8/2/2019 Basics Tec1
1/22
PRINCIPLES OF THERMOELECTRIC COOLING
1. INTRODUCTION
The operation of thermoelectric coolers is based on the phenomenon known as Peltier
effect. This is observed when an electric current flows across a junction of two dissmilar
conductors . The junction region is then found to absorb or evolve heat depending upon thedirection of current flow. In the circuit made of two conductors 'A' & 'B', one of the two junctions
will get heated while the other will be cooled. Electric current, thus, pumps heat out of one
junction, which is consequently cooled and deposits it into the other junction which gets heated.The phenomenon is completely reversible and a reversal of the current direction will interchange
the role of hot and cold junctions.
Peltier effect is one of the three thermoelectric effects which are thermodynamically
interrelated. These are known as Seeback, Peltier and Thomson effects and the threecorresponding thermoelectric coefficients are defined as follows fig. [ 1.1 ].
T1> T2
Fig 1.1
(i) Peltiercoefficient ( ) ; It is the ratio of the rate of heat absorbed or evolved(Q) at ajunction to the current (I) flowing across the junction ,
that is,
= Q/I 1.1(a)
VA
B
B
T1
T2
A
B
B
TC
I
TH
8/2/2019 Basics Tec1
2/22
(2) Seebeck coefficient ( AB): The two junctions are maintained at different temperatures a
thermo emf operates around the circuit giving rise to a closed circuit current or open circuit
voltage. This thermo emf measured as open circuit voltage (V) divided by the temperature
difference between the two junctions is known as Seeback Coeffcient. Thus
=
V
T T H C
3.1 b
AB = V/( TH-TC ) 1.1(b)
(3) Thomson Coefficient (): While the first two coefficients and have been defined for a pair ofconductors A & B, the Thomson coefficient is defined for the individual conductors A & B
separately. When an electric current (I) flows in a conductor across which a temperature gradient(dT/dx) has been maintained parallel to the direction of the current flow, there is reversible
absorption or evolution of heat along the conductor. This is known as Thomson effect. Thomson
Coefficient () is defined as the rate of heat absorbed or evolved per unit length, current andtemperature gradient. That is
1.1c/.
/
dxdTI
dxdQ=
The three coefficients are related thermodynamically through the Kelvin relations given
below :
AB = T . AB 1.2a
A-B=T. dAB/dT 1.2b
Absolute values of Seeback & Peltier coefficients are also defined for individualconductors when the other conductor of the pair is taken to be a superconductor. In that case these
relations can be expressed as :
= T 1.2c
= T. d/dT 1.2d
8/2/2019 Basics Tec1
3/22
2. Basic Thermoelectric Cooler :
The functioning of a thermoelectric cooler can be discussed with reference to a single
thermocouple structure shown in fig 1.2.In a preliminary analysis the Thomson Coefficient (),
being comparatively small in magnitude, is neglected. The two legs of the thermocouple areusually chosen to have n & p type conductivities since their Seeback & Peltier Coefficients have
opposite signs resulting in additive thermo-electric effects.
Fig 1.2
The basic equation for heat pumping in an elementary TE module such as the one shown infig 1.2 expresses the net heat (QC) pumped out per second at the cold junction at temperature (TC)
as the balance of Peltier extracted heat ( ABI) and the heat input originating from joule heating
and thermal conduction from the hot to the cold end. Equation 1.3 represents this relationship.
Q = AB.I- [I2 R/2 +K(TH - TC }] 1.3
Here 'R' is the total electrical resistance offered by the two legs and 'K' is their combined thermal
conductance. Both R & K contain dimensional parameters for lengths (l1 & l2) and areas of cross
section (A1 & A2) for the two legs. Explicit relations for 'R' & 'K' are of the type.
P N
+
TH
TH
TC
QC
V
I
8/2/2019 Basics Tec1
4/22
R = l/A and K= A/l 1.3a
where , are the material specific parameters known as electrical resistivity and thermalconductivity respectively.
For heat balance considerations the total joule heat (RI2) is equally divided between thecold & hot junctions. This division can be analytically justified.
Another equation of interest in this connection relates the totalinput electrical power (W) to
the cooling power (Q) obtained from the device. A coefficient of performance ( ) is defined as :
= Qc/W 1.4
Here W is made up of two parts, one appearing as joule heating (RI2) and the other used in
overcoming the Seeback voltage generated in the thermocouple as a consequence of thetemperature difference between the hot and cold ends. Thus
W= RI2 + AB(TH - TC}I 1.5
It is seen from equations 1.3, 1.4 & 1.5 that the cooling power (QC), temperature lift T (=
TH-Tc) and coefficient of performance ( ) are all functions of the input current (I) which has to
be optimized for a particular application. Using the Kelvin relation we write AB = Tc . AB.
We then have the following equations.
1. Maximum cooling power ( Using dQ / dI = 0 : eqn 1.3)
Qm = (Tc)2/2R-K(TH-TC) (1.6)
Iq= Tc/R (1.6a)
The corresponding coefficient of performance
q= [ZTc -(TH - TC)]/ ZTHTC (1.7)
2. Max coefficient of performance ( Using d /dI=0 : eqn 1.4 & 1.5)
8/2/2019 Basics Tec1
5/22
2/)(TTWhere,
9.1.)11(
)(=
1.811
/1
Hm C
m
CHAB
opt
m
CHm
CH
C
m
T
ZTR
TTI
ZT
TTZT
TT
T
+=
++
++
+
=
3 . Maximum temperature lift ( Using d(T)/dI = 0 : eqn 1.3}
()m= ( TH-TC)m = ZTc2/2 , where Z =
2/R 1.10
The optimum current to obtain ( T)m
Iopt= Tc/ R 1.11
.The voltage applied across the cooler for any current I is given by
V = IR + (TH -TC ) 1.12.The quantity 'Z' appearing in equations 1.4 - 1.11, contains material related parameters
( ,K ,R). This is called the figure of merit for the thermocouple as it determines the ultimateperformance of the TE cooler from the material point of view.
Since 'Z', as defined above, also depends on the dimensions of the thermoelements through
R & K (eqn 1.3a), it can be further optimized by minimizing the product K.R.
It can be shown that KR is minimum when
When the thermoelement dimensions are chosen such that this equation is satisfied , the
expression for the optimized figure of merit (Z) becomes
pn
np
np
pn
AlAl
=
8/2/2019 Basics Tec1
6/22
1.13)( 2
2
ABBA
AB
kkZ
+
=
Since the thermocouples are usually fabricated from materialshaving comparable thermoelectric properties except for the sign of the
Seeback coefficient (AB) , it is useful to define 'Z' for individual materials
independently as follows :
Z = 2/k 1.14
Here parameters , & k pertain to one material alone ( beingmeasured with reference to a common metal usually copper). In a rigorousanalysis of thermoelectric devices, however, the complete expression for Z
has to be used.The material parameter Z as defined in eq. 1.13 is known as the
thermoelectric figure of merit for the material and determines its potential forthermoelectric applications. All efforts in material development are directedat improving Z.
The materials used in thermoelectric coolers are Bi2Te3 based semiconductor alloys with Sb
partially replacing Bi to give p-type material and Se partially replacing Te to give n type materialrespectively
We see that the three parameters , ( or ) and which together decide the figure ofmerit of the material , are not independent and are also strongly temperature sensitive. At a given
temperature they are all related to the carrier concentration (n) of the material as can be seen in the
fig (1.2a). It can be seen that while and rise with the carrier concentration , the thermoelectricpower falls , and the figure of merit Z has a broad maximum around a concentration n 1 X 1019 /cm3 and therefore these alloys are heavily doped.
. The preferred alloy compositions forsingle stage coolers have been those , where the
maxima in Z fall near the room temperature. The adjustment of the carrier concentration in
these alloys is done using dopants ( such as I in n-type alloys) or by stoichiometric deviations( such as excess Te in p-type alloys).
8/2/2019 Basics Tec1
7/22
3. Effect of the Contact Resistance
Under ideal conditions, the size of the thermoelements can be reduced to any limit withoutaffecting the performance of the cooler module. This can be appreciated by examining eq. (1.3 )
& (1.5) reproduced below :
Qc = ITC- IR- K(TH - TC) 1.3
W =IR + (TH - TC) I 1.5
In these expressions 'R' & 'K' are the only dimension - dependent - parameters and these
remains unchanged by keeping l/A (ratio of the length to the area of cross section ofthermoelements) unchanged. For example by reducing l & A by half , R & K remain unchanged
although the volume or mass of the thermoelement becomes 1/4. For a given current I, therefore,
the cooling capacity 'Qc' & input power 'W' remain unchanged provided 'R' & 'K' are maintained
constant. The limitation on dimensions arises as a result of the contact resistance appearing at the
electrical contacts between the thermoelements and the metal links. For small sized
thermoelements the contact resistance becomes comparable to the resistance of the thermoelement.
The joule heating I2R caused by the contact resistance Rc acts as a built in heat load on the
cold stage and can become comparable to the heat load RI2/2 generated by joule heating in the
thermoelement and affecting the heat balance at the cold stage as considered in eq. (1.3). A good
Fig 1.2
8/2/2019 Basics Tec1
8/22
fabrication technology aims at keeping I2Rc
8/2/2019 Basics Tec1
9/22
Fig 1.3Fig 1.4
Now ,according to eq. 1.4, the maximum temperature lift [ ( T)m=ZTc / 2 ] depends upon Tc and also on the figure of merit 'Z' which falls
with decreasing temperature. Therefore, ( T)m for higher stages of a
cascade (operating at lower average temperature) decreases withtemperature due to decrease in both Tc and Z. So adding more and more
stages results in diminishing returns and the gain obtained by increasing thenumber of stages beyond 5 or 6 is not very significant with the thermoelectricmaterials available at present.
The average value of 'Z' for TE materials used in commercial
devices is 2.5 x 10-3K-1 For single stage coolers this gives ( T)m 67 K
(eq 1.4) for the heat sink at 300 K. This value corresponds to zero heat load
(Qc = 0) and therefore zero efficiency ( = Q/W = 0). The maximum heat
pumped by single stage coolers can be of the order of tens of watts using as
many as 127 couples connected thermally in parallel.
The performance parameters of thermoelectric coolers are normally
specified in terms ofTmax ( with Q=0) , Qmax ( with T =0) , Imax ( the current
for Tmax ) and Vmax ( the voltage applied across the cooler at Tmax . Therefore
N PP P P NNN
P
N
P
P
P
N
N
N
N
P
N
P
N
P
N
P
N
P
N
P
N
P
N
P
N
P
thermoelements
Alumina
8/2/2019 Basics Tec1
10/22
a load line of the type ( fig 1.5 ) can be drawn from which the maximum heat
pumped ( Q) at intermediate temperatures T < Tmax can be estimated.
Fig 1.5
For specific applications ,where the heat to be pumped (Q) along with
the temperature (T ) to be maintained is known , one can design a cooler for
optimum performance. The optimization can be in terms of either the
maximum heat pumped or the maximum efficiency ( c.o.p). Other design
constraints can be the voltage ( or current) supply available , the space
available for mounting the cooler , heat sink considerations etc,It has already been mentioned that for higher temperature lifts , a
cascaded structure is required, where the required temperature lift is suitablydivided over different stages. The structure has to be optimised in terms of thenumber of thermocouples per stage, current, dimensions of thethermoelements (thermocouple legs) and of course, the reliability and cost.5A Design of single stage coolers :
The theory of a basic thermoelectric cooler and the terms and equations related to itsdesign and operation have been discussed briefly.
The design of single stage TE cooler modules essentially means that for any application
where the heat load (Qc) , the hot side temperature (TH) , the cold side temperature (TC) and thefigure of merit Z of the material are known , the optimised structure of the TE Cooler has to be
determined in terms of the number of thermocouples (N) , their dimensions and the operational
characteristics of the module like Imax , Vmax (the current and voltage for maximum heat
pumping / cooling ) etc. Any design constraints like the current / voltage available , spaceavailable etc will also have to be taken into consideration .
There are normally two approaches to this , namely maximum heat pumping and
maximum efficiency (coefficient of performance - c.o.p). Maximum heat pumping requires
Q
T
Q max
Tmax
8/2/2019 Basics Tec1
11/22
that the module operate at the current Iq that pumps the maximum amount of heat over a
specified temperature differential with an efficiency q which is less than the maximum c.o.p
max given by equation (1.8 ). The current for max c.o.p , I ,is generally less than that for
maximum heat pumping for a given T. Again , for a given temperature difference T the heat
pumped at maximum c.o.p (Q ) is always less than Qmax .Therefore given a heat load Q and
T , the choice between designing for max and Qmax depends on whether economy in poweror economy in space is desired.
For specific applications , a cooler can also be designed to operate at any efficiency
which is between q and max . This can happen if there is a restriction on the poweravailable to to pump a certain amount of heat Q.
5B. Design of T.E. Cascades
The general principles involved in the design of TE cooling cascades canbe appreciated with reference to fig(1.6) which schematically shows a cascade
of two stages A1 & A2 operating with efficiencies 1 & 2 with power inputs
W1 & W2 respectively.
aa
FIG 1.6
The heat (Qc) extracted per second at the source by A2 is given by :
Qc= 2W2 1.15
The heat (Qm) rejected per second at the intermediate temperature Tmby A
2
and extracted by the lower stage A
1
is given by
Qm = Qc + W2 =Qc(1 + l/ 2) [from eqn 1.15]
Qm/Qc = 1 + 1/ 2 1.16
Qc
Qc (1+1/ 1)
Qc (1 +1/ 1) (1 +1/ 2)
A2
A1W1,
1
W2,
8/2/2019 Basics Tec1
12/22
Finally the heat (Qh) rejected per second at the sink at temperature This given by :
Qh
= Qm
+ W1
= Qm
(1 + 1/1) since
m= Q
m/W
1
substituting for Qm from eq. (1.15)
Qh = Qc(1 + 1/ 2)(1 + 1/ 1) 1.17
The total input power in the two stages is (W1 + W2) and it is used to
extract heat 'Qc' per second at the source. The overall efficiency A of the
two stage cascaded assembly is therefore :
A = Qc/(W1 + W2)
= Qc/(Qh - Qc) since Qh = Qc + W1 + W2
= [(1 + 1/ 2)(1 + 1/ 1)-1]-1 using eq. (1.17) 1.18
A simple extension of this analysis to an n-stage cascade gives the
following expression for A
A = [(1 + 1/ n)(1 + 1/ n-1)...(1+1/ 2)(1+1/ 1)-1]-11.19
If all the stages are assumed to be operating with the same efficiency
the expression for A condenses to
A = [(1+1/ )n-1]-1 1.20
Fig. (1.7) gives a typical representation of the variation of m with
temperature lift (T) for multistage cooler modules.
It can be seen from fig 1.7 that for T
8/2/2019 Basics Tec1
13/22
is twice as efficient as the two stage module. For obtaining T 100 K athree stage module is, therefore, essential.
For estimating the required number (N) of thermocouples in different
stages eq. (1.16) can be used. This expresses the ratio of heat capacities oftwo successive stages. It is easily appreciated that the cooling capacityrequired for the lower stage has to be more because the heat required to bepumped by it is the sum of the heat pumped by the upper stage and thepower fed into that stage. The TE cooling cascades therefore have apyramidal structure with successively lower stages having more and morenumber of couples.
To a first approximation, (ignoring the variation of 'Z' with temperature),the cooling capacity of a single stage module can be taken as proportional to
the number of couples (N) in that modules. Taking = 0.5, equation (1.16)
gives the ratio of cooling capacities or the ratio of number of couples in thetwo successive stages as 3 i.e. N increases by a factor of 3 for successivelower stages .
In a rigorous analysis , however the factors affecting the performance ofa cascaded cooler , like the temperature variation of the thermoelectricproperties , thermal resistance between various stages , contact resistancedue to each of the solders used etc have to be taken into account.
8/2/2019 Basics Tec1
14/22
Preparation of Thermoelectric material
The materials used in thermoelectric coolers are Bi2Te3 based semiconductor alloys with Sb
partially replacing Bi to give p-type material and Se partially replacing Te to give n type materialrespectively.The index of quality of the material is called the figure of merit Z ,which is defined as
Z= 2 / , where is the Seebeck coefficient , is the electrical resistivity and is the thermal
conductivity of the material . (Eqn. 1.14 , Appendix ). These materials have been in use for more
than three decades now and the highest values reported in literature for the figure of merit Z are 3 -
3.5 X 10 -3 / K. Commercial thermoelectric coolers however are made using material with a figure
of merit > 2.5 X 10 -3 / K.
We have from literature the approximate compositions of the p-type and the n-type alloys
which have been found to give the highest figure of merit as:
Bi 0.5Sb 1.5Te 3+ for p-type and Bi2Te 2.7 Se 0.3 for n type thermoelectric material.
It has however been seen that different workers have preferred minor variations of these.
Divergent techniques ranging from single crystal growth to powder metallurgy are used to produce
device grade material . At SSPL , preparation of thermoelectric material has been done using
Vertical Directional freezing technique. Facilities have been set up for the synthesis and growth
of Bismuth Telluride based alloys..Measurement apparatus for the evaluation of the thermoelectric
properties , and at room temperature have been set up. The figure of merit of the
prepared material at room temperature can thus be determined.
8/2/2019 Basics Tec1
15/22
Fabrication of a Thermoelectric Cooler
The schematic diagram of the structure of a basic thermocouple is
shown below.
The steps involved in the fabrication of a thermoelectric cooler can be
broadly classified as follows :
(i) Preparation of the patterned substrates.
(ii) Preparation of the thermoelements.
(iii) Development of jigs and fixtures .
(iv) Development of solders and soldering procedures.
(v) Assembly of cooler modules.
(vi) Cascading and testing.
Preparation of patterned substrates.
The alumina substrate (0.5- 1 mm thick) , which provides electrical
insulation as well as mechanical support to the thermoelements , holds the
copper pads ~25 m thick which provide the electrical interconnections. These
Alumina Substrate
PP NN
Cu pads
Thermoelements
Barrier layer (Ni)
Solder
8/2/2019 Basics Tec1
16/22
pads can be constructed on the alumina substrates using combination of
screen printing ,electroplating and photolithography. The resistance of the Cu
pads , their adhesion to the alumina substrate , as well as their capability of
withstanding repeated thermal cycling within the temperature range of
operation of the cooler are the important issues here. While plating thick
copper it is required that the flatness of the copper layer over the whole
substrate be held to within a tolerance of a few microns. For cascaded
coolers , the substrate has to be metallized and patterned on both the sides.
At SSPL , we have followed the process sequences given below.
(i) Cr / Au by evaporation followed by Cu plating and Photolithography
using a suitable mask.
(ii) Cr / Au by evaporation followed by Photolithography and Selective
plating of Cu on to the Alumina substrate upto the required thickness.
We have found that the second process sequence yields Cu pads
with better adherence. We have achieved upto 30 m thick well adherent Cu
pads on alumina substrates using this .
The suitability of the Cu pads arre tested by the ability of Cu pads
to withstand thermal cycling and prolonged operation.
Preparation of the thermoelements :
Making electrical contacts to the thermoelectric material is a very
important step in the fabrication process . A thin layer of Ni is normally used
to make electrical contacts to the thermoelectric material .The layer of Ni not
only improves the solderability of the material but also acts as a barrier for the
diffusion of Cu into the thermoelectric material.
Our aim is to deposit adherent ~3 m thick Ni layer on TE slices. The
procedure followed for preparation of the thermoelements is as follows:
1. After obtaining the ingot, it is sliced into sections of the required
thickness.
8/2/2019 Basics Tec1
17/22
2. The slices are lapped and polished to obtain the exact thickness as per
the design.
3. Electroless Ni deposition followed by electroplating of Ni on TE slices
upto~3 m is done.
4. The slices are then tinned with the appropriate solder using a suitable
technology.
5. Each slice is then sectioned into thermoelements having a cross-section
as per the design.
6. Surface preparation of the cut ingot has to be done carefully as any
mechanical damage is likely to increase the thermoelement resistance
and affect the performance of the cooler . Tight dimensional tolerances
have to be maintained during the sectioning process so as to ensure
that there is minimum departure from the A/ l ratio required as per the
design.
Development of jigs and fixtures
Suitable mechanical jigs have to be designed for the precise placement
of the thermoelements on the substrates when they are soldered. These jigscan have a cellular structure and may be designed so that they are removable
after the soldering process. As the size of the thermoelements as well as the
distance between them decreases to 1mm and less , the design of these jigs
becomes more and more complex .
At SSPL , we have designed and fabricated jigs for the assembly of
thermoelectric coolers to suit our integration processes .
Development of solders and soldering procedures
Our aim is to develop solders with a range of melting points and
procedures to perform the soldering. The soldering procedures evolved have
8/2/2019 Basics Tec1
18/22
to be such that electrically perfect and mechanically stable joints must be
made , as subsequent repair and rework of such joints is extremely difficult.
At SSPL , we have synthesised solders of melting points 139oC(Bi 58 , Sn
42) and 118oC (In 52 , Sn 48) . The melting points of these solders have been
measured using DTA. These solders are used for TE cooler fabrication.
The contact resistance at the ends of the thermoelements is a very
important technological parameter and has to be measured accurately for
each of these solders . Therefore a measurement set-up for the evaluation of
contact resistance has been developed and measurements performed on test
samples soldered using each of these solders.
8/2/2019 Basics Tec1
19/22
Assembly of cooler modules
Suitable fixtures have to be designed for aligning and holding the substrate and the jig
loaded with the thermoelments during the soldering process. The following procedure can be
followed for TE cooler assembly.
For single stage coolers , the thermoelements are first soldered on to the base alumina
Patterned substrate on a temperature controlled soldering station in Argon atmosphere. The jig
may then be removed and the top plate will be soldered after suitable alignment.
For multistage cooler assembly at least three to four different solders with graded melting
points need to be used . First the individual stages will be assembled using the solder of a higher
melting point (as in the case of single stage TE coolers). The different stages can then be soldered
on top of each other by heating them in a temperature controlled work station in Ar atmosphere.
Suitable fixture for holding the stages together while soldering will be designed and fabricated.
The assembled coolers are tested in a suitable setup. Tests for
performance degradation after thermal cycling are also done.
Heat sinks
The design of suitable heat sinks is a very important part of the development of any
system using thermoelectric coolers . The heat pumping capability of the thermoelectric module
is significantly influenced by the efficiency of the heat sink. The hot side of the module must
interface with an efficient heat removal system to achieve useful temperature differential across the
module.
Natural convection, forced convection and liquid cooled are three of the most common
types of heat sinks. Thermal resistance varies among the different types and sizes of sinks with
natural convection being the least efficient and liquid cooled the most efficient. The
majority of thermoelectric cooling applications use forced convection heat sinks with
thermal resistance values (RQ) ranging from 0.10/W to 0.5/W.
Using TE Coolers.
8/2/2019 Basics Tec1
20/22
Thermoelectric modules can be mounted in the desired system for an application by a
variety of methods including: mechanical clamping, epoxy bonding, and direct solder bonding.
The individual requirements of the application will determine which method is most appropriate,
however, mechanical clamping is by far the most common. Thermoelectric modules are relatively
strong in compression and weak in shear. Whichever method of installation is used, it is
important to avoid excessive mechanical loading of the module. Thermal resistance occurs at each
interface of an assembly and affects the overall system performance. In mechanically-clamped
systems, the recommended flatness of interface surfaces should be within 0.001. Even with this
degree of flatness, interface materials must be used to fill in the small thermal gaps; typical choices
include silicone-based thermal grease, graphite foil, and thermally-conductive pads. Special care
must be taken to ensure that uniform pressure is applied during installation.
8/2/2019 Basics Tec1
21/22
8/2/2019 Basics Tec1
22/22