What it shows
For the methanol blend, the model’s heat-flux figure is a textbook boiling limit for pure methanol, multiplied by assumed boosts; it is not a model of this blend boiling on its own terms. Its median answer is 357.6 W/cm².
It runs 20,000 random draws, each nudging the inputs within an assumed uncertainty, for example by 15% on surface tension. Nine draws in ten land between 203.7 and 556.1 W/cm². In 78.6% of the draws the methanol blend came out ahead of the other three candidates.
Those three are our own as well: two fluorinated blends and a pentane blend. The model does not compare them with any fluid on the market.
The same run gives the pentane blend a median of 184.6 W/cm². Nine draws in ten land between 103.5 and 292.1. It too is modelled by computer, not mixed or measured, and rests on the same assumed boosts.
What the figure is made of
The heat-flux figure is a textbook boiling limit for pure methanol, multiplied by assumed boosts. It is not a model of this blend boiling on its own terms. The published file gives neither that bare limit as a number nor the size of each boost: it marks the bare limit as an estimate and the boosts as taken from published studies and not tested.
Most of the figure, and of the share of draws the blend wins, may come from those boosts. No run with them set to zero is published for this blend.
Why it matters
A model like this sets the order of experiments. It narrows a large field to a short list, and it says in advance which bench results would surprise us. That is its use. It does not say the blend will win a real test.
It also carries two limits a buyer asks about first. Methanol burns, the blend’s flash point has not been measured, and so it is not shown to be non-flammable. Its permittivity is too high for direct immersion, so it fits cooling loops that run beside the chip, not liquid in contact with the electronics.
Who should care
- A data-center operator: treat this as a reason to talk to a lab, not as a product to buy.
- A partner lab: the first bench steps are mixing, checking the liquids stay together, and measuring the flash point.
- An investor: read the label before the number. This figure is modelled, with its assumptions stated.
The post What "modelled, not measured" means walks through the same model in more detail.
The limits, in the lab’s own words
Quoted word for word from the published file named under each quote. Read the result with them.
Monte-Carlo over computational/literature estimates, not measured CHF. Stack uplift unproven until Phase-5. P(top-k) reflects estimate uncertainty, not experimental confidence.
From The modelled heat flux of the four candidate blends, from the seeded draws.
Prior art
The published work this result uses or is checked against. Each link was opened and checked when this page was written.
- Critical heat flux, including Zuber’s textbook limit (Wikipedia): the standard limit on how much heat boiling can carry off a surface, which the model starts from
- Propagation of distributions using a Monte Carlo method, JCGM supplement to the measurement-uncertainty guide (BIPM, PDF): the standard method of pushing input uncertainties through a model by random draws
Evidence
Each number with a dotted underline was read from one of these published files, field by field, when this page was built.