WorksPhysics and Astronomy
Inferring the Total Mass of Galaxy Halos Using Photometry
Abstract
Galaxy size, as defined as the radius at which the stellar surface density reaches (), has been shown to correlate more tightly with stellar mass than the effective radius. Cosmological simulations further predict a tight relation between and dark matter halo mass (), but this prediction has not yet been tested observationally. We present the first observational test of the predicted – relation and assess whether galaxy size, defined through , can serve as a robust estimator of dark matter halo mass. We analyze galaxies from the SPARC database, combining high-quality rotation curves to infer dark matter halo masses with stellar surface density profiles derived from photometric data to measure . We then compare the observed relation with the predictions from state-of-the-art cosmological simulations. We find good agreement between observations and simulations: galaxies follow the same tight – relation predicted by numerical models across the explored mass range, with low intrinsic scatter. We provide the first observational validation of the – relation, demonstrating that is a simple, physically motivated, and robust estimator of dark matter halo mass. Because is determined entirely from photometric observations, it provides a tool for inferring halo masses from imaging data alone, with applications to studies of galaxy formation and evolution.
Introduction
It has long been established that observed galaxies follow a distinct stellar mass-size relation ([21, 4]), where size is traditionally defined by the half-mass radius. However, recent studies indicate that this relation becomes significantly tighter when galaxy size is instead defined by the radius at which the stellar surface density drops to 1 , denoted as ([23]). The surface density threshold is a proxy of the underlying physical process of star formation. It indicates, for spiral galaxies of the size of the Milky Way, the extent of past or current in-situ star formation on the disc.
The tightness of the relation, with scatter 0.06 dex over more than four orders of magnitude in stellar mass, has also been shown to hold true within hydrodynamical cosmological simulations ([1, 6]) independently of the numerical model for galaxy formation. What drives the relation is the gas surface density threshold that dictates the onset of star formation, as theoretically predicted by Schaye [19]. In addition, Sánchez Almeida [18] showed that, at the threshold of stellar surface density that define , Sérsic profiles Sérsic [22] for the same stellar mass and different indices tends to cross, giving similar values of .
Interestingly, analysis of the EAGLE simulations indicates that also closely correlates with the mass of halos hosting galaxies with scatter dex ([6]), significantly less than when using the effective radius ([11]). This hints at the tight relation between halo size, angular momentum and gas cooling for galaxy growth. Trujillo et al. [23] and Dalla Vecchia & Trujillo [6] also showed that, with this definition of size, the relation holds with the same scatter for early-type galaxies, although with different slope.
To date, this predicted relation has not been empirically tested in observed galaxies. In this work, we use rotation curves from the SPARC database ([12]) to derive halo masses, alongside the accompanying photometric maps to derive stellar surface density profiles and determine , in order to establish an observational galaxy size–halo mass relation and test it directly against theoretical predictions from simulations.
We describe the methodology in § 1.1 and the results in § 2. Discussion and conclusions are in § 3.
Methods
Throughout the work, we define the halo mass as , the mass contained in a spherical region of radius , where the density is 200 times the critical density of the universe. The EAGLE simulations ([20, 5]) assume a CDM universe with parameters: , , , , and ([16]).
The SPARC Sample and Stellar Masses
The Spitzer Photometry and Accurate Rotation Curves (SPARC) database ([12]) provides 175 H i rotation curves obtained from aperture synthesis radio observations spanning over three decades. The sample encompasses a variety of disk galaxies, including dwarf irregulars and high-surface-brightness spirals, spanning broad ranges in stellar mass () and rotational velocity ().
Rotation curves within SPARC are categorized into three quality flags (). represents high-quality, symmetric rotation curves. represents intermediate quality, typically constrained by lower spatial resolution or minor kinematic asymmetries. represents low-quality, subject to significant non-circular motions or asymmetric trends. The three quality categories are highlighted throughout our analysis.
SPARC also provides high-quality surface photometry at , which serves as an excellent tracer of the underlying stellar mass distribution, as the stellar mass-to-light ratio () remains nearly constant at this wavelength regardless of variations in stellar mass and morphology. The observed surface-brightness profiles are converted into stellar surface density distributions (), corrected for inclination (see [12]).
Derivation of the Size Metric
We define the size of each galaxy by the outer radius , where the azimuthally averaged stellar surface density reaches ([23]). We derive for SPARC galaxies by finding the radius at which the observationally mapped total stellar surface density profile drops to this threshold value.
Dark Matter Halo Mass Fitting
To determine the total dark matter host halo mass () for each galaxy, we implement the rotation curve decomposition procedure outlined by Katz et al. [10]. We set a minimum of 5 data points in the rotation curve for a reliable fit. The total observed rotational velocity, , is modelled as a quadrature sum of the contributing components to the gravitational potential:
where is the contribution of the cold gas (calculated from H i radial profiles and corrected for helium), and are the stellar disk and bulge contributions derived from the surface brightness profiles, is the disk stellar mass-to-light ratio at adopted as a free parameter in the fit (see below), is the bulge mass-to-light ratio following the ratio adopted by Lelli et al. [12], and represents the contribution from the dark matter halo.
We parameterize the dark matter component using two density models:
Navarro, Frenk & White (NFW) profile: Derived from cosmological -body simulations (Navarro et al. [15]), the NFW profile provides a two-parameter description of the spherically averaged dark matter density:
where is the scale radius at which the logarithmic slope of the density profile equals , and the characteristic density is determined by the halo concentration (see Navarro et al. [15], for details).
Di Cintio et al. (DC14) profile: Formulated from high-resolution hydrodynamical simulations (Di Cintio et al. [7]), the DC14 profile accounts for the modification of dark matter halos driven by supernova feedback and gas outflows. It follows a generalised density distribution:
where the inner slope , outer slope , and transition sharpness are not fixed constants but vary as continuous functions of the stellar-to-halo mass ratio , calibrated directly from hydrodynamical simulations (Di Cintio et al. [7]). This allows the profile to smoothly interpolate between a cuspy NFW-like shape at high and low and a cored profile at intermediate mass ratios, where supernova feedback is most effective at redistributing dark matter.
The free parameters of these mass models are the halo virial velocity (from which is derived), the halo concentration , and the stellar mass-to-light ratio . These are fitted to the observed rotation curves via a Markov Chain Monte Carlo (MCMC) approach following Katz et al. [10], using the emcee ensemble sampler (Foreman-Mackey et al. [9]).
We impose a Gaussian prior on using the concentration–mass relation of Dutton & Macciò [8] with a scatter of 0.11 dex. The stellar mass-to-light ratio is given a flat prior over the physically motivated range , consistent with stellar population synthesis models for a Chabrier IMF (Chabrier [3]) at . Although is nearly constant at this wavelength, allowing it to vary within this narrow range self-consistently accounts for galaxy-to-galaxy variations in stellar population age and metallicity. We verify a posteriori that the resulting posterior distributions cluster around and , well within the prior range and broadly consistent with the fixed values of 0.5 and 0.7 adopted by Lelli et al. [12] for disk and bulge respectively. The factor of approximately two spanned by the prior is generally considered acceptable for stellar mass estimates at ([12, 13]).
To isolate the dominant posterior mode and avoid degenerate secondary solutions, we apply the HDBSCAN clustering algorithm (Campello et al. [2]) to the MCMC chains following Katz et al. [10], retaining only the highest-probability cluster for parameter estimation. Final parameter estimates and uncertainties are taken as the median and 2.5th–97.5th percentile range of the filtered posterior samples.
Figure 1. Stellar surface density profile for ESO116G012, with the dotted horizontal and yellow vertical lines intersecting at , the radius at which
We then compare the resulting observational measurements of galaxy size (), stellar mass, and halo mass to predictions from simulations.
Results
Size
In Figure 1 we show an example derivation for the size, , of the galaxy ESO116-G012. In 36 of 175 SPARC galaxies, observations do not go deep enough to reach the threshold surface density for . These are discarded, leaving 139 galaxies in our sample.
Halo Mass
Two galaxies (D512-2 and UGC00634) are excluded from the halo mass fitting as their rotation curves contain only 4 data points, leaving 137 galaxies with halo mass determinations. Figure 2 shows an example of the fits to rotation curves used to derive halo masses, following Katz et al. [10], as set out in § 1.1. Fits using an NFW profile are on the left, whilst fits using the mass dependent DC14 profile are on the right. The spiral galaxy NGC5055 is shown in the upper panel and the dwarf galaxy NGC3109 is shown in the lower panel.
Figure 2. Red lines are best-fit NFW (left) and DC14 (right) dark matter halo profiles to rotation curves for a spiral (top) and dwarf galaxy (bottom). The black, orange and blue lines are the contribution to the rotation curve of dark matter, stars and gas mass, respectively. These fits are produced using our implementation of the method of Katz et al. [10], and we show the same galaxies as in that work to demonstrate the consistency of our results with theirs.
Figure 3. The relation between our derived and stellar mass for all galaxies, using results from NFW fits (top) and DC14 fits (bottom). Colors indicate the quality of rotation curve data from Q1, high quality, Q2 medium and Q3 low quality in blue, orange and red respectively. The Moster et al. [14] relation is shown as a black line, whilst the Posti linear fit for Spiral galaxies is shown as a green dashed line.
In Figure 3 we show the relation between the derived and stellar mass for all SPARC galaxies, using results from NFW fits (top) and DC14 fits (bottom). Colors indicate the quality of rotation curves from SPARC: Q1, high quality, Q2 medium and Q3 low quality in blue, orange and red respectively. Following Katz et al. [10], which provides full details, error bars represent the 95% credible intervals (2.5{sup}th to 97.5{sup}th percentiles) of the halo mass posterior distribution, determined by isolating the dominant statistical mode of the MCMC chain using HDBSCAN clustering. The black line shows the relation from Moster et al. [14], whilst the dashed green line shows the power law relation from Posti et al. [17] for spiral galaxies.
– relation
Figure 4 shows the relation between and for the 139 SPARC galaxies that have stellar density profiles that reach as low as . Also plotted in the background are simulation data in green for NIHAO (triangles), FIRE (circle), AURIGA (square), HESTIA (diamond), all taken from Arjona-Gálvez et al. 2025, as well as data from the EAGLE simulations (purple triangles, Dalla Vecchia & Trujillo 2025) and observational results (gray circles, Trujillo et al. 2020).
– relation
Figure 5 shows the relation between and for the 137 SPARC galaxies with R1 measurements and halo mass determinations. Also shown is the EAGLE simulation data (gray circles) with binned data as purple triangles, both from Dalla Vecchia & Trujillo [6]. In order to provide a useful prediction that can be used in observations, we provide the best-fit parameters for a double power-law as in Dalla Vecchia & Trujillo [6], but inverting the relation shown in that work. We used the following formulation:
Figure 4. The relation between and for SPARC galaxies, overplotted on data from Arjona-Gálvez et al. 2025, showing simulation data in green for NIHAO (triangles), FIRE (circle), AURIGA (square) and HESTIA (diamond), data from the EAGLE simulations (purple triangles, Dalla Vecchia & Trujillo 2025) and observational results (gray circles, Trujillo et al. 2020).
Figure 5. The relation between and for SPARC galaxies, overplotted on the relation for EAGLE simulation data (grey circles) and fit to those simulations (purple triangles) both from Dalla Vecchia & Trujillo [6].
where is either or . The best-fit parameters are shown in Table 1.
| Parameter | ||
| kpc | ||
| kpc | kpc | |
Table 1. Parameters of the best fits to the EAGLE data for the fit function in Eq. (4).
– relation
Figure 6 shows the relation between and for SPARC galaxies, plotted over EAGLE simulation data (grey circles) with a non-parametric fit shown as a purple line, taken from Dalla Vecchia & Trujillo [6].
Figure 6. The relation between and for SPARC galaxies, plotted over EAGLE simulation data, plotted as grey circles with a non-parametric fit shown as a purple line, taken from Dalla Vecchia & Trujillo [6].
Conclusion
In this work, we have empirically tested the predicted relation between galaxy size (, the outer radius where the stellar surface density drops to ) and dark matter halo mass using observed disk galaxies from the SPARC database. With their high-quality photometry, we determined for 137 galaxies that have surface-density profiles reaching down to the required threshold and rotation curves with 5 or more data points. By decomposing the observed rotation curves of these galaxies via an MCMC framework, we modelled their total host halo masses using both NFW and DC14 dark matter density profiles.
The observational results show that:
The derived versus stellar mass distribution for SPARC galaxies aligns well with standard semi-empirical and empirical expectations.
The observed SPARC sample follows the tight – scaling relation predicted by cosmological hydrodynamical simulations and previous observational work.
The observed galaxies exhibit an – relation that is in good agreement with the theoretical predictions and parameterizations derived from the EAGLE simulations.
An alternative, simpler kinematic indicator, the peak rotational velocity , correlates remarkably well with , matching simulated trends.
Overall, our analysis confirms that the observationally accessible galaxy size metric provides a robust and reliable estimator of both host dark matter halo mass and maximum rotational velocity. This empirical validation establishes as a tool for linking the visible properties of disk galaxies to their underlying dark matter halos.
Acknowledgments
CB and CDV were supported by the Spanish Ministry of Science, Innovation and Universities (MICIU / Agencia Estatal de Investigación, AEI / 10.13039/501100011033) through grant PID2024-156100NB-C22. ADC and EAG acknowledge financial support from MICIU/AEI/10.13039/501100011033 under the Consolidación Investigadora program (grant CNS2023-144669) and grant PID2024-160009NA-I00. This work is also part of grant CEX2025-001609-S, awarded to the Instituto de Astrofísica de Canarias under the Severo Ochoa Centre of Excellence program and funded by MICIU/AEI/10.13039/501100011033.
The authors used Claude (Anthropic) as an AI assistant to help develop and debug analysis code, to assist in drafting and editing the Methods section, and for proofreading and copyediting of the manuscript text. All scientific content, results, and conclusions were verified and are the responsibility of the authors.
This paper was built using the Open Journal of Astrophysics LaTeX template. The OJA is a journal which provides fast and easy peer review for new papers in the astro-ph section of the arXiv, making the reviewing process simpler for authors and referees alike. Learn more at http://astro.theoj.org.
References
- [1]Arjona-Gálvez, E., Dalla Vecchia, C., Trujillo, I., et al. 2025, Astronomy & Astrophysics, in preparation / Companion paper to Dalla Vecchia & Trujillo 2025
- [2]Campello, R. J. G. B., Moulavi, D., & Sander, J. 2013, in Advances in Knowledge Discovery and Data Mining, 160–172
- [3]Chabrier, G. 2003, PASP, 115, 763
- [4]Courteau, S., Dutton, A. A., van den Bosch, F. C., et al. 2007, The Astrophysical Journal, 671, 203
- [5]Crain, R. A., Schaye, J., Bower, R. G., et al. 2015, MNRAS, 450, 1937
- [6]Dalla Vecchia, C. & Trujillo, I. 2025, Astronomy & Astrophysics, 704, A32, arXiv:2510.19926arxiv.org/abs/2510.19926
- [7]Di Cintio, A., Brook, C. B., Macciò, A. V., et al. 2014, MNRAS, 437, 415
- [8]Dutton, A. A. & Macciò, A. V. 2014, Monthly Notices of the Royal Astronomical Society, 441, 3359
- [9]Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306
- [10]Katz, H., Lelli, F., McGaugh, S. S., et al. 2017, Monthly Notices of the Royal Astronomical Society, 466, 1648
- [11]Kravtsov, A. V. 2013, ApJ, 764, L31
- [12]Lelli, F., McGaugh, S. S., & Schombert, J. M. 2016, The Astronomical Journal, 152, 157
- [13]Meidt, S. E. et al. 2014, ApJ, 788, 144
- [14]Moster, B. P., Naab, T., & White, S. D. M. 2013, MNRAS, 428, 3121
- [16]Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2014, A&A, 571, A16
- [17]Posti, L., Fraternali, F., & Marasco, A. 2019, Astronomy & Astrophysics, 626, A56
- [18]Sánchez Almeida, J. 2020, MNRAS, 495, 78
- [19]Schaye, J. 2004, ApJ, 609, 667
- [20]Schaye, J., Crain, R. A., Bower, R. G., et al. 2015, MNRAS, 446, 521
- [21]Shen, S., Mo, H. J., White, S. D. M., et al. 2003, Monthly Notices of the Royal Astronomical Society, 343, 978
- [22]Sérsic, J. L. 1968, Atlas de Galaxias Australes
- [23]Trujillo, I., Chamba, N., & Knepman, J. H. 2020, Monthly Notices of the Royal Astronomical Society, 493, 87