Fu–Kane Scheme: Z2 Topology in Time-Reversal Systems

• Fu–Kane scheme is a framework defining Z2 topological invariants through time-reversal symmetric Bloch bundles and the obstruction to forming global Kramers pairs.
• It employs sewing matrices, Berry curvature and connection integrals, and topological field theory (Wess–Zumino and Chern–Simons actions) to classify phases.
• The approach extends to interacting and magnetic systems as well as semimetal settings, linking conventional band topology with Majorana physics in superconducting heterostructures.

Searching arXiv for recent and foundational papers on the Fu–Kane scheme. The Fu–Kane scheme denotes two closely related constructions associated with topological phases protected by time-reversal symmetry. In band-theoretic form, it is the Fu–Kane–Mele $\mathbb{Z}_2$ scheme for class AII insulators, where a sewing matrix built from a time-reversal-symmetric Bloch frame yields a bulk $\mathbb{Z}_2$ invariant that detects the obstruction to a global Kramers-pair frame (Gawedzki, 2016). In superconducting form, it is the Fu–Kane heterostructure on the surface of a three-dimensional topological insulator, where proximity-induced $s$-wave pairing and magnetic interfaces realize a spinful topological superconductor supporting Majorana modes (Vela et al., 2021). The mathematical literature has recast the insulating version in terms of obstruction theory, Wess–Zumino actions, Chern–Simons functionals, and equivariant cohomology, while more recent work has extended it to interacting many-body systems and to magnetic or semimetallic settings (Monaco, 2017, Monaco et al., 2016, Bachmann et al., 2024, Gao et al., 2022).

1. Band-theoretic definition and the original $\mathbb{Z}_2$ construction

In the insulating setting, one considers a family of Bloch Hamiltonians $H(k)$ on a Brillouin torus $\mathbb{T}^d$, typically with $d=2$ or $3$, satisfying lattice periodicity and fermionic time-reversal symmetry implemented by an antiunitary $\theta$ with $\theta^2=-I$ (Gawedzki, 2016). With a global Fermi energy $\mathbb{Z}_2$0 in a spectral gap, the occupied states define a valence bundle $\mathbb{Z}_2$1, and in $\mathbb{Z}_2$2 this bundle is trivializable and of even rank $\mathbb{Z}_2$3 (Gawedzki, 2016). The key topological issue is not the existence of a global frame as such, but whether one can choose it as Kramers pairs.

For $\mathbb{Z}_2$4, Kane and Mele identified a $\mathbb{Z}_2$5-valued obstruction to choosing a global frame obeying

$\mathbb{Z}_2$6

for all $\mathbb{Z}_2$7 (Gawedzki, 2016). If such a frame exists, the insulator is topologically trivial; if not, the nontrivial value signals the quantum spin Hall phase (Gawedzki, 2016). Monaco later reformulated the same point in obstruction-theoretic language: the Chern number is the obstruction to a continuous periodic Bloch frame, whereas the Fu–Kane–Mele invariant is the obstruction to imposing, in addition, the time-reversal constraint on that frame (Monaco, 2017).

Given a global orthonormal valence frame $\mathbb{Z}_2$8, Fu and Kane define the sewing matrix

$\mathbb{Z}_2$9

which is unitary and satisfies

$s$0

(Gawedzki, 2016). At time-reversal invariant momenta $s$1, the matrix is antisymmetric, so its Pfaffian is defined. The standard two-dimensional Fu–Kane formula is

$s$2

with the product taken over the four TRIM in $s$3 (Gawedzki, 2016). In three dimensions, the strong invariant is given by the analogous product over the eight TRIM in $s$4 (Gawedzki, 2016).

A central structural fact is that, in two dimensions with fermionic time-reversal symmetry, the total Chern number vanishes: $s$5 so the occupied bundle is trivial as a complex bundle, yet it may still be nontrivial as a time-reversal-symmetric or Quaternionic bundle (Monaco, 2017, Monaco et al., 2016). This is the precise sense in which the Fu–Kane scheme supplies a secondary, rather than primary, topological invariant.

2. Sewing matrices, effective Brillouin zones, and obstruction theory

The sewing-matrix formula is only one realization of the invariant. Fu and Kane also gave an equivalent expression in terms of Berry connection and Berry curvature over a half Brillouin zone. In Monaco’s notation, for a half-torus $s$6,

$s$7

where $s$8 is the abelian Berry connection in a boundary time-reversal-symmetric gauge and $s$9 is the Berry curvature (Monaco, 2017). Monaco shows that this quantity equals the parity of the degree of a boundary gauge transformation $\mathbb{Z}_2$0, thereby identifying the Fu–Kane invariant as a mod-2 obstruction to extending a boundary-compatible frame into the effective Brillouin cell (Monaco, 2017).

This obstruction picture clarifies several standard statements. First, a continuous $\mathbb{Z}_2$1-equivariant Bloch frame exists whenever the Chern number vanishes, but a continuous, $\mathbb{Z}_2$2-equivariant, time-reversal-symmetric frame exists iff the $\mathbb{Z}_2$3 invariant vanishes (Monaco, 2017). Second, allowed gauge transformations preserving the boundary time-reversal conditions shift the associated degree only by even integers, so only its parity is gauge invariant (Monaco, 2017). Third, the invariant is deformation-invariant under gap-preserving, symmetry-preserving homotopies (Monaco, 2017).

A later refinement proves a decomposition theorem for time-reversal-symmetric projection-valued maps: any such map can be split into two rank-$\mathbb{Z}_2$4 projectors related by time reversal, and the Fu–Kane–Mele index is the parity of the Chern number of one factor (Ferreri et al., 13 Nov 2025). In that formulation,

$\mathbb{Z}_2$5

so the $\mathbb{Z}_2$6 obstruction is concentrated in a single time-reversal-related factor (Ferreri et al., 13 Nov 2025). The same work shows that, in the nontrivial phase, the obstruction can be localized in a single pseudo-periodic Kramers pair, while the remaining pairs may be chosen periodic and smooth (Ferreri et al., 13 Nov 2025). This suggests a more granular geometric picture of the nontrivial phase than the original Pfaffian formula alone.

3. Wess–Zumino, Chern–Simons, and cohomological reformulations

A major mathematical development is the recasting of the Fu–Kane–Mele invariant as a topological field-theoretic action. Gawędzki, Reyes-Lega, and Walcher proved that for a two-dimensional time-reversal-invariant insulator,

$\mathbb{Z}_2$7

where $\mathbb{Z}_2$8 is the properly normalized Wess–Zumino action of the sewing matrix field $\mathbb{Z}_2$9 (Gawedzki, 2016). After removing the determinant and passing to $H(k)$0, the Wess–Zumino amplitude localizes to a product over TRIM,

$H(k)$1

which reproduces the Fu–Kane Pfaffian formula (Gawedzki, 2016). The proof uses bundle gerbes, an involution $H(k)$2 on $H(k)$3, and localization of gerbe holonomy from the full torus to its time-reversal fixed boundary data (Gawedzki, 2016).

The same paper derives directly the known three-dimensional relation between the strong Fu–Kane–Mele invariant and the Chern–Simons action of the non-Abelian Berry connection. If $H(k)$4 is the Berry connection of a global valence frame on $H(k)$5, then

$H(k)$6

with

$H(k)$7

(Gawedzki, 2016). The two-dimensional Wess–Zumino identity on the slices $H(k)$8 is the key step in this derivation (Gawedzki, 2016). This makes explicit the dimensional hierarchy between two-dimensional $H(k)$9 topology and three-dimensional strong topology.

Monaco and Tauber then established a direct bridge between the Berry/half-Brillouin-zone formula and the Wess–Zumino description by showing that the boundary Wess–Zumino amplitudes equal Berry phases and that their square roots agree in the time-reversal-equivariant setting (Monaco et al., 2016). Their analysis relies on an equivariant adjoint Polyakov–Wiegmann formula for $\mathbb{T}^d$0-valued fields on $\mathbb{T}^d$1, avoiding the full gerbe formalism while retaining the same topological content (Monaco et al., 2016). In that language, the Fu–Kane scheme becomes a statement about a Quaternionic gauge bundle on $\mathbb{T}^d$2 whose $\mathbb{T}^d$3 invariant admits both a Berry-theoretic and a Wess–Zumino-theoretic realization (Monaco et al., 2016).

A complementary perspective is cohomological. De Nittis and Gomi identified the Fu–Kane–Mele index with the FKMM invariant of Quaternionic vector bundles, a class in relative equivariant cohomology

$\mathbb{T}^d$4

which for time-reversal tori reduces to the usual Fu–Kane–Mele $\mathbb{T}^d$5 invariant (Nittis et al., 2016). In dimensions $\mathbb{T}^d$6, this class gives a full classification of Quaternionic bundles over a broad class of involutive spaces, and for $\mathbb{T}^d$7 one recovers the single $\mathbb{T}^d$8 phase of the quantum spin Hall effect (Nittis et al., 2016). This cohomological formulation makes the invariant an obstruction to extending a canonical determinant trivialization from the fixed-point set $\mathbb{T}^d$9 to all of $d=2$0 (Nittis et al., 2016).

4. Extensions beyond free-band, gapped insulators

The original scheme was formulated for free fermions, but later work extended it substantially. A 2024 many-body construction defines a $d=2$1-valued index for symmetric, stably short-range-entangled states of two-dimensional interacting fermionic lattice systems with charge conservation and fermionic time reversal (Bachmann et al., 2024). The defining criterion is defect-based: the index is nontrivial precisely when the $d=2$2-flux insertion state, the “fluxon,” transforms under time reversal as part of a Kramers pair (Bachmann et al., 2024). In the quasi-free limit, this many-body index agrees with the Fu–Kane–Mele index defined by the mod-2 spectral flow under flux insertion (Bachmann et al., 2024).

This fluxon formulation makes concrete one of the physical meanings of the Fu–Kane scheme. In a nontrivial phase, adiabatic insertion of $d=2$3 flux binds localized modes that form a Kramers doublet at half filling; in a trivial phase, the defect is a Kramers singlet (Bachmann et al., 2024). The index is invariant under symmetry-preserving locally generated automorphisms and multiplicative under stacking, so it functions as a bona fide interacting SPT invariant (Bachmann et al., 2024).

The scheme also admits generalizations into semimetallic settings. In time-reversal-invariant Weyl semimetals, two-dimensional Fu–Kane–Mele indices remain definable on time-reversal-invariant planes that avoid the Weyl nodes (Thiang et al., 2017). Deforming such a plane across a Weyl point changes its $d=2$4 index by $d=2$5, where $d=2$6 is the Weyl charge, motivating the description of Weyl points as “Fu–Kane–Mele monopoles” (Thiang et al., 2017). In that framework, the history and connectivity of Weyl-point creation and annihilation encode how insulating phases with different Fu–Kane–Mele indices are connected through semimetallic intermediates (Thiang et al., 2017). A plausible implication is that the Fu–Kane scheme is not confined to fully gapped band topology but also organizes the topology of symmetry-preserving gapless phases.

A different line of extension appears in magnetic band theory. By constructing irreducible corepresentations, compatibility relations, and magnetic band representations for type-III and type-IV magnetic space groups, Zhang and collaborators derived Fu–Kane-like symmetry-indicator formulas expressed purely in terms of high-symmetry-point data (Gao et al., 2022). These formulas generalize the parity-based Fu–Kane logic from nonmagnetic inversion-symmetric insulators to magnetic topological materials with antiunitary crystalline symmetries (Gao et al., 2022).

5. The superconducting Fu–Kane heterostructure and Majorana physics

The second major meaning of the Fu–Kane scheme is the superconducting construction on the surface of a three-dimensional topological insulator. Fu and Kane considered a single helical surface Dirac cone proximized by an $d=2$7-wave superconductor; because of spin–momentum locking, the induced pairing behaves as effective spinless $d=2$8 superconductivity (Grein et al., 2011). In Nambu space, the surface Bogoliubov–de Gennes Hamiltonian takes the standard Dirac-plus-pairing form, and this effective two-dimensional topological superconductor supports Majorana zero modes in vortices and dispersive Majorana modes at appropriate boundaries (Grein et al., 2011).

A microscopic study of a Bi$d=2$9Se$3$0-based topological-insulator/superconductor interface showed that the Fu–Kane surface model is a good low-energy approximation only for energies $3$1, where $3$2 is the superconducting bulk gap (Grein et al., 2011). Near and above $3$3, the interface dispersion is strongly modified and the surface state hybridizes with the superconducting continuum, so the simple effective model breaks down (Grein et al., 2011). The same work found that the induced gap can become comparable to $3$4, but also that excessively strong coupling shifts the interface-state weight into the superconductor, making an intermediate coupling regime most favorable for realizing the Fu–Kane scenario (Grein et al., 2011).

Within this superconducting setting, a particularly important geometry is the Josephson trijunction on a topological-insulator surface. The local Fu–Kane criterion is that a trijunction hosts a Majorana zero mode at its center when the minigap

$3$5

is negative in an odd number of the three constituent Josephson junctions (Zhang et al., 2 Nov 2025). This phase-only control architecture underlies later proposals for topological quantum computation with Majorana surface codes on topological-insulator surfaces (Zhang et al., 2 Nov 2025).

Recent experiments on envelope-shaped multi-trijunction devices built on a topological-insulator surface report migration of in-gap states consistent with these Fu–Kane expectations and interpret the measured trajectories as a preliminary exchange operation of Majorana zero modes (Zhang et al., 2 Nov 2025). The same platform has also been used to study coupling between adjacent trijunctions: in regions of phase space where two trijunctions would independently host Majoranas, the observed reopening of the minigap is consistent with hybridization of the two zero modes (Wang et al., 9 Jul 2025). These works do not constitute a full demonstration of non-Abelian braiding, but they extend the Fu–Kane scheme from static Majorana localization to network-level manipulation protocols (Zhang et al., 2 Nov 2025, Wang et al., 9 Jul 2025).

Another variant of the heterostructure places a magnetic insulator adjacent to a superconducting region on the topological-insulator surface, generating a chiral Majorana edge mode at the interface (Vela et al., 2021). In this spinful topological superconductor, a supercurrent along the magnetic boundary produces a Doppler term $3$6 in the Dirac–BdG Hamiltonian and can invert the chirality of a Majorana edge mode when

$3$7

well below the bulk gap-closing threshold $3$8 (Vela et al., 2021). The predicted signatures are a doubling of thermal conductance from $3$9 to $\theta$0 and the emergence of a charge-carrying Dirac mode on the inverted edge (Vela et al., 2021). The authors emphasize that this chirality inversion is absent in the spinless chiral $\theta$1-wave case, so it is specific to the spinful Dirac structure of the Fu–Kane heterostructure (Vela et al., 2021).

6. Conceptual significance, related invariants, and persistent misconceptions

A recurring misconception is that the Fu–Kane scheme is merely a computational shortcut based on parity eigenvalues or Pfaffians. The broader literature shows instead that it encodes a robust obstruction theory for Quaternionic Bloch bundles, with equivalent expressions in terms of sewing matrices, Berry phases, Wess–Zumino amplitudes, Chern–Simons actions, and equivariant characteristic classes (Monaco, 2017, Gawedzki, 2016, Nittis et al., 2016). Another misconception is that the nontrivial $\theta$2 phase requires a nontrivial complex vector bundle. In fact, time-reversal symmetry forces the first Chern number to vanish in two dimensions, so the occupied bundle is trivial as a complex bundle; the nontriviality lies in the failure of a global time-reversal-compatible trivialization (Monaco, 2017, Monaco et al., 2016).

The entanglement-based literature offers another useful comparison. For the Kane–Mele model, the entanglement spin Chern numbers reproduce the same phase diagram as the Fu–Kane $\theta$3 index in the time-reversal-symmetric regime: the quantum spin Hall phase has entanglement spin Chern numbers $\theta$4, whereas the ordinary insulator has $\theta$5 (Araki et al., 2016). Once time reversal is broken by a Zeeman term, the Fu–Kane index is no longer defined, but the entanglement spin Chern numbers can remain meaningful and obey the sum rule that their sum equals the physical Chern number whenever the entanglement spectrum remains gapped (Araki et al., 2016). This suggests that the Fu–Kane scheme is best viewed as one member of a broader family of symmetry-constrained topological obstructions rather than as an isolated construction.

The relation to classification theory is likewise precise. Real $\theta$6-theory, Quaternionic bundles, and equivariant homotopy all place the two-dimensional class AII phase in a $\theta$7 class, and the Fu–Kane–Mele invariant provides an explicit representative of that class (Nittis et al., 2016, Ferreri et al., 13 Nov 2025). In three dimensions, the same logic yields one strong and three weak $\theta$8 indices, which appear cohomologically for $\theta$9 as $\theta^2=-I$0 (Nittis et al., 2016). This suggests that the persistence of the Fu–Kane scheme across formulations is not accidental but reflects a stable topological datum visible in several mathematical languages.

Taken together, these developments establish the Fu–Kane scheme as both a concrete computational framework and a unifying conceptual structure. In its insulating form, it identifies the quantum spin Hall phase through a time-reversal-sensitive $\theta^2=-I$1 obstruction (Gawedzki, 2016). In its superconducting form, it realizes Majorana physics on topological-insulator surfaces through proximity pairing and phase-engineered junctions (Grein et al., 2011, Zhang et al., 2 Nov 2025). In later extensions, it survives reformulation into interacting many-body indices, magnetic symmetry indicators, and semimetal topology (Bachmann et al., 2024, Gao et al., 2022, Thiang et al., 2017).