R&D Capabilities — Generative Physics

Physics Model Reference

Every object Cosmos Genesis generates — from individual stars to whole galaxies — is produced by physics models drawn from published astrophysics. This page covers the generative engine: stellar populations and evolutionary stages, galactic structure, binary systems, planetary formation, and compact objects. Where a process is not built yet, the page says it is on the roadmap.

For applied capabilities — space weather, radiation environments, satellite tracking, and mission planning — see the R&D Capabilities section.

Go deeper

Public reference. This page explains what Cosmos Genesis does and the science it's based on. Implementation details, such as exact algorithms, tuning values and infrastructure, are kept in our internal documentation.

How aging works today. Generation gives every object its state at a single moment: stellar ages are set when a galaxy is generated, and nothing in the generated galaxy keeps evolving after that. A separate backend aging workflow, built and covered by integration tests but not something a visitor can run, can advance a generated universe through time, evolving its stars and tracking their deaths. Aging a galaxy on demand or on a schedule, and keeping the systems visitors are exploring in step with the wider simulation, are on the roadmap; see the aging capability page.

Galactic Scale

Universe Generation at Scale

The generation pipeline assembles stellar populations for spiral, elliptical, lenticular, and irregular galaxies, some with active galactic nuclei. Every star's mass is drawn from an empirical that runs from the lowest-mass hydrogen-burning red dwarfs up to the most massive O-type stars.

Each star is assigned an age drawn from its host galaxy's star formation history, a that falls with distance from the galactic centre, and an evolutionary stage set by how its age compares with its lifetime. Galactic orbits follow a flat — the flat outer-disk profile observed in real galaxies, which implicates dark matter halos.

Generation runs on , and a single request produces up to a million individually tracked star systems. Scaling to far larger galaxies is on the roadmap.

Outputs & Scale

Mass functionEmpirical
Metallicity modelFalls with distance from the galactic centre
Kinematic modelFlat rotation curve — consistent with dark matter halo models
Pipeline scaleIndividually tracked star systems, up to a million per request
Output formatStored as tables, queryable by position in space

Stellar Evolution

Stellar Life Cycles

Every star in a generated system is placed at the evolutionary stage its mass and age call for. A star's luminosity, radius, and surface temperature are set from its mass using main-sequence relations; they are not recalculated for its giant phase.

Post-main-sequence phases — the , , and — are each assigned as distinct evolutionary stages. The final evolutionary outcome is determined by the star's initial mass: a white dwarf (M < 8 M☉, with masses from a linear initial-final mass relation after Kalirai 20081), a neutron star (8–25 M☉), or a stellar-mass black hole (M ≥ 25 M☉). How luminosity and mass loss change as a star moves through these phases is part of the backend aging workflow, not of generation.

Because stellar ages follow the host galaxy's star formation history, a generated galaxy presents a mix of evolutionary stages rather than a frozen snapshot of zero-age main sequence objects. That state is fixed when the galaxy is generated and does not change until the galaxy is aged.

Outputs & Scale

Stellar agesFrom newly formed stars to the oldest in the galaxy
Stage assignmentMain sequence → SGB → RGB → AGB, by age relative to lifetime
Remnant classificationWhite dwarf (M < 8 M☉) · neutron star (8–25 M☉) · black hole (M ≥ 25 M☉), by initial mass
Output fieldsMass, luminosity, radius, temperature, evolutionary stage

Stellar Death

Supernovae & Transient Events

High-mass stars end their lives in — explosive events that release approximately of energy into the surrounding interstellar medium.2 Each generated supernova is given an affected radius that grows with its 's mass, and every system inside that radius receives a damage rating.

Nearby supernovae strip planetary atmospheres: the closer a system sits to the explosion, the larger the fraction of atmosphere its planets lose. The same damage join covers , which low- and intermediate-mass stars produce when they eject their outer envelopes, so a nearby nebula can also damage the atmospheres of planets in neighboring systems.

Massive young stars ionize the gas around them with light, forming ionized hydrogen regions sized by Strömgren sphere geometry.3

In a generated galaxy, these outcomes are assigned as part of the generation pass — a star already past its death is generated in its post-event state, and a blast wave is not carried forward in time. Only the backend aging workflow expands a blast wave over time. Authoring a death live, on a star that hasn't reached it yet, is on the roadmap; see the stellar evolution capability page.

Outputs & Scale

Supernova energy~ (1 foe) — standard core-collapse yield
Blast radiusScales with progenitor mass
Damage classificationLow · moderate · high · extreme, by fraction of atmosphere lost
Atmosphere lossComputed for planets in systems inside the blast radius
Planetary nebulaeWritten only for stars that shed their envelope within the last 15,000 years, so the table is expected to be empty at today's galaxy sizes
Ionized regionsStrömgren sphere geometry around massive young stars

Multi-Star Systems

Binary & Multi-Star Dynamics

Close to half of Sun-like stars have at least one stellar companion. Generation gives about 45% of systems a companion, and about 5% of those are triples, so a system has one, two, or three stars, with companion orbital periods that follow Kepler's third law.

Generated planets in binary systems are classified as circling one star (S-type) or both (P-type). In the system builder, planetary orbits around binary stars are checked against simplified stability limits after Holman & Wiegert (1999)4, which flag an orbit as stable or unstable.

How binary stars change over time — tidal circularization of their orbits, mass transfer between the stars, and orbital decay through gravitational wave emission — is on the roadmap.

Outputs & Scale

MultiplicitySingle, binary, and triple systems — ~45 % of systems have companions; ~5 % of those are triples
Companion orbitsPeriods from Kepler's third law
Planetary orbit classificationS-type / P-type at generation; stable / unstable check in the system builder
Binary evolutionOn the roadmap — tidal circularization, mass transfer, gravitational wave decay

Disk Physics

Planetary System Formation

Stellar birth is accompanied by a rotating disk of gas and dust — the — from which planets form. Generation creates a disk only around stars young enough to still have one. Each disk has a mass, an rate, and a rate, and its lifetime is a mass-dependent threshold, shorter around more massive stars, informed by observed disk survival times (Hartmann et al. 1998)5. A disk is classed as protoplanetary while it is young and as a debris disk once its gas is gone.

In real systems, disks typically disperse within about ten million years of a star's birth, so the planetary inventory is set early. Habitable zone (HZ) boundaries are computed from each star's luminosity when the system is generated. Evolving disks, and habitable zones that shift as a star ages and brightens, are on the roadmap.

Outputs & Scale

Disk lifetimeMass-dependent threshold, shorter around more massive stars — informed by observed T Tauri disk lifetimes (Hartmann et al. 1998)5
Disk classificationProtoplanetary while young, debris disk later
Output fieldsDisk mass, accretion rate, photoevaporation rate, HZ inner & outer boundary
Habitable zone (HZ)Computed from stellar luminosity at generation
Disk and HZ evolutionOn the roadmap

Planetary Physics

Planets & Moons

Generation gives each planet an orbit, a mass and radius, a state, and an set by its star's luminosity and its orbital distance. Gas giants receive moons (up to 20), and some rocky worlds receive one or two. Giant planets can also receive a ring system of several bands with gaps between them. A planet whose system lies inside a nearby supernova's blast radius carries the atmosphere loss that explosion caused.

In real systems, planets are not static. Hot Jupiters in close orbits lose orbital energy to tidal dissipation, drifting gradually inward over hundreds of millions of years.6 Low-mass planets near luminous stars lose atmospheric gas through -driven photoevaporation, with small worlds potentially shedding entire envelopes on gigayear timescales.7 Moons recede from their parent planets, as the Moon does from Earth at about 3.8 cm/yr. Modeling these slow changes as a planetary system ages, along with surface temperatures that track a star's changing luminosity, is on the roadmap.

Outputs & Scale

Orbits and bodiesPer-planet orbit, mass, and radius; moons for gas giants (up to 20) and some rocky worlds (one or two)
Surface temperatureEquilibrium temperature from stellar luminosity and orbital distance
Supernova damageAtmosphere loss for planets in systems inside a blast radius
Change over timeOn the roadmap — orbital decay, atmospheric escape, moon recession
Output fields, , atmospheric pressure, surface temp

Orbital Mechanics

N-Body Integration & Ephemerides

Gravitational environments with several interacting bodies — planet and moon hierarchies, multi-star systems — require direct numerical integration of the equations of motion for each body. The engine runs in the browser: in a star-system scene it computes the gravitational force between every pair of bodies and advances their orbits step by step.

The same Newtonian gravity is used at a very different scale in galaxy generation, which has an optional step. That step is applied across whole stellar populations rather than to the bodies of one system, so the two are different tools for different jobs rather than one shared mechanism.

Solar system positions are read from JPL's DE430 planetary through the toolkit from the group.8 In an automated test, a Sun, Earth, and Moon scene seeded from that ephemeris keeps the heliocentric positions and velocities of Earth and the Moon within 0.1% of the ephemeris values over about three weeks of simulated time.

Energy conservation is measured on an eccentric orbit, which is the hard case: the body moves fast at and slowly at . Over 100 orbits at eccentricity 0.5 with 1000 steps per orbit, the worst relative energy error is 1.053 × 10⁻⁴. The worst error in the last orbit is 0.99996 times the first orbit's, so it does not accumulate. This is one orbit shape and one step size, not a guarantee for every scene; a close encounter or a coarser step gives a larger error.

Orbit 1: the error swings and comes back

Signed relative energy error through the first orbit, starting at apoapsis. It peaks at periapsis, where the orbit is fastest, then returns to zero.

dE/E, in units of 10⁻⁴ · across the orbit, starting at apoapsis0.00.51.00%25%50%75%100%periapsis 1.053 × 10⁻⁴
100 orbits: the worst error does not grow

The largest relative energy error reached within each orbit. A flat line means no drift: each orbit is as accurate as the first.

dE/E, in units of 10⁻⁴ · across orbit number0.00.51.01255075100orbit 100: 1.053 × 10⁻⁴

Eccentricity 0.5, 1000 steps per orbit (a fixed step of 31554 s), semi-major axis 1 AU, a solar-mass primary and an Earth-mass secondary. Measured by running the integrator itself; the committed series is re-measured and compared in an automated test.

Show the values: signed energy error through orbit 1
signed energy error through orbit 1
Position in orbitdE/E
1%-5.706 × 10⁻¹⁰
2%-2.289 × 10⁻⁹
3%-5.174 × 10⁻⁹
4%-9.258 × 10⁻⁹
5%-1.459 × 10⁻⁸
6%-2.122 × 10⁻⁸
7%-2.924 × 10⁻⁸
8%-3.873 × 10⁻⁸
9%-4.980 × 10⁻⁸
10%-6.259 × 10⁻⁸
11%-7.725 × 10⁻⁸
12%-9.396 × 10⁻⁸
13%-1.129 × 10⁻⁷
14%-1.344 × 10⁻⁷
15%-1.586 × 10⁻⁷
16%-1.859 × 10⁻⁷
17%-2.167 × 10⁻⁷
18%-2.513 × 10⁻⁷
19%-2.903 × 10⁻⁷
20%-3.341 × 10⁻⁷
21%-3.833 × 10⁻⁷
22%-4.387 × 10⁻⁷
23%-5.010 × 10⁻⁷
24%-5.709 × 10⁻⁷
25%-6.494 × 10⁻⁷
26%-7.374 × 10⁻⁷
27%-8.358 × 10⁻⁷
28%-9.453 × 10⁻⁷
29%-1.067 × 10⁻⁶
30%-1.200 × 10⁻⁶
31%-1.343 × 10⁻⁶
32%-1.496 × 10⁻⁶
33%-1.651 × 10⁻⁶
34%-1.801 × 10⁻⁶
35%-1.929 × 10⁻⁶
36%-2.006 × 10⁻⁶
37%-1.986 × 10⁻⁶
38%-1.792 × 10⁻⁶
39%-1.302 × 10⁻⁶
40%-3.179 × 10⁻⁷
41%1.467 × 10⁻⁶
42%4.516 × 10⁻⁶
43%9.491 × 10⁻⁶
44%1.725 × 10⁻⁵
45%2.870 × 10⁻⁵
46%4.438 × 10⁻⁵
47%6.361 × 10⁻⁵
48%8.358 × 10⁻⁵
49%9.924 × 10⁻⁵
50%1.053 × 10⁻⁴
51%9.924 × 10⁻⁵
52%8.357 × 10⁻⁵
53%6.360 × 10⁻⁵
54%4.437 × 10⁻⁵
55%2.870 × 10⁻⁵
56%1.725 × 10⁻⁵
57%9.489 × 10⁻⁶
58%4.515 × 10⁻⁶
59%1.466 × 10⁻⁶
60%-3.183 × 10⁻⁷
61%-1.302 × 10⁻⁶
62%-1.792 × 10⁻⁶
63%-1.986 × 10⁻⁶
64%-2.006 × 10⁻⁶
65%-1.929 × 10⁻⁶
66%-1.801 × 10⁻⁶
67%-1.651 × 10⁻⁶
68%-1.496 × 10⁻⁶
69%-1.343 × 10⁻⁶
70%-1.199 × 10⁻⁶
71%-1.067 × 10⁻⁶
72%-9.453 × 10⁻⁷
73%-8.358 × 10⁻⁷
74%-7.374 × 10⁻⁷
75%-6.494 × 10⁻⁷
76%-5.709 × 10⁻⁷
77%-5.009 × 10⁻⁷
78%-4.387 × 10⁻⁷
79%-3.833 × 10⁻⁷
80%-3.340 × 10⁻⁷
81%-2.902 × 10⁻⁷
82%-2.513 × 10⁻⁷
83%-2.167 × 10⁻⁷
84%-1.859 × 10⁻⁷
85%-1.586 × 10⁻⁷
86%-1.344 × 10⁻⁷
87%-1.129 × 10⁻⁷
88%-9.396 × 10⁻⁸
89%-7.725 × 10⁻⁸
90%-6.259 × 10⁻⁸
91%-4.980 × 10⁻⁸
92%-3.873 × 10⁻⁸
93%-2.924 × 10⁻⁸
94%-2.122 × 10⁻⁸
95%-1.459 × 10⁻⁸
96%-9.257 × 10⁻⁹
97%-5.173 × 10⁻⁹
98%-2.288 × 10⁻⁹
99%-5.703 × 10⁻¹⁰
100%1.120 × 10⁻¹⁴
Show the values: worst energy error in each orbit
worst energy error in each orbit
OrbitdE/E
11.053 × 10⁻⁴
21.053 × 10⁻⁴
31.053 × 10⁻⁴
41.053 × 10⁻⁴
51.053 × 10⁻⁴
61.053 × 10⁻⁴
71.053 × 10⁻⁴
81.053 × 10⁻⁴
91.053 × 10⁻⁴
101.053 × 10⁻⁴
111.053 × 10⁻⁴
121.053 × 10⁻⁴
131.053 × 10⁻⁴
141.053 × 10⁻⁴
151.053 × 10⁻⁴
161.053 × 10⁻⁴
171.053 × 10⁻⁴
181.053 × 10⁻⁴
191.053 × 10⁻⁴
201.053 × 10⁻⁴
211.053 × 10⁻⁴
221.053 × 10⁻⁴
231.053 × 10⁻⁴
241.053 × 10⁻⁴
251.053 × 10⁻⁴
261.053 × 10⁻⁴
271.053 × 10⁻⁴
281.053 × 10⁻⁴
291.053 × 10⁻⁴
301.053 × 10⁻⁴
311.053 × 10⁻⁴
321.053 × 10⁻⁴
331.053 × 10⁻⁴
341.053 × 10⁻⁴
351.053 × 10⁻⁴
361.053 × 10⁻⁴
371.053 × 10⁻⁴
381.053 × 10⁻⁴
391.053 × 10⁻⁴
401.053 × 10⁻⁴
411.053 × 10⁻⁴
421.053 × 10⁻⁴
431.053 × 10⁻⁴
441.053 × 10⁻⁴
451.053 × 10⁻⁴
461.053 × 10⁻⁴
471.053 × 10⁻⁴
481.053 × 10⁻⁴
491.053 × 10⁻⁴
501.053 × 10⁻⁴
511.053 × 10⁻⁴
521.053 × 10⁻⁴
531.053 × 10⁻⁴
541.053 × 10⁻⁴
551.053 × 10⁻⁴
561.053 × 10⁻⁴
571.053 × 10⁻⁴
581.053 × 10⁻⁴
591.053 × 10⁻⁴
601.053 × 10⁻⁴
611.053 × 10⁻⁴
621.053 × 10⁻⁴
631.053 × 10⁻⁴
641.053 × 10⁻⁴
651.053 × 10⁻⁴
661.053 × 10⁻⁴
671.053 × 10⁻⁴
681.053 × 10⁻⁴
691.053 × 10⁻⁴
701.053 × 10⁻⁴
711.053 × 10⁻⁴
721.053 × 10⁻⁴
731.053 × 10⁻⁴
741.053 × 10⁻⁴
751.053 × 10⁻⁴
761.053 × 10⁻⁴
771.053 × 10⁻⁴
781.053 × 10⁻⁴
791.053 × 10⁻⁴
801.053 × 10⁻⁴
811.053 × 10⁻⁴
821.053 × 10⁻⁴
831.053 × 10⁻⁴
841.053 × 10⁻⁴
851.053 × 10⁻⁴
861.053 × 10⁻⁴
871.053 × 10⁻⁴
881.053 × 10⁻⁴
891.053 × 10⁻⁴
901.053 × 10⁻⁴
911.053 × 10⁻⁴
921.053 × 10⁻⁴
931.053 × 10⁻⁴
941.053 × 10⁻⁴
951.053 × 10⁻⁴
961.053 × 10⁻⁴
971.053 × 10⁻⁴
981.053 × 10⁻⁴
991.053 × 10⁻⁴
1001.053 × 10⁻⁴

Outputs & Scale

Energy conservationWorst relative error 1.053 × 10⁻⁴ over 100 orbits (eccentricity 0.5, 1000 steps per orbit), with no growth between orbits
Ground truth anchorNAIF SPICE kernels — JPL DE430 ephemeris
Positional verificationWithin 0.1% of ephemeris positions in an automated Sun–Earth–Moon test
Generation scaleOptional gravitational relaxation across stellar populations
System scaleFull gravitational integration per system
Use casesStar-system scenes with planets and moons

Exotic Objects

Compact Objects & Black Hole Physics

Black holes formed by stellar collapse are generated with a mass and a spin and, for some, an accretion disk with an accretion rate and X-ray output. In real black holes, the maximum accretion rate is limited by radiation pressure — the Eddington limit. Black-hole growth by accretion over time is on the roadmap.

Generation also places an active galactic nucleus at the center of 10–30% of galaxies, depending on galaxy type, some of them jet-producing blazars. Real accreting black holes with strong magnetic fields launch relativistic jets, powered in part by black-hole spin, a mechanism first described by Blandford and Znajek.9 Generated jets take their power from the accretion rate alone, not from spin; spin-powered jets are on the roadmap.

In real physics, lighter than roughly 1012 kg lose mass mainly through — thermal emission arising from quantum field effects near the 10 — with evaporation timescales that can be shorter than the age of the universe for the lightest objects. Modeling black hole evaporation is on the roadmap.

Outputs & Scale

Active galactic nuclei10–30 % of galaxies, depending on galaxy type
Black-hole growthOn the roadmap
Jet powerDerived from accretion rate; spin-powered jets on the roadmap
SpinAssigned at generation; spin evolution on the roadmap
Hawking evaporationOn the roadmap
Output fieldsMass, spin parameter, accretion rate, jet power (/s)

Full Reference List

The published sources behind the claims on this page are listed below, so reviewers can check the science each claim draws on.

  1. 1.Kalirai et al. (2008) — The initial-final mass relationship — Astrophysical Journal 676 ↩
  2. 2.Bethe (1990) — Supernova mechanisms — Reviews of Modern Physics 62:801 ↩
  3. 3.Strömgren (1939) — Ionization of interstellar hydrogen — Astrophysical Journal 89:526 ↩
  4. 4.Holman & Wiegert (1999) — Long-term stability of planets in binary systems — Astronomical Journal 117:621 ↩
  5. 5.Hartmann et al. (1998) — Accretion and the evolution of T Tauri disks — Astrophysical Journal 495 ↩
  6. 6.Jackson et al. (2008) — Tidal heating of terrestrial extrasolar planets — Astrophysical Journal 681:1631 ↩
  7. 7.Lecavelier des Etangs (2007) — Evaporation status of extrasolar planets — Astronomy & Astrophysics 461:1185 ↩
  8. 8.Acton (1996) — NAIF SPICE toolkit — Planetary and Space Science 44:65 ↩
  9. 9.Blandford & Znajek (1977) — Electromagnetic extraction of energy from Kerr black holes — MNRAS 179:433 ↩
  10. 10.Hawking (1974) — Black hole explosions — Nature 248:30 ↩