Stellar Streams and Galactic Potentials

Scrollable formula. Use Left and Right arrow keys to reveal hidden content.

Scrollable table. Use Left and Right arrow keys to reveal additional columns.

Scrollable figure. Use Left and Right arrow keys to inspect hidden content.

A cold stellar stream records how stars stripped from a cluster or dwarf move through the Galactic field. Its long, thin structure samples the potential over radii and times that a single star cannot. Yet a stream is not one orbit: stars leave the progenitor with a distribution of energies and angular momenta, then respond to the disk, bar, satellites, subhalos, and time-dependent Milky Way. Gravity is tested only when those source and perturbation layers are forward-modeled into the same positions, distances, proper motions, radial velocities, widths, and density variations that define the data. The status here is current through 2026-08-19.

Required background. Classical Field Theory supplies gravitational potentials and equations of motion, while Probability and Statistical Inference supplies latent-track likelihoods, selection normalization, covariance, and predictive validation.

Helpful background. Weak-Field Gravity supplies the potential limit, and Black-Hole Astrophysics supplies the Galactic baryonic and satellite environment.

A stream is a distribution, not an orbit

For a regular orbit in a time-independent integrable approximation, an action $J_i=(2\pi)^{-1}\oint p_i\,dq_i$ labels the size of each oscillation and stays constant; its conjugate angle advances with frequency $\Omega_i=\partial H/\partial J_i$ . Escaping stars have slightly different actions from their progenitor, so they gradually get ahead or fall behind it. Linearizing their frequency differences gives

\Delta\boldsymbol\theta(t) \simeq \Delta\boldsymbol\theta_0 +\frac{\partial\boldsymbol\Omega}{\partial\mathbf J}\, \Delta\mathbf J\,(t-t_{\rm strip}),

where $t_{\rm strip}$ is a star’s escape time, $\Delta\mathbf J$ its action offset, and $\Delta\boldsymbol\theta_0$ its initial angle offset. The matrix $D_{ij}=\partial\Omega_i/\partial J_j=\partial^2H/(\partial J_i\partial J_j)$ controls the subsequent spreading. If one eigenvalue dominates and the escaping action distribution populates that direction, the stream mainly lengthens along its eigenvector. An individual orbit instead advances along $\boldsymbol\Omega$ . These directions need not agree, so even noiseless orbit-fitting can bias the potential Sanders and Binney 2013, §§2–4. In a strongly time-dependent or chaotic region, constant actions and this linear expansion are no longer reliable; particle evolution in the time-dependent potential is needed.

For scale, a frequency offset $\Delta\Omega=0.5\,\mathrm{rad\,Gyr^{-1}}$ accumulates an angle separation of one radian after $2\,\mathrm{Gyr}$ , even if the stars escaped at almost the same position. Conversely, a harmonic potential whose frequencies do not depend on actions has $D_{ij}=0$ and no secular spreading at this order. This limit checks the mechanism and shows why stream length cannot be interpreted as age without a force model.

A generative model chooses a Galactic potential or complete modified-gravity force law, a progenitor phase-space point and internal distribution, a mass-loss history, and perturbing structures. It then releases particles or evolves an N-body system and transforms the result to sky coordinates. The observable vector may include stream longitude and latitude, distance modulus, two proper-motion components, line-of-sight velocity, surface density, width, and stellar-population labels. A force-law change must predict these components coherently; moving a spline track through the sky is not enough.

The Milky Way’s bar and disk, giant molecular clouds, dark subhalos, and massive satellites can deflect or heat streams. In particular, a moving external perturber can make a static axisymmetric potential appear wrong. The Large Magellanic Cloud’s influence on stream tracks is quantified in Erkal et al. 2019, §§2–5. Those effects are foreground physics for a gravity test, not optional residuals.

Data release, calibration, and likelihood

A reproducible stream analysis pins the stellar candidate catalog, photometric survey and calibration, Gaia DR3 astrometry and documentation release 1.3, spectroscopy release, extinction map, isochrone and stellar-population filter, Solar position and velocity, coordinate transformation, and retrieval date European Space Agency and Gaia/DPAC 2023, official DR3 documentation. If Gaia’s Focused Product Release contributes epoch or special-source information, it is named separately. Gaia DR4 is scheduled for December 2026 and therefore is not substituted for released data at this page’s cutoff European Space Agency 2026, Gaia mission release status. Astrometric covariance, parallax and proper-motion systematics, radial-velocity zero points, photometric zero points, and extinction uncertainty are calibration parameters.

Selection includes sky footprint and masks, magnitude and color cuts, proper-motion and parallax windows, spectroscopic targeting, radial-velocity quality, stellar-population membership, and the discovery filter used to identify the stream. Because membership is inferred from some of the same coordinates being fitted, hard preselection can artificially narrow a track. A mixture likelihood instead retains contamination:

p(d_i\mid\Theta)=q_i\,p_{\rm stream}(d_i\mid\Phi,\Theta_{\rm prog},\Theta_{\rm pert}) +(1-q_i)\,p_{\rm field}(d_i),

normalized over the survey selection. The mixing weight $q_i$ comes from the population or independent conditioning information; it is not a posterior membership probability computed from $d_i$ and then reused as a prior for the same datum. The stream term can be evaluated with a particle distribution, a density model around a simulated track, or a validated emulator. It must preserve correlations among position, distance, velocity, width, and density. Priors on progenitor mass and orbit, stripping history, Milky Way baryons, halo or modified-gravity parameters, bar, and satellites must be explicit.

Results and limitations through 2026-08-19

Position or evidence	Result or claim	Limitation or counterevidence
Multiple streams constrain a nonspherical, radially varying Galactic potential under GR plus dark matter	Long tracks provide complementary force measurements beyond the rotation curve	Progenitor history, baryonic disk and bar, LMC response, and halo time dependence are correlated with shape
Stream gaps, spurs, and heating can reveal localized perturbations	Some features carry information about compact substructure or encounters	Molecular clouds, bar resonances, epicyclic overdensities, selection, and unknown encounters provide alternatives
Modified dynamics predicts different global tracks or precession from the same baryons	Streams can test field geometry and external-field prescriptions	A complete nonspherical, time-dependent prediction and consistent satellite dynamics are required; simple orbit fits are inadequate

The conservative position uses general relativity’s weak-field limit with a dark halo and explicit baryonic and satellite perturbations. The competing position asks whether a complete modified force law predicts all stream observables with the same luminous sources. Current stream reconstructions constrain Galactic potentials but do not establish a unique gravity law. Localized anomalies and global halo shape remain model-dependent because perturbers and time dependence can mimic force-law changes.

Independent and adversarial checks

An independent check evolves an N-body mock stream in a known time-dependent potential, applies the real survey selection and errors, and recovers the input force field with calibrated interval coverage. Withholding segments tests prediction: a successful model fitted to one arm should predict the other arm’s track, width, density, and velocities. Include a deliberately misaligned mock on which an orbit-only fit is demonstrably biased; an arbitrarily chosen stream need not produce a detectable bias. This control tests the specific missing physics rather than demanding failure by construction.

An adversarial analysis varies progenitor structure and disruption time, bar pattern, disk mass, LMC mass and trajectory, subhalo population, Solar phase space, membership, and extinction. It fits multiple streams simultaneously and asks whether one gravity map explains them without stream-by-stream potential changes. Posterior-predictive checks must cover track, velocity, width, and density; a gain in one observable accompanied by failure in another is not evidence for gravity. The rerun package includes source IDs, queries, selection, covariance, simulation or emulator version, resolution, seeds, and withheld tests.

What could change the conclusion. The interpretation could change after a superseding Gaia or spectroscopy release materially expands six-dimensional coverage, an independently validated time-dependent stream code passes blinded N-body challenges, a new perturbing satellite model removes or creates a coherent multi-stream residual, or one complete gravity law predicts held-out segments across several streams under a common baryonic model. A visually improved sky-track fit alone is insufficient.

Exercise

Why can fitting a single orbit through the centerline of a cold stream bias the inferred Galactic potential even if the centerline is measured without noise?

Solution

Stars escape with small but nonzero offsets in actions, so their angle differences grow along the direction set by the frequency-gradient matrix. That direction is generally not identical to the progenitor’s orbital-frequency vector. The stream centerline and one orbit are therefore misaligned for dynamical reasons, not measurement noise. Forcing an orbit to follow the centerline changes the potential because the fit must compensate for this source effect, which produces a biased force reconstruction.

References

Erkal, D., et al. (2019), “The total mass of the Large Magellanic Cloud from its perturbation on the Orphan stream,” Monthly Notices of the Royal Astronomical Society 487, 2685–2700. DOI.
European Space Agency (2026), “Gaia,” mission status and data-release schedule (accessed 2026-08-19). Official mission page.
European Space Agency and Gaia Data Processing and Analysis Consortium (2023), Gaia Data Release 3 Documentation, release 1.3, 10 July 2023. Official documentation.
Sanders, J. L., and Binney, J. (2013), “Stream–orbit misalignment I: The dangers of orbit-fitting,” Monthly Notices of the Royal Astronomical Society 433, 1813–1825. DOI.