Numerical Methods for Molecular Dynamics

分子动力学数值方法

• 批准号: 9971830
• 负责人: Robert Skeel
• 金额: $ 21.48万
• 依托单位: University of Illinois at Urbana-Champaign
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-07-15 至 2003-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9971830&HistoricalAwards=false
• 关键词: Numerical; Methods; Molecular; Dynamics

The primary goal of the research is to create more efficientpropagators for atomistic computer simulationsand thus to make possible more ambitious scientific calculations.New deterministic and stochastic algorithms for both dynamics and samplingare to be constructed using such techniques as operator splitting, modifiedenergy functions that compensate for finite steps,and optimization of method parameters, together with physical insight.Promising algorithms are tested and compared using mathematical analysesand computer experiments. Tools for mathematical analysis include theconcept of effective accuracy, the method of modified equations, linearanalysis, and KAM theory. Computer experiments are performed on modelproblems chosen to reveal unambiguously the properties of interest.Faster algorithms are to be implemented in molecular modeling softwarebeing developed for widespread use in a couple of projects at theUniversity of Illinois Beckman Institute. Most of the work issufficiently general that is transfers to other types of problems suchas occur in astrophysics, wave phenomena, and mechanical engineering.And many of the techniques can be abstracted and applied to generic problems.Computer simulations of atomic detail are heavily employed in physics,chemistry, materials science, and structural biology.These calculations require the generation of sequences of atomicconfigurations either for the purpose of modeling actual motion or for thepurpose of calculating averaged values and structuresfrom a wide range of representative samples. The computing time rangesfrom hours to months, so it can benefit tremendously from faster algorithms.It is the objective of this research project to do this:to create much more efficient propagators for dynamicsand sampling that reliably achieve acceptable levels of accuracy.The construction of such algorithms employs ideas from mathematicsand computer science together with physical insight.Promising algorithms are tested and compared usingmathematical analyses and computer experiments.The successful ones are implemented in molecular modeling softwarebeing developed for widespread use in a couple of projects at the Universityof Illinois Beckman Institute. These advances in methodology are also to bedisseminated in articles targeted to practitioners.Many of the techniques will apply not only to molecular simulationsbut also to simulations in astrophysics, structural mechanics, and fluids.Potentially, the availability of accelerated propagation algorithmswill lead to a variety of scientific results that otherwise would not beobtained. The performance of the research will bea valuable interdisciplinary experience for a graduate student.

研究的主要目标是为原子计算机模拟创造更有效的传播子，从而使更雄心勃勃的科学计算成为可能。新的动力学和采样的确定性和随机算法将使用算子分裂，修正能量函数补偿有限步骤和优化方法参数等技术来构建，结合物理的洞察力，对有前途的算法进行了数学分析和计算机实验的测试和比较。 数学分析工具包括有效精度概念、修正方程法、线性分析和KAM理论。 计算机实验进行模型问题选择明确地揭示感兴趣的属性。更快的算法将在分子建模软件中实现，该软件正在开发中，广泛用于伊利诺斯大学贝克曼研究所的几个项目。 大多数工作都是一般性的，即转移到其他类型的问题，如天体物理学，波动现象和机械工程。许多技术可以抽象并应用于一般问题。原子细节的计算机模拟在物理，化学，材料科学，这些计算需要生成原子构型的序列，无论是为了模拟实际的运动，还是为了计算平均值和结构从广泛的代表性样品。 计算时间从几个小时到几个月不等，因此它可以从更快的算法中受益匪浅。本研究项目的目标是做到这一点：为动力学和采样创建更有效的传播器，可靠地达到可接受的精度水平。这种算法的构建采用了数学和计算机科学的思想以及物理洞察力。使用数学分析和计算机实验对有前途的算法进行测试和比较。成功的算法在分子建模软件中实现，该软件正在开发中，广泛用于University of Technology的几个项目。伊利诺伊州贝克曼研究所。 这些方法学上的进展也将在针对从业者的文章中传播。许多技术不仅适用于分子模拟，而且适用于天体物理学、结构力学和流体的模拟。潜在地，加速传播算法的可用性将导致各种科学结果，否则将无法获得。 研究的表现将是一个宝贵的跨学科经验的研究生。