ReVIeWS

Axion physics in condensed-​matter

systems
Dennis M. Nenno1,2 ✉, Christina A. C. Garcia   2, Johannes Gooth1,3, Claudia Felser   1,2
and Prineha Narang   2 ✉
Abstract | Axions are hypothetical particles that were proposed to solve the strong charge–parity
problem in high-​energy physics. Although they have long been known in quantum field theory,
axions have so far not been observed as elementary particles in nature. Yet, in condensed-​matter
systems, axions can also emerge as quasiparticles in certain materials such as strong topological
insulators. The corresponding axion field is expected to lead to exciting physical phenomena
in condensed-​matter systems, such as a fractional quantum anomalous Hall effect, the chiral
anomaly, exotic Casimir–Lifshitz repulsion and a linear magnetoelectric response quantized
in units of the fine-​structure constant. First signatures of electronic states that permit axion
dynamics have been reported in condensed-​matter systems. In this Review, we explore the
concepts that introduce axion fields in condensed-​matter systems and present experimental
findings. We discuss predicted and realized material systems, the prospects of using axion
electrodynamics for next-​generation devices and the search for axions as a possible constituent
of dark matter.

Z2 invariant
The concepts of the axion field and its fundamental with respect to perturbations or continuous transfor-
Group of integers 0, 1 first excitation — the axion particle — were first intro- mations. Topological field theories can predict the
introduced in 2D time-​reversal-​ duced more than 40 years ago to explain the absence responses of these materials irrespective of microscopic
invariant systems to distinguish of charge–parity violation in the strong interaction details, making assumptions only about the symmetries
topological from trivial phases.
between quarks1,2. Since then, it has been adopted in of the system. Among them, time-​reversal-​invariant
string theory and cosmology, and the axion particle has insulators in 3D form a class that can be divided into
become a current candidate for dark matter3, although ‘strong’ and ‘weak’ TIs, as well as ‘trivial’ insulators,
no traces of the elementary axion particle have yet been by an integer value derived from their band structure,
observed. Axions are introduced by an additional term the ‘Z 2 invariant’16. This classification was extended to
in the laws of electrodynamics, which rotates the elec- systems that break time-​reversal symmetry, such as
tric and magnetic fields into each other. In 1987, Frank antiferromagnetic TIs17. The electromagnetic coupling
Wilczek proposed that similar laws of electrodynamics governed by the axion field in these structures provides
could potentially be observed in condensed-​matter one method of calculating the Z 2 invariant, with a direct
systems4. Twenty years later, it was discovered that physical interpretation5,17.
1
Max Planck Institute for a term analogous to the one that would be added to In a strong TI, the effective axion term derived from
Chemical Physics of Solids,
Maxwell’s equations to account for elementary axions its electronic states can lead to unique responses under
Dresden, Germany.
does indeed appear in the description of certain top- certain conditions. Among them, the topological mag-
2
John A. Paulson School of
Engineering and Applied
ological materials5–8, giving rise to an effective axion netoelectric effect results in a linear coupling between
Sciences, Harvard University, description of the electrodynamics in these systems. electric and magnetic fields5,6,18. This effect leads to an
Cambridge, MA, USA. The study of topological materials is one of the most induced electric polarization when applying a magne­
3
Institut für Festkörper- und active fields of research in modern condensed-​matter tic field and a magnetization when applying an electric
Materialphysik, Technische science9,10, and it is believed that at least a quarter of field. The fact that this linear coupling is present even in
Universität Dresden, Dresden, all known materials have topological properties11. The centrosymmetric or time-​reversal-​invariant materials is
Germany.
field started with the prediction and realization of top- a striking physical consequence, as there exist only a few
✉e-​mail:
ological insulators (TIs), materials that are insulating in well-studied materials that exhibit such a coupling5,19.
nenno@seas.harvard.edu;
prineha@seas.harvard.edu the bulk but show quantized, dissipationless metallic In addition, the quantum anomalous Hall effect that
https://doi.org/10.1038/ states on their surfaces12–15. At a fundamental level, TIs naturally arises in magnetic TIs can be readily explained
s42254-020-0240-2 are characterized by quantized responses that are stable by the surface current response obtained from axion

Nature Reviews | Physics

Reviews

Key points materials for dark-matter detection, non-reciprocal

emission and their connection to other topological
• 3D insulators can be topologically characterized by the value of their bulk axion field. devices.
• Axion fields introduce additional terms in Maxwell’s equations for condensed-​matter
systems. Topological field theory
• The microscopic expression for the axion field in a crystal is given by the non-​Abelian Particle physics formulates its governing laws in terms of
Chern–Simons integral, which depends on the Berry connection matrix of the band quantum field theories, in which particles are described
structure. as quantized excitations of the field. The standard
• In strong 3D topological insulators, a half-​quantized surface Hall effect appears when model of particle physics can be stated in terms of its
the surface states are gapped, together with linear magnetoelectric coupling in Lagrangian, which controls the dynamics and interac-
their bulk. tion of elementary particles. As mentioned in the intro-
• The axion insulator state can be realized in antiferromagnetic insulators without duction, the term accounting for the axion is added to
external fields. the Lagrangian of the standard model to solve the strong
• Materials with a non-​trivial axion field can be used in dark-​matter detectors and charge–parity problem2,38. We now discuss how this
non-​reciprocal thermal emitters. additional term arises naturally in strong TIs in three
dimensions, how it leads to a topological classification
in materials that break time-​reversal symmetry and how
electrodynamics5. The microscopic expression that it is connected to the formal polarization. The physi-
determines the axion field takes quantized values of 0 cal effects of the coupling can manifest themselves in
or π in time-​reversal-​invariant insulators, providing a half-​quantized surface states and the topological mag-
topological classification equivalent to the Z 2 invar- netoelectric effect. In Weyl semimetals, an axion term
iant in three dimensions. Furthermore, in structures arises from the separation of the chiral band crossings
that lack time-​reversal symmetry, spatial or compos- in momentum space and energy7,8,22. Furthermore, the
ite symmetries can also quantize the axion field5,6,17,20. condensed-​matter realization of axion electrodynamics
Axion domain walls appear at interfaces and surfaces is a fascinating example in which the low-​energy contri-
of these materials and eventually lead to half-​quantized bution to the field (the ‘axion angle’) can be determined
Hall responses5,21. Compared with its elementary version by the microscopic theory.
in high-​energy physics, the condensed-​matter realiza- With the discovery of the quantum Hall effect —
tions of the axion have the advantage of being directly the quantized version of the Hall effect arising in a 2D
accessible in magnetoelectric experiments. electron gas subject to a strong magnetic field — the
It was only a few years after the discovery of 3D TIs concept of topology was introduced in the description
that axion electrodynamics was proposed to govern the of electronic states in condensed-​matter systems39,40.
electromagnetic response in Weyl semimetals as well7,8. David Thouless, Mahito Kohmoto, Marcel den Nijs
The chiral anomaly related to this response had already and Peter Nightingale (TKNN) realized that individual
been predicted to arise in the early 1980s (ref.22). The Bloch bands uj yield quantized contributions to the Hall
realization of Weyl semimetals, as well as their response, conductivity,
has been the subject of several excellent reviews23–26, so

∫BZ d2k ( ),
we centre our discussion around insulating materials. i
nj = ∂ k yuj ∂ k x uj − ∂ k x uj ∂ k yuj (1)
The electronic states in materials that contribute to 2π
the axion field have not only been studied in the origi-
nally proposed bismuth-​based alloys5,27 but have now also where the integral runs over the Brillioun zone (BZ) of
been observed in, among others, charge-​density-​wave the crystal, with the quasi-​momentum axis k. The con-
Weyl semimetals28, antiferromagnetic insulators17,29,30 tributions from occupied bands can be added together
and topological heterostructures31–33. In these mate- to obtain the total Hall conductivity,
rials, experiments have been able to probe signatures
of the peculiar electrodynamics that lead to a quan- e2 e2
tized topological response34–37. However, even though
σxy =
h
∑ j∈occ nj = h C 1. (2)
the basis in terms of suitable materials and structures
is established, a direct observation of the topological The sum of all nj is called the TKNN invariant or
magnetoelectric effect is still missing. Its quantization first Chern number (C1) and provides an integer classifi-
of the interaction in units of the fine-​structure con- cation of the system which is observable in experiment.
stant makes a direct measurement a desirable goal for This result can be obtained from standard perturbation
experimentalists. theory for Bloch states um,n and written in terms of the
Our Review provides an introduction to the topic of Berry connection,
Chern number
axion electrodynamics in condensed-​matter systems.
Berry flux on a closed
manifold, which becomes In particular, we focus on the resulting phenomena in
∂
quantized in time-​reversal-​ TIs, comment briefly on the chiral anomaly as manifes- aμmn(k) = − i um(k) un(k) , (3)
breaking topological insulators. tation of axion electrodynamics in Weyl semimetals and ∂kμ
discuss potential material systems for the realization of
Berry connection
Gauge-​dependent vector
the resulting physics. We highlight recent experiments in as C 1 = 1/(2π) ∫ d2k(∂ k x a y − ∂ k yax ). The quantum Hall
potential connected to the candidate axion insulator systems and close the Review state resides in a topologically non-​t rivial Hilbert
Berry phase. with a discussion of applications of axion insulator space, and its Chern number C1 provides an invariant

www.nature.com/natrevphys
 Reviews

that specifies the system. Its electromagnetic action is Inversion symmetry can further aid the calculation
described by a Chern–Simons term5 of ν, as only special points of the Brillouin zone have
to be considered in this case45. Several excellent reviews
2D C 1 elaborate on the derivation of schemes to calculate this
Seff =
4π
∫ d x ∫ dt ε Aμ ∂νA τ .
2 μντ
(4)
invariant from the entire bulk or specific points of the
band structure9,10. In practice, a feasible way to calcu-
Here, ε denotes the Levi–Civita symbol, and we late the Z 2 invariant is to track centres of the Wannier
sum over spatiotemporal indices in (2 + 1) dimensions charge, elaborated on in refs46,47.
(μ,ν,τ = t,x,y). The external vector field is denoted by A, To realize a Z 2-​insulator in three dimensions, one
and ∂ν takes its spatial derivative along ν. The coupling possibility is to stack layers of quantum spin Hall insu-
constant C1 to the field theory provided by Eq. (4) is lators. The additional dimension yields three different
exactly the first Chern number, which again is identified invariants for each spatial component. This turns out
by the microscopic details of the system, encoded in the to be insufficient to describe the topology of the entire
Bloch bands and the Berry connection, Eq. (3). From system, as six individual numbers for every surface of a
the action, one can recover the response of the current cube are required; however, because these surfaces are
to an external potential A (ref.5), not independent, four quantities are able to fully char-
acterize the topology of a 3D insulator, usually written
μ C 1 μντ as the set (ν0; ν1ν2ν3) (refs14,45). If the first index is odd,
j = ε ∂ν A τ , (5)
2π the material is a strong TI; if it is even and at least one
of the latter three is odd, it is understood to be a weak
from which the quantized Hall effect, Eq. (2), can imme- TI. Weak TIs have an even number of conducting Dirac
diately be obtained. The quantum Hall system is the first cones on the surface, whereas strong TIs show an odd
realization of a class of materials called TIs, as its bulk number of surface cones. Both cases are topologically
states are gapped in the presence of an external mag- non-​trivial states in the clean limit, that is, for systems
netic field, whereas its surface is characterized by chiral with unperturbed bulk states, in which they show the
gapless states. bulk–boundary correspondence, exhibiting metallic
In 2005, the introduction of the quantum spin Hall surface states48,49.
effect generalized the idea of topology to the time-​ The first strong TI was experimentally identified by
invariant case16. The quantum spin Hall insulator was angle-​resolved photoemission in Bi1−xSbx compounds,
proposed as a new type of band insulator with two spin despite a relatively small bulk gap of the order of thermal
channels that are coupled by the spin–orbit interaction excitations50. Using the same technique, a single Dirac
on a graphene lattice. The presence of time-​reversal sym- cone on the surface of the otherwise insulating Bi2Te3
metry in this system results in its electronic structure was identified, providing the first observation in the pre-
having topologically protected non-​trivial character, dicted family of Bi2Te3 and Sb2Te3 compounds51. Bi2Te3 is
making it the first predicted symmetry-​protected top- the heaviest combination in this family and establishes
ological phase. Such a phase was both predicted and a bandgap of 0.3 eV, rendering it a prototype material in
experimentally observed in cladded HgTe structures in this field. Many more topological compounds have been
the following 2 years41,42. Time-​reversal-​invariant TIs identified, synthesized and measured in recent years,
in two and three dimensions can be classified by their opening a wide field of research9–11,52.
Z 2 invariant, an integer value that is either 0 or 1, sepa-
rating trivial from topological states. A brief expression Axion coupling
for the topological invariant ν, sometimes referred to In an effort to unify the different topological quantities in
as the Fu–Kane formula, is based on the Pfaffian of the two and three dimensions, a field theory was intro-
system, Pf (refs43,44), duced based on the nonlinear response of an imaginary
(4 + 1)-​dimensional TI5. In a spirit similar to the deriva-
Det[w(Γj)] tion of the first Chern number from linear perturbation
4
(−1)ν = ∏ j =1 , (6) theory, an expression for the second Chern number is
Pf[w(Γj)] derived, which appears as a nonlinear response coeffi-
cient. For (4 + 1) dimensions, the second Chern number,
where Det is the determinant, Γ j the four time-​ C2, is directly analogous to the TKNN invariant in (2 + 1)
reversal-​invariant momenta in two dimensions and w dimensions in describing the response of a current to
is defined by an external field. By dimensional reduction — choosing
wmn(k) = um(k)∣T ∣un(−k) . (7) one of the four momentum directions and its coupling
to the vector potential as an additional parameter — one
The argument of the Pfaffian, w, calculates the over- obtains both a closed expression for the second Chern
lap between states in different bands under time rever- number and a pseudoscalar polarization term that cou-
sal, T , and is consequently sensitive to degeneracies. ples nonlinearly to the external fields E and B. The cou-
The Pfaffian exists because w in Eq. (7) is an antisym- pling to the vector potential generating these fields, A,
metric matrix. The Z 2 invariant ν can take only two val- is given by the effective action in (3 + 1) dimensions5.
ues, 0 and 1, and, although it remains a bulk quantity
evaluated from its Bloch bands, this bulk topological 1
index decides the existence of non-​trivial surface states.
3D
Seff =
4π
∫ d3x ∫ dt ε μνστθ(x, t )∂μA ν ∂σ A τ , (8)

Nature Reviews | Physics

Reviews

Kramers degeneracy
again using ε to denote the Levi–Civita symbol. these two values in the bulk of the material55. Insulators
In time-​reversal-​symmetric Equation (8) formally introduces the axion field θ, with θ = π protected by such a symmetry that is not
systems, every energy state is which is defined by a Chern–Simons form similar to that time reversal are called axion insulators. The surfaces
at least two-​fold degenerate. of the (2 + 1)-​dimensional action in equation (4): of these materials are naturally gapped56, and a further
elaboration on the equivalence between Z 2 and θ can be
1  2i  found in ref.57. An overview of topological classification
θ=−
4π
∫BZ d3k ε μνσ tr aμ ∂νaσ − 3
aμa ν aσ  .

(9)
schemes is shown in Box 1.
Magnetic materials are not included in conventional
The trace runs over the band index of the Berry classifications for 3D TIs, as their ground states break
connection, a, between occupied bands. We call θ an time-​reversal symmetry, and the key to establish a
axion field in analogy with its interpretation as a fun- Z 2 classification is the existence of Kramers degeneracy58.
damental gauge potential in high-​energy physics. Its However, whereas ferromagnetism breaks time-​reversal
quantized excitations are consequently called axions. symmetry, it can be shown that for almost all states,
In the condensed-​matter context, θ can also be inter- degeneracy is ensured in an antiferromagnet if there is
preted as a contribution to the magnetoelectric polar- an additional operation that connects lattice sites with
ization from extended orbitals5,6,53. This contribution opposite spin58. Topological antiferromagnets are now an
to the polarization is derived from an integral over the active area of research13,59. Usually, a half-​lattice transla-
Berry curvature tensor and arrives at an expression tion τ1/2 connects the different sites that build a staggered
similar to equation (8)6, and ref.54 provides an excellent Zeeman field and can be combined with time reversal
overview about the interpretation of the axion field as to create an anti-​unitary operator. We elaborate on this
a constituent of the microscopic polarizability. using the example of the recently proposed MnBi2Te4
As θ is obtained as the result of integrating a 3D structure later in the text.
Chern–Simons term, it is only defined modulo 2π
(refs5,6,54). In the time-​reversal-​invariant insulator dis- Half-​quantized interface conductance
cussed above, θ can take only two values, 0 for trivial From Eq. (8), a number of exciting physical phenomena
insulators and π for strong TIs, and it thus provides a follow, many of which have been proposed in ref.5. We
Z 2 classification. 5 In fact, in any pseudoscalar-​symmetric now review some of the effects that arise for static axion
crystal, a crystal in which a symmetry is present that fields, before turning to dynamical phenomena. One of
reverses the sign of a pseudoscalar, θ can only take the most interesting features and a direct consequence
of a quantized axion field in the bulk is the occurrence of
a half-​quantized Hall effect on the surface of a strong
Box 1 | Topological classification TI, provided that the interface breaks time-​reversal sym-
metry and gaps the surface states. Such an interface can
Materials can be classified by the topology of their quantum mechanical wavefunc-
be interpreted as an axionic domain wall, as it separates
tions. A general classification scheme categorizes topological insulators (TIs) in the
independent particle regime depending on their non-​spatial, anti-​unitary symmetries.
regions with θ = π and θ = 0 mod 2π (refs4,55). One can
Time-​reversal and charge-​conjugation (particle–hole) symmetry, together with their show that the action, Eq. (8), yields a half-​quantized
combination (chiral symmetry), yield ten different categories. These are the ten surface Hall conductance, which in general is given by
Altland–Zirnbauer symmetry classes (excerpt shown in the table), which depend on the (1/2 + m)e2/h. While the odd number of surface Dirac
sign of the symmetry of the Hamiltonian160,161. The topological classification of a system cones contributes a half Hall quantum, the second part
within one category is then defined by the homotopically inequivalent maps from stems from an arbitrary number m ∈ Z of quantum Hall
momentum space (parametrized by k) to the Hamiltonian. 2D time-​reversal symmetry effects on the surface60. When the surface states are
breaking insulators are characterized by the first Chern number C1 ∈ Z and fall into gapped by a time-​reversal-​breaking perturbation, such
class A. 2D and 3D TIs that have time-​reversal symmetry can both be characterized by as an external magnetic field or cladding with an insu-
a Z 2 number and fall into class AII. The difference between these classes can be under-
lating magnetic material, the surface Hall conductance
stood through the difference in the transformation of the action integral under charge
conjugation and time reversal5, and, as the action integral is exactly a Chern–Simons
becomes5,61
form, such transformations are well understood mathematically162–164. In general, Chern e2
numbers are the coefficients of the Chern–Simons action integral in d dimensions, σxy = ± , (10)
where the coefficient in the latter is the d/2-​Chern number5,60. A physically intuitive 2h
picture of this topic is presented in ref.60. where the sign depends on the time-​reversal-​breaking
Introducing spatial symmetries into the classification scheme makes it possible field. This result can be observed in experiment as fol-
to derive simplified expressions for the topological invariants165. The second Chern lows. Suppose the surfaces of a strong TI are sandwiched
number C2 ∈ Z provides an invariant for the (4 + 1)-​dimensional quantum Hall insulator, between two ferromagnetic insulators. Given that the
and, by dimensional reduction, one can derive the Chern–Simons form for strong TIs magnetizations of the top and bottom ferromagnets are
in three dimensions, which show the topological magnetoelectric effect. By further
parallel, one expects to measure one quantum Hall con-
reducing the dimensionality, we obtain the class of quantum spin Hall insulators and
its classification by, again, a Z 2 invariant5.
ductance, as both surfaces contribute half a quantum.
If the magnetization in one of the layers is now reversed,
Excerpt of Altland–Zirnbauer symmetry classes for 2D and 3D topological the currents on the side surfaces cancel, and a zero Hall
insulators plateau is obtained5. The two scenarios are depicted in
class (2 + 1)d (3 + 1)d (4 + 1)d Fig. 1a,b. In experiment, this effect can only be observed
A Z – – if the magnetically induced gap is larger than the thermal
energy and if the chemical potentials of both the top and
AII Z2 Z2 Z bottom surfaces lie within that gap. Heterostructures of

www.nature.com/natrevphys
 Reviews

a b c E d B

e2 e2
σ xy = σ xy =
FM 2h FM 2h TI TI
––––––
TI e2 TI e2 j
σ xy = σ xy = – FM FM Pt
2h 2h +
+++++
FM FM

Fig. 1 | Quantized responses in topological insulator–ferromagnet heterostructures. a ∣ Topological insulator (TI)

cladded by ferromagnets (FMs) with parallel magnetization. Both top and bottom surface of the TI are gapped and
contribute a half-​quantized quantum Hall conductance σxy for a perpendicular electric field E. The current is carried
by a chiral surface state on the side of the TI. b ∣ For antiparallel configuration, the signs of the induced currents cancel
and no Hall current is measured. c ∣ Induced fields from the topological magnetoelectric effect when an external field
is applied to a cylindrical insulator structure. A magnetization M is induced by an external electric field via a tangential
current j. d ∣ A polarization Pt builds up as a consequence of an external magnetic field B. Adapted with permission
from ref.5, APS.

magnetically doped TIs can aid the search for the zero realizing a quantized coupling of α ≈ 24.3 ps m−1, which
quantum Hall plateau, as their topological band struc- is an order of magnitude stronger than in typical mag-
ture can be tuned more transparently by structural netoelectric materials, such as Cr2O3 (ref.67). Not only
parameters62–64. does the coupling to the axion field allow for a quan-
tized linear electric coupling, but structures based on
Topological magnetoelectric effect wrapped 3D quantum anomalous Hall insulators have
The topological magnetoelectric effect is a particu- been suggested to allow for enormous magnetoelectric
larly intriguing phenomenon that appears when the coupling of the order of 103 in natural units55.
surface of a strong TI is fully gapped, such as by an
outward-​pointing field5. Two examples are depicted in Quantized optical response
Fig. 1c,d, for a system that consists of a cylindrical TI In optical measurements of strong TIs, such as through
sample wrapped by a ferromagnet with its magnetization Kerr and Faraday set-​ups, axion electrodynamics yields
orthogonal to the interface. Applying an electric field a quantized rotation of the field polarization axis5.
to the material induces a quantized current perpendi­ Snell’s law is unmodified in the presence of a jump in
cular to the normal vector of the magnetization and the the axion field, but a non-​trivial Faraday effect arises
field (Fig. 1c). This current then induces a magnetization that was predicted to be observable in experiment and
antiparallel to the electric field, which is quantized in provides direct access to the fine-​structure constant by
units of the fine-​structure constant α, optical means68. The measurement procedure69 requires
that both Kerr and Faraday rotations ΘF should be
 1 measured while varying the external field, although,
M = −n +  α E . (11)
 2  technically, it should be possible to derive the quantiza-
tion from just one of these procedures. This is a gene­
Vice versa, an external magnetic field polarizes the ralization of a previous result that proposed a giant
material along the same axis (Fig. 1d). Thus, one way to Kerr rotation (ΘK ≈ π/2) in an optimized set-​up70. In par-
access this quantized magnetoelectric effect is through ticular, the quantized Kerr and Faraday angles depend
r
magnetic field detection. The induced polarization or on the real part of the optical conductivity σxy , which
magnetization can also be captured by Maxwell’s equa- in the long-​wave limit becomes quantized. In the sam-
tions when properly accounting for Eq. (8). Box 2 intro- ple, the induced rotations can then be expressed as
duces how the constitutive material relations change in Θ K = −tan −1 (c/(2πσxy r
) and Θ F = tan −1(2πσxy
r
/c ) (refs70,71).
the presence of the additional coupling. To gain access to the half-​quantized conductances of the
The topological magnetoelectric effect contributes top and bottom surfaces, one has to measure Kerr and
to the linear magnetoelectric polarizability, which can Faraday rotations independently in films that are much
be written as6 thicker than the probing wavelength72.
Although a well-known occurrence in high-​energy
∂M j ∂Pi α physics, the proposal of an image magnetic monopole
α ij = = = α∼ij + θ 2 δij , (12) mirror charge induced by an electric charge outside
∂E i ∂B j 4π
B =0 E =0 the TI is a more exotic effect in the condensed-​matter
framework18,73. Electrons on the surface and induced
where α∼ij denotes the orbital, spin and ionic contribu- magnetic monopoles form dyons — hypothetical parti-
tions, which vanish in inversion- and time-​reversal-​ cles with both electric and magnetic charge — in the static
symmetric compounds65,66. This is particularly intriguing, limit, and an effective field can be observed with mag-
as strong TIs have been shown to provide a possibility of netic force microscopy. This concept is explained in detail

Nature Reviews | Physics

Reviews

Box 2 | change of the constitutive relations Casimir–Lifshitz repulsion

Axion electrodynamics in topological materials has
The topological field theory (8) predicts an additional also been proposed to have observable consequences
term, Lθ , in the electromagnetic action of the material in the interactions between materials due to electro-
system that scales with the axion angle θ and the magnetic fluctuations. The Casimir–Lifshitz effect79–82
fine-​structure constant α (refs4,5,17),
arises from quantum fluctuations in the electromagnetic
α field and results in a force between two objects that are
Lθ = θ E ⋅ B.
4π 2 micrometres apart or closer. Although in most cases the
As an immediate consequence, this axion term leads Casimir force is attractive, repulsion (while maintain-
to an extension of the Maxwell relations for all materials ing vacuum or air between the two interacting objects)
with an effective coupling described by is desirable for enhanced control over friction and
α applications such as quantum levitation83.
∇ ⋅ E = 4πρ − (∇θ) ⋅ B In 2011, it was proposed that both attractive and
π
4π 1 ∂E repulsive Casimir forces could be possible in a set-​up
∇×B= μ J+
c 0 c ∂t (13) involving two 3D TIs with thin magnetic layers covering
α 1  ∂   each of the two interacting faces84. Because the constit-
+  B θ + (∇θ) × E ,
π  c  ∂t   uent relations describing the electromagnetic response
include extra terms due to the axion field, the authors
effectively altering the constitutive relations of the mate­
concluded that if the magnetizations of the surface lay-
rials. This result can be derived both from the topological
field theory, and from the microscopic topological band ers have opposite signs, there is a characteristic length
theory, which fixes θ in the bulk of a given material5,54. dependent on the dielectric functions. Below this char-
From Eq. (13), it is clear that only a spatial or temporal acteristic length, the Casimir–Lifshitz force is repulsive,
change in θ induces an observable effect. and above it, the force is attractive, with quantum levita-
tion being possible when the plates are separated exactly
by the characteristic length. Many theoretical studies fol-
in ref.74, although so far there exists no experimental lowed that examined the role of temperature85,86, uniaxial
verification of this prediction. anisotropy85, arbitrary object shape86, surface bandgap
size87,88, slab thickness89 and the presence of substrates89
Dynamical axion fields on this model. Further, it was proposed that the repul-
In 2010, two studies presented the idea of topologi- sion could be enhanced by using multilayers of magnetic
cal magnetic insulators, materials that can realize the layer-​coated TIs alternating with normal insulators90.
dynamic axion field by having long-​range magnetic Similar propositions using Chern insulators91 or
order that breaks time-​reversal symmetry 17,27. As Weyl semimetals92 instead of 3D TIs have been put
magnetic fluctuations in these systems couple to the forwards. Recent work, however, suggests that where
axion field, hybridized modes with external fields can repulsion is possible with the Weyl semimetal model, it
emerge75,76. An example is the axion polariton, which is small and hard to obtain and that the Casimir–Lifshitz
couples the axion and the photon field and is predicted force is dominated instead by the metallic nature of the
to have a dispersion relation analogous to the optical semimetals93. Although the models involving TIs seem
phonon polariton. One of the key differences, however, more promising, they still require the surface Hall con-
is that a gap in the spectrum of these axion polaritons ductivity related to the axion field to dominate over
can be tuned by the external magnetic field and thus can the bulk longitudinal conductivity94,95, and it has been
be observed in attenuated total reflection experiments27. argued that this makes the repulsive Casimir–Lifshitz
No signs of this theorized quasiparticle have been effect unlikely to be found using topological materials96.
observed so far, because of the lack of experimental Additional studies explore the Casimir stress on a TI
resolution in the terahertz range and the challenge of placed between two conducting plates97 and the repul-
exciting longitudinal fluctuations in existing materials77. sion between two conductors with a spatially inhomo-
At much lower frequencies, of the order of 1 Hz, the geneous axion field between them — with a suggested
application of a slowly varying external electric field and realization using Weyl semimetal as the axion field
a constant magnetic field to an axion insulator on the tip medium98. However, these proposals do not include
of a cantilever forces the tip of the cantilever to change other material contributions such as the bulk longitu-
its natural frequency. By microcantilever torque mag- dinal conductivity, which can dominate the dielectric
netometry, such a subtle change should be observable in response of these materials. All studies on the possibil-
experiment27. ity of Casimir–Lifshitz repulsion in topological materials
Axion electrodynamics in Weyl semimetals arises so far have been theoretical, leaving this as another
from the separation of band crossings, or Weyl nodes, avenue for experiment to explore.
of opposite chirality in momentum space and energy
(see Box 3). In material systems that exhibit a Fermi-​ Axion field in trivial materials
surface instability, the topological bands can be gapped, The effects discussed so far have the most interesting
and the dynamical axion field effectively couples to the physical consequences in the regime where θ is quan-
Casimir stress
Stress that results in Casimir
phase of the emerging charge-​density wave78. The quan- tized and the field coupling is maximal. In practice,
forces inside inhomogeneous tum of this excitation is called a phason, and its impact however, a non-​vanishing θ can exist in non-​topological
structures. on magnetotransport was detected in Ta2Se8I (ref.28). materials without particular symmetries present. In fact,

www.nature.com/natrevphys
 Reviews

the well-studied magnetoelectric Cr2O3 and multi­ k-​space mesh reveal that Cr2O3 has a relatively weak
ferroic (Fe,Zn)2Mo3O8 may be more promising candi- coupling (θ ≈ π/36) and correctly reproduce the quan-
dates than TIs for measuring the axion coupling and tized value in Bi2Se3 (ref.67). More recently, advances
its contribution to the magnetoelectric polarization, have been made to calculate the axion coupling from
as these compounds exhibit naturally gapped surface Wannier charge centres101. Even though the axion con-
states99,100. The fact that the pseudoscalar contribution tribution to the orbital magnetoelectric effect is thought
to the magnetoelectric tensor exists and is potentially to be a minor one, it should still be included in future
linked to an axion field was already established in 2009, calculations.
although without a microscopic expression for the cou- The preceding discussion focuses on the non-​
pling constant99. First calculations of Chern–Simons interacting case. In strongly interacting systems, quan-
orbital magnetoelectric coupling using Wannier-​based tum Hall states with fractional quantum numbers can
methods in real, non-​topological materials (such as emerge, and a topological field theory has been devel-
Cr2O3, BiFeO3 and GdAlO3) were presented in ref.67. It is oped to capture these effects in 3D insulators102. A prom-
worth mentioning here that the numerical evaluation ising route to detect the quantization in these states is
of the axion coupling parameter remains an intricate again optical spectroscopy68.
task. Although the field theory predicts fixed values for
the bulk of strong and trivial insulators, evaluating the Material realization
microscopic expression, Eq. (9), for systems that lack So far, we have only addressed the conceptual design of
time-​reversal or inversion symmetry is not straightfor- a topological magnetoelectric material. In this section,
ward. Wannier-​function-​based evaluations on a finite we aim to give an overview of real material structures
that have been proposed and observed to host states that
couple to axion fields Fig. 2. Here, we highlight intrin-
Box 3 | chiral anomaly in Weyl semimetals sic axion insulators and materials in which elementary
excitations have axion character. Although some of the
The word ‘anomaly’ in this context refers to when a symmetry required by classical
physical laws is broken in the quantum regime. The chiral anomaly, also referred to
former can be derived by magnetically doping strong
as the axial or Adler–Bell–Jackiw anomaly166,167, is a phenomenon known in relativistic 3D TIs, the latter typically require the presence of
quantum field theory that arises in the axial symmetry of massless, chiral particles such excitations that gap topological semimetal states, such
as Weyl fermions. This anomaly reveals itself in the number of particles of each chirality as a charge-​density wave that is commensurate with
not being separately conserved when the background gauge field is topologically the spacing of Weyl points on the Fermi surface78,103.
non-​trivial, despite this conservation being present classically8. In particular, we focus on Mn- and Cr-​doped Bi2Se3 in
Weyl semimetals are hosts to topologically protected band degeneracies24. These heterostructures, the recently realized intrinsic topo-
degeneracies, or nodes, disperse linearly in momentum and have an associated chiral- logical antiferromagnets, such as MnBi2Te4 and related
ity; thus, they can be described as quasi-​relativistic Weyl fermions, and Weyl semimetals compounds29,30,104–106, and the charge-​density-​wave Weyl
are expected to exhibit the chiral anomaly when coupled to an electromagnetic field7.
semimetal Ta2Se8I (refs28,104).
A few studies7,8,168,169 have shown that the Hamiltonian describing the low-​energy
physics of a Weyl semimetal with a single pair of Weyl nodes naturally leads to the
In many of the structures, experimental characteriza-
addition of an axial four-​vector term to the electromagnetic action of exactly the form tion is achieved by angle-​resolved photoemission spec-
in equation (8), except that the axion field, troscopy, which can access both bulk and surface states
and probe the gap of the system with optical means,
θ(r, t) = − 2bµ x µ = 2b ⋅ r − 2b 0t , (14) as well as by transport measurements that probe the
signatures of axion electrodynamics107.
is in general position-​dependent (r) and time-​dependent (t) and so remains inside the
integral. Here, x is the space–time four-​vector, 2b is the momentum-​space separation
Magnetically doped TIs
of the Weyl nodes, and 2b0 is the energetic separation in the four-​vector b = (b0, b).
The equations of motion describing two known phenomena that can be observed Axion electrodynamics was first proposed to emerge at
in Weyl semimetals were derived by varying this action with respect to the gauge interfaces between 3D TIs and trivial materials5, and in
field A (ref.168): sandwich structures with ferromagnetic materials, as
visualized in Fig. 1. However, the search for an intrin-
e2 sic host structure — that is, a bulk material rather than
jµ = bν ε νµαβ∂α Aβ, ν = 1, 2, 3
2π 2 (15) a heterostructure that breaks time-​reversal symmetry
e2 and thus gaps the Dirac surface states — was guided
j µ = − 2 b 0ε 0µαβ∂α Aβ .
2π early on by the proposal to study magnetically doped
strong TIs, Bi2Te3, Bi2Se3 and Sb2Te3 (refs27). The elec-
Here, ε denotes the Levi–Civita symbol with space–time indices ν, μ, α, β. The first of
tronic structures of these undoped materials each exhibit
these describes the anomalous Hall effect, caused by the breaking of the time-​reversal
symmetry and the separation of Weyl nodes in momentum space. The second
a single Dirac cone on the surface, and the effective
corresponds to the chiral magnetic effect, which can result from the energetic model for these materials yields values 0 and π for θ,
separation of the Weyl nodes or the application of non-​orthogonal electric and depending on parameters. An external magnetic field
magnetic fields, although the former requires a dynamical non-​equilibrium setting in applied to inversion-​symmetric Bi2Se3 alone only breaks
the low-​frequency (d.c.) limit23,24,26,168. Thus, this equation is not a complete description time-​reversal symmetry, such that there cannot be a net
but a simplified representation. Although the anomalous Hall effect has already been macroscopic moment from magnetoelectricity, and
linked to band topology and is straightforward to measure, the chiral magnetic effect inversion has to be broken by other means34. However,
is a more subtle phenomenon that has only recently been definitively captured in it was argued theoretically that a time-​reversal-​breaking
longitudinal magnetoresistance measurements170. However, both phenomena are perturbation can lead to axion electrodynamics in these
exciting, owing to the possibility of dissipationless transport using metallic systems25.
3D TIs. By doping the structure with 3d-​ferromagnets,

Nature Reviews | Physics

Reviews

a b These layers, both insulating, should differ in their coer-

civities, making it possible to access both parallel and
antiparallel magnetization configurations, analogous to
what is shown in Fig. 1a,b, with one external magnetic
field sweep. The zero Hall plateau obtained for antipar-
allel top and bottom layer magnetizations, together with
fully quantized Hall plateaus in the high-​field limit,
yields the requirements for the existence of a topolog-
ical magnetoelectric response in structures where the
magnetization gap dominates the hybridization gap.
Septuple Both the zero Hall plateau and the quantum anomalous
layer Hall effect have to be measured in the same system to
Ir establish its potential to show the topological magneto­
τ1/2 electric response. Finite-​size effects can be eliminated
for large enough layers, and a gap on the dissipative side
c J⊥ surfaces can be induced by ferromagnetic cladding in
Exchange field Magnetization a special set-​up or through quantum confinement63.
A heterostructure with both Cr- and Mn-​doped TIs
JII (Fig.  2c) has been proposed to realize the zero Hall
Top surface plateau62. This idea has further been extended to super-
Van der structures of these Cr- and Mn-​doped TI layers to realize
Cr-doped TI Waals gap electrodynamic axion coupling that was previously
Mn-doped TI proposed for antiferromagnetic TIs27,64.

Bottom Antiferromagnetic TIs

surface
It was not until late 2018 that a stoichiometric material
Exchange field compound, MnBi2Te4, was predicted and experimentally
confirmed as an antiferromagnetic TI104. This study on
Mn Bi Te the topological properties of MnBi2Te4 (Fig. 2b) was soon
followed by a large amount of theoretical and experimen-
Fig. 2 | Material systems proposed to be axion insulators. a ∣ Crystal structure of tal work29,105,108–116. MnBi2Te4 is a layered ternary tetra­
pyrochlore iridate Y2Ir2O7 (Y = yttrium or lanthanide). The magnetic configuration is dymite compound first synthesized as a powder in 2013
all-​in/all-​out on the network of tetrahedrons with Ir corners. b ∣ Van der Waals layered (refs104,117). The crystal structure of MnBi2Te4 is shown
MnBi2Te4 with antiferromagnetic interlayer coupling and combined lattice translation in Fig. 2b and comprises blocks of septuple layers along
and time-​reversal symmetry (S = τ1 / 2T ). c ∣ Magnetic sandwich structure consisting the (0001)-​direction in the trigonal structure of space
of Mn- and Cr-​doped strong topological insulator (TI) layers. Panel a adapted with group R3m (no. 166). The layers are bound by van der
permission from ref.134, APS. Panel b adapted from ref.104, Springer Nature Limited.
Waals forces, and Mn moments couple antiferromagnet-
Panel c adapted with permission from ref.62, APS.
ically across layers. Within the Mn layers, the magnetic
moments show ferromagnetic order, such that an oper-
a magnetic order can be established in the material by ation, S, that combines a half-​layer translation τ1/2 and
means of a staggered Zeeman field. The axion field then time reversal T such that S = T τ1/2 realizes the necessary
couples directly to deviations of the magnetic order symmetry to classify the material as a Z 2 = 1 antiferro-
parameter, and collective excitations of the (anti-)ferro- magnetic insulator17. Both the bulk and the surface states
magnetic phase can change the otherwise constant axion were first reported to be gapped, with values of 220 meV
field by its effective coupling Δθ ∝ ΔMz. This coupling of and 88 meV, respectively104. The Néel temperature of
the axion field is nonlinear, except for the case of a static the compound was calculated to be about 25 K, below
and uniform magnetic field, in which there is a linear which the topological magnetoelectric effect should
coupling to the electric field. be observable. Subsequent experimental investigations
noted the difficulty of accessing gapped surface states, in
Heterostructures of TIs part owing to the strongly n-​type doping and observed
Dynamical axion states in magnetically doped TIs gapless Dirac cones in angle-​resolved photoemission
require long-​range magnetic order27. In contrast to studies115,118,119. It was later proposed to substitute Sb in
magnetically doping strong TIs in such a way that an the position of Bi to tailor its Fermi level to within the
antiferromagnetic order evolves, one can imagine stag- bulk and surface gap110. Irrespective of this controversy,
gering layers with different dopants that eventually cou- MnBi2Te4 was predicted to have a rich phase diagram in
ple antiferromagnetically across different layers. This all of its magnetic phases across 2D and 3D samples120
concept arises as a natural implementation of the idea and, further, to exhibit second-​harmonic generation and
of stacking TIs5 and was further developed in refs62–64, a shift current from infrared photons that is two orders
Shift current partly motivated by the lack of experimental verifica- of magnitude above typical shift currents obtained in
Second-​order optical effect
that results in a d.c. current
tion of the topological magnetoelectric effect up to then. non-​topological ferroelectrics121. Samples can be grown
from incident monochromatic The proposals are united in their idea of using sandwich by molecular-​beam epitaxy or by a self-​flux method to
light. structures with different ferromagnetic capping layers. obtain single crystals105,109. One of the main challenges

www.nature.com/natrevphys
 Reviews

during the growing procedure is the precise temperature Density-​f unctional calculations with self-​
control required122. In measurements, one has to finely interaction-​corrected local spin density approxima-
tune the chemical potential into the insulating surface tion show that the ground state of pyrochlore iridates
gap, and finite-​thickness effects have to be eliminated30. (such as Y2Ir2O7; Fig. 2a) is a 3D semimetal with Fermi
A number of variations of MnBi2Te 4 have been arc surface states and non-​collinear magnetic order134.
proposed as candidate materials for axion insulators On changing the Coulomb interaction parameter U,
and to realize dynamic axion fields. Among them are the material exhibits several phases, among them the
Mn2Bi2Te5, which realizes different surface states from axion insulator state, protected by inversion symmetry,
MnBi2Te4 (ref.120), and MnBi2nTe3n+1, which, depending and a Mott insulator state. This was further confirmed
on the surface termination, was shown to exhibit differ- in a theoretical study in a wider range of compounds135.
ent topological states106. Mn2Bi2Te5 and Eu2Bi2Te5 break By manipulating the surface termination and the mag-
both inversion and time-​reversal symmetry but preserve netic structure in a model system describing the poly­
their product and consequently show a non-​quantized θ chlore iridate lattice, a range of accessible topological
(refs77,123). However, unlike in MnBi2Te4, non-​static axion states have been found, including phases of higher-​order
states can be realized through antiferromagnetic reso- topology56. The axion insulator state in these compounds
nance, opening a path to observe dynamic axion effects77. is stabilized by inversion symmetry. In a similar spirit,
To realize specific stable magnetic configurations, several it was later predicted that osmium compounds such as
studies suggest heterostructures that contain MnBi2Te4 CaOs2O4 and SrOs2O4 can be stabilized in the geomet-
and related compounds124–127. In particular, the coercive rically frustrated spinel crystal structure and show an
fields of these structures are easier to control — both in equally colourful phase diagram when varying U, again
and by external fields — than doped TIs124. including the axion insulator state134,136,137. So far, both
An intrinsic antiferromagnetic insulator state has also structures have eluded experimental verification of the
been proposed to arise in EuIn2As2, which has a hexa­ axion phase.
gonal crystal structure with the space group P63/mmc With a Néel temperature of 9 K, GdPtBi was proposed
(no. 194)128. This material shows a smaller bulk insu- in 2010 as a possible candidate axion insulator17, but it
lating gap and a lower Néel temperature than MnBi2Te4 was verified to be a semimetal with antiferromagnetic
but is a candidate material to realize the axion insula- order in subsequent experimental studies138–140. However,
tor state in its two antiferromagnetic phases as well as a recent experimental work suggests that a negative mag-
higher-​order TI state. For antiferromagnetic order along netothermal resistivity in GdPtBi may be a signature of
the crystallographic c-​axis, one can construct a cylinder the mixed axial–gravitational anomaly141.
with all surfaces being gapped. An effective 1D behavi­
our emerges in the hinges between surface planes. A Experimental realizations
number of studies have followed this proposal and iden- The theoretical prediction of a half-​quantized quan-
tified the in-​plane magnetic configuration as the ground tum Hall effect quickly stimulated experimental efforts
state in EuIn2As2 (refs129,130). More recently, EuSn2As2 to observe this effect. The topological magnetoelectric
(ref.116) and EuSn2P2 (ref.131) have been shown to exhibit effect promises unprecedented coupling strengths even
comparable properties, extending the list of synthesized in inversion-​symmetric or time-​reversal-​invariant mate-
and characterized antiferromagnetic TIs even further. rials. To establish a platform for the axion insulator state,
To aid measurements of the zero Hall plateau and one of the main measurement objectives is to observe
quantum anomalous Hall effect, a stacking similar to the zero Hall plateau along with the quantum anoma-
the proposed sandwich structures depicted in Fig. 2 has lous Hall effect in typical six-​terminal devices and in a
been suggested but with antiferromagnetic MnBi2Se4 material where the magnetization gap dominates the
on one side of the strong TI132. In that way, the expected hybridization gap32,62 (Fig. 3a,b). As the individual sur-
response depends only on the magnetization of one faces are not individually accessible in traditional trans-
ferromagnetic layer. port experiments, measurement of the magnetoelectric
effect, which manifests itself as a bulk property, becomes
Charge-​density-​wave materials a main objective. Optical measurements can provide a
In Weyl semimetals, which host ungapped, chiral topo- direct and non-​contact route to the quantized response
logical states in the bulk of the material, a charge-​density of both bulk and surface states. The main experimental
wave can destroy the intrinsic chiral symmetry and con- challenge remains in the fact that these states need to be
vert the otherwise conducting material into a topologi- sufficiently gapped.
cally non-​trivial insulator78,133. The phase fluctuations of
this induced density wave couple directly to an axion-​like Optical measurements
field. This phase preserves time-​reversal symmetry, such Several groups have presented direct measurements
that no anomalous Hall effect is expected. We elaborate of the Kerr rotation in strong 3D TIs using either high
on the details of this mechanism and comment on recent external magnetic fields or magnetic impurities to break
experimental efforts in Ta2Se8I in the next section28. time-​reversal symmetry34,36,37. In one of these recent
experiments, Bi 2Se3 samples from molecular-​b eam
Other material compounds epitaxy were measured in a terahertz magneto-​optical
Several other materials have been predicted to be Faraday set-​up, and a half-​integer quantization index was
axion insulators but have so far eluded experimental observed34 (Fig. 3c,d. This work claims direct evidence
verification as such. for the optical response altered by the presence of the

Nature Reviews | Physics

 Reviews

a b c
Vxy P1
1.0
60 mK
Magnetically
doped TI 0.5
σxx

σxx, σyx (e2/h)

Current H P2
0.0
Substrate
–0.5 P3
Bi2Se3 on a
σxy Al2O3 substrate
Vxx
–1.0
Vgate –1.0 –0.5 0.0 0.5 1.0
B (T)

d e f
12 QL 0.4 6 2.0
8 μ0H = 0 T
αmeasured (mrad)

6
6.00 T 6.25 T 6.50 T 5 1.5
4
6.75 T 7.00 T 0.2
2 Accepted value 4 1.0
Energy (eV)

d ρyx/dH (kΩ T–1)

0
0.8

ρxx(h/e2)
0.3 0.4 0.5 0.6 0.7 Vg
Frequency (THz) 0 3 0.5
10 QL 30 V
8
αmeasured (mrad)

6 2 0
4 5.75 T 6.00 T 6.25 T –0.2 Axion
2
6.50 T 6.75 T 7.00 T 1 insulator –0.5
Accepted value
0
0.3 0.4 0.5 0.6 0.7 0.8 –0.4 0 –1.0
Frequency (THz) M Γ K 10 20 30 40 50
Vg (V)

Fig. 3 | Experimental set-ups to access topological responses. a ∣ Six-​ constants extrapolated from the induced rotation in samples of different thick-
terminal transport device used for Hall measurements. b ∣ Temperature-​ ness (number of quintuple layers (QLs)) and for varying external fields. e ∣ Bulk
dependent observation of the zero Hall plateau and quantum anomalous Hall band structure of MnBi2Te4 with antiferromagnetic ordering and chemical
effect at fixed gate voltage in sandwich structure of ferromagnetic Cr-doped potential with gate voltage 30 V. f ∣ Longitudinal resistivity and change in the
(Bi,Sb)2Te3, non-​magnetic (Bi,Sb)2Te3 and ferromagnetic V-​doped (Bi,Sb)2Te. Hall resistivity around zero field. TI, topological insulator. Panel b adapted with
c ∣ Terahertz polarization measurement set-​up for Bi2Se3 in high external mag- permission from ref.144, AAAS. Panels c and d adapted with permission from
netic fields. P1, P2, P3, polarizers. d ∣ Measurement results for the fine-​structure ref.34, AAAS. Panels e and f adapted from ref.147, Springer Nature Limited.

axion term, using time-​domain terahertz spectroscopy34. reduce carrier concentration as in ref.34. Control of
The experiment overcame a number of technical chal- the Fermi level by gel gating was achieved, which can
lenges, as the probe frequency has to be much lower provide access to both semiclassical and quantized
than the Landau-​level splitting induced by the external responses in terahertz spectroscopy35.
magnetic field that gaps the surface states. As metallic It is worth noting that whereas a quantized Faraday
gates would induce their own Faraday rotation, bulk rotation was reported in the aforementioned experi-
doping has to be accurately controlled (to eliminate ments, the observation of a half-​quantized surface Hall
the need for gating), which further complicates direct effect, or equivalently the topological magnetoelec­
measurement of the quantized response. By measuring tric effect, has not been achieved yet. Although the struc-
the induced change in polarization of both transmit- tures probed in these recent experiments do not qualify
ted and reflected fields, the fine-​structure constant was as axion insulators per se, and their response can be
extracted with an accuracy of 0.5%. This result does not explained in terms of the quantum Faraday effect, it can
depend on the particular substrate, as it does in opti- also be derived from axion electrodynamics with appro-
cal measurements of 2D electron gases or graphene in priate boundary conditions. Both effects can in princi-
the quantum Hall regime142,143. Thus, the direct obser- ple be realized in antiferromagnetic or ferromagnetically
vation of the quantized Faraday response paves the doped TIs as well54,72.
way towards a definition based purely on topological The proposed axion polariton has so far eluded
invariants in a solid-​state system for three fundamental observation through optical measurements. Together
physical constants (e, h and c)34,69. with the proposal of a cantilever torque magnetom-
Comparable results were presented in strained HgTe etry measurement, the two experiments of dynam-
3D TI structures36,37. Near-​quantization was achieved in ical properties proposed in ref.27 can be addressed in
ref.37 without the need for any external field. In ref.36, future experiments, given clean and large enough crys-
high magnetic fields and semi-​transparent gates were tal samples for both intrinsic and synthetic material
used such that interface doping was not necessary to structures.

www.nature.com/natrevphys
 Reviews

Transport measurements and this coupling yields a positive contribution to the

TI heterostructures. The axion insulator state leads to magnetoconductance. The contribution scales with
the topological magnetoelectric effect. To realize this the applied magnetic field and allows identification of the
state in a 3D heterostructure, the surfaces need to be axion character of the phason. In Fig. 4, we show the unit
gapped and the bulk needs to quantize such that θ = π cell of Ta2Se8I, the gapping mechanism mediated by the
(refs33,62). As shown in Fig. 1a,b, by switching the magnet- formation of a charge-​density wave, and the identifying
ization configuration in such a layered structure, one can measurement of the magnetoresistance. This mechanism
observe both the quantum anomalous Hall regime and should be observable in any Fermi-​surface Weyl semi­
the zero Hall plateau, which provides the moti­vation for metal that exhibits a charge-​density wave with coupl­
further investigation63. Two of the first studies to realize ing commensurate to the spacing of the Weyl points in
the axion insulator state accessed by magneto­transport momentum space. It is worth noting that whether or not
measurements use molecular-​beam epitaxy to grow Ta2Se8I qualifies as an axion insulator is a subject of ongo-
trilayers (Bi,Sb)2Te3 with V-​doping of the top and Cr-​ ing debate133. It does, however, realize a dynamical axion
doping of the bottom layer33,144. For the parallel magneti- field through the chiral anomaly and its charge-​density
zation configuration, the quantum anomalous Hall effect wave. Whether it shows non-​trivial magnetoelectric
is observed, whereas in the antiparallel orientation, the response needs to be established by further studies.
existence of a zero Hall plateau was verified. These stud-
ies move closer to measuring the axion insulator state Applications
than previous work that identified plateaus of zero Hall In addition to the striking quantized responses emerging
conductance32,145, but, in the regime where the hybridi- from the axion coupling and the possibility of construct-
zation gap dominates the magnetization gap, the identi- ing structures with enormous linear magnetoelectric coef-
fication of an antiparallel magnetization configuration ficients, several practical applications have recently been
remains unclear146. proposed for axion insulators. Among them, and cer-
tainly one of the most intriguing possibilities, is the idea
Antiferromagnetic TIs. Antiferromagnetic TIs provide a of using antiferromagnetic TIs and their axion polaritons
platform to measure the quantum anomalous Hall effect to detect the dark-​matter axion by low-​noise terahertz
in stoichiometric compounds. Most of the early studies techniques149. Dark-​matter axions are hypothetical parti-
on MnBi2Te4 focused on characterization of bulk and sur- cles that could account for most or all of the dark matter in
face states by angle-​resolved photoemission104,106,118,120,122 the Universe. One proposal is a device made of Fe-​doped
and exploring its rich magnetic and topological phase Bi2Se3, in which dark-​matter axions can effectively drive
diagram. More recently, an axion-​to-​Chern-​insulator the condensed-​matter axion polariton by electromag-
transition was observed at moderate magnetic fields in netic coupling. As the axion polariton dispersion in the
samples with an even number of layers. This result is antiferromagnetic TI is sensitive to the external magnetic
shown in Fig. 3e,f, together with the dispersion of the field, axion masses between 0.7 meV and 3.5 meV can be
MnBi2Te4 surface states147. In MnBi2Te4 flakes with an scanned, which would fill a gap in the range of detectable
odd number of layers, where the surface layers show fundamental axion energies150. This theoretical proposal
antiparallel magnetization alignment, the quantum assumes relatively large crystals (1 cm3 of antiferromag-
anomalous Hall effect was observed at zero field, and the netically doped (Bi1−xFex)2Se3), which need to exhibit
Hall resistance exhibited finite jumps when individual gapped bandstructures at both bulk and surface, and so
magnetic layers switched due to an external field111. the implementation of such a detector provides ample
challenges to device design and material synthesis.
Charge-​density waves. The first measurement of a Axion electrodynamics can also contribute to the
dynamic axion realization was reported in 2019 and used asymmetry of the dielectric tensor of a Weyl semimetal
transport measurements in quasi-1D Ta2Se8I (ref.28). This via the chiral anomaly (Box 3), given the emergent axion
compound is a Weyl semimetal at room temperature and field’s spatiotemporal nature. If a Weyl semimetal has
opens a bulk gap at lower temperatures owing to the for- broken time-​reversal symmetry and nodes separated in
mation of a charge-​density wave148. A peculiar detail in momentum space, its surface plasmon polariton modes
this system is that the charge-​density-​wave nesting vector can be strongly non-​reciprocal151. Thus, these materi-
coincides with the vectors connecting the 24 Fermi-​ als provide an efficient way to construct non-​reciprocal
surface Weyl points. The phase of this charge-​density thermal emitters. A new method to control thermal
wave can be identified as a contribution to the axion radiation is by exploiting the temperature-​tunable non-
field, and the phason is the axionic mode. The effective reciprocity of radiation in magnetic Weyl semimetals152.
action describing the electrodynamics in this compound This technique makes it possible generally to overcome
shows analogies to that of the pion–photon coupling in accepted photovoltaic performance limits153. The pro-
high-​energy physics78. Ta2Se8I further shows the highest posal assumes a perfect Weyl semimetal with the mini-
net chiral charge (the sum of Weyl node chiralities below mum number of two Weyl nodes at the Fermi level. Such
the Fermi level) reported in a Weyl semimetal so far, up states have already been observed in EuCd2As2, such
to +16. The effect of the axion mode is visible in mag- that future measurements could be used to assess the
Nesting vector netotransport measurements. By applying a magnetic feasibility of such a device154.
Vector connecting pockets of
the Fermi surface, typically
field parallel to the depinning electric field that permits More exotic proposals range from superconducting
related to the formation of a the charge-​density wave to slide, the axion coupling phases that can couple to the topological axion field155
density wave. to the phase of the charge-​density wave is established, to the realization of Majorana fermions at the interface

Nature Reviews | Physics

Reviews

a b –x +x

EF ΔE EF
kq kq
E E

kx kx

CDW

n n
I Se Ta
x a x π/kq

c d e 80 K
B||I 0.6 B||I
30 80 K 9V
0.5 12V
15 V
18 V

Δ(dI/dV) (mS)
0.4
21 V
0T

V (V)
0 24 V
1.5 T 0.3
27 V
3.0 T
4.5 T 0.2
6.0 T
7.5 T 0.1
∝ B2
1 mm –30 9.0 T
0.0

–2 0 2 –9 –6 –3 0 3 6 9
I (mA) B (T)

Fig. 4 | signature of axionic charge-density wave in Ta2se8i. a ∣ Crystal structure of Ta2Se8I with quasi-1D character.
b ∣ Simplified picture of a charge-​density wave (CDW) with a wave vector kq commensurate with the spacing of the Weyl
points, which gaps the states at the Fermi level EF. The CDW appears in the direction of the Ta-​chain (x). c ∣ Grown crystals
of Ta2Se8I. d ∣ Resulting voltage–current (V–I) curve for magnetic fields parallel to depinning field from sample shown in
panel c. e ∣ Quadratic dependence of the magnetoconductance on the external field for various voltages. Adapted from
ref.28, Springer Nature Limited.

between quantum anomalous Hall materials, together with the demand for clean crystal samples and ultra-​high-​
its prospects for topological quantum computing15,156,157. precision measurements. On the other hand, calcu-
lations of the topological band structure and their
Outlook response will be put to a test in highly correlated materi-
Just as quantum Hall insulators were observed within als that undergo phase transitions, where computational
a decade after the initial theoretical prediction, axion approaches based on the single-​particle picture fail.
states are now accessible in a number of condensed-​ Topological antiferromagnets that have sizable gaps
matter systems, pushing theory, material synthesis and and transition temperatures at elevated temperatures
experimental characterization to their limits. are yet to be realized. A direct observation of the top-
Although the axion particle in high-​energy physics ological magnetoelectric effect, for example, in optical
still eludes observation, its low-​energy condensed-​matter experiments that are sensitive to a single surface, or
counterpart can be realized in various ways. Part of this realizations of the dynamical axion field in antiferro-
success comes from the topological field theory of the magnetic TIs is still missing. Bringing about the first
axion insulator allowing for a classification of 3D insu- technologically relevant application of a 3D TI that
lators with and without time-​reversal symmetry. Apart harnesses giant quantized responses, this is a challenge
from its quantized coupling to an external field, the that would be overcome by further material searches,
axion parameter can take continuous values in semime- synthesis and careful experiments158. Many practical
tallic and non-​symmetric systems and thus contributes applications in quantum information science and spin-
significantly to the linear magnetoelectric response. tronics can be enabled from the transport behaviour
The magnetoelectric coupling by the axion term arises, exhibited13,15,159, and fundamental science can also greatly
although with lower magnitude, even in topologically benefit from the prospects of detecting the hypothetical
trivial materials, such that its calculation is necessary to dark-​matter axion through realization of a new type of
accurately predict their response. condensed-​matter detector. The detection of such a dark
So far, experimental discoveries have largely fol- axion by means of the condensed-​matter counterpart
lowed theoretical predictions in this field, in contrast would elegantly close a circle.
with developments in the field of superconductivity.
Excitingly, first principles predictions have fuelled Published online xx xx xxxx

www.nature.com/natrevphys
 Reviews

1. Svrcek, P. & Witten, E. Axions in string theory. J. High 30. Zhang, D. et al. Topological axion states in magnetic 60. Bernevig, B. A. & Hughes, T. L.Topological Insulators
Energy Phys. 2006, 051–051 (2006). insulator MnBi2Te4 with the quantized magnetoelectric and Topological Superconductors (Princeton Univ.
2. Peccei, R. D. & Quinn, H. R. CP conservation in the effect. Phys. Rev. Lett. 122, 206401 (2019). Press, 2013).
presence of pseudoparticles. Phys. Rev. Lett. 38, 31. Feng, Y. et al. Observation of the zero Hall plateau in 61. Nomura, K. & Nagaosa, N. Surface-​quantized
1440–1443 (1977). a quantum anomalous Hall insulator. Phys. Rev. Lett. anomalous Hall current and the magnetoelectric effect
3. Preskill, J., Wise, M. B. & Wilczek, F. Cosmology of 115, 126801 (2015). in magnetically disordered topological insulators.
the invisible axion. Phys. Lett. B 120, 127–132 32. Mogi, M. et al. A magnetic heterostructure of Phys. Rev. Lett. 106, 166802 (2011).
(1983). topological insulators as a candidate for an axion 62. Morimoto, T., Furusaki, A. & Nagaosa, N.
4. Wilczek, F. Two applications of axion electrodynamics. insulator. Nat. Mater. 16, 516–521 (2017). Topological magnetoelectric effects in thin films
Phys. Rev. Lett. 58, 1799–1802 (1987). 33. Xiao, D. et al. Realization of the axion insulator state of topological insulators. Phys. Rev. B 92, 085113
5. Qi, X.-L., Hughes, T. L. & Zhang, S.-C. Topological field in quantum anomalous Hall sandwich heterostructures. (2015).
theory of time-​reversal invariant insulators. Phys. Rev. B Phys. Rev. Lett. 120, 056801 (2018). 63. Wang, J., Lian, B., Qi, X.-L. & Zhang, S.-C. Quantized
78, 195424 (2008). 34. Wu, L. et al. Quantized Faraday and Kerr rotation and topological magnetoelectric effect of the zero-​plateau
Unified topological Chern–Simons field theory in axion electrodynamics of a 3D topological insulator. quantum anomalous Hall state. Phys. Rev. B 92,
phase space for time-​reversal invariant insulators. Science 354, 1124–1127 (2016). 081107 (2015).
6. Essin, A. M., Moore, J. E. & Vanderbilt, D. 35. Mondal, M. et al. Electric field modulated topological 64. Wang, J., Lian, B. & Zhang, S.-C. Dynamical axion
Magnetoelectric polarizability and axion magnetoelectric effect in Bi2Se3. Phys. Rev. B 98, field in a magnetic topological insulator superlattice.
electrodynamics in crystalline insulators. Phys. Rev. 121106 (2018). Phys. Rev. B 93, 045115 (2016).
Lett. 102, 146805 (2009). 36. Dziom, V. et al. Observation of the universal 65. Rivera, J.-P. A short review of the magnetoelectric
7. Grushin, A. G. Consequences of a condensed matter magnetoelectric effect in a 3D topological insulator. effect and related experimental techniques on single
realization of Lorentz-​violating QED in Weyl semi-metals. Nat. Commun. 8, 15197 (2017). phase (multi-) ferroics. Eur. Phys. J. B 71, 299
Phys. Rev. D 86, 045001 (2012). 37. Okada, K. N. et al. Terahertz spectroscopy on Faraday (2009).
8. Zyuzin, A. A. & Burkov, A. A. Topological response in and Kerr rotations in a quantum anomalous Hall state. 66. fiebig, M. & Spaldin, N. A. Current trends of the
Weyl semimetals and the chiral anomaly. Phys. Rev. B Nat. Commun. 7, 12245 (2016). magnetoelectric effect. Eur. Phys. J. B 71, 293
86, 115133 (2012). 38. Peccei, R. D. & Quinn, H. R. Constraints imposed by (2009).
9. Hasan, M. Z. & Kane, C. L. Colloquium: Topological CP conservation in the presence of pseudoparticles. 67. Coh, S., Vanderbilt, D., Malashevich, A. & Souza, I.
insulators. Rev. Mod. Phys. 82, 3045–3067 (2010). Phys. Rev. D 16, 1791–1797 (1977). Chern–Simons orbital magnetoelectric coupling
10. Qi, X.-L. & Zhang, S.-C. Topological insulators and 39. Klitzing, K. V., Dorda, G. & Pepper, M. New method in generic insulators. Phys. Rev. B 83, 085108
superconductors. Rev. Mod. Phys. 83, 1057–1110 for high-​accuracy determination of the fine-​structure (2011).
(2011). constant based on quantized Hall resistance. Numerical calculation of the axion field in
11. Vergniory, M. et al. A complete catalogue of high-​ Phys. Rev. Lett. 45, 494–497 (1980). topologically trivial and non-​trivial materials,
quality topological materials. Nature 566, 480–485 40. Thouless, D. J., Kohmoto, M., Nightingale, M. P. based on maximally localized Wannier functions.
(2019). & den Nijs, M. Quantized Hall conductance in a 68. Karch, A. Electric–magnetic duality and topological
12. Ando, Y. Topological insulator materials. J. Phys. two-dimensional periodic potential. Phys. Rev. Lett. insulators. Phys. Rev. Lett. 103, 171601 (2009).
Soc. Jpn 82, 102001 (2013). 49, 405 (1982). 69. Maciejko, J., Qi, X.-L., Drew, H. D. & Zhang, S.-C.
13. Šmejkal, L., Mokrousov, Y., Yan, B. & MacDonald, A. H. 41. Bernevig, B. A., Hughes, T. L. & Zhang, S.-C. Quantum Topological quantization in units of the fine structure
Topological antiferromagnetic spintronics. Nat. Phys. spin Hall effect and topological phase transition in constant. Phys. Rev. Lett. 105, 166803 (2010).
14, 242 (2018). HgTe quantum wells. Science 314, 1757–1761 70. Tse, W.-K. & MacDonald, A. H. Giant magneto-​optical
14. Moore, J. E. & Balents, L. Topological invariants of (2006). Kerr effect and universal Faraday effect in thin-​film
time-​reversal-invariant band structures. Phys. Rev. B 42. König, M. et al. Quantum spin Hall insulator state in topological insulators. Phys. Rev. Lett. 105, 057401
75, 121306 (2007). HgTe quantum wells. Science 318, 766–770 (2007). (2010).
15. Stern, A. & Lindner, N. H. Topological quantum 43. Mehta, M. L. Random Matrices (Elsevier, 2004). 71. Feng, W. et al. Topological magneto-​optical effects and
computation — from basic concepts to first 44. Fu, L. & Kane, C. L. Time reversal polarization and a Z2 their quantization in noncoplanar antiferromagnets.
experiments. Science 339, 1179–1184 (2013). adiabatic spin pump. Phys. Rev. B 74, 195312 Nat. Commun. 11, 1–9 (2020).
16. Kane, C. L. & Mele, E. J. Z2 topological order and the (2006). 72. Beenakker, C. Topological magnetoelectric effect versus
quantum spin Hall effect. Phys. Rev. Lett. 95, 146802 45. Fu, L. & Kane, C. L. Topological insulators with quantum Faraday effect. J. Club Condens. Matter Phys.
(2005). inversion symmetry. Phys. Rev. B 76, 045302 (2007). https://doi.org/10.36471/JCCM_April_2016_01
17. Mong, R. S. K., Essin, A. M. & Moore, J. E. 46. Soluyanov, A. A. & Vanderbilt, D. Computing (2016).
Antiferromagnetic topological insulators. Phys. Rev. B topological invariants without inversion symmetry. 73. Qi, X.-L., Li, R., Zang, J. & Zhang, S.-C. Inducing a
81, 245209 (2010). Phys. Rev. B 83, 235401 (2011). magnetic monopole with topological surface states.
18. Zirnstein, H.-G. & Rosenow, B. Topological 47. Yu, R., Qi, X. L., Bernevig, A., Fang, Z. & Dai, X. Science 323, 1184–1187 (2009).
magnetoelectric effect: nonlinear time-​reversal- Equivalent expression of Z2 topological invariant 74. König, E. J. et al. Half-​integer quantum Hall effect of
symmetric response, Witten effect, and half-​integer for band insulators using the non-​Abelian Berry disordered Dirac fermions at a topological insulator
quantum Hall effect. Phys. Stat. Solidi B 257, connection. Phys. Rev. B 84, 075119 (2011). surface. Phys. Rev. B 90, 165435 (2014).
1900698 (2020). 48. Ringel, Z., Kraus, Y. E. & Stern, A. Strong side of weak 75. Taguchi, K. et al. Electromagnetic effects induced by a
19. Essin, A. M., Turner, A. M., Moore, J. E. & Vanderbilt, D. topological insulators. Phys. Rev. B 86, 045102 time-​dependent axion field. Phys. Rev. B 97, 214409
Orbital magnetoelectric coupling in band insulators. (2012). (2018).
Phys. Rev. B 81, 205104 (2010). 49. Mong, R. S. K., Bardarson, J. H. & Moore, J. E. 76. Ooguri, H. & Oshikawa, M. Instability in magnetic
20. Turner, A. M., Zhang, Y., Mong, R. S. K. & Vishwanath, A. Quantum transport and two-​parameter scaling at the materials with a dynamical axion field. Phys. Rev. Lett.
Quantized response and topology of magnetic insulators surface of a weak topological insulator. Phys. Rev. Lett. 108, 161803 (2012).
with inversion symmetry. Phys. Rev. B 85, 165120 108, 076804 (2012). 77. Zhang, J. et al. Large dynamical axion field in
(2012). 50. Hsieh, D. et al. A topological Dirac insulator in a topological antiferromagnetic insulator Mn2Bi2Te5.
21. Wilczek, F. Particle physics and condensed matter: quantum spin Hall phase. Nature 452, 970–974 Chin. Phys. Lett. 37, 077304 (2020).
the saga continues. Phys. Scr. T168, 014003 (2008). 78. Wang, Z. & Zhang, S.-C. Chiral anomaly, charge
(2016). 51. Xia, Y. et al. Observation of a large-​gap topological-​ density waves, and axion strings from Weyl
22. Nielsen, H. & Ninomiya, M. The Adler–Bell–Jackiw insulator class with a single Dirac cone on the surface. semimetals. Phys. Rev. B 87, 161107 (2013).
anomaly and Weyl fermions in a crystal. Phys. Lett. B Nat. Phys. 5, 398–402 (2009). 79. Casimir, H. B. G. On the attraction between two
130, 389–396 (1983). 52. Bansil, A., Lin, H. & Das, T. Colloquium: Topological perfectly conducting plates. Proc. K. Ned. Akad. Wet.
23. Vazifeh, M. M. & Franz, M. Electromagnetic response band theory. Rev. Mod. Phys. 88, 021004 (2016). 51, 793 (1948).
of Weyl semimetals. Phys. Rev. Lett. 111, 027201 53. Malashevich, A., Souza, I., Coh, S. & Vanderbilt, D. 80. Lifshitz, E. M. et al. in Perspectives in Theoretical
(2013). Theory of orbital magnetoelectric response. Physics 329–349 (Elsevier, 1992).
24. Armitage, N. P., Mele, E. J. & Vishwanath, A. Weyl and New Journal of Physics 12, 053032 (2010). 81. Dzyaloshinskii, I. E., Lifshitz, E. M. & Pitaevskii, L. P.
Dirac semimetals in three-​dimensional solids. 54. Armitage, N. P. & Wu, L. On the matter of topological The general theory of van der Waals forces. Adv. Phys.
Rev. Mod. Phys. 90, 015001 (2018). insulators as magnetoelectrics. SciPost Phys. 6, 46 10, 165–209 (1961).
25. Ilan, R., Grushin, A. G. & Pikulin, D. I. Pseudo-​ (2019). 82. Rivera, N., flick, J. & Narang, P. Variational theory of
electromagnetic fields in 3D topological semimetals. Review of magnetoelectric responses and their nonrelativistic quantum electrodynamics. Phys. Rev.
Nat. Rev. Phys. 2, 29–41 (2019). connection to the effective polarization. Lett. 122, 193603 (2019).
26. Burkov, A. A. Weyl metals. Annu. Rev. Condens. Matter 55. Vanderbilt, D. Berry Phases in Electronic Structure 83. Capasso, F., Munday, J. N., Iannuzzi, D. & Chan, H. B.
Phys 9, 359–378 (2018). Theory: Electric Polarization, Orbital Magnetization Casimir forces and quantum electrodynamical torques:
27. Li, R., Wang, J., Qi, X.-L. & Zhang, S.-C. Dynamical and Topological Insulators (Cambridge Univ. Press, physics and nanomechanics. IEEE J. Sel. Top. Quantum
axion field in topological magnetic insulators. 2018). Electron. 13, 400–414 (2007).
Nat. Phys. 6, 284–288 (2010). 56. Varnava, N. & Vanderbilt, D. Surfaces of axion 84. Grushin, A. G. & Cortijo, A. Tunable Casimir repulsion
Theory and predictions on the dynamical axion insulators. Phys. Rev. B 98, 245117 (2018). with three-​dimensional topological insulators.
field in magnetically doped TIs. 57. Wang, Z., Qi, X.-L. & Zhang, S.-C. Equivalent Phys. Rev. Lett. 106, 020403 (2011).
28. Gooth, J. et al. Axionic charge-​density wave in the topological invariants of topological insulators. 85. Grushin, A. G., Rodriguez-​Lopez, P. & Cortijo, A.
Weyl semimetal (TaSe4)2I. Nature 575, 315 (2019). New J. Phys. 12, 065007 (2010). Effect of finite temperature and uniaxial anisotropy
Magnetoresistance experiments in axionic 58. Fang, C., Gilbert, M. J. & Bernevig, B. A. Topological on the Casimir effect with three-​dimensional
charge-density-wave material Ta2Se8I. insulators with commensurate antiferromagnetism. topological insulators. Phys. Rev. B 84, 045119
29. Liu, C. et al. Robust axion insulator and Chern insulator Phys. Rev. B 88, 085406 (2013). (2011).
phases in a two-​dimensional antiferromagnetic 59. Tokura, Y., Yasuda, K. & Tsukazaki, A. Magnetic 86. Rodriguez-​Lopez, P. Casimir repulsion between
topological insulator. Nat. Mater. 19, 522–527 topological insulators. Nat. Rev. Phys. 1, 126–143 topological insulators in the diluted regime. Phys. Rev. B
(2020). (2019). 84, 165409 (2011).

Nature Reviews | Physics

Reviews

87. Chen, L. & Wan, S. Casimir interaction between 115. Hao, Y.-J. et al. Gapless surface Dirac cone in 143. Shimano, R. et al. Quantum Faraday and Kerr
topological insulators with finite surface band gap. antiferromagnetic topological insulator MnBi2Te4. rotations in graphene. Nat. Commun. 4, 1–6 (2013).
Phys. Rev. B 84, 075149 (2011). Phys. Rev. X 9, 041038 (2019). 144. Mogi, M. et al. Tailoring tricolor structure of magnetic
88. Chen, L. & Wan, S. Critical surface band gap of repulsive 116. Li, H. et al. Dirac surface states in intrinsic magnetic topological insulator for robust axion insulator.
Casimir interaction between three-​dimensional topological insulators EuSn2As2 and MnBi2nTe3n+1. Sci. Adv. 3, eaao1669 (2017).
topological insulators at finite temperature. Phys. Rev. B Phys. Rev. X 9, 041039 (2019). 145. Grauer, S. et al. Scaling of the quantum anomalous
85, 115102 (2012). 117. Lee, D. S. et al. Crystal structure, properties and Hall effect as an indicator of axion electrodynamics.
89. Nie, W., Zeng, R., Lan, Y. & Zhu, S. Casimir force nanostructuring of a new layered chalcogenide Phys. Rev. Lett. 118, 246801 (2017).
between topological insulator slabs. Phys. Rev. B 88, semiconductor, Bi2MnTe4. CrystEngComm 15, 146. Lachman, E. O. et al. Observation of
085421 (2013). 5532–5538 (2013). superparamagnetism in coexistence with quantum
90. Zeng, R. et al. Enhancing Casimir repulsion via 118. Swatek, P. et al. Gapless Dirac surface states in the anomalous Hall C = ±1 and C = 0 Chern states.
topological insulator multilayers. Phys. Lett. A 380, antiferromagnetic topological insulator MnBi2Te4. npj Quantum Mater. 2, 1–7 (2017).
2861–2869 (2016). Phys. Rev. B 101, 161109 (2020). 147. Liu, C. et al. Robust axion insulator and
91. Rodriguez-​Lopez, P. & Grushin, A. G. Repulsive 119. Li, H. et al. Dirac surface states in intrinsic magnetic Chern insulator phases in a two-​dimensional
Casimir effect with Chern insulators. Phys. Rev. Lett. topological insulators EuSn2As2 and MnBi2nTe3n+1. antiferromagnetic topological insulator.Nat. Mater.
112, 056804 (2014). Phys. Rev. X 9, 041039 (2019). 19, 522–527 (2020).
92. Wilson, J. H., Allocca, A. A. & Galitski, V. Repulsive 120. Li, J. et al. Intrinsic magnetic topological insulators 148. Zhang, Y., Lin, L.-F., Moreo, A., Dong, S. & Dagotto, E.
Casimir force between Weyl semimetals. Phys. Rev. B in van der Waals layered MnBi2Te4-family materials. First-​principles study of the low-​temperature charge
91, 235115 (2015). Sci. Adv. 5, eaaw5685 (2019). density wave phase in the quasi-​one-dimensional Weyl
93. Rodriguez-​Lopez, P., Popescu, A., Fialkovsky, I., 121. Fei, R., Song, W. & Yang, L. Giant photogalvanic chiral compound (TaSe4)2I. Phys. Rev. B 101, 174106
Khusnutdinov, N. & Woods, L. M. Signatures of effect and second-​harmonic generation in magnetic (2020).
complex optical response in Casimir interactions axion insulators. Phys. Rev. B 102, 035440 (2020). 149. Marsh, D. J. E., Fong, K. C., Lentz, E. W., Šmejkal, L.
of type I and II Weyl semimetals. Commun. Mater. 1, 122. Wu, J. et al. Natural van der Waals heterostructural & Ali, M. N. Proposal to detect dark matter using
14 (2020). single crystals with both magnetic and topological axionic topological antiferromagnets. Phys. Rev. Lett.
94. Woods, L. M. et al. Materials perspective on Casimir properties. Sci. Adv. 5, eaax9989 (2019). 123, 121601 (2019).
and van der Waals interactions. Rev. Mod. Phys. 88, 123. Li, Y. et al. Layer-​magnetization-tuned topological Proposal for a dark-​matter axion detector based
045003 (2016). phases in Mn2Bi2Te5 films. Preprint at arXiv https:// on an antiferromagnetic topological insulator.
95. fialkovsky, I., Khusnutdinov, N. & Vassilevich, D. arxiv.org/abs/2001.06133 (2020). 150. Ringwald, A. Exploring the role of axions and other
Quest for Casimir repulsion between Chern–Simons 124. Hou, Y. S., Kim, J. W. & Wu, R. Q. Axion insulator state wisps in the dark Universe. Phys. Dark Universe 1,
surfaces. Phys. Rev. B 97, 165432 (2018). in ferromagnetically ordered Cri3/Bi2Se3/MnBi2Se4 116–135 (2012).
96. Vassilevich, D. On the (im)possibility of Casimir heterostructures. Phys. Rev. B 101, 121401 (2020). 151. Hofmann, J. & Sarma, S. D. Surface plasmon polaritons
repulsion between Chern–Simons surfaces. Mod. 125. Li, Z. et al. Tunable interlayer magnetism and band in topological Weyl semimetals. Phys. Rev. B 93,
Phys. Lett. A 35, 2040017 (2020). topology in van der Waals heterostructures of 241402 (2016).
97. Martín-​Ruiz, A., Cambiaso, M. & Urrutia, L. F. MnBi2Te4-family materials. Phys. Rev. B 102, 081107 152. Zhao, B., Guo, C., Garcia, C. A. C., Narang, P. &
A Green’s function approach to the Casimir effect (2020). Fan, S. Axion-​field-enabled nonreciprocal thermal
on topological insulators with planar symmetry. 126. Wang, H. et al. Dynamical axion state with hidden radiation in Weyl semimetals. Nano Lett. 20,
Europhys. Lett. 113, 60005 (2016). pseudospin Chern numbers in MnBi2Te4-based 1923–1927 (2020).
98. Fukushima, K., Imaki, S. & Qiu, Z. Anomalous Casimir heterostructures. Phys. Rev. B 101, 081109 (2020). Exploitation of the axion field in an ideal Weyl
effect in axion electrodynamics. Phys. Rev. D 100, 127. Fu, H., Liu, C.-X. & Yan, B. Exchange bias and semimetal for non-​reciprocal thermal emitters.
045013 (2019). quantum anomalous Hall effect in the MnBi2Te4/CrI3 153. Green, M. A. Time-​asymmetric photovoltaics. Nano Lett.
99. Hehl, F. W., Obukhov, Y. N., Rivera, J.-P. & Schmid, H. heterostructure. Sci. Adv. 6, eaaz0948 (2020). 12, 5985–5988 (2012).
Magnetoelectric Cr2O3 and relativity theory. Eur. Phys. 128. Xu, Y., Song, Z., Wang, Z., Weng, H. & Dai, X. 154. Soh, J.-R. et al. Ideal Weyl semimetal induced by
J. B 71, 321–329 (2009). Higher-order topology of the axion insulator EuIn2As2. magnetic exchange. Phys. Rev. B 100, 201102
100. Kurumaji, T. et al. Optical magnetoelectric resonance Phys. Rev. Lett. 122, 256402 (2019). (2019).
in a polar magnet (Fe,Zn)2Mo3O8 with axion-​type 129. Zhang, Y. et al. In-​plane antiferromagnetic moments 155. Qi, X.-L., Witten, E. & Zhang, S.-C. Axion topological
coupling. Phys. Rev. Lett. 119, 077206 (2017). and magnetic polaron in the axion topological field theory of topological superconductors. Phys. Rev. B
101. Varnava, N., Souza, I. & Vanderbilt, D. Axion coupling insulator candidate EuIn2As2. Phys. Rev. B 101, 87, 134519 (2013).
in the hybrid Wannier representation. Phys. Rev. B 205126 (2020). 156. Chen, C.-Z., Xie, Y.-M., Liu, J., Lee, P. A. & Law, K. T.
101, 155130 (2020). 130. Regmi, S. et al. Temperature dependent electronic Quasi-​one-dimensional quantum anomalous Hall
102. Maciejko, J., Qi, X.-L., Karch, A. & Zhang, S.-C. structure in a higher order topological insulator systems as new platforms for scalable topological
Fractional topological insulators in three dimensions. candidate EuIn2As2. Preprint at arXiv https://arxiv.org/ quantum computation. Phys. Rev. B 97, 104504
Phys. Rev. Lett. 105, 246809 (2010). abs/1911.03703 (2019). (2018).
103. Shi, W. et al. A charge-​density-wave Weyl semimetal. 131. Gui, X. et al. A new magnetic topological quantum 157. Lian, B., Sun, X.-Q., Vaezi, A., Qi, X.-L. & Zhang, S.-C.
Preprint at arXiv https://arxiv.org/abs/1909.04037 material candidate by design. ACS Central Sci. 5, Topological quantum computation based on chiral
(2019). 900–910 (2019). Majorana fermions. Proc. Natl Acad. Sci. USA 115,
104. Otrokov, M. M. et al. Prediction and observation of an 132. Hou, Y. & Wu, R. Axion insulator state in a 10938–10942 (2018).
antiferromagnetic topological insulator. Nature 576, ferromagnet/topological insulator/antiferromagnet 158. Burkov, A. A. Topological semimetals. Nat. Mater. 15,
416–422 (2019). heterostructure. Nano Lett. 19, 2472–2477 1145–1148 (2016).
Combined theoretical and experimental study on (2019). 159. Keimer, B. & Moore, J. E. The physics of quantum
antiferromagnetic topological insulator MnBi2Te4. 133. Wieder, B. J., Lin, K.-S. & Bradlyn, B. Is the dynamical materials. Nat. Phys. 13, 1045–1055 (2017).
105. Gong, Y. et al. Experimental realization of an intrinsic axion Weyl-​charge-density wave an axionic band 160. Altland, A. & Zirnbauer, M. R. Nonstandard symmetry
magnetic topological insulator. Chin. Phys. Lett. 36, insulator? Preprint at arXiv https://arxiv.org/abs/ classes in mesoscopic normal–superconducting
076801 (2018). 2004.11401 (2020). hybrid structures. Phys. Rev. B 55, 1142–1161
106. Hu, C. et al. A van der Waals antiferromagnetic 134. Wan, X., Turner, A., Vishwanath, A. & Savrasov, S. Y. (1997).
topological insulator with weak interlayer magnetic Electronic structure of pyrochlore iridates: from 161. Kitaev, A. Periodic table for topological insulators
coupling. Nat. Commun. 11, 97 (2020). topological Dirac metal to Mott insulator. Phys. Rev. B and superconductors. AIP Conf. Proc. 1134, 22–30
107. Lv, B., Qian, T. & Ding, H. Angle-​resolved 83, 205101 (2011). (2009).
photoemission spectroscopy and its application to 135. Chen, G. & Hermele, M. Magnetic orders and 162. Chern, S.-S. & Simons, J. Characteristic forms
topological materials. Nat. Rev. Phys. 1, 609–626 topological phases from f–d exchange in pyrochlore and geometric invariants. Ann. Math. 99, 48–69
(2019). iridates. Phys. Rev. B 86, 235129 (2012). (1974).
108. Lee, S. H. et al. Spin scattering and noncollinear spin 136. Yamaura, J. et al. Tetrahedral magnetic order and 163. Witten, E. Topological quantum field theory. Commun.
structure-​induced intrinsic anomalous Hall effect in the metal–insulator transition in the pyrochlore Math. Phys. 117, 353–386 (1988).
antiferromagnetic topological insulator MnBi2Te4. lattice of Cd2Os2O7. Phys. Rev. Lett. 108, 247205 164. Hsiang, W.-Y. & Lee, D.-H. Chern–Simons invariant in
Phys. Rev. Res. 1, 012011 (2019). (2012). the Berry phase of a 2x2 Hamiltonian. Phys. Rev. A
109. Yan, J.-Q. et al. Crystal growth and magnetic structure 137. Shi, Y. G. et al. Continuous metal–insulator transition 64, 052101 (2001).
of MnBi2Te4. Phys. Rev. Mater. 3, 064202 (2019). of the antiferromagnetic perovskite NaOsO3. Phys. 165. Chiu, C.-K., Teo, J. C. Y., Schnyder, A. P. & Ryu, S.
110. Chen, B. et al. Intrinsic magnetic topological insulator Rev. B 80, 161104 (2009). Classification of topological quantum matter with
phases in the Sb doped MnBi2Te4 bulks and thin 138. Liu, C. et al. Metallic surface electronic state in half-​ symmetries. Rev. Mod. Phys. 88, 035005 (2016).
flakes. Nat. Commun. 10, 4469 (2019). Heusler compounds RPtBi (R = Lu, Dy, Gd). Phys. Rev. B 166. Adler, S. L. Axial-​vector vertex in spinor
111. Deng, Y. et al. Quantum anomalous Hall effect in 83, 205133 (2011). electrodynamics. Phys. Rev. 177, 2426–2438
intrinsic magnetic topological insulator MnBi2Te4. 139. Kreyssig, A. et al. Magnetic order in GdBiPt studied by (1969).
Science 367, 895–900 (2020). X-​ray resonant magnetic scattering. Phys. Rev. B 84, 167. Bell, J. S. & Jackiw, R. A PCAC puzzle. Nuovo Cim. A
112. Otrokov, M. et al. Unique thickness-​dependent 220408 (2011). 60, 47–61 (1969).
properties of the van der Waals interlayer 140. Müller, R. A. et al. Magnetic structure of GdBiPt: 168. Chen, Y., Wu, S. & Burkov, A. A. Axion response in
antiferromagnet MnBi2Te4 films. Phys. Rev. Lett. 122, a candidate antiferromagnetic topological insulator. Weyl semimetals. Phys. Rev. B 88, 125105 (2013).
107202 (2019). Phys. Rev. B 90, 041109 (2014). 169. Goswami, P. & Tewari, S. Axionic field theory of
113. Li, H. et al. Antiferromagnetic topological insulator 141. Gooth, J. et al. Experimental signatures of the mixed (3+1)-dimensional Weyl semimetals. Phys. Rev. B 88,
MnBi2Te4: synthesis and magnetic properties. Phys. axial-​gravitational anomaly in the Weyl semimetal 245107 (2013).
Chem. Chem. Phys. 22, 556–563 (2020). NbP. Nature 547, 324–327 (2017). 170. Liang, S. et al. Experimental tests of the chiral
114. Li, Y. et al. Layer-​magnetization-tuned topological 142. Ikebe, Y. et al. Optical hall effect in the integer anomaly magnetoresistance in the Dirac–Weyl
phases in Mn2Bi2Te5 films. Preprint at arXiv https:// quantum Hall regime. Phys. Rev. Lett. 104, 256802 semimetals Na3Bi and GdPtBi. Phys. Rev. X 8,
arxiv.org/abs/2001.06133 (2020). (2010). 031002 (2018).

www.nature.com/natrevphys
 Reviews

Acknowledgements DGE-1745303. Financial support by the European Union Peer review information
The authors thank Y. Wang (Harvard) and S. Roychowdhury (grant no. 742068) is gratefully acknowledged. P.N. is a Nature Reviews Physics thanks Pavan Hosur and the other,
(Max Planck Institute) for input and discussions. This work Moore Inventor Fellow and gratefully acknowledges support anonymous, reviewer(s) for their contribution to the peer
was supported by the US Department of Energy ‘Photonics at through grant no. GBMF8048 from the Gordon and Betty review of this work.
Thermodynamic Limits’ Energy Frontier Research Center Moore Foundation.
under grant DE-​SC0019140, the Army Research Office MURI
(Ab-​I nitio Solid-​S tate Quantum Materials) grant no. Publisher’s note
Author contributions
W911NF-18-1-0431 and by the Science and Technology Springer Nature remains neutral with regard to jurisdictional
The authors contributed equally to all aspects of the article.
Center (STC) Center for Integrated Quantum Materials under claims in published maps and institutional affiliations.
US National Science Foundation (NSF) grant no.
DMR-1231319. C.A.C.G. is supported by the NSF Graduate Competing interests
R e s e a r c h Fe l l o w s h i p P r o g r a m u n d e r g r a n t n o . The authors declare no competing interests. © Springer Nature Limited 2020

Nature Reviews | Physics