Galaxy lacking dark matter in the standard cosmological paradigm

The standard cold dark matter plus cosmological constant model predicts that galaxies form 31 within dark matter halos 1 and that low-mass galaxies are more dark-matter dominated than 32 massive ones 2 . The unexpected discovery of two low-mass galaxies lacking dark matter 3,4 33 immediately provoked concerns about the standard cosmology and ignited explorations of al- 34 ternatives, including self-interacting dark matter and modiﬁed gravity 5–11 . Apprehension 5,11 35 grew after several simulations 5,12–21 using the conventional cosmology failed to adequately 36 form numerical analogs. Here we show that the standard paradigm naturally produces 37 galaxies lacking dark matter, with characteristics largely in agreement with observations. 38 Using a state-of-the-art cosmological simulation and a meticulous galaxy-identiﬁcation tech- 39 nique, we ﬁnd that extreme close-encounters with massive neighbors can transform regular 40 galaxies into dark-matter deﬁcient ones. We predict that ∼ 30% of massive central galaxies 41 (with at least 10 11 solar masses in stars) harbor at least one dark-matter deﬁcient satellite 42

(with 10 8 − 10 9 solar masses in stars). This distinctive class of galaxies opens an additional 43 layer to our understanding of the role of interactions in shaping galactic properties. Future 44 observations surveying galaxies in the aforementioned regime will provide a crucial test of 45 this scenario. 46 For over half a century, cold dark matter has been a key ingredient in our understanding of 47 galaxy formation 22, 23 . In the low-mass regime, the standard cold dark matter paradigm 1 predicts 48 that galaxies should be more dark-matter dominated 2 . This is supported by observations in the 49 nearby Universe 24 . For this reason, the detection of DF2 and DF4, two low-mass galaxies de-50 void of dark matter, was not anticipated 3, 4 . This intriguing discovery immediately sparked several 51 searches of numerical analogs in cosmological simulations 5, 12-21 , with limited success. Unsur-52 prisingly, the absence of dark-matter deficient galaxies in these simulations also raised doubts on 53 the validity of the standard paradigm itself 5, 11 . In this paper, we demonstrate that a novel cosmo-54 logical hydrodynamical simulation (which presupposes this paradigm) naturally creates numerical 55 versions of similar dark matter-deficient galaxies. This simulation utilizes the 'Feedback In Real-56 istic Environments' (FIRE-2) physics model 25 , which successfully reproduces an array of galaxy 57 properties 26 . This run is state-of-the-art in its ability to resolve the internal structure of individual 58 galaxies within a large cosmological environment. We point the reader to the Methods Section for 59 details. By directly comparing with observations, we confirm that some of our simulated galaxies 60 resemble DF2 and DF4 in arresting ways. This finding alleviates the aforementioned concerns -61 and restricts the need for alternative explanations invoking new (non-standard) physics, such as 62 self-interacting dark matter 11 and modified gravity 5-10 .  We identify seven galaxies lacking dark matter within a volume of 10 4 Mpc 3 . Figure 1 shows     with error bars), hereafter we quote quantities within r ⋆ 50 , the radius containing 50% of the stellar 87 mass (we note, however, that our seven galaxies are also dark-matter deficient out to the subhalo 88 radius). We select our sample of seven dark-matter deficient galaxies (pink hexagons) by requiring 89 M dm < M ⋆ (below the 1 : 1 line). Wolf is consistent with both DF2 and DF4 observations. We 90 do not include galaxies with M ⋆ < M dm < M baryon in this sample. These are likely simulated 91 analogs of the recently-discovered baryon-rich galaxies 32, 33 , which we plan to study separately.

93
Hereafter, we only focus on satellites with M ⋆ = 10  10% and one has only 1% of its mass in dark matter -rendering them almost dark-matter free 106 galaxies. See the Methods Section for work by other groups 5, 12-21 . One possible explanation is 107 that our simulation has the distinct advantage in being able to model the small-scale (∼ 20 pc) 108 interstellar medium of individual galaxies within a cosmological region (∼ 20 Mpc) large enough 109 to contain several massive groups of galaxies.  However, our simulation reveals that close encounters with a massive neighbor are responsible for 115 dark-matter deficiency in all of our seven galaxies. They all started out more massive, gas rich and 116 with fairly typical stellar-to-dark-matter mass ratios for galaxies of their mass -but subsequently   To quantify the role of interactions, we evaluate how close satellite galaxies come relative 126 to their hosts and their stellar mass ratios relative to these massive central companions. Figure  Another outstanding issue is the existence of anomalous globular cluster populations 40 in 152 DF2 and DF4. The simulation presented in this work has insufficient resolution to explore this.

153
Using the same physics model employed here, Refs. 41, 42 are capable of re-creating bound star 154 clusters using zoom-in simulations at higher resolution than ours. We suspect that, at some point 155 in their history, the early gas-rich progenitors of our dark-matter deficient galaxies met the requisite 156 conditions to produce bound star clusters. In the future, we plan to conduct higher-resolution zoom-157 in simulations of our sample and ask if there is a correlation between dark-matter deficiency and 158 the presence of a peculiar globular cluster population -and why the stellar component (including 159 globular clusters) in these satellites is more resilient to tidal stripping than the dark matter.

160
Our results demonstrate that dark-matter deficient galaxies can arise naturally within the 161 standard cold-dark-matter based cosmological paradigm. We note that did not expect this to occur 162 a priori (i.e., our simulation was not originally designed for this purpose). Although there certainly 163 is still room for new physics beyond the standard paradigm, discriminating between alternative 164 models will now rely on predicted differences between the expected properties of dark-matter 165 deficient galaxies, rather than their mere existence 5 .

166
In this work we predict that, to efficiently uncover more galaxies devoid of dark matter, ob-      cles. For gas, the force resolution is set to equal the adaptive smoothing length down to a minimum 269 of 1.5 pc, which occurs only in the densest regions of galaxies. We identify galaxies (and merger 270 history trees) with the AMIGA Halo Finder (AHF 1 ), which uses an iterative unbinding proce-271 dure to identify gravitationally-bound objects 2 . We used yt 3 to interface with the particle data. condition, the inter-particle separation in star-forming gas is ∼ 20 pc (or less) and we are able to 281 resolve giant-molecular-cloud complexes within the interstellar medium of individual galaxies.

282
Our simulated galaxy samples. In this work we focus on galaxies with at least 100 stellar parti- address this, first we discard objects with baryon-to-total mass ratios greater than 0.5. This may 289 naturally discard the very objects we seek (and keep false clumps below this threshold). To avoid 290 this, we create surface density maps of every halo with substructures and visually recover objects 291 that might have been discarded erroneously by our 50% cut. The images shown in Figure 1 be-292 long to this set. We focus on manually deleting clumps embedded inside massive galaxies and on 293 recovering galaxies that are either clearly disjoint from their neighbors -or in tidal tails, to avoid 294 rejecting tidally-formed candidates 4 . We highlight that every galaxy lacking dark matter studied in 295 this paper is recovered during this step. Lastly, because we are interested in the low-mass regime, 296 we constrain our sample to have stellar masses under 10 9 M ⊙ (within r ⋆ 80 , the radius containing 297 80% of the stellar mass). This produces our final parent sample, which contains 1,218 resolved 298 galaxies: 886 centrals and 332 satellites.

299
Our dark-matter deficient set consists of seven galaxies with M dm < M ⋆ that are not par-300 ticipating in a major merger (we discard an eighth galaxy meeting the former condition, but not 301 the latter). We adopt this extra condition because the two observed Dragonfly galaxies are not 302 currently merging with a companion of similar mass. We also exclude 19 low-mass galaxies (15 our simulation contains fifteen massive centrals that host 47 satellite galaxies in the target stellar 307 mass range (10 8−9 M ⊙ ). Of these satellites, ∼ 15% belong to our set of seven, which orbit ∼ 30% 308 of the available massive centrals.

309
The lack of dark-matter deficient satellite galaxies at stellar masses below 10 8 M ⊙ could arise 310 from the fact that, at lower masses, galaxies are born as centrals with very high dark matter mass 311 fractions, such that environmental effects cannot act to remove dark matter without completely de-312 stroying the galaxy. The absence of dark-matter deficient galaxies at lower stellar masses is likely 313 not driven by lack of resolution because our simulation produces a large population of galaxies with 314 at least 1,000 stellar particles that are less massive than Wolf (which has 3,280 stellar particles at 315 redshift zero and had a factor of ∼35 higher before becoming a satellite).

316
Galactic sizes. We calculate galactic sizes as follows. For each object, we record the distance to 317 the nearest major companion (with stellar mass of at least a tenth of that of our object of interest). If 318 no such companion exists, we record the virial radius (for centrals) or subhalo radius (for satellites). and record its σ-value. In the main text we report the simple average of these three numbers.

331
Host properties. Figure 5 shows the stellar mass of the central galaxies (hosting galaxies in our 332 low-mass sample) versus that of their satellites. We employ r ⋆ 80 for the former (to better capture 333 the spatial extent of the galaxy 38 ) and r ⋆ 50 for the latter (for easy comparisons with observations).

334
The diagonal lines denote constant stellar-mass ratios. Our seven dark-matter deficient galaxies,     Table 1 for a list of key properties. had enough time at distances far away from its host to become dynamically relaxed. 363 We recognize that it is observationally impossible to directly measure stellar-mass-based 3D 364 sizes. To emulate what is often done, we also measure the effective radii (R 2D e ) of the seven dark-   lost over 99% of their dark-matter content between infall and the present time. See Table 2 for 383 numerical values. We plan to explore the evolution of this set of seven galaxies in more detail in 384 the future.

385
Simulations by other groups. The identification of galaxies lacking dark matter in cosmological 386 simulations has been a challenge 12-14, 17-20 . Whilst many of these studies have shown that galaxies 387 can be produced with dark matter fractions lower than typical, the great majority of them do not dices) in agreement with observations ( Figure 3 and Table 1). Lastly, we predict the existence of 395 dark-matter deficient galaxies with extremely low values of M dm /M ⋆ : three below 10% and one 396 below 1%. Future observations and improved methodologies for measuring M dm /M ⋆ , will help 397 confirm or refute these predictions.

398
There are several possibilities that could explain why results similar to ours have not been 399 reported in past simulations. First, as mentioned in the main text, our simulations are relatively 400 unique in that they explicitly track dense, molecular gas in low-mass galaxies within a volume that 401 is large enough to contain several massive group-size halos. There are published simulations 19, 21 402 with similar formal spatial resolution to our own within comparable cosmological volumes, but 403 31 these utilize different star-formation and galaxy-formation physics. In particular, our star-formation 404 threshold is > 30 times higher than used in these comparably-sized simulations and this allows us 405 to track the internal baryonic structure of galaxies with higher precision. Another possibility is that 406 some simulations do in fact produce such extreme objects, but their automated galaxy-recovery 407 algorithms might pose an impediment to their query (recall that we employ a visual-recovery tech-408 nique, which may not be feasible for larger runs).

409
Alternatively, many authors employ idealized (non-cosmological) simulations 5-7, 11 . These one specific mechanism must be responsible for the creation of dark-matter deficient galaxies.

416
Our work therefore validates this scenario in a more natural, cosmologically-motivated fashion.

417
Namely, that close satellite-host encounters drive the formation of dark-matter deficient galaxies.