AT 2021sdu: TDE Radio Outflow

• The paper establishes that AT 2021sdu is a tidal disruption event characterized by a single, slow, non-relativistic radio outflow (v ≈ 0.03c, E ~ 10^48 erg) amid host contamination.
• Multi-frequency radio observations from VLA, uGMRT, and NOEMA, combined with equipartition modeling, are used to derive key outflow parameters and track its evolution.
• The study underscores the challenge of disentangling transient TDE signals from persistent host galaxy emission in star-forming, composite systems.

Searching arXiv for AT 2021sdu and related TDE radio studies to ground the article in current literature. AT 2021sdu is an optically selected tidal disruption event (TDE) discovered by ZTF/ALeRCE on 2021 July 5 UTC, with TNS discovery MJD 59400.9. It was subsequently classified as a TDE by ZTF-based spectroscopic and photometric analysis, and its optical/UV evolution appears typical for the class, with no exotic line features or light-curve anomalies compared to other thermal TDEs. In radio, AT 2021sdu exhibited transient emission beginning shortly after optical discovery and persisting for several years, but the long-lived signal is not purely transient: the event is best explained by a single slow, non-relativistic outflow combined with diffuse, non-variable host emission that becomes dominant about 500 days after discovery (Christy et al., 17 Sep 2025).

1. Discovery, classification, and host-galaxy context

AT 2021sdu is hosted by WISEA J011123.92+503429.7 at a luminosity distance $D_L = 264\,\mathrm{Mpc}$, corresponding to $z \approx 0.059$ (Christy et al., 17 Sep 2025). The event belongs to the class of thermal TDEs on the basis of its optical and ultraviolet behavior. A late-time Bok 2.3 m spectrum obtained at $\delta t \approx 1234$ d shows narrow emission lines including H$\beta$, [O III] $\lambda\lambda4959,5007$, [N II] $\lambda\lambda6548,6584$, H$\alpha$, and [S II] $\lambda\lambda6717,6731$.

BPT diagnostics place the host in the “composite” region rather than in a pure AGN or pure H II-galaxy locus. Radio and H$\alpha$ measurements indicate significant ongoing star formation, with a derived star-formation rate of order a few $M_\odot\,\mathrm{yr^{-1}}$. This matters directly for the radio interpretation, because the nuclear radio source is not purely transient. The host is therefore both star-forming and possibly weakly AGN-like, and the paper explicitly notes that this ambiguity bears on the late-time radio analysis (Christy et al., 17 Sep 2025).

Within the same study, AT 2021sdu is contrasted with AT 2020zso. AT 2020zso shows two distinct radio outflows, whereas AT 2021sdu exhibits a single radio transient component plus significant host contamination. This distinction is central to the interpretation of AT 2021sdu: the phenomenology is not that of multiple resolved radio outflow episodes, but of one fading transient embedded in a structured host radio environment (Christy et al., 17 Sep 2025).

2. Radio observational campaign and empirical evolution

The radio campaign on AT 2021sdu extends for more than three years after discovery and combines VLA, uGMRT, and NOEMA observations. The VLA observations include $z \approx 0.059$0, $z \approx 0.059$1, $z \approx 0.059$2, $z \approx 0.059$3, $z \approx 0.059$4, $z \approx 0.059$5, $z \approx 0.059$6, and $z \approx 0.059$7 bands, spanning 1–46 GHz and using A, B, and C array configurations. uGMRT observations were obtained in Band 4 at 0.65 GHz and Band 5 at 1.26 GHz. NOEMA observed the 3 mm band centered at 88.5 GHz, yielding mostly upper limits and one detection at $z \approx 0.059$8 d (Christy et al., 17 Sep 2025).

The first radio detection occurred at 15 GHz with the VLA at $z \approx 0.059$9 d. Multi-frequency VLA coverage then followed at $\delta t \approx 1234$0, 146, 205, 314, 349, and 461 d, with later host-dominated VLA epochs at approximately 677, 817, 1040, and 1147 d. uGMRT observations at $\delta t \approx 1234$1, 677, 791, 792, and 965 d constrain the low-frequency behavior of the nuclear source (Christy et al., 17 Sep 2025).

At early times and higher frequencies, the radio source is compact and strongly variable. Fluxes rise to approximately $\delta t \approx 1234$2 at 10–22 GHz before decaying, which is interpreted as the signature of a nuclear synchrotron transient. The host-subtracted light curves at 6, 10, 15, and 22 GHz are fit with broken power laws. The frequency dependence of the peak is explicit: 22 GHz peaks earliest at $\delta t \approx 1234$3 d with $\delta t \approx 1234$4; 15 GHz peaks at $\delta t \approx 1234$5 d with $\delta t \approx 1234$6; 10 GHz peaks at $\delta t \approx 1234$7 d with $\delta t \approx 1234$8; and 6 GHz peaks at $\delta t \approx 1234$9 d with $\beta$0 (Christy et al., 17 Sep 2025).

The fitted light-curve slopes are common across bands, with rise index $\beta$1 and decay index $\beta$2. In the adopted interpretation, this frequency-dependent peak time and amplitude are expected for a synchrotron spectrum with an evolving self-absorption frequency $\beta$3 (Christy et al., 17 Sep 2025).

3. Host contamination and radio decomposition

A defining feature of AT 2021sdu is that the late-time nuclear radio source is diffuse and spatially extended at low frequencies, especially at $\beta$4 GHz. High-resolution 3 GHz VLA imaging shows this directly, and the uGMRT measurements confirm that the low-frequency flux is non-variable over long timescales. At 0.65 GHz, the measured flux density is approximately $\beta$5–$\beta$6 across $\beta$7–965 d; at 1.26 GHz, it is approximately $\beta$8–$\beta$9 and likewise non-variable (Christy et al., 17 Sep 2025).

The non-variability and extended morphology indicate that the low-frequency radio emission is dominated by the host galaxy rather than by the TDE transient. Because the VLA observations were obtained in multiple configurations, the amount of diffuse host emission recovered varies with epoch. The analysis therefore constructs a configuration-dependent host spectrum using the fixed-configuration uGMRT data. The host is modeled as a steep power law,

$\lambda\lambda4959,5007$0

with a common spectral index $\lambda\lambda4959,5007$1 and different normalizations for VLA B-like and C-like uv ranges: $\lambda\lambda4959,5007$2 and $\lambda\lambda4959,5007$3. A-configuration data adopt the B-model, though some are imaged with B uv-range constraints (Christy et al., 17 Sep 2025).

After subtracting this host component from the VLA measurements, the transient is clearly visible from $\lambda\lambda4959,5007$4 d to $\lambda\lambda4959,5007$5 d. Beyond $\lambda\lambda4959,5007$6 d, the host-subtracted flux densities are consistent with zero, meaning that the measured radio emission is fully accounted for by the host model. The study emphasizes that “host contamination becomes dominant $\lambda\lambda4959,5007$7 days after discovery,” particularly below $\lambda\lambda4959,5007$8 GHz, and that array-configuration changes can mimic variability if diffuse emission is not modeled carefully (Christy et al., 17 Sep 2025).

The host contribution is also quantified in terms of star formation. When the uGMRT Band 5 data are imaged with uv cuts corresponding to VLA-C, the inferred 1.26 GHz luminosity is $\lambda\lambda4959,5007$9, implying $\lambda\lambda6548,6584$0 under the Murphy et al. (2011) calibration. The H$\lambda\lambda6548,6584$1 luminosity, $\lambda\lambda6548,6584$2, implies $\lambda\lambda6548,6584$3, broadly consistent with the radio estimate (Christy et al., 17 Sep 2025).

4. Spectral modeling and equipartition framework

The transient radio component is modeled as synchrotron emission from a non-relativistic, roughly spherical blast wave propagating into the circumnuclear medium (CNM). The assumed emitting region is a spherical shell of shocked gas behind the forward shock, with radius $\lambda\lambda6548,6584$4 and thickness $\lambda\lambda6548,6584$5. The electron distribution is taken to be a power law,

$\lambda\lambda6548,6584$6

with $\lambda\lambda6548,6584$7 and $\lambda\lambda6548,6584$8. The microphysics are fixed to equipartition values $\lambda\lambda6548,6584$9, and the geometry is parameterized by filling factors $\alpha$0 and $\alpha$1 (Christy et al., 17 Sep 2025).

Each radio epoch is fit with a smoothed singly broken power-law spectrum,

$\alpha$2

with $\alpha$3 and $\alpha$4. For early epochs, $\alpha$5 d, the preferred spectral ordering is $\alpha$6, giving $\alpha$7 and $\alpha$8. For later epochs, $\alpha$9 d, the ordering is interpreted as $\lambda\lambda6717,6731$0, yielding $\lambda\lambda6717,6731$1 and $\lambda\lambda6717,6731$2 (Christy et al., 17 Sep 2025).

The shift in the optically thin slope from about $\lambda\lambda6717,6731$3 to about $\lambda\lambda6717,6731$4 is attributed to the cooling break $\lambda\lambda6717,6731$5 moving rapidly through the band. The fitted self-absorption frequency and peak flux evolve substantially: $\lambda\lambda6717,6731$6 and $\lambda\lambda6717,6731$7 at $\lambda\lambda6717,6731$8 d; $\lambda\lambda6717,6731$9 and $\alpha$0 at $\alpha$1–167 d; $\alpha$2 and $\alpha$3 at $\alpha$4–220 d; $\alpha$5 and $\alpha$6 at $\alpha$7–314 d; and $\alpha$8 with $\alpha$9 at $M_\odot\,\mathrm{yr^{-1}}$0–367 d. By $M_\odot\,\mathrm{yr^{-1}}$1–461 d only an upper limit, $M_\odot\,\mathrm{yr^{-1}}$2, is obtained (Christy et al., 17 Sep 2025).

Using the equipartition formalism of Barniol Duran et al. (2013), the analysis inverts the observed $M_\odot\,\mathrm{yr^{-1}}$3 and $M_\odot\,\mathrm{yr^{-1}}$4 to infer $M_\odot\,\mathrm{yr^{-1}}$5, $M_\odot\,\mathrm{yr^{-1}}$6, $M_\odot\,\mathrm{yr^{-1}}$7, $M_\odot\,\mathrm{yr^{-1}}$8, and $M_\odot\,\mathrm{yr^{-1}}$9. The synchrotron characteristic frequency is written as

$z \approx 0.059$00

and the free-expansion relation between radius and observer time is approximated by

$z \approx 0.059$01

after noting that for a non-relativistic flow $z \approx 0.059$02 (Christy et al., 17 Sep 2025).

5. Derived outflow properties

The equipartition analysis yields radii of order $z \approx 0.059$03 cm. Specifically, the inferred values are $z \approx 0.059$04 at $z \approx 0.059$05 d, 16.24 at $z \approx 0.059$06 d, 16.53 at $z \approx 0.059$07 d, 16.41 at $z \approx 0.059$08 d, 16.44 at $z \approx 0.059$09 d, and $z \approx 0.059$10 at $z \approx 0.059$11 d. The evolution is almost, but not strictly, monotonic; the apparent decrease between 212 and 297 d is identified in the study as physically suspicious (Christy et al., 17 Sep 2025).

A linear fit to $z \approx 0.059$12 implies a launch time relative to optical discovery of

$z \approx 0.059$13

so the outflow is inferred to have begun 53 days before optical discovery. Using the dynamical age $z \approx 0.059$14, the tabulated velocities are $z \approx 0.059$15 at dynamical age 121 d, 0.033 at 210 d, 0.050 at 265 d, 0.029 at 350 d, 0.027 at 402 d, and $z \approx 0.059$16 at 514 d. The characteristic velocity quoted in the paper is

$z \approx 0.059$17

establishing that the outflow is non-relativistic (Christy et al., 17 Sep 2025).

The total internal energy from equipartition is $z \approx 0.059$18 at $z \approx 0.059$19 d, 48.25 at $z \approx 0.059$20 d, and 48.61 at $z \approx 0.059$21 d, with later epochs remaining at approximately $z \approx 0.059$22 erg but affected by the radius issue. The abstract summarizes the event as a slower, less energetic outflow with $z \approx 0.059$23 and $z \approx 0.059$24 (Christy et al., 17 Sep 2025).

Using

$z \approx 0.059$25

the paper estimates an outflow mass of approximately $z \approx 0.059$26 for $z \approx 0.059$27 and $z \approx 0.059$28. The swept-up CNM mass is much smaller, $z \approx 0.059$29, so $z \approx 0.059$30, consistent with the free-expansion assumption (Christy et al., 17 Sep 2025).

The external density declines slowly with radius. The quoted values are $z \approx 0.059$31 at $z \approx 0.059$32 d, corresponding to about $z \approx 0.059$33; about $z \approx 0.059$34 at $z \approx 0.059$35 d; $z \approx 0.059$36 at $z \approx 0.059$37 d; $z \approx 0.059$38 at $z \approx 0.059$39 d; $z \approx 0.059$40 at $z \approx 0.059$41 d; and $z \approx 0.059$42 at $z \approx 0.059$43 d. The magnetic field decreases from about $z \approx 0.059$44 at 68 d to approximately $z \approx 0.059$45–$z \approx 0.059$46 at later times (Christy et al., 17 Sep 2025).

6. Interpretation, limitations, and relation to the TDE radio population

The inferred combination of $z \approx 0.059$47, $z \approx 0.059$48, and $z \approx 0.059$49 d admits two principal interpretations in the paper. One is an origin in unbound stellar debris ejected during disruption. In that picture, the early launch time, preceding optical discovery, and the relatively low velocity and energy are compatible with a radio-emitting shock driven by unbound debris streams. The alternative is an accretion-driven outflow, such as a disk wind or a collisionally induced outflow, launched early during circularization and close in time to the optical flare (Christy et al., 17 Sep 2025).

The paper does not treat the simple spherical free-expansion model as fully satisfactory. A major difficulty is that between $z \approx 0.059$50 and 297 d, the peak flux density $z \approx 0.059$51 drops by a factor of about 4 and the derived radius decreases, which is unphysical for a forward shock. The authors discuss several possible explanations: changes in source geometry, such as declining emitting-area and filling factors as the outflow becomes patchy or fragmented; complex CNM structure, including clumps; and the limitations of a minimal equipartition model in the presence of anisotropy or non-equipartition effects. They also note that varying $z \approx 0.059$52 and $z \approx 0.059$53 to recover a monotonic radius evolution would require implausibly large changes, such as $z \approx 0.059$54 dropping by about 100 (Christy et al., 17 Sep 2025).

A second difficulty is the rapid change in the optically thin spectral slope, interpreted as the cooling break moving through the band. The paper states that this requires very fast evolution of cooling conditions and is difficult to reproduce in simple models. This suggests that a pure spherical, constant-parameter outflow is too simplistic. Plausible alternatives considered in the study include fragmented unbound debris, an asymmetric outflow interacting with a clumpy CNM, or a mixture of unbound debris and disk winds (Christy et al., 17 Sep 2025).

In the broader TDE radio population, AT 2021sdu occupies the low-velocity, low-energy end of the distribution discussed in the paper. Compared with AT 2020zso, whose outflows have $z \approx 0.059$55–$z \approx 0.059$56 and $z \approx 0.059$57, AT 2021sdu is slower and less energetic. It also differs qualitatively in multiplicity: AT 2020zso shows two physically distinct outflows, whereas AT 2021sdu shows only one transient component, with its apparent late-time disappearance caused by host dominance rather than by the emergence of a second outflow (Christy et al., 17 Sep 2025).

The event therefore exemplifies two points emphasized by the study. First, TDE radio outflows span a wide range of energies and velocities, and AT 2021sdu likely represents a relatively weak, slow outflow. Second, late-time interpretation can be limited by host contamination, especially in star-forming or composite hosts. In AT 2021sdu, the diffuse steep-spectrum host becomes dominant after about 500 d, so the observed late-time radio emission is not evidence for continued transient activity by itself. This suggests that some low-luminosity TDE radio outflows may be systematically difficult to isolate once the transient fades into host emission (Christy et al., 17 Sep 2025).