microchannel cooling device

We propose and analyze a novel two-phase microchannel cooling device that incorporates perforated
side walls for potential use as an embedded thermal management solution for high heat flux semiconductor devices. A dense array of perforated side walls separate alternating liquid and vapor microchannels,
allowing the vapor generated through evaporation of liquid supplied through micro-perforations to flow
only in the dedicated vapor channels. By separating the liquid and vapor flows, these ‘‘perspiring’’ side
walls enable us to circumvent flow instabilities and other challenges associated with conventional
two-phase microchannel cooling while at the same time effectively take advantage of the large extended
surface areas available in high-aspect-ratio microchannels. One implementation of our design is parametrically analyzed using finite element modeling, demonstrating the potential of our proposed device for
handling high heat flux electronic and optoelectronic semiconductor devices.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Continued miniaturization and increasing power ratings of electronic and optoelectronic devices have resulted in the need for active embedded cooling techniques with high heat flux removal
capability. Flow boiling in microchannels has received a lot of
attention for its promise in handling ultra-high heat fluxes
[1–12]. Practical applications of flow boiling in microchannels for
semiconductor device cooling, however, remain challenging due
to issues related to flow instabilities, complexity of two-phase flow
regimes, and excessive pressure drops associated with inlet constrictions. Ability to initiate and localize bubble nucleation in a
controlled manner may help partially mitigate some of these challenges but such ability remains elusive for most dielectric fluids
with very high wettability.
An alternative approach involves separating the liquid and vapor phases as an intrinsic part of the device design to help alleviate
some of the difficulties. An early attempt by David et al.[13]placed
a hydrophobic ‘‘breather’’ membrane on conventional microchannels to partially separate water vapor from liquid water and thereby mitigate flow instabilities. In this study, flow boiling in
microchannels with a 65lm thick porous PTFE membrane (with
220 nm pores) separating the liquid channels from the vapor channels was investigated. The device had 19 channels machined out of
copper with each channel being 19 mm long. The liquid microchannels were 130lm wide and 134lm deep, while the mating
vapor channel were 125lm wide and 132lm deep. Experiments
were performed with water as the test fluid. A thermofluid model
for the device (vented and non-vented) was also developed to predict the performance. The experiments showed that with vapor
venting, the quality of the two-phase flow in the microchannels
was lower. Experimental results showed that compared to a nonventing device, the venting device had 60% reduction in the pressure drop. Model predictions matched well with the experimental
results, except at higher mass flow rates (>600 kg/m
2
s) where the
model overpredicts the pressure drop. This was attributed to the
formation of churn-annular flow pattern observed at high mass
fluxes. Additionally, no two-phase flow instabilities were observed
in these experiments.
In these experiments, heat fluxes up to 70 W/cm
2
were used.
For low mass flux, the heat transfer coefficients measured for the
vented device was lower than that for the non-vented device. On
the other hand at higher mass fluxes, the heat transfer coefficients
for the vented device were approximately the same or slightly
higher than those measured for the non-vented device. Though
the vapor quality in the microchannels were lower in the case of
the vented device, the reduction in the fluid saturation temperature (due to reduction in the pressure drop, up to 4.4 C) and
changes in the flow regime resulted in the heat transfer rates being
similar to that obtained for the non-vented device. Due to lower
vapor quality, stratified flow was typically obtained which generally has lower heat transfer rates compared to annular or churnannular flows. The model predictions matched the experimental
results quite well, except at the lowest mass flux where they were
overpredicted. Though the experimental results obtained using the
0017-9310/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2013.09.022
⇑Corresponding author. Tel.: +1 310 825 9617.
E-mail address:gwarrier@ucla.edu(G.R. Warrier).
International Journal of Heat and Mass Transfer 68 (2014) 174–183
Contents lists available atScienceDirect
International Journal of Heat and Mass Transfer
journal homepage: www.elsevier.com/locate/ijhmt
vented-device did not show much improvement in the heat transfer rates compared to the non-vented device, the results did show a
clear reduction in the pressure drop characteristics. The authors
hypothesized that the much higher heat transfer rates could be
realized with small liquid channel and very large vapor channels
separated by a porous membrane. Since evaporation does not take
place within (or on the surface of) the membrane, this approach relies on passive transport of vapor across the channels and then
across the membrane, which can be inefficient and incomplete.
The lack of comparable ‘‘hydrophobic’’ membranes, furthermore,
makes this approach difficult to apply for well wetting dielectric
fluids.
In another recent study, Narayanan et al.[14]proposed using a
‘‘nano-porous patch’’, where a thin porous membrane is used to
separate the vapor from a thin evaporating liquid layer of a controlled thickness. A gas jet was used to assist the removal of the vapor generated. Both modeling and experiments were conducted.
Modeling results showed that the heat transfer rates increase with
decrease in liquid film thickness and/or membrane thickness. Increase in the air flow rate also increases the heat transfer rate. With
liquid and membrane thicknesses of 1lm, heat fluxes of 700 W/
cm
2
were possible with FC-72 as the test liquid. Results for water
were consistently lower by factors of 2.2–3.
To validate the model results, experiments were conducted
with water as the test fluid. In these experiments, liquid film
thicknesses of 200 and 400lm were tested with a 60lm thick
membrane with 20 nm pores (porosity = 50%). Since the location
of the liquid–vapor interface is a key input parameter in the
model, and because it could not be measured in the experiments, two extreme cases were considered – one where the liquid completely fills the membrane (i.e., evaporation occurs at
the membrane outlet) and the other where the membrane is
completely filled with vapor (i.e., evaporation occurs at the
membrane inlet). Experiments show that compared to singlephase air jet cooling, evaporative cooling enhance heat transfer
by a factor of 2. For surface temperature above 70 C, the experimental results were in agreement with model predictions. At
lower wall temperature, the models underpredict the experimental results. This discrepancy was attributed to leakage through
the membrane.
The results from[14]show that as high heat flux removal can be
achieved with evaporative cooling. Moreover, with the porous
membrane completely filled with liquid, the resistance to mass
transfer is minimized resulting in the highest heat transfer rates.
The ‘‘horizontal’’ porous patch cooling device implemented in
[14], however, suffered from important limitations. The liquid
layer defined by the porous membrane has to be kept very thin
( 1lm) because the thermal energy needs to be first conducted
across this liquid layer. Such a thin liquid layer, however, presents
significant hydraulic resistance to the flow parallel to the membrane, leading to an excessive pressure drop. The horizontal configuration of the cooling device also limits the heat transfer surface
area to only that of the exposed heat source.
In this article, we propose a novel two-phase microchannel
cooling device with perforated side walls. This device helps circumvent the limitations of the previous cooling device by decoupling the main heat conduction paths (along the microchannel
walls) from the liquid supply paths (along dedicated ‘‘wide’’ liquid
channels) and by taking advantage of large extended surfaces
available in high-aspect-ratio microchannel walls for evaporation
heat transfer. We provide details of one practical implementation
of the device concept and thermofluid modeling results.
The cooling device proposed in this study involves a situation
very similar to that studied in[14], where the porous side wall is
completely filled with liquid and evaporation occurring at the vapor channel. Subsequently, mass transfer resistance is expected
to be negligible. The results from[13,14]show that the inclusion
of a porous membrane to separate the liquid and vapor flow paths
improves the heat transfer and pressure drop characteristics of
cooling devices that involve boiling or evaporation. An additional
benefit of separating the two phases is the mitigation of instabilities related to the pressure drop. Due to the differences in the
geometry and the operating conditions, it is not possible to compare the data obtained from these studies to the numerical simulation results obtained in this study. However, the modeling
methodology adopted and the performance of the devices in the
two studies discussed above shows that the concept of our high
heat flux cooling device and the thermofluid modeling approach
used are appropriate.
2. Cooling device concept
We propose a novel embedded evaporative cooling device
based on phase-separated microchannels with perforated (‘‘perspiring’’) side walls for the thermal management of ultra-high heat
flux devices. A schematic of this device is shown inFig. 1. In this
cooling device, each liquid channel is separated from the two adjacent vapor channels by side walls that contain arrays of micro perforations. Liquid flowing in the liquid channels is transported
through the perforations by capillary action and evaporate into
the vapor channels to efficiently remove heat. Conceptually, our
approach may loosely be viewed as a geometric adaptation of transpiration cooling systems investigated to tackle demanding thermal management challenges of hypersonic vehicles. Our unique
perforated channel walls of high aspect ratio offer dual benefits
by serving both as extended heat transfer surfaces (fins) and as micro-porous membranes for phase separation.
We select silicon as our channel wall material in part to take
advantage of its unique micromachinability, such as photo-electrochemical (PEC) etching and deep reactive ion etching (DRIE). This
also helps ensure that our ‘‘embedded’’ cooling solution is compatible with recent trends towards heterogeneous integration of highpower-density circuits built on SiC or diamond substrates with silicon-based CMOS circuits. Examples of the former may include
GaN-based microwave or RF power amplifiers and other power
switching/control devices, and high-performance microprocessors.
Major R&D efforts are currently under way to prepare GaN[15]devices on diamond substrates. The relatively small mismatch in
CTEs between Si and SiC and between Si and diamond helps reduce
thermal stress on the microchannels during thermal cycling.
The microfabrication processes considered for our cooling device can be scaled up to create multiple microchannel groups on
a single substrate to cover large areas. Alternatively, multiple cooling devices may be placed side to side to handle an array of chips
over a large area.
3. Coolant selection and its thermophysical properties
The peak junction temperature specified for wide-bandgap GaN
devices is typically over 200 C. Previous modeling and depth-resolved Raman measurements showed, however, that a large portion of the temperature drop occurs across the device buffer
layers and interfaces rather than the substrate [16]. With a target
top ‘‘substrate’’ temperature of 90 C, we choose as our working
fluid Fluorinert™ FC-72 with a boiling point of 56 C (at 1 atm)
and with material compatibility superior to hydrofluoroethers
(HFEs). Different Fluorinert fluids with higher boiling points (e.g.,
FC-77 at 97 C) may be used instead if higher condenser temperatures are desired.
Heterogeneous bubble nucleation generally occurs at microscopic cavities that entrap gas[17]and the wall superheat required