Wet Extreme Mass-Ratio Inspirals

• Wet extreme mass-ratio inspirals are systems where a stellar-mass black hole inspirals into a supermassive black hole within a gas-rich environment, such as an AGN disk.
• Disk-induced torques, migration, and damping processes shape their nearly circular orbits, leading to distinctive waveform features observable by detectors like LISA.
• These inspirals serve as multi-messenger probes that connect strong-field gravity with galactic-nuclear astrophysics, informing both disk dynamics and AGN duty cycles.

Searching arXiv for papers on wet extreme mass-ratio inspirals and related environmental EMRI literature. Wet extreme mass-ratio inspirals are extreme mass-ratio inspirals that form and evolve in non-vacuum, gas-rich or otherwise environmentally structured settings rather than as isolated vacuum binaries. In current usage, the term most commonly refers to stellar-mass black holes inspiraling into supermassive black holes within active galactic nucleus disks, where capture, inclination damping, eccentricity damping, migration, gas accretion, and compact-object interactions can all affect both source formation and waveform morphology (Pan et al., 2021, Lyu et al., 2024). A broader environmental usage also appears in discussions of EMRIs perturbed by nearby supermassive black holes or dense dark-matter distributions, where the gravitational-wave signal acquires detectable non-vacuum structure [(Yunes et al., 2010); (Wade et al., 28 Aug 2025)]. Across these contexts, wet EMRIs are relevant because long-lived mHz-band signals are exceptionally sensitive to small environmental perturbations, and because such systems couple strong-field gravity to galactic-nuclear astrophysics.

1. Definition and scope

The dry–wet distinction is primarily a formation-channel distinction. Dry EMRIs form in relatively gas-poor nuclear star clusters through multi-body gravitational scattering, loss-cone capture, and subsequent gravitational-radiation-driven inspiral (Pan et al., 2021, Sun et al., 30 Aug 2025). Wet EMRIs form in gas-rich environments, specifically around actively accreting massive black holes surrounded by AGN accretion disks, where stellar-mass black holes are captured by the disk and then migrate inward until gravitational radiation takes over near the central black hole (Pan et al., 2021, Lyu et al., 2024).

This distinction is not merely terminological. In wet systems, the environment actively enters the dynamics through disk-assisted capture, disk-driven radial migration, inclination damping, eccentricity damping, and head wind or accretion drag effects (Pan et al., 2021). The result is that wet EMRIs are generally expected to be much more circular by the time they enter the LISA band than dry EMRIs, although later work emphasizes that they are not necessarily circular because turbulence and multi-body resonances or scattering can re-excite eccentricity (Pan et al., 2021, Sun et al., 30 Aug 2025).

The term is also used more broadly for environmentally contaminated EMRIs whose waveforms encode the surrounding galactic-nuclear environment. In this broader sense, an EMRI in the gravitational field of a nearby secondary supermassive black hole is a “wet” or non-vacuum inspiral because the waveform departs from that of an isolated two-body system (Yunes et al., 2010). Likewise, inspirals embedded in dense dark-matter spikes are environmentally affected through dynamical friction and secondary accretion, producing measurable dephasing relative to vacuum binaries (Wade et al., 28 Aug 2025). This suggests that “wet EMRI” can denote both a specific AGN-disk formation channel and a wider class of non-vacuum EMRIs, with the AGN-disk usage now the dominant one in formation studies.

2. AGN-disk formation channel

The canonical wet channel occurs in active galactic nuclei, where a rapidly accreting massive black hole is surrounded by a gas disk (Pan et al., 2021). In this picture, a nuclear star cluster contains stellar-mass black holes on inclined orbits; when an AGN disk is present, some intersect or enter the disk, are captured into it, become efficiently aligned and circularized, and then migrate inward. If the inward migration time is shorter than the disk lifetime, the object reaches the central black hole and becomes a wet EMRI (Pan et al., 2021).

A central quantitative premise is that AGN activity is not universal. One study adopts a conservative universal active fraction

$f_{\rm AGN}=1\%$

while noting that it can be up to an order of magnitude higher (Pan et al., 2021). Another uses a conservative universal AGN fraction of $1\%$ in population estimates (Lyu et al., 2024). These assumptions are important because the total observable wet-EMRI population is the per-AGN production rate multiplied by the AGN fraction.

The disk–compact-object interaction is organized around density-wave torques and drag. In the type-I regime, the migration torque is written schematically as

$$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$

with associated timescales

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$

so that

$t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$

Since $h\ll 1$, inclination and eccentricity damping are much faster than radial migration (Pan et al., 2021). This is the principal reason the canonical wet channel predicts nearly circular sources in band.

For embedded objects, a head wind from relative motion with the gas contributes an additional torque,

$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$

and the effective migration timescale is summarized as

$$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$

If the stellar-mass black hole opens a gap, type-I migration turns off and type-II migration applies instead (Pan et al., 2021). The broad conclusion remains that disk torques efficiently drive inward drift.

A distinct formulation appears in a later recoil-regulated model, which treats migration and damping in a self-gravitating Sirko–Goodman thin disk implemented with the pAGN package. There the minimal migration prescription is

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$

with $f_{\rm mig}=2$, and for disk crossings the effective drift is reduced by the time spent in gas,

$1\%$0

Gravitational radiation contributes at small radii through

$1\%$1

(Xue et al., 28 May 2026).

3. Orbital and waveform characteristics

Wet EMRIs are commonly distinguished from dry EMRIs by their eccentricity, inclination, and component-mass distributions (Sun et al., 30 Aug 2025). The clearest expectation is eccentricity. Dry EMRIs begin highly eccentric and remain significantly eccentric in band, with a broad distribution $1\%$2 to $1\%$3 and more than $1\%$4 of the population in this range at a reference pericenter $1\%$5 (Sun et al., 30 Aug 2025). Wet EMRIs are generally less eccentric because disk damping is efficient, but can still reach

$1\%$6

with some cases reaching $1\%$7 if close encounters or scattering occur (Sun et al., 30 Aug 2025).

The low-eccentricity expectation follows from the damping timescales

$1\%$8

which imply faster eccentricity damping in thin disks (Sun et al., 30 Aug 2025). However, this baseline can be altered by multi-body resonance effects and stochastic turbulence. Three-body AGN-disk simulations show trapping in mean-motion resonances such as $1\%$9, $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$0, or $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$1, with resonance-maintained systems typically yielding $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$2–$$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$3 when the inner object reaches $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$4, and in some cases strong scattering at very small separations can excite $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$5 (Sun et al., 30 Aug 2025).

Turbulence provides a second eccentricity-pumping channel. The fluctuating potential is modeled as

$$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$6

with effective viscosity parameter

$$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$7

In this framework, long-term N-body evolution gives $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$8 for a $$\dot J_{\rm mig,I} = C_{\rm I}\, \frac{m_{\rm bh}}{M}\,\frac{\Sigma}{M}\,\frac{r^4\Omega^2}{h^2},$$9-disk with fairly strong turbulence, whereas for an $$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$0-disk with similar global parameters the eccentricity can be as small as $$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$1 (Sun et al., 30 Aug 2025). This establishes a direct connection between measured eccentricity and AGN-disk microphysics.

Inclination is likewise channel-sensitive. In Kerr spacetime it is defined by

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$2

For wet EMRIs, inclination traces disk alignment history. Under coherent accretion, the expected distribution is sharply aligned,

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$3

whereas under chaotic accretion the initial disk-spin inclination is isotropic,

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$4

The relevant Bardeen–Petterson warp radius is

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$5

and the inclination distribution depends on whether the disk lifetime is shorter or longer than the Bardeen–Petterson timescale $$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$6 (Sun et al., 30 Aug 2025). Dry EMRIs can also show a prograde bias, but there the origin is the spin dependence of the loss cone and last stable orbit rather than disk alignment physics (Sun et al., 30 Aug 2025).

Component masses furnish a further discriminator. In the wet channel, capture tends to boost the high-mass end, accretion shifts the whole distribution to larger masses, and mergers can create new peaks near

$$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$7

corresponding to combinations such as $$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$8 and $$t_{\rm mig,I} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^2}{\Omega}, \qquad t_{\rm wav} \sim \frac{M}{m_{\rm bh}} \frac{M}{\Sigma r^2}\frac{h^4}{\Omega},$$9 (Sun et al., 30 Aug 2025). This differs from the dry-channel expectation, where mass segregation enhances the $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$0 peak (Sun et al., 30 Aug 2025).

4. Rates, populations, and recoil regulation

Early AGN-disk calculations argued that the wet channel may substantially enhance EMRI formation. One study concluded that the presence of an AGN disk boosts the EMRI intrinsic formation rate by orders of magnitude relative to the dry loss-cone channel, with representative LISA-detectable rates under one mass function of roughly

$t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$1

before applying the AGN fraction, and lower but still significant rates under an alternative mass function (Pan et al., 2021). Another study reported total wet-EMRI rates of about $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$2–$t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$3, LISA detection rates of about $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$4–$t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$5, and resolvable AGN hosts of about $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$6–$t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$7, again assuming a conservative universal AGN fraction of $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$8 (Lyu et al., 2024).

Later work substantially revised this picture by incorporating stellar interactions within AGN disks. In the recoil-regulated model, the disk is treated as a dynamical ecosystem in which binary formation, hierarchical mergers, and recoil kicks from mergers and binary–single encounters repeatedly lift stellar-mass black holes out of the disk plane, temporarily interrupt migration, and suppress EMRI formation in much of parameter space (Xue et al., 28 May 2026). The compact-object population is evolved with coupled single and binary surface-density functions $t_{\rm wav} \approx t_{\rm mig,I}\, h^2.$9 and $h\ll 1$0, including transport in radius and mass plus source and conversion terms. In simplified form,

$h\ll 1$1

$h\ll 1$2

The resulting phenomenology is strongly structured in $h\ll 1$3 space. In low-mass, low-Eddington-ratio AGNs,

$h\ll 1$4

stellar interactions can enhance wet EMRI formation because temporary ejection from the disk can make effective inward transport faster than remaining embedded in the slow type-II regime (Xue et al., 28 May 2026). In intermediate AGNs, interactions strongly suppress EMRI formation, reducing the total mass reaching the EMRI region by roughly $h\ll 1$5–$h\ll 1$6 relative to a model without interactions (Xue et al., 28 May 2026). In high-mass, high-Eddington-ratio AGNs,

$h\ll 1$7

the suppression is milder (Xue et al., 28 May 2026).

The revised LISA detection-rate prediction is therefore much smaller than in smooth-disk models: $h\ll 1$8 depending on the assumed AGN demographics (Xue et al., 28 May 2026). These results identify stellar interactions as a key ingredient in wet-EMRI population synthesis and imply that the observable population is dominated by low-mass AGNs and is highly sensitive to the poorly constrained low-mass, low-luminosity AGN population (Xue et al., 28 May 2026).

A further demographic inference concerns AGN age. In the recoil-regulated picture, detectable EMRIs are preferentially produced in young AGNs, typically within $h\ll 1$9–$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$0 Myr of disk formation, while very old AGNs contribute negligibly because repeated interactions eject stellar-mass black holes from the disk and delay migration too long for EMRI formation before the disk fades (Xue et al., 28 May 2026). This suggests that wet EMRIs can probe AGN duty cycles as well as disk structure.

5. Environmental imprints beyond the canonical AGN channel

A broader wet-EMRI literature examines environmental effects that modify EMRI waveforms even when the formation channel itself is not disk-assisted. One example is the effect of a nearby secondary supermassive black hole on an EMRI waveform. Because EMRIs accumulate millions of radians of phase over a year or more, even a small line-of-sight acceleration of the EMRI center of mass can generate an observable phase drift (Yunes et al., 2010). In the simplest model, the line-of-sight velocity induced by the secondary is

$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$1

with

$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$2

The leading-order phase drift from a uniform acceleration is estimated as

$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$3

For fiducial detectability $$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$4, the characteristic separation is

$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$5

At $$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$6–$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$7 pc, the leading acceleration can produce measurable dephasing; at $$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$8–$$\dot J_{\rm wind}^{\rm id} = - \frac{r\,\delta v_\phi\, \dot m_{\rm gas}}{m_{\rm bh}},$$9 pc, higher derivatives of the motion may also be measurable, enabling separate recovery of perturber mass and distance (Yunes et al., 2010).

In effective-one-body modeling, the orbital phase evolves as

$$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$0

and the external acceleration modifies this to

$$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$1

This is emphasized to be a wave-generation effect rather than a wave-propagation effect (Yunes et al., 2010). For representative systems with perturber masses $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$2–$$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$3 and separations $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$4 pc, the dominant $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$5 mode dephases by $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$6 radians in less than a year, while amplitude changes are only $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$7 or $$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$8, making phase the main observable (Yunes et al., 2010).

A second environmental class involves dense dark-matter spikes around intermediate- or extreme-mass-ratio inspirals. In that case the orbital evolution is written as

$$t_{\rm mig}^{\rm bh,id} = \frac{J}{\left|\dot J_{\rm mig,I,II}+\dot J_{\rm gw}+\dot J_{\rm wind}\right|}.$$9

with dynamical friction and secondary accretion producing gravitational-wave dephasing relative to vacuum (Wade et al., 28 Aug 2025). The initial dark-matter distribution is crucial. A physically motivated angular-momentum cutoff $$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$0 removes particles captured by the primary black hole and yields a density profile

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$1

with $$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$2 (Wade et al., 28 Aug 2025). Replacing this with a sharp position-space cutoff overestimates the dephasing, especially for more extreme mass ratios. A prior merger event depletes the spike further and reduces the dephasing most strongly for less extreme mass ratios (Wade et al., 28 Aug 2025). Although the simulations in that work are for light IMRIs, the same environmental mechanisms are explicitly connected to EMRIs more broadly (Wade et al., 28 Aug 2025).

6. Multi-messenger science and observational applications

Wet EMRIs are of interest not only as formation channels but also as multi-messenger systems. In AGN disks they can be accompanied by transient electromagnetic signals, especially repeated flares produced when a misaligned inspiraling object crosses the disk (Lyu et al., 2024). A central mechanism invokes disk warping and Lense–Thirring precession. If the AGN disk is misaligned with the supermassive black-hole spin, the competition between frame dragging and disk torques produces a decoupling radius $$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$3 at which the stellar-mass black hole stops following the disk and begins precessing around the spin axis (Lyu et al., 2024). The quoted estimates are

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$4

for $$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$5-disks and

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$6

for $$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$7-disks (Lyu et al., 2024).

Inside this radius, twice-per-orbit disk crossings can shock the gas and generate quasi-periodic flares. The orbital energy lost per crossing is given by

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$8

with an estimate

$$\Gamma_{\rm mig} = -\frac{(dr/dt)_{\rm mig}}{r} =2 f_{\rm mig} \left(\frac{M_{\rm BH}}{M_{\rm SMBH}}\right) \left(\frac{\Sigma_{\rm gas}\, r}{M_{\rm SMBH}}\right) \left(\frac{h}{r}\right)^{-2},$$9

These have been proposed as type II quasi-periodic eruptions associated with AGN disks and wet EMRIs, in contrast to type I QPEs associated with tidal disruption events (Lyu et al., 2024).

The same study emphasizes precision gravitational-wave measurements of supermassive black-hole mass and spin, quoting uncertainties of order $f_{\rm mig}=2$0 to $f_{\rm mig}=2$1 in mass and $f_{\rm mig}=2$2 to $f_{\rm mig}=2$3 in spin for wet EMRIs (Lyu et al., 2024). This level of precision is argued to enable calibration of electromagnetic techniques such as broad-line virial mass estimators and X-ray reflection spectroscopy (Lyu et al., 2024). Wet EMRIs have also been proposed as probes of jet formation, because the gravitational waves can constrain the supermassive black-hole spin direction while electromagnetic data may constrain the jet direction, allowing comparison with Blandford–Znajek and Blandford–Payne expectations (Lyu et al., 2024).

Cosmological applications follow from the natural AGN association. If the host AGN is uniquely identified, the source acts as a bright siren; if not, restricting host candidates to AGNs still makes the event a useful dark siren (Lyu et al., 2024). Under one set of assumptions, roughly $f_{\rm mig}=2$4–$f_{\rm mig}=2$5 of detectable wet EMRIs have resolvable AGN hosts, bright-siren rates are about $f_{\rm mig}=2$6–$f_{\rm mig}=2$7 per year, $f_{\rm mig}=2$8 bright sirens can give $f_{\rm mig}=2$9 precision on $1\%$00, and $1\%$01 dark sirens can yield about $1\%$02 precision (Lyu et al., 2024). These are explicitly framed as opportunities for LISA and related mHz observatories.

7. Interpretive issues, uncertainties, and current directions

Several recurrent interpretive issues shape the literature. The first concerns whether wet EMRIs are simply dry EMRIs with gas perturbations. Current formation studies reject that reduction. Wet EMRIs occupy a distinct dissipative channel in which the AGN disk can dominate capture and migration, reshape the eccentricity and inclination distributions, alter the component-mass spectrum through accretion and mergers, and generate electromagnetic counterparts (Pan et al., 2021, Sun et al., 30 Aug 2025, Lyu et al., 2024).

The second concerns circularity. Early rate arguments emphasized that wet EMRIs should be nearly circular in the detector band because $1\%$03 in thin disks (Pan et al., 2021). Later population work qualifies this by showing that turbulence and resonance-driven multibody migration can maintain or pump eccentricity to $1\%$04–$1\%$05, with occasional excursions to $1\%$06 (Sun et al., 30 Aug 2025). A plausible implication is that low but nonzero eccentricity may become a particularly informative observable, because it is sensitive not only to channel classification but also to AGN-disk turbulence and interaction history.

The third concerns rates. Smooth-disk models suggested that wet EMRIs may contribute an important or even dominant fraction of all detectable EMRIs for spaceborne detectors (Pan et al., 2021). Recoil-regulated modeling finds substantially lower rates, $1\%$07–$1\%$08, after including binary formation, hierarchical mergers, and recoil kicks in AGN disks (Xue et al., 28 May 2026). This is not a contradiction in formalism so much as a consequence of different physical assumptions. The current disagreement therefore centers on whether AGN disks can be approximated as smooth migration channels or must be treated as compact-object ecosystems in which stellar interactions regulate the supply of EMRI progenitors.

The fourth concerns the breadth of the term itself. In AGN-disk studies, “wet EMRI” generally denotes the gas-assisted channel. In waveform-environment studies, the same language is extended to EMRIs whose signals encode nearby massive perturbers or dark-matter spikes [(Yunes et al., 2010); (Wade et al., 28 Aug 2025)]. This suggests a useful conceptual hierarchy: a narrow definition tied to AGN-disk formation, and a broader definition encompassing non-vacuum EMRIs whose phase evolution probes the environment.

Across these strands, the unifying theme is that wet EMRIs are not fully specified by the mass ratio and the Kerr background alone. Their dynamics and observables encode disk structure, AGN lifetime, turbulence, compact-object interactions, nearby perturbers, or dark-matter distributions, making them both a source-class for strong-field gravity and a probe of galactic-nuclear astrophysics (Lyu et al., 2024, Xue et al., 28 May 2026).