# A working model of massive star accretion, for game design

This is a compressed synthesis of ~25 years of massive-star-formation theory,
organized as a sequence of stages (→ game levels) and the mechanisms that
either sustain or threaten accretion at each stage (→ game tools/hazards).
It is deliberately simplified for the game; the "Notes" under each item flag
where the game takes dramatic license with the real physics/timescales.

## The core tension

A star more massive than ~10 M☉ reaches the main sequence *while still
accreting*. That means it starts shining — and pushing back — long before
it's done growing. Low-mass star formation doesn't have this problem (the
star finishes accreting while still cool and dim). The entire field of
massive star formation is, in large part, the study of how accretion wins
this fight against the star's own feedback for long enough to build up tens
of solar masses. (Zinnecker & Yorke 2007)

## Stage sequence

### Stage 0 — Hub-filament assembly (parsec scale, "normal" phase)
Parsec-scale filaments within a giant molecular cloud converge into a dense
hub under gravity, funneling gas inward non-isotropically ("clump-fed
accretion"). This sets up the reservoir a massive star will draw from — a
single prestellar core alone usually doesn't contain enough mass; the clump
has to keep resupplying it.
**Source:** Motte, Bontemps & Louvet 2018, ARA&A 56, 41 — [arXiv:1706.00118](https://arxiv.org/abs/1706.00118)

### Stage 1 — Core assembly (two competing pictures, both still debated)
- **Turbulent core / monolithic collapse**: a single massive, turbulent,
  pressure-supported core collapses as a unit, fast (~10⁵ yr), at an
  accretion rate high enough to later outrun radiation pressure.
  **Source:** McKee & Tan 2003, ApJ 585, 850 — [arXiv:astro-ph/0206037](https://arxiv.org/abs/astro-ph/0206037)
- **Competitive accretion**: many low-mass seeds form together in a
  cluster's shared potential well; the one sitting deepest in the well
  accretes disproportionately more of the ambient gas and "wins."
  **Source:** Bonnell & Bate 2006, MNRAS — [arXiv:astro-ph/0604615](https://arxiv.org/abs/astro-ph/0604615); synthesis in Krumholz & Bonnell 2009 — [arXiv:0712.0828](https://arxiv.org/abs/0712.0828)

*Game simplification: both are offered as alternative "feed" mechanisms
rather than mutually exclusive — in reality this is an active, unresolved
debate, not a solved either/or.*

### Stage 2 — Disk-mediated accretion ("normal" phase continues)
Below roughly 15–20 M☉, accretion looks much like scaled-up low-mass star
formation: gas lands through a circumstellar disk, which also launches a
bipolar outflow. This is the last stage where the star's own feedback isn't
yet a major obstacle.
**Source:** reviewed in Tan, Beltrán, Caselli, Fontani, Fuente, Krumholz,
McKee & Stolte 2014, *Protostars & Planets VI*, ch.7 (DOI:
10.2458/azu_uapress_9780816531240-ch007); observational review of disks
specifically in Beltrán & de Wit 2016, A&ARv 24, 6 — [arXiv:1509.08335](https://arxiv.org/abs/1509.08335)

### Stage 3 — The radiation-pressure barrier ("it gets complicated")
Once the protostar is luminous enough (~15–20 M☉), the classic 1D estimate
says radiation pressure on infalling dust should exceed gravity and *halt
spherical accretion* — long before reaching the masses we actually observe.
**Source:** Larson & Starrfield 1971; Kahn 1974; confirmed in 1D by e.g. Kuiper et al.'s review context — [arXiv:1211.7064](https://arxiv.org/abs/1211.7064)

**How accretion survives it (two proposed escape routes, both supported by
3D simulations):**
- **The "flashlight effect" / disk shielding**: a circumstellar disk makes
  the radiation field anisotropic — most of the star's light escapes
  through low-density polar cavities carved by the outflow, while accretion
  keeps flowing in through the shielded, dense equatorial disk.
  **Source:** Kuiper, Klahr, Beuther & Henning 2012 — [arXiv:1211.7064](https://arxiv.org/abs/1211.7064)
- **Radiative Rayleigh–Taylor instabilities**: even where radiation does
  push on infalling gas, the interface is unstable — dense fingers of gas
  survive and fall through a clumpy, radiation-dominated bubble wall.
  **Source:** Krumholz, Klein, McKee, Offner & Cunningham 2009, *Science*
  323, 754 — [arXiv:0901.3157](https://arxiv.org/abs/0901.3157) (3D
  radiation-hydrodynamic simulations); the general finding that radiation
  feedback is not very effective at halting accretion is echoed across
  later RHD studies.
- **Protostellar swelling suppresses the ionizing threat before it starts**:
  at the high accretion rates massive-star formation requires, the
  protostar itself swells to hundreds of solar radii and cools, which
  *delays* when it becomes hot enough to ionize its surroundings in the
  first place — pushing Stage 4's onset later than a non-accreting
  (Kelvin–Helmholtz) track would predict.
  **Source:** Hosokawa & Omukai 2009, ApJ 691, 823 — [arXiv:0806.4122](https://arxiv.org/abs/0806.4122)

### Stage 4 — Ionizing feedback / H II region onset (the real threat)
Around O-type temperatures (~20–30 M☉+), the star emits enough ionizing UV
to carve out an H II region. In 1D/spherical pictures this should blow the
remaining gas away and end accretion. In 3D simulations with multiple
forming stars, this is *not* what stops the most massive star from growing:
- **Fragmentation-induced starvation**: the accretion disk around the
  massive protostar is itself gravitationally unstable and fragments into
  lower-mass companions, which compete for — and steal — gas from the same
  shared reservoir. *This*, not the ionization front, is what caps the
  central star's mass in these simulations; the H II region/outflow cavity
  actually breaks out and vents rather than halting inflow outright.
  **Source:** Peters, Klessen, Mac Low & Banerjee 2010, ApJ 725, 134–145 — [arXiv:1005.3271](https://arxiv.org/abs/1005.3271)
  (not to be confused with the same group's companion paper on H II region
  morphology, Peters, Banerjee, Klessen, Mac Low, Galván-Madrid & Keto
  2010, ApJ 711, 1017).
- **A strong magnetic field suppresses that same fragmentation**: in 3D
  radiation-MHD simulations of massive cores, a magnetic field near or
  above the critical mass-to-flux ratio removes enough angular momentum
  (magnetic braking) to suppress fragmentation into companions, favoring a
  single star over a small cluster; a weaker field only reduces
  fragmentation, it doesn't prevent it outright.
  **Source:** Myers, McKee, Cunningham, Klein & Krumholz 2013, ApJ 766, 97 — [arXiv:1211.3467](https://arxiv.org/abs/1211.3467)
- **Ionized accretion flows persist through evolving H II regions**: even
  once the star is hot enough to ionize its surroundings, the ionized gas
  doesn't simply blow away isotropically — gravity (and later the disk)
  traps and channels it, so the H II region passes through a staged
  hypercompact → ultracompact sequence while accretion continues through a
  thinning equatorial disk, rather than shutting off at first ionization.
  **Source:** Keto 2007, ApJ 666, 976
- **Outflows, radiation pressure and magnetic fields together** set a soft
  cap on final mass and star-formation efficiency rather than a hard wall.
  **Source:** Rosen & Krumholz 2020, AJ 160, 78 — [arXiv:2006.04829](https://arxiv.org/abs/2006.04829)

*Game simplification, flagged explicitly in the tutorial: the game
compresses "the star loses the fight against fragmentation/ionization" into
a single dramatic beat — gas disperses, the star is cut off, cue a
supernova-style flash as the "you ran out of time" animation. In reality a
star that gets starved at say 20 M☉ does **not** immediately explode — it
would live a full, multi-million-year life first. The game's "game over =
supernova" is narrative shorthand for "your accretion stopped here," not a
claim about instantaneous stellar death.*

### Stage 5 (hard mode only) — Dynamical mergers
In the very densest young clusters, close dynamical encounters between
protostars/pre-main-sequence stars can lead to physical collisions and
mergers, proposed as a pathway to the very most massive stars (and to their
high observed binary/multiplicity fraction).
**Source:** Bonnell 2005, IAU Proceedings, "Massive star formation through
accretion and binary mergers"; discussed alongside the two main accretion
models in Zinnecker & Yorke 2007 — [arXiv:0707.1279](https://arxiv.org/abs/0707.1279)

## Mechanism → game element map

| Mechanism | Game role | Stage unlocked |
|---|---|---|
| Hub-filament funnel | passive background inflow | 0 |
| Turbulent core collapse | "feed" tool | 1 |
| Competitive accretion | alternative "feed" tool | 1 |
| Disk channel | "feed" tool | 2 |
| Flashlight vent (disk shielding) | required tool vs. radiation-pressure hazard | 3 |
| Radiative Rayleigh–Taylor leak | alternative required tool vs. radiation-pressure hazard | 3 |
| Protostellar swelling | *not yet implemented* — would delay Stage 4 onset | 3→4 |
| Clump resupply | required tool vs. starvation/ionization hazard | 4 |
| Magnetic shield (fragmentation suppression) | required tool vs. fragmentation hazard | 4 |
| Ionized-flow disk channel (Keto) | *not yet implemented* — alternative tool vs. ionization hazard | 4 |
| Dynamical merger | "Merge" button, hard mode only | 5 |

*The two "not yet implemented" rows are new literature additions (see below)
that strengthen the model's justification for Stage 4 but don't yet have a
corresponding game mechanic — flagged here rather than silently absent.*

## Primary references (chronological)

1. Larson & Starrfield 1971 — original radiation-pressure concern
2. Kahn 1974 — quantified the 1D radiation-pressure limit
3. McKee & Tan 2003, ApJ 585, 850 — [arXiv:astro-ph/0206037](https://arxiv.org/abs/astro-ph/0206037)
4. Bonnell 2005, IAU Symposium 227 — binary mergers
5. Bonnell & Bate 2006, MNRAS 370, 488–494 — [arXiv:astro-ph/0604615](https://arxiv.org/abs/astro-ph/0604615)
6. Beuther, Churchwell, McKee & Tan 2007, *Protostars & Planets V*, p.165–180 — [arXiv:astro-ph/0602012](https://arxiv.org/abs/astro-ph/0602012) (contemporaneous companion review to #7)
7. Zinnecker & Yorke 2007, ARA&A 45, 481–563 — [arXiv:0707.1279](https://arxiv.org/abs/0707.1279)
8. Keto 2007, ApJ 666, 976 — accretion survives through evolving ionized flows / hypercompact H II regions
9. Hosokawa & Omukai 2009, ApJ 691, 823 — [arXiv:0806.4122](https://arxiv.org/abs/0806.4122) — protostellar swelling at high accretion rates
10. Krumholz, Klein, McKee, Offner & Cunningham 2009, *Science* 323, 754 — [arXiv:0901.3157](https://arxiv.org/abs/0901.3157) — radiative Rayleigh–Taylor instabilities
11. Krumholz & Bonnell 2009 — [arXiv:0712.0828](https://arxiv.org/abs/0712.0828)
12. Peters, Klessen, Mac Low & Banerjee 2010, ApJ 725, 134–145 — [arXiv:1005.3271](https://arxiv.org/abs/1005.3271) — fragmentation-induced starvation
13. Kuiper, Klahr, Beuther & Henning 2012, A&A 537, A122 — [arXiv:1211.7064](https://arxiv.org/abs/1211.7064)
14. Myers, McKee, Cunningham, Klein & Krumholz 2013, ApJ 766, 97 — [arXiv:1211.3467](https://arxiv.org/abs/1211.3467) — magnetic suppression of fragmentation
15. Tan, Beltrán, Caselli, Fontani, Fuente, Krumholz, McKee & Stolte 2014, *Protostars & Planets VI*, ch.7
16. Beltrán & de Wit 2016, A&ARv 24, 6 — [arXiv:1509.08335](https://arxiv.org/abs/1509.08335) — disks around massive YSOs
17. Motte, Bontemps & Louvet 2018, ARA&A 56, 41–82 — [arXiv:1706.00118](https://arxiv.org/abs/1706.00118)
18. Rosen & Krumholz 2020, AJ 160, 78 — [arXiv:2006.04829](https://arxiv.org/abs/2006.04829)

**Caveat on completeness:** this is a representative, not exhaustive, tour
of the literature built from targeted searches (not a full systematic
review of 25 years of papers) — it covers the field's major turning points,
the two still-unresolved main pictures (core accretion vs. competitive
accretion), the two contemporaneous 2007 reviews that framed the modern
field (#6, #7), and the main proposed routes past each feedback barrier,
but a real literature review would surface many more observational papers
(ALMA-IMF, individual disk-detection papers, etc.) that the game doesn't
need individually.

**Corrections from a 2026-10-20 verification pass:** the previous version
mis-cited the fragmentation-induced-starvation paper as "ApJ 711, 1017"
(that page number actually belongs to a different paper by an overlapping
author list, Peters, Banerjee, Klessen, Mac Low, Galván-Madrid & Keto 2010)
and dropped two co-authors from Krumholz et al. 2009 along with its actual
venue (*Science*, not an implied ApJ-style ref). Both are fixed above, and
four papers that round out the literature since Zinnecker & Yorke 2007 were
added (#6, #8, #9, #15) — two of them (Keto 2007, Hosokawa & Omukai 2009)
strengthen stages 3–4 but don't yet have a corresponding game mechanic; see
the map note above.