1. Introduction
The standard ΛCDM cosmological model [
1,
2,
3] requires about 81 per cent of the matter content of the Universe to comprise pressureless cold dark matter (CDM) particles, which are not described by the standard model of physics, and about 68 per cent of the total mass–energy density of the Universe to be composed of dark energy, represented by the cosmological constant
[
4,
5]. The related
WDM model is based on dark matter (DM) being made of warm dark matter particles that have a smaller mass than CDM particles but lead to very similar DM halos within which the observed galaxies reside (e.g., [
6,
7,
8,
9]). Detecting DM particles has become a major worldwide effort, which has so far yielded null results (e.g., [
10,
11,
12,
13,
14,
15] and references therein). Finding evidence for DM particles under the assumption that they have a finite interaction cross-section with standard-model particles may lead to null detection if the interaction cross-section is too small to be measurable. However, arguments based on the observed strong correlations between standard particles and dark matter suggest a significant cross-section [
16]. This indicates an impasse in the experimental verification of the existence of DM particles.
An alternative and robust method for establishing the existence of DM particles is to develop techniques that rely solely on their gravitational interaction with ordinary matter. A suitable approach is given by Chandrasekhar dynamical friction (e.g., [
5]): when a satellite galaxy with its own DM halo enters the DM halo of a host galaxy, its orbit decays as a result of dynamical friction, and the satellite merges with the host. The decay of the orbit does not depend on the mass of the DM particle but only on the mass density of DM particles, which is fixed by the cosmological parameters. This is the primary reason for the existence of the merger tree in standard cosmology (e.g., [
17,
18]) and for interacting galaxies to be thought of as merging galaxies. Effectively, a DM halo, being 10–20 times more extended and 50–100 times as massive as the observable part of a galaxy, works like a spider’s web. In contrast, if there were to be no DM halos around galaxies, then galaxy–galaxy encounters would be significantly less dissipative, galaxies would encounter each other multiple times, and mergers would be rare. In the eventuality that DM halos were not to exist, however, non-relativistic gravitational theory would need to be non-Newtonian, and given the correlations that galaxies are observed to obey, this theory would need to be effectively Milgromian [
19,
20,
21,
22]. Galaxy–galaxy encounters and mutual orbits in Milgromian dynamics and without DM halos have been studied (for example, the interacting Antennae galaxy pair [
23], and the Milky Way–Andromeda binary explaining the mutually correlated planes of satellites around both hosts [
24,
25,
26]). The law of gravitation fundamentally defines the formation and evolution of galaxies. For example, by mergers being much rarer in Milgromian gravitation (i.e., without DM), galaxies evolve largely in isolation [
27,
28,
29], and the formation of cosmological structures proceeds differently than in the DM-based models [
5,
30,
31]. It is thus of paramount importance to test for the existence of DM halos around galaxies.
The test based on Chandrasekhar dynamical friction for the existence of DM halos was introduced by Angus, Diaferio and Kroupa [
32] by addressing the question of whether the present-day Galactocentric distances and motion vectors of observed satellite galaxies of the Milky Way (MW) conform to their putative infall many Gyr ago. The solutions for those satellite galaxies for which proper motions were available imply tension with the existence of DM halos since no infall solutions were found. In contrast, without the DM component, the satellite galaxies would be orbiting about the MW, having been most likely born as tidal dwarf galaxies during the MW–Andromeda encounter about 10 Gyr ago [
25,
26,
33]. A further test for the presence of the DM component applying dynamical friction was achieved by Roshan et al. [
34], who found the observed bars of disk galaxies to be too long and rotating too fast in comparison with the theoretically expected bars in the presence of DM halos that absorb the bar’s angular momentum. The reported discrepancy amounts to significantly higher than the 5-sigma threshold, such that the observations are incompatible with the existence of DM halos. Another independent application of the dynamical friction test is available on the basis of the observed distribution of matter within the M81 group of galaxies. The extended and connected tidal material implies multiple past close encounters of the group members. But the dynamical evolution of the M81 group of galaxies is difficult to understand theoretically if the galaxies are contained in the DM halos that are expected in standard cosmology ([
35] and references therein). The problem is that, in the presence of DM halos, the group merges too rapidly to allow the tidal material to be dispersed as observed.
The interesting aspect of these results is that the three independent analyses of the orbital dynamics of the MW satellite galaxies, galactic bars and the M81 group agree in the conclusion that the data are difficult if not impossible to understand in the presence of DM halos. Because the implications of the non-existence of DM bears major implications for theoretical cosmology and galaxy formation and evolution, further tests are important to ascertain the above conclusions, or indeed to question them.
Given the high-accuracy and high-precision position and velocity data available today through the astrometric Gaia mission, the problem of how the existence of the massive and extended DM halos in which galaxies reside can be tested for is revisited here using the process of Chandrasekhar dynamical friction on the triple-galaxy system comprising the MW and the Large (LMC) and Small (SMC) Magellanic Clouds, including the Magellanic Stream. This system is comparable to the M81 system, as it too has a large gaseous structure stemming from the Magellanic Clouds, the Magellanic Stream, which constrains the past orbits of the components of the system. In the following, we assume that the standard cosmological model is correct and that the MW, the LMC and the SMC are each contained in DM halos that are consistent with those obtained from the theory of structure formation in the CMD model of cosmology. That is, we associate the baryonic mass of each galaxy with a DM halo profile as predicted by the theory in order to test the theory.
Given that the LMC and SMC are about 50 and 60 kpc distant from the centre of the MW and about 20 kpc distant from each other and that the radii of the DM halos are such that all three galaxies are immersed in the dark matter halo of the MW and that the SMC is immersed in the dark matter halo of the LMC and vice versa, it is apparent that Chandrasekhar dynamical friction is likely to play a very significant role in establishing the orbits of the three galaxies relative to each other. Recent work [
36,
37,
38,
39,
40,
41] suggests the Magellanic Clouds to be on their first pericentre passage such that the bulk properties of their putative DM halos will not be significantly stripped. Detailed dynamical modelling of the Magellanic Clouds problem has, until now, not come to the conclusion that there is a problem.
For example, Besla et al. [
37] performed simulations of the LMC and SMC in the first-infall scenario by assuming that dynamical friction from the MW DM halo can be neglected. This is not correct and significantly helps the LMC/SMC pair to exist longer before merging. Indeed, Lucchini et al. [
42] studied the formation of the Magellanic Stream using hydrodynamical simulations in the first-infall scenario for the LMC and SMC. Their results, obtained in fully live DM halos of all components such that dynamical friction is correctly computed self-consistently, show that the LMC/SMC binary orbit shrinks significantly more rapidly (their Figure 1) than calculated by Besla et al. [
37]. But Lucchini et al. adopted DM halo masses of the LMC (
), of the SMC (
) and of the MW (
) that are significantly less massive than predicted by the
CDM model (see
Table 1 below). Their calculations also do not consider whether the infalling LMC/SMC binary would have survived for sufficiently long before infall, because it is implausible for the present-day LMC/SMC binary to have formed during or shortly before infall, and whether the infall velocity vector is consistent with the Hubble flow 3.46 Gyr ago at the start of their simulation. Vasiliev [
41] studied a scenario in which the LMC is on its second passage past the MW but ignored the SMC. The calculated orbits of the LMC decay due to dynamical friction consistently to the results presented here. Thus, this past work either neglects important contributions to dynamical friction or members in the problem or uses light-weight DM halos that are not consistent with the
CDM cosmological model. Essentially, past work has demonstrated that the MW/LMC/SMC triple system can broadly be understood in the context of orbital dynamics by effectively suppressing Chandrasekhar dynamical friction, but it does not demonstrate that the calculated orbits are consistent with the standard DM-based cosmological theory.
In order to assess whether the standard DM-based models can account for the existence of the observed MW/LMC/SMC triple system, the present work studies their orbital history by strictly constraining the DM halo of each of the three involved galaxies to be consistent with the
CDM-allowed DM halo masses. The dark matter component is thus assumed to be pressureless dust. The calculations furthermore impose the condition that the LMC and SMC had an encounter within 20 kpc of each other between 1 and 4 Gyr ago in order to allow the Magellanic Stream to form, and they take into account cosmological expansion to constrain the relative positions of the galaxies 5 and more Gyr ago. To check for the consistency of the solutions, two independent statistical methods were applied to search for all allowed initial conditions within the error bars for the transverse velocity components of the LMC and SMC. The two tests yielded indistinguishable results. The calculations based on rigid Navarro–Frenk–White (NFW) DM halo profiles are conservative because equivalent simulations with live DM halos lead to faster merging [
35]. These calculations show there to be no orbital solutions and that the observed system comprising the MW/LMC/SMC plus Magellanic Stream cannot be understood in the presence of DM halos in the context of the
CDM model. The problem that gravitational theory leads to the observed system is left for a future contribution.
3. Observational Data
Table 1.
Stellar masses of the galaxies (model (o)), varied by −
(model (m)) and +
(model (p)), and the derived DM halo masses according to
Section 2.1.
Table 1.
Stellar masses of the galaxies (model (o)), varied by −
(model (m)) and +
(model (p)), and the derived DM halo masses according to
Section 2.1.
Object | Model | Stellar Mass | DM Halo Mass |
---|
[] | []
|
---|
| (o) | | |
MW | (m) | | |
| (p) | | |
| (o) | | |
LMC | (m) | | |
| (p) | | |
| (o) | | |
SMC | (m) | | |
| (p) | | |
As mentioned in
Section 2.1, we derived the DM halo masses of the MW, LMC and SMC according to [
44] based on the stellar masses extracted from the references given in
Table 2. Our approach, therefore, is in full correspondence with the predictions of ΛCDM regarding the DM halo masses. However, in order to overcome possible uncertainties in that regard and to achieve more independent results, we decided to apply our statistical evaluations in
Section 4 individually to 27 mass combinations by varying the stellar mass of each galaxy by plus/minus
, as displayed in
Table 1. The 27 mass combinations are then denoted by the indicators “o” for the original masses, “m” for
and “p” for
, for instance, “o-m-p” for the MW/LMC/SMC mass triple.
The relevant coordinates and velocities are presented in
Table 3,
Table 4 and
Table 5, and the transformation to the Cartesian heliocentric equatorial system is shown in
Table 6 and
Table 7.
The Magellanic Stream consists of
gas, which trails behind both the LMC and the SMC across a large fraction of the sky. It is thought to be the result of a combination of tidal forces and ram pressure stripping through the orbit of the LMC and SMC within the hot gaseous halo of the MW. Studies of the origin of the Magellanic Stream have shown that it was most likely created when the LMC and the SMC had a close encounter, during which some of the gas of the LMC and SMC became less bound to be subsequently removed from the pair through ram pressure stripping [
39,
40,
62,
63].
4. Statistical Methods
In order to cross-check the solutions and to enhance the confidence in them, we utilised two independent statistical methods, a Markov-Chain Monte-Carlo method (MCMC, see
Section 4.1) and the genetic algorithm (GA, see
Section 4.2). The reason why we do so is given in the introductory statement of
Section 4.2.
This is the initial situation for each of the MCMC and GA methods: due to the error bars of the transverse velocity components in RA and DEC for the Magellanic Clouds (see
Table 4), we have to consider four open parameters
regarding the calculation of the three-body orbits of the dark matter halos, as displayed in
Table 8.
This means that by employing the MCMC method and the GA method separately for each of the 27 mass combinations (see
Section 3), we search for solutions of the three-galaxy orbits with best fits to the transverse velocity components of the Magellanic Clouds. It is important to note that our goal is to achieve results independent of the particular masses of the galaxies, not to find a best-fit mass combination.
Further, based on the radio-astronomical observational data concerning the Magellanic
-Stream explained in
Section 3, we specify the following broad condition, which we hereinafter refer to as the condition COND, regarding admissible past orbits of the Magellanic Clouds: the
LMC and SMC encountered each other within the past time interval of [ Gyr,
Gyr
] at a pericentre distance of less than kpc. Incorporating this condition, the algorithms searched for solutions by integrating Equation (
4) backwards in time up to
Gyr.
4.1. Markov-Chain Monte Carlo (MCMC)
We followed a methodology proposed by [
64] employing an affine-invariant ensemble sampler for the Markov-Chain Monte-Carlo method. A detailed guideline for implementation can be found in [
65]. The basics of applying this formalism to our situation are outlined in Appendix D of [
35].
4.1.1. Definition of the Posterior Probability Density
First of all, concerning the open parameters, we need to account for the error bars of the transverse velocity components
of the Magellanic Clouds. This is ensured by an appropriate definition of the prior distribution
, where
symbolises the parameter vector
. With the values from
Table 8, the four contributions to the prior distribution read (
)
Exploiting the minimal distance
between the LMC and SMC within the time period [
Gyr,
Gyr], the condition COND implies the likelihood function
with
kpc and
kpc. The posterior probability density is then given by the product of Equations (
8) and (
9):
Here, the normalising constant for an overall probability of 1 is neglected because it is clear from the comparative search algorithm that the absolute values of
are irrelevant.
4.1.2. The First Ensemble
Generating random integer numbers
separately for each walker and for each open parameter of a given walker, the first ensembles are created according to (
):
with varying
n to account for extended error bars, as explained in
Section 4.3.
4.2. Genetic Algorithm (GA)
The general aspects of the genetic algorithm are explained in detail in [
66]. As pointed out there, the major advantage of this method is perceived to be its capability of avoiding local maxima in the process of searching for the global maximum of a given function (here, the fitness function). This feature makes it worthwhile to employ the GA method as a second independent statistical approach besides the MCMC method.
A precise description of the algorithm, especially instructions for its implementation, can also be found in [
67]. To put it in a nutshell, the open parameters are related to genes, which are concatenated to genotypes. The set of a number of genotypes is considered to be a population, where the members pairs produce a follow-up generation with new features due to randomly performed cross-over and mutation. Winners per generation are determined by means of the fitness function, and by comparing the generations, the overall winner is found.
Comparing it to the more familiar MCMC method, we have the following parallels: walker <-> genotype, ensemble <-> population, set of ensembles <-> generations, posterior probability density <-> fitness function.
4.2.1. The GA Generations
Each open parameter from
Table 8 is mapped to a 4-digit string (“gene”)
(
),
again with varying
n, like in
Section 4.1.2, in order to account for extended error bars, as explained in
Section 4.3. All genes together define the genotypes as 4 × 4-digit strings,
which generate a population of
genotypes. The first generation is established by randomly creating
4 × 4-digit strings.
4.2.2. The Definition of the Fitness Function
Our goal is to establish GA evaluations that are compatible with the evaluations by means of the Markov-Chain Monte-Carlo method. Therefore, we chose the definition of the fitness function to be identical to the definition of the posterior probability density (Equation (
10)),
where the individual components are taken from Equations (
8) and (
9).
4.3. Results
Both statistical methods, MCMC and GA, deliver mutually consistent, and in fact indistinguishable, results.
4.3.1. Step I
As a first step, we tried to find solutions for which all four open parameters
lie within the
error bars of the transverse velocity components of the Magellanic Clouds (see
Table 8). Employing the methods MCMC and GA as search engines, using 1000 ensembles with 100 walkers (MCMC) and, accordingly, 1000 generations with a population of 100 genotypes (GA), we repeated the search 100 times for both statistical methods for each of the 27 mass combinations according to
Table 1. This attempt to find a solution failed. That is, within the here-probed
uncertainty range of the transverse velocity components, neither algorithm found orbits that fulfil the condition COND. In other words, the MW/LMC/SMC plus Magellanic Stream system cannot exist in its observed configuration in the presence of dark matter halos.
To proceed further, we gradually extended the allowed intervals for the parameters
to be
with increasing integers
n; see Equations (
11) and (
12). Neither the MCMC nor the GA statistical method found a solution for
. For
, both methods delivered solutions for a single mass combination only, namely, p-m-p. The details regarding all 27 mass combinations are given in
Appendix A.
4.3.2. Step II
How to go about quantifying the first results obtained in
Section 4.3.1? First of all, for a mass combination with the first solutions based on
error intervals, we chose the
intervals to be allowed for the open parameters. This is motivated by the thought that relaxing the overall
condition for all open parameters may result in solutions with individual smaller-than-
deviations of the parameters.
Furthermore, we constructed a “probability-sigma-grid” in the following sense: If a parameter lies within the error bar (), then its probability weighting factor is 1, a conservative cautious setting. If a parameter lies outside the error bar, then its deviation from is weighted with the remainder of the probability function based on intervals. For instance, if we have , then the corresponding probability weighting factor for this parameter is calculated according to the remainder of the probability function for .
Based on the same search method as in
Section 4.3.1, we searched for best-fit solutions by calculating an overall probability for the combination of the open parameters. The results are displayed in
Table 9 and
Table 10 for the methods MCMC and GA, respectively.
Regarding the MCMC method, it is interesting to note that, in some cases, individual parameters are not confined to the mentioned
intervals because the stretch moves can place walkers outside of the intervals specified by Equation (
11), thus supporting the overall process of maximising the posterior probability density consisting of all open parameters. This is not possible for the GA method, as the genotypes are a priori confined to the preset intervals.
The result in
Section 4.3.1 that the mass combination p-m-p provides the best-fit solution is confirmed by either statistical method, MCMC and GA. Furthermore,
Figure 1 shows the trajectories, displayed as pairwise distances, of the individual best-fit solutions for the mass combinations o-o-o and p-m-p. Even for that detail,
both methods deliver identical results.
4.3.3. Step III
Especially due to the MCMC method not being utilised as a pure search engine only, we established its validity by checking the results obtained in the previous
Section 4.3.2 in terms of the autocorrelation function for the mass combinations o-o-o (original masses) and p-m-p (mass combination with the overall best-fit solutions according to
Section 4.3.1 and
Section 4.3.2) and creating sets of 1000 solutions in each case
3. For the GA method, we undertook broader search runs, establishing 1000 solutions for the mentioned mass combinations.
The autocorrelation functions are presented in
Figure 2 and
Figure 3, demonstrating that convergence is achieved quickly for the MCMC method, and the improved probabilities are displayed in
Table 11 and
Table 12 for the MCMC and GA methods, respectively.
4.3.4. Interpretation
In Step I (
Section 4.3.1), it was shown that no orbital solutions are possible if the observational quantities are allowed to span the
uncertainty range. In Step II (
Section 4.3.2), the uncertainty range was therefore increased to
, with the result that values of
n are needed to obtain viable orbital solutions that lie outside the
confidence range of the quantities such that the orbital solutions are likely with probabilities smaller than
(
Table 9 and
Table 10). Step III (
Section 4.3.3) demonstrates that the MCMC and GA methods yield indistinguishable results, thus confirming the conclusion that orbital solutions do not exist for the observed MW/LMC/SMC plus Magellanic Stream system in the presence of the theoretically expected DM halos.
The dynamical behaviour of the MW/LMC/SMC system demonstrates the importance of dynamical friction in the context of interacting galaxies, as dynamical friction significantly influences the orbits.
Figure 4 shows that the forces due to dynamical friction between the LMC and the SMC are comparable to the pure gravitational force between the overlapping DM halos.
As a thought experiment, we calculated the orbits back in time to
Gyr
with and
without dynamical friction for the overall best-fit solution derived for the mass combination p-m-p and for the best-fit solution for the original mass combination o-o-o. Dynamical friction can be turned off by omitting the terms
in Equation (
5). This retains the gravitational pull of the DM halo but avoids the deceleration through Chandrasekhar dynamical friction. This Gedanken experiment is thus a rough approximation of the situation in Milgromian dynamics, according to which a galaxy generates a Milgromian gravitational potential that can be viewed as stemming from a Newtonian plus phantom DM halo that does not generate dynamical friction. With this approximation of MOND, the GA method as the search engine immediately delivers 17 mass combinations already within the
intervals (see also
Appendix A). In other words, solutions matching the error intervals for the transverse velocities of the LMC and SMC are found readily without Chandrasekhar’s dynamical friction but with the potential generated by the DM halo. The result is displayed in
Figure 5. With the absence of Chandrasekhar dynamical friction on DM halos, the Magellanic Clouds would have a long orbital lifetime as satellites of the MW, possibly being massive tidal dwarf galaxies formed during the MW–Andromeda encounter about 10 Gyr ago.
These results thus suggest that the MW/LMC/SMC plus Magellanic Stream system may have a straightforward orbital solution in Milgromian dynamics, and it is thus of much interest to simulate this system and its possible origin in this non-Newtonian framework.