Biophysical Chemistry, Lund University, POB 124, SE-22100 Lund, Sweden.
J Chem Phys. 2011 Apr 14;134(14):144508. doi: 10.1063/1.3578472.
One of the outstanding challenges presented by liquid water is to understand how molecules can move on a picosecond time scale despite being incorporated in a three-dimensional network of relatively strong H-bonds. This challenge is exacerbated in the supercooled state, where the dramatic slowing down of structural dynamics is reminiscent of the, equally poorly understood, generic behavior of liquids near the glass transition temperature. By probing single-molecule dynamics on a wide range of time and length scales, quasielastic neutron scattering (QENS) can potentially reveal the mechanistic details of water's structural dynamics, but because of interpretational ambiguities this potential has not been fully realized. To resolve these issues, we present here an extensive set of high-quality QENS data from water in the range 253-293 K and a corresponding set of molecular dynamics (MD) simulations to facilitate and validate the interpretation. Using a model-free approach, we analyze the QENS data in terms of two motional components. Based on the dynamical clustering observed in MD trajectories, we identify these components with two distinct types of structural dynamics: picosecond local (L) structural fluctuations within dynamical basins and slower interbasin jumps (J). The Q-dependence of the dominant QENS component, associated with J dynamics, can be quantitatively rationalized with a continuous-time random walk (CTRW) model with an apparent jump length that depends on low-order moments of the jump length and waiting time distributions. Using a simple coarse-graining algorithm to quantitatively identify dynamical basins, we map the newtonian MD trajectory on a CTRW trajectory, from which the jump length and waiting time distributions are computed. The jump length distribution is gaussian and the rms jump length increases from 1.5 to 1.9 Å as the temperature increases from 253 to 293 K. The rms basin radius increases from 0.71 to 0.75 Å over the same range. The waiting time distribution is exponential at all investigated temperatures, ruling out significant dynamical heterogeneity. However, a simulation at 238 K reveals a small but significant dynamical heterogeneity. The macroscopic diffusion coefficient deduced from the QENS data agrees quantitatively with NMR and tracer results. We compare our QENS analysis with existing approaches, arguing that the apparent dynamical heterogeneity implied by stretched exponential fitting functions results from the failure to distinguish intrabasin (L) from interbasin (J) structural dynamics. We propose that the apparent dynamical singularity at ∼220 K corresponds to freezing out of J dynamics, while the calorimetric glass transition corresponds to freezing out of L dynamics.
液态水带来的一个突出挑战是，要理解分子如何在皮秒时间尺度上移动，尽管它们被包含在相对较强的氢键的三维网络中。在过冷状态下，这种挑战更加严重，结构动力学的急剧减缓让人想起同样难以理解的、液体在玻璃化转变温度附近的一般行为。通过在广泛的时间和长度尺度上探测单分子动力学，非弹性中子散射（QENS）有可能揭示水结构动力学的机制细节，但由于解释上的模糊性，这一潜力尚未得到充分实现。为了解决这些问题，我们在这里呈现了一系列来自 253-293 K 范围内的高质量水的 QENS 数据和相应的分子动力学（MD）模拟，以促进和验证解释。我们使用无模型方法，根据两个运动成分来分析 QENS 数据。基于 MD 轨迹中观察到的动态聚类，我们将这些成分与两种不同类型的结构动力学联系起来：动态盆地内的皮秒局部（L）结构波动和较慢的盆地间跳跃（J）。与 J 动力学相关的主导 QENS 成分的 Q 依赖性可以用连续时间随机行走（CTRW）模型进行定量合理化，该模型的表观跳跃长度取决于跳跃长度和等待时间分布的低阶矩。使用简单的粗粒化算法对动态盆地进行定量识别，我们将牛顿 MD 轨迹映射到 CTRW 轨迹上，从中计算出跳跃长度和等待时间分布。跳跃长度分布呈高斯分布，随着温度从 253 升高到 293 K，均方根跳跃长度从 1.5 增加到 1.9 Å。在相同的范围内，均方根盆地半径从 0.71 增加到 0.75 Å。在所有研究的温度下，等待时间分布都是指数分布，排除了显著的动力学异质性。然而，在 238 K 的模拟中发现了一个很小但显著的动力学异质性。从 QENS 数据推断出的宏观扩散系数与 NMR 和示踪结果定量一致。我们将我们的 QENS 分析与现有方法进行了比较，认为拉伸指数拟合函数所暗示的表观动力学异质性是由于未能区分盆地内（L）和盆地间（J）结构动力学。我们提出，在约 220 K 时出现的表观动力学奇点对应于 J 动力学的冻结，而量热玻璃化转变对应于 L 动力学的冻结。
