysis of the AlCl3 complex is usually considered

10.1021/jp108956v  2010 American Chemical Society Published on Web 12/03/2010

17058

J. Phys. Chem. B, Vol. 114, No. 51, 2010

complete, and Cl- ions are missing in the solvation shell.9 The experimental setup used has been described in detail in ref 10. Briefly, the sample was introduced into vacuum as a jet formed by a nozzle with a diameter of 15 µm, backed by a highperformance liquid chromatography (HPLC) pump. Both the nozzle and the pump were obtained from Microliquids GmBH.11 The jet was injected into a differentially pumped compartment, with a 1 mm diameter hole to allow the synchrotron radiation into the ionization region, and with a skimmer with a 0.5 mm diameter opening toward the electron spectrometer. After passing the soft X-ray radiation, the liquid was collected by a liquid nitrogen-cooled trap. The measurements were performed at a distance of 2-3 mm downstream from the nozzle, at a point before the jet spontaneously breaks into a train of droplets.12,13 The jet propagation axis was perpendicular to the exciting radiation and to the detection axis of the spectrometer. The spectrometer was mounted with the lens axis at an angle of 54.7° relative to the plane of polarization from the undulator radiation, to minimize angular distribution effects.14 Photoelectron spectra were recorded for the 2p and 2s levels using photon energies of 140 eV for Na+ and Mg2+ and 220 eV for Al3+. Electron spectra of the energy regions where electrons from conceivable decay processes would be observed were also recorded (see below). Outer valence photoelectron spectra were recorded at the same photon energies, used for energy calibration of the Mg and Al 2p and 2s features against the X state of liquid water, at 11.16 eV.15 For sodium, the binding energy was calibrated versus the Na+ 2p level, at 35.4 eV, reported by Weber et al.16 This energy is an average for the unresolved spin-orbit doublet, and we have in our calibration procedure assumed a 2:1 intensity ratio of the 2p3/2 and 2p1/2 lines, with a splitting of 0.169 eV.19 The total energy resolution in the experiment, set by spectrometer resolution and photon bandwidth, was estimated to be 0.23 eV for the Na+ and Mg2+ measurements and 0.55 eV for the Al3+ spectra, by the width of the (0,0,0) vibrational line of the X (1b1)-1 state of gas phase water from the vapor that surrounds the sample.

¨ hrwall et al. O

Figure 1. Photoelectron spectra of the 2p (right) and 2s (left) levels from aqueous solutions of NaCl, MgCl2, and AlCl3. The solid lines (black) correspond to the total intensities of the fits, and the dot-dashed lines (blue) represent the residues. The data points are represented by filled circles (red). A background has been subtracted in all spectra (see text). In the 2p spectra, the dashed lines (black) represent the Voigt profiles of the 2p spin-orbit components. Γ2s and G2p refer to the Lorentzian lifetime width of the 2s line and the corresponding Gaussian width obtained from curve fitting. The scale on the vertical axis differs between the spectra.

Results and Discussion In Figure 1, our main experimental results are presented. In the right part, photoelectron spectra of the Na (a), Mg (b), and Al (c) 2p level from aqueous solutions are shown and in the left part, photoelectron spectra of the Na (a), Mg (b), and Al (c) 2s line, recorded in direct sequence to the 2p spectra and under the same conditions. The Na 2p feature overlaps with the strong O 2s feature of water, and to facilitate the comparison, we have subtracted a spectrum of pure water in the presented spectrum. The spectra used in the subtraction procedure are shown in the Supporting Information. A direct comparison shows that the 2s lines are significantly broader than the 2p features, even though the latter are split by the spin-orbit interaction in the final state. We have attempted to determine the lifetime of the 2p and 2s vacancies from the spectral widths, but to analyze the data using curve fitting is not a completely straightforward task. In our below analysis, we have used symmetric Voigt profiles in the fits, together with a linear background. The features have essentially Gaussian broadening contributions from the experimental setup and from inherent sources. For instance, large changes in the ion-to-water equilibrium distance can be expected upon ionization, which will lead to extensive vibrational excitation. Variations in ionization potential from differing coordination for the ions in the solutions will also give a significant contribution to the Gaussian width of the observed lines. These contributions may

well be comparable or larger than the lifetime widths, making an accurate determination of the lifetime difficult. The Gaussian widths of the 2p and 2s features, originating due to the aforementioned reasons, should however be equal, which simplifies the analysis. This is because the experimental resolution is the same for both cases; the inherent contribution to the Gaussian width from the initial state must be independent of which orbital is ionized; and the final-state contribution is also identical, since the external electronic relaxation that occurs after ionization should not differ between the two types of core hole. Another difficulty is that the exact value of the spin-orbit splitting for the 2p levels is unknown for the solvated ions and cannot easily be determined from the spectra. In our fittings, we have fixed the spin-orbit splitting to the values obtained by Citrin et al. for metallic magnesium and aluminum: 0.28 and 0.40 eV, respectively.17,18 These are close to those for free Mg3+ and Al4+ ions: 0.276 and 0.427 eV, respectively.19 For sodium, we have used the literature value for a free Na2+ ion: 0.169 eV.19 It is well-known that the chemical environment has little influence on the spin-orbit splitting in core-level electron spectra, which makes us confident that these assumed values will introduce very small errors in the following analysis. In the fitting procedure, the intensity of the 2p1/2 component was set to half that of the 2p3/2 component.

Coster-Kronig Decay Dynamics of Aqueous Ions

J. Phys. Chem. B, Vol. 114, No. 51, 2010 17059

TABLE 1: Vertical Binding Energies and Full Widths at Half Maximum of the 2p and the 2s Lines of the Cations in Aqueous Solutions of NaCl, MgCl2, and AlCl3a

level

binding energy (eV)

Lorentzian width (eV)

Na 2p3/2 Na 2p1/2 Na 2s Mg 2p3/2 Mg 2p1/2 Mg 2s Al 2p3/2 Al 2p1/2 Al 2s Na 2p3/2 Na 2p1/2 Na 2s Mg 2p3/2 Mg 2p1/2 Mg 2s Al 2p3/2 Al 2p1/2 Al 2s

35.34 35.51 68.19 55.51 55.79 94.33 80.27 80.67 125.15 35.34 35.51 68.19 55.51 55.79 94.33 80.27 80.67 125.15

0 (fixed) 0 (fixed) 0.21 0 (fixed) 0 (fixed) 0.45 0 (fixed) 0 (fixed) 0.67 0.21 0.21 0.43 0.17 0.17 0.66 0.11 0.11 0.75

Γ (fs)

Gaussian width (eV)

deconvoluted width (eV)

3.1 1.5 0.98 3.1 3.1 1.5 3.9 3.9 1.0 6.0 6.0 0.88

1.18 1.18 1.18 1.10 1.10 1.10 1.28 1.28 1.28 1.04 1.04 1.04 0.97 0.97 0.97 1.21 1.21 1.21

1.16 1.16 1.16 1.07 1.07 1.07 1.16 1.16 1.16 1.01 1.01 1.01 0.89 0.89 0.89 1.07 1.07 1.07

a The values in the upper section assume no lifetime broadening of the 2p features, and those in the lower section come from fits with free lifetime widths of the 2p features. The deconvoluted widths were calculated by deconvoluting the contribution of spectrometer and photon bandwidth from the total Gaussian width.

In addition to the Gaussian part of the line profile, the spectral lines have a Lorentzian part related to the finite lifetime of the final state. For the free case, the (2p)-1 2P3/2 state is the ground state for the final state Na2+, Mg3+, and Al4+ ions, which has an infinite lifetime. For the (2p)-1 2P1/2 state, the radiative lifetime width is minute in comparison with the experimental broadening in our experiment. The Lorentzian lifetime contribution would thus be completely negligible for ionization of the 2p level in free Na+, Mg2+, and Al3+ ions, which is what we have assumed to apply for the solvated ions in the fits presented in Figure 1. Under this assumption, the Gaussian widths in the fitting of the 2p spectra will be maximized. Using the same Gaussian widths when fitting the 2s spectra will then give a minimum for the lifetime width of the 2s vacancies. From such fits, we obtain Lorentzian widths Γ2s of 0.21, 0.45, and 0.67 eV for the 2s lines of aqueous sodium, magnesium, and aluminum ions, respectively. These lifetime widths show the same trend and are actually quite similar to those found in the metallic cases, 0.28, 0.46, and 0.78 eV, respectively.17,18 In Table 1, the data obtained from the curve fits are collected. The reason for the larger lifetime width of the 2s compared to the 2p level in the metals is the fast Coster-Kronig decay of the 2s hole, involving a 2p electron filling the 2s vacancy and a 3s or 3p electron being ejected, on a time scale in the femtosecond range. In free ions, this decay would be impossible, as there are no 3s or 3p electrons to carry away the excess energy. A super Coster-Kronig decay, involving two 2p electrons, one filling the 2s core hole and one being ejected, is impossible for energetic reasons in the metals as well as for the free or solvated ions. The most obvious explanation of the Lorentzian width of the 2s lines for the solvated ions is that an intermolecular Coster-Kronig process takes place, where instead of the 3s or 3p electron, an electron from the surrounding water molecules or an associated anion is emitted after the 2s vacancy has been filled by an electron from the 2p shell. This process is similar to ICD, in which a vacancy in an inner valence shell of an atom or molecule in a weakly bound complex is

filled by an outer valence electron, while a valence electron from a neighboring site in the complex is emitted into the continuum.20 In atomic Na, Mg, and Al, the radiative contribution to the 2s lifetime is very small.21 Could the radiative lifetime of a 2s vacancy possibly be shorter for the solvated ions, giving a significant contribution to the observed Lorentzian width? The transition probability of X-ray emission after the creation of a 2s vacancy will most likely not be affected much by the loss of the 3s or 3p electrons in the respective ions, as it is the 2p electrons that are dominantly taking part in such transitions, and while the contraction of the orbitals of the ionized species could possibly increase the efficiency of X-ray emission, the contribution to the 2s lifetime due to radiative decay can still be expected to be negligible. For instance, the experimentally determined mean life for the (2s)2(2p)5 2P-(2s)(2p)6 2S transition in free Na III is 0.15 ns,22 and in free Mg IV it is 0.05 ns.23 We have not been able to find any experimental value for this transition for free Al V ions, but calculated lifetimes of the 2s22p5 2P-2s2p5 2S transition for Na2+, Mg3+, and Al4+ ions are 88.5, 64.1, and 49.5 ps.24 These lifetimes correspond to widths of less than 15 µeV and will thus have no bearing on the present discussion. The observed widths correspond to lifetimes of the (2s)-1 inner valence state of 3.1, 1.5, and 0.98 fs for Na+, Mg2+, and Al3+, respectively. The decrease of the lifetime from Na+ to Al3+ correlates well with the increased ion-water interaction. This interaction is dominated by the first solvation shell, which contains approximately the same amount of water molecules (5.4 for Na+, 6 for Mg2+ and Al3+),25 but the ion-water distance decreases with increasing ionic charge (2.4 Å for Na+, 2.1 Å for Mg2+, and 1.9 Å for Al3+).25 First of all, a smaller ion-water distance means a larger overlap between the ion and the water orbitals, which would facilitate this ICD-like Coster-Kronig process. Second, the solvent polarization induced by the positive ions, where the increased charge in the core-ionized intermediate state should be considered, increases the electron density in the vicinity of the ion with the core hole. Since the degree of water polarization in the first solvation shell is very much dependent on the charge of the ion, the discussed ICD-like decay processes should be more efficient for the more highly charged ions, just as observed. These polarization effects, and particularly for the case of Al3+, the possibility of hybridization between the outer water valence orbitals and those of the ions, are thus certainly factors that determine the decay transition rates. To make quantitative estimates of the transition rates to gain insight into these phenomena would most likely require a very complex and most demanding theoretical treatment, but we strongly encourage such efforts. In solid sodium halides, interatomic Coster-Kronig processes have been observed to occur after Na 2s ionization, and the estimated lifetime widths for these processes are 0.75 ( 0.15, 0.45 ( 0.10, and 0.46 ( 0.10 eV for NaF, NaCl, and NaBr, respectively,6 that is, of the same order of magnitude as what we observe here, even though the nature of the interaction with the neighboring sites is quite different. We note that the cation-anion distances in these sodium halide crystals (2.3, 2.8, and 3.0 Å for NaF, NaCl, and NaBr, respectively) are similar or larger than the average distance between the Na+ ion and the oxygen of water in an aqueous solution. The 2p holes in the aqueous ions could conceivably decay by an ETMD process, in which case the above made assumption of a zero Lorentzian width of the 2p feature would not be valid. In this case, both the electron filling the vacancy and the emitted

17060

J. Phys. Chem. B, Vol. 114, No. 51, 2010

electron would have to come from surrounding water molecules or associated anions. As mentioned in the Introduction, this type of decay has theoretically been predicted to occur for 1s ionization of solvated Li+ ions.5 The lifetime for a Li 1s vacancy was estimated to lie in the range 20-100 fs, depending on degree of coordination with water molecules.5 The widths associated with such lifetimes are negligible compared with the total width we observe for the solvated ions. However, in solid sodium fluoride, the lifetime width of a Na 2p vacancy has been observed to be 0.1 eV, which has been attributed to a double interatomic Auger process.6 Even though our data do not allow accurate determination of the lifetime widths, due to the large inherent widths of the features, we have also fitted the 2p spectra with the Lorentzian width in the Voigt profile as a free parameter, to account for possible contributions from ETMDlike processes. The Gaussian width obtained from such fits will instead of maximal be minimal, and then an upper limit of the Lorentzian width of the 2s line can be obtained. From such fits we obtain the Lorentzian widths Γ2p of 0.21, 0.18, and 0.12 eV for sodium, magnesium, and aluminum, respectively, and Γ2s of 0.43, 0.66, and 0.75 eV (see Table 1). Clearly, the assumption that the (2p)-1 state has a significant Lorentzian width affects the estimation of the 2s lifetime, but the obtained values are still in the same range as in the metallic cases. The decreasing trend of Γ2p with increasing ionic charge is interesting, but the uncertainties in the Lorentzian widths Γ2p are such that no definite conclusions can be drawn from these values. The kinetic energy of the electrons emitted in the intermolecular Coster-Kronig decay of a 2s vacancy can be roughly estimated from the binding energies of the 2s, 2p, and water or Cl- valence electrons as Eb(2s)-Eb(2p)-Eb(valence)-V, where Eb are the binding energies of the involved levels and V is the repulsion between the holes in the final state. The latter contribution is difficult to estimate, as the Coulomb repulsion between charges will be screened by the polarized neighboring molecules. We estimated the kinetic energy ranges of 0-24 eV for Na, 0-30 eV for Mg, and 0-36 eV for Al, using the binding energies from Table 1, and the span 9-20 eV for the valence ionization energy of the water molecules and Cl- ions. For the repulsive potential V, we used the Coulomb repulsion calculated for the mean distances between the ions and the oxygen of water25 as the maximum and zero as the minimum. We have recorded electron spectra in these ranges but were not able to observe any structure from final states with vacancies in the 2p shell and a neighboring water molecule or chloride ion, for any of the solutions. At low kinetic energies it is however very difficult to make reliable observations, because of the presence of a strong background of inelastically scattered electrons and to the often varying transmission of the electron spectrometer. Furthermore, the 2s photoelectron line has a low cross section, and the oscillator strength of this level will after the decay be spread out over many different final states, which would make such features difficult to observe even under more fortunate circumstances. An experiment where both the 2s photoelectron and the electron emitted in the decay are detected in coincidence would improve the signal-to-noise ratio, and this was for instance the approach used in the recent work which showed that inner-valence O 2s vacancies undergo ICD.4 Such experiments would be exceedingly difficult to perform, and at present we do not have the capability to perform them, but they would be a challenge for future research. Using the same reasoning as for the 2s decay discussed above for a possible ETMD process after 2p ionization, we have also estimated the range of kinetic energies for electrons emitted in

¨ hrwall et al. O such a process for the three ions: 0-18 eV for Na, 0-38 eV for Mg, and 0-62 eV for Al. Here the contribution from the Coulomb repulsion is even more difficult to estimate than for the 2s decay, as there may be two or three charges involved, which all will be screened by surrounding molecules, but we again used the unscreened Coulomb potential as a maximum value. We used published Auger spectra of liquid water26 to estimate the double ionization potentials of the water molecule for the case when the electrons in the decay originate from the same molecule. We have recorded electron spectra covering these energy ranges but have not been able to discern any structure that can be assigned to such a decay. However, as for the 2s level, the oscillator strength of the 2p photoelectron line would be distributed over a wide range of final states, so the statistics may simply not have been sufficient for these states to be observable on top of the pronounced inelastic scattering background which inevitably will be present in the spectrum. If we compare to the case of K 2p and Cl 2p from aqueous KCl,2 the dominance of the normal intra-atomic Auger decay for the K 2p and Cl 2p core holes strongly suggests that the lifetimes of these core holes are similar to those found in free potassium or chlorine. The lifetime width associated with this is much smaller than the inherent width of the core-level photoelectron spectra, which prohibits its determination with accuracy. In the fitting procedure used in ref 2, we have instead relied on published data for the lifetime. The situation presented here is different: the only decay path, excluding radiative decay, which is too slow to explain the measured difference in width between the 2s and 2p levels, must involve electrons from the solvent or counterion. Effects of the ions on the hydrogen-bonded network around the ion have been observed in X-ray absorption spectra.27,28 For instance, it has been suggested that the presence of Al3+ leads to an increase in strongly tetrahedral structures around the ion, in contrast to alkali ions, where instead a weakening of the hydrogen bonded network occurs.27 This change in the interaction could very well affect the rate of the Auger decay, which would be an exciting issue to investigate further, both theoretically and experimentally, but which we only could speculate about based on the present data alone. To conclude, we have found that, for photoelectron spectra from the 2s and 2p levels of aqueous solutions of NaCl, MgCl2, and AlCl3, the 2s line is broader than the 2p feature for all of the ions. The only conceivable explanation of this observation is that the lifetime of the 2s level is significantly shorter than for the 2p level. The Coster-Kronig decay channels are closed for the corresponding free ions, meaning that the solvated ions must interact so strongly with the coordinated water molecules to make an intermolecular Coster-Kronig process possible. This must occur on the same time scale as in the solid metal, that is, on a femtosecond or even a subfemtosecond time scale. The observed decrease of the lifetime from Na+ to Al3+ correlates well with the increased ion-water interaction due to the decreased size of the first solvation shell. We have not been able to directly observe the electrons emitted in this process, but this study motivates further future measurements of other ions strongly associated with water, where these electrons hopefully can be directly detected. Acknowledgment. The Royal Thai Government and Nakhon Phanom University are gratefully acknowledged for the graduate fellowship of W.P. This work has been financially supported by the Swedish Research Council (VR), the Go¨ran Gustafsson Foundation, the Knut and Alice Wallenberg Foundation, the Foundation for International Cooperation in Research and Higher

Coster-Kronig Decay Dynamics of Aqueous Ions Education (STINT), the Foundation for Strategic Research (SSF), and the Carl Tryggers Foundation. The authors also thank the MAX-lab staff for their helpful assistance during the experiments. Supporting Information Available: Photoelectron spectra in the O 2s, Na 2p, and Na 2s levels for a 3.0 m solution of NaCl and for pure water, recorded under the same conditions. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Coville, M.; Thomas, T. D. Phys. ReV. A 1991, 43, 6053. (2) Pokapanich, W.; Bergersen, H.; Bradeanu, I. L.; Marinho, R. R. T.; Lindblad, A.; Legendre, S.; Rosso, A.; Svensson, S.; Bjo¨rneholm, O.; ¨ hrwall, G.; Kryzhevoi, N. V.; Cederbaum, L. S. J. Am. Tchaplyguine, M.; O Chem. Soc. 2009, 131, 7264. (3) Aziz, E. F.; Ottosson, N.; Faubel, M.; Hertel, I. V.; Winter, B. Nature 2008, 455, 89. (4) Mucke, M.; Braune, M.; Barth, S.; Fo¨rstel, M.; Lischke, T.; Ulrich, V.; Arion, T.; Becker, U.; Bradshaw, A.; Hergenhahn, U. Nat. Phys. 2010, 6, 143. (5) Mu¨ller, I. B.; Cederbaum, L. S. J. Chem. Phys. 2005, 122, 194305. (6) Wertheim, G. K.; Rowe, J. E.; Buchanan, D. N. E.; Citrin, P. H. Phys. ReV. B 1995, 51, 13669. (7) Ba¨ssler, M.; Forsell, J. O.; Bjo¨rneholm, O.; Feifel, R.; Jurvansuu, M.; Aksela, S.; Sundin, S.; Sorensen, S. L.; Nyholm, R.; Ausmees, A.; Svensson, S. J. Electron Spectrosc. Relat. Phenom. 1999, 101-103, 953. (8) Ba¨ssler, M.; Ausmees, A.; Jurvansuu, M.; Feifel, R.; Forsell, J. O.; Fonseca, P. T.; Kivima¨ki, A.; Sundin, S.; Sorensen, S. L.; Nyholm, R.; Bjo¨rneholm, O.; Aksela, S.; Svensson, S. Nucl. Instrum. Methods Phys. Res., Sect. A 2001, 469, 382. (9) Na¨slund, L.-Å.; Cavalleri, M.; Ogasawara, H.; Nilsson, A.; Pettersson, L. G. M.; Wernet, P.; Edwards, D. C.; Sandstro¨m, M.; Myneni, S. J. Phys. Chem. A 2003, 107, 6869.

J. Phys. Chem. B, Vol. 114, No. 51, 2010 17061 (10) Bergersen, H.; Marinho, R. R. T.; Pokapanich, W.; Lindblad, A.; ¨ hrwall, G. J. Phys.: Condens. Matter 2007, Bjo¨rneholm, O.; Sæthre, L. J.; O 19, 326101. (11) Microliquids GmbH. http://www.microliquids.com (accessed April 29, 2010). (12) McCarthy, M. J.; Molloy, N. A. Chem. Eng. 1974, 7, 1. (13) Heinzl, J.; Hertz, C. H. AdV. Electron. Electron Phys. 1985, 65, 91. (14) Cooper, J.; Zare, R. N. J. Chem. Phys. 1968, 48, 942. (15) Winter, B.; Weber, R.; Widdra, W.; Dittmar, M.; Faubel, M.; Hertel, I. V. J. Phys. Chem. A 2004, 108, 2625. (16) Weber, R.; Winter, B.; Schmidt, P. M.; Widdra, W.; Hertel, I. V.; Dittmar, M.; Faubel, M. J. Phys. Chem. B 2004, 108, 4729. (17) Citrin, P. H.; Wertheim, G. K.; Baer, Y. Phys. ReV. Lett. 1975, 35, 885. (18) Citrin, P. H.; Wertheim, G. K.; Baer, Y. Phys. ReV. B 1977, 16, 4256. (19) Ralchenko, Y.; Kramida, A. E.; Reader, J. NIST ASD Team. NIST Atomic Spectra Database (Version 3.1.5); National Institute of Standards and Technology: Gaithersburg, MD, 2008. (20) Cederbaum, L. S.; Zobeley, J.; Tarantelli, F. Phys. ReV. Lett. 1997, 79, 4778. (21) Keski-Rahkonen, O.; Krause, M. O. At. Data Nucl. Data Tables 1974, 14, 139. (22) Buchet, J. P.; Buchet-Poulizac, M. C. Phys. Lett. A 1973, 46, 273. (23) Buchet, J. P.; Buchet-Poulizac, M. C.; Ceyzeriat, P. Phys. Lett. A 1980, 77, 424. (24) Sinanoglu, O. Nucl. Instrum. Methods 1973, 110, 193. (25) Marcus, Y. Chem. ReV. 2009, 109, 1346. (26) Winter, B.; Hergenhahn, U.; Faubel, M.; Bjo¨rneholm, O.; Hertel, I. V. J. Chem. Phys. 2007, 127, 094501. (27) Na¨slund, L.-Å.; Edwards, D. C.; Wernet, P.; Bergmann, U.; Ogasawara, H.; Pettersson, L. G. M.; Myneni, S.; Nilsson, A. J. Phys. Chem. A 2005, 109, 5995. (28) Cappa, C. D.; Smith, J. D.; Wilson, K. R.; Messer, B. M.; Gilles, M. K.; Cohen, R. C.; Saykally, R. J. J. Phys. Chem. B 2005, 109, 7046.

JP108956V