
2
spin-orbit-split Fermi sheets. To demonstrate the impact
of the tunable SOC on the correlated physics of tTLG, we
compute the superconducting order parameter as a func-
tion of D0assuming different dominant pairing states in
the parent tTLG system. We see that D0not only allows
to drive superconducting phase transitions and tune the
order parameter, it also has the potential of being used
as a tool to probe the nature of pairing in tTLG [19–26],
which is currently under debate.
Non-interacting bandstructure.—To model the non-
interacting bandstructure of the system in Fig. 1we use
an appropriate generalization of the continuum-model
description of twisted bilayer graphene [55–60]. The as-
sociated Hamiltonian is diagonal in the graphene layers’
valley degree of freedom, η=±, and consists of four
terms, hη=h(g)
η+h(t)
η+h(SOC)
η+h(D)
η[61]. Here h(g)
η(∇),
∇= (∂x, ∂y), and h(t)
η(r),r= (x, y), capture, respec-
tively, the Dirac cones, rotated by (−1)lθ/2and with op-
posite chirality for η=±, of the three layers, l= 1,2,3,
and the tunneling between them, respectively. The lat-
ter is spatially modulated on the emergent moiré scale
and reconstructs the graphene cones into moiré bands.
As a result of the mirror symmetry [58], σh, the band-
structure of h(g)
η+h(t)
ηis that of twisted bilayer graphene
(σh-even) and single-layer graphene (σh-odd bands) in
valley η. Surrounding tTLG with TMD layers as shown
in Fig. 1induces SOC [47–52] in the outer two layers as
described by
h(SOC)
η=X
l=1,3
(−1)(l−1)
2Pl[λIszη+λR(ηρxsy−ρysx)] ,
where Plprojects onto the lth graphene layer and ρj
(sj), j=x, y, z, are Pauli matrices in sublattice (spin)
space. The relative minus sign between the l= 1,3
terms in h(SOC)
ηis dictated by the inversion symmetry,
I, of the heterostructure. Note that h(SOC)
ηnot only
breaks spin-rotation symmetry, but also σh, which can be
seen in the bandstructure shown Fig. 2(a): the graphene
Dirac cone of the σh-odd sector, located around the K0
point for the valley shown, hybridizes with the σh-even,
twisted-bilayer graphene sector. Despite the presence of
SOC, all bands are still doubly degenerate, which follows
from the momentum-space-local, anti-unitary symmetry
IΘs, with Θsbeing spin-1/2time-reversal, that obeys
(IΘs)2=−.
Applying a displacement field D0—as is routinely done
in experiments on tTLG [19–26] and captured by the
last term, h(D)
η, in the Hamiltonian—breaks Iand IΘs.
As such, the pseudospin-degeneracy of the bands at
D0= 0 will be removed when D06= 0, see Fig. 2(b),
providing direct experimental control over the inversion-
antisymmetric SOC terms, gk6= 0, in the effective Hamil-
tonian, heff
k,η =s0ξη·k+ηgη·k·s, for the bands of tTLG
near the Fermi level. To further discuss the form of gk,
FIG. 2: Band structure, including Fermi surfaces and spin
texture in valley η= +. Here twist angle θ= 1.75◦,
µ=µ∗'16 meV, and λR= 20 meV. The cut through the
band structure in (a) at D0= 0 shows spin-degenerate
bands. The degeneracy is removed when D06= 0 as can be
seen in (b), where D0= 20 meV. Using the same parameters
as in (b), the direction of the two non-zero components of gk
in (c) reveals three vortices (+) and three anti-vortices (−).
When µ=µ∗, we obtain vanishing Fermi-surface splitting,
see (d), at their Cs
3z-related k-space locations, leading to a
Möbius winding of the spin texture (arrows).
let us focus on the limit θTMD = 30◦, where λIvanishes
by symmetry [48–51] and (gk)z= 0 as a consequence of
Cs
2zΘs(Cs
nz is the n-fold rotation symmetry along z). As
can be seen in Fig. 2(c), the remaining two components
of gkexhibit vortices at three, Cs
3z-related, generic posi-
tions k=k∗
j,j= 1,2,3, in the Brillouin zone, compen-
sating those with opposite chirality at Γ,Kand K0. Since
these vortices—and, thus, the associated zeros of |gk|and
type-II Dirac cones in the band structure—cannot be adi-
abatically removed, we are guaranteed to find a chemical
potential µ=µ∗where the Fermi surface of ξkcrosses all
three k∗
j. For µ=µ∗(µ'µ∗), the spin-orbit splitting
of the Fermi surfaces of the system has to (almost) van-
ish at these three points, see Fig. 2(d). Having an odd
number of points on the Fermi surface where the spin
splitting (almost) vanishes also leads to exotic spin tex-
tures: following the spin polarizations of the Bloch states
[also shown in Fig. 2(d)] diabatically along the Fermi sur-
face, one ends up on the other Fermi sheet after one full
revolution, akin to an object traversing a Möbius strip.
For this reason and due to the related, but not identical
concept of “Möbius fermions” [62], which occur as edge