Scientific status

This page is the authoritative public map from SPM-Kit capabilities to the evidence that currently supports them. It deliberately separates implementation, software tests, numerical recovery, external comparison, and physical validation.

Maturity vocabulary

0CLAIMEDDocumented intent only

1SOFTWARE_VERIFIEDScoped automated tests

2NUMERICALLY_VERIFIEDKnown-value recovery

3CROSS_VALIDATEDFrozen external comparison

4PHYSICALLY_VALIDATEDNot generally claimed

5REPRODUCIBILITY_VALIDATEDNot generally claimed

Level	Required meaning
CLAIMED	A claim or intended behavior is documented, but executable evidence is not preserved.
SOFTWARE_VERIFIED	The behavior is exercised by automated software tests.
NUMERICALLY_VERIFIED	Known values are recovered within a stated numerical scope and tolerance.
CROSS_VALIDATED	Results are compared with an external software or reference route under a frozen protocol.
PHYSICALLY_VALIDATED	A physical reference or experiment supports the capability within a stated scope. No general SPM-Kit claim is current.
REPRODUCIBILITY_VALIDATED	An independent party reproduced the result under the declared protocol. No general SPM-Kit claim is current.

The level belongs to a particular claim, campaign, data family, metric, version, and tolerance. It never transfers automatically to an adjacent feature.

Capability matrix

Capability	Implementation	Evidence	Level	Independent reference	Known limitation / next requirement
Sa, Sq, Sz on frozen synthetic matrices	core.analysis.roughness	48 cases, 144/144 comparisons within atol=1e-6 nm + rtol=1e-6	CROSS_VALIDATED	Gwyddion 2.71 library route through a frozen harness-authored wrapper	Shared matrices, no preprocessing, three metrics only; not physical validation or universal equivalence
Sa, Sq, Sz on public experimental GWY matrices	core.analysis.roughness	12 cases, 36/36 shared-matrix comparisons within tolerance	CROSS_VALIDATED for the shared-matrix algorithm track	Gwyddion 2.71	Parser/end-to-end observations are separate: 10 equivalences and 2 preserved channel-count differences
Limited Nanoscope III .spm images	core.io.bruker_spm	Six demonstrated files; 18/18 Sa/Sq/Sz comparisons within tolerance and zero reported pixel delta	NUMERICALLY_VERIFIED	Gwyddion 2.71	ACCIDENTAL_PRE_FREEZE_UNBLINDING; partial variants only, no blind holdout or general Bruker support
NanoSurf .nid mapping and orientation	core.io.nid, core.verify	Synthetic byte-budget/orientation tests and selected lab-context comparisons	SOFTWARE_VERIFIED; selected comparisons do not establish universal format coverage	Gwyddion exports for selected files	Private instrument corpus is not distributed; additional redistributable multi-instrument fixtures are needed
Gwyddion Flatten Base end-to-end trajectory	core.analysis._flatten_base	Focal LM and packed-Cholesky verification plus a frozen Gwyddion 2.71 end-to-end fixture; matching facet/polynomial control flow and corrected-field maximum absolute difference 1.465494e-14	CROSS_VALIDATED	Gwyddion 2.71 executable	Internal Core path and one frozen end-to-end trajectory plus focused numerical cases; no universal equivalence claim across datasets, parameter regimes, platforms or Gwyddion versions
Physical arc-revolution background	core.analysis.background	55 unit and synthetic tests, including a test-local brute-force 1D oracle, inversion duality, physical-unit equivalence, anisotropic spacing, border policies and reconstruction identity	SOFTWARE_VERIFIED	None	Python API only; finite geometric Z data; no masks, Gwyddion equivalence or physical-reference campaign
Gwyddion-compatible Revolve Arc background	core.analysis.background, core.analysis._gwyddion_arc_revolution	Frozen Gwyddion 2.71 source semantics, focal kernel probes, one asymmetric 5×7 directional fixture, 6/6 background routes and 5/5 valid corrected routes within 5e-14; repaired reconstruction for the defective horizontal-inverted wrapper	CROSS_VALIDATED for the frozen campaign	Gwyddion 2.71 source, compiled probes and frozen JSON/NPZ fixture	Radius is in samples; no masks or non-finite data; one-sample processing axes use a documented safe definition; no physical validation, tip reconstruction, performance equivalence or universal-equivalence claim
Physical sphere-revolution background	core.analysis.background	51 unit and synthetic tests, including independent brute-force 2D oracles for nearest and reflect borders, physical anisotropy, non-separability, unit equivalence and reconstruction identity	SOFTWARE_VERIFIED	None	Python API only; finite geometric Z data; no masks, Gwyddion equivalence, performance campaign or physical-reference campaign
Gwyddion-compatible Sphere-revolution background	core.analysis.background, core.analysis._gwyddion_sphere_revolution	Frozen Gwyddion 2.71 source semantics, focal probes, 10 original surfaces, 10 normal executions on negated inputs (20 valid external runs per build), 15/15 inverted runs failing in normal build and under ASan; direct external reference for normal, derived external cross-validation for inverted background, safe deliberate divergence for inverted corrected (atol=5e-14, rtol=0.0); independent Python oracle	CROSS_VALIDATED for the frozen campaign	Gwyddion 2.71 source, compiled probes, independent Python oracle and frozen JSON/NPZ fixture	Radius is in samples; no non-finite data or masks; inverted corrected does not claim equivalence with Gwyddion's crashing wrapper; no physical validation, tip deconvolution or universal-equivalence claim; physical sphere-revolution maintains its independent software verification
Gwyddion 2.71 Median Background	core.analysis.background, core.analysis._median_background	Frozen executable reference campaign: 36 logical cases, 72 executions (36 normal, 36 ASan), radii 1/2/3/4/20/1024, direct and radixtree reference paths; public background and corrected fields 36/36 bitwise exact, maximum absolute difference 0 and maximum ULP 0; input mutation maximum 0 and reconstruction maximum 4.4408920985006262e-16	CROSS_VALIDATED within the frozen 36-case campaign	Gwyddion 2.71 source, executable probe, independent Python oracle, frozen NPZ/JSON fixture	Finite two-dimensional inputs only; no universal equivalence, performance-equivalence, future-Gwyddion, all-radii, or all-matrices claim; rank_backend_reference describes Gwyddion, not an SPM-Kit backend
Gwyddion 2.71 Filter flat-disc morphology	core.analysis.background, core.analysis._gwyddion_flat_disc_morphology	Frozen executable reference campaign: 12 fields, six sizes 2/3/4/5/30/31, 72 Opening and 72 Closing cases; kernels 30/30, Opening 72/72 and Closing 72/72 bitwise exact; maximum absolute difference 0, maximum ULP 0, signed-zero mismatches 0, input mutation 0	CROSS_VALIDATED within the frozen campaign	Audited Gwyddion 2.71 executable, corrected external probe V3, executable reduction trace, independent oracle V2, frozen NPZ/JSON fixture	Finite full-field data with masks ignored; no universal equivalence, NaN/Inf, ROI, masks, ASF, tip morphology, physical rolling-ball, performance, other builds/versions, public erosion/dilation, or source-only tie claim
Gwyddion 2.71 Path Level	core.analysis.leveling, core.analysis._gwyddion_path_level	Audited executable campaign: 18 base families, thicknesses 1/2/3/128, 72 logical cases, 144 fresh external executions and 72 deterministic repeat pairs; private and public arrays 72/72 bitwise exact, 4,652/4,652 elements exact, max absolute/ULP 0, signed-zero mismatches 0, normalized endpoints and mutation/no-op classifications 72/72	CROSS_VALIDATED within the frozen campaign	Audited Gwyddion 2.71 Path Level tool, external probe, independent oracle V1, frozen NPZ/JSON fixture	Finite non-empty full fields and ordered straight selections only; no universal equivalence, NaN/Inf, masks/ROI, paths/splines, profiles, align-rows, volume, GUI, performance, or other-build/version claim
Gwyddion 2.71 Align Rows statistics	core.analysis.leveling, core.analysis._gwyddion_align_rows_statistics	Public 64-case finite campaign: portable source semantics 64/64 arrays and 3,888/3,888 elements bitwise exact; installed fast-math profile 61/64 arrays and 3,757/3,888 elements exact, with only three signed-zero and 128 independently explained reassociation differences	CROSS_VALIDATED within the frozen dual-profile campaign	Gwyddion 2.71 source, external executable probe, independent portable V2 oracle, frozen NPZ/JSON fixture, installed-build diagnosis	Four methods only; finite full fields, frozen masks/directions/trims; no universal, non-finite, performance, other-version/build, GUI, or generic-align_rows compatibility claim
Hertz / conical contact and DMT paths	core.analysis.forcecurve	Unit and synthetic-recovery tests; Hertz/conical modulus recovery gates	NUMERICALLY_VERIFIED within synthetic test scope	Analytical construction	No certified cantilever/tip calibration or broad experimental campaign
Adhesive JKR	core.analysis.experimental	Synthetic recovery of reduced modulus and work of adhesion; Hertz-limit test	NUMERICALLY_VERIFIED within synthetic scope	Analytical construction	Experimental module; no physical-reference campaign
WLC and FJC chain models	core.analysis.chain	Analytical synthetic-recovery tests	NUMERICALLY_VERIFIED within synthetic scope	Analytical construction	No cross-software or experimental population campaign
SLS relaxation	core.analysis.experimental	Synthetic recovery of characteristic time and elastic limit	NUMERICALLY_VERIFIED within synthetic scope	Analytical construction	Loader/UI path for pause segments remains limited
Force-volume property maps	core.analysis.forcevolume*	Scalar/vectorized consistency and synthetic modulus-map recovery tests	NUMERICALLY_VERIFIED within synthetic scope	Internal scalar and analytical controls	Not an independent reference; experimental calibration remains user supplied
KPFM / CPD	core.analysis.kpfm	Unit tests for CPD and work-function calculation	SOFTWARE_VERIFIED	None	No public physical-reference campaign
PSD, correlation, fractal estimates	core.analysis.spectral	Unit and synthetic tests	SOFTWARE_VERIFIED	None	Definitions and preprocessing must be matched before external comparison
Grain detection	core.analysis.grains	Synthetic morphology tests	SOFTWARE_VERIFIED	None	Threshold-dependent; no universal segmentation benchmark
SHO resonance and thermal calibration utilities	core.analysis.resonance, calibration	Unit and numerical tests; limited experimental exercise is documented separately	SOFTWARE_VERIFIED; no general physical claim	None for a frozen public campaign	Requires a redistributable calibrated reference and frozen acquisition context
Fathom workspace	gui/ over public core APIs	Offscreen GUI tests plus architecture tests	SOFTWARE_VERIFIED	None	Platform/Qt behavior can vary; GUI evidence does not upgrade numerical maturity

Canonical campaign records

• 48-case Gwyddion roughness summary
• Public experimental GWY pilot summary
• Nanoscope .spm pilot summary
• Nanoscope incident and final audit
• Flatten Base Gwyddion 2.71 frozen end-to-end fixture
• Sphere Revolution Gwyddion 2.71 frozen fixture
• Median Background Gwyddion 2.71 frozen manifest

Gwyddion Sphere Revolution

SPM-Kit's gwyddion_sphere_revolution implementation is maintained separately from physical sphere revolution. It reproduces Gwyddion 2.71's Revolve Sphere numerical semantics: - Campaign Scope: 10 original surface matrices across WIDE_ASYMMETRIC, TALL_ASYMMETRIC, CONSTANT_ZERO_RMS, and SIGNED_MICRO_GRID families, plus 10 normal executions on explicitly negated inputs (20 valid external runs per build). - Inverted Route Failure Evidence: 15/15 executions of inverted=True crash in Gwyddion 2.71's C module (exit code 139 in normal build, exit code 134 under ASan at sphere-revolve.c:328 gwy_data_field_subtract_fields). - Evidence Classes: - normal route: Direct external reference from Gwyddion 2.71 stdout. - inverted background: Derived external cross-validation from -B(-data). - inverted corrected: Safe deliberate divergence (original - background) reconstructing original data without invoking Gwyddion's crashing subtract wrapper. - Independent Oracle & Tolerances: Verified against an independent Python oracle (atol=5e-14, rtol=0.0). The maximum observed numerical discrepancy across all comparisons is 8.881784e-16 (well below 5e-14). - Claim: LEVEL 3 CROSS_VALIDATED within the frozen fixture scope. No universal equivalence or physical validation is claimed. Physical sphere revolution (estimate_sphere_revolution_background) maintains its independent software verification.

Gwyddion 2.71 Median Background

Claim: CROSS_VALIDATED only for the frozen campaign. Gwyddion 2.71 is the executable external reference; the campaign contains 36 logical cases and 72 executions (36 normal and 36 ASan) over radii 1, 2, 3, 4, 20, and 1024. Both Gwyddion reference paths are represented: direct and radixtree. Public estimate_gwyddion_median_background, remove_gwyddion_median_background, and analyze_gwyddion_median_background reproduce the frozen background and corrected arrays bitwise in 36/36 cases: maximum absolute difference 0, maximum ULP 0, input mutation maximum 0, and reconstruction maximum 4.4408920985006262e-16.

The fixed semantics are the inclusive digital ellipse, gwyddion_border_extend, middle rank kernel_active_count//2, and corrected = input - background. The public API preserves the SPMChannel context and requires finite two-dimensional data. The private kernel is independent of Gwyddion at runtime. The fixture and independent Python oracle remain frozen evidence outside production.

Traceability:

Gwyddion 2.71 source: modules/process/median-bg.c
  → frozen external probe: median_background_behavior_probe.c
  → frozen campaign runner: run_median_background_probe_campaign.sh
  → independent Python oracle recorded by docs/design/GWYDDION_MEDIAN_BACKGROUND_COMPATIBILITY.md
  → tests/validation/fixtures/gwyddion/median_background/median_background_reference.npz
  → tests/validation/fixtures/gwyddion/median_background/median_background_reference.json
  → src/spmkit/core/analysis/_median_background.py
  → src/spmkit/core/analysis/background.py
  → tests/core/test_gwyddion_median_background_private.py
  → tests/core/test_gwyddion_median_background.py
  → tests/validation/test_median_background_fixture_integrity.py
  → docs/scientific-status.md

The evidence was frozen in 818dbd3 (freeze evidence), the private kernel in a53c3bb, and the public API in ed5c837. The focal inventory is 20 fixture-integrity, 67 private Median Background, and 72 public Median Background tests; the preceding combined focal run collected 442 tests. These are focal-campaign counts, not a project-wide total.

Non-claims: no universal equivalence; no guarantee outside the 36 cases; no NaN or infinity coverage and no reproduction of Gwyddion's internal radixtree; no performance-equivalence claim; no claim for future Gwyddion versions; no claim for every radius or matrix; and no validation of configurable border, shape, or rank parameters because the API exposes none.

Non-blocking tooling limitations: the probe runner uses || true during compilation and does not retain the original compiler exit code; its auxiliary parser recognizes broad background_ and corrected_ prefixes. The campaign remains valid because both binaries executed, 72 processes returned, stderr was empty, normal and ASan stdout were byte-identical, outputs were parsed and recalculated independently, oracle/reference results were bitwise exact, and the fixture stores canonical hashes.

Gwyddion 2.71 Filter flat-disc morphology

Claim: CROSS_VALIDATED only within the frozen executable campaign. The audited Gwyddion 2.71 Filter path is represented by 12 deterministic finite fields at sizes 2, 3, 4, 5, 30, and 31. Opening and Closing each match the frozen external outputs bitwise in 72/72 cases; kernels match 30/30, maximum absolute difference and ULP distance are both 0, signed-zero mismatches are 0, and input mutation is 0.

The fixed semantics are the K×K inclusive digital ellipse, nearest-edge extension, asymmetric even-size anchors, and the audited executable Each/Even plus RLE reduction hierarchy. The source strict ternaries and executable MINSD/MAXSD equality behavior are distinct; SPM-Kit reproduces the audited executable path. The rejected uninitialised-kernel microprobe is not evidence; the corrected zero-initialised probe is the valid external record.

Traceability:

Gwyddion 2.71 source: libprocess/filters-minmax.c
  → frozen external probe: flat_disc_probe_v3
  → frozen reduction trace: flat_disc_reduction_trace_v1
  → independent oracle: flat_disc_morphology_oracle.py
  → tests/validation/fixtures/gwyddion/flat_disc_morphology/flat_disc_morphology_reference.npz
  → tests/validation/fixtures/gwyddion/flat_disc_morphology/flat_disc_morphology_reference.json
  → src/spmkit/core/analysis/_gwyddion_flat_disc_morphology.py
  → src/spmkit/core/analysis/background.py
  → tests/core/test_gwyddion_flat_disc_morphology_private.py
  → tests/core/test_gwyddion_flat_disc_morphology.py
  → docs/scientific-status.md

Evidence was frozen in 2ba366e; the private kernel is 05c5ae4; the public API and documentation are recorded in 1b2d081 and c0de811. Non-claims: no universal equivalence; no NaN or infinity coverage; no ROI, masks, ASF, tip morphology, physical rolling-ball equivalence, performance parity, other Gwyddion builds or versions, public erosion or dilation, or claim that source-level C tie semantics alone reproduce the audited binary.

Gwyddion 2.71 Path Level

Claim: CROSS_VALIDATED only within the frozen Path Level campaign against the audited Gwyddion 2.71 tool module tools.so (4711c360dd42e3e16257bf0e86d8bd41852b43d1d34540bf097736a603146237, Build ID 600b16d9857946609b567704b406abcc74aea698). The campaign contains 18 finite, non-empty, full-field base families, thicknesses 1, 2, 3, and 128, 72 logical cases, 144 fresh external executions, and 72/72 deterministic repeat pairs. Private and public gwyddion_path_level arrays are bitwise exact in 72/72 cases and 4,652/4,652 elements: maximum absolute difference 0, maximum ULP 0, signed-zero mismatches 0, normalized endpoints 72/72, mutation/no-op classification 72/72, and input mutation 0.

The operation consumes ordered GwySelectionLine-equivalent straight physical-coordinate segments; duplicates and object order are significant. Its fixed executable semantics include endpoint conversion, horizontal-line exclusion, C truncating division, inclusive thickness windows, and left-to-right cumulative row correction. Gwyddion mutates the selected data field in place and performs GUI publication, undo, and logging; SPM-Kit returns a new SPMChannel and claims no GUI-publication parity.

Traceability:

Gwyddion 2.71 source: modules/tools/pathlevel.c
  → installed Gwyddion 2.71 Path Level tool execution
  → frozen external probe: path_level_probe_v1
  → independent oracle: path_level_oracle_v1
  → tests/validation/fixtures/gwyddion/path_level/path_level_reference.npz
  → tests/validation/fixtures/gwyddion/path_level/path_level_reference.json
  → src/spmkit/core/analysis/_gwyddion_path_level.py
  → src/spmkit/core/analysis/leveling.py
  → tests/core/test_gwyddion_path_level_private.py
  → tests/core/test_gwyddion_path_level.py
  → tests/validation/test_path_level_fixture_integrity.py
  → docs/scientific-status.md

The evidence commit is d3566ce; the private-kernel commit is 4ead95b. No future public or documentation commit hash is claimed. Non-claims: no universal equivalence; no NaN or infinity coverage; no masks or ROI; no GwySelectionPath, splines, or polylines; no profile extraction, align-rows equivalence, volume line-leveling, GUI/undo/logging/selection-widget parity, performance parity, or guarantee for other Gwyddion versions or builds.

Gwyddion 2.71 Align Rows statistics

Claim: CROSS_VALIDATED only within the frozen finite 64-case public campaign, with sixteen cases each for Median, Median of differences, Trimmed mean, and Trimmed mean of differences. The production contract is portable_source_semantics: the public wrappers are bitwise exact to the independent V2 oracle in 64/64 corrected arrays and 3888/3888 elements, retaining all frozen mask modes, absent-mask routes, horizontal/vertical orientations, trim fractions 0.0, 0.05, and 0.5, mutation/no-op classifications, and deterministic output. The wrappers return new context-preserving SPMChannel instances and do not claim Gwyddion GUI, publication, undo, or mutation behavior.

The secondary installed_gwyddion_2_71_fast_math_profile is external executable evidence from process.so (Gwyddion 2.71 installed module) (c21d52375807ae096e34a3469c2f20c4c66ea3197479e13215a6d7b9d465b451). It is bitwise exact for 61/64 arrays and 3757/3888 elements. The complete and bounded exception set is three signed-zero-only Median elements in median__plateaus_signed_zero__10 plus 64 finite elements in each of median_of_differences__irregular__11 and trimmed_mean_of_differences__irregular__11; their maximum absolute difference is 5.329070518200751e-15, with no NaN or infinity discrepancy. All eight requested backgrounds (504/504 elements) are bitwise exact and mutation/no-op classifications agree 64/64.

The installed package build diagnosis is INSTALLED_BUILD_ROOT_CAUSE_CONFIRMED and V3_NOT_JUSTIFIED: GCC 16.1.1 -ffast-math, associative floating-point reassociation, and LTO produce the two irregular difference-method residuals. The portable source-semantic arithmetic is deliberate; SPM-Kit does not emulate that package-specific transformation and introduces no named-case or signed-zero patch. The public functions are explicit alternatives to, not a compatibility claim for, the existing generic align_rows.

Traceability:

Gwyddion 2.71 source: modules/process/linematch.c
  → frozen external probe: align_rows_probe_v1
  → independent oracle: align_rows_oracle_stats_v2
  → tests/validation/fixtures/gwyddion/align_rows_statistics/
  → src/spmkit/core/analysis/_gwyddion_align_rows_statistics.py
  → src/spmkit/core/analysis/leveling.py
  → tests/core/test_gwyddion_align_rows_statistics.py
  → docs/design/GWYDDION_ALIGN_ROWS_STATISTICS_COMPATIBILITY.md

Non-claims: no universal equivalence; no NaN/Inf, other Gwyddion version or build, untested matrix, performance, ROI/GUI, adapter, or other Align Rows method-family claim. This finite campaign does not establish physical validation or general SPMKit parity.

Test-count policy

The collection total is measured with:

python -m pytest --collect-only -q

The final collected value counts pytest test items after parametrization. Skipped items remain collected; a parametrized test contributes one item per parameter combination. The README does not duplicate this changing number.

Claims not made

• no certified metrological traceability;
• no universal reader or instrument compatibility;
• no general equivalence with Gwyddion;
• no blind holdout for the Nanoscope campaign;
• no general LEVEL 4 physical-validation claim;
• no LEVEL 5 independent reproducibility claim;
• no endorsement by UTFSM, the SPM Lab, AFM-SPM, AFMReader, TopoStats, or Gwyddion.

When evidence is incomplete, the narrower statement governs.