Power consumed by integrated circuits (ICs) converts to heat and dissipates through the material of the IC and the surrounding packaging. Heat dissipation from the interior of an IC to the ambient becomes highly constrained when power hungry high performance chips are stacked vertically to create 3D ICs. Pockets of accumulated heat with poor heat conduction paths to the ambient is then trapped in between stacked IC layers. As a result, overheating within ICs threaten the safety and performance of future computing systems that rely on this important new IC design methodology. This project will develop a new thermal sensor design that is especially well suited to be integrated into 3D ICs during semiconductor processing steps. It is smaller by design, i.e., an arbitrarily large number of sensors can monitor an unprecedented part of the IC, and they do not consume additional power. This will impact the sustainability and computational power of future computing systems by enabling them to operate at maximal frequencies without the need for costly active cooling solutions. They also affect the pricing of the chip products and therefore have a direct impact on the industry and economy. The PIs will continue to train graduate and undergraduate students in the basic research that underlies creative technological developments and actively promote science and technology in outreach events to the community. One of the PIs will also leverage her membership in the Diversity Committee at her institution to attract underrepresented minority graduate student applicants.
The project involves a new paradigm for design of thermal sensors using Thin Film Bimetallic Thermocouples (TFBTs), which senses temperature according to an intrinsic material property that is independent of process variation and thermal conditions of the environment. Since TFBTs are passive, they do not consume power or dissipate heat during sensing. Significant challenges remain in better understanding and modeling of the materials as well as the integration of the TFBTs into 3D ICs. The project will study the basic science of these thermocouple metals as a function of layer thickness, since thin film Seebeck coefficients are known to differ significantly from bulk. Particularly, experiments will be undertaken to investigate the dependence of the Seebeck coefficient on film thickness and on the choice of various candidate metals for thermocouples. Samples using integrated resistive heater elements on semiconductor substrates will simulate hot spots underneath arrays of bimetallic thermocouples, generating thermal maps with high spatial resolution. Design optimizations for effective integration into 3D ICs will be developed. Novel thermal management schemes that can make use of the resulting fine-grain, robust, and low cost sensor array will be developed and evaluated.