IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 5, NO. 1, JANUARY 2015 57
Experimental Characterization of Hybrid Solid-State
and Fluidic Cooling for Thermal Management of
Localized Hotspots
Vivek Sahu, Yogendra K. Joshi, Fellow, IEEE , Andrei G. Fedorov, Je-Hyeong Bahk, Xi Wang, and Ali Shakouri
Abstract—A hybrid cooling scheme utilizing superlattice
coolers (SLCs) along with microchannel heat sink for thermal
management of hotspots is presented. In this paper, we have
studied the effect of operating and design parameters on the
performance of the SLC. We have also experimentally inves-
tigated the effect of interface thermal resistance as well as
thermal resistance between the ground electrode and superlattice
using two test configurations: one with on-chip microchannels
and another with off-chip microchannels. We demonstrated heat
removal capability at the localized hotspots of more than over
300 W/cm
2
.
Index Terms— Hotspot cooling, hybrid cooling, microchannel
heat sink, superlattice cooler (SLC), thermal management.
I. I NTRODUCTION
T
HE need to reduce the cost and increase the performance
of ICs has led to shrinkage of the transistor size and
increased transistor packing density immensely. This has not
only increased the power dissipation, but also led to formation
of localized hotspots. Heat flux at these localized hotspots can
be five to ten times higher than the average heat flux dissi-
pated across the die [1]. Even though International Technol-
ogy Roadmap for Semiconductors (ITRS) predicts sustained
power dissipation over the entire die, heat flux dissipated
at the hotspot is expected to rise. Due to the presence of
hotspots, power dissipation becomes highly nonuniform. This
results in large temperature gradient and residual stresses,
and further decreases the reliability and life of the device.
Moreover, the maximum allowable junction temperature for
high-performance devices is projected to drop to 75 °C by
2016 [2], which would reduce the overall thermal budget and
make thermal management even more challenging. Effective
thermal management of electronics will not only require
Manuscript received January 18, 2013; revised May 28, 2013; accepted
June 18, 2014. Date of publication December 1, 2014; date of current
version January 7, 2015. This work was supported in part by the Interconnect
Focus Center through the Focus Center Research Program, and in part by a
Semiconductor Research Corporation Program. Recommended for publication
by Associate Editor B. Sammakia upon evaluation of reviewers’ comments.
V. Sahu, Y. K. Joshi, and A. G. Fedorov are with the George
W. Woodruff School of Mechanical Engineering, Georgia Institute of
Technology, Austin, GA 31546 USA (e-mail: vivek.sahu@gmail.com;
yogendra.joshi@me.gatech.edu; andrei.fedorov@me.gatech.edu).
J.-H. Bahk and X. Wang are with the Baskin School of Engineering,
University of California at Santa Cruz, Santa Cruz, CA 95064 USA.
A. Shakouri is with the Department of Electrical Engineering, Purdue
University, West Lafayette, IN 47907 USA.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TCPMT.2014.2332516
reduction in hotspot temperature, but also minimization of
temperature gradients across the die.
Liquid cooling has shown a great promise to dissipate
high heat fluxes. Kandlikar et al. [3] have provided a good
overview of recent microchannel cooling studies. Tuckerman
and Pease [4] were able to remove 790 W/cm
2
using liquid
cooled microchannel heat sink at the expense of huge pressure
drop (214 kPa) and volumetric flow rate (500 ml/min). Current
microscale pump technologies cannot deliver the required flow
rate at the pressure drop mentioned above [5]. Colgan et al. [6]
used offset strip fin arrangement inside the microchannel and
were able to dissipate close to 275 W/cm
2
of heat flux using
single phase forced convection with water as the working
fluid, while maintaining the junction temperature at 85 °C.
They used multiple entries manifold to keep the channel length
short and the pressure drop small. Several other studies have
been carried out to dissipate high flux using liquid cooled
microchannel heat sinks [7]–[9]. Even though liquid cooling
is capable of removing high heat flux, pressure drop and
volumetric flow rate penalties become severe for heat flux
in excess of 100 W/cm
2
. Moreover, since microchannel heat
sink provides uniform heat flux dissipation, it has to be
designed for the highest heat flux dissipated from the die,
which can be several times higher than the average heat flux.
This will results in highly overdesigned microchannel heat
sink. Furthermore, as microchannel heat sink provides uniform
heat flux removal capability, it will not alleviate temperature
gradients across the die.
Solid-state thermoelectric coolers have been widely used
for several Decades; however, due to poor material properties,
they have not gained much interest in thermal management
community. Solid-state coolers offer significant advantages
over other cooling techniques. They are mechanically robust,
compact, and capable of reducing the temperature below
ambient. However, due to the low coefficient of perfor-
mance (COP), they are not suitable for thermal management
of the entire chip. Since the total power dissipated from
hotspots (usually < 1–2 W) is significantly less than the
power dissipated by the entire chip (around 50–100 W), solid-
state coolers have been utilized for hotspot cooling. Prasher
et al. [10] used a microthermoelectric cooler to provide 15 °C
of cooling at 1300 W/cm
2
hotspot heat flux. In their work,
microthermoelectric coolers were implemented at the package
level between the heat spreader and the thermal interface
material (TIM) layer, due to which significant temperature
gradients were still present over the die.
2156-3950 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.