- Accepted Paper
Counting bounds for flat-band interaction sectors and the superfluid weight
Phys. Rev. B - Accepted 17 November, 2026
DOI: https://doi.org/10.1103/d2b5-3lqb
Phys. Rev. B - Accepted 17 November, 2026
DOI: https://doi.org/10.1103/d2b5-3lqb
For interacting flat bands, our calculations refer to two kinds of quantities, namely counting quantities like the rank of the projected interaction and the number of pairing channels, and second-moment quantities like the quantum metric and the superfluid weight that the metric bounds. We show that both kinds are independent, since in exactly flat finite-range lattices no function of the rank of the projected on-site interaction, and no counting quantity invariant with respect to dilation of the compact-localized-state (CLS) support, can bound the minimal quantum metric integral from above. Thus neither bounds the superfluid weight of an isolated flat band in Bardeen-Cooper-Schrieffer (BCS) mean-field theory for a weak attractive contact interaction with time-reversal symmetry and uniform pairing. Restricting only the hopping range, or the hopping range along with the rank, does not restore an upper bound, and we also need a lower bound on the spectral gap. For a controlled support diameter and CLS conditioning we find an explicit quadratic upper bound. The obstruction is due to a mismatch of symmetry, as the counting ceiling is an affine invariant of the CLS support, whereas the minimal quantum metric integral grows quadratically as that support is dilated. Our exactly flat finite-range models realize this separation in three families, and one of them is a connected-lattice family at constant spectral gap and exact uniform pairing, where the difference-set ceiling holds the pair-sector rank strictly below the orbital count. Since the obstruction originates in symmetry and not in technique, no sharpening of a dilation-invariant counting ceiling removes it. We also derive the counting bounds themselves for the unconjugated pair channel, or the real-CLS time-reversal-symmetric pair channel, as well as for the complex time-reversal-symmetric singlet channel. We prove that equality holds generically on a dense Zariski-open set of CLS coefficients. Whenever the orbital content exceeds the orbit count, the difference set of the CLS support, instead of the orbital count, determines the rank of the projected on-site interaction. The shape of a compact localized state, and not only its size, thus decides the number of interaction channels in a flat band, and moving one CLS site to a new position generically opens one more channel at constant orbital count.
If the author has provided any supplemental materials with this article they will be available upon publication of the version of record.