Skip to content

Berry Geometry and the Quantum Metric

The quantum geometric tensor measures how rapidly an occupied state changes into the unoccupied subspace. Its real part is the quantum metric and its imaginary part is Berry curvature. The connection depends on a phase or occupied frame; the metric, curvature, Wilson loops, and integrated characteristic classes are gauge invariant when their domain and orientation are fixed.

Required background. Bloch and Wannier theory supplies the isolated occupied projector and its U(Nocc)U(N_{\rm occ}) frame freedom.

Helpful background. Characteristic classes and Chern–Weil theory supplies the global interpretation of curvature integrals.

For a normalized nondegenerate state u(λ)|u(\boldsymbol\lambda)\rangle and P=uuP=|u\rangle\langle u|, define

Qij=iu(1P)ju,gij=ReQij,Ωij=2ImQij.Q_{ij}=\langle\partial_i u|(1-P)|\partial_j u\rangle, \qquad g_{ij}=\operatorname{Re}Q_{ij}, \qquad \Omega_{ij}=-2\operatorname{Im}Q_{ij}.

With Ai=iuiu\mathcal A_i=i\langle u|\partial_i u\rangle, direct differentiation gives Ωij=iAjjAi\Omega_{ij}=\partial_i\mathcal A_j-\partial_j\mathcal A_i. The projection 1P1-P removes the phase derivative parallel to u|u\rangle, so QijQ_{ij} is gauge invariant. Infinitesimally,

1u(λ)u(λ+dλ)2=gijdλidλj+O(dλ3),1-|\langle u(\boldsymbol\lambda)|u(\boldsymbol\lambda+d\boldsymbol\lambda)\rangle|^2 =g_{ij}\,d\lambda^i d\lambda^j+O(d\lambda^3),

which makes positivity of gg manifest Provost and Vallée 1980.

For an occupied multiplet, frame-independent projector formulas are

gij=12Tr(iPjP),Ωij=iTrP[iP,jP].g_{ij}=\frac12\operatorname{Tr}(\partial_iP\,\partial_jP), \qquad \Omega_{ij}=i\,\operatorname{Tr}P[\partial_iP,\partial_jP].

The full non-Abelian curvature is matrix-valued; the trace above is the curvature relevant to the first Chern number. At an internal degeneracy, individual-band Abelian curvatures are not defined, but the multiplet projector remains valid if the multiplet stays separated from all other bands.

For H(k)=d0(k)1+d(k)σH(\mathbf k)=d_0(\mathbf k)\mathbf1+\mathbf d(\mathbf k)\cdot\boldsymbol\sigma with d>0|\mathbf d|>0, the lower-band projector is P=(1d^σ)/2P_-=(1-\hat{\mathbf d}\cdot\boldsymbol\sigma)/2. Pauli-matrix algebra yields

gij=14id^jd^,Ωij=12d^(id^×jd^).g_{ij}=\frac14\,\partial_i\hat{\mathbf d}\cdot\partial_j\hat{\mathbf d}, \qquad \Omega_{ij}=\frac12\hat{\mathbf d}\cdot (\partial_i\hat{\mathbf d}\times\partial_j\hat{\mathbf d}).

Thus the metric measures stretching of the map from parameter space to the Bloch sphere, while curvature measures its oriented area. Reversing (kx,ky)(k_x,k_y) reverses Ωxy\Omega_{xy} but not gijg_{ij}. The pointwise inequality detg(Ω12/2)2\det g\ge(\Omega_{12}/2)^2 follows from positivity of QQ in two parameters; equality holds for special two-level geometries.

Finite differences of eigenvector phases are unreliable. On a momentum mesh, use overlaps around closed plaquettes for curvature and singular values of overlap matrices for a multiplet. Refine the mesh until the integrated Chern number and local peaks converge; a curvature hotspot alone is not a topological invariant.

The quantum metric enters localization bounds, superfluid weight in projected multiband settings, and optical sum rules, but each relation has additional dynamical hypotheses Peotta and Törmä 2015. Geometry of a projected band is not by itself evidence for a particular interacting phase.

For H=kxσx+kyσy+mσzH=k_x\sigma_x+k_y\sigma_y+m\sigma_z, evaluate Ωkxky\Omega_{k_xk_y} in the lower band.

Solution

Here d=(kx,ky,m)\mathbf d=(k_x,k_y,m) and D=kx2+ky2+m2D=\sqrt{k_x^2+k_y^2+m^2}. Substitution into the two-band formula gives Ωkxky=m/(2D3)\Omega_{k_xk_y}=m/(2D^3). It changes sign with mm, is concentrated on the scale km|\mathbf k|\sim|m|, and its sign would reverse if the momentum orientation or Berry-connection convention were reversed.

  • Sebastiano Peotta and Päivi Törmä, “Superfluidity in Topologically Nontrivial Flat Bands,” Nature Communications 6 (2015) 8944, doi:10.1038/ncomms9944.
  • J. P. Provost and G. Vallée, “Riemannian Structure on Manifolds of Quantum States,” Communications in Mathematical Physics 76 (1980) 289–301, doi:10.1007/BF02193559.