Many metals have special momenta where the electron energy surface is shaped like a saddle. At the saddle energy, the number of available electron states per energy diverges logarithmically; this divergence is called a van Hove singularity. When the Fermi level, the energy of the highest occupied states, is tuned onto a saddle, transport stops behaving like an ordinary metal. A clear example is the calculation of Breakdown of the Wiedemann-Franz law at the Lifshitz point of strained Sr2RuO4. It predicts that heat and charge currents stop tracking each other: electrical resistivity goes as $T^2\log(1/T)$, thermal resistivity goes as $T^{3/2}$, and the Lorenz ratio (thermal over electrical conductivity, normalized by temperature) vanishes as $T^{1/2}\log(1/T)$. That calculation, like most strain-tuning analyses, makes a quiet assumption. It treats the distance of the saddle from the Fermi level as moving linearly and passively with applied strain, so the singular behavior appears only at one fine-tuned strain. We call this the linear motion assumption for the saddle, borrowing the name from video frame interpolation. There, assuming objects move in straight lines at constant speed between frames fails whenever real forces act (Quadratic Video Interpolation; BiM-VFI).
This proposal claims that the saddle does not move linearly. Two forces act on it as it nears the Fermi level. The first is electron repulsion, which is known to drag saddles toward and pin them at the Fermi level in other materials. The second is lattice relaxation, and the lattice's stiffness itself depends on the magnetic state through spin-phonon coupling, meaning the dependence of vibration frequencies and bond stiffness on spin order and spin fluctuations. The combined effect can turn a single Lifshitz point into an extended strain window of singular transport, or it can shrink the window. The sign and size of this effect is set by whether the lattice and spin contribution reinforces or opposes the Coulomb pinning. The prediction is falsifiable. The width, temperature scaling, and Seebeck-coefficient signature of the Wiedemann-Franz breakdown region should differ sharply between a linearly moving saddle and a self-consistently pinned one.
Core Claim
Write $\delta$ for the energy of the saddle relative to the Fermi level, $\varepsilon$ for applied uniaxial strain, and $\varepsilon^*$ for the strain at which a rigid, non-interacting band would put the saddle exactly at the Fermi level. The linear motion assumption is $\delta = \alpha(\varepsilon-\varepsilon^*)$ with a constant deformation rate $\alpha$. We instead write
$$\delta = \alpha(\varepsilon-\varepsilon^*) + \Sigma(\delta), \qquad \frac{d\delta}{d\varepsilon} = \frac{\alpha}{1-\Sigma'(\delta)},$$
where $\Sigma(\delta)$ is the extra shift of the saddle produced by electrons re-occupying states near it. It has two parts: a Coulomb (Hartree-Fock) part and a lattice-relaxation part whose force constants depend on the spin configuration.
The occupation near a saddle varies as $\delta\log|\delta|$, which makes $\Sigma'(\delta) \approx -u_{\rm eff}\log(\Lambda/|\delta|)$. Here $u_{\rm eff}$ is the net coupling of the saddle occupation to its own energy, and $\Lambda$ is a band-energy cutoff. The consequence is a sign-dependent effect on the saddle's rate of motion $d\delta/d\varepsilon$. If $u_{\rm eff}>0$, that rate collapses logarithmically near $\delta=0$ and the saddle is pinned over a finite strain window. If $u_{\rm eff}<0$, the saddle is pushed through faster, and it can jump discontinuously.
The new content of the proposal is threefold. We compute $u_{\rm eff}$ with a lattice term that includes spin-phonon coupling. We feed the resulting nonlinear saddle trajectory into a Boltzmann transport calculation that keeps the large-momentum and small-momentum scattering channels separate. We then identify transport observables that tell the pinned, linear, and runaway scenarios apart.
Where this sits in prior work
The motivation comes from two results showing that saddle points do not move rigidly. Kim and Son found that electron-electron interactions shift the saddles of monolayer NbSe$_2$ toward the Fermi level. Lewandowski and Nadj-Perge showed that Coulomb interactions pin the van Hove singularity to the Fermi level over wide filling windows in twisted bilayer graphene. Both treat purely electronic pinning versus electron filling. Neither computes transport, and neither includes the lattice. We import their mechanism into a strain-tuned, transport-measured setting and add the lattice term.
Two further papers show why the lattice term must depend on spin. Strong spin-phonon coupling unveiled by coherent phonon oscillations in Ca2RuO4 observed a 1.3 cm$^{-1}$ redshift and a 90-degree phase flip of an octahedral-tilt mode across the Néel temperature, matching DFT distortions that accompany magnetic order. Influence of spin-phonon coupling on antiferromagnetic spin fluctuations in FeSe under pressure showed that lattice vibrations amplify spin fluctuations. Together they indicate that the elastic response of a correlated ruthenate or iron compound depends on its spin state. The lattice part of $\Sigma$ therefore cannot be computed from non-magnetic force constants alone.
The transport method comes from the Sr$_2$RuO$_4$ Wiedemann-Franz calculation, which we reuse and extend. We keep its three-band tight-binding model, which is fitted to angle-resolved photoemission data, and its separation of Umklapp scattering from small-momentum compressive scattering. We replace its passive strain-to-band map with the self-consistent trajectory above. Finally, First principles methodology for studying magnetotransport in narrow gap semiconductors with ZrTe5 example warns that transport anomalies often come from ordinary multiband compensation. Any claimed pinning signature therefore has to survive a multiband, non-interacting control.
Research Program
The first component computes $\Sigma(\delta)$ for Sr$_2$RuO$_4$ under uniaxial strain.
The Coulomb part is a self-consistent Hartree-Fock shift of the saddle-point states in the three-band model, with interaction strengths scanned over a physically plausible range rather than fixed.
The lattice part is computed in two steps. First, we calculate frozen-phonon DFT force constants and deformation potentials for the octahedral rotation and tilt modes and for the in-plane bond-stretch modes. We do this in several spin configurations: non-magnetic, ferromagnetic, and a spin-density-wave pattern at the dominant fluctuation wavevector. Second, we minimize the total energy (elastic plus electronic) at each strain. The spin dependence of the force constants is the spin-phonon ingredient. Its effect on $u_{\rm eff}$ is isolated by switching it off.
The second component solves the Boltzmann transport equation on the self-consistent bands at each strain and temperature. The output is the electrical resistivity, the thermal resistivity, the Lorenz ratio, and the Seebeck coefficient. The Seebeck coefficient matters because it changes sign when the saddle crosses the Fermi level. A pinned saddle should therefore show a Seebeck coefficient that stays near zero over the pinning window, rather than crossing zero at a single strain.
The third component is a second material that makes the result portable. Discovery of charge order above room-temperature in the prototypical kagome superconductor La(Ru1−xFex)3Si2 found charge order up to about 400 K, driven by a Ru-$d_{z^2}$ flat band and saddle near the Fermi level, together with Ru-Ru bond disproportionation. Fe doping adds local moments. We compute $\delta(x)$ versus Fe content $x$ with and without spin-dependent force constants, and ask whether the saddle stays near the Fermi level across $x = 0$, 0.01 and 0.05. That result would help explain why the ordering temperatures are robust. The same comparison speaks to the unresolved question of whether kagome charge order is electronic or structural: a pinning coupling $u_{\rm eff}$ dominated by its lattice part would favor the structural side.
Baselines
The naive reference point is the rigid band. Here the strain-dependent tight-binding hoppings are taken exactly as in the Sr$_2$RuO$_4$ Wiedemann-Franz model, with $\delta$ linear in $\varepsilon$. It is run on the identical momentum grid, temperature grid, and scattering channels as the full model, so every transport difference traces to the saddle trajectory alone.
The strongest existing method is that paper's Boltzmann calculation itself. It is the only work in the file that produces Lifshitz-point transport predictions from a realistic multiband model. It reports scaling forms ($\rho\propto T^2\log(1/T)$, $\rho_Q\propto T^{3/2}$, $L\propto T^{1/2}\log(1/T)$) but not numerical Lorenz-ratio values or window widths. No reported result was found on Lacuna for numerical Lorenz-ratio values, or for the strain width of the Wiedemann-Franz breakdown region. We will therefore reproduce its scalings as a correctness check before any comparison.
Beating this baseline means something specific. The full model must predict a strain window where $L/L_0$ (the Lorenz ratio relative to its standard metallic value $L_0$) stays below a fixed threshold, say 0.8 at the lowest temperature computed. That window must differ in width from the linear model's by more than the strain resolution of typical uniaxial-strain experiments. And it must remain robust across the scanned interaction range.
Three ablations isolate the key ingredient. The first uses Coulomb-only pinning. The second uses lattice-only relaxation with spin-independent force constants. The third is the full model with spin-dependent force constants. The claim that spin-phonon coupling matters succeeds only if the third differs measurably from the second.
Why not just…?
Why not just fit a quadratic $\delta(\varepsilon)$ to DFT calculations at a few strains? This is the transport analogue of the quadratic video prior, which Quadratic Video Interpolation found cut positional error on GOPRO from 1.79 to 0.97 compared with a linear model. It is a genuine threat if the nonlinearity is smooth. The logarithmic form of $\Sigma'$ says it is not smooth: the saddle's motion stalls in a way no low-order polynomial fitted away from $\varepsilon^*$ captures. This is the same failure the video literature reports under abrupt forces between frames (BiM-VFI). We test it directly. We fit a quadratic to self-consistent points outside the window and check whether it predicts the window width. If it does, the pinning mechanism adds nothing practical, and we will say so.
Why not just attribute any widened Wiedemann-Franz breakdown to strain inhomogeneity? This is the most serious competitor. A distribution of local strains smears a single Lifshitz point into an apparent window. The two scenarios differ in temperature dependence. Smearing gives a crossover back to Fermi-liquid scaling below a temperature set by the width of the strain distribution. Pinning keeps the $T^{1/2}\log(1/T)$ suppression of $L$ at every strain inside the window. We will compute both. A Gaussian-averaged rigid-band model is tuned to reproduce the same window width, and we compare low-temperature exponents at fixed strain.
Why not just compute the lattice with non-magnetic phonons? The Ca$_2$RuO$_4$ phonon shift and phase flip and the FeSe spin-fluctuation amplification show that force constants depend on the spin state. Whether that matters for $u_{\rm eff}$ here is an open question, and the spin-dependent versus spin-independent ablation decides it. A null result would be informative: it would show that pinning in ruthenates is electronic.
Decisive Checks
The central check is the predicted strain dependence of $L/L_0$ and of the Seebeck coefficient at low temperature. The linear model predicts a narrow Lorenz dip and a Seebeck sign change at a single strain. The pinned model predicts a flat-bottomed Lorenz dip and a Seebeck coefficient that stays near zero across the window. The runaway model ($u_{\rm eff}<0$) predicts an abrupt, possibly hysteretic jump.
The proposal is falsified in one of two ways. The first is if, over the plausible interaction range, $|\Sigma'|$ never grows large enough near $\delta=0$ to change the window width by more than the inhomogeneous-smearing baseline does. The second is if the quadratic-fit baseline matches the full model.
A second check is internal consistency. The self-consistent lattice relaxation should soften the octahedral modes near the pinning window. That softening is directly measurable by coherent-phonon pump-probe spectroscopy of the kind used on Ca$_2$RuO$_4$, performed under strain, and a predicted redshift and phase change would become the experimental test we hand to collaborators.
A third check applies the full model to La(Ru$_{1-x}$Fe$_x$)$_3$Si$_2$. There we test whether the computed saddle stays within a few meV of the Fermi level across the measured doping range, consistent with the persistence of charge order in the discovery paper.
References
- Breakdown of the Wiedemann-Franz law at the Lifshitz point of strained <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msub><mml:mi mathvariant="normal">Sr</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi mathvariant="normal">RuO</mml:mi><mml:mn>4</mml:mn></mml:msub></mml:math>
- Quadratic Video Interpolation
- BiM-VFI: Bidirectional Motion Field-Guided Frame Interpolation for Video with Non-uniform Motions
- Quasiparticle energy bands and Fermi surfaces of monolayer <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msub><mml:mi>NbSe</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math>
- Does filling-dependent band renormalization aid pairing in twisted bilayer graphene?
- Strong spin-phonon coupling unveiled by coherent phonon oscillations in <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:msub><mml:mi>Ca</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msub><mml:mi>RuO</mml:mi><mml:mn>4</mml:mn></mml:msub></mml:mrow></mml:math>
- Influence of spin-phonon coupling on antiferromagnetic spin fluctuations in FeSe under pressure: First-principles calculations with van der Waals corrections
- First principles methodology for studying magnetotransport in narrow gap semiconductors with ZrTe5 example
- Discovery of charge order above room-temperature in the prototypical kagome superconductor La(Ru1−xFex)3Si2