The abundance of thin dwarf galaxies: a challenge for cosmological simulations

We study the prevalence of thin galaxies as a function of stellar mass in the range $10^7 < M_{\star} / \rm{M_\odot} < 10^{11}$ using data from the GAMA, DESI, ALFALFA, and Nearby Galaxy catalogs. We use the distribution of projected axis ratios, $q$, to infer the abundance of intrinsically flat galaxies needed to reproduce the observed abundance of highly elongated systems in projection. We find that as many as $40%$ of galaxies in the mass range $10^9<M_{\star}/\rm{M_\odot}<10^{10}$ are intrinsically flatter than $1$:$5$ (i.e., $c/a<0.2$), a fraction that rises to $\sim 80%$ for $c/a<0.3$. Although the incidence of thin galaxies decreases towards lower and higher $M_{\star}$, they are still quite common in dwarfs: $\sim 30%$ and $\sim 65%$ of $\sim 10^8 ~ \rm{M_\odot}$ galaxies are inferred to be intrinsically flatter than $c/a=0.2$ and $0.3$, respectively. A comparison of these results with several state-of-the-art cosmological hydrodynamical simulations (TNG50, FIREbox, Romulus25) reveals a distinctive lack of thin simulated dwarfs. In particular, there are no $M_{\star} < 10^9 ~ \rm{M_{\odot}}$ simulated galaxies flatter than $c/a=0.2$, in clear contrast with observational samples. This discrepancy likely reflects limitations in resolution and in the treatment of baryonic physics, suggesting that our understanding of the mechanisms regulating the formation of disk galaxies less massive than the Milky Way is still quite incomplete. Our results present a clear challenge to current numerical models of dwarf galaxy formation, which future models should attempt to meet.

💡 Research Summary

**
The authors investigate how common thin, highly flattened galaxies are across a wide range of stellar masses ($10^{7}$–$10^{11},M_{\odot}$) by analysing projected axis‑ratio ($q$) measurements from four large observational data sets: GAMA, DESI, ALFALFA, and the Nearby Galaxy catalog. Using the assumption of random viewing angles, they de‑project the $q$ distributions to infer the intrinsic three‑dimensional axis‑ratio distribution ($c/a$) of galaxies, treating most systems as either oblate (disk‑like) or prolate. Low $q$ values (e.g., $q<0.2$) correspond to edge‑on disks and thus provide a robust lower limit on the fraction of very flat galaxies.

The main observational findings are:

• In the stellar‑mass interval $10^{9}$–$10^{10},M_{\odot}$, about 40 % of galaxies have $c/a<0.2$ and roughly 80 % have $c/a<0.3$, indicating a very high prevalence of thin disks.
• Even at $M_{\star}\sim10^{8},M_{\odot}$, roughly 30 % of galaxies are flatter than $c/a=0.2$ and 65 % flatter than $c/a=0.3$.
• The fraction of thin systems declines toward both lower and higher masses, but remains non‑negligible down to $M_{\star}\sim10^{7},M_{\odot}$ (∼10 % with $c/a<0.2$) and up to $M_{\star}>10^{10.5},M_{\odot}$ where thick, spheroidal morphologies dominate.

To test theoretical models, the authors compare these results with three state‑of‑the‑art cosmological hydrodynamical simulations: TNG50‑1 (Illustris‑TNG), FIREbox (FIRE project), and Romulus25. All three simulations reproduce the overall galaxy stellar‑mass function and the morphology of massive galaxies, but they fail dramatically for low‑mass systems. In the simulated samples, no galaxy with $M_{\star}<10^{9},M_{\odot}$ attains $c/a<0.2$, and the smallest intrinsic flattenings are typically $c/a\gtrsim0.3$. The simulated thin‑disk fraction therefore falls far short of the observed values.

The authors argue that this discrepancy likely stems from two intertwined issues. First, the current mass and spatial resolution (stellar particle masses $\sim10^{5},M_{\odot}$, gravitational softening $\sim100$ pc) is insufficient to resolve the vertical structure of dwarf disks, which can be only a few tens of parsecs thick. Second, the implementation of stellar feedback (energy, momentum, radiation pressure, and the timing of gas outflows) may be too violent for low‑mass halos, inflating the stellar component and preventing the formation of thin, rotation‑supported disks. Additional factors such as the assumed initial angular momentum distribution, sub‑halo interactions, and the treatment of gas cooling and turbulence may also contribute.

The paper carefully addresses potential observational biases. The HI‑selected ALFALFA sample is less affected by surface‑brightness selection but is biased against edge‑on systems; nevertheless it still shows a high fraction of thin galaxies, reinforcing the robustness of the result. The optical samples (GAMA, DESI) are subject to dust attenuation and surface‑brightness limits, but their large numbers and the use of consistent Sérsic‑fit axis ratios mitigate these effects. The Nearby Galaxy catalog provides a volume‑limited view of the very low‑mass regime, confirming that the trend extends to dwarf spheroidals and irregulars.

In conclusion, the study provides strong empirical evidence that thin stellar disks are common even in dwarf galaxies, contrary to the predictions of current ΛCDM‑based hydrodynamical simulations. The authors call for next‑generation simulations with higher mass resolution (stellar particles $<10^{4},M_{\odot}$), finer gravitational softening (≲30 pc), and more sophisticated, multi‑phase feedback models that can capture the delicate balance between outflows and disk settling in low‑mass halos. Only with such improvements will theoretical models be able to reproduce the observed abundance of thin dwarf galaxies and resolve the highlighted tension.