Low-Frequency Modes of Aqueous Alkali Halide Solutions: Glimpsing the Hydrogen Bonding Vibration

See allHide authors and affiliations

Science  12 Feb 2010:
Vol. 327, Issue 5967, pp. 857-860
DOI: 10.1126/science.1183799

Salty Stretch

What happens at the molecular level when salt dissolves in water? Much of the data characterizing the geometry and dynamics of ion solvation shells has come from indirect observation of the surrounding water structure. Using a time domain Raman technique based on the interference of four ultrashort polarized light pulses, Heisler and Meech (p. 857) have now mapped directly the stretching vibrations associated with the weak hydrogen bonding interactions between bulk water molecules and chloride, bromide, or iodide ions.


The solvation of ions in aqueous media is a fundamental process in biology and chemistry. Here, we report direct time-domain observations of the hydrogen bond vibrational mode formed between a halide ion (chloride, bromide, or iodide) and the surrounding water molecules. The frequency of the hydrogen bond mode is sensitive to both the atomic weight and the concentration of the ion. The peak frequencies fall in the 125 to 175 wave-number range, a spectral region accessed through time-domain polarization-resolved coherent Raman scattering using a diffractive optic method. The polarized Raman response observed is discussed in terms of the structure of the anion’s solvation shell and modeled through calculations on water chloride clusters.

Aqueous solvation of ions is a central process in many chemical and biological reactions. It plays a critical role in determining acid-base equilibria (1), interface structure (2), and ion transport in electrolyte solutions (3) and across membranes (4). Aqueous solutions of alkali halide salts present the archetypal example of solvation and have consequently been studied in great detail through both experiment and simulation (5). Important progress in understanding the microscopic structure and dynamics of solvated ions has recently been made through ultrafast vibrational spectroscopy (69) and molecular dynamics simulation (1014). Most of these experiments focused on the readily observable infrared (IR)–allowed OD stretch of monodeuterated water (HOD) in aqueous solutions of alkali halides. The vibrational relaxation and molecular orientational relaxation times observed are significantly slower compared with the same measurements for HOD in pure H2O and become slower still with increasing concentration of the electrolyte (7, 9). Two-dimensional IR spectroscopy on the OD stretch reveals further information on the molecular dynamics, suggesting a hierarchy of relaxation times. The fastest component arises from local fluctuations in the length of the OD⋯X hydrogen bond (H-bond), where X is the anion. The intermediate (subpicosecond) and longer (picosecond) relaxation components, which increase with increasing halide ion concentration, reflect the transition from local H-bond fluctuations to global reorganization and the reorganization of H-bond networks, respectively (9). From these studies of the OD mode, inferences on the structure and dynamics of the OD⋯X H-bonds are drawn. However, the H-bond mode itself is not well characterized, and a more complete description of it will help to refine the analysis and simulation of aqueous solvation dynamics.

Here, we report direct time-domain experimental observations of the spectra of the OH⋯X hydrogen bond through measurements of the low-frequency isotropic Raman response. To access the low-frequency region of the spectrum, where H-bond modes are expected to appear, we probed in the time domain the decay of polarization induced in the sample by an ultrafast pump pulse (15). In a two-beam pump-probe configuration, this experiment measures the transient optical Kerr effect, which, when combined with heterodyne detection and Fourier transform analysis, yields the Raman spectral density with excellent signal-to-noise ratio in the 0- to 500-cm−1 region (16). In the present experiments, we exploited a four-beam transient grating diffractive optic geometry, described in detail by Goodno et al. (17). Essentially, pump and probe beams are focused onto a diffractive optic element to create two identical pairs of pulses that are brought to a common focus in the sample. The pump-pulse pair generates a transient polarization grating in the sample from which one of a pair of time-delayed probe pulses scatters a signal in a direction that is temporally and spatially overlapped with the second of the probe pulses. Constructive or destructive interference between this third-order nonlinear optical signal and the transmitted probe pulse yields in-phase and out-of-phase components, respectively, with high sensitivity. This geometry has several advantages over the two-beam geometry, and here we exploited in particular its ability to measure the polarizability relaxation with arbitrary pump-probe polarizations (17). By selecting appropriate relative pump and probe polarizations (15), the anisotropic response, Raniso(t), and the isotropic response, Riso(t) are measured. Only these two measurements are required to completely characterize the third-order response of an isotropic medium (18). In a two-beam experiment, only the Raniso(t) response can be determined with high accuracy.

These two measurements are shown in Fig. 1, A and B, for water and 3 M NaCl solution, normalized to the intense response at t = 0 (which arises from the electronic hyperpolarizability and thus contains no information on molecular dynamics). The isotropic and anisotropic measurements [which are formally equivalent to the polarized and depolarized Raman spectral density in the frequency domain (19)] yield fundamentally different information on spectroscopy and dynamics of the liquid phase (20, 21). In terms of molecular vibrational modes, the isotropic spectrum reveals symmetric stretching vibrations, rather than the depolarized modes, such as deformation vibrations and asymmetric stretches, which appear in the anisotropic response. In addition to molecular vibrations, the low-frequency Raman spectrum also contains information on molecular reorientation and intermolecular interactions (16). Molecular reorientation appears in Raniso(t) only and is suppressed in Riso(t), which contains only contributions from intermolecular (interaction induced) components of the sample polarizability (20).

Fig. 1

Measurements of the anisotropic (A) and isotropic (B) third-order responses for water (solid lines) and 3M NaCl solutions (dashed lines). After performing a Fourier transform of the isotropic signal (B) and taking the imaginary part, the corresponding spectral densities (C) are obtained.

The Riso(t) data are shown in Fig. 1B for pure water and a 3 M NaCl solution. For pure water, the isotropic response is an exponential decay with a 56 ± 8 fs time constant, whereas for the NaCl solution a prominent strongly damped oscillatory response is superimposed on the 56 fs exponential decay. Fourier transformation of these data to the frequency domain (Fig. 1C) reveals that aqueous NaCl has a broad but well-defined mode with a peak frequency of 168 ± 4 cm−1. There is no corresponding mode in the pure water spectrum. Because no vibrations can be associated with the monoatomic ions, this band represents a vibration associated with a water-ion interaction. The same measurement was made for aqueous 3 M KCl, and an identical spectrum was recovered (fig. S4), which establishes this mode as arising from an interaction between the anion and water.

The isotropic response of pure water at low frequency (Fig. 1C) has been reported previously (20), with lower signal-to-noise ratio. The signal enhancement associated with the diffractive optic experiment allows us to confirm the appearance of a broad mode in the polarized Raman spectrum around 60 cm−1 and therefore to establish that the depolarization ratio ρ is below 0.75, a point that has previously been unclear (20). A transition at 60 cm−1 has been reported in the depolarized Raman spectrum and was assigned to a bending mode of water molecules linked by hydrogen bonds. A depolarized 180-cm−1 mode was also observed and assigned to an H-bond translational (stretching) mode (22). Both of these modes are delocalized over several water molecules (23). The latter mode is responsible for the oscillation seen in the Raniso(t) for pure water (Fig. 1A) but is absent from the isotropic response (Fig. 1, B and C). This 180-cm−1 depolarized mode is attenuated and shifted to lower frequency in NaCl solutions (Fig. 1A and fig. S7), as described elsewhere (22, 24). The appearance of this mode only in Raniso(t) suggests an assignment to intermolecular translational modes between water molecules, involving interactions of the isotropic components of the molecular polarizability, because this intermolecular interaction will contribute to Raniso(t) but not to Riso(t) (20, 21); for water, the isotropic part of the polarizability has previously been predicted to be the dominant one, consistent with this assignment (21).

We next explored the isotropic response of heavier halide salt solutions. The Raniso(t) responses were also measured for most solutions (figs. S6 and S7), and they agree well with the earlier frequency domain data reported by Tominaga and co-workers (24). The Riso(t) data for 3 M sodium chloride, bromide, and iodide solutions were fit in the time domain to an exponential decay plus a single damped oscillator, Embedded Image, in which τr and τ0 are the exponential and oscillator damping constants, respectively, and ω is the oscillator frequency (Fig. 2). The quality of the fits is good and shows that with increasing size and mass of the ion the damped oscillation shifts to lower frequency (legend to Fig. 2A) while its damping rate remains essentially constant at τ0 = 90 ± 10 fs. For all solutions, a τr value of 56 ± 8 fs is recovered, the time constant associated with pure water. The increasing amplitude of this low-frequency bending mode with increasing anion concentration suggests that the transition undergoes an intensity enhancement in alkali halide solutions. The effect of anion mass supports an assignment to an OH⋯X H-bond. Confirmation of this assignment comes from measurements in deuterated water, which shift the mode to lower frequency by 12 cm−1 (±3 cm−1) (Fig. 2B and fig. S5). In addition, 168 cm−1 is close to the 210-cm−1 frequency reported for a chloride-water complex in the gas phase (25).

Fig. 2

(A) The isotropic signal for 3M NaCl solution (red dot), 3M NaBr solution (green dot) ,and 2M NaI solution (blue dot). Data were fit (black line) to a single exponential (time constant 56 ± 8 fs for all data) plus a damped harmonic oscillation with damping constant of 90 ± 10 fs for all data and frequencies: νNaCl = 168 ± 4 cm−1; νNaBr = 150 ± 5 cm−1; νNaI = 132 ± 6 cm−1. (B) Isotope effect on the isotropic spectral density for 5 M solutions of NaCl in water (solid line) and deuterated water (dashed line). The isotope shift is 12 cm−1 ± 3 cm−1. The spectra were obtained from a direct Fourier transform of the experimental data (the time-domain fits to a single damped oscillator are shown in fig. S5).

The experimentally determined parameters for the oscillation in Riso(t) may be discussed in the light of molecular dynamics simulations of the spectral diffusion and vibrational relaxation of the OH stretch around iodide (12) and chloride ions (11). Both simulations report an oscillatory subpicosecond component assigned to dynamics associated with the OH⋯X H-bond, leading to frequencies of 100 cm−1 and 145 cm−1 for I and Cl, respectively. These frequencies are in reasonable agreement with the present observations (Fig. 2). The 145-cm−1 frequency was assigned to an H-bond stretching vibration (11). The 90-fs damping constant, τ0, recovered from the fit correlates well with the 100- to 200-fs evolution observed in both simulations and measurements of spectral diffusion (9, 11, 12). This time constant is thus associated with fluctuation in the OH⋯X structure. This relaxation time should not be identified with the H-bond lifetime, which is a few picoseconds (11, 12); the measurements reported here cannot separate homogeneous and inhomogeneous contributions to the line width. However, our data are consistent with the ultrafast component in spectral diffusion of the OD oscillator arising from fluctuations in the OH⋯X H-bond.

Steady-state and transient IR measurements of the frequency, vibrational lifetime, and orientational relaxation of the OD stretch have been reported as a function of alkali halide concentration; orientational and vibrational relaxation times become longer with increasing concentration, and the vibration shifts to higher frequency (7, 9). In Fig. 3, the frequency and damping constant of the OH⋯Cl H-bond mode are plotted as a function of the salt concentration. The frequency shifts to a higher value as the concentration increases, whereas τ0 is approximately independent of concentration. At the highest ion concentrations (6 M), there are as few as eight water molecules per ion, so most water molecules are involved in the first solvation shell of an ion [the average coordination number for the Cl ion, for example, is ~6 (5, 14)] and will thus be H-bonded to a water molecule in the solvation shell of an adjacent ion. The blue shift in the OH⋯Cl mode at these high concentrations may thus reflect stronger H-bonding as the water molecules become involved in a network of Cl and Na+ ions.

Fig. 3

Parameters resulting from fitting the concentration dependence of the isotropic response of 1 to 6 M NaCl solutions with a single damped oscillator. Black squares are the recovered frequencies and open dots the damping constants. Error bars reflect the standard deviation of 3 separate experiments.

Finally, we address the question of why the anion gives rise to a single low-frequency mode specifically in the isotropic (polarized) Raman spectrum. Although the low-frequency modes of liquid water are certainly delocalized over several molecules, their assignment has often been aided by analogy with clusters of a few molecules (22, 26). For example, in liquid water much of the low-frequency vibrational spectrum can be assigned from a consideration of a cluster of four water molecules H-bonded to a central molecule in a tetrahedral arrangement (22). We performed density functional theory (DFT) calculations of the Raman spectrum of water clusters around a chloride anion, Cl(H2O)n, searching specifically for polarized modes, which could contribute to Riso(t) and therefore to the spectra reported in Figs. 1 and 2. The Cl(H2O) complex yields an H-bond stretching mode calculated at 184 cm−1 with ρ = 0.4. The frequency is higher than observed experimentally, and there is much evidence from both experiment and calculation that the coordination number for Cl is between 5 and 6 (5, 14, 27). Higher clusters produce a number of low-frequency modes, associated with H-bond stretching and bending (figs. S8 and S9), but the vast majority are largely depolarized (ρ greater than 0.7) and will not contribute strongly to the isotropic response. The Cl(H2O)4 in a square pyramidal structure with Cl at the apex reveals a strongly polarized (ρ = 0.15) symmetric stretching mode at 129 cm−1. This result is particularly important in the light of the Car-Parrinello molecular dynamics simulations of Raugei and Klein (13), who reported a basic square pyramidal structure for the first solvation shell of Br, supplemented by 1 to 3 water molecules, depending on the coordination number. Addition of a fifth water to the cluster yields a symmetric stretching mode at 147 cm−1 with ρ = 0.24. The present experimental data are thus consistent with the picture of asymmetric ion solvation that arises in a number of recent theoretical calculations (13, 14, 28). To obtain a more realistic simulation of the spectrum, DFT calculations were performed on the chloride ion solvated by 24 water molecules, based on the tetrahedral structure given by Raugei and Klein; the final structure is shown in Fig. 4. The calculated polarized Raman spectrum reveals a cluster of polarized (ρ below 0.4) modes between 140 cm−1 and 185 cm−1, consistent with the experimental observations (Figs. 1 and 2); of course, in the liquid, a broad distribution of structures will exist, but the present observations of the hydrogen bonding vibration are consistent with asymmetric solvation of halides reported in simulations.

Fig. 4

Computed structure of a 24–water molecule cluster around the Cl ion. The outer 18 molecules (stick structure) were frozen and the inner solvation shell structure optimized, and the vibrational spectra were then calculated.

Supporting Online Material

Materials and Methods

Figs. S1 to S9


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We are grateful to the Engineering and Physical Sciences Research Council for financial support and to M. Kondo and V. Oganesyan for assistance with the DFT calculations.
View Abstract

Stay Connected to Science

Navigate This Article