Chapter 1 Introduction 1.1 Thermal Management in IC Packaging Heat dissipation in IC packages is becoming very vital with chips getting smaller and running at ever increasing speeds. Intel® microprocessors advanced from having 1.5micron lines on INTEL386TM [1] to 0.13-micron in the current Pentium® 4 [2] within a space of 7 years. The International Technology Roadmap for the Semiconductor (ITRS) is continually studying the trend in circuitry lines, which is also called the technology node, reduction. The figure below is an adaptation from the ITRS reports [3-5]. The thinner these lines go to, the denser the circuitry will be; leading to Technology N ode ( µ m ) extensive increases in heat generation in microprocessor packages. 0.20 0.15 0.10 0.05 0.00 1995 2000 2005 2010 2015 Year Figure 1.1: Trend in technology node over the years The increase in heat flow from IC chips is also apparent in the trend of power dissipation for high-performance chips reported by ITRS as shown in the following figure. Increasing amount of power dissipated has to be offset by the efficient cooling because the maximum junction temperature stays relatively unchanged at 85 - 90oC. 1 Power M ax Junction Temp 95 Power (W) 250 90 200 85 150 100 2000 80 2005 2010 Year 2015 Max Junction Temp ( oC) 300 2020 Figure 1.2: Power dissipation and maximum junction temperature in IC chips A major challenge in the semiconductor field is, therefore, the ability to manage the heat in the IC chips without compromising on the performance of the device. This management of thermal energy is very crucial because heat has many detrimental effects on the device. It has been reported [6] that failure rates have near exponential dependence on device temperature. When a certain upper critical temperature is reached, important parts of the device may cease to function. Furthermore, temperature cycles that result from switching on and off the device can also cause problems, even if the operating temperature does not hit the upper critical temperature. This leads to another problem, the reliability of the device. It is important to dissipate the heat gradually so that the temperature of the device will not reach its ceiling maximum and minimum values. This will reduce the cycling temperature (the difference between the maximum and the minimum). The figure below shows how a gradual cycling reduces the maximum cycling temperature. 2 1 Temperature 8 6 4 2 Time 0 0 10 20 30 40 50 60 70 2 4 6 8 Normal cycling Gradual cycling 1 Figure 1.3: Reduced maximum temperature due to gradual cycling 1.2 Thermal Interface Material IC chips are packaged to provide protection for the intricate chips. There are many layers in a packaged chip with every layer performing crucial roles in ensuring proper functioning. Below is a schematic, not drawn to scale, of a typical packaged IC chip. Heat Flow Heat Spreader Thermal Interface Material Solder Balls Underfill Die Substrate Figure 1.4: Schematic diagram of a packaged chip 3 When the chip is powered up, the die gets hot and heat must effectively be transferred out to prevent overheating. Heat spreaders are incorporated to aid the spreading of heat from the small die to the big heat sink and eventually to the ambient. Improvements on substrates and heat sinks, which are the junction of the actual IC and the ambient, can constructively affect thermal dissipation. It is of equal great advantage that interfacial heat path to the heat sink be improved. The layers in the package create more interfaces for the heat to pass and these solid interfaces become the bottleneck of the heat transfer. Solid surfaces, not being atomically smooth and straight, may be warped and rough. The following figure shows how 3 different surfaces and mated with an atomically smooth surface. (a) Perfectly mated surface (b) Warped surface (c) Rough Surface Figure 1.5: Schematic diagram of surfaces As seen, the warped and rough surface that are common in the surfaces used in the IC packages, create pockets of air trapped in the interface. These poorly mated surfaces cause a rise in contact resistance. As a result of the impedance, there will be a temperature drop across the interface. This phenomenon is discussed in Chapter 2. Air gaps Thermal interface material Figure 1.6: Utilizing thermal interface material to fill air voids 4 In today’s IC packages, thermal interface materials (TIM) are utilized to fill up the micro air voids, as seen in the above figure. The conductivity of the thermal interface is not the only deliberation because the ability of these materials to conform to the micro roughness at the interfaces is just as essential. Filling up of the micro voids with thermal interface material will maximize the heat flow path and hence minimizing thermal contact resistance. The use of a solid, for instance, copper, is therefore not a feasible choice for a TIM. Commonly used TIM in the semiconductor industry include thermal grease, polymeric adhesives and phase change materials (PCMs) [7]. PCM have, in majority cases, replaced grease. [8] This is because they are cleaner to use and provide much greater resistance to the liquid state being pumped out of the interface, unlike thermal greases that are prone to pump out [9]. 1.3 Overview of Research The research is a collaboration between the Department of Chemistry, National University of Singapore and Honeywell Electronic Materials, Honeywell (S) Pte. Ltd. This research focuses on its thermal management using a commercial Thermal Interface Material (TIM), which is a Phase Change Material (PCM), produced by Honeywell. The diagrams in the following figure illustrate the focus of the research. 5 PCM Characterization Thermal Resistance (TR) Measurements Surface Characterization (a) Areas of interest within a packaged chip Chemistry and Interfacial Mechanics of a Phase Change Material on Metal Surfaces PCM Characterization Thermal Resistance Measurement Setup Surface Characterization Interaction of PCM on Metal Surfaces (b) Classification of focus areas Figure 1.7: Illustration of research scope The initial stage of the work is focused in three areas: PCM characterization, design and construction of thermal resistance measurement setup and surface characterization of metals. The PCM and metal surfaces were characterized to understand their basic properties. The thermal resistance measurement setup can then be utilized to understand how the PCM material on metal surfaces perform with varying application methods like temperature and pressure, process conditions, assembly conditions and reliability tests. These studies would enable better understanding of the performance of the PCM as TIM on metal surfaces. 6 Chapter 2 Theory 2.1 Thermal Test Methods and Standards There are various ways by which a packaged chip is tested in the area of thermal management. The following is an account of thermal conduction followed by a short review of the thermal test methods that are employed. Conduction is the exchange of internal energy from one body to another, or from one part of a body to another part. The exchange is the transfer of kinetic energy of the molecules by direct collision or the drift of free electrons, without an appreciable displacement of the matter comprising the body. Convection, on the other hand, is the heat transfer in a fluid by the mixing of one portion of the fluid with another portion due to gross movements. Radiation is the heat transfer via electromagnetic radiation emmited by a thermally excited body. A medium is not always necesary for heat transfer by radiation. The primary mode of transfer of interest in this project is conduction. The relationship governing conduction is Fourier’s law [2]. The following is a demonstration of the derivation of the law, relied on works by Chapman [1], Bar Cohen and Kraus [2]. Considering a thin plate of material (Figure 2.1) with cross-sectional area, A and thickness, dx. Let each of the two sides be uniformly maintained at temperature T1 7 and T2 over the surface.This law indicates that the rate of heat flow by conduction through a material, q is proportional to the cross-sectional area of the material, A , and the temperature gradient dT/dx. dx T1 T2 A• q Figure 2.1: Thin plate of material with uniform temperatures over two opposite surfaces The relationship is as follows: q ∝ -A dT dx (eqn. 2.1) where the negative sign shows that heat flows from a region of higher temperature to one with a lower temperature. The proportionality constant is the thermal conductivity, k. The value k is an intrinsic property of the material. The proportionality hence becomes q = -kA dT dx (eqn. 2.2) Assuming the plate of material is within a homogeneous, isotropic solid as shown in the figure below. 8 S δx x P δq δA Figure 2.2: Schematic of a thin plate of material within a homogeneous and isotropic solid Selecting a point P on surface S, within the solid, we have an area δA, which is part of the surface S containing P, and having a thickness δx in the direction normal to the surface at P. If the difference between the temperature of the back face of the plate and its front face is δT, and if δA is chosen small enough so that δT is essentially uniform over it, the rate of heat flow across the plate, δq is δq = -k δA δT δx (eqn. 2.3) where the negative sign shows that heat flow is taken to be positive if δT is negative in the direction of increasing x, the normal displacement. The heat flow per unit area known as the heat flux, f, can be calculated using δq/δA. When δA → 0, the heat flux through the thickness δx, becomes 9 f= dq δT = -k dA δx (eqn. 2.4) Allowing δx → 0, heat flux across S at point P in terms of temperature gradient across P in the x direction becomes f = -k dT dx (eqn. 2.5) The above equation is Fourier’s conduction law after the French mathematician who first made the an extensive analysis of heat conduction. It states that the flux of heat conducted (energy per unit time per unit area) across a surface is proportional to the temperature gradient taken in a direction normal to the surface at the point under consideration. The total rate of heat transferred across the finite surface S would be q = − ∫k s dT dA dx (eqn. 2.6) In general, the normal gradient dT/dx may vary over the surface. Nevertheless, there are many instances where it is possible to select the surface as one on which the gradient is the same everywhere. This is the situation in the case depicted in the plate of material. (Figure 2.3) Assuming that there are no heat source or sink in within a fixed area of material, 10 q = -kA dT dx (eqn. 2.7) With the Fourier’s law relation, it is possible to derive a general equation of heat conduction by considering the differential volume shown below. y qy+dy qz qs stored within volume qx qx+dx qz+dz qG generated within volume qy x z Figure 2.3: Differential for the derivation of the general heat conduction equation As energy cannot be created or destroyed but can be transformed from one form to another, adding the amount of heat going into the subvolume (qin) and the amount of heat generated within it (qG) equates to total amount of heat coming out of the subvolume (qout) and heat stored within it (qs). qin + qG = qout + qs or qin - qout + qG = qs (eqn. 2.8) 11 Considering the x direction, heat entering the subvolume, qx is qx = -k A δT δT ) dy dz = (-k δx δx (eqn. 2.9) where the temperature gradient is expressed as a partial derivative because the temperature, T, is a function of y, z and the time, t. The heat, leaving the right surface at x+dx is ⎡⎛ δT ⎞ ⎤ δT ⎞ δ ⎛ qx+dx = ⎢⎜ − k ⎟dx ddy dz ⎟ + ⎜− k δx ⎠ ⎥⎦ δx ⎠ δx ⎝ ⎣⎝ (eqn. 2.10) The difference between the amount of heat entering and leaving the subvolume in the x direction is, therefore, qx-qx+dx = δ ⎛ δT ⎞ ⎜k ⎟ddx dy dz δx ⎝ δx ⎠ (eqn. 2.11) qy-qy+dy = δ ⎛ δT ⎞ ⎜k ⎟dx dy dz δy ⎜⎝ δy ⎟⎠ (eqn. 2.12) qz-qz+dz = δ ⎛ δT ⎞ ⎜k ⎟ dx dy dz δz ⎝ δz ⎠ (eqn. 2.13) Similarly Thus, qin-qout = δ ⎛ δT ⎞ δ ⎛ δT ⎞ δ ⎛ δT ⎞ ⎟⎟ + ⎜⎜ k ⎜k ⎟ ⎜k ⎟d+ δx ⎝ δx ⎠ δy ⎝ δy ⎠ δz ⎝ δz ⎠ dx dy dz (eqn. 2.14) 12 The heat transmitted out of the subvolume is related to the heat stored within the subvolume can come from an electrical or electronic heat dissipation or from a chemical reaction. With the volumetric rate of heat generation, qi qG = qi (dx dy dz) (eqn. 2.15) Finally, the amount of heat stored within the subvolume is associated with the rate of change in internal energy, du dT = c dm dt dt (eqn. 2.16) where u is the internal energy, c is the specific heat capacity, dm = ρ dx dy dz, where ρ is the density of the material, du δT =ρc dx dy dz dt δt (eqn. 2.17) where a partial derivative is now employed (due to the infinitesimally small material area) When equations 2.14, 2.15 and 2.17 are inserted into equation 2.8, the following results. δ ⎛ δT ⎞ δ ⎛ δT ⎞ δ ⎛ δT ⎞ δT ⎟⎟ + ⎜⎜ k ⎜k ⎟ + qi = ρ c ⎜k ⎟d+ δx ⎝ δx ⎠ δy ⎝ δy ⎠ δz ⎝ δz ⎠ δt (eqn. 2.18) where the common dx dy dz terms have been cancelled. 13 Assuming that k, c and ρ are independent of temperature, direction, and time, a general equation of heat conduction can be obtained: qg 1 δT δ2T δ2T δ2T + + + = 2 2 2 k α δt δx δy δz (eqn. 2.19) Where α is the thermal diffusivity of the material: α ≡ k / (cρ) At steady state, qg δ2T δ2T δ2T + + + =0 k δx 2 δy 2 δz 2 With an absence of heat sources, δ2T δ2T δ2T + + =0 δx 2 δy 2 δz 2 (eqn. 2.20) (eqn. 2.21) The general equation of heat conduction requires the values of the thermal conductivity, specific heat and thermal diffusivity of the material to be known. Heat conduction path through changes in physical dimensions involves spreading resistance and that across interfaces involves contact resistance. Thermal resistance is a more rigorous physical property, compared to thermal conductivity. It indicates the ease at which heat can be conducted with the consideration of physical size of the material. These physical properties will be explained in some details in the subsections that follow. 2.2 Physical Properties of Materials In most, if not all heat conduction mathematical analyses, at least one of the following physical properties is essential. • Thermal conductivity 14 • Specific heat capacity • Thermal diffusivity • Spreading resistance • Thermal resistance • Contact resistance 2.2.1 Thermal Conductivity Thermal conductivity is often regarded as the primary physical property of the material where heat conduction is concerned. Thermal conductivity of a material indicates how well energy can be exchanged through molecular motion caused by a temperature gradient. It is an intrinsic property of the material [10] and is dependent on the chemical composition, the phase, the structure and the temperature and pressure the material is subjected to. Thermal conductivity was introduced in Fourier’s Law [2] q = -kA. ∆T L (eqn. 2.22) where q = amount of heat transferred k = thermal conductivity of the material A = cross-section area of material ∆T = temperature gradient per unit length of the material L A higher thermal conductivity shows a greater ability of the material to conduct heat. A commonly used unit of thermal conductivity is W/mK. 15 2.2.2 Specific Heat Capacity Specific heat capacity of a material, on the other hand, is the change in temperature of the material with the amount of energy stored in it. Specific heat capacity is used in the general equation of heat conduction (eqn 1.19). From the first law of thermodynamics [10], the changes in enthalpy at constant pressure and changes in internal energy at constant volume in reversible processes represent heat transferred to and from a system. Taking the amount of heat transferred per unit temperature difference during a reversible, constant pressure process as cp, specific heat capacity at constant pressure is ⎛ ∂h ⎞ cp = ⎜ ⎟ ⎝ ∂T ⎠ p (eqn. 2.23) Likewise, cv represent the specific heat capacity during a reversible constant volume process: ⎛ ∂u ⎞ cv = ⎜ ⎟ ⎝ ∂T ⎠ v (eqn. 2.24) where h = specific enthalpy u = specific internal energy T = temperature subscripts p and v = differentiation done at constant pressure and volume, respectively 16 2.2.3 Thermal Diffusivity Thermal diffusivity was introduced in the derivation of the general heat of conduction (eqn 2.19) α= k ρc p (eqn. 2.25) where α = thermal diffusivity k = thermal conductivity cp = specific heat capacity at constant pressure ρ = density Specific heat capacity at constant pressure is normally used as solids, which are a common medium of conduction, are usually not compressible. The pressure in the material is therefore usually constant. Thermal diffusivity incorporates inherent properties of a material; the thermal conductivity, k and the specific heat capacity, c. This integration of intrinsic properties facilitates many calculations because tabulated values of common materials at specific conditions are widely available. 2.2.4 Spreading Resistance Not all thermal conduction takes place across media of the same cross-sectional areas. More often than not, conduction takes place through a solid or across an interface with different cross-sectional area. In an electronic package, for example, 17 heat is conducted through many layers of different cross-sectional areas. The figure below illustrates the example. Heat Flow Heat Spreader Thermal Interface Material Solder Balls Underfill Die Substrate Figure 2.4: Schematic diagram of heat flow in a packaged chip In this type of cases, the spreading resistance becomes a concern. This spreading phenomenon has been discussed [6, 11-13]. Mathematical analyses have been done to predict the spreading resistance. The work of Song, Lee and Au [12] displays equations that have verification with experimental examples. 23 equations were involved in the predictions of the spreading resistance. These equations will not be further elaborated, as it is not in the scope of this study. 2.2.5 Thermal Resistance and Thermal Impedance Thermal resistance and contact resistance are, like thermal conductivity, measures of how well heat is conducted. Thermal conductivity, of homogeneous materials, is independent of physical dimensions while thermal resistance and contact resistance are very much dependent on the physical dimensions of the material. Thermal resistance is [11] the temperature gradient caused by a unit of heat flow through a material of a given size. From Fourier’s Law (equation 2.2) 18 θ= ∆T L = .A q k (eqn. 2.26) where θ = thermal resistance ∆T= temperature gradient q = amount of heat transferred A = cross-section area of material k = thermal conductivity of the material L = length of the material From equation 2.26, it can be seen that thermal resistance of a homogeneous material is proportional to the distance of heat travelled. Thermal impedance, on the other hand, is the product of the thermal resistance and the cross-sectional area. Thermal impedance = θ.A = L ∆T .A = q k where θ = thermal resistance ∆T= temperature gradient q = amount of heat transferred A = cross-section area of material k = thermal conductivity of the material L = length of the material 19 2.2.6 Contact Resistance When heat conduction involves heat travel through 2 solids in contact, an additional temperature gradient exists together with the normal temperature gradient. A significant temperature drop will be observed due to the interface between the two solids. The phenomena that causes this is contact resistance. [11] No matter how well a surface is polished, surface irregularities still exist. Intimate solid contact will therefore inevitably be accompanied with pockets of air. These pockets of air are very much non-conductive relative to the solid in contact and the heat travel through the air, consequently, becomes the bottleneck of the conduction path. This gives rise to the extra temperature drop. hot cool Temperature interface x x x x x x Distance Figure 2.5: Temperature profile of two solids in contact Contact resistance can be accounted for to a certain extent. Mathematical treatments are discussed in works of Kraus & Bar-Cohen [6]. The accountability of contact resistance is dependent on many parameters. These parameters are complex, given the numerous uncertainties present in real surfaces. The parameters of concern include: [6] 20 ƒ Number of intimate contacts ƒ Shape of contact points: circular, elliptic, band, or rectangular ƒ Size and arrangements of contact points ƒ Geometry of contacting surfaces with regard to roughness and waviness ƒ Average thickness and fluid (gas, liquid or vacuum) of void space ƒ Pressure and conductivity of void space ƒ Hardness of contacting surfaces ƒ Average temperature of interface ƒ Contact pressure and contact history of surface ƒ Duration of contact with regard to relaxation effects ƒ Vibrational and directional effects ƒ Contact cleanliness Filling the air gaps with conductive and compliant materials, like thermal grease, and using a very high contact pressure normally reduces contact resistance. Advances in materials technology have seen some other materials fill the air gaps and one such material is the phase change material under study in the present work. Thermal test methods for IC chips such as microprocessors, therefore, may involve any of the above mentioned physical properties. It is of special interest in this project that the thermal resistance be measured. 21 2.3 Methods of Thermal Resistance Measurements In the electronics industry, a range of thermal resistance is measured. The measurements of thermal resistance can be carried on individual components of the chip, like the substrate, die or interface material; or the entire chip itself. Typically, the thermal performance of a packaged chip is measured by the ability of the package to dissipate the heat that is produced within it to the ambient. A number of methods have been used [15] but most of these methods are concerned with the determination of thermal resistance of the entire chip. The types of methods are: • Optical • Chemical • Physical • Electrical The optical method involves IR scanning of the chip surface. This method can only be used on un-encapsulated chips as it presents temperature profile on the surface. The optical method is time consuming and costly. Chemical method of measuring thermal resistance, on the other hand, requires the chip surface to be coated with a thin layer of temperature – indicating chemical like liquid crystal material. Similar to the optical method, the chemical method provides the temperature profile of the surface and therefore cannot be used on encapsulated chips. There are also a number of disadvantages in using the chemical method. Significant expertise is required in selecting and applying the chemical. Although 22 chemical application is more economical compared to the optical method, it may pose resolution and contamination problem; and demands a lot of time. The third method, which is the physical method, makes use of direct attachment of very small temperature-measuring devices (like thermocouples) on the chip surface. This technique is relatively inexpensive but placing a temperature-measuring device on the surface, without affecting the heat source, is a challenge. Resolution will, likely, be compromised. Like the two previous methods, the physical method needs a lot of time and cannot be used on encapsulated chips. The fourth and last approach to thermal measurement is the electrical method. This measurement mode uses pre-calibrated temperature sensitive parameter of the chip. The procedure is the fastest, in comparison with the other 3 methods. The level of proficiency required to obtain accurate and repeatable results is minimal. Encapsulated chips can have thermal resistance measured using this method. This method, however, has even lower resolution than chemical and optical methods, and the set-up is moderately expensive. This method is widely used and has been discussed in several papers. [16-21] A major part of this project, however, focuses on the thermal resistance of a commercial thermal interface material (TIM). As the name implies, TIM is a material sandwiched between a chip and a heat spreader and/or a heat sink. Details of thermal interface materials are discussed in Section 1.2. At first glance, none of the 4 mentioned methods could be used to measure the thermal resistance of TIM. Measurements methodologies have to be simulated that of an actual package and 23 therefore the TIM has to be sandwiched. The optical, chemical and physical methods require the surface of the material of interest to be exposed. Electrical method, on the other hand, requires the TIM to be circuited into a device. The thermal resistance of the TIM is therefore, difficult to obtain using those methods. Methods, using standards (MIL-I-49456A, JESD15, ASTM D5470-95 and ASTM E1530-99) as guidelines, have been created. The ASTM methods have very stringent requirements to suppress the contact resistance contributions to the measured values. For example, the contact requirement for ASTM D5470 is a force of 3.0 ± 0.1 MPa. These values are very high compared to the normal pressure used in semiconductor packaging, which is less than 0.3 MPa. The surface smoothness requirement for the ASTM method is also very stringent. The contact surfaces have to be smoothly finished to within 0.4 µm. A reasonably accepted smoothness by the industry for the surfaces for the measurements is below 25.4µm. It is crucial that the development for a thermal resistance measurement set-up be continued. Some companies have built their own set-up with adjustments deemed necessary, to the standards. No commercial set-up is available in the market and a basic set-up must, therefore, first be built and developed for this study. The thermal resistance measurements carried out in some laboratories [22-27] concentrated on the bulk properties of the materials. Physical properties like thermal conductivity and thermal resistance of the material can be obtained. The experimental methodologies involve sandwiching the material of interest with metal blocks. Temperature drops across the metal blocks are then measured. Due to the 24 sandwiching of the TIM, contact resistance becomes a crucial issue. It can be anticipated to have influence of sizeable proportions to the measured thermal resistance. The ASTM D5470 is the most widely used. Thermal resistance values are important in the process of selecting the right material of the right thickness for a particular application. The work by Rauch [27] exhibits how thermal resistance measurements aid in material selection. One other aspect, which is the effects of different surfaces on the thermal performance, however, has not been touched upon. It is the main aim of this study to investigate the effects of different metal surfaces on the thermal resistance of the TIM. The following chapters will look into the building and development of a set –up for this study. Rigorous validations have been planned to verify the working ability and integrity of the built system. 25 Chapter 3 Instrumentation 3.1 ASTM D5470 The ASTM D5470 exemplifies a method to measure thermal properties thin thermally conductive solid electrical insulation materials at steady state conditions. The conditions that are stated for the analysis, however, have to be modified to suit the needs of the semiconductor industry. The schematic of the apparatus is shown in Figure 4.1 below. Force H Insulator H H Guard Heater Insulator H H T1 T2 T3 Heater Upper metal bar Sample Lower metal bar T4 T5 Reference Calorimeter T6 Cooling Unit Insulator Figure 3.1: Schematic of thermal measurement apparatus according to ASTM D5470 26 General features of the standard include: 1. The sample is placed between two metals bars with high thermal conductivity and smooth finish within 0.4 µm to the approximate true plane. 2. A cooling unit (comprised of one of the metal bar) with constant temperature bath maintained uniformly at ± 0.2°C. 3. A force of 3.0 ± 0.1 MPa pressing the stack to minimize the contribution of contact resistance in the measurements. 4. Upon attaining equilibrium, the temperatures at the bars are taken and the thermal resistance can be calculated using the following equations: θ = (TA-TD) / q (eqn 3.1) where: θ = thermal resistance TA-TD = difference in temperature of the metal bar in contact with the sample q = heat flow For a setup with reference calorimeter: q= λxA d x[T5 − T6 ] (eqn 3.2) where: A = Area of reference calorimeter λ = thermal conductivity of the reference calorimeter T5-T6 = temperature difference between thermocouples of reference calorimeter 27 For setup with no reference calorimeter: q=IxV (eqn 3.3) where: I = electrical potential applied to the heater V = electrical current flow in the heater The temperature on the surface contacting the sample, TA and TD can be derived from the following: TA = T2 − dB (T1 − T2 ) dA TD = T3 − dD (T3 − T4 ) dC (eqn 3.4) (eqn 3.5) where: TA = temperature of the upper metal bar surface in contact with the sample TD = temperature of the lower metal bar surface in contact with the sample dA = distance between temperature sensors in the upper metal bar dB = distance from the lower sensor to the lower surface of the upper metal bar dC = distance between temperature sensors in the lower metal bar dD = distance from the upper sensor to the upper surface of the lower metal bar The features of the ASTM are normally modified [22-23, 28-33] to suit the provisions of the semiconductor industry. The requirements for the smoothness of the metal bars and the applied force are too stringent and are not practical for the applications in the industry. These requirements are added in the ASTM to suppress the contribution of contact resistance in the measurements. The smoothness criterion 28 is very difficult to achieve given the heavy usage of the apparatus and the applied force is too great for the industry to comply with. Furthermore, the measurements at each high pressures would not be meaningful in the industry. The normal applied force is 0.2-0.3 psi but the required pressure in the standard is 3.0 ± 0.1 MPa. A setup has been designed and built based on these issues and the industry specific requirements. 3.2 Instrument Design and Calculations 3.2.1 Physical Design The set up consists primarily of a test stand, 2 intermediate blocks, a temperature and a pressure control system as well as a data acquisition system. A schematic diagram, not drawn to scale, and an actual photograph of the setup is as shown in Figure 3.2a and 3.2b respectively. The test stand is made of stainless steel. There is an opening in the test stand to insert the intermediate blocks. The size of the opening is determined by the extent of screw length after the different blocks and sample, are assembled. The temperature control system is made up of an aluminium heater jacket to house a 200W-cartridge heater and a water cooler jacket that is connected to an 8005 Polyscience water circulator. A Teflon jacket to prevent excessive heat loss insulates the heater jacket. Type-T thermocouples with 0.6mm diameter and accuracy of +0.1oC are used to monitor the temperature of the heater. A 2132 Eurotherm Controls temperature controller controls the cartridge heater with reference to a thermocouple. The pressure applied on the stack is monitored using a Honeywell universal 29 controller, UDC 1000 via a Data Instruments SC500 load cell. The application of pressure can be done via the manual screw press or automatically using the actuator. load cell insulator heater Aluminium intermediate block sample Temperature controller T-type t/c Terminal block Aluminium intermediate block water circulator actuator Data acquisition card LabVIEW Universal controller Water bath valve Pressure controller Figure 3.2a: Schematic of the thermal resistance measurement setup Figure 3.2b: Photograph of the thermal resistance measurement setup 30 Three temperature measurements are taken at each Aluminium block instead of two to maximize the accuracy. The calculation of the thermal resistance is, as a result, slightly modified. The figure below is an example of the temperature profile measured. A 60.5 60.0 58.5 58.0 57.5 57.0 y = -18.271x + 55.677 -0.25 -0.20 -0.15 -0.10 Distance from Sample (cm) 75 56.5 70 56.0 -0.30 80 Temperature (oC) 59.0 Temperature (oC) 59.5 55.5 -0.05 0.00 65 B 45.5 45.0 44.5 44.0 43.5 43.0 42.5 42.0 41.5 41.0 40.5 y = -16.836x + 44.983 0 60 0.05 0.1 0.15 0.2 0.25 0.3 Distance from the sample (cm) 55 50 45 bottom block top block 40 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 Distance from sample (cm) Figure 3.3: Graph of temperature drop across the intermediate blocks As soon as the cartridge heater is switched on, a temperature junction within the intermediate blocks occur. The sample of interest (or often referred to as the interface material) gets heated and cooled by the aluminium blocks above and below it respectively. The thermal resistance value can then be calculated using the following equation: θ = ∆T / q (eqn 3.6) where θ = thermal resistance of sample ∆T = temperature drop across the sample (difference in the two intercepts A & B) q = amount of heat transferred to the sample 31 The amount of heat, q, transferred to the interface is calculated from the equation q = kAl AAl ∆Tblk ∆x (eqn 3.7) where kAl = thermal conductivity of aluminium AAl = cross-sectional area of the aluminium block ∆Tblk = temperature gradient across the top aluminium block. ∆x 3.2.2 LabVIEW The setup is capable of measuring both the transient as well as the steady state conditions. These capabilities have been made possible by specifying the rate of data acquisition parameter of the LabVIEW program. The LabVIEW program is a virtual instrument and is made up of two components. The first one is the front panel that serves as the user interface. The other component is the block diagram that contains the graphical source code that defines the tasking of the virtual instrument. The virtual instrument that has been programmed is shown in Figure 3.4. The virtual instrument has the following main features: 32 1. The temperature readings from the six thermocouples are recorded and displayed in a waveform with digital indicators. The ASTM D5470 states that equilibrium is reached when the difference between two successive sets of temperature readings, taken at 15 minutes intervals, are less than ±0.2oC. Indicators of the equilibrium are placed in the program for an online monitor of the equilibrium status. 2. The calculations for the thermal resistance measurement and the amount of heat supplied to the sample are done to good approximate values as the analyses proceed. 3. Every temperature data point from the analyses can be saved into a spreadsheet with specified frequency. 4. The temperature data were then processed using the LINEST worksheet function to calculate the amount of heat and the thermal resistance. (a) Front panel of the virtual instrument 33 34 3.3 Calibration Methodology Calibration experiments are carried out to verify how reproducible, robust and rugged the setup is. In place of a PCM, brass and Teflon calibration blocks of 25.4mm (1 inch) in diameter of various lengths were made. The calibration experiments are classified in the following areas. 1) Effect of interface material 2) Effect of ambient temperature 3) Effect of water bath temperature 4) Effect of thermal grease at aluminium / sample interfaces 5) Effect of air movement 6) Effect of operator 3.3.1 Effect of interface materials Two types of materials with very different values of thermal conductivities were studied. These two materials were made into blocks with different lengths but the same diameter size as the aluminium blocks. As much as possible, the smoothness of these blocks was kept below 25µm. Apart from the absolute value of the thermal resistance of the material at the particular thickness, this set of experiments enables the values of contact resistance and effective thermal conductivity of the material to be obtained. The thermal conductivity of the material can be determined from at least 2 measurements of the thermal impedance, at the same temperature and contact 35 pressure. Thermal impedance is the product of thermal resistance and the cross section area. From equation 3.8, the reciprocal of the thermal resistance vs. thickness plot will yield the thermal conductivity of the material of interest. θ= ∆T L = .A q k ∴ θ .A = L/k (eqn. 3.8) where θ = thermal resistance ∆T= temperature gradient q = amount of heat transferred A = cross-section area of material k = thermal conductivity of the material L = length of the material The typical unit used for thermal resistance measurements is oC/W. 3.3.2 Effect of ambient temperature The ambient temperature may affect the equilibration of the experiment. For this set of experiments, the air conditioner was set to the lowest temperature (18oC), to a temperature it is normally operating (21oC), the highest temperature (29oC). 3.3.3 Effect of water bath temperature Similar to changing the cartridge heater temperature, changing the water bath temperature determines the amount of heat flow through the intermediate blocks. The difference is that while the heater temperature is increased, the water bath will get 36 heated up as well until equilibrium is achieved. When the water bath is heated, on the other hand, the heater temperature is fixed at 65oC. This temperature was chosen because there will be an appreciable temperature drop across the sample and the blocks will not be too hot to be handled. 3.3.4 Effect of thermal grease at aluminium / sample interfaces Irregular surfaces give rise to contact resistance. This set experiment is similar to the set of experiments involving the study of the effect of interface materials but only brass blocks were used. Thermal grease was applied at the brass/aluminium interface to minimize contact resistance and hence increase the amount of heat supplied to the interface material, the brass block. 3.3.5 Effect of air movement This study analyses the sensitivity of the thermal measurements to air movement around the set up. A fan was placed at 0.5 meters and 2 meters away from the set up. Various fan speeds as well as swinging motions were used to vary the air movements. 3.3.6 Effect of operator The precision of this setup is tested in this study. 3 different operators independently performed measurements of the same sample for 4 times. The nature of the sample and other details of this study are proprietary to the industrial partner. The results are 37 then tabulated and statistically analysed using Gage Repeatability and Reproducibility (GR&R). 3.4 Calibration Results and Discussions 3.4.1 Effect of Interface materials The thermal resistance of the 2 interface materials tested has 1 – 2 orders of magnitude difference. Teflon, a good insulator has much higher thermal resistance than brass, which is a good conductor of heat. The figures below show the results to the experiments. 4 .5 9 0 .0 y w y x 4 .0 = 6 0 .1 3 x + 0 .7 7 h e re = th e r m a l im p e d a n c e x 1 0 = th ic k n e s s 8 0 .0 -4 3 .8 8 7 0 .0 2 .9 3 3 .0 6 0 .0 2 .3 9 -4 2 Thermal Impedance x 10 (Km /W) 3 .5 5 0 .0 q (W) 2 .5 1 .9 4 2 .0 4 0 .0 1 .3 7 1 .5 1 .0 3 0 .0 2 6 .7 4 2 2 .4 6 2 0 .0 2 2 .9 9 2 0 .8 6 1 8 .7 2 0 .5 1 0 .0 T R 0 .0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 0 .0 2 5 0 .0 3 0 .0 3 5 0 .0 4 0 .0 4 5 q 0 .0 0 .0 5 T h ic k n e s s ( m ) (a) Thermal Impedance and amount of heat transferred to brass with varying thicknesses 38 4 .0 y w y x 3 .7 4 3 .5 = 1 0 0 .4 2 x - 0 .2 6 9 1 h e re = th e r m a l im p e d a n c e x 1 0 = th ic k n e s s 4 .0 -2 3 .5 2 .9 5 3 .0 2 .5 2 .5 2 .6 7 2 .0 2 .0 1 .5 0 1 .5 q (W) Thermal Impedance x 10 -2 2 (Km /W) 3 .0 1 .5 0 .6 0 1 .0 1 .0 1 .0 7 0 .4 2 0 .5 0 .5 0 .5 3 TR 0 .0 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 0 .0 2 5 0 .0 3 q 0 .0 0 .0 3 5 T h ic k n e s s ( m ) (b) Thermal Impedance and amount of heat transferred to Teflon of varying thickness Figure 3.5: Graphs of thermal resistance and amount of heat transferred through interface materials with different thickness The thermal resistance of the interface material increases with increasing thickness. The contact resistance of the material is obtained from the thermal resistanceintercept. The brass block has a contact resistance value of 7.70x10-5 Km2/W and Teflon 3.00x10-3 Km2/W. The much higher contact resistance of Teflon can be attributed to the very low thermal conductivity of the material. The effective thermal conductivity, on the other hand, is calculated from the reciprocal of the best fit linear trend line of the thermal resistance vs. thickness graph. The brass block has an effective thermal conductivity of 166 W/mK and Teflon 1.00 W/mK. The reported values of bulk thermal conductivities of brass and Teflon are 120 W/mK [34] and 0.25 W/mK [34] respectively. Theoretically, the effective conductivity is lower than that of the bulk value. On the contrary, the observed experimental values of the effective thermal conductivities are much higher than the reported bulk conductivities. 39 Thermal conductivities, very similar to thermal resistance, can be affected by many factors. They include physical state and chemical composition of the material, temperature and pressure. The condition of the experiment from which the reported value was taken is not specified. The reported brass conductivity has composition 70% copper and 30% zinc. The brass used, on the other hand, has a composition of 57.5% copper and 42.5% of other metals mainly zinc. The variation in the value can result from the difference in the composition. For Teflon, no mention of the material grade mentioned. Teflon, being a soft material may have their properties altered by the firm screwing, resulting in the much higher conductivity. It has been recently reported [47] that proving the accuracy of thermal resistance measurement setup based on the ASTM D5470 is problem. Continuous effort is being done to look for a suitable standard material. For the current work, the accuracy of the setup is validated using reported data from the industrial partner. The results are shown in Section 6.3. 3.4.2 Effect of ambient temperature The effect of ambient temperature on both the amount of heat transferred through the interface and the thermal resistance is insignificant. It did not make any considerable difference if the experiment was carried out with the air conditioner turned on to the lowest or highest possible temperature. It can therefore be ascertained the normal air conditioning will not affect the temperature readings. 40 0.48 0.47 23.53 TR 22.99 0.47 20.86 q 291.16 294.16 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 q (W) Thermal Resistance (K/W) 0.500 0.490 0.480 0.470 0.460 0.450 0.440 0.430 0.420 0.410 0.400 302.16 Room Temperature (K) Figure 3.6: Graph of thermal resistance and amount of heat transferred with different ambient temperatures 3.4.3 Effect of water bath temperature The temperature of the water bath has a more apparent effect as compared to the ambient temperature. The figure below shows a graph of the results obtained. The amount of heat transferred through the interface decreased with increasing water bath temperature. This is because, with the temperature of the cartridge heater fixed at 338.16K, the temperature drop, ∆Tblk, decreases with increasing water bath temperature. From the equation q = kAl AAl ∆Tblk , a decrease in ∆Tblk brings about a ∆x decrease in the amount of heat transferred to the interface. 41 90 0.8 80 0.6 0.5 TR 0.55 0.48 0.47 0.54 0.46 0.46 q 70 60 50 0.36 0.4 40 0.29 0.3 30 19.25 19.25 0.2 14.44 11.23 8.02 0.1 20 4.28 3.74 10 6.42 0 310 q (W) Thermal Resistance (K/W) 0.7 315 320 325 330 0 335 340 Water Bath Heater Temperature (K) Figure 3.7: Graph of thermal resistance and amount of heat transferred with varying cartridge heater temperatures However, it is not clear if the thermal resistance increases or decreases with increasing water bath temperature. The interface material is an alloy. Therefore heat conduction takes place via both movement of free electrons and vibrations of molecules. Mean free path of electrons tends to decrease with increasing temperature [10]. Hence heat transfer by electron movement should be more difficult and TR ought to increases. On the other hand, an increase in temperature enhances molecular vibration and therefore heat transfer by molecular impact should cause a decrease in thermal resistance. Consequently, there is an opposing influence of temperature on thermal resistance and the net result of the effect of temperature is not perceptible. Furthermore, the pressure at the interface may have increased when the temperature of the water bath is increased. The blocks were hand-tightened before the experiments with approximately same amount of force. With an increase in the water bath temperature, the blocks expanded, causing the pressure at the interface to increase. The increased pressure could have reduced the contact resistance and consequently reduced the TR value. When the water bath temperature is set at 65oC, 42 the hot end and cool end are at the same temperature. The net heat flow is, therefore, minimal and the temperature drops across the blocks are small. Hence, the measurements are not as accurate. 3.4.4 Effect of thermal grease at aluminium / sample interfaces Thermal grease is used to minimize the contact resistance at the aluminium/brass interface. From the graph below, it can be seen that the thermal grease decreased the contact resistance (thermal resistance at zero thickness) by more than half the initial value. Without thermal grease, the contact resistance is 7.7x10-5 Km2/W. With the application of grease at the interface, the contact resistance dropped to 3.6x10-5 Km2/W. 4.5 y = 60.13x + 0.77 where -4 y = thermal resistance x 10 x = thickness 3.5 3.88 -4 2 Thermal Resistance x 10 (Km /W) 4.0 2.93 3.0 3.24 2.39 2.5 2.63 1.94 2.0 y = 59.26x + 0.36 where -4 y = thermal resistance x 10 x = thickness 2.15 1.37 1.5 1.44 1.0 0.87 0.5 TR/TG 0.0 0 TR 0.01 0.02 0.03 Thickness (m) 0.04 0.05 0.06 Figure 3.8: Graph of thermal resistance and amount of heat transferred with and without thermal grease at the aluminium/brass interfaces 43 3.4.5 Effect of air movement From the following figure, air movements do not have significant effect on the thermal resistance of the brass block. The 's' in the x-axis indicates swinging motion. When the fan is blowing, the temperature around the setup drops. The drop in temperature caused the overall temperature throughout the heating, cooling and sample blocks to have a reduction in temperature. The drop in temperature, however, is small and approximately equal at both the hot and cool end. As a result, the temperature drop across the interface is approximately the same, giving the same thermal resistance value. 40 0.46 0.449 0.45 0.445 0.440 0.443 0.444 0.439 38 36 0.43 34 0.409 0.42 32 0.41 0.40 31.0 0.39 0.38 27.8 28.3 28.9 q (W) Thermal Resistance (K/W) 0.44 30 29.4 28 27.8 27.3 26 0.37 TR q 24 0.36 1/2m 2/2m 3/2m 3/2m 3s/2m Fan speed / distance from setup 3/0.5m 3s/0.5m Figure 3.9: Graph of thermal resistance and amount of heat transferred with varying air movements 44 3.4.6 Effect of operator The result of the study is shown in the figure below. The measurements preformed repeatedly by the same operator are consistent within the upper critical limit (UCL) and lower critical limits (LCL). The operator-to-operator variations are also in the Sample Range acceptable range. 0.10 Operatorjane A Operator Nor B Susan Operator C UCL=0.1026 0.05 R=0.04854 0.00 LCL=0 Figure 3.10: Data measured independently by 3 different operators This shows that the set-up is capable of producing minimal random errors and is deemed precise. 45 3.5 Conclusions The accuracy and the precision of the thermal resistance measurement setup have been elucidated. The experimental data for the thermal conductivity of the selected standard materials do not agree well with the reported data. This inaccuracy is due to the dissimilar measurement technique. The precision of the setup, however, is well within the acceptable limits. The amount of heat transferred through the interface depends on the temperature difference across the sample interface junction. The type of interface material, its thickness, the cartridge heater temperature as well as the water bath temperatures, can affect the temperature difference. The larger the temperature difference, the more heat will be transferred to the interface. The ambient temperature and the speed of the water circulating through the water jacket do not give any considerable effects. 46 Chapter 4 Material Performance I – Phase Change Materials 4.1 Introduction to Phase Change Material Phase change materials (PCM) can be classified in 3 general categories [8]: organic, inorganic and metallic PCM. Organic PCM can either paraffins or non-paraffin based. The second and third categories consist of inorganic compounds, metals and eutectics. Depending on the type of PCM, the melting point range from 30oC – 90oC. PCM absorbs latent heat to melt and release energy when it solidifies. The absorption and release of heat enabled more gradual temperature cyclings in its applications. In the microprocessor packaging application, PCM prevents the IC chip from hitting the upper critical temperature and improve reliability by reducing temperature cycling while efficiently transferring heat to the heat spreader. PCM have high latent heat of fusion per unit volume and can therefore absorb considerable amount of thermal energy during melting. These properties lead PCM to be an important material to be used as a thermal interface material. Phase change materials, manufactured by Honeywell Electronic Materials are protected by at least 2 patents [35-36]. The material is described as a compliant and crosslinkable thermal interface material that comes in dispensable liquid paste and elastomer film. Also described in the patents are methods to improve thermal conductivity of the polymer system. The material is, in principal, comprised of paraffin wax, liquid rubbers, conductive fillers and some additives. 47 The olefin containing interface materials with appropriate thermal fillers have a thermal capability of less than 0.5 cm2.oC/W. The thermal performance will not degrade after thermal cycling or flow cycling because liquid olefins will crosslink to form a soft gel upon heat activation. The material is also reported to resist ‘squeezed out’ out as thermal grease does in use, and interfacial delamination is not an issue during thermal cycling. A literature review on the available types of paraffin wax and liquid rubber is in Appendix A. 4.1.3 Thermally Conductive Fillers Suitable filler materials for the PCM include [38-39] silver, copper, aluminium, boron nitride, aluminium nitride, silver coated copper, silver coated aluminium and vapour grown carbon fiber (VGCF). VGCF are available in various lengths and diameters. They are highly graphitized type by heat treatment and have thermal conductivity of 1900 W/mK. It is difficult to incorporate large amounts of VGCF in polymer system because they do not mix well. Large amounts of other fillers need to be added if VGCF are used. It is noted that substantial amounts of spherical filler particles is advantageous to maximize packing density and provide some control of the thickness during compaction. The particle size of spherical filler is about 1 to 20 µm. [38-39] 48 4.1.4 Other additives Accelerators in the form of a tertiary amine may be added into the material as an accelerator for room temperature cure. Antioxidants may also be added to inhibit oxidation and thermal degradation of the cured rubber gel. These antioxidants may be a phenol type or an amine type. Typical cure accelerators include tertiary amines such as didecylanethylamine. Dispersion of filler particles can be facilitated by addition of functional organo metallic coupling agent such as organosilane, organotitanate or organozirconium. Coupling agents bridge the matrix-filler boundary and are typically inorganic-organic additives that have one or more hydrolysable groups and a matrix-specific organofunctional group. For example, a lot of organofunctional silanes are capable of silanol-group formation for bonding with mineral surfaces and an epoxy functionality which can react with, or promote adhesion to the matrix. [42] 49 4.2 PCM Characterization The phase change material (PCM) has been characterized in a number of ways to understand its basic properties. The characterization was done to understand its thermal properties and structural properties. 4.2.1 Thermal Analyses Thermal characterization that have been carried out on the PCM include the thermal gravimetry analysis (TGA) and the differential scanning calorimetry (DSC). The amount of weight loss at elevated temperatures measured using TGA is useful to understand the thermal stability of the material. Significant weight losses at temperatures at which the material is normally subjected to, in processes and in applications can affect its thermal performance. Substantial weight loss may affect both the bulk and interfacial thermal performance of the PCM. Essential organics may be lost leading to PCM/metal contact deterioration. The amount of heat the material can sustain may also be reduced due to the weight loss. DSC, on the other hand, determines the amount of enthalpy changes in a substance with temperature as well as the amount of energy needed to cause a unit rise in temperature of the material. Other information that can be obtained from DSC analyses are the range of temperature for melting transition as well as the onset of melting. As melting is an endothermic process, a dip (with an exo up heat flow axis) will be observed when the PCM is brought through the melting temperature. The start and end of the dip give an indication of the temperature range at which the melting is occurring. More information about the polymer can also be obtained from the shape 50 of the dip. [43] A split dip shows melting with crystal orientation, a sharp dip shows melting of oriented polymer and multiple dips show different phases or sizes of crystals of one polymer or a mixture of polymers. A few temperature profiles were used in the TGA and DSC runs. For TGA, the PCM was heated to 300oC to study the weight loss the material at elevated temperatures. A similar study was carried out using the DSC. The highest temperature studied was 500oC as excessive weight loss may contaminate the DSC cell at higher temperatures. 4.2.2 Structural Analysis Fourier Transform Infra-red (FTIR) Spectroscopy was employed to study the structural properties of the PCM. Atoms and molecules vibrate with frequencies in the IR range and these vibrations are able to give useful information on the specimen under study. An IR spectrum can be compared to known databases for identifying an unknown spectrum and the bands can give ideas on the molecular structure of the specimen. The spectrum can be interpreted in two areas: the functional group region and the fingerprint region. Peaks in the functional region (4000-1500 cm-1) are characteristic of specific kinds of bonds, and therefore can be used to identify whether a specific functional group is present. Peaks in the fingerprint region (1500-400 cm-1), on the other hand, arise from complex deformations of the molecule. They may be characteristic of molecular symmetry, or combination bands arising from multiple bonds deforming simultaneously. 51 4.3 Experimental Methodology 4.3.1 TGA Typical sample sizes for thermal analyses like TGA or DSC are 5-10mg. As PCM is a metal filled polymer material, the effective weight of polymer to be analysed is much lesser than the overall weight. As the amount of metal fillers is about 4 times that of polymer, 20-30 mg of PCM samples were used for TGA and DSC analyses. For the TGA analyses, PCM was placed into ceramic sample cells. The sample was then heated in a TA instruments SDT 2960 at a controlled rate of 10oC/min up to 500oC. 4.3.2 DSC 20-30 mg of PCM sample was crimped in a hermetic aluminium pan. The sample pan was then placed in a TA Instruments DSC 2920 cell together with an empty pan as a reference. The pans were then heated from sub-ambient temperature to 300oC at a controlled rate of 10oC/min to determine the melting onset and the melting enthalpy. The specific heat capacity of PCM was also determined using this methodology. The setting of the instrument was changed so that the signals were modulated. A sapphire was then used to calibrate the specific heat capacity function of the cell. The measurements were done at a controlled rate of 5oC/min, which is the maximum rate in the modulated mode. 52 4.3.3 FTIR PCM was pressed into a pellet with KBr matrix using 10 tonnes of force. IR beam of wavenumber 4000 – 400 cm-1 was then passed through the pellet in the Bruker Equinox 55 FTIR instrument. 4.4 Results and Discussions The TGA thermogram of PCM is as shown below. At an onset of about 150oC, PCM loses weight with the highest rate of loss taking place at around 470oC. The PCM is a metal filled polymer and at 500oC almost all organics would have been lost and the weight of the sample reaches a plateau. From the thermogram in the following figure, 13% of the total mass, indicating that the filler loading in PCM to be about 87% by weight. This thermal analysis also shows that the material cannot be brought above 150oC in its applications. Upon reaching this onset of weight loss, material degradation will have occurred and the thermal performance of the material will most likely be affected. 53 onset of weight loss Figure 4.1: TGA thermogram of PCM The DSC thermogram of PCM is as shown in the following figure. The melting temperature of PCM is slightly higher than the ambient temperature. The analyses, therefore, has to be done from a sub-ambient temperature. The DSC cell was loaded with the sample and then brought to a sub-ambient temperature using dry ice. It is apparent, from the thermogram, that the onset of melting is gradual. At temperatures lower than ambient, the PCM is already changing phase as indicated by the deviation form the baseline. This explains why the material is a soft mouldable material at room temperature. The maximum rate of melting took place at temperature ranging from 42oC to 45oC. The phase change is an endothermic process and about 2.2J of energy was absorbed in the process. The melting behaviour of the PCM can be determined by the shape of the melting dip. The gradual melting is reflected by the broad valley-like shape. This is consistent with the fact that the material is made up of amorphous polymer (rubber) that is 54 unlike the melting of a highly oriented polymer which gives a sharp melting onset and steep valley. The material is also of single phase because of the absence of multiple dips. Figure 4.2: DSC thermogram of PCM The trend in specific heat capacity of PCM is shown on the following graph. The specific heat capacity of the material is relatively constant over temperature at values 1.0 to 1.1 J/g/oC. There is a slight increase at the phase change region (40oC – 45oC) due to the absorption of energy. The specific heat capacity values are, however, comparable to the specific heat capacity of carbon (0.7 J/g/oC) and aluminium (0.9 J/g/oC) that makes up the main mass of the PCM material. This amount of heat absorbed by the material to give a unit rise in temperature is an important characteristic of the material. The higher this value is, the better the material will perform in the real-life applications. More experiments have been done (Chapter 5) to demonstrate the importance of this characteristic and the melting enthalpy with respect to the packaging processes and applications. 55 3.0 Instrument : TA Instruments (DSC) Ramp rate : 5oC/min Purge gas : Nitrogen Sample Pan : Hermetic Aluminium Specific Heat Capacity (J/g/oC) 2.5 2.0 1.5 1.0 0.5 0.0 -20 -10 0 10 20 30 40 50 60 70 80 90 100 o Temperature ( C) Figure 4.3: Specific heat capacity of PCM over temperature The IR spectrum of the material is depicted in the following figure. As the PCM is essentially a metal filled polymer, the signals from the analysis are weak. A semitransparent pressed KBr pellet could not be obtained because of the presence of the fillers. This has, in effect, caused the level of transmission to be low and noisy. The situation is made worse with the numerous components present in the PCM specimen, giving are numerous overlapped bands observed in the analysis. In the functional region, it can be determined that the PCM contains the functional groups shown in the diagram below. The O-H groups originate from the liquid rubber where the malenized functional group as well as the reactive groups within the liquid polymer chains. Strong bands resulting from C-H stretching were also observed due to the high hydrocarbon content in the paraffin wax. The fingerprint region, on the other hand, is not conclusive because of the proprietary multi-components of the raw materials in PCM. 56 0.38 Transmittance (%) 0.37 0.36 0.35 2850 C-H, 2920 O-H(acid) 0.34 3440 0.33 1630 water 670 CO 2 water, O-H (alcohol) 0.32 4000 3229 2457 1686 914 Wavenumber (cm-1) Figure 4.4: Signature IR spectrum of PCM 4.5 Conclusions PCM has been characterized for its basic properties like its thermal stability in terms of weight loss and its melting behaviour. The material takes in 2.2J of energy to melt, thereby reducing the preventing the temperature the rise further. It is stable up to 150oC that is well below the predicted maximum junction temperature of a high performance chip at 125oC [3-5]. The specific heat capacity of the material has also been experimentally determined to be 1 J.g-1.oC-1. In melting the material, therefore, keep the temperature down by 2oC. The structure of the material has partially been elucidated using FTIR technique but the results are not conclusive due to the proprietary nature of the material. 57 Chapter 5 Material Performance II – Metal Surfaces 5.1 Metal Surfaces One of the first surfaces that a thermal interface material (TIM) encounters in a typical package is a metal surface. Metals are known to have high thermal conductivity but the surface texture and conformances of their surfaces may result in detrimental effects on the effective conductivity of the material. Metal surfaces, in their pure states, have high surface energy and to reduce the energetic, oxides are formed. The surface features, on the other hand, originate from the roughness of the surface. The interaction of TIM – metal surface is therefore useful to understand their effects on the package thermal performance. Commercial samples that are used in the semiconductor industry were utilized in an effort to study these interactions. Chemical and mechanical properties of these surfaces are analyzed and relations to its overall contribution to the thermal properties are deduced. 5.2 Surface Characterizations There are various types and grades of metals used in the IC packaging. These metals function as heat spreaders and heat sinks for more efficient heat dissipation. Silver and gold are two of the best heat conducting metal but they are not widely used in the semiconductor industry as heat sink or heat spreader materials because they are very expensive and relatively heavy. More widely used metals are aluminium, copper and nickel. 58 The surfaces under study in this research belong to the heat spreader category and five types of heat spreaders were selected to simulate the actual application of the PCM on metal surfaces. These different spreaders surfaces are made of the following materials: a) Anodized aluminium b) Aluminium / silicon carbide composite c) Black copper oxide d) Nickel plated copper (Bright and dull surface finish) Aluminium is a popular metal used in the manufacturing of heat spreaders and heat sinks due to their light weight, high thermal conductivity and corrosion resistance. Aluminium, however, is relatively soft. The two types of aluminium heat spreaders used in this study are hardened to counter that issue. Anodizing aluminum is an electrochemical process that produces a hard and porous oxide layer on the surface of the metal. The pore structure when viewed under a high-powered microscope looks like a honeycomb or metal sponge. The pores are normally sealed and stabilized when color is deposited into the pore structure. Anodized aluminium is a widely used material in heat sinks where weight and material strength are very crucial properties. These different heat spreaders are actual samples being used in the semiconductor industry. Aluminium / silicon carbide composite, on the other hand, is a metalceramic composite that is built to improve the stiffness of the material. 59 Although copper is a best conductor of heat after silver and gold, it is highly prone to oxidation. The semiconductor industry has been able to counter this problem by producing heat spreaders with stable black copper (II) oxide layer. A grade of heat spreader made of this material has been chosen. The other 2 types of chosen heat spreaders are nickel plated copper. One of them has a dull finish while the other is a bright finish. Images of the metal surfaces were taken using the scanning electron microscope (SEM) and mechanically characterized for its surface roughness and surface tension. X-ray photoelectron spectroscopy was then carried to study the surfaces chemically. 5.3 Experimental Methodology 5.3.1 Surface Imaging The metals were mechanically cut to 2cm x 2cm square, without touching the surface area of interest, to fit the SEM sample holder. The sample and its holder were then mounted into the JEOL SEM model JSM-5200. The surface image was then captured in a PC via SEM Afore software, at 2000x magnification. The scan areas are 60µm x 50µm. 60 5.3.2 Surface Roughness The surfaces of the metal heat spreaders were profiled based on a mechanical stylus. This technique uses a pick-up head for converting the height variation into an electrical signal when the contacting stylus moves across the metal surface. The advantage of using this method is that the metal heat spreaders need not be cut because the Kosaka Surfcorder model SE-30C can take a wide variety of sample size. The scan area using this technique is also larger if compared to other techniques like the Scanning Tunneling Microscopy (STM), Tunneling Electron Microscopy (TEM) and Atomic Force Microscopy (AFM). These microscopy techniques have a typical scanning area of 100µm square while the scanning area using the mechanical stylus technique is in the mm ranges. The settings used for the analysis are as follows: Length measure for each run: 2.5 mm, Drive Speed: 0.1 mm/s Vertical Magnification: 5000x, Horizontal Magnification: 20x, Cut - off: 0.8 mm The analyses were recorded by the electro pen and Ra value displayed. 4 scans were made on each metal surface. The scan directions are as follows: Figure 5.1: Scan directions for surface roughness measurements 61 5.3.3 Surface Tension Measurement The free energy of a surface (surface tension) is a good measure of their adhesion properties and the wettability of the surface can be easily measured using contact angle. Contact angle measurements can be done in many ways [44], including the drop-bubble methods, reflected Tensiometric method, level-surface method and capillary rise method. The method employed in this study is the direct observation sessile drop method. The instrument used is the ramé-hart contact angle goniometer model 100-00. Deionized (DI) water and diiodomethane were used as the dropping liquids. In the experiments, liquid was dropped onto the metal surface using a micro syringe and the drop profile directly observed using a microscope. The eyepiece of the microscope has an in-built protractor that facilitates the contact angle measurements. 10 angle measurements were made an average was taken to calculate the surface tension of the metal. 5.3.4 Core Electron Analysis X-ray photoelectron spectroscopy (XPS) is an informative method of investigating the elemental content of a surface. The instrument used for this set of experiment is VG ESCALAB MKII. The heat spreaders were mechanically cut without touching the area of interest (1 cm2) and dried in a vacuum oven for 30 minutes prior to the analysis. 62 5.4 Results and Discussions The micrographs of the metal surfaces taken using the SEM are shown in the Figure 5.1. The surface of anodized aluminium shows a honeycomb-like arrangement. It is evident that the porous structure left behind from the anodizing process is not sealed up by any coloring process. The Al/SiC surface, on the other hand, was observed to have a fiber-like structure. The apparent fiber like may be due to the aluminium being aligned in straight lines with the silicon carbide dispersed between the fiber-like aluminium, or it could be due to machining of the metal spreader. The random dispersion helps in the strengthening of the heat spreader while maintaining aluminium lightweight. The copper oxide and nickel surfaces appear relatively homogeneous, with no apparent patterns. It is likely that the surfaces have coatings that cover any porosity that might have been left behind from the manufacturing of the heat spreaders. The surface of the bright-finished nickel surface is seen to be rougher than the dull surface. Brightness on the surface is apparently not due to the smoothness of the surface but possibly from a brightening surface coating. (a) Anodized Aluminium (b) Al/SiC 63 (c) CuO (d) Dull Ni Plated Cu (e) Bright Ni Plated Cu Figure 5.2: SEM micrographs of heat spreader surfaces The surface roughness measurements of the heat spreader surfaces confirmed the results from the SEM imaging. The following figure shows the 2-dimensional surface profile micrographs from the roughness measurements. Anodized Al Al/SiC Black CuO Dull Ni Bright Ni Figure 5.3: 2-dimensional surface profile of heat spreaders 64 The corresponding roughness values are shown in the figure below. 1.4 1.2 1.13 Ra (m m ) 1.0 0.8 0.6 0.57 0.4 0.33 0.37 0.36 0.2 0.0 Anodized Al Al/SiC Black CuO Dull Surface Ni Bright Surface Ni Figure 5.4: Roughness of heat spreader surfaces The surface of the anodized aluminium is the roughest because of the porosity as seen in the SEM micrograph. It has an average roughness of 1.1 µm. The aluminium silicon carbide composite has the lowest roughness value of 0.3 µm and this is due to the filling up of silicon carbide in the cavities created by the ridges of fiber-like aluminium. Following that, the roughness values of the nickel and copper oxide are effectively the same. The dull nickel is very difficult to handle as it surface is very prone to moisture and dirt settling. This sensitivity may have caused the range of roughness value to be bigger. This same problem may also be the cause the bright nickel surface is more widely used. The wettings on the heat spreaders do not show the same trend as the surface roughness. It was expected that the extent of roughness is proportional to the contact angle. From the figures below, it can be seen that the anodized aluminium exhibit the same wetting for both liquids. The porous surface allowed the liquid to seep in, 65 giving equal contact angles. The bright nickel surface also exhibits the same properties to the liquid. The coating used on the surface is proprietary but it is believed to not only enhance the luster appearance but also an adhesion promoter. It was not possible to pick the inherent coating on the bright nickel surface to do analysis because it was too thin. 90 85 80 78 o Contact Angle ( ) 70 66 60 50 40 55 52 45 32 26 30 20 Diiodomethane DI water 10 0 Anodized Al Al/SiC Black CuO Dull Ni Bright Ni Figure 5.5a: Contact angle on heat spreader surfaces Surface Free Energy (mJ/m2) 100 Polar Non Polar Total 80 60 64 54 40 20 29 25 34 25 33 31 39 37 25 22 17 12 9 0 Anodized Al Al/SiC Black CuO Dull Ni Bright Ni Figure 5.5b: Surface free energy of heat spreaders 66 The chemical analyses of the surfaces are shown in the following figure. The strong oxygen signal and relatively weak aluminium signal from the anodized aluminium shows that the anodizing process created a thick oxide layer. The spectrum from the analysis of the aluminium / silicon carbide (Al/SiC) similarly show weak aluminum signal due to the dispersion of the silicon carbide. There are numerous peaks in all the spectra. This is an indication that there additional surface finishes on all the heat spreader. However, there is no information on the final surface finishes as this information is proprietary. 12000 O 1s 538.9 10000 Intensity (a.u.) 8000 752.5 6000 774.0 1001.9 564.4 4000 Al 2p 81.7 C 1s 292.4 125.8 528.9 2000 0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 Binding energy (eV) (a) Anodized Aluminium 67 7000 O 1s 533.2 6000 746.1 769.7 Intensity (a.u.) 5000 995.5 1013. C 1s 287.0 4000 555.8 Si 2p 104.3 3000 523.5 Al 2p 75.3 2000 120.4 268.8 1000 0 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000 1050 1100 Binding Energy (eV) (b) Aluminium / Silicon Carbide CuO Cu 2p Cu 2p 946.0 16000 O 1s 541.8 14000 Cu 2p 965.4 Cu 2p 974.0 12000 Intensity (a.u.) 348.3 10000 753.6 427.9 504.2 775.1 8000 6000 327.9 89.2 4000 135.5 16.1 C 1s 296.7 2000 0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 Binding Energy (eV) (c) Black Copper Oxide 68 12000 Ni 2p 855.7 O 1s 531.1 10000 994.4 743.9 479.5 8000 Intensity (a.u.) Ni 2p 872.9 407.4 6000 4000 C 1s 284.9 67.7 2000 0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 Binding Energy (eV) (d) Dull Surface Nickel Plated Copper 1003.0 Ni 2p O 1s 540. 14000 755.7 12000 1022.3 776. 489.1 Intensity (a.u.) 10000 417.1 8000 448.3 C 1s 294.6 6000 74.4 4000 403.1 122.6 2000 11 8 0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 Binding Energy (eV) (e) Bright Surface Nickel Plated Copper Figure 5.6: XPS spectra of metal heat spreader surfaces 69 5.5 Conclusions The surface physical structures and morphologies have been revealed. Anodized aluminium heat spreader has a porous surface that leads the contact angles to be low and surface energy to be high. The aluminium / silicon carbide, on the other hand, has comparatively high ordered surface. The silicon carbide addition makes the surface more non-polar, reflected by the low polar surface free energy. The surface of black copper oxide has a high surface free energy that could have resulted from additional surface finish. The nickel plated copper heat spreaders have moderately rough surface and low surface free energy, despite the apparent bright finish of one of the heat spreaders. PCM is essentially an organic material and it is expected to wet better on a non-polar surface. The thermal performance that is affected by how well the material is able to wet the surface is therefore expected to be better on non-polar surfaces. From the measurements of the surface free energy, PCM will have the best thermal performance on black copper oxide heat spreader. 70 Chapter 6 Interaction of PCM on Metal Surfaces The three most important capabilities of PCM in its application as a thermal interface material in the semiconductor packaging is 1) the ability to conform well on the contacting metal surface 2) be able to retain its adhesion to the surface after various process conditioning 3) to continue to exhibit acceptable thermal performance throughout its service life. Various experiments were done to understand the performance of the material as it undergoes various processes in the actual IC packaging process. The effects of different application methods as well as process conditions were studied. The performance of the material throughout its service life was accelerated using various standard reliability conditions. 6.1 Process and Applications of PCM in IC Packaging There are various process steps in IC packaging. After the die has been attached to the substrate, the first thermal interface material (TIM 1) is attached. The protective heat spreader is then attached. The following figure is a schematic of a packaged IC. Figure 6.1: Schematic of a packaged IC 71 There are 2 ways by which the heat spreader can be kept in place: the use of clips or a structural adhesive. An additional step to cure the adhesive is needed if structural adhesives are used. Following this step, another layer of interface material (TIM 2) is applied and finally the heat sink is attached. The method of thermal interface material application can vary a great deal. In the case of clips, the amount of pressure exerted by the clip affects the thickness of the eventual interface material. In some processes, spacers are employed to get the desired thickness. It is therefore instructive to understand how the pressure affects the thermal performance of the material at variable and fixed thicknesses. The type and condition of the metal surface also affects the eventual thermal performance of the material. Different types of metal heat spreaders used in the semiconductor industry have been chosen and tested for their thermal compatibility with the PCM. A specific type of heat spreaders that is widely used is nickel plated copper. It was selected for this study and surface modifications were carried out to evaluate how these changes affect PCM’s contact resistance to nickel plated copper. In its actual applications, PCM is subjected to high temperatures and possibly high humidity. From Figure 1.2, it can be seen that the temperature that the material is constantly subjected to an environment of at least 85oC. It was not possible to subject the PCM to the actual application conditions in this study as it is too time consuming. The alternative that is widely employed by the industry to understand the service life of the material is through reliability tests. These tests allow the material to undergo conditions that would simulate the material having undergone long term application 72 conditions. There are widely available Joint Electron Device Engineering Council (JEDEC) standards that describe tests to evaluate different properties of the material. The JEDEC Solid State Technology is a semiconductor engineering standardization body of the Electronic Industries Alliance (EIA). It is a trade association that represents all areas of the electronics industry. [45] Above all, the PCM was evaluated for its thermal conductivity. Thermal conductivity value is a typical benchmarking property. More advanced comparisons of thermal interface materials involve the thermal resistance of the material at specific pressure, thickness, temperature and possibly on specific surfaces. 6.2 Experimental Methodology Unless otherwise specified, the thermal resistance measurements are done using the following specifications: 1) Heat spreader: 25.4 mm x 25.4 mm nickel plated copper with bright surface finish 2) Spacers: 200+5 µm copper leadframes 3) Interface material: PCM (standard version) 4) Interface temperature: 323K 5) Contact pressure: 0.2 MPa 6.2.1 Sample Preparations The spacers are cut to very small pieces ([...]... of contacting surfaces ƒ Average temperature of interface ƒ Contact pressure and contact history of surface ƒ Duration of contact with regard to relaxation effects ƒ Vibrational and directional effects ƒ Contact cleanliness Filling the air gaps with conductive and compliant materials, like thermal grease, and using a very high contact pressure normally reduces contact resistance Advances in materials... amount of heat transferred k = thermal conductivity of the material A = cross-section area of material ∆T = temperature gradient per unit length of the material L A higher thermal conductivity shows a greater ability of the material to conduct heat A commonly used unit of thermal conductivity is W/mK 15 2.2.2 Specific Heat Capacity Specific heat capacity of a material, on the other hand, is the change. .. cross-section area of material k = thermal conductivity of the material L = length of the material From equation 2.26, it can be seen that thermal resistance of a homogeneous material is proportional to the distance of heat travelled Thermal impedance, on the other hand, is the product of the thermal resistance and the cross-sectional area Thermal impedance = θ .A = L ∆T A = q k where θ = thermal resistance... widely used and has been discussed in several papers [16-21] A major part of this project, however, focuses on the thermal resistance of a commercial thermal interface material (TIM) As the name implies, TIM is a material sandwiched between a chip and a heat spreader and/ or a heat sink Details of thermal interface materials are discussed in Section 1.2 At first glance, none of the 4 mentioned methods... uncertainties present in real surfaces The parameters of concern include: [6] 20 ƒ Number of intimate contacts ƒ Shape of contact points: circular, elliptic, band, or rectangular ƒ Size and arrangements of contact points ƒ Geometry of contacting surfaces with regard to roughness and waviness ƒ Average thickness and fluid (gas, liquid or vacuum) of void space ƒ Pressure and conductivity of void space ƒ Hardness... 25µm Apart from the absolute value of the thermal resistance of the material at the particular thickness, this set of experiments enables the values of contact resistance and effective thermal conductivity of the material to be obtained The thermal conductivity of the material can be determined from at least 2 measurements of the thermal impedance, at the same temperature and contact 35 pressure Thermal... medium of conduction, are usually not compressible The pressure in the material is therefore usually constant Thermal diffusivity incorporates inherent properties of a material; the thermal conductivity, k and the specific heat capacity, c This integration of intrinsic properties facilitates many calculations because tabulated values of common materials at specific conditions are widely available 2.2.4... Spreading Resistance Not all thermal conduction takes place across media of the same cross-sectional areas More often than not, conduction takes place through a solid or across an interface with different cross-sectional area In an electronic package, for example, 17 heat is conducted through many layers of different cross-sectional areas The figure below illustrates the example Heat Flow Heat Spreader... kAl = thermal conductivity of aluminium AAl = cross-sectional area of the aluminium block ∆Tblk = temperature gradient across the top aluminium block ∆x 3.2.2 LabVIEW The setup is capable of measuring both the transient as well as the steady state conditions These capabilities have been made possible by specifying the rate of data acquisition parameter of the LabVIEW program The LabVIEW program is a. .. of Materials In most, if not all heat conduction mathematical analyses, at least one of the following physical properties is essential • Thermal conductivity 14 • Specific heat capacity • Thermal diffusivity • Spreading resistance • Thermal resistance • Contact resistance 2.2.1 Thermal Conductivity Thermal conductivity is often regarded as the primary physical property of the material where heat conduction ... Surface Characterization (a) Areas of interest within a packaged chip Chemistry and Interfacial Mechanics of a Phase Change Material on Metal Surfaces PCM Characterization Thermal Resistance Measurement... characterization, design and construction of thermal resistance measurement setup and surface characterization of metals The PCM and metal surfaces were characterized to understand their basic... space ƒ Hardness of contacting surfaces ƒ Average temperature of interface ƒ Contact pressure and contact history of surface ƒ Duration of contact with regard to relaxation effects ƒ Vibrational