Resonant Auger and ICD spectraWe aim to study the first solvation shell of the Ca2+ ion in water. Electronic decay processes in a 1.5 M CaCl2 aqueous solution are studied by coupling liquid-microjet electron spectroscopy with synchrotron radiation as a tuneable source of soft X-rays36, as detailed in the “Methods” section. The initial step is the resonant excitation of Ca 2p electrons to unoccupied Ca 3d orbitals. The corresponding absorption spectrum at the Ca 2p edge is shown in Fig. 2a. The several features of the spectrum are discussed in the literature37,38 and arise from the reduced symmetry of the hydrated ion compared to the isolated one. Only the two most intense peaks will be probed here.Fig. 2: Resonant absorption and decay of solvated Ca2+.a X-ray absorption spectrum (XAS) at the Ca 2p edge of a 1.5 M CaCl2 solution. The spectrum is recorded in partial-electron-yield mode by integrating electron counts over the 270–350 eV range. IP ionization potential. b Resonant Ca 2p Auger and ICD spectra of the same solution. The two resonant spectra (red and blue traces) correspond to the excitation energies of the two main peaks of the panel (a) (red and blue arrows). The non-resonant Auger/ICD spectrum at 460 eV is also shown (black trace) for comparison. The most important attributions are made in the figure, with dashed black lines following the resonant features as they shift with photon energy. The spectra have been scaled and offset on the intensity scale to improve visibility.In Fig. 2b, electron spectra obtained at these two resonance peaks, in the region where Auger decay and ICD manifest, are shown. A non-resonant Auger and ICD spectrum above the 2p threshold, in good agreement with spectra already reported in the literature39, is also displayed for comparison. In the latter, normal ICD appears at kinetic energy ~20 eV higher than normal Auger features, roughly reflecting the energy difference between the binding energies of the Ca 3p shell and the water valence shell. While resonant spectra have not yet been reported for solvated CaCl2, a comparison can be made with reported 2p resonant spectra of atomic Ca40 and solid CaCl2 films41. In the resonant spectra in Fig. 2, the features in the 280–295 eV range can be assigned to spectator Auger (at similar but slightly blue-shifted energies compared to the normal Auger). Similarly, the feature around 305–310 eV, at slightly blue-shifted energies compared to the normal ICD, is assigned to spectator ICD. Features are also observed in the resonant spectra of atomic Ca and solid CaCl2 in this energy range; there, they correspond to Auger processes in which a Ca 3p electron and a valence electron of the system are involved. This is quite similar to what we observe here, except that the valence electron must be borrowed from a neighbouring water molecule without any chemical bond to calcium, which is characteristic of ICD.Participator ICD (Fig. 1) should manifest itself as an on-resonance enhancement of the valence-band features of water. Indeed, as long as there is no significant nuclear dynamics in the intermediate state, participator ICD is energetically equivalent to direct ionization of the valence-band features, much like participator Auger is energetically equivalent to direct ionization of the outermost orbitals of Ca, and is thus identified by an on-resonance enhancement of the Ca 3s and 3p photoelectron lines in Fig. 2. The on-resonance enhancement of the water valence band is not obvious in Fig. 2; it will become more evident in another figure shown below. From the integration of the peak areas of the different features in the spectra, one can estimate that spectator ICD represents ~4% of all decay processes, while participator ICD represents <0.5%.Spectator ICD: probing the nature of the solvation shellIn order to investigate the sensitivity of resonant ICD to ion pairing, we systematically varied the anion with which calcium is paired in solution (Cl−, I−, ClO−, or \({{{{\rm{NO}}}}}_{3}^{-}\)), as well as, for a CaCl2 solution, the solution concentration. The results of these studies are shown in Fig. 3. We focus here on the spectator ICD feature. All spectra are normalized to the Ca 3s peak. While the main group of broad features in the 305–310 eV kinetic energy (KE) range corresponds to spectator ICD with hydration-shell water molecules, in several cases, one can observe the presence of an additional high-kinetic-energy shoulder (~312 eV KE) in the spectrum. This additional feature can be interpreted as ICD with the counter-anion. Indeed, most anions in solution have the highest occupied molecular orbital (HOMO), which is lower in binding energy than the valence band of water42. The water HOMO, the 1b1, is found at 11.33 eV binding energy43, while, for instance, Cl− 3p is measured at 9.8 eV binding energy in the 1.5 M CaCl2 solution. A lower HOMO binding energy translates into a higher KE of the ICD electron, as can be understood from the energy balance of the process: \({E}_{K}=h\nu -{E}_{B}^{partner}-{E}_{exc}-{E}_{R}\), where EK is the kinetic energy of the ICD electron, \({E}_{{\rm {B}}}^{{\rm {partner}}}\) the binding energy of the orbital of the ICD partner (water or anion), Eexc is the energy required to excite a 3p electron to 3d in Ca2+ and ER is an (unknown) repulsive term. The ICD spectrum thus reflects the valence electronic structure of the ICD partners.Fig. 3: Observation of a signature for ion pairing in the spectator ICD spectra at 349.4 eV (main 2p3/2 → 3d line, see Fig. 2) for several calcium salt solutions.In the top panel, CaCl2 solutions of different concentrations are explored. Concentrations lower than 1.5 M yield spectra similar to the 1.5 M one (see SI). In the bottom panel, 1.5 M Ca2+ solutions with different counter-ions are displayed. Details on the solution preparation [explaining, for instance, the Na+ contamination in the Ca(ClO)2 solution] are available in the “Methods” section.The occurrence of ICD with the counter-anion as a partner implies that this anion is present within the first coordination shell of the cation, indicating contact ion pairing. For instance, in the top panel of Fig. 3, we observe the appearance of the ICD feature with Cl− for a 4 M concentration, and it is even more pronounced at 5 M. In the literature, X-ray diffraction44 and EXAFS45 studies indeed found an onset of contact ion pairing around 4 M, in contradiction with a neutron-diffraction study46 which found no evidence of ion pairing up to the solubility limit of 6.4 M. Our results are in agreement with the first two studies, supporting their conclusion of an onset of ion pairing around 4 M for CaCl2 solutions.In the bottom panel, we kept the Ca2+ concentration constant at 1.5 M but varied the counter-ion. At this concentration, only for \({{{{\rm{NO}}}}}_{3}^{-}\) we observe a trace of counter-ion ICD and thus of ion pairing. This is again in agreement with the literature, where ion pairs between calcium and nitrate were reported47. No data could be found on pairing between calcium and iodide or hypochlorite ions, and our results suggest that no contact ion pairing occurs at 1.5 M.While all types of ICD and other related processes are, in principle, sensitive to ion pairing, spectator ICD is much easier to detect due to its resonant nature. In the present case, it is possible to obtain information from the normal ICD spectra, as we show in the SI, but at the expense of several difficulties, which are also listed in the SI.Participator ICD: probing the electronic structure of the solvation shellWe now turn our attention to the second ICD process: participator ICD. As mentioned above, this process should manifest itself in the valence-band region as an enhancement of the photoelectron lines of the ICD partners, because the final state is the same as for direct photoemission from the valence orbitals. In Fig. 4, we show the valence-band region of water, where the three outer-valence orbitals of water, 1b1, 3a1, and 1b2, as well as the Cl 3s and 3p orbitals, are observed. In condensed-phase water, the 3a1 peak is split into two components, 3a1L and 3a1H, due to intermolecular interactions48,49,50. We display this valence region for on-resonance excitation at 349.4 eV and below the Ca 2p resonance region at 346 eV. The spectra are intensity-normalized at the Cl− peaks. In principle, as outlined in Fig. 1, participator ICD with Cl− counter-ions is also possible. However, we discussed above that no ion pairs are formed in 1.5 M CaCl2 solutions, and therefore, no Cl− ICD is detected in the spectator ICD feature. We can, therefore, also assume the absence of Cl− ICD in the valence/participator ICD signal, and normalize these spectra to the Cl− peak. One can, in this way, clearly observe an on-resonance excess of intensity on the 1b1 and 3a1 peaks, which cannot be explained by the very small variations of relative photoionization cross sections and, therefore, is the signature of participator ICD.Fig. 4: Determination of the valence-band spectrum of water molecules in the solvation shell of Ca2+.In the top panel, the valence-band region of the electron spectra off-resonance (at 346 eV) and on the main 2p3/2 → 3d resonance (at 349.4 eV) for the CaCl2 1.5 M solution are displayed. The valence band spectrum of pure water is also displayed for comparison. The different orbitals are attributed in the figure. The differences are attributed to the resonant participator ICD. The two spectra have been normalized so that the Cl 3p and 3s peaks overlap, as Cl− is not involved in participator ICD. In the bottom panel, the difference between the two spectra, corresponding to the participator ICD contribution only, is displayed along with a fit of the different peaks. This spectrum reflects the valence band of the Ca2+ solvation-shell water molecules. A contamination by Ca 2p photoionization with second-order light is notable in the 346 eV spectrum.The bottom panel of Fig. 4 shows the difference between the on- and off-resonance spectra, thus displaying only the contribution of participator ICD. In this spectrum, the different orbitals of water are easily distinguished, which is not the case in the spectator ICD spectrum and neither was possible in previous non-resonant (normal) ICD studies20. With spectator or normal ICD, the final state also features the excited or ionized neighbouring cation, and the electronic structure of the ICD partner is not revealed as clearly, although it is still underlying the ICD spectrum.In the participator ICD spectrum, we observe relative peak intensities quite different from the direct photoionization spectrum, as well as different relative and absolute peak positions. In addition, fitting the spectrum also reveals some differences in the peak widths, although this parameter was found to be sensitive to the subtraction procedure. The sensitivity of the different parameters to the subtraction method is discussed in the SI. These differences can all be rationalized by qualitative arguments, while comparison to theoretical calculations will be the subject of future work. The intensity of the different features is related to the factors that influence the efficiency of the ICD process (and not just the ionization cross sections, as in regular photoionization). On the other hand, their positions and widths directly reflect the electronic structure of the hydration-shell molecules, something that could not be accessed experimentally before. This point was already highlighted in the calculated results of ref. 20. We will discuss both aspects below.From the relative intensities, we can deduce that participator ICD is more efficient for the 3a1 orbital than for the 1b1 orbital, and it is very weak for the 1b2 orbital. A propensity of non-local decay from the 3a1 orbital has also been seen in previous works on non-resonant ICD20 and ETMD (electron-transfer-mediated decay)9 in solution. As reported here, the ICD efficiency is sensitive to intermolecular distance and, in a regime of close proximity as may be expected for solvation shells (in contrast to, e.g., weakly bound van der Waals dimers), electronic densities overlap. The 1b2 orbital is a bonding orbital with its electronic density mostly localized on the O–H bonds of the water molecule, and thus it is expected that ICD is not very efficient for this channel compared to the mostly non-bonding 3a1 and 1b1 orbitals, with considerable electronic density on the oxygen lone pairs. Indeed, in the solvation shell of a cation like Ca2+, water molecules are oriented with the water dipole facing away from the cation, and thus the oxygen lone pairs towards it. The relative efficiency of ICD of the 3a1 and 1b1 orbitals should depend on their respective orbital overlap with the Ca2+ orbitals. Since these two orbitals are orthogonal, we make the hypothesis that this relative efficiency is directly governed by the distribution of dihedral angles of the water molecules in the solvation shell. The dihedral angle is the angle formed by the Ca–O segment and the plane containing the water molecule. It is the second parameter that, together with the Ca–O distance, fully determines the geometric structure of the hydration shell. An angle of 0° will favour 3a1 orbital overlap, while an angle of 90° will favour 1b1 orbital overlap. The average angle for solvated Ca2+ as deduced from neutron diffraction, is around 34°51. This argument is in the spirit of a recent work on solvated Cu2+, which showed the sensitivity of the Cu 2p photoelectron satellites to the dihedral angle52. This is an example of the fact that the ICD spectrum also contains information on the geometric structure of the hydration shell.As mentioned above, the peak positions and widths of the participator ICD spectrum reflect the electronic structure of the Ca2+ solvation-shell molecules. The decisive question is to what extent this electronic structure differs from that of the average water molecule in the bulk solution, as probed by normal photoelectron spectroscopy. In Table 1, we collected the positions and widths for the different components of the water valence band, measured on pure water and 1.5 M CaCl2 (off-resonance) by photoelectron spectroscopy, and those deduced from the participator ICD spectrum. Absolute positions were determined from the cut-off method outlined in ref. 43 (see also the “Methods” section). Our measurement for pure water is in good agreement with the literature values42,43.Table 1 Peak positions and widths of the valence orbitals of waterThe third row of the table corresponds to normal valence-band photoemission measurements of the 1.5 M CaCl2 solution, i.e., the average water electronic structure in that solution. Only modest changes occur compared to pure water: peaks are blue-shifted on the order of 200 meV, the 1b1 peak is broadened, and the 3a1 peak splitting is reduced. This is in good agreement with previous works on the effect of electrolytes on the electronic structure of water, studied for NaI solutions21,53 over a wide range of concentrations (0.5–8 M).Significant differences are observed in the case of the participator ICD spectrum (fourth row in Table 1). The peak widths are close to the value for pure water within the precision with which they are constrained (see SI). The peak positions, on the other hand, clearly indicate a large blue-shift of about 500 meV of the 1b1 peak and the 3a1L peak, while the 3a1H peak is only slightly shifted, thus making the 3a1 split much smaller (1.1 eV instead of 1.4 eV in pure water). A blue-shift of this magnitude agrees well with the theoretical predictions made for solvated Mg2+ in ref. 20 and, as pointed out in that reference, is the expected behaviour for cation hydration. Interpretation of the 1:0.5 intensity ratio of the 3a1H to 3a1L components will require more theoretical input, as in that case, electronic structure and ICD-efficiency effects are intermingled, but it also indicates significant changes in the hydration shell. We can, therefore, conclude that the electronic structure of the hydration-shell water molecules is markedly different from that of the bulk water molecules. These changes can be directly accessed here, which was not possible before.Summary and conclusionsWe have provided evidence for resonant ICD after core-level excitation of a solvated cation, Ca2+, using liquid-jet photoemission spectroscopy. Resonant ICD has been little explored in liquid systems so far. Two distinct resonant ICD processes are identified here, which we term “spectator” and “participator” ICD, in full analogy to the corresponding resonant Auger processes. Previously, the participator process was only observed in model rare-gas cluster systems and in cases where it is the only expected electron-emission decay pathway, in competition with fluorescence32,34. In contrast, we observe here both spectator and participator processes, and both are in competition with the local resonant Auger decay pathways. Spectator ICD represents 4% of the total relaxation pathways, and participator ICD <0.5%.Aside from this step forward in the fundamental understanding of ICD in liquids, we showed how these processes constitute valuable additions to the liquid-phase photoemission toolbox, providing probes of the solvation shell of the core-excited solutes. While all ICD variants contain, in principle, the same information about the nature, geometry, and electronic structure of the solvation shell, this information is easier to extract from certain processes. Signatures of the presence or absence of ion pairing between Ca2+ and its anion partners are provided by spectator ICD. The resonant character of the process makes this signature easy to measure and unambiguous, providing a useful alternative to normal ICD and other ICD-like processes used thus far for tracking ion pairing18. Although participator ICD is also a resonant process, it is weaker and necessarily has to be detected on top of the regular valence photoemission features, making it less suitable for this purpose.On the other hand, participator ICD provides the even more appealing and novel possibility of directly probing the valence electronic structure of the water molecules in the solvation shell, which we show in the case of Ca2+ to differ from that of the average water electronic structure as probed by direct photoelectron spectroscopy. Important quantities, such as the electron binding energies of these solvation-shell molecules, now become directly measurable. Inferences can also be made on the geometric structure of the water molecules (dihedral angle) from the relative involvement of different water orbitals in the participator ICD process.
