Professor Sachin Sapatnekar

These researchers use MSI supercomputing resources to solve problems in the domain of computer-aided analysis and optimization of VLSI designs. Specifically, the projects will address the three problems described below.

• Impact of thermal-mechanical effects on circuit performance in sub-micron planar and 3D-IC technologies: Thermal issues are growing increasingly critical in deeply-scaled integrated circuits. Newer CMOS device geometries are very compact and generate a large amount of heat in a small volume, with restricted heat removal paths to the ambient. This implies that transistors and wires on a chip experience high temperatures, which affect their performance. Moreover, in emerging 3D-IC technologies, dies are stacked vertically and thick cylindrical shaped copper pillars (through silicon vias, TSVs) are used to connect the circuits on different layers. During manufacturing, both silicon and copper TSVs undergo annealing process with a temperature ramp from 250 degree down to room temperature. However, due to the coefficient of thermal expansion (CTE) mismatch between copper and silicon, thermal residual stress develops inside silicon that impacts the transistor electrical properties and thereby system timing performance. This project is looking into methods for modeling these effects and capturing their impact on circuit performance.
• Power delivery network analysis: The design of power delivery networks is an important step in the design of high-performance integrated systems. The group's work explores various aspects of this problem. First, they address the design of power delivery networks using machine learning. This requires the training of the ML model based on exhaustive simulation, for which MSI machines and infrastructure are very useful. Second, they study the issue of electromigration in power networks, which has become a serious reliability issue while designing VLSI circuits in current technologies. Current models for electromigration are simple and inaccurate, and the group aims to come up with an electromigration model for the interconnects (metal wires) in the power delivery network that is computationally efficient while physically accurate, and further apply the model to analyze the power delivery network, which typically has hundreds of thousands of metal wires for a single chip. Moreover, electromigration is also intimately linked with on-chip residual stress due to CTE mismatches between various constituents of the chip, as well as the stress build-up due to the gradient of atomic concentration. The goal is to analyze the impact of these effects.
• Simulation of nanomagnetic structures: Spintronics has emerged as a promising platform for the next generation of computing hardware. The development of new spintronics-based computing devices requires the detailed, compute-intensive simulation of these magnetic devices. This project uses MSI resources to perform micromagnetic and other simulations to verify, analyze, and optimize spintronic devices.