Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

SEPTUM: what separate parties need to share

The septum is the wall between the heart’s chambers; before birth it keeps a small controlled opening, the foramen ovale. SEPTUM groups the studies in which knowledge belongs to separate parties, a research group screening two-dimensional semiconductors and the computing centre that runs its density functional theory (DFT) jobs, and asks how small the opening between them can be while still helping both.

MuSDL: when sharing pays

μ mu. The septum is measured in micrometres of muscle; the centre meters each job’s memory to learn when the wall is worth opening.

Compound-specific memory demand determines when a research group and a computing centre benefit from sharing information in two-dimensional semiconductor screening

Thirty MX2_2 layers were screened with Quantum ESPRESSO, recording each calculation’s duration and memory. In small cells, demand followed calculation stage and cell size, which the centre already knows, and no arrangement raised throughput. At 8×8 supercells, compounds of equal electron count differed in memory by 23–32%, and one disclosed memory-class bit per job raised the centre’s throughput 1.133-fold.

Use for a campaign: batch a gap screen; disclose one memory class per job only for large defect and heterostructure supercells.

EpsilonSDL: minimal reciprocal disclosure

ε epsilon. The foramen ovale is a small opening, just enough; the disclosure that helps both parties is one bit per job.

Minimal reciprocal disclosure in shared semiconductor screening depends on memory pressure, billing, timing masking and division of work

Across 1,000 simulated scenarios and four preregistered measured comparisons, no agreement, not even full disclosure, benefited both parties unless jobs filled much of node memory, reserved memory was billed, completion times were masked and grading was paired. With all four, one bit per job raised the centre’s valid results per unit charge 1.13-fold.

Use for a campaign: before disclosing, check memory pressure, reserved-memory billing and completion-time visibility.

A third study (⊥ perp: the wall stands square between the chambers, and what still crosses it is the timing), on what job timing reveals about a group’s secret candidate compounds, is under double-blind review; its page will be added to Articles when the review ends.

Code and data

Each published article has a public repository under Sparks-Ackolf-HOLAR-Autonomous-Labs with a Zenodo DOI for every release; the links are on each article page. Both are listed under Articles.