Академический Документы
Профессиональный Документы
Культура Документы
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
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
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.
In these moments, we are blessed when there are others that believe in you,
Thank you to all the people that supported me and the TGP program on this Journey.
DARPA
University of Cincinnati
iv
Table of Contents
Abstract ................................................................................................................................................................................................... ii
Preface .................................................................................................................................................................................................... iv
Table of Contents................................................................................................................................................................................ v
List of Figures........................................................................................................................................................................................ x
1. Introduction.................................................................................................................................................................................1
2.1. Introduction.......................................................................................................................................................................7
v
2.4.6. Pressure Drop due to Body Forces ........................................................................................................... 21
3.1. Introduction.................................................................................................................................................................... 45
3.3.3. Discrete model for Square Shaped Wick Permeability Experiment ............................................... 58
vi
3.3.4. Wick Permeability Results .................................................................................................................................. 60
4.1. Introduction.................................................................................................................................................................... 74
5.1. Introduction.................................................................................................................................................................... 81
vii
5.4. Thermal Model of Experiment ............................................................................................................................... 85
viii
7. Conclusion ............................................................................................................................................................................. 114
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 G: Discrete Wick Permeability for Square Wick with Center Inlet .................................................... 132
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
Figure 11 Fluid meniscus sketch for sintered sphere wicking structure .............................................................. 16
Figure 13 Permeability constant models ( = 0.5) (adapted from Ababneh [19]) Chi refers to the Blake-
Figure 14 Two-dimensional meniscus shape variation simulation along the length of a vapor
chamber ............................................................................................................................................................................................... 22
Figure 17 Wick/Evaporation resistance elements for a sintered spherical particle wicking structure . 28
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
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 31. Measurement of contact angle between wire mesh and water........................................................ 50
Figure 32. Wire mesh bubble point validation test results (IPA) ................................................................................ 50
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 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 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 52. Heat Transfer and Mass Transport Test Setup vacuum chamber .................................................... 66
Figure 53. Heat Transfer and Mass Transport Setup connection diagram ......................................................... 66
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 59 Visualization of wick saturation in copper sintered wick using fluorescent dye over time .... 74
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 68 Saturated wick at 8.9g, excess liquid pools on top (G-force in direction).................................. 80
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 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 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
............................................................................................................................................................................................................... 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
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
xiv
List of Tables
Table 1 DARPA Thermal Ground Plane objectives and phases [5] .............................................................................4
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 =
Table 8 Concept, detailed model comparison (keff=400W/m-K, coolant 20C, ambient 20C) ...................... 89
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 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]
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
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
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
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
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].
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
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
The final goal of the program is to demonstrate performance impact on a relevant system using
Collaborators in this research include GE Global Research(GE), University of Cincinnati(UC), Air Force
1.3. Scope
The scope of this study includes high-level modeling of heat pipes and identification of parameters
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
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.
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
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
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
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
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
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
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.
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
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
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
̇ 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
12
SteamIAPWS
500
400
300
T [°C]
200 1000000 Pa
hfg
100 100000 Pa
50000 Pa
25000 Pa
0
0.0x100 1.0x106 2.0x106 3.0x106 4.0x106
h [J/kg]
The Jakob number, which describes the ratio of sensible versus latent heat, is an important
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.
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
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
⃗⃗⃗⃗⃗ Eq. 4
This assumes:
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
Evaporation, condensation and entrainment effects in the adiabatic region are neglected
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
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
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.
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
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
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
The hydraulic diameter can be defined using the cross section area and the wetted perimeter( ).
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!
̇
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
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
̇
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
parameters such as porosity( ) and sintered particle radius( ) and permeability for sintered wick
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).
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.
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
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
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.
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
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
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
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
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
( ) ( ) ( ) ̇ Eq. 16
Assuming 1-dimensional heat transfer, isotropic material properties, no heat generation and steady
( ) Eq. 17
Eq. 18
Which can be expressed in a thermal resistance ( ) by multiplication with the mean cross section
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
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
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
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
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
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
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
Thin Film
Region
Thin Film
Region
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
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
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
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.
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
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
( ) Eq. 24
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.
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
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
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
Eq. 28
Eq. 28 defines the boundary of feasible design space. This boundary is illustrated in Figure 21 for a
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
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.
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
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
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
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
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
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
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
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
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
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
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
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:
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
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
experimental results.
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
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
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]
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-
The apparatus chosen for performing the bubble point testing is a PMI BPT-101-A bubble point tester
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).
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
1 mm Calibration
sample 1
325 mesh
13 wires/mm
rpore~39 m
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
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
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
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
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
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
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
previously mentioned Chi observed a ratio of pore diameter to packed sphere wick particle diameter
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
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
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
The formulation of the permeability constant for one-dimensional flow can be derived from Eq. 11
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 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.
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
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
̇ 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
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
58
Figure 41. User interface of discretized evaluation of wick permeability model
̇ [ ] 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.
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
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.
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
The intrinsic effective thermal conductivity of a porous sintered copper structure was evaluated
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).
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
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
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,
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
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.
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
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
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
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 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
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
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
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
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
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
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
69
R~4.5K/W
R~4K/W R~0.5K/W
R~0.25K/W
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
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
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.
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.
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
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
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
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.
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.
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
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.
Vapor Chamber
Wick
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
TGP
A A-A
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
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).
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
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).
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.
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
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
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.
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.
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
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
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.
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
TIM
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.
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
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
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
operate the experiment in the low sensitivity region as indicated in Figure 71.
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.
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
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
Chiller
TGP test fixture Rotation fixture
Pump
Temperature safety
controller
DAQ
Power
Supply
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
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
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
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
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%
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.
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
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.
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.
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%
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]
A screening analysis is performed where only the uncertainty contribution of the differential
(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
Temperature measurements in particular are sensitive to uncertainty due to the relative inaccuracy
evaluated as a propagation in uncertainty of the thermocouple combined with the data acquisition
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
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
Eq. 53
91
Note that the Seebeck coefficient is temperature dependent and can be evaluated at the mean
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.
5.0E-03 V c T
i 0
i
i
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]
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
92
[ ] Eq. 54
The function for the Seebeck coefficient is presented in Figure 85. The Seebeck coefficient curve was
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]
The reference junction in the data acquisition system (DAQ) plays an important role in evaluation of
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
errors.
Eq. 55
Eq. 56
√( ) Eq. 57
94
This allows expressing the error in measurement of the temperature at junction 1 as a function of
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
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
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
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
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
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
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
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 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
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.
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
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
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
5
Leak [%]
0
0 10 20 30 40 50 60
T=T_coolant-T_ambient
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.
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
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
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
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
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
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
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
In order to validate the performance of the TGP Thermal Characterization experiment. Tests were
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
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
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)
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
The key focus of the testing team in Phase II was testing the TGP thermal performance at high-g with
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.
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
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
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
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
400
Thermal conductivity [W/m-K]
375
350
325
300
0 2 4 6 8 10
Acceleration [g]
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
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]
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
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]
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
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
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
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
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
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
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.
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
saturated wicking structure at different heat loads. In addition, a novel visual validation experiment
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
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
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
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
6. Perkins, L.P., and Buck, W.E., 1892, “Improvements in Devices for the Diffusion or Transference
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
9. Gerner, F. M. and Henderson, H. T., 1995, “Liquid metal micro heat pipes for space radiator
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
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
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),
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
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
ground plane Proc. of ASME/JSME 8th Thermal Engineering Joint Conf. AJTEC 2011 (Honolulu,
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
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
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,
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
119
32. Adkins, D. R., and Dykhuizen, R. C., “Procedures for Measuring the Properties of Heat-Pipe
33. Feser, J.P., Prasad, A.K., Advani, S.G., “Experimental characterization of in-plane permeability
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
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
37. H. Peter J. de Bock, Shakti Chauhan, Pramod Chamarthy, et al., 2010, “On the Charging and
38. Marotta, Egidio E., Steven J. Mazzuca, and Julian Norley. "Thermal joint conductance for
flexible graphite materials: analytical and experimental study." Components and Packaging
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
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
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
q_flux_left = 0 [W/m2]
q_flux_right = 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
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
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
"
+---------+
^ 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
! 1 htr
! 3 ins
! 4 TGP
! 5 coldplate
! 6 shell
! 9 acr
!11 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
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
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
140