Sublattice-Resolved Perturbative Decoupling

• Sublattice-resolved perturbative decoupling is defined by the selective suppression, redirection, or partial restoration of scattering and spin precession channels based on the internal sublattice composition of Bloch states.
• This mechanism, evidenced in the kagome Hubbard model and fluorinated graphene, shows that sublattice weights can override conventional effects inferred solely from Fermi-surface geometry or density of states.
• Longer-range interactions and impurity-induced sublattice pseudospin polarization relax selection rules, thereby enhancing pairing eigenvalues in superconductivity and extending spin lifetimes in resonant regimes.

Searching arXiv for the cited papers to ground the article. arXiv search query: (Kiesel et al., 2012) Sublattice Interference in the Kagome Hubbard Model “Sublattice-resolved perturbative decoupling” (Editor's term) denotes a class of perturbative low-energy effects in which the internal sublattice structure of Bloch or impurity-dressed states selectively suppresses, redirects, or partially restores scattering and precession channels that would otherwise be inferred from Fermi-surface geometry, density of states, or homogeneous spin-orbit coupling alone. In the available arXiv literature, two distinct realizations exemplify this logic. In the kagome Hubbard model near van Hove filling, “sublattice interference” enters the projected interaction vertex through momentum-dependent sublattice weights and suppresses conventional nesting-driven instabilities (Kiesel et al., 2012). In fluorinated graphene, impurity-induced sublattice pseudospin polarization locally reduces the Rashba spin-precession field and produces a local, resonance-enhanced perturbative decoupling of spin and pseudospin dynamics (Tuan et al., 2016).

1. Conceptual definition and scope

The central organizing idea is that sublattice is not merely a bookkeeping label. In both systems, the low-energy quasiparticles carry a sublattice composition that directly enters the perturbative kernel governing the instability or relaxation problem. The consequence is a sublattice-resolved selection effect: processes that are kinematically plausible in momentum space can become weak if the participating states have incompatible sublattice content.

In the kagome case, this selection effect appears in a weak-coupling renormalization-group treatment of the Cooper channel. The three-sublattice wave-function amplitudes $u_s(\mathbf{k})$ enter the projected two-particle vertex explicitly, so the Fermi surface must be analyzed together with sublattice-resolved Bloch weights (Kiesel et al., 2012). In graphene with fluorine ad-atoms, the relevant object is instead a long-range sublattice pseudospin polarization (SPP) around the impurity resonance. There the perturbative suppression acts on spin dynamics: because Rashba SOC couples spin to the in-plane pseudospin texture, impurity-induced SPP weakens the effective Rashba field locally and prolongs spin lifetime on the hole side of the resonance (Tuan et al., 2016).

A common misconception, directly contradicted by these examples, is that weak-coupling outcomes are determined primarily by Fermi-surface topology, DOS enhancement, or the bare magnitude of SOC. The cited works show that wave-function structure at the sublattice level can be equally decisive.

2. Kagome formulation: three-sublattice structure in weak-coupling RG

The kagome Hubbard model considered in the weak-coupling analysis is

$H = H_0 + H_{\text{int}},$

with

$$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$

and

$$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$

Because the kagome lattice has three sites per unit cell, the noninteracting problem is a three-band model. After Fourier transformation with sublattice index $s$,

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$

the diagonal band basis is

$$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$

The key quantities are the sublattice weights

$$u_{sn}(\mathbf{k}), \qquad \sum_s |u_{sn}(\mathbf{k})|^2=1.$$

The perturbative analysis is performed near van Hove filling, which for the kagome lattice occurs at total filling $n=5/12$, equivalently valence-band filling $n_v=1/4$. At this filling, the Fermi surface touches the $H = H_0 + H_{\text{int}},$0 points of the Brillouin zone and is nested by several wave vectors, including

$H = H_0 + H_{\text{int}},$1

The RG treatment follows the standard weak-coupling scheme of Raghu et al. in the Cooper channel. The pairing vertex

$H = H_0 + H_{\text{int}},$2

is evaluated on the Fermi surface in the infinitesimal-coupling limit, with all diagrammatic contributions summed up to second order in $H = H_0 + H_{\text{int}},$3 and $H = H_0 + H_{\text{int}},$4. The leading instability is obtained by diagonalizing the Cooper kernel; if $H = H_0 + H_{\text{int}},$5 is the most negative eigenvalue, then

$H = H_0 + H_{\text{int}},$6

For onsite interaction, the projected vertex is

$H = H_0 + H_{\text{int}},$7

which is the technical point at which the sublattice structure enters the low-energy problem directly (Kiesel et al., 2012).

3. Sublattice interference as a perturbative selection rule

The central phenomenon identified in the kagome Hubbard model is “sublattice interference.” In a single-sublattice metal, nesting vectors connecting Fermi-surface patches usually enhance particle-hole fluctuations. In the kagome case, the Fermi-surface states are not uniformly distributed over the three sublattices; their dominant sublattice character varies along the Fermi surface. Because the onsite Hubbard term is diagonal in sublattice index, the projected interaction is large only when the interacting states have compatible sublattice content.

If a nesting vector connects regions whose Bloch states have mostly different sublattice weights, the overlap in the projected vertex is small. The result is a suppression of the corresponding interaction matrix element. The paper therefore argues that Fermi-surface topology alone would suggest strong nesting at $H = H_0 + H_{\text{int}},$8, but the sublattice-resolved wave functions weaken or forbid many of those connections. Instead, the enhancement is redistributed into six nesting vectors that connect regions with more compatible sublattice composition; one explicit example is the shift of the naive $H = H_0 + H_{\text{int}},$9 channel to

$$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$0

This mechanism explains why the kagome model differs from the honeycomb model even when the Fermi-surface geometry and DOS appear very similar. In the honeycomb case, the sublattice weights are essentially homogeneous along the Fermi surface, so there is no comparable interference effect. The broader significance is that the kagome lattice behaves, in the authors’ physical interpretation, like a multi-orbital system with “orbitals” replaced by sublattices. Momentum-space nesting alone is therefore insufficient to determine the leading instability (Kiesel et al., 2012).

4. Consequences for superconductivity and interaction range

Near van Hove filling, the kagome Hubbard model still favors chiral $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$1 superconductivity. The weak-coupling analysis finds two degenerate $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$2-wave eigenvalues $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$3, which in mean field combine into the topological $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$4 state. The decisive result, however, concerns scale rather than symmetry: the pairing strength is strongly reduced relative to the honeycomb case. At van Hove filling,

$$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$5

where $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$6 and $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$7 are the leading pairing eigenvalues in the kagome and honeycomb systems, respectively. Since

$$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$8

even a moderate suppression of $$H_0 = t \sum_{\langle i,j\rangle}\sum_\sigma \left(c_{i\sigma}^\dagger c_{j\sigma}^{\phantom{\dagger}} + \text{h.c.}\right) + \mu \sum_{i,\sigma} n_{i,\sigma},$$9 produces a large reduction of $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$0.

This is the principal weak-coupling consequence of sublattice interference: it does not necessarily change the leading pairing symmetry, but it substantially weakens the effective attraction in that channel. The paper therefore states that the kagome sublattice structure is not a spectator; it actively reshapes the character of the Fermi-surface instabilities (Kiesel et al., 2012).

A second result is explicitly described as anomalous. Ordinarily, adding longer-range repulsion $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$1 is expected to frustrate pairing and reduce $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$2. In the kagome model, increasing $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$3 can instead increase the pairing eigenvalue $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$4, and hence raise $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$5, within the weak-coupling RG analysis. The stated reason is that $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$6 makes the interaction vertex no longer purely diagonal in sublattice index: it acquires momentum dependence and mixes sublattices more effectively. This relaxes the sublattice selection rule that suppressed nesting-enhanced scattering in the onsite-only problem. A plausible implication is that longer-range interactions partly compensate for the sublattice mismatch that the onsite term alone cannot overcome.

The authors further conjecture that the suppression of conventional Fermi-liquid instabilities makes the kagome Hubbard model a prototype candidate for hosting exotic electronic states of matter at intermediate coupling.

5. Graphene realization: impurity-induced pseudospin polarization and local decoupling

In graphene, the relevant sublattice degree of freedom is the A/B pseudospin. The cited work analyzes fluorine ad-atoms that chemisorb on one carbon site, denoted $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$7, and thereby break A–B sublattice symmetry without the formation of strong magnetic moment. The central physical effect is the formation of a resonant impurity state whose local density of states is sublattice asymmetric: the sublattice hosting the fluorinated carbon is less occupied than the opposite sublattice, and the imbalance extends over many lattice sites. This long-range SPP is visible in LDOS maps over hundreds of atoms. The resonant level is reported at approximately

$$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$8

with ab initio data also indicating a resonance around $$H_{\text{int}} = U_0 \sum_i n_{i\uparrow}n_{i\downarrow} +\frac{U_1}{2}\sum_{\langle i,j\rangle,\sigma,\sigma'} n_{i\sigma}n_{j\sigma'}.$$9 meV below the Dirac point (Tuan et al., 2016).

The microscopic model is

$s$0

with homogeneous graphene SOC

$s$1

and fluorine impurity terms

$s$2

The fitted parameters quoted in the source are $s$3 eV, $s$4 eV, $s$5 meV, $s$6 meV, and $s$7 meV.

The continuum form

$s$8

makes the mechanism transparent: Rashba SOC couples spin to the in-plane pseudospin components $s$9. Around the fluorine resonance, impurity-induced SPP suppresses the local pseudospin texture that normally feeds the Rashba field. The paper therefore describes a local decoupling of spin and pseudospin dynamics near the impurity resonance. Importantly, it is not full decoupling everywhere in the sample, but a local, resonance-enhanced perturbative decoupling. The transport signature is a large electron-hole asymmetry in spin lifetime, tunable electrostatically from hundreds of picoseconds to several nanoseconds in the dilute impurity limit (Tuan et al., 2016).

6. Spin relaxation, anisotropy, and unifying interpretation

The graphene analysis computes the time-dependent spin polarization as

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$0

with exponential decay

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$1

The dominant relaxation mechanism remains Dyakonov–Perel,

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$2

and the computed anisotropy satisfies approximately

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$3

with deviations near the Dirac point where spin-pseudospin entanglement is strongest. Near the fluorine resonance the numerics are also compared with

$$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$4

The reported concentrations include $$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$5, $$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$6, and a $$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$7 supercell around $$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$8. On the hole side, close to the resonant impurity energy, the spin lifetime reaches the ns regime; on the electron side it is much shorter, around $$c_{i,s,\sigma}^\dagger=\sum_{\mathbf{k},n} u_{sn}^*(\mathbf{k})\,c_{\mathbf{k},n,\sigma}^\dagger e^{-i\mathbf{k}\cdot(\mathbf{R}_i+\mathbf{r}_s)},$$9 ps. The lifetime maximum is shifted slightly to the hole side relative to $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$0, and the shift grows with fluorine concentration; the source attributes this to fluorine-induced SOC terms, because turning off those impurity SOC terms moves the peak back to $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$1 (Tuan et al., 2016).

Taken together with the kagome result, these findings suggest a unifying interpretation. In both cases, perturbative low-energy behavior is filtered by sublattice structure. In the kagome Hubbard model, the filter acts on interaction vertices and nesting-enhanced scattering. In fluorinated graphene, it acts on the pseudospin texture that mediates Rashba spin precession. The former suppresses conventional particle-hole enhancement and lowers the $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$2 superconducting scale; the latter suppresses the effective precession field and increases spin lifetime on the resonant hole side. A common misconception is that added longer-range structure must always increase frustration or disorder. The kagome work shows that $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$3 can increase $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$4 by mixing sublattices more effectively, and the graphene work shows that fluorine-induced SOC does not simply shorten $$H_0=\sum_{\mathbf{k},\sigma,n}\epsilon_n(\mathbf{k})\, c_{\mathbf{k},n,\sigma}^{\dagger} c_{\mathbf{k},n,\sigma}.$$5, because the long-range SPP can reduce the effective Rashba field instead (Kiesel et al., 2012, Tuan et al., 2016).