A GPU runs cool on day one. Weeks later, its hotspot creeps upward. An inverter may show the same pattern after power cycling. Pump-out is an obvious suspect, but do not blame the grey stuff too fast.
PCM thermal pads usually resist pump-out better than many non-curing greases. They are not immune. Material, bond line, clamp, surface shape, temperature, and vibration act as one system. The useful question is: will this PCM keep coverage for the required life?
Pump-Out Is an Interface Failure, Not Just Material at the Edge
Pump-out is the gradual movement of a thermal interface material away from the active contact area. Repeated expansion, warpage, and pressure change may move only a tiny amount per cycle. Eventually, the center can become starved while the edges collect material.
Think of a thin filling between two metal plates that keep breathing at different rates. One plate grows a little more. The other bows. The gap opens at one corner and closes at another. A soft material gets pushed back and forth until it migrates.
That is the basic thermal grease pump-out mechanism. It may create air pockets, reduce contact area, and raise thermal resistance.
Material at an edge is not automatic proof, though. Some appears during the first clamp-down. A teardown image also cannot tell the whole history.
| What you observe | More likely mechanism | Evidence needed |
| Material appears at the edge after assembly | Initial squeeze-out or normal edge witness | Dispensed volume, outline, pressure, and baseline image |
| Center coverage falls after repeated cycles | Pump-out or cyclic migration | Before-and-after thermal data plus aged coverage |
| Material becomes crusty, cracked, or carrier-poor | Dry-out, bleed, or phase separation | Aged surface condition and material analysis |
| Hotspot is high from the first run | Poor placement, thick bond line, low pressure, or incomplete activation | Assembly record, pressure map, and activation history |
Pump-out develops with time. Squeeze-out happens mainly during assembly. Dry-out concerns carrier loss or separation. Mixing them leads to the wrong fix.
Why PCM Usually Resists Pump-Out Better Than Grease
It changes state when the interface needs to wet
A thin phase change pad is easy to handle below its transition range. Heated under pressure, it softens and fills microscopic valleys on the mating surfaces.
When the interface cools, the PCM becomes firmer again. It does not remain as freely mobile as a typical grease. This soften, wet, and re-firm behavior is central to how a PCM thermal pad works.
A die-cut film also controls outline and material amount. A factory coating reduces operator variation. Less excess material means less is available to escape.
Hot rheology matters more than room-temperature feel
Two pads can look alike on a bench and behave differently when hot. Pump-out resistance depends on the matrix, filler network, carrier, adhesion, cohesion, and how they change through transition.
Yield stress, storage modulus, loss modulus, and tan delta describe how much a material acts like a spring or a liquid. The balance is tricky. PCM must wet rough surfaces without letting cyclic shear walk it toward the edge.
So, a high viscosity number at room temperature proves very little. The useful behavior is at the real interface temperature, pressure, cycle rate, and dwell time.
Resistance is not immunity
A PCM can still migrate if its transition range is wrong, it stays hot, the clamp relaxes, or the assembly warps heavily. Formulation and carrier also change the flow path.
“Pump-out resistant” is a design claim to verify. It is not a magic shield.
How a PCM Pad Can Still Move or Lose Coverage
First, phase behavior must match operating temperature. A PCM that stays firm may not wet fully. One that remains very soft through long hot dwell gives gravity and shear more time to work, especially in a vertical assembly.
Volume matters. A thick layer adds thermal resistance and movable material. A poor outline or edge keep-out may create a flow path. Too little material gives incomplete coverage.
Low clamp load may leave patchy contact. Excessive load can squeeze out material, bow a board, or stress a bare die. Uneven force lets one corner become a tiny pump.
Other common contributors include:
- large package, heat-spreader, or baseplate warpage;
- high CTE mismatch between the two mating structures;
- fast temperature swings and large temperature deltas;
- clamp relaxation, loose fasteners, or unsuitable spring hardware;
- vibration combined with a softened interface;
- oil, solvent, release residue, or fingerprints on a mating surface;
- damaged carrier film or poor die-cut conversion;
- incorrect storage, expired stock, or an unapproved substitute;
- counterfeit material with unknown composition.
A little PCM around the die after activation may be normal. Pump-out needs evidence of progressive movement, center depletion, void growth, or thermal drift. Without a timeline, “it leaked” is only a guess.
Assembly Mechanics Often Decide the Outcome
Warpage turns each heat cycle into a small pump
Silicon, copper, aluminum, ceramic, and laminate expand differently. A GPU package and cooler may bow in opposite ways. A power-module baseplate can deform as junction temperature rises, moving local gap and pressure.
その Solstice explanation of TIM pump-out connects expansion, contraction, rapid temperature changes, material displacement, and air-void formation. Its published performance figures belong to the named PTM6880 product and test conditions. They are useful screening data, not a universal PCM lifetime requirement.
A small CPU die, a large IGBT baseplate, and a lidded server processor do not behave alike. Scale, stiffness, fastener position, and thermal gradient matter.
Bond-line thickness and pressure need a window
A thin bond line shortens the heat path and leaves less volume to migrate. Yet thin works only with reliable contact. Roughness, flatness, particle size, carrier, and pressure set the practical minimum.
This is why 熱伝導率対熱インピーダンス matters. A high W/mK value can lose to a modest material that forms a thinner, more stable interface. The assembled resistance is the result that the device feels.
Pressure should be high enough for wetting and stable contact, but stay within the package and board limits. Use a controlled window. Do not solve every thermal problem with another turn of the screw.
Shared coolers make diagnosis messy
Laptop and graphics-card coolers often touch the processor, memory, and VRM at once. A gap pad that is slightly too thick can lift the cold plate from the GPU die. So can a bent frame or uneven screw sequence.
That looks like failing PCM, but the real fault is stack-up. Check the whole cooler before changing the die material again.
PCM, Grease, Gap Pad, Gel, Curable TIM, or Graphite?
No material family wins every interface. Match the choice to gap, heat flux, pressure, production, lifetime, electrical needs, and repair.
| Material family | Pump-out tendency | Main strength | Main trade-off |
| 相変化TIM | Usually low when formulation and assembly are matched | Thin bond line, controlled handling, good wetting after activation | Needs correct transition range, pressure, outline, and commissioning |
| Non-curing grease | Formulation- and assembly-dependent; may migrate under cycling | Excellent micro-gap filling and easy application | Dispense variation, bleed, dry-out, or pump-out may limit life |
| Silicone gap pad | Not classic grease pump-out; may shift, tear, or take compression set | Handles larger gaps and tolerance stack-up | Thicker interface and higher contact resistance |
| Thermal gel or putty | Depends on yield stress and support geometry | Conforms to uneven or changing gaps | Slump, bleed, dispense control, and rework need validation |
| Curable TIM | Often strong migration resistance after cure | Stable location and bond | Cure time, stress, permanence, and repairability |
| Graphite or solid film | No liquid-style pump-out | Thin, clean, and dimensionally stable | Conformability, electrical conductivity, and edge handling |
For large IGBT or SiC interfaces, thermal grease for power modules may still be the right answer when the approved stencil, bond line, mounting system, and lifetime evidence support it. PCM is attractive when a controlled film and better cyclic stability solve a real problem. Curable materials suit other assemblies where rework is less important.
Choose against interface requirements, not reputation alone.
Where Pump-Out Resistance Matters Most
Direct-die CPUs, GPUs, laptops, and AI servers
Direct-die interfaces combine high heat flux with small contact area. Chassis flex and repeated boost cycles add movement. A thin PCM can stay more consistent than grease under those cycles.
The outline, cooler flatness, authentic material, die pressure, and nearby gap pads must still be right. PCM cannot repair a tilted cold plate.
IGBT and SiC power modules
Inverters, onboard chargers, DC/DC converters, and industrial drives use wider interfaces. Their baseplates and heat sinks can distort under load. Vehicles add vibration, humidity, storage extremes, and long service life.
Pre-applied PCM simplifies amount and placement in mass production. The module supplier’s mounting and commissioning procedure remains part of the material system.
Telecom, LED, and industrial equipment
Telecom, 5G, power supplies, and LED modules may run for years with limited service. Some mount vertically or see daily ambient cycling.
Avoiding periodic repaste work may matter more than a small gain in initial laboratory resistance. Stability becomes the business case.
Build a Pump-Out Test Around Real Use
Begin with a controlled baseline
Record PCM grade, lot, storage, thickness, outline, carrier, placement, surface finish, flatness, fasteners, torque, and estimated pressure. Capture bond-line thickness if measurement will not disturb the assembly.
Follow the supplier’s activation process. Some PCMs need defined temperature, load, and dwell. Measure before and after activation so normal seating is not mistaken for drift.
Cycle the mechanisms that exist in the product
Temperature cycling tests ambient expansion. Power cycling creates internal gradients and package warpage. They are not interchangeable. Add vibration, humidity, vertical orientation, or hot dwell when the product sees them.
その Infineon pre-applied PCM application note combines commissioning, power cycling, humidity, vibration, and temperature shock for a defined module system. Its values and results belong to that system.
IEC 60068-2-14 guides temperature-change testing. IEC 60749-34 addresses semiconductor power cycling. JESD22-A104 covers component cycling, while ISO 16750 is relevant to vehicles. None provides one PCM cycle count for every product.
Measure the aged interface, then inspect it
ASTM D5470-17(2024) supports controlled thermal transmission and interface-impedance measurement. Keep pressure, thickness, temperatures, and calculations consistent before and after aging. Its fixture does not reproduce every package, vibration, or cooler.
The broader overview of 一般的なTIM試験規格 helps separate bulk-property data from interface and device tests.
At device level, compare temperatures at matched power and boundaries. Track thermal-resistance drift and hotspot delta. Define limits for Rth change, temperature rise, voids, or coverage before testing.
Inspect representative samples. Teardown can show migration, depletion, cracks, bleed, or poor imprint. Ultrasonic imaging may reveal internal voids. Use controls and enough units to spot real variation.
The evidence chain is simple: baseline, activation, relevant cycling, aged measurement, and inspection. A datasheet number cannot replace it.
Diagnose Rising Temperatures Before Blaming the PCM
Keep power, ambient, cooling, firmware, sensor position, and dwell consistent. Otherwise, a system change can masquerade as interface aging.
| Symptom | Possible cause | Next check |
| Temperature is high on the first run | Incomplete activation, poor placement, thick bond line, or low clamp load | Verify the process record, pressure map, outline, and first-cycle behavior |
| Temperature rises through repeated cycling | Pump-out, dry-out, clamp relaxation, or cooling change | Retest under controlled boundaries and inspect aged coverage |
| Edge material is visible but temperatures stay stable | Initial squeeze-out or designed edge witness | Compare baseline images and monitor before reworking |
| One corner becomes hotter | Uneven pressure, warpage, debris, or tilted cooler | Check flatness, fastener sequence, contact imprint, and local coverage |
| GPU improves after service, then regresses | Grease migration, shared-cooler stack-up, or mounting movement | Inspect die pressure, memory-pad thickness, chassis flex, and interface evidence |
| One production lot behaves differently | Material, carrier, storage, conversion, or assembly variation | Quarantine the change and compare traceable samples |
Before wiping a teardown, photograph both surfaces and mark orientation. A rushed cleaning job erases the failure pattern.
Select and Release PCM for Production
Start with service temperature, power profile, allowable drift, lifetime, orientation, isolation, and maintenance. Then shortlist phase change thermal interface materials by transition behavior, impedance, thickness, carrier, outline, and format.
Build a pressure window using real flatness and fastener tolerance. Define edge clearance and placement. Check liner removal and pad handling. A laboratory sample that tears on the line is not production-ready.
Use application-specific material selection and testing. Run pilot builds with intended hardware and activation. Converted, production-representative samples beat hand-cut coupons when outline affects flow.
The control plan should cover incoming material, storage, liner, lot traceability, placement, torque, activation, and end-of-line checks.
Review changes to carrier, thickness, converter, heat-sink finish, hardware, torque, or site. “Same W/mK” is not enough for automatic substitution.
結論
PCM thermal pads are often a strong answer to pump-out, especially where repeated thermal cycling makes grease coverage unstable. Their advantage comes from controlled wetting, a thin bond line, and a firmer state after cooling.
But the pad does not work alone. Transition range, volume, pressure, flatness, warpage, vibration, and activation all shape the result. A good PCM is a bit like a well-fitted shoe insole: decent material helps, but size and load still have to match.
Define the failure correctly. Then prove the interface with aged thermal data and physical evidence.
FAQs About PCM Thermal Pads and Pump-Out Resistance
What is the pump-out effect in a thermal interface?
Pump-out is the gradual movement of TIM away from active contact during repeated temperature, pressure, and shape changes. It can create air pockets and raise thermal resistance. It differs from material squeezed out during initial assembly.
Do PCM thermal pads completely prevent pump-out?
No. A well-matched PCM usually resists migration better than many non-curing greases, but it is not immune. Wrong transition behavior, excess material, uneven pressure, warpage, vibration, or contamination can still cause movement.
Why do PCM pads resist pump-out better than many thermal greases?
PCM softens under heat and pressure to wet roughness, then firms again when cool. That limits mobility. Die-cut or pre-applied formats also control material amount and outline.
Can a PCM pad still leak or move after phase change?
Yes. Risk increases with long hot dwell, cyclic shear, vertical mounting, vibration, excess volume, poor edge clearance, clamp relaxation, or contamination. Performance depends on the formulation and assembly.
How can you tell pump-out from squeeze-out or dry-out?
Use timing and evidence. Squeeze-out appears during initial clamping. Pump-out progresses with cycles and may deplete the center. Dry-out often leaves cracks or carrier loss. Compare baseline images, aged thermal data, and teardown observations.
Does higher mounting pressure improve pump-out resistance?
Only within the approved range. Enough pressure promotes wetting and a stable bond line. Too much may cause squeeze-out, board bow, cooler distortion, or die damage. Uniform pressure matters more than maximum pressure.
How is PCM pump-out resistance tested?
Measure a baseline, activate the material, run relevant temperature or power cycles, and retest. Inspect coverage, voids, edge migration, and cracks. Add vibration, humidity, orientation, or hot dwell when relevant.
Is there an industry standard or fixed cycle count for pump-out?
There is no universal PCM pump-out standard or cycle count. ASTM D5470 supports impedance measurement; IEC, JEDEC, and automotive methods provide cycling frameworks. Product mission profile and customer specification should set the test.
Where are PCM pads most useful for preventing recurring pump-out?
They suit thin, high-heat-flux interfaces with repeated cycling: direct-die CPUs and GPUs, AI servers, power modules, EV converters, telecom hardware, and industrial equipment. The mounting system must still suit the PCM.
Should a GPU or laptop use PCM when thermal paste keeps pumping out?
PCM may help when grease migration is verified. First check cooler flatness, screw sequence, die pressure, chassis flex, and nearby gap-pad thickness. Use the correct outline and activation, then confirm temperatures across repeated cycles.
