How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
Rank-revealing QR sweeps give exact mixed canonical gauge
Statement
Let a nonzero finite open-boundary TT have order , and choose a center window with and . There is an exact rank-revealing QR transformation of its cores that preserves the represented array, makes every core left-isometric and every core right-isometric, and removes exactly the dependent coordinates on each swept bond. No independent column direction is truncated; only exactly dependent directions are removed.
For each reshaped matrix , the rank-revealing QR used here scans its columns in their fixed order, retaining a column exactly when it increases the span. It applies Gram–Schmidt to the retained columns, writes all other columns in that orthonormal span, and records the resulting permutation and coefficient factor.
Facts & Assumptions
Given: A nonzero open-boundary TT array over , with a selected one-site or two-site center window .
Finite matrix-product states and Hermitian Hamiltonians: an open-boundary MPS/TT presentation is an exact TT core list with one-dimensional boundary bonds; a normalized state is nonzero.
Finite tensor arrays and the Frobenius inner product: the rank of a matrix is the dimension of its column image.
Tensor train cores, bond ranks, and canonical gauges: the left matricization is left-isometric when , and the right matricization is right-isometric when .
Gram–Schmidt turns every finite independent list into an orthonormal list with the same successive spans: a finite linearly independent list in a real or complex inner-product space has an orthonormal list with the same successive spans, produced by Gram–Schmidt.
Proof
Let be any core matricization. Scan its columns in order, starting with an empty list and retaining a column exactly when it is not in the span of those already retained. At termination the retained columns are independent, and every skipped column lies in their final span; hence they span the column image of and their count is by [F2]. If , apply [F4] to obtain a matrix with orthonormal columns spanning those retained columns; their Gram–Schmidt coordinates form an upper triangular with nonzero diagonal. If , then and use the empty factorization.
Move the retained column indices to the front by a permutation . Every remaining column lies in the span of , so it equals times its coordinate vector applied to that column. Thus for some , with ; equivalently where . This is an exact rank-revealing thin QR factorization for every rectangular shape.
Starting at the left boundary, apply step 2.1 to each left matricization of cores . Replace the current core matricization by and absorb into the next core. Since , the shared-index contraction in [F1] is unchanged; makes the replaced core left-isometric by [F3]. Induction over the swept cores gives the required left-isometric prefix.
Starting at the right boundary, apply step 2.1 to the adjoint of each right matricization of cores . If , then . Replace the current right matricization by and absorb into the preceding core. The shared-index contraction in [F1] is unchanged, and makes the replaced matricization right-isometric by [F3]. Induction gives the right-isometric suffix and preserves the represented array, while leaving cores untouched. If , the final absorption into core can change the contracted center coefficients; if , there is no right sweep.
The scan drops only columns already in the span of retained columns, so each bond dimension changes only from its former size to the exact matrix rank. If any swept matricization had rank zero, that core would be zero and the full TT contraction would be zero, contrary to the nonzero-state hypothesis; hence every reduced swept bond has positive dimension. Steps 3.1 and 3.2 establish the exact mixed canonical gauge.
Depends on
Used by
Dependency tree · two levels
8 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.
Sources
- Ulrich Schollwöck, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326 (2011), §§4.1.2, 4.1.3, 4.4 (standard reference, not scraped)
- Grégoire Misguich, Introduction to matrix-product states and tensor networks, complete lecture notes (2026), §§2.4.1–2.4.3 (standard reference, not scraped)