Edge and Surface States

An edge or surface state has wavefunction weight concentrated near a boundary rather than throughout the bulk. In a clean translation-invariant geometry it can disperse along that boundary; disorder or additional confinement can localize it along the boundary as well. In a topological phase, the bulk can obstruct the removal of all such boundary structure under a specified set of local perturbations. The obstruction may enforce net chiral propagation, an odd number of Kramers pairs, an odd number of surface Dirac cones, a chiral Majorana channel, or a Fermi arc that joins projected bulk nodes.

The word protected is easy to overread. It does not mean that every boundary eigenstate has a fixed dispersion, that disorder causes no scattering, that transport is perfectly quantized in every geometry, or that the surface remains gapless after its protecting symmetry is broken. Topology constrains a stable aggregate: net chirality, crossing parity, anomaly, or connectivity to the bulk.

The following models explain how chirality, time reversal, interactions and finite sample thickness affect boundary propagation. Their measured signatures depend on contacts, energy resolution and coupling to bulk states. Bulk–Boundary Correspondence develops the general relation to relative bulk indices; the emphasis here is on the states and what happens when they scatter.

Required background. Topology in Quantum Matter supplies the phase-equivalence and protection distinctions needed before interpreting a boundary feature.

Helpful background. Integer Quantum Hall Effect supplies the chiral branch; Topological Insulators and Time Reversal for Spin-1/2 Particles supply the helical branch; and Conductance Quantization supplies the channel-to-transport bridge.

Boundary spectra and protecting symmetries

An edge is the one-dimensional boundary of a two-dimensional bulk. A surface is the two-dimensional boundary of a three-dimensional bulk. An interface between phases $A$ and $B$ is treated as a boundary after choosing an orientation; vacuum is a topologically trivial phase for the electronic examples below.

Write the coordinate normal to the boundary as $x_\perp$ and conserved boundary momentum as $\mathbf k_\parallel$. A localized state has asymptotic envelope

$$\left| \psi_{\mathbf k_\parallel}(x_\perp) \right| \sim \exp\!\left( -\frac{\lvert x_\perp\rvert}{\xi} \right),$$

where $\xi$ is a penetration depth. In a clean half-space, an isolated boundary eigenstate is normally found in a projected bulk gap. Define the projected bulk spectrum $\mathcal B(\mathbf k_\parallel)$ by varying $k_\perp$ over all bulk bands. Then

$$E\notin\mathcal B(\mathbf k_\parallel).$$

If the boundary energy overlaps a projected bulk continuum, generic coupling turns it into a surface resonance with leakage into bulk states. Exceptional bound states inside a continuum need an additional decoupling mechanism; overlap alone is not an absolute no-go theorem.

For electronic time reversal,

$$\Theta = i\sigma_y\mathcal K, \qquad \Theta^2=-1,$$

where $\mathcal K$ is complex conjugation in the chosen spinor basis. Spin–orbit coupling generally means that $\boldsymbol\sigma$ labels a Kramers pseudospin rather than a conserved physical spin.

Velocities and edge orientations fix signs. Unless a sign is central, conductance magnitudes are quoted. A gap means a spectral gap in the relevant boundary theory, not merely suppressed density of states from matrix elements or disorder broadening.

Boundaries of Topological Phases

Opening a boundary removes translation symmetry in the normal direction. A slab Hamiltonian retains $\mathbf k_\parallel$ and becomes a finite matrix in layer or orbital indices:

$$H_{\mathrm{slab}}(\mathbf k_\parallel) \lvert u_{\alpha\mathbf k_\parallel}\rangle = E_\alpha(\mathbf k_\parallel) \lvert u_{\alpha\mathbf k_\parallel}\rangle.$$

A boundary branch is diagnosed by both its dispersion and its spatial weight. If $P_s$ projects onto the outer few layers, define

$$w_\alpha^{(s)} = \langle u_{\alpha\mathbf k_\parallel} \rvert P_s \lvert u_{\alpha\mathbf k_\parallel} \rangle.$$

Large $w_\alpha^{(s)}$ distinguishes a surface-localized state from a bulk subband. In Green-function language, the surface spectral function is

$$A_s(\mathbf k_\parallel,\omega) = -\frac{1}{\pi} \operatorname{Im} \operatorname{Tr} \left[ P_s G^R(\mathbf k_\parallel,\omega) \right].$$

This is the quantity most directly compared with momentum-resolved surface spectroscopy after matrix elements and resolution are included.

Ordinary boundary states

A state can be localized at a surface without any bulk topology. Mechanisms include:

• a terminated periodic potential, producing Tamm- or Shockley-type states;
• dangling bonds and unsatisfied surface valence;
• electrostatic band bending and accumulation layers;
• chemical adsorption or oxidation;
• reconstruction that changes the surface unit cell;
• confinement in a thin film or quantum well;
• an interface potential well between two ordinary materials.

These states are physically real and can dominate experiments. They are not topological merely because they are two dimensional, metallic, spin split, or located inside part of a projected gap.

What the bulk constrains

For a topological boundary, local surface changes can alter:

• the velocity and chemical potential;
• the detailed spin or orbital texture;
• the position and shape of crossings;
• the penetration depth;
• the number of additional trivial pairs;
• the distribution of spectral weight between surface and bulk;
• edge reconstruction and interaction parameters.

What cannot change continuously, while the bulk remains in the same phase and all protecting assumptions remain valid, is the stable boundary content. A chiral edge cannot lose its net directional imbalance. An odd helical pair cannot be removed by adding ordinary Kramers pairs. A strong topological-insulator surface cannot become a featureless, nondegenerate, symmetry-preserving band insulator by surface terms alone.

The boundary may nevertheless evade a simple gapless-band description by:

1. breaking the protecting symmetry;
2. coupling to a complementary boundary sector;
3. hybridizing opposite sides of a finite sample;
4. closing the bulk gap;
5. developing intrinsic topological order when interactions permit it.

That list is the operational meaning of an anomalous boundary: there is an obstruction only after the assumptions are stated.

Chiral Edge Modes

A minimal charged chiral channel along coordinate $y$ has Hamiltonian

$$H_{\mathrm{ch}} = \int dy\, \psi^\dagger(y) \left( -i\hbar v\,\partial_y -\mu \right) \psi(y),$$

with dispersion

$$E(k) = \hbar vk-\mu.$$

Its group velocity has one sign:

$$v_g = \frac{1}{\hbar} \frac{\partial E}{\partial k} = v.$$

In this fixed-velocity model, a scalar potential changes the phase, not $v$. More general boundary terms can change the dispersion. Neither produces elastic reflection within an isolated one-branch theory because no on-shell state moves in the opposite direction on that edge.

Channel conductance

One ideal charged channel between reservoirs has conductance magnitude

$$\left| G \right| = \frac{q^2}{h}.$$

The result is independent of $v$: a slower channel carries less current per occupied state but has a proportionally larger one-dimensional density of states. Conductance Quantization derives this velocity cancellation and explains contact and transmission corrections.

Scalar disorder as a phase shift

The absence of a counterpropagating state can be seen directly. Add a static scalar potential to one chiral branch:

$$h = -i\hbar v\,\partial_y + V(y).$$

Define

$$\psi(y)=e^{-i\vartheta(y)}\chi(y),$$ $$\partial_y\vartheta=\frac{V(y)}{\hbar v}.$$

Then $h\psi=e^{-i\vartheta}(-i\hbar v\,\partial_y)\chi$. On an open edge, $V(y)$ changes only the accumulated phase. On a closed edge, it can change the boundary-condition twist and level positions, but it still supplies no reflected wave. This argument fails once another counterpropagating channel is coupled in.

For a noninteracting charge-conserving interface, an orientation can be chosen so that

$$N_R-N_L = \Delta C,$$

where $\Delta C$ is the Chern-number difference across the interface. Additional pairs can appear, but they change $N_R$ and $N_L$ equally and leave the net chirality fixed.

Why chirality is not perfect isolation

Real edges can contain:

• several co-propagating channels;
• reconstructed counterpropagating pairs;
• neutral modes;
• compressible strips from electrostatics;
• contacts that fail to equilibrate all channels;
• tunneling to the opposite edge in a narrow sample;
• bulk puddles or metallic parallel paths.

Chirality forbids Anderson localization of one isolated channel by static elastic disorder. It does not prevent energy relaxation, dephasing, interchannel equilibration, leakage into contacts, or backscattering through a remote counterpropagating sector.

Fractional and Majorana edges

An ideal unreconstructed Laughlin edge is a chiral boson theory. At its universal low-energy fixed point, the electron and quasiparticle scaling dimensions are fixed by the Laughlin integer; changing a short-range interaction changes the velocity, not those dimensions. Extra counterpropagating modes or long-range interactions can invalidate the simple power laws. Fractional Quantum Hall Effect develops the action and tunneling limits.

A chiral Majorana channel has a real fermion field and one half of a complex chiral fermion’s degrees of freedom. It contributes $|c_-|=1/2$ and transports heat without an independent conserved electrical charge. Topological Superconductors explains its normalization and thermal response. These examples show why counting boundary branches is not enough to assign an electrical conductance.

Helical Edge Modes

A helical edge contains counterpropagating partners related by time reversal. In the basis

$$\Psi = \begin{pmatrix} \psi_R\\ \psi_L \end{pmatrix},$$

a minimal theory is

$$H_0 = \int dy\, \Psi^\dagger \left( -i\hbar v\,\sigma_z\partial_y -\mu \right) \Psi.$$

The branches have dispersions

$$E_\pm(k) = \pm\hbar vk-\mu.$$

Time reversal exchanges them. At a time-reversal-invariant momentum, Kramers theorem forces a degeneracy.

The forbidden mass

A single-particle gap requires a matrix that anticommutes with $\sigma_z$:

$$H_m = \int dy\, \Psi^\dagger \left( m_x\sigma_x + m_y\sigma_y \right) \Psi.$$

For a spin-half Kramers pair,

$$\Theta\sigma_i\Theta^{-1} = -\sigma_i.$$

Therefore constant $m_x$ and $m_y$ break time reversal. A scalar potential $V(y)I$ preserves it and cannot open a clean single-particle gap at the Kramers crossing. A constant $m_z\sigma_z$ also fails to gap this linear model: it shifts the crossing to $k=-m_z/(\hbar v)$ even though it breaks time reversal. Breaking the protecting symmetry permits a mass; it does not make every perturbation a mass.

This statement concerns one anomalous Kramers pair. Two pairs can be mixed and gapped by time-reversal-symmetric couplings. The stable datum is the parity of the pair count, not the visibility of every crossing in a particular surface band plot.

Kramers backscattering

Let $\lvert u\rangle$ be one state and $\Theta\lvert u\rangle$ its counterpropagating partner. For a Hermitian perturbation $V$ satisfying

$$\Theta V\Theta^{-1} = V,$$

the direct matrix element is

$$M = \langle \Theta u \rvert V \lvert u \rangle.$$

Using antiunitarity, $[V,\Theta]=0$, and $\Theta^2=-1$,

$$\begin{aligned} M &= \langle \Theta(Vu) \rvert \Theta(\Theta u) \rangle\\ &= - \langle V\Theta u \rvert u \rangle\\ &= -M. \end{aligned}$$

Hence $M=0$. This proves that the direct matrix element vanishes. The corresponding elastic scattering restriction is stronger than a first-order result. Choose each outgoing channel as the time reverse of its incoming partner. For a time-reversal-invariant, flux-conserving single-particle scattering problem, the full scattering matrix satisfies $S^T=-S$ (Bardarson). Its reflection block is antisymmetric. With only one incoming channel on a given helical edge, that block is a scalar, so $r=-r=0$ to all orders. Inelastic processes, additional reservoirs and coupling to another edge change the scattering problem.

A finite magnetic barrier

The difference between scalar and magnetic scattering can be calculated exactly. Take the helical Hamiltonian with $v>0$, chemical potential zero and a constant transverse mass $m\sigma_x$ on $0<y<L$. Outside the interval the edge is massless. At $E=0$ the equation inside is

$$\partial_y\Psi=\frac{m}{\hbar v}\sigma_y\Psi.$$

Thus its transfer matrix is

$$\mathcal M=\cosh a\,I+\sinh a\,\sigma_y.$$

Here $a=mL/(\hbar v)$ is the dimensionless barrier strength. Equivalently,

$$\mathcal M=e^{a\sigma_y}.$$

Matching an incoming right mover and reflected left mover on the left to a transmitted right mover on the right gives

$$\begin{pmatrix}t\\0\end{pmatrix} =\mathcal M\begin{pmatrix}1\\r\end{pmatrix}.$$

Consequently

$$r=-i\tanh a,\qquad t=\operatorname{sech}a,$$ $$R=\tanh^2a,\qquad T=\operatorname{sech}^2a.$$

A long mass segment suppresses transmission exponentially without requiring a bulk transition. A scalar barrier alone instead gives $T=1$ in this isolated linear edge. These results assume elastic single-channel leads and do not include contacts that absorb particles or energy.

Helical does not mean conserved spin

In an idealized spin-conserving model one may label the branches $\uparrow$ and $\downarrow$. Generic spin–orbit coupling rotates the spin texture with momentum and orbital composition. The robust statement is Kramers locking:

$$\lvert u_{-\mathbf k}\rangle \propto \Theta \lvert u_{\mathbf k}\rangle,$$

not a momentum-independent eigenvalue of $S_z$.

This distinction matters experimentally. Spin-resolved spectroscopy can find reduced polarization because the measured electron spin is entangled with orbitals, the photoemission matrix element is selective, and the boundary state can hybridize with trivial bands. None of those observations alone destroys the $\mathbb Z_2$ boundary anomaly.

Interactions and inelastic processes

Time reversal forbids one elastic backscattering event between one Kramers pair. It does not forbid every many-body process. Allowed mechanisms can include:

• two-particle backscattering;
• inelastic backscattering through interactions and disorder;
• Kondo scattering from a magnetic impurity;
• coupling to nuclear spins or fluctuating moments;
• spontaneous time-reversal breaking at strong coupling;
• tunneling between opposite edges.

A helical Luttinger liquid is described by

$$H_{\mathrm{HLL}} = \int dy\,\mathcal H(y),$$ $$\mathcal H(y)=\frac{\hbar u}{2\pi} \left[ K \left( \partial_y\theta \right)^2 + \frac{1}{K} \left( \partial_y\phi \right)^2 \right].$$

Here $u$ is the collective velocity and $K=1$ is the noninteracting normalization. Use density fluctuation $\delta n=-\partial_y\phi/\pi$. With generic spin–orbit coupling, a symmetry-allowed two-particle operator has the form $e^{4i\phi}$ and scaling dimension $4K$. Its real part is an umklapp term when momentum conservation permits it. The relevance threshold depends on where the perturbation acts:

Two-particle process	Weak-coupling flow exponent	Relevant when
Uniform, commensurate	$2-4K$	$K<1/2$
One local impurity	$1-4K$	$K<1/4$
Static random amplitude	$3-8K$ for variance	$K<3/8$

The first exponent counts one space and one time integration; the impurity has only a time integration. Averaging static disorder produces two time integrations and two operators. The thresholds apply near the clean Gaussian fixed point with weak perturbations; marginal cases require higher-order flow. Exact spin conservation can forbid the operator, and away from commensurate filling a uniform umklapp oscillates rapidly. Strong uniform umklapp can gap the edge by spontaneous time-reversal breaking at zero temperature. A local impurity can cut transport without producing a uniform bulk edge gap (Wu, Bernevig and Zhang; Xu and Moore).

Luttinger Liquid develops the bosonized fields and scaling framework. The distinction here explains why “time reversal is present” is not a complete interacting transport model.

Ideal conductance and finite length

One ideal helical edge coupled without backscattering to ordinary Fermi-liquid reservoirs gives two-terminal conductance magnitude $e^2/h$. The lead-controlled value should not be confused with the response of an infinite isolated Luttinger liquid. A short two-terminal bar with two equivalent edges can approach $2e^2/h$. Deviations can come from contact resistance, bulk leakage, edge puddles, inelastic scattering, nonuniform gating, or insufficiently short length.

A useful experimental hierarchy is:

1. show conduction persists when the two-dimensional bulk is depleted;
2. demonstrate edge rather than area scaling;
3. test nonlocal response consistent with the perimeter network;
4. vary edge length and temperature;
5. test sensitivity to time-reversal breaking;
6. compare several device geometries and contacts.

A conductance value near $2e^2/h$ on one device is compatible with helical edges, not a self-contained proof.

Surface Dirac Cones

For a surface normal to $\hat{\mathbf z}$, the minimal strong-topological-insulator Hamiltonian is

$$H_D(\mathbf k) = \hbar v_D \left( k_x\sigma_y - k_y\sigma_x \right) - \mu.$$

Its energies are

$$E_\pm(\mathbf k) = \pm \hbar v_D \lvert\mathbf k\rvert - \mu.$$

For the ideal isotropic model, the pseudospin expectation is tangential to the constant-energy contour:

$$\langle \boldsymbol\sigma \rangle_\pm = \pm \frac{ \hat{\mathbf z} \times \mathbf k }{ \lvert\mathbf k\rvert }.$$

This is spin–momentum locking in the effective model. Crystal orbitals and spin–orbit entanglement can reduce or tilt the measured physical-spin polarization.

Time-reversal-breaking mass

The mass term

$$H_{\mathrm{gap}} = m\sigma_z$$

produces

$$E_\pm(\mathbf k) = \pm \sqrt{ \hbar^2v_D^2 \lvert\mathbf k\rvert^2 + m^2 } - \mu.$$

Because $\Theta\sigma_z\Theta^{-1}=-\sigma_z$, this mass breaks time reversal and opens a gap $2\lvert m\rvert$ at the Dirac point. A perpendicular exchange field is one possible source. In the isotropic model a uniform in-plane exchange $h_x\sigma_x+h_y\sigma_y$ instead shifts the cone to $k_x=-h_y/(\hbar v_D)$ and $k_y=h_x/(\hbar v_D)$. Warping and crystal symmetry can change that conclusion away from the minimal model. Structural asymmetry, scalar disorder, and ordinary surface potentials can shift or warp the cone but cannot generate this mass in a single isolated time-reversal-symmetric Dirac theory.

The converse needs care: a spectroscopic gap at the nominal Dirac energy does not by itself prove magnetic mass generation. Hybridization between opposite surfaces, finite momentum resolution, matrix-element suppression, band bending, disorder, or overlap with bulk bands can mimic or obscure a gap.

Warping without gapping

Crystal symmetry permits corrections. On a threefold surface, a common term is

$$H_w = \frac{\lambda}{2} \left( k_+^3+k_-^3 \right) \sigma_z.$$

Here $k_\pm=k_x\pm ik_y$ and $\lambda$ has units of energy times length cubed. Equivalently,

$$H_w=\lambda(k_x^3-3k_xk_y^2)\sigma_z.$$

The cubic momentum factor and $\sigma_z$ are both odd under time reversal, so their product is invariant. It distorts circular constant-energy contours into hexagonal or snowflake-like shapes and introduces an out-of-plane spin component, but it does not create a mass at $\mathbf k=0$.

Exact backscattering versus general scattering

For an ideal single cone, states at $\mathbf k$ and $-\mathbf k$ are Kramers partners. The matrix element of a time-reversal-invariant scalar impurity between them vanishes:

$$\langle u_{-\mathbf k} \rvert V \lvert u_{\mathbf k} \rangle = 0.$$

The direct matrix element is zero, and elastic time-reversal reciprocity forbids scattering into the exact Kramers partner beyond first order as well. Multiple elastic paths do not restore that forbidden amplitude when the same assumptions hold. They can still redistribute momentum through other angles. A two-dimensional surface has many such outgoing directions, unlike a single one-dimensional helical edge.

For the massless isotropic cone, two positive-energy spinors separated by angle $\vartheta$ obey

$$\begin{aligned} F(\vartheta)&=|\langle u_{\mathbf k'}|u_{\mathbf k}\rangle|^2\\ &=\frac{1+\cos\vartheta}{2}. \end{aligned}$$

This follows from the projectors $P_{\mathbf k}=(I+\hat{\mathbf d}_{\mathbf k}\cdot\boldsymbol\sigma)/2$: the overlap is $\operatorname{Tr}(P_{\mathbf k'}P_{\mathbf k})$. The angular weight vanishes at $\pi$, but is positive at almost every other angle.

For dilute short-range scalar impurities in the Born approximation, let $C>0$ contain the common impurity density, squared potential and density-of-states factors. The quantum and transport rates are

$$\tau_q^{-1}=C\langle F\rangle_\vartheta,$$ $$\tau_{\mathrm{tr}}^{-1} =C\langle(1-\cos\vartheta)F\rangle_\vartheta.$$

The angular average is over $0\leq\vartheta<2\pi$. The two averages are $1/2$ and $1/4$, so $\tau_{\mathrm{tr}}=2\tau_q$, both finite. The quantum lifetime counts scattering out of one momentum state; the transport lifetime discounts small changes in current direction. A missing exact-backscattering event therefore does not imply infinite conductivity (Culcer and colleagues).

With a perpendicular mass and positive energy $\varepsilon=E+\mu>|m|$, write $\eta=m/\varepsilon$. The same projector calculation gives

$$F_m(\vartheta)=A+B\cos\vartheta,$$ $$A=\frac{1+\eta^2}{2},\qquad B=\frac{1-\eta^2}{2}.$$

Thus $F_m(\pi)=\eta^2$: breaking time reversal permits exact scalar backscattering between these now non-Kramers-related states. The short-range Born model gives

$$\frac{\tau_{\mathrm{tr}}}{\tau_q} =\frac{2(1+\eta^2)}{1+3\eta^2}.$$

Long-range disorder, warping, additional pockets, bulk leakage and inelastic processes require different transport kernels. Near the band edge, the assumed weak-disorder quasiparticle regime must also be checked.

Top: at zero energy, a scalar barrier transmits the isolated helical edge perfectly, while a transverse magnetic mass gives $T=\operatorname{sech}^2[mL/(\hbar v)]$. Bottom: scalar scattering on a massless Dirac surface has overlap $(1+\cos\vartheta)/2$; a mass with $m/\varepsilon=0.6$ restores a backscattering weight $0.36$. The lower panel shows a spinor-overlap factor, not a conductance or complete scattering probability.

Thin films and paired surfaces

A thick three-dimensional sample has spatially separated top and bottom surfaces. In a thin film, their wavefunctions overlap. A minimal two-surface model adds a pseudospin $\boldsymbol\tau$. Let $h_D=H_D+\mu I$ denote the massless surface Hamiltonian without its chemical-potential shift. At $\mu=0$, take

$$H_{\mathrm{film}}=h_D\tau_z+m_h\tau_x.$$

The term $m_h\tau_x$ preserves time reversal but gaps the pair of cones. Its magnitude typically decays as

$$\lvert m_h(d)\rvert \sim m_0 \exp\!\left( -\frac{d}{\xi} \right),$$

possibly with model-dependent oscillations. This does not contradict protection of one isolated surface: the finite film supplies a second anomalous surface that cancels the first.

Odd and even cone counts

An odd number of Dirac cones is stable under time reversal and charge conservation at the surface of a strong topological insulator. Ordinary surface bands can add cones in pairs. Consequently:

• counting every visible crossing is less reliable than tracking connectivity across the bulk gap;
• two nearby cones can mix and gap without breaking time reversal;
• a cone can merge into the projected bulk continuum and re-emerge elsewhere;
• surface reconstruction can fold the Brillouin zone;
• a weak topological insulator can have orientation-dependent “dark” surfaces.

For interacting three-dimensional SPT phases, a symmetric surface may also gap by developing intrinsic topological order. Symmetry-Protected Topological Phases explains this interacting boundary alternative.

Other Boundary Families

The chiral, helical, and Dirac models are organizing templates, not the whole subject.

Majorana boundaries

BdG particle–hole redundancy permits self-conjugate boundary quasiparticles. Depending on dimension and symmetry class, a topological superconductor can support end Majoranas, a helical or chiral Majorana edge, or a Majorana surface cone. A self-adjoint zero mode has equal electron and hole weight. At finite energy, particle–hole symmetry instead relates the operators at opposite energies; it does not make every BdG eigenstate individually self-adjoint or exactly charge neutral. There is no independently conserved quasiparticle charge, so Andreev reflection, heat transport and parity measurements carry different information from an ordinary charged-channel count.

Fermi arcs

A Weyl-semimetal surface can host an open constant-energy contour. The arc is assembled from chiral edge states of two-dimensional momentum slices whose Chern number is nonzero. Its shape is termination dependent. At the node energy the endpoints reach the projected nodes; at a different chemical potential an arc can terminate on a projected bulk Fermi pocket. If opposite charges project to the same surface momentum, the arc can disappear on that face.

Interacting topological boundaries

Intrinsic topological order can support boundaries selected by anyon condensation. Different boundaries of the same bulk can be gapless, gapped, or separated by localized defect modes. The bulk fixes which condensates and anomalies are allowed, not one microscopic dispersion. Topological Order develops the condensation and chiral-central-charge constraints.

Reconstruction, disorder and local symmetry

Electrostatics and interactions can add boundary channels without a bulk transition. A smooth Hall edge can form compressible strips, and a topological-insulator surface can acquire Rashba-split accumulation bands. Such ordinary pairs may mix and gap; the residual anomalous content must still match the bulk. Contacts and equilibration determine how the remaining channels contribute to a measurement.

Disorder can broaden spectra, create puddles, connect opposite boundaries or move the local chemical potential into bulk bands. In a helical edge, the elastic single-channel theorem remains valid only while that description is valid. On a surface, ordinary-angle scattering already allows finite momentum relaxation. Neither statement substitutes for a transport calculation.

The bulk and boundary can also have different symmetries. Surface magnetic order can permit a mass while the bulk remains in the same topological phase. Conversely, a nominally nonmagnetic coating can alter doping or reconstruction. Determine the actual boundary perturbation and its matrix structure before interpreting an observed gap.

Experimental Probes

No single probe returns “topological boundary state” as a direct observable. Each measures a projection of the boundary spectral function, wavefunction, response, or scattering matrix.

Angle-resolved photoemission

In a sudden-approximation description,

$$I(\mathbf k,\omega) \propto \left| M_{fi} \right|^2 f(\omega) A(\mathbf k,\omega),$$

where $M_{fi}$ is a photon-energy- and polarization-dependent matrix element, $f$ is the Fermi function, and $A$ is the spectral function. Photoelectron Spectroscopy derives the measurement formula. The product $fA$ selects occupied spectral weight.

For a candidate surface state, ARPES can test:

• a branch inside the projected bulk gap;
• connectivity between bulk valence and conduction sectors;
• a Dirac crossing or Fermi arc;
• weak dependence on photon-energy-inferred $k_\perp$;
• surface aging and termination dependence;
• agreement with a slab calculation.

Failure modes include matrix-element extinction, finite $k_\perp$ resolution, surface charging, adsorbates, band bending, unresolved gaps, and overlap with bulk spectral weight. A linear dispersion alone is not a topological diagnosis.

Spin-resolved photoemission

Spin-resolved ARPES tests the reversal of spin polarization around a surface contour. It supports spin–momentum locking when the momentum, band identity, and detector response are controlled.

The measured polarization need not be unity. Orbital entanglement, unresolved bands, final-state effects, domain averaging, and detector calibration matter. Conversely, a Rashba-split ordinary surface state can also have a helical-looking spin texture. The stable distinction requires bulk connectivity and crossing parity.

Scanning tunneling spectroscopy and interference

For a featureless tip at low temperature,

$$\frac{dI}{dV}(\mathbf r,V) \propto \rho_s \left( \mathbf r, E_F+qV \right).$$

Defects generate standing-wave modulations. Their Fourier transform $\delta\rho(\mathbf q,\omega)$ samples scattering between equal-energy states separated by $\mathbf q$, weighted by spinor overlap and the impurity $T$ matrix.

In an ideal helical contour, the exact vector connecting $\mathbf k$ to $-\mathbf k$ is suppressed for a scalar impurity. But a missing Fourier feature is not automatically proof of topology: matrix elements, weak impurity strength, finite field of view, orbital selection, and overlapping bands can also suppress intensity. A strong analysis predicts the complete allowed and forbidden $\mathbf q$ pattern.

Scanning Tunneling Microscopy and Spectroscopy develops the junction, impurity $T$ matrix and Fourier-map limitations.

Electrical transport

Boundary transport is persuasive when geometry and tuning separate it from the bulk. Useful tests include:

• gating through a bulk-depleted regime;
• length, width, thickness, and perimeter scaling;
• local and nonlocal voltage configurations;
• contact permutations;
• temperature and bias dependence;
• response to controlled time-reversal breaking;
• comparison with capacitance or thermodynamic carrier density.

If the same terminal voltage drives independent parallel paths, a useful approximate model is

$$G_{\mathrm{meas}} \simeq G_{\mathrm{boundary}} + G_{\mathrm{bulk}} + G_{\mathrm{other}},$$

with contacts treated separately. In three-dimensional topological insulators, residual bulk carriers often dominate even when ARPES clearly resolves a surface cone. Surface spectroscopy and surface-dominated transport are different achievements.

Thermal and neutral-mode probes

Electrical transport misses charge-neutral Majorana or fractional neutral modes. Thermal conductance, noise, upstream heat propagation, and energy equilibration can probe net chiral central charge. Interpretation requires phonon subtraction, contact calibration, and an equilibration model.

Controlled perturbations

A protection claim becomes stronger when the boundary responds selectively:

• a perturbation with the required transverse or perpendicular mass component opens the predicted gap;
• a nonmagnetic perturbation shifts or broadens it without producing the forbidden mass;
• opposite-surface hybridization decays with thickness;
• changing surface orientation modifies weak-TI or Weyl boundary visibility as predicted;
• domain walls in a boundary mass bind the expected lower-dimensional channel.

The perturbation must be characterized independently. Depositing nominally magnetic atoms can also dope, disorder, or reconstruct a surface, so a spectral change is not automatically an exchange gap.

Combining complementary probes

Observation	What it establishes	Main alternative
Boundary-localized weight	A boundary state	Ordinary surface bands
Bulk-gap connectivity	Anomalous-band candidate	Unresolved bulk branches
Spin-sensitive scattering	A selection rule	Ordinary Rashba texture
Perimeter or thickness scaling	A boundary contribution	Parallel conduction
Mass or overlap response	A model-dependent gap mechanism	Doping or reconstruction

These observations become more discriminating together. For example, a thickness-dependent spectral gap and opposite-surface wavefunction weight support hybridization, while a thickness-independent gap tracking perpendicular magnetization supports an exchange mass. Neither conclusion follows from a gap image alone. Phase-specific experiments and their limits are discussed in Topological Insulators and the corresponding Hall, superconducting and semimetal pages.

Common pitfalls

Equating protection with zero scattering. The protected elastic amplitude is specific. Other surface angles and interacting edge processes still matter.

Treating pseudospin as conserved electron spin. Orbital entanglement and matrix elements affect spin-resolved measurements; a helical spin texture also occurs in ordinary Rashba bands.

Inferring magnetism from one apparent gap. Thickness, trivial-band avoided crossings, resolution and spectral weight can obscure the distinction between a true gap and a missing signal.

Ignoring the other side of a sample. Finite films pair anomalous boundaries. Their overlap can open a gap without changing the bulk phase.

Exercises

1. Scalar phase and magnetic reflection

For the isolated helical edge, compare a scalar barrier $V(y)I$ with the transverse mass segment in the text. Derive $R+T=1$ for the mass segment, and find the large-$|a|$ transmission.

Solution

The two scalar-potential equations are independent first-order equations; each solution acquires a phase and neither generates the other component. Hence scalar reflection is zero. For the mass transfer matrix,

$$\mathcal M=\begin{pmatrix}\cosh a&-i\sinh a\\i\sinh a&\cosh a\end{pmatrix}.$$

The second matching equation gives $r=-i\tanh a$, and the first gives $t=1/\cosh a$. Equal lead velocities mean $R=|r|^2$ and $T=|t|^2$, so

$$R+T=\tanh^2a+\operatorname{sech}^2a=1.$$

For $|a|\gg1$, $T\simeq4e^{-2|a|}$. The result describes an evanescent segment along an otherwise conducting edge, not a change of the two-dimensional bulk invariant.

2. Which magnetic terms gap a surface?

Add a constant $\mathbf h\cdot\boldsymbol\sigma$ to the isotropic surface Dirac Hamiltonian. Locate the minimum gap and explain why broken time reversal is not sufficient by itself to produce it.

Solution

The coefficient vector is

$$\begin{aligned} d_x&=-\hbar v_Dk_y+h_x,\\ d_y&=\hbar v_Dk_x+h_y,\\ d_z&=h_z. \end{aligned}$$

The energy separation is $2|\mathbf d|$. It is minimized at

$$k_x=-\frac{h_y}{\hbar v_D},\qquad k_y=\frac{h_x}{\hbar v_D},$$

where the separation is $2|h_z|$. Thus in-plane exchange shifts the ideal cone and perpendicular exchange gaps it. This statement assumes the linear isotropic model; warping can convert a shift away from the origin into an additional effective mass.

3. Finite transport despite forbidden backscattering

Evaluate the two angular averages for $F=(1+\cos\vartheta)/2$. Repeat with the massive overlap $F_m$ and recover the lifetime ratio in the text.

Solution

Use $\langle\cos\vartheta\rangle=0$ and $\langle\cos^2\vartheta\rangle=1/2$. Then

$$\langle F\rangle=\frac12,\qquad \langle(1-\cos\vartheta)F\rangle=\frac14.$$

For $F_m$, the averages become

$$\langle F_m\rangle=\frac{1+\eta^2}{2},$$ $$\langle(1-\cos\vartheta)F_m\rangle=\frac{1+3\eta^2}{4}.$$

Dividing the quantum rate by the transport rate gives $2(1+\eta^2)/(1+3\eta^2)$. At zero mass the ratio is two; as $|\eta|$ approaches one the spinor overlap becomes nearly angle independent and the ratio approaches one. The density of states and weak-disorder validity must still be checked when taking that band-edge limit.

4. One operator, three relevance thresholds

Suppose $e^{4i\phi}$ has scaling dimension $4K$. Derive the weak-coupling relevance exponents for a uniform perturbation, one impurity and a static random complex amplitude with short-range spatial variance.

Solution

Under a scale change by $b>1$, a space-time integral supplies $b^2$ while the operator supplies $b^{-4K}$. Its coupling grows as $b^{2-4K}$ at commensurate filling. A fixed impurity has only a time integral and grows as $b^{1-4K}$.

Static disorder averaging pairs two operators at the same position but independent times. The resulting measure is $dy\,d\tau\,d\tau'$, and the operator pair has dimension $8K$. The variance therefore grows as $b^{3-8K}$. Positive exponents give the three thresholds in the table. These are weak-coupling tests of a particular operator, not universal critical interaction strengths for every microscopic edge.

5. Extract a boundary penetration depth

A symmetric topological-insulator film has hybridization gaps

$$\Delta(d) = \Delta_0e^{-d/\xi}.$$

Measurements give $\Delta(6\,\mathrm{nm})=20\,\mathrm{meV}$ and $\Delta(10\,\mathrm{nm})=5\,\mathrm{meV}$. Estimate $\xi$.

Solution

Taking the ratio eliminates $\Delta_0$:

$$\frac{ \Delta(6) }{ \Delta(10) } = \exp\!\left( \frac{4\,\mathrm{nm}}{\xi} \right) = 4.$$

Therefore

$$\xi = \frac{ 4\,\mathrm{nm} }{ \ln4 } \approx 2.89\,\mathrm{nm}.$$

With more thicknesses one should fit the full data and test for oscillations or multiple decay lengths rather than trusting a two-point estimate.

Continue studying

Bulk–Boundary Correspondence relates these spectra to bulk indices and gives domain-wall and lattice-strip calculations. Angle-Resolved Photoemission Spectroscopy and Spectral Functions explain how boundary weight and linewidth appear in experiments.

Conductance Quantization and Mesoscopic Transport develop contacts and scattering matrices. Symmetry-Protected Topological Phases and Topological Order explain interacting alternatives to a gapless electron band.

References

• Z. Alpichshev et al., “STM Imaging of Electronic Waves on the Surface of $\mathrm{Bi_2Te_3}$: Topologically Protected Surface States and Hexagonal Warping Effects,” Physical Review Letters 104, 016401 (2010), doi:10.1103/PhysRevLett.104.016401.
• J. H. Bardarson, “A Proof of the Kramers Degeneracy of Transmission Eigenvalues from Antisymmetry of the Scattering Matrix,” Journal of Physics A: Mathematical and Theoretical 41, 405203 (2008), doi:10.1088/1751-8113/41/40/405203.
• M. Büttiker, “Absence of Backscattering in the Quantum Hall Effect in Multiprobe Conductors,” Physical Review B 38, 9375–9389 (1988), doi:10.1103/PhysRevB.38.9375.
• D. Culcer, E. H. Hwang, T. D. Stanescu, and S. Das Sarma, “Two-Dimensional Surface Charge Transport in Topological Insulators,” Physical Review B 82, 155457 (2010), doi:10.1103/PhysRevB.82.155457.
• L. Fu, “Hexagonal Warping Effects in the Surface States of the Topological Insulator $\mathrm{Bi_2Te_3}$,” Physical Review Letters 103, 266801 (2009), doi:10.1103/PhysRevLett.103.266801.
• B. I. Halperin, “Quantized Hall Conductance, Current-Carrying Edge States, and the Existence of Extended States in a Two-Dimensional Disordered Potential,” Physical Review B 25, 2185–2190 (1982), doi:10.1103/PhysRevB.25.2185.
• Y. Hatsugai, “Chern Number and Edge States in the Integer Quantum Hall Effect,” Physical Review Letters 71, 3697–3700 (1993), doi:10.1103/PhysRevLett.71.3697.
• P. Roushan et al., “Topological Surface States Protected from Backscattering by Chiral Spin Texture,” Nature 460, 1106–1109 (2009), doi:10.1038/nature08308.
• W. Shockley, “On the Surface States Associated with a Periodic Potential,” Physical Review 56, 317–323 (1939), doi:10.1103/PhysRev.56.317.
• X.-G. Wen, “Chiral Luttinger Liquid and the Edge Excitations in the Fractional Quantum Hall States,” Physical Review B 41, 12838–12844 (1990), doi:10.1103/PhysRevB.41.12838.
• C. Wu, B. A. Bernevig, and S.-C. Zhang, “Helical Liquid and the Edge of Quantum Spin Hall Systems,” Physical Review Letters 96, 106401 (2006), doi:10.1103/PhysRevLett.96.106401.
• C. Xu and J. E. Moore, “Stability of the Quantum Spin Hall Effect: Effects of Interactions, Disorder, and $\mathbb Z_2$ Topology,” Physical Review B 73, 045322 (2006), doi:10.1103/PhysRevB.73.045322.

Further reading

• B. A. Bernevig and T. L. Hughes, Topological Insulators and Topological Superconductors, Princeton University Press (2013).
• M. Z. Hasan and C. L. Kane, “Colloquium: Topological Insulators,” Reviews of Modern Physics 82, 3045–3067 (2010), doi:10.1103/RevModPhys.82.3045.