Thermal conductivity enhancement of energy storage media

using carbon bers

Jun Fukai*, Makoto Kanou, Yoshikazu Kodama, Osamu Miyatake
Department of Chemical Systems and Engineering, Kyushu University, 6-10-1, Hakozaki, Higashi-ku, Fukuoka 812-
8581, Japan
Received 17 March 1999; accepted 20 September 1999
Abstract
Carbon bers with a high thermal conductivity are used to enhance the thermal conductivities of
energy storage media. Two types of enhancement techniques were studied. One is the technique using
randomly oriented bers, while the other is that of using a ber brush. The carbon bers essentially
enhance the eective thermal conductivity of ber/paran composites. For the random type, the ber
length has little eect on the eective thermal conductivity. The ber brush increases the eective
thermal conductivities to the maximum values predicted which is theoretically. The ber brush is also
useful for enhancing the eective thermal conductivities of packed beds of particles. 7 2000 Elsevier
Science Ltd. All rights reserved.
Keywords: Thermal conductivity enhancement; Carbon ber; Paran; Packed beds
1. Introduction
Phase change materials (PCMs) are prime candidates for use as thermal energy storage
media because of their superior energy storage capacities. They should possess a high
density, high specic heat, high thermal conductivity, congruent melting, small volume
change, little supercooling and so on [1]. However, it is practically dicult to nd or
develop PCMs that satisfy all desirable qualities. In many cases, unsuitable properties are
Energy Conversion & Management 41 (2000) 15431556
0196-8904/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S0196- 8904( 99) 00166- 1
www.elsevier.com/locate/enconman
* Corresponding author. Tel.: +81-92-642-3515; fax: +81-92-642-3519.
E-mail address: [email protected] (J. Fukai).
improved using some techniques. For instance, nucleating agents are added to the PCM
having a poor nucleating property. From a heat transfer point of view, the PCMs are
required to have a high enough thermal conductivity to give the designed heat exchange
rates. Thermal conductivity promoters are frequently used for the PCMs with low thermal
conductivities.
The PCM that has been most often chosen in the studies of thermal conductivity
enhancement is paran. This is probably because, though paran is one of the most useful
materials for thermal energy storage at less than 1008C, its thermal conductivity is
unacceptably low. For paran, many researchers [26] examined thermal conductivity
enhancements using ns, honeycombs and porous media. The enhancement techniques are also
important in space based power system applications because of the large amount of heat
rejection from a power cycle in a short period of time [7,8]. The same problem is found in the
development of chemical heat pump systems, which utilize endothermic/exothermic reactions
between a solid and a gas. These systems use packed beds of solid particles to increase the
Nomenclature
b radius of cylinder (m)
C
P
specic heat (J/kg)
d
P
mean diameter of solid particles (m)
f initial temperature (K)
g wall temperature (K)
k thermal conductivity (W/(m K))
L
f
ber length (m)
r radial distance (m)
T temperature (K)
t time (s)
X volume fraction
a thermal diusivity (m
2
/s)
e void fraction
f parameter representing contact heat transfer rate between particles
r density (kg/m
3
)
Z parameter representing contact heat transfer rate between bers
Subscripts
c composite
cal calculation
exp experiment
g gas phase
f ber
m energy storage media
s solid phase
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1544
surface area. Instead, the eective thermal conductivities of the packed beds are extremely low.
This problem is a major obstacle in the practical applications of these systems. In this eld, the
enhancement techniques using metal ns and wires were also examined [9,10].
The thermal conductivity promoter should not only have high thermal conductivity but also
compatibility with the energy storage media. It is also desirable that its density is too low to
give much extra weight to the energy storage equipment. The aforementioned studies employed
promoters made of aluminum, copper, nickel etc. Although these metals have high thermal
conductivities, they are often incompatible with some of the PCMs. For instance, aluminum is
compatible with paran while copper and nickel are incompatible. Aluminum and copper are
incompatible with some salt hydrates. To this end, the substance of the promoter is frequently
limited by its compatibility with the heat storage media rather than the thermal conductivity.
Carbon bers are strongly resistant to corrosion and chemical attack. Thus, they are
supposed to be compatible with most PCMs. There are many commercial carbon bers whose
thermal conductivities are the same as those of aluminum and copper. The thermal
conductivities of some products are more than 1000 W/(m K). In addition, the densities of
carbon bers are theoretically less than 2260 kg/m
3
, which is lower than those of metals that
are usually used as the promoters. These physical and chemical properties of carbon bers
reveal the possibility that they might become excellent promoters.
In this paper, two thermal conductivity enhancement techniques using carbon bers are
investigated. In the rst technique, the bers are randomly distributed in the PCMs. The
second technique employs a ber brush such that the directions of the bers coincide with
those of the heat ow. The transient temperature response and the eective thermal
conductivity of the composites are measured. The second technique is also applied to a packed
bed to enhance its eective thermal conductivity.
2. Experimental
Carbon bers with a diameter of 10 mm, a density of 2170 kg/m
3
and a thermal conductivity
of 220 W/(m K) at room temperature were provided. The phase change temperature and the
thermal conductivity of the paran wax used in this study are 41438C and 0.26 W/(m K),
respectively, at the solid state. The bers and the paran wax were packed into a steel
cylindrical capsule having a 50 mm diameter and 130 mm height. Two types of enhancement
techniques were studied. In the rst example, the carbon bers, cut to a certain length were
randomly packed in the capsule (Fig. 1(a)). The ber lengths were 5 and 200 mm, which are
much shorter and longer than the diameter of the capsule. In another example, a radial brush
made of carbon bers was inserted into the capsule (Fig. 1(b)). The diameter of the brush is
the same as that of the capsule. In both samples, molten paran was poured into the capsule
after the promoter was inserted in the capsule. Air was left inside the capsule because of the
volume change in the phase change process. The volume fraction of the composite in the
capsule was about 80%. Thick caps made of an acrylic resin covered the top and bottom of
the capsule to suppress any heat ow in the longitudinal direction as much as possible.
For measurement of the transient temperature response, a thermocouple was placed at the
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1545
center of the sample (Fig. 1(b)). The sample at a uniform temperature was heated to melt the
paran wax and cooled to solidify it in a temperature controlled bath.
The eective thermal conductivity of the composite was determined through measurement of
the eective thermal diusivity. Four thermocouples were placed at regular intervals in the
radial direction of the sample (Fig. 1(b)). The sample at the uniform temperature of 258C
was heated to 08C in a water bath. The one-dimensional conduction heat transfer equation for
a homogeneous cylindrical body, initial conditions and boundary conditions are written as
@T
@t
a
1
r
@
@r

r
@T
@r

1
ICs T fr at t 0 2
BCs
@T
@r
0 at r 0 3
T gt at r b 4
where fr is the initial temperature and gt the wall temperature. The thermal diusivity was
determined by using a nonlinear least square technique:
Fig. 1. Congurations of the carbon bers within a cylindrical capsule. In Figure (b), the open circle indicates the
location of the thermocouple for the measurement of the transient temperature response. The crosses indicate those
for the measurement of the thermal diusivity. These thermocouples are also located at the corresponding locations
in the random type.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1546
X
j
X
i

T
cal
a; r
i
, t
j
T
exp, i, j

2
4min 5
where r
i
is the position of the ith thermocouple, t
j
the jth sampling instant, T
cal
the
temperature numerically calculated by using Eqs. (1)(4) and T
exp
the experimental
temperature. The ratio of the thermal diusivity of the composite to that of the PCM is
approximately equal to the thermal conductivity ratio when the ber volume fraction is low.
The composite, including the ber brush, is heterogeneous because the ber volume fraction
depends on the radial distance (Fig. 1(b)). To this end, the model that accounts for this issue
was developed to estimate the eective thermal diusivity. However, the dierences between
the results for the two models were negligible (Appendix A).
3. Results and discussion
3.1. Paran
The thermal conductivity of the composite usually increases with the volume fraction of the
thermal conductivity promoter. Instead, the energy storage density accordingly decreases. The
thermal conductivity enhancement should be attempted under an acceptable reduction in the
energy storage density. In this study, the ber volume fraction was restricted to approximately
<2%.
The photographs shown in Fig. 2 are the typical congurations of the carbon bers packed
into the capsule, which is transparent with the same dimensions as the aforementioned capsule.
Fig. 2. Carbon bers packed into a transparent cylindrical capsule. The volume fraction of the carbon bers is
0.005.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1547
This gure demonstrates that the bers are densely and entirely distributed in the capsule even
at X
f
0:005 (0.5%). There are distinct voids for the brush type because the brush is
handmade. The transparent capsule containing the ber/paran composite was repeatedly
frozen and melted in a water bath to investigate the change in the packing states of the bers.
As the result, the ber state was not changed. That is, the bers did not sink toward the
bottom of the capsule. This is an important property for a thermal conductivity promoter.
Fig. 3 shows the transient temperature response for the random type. In both processes, the
slopes of the curves decrease at 41428C due to the phase change. The thermal response within
the composite must be more sensitive as the ber volume fraction increases. Nevertheless, the
melting times for X
f
< 0:01 are longer than that for the paran wax X
f
0: This is because
the bers suppress natural convection in the liquid phase. However, two percent volume bers
provide almost full recovery of the melting time. This fact shows that the loss in the convection
heat transfer rate is fully compensated by the increase in the conduction heat transfer.
Contrary to the melting process, the bers essentially decrease the freezing time as the volume
fraction increases. At the early stage of the freezing process, the bers must suppress the
natural convection in the liquid phase. This eect is slightly observed in the curves for X
f

0:0054 and 0.01 at T > 428C: However, the convection heat transfer does not play an
important role in the entire freezing process.
Fig. 4 shows the eect of the ber length on the temperature response. Again, these lengths
are much smaller or larger than the radius of the capsule. Although the thermal responses for
the longer ber are slightly sensitive, the ber length does not have a signicant eect on the
response. This fact reveals that for the random type, an arbitrary length of carbon bers is
Fig. 3. The time variation of the transient temperature responses at the center of the ber/paran composites (the
random type).
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1548
Fig. 4. The eect of the ber length on the transient temperature response at the center of the ber/paran
composites (the random type).
Fig. 5. Transient temperature response at the center of the ber/paran composite. Comparison between the
random type and the brush type.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1549
allowed in practical applications as long as the bers are densely and entirely distributed in the
energy storage media.
Fig. 5 compares the results for the dierent congurations. The temperature responses for
the brush type are more sensitive than those for the random type. This is because the bers for
the former are oriented in the heat ow direction.
Fig. 6 shows the eective thermal conductivity of the composites k
c
normalized with the
thermal conductivity of the PCM k
m
: In these measurements, four test pieces were provided
for the individual volume of the bers. The measurements were performed ve times for each
test piece. The errors for these results were within 3%. The average and range of the results for
the four test pieces are indicated in the gure. Apparently, the eective thermal conductivity of
the composite increases as the ber volume fraction increases. It is worthwhile noting that for
paran, two percent bers increase the eective thermal conductivity by a factor of six.
Although the ber length slightly aects the eective thermal conductivity, this eect is not
signicant as observed in Fig. 4. The results for the mixture whose main component is
Na
2
SO
4
10H
2
O [11] are also indicated in the gure. The thermal conductivity of this mixture is
0.8 W/(m K). The thermal conductivity ratio for this mixture is lower because of its high
thermal conductivity. However, an enhancement is still observed for this mixture.
Kunii and Smith [12] developed the following heat transfer model for ber matrices:
X
1=3
m
k
c
k
m

X
1=3
m
X
1=3
m

k
f
Zk
m
1 6
where k
f
is the thermal conductivity of the ber, X
m
the volume fraction of the PCM
1 X
f
and Z the empirical parameter. Based on Eq. (6), the experimental results are
Fig. 6. Thermal conductivity enhancement of the phase change materials using the random type.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1550
Fig. 7. Correlation of the eective thermal conductivity.
Fig. 8. Thermal conductivity enhancement of the phase change material using the brush type.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1551
rearranged in Fig. 7. This gure demonstrates that a straight line ts the experimental results
for each ber length. Thus, the values of Z 1:75 for L
f
200 mm and Z 2:19 for L
f
5
mm were evaluated using a least square technique. It should be noted that the value of Z is
independent of the thermal conductivity of the PCM.
Fig. 8 shows the results for the ber brush. The eective thermal conductivity for this type is
three times as large as that for the random type. The parallel model, theoretically, gives a
maximum for the eective thermal conductivity of the composite media:
k
c
1 X
f
k
m
X
f
k
f
7
A comparison between the experiment and the theory, in Fig. 8, shows that the brush type has
the greatest enhancement eect.
It is dicult to discuss generally the amount of carbon ber to be added in the energy
storage unit using paran. If the thermal conductivity is required to increase to that of ice (2.2
W/(m K)), it should be increased by a factor of 10. If the carbon ber used in this study is
employed, the bers of 3% in volume are needed for the random type while those of 1% are
needed for the brush type. The reductions in the energy storage density corresponding to these
ber volumes are probably acceptable.
3.2. Packed beds
The ber brush was used for a packed bed of glass particles, which are substitutions for the
particles used for chemical heat pump systems. The thermal conductivity of the glass used is
Fig. 9. Eect of the bers on the void fraction of the packed beds.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1552
0.8 W/(m K). The glass particles were packed, with intermittent vibration, into the cylindrical
capsule into which the brush was inserted.
Fig. 9 shows the eect of the ber volume fraction on the void fraction of the packed beds.
d
p
is the mean diameter of the particles. The particle diameters for d
p
0:5 and 1.2 mm range
within 20:15 mm, while the diameter range for d
p
0:1 mm is not known. The bers obstruct
the packing of the particles, thus increasing the void fraction. The void fraction does not
proportionally increase at X
f
> 0:005, while it rapidly increases at X
f
< 0:005: This fact reveals
that the bers do not signicantly reduce the stored chemical energy density even if the ber
volume fraction increases to some degree. The reason that the void fraction for d
p
0:5 mm is
higher than those for the others might be due to the dierence in the particle size distribution.
Fig. 10 shows the normalized eective thermal conductivities. The eective thermal
conductivity of the packed bed without the bers is 0.200.24 W/(m K) (= k
m
). These values
are of the same order as the eective thermal conductivities of the packed beds employed in
the studies on the chemical heat pump system (0.10.3 W/(m K) [9,13]). The eective thermal
conductivity increases as the ber volume fraction increases. There is no eect of the particle
diameter on the thermal conductivity ratio. Ogura et al. [9] attempted to enhance the eective
thermal conductivity of packed calcium hydroxide particles by using coiled copper wires with
an 80 mm diameter. They reported that the wires hardly enhanced the eective thermal
conductivity because the wires obstructed the dense packing of the particles. Contrary to their
results, the carbon bers essentially enhance the eective thermal conductivity of the packed
beds.
The parallel model given by Eq. (7) was also applied for this system. In the model, one
Fig. 10. Thermal conductivity enhancement of the packed beds using the brush type.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1553
phase is the ber phase and the other is the packed bed phase. k
m
in Eq. (7) is dened as the
eective thermal conductivity of the packed bed. Based on the model developed by Kunii and
Smith [12], the eective thermal conductivity of the packed bed phase is modeled by
k
m

e
1 X
f
k
g

1 e=1 X
f

f

2k
g

=3k
s

k
g
8
where k
g
and k
s
are the thermal conductivities of gas and solid phases, respectively, and f is
the parameter representing the heat transfer rate between particles. In this equation, radiation
heat transfer is neglected because of the low temperature in the present experiments. To this
end, Eqs. (7) and (8) give the theoretical value of this system. Eq. (8), including the f value, is
independent of the particle diameter. At this point, this equation explains no dependence of the
experimental results on the particle diameter. However, the theory is higher than the
experimental results as shown in Fig. 10. The reasons for this dierence are probably as
follows: Eq. (7) is available when the temperatures in the two phases located nearby are
assumed to be identical. This assumption is usually available for the composite media. In the
packed beds, on the other hand, the contact between a ber and a particle occurs at a line or a
point. Therefore, the thermal contact resistance between the two phases is probably too high to
maintain them at the same temperature. The second reason is due to the error estimation of
the parameter f in Eq. (8). This parameter is modeled considering the number of contact
points on the surface of a particle [12]. Though the bers must obstruct the contact between
the particles in this system, this eect is not considered. Accordingly, the model for this
parameter has error. The former is probably a major reason for the error in the theory while
the latter is minor. This is because the heat transfer between the particles and the bers, not
that between the particles, mainly contributes to the thermal conductivity enhancement.
4. Conclusion
Carbon bers having a high thermal conductivity are employed to enhance the thermal
conductivity of thermal energy storage media. The random and brush types were examined.
Both types are useful for thermal conductivity enhancement. For the random type, the ber
length has a slight eect on the enhancement. The brush type signicantly enhances the
eective thermal conductivity in the direction of the ber orientation. Although the brush type
is superior to the random one from the standpoint of the enhancement technique, the choice
between the two types also depends on the structure of the energy storage equipment, the cost
and so on. The experimental results reveal that carbon bers are also useful for enhancing the
eective thermal conductivity of the packed beds.
Acknowledgements
We thank the Nippon Graphite Fiber Corporation for providing carbon bers.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1554
Appendix A
The bers are assumed to locate in regular intervals in the azimuthal and longitudinal
directions (Fig. 1(b)). In this case, the local volume fraction of the bers is modeled by
X
f
r

X
f0
=

2R

at 0 < R < X
f0
=2
1 at X
f0
=2 < R < 1
A1
where R r=b is the nondimensional radial distance and X
f0
is the mean volume fraction of
the bers. The local eective thermal conductivity of the composite in the radial direction is
written as
k
c
r k
c0
Fr A2
where k
c0
is the eective thermal conductivity in the radial direction, corresponding to X
f0
: The
thermal conductivities k
c
r and k
c0
are given by
k
c
r

1 X
f
r

k
m
X
f
rk
f
A3
and
k
c0
1 X
f0
k
m
X
f0
k
f
A4
The substitutions of Eqs. (A3) and (A4) into Eq. (A2) yield
Fr

1 X
f
r

k
m
X
f
rk
f
1 X
f0
k
m
X
f0
k
f
A5
On the other hand, the dependence of the heat capacities, C
Pr
, on the radial distance can be
neglected because the dierence between the heat capacities for the bers and paran is about
10%. To this end, the heat conduction equation for the brush system is written as
@T
@t
a
0
1
r
@
@r

Frr
@T
@r

A6
where a
0
is the eective thermal diusivity corresponding to X
f0
: Instead of Eq. (1), Eq. (A6)
was used to estimate a
0
for several experiments. The dierences between the a values from Eq.
(1) and a
0
values from Eq. (A6) were less than 3%. Accordingly, Eq. (1) was used for the
brush type as well as the random type.
References
[1] Abhat A. Solar Energy 1983;30:31332.
[2] De Jong AG, Hoogendoorn CJ. In: den Ouden C, editor. Thermal Storage of Solar Energy. Hague/Boston/
London: TNO and Martinus Nijho, 1981. p. 12333.
[3] Knowles TR, Webb WG. In: Proceedings of AIAA 22nd Thermophysics Conference, Honolulu, USA, AIAA-
87-1489. 1987.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1555
[4] Hasnain SM, Gibbs BM. In: Proceedings of 2nd UK National Conference on Heat Transfer, University of
Strathclyde, UK. 1988. p. 83344.
[5] Al-Jandal SS. In: Proceedings of 2nd World Renewable Energy Congress, Reading University, UK. 1992. p.
1097105.
[6] Bugaje IM. Int J Energy Res 1997;21:75966.
[7] Stovall TK, Arimilli RV. In: Proceedings of 23rd Intersociety Energy Conversion Engineering Conference,
Denver, USA. 1988. p. 1717.
[8] Chow LC, Zhong JK, Beam JE. Int Comm Heat Mass Transfer 1996;23:91100.
[9] Ogura H, Miyazaki M, Matsuda H, Hasatani M. Kagaku Kogaku Ronbunshu 1991;17:91623.
[10] Ogura H, Miyazaki M, Matsuda H, Hasatani M, Yanadori M, Hiramatsu M. Kagaku Kogaku Ronbunshu
1992;18:66976.
[11] Kakiuchi H, Kubokawa S. In: Proceedings of 33rd National Heat Transfer Symposium of Japan, Niigata,
Japan, E215. 1996.
[12] Kunii D, Smith JM. AIChE J 1960;6:718.
[13] Watanabe F, Watabe Y, Katsuyama H, Kozuka J, Hasatani M, Marumo C. Kagaku Kogaku Ronbunshu
1993;19:8390.
J. Fukai et al. / Energy Conversion & Management 41 (2000) 15431556 1556