The Splashback Radius: A Galaxy Cluster's True Edge

A boundary set by orbits, not by a chosen density

For decades, the "size" of a dark-matter halo has been a matter of convention. Cosmologists quote the radius R_Δ inside which the mean density equals some multiple Δ (200, 500, or the virial value) of a reference density. These are convenient but physically arbitrary — R₂₀₀ₘ is not where anything special happens dynamically; it is just where a bookkeeping threshold is crossed.

The splashback radius offers a genuinely physical alternative. Consider a shell of matter that fell into a cluster, passed through the center, and swung back out. It reaches a maximum distance — its apocenter — before falling in again. Because recently accreted shells all pile up near their common apocenter, the cluster boundary is a real, sharp feature: a caustic where the density profile steepens abruptly. Inside it lies multi-stream, orbiting ("virialized") material; outside lies the single-stream cosmic infall. This is the cluster's true edge.

The mechanism: secondary infall and the first-orbit caustic

The physics traces to the secondary infall or self-similar collapse picture developed by Gunn & Gott (1972), Fillmore & Goldreich (1984), and Bertschinger (1985). Around an overdense seed, successive spherical shells decelerate, halt at turnaround, and collapse inward. A given shell falls in, reaches pericenter near the center, and rebounds to its apocenter — this outermost point of the first orbit is the splashback location.

Two effects sharpen it into an edge. First, orbits spend most of their time near apocenter (velocity → 0 there), so matter accumulates in a thin shell. Second, because the halo grows over time, the enclosed mass a shell falls through keeps increasing, so more recently accreted shells reach systematically smaller apocenters. The most recently splashed-back shell defines the outermost caustic. The result: a density profile that transitions from a gentle NFW-like slope inside to a steep drop (logarithmic slope reaching −4 or steeper) right at R_sp, then flattens into the infall regime beyond.

Characteristic numbers, scaling, and the key criterion

For massive clusters, R_sp ≈ 0.8-1.5 × R₂₀₀ₘ, translating to roughly 1-3 Mpc for a 10¹⁴-10¹⁵ M☉ system. The crucial insight from Diemer & Kravtsov (2014) and Adhikari, Dalal & Chamberlain (2014) is that R_sp/R₂₀₀ₘ depends primarily on the halo's mass accretion rate, Γ = d ln M / d ln a. Fast-accreting halos have deeper, faster-changing potentials, so shells reach smaller apocenters: R_sp shrinks (down to ~0.8 R₂₀₀ₘ for Γ ≈ 4-5) and the density drop is sharper. Slowly accreting halos push R_sp out toward ~1.5 R₂₀₀ₘ.

The Diemer-Kravtsov fitting formula parameterizes the whole outer profile: an Einasto-like inner term plus a steepening transition term plus a mean-density infall term, ρ(r) = ρ_inner·f_trans + ρ_outer. There is also a secondary dependence on the peak height ν (how rare the halo is). The splashback radius is defined as the location of the steepest logarithmic slope, d ln ρ / d ln r.

Detecting the edge: galaxy counts and weak lensing

The splashback feature was first detected observationally by More, Miyatake et al. (2016) using ~8,000 clusters from the SDSS redMaPPer catalog. Stacking the projected number density of surrounding galaxies revealed the tell-tale sharp steepening in the galaxy profile at ~1-2 Mpc. Because galaxies trace the collisionless matter, they "splash back" too.

The gold standard is weak gravitational lensing, which maps the total (mostly dark) matter directly via subtle distortions of background-galaxy shapes. Surveys like the Dark Energy Survey (DES), the Hyper Suprime-Cam (HSC) survey on Subaru, and DECaLS have measured the lensing ΔΣ profile out past R_sp and recovered the steepening. A recurring puzzle: early optically-selected samples yielded R_sp measurements ~10-20% smaller than ΛCDM simulations predict — later traced largely to selection effects in cluster-finding (projection, member-galaxy dilution) rather than new physics. X-ray and Sunyaev-Zel'dovich (SZ) selected samples help mitigate this.

Where it operates, and how it differs from related radii

Splashback is universal to gravitationally collapsed, collisionless halos — it appears in galaxy clusters, groups, and even galaxy- and subhalo-scale halos in simulations — but clusters are where it is cleanest and observable, because they are large, recently assembled, and abundant enough to stack.

It should not be confused with neighbors on the density profile. The virial radius and R₂₀₀ₘ are fixed-threshold conventions that always lie interior to (or near) R_sp. The turnaround radius (~4-8 Mpc) is much larger — the shell currently at zero radial velocity, where cosmic infall exactly cancels Hubble expansion. The splashback radius sits between the virialized interior and the turnaround shell. It also marks a physical transition for gas and galaxies: infalling galaxies often begin ram-pressure stripping and quenching near R_sp, and it roughly coincides with an accretion shock in the intracluster medium, though the gas (collisional) edge and the collisionless splashback edge need not coincide exactly.

Open questions and why it matters

The splashback radius matters because it promises a non-arbitrary, self-consistent halo boundary — one that cleanly separates the physical halo from the cosmic web, sharpens definitions of cluster mass and satellite populations, and offers a novel cosmological probe: because R_sp/R₂₀₀ₘ encodes the accretion rate, it is sensitive to the growth of structure and thus to dark energy and modified gravity (which predict altered splashback signatures).

Open questions remain. How large are selection and projection biases in real cluster samples, and can we reconcile the modest observation-simulation tension entirely with them? How does baryonic physics — gas cooling, feedback, dynamical friction — shift the caustic? Is the sharp spherical picture washed out by triaxiality and mergers into a multi-caustic "splashback shell"? Upcoming deep surveys — the Vera C. Rubin Observatory's LSST, Euclid, and SPT-3G/Simons Observatory SZ catalogs — should measure R_sp for tens of thousands of clusters, turning this dynamical edge into a precision tool.

Splashback radius versus conventional cluster boundary definitions for a ~10¹⁴-10¹⁵ M☉ cluster

Boundary definition	Physical basis	Typical radius	Depends on accretion rate?
R₅₀₀ᵥ	Mean density = 500 × ρ_crit	~1.2 Mpc	No (fixed threshold)
R₂₀₀ᵥ (virial)	Mean density = 200 × ρ_crit	~1.9 Mpc	No (fixed threshold)
R₂₀₀ₘ	Mean density = 200 × ρ_mean	~2.4 Mpc	No (fixed threshold)
R_splashback	Apocenter of first-orbit matter	~1.8-3.5 Mpc (~0.8-1.5 R₂₀₀ₘ)	Yes — shrinks with faster accretion
Turnaround radius	Zero net radial velocity (Hubble = infall)	~4-8 Mpc	Yes (cosmology-dependent)