docs: content wave 2 (headline how-tos, tutorials, algorithm landscape)#451
Closed
lkdvos wants to merge 8 commits into
Closed
docs: content wave 2 (headline how-tos, tutorials, algorithm landscape)#451lkdvos wants to merge 8 commits into
lkdvos wants to merge 8 commits into
Conversation
Task recipes for find_groundstate: keyword-form heuristic, explicit DMRG/DMRG2/VUMPS/IDMRG/IDMRG2/GradientGrassmann configuration, &-chaining, verbosity/tolerance control. All 12 @example blocks verified against real MPSKit; physics claims salvaged from man/algorithms.md carry REVIEW flags. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Recipes for timestep/time_evolve with TDVP/TDVP2, bond growth during evolution, infinite-state evolution, imaginary time, and time-evolution MPOs (WII/TaylorCluster/WI) applied via approximate. All 11 @example blocks verified against real MPSKit; physics claims carry REVIEW flags. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Recipes for QuasiparticleAnsatz (infinite gap, dispersion scan, charged Z2 sector), finite-chain excitations (finite QP, FiniteExcited, ChepigaAnsatz/2), and variance quality checks; domain-wall usage is prose-only pending in-repo test coverage. All 11 @example blocks verified against real MPSKit. Note: verification exposed a numerical bug — find_groundstate on InfiniteMPS at g=10, chi=12 hits a LAPACK SVD ArgumentError inside the auto-chained GradientGrassmann step (over-parametrized bond dimension on a near-product state); reproducible across seeds and single-threaded. The recipe uses g=2.0 to avoid it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Router page listing every how-to page and its numbered recipes, grouped by theme, with a missing-recipe issue invitation. All @ref anchors verified; no code blocks. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New Concepts section with the finite/infinite x task decision table and per-task prose salvaged from man/algorithms.md. Zero code blocks. Every physics/convergence claim carries a REVIEW flag (26 total) for maintainer vetting per CLAUDE.md; dropped-claim inventory in a trailing comment. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Hand-held TDVP quench: TFIM ground state at g=0.5, quench to g=2.0, explicit 40-step timestep loop tracking mid-chain transverse magnetization, plotted with Plots. Verified end-to-end twice: the initial sigma-z story was rewritten to track sigma-x after verification showed DMRG at D=16 converges to the symmetric ground state (flat sigma-z trace); the sigma-x trace shows order-1 relaxation dynamics (0.26 -> peak 0.61 -> plateau 0.49). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Hand-held dispersion walk-through on the infinite TFIM at g=2: single momentum gap, 16-momentum sweep, plot against the exact Jordan-Wigner dispersion. Verified end-to-end; the k=0 gap matches 2(g-1) to 1e-11. Physics claims (phase, ansatz framing, exact formula) carry REVIEW flags. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Flagship TFIM redone with explicit Z2 symmetry: graded spaces, energy agreement with the trivial run (1e-14), sector-resolved virtual space and entanglement spectrum, and the charged-sector excitation payoff. Plan deviation noted in-file: MPSKitModels' transverse_field_ising only supports Trivial/Z2Irrep/FermionParity, so the plan's U(1) framing is replaced by Z2; U(1)/SU(2) are pointed to via states how-to and examples. All 13 @example blocks verified twice (unseeded RNG). Also replaces the two TODO(link) stubs in thermodynamic_limit.md with real links to the new excitations and using-symmetries tutorials. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. 🚀 New features to boost your workflow:
|
Contributor
|
After the build completes, the updated documentation will be available here |
Member
Author
|
Folded into #449 by rebasing onto docs/content-wave1 and fast-forwarding; branch removed. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #449 (
docs/content-wave1). Second wave of the docs restructure perdocs/RESTRUCTURE_PLAN.md/docs/IMPROVEMENT_PLAN.mdW2: the headline-workflow task recipes and the back half of the tutorial track, plus the algorithm-landscape decision guide.Pages added (one commit each)
How-to (task recipes)
howto/groundstate_algorithms.md— find_groundstate heuristic, DMRG/DMRG2/VUMPS/IDMRG/IDMRG2/GradientGrassmann config,&-chaining, tolerances/verbosityhowto/time_evolution.md— timestep/time_evolve with TDVP/TDVP2, imaginary time, time-evolution MPOs (WII/TaylorCluster/WI) via approximatehowto/excitations.md— QuasiparticleAnsatz (gap, dispersion, charged Z2 sector), FiniteExcited, ChepigaAnsatz/2, variance checks; domain-wall usage prose-only (no in-repo test coverage to copy)howto/index.md— browsable recipe index (new How-to landing page) with per-recipe listing and a missing-recipe issue invitationTutorials
tutorials/time_evolution.md— hand-held TFIM quench with an explicit TDVP loop and a plotted observabletutorials/excitations.md— dispersion of the infinite TFIM; computed points match the exact Jordan–Wigner dispersion (k=0 gap agrees with 2(g−1) to 1e-11)tutorials/using_symmetries.md— flagship redone with explicit Z2 symmetry; graded spaces, sector-resolved entanglement, charged-sector excitation payoffTODO(link)stubs intutorials/thermodynamic_limit.mdto the new tutorialsConcepts
concepts/algorithm_landscape.md— decision table (task × finite/infinite) + per-task comparison prose salvaged fromman/algorithms.md. 26<!-- REVIEW: -->flags — every physics/convergence claim is gated on maintainer review per CLAUDE.md.Needs maintainer attention
concepts/algorithm_landscape.md(26); each tutorial/how-to carries 3–6 for its physics framing.MPSKitModels.transverse_field_isingonly acceptsTrivial/Z2Irrep/FermionParity(throws otherwise), so the symmetries tutorial uses Z2 and points at states-how-to §9 / XXZ / Haldane examples for U(1)/SU(2). Noted in-file.find_groundstateonInfiniteMPSat g=10 (also g=5), χ=12 deterministically throwsArgumentError: invalid argument #4 to LAPACK callin MatrixAlgebraKit's divide-and-conquer SVD, from the auto-chained GradientGrassmann step — over-parametrized bond dimension on a near-product state; reproduces across seeds, single-threaded. g ≤ 2 is fine. Probably worth an issue on TensorKitManifolds/MatrixAlgebraKit.approximate's docstring says it returns a 2-tuple; every method returns(ψ, envs, ϵ)— docstring fix candidate for the W4 wave.examples/index.mdandlib/public.md's spectral-functions section have no@idanchors, so pages can't deep-link them.Verification
Every
@exampleblock on every new page was executed against real MPSKit before its commit (extracted per-page into a single Julia session,julia --project=docs), with semantic spot-checks (energy agreement trivial-vs-Z2 to 1e-14 across two RNG seeds; charged sector below trivial sector; dispersion vs exact formula). Final gate: fulldocs/make.jlDocumenter run over the whole tree (doctest = true, all cross-references) — green, no new warnings.🤖 Generated with Claude Code