Late-Time HST UV Detections Reveal Eruptive Mass Loss and Circumstellar Interaction in a Quarter of Stripped-Envelope Supernovae

Abstract

We present HST WFC3/UVIS F275W near-UV imaging of 91stripped-envelope supernovae (SE SNe; Types IIb, Ib, Ic) from Snapshot program SNAP-16657, observed at phases of 270-1845 days (median 952 days) after first optical detection. We detect UV counterparts in 13 SE~SNe, of which 6 are classified as secure and 7 as ambiguous after comparison to nearby H\textsc{ii} regions, interpreting the secure sources as signatures of interaction with circumstellar material (CSM). Independent WISE W1/W2 light curves show $>300$ day mid-IR excesses in two of the secure UV sources, corroborating the interaction interpretation, and reveal two additional IR-only candidates without UV counterparts, indicating dust-obscured interaction episodes missed by the UV survey. A forward-modeling MCMC analysis using a physics-based CSM interaction model with three free parameters, the interaction fraction $f_\mathrm{CSM}$, shell mass $M_\mathrm{CSM}$, and thickness fraction $f_\mathrm{thick}$, yields $f_\mathrm{CSM} = 0.23^{+0.17}_{-0.09}$, $M_\mathrm{CSM} \approx 0.013~M_\odot$, and $f_\mathrm{thick} \approx 0.07$. The inferred thin-shell geometry implies an ejection duration of $\sim$6 yr for an outflow velocity of $300$ km s$^{-1}$, two to three orders of magnitude shorter than the thermal timescale of stable Roche-lobe overflow. This result disfavors steady binary mass transfer as the origin of the detected CSM and instead points to eruptive pre-supernova mass ejection in the final years before core collapse, either from wave-driven outbursts or from mass transfer triggered by late-stage progenitor re-expansion.

Figures (5)

• Figure 1: Rest-frame absolute ZTF forced-photometry $r$-band light curves for the HST-observed SE SN sample. The gray and colored curves show the full HST-observed SE SN sample. Colored curves highlight objects with long-lived, multi-peaked, or unusually luminous optical evolution: SN 2019oys (ZTF19abucwzt; Ib; orange), SN 2019obh (ZTF19abqykei; IIb; green), SN 2019oub (ZTF19abtsnyy; Ic; blue), SN 2018dfi (ZTF18abffyqp; IIb; pink), SN 2019dwf (ZTF19aarfkch; IIb; black), SN 2018hmx (ZTF18accnoli; IIb; red), and SN 2019vsi (ZTF19acxpuql; Ib; purple). Boldface legend entries indicate HST UV detection or a strong late-time WISE excess, or both; the bracketed legend tags specify [UV], [IR], or [UV+IR]. SN 2019oys is detected in both HST UV and WISE, SN 2019obh is detected in IR only, SN 2018hmx is detected in UV only, and the remaining four highlighted outliers show neither signature.
• Figure 2: HST WFC3/UVIS cutouts for all candidate detections (confirmed and ambiguous). Each panel shows a $1\hbox{$.\!\!^{\prime\prime}$}0$ radius region centered on the expected SN position. Red dashed ellipses mark the $3.5\sigma$ astrometric uncertainty region derived from the per-axis ZTF and HST WCS uncertainties (Table \ref{['tab:astrometry']}). A thick solid cyan circle marks the adopted primary counterpart used for the reported photometry, while thinner dashed cyan circles mark additional FLC-confirmed sources within the uncertainty region. Red squares mark sources rejected by the FLC comparison as cosmic rays. The panel-title color indicates whether the retained source is classified as secure (cyan) or ambiguous (gray). The grayscale is inverted (black = positive flux).
• Figure 3: Three-dimensional surface representations of HST WFC3/UVIS cutouts for all candidate detections (confirmed and ambiguous). Each panel pair shows a $1\hbox{$.\!\!^{\prime\prime}$}0$ radius region with the DRC image (left) and the PSF-subtracted residual (right) as peak-normalized flux surfaces. Transparent red shading marks the $3.5\sigma$ astrometric uncertainty ellipse projected onto the image surface. Transparent cyan shading marks the primary candidate position. Solid red patches mark sources rejected as cosmic rays by the FLC comparison. For secure detections, the residual panels demonstrate the quality of the PSF fit: a clean subtraction produces a flat residual surface at the candidate position, confirming the point-source nature of the detection. The panel-title color indicates whether the retained source is classified as secure (cyan) or ambiguous (gray).
• Figure 4: Luminosity density $L_\lambda$ (erg s$^{-1}$ Å$^{-1}$) at each filter's pivot wavelength vs. phase (from first ZTF 5$\sigma$ forced-photometry detection) for our HST survey. The SE SN sample is observed in F275W ($\lambda_\mathrm{pivot} \approx 2704$ Å); reference models, SN 2023ixf, and WISE data are plotted at their respective wavelengths. Symbols are defined in the legend; numbered labels identify individual detections. Downward arrows use the adopted $70\%$-coverage astrometry-marginalized PSF injection/recovery limits. Open circles/squares connected by dashed lines show mean late-time WISE W1/W2 excesses (read against the right-hand AB-magnitude axis). Gray curves show reference models: $^{56}$Ni decay of SN 2011fe in UVW1 (dot-dashed), Type IIn SN 2005ip in UVW1 (solid), and the Type Ia--CSM SN 2015cp at 664 d (red circle; GrahamM2019). Green diamonds trace the Swift/UVOT UVW1 light curve of SN 2023ixf from $\sim$24 to $\sim$648 d Jacobson-Galan25; the green shaded band shows our CSM interaction model (Section \ref{['sec:csmmodel']}) for $M_\mathrm{CSM} = 0.005$--$0.03~M_\odot$. Our SE SN sample yields 6 secure and 7 ambiguous late-time UV detections.
• Figure 5: Corner plot of the three-parameter MCMC posterior for the CSM interaction fraction $f_\mathrm{CSM}$, shell mass $\log_{10}(M_\mathrm{CSM}/M_\odot)$, and ejection duration $\Delta t_\mathrm{eject}$ (§\ref{['sec:csmmodel']}). The ejection duration is derived from the shell thickness fraction $f_\mathrm{thick}$ assuming a fiducial outflow velocity $v_\mathrm{w} = 300$ km s$^{-1}$ (motivated by wave-driven outflow models for compact helium stars; Fuller17RoFuller21) and a fiducial onset time of 1000 days; it scales inversely with $v_\mathrm{w}$. The choice of $v_\mathrm{w}$ affects only the ejection-duration axis and does not alter the posteriors for $f_\mathrm{CSM}$ or $M_\mathrm{CSM}$. Diagonal panels show 1D marginal posteriors with 68% credible intervals (shaded). Off-diagonal panels show 2D joint posteriors with 67% and 95% highest-density contours. Rose dashed and dotted lines mark the median and central 67% of the literature ranges for SN 2023ixf: $M_\mathrm{CSM} \approx 0.01~M_\odot$Jacobson-Galan23Hiramatsu23 and an ejection duration of $\sim$3--6 yr inferred from flash spectroscopy Jacobson-Galan23. The SE SN posterior ($\Delta t_\mathrm{eject} \sim 6$ yr) is broadly consistent with the SN 2023ixf timescale and rules out steady binary mass transfer ($10^3$--$10^4$ yr).