ASASSN-18aan: An Eclipsing SU UMa-type Cataclysmic Variable with a 3.6-hour Orbital Period and a Late G-type Secondary Star
Yasuyuki Wakamatsu, John R. Thorstensen, Naoto Kojiguchi, Keisuke Isogai, Mariko Kimura, Ryuhei Ohnishi, Taichi Kato, Hiroshi Itoh, Yuki Sugiura, Sho Sumiya, Hanami Matsumoto, Daiki Ito, Kengo Nikai, Hiroshi Akitaya, Chihiro Ishioka, Kohei Oide, Takahiro Kanai, Yoshinori Uzawa, Yumiko Oasa, Tamás Tordai, Tonny Vanmunster, Sergey Yu. Shugarov, Masayuki Yamanaka, Mahito Sasada, Kengo Takagi, Yuki Nishinaka, Yuina Yamazaki, Ikki Otsubo, Tatsuya Nakaoka, Katsuhiro L. Murata, Ryou Ohsawa, Masahiro Morita, Makoto Ichiki, Sjoerd Dufoer, Masanori Mizutani, Takashi Horiuchi, Miyako Tozuka, Masaki Takayama, Tomohito Ohshima, Tomoki Saito, Pavol A. Dubovsky, Geoff Stone, Ian Miller, Daisaku Nogami
aa r X i v : . [ a s t r o - ph . S R ] F e b Publ. Astron. Soc. Japan (2018) 00(0), 1–17doi: 10.1093/pasj/xxx000 ASASSN-18aan: An Eclipsing SU UMa-typeCataclysmic Variable with a 3.6-hour OrbitalPeriod and a Late G-type Secondary Star
Yasuyuki W
AKAMATSU , John R. T
HORSTENSEN , Naoto K
OJIGUCHI , Keisuke I
SOGAI , Mariko K
IMURA , Ryuhei O
HNISHI , Taichi K
ATO , Hiroshi I
TOH , Yuki S
UGIURA , Sho S
UMIYA , Hanami M
ATSUMOTO , Daiki I TO , Kengo N
IKAI , Hiroshi A
KITAYA , Chihiro I
SHIOKA , Kohei O
IDE , Takahiro K
ANAI , Yoshinori U
ZAWA , Yumiko O
ASA , Tam ´as T
ORDAI , Tonny V
ANMUNSTER , Sergey Yu. S
HUGAROV , Masayuki Y
AMANAKA , Mahito S
ASADA , Kengo T
AKAGI , Yuki N
ISHINAKA , Yuina Y
AMAZAKI , Ikki O
TSUBO , Tatsuya N
AKAOKA , Katsuhiro L. M
URATA , Ryou O
HSAWA , Masahiro M
ORITA , Makoto I
CHIKI , Sjoerd D
UFOER , Masanori M
IZUTANI , Takashi H
ORIUCHI , Miyako T
OZUKA , Masaki T
AKAYAMA , Tomohito O
HSHIMA , Tomoki S
AITO , Pavol A. D
UBOVSKY , Geoff S
TONE , Ian M
ILLER , and Daisaku N OGAMI Department of Astronomy, Kyoto University, Kyoto 606-8502, Japan Department of Physics and Astronomy, 6127 Wilder Laboratory, Dartmouth College,Hanover, NH 03755-3528, USA Okayama Observatory, Kyoto University, 3037-5 Honjo, Kamogatacho, Asakuchi, Okayama719-0232, Japan Extreme Natural Phenomena RIKEN Hakubi Research Team, Cluster for PioneeringResearch, RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan Variable Star Observers League in Japan (VSOLJ), 1001-105 Nishiterakata, Hachioji, Tokyo192-0153, Japan Osaka Kyoiku University, 4-698-1 Asahigaoka, Kashiwara, Osaka 582-8582, Japan Graduate School of Science and Engineering, Saitama University, 255 Shimo-Okubo,Sakura-ku, Saitama, Saitama 338-8570, Japan Hiroshima Astrophysical Science Center, Hiroshima University, 1-3-1 Kagamiyama,Higashi-Hiroshima, Hiroshima 739-8526, Japan Graduate School of Education, Saitama University, 255 Shimo-Okubo, Sakura-ku, Saitama,Saitama 338-8570, Japan Faculty of Education, Saitama University, 255 Shimo-Okubo, Sakura-ku, Saitama, Saitama338-8570, Japan Polaris Observatory, Hungarian Astronomical Association, Laborc utca 2/c, 1037 Budapest,Hungary Center for Backyard Astrophysics Belgium, Walhostraat 1A, B-3401 Landen, Belgium Sternberg Astronomical Institute, Lomonosov Moscow State University, Universitetsky Ave.,13, Moscow 119992, Russia Astronomical Institute of the Slovak Academy of Sciences, 05960 Tatranska Lomnica,Slovakia © 2018. Astronomical Society of Japan.
Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 Department of Physics, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo152-8551, Japan Institute of Astronomy, Graduate School of Science, The University of Tokyo, 2-21-1Osawa, Mitaka, Tokyo 181-0015, Japan Kiso Observatory, Institute of Astronomy, Graduate School of Science, The University ofTokyo 10762-30, Mitake, Kiso-machi, Kiso-gun, Nagano 397-0101, Japan Vereniging Voor Sterrenkunde (VVS), Oostmeers 122 C, 8000 Brugge, Belgium Variable Star Observers League in Japan (VSOLJ), Okayama, Japan Ishigakijima Astronomical Observatory, Public Relations Center, National AstronomicalObservatory of Japan, 1024-1 Arakawa, Ishigaki, Okinawa, 907-0024, Japan Nishi-Harima Astronomical Observatory, Center for Astronomy, University of Hyogo, 407-2,Nishigaichi, Sayo-cho, Sayo, Hyogo 679-5313, Japan Vihorlat Observatory, Mierova 4, 06601 Humenne, Slovakia American Association of Variable Star Observers, 49 Bay State Rd., Cambridge, MA02138, USA Furzehill House, Ilston, Swansea, SA2 7LE, UK ∗ E-mail: ∗ [email protected] Received 201 0; Accepted 201 0
Abstract
We report photometric and spectroscopic observations of the eclipsing SU UMa-type dwarfnova ASASSN-18aan. We observed the 2018 superoutburst with 2.3 mag brightening andfound the orbital period ( P orb ) to be 0.149454(3) d, or 3.59 hr. This is longward of the periodgap, establishing ASASSN-18aan as one of a small number of long- P orb SU UMa-type dwarfnovae. The estimated mass ratio, ( q = M /M = 0 . ), is almost identical to the upperlimit of tidal instability by the 3:1 resonance. From eclipses, we found that the accretion diskat the onset of the superoutburst may reach the 3:1 resonance radius, suggesting that thesuperoutburst of ASASSN-18aan results from the tidal instability. Considering the case oflong- P orb WZ Sge-type dwarf novae, we suggest that the tidal dissipation at the tidal truncationradius is enough to induce SU UMa-like behavior in relatively high- q systems such as SU UMa-type dwarf novae, but that this is no longer effective in low- q systems such as WZ Sge-typedwarf novae. The unusual nature of the system extends to the secondary star, for which we finda spectral type of G9, much earlier than typical for the orbital period, and a secondary mass M of around 0.18 M ⊙ , smaller than expected for the orbital period and the secondary’s spectraltype. We also see indications of enhanced sodium abundance in the secondary’s spectrum.Anomalously hot secondaries are seen in a modest number of other CVs and related objects.These systems evidently underwent significant nuclear evolution before the onset of masstransfer. In the case of ASASSN-18aan, this apparently resulted in a mass ratio lower thantypically found at the system’s P orb , which may account for the occurrence of a superoutburstat this relatively long period. Key words: accretion, accretion disks — stars: novae, cataclysmic variables — stars: dwarf novae —stars: individual (ASASSN-18aan)
Cataclysmic variables (CVs) are close binary systems con-sisting of a white dwarf primary star and a secondary starthat fills its Roche lobe and transfers mass to the primary via the inner Lagrangian point L . Unless the white dwarfis highly magnetized, an accretion disk forms around theprimary. Dwarf novae (DNe) are a subclass of CVs charac-terized by recurrent outbursts of 2-10 magnitude, evidentlycaused by accretion disk instabilities. ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 Over most of the life of a CV, the binary separationgradually decreases due to angular momentum loss fromthe system, resulting shortening of the orbital period. Themass ratio also decreases due to the mass transfer from thesecondary to the primary. CVs therefore generally evolvetoward extreme states with low mass ratios and short or-bital periods. Relatively few non-magnetic CVs are ob-served in a period range from roughly 2 to 3 hr, knownas the period gap (see Knigge et al. 2011 and referencestherein for more on CV evolution).SU UMa-type DNe, a subclass of DNe, show long-lasting, brighter outbursts called superoutbursts in additionto (normal) outbursts. It is a key feature of superoutburststo be accompanied by superhumps, which are variations ofsmall amplitude, typically 0.1-0.5 mag, and have slightlylonger periods than orbital periods. The superoutburst isconsidered to be a result of the tidal instability that is trig-gered when the outer disk reaches the 3:1 resonance radius(Whitehurst 1988; Hirose, Osaki 1990; Lubow 1991; Lubow1991). The expansion of the accretion disk to the 3:1 reso-nance radius can only occur if the mass ratio q = M /M isless than 0.25, (Whitehurst 1988), or 0.33 with reduction ofthe mass transfer from the secondary (Murray et al. 2000).The period of most SU UMa-type DNe lie below the periodgap; at these short periods, the mass rations are typically < .
25 The expansion of the accretion disk is limited bythe tidal truncation radius, where the angular momentumof orbiting material is removed via tidal torque. The 3:1resonance, therefore, is considered to occur when the tidaltruncation radius is larger than the 3:1 resonance radius.The detailed action of the tidal truncation radius, how-ever, is still unknown and it is unclear whether it acts as ahard limit of disk expansion, especially in extremely low- q systems (Osaki, Meyer 2002).A small number of CVs with longer periods have shownthe superhumps and superoutbursts characteristic of SUUMa stars; these long- P orb objects should offer tests ofthe tidal instability model and the effects of tidal torquesat the tidal truncation radius.The longest-known of these long-period SU UMa starsis TU Men (Stolz, Schoembs 1984), which has an orbitalperiod P orb = 0 . q > . P = 0 . ± P orb SU UMa-type DNe, some WZSge-type DNe have been reported that have atypically longorbital periods. ASASSN-16eg (Wakamatsu et al. 2017),which showed superoutburst accompanied by clear earlysuperhumps, has long orbital period and anomalously largemass ratio, almost twice the upper limit of the 2:1 reso-nance. Osaki, Meyer (2002) suggested that in extremelylow q -systems such as WZ Sge-type DNe, the tidal torqueby the secondary at the tidal truncation radius is weak andthe outer disk could reach the 2:1 resonance radius beyondthe tidal truncation radius.In this paper, we report observations of ASASSN-18aan( α = 0 h m s . , δ = +62 ◦ ′ ′′ .
9, from the GaiaCollaboration et al. 2018). The superoutburst was de-tected on 2018 November 30 by the All-Sky AutomatedSurvey for Supernovae (ASAS-SN: Shappee et al. 2014).The inverse of the Gaia DR2 parallax is 675 (+32, − Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 pc (Gaia Collaboration et al. 2018). Green et al. (2018)computed three-dimensional reddening maps across thenorthern sky . Their map gives E ( g − r ) = 0 .
35 at theGaia distance along this line of sight; using expressionsfound in Green et al. (2019) we convert this to A V = 1 . E ( B − V ) = 0 .
36 mag, where we have assumed R = A V /E ( B − V ) = 3 .
32. Nesci et al. (2019) obtained anoptical spectrum indicating the spectral type of the sec-ondary is G–K. They also investigated historical outburstsof ASASSN-18aan from plates and found that the recur-rence time may be about 11 months. The object showedsuperhump-like modulations in the 2018 superoutbursts,as well as clear eclipses. The eclipses revealed on orbitalperiod that was quite long for an SU UMa-type DN, 0.1495d as a tentative value, which attracted special interest(vsnet-alert 22806, 22810). Observations showed the su-perhump period to be much longer than the orbital one,indicating an unusually large mass ratio (vsnet-alert 22816,22817, 22821). We therefore performed world-wide photo-metric and spectroscopic observations. We also performedphotometric and spectroscopic observations in quiescentstate to clarify its nature and binary parameters. Our ob-servations are described in section 2, and their results arein section 3. We describe the analysis in section4, and dis-cuss the nature of this anomalous object in section 5. Wesummarize our conclusions in section 6.
Our most extensive set of spectra is from the OhioState Multi-Object Spectrograph (OSMOS; Martini et al.2011) mounted on the 2.4 m Hiltner telescope at MDMObservatory, on Kitt Peak, Arizona. We obtained a se-quence of eighteen 720 s exposures of ASASSN-18aan on2019 Sept. 06 UT. The spectra span 4.12 hr from startto finish, and cover just over one orbit. The spectra coverfrom 3975 to 6890 ˚A, with 0.7 ˚A pixel − and 3.0 ˚A resolu-tion (full width at half-maximum). We derived a pixel-to-wavelength relation from Hg, Ne, and Xe lamp spectra, andused the [OI] and OH-band features in the night sky spec-trum to derive zero point offsets and linear stretch factorsthat we applied to the wavelength scales of the individualexposures. To convert to absolute flux, we observed spec-trophotometric standard stars in twilight when the sky wasreasonably clear. One can query these maps at http://argonaut.skymaps.info/
Our MDM time series photometry (Table 2) is all fromthe McGraw-Hill 1.3m telescope. In 2019 September weused an Andor frame-transfer camera on three nights; ex-posures were generally 30 seconds with almost no deadtime between exposures. In 2019 October we used a SITe1024 CCD detector cropped to 256 , which resulted in a ∼ λ < r filter. The same set of compari-son stars was used througout. The main comparison star,at α = 0 h m s . , δ = +62 ◦ ′ ′′ . V = 14 .
446 in APASS (Henden et al. 2011). We adjustedour differential magnitudes by this amount to convert themto rough V magnitudes. We performed a world-wide observational campaign viaVariable Star Network (VSNET) collaborations (Katoet al. 2004) and Optical and Infrared Synergetic Telescopesfor Education and Research (OISTER). All of our pho-tometric observations by VSNET and OISTER were de-scribed in barycentric Julian date (BJD). Detailed logsof photometric observations by VSNET and OISTER arelisted in table E2. Our intensive observations of the su-peroutburst of ASASSN-18aan by VSNET was started onDecember 4, 2018 (BJD 2458457), 4 days after the ASAS-SN’s detection.We also performed spectroscopic observations viaOISTER to confirm changes of spectral features throughthe superoutburst on 2018-12-10, 2018-12-24 and 2019-01-05 (UT). We could not observed standard stars due to abad weather and time limitation so not performed a sensi-tivity correction.Table 1 gives a summary of the OISTER observations.
The overall light curve of the superoutburst is shown infigure 1. We also plot photometric data from the ASAS-SN CV Patrol (Shappee et al. 2014; Kochanek et al. 2017;Davis et al. 2015) that constrains the onset of the superout-burst. The superoutburst lasted about 21 days during BJD2458452-2458473. The brightness first rapidly increasedduring BJD 2458448-2458452 and next slightly increasedduring BJD 2458452-2458462, reaching maximum at 15.2 ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 Table 1.
Summary of OISTER observations.
Telescope (Instrument) Date (UT) Exposure (Sec) Filter (Number of data)SaCRA (MuSaSHI) 0.55 m 2018-12-10 30 r (810), i (810), z (810)2018-12-14 60 r (206), i (206), z (206)2018-12-17 60 r (258), i (258), z (258)2018-12-19 60 r (488), i (488), z (488)2018-12-23 60 r (248), i (248), z (248)2019-01-07 60 r (115), i (115), z (115)Kiso (Tomo-e Gozen) 1.05 m 2018-12-10 7.5 No filter (1181)MITSuME Akeno (Tricolor camera) 0.5 m 2018-12-10 60 g (79), Rc (79), Ic (79)2018-12-12 60 g (105), Rc (105), Ic (105)2018-12-13 60 g (25), Rc (25), Ic (25)2018-12-14 60 g (106), Rc (106), Ic (106)2018-12-15 60 g (106), Rc (106), Ic (106)MITSuME Okayama (Tricolor camera) 0.5m 2018-12-08 60 g (52), Rc (52), Ic (52)Kanata (HOWPol) 1.5 m 2018-12-10 40 I (318)2018-12-13 40 I (74)2018-12-15 40 I (141)2018-12-18 40 I (108)2018-12-19 40 I (74)2018-12-27 60 V (82), I (69)2019-01-07 60 V (42), I (53)2019-01-13 60 V (65), I (66)2019-01-14 60 V (37), I (75)2019-01-15 60 V (32), I (41)2019-01-16 60 V (27), I (41)2019-02-02 60 V (62), I (61)Murikabushi (Tricolor camera) 1.05 m 2018-12-20 60 g (218), Rc (218), Ic (218)2018-12-24 60 g (97), Rc (97), Ic (97)Telescope (Instrument) Date (UT) Exposure (Sec) Cover range Resolution Number of dataNayuta (MALLS) 2.0 m 2018-12-10 1200 5130-7970 1800 162018-12-24 1200 4490-9650 800 32019-01-05 1200 4390-9550 800 3
Table 2.
Journal of MDM Time-Series Photometry
Start (UT) HA range Exp. (Sec) N Filter2019-09-05 07:38 − − − − − Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 mag, and then began to decrease. The main superoutburstended at BJD 2458473. There were two rebrightenings(during BJD 2458474-2458477 and 2458483-2458486) af-ter the main superoutburst. After the rebrightenings, themagnitude returned to near the quiescent state around 17.5mag. The superoutburst amplitude was ∼ For analysis, we first subtracted the global trend ofthe light curve of the superoutburst and rebrighteningsby subtracting a smoothed light curve obtained by lo-cally weighted polynomial regression (LOWESS: Cleveland1979) After the subtraction, we performed a phase disper-sion minimization (herafter PDM, Stellingwerf 1978) anal-ysis of whole superoutburst and found the orbital periodto be P orb = 0 . We removed all eclipses to investigate the variation of thesuperhump period during the superoutburst. We calcu-lated the time of eclipses from the above epoch and orbitalperiod and masked eclipses with a range of 0.11 in units oforbital phase.Figure 2 shows the O − C curve (upper panel), theamplitude of the superhumps (middle panel), and thelight curve (lower panel) of ASASSN-18aan during BJD2458458-58474. We determined the times of maxima ofsuperhumps in the same way as in Kato et al. (2009). Theresulting times are listed in table E3.From the variation of the superhump period and theamplitudes of superhumps, we regarded BJD 2458455.0-2458462.5 (0 ≤ E ≤
22) as stage A, BJD 2458462.5-2458467.6 (29 ≤ E ≤
56) as stage B, and BJD 2458467.6-2458473.5 (60 ≤ E ≤
93) as stage C superhumps.The PDM analysis of stage A superhumps (upper panel)and the phase-averaged mean profile (lower panel) areshown in figure 3. Single-peaked, asymmetric modula-tions are evident, with an amplitude of 0.3 mag. We foundthe stage A superhump period to be P stA = 0 . P dot ( ≡ ˙ P sh /P sh ), which is the derivative of the super-hump period during stage B, is 91(34) × − . This valueis quite large, but less reliable because of the large errorand uncertainty of the boundary of each superhump stagestemmed from small number of points. Figure 4 shows the phase-averaged profile folded on theorbital period. We divided the light curve of the su-peroutburst into three stages; superoutburst stage (BJD2458453-2458473.5) including a whole main superoutburst,rebrightening stage (BJD 2458473.5-2458487) includingtwo rebrightenings, and quiescent stage (BJD 2458487-2458520) which is after the rebrightenings. It is evidentthat the eclipse clearly seen at the superoutburst stagearound phase 0.0 becomes shallow at the rebrighteningstage, and then it can be hardly seen at the quiescentstage. This indicates the inclination i of this system issmall for an eclipsing system ( i ∼ ◦ ), and the eclipsesat the superoutburst and rebrightening stages are causedby the secondary star passing the front of the expanded,bright disk. It is clearly seen that the light curve at the su-peroutburst or rebrightening stage is affected by ellipsoidalvariations. We discussed the detail of these ellipsoidal vari-ations based on the MDM observation data as below. Figure 5 shows spectra through the superoutburst. Thespectra were obtained during the brightening state, justafter the first rebrightening, and in the quiescent state af-ter the second rebrightening. As described in section 2.3,although a sensitivity correction was not carried out in allobservations the profile transition can be seen by compar-ing each trace. The disk component rose at shorter wave-length during the superoutburst, while it became weak atthe quiescent state. In addition to this, the H α λ ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 v Fig. 1.
Light curve of ASASSN-18aan. The filled- squared points show the ASAS-SN’s V magnitude. The V-shaped sign represents an upper limit by ASAS-SN’s V and g magnitude, and the quiescent magnitude is shown as a red dashed line. stage A stage B stage C
Fig. 2.
Upper panel: O − C curve of ASASSN-18aan during BJD 2458458-58472. We used an ephemeris of BJD 2458461.485+0.1579 E for drawingthis figure. Middle panel: Amplitude of superhumps. Lower panel: Lightcurve. The horizontal axis in units of BJD and cycle number is common toall of these panels. (d) q P=0.16282 −0.5 0.0 0.5 1.0 1.5−0.2−0.10.00.10.2
Fig. 3.
Upper panel: θ -diagram of our PDM analysis of stage A superhumpsof ASASSN-18aan (BJD 2458455.0-2458462.5). The gray area representsthe 1 σ error of the best-estimated period by the PDM method. Lower panel:Phase-averaged profile of stage A superhumps. Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 −0.5 0.0 0.5 1.0 1.5−1.0−0.50.00.51.0 superoutburst stagerebrightening stagequiescent stage
Fig. 4.
Phase-averaged profiles of ASASSN-18aan during the superoutburststage (BJD 2458453-2458473.5; blue circles), rebrightening stage (BJD2458473.5-2458487; green triangles) and quiescent stage (BJD 2458487-2458520; red diamonds).
Fig. 5.
Average flux. The gray area represents a contamination of the at-mosphere or noise. The spectra shown are from 2018-12-10, 2018-12-24and 2019-01-05 (UT) from top to bottom, respectively. The trace for 2018-12-10 is shifted above by 1600 units for convenience. The spectra are notflux-calibrated (see section 2.3).
Figure 6 shows MDM light curves from 2019 Septemberand October, after the source had returned to quiesence.The light curves show a persistent modulation with twopeaks per orbit. This ellipsoidal variation indicates astrong contribution from a tidally-distorted secondary star.Such strong ellipsoidal variation is seldom observed in CVswith periods this short.Several features of the quiescent light curve are not con-sistent with a pure ellipsoidal variation. Most obviously,shallow eclipses appear around phase zero. The mean mag-nitude is about 0.1 mag fainter in 2019 October than inthe previous month, while the secondary star should benearly constant; evidently the disk faded over the interval
Fig. 6.
Light curves from the MDM 1.3 m telescope, folded on the eclipseephemeris, and repeated for a second cycle for continuity. between observations. In the September data, the peaksnear phase 0.75 are slightly brighter than those near 0.25,while in the October data the opposite is true; in pureellipsoidal variation they should be exactly equal. Also,there are subtle but clear breaks in the slope near phases0.4 and 0.6. These are clearest in the October data, wherethe mean light level is a bit fainter. The interval aroundphase 0.5 was covered on all three nights in the Octoberdata, so this feature is highly reproducible. We considerits interpretation later.Figure 7 gives a magnified view of the eclipse in the qui-escent light curves. The eclipse is only ∼ . The mean spectrum (figure 8) shows weak emission, butalso absorption features characteristic of a late-type star,as might be expected given the strong ellipsoidal modula-tion.To explore how the spectrum behaves through the or-bital cycle, we prepared a phase-averaged representation ofour spectral data. To start, we rectified each spectrum bydividing by its fitted continuum and computed the phaseof each spectrum using the photometric ephemeris. Wethen constructed a grid of equally-spaced phases, and ateach phase computed an average of the phase-adjacent rec-tified spectra, weighted by a narrow Gaussian centered on ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 Fig. 7.
Magnified light curves of the primary eclipses around phase 1.0.Other conditions are same as figure 6.
Fig. 8.
Average flux-calibrated spectrum from 2019 Sept. the fiducial phase. Figure 9 shows a portion of the two-dimensional image formed by stacking these averages. Theabsorption lines vary strikingly in velocity on the orbitalperiod, conclusively demonstrating that they are from thesecondary star rather than some other object. Emissionis visible at H α λ α absorption line, which reveals itselfthrough its velocity modulation.Note that the emission lines at H α and He I λλ , fxcor ; Tonry,Davis (1979) describe the principle. The template was acomposite of 76 spectra of later-type IAU velocity stan-dards that had been shifted to zero apparent radial veloc- Fig. 9.
Phase-resolved spectrum. Note the radial velocity variation of theabsorption lines. ity before averaging. The correlations excluded the regionnear the sodium D lines because of possible confusion withHe I λ v to the reference frame of the solar systembarycenter and fitted them with a sinusoid of the form v ( t ) = γ + K sin (cid:20) π ( t − T ) P (cid:21) , (1)where t is the barycentric time of mid-integration. Wefixed P to the eclipse period, and found γ = 7 ± − (2) K = 273 ± − , (3) T = BJD 2458732 . ± . , and (4) σ = 13 km s − , (5)where the time base is UTC and σ is the standard de-viation around the best fit. Allowing P to float yielded P = 0 . E ( B − V ) = 0 .
36 (see theIntroduction). We then used an interactive program tosubtract spectra of different (known) spectral types andnormalizations from this rest-frame average, with the aimof cancelling the stellar absorption features and leaving thesmooth continuum characteristic of accretion-disk spectrabetween the emission lines. Some of the template spectrawe used were from our own observations of stars classifiedby Keenan, McNeil (1989), and others were originally pub-lished by Jacoby et al. (1984). Figure 11 shows the bestresult, in which the subtracted spectrum is of the G9V starHD29050 (from Jacoby et al. 1984) scaled to an unreddened flux equivalent to V = 16 .
3, which would be V = 17 . Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0
Fig. 10.
Absorption-line velocities folded on the orbital ephemeris, with thebest-fit sinusoid superposed. All points are repeated over a second cycle forcontinuity.
Fig. 11. (Upper trace): Rest-frame average spectrum, dereddened by 1.19magnitudes. (Lower trace): The same, after subtraction of a scaled spectrumof a G9V-type star. The solid vertical lines mark the wavelengths of sodiumfeatures near λ and the very strong NaD doublet near λ . We estimated the mass ratio of ASASSN-18aan in thesame way as proposed in Kato, Osaki (2013) by using thefractional superhump-period excess for the 3:1 resonance, ε ∗ = 1 − P orb /P stA . We calculated ε ∗ = 0 . q = 0 . r p , where r isthe disk radius and p is the exponent. Under the assump-tion of the accretion disk to be the black body and steadyin time, p = − /
4. Since we subtracted the component ofthe primary from the light curve in advance, we assumedthe disk has an inner rim. By using the white dwarf massestimated in section 4.2 and 4.3, we estimated the whitedwarf radius (equal to the inner rim of the disk) to be R inner = 0 . a , where a is a binary separation (Nauenberg1972). The secondary fills its Roche lobe.We set the disk radius r , exponent p and inclination i ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 as free parameters and estimated them by Markov ChainMonte Carlo (MCMC) method. We use the mass ratio, q = 0 . i . At themaxima or minima of superhumps it is difficult to makebetter phase-averaged profile from a few neighboring, con-tinuously changing superhumps, leading to incorrect esti-mation. We thus selected three excellent eclipses whichexist on the hillside of superhumps and have almost lin-ear wings, using eclipse 9a, 9b and 52 in table E1. Weestimated the disk radius, exponent and inclination si-multaneously, and derived the value of inclination to be76 . +0 . − . ◦ , . +0 . − . ◦ , . +0 . − . ◦ (errors are calculatedfrom the 95% credible intervals), and set the inclination tobe 76 . ◦ . Hereafter, in this section we use this valueas the inclination and fixed in our estimation.Figure 12 shows the result of our estimation. We an-alyzed 48 eclipses through the superoutburst and two re-brightenings. We did not investigated the eclipse in quies-cence because of its shallow depth and low signal-to-noiseratio. We additionally note that during the rebrighteningswe obtained limited-quality data on only five eclipses.The effective radius of the Roche lobe (Eggleton 1983),the tidal truncation radius R tidal /a = 0 . / (1+ q ) (approx-imated equation derived from Paczy´nski (1977)), the 3:1resonance radius and the circularization radius (Hessman,Hopp 1990) are indicated in Figure 12. The disk radiusseems to shrink through the superoutburst. Here notethat the disk radius mainly corresponds to the width ofthe ingress and egress of the eclipse, and thus the effectof subtracting the ellipsoidal variations and superhumpsis rather small, implying the estimated values are fairlyreliable. On BJD 2458457, 5 days after the onset of thesuperoutburst inferred from the ASAS-SN observation, thedisk radius is around 0 . a , decreasing almost linearly andthen reaching around 0 . a . The radius at the first andsecond rebrightening is near the circularization radius. Atendency with difference of color are clearly seen; eclipsesobserved in the B-band filter correspond to small radii,those in Ic or Rc-band are wider corresponding to largeradii, and ones by C or V-band filter are intermediate be-tween these two. This results are good agreement with thebasic picture of the accretion disk, which should be hottestin the inner part and cooler at larger radii.The disk radius is largest at the onset of the superout-burst, and comparable to the 3:1 resonance radius. Toinvestigate this, we performed a linear regression analysisto the data only including eclipses by C and V-band fil- Roche lobe radiusTidal truncation radius3:1 reconance radiusCircularization radius
CBVRcIcgri
Fig. 12.
Upper panel: Light curve. The filled-square points show the ASAS-SN’s V magnitude. Middle panels: Radius in units of binary separation.Bottom panel: Exponent. All error bars are calculated from 95 % confidenceintervals of the posterior probability. The horizontal axis in units of BJD iscommon to all of these panels. The points labeled “Ic” include data fromboth the I and Ic passbands. ter in main superoutburst, and estimated the disk radiusat the onset of the superoutburst to be 0 . a . This valueis larger than the 3:1 resonance radius, 0 . a , indicatingthat the superoutburst of ASASSN-18aan may be causedby indeed the tidal instability.The power-law brightness exponent p strongly affectsthe eclipse depth. Our estimated values seem to be con-centrated around − .
5, being small compared with thevalue assuming the steady-state black body, − .
75. Thismight be an artifact of subtracting the ellipsoidal varia-tions and superhumps, which may artificially reduce theeclipse depth. In addition, the MDM observations showshallow eclipses in quiescence at the phase 0.96-1.04 (setphase 1.0 to the mid eclipse). However, the primary eclipsewith the radius of R WD /a = 0 .
011 occurs at the phase 0.98-01.02, indicating the eclipses in quiescence include some-what disk components. Our estimated values thereforemight be systematically small because of oversubtraction. Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0
Fig. 13.
Constraints from K (vertical curves) and the eclipse width(downward-sweeping curves) in the i – q plane. We further constrain the system parameters using ourmeasured velocity amplitude for the secondary star, K .Figure 13 shows the constraints in i - q plane. The solid,roughly vertical curves are for various assumed white dwarfmasses, M WD , for our best fit value of K . The dashedand dotted lines flanking the leftmost (blue) curve illus-trate the effect of a generous 10 km s − uncertainty in K . The curves sweeping down from the upper left arethe constraints for different eclipse half-widths, ∆ φ / .The constraint on q implies M WD ∼ . ⊙ , so that M = qM WD ∼ .
18 M ⊙ . The white dwarf mass is similarto average field white dwarfs, and on the low side com-pared to most CVs, for which Zorotovic et al. (2011) givean average M WD = 0 . ± .
23 M ⊙ , where the standarddeviation is for the distribution rather than the standarddeviation of the mean. The secondary mass is much lessthan that of a main-sequence G9 star; a main-sequencestar of such low mass would have a spectral type aroundM4 or M5 (Pecaut, Mamajek 2013). We built a model of the ellipsoidal variations using a codedeveloped by Thorstensen, Armstrong (2005), which mod-els only the contribution from the Roche-lobe-filling sec-ondary star. The code computes the Roche lobe’s physicalsize and shape for assumed primary and secondary masses,and the surface brightness of each element as viewed fromearth including gravity and limb darkening, using empir-ical surface brightnesses appropriate to the spectral band and surface temperature. After computing the bright-ness at each phase, the code adjusts the normalization byfinding the distance for which the predicted light curvematches the data. We used the 2019 October MDM lightcurve, which has the smallest disk contribution and excel-lent night-to-night reproducibility, scaled the broad-bandpoints to approximate V magnitudes as described above,and subtracted 1.19 mag to correct for extinction. Pointsthat appeared to be affected by the primary or secondaryeclipses were masked. We set the secondary’s T eff to5300 kelvin, appropriate for its G9 type (Pecaut, Mamajek2013), as well as M WD = 0 . M = 0 .
18 M ⊙ and i = 77 ◦ .To the light curve we added a non-variable cwomponentequivalent to F λ = 5 × − erg cm − s − ˚A − , about thesame as the disk contribution shown in Figure 11,Figure 14 shows the result. We found we could notmatch the deep minima around phase 0.5 without a sec-ondary eclipse. Also, without the extra light from the disk,the predicted amplitude of the ellipsoidal variation at thesystem’s inclination was too great. We have one other bitof evidence that supports the secondary-eclipse interpre-tation: referring back to figure 9, we notice transient ab-sorption reversals near phase 0.5 in He I λλ α absorption. This strongly suggeststhat the secondary’s light is passing through a substantialcolumn of high-excitation gas at those phases, as might beexpected in the atmosphere of the disk.The distance obtained by scaling the ellipsoidal varia-tion model to the observed light curve is 624 pc, which, inview of the long chain of reasoning that went into the choiceof parameters, agrees well with the Gaia DR2 parallax de-termination (675 +32, −
29 pc). The overall picture is re-markably self-consistent – M WD ∼ . M ∼ . M ⊙ ,a secondary with T eff above ∼ i ∼ ◦ . As mentioned in section 4.1, the disk radius at the onsetof the superoutburst seems to reach the 3:1 resonance ra-dius. This indicates the superoutburst of ASASSN-18aanis indeed caused by the tidal instability in the accretiondisk, and thus the upper limit of critical mass ratios ofthe 3:1 resonance would be larger than q = 0 . P orb objects also may be caused by the tidal instabil-ity at the 3:1 resonance radius. We therefore investigatedSU UMa-type DNe that have long orbital periods and themass ratios are (even partially) estimated. We summarizedthe results in table 3. There are some objects beyond the ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 Fig. 14.
Light curve from 2019 October. The red curve shows the model ofthe ellipsoidal variation described in the text. The larger blue points wereused in the normalization, and the smaller black points near phase 0.5 andphase 0 were ignored. The modeled light curve has been extinguished by A V = 1 . mag to match the observed points. upper limit of the tidal instability. Most of them, how-ever, are less reliable values. First of all, in almost halfof the listed objects, mass ratios are estimated with spe-cific ranges. Objects except TU Men and NY Ser havelower mass ratio than the upper limit of 0.25 (Whitehurst1988) or 0.33 with reduction of mass transfer (Murray et al.2000).Kato et al. (2019b) estimated a referential value of massratio of NY Ser from the superhump period of the post-superoutburst stage as proposed by Kato, Osaki (2013).In this method, the mass ratio is calculated by assumingignorance of the pressure effect in the accretion disk, whichconditions are considered to be fulfilled during the stageA and post-superoutburst stage, and using the orbital pe-riod, superhump period and disk radius. The range of theestimated mass ratio of NY Ser arose from the supposeddisk radius. The disk radius at the post-superoutburst,however, is uncertain (Kato, Osaki 2013), and thus themass ratio is quite indistinct.TU Men has seriously large mass ratios for the tidalinstability. Stolz, Schoembs (1984) derived the mass ra-tio to be q = 0 .
59 from spectroscopic observations duringthe 1980 superoutburst. Mennickent (1995) estimated themass ratio to be q = 0 . α emission line inquiescent state. He also reported second candidate of itsmass ratio to be q = 0 .
33. Smak (2006) reanalyzed theradial velocity measured by Mennickent (1995) and calcu-lated the mass ratio to be q = 0 . q > . P orb SU UMa-type DNe,excepting TU Men. Kato et al. (2017) indicated that abovethe period gap, DNe on the standard CV evolutionarytrack hardly experience the 3:1 resonance; long- P orb SUUMa-type DNe seem not to be on the standard track. Asdiscussed in section 5.4, ASASSN-18aan has anomalouslywarm secondary, which evidently puts it on an unusualevolutionary track. This is also the case in ASASSN-14ho,another long- P orb SU UMa-type DNe having a warm sec-ondary (Gasque et al. 2019). These results coincide withthe argument by Kato et al. (2017).
Paczy´nski (1977) calculated the restricted three-bodyproblem and derived the last non-intersecting orbit aroundthe primary assuming an inviscid, pressure-free disk. Thislast non-intersecting orbit is considered to be the largestradius of a disk, known as the tidal truncation radius.Several authors have considered the tidal influence at theouter rim of the disk, and obtained nearly identical results(Ichikawa, Osaki 1994; Papaloizou, Pringle 1977; Truss2007). There is, however, some room for discussion aboutthe tidal truncation radius, especially in extremely low- q systems. Lin, Papaloizou (1979) showed that at extremelysmall mass ratios, the outer rim is not truncated and ex-tends beyond the primary Roche lobe. Hellier (2001) sug-gested the multiple rebrightenings called echo outbursts inER UMa systems cause a weak angular momentum dis-sipation in the hot, eccentric disk due to the extremelylow mass ratio. The cause of the discrepancy between theobservation and theoretical predictions in WZ Sge-typeDNe is unknown. The condition that the 3:1 resonanceradius lies inside the tidal truncation radius is realized inthe case of q < .
25, being consistent with observations,most of SU UMa-type DNe having lower than this value. Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0
Table 3.
List of long P orb SU UMa-type DNe candidates which the mass ratios are estimated.
Object P orb ∗ P stB ∗ q ReferencesSDSS J162520.29+120308.7 0.09113(30) 0.09604(3) 0.21(1) 1NY Ser 0.09744(5) 0.10458 0.23-0.43 2, 3CzeV404 0.0980203(6) 0.10472(2) † † † ∗ In units of day. † Averaged period.1.Montgomery et al. (2017); 2. Kato et al. (2019b); 3. Pavlenko et al. (2014); 4. B¸akowska et al. (2014); 5. Kato et al. (2016)6. Pavlenko et al. (2018); 7. Pavlenko et al. (2010); 8. Kato et al. (2014a); 9. Mennickent (1995); 10. Smak (2006)11. Kato et al. (2017) ; 12 Gasque et al. (2019)On the other hand, the condition that the 2:1 resonanceradius lies inside the tidal truncation radius holds onlyif q < . q = 0 . P orb SU UMa-type DNe the mass ratiosseem to be under or near the upper limit, and thus the3:1 resonance radius likely lies in the tidal truncation ra-dius. This means the tidal dissipation at the tidal trunca-tion radius acts effectively and the expansion of the disklimited by this radius. However, when the mass ratio be-comes small and the tidal torque by the secondary alsoweakens, the 2:1 resonance radius is the last position thatthe angular momentum of the disk is effectively extracted.Paczy´nski (1977) showed that the range of unstable orbits,which first appear around q = 0 .
22, shrinks in extremelylow- q systems. This results may indicate that this unstable orbits are unlikely to work as the point of tidal limitationfor extremely low- q systems such as WZ Sge-type DNe. Ifso, it is a reasonable probability that the tidal truncationradius in such extreme systems is no longer a hard limit toterminate disk expansion. We regarded the duration of stage A as BJD 2458455.0–2458462.5 (22 cycles) from the long, uniform superhumpperiod and the amplitude. Considering a delay of startof our observations, the duration of stage A is somewhatlonger. Judged from the ASAS-SN data, the duration ofstage A superhumps is at most about 10 days. The growthtime of superhumps is rather close to the case of WZ Sge-type DNe (Kato 2015).Rebrightenings after the main superoutburst are alsothe intrinsic feature of WZ Sge-type DNe. The rebrighten-ings of ASASSN-18aan are similar to type-B rebrightening,which shows two or more repetitive short brightenings aftera rapid decline of main superoutburst (Kato 2015; Kimuraet al. 2016).Most recently, the similarity of superoutbursts in long- P orb SU UMa-type DNe to those in WZ Sge-type one wasfound in several objects (Antipin, Pavlenko 2002; Mrozet al. 2013; Kato et al. 2016; Kato et al. 2017; Kato et al.2019a; Kato et al. 2019b; Kato 2020) Kato et al. (2016)proposed that such a behavior like WZ Sge-type superout-bursts is responsible for a difficulty in maintaining the tidalinstability due to the weakness of tidal torque at the 3:1resonance radius which lies close to the tidal truncationradius. Kato et al. (2019a) also found that the growth of ublications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0 the 3:1 resonance in systems with the stability border ofthe 3:1 resonance seems to be slow.In the case of ASASN-18aan, the characteristics ofthe superoutburst are consistent with other long- P orb SUUMa-type DNe. The similarity to the WZ Sge-type DNein such a long- P orb SU UMa-type DNe would be a key foran investigation of the effect at the tidal truncation radius.
A 0.18 M ⊙ main-sequence star will have a late M spec-tral type, dramatically cooler than the secondary spec-trum we observe here. Mass loss evidently causes CV sec-ondaries to depart from the main sequence mass-radius-temperature relation (Knigge et al. 2011), but the empiri-cal donor sequence of Knigge (2006) shows CV secondariesnear P orb = 3 . much warmer thanexpected, for both its determined mass and the orbital pe-riod. ASASSN-18aan joins a small group of CVs known tohave anomalously warm secondary stars. We summarizedthem in table 4.An apparent enhancement of the sodium absorptionline can be seen in the averaged spectrum after subtrac-tion of a scaled spectrum of G9V-type star in figure 11.Enhanced sodium absorption is also seen in other warmer-than-expected systems, such as QZ Ser (Thorstensen et al.2002a) and CSS J134052.0+151341 (Thorstensen 2013).Thorstensen et al. (2002a) suggested that this indicates ahistory of significant nuclear evolution of the secondarybefore mass transfer occurred; a side chain of hydro-gen burning via CNO cycle, Ne( p, γ ) Na, leads to en-hanced sodium. CV secondary stars in general departsomewhat from main-sequence mass-radius-luminosity re-lations, but the departures in these warm-secondary sys-tems are much larger. Nuclear evolution prior to masstransfer can account for this for these gross discrepancies,as well, since the strong admixture of fused helium leadsto a mean molecular weight very different from the mainsequence. Evolutionary scenarios for these systems are ex-plored briefly by Thorstensen et al. (2002b).
We report photometry and spectroscopy of the eclipsingSU UMa-type DN ASASSN-18aan in its 2018 outburst andfollowing quiescence. The superoutburst showed a slowgrowth rate of superhumps and rebrightenings, character-istics similar to that of WZ Sge-type DNe. We derived theorbital period to be P orb = 0 . P orb SUUMa-type DNe. We also estimated the mass ratio usingthree different methods and derived nearly identical val-ues; from the superoutburst we find q = 0 . P orb SU UMa-type DNe also could be explainedby the tidal instability model, without the need for otherhypotheses. By considering long- P orb WZ Sge-type DNe(Wakamatsu et al. 2017), we further suggest that the tidaltruncation radius is effective at larger mass ratios suchas those found in SU UMa-type DNe, but in extremelylow- q systems it is no longer viable; in those cases theoutbursting disk may reach the 2:1 resonance, which liesoutside the tidal truncation radius.One of our most striking results is that the secondary ismuch warmer than typical for CVs of this orbital period.Its spectral type is G9, while typical secondary stars at P orb = 3 . ASASSN-18aan is an extremely unusual CV. The vast ma-jority of SU UMa-type DNe have P orb less than about twohours, while the period we find is nearly twice that long.The period is derived from eclipses, and corroborated by alarge modulation in the secondary star’s radial velocities,so it is unquestionably orbital, and not due to some otherphenomenon. In addition, the secondary star is highlyanomalous; it is much hotter than typical CV secondariesin this period range, and is furthermore much less mas-sive than its surface temperature would suggest. Thesegross peculiarities are very likely the consequence of masstransfer that began after the secondary star had undergonesignificant nuclear evolution; we are seeing, in effect, thestripped core of a much larger star. ASASSN-18aan joinsa small number of similar systems already known.Decades of careful study of superhumping systems haveyielded a rich empirical picture of the superhump phenom-ena, which has spawned a correspondingly detailed body oftheory. An anomalous sytem such as this one presents anopportunity for critical tests of such theories in relativelyunexplored regions of parameter space. Publications of the Astronomical Society of Japan , (2018), Vol. 00, No. 0
Table 4.
List of CVs with anomalously warm secondary.
Object P orb ∗ M † M † Spectraltype Spectral type pre-dicted from P orb ‡ ReferencesEI Psc (=1RXSJ232953.9+062814) 0.044567(3) 0.7 0.13 K4 ± ± ± ± ± ∗ In units of day. † In units of M ⊙ . ‡ Knigge (2006).1. Thorstensen et al. (2002b); 2. Thorstensen et al. (2002a); 3. Littlefair et al. (2006); 4. Thorstensen (2013);5. Thorstensen (2015); 6. Gasque et al. (2019)
Acknowledgement
This work was supported by the Optical and Near-infraredAstronomy Inter-University Cooperation Program and theGrants-in-Aid of the Ministry of Education. This researchmade use of the AAVSO Photometric All-Sky Survey(APASS), funded by the Robert Martin Ayers SciencesFund and NSF AST-1412587. JRT thanks the MDM stafffor observing support, and the Tohono O’odham Nationfor allowing their mountain to be used for research intothe sky that surrounds us all. S.S. was supported bythe Slovak Research and Development Agency under thecontract No. APVV-15-0458, by the Slovak Academyof Sciences grant VEGA No. 2/0008/17 and was par-tially supported by the Program of Development of M. V.Lomonosov Moscow State University “Leading ScientificSchools”, project “Physics of Stars, Relativistic Objectsand Galaxies”. We are grateful to the All-Sky AutomatedSurvey for Supernovae (ASAS-SN) project for detectingthe superoutburst of ASASSN-18aan. YW is grateful toH. Kimura at Higashi-Hiroshima Observatory for collabo-ration on observation. YW deeply thanks the referee Dr.Elena P. Pavlenko for her quick review for my doctoraldissertation.
Supplementary figure E1 and tables E1-E3 are reported inthe online version.
Appendix
Our developed modeling method of eclipses during out-bursts uses Metropolis algorithm (Metropolis et al. 1953),which is a kind of MCMC (for detail, see, e.g., Sharma(2017)).We assumed observed magnitudes following a normaldistribution with specific errors. The likelihood function isthus as follows: L = n Y i =1 p πσ i exp (cid:20) − ( x i − y i ) σ i (cid:21) , (6)where n , x i , y i and σ i are a number of data, observed flux,flux of the synthesized light curve and the photometricerror at i -th data of the eclipse, respectively. We practi-cally used a log-likelihood function. We set the map sizeat 201 × σ i = 0 .