THE 19TH INTERNATIONAL CONFERENCE ON COMPOSITE MATERIALS

THERMAL CONDUCTIVITY MEASUREMENT OF GFRP

AT CRYOGENIC TEMPERATURE
T. Inoue1*, K. Hayakawa1, Y. Suzuki1
1
Dept. of Mechanical Sciences and Engineering, Tokyo Institute of Technology, Tokyo, Japan
* Corresponding author ([email protected])

Keywords: Thermal Conductivity, Glass Fiber Reinforced Polymer, FRP, Cryogenics,

Thermal Contact Resistance

1 Introduction mechanical and thermal properties of GFRPs at

Fiber Reinforced Polymer (FRP) has excellent cryogenic temperatures. He revealed discrepancies
mechanical characteristics, such as lightness and among the thermal conductivity data obtained by
high strength. Therefore, it has been extensively some researchers, and pointed out that one of the
investigated as an alternative material to metals, and causes of the discrepancies may be inaccuracy of
recently, it has been used as the structural material of their experiments. In addition, Radcliffe et al. [5]
aircraft, spacecraft, and hydrogen pressure vessels studied effects of fiber-direction and fiber-content of
for automobiles. In addition to such excellent GFRPs and carbon FRPs (CFRPs) on the thermal
properties, FRP has much lower thermal conductivity from 2 to 80 K. The thermal
conductivity than metals and is used as structural conductivity parallel to the fiber was 10-20% higher
material in cryogenic engineering. In designing than that perpendicular to the fiber and both were
cryogenic apparatus such as liquid hydrogen tanks, proportional to fiber-content. Dmitrevsky et al. [6]
both strength design and thermal design are measured thermal conductivities of glass-fabric-base
important to reduce the amount of heat flow from laminates and uniaxial-glass-fiber-base laminates in
the surroundings. the temperature range of 4 to 80K, and obtained
Kamiya et al. [1] studied a conceptual design of a trends similar to Radcliffe’s results. The matrices
large liquid hydrogen (LH2) storage tank. They they adopted were different but the both have almost
adopted Glass Fiber Reinforced Polymer (GFRP) for the same thermal conductivities and the fibers they
supports connecting the cryogenic tank and an outer adopted were E-glass. Therefore, although it was
chamber at room temperature. They also proposed a assumed that thermal conductivities of their
thermal insulation structure of large LH 2 tank, but unidirectional GFRPs should show almost the same
could not show the performance of the tank due to a values, they did not agree well. This result suggests
lack of thermal conductivity data of GFRPs. Tsubaki that we should measure thermal conductivity with
[2] also studied insulation structures of a LH2 tank of higher accuracy under cryogenic condition.
10,000 m3 and proposed the optimum insulation However, in most of the literature mentioned
structure using GFRP to achieve the boil-off rate of previously [3-6], measurement accuracies were not
0.16 wt% per day. However, because the rate is discussed enough.
strongly affected by the thermal conductivity of In the previous study [7], Tanaka analyzed errors in
GFRP, its accurate values from a room temperature measuring temperature after reducing the amount of
to about 20 K are required for the evaluation of the heat flow by convection and radiation, and pointed
boil-off rate. out that we should consider heat flow through the
Thermal conductivity of FRP will be affected by thermocouple wire. He tried to reduce thermal
factors such as fiber material, matrix, fiber direction, contact resistance (TCR) between the thermocouple
fiber content and so on. Reed et al. [3] have and the specimen by improving contact conditions
reviewed the mechanical and thermal properties of but failed to reduce it to a negligible value. In this
unidirectional composites with glass, carbon, boron, research, a temperature compensation method using
alumina, and aramid fibers at cryogenic temperature. TCR between the specimen and the thermocouple
Based on the ratio of strength to thermal was proposed [8] and the thermal conductivity of
conductivity, they concluded that GFRP is the best GFRP was actually measured between 20 and 80 K.
support material among the FRPs above 20K. Kasen
[4] reviewed many works of research on the
2 Temperature Compensation with Thermal specimen and the thermocouple inserted into a hole,
Contact Resistance and therefore the TCR cannot be negligible.
In cryogenic experiments, heat flow from the By considering the situation previously mentioned,
surroundings to the specimen should be reduced we change our way of thinking from reduction of the
enough. In general, the specimen is covered by TCR to acceptance of the resistance. Here, we
radiation shields and experiments are performed propose to compensate temperatures measured with
under vacuum condition. If proper countermeasures thermocouples by a temperature gap based on the
are taken, the amount of heat flow by those heat TCR. Equation 1 shows the relationship between the
transfers will be in negligible levels. However specimen temperature, TS, and the thermocouple
conduction through a conductive wire of the temperature, TTC. In order to use this compensation
temperature sensor still remains as a possible heat method, we must know the TCR, R, in advance and
flow mechanism. control the heat flow through the thermocouple into
A Silicon diode thermometer (SD thermometer) and the specimen, Q̇ con .
a ruthenium oxide resistant thermometer (RuO
thermometer) are commonly used as cryogenic TS = TTC - RQ̇ con (1)
temperature sensor. If we use a manganin wire as the
conductive sensor wire, because its thermal
conductivity is low enough, heat flow through the 3 Experimental
conductive wires can be negligible. But their sensor
heads are not small enough for the measurements of 3.1 GFRP Specimens
In this experiment, GFRP having the lowest thermal
local temperatures. An Au(0.07%Fe)/chromel
conductivity among FRPs above 20 K [3], was used
thermocouple (AF thermocouple) is also commonly
in the thermal conductivity measurements under the
used as a thermocouple for cryogenic temperature
cryogenic region. First, a unidirectional GFRP block
range. As it has a tiny sensor head, it is very suited
was prepared by the hand lay-up method with
for a measurement of a temperature distribution in a
laminated prepreg-sheets of 46% fiber-content (GE
specimen. However, as an Au(0.07%Fe) wire has
352H160S, Mitsubishi Rayon Co., Ltd.). Secondly, a
large thermal conductivity, we must take a proper
load was applied to the GFRP block by clamping it
countermeasure for temperature measurement.
with copper plates, and it was sintered in a
Generally, there is TCR between two solid surfaces.
thermostatic bath at 130 °C. Finally, two pairs of
In a room temperature range, the resistance is
pieces were cut out of the GFRP block. Each pair
usually small enough. But in cryogenic temperatures,
had the same fiber orientation, but consisted of two
it is not ignorable in many cases, because the
pieces of different lengths (30 mm as group “A” and
temperature difference between the specimen and
10 mm as group “B”). Two specimens with a fiber
the thermocouple based on the TCR becomes a
orientation of 0 degree to its longitudinal direction
comparable value with the temperature measured.
were called A0 and B0. Similarly, the other two with
To improve the contact condition, we usually apply
a fiber orientation of 90 degree were called A90 and
a load and make the resistance small enough. Figure
B90 (see Fig. 2). The cross-sectional area of the
1 shows a schematic image of a thermocouple
specimens was 14 mm2.
inserted into a specimen. It is obvious from this
figure that we cannot apply a load between the
Specimen A Specimen B
A0 B0
Fiber direction
Heat flow Qcon Fiber
Specimen
0 degree
Thermal contact
resistance Thermocouple
R Thermocouple SD thermometer

Specimen Measurement A90 B90

temperature temperature
TS TTC 90 degree

Fig. 1. Schematic image of temperature

measurement point. Fig. 2. CFRP specimens (Top view).
For the specimens A0 and A90, three holes of 1 mm S TTA
Q̇ con = ∫ λ dT
diameter were drilled to the center of specimen at L TTC Au (2)
7.5 mm intervals. In addition, bisphenol-A epoxy
resin, which is the same resin as the prepreg-sheets,
Here, Au is thermal conductivity of Au [9], TTC is
was poured into the holes. Then three calibrated AF-
the measured temperature of the thermocouple, TTA
thermocouples (TC1, TC2, and TC3) were inserted
is the temperature of the thermal-anchor, S is the
into the holes, and it was left at room temperature
cross sectional area of Au wire, and L is the wire
until the resin cured. After curing, the TCRs
length between specimen and thermal anchor. Once
between each thermocouple and the specimen were
expected to be constant, but not necessarily equal. the value of Q̇ con is obtained by solving Equation 2,
For the specimens B0 and B90, a hole of 2 mm the TCR can be determined by Equation 1.
diameter was drilled to the center of each specimen,
and a SD thermometer was inserted into the hole 3.3 Thermal Conductivity Measurement
with cryogenic grease (Apiezon-N grease). Using the temperature compensation, the thermal
conductivities of the GFRP specimens (A0 and A90)
3.2 Thermal Contact Resistance Measurement mentioned previously were measured with the one
In order to measure the thermal contact resistance, dimensional steady state method.
which is usually unknown, we developed a new
experimental system shown in Fig. 3. The specimens Q̇ dT
q̇ = =λ (3)
A and B, having the same fiber direction, are A dx
sandwiched tightly with aluminum plates. The
assembly was attached to a cold stage in a cryostat Figure 4 shows the schematics of the test section.
to establish a uniform temperature field on the test The cold stage and a heater were attached to the
section. The pressure in the cryostat was kept at specimen. In order to improve the contact conditions
about 1.3×10-3 Pa. In addition, as the manganin of the specimen, silver paste was used in between
conductive wires for the SD thermometer were the cold stage and the specimen and Apiezon-N
attached to the aluminum plates, the SD thermo- grease in between the specimen and the heater. The
meter could accurately indicate the specimen specimen was insulated by an aluminum radiation
temperature. On the other hand, due to the TCR and shield under vacuum condition, and then the heat
heat flow through the thermocouple, a temperature transfer by radiation and convection became
difference between the specimen and the negligible.
thermocouple occurred. The amount of heat flow, The specimen temperature was controlled by
changing the cold stage temperature from 20 to 80 K.
Q̇ con , through the thermo-couple was controlled
The heater output was in between 15 and 30 mW.
with varying the thermal anchor temperature. It was
The thermal anchor was common to all
assumed that the heat flow only occurred through
thermocouples measuring the temperatures at the
the Au (0.07%Fe) wire, and it can be obtained from
different points of the specimen and its temperature
solving Equation 2.
was set higher than the specimen temperatures. The
amount of heat input through each thermocouple
was in between 0.5 and 1.5 mW. Two of the three

Specimen A Thermocouple SD thermometer Specimen B

TC1, TC2, TC3
Radiation shield
Heater
T : uniform Specimen
Qsd= 0 (GFRP) Thermal anchor
Al-Plate TC1
Qcon TC2
Cold stage
TC3
Thermal anchor: TA Cold stage

Fig. 3. Test section for thermal contact resistance Fig. 4. Test section for the thermal conductivity
measurement. measurement.
thermocouples were selected to estimate the thermal temperature differences, TTC – TS, with respect to the
conductivity with Equation 3. Here, the heat flow amount of the heat flow through the
rate, Q̇ , was considered as the summation of the thermocouples, Q̇ con . The data seemed to be
heat flows from the heater and through each scattered as a whole, and maximum temperature
thermocouple. difference was about 5 K. However, the straight
lines fitting of each data set could be distinguished
4 Results and Discussions as a line from the origin or not. The data of TC1 and
TC3 for A90 could be approximated as the lines
4.1 Thermal Contact Resistance from the origin, and those results indicated that TCR
Thermal contact resistances of the specimens A0 and is constant and independent of temperature. On the
A90 were measured in the temperature range from other hand, the TC2 data for A90 and the data of all
20 to 80 K. The temperatures of specimens B0 and the thermocouples for A0 could not be approximated
B90, which were measured by the SD thermometer, as lines from the origin. From the results, we
matched the cold stage temperature. Thus, the consider two possibilities. The first is that the TCR
uniform temperature field could be established on in Equation 1 will change with temperature. The
the test section. Therefore, we assume that the SD other is that the TCR is constant but the contact
thermometer indicates the temperatures of condition will change with the temperature. Near
specimens A0 and A90. Figure 5 shows the room temperature TCR is generally considered
constant but there may be a possibility that TCR will
change with temperature in low temperature region.
5 In order to examine the temperature dependency, the
data of TC1 for A0 were divided into temperature
4 ranges of TTC by every 10 K as shown in Fig. 6. The
slopes of the fitting lines seem to be almost the same
TTC - TS [K]

3 for all the temperature ranges, and the y-intercepts

increase as the temperature decreases. Considering
2 A0 the results, we adopt Equation 4 instead of Equation
TC1 1. The average value of the slopes can be considered
1 TC2
as R in the equation. We don’t know whether we
TC3
0 should call the whole of these terms as TCR or not,
0.0 1.0 2.0 3.0 but for convenience, hereafter, we will refer to this R
.
Qcon [mW] as TCR.

(a) Fiber direction: 0 degree (A0)

5

4
5 A90
TTC - TS [K]

TC1 3
4 TC2
TTC - TS [K]

TC3 2
3
1
2
0
1 0.0 1.0 2.0 3.0
.
Q
0 con [mW]
0.0 1.0 2.0 3.0
.
Qcon [mW]
20K - 30K 30K - 40K
TTC 40K - 50K 50K - 60K
(b) Fiber direction: 90 degree (A90) 60K - 70K 70K - 80K

Fig. 5. Temperature difference, TTC – TS, versus Fig. 6. Temperature difference, TTC – TS, for each
heat flow rate through thermocouple, Q̇ con . temperature range (A0, TC1).
 dependent term, we observed the cross sections of
TS = T TC - {RQ̇ con + f (T )} (4) the specimen with a scanning electron microscope.
Figure 9 shows the cross sections of the specimens
Next, we investigated the form of f (T). Figure 7 around the junctions of thermocouples. In the photos
clearly shows it can be expressed as a linear of TC1 and TC3 for A90, we observed a void or a
equation of thermocouple temperature, TTC. Then, crack around the thermocouples. But in the others
Equation 4 can be rewrite as Equation 5. we couldn’t recognize them clearly. Of course, from
these photos, it may be very hard to say that the
TS = T TC - {RQ̇ con + (C1 ∙ T TC + C2 ) } (5) cause is contact conditions between the specimen
and thermocouples. But it might be necessary to
Here, the constants, C1 and C2, were determined by study an effect of thermal stress on
the least-square method. To confirm consistency of thermoelectromotive force.
this procedure, we rearranged the data by subtracting We estimated the maximum errors, ΔTmax, and the
this term from the temperature difference, TTC – TS. mean square deviations, ΔTrms, of all the data after
Figure 8 shows the scattered data was well the compensation for each thermocouple by
rearranged and could be expressed by a linear Equations 6 and 7. In the equations, TS is the
equation from the origin. The slope of this line, R’, specimen temperature and TS* is the compensated
should correspond to the TCR. Table 1 shows the temperature of thermocouple. The results for each
values of R, C1, C2, and R’ for each thermocouple. thermocouple were shown in Table 2. By this
As the values of R and R’ agree well, we think this compensation method, the maximum errors were
procedure is reasonable.
In order to study the causes of the temperature- ΔTmax = max | T*S - TS | (6)

5
( TTC - TS ) - RQcon [K]

4 Table 1. Constants in Equation 5.

. 3 R C1 C2 R’
[K/mW] [K/K] [K] [K/mW]
2 TC1 0.267 –0.0369 4.327 0.264
1 A0 TC2 0.315 –0.0200 2.646 0.315
TC3 0.194 –0.0235 2.975 0.191
0
TC1 0.641 — — —
20 40 60 80
TT C [K] A90 TC2 0.750 –0.0309 2.959 0.751
TC3 1.293 — — —
Fig. 7. Temperature-dependent term (A0, TC1).

TC1 TC2 TC3

5
( TTC - TS ) - (C1TTC - C2 ) [K]

4 A0

3 Drilled hole Thermocouple GFRP

2
A90
1

0 Crack Thermocouple ? Void

0.0 1.0 2.0 3.0
.
Qcon [mW]
Fig. 9. SEM images of the cross sections of GFRP
Fig. 8. Thermal contact resistance (A0, TC1). specimens around temperature measurement points.
 proposed is effective to improve temperature
n
1 2 measurement at low temperatures.
ΔTrms =√ ∑(T*S - TS ) (7)
n
4.2 Thermal Conductivity
reduced from 5 K to less than 0.7 K and the mean Figures 10 and 11 show the thermal conductivities
square deviations were about 0.3 K. Therefore, we of the specimens A90 and A0, without and with the
can conclude the temperature compensation method temperature compensation in the cryogenic
temperature region. In the figures, three thermal
Table 2. Errors of temperature measurement conductivities are shown according to a pair of the
with the compensation. thermocouples. The subscript of each  represent a
pair of the thermocouples selected for the calculation.
[K]
A0 A90 In addition, the total errors of 13 at about 70 K is
TC1 TC2 TC3 TC1 TC2 TC3 shown in the figures. Although the three thermal
ΔTmax 0.35 0.54 0.47 0.58 0.68 0.63 conductivities for the specimen should agree, those
for A90 without temperature compensation showed
ΔTrms 0.20 0.23 0.22 0.31 0.29 0.23 unignorable differences, but those obtained with the
compensation agree well (Fig. 10). On the other

0.4 0.4
12
0.3 0.3 23
·K)]

·K)]

13
 [W/(m•

 [W/(m•

0.2 0.2
12
0.1 23 0.1
13
0.0 0.0
20 40 60 80 20 40 60 80
Temperature [K] Temperature [K]

(a) without compensation (a) without compensation

0.4 0.4
12
0.3 23 0.3
·K)]

·K)]

13
 [W/(m•

 [W/(m•

0.2 0.2
12
0.1 0.1 23
13
0.0 0.0
20 40 60 80 20 40 60 80
Temperature [K] Temperature [K]

(b) with compensation (b) with compensation

Fig. 10. Thermal conductivities perpendicular to Fig. 11. Thermal conductivities parallel to
the fiber direction (A90). the fiber direction (A0).
hand, differences among the three thermal conductivities, 13, of A0 and A90 and the thermal
conductivities of the specimen A0, were fairly small conductivity parallel to the fiber, A0, is 10% larger
regardless of the compensation, the total than that perpendicular to the fiber, A90. In the
measurement error was reduced so much by the previous studies, measurement accuracy was
compensation (Fig. 11). Anyway, the thermal insufficient to study the effect, but as the
conductivities of the GFRP increase from 0.1 to 0.3 temperature compensation method improved it
W/(m∙K) with the temperature from 20 to 80 K. significantly, we could show the effect clearly.
The total error of the thermal conductivity Some models for estimating thermal conductivities
measurement is composed of temperature of composites material have been proposed.
measurement error, length measurement error, and Equation 8 is a typical equation for the thermal
error of the heat flow rate in the specimen. As the conductivity parallel to the fiber (Parallel model),
effects of the radiation and the heat conduction can and Equation 9 for that perpendicular to the fiber
be considered negligible, the error of the heat flow (Rayleigh model [11]).
rate is also negligible small. The length
measurement error seems to be mainly caused by a λ = Vf λf + (1-Vf ) λm (8)
thermal shrinkage of the specimen and the most
important factor is the temperature measurement
error. The error of the temperature measurement, ΔT,
with the temperature compensation is ΔTrms 2Vf
λ = λm (1- ) (9)
mentioned previously. On the other hand, ΔT 3V 4
ν + Vf - 2 f 4 (0.032π4 )2
without the compensation is assumed as the mean ν π
temperature error in the range where Q̇ con is smaller
In these equations, f is the thermal conductivity of
than 1.5 mW in Fig. 5, because 1.5 mW is the
maximum heat input through the thermocouple. The the fibers, m is the thermal conductivity of the
total error of the thermal shrinkage is simplified as a matrix, Vf is the volume content of the fibers, and 
sum of the shrinkage of the triaxial components of is a coefficient defined by the thermal conductivities
the matrix from the room temperature to the working of the fiber and matrix ( ν = (λm+λf )/(λm-λf ) ).
temperature [10]. The error of each factor and the Figure 13 shows a comparison of the measured and
total errors of 13 is shown in Table 3. Here, Δ13/13 the estimated thermal conductivities. The measured
is the total error of 13, Q is the heat flow rate in the thermal conductivities were the same in Figure 12.
specimen. The maximum total error of 13 can be In the estimation, Vf = 0.46 was assumed and the
reduced from about 60 % to 15%. Those of both 12 thermal conductivities of E-glass by Radcliffe [5]
and epoxy resin in the Reference 12 were used. The
and 23 with the compensation become about 25%.
estimated values of the thermal conductivities for A0
Then, effect of the fiber direction on the thermal
and A90 don’t agree with the measured values. On
conductivity of GFRP specimens was investigated.
the other hand, Domitrevsky reported a different
As 13 was the most accurate, it was used in the
value for glass fiber. Although the fibers in the
following discussion. Figure 12 shows the thermal
prepreg-sheets we used are E-glass, there might be

Table 3. Errors of 13 with or without the temperature compensation.

ΔTTC1 ΔTTC3 Thermal

Temp. Total error
Q T ΔTTC1 ΔTTC3 shrinkage
[mW]
compensa
[K] TTC1 -TTC3 TTC1 -TTC3 (total) 13 / 13
[K] [K]
-tion [%] [%] [%]
[%]
without 20-40 2.22 1.62 23.0-32.3 16.8-23.5 4.1 44-61
~15
with 20-40 0.20 0.22 2.9-4.1 3.2-4.5 4.1 10-14
A0
without 30-85 2.22 1.62 15.1-28.3 11.0-20.6 4.1 30-55
~30
with 30-85 0.20 0.22 1.8-3.2 2.0-3.6 4.1 8-13
without 30-65 0.60 1.13 7.2-11.7 13.5-22.0 4.1 32-39
~15
with 30-65 0.31 0.23 3.0-4.8 2.2-3.6 4.1 10-15
A90
without 40-100 0.60 1.13 4.5-8.0 8.5-15.0 4.1 18-30
~30
with 40-100 0.31 0.23 2.0-4.0 1.5-3.0 4.1 8-14
 0.4 0.8 A0 Parallel model
A90 Rayleigh model
A0
0.3 A90 0.6
E-glass [5]
·K)]

·K)]
 [W/(m•

 [W/(m•
0.2 0.4

0.1 0.2
Epoxy [12]
0.0 0.0
20 40 60 80 20 40 60 80
Temperature [K] Temperature [K]

Fig. 12. Effect of Fiber direction on Fig.13 Thermal conductivities estimated.

thermal conductivity of GFRP.
3. R. P. Reed and M. Golda "Cryogenic properties of
some possibilities that the thermal conductivity of unidirectional composites". Cryogenics, Vol. 34, No.
the fibers have different value, because of the 11, pp 909-928, 1994.
composition and the crystallinity degree. For further 4. M. B. Kasen “Mechanical and thermal properties of
discussions on a validity of those estimation filamentary-reinforced structural composites at
methods, we need to know the accurate value of the cryogenic temperatures 1: Glass-reinforced
thermal conductivity of glass fiber in GFRP. composites”. Cryogenics, Vol. 15, No. 6, pp 327-
349, 1975.
5. D. J. Radcliffe and H. M. Rosenberg "The thermal
5 Conclusion conductivity of glass-fibre and carbon-fibre/epoxy
A temperature compensation method based on the composites from 2 to 80 K". Cryogenics, Vol. 22 No.
thermal contact resistance between the specimen and 5, pp 245-249, 1982.
a thermocouple was proposed. The TCR between a 6. Yu. P. Dmitrevsky, S. S. Kozub and U. Escher
GFRP specimen and an AF thermocouple was “Thermal conductivity of various glass-reinforced
measured. Thermal conductivities of GFRP were plastics at temperatures below 80 K,” Cryogenics,
measured in the temperature range from 20 to 80 K Vol. 27, No.8 pp 429-432, 1987.
7. S. Tanaka “Accuracy improvement of thermal
and are 0.1-0.3 W/(m∙K). Effect of the fiber- conductivity measurement of FRP in cryogenic
direction is clearly shown that thermal conductivity region,” Bachelor's Thesis, Department of Mechano
of GFRP for 0 degree is about 10% larger than that Aerospace Engineering, Tokyo Institute of
for 90 degree. Technology, March 2011 (in Japanese).
With the temperature compensation, the maximum 8. T. Inoue, K. Hayakawa, S. Tanaka and Y. Suzuki “A
temperature measurement errors could be reduced Method for Accurate Temperature Measurement for
from 5 K to 0.7 K, and the mean-square-deviations GFRP Thermal Conductivity in Cryogenic Region”.
of the error was 0.3 K. Proceedings of the 3rd International Forum on Heat
Transfer, Japan, No. 100, 2012.
9. Y. S. Touloukian ed. “Thermophysical Properties of
Reference High Temperature Solid Materials Volume 1:
1. S. Kamiya, K. Onishi, N. Konshima and K. Elements”.1st edition, Macmillan, 1967.
Nishigaki “Thermal test of the insulation structure 10. G. Hartwig “Thermal expansion of fibre
for LH2 tank by using the large experimental composites”. Cryogenics, Vol. 28, No. 4, pp 255-
apparatus”. Cryogenics, Vol. 40, No. 11, pp 737-748, 266, 1988.
2000. 11. L. Rayleigh “On the influence of obstacles arranged
2. K. Tsubaki “Thermal analysis and optimization of in rectangular order upon the properties of a
the insulation structure for large LH2 storage tank”. medium”. Philosophical Magazine, Vol. 34, No. 211,
Master's Thesis, Department of Mechanical Sciences pp 481-502, 1892.
and Engineering, Tokyo Institute of Technology, 12. S. Kelham and H. M. Roscenberg “The thermal
February 2008 (in Japanese). conductivity and specific heat of epoxy-resin from
0.1-80 K”. Journal of Physics C: Solid State
Physics, Vol. 14, No. 12, pp 1737-1749, 1981.