1: \tolerance = 10000
2: \documentclass[aps,prb,doublespace,twocolumn,showpacs]{revtex4}
3: %\renewcommand{\baselinestretch}{2}
4: \usepackage{amssymb}
5: \usepackage{amsmath}
6: \usepackage{graphicx}
7: \begin{document}
8: \flushbottom
9:
10:
11: \title{Electromagnetic response of layered superconductors with
12: broken lattice inversion symmetry}
13:
14: \author{B.~Uchoa$^1$, A.~H.~Castro Neto$^2$, and G.~G. Cabrera$^1$}
15:
16: \affiliation{$^1$ Instituto de F\'{\i}sica ``Gleb Wataghin''
17: Universidade Estadual de Campinas (UNICAMP),\\
18: C. P. 6165, Campinas, SP 13083-970, Brazil \\
19: $^2$ Department of Physics, Boston University, 590 Commonwealth Ave.,
20: Boston, MA 02215}
21:
22: \date{ July 17, 2003}
23:
24:
25: \begin{abstract}
26: We investigate the macroscopic effects of charge density waves (CDW) and
27: superconductivity in layered superconducting systems with broken lattice
28: inversion symmetry (allowing for piezoelectricity)
29: such as two dimensional (2D) transition metal dichalcogenides (TMD).
30: We work with the low temperature time dependent Ginzburg-Landau theory
31: and study the coupling of lattice distortions and low energy CDW collective modes to
32: the superconducting order parameter in the presence of electromagnetic
33: fields. We show that superconductivity and
34: piezoelectricity can coexist in these singular metals. Furthermore, our study indicates
35: the nature
36: of the quantum phase transition between a commensurate CDW phase
37: and the stripe phase that has been observed as a function of applied pressure.
38: \end{abstract}
39:
40: \pacs{74.70.Ad, 71.10.Hf, 71.45.Lr}
41:
42: \maketitle
43:
44:
45: \section{Introduction}
46:
47:
48: The quasi-two-dimensional (2D) transition metal dichalcogenides (TMD) \cite{Withers}
49: 2H-$\mathrm{NbS_2}$, 2H-$\mathrm{NbSe_2}$, 2H-$\mathrm{TaS_2}$ and 2H-$\mathrm{TaSe_2}$
50: are layered compounds where superconductivity coexists with a charge density wave (CDW)
51: state \cite{Wilson}. It has been documented experimentally that these
52: materials present anomalous
53: properties such as a decrease of the resistivity and
54: decoupling between planar and c-axis transport in the CDW phase \cite{transport},
55: anomalous impurity effects on the superconductivity and non-linear Hall
56: effect \cite{hall}, stripe phases \cite{stripes}, and more recently
57: angle resolved photoemission experiments (ARPES) have shown a quasi-particle lifetime
58: diverging linearly with the energy close to the Fermi surface \cite{valla}
59: in contrast to ordinary Fermi liquids where the lifetime diverges quadratically
60: with energy. Some of these properties are similar to the ones observed in high
61: temperature superconductors (HTc) \cite{valla_htc}. Contrary to HTc,
62: however, TMD are extremely clean systems and therefore the anomalous
63: behavior is undoubtedly intrinsic and sample independent. Therefore
64: the understanding of the anomalous metallic behavior in these 2D materials may provide
65: clues for the understanding of the anomalous physics that occurs in a broad class of
66: 2D systems where superconductivity occurs.
67:
68: One of us (AHCN), has proposed recently an unified picture
69: of the CDW and superconducting transitions that can explain some of the
70: anomalies observed experimentally \cite{Neto}. The theory proposes that
71: the mechanism of superconductivity is related to the pairing of the
72: CDW elementary excitations mediated by acoustic phonons via a piezoelectric coupling.
73: These excitations are Dirac electrons situated in the nodes of a CDW gap,
74: resembling the Fermi surface zone of graphite. In TMD, unlike the case of graphite,
75: the lattice inversion symmetry is broken in the CDW phase.
76: The breaking of the inversion symmetry in TMD,
77: as documented in neutron scattering experiments \cite{neutrons}, allows for
78: a piezoelectric coupling between charge carriers and phonons \cite{piezo}.
79: Since CDW formation is usually related to nested Fermi surfaces in
80: 1D systems, weak 2D nesting combined with strong variations in the electron
81: phonon coupling due to the tight-binding nature of the electronic orbitals,
82: can be responsible for the origin of the nodal CDW order parameter.
83: This model is able to correctly explain some
84: of the anomalous properties of TMD such as the
85: energy dependence of the quasi-particle lifetime (given by the imaginary part of the
86: Dirac fermion self-energy), the critical dependence of the
87: superconducting transition with the lattice parameters,
88: the reduction of the resistivity, and the metallic behavior in the CDW phase.
89:
90: However, a criticism has been raised to this theory based on the fact
91: that piezoelectricity normally occurs in insulating systems since metals
92: can screen the internal electric fields \cite{varma}. However,
93: Dirac fermions in 2D have a vanishing density of states at the Fermi energy and
94: therefore do not screen electric fields \cite{mele} allowing for the existence
95: of a metallic state (that is, a state with gapless fermionic excitations) and
96: piezoelectricity. Nevertheless, the possibility of coexistence of piezoelectricity
97: and superconductivity is indeed surprising. In such a system it would be
98: possible to generate super-currents by simply squeezing the sample.
99: In this work we show that there is no contradiction
100: between the existence of piezoelectricity and superconductivity in materials where
101: the low-lying excitations are Dirac fermions.
102: We investigate the current and charge fluctuations that would arise in these
103: systems under the view point of the collective modes.
104:
105: Using path integrals we derive a
106: semi-classical action that describes the coupling between plasmons (responsible for screening)
107: and acoustic phonons to the superconducting order parameter via the piezoelectric
108: coupling. Various well-known regimes are described by this action: (1) in the case of normal
109: electrons without superconductivity we recover the well-known results for collective modes and
110: screening in 2D and 3D \cite{fetter}; (2) in the case of Dirac fermions without superconductivity
111: we re-obtain the results for collective modes and screening in semi-metals like graphite \cite{Shung};
112: (3) in the absence of piezoelectric coupling, we recover the behavior of an ordinary type II superconductor
113: \cite{tinkham}. However, when we allow the piezoelectric coupling with Dirac fermions and
114: superconductivity, new effects appear in the electromagnetic response.
115: Moreover, our results shed light on the origin of the quantum phase transition
116: between the commensurate CDW phase and the stripe phase in TMD.
117: We also investigate the collective modes that appear when combined with low lying energy bulk plasmons.
118:
119: The article is organized in the following manner: in section II we
120: discuss the plasmons in layered nodal liquids and specifically in TMD;
121: in section III we develop the semi-classical calculation in general grounds
122: and apply it to layered compounds like TMD in section IV. We have added an
123: appendix where the details of the calculation are presented.
124:
125: \section{Nodal liquid plasmons}
126:
127: The field theory of nodal liquids \cite{balents} is built under the basic idea
128: that the nodes of a CDW order parameter lead to two distinct subsystems
129: which correspond to the two components of the Dirac fermion spinor $\Phi_\sigma^\dagger =
130: (\psi_{+\,\sigma}^\dagger , \psi_{-\,\sigma}^\dagger)$, with $+$,$-$ indexing
131: respectively the fermionic particles and anti-particles (holes).
132: The non-interacting low energy Hamiltonian that describes the elementary excitations inside
133: the Dirac cone is:
134: \begin{equation}
135: \mathcal{H}_D = \sum_\sigma \int_{BZ} \frac{\mathrm{d}^3 k}{(2\pi)^3}
136: \Phi_\sigma^\dagger(\mathbf{k}) \,\hbar \left( v_{F}\sigma_x k_x +
137: v_{\Delta}\sigma_y k_y \right)\Phi_\sigma(\mathbf{k})\,, \label{H1}
138: \end{equation}
139: where $\mathbf{k}$ is the momentum, $\sigma_{x,y}$ are Pauli matrices that
140: act in the particle-anti-particle subspace,
141: $v_F$ and $v_\Delta$ are the anisotropic velocities perpendicular and parallel
142: respectively to the Fermi surface, and $BZ$ is the first
143: Brillouin zone of size $2\pi/d$ along the $k_z$ axis ($d$ the inter-plane distance).
144: When the chemical potential, $\mu$, intercepts the Dirac point,
145: the Hamiltonian (\ref{H1}) leads to the zero order polarization function \cite{Vozmediano}
146: at $T=0$, whose complex conjugate is given by:
147: \begin{equation}
148: \Pi^{0\,*}(\omega,\mathbf{q}) = - \frac{v_F}{8 d\,\hbar v_{\Delta}} \,
149: \frac{\bar{q}^2}{\sqrt{v_F^2\bar{q}^2 - \omega^2}} \label{P}\,,
150: \end{equation}
151: where $\omega$ is the frequency,
152: $\bar{\mathbf{q}} = \vec{q}_x + (v_\Delta/ v_F)\vec{q}_y$ the anisotropic
153: in-plane momentum. In the absence of hopping between the planes or interactions
154: with the lattice, the collective excitations of the Dirac fermions are due to the 3D Coulomb
155: interaction, $V_0$, between carriers in different planes:
156: $V_0(q,k_z) = 2\pi\,d\, e^2/(\epsilon_0 q) S(q,k_z)$,
157: with a structure factor $ S(q,k_z) = \sinh(q d)/[\cosh(q d) - \cos(k_z d)]$ identical to
158: the layered electron gas (LEG) case \cite{Hawrylak} ($e$ is the electron
159: charge, and $\epsilon_0$ the dielectric constant). In the random phase approximation (RPA)
160: the electric susceptibility is given by: $\epsilon(\omega,\mathbf{q},k_z) = 1 - V_0(q,k_z)\,
161: \Pi^{0}(\omega,\mathbf{q})$. At the points where the susceptibility vanishes,
162: that is, $\epsilon(\omega_p({\bf q},k_z),\mathbf{q},k_z)=0$, one obtains the plasmon
163: dispersion, $\omega_p({\bf q},k_z)$. From (\ref{P}), we notice that no plasmon modes
164: are allowed for Dirac fermions.
165:
166: The displacement of the chemical potential from the Dirac point generates a pocket around the nodes
167: which drastically changes this picture. The density of states becomes finite
168: at the Fermi
169: surface with Fermi momentum $k_F^*$, and Fermi energy $E_F^* = \hbar k_F^* v_F$,
170: giving rise to intra-band excitations in the cone.
171: In this case we recover the optic plasmon $\omega_p(q) = \sqrt{\omega_0^2 (\bar{q}/q)^2 + (v_0
172: \bar{q})^2}$ for the $k_z = 0$ bulk mode, where $v_0$ is the plasmon speed,
173: and the 2D acoustic mode $\omega_p(q) \propto \sqrt{\bar{q}^2/q}$ for the rest
174: of the plasmon band $0<k_z \leq \pi/d$. The intra-band contribution is satisfactorily
175: understood in the context of doped graphite \cite{Shung}
176: where for $v_F\bar{q} \leq \omega < v_F (2k_F^* - \bar{q})$
177: there is gap in the particle-hole continuum of the
178: inter-band excitations (see fig.1) defined by the imaginary part of (\ref{P}).
179: The optical plasmon is free of Landau damping in
180: the long wavelength limit and its energy is of the order of
181: $E_F^*$. Notice that in the anisotropic case ($v_F \neq v_{\Delta}$)
182: the gap in the plasmon spectrum depends on the direction
183: around the pocket Fermi surface.
184:
185: \begin{figure}[h!]
186: {\centering \resizebox*{2.4 in}{!}{\includegraphics{picture1.eps}} \par}
187: \caption{{\small Schematic drawing of the $k_z = 0$ bulk plasmon in an isotropic pocket
188: with Fermi momentum $k_F^*$. $\nu = \omega/v_F$ and $q$ is the
189: in-plane transfer momentum. The shaded areas correspond to the
190: particle-hole continuum due to intra-band and inter-band excitations. }}
191: \label{excitation_spectrum}
192: \end{figure}
193:
194: If we also add the piezoelectric electron-phonon coupling \cite{Neto}
195: \begin{eqnarray}
196: \mathcal{H}_{EP} = \gamma \sum_\sigma \int \mathrm{d}^3 x\,
197: \phi(\mathbf{x})\,\Phi_\sigma^\dagger(\mathbf{x}) \Phi_\sigma(\mathbf{x})\,,
198: \label{HEP}
199: \end{eqnarray}
200: due to acoustic phonons (with phonon field $\phi(\mathbf{x})$)
201: with energy $\hbar \omega_{\mathbf{q}}$ into the pocket excitations,
202: the RPA electric susceptibility acquires a correction:
203: $\epsilon(\omega,\mathbf{q},k_z) = 1 - [ V_0(q,k_z) +
204: (\gamma^2/\hbar) D^0(\omega,\mathbf{q})] \Pi^{0}(\omega,\mathbf{q},\mu)
205: $ where \cite{fetter}
206: $$
207: D^0(\omega,\mathbf{q}) = \frac{\hbar \omega_{\mathbf{q}}^2 }{\omega^2 - (\omega_{\mathbf{q}} - i \eta)^2}\,,
208: $$ is the phonon propagator
209: that affects very little the optical bulk plasmon in the $q \rightarrow 0$ limit.
210:
211: Besides doping, coherent inter-layer hopping is known to drive a dimensional crossover
212: in direction to a 3D system, leading as well to a pocket formation in a nodal
213: liquid \cite{hopping}. The energy of the pocket is of the same order of the inter-layer hopping energy,
214: which in
215: TMD NbSe$_2$ and NbS$_2$ (we will drop the 2H prefix from now on) were calculated in $\sim$ 0.1 eV
216: \cite{Doran}. We notice that inter-layer interactions were not taken into account into
217: those band calculations and that this energy could be considerably smaller. More experimental studies are required to investigate the nature
218: of the low energy collective modes in these materials. However, we can consider the plasmon
219: modes in real materials to have a finite energy, like in ordinary metals. Nevertheless,
220: the plasmon frequency will be smaller than in ordinary metals because of the low
221: density of states in the system.
222:
223:
224:
225:
226:
227: \section{Semi-classical dynamics}
228:
229: We consider the problem of Dirac fermions described by (\ref{H1}) coupled to
230: lattice vibrations described by the Hamiltonian
231: \begin{eqnarray}
232: H_{PH} = \sum_n \frac{{\bf P}_n^2}{2} + \frac{1}{2} \sum_{n,m} K_{nm}
233: \left({\bf R}_n-{\bf R}_{m}\right)^2
234: \label{HPH}
235: \end{eqnarray}
236: where $n$ labels the lattice site $R^i_n$, $P^{i}_n$ is the lattice momentum
237: operator with $i=x,y,z$ (these operators are canonically conjugated, namely,
238: $[P^i_n,R^j_m]=i \hbar \delta_{ij} \delta_{nm}$), $K_{nm}$ is the coupling
239: matrix. In what follows we will also consider the coupling
240: of the Dirac fermions and lattice degrees of freedom to a {\it classical} electromagnetic field
241: described by a vector potential field, ${\bf A}({\bf r})$, and electromagnetic
242: field tensor, $F^{\mu \nu} = \partial^{\mu} A^{\nu} - \partial^{\nu} A^{\mu}$
243: ($\partial^{\mu} = \partial/\partial x_{\mu}$ with $\mu,\nu=0,i$)
244: with electromagnetic energy density:
245: \begin{eqnarray}
246: E_{el} = - \frac{1}{16 \pi} F_{\mu\nu} F^{\mu\nu}
247: \end{eqnarray}
248: where $F_{0j} = - E_j$ is the electric field strength (we use the metric
249: $g_{\mu,\nu}=(1,-1,-1,-1)$) \cite{jackson}. In a piezoelectric the electric
250: field couples to the lattice distortion, $X^i_n =R^i_n-R^i_{n+1}$,
251: via the piezoelectric tensor, $\Delta^i_{\,\,j}$, by:
252: \begin{eqnarray}
253: H_{P} = \sum_n \Delta^i_{\,\,j} E_i X^j_n
254: \label{HP}
255: \end{eqnarray}
256: where repeated indices are to be summed.
257: The coupling between Dirac fermions and lattice vibrations
258: is given by (\ref{HEP}). The full problem can be written in path integral form
259: for the generating functional of the problem:
260: \begin{eqnarray}
261: Z = \int D{\bar \psi} D \psi \int D{\bf X} \exp\left\{\frac{i}{\hbar} \int dt \int d{\bf r}
262: \mathcal{L}[\psi,{\bf X},A_{\mu}] \right\} \, ,
263: \end{eqnarray}
264: where $\mathcal{L}$ is the Lagrangian (real time) of the problem (${\bar \psi}$, $\psi$ are
265: Grassmann variables).
266:
267: Let us consider first the case of the normal (non-superconducting) state.
268: We are only interested
269: in the long wavelength physics of the electronic problem which, as argued in the previous
270: section, is described by the plasmon mode. The plasmon mode can be separated from the particle-hole
271: continuum using the formalism developed by Bohm and Pines \cite{bohm_pines} and the
272: Lagrangian of the problem reduces to:
273: \begin{eqnarray}
274: \mathcal{L}_{N} &=& \mathcal{L}_{el} + \mathcal{L}_{plasmon} +
275: \mathcal{L}_{phonon} +
276: \mathcal{L}_{piez} \nonumber\\
277: &=& - \frac{1}{16\pi} F_{\mu\nu}F^{\mu\nu} - \frac{1}{c}
278: j_{N}^{\,\,\mu}A_{\mu} +
279: \frac{1}{2} \Omega \left[ \omega_p^{-2}(\partial_{t}\rho_N)^2 - \rho_N^2 \right]\nonumber\\
280: && + \frac{1}{2}\kappa\,[(\partial_{t} X^{i})^2 - v_{ph}^2 (\partial_i X^{i} )^2 ]
281: + \Delta^{i}_{\,\,j}
282: X_{i}F^{0j} ,\qquad \label{L}
283: \end{eqnarray}
284: where $\Omega \approx 4 \pi/k_c^2$, with $k_c$ the 3D electron
285: screening wave vector (for an isotropic metal
286: $k_c = \sqrt{3 \pi n e^2/(2 E_F)}$ where $n$ is the 3D electron density
287: and $E_F$ the Fermi energy \cite{bohm_pines}),
288: $\kappa$ is the lattice mass density,
289: $v_{ph}$ is the sound velocity,
290: $j_{N}^{\,\,0}/c = \rho_{N}$ is the density of normal electrons and
291: $j_{N}^{\,\,i}$ is the normal Ohmic current density (in our notation,
292: $\partial^{0} = c^{-1}\partial_t $ is the time derivative
293: normalized by the speed of light $c$).
294:
295: The collective properties of the electrons in the normal phase
296: are the same of the electrons in the superconducting phase, since the opening
297: of the superconducting gap does {\it not} affect the real part of the electronic polarization function,
298: $\Pi(\mathbf{q})$,
299: in the long wavelength limit \cite{Leggett}. In fact, within RPA one can easily show
300: in the context of BCS theory that $\mathrm{Re}\, \Pi(\mathbf{q})$
301: remains essentially unchanged
302: until the superconducting gap is
303: of the order of the Fermi energy, when the BCS pairing approximation becomes invalid
304: \cite{unpublished}.
305: In other words, for small $q$ the plasmon is not sensitive to the superconductor phase transition.
306: Thus, normal and superconducting electrons screen electric fields exactly in the same way and
307: therefore enter at equal footing in the Lagrangian. Thus, irrespective of the
308: phase, normal or superconducting, $\rho_N$ in (\ref{L})
309: can be replaced by the {\it total} electron density $\rho=\rho_N + \rho_s$,
310: where $\rho_s$ is the density of superconducting
311: electrons ($\rho_s=0$ in the normal state). Normal currents, however, are due to the quasi-particle
312: excitations while the
313: supercurrents, $j_{s}^{\,\,\mu}$, are due to the ground state of the condensate \cite{Bardeen}.
314: Thus, in the superconducting state the vector potential only couples to the normal currents.
315: At finite temperature, $T$, the number of normal and superconducting electrons is
316: not conserved. For simplicity, we have chosen to work at $T=0$,
317: where all the electrons are in the condensate and the total current is $j^{\mu} = j_{s}^{\,\,\mu}$.
318:
319:
320: In the superconducting phase we can introduce the superconducting order parameter, $\Psi$,
321: via a standard Hubbard-Stratanovich transformation and trace over the
322: electrons \cite{hub}. The generating functional reads:
323: \begin{eqnarray}
324: Z &=& \int D{\bar \Psi} D\Psi \int D{\bf X} \exp\left\{\frac{i}{\hbar} \int dt \int d{\bf r}
325: \left(\mathcal{L}_N[\rho,{\bf X},{\bf A}] \right. \right.
326: \nonumber
327: \\
328: &+& \left. \left.
329: \mathcal{L}_{GL}[\Psi,{\bf A}]\right)\right\}
330: \label{Z}
331: \end{eqnarray}
332: where $\mathcal{L}_{GL}$ is the Ginzburg-Landau Lagrangian.
333: In order to describe the macroscopic current-charge fluctuations we must also include a time
334: dependence in the superconductor order parameter. This is not an obvious
335: task but it can be done in two special limits: close to $T = 0$, and near the critical temperature
336: \cite{Tsuneto}. In the later, the validity of the
337: time-dependent Gorkov equations expansion in the gap, $\Delta_s$, require frequencies
338: higher than the binding energy of the Cooper pairs, $\hbar \omega \gg \Delta_s$. In
339: this limit the fluctuations have enough energy to break the Cooper pairs and
340: convert them into single particle excitations, leading to a diffusive regime.
341: As we mentioned above, we will work in the opposite limit $T = 0 $ where the hydrodynamic
342: description at $\hbar \omega < \Delta_s$ is rigorously valid.
343: We introduce the low-temperature time-dependent Ginzburg-Landau Lagrangian
344: \cite{Tsuneto,Parks},
345: \begin{eqnarray}
346: \mathcal{L}_{GL} &=& - \alpha |\Psi|^2 - \beta |\Psi|^4 - \frac{1}{2m^\star} \left|
347: \left( \frac{\hbar}{i} \nabla -
348: \frac{e^\star}{c} \mathbf{A} \right)\Psi \right|^2 \nonumber \\
349: && + \frac{1}{2
350: m^\star v_{s}^2} \left|\left( \frac{\hbar c}{i} \partial^{0} + e^\star \phi
351: \right) \Psi \right|^2\,.
352: \label{GL}
353: \end{eqnarray}
354: which is specially suitable for type II superconductors, where the penetration length is much
355: larger than the coherence length (in (\ref{GL}) $\phi$ is the electrostatic
356: potential and all the other symbols are standard \cite{tinkham}).
357: In the clean limit, $v_s$ is typically of the order of the Fermi velocity $v_F$.
358: Under the assumption that the fields vary much slower than the coherence length, we assume
359: that the magnitude of the superconductor order parameter is
360: constant and therefore all the fluctuations come from the superconducting
361: phase, $\varphi(t,\mathrm{x})$. Thus, the order parameter is written as:
362: \begin{eqnarray}
363: \Psi(\mathrm{x}) = \Psi_{0}\, \mathrm{e}^{i \varphi(t,\mathrm{x})} \, .
364: \end{eqnarray}
365: Despite the gauge invariance of (\ref{GL}), we conveniently assume
366: the transverse gauge, $\nabla \cdot \mathbf{A} = 0$, in what follows.
367: Notice, however, that
368: $\mathcal{L}_{N} + \mathcal{L}_{GL}$ is not complete because
369: the fields and the supercurrents remain decoupled. This problem is solved by
370: assuming the validity of the non-homogeneous Maxwell equations.
371: The Poisson equation is clearly a constraint between the density of
372: superconducting electrons and the electrostatic potential,
373: $$
374: g(t, \mathbf{x}) \equiv \nabla^2 \phi + 4 \pi \rho_{s} = 0 \,,
375: $$
376: and has to be enforced using a Lagrange multiplier $\Lambda(x^{\mu})$
377: in final Lagrangian.
378:
379: In the semi-classical regime, $\hbar \to 0$,
380: the behavior of the fields comes from the minimal action principle
381: $$
382: \delta S = \delta \! \!\int \, \mathrm{d}^4 x \,\left( \mathcal{L}_{N}
383: +\mathcal{L}_{GL} + \Lambda\,g \right) = \delta\!\! \int \, \mathrm{d}^4 x \,
384: \mathcal{L} = 0\, ,
385: $$
386: with respect to the variables $A^{\mu}$, $\rho_{s}$,
387: $X^{i}$, $\varphi$ and $\Lambda$.
388: By minimizing with respect to $A^{\mu}$ and $\Lambda$, we obtain
389: the two non-homogeneous Maxwell's equations,
390: \begin{equation}
391: \nabla^{2} \phi + 4 \pi \rho_{s} = 0 \label{phi1}
392: \end{equation}
393: \begin{equation}
394: - \square \mathbf{A} - \partial_{0} \nabla \phi + \frac{4
395: \pi}{c} \mathbf{j}_{s} = 0 \label{A1}
396: \end{equation}
397: with the supercurrent $j_{s}^{\mu}$ given by
398: \begin{equation}
399: \rho_{s} = - \frac{e^\star \Psi_{0}^2}{m^\star v_{s}^2} \left(\hbar\,c\,
400: \partial_{0} \varphi + e^\star \phi \right)
401: - \,\delta \,\nabla \cdot \mathbf{X} - \nabla^2 \Lambda \label{charge}
402: \end{equation}
403: \begin{equation}
404: \mathbf{j}_{s} = \frac{e^\star \Psi_{0}^2}{m^\star}
405: \left[\hbar\, \nabla \varphi
406: - \frac{e^\star}{c} \mathbf{A}\right] + \, c\,\delta \,\partial_{0}
407: \mathbf{X} \, .\label{j}
408: \end{equation}
409: For simplicity we will assume that $\Delta^{i}_{\,\,j} = \delta \delta_{ij}$ is a diagonal
410: and isotropic tensor. The minimization with respect to the other fields
411: completes the set of equations:
412: \begin{equation}
413: \frac{1}{4\pi}\,\Omega\,\left(\omega_p^{-2} \partial_t^2 \rho_{s} + \rho_{s}\right)
414: = \Lambda \label{rho}
415: \end{equation}
416: \begin{equation}
417: \nabla^2 \varphi
418: - \frac{c^2}{v_{s}^2}(\partial^{0})^2 \varphi -
419: \frac{e^\star c}{\hbar\,v_{s}^2}\, \partial^{0}\phi = 0 \label{varphid}
420: \end{equation}
421: \begin{equation}
422: \kappa \left( \partial_t^2 \mathbf{X} - v_{ph}^2 \nabla^2 \mathbf{X}\right) -
423: \delta ( \nabla \phi + \partial^{0} \mathbf{A} ) = 0 \label{X} \,.
424: \end{equation}
425: Substituting the supercurrent
426: in the continuity equation, $\partial_{\mu} j_{s}^{\,\,\mu} = 0$, we find that
427: \begin{eqnarray}
428: c\,\partial_{0}\rho_{s} + \partial_{i} j_{s}^{\,\,i}
429: = -\, c\,\partial_{0} \nabla^2 \Lambda = 0 \nonumber \,,
430: \end{eqnarray}
431: and therefore, in the absence of external currents $\Lambda$ is either a function of
432: space or of time, but not of both. If $\Lambda$ depends on time, then $\rho$
433: is a stationary time dependent homogeneous (or non-periodic) distribution because of (\ref{rho}),
434: what obviously violates the charge conservation. Therefore,
435: we conclude that $\partial_0 \Lambda = 0$.
436:
437: The Lagrange multiplier defines two distinct classes of solutions for the charge density:
438: ({\it i}) {\it screening-like} when $\Lambda(\mathbf{x}) \neq 0$, and ({\it
439: ii}) {\it plasmon-like} when $\Lambda \equiv 0$.
440: In the first case, $ \rho(\omega,\mathbf{k}) \propto f(\mathbf{k})
441: \delta(\omega) $, where $f({\bf k})$ depends on the boundary conditions, has no dynamics
442: and describes the physical response to a boundary perturbation such as a
443: static squeeze of the crystal or the introduction of a charge probe. In case
444: ({\it ii}), the electrons can oscillate freely with the plasmon frequency,
445: allowing the existence of normal modes.
446:
447:
448: \section{Collective modes}
449:
450: We consider the problem of a layered solid (such as the TMD)
451: made by an infinite number of weakly interacting
452: planes. In the continuum limit this problem becomes an effective
453: 3D model with spatial anisotropy in the direction perpendicular to the
454: planes, say the $z$ axis. It is convenient to define the
455: diagonal anisotropy tensor $\tau^{i}_{\,\,j} = (1,1,\tau )$,
456: where $\tau$ is the anisotropy parameter, (we introduce the notation
457: $\tilde{\partial}^{i} \equiv \tau^{i}_{\,\,j}\partial^{j}$ and
458: $\tilde{A}^{i} \equiv \tau^{i}_{\,\,j} A^{j}$) and rewrite
459: the Ginzburg-Landau Lagrangian as a time-dependent Lawrence-Doniach model \cite{tinkham}:
460: \begin{eqnarray}
461: \mathcal{L}_{GL} &=& - \alpha |\Psi|^2 - \beta |\Psi|^4 - \frac{1}{2m^\star} \left|
462: \left( \frac{\hbar}{i} \tilde{\partial}^{i}\nonumber -
463: \frac{e^\star}{c} \tilde{\mathbf{A}} \right)\Psi \right|^2 \nonumber \\
464: && + \frac{1}{2
465: m^\star v_{s}^2} \left|\left( \frac{\hbar c}{i} \partial^{0} + e^\star \phi
466: \right) \Psi \right|^2\,.
467: \label{GL2}
468: \end{eqnarray}
469: The model leads to the anisotropic version of equation
470: (\ref{varphid}),
471: \begin{equation}
472: \tilde{\nabla}\cdot \tilde{\nabla} \varphi
473: - \frac{c^2}{v_{s}^2}(\partial^{0})^2 \varphi -
474: \frac{e^\star c}{\hbar\,v_{s}^2}\, \partial^{0}\phi -
475: \frac{e^\star}{\hbar\,c}\,\tilde{\nabla}\cdot \tilde{\mathbf{A}} = 0 \,,
476: \label{VarphiAnys}
477: \end{equation}
478: and to the appearance of Josephson currents between the planes,
479: represented by an anisotropic supercurrent:
480: \begin{equation}
481: \mathbf{j}_{s}(t,\mathbf{x}) = \frac{e^\star \Psi_{0}^2}{m^\star}
482: \left[\hbar\, \tilde{\nabla} \varphi(t,\mathbf{x})
483: - \frac{e^\star}{c} \tilde{\mathbf{A}}(t,\mathbf{x})\right] + \, c\,\delta \,\partial_{0}
484: \vec{X}(t,\mathbf{x}) \, .\label{jLD}
485: \end{equation}
486: From now on, we use an overhead symbol to represent the {\it in-plane} vectors $\vec{x}$, $\vec{q}$, $\vec{X}$,$\vec{\nabla}$ and etc. In our notation, we define $\vec{q}$ as the in-plane component of the momentum $\mathbf{k}$, such that $\mathbf{k} = (q,k_z)$ and $\tilde{\mathbf{k}} = (q,\tau k_z)$ .
487:
488: In highly anisotropic layered compounds the absence of piezoelectricity along the $z$
489: axis is justified by the weak distortions of the ions in this direction.
490: The $z$-component of the sound velocity is also very small due to the weak elastic
491: coupling between planes. In the $v_{ph\,z}/v_{ph} \rightarrow 0$ limit, the phonon dispersion
492: is $\omega_{ph} = v_{ph}q $ and
493: the in-plane phonon equation (\ref{X}) has a cylindrical symmetry
494: \begin{eqnarray}
495: && \kappa \left( - \omega^2 + \omega_{ph}^2 \right)\vec{X}(\omega, \mathbf{k}) \nonumber\\
496: && \qquad\qquad - \,i \,\delta \left[ \vec{q}\, \phi(\omega,
497: \mathbf{k}) + \frac{\omega}{c} \vec{A}(\omega, \mathbf{k})\right] = 0\,.\qquad
498: \label{xAnys}
499: \end{eqnarray}
500:
501: Given the plasmon frequency, $\omega_p(\mathbf{k})$,
502: the equation for the charge density in Fourier space becomes:
503: \begin{equation}
504: \Omega
505: \left[ - \omega^2 + \omega_{p}^2(\mathbf{k})\right] \rho_{s}(\omega,\mathbf{k}) =
506: 8\pi^2\,\omega_p^2(\mathbf{k})\,\Lambda(\mathbf{k})
507: \,\delta(\omega)\,.\label{rhoLD0}
508: \end{equation}
509: The screening-like solution is:
510: \begin{equation}
511: \rho_{s}(\omega,\mathbf{k}) = 8\pi^2\,\frac{\Lambda(\mathbf{k})}{\Omega}
512: \,\delta(\omega)\,,\label{rhoLD}
513: \end{equation}
514: for a non-zero $\Lambda(\mathbf{k})$.
515: After a straightforward calculation, the simultaneous solution of equations
516: (\ref{phi1})$-$(\ref{charge}), (\ref{VarphiAnys})$-$(\ref{xAnys}) and (\ref{rhoLD})
517: yields $\rho_{s}(\omega,\mathbf{k}) = \Lambda_{0}(\hat{\mathbf{k}})\,\delta\left(k^2 + k_{0}^2\right)$,
518: for some $\Lambda_0$ function, with the characteristic momentum
519: \begin{equation}
520: k_{0}^2 = 4\pi \left( \frac{e^{\star\,2}\Psi_{0}^2}{m^\star\, v_{s}^2} -
521: \frac{\delta^2}{\kappa\,v_{ph}^2} \label{k0}
522: \right)\,.
523: \end{equation}
524: The $k_0^2 > 0$ case leads to exponentially decaying solutions that describe the screening
525: induced by proper boundary conditions like the squeeze or shear of the crystal.
526: If the piezoelectric coupling is larger than a critical value ($k_0^2 < 0$), the screening is suppressed
527: and a quasi-static charge modulation should be observed, in such way to
528: minimize the elastic energy. This analysis is confirmed by
529: introducing an external charge probe $Q$ at the origin:
530: we re-obtain the well known Thomas-Fermi result \cite{fetter},
531: \begin{equation}
532: \rho_{s}(\omega,\mathbf{k}) = -\,2\pi\, \frac{Q\,k_{0}^2}{k^2 + k_{0}^2}\,\delta(\omega)\, ,
533: \end{equation}
534: which gives a screened potential for $k_0^2 > 0$ and recovers the metallic case when the cut-off of
535: the in-plane band dispersion $s$ (of the order of inverse of the
536: in-plane lattice spacing) is taken to infinity.
537:
538: For a finite cut-off $s$ we have various cases. When $k_0 > s$,
539: the screening is limited to the direction perpendicular to the planes.
540: When $k_0^2 < 0$ and $|k_0| < s$, the system does not show any screening
541: since the potential decays like $\cos(|k_0| r)/r$ for large $r$.
542: The last possible case, $k_0^2 < 0$ and $|k_0| > s$ is not
543: physical in this theory and gives a purely imaginary response.
544: We verify that the phase of the superconducting
545: order parameter is free of fluctuations and satisfies the zero-field vortex equation
546: $(\vec{\nabla})^2 \varphi = 0$ in the whole $\Lambda \neq 0$ class.
547:
548: The collective modes follow from a slightly different calculation.
549: We start from the plasmon solution of (\ref{rhoLD0}),
550: \begin{equation}
551: \rho_{s}(\omega,\mathbf{k}) = \rho_0(\mathbf{k}) \,\delta(\omega - \omega_p(\mathbf{k}))\,,
552: \end{equation}
553: where $\rho_0(\mathbf{k})$ is a function which depends not only
554: on the boundary conditions but also on the initial conditions driven by the perturbation.
555: In the non-relativistic
556: limit $v_F/c \rightarrow 0 $, the field solutions in the momentum space are
557: (the details of the calculation are given in the Appendix):
558: \begin{equation}
559: \rho_{s}(\omega,\mathbf{k}) = \rho(\hat{q}) \,\delta(q - q_{0})\,\delta(k_{z})
560: \,\delta(\omega - \omega_p(\mathbf{k})) \, ,
561: \label{plasmon}\qquad
562: \end{equation}
563: \begin{equation}
564: \phi(\omega,\mathbf{k}) = 4\pi\, \frac{\rho_{s}(\omega,\mathbf{k})}{k^2}
565: \, ,
566: \end{equation}
567: \begin{equation}
568: \varphi(\omega, \mathbf{k}) = 4\pi i \,\frac{\omega\,e^\star}{\hbar}
569: \frac{\rho_s(\omega,\mathbf{k})}
570: {k^2} \frac{1}{\omega^2 - v_s^2 \tilde{k}^2} \, ,
571: \label{VARPHIN}
572: \end{equation}
573: \begin{equation}
574: \vec{X}(\omega, \mathbf{k})
575: = 4\pi \frac{\delta}{i\,\kappa} \frac{\vec{q}}{k^2}
576: \frac{\rho_{s}(\omega,\mathbf{k})}{(\omega^2 - \omega_{ph}^2)}\,\, ,
577: \end{equation}
578: \begin{equation}
579: \mathbf{A}(\omega, \mathbf{k}) = 4\pi\, \frac{\omega}{c} \vec{q}\,
580: \frac{\rho_s(\omega,\mathbf{k})}{k^4}\,
581: \mathcal{D}(\omega, \mathbf{k}) \,,
582: \label{a}
583: \end{equation}
584: where we have labeled
585: \begin{equation}
586: \mathcal{D}(\omega, \mathbf{k}) = 1 - 4\pi \frac{e^{\star\,2}\Psi_{0}^2}{m^\star}
587: \frac{1}{\omega^2 - v_s^2 \tilde{k}^2} + 4\pi \frac{\delta^2}{\kappa}\frac{1}
588: {\omega^2 - v_{ph}^2 q^2} \nonumber \,.
589: \end{equation}
590: Equation (\ref{plasmon}) represents the bulk plasmon mode $k_z = 0$, which prevails
591: over the rest of the plasmon band. This is in agreement with a general result valid for a
592: stack of layers coupled by Coulomb interactions and by coherent hopping terms between adjacent planes.
593: The presence of inter-layer charge transfer induces
594: a dimensional crossover to a 3D system, which dominates the long wavelength spectrum.
595: As we show in the end of the Appendix, this effect is conditioned to the assumption that $\tau \neq 0$.
596: The inverse of the in-plane modulation scale $q_0$ is given by the equation
597: $\mathcal{D}(\omega_p,q_0) = 0$.
598:
599: Noting that at $T = 0$ the electronic density $n$ is twice the density of Cooper pairs $\Psi_0^2$, we realize that
600: the quantity $4\pi e^{\star\,2}\Psi_0^2/m^\star$ can be conveniently written in the form of the expression that
601: gives the square of the plasma frequency, $\Omega_p^2 = 4\pi\,e^2 n/m $.
602: In general grounds, $n$ is given by the sum rule \cite{tinkham}
603: \begin{equation}
604: \omega_p^2(k = 0) =
605: \frac{2}{\pi}\int_0^\infty \mathrm{d}\omega\, \omega\,\mathrm{Im} \,\epsilon(\omega,k) =
606: 4\pi\, \frac{e^2 n}{m} f(\bar{q}/q)\label{SumRule}\,.
607: \end{equation}
608: where $\mathrm{Im}\,\epsilon$ is the imaginary part of the electronic susceptibility, which is given in RPA by
609: $\mathrm{Im}\,\epsilon(\omega,\mathbf{k}) = - V_0(q,k_z) \mathrm{Im}\,\Pi^0(\omega,\vec{q},\mu)$ and
610: $f(\bar{q}/q)$ is the anisotropy function due to the shape of the Fermi surface, with $f(\bar{q}/q) \equiv 1$
611: in the isotropic case.
612: For an anisotropic nodal liquid with
613: a small pocket we find that $f(\bar{q}/q) = \bar{q}^2/q^2$, as it may be easily checked by replacing the
614: leading intra-band polarization function
615: \cite{Shung}
616: $$
617: \mathrm{Im}\, \Pi^{0}(\omega,\vec{q},\mu) \stackrel{q/k_F^* \ll 1}{\longrightarrow} - \frac{k_F^*}{\pi \hbar v_\Delta d}
618: \frac{\omega}{\sqrt{v_F^2\bar{q}^2 - \omega^2}}
619: $$
620: into (\ref{SumRule}), giving
621: $$
622: 4\pi \frac{e^{\star\,2}\Psi_{0}^2}{m^{\star}} = \omega_0^2 \,,
623: $$
624: with $\omega_0 = \sqrt{2k_F^* v_F^2 e^2/( d \epsilon_0^{\prime} \hbar v_\Delta)} $, where
625: $\epsilon_0^\prime $ is the background susceptibility renormalized
626: by the small inter-band terms \cite{Shung}. To simplify the analysis we consider
627: the isotropic case only, where the rigid $\Psi_0$ approximation is more rigorous.
628: Using the $k_z = 0$ bulk plasmon mode $\omega_p^2 = \omega_0^2 + (v_0 q)^2$, we find that
629: \begin{equation}
630: q_0 = \sqrt{ b - \sigma(b) \sqrt{b^2 - \frac{g \omega_0^2 }{(v_s^2 - v_0^2)
631: (v_{ph}^2 -v_0^2 )}}}\,,\label{K0}
632: \end{equation}
633: where $\sigma$ is a sign function, $b = (g + \omega_0^2)/[2(v_{ph}^2 - v_0^2)] $ and
634: \begin{eqnarray}
635: g = 4\pi \, \frac{\delta^2}{\kappa}
636: \label{geff}
637: \end{eqnarray}
638: is the effective piezoelectric coupling.
639:
640: Noting that $v_s > v_{ph}$, because the ions are much slower than the electrons, we identify two distinct cases,
641: (1) $v_0 > v_s$ and (2) $v_0 < v_s$. The normal modes
642: are clearly absent in the regime (1) for any value of $g$. On the other hand, these modes are possible
643: in the regime (2) for non zero $g$.
644: Naturally, they would not be physical for $g < (v_{ph}^2/v_s^2) \omega_0^2$,
645: where the interactions are screened.
646:
647: The clean limit of the superconductor gives exactly
648: $v_s = v_F/\sqrt{3}$. In the particular case of an isotropic nodal liquid with
649: a small pocket, the intra-band bulk plasmon disperses with $v_0 = (\sqrt{3}/2) v_F$ \cite{Shung}
650: and therefore
651: it corresponds to the regime (1) mentioned above. This is also the case of the metal
652: ($v_0 = \sqrt{3/5} v_F$),
653: and of doped graphite ($v_0 = (\sqrt{3}/2) v_F$).
654: Therefore, in the particular case of TMD,
655: we should not observe
656: any essential differences in the plasmon modes in comparison to the LEG (where $g = 0$) because of the piezoelectricity.
657:
658:
659: \subsection{Experimental results}
660:
661: There is a large
662: amount of experimental literature dealing with the observation of
663: commensurate (CDW) or incommensurate charge order in TMD. Neutron and x-ray diffraction
664: measurements in TaSe$_2$ and NbSe$_2$ reveal the existence of Bragg peaks at
665: incommensurate wave vectors $\mathbf{Q}_i = (1 - \delta_i)\mathbf{b}_i/3$,
666: where $\mathbf{b}_i$
667: ($|\mathbf{b}_i| = 4\pi / (\sqrt{3}a)$) are the three reciprocal vectors with hexagonal symmetry, $a$
668: is the lattice spacing
669: and $\delta_i \lesssim 0.02$ is the incommensurability
670: \cite{neutrons,stripes}. This state is called a triple CDW phase.
671:
672: In TaSe$_2$, the phase diagram temperature vs. pressure, $P$, is very rich
673: \cite{McWhan} with three different phases: (1) a high temperature hexagonal incommensurate
674: CDW phase (HCDW) where the three ordering vectors have $\delta_i \neq 0$
675: ($i=1,2,3$); (2) an
676: incommensurate stripe phase where one ordering vector is incommensurate, say
677: $\delta_1 \neq0$, but the other two are commensurate, $\delta_{2,3} = 0$;
678: (3) a commensurate CDW phase (CCDW) where $\delta_i = 0$ ($i=1,2,3$).
679: The transition from the undistorted phase (normal) to HCDW occurs at a
680: pressure independent temperature,
681: $T_{N-H} \approx 120 K$, up to pressures of $4.5$ GPa; the transition between
682: HCDW to stripe phase $T_{H-S} \approx 110 K$ is also roughly pressure independent;
683: the transition between stripe phase and CCDW, $T_{S-C}(P)$, is highly pressure dependent
684: and vanishes at $P=P_c \approx 1.8$ GPa. Thus, there is a quantum phase
685: transition ($T=0$), as a function of pressure at $P=P_c$. By further
686: application of pressure there is a reentrant CCDW phase that has not been
687: fully studied and will be not discussed here. We will concentrate on the
688: nature of the quantum critical point (QCP) at $P=P_c$.
689:
690: The nature of the stripe phase has been discussed by McMillan \cite{mcmillan}
691: and others \cite{perbak} from the phenomenological point of view as a result
692: of the formation of topological defects of the complex CDW order parameter.
693: This stripe phase can be easily observed with electron microscopy \cite{Chen}.
694: The microscopic nature of this phase is still unknown but our results here
695: suggest that it may have its origins on the piezoelectric coupling in
696: the solid. Indeed, we have found that there is a charge modulation in
697: these materials when $k_0^2<0$ (see eq.(26)). This modulation can be thought
698: as an incommensuration $\delta_i = 3 \sqrt{-k_0^2}/|{\bf b}_i|$.
699: Equivalently, using definition
700: (\ref{geff}) there is a critical coupling constant $g_c$ so that an extra charge
701: modulation appears ($\delta_i >0$) when $k_0^2<0$ or
702: \begin{eqnarray}
703: g > g_c = 4\pi\frac{v^2_{ph}}{v_s^2} \frac{e^{\star\,2}\Psi_{0}^2}{m^\star} \, .
704: \end{eqnarray}
705: Recall that TaSe$_2$ ($T_c \approx 0.1 K$) and NbSe$_2$ ($T_c
706: \approx 8.3 K$) are both superconductors \cite{Withers} whose $T_c$ increase
707: under application of pressure \cite{jap} and therefore $\Psi_0 \neq 0$ at
708: $T=0$. Notice that both the piezoelectric coupling, $\delta$, as well as
709: the sound velocity, $v_{ph}$, are also monotonically increasing functions
710: of pressure. If we assume that in TaSe$_2$ we have $g<g_c$
711: at $T=0$ and ambient pressure, then the
712: quantum phase transition can occur as a function of pressure as long as
713: $g/v_{ph}^2$ is an increasing function of pressure. In NbSe$_2$, however,
714: the system is always incommensurate indicating that $g>g_c$ even at
715: ambient pressure.
716: Since the $T=0$ phases seem to be directly connected with the stripe
717: phases that are observed at finite temperature we can immediately conclude
718: that the existence of these stripe phases have to do with the
719: piezoelectric coupling in these materials. Besides, recent experiments
720: report a significant electrostatic modulation of $T_c$ in epitaxial bilayers composed by a
721: HT$_c$ cuprate and a polarized insulator deposited on it \cite{Matthey}.
722: The connections
723: between TMD and HT$_c$ are remarkable and we believe
724: that this experimental result could be numerically calculated
725: by relaxing the rigidity in the amplitude $\Psi_0$ of the superconductor order parameter. In addition, we
726: suggest that similar electrostatic devices combined with neutron scattering could test experimentally
727: the role of the piezoelectricity in TMD stripe phases.
728:
729:
730: Neutron scattering measurements have shown the softening of the
731: phonon optical mode
732: $\Sigma_1$ in NbSe$_2$ and TaSe$_2$ at the $\mathbf{Q}_i$ position for a wide range of temperatures
733: \cite{neutrons}.
734: This behavior was interpreted as a coupling of the optical phonons with the charge order.
735: Constant-$\mathbf{Q}$ scans at $\mathbf{Q}_i$ have shown that the phonon energy gap in both cases
736: is of the order of 10 meV
737: \cite{neutrons}.
738: The coupling of the plasmons with the lattice vibrations
739: in the present theory drives the long wavelength phonons to lock their frequency
740: with the frequency of the $k_z = 0$ bulk plasmon.
741: In comparison to the metallic case where the plasmon
742: modes are rigid (in the sense that their wavelength has exactly the size of the system),
743: we have found that the increase of the elastic energy due to the piezoelectric
744: coupling may give rise to {\it elastic} plasmons which oscillate in resonance with the optical phonon
745: mode. Despite {\it not} observable in TMD, we believe that this effect would be the
746: macroscopic manifestation of the plasmon-phonon resonance observed
747: experimentally in these materials.
748:
749:
750: \section{Conclusions}
751:
752: We have investigated the collective excitations of the electrons in a layered superconductor
753: by a semi-classical calculation in the continuum limit of a highly anisotropic material.
754: This procedure is analogous to the Lawrence-Doniach effective model for an
755: infinite stack of layers.
756: Despite the evident interest in TMD, the calculation is sufficiently general and be could be applied to
757: any superconducting layered compound at zero temperature with broken lattice
758: inversion symmetry.
759:
760: We have demonstrated from the
761: electrodynamic point of view that superconductivity and piezoelectricity can coexist.
762: Metallic screening is observed when the effective piezoelectric coupling $g$
763: is smaller than a critical value $g_c$, with the Thomas-Fermi momentum reduced by the increase
764: of $g$. Above the critical coupling, the system is not screened and a long-range charge modulation is expected
765: to appear as the response to
766: a local quasi-static charge unbalance, which can be created by squeezing the crystal.
767: We have shown that piezoelectricity is possibly related to the mechanism behind the stripe formation in TMD.
768: Besides, we have investigated the existence
769: of zero temperature
770: normal modes arising in the presence of
771: low energy bulk plasmons, which dominate the spectrum of collective excitations. We
772: have also conjectured that the $k_z = 0$ plasmon mode behind these excitations
773: comes from the contribution of the intra-band excitations of a pocket opened around the nodes
774: of TMD by low energy coherent hopping terms between adjacent planes.
775: These pockets could be also generated by doping TMD with intercalating materials.
776:
777:
778: \begin{acknowledgments}
779: A.~H.~C.~N. thanks C.~Varma for stimulating this work. We thank A.~J.~Leggett
780: for many discussions on the problem of screening in metals and superconductors.
781: B.U. and G.G.C. thank E. Miranda and Y. Copelevich for many helpfull discussions. They
782: also acknowledge Funda\c{c}\~ao de Amparo \`a Pesquisa do
783: Estado de S\~ao Paulo (FAPESP), Brazil, project number 00/06881-9, for financial support.
784: B.U. thanks the Dept. of Physics at Boston University for the hospitality.
785: \end{acknowledgments}
786:
787: \appendix*
788: \section{}
789:
790: Here we derive in detail the calculation of equations (\ref{plasmon})$-$(\ref{a}). Starting
791: from the plasmon-like solution
792: \begin{equation}
793: \rho_{s}(\omega,\mathbf{k}) = \rho_{0}(\mathbf{k})\,\delta(\omega - \omega_p(\mathbf{k}))\,,
794: \end{equation}
795: for a general function $\rho_{0}(\mathbf{k})$ and replacing it in the
796: the Poisson equation (\ref{phi1}), we get
797: \begin{equation}
798: \phi(\omega,\mathbf{k}) = 4\pi \frac{\rho_{0}(\mathbf{k})}{k^2}\,
799: \delta(\omega - \omega_p(\mathbf{k}))\,. \label{phiNM}
800: \end{equation}
801:
802: Next we separate the $z$-component of (\ref{A1}) and (\ref{jLD}),
803: \begin{eqnarray}
804: && \frac{\omega}{c}
805: k_{z}\,\phi(\omega,\mathbf{k}) + \left(\frac{\omega^2}{c^2} -
806: k^2
807: - \frac{4\pi}{c^2}\,\tau\,\frac{e^{\star\,2}\Psi_{0}^2}{m^\star} \right)
808: A_{z}(\omega, \mathbf{k}) \nonumber \\
809: && \qquad + \frac{4\pi}{c}\frac{e^\star\Psi_{0}^2}{m^\star} i \, \tau\,k_{z} \, \hbar\,
810: \varphi(\omega,\mathbf{k}) = 0\,
811: ,\nonumber
812: \end{eqnarray}
813: and the $\varphi$ equation (\ref{VarphiAnys})
814: \begin{eqnarray}
815: &&\left(- \tilde{k}^2 + \frac{\omega^2}{v_{s}^2} \right)\varphi(\omega, \mathbf{k})\nonumber\\
816: && \qquad
817: - i\frac{e^\star}{\hbar\,c} \left(\frac{c\,\omega}{v_{s}^2}\phi(\omega, \mathbf{k}) +
818: \tilde{\mathbf{k}} \cdot \tilde{\mathbf{A}}(\omega,\mathbf{k}) \right) = 0\nonumber \, .
819: \end{eqnarray}
820: Noting that $\tilde{\mathbf{k}} \cdot \tilde{\mathbf{A}} = k_{z}(\tau^2 - 1)A_{z} $,
821: for non-identically zero $\rho_{0}(\mathbf{k})$ we find
822: \begin{eqnarray}
823: &&\!\!\!\!\!\!\!\!\!\! A_{z}(\omega,\mathbf{k}) = -\frac{4 \pi}{c} \,\frac{\omega\,k_{z}}{k^2}\,
824: \frac{\rho_{0}(\mathbf{k})}{\mathcal{F}(\omega,\mathbf{k} )}\,\delta(\omega - \omega_p(\mathbf{k}))
825: \nonumber \\
826: &&\!\!\!\!\!\! \times \left[ 1
827: - 4\pi \, \frac{e^{\star\,2}\Psi_{0}^2}{m^\star v_{s}^2 }
828: \frac{\tau\,}{\mathcal{G}(\omega,\mathbf{k})}
829: \left(
830: 1 - \frac{v_{s}^2}{c^2} \frac{(\tau^2 - 1)k_{z}^2}{\mathcal{F}(\omega,\mathbf{k})}\right)
831: \right] \,. \label{AZ3}
832: \end{eqnarray}
833: and
834: \begin{eqnarray}
835: \!\!\!\!\!\!\!\!\!\!\varphi(\omega, \mathbf{k}) &=& 4\pi i \,\frac{\omega\,e^\star}{\hbar\,v_{s}^2}
836: \frac{\rho_{0}(\mathbf{k})}
837: {k^2} \frac{1}{\mathcal{G}(\omega, \mathbf{k})}\, \delta(\omega - \omega_p(\mathbf{k})) \nonumber\\
838: && \times \left[1 - (\tau^2 - 1)
839: \frac{v_{s}^2}{c^2} \frac{k_{z}^2}{\mathcal{F}(\omega,\mathbf{k})}\right]
840: \,, \label{VARPHINN}
841: \end{eqnarray}
842: where we have defined:
843: \begin{eqnarray}
844: \mathcal{F}(\omega,\mathbf{k}) &=& \frac{\omega^2}{c^2} - k^2
845: - \frac{4\pi}{c^2}\,\tau\,\frac{e^{\star\,2}\Psi_{0}^2}{m^\star}\nonumber
846: \end{eqnarray}
847: and
848: \begin{eqnarray}
849: \mathcal{G}(\omega, \mathbf{k}) &=& \frac{\omega^2}{v_{s}^2} - \tilde{k}^2 - 4\pi\,
850: \frac{e^{\star\,2}\Psi_{0}^2}{m^\star c^2}
851: \frac{ k_{z}^2 }{\mathcal{F}(\omega,\mathbf{k})}
852: \,\tau \,(\tau^2 - 1)\, .\nonumber
853: \end{eqnarray}
854:
855: The remaining results are derived from the phonon equation (\ref{xAnys})
856: $$
857: \vec{X}(\omega, \mathbf{k}) = - i\,
858: \frac{\delta}{\kappa}\frac{1}{\omega^2 - \omega_{ph}^2}
859: \left(\vec{q} \,
860: \phi(\omega, \mathbf{k}) + \frac{\omega}{c}
861: \vec{A}(\omega, \mathbf{k}) \right)
862: $$
863: and from the combination of the in-plane components of (\ref{A1}) and (\ref{jLD}),
864: \begin{eqnarray}
865: && \left(\frac{\omega^2}{c^2} -
866: k^2 -
867: \frac{4\pi}{c^2}\frac{e^{\star\,2}\Psi_{0}^2}{m^\star}\right)
868: \vec{A}(\omega, \mathbf{k}) +
869: \frac{\omega}{c}
870: \vec{q}\,\phi(\omega,\mathbf{k})\qquad
871: \nonumber\\
872: && \qquad
873: + \frac{4\pi}{c}\frac{e^\star\Psi_{0}^2}{m^\star} i \,\vec{q}\,\hbar\,
874: \varphi(\omega,\mathbf{k}) + \frac{4\pi}{c}i\,\omega\,\delta\,
875: \vec{X}(\omega,\mathbf{k}) = 0 \,. \nonumber
876: \end{eqnarray}
877: After a straightforward calculation, we encounter that
878: \begin{eqnarray}
879: \vec{A}(\omega, \mathbf{k}) &=& - \frac{4\pi}{c}\omega \vec{q}\,
880: \frac{\rho_{0}(\mathbf{k})}{k^2}\,
881: \frac{\mathcal{D}(\omega, \mathbf{k})}{\mathcal{E}(\omega, \mathbf{k})}\,
882: \delta(\omega - \omega_p(\mathbf{k}))\nonumber\\
883: &&\label{AK}
884: \end{eqnarray}
885: \begin{eqnarray}
886: \vec{X}(\omega, \mathbf{k})
887: &=& - 4\pi i \frac{\delta}{\kappa} \frac{\vec{q}}{k^2}
888: \frac{\rho_{0}(\mathbf{k})}{(\omega^2 - \omega_{ph}^2)}\, \delta(\omega - \omega_p(\mathbf{k}))
889: \nonumber \\
890: && \qquad \times \left[
891: 1 - \frac{\omega^2}{c^2} \frac{\mathcal{D}(\omega, \mathbf{k})}
892: {\mathcal{E}(\omega, \mathbf{k})}\right] \label{XA}\, .
893: \end{eqnarray}
894: where we have labeled
895: \begin{eqnarray}
896: \mathcal{D}(\omega, \mathbf{k}) &=& 1 + 4\pi \frac{\delta^2}{\kappa}\frac{1}
897: {\omega^2 - \omega_{ph}^2} - 4\pi \frac{e^{\star\,2}\Psi_{0}^2}{m^\star v_{s}^2}
898: \frac{1}{\mathcal{G}(\omega, \mathbf{k})}\nonumber \\
899: && \qquad \times \left[1 - (\tau^2 - 1)\,\frac{v_{s}^2}{c^2}
900: \frac{k_{z}^2}{\mathcal{F}(\omega, \mathbf{k})} \right] \nonumber
901: \end{eqnarray}
902: \begin{equation}
903: \mathcal{E}(\omega, \mathbf{k}) = \frac{\omega^2}{c^2} - k^2
904: - \frac{4\pi}{c^2}\frac{e^{\star\,2}\Psi_{0}^2}{m^\star}\nonumber
905: +
906: \frac{4\pi}{c^2}\frac{(\delta\,\omega)^2}{\kappa\,(\omega^2 - \omega_{ph}^2)} \label{E}\, .
907: \end{equation}
908:
909: Next, we substitute these results in the superconducting charge density definition (\ref{charge})
910: \begin{eqnarray}
911: &&\rho_{0}(\omega,\mathbf{k}) + \frac{e^\star \Psi_{0}^2}{m^\star v_{s}^2}
912: \left[\hbar\,i\,\omega\,\varphi(\omega,\mathbf{k}) + e^\star
913: \phi(\omega,\mathbf{k}) \right]\nonumber\\
914: &&\qquad\qquad + i\,\delta\,\vec{q}\cdot \vec{X}(\omega, \mathbf{k}) = 0\,,
915: \nonumber
916: \end{eqnarray}
917: what yields
918: \begin{eqnarray}
919: && \rho_{0}(\mathbf{k})\,\delta(\omega - \omega_p(\mathbf{k})) \left\{ 1 -
920: \frac{4\pi}{k^2} \frac{e^{\star\,2} \Psi_{0}^2}{m^\star v_{s}^2 }
921: \left[ \frac{\omega^2}{v_{s}^2}\,\frac{1}{ \mathcal{G}(\omega,\mathbf{k})}\,
922: \nonumber \right.\right. \\
923: && \qquad \times\left. \left(1 - (\tau^2 - 1)\frac{v_{s}^2}{c^2}
924: \frac{k_{z}^2}{\mathcal{F}(\omega,\mathbf{k})} \right)- 1\right]
925: \nonumber \\
926: && \qquad + \, \left. \, 4\pi \frac{\delta^2 }{\kappa} \frac{1}{k^2}
927: \frac{q^2}{(\omega^2 - \omega_{ph}^2)} \left[
928: 1 - \frac{\omega^2}{c^2} \frac{\mathcal{D}(\omega, \mathbf{k})}{\mathcal{E}(\omega,
929: \mathbf{k})}\right] \right\} \nonumber\\
930: && = 0 \label{Eq2}
931: \end{eqnarray}
932:
933: Applying the transverse gauge to (\ref{AZ3}) and (\ref{AK}),
934: \begin{eqnarray}
935: \mathbf{k}\cdot \mathbf{A}
936: &=& - \frac{4 \pi}{c}\frac{\omega}{k^2}\, \rho_{0}(\mathbf{k})\,\delta(\omega - \omega_p(\mathbf{k}))
937: \nonumber \\
938: && \times \left\{\frac{k_{z}^2}{
939: \mathcal{F}(\omega,\mathbf{k})}\left[ 1 - 4\pi \frac{e^{\star\,2} \Psi^2}{m^\star v_{s}^2}
940: \frac{\tau}{\mathcal{G(\omega,\mathbf{k})}} \right.\right. \nonumber\\
941: && \left.\left. \times \left(
942: 1 - (\tau^2 - 1) \frac{v_{s}^2}{c^2} \frac{k_{z}^2}{\mathcal{F}(\omega,\mathbf{k})}\right)
943: \right] + q^2 \frac{ \mathcal{D}(\omega,\mathbf{k})}
944: {\mathcal{E}(\omega,\mathbf{k})} \right\} \nonumber \\
945: &=& 0 \label{GAUGE}
946: \end{eqnarray}
947:
948: This way, we conclude that $\rho_{0}(\mathbf{k})$ is on the form
949: $$
950: \rho_{0}(\mathbf{k}) = \rho(\hat{q}) \,\delta(q - q_{0})\,\delta(k_{z} - k_{z\,0}) \, ,
951: $$
952: where $q_{0}$ and $k_{z\,0}$ are the zeroes
953: of (\ref{GAUGE}) and (\ref{Eq2}).
954: In the non-relativistic limit $v_F/c \rightarrow 0$,
955: $\mathcal{F}(\omega_p, \mathbf{k}) \sim
956: \mathcal{E}(\omega_p, \mathbf{k}) \rightarrow - k^2$ and
957: \begin{equation}
958: \mathcal{G}(\omega, \mathbf{k}) \rightarrow \frac{\omega^2}{v_{s}^2} - \tilde{k}^2
959: \nonumber
960: \end{equation}
961: \begin{equation}
962: \mathcal{D}(\omega, \mathbf{k}) \rightarrow 1 - 4\pi \frac{e^{\star\,2} \Psi_{0}^2}{m^\star}
963: \frac{1}{ \omega^2 - v_{s}^2 \tilde{k}^2 } + 4\pi \frac{\delta^2}{\kappa}\frac{1}
964: {\omega^2 - \omega_{ph}^2} \,.\nonumber
965: \end{equation}
966: In this limit, (\ref{GAUGE}) and (\ref{Eq2}) simplify respectively to
967: \begin{eqnarray}
968: && \left( k^2 + \, 4\pi \frac{\delta^2}{\kappa}
969: \frac{ q^2}{\omega^2 - \omega_{ph}^2} -
970: 4\pi \frac{e^{\star\,2} \Psi_{0}^2}{m^\star }
971: \frac{q^2 + \tau\,k_{z}^2 }{\omega^2 - v_{s}^2\tilde{k}^2}
972: \right)\nonumber \\
973: && \qquad \times
974: \rho_s(\omega,\mathbf{k})\,\delta(\omega - \omega_p(\mathbf{k})) = 0
975: \end{eqnarray}
976: \begin{eqnarray}
977: &&\left( k^2 + \, 4\pi \frac{\delta^2}{\kappa}
978: \frac{q^2}{\omega^2 - \omega_{ph}^2} -
979: 4\pi \frac{e^{\star\,2} \Psi_{0}^2}{m^\star }
980: \frac{\tilde{k}^2}{\omega^2 - v_{s}^2\tilde{k}^2}
981: \right)\nonumber \\
982: && \qquad \times \rho_s(\omega,\mathbf{k})\delta(\omega - \omega_p(\mathbf{k})) = 0\, .
983: \end{eqnarray}
984: Comparing both and recalling that $\tilde{k}^2 = q^2 + \tau^2 k_{z}^2$, then
985: $ \tau\, (1 - \tau) \,k_{z}^2 = 0$. For $\tau \neq 0,1$, we immediately see that $k_{z\,0} = 0$.
986: Substituting this result in one of the expressions above and integrating in $\omega$, we find
987: that $\mathcal{D}(\omega_p, q_0) = 0$.
988:
989: \begin{thebibliography}{99}
990:
991: \bibitem{Withers}R. L. Withers, and J. A. Wilson, J. Phys. C: Solid State Phys. {\bf 19}, 4809 (1986).
992:
993: \bibitem{Wilson}J. A. Wilson, and A. D. Yoffe, Adv. ~Phys. {\bf 18}, 193
994: (1969); J.~A.~Wilson, F. J. DiSalvo and S. Mahajan, Adv.~Phys. {\bf 24},117 (1975).
995:
996: \bibitem{transport} J. P. Tidman, O. Singh and A. E. Curzon, Philos. Mag. {\bf 30}, 1191 (1974).
997:
998: \bibitem{hall}D. A. Whitney, R. M. Fleming, and R. V. Coleman, Phys. Rev. B {\bf 15}, 3405 (1977).
999:
1000: \bibitem{stripes}R. M. Fleming, D. E. Moncton, D. B. McWhan, and F. J. DiSalvo, Phys. Rev. Lett. {\bf 45}, 576 (1980).
1001:
1002: \bibitem{valla}T. Valla, A. V. Fedorov, P. D. Johnson, J. Xue, K. E. Smith,
1003: and F. J. DiSalvo, Phys. Rev. Lett. {\bf 85}, 4759 (2000).
1004:
1005: \bibitem{valla_htc} T. Valla, A. V. Fedorov, P. D. Johnson, B. O. Wells, S. L. Hulbert, Q. Li, G. D. Gu and N Koshizuka, Science {\bf 285}, 2110 (1999).
1006:
1007: \bibitem{Neto}A. H. Castro Neto, Phys. Rev. Lett. {\bf 86}, 4382 (2001).
1008:
1009: \bibitem{neutrons}D. E. Moncton, J. D. Axe, and F. J. DiSalvo, Phys. Rev. B {\bf 16}, 801 (1977).
1010:
1011: \bibitem{piezo}G. D. Mahan, in {\it Polarons in Ionic Crystals and Polar Semiconductors},
1012: edited by J.~T.~Devreese (North-Holland, Amsterdam, 1971), p.553.
1013:
1014: \bibitem{varma}C. Varma, private communication.
1015:
1016: \bibitem{mele}D. P. DiVincenzo, and E. J. Mele, Phys. Rev. ~B {\bf 29}, 1685 (1984).
1017:
1018: \bibitem{fetter}A. L. Fetter, and J. D. Walecka, {\it Quantum Theory of Many-Particle Systems}
1019: (McGraw-Hill, New York, 1971).
1020:
1021: \bibitem{Shung}K. W. -K. Shung, Phys. ~Rev. ~B {\bf 34} 979 (1986).
1022:
1023: \bibitem{tinkham} M. Tinkham, {\it Introduction to Superconductivity} (McGraw-Hill, New York, 1996).
1024:
1025: \bibitem{balents}L. Balents, M. P. A. Fisher, and C. Nayak, Int. Jou. Mod. Phys. B, {\bf 12}, 1033 (1998).
1026:
1027: \bibitem{Vozmediano}J. Gonzalez, F. Guinea, and M. A. H. Vozmediano, Phys. Rev. Lett. {\bf 77}, 3589 (1996).
1028:
1029: \bibitem{Hawrylak}P. Hawrylak, G. Eliasson, and J. J. Quinn, Phys. Rev. B {\bf 37}, 10187 (1988).
1030:
1031: \bibitem{hopping}J. Gonz\'alez, F. Guinea, and M. A. H. Vozmediano, Phys. ~Rev. ~B {\bf 63}, 134421 (2001).
1032:
1033: \bibitem{Doran} N. J. Doran, B. Ricco, D. J. Titterington and G. Wexler, J. Phys. C: Solid State Phys. {\bf 11}, 685 (1978).
1034:
1035: \bibitem{jackson}J. D. Jackson, {\it Classical Electrodynamics} (John Wiley
1036: \& Sons, New York, 1975).
1037:
1038: \bibitem{bohm_pines}D. Pines, {\it Elementary Excitations in Solids}
1039: (Addison-Wesley, Redwood City, 1963).
1040:
1041: \bibitem{Leggett}A. J. Leggett, Phys.Rev. {\bf 140 }, A1869 (1965).
1042:
1043: \bibitem{unpublished}B. Uchoa, A. H. Castro Neto and G. G. Cabrera (unpublished).
1044:
1045: \bibitem{Bardeen}J. Bardeen, Phys. Rev. Lett. {\bf 1}, 399 (1958).
1046:
1047: \bibitem{hub}S. Weinberg, Nucl. Phys. B {\bf 413}, 567 (1994).
1048:
1049: \bibitem{Tsuneto}E. Abrahams, and T. Tsuneto, Phys. Rev. {\bf 152}, 416 (1966).
1050:
1051: \bibitem{Parks} R. D. Parks, {\it Superconductivity} (Marcel
1052: Dekker Inc., New York, 1969) vol. 1, Chap. 6.
1053:
1054: \bibitem{McWhan} D. B. McWhan, J. D. Axe and R. Youngblood, Phys. Rev. B {\bf 24}, 5391 (1981).
1055:
1056: \bibitem{mcmillan}W. L. McMillan, Phys. Rev. B {\bf 14}, 1496 (1976).
1057:
1058: \bibitem{perbak}P. Bak and D. Mukamel, Phys. Rev. B {\bf 19}, 1604 (1979).
1059:
1060: \bibitem{Chen} K. K. Fung, S. McKernan, J. W. Steeds, and J. A. Wilson,
1061: J. Phys. C:Solid State Phys. {\bf 14}, 5417 (1981);
1062: C. H. Chen, J. M. Gibson and R. M. Fleming, Phys. Rev. Lett. {\bf 47}, 723 (1981).
1063:
1064: \bibitem{jap}K. Yamaya, and T. Sambongi, J. Phys. Soc. Japan {\bf 32}, 1150 (1972).
1065:
1066: \bibitem{Matthey} D. Matthey, S. Gariglio and J. -M. Triscone, Phys. Rev. Lett. {\bf 83},
1067: 3758 (2003); C. H. Ahn, S. Gariglio, P. Paruch, T. Tybell, L. Antognazza and J.-M. Triscone,
1068: Science {\bf 284}, 1152 (1999).
1069:
1070: \end{thebibliography}
1071:
1072:
1073:
1074:
1075:
1076:
1077:
1078:
1079: \end{document}
1080: