Quantifying the Milky Way, LMC and their interaction using all-sky kinematics of outer halo stars

By analyzing all-sky kinematics of outer halo stars out to 160 kpc using Simulation Based Inference on 32,000 rigid MW-LMC simulations, this study quantifies the Milky Way's and LMC's masses while characterizing the reflex motion induced by the LMC's recent pericentric passage, revealing that the LMC's mass is at least 20% of the Milky Way's and that neglecting this interaction biases mass estimates.

The Gist

Here is an explanation of the paper, translated into everyday language with some creative analogies.

The Big Picture: A Cosmic Dance Gone Wrong

Imagine the Milky Way (our home galaxy) as a giant, spinning figure skater. For billions of years, this skater has been spinning smoothly and predictably. Scientists have been trying to figure out exactly how heavy the skater is (the galaxy's mass) by watching how fast the stars on the outer edges are moving.

But recently, a very heavy partner—the Large Magellanic Cloud (LMC), a smaller satellite galaxy—came zooming in and grabbed the skater's hand. This wasn't a graceful waltz; it was a high-speed collision course. The LMC is so massive that as it swung around the Milky Way, it didn't just pull on the stars; it actually jerked the entire Milky Way off its center.

This paper is about measuring exactly how hard that "jerk" was, how heavy the LMC is, and how heavy the Milky Way is, all by looking at the stars that got shaken up during the dance.

The Problem: The "Reflex" Effect

When you are on a bus and the driver suddenly swerves left, your body feels like it's being thrown to the right. You haven't moved on your own; the bus moved, and your body is reacting to that motion.

In space, the Milky Way is the bus, and the LMC is the driver swerving. Because the LMC pulled the Milky Way's center of mass "down," the outer stars of the Milky Way are reacting by moving "up." Astronomers call this reflex motion.

If you try to weigh the Milky Way without realizing the bus is swerving, you will get the wrong answer. You might think the stars are moving fast because the bus is heavy, when actually, they are just reacting to the swerve.

The Solution: A Digital Time Machine

To solve this, the scientists didn't just look at the real stars; they built a giant digital simulation.

1. The Data: They gathered information on 1,296 stars from the outer edges of our galaxy (up to 160,000 light-years away) using three different telescopes (H3, SEGUE, and MagE). Think of this as taking a snapshot of the stars' speed and direction.
2. The Simulations: They created 32,000 different versions of the Milky Way and the LMC on a computer. In each version, they changed the weights (masses) of the galaxies, the speed of the LMC, and how the stars were moving.
3. The AI Detective: They used a type of Artificial Intelligence called Simulation Based Inference (SBI). Imagine feeding the AI all 32,000 simulated scenarios and the real star data. The AI's job is to say, "Out of all these 32,000 possibilities, which ones look most like the real universe we are seeing?"

What They Found

By letting the AI compare the simulations to the real data, they found some very specific answers:

• The Weight of the Galaxy: They calculated that the Milky Way contains about 363 billion times the mass of our Sun within its inner 50,000 light-years.
• The Weight of the LMC: The Large Magellanic Cloud is surprisingly heavy. It weighs about 97 billion Suns. This means the LMC is roughly 20% to 30% as heavy as the Milky Way. That's like a heavyweight boxer fighting a middleweight; the smaller one is still huge!
• The "Jerk" Speed: They measured the speed of the reflex motion (the "swerve"). At a distance of 100,000 light-years, the stars are being pushed at about 39 km/s (roughly 87,000 mph).
• The Shape of the Orbit: Before the LMC arrived, the stars in the outer galaxy were already moving on very stretched-out, oval-shaped orbits (like a racetrack rather than a circle).

Why This Matters

The paper makes a crucial point: If you ignore the LMC, you get the wrong answer.

The scientists ran a test where they pretended the LMC didn't exist. They found that without accounting for the LMC's "jerk," their estimate of the Milky Way's mass was about 5% too high. It's a small number, but in the world of astronomy, that's a huge difference. It's like weighing yourself on a scale that is currently being pushed down by a friend; if you don't account for the friend, you think you are heavier than you really are.

The Takeaway

This study is a breakthrough because it combines real star data with massive computer simulations and AI to finally "weigh" our galaxy while it is in the middle of a chaotic event.

In short: The Milky Way is currently being shaken by a massive neighbor. By measuring how the stars are wobbling, we can finally figure out exactly how heavy both galaxies are and how they are interacting. It turns out the Milky Way is a bit lighter than we thought, and the LMC is a much bigger player in the cosmic neighborhood than we realized.

Technical Summary

Here is a detailed technical summary of the paper "Quantifying the Milky Way, LMC and their interaction using all-sky kinematics of outer halo stars" by Brooks et al. (2026).

1. Problem Statement

The mass of the Milky Way (MW) is a fundamental property required for constructing dynamical models of our Galaxy. Historically, these estimates have relied on the assumption that the MW is in a state of dynamical equilibrium (e.g., using Jeans models). However, the Large Magellanic Cloud (LMC) is currently undergoing its first pericentric passage through the MW. This massive interaction ($M_{LMC} \sim 10^{11} M_\odot$) has induced a state of dynamical disequilibrium throughout the MW, specifically causing a "reflex motion" where the MW's center of mass moves in response to the LMC's gravity.

Previous studies using outer halo stars often ignored this reflex motion or relied solely on mean velocities, which limited the ability to precisely constrain the MW mass (which is sensitive to velocity dispersions). Furthermore, ignoring the LMC in equilibrium models can bias mass estimates. The challenge is to simultaneously constrain the masses of both the MW and LMC, the parameters of their interaction (reflex motion), and the intrinsic properties of the stellar halo (velocity anisotropy) while accounting for this non-equilibrium state.

2. Methodology

The authors employ Simulation Based Inference (SBI), specifically using a Density Estimation Likelihood-Free Inference (DELFI) framework, to bypass the need for an analytic likelihood function, which is intractable for complex, non-equilibrium systems.

• Data Source: The study utilizes a combined all-sky sample of 1,296 outer halo stars (distances $30-160$ kpc) from the H3 + SEGUE + MagE surveys.
  • Summary Statistics: Instead of using individual star positions, the authors use the mean and velocity dispersions of radial ($v_{GSR}$) and tangential ($v_{t,b}$) velocities, binned by Galactocentric distance and sky quadrant (North/South). This approach is robust against differential selection effects between surveys.
• Simulations:
  • A suite of 32,000 rigid MW–LMC simulations was generated.
  • MW Potential: Modeled with an NFW dark matter halo, a stellar disc, and a bulge.
  • LMC Potential: Modeled as a Hernquist dark matter halo with dynamical friction (Chandrasekhar prescription) acting on its orbit.
  • Stellar Halo: Generated using distribution functions with a radially biased velocity anisotropy ($\beta_0$).
  • Evolution: Particles were integrated over the last 2.2 Gyr to evolve the halo to the present day under the combined potential.
• Reflex Motion Modeling:
  • The authors implemented a linear continuity model for the reflex motion, allowing the reflex velocity to vary linearly as a function of Galactocentric distance (anchored at 40 kpc and 120 kpc), improving upon previous distance-independent models.
• Inference Engine:
  • A Masked Autoregressive Flow (MAF) neural network was trained on the simulation suite to learn the conditional probability distribution $p(\theta | D_{obs})$, where $\theta$ are model parameters and $D_{obs}$ are the observed summary statistics.
  • Spurious Data Handling: Data points lying $>3\sigma$ from the simulation mean were identified and removed to prevent bias from unmodeled substructures or simulation limitations.

3. Key Contributions

1. Simultaneous Mass Constraints: First application of SBI to outer halo kinematics that simultaneously constrains the enclosed masses of both the MW and LMC using both mean velocities and velocity dispersions.
2. Radial Reflex Motion Model: Introduction of a linear continuity model to capture the radial dependence of the reflex motion, demonstrating that the reflex velocity magnitude increases with distance.
3. Quantification of Bias: A rigorous assessment of how neglecting the LMC (assuming equilibrium) biases MW mass estimates.
4. Velocity Anisotropy: Constrained the pre-infall velocity anisotropy ($\beta_0$) of the MW stellar halo, showing that mean velocities in a non-equilibrium system can inform anisotropy constraints.

4. Key Results

• Enclosed Masses (at 50 kpc):
  • Milky Way: $M_{MW}(< 50 \text{ kpc}) = 3.63 \pm 0.16 \times 10^{11} M_\odot$.
  • LMC: $M_{LMC}(< 50 \text{ kpc}) = 9.74^{+2.07}_{-1.81} \times 10^{10} M_\odot$.
  • The total LMC mass is estimated to be at least $\sim 20\%$ of the MW's total mass, with an infall-to-present-day mass ratio of $\sim 30 \pm 10\%$.
• Reflex Motion:
  • At a reference distance of 100 kpc, the magnitude of the reflex travel velocity is $v_{travel} = 39.4^{+7.6}_{-7.2}$ km s$^{-1}$.
  • The apex direction is constrained to $(l_{apex}, b_{apex}) \approx (-6.8^\circ, -78.2^\circ)$.
  • The reflex velocity shows a clear radial trend, increasing from $\sim 17$ km s$^{-1}$ at 40 kpc to $\sim 49$ km s$^{-1}$ at 120 kpc.
• Velocity Anisotropy:
  • The average velocity anisotropy prior to LMC infall is constrained to $\beta_0 = 0.61 \pm 0.03$. This indicates the halo was already radially anisotropic before the LMC's recent passage.
• Impact of LMC on Mass Inference:
  • Models that exclude the LMC (assuming equilibrium) bias the MW mass estimate to be $\sim 5\%$ more massive.
  • However, the study finds that while mean velocities are highly sensitive to the LMC's reflex motion, velocity dispersions remain largely insensitive to the LMC's presence, yielding relatively unbiased mass constraints even in equilibrium models (though the equilibrium assumption is physically incorrect).

5. Significance and Outlook

• Methodological Advancement: This work demonstrates the power of SBI in handling complex, non-equilibrium astrophysical systems where traditional Jeans modeling fails. It validates that rigid models, when trained via SBI, can capture sufficient physics to constrain parameters without the computational cost of full $N$-body simulations for every inference step.
• Physical Insight: The results confirm that the LMC's interaction is a dominant driver of the MW's current dynamical state. The precise measurement of the reflex motion provides a direct probe of the MW-LMC mass ratio and interaction history.
• Future Surveys: The framework is designed to be flexible for upcoming datasets from Gaia DR4 and SDSS-V, which will provide significantly larger samples of outer halo stars with improved proper motions, further tightening constraints on the MW and LMC masses and the nature of their interaction.

In summary, this paper provides the most precise joint constraints on the MW and LMC masses to date using outer halo kinematics, explicitly accounting for the dynamical disequilibrium caused by the LMC's infall, and establishes a robust pipeline for future galactic archaeology studies.