NSF-DOE Thermoelectrics Partnership:

Automotive Thermoelectric Modules with Scalable Thermo- and Electro-Mechanical Interfaces

Prof. Ken GoodsonDepartment of Mechanical EngineeringStanford University

Dr. Boris KozinskyEnergy Modeling, Control, & ComputationR. Bosch LLC

Prof. George NolasDepartment of PhysicsUniversity of South Florida

Automotive Thermoelectric Modules with Scalable Thermo-and Electro-Mechanical Interfaces

Project LeadershipProf. Ken Goodson, Stanford Mechanical EngineeringProf. George Nolas, USF Department of PhysicsDr. Boris Kozinsky, Energy Comp. & Modeling, BoschProf. Mehdi Asheghi, Stanford Mechanical EngineeringDr. Winnie Wong-Ng, NIST Functional Properties Group

StaffDr. Yongkwan Dong, USF Department of PhysicsDr. Matt Panzer, Stanford Mechanical Engineering

Students:Yuan Gao, Lewis Hom, Saniya Leblanc, Amy Marconnet

Leveraged Support:Northrop Grumman, AMD/SRC, ONR, AFOSRFellowships from NSF, Sandia National Labs, Stanford DARE

Program Managers: John Fairbanks, Tom Avedisian (DOE). Arvind Atreya (NSF)

Key Challenges for Thermoelectricsin Combustion Systems

…Low-thermal-resistance interfaces with tailored electrical properties, which are stable under thermal cycling.

…High-temperature TE materials that are stable and promise low-cost scaleup.

…Characterization methods that include interfaces and correlate better with system performance.

hot side / heat exchanger

insulation / e.g. ceramic plate

conductor/ e.g. copper

insulation / e.g. ceramic plate

cold side / heat exchanger

conductor/ e.g. copper conductor/ e.g. copper

n p

thermalthermal

electrical& thermal

electrical& thermalthermalthermal

diffusion barrier

joining technology

contact resistanceHeat

Cooling

hot side / heat exchanger

insulation / e.g. ceramic plate

conductor/ e.g. copper

insulation / e.g. ceramic plate

cold side / heat exchanger

conductor/ e.g. copper conductor/ e.g. copper

n p

thermalthermal

electrical& thermal

electrical& thermalthermalthermal

diffusion barrier

joining technology

contact resistanceHeat

Cooling

600 °C

90 °C

Improvements in the intrinsic ZT of TE materials are proving to be very difficult to translate into efficient, reliable TEG systems.

Major needs include…

Thermal Interface Challenges for Thermoelectrics

• Combustion systems experience enormous stresses at interfaces due to large temperature differences.

• Interfaces must offer low thermal resistance, targeted electrical performance, mechanical compliance.

Thermoelectric Interface Challenge

200 µm

Connecting metal

Solder

Hatzikraniotis et al., Proc. Mater. Res. Soc. Symp., 2009.

Clin et al., J. Electronic Materials, 2009

“Nanostructured Interfaces for Thermoelectrics,”Gao, Marconnet, Leblanc, Shakouri, Goodson et al., Proc ICT 2009, J. Electronic Materials, 2010

Pettes, Hodes, Goodson, Trans. Advanced Packaging, 2009

0 2 4 6 8 10 12 14 16 18 200

20

40

60

80

100

120

140

160

180

200

TEM thickness (mm)

Pow

er (W

)

No interfaceThermal greaseSolderCNT array

LeBlanc, Gao, Goodson,Proc. IMECE 2008

Thermal Interface Challenges for Thermoelectrics

Thermoelectric Interface Challenge

200 µm

Connecting metal

Solder

Hatzikraniotis et al., Proc. Mater. Res. Soc. Symp., 2009.

Clin et al., J. Electronic Materials, 2009

TEG (combustion)

“Nanostructured Interfaces for Thermoelectrics,”Gao, Marconnet, Leblanc, Shakouri, Goodson et al., Proc ICT 2009, J. Electronic Materials, 2010

Pettes, Hodes, Goodson, Trans. Advanced Packaging, 2009

• Combustion systems experience enormous stresses at interfaces due to large temperature differences.

• Interfaces must offer low thermal resistance, targeted electrical performance, mechanical compliance.

Automotive Thermoelectric Modules with Scalable Thermo- and Electro-Mechanical Interfaces

GOALSDevelop, and assess the impact of, novel interface and material solutions for TEG systems of interest for Bosch.

Explore and integrate promising technologies including nanostructured interfaces, filled skutterudites, cold-side microfluidics.

Practical TE characterization including interface effects and thermal cycling. METHODS

Multiphysics simulations ranging from ab-initio (band structure) to system scale.

Photothermal metrology including Pico/nanosecond TDTR, cross-sectional IR. MEMS-based mechanical characterization.

System impact assessment considering the interplay of thermal, fluidic, mechanical, electrical, and thermoelectric phenomena.

Removable backer

Nanostructured Film

Mid-temperature binderAdhesion layer

Panzer, Dai, Goodson, et al., Patent Pending (2007)

BoschPrototype TEG in exhaust system

Hu, Fisher, Goodson, et al., J. Heat Transfer (2006)

Area(emphasis)

Specifics Source

Interfaces100%

Nanostructured films & composites, metallic bondingAb initio simulations and optimization

StanfordBosch

Metrology100%

(ZT)eff with independent k&cp, thermal cyclingHigh temperature ZT

StanfordUSF/NIST

Materials100%

Filled skutterudites and half Heusler intermetallicsAb initio simulations for high-T optimization

USFBosch

Durability50%

In-situ thermal cycling tests, propertiesInterface analysis through SEM, XRD, EDS

StanfordBosch

Heat sink50%

Gas/liquid simulations using ANSYS-FluentNovel cold HX using microfluidics, vapor venting

BoschStanford

System50%

System specification, multiphysics codeEvaluation of research impacts

BoschStanford

Outreach TE for vehicles competition, UG Lab, K-12 outreach Stan&USF

Research Overview

Heavy-ion Filling Yields Lower Thermal Conductivity. Low Valence Filling Facilitates Optimization of Power Factor and ZT.

Nolas et al., MRS Bulletin (2006).Nolas, Kaeser, Littleton, Tritt, APL 77, 1855 (2000), Nolas, JAP 79, 4002 (1996)

Lamberton, G.S. Nolas, et al. APL 80, 598 (2001). Nolas, Cohn, and Slack, PRB (1998)

Bulk TE Materials for Automotive Applications

•Skutterudites with partial filling using heavy, low valence “guest” atoms

George S. NolasDepartment of Physics,

University of South Florida

Heavy-ion Filling Yields Lower Thermal Conductivity.

Partial Filling – Optimization of mobility & thermal conductivity

Bulk TE Materials for Automotive Applications

George S. NolasDepartment of Physics,

University of South Florida

•Half-Heusler alloys: from small grain size towards the disordered state

x= 0.3 ; y = 1.5

x= 0.05 ; y=0

x= 0.6 ; y = 2.4

x= 0.23 ; y=0

x= 0.75 ; y = 2.6

x=y=0

x= 0.9 ; y = 2.4

LaxCo4Sb12-ySny

x= 0.3 ; y = 1.5

x= 0.05 ; y=0

x= 0.6 ; y = 2.4

x= 0.23 ; y=0

x= 0.75 ; y = 2.6

x=y=0

x= 0.9 ; y = 2.4

LaxCo4Sb12-ySny

Nolas et al., MRS Bulletin (2006).Nolas, Kaeser, Littleton, Tritt, APL 77, 1855 (2000), Nolas, JAP 79, 4002 (1996)

Lamberton, G.S. Nolas, et al. APL 80, 598 (2001). Nolas, Cohn, and Slack, PRB (1998)

•Skutterudites with partial filling using heavy, low valence “guest” atoms

Ab-initio/BTE computations will assist the optimization of TE material stoichiometry.

Past work at Bosch predicted the effect of Ba filling on CoSb3 skutterudites using DFPT.

Collaborative optimization with Nolas group will focus on filled skutterudites, mobility, seebeck, and interfaces with metallics.

Functional Simulation & Optimization of Materials

Seebeck coefficient

experiment*ab-initio

ab-initio τ(E)

Resistivity

experiment

CoSb

ab-initio

Boris Kozinsky (CR/RTC2-NA) | 12/14/2010 | © Robert Bosch GmbH 2010. All rights reserved, also regarding any disposal, exploitation, reproduction, editing, distribution, as well as in the event of applications for industrial property rights.

Research and Technology Center North America

Dr. Boris Kozinsky, Energy Modeling, Control, and Computation, Bosch LLCWee, Kozinsky et al, Phys. Rev. B 81 (2010)

Interface Modeling & OptimizationKozinsky, Physical Review Letters (2006). Panzer, Goodson et al. JAP (2008)

CTE of TiNiSn Half-Heusler Alloys

(Bosch) exp

ab-initio

Unfilled CoSb3 (stable)

OrthorhombicCoSb2 (Unstable)

MonoclinicCoSb2(Stable)

Co-Sb phase diagramEn

ergy

Diff

eren

ce p

er A

tom

(eV)

Simulations examine thermodynamic stability of TE material phases and assess potential for interdiffusion.

Simulations examine interface electrical conduction and optimize resistance considering band structure.

Mechanical & thermal simulations will focus on the expansion coefficients and transport through low-dimensional contacts.

Panzer, GoodsonJ. Appl. Phys (2008)

Low-dimensional contact resistance

Conduction Physics in CNT Films

SubstrateSubstrateSubstrateSubstrate

Individual CNT conductance

Inter-tube contact

Nanoscale metal-CNT interface resistance (phonons)

Partial nanotube engagement

Spatially varying aligment

Growth interface resistance

Pop,Dai,Goodson et al. Physical Review Letters (2005)Pop,Dai,Goodson et al. Nano Letters (2006)Hu,Fisher,Goodson et al. Journal of Heat Transfer (2006,07)Panzer,Dai,Goodson et al. Journal of Heat Transfer (2008)Panzer,Goodson Journal of Applied Physics (2008)Gao,Shakouri,Goodson et al. Journal of Electronic Materials (2010)Panzer,Murayama,Goodson et al. Nano Letters (2010)

Nanosecond Thermoreflectance Picosecond Thermoreflectance

Cross-sectional IR MicroscopyMechanical Characterization

Sample Film

Heat Sink (Si or metal substrate)

Metal CoatingPulsed Laser

Heating

Probe laser for thermometry

Transparent Substrate (e.g. quartz)

Metal CoatingCNT Catalyst

CNT

Laser Doppler Velocimeter

Vibrometer, ω

Polysilicon CNT

Si substrateOxide

Heater

growth substrate

CNT Film

CNT

kdxdT 1

∝

BRTdxdTk

→∆

∝

−1

500 μm

Distance (μm)

Thermal and MechanicalCharacterization of Aligned CNT Films

010203040506070

0.005 0.006 0.007 0.008 0.009

Metallization and CNT Thermal Resistanceusing Nanosecond Thermoreflectance

RCNT-Sub + RCNT

Rtot

RCNT-Metal

φ= Ceff/Cv,individual

Ther

mal

Res

ista

nce

(m2 K

/MW

)

Ti

Pt Pd

Al

Ni

║0.10

PotentialPerformance

Panzer, Murayama, Wardle, Goodson, et al., Nano Letters (2010)

CNT Engagement Factor

Mechanical Behavior of CNT Films

Maruyama Lab Samples (SWNT), 95 kg/m3

MonanoSamples (MWNT), ~30 kg/m3

Wardle Lab Samples/MIT(MWNT), 45 kg/m3

Students: Yoonjin Won, Yuan Gao, Matt Panzer. Collaborators: Prof. Wei Cai, Stanford ME

Zipping/Velcro Model

Thickness (μm)

Modulus (MPa)

Density (kg/m3)

CNTTop 0.4 140 >29CNTMiddle 0-150 7 29Si 8.7 155e3 2330

Crust Model

Thermal & Mechanical Requirements at Interfaces

Thermal Resistivity (m K / W)1.00.1

Elas

tic M

odul

us (M

Pa)

Greases & Gels

Organic Phase Change

Metallic Alloys

Adhesives

0.01

Nano-gels

Lifetimethermal cycling

104

103

102

101

Thermal & Mechanical Requirements at Interfaces

Thermal Resistivity (m K / W)1.00.1

Elas

tic M

odul

us (M

Pa)

Greases & Gels

Organic Phase Change

Metallic Alloys

Adhesives

0.01

Nano-gels

Lifetimethermal cycling

104

103

102

101

Our Latest CNT Data

Thermal Cycling of CNT-SiGe Composite

0

2

4

6

8

10

12

1 10 100

R" (

m2

K M

W-1

)

N Cycles

R" TotalR" CNT-SubR" CNT-Pt

R”Total R”CNT-Sub + R”CNTR”CNT-Metal

Resistances for 1.5, 2.5, and 40 micron thick CNT films varied between 0.035 and 0.055 cm2 oC/W, with evidence of decreasing engagement with increasing film thickness.

Gao, Leblanc, Marconnet, Shakouri, Goodson et al., “Nanostructured Interfaces for Thermoelectrics,“Proc. ICT 2009. J. Electronic Materials (2010).

Cycles (30 to 200 C, 6min)

(ZT)eff Characterization with Electrical Heating & Cross-Sectional IR Thermometry

p-type

Pellet

Ceramic plate

Ceramic plate

Connecting metal

Cold Side

Cross-sectional IR Microscopy

Th = 300-700 K

AC current source

Rv

Rref

Resistive Heater

Lock-inAmp.

n-type

PelletVSource

RSense

1

2

3

4

10-3 10-2 10-1 1 10 102 103

6.0

5.0

3.5

3.0

Phase (deg)

Frequency (Hz)

Nor

mal

ized

tem

p. A

mpl

itude

0.0

45

90

Thermal

Sapp

hire

IR T

rans

pare

nt w

indo

w

Heater

growth substrate

CNT Film

CNT

kdxdT 1

∝

QFI

Spatial resolution ~ 2 mmTemp. range 300-700 K

IR microscope systemHeating/thermometrySetup

Connecting metal

Boris Kozinsky (CR/RTC2-NA) | 12/14/2010 | © Robert Bosch GmbH 2010. All rights reserved, also regarding any disposal, exploitation, reproduction, editing, distribution, as well as in the event of applications for industrial property rights.

Research and Technology Center North America

HX and System-Level Simulations Bosch-lead system

simulations explore impact of improved parameters on system efficiency

Multiphysics simulations of thermal/thermoelectric transport in TE material, and interface transport incorporating ab initio results.

HX design and optimization accounts for novel pressure drop designs including Stanford Vapor Escape technology

Educational EngagementThermoelectrics for Vehicles Challenge: Multi-University Competition

Undergraduate Thermoelectrics Lab

K-12 Educational Outreach

Long-term vision: Teams of undergraduates work with commercial TE components and heat sinks to extract waste heat from demo vehicle exhaust.

Stanford’s heat transfer course (ME131A) will include a thermoelectrics laboratory experience. Connects theory and practical applications. Recruits undergraduates for research experiences

in thermoelectrics with graduate student mentoring.

High school students and teachers will conduct energy-conversion research in Stanford’s Microscale Heat Transfer Laboratory.

Connects classroom education and research & development.Links students with industry, graduate & faculty advisors.

Interactions and Flow ofSamples & Information

Stanford• Prepares CNTs samples on TE materials• Transport property measurements of CNT-TE pellet combination, thermomechanical reliability tests on interface (300-800 K)• Process development for CNT TIM tape

Bosch• Ab-initio simulations of transport properties of TE materials and interfaces.•System-level simulation and optimization

USF• Develops high-T, high efficiency TE materials• Transport properties (ρ, S and κ) and Hall measurements (10 - 300K)• Structural, morphological and thermal (DTA/TGA) analyses

NIST• Transport properties (ρ, S and κ) and Hall measure- ments (1.8-390K)•Specific heat, Power Factor measurement at 300 K.• Custom-designed precision TE properties measurement system (300 – 1200 K)

1- Interface2- System-level3- Durability4- Materials5- Heat sink6- Metrology

1, 3, 4, 6

4, 6

1, 2, 3, 5