Design for semiconductor manufacturing. Perspective

Abstract

The semiconductor industry is challenged to sustain the historical rate of exponential growth in product innovation and productivity for as long into the future as possible. If this challenge is to be met, substantial investments in chip-design automation are required to improve design capabilities and to narrow the widening gap between projected increases in device complexity and design productivity. Design for manufacture (DFM) is one potential means to leverage the required investment in design automation. DFM focuses on understanding how the design of a product interacts with the various processes and facilities available to make that product, in order to conceive and refine design alternatives that tend to optimize the product in the context of existing or projected manufacturing capabilities. In this paper, we review the current literature on DFM as applied in other industries and consider the relevance of state-of-the-art DFM principles and tools to improved efficiency and productivity within the semiconductor industry. We observe that, overall, many of the organizational structures for implementing DFM already exist within semiconductor firms and the industry more broadly. Moreover, much of the existing spectrum of layout, modeling, simulation, and analysis tools used for semiconductor device, process, equipment, and factory design can be thought of as components of a de facto system of design for semiconductor manufacturing (DFSM). While recent work aimed specifically at DFSM shows great promise, advances in the development of component design tools are lagging and integration of these tools into a coherent system for DFSM is only emerging. Systems integration of tools which characterize future processes and technologies will be needed in order to posit design rules which are sufficiently robust to allow their incorporation in industry-standard automated design and layout tools. The implications are that there is a need for accelerated improvements in DFSM tools and integration if historical semiconductor productivity growth is to be sustained.