Orbital Period Changes of Recurrent Nova U Scorpii Demonstrate that M$_{\rm ejecta}$=26$\times$M$_{\rm accreted}$ and Is Not a Type Ia Supernova Progenitor

TL;DR

U Scorpii is tested as a Type Ia supernova progenitor by directly measuring the ejecta mass through orbital-period changes across eruptions using an extensive eclipse-timing dataset. The authors combine 218 eclipse times (1945–2025) with a refined $\Delta P$–$M_{ m ejecta}$ relation, showing $M_{ m ejecta} = (103 \pm 14)\times10^{-6}\,M_\odot$ across four eruptions and $M_{ m accreted} \approx 4\times10^{-6}\,M_\odot$, giving $M_{ m ejecta}/M_{ m accreted} \approx 26$ and implying net WD mass loss. The results indicate U Sco cannot reach the Chandrasekhar mass via successive eruptions, and spectroscopic evidence of neon overabundances points to an ONe WD rather than a CO WD. This work demonstrates that long-term, precise eclipse timing provides a decisive test of SN Ia progenitor scenarios and has broad implications for CV evolution and SN Ia demographics.

Abstract

Recurrent nova U Scorpii (U Sco) is one of the prototypes for a Type Ia supernova progenitor. The logic is that the white dwarf is near the Chandrasekhar mass and gas is accumulating onto its surface at a near-maximal accretion rate, so it will soon increase its mass to the supernova trigger. But the white dwarf loses mass every nova eruption, so the issue is balancing the mass ejected ($M_{\rm ejecta}$) against the mass accreted between eruptions ($M_{\rm accreted}$). Measuring $M_{\rm accreted}$ can be done in several ways to useable accuracy. But the old methods for measuring $M_{\rm ejecta}$ (involving the flux in hydrogen emission lines) are all with real error bars of 2--3 orders of magnitude. The only solution is to measure the change of the orbital period across the nova eruption ($ΔP$). But this solution requires a vast photometric program of eclipse timings stretching decades. For U Sco, a program started in 1989, now reaches its culmination with measures of $ΔP$ for the eruptions of 1999, 2010, 2016, and 2022. This paper reports on 52 new eclipse times (for a total of 218 eclipses 1945--2025), plus a new theory result allowing for the confident calculation of $M_{\rm ejecta}$ from $ΔP$. The four eruptions ejected a total of (103$\pm$14)$\times$$10^{-6}$ $M_{\odot}$, while the white dwarf accreted 4$\times$$10^{-6}$ $M_{\odot}$ over the four previous eruption cycles. With M$_{\rm ejecta}$=26$\times$M$_{\rm accreted}$, the U Sco white dwarf is losing large masses each eruption cycle, so U Sco can never produce a Type Ia supernova.

Figures (6)

• Figure 1: Three U Sco eclipses. For each eclipse, our measured light curves are in the $CV$ band, with most photometric error bars being smaller than the plotting symbol. For each eclipse, we have fit a parabola by the usual chi-square minimization calculation, with the best-fitting parabola shown superposed on each light curve. To illustrate our eclipse light curves, we show our best case (top curve), typical case (middle curve), and worst case (bottom curve).
• Figure 2: TESS light curve for U Sco. This light curve is from the first orbit of Sector 91 of TESS in 2025 April, with 200 second time bins. The Poisson measurement error is $\pm$2.6 ct/s for each data point, which is comparable to the ordinary flickering noise always present in the U Sco light curve. The constant background flux level is unknown, because the large pixel size (21"$\times$21") contains other stars and a lot of background light, and this makes the amplitude of eclipse to appear to be small. This first orbit shows 5 eclipses, with the best fit parabolas shown as blue curves. The TESS eclipse light curves are poor because U Sco is near the limit of detection and because the background levels are high due to the pixel size. The second orbit of Sector 91 (i.e., the second half of the sector) has 6 eclipses, with these being substantially poorer in quality.
• Figure 3: Mysterious flares just before transition. This light curve shows only $B$, $V$, and $I$ (blue, green, and red diamonds respectively) for Days $+$7 to $+$12 after the start of the eruption, $T_0$, taken to be HJD 2459737.2. The observers contributing to these light curves are F.-J. Hambsch (with AAVSO observer ID of HMB), B. Harris (HBB), S. Dvorak (DKS), B. Monard (MLF), W. Cooney (CWT), M. Richmond (RHM), A. Pearce (PEX), R. Schmidt (SREB), S. M. Brincat (BSM), M. Larsson (LMN), C. Galdies (GCHB), and G. Myers (MGW). The times of zero orbital phase are represented by the thin vertical lines spaced every 1.23 days. The first eclipse seen during the eruption is centered on Day $+$11.5, while no eclipse is seen centered one orbit earlier. So the shell of ejecta is optically thick before Day $+$11. Before Day $+$8, the fading light curve is smooth. But from Days $+$8 to $+$11, U Sco suffered at least 8 flares with amplitudes 0.2--0.5 mags. These flares are fast, with rise times of 0.035--0.090 days, and durations of 0.110--0.165 days. Critically, the extra flare light is extremely blue, with a dereddened $B-V$ from $-$0.6 to $-$0.8. This can be seen by noting that the flare amplitude in $B$ is much larger than in $V$ and $R$. The extremity of this color rules out the various tries at explanation.
• Figure 4: 2010 & 2022 light curves. An important question is whether all eruptions of U Sco have identical light curves. Schaefer (2010) made an exhaustive comparison of 7 U Sco events before 2000 that are identical to within the measurement errors. Now, the 2010 and 2022 light curves are the two all-time best nova light curves (covering the entire eruption, with average frequency of under 2.6 minutes per magnitude throughout, with much $BVRI$ measures), so a detailed comparison will show even small differences. (Here, we do not consider differences due to ordinary flickering, due to the exact phasing of eclipses, nor due to the quiescent level being asymptotically approached.) Here, we plot the light curves for the two eruption, with the 2010 light curve taken from the $V$-band out-of-eclipse trend line from Schaefer (2011), and the green diamonds for all the AAVSO $V$ magnitudes with orbital phases from 0.1--0.9. For most of the duration of the eruption, the red curve is completely hidden under the green diamonds. At later times, the two curves asymptote to slightly different quiescent levels, so that the two light curves are identical when we only consider the eruption light. The small differences between the 2010 and 2022 light curves are consistent with the ordinary problems of comparing observations from groups of observers with differing comparison stars and color terms. That is, the 2010 and 2022 eruptions have identical light curves, with no measurable differences. This creates a conundrum because all the eruptions are photometrically (and spectroscopically) identical, yet the eruptions result in greatly different $\Delta P$ and $\dot{P}$ behavior. So the mechanism that changes the $P$ must be well hidden in the binary.
• Figure 5: U Sco $O-C$ curve for 218 eclipse times. For display purposes only, the plotted red diamonds are seasonal averages. The fiducial ephemeris for creating this $O-C$ curve uses a period of 1.23054695 days, and an epoch for $N=0$ of HJD 2451234.5387. The thin vertical lines indicate the date of an eruption, with the year labeled. The $O-C$ curve must consist of parabolas of steady period change $\dot{P}$ between eruptions, with kinks at the times of eruption due to the sudden period changes $\Delta P$. The best fitting model of broken parabolas is represented by the thick black curve. Startlingly, the $\Delta P$ is seen to change greatly from eruption-to-eruption, and the $\dot{P}$ curvature is also seen to vary greatly from eruption to eruption. We are not aware of any suggestion or understanding for how some mechanism makes for such changes in the nature of the period-changes. For the main purpose of this paper, we see and quantify the sharp upward kinks, where the large-positive $\Delta P$ measures are possible only for a huge $M_{\rm ejecta}$. These ejecta masses are up to 45$\times$ greater than the masses accreted onto the white dwarf in the prior part of the eruption cycle, so the white dwarf must be net losing large masses over each eruption cycle. With the U Sco white dwarf losing mass over the years, U Sco cannot be a Type Ia supernova progenitor.