Thermal Interface Material (TIM)
Theory: thermal resistance of the gap filler
Overview
A CPU and heat sink are both flat metal — why do we need grease between them?
They look flat, but the surfaces are rough on a fine scale. When pressed together, only a tiny fraction of the area actually touches; the rest is air gap. Air conducts about 0.026 W/(m·K), thousands of times less than metal. A thermal interface material (TIM) fills those gaps with something that conducts tens of times better than air. Grease itself conducts far worse than metal, so the goal is to fill, and a thicker layer does more harm than good.
The resistance formula
$R''$ is area-specific resistance (mm²·K/W), BLT the bond-line thickness, $k_{TIM}$ conductivity, $R''_c$ the contact resistance between material and surface, $Q$ the heat load and $A$ the area. It is the sum of a thickness-over-conductivity term and contact terms; the thinner the layer, the larger the contact share.
Main types
| Type | Features | Typical use |
|---|---|---|
| Grease | Goes on thin; pumps out or dries over time | CPUs, GPUs |
| Phase-change material | Solid at room temperature, softens and thins at operating temperature | Power modules |
| Gap pad | Absorbs large or uneven gaps; easy to handle | Multiple components on a board |
| Solder / metal | Very low resistance, but prone to cracking from thermal-expansion mismatch | High-power chips |
Discovering the gap in heat flow
The temperature step across two pressed metal surfaces was studied seriously in the mid-20th century for aircraft and rocket structures and for nuclear fuel rods, where the tiny gap between fuel and cladding strongly affected temperature. Formulas were developed to estimate contact resistance from surface roughness, pressure and hardness. As electronics power rose from the 1990s, the same problem appeared on top of CPUs and TIM development accelerated.
Worked examples
Example 1: temperature drop by material
Resistance and temperature drop for 100 W over 1 cm² (10 mm square). Contact resistance per face was assumed to be 1 mm²·K/W for grease and phase-change material, 5 for the pad and 0.5 for solder.
| Material | Conductivity | Thickness | Resistance | Temperature drop |
|---|---|---|---|---|
| Grease | 3 W/(m·K) | 25 μm | 10.3 mm²·K/W | 10.3 K |
| Phase-change | 5 W/(m·K) | 25 μm | 7.0 mm²·K/W | 7.0 K |
| Gap pad | 6 W/(m·K) | 0.5 mm | 93.3 mm²·K/W | 93.3 K |
| Indium solder | 86 W/(m·K) | 50 μm | 1.6 mm²·K/W | 1.6 K |
The pad conducts twice as well as the grease, yet the drop is nine times larger.
Because it is 20 times thicker. Resistance is thickness ÷ conductivity, so comparing conductivity alone is meaningless. Pads are handy for cooling several components of different heights together, but they don't suit a 100 W chip. Compare total resistance at the real thickness, including contact resistance, not datasheet conductivity.
Example 2: grease thickness
| Thickness | Resistance | Drop (100 W, 1 cm²) |
|---|---|---|
| 10 μm | 5.3 mm²·K/W | 5.3 K |
| 25 μm | 10.3 mm²·K/W | 10.3 K |
| 50 μm | 18.7 mm²·K/W | 18.7 K |
| 100 μm | 35.3 mm²·K/W | 35.3 K |
Grease of 3 W/(m·K) with only the thickness changed. Going from 25 μm to 100 μm multiplies the drop by about 3.4. Low clamping force, warped surfaces or large filler particles make the layer thicker. With a junction-to-ambient budget of 0.65 K/W (100 W, 65 K allowable rise), 25 μm grease takes about 16% of it and 100 μm takes about 54%.
Example 3: no TIM at all
Dry metal-to-metal contact conductance depends on roughness and pressure but is often about 2,000–10,000 W/(m²·K). That is 100–500 mm²·K/W, or a 100–500 K drop at 100 W over 1 cm² — unusable. A TIM cuts it by more than an order of magnitude.
Analysis and design practice
Modelling in CAE
- Apply as a resistance: meshing a layer tens of μm thick inflates element counts. It is usually applied as contact conductance $h_c = 1/R''$ between two surfaces.
- Measure thickness: assembled thickness depends on dispensed amount and clamping force. Check with cross-sections or gauges.
- Watch spreading: with a small hot spot, spreading resistance in the heat spreader grows alongside the TIM resistance.
Planning for degradation
- Thermal cycling makes chip and heat sink expand and contract, pumping grease out. Evaluate resistance after temperature-cycling tests.
- Long exposure to high temperature dries grease out and raises resistance. Check margin with end-of-life values.
- Solder develops cracks and voids from expansion mismatch. Check by cross-section or acoustic inspection.
“More grease must cool better”
A perennial PC-building debate is how much grease to apply. More feels safer, but tightening the heat sink squeezes out most of the excess, so within a sensible range the difference is small. The real problems are the opposite: weak clamping that leaves a thick layer, or too little grease that leaves air gaps. In mass production the dispensed amount is controlled by machine, and assemblies are sometimes X-rayed to check for voids.
Common mistakes
Mistakes and fixes
| Mistake | Effect | Fix |
|---|---|---|
| Choosing by conductivity alone | Thick materials run hot | Compare resistance including thickness |
| Ignoring contact resistance | Thin layers underestimated | Use measured resistance |
| Entering datasheet values directly | Predicts cooler than reality | Calibrate on the assembly |
| Designing on initial values only | Temperatures rise after years | Check margin with aged values |
| Ignoring hot-spot size | Spreading resistance missed | Model the actual heat distribution |
I'd like to try the resistance calculations myself.
Try the heat sink designer and finned heat sink calculator for the heat-sink side and the contact mechanics simulator for surface contact. The overall heat path is covered in thermal resistance, junction temperature and power module cooling.
Related Topics
Experience the theory firsthand with the interactive simulator for this field
All Simulators