Nature of Cations Critically Affects Water at the Negatively Charged Silica Interface

Understanding the collective behavior of ions at charged surfaces is of paramount importance for geological and electrochemical processes. Ions screen the surface charge, and interfacial fields break the centro-symmetry near the surface, which can be probed using second-order nonlinear spectroscopies. The effect of electrolyte concentration on the nonlinear optical response has been semi-quantitatively explained by mean-field models based on the Poisson–Boltzmann equation. Yet, to explain previously reported ion-specific effects on the spectroscopic response, drastic ion-specific changes in the interfacial properties, including surface acidities and dielectric permittivities, or strong ion adsorption/desorption had to be invoked. Here, we use sum-frequency generation (SFG) spectroscopy to probe the symmetry-breaking of water molecules at a charged silica surface in contact with alkaline metal chloride solutions (LiCl, NaCl, KCl, and CsCl) at various concentrations. We find that the water response varies with the cation: the SFG response is markedly enhanced for LiCl compared to CsCl. We show that within mean-field models, neither specific ion–surface interactions nor a reduced dielectric constant of water near the interface can account for the variation of spectral intensities with cation nature. Molecular dynamics simulations confirm that the decay of the electrochemical potential only weakly depends on the salt type. Instead, the effect of different salts on the optical response is indirect, through the reorganization of the interfacial water: the salt-type-dependent alignment of water directly at the interface can explain the observations.


Anion and polarization effect on integrated SFG intensities
To further explore the effect of polarization combinations, ssp or pss, and anion, NaCl or NaI, we also compare the SFG intensities, integrated from 2920 to 3470 cm -1 . For better comparison, the integrated intensities were normalized to the integrated SFG intensity in the absence of salt.
These data in Figure S1 show that, similar to the data at 3150 cm -1 shown in Figure 2 of the main manuscript, also the variation of the integrated intensity with concentration is rather insensitive to the polarization combination and to the anion.

Comparison of ion force fields
We have tested several different ionic force-fields, the parameters of which are listed in Table S1.
The ion force-fields from Loche et al. are based on a simultaneous optimization of NaCl, NaBr, KCl and KBr for solvation free energy and activity. 1 The Lennard-Jones interactions and realspace Coulomb interactions are truncated after 0.9 nm, with long range Coulomb interaction being handled using particle mesh Ewald summation. 2 All other ion force-fields are based on the NaCl force field from Smith and Dang, 3,4 after which Cs + was optimized to reproduce the solvation free energy and the activity coefficients in bulk water. 5 Li + was optimized to reproduce the solvation free energy and enthalpy. 6 As can be seen from Figure S2, and as concluded in the main manuscript, the total interfacial electric fields are insensitive to the nature of the ions and also to the exact choice of the forcefields. However, the ions' distributions markedly depend on the force-field parameters as evident from the ionic displacement fields shown in Figure S2. These different ion distributions also affect the orientation of water (see water electric field in Figure S2), relevant to the analysis of the SFG results in the main manuscript.
For the choice of the force-field parameters for the simulations shown in the main manuscript, we use reproduction of the experimental solvation free energy, as well as the agreement of the simulated activity coefficients and the experimentally determined activity coefficients as criteria.
For NaCl and KCl, it has been shown 1 that the force-fields reported by Loche et al 1 together with SPC/E water can excellently reproduce these experimental quantities. As we show in Figure S3, this is also true for CsCl using the force-field reported by Fyta (set 9, Table S1). 5 Therefore, all data shown in the main manuscript are based on these force-fields. For LiCl, we use the forcefields reported by Horinek, which were optimized to reproduce the solvation free energy and entropy. 6 As we show in Figure S3, this optimization results in an underestimation of the thermodynamic activity coefficients for both force-fields. Set 5a for LiCl is closer to the experimental values. Hence, we show in the main manuscript data obtained using this force-field.
Yet, also set 5a underestimates thermodynamic activity ( Figure S3b). This underestimation indicates that ion-ion interactions are overestimated, which may also result in an overestimation of the adsorption to the silica interface. As such, the ion distributions for LiCl shown in the main manuscript should be interpreted with caution, and we refrain from including these simulations into our orientational analysis ( Figure 8 of the main manuscript).   where the Lennard-Jones parameter of the interaction between water oxygen and itself is given by OO and of the interaction between the oxygen of the OH group and X is given by OX (where O can be either the silanol or water oxygen). Any OH group that does not fulfil the hydrogen bond criterion with either of the parameter sets is considered non-bonded.
For every category of OH groups, we calculate the volumetric number density profile as a function of the distance to the surface weighted by the cosine of the angle between the OH bond vector and the vector normal to the silica surface. That means that OH vectors pointing straight toward the surface contribute 1, OH vectors oriented parallel do not contribute, and OH vectors pointing straight toward the bulk contribute -1 to the orientation density. Figure 8c in the main text shows the orientation density integrated over the coordinate from the center of the water slab to the center of the silica.

Phase-resolved SFG spectrum of the neat silica-water interface
For reference, we have also determined the phase-resolved SFG spectrum of the bare silicawater interface (in the absence of added salts). In Figure S4 we compare this spectrum to the spectra in the presence of 1 mol/L salts (also shown in Figure