You are on page 1of 7

PHY 556 Spring 2007

Note 2
Philip B. Allen
Electrons in graphene – massless Dirac electrons and Berry phase
Graphene is a single (infinite, 2d) sheet of carbon atoms in the graphitic
honeycomb lattice. On the left is a fragment of the lattice showing a primitive unit cell,

with primitive translation vectors a and b, and corresponding primitive vectors G1, G2 of
the reciprocal lattice. On the right is the central part of the reciprocal lattice and the first
Brillouin zone. The corners of the Brillouin zone are the points Ki given by
r
r
r
r
r
r
r
r
r
K1 = G2 − G1 / 3, K 2 = 2G2 + G1 / 3, K 3 = 2G1 + G2 / 3 , etc. Only two are inequivalent.
r
r
r
Notice for example that K 3 = K1 + G1 .
Because the lattice is 2-dimensional, all translations commute with reflection in
the plane of the lattice, so all electron (or vibrational) eigenstates can be chosen to be
either even or odd under this reflection. For this reason, the single-particle electron states
are rigorously separated into two classes, called “σ” and “π,” the even σ states being
derived from carbon s and px, py orbitals, and the odd π states being derived from carbon
pz orbitals. These latter are cylindrically symmetric in the x-y plane, lie near the Fermi
level (half-filled) and are the electrically active states of interest in low energy physics.
A useful picture of electron behavior can be derived by using Hückel theory to
are shown below in
look at the π electrons (pz orbital-derived states.) The two sublattices
r
r
r
different colors, with the “A” sublattice at vectors R = n1a + n2 b , and the “B” sublattice
r r
r
r r
at vectors R + τ , with τ = −(a + b ) / 3 . The Hamiltonian in nearest neighbor Hückel
theory is
r r r
r r r r
r r r r
H = −t ∑
R
R
+
τ
+
R
R
+
a
+
τ
+
R
R + b + τ + h.c.
r

(

)

(

R

)

{

(

)

}

r r r
r
where R is a pz (π) state on an A sublattice atom at site R , R + τ is a similar state on a

B sublattice atom, and t is the “hopping integral” (positive) from a state to an adjacent
similar state. The abbreviation “h.c.” means Hermitean conjugate. The graphite lattice is
“bipartite.” The hopping matrix element couples states on the A sublattice only to states
on the B sublattice, and vice versa. We now transform to the basis of Bloch waves,
r r
r
1
ik ⋅ R r
kA =
e
∑ R
N Rr

and off-diagonal elements r r 1 kA H kB = N ( ) r r r r r r r r r r r ik ⋅ ( R ' − R ) ik ⋅ a ik ⋅b e R H R + τ = − t 1 + e + e = e ( k ). Everywhere else in k-space. K2. K1. r Therefore. The two bands (called π and π*) lie symmetrically above and below the Fermi energy. r 1 + e + i 2π / 3 + e −i 2π / 3 = 0 . from Saito and . the energy e(k ) → 0 at the zone corners. r r r r with diagonal elements k A H k A and k B H k B both zero. ∑ rr RR ' r r The single particle Bloch energies are thus ε (k ) = ± e(k ) . e( k ) / t = 1 + 4 cos⎜⎜ ⎜ ⎟ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠ r r r r iφ ( kr ) . Let us write e(k ) = e(k ) e with the Hamiltonian matrix being r r r ⎛⎜ 0 e(k ) e +iφ ( k ) ⎞⎟ r r Hˆ (k ) = ⎜ e(k ) e −iφ ( k ) ⎟ 0 ⎝ ⎠ Then the eigenvectors ψk are 1 ⎛ e + iφ / 2 ⎞ 1 ⎛ e + iφ / 2 ⎞ ⎟.r 1 kB = N ∑r e r r ik ⋅ R r r R +τ . where ⎛ 3k x a ⎞ ⎛ k y a ⎞ ⎛k a⎞ ⎟ cos⎜ ⎟⎟ + 4 cos 2 ⎜⎜ y ⎟⎟ . The bands are plotted below to the left. The Schrödinger equation is then H (k ) ψ k = ε (k ) ψ k . r r e(k ) ≠ 0 and the splitting of the two graphene π-bands is 2 e(k ) . R This transformation block-diagonalizes the 1-electron Hamiltonian into 2 x 2 sub-blocks. ⎜ −iφ / 2 ⎟ and k + = ⎜ k− = ⎟ 2 ⎜⎝ − e 2 ⎜⎝ e −iφ / 2 ⎟⎠ ⎠ r r r r Note that the phase factors 1 + e ik ⋅a + e ik ⋅b become the 1/3rd roots of unity. when k lies on a corner point of the Brillouin zone. A more elegant two-dimensional presentation. E=0.

2006 The wavefunctions of graphene have attracted a lot of interest. The dispersion seen for the photohole fits amazingly well to the Hückel formula. For example. eds. Dresselhaus. the degeneracy at the K points persists. the Fermi level is between the two symmetrical bands. A beautiful photoemission experiment by Bostwick et.. the figure below from Saito and Kataura shows that even though a more accurate theory does not have exact particle-hole symmetry. al. 2001) is shown below to the right.. the . and is not just an accidental result of the Hückel model. K2. If you move adiabatically in k-space around the K point. from Bostwick et.Kataura (Dresselhaus.10. Let us consider a circular path in r k-space r r around the point K1 or equivalently. Therefore. Nature Physics published online 12. It follows from symmetry. is also shown. A perfect graphene sheet has one electron per carbon in the π-levels. The energy is a linear function of δk = k − K1 . The (ππ*) degeneracy at isolated points K at the Fermi energy is general to the one electron description of graphene. Fitted linewidth versus energy at various dopings n (unit = 1013 cm-2) E(k) for photohole. with zero excitation energy needed to excite an electron from just below the Fermi energy to just above at the K-point. Springer. Carbon Nanotubes. and Avouris. al.

Interestingly. they change phase by π when θ increases by r 2π. r r r r r r r r r ⎛ 1 r r ⎛ 1 e( k ) 3⎞ 3⎞ ⎟ 1 + iδk ⋅ b ⎟ 1 + iδk ⋅ a + ⎜ − + i − = 1 + e i (K +δk )⋅a + e i (K +δk )⋅b ≈ 1 + ⎜⎜ − − i ⎜ 2 t 2 ⎟⎠ 2 ⎟⎠ ⎝ ⎝ 2 1 r r r 3 r r r 3 r − iπ / 2 − iθ ≈ − iδk ⋅ a + b + δk ⋅ a − b = δk ae e 2 2 2 3a 3 r − iπ / 2 −iθ ≈− i δk x − iδk y = δk ae e 2 2 ( ( ) ( [ ] ) ( ) ) r r where θ is the angle from the x-axis in the kx. 2005. The Berry phase of π has the consequence that the quantized cyclotron orbits (Landau levels) will require half-integral numbers of wavelengths to give single-valued wavefunctions. Thus e(k ) equals 3 δk at . and e(k ) has phase φ=-θ-π/2. In Nature vol. Kim et al.” shown below on the left. through zero. Here the formulas a + b = 3axˆ r r r r r and a − b = − ayˆ were used. and ) under a dc magnetic field Bz . This can be r seen by expanding the Hamiltonian matrix to first order in δk .ky plane. and the special points Ki are called the “Dirac points.) and the Columbia group (P. orbiting around K.) reporting the quantum Hall effect in graphene.wavefunctions acquire a “Berry phase” eiπ=-1 when completing a circuit. there were back-to-back papers from the Manchester-Chernogolovka-Nijmegen group (Geim et al. to the energy will be quantized as (n + 1 / 2)hωc where n is an integer. The samples were monolayers of carbon lying on a thin silicon oxide layer on top of a doped silicon substrate which served as a gate electrode. to n-type. Both RH and electrical conductivity σ extrapolate to zero when the Fermi level passes through the Dirac points. that is. will form a cyclotron orbit. The middle and right figures above are from the Geim paper showing how gate voltage dopes graphene p-type (as shown by the positive Hall coefficient RH. it will lie near a K point. 438. σ is actually . The code words “massless Dirac spectrum” are used to refer to the linear ε versus k relation. Since the wavefunctions have phases ±φ/2. If a perfect graphene sheet is given an extra electron or hole. when the k -vector goes once around a loop surrounding a K point. 10 Nov.

The K points in the Brillouin zone are special. and seems not to actually go to zero. not on a “atom” site. Consider the plane wave at the point K1 shown in the r first figure.g. boundaries) r r r r r r where k + G = k + G ' for some pair of reciprocal lattice vectors G and G ' . This has a kinetic energy ε=ε(K1) degenerate with the kinetic [( )] ) ( ( ( ( ) ) ( ) () [ ( )] ( ( () ) ) ( ( ( ) ) ) ( ) ) .pinned at a minimum value near 4e2/h. Both groups unambiguously see half-integer quantization. The weak potential shifts energy only to second order. depending r only on r = r = x 2 + y 2 .Al)As inversion layer with a patterned top electrode in honeycomb morphology. and the Hamiltonian matrix has r disconnected sub-blocks H k of the form r r2 2 r r r r r k h +G H k Gr . Then the single-particle states can be expanded in plane r r r waves k = 1 / V exp ik ⋅ r . The origin is now taken in the center of a r r r hexagon. except at symmetric places in the Brillouin zone (e. and the “atoms” are then at ± τ = ± a + b / 3 . G ' + Vtot G − G ' 2m where the off-diagonal terms Vtot are products of single-site pseudo-potential form factors r r r r V G times structure factors S G = cos G ⋅ τ . suppose we had a two dimensional electron gas (2-deg) with a weak periodic potential applied The applied potential has honeycomb symmetry. Bloch’s theorem applies. I am assuming that the applied potential is the sum of localized cylindrical terms. The band structure is in zeroth r r r r2 approximation just E ≡ ε 0 = ε k + G = h 2 k + G / 2m . Nearly free electron method As an alternative model for graphene. The quantum Hall effect signals are plotted below. we could imagine a GaAs/(Ga. exactly as predicted by two theoretical groups shortly before the measurements.Gr ' = δ G. 1 = K1 . The Hamiltonian is r r r r r r p2 H= +∑ V r − R − τ + V r − R +τ 2m Rr The geometry was shown in the first figure. As a more faithful realization.

The dominant part of the matrix H k () is −V / 2 −V / 2⎞ ⎛ ε r ⎜ 0 ⎟ H k = ⎜ −V / 2 ε0 −V / 2⎟ . Our concern now is with the eigenvalues and eigenvectors that evolve from the E’ doublet (the second and third rows and columns of the matrix. first transform to r the basis of eigenvectors with δk = 0 . 2 . we have to make a Taylor expansion for r r H K + δk . The matrix becomes − V / 2 −V / 2⎞ ⎛ ε r ⎜ 1 ⎟ H k = ⎜−V / 2 ε2 −V / 2⎟ ⎜−V / 2 −V / 2 ε 3 ⎟⎠ ⎝ r2 r where ε1 is h 2 K1 + δk / 2m . and () ( ) the factor -1/2 is the corresponding structure factor S. and with r r r 3 = K1 − G2 which lies at the special point K5.) We need only solve this 2 x 2 submatrix. To see that. namely 0 . and thus generate a representation of. The elements of this matrix are r2 (ε 1 + ε 2 + ε 3 ) / 3 = ε 0 + h 2 δk / 2m () ( (ε 1 ) ) ( ) ( + ε 2 e iφ + ε 3e −iφ / 3 = h 2 / m (2π / 3a )(iδk x + δk y ) Therefore the relevant 2 x 2 submatrix is ) . and similar for ε2 and ε3 . 3 do not close under the operations of D6h because the rotation by 2π/6 takes K1 into the inequivalent point K2. since the influence of the other state is second order.r r energy of 2 = K1 + G1 which lies at the special point K3 in the first figure. But these three functions do close under. and is a Dirac point. The three basis functions 1 . and − φ . To solve this. The new matrix is ( ) () ( ) small small ⎞ ⎛ (ε + ε 2 + ε 3 ) / 3 − V r ⎜ 1 ⎟ iφ − iφ (ε 1 + ε 2 + ε 3 ) / 3 + V / 2 ε 1 + ε 2e + ε 3e / 3 ⎟ H k =⎜ small ⎜ small ε 1 + ε 2 e −iφ + ε 3e iφ / 3 (ε 1 + ε 2 + ε 3 ) / 3 + V / 2 ⎟⎠ ⎝ r where “small” means that the energy shift will be second order in δk . are basis vectors ( ) for an E’ doublet of eigenvalue E1=ε+V/2. ⎜ −V / 2 −V / 2 ε 0 ⎟⎠ ⎝ r r r The notation V is the form factor V G for the relevant vector difference K i − K j . Using φ to denote the angle 2π/3. φ .) It is easy to see that the vector 0 = ( 1 + 2 + 3 ) / 3 has A1 ' symmetry and is an eigenvector with eigenvalue E0=ε0-V. the vectors φ = 1 + e iφ 2 + e − iφ 3 / 3 and − φ . Let us use a little group theory. the complex conjugate of φ . This doublet lies lower in energy if V is negative. the subgroup D3h (known as the “little group” of the wavevector K. The point group of graphene is D6h.

the bands have to evolve back to free electron form. 166402 (2004)) which comes close to addressing this issue. they change sign when the angle θ = tan −1 (δk y / δk x ) goes once around the Dirac point. This is a consequence of the choice of basis functions. Kopelevich (Phys. There is a nice recent paper “Phase analysis of quantum oscillations in graphite. () and the eigenvectors contain exp(±iθ/2). Rev. The question arises whether similar effects should show up in graphite. r2 r ε + h 2 δk / 2m + V / 2 ± (h 2 / m )(2π / 3a )δk . that is.” by I.r2 ⎞ h 2 δk r ⎛⎜ e −i (θ −π / 2 ) ⎞ V ⎟ ) h 2 2π r ⎛ 0 ⎟. 93. At larger separations δk . I am not sure the matter is yet laid to rest. The underlying Hamiltonian is real. but the matrix representations we looked at so far were complex Hermitean. A different basis could have given a real-symmetric Hamiltonian. r This can only be true when the conical energy increments ± (h 2 / m )(2π / 3a )δk are less r than the increments ±V/2 caused by the external potential. Lett. . Luk’yanchuk and Y. It is interesting that the conical dispersion around the Dirac points does not depend on the size of the applied potential. A. δk ⎜⎜ i (θ −π / 2 ) + ⎟1 + H k = ⎜ε + 2m 2⎟ m 3a 0 ⎟⎠ ⎜ ⎝e ⎝ ⎠ The energy eigenvalues are massless Dirac spectra.