Here is the abstract you requested from the Thermal_2014 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.
|Ballistic heat conduction challenges in high-speed electronics|
|Keywords: ballistic effects, dynamic temperature response, reliability|
|The power dissipation in digital ICs and power electronics is highly dynamic. As a result, the junction temperature is composed of a transient fluctuation around a steady state baseline. The static component is affected by the entire heat flow path, from the component level over TIMs and packaging to heat sinks and active cooling systems. The transient response, by contrast, is completely dominated by the dynamic thermal impedance Z of just the semiconductor die. Research over the past few years has shown that at frequencies as low as 10MHz, the measured Z can be nearly two times larger than conventionally predicted. The effect is caused by long-range thermal energy carriers that don't undergo scattering near the semiconductor surface. Such “ballistic” transport modes cause both magnitude and spatial shape of the internal thermal field to deviate from standard Fourier diffusion theory. In this presentation, we analyse the impact of such ballistic effects on the thermal management of electronic systems. A custom developed and experimentally verified ballistic heat model is used to perform two case studies of integrated sub-millimeter heat sources on Si and InGaAs. First we investigate the transient thermal response to pulsed power dissipations with repetition rates ranging from 10MHz to 1GHz and duty cycles between 10% and 50% while maintaining an 80C baseline temperature. Ballistic peak temperatures exceed diffusive predictions by up to 33% (129C vs. 97C). The near threefold increase of the thermal swing around the baseline suggests a significant risk increase for thermal fatigue failures. Next, we analyse the excess ballistic temperature rise over the 10kHz-1GHz bandwidth when the device is set, according to conventional theory, to operate at a 100C peak temperature. The strongest deviations (ballistic peak temperature ~110C) are observed for small duty cycles at intermediate repetition rates ~35MHz in Si and ~300kHz in InGaAs. This is the result of a delicate interplay: as the frequency increases, the transport becomes relatively more anomalous but the absolute Z magnitude is reduced. Overall, our results illustrate that fundamental microscopic physics of heat conduction raise additional thermal challenges in high-speed electronic systems.|
|Bjorn Vermeersch, PostDoctoral Associate
West Lafayette, IN