Tuning crystal-fields by He-irradiation and orbital Widom line in SrVO$_3$ films

Abstract: Helium-ion irradiation of epitaxial SrVO$_3$/SrTiO$_3$ films causes a metal-insulator transition, so far attributed to a Mott localization driven by a reduced kinetic energy. Using density-functional theory plus dynamical mean-field theory, we show that the driving mechanism is instead the crystal-field splitting generated by the irradiation-induced tetragonal expansion, not the interaction-to-bandwidth ratio. The resulting $c$-axis vs. temperature phase diagram mirrors that of the one-band Hubbard model, but its critical end-point spawns an orbital Widom line, rooted in an anomalous compressibility of orbital, rather than charge occupation. At an effectively quarter filling, superexchange-like processes favor orbital over magnetic long-range order. Our results semi-quantitatively reproduce the fluence-dependent spectral gaps and transition thresholds reported experimentally, establishing ion implantation as a route to chemically expand correlated materials.

• The paper demonstrates that He irradiation drives the metal-insulator transition in SrVO3/SrTiO3 films primarily through tetragonal expansion and crystal-field splitting, rather than changes in the interaction-to-bandwidth ratio U/W.
• DFT+DMFT identifies polarized insulating states with gaps of about 350 meV under compression and 250 meV under expansion, while expanded films near c = 4.10 Å show quarter-filled orbital physics and strong orbital-ordering tendencies.
• The study introduces an orbital Widom line that tracks rapid redistribution between t2g orbitals across the metal-to-insulator crossover and provides a framework for controlling correlated phases through reversible chemical expansion.

Motivation and context

Helium-ion irradiation of epitaxial SrVO$_3$/SrTiO$_3$ films was reported by Wang et al. to induce a metal-insulator transition above a critical fluence, an effect previously attributed to Mott localization driven by a reduced kinetic energy. The paper under discussion re-examines this interpretation using density-functional theory combined with dynamical mean-field theory (DFT+DMFT) and reaches a different conclusion: the driving mechanism is the irradiation-induced tetragonal expansion of the unit cell, which generates a crystal-field splitting between the $t_{2g}$ orbitals, rather than a change in the interaction-to-bandwidth ratio $U/W$. This reframing has direct consequences for how the resulting phase diagram should be read: its structure resembles the iconic one-band Hubbard model, but the relevant control parameter is the crystal-field splitting, and the associated long-range order and crossover phenomena are orbital rather than charge or spin based.

The experimental setting is specific. SrVO$_3$ grown on SrTiO$_3$ is tensilely strained in-plane ($a=b=3.95$\,\AA\ versus the bulk value of 3.85\,\AA), producing a Poisson-ratio-driven compression of the out-of-axis lattice constant, $c/a<1$. Because the in-plane lattice constant is locked by the substrate, He implantation acts as a uniaxial chemical pressure that expands only the $c$-axis, continuously tuning $c/a$ from below to above unity. XRD indicates that implanted He occupies interstitial sites without introducing local defects, and the process is reversible by vacuum annealing, supporting its interpretation as a clean structural perturbation.

DFT: crystal-field splitting as the tunable parameter

Relaxing the substrate-strained structure yields an equilibrium out-of-plane constant of $_3$0\,\AA\ with a crystal-field splitting $_3$1\,meV favoring the $_3$2 orbital. Scanning $_3$3 shows that the splitting changes sign at the cubic geometry ($_3$4\,\AA): for $_3$5, the $_3$6/$_3$7 doublet is energetically favored. The orbital bandwidths respond asymmetrically — $_3$8 is considerably more sensitive to $_3$9 than $t_{2g}$0 — which will prove essential for understanding why compression and expansion behave differently.

DMFT: coexistence, hysteresis, and the insulating solutions

At room temperature, a metallic solution can be stabilized for all considered $t_{2g}$1 values when the DMFT loop is initialized from a metallic self-energy; within this branch, orbital occupations vary almost linearly with $t_{2g}$2. The authors argue that this "locked" metallic branch explains why earlier DFT+DMFT work on the same system failed to find insulating solutions. However, both strongly compressed ($t_{2g}$3\,\AA) and expanded ($t_{2g}$4\,\AA) structures lie inside a first-order Mott coexistence region.

To access the insulating branches, the authors employ a technical device: temporarily increasing the interaction parameters (to $t_{2g}$5\,eV) to force the Mott state, then relaxing back to the standard Kanamori values ($t_{2g}$6\,eV, $t_{2g}$7\,eV). Within a finite $t_{2g}$8-range the polarized insulator persists. Its character depends on the sign of the distortion:

• Compressed side ($t_{2g}$9\,\AA): a nearly half-filled $U/W$0 orbital with empty $U/W$1/$U/W$2, i.e., effective one-band Hubbard physics, with a spectral gap of approximately 350\,meV.
• Expanded side ($U/W$3\,\AA): a quarter-filled $U/W$4/$U/W$5 doublet, i.e., effective two-orbital quarter-filled Hubbard physics, with a gap of approximately 250\,meV.

Both gaps are uncharacteristically small for Mott insulators — naively one expects $U/W$6 — because the charge gap opens between states of different orbital character, while same-orbital Hubbard bands remain separated on the scale of $U/W$7. The authors note that analogous small-gap behavior occurs in LaTiO$U/W$8 and V$U/W$9O$_3$0, so this is a generic multi-orbital effect rather than a peculiarity of SrVO$_3$1.

Temperature dependence and the orbital Widom line

The two sides of the tetragonal distortion exhibit a marked asymmetry ("tetragonal dichotomy"). In the compressed structure, the insulating solution exists only up to $_3$2\,K, above which thermal fluctuations collapse the orbital polarization; consequently it cannot be reached by cooling alone. In the expanded structure — the experimentally relevant case — both branches persist up to $_3$3\,K, recombining into a single bad-insulator solution at higher temperature. Cooling from this high-temperature bad insulator stabilizes either a coherent metal (at $_3$4\,\AA) or a coherent Mott insulator (at $_3$5\,\AA), depending on $_3$6.

Above the anticipated critical end point, the metal-to-insulator evolution is a crossover whose location can be tracked by the maximum of the orbital-occupation derivative,

$_3$7

which the authors coin the orbital Widom line, in analogy to the Widom line of the doped one-band Hubbard model defined via the charge compressibility. Crossing this line is accompanied by a drop in double occupancies and a strong enhancement of the low-frequency scattering rate in the $_3$8/$_3$9 channels. Notably, the corresponding eigenvalue of the density response function shows a maximum across the oWL, but — somewhat surprisingly — this is not reflected in any significant change of the physical charge susceptibility $_3$0, underscoring that the anomaly resides in orbital rather than total charge occupation.

Ordering instabilities

With SU(2) symmetry enforced, the paramagnetic solutions hide incipient ordering tendencies that the authors quantify through Bethe-Salpeter eigenvalues computed within DMFT:

• Compressed insulator: leading magnetic eigenvalues $_3$1 and $_3$2 at room temperature already exceed unity, signaling antiferromagnetic order driven by superexchange in the effectively half-filled $_3$3 channel.
• Expanded insulator: the density-channel eigenvalue reaches $_3$4, indicating commensurate orbital order — alternating occupation of $_3$5 and $_3$6 along all three directions — stabilized by virtual hopping processes scaling as $_3$7 with the inter-orbital repulsion $_3$8. Even on the metallic branch, $_3$9 at room temperature suggests orbital order may eventually set in there too.

A key implication follows directly: because the expanded system sits near ideal quarter filling, superexchange-like processes favor orbital over magnetic long-range order, inverting the usual hierarchy of the one-band Hubbard model. The authors also note a practical limitation here: for large negative crystal-field splitting, plain DMFT convergence fails because the occupations attempt to break the local degeneracy, preventing them from probing the far side of the hysteresis without symmetry enforcement or supercell setups.

Phase diagram and comparison with experiment

Combining these results, the authors construct a semi-quantitative temperature-versus-$a=b=3.95$0 phase diagram for the expanded structure. At high temperature ($a=b=3.95$1\,K) the system crosses over smoothly from metal to bad insulator; at intermediate temperatures the orbital Widom line separates coherent metal from bad metal/bad insulator; below the critical end point a wide first-order coexistence region opens, bounded by the Mott transition line, with an orbital-ordering dome expected inside the insulating regime (and possibly a lower-temperature one in the metal).

The comparison with the transport data of Wang et al. is semi-quantitative. Samples grown at $a=b=3.95$2\,K cool from above the critical end point; compressed structures follow the metallic branch, while sufficiently expanded structures land on the insulating branch. The calculated hysteresis onset is compatible with the experimental fluence threshold ($a=b=3.95$3\,\AA\ at $a=b=3.95$4\,He/cm$a=b=3.95$5 and $a=b=3.95$6\,\AA\ at $a=b=3.95$7\,He/cm$a=b=3.95$8); the residual discrepancy is plausibly attributable to the PBE overestimate of the SrTiO$a=b=3.95$9 lattice constant. For the two most irradiated samples, the measured activation gaps are $c/a<1$0\,meV and $c/a<1$1\,meV below $c/a<1$2\,K — much smaller than the DMFT gaps, and the activation-law fit fails above 20\,K. The authors attribute this to lifetime smearing of the very small gap near the transition, possible intrinsic temperature dependence of the cfs-driven gap, and neglected vertex corrections to the conductivity. These are acknowledged caveats rather than resolved issues.

Limitations and open questions

Several assumptions bound the conclusions. The calculations fix the in-plane lattice constant to the PBE value (slightly larger than experiment), assume preservation of tetrahedral symmetry with no octahedral rotations, and use cubic-symmetry interaction parameters despite the broken symmetry, which the authors expect to cause only quantitative changes. The phase diagram is explicitly schematic outside the few computed trajectories, and no free-energy comparison between coexisting solutions was possible with the employed impurity solver, leaving the true thermodynamic stability of the branches undetermined. The precise relationship between the orbital-ordering boundary and the metal-insulator transition line — whether separated, coincident, or overlapping — remains unresolved, as does the real-space spin arrangement of the antiferromagnetic phase in the compressed insulator. Finally, whether the small experimental gaps reflect genuine many-body gap renormalization, disorder, or vertex effects is left open.

Conclusion

This work establishes that the He-irradiation-induced metal-insulator transition in SrVO$c/a<1$3/SrTiO$c/a<1$4 films is governed by the crystal-field splitting generated by uniaxial expansion, not by the interaction-to-bandwidth ratio. The resulting phase diagram mirrors the one-band Hubbard model in form but differs in substance: the Widom-line analog tracks orbital rather than charge redistribution, and the dominant ordered phase at quarter filling is orbital rather than magnetic. Beyond resolving the mechanism of a specific experiment, the study positions neutral-ion implantation as a controlled route to chemically expand correlated materials — effectively negative uniaxial pressure — complementary to conventional compression techniques.