Topological crystalline superconductors with linearly and projectively represented Cn

symmetry
Chen Fang
Beijing National Laboratory for Condensed Matter Physics and Institute of Physics,
Chinese Academy of Sciences, Beijing 100190, China

B. Andrei Bernevig
Department of Physics, Princeton University, Princeton NJ 08544

Matthew J. Gilbert
arXiv:1701.01944v1 [cond-mat.supr-con] 8 Jan 2017

Micro and Nanotechnology Laboratory, University of Illinois, Urbana IL 61801

Department of Electrical and Computer Engineering, University of Illinois, Urbana IL 61801 and
Department of Electronics Engineering, University of Rome ”Tor Vergata”, Rome, Italy 00133
(Dated: January 10, 2017)
We study superconductors with n-fold rotational invariance both in the presence and in the
absence of spin-orbit interactions. More specifically, we classify the non-interacting Hamiltonians
by defining a series of Z-numbers for the Bogoliubov-de Gennes (BdG) symmetry classes of the
Altland-Zimbauer classification of random matrices in 1D, 2D, and 3D in the presence of discrete
rotational invariance. Our analysis emphasizes the important role played by the angular momentum
of the Cooper pairs in the system: for pairings of nonzero angular momentum, the rotation symmetry
may be represented projectively, and a projective representation of rotation symmetry may have
anomalous properties, including the anti-commutation with the time-reversal symmetry. In 1D
and 3D, we show how an n-fold axis enhances the topological classification and give additional
topological numbers; in 2D, we establish a relation between the Chern number (in class D and CI)
and the eigenvalues of rotation symmetry at high-symmetry points. For each nontrivial class in 3D,
we write down a minimal effective theory for the surface Majorana states.

I. INTRODUCTION significant step forward when the Heusler class of materi-

als were predicted as topological candidates29–32 . While
Condensed matter physics has, in recent years, been the specific focus of the original work29,30 had been to
partly focused on the search for new materials that har- explore the existence of additional TRS topological band
bor topological states. A topological state is a gapped insulator candidates in the Heusler class of materials, the
many-body state that cannot be adiabatically connected Heusler compounds exhibit an extremely wide range of
to the atomic limit while preserving a certain symme- physical phenomena such as ferromagnetism33 , which ex-
try group, and yet cannot be associated with any local pressly breaks TRS, and heavy fermion behavior34 . This
order parameter. In place of order parameters, topologi- work provided early indications that multiple symmetries
cal numbers, a global quantity contributed to by all the can be present in materials and establish different topo-
electrons in system, distinguish a topological state from logical orders. Fu35 was the first to provide an explicit
a trivial one. Initial efforts have focused on the eluci- proof that, in 2D and 3D band insulating systems, the
dation of topological states whose existence and global presence of rotational symmetry in the underlying lat-
properties are stabilized by the presence of time-reversal tice structure, namely C4 , together with time-reversal
symmetry (TRS). This search has led to the examination symmetry give rise to a new Z2 classification for such
of a plethora of 2D1–10 and 3D non-interacting band insu- insulators even in the absence of spin-orbit coupling.
lating systems11–25 under a wide variety of experimental Such materials which have topological states whose exis-
conditions seeking to explore the fundamental spin and tence is guaranteed by the presence of underlying crys-
charge behavior of TRS topological band insulators. Yet talline symmetry are commonly referred to as topolog-
the underlying principles of symmetry preservation re- ical crystalline insulators (TCI). From this early work,
quired for the stability of a topological phase within a the search for topological materials beyond those with
given host material are quite general in nature thereby TRS has since been expanded. Predictions of other
making the TRS class of topological non-interacting band types that can be generally classified as topological crys-
insulators but one in a long list of candidate topological talline systems have emerged, including that of inversion
materials. Therefore, while we understand a great deal symmetric topological materials36–38 , and of rotation-
about the physical nature of TRS topological band in- ally invariant topological band insulators39 . Topologi-
sulators, we are at but the beginning in the search for cal crystalline insulators protected by mirror reflection
topological materials26–28 . symmetry40–45 and glide reflection symmetries46–51 have
been theoretically studied and some have been confirmed
From theoretical point of view, the discussion of topo- in experiments52–55 . Most recently, predictions of topo-
logical materials beyond those that preserve TRS took a
 2

logical semimetals whose band crossings are protected their implications.

by rotational invariance56–58 have also seen experimental
confirmation in Na3 Bi59 and in Cd3 As2 60,61 .
II. PRELIMINARIES
One can further consider the case where particle num-
ber is no longer conserved - that is, the case of su-
perconductors, represented by a Bogoliubov-de Gennes A. Enhancement of the AZ Classification by Local
(BdG) Hamiltonian. Ref.[62–64] studied the topologi- Unitary Symmetries
cal classifications of fully gapped superconductors in all
dimensions both in the presence and absence of time- The AZ classification of single particle Hamiltonians76
reversal symmetry and spin-rotation symmetry. This ini- is based on the transformation properties of the sin-
tial work has been followed by further research into topo- gle particle Hamiltonian under two local symmetries,
logical phases in superconductors containing various sym- namely, particle-hole symmetry (PHS, or P ) and time-
metries such as: TRS superconductors65–67 , reflection reversal symmetry (TRS, or T ), and their composi-
symmetric superconductors68,69 , non-centrosymmetric tion which is called chiral symmetry. A non-interacting
superconductors64 , topological superfluid 3 He-B65,70 , Hamiltonian here refers to a Hamiltonian that only has
and Weyl superconductors71,72 . Nevertheless, the su- quadratic couplings among the creation and annihilation
perconducting phase of materials with general rota- operators (in the second quantized form), and can be
tional invariance has remained relatively unstudied (yet put into a matrix, H, in the orbital basis (or Nambu ba-
see Ref.[73] for a discussion of superconductors with sis when charge is not conserved) in the first quantized
twofold symmetries). Rotational symmetries are im- form. The two local antiunitary symmetries are, given a
portant in understanding the superconducting behavior basis, represented by KT and KP , where T, P are uni-
demonstrated in the Heusler alloys LaBiPt74 and, most tary matrices and K complex conjugation. A Hamilto-
recently, in YPtBi75 . Hence there is a need for theoreti- nian H is said to have T if and only if [KT, H] = 0 and
cal elucidation of the possible corresponding topological have P if and only if {KP, H} = 0. We remark that these
nature of such superconducting systems. formulas apply in the first quantized form, where all op-
In this work we study topological superconductors that erators are represented by matrices, while in the second
possess Cn rotational symmetry with and without spin- quantized form, where operators are expanded in terms
orbital interactions, and ask if the presence of Cn symme- of fermion annihilation and creation operators, both P
try can stabilize additional topological crystalline super- and T should commute with H, and P becomes a unitary
conductors. We also explore how the Cn symmetry places operator. Hereafter, we use hatted symbols for second
constraints on the invariants of the original BdG classes. quantized operators, and non-hatted ones for their first
We ask this question broadly by considering the behavior quantized forms.When both P and T are present, we can
of each of the 4 distinct Bogoliubov- de Gennes (BdG) define S = KP ∗ KT = P ∗ ∗ T such that {S, H} = 0,
symmetry classes, namely class C, D, CI and DIII, of and we say the system has chiral symmetry S. However,
random matrices from the Altland and Zimbauer (AZ)76 there are cases where S is a symmetry, i.e., {S, H} = 0,
in 1D, 2D, and 3D. The answer to such a question has while neither KP nor KT is a symmetry.
direct relevance to a wide range of unconventional su- Now we consider a local unitary symmetry added to
perconducting materials such as cuprates and iron-based the system, represented by some unitary matrix L, which
superconductors. In Section II, we discuss the necessary generically satisfies
general background to understand the subsequent anal- [L, H] = 0. (1)
ysis. In Section III, we perform a complete classifica-
tion of 1D superconductors with rotational symmetries. Therefore the Hamiltonian can be block-diagonalized into
In the classification, we find it important to distinguish sectors spanned by eigenvectors of L, namely,
the case where the Cn symmetry is represented projec- H = Hr1 ⊕ Hr2 ⊕ ... + Hrl , (2)
tively from the case where it is linearly represented: in
the former case, time-reversal symmetry anti-commutes where s is the number of eigenvalues of L, and r1,...,l
with the rotation symmetry when acting on a Bogoli- the eigenvalues; in sector ri , the basis vectors are eigen-
ubov quasiparticle. Which case appears depends on the states of L with eigenvalue ri . For each sector, we can
total angular momentum of the Cooper pair. In Section classify each Hri according to its transformation under
IV, we consider Chern superconductors in 2D with ro- time-reversal, particle-hole and chiral symmetries. Phys-
tational symmetries, deriving explicit relations between ically, time-reversal and particle-hole symmetries com-
the Chern number and the rotation eigenvalues of occu- mute with all spatial symmetries, and let us assume that
pied bands at high-symmetry points. In Section V, we [KT , L] = [KP , L] = 0. However, one is reminded that
apply our knowledge of 1D superconductors to classify generally [KT , Hri ] 6= 0 and [KP , Hri ] 6= 0. This is be-
both high-symmetry and generic lines within the Bril- cause if ri ∈
/ Real, KP or KT sends the state to another
louin zone (BZ) of 3D superconductors by determining sector with eigenvalue ri∗ . However the chiral symme-
not only the bulk invariants but also the effective sur- try, represented by a unitary matrix, preserves the block
face theory. In Section VI, we summarize our results and structure of H.
 3

Based on this discussion, we come to a simple con- of an electron. For our purposes, the term ‘spinless’ sim-
clusion: for any r ∈ Real, the Hamiltonian Hr inherits ply refers to the unbroken SU(2) spin rotation symme-
the particle-hole, time-reversal and chiral symmetries the try, while the term ‘spinful’ indicates its absence. When
system may have, thus having the same topological clas- the spin-orbit interaction is ignored, an SU(2) invariant
sification as that found Ref.[62–64]. On the other hand, Hamiltonian can be block-diagonalized into two parts,
if r ∈
/ Real, Hr only inherits the chiral symmetry of the namely those of spin up and the spin down parts, while
system, should the system possess chiral symmetry, while both the TRS and PHS can be composed with a spin-
the time-reversal and particle-hole symmetries relate Hr rotation about a specific directional axis (say, y-axis) so
to Hr∗ . Due to this relation, the topological invariants as to not change the spin state, thus acting within each
for Hr and Hr∗ will be shown to be either the same or block. This signifies that the symmetries of the whole
opposite, depending on S and the symmetry class of the system completely pass to each of the respective spin sec-
Hamiltonian. Therefore, a local unitary symmetry in tors, and it is therefore sufficient to study any one of the
general enhances the topological classifications, as now two independent ‘spinless’ Hamiltonians to understand
the full system is labeled by all the quantum numbers the properties of the system as a whole.
from each sector (if r ∈ Real) and each pair of sectors (if However, this is not quite the complete picture of the
r∈ / Real), rather than the single Z or Z2 number for the proper physics. One needs to take caution in that since
entire Hamiltonian H. PHS and TRS are now combined with a spin rotation,
A point group, or specifically rotational, symmetry is their squares change sign, resulting in KT 2 = −KP 2 =
in general not a local symmetry, as it changes the position 1. This becomes a vital distinction and is needed when
of an electron. The only exceptions are mirror reflection we discuss rotational symmetries. A full rotation of an
in a 2D system when then mirror plane and the system electron gives a −1 factor to the wavefunctions due to
are coplanar, and the rotation in a 1D system when the the inherent π Berry’s phase. This points to the fact
rotation axis coincides with the system. Nonetheless, for that for a general electronic system, we have Cnn = −1
single particle Hamiltonians with translational symme- where Cn is the rotation operator of an n-fold rotation.
try, there are always some sub-manifolds in the k-space When SU(2) is present, the rotation symmetry can be
that are invariant under a point group symmetry. For redefined as a rotation of both spin and space followed
example, in a 3D simple cubic lattice, the tight-binding by a spin rotation in the opposite direction. In such a
Hamiltonian H(kx , ky , kz ) in k-space is invariant under case, we have Cnn = 1. Physically this means that in
mirror reflection Mxy : z → −z when kz = 0 or kz = π; the absence of SOC, a rotation symmetry only operates
and it is invariant under the fourfold rotation about the on the spatial degrees of freedom, leaving the spin part
z-axis C4 : (x, y, z) → (−y, x, z), if (kx , ky ) = (0, 0) or unchanged. Within the context of operators, the terms
(π, π), where the lattice constant is taken to be unity, ‘spinless’ and ‘spinful’ indicate Cnn = KT 2 = −KP 2 = 1
a ≡ 1. For an invariant sub-manifold, the point group and −1, respectively.
symmetry becomes a local symmetry, and hence, we
can use the general scheme described above, i.e., clas-
sifying the sectors labeled by the eigenvalues of L, to C. Rotational Symmetry of a BdG Hamiltonian
find the classification and the corresponding invariants of
the Hamiltonian on the sub-manifold. The set of quan- We are interested in discussing the properties of mean-
tum numbers defined on all these invariant sub-manifolds field superconducting systems, and hence we must un-
characterize a general topological crystalline insulator or derstand how the previously discussed rotational sym-
superconductor. It is this method that we will be using to metries manifest themselves in a Bogoliubov-de Gennes
classify superconductors with rotation symmetry within (BdG) Hamiltonian. The second-quantized form of a
the context of this work. We should note, however, that BdG Hamiltonian reads:
this characterization is incomplete and there exist inter-
ˆ +∆
Ĥ = Ĥ0 + ∆ ˆ †, (3)
esting exceptions35,77,78 . For example, in Ref.[35,77], the
authors show that 3D spinless systems host topological
where
numbers that are protected by rotation symmetry but
yet cannot be defined on any of the high-symmetry lines. Ĥ0 = Hab c†a cb + h.c., (4)
ˆ = ∆ab ca cb ,
∆
B. Spinless Vs. Spinful Systems where a, b are composed indices labeling the site, orbital
and spin in a lattice system. In an n-fold rotation invari-
Having clarified the role of rotational symmetries in ant system, we require that
general Hamiltonians, it is important to mark the dif-
[Ĉn , Ĥ0 ] = 0, (5)
ference between spinless and spinful systems. As super-
conductors are, naturally, electronic systems and their and
constituent electrons are elementary particles with spin
one-half, this distinction does not refer to the spin state ˆ Ĉn−1 = eiθ ∆.
Ĉn ∆ ˆ (6)
 4

Using the fact that Ĉnn = ±1 for spinless and spinful If m = 0, Ĉn,m is the same as Ĉn and also commutes
systems respectively in conjunction with Eq.(6), we find with time-reversal. When m = n/2, which is only possi-
einθ = 1 ,or θ = 2mπ/n where m = 0, ..., n − 1. Consider ble if n ∈ even, the commutation relation between Ĉn,n/2
a gauge transform Û = exp(imπ/nQ̂), where Q̂ is the and T̂ is
total electric charge operator, such that
{Cn,n/2 , KT } = 0, (14)
−1
Ûm ca Ûm = e−imπ/n ca , (7)
where KT is the matrix representation of T where K is
and if we combine Ûm with Ĉn , from Eq. 5, we have complex conjugation. Eq.(14) indicates that the Cn,n/2 ,
KT and KP form a projective representation of the
[Ĉn Ûm , Ĥ] = 0. (8) group generated by T , Cn and P . We again defer the
proof to Appendix B.
Therefore, we may define Ĉn,m ≡ Ĉn Ûm as a symmetry
of the system. In a translation invariant system, we have
III. CLASSIFICATION OF 1D
−1
Cn,m H(k)Cn,m = H(Cn k), (9) SUPERCONDUCTORS WITH ROTATIONAL
SYMMETRIES
where Cn,m is the first quantized matrix representation
of Ĉn,m in the Nambu basis. The significance of Eq.(9) We now proceed to 1D, and some quasi-1D systems,
lies in that the spectrum of H(k) remains rotationally where the rotation along an axis parallel to the system is
invariant even when m 6= 0. Physically, m 6= 0 indi- a symmetry represented by a matrix Cn,m satisfying the
cates that the Cooper pair has total angular momentum commutation relation,
m~ along the rotation axis (modulo n), because the pair
gains a phase of ei2mπ/n after a rotation through 2π/n. [Cn,m , H(k)] = 0. (15)
Nevertheless, this change of phase is not reflected in the
quasiparticle spectrum, which is still n-fold symmetric, It should be noted that in one dimension, there is no
due to Eq.(9). This is because the phase can be com- crystallographic constraint on n, and therefore n ∈ Z +
pensated by a U(1) transform which leaves Ĥ0 invariant. (positive integers). As we classify the respective gapped
Additionally, we should also point out that superconductors with rotational symmetries in 1D below,
we separate the discussions into two distinct sections: one
n
Cn,m = (−1)m Ĉnn = (−1)m+F , (10) in which the total angular momentum of the Cooper pair
is zero (m = 0) and one where the angular momentum is
where F = 0, 1 for spinless and spinful fermions, respec- non-zero (m 6= 0).
tively. Therefore, when m ∈ odd, Cn,m is like a spinless
(spinful) rotation in a spinful (spinless) system. Math-
ematically, when m ∈ odd, Eq.(10) indicates that Cn,m A. Pairing without Cooper Pair Angular
and KP form a projective representation of the group Momentum (m = 0)
generated by Cn and P , a fact that we rigorously prove
in Appendix A. Most generic of BdG Hamiltonians, which belong to
class D in the AZ classification table, have preserved
particle-hole symmetry, which is represented by KP in
D. Time-reversal symmetry and projective the Nambu basis, where K is the complex conjugate and
representation of the group generated by Cn and T P is a unitary matrix. Physically, we have P̂ 2 = 1 for an
electron, which leads to
In this paper we also consider superconductors with
TRS. Naturally, TRS implies that (KP )2 = P P ∗ = 1, (16)
P = PT.
ˆ T̂ −1 = ∆.
T̂ ∆ ˆ (11)
Since PHS changes electrons to holes and vice versa, it
Using Eq.(6), we have anti-commutes with the first quantized Hamiltonian and
the momentum operator. This means that it sends one
ˆ T̂ −1 Ĉn−1 = Ĉn ∆
Ĉn T̂ ∆ ˆ Ĉn−1 = ei2mπ/n ∆,
ˆ (12) single particle state to another with opposite energy and
momentum; symbolically we have
and
(KP )H(k)(KP )−1 = −H(−k), (17)
ˆ Ĉn−1 T̂ −1 = T̂ ei2mπ/n ∆
T̂ Ĉn ∆ ˆ T̂ −1 = e−i2mπ/n ∆.
ˆ (13)
or
However, since [Ĉn , T̂ ] = 0, the m which satisfies both
Eq.(12) and Eq.(13) is m = 0, n/2. P H(k)P † = −H T (−k). (18)
 5

In addition to this, PHS commutes with the rotation also has Z2 classification, the Z2 number of H−1 (k) is
symmetry as the same as the Z2 number of the full Hamiltonian.
In addition to PHS, which is shared by all supercon-
[KP, Cn ] = 0 → P Cn P † = Cn∗ . (19) ductors, we consider the presence of TRS, corresponding
to class DIII in the AZ classification. In the Nambu basis,
In the basis spanned by eigenstates of Cn , (φ1 , φ2 , ...)T , TRS is represented by KT , where T is a unitary matrix.
we have For spinful electron, we have
X
Cn = rIdr , (20) (KT )2 = −1 → T = −T T . (24)
⊕r
The action of TRS reverses the momentum of an electron
where Idr is a dr -by-dr identity matrix and dr is the without changing its energy, or symbolically,
degeneracy of the eigenvalue r. Using Eq.(19) and
Eq.(20), we obtain the expression for P in the basis (KT )H(k)(KT )−1 = H(−k), (25)
(φ1 , φ2 , ..., φ∗1 , φ∗2 , ...)T as those spanned by the eigen- T H(k)T † = H T (−k).
states.
X X  0 Qr  TRS also commutes with all spatial symmetries. Specif-
P = Pr ⊕ , (21) ically,
QTr 0
⊕r∈real ⊕Im[r]>0
[KT, Cn ] = 0 → T Cn T † = Cn∗ . (26)
where Pr and Qr are unitary matrices. This indicates
In the basis spanned by the eigenvectors of Cn , we have
that the PHS operator leaves unchanged the eigenspace
the following block-diagonalization of T
of the rotation matrix, Cn , with a real eigenvalue, but
maps the eigenspace with a complex eigenvalue to its X X  0 Rr 
complex conjugate. T = Tr ⊕ , (27)
−RrT 0
Using Eq.(16) and Eq.(21), we then have for r ∈ Real, ⊕Im[r]=0 ⊕Im[r]>0

Pr Hr (k)Pr† = −HrT (−k), (22) where Tr and Rr are unitary matrices and Tr is anti-
symmetric from Eq.(24). Using Eq.(25) and Eq.(27), we
and for r ∈
/ Real, we have have for each r ∈ Real,

0 Q∗r Tr Hr (k)Tr† = HrT (−k), (28)

   
0 Qr Hr (k) 0
(23)
QTr 0 0 Hr∗ (k) Q†r 0
 T  and for each r ∈
/ Real
Hr (−k) 0
= − 
0 Rr

Hr (k) 0

0 −Rr∗

0 HrT∗ (−k) (29)
−RrT 0 0 Hr∗ (k) Rr† 0
According to the AZ classification, Hr (k) belongs to  T
Hr (−k) 0

class D if r ∈ Real and to class A (i.e., no symme- = .
0 HrT∗ (−k)
try because PHS relates r to r∗ ) if r ∈ / Real, while
Hr (k) ⊕ Hr∗ (k) again belongs to class D. In 1D, class D From Eq.(28), we understand that each sector with real r
has a Z2 classification whereas class A is trivial. There- has both TRS and PHS and hence belongs to class DIII,
fore, each Hr∈Real possesses its own Z2 -index. There- which in 1D has a Z2 number. For r ∈
/ Real, by utilizing
fore, the question that remains to be answered is can a combination of Eq.(23) and Eq.(29), we obtain
Hr (k) ⊕ Hr∗ (k) be Z2 nontrivial? We argue that it  ∗ T   † T 
is impossible by examination of a simple contradiction. Qr Rr 0 Hr (k) 0 Rr Qr 0
(30)
Should Hr (k) ⊕ Hr∗ (k) be Z2 nontrivial, then for an 0 Q†r Rr 0 Hr∗ (k) 0 Rr† Qr
open chain there must be a single Majorana mode at  
Hr (k) 0
each end79 . Due to Cn -symmetry, the Majorana mode =− .
0 Hr∗ (k)
must either have a rotation eigenvalue of r or r∗ , but
either choice breaks the inherent PHS. Based on this dis- If we define Sr = Q∗r RrT and Sr∗ = Q†r Rr for each r ∈
/
cussion, we find that the topological classification of a Real, Eq.(30) leads to
Cn -invariant 1D superconductor without additional sym-
metries is given by a set of Z2 numbers from each Hr (k) {Sr , Hr (k)} = 0, (31)
with r ∈ Real. Using rn = −1, it is obvious that for
n = even all eigenvalues are complex, thus, the classifi- which indicates that Hr∈Real
/ (k) belongs to the chiral
cation is always trivial. Meanwhile for n = odd, r = −1 class AIII. Although a real sector Hr∈Real also has chi-
is the only real eigenvalue of Cn , and the only topological ral symmetry defined as Sr = Pr Tr , one cannot use this
number is the Z2 number of H−1 (k). It must be noted symmetry for classification, as the topological invariants
that since the full Hamiltonian belongs to class D, which protected by Sr are constrained to certain numbers, zero
 6

in this case, by the individual PHS or TRS symmetry. by Cn,m as the rotation symmetry of the system. In the
Class AIII in 1D has a Z number, so each sector with Nambu basis, it is represented by
complex r has a Z number denoted by z (r) . In fact, we mπ
(r) (r ∗ ) Cn,m = Cn eiτz , (34)
can further argue that z = −z . The chiral symme- n

try is the composition of TRS and PHS, thus satisfying

S 2 = P 2 T 2 = −1. This indicates that Sr has eigen- where τz is the Pauli matrix in the particle-hole indices
values ±i. Any class AIII Hamiltonian having z (r) > 0 due to Um . We understand that PHS commutes with
(z (r) < 0) means that there are |z (r) | edge states at each Cn,m because (i) PHS commutes with Cn and (ii) it anti-
end of an open system that are eigenstates of S with commutes with both i and τz , or
eigenvalue +i (−i). But under TRS, an edge state hav-
ing Cn eigenvalue r and Sr eigenvalue +i maps to another [KP, Cn,m ] = 0 → P Cn,m P † = Cn,m
∗
. (35)
edge state having Cn eigenvalue r∗ and Sr eigenvalue −i,
implying that Comparing Eq.(35) and Eq.(19), we see that all preceding
understanding obtained in the previous section where we
∗
z (r) = −z (r ) , (32) ignored the angular momentum of the Cooper pair also
applies to the system with both PHS and Cn,m symme-
as a result of which the number of independent Z num- try. Therefore, we simply apply the results and arrive
bers is determined by one-half the number of complex at the following conclusions: (i) For m, n ∈ even, the
eigenvalues of Cn . Eq.(32) also implies that for r ∈ Real, classification of the system is trivial, as we have already
TRS sets this topological number to zero, as r = r∗ . For stated. (ii) For m ∈ odd and n ∈ even, r may take the
n ∈ even, the full classification is given by n/2 integers value r = ±1, and H±1 (k) belongs to class D and gives
(as all Cn eigenvalues appear in complex pairs), and if (±1)
two Z2 numbers, z2 . (iii) For m ∈ even and n ∈ odd,
n ∈ odd, it is given by one Z2 number and (n − 1)/2 r may take −1 but not +1, and H−1 (k) has a Z2 number
integers (for all eigenvalues except −1 appear in pairs). (−1)
Again, the Z2 number of the full Hamiltonian is given by z2 . (iv) Finally, for m, n ∈ odd, r may take the value
(1)
the same as the Z2 number of H−1 (k) when n ∈ odd, and of +1 but not −1, and H1 (k) gives a Z2 number z2 .
is trivial if n ∈ even. The Z2 -index for the full Hamiltonian, neglecting rota-
Finally, let us consider spinless electrons, or equiva- tion symmetry, is the same as the sum of the Z2 -indices
lently, adding spin-SU(2) symmetry. We can follow all corresponding to each sector with a real eigenvalue of
the steps above to find the classifications, keeping in mind Cn,m .
the distinction that for spinless electrons, we have Now we consider adding TRS to the system. In
Sec.II D we have shown that the only nonzero m that
T = TT, (33) is compatible with TRS is m = n/2 when n ∈ even. In
P = −P T , this case, Cn,n/2 and KT anti-commute, i.e.,
Cnn = 1.
T Cn,n/2 T † = −Cn,n/2
∗
. (36)
Using nearly identical calculations to those presented in
this section, we may derive the following additional re- Eq.(36) indicates that TRS maps a state with Cn,m eigen-
sults: (i) Without TRS, any Hr (r) with real r and any value r to a state with eigenvalue −r∗ . We note that
Hr (k) ⊕ Hr∗ (k) with complex r belong to class C and in this case time-reversal operator anti-commutes with
have a trivial classification. (ii) Additionally, without Cn,m from Eq.(34) due to the fact that: (i) TRS com-
TRS, any Hr (k) with complex r belongs to class A and mutes with Cn yet anti-commutes with imaginary unit
also has a trivial classification. (iii) With TRS, any Hr (r) i and (ii) it commutes with τz for it does not inter-
with real r and any Hr (k) ⊕ Hr∗ (k) with complex r be- change particles and holes. Therefore, for the case when
long to class CI, which has a trivial classification in 1D m = n/2 and we have n ∈ even, we separately discuss
(iv) In the presence of TRS, any Hr (k) with complex the following two constraints on the rotational symmetry:
r belongs to class AIII, having a Z classification, under (i) m ∈ even ⇔ n = 4k and (ii) m ∈ odd ⇔ n = 4k − 2.
the constraint as outlined in Eq.(32). Therefore in class When n = 4k (k being a non-negative integer) and the
CI, for n ∈ even, there are (n − 2)/2 integers to specify general eigenvalue of Cn,m is rs = ei2π(s+1/2)/n (s being
the topological state and for n ∈ odd there are (n − 1)/2 a non-negative integer), then under PHS the rs -sector
integers, corresponding to the number of conjugate pairs and the rn−s−1 -sector are mapped to each other while,
of complex eigenvalues of Cn,m . at the same time, under TRS, the rs -sector and the
rn/2−s−1 -sector are mapped to each other. Therefore,
Hrs ⊕ Hrn/2+s (note that rn/2+s = −rs in this case) be-
B. Pairing with Cooper Pair Angular Momentum longs to class AIII , thereby having a Z-index. (Since
(m 6= 0) P ∗T maps s → n−s−1 → n/2−(n−s−1)−1 = n/2+s,
the direct sum is invariant under the composite symme-
When the angular momentum of the Cooper pair is try.) However, this Z-number must vanish and this can
considered, namely when m 6= 0, Cn must be replaced be shown by contradiction. Suppose this Z number is
 7

z > 0, then on the edge there are z states that are eigen- n ∈ odd n = 4k n = 4k − 2
vectors of P ∗ T having eigenvalue +i, such that P ∗ T C 0 0 0
in the Hilbert space spanned by the zero modes is iI, D, m ∈ even 0 Z2 Z2
where I is the identity matrix the dimension of which
D, m ∈ odd Z22 Z2 Z2
is the number of zero modes. Then we assume that the
rotation symmetry Cn,m be represented by some matrix CI, m = 0 Z (n−1)/2 Z (n−2)/2 Z (n−2)/2
R, and we have [R, P ∗ T ] = 0 contradicting Eq.(35,36). CI, m = n/2 0 0 0
Considering the second case, n = 4k − 2, we know that DIII, m = 0 Z2 × Z (n−1)/2 Z n/2 Z n/2
the general eigenvalue of Cn,n/2 is rs = e2sπ/n . Under DIII, (m = n/2) 0 0 Z2
PHS, the rs -sector and the rn−s sectors are mapped to
each other with the exceptions of s = n/2 and s = n,
where r = −1 and r = +1, respectively. As in the previ- TABLE I: Complete classification of 1D gapped supercon-
ous case, under TRS, the rs -sector and the rn/2−s -sector ductors with rotation symmetry, Cn,m . Within the table, ‘0’
indicates that for the given system the classification is trivial.
are mapped to each other. Therefore, for s 6= n/2, the
Hamiltonian Hrs ⊕ Hrn/2+s belongs to class AIII, while
H+1 ⊕ H−1 belongs to class DIII. Based on above ar-
gument, the Z-index of the class AIII component must IV. 2D SUPERCONDUCTORS WITH
vanish, leaving to an overall classification to be Z2 . The ROTATION SYMMETRIES
Z2 -index for the full Hamiltonian, again without consid-
ering rotation symmetry, is the same as the Z2 -index for With our discussion of gapped 1D superconductors
H+1 ⊕ H−1 . with rotation symmetries complete, we focus our atten-
Finally, we consider adding SU(2) symmetry. With tion in this section on the study of 2D gapped supercon-
PHS and SU(2) the full Hamiltonian is in class C, hav- ductors without TRS (class D and class C)in the presence
ing (KP )2 = −1 as the only symmetry. In this case, of Cn,m invariance, where the rotation axis is assumed to
all sectors of the Hamiltonian belong either to class C be the axis perpendicular to the system. Lattice peri-
or to class A, both being trivial. When we include TRS odicity is compatible with rotation symmetry only when
along with PHS and SU(2), we again need to separately n = 2, 3, 4, 680 . In BZ, there exist high-symmetry points
consider the two cases above with regards to the ro- that are invariant under Cñ , where ñ is a factor of n,
tational symmetry of the given system. We first con- denoted by Kñ . For example, in a C4 -invariant system,
sider the case for n = 4k in which a generic eigenvalue X = (π, 0) and Y = (0, π) are points that are C2 in-
of Cn,n/2 is rs = e2sπ/n . Under PHS, the rs -sector variant; and in a C6 -invariant system, K and K 0 are
and the rn−s sectors are mapped to each other with C3 -invariant. At Kñ , each energy eigenstate is also an
the exceptions of s = n/2 and s = n. Further, under eigenstate of Cñ,m (it being understood that m is a mod
the application of TRS, the rs -sector and the rn/2−s - ñ number). For each eigenvalue r of Cñ,m , we count at
sector are mapped to each other, with the exceptions of Kñ the number of occupied energy eigenstates that are
s = n/4 and s = 3n/4, where rs -sector is mapped to also eigenstates of Cñm with eigenvalue r and denote it
itself. Therefore, for s 6= n/4, n/2, 3n/4, n, the Hamilto- by Nr (Kñ ). We show that these numbers are related to
nian H(rs ) ⊕ H(rn/2+s ) belongs to class AIII and is char- the Chern numbers in superconductors. Physically, the
acterized by a vanishing Z-index. Similarly, the sectors Chern number of a superconductor is determined by both
corresponding to H(+1) ⊕ H(−1) and H(+i) ⊕ H(−i) be- the band structure of the normal state and the symmetry
long to class CI, and possess only a trivial classifica- of the pairing amplitude on the Fermi surface. The for-
tion resulting in an overall classification for the system mer contribution is related to the Cn eigenvalues of the
that is trivial. For the case corresponding to rotational occupied bands at high symmetry points in the normal
symmetries satisfying n = 4k − 2, we have eigenvalue states39,78,81,82 , while the latter contribution is related
of rs = ei2π(s+1/2)/n . As before, under PHS, the rs - to m, namely, the angular momentum of the Cooper pair
sector and the rn−s−1 -sector are mapped to each other, modulo n. In this section, we focus on how these numbers
and the application of TRS maps the rs -sector and the relate to the Chern number in gapped 2D superconduc-
rn/2−s−1 -sector to one another, with the exception of s = tors.
(n/2 − 1)/2. Therefore, for s 6= (n/2 − 1)/2, (n/2 + 1)/2,
the Hamiltonian Hrs ⊕Hrn/2+s belongs to class AIII, hav-
ing a Z-classification but with vanishing Z-index, while
H(+i) ⊕ H(−i) belongs to class CI, having trivial classi- A. Continuum Limit (n = ∞)
fication. Accordingly, the overall classification is again
trivial. To begin our analysis, let us first consider the contin-
In Table I, We summarize the classification of all uum limit with full SO(2) symmetry. In this limit, the
gapped 1D superconductors within four BdG classes (C, angular momentum of the Cooper pair, m, can take any
D, CI and DIII) of the AZ classification enhanced by integer. We choose to work in an orbital basis in which
Cn,m -symmetry. the generator of the rotation operator Jˆ is diagonalized.
 8

Therefore, in the Nambu basis, we have general relation between the Chern number and all jm ’s
at k = 0 and k = ∞:
J˜ = τz ⊗ diag{j1 , j2 , ..., jNorb }, (37)
N
Xorb
i i
where τz is the Pauli matrix acting on the particle-hole C= [jm (0) − jm (∞)]. (47)
i=1
index and jα is the angular momentum of the α-th elec-
tronic orbital. The second quantized form of Jˆ is given If one considers the gapped BdG Hamiltonian to be the
by same as that of an insulator with accidental particle-hole
symmetry, Eq.(47) simply means that its Chern number
ji φ†i φi ,
X
Jˆ = (38) equals the total angular momentum (along z-axis), where
i=1,...,Norb Jˆm is the angular momentum operator, of all occupied
states. To heuristically observe this, we notice that for
where φi is the annihilation operator of angular momen- any occupied state, |ψ(k)i at a generic k 6= 0, ∞, the
tum ji . Under rotation through θ via the application of ˆ
the rotation operator, we have state eiJm θ |ψ(k)i must also be an occupied state with
momentum R(θ)k. One can always construct |ψn i =
R 2π ˆ
ˆ
eiJθ ∆e
ˆ
ˆ −iJθ = eimθ ∆,
ˆ (39) 0
dθeinθ eiJm θ |ψ(k)i for any integer n and, therefore, all
contribution to the total angular momentum from generic
or its infinitesimal version k’s cancel each other, leaving the only contribution from
k = 0, ∞. We then recall that any rotation about k = 0
ˆ J]
[∆, ˆ = −m∆.
ˆ (40) is equivalent to an inverse rotation about k = ∞, so the
total angular momentum is the difference, not the sum,
i
Furthermore, we also know of jm ’s at 0 and ∞.
Now examine the weak pairing limit, where we may
ˆ Q̂] = −2∆,
[∆, ˆ (41) separate the contribution due to the normal state band
structure from that of the pairing on the Fermi surface.
where Q̂ c†α cα is the total charge. Using Eq. (41) in
P
In the weak coupling limit, at each k, the occupied bands
conjunction with Eq. (40), we can prove that in the BdG Hamiltonian consist of two distinct parts: the
occupied bands of the non-superconducting Hamiltonian,
[Jˆm , ∆]
ˆ = 0, (42) and the particle-hole partner of all the unoccupied bands.
Keeping in mind that a hole state has opposite charge and
where Jˆm ≡ Jˆ − m † angular momentum compared with an electron state, we
Q
2 α (1 − 2cα cα ). In the Nambu basis,
Jˆm is represented by have
N orb
m X
J˜m = J˜ − τz ⊗ INorb . (43) i
jm (K) = j 1 (K) + ... + j Nocc (K) (K) (48)
2 i=1

Since Jˆ commutes with the normal part, Ĥ0 of the Hamil- − (j Nocc (K)+1 (K) + ... + j Norb (K))
tonian, using Eq.(42), we know that Jˆm commutes with −
m
[Nocc (K) − Nunocc (K)],
the full Hamiltonian 2
where Nocc and Nunocc are the number of occupied and
[Jˆm , Ĥ] = 0, (44)
unoccupied bands, respectively. Substituting Eq.(48) to
or, in the presence of translational symmetry Eq.(47), we obtain a simple formula

m C = 2[J(0) − J(∞)] − m[Nocc (0) − Nocc (∞)]. (49)

H(k+ eiθ , k− e−iθ ) = exp(i(J˜ − τz )θ)×
2 (45)
m The physical meaning of Eq.(49) is clear as the first term
H(k+ , k− ) exp(−i(J˜ − τz )θ), is simply two times the total angular momentum of the
2
normal state, where the factor of two is because of the
where k± = kx ±iky . In the 2D continuum k-space, k = 0 Fermion doubling in the Nambu basis. The second term
and k = ∞ are the only two points that are invariant is the total angular momentum of the pairing on all Fermi
under rotation. At these points we have surfaces. To see this, we need to notice two separate
facts: (i) Eq.(39) indicates that m is the total angular
[J˜m , H(0)] = [J˜m , H(∞)] = 0, (46) momentum of a Cooper pair, and (ii) Nocc (0) − Nocc (∞)
is the difference in the occupation numbers at k = 0
where we have implicitly assumed that H(∞) is well de- and k = ∞. Suppose Nocc (0) > Nocc , and by traversing
fined. Each state of H(0) or H(∞) is also an eigenstate any path from 0 to ∞ one crosses Ne electron-like Fermi
of J˜m of eigenvalue jm
i
(i denoting the occupied bands surfaces and Nh hole-like Fermi surfaces, then we have
in the BdG Hamiltonian). In Appendix C, we prove a Ne − Nh = Nocc (0) − Nocc (∞).
 9

B. Finite Rotational Symmetry Cn (n = 2, 3, 4, 6) where ζm , θm , ξm , ηm are the product of all eigenvalues

of C2,m , C3,m , C4,m and C6,m at corresponding high-
When considering the more realistic case of a 2D lat- symmetry points on the lower half BdG bands, respec-
tice, the continuous rotation symmetry C∞ breaks down tively. In the weak coupling limit, they again reduce to
to Cn=2,3,4,6 . In our approach here, we closely follow our expressions that only involve the eigenvalues of Cñ , the
previous work39 , in order to obtain the Chern number occupation number at each high-symmetry point and the
up to a multiple of n in terms of the eigenvalues of Cñ,m angular momentum of the Cooper pair (mod n). To be
at k-points invariant under Cñ where ñ divides n. For specific,
Cn=2,3,4,6 we find:

ei2πC/2 = ζm (Γ)ζm (M1 )ζm (M2 )ζm (M3 ), (50)

i2πC/3 Norb (m+1) 0
e = (−1) θm (Γ)θm (K)θm (K ),
i2πC/4 Norb (m+1)
e = (−1) ξm (Γ)ξm (M )ζm (X),
i2πC/6 Norb (m+1)
e = (−1) ηm (Γ)θm (K)ζm (M ),

ei2πC/2 = exp[i(m + 1)π(Nocc (Γ) + Nocc (M1 ) + Nocc (M2 ) + Nocc (M3 )], (51)
θ2 (Γ) 2mπ
ei2πC/3 = 0
exp[i (2Nocc (Γ) − Nocc (K) − Nocc (K 0 ))],
θ(K)θ(K ) 3
ξ 2 (Γ) 2mπ
ei2πC/4 = 2 exp(−i )(Nocc (Γ) + Nocc (M ) − 2Nocc (X)),
ξ (M ) 4
η 2 (Γ)ζ(M ) 2mπ
ei2πC/6 = exp[−i (Nocc (Γ) + 2Nocc (K) − 3Nocc (M ))].
θ(K) 6

and the definition of high-symmetry points is given in If m = 0, then we have (since C22 = −1)
Fig.1. Let us derive the n = 2 case here in detail. In the
weak coupling limit, each occupied state of C2 eigenvalue
mπ
ζ at Γ is an eigenstate of C2,m with eigenvalue ζei 2 ,
and each unoccupied state with eigenvalue ζj at Γ is an
mπ
eigenstate of C2,m with eigenvalue ζj∗ e−i 2 after PHS Y
transform. Therefore, the total product of eigenvalues of ( ζi )2 = (−1)Nocc (Γ) , (53)
i∈occ
C2,m at Γ is
Y mπ
Y −mπ
ζm (Γ) = ( ζi ei 2 )( ζj∗ ei 2 ) (52)
i∈occ j∈unocc
mπ
Y Y
= ( ζi ) (2
ζn∗ )eimπNocc (Γ) e−i 2 Norb . so
i∈occ n∈orb

Y
(−1)C = (−1)Nocc (Γ)+Nocc (M1 )+Nocc (M2 )+Nocc (M3 ) ( ζn∗ )4 (54)
n∈orb

= (−1)Nocc (Γ)+Nocc (M1 )+Nocc (M2 )+Nocc (M3 )

If m = 1, then so
Y
( ζi )2 = 1, (55)
i∈occ
 10

Y
(−1)C = (−1)2Nocc (Γ)+2Nocc (M1 )+2Nocc (M2 )+2Nocc (M3 ) ( ζn∗ )4 e−i2πNorb (56)
n∈orb
= 1.

n/2
In Eq.(51), the contribution to the Chern number either T , P or Cn,m , this momentum is mapped to −k̄,
again decomposes into two parts as promised, but the but under P 0 or T 0 the momentum is mapped back to
physical meaning is not as transparent as in Eq.(49), be- k̄, while sending the momentum along the line to its op-
cause here the angular momentum is well-defined only up posite value. Using the commutation relations similar in
to a multiple of n, and states at high-symmetry points form to those used in Eq. (35) and Eq. (14) , we have
other than Γ contribute to the total angular momentum
in different ways. We hope our eigenvalue formulas for (KP 0 )2 = (−1)m+F (KP )2 = (−1)m+1 , (57)
projector Chern numbers can be useful in the search of (KT )0 2
= (−1) mn/2+m+F 2 F 2
(KT ) = (−1) (KT ) = 1.
topological chiral superconductors.
Eq.(57) makes the class of generic lines different from
the class of the whole system, depending on the parity
V. 3D SUPERCONDUCTORS WITH of m. Let us examine each of the subsequent possibili-
ROTATIONAL SYMMETRIES ties of lines embedded in the 3D BZ in turn: Case (i):
class C and m even, then Eq.(57) states that (KP 0 )2 =
A. Bulk Invariants (KP )2 = −1, thus H(k̄) belongs to class C and has triv-
ial classification. Case (ii): class C and m odd, then
Eq.(57) states that (KP 0 )2 = −(KP )2 = 1, so H(k̄)
We move on to discuss the partial topological classifi-
belongs to class D, which gives it a Z2 classification.
cation of 3D gapped superconductors with rotation sym-
Case (iii): class D and m even, then Eq.(57) states that
metries. In a 3D lattice, crystallographic constraint dic-
(KP 0 )2 = −(KP )2 = −1 indicating that H(k̄) belongs
tates that n = 2, 3, 4, 6 and within a Cn invariant lattice
to class C and has a trivial classification. Case (iv): class
system, there are discrete lines in the 3D BZ that are
D and m odd, Eq.(57) states that (KP 0 )2 = (KP )2 = 1,
invariant under Cñ>1 where ñ is a factor of n. Therefore,
so H(k̄) belongs to class D, having Z2 classification.
in order to classify 3D gapped superconducting systems,
Case (v): class CI and m even and Eq.(57) states that
we can apply the classification of 1D superconductors
(KP 0 )2 = (KP )2 = −1 and (KT 0 )2 = (KT )2 = 1, so
with Cñ invariance to these sub-manifolds, and the set
H(k̄) belongs to class CI, having trivial classification.
of quantum numbers of all Cñ -invariant lines gives the
Case (vi): class CI and m odd thus Eq.(57) states that
enhanced classification of the 3D system. We notice that
(KP 0 )2 = −(KP )2 = 1 and (KT 0 )2 = (KT )2 = 1,
for ñ = 2, 4, and 6 all three local symmetries we have
indicating that H(k̄) belongs to class BDI, which has
discussed in the text, namely time-reversal, particle-hole
a Z classification. Case (vii): class DIII and m even
and spin rotation, of the 3D system are also preserved
so Eq.(57) states that (KP 0 )2 = −(KP )2 = −1 and
on the Cñ invariant lines. The same applies to the line
(KT 0 )2 = −(KT )2 = 1 indicating that H(k̄) belongs
that includes Γ when n = 3. In Fig.1, we schemati-
to class CI, having trivial classification. Case (viii): class
cally represent these special lines by their projections
DIII and m odd, Eq.(57) states that (KP 0 )2 = (KP )2 =
onto the surface BZ, where the surface is perpendicu-
1 and (KT 0 )2 = −(KT )2 = 1, so H(k̄) belongs to class
lar to the rotation axis. For the outset, we show that
BDI and has a Z classification.
special treatment is needed for the following lines: (i) a
We now consider a C3 -invariant line that does not
generic vertical (parallel to the rotation axis) line in BZ
include Γ. If the system under consideration has C3 -
in a system with C2,4,6 -symmetry and (ii) a C3 -invariant
invariance but not C6 -invariance, then the line does not
line that does not include Γ. We must treat these lines
possess PHS. The Hamiltonian on the line hence belongs
specially because, while TRS and PHS are not symme-
to class A and accordingly possesses a trivial classifica-
tries, compositions such as C2 ∗ P and C2 ∗ T , P ∗ T
tion. However, if the system also contains TRS , the
might be symmetries. Due to the lack of PHS on these
line possesses the combined symmetry KP ∗ KT . Since
generic lines, the classification derived in Sec.III does not
both KT and KP commute with C3 , each sector of the
apply for these lines. In a system with C2,4,6 -symmetry,
occupied state has the chiral symmetry or,
a generic vertical line does not have TRS or PHS, but
n/2
it may have the following symmetries: KP 0 ≡ KP Cn,m {Hr , Sr } = 0, (58)
0 n/2
and KT ≡ KT Cn,m , where we should note that TRS is
only possible for m = 0 and m = n/2. To see this, we where the eigenvalues are r ∈ (−1)F {1, ω ≡ ei2π/3 , ω̄ ≡
note that each generic vertical line is labeled by its mo- e−iπ/3 }. Sr here is the matrix representation of the chiral
mentum perpendicular to the rotation axis, k̄, and under symmetry S = P ∗ T projected to the r-sector. Eq.(58)
 11

‫ܭ‬ ‫ܯ‬ଵ
‫ܯ‬ଶ ‫ܯ‬ଷ ‫ܯ‬
‫ܯ‬ଶ
Ȟ ‫ܯ‬ଵ Ȟ ܺ Ȟ

‫ܯ‬ଷ
‫ܭ‬Ԣ
(a) (b) (c)
Ȟത ഥ ത
Ȟത ഥଶ ത 

ഥଵ
 ഥଷ
 ഥ
 ഥ ത Ԣ

  ഥ Ȟ

  Ԣ
 
 
Ȟ Ԣ

Ȟ   Ȟ
 
 

(d) (e) (f)

FIG. 1: (a,b,c) show the Brillouin zones of 2D systems with rotation symmetries, where (a) is for n = 2, (b) for n = 4 and (c)
for n = 3, 6. (d,e,f) show the Brillouin zones of 3D systems with rotation symmetries and the surface Brillouin zones for any
termination perpendicular to the rotation axis, where (d) is for n = 2, (e) for n = 4 and (f) for n = 3, 6.

indicates that each sector is in class AIII, and since class the r = ±1 sector is invariant under both KP 0 and KT 0 ,
AIII has Z-classification, one C3 -invariant line has Z 3 - and therefore belongs to class BDI, which has a Z-index.
classification. Further, we notice that one C3 -invariant The sectors with r = ±ω and r = ±ω̄ are invariant un-
line that does not include Γ is mapped to another C3 - der S 0 = KP 0 ∗ KT 0 and are mapped to each other under
invariant line under either the application of either TRS KP 0 or KT 0 . Therefore each of them is in class AIII,
or PHS. TRS, or equivalently PHS, flips the sign of the which has a Z-index, and the two indices are related to
invariant in each sector as it sends a state with S = +i each other by either KP 0 or KT 0 , therefore, the overall
to a state with S = −i, and maps the sector with r = ±ω classification becomes Z 2 .
to the one with r = ±ω̄. Therefore, if one line has
topological number (z (±ω) , z (±ω̄) , z (±1) ) = (z1 , z2 , z3 ), We move on to discuss the classifications of other high-
the topological number of the other line is fixed to be symmetry lines. The high-symmetry lines are separated
(z (±ω) , z (±ω̄) , z (±1) ) = (−z2 , −z1 , −z3 ). Furthermore, if in k-space, but their quantum numbers are not indepen-
a superconductor has C6,m -symmetry, the C3 -invariant dent of each other in a fully gapped 3D superconductor.
line also possesses KP 0 and if the system has time- This is because the special lines can be adiabatically con-
reversal, it posseses KT 0 . The class of the Hamiltonian nected to each other by an interpolation consisting of a
on this type of line depends on the class of the system series of generic lines which are not invariant under ro-
and the parity of m, determined via Eq.(57). As long tations or time-reversal. In Fig.1(e), for example, ΓZ
as the class of the line is one of the four classes that may continuously move to M A by shifting the two end
we have discussed, C, D, CI or DIII, one can simply use points along the dotted paths in the BZ, through a se-
Table I to obtain the C3 enhanced classification. How- ries of generic lines such as the dotted line that projects
ever, a separate analysis is needed for case (vi) (class CI onto k̄. The fully gapped bulk provides relations be-
and m ∈ odd), and case (viii) (class DIII and m ∈ odd), tween the invariants obtained on high-symmetry lines
where the line belongs to class BDI. With C3 -symmetry, and a generic line and the relationship may be summa-
rized in two simple yet general rules. The first rule states
 12

that if a generic line has trivial classification, the topo- at K̄, K̄ 0 [see Fig.1(f)]. Therefore, when C3 is broken,
logical invariants at high-symmetry lines are not related the total number of surface Majorana modes is given
to (constrained by) each other. It is supported by the by z(k̄) = z (−1) (K̄, K̄ 0 ) + z (−ω) (K̄, K̄ 0 ) + z (−ω̄) (K̄, K̄ 0 ).
following argument without a rigorous proof. Consider For the line that projects to Γ̄, both TRS and PHS are
two high-symmetry generic lines, L1,2 that have nontriv- separately preserved, so the number of +i MZMs must
ial topological invariants thus exhibiting n1,2 Majorana equal the number of −i Majorana modes, dictating that
bound states at high-symmetry points in the SBZ, L̄1,2 , z(k̄) = 0.
for a Cn -invariant termination. As L1 moves to another In 3D superconductors with time-reversal symmetry,
high-symmetry line L2 , since the generic lines are triv- classes CI and DIII, there is a Z-classification without
ial, the n1 Majorana modes at L̄1 move away from zero the existence of any additional symmetry. This Z-index
energy at a generic k̄ into the bulk, and as the path cannot be uniquely determined by the set of the above
reaches L2 , n2 modes emerge from the bulk to meet at indices given by high-symmetry lines. This can be under-
L̄2 with no relation to the L1 line. On the other hand, stood easily from the perspective of surface states: the
when the generic (vertical) lines have nontrivial classi- invariants defined on high-symmetry lines are only re-
fication, the topological invariant of a generic line gives lated to the number and chirality of Majorana cones at
constraints on those of high-symmetry lines (see below high symmetry points in the surface Brillouin zone, while
for an example). The second rule states that if any high- there can be protected Majorana cones away from these
symmetry line has trivial classification, the topological high-symmetry points. From this perspective, it is nat-
invariant of a generic line (given that the presence of ural that the total chirality (a topological index) can be
a nontrivial classification such that an invariant can be determined by the invariants on high-symmetry lines up
defined) must be zero (trivial). This can be proved by to a multiple of n, because any Majorana cone centered
contradiction: if a generic line has Majorana modes at k̄ at a generic k on the surface must coexist with the other
at zero energy, these modes will remain at zero energy as n − 1 Majorana cones of the same chirality due to the Cn
the line moves in the BZ to any high-symmetry line, so symmetry.
the high-symmetry line would have nontrivial classifica-
tion, against our assumption. For cases where both the
generic and high-symmetry lines have a non-trivial classi- B. Effective Surface Theory
fication, that is, when neither of the rules applies, special
analysis is needed. In the following we will derive these On the surface, topologically non-trivial superconduc-
relations for two cases in detail, and the complete result, tors have protected surface Majorana cones containing
obtained via identical methodology, is given in Table II. Majorana modes that have a specific chirality. As we
For our first example, we consider a class D Hamilto- have mentioned previously, these Majorana cones may
nian with C2,1 -invariance. At the high-symmetry lines or may not appear at high-symmetry points in the sur-
in BZ, ΓZ projecting to Γ̄, XU projecting to M̄1 , Y T face BZ, and yet in either case, the specific k · p form of
projecting to M̄2 and SR projecting to M̄3 in the sur- each cone is subject to constraints placed by symmetries.
face BZ [see Fig.1(d)], the Hamiltonian has Z2 × Z2 In this subsection, we study the constraints on these sur-
classification. The two Z2 -indices correspond to the face Majorana cones placed by the presence of rotational
two sectors with C2,1 = ±1. When the two-fold rota- symmetries. In general, for the rotation symmetry to be
tion symmetry is broken, the two Majorana modes from relevant, we require that the surface is invariant under
the two individual sectors may hybridize. Therefore, if Cn . In most cases, the generic form of a surface Majo-
(+) ¯ ) = z (−) (Γ̄, M1,2,3¯ ), there is no Majorana
z2 (Γ̄, M1,2,3 2 rana cones reads
(+) ¯ ) 6=
mode after adding the mass; and if z2 (Γ̄, M1,2,3 X
(−) ¯ h(q) = d0 (q) + di (q)σi . (59)
z2 (Γ̄, M1,2,3 ), the total number of zero energy Majo- i=x,y,z
rana modes at a high-symmetry point projection is one
(nonzero), and it remains at zero energy as one moves The symmetries place constraints on the form of
away from a high-symmetry point to a generic point k̄ d0,x,y,z (k) and it is our goal to find these constraints in
due to PHS and C2 , forming a flat band in the whole the presence of various symmetries including Cn,m , TRS
SBZ. Hence the constraint between z2 (k̄) (the Z2 invari- and SU(2). As before, we are interested in studying two
(±) ¯ ) is found to be
ant for a generic line) and z2 (Γ̄, M1,2,3 separate situations: (i) generic k0 within the surface BZ
z2 (k̄) = z (+1) ¯
(Γ̄, M1,2,3 ) + z (−1) ¯ ).
(Γ̄, M1,2,3 and (ii) k0 at high-symmetry points within the surface
Our second, more intricate, example concerns a Hamil- BZ.
tonian in class DIII with C3,0 -symmetry. For HK or
H 0 K 0 , the index is Z 3 , one integer for each of the three
eigenspaces of C3 . This integer index (for HK or for 1. Generic k0
H 0 K 0 ) equals the number of Majorana modes with cor-
responding eigenvalue of +i of the chiral operator mi- If there is no symmetry other than Cn and PHS, then
nus the number of −i Majorana modes in each sector there is no constraint on the form of h(q) for n ∈ odd,
 13

Class (n, m) k̄ M̄1,2,3 , X̄ K̄, K̄ 0 M̄ Γ̄ Constraints by a bulk gap

C (2, 0) 0 0 N/A N/A 0 None
C (2, 1) Z2 0 N/A N/A 0 z2 (k̄) = 0
C (3, 0), (3, 1), (3, 2) 0 N/A 0 N/A 0 None
C (4, 0), (4, 2) 0 0 N/A 0 0 None
C (4, 1), (4, 3) Z2 0 N/A 0 0 z2 (k̄) = 0
C (6, 0), (6, 2), (6, 4) 0 0 0 N/A 0 None
C (6, 1), (6, 3), (6, 5) Z2 0 Z2 N/A 0 z2 (k̄) = z2 (K̄) = 0
D (2, 0) 0 0 N/A N/A 0 None
(+1) (−1)
D (2, 1) Z2 Z2 × Z2 N/A N/A Z2 × Z2 z2 (M̄1,2,3 , Γ̄) + z2 (M̄1,2,3 , Γ̄) = z2 (k̄)
D (3, 0), (3, 1), (3, 2) 0 N/A 0 N/A Z2 None
D (4, 0), (4, 2) 0 0 N/A 0 0 None
(+1) (−1)
D (4, 1), (4, 3) Z2 Z2 × Z2 N/A Z2 × Z2 Z2 × Z2 z2 (Γ̄, M̄ , X̄) + z2 (Γ̄, M̄ , X̄) = z2 (k̄)
D (6, 0), (6, 2), (6, 4) 0 0 0 N/A 0 None
(+1) (−1)
D (6, 1), (6, 3), (6, 5) Z2 Z2 × Z2 Z2 N/A Z2 × Z2 z2 (Γ̄, M̄ ) + z2 (Γ̄, M̄ ) = z2 (k̄) = z2 (K̄)
CI (2, 0) 0 0 N/A N/A 0 None
CI (2, 1) Z 0 N/A N/A 0 z(k̄) = 0
CI (3, 0) Z N/A Z3 N/A Z z(k̄) = z (+1) (K̄) + z (ω) (K̄) + z (ω̄) (K̄) = 0
CI (4, 0) 0 0 N/A Z Z None
CI (4, 2) 0 0 N/A 0 0 None
CI (6, 0) 0 0 Z N/A Z2 None
CI (6, 3) Z 0 Z2 N/A 0 z(k̄) = z (−1) (K̄) + 2z (−ω) (K̄) = 0
DIII (2, 0) 0 Z N/A N/A Z None
DIII (2, 1) Z Z2 N/A N/A Z2 z(k̄) = 0
DIII (3, 0) Z N/A Z3 N/A Z2 × Z z(k̄) = z (−1) (K̄) + z (−ω) (K̄) + z (−ω̄) (K̄) = 0
DIII (4, 0) 0 Z N/A Z2 Z2 None
DIII (4, 2) 0 Z N/A 0 Z2 None
DIII (6, 0) 0 Z Z N/A Z3 None
DIII (6, 3) Z Z2 Z2 N/A Z2 z(k̄) = z (1) (K̄) + 2z (ω) (K̄) = 0

TABLE II: The complete classification of 1D vertical lines in the BZ of a 3D superconductor with Cn,m -invariance. The column
k̄ contains the classification for a generic line in the bulk. All other columns except the last contain the classification of high-
symmetry lines that project to high-symmetry points in the surface BZ. In the last column, we list the constraints between
the invariants on these lines placed by a full superconducting gap in the 3D BZ. Within the table, ’0’ indicates a trivial entry
while ’N/A’ indicates that the particular constraints considered in a particular tabular entry are not applicable.

2
and a Majorana cone is not protected. Yet for n ∈ even, respectively. Therefore, if m ∈ even, we have KP 0 = −1
P 0 as defined in Eq.(57) is a symmetry, which leads to and the constraint Eq.(64) makes the co-dimension of the
theory 1 and, hence, fully gapped generically. On the
KP 0 h(q)(KP 0 )−1 = −h(q). (60) 2
other hand, if m ∈ odd, we have KP 0 = 1, and the
constraints Eq.(63) makes the co-dimension −1, which
In the k · p basis, P 0 is represented by means the that theory is gapless along a certain direction
and the Majorana modes could appear as nodal rings in
P 0 = I, (61) the surface BZ.
and Beyond this, we consider the addition of TRS in an
effort to understand the manner in which the constraints
P 0 = iσy (62) change. The symmetries, in this case, of h(q) are KP 0
2
and KT 0 if n ∈ even or only the chiral symmetry,
corresponding to situations of P 0 = ∓1, respectively. S 0 = KP 0 ∗ KT 0 if n ∈ odd. For n ∈ odd, h(q) for a
Substituting Eq.(61) or Eq.(62) into Eq.(60), we have generic point q in the surface BZ belongs to class AIII.
For the symmetry representation, we choose S = σz ,
d0 (q) = dx (q) = dz (q) = 0, (63) which leads to the constraint d0 = dz = 0. For n ∈ even,
d0 (q) = 0, (64) using Eqs.(57), we know that: (i) if m ∈ even, then h(q)
 14

belongs to class CI, where P 0 = (iσy ) and T 0 = I and (ii) Majorana cone.
if m ∈ odd, h(q) belongs to class BDI, where P 0 = σx
and T 0 = I. Then by using We proceed to discuss the case in which (n, m) = (2, 1).
Considering class C, we use the fact that KP 2 = C2,1 =
−1 −1 in conjunction with P = (iσy ) and C2,1 = iσz ,
T 0 h(q)T 0 = h∗ (q), (65)
0 0 −1 ∗
thereby allowing a mass term mσz that results in the
P h(q)P = −h (q), formation of a gap in the surface BZ. Moving to class D,
we take advantage of KP 2 = C2,1 2
= 1, where we have
we find the constraints that for m ∈ even, d0 = dy = 0 P = σx and C2,1 = I, or C2,1 = σx , which correspond
and for m ∈ odd, then d0 = dx = dy = 0. When two to the cases where the two bands have the same and the
out of four di ’s are required to vanish, then the resultant opposite C2,1 eigenvalues respectively. With the case of
co-dimension is zero, and a Majorana cone, if exists, is C2,1 = I, mass term mσz may be added, however, for
locally stable within the surface BZ. Furthermore, when C2,1 = σx , mass terms are disallowed and the resultant
three out of four di ’s are required to vanish by the sym- lowest order surface theory is a linear term (vkx +wky )σy ,
metry constraints, then the co-dimension is −1 and there which indicates a nodal line in the SBZ passing through
is a nodal (Majorana) line in the surface bands. In all the high-symmetry point. This is consistent with the pre-
the discussion above, we have implicitly assumed that the vious result that for (n ∈ even, m ∈ odd), the effective
vertical line which projects to a generic point is always theory around a generic point exhibits nodal lines in class
trivial. In the case where a generic vertical line is non- C. With class CI, we have: P = iσy , T = σx and C2,1 =
trivial, there will be Majorana flat bands on the surface. iσx where we have used [KP, C2,1 ] = {KT, C2,1 } = 0. In
By following this identical methodology, one can sweep this case, a mass term m1 σx + m2 σy can be added that
out each of the separate permutations of symmetry con- naturally gaps the effective surface theory. Lastly, in
straints to derive effective theories around a generic point class DIII, we have P = σx , T = K = iσy and C2,1 = σz .
in the surface BZ, the results of which we summarize in The mass terms in class DIII are disallowed and the low-
Table III. est order
P terms allowable in the surface theory are in the
form ij Aij ki σj , which represents a Majorana cone.
2. k0 at high-symmetry points We note that there are several cases where a simple
two-band model cannot adequately describe the symme-
Additional constraints on the effective k·p theory apply try groups. Put another way, this indicates that the
if k0 is invariant under rotation, in other words k0 is symmetry group does not have any 2D irreducible rep-
at a high-symmetry point within the surface BZ. As we resentation. These cases are: (i) a class DIII Hamil-
have noted earlier, these high-symmetry points within tonian with (n, m) = (4, 2) (ii) a class CI Hamiltonian
the surface BZ may be either a two-, three-, four- or with (n, m) = (6, 3) and (iii) a class CI Hamiltonian with
six-fold invariant points. While our goal is to provide a (n, m) = (6, 3), where in each case the smallest represen-
complete classification for the effective surface theories of tation is four-dimensional. As an example, let us discuss
gapped 3D superconductors when the selected line within class DIII with (n, m) = (4, 2); the other cases can be
the surface BZ lies at a high-symmetry point, in what similarly discussed. Suppose we have one state in the
follows we address three specific situations in detail for C4,2 = eiπ/4 -sector, then TRS takes it to the −e−iπ/4 -
(n, m) = (2, 0), (n, m) = (2, 1) and (n, m) = (4, 2). We sector [using {T, C4,2 } = 0 from Eq.(34)], so we have
provide the complete results for arbitrary (n, m) listed in two states with two different C4,2 -eigenvalues. Yet PHS
Table IV, V, VI, VII for C2 , C3 , C4 , and C6 respectively. will send these two states to another two states, in the
For (n, m) = (2, 0), there are four high-symmetry C4,2 = e−iπ/4 -sector and −eiπ/4 -sector. Therefore, there
points in the surface BZ: Γ̄ and M1,2,3 ¯ , each of which must be at least four states, one in each sector, to realize
is C2 -invariant. Therefore the analysis is the same for the full symmetry group. We can choose the symmetries
all four high-symmetry points. In class C, we utilize to be representsd by: P = σ0 ⊗ σx , T = (iσy ) ⊗ σx and
KP 2 = −1 and C2,0 2
= 1, in which case we have C4,2 = σx ⊗ exp(iπσz /4) and notice that a mass term
P = (iσy ) and C2,0 = σ0 . This shows that a mass term mσz ⊗ σz is allowed, thereby rendering the effective sur-
mσx,y,z can be added to the surface theory so that it is face theory gapped. Here the fact that the spectrum can
gapped. Considering a class D system, using P 2 = 1 and be fully gapped and that the irreducible (projective) rep-
2
C2,0 = −1, we have P = σx and C2 = iσz , therefore, in resentation is at least four-dimensional do not contradict,
this case, we may also have a mass term mσz in the the- as they would in a system without particle-hole symme-
ory, effectively gapping the system again. In class CI, in try (non-BdG Hamiltonian). In the latter case, a four-
addition to C2,0 and KP , represented by the same matri- dimensional (or any higher than one-dimensional) repre-
ces as in class C, we have T = I, so once again mass terms sentation implies that the single particle spectrum must
such as mσx,z can be added, and the theory is gapped. be gapless. The distinction is because that for a BdG
However, in class DIII, we have T = (iσy ) and the resul- Hamiltonian, the operator KP anti-commutes, rather
tant theory is massless with the lowest order expansion than commutes with the Hamiltonian, and is hence not
being h(kx , ky ) = Aij ki σj where i, j = x, y, typical of a a real symmetry.
 15

Class KP 0 KT 0 S k·p Node type

C/D, n ∈ odd N/A N/A N/A dx σx + dy σy + dz σz Gapped
C/D, (n ∈ even, m ∈ even) Kiσy N/A N/A dx σx + dy σy + dz σz Gapped
C/D, (n ∈ even, m ∈ odd) K N/A N/A dy σy Nodal line
CI/DIII, n ∈ odd N/A N/A σz dx σx + dy σy Nodal point
CI/DIII, (n ∈ even, m ∈ even) Kiσy K σy dx σx Nodal line
CI/DIII, (n ∈ even, m ∈ odd) K Kσz σz dy σy Nodal line

TABLE III: The effective surface theories in the vicinity of a generic point in the surface BZ for 3D gapped superconductors
constrained by the presence of Cn,m symmetries and in the presence of particle-hole symmetry, time-reversal symmetry or S,
the combination of particle-hole symmetry and time-reversal symmetry.

Class KP KT Cn,m k·p Node type

C K(iσy ) N/A C2,0 = ±I m1 σx + m2 σy + m3 σz Gapped
C K(iσy ) N/A C2,1 = iσz mσz Gapped
D Kσx N/A C2,0 = iσz mσz Gapped
D Kσx N/A C2,1 = ±1 mσz Gapped
D Kσx N/A C2,1 = σx (vkx + wky )σy Nodal line
CI K(iσy ) Kσx C2,0 = ±I m1 σx + m2 σy Gapped
CI K(iσy ) Kσx C2,1 = iσx mσx Gapped
DIII Kσx K(iσy ) C2,0 = iσz Aij ki σj Linear
DIII Kσx K(iσy ) C2,1 = σz Aij ki σj Linear

TABLE IV: The effective theories on the surface in the vicinity of a C2 -invariant point in the surface BZ for 3D gapped
superconductors constrained by the presence of Cn,m symmetries and in the presence of particle-hole symmetry, time-reversal
symmetry or S, the combination of particle-hole symmetry and time-reversal symmetry. An entry of ’N/A’ indicates that the
particular constraints considered in a particular tabular entry are not applicable.

VI. CONCLUSION angular momentum of the Cooper pairs. In 3D, we found

the bulk topological invariants for all rotational symme-
In conclusion, we have examined the topological prop- tries, linearly and projectively represented, defined on
erties of the Bogoliubov - de Gennes Hamiltonians from high-symmetry lines and generic lines; we correspond-
the Altland-Zimbauer classification scheme correspond- ingly found the relevant surface theories for both generic
ing to gapped topological superconductors in one, two and high-symmetry points in the surface Brillouin zone.
and three spatial dimensions in the presence of rotational Acknowledgments
symmetry. In 1D, we complete the classification by block-
diagonalizing the Hamiltonian into sectors labeled by ro-
tation eigenvalues, and finding the topological index, Z MJG would like to thank Dan Arovas for enlightening
or Z2 or trivial, for each sector. The role played by the conversations. CF was supported by Project 11674370
angular momentum of the Cooper pairs is emphasized: it by NSFC and the National Key Research and Devel-
generally leads to a nontrivial projective representation opment Program under grant number 2016YFA0302400
of rotation symmetry, where anomalous commutation re- and 2016YFA0300600. CF, and MJG were supported
lations include Cn2 6= (−1)F and the anti-commutation by ONR - N0014-11-1-0123 and MJG was supported
between the time-reversal and the rotation. In 2D, we by NSF-CAREER EECS-1351871. BAB was supported
explicitly proved the relation between the Chern number by De- partment of Energy de-sc0016239, NSF EAGER
and the rotation eigenvalues at high-symmetry points, Award NOA - AWD1004957, Simons Investigator Award,
and showed that in the weak-coupling limit, the contri- ONR - N00014-14-1-0330, ARO MURI W911NF-12-1-
bution breaks down to two parts. One part is from the 0461, NSF-MRSEC DMR-1420541, Packard Foundation
normal state band structure the other part is from the and Schmidt Fund for Innovative Research.
 16

Class KP KT Cn,m k·p Node type

C K(iσy ) N/A C3,0 = I m1 σx + m2 σy + m3 σz Gapped
2π
C K(iσy ) N/A C3,0 = exp(i 3 σz ) mσz Gapped
C K(iσy ) N/A C3,1 , C3,2 = −I m1 σx + m2 σy + m3 σz Gapped
C K(iσy ) N/A C3,1 , C3,2 = − exp(i 2π 3
σ z ) mσ z Gapped
D Kσx N/A C3,0 = −I mσz Gapped
D Kσx N/A C3,0 = − exp(i 2π 3
σ z ) mσ z Gapped
D Kσx N/A C3,1 , C3,2 = I mσz Gapped
D Kσx N/A C3,1 , C3,2 = exp(i 2π 3 z
σ ) mσ z Gapped
CI K(iσy ) Kσx C3,0 = I m1 σx + m2 σy Gapped
2
CI K(iσy ) Kσx C3,0 = exp(i 2π 3
σ z ) ck+ σ − + h.c. Quadratic
3 3
DIII Kσx K(iσy ) C3,0 = −I c 1 k+ σ+ + c2 k− σ− + h.c. Cubic
2π
DIII Kσx K(iσy ) C3,0 = − exp(i 3 σz ) ck+ σ− + h.c. Linear

TABLE V: The effective theories on the surface in the vicinity of a C3 -invariant point within the surface BZ for 3D gapped
superconductors constrained by the presence of Cn,m symmetries and in the presence of particle-hole symmetry, time-reversal
symmetry or S, the combination of particle-hole symmetry and time-reversal symmetry. An entry of ’N/A’ indicates that the
particular constraints considered in a particular tabular entry are not applicable.

Class KP KT Cn,m k·p Node type

C K(iσy ) N/A C4,0 , C4,2 = ±I m1 σx + m2 σy + m3 σz Gapped
C K(iσy ) N/A C4,0 , C4,2 = iσz mσz Gapped
C K(iσy ) N/A C4,1 , C4,3 = ± exp(i π4 σz ) mσz Gapped
D Kσx N/A C4,0 , C4,2 = ± exp(i π4 σz ) mσz Gapped
D Kσx N/A C4,1 , C4,3 = ±1 mσz Gapped
2 2
D Kσx N/A C4,1 , C4,3 = σx [r1 (kx − ky ) + r2 kx ky ]σz Nodal line
D Kσx N/A C4,1 = iσz mσz Gapped
CI K(iσy ) Kσx C4,0 = ±I m1 σx + m2 σy Gapped
2 2
CI K(iσy ) Kσx C4,0 = iσz c1 k+ σ+ + c2 k− σ− + h.c. Quadratic
2 2
CI K(iσy ) Kσx C4,2 = σz c1 k+ σ+ + c2 k− σ− + h.c. Quadratic
CI K(iσy ) Kσx C4,2 = iσx mσx Gapped
DIII Kσx K(iσy ) C4,0 = ± exp(i π4 σz ) ck+ σ− + h.c. Linear
DIII Kσ0 ⊗ σx K(iσy ) ⊗ σx C4,2 = σz ⊗ exp(i π4 σz ) mσz ⊗ σz Gapped

TABLE VI: The effective theories on the surface in the vicinity of a C4 -invariant point within the surface BZ for 3D gapped
superconductors constrained by the presence of Cn,m symmetries and in the presence of particle-hole symmetry, time-reversal
symmetry or S, the combination of particle-hole symmetry and time-reversal symmetry. Within the table an entry of ’N/A’
indicates that the particular constraints considered in a particular tabular entry are not applicable.

Appendix A: Linear and projective representations of the group generated by P and Cn

A linear representation of the group generated by P and Cn must satisfy

D(P )D∗ (P ) = 1, (1)
Dn (Cn ) = (−1)F ,
D(Cn )D(P ) = D(P )D∗ (Cn ).
All representations that do not obey these relations, but obey them up to some phase factor, are called projective
representations. However, there is a class of projective representations that are trivial, as they can be transformed
back into a linear representation by multiplying each element some phase factor. If a projective representation cannot
be brought back to a linear representation, it is called a nontrivial projective representation.
We prove by contradiction that if a projective representation satisfies
D(Cn )D(P ) = D(P )D∗ (Cn ), (2)
Dn (Cn ) = −(−1)F ,
 17

Class KP KT Cn,m k·p Node type

C K(iσy ) N/A C6,0 , C6,2 , C6,4 = ±I m1 σx + m2 σy + m3 σz Gapped
π
C K(iσy ) N/A C6,0 , C6,2 , C6,4 = ± exp(i 3 σz ) mσz Gapped
C K(iσy ) N/A C6,1 , C6,3 , C6,5 = ± exp(i π6 σz ) mσz Gapped
C K(iσy ) N/A C6,1 , C6,3 , C6,5 = iσz mσz Gapped
π
D Kσx N/A C6,0 , C6,2 , C6,4 = ± exp(i 6 σz ) mσz Gapped
D Kσx N/A C6,0 , C6,2 , C6,4 = iσz mσz Gapped
D Kσx N/A C6,1 , C6,3 , C6,5 = ±I mσz Gapped
3
D Kσx N/A C6,1 , C6,3 , C6,5 = σx ck+ σy + h.c. Nodal line
D Kσx N/A C6,1 , C6,3 , C6,5 = ± exp(i π3 σz ) mσz Gapped
CI K(iσy ) Kσx C6,0 = ±I m1 σx + m2 σy Gapped
2
CI K(iσy ) Kσx C6,0 = ± exp(i π3 σz ) ck+ σ− + h.c. Quadratic
CI Kσz ⊗ (iσy ) Kσx ⊗ σx C6,3 = σz ⊗ exp(i π6 σz ) mσ0 ⊗ σz Gapped
CI K(iσy ) Kσx C6,3 = iσx mσx Gapped
DIII Kσx K(iσy ) C6,0 = ± exp(i π6 σz ) ck+ σ− + h.c. Linear
3 3
DIII Kσx K(iσy ) C6,0 = iσz c 1 k+ σ+ + c2 k− σ− + h.c. Cubic
3
DIII Kσx K(iσy ) C6,3 = σx ck+ σy + h.c. Nodal line
π
DIII Kσ0 ⊗ σx K(iσy ) ⊗ σx C6,3 = σz ⊗ exp(i 3 σz ) mσz ⊗ σz Gapped

TABLE VII: The effective theories on the surface in the vicinity of a C6 -invariant point within the surface BZ for 3D gapped
superconductors constrained by the presence of Cn,m symmetries and in the presence of particle-hole symmetry, time-reversal
symmetry or S, the combination of particle-hole symmetry and time-reversal symmetry. Within the table an entry of ’N/A’
indicates that the particular constraints considered in a particular tabular entry are not applicable.

then it must be nontrivial.

Suppose it is trivial, then we can redefine the generators as
D(P ) → D(P )eiω(P ) , (3)
iω(Cn )
D(Cn ) → D(Cn )e ,
so that Eq.(1) becomes
D(P )D∗ (P ) = 1, (4)
n inω(Cn ) F
D (Cn ) = e (−1) ,
D(Cn )D(P )e iω(Cn )
= D(P )D (Cn )e−iω(Cn ) .
∗

In order to satisfy Eq.(3), one requires

ω(Cn ) = mπ/n, ω(Cn ) = 0, π, (5)
where m ∈ odd. It is obvious that the two equations are contradictory. Therefore, a representation satisfying Eq.(3)
is a nontrivial projective representation.

Appendix B: Linear and projective representation of the group generated by P , T , and Cn

A linear representation of the group should satisfy, besides the equations in Eq.(1)
D(P )D∗ (T ) = D(T )D∗ (P ), (1)
D(T )D∗ (Cn ) = D(Cn )D(T ),
D(T )D∗ (T ) = (−1)F .
We prove that when n ∈ even, representations that satisfy
D(T )D∗ (Dn ) = −D(Cn )D(T ), (2)
D(P )D∗ (Cn ) = D(Cn )D(P ).
 18

FIG. 2: A closed path that encircles a sector of angle θ.

must be nontrivial.
Again we first assume that it is trivial, so that one can find ω(T ) and ω(Cn ) such that Eq.(2) is satisfied. Multiply
D(T ) and D(Cn ) with phase factors eiω(T ) and eiω(Cn ) so that they satisfy

D(T )D∗ (Cn )e−iω(Cn ) = D(Cn )D(T )eiω(Cn ) , (3)

∗ −iω(Cn ) iω(Cn )
D(P )D (Cn )e = D(Cn )D(P )e .

In order to satisfy Eq.(2), one requires, respectively

ω(Cn ) = ±π, (4)

ω(Cn ) = 0, π

which obviously contradict each other. Therefore, a representation satisfying Eq.(2) is indeed a nontrivial projective
representation.

Appendix C: General Proof of Chern Number for C∞ Rotational Invariance

Here we assume that the wavefunction, or more precisely, the projector onto the occupied space at k = ∞ is well
defined, which is denoted by P (∞). This ensures a closed manifold, which is necessary for a well regularized Chern
number calculation.
Consider a loop that consists of three parts: (i) a straight line from k = 0 to a very large k1 , the azimuthal angle
of which is zero, (ii) an arc going counterclockwise by δθ from k1 to k2 and (iii) a straight line going from k2 back to
k = 0 (see Fig.2). Calling them segments 1, 2, 3 and dividing each segment into N1,2,3 smaller segments, the Wilson
loop of the above loop is

W = U1 U2 U3 , (1)

where
Y
U1 = P1,i , (2)
i=1,...,N1
Y
U2 = P2,i , (3)
i=1,...,N2
Y
U3 = P3,i , (4)
i=1,...,N3
 19

where Pn,i is the projection operator onto the occupied states at the i-th point of the n-th segment. Since P (∞) is
well defined, we have that when |k1 | = |k2 | → ∞, we have P2,i = P2,j , or

U2 = P2,1 . (5)

Therefore

lim W = U0k1 Uk2 0 . (6)

|k1 |→∞

Then we notice that

U0k2 = R(δθ)U0k1 R(−δθ), (7)

where R(θ) is the rotation operator through θ, and that

lim R(δθ)U0k1 R(−δθ) = eij(0)δθ U0k1 e−ij(∞)δθ . (8)

|k1 |→∞

From all above equations and using U0k1 Uk1 0 = 1, we have

lim W = exp[i(j(0) − j(∞))δθ]. (9)

|k1 |→∞

According to the relation between the Berry curvature integral and the Wilson loop we know that
Z ∞ Z δθ
kdk dθF (k, θ) = 2πn + (j(0) − j(∞))δθ (10)
0 0
= (j(0) − j(∞))δθ.

The last equality uses the fact that for very small δθ, the integral must also be very small, so the integer part is zero.
The C∞ symmetry, the Berry’s curvature is also rotationally invariant, so
Z
1
C= dk 2 F (k, θ) = j(0) − j(∞). (11)
2π

1
C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 106803 (2007).
12
(2005). X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Phys. Rev. B 78,
2
C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 146802 195424 (2008), URL http://link.aps.org/doi/10.1103/
(2005). PhysRevB.78.195424.
3 13
B.A. Bernevig and S.C. Zhang, Phys. Rev. Lett. 96, D. Hsieh, Y. Xia, D. Qian, L. Wray, J. H. Dil, F. Meier,
106802 (2006). L. Patthey, J. Osterwalder, A. Fedorov, A. B. H. Lin, et al.,
4
B. A. Bernevig, T. L. Hughes, and S.C. Zhang, Science Nature 460, 1101 (2009).
14
314, 1757 (2006). D. Hsieh, Y. Xia, L. Wray, D. Qian, A. Pal, J. H. Dil,
5
M. König, S. Wiedmann, C. Brüne, A. Roth, H. Buhmann, J. Osterwalder, F. Meier, G. Bihlmayer, C. L. Kane, et al.,
L. Molenkamp, X.-L. Qi, and S.-C. Zhang, Science 318, Science 323, 919 (2009).
15
766 (2007). D. Hsieh, Y. Xia, D. Qian, L. Wray, J. H. Dil, F. Meier,
6
A. Roth, C. Brune, H. Buhmann, L. W. Molenkamp, J. Osterwalder, L. Patthey, A. V. Fedorov, H. Lin, et al.,
J. Maciejko, X.-L. Qi, and S.-C. Zhang, Science 325, 294 Phys. Rev. Lett. 103, 146401 (2009).
16
(2009). Y. L. Chen, J. G. Analytis, J. H. Chu, Z. K. Liu, S. K. Mo,
7
C. Liu, T. L. Hughes, X.-L. Qi, K. Wang, and S.-C. Zhang, X. L. Qi, H. J. Zhang, D. H. Lu, X. Dai, Z. Fang, et al.,
Phys. Rev. Lett. 100, 236601 (2008). Science 325, 178 (2009), URL http://www.sciencemag.
8
I. Knez, R. R. Du, and G. Sullivan, Phys. Rev. Lett. 107, org/cgi/content/abstract/325/5937/178.
17
136603 (2011). Y. Xia, L. Wray, D. Qian, D. Hsieh, A. Pal, H. Lin, A. Ban-
9
I. Knez, R. R. Du, and G. Sullivan, Phys. Rev. Lett. 109, sil, D. Grauer, Y. Hor, R. Cava, et al., Nat. Phys. 5, 398
186603 (2012). (2009).
10 18
X. Qian, J. Liu, L. fu, and J. Li, Science 346, 1344 (2014). P. Roushan, J. Seo, C. V. Parker, Y. S. Hor, D. Hsieh,
11
L. Fu, C. L. Kane, and E. J. Mele, Phys. Rev. Lett. 98, D. Qian, A. Richardella, M. Z. Hasan, R. J. Cava, and
 20

40
A. Yazdani, Nature 460, 1106 (2009), URL http://dx. T. Hsieh, H. Lin, J. Liu, W. Duan, A. Bansil, and L. Fu,
doi.org/10.1038/nature08308. Nature Communications 3, 982 (2012).
19 41
Z. Alpichshev, J. G. Analytis, J.-H. Chu, I. R. Fisher, J. Liu, W. Duan, and L. Fu, arXiv:1304.0430 (2013).
42
Y. L. Chen, Z. X. Shen, A. Fang, and A. Kapitulnik, Phys. J. Liu, T. H. Hsieh, P. Wei, D. Wenhui, J. Moodera, and
Rev. Lett. 104, 016401 (2010), URL http://link.aps. L. Fu, arXiv:1310.1044 (2013).
43
org/doi/10.1103/PhysRevLett.104.016401. C. Fang, M. J. Gilbert, and B. A. Bernevig,
20
Y. L. Chen, J.-H. Chu, J. G. Analytis, Z. K. Liu, arXiv:1306.0888 (2013).
44
K. Igarashi, H.-H. Kuo, X. L. Qi, S. K. Mo, R. G. Moore, C. Fang, M. J. Gilbert, and B. A. Bernevig,
D. H. Lu, et al., Science 329, 659 (2010), ISSN 0036-8075, arXiv:1308.2424 (2013).
45
http://science.sciencemag.org/content/329/5992/659.full.pdf, C. Fang, M. J. Gilbert, S.-Y. Xu, B. A. Bernevig, and M. Z.
URL http://science.sciencemag.org/content/329/ Hasan, Phys. Rev. B 88, 125141 (2013).
46
5992/659. C.-X. Liu, R.-X. Zhang, and B. K. VanLeeuwen, Phys.
21
C. L. Liu, X. L. Qi, H. J. Zhang, Z. F. X. Dai, and S. C. Rev. B 90, 085304 (2014).
47
Zhang, Phys. Rev. B 82, 045122 (2010). C. Fang and L. Fu, Phys. Rev. B 91, 161105(R) (2015).
22 48
J. Seo, P. Roushan, H. Beidenkopf, Y. S. Hor, R. J. Cava, K. Shiozaki, M. Sato, and K. Gomi, Phys. Rev. B 91,
and A. Yazdani, Nature 466, 343 (2010). 155120 (2015).
23 49
J. G. Checkelsky, Y. S. Hor, R. J. Cava, and N. P. Ong, L. Lu, C. Fang, S. G. Johnson, J. D. Joannopoulos, and
Phys. Rev. Lett. 106, 196801 (2011), URL http://link. M. Soljacic, Nature Physics 12, 337 (2016).
50
aps.org/doi/10.1103/PhysRevLett.106.196801. Z. Wang, A. Alexandradinata, R. J. Cava, and B. A.
24
Y. Okada, C. Dhital, W. Zhou, E. D. Huemiller, H. Lin, Bernevig, Nature 532, 189 (2016).
51
S. Basak, A. Bansil, Y.-B. Huang, H. Ding, Z. Wang, et al., A. Alexandradinata, Z. Wang, and B. A. Bernevig, Phys.
Phys. Rev. Lett. 106, 206805 (2011), URL http://link. Rev. X 6, 021008 (2016).
52
aps.org/doi/10.1103/PhysRevLett.106.206805. P. Dziawa, B. J. Kowalski, K. Dybko, R. Buczko, A. Szczer-
25
C.-Z. Chang, J. Zhang, X. Feng, J. Shen, bakow, M. Szot, E. Ausakowska, T. Balasubramanian,
Z. Zhang, M. Guo, K. Li, Y. Ou, P. Wei, B. M. Wojek, M. H. Berntsen, et al., Nature Materials
L.-L. Wang, et al., Science 340, 167 (2013), advance online publication (2012).
53
http://www.sciencemag.org/content/340/6129/167.full.pdf, Y. Tanaka, Z. Ren, T. Sato, K. Nakayama, S. Souma,
URL http://www.sciencemag.org/content/340/6129/ T. Takahashi, K. Segawa, and Y. Ando, Nat. Phys. 8, 800
167.abstract. (2012).
26 54
T. Senthil, Annual Review of Condensed Matter Physics S.-Y. Xu, C. Liu, N. Alidoust, D. Qian, M. Neupane, J. D.
6, 299 (2015). Denlinger, Y. J. Wang, L. A. Wray, R. J. Cava, H. Lin,
27
Y. Ando and L. Fu, Annual Review of Condensed Matter et al., Nat Commu 3, 1192 (2012).
55
Physics 6, 361 (2015). J.-Z. e. Ma, arXiv:1605.06824 (2016).
28 56
C.-K. Chiu, J. C. Teo, A. P. Schnyder, and S. Ryu, C. Fang, M. J. Gilbert, X. Dai, and B. A. Bernevig, Phys.
arXiv:1505.03535 (2015). Rev. Lett. 108, 266802 (2012), URL http://link.aps.
29
S. Chadov, X. Qi, J. Kubler, G. H. Fecher, C. Felser, and org/doi/10.1103/PhysRevLett.108.266802.
57
S.-C. Zhang, Nature Materials 9, 541 (2010). G. Xu, H. Weng, Z. Wang, X. Dai, and Z. Fang, Phys.
30
H. Lin, A. Wray, Y. Xia, S. Xu, R. J. Cava, A. Bansil, and Rev. Lett. 107, 186806 (2011).
58
M. Z. Hasan, Nature Materials 9, 546 (2010). Z. Wang, Y. Sun, X.-Q. Chen, C. Franchini, G. Xu,
31
M. Hirschberger, S. Kushwaha, Z. Wang, Q. Gibson, H. Weng, X. Dai, and Z. Fang, Phys. Rev. B 85, 195320
S. Liang, C. A. Belvin, B. A. Bernevig, R. J. Cava, and (2012), ISSN 1098-0121.
59
N. P. Ong (2016). Z. K. Liu, B. Zhou, Z. J. Wang, H. M. Weng, D. Prab-
32
C. Shekhar, A. K. Nayak, S. Singh, N. Kumar, S.-C. hakaran, S. K. Mo, Y. Zhang, Z. X. Shen, Z. Fang, X. Dai,
Wu, Y. Zhang, A. C. Komarek, E. Kampert, Y. Skourski, et al., arXiv:1305.6780 (2013).
60
J. Wosnitza, et al. (2016). M. Neupane, S. Y. Xu, N. Alidoust, G. Bian, C. Liu, I. Be-
33
P. C. Canfield, J. D. Thompson, W. P. Beyermann, A. Lac- lopolski, T. R. Chang, H. T. Jeng, H. Lin, A. Bansil, et al.,
erda, M. F. Hundley, E. Peterson, Z. Fisk, and H. R. Ott, arXiv:1309.7892 (2013).
61
J. Appl. Phys. 70, 5800 (1991). S. Borisenko, Q. Gibson, D. Evtushinsky, V. Zabolotnyy,
34
Z. Fisk, P. C. Canfield, W. P. Beyermann, J. D. Thomp- B. Büchner, and R. J. Cava, arXiv:1309.7978 (2013).
62
son, M. F. Hundley, H. R. Ott, E. Felder, M. B. Maple, A. P. Schnyder, S. Ryu, A. Furusaki, and A. W. W. Lud-
M. A. Lopez de la Torre, P. Visani, et al., Phys. Rev. wig, Phys. Rev. B 78, 195125 (2008).
63
Lett. 67, 3310 (1991), URL http://link.aps.org/doi/ S. Ryu, A. P. Schnyder, A. Furusaki, and A. W. W. Lud-
10.1103/PhysRevLett.67.3310. wig, New Journal of Physics 12, 065010 (2010).
35 64
L. Fu, Phys. Rev. Lett. 106, 106802 (2011). A. P. Schnyder and S. Ryu, Phys. Rev. B 84, 060504
36
T. L. Hughes, E. Prodan, and B. A. Bernevig, Phys. Rev. (2011).
65
B 83, 245132 (2010). X. L. Qi, T. L. Hughes, and S. C. Zhang, Phys. Rev. Lett.
37
A. M. Turner, Y. Zhang, and A. Vishwanath, Phys. Rev. 102, 187001 (2009).
66
B 82, 241102 (2010). X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Phys. Rev. B 81,
38
A. M. Turner, Y. Zhang, R. S. K. Mong, and A. Vish- 134508 (2010), URL http://link.aps.org/doi/10.1103/
wanath, Phys. Rev. B 85, 165120 (2012), URL http: PhysRevB.81.134508.
67
//link.aps.org/doi/10.1103/PhysRevB.85.165120. L. Fu and E. Berg, Phys. Rev. Lett. 105, 097001 (2010),
39
C. Fang, M. J. Gilbert, and B. A. Bernevig, Phys. Rev. B URL http://link.aps.org/doi/10.1103/PhysRevLett.
86, 115112 (2012), URL http://link.aps.org/doi/10. 105.097001.
68
1103/PhysRevB.86.115112. F. Zhang, C. L. Kane, and E. J. Mele, Phys. Rev. Lett.
 21

111, 056403 (2013), URL http://link.aps.org/doi/10. article/pii/S0921452607010447.

75
1103/PhysRevLett.111.056403. N. P. Butch, P. Syers, K. Kirshenbaum, A. P. Hope, and
69
C.-K. Chiu, H. Yao, and S. Ryu, Phys. Rev. B 88, J. Paglione, Phys. Rev. B 84, 220504 (2011), URL http:
075142 (2013), URL http://link.aps.org/doi/10.1103/ //link.aps.org/doi/10.1103/PhysRevB.84.220504.
76
PhysRevB.88.075142. A. Altland and M. R. Zimbauer, Phys. Rev. B 55, 1142
70
G. E. Volovik, Phys. Lett. A 128, 277 (1988). (1997).
71 77
G. Y. Cho, J. H. Bardarson, Y.-M. Lu, and J. E. Moore, A. Alexandradinata, C. Fang, M. J. Gilbert, and B. A.
Phys. Rev. B 86, 214514 (2012). Bernevig, Phys. Rev. Lett. 113, 116403 (2014).
72 78
V. Shivamoggi and M. J. Gilbert, Phys. Rev. B 88, 134504 J. C. Y. Teo and T. L. Hughes, Phys. Rev. Lett. 111,
(2013). 047006 (2013).
73 79
K. Shiozaki and M. Sato, Phys. Rev. B 90, 165114 (2014), A. Y. Kitaev, Physics-Uspekhi 44, 131 (2001), URL http:
URL http://link.aps.org/doi/10.1103/PhysRevB.90. //stacks.iop.org/1063-7869/44/i=10S/a=S29.
80
165114. W. Scherrer, Elemente der Mathematik 1, 97 (1946).
74 81
G. Goll, M. Marz, A. Hamann, T. Tomanic, K. Grube, C. K. Chiu and A. Schnyder, Phys. Rev. B 90, 205136
T. Yoshino, and T. Takabatake, Physica B: Con- (2014).
82
densed Matter 403, 1065 (2008), ISSN 0921-4526, W. A. Benalcazar, J. C. Y. Teo, and T. L. Hughes, Phys.
¡xocs:full-name¿Proceedings of the International Confer- Rev. B 89, 224503 (2014).
ence on Strongly Correlated Electron Systems¡/xocs:full-
name¿, URL http://www.sciencedirect.com/science/