3D-IR spectroscopy of isotope-substituted liquid water reveals heterogeneous dynamics

The dynamics of the hydrogen bond network of isotopically substituted liquid water are investigated with a new ultrafast nonlinear vibrational spectroscopy, three-dimensional infrared spectroscopy (3D-IR). The 3D-IR spectroscopy is sensitive to three-point frequency fluctuation correlation functions, and the measurements reveal heterogeneous structural relaxation dynamics. We interpret these results as subensembles of water which do not interconvert on a half picosecond time scale. We connect the experimental results to molecular dynamics (MD) simulations, performing a line shape analysis as well as complex network analysis. 3D-IR spectroscopy of isotope-substituted liquid water reveals heterogeneous dynamics Sean Garrett-Roe,† Fivos Perakis,† Francesco Rao,‡ and Peter Hamm∗,† Institute of Physical Chemistry, University of Zurich, Zurich, Switzerland, and Freiburg Institute for Advanced Studies (FRIAS), University of Freiburg, Freiburg, Germany E-mail: [email protected] ∗To whom correspondence should be addressed †University of Zurich ‡University of Freiburg 1 Abstract The dynamics of the hydrogen bond network of isotopically substituted liquid water are investigated with a new ultrafast nonlinear vibrational spectroscopy, three-dimensional infrared spectroscopy (3D-IR). The 3D-IR spectroscopy is sensitive to three-point frequency fluctuation correlation functions, and the measurements reveal heterogeneous structural relaxation dynamics. We interpret these results as subensembles of water which do not interconvert on half a picosecond timescale. We connect the experimental results to molecular dynamics simulations, performing a lineshape analysis as well as complex network analysis.The dynamics of the hydrogen bond network of isotopically substituted liquid water are investigated with a new ultrafast nonlinear vibrational spectroscopy, three-dimensional infrared spectroscopy (3D-IR). The 3D-IR spectroscopy is sensitive to three-point frequency fluctuation correlation functions, and the measurements reveal heterogeneous structural relaxation dynamics. We interpret these results as subensembles of water which do not interconvert on half a picosecond timescale. We connect the experimental results to molecular dynamics simulations, performing a lineshape analysis as well as complex network analysis. Introduction Since Roentgen,1 researchers have speculated about inhomogeneous structures in ambient liquid water. It is well understood that as a liquid is supercooled, it develops both structural and dynamical heterogeneities,2,3 which occur when the fluid is no longer uniform, e.g. jamming occurs and different domains develop different structures. In ambient water, small angle x-ray scattering experiments4 have recently been interpreted as evidence of structural heterogeneities on the 1 nm length-scale, though not without controversy.5–7 Liquid water is a highly dynamic medium, of course, and its hydrogen bond network reorganizes on a picosecond time scale.8–18 Here we apply a new multidimensional ultrafast infrared technique, three dimensional infrared spectroscopy (3DIR), to isotopically substituted liquid water at ambient conditions. We observe that the structural relaxation dynamics are heterogeneous, and that water contains a mixture of distinct subensembles which do not intermix on a half picosecond timescale. Whether liquid water should be considered one uniform, homogeneous fluid or a mixture of two or more distinct subspecies is a long standing, unresolved, and controversial question. Many researchers have tried to identify spectroscopically distinct subensembles of hydrogen bonding using multiple pulse IR spectroscopies.8–11,14,15,18,19 Experiments using IR and Raman, for example, have examined wavelength dependent rates of orientational relaxation,8,19 vibrational relaxation,10,15,19,20 and spectral diffusion.21 The evidence in the literature weighs on both sides of 2 the argument. Perhaps foremost, the presence of an isosbestic point in the OH-stretching band of water as a function of temperature19,22 has long been seen as the proverbial smoking gun of two-state behavior. A simple thermodynamic analysis, however, shows that this is not necessarily so;20,23 isosbestic points can appear in a completely homogeneous system. The pathways of vibrational relaxation suggest vibrational subbands associated with strongly and weakly hydrogen bonded waters which are distinct on a 2–10 ps timescale.15 On the other hand, analysis of 2DIR lineshapes and molecular dynamics simulations say that these hydrogen bonds are broken only fleetingly (∼200 fs), much as a transition state or saddle point is only transiently populated.24 Similar fundamental disagreements occur in the context of water around ions,25–27 and water around hydrophobes.28–30 It is meaningful to consider a fluid heterogenous if it has multiple, distinct rates of structural relaxation. Here we demonstrate that 3D-IR spectroscopy can report whether or not structural relaxation rates in liquid water remain distinct through a three-point frequency fluctuation correlation function. The measurements show heterogeneity which persists on half a picosecond timescale. The results are interpreted using a complex network analysis of molecular dynamics (MD) simulations of water which provides a molecular description of the multiple local free-energy minima generating the heterogeneous dynamics. Theoretical foundations Ultrafast vibrational spectroscopy is useful to probe structural dynamics in liquid water because water’s vibrational frequencies depend on the local structure.10,14,18,31,32 The most common vibrational marker modes are the stretching vibrations of isotope diluted water; in this work, we use the OD vibration of HOD in H2O. For example, 2D-IR spectroscopy12,13,16 measures diffusion of the OD vibrator in frequency space (spectral diffusion), which is directly connected to diffusion of a water molecule in real space. Conceptually speaking, the 2D-IR pulse sequence interrogates the OD vibrator by first tagging an initial OD vibrational frequency; then the environment around the OD evolves during a time t2; finally, the sequence reads out the resulting vibrational frequency.