Kondo, Kondo, everywhere, but what is going on in the bulk:

Kondo insulators and strong correlations

Ryan Levy

May 2018

Abstract
A class of materials has been found where quasiparticle excitations have an enormous effec-
tive mass, known as heavy fermion materials. Examples of such materials are Kondo insulators,
in which this behavior is driven by spin-orbit coupling and the presence of localized f orbitals.
Coupling between free electrons and localized spins produces interesting physics, namely a large
effective mass, through screening and hybridization. In this paper, we discuss a toy model real-
izing the physics of these materials, the Kondo Lattice Model, and its connection to experiment.
Finally, we investigate some of the properties of SmB6 , a potential topological Kondo insulator.

1
Part I
Introduction and Effective Mass
In the world of strongly correlated problems, where materials’ electron-electron interactions are
strong and mean field theory can often fail, there exists an emergent behavior where fermions in the
material appear to have a mass which is orders of magnitude greater than the normal electron. Such
heavy fermion materials generally partially consist of rare earth materials, which have f orbitals
and strong spin-orbit coupling. Under a variety of conditions, the f orbitals can localize to the ions,
causing a competition between insulating, semiconductor, and even metallic behavior.
While some of this competition is well understood, there remains a rich set of physics to be
explored. One heavy fermion material discovered in the 1960s [1], SmB6 , has recently created a
flurry of new progress. The material is now believed to be more complex than simply a Kondo
insulator but one with topological properties. Materials with such properties, known as topological
insulators, generally have surface states that are protected from disorder and surface reconstruction
effects . While SmB6 may have resilient surface states, there are confusing, and at times contradic-
tory, set of experimental observations that produce an exotic story for this insulator.

2 How are Fermions Heavy?

Perhaps the most frequent phrase when dealing with “Kondo physics” is the phrase heavy fermions.
The term heavy fermions relates to the effective mass of quasiparticles measured in experiments.
2
The easiest way to think about how one can observe these 10
UBe13 CeAl3
quasiparticles is to look at Fermi Liquid Theory (FLT). In FLT, ~2
Ã/γ = 1 × 10
-5 YbRh2Si2 (0T)
CeCu6

1 CeCu2Si2
problems with strong electronic Coulomb interactions can be 10
µΩcm(K mol/mJ)
2

connected with a description of free or weakly interacting quasi- Ce

UPt3

CeB6 YbRh2Si2 (6T)

particles rather than bare electrons. These quasiparticles have 10
0
Sm CeRu2Si2
Eu
dispersion relations of (k) ∼ ~2 k 2 /2m∗ − µ of a free particle
USn3 YbNi2B2C
Yb UAl2

U UPt2
and a mass given by 10
-1 SmOs4Sb12
à ( µΩ cm / K )

UPt
2

UIn3
SmFe4P12
  UGa3
1 1 ∂(k) 10
-2 YbCu5
= (1) YbCuAl

m∗ pF ∂k k=kF
Yb2Co3Ga9 YbCu4.5Ag0.5

CePd3 CeNi Eu(Pt0.8Ni0.2)2Si2

10
-3 CeNi9Si4
YbCu4Ag
While this quantity is unable to be directly observed experi- YbInAu2
Eu(Pt0.75Ni0.25)2Si2

mentally, it may be extracted from the specific heat 10

-4
YbNi2Ge2

CeSn3

CV
∝ m∗
YbInCu4
γ = lim (2) 10
-5
YbAl2
YbAl3
T →0 T

can be measured with good accuracy. Usual metals, e.g. pure 10

-6

copper, will have γ ∼ 1 − 10 mJ/K 2 mol while heavy fermion

-1 0 1 2 3 4
10 10 10 10 10 10
~ 2
γ ( mJ / mol K )
compounds have observed this to be 100 − 1600 mJ/K 2 mol[3].
Thus these materials appear to have a effective mass, and thus
Figure 1: Kadowaki-Woods ratio for many
density of states, several orders of magnitude higher than ex- Heavy Fermion materials, normalized for
pected. The large effective mass of heavy fermions is in fact ground state degeneracy [2]
an indication for interesting low temperature physics charac-
terized by universal properties lying beyond the FLT paradigm.

2
 One universal property can be calculated within FLT by considering the resistivity. The resistiv-
ity can be calculated as ρ ∼ ρ0 + AT 2 where A is some material dependent constant experimentally
measured, due to forward scattering. For materials that have this T 2 law and T specific heat, we
can define a ratio of A/γ 2 , known as the Kadowaki-Woods ratio. As A ∝ m∗2 and γ ∝ m∗ the
Kadowaki-Woods ratio should be constant [3]. In Fig. 1 this universality is shown for a multitude
of heavy fermion materials.
The remainder of the paper will give the current experimental and theoretical understanding of
the materials with large γ or obtain a heavy effective mass not described by FLT. To do so, two
models are introduced to explain the wide range of heavy fermion behaviors, differentiating between
a single magnetic impurity and a dense lattice of impurities.

Part II
A Single Impurity- the Kondo Problem
3 Experimental Evidence

26.9
Au1
Au2

26.8

26.7
10 4 R/R0 ◦ C

26.6

26.5

26.4
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
T(K)
(a) Resistance data recreated from [4]. The gold (b) Resistance scaling as a function of temperature for MoNb
samples contained Cu and Ag impurities alloys containing 1% of Fe [5]

Figure 2: Experiments showing the rise in resistivity

On the beginning of the path toward the understanding of the materials that were studied in Ref
[2], there were a series of experiments that found interesting behaviors in the low temperature
resistivity. One of the first, illustrated in Fig. 2a, was a 1934 experiment that found an unexpected
rise in resistivity from Gold with a small amount magnetic impurities. Another experiment, shown
in Fig. 2b varied the doping of MoNb with a 1% Fe impurity concentration showing a deviation
from the ρ0 + AT 2 with the addition of − ln T behavior, at a transition temperature Tk (the Kondo

3
temperature). The cross over at the temperature Tk is known as the Kondo effect. Connecting the
high temperature FLT description and the Kondo temperature transition was known as the Kondo
problem.

4 The Kondo Model

To describe the microscopic behavior of these (now more understood) heavy fermion materials, we
begin with the Anderson model, which describes a localized magnetic ion1 interacting with a sea of
conduction electrons of dispersion k

Conduction Conduction-Impurity Impurity

Vk c†kσ fσ + h.c.
X
HAnderson = k nkσ + + Ef nf + U nf ↑ nf ↓ (3)
k,σ

where c†kσ is a conduction electron creation operator and fσ is an f orbital annihilation operator.
Because of the time reversal symmetry and overall spin 1/2 of the impurity, there will be a Kramer’s
doublet or a double degenerate energy state in the atomic limit of the orbital.
To extract the physics of the Kondo effect, we will look at what happens when we remove
the unoccupied and double occupied states of the impurity. With that in mind, we can look at
the virtual processes of conduction electrons interacting with the impurity. There are two import
processes here, which involve two sets of hops (hopping onto then from the impurity) which changes
the spin (↑↔↓) of the conduction electron and the impurity spin. This spin flip interaction can
be modeledPas a spin-spin interaction Hamiltonian, where the conduction electrons spin operator is
written as k,k0 c†kα~σαβ ck0 β such that the new Hamiltonian is

Conduction Conduction-Impurity
 
Jkk0 c†kα~σαβ ck0 β · S
X X
HKondo = k nkσ − ~f (4)
k,σ k,k0

with Sf being the spin of the impurity.

A surprising feature of the spin-spin interaction is the sign of J. Using second order perturbation
theory of the conduction-impurity process from the Anderson Hamiltonian we will obtain an effective
J as Jef f = −4|Vk |2 /U which is inherently negative with U > 0 . This extra negative sign makes the
interaction antiferromagnetic rather than ferromagnetic, a surprise at the time as it was expected
to be a ferromagnetic interaction.
Out of this model we find two interesting results. The first is the asymptotic freedom of the
impurity. At high temperatures the local magnetic moments are free but upon cooling create a
strong coupling to the conduction electrons; this is known as asymptotic freedom, the same physics
governing quarks in QCD [3]. Because this is related to scaling and not emergent physics, we will
gloss over the scaling details and focus on what happens at the lower temperatures.
Hinted by the Kadowaki-Woods ratio, many of the observables can be written in a universal
form [6]. If we let F be some function then both the spin susceptibility, inverse scattering rate, and
1
which we will take, without loss of generality, to be an f orbital in labeling

4
specific heat (at strong coupling) can be written as a universal function T /TK ,
1
χ(T ) = Fχ (T/TK )
T
1 1
= Fτ (T/TK )
τ (T ) τ0
CV ∝ /TK
T

The second result is finding a Kondo resonance, also known as the Abrikosov-Suhl Resonance [3].
Here a singlet between the conduction electrons and magnetic impurity forms at the fermi surface
and creates a peak in the spectral function of width proportional to the Kondo temperature Tk . As
the temperature is lowered, the impurity is screened by the conduction electrons slowly becoming
non-magnetic. Screening of the local magnetic moment is known as the Kondo effect. This effect
provides evidence of a universal behavior of at low temperatures of the resistance minimum due to
magnetic impurities.

Part III
Many Impurities - the Kondo Lattice
5 Experimental Evidence
A key part of the material in Fig. 2b was that
it had a small amount of magnetic impurities
(Fe). If instead there were many magnetic mo-
ments in the lattice, as is the case with SmB6
and other materials, then we might expect new
physics to come out. Experimentally this is seen
with a high temperature Curie-Weiss susceptibil-
ity of χ ∼ (T + θ)−1 from magnetic impurities.
We also see that SmB6 is a strong insulator
when its resistance grows at low temperature as
shown in Fig. 3, unlike the change in resistivity
seen earlier in Part II. There were also a series of
materials like SmB6 , e.g. CeFe4 P12 or YbB12 ,
that were metallic at high temperatures but insu-
lating at low temperatures known as Kondo Insu-
lators [7]. Some of the physics of Kondo Insulators
Figure 3: SmB6 resistance measurement [1, 7]
is now understood, but there are still many open
questions, particularly with SmB6 to be explored
in Part IV.

6 The Kondo Lattice Model

Some of these Kondo insulator experiments lead Doniach [9] to introduce a lattice version of the
Kondo model eq. (4) with many magnetic spins (with the double occupied and unoccupied degrees

5
Figure 4: Differences in typical experiments between materials described by the Kondo model (left) and the Kondo
lattice model (right) [8]. The specific heat includes the − ln T component away from Fermi liquid behavior. Note
that we focus on Kondo Insulators which have a different low temperature resistivity as seen in Fig 3.

removed) interacting with conduction electrons

X † 
k c†kσ ckσ − J
X
HKLattice = ~j
cjβ ~σβα cjα · S (5)
kσ j,αβ

where usually J = −|J| an antiferromagnetic coupling, and S ~j is a local spin 1/2 moment [7]. The
motivation of this model is to describe a dense lattice of magnetic impurities that match the high
temperature behavior of Fig. 4 and explain the many low temperature phenomena observed, such
as a metal-insulator transition. Doniach argued that one still found the Kondo temperature, as seen
experimentally, but there was another fundamental energy (temperature) scale.
Consider the single impurity Kondo model, but with one extra impurity added, such that we
have S~f and S~f . One could write out second order perturbation theory in which we consider the
1 2
two Hamiltonian terms, one from the conduction-spin interaction from impurity 1 and another from
impurity 2. Then upon integrating out the conduction electrons the remaining term goes like

~f · S
~f cos 2kF |r|
HRKKY ∼ JRKKY (r)S 1 2 with JRKKY (r) ∼ J 2 ρ (6)
|r|3
with ρ the conduction electron density of states per spin and r is distance from the impurity. This
is called the RKKY interaction after Ruderman, Kittel, Kasuya and Yosida [6]. While that is a
straightforward way to derive the effective Hamiltonian, there is an alternative picture of description.
Because each local moment acts like an effective magnetic field to the conduction electrons; thus
the electrons will want to anti-align to the impurity spins (recall the antiferromagnetic coupling).
In-between the two impurities will be governed by some periodic function related to the how the
spins are pointed, to maximize the antiferromagnetic interaction but weakened the further from the
impurity the electron is.

6
 Doniach noticed this new effective interaction energy scale, and argued that it scaled like
TRKKY ∼ J 2 ρ. He concluded that TK  TRKKY the spin-spin interaction will create antiferro-
magnetic ordering. On the contrary, when TK  TRKKY then a scattering resonance will stabilize
the singlet ground state considered from the Kondo model, along with heavy fermions. Unlike the
Kondo model, which created a rise in resistivity, the coherent resonant scattering of the lattice model
produces a drop in resistance as shown on the right in Fig 4. When they are on the same order,
something less straightforward must happen. Moreover, for low temperatures, Kondo insulators
have a disappearance of the RKKY interaction [3].
Previously we saw that a single impurity underwent screening into a non-magnetic ground state
and produced a large peak at the fermi surface, yet the Kondo lattice model can produce an insulator
with a charge and spin gap. Following the arguments in Ref. [7], consider the t/J → 0 limit of the
Kondo lattice model. This limit reduces to just spin/electron interactions (the J term), forming a
singlet ground state
Y 1
|Ψ0 i = √ (|⇑j ↓j i − |⇓j ↑j i) (7)
j
2

where ⇑ / ⇓ is the localized spin and ↑ / ↓ are the conduction electrons. Consider the excitations
of this ground state. To switch a singlet to a triplet we’d gain an energy of 2|J|, which means this
model is spin gapped. We can then add or remove an electron to site i, which we will take to be ↑i
√ † Y 1
2ci↑ |Ψ0 i =⇑i (↑i ↓i ) √ (|⇑j ↓j i − |⇓j ↑j i)
i6=j
2
√ Y 1
2ci↑ |Ψ0 i =⇑i √ (|⇑j ↓j i − |⇓j ↑j i)
i6=j
2

For both wave functions we’ve created have extra energy of 3|J|/2 and thus separated by 3|J|.
Therefore the Kondo lattice model at strong coupling gives a ground state that’s a Kondo insulator
with spin gap 2|J| and charge gap 3|J|.
Lastly, there is a hybridization picture of obtaining a gap. Using mean field theory on the Kondo
lattice model, a flat band of localized f electrons can hybridize with the free conduction electrons.
The hybridization opens a gap, which if the chemical potential lies within the gap the material is
insulating. The conduction band is then heavy fermion holes, with a positive charge unlike the high
temperature phase. Mean field theory however is not enough to explain many of the properties of
the Kondo insulator SmB6 , for example.
Additionally, heavy fermion superconductivity can emerge from magnetic impurity interactions
[3, 7, 10]. Like a Cooper pair, the localized spin and high energy electrons form a “composite”
fermion (rather than boson), leading to new phenomena. Those effects won’t be covered here but
is interesting modern emergent behavior that is being studied.

7
Figure 5: The difference between the Kondo model and the Kondo Lattice model. (a) depicts a single impurity
interacting with conduction electrons which opens up the Kondo resonance while (b) shows a dense lattice of impurities
interacting with conduction electrons which can create a hybridization gap of the Kondo insulator. Adapted from [6].

Part IV
More open questions - SmB6
7 What Makes SmB6 Different?
While SmB6 was a prototypical example of a Kondo insu-
lator, even now there are many puzzling questions about
its properties. The first is the resistance saturation some-
where below 4K. At high temperatures the material be-
haves as a metal, and at low temperatures it becomes
insulating as a hallmark Kondo insulator but the − ln T
behavior does not extend to T → 0. This is, however, in
line with what one would expect from a topological Kondo
insulator (TKI).
Different groups set out to measure the surface con-
ductivity compared with the bulk. One group, Wolgast
Figure 6: Crystal Structure of SmB6 [11] et al. [12], constructed an elaborate way to verify bulk
vs surface conduction. In their paper2 , it was found that
below 4K there was indeed a change from bulk to surface
conductance, with many signs pointing to being a TKI. However the experiment was sensitive to
sample preparation (etching the surface for example) and could not definitely conclude another
source of surface conduction.
Why would surface conductance be exciting? Because conducting surface states are sensitive to
disorder or other surface effects (they tend to favor localization under disorder, for example) and
can be difficult to observe in experiments, except when protected by an interplay of symmetry and
topology. Topological order, however, cannot exist in 3D only in 2D, but surface states can be
protected with gapless topologically distinct surface states creating a 3D topological insulator. (TI)
These states are called Dirac surface states [7](For more information see the references within [7]).
Soon after Wolgast et al., Kim et al. [13, 7] measured Hall voltages on a wedge shaped sample
to determine the dependence on thickness. At  4K temperatures the Hall resistance3 was found
to be inversely proportional to the sample’s thickness indicating bulk conductivity. Below 4K, the
2
Originally titled Discovery of the First Topological Kondo Insulator: SmB 6 but then renamed to Low-temperature
surface conduction in the Kondo insulator SmB6
3
resistance taken by measuring the resistance perpendicular to an applied current

8
Hall resistance became thickness independent indicating surface conduction.

Figure 7: Kim et al. Hall measurements on a wedge shaped sample [13, 7]

Still these experiments only showed that SmB6 has resilient surface states, not that they are
topologically protected. An experiment that was used to verify 2D TIs, spin-resolved Angle Resolved
PhotoEmission Spectroscopy (ARPES), has not been successful with this material because of the
small energy gap in the bulk. That isn’t to say there aren’t results for this experiment, but they
are conflicting with other observations, with one group even appears to have found signs of a trivial
insulator rather than a TI from the presence of Rashba-splitting [7].
Even with ARPES out of reach, by applying a magnetic field time reversal symmetry can be
broken. Using weak-antilocalization (WAL) effects, more results have pointed toward a topological
nature of the surface state, but can still be related back to spin-orbit coupling. Measuring a non-
trivial Berry phase using quantum oscillation experiments, a phase of π in particular, would further
the TKI theory and more experiments have shown this to be a case, except a baffling quasiparticle
effective mass measurement of ∼ 0.1me [7].
While there are further experiments shedding more and more light on its properties, each new
turn comes with further questions. Theoretically it has been suggested recently that SmB6 may be
in fact a Skyrme insulator [14] or that the bulk is a Majorana Fermi liquid [15]. The bulk behaving
as a Fermi liquid is backed up by measurements where the linear specific heat was measured after
the sample was ground up into a powder, such that the surface area drastically increased [16].

8 A Model for Topological Kondo Insulators

Instead of focusing on an effective model describing the microscopic mechanisms of SmB6 , Dzero
et al. has provided a theoretical model for a general topological Kondo Insulator [17]. A more
complex approach can be used using the Anderson lattice model, however we will present an effective
Hamiltonian they have introduced instead. A simple TKI Hamiltonian on a cubic lattice consists of
hybridizing conduction electrons (with dispersion c (k)) with the localized f electrons (with f (k)),
! 
c (k) V d~k · ~σ

X  ckc
H= c†kc c†kf (8)
k
V d~k · ~σ f (k) ckf
H(k)

9
where d~k = (sin kx , sin ky , sin kz ). Notice that near k = 0 d~k ≈ k. If we define ± (k) = (c (k) ±
f (k))/2 the energy of the effective Hamiltonian is
q
E = + (k) ± 2− (k) + V 2 |dk |2 . (9)

To determine the topological properties, one needs to consider the 8 high symmetry points in the
Brillouin zone which are invariant under time reversal symmetry. This restricts H to have the
following symmetry properties
H(k) = PH(−k)P −1
H(k) = T H(−k)T −1
where P, T are parity and time-reversal operators respectively with the form
1 0
   
iσ2 0
P= ; T = (10)
0 −1 0 iσ2

where σ2 is the second Pauli matrix. If its assumed that d~k = −d~−k , or that the hybridization form
factor is odd, then at any of the time reversal symmetry points it must be 0 (as d~k = d~−k = −d~k ).
The Hamiltonian H can be written at those points, denoted k∗m as
H(k∗m ) = + 1 + − P. (11)
with m as an index for each of the 8 symmetry points.
What determines the topological properties of the insulator are topological indices; here there
is one strong index and three weak indices. The indices are made up of products of δm =
sgn(c − f )|k∗ , and the overall sign can be classified into strong or weak topological insulators
m
or a conventional band insulator. This is known as a Z2 index as there are two possible descrip-
tors, ±1. With the right combination of band structure one can create a strong topological Kondo
insulator, as well as a weak TI and conventional insulator [7, 17]. With the ability to create a toy
model of a TKI provides hope for discovery a 3D TI with perhaps heavy fermion properties, even if
SmB6 is not believed to be described with this particular toy model.

Part V
Conclusion
Magnetic impurities and localized magnetic moments have shown to create an expansive set of
emergent phenomena. Much of the related effects create quasiparticles with a large effective electron
mass, explained in part by various Kondo or Anderson models. Such massive particles are indicative
of interesting emergent phenomena, mostly at low temperatures. In particular, low temperature
resistivity measurement was of particular importance to understanding the underlying mechanisms,
albeit not all low temperature behaviors have been explained.
We focused on SmB6 , a Kondo insulator that is strongly believed to be a 3D topological Kondo
insulator. While there are many open questions to explain the experimental observation, a plethora
of theories are being created to better understand and test the behavioral mechanisms.
There were two further topics to be explored beyond the scope of this paper. The first, renor-
malization and the scaling of the Kondo model, provide clearer pictures as to the connections of
various toy models to effective models. The second, heavy fermion superconductivity, provides more
questions and another set of emergent physics to be studied.

10
References
[1] A. Menth, E. Buehler, and T. H. Geballe. Magnetic and semiconducting properties of SmB6 .
Phys. Rev. Lett., 22(7):295–297, 1969. doi:10.1103/PhysRevLett.22.295.

[2] N. Tsujii, H. Kontani, and K. Yoshimura. Universality in Heavy Fermion Systems with General
Degeneracy. Phys. Rev. Lett., 94(5):057201, 2005. doi:10.1103/PhysRevLett.94.057201.

[3] Piers Coleman. Heavy Fermions: Electrons at the Edge of Magnetism. In Handb. Magn. Adv.
Magn. Mater., volume 1, pages 95–148. John Wiley & Sons, Ltd, Chichester, UK, dec 2007.
arXiv:0612006, doi:10.1002/9780470022184.hmm105.

[4] W.J. de Haas, J. de Boer, and G.J. van dën Berg. The electrical resistance of gold, copper and
lead at low temperatures. Physica, 1(7-12):1115–1124, 1934. doi:10.1016/S0031-8914(34)
80310-2.

[5] M. P. Sarachik, E. Corenzwit, and L. D. Longinotti. Resistivity of Mo-Nb and Mo-Re Alloys
Containing 1% Fe. Phys. Rev., 135(4A):A1041–A1045, 1964. doi:10.1103/PhysRev.135.
A1041.

[6] Piers Coleman. Introduction to Many-Body Physics. Cambridge University Press, Cambridge,
2015. doi:10.1017/CBO9781139020916.

[7] Maxim Dzero, Jing Xia, Victor Galitski, and Piers Coleman. Topological Kondo Insulators.
Annu. Rev. Condens. Matter Phys., 7(1):249–280, 2016. arXiv:1303.7224, doi:10.1146/
annurev-conmatphys-031214-014749.

[8] Barbara A. Jones. The Kondo Effect. In Handb. Magn. Adv. Magn. Mater., pages 1–15. John
Wiley & Sons, Ltd, Chichester, UK, dec 2007. doi:10.1002/9780470022184.hmm106.

[9] S. Doniach. The Kondo lattice and weak antiferromagnetism. Phys. B+C, 91:231–234, 1977.
doi:10.1016/0378-4363(77)90190-5.

[10] Peter S. Riseborough. Heavy fermion semiconductors. Adv. Phys., 49(3):257–320, 2000. doi:
10.1080/000187300243345.

[11] Rui Yu, Hongming Weng, Xiao Hu, Zhong Fang, and Xi Dai. Model Hamiltonian for topological
Kondo insulator SmB6 . New J. Phys., 17(2):023012, 2015. doi:10.1088/1367-2630/17/2/
023012.

[12] Steven Wolgast, Çağliyan Kurdak, Kai Sun, J. W. Allen, Dae Jeong Kim, and Zachary Fisk.
Low-temperature surface conduction in the Kondo insulator SmB6 . Phys. Rev. B, 88(18):1–5,
2013. arXiv:1211.5104, doi:10.1103/PhysRevB.88.180405.

[13] D. J. Kim, S. Thomas, T. Grant, J. Botimer, Z. Fisk, and Jing Xia. Surface Hall Effect
and Nonlocal Transport in SmB6 : Evidence for Surface Conduction. Sci. Rep., 3:1–4, 2013.
arXiv:1211.6769, doi:10.1038/srep03150.

[14] Onur Erten, Po-Yao Chang, Piers Coleman, and Alexei M. Tsvelik. Skyrme Insulators: In-
sulators at the Brink of Superconductivity. Phys. Rev. Lett., 119(5):057603, 2017. doi:
10.1103/PhysRevLett.119.057603.

11
[15] G. Baskaran. Majorana Fermi Sea in Insulating SmB6 : A proposal and a Theory of Quantum
Oscillations in Kondo Insulators. 2015. arXiv:1507.03477.

[16] N. Wakeham, P. F. S. Rosa, Y. Q. Wang, M. Kang, Z. Fisk, F. Ronning, and J. D. Thompson.

Low-temperature conducting state in two candidate topological Kondo insulators: SmB6 and
Ce3 Bi4 Pt3 . Phys. Rev. B, 94(3):035127, 2016. doi:10.1103/PhysRevB.94.035127.

[17] Maxim Dzero, Kai Sun, Victor Galitski, and Piers Coleman. Topological Kondo Insulators.
Phys. Rev. Lett., 104(10):106408, 2010. doi:10.1103/PhysRevLett.104.106408.

12