OpenPhysics · veillette · Aug 4, 2026 · Aug 4, 2026
@@ -4,9 +4,9 @@ Sim-specific context for AI assistants. General SceneryStack guidance: [OpenPhys
 
 ## Project
 
-A four-screen simulation of flat-spacetime special relativity: **Light Clock**, **Spacetime
-Diagram**, **Twin Paradox**, **Relativistic Doppler**. Original work, not a PhET or NAAP port.
-Forked from `SceneryStackTemplate` on 31 Jul 2026.
+A five-screen simulation of flat-spacetime special relativity: **Light Clock**, **Spacetime
+Diagram**, **Length Contraction**, **Twin Paradox**, **Relativistic Doppler**. Original work, not a
+PhET or NAAP port. Forked from `SceneryStackTemplate` on 31 Jul 2026.
 
 Read [`doc/model.md`](doc/model.md) before changing anything physical, and
 [`doc/implementation-notes.md`](doc/implementation-notes.md) before changing anything structural.
@@ -33,8 +33,9 @@ Read [`doc/model.md`](doc/model.md) before changing anything physical, and
 | `src/common/view/controlHelpers.ts` | `createNumberControl` / `createCheckbox` / `createReadoutRow` — the controls every screen shares |
 | `src/common/TimeModel.ts` | Composable clock: play/pause, speed, `scaledDt`, step forward/back |
 | `src/SpecialRelativityColors.ts` | `ProfileColorProperty` table **and the sim's colour language** — read its header before adding a colour |
-| `src/SpecialRelativityConstants.ts` | Grouped `as const` blocks (`DIAGRAM`, `EVENT`, `LIGHT_CLOCK`, `TWIN`, `DOPPLER`, `FONTS`) |
+| `src/SpecialRelativityConstants.ts` | Grouped `as const` blocks (`DIAGRAM`, `EVENT`, `LIGHT_CLOCK`, `LADDER_BARN`, `TWIN`, `DOPPLER`, `FONTS`) |
 | `src/light-clock/model/lightClockGeometry.ts` | Photon height, tick counts, the zigzag trail, the light-travel triangle |
+| `src/length-contraction/model/ladderBarnGeometry.ts` | Contracted lengths, both frames' snapshots, the two door-slam events, the fitting verdicts, world sheets, and the simultaneity slices in lab coordinates |
 | `src/twin-paradox/model/twinJourney.ts` | Both worldlines, proper times, the simultaneity jump, the pulses the twins exchange |
 | `src/relativistic-doppler/model/dopplerGeometry.ts` | Retarded emission solve (for an arbitrary observer position), received signal, wavefronts, beaming lobe |
 
@@ -72,6 +73,20 @@ Read [`doc/model.md`](doc/model.md) before changing anything physical, and
   exactly that instant, and `β·t_wrap` rounds onto either side of the modulo — the answer can flip to
   the far end of the rail. Use `traverseStartPosition()`, which returns the rail end exactly. This was
   a real bug in `photonTrail`, caught by the light-clock triangle's structural test.
+- **The Length Contraction screen's clock is one number read by two frames.** `sceneTimeProperty` is
+  barn time `ct` or ladder time `ct′` depending on the toggle. This is legitimate only because both
+  frames' clocks are zeroed on the same event — the ladder's centre passing the barn's centre — which
+  is the one instant they can agree to label. Do not add a second clock.
+- **The Length Contraction diagram is always in barn-frame coordinates**, and the frame toggle changes
+  exactly one thing on it: the tilt of the simultaneity slice. Its `betaProperty` is therefore a
+  derived 0-or-β, not the model's β. Switching the diagram's coordinates with the toggle would destroy
+  the screen's point, which is that both frames are describing the same picture.
+- **That screen has no scrubber, on purpose.** At high β in the ladder frame the window is set by the
+  slams (γβB apart), not by the fly-past, so a fixed-range slider would have had a few percent of
+  useful travel. The two "go to slam" buttons replace it and teach better: in the barn frame they land
+  on the same instant.
+- **Its β is capped at 0.95, not the sim-wide 0.99**, and floored at 0.1. Documented in
+  `LADDER_BARN`; both bounds are about the animation window, not the arithmetic.
 - **The Doppler screen uses the retarded emission event**, not the source's current position. That is
   what makes the transverse redshift come out at exactly γ — from wherever the observer is standing.
 - **Beaming is D⁴** (bolometric flux). D³ and D² are also correct, for other measured quantities; the
@@ -96,12 +111,13 @@ Full convention: [Baton/ACCESSIBILITY.md](https://github.com/OpenPhysics/Baton/b
 
 ## Testing
 
-`npm test` — Vitest, `happy-dom`, `--expose-gc`. 129 tests across six files.
+`npm test` — Vitest, `happy-dom`, `--expose-gc`. 151 tests across seven files.
 
 | Path | Purpose |
 |---|---|
 | `tests/lorentz.test.ts` | Kinematics: γ, boosts, invariance, causal structure, velocity addition, Doppler, aberration, beaming |
 | `tests/lightClockGeometry.test.ts` | Photon path, tick counts, and the independent "the photon travels at c" check |
+| `tests/ladderBarnGeometry.test.ts` | Contraction by γ, both frames' fitting verdicts re-derived by sweeping the drawn snapshots, the slams' spacelike separation and invariant interval, and the check that no door shuts through the ladder |
 | `tests/twinJourney.test.ts` | Proper times and the Earth-time accounting identity at the turn |
 | `tests/dopplerGeometry.test.ts` | The retarded solve, the three Doppler limits, wavefronts, the beaming lobe |
 | `tests/TimeModel.test.ts` | The shared clock |
@@ -125,8 +141,17 @@ Query parameters: `?initialBeta=0.8`, `?showRapidity=true`, `?shadeLightCone=tru
 
 After `npm run build`, the sim is installable offline via Workbox (`dist/manifest.webmanifest`).
 
-## Ideas for a fifth screen
+## Ideas for a sixth screen
 
-Length contraction is the one standard topic this sim deliberately avoids — every screen is arranged
-so that it does not enter. A "ladder and barn" screen would reuse `MinkowskiDiagramNode` and
-`lorentz.ts` almost unchanged.
+The four topics still untouched, roughly in order of how much they would reuse:
+
+- **Velocity addition.** `velocityAddition()` already exists in `lorentz.ts` and is unit-tested but is
+  not on screen anywhere. A rocket firing a probe forward, with the two boosts composing on a rapidity
+  scale that *does* add, would need little more than the existing diagram.
+- **Relativity of simultaneity as a train-and-lightning screen.** Cheaper than it sounds — it is the
+  Length Contraction screen's machinery with the two events on a moving object rather than a fixed one.
+- **Momentum and energy.** The one genuinely new module: E = γm, p = γmβ, and E² − p² = m² as a fourth
+  invariant hyperbola to sit beside the one the Spacetime Diagram screen already draws.
+- **Reciprocity of time dilation.** The classic misconception the sim does not yet address head-on: a
+  frame toggle on the Light Clock screen, so that each clock in turn is the one at rest and the *other*
+  is the one running slow. The Length Contraction screen's frame selector is the pattern to copy.
@@ -1,7 +1,8 @@
 # Special Relativity
 
 An interactive simulation of flat-spacetime relativity — a moving light clock, a live Minkowski
-diagram, the twin paradox, and the relativistic Doppler effect — built with
+diagram, the ladder-and-barn paradox, the twin paradox, and the relativistic Doppler effect — built
+with
 [SceneryStack](https://scenerystack.org/), Vite 8, TypeScript 6, and Biome 2.
 
 ## Features
@@ -11,6 +12,10 @@ diagram, the twin paradox, and the relativistic Doppler effect — built with
 - **Spacetime Diagram** — drag two events on a Minkowski diagram while a velocity slider shears the
   primed axes live; light cone, lines of simultaneity, invariant hyperbolas, and a readout that
   contrasts the invariant interval with the frame-dependent order of events
+- **Length Contraction** — a ladder flying through a barn whose two doors slam together, watched
+  from the barn's frame and then from the ladder's; the same two slams, one pair of events, and two
+  frames that answer "did it fit?" differently because they slice spacetime into "nows" at
+  different angles
 - **Twin Paradox** — place the turnaround event and watch both worldlines, both clocks, and the
   jump in the traveller's "now" that resolves the paradox
 - **Relativistic Doppler** — a source flying past an observer, with wavefronts, the received colour
@@ -19,7 +24,7 @@ diagram, the twin paradox, and the relativistic Doppler effect — built with
 - English, Spanish, and French localization via `StringManager`
 - Full keyboard access, screen-reader summaries, and default/projector colour profiles
 - Progressive Web App (installable, offline-capable)
-- 100 unit tests over the pure physics modules, plus a memory-leak suite
+- 151 unit tests over the pure physics modules, plus a memory-leak suite
 
 ## Quick Start
 

@@ -27,6 +27,7 @@ src/
       controlHelpers.ts, chartUtils.ts
   light-clock/          model/{LightClockModel,lightClockGeometry}  view/…
   spacetime/            model/SpacetimeDiagramModel                 view/…
+  length-contraction/   model/{LengthContractionModel,ladderBarnGeometry} view/…
   twin-paradox/         model/{TwinParadoxModel,twinJourney}        view/…
   relativistic-doppler/ model/{RelativisticDopplerModel,dopplerGeometry} view/…
 tests/                  one file per pure module, plus memory-leak
@@ -36,11 +37,12 @@ tests/                  one file per pure module, plus memory-leak
 
 ### Pure functional physics, Property layers on top
 
-`lorentz.ts`, `lightClockGeometry.ts`, `twinJourney.ts` and `dopplerGeometry.ts` are plain functions
+`lorentz.ts`, `lightClockGeometry.ts`, `ladderBarnGeometry.ts`, `twinJourney.ts` and
+`dopplerGeometry.ts` are plain functions
 of plain numbers and `Vector2`s. They import from `scenerystack/dot` and nothing else — no axon, no
 scenery. Everything reactive lives in the model classes that wrap them.
 
-This is what makes the physics testable without SceneryStack, and it is where all 100 unit tests
+This is what makes the physics testable without SceneryStack, and it is where all 151 unit tests
 point. It follows `CarnotHeatEngine/src/common/model/carnotCycleGeometry.ts`.
 
 ### Everything animated is a closed form of elapsed time
@@ -138,6 +140,38 @@ mouse would never see them:
 Both default to **zero** in the pure functions, so the tests check exact behaviour; only the model
 layer passes a positive value. Do not push the tolerance into `lorentz.ts`.
 
+### The Length Contraction screen has one clock, read by two frames
+
+`LengthContractionModel.sceneTimeProperty` is a single number, interpreted as barn time `ct` or as
+ladder time `ct′` depending on which frame is selected. That is not a shortcut. Both frames' clocks
+are zeroed on one event — the ladder's centre passing the barn's centre — so there is exactly one
+instant they can label the same, and it is the one the clock is zeroed on. Flipping the frame toggle
+without touching the clock is then a meaningful operation, and it is the operation the screen exists
+to offer: the number stays put and the scene rearranges itself around it.
+
+The clock wraps rather than scrubs. Its window, `sceneHalfWindow()`, is derived from the setup and
+the frame, and at high β in the ladder frame it is set by the *slams* (γβB apart) rather than by the
+fly-past — so a fixed-range slider would have had a useful travel of a few percent of its track. Two
+push buttons take the clock straight to each slam instead, which is also the better teaching control:
+in the barn frame both buttons land on the same instant, and that is the whole content of "the doors
+are on one switch".
+
+`advance()` uses a modulo rather than a comparison against the ends, so a large `dt` after a
+background tab regains focus lands in the right place instead of skipping the window.
+
+### The Length Contraction diagram never changes frames
+
+The stage is drawn in the selected frame; the spacetime diagram is always in barn-frame coordinates,
+and the toggle changes exactly one thing on it — the tilt of the simultaneity slice. `betaProperty`
+for the shear is therefore *not* the model's β but a DerivedProperty that is β in the ladder frame and
+0 in the barn frame, because in the barn frame the primed mesh would be the unprimed mesh already
+drawn.
+
+That split is the argument the screen makes, and it is why `ladderSliceInLab()` returns **barn-frame**
+events for a measurement taken in either frame: a length is two ends at one instant, and the ladder
+frame's pair of ends lands on the diagram as a tilted segment. Its being shorter against the barn's
+upright strip is the disagreement drawn rather than asserted.
+
 ### The Twin Paradox screen has no `SpecialRelativityModel`
 
 Its β is *derived* from the turn's position, not chosen. A trip is specified by where and when you
@@ -202,7 +236,7 @@ is more than one flat list keeps legible. This is a documented variation on the
 ## Testing
 
 `npm test` runs Vitest over `tests/`, environment `happy-dom`, with `--expose-gc` for the memory-leak
-suite. 129 tests across six files.
+suite. 151 tests across seven files.
 
 The house style is three layers per physics module: