Design for Thermal Issues

Unit 6: Thermal properties of materials

In Unit 5 we described the physical processes of heat transfer, and introduced the idea of thermal resistance as a technique that could be used to create a simple model for quantifying the thermal challenges for component and circuit. It will have become clear that many heat transfer issues are materials-dependent, whether we are looking at heat spreading by conduction away from a die, or at conduction and then convection into a surrounding fluid. In this Unit, therefore, we shift our attention to the materials used for taking away heat.

Unit contents

Introduction

The thermal properties of materials are important throughout the construction of a printed circuit assembly; in the integrated circuit package, the board, and in any external heat sinks. And the approach has to be quite detailed. To take the example of an integrated circuit, there are thermal aspects to consider for the die attach adhesive, the lead frame, and the moulding compound. In power semiconductors, the entire package has to be analysed thermally, and a wide range of materials is used to enhance the thermal performance. For the printed circuit board, the traces in a standard board will affect its thermal properties, but use is also made of board structures in providing thermal vias or additional cores as heat spreaders. For the heat sink, not only is its material important, but so is the way in which it is bonded to the heat-generating component.

There will be more detail about integrated circuits in Unit 10 and about heat sinks in Unit 13 ; our focus in the sections that follow is on the materials themselves, looking first at material properties in general, and at the factors that are important for the application – these are not just the thermal conductivity. We will then examine the materials used for constructional purposes, for boards, packages and heat sinks, and then at what we have called ‘thermal management materials’, a range of frequently quite exotic materials used to improve thermal interfaces.

[ back to top ]


Thermal properties

Thermal energy transfer in a solid takes place by the migration of free electrons and ‘lattice waves’ or phonons, the thermal conductivity k being the sum of contributions from the electrons (ke) and from the lattice (kp):

$k = k_e  + k_p $

ke is proportional to the electrical conductivity, and so inversely proportional to the resistivity ρe. For pure metals with a low ρe, the electron component is much larger than the lattice component and the electrical and thermal conductivities are related by the Wiedemann-Franz Law:

\[L = \frac{k}{{\sigma T}}\]

where L is the Lorentz number, k is the thermal conductivity, σ the conductivity and T the temperature. Theoretically, the Lorenz number should be 2.44×10−8ΩW/K2 for any metal where the heat energy is carried wholly by electrons; in practice, Lorenz numbers for most metals lie within the range 2.24–2.60×10−8ΩW/K2 over 0–100°C, refractory metals such as molybdenum and tungsten having slightly higher values.

By contrast with pure metals, the lattice contribution is no longer negligible for alloys with substantially greater values of ρe, whilst the thermal conductivity of non-metallic solids is determined primarily by the lattice component. A factor that has an important effect on this component is the regularity of the lattice arrangement, so that crystalline materials such as quartz have a higher thermal conductivity than amorphous materials like glass, and certain crystalline non-metals such as diamond and beryllia have values of thermal conductivity that are higher even than good conductors such as aluminium.

material
thermal conductivity (W/m·K)
diamond
2,000–4,000
copper
390–400
beryllia
180–300
aluminium nitride
80–260
aluminium
200–230
silicon
100–150
tin-lead solder
≈50
alumina
18–36
FR-4
0.16–0.24
Table 1: Typical values of thermal conductivity of electronic materials

Note that these values are only indicative, as the exact figure for any material will depend on its purity and the presence of any intentional alloying elements, on the exact crystal structure, and on the temperature of measurement, which is why Table 1 quotes ranges. Always check the measurement conditions, and be suspicious if figures are claimed to be more accurate than ±2%.

For many calculations, we use the convenient simplification that thermal conductivity is independent in temperature. However, the thermal conductivity of many materials varies substantially with temperature. For materials where the bulk of conduction is through electrons, thermal conductivity reduces with temperature, reflecting the increase in electrical resistivity due to the higher level of interaction between electrons and the lattice. But there are anomalies, particularly in materials of higher resistivity and at extremely high temperatures, as you will see from the typical values shown in Figure 1.

Figure 1: The temperature dependence of the thermal conductivity of selected solids

The temperature dependence of the thermal conductivity of selected solids

After Incropera and DeWitt

In liquids, the intermolecular spacing is much larger than in solids, and the motion of the molecules is random, rather than fixed. In consequence, the thermal conductivities of liquids are generally smaller than those of solids. In most cases thermal conductivity decreases both with increasing temperature (water is a notable exception) and with increasing molecular weight. Water, for example, has a thermal conductivity of 0.67W/m·K, and glycol (used for antifreeze) a value of 0.25 W/m·K.

Values of thermal conductivity for most gases and vapours lie between 0.01 and 0.03W/m·K at room temperature; air is 0.026 (source). The values increase roughly in proportion with the absolute temperature, and also increase with pressure, but only at around 1% per bar.

With air, increasing the moisture content reduces its thermal conductivity, contrary to what one might expect, because the thermal conductivity of water vapour is lower than that of air (source). However, the effect is only small, rising from 4% at 50°C to only 25% at 100°C. For practical purposes, therefore, the variation of thermal conductivity with pressure and humidity can be ignored.

Activity

Important factors for the application

Whilst it is easy to think that thermal conductivity is the ultimate factor, this is in fact only one of the issues to be considered. Of course, having read Tony Kordyban’s recommendation to use only that level of thermal control which is required for satisfactory and reliable operation, you will be aware that cost is an issue. But what other factors are important?

Think about this in relation to a wide range of applications before looking at our comments.

 

Apart from the properties contained in our answer, the thermal capacity of the material may be important, because this determines the thermal time constants of the overall structure, and the rate at which changes in dissipation are reflected in its temperature distribution.

In Unit 1 we explained the term specific heat capacity, and it is worth reviewing the figures given for typical materials in Table 2. Note the relatively low figures for materials such as copper and aluminium, compared with water. This is one reason why water is often used for transferring heat from a high-dissipation equipment to an external heat exchanger. On the other hand, if we compare the thermal conductivities of the materials, it becomes very clear why copper and similar materials are used for transferring heat away from the immediate vicinity of the processor.

material
specific heat capacity (J/g·K)
water
4.18
air (100°C)
1.0
aluminium
0.89
alumina
0.76
silicon
0.70
copper
0.39
tin-lead solder
0.17
Table 2: Typical values of specific heat capacity of electronic materials

[ back to top ]


Constructional materials

A wide range of metals and ceramics is used in applications where the thermal properties are important. And, as we will see later, increasing use is being made of composite materials, combining favourable properties of metals with materials such as ceramics and carbon fibre.

Metals

Although composite materials offer certain advantages, their use tends to be confined to specialist applications on account of their significantly higher cost. For most users the most important heat-conducting metals are the various grades of aluminium and copper. Using the term “the various grades” is a reminder that, as with FR-4, the generic term includes many variants with similar but distinct characteristics.

The thermal and mechanical performance of both metals depends on their intrinsic purity and, in the case of aluminium, the presence of alloy constituents:

Activity

Key differences between copper and aluminium

What are the key differences between copper and aluminium that affect the choice between these materials for the purposes of heat management in electronics?

For this you will have to think about the ways in which copper and aluminium are used and relate these to the properties of the materials. You will probably have to carry out some web research in order to obtain sufficient information.

Please think hard about this before looking at our solution.

 

Apart from the many issues raised in our answer to the activity, both aluminium and copper have relatively high coefficients of thermal expansion relative to silicon, as you will see from Table 3. As a result, they are not ideal for direct mounting of die. However, copper alloys such as C19400 (CuFe2P) and C70250 (CuNi3Si1Mg), which have similar CTE values, are frequently used for lead-frames; the reliability under extended temperature cycling will depend on the compliance of the die attach adhesive.

material
CTE (ppm/°C)
thermal conductivity (W/m·K)
Alloy 42 (Fe58Ni42)
4.5
15
alumina 96%
6.5
21
aluminium
23
200
Au80Sn20 hard solder
14
57
copper
16.5
391
epoxy (silver-loaded)
20–40 below Tg
2
kovar (Fe54Ni29Co17)
5.3
16
nickel
13
91
silicon
2.8
146
Sn10Pb90 solder
28
36
Sn63Pb37 solder
24
51
Sn96.5Ag3.5
30
32
mild steel
9.5–12.5
48–52
These figures are mostly from Sergent & Krum Thermal Management Handbook.
Epoxy resin figures are typical, but some vendors report figures as high as 60 W/m·K.
Table 3: Mechanical properties of materials related to component attachment

For applications over wider temperature ranges, alternative materials are often preferred, mostly nickel-iron alloys such as Alloy 42, as these have a reduced CTE. But they also have inferior thermal properties. A final problem is that the alloys need to be protected against corrosion else they tend to rust.

For power semiconductors, especially where hermetic sealing is required, mild steel is a common material, on account of its low cost and its suitability for creating compression seals. This style of construction was first devised for the TO-3 package. For this style of application, a mild steel core may also be provided with a thin copper coating, fusion welded to the surface. [You will also come across other bimetallic structures used in other types of semiconductor packaging.]

We have already mentioned copper as a constructional material, but rather ignored its major thermal use within the board itself. As we will see in the practical exercises, substantial heat is carried through both copper foils and vias, transferring heat from where it is generated either throughout the board or directed towards heat sinks. A common feature is a heat-dissipating component placed above a number of thermal vias, connected through a compliant material to a heat spreader or heat sink.

The thermal properties of boards can be improved by judicious selection of inner planes used for power and ground connections. Further improvement can be made by the addition of metal cores, although these are more frequently employed to reduce the CTE of the overall laminate in order to make it more compatible with devices such as LCCCs.

[ back to top ]


Ceramics

One of the problems of metals is that they conduct electricity as well as heat, whereas many constructions need electrical isolation. In some cases this can be provided by a loose washer or a surface coating, but in both cases there is a danger that the sheet of insulation may be punctured by protrusions that distort and break through the material when pressure is applied. This is why good practice is always to de-burr mounting holes for power semiconductors – a burr can create unwanted electrical continuity, whilst at the same time impairing thermal contact.

Although pre-coated metals are used in electronics for fabricating enclosures, the voltage withstand of the coating is rarely sufficient for professional electronic applications. Aluminium may be anodised, but electrolytic passivation coatings on steel are not isolating. Glass layers have been used on steel, in particular in the form of ‘vitreous enamel’ on the steel pressings that were once commonly used for white goods; the material is cost-effective, and can be pre-formed into useful shapes, but its compatibility with solderable conductors is poor, and the finish will ‘crack up’ under strain. Also, the glass is a relatively poor thermal conductor.

For similar reasons, glasses are little used as constructional materials in thermally-sensitive areas, but ceramics are increasingly popular, especially for high-reliability and high-dissipation applications. The original material used for thick film applications, aluminium oxide (‘alumina’) of 96 or 99% purity has a thermal performance substantially better than FR-4 board. Although rigid and brittle, it is strong in compression and well-matched to silicon. It can be used as a substitute for a printed circuit board, using ‘thick film’ metallization, and this can integrate high quality resistors and low value capacitors, as well as providing crossovers. More details on this if you search for "hybrid microcircuits", and in our topic paper Thick film technology.

Unfortunately, the printed-and-fired conductors used by thick film technology have limited current-carrying capability and high resistance compared with bulk copper. One way of combining the heat-sinking properties of alumina with bulk copper is to use ‘Direct Bonded Copper’; DBC is produced by enabling the reaction between copper and alumina that takes place just below the melting point of copper, creating a permanent bond. The copper-clad alumina is then etched to create a pattern. DBC is very useful for making assemblies of power semiconductors.

Alumina has been used as the basis of thick film hybrids for 40 years, and efforts have been made to find materials that have better thermal performance. Much early work was done on beryllia (beryllium oxide) –its thermal conductivity is extremely high in comparison with other ceramics, particularly below 300ºC (300W/m·K at room temperature). Mechanically less strong than alumina, BeO has only 76% of its density, and is an electrical insulator. Transparent to both microwaves and x-rays, it finds use in microwave communication systems, in x-ray windows and in high-power laser tubes. Beryllia is still used in the form of inserts in a number of power devices. However, it presents a considerable hazard both for manufacture and end-of-life, because fine particles of beryllia dust, such as those produced when the substrate is broken, are extremely toxic, causing respiratory diseases. In consequence, all beryllia-containing components are required to be marked with warnings of the hazard.

Offering no toxic problems, but an improvement on alumina, are a group of materials generally referred to as ‘aluminium nitride’. Because aluminium is very reactive, most will contain at least some proportion of the oxide, so this is a relatively variable material in terms of its thermal characteristics. Particular problems have been associated with metallising aluminium nitride, and it has taken some time to produce thick film materials that can be air-fired and produce reliable interconnections.

Note that, for the purpose of making packages, most ceramic materials need to be joined. When in the green state, the glass-formers within the ceramic (the small percentage that is not the oxide) can be pressed into service to form a glass-rich joint; for fired ceramic, glass can be used, but the ceramic is frequently ‘metallised’ and the parts soldered or brazed together. The first metallising used on alumina was ‘moly-manganese’, a mixed-metal system firing at 1,500°C in a reducing environment (wet hydrogen). This process produces a highly reliable and strongly adherent metal film that can be brazed, for example with Ag72Cu28 eutectic alloy. Alternatively, it can be used without plating as the internal electrode within a multilayer ceramic structure.

Detailed consideration of the many options available lies outside this course, but those interested in the topics should browse for information on multilayer ceramics, hybrid microcircuits and low-temperature co-firing ceramics. Or visit the IMAPS and DuPont Electronic Technologies web sites.

Alumina is used not only as a structural ceramic, but also as a filler for adhesives and rubbers to improve their thermal performance. Boron nitride is another ceramic material, with properties even better than those of aluminium nitride, usually found only in its powdered form as a filler.

[ back to top ]


Other materials

Increasingly, designers trying to get better thermal performance are looking at more exotic materials, especially a range of metal matrix composites (MMCs). One group of these combines metals in ways other than straight alloying, and includes systems made using powder metallurgy to add molybdenum or tungsten to copper to reduce the TCE whilst retaining much of its thermal performance. A second type of MMC combines metals and ceramics, adding as much as 75% by volume of silicon carbide into matrices such as aluminium and copper. The principal aim is to enhance the performance, but an important driver in the case of the ceramic materials is their much lighter weight.

Reinforcements come as continuous fibres, whiskers and particulates; combined in an aluminium matrix, the resulting material has increased stiffness and wear resistance and in some cases strength and fatigue resistance also. Importantly for many applications, the CTE of the base metal is reduced by adding the reinforcement, although the composite retains the high thermal conductivity and low density inherent in the aluminium alloy. The first reinforced aluminium was used in the Space Shuttle, using fibre reinforcement, but the emphasis shifted towards particulate-reinforced materials, aiming at lower-cost, high-volume products for automotive and commercial aerospace applications.

More background information

Two useful websites are MMC-Assess (http://mmc-assess.tuwien.ac.at – funded by the European Community in Brite-Euram III, this group aims to increase market acceptance by collecting and evaluating information on MMCs. For a wider range of applications, see their Applications section. The Aluminum MMC Consortium (www.almmc.com) aims to develop manufacturing process technology and awareness for aluminium matrix composite products. Its ‘Roadmap’ for MMCs gives an interesting insight into the way that progress in this area is being targeted at cost reduction.

 

Of the ceramic materials that can be used as reinforcements, silicon carbide and aluminium oxide are the two that have seen the greatest use, because they are relatively light and cheap and give the best potential for improving material properties.

The glossary of MMC terms prepared by Mortonson and colleagues has much useful information on the different process methods. Manufacturing of these materials typically starts with surface treatment of the reinforcement, after which the composite is created by infiltration, liquid phase sintering or hot pressing. Then the part is shaped by casting or forging before final machining and plating.

An indication of typical MMC issues

The paper by Sundberg et al Copper-silicon carbide for IGBT thermal management (Advancing Microelectronics, November/December 2004, pp 8-12) discusses the use of metal matrix composite (MMC) materials for IGBT power modules. Copper is a highly conductive base-plate material, but has a high CTE; strain caused by expansion mismatch leads to fatigue failure of the solder, component delamination and cracking that ultimately lead to device failure. Of course, it is always possible to make a joint more reliable by making it thicker, or by providing compliant interlayers, but doing this increases the thermal impedance.

Using dispersions of silicon carbide (SiC) in aluminium, the CTE can be adjusted between 7.5 and 12ppm/°C by controlling the percentage of SiC, since the SiC component has a smaller CTE than the matrix. IGBT modules built with this material typically have a CTE of 8.4 ppm/°C between 30–150°C. It has been found that using such a base-plate more than doubles the mean time to fail compared with equivalents with a copper base-plate. However, the typical thermal conductivity of this material is 190W/mK, only half that of copper.

A copper/SiC MMC provides a CTE value that is similarly adjustable, with a value of thermal conductivity greater than 300W/mK. The material has the benefit that it can be brazed with copper-silver eutectic at around 800°C, allowing it to be used to fabricate more complex assemblies. Unfortunately, SiC reacts chemically with copper, leading to degraded performance, unless the powder is treated with a barrier coating – Sundberg and his collaborators found that titanium nitride was a good material choice.

 

Carbon is a cheap basic material that has good thermal conductivity, although most forms also conduct electricity. As carbon fibre, is probably most familiar as a material for strengthening light-weight composites used in sports equipment. However, carbon fibre can be a useful thermal component, either as a strengthening and thermally-enhancing matrix for a polymer, or as a high-conductivity carbon-carbon composite.

Graphite is a form of carbon that is becoming increasingly important for thermal management. Although polycrystalline graphite has unexceptional properties, its normal layered structure is strongly anisotropic: in its basal plane graphite is strong, with extremely high conductivity and very low thermal expansion; across its basal plane it is weak, with low conductivity and higher CTE.

Suggested reading

Julian Norley The role of natural graphite in electronics cooling for an explanation of the structure of graphite and an insight into the range of thermal conductivities available.

Mark Ryals Graphite fiber reinforced Al and Cu alloys for thermal management applications for the way that fibres can be used as reinforcement.

Development of high thermal conductivity, low density graphite foam for information on an extremely lightweight material with promising characteristics.

 

 

Supplementary information

For more detailed technical information on graphite, a useful resource is Poco Graphite’s Properties and characteristics of graphite for the semiconductor industry (1.74MB).

 

For the ultimate in thermal conductivity, it is the allotrope of carbon that we call diamond that gives us superb performance, although at a cost! And diamond is also an excellent insulator – the reason for this uncharacteristic performance is explained at this links. Whilst single-crystal diamond is inappropriately expensive for electronic applications, small heat spreaders made from artificial diamond have been successfully used for high-power critical applications, primarily in fibre optics.

Finally, when considering a constructional material that can be used as a mounting board for semiconductors, we should not forget the possibility of integrating components on silicon itself; the TCE match is perfect, though there may be some strain due to temperature differences, and the thermal performance is good. However, the typical method for such wafer-scale integration uses flip-chip attach of the daughter chips, so the first link in the thermal chain is the comparatively poorly performing set of flip-chip balls.

[ back to top ]


Thermal contact resistance

In systems where two solid materials are in contact, the thermal conductivity of the junction depends primarily on the area of contact. However, as you will probably know in relation to connectors, relays and switches, the effective area of contact is much smaller than the physical area, even if the materials are nominally flat and smooth. This is because the surfaces are rough on a microscopic scale, so that only the peaks touch and spots of intimate contact are interspersed with air-filled gaps as indicated in Figure 2.

Figure 2: Temperature drop due to thermal contact resistance

Temperature drop due to thermal contact resistance

With electrical contacts this results in a ‘contact resistance’, and there is an exact thermal analogue in the form of a temperature drop across the interface. The ‘thermal contact resistance’ is defined as:

$R_{t,c}  = \frac{{T_A  - T_B }}{{J_x }}$

To see how this temperature drop can be measured as part of a simple experiment, see Prediction and mitigation of thermal contact resistance

The heat transfer that takes place is the sume of the conduction across the area actually in contact and of the conduction and/or radiation across the gaps, there being little opportunity for convection. For conforming rough surfaces, the surface-to-surface contact resistance may therefore be viewed as two resistances in parallel:

$\frac{1}{{R_j }} = \frac{1}{{R_c }} + \frac{1}{{R_g }}$

where Rj is the total joint contact resistance, Rc is the resistance through the solid contact between mating surfaces, and Rg the gap resistance through the intermediate material. As the contact area is typically a small percentage, especially if the surfaces are rough, the gaps play a major part in increasing the contact resistance.

At low contact pressure, only a very small portion of the surface area is in intimate contact, so Rg is the major contributor to the resistance and the effects of interface resistance are large. As the applied pressure is increased, a greater proportion will be in contact because the surfaces undergo elastic deformation, depending on their relative hardness.

Supplementary information

Basic information on the microstructure of surfaces at this link.

 

A number of theories have been developed for predicting thermal contact resistance, for example, Rg can be estimated by a complex equation that involves the conductivities of all the materials, the hardness of the softer of the two contact materials, the surface roughness, and the pressure. But the most reliable results are experimental, as in Table 4: note that there is a range of values for each condition, and that increasing the pressure applied substantially reduces the thermal resistance.

contact pressure
100 kN/m2
10,000 kN/m2
stainless steel
6–25
0.7–4.0
copper
1–10
0.1–0.5
magnesium
1.5–3.5
0.2–0.4
aluminium
1.5–5.0
0.2–0.4
Source: Incropera and DeWitt
Table 4: Thermal contact resistance for metallic interfaces under vacuum with different contact pressures

There are two mechanical ways of reducing the contact resistance for the typical case of a solid whose thermal conductivity is higher than that of the gas or other material at the interface. These are by increasing the joint pressure and/or by reducing the roughness and improving the flatness of the mating surfaces.

Lee and his colleagues1 developed an approximate solution for predicting the thermal resistance of bolted joints between two square plates of the same material but of different thicknesses. The analysis was complicated by the large number of factors that determine heat flow, such as the surface characteristics, the nature of the materials, the applied load and the system geometry. Their analysis showed clearly that, despite the larger total area of the plates, the main thermal flow is through the high pressure contact area surrounding the tightened bolt. The radius of this effective contact area is a function of the materials and the pressure applied.

1 Analytical modelling of thermal resistance in bolted joints (HTD-Vol. 263, Enhanced cooling techniques for electronics applications ASME 1993) Lee, Song, Moran and Yovanovich

 

The most useful general point resulting from this work, supported by experiment, is that thermal conductance is considerably improved by using a washer to spread the load, especially when the plates are of unequal thickness. Although the improvement shown in Figure 3 is relatively small, this can still be helpful in reducing system temperature.

Figure 3: Thermal resistance vs washer radius for copper plates of unequal thickness

Thermal resistance vs washer radius for copper plates of unequal thickness

Source: Lee, Song, Moran and Yovanovich op.cit.

More generally, the observation that even perfectly flat surfaces need to be held in contact reinforces the need to consider interface thermal resistance and either pay great attention to applying even pressure or use an appropriate thermal interface material.

Quote

A third way of reducing the contact resistance is to use an interfacial material that has a high thermal conductivity. Koolance make the valid comment that “No matter how flawless a surface may appear, it is highly irregular on a microscopic level. In fact, a CPU cold plate or heat sink may only touch at 0.1% of their total surface areas. Because the rest of the surfaces are separated by air, many high-heat sources would fail without a thermal interface material”.

 

Further reading

M. M. Yovanovich, J. R. Culham and P. Teertstra Calculating interface resistance.

 

[ back to top ]


Solders

Not all metals are used in prefabricated bulk form. For example, soft solders are frequently used for die attach in power semiconductors, and a gold-tin eutectic solder was the material used for all silicon die attach before the adoption of conductive epoxies. Whilst intimate contact is made because solders wet the surfaces, one unfortunate side-effect of soldering, especially when flux is used, is the formation of voids due to gas entrapment. From the point of view of the integrity of the joints, having a percentage of small voids may even improve the life of the joint, and the IPC specifications allow 20% voids by volume in a BGA ball. However, from the thermal conductivity perspective, such voiding immediately increases the thermal resistance, and may lead to local hot-spots under the die. This is one reason why vacuum-assisted soldering is sometimes used when assembling power semiconductors.

Supplementary information

Die attachment using gold-tin eutectic and soft solders are topics in our paper on Die bonding materials which will also be referred to in the next section on adhesives, and again in Unit 10 .

 

Soldering and brazing are also used in making packages for modules and components and in fabricating items such as heat exchangers, especially those made in copper. These too need void-free joints, and should be as thin as practicable in order to minimise their thermal resistance. However, whilst solders have poorer thermal conductivity than copper or aluminium, they are considerably better than filled resins, and their potential use should always be borne in mind.

Of course, soldering normally needs solder-compatible surfaces, so materials such as ceramics need to be metallised, which adds to the cost and introduces further complications. Redd and Smith, in their paper Joining of packaging in thermal management materials using active solders (Advancing Microelectronics, Nov/Dec 2004, pp12–14), comment on the problem faced by designers where it is necessary to join dissimilar or hard-to-bond materials. As they point out, soldering, brazing and adhesives do not work with some combinations of materials, and often the resulting thermal performance is poor.

Although similar approaches have previously been used for high-temperature brazes, Redd and Smith describe ‘active solders’ as a ‘new class of materials’ that can wet and adhere to a wide variety of metals, ceramics, glasses, carbon compounds and ceramic composites. This is achieved by combining a base either of tin/silver or zinc/aluminium with additional elements to create ‘intermetallic micro-domains’. When activated during the joining process, these added elements react or adhere to surfaces to create a joint at 250°–400°C without the use of lead, flux or plating, and without the need to remove oxide layers before joining.

The additives are based on adding titanium, hafnium and zirconium with lanthanide elements such as cerium or lanthanum. Whilst the alloys are made more reactive, the solder fillers become less capable of flowing into joints, because their capillarity is reduced. In order for bonding to take place, the active solder alloy needs to be ‘pre-placed’ in contact with the joint and then ‘activated’ by disturbing the molten solder layer so as to break down the thin layer of tin oxides that form on the surface. Once that is done, by brushing, spraying or ultrasonic action (similar to plastic welding), then the solder will wet and bond to films present on the surface to be bonded. So the solder will bond directly to aluminium oxide on aluminium, to silicon dioxide on silicon, and chromium sesquioxide on stainless steel. The joints are reported to have high thermal conductivities: 48W/mK for tin-silver systems and 90W/mK for zinc-aluminium-silver.

Altogether an interesting technology for bonding dissimilar materials, but one feels it will have greatest potential when used with difficult-to-bond materials of compatible CTE. Possibly a development worth watching out for . . .

[ back to top ]


Thermal management materials

In this category we have brought together a wide range of materials other than solders that are used between structural components. These include both structural adhesives and a number of materials intended to improve thermal connection between parts that are held together mechanically by external means.

Adhesives

Most transistors and integrated circuits use a conductive adhesive to attach the die and lead frame. Traditionally epoxy-based materials, these are heavily loaded with silver particles to make them conductive. Since their introduction in the early 1970s, these materials have improved considerably in terms of the purity of the base resin, the control of flow characteristics and bond line thickness and in the particle size and consistency of the filler. As with solder paste, the upper limit on metal content is set by the rheology of the system; with the small particles used for silver-loaded epoxy, this is typically no more than 60%. Although the silver particles are in contact with each other, which is how the material is able to conduct, contact is limited to a relatively few points, so that the resulting cured material has only of the order of 1% to 5% of the conductivity of pure silver. Fortunately, because the bond lines are thin, this is not a major limitation, except for high-power devices where solders are preferred.

Supplementary information

For more information on die attachment using resins see our paper on Die bonding materials.

 

In the same way that an adhesive can be loaded with silver to create a resin that conducts both heat and electricity, it can be loaded with fine particles of thermally conducting but insulating materials such as alumina or boron nitride.

These adhesives are liquid during the application phase, and able to wet to rough surfaces. Of course, with semiconductor assemblies, the surfaces are usually quite smooth, especially the underside of the silicon, and it is only when we come to use these materials with surfaces that are rough and not quite flat that we really see the benefit of having an intermediate layer.

However, whilst die attach resins are normally used in liquid form, as the surfaces become larger it becomes more difficult to apply an even adhesive coat. For such purposes resins are supplied impregnated onto a support such as glass-fibre cloth and taken to B-stage cure. Although cut-film preforms are fairly easy to use, the assembly needs to be clamped during curing in order to get the best thermal performance, and the limited flow of the resin component means that intimate contact is not achieved. Also, the larger the area of the interface, the greater chance there will be of trapping air when applying the material to the surfaces, and this will increase the interface resistance.

Note that both conductive and non-conductive adhesives are intended to be permanent. Although it is sometimes possible to break the bond by applying simultaneous heat and force, this destroys the adhesive, which needs to be cleaned off and replaced.

Quote

“A great deal of technology development has gone into the engineering of interfaces with low thermal resistance, but this continues to be an important area which is frequently the source of problems. The design of thermal interfaces is also constrained by the need to accommodate mismatch strains between different parts of the assembly which arise either as a result of thermal expansion mismatch combined with thermal excursions during processing or in service.

“The size of the problem increases with the physical dimensions of the interface, the degree of the mismatch of the CTE of the different materials used in the package and the range of temperature change that occurs during operation. The issues are most acute in the design of power control systems such as IGBTs or thyristor devices”.

EPPIC An Introduction to Thermal Management

 

As well as the difficulty with repair, there is also the problem of potential TCE mismatch, as our quotation stresses. For die bonds and mounting to a printed circuit board, there are few options other than solder or adhesive bonding; however, for most heat sink applications, designers prefer to use methods where component and heat sink are held together by clips, bolts, rivets or otherwise, so can be easily separated. The question then comes as to what material to use as a gap filler.

Activity

Before reading further, take a look at Daniel Blazej’s article on Thermal Interface Materials and skim the section on interface materials at the ThermaFlo web site.

 

Thermal greases

Any material that can fill the gap between contacting surfaces, and whose thermal conductivity exceeds that of air, will lower the contact resistance (Figure 4).

Figure 4: Heat flow between two surfaces in contact, showing schematically the effect of an interface material

Heat flow between two surfaces in contact, showing schematically the effect of an interface material

 

Two types of material that are well-suited are soft metals and thermal greases. Metals such as indium, lead, tin and silver may be inserted as a thin foils or applied as a thin coating to one of the pair of materials. A more intimate contact is provided by a thermal grease, and silicone-based greases give a substantial improvement in performance. Some representative figures for thermal interface resistance are given in Table 5.

interface
medium
thermal resistance
(×10−4m2·K/W)
conditions
silicon die to aluminium air
0.3–0.6
27–500kN/m2
aluminium to aluminium indium foil
~0.07
~3,500kN/m2
aluminium to aluminium lead coating
0.01–0.1
aluminium to aluminium silicone grease
~0.07
~100kN/m2
~0.04
~3,500kN/m2
silicon die to aluminium 20µm epoxy
0.2–0.9
Source: Incropera and DeWitt
Table 5: Thermal resistance of selected solid/solid interfaces

 

A note about units

Thermal resistance is expressed in terms of degrees of temperature rise for each unit of power (K/W). However, the thermal resistance of a junction varies inversely as its area, so comparison between materials becomes easier if we normalise our data into units of K per watt per square meter. Unfortunately, the self-explanatory form K/W/m2 is not allowed, and the formal unit is the rather confusing m2·K/W.

Worse is to come . . . Because there aren’t too many metre-square heat-sinks(!), many tables quote units of K/W/(cm)2, which are 104 times larger. This explains the choice of units in Table 5, to make comparison easier.

Thermal calculations generally need extreme care in using consistent units, and in being sure about any data values you import.

 

Conductive greases were among the first thermal interface products, and their main advantage has been the ability to become more liquid under normal operating temperatures, so that they can fill very small voids and eliminate air spaces.

Traces of “white paste” can often be found around power transistors. This is a thermal grease, usually based on high-molecular-weight silicones, with a zinc oxide powder filler. Unfortunately, silicones in particular have a tendency to migrate; not only is this “death to soldering”, but the loss of fluid reduces the effectiveness of the contact, leaving a powdery deposit of filler.

In general, joints that use grease to improve thermal performance tend to have stability problems, especially under power cycling conditions which ‘pump out’ the grease in the interface, so that this eventually dries out. This can also happen under high-temperature operation due to migration of the polymer components.

Both pump-out and phase separation mechanisms are strongly dependent on temperature, approximately doubling for every 10°C increase in average operating temperature. An example quoted in the EPPIC paper An Introduction to Thermal Management is a power cycling of an assembly between 0°C and 100°C for 7,500 cycles, which resulted in a 4–6 fold increase in thermal resistance, compared to a negligible increase for 0–80°C exposure over 2,500 power cycles.

Insulating materials

Thermal greases are very thin, and their insulating properties cannot be relied on. In order to provide an insulation layer, a material traditionally used was mica. A natural laminated mineral mined in India, mica provides a very high dielectric strength with reasonable thermal properties which can be separated into extremely thin sheets. However, being a natural material and very liable to flake, mica has fallen out of favour.

In consequence, a number of different film materials are available, often pre-cut to the required dimensions, and sometimes with an adhesive coating to aid assembly. The filler may be graphite (the cheapest grey/black pads), alumina or boron nitride, and the resin used to make the pad is at least slightly compliant, to take up the non-flatness of the surfaces to be joined. In some cases, the materials are substantially compliant, allowing a single heat sink to make contact with integrated circuits of different heights. See Bergquist S-class for an example of such a product.

Yet, however compliant the material, there will not be the intimacy of contact between a pad and the surfaces to be joined that exists with a liquid system such as a grease. When he investigated the thermal properties of materials used as underfills, Mok found that films filled with boron nitride and alumina produced higher resistance that zinc oxide grease, because the contact pressure was insufficient to force out the trapped air at the interface.

In their paper Thermal performance of interface material in microelectronics packaging applications (1999 IEPS Conference) Early, Lee and Pellilo of Aavid show how thermal resistance can be a strong function of contact pressure (Table 6). Notice that one of the interface materials actually performed worse than using bare surfaces, but a synthetic grease worked well over the entire range.

thermal resistance (K·cm2/W)
contact pressure (kN/m2)
interface material thickness (mm) data sheet 56 251 496 740 985 1,222 2,445
graphite and oil sheet 0.127 1.10 4.00 3.74 2.45 1.87 1.81 1.61 1.29
silicone sheet 0.153 4.84 11.68 11.10 10.32 9.03 8.06 7.61 5.16
fluoroether oil sheet 0.178 n/a 7.10 6.32 5.42 4.58 4.32 4.19 2.84
synthetic grease n/a 1.23 1.77 1.33 1.29 1.25 1.25 1.25 1.16
bare calorimeter surfaces n/a n/a 11.23 10.13 7.94 6.97 5.74 5.35 3.42
metricated data from Early, Lee and Pellilo, op. cit.
Table 6: Thermal performance of interface material at different contact pressures

Early and his colleagues warn that discrepancies between applications and manufacturer’s data can lead to the wrong material being selected and therefore to device failure caused by thermal resistance being higher than expected. The root cause for this is that ASTM test standard D5470-93 uses a high contact pressure (300psi = 2,068kN/m2) to reduce the effect of interface resistance generated by trapped air. But this then becomes the basis of the published thermal performance characteristics, whereas in a typical application the contact pressure may be only 10–50psi (69–345kN/m2), and to apply higher pressure might damage the package.

Phase change films

One way in which this limitation can be overcome is to use a ‘phase change material’ (PCM), where the material changes into a gel above a specifiable working temperature. This allows much more intimate contact with the surfaces to be joined, reducing the thermal resistance.

The PCMs reported by Zhang2 are based on alkyl methyl siloxane (AMS) waxes. Solid at room temperature, these waxes melt at normal operating temperatures (about 50°C) and assembly pressures, so that the surface is completely wetted, resulting in optimal heat transfer. The viscosity of the molten material is high, which helps to eliminate pump-out. Other materials used are manufactured from silicon-organic polymers, carefully tailored to give the required melting temperature and viscosity.

The waxes are usually filled with thermally-conductive particles such as silver, alumina or zinc oxide and are not cured. A typical figure for bulk thermal conductivity is 4–5W/mK at 50°C and above, and rather higher (7–8W/mK) at low temperatures.

PCMs can be supplied as pads with a 35µm thick aluminium mesh reinforcement to give a total pad thickness of around 150µm. The embedded mesh improves bulk conductivity by around 2W/mK, but this gain is offset by the thicker bond line required, so that overall the performance of supported and unsupported material is similar.

Supplementary information

Zhang et al Silicone phase change materials for thermal interface applications in Global SMT & Packaging, November 2002 (PDF file, 232kB). See also his CoolingZone article Introduction to Phase Change Thermal Interface Materials for an indication of the rheology of these materials.

 

Graphite

If you have used any graphite-loaded penetrating oils, or even a soft pencil, you will know that graphite is a form of carbon that has interesting properties. And these extend to the use of graphite materials for thermal management purposes, as you will have seen in an earlier section. Some thermal greases use graphite, rubbers are frequently graphite-impregnated. For more state-of-the-art applications, graphite foams are light and compressible, yet have a very high thermal conductivity. And the open-cell nature of the foam makes it possible to integrate this good conductor with a thermal management system that involves fluid transfer, so that heat can be taken out of the graphite foam and dissipated elsewhere in the system.

Overview

Figure 5 presents a compilation of data for dry interfaces, elastomeric pads, phase change materials and films, greases, metal-filled gels and eutectic attach. There is an embarrassingly long list of potential materials to select from, and the designer has to be careful to make appropriate choices, having regard to the application, the thermal challenge and the budget . . .

Figure 5: Thermal resistance data for different interface materials

Thermal resistance data for different interface materials

EPPIC An Introduction to Thermal Management

Application methods

When a thermal management material such as grease, film, foam or rubber is used between the heat dissipating component and heat sink, the parts need to be held together. In practice, you will need to think about how this is to be done, and the costs of assembling the necessary hardware. As always, keeping it simple keeps the cost down, and there are a number of self-jigging solutions on the market that allow clip, heat sink and phase change film (for example) to be applied as a single activity. The cost of thermal management can be higher than you anticipate!

Quote

“The key criteria for thermal interface materials include high bulk thermal conductivity, low interfacial resistance and long-term stability, without the use of a heavy bond line. Ease of application is also a critical consideration, as added manufacturing steps or difficult processing contribute to higher costs. In many applications, the ability to perform rework and repair operations can be an additional influence”.

Zhang et al, op. cit

 

We have tried in this short section to indicate some of the many opportunities there are for thermal management and the wide variety of materials available. We hope that you will be encouraged to look at alternatives, bearing in mind not only the cost of the material, and the cost of applying it, but also thinking about the ways in which the materials might change with time. Will the foam or rubber degrade with time, for example becoming more brittle or undergoing a permanent set, so as to reduce the quality of the contact between surfaces. Another factor to take into account is the quality of the demountable joint if it is separated after a period of use and then remade.

[ back to top ]


Resources for this Unit

Each of these lists is in the order in which the material is referenced in the Unit text. However, note that links to SAQ answers are not included!

Needed for activities

Recommended supplementary material

Optional links and information

[ back to top ]