Dark Matter Velocity Distributions for Direct Detection: Astrophysical Uncertainties are Smaller Than They Appear

Using the largest sample to date of nearly 100 Milky Way-like galaxies from the TNG50 simulation and a novel phase-space scaling procedure, this study demonstrates that astrophysical uncertainties in dark matter velocity distributions result in cross-section limits varying by only ~60%, a level comparable to or smaller than the systematic uncertainties of current ton-scale direct detection experiments.

The Gist

The Big Picture: Hunting Ghosts in the Dark

Imagine scientists are trying to catch a ghost. This ghost is Dark Matter, an invisible substance that makes up most of the universe. To catch it, they build giant, ultra-sensitive detectors deep underground (like the XENON1T experiment). They wait for a ghost to bump into an atom in their detector, hoping to see a tiny flash of light.

But there's a problem. To know if they've caught a ghost, they need to know how fast the ghosts are running.

If the ghosts are running slowly, they might not have enough energy to make a flash. If they are running fast, they will. The scientists have a standard guess for how fast these ghosts run, called the "Standard Halo Model." It's like assuming all the ghosts in a city are jogging at a steady 10 mph.

However, the authors of this paper say: "Wait a minute. We don't actually know if they are jogging. Maybe some are sprinting, and some are walking. And maybe our 'standard guess' is wrong because we've been looking at the wrong neighborhoods."

The Problem: The "Map" Was Wrong

To figure out the speed of these ghosts, scientists use supercomputer simulations. They build virtual universes and watch how galaxies form. But here's the catch: Our Milky Way is weird.

Think of the Milky Way as a very compact, dense city. Most other galaxies of the same size are sprawling suburbs with loose, spread-out neighborhoods.

• The Simulation Issue: The computer simulations they used (called TNG50) were great at making sprawling suburbs, but they struggled to make a compact city like ours.
• The Result: In these simulations, the "ghosts" (dark matter) and the "stars" were moving too slowly because the virtual city was too spread out. It was like trying to predict traffic in New York City by studying traffic in a small, empty town. The speed limits were all wrong.

Because the simulations were "too slow," the scientists thought the uncertainty in their data was huge. They worried that if they got the speed wrong, they might miss the ghost entirely or claim they found it when they didn't.

The Solution: The "Stretch and Shrink" Trick

The authors realized they didn't need to throw away their simulations. They just needed to fix the map. They invented a clever mathematical trick, which they call "Phase-Space Scaling."

Imagine you have a photo of a galaxy that looks too big and fluffy.

1. The Shrink: They mathematically "shrink" the photo, pulling everything closer to the center, just like zooming in on a map.
2. The Boost: Because gravity gets stronger when things are closer together, they also "boost" the speed of everything in the photo.

The Analogy: Think of a figure skater spinning. When they pull their arms in (shrinking the space), they spin faster (boosting the speed). The authors did this to their virtual galaxies so that the stars and dark matter moved at the exact speed we observe in our real Milky Way.

The Surprising Discovery: It's Not as Scary as We Thought

Once they applied this "shrink and boost" fix to nearly 100 different virtual galaxies, they looked at the results. They expected to find a chaotic mess of different speeds, making it impossible to predict anything.

Instead, they found order.

• The "Standard Guess" Was Actually Good: Even though the individual virtual galaxies looked a bit different, when you averaged them all out, the speeds of the dark matter matched the "Standard Halo Model" (the steady 10 mph jog) surprisingly well.
• The Uncertainty Shrank: Before this fix, scientists thought the uncertainty in their data was huge (like a 60% margin of error). After the fix, the uncertainty dropped to a level that is smaller than the errors caused by the detectors themselves.

The Metaphor: Imagine you are trying to guess the average height of people in a room.

• Before: You were looking at a room full of people wearing stilts and hats, and you were guessing their heights based on a blurry photo. You thought, "I could be off by 2 feet!"
• After: You took off the stilts, fixed the photo, and realized, "Oh, they are all actually within 2 inches of the average height." The uncertainty wasn't the people; it was your measurement tool.

Why Does This Matter?

This is a huge win for the hunt for Dark Matter.

1. Confidence: Scientists can now trust their experiments more. They know that the "noise" from the universe (astrophysical uncertainty) is smaller than the "noise" from their machines.
2. Focus: Instead of worrying about whether the dark matter is moving weirdly, they can focus on improving their detectors and understanding the physics of the particles themselves.
3. History Doesn't Matter Much: They checked if the "family history" of the galaxy (like big crashes with other galaxies in the past) changed the speed. It didn't change the results much. The "Standard Model" holds up even for galaxies with different backstories.

The Bottom Line

The authors took a messy, confusing set of computer simulations, fixed them with a clever math trick to match our real galaxy, and discovered that the universe is actually more predictable than we thought.

The "Standard Halo Model" isn't perfect, but it's good enough. The biggest source of error in catching dark matter isn't the mystery of the galaxy anymore; it's just the limits of our current technology. We are closer to finding the ghost than we thought.

Technical Summary

1. Problem Statement

Direct detection experiments for Dark Matter (DM) rely on the scattering rate of DM particles off terrestrial nuclei. This rate depends critically on the local DM speed distribution ($f(v)$) at the Solar position.

• The Challenge: Theoretical predictions often rely on the Standard Halo Model (SHM), which assumes a Maxwell-Boltzmann distribution truncated at the Galactic escape speed. However, cosmological simulations suggest significant deviations from the SHM due to the complex merger history and baryonic physics of the Milky Way (MW).
• The Bottleneck: Previous studies struggled to quantify the "halo-to-halo" variance (astrophysical uncertainty) because:
  1. Simulated MW-like galaxies often have disk properties (e.g., scale length) that differ from the actual MW, leading to incorrect local circular speeds ($v_{LSR}$).
  2. Finding a simulated galaxy that perfectly matches the MW's compact disk properties is rare, requiring prohibitively large sample sizes to characterize the full variance.
  3. Consequently, uncertainties in DM speed distributions were assumed to be large, potentially dominating experimental systematic errors, especially for low-mass DM where only high-speed particles are detectable.

2. Methodology

The authors utilized the TNG50 simulation (part of the IllustrisTNG suite), which offers high-resolution hydrodynamical modeling of galaxy formation.

• Sample Selection: They identified 98 MW-like halos satisfying specific criteria:
  • Stellar mass consistent with the MW: $(4–7.3) \times 10^{10} M_\odot$.
  • Isolated environment: $>500$ kpc from larger halos and $>1$ Mpc from halos with $M_{dyn} > 10^{13} M_\odot$.
• The Scaling Procedure (Key Innovation):
  • Problem: In the raw TNG50 sample, the average stellar speed at the Solar radius ($R_\odot = 8.3$ kpc) was $\sim 198$ km/s, significantly lower than the observed Local Standard of Rest ($v_{LSR} = 238$ km/s). This discrepancy arises because TNG50 disks are generally more extended than the MW.
  • Solution: The authors introduced an energy-conserving phase-space scaling to align the simulations with the MW's observed $v_{LSR}$.
    1. Define a target mass $M_{LSR} = v_{LSR}^2 R_\odot / G$.
    2. For each halo, find the radius $R_M$ enclosing mass $M_{LSR}$.
    3. Scale positions ($\mathbf{r}$) and velocities ($\mathbf{v}$) such that $R_M \to R_\odot$:
       $$\mathbf{r} \mapsto \frac{R_\odot}{R_M} \mathbf{r}, \quad \mathbf{v} \mapsto \sqrt{\frac{R_M}{R_\odot}} \mathbf{v}$$
  • Justification: This transformation preserves the collisionless Boltzmann equation and the form of the gravitational potential, effectively changing units to match the MW's dynamical scale without violating physical laws.
• Analysis:
  • Constructed DM speed distributions in the "solar annulus" ($|r_{cyl} - R_\odot| < 1$ kpc, $|z| < 1$ kpc) for all 98 halos.
  • Tested sub-samples based on merger history: those with Gaia Sausage–Enceladus (GSE) analogs and those with Large Magellanic Cloud (LMC) analogs.
  • Calculated direct detection limits (spin-independent cross-section $\sigma_{SI}$) using the wimprates package and a simplified XENON1T analysis pipeline.

3. Key Results

A. Velocity Distribution Characteristics

• Alignment with SHM: After scaling, the ensemble of 98 halos aligns remarkably well with the SHM. The best-fit peak speed ($v_0$) for scaled halos is $240 \pm 11$ km/s (vs. $v_{LSR}=238$ km/s), compared to $213 \pm 19$ km/s for unscaled halos.
• Non-Maxwellian Features: While the ensemble average fits the SHM, individual halos exhibit deviations, particularly in the high-speed tail and cylindrical velocity components (excess kurtosis).
• Merger History Impact:
  • LMC analogs: Slightly enhance the high-speed tail (0.6% of particles above escape speed vs. 0.1% in the general population).
  • GSE analogs: Slightly suppress the high-speed tail (0.2% above escape speed).
  • Crucially: These shifts are small and remain within the $1\sigma$ halo-to-halo variance of the full sample.

B. Impact on Direct Detection Limits

• Uncertainty Quantification: The variation in the DM-nucleon cross-section limits ($\sigma_{SI}$) across the 98 halos is approximately 60% around the median.
• Comparison to Systematics: This astrophysical uncertainty ($\sim 60\%$) is comparable to or smaller than the systematic uncertainties of current ton-scale detectors (e.g., XENON1T, LZ) near their energy thresholds.
• Low-Mass Regime: For low-mass DM ($m_\chi \lesssim 10$ GeV), unscaled simulations significantly underestimated limits (due to slower speeds). The scaling procedure corrected this, bringing limits into agreement with SHM predictions.

C. Validation

• The scaled distributions were validated against:
  1. The subset of unscaled halos that naturally had compact disks ($R_M \approx R_\odot$).
  2. Results from the FIRE-2 simulation suite (which uses different subgrid physics), showing consistent speed distributions once circular speeds are matched.

4. Key Contributions

1. Largest Sample to Date: Provided the most comprehensive statistical characterization of local DM velocity distributions using 98 high-resolution MW analogs.
2. Novel Scaling Method: Developed a principled, energy-conserving phase-space scaling technique that allows the use of the entire TNG50 sample (not just rare outliers) to model the MW, correcting for disk size discrepancies.
3. Quantified Uncertainty: Demonstrated that astrophysical uncertainties in the local speed distribution are not the dominant limitation for current experiments; systematic detector uncertainties are often larger.
4. Data Release: Provided tabulated speed distributions and best-fit Maxwell-Boltzmann parameters ($v_0, v_{esc}$) for use in future direct detection studies.

5. Significance and Implications

• Re-evaluating Limits: The paper suggests that previous fears of large astrophysical uncertainties obscuring DM signals may be overstated. The "halo-to-halo" variance is manageable and does not preclude robust interpretation of null results or potential signals.
• Focus Shift: Since astrophysical uncertainties are now understood to be $\lesssim 60\%$ (and often lower), the community should prioritize reducing systematic experimental uncertainties, such as detector response functions near energy thresholds and nuclear form factors.
• Standardization: The authors provide a "global fit" (Table I) for the MW speed distribution ($v_0 = 240 \pm 11$ km/s, $v_{esc} = 567 \pm 43$ km/s) that marginalizes over halo variance, offering a more robust input for experimental analyses than the canonical SHM values.

In conclusion, this work bridges the gap between high-fidelity cosmological simulations and experimental dark matter searches, demonstrating that with proper scaling, simulations can reliably predict local DM kinematics with uncertainties that are no longer the limiting factor in direct detection sensitivity.