Counterrotating Stellar Disks (CRDs)

• Counterrotating stellar disks (CRDs) are galaxies that host two extensive stellar disks rotating in opposite directions, serving as tracers of external gas accretion and merger events.
• Observational diagnostics rely on spectral decomposition and kinematic mapping, such as 2σ signatures and velocity reversals, to clearly distinguish CRDs from other multi-spin systems.
• Detailed studies reveal that the secondary disk is typically younger and chemically distinct from the primary, providing insights into formation channels like retrograde gas accretion and minor mergers.

Searching arXiv for recent and foundational papers on counterrotating stellar disks to ground the article in the literature. Counter-rotating stellar disks (CRDs) are galaxies that host two extended, co-spatial stellar disks rotating in opposite directions around the same axis. Within the broader class of multi-spin galaxies, they are the subset in which the counter-rotating structure is specifically stellar, large-scale, and comparable in radial extent to the main disk rather than a compact core, ring, or orthogonally rotating component (Coccato et al., 2014). Because opposite-sign angular momentum cannot be generated trivially within an already settled stellar disk, CRDs are widely treated as sensitive tracers of external accretion, merger-driven disk assembly, and angular-momentum reorientation in galaxies (Bao et al., 4 Mar 2026).

1. Definition, taxonomy, and scope

In the observational literature summarized by Coccato and collaborators, “counter-rotating galaxies” are systems that host two components rotating in opposite directions, but the components can be stars versus stars, stars versus gas, or gas versus gas, and the scale can range from compact cores to large-scale disks (Coccato et al., 2014). CRDs, in the strict sense used here, are the peculiar case of galaxies hosting two large-scale counter-rotating stellar disks of comparable radial extent. They are therefore distinguished from kinematically decoupled cores, warped inner structures, and polar or orthogonally rotating subsystems.

The prototype is NGC 4550, an E7/S0 galaxy in which Rubin et al. identified two stellar disks rotating in opposite directions (Coccato et al., 2014). Subsequent resolved work extended the class to systems such as NGC 3593, NGC 5197, NGC 5719, IC 719, NGC 4138, and more recently a large MaNGA-based population (Coccato et al., 2014). In nearby-survey terminology, many CRDs also appear as “2σ galaxies,” meaning that their stellar velocity-dispersion maps show two symmetric off-center peaks along the major axis, a consequence of two dynamically cold components with opposite line-of-sight streaming velocities overlapping in projection (Rubino et al., 2021).

The class is not restricted to a single morphology. Counter-rotating galaxies occur in all morphological types, but the incidence of large-scale stellar counter-rotation is higher in early-type systems than in spirals in current survey data (Coccato et al., 2014). In MaNGA DR17, confirmed CRDs occur in ellipticals, lenticulars, and spirals, with raw occurrence rates of $48/2529 \approx 1.9\%$, $36/1598 \approx 2.3\%$, and $42/6156 \approx 0.7\%$, respectively (Piper et al., 2 Mar 2026).

2. Observational diagnostics and spectral decomposition

The central observational difficulty is that the two stellar disks are co-spatial: in each spatial element their absorption-line spectra overlap. In resolved spectroscopy, the integrated stellar line profiles can therefore become double-peaked or strongly asymmetric, especially where the projected rotation amplitudes are high (Coccato et al., 2014). This motivates two complementary approaches. The first is indirect detection through moment maps derived from single-component fitting, especially the 2σ signature and reversals in the stellar velocity field. The second is explicit spectral decomposition into multiple stellar components.

For spatially resolved decomposition, Coccato et al. model each observed spectrum as the sum of two stellar components plus ionized-gas emission. In schematic form,

$$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$

where $T_1$ and $T_2$ are synthetic stellar templates, $w_1$ and $w_2$ are luminosity weights, $\mathcal{G}(V,\sigma)$ is a Gaussian LOSVD, and the gas is represented by Gaussian emission lines including H$\gamma$, H$36/1598 \approx 2.3\%$0, [O III], and N I. The fit recovers, for each component and each spatial bin, the velocity $36/1598 \approx 2.3\%$1, velocity dispersion $36/1598 \approx 2.3\%$2, and relative luminosity; Lick indices measured on the best-fit templates then provide age, metallicity $36/1598 \approx 2.3\%$3, $36/1598 \approx 2.3\%$4-enhancement, mass-to-light ratio, and stellar mass surface density (Coccato et al., 2014).

Related decompositions have been implemented with pPXF in MaNGA-based case studies. In SDSS J074834.64+444117.8, three distinct methods—Ca II $36/1598 \approx 2.3\%$5 double-Gaussian fitting, two-component pPXF spectral decomposition, and an orbit-based Schwarzschild model using non-parametric LOSVDs—gave consistent stellar kinematics (Bao et al., 2024). In that work, the circularity parameter

$36/1598 \approx 2.3\%$6

was used in the dynamical model to separate co-rotating, non-rotating, and counter-rotating orbit families (Bao et al., 2024).

Large statistical surveys often rely on indirect kinematic signatures because explicit two-component decomposition is not always possible at survey resolution. In MaNGA, two main selection channels have been used: the 2σ morphology of the $36/1598 \approx 2.3\%$7 map and direct stellar-velocity reversals in $36/1598 \approx 2.3\%$8 along the photometric major axis (Bao et al., 2022). Automated DR17 pre-selection further formalized this strategy by extracting one-dimensional profiles along the stellar kinematic major axis and searching for multiple stellar-velocity extrema or two off-center $36/1598 \approx 2.3\%$9 peaks before visual confirmation (Piper et al., 2 Mar 2026).

Detectability is strongly inclination- and S/N-dependent. Mock MUSE data show that the strongest signature of two large-scale counter-rotating stellar disks is the symmetric double peak in the velocity-dispersion map, but the size, shape, and slope of the 2σ feature depend strongly on velocity separation and relative light contribution (Rubino et al., 2021). When the 2σ feature is weak, the large-scale structure in the $42/6156 \approx 0.7\%$0 map can still indicate strong or weak counter-rotation (Rubino et al., 2021). This implies that observed CRD frequencies are lower limits rather than complete incidence measurements.

3. Kinematics, stellar populations, and internal structure

Resolved case studies consistently show that the two disks differ not only kinematically but also in stellar population properties. In NGC 3593, the two stellar disks rotate in opposite directions, while the ionized gas rotates in the same direction as the secondary stellar component rather than the main disk (Coccato et al., 2014). The secondary disk is less luminous and less massive than the main stellar disk at almost all radii covered, yet it is a large-scale disk rather than a compact core (Coccato et al., 2014).

The stellar-population contrast in NGC 3593 is especially well constrained. In the central $42/6156 \approx 0.7\%$1 pc, the main disk has age $42/6156 \approx 0.7\%$2 Gyr and the secondary disk $42/6156 \approx 0.7\%$3 Gyr; the paper states that this dates the formation of the counter-rotating stellar disk to $42/6156 \approx 0.7\%$4 Gyr ago, i.e. $42/6156 \approx 0.7\%$5 Gyr after the formation of the main galaxy disk (Coccato et al., 2014). The main component has $42/6156 \approx 0.7\%$6, while the secondary has $42/6156 \approx 0.7\%$7; both show high central metallicity within $42/6156 \approx 0.7\%$8 pc and declining metallicity with radius, and the secondary disk is more $42/6156 \approx 0.7\%$9-enhanced (Coccato et al., 2014). In the broader 3D-spectroscopy sample, the counter-rotating disk is reported as the less massive, the youngest, and chemically distinct from the main disk in all studied cases (Coccato et al., 2014).

SDSS J074834.64+444117.8 shows the same general pattern with different quantitative contrast. The primary disk is more luminous across the whole galaxy and counter-rotates with the ionized gas, while the secondary disk co-rotates with the gas (Bao et al., 2024). The secondary disk has a younger stellar population and poorer stellar metallicity than the primary disk, and the stellar mass ratio between primary and secondary disks is $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$0 (Bao et al., 2024). The average flux ratio is $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$1, and the stellar velocity dispersions from pPXF are $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$2 and $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$3 (Bao et al., 2024).

PGC 66551 provides a chemically unusual variant. Its counter-rotating disk is compact, contains young stars, has stellar mass $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$4 amounting to $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$5 of the galaxy’s total, and is associated with gas rotating in the same direction (Katkov et al., 2023). The main disk is about 8 Gyr old but has a significantly lower metallicity of $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$6 dex than other CR galaxies, while the counter-rotating disk is younger and more metal rich than the main disk (Katkov et al., 2023). The authors use this configuration to constrain the metallicity of the infalling gas that formed the CR disk to $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$7 to $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$8 dex (Katkov et al., 2023).

A recurring observational result is therefore that one disk—often the one aligned with the gas—is younger than the other. This does not imply a universal age ordering across all environments and mass scales, however. In TNG50 Milky Way-mass late-type centrals, most CRDs are older than the co-rotating disk, with $$S_{\mathrm{obs}}(\lambda) \approx w_1 \left[T_1(\lambda)\otimes \mathcal{G}(V_1,\sigma_1)\right] + w_2 \left[T_2(\lambda)\otimes \mathcal{G}(V_2,\sigma_2)\right] + \sum_k G_k(\lambda),$$9 of cases having CRD median stellar age older than that of the co-rotating disk (Bugueño et al., 3 Oct 2025). This is a genuine tension between some observations and some cosmological simulations rather than a contradiction within a single method.

4. Population statistics and host-galaxy properties

MaNGA transformed CRDs from a small set of case studies into a survey-defined population. A visual search of about 4000 MaNGA galaxies yielded 64 CRDs, 61 of them early-type galaxies, with 17 systems where the two stellar components could also be separated spectroscopically (Bevacqua et al., 2021). That work estimated the frequency of CRDs in MaNGA at $T_1$0 for ellipticals, $T_1$1 for lenticulars, and $T_1$2 for spirals at 95% confidence (Bevacqua et al., 2021). A broader MaNGA study based on 101 galaxies with counter-rotating stellar disks and regularly rotating ionized gas classified them into four kinematic types according to whether the gas co-rotates with the inner or outer stellar disk and whether the counter-rotation is directly visible in the stellar velocity field or only inferred from the 2σ signature (Bao et al., 2022).

The largest statistical MaNGA sample to date contains 147 CRDs, accounting for $T_1$3 of the MaNGA survey, with 138 galaxies having reliable global stellar mass and SFR measurements (Bao et al., 4 Mar 2026). Relative to mass- and SFR-matched controls, these CRDs are more bulge-dominated, more gas-poor, and typically more isolated. Median bulge-to-total ratios increase from $T_1$4 in Type CO+IN to $T_1$5 in Type NO-EML, while matched control galaxies have median $T_1$6 in each type (Bao et al., 4 Mar 2026). Molecular gas mass fractions are lower than controls by about 0.5 dex in Type CO+IN, 1.1 dex in Type CT+OUT, and 0.6 dex in Type MIS (Bao et al., 4 Mar 2026). CRDs also tend to inhabit environments with lower tidal strength $T_1$7 and fewer neighbors than controls, with Type MIS showing the lowest tidal strengths, $T_1$8 (Bao et al., 4 Mar 2026).

The MaNGA classification has been extended in two directions. First, the six-type scheme in the 147-galaxy sample distinguishes systems where the gas is aligned with one stellar disk, counter-aligned with the dominant disk, misaligned with both stellar disks, or absent in emission lines (Bao et al., 4 Mar 2026). Second, an automated DR17 pre-selection followed by visual inspection produced 126 confirmed CRDs and 143 candidate galaxies, more than doubling the earlier MaNGA sample (Piper et al., 2 Mar 2026). Among the 126 confirmed DR17 CRDs, 49 are star-forming, 22 AGN-dominated, 10 ambiguous, and 45 have no emission-line data at the adopted thresholds; however, a matched-control comparison showed no statistically significant difference in BPT photoionization classes between CRDs and early-type controls (Piper et al., 2 Mar 2026). Emission-line diagnostics therefore do not by themselves identify counter-rotating disks.

The TNG50 analysis places these statistics in a cosmological context. Selecting 260 central late-type galaxies with $T_1$9, $T_2$0, and $T_2$1, the study finds 26 galaxies with significant CRDs, defined by $T_2$2, corresponding to about 10% of Milky Way-mass late-type centrals in that simulation (Bugueño et al., 3 Oct 2025). The authors explicitly note that most simulated CRDs are compact and low-mass, likely hard to detect observationally, so the simulated abundance can still be described as consistent with the rarity inferred from surveys once detectability is considered (Bugueño et al., 3 Oct 2025).

5. Formation channels and evolutionary pathways

The dominant observational interpretation is that CRDs form primarily through external gas acquisition followed by in-situ star formation. Coccato et al. argue that internal processes such as bar dissolution are ruled out in NGC 3593 because they would produce two counter-rotating disks with the same properties, whereas the observed disks differ clearly in age, metallicity, and $T_2$3-enhancement (Coccato et al., 2014). The natural sequence is that the main stellar disk forms first, the galaxy later accretes gas with opposite angular momentum, that gas settles into a counter-rotating disk, and new stars form from it (Coccato et al., 2014). Alignment of the gas with the younger stellar component is the key empirical prediction, and it is widely observed in case studies (Coccato et al., 2014).

The 101-galaxy MaNGA classification reframed this in terms of two controlling parameters: the abundance of pre-existing gas in the progenitor and the efficiency of angular-momentum consumption during gas–gas interaction (Bao et al., 2022). In gas-rich progenitors, efficient angular-momentum loss drives inflow and central star formation, favoring types where the younger, gas-aligned disk dominates in the inner galaxy. In gas-poor progenitors, the accreted retrograde gas retains more angular momentum and can build a younger outer disk or a fainter gas-aligned component embedded in an old dominant stellar disk (Bao et al., 2022). The 147-galaxy MaNGA study generalizes this further by identifying six types and concluding that the impact of gas accretion on galaxy evolution primarily depends on the abundance of pre-existing gas in progenitors (Bao et al., 4 Mar 2026).

Multiple accretion events are also directly implicated. In the Type MIS MaNGA class, the gas disk is misaligned with both stellar disks, implying at least two distinct gas-accretion episodes with different angular-momentum vectors (Bao et al., 4 Mar 2026). Type NO-EML systems, in which more than 75% of spaxels within $T_2$4 have H$T_2$5 S/N $T_2$6, are interpreted as late-stage CRDs in which the counter-rotating stellar disks persist but the gas has been exhausted or removed (Bao et al., 4 Mar 2026).

Direct evidence for the retrograde gas-accretion pathway comes from atomic hydrogen. In IC 0719, a 43 kpc warped HI disk with a faint tail extending toward the neighboring IC 0718 indicates accretion of misaligned retrograde gas during an encounter (Young et al., 2020). In the inner region, where dynamical timescales are shorter, the HI has settled into the equatorial plane and formed the retrograde secondary stellar disk (Young et al., 2020). The authors describe this as the first direct evidence that a double-disc stellar counterrotator could be formed through the accretion of retrograde gas (Young et al., 2020).

Mergers remain viable in some systems. The 2014 VIMOS-based review states explicitly that external acquisition of gas followed by star formation, or mergers of two galaxies with similar mass, are both viable mechanisms (Coccato et al., 2014). The 2026 MaNGA survey paper likewise notes that some CRD types, especially gas-free or morphologically bulge-dominated systems, may represent more evolved outcomes in which the original gas-accretion event is no longer directly visible (Bao et al., 4 Mar 2026). PGC 66551 is interpreted as favoring a merger with a gas-rich satellite over cold accretion from a cosmic filament, based on the inferred metallicity of the infalling gas, $T_2$7 to $T_2$8 dex (Katkov et al., 2023).

A plausible synthesis is that “gas accretion” and “minor merger” are not mutually exclusive channels. Several papers treat a gas-rich satellite as one of the physically relevant sources of retrograde gas, alongside the intergalactic medium or cosmic web (Coccato et al., 2014). The decisive empirical discriminator is therefore not whether the process is labeled merger or accretion, but whether the younger stellar disk formed in situ from externally supplied retrograde gas.

6. Dynamics, instabilities, and nuclear analogues

Beyond formation, counter-rotation changes disk dynamics. Collisionless $T_2$9-body simulations of two-component stellar disks embedded in fixed halos show that counterrotation can excite strong bending waves and vertical heating even when the initial vertical-to-radial dispersion ratio is $w_1$0, i.e. above the classical fire-hose threshold (Khoperskov et al., 2016). During evolution, $w_1$1 rises toward values close to unity in locally Jeans-stable disks and can exceed that in initially Jeans-unstable disks, especially in the inner regions (Khoperskov et al., 2016). The heating correlates with the strength of the counter-rotating component through the dimensionless kinetic-energy ratio

$w_1$2

and the mean final anisotropy follows an empirical relation

$w_1$3

This implies that even apparently minor retrograde components can vertically heat a disk (Khoperskov et al., 2016).

At interfaces between rotating and counter-rotating collisionless flows, two-stream instability can generate high-$w_1$4 spirals and clumpy rings. In a GADGET-2 simulation of an inner stellar disk plus an outer annular counter-rotating stellar flow, the shear layer excites a multi-armed spiral density wave in the main disk and a high-$w_1$5 density distribution in the counter-rotating flow, with dominant mode $w_1$6–10 in the ring region (Hohlfeld et al., 2014). The result is substantial radial scattering of retrograde material and morphological restructuring of the interface (Hohlfeld et al., 2014).

Nuclear CRDs around massive black holes form a distinct but related dynamical regime. In nearly Keplerian razor-thin disks, Sridhar and Touma’s secular formulation allows two counter-rotating stellar populations to be reduced to a two-degree-of-freedom Hamiltonian system for their centroid eccentricities (Touma et al., 2011). For fixed semi-major axis $w_1$7, the circular state is unstable when

$w_1$8

where $w_1$9 and $w_2$0 are coefficients of the orbit-averaged ring–ring interaction (Touma et al., 2011). The nonlinear dynamics then admits steadily precessing anti-aligned, aligned, and non-aligned eccentric configurations (Touma et al., 2011). This is not the same phenomenon as large-scale galactic CRDs, but it shows that counter-rotation generically introduces a secular pathway to lopsided eccentric structure.

The Galactic Center model of Alig et al. offers a gas-dynamical nuclear analogue. In that simulation, collision of a single molecular cloud with a circum-nuclear gas disk around Sgr A* produces two inclined, counter-rotating sub-parsec accretion disks, with the first forming roughly 1 Myr earlier than the second (Alig et al., 2013). The mechanism is angular-momentum mixing and cancellation between cloud and disk gas rather than stellar dynamical instability, but the result is the same: sequential formation of two stellar systems with opposite spin (Alig et al., 2013).

Theoretical work therefore converges on a broad principle: once a galaxy contains substantial retrograde material, both its subsequent star formation and its dynamical evolution differ qualitatively from those of a single-spin disk. This suggests that CRDs are not merely kinematic curiosities but long-lived records of externally driven angular-momentum restructuring.