Half-light Mass Estimator

Updated 7 July 2026

The half-light mass estimator is an analytic relation that infers the enclosed mass near a galaxy’s half-light radius using observable photometric and kinematic data while minimizing sensitivity to orbital anisotropy.
It applies primarily to dispersion-supported systems and is adapted for flattened galaxies by using circularized radii and alternative calibrations to relate projected measurements to intrinsic mass distributions.
The estimator’s robustness stems from Jeans equation manipulation and supplementary virial-based calibrations, providing a low-complexity yet effective tool for mapping galaxy dynamics.

The half-light mass estimator is a class of analytic relations that infer the mass enclosed near a galaxy’s half-light radius from directly observable photometric and kinematic quantities. In its canonical dynamical form, developed for dispersion-supported systems, the enclosed mass at the 3D deprojected half-light radius $r_{1/2}$ is estimated from the luminosity-weighted square of the line-of-sight velocity dispersion and the projected half-light radius, with the central result that the estimate is only weakly sensitive to the unknown velocity-anisotropy profile. Subsequent work extended the estimator to flattened systems, proposed alternative minimum-variance calibrations, and clarified that half-light radii need not coincide with half-mass radii when strong radial $M/L$ gradients are present (0908.2995, Sanders et al., 2016, Suess et al., 2019).

1. Canonical formulation

The modern half-light mass estimator is usually identified with the result of Wolf et al., who showed that the dynamical mass enclosed within the 3D deprojected half-light radius can be written as

$M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$

where $\sigma_{\rm los}^2$ is the luminosity-weighted square of the line-of-sight velocity dispersion and $R_{\rm eff}$ is the 2D projected half-light radius (0908.2995). For common stellar profiles, $r_{1/2} \simeq (4/3)R_{\rm eff}$ , which makes the projected form directly usable from standard photometric measurements (Wolf, 2010).

The key conceptual advance is that the estimator targets a special radius at which the classic mass–anisotropy degeneracy is minimized. By manipulating the spherical Jeans equation, Wolf et al. showed that the mass inferred from line-of-sight kinematics becomes least sensitive to the unknown orbital anisotropy $\beta(r)$ near $r_{1/2}$ , provided the projected velocity-dispersion profile is fairly flat near the half-light radius (0908.2995). In this sense, the “half-light mass” is not merely a convenient aperture mass; it is the mass at the radius where the data constrain the enclosed mass most robustly.

A related point emphasized in the original derivation is that this formula is not the virial theorem. The virial theorem cannot be used to determine accurate masses unless the radial profile of the total mass is known a priori, whereas the half-light estimator is tied specifically to the behavior of the Jeans solutions near the anisotropy-insensitive radius (0908.2995).

2. Derivation, observables, and domain of validity

The derivation begins from the spherical Jeans equation,

$M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$

with $\gamma_\star=-d\ln n_\star/d\ln r$ and $M/L$ 0. The observational difficulty is that only projected quantities are measured, so different combinations of $M/L$ 1 and $M/L$ 2 can reproduce the same $M/L$ 3. Wolf et al. showed that the difference between masses inferred under different anisotropy assumptions vanishes near the radius where $M/L$ 4, and that for observed systems this radius lies very close to $M/L$ 5 (0908.2995).

The practical observables are therefore straightforward: a projected half-light radius from photometry and a luminosity-weighted line-of-sight velocity dispersion from stellar kinematics. In the numerical form quoted in later exposition,

$M/L$ 6

which makes the scaling explicit in astronomical units (Wolf, 2010).

The estimator is accurate under a specific set of assumptions. The system should be approximately spherical, dispersion-supported rather than rotation-supported, and in dynamical equilibrium. The projected velocity-dispersion profile should also be nearly flat near the half-light radius. When these conditions hold, the remaining uncertainty at $M/L$ 7 is dominated more by observational errors than by anisotropy (0908.2995).

The half-light estimator belongs to a broader class of “simple mass estimators” that avoid computationally expensive detailed modeling. In slowly rotating elliptical galaxies, the global Wolf et al. estimator and the local Churazov et al. estimator both recover an almost unbiased circular-speed estimate with a modest RMS scatter $M/L$ 8, although tests on analytical models and simulated galaxies indicate that the local estimator has a smaller RMS scatter than the global one (Lyskova et al., 2015).

The distinction between global and local estimators is methodological. The global half-light estimator uses the total luminosity-weighted velocity dispersion and evaluates the mass at a radius near the 3D half-light radius. The local estimator instead uses the projected velocity dispersion at a radius $M/L$ 9 where the surface brightness declines as $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 0. In practice, the local formula

$M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 1

and the global formula

$M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 2

are both intended to exploit radii where anisotropy sensitivity is minimized (Lyskova et al., 2015).

This comparison clarifies the epistemic status of the half-light estimator. It is not unique in seeking an anisotropy-insensitive radius, but it is distinguished by its use of global observables and by its especially direct connection to the deprojected half-light radius.

4. Extensions to flattened systems and minimum-variance virial calibrations

A major generalization concerns flattened dispersion-supported galaxies. Sanders and Evans showed how to extend half-light estimators to ellipsoidal systems under the assumption that the dark matter density and the stellar density are stratified on the same self-similar ellipsoids. In that case, the mass within a half-light ellipsoid can be written as

$M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 3

with $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 4 accounting for the relation between the total and observed velocity dispersions and $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 5 relating the projected half-light radius to the intrinsic ellipsoidal scale, with $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 6 for intrinsic axis ratios $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 7 and $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 8 (Sanders et al., 2016).

Their central practical result is a remarkably simple workaround: the spherical estimators remain accurate for the mass within the half-light ellipsoid if $M_{1/2} \equiv M(r_{1/2}) = \frac{3\,\sigma_{\rm los}^2 r_{1/2}}{G} \simeq \frac{4\,\sigma_{\rm los}^2 R_{\rm eff}}{G},$ 9 is replaced by the circularized radius

$\sigma_{\rm los}^2$ 0

so that

$\sigma_{\rm los}^2$ 1

In the near-spherical limit, $\sigma_{\rm los}^2$ 2, which provides the mathematical justification for this prescription (Sanders et al., 2016).

A separate refinement was introduced by Errani et al. using the projected spherical virial theorem. Their estimator was calibrated to minimize uncertainty from ignorance of the inner halo slope and the degree to which the stellar component is embedded within the halo, yielding

$\sigma_{\rm los}^2$ 3

and controlled simulations showed that it provides unbiased enclosed masses with an accuracy of $\sigma_{\rm los}^2$ 4 per cent (Errani et al., 2018).

Estimator	Expression	Intended radius
Walker et al.	$\sigma_{\rm los}^2$ 5	$\sigma_{\rm los}^2$ 6
Wolf et al.	$\sigma_{\rm los}^2$ 7	$\sigma_{\rm los}^2$ 8
Errani et al.	$\sigma_{\rm los}^2$ 9	$R_{\rm eff}$ 0

These variants differ in target radius and calibration philosophy, but they preserve the same basic scaling $R_{\rm eff}$ 1. The main differences are whether anisotropy robustness is achieved through Jeans-based cancellation near $R_{\rm eff}$ 2, through ellipsoidal rescaling, or through projected-virial minimization over halo-profile uncertainties.

5. Half-light radius versus half-mass radius

A recurring misconception is that the half-light radius can always be interpreted as a proxy for the half-mass radius. Resolved photometric analyses show that this is often false because radial $R_{\rm eff}$ 3 gradients make the stellar mass distribution more centrally concentrated than the light distribution.

In a mass-selected sample of 177 galaxies at $R_{\rm eff}$ 4, Szomoru et al. derived stellar mass surface-density profiles from radial rest-frame $R_{\rm eff}$ 5 color profiles and found that stellar half-mass radii are on average $R_{\rm eff}$ 6 smaller than rest-frame $R_{\rm eff}$ 7-band half-light radii. This average size difference is the same at all redshifts, does not correlate with stellar mass, specific star formation rate, effective surface density, Sérsic index, or galaxy size, and for approximately $R_{\rm eff}$ 8 of massive galaxies the half-mass radius is more than a factor two smaller than the half-light radius (Szomoru et al., 2012).

Suess et al. extended this result to $R_{\rm eff}$ 9 galaxies at $r_{1/2} \simeq (4/3)R_{\rm eff}$ 0, comparing three methods for calculating half-mass radii from multi-band HST imaging. All three methods agree, with scatter $r_{1/2} \simeq (4/3)R_{\rm eff}$ 1 dex. They found that most galaxies have negative color gradients, so $r_{1/2} \simeq (4/3)R_{\rm eff}$ 2, and that the normalizations of the star-forming and quiescent $r_{1/2} \simeq (4/3)R_{\rm eff}$ 3- $r_{1/2} \simeq (4/3)R_{\rm eff}$ 4 relations are $r_{1/2} \simeq (4/3)R_{\rm eff}$ 5– $r_{1/2} \simeq (4/3)R_{\rm eff}$ 6 smaller than the corresponding $r_{1/2} \simeq (4/3)R_{\rm eff}$ 7- $r_{1/2} \simeq (4/3)R_{\rm eff}$ 8 relations, while the slopes are $r_{1/2} \simeq (4/3)R_{\rm eff}$ 9– $\beta(r)$ 0 dex shallower (Suess et al., 2019).

A later CANDELS/3D-HST analysis of $\beta(r)$ 1 galaxies at $\beta(r)$ 2 using $\beta(r)$ 3 likewise found significant scatter in individual $\beta(r)$ 4 ratios, correlated with variation in the color gradients, and measured a median $\beta(r)$ 5 ratio evolving from $\beta(r)$ 6 at $\beta(r)$ 7 to $\beta(r)$ 8 at $\beta(r)$ 9 (Miller et al., 2022). In nearby galaxies from CALIFA, the $r_{1/2}$ 0 ratio has a negative radial gradient that is steeper within the central $r_{1/2}$ 1 half-light-radius, and optical $r_{1/2}$ 2-color relations at the half-light radius have scatter of $r_{1/2}$ 3 dex (García-Benito et al., 2018).

The implication for half-light-based dynamical scaling is explicit in the high-redshift structural studies: estimators using the half-light radius in relations of the form $r_{1/2}$ 4 can overestimate galaxy size and, thus, mass and density for systems with strong bulge/disk contrasts or dust gradients (Szomoru et al., 2012). In globular-cluster $r_{1/2}$ 5-body models, no significant difference was found for the half-mass radii of metal-rich and metal-poor systems, yet the fitted half-light radii differed by $r_{1/2}$ 6 on average because metallicity-dependent stellar evolution and mass segregation altered the light profile rather than the mass profile (Sippel et al., 2012). Half-light estimators therefore remain dynamical tools tied to the observable light distribution, not direct measurements of stellar mass structure.

6. Applications, limitations, and recurrent points of confusion

The canonical estimator has been used to derive broad astrophysical inferences. Wolf et al. showed that all of the Milky Way dwarf spheroidal galaxies are consistent with having formed within a halo of mass approximately $r_{1/2}$ 7 in $r_{1/2}$ 8CDM cosmology. Using the same framework across dispersion-supported systems, they found that the $r_{1/2}$ 9-band $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 0 versus $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 1 relation follows a U-shape, with a broad minimum near $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 2, a steep rise to $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 3 for ultra-faint dSphs, and a more shallow rise to $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 4 for galaxy-cluster spheroids (0908.2995).

Alternative calibrations reinforce these conclusions while shifting the emphasis from anisotropy to halo-profile systematics. Errani et al. found that application of their projected-virial estimator to Milky Way dSph data reveals a tight correlation between enclosed mass and luminosity, that tidal stripping has little effect on this relation, and that the high mass densities of ultrafaint galaxies are not compatible with large dark matter cores (Errani et al., 2018).

Three misunderstandings recur in the literature. First, the half-light mass estimator is not a generic virial estimator; its robustness depends on the anisotropy-insensitive radius identified through Jeans analysis (0908.2995). Second, it is not universally valid for flattened, rotating, or non-equilibrium systems; flattened galaxies require either explicit ellipsoidal treatment or the circularized-radius substitution $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 5 under the common-ellipsoid assumption (Sanders et al., 2016). Third, the observable half-light radius is not identical to the half-mass radius of the stellar distribution when color gradients, dust, metallicity effects, or remnant segregation generate strong $M(r) = \frac{r\,\sigma_r^2}{G}\left[\gamma_\star+\gamma_\sigma-2\beta\right],$ 6 gradients (Suess et al., 2019).

Within those limits, the half-light mass estimator remains one of the central low-complexity tools for converting resolved photometric and kinematic data into enclosed dynamical masses in dispersion-supported systems. Its continuing development has not altered its basic logic: mass is best constrained at a characteristic radius tied to the light distribution, but the interpretation of that radius depends on geometry, equilibrium, and the distinction between where the light is and where the mass is.