Electronic and Transport Properties of Weyl Semimetals
Dissertation
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University
By
Timothy M. McCormick, B.S.
Graduate Program in Physics
The Ohio State University
2018
Dissertation Committee:
Professor Nandini Trivedi, Advisor Professor Yuri Kovchegov Professor Mohit Randeria Professor Rolando Valdes Aguilar c Copyright by
Timothy M. McCormick
2018 Abstract
Topological Weyl semimetals have attracted substantial recent interest in con- densed matter physics. In this thesis, we theoretically explore electronic and transport properties of these novel materials. We also present results of experimental collabora- tions that support our theoretical calculations. Topological Weyl semimetals (TWS) can be classified as type-I TWS, in which the density of states vanishes at the Weyl nodes, and type-II TWS, in which an electron pocket and a hole pocket meet at a singular point of momentum space, allowing for distinct topological properties. We consider various minimal lattice models for type-II TWS. We present the discovery of a type II topological Weyl semimetal (TWS) state in pure MoTe2, where two sets of WPs (W2±, W3±) exist at the touching points of electron and hole pockets and are located at different binding energies above EF . Using ARPES, modeling, DFT and calculations of Berry curvature, we identify the Weyl points and demonstrate that they are connected by different sets of Fermi arcs for each of the two surface terminations.
Weyl semimetals possess low energy excitations which act as monopoles of Berry curvature in momentum space. These emergent monopoles are at the heart of the extensive novel transport properties that Weyl semimetals exhibit. We show how the Nernst effect, combining entropy with charge transport, gives a unique signature for the presence of Dirac bands. The Nernst thermopower of NbP (maximum of 800
ii µV·K−1 at 9 T, 109 K) exceeds its conventional thermopower by a hundredfold and is significantly larger than the thermopower of traditional thermoelectric materials. The
Nernst effect has a pronounced maximum near TM = 90±20K = µ0/kB (µ0 is chemical potential at T = 0 K). A self-consistent theory without adjustable parameters shows that this results from electrochemical potential pinning to the Weyl point energy at
T ≥ TM , driven by charge neutrality and Dirac band symmetry.
We propose that Fermi arcs in Weyl semimetals lead to an anisotropic magne- tothermal conductivity, strongly dependent on externally applied magnetic field and resulting from entropy transport driven by circulating electronic currents. The circu- lating currents result in no net charge transport, but they do result in a net entropy transport. This translates into a magnetothermal conductivity that should be a unique experimental signature for the existence of the arcs. We analytically calculate the Fermi arc-mediated magnetothermal conductivity in the low-field semiclassical limit as well as in the high-field ultra-quantum limit, where only the chiral Landau levels are involved. By numerically including the effects of higher Landau levels, we show how the two limits are linked at intermediate magnetic fields. This work pro- vides the first proposed signature of Fermi arc-mediated thermal transport and sets the stage for utilizing and manipulating the topological Fermi arcs in experimental thermal applications.
iii To my parents, who inspired my love of science.
iv Acknowledgments
There are many people who have been indispensible during my time in graduate school. First and foremost, I extend my deepest thanks to my advisor, Professor
Nandini Trivedi. Her guidance and support were invaluable during my time here and she was a superb role model for how one should attack an unstructured problem. I can only hope that I picked up some of her physical insight during my time here. I am particularly indebted to her for the many opportunities that I had to collaborate with excellent experimentalists and theorists, as well as to attend several truly amazing conferences in the US and abroad.
My time here would have been much less productive had it not been for the won- derful collaborations that I have been apart of. I thank Professor Adam Kaminski for the opportunity to collaborate on the ARPES discovery of type-II Weyl semimetal
MoTe2. I am deeply grateful to Professor Jos Heremans and his student Sarah Watz- man for our many collaborations on transport phenomena in Weyl semimetals. From them I learned how truly messy yet rewarding studying transport can be. Much of this thesis was inspired by their excellent experiments. I am thankful to Professor Mohit
Randeria, whose precise questions always got to the heart of a matter and were the source of many illuminating discussions on anomalous transport in Skyrmions. I also had the pleasure of collaborating with a very hard-working undergraduate student,
Robert McKay, who was as ideal a mentee as I could ask for.
v The Trivedi and Randeria groups were close to a second family to me during my time here. I follow in the footsteps of Will Cole, Mason Swanson, Eric Duchon,
Onur Erten, and Nganba Meetei, who all helped to shape my physical intuition. I am particularly thankful for their infinite patience in answering the many questions of a new member of the group. I thank Hasan Khan, James Rowland, Tamaghna
Hazra, Po-Kuan Wu, Wenjuan Zhang, Tim Gao, Kyungmin Lee, Kyusung Hwang,
David Nozadze, Mehdi Kargarian, and Sumilan Banerjee for their discussions and companionship. I was also fortunate to have the best office-mates that I could ask for in Blythe Moreland and Jiaxin Wu.
The condensed matter theory group is particularly lively at Ohio State. My time here was greatly enriched by the many interactions with other students, postdocs, and professors. I extend my thanks to Professor Ilya Gruzberg, Professor Jason Ho, and Professor Yuan-Ming Lu for teaching several excellent classes. I always found their doors open for questions and discussion.
My friends and family provided a backbone of support while I completed my dissertation. I thank Jon Zizka, Andrew Hausman, Jake Kerrigan, Drew Gallagher,
Chris Wolfe, Marci Howdyshell, Adam Ahmed, Simran Singh, Jyoti Katoch, Igor
Pinchuk, Chris Ehemann, Dennis Bazow, and Dante O’Hara for always always being there for me. I thank my wife, Beth McCormick, for her love and support. I also thank my parents for their constant love and encouragement.
Last but certainly not least, I would like to thank the NSF and the Center for
Emergent Materials for funding the majority of my research. I also acknowledge the
OSU presidential fellowship for funding my final year.
vi Vita
January 24, 1990 ...... Born - Wilmington, DE, USA
2012 ...... B.S. Physics
Publications
Research Publications
Timothy M. McCormick, Sarah J. Watzman, Joseph P. Heremans, Nandini Trivedi, “Fermi arc mediated entropy transport in topological semimetals” Phys. Rev. B 97, 195152 (2018). Sarah J. Watzman, Timothy M. McCormick, Chandra Shekhar, Shu-Chun Wu, Yan Sun, Arati Prakash, Claudia Felser, Nandini Trivedi, Joseph P. Heremans, “Dirac dispersion generates unusually large Nernst effect in Weyl semimetals” Phys. Rev. B 97, 161404 (2018). Timothy M. McCormick, Robert C. McKay, Nandini Trivedi, “Semiclassical the- ory of anomalous transport in type-II topological Weyl semimetals” Phys. Rev. B 96, 235116 (2017) Joseph R. Smith, Amber Byrum, Timothy M. McCormick, Nathan Young, Christo- pher Orban, and Christopher D. Porter, “A controlled study of stereoscopic virtual re- ality in freshman electrostatics” Physics Education Research Conference Series 2017, 363 (2017). Timothy M. McCormick, Itamar Kimchi, Nandini Trivedi, “Minimal models for topological Weyl semimetals” Phys. Rev. B 95, 075133 (2017). Lunan Huang, Timothy M. McCormick, Masayuki Ochi, Zhiying Zhao, Michi-to Suzuki, Ryotaro Arita, Yun Wu, Daixiang Mou, Huibo Cao, Jiaqiang Yan, Nandini
vii Trivedi, Adam Kaminski, “Spectroscopic evidence for type II Weyl semimetal state in MoTe2” Nature Materials 15, 1155-1160 (2016). Timothy M. McCormick, Nandini Trivedi, “Tuning the Chern number and Berry curvature with spin-orbit coupling and magnetic textures” Phys. Rev. A 91, 063609 (2015). Fields of Study
Major Field: Physics
viii Table of Contents
Page
Abstract ...... ii
Dedication ...... iv
Acknowledgments ...... v
Vita...... vii
List of Tables ...... xii
List of Figures ...... xiii
1. Introduction ...... 1
2. Topology in Solid State Systems ...... 5
2.1 Berry Phase ...... 6 2.1.1 Berry Curvature ...... 8 2.1.2 Chern Number ...... 10 2.2 Chern Insulators ...... 11 2.2.1 Berry Curvature of a 2-Band Hamiltonian ...... 12 2.2.2 Chern Insulator on the Square Lattice ...... 13 2.2.3 Chiral Edge Modes ...... 15
3. Introduction to Weyl Semimetals ...... 18
3.1 Weyl Fermions ...... 19 3.1.1 Weyl Fermions in Quantum Materials ...... 21 3.2 Lattice Models for Weyl Semimetals ...... 26 3.3 Type II Weyl Semimetals ...... 30
ix 3.3.1 Time Reversal Breaking Model ...... 33 3.3.2 Inversion Breaking Model ...... 42 3.3.3 Surface States: Topological and Track ...... 46 3.3.4 Comparison with Experiments ...... 53 3.3.5 Conclusions ...... 55
3.4 Experimental Discovery of Weyl Semimetal MoTe2 ...... 56 3.4.1 ARPES Results ...... 59 3.4.2 DFT and Topological Analysis ...... 65
4. Thermoelectric Transport in Weyl Semimetals ...... 67
4.1 Boltzmann Transport Theory ...... 68 4.1.1 Definition of Transport Coefficients ...... 68 4.1.2 Boltzmann formalism ...... 71 4.2 Nernst Effect of Isotropic Weyl Nodes ...... 75 4.2.1 Numerical Results ...... 80 4.2.2 Analytic Results ...... 81 4.2.3 Nernst Thermopower in Weyl Semimetal NbP ...... 83 4.3 Anomalous Transport in Type-II Weyl Semimetals ...... 86 4.3.1 Model ...... 88 4.3.2 Anomalous Transport ...... 94 4.3.3 Relation to Measurable Quantities ...... 102 4.3.4 Discussion and Conclusion ...... 103
5. Fermi Arc-Mediated Entropy Transport in Weyl Semimetals ...... 106
5.1 Fermi arc-mediated magnetothermal transport ...... 110 5.1.1 Model ...... 110 5.1.2 Semiclassical Regime ...... 111 5.1.3 Effect of disorder ...... 119 5.1.4 Ultra-quantum Regime ...... 120 5.1.5 Intermediate regime ...... 123 5.2 Comparison with bulk thermal conductivity of WSM ...... 126 5.3 Fermi Arc-Mediated Entropy Transport in Dirac Semimetals . . . . 128 5.4 Discussion and Summary ...... 129 5.5 Conclusion and further directions ...... 133
Appendices 135
A. Isothermal Heat Transport and the Adiabatic Nernst Effect ...... 135
x A.1 Heat transport under isothermal conditions ...... 135 A.2 The adiabatic Nernst effect ...... 137
B. Landau Levels of a Weyl node ...... 140
C. Scattering in Weyl semimetals ...... 143
D. Specific Heat of Weyl Nodes ...... 145
D.1 Semiclassical regime ...... 145 D.2 Ultra-quantum regime ...... 146 D.3 Crossover regime ...... 147
Bibliography ...... 148
xi List of Tables
Table Page
3.1 Experimental realizations of Weyl semimetals...... 54
3.2 The locations (kx, ky, E) of the Weyl points from DFT and ARPES for weyl semimetal MoTe2...... 66
4.1 Summary of properties of anomalous transport coefficients in type II Weyl semimetals...... 100
5.1 Summary of temperature and magnetic field dependence of arc-mediated
κzzz in the semiclassical and ultra-quantum limits as well as the bulk semiclassical magnetothermal conductivity...... 126
5.2 Several topological semimetal candidates for Fermi arc-mediated en-
tropy transport. Since Cd3As2 and Na3Bi are Dirac semimetals, the number of Weyl nodes Np reported is double the number of Dirac nodes. 132
xii List of Figures
Figure Page
2.1 Band structure for the Hamiltonian given in Eqn. (2.19) for B = t and various M: (a) M = −4 where we see the transition between the trivial state and the first topologically nontrivial sector. (b) M = −2 where we see the transition between the two distinct topological regimes. (c) M = 0 showing a gap closing at the final topological tranisition to the trivial state. (d) M = 1 where we see the gapped topological trivial sector. As M increases, this gap grows and the bands flatten...... 14
2.2 Berry curvature (in units of a2 of the lower band and edge states for the Hamiltonian given in Eqn. (2.19) for B = t and various M. Edge states are calculated using strip geometry, finite in the y-direction with N = 50 layers. In the strip calculations, states localized to the top edge are colored red, states localized to the bottom surface are colored blue, and states in the bulk are colored black. (a) Berry curvature averages to zero over the either band for M = −5 in one of the topologically trivial sectors. (b) Berry curvature in the lower band M = −2.5 where
the Chern number is given by c− = +1. (c) Berry curvature in the lower band M = −1.5 where the Chern number is given by c− = −1. (d) Berry curvature is again zero in the trivial sector where M = 1. (e) Strip geometry calculation for M = −5 in the trivial sector showing a lack of edge states. (f) Edge states for M = −2.5 crossing zero energy
at the kx = ±π point. (f) Edge states for M = −1.5 crossing zero energy at the kx = 0 point. We note that the chirality, or direction, of the edge state on a given surface changes between (e) and (f) due to the change in Chern number. (g) Strip geometry calculation for M = 0 in the trivial sector showing a lack of edge states...... 16
3.1 (a) Continuum dispersion of a Weyl node. (b) Dependence of the chemical potential on temperature µ(T ). We see that on a temperature
scale TW ∼ µ(T = 0)/kB, the chemical potential reaches the Weyl node. 25
xiii 3.2 Band structure for Weyl semimetal in Eqn. (3.19) for tx = 0.5t, m = 2t, k0 = π/2, and lattice constant set to unity a = 1. (a) Bulk energy dispersion with ky = 0. (b) Cut through the Weyl nodes along kx with ky = kz = 0. (c) Constant energy EF = 0.2t cut for a system with slab geometry with N = 50 layers in the y-direction. Surface states are colored red (top) and blue (bottom). We see that small bulk Fermi pockets (shown in black) enclosing the Weyl points (green points) are connected by the Fermi arcs on the top and bottom states...... 28
3.3 Bulk band structure for the “Hydrogen atom” of type-I and type-II Weyl semimetal. a-c The bulk band structure for the Hamiltonian in Eqn. (3.25). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown in green. Here
we have chosen parameters ky = 0 with parameters k0 = π/2, tx = t, m = 2t for (a) type-I Weyl semimetal with γ = 0, (b) the critical point between type-I and type-II Weyl semimetal with γ = 2t and (c) type-II Weyl semimetal with γ = 3t. The cones comprising the Weyl nodes develop a characteristic tilt of the type-II TWS as γ is increased. d-f
Cuts through the Weyl nodes at ky = kz = 0 for the same parameters as (a-c). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). We see that for a type-I TWS, there are no states at the Fermi energy. At the critical point between a type- I and type-II TWS, we see lines of bulk states appearing between the nodes. These lines open into bulk hole and electron pockets (in the repeated zone scheme) when the system becomes a type-II TWS. . . 35
xiv 3.4 Fermi surface and arc configuration for the “Hydrogen atom” of type-I and type-II TWS. a-c Bulk Fermi surfaces and surface Fermi arcs for a type I TWS with the same bulk parameters as in Fig. 3.3a,d,g calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the following constant energy: (a) E = −0.2t, (b) E = 0, (c) E = 0.2t. We color the states which are exponentially localized to the y = 1 (y = L) surface red (blue) and note that such surface states form topological arcs connect- ing the two Weyl nodes (shown as green dots and marked with pink arrows). We note that at E = 0 the two Fermi arcs are degenerate
along kz = 0 and we color them purple to signify this. d-f Bulk Fermi surfaces and surface Fermi arcs for a type-II TWS with the same bulk parameters as in Fig. 3.3c,f,i calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the same constant energies as above: (d) E = −0.2t, (e) E = 0, (f) E = 0.2t. 37
3.5 Bulk band structure for type-I and type-II TRB model with separate pockets (the “Helium atom”). a-c The bulk band structure for the Hamiltonian in Eqn. (3.26). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown
in green. Here we have chosen parameters ky = 0 with the parameters k0 = π/2, tx = t, m = 2t and γx = t/2 for (a) type-I TWS with γ = 0, (b) type-II TWS with γ = t and (c) type-II TWS with γ = 1.5t. The cones comprising the Weyl nodes again develop a characteristic tilt of the type-II TWS as γ is increased. d-f Cuts through the Weyl nodes
at ky = kz = 0 for the same parameters as (a-c). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). Note that for a type-I TWS, there are no states at the Fermi energy while in the type-II regime, there are two sets of electron and hole pockets on either side of the Weyl nodes. We see that unlike the Hydrogen-atom model, the Helium-atom model has disjoint pairs of electron and hole pockets and a pair of each meet at the two Weyl nodes...... 40
xv 3.6 Fermi surface and Fermi arc configuration for type I and type- II time-reversal-breaking model with separate pockets (the “Helium atom”). a-c Bulk Fermi surfaces and surface Fermi arcs for a type-II Weyl semimetal given by Eqn. (3.26) with the same bulk parameters as in Fig. 3.5b,e,h calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the constant energies: (a) E = −0.2t, (b) E = 0, (c) E = 0.2t. As in Fig. 3.4, we color the states that are exponentially localized to the y = 1 (y = L) surface red (blue) and note that such surface states form topological arcs connecting the two Weyl nodes (shown as green dots). We note unlike in Fig. 3.4, each node is isolated in its own hole (a) or electron (c) pocket when the chemical potential is away from E = 0. These pockets are connected by arcs confined to the surface in the y- direction. However, in this type-II TWS the Fermi pockets enclosing a Weyl node can be quite extended, unlike a type-I TWS, the arcs can terminate on a pocket quite far away from the projection of the nodes. d-f Bulk Fermi surfaces and surface Fermi arcs for a type-II TWS with the same bulk parameters as in Fig. 3.5c,f,i calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the same constant energies as above: (d) E = −0.2t, (e) E = 0, (f) E = 0.2t. We see that as the tilt grows, so do the pockets enclosing the nodes. We note that a trivial electron pocket appears
around the (kx, kz) = (π, π) point. This pocket encloses no Weyl nodes and so is not connected via Fermi arcs to any other pockets...... 41
xvi 3.7 Bulk band structure for type-I and type-II inversion break- ing TWS. a-c The bulk band structure for the Hamiltonian in Eqn. (3.28). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown in green. Here we have chosen param-
eters ky = 0 with the parameters k0 = π/2, tx = t/2, m = 2t for (a) type I TWS with γ = 0, (b) the critical point between a type-I and a type-II TWS with γ = 2t and (c) type-II TWS with γ = 2.4t. The cones comprising the Weyl nodes develop a characteristic tilt of the type-II Weyl node as γ is increased. d-f Cuts through the Weyl nodes
at ky = 0 and kz = −π/2 for the same parameters as (a-c). These cuts are shown as the green lines in (g-i). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). We see that for a type-I Weyl semimetal, there are no states at the Fermi energy. At the critical point between a type-I and type-II TWS, the density of states still vanishes. In the type-II regime, electron and hole pockets form near the Weyl nodes. These pockets enclose the projections of the Weyl nodes when the chemical potential is shifted away from E = 0. Trivial pockets also appear at k = (0, 0, 0) and k = (0, 0, π)...... 45
3.8 Fermi surface and Fermi arc configuration for type-I and type- II inversion-breaking Weyl semimetal. a,b The Fermi surface and Fermi arc configuration for the Hamiltonian given in Eqn. (3.28) in the type-I limit (γ = 0) calculated in a slab geometry with L = 50 layers and with bulk parameters the same as in Fig. 3.7a,d,g. We show this calculation at constant energies: E = −0.25t (a) and E = 0.25t (b).
Here we see that Weyl nodes located at (kx, kz) = (±π/2, ±π/2) are connected by surface states (red and blue lines) to one of opposite
chirality across the Brillouin zone in the kx-direction. c,d The Fermi surface and Fermi arc configuration for the Hamiltonian given in Eqn. (3.28) in the type II limit (γ = 2.4t) calculated in a slab geometry with L = 50 layers and with bulk parameters the same as in Fig. 3.7c,f,i. We show these for the same constant energies as above: (c) and E = 0.25t (d). The locations of the Weyl nodes are marked with pink arrows. We term the exponentially localized surface states that form closed loops “track states”. Fermi arcs are shown as bold lines and connect Weyl
nodes in the kz-direction...... 47
xvii 3.9 Evolution of Fermi surface and Fermi arc configuration for inversion-breaking Weyl semimetal as a function of γ. a-d The evolution of the Fermi surface and Fermi arc configuration in a slab geometry for Eqn. (3.28). Bulk states are down in black, surface states are shown in red and blue. We have chosen the parameters
k0 = π/2, tx = t/2, m = 2t. The calculations are done at constant energy E = −0.25t for γ = 0 (a), γ = 0.8t (b), γ = 1.4t (c), and
γ = 2. (d) shown in an extended Brillouin zone where both kx and kz range from −1.5π to 1.5π. We see that at the critical point between a type-I and type-II (d), the Fermi arcs that previously connected
Fermi pockets in the kx-direction now connect Fermi pockets in the kz-direction and track states have formed on the bottom surface (blue) around the (kx, kz) = (π, π) point...... 48
3.10 Sketch of the three types of surface states in a topological Weyl semimetal. a Two type-I Weyl nodes of opposite chirality connected by a Fermi arc on the top (red) and bottom (blue) surfaces. In an arbitrary type-II TWS at an energy away from the Weyl energy, these arcs would connect Fermi pockets instead of nodes. b A sin- gle Fermi pocket enclosing two nodes of opposite chirality. Since no Gaussian surface can be constructed in a region that is both gapped and encloses only one node, the only possible surface states are trivial ones, shown in red and blue at the boundary of the pocket that hy- bridize with bulk states due to lack of topological protection. c Pairs of Weyl nodes, two of each chirality with each node surrounded by a Fermi pocket. The pockets are connected by Fermi arcs (thinner red and blue contours) as well as track states (thicker blue lines) on the bottom surface. Note that states on opposite sides of a given loop of track states will disperse in opposite directions and so a Gaussian sur- face enclosing a given Fermi pocket will still have one net surface state of each chirality...... 51
xviii 3.11 Simple model of type II Weyl semimetal described by a two band model given by Eq. 3.28 which exhibits four Weyl nodes. a Electronic band structure for µ = ±0.1t indicated by the blue translucent plane. b,c The topological surface states and Fermi arcs on surface A (in red) and B (in blue) are calculated for a slab geometry confined along the y-direction. The bulk bands are shown in black. When µ = 0 exactly, the electron and hole pockets touch and the arcs terminate on the node (green dot) itself. For Fermi energy below (above) the nodal energy, arcs of surface states connect the Fermi hole (electron) pockets surrounding a node rather than terminating on a node. d,e Energy
dispersion along kz at fixed kx as shown by cuts in panels (b, c). Cut 1 along kx = π/2 shows the bulk electron and hole bands touching at the node and the merging of surface states into the bulk away from
the Weyl node. Cut 2 along kx = 0.63π shows a gap between the bulk bands and a surface state that disperses with opposite velocities at
the projections of the two WPs. The WPs are located at (kx, kz) = (±π/2, ±π/2) indicated by pink arrows pointing to green dots. . . . . 58
3.12 Experimental Fermi surface and band structure of MoTe2. a Constant energy intensity plot measured at EF using 6.7 eV photons for a sample with termination A. The calculated (DFT) positions of Weyl points W2 are marked as pink dots, while experimentally determined locations of
W2 and W3 points are marked as red dots. The chiralities of Weyl points are marked with “+” and “-” and their locations (kx, ky, E) are summarized in Table 3.4.2. b Same as in a above but taken at 10 meV
above EF . c A sketch of constant energy contours of electron and hole bands showing the locations of Weyl points and Fermi arcs. d Constant
energy contour measured at 30 meV above EF using 5.9 eV photons for a sample with termination B. Positions of calculated and measured Weyl points are marked as above. e Same surface termination and
photon energy as d but at 30 meV below EF . f - i Experimental band dispersion along cuts at kx = 0.24, 0.28, 0.32 and 0.36 π/b. j - m Calculated band dispersion for a sample with termination A along kx = 0.24, 0.28, 0.32 and 0.36 π/b...... 60
xix 3.13 Identification of Weyl points and Fermi arcs from experimental data.
a Constant energy contour at EF , measured by 6.7 eV photons for surface termination A. DFT predicted locations for Weyl points W2 and measured Weyl points W2, W3 are marked as red and pink dots respectively. b The same panel as a except for surface termination B. c The same panel as b except for using 5.9 eV photons. d - g Energy
dispersion for surface termination A along ky = 0, 0.05, 0.10 and 0.20 π/a. The projections of Weyl points W2 are marked as dots. h - k The same panels as (d - g) except for surface termination B. l - o The same panels as (h - k) except for using 5.9 eV photons. p - s
Calculated band dispersion for surface termination A along cuts at ky = 0, 0.05, 0.10 and 0.20 π/a. Positions of W2 are marked similarly as above. t - w The same as (p - s) except for surface termination B. Bands plotted with darker lines have more surface weights...... 62
3.14 Results of DFT calculations. a Calculated bulk Fermi surface of MoTe2 for kz = 0.6π/c and projections of W2 (kx, ky) = (±0.17 π/b, ±0.06π/a) are marked with pink dots. b Bulk band dispersion along W2-W2 direction (the vertical dashed line in a). DFT predicted positions of
W2 (ky, E) = (±0.06π/a, 0.028 eV) are marked. c The dominant DD DD contribution for the divergence of the Berry curvature (Ωn,yz, Ωn,zx) for the n = N + 1 th band where N is the number of electrons in
the unit cell with kz = 0. Red and blue indicate different chiralities of the two Weyl points. d - g Calculated constant energy contours
of MoTe2. Darker bands are surface bands and lighter bands are bulk bands. d, e are at Fermi level for surface termination A and B. f, g are at Fermi level + 28 meV of surface termination A and B, respectively.
h, i Surface band dispersions of termination A and B along W2-W2 direction. j, k Surface band dispersions of termination A and B along
ky = 0.05 π/a direction, which is very close to the ky position of W2 (0.06 π/a). Positions of calculated Weyl points W2 are marked and darker bands have more surface weights in d - k...... 64
4.1 Geometry for measuring the Nernst effect αxyz. Temperature gradient ∇rT , electric field E, and magnetic field B are all mutually perpendicular. 75
xx 4.2 Magnetic field dependence of Nernst thermopower αxyz, and temperature dependence of Nernst coefficient Nxyz (A) Data for the Nernst thermopower plotted as a function of applied magnetic field at the various temperatures indicated. The insert shows the mag- netic field dependence of the Nernst voltage measured at 4.92 K in a temperature gradient of 2.17 mK. SdH oscillations are plainly visible that correspond to the periods measured in the magnetization. (B) Nernst coefficient plotted as a function of temperature, with low-field Nernst effect in red and high-field Nernst effect in blue. The low field curve peaks near 50K; the high-field curve peaks near 90K, which is the temperature at which the chemical potential touches the Weyl nodes. Error bars represent a 95% confidence interval on the standard devia- tion of the systematic error, excluding geometrical error on the sample mount itself...... 79
4.3 Nernst coefficient Nxyz as a function of temperature calculated using Eqn. (4.47) using the temperature dependent chemical poten- tial found by solving Eqn. (3.10) self-consistently. We see agreement with the numerical results above and we note that the temperature
dependence of Nxyz is independent of τ...... 82
4.4 Cuts through the band structure given by the Hamiltonian in Eqn.
(4.48). In (a-d), we show energy versus kz cuts for kx = ky = 0. Here we have chosen m = 3t; tz = t; k0a = π/2; γz = 0.5t for γ = 0 (a), γ = 1.2t (b), γ = 2t (c), and γ = 2.8t (d). In (e-h), we show constant energy cuts for the band structure defined by Eqn. (4.48). (a) and (e) are in the type-I limit; (b) and (f) are in the type-II limit with distinct pockets making up each nodes; (c) and (g) are in the type-II limit after the electron pockets have merged; and (d) and (h) are in the type-II regime where the Weyl nodes share only a single electron and single hole pocket. Thus, as γ is increased we can see the successive Lifshitz transitions described in the text...... 90
xxi 4.5 Each column corresponds to a particular γ with: γ = 0 (a, e, i), γ = 1.2t (b, f, j), γ = 2t (c, g, k), and γ = 2.8t (d, h, l). (a-d) show the net net Berry curvature in the z-direction Ωz (E) defined by Eqn. (4.50). We see that for nonzero γ, the net Berry curvature around the nodes
is of the same sign. (e-h) the density of states for m = 3t, tz = t, k0a = π/2, and γz = 0.5t, for different values of the tilt parameter. (i-l) illustrates the temperature dependence of the chemical potential,
µ(T ). Each plot shows three separate values of EF : EF = 0 (blue), EF = 0.1 (green), and EF = 0.2 (red). We see that for smaller values of γ, g(E) has a minimum close to the Weyl energy E = 0, but for larger values of γ, this minimum shifts far from the nodes. This has a strong effect on the shift of the chemical potential with temperature. 93
EE ET 4.6 In (a-c), we plot each anomalous transport coefficient Lxy , Lxy , and TT Lxy for the Hamiltonian given by Eqn. (4.48) with parameters m = 3t, tz = t, k0a = π/2, and γz = 0.5t, as a function of γ for the following temperatures: T = 0.05t (purple), T = 0.1t (blue), T = 0.15t (green), EE ET TT and T = 0.2t (red). In (d-f), we show Lxy , Lxy , and Lxy for the same values as in (a)-(c), with γ = 1.2t, plotted as functions of temperature
for various Fermi energies: EF = 0 (black), EF = 0.1t (magenta), and EF = 0.2t (blue)...... 96
5.1 (a) Mixed real-space and momentum-space depiction of a Weyl semimetal in a slab geometry with thickness L in the z-direction. Bulk Weyl nodes
are labeled with their chirality χ = ± and separated in the kx-direction. The projections of the Weyl nodes on the surface Brillouin zone define the end points of the Fermi arcs on the top and bottom surfaces. (b) Schematic of the “conveyor belt” motion of charge leading to a net heat flux. When B is aligned in the z-direction in (a), charge current den- sity circulates Je in the clockwise direction shown in a mixed real and momentum space orbit. When ∇T is also aligned in the z-direction, this circulation of charge leads to a net flow of heat current density JQ
in the direction shown. Unit tangent vectors et are shown for the arcs. 108
xxii 5.2 (a) Thermal occupations for the minimal model in Eqn. (5.2) with
m = 2t, tz = t, and kW a = π/2 for cuts along ky at fixed kxa = 0.75π (top and bottom) and kxa = 0.5π (bulk). We show the occupations of the top two layers (top), the bottom two layers (bottom) and the other layers (bulk), weighting each point’s color and thickness in the figure
with the Fermi function f0 at that layer for a temperature gradient of dT 0.8t = at an average temperature of kBT = 0.6t. (b) Orbits for the dz kB a minimal model in Eqn. (5.2) with the same parameters as in (a) for
cuts along kx at ky = 0. We again show the occupations of the top two layers (top), the bottom two layers (bottom) and the other layers (bulk), weighting each point as in (a) with the same temperature and temperature gradient. We show the mixed real- and momentum-space orbits and see that the the states most occupied on the top and bottom are the arcs, while the bulk states merge into the arc and carry heat through the bulk...... 112
5.3 (a) Schematic of the Landau levels. Chiral n = 0 Landau levels are shown in red. The sign of the slope of the n = 0 Landau levels is posi- tive (dashed line) for χ = 1 nodes and negative (solid line) for χ = −1. Non-chiral Landau levels (n 6= 0) are shown in black. (b) Specific heat for a single pair of Weyl nodes in the intermediate quantum limit as a
function magnetic field. We have set the Landau level cutoff Nmax = 50 and we have considered an electron density such that nB = 1. (c) Mag- netothermal conductivity κzzz for a single pair of Weyl nodes in the intermediate quantum limit as a function of magnetic field. We have
set the Landau level cutoff Nmax = 50 (see Appendix C) and n = 1. We have fixed the temperature such that kB T = 0.02 (blue), kB T = 0.04 ~vF Λ ~vF Λ (green), and kB T = 0.06 (red). At higher temperatures (red), we see a ~vF Λ crossover between the linear low-field behavior and the quadratic field dependence when the ultra-quantum limit is reached. At lower tem- peratures (green, blue) only the lowest Landau level is populated in the field range shown. (d) Fermi arc-mediated magnetothermal con- ductivity for a single pair of Weyl nodes in the intermediate quantum limit as a function of magnetic field. Parameters are the same as in (b,c)...... 125
xxiii 5.4 Schematic of the different cases for surface states in a Dirac semimetal with a single pair of Dirac nodes. Green dots represent projections of the Dirac nodes on the surface Brillouin zone. Red contours are surface states on a single surface. Purple filled regions represent projections of bulk Fermi pockets which enclose Dirac nodes. In (a-b), the Fermi energy is at the Dirac nodes, while in (c-d) the Fermi energy is doped away from the Dirac nodes. (a) A pair of surface Fermi arcs terminates on the Dirac nodes on a single surface. (b) A perturbation containing the full symmetry of the full Dirac Hamiltonian which does not shift the bulk Dirac nodes can deform the Fermi arcs into a closed Fermi pocket on the surface, much like a topological insulator. (c) When the Fermi energy is shifted away from the Dirac nodes, the surface states may still form a closed Fermi pocket. (d) Sufficient doping may cause the Fermi pockets to grow large enough for the surface states to merge with the bulk pockets...... 130
xxiv Chapter 1: Introduction
The intersection of topology and the band theory of solids has attracted extensive recent interest in condensed matter physics. Topological insulators [1, 2] opened the door to novel phases of matter with bulk energy gaps and protected, gapless edge states. Weyl semimetals[3, 4, 5] were the first topological material that were no longer constrained by a bulk energy gap that protected the topological nature of the energy bands. In this thesis, we predict new electronic and transport properties of
Weyl semimetals. We also present experimental discoveries of the first type-II Weyl semimetal and the first measurement of the Nernst effect in a Weyl semimetal for which we have provided theoretical support. For the remainder of this chapter, we shall outline the historical discovery of Weyl semimetals and provide context and motivation for our work.
Weyl first predicted his massless fermions as candidates for the neutrino[6], how- ever the discovery of the neutrino mass[7, 8] eliminated this possibility. It was decades until Weyl fermions were again proposed to exist, this time as low-energy, linearly dispersing modes in pyrochlore iridates[9]. These antiferromagnetic materials with strong electron-electron interactions and large spin-orbit coupling attracted intense experimental interest searching for these Weyl fermions, however to this date, no convincing signatures have been found in this class of materials.
1 Other theoretical predictions for the emergence of Weyl fermions were made for
superlattices of topological insulators and trivial insulators [10], magnetic spinels
[11], and in liquid 3He [12]. Further density functional theory predictions of Weyl
semimetals were made, particularly in the TaAs class of materials[13, 14] where a
Weyl semimetal was discovered experimentally in TaAs [15, 16, 17] by angle-resolved
photo-emission spectroscopy (ARPES).
In a Weyl semimetal, the linearly dispersing Weyl nodes are monopoles of Berry
curvature, which acts like a momentum-space magnetic field. The Berry curvature
is responsible for the topological nature of these materials. We devote Chapter 2 to
a pedagogical introduction of topology in quantum materials as well as the Berry
curvature and associated topological quantities such as the Berry connection and
phase. We also introduce the Chern insulator, a 2D gapped phase with nontrivial
topology which we will see plays a key role in understanding Weyl semimetals.
Like topological insulators, Weyl semimetals also have highly nontrivial surface
states. These so-called Fermi arcs are perhaps the most fascinating aspect of these
materials. In Chapter 3, we formally introduce Weyl fermions as solutions to the
Dirac equation and show that, on a lattice, these Fermi arcs must exist. In Chapter
3, we also introduce the tilted type-II class of Weyl semimetals and introduce the first
lattice models describing this class of materials. We also present the first experimental
evidence for a Weyl semimetal in MoTe2, where our theoretical work plays a central role.
The bulk Berry curvature of Weyl fermions is known to result in a plethora of unique transport phenomena in Weyl semimetals. In parallel magnetic and electric
fields, Weyl semimetals exhibit negative longitudinal magnetoresistance as a result
2 of the chiral anomaly[18, 19, 20, 21, 22, 23, 24]. In Chapter 4, we predict that the thermoelectric Nernst response of Weyl semimetals contains a key signature that allows for a quantitative determination of the Fermi energy. The Nernst effect is the transverse conversion of a thermal gradient to a voltage and has technological applications [25]. In Chapter 4, we also present the first experimental study of the
Nernst thermopower in a Weyl semimetal and show that this matches our prediction quantitatively. The Berry curvature of time-reversal breaking Weyl semimetals is also known to lead to an anomalous Hall coefficient[26]. In Chapter 4, we show the first theoretical predictions for how the tilt of a type-II Weyl semimetal affects the anomalous Hall coefficient, the anomalous transverse thermoelectric coefficient, and the anomalous thermal Hall coefficient. These results also have broad possible technological applications.
The role of Fermi arcs in transport was first predicted to lead to quantum oscil- lations in conductivity that are dependent on the thickness of a finite slab of Weyl semimetal [27]. Putative signatures of these mixed bulk and surface coherent os- cillations have been observed in Dirac semimetal Cd3As2 [28]. However, there are limitations to detecting and utilizing these mixed bulk and arc oscillations since they rely on quantum coherence across the orbits. Motivated by these previous works, in
Chapter 5, we propose a novel Fermi-arc mediated pathway for thermal transport in Weyl semimetals. Unlike the Shubnikov-de-Haas oscillations discussed above, the
Fermi-arc mediated thermal conductivity that we predict does not rely on quantum coherence. Signatures of this heat transport should be more readily detectable due to the characteristic sample thickness independent conductance. Our predictions open
3 the door to the possibility utilizing this pathway in technological applications such as magnetically tunable thermal switches.
4 Chapter 2: Topology in Solid State Systems
Beautiful examples of symmetry abound in the natural world, ranging from closely spherical stars to the spectacular six-fold patterns of snowflakes. Symmetry also proves to be a strikingly useful tool in theoretical physics. In condensed matter physics, where one considers vast collections of matter organized in a plethora of states, symmetry plays an especially crucial role in the classication of states of matter.
The Ginzburg-Landau[29] theory of phase transitions characterizes a stable phase of matter with a local order parameter, which is zero in the disordered phase but nonzero in the ordered phase. Different ordered phases can be organized by how their local order parameters transform under symmetry operations. Prior to 1980, this phenomenological theory of phase transitions, along with microscopic many-body theories, formed the backbone of the study of condensed matter physics.
The 1980 discovery of the quantum Hall effect (QHE) [30] shattered this paradigm and presented the first example of a system that had no spontaneously broken symme- try. Instead, its behavior depends only on a topological invariant that is independent of the system’s geometry or microscopic properties. In the QHE, the transverse cur- rent is characterized by a quantized conductivity given by
e2 σxy = cn, (2.1) ~
5 where e is the fundamental unit of electronic charge, ~ is Planck’s constant, and cn is an integer. This integer quantization has been experimentally verified to one part in 107[31] and is a direct manifestation of nontrivial topology of the bulk elec- tronic wavefunctions. The quantum Hall system possesses cn topologically protected
edge modes, each carrying a quantum of conductance e2 . This correspondence of ~ bulk topology with protected edge modes is characteristic of topological systems in condensed matter physics.
Since the discovery of the quantum Hall effect, many examples of topological states of matter have emerged in condensed matter physics. Topological insulators
[1, 2] extended this paradigm to three dimensional materials. Since then, many exotic states that possess a gapped, nontrivial bulk topology and corresponding gapless edge modes have been theoretically and experimentally discovered, such as topological crystalline insulators[32] and topological superconductors [33, 34].
In this chapter, we show that an electron will in general pick up a geometric phase known as the Berry phase under time evolution. The Berry phase is a global, topo- logical invariant related to a local geometrical quantity known as the Berry curvature.
We derive general facts about these related quantities. Finally, we introduce a simple model for a Chern insulator that will provide key insights in our later discussions of
Weyl semimetals.
2.1 Berry Phase
Consider a system with a time-dependent Hamiltonian H(k) that depends on time through several parameters k = (k1, k2, ...), where kj = kj(t). We are interested in
how the wavefunctions |ψ(t)i vary adiabatically as the system evolves in time. The
6 time evolution of the system is given by
d H(k(t)) |ψ(t)i = i |ψ(t)i . (2.2) ~dt
We introduce the instantaneous eigenstates |n(k)i by solving
H(k) |n(k)i = En(k) |n(k)i , (2.3) at each point k. If the system begins in an instantaneous eigenstate |n(k)i and evolves at a rate slow compared to the energy separation |En − En±1|, then by the adiabatic theorem[35], the system will remain in this state up to a phase |ψ(t)i = e−iθ(t) |n(k)i.
When we insert this wavefunction into Eqn. (2.2), we obtain that θ(t) is given by
Z t Z t 1 0 0 0 d 0 0 θ(t) = En(k(t ))dt − i hn(k(t ))| 0 |n(k(t ))i dt . (2.4) ~ 0 0 dt
The first term of the phase is the conventional dynamical phase. The negative of the second term is known as the Berry phase. Since the states only depend on time through the parameter k(t), we can write the Berry phase as
Z t Z 0 0 dk 0 γn = i hn(k(t ))| ∇k |n(k(t ))i · 0 dt = i hn(k)| ∇k |n(k)i · dk, (2.5) 0 dt C where C is the path that the electron traverses in time t. We can define a vector potential called the Berry connection:
An(k) = i hn(k)| ∇k |n(k)i , (2.6) which acts analogous to the vector potential from electromagnetic field theory. There- fore, we can write the Berry phase as a line integral of this vector potential such that
Z γn = dk · An(k). (2.7) C
7 Continuing the analogy with real-space electrodynamics, we see that γn plays the role
of a magnetic flux.
The Berry connection is gauge dependent. If the states are transformed by a U(1) rotation |n(k)i → eiξ(k) |n(k)i for some real-valued function ξ(k), then the Berry connection transforms to
∂ A (k) → A (k) − ξ(k). (2.8) n n ∂k
Hence the Berry phase is changed by an amount ξ(k(t)) − ξ(k(0)) over the contour
C. It was originally thought that any such phase factor could be eliminated with a
clever choice of gauge. However, Berry showed that[35] that for closed paths where
k(t) = k(0), since the wavefunctions must be single-valued, it must be true that
ξ(k(t)) − ξ(k(0)) = 2πm, with m an integer. Hence the Berry phase itself cannot be
canceled for closed paths unless it is an integral multiple of 2π.
2.1.1 Berry Curvature
We have shown above that the Berry connection is gauge dependent and it is often
convenient and instructive to consider the gauge independent Berry curvature defined
by
Ωn(k) = ∇k × An(k). (2.9)
This quantity is analogous to the magnetic field and is a similarly observable quantity,
unlike the Berry connection or real-space electromagnetic vector potential. By Stokes’
theorem, we can write the Berry phase as
Z γn = dS · Ωn(k), (2.10)
where dS is the differential area element of the region enclosed by C.
8 We briefly discuss how Ωn(k) transforms under some basic symmetry operations.
Henceforth, we let k be the wavenumber of a particle p = ~k. A system with no internal degrees of freedom transforms under inversion (or parity) Pˆ by sending each
momentum k → −k. Similarly, a spinless system under time-reversal symmetry Tˆ by the complex conjugation operator composed with k → −k. The gradient operator ˆ ˆ in momentum-space therefore transforms ∇k → −∇k under either P or T . Hence the Berry connection transform as An(k) → An(−k) under inversion and transforms as An(k) → −An(−k) under time-reversal. Hence, using Eqn. (2.9), we obtain the following useful relations
Pˆ Tˆ Ωn(k) −→ Ωn(−k), Ωn(k) −→−Ωn(−k). (2.11)
Hence in the presence of inversion symmetry, the Berry curvature is an even function
of k, while in the presence of time-reversal symmetry it is is an odd function of
k. Therefore in the presence of both, the Berry curvature must be identically zero
everywhere.
In principle, the Berry curvature is straightforward to calculate directly from the
quantum mechanical wavefunctions following Eqns. (2.6) and (2.9). However, in
many cases, it is difficult or impossible to obtain an analytic result for the eigenvec-
tors of a Hamiltonian with Hilbert space dimension ≥ 5. One can naively calculate
the Berry connection and curvature for a generic Hamiltonian whose states have been
numerically solved for, but this can present serious challenges if the numeric methods
used to obtain the spectrum do not consistently choose a gauge for different parame-
ters k. Specifically, if the matrix diagonalization routine chooses different numerical
phases for nearby eigenvectors, a simple-minded calculation of the Berry connection
as a finite difference will introduce substantial errors in the calculation. In this case,
9 it is best practice to use the more sophisticated method of Fukui, Hatsugai, and
Suzuki[36].
2.1.2 Chern Number
The Berry curvature represents a local, gauge-invariant geometric property of the spectrum for a particular Hamiltonian. Much like the physical Gaussian curvature of a manifold, the integral of the Berry curvature is a quantized topological invariant, known as the TKNN invariant[37] or the Chern number[38]. The Chern number for the nth electronic band is given by
1 Z c = dk · Ω (k), (2.12) n 2π n where k = (kx, ky) is the crystal momentum for a two-dimensional periodic system integrated over the entire Brillouin zone. The Chern number for a single band is well- defined only if there exists a local gap in energy |En(k)−En±1(k)| > 0 for all k. When electronic bands are degenerate at some subset of points in the Brillouin zone, then the Chern number given by Eqn. (2.12) is not well-defined for any of those bands.
It is clear that the Chern number can only be non-zero if time-reversal symmetry is broken.
The integer in Eqn. (2.1) for a QHE system on a lattice is the Chern number given above and, generically, the Hall conductivity of a filled band is given by Eqn.
(2.1), even for systems with zero external magnetic field. Just as in the quantum
Hall system, there exists a number of dissipationless edge modes equal to the sum of Chern numbers for all occupied bands. We shall see in Chapter 4 that even when bands are not completely occupied, the anomalous Hall conductivity can be related to the net Berry curvature of fully and partially occupied bands.
10 There is a rather sophisticated mathematical argument based on homotopy theory by Avron, Seiler, and Simon [39] that for a system with a finite number of bands, the sum of the Chern numbers of all bands must be zero. Since one is often interested in the sum of the Chern number of all occupied bands, this sum rule tells us that the sum of the Chern numbers of all unoccupied bands must therefore be equal and opposite. This sum rule also tells us that a system with only a single band cannot have a non-trivial Chern number. Hence, we must look to systems with at least two bands to find the simplest example of non-trivial topology in the band theory of solids. In the next section, we will discuss a specific example of a two-band system with non-zero Chern number for its bands.
2.2 Chern Insulators
We can write the generic Hamiltonian for any two-band system in the following way 3 X H(k) = d0(k)ˆσ0 + dj(k)ˆσj, (2.13) j=1 where dα(k) is a periodic function of k,σ ˆ0 is the 2 × 2 identity matrix, andσ ˆj is the j-th Pauli matrix. Here, the space over which the Pauli matrices span need not be electronic spin. Instead, “up” and “down” may represent the states of any two-orbital system. H(k) is the kernel of the momentum space Hamiltonian given by
ˆ X ˆ † ˆ H = Ψk ·H(k) · Ψk, (2.14) k ˆ ˆ where Ψk is a spinor of electron creation and annihilation operators given by Ψk =
(†) (ˆck,↑, cˆk,↓) and wherec ˆk,s destroys (creates) an electron in state s.
11 2.2.1 Berry Curvature of a 2-Band Hamiltonian
The energy spectrum of the Hamiltonian in Eqn. (2.13) is straightforwardly found to be q 2 2 2 E(k) = d0(k) ± dx(k) + dy(k) + dz(k). (2.15)
q 2 2 2 We henceforth denote d ≡ dx(k) + dy(k) + dz(k). The eigenstates are found to be 1 dz ± d ψ± = p . (2.16) 2d(d ± dz) dx − idy
Hence, we can directly compute the j-th component of the Berry connection to be
(j) −1 A− (k) = i hψ−| ∂kj |ψ−i = dy∂kj dx − dx∂kj dy . (2.17) 2d(d + dz)
We can similarly calculate the Berry connection for the band with energy E+. We can then write the j-th component Berry curvature for each band compactly as
d(k) · (∂ d(k) × ∂ d(k)) Ω(j)(k) = ∓ kl km . (2.18) ± jlm 2d3(k)
We notice that the field strength of the Berry curvature is the Jacobian of the map
d 2 k → d . This Jacobian is a map between 2-D Brillouin zone - the two torus T - and the target manifold given by the 2-D sphere S2 given by d2 = 1. We see that for a two-band Hamiltonian, the Berry curvature field strength, integrated over momentum-space, simply counts the number of times d winds wraps the unit sphere.
We define a Chern insulator as a system where: (1) the Hamiltonian is of the form given by Eqn. (2.13) whose d-vector wraps the unit sphere a non-zero integral number of times and (2) where there exists a nonzero gap between the E+ and E− bands throughout the Brillouin zone. There are several examples of such systems in the literature. Historically, the first example was Haldane’s model[40] for the quantum
12 Hall effect without Landau levels on the honeycomb lattice. However, since no system that we will be interested in possesses the C3 symmetry of the honeycomb lattice, we
find it more insightful to introduce a Chern insulator on a square lattice.
2.2.2 Chern Insulator on the Square Lattice
Consider the following Hamiltonian
H(k) = t sin (kxa)ˆσx + t sin (kya)ˆσy + B (2 + M − cos (kxa) − cos (kya))σ ˆz, (2.19) where t and B are parameters with the dimension of energy, M is a dimensionless parameter, and a is the spacing of the real-space lattice. We note that this Hamil- tonian reduces to a linear Hamiltonian around kx = ky = 0. Physically, we may think of this as the Hamiltonian for some spinless two-orbital system with one s- type orbital and one p-type orbital (more generally, we require that the orbitals must have different parity). Hence, the off-diagonal terms Eqn. (2.19) represent an L = 1 angular-momentum coupling of the form sin kx + i sin ky and we have included and intraorbital dispersion of the form 2 − cos kx − cos ky, which is allowed by the C4 rotational symmetry of our model.
It is straightforward to see that the energy dispersion for the system described by
Eqn. (2.19) is fully gapped, other than at several values of M: for M = 0, the system is gapless at kx = ky = 0; when M = −2, it is gapless at (kxa, kya) = (π, 0) and
(kxa, kya) = (0, π); and for M = −4, the gap closes at kxa = kya = π. By adiabatic continuity, the system must remain in the same topological phase so long as its gap does not close. For M → ±∞, the system is in the atomic limit. Since a gap does not close as we tune M from +∞ to M = 0 or from −∞ to M = −4, the system is
13 (a) (b) (c) (d)
Figure 2.1: Band structure for the Hamiltonian given in Eqn. (2.19) for B = t and various M: (a) M = −4 where we see the transition between the trivial state and the first topologically nontrivial sector. (b) M = −2 where we see the transition between the two distinct topological regimes. (c) M = 0 showing a gap closing at the final topological tranisition to the trivial state. (d) M = 1 where we see the gapped topological trivial sector. As M increases, this gap grows and the bands flatten.
a trivial insulator for M > 0 or M < −4. We show the band structure for various regimes in Fig. 2.1.
We can explicitly compute the Berry curvature, which we find to be
2 (z) 2 Bt (cos (kxa) ((2 + M) cos (kya) − 1) − cos (kya)) Ω± (k) = ±a 3 . 2 2 2 2 2 2 2 t (sin (kxa) + sin (kya)) + B (2 + M − cos (kxa) − cos (kya)) (2.20)
Although a closed form anti-derivative for Eqn. (2.20) does not exist, it is straight-
forward to numerically compute the integral for the Chern number given by Eqn.
(2.12). Doing so, we obtain that for −4 < M < −2, c± = ∓1. For −2 < M < 0,
the Chern numbers for each band are given by c± = ±1. Hence, we see that there
are four separate, topologically distinct sectors: two topological and two trivial. We
show the Berry curvature for all four regimes in Fig. 2.2(a-d). For the topologically
trivial regions, the Berry curvature averages to zero, while for the nontrivial regions,
it integrates to the Chern numbers given above. We note that the Berry curvature in
the non-trivial regions is concentrated near the points where the gap closings occur
14 (see Fig. 2.1). This can be understood intuitively by a careful examination of Eqn.
(2.18): since the denominator d3 is essentially half the energy gap, as the gap shrinks, the Berry curvature is enhanced. The Berry curvature is singular at the gap closings.
2.2.3 Chiral Edge Modes
As discussed above, in topological systems, there exists a “bulk-boundary” corre- spondence between bulk topological invariants and gapless edge modes. In the case of the Chern insulator given by Eqn. (2.19), a positive-(negative)-definite Chern num- ber corresponds to the net number of right-(left)-moving chiral edge modes on the real-space edge of the system. In order to directly calculate the dispersions of these edge modes, we employ the standard procedure of considering a system periodic in one direction and with open, finite boundary conditions in the other. Without loss of generality, let us consider the Chern insulator in the strip geometry which is periodic in the x-directions and finite in the y-direction with N sites. To do so, we transform the operators in Eqn. (2.14) as follows:
1 ˆ X −ikyy ˆ Ψk ,y = √ e Ψk ,k , (2.21) x N x y ky
2πny where ky = N for ny = 0, 1, ..., N − 1. We obtain a 2N × 2N Hamiltonian at every momentum kx.
In Fig. 2.2(e-h), we show the edge states calculated in the above manner for the Chern insulator with N = 50 layers in the finite y-direction. For states which are localized exponentially on the top edge, we color them red while we color states on the bottom edge blue. States whose expectation values in the y-direction are predominantly found in the bulk are colored black. We see that in the topologically trivial sectors (M < 4 or M > 0) shown in Fig. 2.2(e) and 2.2(h), there are no
15 (a) (b) (c) (d)
(e) (f) (g) (h)
Figure 2.2: Berry curvature (in units of a2 of the lower band and edge states for the Hamiltonian given in Eqn. (2.19) for B = t and various M. Edge states are calculated using strip geometry, finite in the y-direction with N = 50 layers. In the strip calculations, states localized to the top edge are colored red, states localized to the bottom surface are colored blue, and states in the bulk are colored black. (a) Berry curvature averages to zero over the either band for M = −5 in one of the topologically trivial sectors. (b) Berry curvature in the lower band M = −2.5 where the Chern number is given by c− = +1. (c) Berry curvature in the lower band M = −1.5 where the Chern number is given by c− = −1. (d) Berry curvature is again zero in the trivial sector where M = 1. (e) Strip geometry calculation for M = −5 in the trivial sector showing a lack of edge states. (f) Edge states for M = −2.5 crossing zero energy at the kx = ±π point. (f) Edge states for M = −1.5 crossing zero energy at the kx = 0 point. We note that the chirality, or direction, of the edge state on a given surface changes between (e) and (f) due to the change in Chern number. (g) Strip geometry calculation for M = 0 in the trivial sector showing a lack of edge states.
16 edge states, reflecting the lack of bulk topology. However, we see in Fig. 2.2(f) and
2.2(g), we note the presence of topological edge states. These edge states on a given surface (say, the top shown in red) have disperse linearly and are said to be chiral.
The chirality, or sign of Fermi velocity, of these edge states on a given edge is directly related to the bulk topology. Hence, on the top (bottom) surface the edge states transitions from a left (right) mover for −4 < M < −2 to a right (left) mover for
−2 < M < 0. Hence we see that the chiral edge states are a direct manifestation of the bulk topology. Each of these edge states carries a conductance of e2 [25] and ~ consequently, the Hall conductance of this system is a consequence of these edge modes.
Although there are other nontrivial topological systems in two dimensions, the
Chern insulator is perhaps the simplest to understand. It is also important in our story. In the next chapter, we shall see that although the Chern insulator is a 2D system, it plays a crucial role in 3D topological Weyl semimetals.
17 Chapter 3: Introduction to Weyl Semimetals
In the previous chapter, we showed how a 2-D gapped Hamiltonian that breaks time-reversal symmetry could be classified by its bulk topology using the Chern num- ber. This bulk topological invariant was a global property of the local geometrical quantity known as the Berry curvature. Using a specific model for a Chern insulator, we showed that the Chern number corresponded directly to the number chirality of gapless edge modes on the real-space boundary of the system. A fundamental prop- erty of the Chern insulator - as well as 3-D topological insulators - is the necessity for a bulk gap to protect the topology of the system.
Weyl semimetals introduced a new paradigm for topological band theory, where the excitations in the bulk are gapless and yet the system still possesses a topological invariant as well as gapless, topologically protected edge modes. The bulk modes of
Weyl semimetals are linearly dispersing Weyl fermions that share the dispersion of
c light, Fermi velocity vF ∼ 100 . The three dimensional nature of this system is crucial. We have seen that for the 2-D Chern insulator given by Eqn. (2.19), the energy bands became degenerate for specific, fine-tuned parameter regimes.
In this chapter, we will show that Weyl semimetals are broadly characterized by three criteria: (1) linearly dispersing Weyl nodes (or Weyl points), (2) monopoles of
Berry curvature originating at the location of the Weyl points in momentum space,
18 (3) unique boundary modes known as Fermi arcs that connect projections of these
Weyl nodes in the surface Brillouin zone. Using the simplest two-band lattice model
that one can write down by symmetry, we will show how these features are closely
related. A central effort of this thesis has been to explore the consequences of the tilt
of a type-II Weyl semimetal. This class of Weyl semimetals breaks Lorentz invariance
and we show using several lattice models how the connectivity of the Fermi arcs can
be strongly modified by the tilt of type-II Weyl semimetals. We also show that in
inversion-breaking type-II Weyl semimetals, where there are necessarily multiple pairs
of Weyl nodes, novel surface states that we have termed “track states” may emerge.
We conclude this chapter with a case study of how our relatively simple models were
applied to provide insight into the experimental discovery of the first type-II Weyl
semimetal MoTe2.
3.1 Weyl Fermions
We consider the Dirac equation for an electron with mass m in d + 1 dimensions given by
µ (iγ ∂µ − m) ψ = 0, (3.1)
where we have set the speed of light and Planck’s constant to unity ~ = c = 1, and where ψ is the spinor of the electronc and µ runs from 0, ..., d. The d + 1 dimensional gamma matrices satisfy the anticommutation relations {γµ, γν} = 0 for µ 6= ν and
2 2 (γ0) = − (γj) = I, for j = 1, ..., d. The minimum size of the matrices that satisfy these relations depends on the spatial dimension.
Hermann Weyl [6] noticed that in odd spatial dimensions, the Dirac equation can sometimes be written in a simplified form. In the simplest case of d = 1, we only
19 0 need two 2 × 2 gamma matrices. We choose the following representation: γ =σ ˆz
1 and γ = iσˆy. With some rearrangement, we can write the 1 + 1-D Dirac equation as
0 1 0 i∂tψ = (γ γ px + mγ )ψ, (3.2)
where the momentum in the x-direction is given by px ≡ −i∂x. We can introduce the
0 1 matrix γ5 = γ γ =σ ˆx, where the last equality holds for the particular representation that we have chosen above. When the electron described by Eqn. (3.2) has no mass m = 0, if ψ is an eigenstate of γ5 such that γ5ψ± = ±ψ±, then
i∂tψ± = ±pxψ±, (3.3) which describes a right-moving and left-moving chiral electrons with linear dispersion
E±(px) = ±px. These chiral electrons are precisely the edge states we found at the boundary of the Chern insulator and a single mode of a given chirality may only exist at the boundary of such a topological state. In a truly 1-D system, massless
Dirac electrons must come in pairs of opposite chirality. We see that the mass can be understood as the term causing the chiralities to mix.
In higher odd spatial dimensions d = 2k + 1, where k is a positive integer, one
k 0 1 d can form the Hermitian operator γ5 = i γ γ ...γ . We note that in even directions,
γ5 is given rather trivially by the identity matrix. However, in odd dimensions, γ5 anticommutes with γ0. The Dirac equation in this dimension can be written as
d 0 X 0 j i∂tψ = γ mψ + γ γ pjψ (3.4) j=1 µ From the anticommutation properties of γ , it is straightforward to show that γ5 commutes with the prefactors of the j-th momenta given by γ0γj.
Now we take d = 3. In this case, we can represent each γµ as a 4 × 4 matrix in
0 j 5 the following way: γ = I ⊗ σˆx, γ = iσˆj ⊗ σˆy, and γ = −I ⊗ σˆz. We note that the
20 Kronecker product is between two distinct 2-component subspaces. When m = 0, we
can write Eqn. (3.4) as
i∂tψ± = ∓p · σψˆ ±, (3.5)
where we see that the fermion propagates with its spin parallel or antiparallel to its
momentum. This defines the chirality of the electron. Hence, we obtain the Weyl
Hamiltonian for the massless electron with chirality χ = ±1, given by
Hχ(k) = χ~vF k · σ,ˆ (3.6)
where we have restored Planck’s constant ~, the speed of the fermion vF , and written the fermion’s momentum as p = ~k.
3.1.1 Weyl Fermions in Quantum Materials
We have seen that Weyl fermions emerge from a massless solutions of the rela- tivistic Dirac equation in three spatial dimensions. Hermann Weyl first proposed that neutrinos were described by Eqn. (3.6), however we now understand that this cannot be the case, since it has been shown that neutrinos possess mass[7]. For the remain- der of this thesis, we will discuss Weyl fermions in the context of condensed matter physics. While the Dirac equation is the correct description of electrons in vacuum, we know that the Schrodinger equation is adequate to treat electrons with kinetic energy much less than their rest mass, such as in the band structures of metals and semimetals. In this sense, it is rather remarkable that Weyl fermions with a linear relativistic spectrum emerge as the low energy excitations of quantum materials.
It would be too substantial a digression here to recapitulate all of the electronic band theory of solids. Instead, we wish to remind the reader of several facts that will be key to our discussion:
21 1. Bloch’s theorem states that an electron in a periodic potential U(r+R) = U(r)
ik·r is described by a wavefunction ψn,k(r) = e un,k(r), where un,k(r) = un,k(r + R) is a function with the same periodicity as the potential and n is known as the band index.
2. The wavevector k in the Bloch eigenstates plays the same role as the wavevector k = p/~ in Sommerfeld’s theory of the free electron. However, in the case of a periodic potential, the above simple relation between p and k does not hold.
3. For a real-space lattice vector R, a reciprocal lattice vector K satisfies eiK·R = 1.
Any k0 can be written as k0 = k + K where k is confined to the first Brillouin zone.
4. Since the periodic Schrodinger equation for the Bloch function un,k(r) takes place in a finite volume subject to the boundary conditions given in the first point above, we expect on fairly general grounds that the solutions will take the form of an infinite spectrum of discretely spaced energy levels. It is these levels that we have labeled by the band index n.
5. The eigenstates and eigenvalues are periodic function of k such that ψn,k(r) =
ψn,k+K(r) and En(k) = En(k + K).
For the details, we direct the reader to one of any of the standard texts [41, 42].
In the electronic band structure, if two nondegenerate bands meet at isolated zero- dimensional crossings in the Brillouin zone, then close to the so-called Weyl point at the crossing k = 0, they may be described by the Hamiltonian in Eqn. (3.6). In general, it can be useful to consider the case where the Weyl fermions disperse with different velocities in each direction, however for now we restrict ourselves to the isotropic limit. In 2D, where k = (kx, ky), there is no coefficient ofσ ˆz, which is only possible if both time-reversal and inversion symmetry are present. Graphene[43], a two-dimensional hexagonal array of covalently bonded Carbon atoms, possesses
22 the required symmetries and its Brillouin zone contains two doubly degenerate Dirac nodes. The Chern insulator in the previous section illustrates the 2D case where time- reversal symmetry is broken and there exists a massive, gapped Weyl node. Since by our definition, a Weyl node is comprised of non-degenerate bands, we must break either time-reversal or inversion symmetry to remove the degeneracy of Kramer’s pairs everywhere in the Brillouin zone. Unlike in two-dimensions, when k = (kx, ky, kz), the Weyl node described by Eqn. (3.6) cannot be gapped out by a broken symmetry.
Such a broken symmetry will lead to the introduction of “mass term” M(kx, ky, kz)ˆσz to the Hamiltonian. However, it is simple to show that the presence of any such term in three dimensions will only shift the location of the zero crossing.
For the remainder of this thesis, we will restrict ourselves to the three-dimensional case. Consider a material that breaks either inversion or time-reversal symmetry and that contains Weyl nodes described by Eqn. (3.6). We note that the associated energies are independent of chirality χ and are given by
q 2 2 2 E±(k) = ±~vF kx + ky + kz . (3.7)
This dispersion shown in Fig. 3.1a. The eigenstates around a Weyl node with chirality
χ are given by χ kz ± χk ψ± = p , (3.8) 2k(k ± χkz) kx − iky
p 2 2 2 where k = kx + ky + kz .
Density of states and temperature dependence of chmical potential
The density of states for the Hamiltonian in Eqn. (3.6) is found to be
1 E 2 g(E) = 2 3 , (3.9) π (~vF )
23 for energies E measured from the nodal energy. The density of states in Eqn. (3.9) vanishes at the Weyl nodes and this is responsible for a number of novel phenomena in Weyl semimetals. Later in this thesis, we will discuss how this density of states reveals itself in transport phenomena.
In the usual way, we can calculate the temperature dependence of the chemical potential by fixing the density and requiring
Z g(E) 1 Z E 2 n = dE β(E−µ(T )) = 2 3 dE β(E−µ(T )) , (3.10) 1 + e π (~vF ) 1 + e where β = 1 , n is the density and µ(T ) is the temperature dependent chemical po- kB T tential. We can self-consistently attain the chemical potential by numerically solving for the chemical potential of Eqn. (3.10). We plot the results in Fig. 3.1, though we obtain the same upon carefully performing the continuum calculation. We note that the chemical potential always shifts to the nodal energy µ = 0 on a temperature scale
TW ∼ µ(T = 0)/kB. This has the interesting consequence that no matter the Fermi energy EF ≡ µ(T = 0), the chemical potential will always go to zero for T & TW . This is illustrated in Fig. 3.1b.
Berry curvature of a Weyl fermion
Since the Weyl Hamiltonian given by Eqn. (3.6) is a 2-band Hamiltonian, we can analyze its associated Berry curvature using the formalism developed for the
Chern inslator in the previous chapter. From Eqn. (2.18), we can show that the momentum-space Berry curvature for the n = ± band of a node with chirality χ is given by χ k Ωχ (k) = ± . (3.11) ± 2 k3
24 (a) (b)
Figure 3.1: (a) Continuum dispersion of a Weyl node. (b) Dependence of the chemical potential on temperature µ(T ). We see that on a temperature scale TW ∼ µ(T = 0)/kB, the chemical potential reaches the Weyl node.
Furthering the analogy developed in the previous chapter of the Berry curvature as a magnetic field in momentum space, we see that Weyl nodes act as monopoles in the
χ Brillouin zone, with the charge of the monopole given by 2 . We now present a simple argument that the total chirality of Weyl nodes in a condensed matter system must be zero[44]. Hence, if Weyl nodes exist in the band structure of a solid, they must come in pairs of opposite chirality. Consider a system with N > 0 Weyl nodes each with chirality χj, then for momenta around the Weyl nodes, the Berry curvature is given by
N 1 X (k − kj) Ω(k) = χ , (3.12) 2 j k3 j=1 where the j-th Weyl node is at momentum kj. Then the integrated divergence of this
Berry curvature is given by Z N Z N 3 X 3 X d k ∇k · Ω(k) = 2π χj d k δ(k − kj) = 2π χj, (3.13) BZ j=1 BZ j=1
25 where the integrals are over the Brillouin zone. Then, from the Divergence Theorem, it is true that Z Z 3 2 d k ∇k · Ω(k) = d k · Ω(k), (3.14) BZ ∂BZ where ∂BZ denotes the boundary of the Brillouin zone. However, the Brillouin zone is a 3-torus and tori have no boundaries, so the integral above is zero. Hence we have that N X χj = 0. (3.15) j=1 We see that for N > 0, the net chirality of the Weyl nodes must be zero and hence if
Weyl nodes exist in a system, they must come in pairs of opposite chirality.
3.2 Lattice Models for Weyl Semimetals
The defining features of a TWS are the nodal energy crossings in the Brillouin zone and so a minimal lattice model for a TWS must have at least two bands. A general two-band Hamiltonian takes the form
ˆ X † ˆ H = cˆkα H(k) cˆkβ (3.16) αβ k
(†) wherec ˆkα annihilates (creates) an electron at momentum k in orbital α and
ˆ X H(k) = di(k)σ ˆi (3.17) i=0,1,2,3 as was true for the Chern insulator of the previous chapter.
As we saw last chapter, in the presence of both inversion and time reversal sym- metry the Berry curvature is identically zero throughout the Brillouin zone. Hence, the presence of Weyl nodes relies on breaking either inversion (henceforth labeled Pˆ) or time reversal symmetry (labeled Tˆ). For spinless fermions, we choose a definite
26 representation for the Pˆ and Tˆ operators,
ˆ ˆ ˆ P ↔ σˆ1, T ↔ K, (3.18)
where Kˆ is the anti-Hermitian complex conjugation operator. Each of Pˆ and Tˆ also reverse the sign of the momentum such that k → −k. In this paper we investigate
lattice models for Weyl semimetals that break either Tˆ or Pˆ, and using the definitions
in Eqn. (3.18) it will be straightforward to show this symmetry breaking explicitly
for each model we consider.
Under inversion symmetry, k → −k and σ → σ. Hence a time-reversal breaking
Weyl semimetal (with inversion symmetry present) may have only two nodes. On the other hand, since under time-reversal symmetry k → −k and σ → −σ, a system that breaks time-reversal symmetry but not inversion symmetry must have at least four nodes, since under time-reversal symmetry a node does not change chirality. Hence there must be an additional pair. The simplest model for a Weyl semimetal therefore breaks time-reversal symmetry, preserves inversion symmetry and possesses a single pair of Weyl nodes.
The first two-band model that appears in the literature is due to Yang, Lu, and
Ran[45] and is given by
TRB H (k) = − (m[2 − cos (kya) − cos (kza)] + 2tx[cos (kxa) − cos (k0a)])σ ˆx
− 2t sin (kya)ˆσy − 2t sin (kza)ˆσz, (3.19)
where m, tx, and t are parameters with the dimension of energy and a is the lattice constant. This model explicitly breaks time-reversal symmetry and contains a single pair of Weyl nodes at k = (±k0, 0, 0). Around these points, it is straightforward
to see that the linearized Hamiltonian reduces to that of Eqn. (3.6). Each of these
27 (a) (b) (c)
Figure 3.2: Band structure for Weyl semimetal in Eqn. (3.19) for tx = 0.5t, m = 2t, k0 = π/2, and lattice constant set to unity a = 1. (a) Bulk energy dispersion with ky = 0. (b) Cut through the Weyl nodes along kx with ky = kz = 0. (c) Constant energy EF = 0.2t cut for a system with slab geometry with N = 50 layers in the y-direction. Surface states are colored red (top) and blue (bottom). We see that small bulk Fermi pockets (shown in black) enclosing the Weyl points (green points) are connected by the Fermi arcs on the top and bottom states.
Weyl nodes acts as a monopole of Berry curvature as described above. We show an
example of the bulk band structure in Fig. 3.2(a) and 3.2(b) for tx = 0.5t, m = 2t, k0 = π/2, and lattice constant set to unity a = 1.
We note that this model, for fixed kx, is very similar to the model for the Chern insulator given by Eqn. (2.13). The mass term M of this Chern insulator is given t by M(k ) = x [cos (k a) − cos (k a)]. We take what will be typical values for the x m x 0
remainder of this thesis of m = 2t and tz = 0.5t. We see that in the first Brillouin
zone, the 2D subspace of the Brillouin zone for constant |kx| < k0 describes a trivial gapped insulator with Chern number zero. However, the 2D subspace of the Brillouin zone for constant k0 < |kx| < π describes a Chern insulator and therefore possesses chiral edge states in a finite geometry.
28 We can therefore understand the simple model for a Weyl semimetal in Eqn.
(3.19) as a ”stack” (in momentum space) of Chern insulators and a ”stack” of trivial insulators. We know that the boundary of systems with different Chern numbers must be gapless and it is precisely at the momenta of the Weyl nodes where the energy gap goes to zero. For a system which is finite in the y-direction, much like the
Chern insulator in Chapter 2, but periodic in x and z, as kz evolves (k0 < |kx| < π) from one node to the other, the chiral edge states of the Chern insulator at each kx trace out an arc in the surface Brillouin zone. This arc is known as a Fermi arc and is perhaps the most unique property of Weyl semimetals.
In Fig. 3.2(c), we show a constant energy EF = 0.2t cut for a system with slab geometry with N = 50 layers in the y-direction. The states on the surface are colored red (top) and blue (bottom). We see that small bulk Fermi pockets (shown in black) enclosing the Weyl points (green points) are connected by the Fermi arcs on the top and bottom states. The Fermi arcs disperse linearly in kz. Strikingly, we see that these states form open contours in the surface Brillouin zone. These open contours are completely unlike traditional electronic states in solids and reflect the bulk topology of the Weyl nodes.
Three features clearly identify Weyl semimetals:
(1) Pairs of linearly dispersing Weyl nodes in momentum space.
(2) Monopoles of Berry curvature centered on the Weyl nodes
(3) Fermi arcs in the surface Brillouin zone connecting projections of Weyl nodes of opposite chirality.
29 In the following chapters, we shall see how each of these features reveal themselves in thermoelectric and thermomagnetic transport. For the remainder of this chapter, we will highlight important aspects of the band structure of Weyl semimetals.
Both the bulk Weyl nodes and the surface Fermi arcs have unique signatures in angle resolved photoemission spectroscopy (ARPES) experiments. Searching for these signatures has proven to be extremely fruitful and several groups[15, 16, 17, 46,
47, 48, 49] have discovered a Weyl semimetal phase in the transition metal pnictide family: TaAs, TaP, NbP and NbAs. These materials were the first experimentally discovered Weyl semimetals and, unlike the simple model above, they break inversion symmetry and contain twelve pairs of Weyl nodes. However, both the model above and the transition metal pnictide family are known as type-I Weyl semimetals, where the density of states of the Weyl fermions goes to zero at the Weyl points. In some sense, we can think of these as the limiting case of a direct gap semiconductor where the conduction and valence bands meet at the Weyl nodes. In the next section, we will explore the so-called type-II Weyl semimetals, where the bands comprising the
Weyl nodes have a finite density of states at the Weyl energy.
3.3 Type II Weyl Semimetals
In a type-I Weyl semimetal, the valence and conduction bands that meet at the
Weyl node have a finite density of states at the Weyl energy. We may picture this system as the limit of an indirect gap semiconductor, where offset bands meet at isolated pairs of points in the Brillouin zone. We will present a formal definition shortly. Type-II Weyl semimetals were recently theoretially predicted in a variety of compounds[50, 51, 52, 53]. Recently, signatures of a type-II Weyl semimetals
30 have been reported[54, 55, 56] in MoxW1−xTe2, stoichiometric MoTe2, and LaAlGe,
opening the door for further experimental study of the type-II Weyl semimetals. Later
in this chapter, we will outline our role in the discovery of MoTe2, the first of this
family of materials.
We extend the Hamiltonian given by Eqn. (3.6) to the case of an anisotropic Weyl
node with a momentum-dependent tilt (shift) of both energy bands:
ˆ X X HWP(k) = γikiσˆ0 + kiAijσˆi, (3.20) i=x,y,z i,j=x,y,z whereσ ˆ0 is the 2 × 2 identity matrix. Eqn. (3.20) describes a Weyl fermion with nodes of chirality χ = det(Aij). The energy spectrum for the Hamiltonian in Eqn.
(3.20) is given by v u !2 X u X X E±(k) = γiki ± t kiAij i=1,2,3 j=1,2,3 i=1,2,3 = T (k) ± U(k), (3.21) where T (k) tilts the Weyl cone. The definition[50, 57] of a type-II Weyl node is one where there exists a direction ek in the Brillouin zone such that
T (ek) > U(ek). (3.22)
The simplest Hamiltonian for a type-II Weyl Fermion is given by
Hχ(k) = γkxσˆ0 + χ~vF k · σ,ˆ (3.23)
where we have chosen the tilt direction to be the x-direction. Eqn. (3.23) describes
a type-II Weyl semimetal so long as γ > ~vF . Although there have been some studies of lattice models for TWS,[58, 59, 60]
much of the theoretical work on topological Weyl semimetals has focused on low
31 energy effective models of single Weyl nodes. In a type-I TWS, where the density of states vanishes at the energy of the Weyl nodes, these effective models capture much of the essential physics including electro- and magnetotransport, [61, 62, 63, 64,
65, 66, 67, 68] thermoelectric properties, [69, 70, 71, 20] magnetic properties,[72] and effects of disorder[73, 74, 75]. In a type-I TWS, when the chemical potential is shifted slightly away from the nodal energy, the Fermi pockets enclosing the projections of the Weyl nodes are very small. However, in a type-II TWS extended pockets of holes and electrons exist already at the node energy. Doping away from the node energy then results in the surface projections of the Weyl nodes, for typical crystal surfaces, becoming enclosed within large Fermi pockets. Understanding the interplay of these large Fermi pockets and any topological properties associated with the type-II nodes can require explicit lattice models, rather than just a low-energy theory. Here we present a study of a few such relatively simple lattice models for type-II TWS.
We begin by discussing models for time-reversal-breaking type-II TWS. We distin- guish between two types of basic models: the simplest model (“hydrogen-like model”) has a single pair of Weyl nodes which share a single electron pocket and a single hole pocket. However, we argue that this simplest model fails to capture some impor- tant properties. These are instead captured by the next-simplest model (“helium-like model”), with an additional term that splits both the electron pocket as well as the hole pocket into pairs of separate pockets. Each Weyl node is now formed from the intersection of an isolated pair of electron and hole pockets. The hydrogen-like model has no topologically-protected Fermi arcs, though it exhibits relics of them away from the Fermi energy; in the helium-like model, the topological Fermi arcs are restored.
We also study inversion-breaking type-II TWS models, and find that even simple toy
32 models support an additional set of surface states (”track states”) which are not topo- logical but nonetheless play a role in how the Fermi arc connectivity changes when either the Fermi energy is changed or when the tilt of the Weyl nodes is changed.
3.3.1 Time Reversal Breaking Model
We begin by investigating a lattice model given by a Hamiltonian Hˆ(k) that hosts
Weyl nodes and breaks time reversal symmetry but preserves inversion symmetry such that
Pˆ†Hˆ(−k)Pˆ = Hˆ(k), Tˆ †Hˆ(−k)Tˆ 6= Hˆ(k). (3.24)
The minimal number of Weyl nodes for such a Hamiltonian is two and we find that such a minimal model can be used to investigate a wide range of possible TWS Fermi surface and arc connectivity. We begin by writing down the simplest possible two node time-reversal breaking (TRB) Hamiltonian with a type-II tilt and investigating its band structure. A pair of Weyl nodes are formed from the nodal crossing of exactly one electron band with one hole band. By calculating the band structure for the system in a finite slab geometry, we investigate the surface Fermi arc behavior.
We then show that this minimal model can be modified with a term that splits these electron and hole pockets into pairs that exist around each node.
The “Hydrogen atom” for a type II time reversal breaking TWS
The following Hamiltonian
ˆTRB HA (k) = γ cos(kx) − cos(k0) σˆ0 − m(2 − cos(ky) − cos(kz)) + 2tx(cos(kx) − cos(k0)) σˆ1
− 2t sin(ky)ˆσ2 − 2t sin(kz)ˆσ3 (3.25)
33 satisfies the symmetry conditions in Eqn. (3.24) and possesses two Weyl nodes at k = (±k0, 0, 0). When γ = 0, this Hamiltonian is known[26] to host nodes of type-I. However, the addition of the term γ cos(kx)−cos(k0) σˆ0 bends both bands and when
γ > 2tx it is simple to see these nodes become type-II as defined by Eqn. (3.22). We see this evolution from type-I to type-II very clearly in Fig. 3.3. When γ = 0, the hole band (blue) touches the electron band (red) at the two Weyl points where the density of states vanishes, as seen in Fig. 3.3a,d,g. When the system is in the type-II regime, the Weyl cones are tilted and this leads to a nonzero density of electron and hole states at the node energy, as seen clearly in Fig. 3.3c,f,i. When γ = 2tx exactly, the system is at a critical point between a type-I and a type-II Weyl semimetal. This is clearly seen in Fig. 3.3b,e,h, where a single line of bulk states connect the Weyl points at E = 0. The states seen in Fig. 3.3h open up into the electron and hole pockets seen at E = 0 for the type II case in Fig. 3.3i.
In a type-II TWS, it is important to consider the net chirality enclosed by the bulk Fermi pockets when determining the Fermi arc connectivity. If one encloses a bulk pocket by a Gaussian surface in a region where the bulk band structure is gapped, the number of Fermi arcs impinging on the Gaussian surface are quantized and equal to the net chirality of Weyl nodes enclosed. When the model in Eqn. (3.25) is in the type-II regime and the chemical potential is shifted away from E = 0, the projections of both Weyl nodes are either enclosed in the electron pocket (E > 0) or the projections are both enclosed in the hole pocket (E < 0). Since the projections of both nodes lie within the same Fermi pocket, we expect that Fermi arcs in this system are not topologically protected in general. Surface states may exist, but their lack of topological protection stems from the fact that there are no isolated Fermi
34 Type I Critical Point Type II
a b c
d e f
g h i
Figure 3.3: Bulk band structure for the “Hydrogen atom” of type-I and type-II Weyl semimetal. a-c The bulk band structure for the Hamiltonian in Eqn. (3.25). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown in green. Here we have chosen parameters ky = 0 with parameters k0 = π/2, tx = t, m = 2t for (a) type-I Weyl semimetal with γ = 0, (b) the critical point between type-I and type-II Weyl semimetal with γ = 2t and (c) type-II Weyl semimetal with γ = 3t. The cones comprising the Weyl nodes develop a characteristic tilt of the type-II TWS as γ is increased. d-f Cuts through the Weyl nodes at ky = kz = 0 for the same parameters as (a-c). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). We see that for a type-I TWS, there are no states at the Fermi energy. At the critical point between a type-I and type-II TWS, we see lines of bulk states appearing between the nodes. These lines open into bulk hole and electron pockets (in the repeated zone scheme) when the system becomes a type-II TWS.
35 pockets that enclose Weyl nodes with nonzero net chirality. As a result, the surface
states can hybridize with bulk states and are therefore trivial.
In order to investigate the structure of the Fermi arcs, we introduce an edge
by considering a slab with a finite thickness in one direction. We partially Fourier
transform the Hamiltonian in Eqn. (3.25) into real space for a L layer system in the
y-direction, while keeping the system infinite in the x- and z-directions. In Fig. 3.4,
we show the results of such a slab calculation for the model given by Eqn. (3.25) in
the type I regime (γ = 0) with the same bulk parameters as in Fig. 3.3a,d,g and in
the type II regime (γ = 3tx) with the same bulk parameters as in Fig. 3.3c,f,i for
L = 50 layers. We calculate the expectation of the finite position hyi and label the
states as “top” (“bottom”) if they are exponentially localized at hyi = 1 (hyi = L).
We color these top and bottom states red and blue respectively.
As we expect, for the type-I case when γ = 0, a Fermi arc on each surface connects
the Weyl nodes, as seen in Fig. 3.4a-c. This is seen clearly in Fig. 3.4b where two
Fermi arcs connect the two nodes from (kx, kz) = (−π/2, 0) to (kx, kz) = (π/2, 0). At
E = 0, both the top and bottom arcs are degenerate at kz = 0, shown as a purple
line. When we lower the Fermi energy below the node energy, each node is enclosed
in a small isolated Fermi pocket. Since each pocket encloses a net chirality χ = ±1,
the pockets are connected by an arc on each surface, as seen in Fig. 3.4a. The same
is seen at higher energies E > 0 in Fig. 3.4c.
We calculate the band structure in the slab geometry for a type-II TWS (γ = 3tx)
and find that there are marked differences in the surface state behavior (see Fig.
3.4d-f). Since both nodes are formed from a single electron and a single hole pocket,
we cannot construct a simply connected 2D Gaussian surface in the Brillouin zone
36 a b c
Type I
d e f
Type II
Figure 3.4: Fermi surface and arc configuration for the “Hydrogen atom” of type-I and type-II TWS. a-c Bulk Fermi surfaces and surface Fermi arcs for a type I TWS with the same bulk parameters as in Fig. 3.3a,d,g calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the following constant energy: (a) E = −0.2t, (b) E = 0, (c) E = 0.2t. We color the states which are exponentially localized to the y = 1 (y = L) surface red (blue) and note that such surface states form topological arcs connecting the two Weyl nodes (shown as green dots and marked with pink arrows). We note that at E = 0 the two Fermi arcs are degenerate along kz = 0 and we color them purple to signify this. d-f Bulk Fermi surfaces and surface Fermi arcs for a type-II TWS with the same bulk parameters as in Fig. 3.3c,f,i calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the same constant energies as above: (d) E = −0.2t, (e) E = 0, (f) E = 0.2t.
37 that encloses a single node. When the energy is lower than the Weyl energy in Fig.
3.4d, we see that the projections of both nodes are enclosed by the same hole pocket.
Although there are two sets of surface states connecting the hole and electron pockets,
they are trivial in a topological sense. When one considers a Gaussian surface that
encloses the central hole pocket, it is pierced by four arcs, two on each real-space
surface. The Fermi velocity of each arc is opposite on a given real-space surface and
so the net chirality of the arcs is zero. We see that as we raise the chemical potential
to the node energy and above, these arcs disappear completely. This is completely
different from the type-I case where the arcs exist at all energies since the nodes were
always isolated in separate Fermi pockets.
The “Helium atom” for a type II time reversal breaking TWS
In order to study the physics of type-II Weyl nodes surronded by isolated Fermi
pockets which they do not share, we seek to introduce a term to the Hamiltonian
separates the single pair of pockets possessed by the ”Hydrogen-atom” model. In
particular, this new term must gap out the bulk band structure in the kx = 0 plane
and the kx = π plane. Due to the pairs of electron and hole pockets supported by this
model, we call it the “Helium model” for a type-II time-reversal-breaking TWS in
analogy with the “Hydrogen model” above. We consider the following Hamiltonian
ˆTRB ˆTRB HB (k) = HA (k) − γx(cos(3kx) − cos(3k0))ˆσ1, (3.26)
where we have added to Eqn. (3.25) the term proportional to γx. In general, this model supports up to six Weyl nodes. However, so long as |2tx| > |3γx|, there are only two Weyl nodes in the Brillouin zone. These nodes are located at E = 0 and k = (±k0, 0, 0) and they are type-II if γ > 3γx − 2tx. The addition of the term
38 γx(cos(3kx) − cos(3k0)) gaps out the bulk spectrum along the lines (ky, kz) = (0, 0)
and (ky, kz) = (0, π) at the nodal energy. This leads to a pair of isolated hole pockets touching a pair of isolated electron pockets at the Weyl nodes when the system is type-II. In Fig. 3.5, we find that as γ grows relative to 3γx − 2tx, the Fermi pockets grow in size. This is because as the tilt of the nodes gets larger, more electron and hole states exist at the Fermi energy. As we shift the chemical potential away from the node energy, the projections of the nodes are now isolated with each node in a single electron (hole) pocket when the chemical potential is raised (lowered).
We again consider the slab geometry described in the section above in order to investigate the interplay of the bulk pockets and the Fermi arcs for the model given by Eqn. (3.26). Unlike the simpler model described by Eqn. (3.25), we see in Fig.
3.6 that Eqn. (3.26) supports isolated Fermi pockets enclosing the Weyl nodes in the type-II regime when γ = t (Fig. 3.6a-c) and γ = 1.5t (Fig. 3.6d-f). Unlike the
Fermi surfaces and arcs generated by Eqn. (3.25), in Fig. 3.6 we see that each node
is isolated in its own hole (Fig. 3.6a,d) or electron (Fig. 3.6c,f) pocket when the
chemical potential is away from E = 0. We emphasize that this is due to the extra
σˆ1 term in the Hamiltonian in Eqn. (3.26). These isolated pockets in Fig. 3.6 are
connected by arcs confined to the surface in the y-direction. However, in this type-II
TWS the Fermi pockets enclosing a Weyl node can be quite extended and, unlike a
type-I TWS, the arcs can terminate on a pocket quite far away from the projection
of the nodes. We see that as the tilt grows in Fig. 3.6d-f, so do the pockets enclosing
the nodes. We note that a trivial electron pocket appears around the (kx, kz) = (π, π)
point. This pocket encloses no Weyl nodes and therefore it is not connected via Fermi
arcs to any other pockets.
39 Type I Type II Type II
a b c
d e f
g h i
Figure 3.5: Bulk band structure for type-I and type-II TRB model with separate pockets (the “Helium atom”). a-c The bulk band structure for the Hamiltonian in Eqn. (3.26). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown in green. Here we have chosen parameters ky = 0 with the parameters k0 = π/2, tx = t, m = 2t and γx = t/2 for (a) type-I TWS with γ = 0, (b) type-II TWS with γ = t and (c) type-II TWS with γ = 1.5t. The cones comprising the Weyl nodes again develop a characteristic tilt of the type-II TWS as γ is increased. d-f Cuts through the Weyl nodes at ky = kz = 0 for the same parameters as (a-c). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). Note that for a type-I TWS, there are no states at the Fermi energy while in the type-II regime, there are two sets of electron and hole pockets on either side of the Weyl nodes. We see that unlike the Hydrogen-atom model, the Helium-atom model has disjoint pairs of electron and hole pockets and a pair of each meet at the two Weyl nodes.
40 a b c
Type II
d e f
Type II
Figure 3.6: Fermi surface and Fermi arc configuration for type I and type-II time-reversal-breaking model with separate pockets (the “Helium atom”). a-c Bulk Fermi surfaces and surface Fermi arcs for a type-II Weyl semimetal given by Eqn. (3.26) with the same bulk parameters as in Fig. 3.5b,e,h calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the constant energies: (a) E = −0.2t, (b) E = 0, (c) E = 0.2t. As in Fig. 3.4, we color the states that are exponentially localized to the y = 1 (y = L) surface red (blue) and note that such surface states form topological arcs connecting the two Weyl nodes (shown as green dots). We note unlike in Fig. 3.4, each node is isolated in its own hole (a) or electron (c) pocket when the chemical potential is away from E = 0. These pockets are connected by arcs confined to the surface in the y-direction. However, in this type-II TWS the Fermi pockets enclosing a Weyl node can be quite extended, unlike a type-I TWS, the arcs can terminate on a pocket quite far away from the projection of the nodes. d-f Bulk Fermi surfaces and surface Fermi arcs for a type-II TWS with the same bulk parameters as in Fig. 3.5c,f,i calculated in a slab geometry with L = 50 layers in the y-direction. The slab calculations are done at the same constant energies as above: (d) E = −0.2t, (e) E = 0, (f) E = 0.2t. We see that as the tilt grows, so do the pockets enclosing the nodes. We note that a trivial electron pocket appears around the (kx, kz) = (π, π) point. This pocket encloses no Weyl nodes and so is not connected via Fermi arcs to any other pockets.
41 Although the local linearized Hamiltonian describing the spectrum close to a node in Eqn. (3.26) is identical to the effective Hamiltonian of nodes of the model described by Eqn. (3.25), the full lattice models describe topologically distinct configurations of bulk Fermi surfaces and surface Fermi arcs. When there is only one electron pocket and one hole pocket with the projections of the Weyl nodes enclosed by the same pocket, the topological protection of the Fermi arcs is lost. However, we see that once each node is enclosed in its own isolated pocket, the topological protection of the Fermi arcs is restored.
Finally, we consider the energy dispersion of the Fermi arcs near a node. Again using the slab geometry as above, we calculate the energy spectrum, this time at a constant kz. We see that for the simplest type-I case (Eqn. (3.25) with γ = 0), the surface arcs do not disperse in kx for a fixed kz. This changes in the type-II case for both the simple Hamiltonian in Eqns. (3.25) and (3.26). At fixed kz, the arcs connecting the node inherit the tilt proportional to γ and now bend. This characteristic bend of the Fermi arc dispersion has been observed in ARPES studies of type-II Weyl semimetal[55].
3.3.2 Inversion Breaking Model
We now turn to a lattice model for a topological Weyl semimetal that breaks inversion symmetry but is invariant under time-reversal. Analogous with Eqn. (3.24), we seek a Hamiltonian Hˆ(k) that satisfies the following symmetry conditions
Pˆ†Hˆ(−k)Pˆ = 6 Hˆ(k), Tˆ †Hˆ(−k)Tˆ = Hˆ(k), (3.27) where Pˆ and Tˆ are again given by Eqn. (3.18). Unlike a time-reversal-breaking Weyl semimetal, the minimum number of Weyl nodes for a spinless inversion-breaking (IB)
42 TWS is four. More importantly, the lattice model for an IB TWS exhibits what we term ”track states” that are loops of states that live on the surface of the TWS and are degenerate with the states forming the topological Fermi arcs. However, unlike topological Fermi arcs, these track states form closed contours rather than open ones; they are not topological, but do evolve from the topological arc states upon the transition from type-I to type-II.
It is easy to show that the Hamiltonian
ˆIB H (k) = γ(cos(2kx) − cos(k0))(cos(kz) − cos(k0))ˆσ0
2 − (m(1 − cos (kz) − cos(ky)) + 2tx(cos(kx) − cos(k0)))ˆσ1
− 2t sin(ky)ˆσ2 − 2t cos(kz)ˆσ3 (3.28) satisfies the conditions in Eqn. (3.27). When γ = 0, Eqn. (3.28) describes a TWS with four nodes located at kW = (±k0, 0, ±π/2) that breaks inversion but preserves time-reversal symmetry. The term γ(cos(2kx)−cos(k0))(cos(kz)−cos(k0))ˆσ0 causes a different shift in both band than those considered in the time reversal breaking cases and results in both bands bending in both the kx- and kz-directions. This can produce isolated Fermi pockets around the Weyl points without having to add an additional
σˆ1 term like in the time-reversal-breaking case in Eqn. (3.26). The inversion-breaking model above also easily generates trivial Fermi pockets that exist in isolation from those that meet at the Weyl nodes.
We show the bulk band structure for Eqn. (3.28) in Fig. 3.7. We see that indeed when γ = 0 (Fig. 3.7a,d,g), the electron band meets the hole band at four isolated type-I Weyl points and the density of states vanishes at the nodal energy. As γ increases, the Weyl nodes begin to tilt in the kz-direction. When γ is tuned to the
43 critical point between the type-I and type-II phases (Fig. 3.7b,e,h), the electron and
hole pockets still meet at the four Weyl nodes with a vanishing density of states, but
we see in Fig. 3.7e that the Weyl nodes are now tilting in the kz-direction. As γ is
further increased into the type-II limit (Fig. 3.7c,f,i), we now see that the nodes are
tilted as seen in Fig. 3.7f and the electron (hole) pockets are shifted below (above)
the node energy. In particular, we see in Fig. 3.7i that there are four electron and
four hole pockets that exist at E = 0 and meet at the Weyl nodes. There is also a trivial hole pocket centered at k = (0, 0, 0) and a trivial electron pocket centered at
k = (π, 0, 0).
In order to study the Fermi arcs, we again construct a slab geometry by trans-
forming the terms dependent on ky in Eqn. (3.28) into real space and considering a
system with L layers in the y-direction and infinite in the x- and z-directions. In the
type-I limit with γ = 0 shown in Fig. 3.8a and b, we find that away from E = 0, the
projections of the nodes are enclosed by isolated small Fermi pockets. These pockets
are connected to one another by topological Fermi arcs in the kx-direction. At E = 0,
the top and bottom arcs are degenerate along the lines kz = ±π/2. In a sense, this
type-I (γ = 0) limit in the inversion-breaking model is effectively composed of two
copies of a time-reversal-breaking Weyl semimetal separated by π reciprocal lattice
vectors along the kz direction.
When γ is increased to the type-II limit, the Fermi arc and bulk Fermi surface
configuration in the inversion-breaking case is very different from the time-reversal-
breaking model as we see in Fig. 3.8c and d. The projections of the Weyl nodes are
now enclosed by extended hole pockets for E < 0 (Fig. 3.8c) and electron pockets
for E > 0 (Fig. 3.8d). These Fermi pockets are connected by topological Fermi
44 Type I Critical Point Type II
a b c
d e f
g h i
Figure 3.7: Bulk band structure for type-I and type-II inversion breaking TWS. a-c The bulk band structure for the Hamiltonian in Eqn. (3.28). Electron pockets shown in red and hole pockets shown in blue merge at the Weyl nodes shown in green. Here we have chosen parameters ky = 0 with the parameters k0 = π/2, tx = t/2, m = 2t for (a) type I TWS with γ = 0, (b) the critical point between a type-I and a type-II TWS with γ = 2t and (c) type-II TWS with γ = 2.4t. The cones comprising the Weyl nodes develop a characteristic tilt of the type-II Weyl node as γ is increased. d-f Cuts through the Weyl nodes at ky = 0 and kz = −π/2 for the same parameters as (a-c). These cuts are shown as the green lines in (g-i). g-i Constant energy cuts through the nodal energy (E = 0) for the same parameters as (a-c). We see that for a type-I Weyl semimetal, there are no states at the Fermi energy. At the critical point between a type-I and type-II TWS, the density of states still vanishes. In the type-II regime, electron and hole pockets form near the Weyl nodes. These pockets enclose the projections of the Weyl nodes when the chemical potential is shifted away from E = 0. Trivial pockets also appear at k = (0, 0, 0) and k = (0, 0, π).
45 arcs, shown by thick red and blue lines, to pockets containing Weyl nodes of opposite
chirality. Unlike in the type-I limit, here the Fermi arcs connect pockets along the
kz-direction rather than the kx-direction. One might expect that the transition point where the Fermi arcs connect nodes in one direction rather than another is concurrent with the transition point between a type-I and type-II Weyl semimetal and indeed our numerical calculations show that is the case (see Fig. 3.9). Hence we see that for the same model with all other parameters held constant, merely tilting the nodes can lead to a dramatic recombination of the Fermi arcs and a qualitatively different pocket connectivity.
In Fig. 3.8c and d, we see that there are many states that are exponentially localized on the surface, however many of them form closed loops. We term these closed loops “track states”; they are degenerate in energy with the Fermi arcs but do not share their topology. Unlike Fermi arcs, track states form closed rather than open contours of surface states. By investigating the evolution of the Fermi arc and Fermi surface configuration as a function of γ (Fig. 3.9), we see that when the Fermi arc connectivity changes from the the kx-direction to the kz-direction, they leave behind
track states around the (kx, kz) = (π, π) point.
3.3.3 Surface States: Topological and Track
We briefly recapitulate the argument[9] for the existence of topologically protected
Fermi arcs in a Weyl semimetals. It can be shown that a Weyl node is a monopole
source of Berry curvature with charge equal to its chirality χ. We enclose an isolated
Weyl node by a closed 2D subspace of the Brillouin zone. It is well known that
the integral of Berry curvature over a 2D manifold is a quantized integer known
46 a b
Type I
c d
Type II
Figure 3.8: Fermi surface and Fermi arc configuration for type-I and type- II inversion-breaking Weyl semimetal. a,b The Fermi surface and Fermi arc configuration for the Hamiltonian given in Eqn. (3.28) in the type-I limit (γ = 0) calculated in a slab geometry with L = 50 layers and with bulk parameters the same as in Fig. 3.7a,d,g. We show this calculation at constant energies: E = −0.25t (a) and E = 0.25t (b). Here we see that Weyl nodes located at (kx, kz) = (±π/2, ±π/2) are connected by surface states (red and blue lines) to one of opposite chirality across the Brillouin zone in the kx-direction. c,d The Fermi surface and Fermi arc configuration for the Hamiltonian given in Eqn. (3.28) in the type II limit (γ = 2.4t) calculated in a slab geometry with L = 50 layers and with bulk parameters the same as in Fig. 3.7c,f,i. We show these for the same constant energies as above: (c) and E = 0.25t (d). The locations of the Weyl nodes are marked with pink arrows. We term the exponentially localized surface states that form closed loops “track states”. Fermi arcs are shown as bold lines and connect Weyl nodes in the kz-direction.
47 a b c d
Figure 3.9: Evolution of Fermi surface and Fermi arc configuration for inversion-breaking Weyl semimetal as a function of γ. a-d The evolu- tion of the Fermi surface and Fermi arc configuration in a slab geometry for Eqn. (3.28). Bulk states are down in black, surface states are shown in red and blue. We have chosen the parameters k0 = π/2, tx = t/2, m = 2t. The calculations are done at constant energy E = −0.25t for γ = 0 (a), γ = 0.8t (b), γ = 1.4t (c), and γ = 2. (d) shown in an extended Brillouin zone where both kx and kz range from −1.5π to 1.5π. We see that at the critical point between a type-I and type-II (d), the Fermi arcs that previously connected Fermi pockets in the kx-direction now connect Fermi pockets in the kz-direction and track states have formed on the bottom surface (blue) around the (kx, kz) = (π, π) point.
as the Chern number[76] when the bulk band structure is gapped over the region of integration. In the case of a surface enclosing a Weyl node, the Chern number calculated in this way is equal to the chirality χ of the node enclosed. By definition, such a surface enclosing a Weyl node defines a 2D Chern insulator and therefore possesses |χ| chiral edge modes on its boundary. As we consider various families of such closed surface in the Brillouin zone, these chiral edge modes trace out the open contours of surface states known as Fermi arcs that must terminate on Weyl nodes. In this way, there is a correspondence between the Berry curvature of the Weyl nodes, a topological property of the bulk, and the surface Fermi arcs (see sketch in Fig. 3.10a) that are also topological in nature.
48 Topological Protection of Fermi Arcs in Type-II Weyl Semimetals
The chirality and Berry curvature of a Weyl node are unaffected by its type[50].
In the case of the lattice models we consider in the sections above, this can be shown
explicitly by noting that the ith component of the Berry curvature of each band (E+ and E−) is given by ∂dk ∂dk dk · × ∂kj ∂kl Ωk,±,i = ±ijl 3 , (3.29) 4|dk|
where ijl is the rank 3 Levi-Civita tensor and dk ≡ (d1(k), d2(k), d3(k)) as defined in
Eqn. (3.17). Since the type of the Weyl node is determined by d0(k) which does not
enter Eqn. (3.29), the Berry curvature around a node is indeed manifestly invariant
with respect to its type.
The presence of topologically protected Fermi arcs relies on the quantized edge
modes of 2D surfaces enclosing Weyl nodes. We again emphasize that it is necessary
for such 2D surface to exist in a region which is fully gapped in the bulk. If one
constructs such a surface which intersects a bulk pocket, then it no longer describes
a Chern insulator and the quantization of the edge modes is destroyed. It is clear
that the extended pockets around type-II Weyl nodes play an important role in the
nature of the connectivity of the Fermi arcs and the pockets, since by definition one
necessarily cannot take a gapped 2D surface to lie within these pockets. Therefore,
the presence of Fermi arcs in a type-II Weyl semimetal is only guaranteed by ensuring
that the Gaussian surfaces one constructs in the Brillouin zone enclose Fermi pockets
rather than bare nodes.
We provide a simple counting argument that limits the possible connectivity of
Fermi arcs in a Weyl semimetal of either type:
49 1. If a Weyl node is type-I with chirality χ, then |χ| pairs of Fermi arcs will
terminate on the Weyl point when the Fermi energy lies at the nodal energy.
This well-known result is illustrated for the lattice models in Fig. 3.4b.
2. If an isolated Fermi pocket fully encloses n Weyl node of either type such that
a closed 2D subspace where the bulk band structure is gapped can completely
surround the pocket, then the Fermi arcs on a given surface will have net chiral-
ity χtot and terminate on the pocket. Here χtot is the total chirality of all Weyl Pn nodes enclosed by the Fermi pocket such that χtot = i=1 χi. For type-I nodes, this is illustrated by Fig. 3.4a,c and Fig. 3.8a,b. The lattice models illustrate
Fermi pockets enclosing the projections of isolated type-II Weyl nodes in Fig.
3.6a,c,d,f and Fig. 3.8c,d. We see in Fig. 3.4d,f that when the net chirality
enclosed is zero, Fermi arcs are not present.
3. When the chemical potential lies at the energy of a type-II Weyl node at least
two Fermi pockets are connected at the Weyl node. In this case, it is necessary to
consider the set of all connected Fermi pockets when determining the possible
Fermi arc configuration. When multiple Weyl nodes connect a set of Fermi
pockets such that the only gapped 2D subspace of the Brillouin zone surrounding
it contains a net chirality χtot = 0, then the net chirality of Fermi arcs on a
surface is also zero, even though Weyl nodes are present at the Fermi energy.
This is illustrated in Fig. 3.4e where two Weyl nodes of opposite chirality
connect a single pair of hole and electron pockets and Fermi arcs are absent
even at the nodal energy. However, when the net chirality of nodes connecting
the set of pockets is nonzero, then a set of Fermi arcs with net chirality χtot must
50 a Topological Arcs b Trivial c Track States
Figure 3.10: Sketch of the three types of surface states in a topological Weyl semimetal. a Two type-I Weyl nodes of opposite chirality connected by a Fermi arc on the top (red) and bottom (blue) surfaces. In an arbitrary type-II TWS at an energy away from the Weyl energy, these arcs would connect Fermi pockets instead of nodes. b A single Fermi pocket enclosing two nodes of opposite chirality. Since no Gaussian surface can be constructed in a region that is both gapped and encloses only one node, the only possible surface states are trivial ones, shown in red and blue at the boundary of the pocket that hybridize with bulk states due to lack of topological protection. c Pairs of Weyl nodes, two of each chirality with each node surrounded by a Fermi pocket. The pockets are connected by Fermi arcs (thinner red and blue contours) as well as track states (thicker blue lines) on the bottom surface. Note that states on opposite sides of a given loop of track states will disperse in opposite directions and so a Gaussian surface enclosing a given Fermi pocket will still have one net surface state of each chirality.
satisfy is that they must terminate somewhere on the set of Fermi pockets. This
has the striking consequence that even when the Fermi level lies at the node
energy and topologically protected Fermi arcs are present, the termination of
the surface arcs on the bulk pockets can occur at any point on the surrounding
Fermi pockets. We see this illustrated for type-II Weyl nodes connecting isolated
pairs of electron and hole pockets in Fig. 3.6b,e, where the Fermi level is at the
Weyl energy but Fermi arcs terminate on a bulk pocket a substantial fraction
of a reciprocal lattice vector from the Weyl nodes.
51 Surface ”Track States” in Type-II Weyl Semimetals
Topological Fermi arcs are not the only novel surface states possible in a type-II
Weyl semimetal. We have shown that due to the finite density of states available at type-II Weyl nodes, new surface states can be appear which we term “track states.”
These track states are degenerate with the Fermi arcs but do not share the topological properties of the arcs; instead track states form closed contours on a given surface which are contractible to points in the Brillouin zone. Although topologically trivial, track states appear to play an important role in determining the locations Fermi arcs may appear in the surface Brillouin zone.
Track states are generated when the connectivity of Weyl nodes changes as we tune the parameters of a system with multiple pairs of Weyl nodes. In Fig. 3.9, we see that by tuning the parameter γ in the Hamiltonian in Eqn. (3.28) through the type-I to type-II transition, the Fermi arcs shift locations. When the nodes are type-I,
Fermi arcs pair up nodes of opposite chirality in the kx-direction; when the nodes are type-II, Fermi arcs pair up nodes in the kz-direction. Because the Berry curvature is invariant with respect to γ, the Chern number of a bulk-gapped 2D subspace of the
Brillouin zone surrounding an isolated node does not change. Although the Fermi arcs can shift locations in the Brillouin zone, the net chirality of modes on a given surface is conserved. When γ = 2t, track states appear at the (kx, kz) points where
Fermi arcs were located in the type-I limit.
In a type-II Weyl semimetal, track states can also appear as the Fermi energy shifts. It is shown in Fig. 3.8c,d that as the Fermi energy changes from below the
Weyl energy in Fig. 3.8c to above the Weyl energy in Fig. 3.8d, the locations of the arcs shift. For E < 0, the arcs on the bottom surface (shown as thick blue contours)
52 connect across the kz = π line while track states are seen as closed blue contours
encircling the points (kx, kz) = (±π/2, 0) points. For E > 0, this pair of track states
have become a single track state encircling the (kx, kz) = (±π, 0) point and a pair of arcs connecting electron pockets across the kz = 0 line. A precisely analogous reconfiguration of states on the top surface also occurs as shown by the reorientation of the red contours.
We note that these track states can appear very similar to Fermi arcs when track states and arcs lie close together. Caution must therefore be taken when analyzing the surface Fermi state configurations of type-II Weyl semimetals in DFT calculations or in ARPES data. There is experimental evidence for the existence of track states in
MoTe2[55], WTe2[77, 78, 79], and a recent ARPES study of Ta3As2[80] has revealed closed contours of surface states which are strong track state candidates. The Ta3As2 system is particularly promising as it has been predicted that pressure can tune a type-I to type-II transition where track states are likely to appear.
3.3.4 Comparison with Experiments
In this section, we describe the current state of experimental realizations of topo- logical Weyl semimetals. Our results are summarized by Table 3.1. Although various ab-inito studies have proven useful in the study of the materials in Table 3.1, as well as the prediction of a variety of Weyl semimetals yet to be discovered experimentally, it is clear that there is a distinct need for a set of minimal models which describe the general features of topological Weyl semimetals. From the abundance of type II TWS in Table 3.1, it is particularly evident that our models provide a general framework
53 Material Type Broken symmetry Pairs of Weyl nodes Surface states TaAs[16, 17, 46] I Inversion 12 MoTe2[55] II Inversion 4 track states WTe2[77, 78, 79] II Inversion 4 track states LaAlGe[56] II Inversion 20 Ta3S2[81] II Inversion 4 track states
Table 3.1: Experimental realizations of Weyl semimetals.
for understanding the topological features of type II TWS which is complimentary to
DFT.
In Table 3.1, we note that other than the transition metal monophosphides, all of the Weyl semimetals which have been uncovered by spectroscopic experiments are of type II. Additionally, they all break inversion symmetry with strictly more than the minimum of two pairs of nodes. For this reason, we expect track states may be common in Weyl semimetals. Indeed, we have found that a detailed examination of the spectroscopic results indicate that evidence of track states is found in nearly all of the type II Weyl semimetals so far discovered.
The transition metal dichalcogenides MoTe2 and WTe2 each feature long surface states which begin on bulk electron pockets and terminate on bulk hole pockets.
These bulk pockets each enclose the projections of a net zero chirality of Weyl nodes and, by the arguments in Section 3.3.3 above, cannot have a nonzero net chirality of
Fermi arcs terminating on it. This is borne out in both the ab-initio calculations as well as the ARPES results[77, 78, 79, 55]. The long surface state in WTe2 has been shown[79] to have both topological and trivial character, depending on the material
54 parameters used in the ab-initio calculations and therefore the configurations of the
Weyl nodes. This is manifestly a characteristic of a track state.
The transition metal pnictide Ta3S2 features[81] 4 pairs of Weyl nodes which are formed from the merging of two hole pockets with an electron pocket. As we have shown in the section above, in such a configuration, there cannot exist a closed and gapped region of the Brillouin zone which encloses a net chirality of Weyl nodes. In this way, all surface states shown in Fig. 4 of Ref. [81] are in fact trivial in a topological sense. Additionally, Ta3S2 has set of surface states that lie close in momentum to the bulk hole pockets. The bulk band structure of Ta3S2 is predicted to be highly tunable and is has been predicted [81] that strain can tune transitions between type I and type II Weyl semimetals as well as between these semimetal phases and a strong topological insulating phase. It is possible that the track state nature of these surface states will be revealed by such an experiment.
3.3.5 Conclusions
The models we present here comprehensively describe the four classes of Weyl semimetals which can be delineated by the type of the nodes and whether they break inversion or time-reversal symmetry. Examples from each class have been predicted by theory and have been experimentally observed in quantum materials. Particular realizations obey point group symmetries different in general than those presented here. It is straightforward to extend the models we present here to study a Weyl semimetal with a chosen point group symmetry.
This summary of minimal models for type-I and type-II Weyl semimetals for both time-reversal-breaking and inversion-breaking cases may contribute to future
55 investigations of their properties in applied electric and magnetic fields. In particular, we expect our models to shed light on the nature of quantum oscillations in type-II
Weyl semimetals. Preliminary calculations[50] show the absence of a chiral zero- energy Landau level when the direction of the applied magnetic field lies outside of the tilt cone of the type-II Weyl node. However, these calculations rely on a linearized model for type-II Weyl nodes and a proper treatment should include the full Fermi pockets surrounding the Weyl nodes. The models presented here provide an ideal framework for such a calculation which we leave for future study. These models also provide a foundation for additional effects of repulsive and attractive interactions.
Experimental discoveries of magnetism and superconductivity in Weyl semimetals could provide impetus for such theoretical studies. In the following section, we shall see how the model given by Eqn. (3.28) played a key role in the discovery of the first type-II Weyl semimetal MoTe2.
3.4 Experimental Discovery of Weyl Semimetal MoTe2
To set the stage for interpretation of the experimental results, we investigate the two-band lattice model given by Eqn. (3.28) which breaks inversion symmetry but is invariant under time-reversal symmetry. The main lessons learned by examining this model are shown in Fig. 3.11 and summarized here: (1) The minimum number of four
Weyl nodes in this type II TWS occur at E = 0 at the touching point of electron and hole pockets in contrast with a type I TWS that has a zero density of states at E = 0.
The touching of electron and hole bands in our model is similar to the touching of the electron and hole bands in the experimental data shown in Fig. 3.12a and b.
(2) For a slab geometry, constant energy cuts at E = 0 show Fermi arcs on surface
56 termination A and B that connect Weyl points of opposite chirality. In addition there
are what we term “track states” that exist on the surface and pass through the WPs
but, unlike Fermi arcs, form closed loops. For E < 0, the projections of the WPs are
within the hole pocket, and at the surface the arc states connect the two hole pockets
and the track states loop around the electron pockets. The opposite is true for E > 0.
(3) The energy dispersion clearly shows a surface state dispersing separately from the
bulk bands and merging with the bulk bands close to the WP in Fig. 3.11d. This is
corroborated by the experimental data around the Weyl nodes in Fig. 3.12i and Fig.
3.13n where the arc merges with the bulk states.
Henceforth, we set the parameters of Eqn. (3.28) to m = 2t, tx = t/2, k0 = π/2,
and γ = 2.4t. The bulk band structure for this parameter choice was explored in
detail in the previous section and can be seen in Fig. 3.11a which shows hole and
electron pockets touching at the Weyl nodes as well as pockets disconnected from
the nodes. Similar Fermiology is also present in the MoTe2 system and we can gain
insight into this and other related materials by taking advantage of the lattice model’s
simplicity and tunability.
We again consider a slab geometry finite in the y- direction with L layers but infinite in the x- and z-directions. We label the states as “surface termination B”
(“surface termination A”) if they are exponentially localized at hyi = 1 (hyi = L).
Fig. 3.11 also shows the surface states at µ = ±0.1t overlaid on the bulk band structure. We show constant energy cuts through the band structure of the slab geometry in Fig. 3.11b and c for µ = ±0.1t. When µ < 0, the projections of the
Weyl nodes (shown by green dots) are enclosed by hole pockets. Each of these hole pockets are connected to another pocket containing a node of opposite chirality by
57 a d
Cut 1
b 1 2 c 1 2 e
Cut 2
Figure 3.11: Simple model of type II Weyl semimetal described by a two band model given by Eq. 3.28 which exhibits four Weyl nodes. a Electronic band structure for µ = ±0.1t indicated by the blue translucent plane. b,c The topological surface states and Fermi arcs on surface A (in red) and B (in blue) are calculated for a slab geometry confined along the y-direction. The bulk bands are shown in black. When µ = 0 exactly, the electron and hole pockets touch and the arcs terminate on the node (green dot) itself. For Fermi energy below (above) the nodal energy, arcs of surface states connect the Fermi hole (electron) pockets surrounding a node rather than terminating on a node. d,e Energy dispersion along kz at fixed kx as shown by cuts in panels (b, c). Cut 1 along kx = π/2 shows the bulk electron and hole bands touching at the node and the merging of surface states into the bulk away from the Weyl node. Cut 2 along kx = 0.63π shows a gap between the bulk bands and a surface state that disperses with opposite velocities at the projections of the two WPs. The WPs are located at (kx, kz) = (±π/2, ±π/2) indicated by pink arrows pointing to green dots.
58 one Fermi arc on surface A (B) shown as a thick light red (blue) line. When µ > 0, the projections of the Weyl nodes are enclosed by electron pockets which are similarly connected by Fermi arcs on the surfaces. At precisely µ = 0, because all of the nodes lie at E = 0, all Fermi arcs terminate on the nodes themselves as in a type I TWS.
The slab configuration energy dispersion for fixed kx is shown in Fig. 3.11d and
1e. These cuts are shown as green dashed lines labeled cut 1 and cut 2 respectively.
We can see that at the Weyl nodes, the red surface bands in 1d disappear into the bulk. As we move past the Weyl points in e, we see that these two red bands combine into a single continuous band.
3.4.1 ARPES Results
MoTe2 is a semimetal that crystallizes in a orthorhombic lattice. The Fermi sur- face of MoTe2 also has two 2-fold symmetry axes, along Γ - X and Γ - Y directions.
The lattice constants are a = 6.33 A,˚ b = 3.469 A.˚ Due to breaking of the inversion symmetry there are two different possible terminations of the cleaved sample surface, referred to as termination “A” and “B” respectively. The two different terminations also have different surface band structures as seen by laser-based angle resolved pho- toemission spectroscopy (ARPES) and corroborated by DFT calculations.
We identify electron and hole bands in the spectroscopic data shown in Figs. 3.12 and 3.13. The hole bands at the center of the Brillouin zone have a “butterfly” shape.
The electron pockets shaped like ovals are located on each side of the butterfly. There are also two banana like hole pockets partially overlapping the oval electron pockets.
The configuration of these pockets can be seen at the Fermi energy in Fig. 3.12a and
10 meV above the Fermi energy in Fig. 3.12b and their electron or hole character
59 0.4 a FS, 6.7eV f 40 π / W3exp kx = 0.36 ( b) j + 20 0.2 + 0 a) + -20 (meV)
0.0 f π /
( + W2calc -40 y
k + W2exp -60 -0.2 E - W3exp + -80 π / kx = 0.36 ( b) -100 -0.4 g 40 k = 0.32 (π / b) b FS + 10 meV x k 20 + 0 + a) -20
+ (meV) f
π / -40
( + y
k + -60 E -
-80 π / + kx = 0.32 ( b) -100
h 40 π / kx = 0.28 ( b) l c 0.4 Hole Band (HB) 1 Weyl points 20 HB 2 EB 1 0 0.2 -20 a) (meV) f 0.0 Fermi Arcs -40 π / (
y -60 E - k -0.2 -80 π / kx = 0.28 ( b) EB 2 Arc W2-W3 -100 -0.4 Arc W2-W2 i 40 π / kx = 0.24 ( b) W2exp -0.4 -0.2 0.0 0.2 m 20 π / kx ( b) 0 d 0.4 FS + 30meV FS - 30 meV -20 (meV) e f -40 + + a) 0.2 -60 + E - π / -80 π / ( 0.0 kx = 0.24 ( b) y + + k -100 + + -0.2 0.0 0.2 -0.2 0.0 0.2 -0.2 + π / π / -0.4 -0.2 0.0 -0.4 -0.2 0.0 ky ( a) ky ( a) π / π / kx ( b) kx ( b)
Figure 3.12: Experimental Fermi surface and band structure of MoTe2. a Constant energy intensity plot measured at EF using 6.7 eV photons for a sample with termina- tion A. The calculated (DFT) positions of Weyl points W2 are marked as pink dots, while experimentally determined locations of W2 and W3 points are marked as red dots. The chiralities of Weyl points are marked with “+” and “-” and their locations (kx, ky, E) are summarized in Table 3.4.2. b Same as in a above but taken at 10 meV above EF . c A sketch of constant energy contours of electron and hole bands showing the locations of Weyl points and Fermi arcs. d Constant energy contour measured at 30 meV above EF using 5.9 eV photons for a sample with termination B. Positions of calculated and measured Weyl points are marked as above. e Same surface termi- nation and photon energy as d but at 30 meV below EF . f - i Experimental band dispersion along cuts at kx = 0.24, 0.28, 0.32 and 0.36 π/b. j - m Calculated band dispersion for a sample with termination A along kx = 0.24, 0.28, 0.32 and 0.36 π/b. 60 is easily identified because hole (electron) pockets shrink (expand) with increasing
energy. A simplified sketch of constant energy contours of electron and hole bands is
shown in Fig. 3.12c.
The central hole pocket touches the electron pockets at four Weyl points shown
as red dots in Fig. 3.12a-c which we label as W2. The outer banana shaped hole
pockets also touch the oval electron pockets at two other Weyl points labeled as W3.
At surface termination A, Fig. 3.12b, those two types of Weyl points are connected by topological arcs seen as white-gray high intensity areas. For this surface termination there is no strong evidence for arcs connecting positive and negative chirality W2 nor positive and negative chirality W3 points. The situation for surface termination B is more complicated as shown in Fig. 3.12d. There seems to be a sharp contour connecting both sets of W2 and W3 points. Most likely this is a track state discussed above. The examination of constant energy plot at energy of 30 meV below EF
(Fig 3.12e), reveals that there are actually two bands present. In addition to the track state, there is also an arc present that connects positive and negative chirality
W2 points. Although present data does not allow us to definitely demonstrate a connection between positive and negative chirality W3 points, we can deduce that they are likely connected, so the arcs on surface A between W2-W3 together with arcs
+ − + − on surface B W2 -W2 and W3 -W3 form a closed loop when connected via the bulk of the sample.
We now examine the locations of the Weyl points in the band dispersion. In Fig.
3.12f-i we plot the band dispersion along ky cut for selected values of kx. At ky=0.36
π/b (panel f) two bands are clearly visible: an “M” shaped band at higher binding
energy and a “U” shaped band at slightly lower binding energy. Both bands appear
61 6.7 eV, FS, Top k = 0 (π / a) k = 0.05 (π / a) ky = 0.1 (π / a) k = 0.2 (π / a) a y y y 0.0 d e f g Cut d e f g d - g -0.1 6.7 eV 0.0 W2exp (eV) W2calc A f -0.2 W2exp E - / b) -0.3 π -0.1 (
x W3exp k -0.4 h - k h i j k W2exp 6.7eV 0.0 (eV)
-0.5 f B -0.6 b 0.0 E - 6.7 eV, FS, Bottom -0.1 -0.1 l m n o W2calc l - o 0.0 W2exp -0.2 W2exp 5.9 eV (eV) f
/ b) B -0.3 π E - ( W3exp x
k -0.4 -0.1 0.10 0.25 0.10 0.25 0.10 0.25 0.10 0.25 π π π π -0.5 kx ( /b) kx ( /b) kx ( /b) kx ( /b) 0.1 Band 3 -0.6 p - s p W2calc q r s 0.2 c A
5.9 eV, FS, Bottom (eV) f 0.0 0.1 Band 2
0.0 E - Band 1 -0.1 0.1 / b) -0.1 Band 3 t u v w π W2calc t - w W2calc ( x -0.2 B (eV) k W2exp f 0.0
-0.3 Band 2
W3exp E - Band 1 -0.4 -0.1 0.10 0.25 0.10 0.25 0.10 0.25 0.10 0.25 -0.4 -0.2 0.0 0.2 0.4 π k (π/b) k (π/b) k (π/b) k (π/b) ky ( / a) x x x x
Figure 3.13: Identification of Weyl points and Fermi arcs from experimental data. a Constant energy contour at EF , measured by 6.7 eV photons for surface termination A. DFT predicted locations for Weyl points W2 and measured Weyl points W2, W3 are marked as red and pink dots respectively. b The same panel as a except for surface termination B. c The same panel as b except for using 5.9 eV photons. d - g Energy dispersion for surface termination A along ky = 0, 0.05, 0.10 and 0.20 π/a. The projections of Weyl points W2 are marked as dots. h - k The same panels as (d - g) except for surface termination B. l - o The same panels as (h - k) except for using 5.9 eV photons. p - s Calculated band dispersion for surface termination A along cuts at ky = 0, 0.05, 0.10 and 0.20 π/a. Positions of W2 are marked similarly as above. t - w The same as (p - s) except for surface termination B. Bands plotted with darker lines have more surface weights.
62 connected at zero momentum with Dirac-like structure. As we move towards the
zone center, both bands move to lower binding energy and their energy separation
decreases. In panel h, the tips of the “M” shaped band (red dotted line) touches the
EF and form parts of the butterfly hole pockets. As these tips move above EF , they
touch merge with wings of the “U” shaped electron band (white dotted line) forming
two Weyl points approximately 20 meV above EF marked by black dots. At each
side of the symmetry line, they form two tilted cones characteristic of a type II Weyl
node. The data along kx direction are shown in Fig. 3.13d-o along with results of calculations (Fig. 3.13p-w) for the two surface terminations. The surface termination
A is characterized by lower binding energy of electron pocket in panels d-g, when compared to the data from surface termination B shown in panels h-k and l-o. The data in panels l-o best illustrates the formation of the W2 points. In panel l, the hole band is marked with red dashed line, while the electron band is marked with white dashed line. As we move away from the symmetry line, the separation between those bands becomes smaller and they merge at a point located ∼20 meV above EF marked by red dot in panel n. For higher values of ky momentum they separate again as seen in panel o. The DFT calculation also demonstrates the energy difference of the band locations for the two terminations and formation of the W2 Weyl point that agrees with experiment on a qualitative level.
The momentum location of the experimentally determined Weyl points is some- what different from DFT predictions (marked as pink dots in 3.12a,b) most likely due to high sensitivity of the band calculation to structural parameters. Table I summarizes the positions of WPs determined from experiment and DFT. Despite the discrepancy between the predicted locations of the Weyl nodes from DFT and where
63 Figure 3.14: Results of DFT calculations. a Calculated bulk Fermi surface of MoTe2 for kz = 0.6π/c and projections of W2 (kx, ky) = (±0.17 π/b, ±0.06π/a) are marked with pink dots. b Bulk band dispersion along W2-W2 direction (the vertical dashed line in a). DFT predicted positions of W2 (ky, E) = (±0.06π/a, 0.028 eV) are marked. DD DD c The dominant contribution for the divergence of the Berry curvature (Ωn,yz, Ωn,zx) for the n = N + 1 th band where N is the number of electrons in the unit cell with kz = 0. Red and blue indicate different chiralities of the two Weyl points. d - g Calculated constant energy contours of MoTe2. Darker bands are surface bands and lighter bands are bulk bands. d, e are at Fermi level for surface termination A and B. f, g are at Fermi level + 28 meV of surface termination A and B, respectively. h, i Surface band dispersions of termination A and B along W2-W2 direction. j, k Surface band dispersions of termination A and B along ky = 0.05 π/a direction, which is very close to the ky position of W2 (0.06 π/a). Positions of calculated Weyl points W2 are marked and darker bands have more surface weights in d - k.
64 they are located experimentally, in each case they are at the touching points of the electron and hole bands. In the ky = 0 cuts shown in Fig. 3.13d,h,l,p,t, band 1 is connected to bulk states below the Fermi level, while band 3 dips down and goes into bulk just before it reaches the Weyl point. As we increase ky, band 1 and band 3 merge together. In the ky = 0.1π/a cuts, the two bands merge into one band which goes through the position of the projection of W2. This behavior is exactly the the behavior predicted in Fig. 3.11d and 3.11e.
3.4.2 DFT and Topological Analysis
Fig. 3.14 is the DFT calculation of the band structure of MoTe2 performed by
Ryotaro Arita’s group. Fig 3.14a is the bulk Fermi surface for kz = 0.6π/c and calculated positions of four Weyl points are marked. The shapes of outermost electron and hole bands are very similar to our experiment result in Fig. 3.12b. Pink dots are projections of the calculated Weyl points on the kz = 0 plane from energy +28 meV above Fermi level, thus the electron band is not touching the two Weyl point projections. The surface weighted constant energy contours are shown in 3.14d -
3.14g. Fig. 3.14d and 3.14e are at Fermi surfaces of termination A and B, while
3.14f and 3.14g are at Fermi level + 28 meV, the DFT predicted energy of W2. In the calculations, W2 is not directly connected to another W2 by surface states on the
Fermi surface of termination A calculation while they are connected by weak and short surface states in termination B calculation. However, the W2 points are connected by bulk electron bands in termination A. This is consistent with our experimental results shown in Fig. 3.12a-e. Fig. 3.14b is the bulk band dispersion at W2-W2 direction, as the vertical dashed line shown in 3.14a. The two W2 points from DFT are right
65 kx (π/b) ky (π/a) E (meV)
W2 DFT ±0.17 ±0.06 28 W2 Exp ±0.24 ±0.12 20 W3 Exp ±0.37 ±0.25 30
Table 3.2: The locations (kx, ky, E) of the Weyl points from DFT and ARPES for weyl semimetal MoTe2.
at the touching points of one hole band and one electron band. Fig. 3.14h and 3.14i show termination A and B surface band dispersions along the same direction as in
Fig. 3.14b. The surface bands are to connect bulk states near the positions of the
Weyl points. Fig 3.14j and 3.14k are termination A and B surface band dispersions along ky = 0.05 π/a direction, as the horizontal dashed line shown in 3.14a. We also calculated the Berry curvature on Fermi surface. The bright points in Fig. 3.14c are possible singular points of the Berry curvature and DFT calculated W2 points are marked in red and blue, indicating different chiralities of the Weyl points. The summary of energy and momentum locations of Weyl points based on calculations and experiment are provided in Table 3.2.
66 Chapter 4: Thermoelectric Transport in Weyl Semimetals
In the previous chapter, we saw that Weyl semimetals have a truly unique fermi- ology. In particular, we saw that they feature three defining and interwoven char- acteristics: linearly dispersing low energy modes, monopoles of Berry curvature in momentum space centered at these nodes, and Fermi arcs on in the surface Bril- louin zones connecting projections of Weyl points. Since the first theoretical predic- tion of Weyl semimetals in solid state systems, various transport formalisms have been used to predict how charges move in these materials. Most notably, Weyl semimetals exhibit negative longitudinal magnetoresistance as a result of the chi- ral anomaly[18, 19, 20, 21, 22, 23, 24]. This phenomena is due to the chiral Landau levels of the Weyl nodes and relies on sensitive measurements at quite high magnetic
fields. Although theoretically interesting, we seek other transport quantities that will reveal the unique bulk electronic structure of Weyl semimetals.
The thermoelectric response of a material is in general an excellent way to probe the bulk electronic characteristics. In this chapter, we present the first comprehen- sive theory for the Nernst thermopower in Weyl semimetals. We find that the linear band structure of a Weyl semimetal with its unique temperature dependent chemical
67 potential (see previous chapter) leads to a distinctly identifiable peak in the temper- ature dependence of the Nernst coefficient. We also present experimental data from collaborators that show remarkable agreement with our prediction.
We also find that the Berry curvature leads to anomalous transverse currents in the absence of a magnetic field. Although this was first predicted in the type-I
Weyl semimetals [82], we extend this formalism to type-II Weyl semimetals for the
first time. In particular, we find that, due to the Berry curvature, the anomalous transverse thermoelectric coefficient is enhanced with increasing tilt in type-II Weyl semimetals.
Thus in this chapter, we identify clear signatures in thermoelectric transport for the defining bulk characteristics of Weyl semimetals. Both of these features can be well-explained by Boltzmann transport calculations and are readily accessible to experimental probes. In the next chapter, we shall also see that the third defining characteristic of Weyl semimetals, the presence of Fermi arcs, leads to a topological thermomagnetic conductivity.
4.1 Boltzmann Transport Theory
In this section, we define the bulk thermoelectric transport coefficients and intro- duce the Boltzmann transport formalism. We apply it to the general case of a lattice model with Berry curvature. In subsequent sections, we will apply these results to the lattice models for Weyl semimetals of the previous chapter.
4.1.1 Definition of Transport Coefficients
We define the thermomagnetic coefficients in the following way:
68 JE LEE LET E = · , (4.1) JQ LTE LTT −∇T where J is the charge current density, Q is the heat current density, E is the electric
field and ∇T is the gradient in temperature. Each of the thermomagnetic transport
coefficients is a tensor of the form
αβ αβ αβ Lxx Lxy L = αβ αβ . (4.2) Lyx Lyy
αβ αβ Due to symmetry, these must be antisymmetric tensors with Lxy = −Lyx .
The Isothermal Nernst Effect
We are interested in calculating the Nernst effect defined by
Ey αxyz = , (4.3) −∇xT
under the isothermal conditions
Jx = Jy = 0 (4.4)
and
∇yT = 0. (4.5)
From Eqn. (4.1) and Eqn. (4.63), it is clear that
J = LEE · E + LET · (−∇T ) = 0. (4.6)
Solving for the electric field, we obtain
E = −(LEE)−1 · LET · (−∇T ). (4.7)
Now, due to the isothermal condition in Eqn. (4.64), we have that
X EE −1 ET EE −1 ET EE −1 ET Ey = − (L )yj Ljx (−∇xT ) = − (L )yx Lxx + (L )yy Lyx (−∇xT ) (4.8) j
69 It is straightforward to obtain the inverse of LEE as
EE −1 EE −1 EE EE EE −1 (L )xx (L )xy 1 Lyy −Lxy (L ) ≡ EE −1 EE −1 = EE EE EE EE EE EE . (L )yx (L )yy Lxx Lyy − Lxy Lyx −Lyx Lxx (4.9)
EE EE We now take advantage of the fact that Lyy = Lxx and the antisymmetry property
EE EE Lxy = −Lyx to obtain
EE −1 EE −1 EE EE (L )xx (L )xy 1 Lxx Lyx EE −1 EE −1 = EE 2 EE 2 EE EE . (4.10) (L )yx (L )yy (Lxx ) + (Lxy ) Lxy Lxx
The y-component of the electric field now becomes