Fermi arcs from holography
Abstract
In this paper, we find mechanisms for the generation of Fermi arcs using the gauge/gravity correspondence. The gravity background is taken to be a charged black hole with vector hair in asymptotically spacetime. The response function of fermion probes exhibits a pwave gap in the dual superconductor. We couple the fermions to a charged ranktwo antisymmetric field. Assuming that its spatial components condense, a novel type of open Fermi surface is produced. We derive an analytical formula for the Green’s function and study its unique properties. The results are confirmed by separate numerical computations.
In the Appendix, we study the effect of a neutral scalar field on the fermionic spectral functions. A suitable interaction term shifts the original spinup / spindown Fermi momenta in opposite directions and thus the two nodal points of the pwave gap extend into Fermi pockets.
I Introduction
Strongly coupled systems exhibit various interesting phenomena which often cannot be understood using intuition from weakly coupled physics. Among these are many unconventional properties of hightemperature superconductors J. G. Bednorz and K. A. Müller (1986). Understanding these materials remains one of the major challenges in physics.
In the superconducting phase of cuprates, an anisotropic energy gap opens up near the Fermi level. The interesting physics is essentially twodimensional and the order parameter has a dwave symmetry. This means that the gap is approximately a function of the angle in momentum space and thus it vanishes at the four nodal points on the diagonals of the Brillouin zone. The phase transition temperature () depends on the amount of electron or hole doping (). At zero doping, these materials are antiferromagnetic Mott insulators. The superconducting phase covers a domeshaped region in the phase diagram. Optimal doping is achieved when is maximal. Underdoping and overdoping refer to superconductors with doping levels below and above optimal doping, respectively.
Underdoped cuprates exhibit fascinating phenomena including charge stripes, a large Nernst effect, unusual specific heat, spin susceptibility and transport properties. Above , in the metallic pseudogap phase, the density of states near the Fermi energy is partially suppressed and the four nodal “Fermi points” extend into Fermi arcs: gapless excitations in disconnected arcshaped regions of momentum space. The length of the arcs grows with temperature and doping until a large connected Fermi surface is recovered. The existence of open Fermi arcs is rather interesting since in conventional metals Fermi surfaces are closed boundaries separating occupied and unoccupied states. The origin of Fermi arcs is not a settled question, although different models exhibiting similar features have been proposed^{1}^{1}1For angleresolved photoemission spectroscopy (ARPES) and quantum oscillation data, see e.g. Chatterjee et al. (2006); Kanigel et al. (2006); DoironLeyraud et al. (2007); Kanigel et al. (2008) and the reviews Carlson et al. (2002); Campuzano et al. (2002); Damascelli et al. (2003). For theoretical models, see Lee et al. (2004); Norman et al. (2007) and references therein..
Motivated by cuprate phenomenology, we will study anisotropic Fermi surfaces using the gauge/gravity correspondence Maldacena (1998); Gubser et al. (1998); Witten (1998)^{2}^{2}2For recent developments in the topic of holographic Fermi liquids, see Lee (2008); Liu et al. (2009); Cubrovic et al. (2009); Faulkner et al. (2009); Denef et al. (2010); Faulkner and Polchinski (2010). For holographic superconductors, see Gubser (2008); Hartnoll et al. (2008a, b); Horowitz and Roberts (2009); Gubser and Nellore (2009); Faulkner et al. (2010a); Gubser et al. (2009); Chen et al. (2009); Gubser et al. (2010). For further references, see the recent reviews Hartnoll (2009); Herzog (2009); McGreevy (2009); Sachdev (2010); Faulkner et al. (2010b); Horowitz (2010).. The purpose of the paper is to find Fermi arclike phenomena using ingredients which are natural from a holographic point of view.
In Section II, we describe the (3+1)dimensional gravity background. In order to introduce anisotropy in the system, we consider an antide Sitter black hole with a condensed vector field. A nonzero spatial component of the vector field breaks rotational invariance in the boundary theory and gives a pwave holographic superconductor^{3}^{3}3Alternatively, one can substitute the vector field with spintwo (or higher spin) fields as in the recent papers Chen et al. (2010); Benini et al. (2010). In the latter paper, Fermi arcs were due to temperature broadening.. Probe fermions are coupled to the vector field. The pwave gap is seen in their response functions: there are lowenergy excitations near the two nodal points.
In Section III, we couple the fermions to an antisymmetric tensor field background. The fermionic response functions exhibit Fermi arcs. Note that these arcs exist in the superconducting phase. Using approximations whose results are confirmed by numerical computations, we derive an analytical formula for the Green’s function. We show that the antisymmetric coupling enhances temperature broadening in the spectral function. More interestingly, it decreases the distance between the quasiparticle poles on the complex frequency plane. Thus, it provides a heretofore unknown mechanism for generating Fermi surfaces that are not closed.
In the Appendix, we describe a different model where the fermions couple to a neutral scalar field. The coupling behaves as an effective mass term and it shifts the original “spinup” and “spindown” Fermi momenta in different directions. As a result, near the pwave nodal directions the two gapless points extend into “Fermi pockets”.
Ii The setup
ii.1 Anisotropic background
Let us consider a massive vector field in the bulk. This will play the role of the pwave order parameter^{4}^{4}4I thank Michal Heller for collaboration. Gubser and Pufu (2008); Roberts and Hartnoll (2008); Basu et al. (2009); Ammon et al. (2010). We assume that it has charge under the gauge symmetry. The action takes the form,
(1)  
with and .
For the condensed phase, we take the ansatz,
The and functions can be determined numerically from the equations of motion Heller and Vegh () such that the metric describes a static asymptotically spacetime with a charged black hole of Hawking temperature and a condensed vector field hair. The horizon is located at . The metric is anisotropic in the spatial directions.
ii.2 Spinor probes
Let us now introduce two^{5}^{5}5In even dimensions, we could use a single spinor instead and couple it to its charged conjugate as in Faulkner et al. (2010a). probe bulk Dirac fields and with opposite charges, , under the gauge symmetry. The fermionic action takes the form^{6}^{6}6In order to write down physically interesting interaction terms, as an organizing principle, it is useful to think of (1) as the result of a broken “pseudospin” bulk gauge symmetry. The bulk gauge symmetry is interpreted as the unbroken diagonal subgroup and is the pseudospin Wboson. The fields then form a pseudospin doublet.,
where and parametrizes the coupling of the fermions to the pwave order. We are going to use the following basis for gamma matrices,
For convenience, we also give the following matrices,
Since in the above background the spin connection satisfies
we can rescale the Dirac fields and remove the spin connection from the equations. For a given mode, we introduce the notation in the above gamma matrix basis,
(2) 
The fourcomponent spinor has been split into the rescaled and twocomponent spinors.
ii.3 Green’s function
If we restrict the spinor momentum to be in the direction (perpendicular to the condensed field), then the equations decouple into two sets of equations containing and , respectively^{7}^{7}7I thank Hong Liu for pointing out that this can also be done in the general case by applying a change of basis.. Thus, we can consider these variables separately. In the rest of the paper, we will focus on and , suppressing their 1, 2 indices. (, can be treated similarly.) Let us combine these spinors into the fourcomponent NambuGor’kov spinor . The Dirac equation,
Here we used the notation,
The offdiagonal terms are subdominant at the UV and IR boundaries. The two linearly independent solutions with ingoing boundary conditions at the horizon will be denoted by and . At the UV boundary, the two independent solutions for a twocomponent spinor () are,
The Green’s function matrix is given by Henningson and Sfetsos (1998); Mueck and Viswanathan (1998); Iqbal and Liu (2009); Faulkner et al. (2009); Faulkner et al. (2010a),
(3) 
where
is the retarded correlator and denote the fermionic boundary operators dual to the bulk spinor fields.
The condensed vector field induces a mixing of positive and negative frequency modes of the probe spinors. This mixing is maximal when their momenta lie in perpendicular () direction. The eigenvalue repulsion between particles and holes at produces a pwave gap in the fermion spectral function, see FIG. 5 (i).
Iii Fermi arc
In this section, we describe an antisymmetric coupling that can be used to produce a Fermi arc. We leave the issues related to building a full model to future works.
Let us consider an antisymmetric tensor field in the gravity background. We introduce the fermion interaction,
(4) 
For this term to be gauge invariant, the antisymmetric field must have twice the charge of the spinor, i.e. .
In the following, we will assume that the spatial tensor components condense. The qualitative features of the results will not depend on the exact details of the profile. We will not consider the backreaction of on the metric.
In order to obtain (approximate) analytical results, we consider a simplified system with . We let condense. For simplicity, consider spinor momentum near a Fermi surface, along the direction: . We turn on a finite coupling to the condensed field, but treat , and as small perturbations in the Dirac equation. Both the equation of motion and the boundary condition for a charged bosonic probe are real at zero frequency. Thus, for simplicity, we will consider real , in the following.
Let us collect the coupled twocomponent spinors of opposite charges into the modified NambuGor’kov spinor . The Pauli matrix is included in order to make the equations more symmetric: it changes into . The indices on and will be suppressed in the following. In order to simplify the computation, we use a basis of ingoing solutions () having and , respectively, when the and couplings are turned off.
iii.1 Finite Bfield background
If a finite, possibly large, Bfield is turned on, the ingoing wavefunctions change in the following way,
The Dirac equation for the modified Nambu spinor,
(5) 
and similarly for . Tilde will indicate solutions. If the solutions (satisfying ) are known, then (5) can surprisingly be solved by setting
(6) 
where and . We will use the notation, and . At finite temperature, the limit is convergent because has no boundary sources and the horizon provides an IR cutoff. (At zero temperature, may diverge.) Compared to the case, the expectation value matrix in (3) gets multiplied by from the left.
iii.2 Perturbation theory
After turning on a finite , let us perturb the system by a small and . The wavefunctions change,
Here is a small perturbation to the rotated wavefunction. Plug this ansatz into the Dirac equation to get,
(7)  
Let us now integrate the equation using . After integration by parts, the differential operator will act on the integrand and vanish. Thus, the only contribution from will come from the Wronskians computed at the boundary (, vanish at the horizon),
iii.3 Zero temperature
At zero temperature at the Fermi surface, the source component of the spinor vanishes (). Thus,
For simplicity, we will pretend that at . Using (6) and thus , this can
be further written in matrix form as,
Here we have defined to be the matrix without the minus signs. The null subscript refers to zero temperature. Since at we have , the source matrix in (3) is in fact equal to . Integration of the perturbed Dirac equation (7) gives
The RHS comes from the second term in (7). The gap parameter is proportional to , is the Fermi velocity and is a constant. Multiplying from the left by we get,
The source matrix is related to the Green’s function through where is the expectation value matrix. The quasiparticle poles are located where . Since is only rotated compared to the case, the poles remain at the same place in the complex plane.
iii.4 Finite temperature
At small temperatures, the pole in the Green’s function at the Fermi surface is not at , but it is located on the lower half plane. In the spectral function this manifests itself as temperature broadening. The original ( and ) spinor wavefunctions now have a nonzero source component at the boundary. Let us denote it by . We assume that and thus . After integrating the (7) Dirac equation, the Wronskians give
Since and are proportional to , they are much smaller than and . Thus, we can neglect the second line. Similarly to the case, the Wronskians give,
(8) 
To first order, the integrated equation (7) still gives,
(9) 
At finite temperatures, however, the source matrix in (3) will be different from . Let us denote it by ,
(10) 
The second matrix is just . Using (6), the first matrix can be written as,
Thus, (9) and (10) combines into,
Importantly, the perturbations, which are encoded in , are rotated in the opposite direction compared to the width term. This implies that the antisymmetric coupling does alter the finite temperature correlators.
The Green’s function can be computed from where is the diagonal expectation value matrix of the solution, rotated by ,
(11) 
The Green’s function matrix,
(12) 
where . The function may diverge as . The location of the two poles is readily computed,
The Green’s function and the dispersion relation are our central technical results. In the following, we will analyze their properties and show how they can give Fermi arcs.
iii.5 Properties of the Green’s function

At , the familiar BCS formula is recovered,

The limit also gives back the BCS Green’s function unless diverges too quickly.

When is large (with fixed ), the peak is stuck near and disperses very slowly. On the complex plane, the quasiparticle poles move on the negative imaginary axis.

By extracting the imaginary part from (12), one can show that the spectral function is always positive. The poles always stay on the lower half plane.

FIG. 1 shows plots of Im . The left plot shows the BCS case where the various curves have different . At small , temperature broadening kicks in and the maxima of the two peaks coalesce as shown by the orange lines.
The middle plot shows the same figure at finite value of . The peaks at larger become wider and at small taller. The gap now vanishes even at intermediate values of .
In case of a pwave gap, and thus different values of the parameter correspond to different angles in momentum space. The second figure thus shows that there is an extended “Fermi arc” region where the gap vanishes. This arc is longer than what is justified by temperature broadening.
Finally, the third figure shows the dispersion of the peak at large . The gap vanishes and the dispersion is nonlinear, .

FIG. 2 shows the paths of the poles in the Green’s function as the momentum is varied. This figure has been separately confirmed by numerical computations (using a real Bfield profile). For smaller values of , the size of the gap decreases. The effective quasiparticle width is . For large enough , the poles actually collide and then move on the imaginary axis.

When , the spectral function simplifies,
Since only appears through a multiplicative factor, it does not have a significant effect on an twodimensional ARPEStype figure of the “Brillouin zone”. Any visible arcs in such a figure will be similar to arcs caused by temperature broadening (see FIG. 5 (i)).

Even though the (normalized) spectral function does not change at , the gap does vanish in an extended arc region. FIG. 4 shows the gap as a function of the angle, computed numerically. Purple curve shows the pwave gap. The temperature is rather small and its broadening effects cannot be seen. There are, however, larger, visible effects when the Bfield is turned on (blue curve): the gap vanishes for in the Fermi arc region.
Iv Discussion
In this paper, we found Fermi arcs in holographic superconductors using a phenomenological approach. A condensing vector field (which could be substituted with higher spin fields) produces an anisotropic gap in the probe fermion spectral functions. A coupling between the fermions and a charged antisymmetric tensor field reduces the size of the gap. If the original gap at a certain point in momentum space was small enough, then it gets completely eliminated and a nonlinear dispersion is produced. This happens near the nodal points where the pwave gap is small. Thus, the gapless points will become a finite length Fermi surface: a Fermi arc.
An approximate analytical formula (12) for the fermionic Green’s function has been derived and separately confirmed by numerical calculations. We are currently lacking a purely boundary field theory (or “semiholographic” Faulkner and Polchinski (2010)) interpretation of this phenomenon.
Increasing the temperature increases the width of the quasiparticles, but presumably reduces . It would be interesting to construct a full model where the backreaction of the antisymmetric field is also taken into account. This would allow for a study of the arc length as a function of temperature and other parameters.
In an holographic superconductor with spinor doublet , a similar effect may arise from the coupling
where is the field strength. When expanded, this contains a term
with . If there are spatial (amplitude) fluctuations in the order parameter, this term may produce similar effects to the antisymmetric coupling that we introduced in section III.
In the Appendix, we describe another model which deals with “spin order” represented by a neutral scalar field in the bulk. The coupling of this field to the fermions creates a closed Fermi pocket. It would be interesting to build a more realistic holographic model using this idea.
Finally, it is worth emphasizing that in both cases, the system is in the superconducting phase. In real materials, arcs appear in the nonsuperconducting metallic phase. It would be interesting to understand whether longrange order could be eliminated in holographic systems while preserving Fermi arcs.
Acknowledgments
I thank Hong Liu for collaboration, many suggestions and comments on the manuscript. I further thank Tom Faulkner and John McGreevy for many comments, and Michal Heller, Nabil Iqbal and Márk Mezei for collaborations on related projects.
Appendix: Fermi pocket
In this appendix, we demonstrate how a neutral field corresponding to spin (density wave) order can modify the anisotropic Fermi surface of the fermions.
Let us introduce “spin” in the boundary theory by considering two identical bulk Dirac fields, and . Here up and down refer to a new direction which is perpendicular to the 2d superconductor and is not be confused with the radial direction of .
We can arrange the spinors into a spin doublet . Naturally, we only want to consider Lagrangians which are invariant under the spin .
In order to write down physically interesting interaction terms, it is useful to consider another symmetry called the pseudospin. The bulk gauge symmetry may be interpreted as the unbroken diagonal subgroup and the field is the pseudospin Wboson. The pseudospin acts on the doublet where is the charge conjugate spinor. Thus, and have opposite charges, , under the gauge symmetry. These are the spinors that we used in the rest of the paper.
The spin order parameter is a spin triplet Iqbal et al. (2010); Metlitski and Sachdev (2010). Its condensate breaks the spin down to . Since the fermions form a spin doublet , a natural coupling between the fields is
We assume that only condenses. Hence, we can rewrite the interaction as . Here we used the fact that for a Dirac spinor.
The action then takes the form,
Note that the two spinors get contributions to their effective masses with opposite signs. As a result, the degeneracy of the Fermi momenta of the two spinors is resolved: . Originally, the poles of and collided at as the momentum increased. Since the Fermi momenta are different now, one of the peaks will cross if the coupling to is large enough compared to the gap. This results in a Fermi arc.
We emphasize that the arc here is only present if one plots the spin components separately. Since the other spin component has its arc on the other side of the “banana”, the trace will be a full oval, a Fermi pocket.
The shift in the Fermi momenta can be modeled by the following Green’s function,
where are the shifts in the Fermi momenta for the two spinors. Depending on the parameters, this function gives an arc similar to the one in FIG. 5 (ii).
References
 J. G. Bednorz and K. A. Müller (1986) J. G. Bednorz and K. A. Müller, Z. Phys. B64, 189 (1986).
 Chatterjee et al. (2006) U. Chatterjee et al., Physical Review Letters 96, 107006 (2006), eprint arXiv:condmat/0505296.
 Kanigel et al. (2006) A. Kanigel et al., Nature Physics 2, 447 (2006), eprint arXiv:condmat/0605499.
 DoironLeyraud et al. (2007) N. DoironLeyraud et al., Nature (London) 447, 565 (2007), eprint 0801.1281.
 Kanigel et al. (2008) A. Kanigel et al., Physical Review Letters 101, 137002 (2008), eprint 0803.3052.
 Carlson et al. (2002) E. W. Carlson et al. (2002), eprint arXiv:condmat/0206217.
 Campuzano et al. (2002) J. C. Campuzano, M. R. Norman, and M. Randeria (2002), eprint arXiv:condmat/0209476.
 Damascelli et al. (2003) A. Damascelli, Z. Hussain, and Z. Shen, Reviews of Modern Physics 75, 473 (2003), eprint arXiv:condmat/0208504.
 Lee et al. (2004) P. A. Lee et al. (2004), eprint arXiv:condmat/0410445.
 Norman et al. (2007) M. R. Norman et al., Phys. Rev. B 76, 174501 (2007), eprint 0708.1713.
 Maldacena (1998) J. M. Maldacena, Adv. Theor. Math. Phys. 2, 231 (1998).
 Gubser et al. (1998) S. S. Gubser, I. R. Klebanov, and A. M. Polyakov, Phys. Lett. B428, 105 (1998), eprint hepth/9802109.
 Witten (1998) E. Witten, Adv. Theor. Math. Phys. 2, 253 (1998).
 Lee (2008) S.S. Lee (2008), eprint 0809.3402.
 Liu et al. (2009) H. Liu, J. McGreevy, and D. Vegh (2009), eprint 0903.2477.
 Cubrovic et al. (2009) M. Cubrovic, J. Zaanen, and K. Schalm, Science 325, 439 (2009), eprint 0904.1993.
 Faulkner et al. (2009) T. Faulkner, H. Liu, J. McGreevy, and D. Vegh (2009), eprint 0907.2694.
 Denef et al. (2010) F. Denef, S. A. Hartnoll, and S. Sachdev, Class. Quant. Grav. 27, 125001 (2010), eprint 0908.2657.
 Faulkner and Polchinski (2010) T. Faulkner and J. Polchinski (2010), eprint 1001.5049.
 Gubser (2008) S. S. Gubser, Phys. Rev. D78, 065034 (2008), eprint 0801.2977.
 Hartnoll et al. (2008a) S. A. Hartnoll, C. P. Herzog, and G. T. Horowitz, Phys. Rev. Lett. 101, 031601 (2008a), eprint 0803.3295.
 Hartnoll et al. (2008b) S. A. Hartnoll, C. P. Herzog, and G. T. Horowitz, JHEP 12, 015 (2008b), eprint 0810.1563.
 Horowitz and Roberts (2009) G. T. Horowitz and M. M. Roberts, JHEP 11, 015 (2009).
 Gubser and Nellore (2009) S. S. Gubser and A. Nellore, Phys. Rev. D80, 105007 (2009), eprint 0908.1972.
 Faulkner et al. (2010a) T. Faulkner, G. T. Horowitz, J. McGreevy, M. M. Roberts, and D. Vegh, JHEP 03, 121 (2010a), eprint 0911.3402.
 Gubser et al. (2009) S. S. Gubser, F. D. Rocha, and P. Talavera (2009), eprint 0911.3632.
 Chen et al. (2009) J.W. Chen, Y.J. Kao, and W.Y. Wen (2009), eprint 0911.2821.
 Gubser et al. (2010) S. S. Gubser, F. D. Rocha, and A. Yarom (2010), eprint 1002.4416.
 Hartnoll (2009) S. A. Hartnoll, Class. Quant. Grav. 26, 224002 (2009), eprint 0903.3246.
 Herzog (2009) C. P. Herzog, J. Phys. A42, 343001 (2009), eprint 0904.1975.
 McGreevy (2009) J. McGreevy (2009), eprint 0909.0518.
 Sachdev (2010) S. Sachdev (2010), eprint 1002.2947.
 Faulkner et al. (2010b) T. Faulkner, N. Iqbal, H. Liu, J. McGreevy, and D. Vegh (2010b), eprint 1003.1728.
 Horowitz (2010) G. T. Horowitz (2010), eprint 1002.1722.
 Chen et al. (2010) J.W. Chen, Y.J. Kao, D. Maity, W.Y. Wen, and C.P. Yeh, Phys. Rev. D81, 106008 (2010), eprint 1003.2991.
 Benini et al. (2010) F. Benini, C. P. Herzog, and A. Yarom (2010), eprint 1006.0731.
 Gubser and Pufu (2008) S. S. Gubser and S. S. Pufu, JHEP 11, 033 (2008).
 Roberts and Hartnoll (2008) M. M. Roberts and S. A. Hartnoll, JHEP 08, 035 (2008).
 Basu et al. (2009) P. Basu, A. Mukherjee, and H.H. Shieh, Phys. Rev. D79, 045010 (2009), eprint 0809.4494.
 Ammon et al. (2010) M. Ammon, J. Erdmenger, V. Grass, P. Kerner, and A. O’Bannon, Phys. Lett. B686, 192 (2010), eprint 0912.3515.
 (41) M. Heller and D. Vegh, Work in progress.
 Henningson and Sfetsos (1998) M. Henningson and K. Sfetsos, Phys. Lett. B431, 63 (1998), eprint hepth/9803251.
 Mueck and Viswanathan (1998) W. Mueck and K. S. Viswanathan, Phys. Rev. D58, 106006 (1998), eprint hepth/9805145.
 Iqbal and Liu (2009) N. Iqbal and H. Liu, Fortsch. Phys. 57, 367 (2009).
 Iqbal et al. (2010) N. Iqbal, H. Liu, M. Mezei, and Q. Si (2010), eprint 1003.0010.
 Metlitski and Sachdev (2010) M. Metlitski and S. Sachdev (2010), eprint 1005.1288.