03 · X-Score Lab

Scoring methodology

The X-Score is a weighted aggregation across eight subscores. Subscores are independent, on a 0–100 scale, and reflect physical, computational, and feasibility considerations. Adjust subscores to interrogate how candidate ranking changes.

01

Methodology

Subscore definitions & weights

Weights sum to 1.00. The overall X-Score is the weighted mean rounded to the nearest integer.

#SubscoreWeightDescriptionValue
  1. 01Thermodynamic Stability15%Convex-hull plausibility and formation enthalpy at the target pressure.70
  2. 02Phonon Stability15%Estimated absence of imaginary modes across the Brillouin zone.65
  3. 03EPC Potential20%Plausibility of strong electron-phonon coupling from hydrogen modes.75
  4. 04DOS / Fermi-Level Relevance10%Estimated weight of hydrogen-derived states at the Fermi level.60
  5. 05Pressure Feasibility15%How accessible the target pressure is for synthesis and measurement.55
  6. 06Synthesis Feasibility10%Practical feasibility of the proposed synthesis pathway.50
  7. 07Novelty5%Distance from already-screened hydride candidates in the literature.70
  8. 08Validation Readiness10%Maturity of the workflow needed to take the candidate to validation.65
02

Adjust

Interactive scoring panel

Move sliders to interrogate the scoring surface. Optional: load subscores derived from an existing candidate's headline X-Score.

Load from candidate

Thermodynamic Stability

Weight 15%

70

Phonon Stability

Weight 15%

65

EPC Potential

Weight 20%

75

DOS / Fermi-Level Relevance

Weight 10%

60

Pressure Feasibility

Weight 15%

55

Synthesis Feasibility

Weight 10%

50

Novelty

Weight 5%

70

Validation Readiness

Weight 10%

65
03

Result

Aggregate

X-Score65·Promising
Overall X-Score65/ 100

Recommended next gate

Experimental validation planning

ThermoPhononEPCDOS/EFPressureSynthNoveltyValidate
04

Risks

Risk flag panel

Conservative, threshold-based advisories. Flags are not scientific validation.

  • 01No high-severity risk flags raised.

Validation required

AI-generated research candidates are exploratory hypotheses and require DFT, DFPT, EPW, Eliashberg, RPA, and experimental validation before scientific claims can be made.