New Simulations Challenge Milky Way–Andromeda Collision Odds

Introduction

For decades, textbooks have taught that our Milky Way galaxy is destined to collide and merge with its larger neighbor, Andromeda (M31), in roughly 4–5 billion years. This classic scenario—sometimes dubbed “Milkomeda”—has become a staple illustration of cosmic evolution. However, a new study published in Nature Astronomy challenges that narrative, showing only a ~50% chance of collision over the next 10 billion years when the full Local Group dynamics are modeled.

Methodology: Data Fusion and Monte Carlo Simulations

The team, led by Till Sawala (University of Helsinki), merged high-precision astrometric and photometric data from NASA’s Hubble Space Telescope (HST) with ESA’s Gaia Early Data Release 3 (EDR3). They used:

• Proper motions: Sub-microarcsecond measurements over multi-year baselines.
• Radial velocities: Spectroscopic data with uncertainties < 2 km/s.
• Mass profiles: Virial mass estimates for dark matter halos and baryonic components, spanning 0.5–2×10¹² M_⊙.

These inputs fed into 100,000 Monte Carlo realizations, each evolving the full Local Group (LG) over 10 Gyr using N-body codes (GADGET-2 and REBOUND) with explicit dynamical friction terms. The simulations accounted for tidal forces, halo shape asymmetries, and satellite perturbations.

Results: Collision Probabilities Revised

Contrary to prior estimates (>90% chance of merger), the study finds:

1. ~50% probability of a direct Milky Way–Andromeda collision within 10 Gyr.
2. ~2% chance of impact in the next 4–5 Gyr.
3. Remaining trajectories involve multiple close passages but no full merger, leading to prolonged orbital dance.

“Based on the best available data, the fate of our galaxy is still completely open,”

Key to this revision are the gravitational influences of M33 and the Large Magellanic Cloud (LMC). M33’s mass (~5×10¹⁰ M_⊙) slightly enhances merger likelihood by pulling both giants inward, while the LMC (~1.5×10¹¹ M_⊙) perturbs the Milky Way off the collision plane, reducing direct impact odds.

Modeling Techniques and Computational Infrastructure

The study leveraged Europe’s PRACE & US XSEDE supercomputing centers, deploying:

• 4,000 CPU cores per simulation batch and GPU acceleration for force calculations.
• Adaptive time stepping (∆t ~0.1 Myr) and softening lengths tuned to 100 pc.
• Customized modules for Chandrasekhar dynamical friction and mass loss from tidal stripping.

By sampling halo spin parameters, triaxiality, and satellite infall angles, the team estimated uncertainties in orbital decay timescales (±1 Gyr) and merger cross-sections.

Implications for Galaxy Evolution Theory

If a collision does occur, it will drive intense starbursts, transform spiral disks into an elliptical remnant, and generate extended tidal streams. However, the revised 50/50 odds suggest the Local Group may remain a multi-galaxy ensemble far longer than expected, affecting predictions of intragroup gas heating, metal enrichment, and future gravitational wave signatures from supermassive black hole interactions.

Future Observational Campaigns and Data Needs

To further refine collision forecasts, astronomers point to:

• James Webb Space Telescope: Deep proper motion studies of faint LG satellites.
• Nancy Grace Roman Telescope: Wide-field infrared mapping of halo substructure.
• Vera C. Rubin Observatory (LSST): Time-domain constraints on LG dwarf galaxy orbits.

Expected Advances in Proper Motion Precision

Next-generation astrometry aims for tens of nanoarcsecond accuracy, reducing velocity uncertainties to <1 km/s and tightening orbital phase space constraints.

Role of AI in Simulation and Data Analysis

Machine learning surrogates (e.g., Gaussian process emulators, neural network potentials) are increasingly adopted to explore parameter spaces more efficiently, complementing traditional N-body integration.

Conclusion

While the iconic Milky Way–Andromeda collision scenario may still unfold, the updated 50% probability underscores the complexity of galactic dynamics and the crucial role of high-precision data. As new observatories come online and computational methods evolve, our cosmic forecast will become ever clearer—though for now, the ultimate fate of our galaxy remains uncertain.

DOI: 10.1038/s41550-025-02563-1