Fabrication Steps and Thermal Modeling of Three-Dimensional Asynchronous Field Programmable Gate Array (3D-AFPGA) With Through Silicon Via and Copper Pillar Bonding Approach

Abstract

This work will specifically detail the development of a processing and fabrication route for a three-dimensional asynchronous field-programmable gate array (3D-AFPGA) design based on an extension of pre-existing two-dimensional-field-programmable gate array (2D-FPGA) tile designs. The periodic nature of FPGAs permits the use of an alternative approach, whereby the design entails splitting the FPGA design along tile borders and inserting through silicon vias (TSVs) at regular spatial intervals. This serves to enable true 3D performance (i.e., full 3D signal routing) while leaving most of the 2D circuit layouts intact. 3D signal buffers are inserted to handle communication between vertical and adjacent neighbors. For this approach, the density of vertical interconnections was shown to be determined by the size of the bond pads used for tier–tier communications and bonding. As a consequence, reducing bond pad dimensions from 25 μm to 15 μm, or 10 μm, bond pads are preferred to increase the connectivity between layers. A 3D-AFPGA mockup test structure was then proposed for completing development and exercising the 3D integration process flows. This mockup test structure consists of a three-tier demonstration vehicle consisting of a chip-to-wafer and a subsequent chip-to-chip bond. Besides, an alternate copper bonding approach using pillars was explored. Although the intended application is for the 3D integration process compatible with the 3D AFPGA design, the test structure was also designed to be generally applicable to various applications for 3D integration. Because of the importance of thermal management of 3D-AFPGA, it is important to predict the temperature distribution and avoid the maximum junction temperature. The numerical thermal modeling for predicting the equivalent thermal conductivity in every layer and the 3D temperature distribution in the 3D-AFPGA are developed and discussed as well.