# A Modern, Volume-Complete Census of Galactic High-Mass Star Formation

*Plan document — successor to the RMS (Red MSX Source) survey. Drafted 2026-06-14.*

This is a prelude/companion to the UC HII accretion project. The accretion-timescale test
(testing Padoan & Gieles 2026) is ultimately limited by the **selection function and distance
quality of the parent catalog**. The RMS survey is the best existing parent catalog, but it is a
mid-IR **flux-limited** sample with **coarse kinematic distances**. This document (1) characterizes
exactly what RMS is and is not complete to, (2) identifies distances as the binding constraint, and
(3) lays out a path to a modern, *genuinely volume-complete* census from which carefully selected,
unbiased subsamples can be drawn.

---

## 1. What kind of survey is RMS?

**RMS is flux-limited and color-selected — it is neither mass-limited nor volume-limited.**

- **Parent instrument/band:** the MSX (Midcourse Space Experiment) Point Source Catalogue, selected
  primarily in **band E = 21.3 μm** (with bands A = 8.3 μm and D = 14.7 μm used for colors).
  [Lumsden et al. 2013, ApJS 208, 11]
- **Flux limit:** band E **95% completeness at 2.7 Jy** (band A point-source sensitivity ~0.1 Jy).
- **Color selection** (reproduces Wood–Churchwell UC HII colors, rejects PNe/AGB):
  `F_E > 2·F_A ; F_E > F_D ; F_D > F_A ; F_A > 5·F_K ; F_K > 2·F_J`.
- **Sky coverage:** |b| < 5°, longitude **10° < l < 350°**. The inner **±10° around the Galactic
  Centre is excluded** (source confusion + kinematic-distance ambiguity), so RMS is insensitive to
  SF within ~3 kpc of the GC.
- Final sample: **~1750 embedded young massive sources** (MYSOs + compact/UC HII, roughly balanced;
  Lumsden 2013 has the clean per-class split — Urquhart 2014 only quotes rounded totals).

**Why this matters:** a single-band 21 μm flux cut translates to a luminosity limit that scales as
d². So RMS *is* effectively luminosity-limited, but the luminosity floor it reaches moves outward
with distance. It is volume-complete only for the most luminous sources, and only as well as the
distances allow you to convert flux → L → M.

## 2. How complete is it?

- **Headline (Urquhart et al. 2014, MNRAS 437, 1791):** complete to embedded sources with
  **L_bol ≳ 2×10⁴ L⊙ out to ~18 kpc** — i.e. across essentially the whole Galaxy for that
  luminosity floor. ~100% complete above ~1×10⁴ L⊙ at smaller distances.
- **Partial floor (Mottram et al. 2011, ApJL 730, L33):** ~50% complete down to ~8×10³ L⊙ *locally*.
- **Physical threshold:** L_bol ≈ 2×10⁴ L⊙ corresponds to roughly **ZAMS B0 / late-O, M ≈ 10–15 M⊙**
  — i.e. the ionizing massive stars. *(This L→type mapping is the Mottram 2011 interpretation, not a
  verbatim Urquhart 2014 number.)*
- **UC HII:** Lumsden 2013 estimate **<5% of Galactic UC HII regions missing**.
- **Where it degrades:** MSX's ~18″ beam → confusion in the crowded inner plane (~1500 candidates
  rejected as extended/filamentary), which drives both the 350°–10° exclusion and the falling
  completeness toward the inner Galaxy.

**Bottom line:** RMS is a clean, near-complete *luminosity-limited* census of the ionizing massive
population (≳ B0) over most of the Galaxy. Its weaknesses are (a) the inner-Galaxy hole, (b)
mid-IR confusion, and (c) — the big one — distances.

## 3. Distances are the binding constraint

- ~1650 of ~1750 sources have distances (~7% lack them).
- **Method:** kinematic, on the Reid et al. 2009 rotation model (R₀ = 8.4 kpc, Θ₀ = 254 km/s),
  using RMS's own **¹³CO(1–0)/(2–1)** follow-up (~88% detection) + CS(2–1)/NH₃.
- **Kinematic distance ambiguity** affects **~80%** of the sample (inside the solar circle), resolved
  per-source via HI self-absorption / HI emission–absorption against continuum (SGPS/VGPS), IRDC
  association (GLIMPSE/MIPSGAL), and dedicated HI-absorption work (Urquhart et al. 2012).
- **Maser parallax** (BeSSeL/VERA) anchored only a few percent of sources at publication time.
- **Typical uncertainty:** kinematic **~±1 kpc** (from a ±7 km/s streaming-motion allowance) — small
  fractional error for distant sources, but tens of percent nearby, plus unresolved-KDA outliers.

**Synthesis (our inference, not a verbatim claim):** the *global completeness* of RMS is set by the
flux/luminosity limit, but the *per-source* mass and the ability to build a volume-complete subsample
are limited by the ±1 kpc kinematic distances and residual KDA. L ∝ d² and M(L) propagate that error
directly. **Distance is the lever a modern redo pulls hardest.**

---

## 4. The modern photometric replacement for MSX

MSX 21 μm at ~18″ is the weak link in *both* sensitivity and resolution. The combination below
replaces it with ~1–6″ data that reach 2–3 orders of magnitude fainter and that actually constrain
bolometric luminosity.

| Survey | Bands | Resolution | Depth | Role |
|---|---|---|---|---|
| **GLIMPSE / IRAC** (Churchwell+2009) | 3.6–8.0 μm | <2″ | 3σ ≈ 15.5/15.0/13.0/13.0 mag | point sources, photospheres, embedded SEDs |
| **MIPSGAL** (Carey+2009) | 24, 70 μm | 6″/18″ | 3σ ≈ 2 mJy / 75 mJy | the MSX-21 μm replacement at ~2500× depth |
| **Hi-GAL** (Molinari+2010,2016) | 70–500 μm | 5.6–35″ | compact-source cat. | cold-dust SED peak → **L_bol** and protostellar flag |

Key points:
- **MIPSGAL 24 μm replaces MSX band E directly** but ~2500× deeper and 3× sharper. Caveat: 24 μm
  **saturates** on the brightest sources (~1–2 Jy), exactly the UC HII regions — these must be
  recovered from Hi-GAL 70 μm / WISE / MSX itself.
- **Hi-GAL 70 + 160 μm bracket the cold-dust SED peak**, fixing **bolometric luminosity** of embedded
  sources and giving a clean **protostellar vs starless** discriminator (70 μm detection) — exactly
  what MSX could not do. Compact-source catalogs already exist: **Elia et al. 2017** (100,922 sources;
  24,584 protostellar) and **Elia et al. 2021** (150,223 entries; ~23% protostellar).
- Coverage: GLIMPSE/MIPSGAL/Hi-GAL all tile the inner plane at |b| ≲ 1° (Hi-GAL is full-longitude),
  so an MSX-equivalent footprint is fully covered with far better fidelity.

**Selection-function upgrade:** instead of a single 21 μm flux cut, define the sample on a
**Hi-GAL-derived L_bol** with MIPSGAL+IRAC SED constraints. This makes the primitive selection a
*luminosity* selection (modulo distance) rather than a monochromatic flux selection — a cleaner basis
for completeness modeling.

## 5. Distances: the path to volume-completeness

The goal is to replace ±1 kpc kinematic distances with **few-percent stellar-parallax / 3D-dust
distances** over as large a volume as possible, and to characterize exactly where that volume ends.

- **Gaia DR3** (Vallenari+2023): optical, so the embedded high-mass YSOs themselves (A_V ≳ 30–1000)
  are mostly invisible. Used **indirectly** — emerged O/B stars, cluster members, and foreground
  reddening stars. Reliable individual parallaxes (σ_ϖ/ϖ ≲ 0.1) to **~2–3 kpc** for typical OB stars.
  (HII-region anchor example: Shen et al. 2025 compiled 572 HII regions with mostly Gaia distances.)
- **3D dust "cloud distance" via the stellar reddening break** — assign a distance to a molecular
  cloud / SF region from where background stars redden:
  - Zucker et al. 2019/2020: ~1–5% statistical, ≲10% vs maser, to ~2.5 kpc.
  - Green et al. 2019 (Bayestar19, PS1+2MASS+Gaia): reliable to several kpc, δ > −30°.
  - **DECaPS2 deep (Zucker/Saydjari/Speagle 2025):** southern plane 239° < l < 6°, |b| < 10°,
    >700M stars, reaches A_V ≈ 12, **max reliable distance ≈10 kpc** — the deepest current
    inner-plane horizon.
- **Realistic combined volume-complete horizon:** **~2.5–4 kpc robust everywhere** (Gaia OB/cluster +
  3D-dust break), extendable to **~5 kpc off-plane / anticenter**, and opportunistically to **~10 kpc**
  along deep low-extinction southern DECaPS2 sightlines. The 10 kpc figure is a *max-reliable*, not a
  *uniform-completeness*, limit; beyond it, fall back to maser parallax + kinematic.

**Design consequence:** a modern survey can be **genuinely volume-complete (every source, all
luminosities down to the photometric floor) within ~3 kpc**, and *mass-complete for the massive
population* (≳ B0) to considerably farther because those sources stay above the flux floor. The clean
science sample is therefore a **distance-bounded cylinder (~3 kpc)** for unbiased volume statistics,
plus a **luminosity-bounded all-Galaxy sample** (RMS-like, but with Hi-GAL L_bol and improved
distances) for the rare luminous end.

## 6. Upcoming surveys that change the game

- **SPHEREx** (launched 2025-03-11; science ops since 2025-05-01; all-sky map Dec 2025): all-sky NIR
  **spectro-photometry 0.75–5.0 μm, ~102 bands, R ≈ 41–135**, 6.2″ pixels, 5σ ≈ AB 18–20.
  Adds **ice/SED diagnostics** (3.0 μm H₂O, 4.27 μm CO₂, 4.67 μm CO), low-res SEDs, and A_V/luminosity
  for every source — Galaxy-wide. Caveat: 6.2″ pixels → confusion in the crowded inner plane/CMZ.
- **Roman Galactic Plane Survey** (community GA survey, ~700 hr; approved, pre-execution; launch
  ~late 2026/2027): WFI **0.11″/pix, 0.5–2.3 μm**, 691 deg² (|b| < 2°, l = +50° to −79°) + CMZ/NSD.
  **0.11″ resolution defeats inner-plane confusion**, resolving embedded cluster members and giving
  deep NIR photometry + proper motions (complementary to Gaia in high-A_V crowded fields) → membership
  and distances for embedded high-mass clusters.

These two fill the exact gaps that limit MSX/Spitzer: SPHEREx adds Galaxy-wide spectral/extinction
information; Roman adds the angular resolution and NIR astrometry needed where confusion currently
defeats both selection and distance.

---

## 7. Proposed program

**Phase 0 — Characterize the RMS selection function quantitatively.** Re-derive the RMS completeness
in the (L_bol, distance) plane using its own catalog, and map where the inner-Galaxy hole and the
24 μm-saturation gap fall. Deliverable: a completeness function C(L_bol, d, l, b) for RMS.

**Phase 1 — Build the modern parent catalog.** Cross-match GLIMPSE + MIPSGAL + Hi-GAL (Elia
compact-source catalogs as the backbone), fit SEDs, derive L_bol and a protostellar/starless +
MYSO/HII classification. Replace the MSX 21 μm flux cut with a Hi-GAL L_bol selection. Recover
saturated bright sources from 70 μm/WISE/MSX.

**Phase 2 — Assign modern distances.** For every source, attempt in order: (a) maser parallax;
(b) Gaia-anchored cluster/OB-association membership; (c) 3D-dust cloud distance (Bayestar/DECaPS2/
Edenhofer); (d) kinematic + KDA resolution as fallback. Record method and uncertainty per source.

**Phase 3 — Define the completeness-controlled samples.**
  - *Volume-complete cylinder:* all sources within the robust 3D-dust/Gaia horizon (~3 kpc),
    complete down to the photometric luminosity floor → unbiased volume statistics, IMF of forming
    massive stars, multiplicity, evolutionary-stage lifetimes.
  - *Luminosity-complete all-Galaxy sample:* RMS-like (≳ B0) but with Hi-GAL L_bol and upgraded
    distances → rare luminous end, accretion-timescale tests (the P&G question), etc.

**Phase 4 — Fold in SPHEREx + Roman.** SPHEREx spectra → uniform A_V/ice/SED for the whole catalog;
Roman GPS → resolve confusion-limited inner-plane and CMZ sources and pin embedded-cluster distances.
Re-derive the completeness function with the confusion limit lifted.

**Phase 5 — Science with controlled subsamples.** With C(L_bol, d, l, b) known, draw carefully
selected subsamples (volume-limited, mass-limited, evolutionary-stage-matched) and use them to draw
Galaxy-wide conclusions about high-mass SF — including a definitive, selection-corrected version of
the UC HII accretion test this project began.

---

## Key references
- Lumsden et al. 2013, ApJS 208, 11 — RMS catalog & selection
- Mottram et al. 2011, ApJL 730, L33 — RMS luminosity completeness & L↔type
- Urquhart et al. 2014, MNRAS 437, 1791 — RMS distances & Galactic distribution
- Urquhart et al. 2012, MNRAS 420, 1656 — HI-absorption KDA resolution
- Carey et al. 2009, PASP 121, 76 — MIPSGAL; Churchwell et al. 2009, PASP 121, 213 — GLIMPSE
- Molinari et al. 2010, PASP 122, 314 / 2016, A&A 591, A149 — Hi-GAL; Elia et al. 2017, 2021 — catalogs
- Gaia DR3: Vallenari et al. 2023, A&A 674, A1
- Zucker et al. 2019/2020; Green et al. 2019 (Bayestar19); Edenhofer et al. 2024; Zucker/Saydjari/Speagle 2025 (DECaPS2) — 3D dust distances
- SPHEREx: Doré et al. 2014 (arXiv:1412.4872) + Mission Paper (arXiv:2511.02985)
- Roman GPS: Definition Committee Report (arXiv:2511.07494)

*Caveats flagged in sourcing:* the L_bol→spectral-type mapping is from Mottram 2011 (not Urquhart
2014); "distance = dominant systematic" is our synthesis, not a verbatim claim; the ~10 kpc DECaPS2
horizon is a max-reliable, not uniform-completeness, limit; exact RMS MYSO/HII integer split should be
pulled from the Lumsden 2013 database if load-bearing.