ROOM-TEMPERATURE SUPERCONDUCTIVITY / RESEARCH WORKBENCH
My working hypothesis: an ordered hydrogen-bearing solid, stabilized by a light-element framework. This page explores what pairing at 293 K would demand, maps that demand with a numerical solver, prepares the calculations, and lays out the experiments that could disprove the idea.
Explore a two-mode spectrum or analyze your own electron–phonon spectrum.
ALLEN–DYNES ESTIMATE
Scale all spectral weights together; keep mode energies fixed.
| Scenario | λ | ωlog | μ* | Model Tc | Origin |
|---|
Up to eight scenarios, saved in this browser. These are model comparisons, not a ranking of real materials.
A documented hydride benchmark helps validate the tools before testing the framework hypothesis.
Drag or use arrow keys to rotate. Cell edges show periodicity; no chemical bonds are inferred.
Xu et al. predict about 130 K for ordered MgAlFeH₆. Its published lattice (6.24 Å, rounded) and sites supply a starting cell. This workbench reconstructs rounded coordinates; it has not reproduced their calculation. Primary paper ↗
SrB₃C₃ is the framework without the hydrogen: a synthesized Pm-3n clathrate (4.88 Å calculated at 0 GPa, one 1.73 Å B–C bond) that superconducts near 20 K at 40 GPa. Its starting cell is selectable above, with Sr on 2a, B on 6c and C on 6d. So is SrNH₄B₆C₆, the hydride-filled version the concept cites: the Fmmm phase from Sun & Zhu’s supplementary table (9.7937 Å along a, 36 atoms in the primitive cell, ammonium in every other cage), predicted at 85 K and never synthesized. Both are DFT geometries, not measured ones, and neither has been relaxed or calculated here. Structure paper ↗ Measurement ↗
| Atom | Element | x | y | z |
|---|
The browser explores pairing. Quantum ESPRESSO, EPW and SSCHA run on a configured workstation or cluster.
EXECUTED ON THIS SITE’S RASPBERRY PI · TOOLKIT CLI
EXECUTED ON THIS SITE’S RASPBERRY PI
The scf, phonon and electron–phonon stages of the pipeline, run with Quantum ESPRESSO on materials whose answers are known, then fed through the toolkit’s solver. Aluminium checks that the pipeline runs and that its α²F round-trips into the pairing model. MgB₂ is experiment 1 in its computational half: the first literature anchor on the T꜀ landscape, recomputed here. The MgB₂ recipe was then re-run unchanged on a rented 32-core node, with the same conda-forge build, to check that the numbers do not depend on the machine. None of this is a hydride, the hypothesis, or converged.
RUNNABLE PYTHON TOOLKIT
Includes structures, variants, checksummed job preparation, staged QE execution, spectral integration, a numerical isotropic Eliashberg solver, and competing-phase enthalpy comparisons.
Download toolkit ↓PyPI: hydrogen-rtlab ↗Read the run guide# From PyPI (the same code as the zip's simulation/ directory): python -m pip install hydrogen-rtlab # or from the zip: cd simulation python -m pip install -e . python -m rtlab doctor python -m rtlab analyze --modes examples/two-mode.json --eliashberg # For DFT, supply UPF files and converge settings. # No paid compute or remote jobs are launched.
Progress is evidence for a reproducible phase, not a single resistance anomaly. Each experiment carries a kill criterion, pre-registered on 9 September 2026 before any of them has run: the result that ends the hypothesis.
Checkmarks record your review, not experiment completion. Notes stay in this browser. Status chips are set in the page source, dated, and change only when work is done.
PRESSURE-RELEASE HYPOTHESIS
Explore a single-event Arrhenius model. The barrier must come from a pathway calculation or experiment.
τ = ν⁻¹ exp(Eb / kBT). This is not a bulk lifetime prediction: nucleation sites, diffusion and changing decomposition pathways are omitted. Short molecular dynamics cannot establish long-term retention.
The toolkit's isotropic Eliashberg solver, run on this site's own Raspberry Pi over synthetic one- and two-mode spectra. The map is this experiment's data. Published materials sit on it as single-mode proxies for comparison. Nothing on it is a candidate.
For a single Einstein mode at energy ω, the smallest λ whose solver T꜀ reaches the target.
Each cell is one solved spectrum. The gold line is the 293 K contour.
REAL MATERIALS ON THE MAP
| Material | λ | ωlog · meV | Paper T꜀ · K | Measured T꜀ · K | Proxy Eliashberg · K | Proxy Allen–Dynes · K | λ multiple for 293 K |
|---|
Proxy columns are this site’s solver on one Einstein mode at the published ωlog with the published λ, μ* = 0.10. The difference between the proxy and the paper is the price of collapsing a spectrum to one number.
Two synthetic modes share one total coupling: a framework mode at 40 meV and a hydrogen mode at 140 meV. The central mechanism of the bet, put to the solver.
Each cell is one solved two-mode spectrum at μ* = 0.10. Gold: the 293 K contour.
Against the hydrogen share, for three hydrogen-mode energies.
WHAT THE SPLIT SAYS
READ THE MAP
WHAT THE MAP SAYS
Loading the solver record.
The coupling needed at μ* = 0.10, solved on 192 and then 384 positive Matsubara frequencies. The split map and the anchors use the same solver and settings.
| ω · meV | λ at N = 192 | λ at N = 384 | Difference |
|---|
| ω · meV | Eliashberg λ · μ* 0.10 | Eliashberg λ · μ* 0.15 | Allen–Dynes λ · μ* 0.10 | Allen–Dynes λ · μ* 0.15 |
|---|
Download the full record as tc-map.json: both grids, the anchors, the toolkit checksum, solver settings and library versions.
Sources checked 9 September 2026. No room-temperature, ambient-pressure material is demonstrated by this project.
The bet is to preserve useful hydrogen-related electron–phonon coupling while a framework stabilizes the structure after pressure release. Ordering, carrier tuning and retention are hypotheses to test. No winning composition has been identified.
Original annotated concept ↗Sun & Zhu predict about 85 K for SrNH₄B₆C₆ and up to 115 K for PbNH₄B₆C₆ at ambient pressure. These motivate the route, but do not establish 293 K superconductivity. High-frequency hydrogen motion alone does not guarantee strong pairing. This project has run their SrNH₄B₆C₆ cell through the toolkit’s stages on a rented node: the LDA cell shrinks to a 6.70 Å primitive vector and every Γ-point phonon is real, up to 98 THz for the N–H stretch. Stable at the zone centre at that level. Its electron–phonon spectrum then took 5.8 h of rented-node time on a 2×2×2 mesh: λ 1.07, ωlog 44 meV, 54 K from the isotropic solver, against the paper’s λ 2.01, ωlog 377 K and 68 to 73 K by the same interpolation method (its tetrahedron numbers differ from its own by half again). The coupling is about half the paper’s at these meshes, which is the same shortfall the MgB₂ reproduction showed before its dense run. The split is the robust part: 80% of λ sits below 700 cm⁻¹ in framework, Sr and NH₄ libration modes, 20% in the band where B–C stretches and NH₄ bends mix, and 0.0% in the N–H stretches. On the hydrogen-share map that would be the bottom edge. Correction, 23 September 2026: this spectrum cannot carry that verdict. The electron–phonon run relaxed the cell again at a lower cutoff (40 Ry, against 50 Ry for the zone-centre job), and the stress recomputed after that relaxation is -41 kbar: the cell was not relaxed, and every computed q-point has imaginary branches, down to -14.6 THz, which the α²F leaves out. The split above describes that strained cell, not the material. Whether the bet holds in this cell is open until the spectrum is recomputed on a relaxed cell (the check). Recomputed, 24 September: relaxed properly at 50 Ry (stress -2 kbar) and with a tight phonon threshold, the cell is stable at the zone centre but has one soft branch at the zone boundary along each axis, down to -2.3 THz. Its coupling table reads λ 3.3 with ωln 66 K, which is what a soft branch does to λ, not strong pairing. At this level of theory the published cell is weakly unstable, so no transition temperature and no hydrogen-share verdict come from it; following the soft mode to the distorted cell is the next calculation (the recompute).
Communications Physics, 2024 ↗Xu et al. predict roughly 130 K for ordered MgAlFeH₆. Their random-alloy estimate is around 40 K. The ordered phase is metastable, with a reported energy above hull near 0.11 eV/atom. These are literature results, not workbench outputs.
npj Computational Materials, 2026 ↗SrB₃C₃, the boron–carbon clathrate, has been made and measured: onset near 20 K at 40 GPa, with λ of 0.95 and 27 to 43 K predicted at ambient pressure. Most of its coupling sits in the framework modes. It is the structural half of the bet with the pairing half missing. This project has now run it through the toolkit’s own stages: the LDA cell relaxes to 4.755 Å and every phonon on a 2×2×2 mesh is real, up to 26 THz, so the framework is dynamically stable at that level. Its electron–phonon spectrum has not been computed here.
Physical Review Research, 2023 ↗Dolui et al. predict 160 to 175 K at ambient pressure for cubic Mg₂IrH₆ (6.66 Å), with λ 2.5 and 91% of it in hydrogen bands, because the IrH₆ units’ antibonding states sit at the Fermi level. Ca₂IrH₆, same structure, is predicted not to superconduct. This project ran both through the same α²F recipe as SrNH₄B₆C₆. Mg₂IrH₆ puts 69% of its coupling in Ir–H modes at every mesh and smearing; Ca₂IrH₆ puts 32%, with λ 0.39, ωlog 26 meV and no transition above the solver’s 8 K floor. The pipeline sees hydrogen-carried coupling where the literature says it is and not where it says it is absent. Mg₂IrH₆’s total took three k meshes to pin. At 16×16×16 against 12×12×12 the mesh-agreement rule lands at λ 1.72 with ωlog 63 meV, which the isotropic solver turns into 131 to 168 K from the pick to the smallest smearing, against the paper’s λ 2.5 and 160 to 175 K. The table still spans λ 0.77 to 2.16, so the number is a criterion, stated; but the criterion now sits within a third of the paper instead of a factor of three below it, which is what a finer mesh was supposed to do. Two more published members of the family went through the same recipe afterwards. Mg₂RhH₆, predicted at 45 to 80 K in the Mg₂XH₆ family and since made, superconducting at 24 K under 30 GPa: 79% of λ in the Rh–H frequency window (78% by atomic participation), λ 0.68 to 1.43 across smearings, 32 to 113 K. Li₂AuH₆, predicted at 140 K: 53% of λ in the Au–H frequency window (77% by atomic participation), λ 1.66 to 1.99, 97 to 119 K. LiMgAuH₆, listed above 110 K in a January 2026 screen of quaternary MAXH₆ cells: 66% of λ in the Au–H frequency window (82% by atomic participation), λ 1.97 to 2.75 across smearings, 119 to 153 K, ω_log 55 meV. Only two of these λ are converged in the k mesh at the smearings quoted: Mg₂IrH₆ and Mg₂RhH₆ agree between their two densest meshes within 3% from 0.01 Ry up. Li₂AuH₆ and LiMgAuH₆ move by 5 to 36% between the 8 and 12 k-meshes, so their ranges are first looks, and a 16 k-mesh run of Li₂AuH₆ is queued. It has the strongest coupling of any control on this page, and its soft ω_log is why that coupling stops where the Tc range above says: the demand map wants the same λ on a spectrum three times harder. Four cells of this class, sorted the way the literature sorts them, at a few dollars each.
Physical Review Letters, 2024 ↗ Angewandte Chemie, 2024 ↗ npj Computational Materials, 2024 ↗ JACS, 2026 ↗ Physical Review B, 2025 ↗Quantum ESPRESSO ran end to end on this site’s Raspberry Pi. Aluminium first, as a pipeline check: λ 0.35 and 1.0 K against 0.42 and 1.2 K in the literature. Then MgB₂, the first anchor on the T꜀ landscape, as experiment 1’s computational half, at three meshes over five hours of Pi time. At the densest, reading λ where the two finest meshes agree: λ 0.90, ωlog 56 meV and 44 K against the published isotropic λ 0.87, 504 cm⁻¹ and 40 K, and 39 K measured. That meets experiment 1’s 25% criterion for the coupling. It does so by a criterion: across the smearing table λ still runs 0.64 to 0.97, and the frequency scale sits a tenth low. Experiment 1 stays open for its measurement half. “Executed” is a different claim from “documented”, and now the page can show the difference.
Pipeline tab ↗| Quantity | Calculation | Limit |
|---|---|---|
| λ, ωlog, ω₂ | Einstein-mode moments or trapezoidal integration of α²F. | No link from atomic coordinates to the spectrum is assumed. |
| Browser Tc | Allen–Dynes interpolation, including f₁ and f₂. | No anisotropy, explicit anharmonicity or phase-coherence calculation. |
| Python Tc | Linearized isotropic Matsubara equation; largest pairing eigenvalue reaches one. | Finite frequency grid; symmetric Coulomb cutoff. Convergence and a valid spectrum are required. |
| Phase energy | Balanced mixture of supplied reference enthalpies. | Not a global convex hull or decomposition rate. |
| Retention | One-event Arrhenius waiting time. | Not a macroscopic sample lifetime. |
| T꜀ landscape | Python Eliashberg solver over synthetic single-mode spectra, 192 Matsubara frequencies, checked at 384. | Isotropic and synthetic. Cells below the grid's resolved floor are left blank. No spectrum here belongs to a real material. |
| Hydrogen-share map | The same solver over two Einstein modes (40 and 140 meV) sharing one total coupling; coupling needed solved by bisection for hydrogen modes at 100, 140 and 200 meV. | Two modes are not a spectrum. The framework and hydrogen energies are the workbench defaults, not a calculation. |
| Literature anchors | Published λ and ωlog quoted from the ledger; solved here as one Einstein mode at ωlog. | A proxy. Its T꜀ differs from the paper’s by the weight the real spectrum spreads around ωlog. SrB₃C₃’s ωlog is back-solved, not quoted. |
| Aluminium benchmark | Quantum ESPRESSO 7.4: scf, DFPT phonons with electron–phonon coefficients on a 4×4×4 q mesh, matdyn α²F; toolkit solver at 2048 Matsubara frequencies. | Tutorial cutoffs and meshes; λ still moves with the double-delta smearing. A pipeline check, not a converged calculation. |
λ = 2 ∫ α²F(E) / E dE ln(Elog / 1 meV) = (2/λ) ∫ α²F(E) / E ln(E / 1 meV) dE E₂² = (2/λ) ∫ α²F(E) E dE Tc = f₁ f₂ Elog / (1.2 kB) × exp[−1.04(1+λ)/(λ−μ*(1+0.62λ))] f₁ = [1 + (λ / (2.46(1+3.8μ*)))^(3/2)]^(1/3) f₂ = 1 + (E₂/Elog−1)λ² / [λ² + (1.82(1+6.3μ*)E₂/Elog)²] kB = 0.08617333262145 meV/K
The μ* range is a sensitivity interval, not a confidence interval. Ideal isotope scaling changes H-mode energies by 1/√2 and holds λ fixed; real isotope effects can alter structure and anharmonicity.
Every number on this page and the concept page that is not the target sits in claims.json with one of these sources. Hover a dotted number for its citation. The test suite fails on an unledgered number.