Purdue University
Purdue e-Pubs
CTRC Research Publications Cooling Technologies Research Center
2016
A Method for %ermal Performance
Characterization of Ultra-%in Vapor Chambers
Cooled by Natural Convection
G. Patankar
S. Mancin
J. A. Weibel
Purdue University, jaweibel@purdue.edu
M. A. McDonald
S V. Garimella
Purdue University, sureshg@purdue.edu
Follow this and additional works at: h=p://docs.lib.purdue.edu/coolingpubs
<is document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for
additional information.
Patankar, G.; Mancin, S.; Weibel, J. A.; McDonald, M. A.; and Garimella, S V., "A Method for <ermal Performance Characterization of
Ultra-<in Vapor Chambers Cooled by Natural Convection" (2016). CTRC Research Publications. Paper 301.
h=p://dx.doi.org/10.1115/1.4032345
Gaurav Patankar
Cooling Technologies Research Center,
an NSF I/UCRC,
School of Mechanical Engineering,
Purdue University,
585 Purdue Mall,
West Lafayette, IN 47907
e-mail: gpatank@purdue.edu
Simone Mancin
Cooling Technologies Research Center,
an NSF I/UCRC,
School of Mechanical Engineering,
Purdue University,
585 Purdue Mall,
West Lafayette, IN 47907
e-mail: simone.mancin@unipd.it
Justin A. Weibel
Cooling Technologies Research Center,
an NSF I/UCRC,
School of Mechanical Engineering,
Purdue University,
585 Purdue Mall,
West Lafayette, IN 47907
e-mail: jaweibel@purdue.edu
Suresh V. Garimella
1
Cooling Technologies Research Center,
an NSF I/UCRC,
School of Mechanical Engineering,
Purdue University,
585 Purdue Mall,
West Lafayette, IN 47907
e-mail: garimell@purdue.edu
Mark A. MacDonald
Intel Corporation,
5200 Elam Young Pkwy,
Hillsboro, OR 97124
e-mail: mark.macdonald@intel.com
A Method for Thermal
Performance Characterization
of Ultrathin Vapor Chambers
Cooled by Natural Convection
Vapor chamber technologies offer an attractive approach for passive cooling in portable
electronic devices. Due to the market trends in device power consumption and thickness,
vapor chamber effectiveness must be compared with alternative heat spreading materials
at ultrathin form factors and low heat dissipation rates. A test facility is developed to
experimentally characterize performance and analyze the behavior of ultrathin vapor
chambers that must reject heat to the ambient via natural convection. The evaporator-
side and ambient temperatures are measured directly; the condenser-side surface temper-
ature distribution, which has critical ergonomics implications, is measured using an
infrared (IR) camera calibrated pixel-by-pixel over the field of view and operating tem-
perature range. The high thermal resistance imposed by natural convection in the vapor
chamber heat dissipation pathway requires accurate prediction of the parasitic heat
losses from the test facility using a combined experimental and numerical calibration
procedure. Solid metal heat spreaders of known thermal conductivity are first tested, and
the temperature distribution is reproduced using a numerical model for conduction in the
heat spreader and thermal insulation by iteratively adjusting the external boundary con-
ditions. A regression expression for the heat loss is developed as a function of measured
operating conditions using the numerical model. A sample vapor chamber is tested for
heat inputs below 2.5 W. Performance metrics are developed to characterize heat
spreader performance in terms of the effective thermal resistance and the condenser-side
temperature uniformity. The study offers a rigorous approach for testing and analysis of
new vapor chamber designs, with accurate characterization of their performance relative
to other heat spreaders. [DOI: 10.1115/1.4032345]
Introduction
A heat pipe or vapor chamber can passively transport heat from
a localized generation source to a diffuse heat rejection surface at
a low temperature gradient. There is an extensive body of research
on the investigation of heat pipes and vapor chambers for the ther-
mal management of electronics [1]. A vapor chamber is used to
mitigate the temperature rise of sensitive components by spread-
ing heat away from local hot spots. The sealed vapor chamber
contains a working fluid, and vapor is generated at the evaporator
section located over the hot spot. The vapor is driven outward and
condenses on the inner surface of the opposing wall. A porous
wick passively pumps the condensed liquid back to the evaporator
(Fig. 1). Portable electronic device platforms such as smartphones
and tablets are trending toward thinner, compact designs with
more embedded functionality (and thereby more waste heat gener-
ation from active components). Due to power consumption and
size constraints, it is not practical to use active air cooling meth-
ods, or heat rejection surfaces with large area enhancement, to dis-
sipate heat. In such instances, ultrathin vapor chambers may offer
a viable solution.
Prior studies have experimentally assessed the performance of
heat pipes and vapor chambers using a standard testing approach
in which the heat is spread from a heat source to a cold plate or
air-cooled heat sink, and the heater-to-ambient thermal resistance
is determined [2]. The transport behavior of internal components
of a vapor chamber, such as the effective thermal resistance across
Fig. 1 Schematic diagram of vapor chamber operation
1
Corresponding author.
Contributed by the Electronic and Photonic Packaging Division of ASME for
publication in the J
OURNAL OF ELECTRONIC PACKAGING. Manuscript received
September 12, 2015; final manuscript received November 16, 2015; published online
March 10, 2016. Assoc. Editor: Ashish Gupta.
Journal of Electronic Packaging MARCH 2016, Vol. 138 / 010903-1Copyright
V
C
2016 by ASME
Downloaded From: http://electronicpackaging.asmedigitalcollection.asme.org/ on 03/16/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
the evaporator wick [3,4], can also be assessed. In a testing config-
uration closely related to the current study, Wang et al. [5] investi-
gated copper vapor chambers that contained interlaced grooves
and channels as the wick structure. The hot spot heat input was
supplied through a copper platen with an embedded rake of ther-
mocouples for heat flux measurement. Heat was rejected on the
condenser side using a finned heat sink cooled by forced air con-
vection. Thermal grease was used to reduce contact resistance
between the vapor chamber and the heat sink; thermocouples
were embedded in this grease layer between the heat sink and the
vapor chamber to measure surface temperatures. The thermal per-
formance of the vapor chamber was assessed based on its thermal
resistance and condenser-side temperature uniformity [5]. This
vapor chamber testing configuration has commonly been used
[6–9]. Other variants use a liquid-cooled cold plate on the con-
denser side [10]. This testing configuration is tailored for high-
power or high-density cooling applications for which the heat flux
induces a large temperature gradient in the heater platen that can
be accurately measured, and where the condenser-side heat rejec-
tion method mimics the intended application. For example,
Mochizuki et al. [9] tested input heat fluxes from 20 to 100 W
over 1 cm
2
, Wong et al. [6] from 300 to 400 W over 1–4 cm
2
,
and Chen et al. [8] from 20 to 80 W over 2cm
2
. The lowest
reported heat fluxes investigated under this vapor chamber testing
configuration include measurements by Koito et al. [11] from 16
to 32 W/cm
2
and Wang et al. [5] from 4 to 10 W/cm
2
.
In portable electronic platforms, heat is rejected to the ambient
via natural convection directly from the device surface (typically
a smooth, flat surface that does not incorporate a finned heat sink
due to lack of space and the low operating power). Hence, it is
necessary to develop good characterization techniques for vapor
chambers cooled by natural convection. To the authors’ knowl-
edge, experimental investigation of vapor chambers operating at a
low power density where the condenser-side boundary condition
is one of natural convection has not been considered in the litera-
ture. Also, due to the proximity of the vapor chamber condenser
surface to the device skin in thin form factor platforms, an assess-
ment of the condenser-side surface temperature distribution is
extremely important. This product sector necessitates a paradigm
shift in thermal management, where the external surface tempera-
ture threshold is dictated by user considerations, rather than by
device operating temperature limits. According to Moritz and
Henriques [12], roughly 5 mins of contact with temperatures of
50
C can cause skin tissue damage; to avoid this condition, per-
formance throttling would be dictated by user comfort standards.
Berhe [13] defined ergonomic temperature limits on handheld
devices of 41
C for aluminum surfaces and 43
C for plastic
surfaces. It is clear from this review that existing metrology
approaches stress a characterization of the total thermal resistance
of the vapor chamber, while few studies analyze the condenser-
side temperature distribution. A rigorous mapping of the surface
temperature distribution is necessary to characterize vapor cham-
ber performance for portable electronics applications.
This paper presents an approach for characterizing the perform-
ance of ultrathin vapor chambers for portable electronics plat-
forms operating at low power. An experimental test facility is
developed that subjects the vapor chamber to a hot spot on the
evaporator side and rejects heat from the condenser side by natu-
ral convection. Precise evaluation of performance at very low
power densities (1 W/cm
2
) is enabled by a combined experimen-
tal and numerical approach for calibration of the heat transport
through the vapor chamber. Keeping the heat source at the mini-
mum possible temperature and mitigating hotspots on the con-
denser surface are key functional requirements. Hence, in addition
to the conventional thermal resistance metric, IR measurement of
the condenser-side surface temperature allows characterization of
the vapor chamber performance in terms of temperature distribu-
tion. The assessment is based on the performance of an ultrathin
vapor chamber relative to a solid heat spreader with identical
outer dimensions.
Experimental Facility
An experimental facility is developed to evaluate the perform-
ance of ultrathin vapor chambers at low heat loads. The intrinsic
challenge in vapor chamber characterization under such condi-
tions is estimation of the percentage of heat input rejected through
the vapor chamber versus parasitic losses through other pathways.
To measure extremely low heat loads, a test section is typically
designed to eliminate heat losses (an isolated system, e.g.,
Ref. [2]). To evaluate performance of a vapor chamber rejecting
heat to the ambient via natural convection, which inserts a large
associated thermal resistance in the primary heat rejection path-
way, it is difficult to create a sufficiently isolated system. An alter-
native approach is to control the heat losses in a manner that
allows for accurate estimation and calibration, as implemented in
the current study.
Test Section Design and Instrumentation. A schematic dia-
gram of the test section configuration is shown in Fig. 2. The test
section is comprised of the heat spreader sample, with insulation
and a centered heater block on the underside; the top side of the
heat spreader is exposed to ambient air. The test section insulation
is made of PEEK (k ¼ 0.25 W m
1
K
1
) with outer dimensions of
150 mm 115 mm 25.4 mm. A 92 mm 52 mm 0.7 mm deep
recess milled into the top surface of the insulation seats the heat
spreader sample. In the center, a 10 mm 10 mm square pocket
was machined to insert the heater block assembly. As shown in the
inset of Fig. 2, the hot spot heat input is simulated using a
10 mm 10 mm thin-film polyimide heater attached using thermally
conductive paste to the base of a 10 mm copper heater block that
ensures uniform distribution of the heat load imposed on the
spreader. A uniform, thin layer of high-conductivity epoxy was
applied onto the bottom surface of the heat spreader to cover the area
overlapping the copper heater block. This allowed a consistent joint
to be formed between the heat spreader and heater block across all
samples to yield consistent calibration. The top surface of the vapor
chamber is cooled by natural convection to the surrounding air.
A photograph of the experimental facility is shown in Fig. 3.A
sample is shown inserted into the test section, and the auxiliary
components for temperature and power measurements are visible.
The spatial temperature distribution on the top surface of the
heat spreader is measured by a mid-wave IR camera (Indigo
Fig. 2 Schematic diagram of the test section (top inset shows
the heater block assembly)
010903-2 / Vol. 138, MARCH 2016 Transactions of the ASME
Downloaded From: http://electronicpackaging.asmedigitalcollection.asme.org/ on 03/16/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Merlin MID, FLIR) positioned above the test section. Visualization
of the surface temperature via an IR camera allows for the develop-
ment of performance metrics based on the surface temperature distri-
bution. Calibration of the IR camera measurement was carried out
using a reference black body (Blackbody Source Model 2004, SBIR)
with a known emissivity (0.985 6 0.015). The temperature of the
black body was incremented in steps of 5
C from 20
Cto100
C.
From the recorded images, a pixel-by-pixel calibration of the surface
temperature versus sensor output was performed. A sixth-o rder poly-
nomial curve was fit to the data to obtain a correlation between sensor
output and temperature. Subsequent evaluation of the calibration at
selected blackbody reference temperatures in the 20
Cto100
C
range showed a maximum mean error averaged over the field of view
of 0.2
C. The top surface of the heat spreader sample is painted black
(#1602, Krylon) to impart a known emissivity of 0.96 [14]; the ratio
of the cal ibration black body emissivity to the surface emissivity is
used to correct the IR temperature measurement.
As shown in Fig. 2, a total of 30 thermocouples are embedded
throughout the insulation block to monitor the temperature. In par-
ticular, the thermocouple locations are classified into groupings of
those embedded under the top, side, and bottom surfaces (four
each) and in the middle of the insulation block (nine). Eight ther-
mocouples are placed in grooves along the surface in contact with
the bottom of the heat spreader. Finally, a thermocouple is inserted
at the center of the copper heater block to measure the junction
temperature. This deployment of thermocouples is essential to the
calibration procedure used for estimation of the heat loss from the
insulation block, as described in the following section. Each ther-
mocouple was individually wired to a reference junction that is
placed in a dry-block ice point reference (TRCIII, Omega). The
thermocouples were individually calibrated using a thermostatic
oil bath (7103 Micro-Bath, Fluke) and two factory-calibrated
resistance temperature detectors (RTD, 6 0.1 K), one each for the
ice point and the oil bath. Following calibration, the thermocouple
temperature measurements have an absolute uncertainty of
6 0.3 K. The ambient temperature is measured using an RTD.
The electrical power supplied to the film heater is determined
by measuring the voltage drop across the resistance heating ele-
ment and across a shunt resistance placed in series with the film
heater. The electrical input power has a measured uncertainty of
0.2% (governed by the shunt resistance uncertainty).
Test Procedure. A strict experimental procedure is followed
for all tests to ensure repeatability of the measurements. The IR
camera is switched on at least 1 hr prior to starting the test to
ensure that the sensor cools down to a steady temperature for
reduced noise in the images. Boards are placed around the test
section so as to prevent air flow disturbances in the surrounding
ambient. To acquire each data point, the electrical power input to
the heater is set at the desired value; all the monitored data are
recorded every 4 s using an NI cDAQ 9178 data acquisition chas-
sis with NI 9124 thermocouple, NI 9217 RTD, and NI 9205 volt-
age input modules. Active data processing is performed in a
LABVIEW interface to determine when steady-state conditions have
been reached, defined as when the standard deviation of the junc-
tion temperature for the last 150 data points is less than 0.02 K.
The time usually taken to reach steady-state conditions is approxi-
mately 3 hrs. After steady-state conditions are reached, the per-
formance is monitored for an extended period ( 30 min) to
obtain a large steady-state data set; IR images are acquired at
5-min intervals during this period. Due to the small fluctuations in
ambient temperature that affect the test section temperatures at
steady state, a set of 150 data points is selected from the steady-
state data set which has the lowest standard deviation in junction
temperature. An average over these data is used for subsequent
analysis, and associated with the specific steady-state IR image
taken during this interval. This procedure for acquiring a single
data point is repeated for each heat input power.
Calibration of the Test-Section Heat Loss
A calibration procedure is implemented that predicts heat loss
from the test section. The experimental step of the calibration pro-
cedure evaluates heat spreading in two thin metal plates of known
thermal conductivity, viz., copper and aluminum. The test section
temperatures were recorded for heat loads in the nominal range of
0.15–4 W. Key characteristics of the metal heat spreaders used for
the calibration process are listed in Table 1.
A numerical model of the test section is generated to simulate
conduction in the heater block assembly, insulation block, and
heat spreader. As shown in Fig. 4, the model boundary conditions
have a constant heat flux applied at the base of the heater block, a
thermal resistance at the interface between the insulation and heat
Table 1 Heat-loss calibration data set
Copper Aluminum
Thermal conductivity (W m
1
K
1
) 387.6 202.4
Outer dimensions (mm) 90 55 0.7 90 51 0.635
Electrical heat input (W) 0.17–4.16 0.16–3.88
# of data points 8 10
Fig. 3 Photograph of the experimental facility
Fig. 4 Exploded view of the numerical conduction model do-
main and boundary conditions
Journal of Electronic Packaging MARCH 2016, Vol. 138 / 010903-3
Downloaded From: http://electronicpackaging.asmedigitalcollection.asme.org/ on 03/16/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
spreader, and convection coefficients on each external surface. A
grid-independent rectangular mesh is used to discretize the geom-
etry using a total of 325,000 cells. The peripheral regions of the
insulation block have a uniform coarse mesh; the solution is
insensitive to further refinements because of the low temperature
gradients in these regions. The copper heater block assembly has
a finer mesh. In the heat spreader, a gradient-based mesh is used
for refinement near the hotspot. The lateral cell lengths increase in
the outward direction from 0.25 mm to 2.25 mm. There are 20
cells across the thickness of the heat spreader near the hotspot.
The properties of the heat spreader are specified for the sample
being tested according to Table 1. The governing energy equation
is solved using the finite-volume method implemented in the com-
mercial software
ANSYS FLUENT [15].
The primary objective of the numerical model is to predict the
boundary conditions and overall heat losses that cannot be deter-
mined directly from the available experimental data. A formal
procedure is implemented in order to iterate on the boundary con-
ditions in the model in order to produce good agreement between
the experimental and numerical values of temperature at the loca-
tions in the test section measured by thermocouples. For each cali-
bration data point, the free variables in the numerical simulation
are the heat transfer coefficients on the top, side, and bottom
surfaces of the insulation block and the heat spreader top surface.
Tuning of the thermal resistance at the interface between the insu-
lation and heat spreader to a fixed value of 0.02 m
2
KW
1
across
all test cases yielded the best agreement with experimental data
(equivalent to an air gap thickness of 0.5 mm).
For the initial guess value, a prediction of the natural convec-
tion heat transfer coefficient at each surface with a different orien-
tation obtained from standard correlations was imposed, and then
was subsequently iterated to generate a match with the thermocou-
ple data. Priority was given to first match the junction temperature
closest to the heat source, and then finer adjustments to the bound-
ary heat transfer coefficients (increments of 0.5 W m
2
K
1
)were
made to minimize the overall average deviation from the experi-
mental temperature data. Simple rules were applied that ensure the
heat transfer coefficients increased from the downward to upward
facing surfaces according to the physical behavior expected. With a
sufficient match to the experimental data, the heat transported
through the heat spreaders and the heat loss through the insulation
block can be easily extracted from the numerical data.
The values of the external heat transfer coefficients were found
to be in the range from 4 to 16 W m
2
K
1
. In the current study,
where low heat loads are applied and the overall heat loss is a sig-
nificant percentage of the overall heat input, a single value for the
heat transfer coefficient on all exposed surfaces did not yield suffi-
cient accuracy in the match between experimental and numerical
temperatures. When these values were allowed to independently
vary, the temperature mismatch between the measured and com-
puted values was significantly decreased for all test cases.
The temperature mismatch between the test and the simulation,
averaged over all the cases, is 0.34 K, with a standard deviation of
0.56 K. Figure 5 shows a comparison between the simulated tem-
peratures after iterating on the boundary conditions compared
with the measured values for a selected copper spreader test case.
The thermocouple groups (as discussed under “Test Section Design
and Instrumentation” section) are on the bottom, side, and top
surfaces of the insulation block, inside the insulation block
(internal), embedded below the heat spreader, and at the junction.
For the selected case, the junction temperature is matched most
closely (difference of 0.02 K); the maximum difference is observed
for the heat spreader group of thermocouples (difference of 0.71 K).
Using this calibration procedure, the uncertainty in the eval-
uated heat loss from the test section is roughly estimated based on
both the resolution of the heat transfer coefficient increments used
during the iteration process and the ultimate temperature mis-
match at the surface-embedded thermocouple locations. Using
these component uncertainty values for each case, and expressing
the heat loss as a single equation of the form
Q
loss
¼ h A ðT T
amb
Þ (1)
a standard propagation of errors can be used estimate the uncer-
tainty in the predicted heat loss; this uncertainty varies from 3%
to 14% of the calculated heat loss for the test cases described in
Table 1.
The heat loss values extracted from the calibration of the cop-
per and aluminum heat spreaders are plotted in Fig. 6(a). A gener-
alized regression is developed for the heat loss value as a function
of the electrical input power and the junction-to-ambient tempera-
ture difference, as given by
Q
loss
¼ a
T
j
T
amb
Q
ele
b
"#
Q
ele
(2)
This form of the equation assumes that the heat loss value is pro-
portional to the electrical input power (and that there is no heat
loss at zero input power). This relationship can be clearly observed
in Fig. 6(a) (dashed lines indicate best linear fit to the data points).
The proportionality constant would then depend on the thermal re-
sistance of the sample being tested. The ratio of junction-to-
ambient temperature difference and the electrical input reflects this
thermal resistance. By evaluating the thermal resistance of both the
copper and aluminum heat spreaders, as shown in Fig. 6(b),the
influence of the junction-to-ambient temperature on the overall
heat loss can be incorporated into the regression. The result of the
calibration yields the constants a ¼ 0.14 and b ¼ 0.57, which can
subsequently be used to calculate the heat losses through the insula-
tion block when evaluating heat spreading devices that have an
unknown thermal resistance and heat spreading behavior. The val-
ues of these constants are specific to the current test section design;
a similar calibration procedure would need to be employed with
any change in the experimental setup.
Results and Performance Metrics
A representative vapor chamber device with outer dimensions
of 90 mm 50 mm 0.8 mm, obtained from a commercial ven-
dor, is characterized to demonstrate the testing approach devel-
oped. The copper vapor chamber has 0.2-mm-thick copper walls,
uses water as the working fluid, and is lined with a single layer of
copper mesh (pore sizes of approximately 50–100 lm). The heat
spreading behavior of the vapor chamber is evaluated for 12
device power levels (electrical heat input minus losses) ranging
from 0.4 to 2.2 W, and resulting in vapor chamber area-weighted
mean condenser-side surface temperatures from 24.2
Cto
50.3
C, and maximum condenser-side surface temperatures in the
range of 32.8
C to 55.9
C. During testing, the ambient air
Fig. 5 Comparison of thermocouple temperatures obtained
from experiments against those from the simulations at an elec-
trical heat input of 1 W and ambient temperature of 298.2 K.
Each bar is an average temperature from each grouping of
thermocouples.
010903-4 / Vol. 138, MARCH 2016 Transactions of the ASME
Downloaded From: http://electronicpackaging.asmedigitalcollection.asme.org/ on 03/16/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use