Peltier Effect in Semiconductors: April 2014

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/262179901
Peltier Effect in Semiconductors
Chapter · April 2014

DOI: 10.1002/047134608X.W8206
CITATIONS READS
20 32,035
2 authors:
Yuri G. Gurevich Jesús Enrique Velázquez-Pérez

Center for Research and Advanced Studies of the National Polytechnic Institute Universidad de Salamanca
437 PUBLICATIONS 2,016 CITATIONS 135 PUBLICATIONS 781 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Terahertz detectors View project
Mono- and bipolar transport in semiconductor structures. View project
All content following this page was uploaded by Yuri G. Gurevich on 23 November 2017.
The user has requested enhancement of the downloaded file.

3GW8206 03/26/2014 0:6:44 Page 1
PELTIER EFFECT IN SEMICONDUCTORS are not acceptable solutions to allow the mobility of persons
with certain chronic diseases. Compact refrigerators with
eased low power supply requirements (low-power DC
instead of wall alternating current [AC] or lead batteries
INTRODUCTION used in cars) would be very useful. They would be used in
many different applications, such as in cryogenic skin
In recent years, devices based on the Peltier effect, which is surgery, high-availability electronic systems, and defense.
the basis for solid-state thermoelectric cooling, have In addition, those refrigerators should have a long life and
evolved rapidly to meet the fast-growing industry of elec- be maintenance-free.
tronics. The main point arises from the fact that the heat Thermoelectric cooling may advantageously replace the
extraction or absorption occurs at the contact between two current cooling methods based on chlorofluorocarbons
different conducting media when a direct current (DC) (CFCs) that are suspected as responsible for the break-
electric current flows through this contact. A comprehen- down of the ozone layer. Thermoelectric cooling has some
sive study of the mechanisms of heating and cooling origi- distinct benefits, as follows:
nated by an electrical current in semiconductor devices is
reported. The thermoelectric cooling in n-n, p-p, and p-n 1. Switch from heating to cooling by inverting the
junction contacts, as well as inhomogeneous bulk semi- polarity of the applied bias.
conductors, are analyzed. Both degenerate and nondegen-
2. Vibration and acoustic noise free.
erate electron and hole gases are considered. The role of
recombination and nonequilibrium charge carriers in the 3. Modulation of the cooling power through the supply
contact cooling (heating) effect is discussed. Along with the bias.
above, special attention is paid to several aspects of non- 4. Maintenance free.
equilibrium thermodynamics of thermoelectric phe- 5. Free of moving parts.
nomena involved in Peltier effect in semiconductors that 6. Functioning freely and independently of the
demand a careful examination. The formulation of an position.
adequate self-consistent theoretical model describing the
Peltier effect is also presented. Several interesting results have been reported on the
The Peltier effect was discovered in 1834 in a contact of progress in materials science with applications in thermo-
two metals, and it was described as the heat extraction or electrics: Skutterudites, clathrates compounds, half-Heus-
absorption that occurs at the contact metals when a DC ler alloys, compounds of layered bismuth telluride
electric current flows through it (1). As long as the structures, and low-dimensional structures (such as quan-
attempts to repeat the experiments to confirm this new tum dots and superlattices) have been investigated as a
effect failed to give positive results, it was not recognized means to increase the thermoelectric figure of merit.
by physicists until 1838 when E. H. Lentz managed to The traditional theory of the Peltier effect is presented
verify visually the Peltier effect (2). in several books (5–9, 14) and reviews (10, 11, 15, 16), but
The first theoretical descriptions of thermoelectric many important aspects of the Peltier effect are far from
effects were based on the thermodynamic ideas being covered satisfactorily in common books. In this
applied to the simplest models of thermoelectric circuits. article, we will present the current status of the theory
R. Klausius and W. Thomson initiated the development of of Peltier effect.
a theory almost simultaneously. The first result of these
investigations was the paper by W. Thomson (3). This
paper marked the beginning of the development of TRADITIONAL THEORY OF PELTIER EFFECT
the phenomenological theory of thermoelectricity. In
1931, L. Onsager used the thermodynamics of irreversible From the earliest papers on thermoelectrics right up to the
processes to describe thermoelectric effects, but his results latest publications, this effect has been defined as an
were not different from those obtained by W. Thomson (4). absorption of heat (in addition to the Joule heat) of the
In the middle of the twentieth century, A. F. Ioffe and his junction of two conductors or its cooling through which a
colleagues led the revival of interest in thermoelectricity (5). DC electric current flows (see, for example, References 5, 6,
One of the first materials they studied was PbTe, and they 14, and 16–18). The absorption of this heat or its evolution
quickly understood that the most prospective materials for depends on the direction of the electric current, and per
Peltier effect were semiconductors. unit time, it is equal to
The Peltier effect can be used to create a refrigerator
(see Figure 1) that is compact and has no circulating fluid QP ¼ ðP1 P2 ÞJ (1)
or moving parts; such refrigerators are useful in applica- where P1;2 are the Peltier coefficients of the conducting
tions where their advantages outweigh the disadvantage of materials at each side of the junction and J is the total
their very low efficiency (6–14). electric current.
Cooling devices play a crucial role in everyday life. We For a nondegenerate system of electrons (subscript n)
use them to preserve food and drugs, among many other and holes (subscript p) (14), the Peltier coefficient is
purposes. Often, persons cannot travel freely because they
use a medicine that must be kept in a refrigerator. It is easy h i
to understand that a 10-cubic-foot refrigerator or an icebox Pn;p ¼ 1=e ðgn;p þ 5=2ÞT mn;p (2)
J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 2014 John Wiley & Sons, Inc.
3GW8206 03/26/2014 0:6:45 Page 2
2 Peltier Effect in Semiconductors
Figure 1. Schematic of a thermoelectric couple.
where T is the temperature in energy units; mn,p are the while for the degenerate case (jmj T, m > 0)
chemical potentials of electrons and holes that are meas-

ured from the bottom of the conduction band and the top of p2 T 2 3
the valence band, respectively (mp ¼ eg mn, where eg is hei ¼ m þ gþ (8)
3m 2
the band gap); e is the hole charge; and gn,p are the
exponents in the momentum relaxation times (19): In some papers (see, for example, References 15 and 17), it
is emphasized that the Peltier effect manifests itself in the
e gn;p
thermally uniform systems, i.e., when the temperature
t n;p ðeÞ ¼ t 0n;p (3)
T gradient is absent.
Thus, the authors of Reference 17, when describing the
where e is the carrier energy. The constant quantities t 0n;p Peltier effect, wrote the following:
and gn;p for different relaxation mechanisms can be found
in Reference 19. If an electric current is driven in a bimetallic
For degenerate electron and hole gases circuit maintained at uniform temperature,
then heat is produced at one junction and
p2 3 T2 absorbed at the other (Figure 2). This is because
Pn;p ¼ gn;p þ (4)
3e 2 mn;p an isothermal electric current in a metal is accom-
panied by a thermal current q ¼ Pj, where j is the
To understand the physical meaning of equations 2 and 4, electric current density. Because the electric cur-
it is convenient to rewrite these expressions as follows: rent is constant in the closed circuit and the
Peltier coefficients differ from metal to metal,
1h i the thermal currents in the two metals are not
Pn;p ¼ hein;p mn;p (5) equal, and the difference must be produced at one
e
junction and supplied to the other to keep the
temperature uniform.
where hein;p is the mean kinetic energy of the carriers in the
flux,
Once the discussion of the physical principles of the
R1 df n;p
0 ðe Þ Peltier effect is concluded, we will briefly revise its
tn;p ðeÞe5=2 de
0 de
hein;p ¼1 (6)
R df n;p ðeÞ
tn;p ðeÞe3=2 0 de
de
0 jq = ∏ j
T0 T0
where f n;p
0 ðeÞ are the Fermi-Dirac distribution functions.
j
Notice that, similar to the case of the chemical potentials,
the mean energies hein;p are measured from the bottom of
the conduction band and from the top of the valence band, jq jq
respectively. Often, the chemical potentials mn;p are inter-
preted as the potential energies of electrons and holes (14). j
j
For nondegenerate statistics (jmj T, m < 0)
T0 T0
5
hei ¼ gþ T (7)
2 Figure 2. Schematic circuit of Peltier effect for constant T.
3GW8206 03/26/2014 0:6:45 Page 3
Peltier Effect in Semiconductors 3
applications. Actually there are many of them, for several The answer given by the cited works to the second
reasons. Thermoelectric devices use DC electric current, question is negative. But, it can be readily shown that
and changing its direction switches the device from the the Peltier effect in an adiabatic structure may take place
cooling mode to the heating one or vice versa. Precise beyond the traditional interpretation of the Peltier effect.
temperature control ð< 0:1 CÞ can be achieved with
Peltier coolers. Having such advantages, thermoelectric
coolers find a lot of applications in areas such as consumer PHYSICS OF PELTIER EFFECT IN n-n AND p-p STRUCTURES
products (recreational vehicle refrigerators, automobile
seat coolers, portable picnic coolers, motorcycle helmet The origin of the Peltier effect is not associated with the
refrigerators, and residential water coolers/purifiers), lab- external Peltier heat sources or the Peltier heat sinks. This
oratory and scientific equipment (coolers of infrared detec- effect can also exist in an adiabatically insulated thermo-
tors, laser diode coolers, integrated circuit coolers, heat electric structure. It is necessary to understand the con-
density measuring), industrial equipment (personal com- cepts of “cooling” or “heating” of the junctions as the
puter [PC] microprocessors, fine temperature control), decrease or increase in the junction’s temperature com-
medical instruments (portable and fixed pharmaceutical pared with the equilibrium one in absence of electric
refrigerators, blood analysis, insulin coolers), restaurants, current. Of course, this statement contradicts the hypoth-
and restaurant industry products (hotel room refrigera- esis isothermal conditions, from where a natural question
tors, noiseless air conditioners, ice makers). This list is not immediately arises: Why is a nonuniform temperature
complete and can go on and on. profile established across the structure in the absence of
At the same time, it is necessary to point out some external heat sources when an electric current flows per-
disadvantages of thermoelectric coolers. First, they have pendicularly to the two media interface?
a low efficiency compared with conventional refrigerators. Let us consider the simplest structure composed of two
Current thermoelectric devices operate at about 10% of uniform semiconductors through which the electric cur-
Carnot efficiency, whereas the efficiency of a compressor- rent flows in the direction 0x (Figure 3). Assume that the
based refrigerator is about 30%. Thus, the use of thermoelectric contacts x ¼ d1 and x ¼ d2 are kept at the equi-
electric coolers is restricted to applications where their librium temperature T 0 , that the lateral surfaces are
unique advantages outweigh their low efficiency. Although insulated adiabatically, and that the structure’s cross-
some large-scale applications have been considered, sectional area is equal to the unit measurement every-
Peltier coolers are generally used in applications where where. For simplicity, we shall assume initially that a heat
small size is needed and the cooling demands are not too resistance of the junction located at x ¼ 0 can be neglected
great, such as cooling electronic components (see, for the (this assumption will be removed later on); this assump-
sake of example, References 20 and 21). tion implies temperature continuity at the interface in
Summing up, one could say that the traditional x ¼ 0.
explanation of the Peltier effect presupposes the following
(22, 23): T 1 ðx ¼ 0Þ ¼ T 2 ðx ¼ 0Þ (9)
1. An isothermal state of the structure (bimetallic cir- where T 1;2 ðxÞ are the temperature distributions in each
cuit) exists, through which an electric current flows; region of the structure.
consequently, there is total absence of thermal fluxes Usually, the thermal flux Pj is named the Peltier heat
that would be proportional to a temperature gradi- flux (16). For the theory stated, it is more convenient to
ent. The only thermal fluxes are those accompanying name this heat flux a “drift heat flux” because it is associ-
the electric current. ated with the charge drift transport in the external electric
field, and accordingly, for holes the “drift heat flux” and
2. If a thermal interaction between this structure and
current have the same direction while they are opposite for
the external heat reservoirs is necessary to maintain
electrons. An important feature of the drift heat flux is that
the previously described isothermal state (a Peltier
heat source and a Peltier heat sink), then it becomes
necessary to take into account the external sources of
heat.
An important question arises regarding these consider-

ations: What do authors imply by the concepts of thermo-
electric “cooling” or “heating” in the traditional theories?
The presence of heat fluxes from the external reservoirs
into the structure and vice versa from the structure to the
external reservoirs does not describe heating or cooling
processes under an isothermal condition for the structure.
Another question is as follows: Can the Peltier effect
occur in the thermoelectric circuit without a thermal
interaction with the surroundings at the junctions; i.e.,
can the Peltier effect occur as a physical phenomenon in a Figure 3. Schema of the simplest Peltier device consisting of two
structure adiabatically insulated? different media.
3GW8206 03/26/2014 0:6:45 Page 4
it is associated with particle drift in a single direction; this j

is in strong contrast with the movement in two opposite
directions in the thermal diffusion flux that will be ana-
(1) dT1 dT2
lyzed below. qdif = – κ1 (2)
qdif = – κ2
dx dx
According to the Le Ch^ atelier-Braun principle in the
irreversible thermodynamics (24), “some internal fluxes
appear in the system in the stationary state when an
external influence affects this system, and these internal
fluxes weaken the results of this influence.” Applying this (1) (2)
principle to our problem, one can say that the discontinuity ddr = ∏1 j ddr = ∏2 j
in the drift fluxes at the junction x ¼ 0 that appears because
of the different Peltier coefficients (P1 6¼ P2 ) must lead to
other thermal fluxes tending to reduce this discontinuity
(22).
These thermal fluxes can only be thermal diffusion x=0
fluxes (qdif ¼ kdT=dx, k being the thermal conductivity)
because the other drift heat fluxes are absent. The Figure 4. Heat fluxes in an n-n structure assuming jP2 j < jP1 j,
“diffusion heat flux” nature greatly differs from the one k1;2 is thermal conductivity in each region.
of the drift heat flux. To understand the origin of the
diffusion heat flux, let us consider a material with a
nonhomogeneous temperature profile across it; under second side too, as compared with the temperature on the
these conditions, the equilibrium in the microscopic fluxes surface x ¼ d2 . As a result, an induced thermal diffusion
and balance will be removed. Inside the material, two flux q2dif ¼ k2 dT 2 =dx arises in the second sample, where
fluxes of quasiparticles will take place: one of particles k2 is the thermal conductivity of the right side. Let us
moving from the hotter area into the cooler one and a notice that this thermal flux direction coincides now with
second flux of particles in opposite direction. The total the direction of a drift heat flux q2dr ¼ P2 j in this sample
particle flux balance is null in the absence of electrical (Figure 4), tending to decrease the discontinuity in drift
current or when particles are not charged (phonons). But heat fluxes.
because the energy of carriers in the hotter area is larger The corresponding temperature distributions in the
than the ones in the cooler one, the total heat flux is structure are qualitatively represented in Figure 5. It is
nonzero and is directed from the hot side toward the obvious that for jP2 j < jP1 j and the chosen electric current
cold side. This flux is the diffusion heat flux. direction, the junction temperature T (x ¼ 0) decreases as
Thus, a temperature heterogeneity arises inevitably in compared with the equilibrium temperature T 0 . The low-
the structure in the presence of an electric current. ering of the temperature caused by the emergence of the
Because of this temperature distribution, the temperature induced thermal diffusion fluxes is the essence of the
of the junction can be lower (thermoelectric cooling) or Peltier thermoelectric cooling.
higher (thermoelectric heating) than the equilibrium one. If the condition jP2 j > jP1 j takes place and the electric
Because we are not interested in carrying calculations current has the same direction, then the drift heat flux
in real structures, let us keep considering the simple decreases when passing through the junction. In this case,
structure of Figure 3, assuming that it is composed of the induced thermal diffusion flux arises in the first sample
two n-type semiconductors and using the same subscript again, but its direction changes to the opposite compared
1;2
convention. Both drift heat fluxes qdr ¼ P1;2 j flow in this with the previous case (Figure 6). Now this flux tends to
structure in the direction opposite to the electric current
direction, as it is pointed in Figures 4 through 6, because
P1;2 < 0 in the n-type materials (14).
T(x)
Let us suppose that jP2 j < jP1 j for definiteness. In this
case, the drift thermal flux increases when coming from
side 2 into side 1 (Figure 4). According to the Le Ch^ atelier-
Braun principle, a thermal diffusion flux q1dif ¼ k1 dT 1 =dx
(k1 being the thermal conductivity of the left side) should
arise in the first side, tending to compensate this increase T0
of drift heat flux. It is clear that the direction of this flux –d1 0 d2 x
should be opposite to the direction of the drift flux
q1dr ¼ P1 j. We name these arising thermal diffusion fluxes T1(x) T2(x)
the “induced thermal diffusion fluxes.”
In turn, this means that the temperature of the junction
from the side of the first sample decreases in comparison
with the temperature on the surface x ¼ d1 . The condi-
tion of the temperature continuity (eq. 9) requires a reduc- Figure 5. Temperature distribution in an n-n structure assuming
tion in temperature in the junction from the side of the jP2 j < jP1 j.
3GW8206 03/26/2014 0:6:45 Page 5
(1) dT1 (2) dT2

qdif = – κ1 qdif = – κ2
dx dx
(1) (2)
ddr = ∏1 j ddr = ∏2 j
Figure 8. Heating or cooling of a metal-n-type semiconductor

contact.
x=0
What does the potential energy for the degenerate gas
Figure 6. Heat fluxes in an n-n structure assuming jP2 j > jP1 j. of the charge carriers mean?
What is the physical meaning of the energy change in
increase the drift thermal flux q1dr ¼ P1 j, which has a metal-p-semiconductor contact and in a p-n
decreased compared with q2dr ¼ P2 j. junction?
Further reasoning is similar to that used in the analysis
Why does the approach stated above ignore the work
of the previous situation, and it leads to the temperature
of the built-in electric field within the contact?
distributions in the structure presented in Figure 7. In this
It is easy to show that the value hei m in any type of
case the junction temperature increases in comparison
with the equilibrium temperature T 0 , and this situation contacts is different on the left and on the right from
corresponds to the Peltier thermoelectric heating. the contact plane. It means that single electron tran-
Thus, we have to understand the Peltier effect as the sitions cannot determine the work of the refrigerator.
lowering or raising of the junction temperature (depending Therefore, the description of the contact refrigeration
on the current direction) as a result of the appearance of based on the consideration of the separate electron
the induced thermal diffusion fluxes in the structure, but transitions is not valid. One needs to take into
not as an evolution or absorption of the additional heat on account the statistics of these transitions. How can
the junction. one reflect this fact in the preceding scheme?
To demonstrate an additional physical meaning (25) of
the Peltier effect, a metal-n-type semiconductor contact is In the usual scheme of n-n or p-p junctions, the contact
normally used (see Figure 8). It is easy to verify that the heating or cooling is explained by the electron (hole) tran-
surplus of the electron energy in the semiconductor in sitions between hei–levels (see Figure 9). The latter are
comparison with the metal is De ¼ hei mn ¼ ðg þ 5=2ÞT counted from the common Fermi level.
mn and P1 P2 ¼ De=e. Some additional assumptions will be introduced. Let us
The obtained result is clear for this type of contact. At consider first the n-n contact between nondegenerate semi-
the same time, these contacts are not really used in appli- conductors (Figure 10). If the momentum scattering in
cations. The consideration of other contacts does not have both semiconductors is the same, then (see eq. 2):
the same obvious understanding, and a priori the following
questions arise: m1 m2 x01 x1 x02 x2 Dec
P1 P2 ¼ ¼ ¼ wc þ (10)
e e e
T(x)
T1(x) T2(x)
–d1 T0 0 d2 x
Figure 7. Temperature distribution in an n-n structure assuming Figure 9. Energy band diagram of n-n or p-p contacts in equili-
jP2 j > jP1 j. brium (j ¼ 0).
3GW8206 03/26/2014 0:6:46 Page 6
and the work of the built-in electric field jointly with the
work of the valence forces occurs.
Now,

p2 3 T2
P1 P2 ¼ gþ ðm m2 Þ (12)
3e 2 m1 m2 1
Comparing equations 10 and 12, we see that the Peltier

heat in the contact between degenerate materials is essen-
tially less than that in the contact between the nondegen-
erate semiconductors.
Let us also note that the work of the built-in electric
field and the valence forces will give rise to the qualita-
tively different temperature of the contacts in the cases of
nondegenerate and degenerated branches of the circuit. In
Figure 10. Energy band diagram of the contact of two nonde- the former
generate semiconductors. e0 is the vacuum level, and xi and xi0 are
the work functions and electron affinities, respectively. 3
Wc / ðT ðEÞ T 0 Þ (13)
2
where xi and x0i (I ¼ 1, 2) are work functions and electron where T(E) is the nonequilibrium temperature, E is the
affinities of materials 1 and 2, wc ¼ ðx2 x1 Þ=e is the built-in electric field, and T0 is the room temperature. In
contact voltage, and Dec ¼ x01 x02 is the work of the the latter,
valence force. As is known, this is a result of the length
of the forbidden region that is position dependent. It means p2 T 2 ðEÞ T 20
Wc / (14)
that the contact cooling or heating is connected with the 4 m
work of the built-in electric field Wc and the work of the
valence forces Wn. The total work is W ¼ QP ¼ W c þ W n , The quantitative calculation of the thermoelectric cooling
where W c ¼ Jwc and W n ¼ JDec =e. Here, the valence force or thermoelectric heating requires a solution to the heat
is caused by the abrupt change of the conduction band. transfer equation with the corresponding boundary condi-
Similar reasoning is valid for the p-p junction. tions. The next section is devoted to this calculation.
If we have a contact between degenerate and nonde-
generate semiconductors, i.e., the nþ-n or pþ-p junction,
then the Peltier effect does not depend on the contact ENERGY AND HEAT BALANCE EQUATIONS FOR
properties, and it is determined only by the parameters UNIPOLAR SEMICONDUCTORS
of the nondegenerate semiconductor (see Figure 8).
Let us now remove the assumption about the absence of the
mn thermal resistance at the junction x ¼ 0. Instead, we shall
P1 P2 ¼ (11)
e describe this junction by the coefficient of the surface
In the case of a contact between two degenerate n- or p-type thermal conductivity (26–29), which can take arbitrary
semiconductors (or two metals), the Peltier effect depends values. Later, we shall show that the junction is described
on the contact properties again (see Figure 11 and eq. 4), not only by the surface thermal conductivity but also by
some other surface parameters.
The searched temperature distribution in the structure
can be obtained from the energy balance equation. In a
steady state, it is given by (16)
rw ¼ 0 (15)
where w ¼ q þ cj is the energy flux of charge carriers

(electrons or holes), q ¼ krT þ Pj is a total heat flux of
these carriers, c ¼ w m=e is the electrochemical poten-
tial, w is the electric potential, m is the chemical potential,
and e is the electron charge.
Now equation 15 can be rewritten in the form
rq þ jrc ¼ 0 (16)
We have made the assumption that nonequilibrium carri-

ers are absent in the structure (30), i.e., rj ¼ 0. Let us note
Figure 11. Energy band diagram of the contact between two
that it is necessary to take into account the nonequilibrium
metals. carriers in p n structures (17, 18).
3GW8206 03/26/2014 0:6:46 Page 7
Using the equation for the electric current (31) generation loses its physical sense in the bulk of the
structure (the analysis of this situation at the junction is
j ¼ sðrc þ arTÞ (17) presented next).
Let us now obtain the boundary conditions for equation
to eliminate the electrochemical potential c (s is electric
18 (22, 23). Because equation 18 is an equation of the
conductivity and a is the Seebeck coefficient), we arrive at second order in relation to unknown temperatures T 1 ðxÞ
the heat balance equation and T 2 ðxÞ for two-layer samples, it is necessary to obtain
rq ¼ j2 =s þ ajrT (18) four boundary conditions. Two of them correspond to the
assumptions that were made before
It is easy to see, from the shape of this equation, that there
are only two sources of heat: one of them is the Joule heat, T 1 ðx ¼ d1 Þ ¼ T 0 (19)
and the remaining source of heat can be named the Thom-
son heat because of its proportionality to the product of j
and rT (32). T 2 ðx ¼ d2 Þ ¼ T 0 (20)
Because the temperature gradient has appeared from The next boundary condition describes the total heat flux
the Peltier effect, we can say that the Thomson effect continuity at the boundary of the two media
occurs as a consequence of the Peltier effect in the ther- ðq1 ðx ¼ 0Þ ¼ q2 ðx ¼ 0ÞÞ in the most simple case of the
moelectric cooling (heating) phenomenon. absence of heat sources and sinks on the interface x ¼ 0.
Let us note that according to our definition following
from the heat balance equation (eq. 18), the Thomson k1 rTðx ¼ 0Þ P1 j ¼ k2 rT 2 ðx ¼ 0Þ P2 j (21)
coefficient is equal to the Seebeck coefficient a. From
this definition, it follows that the Thomson heat occurs Last, the fourth boundary condition determines the heat
even in the case when the Seebeck coefficient does not flux through the interface x ¼ 0 with the surface thermal
depend on temperature. Moreover, it is possible to claim conductivity h (33).
that thermoelectrics contains only one fundamental
parameter, which is the Seebeck coefficient. h½T 1 ðx ¼ 0Þ T 2 ðx ¼ 0Þ ¼ k1 rTðx ¼ 0Þ þ P1 j (22)
Thus, only one thermoelectric relationship was named The right-hand side of equation 22 determines the total
the second thermoelectric relationship, P ¼ aT (15, 16). As thermal flux flowing into the interface of the two media.
it was shown in Reference 16, it follows from the Onsager’s The expression on the left-hand side of the equation deter-
principle of kinetic coefficients symmetry. The first ther- mines the same thermal flux in the surface layer x ¼ 0 (for
moelectric relationship actually determines the Thomson simplification, we neglect the surface Peltier coefficient
coefficient t. As it follows from equation 18, it can be (33) in the given work, supposing it negligibly small in
reduced to the trivial equality t ¼ a. comparison with the bulk Peltier coefficient).
The Thomson coefficient was determined by the
equation t ¼ @P=@T a in the previous publications
(5–9, 14, 18, 31, 41). The authors of these works formed TEMPERATURE DISTRIBUTION IN n-n AND p-p
the term by combining j@P=@TrT from divq with the term STRUCTURES
ajrT presented on the right-hand side of equation 18. This
is incorrect because ajrT is the source of heat while To obtain the required temperature distributions we will
j@P=@TrT is just a part of the thermal flux change. use the heat flux continuity equation (eq. 18) in the linear
This is trivially wrong from the formal point of view. We approximation of j, which for a homogeneous medium in
cannot state that the left-hand side of equation 18 one dimensional case reduces to
is still the complete divergence of the thermal flux vector
q ¼ krT þ Pj when the term j@P=@TrT is moved to the d2 T=dx2 ¼ 0 (23)
right-hand side of equation 18 or that it is the divergence of
another vector in this case. Together with the boundary conditions (19–22), equation
Now, the sense of the Thomson heat is very transparent. 23 results in the following temperature distributions:
It is the charge carriers heating or cooling in the thermo- (
electric field Ete ¼ arT. In a certain sense, it can be named hd2;1
T 1;2 ¼ T 0 1 P1;2 ðP1 P2 Þ
the Joule effect associated with the thermoelectric field. k2;1
The Joule heating takes place in the electric field. It is )
important to note that the Joule effect leads to heating in j d1;2 x d1 x 0
h i ; : ð24Þ
any case, whereas the Thomson effect may lead to either T 0 k1;2 1 þ h dk11 þ dk22 0 x d2
heating or cooling depending on the relative directions of
the electric current and the temperature gradient. Subse- The second term in the square brackets of equation 24
quent development of these ideas can be found in Refer- determines the contribution of the Peltier effect to the
ence 32. temperature distribution in the structure with a finite h.
Equation 18 demonstrates that the Peltier source of This term defines the Peltier thermoelectric heating or
heat is absent everywhere in the structure volume, as it cooling considered in the introduction. As it has been
was pointed out previously in this article. For this reason, shown in Reference 25, the value ðP1 P2 Þj determines
the statement about the Peltier’s heat absorption or the change in the kinetic energy flux on the junction. This
3GW8206 03/26/2014 0:6:46 Page 8
change is completely dependent on the electric potential conductivity is characterized by the Peltier coefficient
barrier parameters on the junction, such as the work Pn < 0, whereas the material with the p-type conductivity
functions of the contacting semiconductors and their elec- is characterized by the Peltier coefficient Pp > 0. In the
tron affinities. last case, both areas close to the junction will be heated or
Let us note, as it has not been previously mentioned, cooled (depending on the direction of the electric current)
that the Peltier effect depends strongly on the junction simultaneously and the effect of thermoelectric cooling
surface thermal conductivity. The contribution of this (heating) is more intense.
effect to the total effect of thermoelectric cooling increases In fact, let us assume for simplicity that the structure
with an increase in surface thermal conductivity and it represented in Figure 3 is composed of p- and n- materials
slackens with a decrease in surface thermal conductivity. identical in length and with equal values of the coefficients
The term in the square brackets of the numerator of of thermal conductivity d1 ¼ d2 ¼ d; k1 ¼ k2 ¼ k; that the
equation 24 determines a new cooling or heating effect that recombination rate on p-n junction is infinitely high; and
cannot be reduced to the Peltier effect. First, it is deter- that the Peltier coefficients of both materials are equal in
mined not by the difference between the Peltier coefficients absolute magnitude (P1 ¼ P2 ).
but by each Peltier coefficient separately. This implies that In this case, the Peltier effects are
this effect can be observed in a structure composed of a
single material and radically distinguishes it from the
jPjd
Peltier effect. Besides, this effect is not associated with Tðx ¼ 0Þ ¼ T 0 1 j (25)
T0 k
the electric potential barrier on the interface x ¼ 0, being
determined exclusively by the junction surface thermal
properties. where jPj ¼ jP1 j ¼ jP2 j. The last equation shows that the
The heat balance equation (eq. 18) is in the general case Peltier effects are commensurable in the structure
a nonlinear equation of the density of the electric current j. considered.
This follows from the nonlinear terms describing the Joule
and the Thomson heat. In the case of the Thomson heat, it
is necessary to take into account that the temperature COOLING IN INHOMOGENEOUS SEMICONDUCTORS
gradient in the problem considered is not created by any (BULK PELTIER EFFECT)
external sources. It arises only as a result of an electric
current flow. Thus, both the Joule and the Thomson heat Under certain conditions charge carriers in inhomogeneous
sources are nonlinear terms in j. As it was shown in semiconductors can be cooled by a static electric field instead
Reference 32, the magnitudes of both heat sources are of of being heated, and consequently the carrier average
the same order. The Thomson heat even exceeds the Joule energy of drops below the lattice temperature T 0 (37).
heat in good thermoelectric materials, i.e., those materials This effect can be explained as follows. Let assume that
that exhibit a figure-of-merit ZT, where Z is Z ¼ a2 s=k (5), the impurity concentration N in a semiconductor depends on
exceeding the unity (more details about ZT will be given in a coordinate (x, for example). If the Debye radius is the
the section titled, “Applications”). Besides, both the smallest parameter having the dimension of length in the
coefficient of thermal conductivity and the Seebeck problem, the carrier concentration at each point will coin-
coefficient depend on temperature, which is also a function cide with the impurity concentration (26, 27), thus establish-
of the electric current. ing an electron concentration gradient (dN/dx).
At the same time, it is easy to observe that the non- It is well known that an associated electrostatic internal
equilibrium temperature distributions occur even in the field is then directed along the gradient. An external
linear approximation with respect to the electric current. electric field applied to the sample will induce a current
So, thermoelectric cooling or heating are linear effects in in the same direction as this field (31). If the current
an electric current and manifested themselves “in the pure direction is opposite to the direction of the concentration
state” only at small values of current when the Joule and gradient and the external field is weaker than the field
the Thomson heat do not play a noticeable role. For this associated to the gradient dN/dx, then the carriers will
reason, it is convenient to consider the problem of thermo- move in a (total) field that is oriented against their motion.
electric cooling just in the linear approximation to the Then, the electric field will obviously remove energy from
electric current j. the carriers, which are thus “cooled.” The energy removed
Let us notice that all the equations obtained above from the carriers must be liberated outside the sample, to
remain true for the structures composed of semiconductors the current source, for example.
with the p-type conductivity. It is only necessary to con- The described effect is somewhat analogous to the
sider that P1;2 > 0 in this case. Peltier effect, although the former is of spatial character.
The Peltier effect occurs at the junction of semiconduc- Let us as assume an isotropic quadratic electron dis-
tors with electron and hole conductivities too. However, persion law and that collisions between carriers and scat-
the thermoelectric processes in this case depend essen- tering centers are quasielastic (with the momentum-
tially on the recombination rates in areas close to the transfer average free carrier path lp le (38)).
junction (see References 25, 30, 34–36). In the case of an The concentration of impurities, and therefore that of
infinitely strong recombination, all the equations above are the carriers, depends on x and it is given, for simplicity, by
correct. The essential difference between this case and the N ¼ N 0 expðx=lÞ. The magnitude of N0 is such that the
previous ones is that now the material with the n-type frequency nee of interelectronic collisions greatly exceeds
3GW8206 03/26/2014 0:6:46 Page 9
the frequency ne of collisions between electrons and scat- approximation are usually not discussed. Nevertheless,
tering centers accompanied by energy transfer. there are many phenomena in which the temperatures
Under the previous assumptions, our problem may be of current carriers and phonons are different from each
considered one dimensional (i.e., all quantities will depend other (as it was shown in this section).
only on the x coordinate). As it is well known (22), the Peltier effect, which under-
The electron temperature as a function of the coordi- lies thermoelectric cooling, is a contact phenomenon. It
nates can be determined by the equation of heat balance, occurs in a subsystem of electrical-charged quasiparticles
which can be obtained from the following kinetic equation (electrons and holes). That is, the electron and/or the hole
(see References 26 and 27, for example): cooling takes place near an interface. Then, the lattice is
also cooled by means of energy interactions between elec-
dq=dx ¼ jE NðxÞne ðT e T 0 Þ (26) trons (holes) with phonons near the interface. As a result of
this two-step process, different temperatures are set in the
where E ¼ dw=dx, subsystems of current carriers and phonons.
Thus, it is obvious that the multitemperature approxi-
q ¼ N=N 0 ke T e 1þg ½eE T e =NdN=dx ð2 þ gÞdT e =dx mation, when each subsystem of quasiparticles is charac-
(27) terized by its own temperature, is more appropriate for the
investigation of the thermoelectric cooling phenomenon
is the heat flux in electron subsystem and E is the electric
(39, 40).
field strength in the sample (E has only an x component).
j ¼ N=N 0 sT e g ½E T e =e 1=NdN=dx ð1 þ gÞ 1=edT e =dx

PELTIER EFFECT IN p-n JUNCTIONS
(28)
is the current and T e is the electron temperature. The Conventionally, a p-n structure is used to make a thermo-
parameters electric refrigerator (5, 6, 9, 41) because of the thermo-
electric drift fluxes directed (at a corresponding direction of
4Gð7=2 þ gÞN 0 4Gð5=2 þ gÞN 0 e2 a current from the n-region to p-region) from the interface
ke ¼ pffiffiffiffi ; s¼ pffiffiffiffi toward the edge in both layers of the p-n structure that
3 pmn0 T g0 3 pmn0 T g0
strengthens the cooling phenomenon (22). Traditionally,
are, respectively, the electronic heat and the electric con- the studies of the Peltier effect do not consider the non-
ductivity at T e ¼ T 0 . n ¼ n0 ðT 0 Þðe=T 0 Þg is the frequency of equilibrium charge carriers, so that only majority charge
collisions with momentum transfer (26, 27) between carriers and their electric current are taken into account in
electrons and scattering centers. ne ¼ n0e ðT 0 ÞðT e =T 0 Þr1 the expressions for heat fluxes in n- and p-regions, even
(26, 27) is the frequency of collisions with energy transfer though the current of minority charge carriers near the p-n
(26, 27) between electrons and scattering centers. r and g junction has the same order of magnitude than the current
are numerical coefficients with magnitudes close to one (for of majority charge carriers (25, 42) (see Figure 12).
which explicit expressions are given in Reference 19), In this way, the thermal generation and extraction of
whose values depend on the scattering mechanism with minority charge carriers must take place near the interface
respect to the carrier energy and momentum, respectively. to allow the flow of electric current (42). As a consequence,
Because under stationary conditions the current nonequilibrium charge carriers will settle (25, 35, 36).
density j does not depend on the coordinate, it is Let us examine the p-n junction (Figure 13). It is easy to
convenient to express the field E in terms of j. Using see From Figure 13 that
equation 28, we obtain E and substituting it and q into
equation 27, we can finally obtain the expression for the mn þ mp
P1 P2 ¼ (29)
electron temperature. e
This equation must be supplemented with the boundary
conditions for temperature in the planes x ¼ 0 and x ¼ L,
where L is length of sample (see eq. 21).
In the current range 0 < j < jc, where jc ¼ 2le =lð5=2 þ
expðL=lÞ1 ek
gÞ expð2L=lÞ1 le T 0 is the average temperature hT e i < T 0
RL
where hT e i ¼ 1=L T e ðxÞdx, if le < l (37). As the ratio
0
le =l increases, the minimum of hT e i as a function of j is
shifted to higher values of j; hT e imin can be of the order of
unity when le/l 1 (37).
Note that the one-temperature approximation consists
of assuming that the temperatures of all quasiparticles
(electrons, holes, and phonons) in each spatial point
coincide. Until now, this approximation has been widely
used for research of the thermoelectric cooling phenome-
non. However, the criteria for applicability of such an Figure 12. Distribution of currents in p-n junctions.
3GW8206 03/26/2014 0:6:46 Page 10
Figure 13. Energy band diagram of a p-n junction.

Recombination.
Figure 14. Energy band diagram a p-n junction without

The generation or recombination at the contact causes
recombination.
the Peltier heating or cooling in this case.
At the same time, there are a lot of situations where this
rate is weak enough. What has happened with the Peltier by Ponomarenko and Stafeev (44). The investigation of
effect in this case? The situation is similar to the electric the thermoelectric effect was undertaken in the following
current or thermoelectric current in bipolar semiconduc- years in a number of studies (45–53). Unfortunately in
tors. The Fermi quasilevels will set up causing the all these articles, several assertions were incorrect and
redistribution of carriers concentrations and, accordingly, led to erroneous results. Some of the mistakes were
modifying the electric field profile. As a result, the nature already pointed out and discussed in Ref. 54; this short
of the Peltier effect can drastically change. paper is free from fundamental errors, but it presents
It follows from Figure 14 that under the chosen current neither new equations nor a quantitative analysis of the
direction cooling of the junction switches to heating in thermoelectric phenomena. Common mistakes in previ-
absence of recombination. ous studies, additional to those discussed in Ref. 54, can
The equivalent electric circuits are presented in be found in the modeling of the recombination used, in
Figures 15 and 16 for the two limiting cases, very strong the energy balance equation, in the boundary conditions,
and weak recombination, respectively. and so on (35, 36). The analysis of the thermoelectric
The thermoelectric behavior of p-n junction for Peltier cooling that is given in References 43–54 is only approxi-
effect was investigated for the first time in Reference 43. mate because it ignores several factors as we will discuss
Experimentally, this phenomenon was discovered in 1974 next.
Figure 15. Strong recombination: Carrier movement (left) and schematic of the equivalent circuit (right).
Figure 16. Weak recombination: Carrier movement (left) and schematic of the equivalent circuit (right).
3GW8206 03/26/2014 0:6:47 Page 11
HEAT BALANCE EQUATION IN BIPOLAR The expression for the energy flux w in a bipolar semi-
SEMICONDUCTORS conductor can be written in a reduced form as follows:
The energy balance equation has never been sufficiently w ¼ q þ cn jn þ cp jp (33)

investigated in bipolar semiconductors; nevertheless, even
if in the linear approximation with respect to the gradient where
q ¼ qdr þ qdiff (34)
of the temperature (i.e., assuming a small mismatch of the
temperatures in both regions) a significant population of If the transport energy processes take place in a bipolar
nonequilibrium charge carriers will always set up and, semiconductor, then the rate of change of the energy
consequently, generation-recombination processes must density u in the semiconductor is as follows:
be taken also into account (55).
Let us note that in this case (as it will be shown next), @u
¼ divw (35)
the recombination has a dual role. On the one hand, it is @t
the source or the absorbent (sink) of heat, and on the other The right-hand side of equation 35 corresponds to an
hand, it bears a significant influence on the drift heat energy change caused by energy transport. If the processes
fluxes. Consequently, the expressions for the sources of that take place in the semiconductor are stationary, then
the total heat flux and the diffusion heat flux are distinctly the energy of each region does not change with time and
different. equation 35 reduces to the following:
The energy flux density in a unipolar semiconductor is
(see eqs. 15 and 16): divw ¼ 0 (36)
w ¼ q þ cj (30) Now a question arises: “What value plays the role of the
Let us recall that, under the approximation of quasineu- source of the heat flux?” To answer to this question, the
trality, nonequilibrium charge carriers in a homogeneous expression for divq needs to be found. By substitution of
unipolar semiconductor will not exist in the linear approxi- the energy flux, w, from equation 33 into equation 36, we
mation with respect to the perturbation (55). Consequently, obtain the following expression (let us recover the vector
the variation of the electrochemical potential dc reduces to nature of several magnitudes to avoid misunderstand-
the variation of the electrical potential dw (56) because ings):
cn;p ¼ w ð1=eÞmn;p , where w is the electrical potential, e
is the elementary positive charge, and mn ðmp Þ is the chemical divq ¼ rðcn Þ jn cn ðr jn Þ r cp jp c p r jp
potential of electrons (holes). This explains why, as it follows (37)
directly from equation 30, the variation of the energy flux dw
Taking into account the carrier continuity equations (55),
is equal to the variation of the heat flux dq, dw ¼ dq under
in which the rates of the total recombination are the same
steady-state conditions (divj ¼ 0).
and under total absence of external generation (see next
The heat flux density q is equal to the sum of the density
section)
of the drift heat flux qdr ¼ Pj and the density of the
diffusion one qdiff ¼ krT. Let us remark that formally rjn ¼ eR; rjp ¼ eR (38)
heat flux is a vector (q), in here we will drop the vector
where R is the recombination rate. Using equation 38,
notation as, for the sake of simplicity, we are considering a
equation 37 transforms into the following:
one-dimensional case.
In a bipolar semiconductor (57, 58), the drift heat flux is
divided in the drift heat fluxes of electrons and holes, divq ¼ mn þ mp R rðcn Þ jn r cp jp (39)
qdr;n ¼ Pn jn and qdr;p ¼ Pp jp , whereas the diffusion heat
flux is equal to the sum of the diffusion heat fluxes of Using the expressions for the electrical current (34)
electrons (holes) qdiff ;n ¼ kn rT (qdiff ;p ¼ kp rT) and the
diffusion heat flux of phonons, qdiff ;ph ¼ kph rT. jn;p ¼ sn;p r cn;p þ an;p rT (40)
where sn ðsp Þ is the electrical conductivity of electrons
qdiff ¼ qdiff ;n þ qdiff ;ph ¼ kn þ kph rT
(31) (holes), and Pn;p ¼ Tan;p . The gradients of the electrochem-
qdiff ¼ qdiff ;p þ qdiff ;ph ¼ kp þ kph rT ical potentials can be written as follows:
From equation 31, it follows that the energy flux in a
1
bipolar semiconductor is r cn;p ¼ j an;p rT (41)
sn;p n;p
w ¼ qdr;n þ qdiff ;n þ cn jn þ qdr;p þ qdiff ;p þ cp jp þ qdiff ;ph Substituting the expression in equation 41 into equation
(32) 39, the divergence of the heat flux adopts the following
In this expression of the energy flux in a bipolar semi- form:
conductor is the energy flux of electrons, holes, and pho- 1 1
nons. Let us note that cn and cp are independent Fermi divq ¼ mn þ mp R þ j2n þ j2p þ an jn rT þ ap jp rT
quasilevels of electrons and holes (in equilibrium sn sp
cn cp ¼ eg =e, where eg is the energy gap) (14). (42)
3GW8206 03/26/2014 0:6:48 Page 12
From equation 42, we can see that recombination creates Because the current densities can be calculated as
the source of the heat flux ðmn þ mp ÞR. Because mn þ mp is (see eq. 40) follows:
negative in nondegenerate semiconductors, if the
recombination rate is positive, then this source of heat jn ¼ sn ðrw rmn =e þ an rT Þ
flux is negative ðmn þ mp ÞR < 0, while in the case of ther- (49)
mal generation (R < 0), it is positive ðmn þ mp ÞR > 0. These jp ¼ sp rw þ rmp =e þ ap rT
paradoxical phenomena will be explained subsequently.
Because under a linear approximation mn þ mp ¼ eg , the the calculation of mn ; mp , and w is required.
source of heat flux caused by recombination transforms The macroscopic description of the transport of non-
into eg R and we obtain the following: equilibrium charge carriers is performed using the conti-
nuity equations for the electron and hole current densities
divq ¼ eg R (43) (see eq. 38) and the Poisson equation (30)
To find the sources of the diffusion heat flux, the diver- r jn ¼ eR (50)
gence of the diffusion heat flux divqdiff must be obtained.
Substituting the heat flux q from equation 34 into equation
42 and taking into account equation 38, we obtain in the r jp ¼ eR (51)
linear approximation the following:
r
divqdiff ¼ ðgn þ gp þ 5ÞT 0 R (44) Dw ¼ 4p (52)
e
Taking into account, that (58) where r is the space charge, e is the permittivity, and R is
the recombination rate in n- and p-regions.
Pn;p ¼ ð1/eÞ gn;p þ 5/2 T mn;p
(45)
t n;p ðeÞ ¼ t 0n;p ðe/T Þgn;p RECOMBINATION IN SEMICONDUCTORS
It is worth noting that the source of the diffusion heat flux
caused by the recombination is positive, ðgn þ gp þ To make use of the system of equations 50–52, we should
5ÞT0 R > 0 because gn þ gp < 5 (19). This means that specify the dependences of partial currents, the space-
recombination causes the increment of the temperature charge density, and the recombination rates on the non-
(heating) in agreement with common sense. equilibrium-carrier concentration.
It should be borne in mind that the absolute value of the Unfortunately, many publications devoted to these
Peltier coefficient, equation 45, of minority carriers can problems contain obvious errors; these fallacies are caused
largely exceed the one of majority carriers. Because the by the recombination itself that is often described by the
Fermi quasi-levels depend on the concentration of majority incorrect expressions as follows:
and minority charge carries, the Peltier coefficients will
depend on the coordinate in the space-charge layer near divjn ¼ Rn ; divjp ¼ Rp (53)
the p-n junction (rnD < x < rpD , where rn;p
D is the Debye with Rn ¼ dn=tn ; Rp ¼ dp=tp .
radius in the n- and p-regions as indicated by the super-
Here n ¼ n0 þ dn and p ¼ p0 þ dp are the electron and
script) even in linear approximation with respect to the
hole concentrations, where dn and dp are the concentra-
current because of the spatial dependence of the equili-
tions of nonequilibrium charge carriers, and tn and tp are
brium concentrations near the p-n junction (59).
the lifetimes of nonequilibrium charge carriers, which are
The expression for the diffusion heat fluxes qn;p
diff is as key parameters of the semiconductor under consideration
follows:
(60, 61). However, because the condition for the total
current continuity div j ¼ 0 should be satisfied under static
qn;p n;p n;p n;p
diff ¼ kn þ kp þ kph rT (46) conditions, an additional condition dn=tn ¼ dp=tp arises;
n;p
the latter condition does not follow from any physical
n ; kp , and kph are, respectively, the electron,
where kn;p n;p
concepts and is difficult to interpret. In certain cases
hole, and phonon heat conductivities in the n- and the (see, for example, monograph in Reference 62), this condi-
p-regions as indicated by the superscript. tion is used to reduce the number of variables, which is
Because kn;p kph in nondegenerate semiconductors, completely wrong. Sometimes, this condition is also con-
equation 46 reduces to sidered as an expression relating to the lifetimes of charge
carriers (60). The latter approach is not physically mean-
qn;p n;p
diff ¼ kph rT (47) ingful and, in addition, is hardly constructive because the
carrier lifetimes no longer constitute the semiconductor
Taking into account the considerations presented previ- parameters; rather, the lifetimes are functions of the non-
ously, the heat balance in equation 43 can be rewritten as equilibrium carrier concentrations that, in turn, should be
follows: determined, which leads to considerable mathematical
difficulties.
kph DT þ Pn r jn þ Pp r jp þ jn rPn þ jp rPp ¼ eg R
Another approach based on the assumption that Rn ¼
(48) Rp ¼ dp/t p, where dp and t p are the concentration and
3GW8206 03/26/2014 0:6:48 Page 13
lifetime, respectively, of the nonequilibrium minority car- where Nt is the impurity concentration; xn and xp are,
riers, is widely accepted (63). In this case, the condition for respectively, the electron and hole capture coefficients
the total-current continuity is identically satisfied; how- n1 ¼ nn ðTÞexp½et =T; p1 ¼ np ðTÞexp½ðet eg Þ=T, where
ever, another basic contradiction takes place. This contra- et is the impurity energy level; nn ðTÞ ¼ 1=4ð2mn T=
diction becomes especially evident if we consider a majority ph2 Þ3=2; np ðTÞ ¼ 1=4ð2mp T=ph2 Þ3=2 ; and mn and mp are
carrier injection. Physically, it is obvious that the injected the electron and hole effective masses. The concentrations
nonequilibrium majority carriers recombine. At the same n1 and p1 are the parameters characterizing the impurity
time, formally, the recombination rate is equal to zero level and, physically, represent the electron or hole con-
because there are no nonequilibrium charge carriers centrations when the Fermi level in the semiconductor
(dp ¼ 0). coincide with the impurity level. nn and np are the densities
The problem is appreciably complicated if the tempera- of state at the bottom of the conduction band and top of the
ture distribution in the sample is nonuniform (i.e., the valence band.
temperature is coordinate dependent). As a result, it Unless otherwise indicated, and with no loss of gener-
becomes unclear from which value the nonequilibrium- ality, we shall refer to a semiconductor that contains
carrier concentration is measured (i.e., what is meant by impurities with a single energy level able to capture elec-
the quantities n0 and p0). A method to overcome this trons. It follows from equation 56 that one more unknown
difficulty has been suggested recently (34); however, this value nt (nt is the concentration of electrons captured by
method is rather artificial and is applicable only in the case the impurities) arises when the recombination takes place
of a known (fixed) temperature field. through the impurity centers.
It is relatively straightforward to obtain the following Subtracting the second equation 53 from the first, we
expressions for recombination processes (64) by rigorous obtain (30, 55) the following:
consideration of transitions between valence and conduc-
tion bands: divðjn þ jp Þ ¼ eðRn Rp Þ (57)
Rn ¼ Rp ¼ R ¼ xðnp n2i Þ (54)

The charge conservation in steady state can be written as
where x is the recombination factor and ni is the carrier follows (31):
concentration of the intrinsic semiconductor. By lineariza-
tion of these expressions for the case of weak deviation divj ¼ divðjn þ jp Þ ¼ 0 (58)
from the condition of thermodynamic equilibrium (the
changes in carrier concentrations as a result of current From equations 57 and 58, we obtain the following rela-
are small in comparison with their equilibrium values), we tionship (30):
have (30) (TðxÞ ¼ T0 þ dTðxÞ and dTðxÞ T 0 ):
Rn ¼ Rp (59)

p0 n0
R ¼ 1=t dn þ dp ldT (55) As it follows directly from equations 59 and 56, the
n0 þ p0 n0 þ p0
deviation of the concentration of the electrons trapped
(65, 66) t ¼ ½xðn0 þ p0 Þ1 , l ¼ n02n in the impurity level dnt from the equilibrium value n0t
where þp0 dni =dT ¼
i
n0 p0 1
ð3 þ eg =T 0 Þ. depends on the deviations of the electron and hole
n0 þp0 T 0
concentrations from their equilibrium values through
Let us note that this baseless idea is widespread; Rn and Rp.
namely, the presence of only interband recombination is The recombination rates Rn and Rp are actually defined
a sufficient condition for dn ¼ dp to be fulfilled (42, 61, 62). as the difference between the rates of capture of electrons
However, there is no proof of this conclusion; moreover, the and holes and their thermal generation. These two mech-
case of injection obviously contradicts it. anisms cannot create nonequilibrium carriers without an
Thus, the presence of a temperature gradient results in external source of excitation. From equations 59 and 56, we
the appearance of an additional term in expressions for obtain the following expression for dnt :
recombination rates. This term takes into account the
change in the rate of thermal generation (which, as it is n2t0 p
well known, is proportional to the squared concentration of dnt ¼ xn 0 dn xp dp
N t ðxn no þ xp p10 Þ p10
the intrinsic semiconductor at a given temperature).

The situation becomes even more complicated if eg et
recombination resulting through the impurity centers xn n10 ðet =T 0 þ 3=2Þ xp p0 þ 3=2 dT=T 0
T0
(traps) is taken into account. Within the framework of (60)
the Shockley–Read–Hall model and with the assumption
that the carriers of impurity centers are characterized by
To obtain equation 60, we showed that n1 and p1 can be
the temperature T(r), the recombination can be given by
presented in the following form: n1 ¼ n10 þ dn1 and
the following equations (see References 61, 63, 64, and 67):
p1 ¼ p10 þ dp1 . Here, dn 1 ¼ n10 =T 0 ð3=2 þ et =T 0 ÞdT and
e e
Rn ¼ xn ½nðN t nt Þ n1 nt dp1 ¼ p10 =T 0 3=2 tT0 g dT.
(56) By substitution of equation 60 into equation 56, we
Rp ¼ xp ½pnt p1 ðN t nt Þ þp0
obtain equation 55, where t1 ¼ xn xp N t x ðn0 þn10n0Þþx
n ðp þp Þ.
p 0 10
3GW8206 03/26/2014 0:6:48 Page 14
QUASINEUTRAL APPROXIMATION loss of physical information. Furthermore, in the limiting

case of rd ! 0 we are in QN, and the Poisson equation does
We can rewrite the Poisson equation (eq. 52) for nonequi- not need to be solved; instead, the value of w0 can be
librium variations in the following form: obtained by taking advantage of the fact that the electro-
chemical potential is constant across the structure in
d2 dw=dx2 ¼ 4pdr=e (61) equilibrium (see Reference 72).
where the electrical potential w ¼ w0 þ dw, bulk electrical It should be stressed that beyond the fact that the
charge r ¼ r0 þ dr, and w0 and r0 are the equilibrium quasineutrality concept is useful when dealing with trans-
electrical potential and bulk charge. port phenomena, QN is important in semiconductor
For intrinsic semiconductors dr ¼ eðdp dnÞ, and for devices because it is the basis of the operation of most
semiconductors of n- or p-type, dr ¼ eðdp dn dnt Þ. In semiconductor devices. In bipolar semiconductors, the
the last case, it is necessary to use equation 60. movement of charge packets requires a quasineutral
Quasineutrality (QN) is a basic concept in semiconductor nature of the packet (73, 74).
device analysis and is used widely in the literature on
transport phenomena. Its physical meaning has been exten-
sively reviewed (see, for instance, Reference 68). It can be EQUATIONS FOR COOLING AND HEATING IN p-n
readily seen from Poisson’s equation that dr=r0 ðrd =LÞ2 JUNCTIONS
(rd is the Debye length and L is the characteristic length of
the change of the chemical potential). If ðrd =LÞ2 1, then In the linear approximation with respect to the electric
dr ! 0. Under these conditions, the differential Poisson current, the heat balance equation is given by (see eq. 43)
equation becomes the following algebraic equation: the following:
r q ¼ eg R (64)
dr 0 (62)
which does not need boundary conditions. Additionally, Equation 64 represents the total heat flux conservation
this algebraic equation establishes a relationship between law that can be formulated as follows: Any change in the
the excess of both kinds of carrier and renders the Poisson drift component of the heat flux is accompanied by a
equation redundant in the system of equations 50–52. change in the thermal diffusion component of the same
Despite the widespread use of the QN hypothesis, it is in heat flux, and the heat bulk sources and sinks are given by
general mistakenly reduced to a simplified situation in the recombination processes.
which both excesses of majority and minority carriers The expression for qdr in bipolar semiconductors is
are the same: dn dp (69, 70). This assumption, which is
valid in some situations, can also lead to important errors in qdr ¼ Pn jn þ Pp jp (65)
other cases (55). From the equation dr ’ 0, it follows natu-
As introduced previously, the Peltier coefficients in non-
rally that quasineutrality does not necessarily imply
degenerated semiconductors are
dn dp, but
dp ¼ Adn Bðn0 þ p0 ÞdT=T 0 (63) 1 h
Pn;p ¼ gn;p þ 5 =2 T mn;p (66)
where e
xn ðN t n0 þ n2t0 p0 =p10 Þ þ xp N t p10 The expression for the diffusion heat fluxes qn;p
diff is as
A ¼ ; follows (see eq. 47):
xn N t n0 þ xp ðN t p10 þ n2t0 Þ

eg et qn;p n;p
diff ¼ kph rT (67)
xn n10 ðet =T 0 þ 3=2Þ xp p0 þ 3=2
T0 n2t0
B ¼
xn N t n0 þ xp ðN t p10 þ nt0 Þ
2 n0 þ p0 where kn;p
ph is the phonon heat conductivity in the n- and the
p-regions as indicated by the superscript.
Therefore, quasineutrality can be identified with the
Taking into account the considerations presented pre-
expression dn ¼ dp only in intrinsic semiconductors or in
viously, the heat balance equation can be rewritten as
doped ones if the variation in the trapped charge is weak.
follows:
From Reference 55, it follows that if rd ! 0, then w0 ðxÞ
tends to a constant value and is null elsewhere. Therefore,
kph DT þ Pn r jn þ Pp r jp þ jn rPn þ jp rPp ¼ eg R
w0 ðxÞ can be asymptotically approximated in the p-n junc-
tion by a step function with a discontinuity at the semi- (68)
conductor interfaces. In electrical terms, this means that Because the current densities can be calculated as (see
in quasineutrality, if the Debye length is very small (i.e., eq. 49),
when considering highly doped semiconductors), then the
jn ¼ sn ðrw rmn =e þ an rT Þ
space charge spread over a length comparable with the (69)
Debye length in the semiconductor can be replaced by a jp ¼ sp rw þ rmp =e þ ap rT
surface space charge (boundary layer function (71)) that
supports the discontinuity of the electric potential. the calculation of mn , mp , and w is required. It means
Because the Poisson equation cannot provide w0 , removing that we need equations 50–52. Under quasineutral approx-
it from the system of equations 50–52 will not lead to any imation, instead of equation 52 we have equation 62.
3GW8206 03/26/2014 0:6:48 Page 15
Because by definition (see Reference 72), n ¼ nn exp dT n ð0Þ ¼ dT p ð0Þ (79)

½mn =T; p ¼ np exp½mp =T, where mn;p ¼ m0n;p þ dmn;p , and
dn ¼ n0 =T 0 ½dmn þ ð3=2 m0n =T 0 ÞdT, dp ¼ p0 =T 0 ½dmp þ In the preceding equations, the superscript n(p) in mag-
ð3=2 m0p =T 0 ÞdT. nitudes refers to the n (p) region and subscript (n or p)
The system of equations 50, 51, 68, and 63 defines refers to the carriers (electrons or holes); additionally,
the mathematical framework of thermoelectric cooling, subscript 0 in a magnitude denotes equilibrium. Because
and it must be complemented with the appropriate bound- the total current density is constant across the junction, it
ary conditions that describe the electric currents, the is enough to formulate boundary conditions for one of the
heat flux, and the electric potential through the interfaces two current densities. These boundary conditions are
(75–77). Let us assume that in the y- and z-direction the p-n obtained assuming, respectively, continuity of the electro-
junction is adiabatically isolated. Then, the boundary con- chemical potential at the interface, that both heat and
ditions in the remaining direction (i.e., the p-n junction electrical conductivities are very large at the junction,
interface is orthogonal to the x-axis, and assuming that the and the absence of surface recombination. In fact, because
interface is located at x ¼ 0, the n region located between the p-n interface is inside the depletion region, this latter
x ¼ ln and x ¼ 0, the p region between x ¼ 0 and x ¼ lp) are assumption is not realistic, and boundary conditions with
given below. finite conductivities need to be used (75). In this work, we
Assuming that an ideal metal–semiconductor contact is use equations 76–79 for the sake of simplicity.
placed at x ¼ ln, we can write the following boundary
conditions for the excess of temperature and carrier densi-
ties (hereafter a superscript n or p in a magnitude refers to SIMPLIFICATION OF THE MODEL OF PELTIER EFFECT
the n or p region, respectively): IN TWO LIMITING CASES
dT n ðln Þ ¼ 0 (70) In this section, we will analyze the thermoelectric cooling

in a p-n junction in the two limiting cases: strong and weak
recombination.
dnðln Þ ¼ 0 (71)
dwðln Þ ¼ 0 (72) Weak Recombination
These boundary conditions are justified because of the high Let us now consider that volume recombination is weak. In
value of the thermal conductivity of metals and the intense this case, the conditions ln;p n;p n;p
D ln;p rD (lD are the dif-
recombination at the metal–semiconductor interface. Sim- fusion lengths in n-p regions) are fulfilled, which means
ilar boundary conditions may be written as follows at the that the weak recombination is correct for thin film p-n
metal–semiconductor interface at x ¼ lp: structures. Formally R ¼ 0 when t ! 1. Under this condi-
tion, the right-hand side of equation 43 becomes trivially
null and along with equations 65 and 67 transforms into
dT p lp ¼ 0 (73)
DT ¼ 0 (80)
dp lp ¼ 0 (74)
Equations 50 and 51 transform into
dwðlp Þ ¼ V (75)
r jn;p ¼ 0 (81)
where V is the applied voltage. These boundary conditions
assume that the semiconductor is at equilibrium in x ¼ ln From equation 81, it follows that jn;p are not spatially
and in x ¼ lp; in other situations, the electric potential (i.e., dependent and jnn þ jnp ¼ jpn þ jpp ¼ j0 , where j0 is the whole
eqs. 72 and 75) cannot be rigorously defined (78). At the p-n current through the p-n structure. From the boundary
junction interface, we can introduce the following addi- conditions for currents (75, 76), it follows that
tional boundary conditions (75, 76): jnn þ jpn ¼ jn0 , jnp þ jpp ¼ jp0 (jn0 þ jp0 ¼ j0 ).
It is not difficult to understand that the concentrations
dmnn ð0Þ 1 @mnn0 of the nonequilibrium carriers (dn and dp) are maxima in
dwn ð0Þ dT n ð0Þ this case.
e e @T
It may seem that the calculation of thermoelectric cool-
dmp ð0Þ 1 @mpn0 ing does not require the use of equation 81 in the absence of
¼ dwp ð0Þ n dT p ð0Þ
e e @T recombination because there are no other unknown func-
ð76Þ
dmnp ð0Þ 1 @mnn0 tions in equation 80. So, it seems that thermoelectric
dwn ð0Þ þ dT n ð0Þ
e e @T cooling does not depend on the nonequilibrium carrier
dmpp ð0Þ 1 @mpn0 concentrations. However, the boundary conditions to equa-
¼ dwp ð0Þ þ dT p ð0Þ tion 80 must be formulated for heat fluxes (eqs. 65 and 67).
e e @T
The drift heat fluxes depend on the current of majority and
minority carriers (eq. 65). The latter essentially depends
qn ð0Þ ¼ qp ð0Þ (77)
on the distribution of nonequilibrium carrier concentration
because of the terms ðrmn;p Þ=e. Therefore, there are no
jnn ð0Þ ¼ jpn ð0Þ (78) reasons to assume a priori that jnn jnp and jpp jpn .
3GW8206 03/26/2014 0:6:49 Page 16
The problem is reduced to the calculation of the currents Adding equation 38 we have
in the electrical circuit composed of two circuits connected
in parallel (see Figure 16). One of them is composed of two r jnn þ jnp ¼ 0; r jpn þ jpp ¼ 0 (86)
n-type semiconductors connected in series with the con-
centrations nn and np , whereas the other is composed of It is important to emphasize that the nonequilibrium
two p-type semiconductors connected in series with the charge carrier concentrations (dn and dp) are not equal
concentrations pn and pp . At a given direction of the to zero in the considered approximation. Therefore, there
current (from n- to p-region), heating instead of cooling is no reason to state that jnn jnp and jpp jpn .
takes place at weak recombination. Therefore, because ln;p D ! 0, R differs from zero only at
With respect to what was said previously, let us notice the interface. The volume equation (eq. 43) transforms
that the classic theory for current-voltage characteristic again into equation 80.
through the p-n junction (6, 14) obeys the following Once again, just like in the weak recombination case,
expression: the right-hand side of equation 43 also becomes zero but for

eV different physical reasons. But as in the case of weak
j0 ¼ js exp 1 (82) recombination, the heat flux depends on the nonequili-
T
brium carrier concentrations. The latter are defined by
where the saturation current (js) varies in direct proportion equations 85, 86, and 80 with the corresponding boundary
to the capture coefficients. It means the current j0 through conditions. In the previous case, it was noted that equation
the p-n junction is equal to zero when the recombination is 82 is not correct when the recombination is weak enough.
absent under any voltage. Also, it is not difficult to understand that equation 82 is not
The above means that the model (eq. 82) is not correct correct in the case of strong recombination. It follows from
for weak recombination. At the same time, equation 81 the expression for js that js ! 1 when t ! 0 at any applied
(together with eqs. 80 and 63) will give the correct expres- voltage V. The last statement is not correct from a physical
sion for the current density j0 (at least under weak bias point of view.
voltage).
Pnn s pn
p p
A main result is that the temperature deviation from
equilibrium at the junction may be obtained as follows: Pn sp
2 2 l The method described allows the calculation of the

2 lp n
dT n ð0Þ / j0 ln Ppn þ Pnp Ppn
þ Pnp current-voltage characteristic of the p-n junction in the
spp snn
" ! !# case of strong recombination in the linear regime with
p 2 lp ln ln lp respect to the applied voltage V.
þ j0 ln H Pn þ þ The temperature deviation at the junction has been
spn snp snp spp
" ! analytically obtained as follows:
2 l #
n p ln lp lp h i
þ j0 ln H Pp þ þ
spn snp spn snn dT n ð0Þ / j0 ln snn spp Pnn Ppp þ snp spp Pnp snn spn Ppn (87)
" #
2 l l This expression clearly differs from the commonly used
n p n p
þ j0 ln H Pp þ Pn
spn snp expression as follows (6, 9, 14, 18):
(83)
dT n ð0Þ / j0 snn spp Pnn Ppp ln (88)
The expression of H is
T0 n
The differences are not only in magnitude but also in sign.
H¼ kph lp þ kpph ln (84)
ln lp Contrary to equation 88, indicating that the positive values
of j0 only predict a decrease in temperature with j0, equa-
From equation 83, it follows that a positive current will tion 87 predicts that a p-n junction under the same bias
generate heat instead of cooling at the junction, which is in conditions (positive values of j0) may be heated or cooled
strong contrast with the conventional results. depending on the values of the Peltier coefficients and
electrical conductivities in the p-n junction. Moreover,
Strong Recombination equation 87 clearly shows the paramount importance of
the nonequilibrium carriers at both sides of the junction
Let us assume that the recombination is very strong. The
(neglected in equation 88) because they control the sign of
physical meaning is that ln;p ln:p n;p
D ! 0 rD ! 0. From a dT n ð0Þ.
mathematical point of view, we have t ! 0 in equation 55.
Finally, let us emphasize that only when the following
Because the recombination rate (R) cannot be infinite at
two criteria are met simultaneously:
t ! 0, from equation 55 follows that
p0 n0 Pnn spn Ppp snp

dn þ dp ¼ ldT (85) p p; n n (89)
n0 þ p0 n0 þ p0 Pn sp Pp sn
At the same time, the magnitude R is finite but not defined. Equation 87 reduces to equation 88.
3GW8206 03/26/2014 0:6:49 Page 17
APPLICATIONS magnitudes in the figure of merit (a2 s). Although this

ignited intense research into nanostructured thermo-
Research and development of bulk homogeneous materials electric materials, there are still many debates about the
for thermoelectric applications began to increase starting exact role that low dimensionality and nanostructures
in the 1950s and resulted in commercial solid-state power could play in improving thermoelectrics. Table 1 summa-
generation and cooling systems (5). With the introduction rizes major ZT milestones achieved over the past decades
of materials and concepts based on nanostructuring, the (79).
field has witnessed truly dramatic growth over the past 15 It is important to note that some of the ZT values in
years. Nanostructuring of semiconductors in the form of Table 1 have not yet been independently verified or repro-
quantum wells and superlattices started in the 1960s and duced, and corroboration of these key breakthroughs
early 1970s for electrical and optical devices. In contrast, would be helpful to the thermoelectrics community.
the primary benefit to date of nanostructuring for thermo- The field of thermoelectrics presents an important chal-
electric materials has been the impact on thermal propert- lenge to synthetic chemists, physicists, and materials sci-
ies rather than electronic (79). entists. The discovery of new and promising materials
The performance of thermoelectric materials is quanti- requires a combination of theoretical guidance, keen chem-
fied by the previously introduced unitless figure of merit ical intuition, synthetic chemistry expertise, materials
ZT, which is defined as Z ¼ a2 s=k. The theory of thermo- processing, and good measurement skills. This powerful
electric efficiency was outlined by Altenkirch (80, 81). It combination can be effectively achieved by reaching across
was based on a simple procedure of taking into account scientific disciplines (99).
Joule losses and losses caused by thermal conductivity. A common theme among many emerging thermoelectric
The resultant expression was given in the form proposed materials is the concept of nanostructuring to improve
by Ioffe (5). The process of energy conversion was studied the thermoelectric performance. In these systems, the
by Harman and Honig (82), Anatychuk (9), Nolas et al. (6), enhanced thermoelectric performance is attributable to a
and Tauc (14). strong decrease in lattice thermal conductivity rather than
In the 1950s, it was discovered that alloys of Bi2Te3 had an increase in the electrical power factor (a2 s). Thus, in
a value of ZT 1 near room temperature, and they have certain cases, nanodots clearly play a significant role in
played a dominant role in the field of thermoelectrics reducing lattice thermal conductivity, probably by effec-
through today. Although each magnitude in the expression tively scattering phonons that otherwise would have rela-
of ZT can individually be changed by several orders of tively long mean free paths. In many of these cases, it has
magnitude, the interdependence and coupling between been clearly demonstrated that the reduction in thermal
these properties have made it extremely difficult to conductivity far exceeds any concomitant reduction in the
increase ZT above the unity despite five decades of active power factor caused by electronic carrier scattering, thus
research. The thermoelectrics community is widely target- resulting in enhanced ZT values (99). Thin-film super-
ing ZT > 3 to make these solid-state systems competitive lattices and nanowires make up a significant part of cur-
with traditional mechanical energy conversion systems rent thermoelectric research.
(79). In review (100), strategies to improve the thermo-
In 1993, Hicks and Dresselhaus (13, 83) pioneered the electric figure of merit, new discussions on device physics
concept that quantum confinement of electrons and holes and applications, and assessments of challenges on
in low-dimensional materials could dramatically increase these topics were presented. Understanding of phonon
ZT above the unity by independently changing the transport in bulk materials has advanced significantly
Table 1. Thermoelectric Figure-of-Merit ZT as a Function of Temperature and Year Illustrating Important Milestones
Material ZT T (K) Year Reference

Bi2Te3/Sb2Te3 (supperlattice) 2.4 300 2001 84
PbSeTe (nano dot superlattice) 1.6 300 2002 85
AgPbmSbTemþ2 2.2 800 2004 86
Bi2Te3 0.77 300 1958 87
Pb1xEuxTe/PbTe (multi quantum well) 1.2 300 1996 88
CsBi4Te6 0.8 225 2000 89
Si NW (nanowire) 1.0 200 2008 90
Si NW (nanowire) 0.6 300 2008 91
BiSbTe 1.4 373 2008 92
Na1xSbmSbyTemþ2 1.7 700 2006 93
In4Se3d 1.48 705 2009 94
PbTe (Ta-doped) 1.5 773 2008 95
Ba8Ga16Ge30 1.35 900 2006 96
SiGe 1.3 1173 2008 97
SiGe NW 0.46 450 2012 98

Although there have been several demonstrations of ZT > 1 in the past decade, no material has yet achieved the target goal of ZT 3. The
material systems that have achieved ZT > 1 have all been based on some form of nanostructuring.
3GW8206 03/26/2014 0:6:49 Page 18
Table 2. Main Advantages and Disadvantages of Thermoelectric Thermoelectric phenomena, such as the Peltier effect,
Coolers are at the crossroads of several fields of physics and
Favorable Unfavorable
engineering such as solid-state electronics, nonequili-
brium thermodynamics, transducer devices, and so on.
CFC-free operation A comprehensive study of the mechanisms of heating and
Free of moving parts and fluids cooling originated by an electrical current in semi-
Vibrations and acoustic noise conductor devices has been presented in this article.
DC power instead of wall AC
Thermoelectric cooling in n-n, p-p, and p-n junction con-
or lead batteries
Maintenance tacts and inhomogeneous bulk semiconductors was ana-
Efficiency lyzed. Both degenerate and nondegenerate electron and
Cooling power hole gases were considered. The eminent roles of
recombination and nonequilibrium charge carriers in
heating/cooling of solid-state devices, usually ignored
in the Peltier effect’s literature, has been discussed.
as the first-principles calculations are applied to thermo- Along with the above, special attention has been paid
electric materials, and new experimental tools are being to several aspects of nonequilibrium thermodynamics of
developed. As a result, some new strategies have been the thermoelectric phenomena involved in Peltier effect
developed to improve electron transport in thermoelectric in semiconductors that demanded a careful examination.
materials. In particular, some inaccuracies in the traditional theory
A thermoelectric device consists of heavily doped semi- widely extended in the literature were discussed, and a
conductor legs that are connected electrically in series and recently formulated self-consistent theoretical model,
thermally in parallel. A fundamental understanding of better describing the Peltier effect, was presented.
heat and charge carrier transport inside the thermoelectric Finally, a glimpse of applications was given with special
legs will lead to new strategies to design and fabricate emphasis on the evolution of experimental values of the
high-efficiency thermoelectric materials. The device effi- figure of merit ZT in the past decades. A reflection was
ciency depends not only on materials but also on an opti- made about the lack of figure of merit for thermoelectric
mum choice of the legs’ size, configuration, and contacts. devices involving two or more materials (such as those
Finding new applications for thermoelectric devices in based on the Peltier effect).
places that they are superior to other technologies is
another challenge for the thermoelectric community.
Table 2 summarizes the main advantages and disad- ACKNOWLEDGMENTS
vantages of thermoelectric coolers.
As a motivating conclusion of this section, it is impor- Yu. G. Gurevich thanks CONACYT-Mexico for financial
tant to examine more closely the meaning of ZT and its support. J. E. Velazquez-Perez wants to thank Spanish
limits as a figure-of-merit in thermoelectric devices. ZT is a Ministerio de Economıa y Competitividad (MINECO) and
good parameter to characterize thermoelectric materials. FEDER for financial support under grant TEC2012-32777.
When a thermoelectric device consists of just one homoge- We would like also thank Associate Editor Ms. Cassandra
neous material (as it is often the case in thermoelectric Strikland for her valuable help and assistance in the
generators), ZT will be truly useful to characterize the processing of this manuscript.
efficiency of the device. Nevertheless, in thermoelectric
cooling (Peltier effect), it is necessary to have a contact
between two materials. In this situation, nonequilibrium BIBLIOGRAPHY
electrons and holes will set up. Which parameter that could
characterize the efficiency of cooling system as a whole was 1. J. C. Peltier. Nouvelles Experiences sur la Caloricite
not known until today. des Courants Electriques. Ann. Chim. Phys. 1834, 56,
pp 371–386.
2. E. H. Lentz. Electrical Experiment of Academician Lentz.
SUMMARY Water Freezing by Galvanic Stream. Bibliotec. Chten. 1838,
28, pp 44–48.
The Peltier effect is the heat extraction or absorption that 3. W. Thomson. On Mechanical Theory of Thermoelectric
occurs at the contact between two different conducting Currents, in Proc. Roy. Soc. Edinburgh, 1851, pp 9198.
media when a DC electric current flows through this 4. L. Onsager. Reciprocal Relations in Irreversible Processes I.
contact. Devices based on the Peltier effect, which is the Phys. Rev. 1931, 37, pp 405–426; Reciprocal Relations in
basis for solid-state thermoelectric cooling, have evolved Irreversible Processes II. 1931, 38, pp 2265–2279.
rapidly to meet the pace of the ever-growing industry of 5. A. F. Ioffe. Semiconductor Thermoelements and Thermo-
electronics. In recent years, new solid-state cooling solu- electric Cooling. London, U.K.: Infosearch, 1957.
tions based on new materials and devices (such as gra- 6. G.S. Nolas, J. Sharp, and H. J. Goldsmid. Thermoelectrics.
phene, Si and SiGe nanowires, etc.) have been proposed. To Basic Principles and New Materials Development. Springer:
drive the quest for new high-efficiency cooling devices, a Berlin, Germany, 2001.
suitable knowledge of the theory behind thermoelectric 7. H. J. Goldsmid. Thermoelectric Refrigeration. Plenum:
cooling is of paramount importance. New York, 1964.
3GW8206 03/26/2014 0:6:49 Page 19
8. D. M. Rowe and C. M. Bhandari. Modern Thermoelectric. 30. I. N. Volovichev, G. N. Logvinov, O. Y. Titov, and Y. G.
Rinehart & Winston: London, U.K., 1983. Gurevich. Recombination and Lifetimes of Charge
9. L. I. Anatychuk. Physics of Thermoelectricity. Institute of Carriers in Semiconductors. J. Appl. Phys. 2004, 95,
Thermoelectricity: Kyiv, Chernivtsi, 1998. pp 4494–4496.
10. G. D. Mahan. Good Thermoelectrics. Solid State Phys. 1998, 31. L. D. Landau, E. M. Lifshitz, and L. P. Pitaevskii. Electro-
51, pp 81–157. dynamics of Continuous Media, 2nd ed., Course of Theoreti-
11. G. D. Mahan and J. O. Sofo. The Best Thermoelectric. Proc. cal Physics, Vol. 8, eds. L. D. Landau and E. M. Lifshitz.
Natl. Acad. Sci. USA 1996, 93, pp 7436–7439. Pergamon Press: New York, 1984.
32. G. N. Logvinov, J. E. Vel
azquez, I. M. Lashkevich, and Y. G.
12. T. M. Tritt, M. G. Kanatzidis, H. B. Lyon, Jr., and G. D.
Mahan. Thermoelectric Materials–New Directions and Gurevich. Heating and Cooling in Semiconductor Structures
Approaches. Mater. Res. Soc. Proc. 1997, 478, pp 73–84. by an Electric Current. Appl. Phys. Lett. 2006, 89, p 092118.
13. L. D. Hicks and M. D. Dresselhaus. Effect of Quantum-Well 33. Y. G. Gurevich and G. N. Logvinov. Formulation and Mea-
surement of the Thermo-e.m.f. in Unipolar Semiconductors.
Structures on the Thermoelectric Figure of Merit. Phys. Rev.
Phys. Rev. B 1992, 46, pp 15516–15517.
B 1993, 47, pp 12727–12731.
14. J. Tauc. Photo and Thermoelectric Effects in Semiconductors. 34. Y. G. Gurevich, O. Yu. Titov, G. N. Logvinov, and O. I.
Pergamon Press: New York, 1962. Lyubimov. Nature of the Thermopower in Bipolar Semi-
conductor. Phys. Rev. B 1995, 51, pp 6999–7004.
15. H. B. Callen. The Application of Onsager’s Reciprocal Rela-
35. I. Lashkevych, C. Cortes, Y. G. Gurevich. Physics of Ther-
tions to Thermoelectric, Thermomagnetic, and Galvanomag-
netic Effects. Phys. Rev. 1948, 73, pp 1349–1358. moelectric Cooling: Alternative Approach. J. Appl. Phys.
2009, 105, p 053706.
16. A. G. Samoilovich and L. L. Korenblit. Current State of
Theory of Thermoelectric and Thermomagnetic Phenomena 36. Y. G. Gurevich and J. E. Vel
azquez-Perez. The Role of Non-
Equilibrium Charge Carriers in Thermoelectric Cooling.
in Semiconductors. Uspekhi Fiz. Nauk 1953, 49, pp 243–272
J. Appl. Phys. 2013, 114, p 1033704.
(in Russian), German version: Fortschr. Phys. 1953, 1,
pp 486–512. 37. Y. G. Gurevich and S. I. Shevchenko. Cooling of Electrons in
17. N. W. Ashcroft and N. D. Mermin. Solid State Physics. Semiconductors by a Static Electric Field. Sov. Phys. JETP
Cornell University: Ithaca, NY, 1976. 1972, 35, pp 426–428.
38. B. I. Davydov. To the Theory of Electron Motion in Gases
18. D. M. Rowe. Thermoelectrics Handbook, Macro to Nano. CRC
Press, LLC: Boca Raton, FL, 2006. and Semiconductors. Sov. Phys. JETP 1937, 7, pp 1069–
1072; B. I. Davydov and I. Shmushkevich. Electrical
19. V. F. Gantmakher and I. B. Levinson. Carrier Scattering in
Conduction in Semiconductors with a Complete Lattice
Metal and Semiconductors, Modern Problems in Condensed Embedded in a Strong Field. Sov. Phys. JETP 1940, 10,
Matter Science, Vol. 19. North-Holland: Amsterdam, the pp 1043–1048.
Netherlands, 1987.
39. I. Lashkevich, O. Angeles Fragoso, and Y. G. Gurevich.
20. F. J. DiSalvo. Thermoelectric Cooling and Power Generation.
Peculiarities of Thermoelectric Cooling in p-n Structures.
Science 1999, 285, pp 703–706. Int. J. Thermophys. 2009, 30, pp 635–647.
21. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B. 40. Y. G. Gurevich and I. Lashkevych. Non-Typical Temperature
O’Quinn. Thin-Film Thermoelectric Devices with High Distribution in p-n Structure Under Thermoelectric Cooling.
Room-Temperature Figures of Merit. Nature 2001, 413, pp
Int. J. Therm. Sci. 2009, 48, pp 2080–2084.
597–602.
41. D. M. Rowe, CRC Handbook of Thermoelectrics. CRC Press:
22. Y. G. Gurevich and G. N. Logvinov. Physics of Thermo- London, 1996.
electric Cooling. Semicond. Sci. Technol. 2005, 20,
42. K. Seeger. Semiconductor Physics. Springer: Berlin, Ger-
pp R57–R64.
many, 1985.
23. Y. G. Gurevich and G. N. Logvinov. Theory of Thermoelectric
Cooling in Semiconductor Structures. Revista Mexicana 43. V. I. Stafeev. Modulation of the Diffusion Length as a New
Fisica 2007, 53, pp 337–349. Working Principle of Semiconducting Devices. Sov. Phys.
Solid State 1959, 1, pp 841–847.
24. S. R. de Groot and P. Mazur. Non-Equilibrium Thermo-
dynamics. Dover: New York, 1984. 44. V. I. Stafeev and E. I. Karakushan. Magnetodiodes: New
Semiconductor Devices with High Sensitivity to a Magnetic
25. Y. G. Gurevich, G. N. Logvinov, O. Yu. Titov, and J. Giraldo.
Field. Nauka: Moskow, Russia, 1975.
New Physical Principles of Contact Thermoelectric Cooling.
Surf. Rev. Lett. 2002, 9, pp 1703–1708. 45. R. N. Hall. An Analysis of the Performance of Thermoelectric
Devices Made from Long Lifetime Semiconductors. Solid
26. F. G. Bass, V. S. Bochkov, and Y. G. Gurevich. Electrons and State Electron. 1961, 2, pp 115–122.
Phonons in Bounded Semiconductors. Nauka: Moscow, Rus-
sia, 1984. 46. M. Cutler. Thermoelectric Behavior of p-n Junctions. J. Appl.
Phys. 1961, 32, pp 222–225.
27. Y. G. Gurevich and F. Perez Rodriguez. Fenomenos de Trans-
47. M. A. Melehy. Theory of Minority Carrier Thermoelectric
porte en Semiconductores. Fondo de Cultura Economica:
Mexico, 2007. Cooling. J. Appl. Phys. 1962, 33, pp 2100–2106.
28. Y. G. Gurevich and G. N. Logvinov. Thermo-EMF and Ther- 48. A. L. Vainer, N. V. Kolomoets, E. M. Lukishker, and V. M.
moelectric Current in Unipolar Semiconductors with Finite Rzhcvsky. To the Theory of Combined Thermoelement. Sov.
Dimensions. Sov. Phys. Semicond. 1992, 26, pp 1091–1094. Phys. Semicond. 1977, 11, pp 546–552.
49. U. Lindefelt. Heat Generation in Semiconductor Devices.
29. G. N. Logvinov, Y. G. Gurevich, and I. M. Lashkevich.
Surface Heat Capacity and Surface Heat Impedance: An J. Appl. Phys. 1994, 75, pp 942–957.
Application to Theory of Thermal Waves. Jpn. J. Appl. 50. G. D. Mahan. Density Variations in Thermoelectrics. J. Appl.
Phys. 2003, 42, pp 4448–4452. Phys. 2000, 87, pp 7326.
3GW8206 03/26/2014 0:6:49 Page 20
51. K. P. Pipe, R. J. Ram, and A. Shakouri. Bias-Dependent Carriers in the Formation of Thermo-e.m.f. in Bipolar Semi-
Peltier Coefficient and Internal Cooling in Bipolar Devices. conductors. Phys. Status Solidi B 2002, 231, pp 278–293.
Phys. Rev. B 2002, 66, p 125316. 73. W. van Roosbroeck. Current-Carrier Transport with Space
52. V. I. Stafeev. Thermoelectric and Other Phenomena in Struc- Charge in Semiconductors. Phys. Rev. 1961, 123, pp 474–490.
tures with Nonequilibrium Charge Carriers and Nanopar- 74. Y. Moureau, J.-C. Manifacier, and H. K. Henisch. Minority-
ticles. Semiconductors 2009, 43, pp 1280–1287. Carrier Injection into Relaxation Semiconductors. J. Appl.
53. D. Fu, A. X. Levander, R. Zhang, J. W. Ager, III, and J. Wu. Phys. 1986, 60, pp 2904–2909.
Electrothermally Driven Current Vortices in In- 75. O. Y. Titov, J. Giraldo, and Y. G. Gurevich. Boundary Con-
homogeneous Bipolar Semiconductors. Phys. Rev. B 2011, ditions in an Electric Current Contact. Appl. Phys. Lett.
84, p 045205. 2002, 80, pp 3108–3110.
54. W. M. Bullis. Minority Carrier Thermoelectric Cooling. 76. I. N. Volovichev, J. E. Velazquez-Perez, and Y. G. Gurevich.
J. Appl. Phys. 1963, 34, pp 1648–1649. Transport Boundary Condition for Semiconductor Struc-
55. Y. G. Gurevich, J. E. Velazquez-Perez, G. Espejo-Lopez, I. N. tures. Solid-State Electron. 2008, 52, pp 1703–1709.
Volovichev, and O. Y. Titov. Transport of Nonequilibrium 77. I. Lashkevich and Y. G. Gurevich. Boundary Conditions for
Carriers in Bipolar Semiconductors. J. Appl. Phys. 2007, Thermoelectric Cooling in p-n Junction. Int. J. Thermophys.
101, p 023705. 2011, 32, pp 1086–1097.
56. Y. G. Gurevich and I. N. Volovichev. Forgotten Mechanism of 78. Y. G. Gurevich and V. B. Yurchenko. Problem of Formation of
Nonlinearity in the Theory of Hot Electrons. Phys. Rev. B an EMF in a Semiconductor and its Transfer to an External
1999, 60, pp 7715–7717. Circuit. Sov. Phys. Semicond. 1991, 25, pp 1268–1273.
57. I. Lashkevich, O. Titov, and Y. G. Gurevich. Recombination 79. C. J. Vineis, A. Shakouri, A. Majumdar, and M. G. Kanatzi-
and Temperature Distribution in Semiconductors. Semi- dis. Nanostructured Thermoelectrics: Big Efficiency Gains
cond. Sci. Technol. 2012, 27, pp 055014. from Small Features. Adv. Mater. 2010, 22, pp 3970–3980.
58. Y. G. Gurevich and I. Lashkevych. Sources of Fluxes of €
80. E. Altenkirch. Uber den Nutzeffekt der Thermos€ aulen.
Energy, Heat, and Diffusion Heat in a Bipolar Semi- Physik. Z. 1909, 10, pp 560–580.
conductor: Influence of Nonequilibrium Charge Carriers. 81. E. Altenkirch. Elektrothermische K€ alteerzeugung und
Int. J. Thermophys. 2013, 34, pp 341–349. Reversible Elektrische Heizung. Physik. Z. 1911, 12,
59. W. Shockley. The Theory of p-n Junctions in Semiconductors pp 920–924.
and p-n Junction Transistors. Bell Syst. Tech. J. 1949, 28, pp 82. T. C. Harman and J. M. Honig. Thermoelectric and Thermo-
435–489. magnetic Effects and Applications. McGraw Hill: New York,
60. R. A. Smith. Semiconductors. Cambridge University Press: 1967.
Cambridge, U.K., 1961.
83. L. D. Hicks, T. C. Harman, and M. S. Dresselhaus. Use of
61. D. A. Neamen. Semiconductor Physics and Devices: Basic Quantum-Well Superlattices to Obtain a High Figure of
Principles. Irwin: Homewood, IL, 1992. Merit From Nonconventional Thermoelectric Materials.
62. V. L. Bonch-Bruevich and S. G. Kalashnikov. Physik fur Appl. Phys. Lett. 1993, 63, pp 3230–3232.
Halbleiter. VEB Deutscher Verlagder Wissenschaften: 84. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B.
Berlin, Germany, 1982. O’Quinn. Thin-Film Thermoelectric Devices with High
63. S. Sze. Physics of Semiconductor Devices. Wiley: New York, Room-Temperature Figures of Merit. Nature 2001, 413,
1981. pp 597–602.
64. P. T. Landsberg. Recombination in Semiconductors. 85. T. C. Harman, P. J. Taylor, M. P. Walsh, and B. E. LaForge.
Cambridge University Press: Cambridge, U.K., 1991. Quantum Dot Superlattice Thermoelectric Materials and
65. I. N. Volovichev, G. Espejo, Y. G. Gurevich, O. Yu. Titov, and Devices. Science 2002, 297, pp 2229–2232.
A. Meriuts. Recombination in Semiconductors: Appearance 86. K. F. Hsu, S. Loo, F. Guo, W. Chen, J. S. Dyck, C. Uher, T.
of Nonequilibrium Carriers due to Injection or Redistribution Hogan, E. K. Polychroniadis, and M. G. Kanatzidis. Cubic
in the Sample. Jpn. J. Appl. Phys. 2002, 41, Part 1, AgPb(m)SbTe(2þm): Bulk Thermoelectric Materials with
pp 182–184. High Figure of Merit. Science 2004, 303, pp 818–821.
66. G. Espejo-Lopez, O. I. Lyubimov, O. Yu. Titov, and Y. G. 87. H. J. Goldsmid, A. R. Sheard, and D. A. Wright. The Per-
Gurevich. Corriente Electrica de Portadores Calientes en formance of Bismuth Telluride Thermojunctions. Br. J. Appl.
Semiconductores Intrinsecos: Desequilibrio en la Concentra- Phys. 1958, 9, pp 365–370.
cion. Revista Mex. Fisica 2004, 50, pp 620–624. 88. T. C. Harman, D. L. Spears, and M. J. Manfra. High
67. W. Shockley and W. T. Read. Statistics of the Recombinations Thermoelectric Figures of Merit in PbTe Quantum Wells.
of Holes and Electrons. Phys. Rev. 1952, 87, pp 835–842. J. Electron. Mater. 1996, 25, pp 1121–1127.
68. J. N. Chazalviel. Coulomb Screening by Mobile Charges. 89. D.-Y. Chung, T. Hogan, P. Brazis, M. Rocci-Lane, C.
Birkhauser: Boston, MA, 1999. Kannewurf, M. Bastea, C. Uher, and M. G. Kanatzidis.
69. S. R. in t’Hout. Quasineutrality in Semiconductors. J. Appl. CsBi(4)Te(6): A High-Performance Thermoelectric Material
Phys. 1996, 79, pp 8435–8444. for Low-Temperature Applications. Science 2000, 287,
70. M. Krcmar and W. Saslow. Exact Surface Solutions for pp 1024–1027.
Semiconductors: The Dember Effect and Partial Currents. 90. A. I. Boukai, Y. Bunimovich, J. Tahir-Kheli, J. K. Yu, W. A.
Phys. Rev. B 2002, 65, p 233313. Goddard, and J. R. Heath. Silicon Nanowires as
Efficient Thermoelectric Materials. Nature 2008, 451, pp
71. C. Miranda. Partial Differential Equations of Ellyptic Type.
Springer: Berlin, Germany, 1970. 168–171.
72. Y. G. Gurevich, G. N. Logvinov, I. N. Volovichev, G. Espejo, 91. A. I. Hochbaum, R. K. Chen, R. D. Delgado, W. J. Liang, E. C.
O. Yu. Titov, and A. Meriuts. The Role of Non-Equilibrium Garnett, M. Najarian, A. Majumdar, and P. D. Yang.
3GW8206 03/26/2014 0:6:49 Page 21
Enhanced Thermoelectric Performance of Rough Silicon 97. X. W. Wang, H. Lee, Y. C. Lan, G. H. Zhu, G. Joshi, D. Z.
Nanowires. Nature 2008, 451, pp 163–167. Wang, J. Yang, A. J. Muto, M. Y. Tang, J. Klatsky, S. Song,
92. B. Poudel, Q. Hao, Y. Ma, Y. C. Lan, A. Minnich, B. Yu, M. S. Dresselhaus, G. Chen, and Z. F. Ren. Enhanced
X. Yan, D. Z. Wang, A. Muto, D. Vashaee, X. Y. Chen, Thermoelectric Figure of Merit in Nanostructured n-Type
J. M. Liu, M. S. Dresselhaus, G. Chen, and Z. Ren. High- Silicon Germanium Bulk Alloy. Appl. Phys. Lett. 2008, 93,
Thermoelectric Performance of Nanostructured Bismuth pp 193121.
Antimony Telluride Bulk Alloys. Science 2008, 320, 98. E. K. Lee, L. Yin, Y. Lee, J. W. Lee, S. J. Lee, J. Lee, S. N. Cha,
pp 634–638. D. Whang, G. S. Hwang, K. Hippalgaonkar, A. Majumdar, C.
93. P. F. P. Poudeu, J. D’Angelo, A. D. Downey, J. L. Short, Yu, B. L. Choi, J. M. Kim, and K. Kim. Large Thermoelectric
T. P. Hogan, and M. G. Kanatzidis. High Thermoelectric Figure-of-Merits from SiGe Nanowires by Simultaneously
Figure of Merit and Nanostructuring in Bulk p-type Measuring Electrical and Thermal Transport Properties.
Na1xPbmSbyTemþ2. Angew. Chem. Int. Ed. 2006, 45, Nano Lett. 2012, 12, pp 2918–2923.
pp 3835–3839. 99. J.R. Sootsman, D. Y. Chung, and M. G. Kanatzidis. New and
94. J.-S. Rhyee, K. H. Lee, S. M. Lee, E. Cho, S. I. Kim, E. Lee, Old Concepts in Thermoelectric Materials. Angew. Chem.
Y. S. Kwon, J. H. Shim, and G. Kotliar. Peierls Distortion as a Int. Ed. 2009, 48, pp 8616–8639.
Route to High Thermoelectric Performance in In4Se3d 100. M. Zebarjadi, K. Esfarjani, M. S. Dresselhaus, Z. F. Ren and
Crystals. Nature 2009, 459, pp 965–968. G. Chen. Perspectives on Thermoelectrics: From Fundamen-
95. J. P. Heremans, V. Jovovic, E. S. Toberer, A. Saramat, K. tals to Device Applications. Energy Environ. Sci. 2012, 5,
Kurosaki, A. Charoenphakdee, S. Yamanaka, and G. J. pp 5147–5162.
Snyder. Enhancement of Thermoelectric Efficiency in
PbTe by Distortion of the Electronic Density of States.
YU. G. GUREVICH
Science 2008, 321, pp 554–557.
96. A. Saramat, G. Svensson, A. E. C. Palmqvist, C. Stiewe, E. CINVESTAV-IPN
Mueller, D. Platzek, S. G. K. Williams, D. M. Rowe, J. D.
Bryan, and G. D. Stucky. Large Thermoelectric Figure of J. E. VELAZQUEZ-PEREZ
Merit at High Temperature in Czochralski-Grown Clathrate Universidad de Salamanca
Ba8Ga16Ge30. J. Appl. Phys. 2006, 99, pp 023708.
View publication stats

Peltier Effect in Semiconductors: April 2014

Uploaded by

Copyright:

Available Formats

Peltier Effect in Semiconductors: April 2014

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Peltier Effect in Semiconductors: April 2014

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Peltier Effect in Semiconductors

Chapter · April 2014

Yuri G. Gurevich Jesús Enrique Velázquez-Pérez

SEE PROFILE SEE PROFILE

Terahertz detectors View project

Mono- and bipolar transport in semiconductor structures. View project

The user has requested enhancement of the downloaded file.

2 Peltier Effect in Semiconductors

Figure 1. Schematic of a thermoelectric couple.

Peltier Effect in Semiconductors 3

An important question arises regarding these consider-

4 Peltier Effect in Semiconductors

it is associated with particle drift in a single direction; this j

Peltier Effect in Semiconductors 5

(1) dT1 (2) dT2

Figure 8. Heating or cooling of a metal-n-type semiconductor

6 Peltier Effect in Semiconductors

Comparing equations 10 and 12, we see that the Peltier

where w ¼ q þ cj is the energy flux of charge carriers

We have made the assumption that nonequilibrium carri-

Peltier Effect in Semiconductors 7

8 Peltier Effect in Semiconductors

Peltier Effect in Semiconductors 9

j ¼ N=N 0 sT e g ½E T e =e 1=NdN=dx ð1 þ gÞ 1=edT e =dx

10 Peltier Effect in Semiconductors

Figure 13. Energy band diagram of a p-n junction.

Figure 14. Energy band diagram a p-n junction without

Peltier Effect in Semiconductors 11

The energy balance equation has never been sufficiently w ¼ q þ cn jn þ cp jp (33)

12 Peltier Effect in Semiconductors

Peltier Effect in Semiconductors 13

Rn ¼ Rp ¼ R ¼ xðnp n2i Þ (54)

14 Peltier Effect in Semiconductors

QUASINEUTRAL APPROXIMATION loss of physical information. Furthermore, in the limiting

Peltier Effect in Semiconductors 15

Because by definition (see Reference 72), n ¼ nn exp dT n ð0Þ ¼ dT p ð0Þ (79)

dT n ðln Þ ¼ 0 (70) In this section, we will analyze the thermoelectric cooling

dwðln Þ ¼ 0 (72) Weak Recombination

16 Peltier Effect in Semiconductors

 2  2 l  The method described allows the calculation of the

p0 n0 Pnn spn Ppp snp

Peltier Effect in Semiconductors 17

APPLICATIONS magnitudes in the figure of merit (a2 s). Although this

Material ZT T (K) Year Reference

18 Peltier Effect in Semiconductors

Peltier Effect in Semiconductors 19

20 Peltier Effect in Semiconductors

Peltier Effect in Semiconductors 21

View publication stats

You might also like

j ¼ N=N 0 sT e g ½E T e =e 1=NdN=dx ð1 þ gÞ 1=edT e =dx

2 2 l The method described allows the calculation of the