Advanced Energy Conversion. Vol. 2, pp. 241-248. Pergamon Press, 1962.

Printed in Great Britain

THERMOELECTRIC AIR CONDITIONER FOR SUBMARINES*

J. R. ANDERSENt
Summary--This paper summarizes the results thus far obtained in a program of design and experimental
development of a thermoelectric air conditioner for submarines. The basic design concept, intended to
satisfy the application criteria, is discussed in terms of performance, pressurization, sea water corrosion, and
design simplicity. A working theory which expresses the performance of a "real" thermoelectric refrigeration
machine leads to a discussion of the design parameters for such a machine. The results of design optimization
studies are then displayed showing, specifically, the size of the machine and the weight of thermoelectric
material required as a function of maximum coefficient of performance for a variety of thermopile configura-
tions. The final design, based on these considerations, is presented and its expected performance is discussed.
A one-ton air conditioner model based on this design has been constructed. Experimental performance data
so far obtained is discussed and compared with predicted performance. Some difficulties in implementing the
design concepts in practice are mentioned. The conclusion is that large capacity, compact thermoelectric
air conditioners having adequate performance are now practical at the present state of the thermoelectric
art.
NOMENCLATURE
a = Seebeck coefficient (V/°C)
i = current density ( A / c m 2)
k = t h e r m a l conductivity o f thermoelectric m a t e r i a l ( W / c m °C)
k~ = t h e r m a l conductivity o f e n c a p s u l a n t ( W / c m °C)
l = t h e r m o e l e m e n t length (cm)
p = p o w e r i n p u t density ( W / c m z)
q = refrigeration density ( W / c m z)
R = c o n t a c t resistance (t2cm 2)
To ~- h o t j u n c t i o n t e m p e r a t u r e (°K)
T1 =- cold j u n c t i o n t e m p e r a t u r e (°K)
To' -----h o t fluid t e m p e r a t u r e (°K)
7"1' = cold fluid t e m p e r a t u r e (°K)
Z = figure o f merit o f a thermoelectric m a t e r i a l (°K) -1
/31 = cold side over-all h e a t transfer coefficient ( W / c m 2 °C)
/30 = h o t side over-all loot transfer coefficient ( W / c m 2 °C)
¢ = coefficient o f p e r f o r m a n c e
v~ = t h e r m o p i l e p a c k i n g density
p = resistivity o f thermoelectric m a t e r i a l (O cm)

INTRODUCTION
THE OBJECTIVE o f this p r o g r a m was to p e r f o r m an a p p l i c a t i o n analysis, design, a n d experi-
m e n t a l d e v e l o p m e n t o f a thermoelectric air conditioner for use on submarines. Specifically,
it was p r o p o s e d to design a central refrigeration p l a n t to replace the m e c h a n i c a l unit in a n
existing s u b m a r i n e air conditioning system. This p l a n t w o u l d p r o d u c e chilled water a t
50°F with an inlet t e m p e r a t u r e o f 55°F, a n d the heat w o u l d be dissipated to sea water at
a m a x i m u m inlet t e m p e r a t u r e o f 85°F.

*Supported by U.S. Navy, Bureau of Ships, Contract No. NObs. 77123.

tApplied Research, Defense Electronic Products, Radio Corporation of America, Camden, New Jersey.
241
242 J.R. ANDERSEN

BASIC DESIGN CONCEPT

The design configuration that appeared to best fulfill the requirements was a unit in the
form of a modified shell-and-tube heat-exchanger, the thermoelectric couples being built
into the tubes. This concept is illustrated in Fig. 1. The chilled water flows inside the tubes
and the sea water is on the shell side. The final design consists of several identical units,
each having a nominal capacity of 10 tons of refrigeration. Each unit will be contained in
a pressure vessel 20 in. in diameter and 10 ft long and will contain 30 identical tube sub-
units, substantially identical to the units used in the model to be described later.

IDEALIZED THERMOELECTRIC REFRIGERATOR

The basic relations for the steady performance of an idealized thermoelectric refrigerator
are well known. Assuming that the thermoelectric properties are independent of tempera-
ture, these relations are
k (7"0 -
q = - - ½ pli 2 -4- a T l i - - -i- rl) (1)

p = pli 2 + a(To - - 7'1) i (2)

¢ - q (3)
P
The first term of equation (1) expresses that part of the Joule heat generation caused by
the current passing through the arm that is returned to the cold junction. The other half
of the Joule heat appears at the hot junction. The second term is the Peltier cooling at the
cold junction. The third term is the rate of heat conduction to the cold junction due to the
temperature difference between the hot and cold junctions.
The first term of equation (2) is the total rate of Joule heat generation. The second term
is the power required to overcome the Seebeck voltage generated by the temperature
differential between the hot and cold junctions of the circuit. Equation (3) is the definition
of the coefficient of performance of a refrigeration machine.
While equations (1) through (3) completely describe the performance of an idealized
thermoelectric refrigeration machine, they are not in suitable form to display the per-
formance characteristics in terms of convenient experimental parameters.
The relevant parameter is the figure of merit, Z, of an idealized thermocouple, defined
by Z = a2/pk. T h e maximum cooling for a given temperature differential, (To -- T0, is
obtained with a current density for which
dg _ 0
di
This gives i = aT1/pl. The maximum temperature differential obtainable, (To -- T1)~ax, 'is
then obtained by setting q = 0. The result is
I a2
(Z0 -- Zl)max = -~-~ Zl (4)
or
(To -- T1)max = ½ Z ~ (5)
The figure of merit is an experimental measure of the quality of thermoelectric materials
and is easily obtained by measurement of the maximum temperature differential under
conditions of no net heat transfer at the cold junction.
 Thermoelectric Air Conditioner for Submarines 243

In a similar way, an expression for the maximum coefficient of performance of an

idealized thermoelectric refrigerator is obtained in the well known form
5bmax __ T1 (1 ÷ Z(To ÷ T1)/2) 1/2 -- To~T1 (6)
(To -- 7"1) (1 ÷ Z(To q- I'1)/2) 1/2 ÷ 1

Performance o f a "real" thermoelectric refrigerator

Equations (5) and (6) show that the performance of an idealized thermoelectric refri-
gerator is independent of the length of the arms. This conclusion is not true if the contact
resistance at the junctions is considered. Including the contact resistance, assumed equally
divided between the hot and cold junctions, equations (1) and (2) become respectively

g = -- ½ (pl q- 2R) i 2 ÷ aTli -- K (To -- T1) (7)

p ---- (pl ÷ 2R) i 2 + a(To -- I"1) i (8)

These equations permit definition of an effective figure of merit, Zeer defined as
Z

such that the maximum temperature differential obtainable is

(To -- T1)max -~ ½ Zeff T~ (10)
and the maximum coefficient of performance is
_ T1 (1 + [Zeer (To + I"1)]/2) 1/z - - To/T1 (11)
~max ( T 0 ~ ) (1 ÷ [Zar (To + T1)]/2) lie + 1
Note that, with the assumption of equal contact resistance at the hot and cold junctions,
the introduction of the effective figure of merit (defined by equation (9)) leads to expressions
identical in form to the idealized-performance equations. The above relations assume that
the thermoelectric properties are independent of temperature. However, equation (10) can
be used as a definition of a mean figure of merit over the temperature range (To -- T1)max
and with this understanding, equation (11) can be used to explore with sufficient accuracy
the performance characteristics of a real thermoelectric refrigerator.
Making use of the definition of effective figure of merit, equation (9), we can eliminate
the current density, i, between equations (7) and (8). The resulting expression can be put
in the form
i[ T1)]j 1/2 _ 1}.
q/(k/l)(To - - T 1 ) = [Zo~ (T0 ÷ T1)]/4 [/1 ÷ 4p/(k/l)(To~____T~--
-- [p/2(k/l)(To -k /'1)] -- 1 (12)
This relation suggests the definition of the following dimensionless parameters

q (13)
q* -- k/l(To -- T1)

p* -- P (14)
k / l ( r o - T1)
0 = Zoo(to + T1)/4 05)
244 J.R. ANDERSEN

By using these parameters, we can put equation (12) in the form

(q* + 1)/0 =
[1 + t~P *"~'[(T°/T1)+I)]1/2
/ ~ ) ~ _ -- (p*/20) -- i (16)

The virtue of equation (16) is that the dimensionless refrigerating capacity, (q* + 1)/0,
can be calculated in terms of the dimensionless power input, p*/O, and the operating
absolute temperature ratio, To~T1, without explicit reference to any specific properties of
the thermoelements. Equation (16) is displayed graphically in Fig. 2.

50

40
O3
ta.I
.J
Z
0 30
03
Z
U.I
2
20

IO

0 1 \ ~I \ I N 1 I I I
0 50 I00 150 200 250 300 550 400
p~
•~- (DIMENSIONLESS)

FIG. 2. Generalized thermoelectric performance function

The following expressions for performance at particular operating conditions follow

from equations (13) through (16).
The cooling capacity at maximum coefficient of performance is

(q*)Qmax : 4 0 { M - [(To~T1) + 1]/2} (17)

(M -- 1)2 [(To~T1) -~ 1]
where
M = (1 + 20) 1/2
The maximum coefficient of performance is
•max = {[II(TolT1) - - I][(M -- ( T o l T I ) ) I M + 1] } (18)
The maximum cooling capacity is
(q*)max = {201[(TolT1) ~ i 1 ] } - 1 (19)
The power required at zero cooling capacity is
(p*)a.:0 = { ( 4 0 W ) I [ ( T o / T 1 ) 2 - - I]}{[(TolT1) - 11 + w} (20)
where
W = 1 - - {I - - [2([To/TI] 2 - - 1)/0]} 1/z

The cooling capacity at zero power input is

(q*)v,=-o = -- 1 (21)
FIG. 1. Thermoelectric air conditioner--conceptual design.

[facing p. ~ ]
FIG. 5. Thermoelectric air conditioner sub-unit.
 ThermoelectricAir Conditionerfor Submarines 245

Equations (16) through (21) are in convenient form to explore the performance charac-
teristics and control aspects of a thermoelectric refrigerator.
The foregoing performance equations are expressed in terms of junction temperatures.
Although these equations are adequate for some design considerations, one must ultimately
consider the effect of the over-all heat transfer coefficients at the hot and cold junctions in
order to determine adequately the performance of a “real” thermoelectric refrigeration
machine. As has been seen the performance of a module, with known junction tempera-
tures, depends only on the effective figure of merit and current density. When a module and
associated heat transfer structures are operated between fluids of known temperature, the
packing density of thermoelements in the module becomes an important parameter of
thermopile design. The over-all heat transfer coefficient, based on unit area of thermo-
element material, is inversely proportional to the packing density for a given heat transfer
structure. When performance is required to be expressed in terms of fluid temperatures, two
additional relations must be satisfied
4= w77m; - Tl) (22)

4 + P = (rBOlrlPP0 - q (23)
and equation (7) is modified to become
q = - + (pl + 2R) i2 + aTli - (ke/l)(TO - 271) (24)
where k, = k[l + kp/k(l/yp - l)]. The back flow of heat through the interstices between
the thermopile elements is represented by k,.
Success was not achieved in obtaining a compact representation of performance in
terms of given fluid temperatures such as was possible using junction temperature. There-
fore, a scheme of successive approximations was used to obtain the maximum coefficient
of performance for a given set of design parameters.

THERMOPILE OPTIMIZATION
Of all the parameters appearing in the performance equations, only two can be arbi-
trarily chosen by the designer. These are the arm length, I, and the thermopile packing
density, qp. The hot and cold fluid temperatures are given. The thermoelectric material
properties and the junction contact resistance are determined by the state of the thermo-
electric art. An extensive experimental evaluation of the performance of available thermo-
electric materials and an investigation of couple joining techniques led to the choice of the
following properties as representative mean values of the best that can now be achieved:
Z = 2.80 x 10-a “K-l
k = 0.017 W/cm “K
p = 10.0 X 1O-4 Sz cm
a = 218 x 1O-6 V/OK
R = 10 x 10-G Q cm2
k, = 0 ~0003 W/cm “K

The hot and cold side over-all heat transfer coefficients require further discussion.
The most important one, as well as the one must subject to limitation by structural re-
quirements over-all heat transfer coefficient at the hot side, or sea-water side. The preferred
material is 70-30 Cupro-nickel and the calculations are based on its properties. The thermal
246 J. R. ANDE~SEN

resistance of the electrical insulation between the connector bars and the heat-transfer
structure is another important limiting factor on the over-all heat transfer coefficient. The
scheme adopted was to use a layer of aluminum oxide (Martin hardcoat) formed on an
aluminum plate and attached to the thermopile by a thin layer of epoxy cement. The fluid
velocity between the finned heat-transfer surfaces was adjusted to give the optimum balance
between increased surface heat transfer coefficient and pressure drop. The over-all maximum
coefficients of performance presented include the fluid pumping power as part of the power
input to the machine.

I’
/,T /
8’
,’ IQ7CM I’
.’ ,’
?)p=0.9,-' ,I’
C*
*cc m?3cM ,,’
*--
__-* _.-*---Ap=o.3
__-- --‘:.0.3cM _______---*
I I 1 I I
1.1 I.4 I5
&0MP0S,T:.3C0P

FIG. 3. Thermoelectric material and coefficient of performance vs. module parameters.

3.0 -

2.5 -
b

PE 2.0 _

I I I I

I.2 I.3 I.4 I*5

4” COMPOSITE COP

FIG. 4. Volume and coefficient of performance vs. module parameters.

The results of these optimization computations are shown in Figs. 3 and 4. Figure 3
shows the weight of thermoelectric material required per ton of refrigeration and the
over-all maximum coefficient of performance as function of the two module design para-
meters, thermoelement length and module packing density. Figure 4 shows the volume per
ton of refrigeration and the overall maximum coefficient of performance as function of
these same two parameters. It is interesting to note the large changes in performance,
volume and amount of thermoelectric material required for a given refrigeration capacity
resulting from relatively small changes in thermopile configuration. The final configuration
chosen as best compromise between these values was I = 0 -5 cm and rlP = 0 - 5.
 Thermoelectric Air Conditioner for Submarines 247

FINAL DESIGN AND MODEL

The final refrigeration plant will consist of several individual units, each having a
nominal cooling capacity of ten tons of refrigeration at design conditions, containing 30
identical tube units in a pressure vessel 20 in. in diameter and 10 ft long.
A one-ton model of this design has been constructed. This model contains six tube
units, identical in design to that for the final unit, 5 ft long, and contained in a pressure
vessel 10 in. in diameter. One tube unit partially assembled is shown in Fig. 5. Each 5 ft
tube of the one-ton model contains 48 modules, each module containing 18 couples. Each
module is two inches square, with integral heat-transfer extended surface on the chilled-
water side. The thermoelectric arms are 0.7 cm in diameter and 0.5 cm long. The sea-water
tube structure is aluminum bronze 3.06 in. by 2.29 in. in cross-section. The hot-side fin
plates and end-bells are pressure-sealed with O-rings. The modules are electrically connected
in series and have a plug-in design for easy disassembly and maintenance.

EXPERIMENTAL PERFORMANCE

One tube of the one-ton model has been tested for performance. Figure 6 shows the
performance data thus far obtained with this tube and permits comparison with the per-

Y ,
I 2 3 4
P0WER.P kW

FIG. 6. Anticipated and observed performance characteristics.

formance predicted by the design computations. The substantial agreement between the
predicted and realized performance is gratifying.

DIFFICULTIES

The only difficulties encountered in implementing this design were some minor water-
sealing problems on the chilled-water side and unreliability of the electrical insulation on
the thermoelectric modules. The sealing problem has been corrected. A short circuit to
ground through the thermopile insulation on one of the modules occurred when the tube
was being operated near its maximum cooling capacity and after the tube had been under
248 J. R. ANDERSEN

test for about 40 hr. The adequacy of aluminum oxide films for electrical insulation is being
investigated.

CONCLUSION

The conclusion is that practical, large capacity, compact thermoelectric refrigeration

machines having adequate performance can now be built utilizing current thermoelectric
art, and this performance can be reliably predicted.

Acknowledgement-The project on which this report is based, was supported by the U.S. Navy, Bureau
of Ships. Contract NObs 77123. This support is gratefully acknowledged. In addition, the work could not
have been accomplished iyithout the contributions of my co-workers Dr. L. Pessel, E. Farmer, R. Johnston,
M. Crouthamel, B. Shelpuk, and H. Walter.