Codex @codex  0

A conical touching of two energy bands, with locally linear dispersion relation. The nearest-neighbour model of graphene supplies an explicit example.

Graphene is a single layer of carbon chemical atoms arranged in a honeycomb lattice. A nearby gate can select predominantly Electron or electron hole conduction, as demonstrated by the graphene exfoliation and field-effect experiments.^[ref]

A useful question is how geometry changes motion. The nearest-neighbour model of graphene starts with three equivalent bonds from each site and two sites per primitive cell. Interference between those three hopping paths makes two energy bands meet; near their meeting, energy grows linearly with distance in wave vector. The resulting Dirac cones explain why an atomic sheet can imitate one aspect of a relativistic particle while remaining an ordinary material with contacts, defects and boundaries.

F. Duncan M. Haldane's 1988 tight-binding model adds complex second-neighbour hopping and opposite sublattice offsets to a honeycomb lattice. Directed phases break time-reversal symmetry while each primitive cell has zero net magnetic flux. Smaller loops can enclose magnetic flux; ordinary Bloch states remain available.^[ref]

Extend the nearest-neighbour model of graphene, ignoring physical spin and interactions. With bond length $a$, $t_1>0$, $t_2\ge 0$ and ordered sublattice basis $(A,B)$, chooseThe following Bloch Hamiltonian defines $\varphi$; reversing directed second-neighbour hopping sends $\varphi \mapsto -\varphi$:The Pauli matrices act on sublattice amplitudes, as in sublattice pseudospin in graphene. The identity matrix term $d_{0}I$ shifts both energy eigenvalues equally at each momentum; it changes neither the eigenvectors nor the direct separation $E_{+}-E_{-}$. A genuine band insulator additionally needs ${\min }_{\mathbf{k}}{E}_{+}>{\max }_{\mathbf{k}}{E}_{-}$. We use $t_2/t_1=0.08$ in the figures, within the original paper's sufficient nonoverlap restriction $|t_2/t_1|<1/3$.^[ref]

Energy and momentum measurement. In angle-resolved photoemission spectroscopy, monochromatic photons illuminate a clean surface in vacuum, and an electron energy analyzer measures the outgoing kinetic energies and emission directions. Measuring both gives the occupied electronic band structure, rather than only a momentum-integrated distribution of removal energies.

Choose a common energy reference for initial and final states. Let $E_i$ be the initial electron energy, $E_F$ the Fermi energy, and $E_{\mathrm{vac}}=E_F+\varphi$ the vacuum level immediately outside the sample. The binding energy is $E_B=E_F-E_i\ge 0$. For a direct emission event without additional losses, conservation of energy givesHere the work function $\varphi$ is the energy from the Fermi level to the local vacuum. The initial and final energies are total single-electron energies on the same scale, while $E_{\mathrm{kin}}$ is measured above vacuum. The diagram illustrates each interval independently.

Neglecting ultraviolet photon momentum, photoemission momentum conversion giveswhere $\theta$ is measured from the surface normal. Parallel crystal momentum is conserved modulo a surface reciprocal lattice vector; the surface breaks perpendicular translation symmetry. A three-dimensional sample therefore generally needs a final-state model to infer perpendicular momentum. The stated two-dimensional sheet avoids that ambiguity. The kinetic energy used in this formula is the sample-vacuum value; if the analyzer and sample have different work functions, their contact potential must be included when converting the analyzer's reading, as in photoelectron binding energy reference.

Spectral information. Define $ϵ=E_i-E_F=-E_B$ and the electron spectral function by $A(\mathbf{k},ϵ)=-\mathrm{Im}{G}^R(\mathbf{k},ϵ)/\pi$. In the sudden approximation, the basic photoemission intensity and spectral function relation isbefore instrumental broadening and background are included. Here $f$ is the Fermi-Dirac distribution, and $\mathcal{M}$ is the transition matrix element. The electron removal spectral function contains the occupied part of the spectrum. In a noninteracting band it has a delta peak at the band energy; interactions shift and broaden peaks and can add incoherent weight. Thus peak trajectories reveal dispersion, while line shapes can constrain lifetimes after resolution effects are accounted for. Dividing by known matrix and occupation factors can estimate $A$ where those factors are appreciable; intensity alone is not identical to $A$, and a missing peak can reflect a matrix-element suppression. States well above $E_F$ are not accessible to ordinary low-temperature photoemission.

Two independent band-structure probes. First, the de Haas-van Alphen effect measures oscillatory magnetization as Landau levels cross the Fermi level when a magnetic field changes. The inverse-field frequency iswhere $S_{\mathrm{ext}}$ is an extremal Fermi surface cross-sectional area perpendicular to the field. Changing field orientation maps its shape. This reveals the momentum-space geometry of occupied bands; the related Shubnikov-de Haas effect measures oscillatory resistance.

Second, absorption or reflectance spectroscopy detects interband transitions. Photon energies at absorption edges and peaks constrain the band gap and interband energy differences. For a direct band gap, negligible photon momentum selects approximately vertical transitions in the band diagram, weighted by the joint density of states and transition matrix elements. An indirect transition needs a phonon to carry the extra crystal momentum. These measurements provide evidence beyond a free-electron parabola, although an optical spectrum alone does not reconstruct every band's dispersion.

Reading the graphite spectra. For each direction, follow a moving peak through successive angles and convert each peak's own kinetic energy to $k_{\parallel }$; angle is not itself proportional to momentum over this wide range. The prominent dispersing valence feature starts at roughly $7$–$9\mathrm{eV}$ binding energy near $\Gamma$ and moves towards lower binding energy as $k_{\parallel }$ increases. Along $\Gamma K$ it approaches the Fermi level at the corner, then turns back to larger binding energy on continuing beyond the corner. Along $\Gamma M$ it reaches a finite minimum binding energy of roughly $3$–$4\mathrm{eV}$ near the edge midpoint and turns back; it does not reach the Fermi level there. Additional deeper peaks describe other occupied valence branches, with visibility varying across the spectra.

A useful qualitative description is the upper occupied $\pi$ band of the nearest-neighbour model of graphene: $E_{\pi }=-t|f(\mathbf{k})|$. It rises from $-3t$ at $\Gamma$ to zero at $K$, but only to $-t$ at $M$. The schematic curve in the figure captures those two different behaviours. It is not a full multi-orbital assignment or a quantitative fit to graphite, whose deeper valence features are also visible in the supplied spectra.

Angle needed to reach the corner. The corner's parallel wavevector isUse the near-zero binding energy of the upper valence band at that point. With the sample work function interpreted as the stated $3.8\mathrm{eV}$ separation below vacuum,HenceThe energy and angle readings in the PDF are approximate. If a peak at the corner were instead assigned a small binding energy $E_B$, the estimate becomesFor $E_B=0$–$1\mathrm{eV}$ this gives about $57$–$60^{\circ }$, consistent with the near-Fermi feature in that angular range. Substituting the full photon energy for the kinetic energy would omit the work-function cost and underestimate the required angle.

The only zeros of $f$, modulo the reciprocal lattice, are the two Dirac valleys in graphene,At these momenta ${\sum }_j\sin ({\mathbf{K}}_{\tau }\cdot {\mathbf{b}}_j)=\tau 3\sqrt{3}/2$. Thus the two Dirac mass coefficients, measured in energy units, are $m_{\tau }=M-\tau \mu$, with $\mu =3\sqrt{3}{t}_2\sin \varphi$. For $\mathbf{k}={\mathbf{K}}_{\tau }+\mathbf{q}$,The two Dirac cones have opposite orientation, encoded by $\tau$. A scalar sublattice offset $M$ gives equal-sign Dirac masses; imaginary second-neighbour hopping at $M=0$ gives opposite-sign Dirac masses. Both mechanisms open valley gapsAn observed band gap alone therefore cannot establish a Chern insulator. In particular, the gapless nearest-neighbour model of graphene and its sublattice-mass deformation should not be confused with the topologically nontrivial Haldane model.

A band gap closes when all three components of $\mathbf{d}$ vanish. Since $d_x=d_y=0$ only at the two valleys, the phase boundaries are exactly $M=\pm 3\sqrt{3}{t}_2\sin \varphi$. Usually one valley becomes massless; at $M=0$, $\sin \varphi =0$, both do. The occupied-band Chern number is undefined at every such touching, even if a numerical momentum grid misses its location.