PRL 104, 076802 (2010)

PHYSICAL REVIEW LETTERS

week ending 19 FEBRUARY 2010

Charge Transport in Graphene with Resonant Scatterers

M. Titov,1,2 P. M. Ostrovsky,3,4 I. V. Gornyi,3,5,2 A. Schuessler,3 and A. D. Mirlin3,6,7,2
School of Engineering & Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, United Kingdom DFG Center for Functional Nanostructures, Karlsruhe Institute of Technology, 76128 Karlsruhe, Germany 3 r Nanotechnologie, Karlsruhe Institute of Technology, 76021 Karlsruhe, Germany Institut fu 4 L. D. Landau Institute for Theoretical Physics RAS, 119334 Moscow, Russia 5 A. F. Ioffe Physico-Technical Institute, 194021 St. Petersburg, Russia 6 r Theorie der kondensierten Materie, Karlsruhe Institute of Technology, 76128 Karlsruhe, Germany Institut fu 7 Petersburg Nuclear Physics Institute, 188300 St. Petersburg, Russia (Received 26 August 2009; revised manuscript received 6 November 2009; published 18 February 2010)
2 1

The full counting statistics for the charge transport through an undoped graphene sheet in the presence of strong potential impurities is studied. Treating the scattering off the impurity in the s-wave approximation, we calculate the impurity correction to the cumulant generating function. This correction is universal provided the impurity strength is tuned to a resonant value. In particular, the conductance of the sample acquires a correction of 16e2 =2 h per resonant impurity.
DOI: 10.1103/PhysRevLett.104.076802 PACS numbers: 73.63.b, 73.22.f

Since the discovery of graphene [1] its transport properties have become a subject of intense studies [2,3]. The most remarkable effects arise when the chemical potential is tuned into a close vicinity of the Dirac point. In particular, a short and wide sample of clean graphene exhibits a pseudodiffusive charge transport [4], with the counting statistics equivalent to that of a diffusive wire [5,6]. This equivalence has been conrmed in recent measurements of conductance and noise in ballistic graphene akes [7,8]. In contrast to conventional metals, ballistic graphene near the Dirac point conducts better when potential impurities are added [911]. Quantum interference in disordered graphene is also highly peculiar due to the Dirac nature of carriers. In particular, in the absence of intervalley scattering, the minimal conductivity $e2 =h [2] is protected from quantum localization [12]. Strong impurities creating resonances near the Dirac point (midgap states) are one of the most plausible mechanisms limiting the electron mobility and can be used and exploited for the functionalization of graphene. As was shown in Refs. [13,14], such scatterers provide the concentration dependence of the conductivity,  / nln2 n, which is consistent with most of the experimental observations. Possible realizations of such scatterers are vacancies, adsorbed atoms, molecules, or impurity clusters [15,16]. In particular, an important example is hydrogen atoms that can be controllably added to the graphene sample [17]. Resonant scattering may also be intentionally induced by metallic islands deposited on the graphene surface [18]. In this case the strength of impurity can be controlled by a local external gate. In this Letter we study the effect of strong impurities on electron transport in graphene. We consider the ballistic transport regime in which the sample size is smaller than the electron mean free path (low impurity concentration). This model was experimentally implemented in the 0031-9007= 10 =104(7)=076802(4)

samples on the SiO2 substrate [7,8,19]. The ballistic transport is particularly relevant for suspended samples where much higher mobilities have been achieved [20,21]. We evaluate analytically the effect of strong impurities on the full counting statistics (FCS) in a rectangular sample of length L and width W (Fig. 1). In particular, for the short-and-wide setup, W ) L, we nd the universal corrections to the conductance G and to the shot noise S (divided by 2eV , where V is the bias voltage),   16 e2 S 1 3 16 e2 G 2 ; 2 2 ; (1) 2eV 3   h  h per resonant potential scatterer of the round shape. We consider an effective model described by the Dirac Hamiltonian, H i@v r Ur, where  x ; y is the vector of Pauli matrices, v is the velocity, and Ur is the impurity potential. Below, we set @v 1. Two metallic leads at x < 0 and x > L are dened by adding a large chemical potential to the graphene Hamiltonian. Inside the sample, i.e., for 0 < x < L, the chemical potential is set to zero (Dirac point). The potential Ur represents a collec-

FIG. 1 (color online). Ballistic graphene setup with various strong scatterers. Vacancies as well as atomic or molecular impurities can create midgap states [15]. Metallic islands support quasibound states that can be tuned to the resonance.

076802-1

2010 The American Physical Society

PRL 104, 076802 (2010)

PHYSICAL REVIEW LETTERS

week ending 19 FEBRUARY 2010

tion of randomly distributed sharp scatterers of arbitrary strength. We will study FCS of ballistic electron transport through the sample using two complementary approachesscattering theory and Green function formalism [22]. The rst approach works for any aspect ratio of the sample while impurities are treated as pointlike. The second method is particularly suitable for W ) L and allows for arbitrary impurity prole. We begin with the scattering approach for a single impurity [9]. We discretize y coordinate, y W=N , where  1; 2; . . . ; N is an integer, N W=a ) 1, and a ( L plays the role of the impurity size. The impurity potential in this model is dened as Ur x x0 0 ; (2) where x0 and y0 W0 =N specify the impurity position. Generically,  is a 4 4 matrix in sublattice and valley spaces. Below we concentrate on the case of scalar potential with  being a number. In order to study transport properties, we perform a standard unitary rotation [9]pof the Hamiltonian, H Ly H L with L x z = 2. After this transformation the upper (lower) element of the spinor wave function represents the right- (left-) propagating mode in the leads. We then perform the discrete Fourier transformation with respect to  and arrive at the Dirac equation @x ^ xx; ^ iz U x q @x ^ is a diagonal matrix with entries qn 2n=W where q being the discrete transverse momenta and n N=2; . . . ; N=2 (for deniteness, we assume periodic boundary conditions). The impurity potential is repre^ x x x0 jihj that sented by the operator U projects onto the state with the wave function qn N 1=2 eiqn y0 . Using this separable form of the impurity ^ , we can explicitly calculate the transfer matrix potential U T that relates the wave function amplitudes in the opposite leads, L T 0. The result is given by
^ Lx0 iz jihj x qx e e ^ 0: T ex q

is W times the local density of states of a clean system at the position of impurity. The statistics of the charge transport is described by the cumulant generating function F  [23] as cn lim!0 @n F =@n , where c1 G=g0 is the dimensionless conductance, c2 S=2eV g0 is the dimensionless noise, etc. For our purposes it is more convenient to use the variable  dened via  2 lnj cos=2j, which yields ^t ^y sin2 =2: F  lndet1 t (5)

Using Eq. (3), we calculate conductance and noise of the sample with a resonant impurity placed in the center, see Fig. 2. The general expression for the generating function applicable for any strength and position of the impurity is given in Ref. [24]. In the limit W ) L the summations over transversal modes can be replaced by momentum integrals. For the clean system this yields F 0 W2 =4L. The impurity correction to F takes the form [24]    2 2 1 F 2 ln 1 ; (6) 4L2 2 sin2 x0 =L where 2a tan=2 (as we show below, is the scattering length at low energies). In particular, the correction to the conductance G0 g0 W=L reads G 8g0 =2 sin2 x0 =L 4L2 =2 1 : (7)

This completes our analysis of the discretized model (2). Let us now turn to the microscopic description of the impurity within the continuous Dirac model, assuming that the impurity potential is rotational symmetric, Ur ujr r0 j, where r0 is the impurity position. To calculate FCS we use the matrix Green function approach [6,22]. The Green function in the retarded-advanced space satises the equation   x H i0 x x Gr; r0 r r0 ; x x L x H i0 (8)
0.8 0.6 0.4 0.2
2.5 2.0 1.5 1.0 0.5 0.0 0 2 4 6 8 10

Inverting the element T 11 (where indices refer to the  space), we obtain an exact expression for the transmission amplitude from the nth channel in the left lead to the mth channel in the right lead, tnm nm 2 z iN cot=21 nm ; coshqn L coshqn L coshqm L
iqn qm y0

(3)
0.0 0 5

with nm e z

X
n

coshqn x0 qm L x0 and (4)

10

15

coshqn L 2x0 = coshqn L:

The rst term in Eq. (3) describes the clean system ( 0). The second term represents the effect of the impurity which is particularly important in the vicinity of resonant values n 2n 1. The quantity z dened by Eq. (4)

FIG. 2 (color online). Contribution to the conductance G (solid line) and to the shot noise S (dashed line) from a resonant impurity placed at x0 L=2, as a function of the aspect ratio W=L. Dotted lines show asymptotic values from Eq. (1). Inset: conductance and noise of a clean sample, cf. Ref. [5].

076802-2

PRL 104, 076802 (2010)

PHYSICAL REVIEW LETTERS

week ending 19 FEBRUARY 2010

where  sin=2 is the counting eld and  is the chemical potential which is zero inside the sample and innite in the leads. An explicit solution to the above equation is given in [24]. The generating function can be expressed through G as F  Tr lnG1 , where Tr is the full operator trace. Expanding F  in Ur, we obtain the impurity correction to the generating function as a series F
1 X 1 TrUG0 p ; p p1

(9)

where G0 is the Green function of the clean system. Evaluation of Eq. (9) essentially simplies if the impurity size a is small compared to L. In this limit the arguments of the Green function are close to the position of the impurity, r0 . Therefore it is useful to split G0 into the singular and regular parts, G0 g Greg , where gr; r0 i=2 r r0 =jr r0 j2 is the zero-energy Green function of the free Dirac fermion. The regular part of the Green function, Greg , can be taken with equal arguments r r0 r0 . Explicit expression for Greg r is given in [24]. We characterize the impurity by its T matrix T
1 Z X p 1

u0 a ) 1, one nds 2a tanu0 a =4, which corresponds to  2u0 a =2 in Eq. (2). In Fig. 3 we compare the results (7) and (10) to the conductance calculated numerically in Ref. [26] for the case of the disk impurity. In addition to the s-wave resonances, that are perfectly described by our theory, there exist higher resonances that can be included into our consideration in a similar manner, see [24]. The position of the resonances are given by the zeroes of Jm u0 a, where the corresponding index m 0; 1; 2; . . . is specied in Fig. 3. The resonance widths scale as a=L2m1 . Now we turn to the case of disordered graphene with a small concentration of impurities randomly distributed " can over the sample. The average generating function F be found to the linear order in nimp [see Fig. 4(a)] by the integration of Eq. (6) over the sample area, s  W2 16L2 2 " F 4Nimp ln 1 1 2 2 ; 4L  where Nimp nimp WL is the total number of impurities. Resonant impurities ( 1) contribute on average the universal correction (1) to the conductance and noise, yielding the conductivity  GL=W and Fano factor   4e2 4 1 12  1 nimp L2 ; F 3 nimp L2 :  3  h Away from a resonance, a ( ( L, we nd   nimp 2 4e2 1 nimp 4 1 ; F 3 2:  3 8 L h 2

d2 r1 . . . d2 rp Ur1 gr1 ; r2 . . . Urp :

Rearranging singular and regular parts in Eq. (9) yields F Tr ln1 T Greg r0 : Note that this result holds for any T matrix, including possible valley mixing. In the case of rotational invariant impurity, the T matrix reduces to the scattering length in the s channel. Since TrGreg r0 0 and G2 reg r0 2 =2 sin2 x0 =L=4L2 [24], the impurity correction to the generating function reproduces the result (6). This establishes a relation between the phenomenological parameter  in Eq. (2) and the actual prole of the impurity potential. It is convenient to derive the scattering length from the solution of the Dirac equation in an innite system [25] at a low-energy k. Specically, is related to the scattering cross section as k2 =2. As an example, let us calculate for an impurity with the rectangular prole ur u0 a r. Matching the solutions of the free Dirac equation at r < a and r > a, we obtain 2aJ1 u0 a=J0 u0 a; (10)

(11)

To establish the limits of validity of these ballistic results, we estimate the contribution of the order n2 imp in the virial expansion. The diagrams involving two resonant impurities, Fig. 4(b), yield the free-energy correction 1 1 1 2 F Trln1 G reg r1 G0 r1 ; r2 Greg r2 G0 r2 ; r1 ; 2 which is a contribution of atypical pairs of scatterers separated by a distance & L. The corresponding correc2.5 2 2.0 1.5 1.0 0.5 0.0 0 2 4 6 8 10 0 0 0 1 3 1 4 2

where Jn stands for the Bessel function. The scattering length diverges when the impurity potential develops a quasibound state (one component of the spinor wave function is localized) at zero energy. This implies resonant scattering, which is a generic feature of any strong impurity independent of its prole. In the quasiclassical limit,

FIG. 3 (color online). Correction to the conductance from a circular impurity at x0 L=2 as a function of u0 a. Solid line: analytic result, Eqs. (7) and (10); dotted line: numerical simulations [26] with the parameters a=L 0:2, W=L 6.

076802-3

PRL 104, 076802 (2010)

3.0 2.5
(a)

PHYSICAL REVIEW LETTERS

week ending 19 FEBRUARY 2010

2.0 1.5 0 1
(b)

FIG. 4 (color online). Conductance vs L for resonant impurities. Dotted lines: (a) one- and (b) two-impurity contributions. Inset: corresponding typical diagrams. Dashed line illustrates schematically the crossover to diffusive regime.

In conclusion, we have computed FCS for charge transport through an undoped graphene sheet in the presence of strong potential impurities. The impurity correction to FCS, Eq. (6), is determined by the position of impurity and the low-energy scattering length. The latter diverges when the impurity potential has a quasibound state at zero energy. At such resonant conditions the impurity correction becomes universal, Eq. (1). Our results are fully supported by numerical simulations, Fig. 3. We are grateful to J. Bardarson and T. O. Wehling for valuable discussions. We thank Centro de Ciencias de Benasque, Spain where this work was completed. The work was supported by the EUROHORCS/ESF EURYI grant (I. V. G.) and Rosnauka Grant No. 02.740.11.5072.

tions to the conductance and Fano factor are (see [24]) e2 L4 ;   1:032 n2 h imp
2

 F

4 0:8623n2 imp L :

Two closely located strong impurities detune each other from the resonance, hence the negative correction to the conductance. For off-resonance impurities, ( L, the multiple scattering correction to conductance can be derived from the rst-order result Eq. (11) supplied by the logarithmic renormalization of the effective impurity strength 0 nimp 2 =2 according to Refs. [10,13],   nimp 2 4e2 1  : h 2 nimp 2 logL= When  $ e2 =h (which corresponds to mean free path $L), the system enters the diffusive regime, Fig. 4. This happens at nimp minfL2 ; 2 j logL=jg $ 1. Remarkably, for resonant impurities this implies just one impurity per square L L. The scaling of conductivity in the diffusive regime is determined by the symmetries of impurities in the sublattice and valley space [12]. In particular, for vacancies the quantum interference is suppressed in view of chiral symmetry [13]. This should yield the universal (as long as Nimp is small compared to the number of carbon atoms) minimal conductivity $e2 =h. The situation with strong potential scatterers (e.g., adsorbants inducing midgap states) is similar for intermediate L where the chiral symmetry is approximately preserved, Fig. 4. At larger scales, L > Lc , violation of this symmetry becomes important, and quantum interference comes into play. The symmetry-breaking length Lc is given by Lc $
1=2 2 2 n imp minfnimp ; 1=nimp a g: the chiral symmetry is broken due to detuning from resonance as well as due to higher (nonresonant) scattering channels. For impurities that do not mix the valleys (i.e., smooth on the scale of the lattice constant) this implies antilocalization (symplectic symmetry class) leading to logarithmic increase of conductivity with L. The valley mixing restores conventional (orthogonal) 2D localization.

[1] K. S. Novoselov et al., Science 306, 666 (2004). [2] K. S. Novoselov et al., Nature (London) 438, 197 (2005); Y. Zhang et al., Nature (London) 438, 201 (2005); Y.-W. Tan et al., Eur. Phys. J. Special Topics 148, 15 (2007). [3] A. H. Castro Neto et al., Rev. Mod. Phys. 81, 109 (2009). [4] M. I. Katsnelson, Eur. Phys. J. B 51, 157 (2006). [5] J. Tworzydo et al., Phys. Rev. Lett. 96, 246802 (2006); C. W. J. Beenakker, Rev. Mod. Phys. 80, 1337 (2008). [6] S. Ryu et al., Phys. Rev. B 75, 205344 (2007). [7] F. Miao et al., Science 317, 1530 (2007). [8] R. Danneau et al., Phys. Rev. Lett. 100, 196802 (2008). [9] M. Titov, Europhys. Lett. 79, 17 004 (2007). [10] A. Schuessler et al., Phys. Rev. B 79, 075405 (2009). [11] J. H. Bardarson et al., Phys. Rev. Lett. 99, 106801 (2007); K. Nomura, M. Koshino, and S. Ryu, ibid. 99, 146806 (2007); P. San-Jose, E. Prada, and D. S. Golubev, Phys. Rev. B 76, 195445 (2007); C. H. Lewenkopf, E. R. Mucciolo, and A. H. CastroNeto, ibid. 77, 081410 (2008); J. Tworzydlo, C. W. Groth, and C. W. J. Beenakker, ibid. 78, 235438 (2008). [12] P. M. Ostrovsky, I. V. Gornyi, and A. D. Mirlin, Phys. Rev. Lett. 98, 256801 (2007); Eur. Phys. J. Special Topics 148, 63 (2007). [13] P. M. Ostrovsky, I. V. Gornyi, and A. D. Mirlin, Phys. Rev. B 74, 235443 (2006). [14] T. Stauber, N. M. R. Peres, and F. Guinea, Phys. Rev. B 76, 205423 (2007). [15] T. O. Wehling et al., Phys. Rev. B 75, 125425 (2007). [16] F. Schedin et al., Nature Mater. 6, 652 (2007). [17] D. C. Elias et al., Science 323, 610 (2009). [18] B. M. Kessler et al., arXiv:0907.3661. [19] H. B. Heersche et al., Nature (London) 446, 56 (2007). [20] K. I. Bolotin et al., Solid State Commun. 146, 351 (2008). [21] Xu Du et al., Int. J. Mod. Phys. B 22, 4579 (2008). [22] Yu. V. Nazarov, Phys. Rev. Lett. 73, 134 (1994). [23] L. S. Levitov and G. B. Lesovik, JETP Lett. 58, 230 (1993). [24] See supplementary material at http://link.aps.org/ supplemental/10.1103/PhysRevLett.104.076802. [25] M. Hentschel and F. Guinea, Phys. Rev. B 76, 115407 (2007); D. S. Novikov, ibid. 76, 245435 (2007); D. M. Basko, ibid. 78, 115432 (2008). [26] J. H. Bardarson, M. Titov, and P. W. Brouwer, Phys. Rev. Lett. 102, 226803 (2009).

076802-4