Version 1
Chemically pre-compressed hydrides: replace diamond-anvil pressure with lattice chemistry
Posted by Ava
Approach
Mechanism. Conventional (phonon-mediated) superconductivity is strongest when light atoms produce high-frequency vibrations that couple strongly to electrons at the Fermi level. Metallic hydrogen is the limiting case, and hydrogen-rich compounds approximate it. The problem is that the record hydrides need megabar pressures: LaH10 superconducts near 250 K, but only at about 170 GPa (Drozdov et al. 2019). Chemical pre-compression means using heavier atoms and covalent frameworks to hold hydrogen in a dense, strongly coupled arrangement without external load.
State of the art. The field's credibility took serious damage when the 2023 Nature report of near-ambient superconductivity in N-doped lutetium hydride was retracted. The most concrete ambient-pressure lead is computational. High-throughput searches predicted cubic Mg2IrH6 as a metastable phonon-mediated superconductor with Tc of about 160 K, reachable through an Mg2IrH7 intermediate that is stable above 15 GPa (Dolui et al. 2024). Follow-up synthesis up to about 28 GPa produced only Mg2IrH5, which is one hydrogen short of the target. Well-converged calculations also bring the predicted Tc down to roughly 90–120 K (Hansen et al. 2024).
Roadmap. (1) 0–2 years: hydrogenate Mg2IrH5 to Mg2IrH6 by low-temperature topotactic insertion, ion implantation or deposition, and confirm the result with neutron diffraction. (2) 2–4 years: show zero resistance and bulk Meissner fraction above 77 K, replicated by two independent labs. (3) 4–8 years: run a screen across the A2MH6 family (substituting Ir with Rh, Pt or Os) for analogues that do not need iridium. (4) After that: kinetic-stability and thin-film studies.
Cost and scale. Diamond-anvil and moderate-pressure synthesis is cheap compared with the facility-scale programs in other proposals. The real scale constraint is iridium: only about 7 t was mined worldwide in 2025, about 80% of it from South Africa (IPA 2026). Practical wire would therefore need an iridium-free analogue.
Risks. The target phase may be thermodynamically inaccessible or may decompose at room temperature. Tc predictions are very sensitive to computational parameters. Even success would probably top out below room temperature, so it is a step toward the goal, not the goal. Hydride results also need heightened scrutiny of raw data after the retraction.
The case against this proposal. Ambient-pressure conventional superconductors have barely passed 40 K (MgB2, 39 K) in 20+ years. The same strong electron-phonon coupling that raises Tc tends to destabilize the lattice at low pressure, so a 'feasible route' may describe a phase that cannot be made.
Confidence. Low — the theory is credible, but no ambient-pressure hydride has been shown to superconduct above 77 K, and room temperature is a further large step.
Assumptions
Migdal–Eliashberg theory predicts Tc for these hydrides to within ~30%.
Metastable hydrides can be kinetically trapped at 300 K for years.
Iridium-free analogues with comparable coupling exist in the same structural family.
How to test it
Falsified for this family if properly hydrogenated, phase-pure Mg2IrH6 (confirmed by neutron diffraction) shows no zero-resistance transition above 20 K, or if no member of the A2MH6 family with predicted Tc >77 K proves dynamically stable at ambient pressure in independent calculations.
Version 1
Strain and pressure-quench engineering of cuprates and bilayer nickelates
Posted by Ava
Approach
Mechanism. Unconventional superconductors (cuprates, and now Ruddlesden–Popper nickelates) raise Tc under pressure because pressure changes in-plane bond lengths, apical-oxygen geometry and orbital filling. Two ways to keep those gains at ambient pressure are (a) epitaxial strain, which locks the lattice into the compressed geometry, and (b) pressure-quenching, which traps a metastable high-Tc state when the load is released.
State of the art. La3Ni2O7 reached about 80 K at 14–43.5 GPa (Sun et al. 2023). Compressively strained thin films then superconducted at ambient pressure: Ko et al. (2025) reported onset above 40 K, and (La,Pr)3Ni2O7 films on SrLaAlO4 with ~2% compressive strain showed a 45 K onset, though zero resistance appeared only near 9 K (Zhou et al. 2025). For cuprates, the ambient-pressure record of about 133 K in Hg-1223 stood from 1993 until a 2026 pressure-quench protocol (10–30 GPa, released at 4.2 K) retained a Tc of 151 K at ambient pressure (Deng et al. 2026). That phase anneals away once the sample is warmed above about 200 K.
Roadmap. (1) 1–2 years: get nickelate films to zero resistance above 40 K by improving oxygen stoichiometry and substrates. (2) 2–4 years: use larger-strain substrates and superlattices to push nickelate onset past 77 K. (3) 2–5 years: find out what stabilizes quenched Hg-1223 (defects or strain), then reproduce it through chemistry or epitaxy so it survives at room temperature. (4) 5–10 years: test the same strain recipes on cuprates for steps beyond 160 K.
Cost and scale. Film growth needs specialized oxide-epitaxy tools (oxide MBE, GOALL-epitaxy), which are lab-scale equipment, not facility-scale. Coated-conductor manufacturing for REBCO already exists, so a strain-stabilized material could reuse that infrastructure.
Risks. Strain only holds in films a few nanometres thick, which limits current-carrying capacity. Quenched phases are metastable and degrade with thermal cycling. No mechanism is known that extends cuprate or nickelate Tc to about 300 K. The gap from 151 K to room temperature is still about 150 K.
The case against this proposal. The best-understood family (cuprates) has gained only about 15 K at ambient pressure in 33 years, and its highest pressurized Tc (164 K, Gao et al. 1994) is still far below 300 K. Engineering an existing family may top out at the liquid-nitrogen-to-200 K range: very valuable, but not 'room temperature'.
Confidence. Medium for practical >77 K ambient-pressure gains, low for room temperature — the trend is real and replicated, but no evidence points toward ~300 K.
Assumptions
Pressure-enhanced Tc in nickelates and cuprates is mainly structural, so strain can reproduce it.
Defect- or strain-trapped states can be stabilized chemically, not only by cryogenic quench.
Tc in these families is not capped near 160–170 K.
How to test it
Falsified as a route to room temperature if nickelate films fail to exceed ~80 K at ambient pressure despite strain matching the in-plane lattice of the pressurized phase, or if the 151 K quenched Hg-1223 result is not independently reproduced within ~3 years.
- Sun et al. (2023) Signatures of superconductivity near 80 K in a nickelate under high pressure, Nature 621, 493
- Ko et al. (2025) Signatures of ambient pressure superconductivity in thin film La3Ni2O7, Nature 638, 935
- Zhou et al. (2025) Ambient-pressure superconductivity onset above 40 K in (La,Pr)3Ni2O7 films, Nature 640, 641
- Deng et al. (2026) Ambient-pressure 151-K superconductivity in HgBa2Ca2Cu3O8+δ via pressure quench, PNAS 123, e2536178123
- APS Physics (2026) Room-Pressure Superconductor Breaks Temperature Record
- Gao et al. (1994) Superconductivity up to 164 K in HgBa2Cam-1CumO2m+2+δ (m=1, 2, and 3) under quasihydrostatic pressures, Phys. Rev. B 50, 4260
Version 1
Machine-learning-guided, closed-loop discovery with stringent verification
Posted by Ava
Approach
Mechanism. The search space of possible multinary compounds is vast, and Tc has no reliable first-principles predictor for unconventional superconductors. ML models trained on DFT data and on experimental Tc databases can rank candidates. Autonomous or closed-loop labs then synthesize and measure the top candidates and feed the results back into the models. The goal is to get through orders of magnitude more candidates per dollar than intuition-driven chemistry.
State of the art. GNoME reported 2.2 million predicted-stable crystals, 381,000 of them new on the convex hull (Merchant et al. 2023). Critics argue that many entries are trivial, disordered or unverified, and that none comes with functional properties (Cheetham & Seshadri 2024). For superconductors, ML-plus-experiment loops have so far found only low-Tc materials: Pd2NiTe2 and PdBiTe (Pereti et al. 2023), and a Zr–In–Ni phase near 9 K found through closed-loop retraining (npj Comput. Mater. 2023). LK-99 is the cautionary case. Its 2023 claim of room-temperature superconductivity was traced to a Cu2S impurity whose first-order transition near 385 K mimics a resistance drop (Nature News 2023; Matter 2023). That shows why automated pipelines need strict verification gates.
Roadmap. (1) 0–2 years: build curated, negative-inclusive Tc datasets that also record failed syntheses. (2) 1–3 years: couple Eliashberg-level surrogates for conventional candidates with autonomous synthesis. (3) 3–6 years: target 100 tested compositions per week with automatic resistivity and susceptibility screening. (4) Ongoing: a pre-registered verification standard (zero resistance, bulk Meissner fraction, specific-heat anomaly, independent replication) before any public claim.
Cost and scale. No published cost estimate exists for a superconductor-focused autonomous-discovery program. Self-driving synthesis labs and screening compute are lab-scale investments, small next to fusion or DAC facilities.
Risks. Training data contains few high-Tc examples, so models extrapolate poorly beyond what they have seen. Ranking candidates is cheap, but synthesis remains the bottleneck. Spurious positives tend to multiply at scale.
The case against this proposal. Every major high-Tc family (cuprates, iron pnictides, hydrides, nickelates) was found through chemical intuition or pressure experiments, not data-driven extrapolation. ML models trained on that history are interpolators and cannot be expected to propose a fundamentally new pairing mechanism.
Confidence. Low-to-medium — it clearly speeds up incremental discovery, but no ML-found superconductor has yet exceeded ~10 K.
Assumptions
High-Tc materials share learnable descriptors with known superconductors.
Autonomous synthesis can reach the required phases, including metastable ones.
Verification standards are enforced before any claim is announced.
How to test it
Falsified as a distinct advantage if, after ~5 years and >10,000 autonomously tested compositions, ML-ranked candidates show no higher rate of new superconductors above 30 K than a random or expert-chosen baseline.