Accelerating Nano-scale Transistor Innovation though Petascale Simulation

通过千万亿次模拟加速纳米级晶体管创新四十年的晶体管缩小已经导致了原子尺度的特征，使器件在原子尺度上受到不可避免的制造不规则性的影响。一种新的设计方法包含了组成材料的原子性、量子力学（QM）性质，这对于开发更强大但又节省能量的设备是必要的。这就要求在超过数千万个复自由度的系统中求解非平衡统计QM。由于缺乏足够强大的计算机，这种量级的计算是不可能的。千兆级计算创造了追求多尺度设计方法的机会。我们将开发一个通用的模拟引擎，它将在包含数百万原子的原子材料描述中使用非平衡格林函数（NEGF）形式主义模拟实际扩展设备中的非平衡电子传输。首席研究员开发了NEMO1D，这是第一个基于NEGF的商业级设备模拟器，以及NEMO3D，用于计算大于5200万个原子的系统的电子结构（无NEGF）。这些代码已经被证明可以分别扩展到超过16000和8000核。该项目希望通过NEMO3D电子结构描述扩展NEMO1D的nef功能，形成下一代工程工具包OMEN。这项工作希望能够发现更快的开关，更小的特征尺寸和减少热量产生的新技术。一种新型开关的发明有可能在2015年重振半导体行业。现有的建模和模拟工具已经从宏观的理解（材料是光滑的连续体，电子是经典的粒子）开始缩小尺度。在纳米尺度上，材料必须用离散的原子分辨率来描述，电子遵循量子力学，没有现实设备的工具。本项目旨在构建这样一个工具。OMEN将是一个通过nanoHUB.org发布的开源代码社区代码。NEMO3D已经体现了对OMEN未来使用的设想，一个高度可扩展的工具可以回答广泛的问题，涵盖计算强度范围，为广泛的用户。自发布以来，已有超过1000名用户使用了NEMO3D的教育版本。

Accelerating Nano-scale Transistor Innovation though Petascale SimulationGerhard KlimeckPurdue UniversityForty years of transistor downscaling has led to atomic-scale features, making devices subject to unavoidable manufacturing irregularities at the atomic scale. A new approach to design that embraces the atomistic, quantum mechanical (QM) nature of the constituent materials is necessary to develop more powerful yet energy miserly devices. This demands the solution of non-equilibrium statistical QM in systems in excess of tens of million complex degrees of freedom. Computations of this magnitude have been impossible, due to the lack of sufficiently powerful computers. Petascale computing creates an opportunity to pursue a multi-scale design approach. We will develop a general-purpose simulation engine, which will model out-of-equilibrium electron transport in realistically extended devices in an atomistic material description containing millions of atoms using the non-equilibrium Green function (NEGF) formalism. The principal investigator developed NEMO1D, the first NEGF-based commercial-grade device simulator, and NEMO3D, which calculates electronic structure (no NEGF) for systems larger than 52 million atoms. The codes have been shown to scale to more than 16,000 and 8,000 cores, respectively. The project expects to extend the NEGF capabilities of NEMO1D with the NEMO3D electronic structure description to form the next generation engineering toolkit, OMEN.This work hopes to enable the discovery of new technologies for faster switching, smaller feature size, and reduced heat generation. The creation of a new switch has the potential to revitalize the semiconductor industry in 2015. Existing modeling and simulation tools have followed the path of downscaling from a macroscopic understanding where materials are a smooth continuum and electrons are classical particles. At the nanometer scale the material must be described with discrete atomic resolution and electrons follow quantum mechanics and no tools exist for realistic devices. This project aims to build such a tool. OMEN will be an open source community code released through nanoHUB.org. NEMO3D already embodies the envisioned future use of OMEN where a highly scalable tool can answer a broad range of problems, covering ranges of computational intensity, for a broad range of users. Over 1,000 users have used an educational version of NEMO3D since its release.