KYAMOS / MAT4EU study
PHASE-CHANGE HEAT STORAGE
Explore how a material stores heat as it melts, then releases it as it freezes.
Original figures from a checked aluminium-silicon calculation. A clear view of what the model shows and what still needs a real sample test.
Presentation artwork: a concept illustration, not a calculated field or laboratory image.
01 / The idea
Think of ice melting: heat goes into the change from solid to liquid. Here, the assumed material is AlSi12, an aluminium-silicon alloy, with a much higher melting temperature.
Heat moves from the controlled wall into the sample.
Melting absorbs latent heat: energy used to change solid into liquid.
The material gives back heat as it returns to a solid.
The assumed sample
31.704 grams of dense alloy. Starts at 500 degrees Celsius. The assumed melting point is 576 degrees Celsius.
°C means degrees Celsius. One wall changes temperature; all other walls are insulated. Heat moves by conduction, meaning heat spreading through the material. Density stays fixed. The working grid divides the sample into 5 × 3 × 33 small calculation blocks.
02 / What the calculation predicts
kJ means kilojoules, a unit of heat energy; s means seconds. These are conditional simulation predictions. Near-complete recovery follows from the ideal insulated setup and does not establish device efficiency. Time brackets come from saving results every 0.25 seconds, not measured uncertainty.
03 / The original evidence
Select a figure to explore the saved results, learn how to read the plot, and see its limits. Every plot is the unchanged original from the report and figure presentation.
Figure 01of 06
Follow one complete heat-storage cycle in the assumed aluminium-silicon sample.
24.293 kilojoules are stored at 40 seconds. The sample reaches 99% melted in 25.25-25.50 seconds and 99% frozen 6.00-6.25 seconds after cooling begins.
Read left to right. The dotted line at 40 seconds marks the start of cooling. Liquid fraction means the share that has melted. Latent heat is heat used for melting; sensible heat raises the temperature.
Near-complete heat recovery follows from the ideal insulated setup. It is not a storage-device efficiency claim. The time brackets reflect results saved every 0.25 seconds.
Figure 02of 06
See how heating, melting and freezing move from the controlled wall into the material.
Melting progresses inward during heating. After 40 seconds, freezing progresses inward from the same wall.
Move right to follow time and down to look deeper. The controlled wall is at depth zero, at the top of each map. In the lower map, 0 is solid and 1 is liquid.
These maps show calculated values, not camera images. Liquid movement, the container and imperfect contact are outside this model.
Figure 03of 06
Repeat the same cycle while changing one material assumption at a time.
Changing melting heat from 560 to 470 kilojoules per kilogram lowers stored heat from 24.293 to 21.445 kilojoules. Changing heat conduction by 20% shifts melting time substantially.
Colours identify seven scenarios. Compare heat stored after 40 seconds at lower left and melting times at lower right. Heat capacity means heat needed per degree of warming.
These are chosen alternatives, not measured supplier limits or statistical error bars. Sample volume stays fixed, so changing density also changes mass.
Figure 04of 06
Check whether smaller calculation blocks and shorter time steps change the answer.
The 17-cell grid is rejected: its centre-temperature difference is 5.7487 degrees, above the 3-degree limit. The 33-versus-65-cell difference is 1.4207 degrees; halving the time step gives 0.0092 degrees.
Compare centre temperatures and melted fractions at matching times. A grid divides the sample into small blocks; a time step is how far the calculation advances each time.
The raw campaign remains FAIL: 219 of 220 checks passed. The separate working-grid acceptance retains that failure. Temperature differences in kelvin (K) equal differences in degrees Celsius.
Figure 05of 06
An additional example holds three faces at different temperatures to create a pattern in all three directions.
A heated face and two cooler neighbouring faces produce different melting patterns near their shared corner. Four graphics processors share the calculation.
These slices show the sample at 20 seconds. The z=0 face is held at 620 degrees Celsius, x=0 at 500 and y=0 at 520. The other faces are insulated. The grid has 17 by 13 by 33 cells.
This is a three-dimensional demonstration. It has no separate grid-accuracy claim and does not predict a complete storage unit.
Figure 06of 06
Test simplified melting and freezing problems against an exact mathematical answer.
The earlier mathematical test set passed 434 checks across 31 cases. At 160 depth cells, the largest melting or freezing front error is 0.00773%. Six selected cases are shown here.
The upper plots compare temperatures at 0.25 seconds with the dashed exact answer. Below, compare the moving solid-liquid boundary and the error in its position.
These tests use an invented test material, separate from the alloy study. Error does not decrease steadily with finer grids, so no measured convergence order is claimed. Real sample validation is still needed.
04 / Material sensitivity
Seven scenarios change one material assumption at a time. Melting heat affects stored energy; heat conduction strongly affects how quickly the sample melts.
| Material assumption | Heat at 40 seconds kilojoules | Time to 99% melted seconds |
|---|---|---|
| Nominal published-property mixReference case | 24.293 | 25.25-25.50 |
| Melting heat: 470 kilojoules per kilogram | 21.445 | 21.50-21.75 |
| Original table heat capacities | 22.679 | 24.75-25.00 |
| Density: 2,460 kilograms per cubic metre | 22.624 | 23.50-23.75 |
| Heat conduction: 20% lower | 24.169 | 31.50-31.75 |
| Heat conduction: 20% higher | 24.300 | 21.00-21.25 |
| Melting spread over 10 degrees | 24.298 | 22.50-22.75 |
These alternatives are not statistical confidence intervals or measured supplier limits. Changing density at fixed volume also changes mass. The actual material needs to be measured.
05 / Evidence, with its boundaries
An earlier test set compared simplified melting and freezing against exact mathematics. The finest-grid boundary-position error is 0.00773%. These tests use an invented test material.
Error did not fall steadily with finer grids. No measured convergence order is claimed.
The 17-cell grid exceeded the centre-temperature limit. The originally selected 33-cell working grid passes the finer-grid and shorter-step comparisons.
219 of 220 screening checks passed. The coarse grid is not used; the original failed check remains in the evidence below.
Compare measured heating and cooling curves with the same sample conditions. Record the material batch, sensor calibration, repeat cycles and heat losses.
Material-specific accuracy, storage-unit charging time and device efficiency are not established.
Raw screening status: FAIL, with 219 of 220 checks passed. The exploratory 17-cell grid differed by 5.748698 K, above the 3 K limit. The selected 33-cell grid differs from the finer 65-cell grid by 1.420734 K. Its separate acceptance was added after the coarse-grid failure; the original result and tolerances are unchanged.
The model accounts for heat spreading, warming, melting and freezing at fixed density. It does not include liquid movement, a container, imperfect thermal contact, radiation, expansion, shrinkage or cycling damage.
Complete saved histories and final fields match across calculations shared by one, two and four graphics processors. Direct exchange between graphics-memory buffers was demonstrated. This campaign makes no universal speed or scaling claim.
Earlier checks of the communication software's memory and programming interface were not clean. The numerical results do not resolve those separate diagnostic issues.
06 / From the presentation
The opening image and these four illustrations come from the supplied PowerPoint. They explain the concept; their shapes, equations and curves are not the numerical evidence. The six result plots above provide that evidence.
The porous textures and laboratory setup in the artwork are illustrative. The calculated sample is dense, and independent laboratory validation is a future milestone.
07 / Take the evidence with you
The source documents and numerical table behind this page.
Saved simulation results: 15 September 2026. Figure companion and webpage: 28 September 2026. This page reuses the saved evidence; no new simulation or sample measurement is introduced.
The next milestone
Independent material sample testing is the next step toward material-specific predictions.
For material researchers and thermal-storage developers. Tell us about your material, heating cycle, sample measurements and validation needs.
Pricing: quoted per project. Scope, computing needs, validation checks and support determine the quote. Contact us to agree access and deliverables before work starts.
This page presents saved numerical evidence. Independent sample testing is still required for material-specific design decisions.