Skip to main content

Full text of "Skyrmions with quadratic band touching fermions: A way to achieve charge 4e superconductivity"

See other formats


Skyrmions with quadratic band touching fermions: 
A way to achieve charge 4e superconductivity. 



Eun-Gook Moon 

Department of Physics, University of California, Santa Barbara, CA 93106 

(Dated: April 10, 2012) 

We study Skyrmion quantum numbers, charge and statistics, in (2 + 1) dimension induced by 
quadratic band toucing(QBT) fermions. It is shown that induced charge of Skyrmions is twice 
bigger than corresponding Dirac particles' and their statistics are always bosonic. Applying to the 
Bernal stacking bi-layer graphene, we show that Skyrmions of quantum spin Hall(QSH) are charge 4e 
bosons, so their condensation realizes charge 4e superconductivity(SC). The phase transition could 
be a second order, and one candidate theory of the transition is 0(5) non linear sigma model(NLSM) 
with non-zero Wess-Zumino-Witten(WZW) term. We calculate renormalization group beta function 
of the model perturbatively and propose a possible phase diagram. We also discuss how QBT 
fermions are different from two copies of Dirac particles. 



Introduction : Electronic strucutres with helicity such 
as ones of Dirac particles and quadratic band touch- 
ing(QBT) fermions in two-spatial dimension receive a lot 
of attention especially with the discovery of graphene^ 
The helicity and its associated Berry phase play an im- 
portant role in quantum mechancal phenome na such as 
half-integer Quantum Hall effects in graphene P^l In par- 
ticular, QBT fermions lead more interesting and puz- 
zling physics due to a stronger interaction effect and 
are not well understood yetP"^ Not only graphene but 
also some spin liquids and topological insulator also have 
QBT fermions structures EES! 

In (2+1) dimension, there is another common and 
interesting excitation, Skyrmion of 0(3) vector order 
parameters ! 11 * 12 ! For example, the Neel order phase 
has Skyrmions because its ground state manifold is 
a two sphere due to its broken symmetry structure, 
SU(2)/U(1) — S 2 . Basically, a Skyrmion number is a 
wrapping number of a given configuration on a two di- 
mensional space with identified boundaries. 

One natural question with the two excitations, QBT 
fermion and Skyrmion, is how they interact with each 
other. In this paper we answer the question focusing on 
Skyrmions' induced quantum numbers by QBT fermions. 
Extensiv e rese arch already exists on Skymions with Dirac 
particlesPEsl 

Our goal is two-fold. First, we show Skyrmions receive 
quantum numbers by interacting with QBT fermions, 
which are doubled compared to Dirac particles'. In ad- 
dition, we apply the mechanism of the induced quan- 
tum numbers of Skyrmions to the Bernal stacking bi- 
layer graphene structure and study an exotic quantum 
phase transition in the bi-layer graphene. We report two 
findings. First, quantum spin hall phase in the bi-layer 
graphene is connected to charge 4e SC by QSH Skyrmions 
condensation. And second, the phase transiiton could 
be a second order and we propose one candidate the- 
ory, 0(5) non linear sigma model with the Wess-Zumino- 
Witten term. 

In terms of charge 4e SC, one can see that it is almost 
impossible to realize it with the conventional BCS theory 
because two pairs of charge 2e order parameters have a 



better chance of forming than charge 4e SC. Therefore, 
one key factor to achieve charge 4e SC is how to prohibit 
charge 2e order parameters. One possibility of charge 4e 
superconductivity was suggested by Berg, et, aZ.P^Main 
idea of their study is to prohibit 2e pairings thermally 
to make 4e superconductors, so they only find non-zero 
temperature charge 4e SC. Our research is a significant 
departure from the previous one by Berg in that our 4e 
SC is a ground state at zero temperature. 

Abanov and Wiegmann studied a non-BCS way to 
achieve superconductivity focusing on acquired quantum 
numbers of topological objects interacting with Dirac 
fermionsPS If a quantum number-endowed topological 
object is a boson, it is possible to condense the object, 
and it induces quantum phase transition as a textbook 
example in the two dimensional XY model. Applying the 
idea to monolayer graphene, Grover and Senthil argued 
that quantum spin hall phase in monolayer graphene can 
be connected to charge 2e superconductor through the 
CP{\) deconfined quantum criticalityPil Below we show 
how QBT fermions change Skyrmions' properties and as- 
socited Quantum phase transition. 

Model Hamiltonian : Before going into QBT fe rmoins, 
let us briefly review results of Dirac particles ! 14 * 15 * 21 ! 
Hamiltonian consists of Dirac fermions with 0(3) vector 
order parameter. 

H D irac = ^ [Ko x + k y a v + ga z n ■ f] ip. (1) 

, where a 1 , t j are different kinds of Pauli matrix, so it is 
necessary to introduce at least four component spinors. It 
is well known that the 0(3) vector fields allow non-trivial 
topological configuration, Skyrmion, due to ^(S 2 ) = Z. 
Its topological current is = ^e^xri ■ [d v n x d\n) 
and charge is spatial integration of its time component, 
gSkyr = j d 2 x±n ■ (d x n x d y n). 

It was shown that the induced fermionic current, < 
jf erm > ; i s proportional to Skyrmion current and its 
charge is exactly determined by Skyrmions number upto 
number of flavors. 

Q/erm = Nf QSkyr ^ 



Also, it was well known that Skyrmions' statitical prop- 
erties could be read off by examining two Skyrmions 
exchange property or measureing a phase of 2tt rota- 
tion. Those information is contained in the so-called Hopf 
term, which corresponds to ^(S* 2 ) = Z mathematically 



cHofp _ 

Dirac 



iNfir • sgn( 5 ) / 



, where Nf is the number of flavors and the gauge 
field(v4 M ) is defined by the relation, J M = s^ u \dvA\. 

Massage of the above discussion is manifest. 
Skyrmions of order paprameters receives charge and 
statistics by Dirac fermions. Dirac particles with even 
number flavors make Skyrmion bosonic charged parti- 
cles and odd number's make it fermionic charged par- 
icles. Notice that there is one dimensionful coupling 
constant (g) but it does not appear in our final results 
except its sign. It is a signal of topological origin of in- 
duced quantum numbers of Skyrmions. 

Now, let us consider quadratic band touching fermions 
interacting with 0(3) vector. In general, Hamiltonian is 



H = ^ 



(k\ - k 2 y )a x + (2k x k y )gy 
2m 



+ ga n ■ f 



This is a generalization of the Eqn. [T] from the p wave 
helicity to the d wave helicity with introducing one more 
dimensionful scale, effective mass(m). Obviously, the 
Lorentz symmetry is broken by quadratic dispersion re- 
lation. Such systems can be realized, for example, in 
bi-layer graphene or in the hole-doped semiconductors 
with the inversion symmetry. 

One way to see induced charge of Skyrmion is to eval- 
uate fermion's energy spectrums under Skyrimon back- 
ground. Turning on Skyrmion configuration adiabati- 
cally, one can show there are zero energy level crossing 
states. For one spatial dimension with 0{2) symmetry, 
Yao and Lee showed there were two level crossing for the 
QBTP In principle, one can do similar calculation in 
two spatial dimension. Instead, we use a field theoritic 
method, the graident expansion, to see both Skyrmion's 
charge and statistics also. Similar calculation method has 
been applied to Dirac particles in various literatures.^ 

For QBT fermions, broken Lorentz symmetry makes 
the expansion little more tedious but it is straightforward 
to do if one only keeps the lowest term. Partition function 
is 



Tr(e" 



- J d 2 xdrC 



, where the Lagrangian is 

C = ^Dxjj = ip^d T ip - 



H. 



(3) 



One key step of the gradient expansion is to consider 
variation of the effective action. 



SS, 



eff 



-triD^SD) = -tr(SDD\DD^y 



(4) 





(b) 



FIG. 1. Induced current(a) and Hopf term(b) diagram. The 
plain line is for the propagator. (Go) The square vertex is for 
the variation of the D operator. (SDD^) The circle vertex is 
the coupling between the fermion and the order parameter 
variations. (M) 



Following commutation relations are useful. 

[kl - k 2 y ,h] = 2{-i){k X d X h - kyOyh) + {dl - d 2 y )h 

[k x k y + k y k x , n\ — 2(—i)(k x d y h + k y d x n) + 2d x d y h 
If we keep only the lowest order terms, we get 

k 2 



DD^ = Gq 1 



M 



lo 2 + ( — ) 2 + g 2 + M 
Zm 



M = g 



a z d T n 



k v (j x — k x a l 



-dyfl 



With these expansion, one can do perturbative calcula- 
tion with Green's function, Go with perturbation, M. In 
Fig. [Tj we illustrate two diagrams for induced current (a) 
and Hopf term(b) calculations. Note that to calculate 
induced currents it is necessary to couple external gauge 



field as in an usual way, {k^ 
the first diagram, we find 



— > fc M — a M ). By evaluating 



2N f Q skr . (5) 

Here, we put Nf back to the formular for notational pur- 
pose. 

Few remarks are in order. First, there is a factor two 
compared with the Dirac fermion's of Eqn. [2j The in- 
duced charge of the Skyrmion by QBT fermions is dou- 
bled compared to the Dirac case. Second, the result in- 
dicates that Lorentz symmetry is not a crucial factor for 
topologically induced currents. In other words, the vac- 
uum of the non relativistic fermion with the topologi- 
cal background could be polarized and its polarization is 
twice bigger. Also, one notice that there is no effective 
mass dependence in the current calculations. This fact 
is another signature of its topological origin. Otherwise, 
one expect that the effective mass term would appear in 
momentum integration. 

To complete induced quantum numbers, let us study 
statistics of topological defects. It was shown by Wilczek 
and Zee that the Hopf term of the non-linear sigma 
model, so-called the "theta" term, describes the statistics 



3 



of the Skyrmion. 



16 



gHopf 



(6) 



, where 9 term is determined by microscopic models. If 
we consider the rotation(or exchange) of the soliton, then 
its current naturally induces the non-trivial imaginary 
factor depending on its spin(J). 



S^S + ie (S^S + i2irJ) 



(7) 



So the Skyrmion spin is J = and tt(2tt) corresponds 
to fermion(boson). Other values indicate anyonic statis- 
tics of the Skyrmion. 

Following the same steps as in the current calculaiton, 
it is straighforward to obatin $S q °*J using the second 
diagram of Fig. [ljb). We find that imaginary part of ef- 
fective action's variation corresponding to the Hopf term 
is 



5S. 



Hopf 



2i 



32tt 



^hrlSnnd^hdvfidxn} (8) 



, where h = f • ft is used. 

It is obvious that the variation becomes zero, which 
naively looks no Hopf-term contribution. However, it 
is well known that the Hopf term caluculaiton is more 
subtle due to its non-trivail homotopy group, ^(S 2 ) = 
Z. In other words, the zero variation is an artifact of our 
perturbative method. To ove rcome this difficulty, we use 
the embedding method? 15 * 23 ! Main idea is to extend the 
order parameter group, CP 1 
avoid the ambiguity. 



S 2 , to enlarged CP M to 



n ■ t — > 



2zz f - 1 



(9) 



with z t — (zi, • • • , zm+i) and z^ z = 1. Due to the fact, 
n^(CP M ) = 0, for M > 1 it is safe to use a perturbation 
theory. Variation of the effective action has no ambiguity 
and our perturbative calculation is well-defined. Then, 
we take the limit M — > 1 and read off the induced Hopf 
term. And the induced Hopf term is 



(10) 



We put Nf back to the final forms for notational pur- 
poses. As we can see, the factor two again appears in 
the Hopf term calculation, which is consistent with the 
charge current calculation. Therefore, Skyrmions with 
the QBT fermions are always bosons and their charge 
is determined by number of flavors(2iV/). Again, the 
dimensionful effective mass does not appear due to its 
topological origin. 

How do we understand Skyrmion charge doubling ef- 
fect in QBT fermions? One way to understand is to add 
small perturabation which do not break other symme- 
tries such as gauge potential disorder with strengh(q). 
Then, it is easy to see the QBT point at the minium 
splits into two Dirac points below the energy sclae cor- 



responding (q). If only low energy properties are dom- 
inant to determine topological charges, we can just add 
two Dirac particles to make QBT fermions, which ex- 
plains the doubling effect. It seems the argument that 
QBT is equivalent to two Dirac particles is compelling, 
but they show different fermionic mass structures. Espe- 
cially, competinig order parameter descriptions based on 
the mass term structure are different! 3 -^ 

Physical Realization : One of the well-known examples 
of the QBT fermion is a bi-layer graphene with Bernal 
stacking. Also, hole-doped GaAs with the inversion sym- 
metry along the perpendicular axis has similar structure. 
In this paper, let us focus on the bi-layer graphene. Its 
tight binding Hamiltonian is 



H = -t i c \,i c hi + 4,j c 2,i) + * ( c m c 2,* + h.c.) 

(11) 



<i,j> 



iGA-B 



, where c\^) % ^ s a creation operator of the first (second) 
layer at the site(i), and the second summation is only for 
one sublattice where sites on one layer is on top of sites 
on the other layer, {A — B). 

Among four band dispersions, only two bands become 
important and the low energy Hamiltonian becomes 



{2k x k y )o-y p z 



2m 



V; (12) 



, where a, t, p correspond to sub-lattice(or layer), spin, 
and valley Pauli matrix. This Hamiltonian has the helical 
quadratic dispersion structure as our model Hamiltonian. 
Only considering non-valley mixing states, there are two 
order parameters with 0(3) vector properties. 



Oqsh = V'Vfp 2 ?/' 



O 



N 



— ,/it 



ipo'ftl) (13) 



The first one corresponds to the Quantum Spin 
Hall(QSH) order parameter because it does not break 
the time reversal symmetry even though the spin rota- 
tional symmetry is broken. The second one breaks both 
the spin rotational symmetry and time-reversal symme- 
try, so it's a Neel order parameter. To make connnection 
to our model Hamilnotian manifest, let us rotate the ba- 
sis as follows. 



i 



1 - a x 



P z )i> 



(14) 



Then the p dependence of the non-interacting Hamilto- 
nian i?G,o dropped. Of course, the order parameter terms 
also rotate under the transformation and become 



O 



QSH 



l/j^ (J Z Tip , N ->^<T Z T P Z ll> (15) 



Therefore, the QSH order parameter exactly corresponds 
to our model Hamiltonain except doubling of the number 
of flavors (valley index). 

Now we can apply our model Hamiltonian results di- 
rectly. Skyrmion of the QSH order parameter is a boson 



4 



with charge four considering the valley index. On the 
other hand, if we consider the Neel order parameter, it 
is easy to see the fermionic charge is zero and the Hopf 
term becomes trivial because the signs of the coupling 
constant of different valley fermions are opposite and 
their contributions are cancelled out. It indicates that 
the Skyrmion of the Neel order parameter is a charge 
neutral boson. Therefore, it is manifest that Skyrmion's 
induced quantum number depends on order parameters' 
property, especially how an order parameter couples to 
Dirac particles with different hclicity. 

Let us focus on the QSH phase. By tuning interac- 
tions, one can destroy the QSH order parameter. Similar 
to vortex proliferation in XY model in (1+1) dimension, 
it is possible to eliminate the QSH phase by condensing 
the Skyrmion due to its bosonic properties. This phase 
cannot be the same as the semimetal phase because the 
Skyrmion's electric charge is four. Therefore one can re- 
alize the charge 4e superconductors by condensing topo- 
logical defects of the QSH order parameter. This scenario 
does not require any Fermi surfaces and the Cooper pair 
formation, so there is no way to form charge 2e order 
parameters here. In other words, we can circumvent the 
instability problem of the charge 4e SC into formation of 
two charge 2e Cooper pairs. Therefore, it is one concrete 
way to achieve charge 4e SC through the quantum phase 
transition from QSH. 

What kind of a quantum phase transition is the tran- 
sition from QSH to charge 4e SC? Because two phases 
break different symmetries, it should be a first order 
phase transition under the Landau-Ginzburg paradigm. 
But, a second order phase transition is also possible via a 
deconfined quantum criticality! 20 * 21 ! In quantum mechan- 
ics, topological defects of one order parameter can carry 
the other order parameter's quantum numberPSl Their 
condensation suppresses the original ordered phase and 
induces the different symmetry-broken phase simultane- 
ously. Thus, a continuous phase transition can naturally 
appears in this process. 

One candidate critical theory of the phase tran- 
sition is 0(5) model with the Wess-Zumino-Wittcn 
term(WZW) ! 24 l 27 l Introducing a multiplet, <f> — 
(ft, Re ip ie , Im^ 4e ), the model is 

S = [ drd 2 x ^(d$) 2 + 2mkT wzw ' ($) (16) 

J %g 

T wzw ($) = 

, where k is an integer number level. 

Without the WZW term, the beta function of g is well 
understood. There are two fixed points; g = corre- 
sponds to the ordered phase and g = g c describes a criti- 
cal fixed point. Beyond the critical fixed point, the beta 
function monotonically increases. One can expect the 
WZW term plays non-trivial effect due to its topological 
orgin. Indeed Witten showed the WZW term in (1 + 1) 



dimension induces stable fixed points beyond an unstable 
fixed point PU 

Here, we follow his method and do the beta function 
calculation perturbatively. We find that 

^=/3 (5 2 )"O fcV° + O(g 12 ,fc 4 ). (17) 

The /3q function is one without the WZW term, which is 
obtained from the literature!^ 

/3o(5 2 ) = -.9 2 +3.g 4 + 3g 6 + |5 8 + 365 10 

, where b = ^ + 3£(3) and dimensionless coupling con- 
stant g is introduced to match higher loop calculation. 

We note that our calculation is not controlled basi- 
cally because of existence of the original mass scale of 
the coupling constant. In contrast, 0(4) NLSM on 1 + 1 
dimension calculation is controlled by 1/k expansion, and 
one can show that there are stable fixed points with non- 
zero k values, the so-called SU(2) k WZW fixed point.^ 
One physic al realization of such fixed points is spin 1/2 
chain! 28 * 29 ! Even though our calculation is not controlled, 
it clearly shows that the topological term induces a non- 
trivial effect. The topological term's contribution is of 
the opposite sign to the other terms in the beta func- 
tion. Physically, this means that the topological term 
suppresses the tendency of becoming a disordered state. 

There are three possible scenarios based on our per- 
turbative calculation. First, the topological term con- 
tribution is not big enough, so it does not induce any 
stable fixed point. With a small enough value of k, the 
beta function is barely affected by the topological term. 
Phase transition is described by the 0(5) NLSM model 
without the WZW term and the topological contribution 
only gives a small correction. The second scenario is that 
the topological term is so big that the beta function be- 
comes always negative. An implication of this case is 
that the system is always in the ordered phase. Both 
cases can be described by the 0(5) NLSM without the 
WZW term, so there is no new physics. The final case 
is that the topological term and the higher order fluc- 
tuation of the order parameter have similar amounts of 
effects. Thus, one new stable fixed point appears beyond 
the usual critical fixed point. 

This new fixed point is our candidate for the deconfined 
quantum criticality between QSH and charge 4e SC. By 
introducing symmetry breaking term such as y(ft ■ ft — 
IV'ScI 2 ), each phase can be accessed from the stable fixed 
point. Numerically, we find Co = J+ and k c = 2 with a 
hard-cutoff regularization schemePS If one believes that 
the SU (2)k WZW fixed points in 1 + 1 dimension have its 
extension to ones in the 2 + 1 dimension, our new fixed 
point would be a natural candidate. Figure [2] illustrates 
both the three possible cases in terms of renormalization 
group flow and the phase diagram at the critical level. 

Finally, QBT fermions are different from two copies of 
Dirac fermions in terms of fermion mass term structures, 
which usually explain competing order parameters. As 



5 



(a) 



fj 0.0 




0.0 0.1 0.2 0.3 0.4 0.5 



0(2) ordered 
(b) P 1135 " 9 y 

Critical Point 

o < c ) i < a 

OC5) ordered I Interacting stable 

phase I fixed point 

OC3) ordered 

FIG. 2. (Color Online) (a) Beta function behaviors of 0(5) 
NLSM model with different levels WZW level, k. The dot- 
ted(black), normal(red), and dashed(blue) lines correspond to 
fc c cases. At k — k c , there is one stable 
interacting fixed point(balck dot), (b) Phase diagram with 
RG flows for the critical level at k = k c . 

shown by HerbuP^ recently, there is a "hidden" internal 
structure(pseudo-spin) of competing order parameters of 
Dirac particles. For mono-layer graphene with Dirac par- 
ticles, SC order parameter can make competing 0(5) vec- 
tor with either Quantum spin Hall or density wave order 



parameters. In other words, there are two possible ways 
to make competing 0(5) order parameters. However, one 
can easily check that QBT fermions do not allow the same 
two possibilities. Only one choice is allowed. For the SC 
order parameter, only the QSH order parameter is pos- 
sible to make 0(5) vector. In that sense, QBT fermions 
describe a more restricted competing orders' structure, 
which originates from the difference of Clifford algebra 
representations of Dirac particles and QBT fermions.^ 
We leave this interesting problem for the future work. 

Conclusion: In this paper, we showed how Skyrmions 
interacting with QBT fermions obtains their quantum 
numbers. It is shown that QBT fermions double induced 
charges compared to Dirac fermions' and make all the 
Skyrmions bosonic. Applying the mechanism to bi-layer 
graphene, we also showed that QSH Skyrmions in bi- 
layer graphene are charge 4e bosons. By condensing the 
Skyrmions, it is possible to achieve charge 4e supercon- 
ductivity. Such mechanism is far different from the gen- 
eral BCS-like description, and we argue that it might 
be described by the deconfined quantum criticality. We 
propose the 0(5) non-linear sigma model with the WZW 
term at level k c as a critical theory. 

Acknowledgement : The author thanks to Igor Herbut, 
Laing Fu, and Cenke Xu for invaluable dicussion. The 
author is especially grateful to Cenke Xu for introducing 
the subject. 



1 K. S. Novoselov, et al., Science 306, 666 (2004). 

2 A. H. C. Neto, et al, Rev. Mod. Phys. 81, 109 (2009) 

3 Y. Zhang, Y. Tan, H. Stormer, P. Kim, Nature 438, 201 
(2005) 

4 Wenzhong Bao, Jairo Velasco Jr, Fan Zhang, Lei Jing, 
Brian Standley, Dmitry Smirnov, Marc Bockrath, Allan 
MacDonald, and Chun Ning Lau, arXiv: 1202.3212. 

5 Y. Zhao, P. Cadden-Zimansky, Z. Jiang, and P.Kim, Phys. 
Rev. Lett. 104, 066801 (2010). 

6 Javier D. Sanchez- Yamagishi, Thiti Taychatanapat, Kenji 
Watanabe, Takashi Taniguchi, Amir Yacoby, and Pablo 
Jarillo-Herrero Phys. Rev. Lett. 108, 076601 (2012) 

7 L. Fu, Phys. Rev. Lett. 106, 106802 (2011) 

8 K. Sun, H. Yao,E. Fradkin, and S. Kivelson, Phys. Rev. 
Lett. 103, 046811 (2009). 

9 C. Xu, et al. JarXiv: 1110.3328| 

E. Moon and C. Xu, in preparation 

Jae-yoon Choi, Woo Jin Kwon, and Yong-il Shin, Phys. 
Rev. Lett. 108, 035301 (2012). 

2 Su Do Yi, Shigeki Onoda, Naoto Nagaosa, and Jung Hoon 
Han, Phys. Rev. B. 80, 054416 (2009). 

3 E. Berg, E. Fradkin, and S. Kivelson, Nature Physics 5, 
830 (2009). 

4 A.G. Abanov and P.B. Wiegmann, Phys. Rev. Lett. 86, 
1319 (2001). 

5 A.G. Abanov and P.B. Wiegmann, Nucl. Phys. B 570, 685 



(1994), and references therein. 

16 F. Wilczek and A. Zee, Phys. Rev. Lett. 51, 2250(1983). 

17 Y. Wu and A. Zee, Phys. Lett. 147B, 325 (1984). 

18 J. Goldstone and F. Wilczek, Phy. Rev. Lett. 47, 
986(1981). 

19 H. Yao and D. H. Lee, Phys. Rev. B 82, 245117 (2010) . 

20 T. Senthil, A. Vishwanath, L. Balents, S. Sachdev, and 
M. P. A. Fisher, Science 303,1490 (2004) 

21 T. Grover and T. Senthil, Phys. Rev. Lett 100, 156804 
(2008) 

22 S. Raghu, X. Qi, C. Honerkamp, and S. Zhang, Phys. Rev. 
Lett 100, 156401 (2008) 

23 E. Witten, Phys. Lett. B 117, 324 (1982) 

24 T. Senthil, and M. Fisher, Phys. Rev. B 74, 064405 (2006) 

25 E. Witten, Comm. Math. Phys. 92, 455 (1984) 

26 W. Bernreuther and F. Wegner, Phys. Rev. Lett., 57 1383 
(1986) 

27 P. Hosur, S. Ryu, and A. Vishwanath, Phys. Rev. B 81 
045120 (2010) 

28 D. Haldane, Phy. Lett. 93, 464 (1982). 

29 I. Affleck, Nucl. Phys. B 265, 409 (1986). 

30 We define k c as the smallest value of k to have the stable 
non-zero fixed point. It is difficult to obtain the exact value 
of fee which might depend on regularization schemes due 
to its uncontrolled property. 

31 I. Herbut, Phys. Rev. B 85, 085304 (2012)