The Shape of FIREbox Galaxies and a Potential Tension with Low-mass Disks

We study the intrinsic and observable shapes of approximately 700 star-forming galaxies with stellar masses of $10^8 - 10^{11}$ M$\odot$ from the FIREbox simulation at $z=0$. We calculate intrinsic axis ratios using inertia tensors weighted by three morphology types: “All Stars,” “Young Stars,” and “Luminosity-weighted Stars.” Young Stars shows mass-dependent 3D configurations, with spheroidal, elongated, and disky shapes dominant at stellar masses of $10^{8.5}$ M$\odot$, $10^{9.5}$ M$\odot$, and $10^{10.5}$ M$\odot$, respectively. Using the radiative transfer code SKIRT, we construct mock images for each galaxy and show that projected short-to-long axis ratios, $q$, inferred from 2D Sérsic fits are most closely related to Luminosity-weighted Stars tensor shapes and least resemble the All Stars’ shapes. This suggests observed 2D shape distributions should not be compared to predictions based on 3D stellar mass shapes. Next, we construct a sample of mock images projected in random orientations and compare them to observed axis ratio distributions from the GAMA survey. At stellar masses below $10^{10}$ M$\odot$, we produce too few galaxies with observed $q<0.4$ and none with $q<0.2$, suggesting that FIREbox does not produce enough low-mass disk galaxies. At higher masses, $10^{10} - 10^{11}$ M$\odot$, we find that the predicted q distribution is sensitive to the dust-to-metal ratio; using our fiducial model, the distribution of $q$ values is formally consistent with observations, but there is tension with our ability to produce enough very thin systems with $q<0.2$. Future observational and theoretical programs aimed at understanding disk and thin-disk fractions will provide crucial tests of galaxy formation models.

💡 Research Summary

This paper presents a comprehensive study of the intrinsic three‑dimensional (3D) shapes and the observable two‑dimensional (2D) axis‑ratio distributions of star‑forming galaxies drawn from the FIREbox cosmological simulation at redshift z = 0. The authors select 698 central galaxies with stellar masses between 10⁸ M⊙ and 10¹¹ M⊙, ensuring each galaxy contains at least ~1,600 star particles. They exclude interacting systems to focus on isolated evolution.

3D Shape Measurement:
The 3D shape of each galaxy is quantified using a weighted inertia tensor. Three distinct weightings are applied: (1) “All Stars,” where every star particle contributes with its mass; (2) “Young Stars,” which includes only stars younger than 0.5 Gyr and is also mass‑weighted; and (3) “Luminosity‑weighted Stars,” where the r‑band luminosity of each particle replaces its mass as the weight. The tensor is iteratively diagonalized within an ellipsoidal volume until the axis ratios B/A and C/A converge to 10⁻⁴. This yields principal axes A > B > C for each weighting scheme.

The “Young Stars” weighting reveals a clear mass‑dependent transition in shape: low‑mass galaxies (≈10⁸·⁵ M⊙) are predominantly spheroidal (C/A ≈ B/A ≈ 1), intermediate‑mass systems (≈10⁹·⁵ M⊙) tend to be prolate (B/A ≈ C/A < A/B), and high‑mass galaxies (≈10¹⁰·⁵ M⊙) are mainly oblate disks (C/A ≪ B/A ≈ A/B). This suggests that recent star formation becomes increasingly confined to thin rotating structures as galaxy mass grows.

Mock Imaging and Radiative Transfer:
To connect intrinsic shapes with observable quantities, the authors generate synthetic images using the Monte‑Carlo radiative‑transfer code SKIRT. Images are produced in the SDSS u, g, and r bands, with a focus on the r‑band for statistical robustness. Gas particles are assigned a THEMIS dust mixture, and a dust‑to‑metal ratio (f_dust) is adopted: f_dust = 0.3 for galaxies below 10¹⁰ M⊙ and f_dust = 0.1 for more massive systems to compensate for the known over‑metallicity of FIREbox at high mass. Stellar particles are assigned Bruzual & Charlot (2003) spectral energy distributions with a Chabrier IMF, and their light is smoothed over a Gaussian kernel scaled to 1.4 × the gravitational softening length.

For each galaxy three random viewing angles are generated, plus three “principal‑axis orientations” (PAOs) aligned with the eigenvectors of the All‑Stars inertia tensor (face‑on, edge‑on, and an intermediate view). This yields a total of 2,094 mock images.

2D Shape Extraction:
The 2D projected axis ratio q = b/a is measured by fitting a Sérsic surface‑brightness profile to each mock r‑band image using AstroPhot. The Sérsic index n, effective radius R_e, and axis ratio q are free parameters. The authors find that the measured q correlates most strongly with the C/A ratio derived from the Luminosity‑weighted Stars tensor, and only weakly with the All‑Stars mass‑weighted shape. Consequently, observed axis ratios are best interpreted in terms of light‑weighted, not mass‑weighted, intrinsic geometry.

Comparison with Observations:
The authors construct a statistical sample of q values from the randomly oriented mock images and compare it to the axis‑ratio distribution measured in the GAMA survey for star‑forming galaxies in the same stellar‑mass bins.

• Low‑mass regime (M_ < 10¹⁰ M⊙):* The simulated q distribution is deficient in flattened systems. Fewer than a few percent of simulated galaxies have q < 0.4, and none reach q < 0.2, whereas GAMA shows a substantial tail toward very thin shapes. This indicates that FIREbox does not produce enough low‑mass thin disks or highly inclined disks.
• High‑mass regime (10¹⁰ M⊙ < M_ < 10¹¹ M⊙):* The agreement improves. With the fiducial f_dust = 0.1, the simulated q histogram matches the GAMA data within statistical uncertainties, but the simulated sample still lacks a pronounced population of ultra‑thin disks (q < 0.2). Raising f_dust to 0.3 increases the number of thin systems but simultaneously over‑produces low‑q galaxies, degrading the overall fit.

Interpretation and Implications:
The study demonstrates that (i) intrinsic shapes derived from stellar mass distributions cannot be directly compared to observed axis‑ratio statistics; light‑weighted shapes provide a more faithful bridge; (ii) the FIREbox model reproduces the overall trend of increasing diskiness with stellar mass but fails to generate a sufficient fraction of very thin disks at low masses, suggesting possible shortcomings in the implementation of stellar feedback, angular momentum transport, or resolution limits; (iii) at higher masses, the predicted q distribution is sensitive to the assumed dust‑to‑metal ratio, highlighting the importance of realistic dust modeling when forward‑modeling galaxy morphologies.

Conclusions and Future Directions:
The authors conclude that while FIREbox captures many qualitative aspects of galaxy morphology, a tension remains regarding the prevalence of thin, low‑mass disks. They recommend higher‑resolution zoom‑in simulations, exploration of alternative feedback prescriptions, and more sophisticated dust treatments. Observationally, deeper imaging and spectroscopic surveys that can robustly identify thin disks (e.g., with JWST, LSST, or integral‑field spectroscopy) will provide critical tests for the next generation of galaxy formation models.