Microscopic theory of pair density waves in spin-orbit coupled Kondo lattice

Introduction. Spatially modulated pair condensates, or pair density wavesAgterberg et al. (2020) (PDWs) were originally envisioned by Fulde, FerrellFulde and Ferrell (1964), Larkin and OvchinikovLarkin and Ovchinnikov (1964)(FFLO), as a high-field superconducting response to the Zeeman splitting of the Fermi surface. The recent observation of PDWs via scanning tunneling microscopyRanderia et al. (2016) measurements in cuprateHücker et al. (2011), iron-based Zhao et al. (2023); Liu et al. (2023) and heavy fermion superconductorsGu et al. (2023) has sparked a resurgence of interest in this phenomenon. Crucially, in each of these unconventional superconductors the PDW develops without a magnetic field. PDWs naturally form as a response of a uniform pair condensate to a charge or spin density wave(CDW)Agterberg et al. (2020); Fradkin et al. (2015). However, the recent PDW observation in the novel triplet superconductor, UTe₂Gu et al. (2023) found that the PDW survives the melting of CDW orderAishwarya et al. (2024), suggesting that in this case, the PDW is a primary order parameter, developing without the aid of a charge or spin density wave. These observations motivate us to seek a natural mechanism for the spontaneous development of pair density wave order.

The condensation of a PDW requires a pair susceptibility that peaks at a finite momentum. This does not occur naturally in a BCS superconductorBardeen et al. (1957), where the logarithmic divergence of the uniform pair susceptibility is cut off by the momentum, or more precisely by $v_{F}Q$, where $v_{F}$ is the Fermi velocity and $Q$ the wavevector of the PDW. In a BCS superconductors, a Zeeman field polarizes the Fermi surfaces, cutting off the zero-momentum instability and favoring a finite-momentum peak in the pair susceptibility, leading to an FFLO-like pair density wavesFulde and Ferrell (1964); Larkin and Ovchinnikov (1964). While experiment suggests that pair density waves can spontaneously develop in unconventional superconductors even in the absence of a magnetic field, there is currently no current consensus on the underlying mechanism.

In this paper we explore PDW formation in the context of the Kondo lattice model for heavy fermions. The current authors (CPT) have recently proposed a three dimensional, solvable model in which the Kondo screening of a $\mathbb{Z}_{2}$ spin liquid induces superconductivity in the surrounding conduction seaColeman et al. (2022); Tsvelik and Coleman (2022). The underlying spin liquid is a three dimensional generalization of the spin-orbital derivative of the well-known $\mathbb{Z}_{2}$ Kitaev spin liquid, known as the Yao-Lee spin liquidYao and Lee (2011)). At half-filling and in the weak coupling limit, Kondo coupling between the spin liquid and the conduction electrons induces odd-frequency triplet superconductivity.

In this letter, we demonstrate that our modelColeman et al. (2022) provides a natural mechanism for the emergence of incommensurate pair-density waves when the system is doped away from half-filling. In particular, its weak coupling description offers a long-sought pathway to understanding the spontaneous formation of pair-density waves beyond one dimension Agterberg et al. (2020) in the absence of a field.

At half-filling, the model exhibits nesting between electron, hole, and Majorana Fermi surfaces, leading to a logarithmic instability into an electron-Majorana pair condensate under infinitesimal Kondo coupling. Our key result is that doping the system away from half-filling, through a shift in the chemical potential, modifies the electron and hole Fermi surfaces while leaving the Majorana Fermi surface unchanged. This imbalance induces a finite-momentum FFLO-like electron-Majorana condensation, giving rise to amplitude-modulated pair-density waves.

A key advantage of our model over general parton theories is that it does not rely on an approximate Gutzwiller projection to enforce the single-occupancy constraint by mean-field methods. Instead, in the Yao-Lee spin liquid, this constraint is inherently encoded in the Majorana representation, offering a more precise understanding of how the chemical potential influences the order parameter.

Model. The CPT modelColeman et al. (2022) is defined on the hyperoctagon lattice (Fig. 1), Kondo-couples conduction electrons to the spins of a Yao-Lee spin liquid. Each site possesses three degrees of freedom: conduction electrons, localized spins, and localized orbitals, leading to a Hamiltonian with three components, $H_{CPT}=H_{c}+H_{YL}+H_{K},$ where $H_{c}$ describes the nearest-neighbor hopping of the conduction electrons, $H_{YL}$ captures the Yao-Lee spin-spin interaction, and $H_{K}$ couples the conduction sea to the Yao-Lee spin liquid through the Kondo interaction,

Yao-Lee spin liquid: The Yao-Lee term involves an anisotropic Kitaev interaction between the orbitals $\lambda_{j}$’s components $\alpha_{ij}=x,y,z$ on nearest-neighbor sites $ij$, decorated by a Heisenberg interaction between the spins $\vec{S}_{j}$. The Hamiltonian can be significantly simplified through fermionization, where the commutation relations of the orbital and spin operators are encoded by expressing the orbitals as $\vec{\lambda}_{j}=-i\vec{b}_{j}\times\vec{b}_{j}$ and the spins as $\vec{S}_{j}=-\frac{i}{2}\vec{\chi}_{j}\times\vec{\chi}_{j}$, in terms of Majorana fermions $b$ and $\chi$. The physical Hilbert space is constrained by ${\cal D}=8ib_{1}b_{2}b_{3}\chi_{1}\chi_{2}\chi_{3}=1$, which commutes with the Hamiltonian.

The orbital frustration in the Yao-Lee Hamiltonian induces a Majorana fractionalization of spins, which couple to static $\mathbb{Z}_{2}$ gauge fields $\hat{u}_{ij}=ib^{\alpha_{ij}}_{i}b^{\alpha_{ij}}_{j}$, $H_{YL}=K\sum_{\left\langle i,j\right\rangle}\hat{u}_{ij}(i\vec{\chi}_{i}\cdot\vec{\chi}_{j}).$, An Ising deconfinement transition occurs at a finite temperature $T_{c1}$, where the gauge fields freeze into a flux-free configuration. This leads to a four-band Majorana Hamiltonian on the hyperoctagon lattice. At low energies, the Yao-Lee spin liquid can be effectively described by a single-band with dispersion $\epsilon_{\chi}({\bf k})=K\tilde{\epsilon}({\bf k})$, where

where, $c_{i}=\cos(k_{i}/2)$ and $s_{i}=\sin(k_{i}/2)$. The projected Hamiltonian is then expressed as,

where the momentum sum takes place over the cubic Majorana Brillouin zone \mancube.

Conduction electrons: Similarly, the low energy physics of the conduction band following a non-singular gauge transformationColeman et al. (2022), can be described by a projected single band $\epsilon_{c}({\bf k})=-t\tilde{\epsilon}({\bf k})$, with $\tilde{\epsilon}({\bf k})$ taking the form (2). Further, the conduction sea can be restricted to $1/2BLZ$ \mancube, by rewriting the conduction Hamiltonian in terms of the Balian-Werthammer spinor $\psi_{\bf k}=(c_{{\bf k}},-i\sigma^{2}c^{{}^{\dagger}}_{-{\bf k}})^{T}$ as,

Kondo interaction: Below the Ising phase boundary of Yao-Lee spin liquid, the Majorana deconfinement of spins enables the re-expression of Kondo interactions in terms of Majorana fermions and conduction electrons. This reformulation facilitates an analytic weak coupling mean-field treatment of the Kondo interaction without relying on Gutzwiller projection. In this representation, the spins are expressed as $\vec{S}_{j}=-\frac{i}{2}\vec{\chi}_{j}\times\vec{\chi}_{j}$, allowing the Kondo interaction to be rewritten as,

$H_{K}$ can be decoupled via a Hubbard-Stratonovich transformation in terms of a spinor order parameter field $V_{j}=-J\langle\vec{\chi}_{j}\cdot\vec{\sigma}c_{j}\rangle=(V_{j\uparrow},V_{j\downarrow})^{T}$, which is solved self-consistently. Expressing the conduction electrons and the spinor order parameter using the Balian-Werthamer representation, $\psi_{\vec{k}}=(c_{\vec{k}},-i\sigma^{2}c^{\dagger}_{-\vec{k}})^{T}$ and $\mathcal{V}_{j}=(V_{j},-i\sigma^{2}V^{*}_{j})^{T}$, respectively, provides a compact mean-field solution for the Kondo interaction.

In previous work Coleman et al. (2022), we showed that at half-filling, the model exhibits a logarithmic singularity in the electron-Majorana pair susceptibility, leading to odd-frequency triplet superconductivity.

Pair density waves. We now show that doping this analytically tractable Kondo lattice model provides a natural mechanism for the spontaneous formation of a pair density wave (PDW). Here, the chemical potential plays a role analogous to the Zeeman field in conventional superconductors. Specifically, the addition of a chemical potential splits the electron and hole Fermi surfaces while leaving the Majorana Fermi surface unaffected (Fig 2a). This then causes the fractionalized spinor order parameter $\mathcal{V}_{\vec{q}}$ to acquire finite momentum $\vec{q}$ (Fig. 2b), resulting in the formation of a PDW state. In this scenario, the Fourier transform of the Kondo interaction with finite momentum order parameter is given by,

The tendency to form a finite-momentum spinor order arises from the static electron-Majorana pairing susceptibility (Fig. 3) being maximized at a finite momentum $\vec{q}$,

Which is expressed in terms of the conduction and Majorana Green’s functions, $G_{\psi}(\vec{k}+\vec{q},i\omega_{n})$, and $G_{\chi}(\vec{k},i\omega_{n})$ respectively, with $\mathcal{Z}=\frac{2\mathcal{V}}{V}$ representing the orientation of the spinor order. Further simplifications to trace conditions are made by noting that $\mathcal{P}=\mathcal{Z}\mathcal{Z}^{\dagger}$ forms a projection operator, which decouples a conduction Majorana as a result of Kondo screening.

Following simplifications, outlined in the supplementary material, we can rewrite the susceptibility for the hyperoctagon lattice in the following closed form,

Here, $\rho(0)$ is the density of states at the Fermi surface, and $\vec{v}_{F}=\vec{\nabla}_{\bf k}\epsilon_{c}({\bf k})$ is the Fermi velocity of the conduction sea, which, for the hyperoctagon lattice, depends on the orientation $(\theta,\phi)$. The logarithmic instability of the electron-Majorana susceptibility in equation (9) is cut off by the chemical potential, which splits the nesting between the conduction sea and the Majorana Fermi surfaces. This results in the susceptibility maxima occurring at finite momenta $\vec{q}$ (Fig. 3).

To evaluate the susceptibility in angular coordinates, we rewrite the Fermi-velocity of the conduction sea in the momentum basis,

where, $c_{i}=\cos(k_{i}/2)$ and $s_{i}=\sin(k_{i}/2)$. We then parameterize the $c_{i}$ of the Fermi surface in spherical coordinates

in terms of angular variables $\theta$ and $\phi$ and radius $r=\sqrt{3}/2$. Carrying out the angular integral in equation (9), we numerically find that the susceptibility is maximized (Fig. 3) for:

along the principal axes where the magnitude of the wavenumber at the susceptibility maxima $q_{m}=0.77\pi\mu$ (Fig. 3) is proportional to the chemical potential. Here, the cubic symmetry of our model leads to six degenerate wave vectors along which the spinor order prefers to modulate. Such degeneracy is often lifted, as in the case of FFLOLarkin and Ovchinnikov (1964) states, where the order parameter favors amplitude modulation due to attractive interactions between $\mathcal{V}_{\vec{q}}$ and $\mathcal{V}_{-\vec{q}}$. This results in ordering of the form $\mathcal{V}_{\vec{q}}+\mathcal{V}_{-\vec{q}}$, or other superpositions, depending on the level of doping.

Ginzburg Landau Theory. The criteria for amplitude modulation of the pairing is most easily accessible from the Ginzburg Landau theory where the free energy is a function of the magnitude of the spinor order $|V_{q}|^{2}=\mathcal{V}^{{}^{\dagger}}_{q}\mathcal{V}_{q}$ given by,

The quadratic degeneracy amongst the order parameters $\mathcal{V}_{\vec{q}_{i}}$ is broken by the quartic coefficients $\beta_{i}$. Depending on the values of $\beta_{i}$, the model exhibits various amplitude-modulated phases. For $\beta_{1}>2\beta_{2}>2\beta_{3}$, the order parameter displays unidirectional modulation and takes the form $V(\vec{r})=V\cos(\vec{q}_{i}\cdot\vec{r})$. In contrast, for $\beta_{1}>2\beta_{2}=2\beta_{3}$, the degeneracy of $\vec{q}_{x}$, $\vec{q}_{y}$, and $\vec{q}_{z}$ is lifted by a sixth-order term, resulting in a pair-density wave crystal of the form $V(\vec{r})=V[\cos(\vec{q}_{x}\cdot\vec{r})+\cos(\vec{q}_{y}\cdot\vec{r})+\cos(\vec{q}_{z}\cdot\vec{r})]$, arising through the FFLO mechanism Fulde and Ferrell (1964); Larkin and Ovchinnikov (1964).

Conduction electron response. Thus far, we have shown that doping the CPT Kondo lattice model induces amplitude modulation in the electron-Majorana spinor order parameter. This in turn, generates amplitude modulation in the self-energy, $\Sigma=\mathcal{V}^{\dagger}G_{\chi}\mathcal{V}$, specifically affecting the triplet pairing component of the electronic self-energy:

where $\Sigma_{N}$ denotes the normal component of the self-energy, representing the odd-frequency magnetic component of the electrons, while $\Delta$ corresponds to the triplet pairing component. As both $\Sigma_{N}$ and $\Delta$ are proportional to $V^{2}(\vec{r})$, they inherit the amplitude modulation of the spinor order parameter. Thus doping the CPT Kondo lattice Coleman et al. (2022) provides a natural mechanism for the spontaneous formation of pair-density waves beyond one dimension, even in the absence of an external magnetic field.

Discussion. In this paper, we have proposed a natural mechanism for the spontaneous formation of pair-density waves, where the underlying superconductivity arises from pairing with a $\mathbb{Z}_{2}$ spin liquid. Screening of the spin liquid by the conduction sea then gives rise to triplet pair-density waves. While our model represents a specific case, the underlying concept extends to other forms of unconventional superconductivity believed to result from spin fractionalization, such as the RVB theory of high-$T_{c}$ superconductivity.

In heavy fermion materials, such pair-density wave (PDW) order has been observed in STM experiments in UTe₂ Gu et al. (2023). This PDW order has subsequently been identified as the parent order underlying the charge-density wave (CDW) order through melting transitions Aishwarya et al. (2024). Furthermore, time-reversal symmetry breaking observed in chiral step-edge experiments Jiao et al. (2020) and the onset of re-entrant superconductivity at 40T Aoki et al. (2022), alongside the absence of a two-stage transition and the fully gapped spectrum revealed by transport measurements Suetsugu et al. (2024); Theuss et al. (2024), have fueled debate Jiao et al. (2020); Ishizuka et al. (2019); Xu et al. (2019); Machida (2020); Bae et al. (2021); Hayes et al. (2021); Machida (2021); Nakamine et al. (2021); Thomas et al. (2021); Fujibayashi et al. (2022); Girod et al. (2022); Rosa et al. (2022); Shaffer and Chichinadze (2022); Wei et al. (2022); Hazra and Volkov (2024); Hazra and Coleman (2023); Iguchi et al. (2023); Ishihara et al. (2023); Matsumura et al. (2023); Chang et al. (2024); Suetsugu et al. (2024); Theuss et al. (2024) over the nature of superconducting order of UTe₂. This controversy arises because no known two-dimensional representations of superconducting order for UTe₂ can fully reconcile these experimental findings.

While our work does not provide a microscopic mechanism for the superconductivity in UTe₂, it does open a way out of this dilemma. Superconductivity breaks the U(1) isospsin symmetry associated with charge conservation. In principle, group theory allows for both singly connected representations of the order parameter, exemplified by the charge 2e BCS pairing, but it also allows for double-group representations of the order parameter, with half-integer isospin. Majorana-mediated superconductivity is a first concrete example of this new class of broken-symmetry. Here the electron-Majorana spinor plays the role analogous to the quark in strong interaction physics. By analogy, new families of superconducting phases are now possible, generated by product group representations analogous to the eight-fold way of mesons. This opens up the a zoo of enriched superconducting phases, with a corresponding diversity of new topological superconductors. Recent work has confirmed Zhuang and Coleman (2024) the potential to form topological states out of such fractionalized order. Importantly, double groups introduce Kramers degeneracy, allowing for broken time-reversal in low-symmetry crystalline environments, providing a clue for for the emergence of time-reversal symmetry-breaking, with a single-step transition emerging out of a heavy fermi liquid Panigrahi et al. (2024), observed in UTe₂, despite its low orthorhombic Immm crystal symmetry of UTe₂.