Dynamical current-current correlation of the hexagonal lattice and graphene

T. Stauber1,2 and G. Gómez-Santos1

1
Departamento de Fı́sica de la Materia Condensada and Instituto Nicolás Cabrera,
Universidad Autónoma de Madrid, E-28049 Madrid, Spain and
2
Centro de Fı́sica e Departamento de Fı́sica, Universidade do Minho, P-4710-057, Braga, Portugal
(Dated: October 25, 2018)
We discuss the dynamical current-current correlation function of the hexagonal lattice using a
local current operator defined on a continuum-replica model of the original lattice model. In the
arXiv:1009.3812v2 [cond-mat.mes-hall] 23 Oct 2010

Dirac approximation, the correlation function can be decomposed into a parallel and perpendicular
contribution. We show that this is not possible for the hexagonal lattice even in the Dirac regime.
A comparison between the analytical isotropic solution and the numerical results for the honeycomb
lattice is given.

PACS numbers: 81.05.ue, 75.20.-g, 75.70.Ak, 73.22.Pr

I. INTRODUCTION for finite chemical potential µ at zero temperature.8,9 In

Graphene is a two-dimensional carbon allotrope which
was isolated in 20041 and has attracted immense re-
search activities due to its novel mechanical and elec-
tronic properties.2–5 Whereas the mechanical properties
are determined by electrons with sp2 -hybridization, the
electronic properties can be mainly deduced considering
only the π-electrons. The simplest model to study the
electronic response of graphene to an external field or po-
tential is thus given by a one-orbital tight-binding model
on a hexagonal lattice.
Most of the novel electronic properties of graphene
originate from the fact that there are two equivalent
atoms in the Wigner-Seitz cell which give rise to two
gapless bands with linear density of states close to the
neutrality point. Most standard results of solid state text
books can thus not be applied to the case of graphene due
to the different dispersion and/or dimensionality, but also
due to the two coherently coupled bands.
An example is the density-density correlation or Lind-
hard function which in the case of the honeycomb lattice
is given by6
−gs
Z
X 0,0
0,0 2
π (q, ω) = d k fs·s′ (k, q) (1)
(2π)2 1.BZ
s,s =±
′
the static case, it shows differences to the one-band result
by Stern10 for |q| ≥ 2kF with kF the Fermi wave vector
due to the contribution of interband processes. For finite
frequencies, these differences are even more pronounced
and lead to a logarithmic singularity at ~ω = 2µ.
The density-density correlation function or polariz-
ability of graphene was calculated in a number of pa-
pers using different formalisms and introducing various
modifications to the original Dirac Hamiltonian.11–23 Us-
ing these results, plasmons,8,9,20,24 wrinkles,25 van-der-
Waals interactions26 and forces due to moving external
charges27 were discussed. In this paper, we will focus on
the related current-current correlation function π i,j (q, ω)
of graphene (i, j = x, y) starting from the tight-binding
model of the honeycomb lattice.
In the Dirac approximation, the system is rotationally
invariant and the current-current correlation function can
be decomposed in a parallel π k and perpendicular con-
tribution π ⊥ . The parallel contribution is related to the
density-density correlation function via the continuity
equation and thus determines the dielectric properties of
the system. The perpendicular contribution is related to
the magnetic susceptibility which in the static case has
been first discussed by McClure28 via the Helmholtz free
energy and recently by Ando and co-workers using π ⊥ .29

′ In view of new experiments on the magnetic behavior of

nF (E s (k)) − nF (E s (k + q)) graphene30 , the magnetic susceptibility was also calcu-
× s ,
E (k) − E s′ (k + q) + ~ω + iδ lated including electron-electron interactions to first or-
der which results in a paramagnetic response away from
with the eigenenergies E ± (k) = ±t|φ(k)| (t ≈ 2.7eV is half-filling.31
the hopping amplitude), nF (E) the Fermi function, gs =
Here, we shall mainly discuss π ⊥ (q, ω) for finite fre-
2 the spin-degeneracy and φ(k) the complex structure
quencies. In the Dirac approximation, this was first
factor defined below. Due to the two gapless bands, the
done in Ref.32 . We will summarize their results and
above expression contains the band-overlap function
compare the analytical solution of the isotropic system
0,0 1
 
φ(k) φ∗ (k + q)
 with the numerical solution of the hexagonal lattice. For
f± (k, q) = 1 ± Re , (2) that, we will define a local current operator defined for a
2 |φ(k)| |φ(k + q)|
continuous-replica model of the original lattice Hamilto-
which marks the crucial difference to the standard text- nian. This formalism permits deeper insight in the lattice
book results containing only one band.7 effects and can be used to calculate corrections which are
In the linear (Dirac) approximation of the band dis- lost in the scaling limit, i.e., the Dirac model.
persion, the above expression can be solved analytically The paper is organized as follows. In section II, we will
 2

define the continuum model and derive the local current where c = a, b and A the area of the sample.
operator of this model. We will further show that this op- The continuous version of the Hamiltonian thus reads
erator satisfies the continuity equation with respect to the Z
density operator defined on the lattice. In section III, we H = −t d2 r a† (r)φ(−i∇)b(r) + H.c. .
 
(9)
will present general expressions for the current-current
correlation function and introduce the parallel and per- The gauged Hamiltonian is obtained by replacing −i∇ →
pendicular contribution defined for the Dirac model. In −i∇ + ~e A(r) (e > 0). Notice that by going back in
section IV, we summarize the analytical results and com- Fourier space, we obtain the correct Peierls substitution
pare them with the numerical results obtained from the
R R+δ
hexagonal lattice. We close with a summary and con- tR,R+δ → tR,R+δ ei ~
e
R
dlA(l)
(10)
clusions and give real expressions for the current-current
correlation function in an appendix. in the case of a gauge field which is constant over one
lattice spacing, i.e., for a spatially weakly varying field.
Because e∇·δ b(r) = b(r + δ), we can write Eq. (9) as
II. CONTINUUM MODEL AND CURRENT
OPERATOR
Z X
H = −t d2 r a† (r)b(r + δ) + H.c. .

(11)
δ
To calculate the current-current correlation function
for a lattice model for finite wave vector q, we are con- The continuous model thus consists of infinitely many
fronted with the following problem. The current operator replica of the original lattice model. The unperturbed
for a lattice model, as given by the continuity equation, Hamiltonian is homogeneous, real (not crystalline) mo-
describes the flow from site i to j per unit time.33 In or- mentum is conserved and yet, each particle is bound to
der to define a vector which depends on one lattice site hop in the replica where it lives with strict fidelity to the
instead of two, one needs to define a continuous model original lattice Hamiltonian. In particular, the lattice
based on the Hamiltonian in reciprocal space. If the vec- anisotropy is fully preserved. Also, the minimal substi-
tor potential A is a smooth function of r, then the cou- tution used to include the perturbing vector potential
pling between A and the current can only see the smooth guarantees gauge invariance to all orders.
part and the continuous limit is justified. For this model, the current can be defined by
Let us start with the tight-binding Hamiltonian of
a general bipartite lattice with Nc lattice sites R and δH
j(r) = − = jP (r) + jD (r) + O(A2 ) , (12)
nearest-neighbor lattice vectors δ: δA(r)
Xh i
H=− tR,R+δ a†R bR+δ + H.c. (3) where the diamagnetic contribution jD is linear in the
R,δ gauge field A.
For the paramagnetic operator, we obtain
The spin-index on the operators shall be suppressed
throughout this work. With the Fourier components ite X
Z 1
jP (r) = ds a† (r − sδ)b(r + (1 − s)δ) + H.c.
 
δ
1 X ik·R ~ 0
aR = √ e ak , (4) δ
Nc k (13)
1 X ik·(R+δ)
bR+δ = √ e bk , (5) which consists of a symmetrized version of the paramag-
Nc k netic current given in Refs.35,36 which is obtained from
this reads for tR,R+δ = t the above formula by setting s = 0.
For the diamagnetic contribution, we obtain with
(summation over j is implied)
Xh i
H = −t φ(k)a†k bk + H.c. (6)
k
δj i (r) 
Z 
j D,i (r) = d2 r′ Aj (r′ ) (14)
with φ(k) = δ eik·δ the complex structure factor where
P
δAj (r′ ) A=0
the sum goes over all nearest-neighbor vectors δ. Notice
that the phase factor eik·δ in Eq. (5) is important for the the following expression:
definition of the current.34
te2 X i j 1
Z
We will now define a continuous model by introducing j D,i (r) = − 2 δδ dsds′ a† (r − sδ)b(r + (1 − s)δ)

the following Fourier components ~ 0
δ
× sA (r − ss δ) + s′ Aj (r + ss′ δ) + H.c.
′
 j
1
Z 
ck = √ d2 re−ik·r c(r) , (7)
A (15)
1 X ik·r
c(r) = √ e ck , (8) which again resembles a symmetrized version of the dia-
A k magnetic current given in Refs.35,36 which this time is
 3

obtained from the above formula by setting s = 0 and We obtain the general expression for the current-
s′ = 1. Notice that the diamagnetic current is non-local current correlation function
in the external gauge field.
In linear response, only ground-state averages enter in  2
the diamagnetic current. With the energy per bond per te −gs
Z X i,j
unit area π i,j (q, ω) = d2 k fs·s ′ (k, q)
~ (2π)2 1.BZ s,s′ =±
† s′
hbond = −2tha (r) b(r + δ)i , (16) s
nF (E (k)) − nF (E (k + q))
× , (22)
E s (k) − E s′ (k + q) + ~ω + iδ
which is independent of both r and δ, the Fourier trans-
form of the paramagnetic and diamagnetic current are
given by:
with E ± (k) = ±t|φ(k)| and nF (E) = (eβ(E−µ) + 1)−1
te X i the Fermi function.
jqP,i = φ̃ (k, q)a†k bk+q + (φ̃i (k, q))∗ b†k ak+q , (17)
~ This is the same expression as for the density-density
k
hjqD,i i = D,i,j
χq Ajq (18) correlation function of Eq. (1), but the band-overlap is
now given by
with
X δi  
i,j 1 h i i
φ̃i (k, q) = ei(k+q)·δ − eik·δ , (19) f± (k, q) = Re φ̃ (k, q)(φ̃j (k, q))∗ (23)
q·δ 2 
δ
φ∗ (k) φ∗ (k + q)

e2 X 4 q·δ ±Re φ̃i (k, q)φ̃j (k, q) .
χD,i,j
q = 2 hbond δi δj sin2 ( ). (20) |φ(k)| |φ(k + q)|
~ (q · δ)2 2
δ

For q → 0, we obtain the same expression as in Refs.36,37 .

The Fourier transform of the particle density of the Due to charge conservation, we have
lattice model is given by nq = k (a†k ak+q + b†k bk+q ).38
P e2 ω 2 Imπ 0,0 (q, ω) = qi qj Imπ i,j (q, ω) where summa-
For the charge density ρq = enq , the continuity equation tion over double indices is implied. To see this within
ρ̇q − iq · jq = 0 is obeyed for the paramagnetic current our notation, we note that qi φ̃i (k, q) = φ(k) − φ(k + q)
operator of Eq. (17). We can thus consider this operator and thus
to be the current operator of the lattice model for general
q. In the same manner, the diamagnetic term is also i,j
correct for arbitrary q. 0,0 f± (k, q)
f± (k, q) = q i q j (24)
(|φ(k)| ∓ |φ(k + q)|)2

III. CORRELATION FUNCTION

which proves the relation since (~ω/t)2 = (|φ(k)|∓|φ(k+
We can now determine the current-current correlation q)|)2 .
function. In terms of the bosonic Matsubara frequencies In the Dirac cone approximation, the expressions sim-
~ωn = 2πn/β (β = 1/kB T ), it is defined by plify considerably. Denoting the angle between k and q
~β by ϕ and neglecting terms proportional to sin ϕ which
1
Z
π i,j (q, iωn ) = dτ eiωn τ hjP,i P,j
q (τ )j−q i . (21) cancel to zero due to the angle integration, we have for
~A 0 the effective band overlap

2 !
2 (q 2 − q 2 ) h

i,i 1 3a k x y q i
f± = 1 ± (−1)δi,y 2 1 − 2 sin2 ϕ + cos ϕ , (25)
2 2 q |k||k + q| k
 2 
i,j 1 3a k 2 2qx qy h 2 q i
f± = ±(1 − δi,j ) 2 1 − 2 sin ϕ + cos ϕ , (26)
2 2 q |k||k + q| k

where we introduced the carbon-carbon distance a = The system linearized around the Dirac point is ro-
0.14nm. tationally invariant. We can thus decompose π i,j into
 4

a longitudinal component π k and transverse component where the energy per bond per unit area of the hexagonal
π ⊥ . These are defined by Eq. (22) after substitution of lattice is given by
i,j
the overlap function f± by gs X +
E (k) nF (E − (k)) − nF (E + (k)) ,
 
hbond =
2  3A
k
k + q cos ϕ − 2k sin2 ϕ
 
k(⊥) 1 3a
f± = 1 ± (∓) . (34)
2 2 |k + q|
(27) and the band cutoff ΛE = 6t.

We then recover the general relation

IV. RESULTS
qi qj k qi qj ⊥
π i,j (q, ω) = π (|q|, ω) + (δi,j − )π (|q|, ω) .
|q|2 |q|2 We will now summarize the analytical results obtained
(28) for the Dirac approximation at zero temperature first pre-
sented in Ref.32 and compare them with the numerical
0,0
We note that the overlap function f± in the Dirac ap- results obtained from the hexagonal lattice.
k
proximation is proportional to f± , but with the last term,
2k sin2 ϕ, missing.8
A. Analytical results
Due to current conservation and q 2 π k = qi π i,j qj , the
parallel component of the current-current correlation is
related to the density-density correlation by In order to present the analytical results, we express
the current-current correlation function of Eq. (22),
q 2 π k (|q|, ω) = −h[ρq , q · j−q ]i/(~A) + e2 ω 2 π 0,0 (|q|, ω) . π ± (q, ω), by two dimensionless functions
(29)  2 
e t  ±
π ± (q, ω) = π0 (q, ω) + ∆πµ± (q, ω) ,

2
(35)
~
Apart from the constant surface or contact term, which
was determined in Ref.39 for the linearized Dirac model, where we will use the the superindex + to denote the lon-
we are thus left with the calculation of the perpendicular gitudinal component (k) and the superindex − to denote
component π ⊥ which is related to the magnetic suscepti- the transverse component (⊥). We restrict the discus-
∗
bility χM (q, ω)/µ0 = π ⊥ (q, ω)/|q|2 for ω ≪ |q| with µ0 sion to ω ≥ 0 since π ± (q, −ω) = [π ± (q, ω)] and to µ ≥ 0
the magnetic permeability.33 due to particle-hole symmetry. π0± contains the contri-
For the full dispersion, we have −h[ρq , q · j−q ]i/(~A) = bution for the system at half-filling, i.e., interband con-
qi χD,i,j
q qj . It is thus often more transparent to deal with tributions, whereas ∆πµ± contains the contributions due
the physical response, Πi,j , which includes the diamag- to the finite chemical potential µ, i.e., intraband contri-
netic contribution: butions. The formulas are given in terms of the Fermi
velocity ~vF = 32 at.
Πi,j (q, ω) = π i,j (q, ω) + χD,i,j
q (30) The results can be written in compact form using two
dimensionless, complex functions defined as
Charge conservation then implies
g ~ω
  v q 2 ∓ 21
± F
F (q, ω) = 1− , (36)
qi Πi,j (q, ω) qj = e2 ω 2 π 0,0 (q, ω) . (31) 16π t ω
p  p 
Notice that the anisotropy of the response for finite q G± (x) = x x2 − 1 ∓ ln x + x2 − 1 . (37)
requires the full tensorial structure of Πi,j . In particu-
Let us first present the results for the undoped system.
lar, the relation between polarizability and conductivity
For large energy cutoff ΛE ≫ 1, we have
reads
 
g ΛE
qi σ i,j (q, ω) qj = iωe2 π 0,0 (q, ω) . (32) π0± (q, ω) = + iπF ± (q, ω) . (38)
8π t
We will show in the next section that the often used Notice that the constant cutoff term can be obtained ei-
scalar version of Eq. (32) would not hold for the lattice ther from the Kramers-Kronig relation or from the conti-
model even in the regime where the Dirac approximation nuity equation. This connection gives rise to the so-called
is justified. f-sum rule.39
We finally state the general f-sum rule for a bipartite The contribution due to the finite chemical potential
tight-binding model: reads
g µ ω2
∆πµ± (q, ω) = ± ∓ F ± (q, ω) G± (x+ ) (39)

ΛE
2 4hbond X 2 q · δ
Z
2π t (vF q)2
dωωImπ 0,0 (q, ω) = sin ( ) (33)
π ~2 2
0
−Θ (x− − 1) G± (x− ) ∓ iπ − Θ (1 − x− ) G± (−x− )
 
δ
 5

|q|a=0.1, µ/t=0.05
where we defined x± = 2µ±~ω
~vF q . 4

The above expression for graphene shall be contrasted i=x,qx=q,qy=0

3

Im π (qx,qy,ω)
with the expression for the two-dimensional electron gas. i=x,qx=0,qy=q
i=y,qx=q,qy=0
For quadratic dispersion ǫk = ~2 k2 /(2m), we have 3
i=y,qx=0,qy=q
2

Im π (qx,qy,ω)
||
π

i,i
 v q 2 ∓ 21 _|_
( !) π
2
 e 2 µ 
ω m F
π ± (q, ω) = ± 2 1− 1− , 2

i,i
~ 2π q 2π ω 1
1.9 2 2.1
ω/t
(40) 1

where the term proportional to µ = ǫkF corresponds to

the contact term which is canceled by the diamagnetic 0
0 1 2 3 4 5 6
contribution. ω/t
Eq. (39) can be written as real and imaginary part in
terms of three real dimensionless functions FIG. 1: The imaginary part of the current-current correla-
tion function Imπ i,i (qx , qy , ω) as function of the energy ω at

g ~ω   v q 2 ∓ 12 kB T /t = 0.01 for different directions with |q|a = 0.1. The
± F
f (q, ω) = 1− , (41)
results obtained from the Dirac-cone approximation π k (q, ω)

16π t  ω 
p
−1
and π ⊥ (q, ω) are also shown (dashed lines).
G± 2
> (x) = x x − 1 ∓ cosh (x) , x > 1 ,
p
± 2 −1
G< (x) = ±x 1 − x − cos (x) , |x| < 1 .

The lengthy expressions are given in the appendix.
Let us now discuss two limiting cases. For the long
wavelength limit q → 0, we obtain
e2 g
the imaginary part of the current-current correlation
function Imπ i,i (qx , qy , ω) as function of the energy ~ω at
kB T /t = 0.01 for different directions with |q|a = 0.1.
The results obtained from the Dirac-cone approxima-
  
2µ  2µ + ~ω 
±
π (q → 0, ω) = ω − + ln 
  tion π k (q, ω) and π ⊥ (q, ω) are also shown (dashed lines).
~ 8π ~ω 2µ − ~ω  Clearly, there are strong differences for energies ~ω > t
π i
due to the van Hove singularity. The inset shows that
+ i Θ(~ω − 2µ) . (42)
2 there is a peak splitting for the different directions due
Using the RPA-approximation for the longitudinal part, to the different contributions of the three M -points, also
the above expansion leads to plasmon excitations for present in the charge response.21
which the logarithmic term is usually neglected.8,9 Due In Fig. 2, the same curves are shown, but for lower en-
to the sign change of the photon propagator in the ergies. On the left hand side, the wave vector q is parallel
case of transverse modes, the denominator of the RPA- to the current and on the right hand side perpendicular.
approximation cannot become zero for the perpendicular The results obtained from the Dirac-cone approximation
part without the logarithmic term. But including it leads π k (q, ω) and π ⊥ (q, ω) are also shown (dashed lines).
to a new transverse electromagnetic mode in graphene.40 For the perpendicular contribution of π i,i (right hand
For the static case, we obtain the following formula side), clear differences are seen for lower energies due to
which was first given in Ref.29 : the finite temperature kB T /t = 0.01 used in the numer-
ical calculation. This results in a thermal broadening of
e2 g 2kF
π − (q, ω = 0) = vF qΘ(q − 2kF )G−
<( ) (43) the delta-function of Eq. (44) and is responsible for the
~ 8π q diamagnetism found in graphene30 since intrinsic dop-
ing leads to µ 6= 0. The dotted lines on the right hand
where kF = µ/(~vF ). The parallel component π + is zero.
side refer to the same curves but at lower temperature
For fixed q, πµ− is only non-zero for µ < ~vF q/2 and since
R1 kB T /t = 0.001 which agrees well with the Dirac cone
−
0 dxG< (x) = 4/3, the limit q → 0 leads to the well approximation now also at low energies.
known delta function for the diamagnetic susceptibility
When q is in x-direction which was chosen to be the
of graphene:
high symmetry axis which connects the Γ- and M -point
g 2 2 of the Brillouin zone, there is good agreement with the
χM = −µ0 e vF δ(µ) (44)
6π result coming from the Dirac cone approximation (except
for the deviations in π ⊥ due to temperature, mentioned
before). When q is in y-direction, we observe a peak
B. Numerical Results splitting around the resonant energy ~ω = ~vF q (see in-
set on the left hand side). There are thus lattice effects
We shall now compare the analytical results of the lin- which show up even in the regime where the Dirac cone
earized, isotropic Dirac model with the numerical results approximation and where the system should be isotropi-
obtained from the hexagonal lattice. In Fig. 1, we show cally invariant.
 6
|q|a=0.1 , µ=0.05t |q|a=0.1 , µ=0.05t

0.25 qx=q,qy=0,i=x
0.04 onant energy ~ω = ~vF q. This is reminiscent to the
qx=0,qy=q,i=y qx=0,qy=q,i=x fact that also the polarizability is not well described by
the Dirac approximation for these energies.21 The scalar
||
π (Dirac) qx=q,qy=0,i=y
0.2 _|_
0.03 π (Dirac)
relation between the conductivity and the polarizability
Im π (qx,qy,ω)

Im π (qx,qy,ω)
0.15
0.3
which makes use of the fact that there is a parallel compo-
0.2 0.02 nent does thus not hold for the lattice model. This might
be important for first principle studies which make use
i,i

i,i
0.1
0.1
of this relation.
0.01
0
0.05 0.14 0.16
3B

ω
0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
ω/t ω/t 2B
2µ

FIG. 2: The imaginary part of the current-current correla- 2A

tion function Imπ i,i (qx , qy , ω) as function of the energy ω 1B
at kB T /t = 0.01 for different directions with |q|a = 0.1.
Left hand side: The wave vector q is parallel to the cur- 1A 3A
rent. Right hand side: The wave vector q is perpendicular to
2µ
the current.The results obtained from the Dirac-cone approx- q
imation π k (q, ω) and π ⊥ (q, ω) are also shown, respectively
(dashed lines). Inset: Energy region around the resonant en-
ergy ω = vF q. Right hand side: Also the curves for lower FIG. 3: Display of the different regions characterizing the
temperature kB T /t = 0.001 are shown (dotted lines). current-current correlation function given in the appendix.
The regions are limited by straight lines ω = q (solid), ω =
q − 2µ (dashed) and ω = 2µ − q (dotted) where we set ~ =
vF = 1.
V. SUMMARY AND CONCLUSIONS

VI. ACKNOWLEDGMENTS
We have discussed the dynamical current-current cor-
relation function of the hexagonal lattice and of graphene
modeled by the linearized Dirac model. To define a local This work has been supported by grants
current operator, we introduced a continuum-replica of PTDC/FIS/101434/2008 and FIS2010-21883-C02-02.
the original lattice model. The resulting paramagnetic
current operator obeys the continuity equation with re-
spect to the density operator defined on the original lat- VII. APPENDIX
tice. The diamagnetic response is non-local.
We then gave explicit expressions of the current- Here, we shall present the real and imaginary part of
current correlation function for the honeycomb lattice π ± in terms of the three real dimensionless functions
and defined the longitudinal and transverse component 
g ~ω   v q 2 ∓ 12
in case of the rotationally invariant Dirac model. For the ±
f (q, ω) = 1−
F  , (45)
Dirac model, explicit analytical expressions were given 16π t  ω 
where the results for the longitudinal component can be p
−1
G± 2
> (x) = x x − 1 ∓ cosh (x) , x > 1 ,
obtained via the continuity equation from the density- p
density correlation function, as was discussed in detail. ± −1
G< (x) = ±x 1 − x2 − cos (x) , |x| < 1 .
In the last part of this paper, we showed that in the
honeycomb lattice, the longitudinal and transverse com- For the imaginary part, the additional terms at finite
ponent cannot be defined for energies around the res- doping then read in the language of Fig. 3:

 ± 2µ−~ω 2µ+~ω

 G> ( ~vF q ) − G±
> ( ~vF q ) , 1A


 π ± 2µ+~ω , 1 B



Im∆πµ (q, ω) = −f (q, ω) × −G> ( ~vF q )
± ± , 2 A (46)
 −G±


~ω−2µ
< ( ~vF q ) , 2 B

0 , 3 A



0 , 3 B
 7

For the real part, we get in the language of Fig. 3:

π , 1A




2µ−~ω ± 2µ+~ω
−G±

> ( ~vF q ) + G> ( ~vF q ) , 1 B





 −G± ~ω−2µ

gµ ω 2 < ( ~vF q ) , 2A

Re∆πµ± (q, ω) = ± ∓ f ± (q, ω) × 2µ+~ω
(47)
2πt (vF q)2 
 G±
> ( ~vF q ) , 2B


± ~ω−2µ ± 2µ+~ω

−G< ( ~vF q ) + G< ( ~vF q ) , 3 A





 ± 2µ+~ω
G> ( ~vF q ) − G± ~ω−2µ

> ( ~vF q ) , 3B

Since this agrees with the complex expression given in Eq. (39).
For more details, see Ref.8 and Ref.32 .
G±

> (x) ,x > 1
G± (x) = (48)
±iG±< (x) , |x| < 1

1 22
K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. T. Tudorovskiy and S. A. Mikhailov, Phys. Rev. B 82,
Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, 073411 (2010).
23
Science 306, 666 (2004). M. van Schilfgaarde and M. I. Katsnelson,
2
A. K. Geim, Science 324, 1530 (2009). arXiv:1006.2426.
3 24
A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. R. A. Muniz, H. P. Dahal, A. V. Balatsky, and S. Haas,
Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 Phys. Rev. B 82, 081411(R) (2010).
25
(2009). D. Gazit, Phys. Rev. B 79, 113411 (2009).
4 26
S. Das Sarma, Shaffique Adam, E. H. Hwang, and Enrico J. F. Dobson, A. White, and A. Rubio, Phys. Rev. Lett.
Rossi, arXiv:1003.4731. 96, 073201 (2006); G. Gómez-Santos, Phys. Rev. B 80,
5
N. M. R. Peres, Rev. Mod. Phys 82, 2673 (2010). 245424 (2009).
6 27
S. L. Adler, Phys. Rev. 126, 413 (1962). K. F. Allison, D. Borka, I. Radovic, L. Hadzievski, and Z.
7
Note that the functional form of the band-overlap function L. Miskovic, Phys. Rev. B 80, 195405 (2009).
28
depends on the phase of the structure factor. J. W. Clure, Phys. Rev. 104,666 (1956).
8 29
B. Wunsch, T. Stauber, F. Sols, and F. Guinea, New J. M. Koshino, Y. Arimura, and T. Ando, Phys. Rev. Lett.
Phys. 8, 318 (2006). 102, 177203 (2009).
9 30
E. H. Hwang and S. Das Sarma, Phys. Rev. B 75, 205418 M. Sepioni, S. Rablen, R. R. Nair, J. Narayanan, F.
(2007). Tuna, R. Winpenny, A. K. Geim, and I. V. Grigorieva,
10
F. Stern, Phys. Rev. Lett. 18, 546 (1967). arXiv:1007.0423.
11 31
Kenneth W.-K. Shung, Phys. Rev. B 34, 979 (1986); ibid. A. Principi, M. Polini, G. Vignale, and M. I. Katsnelson,
34, 1264 (1986). Phys. Rev. Lett. 104, 225503 (2010).
12 32
J. González, F. Guinea, and M. A. H. Vozmediano, Phys. A. Principi, M. Polini, and G. Vignale, Phys. Rev. B 80,
Rev. B 59, R2474 (1999). 075418 (2009).
13 33
T. Ando, J. Phys. Soc. Jpn. 75, 074716 (2006). X.-G. Wen, Quantum Field Theory of Many-Body Systems.
14
O. Vafek, Phys. Rev. Lett. 97, 266406 (2006). (Oxford U. P., 2004).
15 34
S. Gangadharaiah, A. M. Farid, E. G. Mishchenko, Phys. I. Paul and G. Kotliar, Phys. Rev. B 67, 115131 (2003).
35
Rev. Lett. 100, 166802 (2008) D. J. Scalapino, S. R. White, and S. C. Zhang, Phys. Rev.
16
M. R. Ramezanali, M. M. Vazifeh, R. Asgari, M. Polini, B 47, 7995 (1993).
36
and A. H. MacDonald, J. Phys. A : Math. Theor. 42, V. P. Gusynin, S. G. Sharapov, and J. P. Carbotte, Phys.
214015 (2009). Rev. B 75, 165407 (2007).
17 37
R. Roldan, J.-N. Fuchs, M. O. Goerbig, Phys. Rev. B 80, N. M. R. Peres and T. Stauber, Int. J. Mod. Phys. B 22,
085408 (2009). 2529 (2008); T. Stauber, N. M. R. Peres, and A. K. Geim,
18
P. K. Pyatkovskiy, J. Phys.: Condens. Matter 21, 025506 Phys. Rev. B 78, 085432 (2008).
38
(2009). C. Bena and G. Montambaux, New J. Phys. 11, 095003
19
T. G. Pedersen, A-P. Jauho, and K. Pedersen, Phys. Rev. (2009).
39
B 79, 113406 (2009). J. Sabio, J. Nilsson, and A. H. Castro Neto, Phys. Rev. B
20
A. Hill, S. A. Mikhailov, and K. Ziegler, Europhys. Lett. 78, 075410 (2008).
40
87, 27005 (2009). S. A. Mikhailov and K. Ziegler, Phys. Rev. Lett. 99, 016803
21
T. Stauber, J. Schliemann, and N. M. R. Peres, Phys. Rev. (2007).
B 81, 085409 (2010).