Design and Experimental

Validation of a Micro-Nano Structured

Thermal Ground Plane
for High-g Environments

A dissertation submitted to the

Graduate School of the University of Cincinnati
in partial fulfillment of the requirements for the degree of

DOCTORATE OF PHILOSOPHY (Ph. D.)

in the School of Dynamic Systems – Mechanical Engineering

of the College of Engineering and Applied Science
April 2013

by Hendrik Pieter Jacobus de Bock

M.S. (Mechanical Engineering), Mechanical Engineering,
University of Twente, Netherlands

Committee Chair: Dr. Frank M. Gerner

Abstract

As capabilities and associated power consumption of computers, portable devices, avionics and

other electronics systems have risen rapidly over the last decade, chip heat fluxes have risen sharply.

The function of the thermal management solution is to provide sufficient low thermal resistance

spreading such that the heat generated can be rejected at minimal temperature rise. Heat pipes and

vapor chambers are efficient mechanisms for spreading heat as they employ two-phase heat

transfer and capillary transport for operation. A limitation of these devices is that their operation

capability is affected by body forces, such as gravity.

This work describes the relevant physics related to the gravitational dependence of a Thermal

Ground Plane(TGP). These relations are developed into a design model for TGPs where a design space

is identified for successful operation in high-g force environment. From this evaluation, four factors

are identified that require empirical evaluation in order to evaluate design performance. The factors

are, the capillary pressure, permeability and thermal conductivity of the wick structure and the

evaporator and condenser heat transfer performance. Experiments and methods were designed to

evaluate each one of these factors.

Using the design model and the empirical inputs, 3cm TGP prototypes were developed for testing. As

the thermal conductivity of TGP can be very high, significant effort was put into evaluating the

uncertainty of the experimental setup. A TGP prototype was evaluated with effective thermal

conductivity in excess of 461 W/m-K, exceeding common solid copper thermal conductivity. The

experimental setup was modified for operation on a high-g centrifuge at Air Force Research

Laboratory (AFRL). On this centrifuge, the TGP demonstrated operation at a thermal conductivity of

436 W/m-K at 10.5 g’s. This validates that it is feasible to have a vapor chamber operational in a

high-g environment, enabling additional application space such as avionics systems. Subsequent

work demonstrated 15cm TGP with effective thermal conductivity in excess of 5000 W/m-K.

ii
iii
Preface

I dedicate this work to my wife Megan and our children Jacob and Mara.

Through our life Journey, we encounter challenges and opportunities.

What defines us on how we respond to these.

Sometimes we embark on a Journey where we cannot see the end clearly,

a Journey where we cannot fully control our destiny.

In these moments, we are blessed when there are others that believe in you,

People who support you and encourage you

Thank you to all the people that supported me and the TGP program on this Journey.

Your support meant a lot to me.

My family and friends

Prof. Frank Gerner

The General Electric Company

DARPA

University of Cincinnati

Air Force Research Laboratory

iv
Table of Contents

Abstract ................................................................................................................................................................................................... ii

Preface .................................................................................................................................................................................................... iv

Table of Contents................................................................................................................................................................................ v

List of Figures........................................................................................................................................................................................ x

List of Tables .......................................................................................................................................................................................xv

Nomenclature .................................................................................................................................................................................. xvi

1. Introduction.................................................................................................................................................................................1

1.1. Motivation ..........................................................................................................................................................................1

1.2. Thermal Ground Plane Program .............................................................................................................................4

1.3. Scope ....................................................................................................................................................................................5

2. Thermal Ground Plane Design Approach & Model ..................................................................................................7

2.1. Introduction.......................................................................................................................................................................7

2.2. Heat Pipe Operation......................................................................................................................................................7

2.3. Heat Pipes, Thermosyphons and Vapor Chambers.......................................................................................9

2.4. Thermal Ground Plane Transport Model .......................................................................................................... 11

2.4.1. Working Fluid Mass Transport Rate ......................................................................................................... 11

2.4.2. Criterion for Working Fluid Transport ...................................................................................................... 13

2.4.3. Capillary Pressure Rise ................................................................................................................................... 16

2.4.4. Vapor Transport Pressure Drop ................................................................................................................. 17

2.4.5. Liquid Transport Pressure Drop ................................................................................................................. 19

v
 2.4.6. Pressure Drop due to Body Forces ........................................................................................................... 21

2.5. Heat Transfer Model .................................................................................................................................................. 23

2.5.1. Substrate conduction thermal resistance ............................................................................................ 26

2.5.2. Wick/Evaporation resistance ...................................................................................................................... 28

2.5.3. Vapor Transport Resistance......................................................................................................................... 31

2.5.4. Wick Condensation Resistance .................................................................................................................. 33

2.6. Combined Transport Model and Pressure Diagram................................................................................... 33

2.7. TGP Vapor Chamber Design for High-g Operation ..................................................................................... 35

2.8. Effective Thermal Conductivity ............................................................................................................................. 39

2.9. Need for empirical factors ...................................................................................................................................... 44

3. Experimental Evaluation of Thermal Ground Plane Empirical Factors ....................................................... 45

3.1. Introduction.................................................................................................................................................................... 45

3.2. Evaluation of Wick Capillary Performance ..................................................................................................... 45

3.2.1. Capillary Performance Measurement Methods ...................................................................................... 45

3.2.2. Bubble Point Measurement Setup Description ........................................................................................ 47

3.2.3. Experiment Validation Using Wire Mesh Structures ............................................................................. 49

3.2.4. Measuring Capillary Performance of Sintered Wicking Structures................................................ 51

3.3. Permeability ................................................................................................................................................................... 55

3.3.1. Wick Permeability Formulation ....................................................................................................................... 55

3.3.2. Experimental Evaluation of Wick Permeability ........................................................................................ 56

3.3.3. Discrete model for Square Shaped Wick Permeability Experiment ............................................... 58

vi
 3.3.4. Wick Permeability Results .................................................................................................................................. 60

3.4. Wick Thermal Conductivity ..................................................................................................................................... 61

3.4.1. Laser Flash Method ............................................................................................................................................... 61

3.4.2. Wick Thermal Conductivity Results ............................................................................................................... 62

3.5. Experimental Evaluation of Heat Transfer in the Evaporator ................................................................ 63

3.5.1. Evaporation Heat Transfer Experimental Setup ..................................................................................... 64

3.5.2. Evaporation Heat Transfer Experiment Procedure................................................................................ 67

3.5.3. Evaporation Heat Transfer Experiment Results ...................................................................................... 69

3.6. Evaluation of Hysteresis Effects ........................................................................................................................... 71

3.7. Experimental Evaluation of Heat Transfer in the Condenser................................................................. 72

4. Visual Validation of TGP Performance under High-g Conditions ................................................................... 74

4.1. Introduction.................................................................................................................................................................... 74

4.2. Wick Saturation Visualization using Fluorescent Dyes ............................................................................. 74

4.3. TGP prototype Wick Saturation Visualization ................................................................................................ 75

4.4. Wick Saturation Visualization Validation ......................................................................................................... 76

4.5. Development of High-G TGP Wicking Visualization Experiment .......................................................... 77

4.6. High-G TGP Wicking Visualization Experiment Results ............................................................................. 78

5. Evaluation of TGP Thermal Performance................................................................................................................... 81

5.1. Introduction.................................................................................................................................................................... 81

5.2. TGP Thermal Characterization Experiment Description ........................................................................... 81

5.3. Setup components ..................................................................................................................................................... 82

vii
 5.4. Thermal Model of Experiment ............................................................................................................................... 85

5.4.1. Screening Thermal Model Results .................................................................................................................. 86

5.4.2. Detailed thermal model results ....................................................................................................................... 87

5.5. Measurement Uncertainty ...................................................................................................................................... 89

5.5.1. Temperature Measurement Uncertainty.................................................................................................... 90

5.5.1.1. Thermocouple Measurement Uncertainty................................................................................................. 91

5.5.1.2. Data Acquisition Contribution to Thermocouple Uncertainty .......................................................... 93

5.5.1.3. Propagation of Temperature Uncertainty.................................................................................................. 94

5.5.1.4. Temperature Measurement Uncertainty.................................................................................................... 95

5.5.2. Heat Input Uncertainty ........................................................................................................................................ 97

5.5.2.1. Power Supply Uncertainty ................................................................................................................................. 97

5.5.2.2. DAQ and Power Supply Uncertainty Propagation.................................................................................. 98

5.5.2.3. Heat Leakage ........................................................................................................................................................... 99

5.5.3. Effective Thermal Conductivity Uncertainty.............................................................................................. 99

5.6. Thermal Characterization Experiment Validation .................................................................................... 102

5.7. TGP Prototype Thermal Characterization ..................................................................................................... 104

5.8. High-g Thermal Performance Characterization Experiment Development ................................ 105

5.8.1. High-G Thermal Performance Characterization Hardware Description .................................. 105

5.8.2. High-g Thermal Performance Characterization Experiment Validation .................................. 108

5.9. High-G TGP Thermal Performance Characterization.............................................................................. 108

6. Subsequent Work and Application ............................................................................................................................ 110

viii
7. Conclusion ............................................................................................................................................................................. 114

Acknowledgements ..................................................................................................................................................................... 116

References ....................................................................................................................................................................................... 117

Appendix A: Fricton factor coefficient for non-circular geometries..................................................................... 122

Appendix B: TGP Design for 10G operation EES model .............................................................................................. 123

Appendix C: EES Finite difference model for Shape Factor derivation ............................................................... 124

Appendix D: Shape factor FDM results and correlation derivation ...................................................................... 128

Appendix E: Bubble point validation test results ........................................................................................................... 129

Appendix F: Wick on copper IPA test results ................................................................................................................... 131

Appendix G: Discrete Wick Permeability for Square Wick with Center Inlet .................................................... 132

Appendix H: TGP Thermal Performance Experiment model (APDL) ..................................................................... 134

Appendix I: Thermal Performance Characterization Experiment Validation ................................................... 140

ix
List of Figures

Figure 1 Trends in heat flux for common CPU, adapted from [2] ...............................................................................1

Figure 2 Noctua NH-D14 CPU cooler (surface area ~ 1m2) Source [4] .....................................................................2

Figure 3 Approximate heat flux and temperature rise profile estimated for NH-D14 heat sink (100

Watts heat input from chip)...........................................................................................................................................................2

Figure 4 Thermal Ground Plane Sketch ...................................................................................................................................5

Figure 5 Heat pipe operation ........................................................................................................................................................7

Figure 6 Vapor chambers (Source: Advanced Cooling Technologies (ACT)) ........................................................ 10

Figure 7 Loop Heat Pipe [11] ...................................................................................................................................................... 11

Figure 8 Pulsating Heat Pipe [12] ............................................................................................................................................. 11

Figure 9 Enthalpy diagram for water ..................................................................................................................................... 13

Figure 10 Simplified vapor chamber with one sided wick ........................................................................................... 14

Figure 11 Fluid meniscus sketch for sintered sphere wicking structure .............................................................. 16

Figure 12 Fricton factor coefficient and regression function..................................................................................... 18

Figure 13 Permeability constant models ( = 0.5) (adapted from Ababneh [19]) Chi refers to the Blake-

Kozeny relation(Eq. 12) .................................................................................................................................................................. 21

Figure 14 Two-dimensional meniscus shape variation simulation along the length of a vapor

chamber ............................................................................................................................................................................................... 22

Figure 15 Detailed TGP thermal resistance diagram ..................................................................................................... 24

Figure 16 Simplified TGP Thermal resistance diagram ................................................................................................. 25

Figure 17 Wick/Evaporation resistance elements for a sintered spherical particle wicking structure . 28

Figure 18 Heat transfer at the fluid vapor interface....................................................................................................... 30

Figure 19 Normalized saturation curve for water ........................................................................................................... 32

Figure 20. Modeled pressure diagram for water based 3 cm TGP vapor chamber ........................................ 34

x
Figure 21. Design Space as a function of body force multiplier for water at zero contact angle

(adapted from de Bock et al. [24])............................................................................................................................................ 36

Figure 22. Normalized loss in heat transport capability as a function of g-force multiplier for a 3cm

TGP using water as working fluid, assuming zero contact angle ............................................................................ 37

Figure 23. Anticipated Mean heat flux path for conduction through solid material (baseline) with

uniform in and out heat flux boundary conditions and adiabatic sidewalls ...................................................... 40

Figure 24. Temperature and finite elements plots for evaluation of the effective thermal conductivity

of a solid material for shape factor evaluation (not a vapor chamber) ................................................................ 41

Figure 25. Agreement between regressed and FDM evaluated effective length ............................................. 43

Figure 26. Liquid column height for different fluids (properties at P=1[atm], T=293 [K]) .............................. 46

Figure 27. Bubble point test machine. PMI BPT-1100-ABPT ....................................................................................... 48

Figure 28. Sketch of bubble point test fixture .................................................................................................................... 48

Figure 29. Bubble point pressure, time profile ................................................................................................................... 48

Figure 30. Image of 325 mesh ................................................................................................................................................... 49

Figure 31. Measurement of contact angle between wire mesh and water........................................................ 50

Figure 32. Wire mesh bubble point validation test results (IPA) ................................................................................ 50

Figure 33. SEM image of wick structure[27] ....................................................................................................................... 51

Figure 34. Wick sample preparation method .................................................................................................................... 52

Figure 35. Prepared sample, copper wick on copper adapter .................................................................................. 53

Figure 36. Effective pore radii results for copper wick sample using IPA ............................................................. 54

Figure 37. Bubble point pressure results for copper wick sample using IPA ...................................................... 55

Figure 38. Planar Wick Permeability configuration ......................................................................................................... 56

Figure 39. Feser[33] and proposed experimental configuration for in-plane permeability

measurement .................................................................................................................................................................................... 57

Figure 40. Radial flow permeability testing apparatus from Feser et al. [33] .................................................... 57

xi
Figure 41. User interface of discretized evaluation of wick permeability model .............................................. 59

Figure 42. Pressure drop as a function of Air Massflow through 26mm x 26mm x 1mm wicking

structure with center inlet and exit around periphery for range of Permeability coefficients .................. 60

Figure 43. Perpendicular Wick Permeability configuration ......................................................................................... 61

Figure 44. Laser Flash Method .................................................................................................................................................. 62

Figure 45. Laser Flash Test Results (Copper Wick sample result averaged over 45 tests) .......................... 63

Figure 46. Principle of Evaporation Heat Transfer and Mass Transport Test Setup ....................................... 64

Figure 47. Base fixture carroussel for three simultanous tests ................................................................................. 65

Figure 48. Base fixture carroussel with Window fixture installed ........................................................................... 65

Figure 49. Window fixture back ................................................................................................................................................ 65

Figure 50. Window fixture front ................................................................................................................................................ 65

Figure 51. Window fixture Section........................................................................................................................................... 65

Figure 52. Heat Transfer and Mass Transport Test Setup vacuum chamber .................................................... 66

Figure 53. Heat Transfer and Mass Transport Setup connection diagram ......................................................... 66

Figure 54. Heat Transfer and Mass Transport Setup ..................................................................................................... 67

Figure 55. Thermal Resistance Network mode1: Vessel Vapor in equilibrium with water........................... 68

Figure 56. Thermal Resistance Network mode2: Vessel Open, Limited evaporation ..................................... 68

Figure 57. Mass Transport Thermal Response data ....................................................................................................... 70

Figure 58. Mass Transport Thermal Response data ....................................................................................................... 71

Figure 59 Visualization of wick saturation in copper sintered wick using fluorescent dye over time .... 74

Figure 60 TGP Sample with Quartz Lid on filling station ............................................................................................... 75

Figure 61 UV Visualization setup.............................................................................................................................................. 75

Figure 62 TGP Sample and light source holder fixture .................................................................................................. 76

Figure 63 Baseline TGP UV lighting results.......................................................................................................................... 76

Figure 64 AFRL centrifuge table ............................................................................................................................................... 77

xii
Figure 65 TGP Sample and Light holder mounted on Aluminum structure ........................................................ 78

Figure 66 Saturated wick at 2.6g, excess liquid pools on top (G-force in direction).................................. 79

Figure 67 Sample is decelerated, excess liquid returns (G-force in direction).............................................. 79

Figure 68 Saturated wick at 8.9g, excess liquid pools on top (G-force in direction).................................. 80

Figure 69 Sample is decelerated, excess liquid returns (G-force in direction).............................................. 80

Figure 70 TGP Thermal Characterization Setup................................................................................................................ 81

Figure 71 Fujipoly 50Xr-M thermal resistivity as a function of contact pressure ............................................. 82

Figure 72 Lytron CP25 high performance cold plate ..................................................................................................... 83

Figure 73 Insulation shells with installed cold plate ....................................................................................................... 84

Figure 74 Overview of TGP Thermal Characterization Setup Components and interaction....................... 84

Figure 75 TGP thermal performance characterization experiment ........................................................................ 85

Figure 76 Relative contributions to resistance budget TGP keff=400 W/m-K..................................................... 87

Figure 77 Relative contributions to resistance budget TGP keff=2000 W/m-K .................................................. 87

Figure 78 TGP Thermal Characterization Setup................................................................................................................ 87

Figure 79 Finite Element Analysis Model vector plot ...................................................................................................... 88

Figure 80 Temperatures at 40W heat input (keff=400W/m-K, coolant 20C, ambient 20C) .......................... 89

Figure 81 Temperatures at 40W heat input (keff=400W/m-K coolant 20C, ambient 20C) ........................... 89

Figure 82 Effect of temperature measurement uncertainty on ............................................................................... 90

Figure 83 Circuit for coupling of a T-type thermocouple to a data acquisition system. Adapted from

[39] .......................................................................................................................................................................................................... 91

Figure 84 Type-T thermocouple voltage with 0°C reference, source NIST [40]................................................. 92

Figure 85 Type-T thermocouple Effective Seebeck coefficient ................................................................................. 93

Figure 86 Type-T thermocouple Effective Seebeck coefficient ................................................................................. 93

Figure 87 Circuit for coupling of a T-type thermocouple to a.................................................................................... 96

Figure 88 Heat leakage (results from FEA model used in paragraph 5.4.2) ........................................................ 99

xiii
Figure 89 Effective Thermal Conductivity uncertainty and contributors (k eff=400W/m-K, 2 x

Thermocouples) ............................................................................................................................................................................. 101

Figure 90 Effective Thermal Conductivity uncertainty and contributors (k eff=10000W/m-K, Thermopile)

............................................................................................................................................................................................................... 101

Figure 91 Relative effective thermal conductivity uncertainty at 40W and 120W input .......................... 102

Figure 92 Heat leakage (results from FEA model used in paragraph 5.4.2) ..................................................... 103

Figure 93 3cm TGP Effective Thermal Conductivity as a function of input power at 15ºC condenser

temperature .................................................................................................................................................................................... 104

Figure 94: High-g TGP Sample fixture................................................................................................................................. 106

Figure 95: Power Supply on Table ........................................................................................................................................ 107

Figure 96: Power Supply performance under acceleration ..................................................................................... 107

Figure 97: Test enclosure on centrifuge table ................................................................................................................ 107

Figure 98: Heater cold plate configuration on centrifuge ........................................................................................ 107

Figure 99: Test fixture on centrifuge validation ............................................................................................................. 108

Figure 100: High-G TGP Performance Results ................................................................................................................ 109

Figure 101: hevap as a function of ΔT, adapted from Ababneh et al.[35] ............................................................ 110

Figure 102: FEA model temperature contour of 15 cm TGP, adapted from Ababneh et al.[35] ........... 111

Figure 103: Comparison of model and experimental data for 15cm TGP prototype, adapted from

Ababneh et al.[35] ........................................................................................................................................................................ 111

Figure 104: GE IP Ruggedized Computing System ....................................................................................................... 113

Figure 105: Ruggedized System Typical Thermal Resistance breakdown........................................................ 113

xiv
List of Tables

Table 1 DARPA Thermal Ground Plane objectives and phases [5] .............................................................................4

Table 2 Relevant geometry parameters for typical device ......................................................................................... 37

Table 3 TGP Transport model results for 10 G design.................................................................................................... 38

Table 4 Parameters for Empirical Evaluation..................................................................................................................... 44

Table 5 High-G TGP Wicking Visualization Testing plan ............................................................................................... 79

Table 6 TGP testing setup performance, contact pressure > 40PSI, cold plate flow rate 1 GPM, k eff =

400 W/m-K.......................................................................................................................................................................................... 86

Table 7 TGP testing setup performance, contact pressure > 40PSI, cold plate flow rate 1 GPM, k eff =

2000 W/m-K ....................................................................................................................................................................................... 86

Table 8 Concept, detailed model comparison (keff=400W/m-K, coolant 20C, ambient 20C) ...................... 89

Table 9 Agilent 34970A Accuracy summary ...................................................................................................................... 94

Table 10 Temperature accuracy using Agilent 34970A and internal reference ............................................... 95

Table 11 Temperature accuracy using external reference at 0ºC with 0.06ºC accuracy ............................ 96

Table 12 Temperature accuracy using external reference at 95ºC with 0.06ºC accuracy ......................... 97

Table 13 Heat Input Accuracy using Agilent 3634A Power supply ......................................................................... 98

Table 14 Temperature and Heat input Uncertainty..................................................................................................... 100

Table 15 Validation Test Samples ......................................................................................................................................... 103

Table 16 Friction factor and Nusselt numbers for different geometries (adapted from [15]) ................. 122

xv
Nomenclature

area, [m2]
Biot number
Hagen-Poiseuille constant
friction factor coefficient
correction factor for 2D geometry
specific heat, [J/kg-K]
Diameter, [m]
hydraulic diameter, [m]
friction coefficient
gravitational acceleration, [m/s2]
Heat transfer coefficient, [W/m2-K]
latent heat of evaporation, [J/kg]
current, [A]
Jacob number
thermal conductivity, [W/m-K]
permeability factor, [m2]
length, [m]
̇ mass flow rate, [kg/s]
multiplier
mesh number, [1/m]
Pressure, [Pa]
perimeter, [m]
̇ volumetric heat generation, [W/m3]
heat load, [W]
radius, [m]
thermal resistance, [K/W]
Reynolds number
shape factor, [m]

xvi
 time, [s]
substrate thickness, [m]
wick thickness, [m]
temperature, [K]
velocity, [m/s]
Voltage, [V]
distance in x-direction, [m]
distance in x-direction, [m]
distance in z-direction, [m]

 body force angle, [deg]

Seebeck coefficient [V/K]
 temperature gradient, [K]
pressure gradient, [Pa]
 porosity
 liquid dynamic viscosity [Pa-s]
 density, [kg/m3]
 surface tension, [N/m]
contact angle, [deg]

xvii
1. Introduction

1.1. Motivation

The ongoing trend of miniaturization of semiconductor components has led to a dramatic increase in

the capability of electronics devices. This trend has enabled powerful modern devices such as smart

phones, tablets and laptops. The trend of component miniaturization has led to a point where

transistor count for a typical chip has exceeded one billion transistors without significantly increasing

chip size[1]. With this increase in transistors per area the heat production per area, or heat flux, has

also increased significantly. Heat fluxes in common microprocessors exceed levels of 100 W/cm2, as

illustrated by Figure 1. This represents nearly a two order of magnitude increase over the past two

decades.

Figure 1 Trends in heat flux for common CPU, adapted from [2]

The typical purpose of the thermal management solution is to remove heat efficiently from the

system at minimal temperature rise. In order to effectively maintain transistor junction temperature

1
of higher heat flux devices within design specifications[3], advances in thermal management

technologies are required.

As the typical objective of the thermal management solution is to reject heat to a coolant with

minimal temperature rise, a method is required to spread heat effectively from a high heat flux

source to a large convective surface area at low heat flux. For convection cooled solutions, since it is

challenging to increase the convective heat transfer coefficient without adding penalties such as

added noise or excessive power consumption, it is desired to employ as large a surface area as

possible. State of the art chips have a surface area of approximately 1 cm2 while state of the art

compact heat sinks employ a total convective surface area on the order of a square meter[4]. An

example of such a heat sink is presented in Figure 2.

Figure 2 Noctua NH-D14 CPU Figure 3 Approximate heat flux and temperature rise profile
cooler (surface area ~ 1m2) Source estimated for NH-D14 heat sink (100 Watts heat input from chip)
[4]

Figure 3 illustrates different levels of heat flux as can be observed for typical CPU cooling thermal

problem versus the typical temperature rise in each element of the heat transfer path (package,

2
thermal interface material (TIM), heat spreader, convection). Heat fluxes start at 100 W/cm2 at the

CPU die. Typical heat spreading in the chip package leads to significant reduction in heat flux

resulting in approximately 10 W/cm2 at the exposed CPU lid surface. In the typical solution, heat is

then transferred to a CPU cooler as depicted in Figure 2. Heat pipes are an important part of these

cooling systems. Heat pipes are highly effective heat spreaders that can spread heat at a

significantly smaller thermal resistance than highly thermally conductive metals such as Aluminum

or Copper. Figure 3 illustrates that although some of the heat spreading takes place in the chip

package and thermal interface layers, the majority of heat spreading occurs by means of heat pipes.

The heat pipe heat spreader allows for a reduction of the heat flux of nearly four orders of

magnitude, typically at very low thermal resistance (as illustrated by Figure 3). The high spreading

performance, low cost and passive nature of heat pipes have contributed to the fact that air cooling

has remained competitive in performance with liquid cooled solutions (also sketched in Figure 3).

Heat pipe operation and performance will be discussed in more detail in subsequent chapters,

current state of the art heat pipe performance is dependent upon orientation with respect to body

forces such as gravity. This limits application of heat pipes to those with fixed orientation and/or

applications where minimal body forces are exerted on the device. This has resulted in slow to

minimal adaption of heat pipes in avionics applications where significant g-forces might be exerted

upon the device. Due to their important role in heat spreading, development of next generation heat

pipes with less or minimal orientation dependence is important for cooling electronics in high-g force

applications.

The research presented in this work strives to advance the state of science for high performance

two-phase heat transfer devices such as heat pipes or vapor chambers for use in high g-force

environments. A combination of exploration of the relevant physics, derivation of parameter

combinations of importance, development of design practices for high g-force device design and

3
 experimental validation will be presented. In conclusion, the performance impact of such heat

spreader device will be evaluated in a high-g force environment.

1.2. Thermal Ground Plane Program

The Defense Advanced Research Agency (DARPA) has identified limitations of two-phase heat

transfer devices used in high g-force environments and in response issued a challenge to advance

the state of the art in heat spreading technology by development of a Thermal Ground Plane(TGP)[5].

A list of program objectives is given in Table 1.

Metric Unit SOA Phase 1 Phase 2 Phase 3

Hermeticity % fluid loss/year 0 10 1 0.1
@ 100C
Wicking g (body force) 3 2 10 >10
Thermal Expansion Mismatch*) % 10 10 5 3
Wick Thermal Conductivity W/m-K 23 100 - -
TGP Thermal Conductivity W/m-K 200 - 500 20.000
TGP Thickness Mm 2 2 3 <1
Area cm x cm 15 x 30 - 3x3 10 x 20
Weight Gm 270 - 20 50
Duration Hours >30.000 - 100 1000
*) CTE mismatch with respect to semiconductor devices
Table 1 DARPA Thermal Ground Plane objectives and phases [5]

Besides requirements on reliability and specific dimensions and weight, the relevant performance

metrics for the device are to reach effective axial thermal conductivity equivalent to 20,000 W/m-K in

a thin planar coefficient of thermal expansion (CTE)-matched package operating in a >10g’s

acceleration environment. The CTE-match is targeted relative to typical semiconductor materials

such as Si, SiGe, AlN and SiC.

The objective of the program is for the TGP to serve as an isothermal “Thermal Ground Plane” similar

to an electrical ground plane in common electronic printed circuit boards. The TGP is to spread heat

from electrical components with high heat flux input to heat rejection regions on the sides of the

system. The CTE-match with semiconductor devices objective aligns the system such that electronic

4
components can be mounted on the surface to the TGP with minimal bondline thickness, therefore

with minimal thermal interface resistance. An artist sketch of this concept is given in Figure 4. Figure

4 also illustrates suggested methods for enhancing evaporator and condenser performance using

tailored superhydrophobic and superhydrophilic structures.

Figure 4 Thermal Ground Plane Sketch

The final goal of the program is to demonstrate performance impact on a relevant system using

thermal ground plane prototypes.

Collaborators in this research include GE Global Research(GE), University of Cincinnati(UC), Air Force

Research Laboratory(AFRL) and GE Intelligent Platforms(GE IP).

1.3. Scope

The scope of this study includes high-level modeling of heat pipes and identification of parameters

critical to reaching the performance objectives as stated by Table 1. A semi-empirical approach is

presented in which historical results from the literature are compared to experimental studies on

structures developed under the TGP program. These empirical results are integrated into a high level-

model and are presented to evaluate performance trends for a TGP system. Experimental setups for

5
TGP benchmark and high-g performance evaluation are developed and results are presented. Finally

an overview of subsequent work and application relevance is presented.

6
2. Thermal Ground Plane Design Approach & Model

2.1. Introduction

In order to design a Thermal Ground Plane for high-g operation, the relevant physics are identified.

Significant simplifications are made in order to arrive at a design model which can be used to design

a Thermal Ground Plane for high-g operation that meets DARPA’s requirements.

2.2. Heat Pipe Operation

Heat pipes and/or vapor chambers are two-phase heat transfer devices that spread heat with

significantly greater effective axial thermal conductivity than solid conductors. Common features of a

heat pipe include; a substrate with a hollow cavity, a wicking structure lining the inside of this cavity

and a working fluid (Figure 5).

Evaporator Transport Condenser

Section (adiabatic) Section
Section

evaporation liquid condensation

Heat vapor
Heat
In Out
liquid

Heat pipe casing Wicking structure Vapor channel

Figure 5 Heat pipe operation

In a functional heat pipe the wicking structure is fully saturated with a working fluid while the

remainder of the hollow cavity, also known as the vapor space, contains the working fluid in the

vapor phase. As an ideal heat pipe is a closed volume filled with a single fluid, the fluid in the

saturated wick and in the vapor space will be at equilibrium with a saturation pressure roughly

corresponding to the average temperature of the device.

7
Figure 5 describes the working fluid transport inside the thermal ground plane. During operation,

when a small amount of heat is input into the device, the temperature of the evaporator region rises.

Heat from here is transported by two separate mechanisms towards the cooler areas of the device.

The first is transverse conduction through the wick and substrate and the second two-phase heat

transport evaporation and condensation of the working fluid. The contribution of heat transport to

each of these mechanisms can be estimated by evaluating dimensionless numbers such as the Biot

number, which will be discussed in more detail later in this chapter. The mechanism for two-phase

heat transport starts with evaporation of a portion of liquid from the wick. This occurs when the liquid

temperature exceeds the local required temperature gradient for evaporation (superheat) and a

nucleation site is present. The evaporation of liquid creates a local increase in vapor pressure. The

pressure gradient that is created drives the vapor from the heat input location to areas of lower

pressure, which are at lower temperature. In these areas the vapor rejects its heat by condensing to

its liquid state. When the mass flow in the system is known (which can be estimated from a known

heat input), the pressure difference from evaporator to condenser can be evaluated. As the vapor

pressure reduces, its temperature reduces as described by the Clausius-Clapeyron relation. In

practice however, this vapor temperature gradient is small. Vapor condenses in the condenser

region. The capillary structure of the wick captures the liquid that condenses. In the evaporator,

where liquid departs the wick, the capillary structure and fluid surface meniscus maintain a pressure

gradient driving liquid from the condenser back to the evaporator. From here the cycle is repeated.

The thermal resistance of a heat pipe is commonly dominated by the resistance from the outer

substrate to the vapor phase in the evaporator and the rejection of heat from the vapor phase to the

outer substrate in the condenser. In contrast, the effective thermal resistance of the vapor transport

from evaporator section to the condenser section of the device is typically minimal. This means that

a heat pipe, in contrast to a solid conductor, can be elongated without substantially increasing the

thermal resistance. As a result, the effectiveness of heat transport can be many times greater than

8
that of conduction through solid materials. This can be expressed as a high effective axial thermal

conductivity. Note also, that the opposite is also true. Short heat pipes or heat pipes with excessive

thick walls typically have relatively poor performance.

2.3. Heat Pipes, Thermosyphons and Vapor Chambers

The concept of a heat pipe was originally described in a patent by Perkins and Buck[5] at the end of

the 19th century. This device however, did not have a wicking structure and should technically be

classified as a thermosyphon. A thermosyphon is a device similar to a heat pipe but without a

wicking structure and therefore reliant on body forces such as gravity or other acceleration for liquid

to return to the evaporator section. For example in gravity operation, thermosyphons can only

operate when the condenser is located at a higher elevation than the evaporator. Body forces can

also be used to enhance liquid return such as in applications in electric motors or other rotating

machinery.

The introduction of the capillary wicking structure to assist liquid return to the evaporator was first

conceived by Gaugler of General Motors in 1944[7]. Work by Grover and Cotter[8] introduced the

name ”heat pipe” in 1964 and accelerated the development and application of heat pipes in space

and ground applications[9,10].

Vapor chambers have the same components as a heat pipe. However, they are flat two-dimensional

structures in configuration compared to a round circular heat pipe. They are also sometimes referred

to as “flat heat pipes”. The definition of a vapor chamber is not yet strictly defined and sometimes

include the application configuration where heat is input on one side and a rejected at larger location

on the other side of the device. This allows for efficient heat spreading from one or more heat

sources.

9
 Figure 6 Vapor chambers (Source: Advanced Cooling Technologies (ACT))

In this study, the Thermal Ground Plane is referred to as a variant of a vapor chamber, even though

it’s tested configuration differs from the definition above. As this research explores the fundamentals

of heat transport in two phase heat transfer devices under high g-force accelerations, other similar

heat pipe devices are presented as they might benefit from the presented work and will be referred

to in subsequent chapters. Similar two-phase heat transfer devices are loop heat pipes(LHPs) and

pulsating or oscillating heat pipes (PHPs or OHPs). The former describes a looped heat pipe in which

liquid transport resistance is reduced by limiting the application of the wicking structure to the

evaporator such that the transport region can achieve maximum transport with minimal liquid

resistance(Figure 7). The pulsating heat pipe describes a looped tube in serpentine configuration,

where vapor bubbles in the device heat up and cool down resulting in a semi-random fluid

movement in which heat is transferred from the evaporator to the condenser region(Figure 8).

10
 Figure 7 Loop Heat Pipe [11] Figure 8 Pulsating Heat Pipe [12]

In order to achieve the desired TGP objective of 20,000 W/m-K effective transverse thermal

conductivity at 20 g, the relationships between design parameters and effective performance needs

to be identified. Therefore a high level thermal and hydraulic model of a thermal ground plane is

developed.

2.4. Thermal Ground Plane Transport Model

Vapor chamber operation can be described at the most basic level by hydraulic transport and

thermal resistance network models. As the transport of the working fluid is the fundamental driving

principle behind vapor chamber operation, this will be described first, followed by the effective

thermal transport efficiency of such systems. The transport of the working fluid by the device

ultimately determines how the device operates in high g-force environments.

2.4.1. Working Fluid Mass Transport Rate

The mass flow rate of a vapor chamber is ideally governed solely by the amount of heat input into

the device. In general heat is input in one location, which will be referred to as the evaporator, while

heat is removed at another location, which will be referred to as the condenser (see Figure 5). This

11
assumes that heat rejection in other areas, such as the adiabatic region, is negligible. To satisfy the

energy balance, heat has to be removed either by evaporation of fluid at the meniscus interface or

by conduction (diffusion) through the substrate, wick structure and liquid towards the heat sink as

indicated by Eq. 1.

Eq. 1

For a typical heat pipe application, heat transfer is dominated by , meaning that most heat is

removed by two phase heat transfer and that diffusion heat transport is small. For short heat pipes,

heat pipes with thick substrates or wicks, or if the substrate or wicking materials have high thermal

conductivity heat transfer by conduction or phonon diffusion, , can be significant.

In the simplified model it is assumed that diffusion heat transport is negligible, therefore the fluid

mass flow ( ̇ ) is at maximum for the given heat input and can be defined by the ratio of total heat

input ( ) and latent heat of evaporation ( ) as illustrated by equation 1.

̇ Eq. 2

The latent heat of evaporation ( ) is a strong function of temperature and can vary significantly

depending upon what mean vapor temperature is reached as illustrated by the enthalpy diagram for

water, a common vapor chamber working fluid, as illustrated by Figure 9.

12
 SteamIAPWS
500

400

300
T [°C]

200 1000000 Pa

hfg
100 100000 Pa
50000 Pa
25000 Pa

0.2 0.4 0.6 0.8

0
0.0x100 1.0x106 2.0x106 3.0x106 4.0x106
h [J/kg]

Figure 9 Enthalpy diagram for water

The Jakob number, which describes the ratio of sensible versus latent heat, is an important

dimensionless metric in the evaluation of working fluid effectiveness.

Eq. 3

Typical Jakob numbers for a water based heat pipe are on the order of 0.008 (liquid properties at 55

°C, evaporator-condenser temperature gradient ΔT~10 °C). A Jakob number of 0.008 illustrates that

the latent heat of the working fluid is high with respect to a typical sensible rise. This indicates that

the working fluid can absorb a significant amount of heat by phase change per unit mass flow. The

Jakob number will reduce towards the triple point as the latent heat ( ) reduces.

2.4.2. Criterion for Working Fluid Transport

In steady state operation, a transport network can be used to describe fluid motion from the

condenser to the evaporator and vapor motion from the evaporator to the condenser. Figure 10

13
illustrates such a network by use of a simplified vapor chamber model. Figure 10 shows the cross-

section of a one-sided wick device with one-sided heat input and output boundary conditions. This is

the configuration that is used for qualification of Thermal Ground Plane performance. If the device is

of sufficient width, the geometry effectively allows for reduction to a simplified transport model. This

is described as phase change transport from 1 to 2, vapor transport from 2 to 3, phase change

transport from 3 to 4 and liquid transport from 4 to 1. Note that this simplification suggests that

transverse substrate and wick thermal transport can be neglected which needs to be verified by

evaluation of the Biot number.

Figure 10 Simplified vapor chamber with one sided wick

As liquid and vapor are transported, pressure gradients exist within the network. The liquid transport

pressure loss can be subdivided into a part relating the viscous losses of moving a liquid from

condenser to evaporator (losses related to moving liquid through porous wick and sheer losses at

substrate and vapor space surfaces) ( ) and into a part relating the total head required to move

a liquid upon which body forces ( ) are exerted at angle( ) as shown in Figure 10. In

steady state, the capillary pressure gradient across the wick meniscus ( ) is equal to the sum of

14
total pressure losses in the system such that fluid transport can be sustained and the device can

transport heat (Eq. 4).

⃗⃗⃗⃗⃗ Eq. 4

This assumes:

 Steady state operation

 Liquid density significantly greater than vapor density such that body forces acting on the

vapor can be neglected, which typically is true except near the triple-point. ( ⁄ )

 Viscous losses in liquid are caused by shear friction with porous wicking structure, losses

caused by vapor sheer (no slip)

 Flows are fully developed

 Evaporation, condensation and entrainment effects in the adiabatic region are neglected

(no blowing or suction)

The system pressure losses can be subdivided into vapor transport pressure losses ( ), moving

vapor from location 2 to location 3 and liquid transport pressure losses ( ⃗⃗⃗⃗⃗ ),

moving liquid from location 4 to location 1, as defined by Figure 10. The capillary pressure head

( ) is commonly only evaluated at the evaporator as the pressure gradient in the condenser

region is minimal, resulting in the vapor pressure at location 3 to be approximately equal to the liquid

pressure at location 4.

15
 2.4.3. Capillary Pressure Rise

The capillary pressure rise describes the pressure gradient between locations 1 and 2 as sketched in

Figure 10. This pressure gradient is required in order to sustain fluid transport and is provided by the

surface tension at the meniscus of the fluid in the evaporator region.

Figure 11 Fluid meniscus sketch for sintered sphere wicking structure

The surface that describes the intersection between vapor and liquid phases upon a wicking

structure is commonly referred to as the contact line. A force balance along the contact line

integrating the liquid-solid, liquid-vapor and vapor-solid surface tensions can be used to evaluate the

contact angle( ) between the fluid and the solid surface at the interface as demonstrated by

Neumann’s formula or Young’s equation [13, 14].

Eq. 5

As liquid evaporates, the meniscus has a tendency to recede into the wicking structure which is

counteracted by a liquid “pull” by the surface tension forces to maintain contact angle as defined by

Eq. 5. Similarly at the condenser end, liquid is added to the meniscus causing the meniscus to expand

tending towards a flat interface. The difference in meniscus angle between evaporator and

16
condenser creates an effective liquid “pull” whose magnitude can be evaluated from the surface

tension( ), contact angle( ) and wick structure pore radius( ). The wick structure pore radius is a

non-physical parameter that describes the effective pore radius of a cylindrical capillary structure. As

typical wicking structures, such as sintered sphere wicks, have complex three-dimensional

geometries, an experimentally evaluated effective pore radius is typically used. The understanding of

the capillary pressure head ( ) provided by this pull is essential as it is the driving force behind

liquid transport in the system and the fundamental working principle of a vapor chamber.

Eq. 6

In addition to Young-Laplace’s formula, the Kelvin equation is often used to describe the inertial

pressure gradient that can occur at the evaporation interface[14]. As these effects are typically

significantly smaller than the pressure effects described by Eq. 6, they will be neglected in this study.

2.4.4. Vapor Transport Pressure Drop

The vapor transport pressure drop describes the viscous pressure losses as fluid interacts with the

walls and wicking structure of a vapor chamber while moving from evaporator to condenser, from

location 2 to location 3 as sketched in Figure 10. As vapor is released from the meniscus surface in a

direction perpendicular to the meniscus surface, the flow has to turn and align with the vapor space

to move in the direction of the condenser region. In general the direction change that occurs in the

evaporator region is of little impact to the overall vapor pressure drop and commonly neglected. This

allows for the flow in the adiabatic section to be approximated by steady one-dimensional flow. The

one-dimensional steady state Navier-Stokes equations describe the viscous losses in the system.

Darcy’s law describes how the mean flow velocity is proportional to the pressure gradient. The

Darcy-Weisbach equation is a simplified form derived from the Navier-stokes relation which offers a

17
friction coefficient depending upon the character of the flow field. Using these equations a simplified

formula for the vapor pressure gradient( ) for fully developed laminar flow can be derived.

Darcy-Weisbach Eq. 7

Friction coefficient (Re<2300, Mav<0.2) Eq. 8

Geometric parameters are introduced to characterize the vapor transport channel. These are the

hydraulic diameter of the vapor space( ), cross section area( ) and mean vapor channel

length( ). The friction factor coefficient( ) varies depending on cross section and is typically 64 for

circular geometry and in the range of 57 to 96 for rectangular geometry as can be observed in

Appendix A[15]. Given the TGP vapor space cross section dimension of approximately 2 mm x 26

mm, the TGP has an aspect ratio of ~13. Since this is out of the range of data provided by Appendix A,

an extrapolation is required.

Extrapolated

Appendix A

Figure 12 Fricton factor coefficient and regression function

18
Figure 12 shows a logarithmic extrapolation (R2~99%) of the factor. For an aspect ratio of 13 this

correlation gives a value of approximately 88. Therefore it is recommend to use a factor of 88 for

evaluation of the vapor space pressure drop.

The hydraulic diameter can be defined using the cross section area and the wetted perimeter( ).

Vapor channel hydraulic diameter Eq. 9

The mean vapor channel length( ) is defined as the mean distance the vapor particles have

travelled from evaporator to condenser. For a typical device this is approximately half the length of

the evaporator and condenser sections plus the adiabatic section length. These equations can be

combined to give a formulation for the vapor pressure gradient as a function of vapor space

geometry, fluid properties vapor density( ) viscosity( ) and the mass flow( ̇ ) as evaluated by Error!

eference source not found..

̇
Vapor pressure drop (Re<2300, Mav<0.2) Eq. 10

Note that typical vapor chambers used for electronics cooling applications are designed for Mach

numbers less than 0.2 as to avoid compressibility effects and potential choking of the flow. Devices

with higher operating temperature ranges, such as sodium heat pipes, often do operate at Mach

numbers closer to the sonic limit[16,17].

2.4.5. Liquid Transport Pressure Drop

As liquid is transported through the wicking structure from the condenser (location 4) to the

evaporator (location 3) the viscous pressure losses can be significant. In typical vapor chambers the

liquid transport pressure drop dominates the transport model. The liquid pressure drop, excluding the

body force related pressures (which will be discussed in detail in the next paragraph), can be defined

19
similarly to Error! Reference source not found. by assuming steady-state, laminar, one-dimensional

low. This results in an equation which is a function of the liquid density( ) and viscosity( ), mass flow

( ̇ ) as well as geometric parameters such as the wick cross sectional area( ) the mean liquid

transport distance( ) and the wick permeability factor( ).

̇
Eq. 11

The wick structure is commonly a porous media such as a series of parallel channels or a sintered

particle structure. The permeability of the porous media is commonly a function of geometric

parameters effective sphere radius( ) and porosity( ). The first describes the qualitative dimension of

the average flow passage, the second the quantity of these flow passages with respect to the overall

cross section. These parameters can be combined in a single parameter that describes the pressure

required to move a given amount of fluid through a porous media, the permeability factor( ).

Although this parameter will be experimentally derived later, several models exist that can be used

as approximation. Blake-Kozeny[18] suggested a commonly used relation between geometric

parameters such as porosity( ) and sintered particle radius( ) and permeability for sintered wick

structures as described by Eq. 12.

Eq. 12

Ababneh[19] illustrated the shortcomings of this correlation and suggested use of the Hagen-

Poiseuille relation with a modified constant of 1.7 (Eq. 13) with improved success.

Eq. 13

20
Figure 13 illustrates these permeability models for a range of pore radii at constant porosity ( = 0.5).

Figure 13 Permeability constant models ( = 0.5) (adapted from Ababneh [19])

Chi refers to the Blake-Kozeny relation(Eq. 12)

Given the permeability factor, either derived from experimental data or established by Eq. 13, the

viscous liquid pressure drop at a given mass flow rate can be calculated.

2.4.6. Pressure Drop due to Body Forces

Body forces acting upon the fluid inside the vapor chamber can either assist or hinder fluid

movement inside the device. Since density of the vapor is significantly lower than for the liquid state,

body force action on the vapor transport is neglected. The body force contribution to the pressure

loss is a function of the angle( ) with respect to the liquid transport direction as sketched in Figure

10.

Eq. 14

21
If the body forces such as gravity or other acceleration forces align with the working fluid motion

( ), the sign of the body forces pressure difference will be negative, assisting in the fluid motion,

while if the body forces oppose the working fluid motion, the sign will be positive. Most heat pipes or

vapor chambers are very sensitive to external body forces and experience reduced performance

under body forces that oppose the flow of the liquid.

As both Eq. 11 and Eq. 14 describe a liquid pressure gradient component across the fluid interface

that varies linearly with the axial distance along the length of a vapor chamber from condenser to

evaporator, the meniscus profile can be anticipated to vary similarly in linear fashion along this

length. Figure 14 illustrates this in simplified fashion between ~0 Pa in the condenser to 3.2 kPa in the

evaporator using water as working fluid. The contact angle and respective meniscus shape are

calculated for each pressure condition and sketched between spherical particles.

Figure 14 Two-dimensional meniscus shape variation simulation along the length of a vapor chamber
from 0 Pa in the condenser to 3.2 kPa in the evaporator using water as working fluid

Note that Figure 14 represents a two-dimensional representation while in reality particles are

spherical in nature providing a three-dimensional meniscus shape. From Figure 14 it can be clearly

observed that although the pressure gradient varies linearly, the meniscus surface area reduces

more rapidly. This has a significant impact on the wetted area of the sphere as well as the film

thickness. This particular aspect will be revisited when analyzing the heat transfer performance as a

function of meniscus shape.

The combination of the pressure drop terms forms a simplified hydraulic model that is the basis for

operation of the vapor chamber operation. For a typical application the input heat flux drives the

22
mass transport rate while the condenser wall temperature acts as the reference temperature of the

system.

2.5. Heat Transfer Model

The heat transfer performance of a vapor chamber can be described in a simplified fashion by the

total thermal resistance of the device as used in a specific configuration. This total thermal resistance

can be broken down in a number of component resistances. The approach taken describes the

thermal network in terms of effective thermal resistances and then uses these to evaluate an

effective axial thermal conductivity. As the resistances in the network are highly coupled, it becomes

imperative to understand the dependencies of each of these elements. An overview of a detailed

resistance model is sketched in Figure 15. Relevant physics in a TGP include conduction thermal

resistances through substrate walls, interface thermal resistances, wick and liquid thermal

resistances as well as an effective thermal resistance to describe the temperature gradient in the

vapor space. Typical vapor chamber/heat pipe models neglect axial diffusion heat transfer from

evaporator to condenser through the substrate and wicking structure as for long devices, this

resistance is large compared to the parallel path provided by the two-phase heat transfer

mechanism. For short vapor chambers, or for vapor chambers that have substrate and wick made

out of high material thermal conductivity, the diffusion path will be important and might even

dominate. Technically one could argue that at the evaporator the energy is absorbed by an energy

sink and released at the condenser. As the magnitude of the sink and source will be approximately

the same (the vapor cannot endlessly absorb or reject heat and has to reach equilibrium state), the

assumption of using an equivalent resistance for the vapor space is reasonable as it describes a

temperature gradient related to an amount of heat transferred.

23
 Figure 15 Detailed TGP thermal resistance diagram

Due to the complexity of the structure, the exact breakdown in the vertical path of substrate,

interface, wick and liquid thin film thermal resistance is challenging to evaluate by experiments or

numerical simulation. Therefore a simplification is proposed to take the sum of these resistances as a

simplification of the model. This greatly simplified 1D heat transfer model is proposed to deliver an

estimate of TGP heat transfer performance in an ideal case where heat losses to the ambient are

neglected (Figure 16).

24
 Figure 16 Simplified TGP Thermal resistance diagram

This simplified model assumes that all heat is transferred by vapor transport alone and that heat

conduction through the vapor chamber wick and wall are negligible. This assumption is challenging

for short devices with high thermal conductivity wall and wick materials that were tested but more

reasonable for long devices with thin side walls. However, this simplification allows for separation of

the heat transfer path in five distinct sections from station 1 to 6 as sketched in Figure 16. The

distinct segments of this heat transfer path are the heat transfer from outer to inner wall of the

substrate ( ), heat transport through the wick and to the vapor state ( ), temperature gradient

due to transport in the vapor space ( ), heat transport from the vapor state to the inner wall of

the condenser ( ) and heat conduction from the inner to the outer wall of the condenser ( ).

These elements will be discussed separately. It is clear that this is a simplification of a complicated

geometry. For instance ( ), could be broken down into separate resistances for substrate,

interface, wick structure and liquid layer. However, in order to give the first description of the heat

pipe thermal path, this is simplified for now. The combined simplified total thermal resistance is

defined by Eq. 15.

Eq. 15

25
 2.5.1. Substrate conduction thermal resistance

Assuming uniform heat input and output flux, steady state without heat generation and isotropic

material properties, the heat transport through the substrate in the evaporator and condenser

regions can be approximated by the Fourier equation (Eq. 16).

( ) ( ) ( ) ̇ Eq. 16

Assuming 1-dimensional heat transfer, isotropic material properties, no heat generation and steady

state (Eq. 17).

( ) Eq. 17

Integration with heat flux boundary condition gives (Eq. 18).

Eq. 18

Which can be expressed in a thermal resistance ( ) by multiplication with the mean cross section

area ( ) (Eq. 19)

Eq. 19

The thermal resistance for the evaporation and condensation regions of the chamber(Eq. 20, Eq. 21)

are evaluated using the substrate thickness and the product of the respective effective evaporator

and condenser cross section areas with the substrate thermal conductivity. The term of “effective”

areas is used here to indicate that the evaporator and condenser areas are not limited to the heater

26
and cold plate dimensions. Due to conduction in the substrate and wick, these areas are effectively

enlarged.

Eq. 20

Eq. 21

The simplified model applicability is limited to cases with uniform heat flux input in the evaporator

and output in the condenser. In a standard chip package, heat generating components are

distributed in the silicon die. The silicon die is then packaged with a heat spreading lid that assists in

protecting the die as well as spreading the heat to generate a more uniform profile. This is then

attached to a heat spreader which can be a vapor chamber or heat pipe that distributes the heat to

the convective surface area. Arguably, the assumption of uniform heat absorption in the evaporator

is only valid when the wick is fully saturated (no on-set of dryout). When dry-out starts to occur, it

usually appears at the end of vapor chamber near the wall. This as the liquid feed mechanism is

unable to drive enough liquid to the far end of the evaporator. Naturally when partial or full dry-out

occurs, the heat flux input cannot be assumed to be uniform.

In order to use a uniform heat flux assumption in the condenser region, the two phase heat transfer

mechanism of condensation must take place at a similar rate in the condenser region. This is only

applicable if liquid transport in the condenser is not impeded allowing for a uniformly wetted profile

with equal liquid removal throughout the condenser space. As liquid near the far edge of the vapor

chamber has to transfer a longer path through the condenser to the adiabatic region than liquid that

condenses on the edge of the adiabatic region, it might be more challenging to maintain uniform

condensation. However, as the developed model aims to evaluate vapor chamber trends using

operation of heat loads that are significantly below dryout, the assumptions are used as they capture

the relevant physics in a simplified manner.

27
 2.5.2. Wick/Evaporation resistance

The thermal resistance from point 2 to point 3, as sketched in Figure 17, describes the temperature

gradient from the substrate inner wall to the vapor in the evaporator. Although a representation by a

single thermal resistance is suggested, heat travels over a number of interfaces and layers in this

area, involving different physics. Modeling the conduction of heat through the wick structure and

evaporation of liquid from the wick to vapor in the vapor space as a summed total thermal

resistance is a grand, but not unprecedented[14,18], simplification. The linear relationship, as

suggested by a thermal resistance formulation, only holds within a select window of operating

conditions that should be empirically evaluated. A number of elements that are part of this thermal

conduction path are discussed. First, coating layers for adhesion or chemical stability can be present

and technically have an impact. However, these layers are in general thin, i.e. on the order of

microns, and therefore their contribution to the thermal resistance is usually practically negligible.

Figure 17 Wick/Evaporation resistance elements for a sintered spherical particle wicking structure

From the surface of the interface layer, heat travels through a number of parallel paths through the

wicking structure to the liquid/vapor surface. Elements of this parallel thermal network include

conduction through the spherical particles, particle to particle interface conduction and conduction

though the working fluid (as illustrated in Figure 17. Typical materials for wicking structures are

metals, such as copper or aluminum. Typical working fluids for devices that are to operate around

28
room temperature are water or ammonia. This means that wicking material thermal conductivity

properties are typically significantly greater than the working fluid. Hence, conduction through the

submerged wicking structure is typically dominated by the wick material. However, for thin wicks the

liquid thermal conductivity can also be dominant. Maxwell[20] developed an empirical relation for an

effective thermal conductivity of sintered metal powered wicks( ) as a function of solid and

liquid thermal conductivity( and ) and porosity( ) with good success as confirmed by Kozai[21].

( )
( ) Eq. 22
( )

Eq. 22 illustrates that as the solid thermal conductivity is significantly greater than the liquid thermal

conductivity or if the porosity nears zero, the effective thermal conductivity tends towards the solid

thermal conductivity. Reciprocally, if the porosity is near unity, the effective thermal conductivity

tends towards the thermal conductivity of the working fluid.

Heat conduction resistance through the sintered sphere wick is small compared to the surface heat

transport resistance. An evaluation of the Biot number formulation (Eq. 36) gives that for anticipated

effective thermal conductivity values of around 100 W/m-K (copper wick), and wick thicknesses of

200 to 500 µm, convective heat transfer coefficients would have to be in excess of 20.000 W/m 2-K in

order for the Biot number to be greater than 0.1.

Eq. 23

At Biot values of less than 0.1, it is acceptable to treat the wicking structure as being at uniform

temperature. Although the evaporation heat transfer coefficient in the thin film region can be high,

the bulk of the wick will observe heat transfer coefficients that are significantly lower. This suggests

29
that the wick heat transfer up to the thin film region can be assumed at little thermal gradient. Due to

the high thermal conductivity wicking material and thin dimensions in the TGP vapor chamber, the

heat transfer is more interface than diffusion driven.

The top layer of the wick spheres can be exposed to the vapor as the area between the spherical

particles is connected by the liquid meniscus at the fluid/vapor interface as sketched in Figure 11. A

variety of mechanisms are present. First, heat can conduct from the exposed dry part of the

spherical wick directly towards the vapor. Secondly heat can conduct through the liquid towards the

liquid vapor interface. Evaporation occurs at the liquid vapor interface and heat is absorbed as the

liquid changes phase.

Thin Film
Region

Thin Film
Region

Figure 18 Heat transfer at the fluid vapor interface

If heat were to pass through the liquid to the liquid/vapor interface, it would encounter a significant

thermal resistance due to the poor thermal conductivity of the liquid film. However, the liquid layer

thickness is not constant. The contact angle between the solid spherical particles and the liquid

drives the meniscus shape to zero thickness at its ends. This adsorbed film region for very thin fluid

layers is called the thin-film region. As this region has the benefit of heat being removed by two

phase evaporation with a minimal penalty of heat having to transfer through the liquid the heat

transfer coefficients are high. The thin-film region is one of the critical regions for performance of a

vapor chamber. The fully submerged section of the particle is anticipated to experience lower heat

transfer coefficients similar to forced convection around a submerged particle.

30
As the liquid resistance in the thin film region is the lowest and the heat transfer coefficients are th

highest, it is assumed that most of the heat is transferred through this area. Due to the three-

dimensional nature of a sintered wick, the area of the thin-film region depends on the meniscus

shape and on the pressure difference between the liquid and the vapor state forming the meniscus.

This implies that the thermal resistance between the wick structure and the vapor state varies with

local pressure gradient. As illustrated by Figure 14, the meniscus shape changes along the length of

a vapor chamber and will also change depending on the pressure gradient that is established due to

the requirement of attaining a given mass transport rate.

Different research groups have aspired to provide detailed analytical and numerical models related

to the heat transfer mechanisms from wicking structures through the thin film region. Ranjan, Murthy

and Garimella first explored this in detail using geometric analysis[22] followed by a more detailed

numerical analysis by Ranjan et al.[23]. Although these studies help greatly to understand the

relevant physics and their relationships, the accuracy of such models compared to actually

fabricated structures is hard to validate due to the complexity and irregularity of real wicking

structures at these scales.

Because of the lack of validation of numerical models, the complexity of the heat transfer

mechanisms in this area and the geometric uncertainty due to imperfect manufacturing processes,

this study will take an empirical approach to evaluate the combined thermal resistance from the

inner substrate surface out to the vapor state in the evaporator region.

2.5.3. Vapor Transport Resistance

As vapor is generated at the liquid/vapor interface in the evaporator, local vapor pressure increases

relative to other areas of the vapor chamber. This mechanism drives vapor through the system to

areas of lower pressure as the vapor pressure tries to achieve equilibrium. The change in pressure

between the evaporator region and the condenser region can also be described by a change in

31
temperature. The magnitude of a temperature gradient corresponding to a given pressure gradient

depends on the slope of the saturation curve as given by the Clausius-Clapeyron equation at the

approximate mean temperature of the vapor(Figure 19).

Figure 19 Normalized saturation curve for water

In addition to the Clausius-Clapeyron relation, the amount the temperature gradient also depends on

the amount of superheat required for the working fluid to change phase in the evaporator and the

amount of subcooling required in the condenser. The greater the temperature gradient, the greater

the driving force for the vapor to move from evaporator to condenser. If superheat and sub cooling

temperature gradients are assumed zero, the temperature gradient of the vapor relates to the

pressure gradient as described by Eq. 24.

( ) Eq. 24

Substitution of Eq. 2 and Error! Reference source not found. gives:

32
 ( ) Eq. 25

Commonly the vapor flow pressure drop relative to the amount of heat transported is low and this

effective thermal resistance is neglected, however Eq. 25 illustrates that this can be a factor of

importance if the vapor chamber operates near the critical point of the fluid.

2.5.4. Wick Condensation Resistance

In the condenser region a variety of physical mechanisms occur to reject the heat from the vapor to

the substrate wall similar to the evaporator. Condensation resistance is similarly complicated to

evaporation resistance. As vapor cools down when it comes in contact with a cooler wick surface or

meniscus interface, it can condense if a nucleation site is present. At this location, heat is then

rejected to the wicking structure as the vapor changes phase to its liquid state. The heat then

conducts through a series of parallel paths to the substrate inner surface similar to the evaporator.

Although the meniscus interface is anticipated to be more clearly defined as flat, given there is

minimal pressure difference between vapor and liquid in this region, the heat rejection path is

challenging to predict using analytical or numerical models. Therefore, the effective thermal

resistance in the condenser will be evaluated empirically.

2.6. Combined Transport Model and Pressure Diagram

A typical boundary condition for a vapor chamber is one where the condenser temperature is known

due to the knowledge of the reference temperature (the temperature of the medium to which heat is

rejected) and the heat input is known due to the need for heat dissipation of an electrical device.

Assuming zero heat is lost to the environment, in steady state, the heat rejection in the condenser

must equal the heat input in the evaporator. The thermal resistance in the condenser can now be

used to evaluate the approximate mean temperature of the vapor in the condenser region. Using the

saturation curve (Figure 19) the pressure in this region can be evaluated.

33
 Figure 20. Modeled pressure diagram for water based 3 cm TGP vapor chamber

The required mass transport rate to dissipate the input heat load can be evaluated using (Eq. 2). This

rate can then be combined with geometric input and body forces to find the pressure gradients in

the vapor space and wicking structure (Error! Reference source not found., Eq. 11 & Eq. 14). The

otal pressure drop drives the meniscus shape in the evaporator. If the capillary pressure drop

associated with this meniscus shape is less than the maximum capillary pressure gradient the vapor

chamber can sustain operation (Eq. 4). If the required pressure drop is greater than the maximum

capillary pressure drop, the mass transport rate cannot be sustained and the evaporator will partially

or fully dryout resulting in a larger temperature gradient until a new equilibrium is reached (a

situation can be envisioned where dryout of the evaporator leads to increase in the significance of

substrate conduction). The pressures evaluated at each location can be combined in a pressure

diagram which can be used to evaluate device performance at a given mass transport rate (Figure

20).

34
 2.7. TGP Vapor Chamber Design for High-g Operation

It is the intent to design a TGP vapor chamber for use in high-g environments. De Bock et al[24]

presented an approach to identify the feasible design space for high-g operation. For high-g

operation body forces acting upon the fluid can result in a lower mass flow rate throughout the

system to a point where the mass flow rate is effectively zero and the system no longer functions. At

this limiting condition, mass flow rate is zero, therefore liquid and vapor pressure drops are also zero

as hydrodynamic pressure and capillary pressure are in balance. Re-arrangement of Eq. 6 and Eq. 14

and introduction of a g-force multiplier ( ) can give this limiting condition:

Eq. 26

Eq. 26 demonstrates the reciprocal relationship between pore radius and device length. Chi [5] has

given an approximate relationship between pore radius( ) and sintered wick particle radius( ):

Eq. 27

Combination of Eq. 26 with Eq. 27 gives:

Eq. 28

Eq. 28 defines the boundary of feasible design space. This boundary is illustrated in Figure 21 for a

device using water as working fluid assuming zero contact angle.

35
 Figure 21. Design Space as a function of body force multiplier for water at zero contact angle
(adapted from de Bock et al. [24])

A zero contact angle between water and a copper wick structure can be achieved by making the

structure super hydrophilic using nano features as proven by Tadanaga et al.[25] and Chio et al.[26]

and was reproduced by GE as presented by de Bock et al.[27]. Eq. 28 gives an upper bound to the

product of particle diameter and length using Chi’s correlation, such that for a given particle

diameter and maximum length, the potential operational space for the device can be observed as a

function of g-force multiplier ( ). It must be noted that Eq. 28, as presented by Figure 21, depicts the

limiting case where capillary and hydrostatic pressures are in balance but at zero mass flow. In order

to provide a practically useful design, it is recommend to design well within the region marked as

“feasible design space” in Figure 21 as to have sufficient capillary pressure to not only withstand the

hydrostatic pressure, but also the hydrodynamic pressure as liquid is transported by an operating

device with effective heat transport.

Combing the provided relations of the Thermal Ground Plane transport model as provided by

paragraph 2.4, allows for TGP prototype design and evaluation of heat transport performance at

36
different g-forces. The equations from paragraph 2.4 were evaluated in Engineering Equation Solver

(EES). This model is available in Appendix B. To solve the system of equations, geometric parameters

must be chosen. Relevant geometry parameters for a typical TGP design are presented in Table 2.

Design parameter Value

Length 3E-2 [m]

Wick porosity 50 [%]
Mean particle diameter 75 [m]
Vapor space cross section area 26E-5 [m2]
Wick cross section area 13E-5 [m2]
Contact angle 0 [deg]
Table 2 Relevant geometry parameters for typical device

A series of results from the heat transport model as a function of wick particle diameter are given in

Figure 22.

100
d_part=241E-6m
loss in heat transport

d_part=100E-6m
80
capability [%]

d_part=50E-6m
d_part=10E-6m
60

40

20

0
0 2 4 6 8 10
g-force factor mp [-]

Figure 22. Normalized loss in heat transport capability as a function of g-force multiplier for a 3cm TGP
using water as working fluid, assuming zero contact angle

Figure 22 describes the loss in heat transport capability as a function of g-force multiplication factor

for a variety of sintered wick particle diameters as evaluated by the model as presented in appendix

B. As the particle size reduces, wicks have a stronger capillary performance, hence are more

37
insensitive to changes in g-force. Where a 100 µm particle sintered wick structure is anticipated to

lose 40% of it’s heat transport capability at 10 g’s, a 10 µm particle sintered wick structure will

experience a loss of heat transport capability of less than 3% at 10 g’s. At the limiting case of 241 m

particle diameter, the heat transport capacity reduces to zero at a g-force multiplier of 10. For a

practical device, a desired heat transport capability at the high-g force condition must be specified

such that the right particle diameter can be chosen.

The challenge in the design of a TGP is to balance the need to have minimal heat transport capability

at high-g operation (capillary pressure high, wick particle diameter small), while designing for

minimal hydrodynamic losses in the wick (friction factor small, wick particle diameter large). Careful

balancing of these design parameters is required to achieve an optimal design.

Given the geometry parameters from Table 2 and Eq. 28, it can be shown that if it is desired to have

heat transport capability in excess of 100 W at 10g operation, the minimum wick particle diameter

should be approximately ~ 75 m.

Table 3 TGP Transport model results for 10 G design

The results of this evaluation using the TGP transport model in Appendix B are presented in Table 3.

38
 2.8. Effective Thermal Conductivity

Effective axial thermal conductivity is a metric that can be used to describe the performance of a

heat pipe or vapor chamber. In this study the effective thermal conductivity is defined per agreement

from DARPA, as:

“The thermal conductivity a virtual material needs to have

in order to have a similar thermal resistance to the vapor chamber”

Note that this definition of effective axial conductivity is not an intrinsic material or device property

but a metric that depends on the dimension of the device and the application boundary conditions.

This poses a number of challenges. First, an effective axial thermal conductivity is not easily

measured due to the multi-dimensional challenge of a typical problem, this leads to the required use

of a shape factor to correct for non-one-dimensional heat transfer. Secondly, effective axial thermal

conductivity is derived by measuring the effective thermal resistance and correcting for device

geometry and shape factor. This leads to the situation where boundary conditions and geometry can

have an impact on the effective thermal conductivity without an actual change in thermal resistance.

Last, effective thermal conductivity can easily be mistakenly interpreted as a material property.

The classical 1D conduction equation with shape factor correction( ) can be observed in Eq. 29.

Eq. 29

Rearrangement can give a formulation of the effective thermal conductivity as a function of the

thermal resistance and geometry parameters.

Eq. 30

The shape factor( ) can be expanded by use of known or measurable geometry parameters.

39
 Eq. 31

Eq. 31 gives the relationship for effective thermal conductivity. It can be observed that effectively Eq.

31 reflects a modified one-dimensional steady state form of the Fourier equation where a correction

factor is applied to account for non-one-dimensional heat transfer as illustrated in Figure 23.

Figure 23. Anticipated Mean heat flux path for conduction through solid material (baseline)
with uniform in and out heat flux boundary conditions and adiabatic sidewalls

The dimensions of the evaporator and condenser areas and their contribution to the total thermal

resistance( ) introduces boundary condition sensitivity to Eq. 31. In addition the device mean

length( ) over cross section area( ) times the shape factor( ) introduces geometry

dependence. Ideally the length used would reflect the mean heat flux path from heat input to heat

output region while the cross section area would reflect a weighted mean cross section area of this

path. Effectively the shape correction factor corrects the actual (measurable) geometry to the

effective heat transfer path.

To interrogate the shape of the heat transfer path a finite element model was built in ANSYS. The

element mesh and the cross section temperature contour can be observed in Figure 24. Isotherms

40
can be observed that are sloped in the heat in and heat output regions and near vertical in the

adiabatic region.

Figure 24. Temperature and finite elements plots for evaluation of the effective thermal conductivity of a
solid material for shape factor evaluation (not a vapor chamber)

For a short vapor chamber the slope in turning of the heat path in the evaporator and condenser is

significant. For a long vapor chamber the effective contribution of the heat path turning is small and

the shape factor tends to unity.

In order to derive the shape factor correlation between shape and boundary conditions for the

baseline geometry as given in Figure 23, it is necessary to evaluate the thermal resistance of a solid

conductor with known thermal conductivity property for a number of geometries and boundary

conditions. Due to challenges performing such an evaluation with a finite element analysis model, a

coarse finite difference model was authored in Engineering Equation Solver (EES). Source code for

this model can be found in Appendix C.

The non-dimensional vapor chamber height, (or aspect ratio) is given by Eq. 32:

Eq. 32

The non-dimensional heat input length boundary conditions is given by Eq. 33:

Eq. 33

41
From the temperature result in each run, the mean total resistance is calculated (Eq. 34), which is

used to evaluate the effective length of the heat path.

Eq. 34

By dividing the effective length by the actual mean length between heat input and heat extraction

areas (as shown in Figure 23), the shape factor can be derived (Eq. 35).

Eq. 35

As for this geometry no available analytical equation could be found, a numerical design of

experiments was performed where input geometry and boundary conditions were varied for a total

of 36 runs. Input range values for varied from to . Input range values for varied from to .

Results from this design of experiments can be found in Appendix D. From data interrogation, it was

found that a correlation could be derived between the dimensionless aspect ratio, heat input area

and shape factor.

Eq. 36

A quality test showed excellent agreement (R2>99%) between the regressed effective length and the

effective length as evaluated from the finite difference model results. Note that this result was

achieved by eliminating 2 outliers from the results. With these outliers included the R 2 value drops to

96%, still an acceptable result, especially considering that no empirical constant was used in the

regression.

42
 Figure 25. Agreement between regressed and FDM evaluated effective length

Once a shape factor is known, the effective thermal conductivity can be evaluated from Eq. 31 once

the shape factor has been evaluated using Eq. 36.

43
 2.9. Need for empirical factors

Although a simplification, a vapor chamber can be effectively represented by the presented hydraulic

and thermal models. As wicking structures can exhibit a semi-random nature, and due to the fact

that micro- and nano-scale surface features can impact physical mechanisms such as permeability

and wetting, empirical evaluation of parameters that are connected to the wicking structures is

recommend. This requires for the development of experiments to determine the following properties:

Eq. 6 Capillary performance of the wicking structure

Eq. 12 Permeability constant of the wicking structure

/ Eq. 15 Thermal resistance of the evaporator and condenser sections

Eq. 22 Wick effective thermal conductivity

Table 4 Parameters for Empirical Evaluation

44
3. Experimental Evaluation of Thermal Ground Plane Empirical Factors

3.1. Introduction

In the previous chapter it was discussed that in order to correctly anticipate and design the thermal

performance of a TGP a number of empirical factors for device components are required. This

chapter introduces and discusses measurement methods in order to obtain these factors. The

respective factors are; capillary pressure, wick permeability, wick effective thermal conductivity and

evaporator/condenser thermal resistance.

3.2. Evaluation of Wick Capillary Performance

The capillary performance of the wick is one of the most important parameters in a heat pipe or

vapor chambers. Given the unique desired performance metrics of TGP, some unique challenges

were identified in establishing capillary performance. This section describes an overview of available

methods for capillary performance evaluation, description of a measurement setup and

experimental results.

3.2.1. Capillary Performance Measurement Methods

The wick structure is a critical component of the vapor chamber, both for its mass transport and heat

transfer characteristics[18,28]. Capillary forces in the wick structure drive the liquid transport that is

at the heart of the operation of the device. A method is necessary to characterize the capillary

performance of wick structures of micro- and nano- dimensions.

Several methods have been described in the literature to study the capillary performance of wick

structures[29] The simplest method is the rising meniscus method[30]. In a variant of this method, a

section of the wick is submerged in an open reservoir and the rise of liquid in the wick is measured.

The height of the liquid column is a metric for the capillary performance of the wick structure.

45
This height is a function of the equilibrium between the capillary ( , Eq. 6) and the hydrostatic

( ,Eq. 37) pressure at the liquid interface [14,29].

Eq. 37

The capillary height ( ) is expressed by Eq. 38.

Eq. 38

In a given gravitational environment, the liquid column height is related to geometric properties as

contact angle ( ) and effective pore radius ( ) as well as fluid properties surface tension ( ) and

density ( ).

10.00
liquid column height [m]

1.00

0.10

0.01
1.00E-03 1.00E-04 1.00E-05 1.00E-06 1.00E-07
effective pore radius [m]

water mercury silicon_oil

Water mercury silicon oil

 [N/m] 73E-3 500E-3 16E-3
 [kg/m3] 0.998E+3 13.5E+3 0.76E+3
  [Nm2/kg] 73.1E-6 37.0E-6 21.1E-6
Figure 26. Liquid column height for different fluids (properties at P=1[atm], T=293 [K])

Figure 26 gives the theoretical liquid column height for different fluids and wick effective pore radius

from Equation 3. It must be noted that this is an idealization as it is unrealistic to expect exact

46
uniform capillary performance throughout a wick. From Figure 26 it can be observed that the rising

meniscus method can be impractical for wick structures with micro- or nano-sized effective pore

radius, as column heights can easily exceed practical laboratory dimensions (i.e. > ~2 meters).

Adkins et al. developed a method called bubble point testing for measurement of the largest pore

size in a porous sample [31,32]. In this bubble point test, the wick structure is saturated with a fluid

and gas is forced through the wick by applying a pressure difference across the sample. If one side of

the sample is completely submerged, this pressure can be identified by the formation of gas bubbles,

giving the test its name. The effective pore radius is inversely proportional to the required pressure

difference at the appearance of the first bubbles, which can be derived from Eq. 39.

Eq. 39

The bubble point test idealizes the pore as a circular hole and measures the largest pore flow path.

This technique is only able to characterize the minimum capillary performance of a wick structure.

Due to variation in pore sizes, the maximum capillary performance can be substantially different [1].

However, for a complex wicking structure, it is practical to assume that the largest pore in the

structure will determine the limit of the capillary behavior. As the aforementioned rising meniscus

method is practically infeasible, the bubble point method provides a practical alternative. As the

wicking structures developed for the Thermal Ground Plane vapor chamber require micro-/nano-

pore radius, the bubble point method is used in this study.

3.2.2. Bubble Point Measurement Setup Description

The apparatus chosen for performing the bubble point testing is a PMI BPT-101-A bubble point tester

(Figure 27). The accuracy of the PMI tester is rated at 0.15%.

47
 Figure 27. Bubble point test machine. PMI BPT-1100-ABPT

This apparatus allows a sample to be placed between two O-rings. Subsequently the sample is

wetted with a testing fluid. A sketch of the test setup is presented in Figure 28.

Volume 1

Volume 2

Figure 28. Sketch of bubble point test fixture Figure 29. Bubble point pressure, time profile

The pressure of the volume on one side of the sample is increased in small steps. This pressure is

raised slowly as to minimize dynamic pressure head. The point at which air passes through the

sample can be identified by the inability to increase the pressure in volume 1 any further. Historically

this point was identified by the appearance of bubbles in Volume 2, hence lending the method its

48
name. The pressure difference across the sample at this point is the bubble point pressure and can

be related to the largest pore diameter using Eq. 39 (as sketched in Figure 29).

3.2.3. Experiment Validation Using Wire Mesh Structures

In order to validate the bubble point method and test setup, a structure with known pore diameter is

tested. For this validation test, wire meshes with known structure were acquired and tested using the

bubble point machine.

1 mm Calibration
sample 1

325 mesh
13 wires/mm

rpore~39 m

Figure 30. Image of 325 mesh

Samples of 325 and 150 screen mesh number woven stainless steel mesh were prepared for

calibration (Figure 30). The effective pore radius of pores in such a mesh is estimated using the

screen mesh number ( ) and Equation 5 given by Chi[18].

Eq. 40

Using Eq. 40, the 325 mesh and 150 mesh screen effective pore radii are estimated to be 39 and 84

m respectively.

A parameter required for the evaluation of the measured effective pore radius is the contact angle

between the sample material and the testing fluid. Contact angles between fluids and wire meshes

were measured (Figure 31) and data was corrected using the measured contact angles. Although the

49
contact angle for a low surface tension fluid like isopropyl alcohol (IPA) is near zero, the contact angle

of water was hard to measure but estimated to be between 30 and 40 degrees.

Figure 31. Measurement of contact angle between wire mesh and water

Meshes were tested using distilled water and IPA. Tests were repeated at least ten times with

different samples of the same mesh size. The data from these experiments can be found in Appendix

E.

Figure 32. Wire mesh bubble point validation test results (IPA)

Since there is greater uncertainty in post processing the water data due to the challenging contact

angle measurement, IPA data was used for post processing. It is expected that there is some error in

the calculated effective pore radius due to the transition from square pore geometry to a simplified

effective radius. In addition, the fact that the mesh is woven adds three-dimensional offset to one

50
half of the squares sides. Nevertheless, the test data shown in Figure 32 fell within one standard

deviation of the calculated estimate, confirming the ability to measure capillary structures with

effective pore radii in the sub 100 micron range.

3.2.4. Measuring Capillary Performance of Sintered Wicking Structures

Iverson[28] points out some of the challenges in characterizing the mean effective pore radius of

porous samples using the bubble point method. Due to the non-homogenous nature of porous

structures, the bubble point method is expected only to give information on the minimum capillary

performance of a structure.

In order to test samples of porous wick structures with small effective pore radii, a method was

developed for preparing the samples for the bubble point test. A challenge that had to be addressed

is that porous structures with micro- or nano-sized pores (Figure 33) can be extremely fragile and

therefore cannot be clamped into the test fixture as sketched in Figure 28. In addition, the bubble

point test method requires the structure to have sufficient strength to withstand the mechanical

stress due to the pressure difference across the sample.

Figure 33. SEM image of wick structure[27]

51
This challenge is addressed by use of an adapter plate with a smaller diameter. The sum of the

distributed load force on the sample scales inversely with area and therefore inversely with the

diameter squared. By reducing the opening diameter, the stresses on the sample can be greatly

reduced. The porous structure then sits as a “bridge” across this smaller diameter opening.

A four-step procedure is used to prepare the sample on the adapter disc (Figure 34).

1 2
disc with hole filled
centerhole with filler

3 4
wick placed wick sintering
on disc filler melts

Figure 34. Wick sample preparation method

First, a small center hole is drilled in the disc by laser. The diameter of this hole is chosen to be at

least one order of magnitude greater than the expected largest pore radius of the wick sample. The

center hole is then filled with wax filler. The wax plug facilitates the placement of the wick structure

on the disc and prevents wick particles from entering the hole. After the wick is placed on the disc,

the disc and wick are sintered together to eliminate leakage paths between the wick and disc. During

this process, the filler melts and the sample is ready for testing (Figure 35).

52
 Figure 35. Prepared sample, copper wick on copper adapter

An advantage of this method is that if the diameter of the clamp is greater than the wick sample, the

disc will be clamped between the o-rings, as illustrated by Figure 28, while the sample remains

uncompressed. One must however note that this method provides only a single point measurement

over the wick “bridge”. If large variations in performance are expected, it is possible to use multiple

holes in the disc for characterization.

A porous wick sample was prepared using copper particles of 75 m mean diameter (Figure 35). The

prepared wick sample was sintered to a larger support disc structure using the described method.

The bubble point test was performed using isopropyl alcohol (measured near zero contact angle) and

repeated six times (Figure 36). The experimental data for this experiment can be found in Appendix F.

53
 Figure 36. Effective pore radii results for copper wick sample using IPA

The mean measurement pore diameter of the sample was found to be 8.8 m. A modest deviation in

measurement results was observed. A mean pore diameter of 8.8 m for a sample with particle

diameter of 75 m is significantly smaller than initially expected based on work by Chi[18]. As

previously mentioned Chi observed a ratio of pore diameter to packed sphere wick particle diameter

of 41% (Eq. 27).

Taken into account that factors such as particle shape, necking between particles and porosity all

potentially affect the capillary performance of porous structures; a deviation from the simplified

correlation by Chi is not unexpected. Furthermore, although pore radii is interesting to evaluate from

a validation perspective, the more important metric to be used is the Bubble Point pressure of the

wicking structure. Figure 37 gives the measured effective Bubble Point pressure for a single copper

sintered wick sample measured six times.

54
 Figure 37. Bubble point pressure results for copper wick sample using IPA

From Figure 37 it can be evaluated that the average measured bubble point pressure is

approximately 22.4 Pa with a standard deviation of 2.8 Pa. The bubble point method was found to be

an effective method of evaluation wick capillary performance.

3.3. Permeability

Wick permeability is the factor that relates the particular mass flow rate through a structure to the

imposed pressure drop. This factor is important in establishment of the flow rates and pressures

within the vapor chamber. Typically the largest pressure drop in a vapor chamber is in the liquid

saturated wick between the evaporator and condenser.

3.3.1. Wick Permeability Formulation

The formulation of the permeability constant for one-dimensional flow can be derived from Eq. 11

and is presented in Eq. 41.

55
 ̇
Eq. 41

This relation describes that if a pressure gradient is applied across a planar wicking structure (Figure

38) of known geometry saturated with a known fluid, the permeability constant can be evaluated by

measurement of the mass flow rate and mean medium transfer length.

liquid liquid
wick ̇

Figure 38. Planar Wick Permeability configuration

Figure 38 illustrates an ideal, but technically challenging experiment. In order to measure the

permeability constant of a planar structure in-plane, sufficient sealing along the wick sides is

required. It is challenging to seal the wick tightly without damaging the fragile structure.

3.3.2. Experimental Evaluation of Wick Permeability

As the TGP geometry is of square geometry and development of a linear test requires the

development of a new experiment, an alternative test method is sought. In effect, the bubble point

measurement setup can be adapted for use as method for evaluating permeability by covering the

top of the wicking structure with a cover with a center hole. If the bottom is covered with a plate with

an open periphery (as illustrated in Figure 39), the effective fluid transfer path is now through the

wicking structure.

56
 Figure 39. Feser[33] and proposed experimental configuration for in-plane permeability measurement

A formulation is now required for the permeability constant of a structure with a center hole inlet and

outer periphery outlet. Feser et al.[33], described a similar method for measurement of in-planar gas

permeability through circular porous structures.

Figure 40. Radial flow permeability testing apparatus from Feser et al. [33]

57
In the experimental setup, as depicted in Figure 39 and Figure 40, flow enters the system around the

periphery of a circular disk. The medium then penetrates the porous sample to the center from which

it exits the system. For the circular geometry (outer radius( ), inner radius( )), Feser derived the

relationship between flowrate and pressure (for ) as given by Eq. 42.

̇ Eq. 42

This in turn results in a wick permeability constant formulation for circular geometry of:

̇
Eq. 43

Although this formulation will work for circular structures, the wick in the TGP is of square geometry

and a different formulation is required.

3.3.3. Discrete model for Square Shaped Wick Permeability Experiment

As the requirement for DARPA is to evaluate a vapor chamber of rectangular geometry, wick samples

of square dimensions were available for evaluation. Figure 39 illustrates that the formulation as

derived by Feser cannot apply to the square geometry as the mean liquid path is significantly

different. As a square geometry does not lend itself for an analytical solution, a discretized model of

the mass transport equations were formulated in Engineering Equation Solver (EES) using a 23 x 23

grid with a source input in the center cell. A reduced version of this grid can be observed in Figure 41

(sketched as 5 x 5 grid, a 23 x 23 grid was used).

58
 Figure 41. User interface of discretized evaluation of wick permeability model

The equations used to evaluate a unit cell are presented below:

̇ [ ] Eq. 44

̇ [ ] Eq. 45

̇ [ ] Eq. 46

̇ [ ] Eq. 47

̇ ̇ ̇ ̇ ̇ Eq. 48

Where the first four equations(Eq. 44-Eq. 47) describe the mass flux relations between each cell and

it’s neighboring cell and the last equation(Eq. 48) represents the conservation equation with an

additional source term. The source term allows for injection in the center node. From this discrete

59
model an effective permeability can be derived, even if the experiment is performed on a rectangular

sample. The full set of equations of the model can be found in Appendix G.

Figure 42. Pressure drop as a function of

Air Massflow through 26mm x 26mm x 1mm wicking structure with center inlet
and exit around periphery for range of Permeability coefficients

Model results are presented in Figure 42. The data in Figure 42 represents a series of numerical

evaluations of pressure drop as a function of mass flow through a 26mm x 26mm x 1 mm wicking

structure where mass flow enters the center and exits around the periphery for a number of different

values of wick permeability coefficient. Figure 42 can be used as a look up chart.

3.3.4. Wick Permeability Results

Considerable challenges were faced when attempting to evaluate the permeability of a square

26mm x 26 mm x 1mm wicking sample in a center inlet, periphery outlet test configuration similar to

the work form Feser et al.[33] as depicted in Figure 40. Most challenging aspects were the sealing

around the edges of the wicking structure without damaging or compressing the wick such that the

results were affected. In lieu of full completion of these tests, a permeability test was performed to

test the permeability of wicking sample perpendicular to its planar direction as depicted in Figure 43.

60
 liquid
̇

wick

liquid
Figure 43. Perpendicular Wick Permeability configuration

As Ababneh[19] already illustrated the use of the Hagen-Poiseuille relation with a modified

constant of 1.7 with considerable success(Figure 13), it is suggested to use this formulation for the

evaluation of permeability.

3.4. Wick Thermal Conductivity

Wick thermal resistance plays an important part in the thermal path as it aids in the diffusion of heat

from the wick substrate to the meniscus surface as well as the planar heat diffusion leakage path

between evaporator and condenser. Although empirical models of effective wick thermal

conductivity are available (as presented by Eq. 22), the uncertainty of the inputs for a sintered wicking

structure warrant an experimental validation of wick thermal conductivity.

3.4.1. Laser Flash Method

The intrinsic effective thermal conductivity of a porous sintered copper structure was evaluated

using the laser flash method.

61
 Figure 44. Laser Flash Method

In this method, a sample is heated from the bottom with a laser pulse while the thermal transient

response is measured from the other side using an IR-detector. The transient thermal response is

used to calculate the thermal diffusivity, which relates to the thermal conductivity of the sample

(Figure 44).

3.4.2. Wick Thermal Conductivity Results

An effective thermal conductivity of 165 W/m-K was measured on the first copper wick samples

produced by the fabrication team using a Netszch µflash system. Samples were prepared by cutting

out 8 mm x 8 mm x 1mm sections from a larger porous wick structure. Aluminum, copper and porous

copper foam structures were used as reference baselines. These reference data points provide

additional confidence in the quality of the porous copper wick sample.

62
 500

450
Measured Thermal Conductivity (W/mK)...

400

350

300
Copper wick
sample
Copper foam
250 baseline

200

150

Phase1 target
100

50

0
Aluminum Copper Reference Copper Foam Copper Foam Copper Foam Copper Wick
Reference Sample Sample 46.2% Porosity 57.8% Porosity 65.5% Porosity 33.2% Porosity

Figure 45. Laser Flash Test Results (Copper Wick sample result averaged over 45 tests)

A total of 15 samples were tested three times to provide sufficient statistical data. Variation in

measured effective thermal conductivity is attributed to local porosity variation within the sample.

The measured thermal conductivity is presented in Figure 45. The measured effective thermal

conductivity of 165 W/m-K significantly exceeded a target of 100 W/m-K as imposed by DARPA as a

requirement of the first phase of the program.

3.5. Experimental Evaluation of Heat Transfer in the Evaporator

The thermal resistances from the outer substrate wall to the vapor in the evaporator area and from

the vapor to outer substrate wall in the condenser area are crucial to the performance of the TGP.

Although modeled as simplified resistances ( , Eq. 20 and , Eq. 21), the thermal path in these

areas is complicated and consists of multiple elements. The thermal path for heat going into the

vapor state and vice versa includes the following elements: Thermal resistance of the substrate,

wick-substrate interface thermal resistivity, thermal conduction through the wicking structure,

63
thermal conduction through the liquid and finally evaporation or condensation at the liquid-vapor

interface. This path is described in more detail in paragraph 2.5.2. Due to the complexity of this path,

measurement of each of the component resistances of this path is challenging. A method is

developed to measure the thermal resistance from the outer substrate surface to the vapor space in

its entirety. This total resistance can serve as an important input to the thermal model while an

individual breakdown of subcomponent thermal resistances is of lesser importance.

3.5.1. Evaporation Heat Transfer Experimental Setup

An experiment was developed to evaluate both the mass transport capability as well as the

evaporation heat transfer coefficient of a wicking structure. The setup is based on the principal of

creating an “open instrumented heat-pipe” in a large vessel that is under evacuated conditions . As

the experiment is done in a large vessel, instrumentation of the sample is feasible using vacuum

feed-through connections. As conditions in the evaporator area are provided that are similar to those

inside a working device, an attempt is made to copy and measure the heat transfer behavior.

Retainment fixture base saturated vapor Window fixture

heat temperature
input(Q) sensor vapor
at
Vent window in retainment
wick window
Substrate

Figure 46. Principle of Evaporation Heat Transfer and Mass Transport Test Setup

64
Figure 46 describes the key elements of the experiment. A TGP open substrate with wicking structure

is placed vertically in a pool of liquid. The liquid pool height covers the condenser section of the wick

(about 8 mm). A base fixture contains a heater to heat the TGP in the evaporator region. The base

fixture also provides thermal insulation to the back of the heater (Figure 47 and Figure 48).

Figure 47. Base fixture carroussel for three Figure 48. Base fixture carroussel with
simultanous tests Window fixture installed
The fixture was produced out of Polycarbonate using rapid-prototyping. The test fixtures and

electrical passages to the vacuum vessel were expanded to accommodate a carrousel where three

wick simples can now be tested simultaneously.

Figure 49. Window fixture back Figure 50. Window fixture front Figure 51. Window
fixture Section

65
The back of the window fixture contains a cutout for the TGP sample (Figure 49). The fixture covers a

2mm section of the device for retainment. The front of the window fixture has an opening for vapor

to escape(Figure 50). A feed channel in the bottom provides liquid replenishment for the wick as

vapor escapes(Figure 51). The fixture is placed in a vessel that can be evacuated using a vacuum

pump (Figure 52).

Figure 52. Heat Transfer and Mass Transport Test Setup vacuum chamber

Heater lead wires and thermocouples are connected to the samples inside the vessel using vacuum

feed throughs. An overview of the connected hardware can be observed in Figure 53.

Power Supply

E3634A
heater
Vacuum DAQ Labview PC
TC
pump 34970A

Figure 53. Heat Transfer and Mass Transport Setup connection diagram

66
An Agilent E3634A powersupply was used to power one or more heaters in parallel. K-type

thermocouples were connected to an Agilent 34970A Data Acquisition System.

Vacuum Vessel
Vacuum pump
PC
DAQ

Chiller
Power supply1 Power supply2
Figure 54. Heat Transfer and Mass Transport Setup

A picture of the setup is given in Figure 54. A pressure gauge is used to read out the pressure inside

the vacuum vessel.

3.5.2. Evaporation Heat Transfer Experiment Procedure

The focus of the experiment is to evaluate the heat transfer performance in the evaporator. In a TGP

liquid is provided to the evaporator by capillary forces acting on the fluid moving liquid from the

condenser. To isolate evaporator effects and eliminate condenser effects, the bottom 8 mm of the

wick are submerged in water. This simulates a flooded or infinite condenser. The vessel is then sealed

by placing a plexiglass lid on top. Vacuum grease is used to create a tight seal between the rubber

gasket and the lid. The vessel is evacuated by a roughing pump able to provide sufficient vacuum for

water to start evaporating. Initially the pump is able to draw a vacuum in the vessel lower than the

water saturation pressure at room temperature (approximately 24 Torr, 3.2kPa at 25ºC.). As the water

67
starts evaporating, the valve between the vessel and the vacuum pump is closed. The vessel

pressure then reaches equilibrium at the saturation pressure at room temperature.

T htr T wick T vapor T htr ambient

Test2: T wick Test2: ambien
Rcontact Revap Assume RevapR
~0
contact T vapor Assume Revap~0
Rcond~4 K/W Rcond~4 K/W

Test4: saturation conditions Test4: saturat

1/Rtot=1/Revap+ 1/Rcond (assume Tvapor ~Twater ) 1/Rtot=1/Revap+ 1/
Rtot~0.4 K/W Rtot~0.4 K/W

Rcond Revap~0.44 K/W Rcond Revap~0.44 K/W

Revap=1/(htc*A) Revap=1/(htc*A)
htc=1 /(R evap *A) htc=1 /(R evap *A)

Alow=6 mm * 22 mm = 132E-6 m2 Alow=6 mm * 22 m

Ahigh=16 mm * 22 mm = 352E-6 m2 Ahigh=16 mm * 22

Htclow=1/(Revap * Ahigh)~ 6392 W/m2-K Htclow=1/(Revap * A

T water T water
Htchigh=1/(Revap * Alow)~17045 W/m2 -K Htchigh=1/(Revap *

Figure 55. Thermal Resistance Network Figure 56. Thermal Resistance Network
mode1: Vessel Vapor in equilibrium with water mode2: Vessel Open, Limited evaporation

It is the objective of the experiment to get a ball park estimate (not exact) of the evaporation

resistance( ). A simplified thermal network of the sample in a saturated environment (mode 1) is

depicted in Figure 55. The temperature at the heater and water are measured. As water and vapor

are in thermal equilibrium it is not unreasonable to assume that the temperature of the water( )

will be nearly equal to that of the vapor( ). Thermal contact resistance( ) is small and is

assumed to be negligible. The remaining 2 unknowns are the evaporator( ) and

conduction( ) thermal resistances. If it assumed:

Eq. 49

The thermal resistance network from Figure 55 then reduces to network two thermal resistors that

are in parallel. Effectively the relationship between temperature rise and power input can then be

resolved as (mode 1):

68
 Eq. 50

The experiment can now be operated in two different modes. When the lid is opened and the

experiment is run open to the environment (mode 2), it can be assumed (for low input powers) that

evaporation resistance is high compared to the conduction resistance as the liquid temperature is

significantly below the saturation temperature at room pressure (compare ~25ºC to the saturation

temperature at ambient pressure of approximately 100 ºC). Effectively this reduces the thermal

network to a single thermal resistor equation (mode 2) (Eq. 51):

Eq. 51

If the conduction thermal resistance is resolved during the mode 2 operation run, it is likely to

assume that this conduction thermal resistance does not change whether the experiment is

operated in mode 1 or mode 2. Therefore, this value can be used to approximate an estimate of the

evaporation thermal resistance after the experiment is subsequently run in mode 1.

3.5.3. Evaporation Heat Transfer Experiment Results

Initially a baseline test is performed to establish the thermal resistance of the conduction path

through the sample (mode 2). This is done by opening up the vessel and measuring the thermal

response as a function of power input at ambient pressure (Figure 57, closed markers). Note that the

data in Figure 57 represents two different tests with different wick types. Differential thermocouple

accuracy is evaluated as ±0.78ºC (see paragraph 5.5.1.4). It can be observed as the samples reach

the saturation temperature at ambient pressure (~100ºC), the gradient of the curve reduces as more

and more liquid is evaporated.

69
 R~4.5K/W

R~4K/W R~0.5K/W

R~0.25K/W

Figure 57. Mass Transport Thermal Response data

By evaluating the constant initial slope of the temperature response, an estimate can be made for

the conduction resistance of the sample. The conduction resistance of the sintered and tape cast

samples shown in Figure 2 was found to be 4 and 4.5 K/W respectively.

The vessel was subsequently sealed and evacuated (charged). After initial evacuation, the valve to

the chamber was closed and in about half an hour the vessel reached equilibrium pressure at room

temperature. Figure 57 shows the response of two samples under these new conditions (open

markers). Under saturation conditions, evaporation effects dominate and, as discussed, the effective

conduction through the wick down to the water and evaporation resistance to the vapor can now be

considered as parallel thermal resistances. This results in a significantly reduced slope of the total

thermal resistance as shown in Figure 57. The effective total thermal resistances in these cases are

0.5 K/W and 0.25 K/W for the sintered and tape cast wick samples respectively. By evaluation of Eq.

50, it was found that this results in evaporation thermal resistance approximations of ~0.57

K/W and ~0.26 K/W for the sintered and tape cast wicks respectively. Although these are

approximate and not exact results (with a great number assumptions), the data seems to suggest

70
that evaporation performance of tape cast wicks is superior to sintered wicks. Results from this

method have been successfully used by Ababneh et al. [34, 35] as inputs to a finite element model

prediction of TGP performance.

3.6. Evaluation of Hysteresis Effects

Another interesting effect was observed when testing wicking structures in the evaporation heat

transfer experiment. Figure 58 shows data of a single wick sample tested at different heat input

levels. Heat input to the sample was gradually increased and decreased from zero to 25, 18 and 11

watts respectively.

Figure 58. Mass Transport Thermal Response data

At low input powers, a startup region is observed with a linear relation between power input and

temperature rise. After the startup region the slope of temperature rise decreases, which is attributed

to water evaporating from the wick structure. As heat is increased further, a linear region exists with

71
a lower slope than the initial startup region. After the linear region, the temperature curve slope

increases to a point where it approaches the intersection of the initial startup curve. This is attributed

to partial and finally full dryout of the wick structure. It was observed that for heat input levels that

have exceeded the estimated dryout onset point, hysteresis occurs where the power return curve

follows a higher temperature rise path than the power up path. This is attributed to the fact that the

portion of the wick that dries out experiences significantly higher temperatures making it challenging

for a fluid to wet in saturated conditions. However, when the power level is reduced to zero and the

wick is cooled down, the performance of the wick again follows the “wetted” power up curve. This

hysteresis effect is similar to effects that have been observed in boiling experiments. In fact, if the

axis of Figure 58 were reversed, it can be appreciated that the “up slope” resembles a typical boiling

curve with an initial sensible heat rise region (until superheat is sufficient for evaporation to start)

followed by a region of high heat transfer coefficient which levels off near a critical heat flux (dry-out)

point.

3.7. Experimental Evaluation of Heat Transfer in the Condenser

Evaluation of heat transfer performance in the condenser area is significantly more challenging. In

order to evaluate condenser heat transfer thermal resistance in a method similar to the

aforementioned “open heat-pipe” type experiment, the wicking structure needs to be placed in a

saturated vapor chamber while the condenser temperature is controlled to initiate condensation. By

subcooling the condenser area (i.e. by use of a Thermoelectic element) and by measurement of the

extracted heat flux (i.e. by use of a heat flux sensor), the condenser thermal resistance could

technically be evaluated. However, it must be taken into consideration that a cycle of condensation

and evaporation would need to be sustained to perform this measurement in steady state. This

would mean an equivalent amount of heat would need to be put in the evaporator to evaporate the

condensate. Due to the significant challenge of this experiment and the limited resources and time

constraint on the program, an alternative was sought to approximate .

72
As mentioned before, the resistances and are grouped total resistances combining a

number of physical layers and effects (substrate conduction, thermal interface resistance, wick

conduction and two phase heat transport). Since both resistances share most of the layers in

common (except for the different modes of two phase heat transport), it is assumed that the

previously evaluated will provide a reasonable estimate for evaluation an approximation of

Eq. 52

73
4. Visual Validation of TGP Performance under High-g Conditions

4.1. Introduction

Per requirement of the first phase of the TGP program, a visual demonstration had to be performed

to confirm wick saturation capability at a minimum of 2g acceleration. An experiment using an

ultraviolet fluorescent dye as contrast agent was developed. The experiment was mounted and

executed on a high g centrifuge table at the Air Force Research Laboratory(AFRL) in Dayton, OH. This

work was published and presented in greater detail by Chamarthy and de Bock et al. [36]. In this

work, the selection of the fluorescent dye was led by Chamarthy and Russ. The experimental

evaluation, test fixture design and high-g testing work was performed by de Bock.

4.2. Wick Saturation Visualization using Fluorescent Dyes

DARPA required visual demonstration of wicking performance at 2-g’s to validate proper functioning

of the wick. A method was developed to add small amounts of fluorescent dye to the working fluid to

visualize the location of the fluid in the wicking structure. Figure 59 shows the results of initial

experiments that were used to validate this approach. Figure 59 shows the saturation front

progression of a circular wick structure that is touching a pool of liquid in vertical orientation.

t=0s t=14s t=28s t=42s t=56s t=1m10s

Figure 59 Visualization of wick saturation in copper sintered wick using fluorescent dye over time

74
 4.3. TGP prototype Wick Saturation Visualization

As Figure 59 demonstrated the feasibility of using fluorescent dyes for liquid visualization in wicking

structures, a TGP was developed with a transparent quartz lid. The lid functions to limit unwanted

evaporation of the liquid while enabling visual access to the wick.

Figure 60 TGP Sample with Quartz Lid on filling station

When a UV light source shines on the wick structure that is saturated with water/UV dye solution,

visual light is emitted. The fluorescent dye absorbs the UV light and emits in the visual spectrum.

UV light source 254 nm Regular video camera

Filter sheet that blocks UV

but passes visual light

Vapor Chamber
Wick

Wick saturated with water/UV dye

solution
Figure 61 UV Visualization setup

A kapton filter was used to pass only the visual emission from the dye such that the camera only

registers the location of the solution in the wick structure. A schematic of this experiment is sketched

75
in Figure 61. A fixture was designed and produced out of polycarbonate using rapid prototyping

(Figure 62).

A
Camera window

2xUVP 254 nm light source

Pockets for flexible strip to

clamp on penlight source
Edge to clamp on TGP

TGP

A A-A

Figure 62 TGP Sample and light source holder fixture

The fixture serves to retain the TGP sample as well as fix the position of two UV pen-light sources that

irradiate the sample while providing visual access for a camera. The Kapton filter can be attached

over the top surface of the camera window.

4.4. Wick Saturation Visualization Validation

Optimization of the dye selection (absorption and emission bands) resulted in a method where a clear

contrast between saturated and unsaturated regions is achieved (see Figure 63).

Baseline: UV light on Test: UV light on Test: UV light &hood light on

unsaturated sample saturated sample saturated sample

Figure 63 Baseline TGP UV lighting results

Figure 63 shows a clear contrast between the unsaturated and saturated samples that are

76
illuminated using the UV light. For reference Figure 63 also shows a sample lighted with regular room

lighting (visual spectrum). Excess liquid pools at the sides and corners between the package and wick

structure and can be observed as bright spots.

4.5. Development of High-G TGP Wicking Visualization Experiment

The AFRL high-g centrifuge facility at Wright Patterson Air Force Base(WPAFB) in Dayton consists of

an 8-feet diameter spin table that is rated to provide acceleration up to 10 g’s (Figure 64).

Figure 64 AFRL centrifuge table

The centrifuge table is located in a safety bunker at WPAFB. A light-weight aluminum profile structure

was developed to mount the TGP Sample and Light Holder(Figure 62) fixture and a Sony portable

camera to the table (Figure 65). The aluminum structure vertical beams were designed to hold the

camera in position without exceeding elastic limits with a 10x safety factor. Additional diagonal

braces were added to ensure the structure would not deflect significantly under high-g loading.

77
 Figure 65 TGP Sample and Light holder mounted on Aluminum structure

Mounting holes in the aluminum structure allowed for securing to the centrifuge table. The structure

was designed out of aluminum for its favorable strength/weight ratio. The designed structure was

installed on the table by the AFRL team (Quinn Leland, Kirk Yerkes, Jamie Ervin and Doug Johnson)

who performed balancing of the centrifuge prior to testing. An accelerometer was added to the test

fixture by the AFRL team. A live video feed of the sample was available in the control room as well as

live data feed from the accelerometer. The acceleration of the sample was calculated by correcting

for the difference in radius between the accelerometer and the sample. The mean sample radius was

found to be 44 inches.

4.6. High-G TGP Wicking Visualization Experiment Results

In order to exceed DARPA’s 2g acceleration requirement, a testing plan was developed of increasing

acceleration up to 2.5 g’s followed by a hold at 2.5 g’s for a minimum 30 seconds. This pattern was

repeated 3 times followed by an acceleration to the table maximum (Table 5).

78
 Step Mode Start g [m/s2] End g [m/s2] Minimum
Duration [s]
1 Acceleration 0 2.5 30
2 hold 2.5 2.5 60
3 deceleration 2.5 0 30
4 hold 0 0 30
5 Acceleration 0 2.5 30
6 hold 2.5 2.5 60
7 deceleration 2.5 0 30
8 hold 0 0 30
9 Acceleration 0 2.5 30
10 hold 2.5 2.5 60
11 Acceleration 0 Table maximum
12 deceleration Table maximum 0 30

Table 5 High-G TGP Wicking Visualization Testing plan

Both video and accelerometer data were recorded during the test. Postprocessing of the data

allowed for reconstruction of the test in a series of snapshots. Figure 66 and Figure 67 present

images of the wick sample as recorded by the camera at different acceleration conditions. A graph

presents the acceleration profile. A slight excess pool of liquid can be observed on the sides of the

sample at 0.0g’s. At higher acceleration, the excess liquid pools at the top.

Excess liquid pools at top

Figure 66 Saturated wick at 2.6g, Figure 67 Sample is decelerated,

excess liquid pools on top excess liquid returns
(G-force in direction) (G-force in direction)

After a 30-second acceleration, the sample was held at around 2.6g’s for 60 seconds. It is important

to note that the main square area of the wick remains visible. This indicates that the UV dye solution

79
is present inside the square wick throughout the acceleration to 2.6 g’s, meeting the DARPA

requirement.

Figure 68 Saturated wick at 8.9g, Figure 69 Sample is decelerated,

excess liquid pools on top excess liquid returns
(G-force in direction) (G-force in direction)

After a deceleration to 0.0 g’s the liquid pool from the top re-floods the edges of the wick sample

(Figure 67). This acceleration profile was repeated. During the test, confidence was gained in the

strength of the setup and the final acceleration profile was approved. In this profile the sample is

accelerated to around 2.5 g’s over a period of 30 seconds after which the sample is accelerated an

additional 1g every 30 seconds. This was repeated until the accelerometer reading was at 10 g’s. As

the accelerometer was at larger radius than the sample the effective corresponding sample

acceleration was found to be 8.9 g’s(Figure 68). After completion of the test, the liquid distributed

itself evenly over the wicking structure (Figure 69).

As a result, it is demonstrated that the developed wicking sample meets and exceeds the DARPA

phase 1 requirement of visual wicking capability at 2 g’s acceleration. In fact, wicking performance

was demonstrated up to 8.9’s acceleration.

80
5. Evaluation of TGP Thermal Performance

5.1. Introduction

The main purpose of the TGP is to provide a low thermal resistance connection between a heat

source, such as a chip or power device, and a heat sink. Using a detailed design approach a

measurement setup was developed to characterize TGP thermal performance. As heat pipes and/or

vapor chambers can have an effective thermal conductivity several times greater than copper,

temperature gradients across a TGP can be small. This makes measurement uncertainty is an

important aspect of thermal characterization, which is therefore discussed in detail. A variant of this

setup is then adapted for evaluation of thermal performance under high-g conditions.

5.2. TGP Thermal Characterization Experiment Description

An experiment was designed in which heat is added to the TGP by a thick film heater and the heat is

removed by a high performance micro channel cold plate. A sketch of the test setup with its

components is given in Figure 70.

Load cell Test sample Insulation

evaporation transport condensation

TIM

Thick film Micro channel

heater coldplate

Figure 70 TGP Thermal Characterization Setup

The purpose of the setup is to provide a heat path between heat source and sink through the test

sample, with minimal heat loss to the ambient. Heat flows from the heater element through the

81
heater substrate and thermal interface material to the TGP sample. Heat is removed from the TGP

through the thermal interface layer and rejected by the cold plate. A chiller controls the temperature

of the fluid in the cold plate. Insulation is provided around this setup. As thermal interface material

performance varies greatly with contact pressure, the applied compressive load is measured using a

load cell. Care was taken to supply sufficient compressive load to reduce the contact pressure

sensitivity. Measuring the contact load also allowed for improved experimental repetition accuracy.

5.3. Setup components

The individual components in the setup and their relevant parameters are:

 Heater element: the heater consists of a custom 30mm x 10mm x 2.5mm Aluminum Nitrate

(AlN) Watlow Ultramic thick film heater with embedded k-type thermocouple. The heater is

rated to perform up to 240°C, which is the melting temperature of the solder connections.

 Thermal interface material: FujiPoly 50XR-m is used as thermal interface material to ensure

a good contact between the heater substrate and the TGP. The chosen material has a

thickness of 500µm and a nominal through thickness thermal conductivity of 17 W/m-K.

Figure 71 Fujipoly 50Xr-M thermal resistivity as a function of contact pressure

82
 The XR-m material has a putty like structure. Figure 71 gives the thermal resistivity as a

function of contact pressure for a 50 Xr-m which is to typical for a thermal interface

material[38]. As the thermal interface resistance is expected to be a significant part of the

resistance budget, the setup includes a pressure transducer to evaluate the contact pressure

placed on the interface material. This is also expected to enhance the repeatability of the

measurements as the contact pressure can be registered and reproduced. It is preferable to

operate the experiment in the low sensitivity region as indicated in Figure 71.

 Load cell: The load cell used is an Omega LCKD-1000

 Cold plate: A lytron CP25 high performance microchannel cold plate is selected for this

program for its superior low thermal resistance and dimensional compatibility. The

performance of the cold plate scales with flow rate. Therefore it is important that a similar

flow rate is used when multiple tests are performed to ensure compatible results.

Figure 72 Lytron CP25 high performance cold plate

 Insulation material: To eliminate interior convection cells, a tight fitting shape around the

components is required. Polycarbonate shell fixtures were produced using rapid prototyping.

83
 These are formfitting and eliminate air pockets < 1mm3 that could act as interior convection

cells. The thermal conductivity of the PC shells is 0.2 W/m-K, the maximum temperature of

this material is limited by its glass temperature of 160°C.

Figure 73 Insulation shells with installed cold plate

The components are connected as depicted in the diagram in Figure 74.

Figure 74 Overview of TGP Thermal Characterization Setup Components and interaction.

84
Although the Cole-Palmer Chiller has its internal circulation pump, an additional pump was used in

series to provide sufficient pumping head to achieve the desired flow rate through the micro-channel

cold plate. A picture of the experimental setup can be observed in Figure 75

Chiller
TGP test fixture Rotation fixture

Flow bypass valve

Pump
Temperature safety

controller

DAQ

Power

Supply

Figure 75 TGP thermal performance characterization experiment

5.4. Thermal Model of Experiment

In order to understand the relative importance of the individual components of the experiment, a

model is developed of the experiment in two stages. First a screening model is developed to provide

a quick guidance on which areas are of most importance. The screening model is followed by a more

detailed model which evaluates the experiment using conjugate FEM. This conjugate model is used

to obtain an estimate of the system heat losses.

85
 5.4.1. Screening Thermal Model Results

As it was uncertain what the initial effective thermal conductivity of the TGP was going to be, A

simple one-dimensional thermal analysis was performed using 400 W/m-K and 2000 W/m-K TGP

effective thermal conductivity to give an estimate of the significant contributors to the thermal path.

The results from this model drive the detailed requirements for a consecutive detailed finite element

analysis (FEA) model. Table 6 and Table 7 provide overviews of the expected thermal resistances.

R A k dx R" R"
K/W m2 W/m-K m e+4 m2-K/W e+2 in2-K/W
R_substrate R1 8.33E-02 3.00E-04 40.0 1.00E-03
R_interface1 R2 8.47E-02 3.00E-04 10.0 2.54E-04 2.54E-01 3.94
R_tpg R_TPG 5.56E-01 9.00E-05 400 2.00E-02
R_interface2 R3 8.47E-02 3.00E-04 10.0 2.54E-04 2.54E-01 3.94
R_coldplate R4 1.90E-01 3.00E-04
R_total R_total 1.00
Temperature rise @ Q=40W dT_40 39.9
Temperature rise @ Q=120W dT_120 119.8

Table 6 TGP testing setup performance,

contact pressure > 40PSI, cold plate flow rate 1 GPM, k eff = 400 W/m-K

R A k dx R" R"
K/W m2 W/m-K m e+4 m2-K/W e+2 in2-K/W
R_substrate R1 8.33E-02 3.00E-04 40.0 1.00E-03
R_interface1 R2 8.47E-02 3.00E-04 10.0 2.54E-04 2.54E-01 3.94
R_tpg R_TPG 1.11E-01 9.00E-05 2000 2.00E-02
R_interface2 R3 8.47E-02 3.00E-04 10.0 2.54E-04 2.54E-01 3.94
R_coldplate R4 1.90E-01 3.00E-04
R_total R_total 0.55
Temperature rise @ Q=40W dT_40 22.2
Temperature rise @ Q=120W dT_120 66.5

Table 7 TGP testing setup performance,

contact pressure > 40PSI, cold plate flow rate 1 GPM, k eff = 2000 W/m-K

The expected temperature rises are within acceptable temperature limits. The relative contributions

of the individual components to the resistance budgets are broken down in Figure 76 and Figure 77.

Figure 76 and Figure 77 illustrate that each of the thermal resistances is significant and that they are

of similar magnitude. However, at lower TGP effective thermal conductivity, the TGP itself becomes

the dominant resistance. At higher TGP effective thermal conductivities, the cold plate and the two

interface resistances are dominant.

86
 R_substrate R_substrate
R_coldplate 8.3% 15.0%
19.0%
R_interface1
8.5%
R_coldplate
34.3%

R_interface1
R_interface2
15.3%
8.5%

R_tpg
R_interface2
R_tpg 20.1%
15.3%
55.6%

Figure 76 Relative contributions to resistance Figure 77 Relative contributions to resistance

budget TGP keff=400 W/m-K budget TGP keff=2000 W/m-K

5.4.2. Detailed thermal model results

As the conceptual model gave acceptable temperature rises, a higher level of detail model is

constructed to validate the amount of heat leakage. Heat leakage is defined as the amount of heat

that does not follow the prescribed one-dimensional path but exits the setup from the sidewalls by

other means (such as free convection). A sketch of the three-dimensional setup can be observed in

Figure 78.

Insulating top
shell

TGP
prototype Cold plate

Heater
Insulating bottom
shell
Figure 78 TGP Thermal Characterization Setup

87
A three-dimensional finite element analysis model has been created in ANSYS modeling the

individual components of the setup. Thermal interfaces were added as thermal conductances.

Convection boundary conditions were added on the exterior walls as well as inside the cold plate.

The heater is modeled as a volumetric heat source and the cold plate as a constant temperature

solid.

Figure 79 Finite Element Analysis Model vector plot

The vector plot in Figure 79 shows the heat transfer from the heater to the cold plate through the

TGP. The effective temperature rises at two different heat loads are presented in Figure 80 and Figure

81. Temperature rises at 40W and 120W heat input respectively are 33.3 and 99.7 degrees. The

ANSYS APDL input for this model can be found in Appendix H.

88
 Figure 80 Temperatures at 40W heat input Figure 81 Temperatures at 40W heat input
(keff=400W/m-K, coolant 20C, ambient 20C) (keff=400W/m-K coolant 20C, ambient 20C)

Table 8 presents the temperature rise data for the screening and the detailed model. Table 8 shows

that expected heat leakage is around 1.3% for the condition that ambient and coolant are at the

same temperature and the effective thermal conductivity of the TGP is around 400 W/m-K.

Screening model detailed model

Temperature rise at 40W 39.9 K 33.3 K

Temperature rise at 120W 119.8 K 99.7 K
Heat to cold plate (Qin=40W) 40.0 W 39.48 W
Heat to environment (Qin=40W) 0.0 W 0.52 W
Heat leakage 0.0 % 1.3 %
Heat to cold plate (Qin=120W) 120.0 W 118.45 W
Heat to environment (Qin=120W) 0.0 W 1.55 W
Heat leakage 0.0 % 1.3 %
Table 8 Concept, detailed model comparison
(keff=400W/m-K, coolant 20C, ambient 20C)

If the coolant temperature is raised higher than ambient, this leakage is expected to increase and will

have to be re-evaluated. For higher TGP effective thermal conductivities the leakage is expected to

decrease.

5.5. Measurement Uncertainty

Measurement uncertainty plays a role in any experiment, but is especially important when an

attempt is made to measure devices with high effective thermal conductivity such as heat pipes or

vapor chambers.

89
 100%

Relative error in thermal

Q=20W

conductivity dK/K
75%

50%
dT=+/- 1.41K

25%
Target <10% dT=+/- 0.15K

0%
0 500 1000 1500 2000
TGP effective thermal conductivity K [W/m-K]

Figure 82 Effect of temperature measurement uncertainty on

thermal conductivity measurement error

A screening analysis is performed where only the uncertainty contribution of the differential

temperature measurement on the relative effective thermal conductivity uncertainty is evaluated

(assuming all other contributions zero). The results from this analysis can be observed in Figure 82. As

Figure 82 illustrates that it is significantly challenging to reach the objective if less than 10% effective

thermal conductivity error even at effective thermal conductivities similar to copper (~400 W/m-K).

This challenge only increases as higher effective thermal conductivity prototypes are developed.

Therefore, a significant effort is put into understanding the uncertainty and contributions to

uncertainty of the measurement.

5.5.1. Temperature Measurement Uncertainty

Temperature measurements in particular are sensitive to uncertainty due to the relative inaccuracy

of absolute temperature measurement instruments. Temperature measurement uncertainty can be

evaluated as a propagation in uncertainty of the thermocouple combined with the data acquisition

reference junction uncertainty.

90
 5.5.1.1. Thermocouple Measurement Uncertainty

Thermocouples produce a voltage based on the temperature difference of two junctions at dissimilar

temperature. The magnitude of the voltage induced by the junction is based upon the principle of the

Seebeck effect. The Seebeck effect governs the direct conversion of heat differentials to electric

voltage and vice versa. If a thermocouple is connected to a data acquisition system, a circuit is

created which is illustrated in Figure 83.

Figure 83 Circuit for coupling of a T-type thermocouple to a

data acquisition system. Adapted from [39]

A T-type thermocouple consists of a junction between a constantan and copper lead. A junction is

created at the measurement location (1) and at the copper connector in the data acquisition

system(2). The voltage( ) read-out is proportional to the temperature difference between junction 1

and junction 2 and scales with the Seebeck Coefficient( ).

Eq. 53

91
Note that the Seebeck coefficient is temperature dependent and can be evaluated at the mean

temperature between junction 1 and 2.

For the temperature range 10°C - 100°C, E- and T-type thermocouples are attractive as they both

exhibit a large effective Seebeck coefficient in this temperature range giving maximum sensitivity. T-

type thermocouples have the advantage that the positive lead is made of copper reducing the

number of junctions by one. For this reason, T-type thermocouples were used in this study. Seebeck

coefficients standards for the T-type thermocouple are defined by NIST and are presented in Figure

84.

NIST T-type Thermocouple voltages [0-400C]

Nist coefficients:
n

5.0E-03 V  c T
i 0
i
i

4.0E-03 c0= 0.000000000000E+00

c1= 0.387481063640E-01
Voltage [V]

3.0E-03
c2= 0.332922278800E-04
2.0E-03 c3= 0.206182434040E-06
c4=-0.218822568460E-08
1.0E-03
c5= 0.109968809280E-10
0.0E+00
c6=-0.308157587720E-13
0 20 40 60 80 100 120 c7= 0.454791352900E-16
Temperature [C]
c8=-0.275129016730E-19

Figure 84 Type-T thermocouple voltage with 0°C reference, source NIST [40]

As thermoelectric properties can vary batch-to-batch, calibration of thermocouples will be

performed before use. Figure 84 can be used to back-calculate the effective Seebeck coefficient as a

function of temperature for this thermocouple. Note that the voltages shown in Figure 84 correspond

to the temperature difference between measured temperature and a zero reference. Therefore the

voltage induced at a temperature can be used to calculate the Seebeck coefficient where the

material is effectively at a temperature between and the reference temperature .

92
 [ ] Eq. 54

The function for the Seebeck coefficient is presented in Figure 85. The Seebeck coefficient curve was

fitted using a 2nd order polynomial with R2=99.99%

Effective Seebeck coefficient for T-type thermocouple 0-400C

from NIST

4.300E-05
y = 3.4458E-11x 2 + 6.1865E-08x + 1.9209E-05
R2 = 9.9992E-01
4.200E-05
alpha [V/K]

4.100E-05

4.000E-05

3.900E-05

3.800E-05
275 280 285 290 295 300 305 310 315 320 325 330
Temperature [K]

Figure 85 Type-T thermocouple Effective Seebeck coefficient

5.5.1.2. Data Acquisition Contribution to Thermocouple Uncertainty

The reference junction in the data acquisition system (DAQ) plays an important role in evaluation of

the temperature measurement uncertainty.

Figure 86 Type-T thermocouple Effective Seebeck coefficient

93
The temperature measurements in this report were performed using an Agilent 34970A Data

Acquisition system using a 34901A 20-channel module. The most important characteristics for

accuracy evaluation are the range and reading error of DC voltage as produced by thermocouples

as well as the temperature accuracy of the reference junction these parameters can be observed in

Table 9.

Agilent 34970A
Minimum range:  100 mV
Range error: *) 0.0040 % of range   V
Reading error: *) 0.0050 % of reading
Reference junction accuracy:  0.5 C
*) at 1 year calibration and operating temperature within 23  5 C
Table 9 Agilent 34970A Accuracy summary

5.5.1.3. Propagation of Temperature Uncertainty

A temperature measurement error is a combination of data acquisition, and temperature reference

errors.

The measured voltage is expressed as:

Eq. 55

The measured temperature can then be expressed as:

Eq. 56

Propagation of errors gives:

√( ) Eq. 57

94
This allows expressing the error in measurement of the temperature at junction 1 as a function of

data acquisition and temperature reference error.

Yokogawa MX100 dTr 0.8 C Reference temperature accuracy

T_1
5.5.1.4.
T_ref
Temperature
T_alpha alpha
Measurement
V Verr_rng
Uncertainty
Verr_rdng Verr_total dT+ dT- Vref
C C K V/K V V V V C C V
10 20 288.15 3.990E-05 -3.986E-04 5.00E-06 -1.993E-07 4.801E-06 0.809 -0.809 7.896E-04
The expected temperature measurement accuracy of a temperature between 10 - 100C for a T-type
20 20 293.15 4.031E-05 5.925E-08 5.00E-06 2.963E-11 5.000E-06 0.810 -0.810 7.896E-04
30 20 298.15 4.072E-05 4.069E-04 5.00E-06 2.034E-07 5.203E-06 0.810 -0.810 7.896E-04
40 20 303.15 4.113E-05 8.222E-04 5.00E-06 4.111E-07 5.411E-06 0.811 -0.811 7.896E-04
thermocouple using an Agilent 34970A system using the internal reference junction is given in Table
50 20 308.15 4.154E-05 1.246E-03 5.00E-06 6.231E-07 5.623E-06 0.811 -0.811 7.896E-04
60 20 313.15 4.196E-05 1.679E-03 5.00E-06 8.393E-07 5.839E-06 0.812 -0.812 7.896E-04
70 20 318.15 4.238E-05 2.119E-03 5.00E-06 1.060E-06 6.060E-06 0.813 -0.813 7.896E-04
10 80 20 323.15 4.280E-05 2.568E-03 5.00E-06 1.284E-06 6.284E-06 0.813 -0.813 7.896E-04
90 20 328.15 4.322E-05 3.025E-03 5.00E-06 1.512E-06 6.512E-06 0.814 -0.814 7.896E-04
100 20 333.15 4.364E-05 3.489E-03 5.00E-06 1.744E-06 6.744E-06 0.815 -0.815 7.896E-04

Agilent 34970A dTr 0.5 C Reference temperature accuracy

T_1 T_ref T_alpha alpha V Verr_rng Verr_rdng Verr_total dT+ dT- Vref
C C K V/K V V V V C C V
10 20 288.15 3.990E-05 -3.986E-04 4.00E-06 -1.993E-08 3.980E-06 0.510 -0.510 7.896E-04
20 20 293.15 4.031E-05 5.925E-08 4.00E-06 2.963E-11 4.000E-06 0.510 -0.510 7.896E-04
30 20 298.15 4.072E-05 4.069E-04 4.00E-06 2.034E-07 4.203E-06 0.511 -0.511 7.896E-04
40 20 303.15 4.113E-05 8.222E-04 4.00E-06 4.111E-07 4.411E-06 0.511 -0.511 7.896E-04
50 20 308.15 4.154E-05 1.246E-03 4.00E-06 6.231E-07 4.623E-06 0.512 -0.512 7.896E-04
60 20 313.15 4.196E-05 1.679E-03 4.00E-06 8.393E-07 4.839E-06 0.513 -0.513 7.896E-04
70 20 318.15 4.238E-05 2.119E-03 4.00E-06 1.060E-06 5.060E-06 0.514 -0.514 7.896E-04
80 20 323.15 4.280E-05 2.568E-03 4.00E-06 1.284E-06 5.284E-06 0.515 -0.515 7.896E-04
90 20 328.15 4.322E-05 3.025E-03 4.00E-06 1.512E-06 5.512E-06 0.516 -0.516 7.896E-04
100 20 333.15 4.364E-05 3.489E-03 4.00E-06 1.744E-06 5.744E-06 0.517 -0.517 7.896E-04

Table 10 Temperature accuracy using Agilent 34970A and internal reference

From Table 10 it can be concluded that the largest contributor to the measurement error is the

internal reference from the data acquisition system. The reference temperature error is significantly

larger than the error of an accurate RTD measurement (typically ±0.1K). It is suggested that this

might be caused by the limitations and constraints of the design of the reference junction in the data

acquisition system package.

95
 Figure 87 Circuit for coupling of a T-type thermocouple to a
data acquisition system using an external reference. Adapted from [39]

For example the size of the reference junction is limited as well as its proximity to heat sources and

ventilation passages. Preferably a reference junction has a large thermal mass, is well insulated and

instrumented with an accurate RTD or thermistor.

If there is a need for increased temperature measurement accuracy an external reference can be

used (i.e. when very small temperature gradients are measured). The temperature of the external

reference can be either kept a constant temperature such as the freezing temperature of water (0ºC

depending on ambient pressure) or measured accurately using an RTD. Agilent specifications list that

the accuracy of a calibrated RTD is within  0.06°C. This results in the temperature measurement

accuracies as listed in Table 11.

Agilent 34970A dTr 0.06 C External Reference temperature accuracy

T_1 T_ref T_alpha alpha V Verr_rng Verr_rdng Verr_total dT+ dT- Vref
C C K V/K V V V V C C V
10 0 278.15 3.908E-05 3.910E-04 4.00E-06 1.955E-08 4.020E-06 0.119 -0.119 -3.833E-08
20 0 283.15 3.949E-05 7.896E-04 4.00E-06 3.948E-07 4.395E-06 0.126 -0.126 -3.833E-08
30 0 288.15 3.990E-05 1.196E-03 4.00E-06 5.982E-07 4.598E-06 0.130 -0.130 -3.833E-08
40 0 293.15 4.031E-05 1.612E-03 4.00E-06 8.059E-07 4.806E-06 0.133 -0.133 -3.833E-08
50 0 298.15 4.072E-05 2.036E-03 4.00E-06 1.018E-06 5.018E-06 0.137 -0.137 -3.833E-08
60 0 303.15 4.113E-05 2.468E-03 4.00E-06 1.234E-06 5.234E-06 0.141 -0.141 -3.833E-08
70 0 308.15 4.154E-05 2.909E-03 4.00E-06 1.454E-06 5.454E-06 0.144 -0.144 -3.833E-08
80 0 313.15 4.196E-05 3.358E-03 4.00E-06 1.679E-06 5.679E-06 0.148 -0.148 -3.833E-08
90 0 318.15 4.238E-05 3.814E-03 4.00E-06 1.907E-06 5.907E-06 0.152 -0.152 -3.833E-08
100 0 323.15 4.280E-05 4.279E-03 4.00E-06 2.139E-06 6.139E-06 0.155 -0.155 -3.833E-08

Table 11 Temperature accuracy using external reference at 0ºC with 0.06ºC accuracy

96
Interestingly enough the measurement accuracy is a weak function of the absolute reference

temperature due to the more accurate evaluation of the Seebeck coefficient at larger temperature

differences. This is illustrated by raising the external reference temperature to 95ºC and measuring

using the Agilent system as shown in Table 12.

Agilent 34970A dTr 0.06 C External Reference temperature accuracy

T_1 T_ref T_alpha alpha V Verr_rng Verr_rdng Verr_total dT+ dT- Vref
C C K V/K V V V V C C V
10 95 325.65 4.301E-05 -3.655E-03 4.00E-06 -1.827E-07 3.817E-06 0.107 -0.107 4.046E-03
20 95 330.65 4.343E-05 -3.256E-03 4.00E-06 -1.628E-06 2.372E-06 0.081 -0.081 4.046E-03
30 95 335.65 4.386E-05 -2.849E-03 4.00E-06 -1.425E-06 2.575E-06 0.084 -0.084 4.046E-03
40 95 340.65 4.428E-05 -2.434E-03 4.00E-06 -1.217E-06 2.783E-06 0.087 -0.087 4.046E-03
50 95 345.65 4.471E-05 -2.010E-03 4.00E-06 -1.005E-06 2.995E-06 0.090 -0.090 4.046E-03
60 95 350.65 4.514E-05 -1.577E-03 4.00E-06 -7.887E-07 3.211E-06 0.093 -0.093 4.046E-03
70 95 355.65 4.557E-05 -1.137E-03 4.00E-06 -5.683E-07 3.432E-06 0.096 -0.096 4.046E-03
80 95 360.65 4.600E-05 -6.879E-04 4.00E-06 -3.439E-07 3.656E-06 0.100 -0.100 4.046E-03
90 95 365.65 4.644E-05 -2.312E-04 4.00E-06 -1.156E-07 3.884E-06 0.103 -0.103 4.046E-03
100 95 370.65 4.687E-05 2.329E-04 4.00E-06 1.165E-07 4.116E-06 0.106 -0.106 4.046E-03

Table 12 Temperature accuracy using external reference at 95ºC with 0.06ºC accuracy

Raising the reference temperature to 95ºC yields an additional reduction in measurement

uncertainty of ±~0.05 ºC. Although interesting, keeping a reference temperature at 95ºC is

challenging and therefore not recommended.

5.5.2. Heat Input Uncertainty

Heat input uncertainty is another important factor in establishing the uncertainty of the effective

thermal conductivity of a vapor chamber. The heat input uncertainty is a combination of heat input

uncertainty of power supply uncertainty, data acquisition voltage measurement uncertainty and

heat leakage uncertainty.

5.5.2.1. Power Supply Uncertainty

Heat will be put into the system by flowing a current through the resistive heater (Joule heating). The

heat load is defined by Eq. 58. By driving the circuit with a current source, the current is known and

measurement of the heater voltage differential( ) using the data acquisition system will give

the dissipated heat load ( ).

97
 Eq. 58

By measuring this voltage differential with a separate set of leads (4-wire method) is the most

accurate as no current will flow through the measurement leads due to the high DAQ impedance.

The accuracy of the heat load is dependent on the accuracy of the driving current and the voltage

measurement.

5.5.2.2. DAQ and Power Supply Uncertainty Propagation

Both voltage measurement and current source contribute to the heat load input uncertainty as given

by Eq. 59.

√ Eq. 59

Table 13 gives the combined uncertainty for heat input levels of 5, 10 and 25 Watts using an Agilent

3634A Power supply.

Q DQ V DV I DI R
W mW V mV A mA ohm
5 6.1 22.4 1.6 0.224 0.27 100
10 8.9 31.6 2.0 0.316 0.28 100
25 15.1 50.0 2.9 0.500 0.30 100

Table 13 Heat Input Accuracy using Agilent 3634A Power supply

In example, assuming heat resistance of 100 Ω, the required current for 25W heat dissipation is 0.5A.

At this setting the noise of the Agilent 3634A current source is found to be <30 mA in load regulation

mode. The voltage measurement accuracy is found to be <3mV within this range(Table 9).

Propagation of these uncertainties using Eq. 59 gives a power input accuracy of <±15.1 mW.

98
 5.5.2.3. Heat Leakage

The system analysis in paragraph 5.4.2 concluded that heat leakage is expected to be less than

±1.3% at a point where coolant and ambient temperature are comparable and the TGP effective

conductivity is 400 W/m-K. Figure 88 is an expansion on the work from paragraph 5.4.2 that

illustrates that the heat leakage percentage increases when the coolant temperature is greater than

the ambient environment temperature.

5
Leak [%]

0
0 10 20 30 40 50 60
T=T_coolant-T_ambient

k_eff = 400 W/m-K k_eff = 2000 W/m-K

Figure 88 Heat leakage (results from FEA model used in paragraph 5.4.2)

It also illustrates the effect of different TGP effective thermal conductivity on heat leakage. Figure 88

was used to correct the amount of heat flowing through the TGP.

5.5.3. Effective Thermal Conductivity Uncertainty

As it is the purpose of the experiment to establish the effective thermal conductivity of the TGP with

less than 10% uncertainty, the heat input and temperature measurement uncertainty can be

propagated to estimate their contribution to effective thermal conductivity uncertainty.

99
The effective thermal conductivity, as defined by Eq. 31, is a linear function of both heat input and

differential temperature measurement combined with geometric factors. This means the relative

uncertainty can be evaluated as

Eq. 60

Which includes the previously evaluated temperature and heat input uncertainties. Geometric input

uncertainties are low as they can be effectively measured by micrometer. Table 14 gives an overview

of the previously evaluated uncertainties. Table 14 illustrates the importance of accuracy of the

differential temperature measurement.

Source Absolute Uncertainty Typical Typical

Value Relative
Uncertainty
Differential Temperature Measurement ±0.78 [C] 10 [C] 7.8%
1: Differential Thermocouple
Differential Temperature Measurement ±0.15 [C] 10 [C] 1.5%
2: Thermopile
Heat input ±0.015 [W] 25 [W] 0.06%

Table 14 Temperature and Heat input Uncertainty

Note that the relative uncertainties change as a function of operating conditions and performance

(effective thermal conductivity) of the TGP. Therefore these uncertainties are evaluated in greater

detail using a numerical model.

100
 Figure 89 Effective Thermal Conductivity Figure 90 Effective Thermal Conductivity
uncertainty and contributors uncertainty and contributors
(keff=400W/m-K, 2 x Thermocouples) (keff=10000W/m-K, Thermopile)

Figure 89 and Figure 90 illustrate numerical evaluation of the contributors to effective thermal

conductivity uncertainty. It was found that, when effective thermal conductivity of TGP is in on the

order of 400 W/m-K, measuring using two thermocouples allows for establishment of effective

thermal conductivity with less than 10% error (about ~3%, Figure 89). However when effective

thermal conductivity of TGP is around 10kW/m-K, the relative contribution error of the temperature

measurement can be 50% if thermocouples were used. If instead a more accurate thermopile

measurement is used, the error can be reduced to less than 8% (Figure 90). The analysis also takes a

spatial thermocouple placement accuracy of ±0.5mm into account. Depending on the local surface

temperature gradient, this factor can be significant as shown in Figure 89.

Further evaluation using this model finds that a switchover point exists at each power level, at which

a 2 thermocouple measurement becomes in sufficient and when one would have to switch to a

thermopile measurement in order to satisfy the objective of less than 10% uncertainty on the

effective thermal conductivity measurement.

101
 Figure 91 Relative effective thermal conductivity uncertainty at 40W and 120W input

Figure 91 illustrates that for heat loads at 40W and 120W, the switchover points for required

measurement technique are at 750 and 2252 W/m-K effective thermal conductivity respectively.

Figure 91 also illustrates the strong impact the temperature measurement has on this experiment. If

even greater accuracy than a thermopile is required, a number of thermopiles can be configured in

parallel to boost the voltage reading or the TGP can be evaluated at higher power input (as long as

the device does not experience dry-out).

5.6. Thermal Characterization Experiment Validation

In order to validate the performance of the TGP Thermal Characterization experiment. Tests were

performed using materials with known thermal conductivity.

102
 Sample Material Dimensions Thermal Measured Modeled
[mm x mm x Conductivity Thermal Thermal
mm] [W/m-K] Resistance [K/W] Resistance –
ICEPAK [K/W]
Copper 3 x 3 x 4.8 390 0.37 0.35

Aluminum Nitrate 3 x 3 x 2.0 170 1.68 1.73

Table 15 Validation Test Samples

These samples were tested in the TGP thermal characterization experiment as described in

paragraph 5.2 at different heat loads. A summary of the test data is presented in Appendix I. The

effective thermal conductivity of the samples was evaluated using the shape factor method as

described in paragraph 2.8. The retrieved effective thermal conductivities were 363 and 175 for the

copper and AlN samples respectively. As some uncertainty remained over the accuracy of the

estimated shape factors, a finite element model was developed in ICEPAK evaluating the full 3D flux

path, The results of the comparison to this model are presented in Figure 92.

100
Temperature rise across sample [C]

80

60

40

20

0
0 10 20 30 40 50
Power [W]
AlN - Icepak model AlN - measurement
Cu - Icepak model Cu - measurement

Figure 92 Heat leakage (results from FEA model used in paragraph 5.4.2)

103
Figure 92 gives confidence that even with the shape factors evaluated from the correlations as

derived in paragraph 2.8, a good match is provided compared to a 3D finite element model. In effect

the measured effective thermal conductivities for the copper and AlN sample were found to have an

error of -8.0% and 3.2% respectively. As this is within the objective of measuring effective thermal

conductivity within 10% error, it is deemed that the experiment is appropriate for measurement of

TGPs with similar effective thermal conductivity values.

5.7. TGP Prototype Thermal Characterization

As the validation effort produced confidence in the capability of the measurement method, a TGP

prototype of 30 mm x 30 mm x 3 mm was tested over a range of heat loads. The substrate material

of this device was made out of AlN and the wick material out of sintered copper. Effective thermal

conductivity was evaluated and the results are presented in Figure 93.

600

500
Effective Conductivity ( W/mK)

400

300

200

100

0
0 10 20 30 40 50 60

Power (W)

Figure 93 3cm TGP Effective Thermal Conductivity

as a function of input power at 15ºC condenser temperature

104
From Figure 93 it can be observed that for low power inputs, the effective thermal conductivity starts

low, then rises to a peak condition around 25W, remains there for a short interval before a decline at

around 40W. The initial startup range is contributed to a similar mechanism as observed in the

evaporator heat transfer experiment shown in Figure 58. In this test range, some sensible

temperature rise of the liquid is anticipated to take place in order to overcome the required

superheat. This causes more heat to diffuse to the condenser by the parallel conduction path, hence

resulting in a lower effective thermal conductivity. The peak region is the area with the highest heat

transfer coefficient in the evaporator. This is the area where the mass transport is high but where the

evaporator is fully flooded maximizing the area of evaporation. In the last section of the experiment

the onset of dryout can be observed similar to Figure 58. As the liquid layer in the evaporator reduces

to the point of dryout, additional heat will cause the local evaporator temperature to rise, increasing

the heat transported by the diffusion path as the device looks to find a new thermo- and

hydrodynamic equilibrium. As the heat load further increases the device dries out more and the

effective thermal conductivity reduces. The device was not tested at higher heat loads as the

operating limit of the insulation was reached. It is suggested to evaluate a high temperature capable

fixture in subsequent work.

5.8. High-g Thermal Performance Characterization Experiment Development

The key focus of the testing team in Phase II was testing the TGP thermal performance at high-g with

the centrifuge based at the Air Force Research Lab(AFRL).

5.8.1. High-G Thermal Performance Characterization Hardware Description

The hardware that was mounted on the centrifuge mimicked to a great extent the hardware

developed for the stationary test. Liquid cooling was available on the centrifuge. A slip ring provided

electrical connection from to the rotating centrifuge table. Temperatures were measured using

calibrated thermocouples and an on-table reference junction. Signals were amplified and connected

105
through a slip ring before being acquired by the data acquisition system. An improved TGP sample

fixture was developed for the high-G experiment as shown in Figure 94.

Figure 94: High-g TGP Sample fixture

The shape ensures all mechanical clamping load is forced through the TGP sample. Load is

measured with a load cell placed on top of the stack. As thermal interface material performance can

be a strong function of contact load, contact load is ensured to be within specifications before each

run by use of this load cell.

Initial attempts to power the heater on the centrifuge through the slip ring ran into difficulties as it

was found challenging to provide a clean power signal. Improvements were suggested in the form of

acquisition of a similar power supply to the setup(Figure 95) and improved thermocouple calibrations.

106
 5W spin, acceleration up to 6g

14 18
16
12
14

Voltage [V], Current [A]

10
12 power watt

Power [W]
8 10 voltage v
6 8 current i
6
4
4
2
2
0 0
0 50 100 150 200 250 300
Time [min]

Figure 95: Power Supply on Table Figure 96: Power Supply performance under
acceleration

As a result, reliable and accurate power supply was available for the experiment(Figure 96). Labview

software was developed by the AFRL team to control power and measure temperatures and

acceleration of the setup.

Insulated test enclosure

Accelerometer

Heater

Cold plate
Power supply

Figure 97: Test enclosure on centrifuge table Figure 98: Heater cold plate configuration on centrifuge

An enclosure, comprised of an aluminum frame with plastic walls covered with Styrofoam, was built

around the fixture(Figure 97). The test enclosure also minimized convective losses to the exterior. At

high-g acceleration, a significant amount of air flowed around the test enclosure. The accelerometer

was placed at a smaller radius than the test sample(Figure 98). Figure 98 also illustrates how the

condenser is placed at a larger radius than the evaporator. This tests the vapor chamber in the worst

case configuration such that liquid has to wick against gravity back to the evaporator.

107
 5.8.2. High-g Thermal Performance Characterization Experiment Validation

In order to validate performance of the test fixture, a test run was performed using a solid copper

sample with known conductivity (Figure 99).

Copper 110 validation test

400
Thermal conductivity [W/m-K]

375

350

325

300
0 2 4 6 8 10
Acceleration [g]

AFRL measurement GRC Measurement Supplier data

Figure 99: Test fixture on centrifuge validation

As anticipated, the conductivity of a copper sample does not change with increased acceleration

giving confidence in the capability of the test setup to measure effective thermal conductivity within

reasonable accuracy. From the validation test it was concluded that aforementioned accuracy is not

impacted by the acceleration of the test fixture and sample.

5.9. High-G TGP Thermal Performance Characterization

High-g thermal tests were performed in February 2010. Performance of three samples is shown in

Figure 100.

108
 500 500
436 W/m-K
450 450
Effective Thermal Conductivity [W/m-K]

Effective Thermal Conductivity [W/m-K]

400 400

350 350

300 300

250 250

200 200

150 150

100 100

50 50

0 0
0.0 2.0 4.0 6.0 8.0 10.0 10 15 20 25 30 35 40 45 50
Acceleration [g] Pow er input [W]
Sample 1 Sample 2 Sample 3 Sample 1 Sample 2 Sample 3

Constant Power (10W) acceleration to 10+g’s At 10+g’s power increased until dryout
Figure 100: High-G TGP Performance Results

All devices operated without problem at 10 W load up to >10 g’s. Effective thermal conductivity did

not change as a function of g-load, which was the expected result, as effective thermal conductivity

is based on the individual resistances of the components of the TGP as well as the heat transfer

coefficients in the evaporator and condenser regions and should change only minimally with g-

forces. (It is expected, however, that the maximum heat transport capability reduces as the device is

accelerated as the capillary forces are opposed by the high-g forces). The samples achieved effective

thermal conductivities in the range of 225-436 W/m-K. Operation just below dry-out reached the

greatest effective axial thermal conductivity. Heat fluxes were observed as high as 15-20 W/cm2 at

>10 g. Heat input was limited to 50 W as to not exceed safety limits. At this heat transport rate and

10 g’s there was no indication that the TGPs were nearing the capillary limit for samples 2 and 3

(Figure 100). This demonstrates the potential for future heat flux increases, especially at lower g

forces.

109
6. Subsequent Work and Application

The fundamental empirical experiments and modeling methods that are developed in this work have

also proven useful for other researchers. Ababneh et al.[34], utilized inputs from the wick thermal

conductivity measurements to prove the validity of the Maxwell’s equation (Eq. 22). In addition,

Ababneh et al.[19] demonstrated the effectiveness in evaluation of wick permeability using Hagen-

Poiseuille relation with a modified constant. Follow-up measurements of effective evaporator heat

transfer performance using the setup as described in paragraph 3.5 yielded a relation for the

evaporative heat transfer coefficient as a function of superheat as shown in Figure 101.

Figure 101: hevap as a function of ΔT, adapted from Ababneh et al.[35]

Using the empirically evaluated evaporative heat transfer coefficient, Ababneh et al.[35]

demonstrated the ability to de-couple the temperature and velocity fields within the liquid saturated

wick resulting in a simplified finite element model for evaluation of TGP performance (Figure 102).

110
 Figure 102: FEA model temperature contour of 15 cm TGP,
adapted from Ababneh et al.[35]

Chamarthy and Shauhan[35] experimentally characterized thermal performance of 15 cm TGP

prototypes using a variant of the thermal characterization experiment described in paragraph 5.2.

The 15 cm TGP was evaluated to have an effective axial thermal conductivity of 5000 W/m-K[35].

Figure 103: Comparison of model and experimental data for 15cm TGP prototype,
adapted from Ababneh et al.[35]

111
Results of these experiments evaluating the geometric thermal profile of a 15cm TGP using a series

of temperature measurements and comparison to the aforementioned model showed fair

agreement as shown in Figure 103. Note that additional evaluation of measurement error would be

required to fully assess the quality of agreement of the experimental data with the model. Figure 103

demonstrates that not only absolute measurements were matched with fair agreement, but also

that temperature profiles in the evaporator and condenser agreed with predictions from the model

at different heat loads. The quality of agreement with the model is in part attributed to the empirical

coefficients obtained from experiments that were presented in this work.

Considering that Ababneh made significant simplifications in his model, such as omitting modeling of

the wicking structure or microfluidics, the fair agreement as demonstrated in Figure 103

demonstrates the effectiveness of simplified models with empirical inputs.

This work demonstrates the design and experimental methods for development of a Thermal Ground

Plane that can operate in high-g environment. The presented work opens up the ability for use of

Thermal Ground Planes in applications were high-g force operation is important such as avionics and

space. Furthermore, by identification of the relevant parameters for high-g operation, application of

the developed approach allows for standard heat pipes for consumer electronics applications to be

made nearly insensitive to device orientation.

112
 Figure 104: GE IP Ruggedized Computing System Figure 105: Ruggedized System Typical Thermal
Resistance breakdown

Within the General Electric Company, a large number of applications have the opportunity to benefit

from improved heat spreading as can be provided by the Thermal Ground Plane. De Bock[41]

demonstrated that a typical ruggedized computing platform product(Figure 104), such as developed

by GE Intelligent Platforms and GE Aviation Systems, relies significantly on heat spreading to conduct

heat from the internal electronics to the exterior convective surface. Spreading thermal resistance

contributes to a significant portion of the total thermal resistance of such a system(Figure 105).

Utilization of advanced spreading technologies, such as the Thermal Ground Plane, therefore yields

the potential more powerful, faster or smaller systems.

113
7. Conclusion

Efficient heat spreading is crucial for enabling high power density electronic systems. Heat pipes and

vapor chambers play an important role in providing the effective spreading needed as they have the

ability to transport heat at an effective thermal conductivity significantly greater than solid materials.

Until now, due to the nature of the internal physics of such devices, operation under high-g

conditions remained an elusive challenge. In this study a Thermal Ground Plane (TGP) has been

designed, built and tested to overcome this challenge.

The presented work describes and discusses the relevant physics and relationships relevant to

high-g operation of the TGP. These relationships are transformed to develop a design model. The

design model identifies a design space for devices to operate under high-g conditions.

Characterization experiments were developed and executed to empirically establish component

performance. The bubble point method was found to be the most effective method for establishment

of the capillary performance of micro-nano wick structures. Laser flash was used to evaluate the

effective thermal conductivity of the wicking structure. A novel “open heat-pipe” evaporator heat

transfer performance experiment was developed to evaluate evaporation performance of a

saturated wicking structure at different heat loads. In addition, a novel visual validation experiment

was performed demonstrating high-g wick saturation using UV fluorescent dye.

A detailed TGP thermal characterization experiment was developed. Significant effort was spent on

evaluating the uncertainty and capability of the setup. After validation using known thermal

conductivity samples, TGP devices were tested. Prototype TGP devices demonstrated effective axial

thermal conductivity of 461 W/m-K, exceeding the thermal conductivity of solid metals such as

copper. A version of the TGP thermal characterization experiment was developed for operation on

the high-g centrifuge table at AFRL. Operation of this experiment on this table demonstrated a

working TGP with effective axial thermal conductivity up to 436 W/m-K at 10.5 g’s operation.

114
Effective axial thermal conductivity did not prove to be a function of g-force in acceleration from 0 to

10 g’s. This confirms the potential for de-coupling of the thermal and fluid transport effects.

Subsequent work utilized experimental data and testing methods developed in this work to develop a

simplified but effective finite element modeling method using empirical inputs. Ababneh et al.[35]

further demonstrated the ability to de-couple the temperature and velocity fields within the liquid

saturated wick using methods developed in this work. The subsequent work demonstrated effective

modeling of a 15 cm TGP prototype, which had an effective axial thermal conductivity of 5000

W/m-K.

Application of heat pipes and vapor chambers to high-g applications has proven to be feasible by

this study. Applications such as avionics, space vehicles and high speed platforms can now utilize

this work for development of TGP systems with the same effective heat pipe and vapor chamber

spreading technology as used by their low-g consumer electronics counterparts.

115
Acknowledgements

This work is supported by DARPA under SSC SD Contract No. N66001-08-C-2008. Any opinions,

findings and conclusions or recommendations expressed in this material are those of the author(s)

and do not necessarily reflect the views of the SSC San Diego

116
References

1. M. Bohr, “The new era of scaling in the SOC world,” presented at the Int. Solid-State Circuits

Conf. [Online]. Available: http://download.intel.com/technology/architecture-silicon/ISSCC_09

plenary_bohr_presentation.pdf

2. Bailey, C., 2008, “Thermal Management Technologies for Electronic Packaging: Current

Capabilities and Future Challenges for Modeling Tools”, Proceedings of Electronics Packaging

Technology Conference (EPTC 2008), Singapore, Singapore.

3. http://www.pantherproducts.co.uk/Articles/CPU/CPU%20Temperatures.shtml

4. Noctua NH-D14.

http://www.noctua.at/main.php?show=productview&products_id=34&lng=en

5. Thermal Ground Plane (TGP) DARPA Solicitation, BAA 07-36.

6. Perkins, L.P., and Buck, W.E., 1892, “Improvements in Devices for the Diffusion or Transference

of Heat,” UK Patent 22,272, London, England.

7. Gaugler, Richard (1944). Heat Transfer Devices. Dayton, Ohio: U.S. Patent Office. pp. 4.

2350348

8. Grover, G.M., Cotter, T.P., and Erikson, G.F., 1964, “Structures of Very High Thermal

Conductivity“, Journal of Applied Physics, Vol. 35, pp. 1190-1191.

9. Gerner, F. M. and Henderson, H. T., 1995, “Liquid metal micro heat pipes for space radiator

applications,” Final Report NASA-CR-199122 15 pp.

10. Henderson, H.T., Gerner, F.M., Medis, P., Parimi, S., Shuja, A., Suh, J., Ponugoti, P., and Ogirala,

K., 2005, “Proof of Concept LHP,” NASA Glenn contract #NNC04CB44C annual report.

117
11. Maydanik, Y.F., 2005, “Loop heat pipes”, Applied Thermal Engineering, 25, pp. 635-657

12. Qu W., Ma H.B. “Theoretical analysis of startup of a pulsating heat pipe”, 2007, International

Journal of Heat and Mass Transfer, 50 (11-12) , pp. 2309-2316.

13. Adamson, A.W., 1990, “Physical Chemistry of Surfaces”, 5th ed., Wiley, New York

14. Peterson, G.P., 1994, “An Introduction to Heat Pipes”, John Wiley & Sons, Inc

15. Peles. Y. “Internal Forced Convection”, Department of Mechanical, Aerospace and Nuclear

Engineering Rensselaer Polytechnic Institute, Heat Transfer Course, Chapter 8.

http://wwwme.nchu.edu.tw/Enter/html/lab/lab516/Heat%20Transfer/chapter_8.pdf

16. Faghri, A., 1995, Heat Pipe Science and Technology, Taylor and Francis, Washington DC.

17. Reay, D.A. and Kew, P. A., 2006, Heat Pipes: Theory Design and Applications. Fifth ed. Oxford,

Great Britain.

18. Chi, S. W., 1976, “Heat pipe Theory and Practice”, McGraw-Hill, New York, USA

19. Ababneh, M.T., 2012, “Novel Charging Station and Computational Modeling for High Thermal

Conductivity Heat Pipe Thermal Ground Planes”, PhD. Dissertation, University of Cincinnati.

20. Maxwell, J.C. “A treatise on electricity and magnetism,” Vol. 1, 3rd Edn. OUP, 1954 (1981),

reprinted by Dover, New York.

21. H. Kozai, H. Imaura, and K. Takashima, 1989, “The effective Thermal Conductivity of Screen

Wicks,” symposium 3rd international heat pipe symposium, Tsukuba, Japan, pp. 104-109.

22. R. Ranjan, J.Y. Murthy and S. V. Garimella, 2009, “Analysis of the Wicking and Thin-Film

Evaporation of Microstructures”, Journal of Heat Transfer, Vol. 131.

118
23. R. Ranjan. J.Y. Murthy and S. V. Garimella, “A micro-scale model for thin-film evaporation in

capillary wick structures” International Journal of Heat and Mass Transfer 54 (2011) 169–179

24. de Bock H P et al 2011 Development and experimental validation of a micro/nano thermal

ground plane Proc. of ASME/JSME 8th Thermal Engineering Joint Conf. AJTEC 2011 (Honolulu,

HI, USA,) pp T10249–T10249-7

25. Tadanaga, Kiyoharu, et al. "Superhydrophobic-superhydrophilic micropatterning on flowerlike

alumina coating film by the sol-gel method." Chemistry of materials 12.3 (2000): 590-592.

26. Chiou, Nan-Rong, et al. "Growth and alignment of polyaniline nanofibres with

superhydrophobic, superhydrophilic and other properties." Nature nanotechnology 2.6 (2007):

354-357.

27. de Bock, H., Kripa Varanasi, Pramod Chamarthy, Tao Deng, Ambarish Kulkarni, B. Rush, B.

Russ, S. Weaver, and F. Gerner. "Experimental investigation of micro–nano heat pipe wick

structures." In Proc. ASME IMECE, pp. 1-6. 2008.

28. Iverson, B.D., Davis, T.W., Garimella, S.V., North, M.T., Sukhvinder, S.K., 2007 “Heat and Mass

Transport in Heat Pipe Wick Structures,” Journal of Thermophysics and Heat Transfer, Vol. 21,

No. 2, April–June 2007

29. Faghri, A, 1995, Heat Pipe Science and Technology, Taylor and Francis, New York

30. Semenic, T, Lin, Y,Y, Catton, I, 2008, “Thermophysical Properties of Biporous Heat Pipe

Evaporators”, Journal of Heat Transfer, Vol. 130, Issue 2, 022602, February 2008.

31. Adkins, D.R., Moss, T., 1990, “Measuring Flow Properties of Wicks for heat Pipe Solar

Receivers,” Solar Engineering

119
32. Adkins, D. R., and Dykhuizen, R. C., “Procedures for Measuring the Properties of Heat-Pipe

Wick Materials,” 28th Intersociety Energy Conversion Engineering Conference, Society of

Automotive Engineers, Warrendale, PA, Vol. 2, 1993, pp. 911–917.

33. Feser, J.P., Prasad, A.K., Advani, S.G., “Experimental characterization of in-plane permeability

of gas diffusion layers”, Journal of Power Sources 162 (2006) 1226-1231

34. Ababneh, M.T., Gerner, F.M. et. al 2012, “Thermo-Fluid Model for High Thermal Conductivity

Thermal Ground Planes,” Proceedings of the ASME 2012 3rd Micro/Nanoscale Heat & Mass

Transfer International Conference, March 3-6, 2012, Atlanta, GA, USA.

35. Ababneh, M.T., Chamarthy, P., Chauhan, S., Gerner, F.M., De Bock, P., Deng, T., “Thermal

Modeling For High Thermal Conductivity Thermal Ground Planes”, Proceedings of the ASME

2012 Summer Heat Transfer Conference, HT2012, July 8-12, 2012, Rio Grande, Puerto Rico.

36. Pramod Chamarthy, H. Peter J. de Bock, Boris Russ, et al. 2009, “Novel Fluorescent

Visualization method to characterize transport properties in micro/nano heat pipe wick

structures”, Proceedings of IPACK2009, San Francisco, CA, USA

37. H. Peter J. de Bock, Shakti Chauhan, Pramod Chamarthy, et al., 2010, “On the Charging and

Thermal Characterization of a micro/nano structured Thermal Ground Plane”, Itherm 2010,

Las Vegas, NV, USA

38. Marotta, Egidio E., Steven J. Mazzuca, and Julian Norley. "Thermal joint conductance for

flexible graphite materials: analytical and experimental study." Components and Packaging

Technologies, IEEE Transactions on 28.1 (2005): 102-110.

39. The Omega® Temperature Measurement Handbook™ and Encyclopedia, Vol. MMV™ 5th

Edition.

120
40. National Institute of Standards and Technology (NIST). T-type Thermocouple Data Table

http://srdata.nist.gov/its90/download/type_t.tab

41. De Bock, H.P.J., “Avionics Thermal Management”, presentation at Association for Unmanned

Vehicle Systems International(AUVSI) Conference, August 11, 2012, Las Vegas, NV.

42. Ababneh, M.T., Gerner, F.M., Hurd, D., De Bock, P., Chauhan, S., Deng, T., 2011, “Charging

Station of a Planar Miniature Thermal Ground Plane,” Proceedings of the ASME/JSME 2011

8th Thermal Engineering Joint Conference, Honolulu, Hawaii, USA.

121
Appendix A: Fricton factor coefficient for non-circular geometries

Table 16 Friction factor and Nusselt numbers for different geometries (adapted from [15])

122
Appendix B: TGP Design for 10G operation EES model
DELTAP_cap=2*sigma/r_pore
DELTAP_liq=(32*mu_l*mdot*L)/(rho_l*A_l*epsilon*d_pore^2)
DELTAP_vap=(32*mu_v*mdot*L)/(rho_v*A_v*d_chan^2)
DELTAP_bf=mp*rho_l*g*L

DELTAP_cap=DELTAP_liq+DELTAP_vap+DELTAP_bf

fluid$ = 'steam_iapws'
T0 = 293 [K] {T0 is used for fluid properties}

{evaporator}
Q = mdot * h_fg

h0=enthalpy(fluid$,T=T0,x=0.0) {at the evaporator}

h1=enthalpy(fluid$,T=T0,x=1.0)
h_fg = h1-h0

sigma = SURFACETENSION(fluid$, T=T0)

mu_l = VISCOSITY(fluid$,T=T0,x=0.0)
rho_l = DENSITY(fluid$,T=T0,x=0.0)

mu_v = VISCOSITY(fluid$,T=T0,x=1.0)
rho_v = DENSITY(fluid$,T=T0,x=1.0)

g = 9.81 [m/s^2]
{mp = 10 [-]}

{r_pore=1E-5 [m]}
r_pore=d_pore/2
L=0.03[m]
epsilon = 0.5 [-]
A_v=1E-3[m]*26E-3[m]
A_l=A_v/2
d_chan=26E-3[m]
d_pore=0.41*d_part
{d_part=75E-6[m]}

D*L=9.76*sigma/(rho_l*g*mp)
DP=0.41*D

Solution:

123
Appendix C: EES Finite difference model for Shape Factor derivation
“ 2D Numerical conduction model (im x jm nodes HxW)”
ly_star = ly/lin
lx=30[mm]*Convert('mm','m')
im=10
jm1=10
jm2=20
jm3=30
jm=jm3

dy=ly/im
dx=lx/jm
dz=30[mm]*Convert('mm','m')
l_in_star=1/lin_over_lx
lin_over_lx=2*lin/lx

dx_io=lin/jm1
dx_c=(lx-2*lin)/(jm2-jm1)
dx_m = dx_io/2+dx_c/2

T_wall = 293 [K]

q_flux_cm2 = 10 [W/cm2]

q_flux_left = 0 [W/m2]
q_flux_right = 0 [W/m2]

q_flux_bot1 = q_flux_cm2*Convert('W/cm2','W/m2') “constant heatflux inlet BC”

q_flux_bot2 = 0 [W/m2]
{q_flux_bot3 = 0 [W/m2]}

duplicate j=jm3,jm3 “constant temperature outlet BC”

duplicate i=0,0
T[i,j]=T_wall
end
end
“adiabatic top”
q_flux_top1 = 0 [W/m2]
q_flux_top2 = 0 [W/m2]
q_flux_top3 = 0 [W/m2]

k = 40 [W/m-K]

"
+---------+
^ dy |
i | |
+-----dx--+
j->

"
{finite difference model}
"left"
duplicate j=0,0
duplicate i=1,im-1
q_bottom[i,j] = k*dx_io/2*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_io/2*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = q_flux_left*dy*dz
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
"center nodes"
duplicate j=1,jm1-1
duplicate i=1,im-1
q_bottom[i,j] = k*dx_io*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_io*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm1,jm1
duplicate i=1,im-1
q_bottom[i,j] = k*dx_m*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_m*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm1+1,jm2-1

124
 duplicate i=1,im-1
q_bottom[i,j] = k*dx_c*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_c*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm2,jm2
duplicate i=1,im-1
q_bottom[i,j] = k*dx_m*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_m*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm2+1,jm3-1
duplicate i=1,im-1
q_bottom[i,j] = k*dx_io*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_io*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
"right wall"
duplicate j=jm3,jm3
duplicate i=1,im-1
q_bottom[i,j] = k*dx_io/2*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = k*dx_io/2*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = q_flux_right*dy*dz
end
end

"boundary conditions"
"bottom"
duplicate j=0,0
duplicate i=0,0
q_bottom[i,j] = q_flux_bot1*dx_io/2*dz
q_top[i,j] = k*dx_io/2*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = q_flux_left*dy/2*dz
q_right[i,j] = k*dy/2*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=1,jm1-1
duplicate i=0,0
q_bottom[i,j] = q_flux_bot1*dx_io*dz
q_top[i,j] = k*dx_io*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm1,jm1
duplicate i=0,0
q_bottom[i,j] = q_flux_bot1*dx_io/2*dz+q_flux_bot2*dx_c/2*dz
q_top[i,j] = k*dx_m*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm1+1,jm2-1
duplicate i=0,0
q_bottom[i,j] = q_flux_bot2*dx_c*dz
q_top[i,j] = k*dx_c*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm2,jm2
duplicate i=0,0
q_bottom[i,j] = q_flux_bot2*dx_c/2*dz+q_flux_bot3*dx_io/2*dz
q_top[i,j] = k*dx_m*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm2+1,jm3-1
duplicate i=0,0
q_bottom[i,j] = q_flux_bot3*dx_io*dz
q_top[i,j] = k*dx_io*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io

125
 end
end
duplicate j=jm3,jm3
duplicate i=0,0
q_bottom[i,j] = q_flux_bot3*dx_io/2*dz
q_top[i,j] = k*dx_io/2*dz*(T[i+1,j]-T[i,j])/dy
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = q_flux_right*dy/2*dz
end
end
"top"
duplicate j=0,0
duplicate i=im,im
q_bottom[i,j] = k*dx_io/2*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top1*dx_io/2*dz
q_left[i,j] = q_flux_left*dy/2*dz
q_right[i,j] = k*dy/2*dz*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=1,jm1-1
duplicate i=im,im
q_bottom[i,j] = k*dx_io*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top1*dx_io*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm1,jm1
duplicate i=im,im
q_bottom[i,j] = k*dx_m*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top1*dx_io/2*dz+q_flux_top2*dx_c/2*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm1+1,jm2-1
duplicate i=im,im
q_bottom[i,j] = k*dx_c*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top2*dx_c*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_c
end
end
duplicate j=jm2,jm2
duplicate i=im,im
q_bottom[i,j] = k*dx_m*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top2*dx_c/2*dz+q_flux_top3*dx_io/2*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_c
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm2+1,jm3-1
duplicate i=im,im
q_bottom[i,j] = k*dx_io*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top3*dx_io*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = k*dy*dz/2*(T[i,j+1]-T[i,j])/dx_io
end
end
duplicate j=jm3,jm3
duplicate i=im,im
q_bottom[i,j] = k*dx_io/2*dz*(T[i-1,j]-T[i,j])/dy
q_top[i,j] = q_flux_top3*dx_io/2*dz
q_left[i,j] = k*dy*dz/2*(T[i,j-1]-T[i,j])/dx_io
q_right[i,j] = q_flux_right*dy/2*dz
end
end
"energy balance, all nodes"
duplicate j=0,jm
duplicate i=0,im
y[i,j]=i*dy
q_bottom[i,j]+q_top[i,j]+q_left[i,j]+q_right[i,j]=0
end
end
duplicate i=0,im
duplicate j=0,jm1
x[i,j]=j*dx_io
end
end
duplicate i=0,im
duplicate j=jm1+1,jm2

126
 x[i,j]=jm1*dx_io+(j-jm1)*dx_c
end
end
duplicate i=0,im
duplicate j=jm2+1,jm3
x[i,j]=jm1*dx_io+(jm2-jm1)*dx_c+(j-jm2)*dx_io
end
end

A1 = ly *dz
A2 = lx *dz

T_in_mean = AVERAGE(T[0,0..jm1])
T_out_mean = AVERAGE(T[0,jm2..jm3])

q_in=SUM(q_bottom[0,0..jm1])
q_in=k_eff*A1*(T_in_mean-T_out_mean)/(lx-lin)
q_in=k*A1*(T_in_mean-T_out_mean)/(l_eff)

h=ly

{i=1
j=1
returnX=X[i,j]
returnY=Y[i,j]
returnT=T[i,j]}

{q_in= SUM(q_bottom[0,0..jm1])
jstart=jm2
q_out= SUM(q_bottom[0,jstart..jm])

T_in_mean = AVERAGE(T[0,0..jm1])
T_out_mean = AVERAGE(T[0,jstart..jm])
mean_path = 20[mm]*Convert('mm','m')
mean_path2 = 10[mm]*Convert('mm','m')
k_eff = q_in/A_cross*mean_path/(T_in_mean-T_out_mean)
k_eff2 = q_in/A_cross*mean_path2/(T[0,jm1]-T[0,jm2])

T$='out.csv'
$EXPORT T$ T[0..im,0..jm]
}

127
Appendix D: Shape factor FDM results and correlation derivation

Geometry L_eff Aspect Heat Shapefactor Center L_eff from

from ratio input (eq. 27) Length regression
FDM ratio
Run lin lx ly l_eff h* L* c L_c L_eff_reg
# [m] [m] [m] [m] [-] [-] [-] [m] [m]
1 0.01 0.03 0.0025 0.01689 0.083 0.333 8.27E-01 2.00E-02 1.65E-02
2 0.0075 0.03 0.001875 0.02016 0.063 0.250 8.84E-01 2.25E-02 1.99E-02
3 0.005 0.03 0.00125 0.02344 0.042 0.167 9.31E-01 2.50E-02 2.33E-02
4 0.00375 0.03 0.000938 0.02508 0.031 0.125 9.50E-01 2.63E-02 2.50E-02
5 0.003 0.03 0.00075 0.02606 0.025 0.100 9.61E-01 2.70E-02 2.60E-02
6 0.0015 0.03 0.000375 0.02803 0.013 0.050 9.82E-01 2.85E-02 2.80E-02
7 0.01 0.03 0.005 0.01802 0.167 0.333 8.27E-01 2.00E-02 1.65E-02
8 0.0075 0.03 0.00375 0.02101 0.125 0.250 8.84E-01 2.25E-02 1.99E-02
9 0.005 0.03 0.0025 0.02399 0.083 0.167 9.31E-01 2.50E-02 2.33E-02
10 0.00375 0.03 0.001875 0.02549 0.063 0.125 9.50E-01 2.63E-02 2.50E-02
11 0.003 0.03 0.0015 0.02639 0.050 0.100 9.61E-01 2.70E-02 2.60E-02
12 0.0015 0.03 0.00075 0.02819 0.025 0.050 9.82E-01 2.85E-02 2.80E-02
13 0.01 0.03 0.01 0.02221 0.333 0.333 1.00E+00 2.00E-02 2.00E-02
14 0.0075 0.03 0.0075 0.02416 0.250 0.250 1.00E+00 2.25E-02 2.25E-02
15 0.005 0.03 0.005 0.02607 0.167 0.167 1.00E+00 2.50E-02 2.50E-02
16 0.00375 0.03 0.00375 0.02702 0.125 0.125 1.00E+00 2.63E-02 2.63E-02
17 0.003 0.03 0.003 0.02759 0.100 0.100 1.00E+00 2.70E-02 2.70E-02
18 0.0015 0.03 0.0015 0.02876 0.050 0.050 1.00E+00 2.85E-02 2.85E-02
19 0.01 0.03 0.02 0.03562 0.667 0.333 1.69E+00 2.00E-02 3.39E-02
20 0.0075 0.03 0.015 0.03479 0.500 0.250 1.46E+00 2.25E-02 3.29E-02
21 0.005 0.03 0.01 0.03319 0.333 0.167 1.28E+00 2.50E-02 3.19E-02
22 0.00375 0.03 0.0075 0.0323 0.250 0.125 1.20E+00 2.63E-02 3.14E-02
23 0.003 0.03 0.006 0.03176 0.200 0.100 1.15E+00 2.70E-02 3.12E-02
24 0.0015 0.03 0.003 0.03069 0.100 0.050 1.07E+00 2.85E-02 3.06E-02
25 0.01 0.03 0.04 0.06792 1.333 0.333 3.77E+00 2.00E-02 7.55E-02
26 0.0075 0.03 0.03 0.06341 1.000 0.250 2.85E+00 2.25E-02 6.41E-02
27 0.005 0.03 0.02 0.05423 0.667 0.167 2.11E+00 2.50E-02 5.27E-02
28 0.00375 0.03 0.015 0.04828 0.500 0.125 1.79E+00 2.63E-02 4.70E-02
29 0.003 0.03 0.012 0.04447 0.400 0.100 1.62E+00 2.70E-02 4.36E-02
30 0.0015 0.03 0.006 0.03665 0.200 0.050 1.29E+00 2.85E-02 3.68E-02
31 0.01 0.03 0.08 0.1277 2.667 0.333 9.32E+00 2.00E-02 1.86E-01 outlier
32 0.0075 0.03 0.06 0.1211 2.000 0.250 6.55E+00 2.25E-02 1.47E-01 outlier
33 0.005 0.03 0.04 0.1025 1.333 0.167 4.33E+00 2.50E-02 1.08E-01
34 0.00375 0.03 0.03 0.08801 1.000 0.125 3.38E+00 2.63E-02 8.86E-02
35 0.003 0.03 0.024 0.07747 0.800 0.100 2.85E+00 2.70E-02 7.69E-02
36 0.0015 0.03 0.012 0.05299 0.400 0.050 1.88E+00 2.85E-02 5.35E-02

128
Appendix E: Bubble point validation test results

Test sample date Time Operator dP fluid pore diam Meas. Radius
# # mm-hh-yy hh:mm - PSI - m m
1 mesh 325 1 4/11/2008 2:32 Peter 0.275 water 151.692 75.85
2 mesh 325 1 4/11/2008 2:37 Peter 0.271 water 154.148 77.07
3 mesh 325 1 4/11/2008 2:42 Peter 0.287 water 145.694 72.85
4 mesh 325 1 4/11/2008 2:54 Peter 0.265 water 157.987 78.99
5 mesh 150
6 mesh 150 1 4/11/2008 2:59 Peter 0.188 water 222.658 111.33
7 mesh 150 1 4/11/2008 3:03 Peter 0.135 water 308.713 154.36
8 mesh 150 1 4/11/2008 3:06 Peter 0.128 water 327.290 163.65
9 mesh 325 1 4/11/2008 3:11 Peter 0.281 water 148.923 74.46
10 mesh 325 1 4/11/2008 3:15 Peter 0.301 water 138.79 69.40
11 mesh 325 2 4/14/2008 11:03 Peter 0.196 water 213.503
12 mesh 325 2 4/14/2008 11:06 Peter 0.388 water 107.657
13 mesh 325 2 4/14/2008 11:09 Peter 0.373 water 112.094
14 mesh 325 2 4/14/2008 11:11 Peter 0.291 water 143.678
15 mesh 325 2 4/14/2008 11:14 Peter 0.311 water 134.224
16 mesh 150 2 4/14/2008 11:17 Peter 0.119 water 351.459
17 mesh 150 2 4/14/2008 11:19 Peter 0.134 water 312.941
18 mesh 150 2 4/14/2008 11:21 Peter 0.220 water 190.057
19 mesh 150 2 4/14/2008 11:24 Peter 0.124 water 337.942
20 mesh 150 2 4/14/2008 11:28 Peter 0.145 water 287.718
21 mesh 100 2 4/14/2008 11:32 Peter 0.115 water 362.615
22 mesh 100 2 4/14/2008 11:34 Peter 0.137 water 305.412
23 mesh 100 2 4/14/2008 11:35 Peter 0.121 water 344.047
24 mesh 100 2 4/14/2008 11:37 Peter 0.096 water 435.970
25 mesh 100 2 4/14/2008 11:39 Peter 0.138 water 302.982
26 mesh 80 2 4/14/2008 11:41 Peter 0.086 water 484.004
27 mesh 80 2 4/14/2008 11:43 Peter 0.111 water 375.736
28 mesh 80 2 4/14/2008 11:45 Peter 0.113 water 369.657
29 mesh 80 2 4/14/2008 11:47 Peter 0.09 water 462.444
30 mesh 80 2 4/14/2008 11:49 Peter 0.107 water 391.179
31 AA0 200nm 1 4/14/2008 4:46 Peter water
32 AA0 200nm 1 4/14/2008 Peter 26.166 IPA 0.506
33 AA0 200nm 1 4/14/2008 Peter water
34 AA0 200nm 1 4/14/2008 Peter water
35 AA0 200nm 1 4/14/2008 Peter water
37 325 mesh 2 4/14/2008 5:19 Peter 0.144 IPA 92.038
38 325 mesh 2 4/14/2008 5:22 Peter 0.176 IPA 75.199
39 325 mesh 2 4/14/2008 5:24 Peter 0.163 IPA 81.283
40 325 mesh 3 4/15/2008 10:08 Peter 0.175 IPA 75.513
41 325 mesh 3 4/15/2008 10:10 Peter 0.195 IPA 67.736
42 325 mesh 3 4/15/2008 10:13 Peter 0.200 IPA 66.126
43 325 mesh 3 4/15/2008 10:15 Peter 0.195 IPA 67.736
44 325 mesh 3 4/15/2008 10:17 Peter 0.194 IPA 68.119
45 325 mesh 3 4/15/2008 10:19 Peter 0.205 IPA 64.476
46 325 mesh 3 4/15/2008 10:20 Peter 0.202 IPA 65.409

129
47 325 mesh 3 4/15/2008 10:22 Peter 0.176 IPA 75.044
48 325 mesh 3 4/15/2008 10:26 Peter 0.198 IPA 66.736
49 325 mesh 3 4/15/2008 10:29 Peter 0.207 IPA 63.906
50 150 mesh 3 4/15/2008 10:35 Peter 0.068 IPA 194.467
51 150 mesh 3 4/15/2008 10:37 Peter 0.087 IPA 151.979
52 150 mesh 3 4/15/2008 10:39 Peter 0.086 IPA 153.919
53 150 mesh 3 4/15/2008 10:41 Peter 0.072 IPA 182.683
54 150 mesh 3 4/15/2008 10:43 Peter 0.089 IPA 148.243
55 150 mesh 3 4/15/2008 10:45 Peter 0.085 IPA 155.241
56 150 mesh 3 4/15/2008 10:47 Peter 0.088 IPA 150.087
57 150 mesh 3 4/15/2008 10:48 Peter 0.080 IPA 165.922
58 150 mesh 3 4/15/2008 10:51 Peter 0.081 IPA 162.932
59 150 mesh 3 4/15/2008 10:52 Peter 0.092 IPA 144.106

130
Appendix F: Wick on copper IPA test results

Test date Operator dP fluid pore diam Meas. Radius

# mm-hh-yy - Pa - m m
1 080711_cu75um_p_cudisk_2 7/21/2008 Peter 26.8 IPA 7.266 3.63
2 080711_cu75um_p_cudisk_2 7/21/2008 Peter 22.8 IPA 8.547 4.27
3 080711_cu75um_p_cudisk_2 7/21/2008 Peter 18.3 IPA 10.623 5.31
4 080711_cu75um_p_cudisk_2 7/21/2008 Peter 23.5 IPA 8.283 4.14
5 080711_cu75um_p_cudisk_2 7/21/2008 Peter 22.4 IPA 8.685 4.34
6 080711_cu75um_p_cudisk_2 7/21/2008 Peter 20.8 IPA 9.371 4.69

131
Appendix G: Discrete Wick Permeability for Square Wick with Center Inlet
fluid$='air'

{T_avg=293[K]}
{P_out=101325 [Pa]}
{mu = viscosity(fluid$,T=T_avg,P=P_out)}
{DELTAP=800 [Pa]}
{k = 1 [m2]}
{rho =density(fluid$,T=T_avg,P=P_out)}
{mdot_in = 1 [kg/s]}

mdot_source[mid,mid]= mdot_in

"Blake-Kozeny equation - Chi 1976"

k_bk = (r_s^2*epsilon^3)/(37.5*(1-epsilon)^2)

"
+---------+
^ dy |
i | |
+----dx--+
j - >

"
nn=24 " must be even number greater than 0 less than 42"

l_x=1 [in]*Convert('in','m')
l_y=l_x
l_z=1e-3[m]
dx=l_x/nn
dy=l_y/nn
dz=l_z

"center nodes"
duplicate j=1,nn-1 "x"
duplicate i=1,nn-1 "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end
"right wall"
duplicate j=nn,nn "x"
duplicate i=1,nn-1 "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P_out)/dx/2*rho*(dz*dy)
end
end
"right top corner"
duplicate j=nn,nn "x"
duplicate i=nn,nn "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx/2)
mdot_top[i,j]=(k/mu)*(P[i,j]-P_out)/dy/2*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P_out)/dx/2*rho*(dz*dy)
end
end
"top wall"
duplicate j=1,nn-1 "x"
duplicate i=nn,nn "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P_out)/dy/2*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end
"top left corner"
duplicate j=0,0 "x"
duplicate i=nn,nn "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P_out)/dy/2*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P_out-P[i,j])/dx/2*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end

132
"left wall"
duplicate j=0,0 "x"
duplicate i=1,nn-1 "y"
mdot_bot[i,j]=(k/mu)*(P[i-1,j]-P[i,j])/dy*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P_out-P[i,j])/dx/2*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end
"left bottom corner"
duplicate j=0,0 "x"
duplicate i=0,0 "y"
mdot_bot[i,j]=(k/mu)*(P_out-P[i,j])/dy/2*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P_out-P[i,j])/dx/2*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end
"bottom center"
duplicate j=1,nn-1 "x"
duplicate i=0,0 "y"
mdot_bot[i,j]=(k/mu)*(P_out-P[i,j])/dy/2*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P[i,j+1])/dx*rho*(dz*dy)
end
end
"bottom right corner"
duplicate j=nn,nn "x"
duplicate i=0,0 "y"
mdot_bot[i,j]=(k/mu)*(P_out-P[i,j])/dy/2*rho*(dz*dx)
mdot_top[i,j]=(k/mu)*(P[i,j]-P[i+1,j])/dy*rho*(dz*dx)
mdot_left[i,j]=(k/mu)*(P[i,j-1]-P[i,j])/dx*rho*(dz*dy)
mdot_right[i,j]=(k/mu)*(P[i,j]-P_out)/dx/2*rho*(dz*dy)
end
end
"mass flow is conserved"
duplicate j=0,nn "x"
duplicate i=0,nn "y"
mdot_bot[i,j]-mdot_top[i,j]+mdot_left[i,j]-mdot_right[i,j]+mdot_source[i,j]=0
end
end
"center source"
mid=(nn+2)/2-1
duplicate j=0,nn "x"
duplicate i=0,mid-1 "y"
mdot_source[i,j]=0
end
end
duplicate j=0,nn "x"
duplicate i=mid+1,nn "y"
mdot_source[i,j]=0
end
end
duplicate j=0,mid-1 "x"
duplicate i=mid,mid "y"
mdot_source[i,j]=0
end
end
duplicate j=mid+1,nn "x"
duplicate i=mid,mid "y"
mdot_source[i,j]=0
end
end
mdot_out_top = sum(mdot_top[nn,j],j=0,nn)
mdot_out_bot = sum(mdot_bot[0,j],j=0,nn)
mdot_out_left = sum(mdot_left[i,0],i=0,nn)
mdot_out_right = sum(mdot_right[i,nn],i=0,nn)
mdot_out=mdot_out_top+mdot_out_right-mdot_out_bot-mdot_out_left

P[mid,mid]=P_mid
DELTAP=P_mid-P_out

133
Appendix H: TGP Thermal Performance Experiment model (APDL)
!
! ===============================
! Thermal performance test model
!
! Peter de Bock
!
! GE Global Research
! October 2008
! v1.0
! ===============================
!
! Q_in=120.000 W
! Q_cp=118.453 W
! Q_ex= 1.547 W
! Leak_perc = 1.306%
!
finish
/clear
/config,nproc,4
!
! Boundary conditions
!
Q = 120 ! [W]
htc_free = 5 ! [W/m2-K]
T_amb = 293 ! [K]
T_fluid = 293 ! [K]
!
t_acr = 0.0127! [m]
t_shl = 0.010 ! [m]
t_ins = 0.0127! [m]
t_ins2 = 0.0016! [m]
!
K_cu = 400 ! [W/m-K]
K_TGP = 400 ! [W/m-K]
K_ins = 0.06 ! [W/m-K]
K_shl = 0.20 ! [W/m-K]
K_acr = 1.90 ! [W/m-K]
K_htrs = 40 ! [W/m-K]
!
! Thermal contact conductances
TCC1 = 39370 ! [W/K-m2] = 1/0.2540E-4
TCC2 = K_ins/t_ins2
TCC3 = 12356 ! [W/K-m2] = 1/(0.2540E-4+0.57E-4(coldplate))
TCC4 = K_ins/t_ins2
!
n = 0.5 ! refinement parameter
!
!! Geometry
!
! Heater block
!
h_htr = 0.001 ! [m]
w_htr = 0.010 ! [m]
d_htr = 0.030 ! [m]
!
V_htr = h_htr*w_htr*d_htr
!
h_htrs = 0.001
w_htrs = 0.020 ! [m]
d_htrs = 0.050 ! [m]
!
x_htr = -d_htr/2
y_htr = -w_htr/2
z_htr = 0 ! [m]
!
x_htrs = -d_htrs/2
y_htrs = -w_htrs/2
z_htrs = 0 ! [m]
!
! Cold plate
!
h_clp = 0.003 ! [m]
w_clp = 0.010 ! [m]
d_clp = 0.030 ! [m]
!
h_clps = h_clp
w_clps = 0.033 ! [m]

134
d_clps = 0.050 ! [m]
!
x_clp = -d_clp/2
y_clp = 0.015 ! [m]
z_clp = h_htr-h_clp
!
x_clps = -d_clps/2
y_clps = 0.015 ! [m]
z_clps = h_htr-h_clp
!
! TGP
!
h_TGP1 = 0.003 ! [m]
w_TGP1 = 0.010 ! [m]
d_TGP1 = 0.030 ! [m]
!
h_TGP2 = h_TGP1
w_TGP2 = w_TGP1
d_TGP2 = d_TGP1
!
h_TGP3 = h_TGP1
w_TGP3 = w_TGP1
d_TGP3 = d_TGP1
!
x_TGP1 = -d_TGP1/2
y_TGP1 = -w_htr/2
z_TGP1 = h_htr
!
x_TGP2 = -d_TGP1/2
y_TGP2 = -w_htr/2+w_TGP1
z_TGP2 = h_htr
!
x_TGP3 = -d_TGP1/2
y_TGP3 = -w_htr/2+w_TGP1+w_TGP2
z_TGP3 = h_htr
!
! Shell
!
h_shl = h_clp+h_TGP1+2*t_shl
w_shl = w_TGP1*2+2*t_shl+w_clps
d_shl = d_htrs+2*t_shl
!
x_shl = -d_shl/2
y_shl = -w_htr/2-t_shl
z_shl = h_htr-h_clp-t_shl
!
! Acrylic plate
!
h_acr = t_acr
w_acr = w_shl
d_acr = d_shl
!
x_acr = x_shl
y_acr = y_shl
z_acr = z_shl+h_shl
!
! Cotronics box
!
h_ins = h_shl+h_acr+2*t_ins
w_ins = w_shl+2*t_ins
d_ins = d_shl+2*t_ins
!
x_ins = x_shl-t_ins
y_ins = y_shl-t_ins
z_ins = z_shl-t_ins
!
! ===============================
!
/title,test
/graph,power
/triad,lbot
/view,1,0.5,-0.8, 0.326
/ANG, 1, -52
/PSYMB,XNODE,0
/auto
/pnum,type,1
/num,1
/prep7
!
n,1e6,x_ins - w_ins*0.5,y_ins + d_ins*0.5,z_ins + h_ins*0.5
n,2e6,x_ins + w_ins*0.5,y_ins - d_ins*0.5,z_ins + h_ins*0.5

135
n,3e6,x_ins + w_ins*1.5,y_ins + d_ins*0.5,z_ins + h_ins*0.5
n,4e6,x_ins + w_ins*0.5,y_ins + d_ins*1.5,z_ins + h_ins*0.5
n,5e6,x_ins + w_ins*0.5,y_ins + d_ins*0.5,z_ins + h_ins*1.5
n,6e6,x_ins + w_ins*0.5,y_ins + d_ins*0.5,z_ins - h_ins*0.5
!
! ===============================
! 1,2
BLOCK,x_htr ,x_htr +d_htr ,y_htr ,y_htr +w_htr ,z_htr ,z_htr +h_htr
BLOCK,x_htrs,x_htrs+d_htrs,y_htrs,y_htrs+w_htrs,z_htrs,z_htrs+h_htrs
! 3,4
BLOCK,x_clp ,x_clp +d_clp ,y_clp ,y_clp +w_clp ,z_clp ,z_clp +h_clp
BLOCK,x_clps,x_clps+d_clps,y_clps,y_clps+w_clps,z_clps,z_clps+h_clps
! 5,6,7
BLOCK,x_TGP1,x_TGP1+d_TGP1,y_TGP1,y_TGP1+w_TGP1,z_TGP1,z_TGP1+h_TGP1
BLOCK,x_TGP2,x_TGP2+d_TGP2,y_TGP2,y_TGP2+w_TGP2,z_TGP2,z_TGP2+h_TGP2
BLOCK,x_TGP3,x_TGP3+d_TGP3,y_TGP3,y_TGP3+w_TGP3,z_TGP3,z_TGP3+h_TGP3
! 8,9,10
BLOCK,x_shl ,x_shl +d_shl ,y_shl ,y_shl +w_shl ,z_shl ,z_shl +h_shl
BLOCK,x_acr ,x_acr +d_acr ,y_acr ,y_acr +w_acr ,z_acr ,z_acr +h_acr
BLOCK,x_ins ,x_ins +d_ins ,y_ins ,y_ins +w_ins ,z_ins ,z_ins +h_ins
!
VSBV, 2, 1,,DELETE,KEEP ! --> 11 htrs
VSBV, 4, 3,,DELETE,KEEP ! --> 2 clp
VADD, 5, 6,7 ! group TGP --> 4
VADD, 3, 2 ! group clp --> 5

VSBV,10, 9,,DELETE,KEEP ! substract acr from ins --> 2

VSBV, 2, 8,,DELETE,KEEP ! substract shl from ins --> 3

VSBV, 8, 4,,DELETE,KEEP ! substract tgp from shl --> 2

VSBV, 2, 1,SEPO,DELETE,KEEP ! substract htr from shl --> 6
VSBV, 6,11,SEPO,DELETE,KEEP ! substract htrs from shl --> 2
VSBV, 2, 5,SEPO,DELETE,KEEP ! substract clp from shl --> 6

! 1 htr
! 3 ins
! 4 TGP
! 5 coldplate
! 6 shell
! 9 acr
!11 htrs

VGLUE,6,9,3,4 ! glue shell,acr,ins & TGP

VGLUE,1,11 ! glue htr & htrs

! 1 htr
! 2 TGP
! 3 ins
! 5 coldplate
! 7 shell
! 9 acr
!11 htrs
!
! ===============================
!
et,1,conta174,2
et,2,targe170
et,3,solid87
et,4,152
keyopt,4,4,1 ! lower order
keyopt,4,5,1 ! use extra node, bulk temp
keyopt,4,8,2 ! use convection only
et,5,152
keyopt,5,4,1 ! lower order
keyopt,5,5,1 ! use extra node, bulk temp
keyopt,5,8,2 ! use convection only
!
!Thermal contact resistance
!
R,1, ! interface 1: TGP - htr
RMORE,
RMORE,,TCC1
R,2, ! interface 2: htr - shell
RMORE,
RMORE,,TCC2
R,3, ! interface 2: TGP - clp
RMORE,
RMORE,,TCC3
R,4, ! interface 2: clp - shell
RMORE,
RMORE,,TCC4

136
!
! Material properties
!
mp,kxx,1,K_cu
mp,kxx,2,K_TGP
mp,kxx,3,K_ins
mp,kxx,4,K_shl
mp,kxx,5,K_acr
mp,kxx,6,K_htrs
!
! ===============================
! 1 htr
! 2 TGP
! 3 ins
! 5 coldplate
! 7 shell
! 9 acr
!11 htrs
!
type,3
real,6
esize,h_htr/n
!
! heater
mat,1
vmesh,1
esize,0.001/n
!
! TGP
mat,2
vmesh,2
!
! clp
mat,1
vmesh,5
!
! htrs
mat,6
vmesh,11
!
esize,0.005/n
!
! shell
mat,4
vmesh,7
!
! acr
mat,5
vmesh,9
!
esize,0.01/n
mat,3
vmesh,3
!
! ===============================
! Interfaces
!
! Interface 1: TGP (A25) -> HTR (A2)
!
TYPE,1 ! CONTACT ELEMENTS
REAL,1
!
asel,s,area,,25
nsla,s,1
esurf
!
TYPE,2 ! TARGET ELEMENTS
REAL,1
!
asel,s,area,,2
nsla,s,1
esurf
!
! Interface 2: HTR,HTRS (A1,A61) -> SHELL (A75,A113)
!
TYPE,1 ! CONTACT ELEMENTS
REAL,2
!
asel,s,area,,1
asel,a,area,,61
nsla,s,1

137
esurf
!
TYPE,2 ! TARGET ELEMENTS
REAL,2
!
asel,s,area,,75
asel,a,area,,113
nsla,s,1
esurf
!
! Interface 3: TGP (A66) -> CLP (A14)
!
TYPE,1 ! CONTACT ELEMENTS
REAL,3
!
asel,s,area,,66
nsla,s,1
esurf
!
TYPE,2 ! TARGET ELEMENTS
REAL,3
!
asel,s,area,,14
nsla,s,1
esurf
!
! Interface 4: CLP (A13,A63) -> HTR (A39,A110)
!
TYPE,1 ! CONTACT ELEMENTS
REAL,4
!
asel,s,area,,13
asel,a,area,,63
nsla,s,1
esurf
!
TYPE,2 ! TARGET ELEMENTS
REAL,4
!
asel,s,area,,39
asel,a,area,,110
nsla,s,1
esurf
!
! Convection elements
type,4
real,7
!
asel,s,area,,55
nsla,s,1
esurf,6e6
!
asel,s,area,,56
nsla,s,1
esurf,5e6
!
asel,s,area,,57
nsla,s,1
esurf,2e6
!
asel,s,area,,58
nsla,s,1
esurf,4e6
!
asel,s,area,,59
nsla,s,1
esurf,3e6
!
asel,s,area,,60
nsla,s,1
esurf,1e6
!
!
! ===============================
! 1 htr
! 2 TGP
! 3 ins
! 5 coldplate
! 7 shell
! 9 acr
!11 htrs

138
!
! Loads
bfv,1,hgen,Q/V_htr
!
vsel,s,,,5
nslv,s,1
D,ALL,TEMP,T_fluid
!
esel,s,type,,4
sfe,all,1,conv,1,htc_free
!
allsel
d,1e6,temp,T_amb
d,2e6,temp,T_amb
d,3e6,temp,T_amb
d,4e6,temp,T_amb
d,5e6,temp,T_amb
d,6e6,temp,T_amb
!
! ===============================
!
/solu
allsel,all
SOLVE
FINISH

/post1
allsel,all
plnsol,temp

wplane,1,0,0,0,0,y_clp,0,0,0,h_htr
sucr,cut1,cplane
susel,s,cut1
sumap,temp,temp,sum
supl,cut1,temp,1

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! exterior walls
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
allsel,all
esel,s,type,,4
etable,area,nmisc,1
etable,hfctot,smisc,2
etable,hflux,smisc,3
smult,flux,area,hfctot

aaa= ! kill the array to be used

*get,xmax,elem,,num,max
*get,xmin,elem,,num,min
*dim,aaa,array,xmax-xmin+1
*vget,aaa(1),elem,xmin,esel
*vmask,aaa(1)
*vget,aaa(1),elem,xmin,geom
etable,hfctot,smisc,2
*vmask,aaa(1)
*vget,aaa(1),elem,xmin,etab,hfctot!
*VSCFUN,hf_ext,sum,aaa(1)
! what is the highest element number selected?
! element number compression will be desirable
! minimum element number
! array to hold areas has to be this big
! fill array with info on whether element is selected
! -1=not selected, 0=undefined, 1=selected
! offset with xmin (see the manual)
! use element selection info as a mask
! fill the array with geom info on the elements
! for shell elements this is AREA. Offset with xmin
! create element table column with element volume
! use array as a mask (geom data is positive or zero)
put data into ETABLE "geom" column. Offset with xmin

hf_ext=hf_ext*1
hf_coldplate=Q-hf_ext

leak_perc=hf_ext/hf_coldplate*100

/eof

139
Appendix I: Thermal Performance Characterization Experiment Validation

Test Date: 08/06/2009 - Sample: Copper

V across Cold Heat Delta

Watts sample Plate er-T T R_tot alpha dT_sample R_sample
0 0.00E+00 20.32 20.35 0.04
1 2.00E-05 20.32 21.08 0.76 0.763017 3.9459E-05 0.51 0.506856
5 7.00E-05 20.34 23.99 3.65 0.729953 3.95198E-05 1.77 0.354253
10 1.40E-04 20.41 27.67 7.25 0.725056 3.95977E-05 3.54 0.353556
20 2.90E-04 20.55 35.03 14.49 0.724253 3.97534E-05 7.29 0.364749
30 4.40E-04 20.72 42.57 21.85 0.728464 3.99133E-05 11.02 0.367463
50 7.40E-04 21.02 57.66 36.65 0.732933 4.02326E-05 18.39 0.36786
75 1.13E-03 21.38 76.50 55.12 0.734951 4.0631E-05 27.81 0.370817
90 1.37E-03 21.57 87.91 66.34 0.737075 4.08717E-05 33.52 0.372439

Test Date: 09/24/2009 - Sampel: AlN

Date

Comment

TC_1

TC_2

TC_3

TC_4

TC_5

Current

Voltage

Power

DELTAT

R
9/24/2009 0W 2mm AlN 4 in. lb 15.3 15.4 15.2 15.0 15.2 0.0 0.0 0.0
9/24/2009 0W 2mm AlN 4 in. lb 24.2 24.4 24.2 15.7 15.2 0.5 9.5 5.0 8.4 1.67
9/24/2009 10W 2mm AlN 4 in. lb 33.2 33.5 33.2 16.5 15.3 0.7 13.5 10.0 16.7 1.66
9/24/2009 15W 2mm AlN 4 in. lb 42.2 42.8 42.3 17.2 15.4 0.9 16.7 15.0 25.1 1.67
18.9W 2mm AlN 4 in.
9/24/2009 lb 49.7 50.4 49.8 17.8 15.4 1.0 18.9 18.9 32.0 1.69

copper test
copper ALN 2mm 2 TC's
Test name thermopile test (insul)
Date 8/6/2009 9/24/2009 12/15/2009
length [m] 3.00E-02 3.00E-02 3.00E-02
width [m] 3.00E-02 3.00E-02 3.00E-02
thickness [m] 0.004763 2.00E-03 0.0047625
k [W/m-k] 390 170 390
R_measured [K/W] 0.371 1.679 0.583

R_icepak [K/W] 0.345 1.731

R_simple [K/W] 0.538 2.941

Shape_factor [-] 0.638 0.590
R_corr [K/W] 0.343 1.735

Error meas [%] -8.0% 3.2%

Error icepak [%] -0.4% 0.3%
k_measured 362.8 175.2

140