Pith. sign in

REVIEW 4 major objections 3 minor 57 references

Machine learning-accelerated search of superconductors in B-C-N based compounds and R3Ni2O7-type nickelates

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that an ML-plus-DFT pipeline finds 12 new B-C-N superconductors with computed TC at or above 10 K—including B2CN near 45 K—and flags two nickelates, Tb3Ni2O7 and Ac3Ni2O7, as high-pressure high-TC candidates.

desk verdict Useful ML+DFT screening workflow and a few genuinely new ternary nitride candidates, but the headline TC values are uncalibrated, the stability filter is too weak, and the nickelate numbers are ML extrapolations rather than first-principles results. read the letter →

arxiv 2509.03081 v1 pith:NUD74L5Z submitted 2025-09-03 cond-mat.supr-con cond-mat.mtrl-sci

Xiaoying Li , Wenqian Tu , Run Lv , Li'e Liu , Dingfu Shao , Yuping Sun , Wenjian Lu This is my paper
classification cond-mat.supr-concond-mat.mtrl-sci
keywords machinelearningsuperconductivitydensityfunctionaltheoryB-C-Ncompoundsbilayernickelateselectron-phononcouplingLa3Ni2O7highpressure
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a machine-learning model trained on known superconductors, followed by density-functional-theory verification, can find new conventional superconductors in B-C-N chemistry and flag promising nickelate candidates. Screening more than 150,000 materials, it reports 12 binary and ternary B-C-N compounds with computed TC at or above 10 K, including two structural forms of B2CN with TC near 45 K and TiNbN2 at 26.2 K, all sharing strong sigma-bonding bands similar to MgB2. It also predicts that Tb3Ni2O7 and Ac3Ni2O7 could superconduct at about 62 K and 70 K under 30 GPa, based on electronic structures resembling La3Ni2O7, especially the Ni-3dz2 hole band. If these predictions hold, the work would expand the known families of conventional and nickelate superconductors and give practical descriptors for further searches.

What carries the argument

The carrying mechanism is a two-stage pipeline. First, a Gradient Boosting regression model, trained on experimentally known superconductors with 45 features built from elemental properties (mass, radius, valence, electronegativity, ionization energy, shell count, d-electron count) via operations like weighted means, entropies and standard deviations, assigns a TC to every candidate formula. Second, density-functional-theory calculations compute phonon dispersions and the electron-phonon coupling lambda; the Allen-Dynes-modified McMillan equation then converts lambda into TC using a fixed Coulomb pseudopotential mu* = 0.10. For nickelates, the key object is the hole-type gamma band dominated

What would settle it

Measure resistivity and magnetic susceptibility of the R3m phase of B2CN: no superconducting transition near 45 K would directly contradict the headline prediction. Alternatively, recompute the 100 screened candidates with mu* = 0.15; a material shrinkage of the TC_DFT >= 10 K set would show the candidate count is not robust.

Watch

Extended reading notes

Core claim

The central claim is that combining machine learning with DFT-based validation is an effective route to new superconducting materials. From 23,635 metallic, nonmagnetic, thermodynamically stable candidates in a large structure database, the authors select 100 B-C-N binaries and ternaries with ML-predicted TC ≥ 10 K, then compute electron-phonon coupling for each. Twelve survive with Allen-Dynes-modified McMillan TC ≥ 10 K, three above 25 K: B2CN in two phases (44.8 K and 41.5 K) and TiNbN2 (26.2 K). The paper argues their common feature is metallized strong sigma-bonding bands, the same mechanism proposed for MgB2. For nickelates, since DFT cannot give TC in strongly correlated systems, the

Load-bearing premise

The load-bearing premise is that the Allen-Dynes-modified McMillan formula with mu* fixed at 0.10 gives reliable TC from computed electron-phonon coupling; if the coupling is overestimated or the true mu* is higher, several of the 12 candidates, and possibly the TC >= 25 K count, fall below the claimed thresholds.

Editorial extensions

If this is right

  • B2CN in its R3m and P3m1 phases becomes a concrete, synthesizable candidate for conventional superconductivity near 45 K, comparable to MgB2.
  • The 12 predicted B-C-N superconductors broaden the known chemical space of borides, carbides and nitrides, several with simple structures accessible to experiment.
  • Tb3Ni2O7 and Ac3Ni2O7, if synthesized under pressure, would expand the R3Ni2O7 nickelate family beyond La3Ni2O7 with comparable electronic features.
  • The feature-importance analysis gives physically interpretable descriptors (weighted mean mass, valence, ionization energy for conventional; valence and radius dispersion for high-TC) that can guide future searches without full DFT.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The nickelate TC values of 61.6 and 70.3 K are ML estimates, not DFT transition temperatures; the paper's case for them rests on electronic-structure similarity, so the actual numbers should be read as screening priorities rather than quantitative predictions.
  • Because the conventional-superconductor model is trained on historical experimental TC, its screening is biased toward chemistries already represented in the database; genuinely novel pairing mechanisms may be missed.
  • The sigma-bonding criterion, if it holds, could be turned into a direct structural filter—searching for frameworks with head-on p-orbital overlaps at the Fermi level—making future screens cheaper than full electron-phonon calculations.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 3 minor

Summary. The paper combines machine-learning regression with DFT-based electron-phonon coupling calculations to screen for new conventional superconductors among B-C-N binary/ternary compounds and to propose candidate high-TC nickelate superconductors in the R3Ni2O7 family. A Gradient Boosting model trained on SuperCon conventional superconductors predicts TC for 2,249 Materials Project candidates; 100 of these are selected for DFT+EPC computations, yielding 12 compounds with isotropic Allen-Dynes-modified McMillan TC_DFT >= 10 K (e.g., two B2CN structures at 44.8/41.5 K and TiNbN2 at 26.2 K). A second Gradient Boosting model trained on cuprate and nickelate data predicts TC for R3Ni2O7 at 30 GPa, identifying Tb3Ni2O7 (61.6 K) and Ac3Ni2O7 (70.3 K) as local maxima; DFT electronic-structure analysis shows their Ni-3dz2 hole pockets resemble that of La3Ni2O7.

Significance. If the predictions are reliable, the B-C-N candidates would be computationally discovered conventional superconductors with TC values in the MgB2 class, and the nickelate candidates would guide high-pressure experiments in an active field. The paper has several strengths: the ML model is validated against 12 recently discovered superconductors not in training (Table II); the B-C-N TC_DFT values come from independent electron-phonon calculations, not from the ML fit; the orbital-projected band structures provide a physical interpretation in terms of sigma-bonding bands; and the electronic-structure comparison for R3Ni2O7 is careful and useful. However, the headline claims rest on a thermodynamic-stability filter that is weaker than convex-hull stability, an unbenchmarked isotropic McMillan treatment that badly underestimates the MgB2 reference, and ML extrapolations for the nickelates presented without uncertainty. These issues are correctable but require substantive revision before the claims are supported.

major comments (4)
  1. [§3.3 and Abstract] Section 3.3 states that the criterion Eform < 0 eV/atom 'ensures the candidate materials have ... thermodynamic stability.' This is not a convex-hull condition: many compounds with negative formation energy are metastable with respect to decomposition into competing phases. The 12 candidates in Table III, e.g., B2CN (mp-1008525/6), TiNbN2 (mp-35869), CaB2 (mp-1009695), and KB6 (mp-1076), may be off the convex hull and therefore not synthesizable. The abstract's 'identified 12 new ... superconductors' is accordingly too strong. The authors should compute the energy above hull (e_above_hull) for all Table III candidates, report the values, and either restrict the claim to hull-stable compounds or explicitly reclassify the others as predicted metastable phases.
  2. [Table SI, row 13; §2.3] The same Allen-Dynes-modified McMillan equation used for all B-C-N candidates gives TC_DFT = 11.8 K for MgB2 (Table SI row 13), while the experimental TC is 39 K. This factor-of-3 underestimate is not discussed, yet the paper relies on this method to claim B2CN TC = 44.8 K and the count of TC >= 10 K. Without a benchmark against known superconductors (e.g., MgB2, CaB2, NbN) and a discussion of the discrepancy, the absolute TC values of the predicted compounds are uncalibrated. The authors should either employ a more accurate anisotropic Eliashberg treatment for the high-TC candidates or clearly present the TC values as qualitative estimates with a systematic-error caveat.
  3. [§3.1, §3.3 and Abstract] The quoted TC values for Tb3Ni2O7 (61.6 K) and Ac3Ni2O7 (70.3 K) are outputs of the Gradient Boosting regression model fitted to the cuprate/nickelate dataset, not results of a first-principles calculation. The manuscript itself states that 'one cannot directly calculate their TC by the DFT method' and instead uses electronic-structure similarity as support. However, the abstract and conclusions present these numbers without clarifying their ML origin or the absence of uncertainty quantification. Because Tb and Ac are not in the training set, these are extrapolations. The authors should explicitly label the values as TC_ML, report the model's prediction interval or variance, and rephrase the claims to say the ML model proposes these as candidates, with DFT used only to verify the similarity of the electronic structure to La3Ni2O7.
  4. [§2.3, Eq. (Allen-Dynes); Table III] All TC_DFT values are computed with a fixed Coulomb pseudopotential mu* = 0.10, and no sensitivity analysis is provided. The count of 12 candidates with TC_DFT >= 10 K includes borderline cases such as HfZrN2 (10.8 K) and HfTiN2 (9.9 K) in Table III/Table SI. The Allen-Dynes equation is sensitive to mu*, particularly for moderate lambda, and to the numerical accuracy of the computed lambda. The authors should report TC for a range of mu* values (e.g., 0.08, 0.12, 0.16) or otherwise estimate the uncertainty; without this, the precise count of 12 and the specific TC values should not be stated as definitive.
minor comments (3)
  1. [Eq. (3)] The formula for R2 as printed has the same sum in the numerator and denominator, which would make R2 identically zero. The denominator should be sum_i (y_i - ybar)^2. Please correct the typo.
  2. [Table III and main text] The table caption says 'TC_DFT >= 5 K' and the table lists 18 entries, while the text and abstract refer to '12 compounds with TC_DFT >= 10 K.' This inconsistency is confusing; please make the selection criterion and table content match, or explicitly state that the table includes lower-TC entries for completeness.
  3. [Table I] The list of 'eight fundamental atomic attributes' includes 'number of elements,' which is not an atomic property but a compound-level quantity. Consider renaming this feature or moving it to the compound-level descriptor set.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: B-C-N candidates are validated by independent DFT electron-phonon calculations, and the nickelate TC values are transparently presented as ML outputs rather than first-principles results.

full rationale

The paper's central screening chain for B-C-N compounds is not circular. The Gradient Boosting model is used only to prioritize candidates; the final 'identified' list (Table III) is based on TC_DFT computed with the Allen-Dynes-modified McMillan equation from independently calculated electron-phonon coupling in QUANTUM ESPRESSO. This is a separate, parameter-free (apart from standard mu*=0.10) first-principles calculation, so the DFT results do not reduce to the ML training data. The nickelate TC values (Tb3Ni2O7: 61.6 K, Ac3Ni2O7: 70.3 K) are explicitly stated to be ML predictions ('Using the ML-trained model, we predicted the TC...'), and the paper acknowledges 'one cannot directly calculate their TC by the DFT method.' Thus the quantitative nickelate TCs are model outputs, not disguised first-principles derivations; this is a limitation in inferential strength, not a circular derivation. There are no load-bearing self-citations: the references to prior work (e.g., DFT+U parameters, La3Ni2O7 models) are external and not used to define the paper's own predictions. The feature-importance discussion is a post hoc interpretation, not a derivation. Overall, the claimed derivation chain is self-contained: ML suggests, DFT validates for B-C-N; ML predicts and DFT electronic-structure comparison supports qualitatively for nickelates. No step exhibits Eq X = Eq Y by construction or a fitted parameter renamed as an independent prediction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claims rest on a handful of input parameters (mu*, U, J, ML hyperparameters, screening thresholds) and on several domain assumptions about the reliability of ML extrapolation and DFT/EPC calculations. No new physical entities are introduced.

free parameters (5)
  • Coulomb pseudopotential mu* = 0.10
    Used in the Allen-Dynes-modified McMillan equation to compute TC_DFT; a standard but adjustable value that directly affects predicted TC.
  • Hubbard U = 3 eV
    DFT+U parameter for Ni-3d electrons in R3Ni2O7, taken from ref [23]; changes the band structure and hole concentrations.
  • Hund's J = 0.4 eV
    DFT+U parameter from ref [23]; affects the electronic structure and hole pocket analysis.
  • ML model hyperparameters = not specified
    Gradient Boosting, Random Forest, and Gaussian process hyperparameters are not reported; they are fitted during training and affect TC_ML values.
  • Screening thresholds = band gap = 0 eV, m < 0.01 mu_B, Eform < 0 eV/atom
    Criteria in Section 3.3 that select 23,635 from 155,361 Materials Project compounds; changing these thresholds changes the candidate pool.
assumptions (5)
  • domain assumption The Allen-Dynes-modified McMillan equation with mu* = 0.10 yields quantitatively reliable TC for these B-C-N compounds.
    Used in Section 2.3 to convert lambda and omega_log to TC; no error bars or mu* sensitivity analysis is provided.
  • domain assumption DFT+U with U = 3 eV, J = 0.4 eV adequately describes Ni-3d correlations in R3Ni2O7.
    Chosen from ref [23]; affects the band structure and hole concentrations in Section 3.3.
  • domain assumption The Gradient Boosting model trained on SuperCon and cuprate/nickelate data generalizes to out-of-distribution compositions such as Tb3Ni2O7 and Ac3Ni2O7.
    The model is evaluated on a random split and a few recent superconductors, but extrapolation from La-based nickelates to Tb/Ac is assumed in Fig. 7(a).
  • domain assumption Materials Project structures and formation energies (PBE) are reliable for the candidate phases.
    Selection relies on mp-ids with Eform < 0 eV/atom; no crystal structure prediction or experimental synthesis is performed.
  • domain assumption The phonon and electron-phonon coupling calculations are converged and pseudopotential choices do not bias TC_DFT.
    Cutoffs and k-point resolutions are stated, but no convergence tests are shown in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine learning-accelerated search of superconductors in B-C-N based compounds and R3Ni2O7-type nickelates." pith.science (2026). https://pith.science/paper/NUD74L5Z

@misc{pith2026250903081,
  author       = {Pith},
  title        = {Pith review of: Machine learning-accelerated search of superconductors in B-C-N based compounds and R3Ni2O7-type nickelates},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NUD74L5Z}},
  note         = {Machine review of arXiv:2509.03081}
}
read the original abstract

Superconductor research has traditionally depended on experiments and theoretical approaches. However, the rapid advancement of data-driven methods and machine learning (ML) has opened avenues for accelerating superconductor discovery. Here, we integrated ML with density functional theory (DFT) calculations to efficiently screen conventional B-C-N based superconductors and identify potential high-TC candidates among R3Ni2O7-type bilayer nickelates. We identified 12 new binary and ternary B-C-N based superconductors with TC >= 10 K, including 3 with TC >= 25 K, such as two structural forms of B2CN (TC = 44.8 K and 41.5 K) and TiNbN2 (TC = 26.2 K). These materials share a common feature of strong {\sigma}-bonds, which is key to achieving relatively high TC. Moreover, we proposed Tb3Ni2O7 (TC = 61.6 K) and Ac3Ni2O7 (TC = 70.3 K) as potential high-TC nickelate superconductors under high pressure. Their electronic structures closely resemble those of La3Ni2O7, especially in the hole-type band dominated by Ni-3dz2 orbital character. We also analyzed feature importance in the ML results for both conventional and high-TC superconductors. These results advance the search for new superconductors and enhance the fundamental understanding of superconducting mechanisms.

Figures

Figures reproduced from arXiv: 2509.03081 by the authors.

Figure 4
Figure 4. The weighted standard deviation of valence electrons (valence_std_wtd) has the highest weight, indicating that the dispersion of valence electron distributions within compounds plays a dominant role in regulating TC. Other important features include the weighted standard deviation of atomic radius (radius_std_wtd) and the number of unfilled d-orbital electrons (d_unfill). Figures 4(b─d) illustrate how the three key … view at source ↗
Figure 1
Figure 1. [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

57 extracted references · 53 canonical work pages

  1. [1]

    H. K. Onnes, The Resistance of Pure Mercury at Helium Temperatures, Commun. Phys. Lab. Univ. Leiden 12, 1 (1911)

    work page 1911

  2. [2]

    Bardeen, L

    J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Theory of superconductivity, Phys. Rev. 108, 1175 (1957)

    1957

  3. [3]

    Nagamatsu, N

    J. Nagamatsu, N. Nakagawa, T. Muranaka, Y. Zenitani, and J. Akimitsu, Superconductivity at 39 K in magnesium diboride, Nature 410, 63 (2001)

    work page 2001

  4. [4]

    H. L. Sun, M. W. Huo, X. W. Hu, J. Y. Li, Z. J. Liu, Y. F. Han, L. Y. Tang, Z. Q. Mao, P. T. Yang, B. S. Wang, J. G. Cheng, D. X. Yao, G. M. Zhang, and M. Wang, Signatures of superconductivity near 80 K in a nickelate under high pressure , Nature 621, 493 (2023)

    work page 2023

  5. [5]

    Park and R

    C. Park and R. L. Snyder, Structures of High- Temperature Cuprate Superconductors, J. Am. Ceram. Soc. 78, 3171 (1995)

    work page 1995

  6. [6]

    Kitchin, Big data, new epistemologies and paradigm shifts , Big Data Soc

    R. Kitchin, Big data, new epistemologies and paradigm shifts , Big Data Soc. 1, 1 (2014)

    work page 2014

  7. [7]

    G. R. Schleder, A. C. M. Padilha, C. M. Acosta, M. Costa, and A. Fazzio, From DFT to machine learning: r ecent approaches to materials science — a review, J. Phys. Mater. 2, 032001 (2019)

    work page 2019

  8. [8]

    Z. Yu, T. Bo, B. Liu, Z. D. Fu, H. Wang, S. Xu, T. L. Xia, S. L. Li, and S. Meng, Superconductive materials with MgB2-like structures from data-driven screening, Phys. Rev. B 105, 214517 (2022)

    work page 2022

Show all 57 references
  1. [10]

    Gashmard, H

    H. Gashmard, H. Shakeripour, and M. Alaei, Predicting superconducting transition temperature through advanced machine learning and innovative feature engineering, Sci. Rep. 14, 3965 (2024)

    2024

  2. [11]

    M. R. Quinn and T . M. McQueen, Identifying new classes of high temperature superconductors with convolutional neural networks, Front. Electron. Mater . 2, 2022 (2022). 18

    2022

  3. [12]

    Hoffmann, T

    N. Hoffmann, T. F. T. Cerqueira, J. Schmidt, and M. A. L. Marques , Superconductivity in antiperovskites, npj Comput. Mater. 8, 150 (2022)

    2022

  4. [13]

    Hoffmann, T

    N. Hoffmann, T. F. T. Cerqueira, P. Borlido, A. Sanna, J. Schmidt, and M. A. L. Marques, Searching for ductile superconducting Heusler X 2YZ compounds, npj Comput. Mater. 9, 138 (2023)

    2023

  5. [14]

    L. Gu, Y. Liu, P. Chen, H. Y. Huang, N. Chen, Y. Li, T. Lookman, Y. T. Lu, and Y. T. Su, Bond sensitive graph neural networks for predicting high temperature superconductors, MGE Adv. 2, e48 (2024)

    2024

  6. [15]

    B. W. Jiang, X. S. Luo, Y . Sun, X. Zhong, J. Lv, Y. Xie, Y. M. Ma, and H. Y. Liu, Data-driven search for high-temperature superconductors in ternary hydrides under pressure, Phys. Rev. B 111, 054505 (2025)

    2025

  7. [16]

    L. L. Zha, J. J. Jiang, Y . M. Xue, Z. B. Cheng, S. W. Yao, W. J. Hu, L. Peng, T. T. Shi, J. Chen, X. L. Liu, and J. Lin, Stability and superconductivity of hexagonal prism-structured polyhydrides X2MgH18 (X = Li, Na, K, Rb, Cs) under moderate pressure, Phys. Chem. Chem. Phys....

    2025

  8. [17]

    Y . Du, Z. F. Wang, H. Y . Liu, G . J. Liu, and X . Zhong, Superconductivity of electron-doped chalcohydrides under high pressure , Phys. Rev. Res . 7, 013049 (2025)

    2025

  9. [18]

    National Institute of Materials Science, Materials Information Station, SuperCon (2011)

    2011

  10. [19]

    Hamidieh, A data-driven statistical model for predicting the critical temperature of a superconductor, Comput

    K. Hamidieh, A data-driven statistical model for predicting the critical temperature of a superconductor, Comput. Mater. Sci. 154, 346 (2018)

    2018

  11. [20]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996)

    1996

  12. [21]

    P. E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50, 17953 (1994)

    1994

  13. [22]

    Giannozzi, S

    P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Ma...

    2009

  14. [23]

    Z. H. Luo, X. W. Hu, M. Wang, W. Wu, and D. X. Yao, Bilayer Two-Orbital Model 19 of La3Ni2O7 under Pressure, Phys. Rev. Lett. 131, 126001 (2023)

    2023

  15. [24]

    P. B. Allen and R. C. Dynes , Transition temperature of strong- coupled superconductors reanalyzed, Phys. Rev. B 12, 905 (1975)

    1975

  16. [25]

    J. Li, X. P. Cui, Y . Jin, X. Q. Yin, S. X. Cao, Z. J. Feng, and J. C. Zhang, Appearance of superconductivity at 15.9 K in layered V2AlN, J. Supercond. Nov. Magn. 30, 1 (2017)

    2017

  17. [26]

    J. N. Graham, H. Liu, V . Sazgari, C. Mielke, M. Medarde, H. Luetkens, R. Khasanov, Y . Shi, and Z. Guguchia, Microscopic probing of the superconducting and normal state properties of Ta2V3.1Si0.9 by muon spin rotation, Commun. Mater. 5, 225 (2024)

    2024

  18. [27]

    Y . J. Lee, O. Inturu, J. H. Kim, and J. S. Rhyee, Robust bulk superconductivity by giant proximity effect in Weyl semimetal-superconducting NbP/NbSe2 composites, J. Appl. Phys. 132, 123903 (2022)

    2022

  19. [28]

    Ishikawa, U

    H. Ishikawa, U. Wedig, J. Nuss, R. K. Kremer, R. Dinnebier, M. Blankenhorn, M. Pakdaman, Y. Matsumoto, T. Takayama, K. Kitagawa, and H. Takagi, Superconductivity at 4.8 K and violation of Pauli l imit in La2IRu2 comprising Ru honeycomb layer, Inorg. Chem. 58, 19 (2019)

    2019

  20. [29]

    D. Tay, T. Shang, Y. P. Qi, T. P. Ying, H. Hosono, H. R. Ott, and T. Shiroka, s-wave superconductivity in the noncentrosymmetric W 3Al2C superconductor: an NMR study, J. Phys. Condens. Matter 34, 194005 (2022)

    2022

  21. [30]

    K.W. Chen, G. Chappell, S. Zhang, W. Lan, T. Besara, K. Huang, D. Graf, L. Balicas, A. P. Reyes , Superconductivity in single crystals of ZrP 1.27Se0.73, Phys. Rev. B 102, 144522 (2020)

    2020

  22. [31]

    Zheng, X

    B. Zheng, X. K. Feng, B. Liu, Z. F. Liu, S. S. Wang, Y. Zhang, X. Ma, Y. Luo, C. L. Wang, R. M. Li, Z. Y. Zhang, S. T. Cui, Y. L. Lu, Z. Sun, J. F. He, S. Y. A. Yang, and B. Xiang, The coexistence of superconductivity and topological order in Van der Waals InNbS2, Small 20, 5 (2024)

    2024

  23. [32]

    Z. H. Yang, Z. Yang, Q. P. Su, E. D. Fang, J. H. Yang, B. Chen, H. D. Wang, J. H. Du, C. X. Wu, and M. H. Fang, Superconductivity in TlBi2 with a large Kadowaki- Woods ratio, Phys. Rev. B 106, 224501 (2022)

    2022

  24. [33]

    Shang, D

    T. Shang, D. Tay, H. Su, and T. Shiroka , Evidence of fully gapped superconductivity in NbReSi: A combined μ SR and NMR study, Phys. Rev. B 105, 144506 (2022)

    2022

  25. [34]

    Q. Q. Zhu, G. R. Xiao, W. Z. Yang, S. J. Song, G. H. Cao, and Z. Ren, Mo3ReRuC: 20 A Noncentrosymmetric superconductor formed in the MoReRu–Mo 2C System , Inorg. Chem. 61, 43 (2022)

    2022

  26. [35]

    Y . Liu, J. Li, W. Z. Yang, J. Y. Lu, B. Y. Cao, H. X. Li, W. L. Chai, S. Q. Wu, B. Z. Li, Y . L. Sun, W . H. Jiao, C . Wang, X . F. Xu, Z . Ren, and G . H. Cao, Superconductivity in Kagome metal ThRu3Si2, Chin. Phys. B 33, 057401 (2024)

    2024

  27. [36]

    Olsen, Superconductivity of Lead Isotopes, Nature 168, 245 (1951)

    M. Olsen, Superconductivity of Lead Isotopes, Nature 168, 245 (1951)

    1951

  28. [37]

    M. Pop, G . Borodi, and S. Simon, Correlation between valence electron concentration and high-temperature superconductivity, J. Phys. Chem. Solids 61, 12 (2000)

    2000

  29. [38]

    B. T. Matthias, Empirical relation between s uperconductivity and the number of valence electrons per atom, Phys. Rev. 97, 74 (1955)

    1955

  30. [39]

    Stanev, C

    V. Stanev, C. Oses, A. G. Kusne, E. Rodriguez, J. Paglione, S. Curtarolo, and I. Takeuchi, Machine learning modeling of superconducting critical temperature, npj Comput. Mater. 4, 29 (2018)

    2018

  31. [40]

    S. P. Ong, S. Cholia, A. Jain, M. Brafman, D. Gunter, G. Ceder, and K. A. Persson, The Materials Application Programming Interface (API): A simple, flexible , and efficient API for materials data based on REpresentational State Transfer (REST) principles, Comput. Mater. Sci. 9...

    2015

  32. [41]

    X. X. Zhang, G. L. Yu, H. Chen, Y. Zhao, T. M. Cheng, Q. Li, Electron deficiency but semiconductive diamond-like B2CN originated from three-center bonds, Phys. Chem. Chem. Phys. 23, 4 (2021)

    2021

  33. [42]

    N. I. A. Ahmed and C. Parlak, First-principles calculations on mechanical and electronical properties of AlB 2-type CaB 2, Mater. Today Commun. 38, 108290 (2024)

    2024

  34. [43]

    H. J. Choi, S. G. Louie, and M. L. Cohen, Prediction of superconducting properties of CaB2 using anisotropic Eliashberg theory, Phys. Rev. B 80, 064503 (2009)

    2009

  35. [44]

    M. Gao, Z. Y . Lu, and T. Xiang, Finding high- temperature superconductors by metallizing the σ-bonding electrons, Physics, 44, 421 (2015)

    2015

  36. [45]

    J. M. An and W. E. Pickett , Superconductivity of MgB2: Covalent Bonds Driven Metallic, Phys. Rev. Lett. 86, 4366 (2011)

    2011

  37. [46]

    Y . Kong, O. V . Dolgov, O. Jepsen, and O. K. Andersen, Electron-phonon interaction in the normal and superconducting states of MgB 2, Phys. Rev. B 64, 020501(R) (2011)

    2011

  38. [47]

    Geisler, J

    B. Geisler, J. J. Hamlin, G. R. Stewart, R. G. Hennig, and P. J. Hirschfeld, 21 Structural transitions, octahedral rotations, and electronic properties of A 3Ni2O7 rare-earth nickelates under high pressure, npj Quantum Mater. 9, 38 (2024)

    2024

  39. [48]

    S. Q. Wu, Z. H. Yang, X. Ma, J. H. Dai, M. Shi, H. Q. Yuan, H. Q. Lin, and C. Cao, Ac3Ni2O7 and La 2AeNi2O6F ( Ae=Sr, Ba): Benchmark Materials for Bilayer Nickelate Superconductivity, arXiv:2403.11713

    arXiv

  40. [49]

    L. C. Rhodes and P. Wahl, Structural routes to stabilize superconducting La3Ni2O7 at ambient pressure, Phys. Rev. Mater. 8, 044801 (2024)

    2024

  41. [50]

    P. Li, G. D. Zhou, W. Lv, Y . Y . Li, C.M. Yue, H. L. Huang, L. Z. Xu, J. C. Shen, Y . Miao, W. H. Song, Z. H. Nie, Y . Q. Chen, H. Wang, W. Q. Chen, Y . B. Huang, Z. H. Chen, T. Qian, J. H. Lin, J. F. He, Y . J. Sun, Z. Y . Chen, and Q. K. Xue, Angle- resolved photoemission s...

    2025

  42. [51]

    Z. Luo, X. Hu, M. Wang, W. Wu, and D. X. Yao, Bilayer two-orbital model of La3Ni2O7 under pressure. Phys. Rev. Lett. 131, 126001 (2023)

    2023

  43. [52]

    Q. G. Yang, D. Wang, and Q. H. Wang, Possible s ± wave superconductivity in La3Ni2O7, Phys. Rev. B 108, L140505 (2023)

    2023

  44. [53]

    Y. H. Gu, C. C. Le, Z. S. Yang, X. X. Wu, and J. P. Hu, Effective model and pairing tendency in the bilayer Ni -based superconductor La 3Ni2O7, Phys. Rev. B 111, 174506 (2025)

    2025

  45. [54]

    Z. Y. Shao, Y. B. Liu, M. Liu, and F. Yang, Band structure and pairing nature of La3Ni2O7 thin film at ambient pressure, Phys. Rev. B 112, 024506 (2025). SUPPLEMENTAL MATERIAL Machine learning-accelerated search of superconductors in B-C-N based compounds and R3Ni2O7-type nick...

    2025

  46. [55]

    S1 and Fig

    Machine learning for TC prediction: supplementary figures of model metrics and comparative benchmarks We conduct a series of analyses, including model metrics, as shown in Fig. S1 and Fig. S2, clearly showing the performance differences of different models. Benchmarking charts...

  47. [56]

    There are total 100 compounds with TC ML ≥ 10 K

    Superconducting candidates predicted via the Gradient Boosting regression model We illustrate the relevant information of superconducting candidates predicted via the Gradient Boosting regression model. There are total 100 compounds with TC ML ≥ 10 K. We have also conducted DF...

  48. [57]

    Ⅲ of the main text

    Phonon spectra of compounds listed in Table III of the main text We calculate the phonon spectra of all the compounds listed in Table. Ⅲ of the main text. Each figure contains the Materials Project ID number (mp-id), electron-phonon coupling strength ( λqv), and the Eliashberg...

  49. [58]

    S8 and Figs

    Electronic structures of R3Ni2O7 We calculated the electronic structures of R3Ni2O7, including orbital-projected band structures and density of states under 0 GPa (Figs. S8 and Figs. S9) and 30 GPa (Figs. S10 and Figs. S11), and Fermi surfaces under 0 and 30 GPa (Figs. S12 and...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.