Symmetry-Protected Structure Preview

A protected feature survives a specified class of changes. A Kramers pair cannot split while its time-reversal assumptions hold; a Chern number cannot change while the relevant band subspace remains isolated. These are different mechanisms. Learning to identify the permitted perturbations explains when a degeneracy, boundary excitation, or phase distinction is robust, and when a small-looking change can remove it. This page introduces that reasoning; symmetry-protected topological phases develops the many-body theory.

Helpful background. Exact symmetry specifies the invariant Hamiltonian, degeneracy lifting explains restricted perturbation matrices, Kramers degeneracy supplies antiunitary pairing, and Chern numbers explains a global band invariant.

Protection means restricting the allowed changes

For a spectral degeneracy, ask whether an allowed perturbation $V$ can split the relevant energy level. For a gapped phase, ask whether there is a continuous path $H(s)$, $0\leq s\leq1$, to a reference phase while preserving the required symmetry, locality, and bulk gap. “Protected” always refers to these conditions, including spatial dimension and whether interactions are allowed.

Smallness by itself is not the protection. An arbitrarily weak symmetry-breaking term may split an exact doublet, while a symmetry-preserving term of substantial size can keep it paired. Conversely, a sufficiently large symmetric deformation can close a bulk gap and cross a phase transition. Robustness near one Hamiltonian does not allow every finite deformation.

If $P$ projects onto a degenerate eigenspace, $PVP$ determines its first-order shifts. A scalar $PVP$ for one perturbation establishes only first-order equality. Exact protection requires an argument that applies to the full allowed Hamiltonian. The three-state example shows why second-order splitting can occur when first order vanishes.

Kramers degeneracy is an algebraic constraint

Let $\Theta$ be antiunitary and suppose

$$\Theta H\Theta^{-1}=H,\qquad \Theta^2=-I$$

in the sector under consideration. Every normalizable energy eigenstate then has an orthogonal partner at the same energy. The Kramers theorem page proves this result.

Within a two-dimensional Kramers subspace, choose $\Theta=i\sigma_yK$. A Hermitian effective Hamiltonian is $H_{\rm eff}=a_0I+\mathbf a\cdot\boldsymbol\sigma$ with real coefficients. Since time reversal reverses all three Pauli matrices,

$$\Theta H_{\rm eff}\Theta^{-1}=a_0I-\mathbf a\cdot\boldsymbol\sigma.$$

Invariance forces $\mathbf a=0$. The doublet can move in energy but cannot split. Coupling to additional levels can rearrange doublets without removing the pairing theorem. A fixed magnetic field can allow a nonzero $\mathbf a$ because it breaks the assumed time reversal.

The sector qualification matters. Two spin-$1/2$ degrees of freedom together have $\Theta^2=+I$, so a time-reversal-invariant exchange interaction can give a unique singlet ground state. Protection of one isolated Kramers doublet is not a claim that every pair of such degrees of freedom must remain independently degenerate.

A Chern number remains fixed while the subspace stays isolated

For an isolated band or group of bands over a closed oriented two-dimensional parameter space $M$, the first Chern number is

$$C=\frac{1}{2\pi}\int_M\operatorname{Tr}F,$$

where $F$ is the Berry curvature of that subspace; for one band the trace is unnecessary. Smooth deformations preserve the integer while the spectral projector remains well-defined and smooth. An integer cannot vary continuously along a connected path.

For a band insulator, the occupied subspace must stay separated from the unoccupied one. A crossing between occupied bands can change how individual band labels are assigned without changing the Chern number of their full occupied subspace. A transition that changes this total invariant requires loss of the separating gap or another hypothesis.

A Chern insulator does not need time reversal for its protection. Indeed, ordinary time reversal, whether spinless or spinful, forces the total occupied first Chern number to vanish in a two-dimensional time-reversal-invariant band insulator. Time-reversal-protected insulators instead use other structure, such as a $\mathbb Z_2$ invariant. Zero Chern number does not mean every such phase is trivial. The condensed-matter application explains the curvature constraints, and Hasan–Kane reviews topological insulators.

Boundary structure and finite-size effects

A bulk invariant can enforce boundary behavior, but it need not fix the energy of every edge-localized state. For a Chern insulator, the net chiral edge structure across the gap is tied to the difference of bulk Chern numbers on the two sides of the interface. Local boundary changes can alter the dispersion or add ordinary pairs of edge branches without removing that net structure under the applicable assumptions.

The SSH model provides a different example. Its chiral symmetry can protect an end state at zero energy at an isolated boundary with the appropriate termination. In a finite chain, the two end states can overlap. In their effective basis let the chiral operator be $\Gamma=\sigma_z$. Then

$$H_{\rm ends}=\begin{pmatrix}0&\varepsilon\\\varepsilon^*&0\end{pmatrix}, \qquad \Gamma H_{\rm ends}\Gamma^{-1}=-H_{\rm ends},$$

yet the energies are $\pm|\varepsilon|$. Chiral symmetry pairs the energies about zero; it does not forbid coupling between opposite ends. For exponentially localized modes separated by length $L$, the overlap generally becomes exponentially small as $L$ grows. Exact zero modes at a special finite-chain point require additional conditions, such as zero overlap or sublattice imbalance.

Ordinary boundary potentials can also create localized states unrelated to a bulk invariant. Their removal need not change the bulk phase. Boundary termination, boundary symmetry, coupling to extra degrees of freedom, and the large-system limit are therefore part of a useful protection statement.

Symmetry-protected phases and intrinsic topological order

In the usual sense, a symmetry-protected topological phase is a gapped phase that can be made trivial by a suitable local gapped deformation if the protecting symmetry is relaxed, but cannot be made trivial under the corresponding symmetric restrictions. The bulk has no intrinsic topological order. Boundary consequences depend on dimension and interactions: an obstruction to a trivial symmetric boundary need not mean every boundary must stay gapless. It may instead break symmetry, have degeneracy, or, in appropriate dimensions, support its own topological order.

Intrinsic topological order is a different kind of bulk structure, with long-range entanglement and, in suitable dimensions, fractionalized excitations and characteristic ground-state degeneracy. It does not generally disappear merely by relaxing a protecting symmetry. Chern phases also illustrate why every nontrivial topological band phase should not automatically be called symmetry-protected in this narrower sense.

The topology introduction connects phases to boundaries and responses. Symmetry classification organizes the allowed single-particle constraints; dimension and interaction assumptions are still needed to obtain a classification. The review by Chiu and collaborators develops these distinctions.

Exercises

1. Two spin-$1/2$ degrees of freedom interact through $H=J\mathbf S_1\cdot\mathbf S_2/\hbar^2$, with $J>0$. Find the singlet and triplet energies. Why does a unique ground state not violate Kramers theorem?

Solution

Using $2\mathbf S_1\cdot\mathbf S_2=S_{\rm tot}^2-S_1^2-S_2^2$ gives $E_{s=0}=-3J/4$ and $E_{s=1}=J/4$. The singlet is unique. Time reversal squares to $(-1)(-1)=+1$ on the two-spin system, so the hypothesis $\Theta^2=-I$ is absent. The interaction respects time reversal while coupling the two individual doublets.

2. For $H_{\rm ends}$ above, compute the energies and show that an allowed chiral-symmetric change can split two finite-chain zero modes without closing the bulk gap.

Solution

The characteristic equation is $E^2-|\varepsilon|^2=0$, giving $E=\pm|\varepsilon|$. The off-diagonal coupling anticommutes with $\Gamma$ for every complex $\varepsilon$. Thus changing a small overlap from zero to nonzero preserves the symmetry while splitting this boundary pair. For a sufficiently long chain the splitting can remain much smaller than the bulk gap; the isolated-boundary limit and finite-chain spectrum are distinct claims.

References

• C.-K. Chiu, J. C. Y. Teo, A. P. Schnyder, and S. Ryu, “Classification of topological quantum matter with symmetries,” Reviews of Modern Physics 88, 035005 (2016). doi:10.1103/RevModPhys.88.035005.
• F. D. M. Haldane, “Model for a quantum Hall effect without Landau levels,” Physical Review Letters 61, 2015–2018 (1988). doi:10.1103/PhysRevLett.61.2015.
• M. Z. Hasan and C. L. Kane, “Colloquium: Topological insulators,” Reviews of Modern Physics 82, 3045–3067 (2010). doi:10.1103/RevModPhys.82.3045.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020. See time reversal and angular-momentum addition.
• D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs, “Quantized Hall conductance in a two-dimensional periodic potential,” Physical Review Letters 49, 405–408 (1982). doi:10.1103/PhysRevLett.49.405.