betaFeedback

WorksPhysics and Astronomy

Predictions of superconducting properties from chemical bonding interactions

Abstract

The need for rapid estimates of superconducting critical temperatures, Tc s, is becoming increasingly urgent as first-principles and artificial intelligence based methods enable the exploration of ever-larger and more chemically complex materials spaces. In this context, chemical bonding based quantities have emerged as promising descriptors for predicting superconductivity. To date, these approaches have mainly been developed for hydride superconductors, notably through the networking value, a descriptor based upon the electron localization function, whose topological properties have been found to correlate with Tc . In this perspective, we review and consolidate the method's theory, and we discuss how the estimate of the Tc is achievable within a real-space view. We conclude by proposing a framework that enables Tc prediction for the broad family of conventional superconductors from easily computable intermediate physical quantities.

Introduction

Since its discovery in 1911 [1], superconductivity has fascinated physicists, chemists, and materials scientists, as evidenced by five Nobel prizes (the first awarded to Onnes in 1913), tens-of-thousands patents filed, and intense ongoing research [2, 3, 4, 5, 6, 7]. The potential rewards for finding “better” superconductors are immense due to these materials real-world technological uses: as wires conducting electricity without energy loss, powerful and compact magnets in wind turbines and in medical imaging, high precision magnetic measuring devices, and as electronic circuit qubits for quantum computing. While “better” is often associated with higher superconducting critical temperatures, TcT_\mathrm{c}s, other properties, including ductility [8] and malleability, ease of synthesis, and ability to sustain a high superconducting critical current or magnetic field [9], can be key factors driving industrial adoption. The discovery of an easy-to-process room-temperature superconductor would change our lives in ways that cannot be imagined.

Following Onnes’ discovery, superconductivity remained an enigma. Early experimental efforts identified only a few simple superconductors, all with relatively low TcT_\mathrm{c}s, such as Pb (7.2 K) [10], Nb (9.2 K), Nb3_3Sn (18.3 K) [11], and NbTi (10 K) [12]. A giant leap forward occurred in 1986 with the discovery of cuprate superconductors [13], a group of perovskite-type compounds that contain CuO2_2 layers. Perhaps the most famous example is YBa2_2Cu3_3O7−x_{7-x} (YBCO), the first compound to possess a TcT_\mathrm{c} surpassing the boiling point of liquid nitrogen (77 K) [14]. Prior to the discovery of this class of so-called high-temperature superconductors (HTSC), expensive cryogens were required to induce the superconducting state. The highest-TcT_\mathrm{c} cuprate known is HgBa2_2Ca2_2Cu3_3O8_8, with TcT_\mathrm{c} = 133 K (164 K under 31 GPa of pressure [15], and 151 K after pressure quench to ambient [16]). As time passed, other classes of superconductors with interesting physics were discovered, but it took ∼30\sim30 years until the TcT_\mathrm{c} record was broken again, first by H3_3S [17] and then by LaH10_{10} [18, 19], with measured TcT_\mathrm{c}s of 200 K at 150 GPa and 250-260 K near 180 GPa, respectively, en route towards room temperature. However, the brittleness of the cuprates and the high pressures required to stabilize the hydrides hinder their wide scale adoption within applications. Examples of the idealized atomistic structures of some superconducting compounds are shown in the left hand side of Figure 1.

Figure 1. Examples of cuprate, iron based, and heavy fermion (unconventional) superconductors are shown under the cyan, orange and green headers, while examples of conventional electron-phonon (e-ph) mediated systems are shown under the red header. The e-ph mechanism is illustrated under the pink header: atomic vibrations (phonons) generate a displacement-potential that perturbs the electron density, which, in turn, induces a modification of the band structure near the Fermi energy (EFermiE_{\mathrm{Fermi}}) on the scale of the phonon energy (EphE_{\mathrm{ph}}). A sketch of the mechanism resulting in the formation of Cooper pairs in conventional superconductors is shown under the blue header; an electron moving through the lattice creates a local distortion, producing a charge imbalance that attracts a second electron. An example of a potential energy surface (PES) is provided under the purple header, highlighting candidates that could be obtained in a crystal structure prediction (CSP) search. In a CSP run, structures are generated and locally optimized until the lowest enthalpy structure (blue star) is found multiple times.

While these experimental advances unfolded, the theoretical side saw the development of the Nobel-prize winning Bardeen-Cooper-Schrieffer (BCS) theory [20], which provided the first plausible microscopic mechanism for superconductivity. Within BCS theory the superconducting state is associated with the pairing of electrons (with opposite momentum, and usually opposite spin) via an attractive phonon-mediated interaction capable of overcoming their mutual Coulomb repulsion (Figure 1, top right). The resulting Cooper pairs behave as bosons; they are therefore able to condense into the same ground quantum state. This framework was later formalized in Migdal-Eliashberg theory [21, 22, 23] by incorporating the Green’s function approach developed by Migdal for the electron-phonon (e-ph) interaction in the normal metallic state. Theories deriving from BCS, however, were limited to e-ph mediated (or conventional) superconductivity. They are therefore not able to predict the properties of unconventional superconductors (examples of which are illustrated in Figure 1 top left) including cuprates [24, 25], nickelates [26], as well as iron-based [27] and heavy-fermion [28] systems, whose superconducting mechanisms are still under debate.

The development of Migdal-Eliashberg theory represented an important paradigm shift: for the first time, it enabled the prediction of the superconducting critical temperature directly from first-principles calculations. Once it became possible to calculate TcT_\mathrm{c}, experimental results could be used as benchmarks to refine the theory and establish a more solid foundation. These advances paved the way for the investigation of subtle effects that influence superconducting properties including spin-orbit coupling and/or spin fluctuations, which are important for superconductivity in some elemental metals [29, 30], and the influence of configurational entropy on the TcT_{\mathrm{c}} of disordered systems and alloys [31]. Theoretical studies have yielded a deeper understanding of the dynamic stabilization, structural renormalization, and anomalous trends in TcT_{\mathrm{c}} (including the inverse isotope effect in PdH [32]), for various ambient and high-pressure hydrides, induced by quantum-nuclear-driven anharmonicity [33, 34]. Furthermore, advances in computational methods have enabled investigations of subtle phenomena such as: anisotropic contributions to the superconducting gap (crucial in layered materials, such as MgB2_2 [35]), full-bandwidth description (important for systems with van Hove singularities near the Fermi energy, such as H3_3S [36]), and the addition of higher-order terms to the e-ph vertex expansion (important for hydrides [37, 38, 39]). All these developments have resulted in an incredibly precise computational theory able to estimate TcT_{\mathrm{c}} within a few Kelvin of experiments [37].

A second advance, important for materials prediction, has been the development of crystal structure prediction (CSP) methods [40, 41]. In the last two decades algorithms that can be used to computationally find the (putative) thermodynamically stable, and important metastable structures for a given chemical composition have become widely adopted. Focusing on

inorganic systems, and in particular those under high pressure, some of the most well-known CSP methods include random structure searches (e.g. the AIRSS code [42]), particle swarm methods (e.g. CALYPSO [43]), evolutionary algorithms or EAs (e.g. USPEX [44] or XtalOpt [45]), along with metadynamics [46] and minima hopping [47] algorithms. Though these methods vary in how structures are created or evolved, all aim to explore the low-lying, and therefore chemically sensible, regions of the potential energy surface (PES) of a user defined chemical combination (as illustrated schematically in Figure 1, bottom right).

A number of the aforementioned CSP methods have been further refined towards the discovery of materials with specific functionalities [48]. In standard EAs, which aim to find the global minimum in the PES, thermodynamic quantities are used to determine a structure’s fitness, which is related to its likelihood to be chosen for procreation. Multi-objective EA schemes, on the other hand, allow for the incorporation of additional criteria including physical properties (e.g. hardness [49]), electronic structure (e.g. degree of electride character [50]) or structural information (e.g. similarity with an experimental X-ray diffraction pattern [51] or desired crystalline features [52]) within the search. These properties can be employed to remove structures from the gene pool (filtration [52]), be used within a generalized scalar function [53] that merges all objectives to obtain a single measure of fitness for each structure or within the well-known Pareto optimization scheme [54], where optimization proceeds along a multidimensional front of optimal trade-offs. Instead of minimizing a single variable (e.g., energy), a Pareto search explores the best compromise among multiple relevant properties with the Pareto front representing the optimal hypersurface balancing all desired quantities.

CSP methods have been widely used in the search for novel inorganic superconducting materials, both to identify the lowest enthalpy structures for a given composition, and to locate all those that comprise the convex hull of thermodynamic stability (focusing on hydrides examples include Refs. [55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65]). To mitigate the large cost associated with ab initio CSP searches, approaches employing machine learning (ML) interatomic potentials for filtering of the initial identified set of structures, followed by DFT relaxation of the most promising species, have been put forward [66, 67]. ML generative models trained on high pressure phases [68], data-driven techniques [69], as well as workflows for high-throughput screenings that focus on structural templates [70, 71], have recently been developed.

The search for superconducting materials via CSP-based multi-objective optimization schemes, which not only optimize thermodynamic stability, but also maximize superconducting properties either by considering features enhancing TcT_c, such as specific geometric motifs or increased density of states (DOS) at the Fermi level, or by embedding the calculation of TcT_c into the multi-objective scheme itself, are the emerging frontier. A quick and reliable way to estimate TcT_c remains the major computational bottleneck towards achieving this goal. While the cost of CSP has been drastically reduced through the aforementioned techniques, calculating the TcT_c across a large number of candidate structures is still prohibitively demanding.

As a result, significant effort is being devoted to find ways to accelerate TcT_c calculations, for example through simplified Muffin-tin approximations for the potential [72], or via ML models. With regards to ML, deep neural-network models have been developed and employed to screen structures found in public databases for superconductivity, and to predict the magnitude of their TcT_c [73, 74, 75, 76, 77, 78, 79, 80, 81, 82]. Alternative screening approaches have employed ML models trained on the Eliashberg spectral functions of low temperature superconductors [83], while other studies have derived trends or descriptors based on the analysis of properties that are easy to compute or obtain, such as stoichiometry, electronegativity, atomic mass, composition, space group, volume, DOS, and interatomic distances [69, 77, 84].

An alternative path towards estimating TcT_c – and the one we are concerned with here – makes use of methods that can quantify chemical bonding in the solid state. In conventional superconductors the onset of superconductivity is governed by the e-ph interaction, which is related to the vibration-induced change in the local overlap of electronic orbitals comprising the occupied conduction bands. For systems with appreciable TcT_cs, this overlap is often highly localized, suggesting that real-space bonding descriptors encompassing the first few neighboring atoms could capture the essential ingredients of the (Cooper) pairing interaction.

Tools for the analysis of chemical bonding and electronic structure such as the Electron Localization Function (ELF) [85], bond valence functions [86], and bonding analysis tools such as Crystal Orbital Overlap or Hamilton Populations (COOP [87] or COHP [88]), or the Crystal Orbital Bond Index (COBI [89]) can be employed to quantify the local orbital overlap. The values associated with these descriptors represent intermediate physical quantities that could be used for TcT_c prediction, and they offer direct information about the bond strengths and related force constants – information that can help determine the characteristic vibrational energy scale [90] entering the e-ph coupling. Together these quantities estimate both the electronic and vibrational contributions controlling the magnitude of TcT_c.

In what follows we describe how TcT_c can be estimated in certain classes of superconductors starting from a real-space perspective, thereby providing a theoretical justification for the recently developed models that estimate TcT_c using chemical bonding concepts [91, 92, 93]. Furthermore, we illustrate how bonding analysis descriptors such as the ELF and COHP can be connected to e-ph theory for conventional superconductivity, and propose ways to employ this approach systematically to accelerate the discovery of new superconductors in conjunction with multi-objective CSP methodologies.

Mathematical Formalism

Superconductivity arises from the formation of correlated electron pairs extending over length scales ranging from a few, to thousands of nanometers. At first glance, such nonlocal interactions may appear challenging to characterize. Remarkably, TcT_c can be expressed in terms of effective, Fermi-surface-averaged quantities calculated for the primitive unit cell. Advanced theoretical approaches have been developed for estimating TcT_c [35]. However, the most widely used expression is the semi-empirical Allen-Dynes modified McMillan (ADM) formula [94]:

kBTc=ℏωln⁡1.2exp⁡[−1.04(1+λ)λ−μ∗(1+0.62λ)],(1)k_{\mathrm{B}}T_c = \frac{\hbar\omega_{\ln}}{1.2}\exp\left[\frac{-1.04(1+\lambda)}{\lambda-\mu^*(1+0.62\lambda)}\right], \tag*{(1)}

where kBk_{\mathrm{B}} is the Boltzmann constant, and μ∗\mu^*, λ\lambda and ωln⁡\omega_{\ln} represent the screened single electron repulsive Coulomb interaction, the attractive e-ph coupling constant, and the logarithmic average phonon frequency, respectively, given as

μ∗=NF⟨V⟩1+NF⟨V⟩ln⁡(εFωph),(2)\mu^* = \frac{N_{\mathrm{F}}\langle V\rangle}{1+N_{\mathrm{F}}\langle V\rangle\ln\left(\frac{\varepsilon_{\mathrm{F}}}{\omega_{\mathrm{ph}}}\right)}, \tag*{(2)}
λ=∫0∞α2F(ω)ω dω,(3)\lambda= \int_{0}^{\infty}\frac{\alpha^2F(\omega)}{\omega}\,d\omega, \tag*{(3)}

and,

ωln⁡=exp⁡[2λ∫0∞α2F(ω)ωln⁡(ω) dω].(4)\omega_{\ln}=\exp\left[\frac{2}{\lambda}\int_{0}^{\infty}\frac{\alpha^{2}F(\omega)}{\omega}\ln(\omega)\,d\omega\right]. \tag*{(4)}

In these equations, NFN_{\mathrm{F}} represents the electronic DOS at the Fermi level, εF\varepsilon_{\mathrm{F}}, ⟨V⟩\langle V\rangle is the average of the Coulomb matrix elements on the Fermi surface, and ωph\omega_{\mathrm{ph}} is the characteristic phonon frequency of the system. Equation (1) provides a good estimate of TcT_{\mathrm{c}} in the limit of λ≤1\lambda\le1. For larger values of λ\lambda this equation underestimates TcT_{\mathrm{c}} as it lacks the strong-coupling and shape corrections included in the less intuitive but more advanced Allen-Dynes approach [94]. As with many semi-empirical schemes, the strength of the ADM formulation stems from the ability to interpret the system-specific variables, λ\lambda, ωln⁡\omega_{\ln}, and (to a lesser extent) μ∗\mu^{*}, in terms of physical properties of the compound.

In Equations (3) and (4) the quantity α2F(ω)\alpha^{2}F(\omega) represents the Eliashberg spectral function, which describes the scattering of electronic states near the Fermi energy induced by atomic vibrations and takes the form

α2F(ω)=1NFNqNk∑kq,nmν∣gnmν(k,q)∣2×\alpha^{2}F(\omega)=\frac{1}{N_{\mathrm{F}}N_{\mathrm{q}}N_{\mathrm{k}}}\sum_{\mathbf{kq},nm\nu}\left|g_{nm\nu}(\mathbf{k},\mathbf{q})\right|^{2}\times
δ(εnk−εF)δ(εmk+q−εF)δ(ℏω−ℏωqν).(5)\delta(\varepsilon_{n\mathbf{k}}-\varepsilon_{\mathrm{F}})\delta(\varepsilon_{m\mathbf{k+q}}-\varepsilon_{\mathrm{F}})\delta(\hbar\omega-\hbar\omega_{\mathbf{q}\nu}). \tag*{(5)}

Here

gnmν(k,q)=⟨ψkn∣ΔqνVKS∣ψk+qm⟩(6)g_{nm\nu}(\mathbf{k},\mathbf{q})=\langle\psi_{\mathbf{k}n}|\Delta_{\mathbf{q}\nu}V^{\mathrm{KS}}|\psi_{\mathbf{k+q}m}\rangle \tag*{(6)}

corresponds to the e-ph matrix elements between a state ∣ψkn⟩|\psi_{\mathbf{k}n}\rangle with crystalline momentum k\mathbf{k}, band index nn and energy εnk\varepsilon_{n\mathbf{k}}, and a state ∣ψk+qm⟩|\psi_{\mathbf{k+q}m}\rangle with crystalline momentum k+q\mathbf{k}+\mathbf{q}, band index mm and energy εnk\varepsilon_{n\mathbf{k}}. ΔqνVKS\Delta_{\mathbf{q}\nu}V^{\mathrm{KS}} is the derivative of the Kohn-Sham potential with respect to an atomic displacement induced by the phonon mode (q,ν)(\mathbf{q},\nu) with frequency ωqν\omega_{\mathbf{q}\nu} given as

ΔqνVKS=1Np∑aγ,pℏ2Maωqνϵaγ,ν(q)eiq⋅RpΔVaγ,RpKS.(7)\Delta_{\mathbf{q}\nu}V^{\mathrm{KS}}=\frac{1}{N_{\mathrm{p}}}\sum_{a\gamma,p}\sqrt{\frac{\hbar}{2M_{a}\omega_{\mathbf{q}\nu}}}\epsilon_{a\gamma,\nu}(\mathbf{q})e^{i\mathbf{q}\cdot\mathbf{R}_{p}}\Delta V^{\mathrm{KS}}_{a\gamma,\mathbf{R}_{p}}. \tag*{(7)}

In Equation (7) MaM_{a} is the mass of atom aa, Rp\mathbf{R}_{p} is the lattice vector of the pp-th unit cell, ΔVaγ,RpKS\Delta V^{\mathrm{KS}}_{a\gamma,\mathbf{R}_{p}} is the derivative of the Kohn-Sham potential with respect to the displacement of atom aa along direction γ\gamma in cell Rp\mathbf{R}_{p}, and ϵaγ,ν(q)\epsilon_{a\gamma,\nu}(\mathbf{q}) is the phonon eigenvector. Furthermore, NkN_{\mathrm{k}}, NqN_{\mathrm{q}} and NpN_{\mathrm{p}} represent the number of points in the k\mathbf{k}-mesh and q\mathbf{q}-mesh used in the integrations, and the number of unit cells, respectively.

At first glance, Equation (5) may appear complicated, but it can be understood in terms of Fermi’s golden rule, which describes the rates of scattering between quantum states induced by a weak perturbation, averaged over the Fermi surface. Specifically, it describes transitions between electronic states, ψkn→ψk+qm\psi_{\mathbf{k}n}\rightarrow\psi_{\mathbf{k+q}m}, induced by the phonon-mediated variation of the Kohn-Sham potential, ΔqνVKS\Delta_{\mathbf{q}\nu}V^{\mathrm{KS}}. This perturbation arises from the atomic displacement associated with phonon mode (q,ν)(\mathbf{q},\nu) given by the normalized eigenvector ϵaγ,ν(q)\epsilon_{a\gamma,\nu}(\mathbf{q}) with amplitude ℏ2Maωqν\sqrt{\frac{\hbar}{2M_{a}\omega_{\mathbf{q}\nu}}}.

Real Space Interpretation

A commonly used approach to evaluate the e-ph matrix elements entering Equation (5) relies on a real-space formulation based on localized functions, obtained by performing a Wannier transformation [95] of the plane-wave Bloch states as:

∣ϕn,Rp⟩=1Nk∑n′ke−ik⋅RpUnn′,k∣ψn′k⟩.(8)|\phi_{n,\mathbf{R}_{p}}\rangle=\frac{1}{N_{\mathrm{k}}}\sum_{n'\mathbf{k}}e^{-i\mathbf{k}\cdot\mathbf{R}_{p}}U_{nn',\mathbf{k}}|\psi_{n'\mathbf{k}}\rangle. \tag*{(8)}

Here, Unn′,kU_{nn',\mathbf{k}} is a k-dependent unitary transformation, and ϕn,Rp\phi_{n,\mathbf{R}_{p}} is a localized function centered at cell Rp\mathbf{R}_{p}. The index nn labels the Wannier function, encoding both its atomic site and orbital character. Using this transformation, along with Equations (6) and (7), the e-ph matrix elements in reciprocal space can be expressed as:

gnmν(k,q)=∑Rp,Rp′∑n′m′aγei(k⋅Rp+q⋅Rp′)ℏ2Maωqνϵaγ,ν(q)×Un′n,k+qgn′m′aγ(Rp,Rp′)Um′m,k†.(9)\begin{aligned} g_{nm\nu}(\mathbf{k},\mathbf{q})=\sum_{\mathbf{R}_{p},\mathbf{R}_{p'}}\sum_{n'm'a\gamma}e^{i(\mathbf{k}\cdot\mathbf{R}_{p}+\mathbf{q}\cdot\mathbf{R}_{p'})}\sqrt{\frac{\hbar}{2M_{a}\omega_{\mathbf{q}\nu}}}\epsilon_{a\gamma,\nu}(\mathbf{q}) \\ &\times U_{n'n,\mathbf{k+q}}g_{n'm'a\gamma}(\mathbf{R}_{p},\mathbf{R}_{p'})U^{\dagger}_{m'm,\mathbf{k}}. \tag*{(9)} \end{aligned}

where

gn′m′aγ(Rp,Rp′)=⟨ϕn′,0∣ΔVaγ,Rp′KS∣ϕm′,Rp⟩(10)g_{n'm'a\gamma}(\mathbf{R}_{p},\mathbf{R}_{p'})=\langle\phi_{n',0}|\Delta V^{\mathrm{KS}}_{a\gamma,\mathbf{R}_{p'}}|\phi_{m',\mathbf{R}_{p}}\rangle \tag*{(10)}

are the real-space counterparts of the e-ph matrix elements. A sketch of the quantities entering Equation (10) is shown in Figure 2, providing an intuitive real-space picture of the e-ph interaction. Here, gn′m′aγ(Rp,Rp′)g_{n'm'a\gamma}(\mathbf{R}_{p},\mathbf{R}_{p'}) is the matrix element that describes the coupling between two localized electronic states, ϕm′,Rp\phi_{m',\mathbf{R}_{p}} and ϕn′,R0\phi_{n',\mathbf{R}_{0}}, induced by a lattice displacement associated with a particular phonon mode. The change in the Kohn-Sham potential, ΔVaγ,Rp′KS\Delta V^{\mathrm{KS}}_{a\gamma,\mathbf{R}_{p'}}, resulting from this vibration mediates the coupling between these two electronic states, and the strength of this coupling determines the probability of scattering between them via Fermi’s golden rule.

Schematic representation of the quantities entering Equation 10 for the calculation of the real space electron-phonon matrix elements

Figure 2. Schematic representation of the quantities entering Equation (10) for the calculation of the real space electron-phonon (e-ph) matrix elements on a 2-dimensional plane. The black lines describe the e-ph scattering processes from ϕm′,Rp\phi_{m',\mathbf{R}_p} to ϕn′,R0\phi_{n',\mathbf{R}_0} induced by the atomic displacement ΔVαγ,Rp′KS\Delta V^{\mathrm{KS}}_{\alpha\gamma,\mathbf{R}_{p'}}. The localized (red-yellow-green-blue) functions describe the electronic states in the proximity of atomic sites (red dots), while the red and blue functions describe the perturbation induced in the potential due to an atomic displacement.

A key advantage of the Wannier formulation lies in the spatial localization of the basis functions. For Equation (10) to be nonzero, the two Wannier functions ϕn′,0\phi_{n',0} and ϕm′,Rp\phi_{m',\mathbf{R}_{p}} must have a non-negligible spatial overlap. In practice, the function ϕ\phi decays rapidly with distance [96], implying that contributions are significant when the orbitals involved in the scattering and potential variation are located within a few neighboring atomic sites. This locality implies that the e-ph interaction is short-ranged in real space.

Equation (10) can be interpreted as an overlap between atom or bond-centered localized orbitals, modulated by a potential variation induced by an atomic displacement. However, Wannier functions are not unique and their functional forms are determined by the choice of the unitary transformation, Unn′,kU_{nn',\mathbf{k}}. Maximally localized Wannier functions (MLWFs) proposed by Marzari and Vanderbilt [97] are among the most commonly used. In MLWFs Unn′,kU_{nn',\mathbf{k}} is chosen to minimize a cost

E-Ph Matrix Elements

gnm′αγ(Rp,Rp′)=⟨ϕn,R0∣ΔVαγ,Rp′KS∣ϕm′,Rp⟩g_{nm'\alpha\gamma}(\mathbf{R}_p,\mathbf{R}_{p'})=\langle\phi_{n,\mathbf{R}_0}\mid\Delta V^{\mathrm{KS}}_{\alpha\gamma,\mathbf{R}_{p'}}\mid\phi_{m',\mathbf{R}_p}\rangle

function measuring the orbital spread, equivalent to the Foster-Boys criterion from quantum chemistry [98]. Alternatively, Wannier functions may be generated by minimizing weighted sums of both the spatial and energy variance [99], projecting onto localized atomic orbitals [100], or via symmetrical orthonormalization of order-NN Muffin-Tin Orbitals (NMTOs), which do not rely on the plane wave orbital projection [101]. The resulting shape and extent of localization depends upon the method used; nevertheless an important requirement is that the function retains most of the spatial weight in regions characterized by significant electron density [97].

The conventional superconductor with the highest TcT_c at 1 atmosphere (39 K) is MgB2_2 (whose unit cell is shown in Figure 1). This compound can be formally described as Mg2+^{2+}(B−^-)2_2, adopting a layered structure analogous to it’s isoelectronic cousin, graphite. The boron atoms in MgB2_2 form a honeycomb lattice with magnesium atoms located above and below the center of the hexagons. Analogous to graphite, each boron atom’s in-plane pxp_x- and pyp_y-orbitals hybridize with the ss-orbital forming a set of three σ\sigma-bands, in addition to a π\pi-band from the out-of-plane pzp_z-orbital [102]. However, unlike graphite, MgB2_2 has an intricate Fermi surface because the electrostatic potential from Mg2+^{2+} lowers the energy of the π\pi-bands relative to the σ\sigma-bands, such that both cross the Fermi level. As a result both bands contribute towards

superconductivity, rendering MgB2_2 a prototypical example of a multiband superconductor, with the stronger e-ph coupling arising from the B-B stretching mode within the honeycomb layers. The MLWFs associated with the σ\sigma-bonds within this layer (one of which is illustrated in Figure 3) resemble the localized σ\sigma molecular orbitals computed for benzene [103]. Contour plots of the valence electron density, and ELF of MgB2_2 (also shown in Figure 3) clearly possess features connecting the B-B bonds. Thus, plots of the MLWF, electron density and ELF describing the B-B σ\sigma-bonds within MgB2_2, look qualitatively similar, and one could say that they are topologically consistent. The e-ph interaction associated with the vibrations of the σ\sigma-bonds, λσ\lambda_\sigma, is larger than λπ\lambda_\pi.

Contour plots and/or isosurfaces for MgB2 and fcc-lithium

Figure 3. Contour plots and/or isosurfaces for a characteristic maximally localized Wannier function or MLWF (red header), valence electron density (blue header), ELF (yellow header), and potential variation (purple header) for MgB2 (top) and fcc-lithium at 35 GPa (bottom). An isosurface of one of the MLWFs computed for lithium is shown in the inset. Notice the topological resemblance of the illustrated MLWF, electron density and ELF computed for MgB2. In lithium, on the other hand, the region of space where the contour values for the MLWF are highest do not correspond to regions that possess large values of the electron density or ELF.

Near 7.5 GPa and at low temperature lithium’s body-centered cubic (bcc) phase transitions to a face-centered-cubic (fcc) structure, whose atom-projected electronic DOS is characterized by spsp hybridization. Pressure induces superconductivity within lithium, with the TcT_c of the fcc phase computed to be 13 K at 35 GPa, in good agreement with experiment [104]. Despite its nearly-free-electron-like DOS, quantum mechanical calculations reveal that fcc Li possesses non-nuclear maxima (NNM) found within octahedral interstitial sites [104, 105]. In fact, at this pressure fcc lithium is an electride – a compound where the electrons, which are localized in interstitial regions, act as anions [106]. Because generating Wannier Functions that span only the occupied bands in a metal can require the construction of large supercells that result in band back-folding [101], we generated MLWFs that spanned the four lowest energy bands, extending up to 6 eV above the Fermi level, by placing ss-type orbitals on the lithium atoms as initial guesses. Four qualitatively similar MLWFs were obtained; one is illustrated in Figure 3.

The generated MLWFs is overall more delocalized but posses strong localization on the lithium atoms with additional character in the interstitial regions where the NNM are found. In fact, they resemble the NNM-centered WF (generated by the NMTO method) for another high-pressure-electride, Na-hP4hP4, which was interpreted as originating from the overlap of Na-centered spdspd hybrids resulting in the formation of weak multi-centered bonds [107]. A similar interpretation is likely relevant for fcc-Li, where the interstitial electron localization originates from the overlap of Li spsp hybrids within multi-centered bonds. Comparison of the MLWFs obtained for Li with plots of the electron density and the ELF all show similarities, though low isosurfaces in the density are required to observe the contributions of the interstitial regions, which are much more pronounced in plots of the ELF that coincide with the location of the NNM. Despite the low electron polarization in the NNM, the density in these regions is strongly localized, resulting in high ELF values at the NNM sites.

The different topology of the MLWFs in MgB2_2 and Li are noteworthy and provide insights on the type of bonding interactions featured in these systems. In MgB2_2, the Wannier functions are characteristic of localized bonding σ\sigma orbitals [103]. In contrast, the MLWFs spanning the valence and some unoccupied states in lithium are significantly more delocalized, reflecting the metallic nature of the system.

Site overlap and networking

Equation 10 includes the derivative of the Kohn-Sham potential induced by an atomic displacement. The magnitude, shape, and spatial extent of this potential variation is determined by the underlying electronic structure and bonding topology of the system. The right hand panel of Figure 3 illustrates ΔVaγ,Rp′KS\Delta V^{\mathrm{KS}}_{a\gamma,\mathbf{R}_{p'}} calculated for the displacement of a single atom along a Cartesian direction chosen so as to maximize the overlap of the derivative of the potential with regions of the electron density arising from states near the Fermi level.

In MgB2_2 the σ\sigma-bands, which can be described by the bond-centered MLWF in Figure 3, form hole-pockets resulting in two nearly cylindrical Fermi surfaces extending along the Γ−A\Gamma- A direction. The potential derivative in this system decays slowly, retaining non-negligible values in the MLWF centers, reflecting the localized and directional nature of the covalent bonding and the large interaction between the potential and the electronic states.

In the spsp metal lithium, on the other hand, the states near the Fermi level correspond to the maxima in the electron density located in the octahedral interstitial regions characterised by high ELF values that are not well described by the generated MLWFs (Figure 3). In contrast to the results obtained for MgB2_2, the perturbation decays rapidly within a distance comparable to the core radius, consistent with lithium’s nearly free-electron character. This fast decay is evident in Figure 3, where the bulk of the potential variation is localized around the lithium atom, exhibiting minimal overlap with the electronic density residing in the octahedral interstitial regions.

The previous discussion illustrates that covalently bonded metals are characterized by factors favoring strong e-ph coupling in conventional superconductors: electronic states near the Fermi energy that are well-described by atom- and/or bond-centered Wannier functions, and that overlap with regions of significant potential variation, thereby maximizing the e-ph coupling matrix elements described in Equation (10). These directional bonding/antibonding states are highly sensitive to atomic displacements, so even small vibrations generate strong variations in the electronic structure and, consequently, large e-ph matrix elements. This phenomenon, first discussed by Cohen in Ref. [108], is well documented for hydrogen-rich superconductors where covalently bonded networks, strengthened by pressure and electron donation, are responsible for values of TcT_c close to room temperature [91, 92], in stark contrast to the TcT_c achievable in simple spsp-metals and transition metals.

Because the e-ph interaction can be viewed as a local bonding-driven phenomenon, descriptors quantifying the strength and spatial extent of chemical bonding should be able to provide effective proxies for the magnitude of the dominant e-ph matrix elements, without requiring an explicit evaluation of the potential derivative. Potentially suitable descriptors include bond valence functions, the integrated COHP and COBI (ICOHP and ICOBI), and the ELF, which characterize the strength of chemical bonding or the degree of local electron sharing.

In systems with strongly localized covalent bonds, the topology of the ELF attractors identify regions of localized bonding electrons, and the critical ELF value along the pathway connecting neighboring attractors provides a measure of the local orbital overlap between the bonded atoms. The “networking value” has been defined as the largest ELF isovalue, ΦNV\Phi_{\mathrm{NV}}, for which the corresponding isosurface forms a three-dimensional network spanning the crystal (Figure 4). For hydride superconductors it has been shown that combining ΦNV\Phi_{\mathrm{NV}} with the hydrogen orbital-derived density of states at the Fermi level, HDOSH_{\mathrm{DOS}}, and the hydrogen atomic fraction, HfracH_{\mathrm{frac}}, can provide an estimate of the superconducting critical temperature via

Figure 4. A schematic illustration of the procedure used to obtain the networking value, ΦNV\Phi_{\mathrm{NV}}, for I4/mmmI4/mmm YH4YH_{4}. First the electron localization function (ELF) attractors are identified (left), next the value of the ELF is lowered until the basins connect and form an extended 3-dimensional network throughout the crystal (middle). The explicit connections between the ELF attractors that form the network are shown in the last panel (right).

Tc=750ΦNVHfracHDOS1/3−85,(11)T_{\mathrm{c}} = 750\Phi_{\mathrm{NV}}H_{\mathrm{frac}}H_{\mathrm{DOS}}^{1/3} - 85, \tag*{(11)}

with a root-mean-square error of ∼60\sim60 K [91]. The predictive power of this method is illustrated for a set of hydride superconductors in Figure 5(a). Although the quantitative agreement is imperfect, the results demonstrate that chemically motivated descriptors derived from the ELF can capture the dominant chemical factors governing TcT_{\mathrm{c}}.

Figure 5. (a) The prefactor of the slope in Equation (11) obtained by calculating the networking value (ΦNV\Phi_{\mathrm{NV}}), the projection of the DOS onto the hydrogen orbitals at the Fermi energy (HDOSH_{\mathrm{DOS}}) and the fraction of hydrogen atoms (HfracH_{\mathrm{frac}}) comprising the compound. (b) The correlation of the TcT_{\mathrm{c}} with the integral of the Eliashberg spectral function, illustrating the relationship suggested by Equation (12). Both panels include data for AlH3_3 (150 GPa) [115], PdH (ambient) [32], YH6_6 (150 GPa) [116], YCaH8_8 (150 GPa) [117], H-I41_1/amd (atomic, 500 GPa) [118], H-Cmca-4 (molecular, 450 GPa) [119], LaH10_{10} (250 GPa) [120], PtH (100 GPa) [121], ScH6_6-P63_3/mmc (140 GPa) [122], Mg2_2IrH6_6 (ambient) [114], and H3_3S-Im3ˉ\bar{3}m (150 GPa) [37, 123]. The horizontal line in panel (a) between the blue AlH3_3 points illustrates variations in the prefactor arising from the choice of density functional.

Improvements to the original fit [91] have been proposed by refitting using extended datasets [92], development of automated methods for determining the networking value [109], introduction of the molecularity index to account for the presence of H2H_{2} units [93], and incorporation of a descriptor that includes the derivatives of the KS potential with respect to the displacement of each atom along the Cartesian directions [110]. However, it has recently been shown that (some of) the revised fits are still unable to resolve large TcT_{\mathrm{c}} differences within a given family of multi-metal hydrides [111].

Although the predictive accuracy of the aforementioned fits could potentially be improved by introducing further chemically-inspired parameters, the intrinsic limitations of ELF-based methods must be noted. For example, the ELF is constructed from all occupied electronic states, not just those near the Fermi level. Secondly, the current approaches assume that TcT_{\mathrm{c}} is predominantly driven by hydrogen-derived states; while this assumption is sensible the inability of the RS5 [92] fit to predict the 150 K variations in TcT_{\mathrm{c}} for different YCaH12YCaH_{12} structures suggests that it is insufficient [111]. Thirdly, as in the case of electride fcc-lithium, the magnitude of the ELF attractors does not reflect the amount of electron density found within the NNM basins. Furthermore, the planewave based codes typically employed for extended systems calculate the atom projected DOS by associating a spherical volume to each atom, and the results of the projection are sensitive to the size of the sphere(s) employed, especially for systems with interstitial voids [112] and molecular motifs (such as in Mg2IrH6Mg_{2}IrH_{6} [55, 113, 114]). In addition, the ELF-based approach remains largely phenomenological and lacks a rigorous theoretical foundation.

In what follows, we derive a real-space framework that provides the missing microscopic justification for ELF-based descriptors, explains both their successes and failures, and enables the construction of more accurate chemically motivated descriptors based on quantities such as the ICOHP and ICOBI.

Theoretical framework.

A natural starting point for developing a formula that can estimate TcT_{\mathrm{c}} from chemical-bonding descriptors is the Leavens and Carbotte expression

Tc=0.148∫0∞α2F(ω) dω=0.148 α2Fˉ.(12)T_{\mathrm{c}} = 0.148 \int_{0}^{\infty} \alpha^{2}F(\omega)\,d\omega= 0.148\ \alpha^{2}\bar{F}. \tag*{(12)}

This expression is based on the assumption that, in the strong-coupling regime (λ>0.6\lambda> 0.6), the e–ph interaction is larger than the opposing Coulomb repulsion. Under these conditions, TcT_{\mathrm{c}} is determined primarily by the e–ph interaction strength.

Equation (12) was originally fit using the properties of low temperature superconductors, but the proposed parametrization still holds for known high temperature conventional superconductors [124]. Furthermore, recent studies have shown that this relationship is more general, holding across a broad range of critical temperatures [34, 124]. From calculations on a set of hydride superconductors a near-linear relationship between the integral of the Eliashberg spectral function and TcT_{\mathrm{c}} is evident (Figure 5b). Although the correlation between the two quantities is not perfect, it is sufficiently strong to allow the Eliashberg-spectral function to serve as a proxy for TcT_{\mathrm{c}}.

With this framework in mind, the explicit form of the spectral function of Equation (5) can be factored by separating the contributions associated with each displaced atom, corresponding to the sum over aa in Equation (9), and subsequently integrating over the Brillouin zone,

α2F∝NF∑ab⟨gagb⟩MaMb⟨ω⟩.(13)\alpha^2 F \propto N_{\mathrm{F}} \sum_{ab} \frac{\langle g_a g_b\rangle}{\sqrt{M_a M_b \langle\omega\rangle}}. \tag*{(13)}

Here gag_a and gbg_b denote the e–ph matrix elements associated with displacements of atoms aa and bb, averaged over the Fermi surface, while MaM_a and MbM_b are the corresponding atomic masses. In deriving Equation (13), it is assumed that the phonon frequencies associated with the different atomic displacements can be represented by a single average phonon frequency, ⟨ω⟩\langle\omega\rangle. This expression, formally derived for the strong-coupling regime, makes explicit the quantities governing conventional superconductivity: the electronic density of states at the Fermi level, NFN_{\mathrm{F}}, the atomic masses, the characteristic phonon frequency, and, most importantly, the averaged e–ph matrix elements ⟨gagb⟩\langle g_a g_b\rangle.

Equation (13) also provides a theoretical justification for the networking value approach previously proposed for hydrogen-rich superconductors. As discussed above, it is reasonable to expect that, in these materials, a substantial fraction of the Eliashberg spectral function arises from phonon modes dominated by hydrogen motion. Equation (13) can therefore be approximated by retaining the terms due to hydrogen,

α2F∝HDOS⟨gH2⟩Hfrac2MH⟨ω⟩.(14)\alpha^2 F \propto H_{\mathrm{DOS}} \frac{\langle g_{\mathrm{H}}^2\rangle H_{\mathrm{frac}}^2}{M_{\mathrm{H}}\langle\omega\rangle}. \tag*{(14)}

Here, the factor Hfrac2H_{\mathrm{frac}}^2 accounts for the fraction of atoms contributing to the hydrogen-dominated e–ph coupling, ensuring the correct scaling of the Eliashberg spectral function with system size and composition. Since the hydrogen mass is constant, MHM_{\mathrm{H}} does not influence trends across different hydrides, while the average phonon frequency ⟨ω⟩\langle\omega\rangle is expected to vary only moderately among chemically and structurally similar compounds. Consequently, the dominant quantities controlling the variation of α2F\alpha^2 F are the hydrogen-projected electronic density of states at the Fermi level, the hydrogen e–ph matrix elements, and the hydrogen fraction.

In this model, the networking value serves as a proxy for ⟨gH2⟩\langle g_{\mathrm{H}}^2\rangle, as it reflects the degree of local orbital overlap involving hydrogen atoms. Moreover, because stronger covalent bonding generally leads to both larger orbital overlap and stiffer bonds, the networking value is also expected to correlate with the characteristic phonon frequency. Equation (14) captures the essential ingredients already present in Equation (11), with the latter differing primarily in that the exponents have been empirically optimized to achieve the best quantitative agreement with the TcT_{\mathrm{c}}s computed using DFT.

Generalization of the Method

Equation (13) suggests that this approach can be generalized to all conventional superconductors, provided the atom-resolved e–ph matrix elements and the characteristic phonon frequency can be estimated. A particularly useful quantity for these estimates is the (projected) crystal orbital Hamilton population (pCOHP) [88]. To calculate this quantity, the wavefunctions obtained from DFT computations are projected onto a localized atomic basis of Slater-type orbitals (ϕ\phi), with the projection coefficients defined as

Tnm(k)=⟨ψkn∣ϕm⟩(15)T_{nm}(\mathbf{k}) = \langle\psi_{\mathbf{k}n} \mid\phi_m \rangle \tag*{(15)}

Using these projections, the Hamiltonian in the local orbital basis is constructed as

Hmm′(k)=∑nεnkTnm∗(k)Tnm′(k).(16)H_{mm'}(\mathbf{k}) = \sum_n \varepsilon_{n\mathbf{k}} T_{nm}^{*}(\mathbf{k}) T_{nm'}(\mathbf{k}). \tag*{(16)}

The pCOHP is then obtained by projecting the Hamiltonian onto the density matrix and integrating over reciprocal space,

pCOHP⁡mm′(ε)=∑kn[Tnm∗(k)Tnm′(k)Hmm′′(k)]×δ(εnk−ε);(17)\operatorname{pCOHP}_{mm'}(\varepsilon) = \sum_{kn} \left[ T_{nm}^{*}(\mathbf{k}) T_{nm'}(\mathbf{k}) H_{mm'}'(\mathbf{k}) \right] \times\delta(\varepsilon_{nk} - \varepsilon); \tag*{(17)}

it can further be summed over all orbitals belonging to a specific pair of atoms to derive a bond resolved pCOHP

pCOHP⁡ab(ε)=∑m∈a∑m′∈bpCOHP⁡mm′(ε).(18)\operatorname{pCOHP}_{ab}(\varepsilon) = \sum_{m \in a} \sum_{m' \in b} \operatorname{pCOHP}_{mm'}(\varepsilon). \tag*{(18)}

Since the pCOHP is energy resolved, it can be used to provide a direct measure of the bonding (or antibonding) interactions between specific pairs of atoms in a chosen energy window, including the vicinity of the Fermi level where the e-ph interaction is determined.

A recent study reported a correlation between the largest (most negative) value for the integrated pCOHP (ICOHP) for a pair of atoms, corresponding to the strongest bond in a compound, and the crystal’s highest phonon frequency [90]. This chemically intuitive trend suggests that a COHP-based descriptor can serve not only as a measure of the e-ph matrix elements, but also as a proxy for the characteristic phonon frequencies entering Equation (13). Importantly, the pCOHP can be obtained from a single static DFT calculation, requiring only a small fraction of the computational cost of a full phonon calculation, or an even more expensive e-ph coupling calculation.

To develop a fit for TcT_{\mathrm{c}}, considering its direct dependence on the Eliashberg spectral function (Equation (12)), the bond resolved pCOHP can be used to define a volume-normalized mean absolute interaction index for an atom as:

Ka=a03V∑b∫ε1εF∣pCOHP⁡ab(ε)∣ dε2(εF−ε1).(19)K_a = \frac{a_0^3}{V} \frac{\sum_b \int_{\varepsilon_1}^{\varepsilon_{\mathrm{F}}} \left| \operatorname{pCOHP}_{ab}(\varepsilon) \right| \, \mathrm{d}\varepsilon}{2(\varepsilon_{\mathrm{F}}-\varepsilon_1)}. \tag*{(19)}

In this equation, a0a_0 is the Bohr radius (introduced to render KaK_a dimensionless), ε1\varepsilon_1 is the lower bound for the integration window below the Fermi energy εF\varepsilon_{\mathrm{F}}, and VV is the volume of the cell (introduced to normalize the sum of KaK_a over all atoms within the cell). The factor of 12\frac{1}{2} is introduced to prevent double counting, and the absolute value for the pCOHP is employed to take into account both the bonding and antibonding states, which prevents them from cancelling each other, since they have different signs. The choice of ε1\varepsilon_1 is not arbitrary. If chosen to be too close to εF\varepsilon_F, KaK_a will become artificially small as ∣pCOHPab(ε)∣|p\mathrm{COHP}_{ab}(\varepsilon)| decreases due to the sign inversion of the pCOHP that typically occurs in the proximity of εF\varepsilon_F. Conversely, choosing a too large value of ε1\varepsilon_1 results in the inclusion of contributions from electronic states that may not be relevant for superconductivity. In the tests performed here, the optimal choice for ε1\varepsilon_1 was found to be on the order of about 0.5 eV.

KaK_a is used to construct two approximants for the quantities entering Equation (13), which when combined with Equation (12), provide a direct measure of TcT_c in the strong coupling regime. The first is the average phonon frequency weighted by the density of states near the Fermi level

⟨ω⟩NF∼∑aKa(∑s∈speciesM0Ms)1/2,(20)\langle\omega\rangle_{N_F} \sim\sum_a K_a \left(\sum_{s\in\mathrm{species}} \frac{M_0}{M_s}\right)^{1/2}, \tag*{(20)}

where the sum of KaK_a is over each atom in the unit cell, M0M_0 is the mass of the hydrogen atom introduced to retain the adimensionality, and where the sum over ss runs over each distinct chemical element, or species, present in the cell. The second quantity provides an estimate for the e-ph matrix elements, considering their dependence on the atomic orbital overlap

NF2∑ab⟨gagb⟩MaMb∼(∑aKaM0Ma)2.(21)N_F^2 \sum_{ab}\frac{\langle g_a g_b\rangle}{\sqrt{M_a M_b}} \sim\left(\sum_a K_a\sqrt{\frac{M_0}{M_a}}\right)^2. \tag*{(21)}

By substituting Equations (20) and (21) into Equation (13) we propose an estimator for α2F\alpha^2F, denoted as η\eta,

η=(∑aKaM0Ma)2(∑aKa)−1(∑sM0Ms)−1/2,(22)\eta= \left(\sum_a K_a\sqrt{\frac{M_0}{M_a}}\right)^2\left(\sum_a K_a\right)^{-1}\left(\sum_s\frac{M_0}{M_s}\right)^{-1/2}, \tag*{(22)}

which can, considering Equation (12), be used to predict TcT_c in the case of strong coupling.

To determine the predictive power of the estimator η\eta, we plot its value for a set of conventional superconductors versus TcT_c in Figure 6. These results were obtained by either setting ε1\varepsilon_1 to 0.5 eV, or to the energy of the bottom of the conduction band (if it was within 0.5 eV of εF\varepsilon_F). This dataset included a number of elemental metals namely Nb, Li-fcc (35 GPa), Li-bcc, Cu, Ir, V, Tc, Sn, Al, and Pb (green squares) [30]; the hydrides listed in Figure 5 (orange triangles); and a set of ambient pressure conventional superconductors that contain covalent bonds consisting of MgB2_2 [125], NbN [126], Nb3_3Ge [127], KPbB6_6C6_6 [128], SrB6_6C6_6 [128], NaC4_4 [129], and YB6_6 [130] (blue circles). The dotted black line represents the best fit of the form

Values of $T_{\mathrm{c}}$ with respect to $\eta$ for hydrides, general multi-element compounds, atomic metals, and a test set

Figure 6. Values of TcT_{\mathrm{c}} (a) with respect to the parameter (η\eta) proposed in Equation (22), and (b) a magnified view of this plot in the range from 0-60 K. The correlation is reported for hydrides (yellow triangles), general multi-element compounds (blue circles), atomic metals (green squares), and a test set not used to develop the fit (red stars). The dotted black line represents the best fit obtained through the equation Tc=a/(eb/η−1)T_{\mathrm{c}} = a/(e^{b/\eta} - 1).

Tc=aeb/η−1,(23)T_c = \frac{a}{e^{b/\eta}-1}, \tag*{(23)}

constructed by ensuring the exponential decay for small values of η\eta, and linear behavior for high values of η\eta, with a=13.1587a = 13.1587 and b=0.00057578b = 0.00057578, resulting in a coefficient of determination of R2=0.957R^2 = 0.957, and root mean square error of 20 K. The prefactor obtained here, though universal across all conventional superconductors, is expected to be sensitive to the choice of the pseudopotential and exchange correlation functional used. The TcT_c data originated from a variety of sources, including both theory and experiment, and therefore a perfect correlation is not to be expected. A subsequent set of experimentally synthesized superconductors consisting of Nb3_3Al [105], V3_3Si [131], V3_3Ge [132], TiN [133], VN [134], LuB12_{12} [135] and ZrB12_{12} [136], was chosen as testset for the validation of the fit, resulting in a root mean square error of 12 K.

As anticipated, and based upon the discussion in the previous sections, the model proposed here is able to provide a sensible value of TcT_c for most of the conventional superconductors we considered, with only V and AlH3_3 as outliers. For vanadium, standard DFT calculations result in a TcT_c of ∼\sim22 K, while spin-orbit and spin-fluctuations play an important role in bringing this prediction close to its experimental value of 6 K [30]. Because these effects are not included in our analysis of the electronic structure, the η\eta parameter we obtain yields a TcT_c that is closer to the PBE value. For AlH3_3 standard DFT calculations using the PBE exchange-correlation functional result in TcT_cs that are about one order of magnitude higher than those obtained including anharmonicity, and are closer to those predicted by our η\eta parameter. It should however be noted that more advanced DFT calculations predict the opening of a gap in AlH3_3 at 150 GPa [137], and experimentally no superconductivity was found above 4 K in AlH3_3 at this pressure [138].

Because η\eta is quick to compute, we imagine it will become useful as an intermediate physical quantity to screen for superconductivity, or as a secondary objective to be used in Pareto-type CSP searches. Beyond its predictive capability, the descriptor η\eta (Equation (22)) provides a physical perspective on the factors that fundamentally constrain the superconducting transition temperature. In the present formulation, TcT_c is solely related to the strength of the interatomic interactions associated with the electronic states near the Fermi energy, and the magnitude of the atomic masses. The strength of the interatomic interactions is, in turn, governed by the chemistry of the constituent atoms, which therefore limits the maximum achievable TcT_c for a compound that contains a particular set of bonds. Pressure, on the other hand, provides a mechanism for modifying these interactions by reducing some interatomic distances thereby enhancing their orbital overlap and hardening the phonon modes with which they are associated, weakening other interactions and softening phonons, and finally by bringing certain electronic states to the Fermi level. Pressure can therefore increase both the characteristic phonon energy scale and the e-ph interaction strength. From this perspective, the pressure dependence of the proposed descriptor provides information not only on the expected variation of TcT_c, but also on the extent to which the underlying interactions can be engineered.

It has recently been shown that the contribution arising from the potential variation of the e-ph matrix elements plays an important role in suppressing the magnitude of the orbital overlap and restoring the local e-ph properties [110]. Such contributions can in general not be neglected, and it is possible that further refinements of our model that take into account this potential variation could further improve TcT_c prediction.

Conclusions.

This perspective investigates the relationship between chemical bonding and the physics governing conventional superconductivity. The magnitude of the superconducting critical temperature, TcT_c, is governed by the total integral of the Eliashberg spectral function, which we show is related to the magnitude of the local bonding strength between atoms integrated in a window around the Fermi energy. Previous formulations of this idea were based on heuristic application of the electron localization function as a means to capture the electron-phonon coupling in hydrides.

Here, we justify why chemical-bonding based descriptors can be used as intermediate physical quantities for the prediction of TcT_c, and extend this methodology to all conventional superconductors. Specifically, we develop a descriptor that is based upon the integrated crystal orbital Hamilton population (COHP), which is chemically intuitive and computationally efficient, and therefore extremely useful for high throughput or crystal structure prediction searches of superconducting systems. It remains to be seen if this methodology can be extended to superconductors where the pairing mechanism is unclear. As a matter of fact, similar approaches for the description of unconventional superconductors based on the real space overlap of orbitals have been proposed [139], but they are not yet capable of providing concrete TcT_c estimates.

In conclusion, the method presented here constitutes a powerful and versatile tool for the investigation of superconductivity in a conventional setting. Its accuracy, efficiency, and chemical interpretation make it suitable for practical applications, with the possibility that developments will broaden its predictive capabilities and enable its application to a wider range of superconducting materials.

Methods

DFT calculations were performed with the Vienna ab-initio Simulation Package (VASP) version 6.4.2 [140, 141]. Projector Augmented-Wave (PAW) potentials [72] were employed together with the Perdew-Burke-Ernzerhof (PBE) [142] exchange-correlation functional. The employed PAW electronic valence configurations were Al 3s23p13s^{2}3p^{1}, B 2s22p12s^{2}2p^{1}, C 2s22p22s^{2}2p^{2}, Ca 3s23p64s23s^{2}3p^{6}4s^{2}, Cu 3d104s13d^{10}4s^{1}, Ge 4s24p24s^{2}4p^{2}, H 1s12p01s^{1}2p^{0}, Ir 5s16d85s^{1}6d^{8}, K 3s23p64s13s^{2}3p^{6}4s^{1}, La 5s26s25p65d15s^{2}6s^{2}5p^{6}5d^{1}, Lu 4f145s26s25p65d14f^{14}5s^{2}6s^{2}5p^{6}5d^{1}, Mg 2p63s22p^{6}3s^{2}, Mo 4p65s04d04p^{6}5s^{0}4d^{0}, N 2s22p32s^{2}2p^{3}, Na 3s13p03s^{1}3p^{0}, S 3s23p43s^{2}3p^{4}, Sc 3p64s13d03p^{6}4s^{1}3d^{0}, Sn 5s25p25s^{2}5p^{2}, Sr 4s24p65s24s^{2}4p^{6}5s^{2}, Tc 4p65s24p^{6}5s^{2}, Ti 3d34s23d^{3}4s^{2}, V 3p63d44s13p^{6}3d^{4}4s^{1}, Y 4s24p65s24d24s^{2}4p^{6}5s^{2}4d^{2}, and Zr 4s25s24p64d34s^{2}5s^{2}4p^{6}4d^{3}. Converged parameters for the calculations considered an energy cutoff of 600 eV, a k-point grid density of 2π×0.0252\pi\times0.025 Å−1^{-1}, and a first-order Methfessel-Paxton smearing of 0.2 eV. The energy resolved bond projected COHP was obtained through the Local Orbital Basis Suite Towards Electronic-Structure Reconstruction (LOBSTER) version 5.5.1 [144] using the tetrahedron method for the mesh construction and the vasp2015 basis/projector set. The obtained charge spilling was found to be greater than 4% but lower than 10% for Ir, LaH10_{10}, Li-bcc, MgIrH6_{6}, Pb, Sn,V3_{3}Ge,V3_{3}Si and V, but less than 4% for the rest of the systems studied.

Acknowledgements

We thank Trevor Higgins and Chris Walsh for their insightful suggestions and benchmarking of the pCOHP descriptor. We further acknowledge the National Science Foundation (Grant No. DMR-2136038) and the U.S. Department of Energy, Office of Science, Fusion Energy Sciences, funding the award entitled “High Energy Density Quantum Matter,” Award No. DE-SC0020340 for financial support. Calculations were performed at the Center for Computational Research at SUNY Buffalo (https://hdl.handle.net/10477/79221).

References

References

References

  1. [1]Onnes, K. H. T. The resistance of pure mercury at helium temperatures. Commun. Phys. Lab Univ Leiden B 12 (1911).
  2. [2]Flores-Livas, J. A. et al. A perspective on conventional high-temperature superconductors at high pressure: Methods and materials. Phys. Rep. 856, 1–78 (2020).
  3. [3]Hilleke, K. P. & Zurek, E. Tuning chemical precompression: Theoretical design and crystal chemistry of novel hydrides in the quest for warm and light superconductivity at ambient pressures. J. Appl. Phys. 131, 070901 (1–19) (2022). URL https://doi.org/10.1063/5.0078478.
  4. [4]Zeng, L., Li, K., Li, Chen, R. & Luo, H. Recent advances in high-entropy superconductors. npg Asia Materials 16, 60 (2024).DOI
  5. [5]Yao, C. & Ma, Y. Superconducting materials: Challenges and opportunities for large-scale applications. Iscience 24 (2021).DOI
  6. [6]Wang, Y. et al. Recent progress in nickelate superconductors. national Science Review 12, nwaf373 (2025).DOI
  7. [7]Prasankumar, R. P. et al. The path to room-temperature superconductivity: A programmatic approach. Proc. Natl. Acad. Sci. U.S.A. 123, e2520324123 (2026). URL https://doi.org/10.1073/pnas.2520324123.
  8. [8]Zhang, Y. & Poncé, S. In search of novel ductile superconductors. arXiv preprint arXiv:608.04789 (2026). URL https://arxiv.org/abs/2608.04789.
  9. [9]Geisler, B. et al. Towards the discovery of high critical magnetic field superconductors. npj Comput. Mater. (2026). URL https://doi.org/10.1038/s41524-026-02266-9. In press.
  10. [10]Boorse, H. A., Cook, D. B. & Zemansky, M. W. Superconductivity of lead. Phys. Rev. 78, 635–636 (1950). URL https://link.aps.org/doi/10.1103/PhysRev.78.635.DOI
  11. [11]Matthias, B. T., Geballe, T. H., Geller, S. & Corenzwit, E. Superconductivity of Nb₃Sn. Phys. Rev. 95, 1435–1435 (1954). URL https://link.aps.org/doi/10.1103/PhysRev.95.1435.
  12. [12]Scanlan, R. M., Malozemoff, A. P. & Larbalestier, D. C. Superconducting materials for large scale applications. Proc. IEEE 92, 1639–1654 (2004).
  13. [13]Bednorz, J. G. & Müller, K. A. Possible high t_c superconductivity in the ba-la-cu-o system. Z. Phys. B - Cond. Mat. 64, 189–193 (1986).DOI
  14. [14]Wu, M. K. et al. Superconductivity at 93 k in a new mixed-phase y-ba-cu-o compound system at ambient pressure. Phys. Rev. Lett. 58, 908–910 (1987).
  15. [15]Gao, L. et al. Superconductivity up to 164 k in hgba₂cam−1cumO₂m+2+δ (m=1, 2, and 3) under quasihydrostatic pressures. Phys. Rev. B 50, 4260–4263 (1994).
  16. [16]Deng, L., Habamahoro, T., Safezოდdeh, A. & Chu, C. W. Ambient-pressure 151 k superconductivity in HgBa₂Ca₂Cu₃O₈+δ via pressure quench. Proc. Natl. Acad. Sci. U.S.A. 123, e2536178123 (2026). URL https://doi.org/10.1073/pnas.2536178123.
  17. [17]Drozdov, A. P., Eremets, M. I., Troyan, I. A., Ksenofontov, V. & Shylin, S. I. Conventional superconductivity at 203 kelvin at high pressures in the sulfur hydride system. Nature 525, 73–76 (2015).
  18. [18]Somayazulu, M. et al. Evidence for superconductivity above 260 k in lanthanum superhydride at megabar pressures. Phys. Rev. Lett. 122, 027001 (2019). URL https://link.aps.org/doi/10.1103/PhysRevLett.122.027001.
  19. [19]Drozdov, A. et al. Superconductivity at 250 k in lanthanum hydride under high pressures. Nature 569, 528–531 (2019).DOI
  20. [20]Bardeen, J., Cooper, L. N. & Schrieffer, J. R. Theory of superconductivity. Phys. Rev. 108, 1175–1204 (1957). URL https://link.aps.org/doi/10.1103/PhysRev.108.1175.DOI
  21. [21]Migdal, A. Interaction between electrons and lattice vibrations in a normal metal. Sov. Phys. JETP 7, 996–1001 (1958).
  22. [22]Eliashberg, G. Interactions between electrons and lattice vibrations in a superconductor. Sov. Phys. JETP 11, 696–702 (1960).
  23. [23]Eliashberg, G. Temperature green’s function for electrons in a superconductor. Sov. Phys. JETP 12, 1000–1002 (1961).
  24. [24]Taillefer, L. Scattering and pairing in cuprate superconductors. Annu. Rev. Condens. Matter Phys. 1, 51–70 (2010). URL https://www.annualreviews.org/content/journals/10.1146/annurev-conmatphys-070909-104117.DOI
  25. [25]Negi, M., Singh, A. R., Vikram, K., Roy, S. K. & Singh, A. K. Cuprate superconductors: An overview. In International Conference on Recent Advances in Functional Materials, 43–70 (Springer, 2024).
  26. [26]Kitatani, M. et al. Nickelate superconductors-a renaissance of the one-band hubbard model. npj Quantum Mater. 5, 59 (2020).
  27. [27]Stewart, G. R. Superconductivity in iron compounds. Rev. Mod. Phys. 83, 1589–1652 (2011). URL https://link.aps.org/doi/10.1103/RevModPhys.83.1589.DOI
  28. [28]Brison, J.-P. et al. Heavy fermion superconductivity. Phys. Rev. B Condens. Matter 280, 165–171 (2000).
  29. [29]Tresca, C. et al. Why mercury is a superconductor. Phys. Rev. B 106, L180501 (2022). URL https://link.aps.org/doi/10.1103/PhysRevB.106.L180501.DOI
  30. [30]Kawamura, M., Hizume, Y. & Ozaki, T. Benchmark of density functional theory for superconductors in elemental materials. Phys. Rev. B 101, 134511 (2020).
  31. [31]Ferreira, P. N. et al. Ab initio modeling of superconducting alloys. Mater. Today Phys. 48, 101547 (2024).DOI
  32. [32]Errea, I., Calandra, M. & Mauri, F. First-principles theory of anharmonicity and the inverse isotope effect in superconducting palladium-hydride compounds. Phys. Rev. Lett. 111, 177002 (2013). URL https://link.aps.org/doi/10.1103/PhysRevLett.111.177002.
  33. [33]Belli, F. & Zurek, E. Efficient modelling of anharmonicity and quantum effects in pdcuh₂ with machine learning potentials. npj Comput. Mater. 11, 87 (2025).
  34. [34]Belli, F., Zurek, E. & Errea, I. A chemical bonding based descriptor for predicting the impact of quantum nuclear and anharmonic effects on hydrogen-based superconductors. arXiv preprint arXiv:2501.14420 (2025).
  35. [35]Margine, E. R. & Giustino, F. Anisotropic migdal-eliashberg theory using wannier functions. Phys. Rev. B 87, 024505 (2013). URL https://link.aps.org/doi/10.1103/PhysRevB.87.024505.
  36. [36]Lucrezi, R. et al. Full-bandwidth anisotropic migdal-eliashberg theory and its application to superhydrides. Commun. Phys. 7, 33 (2024).
  37. [37]Mishra, S. B., Mori, H. & Margine, E. R. Electron-phonon vertex correction effect in superconducting h₃s. npj Computational Materials 11, 342 (2025).arxiv.org/abs/2501.01866
  38. [38]Bianco, R. & Errea, I. Non-perturbative theory of the electron-phonon coupling and its first-principles implementation. arXiv preprint arXiv:2303.02621 (2023).
  39. [39]Mishra, S. B., Belli, F., Zurek, E. & Margine, E. R. Anharmonicity and nonadiabaticity in hydride superconductors (2026). https://doi.org/10.48550/arXiv.2608.14494.
  40. [40]Zurek, E. Discovering new materials via A Priori crystal structure prediction. In Parrill, A. L. & Lipkowitz, K. B. (eds.) Reviews in computational chemistry, vol. 29, 274–326 (John Wiley & Sons, Inc., Hoboken, New Jersey, 2016). URL https://doi.org/10.1002/9781119148739.ch5.
  41. [41]Falls, Z., Avery, P., Wang, X., Hilleke, K. P. & Zurek, E. The xtalopt evolutionary algorithm for crystal structure prediction. J. Phys. Chem. C 125, 1601–1620 (2021). URL https://doi.org/10.1021/acs.jpcc.0c09531.
  42. [42]Pickard, C. J. & Needs, R. Ab initio random structure searching. J. Phys.: Condens. Mat. 23, 053201 (2011).DOI
  43. [43]Wang, Y., Lv, J., Zhu, L. & Ma, Y. CALYPSO: A method for crystal structure prediction. Comput. Phys. Commun. 183, 2063–2070 (2012).DOI
  44. [44]Oganov, A. R. & Glass, C. W. Crystal structure prediction using ab initio evolutionary techniques: Principles and applications. J. Chem. Phys. 124 (2006).
  45. [45]Lonie, D. C. & Zurek, E. Xtalopt: An open-source evolutionary algorithm for crystal structure prediction. Comput. Phys. Commun. 182, 372–387 (2011).
  46. [46]Laio, A. & Parrinello, M. Escaping free-energy minima. Proc. Natl. Acad. Sci. 99, 12562–12566 (2002).DOI
  47. [47]Gödecker, S. Minima hopping: An efficient search method for the global minimum of the potential energy surface of complex molecular systems. J. Chem. Phys. 120, 9911–9917 (2004).
  48. [48]Hajinazar, S. & Zurek, E. Xtalopt version 14: Variable-composition crystal structure search for functional materials [49] Avery, P. et al. Predicting superhard materials via a machine learning informed evolutionary structure search. npj Comput. Mater. 5, 89 (1–11) (2019). URL https://doi.org/10.1038/s41524-019-0226-8.
  49. [50]Racioppi, S. & Zurek, E. Using topology to predict electrides in the solid state. J. Phys. Chem. A 129, 10031–10038 (2025). URL https://doi.org/10.1021/acs.jpca.5c05317.
  50. [51]Racioppi, S., De la Roza, A. O., Hajinazar, S. & Zurek, E. Powder x-ray diffraction assisted evolutionary algorithm for crystal structure prediction. Digital Discovery 4, 73–83 (2025). URL https://doi.org/10.1039/D4DD00269E.
  51. [52]Wang, B., Hilleke, K. P., Hajinazar, S., Frapper, G. & Zurek, E. Structurally constrained evolutionary algorithm for the discovery and design of metastable phases. J. Chem. Theory Comput. 19, 7960–7971 (2023). URL https://doi.org/10.1021/acs.jctc.3c00594.
  52. [53]Hajinazar, S. & Zurek, E. Xtalopt version 13: Multi-objective evolutionary search for novel functional materials. Comput. Phys. Commun. 304, 109306 (2024). URL https://doi.org/10.1016/j.cpc.2024.109306.
  53. [54]Hajinazar, S. & Zurek, E. Xtalopt version 14: Variable-composition crystal structure search for functional materials through pareto optimization. Comput. Phys. Commun. 320, 109910 (2026). URL https://doi.org/10.1016/j.cpc.2025.109910.
  54. [55]Sanna, A. et al. Prediction of ambient pressure conventional superconductivity above 80 k in hydride compounds. npj Comput. Mater. 10, 44 (2024).DOI
  55. [56]Wang, H., Tse, J. S., Tanaka, K., Iitaka, T. & Ma, Y. Superconductive sodalite-like clathrate calcium hydride at high pressures. Proc. Natl. Acad. Sci. U. S. A. 109, 6463–6466 (2012). URL https://doi.org/10.1073/pnas.1118168109.
  56. [57]Shipley, M., Alice M. Hutcheon, Needs, R. J. & Pickard, C. J. High-throughput discovery of high-temperature conventional superconductors. Phys. Rev. B 104, 054501 (2021).DOI
  57. [58]Liu, H., Naumov, I. I., Hoffmann, R., Ashcroft, N. W. & Hemley, R. J. Potential high-tc superconducting lanthanum and yttrium hydrides at high pressure. Proc. Natl. Acad. Sci. U. S. A. 114, 6990–6995 (2017). URL https://doi.org/10.1073/pnas.1704505114.
  58. [59]Gao, G. et al. Superconducting high pressure phase of germane. Phys. Rev. Lett. 101, 107002 (1–4) (2008).pubmed.ncbi.nlm.nih.gov/18851245
  59. [60]Bi, T., Miller, D. P., Shamp, A. & Zurek, E. Superconducting phases of phosphorus hydride under pressure: Stabilization by mobile molecular hydrogen. Angew. Chem. Int. Ed. 56, 10192–10195 (2017). URL https://doi.org/10.1002/ange.201701660.
  60. [61]Geng, N., Bi, T. & Zurek, E. Structural diversity and superconductivity in s-p-h ternary hydrides under pressure. J. Phy. Chem. C 126, 7208–7220 (2022).DOI
  61. [62]Yan, Y., Bi, T., Geng, N., Wang, X. & Zurek, E. A metastable cash₃ phase composed of hs honeycomb sheets that is superconducting under pressure. J. Phys. Chem. Lett. 11, 9629–9636 (2020).DOI
  62. [63]Redington, M. & Zurek, E. Predicted high-pressure hot superconductivity in li₂cah₁₆ and li₂cah₁₇ phases that resemble the type-ii clathrate structure. Chem. of Mater. 36, 8412–8423 (2024).DOI
  63. [64]Wang, X., Geng, N., de Villa, K., Militzer, B. & Zurek, E. Superconductivity in dilute hydrides of ammonia under pressure.arxiv.org/abs/2405.11438
  64. [65]Di Cataldo, S., von der Linden, W. & Boeri, L. First-principles search of hot superconductivity in la-x-h ternary hydrides. npj Comput. Mater. 8, 2 (2022).DOI
  65. [66]Salzbrunner, P. T. et al. Developments and further applications of ephemeral data derived potentials. J. Chem. Phys. 159 (2023).
  66. [67]Pickard, C. J. Ephemeral data derived potentials for random structure search. Phys. Rev. B 106, 014102 (2022).DOI
  67. [68]Luo, X. et al. Deep learning generative model for crystal structure prediction. npj Comput. Mater. 10, 254 (2024).DOI
  68. [69]Jiang, B. et al. Data-driven design of high-temperature superconductivity among ternary hydrides under pressure. Phys. Rev. B 111, 054505 (2025).arxiv.org/abs/2410.20114
  69. [70]Lucrezi, R., Di Cataldo, S., von der Linden, W., Boeri, L. & Heil, C. In-silico synthesis of lowest-pressure high-tc ternary superhydrides. npj Comput. Mater. 8, 119 (2022).DOI
  70. [71]Geng, N., Hilleke, K. P., Belli, F., Das, P. K. & Zurek, E. Superconductivity in ch₄ and bh₄− containing compounds derived from the high-pressure superhydrides. Mater. Today Phy. 44, 101443 (2024). URL https://www.sciencedirect.com/science/article/pii/S2542529324001196.arxiv.org/abs/2401.00088
  71. [72]Radychev, D. et al. Rigid muffin-tip approximation in plane-wave codes for fast modeling of phonon-mediated superconductors. arXiv preprint arXiv:2508.21281 (2025).
  72. [73]Wines, D. & Choudhary, K. Data-driven design of high pressure hydride superconductors using dft and deep learning. Mater. futures 3, 025602 (2024).DOI
  73. [74]Wines, D., Xie, T. & Choudhary, K. Inverse design of next-generation superconductors using data-driven deep generative models. J. Phys. Chem. Lett. 14, 6630–6638 (2023).DOI
  74. [75]Wines, D., Choudhary, K., Biacchi, A. J., Garrity, K. F. & Tavan, Z. High-throughput dft-based discovery of next generation two-dimensional (2d) superconductors. Nano lett. 23, 969–978 (2023).DOI
  75. [76]Choudhary, K. & Garrity, K. Designing high-tc superconductors with bcs-inspired screening, density functional theory, and deep-learning. npj Comput. Mater. 8, 244 (2022).DOI
  76. [77]Hutcheon, M. J., Shipley, A. M. & Needs, R. J. Predicting novel superconducting hydrides using machine learning approaches. Phys. Rev. B 101, 144505 (2020). URL https://link.aps.org/doi/10.1103/PhysRevB.101.144505.DOI
  77. [78]Quinn, M. R. & McQueen, T. M. Identifying new classes of high temperature superconductors with convolutional neural networks. Front. Electron. Mater. 2, 893797 (2022).DOI
  78. [79]Gibson, J. B. et al. Developing a complete AI-accelerated workflow for superconductor discovery. npj Comput. Mater. 12, 95 (2026).arxiv.org/abs/2503.20005
  79. [80]Prakash, P. et al. Guided diffusion for the discovery of new superconductors. arXiv preprint arXiv:2509.25186 (2025).arxiv.org/abs/2509.25186
  80. [81]Cerqueira, T. F., Sanna, A. & Marques, M. A. Sampling the materials space for conventional superconducting compounds. Adv. Mater. 36, 2307085 (2024).DOI
  81. [82]da Silva, T. H., Cavignac, T., Cerqueira, T. F., Wang, H.-C. & Marques, M. A. Machine-learning accelerated prediction of two-dimensional conventional superconductors. Mater. Horiz. 12, 3408–3419 (2025).DOI
  82. [83]Gibson, J. B. et al. Accelerating superconductor discovery through tempered deep learning of the electron-phonon spectral function. npj Comput. Mater. 11, 7 (2025).DOI
  83. [84]Ma, T. et al. High-throughput calculation for superconductivity of sodalite-like clathrate ternary hydrides mxh₁₂ at high pressure. Mater. Today Phys. 38, 101233 (2023). URL https://www.sciencedirect.com/science/article/pii/S2542529323002699.DOI
  84. [85]Savin, A., Nesper, R., Wengert, S. & Fässler, T. F. Elf: The electron localization function. Angew. Chem. 36, 1808–1832 (1997).
  85. [86]O’Keeffe, M. The Bond Valence Method in Crystal Chemistry, 163–175 (Springer Netherlands, Dordrecht, 1992). URL https://doi.org/10.1007/978-94-011-2726-4_9.
  86. [87]Hughbanks, T. & Hoffmann, R. Chains of trans-edge-sharing molybdenum octahedra: Metal-metal bonding in extended systems. J. Am. Chem. Soc. 105, 3528–3537 (1983).DOI
  87. [88]Dronskowski, R. & Bloechl, P. E. Crystal orbital hamilton populations (cohp): Energy-resolved visualization of chemical bonding in solids based on density-functional calculations. J. Phys. Chem. 97, 8617–8624 (1993).DOI
  88. [89]Müller, P. C., Ertural, C., Hempelmann, J. & Dronskowski, R. Crystal orbital bond index: Covalent bond orders in solids. J. Phys. Chem. C 125, 7959–7970 (2021).
  89. [90]Naik, A. A., Ertural, C., Dhamrait, N., Benner, P. & George, J. A quantum-chemical bonding database for solid-state materials. Sci. Data 10, 610 (2023).DOI
  90. [91]Belli, F., Novoa, T., Contreras-García, J. & Errea, I. Strong correlation between electronic bonding network and critical temperature in hydrogen-based superconductors. Nat. Commun. 12, 5381 (2021).DOI
  91. [92]Belli, F., Torres, S., Contreras-García, J. & Zurek, E. Refining t_c prediction in hydrides via symbolic-regression-enhanced electron-localization-function-based descriptors. Annalen der Physik 537, e00280 (2025).DOI
  92. [93]di Mauro, M. E., Braïda, B., Errea, I., Novoa, T. & Contreras-García, J. Molecularity: A fast and efficient criterion for probing superconductivity. Phys. Rev. B 110, 174515 (2024). URL https://link.aps.org/doi/10.1103/PhysRevB.110.174515.DOI
  93. [94]Allen, P. B. & Dynes, R. C. Transition temperature of strong-coupled superconductors reanalyzed. Phys. Rev. B 12, 905–922 (1975). URL https://link.aps.org/doi/10.1103/PhysRevB.12.905.DOI
  94. [95]Wannier, G. H. The structure of electronic excitation levels in insulating crystals. Phys. Rev. 52, 191–197 (1937). URL https://link.aps.org/doi/10.1103/PhysRev.52.191.DOI
  95. [96]Giustino, F., Cohen, M. L. & Louie, S. G. Electron-phonon interaction using wannier functions. Phys. Rev. B 76, 165108 (2007). URL https://link.aps.org/doi/10.1103/PhysRevB.76.165108.DOI
  96. [97]Marzari, N., Mostofi, A. A., Yates, J. R., Souza, I. & Vanderbilt, D. Maximally localized wannier functions: Theory and applications. Rev. Mod. Phys. 84, 1419–1475 (2012). URL https://link.aps.org/doi/10.1103/RevModPhys.84.1419.DOI
  97. [98]Foster, J. M. & Boys, S. F. Canonical configurational interaction procedure. Rev. Mod. Phys. 32, 300–302 (1960). URL https://link.aps.org/doi/10.1103/RevModPhys.32.300.
  98. [99]Mahler, A., Williams, J. Z., Su, N. Q. & Yang, W. Wannier functions dually localized in space and energy. Phys. Rev. B 112, 075137 (2025). URL https://link.aps.org/doi/10.1103/16r3-hjc7.DOI
  99. [100]Solovyev, I. V., Pchelkina, Z. V. & Anisimov, V. I. Construction of wannier functions from localized atomiclike orbitals. Phys. Rev. B 75, 045110 (2007). URL https://link.aps.org/doi/10.1103/PhysRevB.75.045110.DOI
  100. [101]Zurek, E., Jepsen, O. & Andersen, O. K. Muffin-tin orbital wannier-like functions for insulators and metals. Chem. Phys. Chem. 6, 1934–1942 (2005).pubmed.ncbi.nlm.nih.gov/16144010
  101. [102]An, J. M. & Pickett, W. E. Superconductivity of mgb₂: Covalent bonds driven metallic. Phys. Rev. Lett. 86, 4366–4369 (2001). URL https://link.aps.org/doi/10.1103/PhysRevLett.86.4366.DOI
  102. [103]Autschbach, J. Orbitals: Some fiction and some facts. J. Chem. Educ. 89, 1032–1040 (2012).
  103. [104]Racioppi, S. et al. Lithium’s low-temperature phase transitions: Insights into quantum lattice dynamics and superconductivity. Phys. Rev. B. 111, 054111 (2025). URL https://doi.org/10.1103/PhysRevB.111.054111.
  104. [105]Naumov, I. I., Hemley, R. J., Hoffmann, R. & Ashcroft, N. Chemical bonding in hydrogen and lithium under pressure. J. Chem. Phys. 143 (2015).DOI
  105. [106]Racioppi, S. & Zurek, E. High-pressure electrides: A quantum chemical perspective. Annu. Rev. Condens. Matter Phys. 55, 421–442 (2025). URL https://doi.org/10.1146/annurev-matsci-080423-124050.
  106. [107]Racioppi, S., Storm, C. V., McMahon, M. I. & Zurek, E. On the electride nature of na-hp4. Angew. Chem. 62, e202310802 (2023).DOI
  107. [108]Cohen, M. L. & Anderson, P. Comments on the maximum superconducting transition temperature. In A Career In Theoretical Physics, 287–298 (World Scientific, 1994).
  108. [109]Novoa, T. et al. Predicting high-temperature hydrogen-based superconductors. Chem. Sci. 16, 57–68 (2025). URL http://dx.doi.org/10.1039/D4SC04465G.
  109. [110]Novoa, T., Belli, F., Bianco, R., Gian-...DOI
  110. [111]Kimura, M. W., Jang, S. W., Geng, N. & Zurek, E. Metal atom (dis)order and superconductivity in yehn (n = 8 – 20) high-pressure superhydrides. Inorg. Chem. 65, 17538–17551 (2026). URL https://doi.org/10.1021/acs.inorgchem.6c02106.
  111. [112]Racioppi, S., Storm, C. V., McMahon, M. I. & Zurek, E. On the electride nature of na-hp4. Angew. Chem. Int. Ed. 135, e202310802 (1–5) (2023). URL https://doi.org/10.1002/anie.20231802.
  112. [113]Wang, X., Pickett, W. E., Hutcheon, M., Prasankumar, R. P. & Zurek, E. Why mg₂irh₆ is predicted to be a high-temperature superconductor, but ca₂irh₆ is not. Angew. Chem. 136, e202412687 (2024).DOI
  113. [114]Dolui, K. et al. Feasible route to high-temperature ambient-pressure hydride superconductivity. Phys. Rev. Lett. 132, 166001 (2024). URL https://link.aps.org/doi/10.1103/PhysRevLett.132.166001.DOI
  114. [115]Hou, P., Belli, F., Bianco, R. & Errea, I. Strong anharmonic and quantum effects in pm₃n₃ under high pressure: A first-principles study. Phys. Rev. B 103, 134305 (1–8) (2021).
  115. [116]Troyan, I. A. et al. Anomalous high-temperature superconductivity in yh₆. Adv. Mater. 33, 2006832 (2021).
  116. [117]Zhao, J. et al. Prediction of pressure-induced superconductivity in the ternary systems cayh₂n (n=3-6) at moderate pressures. J. Mater. Chem. C 13, 4128–4136 (2025).
  117. [118]Borinaga, M., Errea, I., Calandra, M., Mauri, F. & Bergara, A. Anharmonic effects in atomic hydrogen: Superconductivity and lattice dynamical stability. Phys. Rev. B 93, 174308 (2016).DOI
  118. [119]Borinaga, M. et al. Anharmonic enhancement of superconductivity in metallic cmca4 hydrogen at high pressure: A first-principles study. J. Phys. Condens. Matter 28, 494001 (2016).
  119. [120]Errea, I. et al. Quantum crystal structure in the 250-kelvin superconducting lanthanum hydride. Nature 578, 66–69 (2020).DOI
  120. [121]Errea, I., Calandra, M. & Mauri, F. Anharmonic free energies and phonon dispersions from the stochastic self-consistent harmonic approximation: Application to platinum and palladium hydrides. Phys. Rev. B 89, 064302 (1–16) (2014).DOI
  121. [122]Hou, P., Belli, F., Bianco, R. & Errea, I. Quantum anharmonic enhancement of superconductivity in pδ3/mmc sch6 at high pressures: A first-principles study. J. Appl. Phys. 130, 175902 (2021).DOI
  122. [123]Errea, I. et al. High-pressure hydrogen sulfide from first principles: A strongly anharmonic phonon-mediated superconductor. Phys. Rev. Lett. 114, 157004 (2015).pubmed.ncbi.nlm.nih.gov/25933334
  123. [124]Pickett, W. E. Why compressed hydrides are near-room-temperature superconductors. Ann. Phys. (Berl.) 538, e00362 (2026).arxiv.org/abs/2508.18429
  124. [125]Nagamatsu, J., Nakagawa, N., Muranaka, T., Zenitani, Y. & Akimitsu, J. Superconductivity at 39 k in magnesium diboride. nature 410, 63 (2001).pubmed.ncbi.nlm.nih.gov/11242039
  125. [126]Shoji, A., Kiryu, S. & Kohjiro, S. Superconducting properties and normal-state resistivity of single-crystal nbn films prepared by a reactive rf-magnetron sputtering method. Appl. phys. lett. 60, 1624–1626 (1992).DOI
  126. [127]Gavaler, J. Superconductivity in nb-ge films above 22 k. Appl. Phys. Lett. 23, 480–482 (1973).DOI
  127. [128]Geng, N. et al. Conventional high-temperature superconductivity in metallic, covalently bonded, binary-guest c-b clathrates. J. Am. Chem. Soc. 145, 1696–1706 (2023).DOI
  128. [129]Mishra, S. B., Marcial, E. T., Debata, S., Kolmogorov, A. N. & Margine, E. R. Stability-superconductivity map for compressed na-intercalated graphite. Phys. Rev. B 110, 174508 (2024). URL https://link.aps.org/doi/10.1103/PhysRevB.110.174508.DOI
  129. [130]Lortz, R. et al. Superconductivity mediated by a soft phonon mode: Specific heat, resistivity, thermal expansion, and magnetization of Yb6. Phys. Rev. B 73, 024512 (2006). URL https://link.aps.org/doi/10.1103/PhysRevB.73.024512.DOI
  130. [131]Ebihara, T. et al. High-field specific heat and fundamental superconducting properties of the single-phase type-ii superconductor v2si. J. Phys. Soc. of Jpn. 95, 104707 (2026).
  131. [132]Ahmed, E. et al. Superconducting properties of the a15 structure compound v3ge. Phys. C: Supercond. Appl. 602, 1354140 (2022).DOI
  132. [133]Krockenberger, Y., Karimoto, S.-i., Yamamoto, H. & Semba, K. Coherent growth of superconducting tin thin films by plasma enhanced molecular beam epitaxy. J. Appl. Phys. 112 (2012).DOI
  133. [134]Luo, H., Wolf, S., Fuller, W., Edelstein, A. & Huang, C. Superconductivity of vn under pressure. Phys. Rev. B 29, 1443 (1984).
  134. [135]Bolotina, N. B. et al. The lower symmetry electron-density distribution and the charge transport anisotropy in cubic dodecaboride lub12. Jo. Phys. Condens. Matter 30, 265402 (2018).DOI
  135. [136]Bolotina, N. et al. Low temperature singularities of electron density in a two-gap superconductor zrb12. Solid State Sci. 142, 107245 (2023).DOI
  136. [137]Shi, H., Zarifi, N., Yim, W.-L. & Tse, J. Electron band structure of the high pressure cubic phase of alh3. In J. Phys.: Conf. Ser., vol. 377, 012093 (2012).DOI
  137. [138]Goncharenko, I. et al. Pressure-induced hydrogen-dominant metallic state in aluminum hydride. Phys. Rev. Lett. 100, 045504 (2008). URL https://link.aps.org/doi/10.1103/PhysRevLett.100.045504.DOI
  138. [139]Bassen, G. et al. Real-space orbital tiling approach for the classification and design of superconductors. J. Phys. Condens. Matter 37, 451501 (2025).DOI
  139. [140]Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B 47, 558–561 (1993).pubmed.ncbi.nlm.nih.gov/10004490
  140. [141]Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758–1775 (1999).DOI
  141. [142]Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865 (1996).pubmed.ncbi.nlm.nih.gov/10062328
  142. [144]Maintz, S., Deringer, V. L., Tchougréeff, A. L. & Dronskowski, R. Lobster: A tool to extract chemical bonding from plane-wave based dft. J. Comput. Chem. 37, 1030–1035 (2016). Software version 5.5.1.DOI

Paper details

vlnc:W_gru8ca2r
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents