Advancing Stellar Streams as a Dark Matter Probe -- I: Evolution of the CDM subhalo population

This paper develops a more detailed semi-analytic model for the statistics of gaps in stellar streams caused by dark matter subhalos, with the goal of improving the accuracy of predictions for future surveys.

The authors state: "Prior models for the statistics of such gaps have relied on several simplifying assumptions for the properties of the subhalo population in the cold dark matter scenario. With the expected forthcoming increase in the number of streams and gaps observed, this work develops a more detailed model for the statistics of subhalos interacting with streams and tests some of the assumptions made in prior works."

The key methodological improvement is the use of a physical, time-dependent subhalo population model: "Instead of using simple fits to N-body estimates of subhalo population statistics at z = 0 as in previous work, we make use of realizations of time-dependent subhalo populations generated from an entirely physical model, incorporating structure formation and subhalo orbital evolution, including tidal heating and stripping physics, which has been carefully calibrated to match results of cosmological N-body simulations."

The main result is that this new model predicts more gaps than previous work: We find that this model predicts 20% more gaps (up to 60% for deep gaps) on average in Pal-5-like streams than prior works.

The paper details the methodology, which is divided into two main parts: generating subhalo populations and modeling gap formation. Subhalo populations are generated using the Galacticus semi-analytic model, which constructs merger trees for a Milky Way-mass halo and evolves subhalos including tidal stripping and heating. The authors generate 4,000 merger tree realizations, and through random rotations, effectively create 1,600,000 realizations to achieve 1% statistical precision.

For gap formation, the model treats interactions as impulsive scattering events. A key new formula for velocity kicks is derived (eqn. 17), which is separable into geometric and profile-specific components. The paper states: We find that this model predicts 20% more gaps (up to 60% for deep gaps) on average in Pal-5-like streams than prior works. The model also includes an exact solution for the orbital shift and a numerical evaluation of gap sizes valid at all times, avoiding previous approximations.

The paper tests six different subhalo density profiles: a point mass, a thin shell, a constant density model, a Plummer sphere, a pure NFW profile, and a tidally stripped NFW profile. The results show significant model dependence. For example, The cuspy NFW and tidally stripped NFW profiles produce more gaps (of given depth and size) than the cored Plummer profile due to their more concentrated mass distributions. The point mass naturally produces the most gaps, while the Plummer model produces the fewest. The paper notes: "Comparing our Plummer and Truncated profiles in Table 3 and Figure 8, the Plummer model gives 40% fewer gaps at rho˜ c = 0.5, and 7 times fewer gaps at rho˜ c = 0.1. Therefore, the proper mass profile is vital for realistic gap modelling."

The paper also finds that lower-mass subhalos are more important than previously thought: "Figure 7 shows that with a cuspy mass profile, 105 −106 M⊙ subhalos outperform 108 −109 M⊙ subhalos at all gap depths, and rival 107 − 108 M⊙ subhalos for small rho˜ c. Further work is required to place a lower bound on gap-producing subhalos, but it is very possible that perturbers ≲ 104 M ⊙ can be detected."

The paper concludes that semi-analytic modeling is a viable path forward for large-scale tests of dark matter using stellar streams, and outlines future improvements such as including baryonic effects and realistic stream orbits.

Based on this paper, I can identify several specific improvements to AI systems, particularly in the areas of scientific simulation, statistical modeling, and data analysis.

1. Improved Subhalo Population Modeling for Dark Matter Simulations

• Improvement: Replace static, redshift-zero subhalo population fits (as used in prior works) with a time-dependent, physics-based model that incorporates structure formation, orbital evolution, tidal heating, and stripping.

• What the improved AI system can do: Generate realistic, time-resolved realizations of dark matter subhalo populations (e.g., using Galacticus) that accurately match N-body simulation results, including mass functions, radial distributions, and density profile evolution. This enables more accurate predictions of gravitational interactions with stellar streams, leading to a 20% increase in predicted gap counts (up to 60% for deep gaps) compared to previous models.

2. Faster and More Accurate Velocity Kick Calculations

• Improvement: Replace brute-force numerical integration of velocity kicks with a new analytical form (Eq. 17) that separates geometric and profile-specific components, and use hypergeometric functions for non-integrable profiles.

• What the improved AI system can do: Compute velocity kicks for subhalo-stream interactions up to 3,000 times faster than previous methods (Table 1). This enables large-scale Monte Carlo simulations (e.g., 1.6 million realizations) with 1% statistical precision, which is critical for distinguishing between dark matter models. The system can also handle realistic density profiles (e.g., tidally stripped NFW) that were previously computationally prohibitive.

3. Exact Orbital Shift and Gap Evolution Modeling

• Improvement: Solve the orbital shift equation (Eq. 32) exactly, rather than using series expansions, and compute gap sizes numerically at all times (not just early/intermediate/late approximations).

• What the improved AI system can do: Predict the evolution of gaps in stellar streams with higher accuracy, including the transition between regimes. The system can also model the refilling of gaps due to stellar velocity dispersion, which was previously neglected. This leads to more realistic predictions of observable gap sizes and depths, improving the sensitivity of stream-based dark matter probes.

4. Efficient Subhalo Pre-Screening and Impact Parameter Computation

• Improvement: Use a series of algebraic filters and Sturm chain methods to eliminate 98% of irrelevant subhalos before computing impact parameters, and solve for impact parameters via quartic roots instead of global minimization.

• What the improved AI system can do: Reduce post-processing time from 1 day per timestep to 3 minutes, enabling the analysis of millions of subhalo-stream encounters. The system can precisely identify the closest approach distance and time for each interaction, which is essential for accurate gap formation modeling.

5. Model-Dependent Gap Statistics and Sensitivity Analysis

• Improvement: Systematically compare six different subhalo density profiles (point, shell, constant, Plummer, NFW, tidally stripped NFW) and quantify their impact on gap size distributions and counts.

• What the improved AI system can do: Quantify the sensitivity of gap statistics to subhalo internal structure, revealing that cuspy profiles (NFW) produce significantly more gaps than cored profiles (Plummer). The system can also identify which observables (e.g., gap width, depth, slope) are most discriminative between dark matter models, guiding future observational strategies.

6. Scalable Framework for Multi-Stream Analysis

• Improvement: Extend the model to handle arbitrary stream radii, lengths, and ages, and use a circular-orbit approximation with length-weighting to increase statistical power.

• What the improved AI system can do: Predict gap counts as a function of galactocentric radius (Figure 9), showing that inner streams are more sensitive to subhalos. The system can be applied to the full ensemble of 250 known streams (and thousands expected from future surveys) to provide statistically robust constraints on dark matter, rather than relying on individual streams.

7. Identification of New Physical Insights

• Improvement: The system reveals unexpected results, such as the shell profile producing more gaps than the Plummer model, and the importance of subhalos below 10 5 M sun for deep gap formation.

• What the improved AI system can do: Automatically flag and investigate such counterintuitive results, potentially uncovering new physics or model dependencies. For example, the system can determine that gap width and depth alone are insufficient to characterize interactions, suggesting the need for new observables like gap slope or integrated underdensity.

8. Forward Modeling and Likelihood-Free Inference

• Improvement: The semi-analytic framework is designed for forward modeling, enabling the use of approximate Bayesian computing or simulation-based inference.

• What the improved AI system can do: Generate millions of synthetic stream observations under different dark matter models and compare them to real data without requiring explicit likelihood functions. This is crucial for combining information from multiple streams and for handling complex, multi-parameter dark matter models.

Stellar streams, long, thin streams of stars, have been used as sensitive probes of dark matter substructure for over two decades. Gravitational interactions between dark matter substructures and streams lead to the formation of low-density ``gaps'' in streams, with any given stream typically containing no more than a few such gaps. Prior models for the statistics of such gaps have relied on several simplifying assumptions for the properties of the subhalo population in the cold dark matter scenario. With the expected forthcoming increase in the number of streams and gaps observed, this work develops a more detailed model for the statistics of subhalos interacting with streams and tests some of the assumptions made in prior works. Instead of using simple fits to N-body estimates of subhalo population statistics at z=0 as in previous work, we make use of realizations of time-dependent subhalo populations generated from an entirely physical model, incorporating structure formation and subhalo orbital evolution, including tidal heating and stripping physics, which has been carefully calibrated to match results of cosmological N-body simulations. We find that this model predicts 20% more gaps (up to 60% for deep gaps) on average in Pal-5-like streams than prior works.