Instituto de Química (Universidade Federal do Rio Grande do Sul), Avenida Bento Gonçalves 9500, CEP 91501-970Porto Alegre, Brazil.
Dipartimento di Biotecnologie, Chimica e Farmacia (Università di Siena), via Aldo Moro 2, 53100Siena, SI, Italy.
J Chem Theory Comput. 2022 Nov 8;18(11):6905-6919. doi: 10.1021/acs.jctc.2c00747. Epub 2022 Oct 19.
The wide range of time/length scales covered by self-assembly in soft matter makes molecular dynamics (MD) the ideal candidate for simulating such a supramolecular phenomenon at an atomistic level. However, the reliability of MD outcomes heavily relies on the accuracy of the adopted force-field (FF). The spontaneous re-ordering in liquid crystalline materials stands as a clear example of such collective self-assembling processes, driven by a subtle and delicate balance between supramolecular interactions and single-molecule flexibility. General-purpose transferable FFs often dramatically fail to reproduce such complex phenomena, for example, the error on the transition temperatures being larger than 100 K. Conversely, quantum-mechanically derived force-fields (QMD-FFs), specifically tailored for the target system, were recently shown (. . , 243) to allow for the required accuracy as they not only well reproduced transition temperatures but also yielded a quantitative agreement with the experiment on a wealth of structural, dynamic, and thermodynamic properties. The main drawback of this strategy stands in the computational burden connected to the numerous quantum mechanical (QM) calculations usually required for a target-specific parameterization, which has undoubtedly hampered the routine application of QMD-FFs. In this work, we propose a fragment-based strategy to extend the applicability of QMD-FFs, in which the amount of QM calculations is significantly reduced, being a single-molecule-optimized geometry and its Hessian matrix the only QM information required. To validate this route, a new FF is assembled for a large mesogen, exploiting the parameters obtained for two smaller liquid crystalline molecules, in this and previous work. Lengthy MD simulations are carried out with the new transferred QMD-FF, observing again a spontaneous re-orientation in the correct range of temperatures, with good agreement with the available experimental measures. The present results strongly suggest that a partial transfer of QMD-FF parameters can be invoked without a significant loss of accuracy, thus paving the way to exploit the method's intrinsic predictive capabilities in the simulation of novel soft materials.
自组装在软物质中涵盖的时间/长度尺度范围很广，这使得分子动力学（MD）成为在原子水平上模拟这种超分子现象的理想选择。然而，MD 结果的可靠性在很大程度上依赖于所采用的力场（FF）的准确性。液晶材料中的自发重排就是这种集体自组装过程的一个明显例子，它是由超分子相互作用和单分子柔性之间的微妙平衡驱动的。通用可转移 FF 通常无法重现这种复杂现象，例如，转变温度的误差大于 100 K。相反，最近的量子力学衍生力场（QMD-FF）（例如， 243）专门针对目标系统进行了调整，被证明可以达到所需的精度，因为它们不仅很好地重现了转变温度，而且在大量结构、动态和热力学性质上与实验结果定量一致。这种策略的主要缺点在于与目标特定参数化相关的计算负担，这无疑阻碍了 QMD-FF 的常规应用。在这项工作中，我们提出了一种基于片段的策略来扩展 QMD-FF 的适用性，其中大大减少了 QM 计算的数量，仅需要一个单分子优化的几何形状及其 Hessian 矩阵作为唯一的 QM 信息。为了验证这条途径，我们利用两个较小的液晶分子获得的参数，为一个大介晶组装了一个新的 FF。利用新的转移 QMD-FF 进行了长时间的 MD 模拟，再次观察到在正确温度范围内的自发重取向，与可用的实验测量结果吻合良好。目前的结果强烈表明，可以在不显著降低准确性的情况下调用 QMD-FF 参数的部分转移，从而为在新型软物质的模拟中利用该方法的内在预测能力铺平道路。
