See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/315489641

Theory of 2D crystals: Graphene and beyond

Article in Chemical Society Reviews · March 2017

DOI: 10.1039/C7CS00210F

CITATIONS READS
137 5,008

6 authors, including:

Luca Chirolli Elsa Prada

Italian National Research Council Universidad Autónoma de Madrid
60 PUBLICATIONS 1,285 CITATIONS 87 PUBLICATIONS 5,936 CITATIONS

SEE PROFILE SEE PROFILE

Jose Angel Silva-Guillén Pablo San-Jose

Madrid Institute for Advanced Studies Spanish National Research Council
36 PUBLICATIONS 1,703 CITATIONS 118 PUBLICATIONS 6,892 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Luca Chirolli on 25 December 2017.

The user has requested enhancement of the downloaded file.

 Jour
nal
Name

Theory of 2D crystals: graphene and beyond

Rafael Roldán,∗a Luca Chirolli,b Elsa Prada,c Jose Angel Silva-Guillén,b Pablo San-
arXiv:1703.07200v3 [cond-mat.mes-hall] 2 Aug 2017

Josea and Francisco Guineab,d

This tutorial review presents an overview of the basic theoretical aspects of two-dimensional (2D)
crystals. We revise essential aspects of graphene and the new families of semiconducting 2D
materials, like transition metal dichalcogenides or black phosphorus. Minimal theoretical models
for various materials are presented. Some of the exciting new possibilities offered by 2D crys-
tals are discussed, such as manipulation and control of quantum degrees of freedom (spin and
pseudospin), confinement of excitons, control of the electronic and optical properties with strain
engineering, or unconventional superconducting phases.

Key learning points: Novel two dimensional crystals. Models for the electronic structure. Spin
and pseudospin. Excitons and optical absorption. Superconductivity.

1 Introduction
In more practical terms, the all-surface nature of 2D crystals
Graphene is the first truly 2D crystal that has been isolated in a exposes them much more directly to influences of all sorts from
controlled manner, initiating a field of research known as "2D ma- the environment. As a consequence, their electronic and optical
terials". 1 2D crystals are materials of atomic thickness that, as a properties can be tuned with a particularly high degree of flexibil-
result of their reduced dimensionality, exhibit unique physical and ity. Parameters associated to the electronic structure of crystals,
chemical properties that strongly differ from their 3D counter- like effective masses, Fermi energy, Fermi velocity or band gap,
parts. If the crystalline structure of a 3D crystal is preserved as its can be efficiently tuned by controlling the number of layers, by
thickness is reduced down to atomic scales, as happens with lay- chemical functionalization, by gating or by applying strain to the
ered materials, it typically exhibits dramatic changes in its physi- samples or the substrate on which 2D materials are deposited. As
cal properties. For example, only when graphene is isolated from compared to other 2D systems such as conventional thin films,
3D graphite into its one-atom-thick form, do its carriers behave as 2D crystals also exhibit a much higher quality and overall coher-
massless relativistic electrons. Another example is MoS2 , which in ence, as their strong covalent in-plane bonds allow them to keep
its monolayer form presents a direct band gap and spin-polarized disorder under control as their thickness is reduced.
valleys, two key features that make this 2D semiconductor much Graphene is the best studied among 2D materials, with recog-
better suited for photonics and optoelectronics than its 3D ver- nised hallmark properties that make it particularly attractive,
sion. From a fundamental point of view, moreover, 2D is a critical both from a fundamental viewpoint and because of its potential
dimensionality for many physical effects, and marks a threshold applications. 2 However, it lacks a band gap, which is necessary to
wherein thermal and quantum fluctuations acquire a much more switch between insulating and metallic states, making graphene
dominant role than in 3D. This alone often makes the physics of unsuitable for some electronic devices. Furthermore, a band gap
2D crystals highly non-trivial. in the visible or infrared range of the spectrum is also required for
solar cell and telecommunication applications. Consequently, sig-
nificant efforts have been devoted to identifying possible 2D semi-
a
Instituto de Ciencia de Materiales de Madrid, ICMM-CSIC, Cantoblanco, E- conducting crystals. 3 Several classes of layered compounds have
28049 Madrid, Spain. Fax: +34 913720623; Tel: +34 913349000; E-mail:
received attention recently, including hexagonal boron nitride (h-
rroldan@icmm.csic.es
b
Fundación IMDEA Nanociencia, C/Faraday 9, Campus Cantoblanco, 28049 Madrid, BN), silicene, MoS2 , black phosphorus (BP), etc. Single layers of
Spain h-BN are stable insulators with a large gap. Conversely, silicene
c
Departamento de Física de la Materia Condensada, Condensed Matter Physics Cen- is highly unstable because single layers can react with air. Among
ter (IFIMAC) & Instituto Nicolas Cabrera, Universidad Autónoma de Madrid, E-28049
the stable 2D crystals with semiconducting behaviour, some of
Madrid, Spain
d
Department of Physics and Astronomy, University of Manchester, Oxford Road,
the best known are transition metal dichalcogenides (TMDs), 4
Manchester M13 9PL, UK which also exhibit strong spin-orbit coupling (SOC) effects. BP

J
our
nal
Name,
[yea
r][
,vol
.
],1–13 | 1
is another layered crystal that has been recently synthesized in due to elastic properties that are dominated by out-of-plane
its single layer form, also known as phosphorene. BP is a stable flexural phonons. 8 Despite its one-atom-thickness, graphene
allotrope of phosphorus and an elemental semiconductor, with is highly impermeable to gases.
a high degree of anisotropy in its electronic and optical prop-
erties. 5 Much less understood are other families of transition Some of these attributes are shared by other 2D crystals that
metal trichalcogenides like TiS3 , single layer Sb (antimonene) or have been exfoliated after graphene.
monochalcogenides like GeSe, that have been recently synthe-
sised in their single layer form, joining the growing catalogue of 2.2 Transition Metal Dichalcogenides MX2
2D materials. Herein, we review current knowledge on the phys- TMDs form a very large and rich family of crystals, whose mem-
ical properties of graphene and related 2D crystals. We survey bers present several kinds of lattice structures and many differ-
their main electronic and structural features and provide minimal ent physical properties. The most studied are the semiconduct-
theoretical models that capture their low energy physics. Finally, ing compounds with the general formula MX2 (M = Mo, W; X =
we discuss some of the novel possibilities afforded by 2D crystals, S, Se, Te) that are composed, in its bulk configuration, of two-
including spintronics and valleytronics, control of excitons, strain dimensional X − M − X layers stacked on top of each other, cou-
engineering and new aspects of superconductivity. pled by weak van der Waals forces. A MX2 single layer is a sand-
wich structure in which the M atoms are ordered in a triangular
2 General description lattice, each of them bonded covalently to six X atoms, three of
In this section we review the crystalline order and electronic band them in the top and three in the bottom layer, see Fig. 1. Some of
structure of the most relevant 2D materials, and highlight some the main features of TMDs include: 9
of their main features.
• Thickness dependence of the nature of the gap: TMDs
2.1 Graphene have an electronic band structure which strongly depends
on the number of layers. Single layer TMDs are direct gap
Graphene is a one-atom-thick layer of carbon atoms, which when
semiconductors, with a gap (∼ 1.9 eV) located at the K and
stacked constitute 3D graphite. As in other carbon structures like
K’ points of the hexagonal BZ. The energy of the band gap
0D fullerenes (C60 ) or 1D carbon nanotubes, 2D graphene is a
lies in the visible range of the spectrum. Multilayer samples
crystal composed of sp2 -hybridized carbon atoms arranged in a
are indirect gap semiconductors (of ∼ 1.3 eV gap) with the
honeycomb lattice. However, the physical and chemical proper-
maximum of the valence band at Γ and the minimum of the
ties of graphene strongly differ from the other sp2 and sp3 carbon
conduction band at an intermediate point between Γ and K.
compounds of different dimensionality. Its electronic bands de-
fine it as a semimetal, at the boundary between the two most fre- • Multi-Orbital character: TMDs have valence and conduc-
quent phases of materials, metals and insulators. The electronic tion bands with a rich and complex orbital character. The
band structure of graphene is characterised by conical valence edge of the valence band is formed by a mixture of dx2 −y2
and conduction bands that touch at only two inequivalent points and dxy orbitals of the metal M, and px and py orbitals of the
in the Brillouin zone (BZ), dubbed Dirac points. The low energy chalcogen atom X. The edge of the conduction band, on the
properties are governed by out-of-plane pz -orbitals. Some of the other hand, is formed by a combination of d3z2 −r2 of M, plus
main characteristics of graphene are: 6 some minor contribution of chalcogen px and py orbitals.
• Massless carriers: The characteristic linear dispersion re-
lation of electrons in graphene in the vicinity of the Dirac • Strong spin-orbit coupling: TMDs present a strong SOC
points makes them behave as relativistic quasiparticles with which, together with the lack of inversion symmetry, leads
zero effective mass (see Sec. 3). As a result, the electronic to a splitting of the valence band of ∼ 140 meV (for Mo com-
and optical properties of graphene are completely differ- pounds) and ∼ 400 meV (for W compounds). The conduc-
ent from those of a standard 2D electron gas with a mas- tion band is also split by a few tens of meV.
sive parabolic dispersion relation, as e.g. Si and GaAlAs
Other characteristics of TMDs that we will discuss in Sec. 4
heterostructures. When graphene is exposed to a strong
include the coexistence of spin and valley Hall effects, the possi-
magnetic field, the massless character of the carriers man-
bility of tuning the band gap by different means, like strain en-
ifests in an unconventional quantum Hall effect. 7 Graphene
gineering or application of external electric fields, generation of
electrons can, moreover, propagate over large (micrometers)
highly stable excitons (electron-hole pairs bonded by Coulomb in-
distances without scattering due to the chirality of an inter-
teraction), or the emergence of superconductivity in highly doped
nal degree of freedom of carriers known as pseudospin.
samples.
• High stiffness and impermeability: Graphene is simulta- While the most commonly studied TMDs in the single-layer
neously flexible and extraordinarily rigid, with the highest form are those with TMs from group VI-B and are semiconduct-
elastic constants ever measured in any material. It can be ing, those with TMs from groups IV-B and V-B have also been stud-
stretched elastically up to ∼ 20% without rupture. Graphene ied. The ones from group V-B with trigonal prismatic coordination
(as other 2D membranes) presents a negative thermal ex- are metallic. 10 The change from semiconductor to metallic phase
pansion coefficient: it shrinks with increasing temperature, can be easily explained, since group V-B TMs have one valence

2| J
our
nal
Name,
[yea
r][
,vol
.
],
1–13
Fig. 1 Crystalline lattice and band structures of different 2D materials, ordered by increasing size of their gap (red shadowed area) as computed
within DFT without correlation effects: metallic (NbSe2 ), semi-metallic (graphene), semiconducting (TiS3 , antimonene, phosphorene, SnS, MoS2 ) and
insulating (h-BN).

electron less than the TMs from group VI-B and, therefore, do not a high level of structural anisotropy (Fig. 1). The edges of the va-
have enough electrons in the unit cell to occupy the valence band lence and conduction bands are formed from pz orbitals. 11 Some
like the group VI-B TMDs. The most studied TMDs from group V- characteristics of BP to be highlighted are:
B are NbS2 , NbSe2 and TaS2 . These materials present electronic
• Evolution of the band gap with thickness and applied
correlation phenomena such as superconductivity and even, in
strain: The gap in single-layer as well as in multi-layer sam-
the case of NbSe2 and TaS2 , a charge density wave phase that
ples is direct and located at the Γ point of the BZ. However,
can compete with the superconducting state. Those of group IV-B
while the energy of the gap in bulk BP is ∼ 0.3 eV, its value
are semimetals and, for example, doped TiSe2 exhibits a charge
increases with decreasing the layer number to ∼ 1.5 eV for a
density wave as well as superconductivity.
single layer phosphorene. For a given sample thickness, the
band gap is extremely sensitive to external strain. 12 There-
2.3 Phosphorene fore, BP provides a high feasibility for photonics and opto-
electronics devices that can operate at different frequencies.
Black phosphorus (BP) is another layered material that has been
BP is, in this sense, complementary to some of the most stud-
recently synthesized in its single layer form, also known as phos-
ied 2D crystals, namely graphene and TMDs, that have band
phorene. 5 BP is a stable allotrope of phosphorus and an elemental
gaps ranging from zero in graphene and the ∼ 2 eV in TMDs
semiconductor. As in graphene, each P atom in single layer phos-
(∼ 2.5 eV if we include correlation effects).
phorene is coordinated to three neighbouring atoms. But contrary
to graphene, the orbital hybridization is sp3 like, yielding a crys- • In-plane anisotropy: The peculiar puckered structure of BP
tal structure characterized by a puckered honeycomb lattice with layers leads to highly anisotropic optical and electronic prop-

J
our
nal
Name,
[yea
r][
,vol
.
],1–13 | 3
 erties in-plane. The non-isotropic band dispersion also yields M. While single layer antimonene is a topologically trivial semi-
highly anisotropic excitons and plasmons. 13 conductor, the possibility has been discussed that a transition to
a topological insulator may take place upon increasing the num-
• Semiconductor-to-semimetal transition: Thanks to the ber of layers, including the development of quantum spin Hall
strong response of BP to electric and strain fields, it is possi- phases. 22
ble to drive a semiconductor-to-semimetal transition in this
material with the appearance of a pair of Dirac-like cones in
the spectrum, similarly to graphene. 14,15 Such transition is 2.7 Transition metal trichalcogenides MX3
accompanied by a change in the topology of the system, due
to generation of ±π Berry phases around the Dirac points. Transition metal trichalcogenides (TMT) with the common for-
mula MX3 (M = Ti, Zr, Hf and X = S, Se, Te) are formed from
2.4 Hexagonal boron nitride h-BN trigonal prismatic MX3 chains such that two rectangular faces of
h-BN is a band insulator with B and N atoms sitting on different a MX6 trigonal prism are capped by X atoms of the neighboring
sublattices of a honeycomb lattice. It presents a direct band gap of chains. Thus, each transition metal atom M is coordinated to
∼ 4.5 eV located at the K and K’ points of the BZ. Few layer h-BN eight chalcogen atoms, X (see Fig. 1). Among the TMTs fam-
is widely used in 2D materials research because it is an excellent ily, TiS3 is attracting special attention because of its direct band
substrate to support and encapsulate graphene, black phospho- gap. 23 Contrary to TMDs, the direct nature of the TiS3 band gap
rus, etc. Due to its atomic flatness and very low concentration of is robust, with only a weak dependence of its magnitude with the
trapped charges, the use of h-BN as dielectric substrate instead number of layers or their stacking order. The reduced in-plane
of, e.g. SiO2 , considerably enhances graphene mobility with the structural symmetry of the crystal confers a strong anisotropy to
corresponding improvement of device performance. On the other its band dispersion. Interestingly, the in-plane anisotropy is op-
hand, due to its small lattice mismatch with graphene (∼ 1.8%), posite for the valence and the conduction bands. 24 The ultimate
graphene on h-BN is a perfect platform to study moiré pattern origin of such different anisotropy resides in the different orbital
effects on the electronic and optical properties. 16 h-BN is also character of the valence and conduction band edges. The top of
a fundamental piece in so called van der Waals heterostructures, the valence band is mostly made of the 3px orbitals of S belonging
which is the name commonly used to refer to heterostructures and to the ‘inner’ sulfur atoms, dispersing along the a-crystallographic
devices whose properties are engineered by stacking different 2D direction. On the other hand, the conduction band is mainly
crystals, with the desired properties, on top of each other. 17 Fur- composed of d3z2 −r2 orbitals of Ti atoms, oriented along the b-
thermore, h-BN is interesting in its own right as it supports mixed direction. The electronic and optical properties are therefore
photon-phonon modes (polaritons) with an hyperbolic dispersion highly anisotropic, with a strong directional dependence of con-
in a finite range of frequencies. These modes lead to highly con- ductance and a large linear dichroism.
fined electric fields. 18

3 Theoretical models
2.5 Monochalcogenides MX
Group IV monochalcogenides with the common formula MX (M =
As with any new class of materials, theoretical models have
Ge, Sn and X = S, Se) are bi-elemental crystals that, as phospho-
proved to be a powerful tool to understand the properties of many
rene, presents a puckered orthorhombic structure (Fig. 1). They
2D crystals. One may even say that the predictive role of theory is
are indirect gap semiconductors with a quasiparticle gap (includ-
particularly important in these systems, given their rather unique
ing correlation effects) ranging from 1.2 eV to 2.7 eV. As TMDs.
combination of interrelated physical properties, of which there is
Unlike phosphorene, they are not inversion-symmetric crystals.
hardly any precedent in more conventional materials. For exam-
Therefore the monochalcogenides present piezoelectric proper-
ple, many 2D crystals are so stable despite their reduced dimen-
ties (i.e. possibility to convert mechanical to electric energy),
sionality that they may be suspended to behave as a robust elas-
with piezoelectric coefficients that can be two orders of magni-
tic membrane of ultimate thinness, with a host of associated and
tude larger than in MoS2 or h-BN. Indeed, their puckered struc-
highly non-trivial anharmonic effects. At the same time, a sus-
ture makes them very soft along the armchair direction, a prop-
pended 2D crystal is often of such a pristine electronic quality as
erty that can further improve the piezoelectric performance. 19
to rival many semiconducting heterostructures. Crucially, more-
over, the electronic and elastic degrees of freedom of the crystal
2.6 Antimonene are often strongly coupled. The interplay of these two physical
Single layer antimonene is an elemental 2D crystal formed by Sb sectors gives rise to unexpected possibilities, both fundamental
atoms ordered in a buckled honeycomb structure. 20 It is an indi- and applied. Other examples of this synergetic character include
rect gap semiconductor with strong SOC (λ ∼ 0.34 eV) and a gap the interplay between strain and optical activity in MoS2 , between
of ∼ 1 eV. The valence band edge (composed of px and py orbitals) electronic structure, deformations and interlayer registry in var-
is located at the Γ-point of the BZ. 21 The bottom of the conduc- ious multilayers, 16 or between topological order, electric fields
tion band (formed by a combination of the three p orbitals) is and pressure in black phosphorus. 25 Theoretical models allow to
placed at a non-high symmetry point of the BZ, between Γ and navigate all these numerous possibilities much more easily.

4| J
our
nal
Name,
[yea
r][
,vol
.
],
1–13
� ���� � ���� and β as
|~ri j −~ri0j |
!
� � αβ αβ αβ
ti j = ti j,0 1−Λ . (1)
|~ri0j |
(0)
� � Here ~ri j and ~ri j are the vectors connecting sites i and j under
strain and at equilibrium, respectively. Dimensionless parameters
�������� [��]

�������� [��]
Λαβ may be adjusted to ab-initio results (even making them de-
������������ ��� ������������ ��� pend on i, j), or relating them to the angular momentum of α and
� �
������-������ ��� �������������� β using TB theory (Harrison rule). However, the quality of the fit
using the Slater-Koster approach is often not excellent. Also, the
-� -�
fit procedure may become very delicate given the large number of
free parameters, and the result may depend on which parts of the
bandstructure one chooses to emphasise. For a more systematic
and deterministic way to build TB models it may be preferrable
-� -�
to employ ‘Wannierization’ techniques instead. In essence, Wan-
� � Γ � � Γ
nierization consists on a projection of the DFT bandstructure onto
Fig. 2 MoS2 bandstructure comparison using tight-binding (red) and maximally localized states in real space of a given orbital charac-
relativistic Density Functional Theory (black). Panel (a) shows a fit of a ter. The result is a TB Hamiltonian that is not of a Slater-Koster
Slater-Koster model up to next-nearest-neighbors, adapted with
form, but that, if allowed to range beyond nearest neighbors, can
permission from Ref. 26 . Panel (b) shows a Wannierization calculation
with hoppings up to sixth-order neighbors, adapted with permission from often yield a precise fit of both bands and orbital character in a
Ref. 27 . controlled way. Figure 2 shows a comparison of both methods in
the case of MoS2 , taken from Refs. 26,27 .

The final level of the model hierarchy consists of analytical ap-

proaches. These are effective models useful to describe in a trans-
3.1 A threefold approach to modelling 2D crystals
parent way the essential physical mechanisms at play in a 2D crys-
When a clean experimental characterisation is lacking, a power- tal without aiming for a quantitative description. The relevant
ful strategy to modelling the electronic and elastic properties of model for a given crystal depends on the specific space symme-
2D crystals is to combine three types of theoretical approaches, tries of its lattice. The most archetypal of these, relevant for hon-
with a decreasing level of computational complexity. At the top eycomb lattices (mono- and dichalcogenides, h-BN, graphene) is
of the stack are ab-initio methods, that range from Density Func- the massive Dirac equation in two dimensions, in effect a k · p
tional Theory (DFT) to more elaborate extensions such as GW extension of the effective mass approximation of semiconductors.
or the Bethe-Salpeter equation to capture some of the electronic Honeycomb crystals typically possess a bandstructure with a gap
correlations. These methods are very successful in some of these centred at two identical ± valleys located at the K and K’ points
systems, and remain computationally viable given the relatively in the BZ. Graphene is a special case, with gapless valleys, see
small number of orbitals in the unit cell of typical 2D crystals. red cones in Fig. 3a. The idea is to consider a neighborhood
They can often predict the single-particle bandstructure and elas- of the bandstructure around the two valleys, neglecting all but
tic coefficients with good accuracy. the (spinful) valence and conduction subbands. Expanding these
to second order in wavevector ~k = (kx , ky ) around K and K’ and
For larger-scale nanostructures and devices, inaccesible to ab-
imposing invariance under the C3 symmetry group of the lattice
initio approaches, the method of choice is typically tight-binding
(120◦ rotations), we arrive at a valley-degenerate Hamiltonian of
(TB), which occupies an intermediate level of computational com-
the form H = H+ ⊕ H− , where each τ = ± valley is described, in
plexity. The TB models for these systems can become moderately
the absence of strains, by
complicated, and may include many different crystal fields (onsite
energies) and inter-orbital overlaps (hopping amplitudes) beyond
!
m0 + (α + β )|~k|2 h̄v(τkx − iky )
nearest neighbors. All these parameters are often tuned to match Hτ = + Hso (2)
h̄v(τkx + iky ) −m0 + (α − β )|~k|2
the electronic structure of ab-initio calculations. There are two
main ways to do this. On the one hand, one may choose a set
of relevant atomic orbitals in the unit cell of the material, fix the The gap is Eg = 2m0 , v is a velocity and α and β control the
range of hoppings to consider and derive a general Slater-Koster valence and conduction effective masses. The Hτ matrix above
model for the system, parametrized by a set of overlap integrals is expressed in the basis of ~k = 0 valence and conduction states,
and crystal fields. 28 These can then be fitted to match the ab-initio which define a ‘pseudospin’. The ~k = 0 conduction band state
bandstructure as closely as possible, potentially taking into ac- (‘peudospin-up’) becomes mixed with the~k = 0 valence band state
count also the orbital character of the bands at different k-points. (‘peudospin-down’) due to the off-diagonal h̄v(τkx ± iky ) terms,
Such approach produces a sensible Slater-Koster TB model that which makes pseudospin a non-trivial degree of freedom in any
allows in particular to incorporate the effect of strain in a natu- scattering (i.e. ~k-changing) process. The pseudospin should not
αβ
ral way, by modifying hopping amplitudes ti j,0 between orbitals α be confused with the real electron spin sz that enters the SOC

J
our
nal
Name,
[yea
r][
,vol
.
],1–13 | 5
 � under inversion ~ri → −~ri of all its site positions. The implications
are profound. In particular the model cannot contain terms that
distinguish the two sublattices, so m0 = 0, and the system must be
gapless. This argument points to a connection between the size of
the gap and the chemical imbalance between the two sublattices
in honeycomb crystals.
Graphene’s low energy spectrum is thus composed of two
�
isotropic massless Dirac cones (the α, β corrections in (2) can
�� |↑〉 be neglected for most purposes, and spin-orbit is also negligible
�������� [��]

due to the low atomic mass of Carbon). The massless Dirac spec-
|↓〉 trum exhibits pseudospin chirality and a unique scale invariance
(~r → Λ~r and ε → ε/Λ), responsible for many of graphene’s re-
markable electronic properties. Amongst these are an absence of
-�� Anderson localization (provided disorder is smooth on the scale of
the lattice spacing), the associated Klein tunneling phenomenon,
� � �� or a logarithmic Fermi velocity renormalization from interac-
tions. 2 One of the most remarkable consequences of graphene’s
Fig. 3 (a) Evolution of graphene’s Dirac cones (red) under uniaxial scale invariance is the way strain couples to electrons. As men-
tension along the armchair direction (green). If the Dirac cones fuse at tioned above, a given (possibly position dependent) strain ten-
the Γ point (blue), a gap opens, as in the case of phosphorene. (b) sor εi j applied to the system enters the effective massless Dirac
Transition metal dichalcogenides have gapped Dirac cones with a strong
spin-orbit splitting of the valence band. Circularly polarized light may
model as an effective pseudogauge field that shifts the position
selectively excite electrons from a specific valley. of the two Dirac cones in opposite directions, see green cones in
Fig. 3a. If a sufficiently strong uniaxial strain is applied along the
armchair direction, it can in principle have the dramatic effect of
term Hso , which reads fusing the two Dirac cones at the Γ point of the BZ. This possibil-
! ity requires unrealistic deformations beyond graphene’s point of
λ0 +λ
2 + (λ00 + λ 0 )|~k|2 0 rupture, but signals a possibility that is actually materialized in a
Hso = τsz λ0 −λ , (3)
0 2 + (λ00 − λ 0 )|~k|2 different 2D crystal: phosphorene.

for some λ0 , λ00 , λ and λ 0 . This SOC polarizes the real spin in op-
posite directions in opposite valleys. All parameters in the model 3.3 Phosphorene: fused Dirac cones
may be obtained from experiment, from a TB or from ab-initio The lattice of black phosphorus monolayers is not of the honey-
models. States within each valley are characterized by an oppo- comb type, as its buckling along the armchair direction breaks the
site angular momentum associated to the circulation of the pseu- C3 symmetry, see Fig. 1. Its gap is not located at two valleys. In
dospin around the K and K’ points in the BZ. This follows from the terms of symmetries, black phosphorus monolayers are analogous
form of Hτ , which has a non-zero ‘Berry curvature’, 29 and leads to graphene with a uniform uniaxial strain along the armchair di-
to valley selectivity of various properties, such as transport or the rection, i.e. C3 symmetry is broken but inversion is not. More-
absorption of circularly polarised light, to be discussed below. over, spin-orbit coupling is also rather small. The TB model of
The effect of a strain field εi j at low energies can also be de- the system is therefore much like graphene’s, with the exception
rived analytically. 30 Its dominant contribution enters as a val- that of the three hopping amplitudes connecting a phosphorus π
ley shift in graphene ~k → ~k − ητ ~A or as a deformation poten- orbital to its nearest neighbours, two are equal (t1 < 0), but the
tial Hτ → Hτ + η 0 D in other gapped honeycomb crystals. Here, other (t2 > 0) is different and of opposite sign, 11 as would cor-
~A = (εxx − εyy , −2εxy ) is a pseudo-gauge vector field, expressed in respond to Eq. (1) with an (unrealistically) large |~ri j −~ri0j | along
terms of the strain tensor components, D ≈ εxx + εyy + O(εi2j ) is a the armchair direction. One may thus continuously transform the
scalar and η and η 0 are parameters that control the strength of graphene TB model with t2 = t1 < 0 into that of black phospho-
the electromechanical coupling. rus by increasing t2 , which introduces a pseudogauge-like shift
of the two Dirac cones that approach each other. For t2 < −2t1
3.2 Graphene: massless Dirac the model remains gapless, but as t2 ≥ −2t1 the two cones fuse at
Graphene, the carbon honeycomb 2D crystal and forefather of all the Γ point. A gap Eg ≈ 2t2 + 4t1 like phosphorene’s then opens,
other 2D materials, corresponds to a particularly simple instance abruptly changing the topology of the bandstructure, as depicted
of the model in Eq. (2), as a result of its enhanced lattice sym- in Fig. 3a.
metry group C3v . The two orbitals that form the pseudospin in This process can actually be reversed using strain. If we apply
the Dirac model are π orbitals of carbon atoms arranged into two uniaxial compression to phosphorene along the armchair direc-
triangular sublattices. Unlike in most other honeycomb 2D crys- tion, |t1 | is increased, while t2 slightly decreases, so that the quasi-
tals, these two orbitals are indistinguishable in graphene, so that particle gap (Eg ≈ 1.8 eV using DFT-GW at equilibrium) decreases
in addition to the 120◦ rotations of C3 , the lattice is also invariant (increases) by a huge ∼ 6% per 1% of uniaxial compression (ex-

6| J
our
nal
Name,
[yea
r][
,vol
.
],
1–13
pansion). For a sufficient compression, we may reach t2 ≤ −2t1 , shown by some crystals when they are reduced to their monolayer
which corresponds to a gapless spectrum. This situation may be form.
achieved in practice for multilayers, 25 which then transition into
a semimetalic phase with shifted Dirac cones.
4.2 Excitons
By shining materials with light and analysing their optical prop-
3.4 Transition metal dichalcogenides erties one can in turn acquire information about their electronic
The family of TMDs, of which MoS2 is probably the most studied, structure. When photons of sufficient energy hit a semiconductor,
also share a honeycomb lattice structure. The difference between they can excite electrons from the valence band to the conduction
the metal Mo atom and the chalcogen S atom makes them semi- band creating electron-hole (e-h) pairs. Within a one-electron ap-
conducting, with a spectral gap Eg = 2m0 of the order of 1.9 eV at proximation, these pairs can be described in terms of the mate-
the K and K’ points. The response of the gap to uniaxial strain is rial’s band structure. An excited e-h pair then has a minimum
opposite to that of phosphorene, decreasing by around 1.5% per energy equal to the bandgap Eg . However, since these quasiparti-
1% of uniaxial tension. cles are charged, they interact through the electrostatic Coulomb
The large atomic mass of the metallic species endows these force and a proper description of the excitation spectrum has to
materials with a very sizeable spin-orbit coupling. This leads to take into account the many-body nature of their wave function.
a significant splitting of the valence bands (the splitting in the It turns out that the attraction between the excited electron with
conduction band is quite smaller). The corresponding spin struc- charge −e and the hole left behind with charge +e, causes their
ture around the gap is sketched in Fig. 3b. The opposite spin- motion to be correlated and decreases the energy of the pair with
polarization of opposite valleys opens unique opportunities for respect to the free e-h excitation, forming a bound entity with a
spintronics in these materials, as discussed below. minimum energy Eex < Eg inside the gap that is called an exciton,
see Fig. 4a. The exciton has a finite lifetime, as the electron and
4 Novel developments the hole can undergo recombination into the original unexcited
4.1 Control of spin and valley degrees of freedom ground state. An exciton is thus a metastable condensed matter
elementary excitation that carries energy and momentum with
Electron spin is an internal quantum degree of freedom associated
zero net charge. There are other higher order excitations, such as
with a magnetic moment, whose manipulation and control for
trions, biexcitons, polaritons, etc. that we will not discuss here.
electronic applications are studied in the field known as spintron-
ics. The pseudospin and the valley index are additional degrees of The typical quantities that describe an exciton are its binding
freedom that can play a role similar to spin in spintronics. 31 We energy, Eb ≡ Eg − Eex and its size, parametrized by the exciton
have seen that the band structure of graphene and TMDs present radius, a. In a few materials where the Coulomb interaction is
two degenerate and inequivalent valleys at the K and K’ points very strong, as is the case of fullerenes, small excitons of a size
of the BZ. While graphene is centrosymmetric, inversion symme- comparable to the lattice constant may form, which are known
try is broken in single layer MoS2 and related TMDs. The lat- as Frenkel excitons. These have binding energies of the order
ter present a strong SOC, lifting spin degeneracy in both valence of 1 eV. In the most common cases, the semiconductor dielec-
and conduction bands. Time reversal symmetry imposes a rela- tric constant is large and electric field screening tends to reduce
tion between valley and spin quantum numbers. Additionally, the the Coulomb interaction, producing bigger exciton radii, smaller
opposite pseudospin angular momentum in each valley imposes binding energies of the order of 10 − 100 meV, and longer exciton
selection rules for inter-band optical transitions under circularly lifetimes. These are called Wannier-Mott excitons.
polarized light, so that opposite light polarizations are absorbed To find the exciton solution exactly is certainly complicated, but
by opposite valleys, see Fig. 3b. The combination of both effects can in principle be done by solving the so-called Bethe-Salpeter
allows the use of optical helicity to control the spin polarization of equation for excitons. In many cases, though, we can perform a
photoexcited carriers at each valley independently. 29 Since inter- series of simplifications valid for most common semiconductors.
valley scattering is highly suppressed due to the breaking of spin First, we consider that the electron that is excited into the con-
degeneracy, long spin and valley polarization lifetimes are possi- duction band can be simply described as a quasiparticle with mass
ble in single layers of TMDs. A related phenomenon is the spin me as obtained by the conduction band minimum. Its interaction
(valley) Hall effect, whereby opposite Hall currents for opposite with the rest of the electrons of the valence band is replaced by its
spins (valleys) emerge. In single layer TMDs, the valley Hall ef- interaction with a hole with opposite charge and mass mh given
fect can be realised in optoelectronic setups by using the optical by the valence band structure at its maximum. This is known as
valley-selection rules to create a population imbalance between the effective mass approximation. Second, we consider that dif-
different valleys. This produces a valley-polarized Hall current ferent e-h pairs are so far apart (in space or in time), that they
that switches sign with the light polarisation. Therefore, 2D TMDs can be considered independent. These two assumptions allow us
are materials with potential to use the valley index in an equiv- to neglect exchange and correlation contributions to the Coulomb
alent way as the spin degree of freedom is used in spintronics, interaction. The two-particle problem is then solved by going to
leading to the concept of valleytronics. 32 Note that inversion sym- the center of mass and relative coordinate reference system.
metry and spin degeneracy at each valley are restored in bilayer The wave function of the resulting bound state is said to be hy-
and bulk TMD samples, demonstrating again the singular physics drogenic because it is similar to the one of the hydrogen atom.

J
our
nal
Name,
[yea
r][
,vol
.
],1–13 | 7
� � �
conduction photocarriers length and Y0 and H0 are second-kind Bessel and Struve functions,

��������-���� �����������
Eb respectively. This potential presents a logarithmic divergence for
dark
�����

Eg
Eex
excit. bright r → 0 (like the potential of a charged string) and reduces to the
excitons
unscreened Coulomb potential at large distances,
light cone
valence (  
���������� ���������� 2a0 R∞ γ + log 2rr 0 r  r0
� VK (r) ≈ , (5)
ε̄r0 −r0 /r r  r0

where γ is Euler’s constant. The crossover between these two

behaviours is characterised by the length scale r0 . Cudazzo et
al. 38 later considered the problem of strictly d = 0 2D crystals
(embedded in a dielectric environment) and arrived at the same
Keldysh potential, but expressed in terms of the 2D polarizability
α2D of the dielectric sheet. The polarizability determines the scale
at which the two asymptotic forms match, r0 = 2πα2D .
Fig. 4 (a) Schematic representation of the single-particle energy bands,
For excitons larger than r0 , the binding energy and the exciton
with an exciton state at energy Eb below the gap Eg . (b) Schematic
representation of the particle-hole excitation spectrum. Excitons within radius have the same form as Eb2D and a2D , but with ε replaced by
the light cone can decay radiatively and are thus "bright", see Ref. 33 . (c) ε̄ (note that ε̄ = 1 for the 2D dielectric in vacuum). This approxi-
Excitons in phosphorene from Ref. 34 . Top: schematic plot showing the mation is good for excitons whose size is large as compared to r0 .
measured ground-state exciton energy (red) and the energy
In the opposite limit of small excitons, the e-h pair experiences
corresponding to the quasiparticle band edge (blue). Bottom:
Calculated optical absorption from first-principles of monolayer black mostly the logarithmic part of the Keldysh interaction, and 39
phosphorus with e-h interactions (excitonic absorption, red curve) and     s
without e-h interactions (quasiparticle absorption, blue curve). (d) 2D a0 R∞ 4r0 2D ε̄m
Eb = 2 log 2D − 3 , a = a0 r0 . (6)
Photoluminescence spectra (red, green, blue and purple curves) and ε̄r0 a µ
differential reflectance spectra (grey curves) from Ref. 35 for a variety of
TMDs exhibiting a number of different excitonic peaks. Adapted from This is the case of monolayer MoS2 or phosphorene, with binding
Ref. 33 with permission of the American Physical Society [panel a)], from energies around 600 meV, 33,39,40 mid-way between the typical
Refs. 34 and 35 with permission from Nature Publishing Group [panels b)
and c)].
Wannier-Mott and Frenkel scales.
The excitons influence the optical properties of semiconductors
and can be detected in experiments. One such property is the
optical absorption, which is the conversion of a photon into an
However, the Coulomb interaction in a typical 3D semiconductor
exciton. Due to conservation of energy and momentum, this pro-
VC (r) = −e2 /(4πε0 εr) is screened by all other electrons in the va-
cess occurs at points in momentum space where the photon light
lence band. This effect is captured by a dielectric constant ε  1.
cone overlaps with the exciton dispersion surface, see Fig 4b. The
The effective masses me and mh are, moreover, much smaller than
existence of excitons lowers the threshold of photon absorption,
the free electron mass m. This results in an exciton binding en-
which shows peaks at the energies of different internal states of
ergy Eb much smaller than the Rydberg energy, and an exciton
the exciton, see Fig. 4c for the case of phosphorene. Said ab-
radius a much bigger than the hydrogen atom, 36
sorbance may be measured through the differential reflectance
Eb3D =
µ
R∞ , a3D =
mε
a0 . (4) ∆R/R, 35 which thus exhibits signatures from different internal
mε 2 µ exciton states that may decay radiatively or non-radiatively, see
e 1 2 Fig. 4d. It is also possible to measure the opposite phenomenon,
Here R∞ ≡ 4πε 0 2a0
= 13.6 eV is the Rydberg energy, µ =
the material’s photoluminescence (PL). In PL, the electrons in the
me mh /(me + mh ) is the exciton reduced mass and a0 =
sample are excited electrically or optically and, after some energy
4πε0 h̄2 /me2 = 0.53 Å is the Bohr radius.
loss (relaxation), they recombine and return to the ground state
Now, in a 2D crystal, the electron and hole are confined to move radiatively, i.e. by emitting light. Fig. 4d shows different PL spec-
in lower dimensions, thus increasing their attraction. Constrained tra for a series of TMDs.
to 2D space, the same Coulomb interaction VC (r) translates into
a bigger Eb2D = 4Eb3D and a smaller a2D = a3D /2. This is a rather
rough simplification of the problem, however, that neglects the 4.3 Strain engineering
different screening between the 2D material and its surrounding The coexistence of high stiffness and flexibility in 2D crystals
environment. Keldysh showed that in thin films of thickness d, make these materials excellent platforms for strain engineering,
the actual interaction VK (r) between charges at distances bigger i.e. applying strain externally to tune and control the electronic
than d is influenced by the the dielectric constant of the medium and optical properties of materials. 41 Some possibilities are:
surrounding the film, ε1 and ε2 (typically the substrate and vac- Generation of pseudo-magnetic fields– As discussed in Sec.
uum). Solving the electrostatic problem with the use of image 3, when graphene is subjected to strain fields, the modification of
charges, 37 he obtained VK (r) = R∞ (πa0 /ε̄r0 ) [Y0 (r/r0 ) − H0 (r/r0 )], hopping amplitudes between carbon atoms gives rise to effective
where ε̄ = (ε1 + ε2 )/2, r0 = dε/(ε1 + ε2 ) is an effective screening gauge fields, whose effect is similar to that of a magnetic field ap-

8| J
our
nal
Name,
[yea
r][
,vol
.
],
1–13
 more, by applying higher amounts of strain, but still below the
fracture limit, it is possible to drive a semiconducting-to-metallic
transition. Such a huge modification of the gap in 2D semicon-
ductors (from ∼ 1.9 eV to 0 eV for the case of MoS2 ) has to be
compared with the poor tunability of 3D semiconductors like
silicon of only ∼ 0.25 eV under ∼ 1.5% of biaxial strain.
Strain induced funnel of excitons– Strain engineering can be
used for the creation of a broad-band optical funnel for excitons
in semiconducting 2D crystals. 45 By continuously changing the
strain across a sheet of MoS2 or black phosphorus, for example,
a continuous variation of the optical band gap is produced, al-
lowing for the capture of photons with different energies. Fur-
thermore, it is possible to tune the strain profile so as to drive
the photogenerated excitons towards the regions of minimal gap,
creating a funnel for excitons. 46 The funnel effect in BP is much
stronger than in MoS2 and of opposite sign. 33 While excitons in
MoS2 are driven isotropically towards regions of maximum ten-
Fig. 5 Strained induced funnel effect of excitons in 2D crystals. An sion, excitons in BP move away from tensile regions. This differ-
indenter creates an inhomogeneous profile of strain in single layer MoS2
ence stems from the fact that MoS2 and BP respond differently
(a) and phosphorene (b). The strain gradient modulates the gap of the
two crystals as sketched in the bottom panel. Photogenerated excitons to the application of strain: the gap in MoS2 (BP) is reduced
(green arrows) are pushed isotropically towards the indenter center in (increased) with tensile strain. Furthermore, the exciton drift in
MoS2 (funnel effect), while they are pushed anisotropically away it in MoS2 is isotropic, while the funnelling in BP is highly anisotropic,
black phosphorus (inverse funnel effect). Adapted from Ref. 33 with with much larger drift lengths along one crystallographic (arm-
permission of the American Physical Society.
chair) direction (see Fig. 5). The inverse funnel effect can be
beneficial for manipulation and harvesting of light, and in partic-
plied perpendicular to the graphene plane, albeit opposite in dif- ular for the design of more efficient solar cells.
ferent valleys. 42 Therefore, strain engineering can be used to dis- Piezoelectricity– A large number of 2D crystals lack inversion
cretize the graphene band structure into a set of pseudo-Landau symmetry. This gives them piezoelectric properties, that is the
levels, corresponding to counterpropagating cyclotron orbits in potential ability to convert mechanical to electric energy. Stretch-
opposite valleys. This effect has been observed experimentally ing or compressing a piezoelectric crystal generates an electrical
and pseudo-magnetic fields exceeding 300 Tesla have been re- voltage. Conversely, an applied voltage produces expansion or
ported in trigonally distorted graphene nanobubbles. 43 MoS2 and contraction of the crystal. MoS2 , which is centrosymmetric in its
related TMDs have, like graphene, an hexagonal lattice structure. 3D bulk configuration, has been shown to become piezoelectric
However, the simple pseudo-magnetic field picture of strained when it is reduced to its monolayer form. 47 Recently, single layer
graphene does not carry over to deformed monolayer TMDs. Due monochalcogenides have been predicted to show an anomalously
to the complex orbital contributions leading to the formation of strong piezoelectric response, with piezoelectric coefficients that
the valence and conduction bands, we have seen in Sec. 3 that can be two order of magnitude larger than those in MoS2 . Fur-
the effect of strains in the low energy model is dominated by dif- thermore, due to their anisotropic puckered lattice, the piezo-
ferent terms than in the massless Dirac Hamiltonian relevant for electric properties are strongly angle-dependent. 48 Considering
graphene. As a consequence, not only one gauge field but several the rapid advances in nanofabrication techniques, their amazing
pseudo-vector potentials and scalar fields appear in strained sin- elastic properties and the possibility to withstand large strains, 2D
gle layer TMDs. 30 The inclusion of strain-displacement relations piezoelectric crystals are viewed as promising platforms for appli-
from valence force-field models leads to additional correction to cations in nano-sensors or portable electronic devices for energy
the strain-induced pseudo-magnetic fields. 44 harvesting.
Direct-to-indirect gap and semiconducting-to-metallic 5 Superconductivity
transitions– The outstanding stretchability of 2D semicondcutors
can be used to tune the size and the nature of their band gap. Superconductivity is one of the most fascinating property of mat-
While single layer TMDs are direct gap semiconductors, uniaxial ter of pure quantum mechanical origin. It is well known that,
strain can produce a shift of the band edges and drive a transition by applying a current through a metal, a voltage drop across the
to an indirect gap. The opposite trend takes place in group IV-B sample is usually generated that is proportional to the applied
monochalcogenides, which are indirect gap semiconductors that current via the well known Ohmic law, V = RI, with R the resis-
can become direct gap crystals under tensile strain. Controlling tance. In ordinary metals, the resistance diminishes as tempera-
the nature of the gap with strain (straintronics) may become a ture is decreased, but eventually saturates to a finite value that is
powerful strategy for photonics applications. One may envision due to electrons scattering off impurities and imperfections of the
combining dark and bright regions for photoexcitation within the sample. In a superconductor, in contrast, an abrupt drop to zero
same sample by tailoring substrate-induced tensions. Further- resistivity occurs at a certain critical temperature, Tc , below which

J
our
nal
Name,
[yea
r][
,vol
.
],1–13 | 9
current flows through the system without resistance (see Fig. 6a). a b 30

A finite resistance generates dissipation, but in superconductors

supercurrents can flow indefinitely without dissipating. The local
20
density of states measured in Scanning Tunneling Spectroscopy

dI/dV(a.u)
reveals that a superconductor is characterized by a full gap ∆
in the quasiparticle spectrum, with highly pronounced coherence 10
peaks at energy ±∆ (see Fig. 6b). The absence of single-particle
states at energies below the gap and the presence of supercur-
rents point to a non-trivial state of matter, where charge carriers 0
-0.5 0.0 0.5
are different from the conventional elementary quasiparticles. Vbias (mV)

The celebrated Bardeen-Cooper-Schrieffer (BCS) theory of su-

Fig. 6 (a) Experimental variation of the resistance vs. temperature for
perconductivity (SC) explains the phenomenon as a many body
different number of layers of NbSe2 . Shown in the inset is the
phase-coherent state, where electrons with opposite spin and mo- dependence of Tc with the number of layers. Adapted with permission
mentum pair up across the Fermi surface via a phonon-mediated from Ref. 49 (b) Experimental tunnelling differential conductance dI/dV of
attractive interaction. This pairing leads to the formation of a superconducting monolayer of Pb for different representative tip
new elementary entities, the so-called Cooper pairs, that Bose- positions (red and blue solid lines refer to regions of high and low
coherence peak, respectively, the and the black solid line is the BCS
condense to form a new ground state with a finite excitation gap
best fit). The dashed orange line corresponds to the theoretical
∆ that reflects the binding energy of Cooper pairs. The BCS theory zero-temperature dI/dV spectra of a BCS superconductor, proportional
was a tremendous success, and was able to explain and predict all to the density of states (DoS). The DoS is zero within the gap ∆ and
the essential properties of many 3D superconductors. exhibits pronounced coherence peaks at ±∆. Adapted with permission
from Ref. 50 .
Thermal fluctuations break Cooper pairs apart. This makes the
gap temperature-dependent, with ∆(T = 0) = ∆0 and ∆(Tc ) = 0.
For T < Tc , the coherence length ξ (T ) ∝ 1/∆(T ) measures the size
of the Cooper pairs. At T = Tc , ξ → ∞ and the Cooper pairs un- be well-defined. It follows that the BCS ground state is a superpo-
bind. According to Anderson theorem, Tc for s-wave SC is not sition of states with different number of particles. The quantum
affected by weak conventional disorder due to time-reversal (TR) uncertainty on the number of particles fixes the phase of the BCS
invariance. A magnetic field breaks TR in a way that makes states ground state. It is thus said that SC spontaneously breaks the
of opposite spin and momentum on the Fermi surface no longer global gauge symmetry of the system, and develops a complex
degenerate. For low fields the resulting non-dissipative Cooper order parameter ∆(r) = ∆0 eiφ (r) . In this language, the spectral
pair motion screens the external field completely. The magnetic gap is given by the average |h∆i|. Thermal fluctuations of φ are
flux is thus completely expelled from the superconductor, which responsible for the suppression of the gap with temperature. In
behaves as a perfect diamagnet. When the magnetic field is in- a 3D superconductor below Tc , the phase φ (r) fluctuates around
creased beyond a critical value, Hcr , SC is destroyed. The value a well-defined average that is fixed across the entire sample at
of Hcr corresponds to a magnetic length for Cooper pairs equal to equilibrium, so that the gap is finite and SC remains stable against
their size ξ , Hcr = Φ0 /(2πξ 2 ), with Φ0 = h/2e the superconduct- fluctuations.
ing flux quantum. In 2D systems a dramatic change takes place. Thermal fluctu-
When lowering the dimensionality of the system, these well- ations in this case have a stronger effect, and destroy the long-
defined results must be revised. The density of states (DoS) at the range rigidity of the phase across the system. The non-local cor-
0
Fermi level decreases, and so does the density of electrons avail- relation function h∆∗ (r)∆(r0 )i = ∆20 he−iφ (r) eiφ (r ) i, which remains
able for pairing, and consequently Tc . In a 2D material with N finite at long distances |r0 − r|  ξ in 3D, is in contrast suppressed
layers, the BCS theory predicts that Tc (N) = Tc exp (−1/(Uρ0 N)) in 2D, reflecting a lack of long-range superconducting order,
(Cooper law), with ρ0 the single layer DoS at the Fermi level, −η
|r − r0 |

and U the pairing interaction strength. For a quasi-2D slab of h∆∗ (r)∆(r0 )i = ∆20 in 2D. (7)
r0
thickness d  ξ , the in-plane critical field changes to Hcr,k ∝
Φ0 /(2πξ d), so that the system may support much higher in- Here, η(T ) > 0 is a so-called critical exponent and r0 a cutoff
plane fields than the 3D case. In the extreme 2D limit d → 0, length scale of the order of the coherence length ξ . The super-
Hcr,k formally diverges. In this case, a second critical field ap- conductor correlation function of the gap decays as a power-law
pears, that is related to the pair-breaking action of the applied of distance in 2D. This is a reflection of the Mermin-Wagner the-
field via Zeeman spin polarization. Thus, the critical field in a orem, that states that there cannot exist long-range order in a
strictly 2D BCS superconductor is the Pauli paramagnetic limit, finite-temperature 2D system by spontaneously breaking a contin-
√
Hcr,k = H p ≡ ∆/( 2µB ), with µB the Bohr magneton. Above the uous symmetry. The above power-law decay is still much slower
Pauli field H p , the Zeeman splitting of the Cooper pairs compen- than an ordinary exponential, so we may talk about so-called
sates the energy gained from creating the BCS condensate, and quasi-long-range order in 2D superconductors. In fact, 2D SC
2D SC is suppressed. may survive in the form of a Berezinskii-Kosterlitz-Thouless (BKT)
The Bose condensation of Cooper pairs into the same state im- phase. This phase is a very complex state of matter that will not
plies that the number of pairs (and hence of electrons) ceases to be discussed here.

10 | J
our
nal
Name,
[yea
r][
,vol
.
],1–13
 The fundamental arguments sketched above suggest that 2D
systems should be host to particularly non-trivial superconduct-
ing phases. Superconducting thin films have been the subject of
intense investigation during the last decades of the past century K K0
and most of their basic superconducting properties have been un-
veiled. It has been found that, as their thickness is reduced, thin
films usually exhibit disordered structures, mostly amorphous and
granular, that do not favor SC, or that show unusual behaviours, HSO HSO
such as strong spatial fluctuations of the quasiparticle peaks (see
SOC splitting and Ising pairing
Fig. 6 b). In contrast, 2D crystals fabricated using methods such
as exfoliation are very pure and clean down to atomic scales.
Fig. 7 Schematics of the Ising pairing on the spin-orbit split Fermi
They thus allow us to pursue 2D superconductivity into new terri- surface of TMDs.
tory. The newly emerging field of 2D superconductors 51,52 is now
actively exploring the fundamentally new physics in these sys-
tems and their possible applications. We now present an overview crease the electronic concentration and enhance the electron-
of some of the recent experimental achievements in this domain phonon coupling. The T -dependent Hcr,k (T ) exhibits a positive
with simple theoretical explanations. curvature that is consistent with the behavior of superconductors
made of weakly coupled superconducting layers. This system re-
5.1 Quantum Confinement alises layered SC, in which the vortex lines move between the
layers, allowing the individual layers to remain superconducting
According to BCS theory, finding a quasi 2D superconducting sys-
at much higher fields.
tem at relatively high temperatures would be very unlikely. Ex-
periments have nonetheless demonstrated that superconductivity
5.2 Unconventional properties
persists when reducing the dimensionality of the samples. In par-
ticular, the controlled exfoliation of van der Waals materials like Low dimensionality also allows for the emergence of novel SC ef-
TMDs has made it possible to study the evolution of the supercon- fects associated to the shape of the Fermi surface (which is often
ducting properties of layered systems with the number of layers. nested), the reduction of symmetry, the appearance of van Hove
In NbSe2 a clear trend is reported, 49 whereby Tc decreases mono- singularities and the enhancement of Coulomb interactions, see
tonically upon reducing the number of layers, with a law that ap- Sec. 4.2. One such example is the emergence of Ising supercon-
proximately follows the BCS predictions. SC has been eventually ductivity in system characterized by broken inversion symmetry
reported down to the single layer limit (see Fig. 6a). However, and strong spin-orbit coupling, such as TMDs monolayers.
deviations from the Cooper law have been reported. Ising SC – We have seen that, for 2D crystals, the in-plane
critical field is given by Pauli paramagnetic limit H p . In MoS2
Tc vs thickness – Besides an overall decrease of Tc with sam-
and NbSe2 , 51 it has been shown experimentally that Hcr,|| greatly
ple thickness, experiments in thin slabs of Pb 52 have revealed
exceeds the expected value from the Pauli limit. This behaviour
oscillations of Tc with the number of layers. This phenomenon
has been ascribed to the strong spin-orbit coupling that arises in
is explained as follows. As the thickness of a film is reduced to
monolayer TMDs, see Sec. 2, which produces a strong spin–valley
the nanometer scale, the film surface and interface confine the
locking. The Fermi surface of this materials is formed by pockets
motion of the electrons, leading to the formation of discrete elec-
around the K and K’ points (in the case of MoS2 this is achieved
tronic states known as quantum well states. This quantum size
upon doping). As discussed in Sec. 3, the SOC produces an effec-
effect changes the overall electronic structure of the film and de-
tive Zeeman field HSO with opposite sign at the K and K’ pockets
termines oscillations of Tc . A much more dramatic deviation from
(see Fig. 7). The SOC spin-splits the Fermi pockets and polar-
Cooper law has been reported in a TMD layered material, TaS2 ,
izes the spin along the out-of-plane direction. Singlet supercon-
for which Tc shows an enhancement upon reducing the number
ductivity pairs electron of opposite spin across the Fermi surface,
of layers, 53 contrary to the behavior of NbSe2 . These findings
and Cooper pairs may form either as |K, ↑; K’, ↓i − |K’, ↓; K, ↑i or as
point to an interplay between dimensionality, strong Coulomb re-
|K, ↓; K’, ↑i − |K’, ↑; K, ↓i, resulting in a so-called Ising pairing, as
pulsion and existence of van Hove singularities in the DoS (i.e.
schematically depicted in Fig. 7. An external in-plane field Hk
logarithmic divergences in the DoS associated to saddle points in
tends to tilt the spin towards the plane, in competition with the
the band structure of 2D crystals).
out-of-plane effective SOC field HSO , that tends to keep the spin
Layered SC – Despite many theoretical predictions, supercon- aligned to the out-of-plane direction. This allows greater fields to
ductivity in graphene has long remained elusive. At the Dirac be applied before destroying SC. The in-plane critical field may be
point the DoS is zero, so that SC is not expected at charge neu- estimated by noting that the in-plane component of the spin mag-
trality. The high Fermi velocity at the Fermi level also requires netic moment is reduced to ∼ Hk /HSO . Pair breaking occurs when
strong doping in order to achieve a sizeable DoS and electron den- the modified Zeeman energy ∼ (Hk2 /HSO )µB (known as van Vleck
sity. Superconductivity has been recently reported in graphene paramagnetism) overcomes the superconducting gap. For strong
decorated with alkali metals, 54,55 with the dopants intercalated SOC, an Ising superconductor therefore exhibits an enhanced up-
between effectively decoupled graphene layers. The dopants in- p
per critical field Hcr,|| ' H p HSO .

J
our
nal
Name,
[yea
r][
,vol
.
],
1–13 | 11
6 Summary and Outlook E. Cappelluti and P. Ordejón, Ann. Phys. (Berlin), 2014, 526,
347–357.
Research into 2D crystals, a field born with the discovery of
10 M. Chhowalla, H. S. Shin, G. Eda, L.-J. Li, K. P. Loh and
graphene, is growing at a fast pace with each new atomically thin
H. Zhang, Nature chemistry, 2013, 5, 263–275.
material that is isolated or synthesised. Here, we have reviewed
11 A. N. Rudenko, S. Yuan and M. I. Katsnelson, Phys. Rev. B,
the main features of the currently most studied 2D crystals, in-
2015, 92, 085419.
cluding metals (NbSe2 ), semimetals (graphene), semiconductors
12 J. Quereda, P. San-Jose, V. Parente, L. Vaquero-Garzon, A. J.
(TMDs, phosphorene, etc.) and insulators (h-BN). Individually
Molina-Mendoza, N. Agraït, G. Rubio-Bollinger, F. Guinea,
or combined with other layered materials to form van der Waals
R. Roldán and A. Castellanos-Gomez, Nano Lett., 2016, 16,
heterostructures, 2D materials are demonstrating new physics of
2931–2937.
fundamental interest, but also with significant potential for future
13 T. Low, A. Chaves, J. D. Caldwell, A. Kumar, N. X. Fang,
applications in photonics and nanoelectronics. Some of the new
P. Avouris, T. F. Heinz, F. Guinea, L. Martin-Moreno and
possibilities discussed in this tutorial include the manipulation of
F. Koppens, Nat. Mater., 2017, 16, 182–194.
spin and valley degrees of freedom for application in spintronics
and valleytronics, the generation and confinement of excitons, 14 S. Yuan, E. van Veen, M. I. Katsnelson and R. Roldán, Phys.
the tuning of the optoelectronic properties of 2D materials by Rev. B, 2016, 93, 245433.
strain engineering, or the emergence of novel superconducting 15 C. Dutreix, E. A. Stepanov and M. I. Katsnelson, Phys. Rev. B,
phases, such as Ising superconductivity, only possible in 2D. The 2016, 93, 241404.
intense activity in this field of research promises the discovery of 16 M. Yankowitz, K. Watanabe, T. Taniguchi, P. San-Jose and
new exciting phenomena, that will continue to demonstrate the B. J. Leroy, Nat. Comm., 2016, 7, 13168.
unique physics of crystalline solids as they are reduced to a two- 17 A. K. Geim and I. V. Grigorieva, Nature, 2013, 499, 419–425.
dimensional, atomically thin form. 18 J. D. Caldwell, A. V. Kretinin, Y. Chen, V. Giannini, M. M.
Fogler, Y. Francescato, C. T. Ellis, J. G. Tischler, C. R. Woods,
Acknowledgements A. J. Giles, M. Hong, K. Watanabe, T. Taniguchi, S. A. Maier
We thank A. N. Rudenko and J. L. Lado for kindly providing and K. S. Novoselov, Nat. Comm., 2014, 5, 5221.
band structures of phosphorene, antimonene and MoS2 . We ac- 19 R. Fei, W. Li, J. Li and L. Yang, Appl. Phys. Lett., 2015, 107,
knowledge financial support from the Spanish Ministry of Econ- 173104.
omy and Competitiveness through Grant Nos. FIS2014-58445- 20 P. Ares, F. Aguilar-Galindo, D. Rodríguez-San-Miguel, D. A.
JIN, FIS2015-65706-P and FIS2016-80434-P (MINEICO/FEDER), Aldave, S. Díaz-Tendero, M. Alcamí, F. Martín, J. Gómez-
the Ramón y Cajal programme, Grant Nos. RYC-2011-09345 and Herrero and F. Zamora, Adv. Mater., 2016, 28, 6515–6515.
RYC-2013-14645, the “María de Maeztu” Programme for Units 21 A. N. Rudenko, M. I. Katsnelson and R. Roldán, Phys. Rev. B,
of Excellence in R&D (MDM-2014-0377), the European Union’s 2017, 95, 081407.
Seventh Framework Programme (FP7/2007-2013) through the 22 P. Zhang, Z. Liu, W. Duan, F. Liu and J. Wu, Phys. Rev. B, 2012,
ERC Advanced Grant NOVGRAPHENE (GA No. 290846) and 85, 201410.
European Commission under the Graphene Flagship, contract 23 J. O. Island, M. Buscema, M. Barawi, J. M. Clamagirand, J. R.
CNECTICT-604391. Ares, C. Sánchez, I. J. Ferrer, G. A. Steele, H. S. J. van der
Zant and A. Castellanos-Gomez, Adv. Opt. Mater., 2014, 2,
References 641–645.
1 K. Novoselov, D. Jiang, F. Schedin, T. Booth, V. Khotkevich, 24 J. Silva-Guillén, E. Canadell, P. Ordejón, F. Guinea and
S. Morozov and A. Geim, PNAS, 2005, 102, 10451–10453. R. Roldán, 2D Mater., 2017, 4, 025085.
2 A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov 25 J. Kim, S. S. Baik, S. H. Ryu, Y. Sohn, S. Park, B.-G. Park,
and A. K. Geim, Rev. Mod. Phys., 2009, 81, 109–162. J. Denlinger, Y. Yi, H. J. Choi and K. S. Kim, Science, 2015,
3 A. Castellanos-Gomez, Nat. Photon., 2016, 10, 202–204. 349, 723–726.
4 G.-B. Liu, D. Xiao, Y. Yao, X. Xude and W. Yao, Chem. Soc. Rev, 26 J. Á. Silva-Guillén, P. San-Jose and R. Roldán, Appl. Sci., 2016,
2015, 44, 2643–2663. 284.
5 A. Castellanos-Gomez, J. Phys. Chem. Lett, 2015, 6, 4280– 27 J. L. Lado and J. Fernández-Rossier, 2D Mater., 2016, 3,
4291. 035023.
6 M. I. Katsnelson, Graphene: carbon in two dimensions, Cam- 28 J. C. Slater and G. F. Koster, Phys. Rev., 1954, 94, 1498–1524.
bridge University Press, 2012. 29 X. Xu, W. Yao, D. Xiao and T. F. Heinz, Nat. Phys., 2014, 10,
7 M. O. Goerbig, Rev. Mod. Phys., 2011, 83, 1193–1243. 343–350.
8 B. Amorim, A. Cortijo, F. de Juan, A. Grushin, F. Guinea, 30 H. Rostami, R. Roldán, E. Cappelluti, R. Asgari and F. Guinea,
A. Gutiérrez-Rubio, H. Ochoa, V. Parente, R. Roldán, P. San- Phys. Rev. B, 2015, 92, 195402.
Jose, J. Schiefele, M. Sturla and M. Vozmediano, Phys. Rep., 31 P. San-Jose, E. Prada, E. McCann and H. Schomerus, Phys.
2016, 617, 1 – 54. Rev. Lett., 2009, 102, 247204.
9 R. Roldán, J. A. Silva-Guillén, M. P. López-Sancho, F. Guinea, 32 J. R. Schaibley, H. Yu, G. Clark, P. Rivera, J. S. Ross, K. L.

12 | J
our
nal
Name,
[yea
r][
,vol
.
],1–13
 Seyler, W. Yao and X. Xu, Nat. Rev. Mater., 2016, 1, 16055. F. Guinea, H. S. J. van der Zant and G. A. Steele, Nano Lett.,
33 P. San-Jose, V. Parente, F. Guinea, R. Roldán and E. Prada, 2013, 13, 5361–5366.
Phys. Rev. X, 2016, 6, 031046. 47 W. Wu, L. Wang, Y. Li, F. Zhang, L. Lin, S. Niu, D. Chenet,
34 X. Wang, A. M. Jones, K. L. Seyler, V. Tran, Y. Jia, H. Zhao, X. Zhang, Y. Hao, T. F. Heinz et al., Nature, 2014, 514, 470–
H. Wang, L. Yang, X. Xu and F. Xia, Nat. Nano., 2015, 10, 474.
517–521. 48 H. Tian, J. Tice, R. Fei, V. Tran, X. Yan, L. Yang and H. Wang,
35 D. Kozawa, R. Kumar, A. Carvalho, K. Kumar Amara, W. Zhao, Nano Today, 2016, 11, 763 – 777.
S. Wang, M. Toh, R. M. Ribeiro, A. H. Castro Neto, K. Matsuda 49 Y. Cao, A. Mishchenko, G. L. Yu, E. Khestanova, A. P. Rooney,
and G. Eda, Nat. Comm., 2014, 5, 4543. E. Prestat, A. V. Kretinin, P. Blake, M. B. Shalom, C. Woods,
36 P. Y. Yu and M. Cardona, Fundamentals of semiconductors, J. Chapman, G. Balakrishnan, I. V. Grigorieva, K. S. Novoselov,
Springer, 2005. B. A. Piot, M. Potemski, K. Watanabe, T. Taniguchi, S. J.
37 L. V. Keldysh, JETP Lett., 1979, 29, 658. Haigh, A. K. Geim and R. V. Gorbachev, Nano Lett., 2015, 15,
38 P. Cudazzo, I. V. Tokatly and A. Rubio, Phys. Rev. B, 2011, 84, 4914–4921.
085406. 50 C. Brun, T. Cren, V. Cherkez, F. Debontridder, S. Pons,
39 E. Prada, J. V. Alvarez, K. L. Narasimha-Acharya, F. J. Bailen D. Fokin, M. C. Tringides, S. Bozhko, L. B. Ioffe, B. L. Altshuler
and J. J. Palacios, Phys. Rev. B, 2015, 91, 245421. and D. Roditchev, Nature Physics, 2014, 10, 444–450.
40 A. Castellanos-Gomez, L. Vicarelli, E. Prada, J. O. Island, 51 Y. Saito, T. Nojima and Y. Iwasa, Nat. Rev. Mater., 2016, 2,
K. L. Narasimha-Acharya, S. I. Blanter, D. J. Groenendijk, 16094.
M. Buscema, G. A. Steele, J. V. Alvarez, H. W. Zandbergen, 52 C. Brun, T. Cren and D. Roditchev, Supercond. Sci. Technol.,
J. J. Palacios and H. S. J. van der Zant, 2D Mater., 2014, 1, 2017, 30, 013003.
025001. 53 E. Navarro-Moratalla, J. O. Island, S. Mañas-Valero, E. Pinilla-
41 R. Roldán, A. Castellanos-Gomez, E. Cappelluti and F. Guinea, Cienfuegos, A. Castellanos-Gomez, J. Quereda, G. Rubio-
J. Phys.: Condens. Matter, 2015, 27, 313201. Bollinger, L. Chirolli, J. A. Silva-Guillén, N. Agraït, G. A.
42 F. Guinea, M. Katsnelson and A. Geim, Nat. Phys., 2010, 6, Steele, F. Guinea, H. S. J. van der Zant and E. Coronado, Nat.
30–33. Comm., 2016, 7, 11043.
43 N. Levy, S. Burke, K. Meaker, M. Panlasigui, A. Zettl, 54 B. M. Ludbrook, G. Levy, P. Nigge, M. Zonno, M. Schnei-
F. Guinea, A. C. Neto and M. Crommie, Science, 2010, 329, der, D. J. Dvorak, C. N. Veenstra, S. Zhdanovich, D. Wong,
544–547. P. Dosanjh, C. Straßer, A. Stöhr, S. Forti, C. R. Ast, U. Starke
44 D. Midtvedt, C. H. Lewenkopf and A. Croy, 2D Materials, and A. Damascelli, PNAS, 2015, 112, 11795–9.
2016, 3, 011005. 55 J. Chapman, Y. Su, C. A. Howard, D. Kundys, A. N. Grig-
45 J. Feng, X. Qian, C.-W. Huang and J. Li, Nat. Photon., 2012, orenko, F. Guinea, A. K. Geim, I. V. Grigorieva and R. R. Nair,
6, 866–872. Sci. Rep., 2016, 6, 23254.
46 A. Castellanos-Gomez, R. Roldán, E. Cappelluti, M. Buscema,

J
our
nal
Name,
[yea
r][
,vol
.
],
1–13 | 13

View publication stats