Professional Documents
Culture Documents
The Performance of A Novel Impinging Two-Phase Jet Heat Sink Oliveira2017
The Performance of A Novel Impinging Two-Phase Jet Heat Sink Oliveira2017
1 Introduction them. Additional design benefits associated with this approach are
the reduction of weight and volume of the cooling system [8].
Throughout the years, ground breaking advances in design,
The viability of a direct liquid cooling configuration is highly
microfabrication, and performance of electronic systems (e.g., com-
dependent on its ability to produce very high convective heat
puter technology and power modules) have been accompanied by
transfer coefficients in order to compensate for the inferior
higher rates of heat dissipation (and temperature rise), which are a
thermophysical properties of dielectric coolants. Two-phase cool-
challenging obstacle for technological innovation and development
ing schemes, such as sprays, impinging jets, and boiling in micro-
[1]. According to the 2015 International Electronics Manufacturing
channels, have been attracting the attention of the research
Initiative (iNEMI) roadmaps, contemporary microprocessors for
community for a number of years due to their enhanced heat trans-
desktop machines are designed to operate above 150 W, server and
fer characteristics [1,7,9–15].
computer cluster chips are likely to exceed power dissipation of
Mechanical vapor compression refrigeration can lower the
500 W per chip, and the power dissipation of chips in automotive
component (e.g., junction) temperature to a value inferior to that
applications, whose dimensions are typically smaller than desktop
of its surroundings. The integration of vapor compression systems
computer and servers, is reaching 300 W with heat fluxes of
with two-phase enhanced heat transfer schemes (jets, sprays,
200–300 W/cm2 [2]. Besides, the demand for reducing the size of
microchannels) has proven to be an innovative and effective way
insulated gate bipolar transistor power modules for hybrid and elec-
of removing highly concentrated heat loads [16–20].
tric vehicles was expected to increase the volumetric power rating
Jet impingement cooling has been extensively studied over the
in 10–50 times compared to the larger modules commonly used in
past two decades. The effects of several parameters such as nozzle
railway applications [3].
shape and diameter, impingement height, impact velocity, nozzle
Several cooling technologies have been studied with a focus on
pitch distance, jet array, subcooling degree, surface structure and
advanced military avionics and high-performance computers and
aging, cross flow and drainage conditions on the fluid dynamics,
servers [4–6] since novel efficient alternatives for high heat flux
and heat transfer of single- and two-phase impinging jets have
removal are required in these applications. Among the existing
been widely reported [21–28]. The use of multiple impinging jets
thermal management solutions, direct liquid cooling [7] enables
has been proposed in many studies. However, this cooling princi-
the contact between a dielectric liquid coolant and the electronic
ple has not yet been integrated into mechanical vapor compression
components taking advantage of the complete elimination of the
refrigeration loops [29–40]. Impinging jets have also been com-
thermal resistances of the different layers of materials separating
bined with microchannels aiming to create high-performance
hybrid cooling schemes that benefit from the attributes of both
enhanced heat transfer techniques [41–46]. A few works have pre-
1
Corresponding author. sented mechanical vapor compression systems that use thermal
Contributed by the Electronic and Photonic Packaging Division of ASME for
publication in the JOURNAL OF ELECTRONIC PACKAGING. Manuscript received January
devices based on spray cooling, but were not purposely designed
25, 2017; final manuscript received April 24, 2017; published online June 28, 2017. to be compact or miniaturized [47–53]. Oil lubricated compres-
Assoc. Editor: Dong Liu. sors [48,51–53], ancillary expansion valves [47–49,51], and tube
Fig. 1 Schematic diagram of the experimental apparatus: (1) compressor inlet (suction),
(2) compressor outlet (discharge), (3) condenser inlet (refrigerant side), (4) condenser
outlet (refrigerant side), (5) jet cooler inlet, (6) jet cooler outlet, (7) condenser inlet (WEG
side), (8) condenser outlet (WEG side), and (9) secondary evaporator inlet (not used in
this study)
Fig. 3 Orifice plenum: (a) assembled components and (b) outer and inner orifice plates
array configurations can be tested with the same orifice plenum. by a Teflon base plate. This arrangement guarantees a good ther-
As the subcooled liquid passes through the nozzle, it expands gen- mal contact between the heater and the block. The thermal load
erating a downward-oriented jet inside a polycarbonate jet cham- was finely controlled by a DC digital power supply (Agilent
ber. The jet length, i.e., the orifice-to-heater distance, is defined as N5770A). During the experimental runs, the thermal load was
the vertical distance from the outlet of the orifice to the base of augmented in steps by increasing the voltage or the current pro-
the chamber, which is at the same level as the impingement vided by the power supply.
surface. The copper block has six RTD wells, as shown in Fig. 6. Five
The impingement surface is the top surface of a heated cylindri- RTDs are used to measure the temperature and allow for an estimate
cal copper block mounted inside a polycarbonate bottom piece of the surface temperature to be made using Fourier’s law. The
(the so-called insulation and drainage unit). This unit was spe- remaining RTD is connected to a commercial proportional–integra-
cially designed to thermally insulate the lateral surface and the l–derivative controller that functions as a safety system cutting the
bottom of the copper block and facilitate drainage of the two- heating power supply in the event of a temperature runaway associ-
phase mixture from the jet chamber into the reservoir just below ated with the critical heat flux. In the present experiments, the cutoff
it. The assembly procedure of the copper block in the unit is temperature was set at 68 C. The area of the target surface is 6.36
shown in Fig. 5. The small gap between the copper block and the cm2 (D ¼ 28.54 mm) and the distance from the plane of the RTDs to
insulation and drainage unit is sealed by two O-rings (the grooves the impingement surface is 10.03 mm [54,55].
are shown in Fig. 6). This prevented the flow of refrigerant into The copper block was designed so that its diameter is circum-
the skin heater cavity. The heat load is provided by a 200 W thin scribed the skin heater surface area (20 mm 20 mm), as shown
film heater pressed firmly against the bottom of the copper block in Fig. 5(a). Numerical simulations of the heat transfer in the
Fig. 6 (a) Copper block and (b) cross-sectional view of the copper block (RTD’s plane, dimen-
sions in millimeters)
copper block/insulation and drainage unit (the skin heater was subcooling degree at the outlet of the condenser, vapor mass qual-
modeled and implemented as a source term) revealed that the dif- ity at the outlet of the two-phase jet cooler, heat transfer rates, and
ference between the power dissipated at the skin heater and the compressor power. Based on the output variables, the jet impinge-
heat transfer rate effectively dissipated at the jet impingement sur- ment heat transfer coefficient can be calculated.
face (top surface of the copper block) was lower than 5%. Further All output data were recorded during a 15 min time interval,
details regarding this analysis can be found in Ref. [54]. Ds, after guaranteeing that the following steady-state criteria were
simultaneously satisfied:
2.3 Experimental Conditions and Procedure. The experi- sðXÞjDs jUðXÞj (1)
mental tests were run at the maximum compressor displacement.
The diameter of the orifices was 300 lm and the orifice-to-heater
distance (jet length) was 28.84 mm. The secondary fluid tempera- @X
max Ds jsð XÞ (2)
ture was set at 25 C in all tests. The refrigerant charge in the sys- @s Ds
tem was adjusted so that the tests for each multiple jet array had
the same refrigerant mass flow rate at the lowest thermal load where sðXÞ is the standard deviation of a generic output variable
(75 W). The mass flow rate of the WEG mixture was kept fixed X and U expresses its overall experimental uncertainty. j is an
at 180 kg/h. This value was sufficient to produce a high overall integer corresponding to each one of the three divisions of the
thermal conductance in the condenser, thus guaranteeing large Gaussian probability distribution, i.e., j ¼ 1, 2 or 3. The left-hand
condenser heat transfer rates (up to 301 W). The room temperature side of Eq. (2) was estimated numerically by a second-order cen-
and the calorimeter temperature were set at 25 C. The compressor tered finite difference scheme.
inlet superheat, defined as DTsh ¼ T1 Tevap , was kept fixed at Before recording the data, Eqs. (1) and (2) were verified during
10 C. three consecutive time intervals, Ds, to ascertain that the steady-
The influence of the following variables was investigated: (i) state was actually reached. The output data were recorded only
applied thermal load and (ii) orifice array configuration (for the when the left-hand side of Eq. (2) was lower than 2sðXÞ. In the
tests with multiple jets). Figure 7 presents the single and the mul- following, three distinct cases are presented as numerical exam-
tiple jet arrays explored, which are different combinations of a ples of the computation procedure. Each case considers a different
five-orifice configuration. For the single-jet test, the orifice screw output variable, i.e., mass flow rate, temperature, and pressure, for
was positioned at the center of the orifice plate assembly and the single-jet, 75-W test.
dummy screws occupied the remaining positions. Case 1: Mass flow rate: Considering that sðm_ r ÞjDs ¼ 0:006 kg/h,
All experimental runs were carried out in the heating-up mode, Uðm_ r Þ ¼ 60:05 kg/h, ðmaxð@ m_ r =@sÞÞjDs Ds maxðDm_ r =2Þ ¼
i.e., increasing the heat load until the critical heat flux was 0:003 kg/h.
reached, which explains the different number of tests for the Case 2: Temperature: Considering that sðT2 ÞjDs ¼ 0:16 C,
single- and multiple-jet configurations. The critical heat flux was UðT2 Þ ¼ 60:16 C, ðmaxð@T2 =@sÞÞjDs DsmaxðDT2 =2Þ¼0:15 C.
characterized experimentally by a sudden increase of the heater Case 3: Pressure: Considering that sðP2 ÞjDs ¼ 0:016 bar, UðP2 Þ
surface temperature and was confirmed by visual observations to ¼ 60:05 bar, ðmaxð@P2 =@sÞÞjDs Ds maxðDP2 =2Þ ¼ 0:013 bar.
coincide with a partial or total dryout of the liquid film on the
heater surface.
The output variables of the experimental apparatus are the pres- 2.4 Dependent Variables. The dependent variables were cal-
sures, temperatures, refrigerant mass flow rate, refrigerant culated based on the experimental measurements, as described in
Q_ c (W) 2.01
Q_ sh ¼ m_ r ðh1 h6 Þ (4) W_ comp (W) 4.53
COP 0.49
where m_ r is the refrigerant mass flow rate and the refrigerant x6 0.013
enthalpy at the compressor inlet, h1 , condenser inlet, h3 , and con- hs (W/m2 K)
433.0
denser outlet, h4 , were computed via REFPROP 8.0 [57] using the Ts ( C) 0.11
local experimental values of pressure and temperature. Tcond ( C) 0.09
Tevap ( C) 0.26
The refrigerant enthalpy at the outlet of the two-phase jet DTsh ( C) 0.31
cooler, h6 , was calculated from an energy balance on the cooler, DTsc ( C) 0.17
neglecting the kinetic and potential energy contributions. Thus DP (bar) 0.05
Q_ c
h6 ¼ h5 þ (5)
m_ r
where Q_ c is the cooling capacity and the refrigerant The temperature of the surface of the copper block, Ts , was
specific enthalpy at the inlet of the jet cooler, h5 , was also determined through a linear extrapolation of Fourier’s law consid-
determined from local measurements of pressure and temperature. ering one-diimensional heat conduction in the axial direction
At steady-state, the cooling capacity is equal to the thermal load
provided by the skin heater, considering a perfectly insulated test LQ_ c
section. Ts ¼ TRTD (9)
ks As
The vapor mass quality at the exit of the jet cooler, x6 , was
determined from h6 and the measured outlet pressure, P6 . The sat- where L is the distance between the copper block surface and the
uration pressures were calculated directly from the experimental plane of the RTDs and ks is the thermal conductivity of copper
data, i.e., Pevap ¼ P6 and Pcond ¼ P4 , both taken at the outlets of evaluated at TRTD , which is the arithmetic mean of the five RTDs
the jet cooler and condenser, respectively. used to measure the copper block temperature.
The indicated power, i.e., the useful work performed on the
refrigerant by the piston per unit time, can be determined by the 2.5 Energy Balances and Experimental Uncertainties. As
following equation: discussed in Ref. [55], energy balances were executed on the
experimental apparatus for each test to verify the consistency of
W_ ind ¼ m_ r ðh2 h1 Þ (6) the experimental measurements. The energy balance residue, deb ,
is defined as the difference between the heat transfer rates out of
where h2 is the enthalpy at the compressor outlet computed using and into the refrigeration system
the local experimental values of pressure and temperature.
The coefficient of performance (COP) has been defined consid-
jQ_ cond Q_ c þ Q_ sh þ W_ ind j
ering the energy consumption strictly necessary to remove the deb ¼ 100% (10)
imposed heat load upon the jet cooler Q_ c þ Q_ sh þ W_ ind
Variable Single-jet test, 125 W Jet array test #1, 100 W Jet array test #2, 75 W Jet array test #3, 200 W
3 Results
The coefficient of performance is shown in Fig. 8. A quasi-
linear increase of COP with Q_ c is observed, which results mainly
from the increase of Q_ c since W_ comp increases only slightly with
the thermal load, as depicted in Fig. 9. The multiple jet strategy
not only allows for larger thermal loads to be removed from the
heated surface (up to 200 W), but it also does so with a higher
COP compared to the single-jet case.
The physical explanation for the lower values of compressor
power consumption in the multiple jet cases lies in the corre-
sponding smaller value of pressure lift in comparison to the single
jet case. The behavior of the pressure lift (the difference between
the suction and discharge pressures) is shown in Table 3 for all
cases. For all the multiple jet arrays, the pressure lift is approxi-
mately equal to 4 bar (considering a combined experimental
uncertainty of 60.05 bar), which is almost half the value of pres-
sure lift observed in the single-jet case. As more orifices are pres-
ent in the multiple jets, the restriction imposed on the expanding
fluid is lower than that provoked by the single orifice (300 lm
diameter). As a result, a smaller refrigerant mass is accumulated
in the components on the high-pressure side of the system,
which reduces the pressure upstream of the orifice. The lower DP
Fig. 8 Coefficient of performance as a function of the applied through the orifice results in a significantly higher pressure down-
heat load (cooling capacity) for single and multiple jet impinge- stream (evaporating pressure). The behavior of the saturation tem-
ment cooling peratures (evaporating and condensing) is shown in Fig. 10, which
obviously follows that of the suction and discharge pressures. As
the condensing temperature is significantly reduced for tests with
multiple jet arrays, the subcooling degree at the condenser outlet,
i.e., DTsc ¼ Tcond T4 , drastically diminished from above 21 C,
for the single-jet tests, to values ranging from 0.2 to 3.5 C, for the
multiple-jet tests.
In Figs. 8 and 9, no noticeable differences were perceived
between the values of COP and W_ comp for the multiple-jet arrays
#2 and #3. However, a higher compressor power consumption
was observed for array #1. Although the differences between val-
ues of W_ comp for arrays #1 and #2 (from 4.1 to 7.5 W) and for
arrays #1 and #3 (from 3.7 to 7.7 W) were within the expanded
experimental uncertainty reported in Table 1 (64.53 W), the exis-
tence of a deviation is clear. As can be seen in Fig. 11, the refrig-
erant mass flow rate was very similar for arrays #1, #2, and #3, so
the higher power consumption behavior for jet array #1 can be
attributed to larger values of heat dissipation rate through the
compressor shell compared to the other multiple jet cases. The
exact reason for this higher heat transfer rate is unknown—a
slightly higher air temperature (1 C) inside the calorimeter
chamber is a possible cause—but the fact is that consistently
higher compressor gas discharge temperatures have been meas-
ured in the tests with array #1, in comparison with the other multi-
ple jet tests (see Table 4). These higher discharge temperatures
resulted in higher values of W_ ind , Q_ shell;eb , and Q_ shell;cal .
Fig. 9 Compressor power consumption as a function of the The behavior of the average heater surface temperature is
applied heat load (cooling capacity) for single and multiple jet shown in Fig. 12. As can be seen, a jet configuration with more
impingement cooling spaced orifices (arrays #2 and #3) is advantageous for reaching
Q_ c (W) Single jet (bar) Multiple jet array #1 (bar) Multiple jet array #2 (bar) Multiple jet array #3 (bar)
lower surface temperatures, particularly at high cooling capacities. 150 W, it is possible to see that the three multiple jet arrays give
Since the multiple jet arrays #2 and #3 differ by a 45 deg rotation very similar results. However, a gradual temperature increase is
of the orifice configuration, the similar temperature behavior of clearly perceived for array #1 at higher cooling loads (175 and
these two cases was somewhat expected. For cooling loads up to 200 W). At the more extreme conditions, partial or total dryout of
Nomenclature
As ¼ impingement surface area, cm2
D¼ copper block diameter, cm
h¼ specific enthalpy, J kg1
hs ¼ jet impingement heat transfer coefficient, W m2 K1
ks ¼ thermal conductivity of the heated surface, W m1 K1
L¼ distance between the impingement surface and the
RTDs’ plane, mm
m_ r ¼ refrigerant mass flow rate, kg h1
Fig. 14 Vapor mass quality at the outlet of the jet cooler (x6 ) as
a function of the applied heat load (cooling capacity) for single
P¼ pressure, bar
and multiple jet impingement cooling Q_ c ¼ cooling capacity, W
Q_ cond ¼ condenser heat transfer rate, W
Q_ sh ¼ superheating heat transfer rate, W
Q_ shell ¼ heat transfer rate through the compressor shell, W
The outlet vapor mass quality, x6 , is shown in Fig. 14. As Q_ c ¼ heat transfer rate at the copper block surface, W
expected, the vapor quality increases steadily, exhibiting very s¼ standard deviation
close values for all multiple jet array configurations as well as T¼ temperature, C
reaching very high values (up to 75%). According to Ref. [50], Tamb ¼ temperature of the hot reservoir (ambient), C
the outlet vapor quality can be interpreted as an evaporation effi- TRTD ¼ mean temperature of the RTDs in the copper block, C
ciency, since the liquid that leaves the jet cooler does not produce Ts ¼ impingement surface temperature, C
a cooling effect. Therefore, high evaporation efficiencies are U¼ expanded uncertainty
observed (up to 75%), showing that the designed heat sink was W_ ¼ total input work rate, W
capable of converting a large amount of the impinging liquid jet W_ comp ¼ electrical power consumption of the compressor, W
into vapor, which is the main physical mechanism responsible for W_ fan ¼ electrical power consumption of the fan, W
the heat removal from the test surface. W_ ind ¼ indicated power, W
x¼ vapor mass quality
4 Conclusions X¼ generic output variable
d¼ heat balance residue, %
A novel two-phase jet heat sink that integrates the evapora-
DP ¼ pressure drop through the jet cooler, bar
tor and the expansion device into a single cooling module was
DTsc ¼ refrigerant subcooling degree, C
combined with a refrigeration system which operates with a
DTsh ¼ refrigerant superheating degree, C
compact R-134a oil-free linear-motor compressor. The applic-
Ds ¼ sampling interval, s
ability of the system in the removal of highly concentrated
heat loads has been demonstrated. Experiments have been con-
ducted for distinct jet configurations, i.e., single and multiple Subscripts
jet arrays. The influence of the applied thermal load and orifice
cal ¼ calorimeter energy (heat) balance
geometric configuration on the system performance was quanti-
cond ¼ condensing
fied using thermodynamic performance and heat transfer
eb ¼ first-law energy (heat) balance
parameters. For a fixed orifice diameter and a fixed jet length,
evap ¼ evaporating
operating the active cooling system with multiple orifice con-
figurations resulted in a better thermodynamic performance
than the single orifice configuration. Although more research is References
needed to determine the optimal multiple jet configurations, [1] Smakulski, P., and Pietrowicz, S., 2016, “A Review of the Capabilities of High
these seem to be the way forward to achieve higher critical Heat Flux Removal by Porous Materials, Microchannels and Spray Cooling
heat fluxes and, therefore, higher evaporation efficiencies (out- Techniques,” Appl. Therm. Eng., 104, pp. 636–646.
let vapor qualities). [2] Bar-Cohen, A., and Holloway, C. A., 2014, “Thermal Science and Engineering—
From Macro to Nano in 200 Years,” 15th International Heat Transfer Conference
The two-phase jet heat sink was capable of dissipating cool- (IHTC-15), Kyoto, Japan, Aug. 10–15, Paper No. IHTC15-FL01.
ing capacities of up to 160 W and 200 W from a 6.36 cm2 sur- [3] Nakayama, W., Suzuki, O., and Hara, Y., 2009, “Thermal Management of
face for single and multiple orifice configurations (#2 and #3), Electronic and Electrical Devices in Automobile Environment,” Fifth Interna-
respectively. For these cases, the temperature of the impinge- tional IEEE Vehicle Power and Propulsion Conference (VPPC), Dearborn, MI,
Sept. 7–10, pp. 601–608.
ment surface was kept below 40 C and the heat transfer coeffi- [4] Mudawar, I., 2001, “Assessment of High-Heat-Flux Thermal Management
cient reached values between 14,000 and 16,000 W/(m2 K). Schemes,” IEEE Trans. Compon. Packag. Technol., 24(2), pp. 122–141.
The compact vapor compression cooling solution introduced [5] Chu, R. C., Simons, R. E., Ellsworth, M. J., Schmidt, R. R., and Cozzolino, V.,
here can be further developed for specific applications in ther- 2004, “Review of Cooling Technologies for Computer Products,” IEEE Trans.
Device Mater. Reliab., 4(4), pp. 568–585.
mal management of power electronics for a variety of station- [6] Kheirabadi, A. C., and Groulx, D., 2016, “Cooling of Server Electronics:
ary and mobile systems (for instance, hybrid and electric A Design Review of Existing Technology,” Appl. Therm. Eng., 105, pp.
vehicles). 622–638.