# Cosmic Atlas: models and sources

Nine original educational 3D exhibits. No Einstein Toolkit, GADGET or Athena++
binary is running in the headset. The named projects are research references,
not dependencies or endorsements. The galaxy encounter is captured by our own
headless N-body solver; the other exhibits are analytic or schematic models.

| Exhibit | Model actually implemented | Limits |
|---|---|---|
| Black hole | Procedural emissive accretion disk with differential Keplerian rotation, horizon and photon-sphere markers | Curvature mesh is an analogy. No ray-traced relativistic disk, Doppler spectrum or Einstein-equation solve. Mass uses display scaling. |
| Binary merger | Prescribed shrinking binary orbit, quadrupole-style chirp, damped ringdown | Not an observed event or a calibrated GR waveform. |
| Lensing | Two solutions of the point-mass thin-lens equation; aligned Einstein ring | Broken straight rays through a lens plane, exaggerated angles, no strong-field geodesics. |
| Galaxy encounter | 514 mutually gravitating particles; Plummer softening; kick–drift–kick leapfrog; 241 binary snapshots generated on Hetzner | Normalized G=1; two massive centres; no hydrodynamics, halo distribution or expansion. Not a GADGET dataset. |
| Cosmic web | 6,859 displaced points in a deterministic periodic potential | Zel’dovich teaching approximation, no calibrated cosmic age or post-shell-crossing accuracy. |
| Dark matter halo | Circular test-particle speeds from luminous and extended halo contributions | Parametric rotation curves, no observed galaxy fit or self-consistent dynamics. Visible halo colour is explanatory. |
| Stellar cutaway | Four nested Sun-like zones, granulation shader, energy and convection markers | Not a stellar-evolution or nuclear-reaction solver. Motion is accelerated and illustrative. |
| Reconnection | Prescribed current sheet, field geometry and plasma inflow/exhaust | Does not solve resistive MHD, calculate heating, or reproduce a measured flare. |
| Alfvén wave | Transverse analytic linear ideal-MHD wave; phase speed varies with background field | Normalized density/permeability, visually enlarged displacement; no dissipation or nonlinear coupling. |

The capture metadata is in [data/encounter.json](./data/encounter.json), including
the timestep, softening, seed, energy drift and SHA-256 digest. Snapshot playback
interpolates coordinates and never reruns the N-body dynamics on the headset.
Changing trail persistence changes only the view.

## Research references

- [Einstein Toolkit](https://einsteintoolkit.org/) and its [community infrastructure paper](https://arxiv.org/abs/1111.3344): numerical relativity and relativistic astrophysics.
- [GADGET-4](https://wwwmpa.mpa-garching.mpg.de/gadget4/) and [method paper](https://arxiv.org/abs/2010.03567): parallel cosmological N-body and SPH simulation.
- [Athena++](https://www.athena-astro.app/) and [method paper](https://arxiv.org/abs/2005.06651): adaptive mesh refinement and astrophysical MHD solvers. Nuclear fusion and stellar evolution require additional physics; they are not implied by basic MHD support.
- [Meta mixed reality in Browser](https://developers.meta.com/horizon/documentation/web/webxr-mixed-reality/): `immersive-ar` sessions for passthrough.
- [Three.js WebXRManager](https://threejs.org/docs/pages/WebXRManager.html): renderer session, reference-space and framebuffer controls.

## Rendering and interaction

Three.js r180 is vendored under the [MIT license](./vendor/THREE-LICENSE.txt).
Shaders and geometry are procedural. No external texture CDN, model download,
camera image access, account, inference API or paid service is needed.

White is used in the normal browser and studio VR. Passthrough has no background
mesh or fog, and the renderer clears with alpha zero. Only the exhibits, their
labels, and three floating interface panels are opaque. Real-room imagery is
provided by the Quest compositor. Placement is relative to the viewer, not a
persistent physical-world anchor. Physical Quest performance and tracking still
require on-device validation.