The tunneling density-of-states of interacting massless Dirac fermions
We calculate the tunneling density-of-states (DOS) of a disorder-free two-dimensional interacting electron system with a massless-Dirac band Hamiltonian. The DOS exhibits two main features: i) linear growth at large energies with a slope that is suppressed by quasiparticle velocity enhancement, and ii) a rich structure of plasmaron peaks which appear at negative bias voltages in an n-doped sample and at positive bias voltages in a p-doped sample. We predict that the DOS at the Dirac point is non-zero even in the absence of disorder because of electron-electron interactions, and that it is then accurately proportional to the Fermi energy. The finite background DOS observed at the Dirac point of graphene sheets and topological insulator surfaces can therefore be an interaction effect rather than a disorder effect.
Introduction: A V-shaped density-of-states (DOS) curve, where is energy measured from the Dirac point, is a distinctive feature of graphene and of topologial insulator (TI) surface states. Both two-dimensional electron systems are described by free-particle massless Dirac fermion (MDF) Hamiltonians: graphenetheoreticalreviews (); TIreviews (). Here is a 2D vector of Pauli matrices and is the MDF velocity, which is roughly three-hundred times smaller than the speed of light in the graphene sheet case graphenetheoreticalreviews (); grapheneexperimentalreviews (), and six-hundred times smaller TIreviews () in the case of the currently-studied chalcogenide TI’s.
The DOS of electronic systems has often been measured using tunneling, and in recent decades often using scanning tunneling microscopy (STM) which probes locally to mitigate the influence of inhomogeneities and some types of disorder. In STM sts_review () the DOS is proportional to the differential conductivity , where and are respectively the tunneling current and the bias voltage between the STM tip and the sample. When graphene sheets STMgraphene () or TI surfaces STMTI () are studied by STM spectroscopy, the experimental DOS curves are never simple V-shaped curves. This is unsurprising since disorder, electron-phonon, and electron-electron interactions manoharangroup () can all play a role in altering the tunneling DOS. Brar et al. brar_prl_2010 () have however recently discovered some very specific gate-voltage dependent, i.e. carrier density dependent, features in the spectra of doped graphene sheets deposited on which they attributed to electron-plasmon interactions. Similar features with an enhanced amplitude have been observed by the same group for graphene sheets deposited on h-BN decker_nano_2011 (), suggesting that the effect is rather universal.
In this Rapid Communication we propose electron-electron interactions as the source of these DOS spectral features. Indeed theory has already predicted polini_prb_2008 (); hwang_prb_2008 () that electron-plasmon interactions have a large impact on the one-body spectral function of doped graphene. After carrying out extensive studies of in quasi-freestanding doped graphene sheets on (hydrogen-terminated) SiC using angle-resolved photoemission spectroscopy (ARPES), Bostwick et al. bostwick_science_2010 () recently discovered experimentally that coupling between electrons and plasmons leads to a substantial reconstruction of the MDF conical spectrum. The reconstruction is related to the appearance of new composite quasiparticles, known as plasmarons earlywork (), which are expected to appear when interactions between electrons and the collective charge-density (plasmon) oscillations of an electron liquid Giuliani_and_Vignale () are strong enough. Plasmarons are signaled theoretically by multiple solutions of the one-particle Dyson equation, and experimentally by satellite bands in the ARPES spectra. Since the tunneling DOS is given by an integral over all momenta of , plasmaron-related features must also be present in the STM spectra.
Fig. 1a) summarizes our theoretical results for the tunneling DOS of electrons with MDF bands which were obtained using a continuum model (random phase) approximation polini_prb_2008 (); hwang_prb_2008 (), with the dimensionless Coulomb coupling constant and the flavor number set to values appropriate for graphene on h-BN. The flavor number accounts for graphene’s spin and valley degeneracies. ( for TI surfaces.) The DOS differs from the non-interacting result , in three ways: i) at large energies the DOS is linear, but its slope is altered because the quasiparticle velocity is renormalized elias_natphys_2011 (); borghi_ssc_2009 (); ii) the DOS does not vanish at any energy; and iii) the DOS displays a rich structure of plasmaron peaks separated by local minima located near the energies of the Dirac crossings bostwick_science_2010 () in the corresponding ARPES spectrum [Fig. 1b)].
In what follows we introduce and discuss qualitative features of the continuum model Hamiltonian and the random phase () approximation that we have employed in combination to approximate 2D MDF interaction physics to obtain the results illustrated in Fig. 1. We then compare the roles of electron-electron and electron-disorder interaction in filling in the Dirac model DOS V, concluding that interaction effects play the dominant role in high quality graphene samples.
The Coulomb MDF model: We describe 2D electron systems with MDF bands using the following Hamiltonian (setting from now on):
Here is the bare Fermi velocity , is the sample area, and () creates (destroys) an electron with momentum and two-valued (pseudo)spin index , and flavor indices are implicit. In the case of graphene, is a sublattice-pseudospin index, whereas in the case of TI surface states it is a true spin index. In Eq. (1) is the density operator, and is the 2D Fourier transform of Coulomb potential. Here depends on the dielectric environment surrounding the 2D MDF fluid. In the case of graphene , where () is the dielectric constant of the medium above (below) the sheet. For the surface states of a thick TI where is the TI bulk dielectric constant. Because chalcogenide TI’s have large dielectric constants, interaction effects on their properties will tend to be weak. For this reason we concentrate on graphene MDF’s in what follows. Interaction effects can be made stronger in chalcogenide TI MDF’s by preparing thin film samples, but interactions between top and bottom surfaces will then play an essential role.
Below we describe an electron-doped system with an excess electron density ; the properties of hole-doped systems can be obtained by appealing to the model’s particle hole symmetry as we explain below. The corresponding Fermi wave number and Fermi energy are and , respectively. In calculating the DOS it will be convenient to use the Fermi level rather than the Dirac point as the zero of energy so that the Dirac-band single-particle energies are where distinguishes the conduction and valence bands. Note that the band and interaction terms scale in the same way with changes in length scale. The spectra we calculate will therefore be independent of if energies are measured in terms of , apart from weak logarithmic dependences on a -band width ultraviolet cutoff scale which must be included in the theory. It follows in particular that the DOS at the Dirac point is accurately proportional to the Fermi energy.
approximation spectral functions: The DOS is related to the spectral function by
where and (suppressing the variables for simplicity)
The quantity denotes the electron-electron interaction contribution to the quasiparticle self-energy. The real part of is measured from its value at the Fermi surface, which physically represents the interaction contribution to the chemical potential, . Correspondingly, the variable in Eqs. (2)-(3) [and also in Eq. (The tunneling density-of-states of interacting massless Dirac fermions) below] represents energy measured from the chemical potential of the interacting system and in units of the Fermi energy . Physically, represents Giuliani_and_Vignale () the probability density for increasing or decreasing the energy of the -particle system by upon adding or removing a single particle in state . An STM experiment probes both occupied () and empty () states, whereas ARPES probes the occupied portion of the spectrum only.
For the quasiparticle self-energy we use the (or random phase) approximation Giuliani_and_Vignale (); polini_prb_2008 (); hwang_prb_2008 (); bostwick_science_2010 (), in which the self-energy is expanded up to first order in the dynamically-screened interaction :
where , is the MDF model chirality factor, and is a shorthand for . The RPA dielectric function is , where is the well-known Lindhard response function of non-interacting 2D MDFs wunsch_njp_2006 () at arbitrary doping . The real part of the self-energy is obtained from Eq. (The tunneling density-of-states of interacting massless Dirac fermions) by a Kramers-Kronig transform.
Note that, since the MDF model is particle-hole symmetric, the real and imaginary [ and ] parts of the quasiparticle self-energy for electron doping are equal to and for hole doping. Thus and for electron doping are equal to and for hole doping.
Interaction Strength and Disorder Dependence: As seen in Fig. 1b), the plasmaron bands loose spectral weight as they approach the Fermi energy but are otherwise remarkably well defined, providing a lower energy echo of the quasiparticle Dirac point. The diamond-like shape in space bostwick_science_2010 () that can be clearly recognized in Fig. 1b) is bounded by plasmaron and quasiparticle peaks which cross at a finite value of at an energy which lies between the quasiparticle band crossing at energy and the plasmaron band crossing at energy .
The diamond feature represents a massive reconstruction of the MDF spectrum. As shown in Figs. 1a) and 2, it has a striking impact on the DOS at low energies. Note that for a n-doped system the strongest plasmaron features appear in the filled portion of the spectra. The particle-hole symmetry properties explained above imply that the opposite occurs for a p-doped system. At large energies we clearly see that the DOS grows linearly with a slope which decreases as increases, an effect which can be understood qualitatively in terms of the interaction enhancement borghi_ssc_2009 () of the quasiparticle Fermi velocity .
In Fig. 2 we plot the DOS of an interacting 2D MDF system for three values of the coupling constant . Due to the increase of screening with increasing , the amplitude of the plasmaron peaks in the energy range becomes quite small when . The DOS does not vanish at any energy when interactions are included, as observed experimentally in Ref. brar_prl_2010, . The strength of the interaction induced background in the non-interacting V can be characterized by the absolute minimum value of : . To estimate how important this background typically is for the interpretation of experiment, we can compare it with the finite DOS background due to disorder. To this end, we calculate the DOS of a 2D MDF system in the presence of a model disorder potential, including electron-electron interactions only through the static RPA screening function . For the sake of definiteness, we consider Coulomb impurities (CI) located at random in a plane parallel to the 2D electron system that is separated by distance . Let be the impurity density. For our qualitative goals it is sufficient to use the self-consistent Born approximation (SCBA) vozmediano_prb_2010 () for which the disorder self-energy reads
where is the statically-screened electron-impurity potential, and is the 2D Fourier transform of the bare electron-impurity potential (assuming that all impurities have charge equal to , in absolute value). The DOS in the presence of disorder, , can be computed by iteratively evaluating the integral in Eq. (5) numerically until is self-consistent.
In Fig. 3 we plot the DOS background due to long-range scatterers, , as a function of the impurity concentration and for different values of the fine-structure constant . Note, indeed, that since we have assumed Coulomb disorder, the strength of the electron-impurity interaction is controlled by the same dimensionless coupling constant which determines the strength of electron-electron interactions. We clearly see from this plot that the impurity concentration needed to give a background comparable to is rather high. For example, for graphene on the impurity density necessary to make disorder more important than interactions is more than twice the electron density. Typical impurity densities, extracted from a comparison between theory and transport data dassarma_rmp_2011 () which assumes Coulomb scatterers, are for comparison on the order of , giving typical values of . We thus conclude that the background due to electron-electron interactions is very significant, at least in comparison with that due to long-range Coulomb impurities. Of course, the disorder effect will always dominate at very small carrier densities.
In summary, we have calculated the tunneling DOS of a 2D system of interacting MDFs. We have found that plasmarons in the one-body spectral function give rise to a non-trivial structure of peaks and valleys in the DOS. The spectral weight carried by plasmarons gives rise to a finite background in the DOS, which does not vanish at any energy and instead displays two peaks associated with the electron and plasmaron Dirac points bostwick_science_2010 (). A more direct comparison with the experimental results in Ref. decker_nano_2011, is necessary to determine convincingly whether or not the DOS peak has already been observed in the STM spectra of graphene on h-BN. Our predictions can also be tested by low-temperature STM spectroscopy on quasi-freestanding doped graphene sheets on (hydrogen-terminated) SiC NEST (). Finally, our findings are also relevant to STM studies of TI surface states. A finite background has, for example, recently been observed by Alpichshev et al. STMTI () in the STM spectra of the surface states of which could be due in part to interactions.
Acknowledgements. — Work in Pisa was supported by the Italian Ministry of Education, University, and Research (MIUR) through the program “FIRB - Futuro in Ricerca 2010” (project title “PLASMOGRAPH: plasmons and terahertz devices in graphene”). A.H.M. was supported by the Welch Foundation under grant TBF1473 and by the NRI SWAN program. As this manuscript was being completed we became aware of recent closely related work by LeBlanc et al. leblanc_prb_2011 (). The authors of this work also predict distinct plasmaron features in the DOS of an interacting system of MDFs.
- (1) A.H. Castro Neto et al., Rev. Mod. Phys. 81, 109 (2009).
- (2) M.Z. Hasan and C.L. Kane, Rev. Mod. Phys. 82, 3045 (2010); J.E. Moore, Nature 464, 194 (2010); X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057 (2011).
- (3) A.K. Geim and K.S. Novoselov, Nature Mater. 6, 183 (2007); A.K. Geim, Science 324, 1530 (2009).
- (4) J. Tersoff and D.R. Hamann, Phys. Rev. B31, 805 (1985); C.J. Chen, Introduction to Scanning Tunneling Microscopy (Oxford University Press, New York, 2008).
- (5) For a recent review see e.g. M. Morgenstern, Phys. Status Solidi (b) 248, 2423 (2011).
- (6) See e.g. P. Roushan et al., Nature 460, 1106 (2009); T. Zhang et al., Phys. Rev. Lett. 103, 266803 (2009); Z. Alpichshev et al., Phys. Rev. Lett. 104, 016401 (2010); Y. Okada et al., ibid. 106, 206805 (2011); H. Beidenkopf et al., Nature Phys. advance online publication, 09 October 2011 (DOI 10.1038/nphys2108).
- (7) To the best of our knowledge, the only 2D MDF systems to date that display an almost perfect V-shaped DOS are graphene flakes on graphite [G. Li, A. Luican, and E.Y. Andrei, Phys. Rev. Lett. 102, 176804 (2009)] and artificial graphene lattices created by placing atoms one-by-one on the surface of a metal with a honeycomb lattice arrangement [K.K. Gomes et al., http://meetings.aps.org/link/BAPS.2011.MAR.A21.8].
- (8) V.W. Brar et al., Phys. Rev. Lett. 104, 036805 (2010).
- (9) R. Decker et al., Nano Lett. 11, 2291 (2011).
- (10) M. Polini et al., Phys. Rev. B77, 081411(R) (2008); see also A. Principi, R. Asgari, and M. Polini, Solid State Commun. 151, 1627 (2011).
- (11) E.H. Hwang and S. Das Sarma, Phys. Rev. B77, 081412(R) (2008).
- (12) A. Bostwick et al., Science 328, 999 (2010); A.L. Walter et al., Phys. Rev. B84, 085410 (2011).
- (13) B.I. Lundqvist, Phys. Kondens. Mater. 6, 193 (1967); L. Hedin et al., Solid State Commun. 5, 237 (1967); for recent experimental work on plasmarons see R. Tediosi et al., Phys. Rev. Lett. 99, 016406 (2007) and O.E. Dial et al., arXiv:1009.0519.
- (14) G.F. Giuliani and G. Vignale, Quantum Theory of the Electron Liquid (Cambridge University Press, Cambridge, 2005).
- (15) D.C. Elias et al., Nature Phys. 7, 701 (2011).
- (16) J. González, F. Guinea, and M.A.H. Vozmediano, Nucl. Phys. B 424, 595 (1994) and Phys. Rev. B59, R2474 (1999); G. Borghi et al., Solid State Commun. 149, 1117 (2009).
- (17) See e.g. B. Wunsch et al., New J. Phys. 8, 318 (2006).
- (18) F. de Juan, E.H. Hwang, and M.A.H. Vozmediano, Phys. Rev. B82, 245418 (2010).
- (19) S. Das Sarma et al., Rev. Mod. Phys. 83, 407 (2011).
- (20) S. Goler et al., submitted.
- (21) J.P.F. LeBlanc, J.P. Carbotte, and E.J. Nicol, Phys. Rev. B84, 165448 (2011).