550 AU

Interactive real-physics visualisation of the Solar Gravitational Lens Telescope (SGLT) mission concept: solar-sail swarms flying to 650+ AU to use the Sun as a gravitational lens and image an exoplanet's surface. Named for the minimum solar gravitational focal distance, 547.8 AU.

Status: all eight acts complete (Tour and Explore), physics tests green, generative audio, credits overlay, honest-scale system with a true-scale toggle. GitHub Pages workflow is prepared but not enabled.

Repo: m4cd4r4/550-AU.

Run it

npm install
npm run dev      # dev server
npm test         # physics test suite (vitest)
npm run build    # type-check + production build

The dev server also serves /spike.html: the standalone Act 5 lensing spike, a permanent dev harness for the Einstein-ring shader with live controls (heliocentric distance, source offset, J2 quadrupole, exposure).

Optional: npm run fetch-assets re-downloads and re-trims the bundled assets (HYG star subset, Sun texture, LASCO C2 corona frame, and the NASA JWST and Voyager render loupes). The repo already includes them.

What it is

Eight acts, from "why JWST cannot image an exoplanet surface" to the Einstein-ring imaging dance at 650 AU and the epilogue's lens-to-lens internet. Each act has a Tour mode (scripted camera, captions) and an Explore mode (free camera, clickable objects). A chapter rail switches between them; global controls pause, scrub, and time-warp.

• Act 0 - The Problem. NASA's 3D JWST model at true 6.5 m scale beside the ghosted outline of the roughly 90 km aperture the same image would need by brute force, plus the contrast inset (a planet is about 1e10 times fainter than its star; a coronagraph buys about 1e6 of that).
• Act 1 - Einstein's Lens. Photon rays bending past the Sun, converging on the focal line that draws itself outward as impact parameter grows. The deflection is dramatised; alpha(b) and z(b) on the HUD are the true values.
• Act 2 - The Focal Line. A story-paced pullback from the Sun to the gravitational focus along a compressed ruler, with the planets on their orbits as real textured surfaces, the Kuiper belt, the heliopause, Voyager 1, and the two lens milestones.
• Act 3 - The Sundiver. The integrated trajectory: sail deployment, perihelion slingshot at 0.1 AU, jettison at year 2, and a speed race against Voyager 1 and chemical propulsion.
• Act 4 - The String of Pearls. Yearly launches marching out at ~25 AU spacing, the pearl inspector (six CubeSats to a 1 m telescope), and the laser relay hopping images home.
• Act 5 - Imaging the Ring. The payoff: the live lensing shader at true angular scale, the 32 km image cylinder, the four-telescope raster dance with corona subtraction, and the map reconstructing pixel by pixel.
• Act 6 - Many Worlds. The real stellar neighbourhood, with focal lines radiating anti-target for Proxima b, TRAPPIST-1 and GJ 273b.
• Act 7 - Epilogue. The lens works both ways: a Sun-to-Alpha-Centauri comm link, then the Sagittarius A* thought experiment.

Honest-scale rules: distances may be compressed for comprehension, but the scale ribbon always declares the mapping, the local compression factor and the true distance, and Act 2's Explore mode offers a true-scale toggle where the solar system shrinks to a dot 550 AU from the focal line. That emptiness is the point.

Audio is generative: a warm, evolving WebAudio pad, a consonant chord of pure-ratio voices with a slow filter drift, a synthetic reverb and sparse chord-tone bells. It shifts colour per act. No sample files, so the repo stays licence-clean. Muted state persists; audio starts only after a user gesture, per browser autoplay policy.

In Tour mode the acts auto-advance: when one act's tour finishes it holds briefly, then the show rolls on to the next (looping back at the end). Explore mode, a chapter jump or a scrub hand control back. Two acts add a circular loupe with a real NASA render (public domain): the James Webb telescope in Act 0, and Voyager 1 and 2 in Act 4 as the pearl string overtakes them.

Physics model and what is dramatised

Every on-screen number comes from src/data/mission-facts.json or falls out of the model in src/sim/, which is unit-tested (vitest) against these anchors:

Quantity	Equation	Anchor (test)
Light deflection	alpha = 4GM / (c^2 b)	1.751 arcsec at the solar limb
Focal distance	z(b) = b^2 c^2 / 4GM	z(R_sun) = 547.8 AU
Einstein ring radius	theta_E = sqrt(4GM / (c^2 z))	1.61 arcsec at 650 AU (solar disc 1.48 arcsec)
Lens mapping (shader)	beta = theta - theta_E^2 theta / abs(theta)^2 + J2 quadrupole	ring radius = theta_E within 1%; two arcs off-axis
Sundiver trajectory	RK4: two-body + solar radiation pressure, face-on sail from perihelion	perihelion ~0.1 AU; exit 25-26 AU/yr; min focus in 19-22 yr
Pearl string	yearly launches on the shared trajectory	cruise spacing ~25 AU within 10% of the published figure
Image cylinder	D = D_planet x z / d_source	32 km at 650 AU, 57 km at 1200 AU (Proxima b)
Planet positions	Kepler propagation, JPL approximate elements	Earth at 1 AU, closes orbit in 365.25 d
Star positions	HYG RA/Dec/distance	targets at their true sky directions

Appendix: which visuals are simulated vs dramatised

Simulated live, straight from the equations above:

• The Einstein ring and its off-axis arcs (Act 5 and the spike) are a fullscreen fragment shader inverting the thin-lens mapping per pixel; the GLSL is a transcription of the unit-tested src/sim/lensing.ts.
• The sundiver trajectory (Act 3) is a live RK4 integration; the pearl string (Act 4) is that same trajectory offset by each launch year.
• Ring radius, solar-limb radius, focal distances, image-cylinder diameter and pixel pitch are computed from the current heliocentric distance.
• Star and target directions (Acts 2, 6, 7) use real RA/Dec.

Dramatised, and labelled as such in-app:

• Act 0 draws JWST and the 90 km ring to one true scale (the ratio, 13,846 to 1, is the point), but the host star and planet in the contrast inset are schematic dots.
• Act 1 exaggerates the deflection angle so the bend is visible at all (the true limb deflection is 1.75 arcsec, invisible at this zoom); the HUD carries the true alpha(b) and z(b), and the ribbon reads SCHEMATIC.
• Act 3 draws the spacecraft far above true scale (a metre-class craft is sub-pixel at these distances); the HUD declares the factor and marks it NOT TO SCALE once only the bus remains. The sail lightness number is a tuned effective parameter (recorded in mission-facts.json), because the published exit speed exceeds what the quoted sail loading gives physically. The published timeline figures are quoted as claims; the integrator's own timeline is documented in a clarification appended to docs/BUILD-PROMPT.md.
• Act 4 compresses the 25 AU pearl spacing for the frame; the ribbon declares the compression, and the true spacing is on the HUD. The laser pulse cadence is visual; the true 13 h transmit and 3.2 day light lag are on the HUD and captions.
• Act 5 exaggerates the source-planet angular size and the J2 caustic strength for visibility (both stated on the HUD), draws the imaging grid far coarser than the true 1000 x 1000 raster (declared), and reconstructs the map as a progressive reveal standing in for the deconvolution. The planet surface is procedural and watermarked speculative throughout.
• Act 6 log-compresses the light-year distances so all three targets fit one frame; the directions are real and the ribbon says so.
• Act 7 is schematic: the comm link and the Sagittarius A* scene carry the real figures on cards, with the black hole flagged as a thought experiment beyond current engineering.

Architecture notes

• Vite + TypeScript strict + Three.js, no UI framework. Vitest for physics.
• World positions are double-precision AU on the CPU; rendering is camera-relative (floating origin), so only small float coordinates reach the GPU. The scene spans 0.1 to 1200 AU.
• Star positions are real: the bundled HYG subset (8,922 stars to mag 6.5 plus the mission targets) renders at true RA/Dec, so each target's focal line points anti-target against the real sky.
• One controller per act implements a small Act interface; main.ts routes between them and owns the shared renderer, Sun, starfield, HUD and audio.

Sources

• AstroKobi, "NASA's $40 BILLION Plan To Image Alien Worlds" - the video that surfaced this mission concept and shaped which facts and story beats to research. No narration was reproduced; all copy in this app is original (see the IP note in docs/SIM-PLAN.md)
• Turyshev et al 2020, NIAC Phase III report, arXiv:2002.11871
• Turyshev and Andersson 2002, arXiv:gr-qc/0205126
• FOCAL mission heritage (Maccone, Eshleman)
• NASA MMS formation-flying results
• Plan and build contract: docs/SIM-PLAN.md, docs/BUILD-PROMPT.md

Asset licences: see CREDITS.md. The HYG database is CC BY-SA 4.0 (David Nash, astronexus).