jb: [1]˝: #1blue mb:

Chern number landscape of spin-orbit coupled chiral superconductors

Abstract

Chiral superconductors are one of the predominant quantum electronic states of matter where topology, symmetry, and Fermiology intertwine. This is pushed to a new limit by further invoking the coupling between spin and charge degrees of freedom, which fundamentally affects the principal nature of the Cooper pair wave function. We investigate the onset of superconductivity in the Rashba-Hubbard model on the triangular lattice, which is symmetry-classified by the associated irreducible representations (irrep) of the hexagonal point group. From an instability analysis by means of the truncated-unity functional renormalization group (TU-FRG) we find the $E_{2}$ irrep to dominate a large fraction of phase space and to lead up to an energetically preferred gapped, chiral superconducting state. The topological phase space classification associated with the anomalous propagators obtained from TU-FRG reveals a fragmentation of the $E_{2}$ domain into different topological sectors with vastly differing Chern numbers. It hints at a potentially applicable high sensitivity and tunability of chiral superconductors with respect to topological edge modes and phase transitions.

Introduction.—Topological superconductors (TSCs) are amongst the most desired states of topological matter, as their ability to host Majorana zero modes is believed to be key for future implementation of topological quantum computing platforms. This immense technological potential is hindered by the lack of cfundamental understanding of TSCs: (i) Theoreticians are still unable to provide recipes or guidance for crystal synthesis or to predict materials hosting TSC phases. (ii) Experimental discoveries of TSC candidate materials often come as a surprise; and often enough after some initial excitement, subsequent studies suggest that trivial superconducting states might be a more likely explanation. To remedy the former, it is essential that the many-body techniques which predict unconventional superconducting ground states are also capable of shedding light on the topological features of these many-body instabilities.

We are particularly interested in topological superconductivity in the two-dimensional triangular lattice as the simplest case with hexagonal symmetry. Here we can also draw on extensive and noteworthy literature on unconventional superconductivity in the triangular lattice Hubbard model, motivated by the doped Mott insulator $\kappa$-(BEDT-TTF)₂Cu₂(CN)₃ Vojta and Dagotto (1999) as well as water-intercalated sodium cobaltates Na_xCoO₂$\cdot y$H₂O Honerkamp (2003); Ikeda et al. (2004); Zhou and Wang (2008); Chen et al. (2013). Another important class of materials are the $\sqrt{3}\times\sqrt{3}$ reconstructed adatom lattices on semiconductor substrates such as Sn/Si(111), Pb/Si(111) and Sn/SiC(0001). These are believed to be good realizations of one-orbital Hubbard models, and their observed ground states range from spin and charge density waves to chiral superconductivity Glass et al. (2015); Badrtdinov et al. (2016); Ming et al. (2017); Cao et al. (2018); Nakamura et al. (2018); Tresca et al. (2018); Adler et al. (2019); Wolf et al. (2022); Machida et al. (2022); Biderang et al. (2022a); Ming et al. (2023). Most recently, the advent of twisted two-dimensional materials led to a renewed interest in unconventional superconductivity on the triangular lattice Classen et al. (2019); Venderley and Kim (2019); Gneist et al. (2022); Scherer et al. (2022).

Inversion-symmetry breaking is ubiquitous in materials, e.g. due to crystalline absence of an inversion center, as in so-called non-centrosymmetric materials, or heavy atom superlattices, heterostructures and surface systems. Nevertheless, the effect of inversion symmetry breaking, usually manifested as Rashba spin-orbit coupling (RSOC), on the correlated and superconducting phase diagrams of such materials and corresponding models remains relatively unexplored.

There have been a few studies on the paradigmatic square lattice case Greco and Schnyder (2018); Wolf et al. (2022); Beyer et al. (2023). As we will see later in the group-theoretical discussion, the possible mixed states in the triangular lattice case are significantly more interesting. Notable exceptions where spin-orbit effects were considered in interacting triangular lattice systems are works on $\sqrt{3}\times\sqrt{3}$-Pb/Si(111) Badrtdinov et al. (2016); Tresca et al. (2018); Machida et al. (2022); Biderang et al. (2022b), on $\sqrt{3}\times\sqrt{3}$-Sn/SiC(0001) Glass et al. (2015) as well as on twisted bilayer systems WSe2 Klebl et al. (2023) and PtSe2 Klebl et al. (2022).

In this Letter, we close this gap in the literature by presenting a paradigmatic case study for the triangular-lattice Rashba–Hubbard model (TLRHM). Despite being most basic model, it turns out to feature a surprisingly rich phase diagram. We employ the truncated united extension of the functional renormalization group (TUFRG). TUFRG allows the investigation of competing many-body instabilities in a way that is unbiased to any one particular instability, treating particle-particle and particle-hole instabilities on equal footing. (For more details, see SM.) This method is particularly suited to study the effect of RSOC on the interacting phase diagram, as the truncated unity approach retains the details of the involved lattice harmonics of the superconducting pairing. That allows us to easily extract the degree of singlet-triplet mixing of superconducting states (expected when inversion symmetry is broken). By feeding the TUFRG output into a Bogoliubov–de Gennes formalism, we derive the topological properties of the unconventional superconducting phases.

Model.—In the following, we investigate the TLRHM defined as

$$H=-\!\!\sum_{\langle ij\rangle,ss^{\prime}}\!\left[t\delta_{ss^{\prime}}+\alpha\left({\boldsymbol{\sigma}}\times\boldsymbol{\rho}_{ij}\right)_{ss^{\prime}}^{z}\right]c^{\dagger}_{is}c^{\phantom{{\dagger}}}_{js^{\prime}}+U\sum_{i}n_{i{\uparrow}}n_{i{\downarrow}}.$$ (1)

The sum $\braket{ij}$ runs over nearest neighbor sites, $s$, $s^{\prime}$ are spin indices, $t$ is the hopping amplitude and $\alpha$ quantifies RSOC. The bond vector is given by $\boldsymbol{\rho}_{ij}=\boldsymbol{r}_{i}-\boldsymbol{r}_{j}$, ${\boldsymbol{\sigma}}$ is the vector of Pauli matrices, and $n_{is}=c^{\dagger}_{is}c^{\vphantom{\dagger}}_{is}$ is the number operator.

The band structure of (1) is readily derived from the Bloch matrix

$$h_{0}^{ss^{\prime}}(\boldsymbol{k})=\begin{pmatrix}{\varepsilon}_{0}(\boldsymbol{k})&\alpha\left[\frac{\partial{\varepsilon}_{0}}{\partial k_{y}}+i\frac{\partial{\varepsilon}_{0}}{\partial k_{x}}\right]\\ \alpha\left[\frac{\partial{\varepsilon}_{0}}{\partial k_{y}}-i\frac{\partial{\varepsilon}_{0}}{\partial k_{x}}\right]&{\varepsilon}_{0}(\boldsymbol{k})\end{pmatrix}_{ss^{\prime}}$$

with ${\varepsilon}_{0}(\boldsymbol{k})=-2t[\cos(\boldsymbol{a}_{1}\boldsymbol{k})+\cos(\boldsymbol{a}_{2}\boldsymbol{k})+\cos((\boldsymbol{a}_{2}-\boldsymbol{a}_{1})\boldsymbol{k})]$; $\boldsymbol{a}_{1}$ and $\boldsymbol{a}_{2}$ are the primitive lattice vectors.

We analyse the leading many-body instabilities of the Hubbard model through the use of TUFRGMetzner et al. (2012); Platt et al. (2013). FRG interpolates between the bare Hubbard interaction and a low-energy effective two-particle interaction vertex, by means of iteratively integrating a flow equation. In the TU formalism, the interaction vertices are expressed in terms of the lattice harmonics Lichtenstein et al. (2017); Husemann and Salmhofer (2009); Wang et al. (2012), which gives the advantage of explicit formulations of leading instabilites. We can then track changes in these tight-binding parameters with changes in normal state parameters across our phase diagrams.

Superconducting instabilities on the triangular lattice can generically be written as Sigrist and Rice (1987):

$$\hat{\Delta}(\boldsymbol{k})=[\Psi(\boldsymbol{k})\hat{\mathbb{1}}+\boldsymbol{d}(\boldsymbol{k})\cdot{\hat{\boldsymbol{\sigma}}}]i\hat{\sigma}_{y}\,,$$ (2)

where the Pauli matrices $(\hat{\mathbb{1}},\hat{\sigma}_{i})$ act on the spin subspace.

In the absence of RSOC, spin rotational symmetry admits spin-singlet $\Psi$ and triplet $\boldsymbol{d}$ states, and all possible superconducting instabilities are then characterized according to the irreps of the point group $\text{D}_{6h}$.

For finite RSOC, the point group symmetry is reduced to $\text{C}_{6v}$ from broken inversion symmetry. The spins are also “frozen” to the lattice and must rotate with it, i.e., the system is now spin-orbit coupled. Thus the superconducting states can no longer characterized by the spatial irreps but only by total irreps, i.e., space and spin combined. A complete discussion of group theory is presented in the SM; here we wish to emphasize that amongst the total irreps there are three different possibilities for the two-dimensional irrep $E_{2}$, which are permitted to mix among possible tight-binding superconducting instabilities. These three states are: the standard $d$-wave spin-singlet $\{\Gamma_{d_{x^{2}-y^{2}}},\Gamma_{d_{xy}}\}$; and two spin-triplet states $\{(\Gamma_{f},0,0),(0,\Gamma_{f},0)\}$ and $\{(\Gamma_{p_{x}},-\Gamma_{p_{y}},0),(\Gamma_{p_{y}},\Gamma_{p_{x}},0)\}$, where the 3-vector corresponds to $\boldsymbol{d}=(d_{x},d_{y},d_{z})$ .

Results I: competing many-body phases.—In the following, we present our TUFRG results for $U=8t$ and first focus on different types of instabilities. Phase diagram Fig. 1 (a) reveals for zero RSOC ($\alpha=0$) and filling $\nu\geq 0.34$ an extended superconducting phase with $E_{2g}$ irrep (spin-singlet), adjacent to a magnetically ordered phase at lower fillings. For even lower filling around $\nu=0.2$, we find another narrow superconducting phase in the $B_{1u}$ irrep (spin-triplet). The extended $E_{2g}$ phase is not unexpected, as this is the preferred type of unconventional superconductivity on hexagonal lattices with a spherical Fermi surface.

Figure 1: (a) Phase diagram with RSOC ($\alpha$) vs. filling ($\nu$). Colored areas correspond to superconductivity, grey to magnetic order, and white to metal/Fermi liquid. Color corresponds to singlet-triplet weight of the divergent superconducting instability (see main text). The dashed line indicates van-Hove filling. (b, c) Two degenerate superconducting states corresponding to the parameters as indicated by the circle in (a) $\nu=0.5,~{}\alpha=0.4$, realizing a superposition of the three possible $E_{2}$ superconducting states for Cooper-pairs across the nearest-neighbor bonds. The imaginary component of the two superconducting states is zero.

The two-dimensional $E_{2g}$ irrep allows for arbitrary complex superpositions of the two degenerate superconducting states that constitute the diverging instability. Free energy calculations reveal that the equal weight complex superposition “$d+id$ ” has the lowest energy (see SM for explicit calculations), due to the largest condensation energy (as previously found in Refs. Sigrist and Rice, 1987; Sigrist and Ueda, 1991; Kuznetsova and Barzykin, 2005 or often assumed without explicit calculation). We stress that the $d+id$ SC is a topological state with chiral edge states and finite Chern number, as explicitly confirmed below. We note that our $\alpha=0$ results are compatible with previous work Honerkamp (2003); Klebl et al. (2023); Classen et al. (2019).Outside of the shown fillings $\nu$, TUFRG does not diverge, signaling stability of the Fermi liquid phase.

Increasing the RSOC $\alpha$ does not change much on the interplay of superconducting and magnetic phases in the considered range $0\leq\alpha<0.5$ (see Fig. 1 a). We do find commensurate and incommensurate spin-density waves, but leave the detailed analysis of the magnetically ordered phase for future work.

Results II: singlet-triplet mixing.— In line with previous work in Rashba systems Vafek and Wang (2011); Wolf and Rachel (2020); Beyer et al. (2023), we find the fraction of singlet/triplet mixing to increase with increasing RSOC, so long as the SC phase is spin singlet at $\alpha=0$. One of the major advantages of TUFRG is to quantitatively determine this singlet-triplet mixing (see SM for details). In particular, for $0.34\leq\nu\leq 0.44$ we observe significant triplet contributions in the otherwise singlet-dominated phase. For larger values of RSOC, such as $\alpha=0.5$, singlet and triplet contributions are of similar weight. In contrast, the low-filling superconducting phase remains fully spin-triplet regardless of $\alpha$. It is a mixture of the two possible triplet $E_{2}$ superconducting states. We attribute this to the particular band structure (see SM for details): the Fermi energy is in the vicinity of the Kramer’s-degeneracy protected $M$-point, where the dispersion is essentially flat and the effect of Rashba band splitting is suppressed. This allows the persistence of the triplet-dominated superconductivity for finite RSOC, yielding the extended triplet phase. RSOC does, however, change the nature of the triplet state. At $\alpha=0$ we find $B_{1u}$ irrep (gapless), and for any finite $\alpha$ the $E_{2}$ irrep containing $f$-wave and $p$-wave pairings. Roughly speaking, the $p$-wave contribution, responsible for the finite gap, increases with $\alpha$. At $\alpha\sim 0.4$ this $E_{2}$ phase is dominated by $p$-wave pairing. We stress that there is hardly any method which can resolve spin-triplet or triplet-triplet mixing as easily as TUFRG can. To determine whether the mixed-in spin-triplet components trigger any additional topological features we need to push the analysis further.

Results III: BdG analysis, Chern numbers and ribbon spectra.—TUFRG provides us with details about the pairing symmetry and relative strength for shells with different distance (e.g. on-site pairing vs. nearest-neighbor pairing), thanks to its truncated unity extension. In the following, we directly make use of this information and feed it into a Bogoliubov–de Gennes (BdG) Hamiltonian (see SM) where all other bandstructure parameters are kept identical to the TUFRG input. Since the superconducting states are all $E_{2}$ and doubly degenerate, we calculate the free energy to determine the stable ground state configuration Platt et al. (2013); Klebl et al. (2022); Classen et al. (2019). Knowledge of the BdG Hamiltonian and its eigenvectors allows us to compute Chern numbers of the corresponding gapped superconducting phases and compute ribbon (and real space) spectra to reveal topological edge states. For instance, by choosing $\alpha=0$ and $\nu=0.5$ we recover the earlier result Crépieux et al. (2023) that the chiral $d+id$ superconductor ($E_{2}$ irrep) possesses a Chern number $C=4$. The corresponding ribbon spectra is shown in Fig. 2 (a), which at $\alpha=0$ manifests as only two distinguishable edge states due to spin degeneracyBlack-Schaffer (2012); Scherer et al. (2022).

Figure 2: Ribbon spectra from TUFRG and BdG methods. (a) $\alpha=0$, $\nu=0.5$ resulting in $C=4$, as evidenced by two doubly degenerate chiral edge modes per edge. (b) $\alpha=0$, $\nu=0.4$ resulting in $C=-8$, as evidenced by four doubly degenerate chiral edge modes per edge. (c) $\alpha=0.35$, $\nu=0.4$ resulting in $C=-6$, as evidenced by six chiral edge modes per edge. All spectra calculated for a SC amplitude of $|\Delta|=0.05$, on a ribbon of 900 atoms, and a $k$-momentum resolution of 3000 points.

Figure 3: Topological transition in filling at $\alpha=0$. (a) Critical TUFRG scale $\Lambda_{c}$ as a function of filling $\nu$. (b) BdG energy $E$ as a function of filling $\nu$, Chern numbers color-coded relative to Fig. 4. (c) Relative weights in the TUFRG superconducting instability for the $n$-th nearest neighbor pairing lengths as a function of filling. In all plots, the topological transition at filling $\nu=0.445$ is shown by the dashed grey line.

Topological phase transitions.—By virtue of combining the TUFRG and BdG methods, we analyze the entire superconducting phase space. First, we focus on constant RSOC $\alpha$ and vary the filling $\nu$. At $\nu_{c}=0.445$ (irrespective of $\alpha$), we observe a dip in critical scale $\Lambda_{c}$ (see Fig. 3 (a)). We note that this feature should also be observable within other functional renormalization group approaches. We observe this non-analytical behavior also in other quantities such as the gap energy of the BdG Hamiltonian (Fig. 3 (b)), as well as a shift in the spectral weight of the Cooper pairing from second- and third-nearest neighbor bonding to primarily nearest-neighbor. We visualize this by calculating the spectral weight on each interaction shell $w_{n}$ in Fig. 3 (c) (for details see SM). We find that for fillings larger than $\nu_{c}$ the BdG Hamiltonian possess a Chern number $C=4$, however, for fillings smaller than $\nu_{c}$ we obtain $C=-8$; see the corresponding ribbon spectra in Fig. 2 (a, b), revealing a topological phase transition.

We can understand this topological phase transition by focusing on the purely singlet chiral $d+id$-wave superconducting phases at $\alpha=0$. In this limit, the gap in the BdG spectrum closes when the Fermi surface intersects vortices in the superconducting pairing $\Gamma_{d_{x^{2}-y^{2}}}+i\,\Gamma_{d_{xy}}$. The location and number of these vortices depends on the precise mix of the three interaction shell pairings. With a longer pairing, higher harmonics are introduced into the superconducting pairing function, which means more vortices are introduced for longer range pairing. These vortices can then be shifted in reciprocal space by changing the admixture of the superconducting pairings, with the total number changing when vortices recombine and are moved in and out the Brillouin zone. That is, the number of vortices can change and thus the topological invariant can change too. The bottomline is: even in the absence of RSOC, topological phase transitions are to be expected, as demonstrated above, although we find $E_{2}$ irreps on both sides of the phase transition. We note that a similar phase transition was observed at $\alpha=0$ in extended Hubbard and $tJ$ models Scherer et al. (2022); Val’kov et al. (2015); Zhou and Wang (2008). Our work establishes the plain-vanilla onsite Hubbard model as the minimal model to find such a phase transition, as well as its stability towards RSOC.

Figure 4: (a) Topological phase diagram: superconducting phases from Fig. 1 a with their Chern numbers. White regions correspond to the Fermi liquid, grey is magnetic ordering, and black states are gapless. (b) Varying Chern number with increasing RSOC $\alpha$ and superconducting order parameter amplitude $|\Delta|$, for fixed filling $\nu=0.4$.

We also encounter topological phase transitions when varying $\alpha$. For instance, for fixed $\nu=0.4$, we previously found $C=-8$ at $\alpha=0$ (see a corresponding ribbon spectra in Fig. 2 (b)) and for larger $\alpha=0.35$ we find $C=-6$, as corroborated by Fig. 2 (c). The entire analysis of Chern number and edge states from ribbon spectra culminates in the topological phase diagram Fig. 4.

Clear critical signatures as those shown in Fig. 3 (a) are absent, however, the BdG energy gap clearly reveals the topological phase transition when changing $\alpha$. This difference can be traced to the nature of the BdG gap closing. Unlike the purely singlet case, the characteristic polynomial of the BdG matrix has some constant terms, which mean that the BdG energies can have zeros and the BdG gap can close where the superconducting amplitude $|\Delta|$ is not zero; Fig. 4 (b) plots the relationship between $\alpha$ and $|\Delta|$. If we assume that $|\Delta|/t<0.1$ (which is reasonable as we follow the educated guess $|\Delta|\propto\Lambda_{c}$) then this fixes $\alpha_{c}=0.17$ to two significant figures regardless how small $|\Delta|$ will be. This also justifies omitting an additional self-consistent treatment of the magnitude of $|\Delta|$. We then compute topological invariants and derive ribbon spectra resulting in Fig. 4.

Outlook.—There are many examples of studying unconventional superconductivity as the groundstate of correlated electron systems in the literature. In particular, for hexagonal lattices with a circular Fermi surface, the $E_{2}$ irrep is the most likely candidate. While it is well-known that such a state usually corresponds to the chiral $d+id$ superconductor, here we show that the phenomenology is much richer, in particular, in the presence of spin-orbit coupling. We find Chern numbers $C=4,-2,-6,-8$ within an otherwise homogeneous $E_{2}$ irrep phase. Only the employed TUFRG formalism allows to reveal these topological features and the associated topological phase transition within that phase. In the light of ongoing experimental activities with the prospect of unconventional superconductivity in triangular lattice materials such as B-doped Sn/Si(111) and related compounds, we hope that our work will spark further activity for these systems since different topological states and possibly even topological phase transitions might be observable.