[MPS] 🌠 Wander/漫步者: Issue #123 Non-Markovian Floquet Heat Transport (N=1–4)

tracks/mps/solutions/Ranger-123/.gitignore

@@ -0,0 +1,15 @@
+.venv*/
+__pycache__/
+*.py[cod]
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+.omx/
+work/
+build/
+dist/
+*.egg-info/
+/results/
+/docs/plans/
+/docs/superpowers/
+/julia/Manifest.toml

tracks/mps/solutions/Ranger-123/CITATION.cff

@@ -0,0 +1,29 @@
+cff-version: 1.2.0
+title: "Floquet-IF Many-body: N=2,3"
+message: "If you use this research implementation, cite this software and the upstream methods."
+type: software
+authors:
+  - family-names: "Wang"
+    given-names: "Thomas"
+version: 0.1.0
+date-released: 2026-07-29
+license: AGPL-3.0-or-later
+references:
+  - type: article
+    title: "Exact Floquet Dynamics of Strongly Damped Driven Quantum Systems"
+    authors:
+      - family-names: Mickiewicz
+        given-names: Konrad
+      - family-names: Link
+        given-names: Valentin
+      - family-names: Strunz
+        given-names: Walter T.
+    year: 2026
+    doi: 10.1103/5z1m-122d
+  - type: software
+    title: "UniformTEMPO.jl"
+    url: "https://github.com/uniformTEMPO/UniformTEMPO.jl"
+    version: "b76a018c32e5415989761d902b1b0e95f1a337da"
+  - type: software
+    title: "OQuPy"
+    url: "https://github.com/tempoCollaboration/OQuPy"

tracks/mps/solutions/Ranger-123/README.md

@@ -0,0 +1,146 @@
+# Floquet-IF Many-body: N=1--4 validation and N=2,3 production
+
+## Submission
+
+- Team: Ranger
+- Members: Chenxi Wan, Yedi Shen, Junkai Wang
+- Challenge: #123
+- Track: MPS
+- Completed scope: published single-spin validation, reproducible collective
+  common-bath calculations for \(N=2,3\), a convergence-gated \(N=4\) point,
+  and quantitative non-Markovian--Floquet-Markov benchmarks.
+
+This repository is a reproducible implementation of the two- and three-spin
+part of [QuantumBFS/quantum.harness issue #123](https://github.com/QuantumBFS/quantum.harness/issues/123).
+It combines exact symmetry reduction, a public uniform influence-functional
+solver, frequency-resolved bath heat currents, and a Floquet-Markov benchmark.
+
+The production backend is
+[UniformTEMPO.jl](https://github.com/uniformTEMPO/UniformTEMPO.jl), pinned to
+revision `b76a018c32e5415989761d902b1b0e95f1a337da`. The Python layer projects
+the physical model into its symmetry sectors, runs Julia, caches the uniform
+process tensor atomically, inserts operators into the extended process-tensor
+state, separates coherent delta peaks from the decaying correlation, and
+records every convergence comparison in JSON.
+
+## Completed results
+
+- Exact \(N=2\) singlet/triplet reduction, dark singlet, bright gaps, and
+  transition weights.
+- Exact \(N=3\) reflection \(6\oplus2\) reduction and the
+  \(J\)-independent embedded single-spin odd sector.
+- A converged six-point \(N=3\) heat-spectrum grid for
+  \(J/\Omega=0.25,0.5,1\) in both reflection sectors.
+- A converged \(3\times3\) \(N=2\) calibration grid comparing uniform TEMPO
+  with Floquet-Markov/QRT using state, correlation, and heat-spectrum errors.
+- A converged 6/6 same-model \(N=3\) benchmark comparing each existing
+  reflection-sector UniformTEMPO point directly to Floquet-Markov/QRT.
+- Floquet matrix-element, collective-variance, counterterm, and normalization
+  diagnostics.
+- Full convergence of the two bounded-normalization model variants.
+- Compression convergence of both Kac-normalized variants, with their
+  remaining timestep/phase refinement explicitly marked as cluster work.
+- An independent 3/3 reproduction of the published transversal-drive Fig. 3
+  bottom panel, plus a coarse UniformTEMPO--OQuPy cross-check.
+- A converged reflection-odd \(N=4\) pilot with a same-model
+  Floquet-Markov/QRT heat-spectrum error of 3.405 and resolved collective
+  spectral peaks.
+- A converged reflection-even \(N=4\) pilot using an automatically extended
+  six-period correlation window; its same-model heat-spectrum error is 5.494.
+- A pole-resolved, fixed-frequency \(N=1,2,3\) heat-valve pre-scan and
+  three-point \(N=3\) UniformTEMPO go/no-go pilot. The pilot rejects—not
+  confirms—the dark-channel hypothesis: the quasienergy gap collapses while
+  the observable transfer-pole residue increases.
+
+The main \(N=2\) and \(N=3\) results do **not** require a cluster. The failed
+heat-valve pilot does not justify a cluster-scale nine-point continuation.
+Only the optional full Kac refinement does. Reflection-resolved \(N=4\)
+support is implemented and both sector pilots pass all declared gates; the
+even sector records the required six-period tail-window extension explicitly.
+The project makes no thermodynamic-limit, continuum, or critical-exponent
+claim.
+
+## Reproduce
+
+Python 3.12 and Julia 1.12 are the recorded production versions.
+
+```bash
+python3.12 -m venv .venv
+.venv/bin/python -m pip install -e '.[dev,nonmarkov]'
+julia --project=julia -e 'using Pkg; Pkg.instantiate()'
+
+# Exact and legacy validation baselines
+PYTHON_BIN=.venv/bin/python scripts/run_baselines.sh
+PYTHON_BIN=.venv/bin/python scripts/run_pt_baselines.sh
+
+# Resumable production calculation and strict audit
+PYTHON_BIN=.venv/bin/python scripts/run_paper_extension.sh all
+
+# Independent backend checks
+.venv/bin/python scripts/run_uniform_validation.py
+.venv/bin/python scripts/run_fig3_validation.py --drive-frequency 1
+.venv/bin/python scripts/run_fig3_validation.py --drive-frequency 1.5
+.venv/bin/python scripts/run_fig3_validation.py --drive-frequency 2
+.venv/bin/python scripts/run_fig3_validation.py --plot-summary
+
+# Convergence-gated N=4 sector points
+.venv/bin/python -m floquet_if_manybody.cli n4-pilot --sector odd --j 0.25
+.venv/bin/python -m floquet_if_manybody.cli n4-pilot --sector even --j 0.25
+
+# Software and result verification
+.venv/bin/python -m pytest -q
+.venv/bin/python -m ruff check src tests scripts/run_uniform_validation.py scripts/run_fig3_validation.py
+.venv/bin/python -m mypy src scripts/run_fig3_validation.py
+.venv/bin/python -m floquet_if_manybody.cli audit results
+.venv/bin/python -m floquet_if_manybody.cli paper-audit results/paper
+
+# Pole-resolved heat-valve pilot. The audit intentionally exits nonzero
+# because the dark-channel claim gates are not met.
+.venv/bin/python -m floquet_if_manybody.cli heat-valve --pilot
+.venv/bin/python -m floquet_if_manybody.cli \
+  heat-valve-audit results/heat-valve
+```
+
+The production command is resumable. Its cache is content-addressed by the
+physical model, numerical controls, projected operators, solver revision, and
+source revision. Cache files are excluded from the research archive.
+
+The upstream harness excludes machine-generated `results/` data from Git.
+The commands above reproduce those local files. This submission commits the
+publication figures, reports, test suite, and compact validation snapshots;
+`validation/ARTIFACT_PROVENANCE.json` records hashes for the audited local
+artifact set.
+
+To continue the Kac variants on a cluster:
+
+```bash
+FULL_KAC=1 PYTHON_BIN=.venv/bin/python \
+  scripts/run_paper_extension.sh models
+```
+
+## Result guide
+
+| Artifact | Meaning | Status |
+|---|---|---|
+| `figures/n2_exact.*` | Interacting-triplet gaps and bright weights | Exact |
+| `figures/n3_exact.*` | Collective cat gap and weight | Exact |
+| `figures/paper/n3_sector_heat.*` | \(N=3\) even/odd heat spectra | 6/6 converged |
+| `figures/paper/n3_odd_difference.*` | Odd-sector \(J\)-invariance residual | Exact zero on the projected grid |
+| `figures/paper/error_maps.*` | Uniform TEMPO vs Floquet-Markov/QRT | 9/9 converged |
+| `figures/paper/n3_error_maps.*` | Same-parameter \(N=3\) UniformTEMPO vs Floquet-Markov/QRT | 6/6 converged |
+| `figures/validation/fig3_transversal_summary.*` | Published Fig. 3 bottom data vs independent calculation | 3/3 physical gates passed |
+| `figures/paper/n4_odd_j0p25_comparison.*` | Same-model \(N=4\) odd-sector comparison | Converged; heat error 3.405 |
+| `figures/paper/n4_even_j0p25_comparison.*` | Same-model \(N=4\) even-sector comparison | Converged; heat error 5.494 |
+| `figures/paper/dark_diagnostics.*` | Floquet matrix elements, heat, and \(\mathrm{Var}(S)\) | Converged heat plus exact Floquet diagnostic |
+| `figures/paper/model_variants.*` | Bounded/Kac and counterterm comparison | Bounded converged; Kac compression-audited |
+| `figures/heat-valve/heat_valve_hero.*` | Quasienergy collapse tested against exact transfer-pole residues | Pilot complete; dark-channel claim rejected |
+| `docs/heat-valve-result-zh.md` | Fixed-frequency go/no-go evidence and independent claim audit | 3/3 \(N=3\) pilot points; full grid intentionally not run |
+| `validation/*.json` | Artifact provenance and independent backend checks | Diagnostic validation |
+
+See [the Chinese report](docs/report-zh.md),
+[the heat-valve pilot result](docs/heat-valve-result-zh.md),
+[theory conventions](docs/theory.md), [numerical methods](docs/methods.md), and
+the [completion matrix](docs/completion-matrix.md).
+
+This integrated project is distributed under the repository's
+AGPL-3.0-or-later license. Third-party solvers retain their own licenses.

tracks/mps/solutions/Ranger-123/configs/n2_baseline.yaml

@@ -0,0 +1,13 @@
+model:
+  n: 2
+  j: 0.5
+  omega: 1.0
+  drive_amplitude: 0.2
+  drive_frequency: 0.6180339887498949
+  normalization: bounded
+bath:
+  alpha: 0.05
+  cutoff: 2.5
+  temperature: 0.0
+sector: triplet
+method: finite_memory_if

tracks/mps/solutions/Ranger-123/configs/n3_even_baseline.yaml

@@ -0,0 +1,13 @@
+model:
+  n: 3
+  j: 0.5
+  omega: 1.0
+  drive_amplitude: 0.2
+  drive_frequency: 0.4450418679126287
+  normalization: bounded
+bath:
+  alpha: 0.05
+  cutoff: 2.5
+  temperature: 0.0
+sector: even
+method: floquet_markov

tracks/mps/solutions/Ranger-123/configs/n3_odd_check.yaml

@@ -0,0 +1,9 @@
+model:
+  n: 3
+  j: 1.0
+  omega: 1.0
+  drive_amplitude: 0.2
+  drive_frequency: 1.0
+  normalization: bounded
+sector: odd
+method: closed

tracks/mps/solutions/Ranger-123/docs/completion-matrix.md

@@ -0,0 +1,70 @@
+# Completion matrix
+
+| Requirement | Evidence | Status |
+|---|---|---|
+| Reproducible Python/Julia package | `pyproject.toml`, `uv.lock`, pinned source in `julia/Project.toml` | Complete |
+| \(N=2\) singlet dark sector | `tests/test_symmetry_n2.py` | Complete |
+| \(N=2\) exact gaps and weights | `tests/test_spectra_n2.py`, `figures/n2_exact.*` | Complete |
+| \(N=3\) \(6\oplus2\) reflection split | `tests/test_symmetry_n3.py` | Complete |
+| \(N=3\) odd-sector \(J\)-independence | projected operator hashes and zero curve difference | Complete |
+| \(N=3\) cat-gap coefficient | `tests/test_spectra_n3.py` | Complete |
+| Closed Floquet solver | `tests/test_floquet.py` | Complete |
+| Public uniform process-tensor backend | pinned UniformTEMPO revision, Julia runner | Complete |
+| Extended-state multi-time insertions | Julia runner and backend tests | Complete |
+| Delta/continuum separation | correlation and heat-current tests | Complete |
+| Atomic, content-addressed cache | convergence/backend tests | Complete |
+| Nested compression/timestep/phase controller | adaptive tests and per-point evidence | Complete |
+| \(N=3\) even/odd heat grid | `figures/paper/n3_sector_heat.*`, audited local manifest | 6/6 converged |
+| Odd-sector spectral invariance | relative maximum difference 0 | Complete |
+| \(N=2\) exact-vs-Markov grid | `error_map_manifest.json` | 9/9 converged |
+| Three Markov-error metrics | all nine cells | Complete |
+| \(N=3\) same-model exact-vs-Markov grid | `validation/n3_same_model_error_map.json`, `figures/paper/n3_error_maps.*` | 6/6 converged |
+| Floquet dark-channel diagnostics | \(|m|\le40\), Parseval tests | Complete diagnostic |
+| Independent drive/bath normalization | model and projected-sector regression tests | Complete |
+| Floquet transfer eigenvalues and residuals | Julia Krylov extraction plus real \(N=1\) smoke | Complete |
+| Observable pole-residue fit and mode tracking | synthetic exact-recovery and matching tests | Complete |
+| Fixed-frequency \(N=1,2,3\) coherent-destruction scan | `docs/heat-valve-result-zh.md`, hero figure | Complete |
+| \(N=3\) pole-resolved heat-valve pilot | report, hero figure, and audited local manifest | 3/3 executed; claim gates rejected |
+| Nine-point \(N=1,2,3\) heat-valve grid | go/no-go required threefold heat and residue suppression | Intentionally not run: pilot failed |
+| Bounded normalization/counterterm variants | full adaptive evidence | 2/2 converged |
+| Kac normalization/counterterm variants | passed compression evidence | Local endpoint complete; full timestep/phase requires cluster |
+| Independent single-spin validation | `validation/uniform_tempo_single_spin.json` | Passed |
+| Published Fig. 3 bottom-panel reproduction | `validation/fig3_transversal_summary.json`, `figures/validation/fig3_transversal_summary.*` | 3/3 physical gates passed |
+| Independent OQuPy cross-check | `validation/uniform_tempo_oqupy_crosscheck.json` | Diagnostic, not convergence |
+| Publication PNG/PDF figures | `figures/paper` | Visually checked |
+| Pole-resolved hero figure | `figures/heat-valve/heat_valve_hero.{png,pdf}` | Visually checked; labeled candidate/rejected |
+| Chinese report | `docs/report-zh.md` | Complete |
+| Heat-valve negative-result report | `docs/heat-valve-result-zh.md` | Complete |
+| \(N=4\) reflection implementation | `tests/test_symmetry_n4.py`, convergence-gated `n4-pilot` command | Complete, \(10\oplus6\) blocks |
+| \(N=4\) reflection-odd pilot and same-model benchmark | `figures/paper/n4_odd_j0p25_comparison.*` | Converged; \(\epsilon_j=3.405\) |
+| \(N=4\) reflection-even pilot and same-model benchmark | six-period tail-window evidence, comparison figure | Converged; \(\epsilon_j=5.494\) |
+| Continuum and critical exponent | Outside finite-system scope | Not claimed |
+
+## Final gates
+
+```bash
+.venv/bin/python -m pytest -q
+.venv/bin/ruff check src tests scripts/run_uniform_validation.py
+.venv/bin/mypy src
+.venv/bin/python -m floquet_if_manybody.cli audit results
+.venv/bin/python -m floquet_if_manybody.cli paper-audit results/paper
+```
+
+`paper-audit` requires all six \(N=3\) points and all nine error-grid points to
+be converged. It also requires both bounded model variants to be fully
+converged. The two Kac variants are accepted only as an explicitly declared
+timestep resource ceiling after a passed compression comparison; they are
+never re-labeled as converged.
+
+The additional command
+
+```bash
+.venv/bin/python -m floquet_if_manybody.cli \
+  heat-valve-audit results/heat-valve
+```
+
+is expected to return nonzero for the archived pilot. This is a scientific
+result, not a software failure: heat is not suppressed against both flanks,
+the visible pole residue increases, the full nine-point grid was therefore
+not authorized by the predeclared resource gate, and the figure is labeled
+`candidate; claim gates not met`.