Superconductivity mediated by polar modes in ferroelectric metals

Strontium titanate (SrTiO3) is an incipient ferroelectric that can be tuned into a ferroelectric phase by quantum tuning parameters such as pressure, isotopic or chemical substitution. Its phase diagram features a ferroelectric quantum critical point (QCP) where the soft transverse optical mode gap vanishes and characteristic crossovers in the dielectric response. Electron doping (e.g., by Nb substitution) introduces carriers into t2g bands, producing superconductivity at very low carrier densities—an outstanding puzzle due to the implied strong pairing interaction. The research question is whether proximity to the ferroelectric QCP enhances superconductivity and what bosonic modes mediate pairing. This study measures the pressure dependence of T_c near optimal doping and develops a microscopic model based on dynamical screening to test the hypothesis that longitudinal hybrid polar modes mediate pairing and that T_c is enhanced as the ferroelectric soft-mode gap decreases.

Prior work established superconductivity in doped SrTiO3 at very low carrier densities and identified SrTiO3 as a quantum paraelectric near a ferroelectric QCP. Numerous mechanisms have been proposed: coupling to critical transverse optical (TO) ferroelectric modes; plasmon-mediated pairing in the conduction electron system; multivalley processes; nonpolar acoustic and soft optical phonons; contributions from transverse polar optical currents; non-cancelling intermediate polar optical modes; impurity-localized phonons; disorder-related effects; polaronic/bipolaronic and preformed pair scenarios. Earlier dielectric screening approaches (Gurevich–Larkin–Firsov; Takada) suggested attraction from dynamically screened Coulomb interactions between Ω_TO and Ω_LO. Ruhman and Lee analyzed cancellations among intermediate polar modes and emphasized low-density regimes. Experiments showed T_c enhancements with strain and isotope substitution and lower T_c in KTaO3, consistent with distance from the QCP. However, a quantitative, parameter-free explanation of the dome-shaped T_c versus carrier density and its pressure dependence remained unsettled.

Experimental: Single crystals of SrTi1−xNbxO3 with nominal Nb concentrations 0.02, 0.2, and 1 at.% (log n/cm−3 ≈ 18.5, 19.5, 20.3) were obtained commercially and cut into Hall bar geometries. Ohmic contacts were prepared by argon-ion etching and Ti/Au sputtering. Hall measurements at liquid-helium temperatures in up to 9 T provided carrier densities (log n/cm−3 ≈ 18.0, 19.3, 20.6). Four-terminal resistivity was measured down to 50 mK; residual resistance ratio exceeded 600. For pressure studies, the 0.2 at.% Nb sample (T_c ≈ 0.4 K) was measured in two piston-cylinder clamp cells with Fluorinert pressure medium, in two laboratories, down to 50 mK. Pressure was calibrated via a tin manometer. T_c was defined as the temperature at which resistivity dropped by 10% from the normal-state residual value. Temperature sweeps were at ±1 K/h; uncertainties are shown in the figures. The normal-state resistivity A T^2 coefficient was also tracked versus pressure.

Theory: A dielectric screening model describes the effective electron-electron interaction V(q,ω) = 4πe^2/[q^2 ε(q,ω)], where ε(q,ω) includes contributions from ionic dipolar fluctuations and conduction electrons. The ionic sector is modeled as a gapped oscillator with bare transverse frequency Ω(q) = [Ω(0)^2 + v_s^2 q^2]^{1/2}, longitudinal Ω_LO ~ 100 meV, and background dielectric constant ε_∞; the conduction electron susceptibility is taken from the Lindhard function, approximated at low q by an interpolation involving the electron plasma frequency ω_p and characteristic frequency ω(q) = v_F q/√3. The coupled system yields two longitudinal hybrid polar modes ω_±(q) (solutions of ε(q,ω)=0) with coupling strengths γ_±(q); at low n, ω_− is a carrier plasma mode screened by ions, while at high n, ω_− tends toward the ionic mode screened by carriers. The effective interaction can be written as a sum over the two hybrid longitudinal resonances with weights determined by γ_±(q). To compute superconducting instabilities, the linearized gap equation is solved in weak coupling using a kernel U(k,k′) derived from Eliashberg theory in the Kirzhnits–Maksimov–Khomskii (KMK) approximation, which averages V(q,ω) along the imaginary axis and depends on individual quasiparticle energies rather than only their difference. A physically motivated wavevector (frequency) cutoff is applied; results for normalized T_c and the highest eigenvalue λ_h are insensitive to reasonable cutoff choices. Material parameters (effective mass, soft-mode dispersion, ε_∞, Ω_LO, velocity v_s, and the dependence of Ω(0) on n and pressure) are taken from independent measurements; the soft-mode gap Ω(0) evolves with pressure and density per an empirical relation capturing approach to the ferroelectric QCP. Numerical solutions use dense k- and frequency grids until convergence; comparisons with full Lindhard calculations including dissipation show qualitatively similar eigenvalue trends.

• Experimentally, in a 0.2 at.% Nb-doped SrTiO3 sample near the dome maximum (T_c ≈ 0.4 K at ambient pressure), T_c decreases sharply with modest hydrostatic pressure and collapses toward zero above approximately 5 kbar. This corresponds to increasing 1/ε_0 (increasing soft-mode gap), i.e., moving away from the ferroelectric QCP. Conversely, T_c increases as 1/ε_0 decreases (approaching the QCP).
• The normal-state resistivity retains an approximate T^2 dependence; the A coefficient changes by about 30% over the pressure range where T_c is suppressed by more than an order of magnitude, constraining scattering mechanisms.
• The observed pressure trend aligns with reports that strains or isotope substitution (oxygen-18) that soften the TO mode enhance T_c, and with lower T_c in KTaO3, which is farther from a ferroelectric QCP.
• Theoretical modeling with dynamically screened Coulomb interactions mediated by longitudinal hybrid polar modes reproduces: (i) a superconducting dome versus carrier density with a maximum near n ≈ 10^19–10^20 cm−3 and collapse at higher n due to enhanced screening; and (ii) a rapid suppression of T_c with pressure at fixed n around 10^20 cm−3, consistent with experiment.
• The model shows that attractive interaction arises primarily from virtual exchange of longitudinal hybrid polar modes; direct coupling to the transverse critical TO mode is not required to explain the enhancement near the QCP.
• Frequency-domain analysis reveals two characteristic steps in the interaction on the imaginary axis associated with the two hybrid modes; tuning toward the QCP changes the relative step sizes, enhancing or suppressing pairing depending on the Fermi energy position relative to these features.
• In the low-density limit, the pairing interaction amplitude scales with the soft-mode gap and can decrease on approaching the QCP, predicting a qualitative change: at very low n (<10^18 cm−3) T_c may decrease as the QCP is approached, unlike the intermediate-density regime near the dome maximum.

The results address whether ferroelectric criticality enhances superconductivity in SrTiO3 and identify the mediating bosons. Experimentally, T_c is strongly enhanced as the system is tuned toward the ferroelectric QCP (decreasing 1/ε_0), analogous to magnetic QCP systems but with different microscopic coupling. Theoretically, incorporating the dielectric response of both ions and conduction electrons yields two hybrid longitudinal polar modes that mediate attraction via dynamical screening. This framework captures the dome-shaped T_c versus carrier density and the rapid pressure-induced suppression at fixed n without invoking strong coupling to transverse critical modes or additional ad hoc mechanisms. The KMK-derived kernel, including retardation and realistic screening, is crucial for quantitative trends. The findings imply that ionic polarizability near a ferroelectric instability can provide a robust pairing glue in dilute metals. The model predicts anisotropic stress effects via their impact on ε_0 and suggests that in the very low-density regime, approaching the QCP could suppress T_c, a qualitative departure from the behavior near optimal doping. The work constrains alternative proposals (e.g., pure TO-mode critical pairing, impurity-localized modes, purely plasmonic mechanisms) by showing they are not required to explain the main experimental trends, though they may contribute away from optimal conditions.

This study combines pressure-dependent transport in Nb-doped SrTiO3 with a quantitative dielectric-screening theory to demonstrate that superconductivity is mediated predominantly by longitudinal hybrid polar modes arising from coupled ionic and electronic charge fluctuations. Experimentally, T_c is maximized near the ferroelectric QCP and collapses rapidly under modest pressure, while the normal-state T^2 coefficient changes only weakly. The theoretical model, using independently measured parameters and an Eliashberg/KMK-based kernel, reproduces the superconducting dome versus carrier density and the strong pressure dependence without invoking strong coupling to transverse critical modes. The work highlights polar-fluctuation-mediated pairing as a distinct route to superconductivity in highly polarizable, low-density metals and suggests broader applicability to related oxides and low-carrier systems. Future directions include probing much lower carrier densities to test the predicted inversion of pressure dependence, exploring uniaxial and biaxial strain in films and heterostructures, refining theoretical treatments beyond weak-coupling/Eliashberg to assess vertex and nonadiabatic effects, and investigating materials where increased Fermi energy does not strongly diminish interaction strength (e.g., bismuthates).

• The theoretical analysis relies on the weak-coupling Eliashberg framework and the KMK kernel; vertex corrections and nonadiabatic effects at energies comparable to or exceeding the Fermi energy may alter quantitative predictions, especially on the overdoped side.
• Absolute T_c values depend on cutoff choices and approximations; while trends are robust, magnitudes are estimated with logarithmic accuracy and may be overestimated at high carrier densities.
• The model employs effective-mass and isotropic approximations and simplified forms of the dielectric function; full band anisotropy and dissipation are only approximately included.
• Experimental pressure studies focus on a limited doping range near optimal Nb content; very low n (<10^18 cm−3) regimes where different behavior is predicted were not accessible with these samples.
• Possible inhomogeneities (especially in oxygen-depleted samples reported elsewhere) can complicate interpretation at low n; stress/strain experiments can introduce anisotropy and inhomogeneity not fully controlled here.