DIVERGENCE RATES OF TENSOR DEGENERATIONS
We study the divergence rates of approximate tensor decompositions. We establish a precise connection between the algebraic geometry of tensor degenerations and quantum information theory, demonstrating how the optimal probability of converting a GHZ state into a target state decays with respect to the approximation error. The main results are: • Exact rates for pencils: For tripartite tensors with polynomial slices (including all regular pencils), the divergence exponent is exactly (d−1)/d, where d is the size of the largest Jordan block. The matching lower bound holds for every approximating decomposition. • Stability: Adding a product block on new directions to a tensor with an invertible slice leaves the divergence exponent unchanged. • W-states: For the W state, an r-node difference formula yields an exponent bounded by 1/r, which is sharp for r=2,3 according to Hackbusch's results. • The Hackbusch conjecture: The general conjecture θ≥1/r is restated in operational form and reduced to a statement about Laurent arcs of product vectors on the Segre variety.
Authors
- JOSÉ TORREGROSA JIMÉNEZ
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.5281/zenodo.24308624
- Primary Topic
- Tensor decomposition and applications
- Type
- preprint