The 100-Parsec Map of the Solar Neighbourhood

About this map

The Gaia Catalogue of Nearby Stars (GCNS, Smart+ 2021) is the most complete vetted census of the solar neighbourhood within 100 pc — 331,312 sources rebuilt from Gaia EDR3 with a random-forest cleaning of spurious astrometry and Bayesian distance posteriors. This map is that catalogue rendered in 3D, exactly the way the catalogue's Galactic Cartesian columns are tabulated: Sun at the origin, X toward the Galactic centre, Y toward the direction of Galactic rotation, Z toward the north Galactic pole.

Five class layers scaffold the volume. Main sequence dominates everywhere. Late-type dwarfs (BP−RP > 3) number 88,142 here, with a median distance of 81.4 pc - only 2.9% lie inside 30 pc. The apparent-magnitude limit does truncate the intrinsically faintest of them, but the layer as a whole is not concentrated nearby. White dwarfs (WDprob ≥ 0.5) trace the local degenerate population. Giants (M_G < 4) are sparse and far. Halo kinematics (|U|, |V|, or |W| > 200 km/s in the Galactic frame) pulls out the local manifestations of streams and the Gaia-Enceladus members reported in §5.4 of Smart+ 2021.

Spectral types

The GCNS publishes no spectral classification. It does publish Gaia EDR3 G_BP − G_RP for every source, and this volume is shallow enough that the observed colour can be read as the intrinsic colour without an extinction correction. That approximation is good, but it is an approximation, and it is worth being exact about how far the literature actually carries it. Gaia Collaboration, Babusiaux et al. (2018) state that "the reddening is almost negligible within ∼60 pc of the Sun" — 60 pc, not 100. Most of this volume lies inside the Local Cavity, whose boundary Lallement et al. (2014) mark at a differential opacity of dE(B−V)/dr = 0.0002 mag pc⁻¹, with "the closest dense and cold gas … at ≃80 pc". We found no published statement setting a numerical reddening bound at 100 pc, and this page does not invent one. No extinction correction is applied here; the residual effect reddens colours slightly, so it pushes types marginally later, and it does so most at the outer edge of the sample.

Every point is binned against the Pecaut & Mamajek (2013) dwarf sequence in the form maintained by E. Mamajek as the Modern Mean Dwarf Stellar Color and Effective Temperature Sequence (table version 2022.04.16), Bp-Rp column. The anchor values are the table's; the bin edges are adopted here, not published as boundaries by anyone — each is the midpoint in BP−RP between the last subtype of one letter and the first of the next: B9.5V/A0V at −0.062, A9V/F0V at 0.352, F9V/G0V at 0.752, G9V/K0V at 0.967, K9V/M0V at 1.815. The M/L edge at 4.800 is adopted too, and less firmly — see limit (3). Hovering any star reports a subtype linearly interpolated between the same published anchors.

One check is worth stating, because it is the failure that would silently ruin everything downstream: the EEM file is fixed-width with 30 columns and no delimiter, so reading Bp-Rp off by one column yields a plausible-looking but entirely wrong sequence. Two things pin it. The decisive one is external: EEM's G2V row gives Bp-Rp = 0.823 and M_G = 4.635, against the Sun's published Gaia values of (G_BP − G_RP)_☉ = 0.82 and M_G,☉ = 4.67 from Casagrande & VandenBerg (2018) — an independent measurement of the same quantity, agreeing to 0.003 mag. The second is internal: the table satisfies its own identity M_G = M_V + (G−V), exactly (to 0.000) across the nine consecutive rows F0V through K0V, and to ≤ 0.015 mag on 21 of the 27 anchor rows used here. Six rows do exceed that — K8V (0.065), K9V (−0.130), M0V (−0.060), M6V (−0.310), M8V (+0.290), M9V (+0.200) — all at the cool end where the table prints G−V to two decimals. Those are rounding residuals, two orders of magnitude smaller than the ~1 mag error a column shift would produce. Worth being precise about, since the K9V/M0V pair is exactly what the K/M edge rests on.

What the distribution says. Of the 302,469 sources this types, 78.2% are M, another 262 are L or cooler, and just 8.4% are G or earlier. That is what a volume-limited sample looks like, and it is nothing like the naked-eye sky, which is dominated by intrinsically luminous stars at much greater distances. As a check that the binning is right, the median absolute magnitude of each bin — measured from this payload, from its own G and distance columns — comes out A 1.75, F 3.59, G 4.82, K 6.74, M 11.22. The published EEM values are M_G = 1.00 at A0V, 3.26 at F5V, 4.635 at G2V, 6.83 at K5V and 11.21 at M4V. The bins land on the published sequence without M_G having entered the binning at any point. Note this is a consistency check, not a proof: each bin spans several subtypes and is not uniformly populated, so agreement to a few hundredths at the M end is closer than the method deserves.

On the colours

Every stellar colour on this page is copied from a published table. None of it is computed here.

The source is Harre & Heller (2021), Digital color codes of stars, which convolves PHOENIX and TLUSTY model spectra with the CIE colour-matching functions and publishes an RGB and hex code per spectral type. Model spectra, not black bodies — and that distinction is the paper's own headline result: it reports "significant deviations between the color codes of stars computed from stellar spectra and from a black body radiator of the same effective temperature." Each bin's published swatch in the table above names the exact row it came from — spectral type, effective temperature and log g — so any value here can be checked against Table 5 of the paper, or against the machine-readable version on Zenodo, which we joined on both T_eff and log g to confirm every value.

Two gaps in that source are the reason two bins have no published colour. Table 5 tabulates main-sequence stars only, so there is nothing for white dwarfs; and its cool end stops at M9.5V / 2300 K, the floor of the PHOENIX grid, so there is nothing for L dwarfs. Those bins are shown neutral, or borrow the coolest published row and say so. A third caveat belongs to the source and we pass it on rather than quietly fixing it: the tables are labelled "Linear RGB", and the paper never names a transfer function, white point or colour space for its hex codes. We reproduce them as published.

Why there are two palettes. The published colours do not separate — and that is a finding, not a defect. Harre & Heller conclude that "there are no yellow, green, cyan, or purple stars", and that "Red dwarf stars (spectral types M0V - M9V) actually look orange to the human eye." Real stellar colours occupy two narrow families, blue-white and orange, so G, K and M land almost on top of one another and nine bins cannot be told apart on a crowded screen. Rather than doctor the published values and still call them physical, the map carries both and labels which is which: Spectral swatches: published uses Harre & Heller verbatim, and Spectral swatches: separated uses a display palette chosen only so the bins can be distinguished. The separated palette is ordered along the temperature sequence so the legend reads correctly, but no swatch in it is a claim about how any star looks. The button in the Color mode panel switches between them.

The BP−RP colour mode is continuous rather than binned, and its stops are the same two published tables joined on spectral type: the BP−RP of each stop from Pecaut & Mamajek / EEM, the colour from Harre & Heller at the nearest temperature that table prints for that type. Two of the seven stops match on T_eff exactly; the rest agree to within 70 K.

One rendering change came with this work. The point cloud now composites with normal blending in every mode except Density. It previously used additive blending everywhere, which sums the colour of every point landing on a pixel — and with 208,967 warm-white main-sequence points that drove the categorical views to flat white, so no class could be distinguished at all. Density still blends additively, because there the summing is the measurement.

What this map is — and isn't

It is the GCNS rendered in 3D, accurate to the spatial resolution your monitor can show. Every point is a real Gaia EDR3 source with a measured parallax, GCNS_prob ≥ 0.38, and dist_1 ≤ 100 pc — i.e. the published Smart+ 2021 selection.

It is not a kinematic study tool. The U/V/W velocities are tabulated and used here only to flag halo kinematic candidates; for proper-motion-aware visualizations or stream membership analysis use AstroPy + the GCNS table directly.

What's missing in v1.2. Luminosity classes (this map types colour, and colour alone cannot separate a K giant from a K dwarf — M_G is carried in the tooltip but is not yet used to split the sequence). Distance-PDF clouds (rendering each star as its full posterior rather than the median dist_50). Hyades-cluster overlay from J/A+A/649/A6/hyacomb. Wide-binary catalogue. Local-bubble ISM contours. WD cooling tracks. All on the v1.1 list.