Vapor Chamber
Principle: spreading heat by evaporation and condensation
Overview
I hear vapor chambers are used to cool smartphones and GPUs. How are they different from a copper plate?
It's a hollow, flat enclosure with a little liquid such as water inside. Liquid evaporates at the hot spot, the vapour spreads through the whole chamber almost instantly and condenses on cooler surfaces, depositing its heat. The condensate returns to the hot spot by capillary action in the wick on the inner walls. Because heat travels with moving mass, it spreads in-plane many times faster than through copper. Think of it as a flattened heat pipe.
Why spreading helps
When a small chip's heat is removed through a large heat sink, it must spread sideways in the spreader. If it doesn't spread fully, only part of the heat sink works and the area above the chip runs hot (spreading resistance). The ideal case of perfectly uniform spreading gives:
Modelling approach
System-level thermal analyses almost never solve the internal two-phase flow. The VC is replaced by an anisotropic solid with very large in-plane effective conductivity $k_{xy}$ and a small through-thickness value $k_z$ reflecting wick resistance at the evaporator and condenser. Values are set from manufacturer tests or matched to measured temperatures under the same conditions.
A cooling technology from space
The heat-pipe principle was patented in the 1940s and developed in earnest at Los Alamos in the 1960s. The goal was to move heat without pumps in environments such as spacecraft and reactors, even without gravity — the wick's capillary action returns the liquid regardless of orientation. Over about half a century it was shrunk and thinned, and today it sits inside smartphones as plates only around 0.3 mm thick.
Worked example: copper versus a vapor chamber
Conditions
A 50 W source sits at the centre of the top of a 60 mm square, 3 mm thick spreader, cooled underneath by a heat sink (including TIM) represented as 5000 W/(m²·K). A 3D finite-volume conduction model (60 × 60 × 6 cells) compared the peak top-surface temperature rise. The vapor chamber was an anisotropic solid with 10,000 W/(m·K) in-plane and 40 W/(m·K) through-thickness (illustrative values). Perfect uniform spreading would give a rise of 2.8 K.
| Spreader | 10 mm source | 30 mm source |
|---|---|---|
| Aluminium (200 W/(m·K)) | 20.3 K | 7.2 K |
| Copper (390 W/(m·K)) | 12.7 K | 5.5 K |
| Vapor chamber | 5.5 K | 4.2 K |
| Ideal (perfectly uniform) | 2.8 K | 2.8 K |
With the small source it's less than half of copper, but with the large source the gap is small.
A large source doesn't need much lateral spreading in the first place. A VC shines when the source is small and the cooled area large. Also, a VC's through-thickness resistance is higher than copper's, so when source and cooled area are similar in size, that penalty can outweigh the in-plane gain. Decide by comparing with a copper plate using calculations like this.
Temperature uniformity
With the same 10 mm source, the top-surface temperature spread (max minus min) was 11.4 K for copper and 2.0 K for the vapor chamber. With multiple chips, or optics and batteries where temperature variation matters, uniformity is an assessment criterion alongside peak temperature.
Design and analysis cautions
Vapor chamber limits
- Capillary limit: the wick can return only so much liquid; beyond that the hot spot dries out and temperature jumps. An effective-conductivity model cannot show this, so check against the manufacturer's maximum heat load.
- Local heat flux: even within the total heat limit, a very small source with high heat flux can hit a local boiling limit.
- Orientation and low temperature: returning liquid against gravity reduces performance. At low temperatures vapour pressure is low, and if the working fluid freezes, start-up is slow.
Checking the model
- Reproduce the manufacturer's test conditions (thermal resistance or temperature distribution) and tune the effective conductivities to match.
- Solve the actual product conditions with those values and confirm the heat load is within the maximum.
- Measure temperatures near and far from the source on a prototype to confirm in-plane spreading matches the analysis.
“We switched to a VC and it didn't cool better”
A common complaint is that switching to a VC barely lowered the temperature. Investigation often finds the heat sink itself was small, leaving nowhere to spread to, or a thick interface between the VC and heat sink was the largest resistance. A VC reduces only spreading resistance; heat-sink-to-air and interface resistances stay the same. Calculate the resistance breakdown first, find the largest term, and choose parts accordingly.
Common mistakes
Mistakes and fixes
| Mistake | Effect | Fix |
|---|---|---|
| Using a large isotropic conductivity | Through-thickness resistance ignored; temperatures too low | Specify in-plane and through-thickness separately |
| Not checking the maximum heat load | Dry-out causes a sudden temperature jump | Compare with the supplier's limits |
| Choosing a VC for a large source | Benefit doesn't justify the cost | Compare with copper by calculation first |
| Overlooking heat-sink-side resistance | VC doesn't lower the temperature | Calculate the resistance breakdown first |
I'd like to explore the heat-pipe principle too.
Use the heat pipe simulator for heat transport, the capillary rise simulator for wick capillary action and the Clausius–Clapeyron simulator for vapour pressure versus temperature of the working fluid. Heat-sink design is in the heat sink design calculator. Related material is in the articles on heat pipes in electronics and thermal resistance (Rth) models.