Applying TIMs Correctly: Thickness, Pressure and Pump-Out
Short answer: a thermal interface material (TIM) should be as thin as the two surfaces allow — for paste on a flat 20×20 mm die, a 50 µm bond line at 5 W/m·K adds only about 0.025 °C/W, while the same paste at 200 µm thick adds 0.1 °C/W and silently overheats the junction. Mounting pressure for a standard clip or screw layout should land in the roughly 10–50 psi (70–350 kPa) range, applied evenly. Use just enough paste to cover the contact area with no air gaps, cure or clamp per the datasheet, and design against pump-out: the cyclic squeezing of paste out from under the die as the package expands and contracts with temperature. Pump-out, not raw conductivity, is the most common field failure of an otherwise correct thermal design.
Buyers spend serious money selecting a high-conductivity TIM and then undo it in five minutes of assembly. The interface between component and heat sink is where the thermal budget is won or lost, because even polished metal surfaces touch at only a small fraction of their apparent area — the microscopic valleys between contact points are filled with air at roughly 0.026 W/m·K, an excellent insulator. A TIM's only job is to replace that air with something better. Everything about correct application follows from that job: fill the voids, keep the layer thin, keep it in place, and keep it there for the life of the product.
Bond Line Thickness: Thin Is the Whole Point
Thermal resistance through the TIM layer is thickness divided by conductivity divided by area, R = t ÷ (k × A). Since conductivity is fixed by the material you chose, thickness is the only lever the assembler controls, and it is a powerful one. Take a 20×20 mm die (A = 4 cm²) with a paste of 5 W/m·K: at a 50 µm bond line the interface adds 0.025 °C/W; at 200 µm it adds 0.1 °C/W — four times the temperature rise for the same component and sink, typically 5–15 °C extra junction temperature. That difference decides between a reliable design and a returned product.
The correct thickness is therefore the minimum that still fills the roughness and flatness gaps between the two surfaces. Typical flat surfaces need 25–100 µm of paste; warped or poorly machined bases need more, which is exactly the wrong way to fix a flatness problem — fix the flatness instead. More paste is never a safety margin: excess paste squeezes out, makes a mess, and because paste conducts roughly forty times worse than aluminum, every extra tenth of a millimeter is a thermal penalty you paid for with a thicker layer. The same logic applies to choosing a material: phase-change materials and solders achieve thinner bond lines than greases, which is part of why they outperform despite similar datasheet conductivity.
| TIM type | Typical conductivity | Typical bond line | Typical mounting pressure | Notes |
|---|---|---|---|---|
| Silicone grease | 1–8 W/m·K | 25–100 µm | 10–50 psi | Cheap, pump-out risk, needs retainer |
| Ceramic-filled paste | 3–8 W/m·K | 25–75 µm | 10–50 psi | Common, electrically insulating grades |
| Phase-change material | 3–8 W/m·K | 25–75 µm after melt | 10–50 psi | Flows at first power-up, low pump-out |
| Thermal pad | 1–6 W/m·K | 0.5–3 mm | 5–30 psi (light) | Fills big gaps, higher resistance |
| Gap filler | 1–5 W/m·K | 0.5–5 mm | Light, no hard clamp | For uneven or tall gaps |
| Cured adhesive TIM | 1–3 W/m·K | 25–100 µm | Cure under light clamp | Bonds parts, rework hard |
Conductivity ranges are typical datasheet values across commercial grades and the pressure figures are general assembly guidance, not a spec for any particular product. The pattern to notice: materials that achieve thin bond lines (paste, phase change) beat thick gap-filling materials on resistance despite similar conductivity numbers, because thickness matters as much as the material itself.
Mounting Pressure: Even and Within the Window
The heat sink must press the TIM to its working thickness and hold it there, and the pressure has a window with real penalties on both sides. Too little pressure leaves the TIM thick and voids present, especially if the base is slightly bowed; the interface runs hot and the paste may never wet the whole surface. Too much pressure can squeeze paste out from under the die until the gap is metal-to-metal at the edges, distort a thin base plate, or over-compress a pad past its working range. The practical target for screw or clip layouts is roughly 10–50 psi (70–350 kPa) at the component, which for a typical module means spring clips or screws torqued to the module maker's spec — commonly in the 0.4–0.6 N·m range for M3-class fasteners, always per the datasheet.
Evenness matters as much as the average value. Four screws torqued in the wrong order can bow the assembly and create a gap under the die center that no TIM can fill. The assembly habits that protect you are cheap: apply screws in a star or criss-cross pattern in two or three passes, use a torque driver rather than feel, and check that the sink base does not rock on the component before final torque. If the base rocks, the flatness or the screw layout is wrong, and no amount of paste fixes it — this is one of the first checks in any heat sink troubleshooting session.
Coverage Patterns and Dispensing Volume
Coverage is a simple goal with practical subtleties: the entire apparent contact area must end up wet, with no trapped air pockets, and without so much excess that paste floods the board. For a small die, a single dot of paste at the center works because clamping spreads it; for a long rectangular die or module, a line or an X pattern spreads more evenly and avoids trapping air at the far ends. The correct volume is approximately the contact area times the target bond line, plus a small allowance for squeeze-out; dispensing by weight or by programmed volume beats eyeballing, because eyeballing drifts with every batch.
| Application fault | What it does | The fix |
|---|---|---|
| Too much paste | Thick layer, thermal penalty, board contamination | Dispense area × target thickness, no more |
| Too little paste | Corner voids run hot, uneven temperature | Use X or line pattern for rectangular parts |
| Uneven screw torque | Bowed base, gap under die center | Star pattern, two passes, torque driver |
| No cure or clamp time | Adhesive TIM never bonds; pad creeps | Follow datasheet cure time and pressure |
| Pad compressed too far | Over-compression, spring-back loss over time | Choose pad thickness for the real gap |
The failure column reads like a checklist, because every one of these is a recurring finding when thermally failed products come back to the factory. Note that pads and gap fillers invert some rules: a pad must be slightly thicker than the gap so it compresses 10–30% to its working range, and it must never be forced into a gap far smaller than its uncompressed thickness. Read the datasheet's compression curve before choosing pad thickness, not after the first thermal test fails.
Pump-Out, Dry-Out and Keeping the TIM in Place for Years
Pump-out is the field killer. Every power cycle makes the component and the sink expand and contract at different rates — the die and its lid move relative to the cold plate or heat sink base — and that cyclic motion literally pumps soft grease out from under the component, a few microns per cycle, until the interface is starved and the junction runs away. It is a fatigue mechanism, which is why it appears after months in service rather than in factory test. Three countermeasures are standard: choose a material resistant to it (phase-change materials and cured adhesives do not pump out like greases do), keep the thermal expansion difference small by using a mounting method that does not over-constrain the parts, and verify with thermal cycling rather than a single hot run.
Related failure modes share the same assembly roots. Dry-out happens when the carrier oil evaporates or bleeds out, accelerated by sustained high temperature — silicone greases can bleed oil that migrates across the board. Voids appear when paste is applied in blobs that never merge during clamping. Over time, a pad can take a compression set and lose its spring-back, thinning the interface it was meant to fill. None of these are solved by buying a more expensive paste; all of them are solved by correct thickness, correct pressure, a pump-out-resistant material choice for cyclic loads, and a thermal-cycle test before production. When a design passes a 1,000-cycle thermal test with the junction stable, the interface is done; when it fails, the fix is usually in the assembly spec, not the material datasheet.
Frequently Asked Questions
Q: How much thermal paste should I apply to a CPU or power module?
A: Just enough to cover the contact area at the target bond line, typically 25–100 µm: a small center dot for a small die, an X or line pattern for rectangular parts. Excess paste adds thermal resistance instead of removing it, because the paste conducts far worse than the metal it separates.
Q: What mounting pressure does a heat sink need for good TIM performance?
A: Roughly 10–50 psi (70–350 kPa) at the component is the typical working window for paste and phase-change materials. Apply it evenly — star-pattern screw torque in two passes with a torque driver — because uneven pressure bows the base and opens a gap no TIM can fill.
Q: What is pump-out and how do I prevent it?
A: Pump-out is the cyclic squeezing of soft grease out from under a component as the package and heat sink expand and contract at different rates through power cycles. Prevent it by choosing phase-change or cured materials for cyclic loads, keeping mounting pressure within spec, and validating with thermal-cycle testing.
Q: Is a thicker thermal pad better for filling a large gap?
A: Only up to a point. A pad must be slightly thicker than the gap so it compresses roughly 10–30% to its working range; too thick a pad adds resistance, and forcing an over-thick pad into a small gap over-compresses it so it loses spring-back over time. Match pad thickness to the real gap with the datasheet compression curve.
Q: Why is my junction temperature higher than the datasheet predicts?
A: Check the interface before blaming the sink: bond line thickness, screw torque and sequence, coverage, and cure time are the usual culprits. Measure with the recommended torque and a correct paste volume, then retest — most "underperforming heat sinks" are actually underperforming TIM applications.
Related Resources
- Thermal interface materials guide — choosing between grease, pads, phase change and adhesives.
- Heat sink mounting methods — clips, screws and hardware that hold the interface at its working pressure.
- About BQUQ: an ISO9001 source factory in Dongguan machining heat sinks and validating thermal assemblies under one roof.
- Contact us: send your thermal requirements and drawings for a design review within 12 working hours.
Authored by the BQUQ Engineering Team. BQUQ is an ISO9001-certified source factory in Dongguan, China, running CNC machining, metal stamping, custom springs, heat sink and collet lines under one roof. Send drawings to sc@bquq.com or WhatsApp +86 13713157787 for a quote within 12 working hours. www.bquq.com
