solar thermoelectric generators fabricated on a silicon-on-insulator substrate

8/20/2019 Solar Thermoelectric Generators Fabricated on a Silicon-On-Insulator Substrate

1/13

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 130.236.167.210

This content was downloaded on 10/03/2015 at 15:25

Please note that terms and conditions apply.

Solar thermoelectric generators fabricated on a silicon-on-insulator substrate

View the table of contents for this issue, or go to the journal homepage for more

2014 J. Micromech. Microeng. 24 085011

(http://iopscience.iop.org/0960-1317/24/8/085011)

ome Search Collections Journals About Contact us My IOPscience

http://localhost/var/www/apps/conversion/tmp/scratch_7/iopscience.iop.org/page/termshttp://iopscience.iop.org/0960-1317/24/8http://iopscience.iop.org/0960-1317http://iopscience.iop.org/http://iopscience.iop.org/searchhttp://iopscience.iop.org/collectionshttp://iopscience.iop.org/journalshttp://iopscience.iop.org/page/aboutioppublishinghttp://iopscience.iop.org/contacthttp://iopscience.iop.org/myiopsciencehttp://iopscience.iop.org/myiopsciencehttp://iopscience.iop.org/contacthttp://iopscience.iop.org/page/aboutioppublishinghttp://iopscience.iop.org/journalshttp://iopscience.iop.org/collectionshttp://iopscience.iop.org/searchhttp://iopscience.iop.org/http://iopscience.iop.org/0960-1317http://iopscience.iop.org/0960-1317/24/8http://localhost/var/www/apps/conversion/tmp/scratch_7/iopscience.iop.org/page/terms


2/13

1 © 2014 IOP Publishing Ltd Printed in the UK

1. Introduction

A promising way of generating power by alternative means

is through thermoelectric conversion, where electrical

energy is provided by the direct conversion of heat to elec-

tricity with the use of thermoelectric generators (TEGs).

The basic principle behind the operation of a TEG is the

Seebeck effect. The Seebeck effect produces voltage when

a temperature difference is applied between the junctions

of two different materials. For a TEG to supply a signifi-

cant amount of power, several thermocouples are connected

electrically in series and thermally in parallel. It is attractive

to use TEGs because they have no mechanical parts; hence

resulting in an alternative power system that is silent, stable,

Journal of Micromechanics and Microengineering

Solar thermoelectric generators fabricated

on a silicon-on-insulator substrate

Maria Theresa de Leon1,3, Harold Chong1 and Michael Kraft2

1 Nano Research Group, Faculty of Physical Sciences and Engineering, University of Southampton,

Highfield, Southampton, SO17 1BJ, UK2 Faculty of Engineering, University of Duisburg-Essen, 47057 Duisburg, Germany3 Electrical and Electronics Engineering Institute, University of the Philippines, Diliman,

Quezon City 1101, Philippines

E-mail: [email protected]

Received 14 April 2014, revised 27 May 2014

Accepted for publication 16 June 2014

Published 28 July 2014

Abstract

Solar thermal power generation is an attractive electricity generation technology as it is

environment-friendly, has the potential for increased efficiency, and has high reliability.

The design, modelling, and evaluation of solar thermoelectric generators (STEGs) fabricated

on a silicon-on-insulator substrate are presented in this paper. Solar concentration is achieved

by using a focusing lens to concentrate solar input onto the membrane of the STEG. A thermal

model is developed based on energy balance and heat transfer equations using lumped

thermal conductances. This thermal model is shown to be in good agreement with actual

measurement results. For a 1 W laser input with a spot size of 1 mm, a maximum open-circuit

voltage of 3.06 V is obtained, which translates to a temperature difference of 226 °C across the

thermoelements and delivers 25 µW of output power under matched load conditions. Based on

solar simulator measurements, a maximum TEG voltage of 803 mV was achieved by using a

50.8 mm diameter plano-convex lens to focus solar input to a TEG with a length of 1000 µm,

width of 15 µm, membrane diameter of 3 mm, and 114 thermocouples. This translates to

a temperature difference of 18 °C across the thermoelements and an output power under

matched load conditions of 431 nW.

This paper demonstrates that by utilizing a solar concentrator to focus solar radiation

onto the hot junction of a TEG, the temperature difference across the device is increased;

subsequently improving the TEG’s efficiency. By using materials that are compatible with

standard CMOS and MEMS processes, integration of solar-driven TEGs with on-chip

electronics is seen to be a viable way of solar energy harvesting where the resulting microscale

system is envisioned to have promising applications in on-board power sources, sensor

networks, and autonomous microsystems.

Keywords: solar thermoelectric generators, silicon-on-insulator, thermal modelling,

solar simulator

(Some figures may appear in colour only in the online journal)

0960-1317/14/085011+12$33.00

doi:10.1088/0960-1317/24/8/085011J. Micromech. M icroeng. 24 (2014) 085011 (12pp)

mailto:[email protected]://dx.doi.org/10.1088/0960-1317/24/8/085011http://dx.doi.org/10.1088/0960-1317/24/8/085011mailto:[email protected]


3/13

M T de Leon et al

2

reliable, environment-friendly, and possess virtually unlim-

ited lifetime [1, 2].

A typical thermoelectric generator exhibits only 5–10%

conversion efficiency depending on the materials used and the

temperature difference involved [3]. Meanwhile, the best solar

cell is 3–5 times more efficient than thermoelectric devices

[4]. Several implementations of TEGs focus on improving

its efficiency by exploring advanced thermoelectric mate-

rials such as skutterudites [5, 6], clathrates [7], Zn-Sb alloys

[8], Pb-Te alloys [9–11], InGaN alloys [12], and ZnO alloys

[13, 14]. Higher efficiency TEGs have also been designed by

using segmented thermoelectric legs to exploit the operating

temperatures of several materials; thereby optimizing heat

flow across the thermoelements [15–18]. Meanwhile, other

researchers emphasized nanostructuring of bulk materials to

improve the material’s thermoelectric figure merit [19–21]. In

nanostructuring, the material’s figure of merit is increased by

creating materials composed of nanosized grains. By doing

so, the thermal conductivity of the material is decreased

while maintaining its electrical conductivity. The technique ofnanostructuring has been applied to silicon [22–28], silicon

germanium [29, 30], bismuth telluride alloys [31–34], and

complex cobalt oxides [35].

The above-mentioned techniques focus on improving the

thermoelectric properties of the materials to improve the effi-

ciency of the TEG. While results obtained by these techniques

are promising, synthesizing novel compounds, fabricating

segmented thermoelements, and creating nanostructured

materials are quite complex. Another aspect that can be

explored to improve the efficiency of a TEG is by increasing

the temperature difference across the thermoelements [36].

An increase in temperature difference can be realized by using

a high input heat flux such as that coming from the sun [5].

Direct solar thermal power generation is an attractive elec-

tricity generation technology since it can achieve a flexible

power generation scheme that is environment-friendly, has

high efficiency, and has high reliability characteristics [1].

STEGs are also scalable, making it suitable for both small-

and large-scale applications [37]. Moreover, photovoltaics

are limited to the fraction of incident solar radiation above

the bandgap while STEGs utilize a larger portion of the solar

spectrum. In this regard, a solar concentrator can be used

to concentrate solar radiation onto the hot side of the TEG.

Several researches have demonstrated the functionality of

such systems on a large scale by using commercially-availablesolar concentrators and TEG modules [38, 39].

At chip scale, the use of a lens to concentrate light onto

a TEG that serves as power supply to a microactuator has

already been proposed [40]. More recently, an improvement

in TEG efficiency has been achieved by employing both

solar and thermal concentration on a flat-panel solar ther-

moelectric generator composed of a pair of n- and p-type

thermoelectric materials based on nanostructured Bi2Te3

alloys [41]. Another study where a cylindrical lens was used

to focus solar light onto TEGs with p-type Bi0.5Sb1.5Te3 and

n-type Bi2Te2.7Se0.3 thermoelements reported an efficiency of

8.75 × 10−4

% [42]. Despite having the best thermoelectricfigure of merit, the use of Bi2Te3 alloys in MEMS systems is

hindered by challenges in technological compatibility [43].

It is therefore more practical to use materials like silicon or

polysilicon as thermoelectric materials as they have better

compatibility with standard CMOS and MEMS processes.

Thus, it is worthwhile to investigate the feasibility of imple-

menting solar thermoelectric generator (STEG) systems

utilizing conventional materials in MEMS and CMOS pro-

cessing and characterize its improvement in efficiency as this

gives way to future advancements in solar energy harvesting.

It is important to choose materials and processes that will

enable easy integration of the solar-driven TEG with on-chip

electronics as this microscale system is envisioned to have

promising applications in on-board power sources, sensor

networks, and autonomous microsystems.

This paper presents the design, modelling, fabrication, and

evaluation of STEGs fabricated on a silicon-on-insulator (SOI)

substrate. Section 2 presents the design of the STEG, along

with the thermal model developed to predict the performance

of the device based on its geometry and input conditions. The

thermal model is developed based on energy balance and heattransfer equations using lumped thermal conductances. Section

3 provides details on the fabrication process of the STEG.

Section 4 presents the results of the measurements performed

on fabricated test structures to determine the electrical and

thermoelectric properties of the SOI’s device layer, which is

utilized as one of the thermoelement materials. The fabricated

STEGs are also tested using a laser set-up with a modulated

input power to compare the actual performance of the fabricated

STEGs to that of three-dimensional heat transfer simulations

and the developed thermal model. Section 5 gives the results

of measurements using a solar simulator set-up where lenses of

different diameters are used to concentrate solar input onto the

membrane of the STEG. Lastly, conclusions drawn from the

results presented in this paper are given in section 6.

2. STEG design and thermal modelling

Figure 1 shows an illustration of the STEG design investigated

in this work. The device is formed on an SOI substrate with

a 3 µm thick heavily boron-doped device layer. This p-type

silicon device layer is utilized as one of the thermoelement

materials. To simplify the fabrication process, aluminum is

used as the second thermoelement material. The membrane

acts as the heat absorber while the rim acts as the heat sink.The thermoelements are oriented radially around the circular

membrane to insure optimum transfer of heat from the center

of the membrane to the tip of the thermoelements. Isolation

trenches are etched on both sides of the thermoelements to

electrically insulate the thermoelements from the membrane

and the rim. Although etching the buried oxide and handle

layers directly below the membrane and thermoelements

would result in an optimum heat flux path across the thermo-

elements, the isolation trenches would have to be refilled with

a suitable material that will make the device mechanically

stable. Hence, the STEG design retains the buried oxide layer

and a thin part of the handle layer under the membrane and thethermoelements for structural stability.

J. Micromech. Microeng. 24 (2014) 085011


4/13

M T de Leon et al

3

After settling on the STEG design, design parameters that

can be investigated in this work are identified. The design

space essentially includes the length and width of the ther-

moelements, and the diameter of the membrane. Higher

efficiencies can be achieved by using STEGs with longer

lengths and narrower widths. However, the mechanical sta-

bility of the STEG must also be considered. It is desirable for

the lengths of the thermoelements to be long so as to achievea larger temperature difference across the device. However,

longer thermoelements can be mechanically unstable once the

handle layer is thinned. As such, the lengths are varied from

200 µm to 1 mm. With regards to the thermoelement width,

narrower thermoelements are ideal for a larger open-circuit

voltage across the device. However, a tradeoff in mechan-

ical stability also exists with narrow thermoelements. In this

regard, the thermoelement widths are varied from 15 µm to

30 µm. The membrane diameter also plays a crucial part in

the overall performance of the STEG. It is desirable to have

a smaller membrane diameter so as to have a higher tempera-

ture difference across the device for a specific amount of solar

input. The area of the suspended membrane must also be kept

small to achieve good mechanical stability [44]. However, it is

more difficult to focus solar light onto a device with a smaller

membrane. Hence, the membrane diameter is varied from

1 mm to 5 mm. Each STEG chip is also set to have a dimen-

sion of 1 × 1 cm2.

Suppose the sun uniformly irradiates an energy density qs

onto the lens, then the heat power density qh of the incoming

heat flux to the TEG membrane is given by:

=q qh slens mem (1)

where γ is the concentration factor, τ lens is the lens trans-

mittance, and α mem is the membrane absorptance. Theconcentration factor is proportional to the ratio of the effective

lens diameter to the spot size diameter on the membrane as

given by:

=

d

d

lens,eff

spot

2

(2)

With this approach, an input heat flux in the order of hun-

dreds of kW m−2 can be generated. Based on the general heat

transfer equation [45], an increase in the input heat flux would

translate to a corresponding increase in the temperature dif-

ference between the temperature at the absorber side of the

thermoelements (T H) and the temperature at the rim side of the

thermoelements (T C). This results in an open-circuit output

voltage, V TEG, described by:

= − − = − =V N T T S T T S T ( ) ( ) ( )TEG Si Al H C TEG H C TEG (3)

where N is the number of thermocouples and αSi and αAl are

the Seebeck coefficients of p-type silicon and aluminum,

respectively. S TEG is the total Seebeck coefficient of the TEG.

If a load resistance RLOAD is attached to the output of the TEG,

the maximum power delivered to the load is:

= =P

V

R

S T

R4

( )

4OUT

TEG2

TEG

TEG2

TEG

(4)

where RTEG is the electrical resistance of the TEG and

= R RTEG LOAD

Figure 2(a) shows the thermal equivalent model of the

STEG developed in this study based on energy balance and

heat transfer equations using lumped thermal conductances.

QIN refers to the input power to the TEG, which is equiva-

lent to the product of the heat power density of the incoming

heat flux (qh) and the surface area of the solar spot size on

the membrane. QMEM refers to the rate of heat flow from themembrane to the TEG while QRIM refers to the rate of heat

Figure 1. (a) STEG design using an SOI substrate with p-type polysilicon and aluminum as thermoelement materials, with (b) top viewshowing the placement of isolation trenches and (c) cross-section view showing the thinned SOI handle layer for improved heat flux path.

J. Micromech. M icroeng. 24 (2014) 085011


5/13

M T de Leon et al

4

flow from the TEG to the rim. The corresponding Thevenin

equivalent circuit is also shown in figure 2(b). The thermal

equivalent model also includes heat contributions due to

Peltier effect (S TEGT C I+ and S TEGT H I ) and Joule heating

(POUT /2) in the generator. The current I is the electric current

flowing through the thermocouples. K TEG is the total thermal

conductance of the TEG, K MEM is the thermal conductance

between the thermocouples and the membrane, K RIM is the

thermal conductance between the thermocouples and the rim,

K BOX is the thermal conductance of the portion of the buried

oxide layer directly below the heated area of the device, and

K HAN is the thermal conductance of the portion of the handle

layer that is also directly below the heated area of the device.

The temperature node, T HAN, is the temperature at the bottom

of the thinned area of the handle layer. QCONV’s and QRAD’s

in the thermal model represent heat losses due to convection

and radiation, respectively, at different areas of the device.

These heat losses are dependent on the geometry of the device

and on the temperature, as given by the following general

expressions:

= −Q h A T T ( ) X X CONV, conv AMB (5)

= −( )Q A T T X X RAD, 4 AMB4 (6)

where the subscript A X refers to the surface area of the specific

part of the STEG structure (MEM, RIM, TEG, or HAN) and T

refers to the corresponding temperature node in figure 2(a). ε

is the surface emissivity; σ is the Stefan–Boltzmann constant;

and hconv is the convective heat transfer coefficient. Free con-

vection in air would have hconv values between 5–50 W m−2K−1

while forced convection in air would have hconv values between

25–250 W m−2K [46].

Referring to each temperature node in figure 2(a), the heat

balance equations can be derived as:

= + +

+ + −

T Q Q Q Q

K K T T

at :

( ) ( )

M M 1 IN MEM RAD, CONV,

BOX HAN 1 HAN

(7)

= − +

− +

T Q K T T S T I

P Q

at : ( )

1

2

H MEM TEG H C TEG H

OUT CONV,TH (8)

= − +

+ −

T Q K T T S T I

P Q

at : ( )

1

2

C RIM TEG H C TEG C

OUT CONV,TC (9)

= +T Q Q Qat : R R2 RIM RAD, CONV, (10)

= + −

T Q K K T T at : ( ) ( )HAN CONV, HAN BOX HAN 1 HAN (11)Considering heat flow through K MEM and K RIM, we derive

equations for T H and T C in terms of T 1 and T 2, respectively.

= −T T Q

K H 1

MEM

MEM (12)

= +T T Q

K C 2

RIM

RIM (13)

Incorporating equations (5), (6), (12) and (13) into the heat

balance equations in equations (7)–(11), the following expres-

sions can be derived:

= + − + −

+ + −

( )Q Q A T T h A T T K K T T

( )

( ) ( )

IN MEM MEM 1 4 AMB4 conv MEM 1 AMB

BOX HAN 1 HAN

(14)

= + − −

+ − −

Q K T S T Q

K I P

h A T Q

K T

1

2MEM TEG TEG 1

MEM

MEM

OUT

conv TEG 1

MEM

MEM

AMB

(15)

= + + +

− + −

Q K T S T Q

K I P

h A T Q

K T

1

2RIM TEG TEG 2

RIM

RIM

OUT

conv TEG 2

RIM

RIM

AMB

(16)

Figure 2. (a) Thermal equivalent model and (b) Thevenin equivalent circuit of the STEG.



6/13

M T de Leon et al

5

= − + −( )Q A T T h A T T ( )RIM RIM 24 AMB4 conv RIM 2 AMB (17)

− = + −h A T T K K T T ( ) ( ) ( )conv HAN HAN AMB BOX HAN 1 HAN (18)

From the previous equation, we can derive an expression

for T HAN in terms of T 1:

=

+ +

+ +T

K K T h A T

h A K K

( )HAN

BOX HAN 1 conv HAN AMB

conv HAN BOX HAN (19)

Substituting T HAN in equation (14) with equation (19) gives:

= + − + −

+ +

× −+ +

+ +

( )Q Q A T T h A T T

K K

T K K T h A T

h A K K

( )

( )

( )

IN MEM MEM 14

AMB4

conv MEM 1 AMB

BOX HAN

1BOX HAN 1 conv HAN AMB

conv HAN BOX HAN

(20)

Referring back to the thermal equivalent circuit in figure 2(a),

we can express the temperature difference between T 1 and T 2 as:

− = + +T T

Q

K T

Q

K 1 2

MEM

MEM

RIM

RIM (21)

In equations (15) and (16), the current I and output power

POUT can be expressed in terms of ΔT using equations (3) and

(4), respectively. Hence, combining equations (15) and (20)

gives a fourth-order polynomial in T 1 with only T 1 and ΔT as

unknown variables.

+ + = A T C T C 0M EM 14

3 1 4(22)

where

= + +

+ + −

+

+ +

C S I h A A

K K

K K

h A K K

( )

( ) 1

( )

3 TEG conv TEG MEM

BOX HAN

BOX HAN

conv HAN BOX HAN

(23)

and

= − − −

−+

+ +

− + + −

C K T S Q

K I P A T

K K h A T

h A K K

h Q

K A A A T Q

1

2

( )

( )

4 TEG TEGMEM

MEM

OUT MEM AMB4

BOX HAN conv HAN AMB

conv HAN BOX HAN

convMEM

MEM

TEG TEG MEM AMB IN

(24)

By using Ferrari’s solution to a quartic function [47] in

solving the roots of equation (22), four expressions for T 1 as functions of ΔT can be derived. By plugging in a positive

value for ΔT and calculating the roots, the expression that

gives a real and positive value for T 1 is selected. Similarly,

equations (16) and (17) can be combined to form the fol-

lowing fourth-order polynomial equation:

+ + − + = A T h A A S I T C ( ( ) ) 0RIM 24

conv TEG RIM TEG 2 2 (25)

where

= − − − −

+ − +

C K T S Q

K I P A T

h

Q

K A A A T

1

2

( )

2 TEG TEGRIM

RIM

OUT RIM AMB4

conv

RIM

RIMTEG RIM TEG AMB

(26)

The roots of equation (25) can then be solved and T 2 can

be expressed in terms of ΔT . The derived equations for T 1 and

T 2 can then be substituted into equation (21), which gives an

equation with only ΔT as the unknown variable. The tempera-

ture difference across the thermoelements, ΔT , can then be

solved numerically using Matlab.

Of the model parameters, the temperature difference has the

most sensitivity to the input powerQIN, which is proportional to

both the concentration factor and the membrane absorptance.

The temperature difference increases linearly with the input

power at a rate of 272 °C W−1. The convective heat flux, on

the other hand, has an inverse effect on the temperature dif-

ference. When the convective heat flux is increased, both the

hot and cold side temperatures decreases, which results in a

decrease in temperature difference of up to 30 °C when the

convective heat flux is increased to 250 W m−2K−1. Increasing

the surface emissivity also results in a decrease in both the hot

and cold side temperatures. However, since both temperatures

decrease at about the same rate, there is no significant change

in the temperature difference when the surface emissivity isincreased. Based on the thermal model, heat transfer due to

conduction contributes to 65.6% of the total heat flow whereas

heat transfer due to convection and radiation contribute 34%

and 0.35%, respectively.

3. Device fabrication

The fabrication process for the STEG implemented on a SOI

substrate is illustrated in figure 3. The fabrication process

starts with a 6 inch SOI wafer from Ultrasil Corporation with

the following thicknesses: 500 ± 15 µm handle layer, 400 nm

buried oxide layer, and 3 ± 0.5 µm device layer. The handlelayer is boron-doped with a resistivity of 1–30 Ω cm while

the device layer is also boron-doped and has a resistivity of

0.005–0.02 Ω cm. Material properties for the SOI wafer were

chosen with special consideration given to the heavily-doped

device layer. Heavily-doped silicon is seen to be a viable

choice for thermoelement material since it has high Seebeck

coefficients at doping levels between 3.5 × 1019 cm−3 to

1.6 × 1020 cm−3 [48, 49].

The first step in the fabrication process is to deposit sil-

icon dioxide by plasma enhanced chemical vapor deposition

(PECVD) on both sides of the SOI wafer, which will act as

hardmasks. For the front side, 500 nm thick SiO2 was depos-ited whereas 3.6 µm SiO2 was deposited at the back. The gas

flow rates of silane-based PECVD SiO2 were 4.2 sccm SiH4,

350 sccm N2O, and 80 sccm N2. The deposition was per-

formed at a table temperature of 350° C, chamber pressure of

1000 mTorr, and RF power of 20 W. The deposition rate was

about 1 nm s−1. After coating both front and back sides with

PECVD SiO2, lithography of 6 µm thick AZ® 9260 photoresist

(PR) was performed to define the frontside hardmask. After

lithography, the wafer was placed inside the chamber of an

OIPT SYS380 inductively-coupled plasma (ICP) etcher for

SiO2 etching. The gases used for this purpose were 37.4 sccm

CHF3, 34 sccm C4F8, and 8.5 sccm O2. Etching was done at a

table temperature of 15° C, chamber pressure of 7 mTorr, RF



7/13

M T de Leon et al

6

power of 100 W, and ICP power of 1500 W. The same lithog-

raphy procedure was also performed to define the backside

hardmask. The PECVD SiO2 at the backside was etched

using the Plasmatherm Versaline Deep Silicon Etcher as it

was observed that it etches SiO2 more uniformly than the ICP

etcher in cases where larger areas of SiO2 are exposed. The

etching parameters used for this purpose were as follows: 50

sccm CF4, chamber pressure of 5 mT, high-frequency bias of

100 W, and ICP power of 400 W. The SiO2 etch rate was about

180 nm min−1 with a selectivity to AZ® 9260 photoresist of 3:4.

Figure 3. STEG fabrication process.

(a)Deposit PECVD SiO2 on both

sides of SOI substrate.

(b)Deposit, pattern, and develop

PR at frontside.

(c)

Etch PECVD SiO2 to form

frontside hardmask. Strip PR.

(d)

Deposit, pattern, and develop

PR at backside.

(e)

Etch PECVD SiO2 to form

backside hardmask. Strip PR.

(f)Etch p-type Si device layer.

Strip frontside SiO2 hardmask.

(g)Deposit, pattern, and develop

PR. Deposit Al.

(h) Lift-off Al.

(i) Etch Si handle layer.



8/13

M T de Leon et al

7

Once the front and back hardmasks were patterned, the

exposed silicon device layer was etched up to the buried oxidelayer using an OIPT reactive ion etching (RIE) tool. The RIE

conditions were: 18 sccm SF6, 22 sccm O2, table tempera-

ture of 20° C, chamber pressure of 30 mTorr, and RF power

of 100 W. Under these settings, the silicon etch rate was

280 nm min−1 whereas the SiO2 etch rate was 33 nm min−1,

making it an appropriate masking material with a selectivity

of about 8.5. Then, the remaining SiO2 mask at the front-

side was stripped in 7:1 HF solution for 2 minutes. An AZ®

nLOFTM 2070 negative photoresist mask was then patterned

to define the aluminum thermoelements and bonding pads.

After which, 3 µm thick aluminum was e-beam deposited

at a rate of 0.5 nm s−1. Next, the wafer was submerged in a

beaker with N-Methyl-2-pyrrolidone (NMP) solvent, whichacts as the lift-off medium. After depositing aluminum, back-

side deep reactive ion etching (DRIE) using the Plasmatherm

Versaline Deep Silicon Etcher is performed until the silicon

handle layer is thinned to about 5 µm. Once the handle layer of

the STEG chip was suitably thinned, it was bonded to a chip

carrier using a Delvotek wirebonder. An optical micrograph

of a STEG device with thermoelement length of 200 µm, ther-

moelement width of 15 µm, absorber diameter of 1 mm, and

with 31 thermocouples is shown in figure 4.

4. Device characterization

To properly characterize the device, the electrical and thermal

properties of the silicon device layer were first determined

experimentally. Several structures based on the characteriza-

tion study in [50] for polysilicon films used as thermoelectric

material in a CMOS-MEMS thermoelectric power generator

were implemented. The Van der Pauw structure is used to

determine its electrical resistivity; the planar structure is used

to determine its Seebeck coefficient; and the cantilever struc-

ture is used to determine its thermal conductivity. The results

of the tests performed on the fabricated test structures are

listed in table 1, along with the electrical and thermal proper-

ties of aluminum obtained from literature.

To characterize the fabricated TEGs, they are first tested

using a laser set-up where the input power is varied at a con-stant spot size. This depicts a scenario where there is precise

control of the solar spot size and the variation in the laser’s

input power represents varying the concentration ratio. In this

set-up, the laser source is a 488 nm Innova™ 300C FreD™

ion laser source from Coherent, Inc. The laser power was

modulated by using a neutral density filter (NDF) wheel. At

each change in input power, the corresponding output voltage

was measured with a digital volt meter. An infrared thermom-

eter with an emissivity setting at 0.6 was also used to measure

the temperature at the cold side of the TEG for each iteration

of the input power.

To compare the actual measurements from the laser set-up

to those derived from the thermal model, the parameters listed

in table 1 are used in the thermal model, along with the fol-

lowing parameters: αmem = 0.63, ε = 0.6, hconv = 250 W m−2K−1,

and T AMB = 20 °C. Figure 5 shows a comparison between the

simulated, thermal model, and measured parameters. The sim-

ulated plots are based on 3-dimensional finite-element heat

transfer simulations performed in COMSOL. To illustrate the

simulations performed in COMSOL, figure 6 shows the tem-

perature profile of a TEG device for an input power of 1 W.

From figure 5(a), the hot and cold side temperatures derived

from the thermal model is within 5.9% and 1.2%, respectively,

of those obtained from measurements. The simulated cold side

temperature, however, deviates from the measurements and the

Table 1. Electrical and thermal properties of p-type silicon device

layer and aluminum.Seebeckcoefficient(µV K−1)

Thermalconductivity(W mK−1)

Electricalresistivity(Ω m)

P-type silicon 397 146 8.94 × 10−5

Aluminum −1.8a 237 b 2.65 × 10−8 c

a [51].b [52].c [53].

Figure 4. Optical micrograph of STEG with thermolength of 200 µm, thermowidth of 15 µm, absorber diameter of 1 mm, and with31 thermocouples.



9/13

M T de Leon et al

8

thermal model as the input power increases. This results in the

simulated temperature difference ΔT being up to 30 °C lower

than the thermal model and the measured values as is evidentfrom the temperature difference graphs in figure 5(b). The

open-circuit TEG voltage shown in figure 5(c) and the output

power under matched load conditions shown in figure 5(d )

also show good agreement between the simulated, thermal

model, and measured values. Note that due to the deviation

of the simulated temperature difference at higher input power,

the simulated values of the open-circuit voltage and output

power are also lower than the thermal model and measured

data at higher input power levels. Based on these graphs, we

can infer that the thermal model developed can reasonably

predict the performance of the actual device. The tempera-

ture difference, open circuit voltage, and output power derivedfrom the thermal model deviates from their measured values

by ±19 °C, ±0.258 V, and ±1.35 µW, respectively. For a 1 W

laser input with a spot size of 1 mm, the open circuit voltage is

3.06 V, which translates to a temperature difference of 226 °C

across the thermoelements and delivers 25 µW under matched

load conditions. This gives a conversion efficiency of 0.026%,

which is nine times better than when the input power is 1 mW

corresponding to no solar concentration.

5. Solar simulator test results and discussion

The solar to electric power generation properties ofthe fabricated TEGs were evaluated using an Abet

Technologies Sun 3000 Solar Simulator Model 11016A.

Collimated light of 100 mW cm−2 was irradiated onto

spherical convex lenses with varying diameters to emu-

late varying concentration ratios. The test set-up with the

solar simulator is shown in figure 7. A lens was used to

focus the solar input onto the center of the device. The

voltage and current measurements were monitored using a

Keithley Model 2400 SourceMeter® and data was captured

through LabView. Although the solar simulator set-up did

not have enough clearance to allow measurement of the

cold side temperature using an infrared thermometer, theconversion efficiency can be estimated by assuming that

Figure 5. Comparison between simulations, thermal model, and measured parameters: (a) hot and cold side temperature, (b) temperaturedifference, (c) open-circuit TEG voltage, and (d ) matched output power for a TEG implemented on an SOI substrate with thermolength of500 µm, thermowidth of 5 µm, membrane diameter of 1 mm and 34 thermocouples.

Figure 6. Temperature profile of TEG with input power of 1 Wobtained from COMSOL heat transfer simulations.



10/13

M T de Leon et al

9

the cold side temperature is fixed at 30 °C, close to room

temperature based on the laser characterization initially

performed.

The V-I characteristics of a solar TEG with varying lens

diameters used in focusing solar light onto the center of the

device are displayed in figure 8. The x-intercept of the line

defines the open-circuit voltage. The open circuit voltageincreases as the lens diameter increases. This implies an

increase in the input heat flux as the lens diameter is increased.

The slope of the line also slightly decreases as the amount of

solar input is increased, translating to a slight increase in the

TEG’s electrical resistance. The electrical resistance changes

from 355.2 k Ω when using a 12.7 mm lens, to 360 k Ω when

using a 30 mm lens, and to 374.3 k Ω when using a 50.8 mm

lens. This increase in TEG resistance can be attributed to anincrease in the silicon’s electrical resistivity brought about

Figure 7. Measurement set-up with a solar simulator providing solar input and a Keithley Model 2400 SourceMeter® sensing the TEG’soutput.

Figure 8. V-I characteristics of TEG implemented on a silicon-on-insulator substrate with thermolength of 1000 µm, thermowidth of15 µm, membrane diameter of 3 mm, and 114 thermocouples using three different lens diameters.



11/13

M T de Leon et al

10

by the increase in temperature as the amount of solar input

increases.

Table 2 lists the temperature difference, open-circuit

voltage, and output power under matched load conditions of

six fabricated TEGs using two lenses with diameters of 30 mm

and 50.8 mm. Each parameter value is the result of the average

of five measurements. No measurements were taken using the

50.8 mm lens for TEGs with 1 mm membrane diameters as the

resulting spot size is larger than the membrane diameter. For

each TEG, the open-circuit voltage increases as the diameter

of the lens used is increased. This is due to the increase in

temperature difference as the input heat flux increases with

the increasing lens diameter. As expected, the open-circuit

voltage and output power increases as the thermoelement

length increases. This is attributed to the increase in tempera-

ture difference across the device brought about by the decrease

in thermal conductance. As a result of the increase in open-

circuit TEG voltage, the output power under matched load

conditions also increases with increasing lens diameter. As

the thermoelement width is increased, the voltage decreaseswhile the output power increases. The decrease in voltage is

consistent with the increase in thermal conductance brought

about by increasing the thermoelement width. The increase in

output power is mainly due to the decrease in the TEG’s series

resistance as the thermoelement width is increased. Due to the

higher number of thermocouples, the temperature difference

decreases as the membrane diameter is increased. Since the

output voltage is proportional to both the number of thermo-

couples and the temperature difference, it can be inferred that

the increase in the number of thermocouples dominates as is

evident in the increase in the open-circuit TEG voltage as the

membrane diameter is increased. This increase in the open-

circuit TEG voltage results in a corresponding increase in the

output power as the number of thermocouples is increased

with increasing membrane diameter. As for the estimated

conversion efficiency, it is observed that the conversion effi-

ciency improves as the lens diameter is increased. The best

conversion efficiency obtained with the solar simulator set-up

is 0.0020% when a 50.8 mm diameter lens is used on a STEG

with a length of 1 mm, width of 15 µm, and membrane diam-

eter of 3 mm.

6. Conclusions

This paper demonstrated the design, modelling, and evalu-

ation of solar thermoelectric generators fabricated on a SOI

substrate with a heavily boron-doped device layer. The device

layer is used as one of the thermoelements and for ease of fab-

rication, aluminum was utilized as the second thermoelement.

A thermal model was developed based on energy balance and

heat transfer equations using lumped thermal conductances.

Several test structures have also been fabricated to allow

characterization of the SOI device layer. Based on the tests

performed on these test structures, it was determined that the

Seebeck coefficient is 397 µV K−1, the thermal conductivity is

146 W mK−1

, and the electrical resistivity is 8.94 × 10−5

Ω m.These values are consistent with the properties of silicon T

a b

l e

2 .

T e m p e r a t u r e d i f f e r e n c e , o p e n - c i r c u i t v o l t a g e , m a t c h e d o u t p u t p o w e r , a n d e s t i m a t e d c o n v e r s i o n e f fi c i e n c y o f T E G s f a b r

i c a t e d o n a n S O I s u b s t r a t e .

L e

n g t h

( µ m )

W i d t h

( µ m )

d m e m

( m m )

N

T ( ° C )

V T E G

( m V )

P O U T ( n W )

E f fi c i e n c y ( % )

L e n s A

L e n s B

L e n s A

L e n s B

L e n s A

L e n s B

L e n s A

L e n

s B

2 0

0

1 5

1

3 1

2 . 1 6 ± 0 . 0 3

—

2 6 . 7 ± 0 . 3 6

—

3 . 3 2 ± 0 . 1 0

—

0 . 0 0 0 2 5 ± 3 . 3 8 × 1

0 − 6

—

5 0

0

1 5

1

3 4

1 7 . 2 ± 0 . 0 9

—

2 3 2 . 7 ± 1 . 2 7

—

1 4 0 . 9 ± 1 . 8 7

—

0 . 0 0 2 ± 1 . 0 8 × 1

0 − 5

—

5 0

0

1 5

3

1 1 1

9 . 9 ± 0 . 0 1

1 3 . 1 ± 0 . 0 3

4 3 7 . 1 ± 0 . 5 4

5 8 1 . 2 ± 1 . 3 2

1 8 7 ± 0 . 8 2

3 2 8 . 4

± 1 . 4 2

0 . 0 0 1 1 ± 1 . 4 1 × 1

0 − 6

0 . 0 0 1 5 ± 3 . 4 3 × 1

0 − 6

5 0

0

3 0

3

8 1

1 1 . 5 ± 0 . 0 5

1 4 ± 0 . 1 4

3 7 2 . 4 ± 1 . 7 0

4 5 1 . 6 ± 4 . 5 0

3 3 1 . 8 ± 2 . 1 1

4 6 5 . 1

± 9 . 7 0

0 . 0 0 1 3 ± 6 . 0 4 × 1

0 − 6

0 . 0 0 1 6 ± 1 . 6 0 × 1

0 − 5

1 0

0 0

1 5

3

1 1 4

1 2 ± 0 . 0 5

1 7 . 7 ± 0 . 4 2

5 4 6 . 1 ± 2 . 3 6

8 0 3 . 2 ± 1

8 . 9

2 0 7 ± 1 . 3 2

4 3 0 . 9

± 1

4 . 4

0 . 0 0 1 4 ± 5 . 9 8 × 1

0 − 6

0 . 0 0 2 0 ± 4 . 7 8 × 1

0 − 5

2 0

0

1 5

5

1 8 8

1 . 3 2 ± 0 . 0 1

2 . 9 ± 0 . 0 6

9 9 . 1 ± 0 . 6 6

2 1 7 ± 4 . 4 6

8 . 5 ± 0 . 0 1

3 9 . 9

± 1 . 3 0

0 . 0 0 0 1 5 ± 1 . 0 2 × 1

0 − 6

0 . 0 0 0 3 3 ± 6 . 8 4 × 1

0 − 6



12/13

M T de Leon et al

11

having a doping concentration of about 1 × 1019 cm−3. To

properly characterize the fabricated TEG, a laser test set-up

was employed where the input power is varied at a constant

1 mm diameter spot size. Measurement results are in good

agreement with the thermal model developed in this study. A

maximum TEG voltage per watt of input power was generated

for a TEG with a length of 500 µm, width of 15 µm, membrane

diameter of 1 mm, and 34 thermocouples. For a 1 W laser input

with a spot size of 1 mm, the open circuit voltage is 3.06 V,

which translates to a temperature difference of 226 °C across

the thermoelements and delivers 25 µW of output power under

matched load conditions. To evaluate the solar to electric gen-

eration properties of the fabricated TEGs, a solar simulator

test set-up was employed where lenses having different diam-

eters are used to concentrate solar light onto the membrane

of the TEG. Based on the solar simulator measurements, a

maximum TEG voltage of 803 mV was achieved by using a

50.8 mm diameter plano-convex lens to focus solar input to

a TEG with a length of 1000 µm, width of 15 µm, membrane

diameter of 3 mm, and 114 thermocouples. This translatesto a temperature difference of 18 °C across the thermoele-

ments and an output power under matched load conditions

of 431 nW. The temperature difference obtained in the solar

simulator set-up is lower than in the laser set-up primarily due

to the varying intensity of the spot size. In the solar simulator

set-up, light rays strike the lens from several angles, which

results in a spot size having an intensity that is maximum at

the center and gradually decreasing as you go further from

the center. In fact, the spot size can be larger than the mem-

brane diameter, which results in thermal losses and a lower

temperature difference. One way to resolve this issue is to use

lenses with shorter focal lengths. Although this would solve

the problem with the spot size, it should be noted that this

would mean positioning the lens more closely to the STEG. In

this case, it would be practical to either use translation stages

for both the lens and the STEG to allow for better alignment

and precise distance control.

Acknowledgements

M T de Leon is supported by the Engineering Research and

Development for Technology program of the Department of

Science and Technology, Philippines and the University of the

Philippines. We would like to thank Sakellaris Mailis and Gre-gorio Martinez for their assistance with the laser test set-up.

References

[1] Deng Y and Liu J 2009 Recent advances in direct solarthermal power generation J. Renew. Sust. Energy 1 052701

[2]

Strasser M, Aigner R, Lauterbach C, Sturm T, Franosch Mand Wachutka G 2004 Micromachined CMOSthermoelectric generators as on-chip power supply Sensors Actuators A 114 362–70

[3]

Matsubara K and Matsuura M 2006 Thermoelectric Applications to Vehicles in Thermoelectric Handbook:

Macro to Nano ed D Rowe (Boca Raton, FL: CRC Press/ Taylor and Francis)

[4]

Savage N 2011 Photon recycling breaks solar power record IEEE Spectr. 48 16

[5]

Chen G, Kraemer D, Muto A, McEnaney K, Feng H, Liu W,Zhang Q, Yu B and Ren Z 2011 Thermoelectric energyconversion using nanostructured materials Proc. SPIE 8031 80311J

[6]

Wang J, Liu H, Hu X, Jiang H, Zhao S, Li Q, Boughton R andJiang M 2001 Progress in skutterudite-based thermoelectric

materials Proc. Int. Conf. Thermoelectrics (Beijing, Jun.2001) pp 89–92

[7]

Kleinke H 2010 New bulk materials for thermoelectric powergeneration: clathrates and complex antimonides Chem. Mater . 22 604–11

[8]

Caillat T and Fleurial J 1996 Zn-Sb alloys for thermoelectricpower generation Proc. IECEC (Washington DC, August1996 ) pp 905–9

[9]

Gelbstein Y, Dashevsky Z and Dariel M 2008 The search formechanically stable PbTe based thermoelectric materials J. Appl. Phys. 104 033702

[10] Heremans J, Jovovic V, Toberer E, Saramat A, Kurosaki K,Charoenphakdee A, Yamanaka S and Snyder J 2008Enhancement of thermoelectric efficiency in PbTe bydistortion of the electronic density of states Science 321

554–7

[11]

Pei Y, Shi X, LaLonde A, Wang H, Chen L and Snyder J 2011Convergence of electronic bands for high performance bulkthermoelectrics Nature 473 66–9

[12] Pantha B, Dahal R, Li J, Lin J, Jiang H and Pomrenke G 2009Thermoelectric properties of In0.3Ga0.7N alloys J. Electron. Mater . 38 1132–5

[13]

Schaeuble N, Aguirre M, Weidenkaff A, Trottmann M,Haemmerli A, Bocher L and Hug P 2008 Aluminium-substituted zinc oxide for thermoelectric energy conversionProc. Eur. Conf. Thermoelectronics (Paris, July 2008)

[14]

Ong K, Singh D and Wu P 2011 Analysis of the thermoelectricproperties of n-type ZnO Phys. Rev. B 83 115110

[15]

Fleurial J, Borshchevsky A, Caillat T and Ewell R 1996 Newmaterials and devices for thermoelectric applications Proc.

IECEC (Washington DC, August 1996 ) pp 1080–5

[16]

Caillat T, Fleurial J, Snyder J and Borshchevsky A2001 Development of high efficiency segmentedthermoelectric unicouples Proc. Int. Conf. Thermoelectrics( Beijing, June 2001) pp 282–5

[17]

Macia E 2004 Compatibility factor of segmentedthermoelectric generators based on quasicrystalline alloysPhys. Rev. B 70 100201

[18]

Snyder J 2004 Application of the compatibility factor tothe design of segmented and cascaded thermoelectricgenerators Appl. Phys. Lett . 84 2436–8

[19] Majumdar A 2004 Thermoelectricity in semiconductornanostructures Science 303 777–8

[20]

Singh D and Terasaki I 2008 Thermoelectrics: nanostructuringand more Nature Mater . 7 616–7

[21]

lan Y, Minnich A, Chen G and Ren Z 2010 Enhancement ofthermoelectric figure-of-merit by a bulk nanostructuringapproach Adv. Funct. Mater . 20 357–76

[22]

Hochbaum A, Chen R, Delgado R, Liang W, Garnett E,Najarian M, Majumdar A and Yang P 2008 Enhancedthermoelectric performance of rough silicon nanowires Nature 451 163–7

[23]

Lee J, Galli G and Grossman J 2008 Nanoporous silicon as anefficient thermoelectric materials Nano Lett . 8 3750–4

[24]

Bux S, Blair R, Gogna P, Lee H, Chen G, Dresselhaus M,Kaner R and Fleurial J 2009 Nanostructured bulksilicon as an effective thermoelectric material Adv. Funct. Mater . 19 2445–52

[25]

Ramayya E and Knezevic I 2009 Ultrascaled silicon nanowiresas efficient thermoelectric materials Proc. Int. WorkshopComputers and Electronics ( Beijing, May 2009) pp 1-4


http://dx.doi.org/10.1063/1.3212675http://dx.doi.org/10.1063/1.3212675http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1109/MSPEC.2011.5960150http://dx.doi.org/10.1109/MSPEC.2011.5960150http://dx.doi.org/10.1117/12.885759http://dx.doi.org/10.1117/12.885759http://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1063/1.2963359http://dx.doi.org/10.1063/1.2963359http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1103/PhysRevB.70.100201http://dx.doi.org/10.1103/PhysRevB.70.100201http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1002/adfm.200900250http://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1021/nl802045fhttp://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1038/nature06381http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1002/adfm.200901512http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1038/nmat2243http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1126/science.1093164http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1063/1.1689396http://dx.doi.org/10.1103/PhysRevB.70.100201http://dx.doi.org/10.1103/PhysRevB.70.100201http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1007/s11664-009-0676-8http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1038/nature09996http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1126/science.1159725http://dx.doi.org/10.1063/1.2963359http://dx.doi.org/10.1063/1.2963359http://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1021/cm901591dhttp://dx.doi.org/10.1117/12.885759http://dx.doi.org/10.1117/12.885759http://dx.doi.org/10.1109/MSPEC.2011.5960150http://dx.doi.org/10.1109/MSPEC.2011.5960150http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1016/j.sna.2003.11.039http://dx.doi.org/10.1063/1.3212675http://dx.doi.org/10.1063/1.3212675


13/13

M T de Leon et al

12

[26]

Cerofolini G, Ferri M, Romano E, Roncaglia A, Selezneva E,Arcari A, Suriano F, Veronese G, Solmi S and Narducci D2010 Industrially scalable process for silicon nanowiresfor seebeck generation Proc. Eur. Conf. Thermoelectrics (Como, Italy, Septembr 2010) pp 147–51

[27]

Hao Q, Zhu G, Joshi G, Wang X, Minnich A, Ren Z andChen G 2010 Theoretical studies on the thermoelectricfigure of merit of nanograined bulk silicon Appl. Phys. Lett .

97 063109

[28]

Kessler V et al 2014 Fabrication of high-temperature-stablethermoelectric generator modules based on nanocrystallinesilicon J. Electron. Mater . 43 1389–96

[29]

Wang X et al 2008 Enhanced thermoelectric figure ofmerit in nanostructures n-type silicon germanium bulk alloy Appl. Phys. Lett . 93 193121

[30] Venkatasubramanian R, Siivola E, Colpitts T andO’Quinn B 2001 Thin-film thermoelectric devices with highroom-temperature figure of merit Nature 413 597–602

[31] Wang W, Jia F, Huang Q and Zhang J 2005 A new type oflow power thermoelectric micro-generator fabricated bynanowire array thermoelectric material Microelectron. Eng.77 223–9

[32]

Koukharenko E, Nandhakumar I, Frety N, Beeby S,Cox D, Tudor M, Schiedt B, Trautmann C, Bertsch A andWhite N 2008 Towards a nanostructured thermoelectricgenerator using ion-track lithography J. Micromech. Microeng. 18 104015

[33] Lan Y, Poudel B, Ma Y, Wang D, Dresselhaus M, Chen Gand Ren Z 2009 Structure study of bulk nanograinedthermoelectric bismuth antimony telluride Nano Lett .9 1419–22

[34] Minnich A, Dresselhaus M, Ren Z and Chen G 2009 Bulknanostructured thermoelectric materials: current researchand future prospects Energy Environ. Sci. 2 466–79

[35]

Robert R, Romer S, Reller A and Weidenkaff A 2005Nanostructured complex cobalt oxides as potentialmaterials for solar thermoelectric power generators

Adv. Eng. Mater . 7 303–8

[36]

Rowe D 2006 Thermoelectric waste heat recovery as a renewableenergy source Int. J. Innov. Energy Syst. Power 1 13–23

[37]

Baranowski L, Snyder J and Tobere E 2012 Concentratedsolar thermoelectric generators Energ. Environ. Sci. 5 9055–67

[38]

Telkes M 1954 Solar thermoelectric generators J. Appl. Phys.25 765–77

[39] Amatya R and Ram R 2010 Solar thermoelectricgenerator for micropower applications J. Electron. Mater .39 1735–40

[40]

Baglio S, Castorina S, Fortuna L and Savalli N 2002Development of autonomous, mobile micro-electro-mechanical devices Proc. IEEE Symp. Circuits Syst . 4 285–8

[41]

Kraemer D et al 2011 High-performance flat-panel solarthermoelectric generators with high thermal concentration Nature Mater. 10 532–8

[42]

Mizoshiri M, Mikami M, Ozaki K and Kobayashi K 2012Thin-film thermoelectric modules for power generationusing focused solar light J. Electron. Mater . 41 1713–9

[43] Roncaglia A and Ferri M 2011 Thermoelectric materialsin MEMS and NEMS: a review Sci. Adv. Mater . 3 401–19

[44]

Korvink J and Paul O 2006 MEMS: a Practical Guide to Design, Analysis, and Applications (Norwich, NY: WilliamAndrew)

[45]

Bejan A and Kraus A 2003 Heat Transfer Handbook (New York: Wiley)

[46]

Welty J, Wicks C, Wilson R and Rorrer G 2008 Fundamentalsof Momentum, Heat, and Mass Transfer 5th edn (New York:

Wiley)

[47]

Weisstein E Quartic equation (MathWorld—a Wolfram Web Resource) http://mathworld.wolfram.com/QuarticEquation.html. Accessed Oct. 2013

[48]

Ikeda H and Salleh F 2010 Influence of heavy doping onseebeck coefficient in silicon-on-insulator Appl. Phys. Lett .96 012106

[49] Salleh F, Asai K, Ishida A and Ikeda H 2009 Seebeckcoefficient in ultrathin silicon-on-insulator layers Appl.Phys. Express 2 071203

[50]

Xie J, Lee C, Wang M, Liu Y and Feng H 2009Characterization of heavily doped polysilicon filmsfor CMOS-MEMS thermoelectric power generators J. Micromech. Microeng. 19 125029

[51]

Kasap S 2001 Thermoelectric Effects in Metals:

Thermocouples (http://materials.usask.ca/samples/ Thermoelectric-Seebeck.pdf )

[52]

Shackelford J and Alexander W 2001 CRC Materials Scienceand Engineering Handbook (Boca Raton, FL: CRC Press)

[53]

The Physics Hypertextbook (http://physics.info/ electric-resistance)

http://dx.doi.org/10.1063/1.3478459http://dx.doi.org/10.1063/1.3478459http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1063/1.3027060http://dx.doi.org/10.1063/1.3027060http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1088/0960-1317/18/10/104015http://dx.doi.org/10.1088/0960-1317/18/10/104015http://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1166/sam.2011.1168http://dx.doi.org/10.1166/sam.2011.1168http://dx.doi.org/10.1166/sam.2011.1168http://mathworld.wolfram.com/QuarticEquation.htmlhttp://mathworld.wolfram.com/QuarticEquation.htmlhttp://dx.doi.org/10.1063/1.3282783http://dx.doi.org/10.1063/1.3282783http://dx.doi.org/10.1143/APEX.2.071203http://dx.doi.org/10.1143/APEX.2.071203http://dx.doi.org/10.1088/0960-1317/19/12/125029http://dx.doi.org/10.1088/0960-1317/19/12/125029http://materials.usask.ca/samples/Thermoelectric-Seebeck.pdfhttp://materials.usask.ca/samples/Thermoelectric-Seebeck.pdfhttp://physics.info/electric-resistancehttp://physics.info/electric-resistancehttp://physics.info/electric-resistancehttp://physics.info/electric-resistancehttp://physics.info/electric-resistancehttp://physics.info/electric-resistancehttp://materials.usask.ca/samples/Thermoelectric-Seebeck.pdfhttp://materials.usask.ca/samples/Thermoelectric-Seebeck.pdfhttp://dx.doi.org/10.1088/0960-1317/19/12/125029http://dx.doi.org/10.1088/0960-1317/19/12/125029http://dx.doi.org/10.1143/APEX.2.071203http://dx.doi.org/10.1143/APEX.2.071203http://dx.doi.org/10.1063/1.3282783http://dx.doi.org/10.1063/1.3282783http://mathworld.wolfram.com/QuarticEquation.htmlhttp://mathworld.wolfram.com/QuarticEquation.htmlhttp://dx.doi.org/10.1166/sam.2011.1168http://dx.doi.org/10.1166/sam.2011.1168http://dx.doi.org/10.1166/sam.2011.1168http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1007/s11664-012-2047-0http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1038/nmat3013http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1007/s11664-010-1190-8http://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1063/1.1721728http://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1039/c2ee22248ehttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1021/nl803235nhttp://dx.doi.org/10.1088/0960-1317/18/10/104015http://dx.doi.org/10.1088/0960-1317/18/10/104015http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1016/j.mee.2004.11.005http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1038/35098012http://dx.doi.org/10.1063/1.3027060http://dx.doi.org/10.1063/1.3027060http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1007/s11664-014-3093-6http://dx.doi.org/10.1063/1.3478459http://dx.doi.org/10.1063/1.3478459

solar thermoelectric generators fabricated on a silicon-on-insulator substrate

Documents