Page 1 of 7
10/19/2016
2017-01-0121
Improved Thermoelectric Generator Performance using High Temperature
Thermoelectric Materials
Author, co-author (Do NOT enter this information. It will be pulled from participant tab in
MyTechZone)
Affiliation (Do NOT enter this information. It will be pulled from participant tab in MyTechZone)
Abstract
Thermoelectric generator (TEG) has received more and more
attention in its application in the harvesting of waste thermal energy
in automotive engines. Even though the commercial Bismuth
Telluride thermoelectric material only have 5% efficiency and 250°C
hot side temperature limit, it is possible to generate peak 1kW
electrical energy from a heavy-duty engine. If being equipped with
500W TEG, a passenger car has potential to save more than 2% fuel
consumption and hence CO
2
emission reduction. TEG has advantages
of compact and motionless parts over other thermal harvest
technologies such as Organic Rankine Cycle (ORC) and Turbo-
Compound (TC). Intense research works are being carried on
improving the thermal efficiency of the thermoelectric materials and
increasing the hot side temperature limit. Future thermoelectric
modules are expected to have 10% to 20% efficiency and over 500°C
hot side temperature limit. This paper presents the experimental
synthesis procedure of both p-type and n-type skutterudite
thermoelectric materials and the fabrication procedure of the
thermoelectric modules using this material. These skutterudite
materials were manufactured in the chemical lab in the University of
Reading and then was fabricated into modules in the lab in Cardiff
University. These thermoelectric materials can work up to as high as
500°C temperature and the corresponding modules can work at
maximum 400°C hot side temperature. The performance loss from
materials to modules has been investigated and discussed in this
paper. By using a validated TEG model, the performance
improvement using these modules has been estimated compared to
commercial Bisemous Telluride modules.
Introduction
More and more research efforts have been invested in waste thermal
energy recovery in automotive engine in endeavor to improve the fuel
economy and reduce CO
2
emissions. There are multiple choices of
technologies for waste thermal energy recovery in automotive
applications. Among them there are Mechanical Turbo-Compounding
(MTC) , electrical turbo-compounding (ETC), Thermoelectric
Generator (TEG) and Organic Rankine Cycle (ORC) [1-5]. Each
technique has its own advantages and disadvantages. It is also
possible to use different harvest techniques within a same engine
system. Among them only TEG can claim all the following three
advantages: 1) without moving parts; 2) harvest the thermal energy
from multiple locations; 3) potential to be integrated to aftertreatment
system.
The state-of art commercial Thermoelectric Module (TEM) are made
from Bismuth Telluride thermoelectric material with maximum 5%
efficiency and 250°C hot side temperature limit. It was reported that
the TEG made of these modules could generate peak 1kW electrical
energy from a heavy-duty engine [6]. There is conservative
prediction that if a passenger car is equipped with a 500W TEG, a
there is potential to save more than 2% fuel consumption and hence
CO
2
emission reduction [7]. Since the commercial TEMs have hot
side temperature limit which is 250°C, a by-pass solution was used in
most experimental TEG project for automotive engine especially
gasoline engine which has higher exhaust temperature than that of
diesel engine [8-11]. However, this solution is not ideal as there will
be a lot of thermal energy at high temperature escaped without any
recovery. So that developing high temperature thermoelectric
materials is important for automotive applications. Meanwhile,
segmenting the elements within the TEM using different
thermoelectric materials can further improve the TEG performance
[12]. Future thermoelectric modules are expected to have 10% to
20% efficiency and over 500°C hot side temperature limit. The
possible solution is using segmented thermoelectric elements which
consist of low temperature (Bismush Telluride) , mid-range
(skutterudite) and high-range(half-Heuslers [26]) thermoelectric
materials.
The efficiency of a TEM is determined by the materials that
constitute the TEM. Materials performance is expressed in terms of
the dimensionless figure-of-merit ZT, which is the ratio of electrical
power to heat flux. High performance requires the unusual
combination of a high electrical conductivity (σ), usually associated
with metallic phases, together with a high Seebeck coefficient (S) and
low thermal conductivity (κ), typical in non-metallic systems [13,14].
Materials with the skutterudite crystal structure possess attractive
transport properties and have a good potential to achieve high ZT
values at high temperature [15, 16].
A thermoelectric module can either be run with an applied bias to use
electrical energy to move heat across the module and create a
temperature difference (Peltier mode), or to generate electrical power
from a heat source by creating a temperature gradient across the
module and harvest power from the resulting heat flow (Seebeck
mode). It is this second mode which makes thermoelectric modules
an attractive technology for energy recovery from heat sources such
as automotive exhaust systems. Bass et al demonstrated a 1kW
bismuth telluride system [6] in the mid-1990s and subsequently many
others have followed [17]. Although ring structured thermoelectric
modules to generate power from radial sources have been
Page 2 of 7
10/19/2016
demonstrated in the lab [18] and commercially [19], most modules
are fabricated with a planar design with heat flow from one side of
the plate to another.
The rest of the paper was organized as: The basic theory of TEG was
given in Section 2 which was followed by the introduction of
manufacture of skutterudite material in Section 3. The introduction of
fabrication of TEM made from skutterudite materials was presented
in Section 4. The property loss caused by the module fabrication
procedure was discussed in Section 5. Section 6 is about how the
improved TEG performance using high temperature skutterudite
materials was estimated using a validated TEG model. The summary
and conclusions can be found in Section 7.
Theory of Thermoelectric Generator
A helpful analogy for understanding the workings of the
thermoelectric module is to consider it as a ‘thermal battery’ [23].
The Seebeck voltage (related to the Seebeck coefficient α by the
equation V=αΔT,) may then be considered like an EMF, while the
internal resistance of the module R, similar to the internal resistance
of a battery. When a load RL is placed across the thermoelectric
module, the voltage VL generated across the load is given by;
, (1)
and the current IL by the equation;
. (2)
For a given temperature difference, the peak power will be achieved
when the external load matches the internal resistance of the module
i.e., when RL=R and the maximum power is given by;
. (3)
Module efficiency is an important parameter to consider, especially if
there is limited heat available. The maximum conversion efficiency η
of the module is given by the equation
, (4)
where
hot is the heat flowing through the hot side of the module
and the peak power is measured under load conditions with the
system in thermal equilibrium. When the module is under open
circuit conditions, the heat flow is given by the equation
, (5)
where A and L are the area and length of the thermoelectric legs
respectively. Under load conditions
hot becomes
. (6)
Substituting equation (6) into equation (5) and using the relation
ZT=α2σT/κ, the efficiency can be written in terms of the
dimensionless figure of merit
Δ
. (7)
The efficiency is strongly dependent on a large ΔT and also large
average ZT. Skutterudites are designed to work in temperatures up to
600 ºC which has the potential to be a highly efficient module.
Manufacture of Skutterudite Material
The manufacture of Skutterudite material has been carried out in the
lab of chemistry department in the University of Reading, UK. The
skutterudite materials was produced during the synthetic process and
then was characterized to find out its real three thermoelectric
properties during the characterization stage.
Figure 1 shows the crystal structure of a filled skutterudite. The
skutterudite structure consists of a framework of stoichiometry
M8X24, (M = transition metal, X = pnictogen atom) formed by
corner-sharing MX6 octahedra, and contains larges cages that can be
filled with a heavy filler atom, i.e. rare earth [20-22].
Figure 1. Crystal structure of a filled skutterudite. Transition metals are
represented by yellow spheres, pnictogen atoms by red spheres and the filler
atoms by blue spheres.
Synthetic Procedures
Pure elements of constituent elements were weighted according to the
stoichiometric formula of CoSb
2.75
Sn
0.05
Te
0.20
(n-type material) and
Ce
0.5
Yb
0.5
Fe
3.25
Co
0.75
Sb
12
(p-type material): The appropriate amounts
of the powdered elements were loaded together with 13 Zirconium
dioxide balls into a 250 ml Zirconium dioxide grinding vessel. To
prevent oxidation, the vessel was properly sealed inside an Argon-
filled glove-box. The grinding vessel was loaded into a Fritsch P6
Planetary Ball Mill. Grinding was carried out at 400 rpm for 10 h.
This methodology allows the preparation of > 60 g per batch at the
laboratory scale. Figure 2 shows the Ball Mill device together with
the grinding vessel.
Page 3 of 7
10/19/2016
Figure 2. Ball Mill device and Grinding Vessel
The ball milled powders were consolidated by hot-pressing (90 MPa,
600° C for 30 minutes). The densities of the consolidated discs were
measured using an Archimedes balance ADAM PW184. The
compacted samples have densities above 98 % of the crystallographic
values. The consolidated pellets show a dense microstructure, void-
free and with a relatively uniform particle size (150 – 200 nm). The
pictures of skutterudite material powder and pellet made from this
powder are shown in Figure 3.
Figure 3. (a) Manufactured skutterudite material power; (b) Skutterudite
thermoelectric pellet.
Characterization techniques
X-Ray diffraction
X-ray diffraction (XRD) data for both CoSb
2.75
Sn
0.05
Te
0.20
and
Ce
0.5
Yb
0.5
Fe
3.25
Co
0.75
Sb
12
pellets were collected using a Bruker D8
Advance Powder X-ray diffractometer, operating with Ge
monochromated Cu Kα1 radiation (λ= 1.54046 Å) and fitted with a
LynxEye detector. Data were collected over the angular range 5
2Θ/° 90 for 6 hours.
Figure of merit ZT
Seebeck coefficient (S), electrical (σ) and thermal (κ) conductivities
determine the figure of merit ZT (ZT=S2σT/κ) of a material. A
Linseis LSR-3 instrument was employed to measure simultaneously
Seebeck coefficient and electrical conductivity over the temperature
range 25 ≤ T/°C ≤ 550 with a partial pressure of Helium. A Netzsch
LFA 447 NanoFlash instrument was employed to determine the
thermal conductivity of the samples over a temperature range of 25 ≤
T/°C ≤ 300 in 25 °C steps. Higher Temperature measurements (300 ≤
T/°C ≤ 550) in 50 °C steps were performed using an Anter Flashline
3000 instrument.
Measured thermoelectric properties of the skutterudite
pellet
X-ray diffraction based on Bragg's Law which explains that the
cleavage faces of crystals appear to reflect X-ray beams at certain
angles of incidence. It was used here to verify the crystal structure of
the fabricated skutterudite thermoelectric materials. X-ray diffraction
patterns of the two synthesized materials after consolidation by hot-
pressing present the basic reflections corresponding to the
skutterudite structure and can be indexed in the cubic Im
crystallographic space group. See Figure 4.
Figure 4. X-ray diffraction pattern of the skutterudite samples after
consolidation by hot pressing.
The figure of merit ZT of the n and p thermoelectric materials versus
the applied temperature is illustrated in Figure 5. It can be observed
that the synthesized materials are appropriate for mid - high
temperature range application, with maximum ZT of 1.13 at 405 °C
for the n-type material and 0.93 at 550 °C for the p-type material.
Figure 5. Temperature dependence of the figure of merit, ZT, of the
skutterudites
Page 4 of 7
10/19/2016
Fabrication of Skutterudite Thermoelectric
Module
The characterized pellets manufactured in Reading University then
was fabricated into TEM in Cardiff University in UK.
Module Fabrication
Modules consisting of 9 couples have been fabricated from 13mm
diameter skutterudite pellets which have the chemical composition
Co
1
Sb
2.75
Sn
0.05
Te
0.20
(n-type) and Ce
0.5
Yb
0.5
Fe
3.25
Co
0.75
Sb
12
, (p-type).
The pellets are ground and polished to an equal height of
approximately 1.5 mm and a metal diffusion barrier is then applied to
the top and bottom surfaces via electroplating. This consists of a
palladium layer followed by a nickel layer, with the total thickness of
the layer verified to be approximately 10 μm using a microscope. The
pellets are then cut into legs using an Accutom-100 automated
diamond cutting wheel, with each leg having the approximate
dimensions 1.7 mm x 1.7 mm x 1.5 mm.
Two alumina ceramic plates (16 mm x 13 mm x 0.63mm) with
bonded 2mm x 5mm x 0.35mm copper strips (purchased from
European Thermodynamics Ltd) are used to provide the thermal and
electrical contacts. Junctions between the copper strips and the
thermoelements are formed using the high temperature solder Pb
93.5
Sn
5
Ag
1.5
. The joining materials have a melting point of around
296°C, which restricts the hot-side temperature and is in reality a low
temperature joining material for skutterudites, but is useful for basic
characterization of the materials in a module formation. Lead based
solders wet readily to the surface of the barrier layer and to the
electrode and so a lower contact resistance is guaranteed. Also, the
lower soldering temperatures makes inter-diffusion into the bulk less
likely. All the components (i.e., thermoelements, alumina plates, and
solder pieces) are sandwiched using a homemade assembly holder
made from stainless steel. A good bond is achieved with the flux
paste spread between both the thermoelement/solder and solder/Cu
strip interfaces. The assembled module in the holder is then inserted
into a quartz tube furnace (Carbolite furnaces) and isolated in an inert
argon atmosphere. The holder is held in place until a temperature of
350°C is reached to melt the soldered joints. This ensured any
difference between the temperature at the bond and at the
thermocouple is mitigated. The holder is than removed and allowed
to cool in an argon atmosphere. Finally, two contacting wires are
soldered to the cold side of the module to provide the contacts. Figure
6 is the picture of this skutterudite TEM.
Figure 6. A picture of the skutterudite thermoelectric module.
Module Testing
Measurements are undertaken in a custom-made module testing
vacuum system which is typically evacuated to approximately 5x10-6
mbar at the top of the turbo-pump. A copper block with inserts for
electrical cartridge heaters is attached to the hot-side of the module
while the cold-side is attached to a water cooled aluminum block
which forms part of the vacuum chamber. The hot side heater block
is approximately the same size as the module to minimize any
thermal losses. A clamping plate is used to hold the heater against the
module with 20mm of thermal insulation to minimize thermal losses
between the heater block and the plate. The plate is held in position
by two screws embedded in the cold side and clamped in place using
nuts and spring loaded washers to apply pressure. Both the module
output power and heater input power measurements are taken using
separate voltage and current leads to negate any resistances other than
the module and its leads. Once the system has been evacuated, a
B&K Precision 9183 power supply is used to ramp the heater to a set
temperature and to maintain a constant heat flow through the module.
When the temperature is reached and has been stabilised, fast I-V
curves are taken from open circuit to short circuit conditions using an
Autolab potentiastat PGSTAT302N. The curve is taken with a scan
rate of 5V/s, to ensure that a constant temperature is maintained
throughout the measurement. In all, five modules were fabricated and
tested in this way. A module with characteristics similar to the
average is then measured again to calculate the efficiency of the
module. This measurement takes more time to complete, as it
requires the module to be in thermal equilibrium at each point of the
I-V curve. When current is allowed to flow through the module,
additional heat is moved from the hot side to the cold side due to the
Peltier effect (equation 6). Since the amount of heat supplied to the
module is fixed, the result is a drop in the ΔT across the module
which is dependent on the amount of current flowing within the
module.
Results about the Module
Figure 7 shows the nodule output power for the five modules with the
coldside kept around 25°C. The hotside was kept below the melting
point of the solder as much as possible. Three of the five modules
have a power output above the 1 W mark at ΔT=300K and this
represents a power density of 0.48 Wcm
2
. This high power density is
possibly due to the leg length being quite short, which increases the
power output at the expense of efficiency. This is reflected in the
lower efficiency measured for module 4. The efficiency of module 4
can be found in Figure 8. This module was selected from the five
produced since its performance properties approximately match the
average of the five. An efficiency value of around 3% at ΔT=273 K
was obtained, which is much lower than the value of 7% obtained by
Salvador et al [24]. Their higher efficiency value is due to the much
larger temperature difference across the module (ΔT =460 K). With a
high temperature braze which can maintain a 500°C hotside
temperature, it is expected that the skutterudite materials used in
module 4 will achieve similar or better efficiency values at larger
temperature difference.
The device used for measuring I-V curve has ±0.2% uncertainty for
both voltage and current measurement. Then according to the
uncertainty analysis theory, the uncertainty of the computed electrical
power shown in Figure 7 is about ±0.283%.
Page 5 of 7
10/19/2016
Figure 7. Electrical power output plotted against temperature difference across
the module.
Figure 8. Module efficiency plotted as a function of temperature difference
across the module. One module was chosen test the efficiency and module
number 4 was selected since it represented the average performance properties
of the five modules.
Property Loss via Module Fabrication
Three module thermoelectric properties which are module seebeck
coefficient, module internal electric resistance and module thermal
resistance can be estimated from measured pellet thermoelectric
properties. These values were plotted together with the corresponding
measured properties. They are shown in Figure 9 to Figure 11. Figure
9 shows that the seebeck coefficient was reduced from the module
fabrication process. The reduction percentage increases with the delta
temperature. The higher delta temperature, the higher loss of seebeck
effect. Maximum percentage is about 23%. Figure 10 shows that the
module fabrication process increased the module internal electric
resistance. The higher delta temperature, the higher percentage of the
increased internal resistance. The maximum percentage of increased
internal resistance is around 11%. The module fabrication process has
big impact on the module thermal resistance, See Figure 11. The
reduction percentage is nearly constant at 44% for all delta
temperature conditions.
Figure 9. (a) Estimated and measured module seebeck coefficient; (b)
Percentage of reduced seebeck coefficient.
Figure 10. (a) Estimated and measured module internal electric resistance; (b)
Percentage of increased electric resistance.
Figure 11. (a) Estimated and measured module thermal resistance; (b)
Percentage of reduced thermal resistance.
The change of three thermoelectric properties caused by the module
fabrication process all end up at property loss. The biggest property
loss is thermal resistance. The second is the seebeck coefficient.
Further investigation needs to be carried out to reduce the property
loss of these three thermoelectric properties during the module
fabrication process.
Improved Thermoelectric Generator
Performance Using Skutterudite Modules
A MATLAB function based TEG dynamic model has been validated
using engine test data. The model structure and validation results
were discussed in another paper [25]. Use this model to estimate the
maximum power output at steady state condition for a TEG device.
Two TEG devices were assembled using two types of commercial
modules which are ETL-TEM and HZ-TEM respectively. ETL-TEM
and HZ-TEM are both made from Bismuth Telluride thermoelectric
materials. But they have different leg length, module fill ratio and
number of couples. The corresponding TEG devices were named as
ETL-TEG and HZ-TEG in this paper. The third TEG device was
