You are on page 1of 7

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A, VOL. 18, NO.

3, SEPTEMBER 1995 521

Jet Impingement Boiling of a Dielectric


Coolant in Narrow Gaps
Gregory M. Chrysler, Richard C. Chu, and Robert E. Simons, Member, ZEEE

Abstract- An experimental investigation into the effect of the because of the complex plumbing that would be required to
chip-to-orifice gap was conducted for jet-impingement boiling of properly channel the large flows.
FC-72 at the back surface of a 6.5-mm square thermal chip on a In addition to simple single phase jet impingement, if the
28-mm square ceramic substrate. Four different types of jets were
used, all of them employing a single 0.50-mm diameter orifice. A coolant is allowed to boil as it passes over the heat source,
fixed-positionjet impingement pin with the orifice centered on a an even greater heat transfer coefficient can be realized. This
flat face measuring 7.62 mm on an edge was used as a base case. method takes advantage of the high heat transfer coefficients
This jet supported a heat flux of about 120 W/cm2 at a reasonable associated with nucleate boiling. Ma and Bergles [7] reported
flowrate. Variable-positionjet impingement pistons were shown to heat fluxes in excess of 100 W/cm2 using a submerged jet to
perform as well as, but no better than the base case. Additional
tests were conducted to investigate performance with variation enhance nucleate boiling on a simulated microelectronic chip.
in the chip-to-orifice gap. Thermal performance was found to be Nonn er al. [8], [9] reported subcooled boiling data for single
insensitive to the gap at large spacing, but below some specific and multiple jets of FC-72 and mixtures of FC-72 and FC-87
gap it degraded with decreasing gap. A sudden jump in the chip liquids impinging on single or multiple heat sources. Copeland
temperature was discovered to occur at a specific gap. This gap [ 101 reported single phase heat transfer coefficients of over 1.6
length was found to vary with jet flowrate.
W/cm2.K and critical heat fluxes of over 160 W/cm2 using
multiple jet impingement on a single silicon heat source.
I. INTRODUCTION One school of thought in the quest for high heat transfer
coefficients at moderate flowrates is to shrink the flow channel
J ET IMPINGEMENT heat transfer has been studied by
many investigators over the past 40 years. Martin [ l ] pre-
sented a thorough review of the literature dealing with single
dimensions. It is well known that small characteristic lengths
will result in large heat transfer coefficients. By this line of
thought, small channels in flow boiling should also provide
and multiple gas jets. More recently several investigators have
studied the potential application of liquid jet impingement heat large coefficients. The problem with small channels in flow
transfer for cooling microelectronic chips. Kiper [2] used the boiling is that as the coolant changes phase, the large ratio in
correlations developed by Martin for a conceptual study of specific volume could result in excessive back pressure and
jet impingement cooling of arrays of electronic chips. Jiji choke off the flow path.
and Dagan [3] used FC-77 and water to investigate single The work reported here investigates the effect of small flow
phase multijet impingement cooling of an array of simulated channels on jet impingement boiling of FC-72 at the surface of
microelectronic chips. Womac et al. [4] investigated single a 6.5-mm square thermal chip. Most of the data was collected
circular jets impinging onto a single heat source. Wadsworth for submerged jets, where the area around the thermal chip
and Mudawar [5] conducted experiments to investigate single- was filled with the coolant. Some data was taken under free
phase heat transfer from an array of simulated microelectronic jet conditions where the area around the chip was not filled
chips to confined two-dimensional jets of FC-72. Incropera and with liquid.
Ramadhyani [6] provided an extensive review of single-phase,
liquid jet impingement cooling for microelectronic chips. 11. APPARATUS
The results of these studies have shown that jet impingement A schematic diagram of the single chip jet impingement
heat transfer promises to provide high heat transfer coefficients apparatus is shown in Fig. 1. Subcooled liquid is drawn from
for electronics cooling applications. The main problem is the expansion tank by a LiquiFlo gear pump (series 35). Most
to provide a high heat transfer coefficient at a moderate or of the liquid flows through the by-pass directly back to the
even low coolant flowrate per chip. Many schemes have been expansion tank. Only a very small amount is supplied to the
devised which will provide very high heat transfer coefficients, test chamber, typically less than 3.2 cm3/s.
but at an excessively high coolant flowrate. These would be Just upstream of the test chamber is a metering valve
difficult to use in the cooling of a multichip electronics module (Nupro SS4MG). This valve allows for very precise control
of the very small flowrates employed in the single chip jet
Manuscript received May 1994; revised March 1995. This paper was
presented at the Intersociety Conference on Thermal Phenomena, Washington, impingement apparatus. Upstream of the metering valve is
DC, May 5-1, 1994. a turbine flowmeter (Flow Technology model FTO-2AIYB-
G. M. Chrysler and R. C. Chu are with the IBM Corporation,Poughkeepsie, LHC-1) capable of accurately measuring flowrates in the range
NY 12601 USA.
R. E. Simons is an Independent Consultant, Poughkeepsie, NY 12603 USA. 0.50-5.0 cm3/s. Just upstream of this flowmeter is a shut-off
IEEE Log Number 9412384. valve, the sole use of which is to prevent backflow through the
107~9886/95$04.000 1995 IEEE
528 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A. VOL. 18, NO. 3, SEYEMBER 1995

Control valve
Turbine flowmeter

Chip 1

XYZ Translation
stage
Exvansion tank -,

Substrate ’
Mounting collar Jet pin
Water-cooled
heat exchanger
(a)

Fig. 1. Experimental flow loop. 1 I


1

IK
Supply plenum

turbine flowmeter when the pump is powered off. Before the


coolant passes through any of these devices, it passes through
an inline filter (Nupro SS-4FW-2, 2.0-pm filtration). Filtration
protects both the flowmeter and the chip in the test chamber.
The test chamber is an aluminum structure with two win-
dows on opposite sides (front and back) for viewing, a
removable right side wall (the jet plenum), and a left side wall
fitted with a rubber boot. The single chip module is positioned
in the test chamber on an arm that extends through the rubber
boot. External to the chamber the arm is mounted to a XYZ Fig. 2. Jet orifice sectional views.
translation stage. The rubber boothranslation stage allows three
degrees of freedom for positioning the chip within the test
chamber. Each stage has a linear displacement of 25 mm so
the chip can be positioned over a range of f12.5 mm from
a nominal position in each of the three coordinate directions,
in increments of 0.01 mm.
The liquid return is located at the bottom of the chamber.
‘ Single 0.50”
dia orifice
Slots for ’
alignment pins
From the chamber, the coolant flows back to the expansion Fig. 3. Jet pin 105.
tank and is cooled by a water cooled heat exchanger.
The jet plenum chamber is made up of two parts; a common
via the XYZ translation stage as with the jet pin. However, the
wall section which is also the right wall of the test chamber,
gap through which the coolant flows across the chip surface
and a removable plenum cover. The internal dimensions of the
is controlled by the piston design-offset fins in the figure.
plenum chamber measured 72.6-mm square by 30.5-mm deep.
The chamber was lined with 6.3-mm thick Lexan to provide
A. Jet Impingement Pins
some thermal isolation from the surrounding test section. An
O-ring was used to seal the plenum against the surrounding A jet impingement pin (jet 105) is presented in Fig. 3. This
test section chamber and another O-ring was used to seal the device is a fixed position orifice. It is O-ring sealed in the jet
plenum cover against the plenum and sidewall assembly. Two plenum cover. The jet orifice is centered in a square, flat face
covers were used for the experiments. One accepted jet pins, measuring 7.62 mm on an edge, slightly larger than the 6.5-mm
the other jet pistons. Both the jet pin and the jet piston have chip. The gap between the chip and the orifice is controlled
a single orifice with a diameter of 0.50 mm. Each cover will by the z-axis of the translation stage. The plenum cover used
be described along with the appropriate orifice. with the jet pin was an aluminum wall measuring 101.6-mm
Conceptually, the method of adjusting the chip to jet orifice square and 7.6-mm thick. A center 12.7-mm diameter through
gap is shown in Fig. 2(a) and (b). Fig. 2(a) shows the jet pin hole receives the pin. A collar slides over the pin and along
approach. A fixed position pin is attached to the plenum cover with three screws, secures the pin to the plenum cover. The
by a collar. Once so attached, the pin is fixed in place. The seal between the jet pin and the plenum cover is made through
gap between the jet orifice and the chip is changed by moving an O-ring positioned in the 2.1 1-mm wide, 9.50-mm diameter
the chip. As mentioned above the chip can be positioned gland in the jet pin.
in increments of 0.01 mm in three coordinate directions by
adjustment of the XYZ translation stage. The other approach B. Jet Impingement Pistons
used to adjust the gap is through the jet piston; this is shown The jet impingement piston is a somewhat more complicated
in Fig. 2(b). Using this method the orifice to chip gap is set part. The orifice is actually part of the plenum cover. One
CHRYSLER et al.: JET IMPINGEMENT BOILING OF A DIELECTRIC COOLANT 529

1
loowedge offset tins
Comer

e
Quarter

for jet ontice

Fig. 4. Jet piston 502.

,ffomd surface at 1.76' Zone 1


...........

Comer ~

~ ; Zone2
....................
:-4
:

Fig. 6. Position of thermal diodes and heating zones.

for jet orifice


TABLE I
Fig. 5. Jet piston 505. STEADY-STATE
PARAMETERS

surface on the piston actually makes contact with the chip. Parameter -
Value
The piston fits over the orifice of the plenum cover. The piston
diode temperature 0.1O'C
position is varied by compression of a spring that is located in
the base of the opening in the plenum cover. A conical spring zone heat flux 0.20 W/cmz
was used in all tests. It is designed to provide a force that is
nearly constant with compression length. The spring is 6.25- coolant flowrate 0.06 cm3/s
mm long and has a major diameter of 7.7 mm and a minor coolant temperature 0.20'c
diameter of 3.7 mm. The specifications for the spring are that
the force be 1.9 to 3.5 N when the spring is compressed 4.45
to 5.8 mm. The thermal chip has two separately powered heating zones,
For each piston, the body is a hollow cylinder with OD 8.18 indicated in Fig. 6, which allows the chip to be powered either
mm and ID 3.30 mm. Figs. 4 and 5 show jet impingement uniformly, or with a simple heat flux distribution (center zone
pistons 502 and 505, respectively. The mating plenum cover different from the perimeter zone). All tests in this study were
for the jet impingement pistons actually provides the jet orifice. conducted with the two zones powered to provide a uniform
For the tests reported the "jet" is created through a single 0.50- heat flux across the chip.
mm diameter orifice. This orifice is located at the end of a Fig. 6 shows the thermal module, viewed from the back
3.18-mm diameter shaft centered in a 8.20-mm diameter by side-the side onto which the jet impinges. The thermal chip is
10.70-mm deep cavity in the jet pin plenum cover. actually instrumented with six diodes for sensing temperature,
Piston 502 maintains a 0.10-mm channel across the chip but only five are accessible through the pins on the 28-
surface through which the coolant will flow. This channel is mm ceramic substrate. There are two centrally located diodes
maintained through an off-set machined into the active end (neither of which is truly in the center of the chip, and only
of the piston. Assuming that the'pressure at the chip surface one of which is accessible), two diodes located in opposite
is not high enough to float the piston, the gap is uniform comers, one edge diode, and a diode located in roughly the
across the chip, independent of flowrate and independent of center of a quadrant of the chip. This arrangement gives an
chip-to-orifice distance. excellent distribution of the chip temperature. The positions of
Piston 505 is similar in basic design to piston 502. The the diodes and the numbering scheme used in data collection
major difference is that for this piston the active surface is are displayed in the figure. During testing the chip is positioned
conical with an angle of 1.76" (from flat). For this piston, upside down in the test chamber.
it is desired that the pressure developed at the surface of
the chip is sufficient to float the piston off the chip surface. 111. EXPERIMENTAL
PROCEDURE
The gap generated will be a function of the flowrate and the
compression strength of the spring. The setting of power supplies (two required for the single
chip thermal module), as well as monitoring and collection
of all data and determining steady-state were controlled by a
C. Heat Source data acquisition and control program running on a personal
All data was collected on an uncapped ceramic thermal computer. The program used a simple criteria for determining
module. This module has a single thermal chip measuring 6.5 steady state: it checked the difference in the value of four
mm on an edge, centered in a ceramic substrate measuring 28 parameters over a fixed monitoring period. The absolute value
mm on edge. The substrate was exposed to the coolant (after it of the difference of each parameter between two monitoring
impinged against the back of the chip) and thus could provide periods could not exceed a pre-selected value. The parameters
an additional path for heat transfer. and the pre-selected values are listed in Table I.
530 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A, VOL. 18, NO. 3, SEPTEMBER 1995

In all cases the parameter checked was the integrated TABLE I1


average over the 60-s monitoring period, not the instanta- DIODESYMBOLS
THERMAL

neous value. Many trial runs have shown that the above Symbol Position
values provide an accurate method of checking for steady
state+specially in flow boiling where there can be large 0 Edge
instantaneous temporal variations in diode temperatures.
A Quarter
Most of the data was obtained for submerged jet conditions.
For these, the test chamber was filled with FC-72 to a height U Comer
of about 5 cm above the top edge of the substrate. Since the
retum is at the bottom of the test chamber, the coolant can V Comer
become stratified in the chamber during long, high power runs. 0 Center
The 5-cm high region of liquid above the substrate provided
a region where the stratified liquid could accumulate.
In the jet impingement data presented in the results section, T,hP - T
,.c
the FC-72 jet temperature for the runs displayed in Figs. 9 --
20 30 40 50 60 70 80 90
7
7-
through 13 was maintained at a mean temperature of 20.5"C, Natural ConvectionI Pool Boiling
and for any set of data (any plot) the standard deviation was FC-72 I
Run did not reach CHF
less than 0.5"C. chip temp exceeded 85'C I

I
I
IV. DATAREDUCTION
The presentation of the results in this report is in the form of
heat flux (W/cm2) and temperature or temperature difference
(both in "C). Since the thermal chip has two heating zones,
the heat flux is calculated based on the chip average. As was
pointed out in the Section 111, a software program controlled
the setting of the power supplies. For the data presented in this
report, the outer zone was powered to a level 1.91 times that Tchrp foe)
of the inner zone. This corresponds to the area ratio between Fig. 7. Natural convection and pool boiling results
the inner and outer zones. Thus the chip was powered to a
uniform heat flux.
The diode temperatures were determined from the mea- of about 40"C, where, as the heat flux is increased (the next
surement of the diode resistance. Each diode has a unique data point), the temperature difference decreases.
calibration curve which was determined in separate experi- The actual diode temperatures are plotted in the same figure.
ments prior to this investigation. The chip average temperature They are shown by the grouping of data at the right. All
was calculated from the average of the five thermal diodes. four diodes record the phenomena of temperature over-shoot
Chip-to-jet temperature differences were then determined from (one comer diode on the chip was dead). Starting at the
the average chip temperature minus the jet temperature. The lowest heat flux and following the data up shows that the
jet temperature is measured by a single copper-constantan temperature variation across the chip increases with heat flux.
thermocouple located in the jet plenum. This variation is almost nonexistent at low fluxes but becomes
The data files written by the controlling software were about 8°C at the highest flux.
converted into a form suitable for plotting (i.e., chip average Finally, the maximum heat flux obtained before the 85"C
heat flux and the chip average temperatures). high temperature limit was reached was only 45 W/cm2
corresponding to a 19-W chip. Although critical heat flux in
pool boiling would be expected to occur in the range of 15 to
V. RESULTS 20 W/cm2, as shown in Fig. 7 a heat flux of 45 W/cm2 was
Before discussing the results, a description of the symbols attained without experiencing departure from nucleate boiling.
used in the data plots is needed. In all plots, the lone-filled This higher heat flux is due to natural convection off the
diamond symbols represents the chip average temperature. The substrate surface and, in the immediate vicinity of the chip,
collection of symbols is shown in Table 11. nucleate boiling.
A starting point in the discussion of jet impingement boiling Before proceeding with the results of this study a few
results is natural convection/pool boiling. This provides a words must be said about the effect of the substrate on the
base case for comparisons. Subcooled pool boiling results performance of jet impingement boiling. First it should be
are presented in Fig. 7. The results are for a chip facing noted that since the coolant is impinged directly against the
horizontal (the same orientation as used in the jet impingement back of the heat producing device, the thermal chip, the
runs). In the figure, the lone symbols are the chip heat flux substrate will experience only indirect flow of the coolant. That
(W/cm2) plotted against the driving temperature difference being the case, it might be expected that the substrate would
in the typical boiling curve fashion. A small temperature play a relatively small role in the overall thermal performance.
overshoot is apparent at a chip to pool temperature difference Fig. 8 illustrates the truth of this expectation. Heat transfer data
CHRYSLER er al.: JET IMPINGEMENT BOILING OF A DIELECTRIC COOLANT 53 1

-Pm 105 '


00
i
- Plston500
Plston 502
I
i
o chipwithmask 1 I
vo
v chip without mask

I 4

I
L - -~ I
0' '
IO 20 30 40 50 60 70 80 90

Fig. 8. Effect of substrate masking. Tchip (OC)

Fig 10 Results for jet pin and pistons at 2 2 cm3/s


Tchlp T~et(OC)
10 20 30 40 50 60 70 80 90
Tchmp Tje~(OC)
140 Jet Pin ;05 ' ' ' ' ' I 10 20 30 40 50 60 70 EO 90

I
2.2 cm'ls FC-72 140
2.951" gap Plston 500
100
I2O
mv 0"
120 I --- plston 505
-I
I

20 1 m

O L ' 1 1 1 .I
10 20 30 40 50 60 70 80 90
0 1 '
TCh,Q COc) 10 20 30 40 50 60 70 EO 90

Fig. 9. Results for jet pin 105 at 2.2 cm3/s Tchip (OC)

Fig. 11. Results for pistons 500 and 505 at 1.6 cm3/s.
was collected with the substrate encased in plastic to determine
the effect of substrate cooling. This data was then compared
500 is in the middle of a circular surface of diameter 3.18
with the normal exposed substrate tests. In both cases, the
mm is immaterial, at least for the average temperature. The
coolant was impinged against the back of the exposed chip.
gap is simply too large for any surface effect to be felt at the
As can be seen, the effect of the substrate is very small, about
10%. The heat transfer due to jet impingement boiling at the chip. There is however, a slight variation in the temperature
back side of the chip dominates the thermal performance even distribution across the chip with pin 105 providing a somewhat
with an exposed 28-mm square substrate. smaller ma-min temperature than piston 500. The effect of
Jet impingement provides the required method of enhance- reducing that gap will be discussed shortly.
ment. The results of jet impingement boiling is presented The first attempt at providing a controlled gap was with jet
in Fig. 9. The format is identical to that of Fig. 7. Jet pin piston 502. Recall that this piston provided a narrow channel
105 (single 0.5-mm orifice), centered over the back of the for flow across the chip surface through 0.10-mm offset on the
chip provided 2.2 cm3/s of FC-72 to the chip surface. The piston surface. The results of a series of runs at a flowrate of
significantly higher heat flux (over pool boiling) is obvious. 2.2 cm3/s are presented in Fig. 10. Close inspection reveals
Chip heat fluxes of almost 120 W/cm2 can now be realized that at any given temperature difference, the use of piston 502
before the 85OC maximum temperature is reached. results in a chip heat flux which is lower than with no piston.
Fig. 10 presents the results of the jet pin 105, and all jet The poor performance of this piston is obvious at higher heat
pistons. In this figure as well as Fig. 11 only the chip average fluxes, but also exists at lower heat fluxes where the flow is
temperature difference and the temperature of the center diode single phase. Comparison with Fig. 11 indicates that piston
are shown. The results for piston 500 (no piston, just the 502 at 2.2 cm3/s performs about the same as no piston at 1.6
orifice) plot practically on top of those of jet pin 105. This cm3/s, especially at the higher heat fluxes where more vapor
is expected since both pin 105 and piston 500 have a single generation occurs.
0.50-mm orifice. At the chip-to-orifice gap of 2.95 mm, the There are a number of possible explanations for this poor
fact that the orifice of pin 105 is in the middle of a flat plate performance. It is possible that not all the coolant delivered
measuring 7.62 mm on an edge whereas the orifice of piston to the orifice actually flows over the surface of the chip.
532 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A, VOL. 18, NO. 3, SEPTEMBER 1995

!3- ,- 90
Chip powered at ?OW Clup powered at $OW
let 105 at 2 2 cm Is FC-72 let 105 at 2 2 cm Is FC-72
80

P
2
+ .
..
5 j

40
0.0 0.5
* . ...
1.0
4

1.5
t
2.0
t t
2.5
, 1'

3.0
:50

140
60

50 1 .-. * w . . .
, , I
.Iso
0.0 0.5 1.0 1.5 2.0 2.5 3.0
zchpjet "(1 Zchlp-pi (")

Fig. 12. Thermal jump at 30 W, jet pin 105 at 2.2 cm3/s. Fig. 13. Thermal jump at 40 W, jet pin 105 at 2.2 cm'k

After issuing from the orifice, some of the coolant could


Prior to the jump there is vapor generation at the surface of
flow backwards through the center of the piston and then
the chip (in the gap region) but not so much so as to totally
down over the outer wall of the piston completely bypassing
block light transmission through the region. After the gap, the
the chip surface. Another possible explanation for the poor
vapor generation totally blocks out light transmission making
performance is that the 0.1-mm gap is not small enough to
the region appear black.
really be considered a small gap when compared with the thin
The same thing occurs at a higher chip power. Fig. 13 shows
boundary layers associated with jet impingement.
the results for a chip power of 40 W. In this case, due to the
A smaller gap was established with piston 505. This piston
higher chip power there is more vapor generation prior to the
would float above the chip surface with a very small gap
temperature jump. The temperature jump is again accompanied
dimension. Fig. 10 shows the boiling curve for piston 505
by a sudden increase in vapor generation totally blocking light
at a flowrate of 2.2 cm3/s and an chip-to-orifice gap of 2.95
transmission through the gap region. More data, spaced 0.05
mm. This gap compressed the spring by 1.O mm. These results
mm apart, was collected between the pre-jump 1.75 mm and
compare well with those of piston 500 (no piston). One major
post-jump 1SO-mm gap to try to see if the transition is sharp or
advantage of piston 505 (although not shown in the figure)
spread out over length. The results indicate that the transition
is that the max-min temperature is much smaller than that
is indeed sharp and that the difference between pre and post
for piston 500. It was discomforting that even the small flow
jump is less than 0.05 mm. After the jump, decreasing the gap
gap provided by the floating piston did not result in better heat
further results in a gradual increase in the chip temperature.
transfer performance, although it did do better than piston 502.
This is different than the pre-jump effect of gap. Prior to the
Results for jet piston 500 and 505 at the same gap but at
jump the chip temperature is very insensitive to gap.
a lower flowrate of 1.6 cm3/s are presented in Fig. 11. The
Finally, the effect of flowrate was investigated by running
overall appearance is the same as that at the higher flowrate,
the experiment at a lower flowrate. At 1.6 cm3/s and 30 W
but the performance is poorer. At any given chip heat flux, the
the effect is the same. Prior to the jump, the temperature does
lower flowrate jet results in a higher chip-to-jet temperature
not change with gap. However, after the jump the temperature
difference. The maximum heat flux before reaching the 85°C
again increases as the gap is decreased. It appears that the jump
maximum temperature limit is reduced to under 100 W/cm2.
position is a function of flowrate rather than chip power. Both
As it stands, the jet impingement piston can perform as well
the 30-W and 40-W chips at 2.2 cm3/s have a jump between
as, but no better than no piston at all. Because of this finding, a
1.75 mm and 1.50 mm whereas the 30-W chip at 1.6 cm3/s
series of experiments were initiated to investigate the effect of
has a jump at a larger gap, between 2.00 mm and 1.75 mm.
the gap. The results of these experiments are presented in Figs.
It is interesting to notice the individual diode temperatures
12 and 13. In these the average temperature difference (lower
before and after the jump. At large gaps the temperature of
single symbols) and the actual diode temperatures (upper
the center diode (under the stagnation point of the jet) is not
group of symbols) were plotted against the jet orifice-to-chip
the lowest. It appears that the stagnation point heat transfer is
gap. All runs were with pin 105.
not the best. The heat transfer may be better at some distance
Fig. 12 presents the data for a chip power of 30 W with
from the stagnation point where boiling actually takes place.
a jet flowrate of 2.2 cm3/s. Reading the curve from right to
The stagnation point heat transfer, at least at this flowrate and
left shows how the heat transfer performance changes as the
chip power, may be single phase. After the temperature jump,
gap is decreased. The starting gap was 2.95 mm, the same
the center diode temperature is the lowest. Perhaps boiling is
as all the runs described earlier. The performance is more or
now taking place over the entire back surface of the chip.
less constant at large gaps; gaps down to 1.75 mm. At this
point, reduction of the gap by another 0.25 mm results in a
sudden jump in the chip surface temperature. This jump in VI. CONCLUSION
temperature is accompanied by a sudden increase in the vapor Jet impingement boiling experiments with FC-72 were
generated in the gap between the chip surface and the orifice. conducted at two flowrates and chip-to-orifice gaps of 0.25
CHRYSLER er al.: JET IMPINGEMENT BOILING OF A DIELECTRIC COOLANT 533

mm to 2.95 mm. At 2.2 cm3/s the average temperature of [lo] D. Copeland, “Single-phase and boiling cooling of a small heat source
the chip remained relatively unchanged with gap, for gaps by multiple nozzle jet impingement,” ASME Paper No. 92-WAEEP-4,
1992.
greater than 1.75 mm. Below this gap, the average chip
temperature increases with decreasing gap. The same form
is true at the lower flowrate, 1.6 cm3/s, but the change from
gap insensitivity to increasing temperature with decreasing gap
occurs at a larger gap, between 2.00 mm and 1.75 mm. Gregory M. Chrysler received the B.M.E. in 1977,
the M.S.M.E. in 1981, and the Ph.D. degree in 1984,
The jet impingement piston can perform as well as, but no all from the University of Minnesota.
better than no piston at all. This is because of the very small In 1984, he joined the IBM Corporation, where he
gaps established with the jet impingement pistons. The work is presently a senior engineer for IBM Systed390
Division at Poughkeepsie, NY,working in the Ad-
reported here shows that the chip surface temperature increases vanced Thermal Laboratory. His areas of interest
as the chip-to-orifice gap decreases for gaps typically below include jet impingement, boding heat transfer, and
1.5 mm to 2.0 &. The very small chip-to-orifice gaps created forced-air convection for electronics cooling.
with the jet impingement pistons (0.10 mm) were too small
for boiling heat transfer to- perform well.

REFEENCES

H. Martin, “Heat and mass transfer between impinging gas jets and solid Richard C. Chu received the B.S. degree in me-
surfaces,” in Advances in Heat Transfer, vol. 13, J. P. Hartnett and T. chanical engineering from National Cheng-Kung
F. Irvine Jr., Eds. New York Academic Press, 1977, pp. 1-60. University, in 1958, and the M.S. degree from
A. M. Kiper, “Impingingwater jet cooling of VLSI circuits,” Int. Comm.
’ Purdue University, in 1960. He also received an
in Heat and Mass Transfer, vol. 11, pp. 517-526, 1984.
L. M. Jiji and Z. Dagan, “Experimental investigation of single-phase Honorary Doctor of Science degree from the Amer-
multi-jet impingement cooling of an array of microelectronic heat ican University of the Caribbean, in 1992.
sources,” in Cooling Technology for Electronic Equipment, W. Aung, He joined IBM in 1960 and has held a variety
Ed. New York Hemisphere, 1988, pp. 333-351. of technical and managerial assignments in the
D. J. Womac, G. Aharoni, S. Ramadhyani, and F. P. Incropera, “Single Development Laboratory at Poughkeepsie, NY. He
phase liquid jet impingement cooling of small heat sources,” in Proc. has made sustained contributionsto the development
Ninth Int. Heat Trans. Cont. 1990, pp. 149-154. of cooling technology and thermal systems for large-
D. C. Wadsworth and I. Mudawar, “Cooling of a multichip electronic scale computers. His achievements are documented in 30 patents and over
module by means of confined two-dimensional jets of dielectric liquid,” 100 patent disclosure publications.
ASME J. Heat Trans., vol. 112, pp. 891-898, 1990. Dr. Chu has received 18 IBM Invention Achevement Awards, 4 IBM
161 F. P. Incropera and S. Ramadhyani, “Single-phase, liquid jet impinge- Outstanding Innovation Awards, and an IBM Corporate Award. He was
ment cooling of high performance chips,” in Proc. NATO Adv. Study appointed an IBM Fellow in 1983, the company’s highest technical honor.
Institute on Cooling of Electronic Systems, 1994, pp. 277-327. He is also an ASME Fellow, A A A S Fellow, and a member of the National
[71 C.-F. Ma and A. E. Bergles, “Jet impingement nucleate boiling,” Int. J. Academy of Engineering.
Heat and Mass Transfer, vol. 29, no. 8, pp. 1095-1 101, 1986.
P I T. Nonn, Z. Dagan, and L. M. Jiji, “Boilingjet impingement cooling of
simulated microelectronic components,” ASME Paper No. 88-WAEEP-
3, 1988.
r91 T. Nonn, Z. Dagan, and L. M. Jiji, “Jet impingement flow boiling of a
mixture of FC-72 and FC-87 liquids on a simulated electronic chip,” in Robert E. Simons (M’88) for a photograph and biography, see p. 30 of the
Hear Transfer in Electronics, HTD-111, ASME, 1989, pp. 121-128. March issue of this TRANSACTIONS.

You might also like