Confirmation of a Second Propeller: A High-Inclination Twin of AE~Aquarii
Peter Garnavich, Colin Littlefield, R. M. Wagner, Jan van Roestel, Amruta D. Jaodand, Paula Szkody, John R. Thorstensen
SSubmitted for publication in the Astrophysical Jorunal, February 8, 2021
Typeset using L A TEX twocolumn style in AASTeX63
Confirmation of a Second Propeller: A High-Inclination Twin of AE Aquarii
Peter Garnavich, Colin Littlefield,
1, 2
R. M. Wagner, Jan van Roestel, Amruta D. Jaodand, Paula Szkody, and John R. Thorstensen University of Notre Dame, Notre Dame, IN 46556, USA Department of Astronomy, University of Washington, Seattle, WA 98195, USA Department of Astronomy, Ohio State University, Columbus, OH 43210, USA Departments of Astronomy and Physics, California Institute of Technology, Pasadena CA 91125, USA Department of Physics and Astronomy, Dartmouth College, Hanover, NH 03755 USA
ABSTRACTFor decades, AE Aquarii (AE Aqr) has been the only cataclysmic variable star known to contain amagnetic propeller: a persistent outflow whose expulsion from the binary is powered by the spin-downof the rapidly rotating, magnetized white dwarf. In 2020, LAMOST J024048.51+195226.9 (J0240)was identified as a candidate eclipsing AE Aqr object, and we present three epochs of time-seriesspectroscopy that strongly support this hypothesis. We show that during the photometric flares notedby Thorstensen, the half-width-at-zero-intensity of the Balmer and He I lines routinely reaches amaximum of ∼ − , well in excess of what is observed in normal cataclysmic variables. Thisis, however, consistent with the high-velocity emission seen in flares from AE Aqr. Additionally, weconfirm beyond doubt that J0240 is a deeply eclipsing system. The flaring continuum, He I and muchof the Balmer emission likely originate close to the WD because they disappear during the eclipsethat is centered on inferior conjunction of the secondary star. The fraction of the Balmer emissionremaining visible during eclipse has a steep decrement and it is likely produced in the extended outflow.Most enticingly of all, this outflow produces a narrow P Cyg absorption component for nearly half ofthe orbit, and we demonstrate that this scenario closely matches the outflow kinematics predicted byWynn, King, & Horne. While an important piece of evidence for the magnetic-propeller hypothesis—arapid WD spin period—remains elusive, our spectra provide compelling support for the existence of apropeller-driven outflow viewed nearly edge-on, enabling a new means of rigorously testing theories ofthe propeller phenomenon. Keywords:
Cataclysmic variable stars: Intermediate Polars; White dwarf stars; Magnetic stars; Eclips-ing binary stars; AE Aquarii INTRODUCTIONIntermediate polars (IPs aka DQ Herculis stars) area type of cataclysmic variable star (CV) consisting of aa cool, non-degenerate companion transferring mass toa magnetized white dwarf (WD). In an IP, the spin pe-riod of the WD is significantly shorter than the binaryorbit, so that the mass transfer is complicated by thechanging aspect of the WD magnetic field relative tothe source of the donated gas (Patterson 1994). If themagnetized WD is spinning sufficiently fast, donated gasmay be ejected from the system via a “magnetic pro-peller” mode (Wynn et al. 1997), in which the rapidlyrotating magnetosphere of the WD acts as a centrifugalbarrier that inhibits accretion onto the WD. The mag-netic propeller mechanism is believed to operate over awide range of accreting systems, including neutron stars and young stellar objects (Campana et al. 2018), butin WD systems, the propeller mode appears extremelyrare.Until now, the only confirmed IP in a propeller modehas been AE Aqr (Eracleous & Horne 1996), which dis-plays unique photometric and spectroscopic properties.Meintjes et al. (2015) review the major observationaland theoretical studies of AE Aqr.Recently, Thorstensen (2020) pointed to the flar-ing light curves and spectral properties of LAM-OST J024048.51+195226.9 (J0240 hereafter) as similarto AE Aqr and suggested that it might be only the sec-ond IP in the propeller mode. Littlefield & Garnavich(2020) analyzed its long-term light curve from sky sur-veys that suggested that the secondary star eclipses the a r X i v : . [ a s t r o - ph . S R ] F e b location of the flaring activity, implying a high orbitalinclination.Here, we present fast cadence, time-resolved spec-troscopy and photometry to compare J0240 with theobservational properties of AE Aqr. DATA2.1.
Spectroscopy
We observed J0240 with the Large Binocular Tele-scope (LBT ) and twin Multi-Object Dual Spectro-graphs (MODS; Pogge et al. 2012) on 2020 September12, October 13, and November 20 (UT). The spectro-graphs were set up in dual grating mode providing wave-length coverage from 320 nm to 1.01 µ m, with a dichroicsplitting the light into the red and blue optimized armsat 565 nm. A 0.8 arcsec slit was employed on all thenights giving a spectral resolution of 1350 near H α . Thetypical seeing during the observations was between 0.8and 1.0 arcsec. The sky was clear and seeing was steadyaround 1.0 arcsec for the September and October runs.In November, there were some clouds and the seeingvaried between 1.2 and 1.5 arcsec.The four MODS spectrographs (SX/MODS1,DX/MODS2 plus the red and blue arms for both) wererun independently and each have slightly different read-out and overhead costs. So, despite all the spectrographsset to take 180s integrations, the start times for a longseries of exposures soon become unsynchronized. Thisresults in an improved temporal resolution when datafrom the two telescopes are combined. On the Septem-ber night, 135 red exposures were obtained covering4.7 hours while on the October, 72 red-side exposureswere taken covering 2.5 hours. The average cadence was2 minutes for the red spectra. Serendipitiously, the tworuns combined covered an entire 0.306 day binary orbit.The November run generated 75 red-side spectra thatcovered a similar orbital phase as the October data.The two-dimensional CCD images were processed andone-dimensional spectra extracted using a 1.0 arcsecaperture. The spectra were wavelength calibrated us-ing neon, argon, and mercury emission lamps. To cor-rect for flexture over the time series, the wavelengthscale was adjusted slightly to shift the airglow emis- The LBT is an international collaboration among institutions inthe United States, Italy and Germany. LBT Corporation part-ners are: The University of Arizona on behalf of the ArizonaBoard of Regents; Istituto Nazionale di Astrofisica, Italy; LBTBeteiligungsgesellschaft, Germany, representing the Max-PlanckSociety, The Leibniz Institute for Astrophysics Potsdam, andHeidelberg University; The Ohio State University, representingOSU, University of Notre Dame, University of Minnesota andUniversity of Virginia. sion lines to their rest wavelengths. The J0240 spectrawere flux calibrated using the spectrophotometric stan-dard star BD+28 ◦ B , V , and R mag-nitudes were estimated by averaging the flux densitiesover the Johnson-Cousins bandpasses, and the R -bandbrightness is shown in the right panel of Figure 1.2.2. Photometry
We obtained time-resolved photometry with the MDM2.4m Hiltner telescope and OSMOS camera on fivenights in late 2020 and early 2021. A log of the ob-servations is given in Table 1. The first run coverednearly an entire binary orbit with an average time be-tween exposures (the cadence) of 16.5 s. The observa-tions in 2021 January were timed to cover two hours cen-tered around inferior conjunction with a slightly longercadence of 20.4 s. The data were obtained through aJohnson B filter and calibrated assuming that a com-parison star 91 (cid:48)(cid:48) E and 25 (cid:48)(cid:48)
N of J0240 has an apparentbrightness of B = 16 . ± . mag.Very high cadence photometry was also obtained withthe Palomar 200-inch telescope (P200) and CHIMERAfast photometer (Harding et al. 2016) on 2020 November13 and 14 (UT). Because the frame-transfer camera hadzero readout time, the cadence matched the 3 s expo-sure time. Images were obtained simultaneously throughSloan g (cid:48) and i (cid:48) filters. The length of the first time serieswas nearly an hour and the next night covered 1.6 hours.The images were bias subtracted and divided by twilightflat fields using the standard CHIMERA pipeline . Weused a variable aperture applied to the target and ref-erence star of 1.5 times the seeing width to generate adifferential light curve of the target. ANALYSIS3.1.
Photometry
The original propeller system, AE Aqr, shows a WDspin period of 33 s (Patterson 1979), so we might expectthe J0240 WD to have a similarly fast rotation rate.We obtained high cadence photometry of J0240, withthe goal of detecting the spin modulation. Thorstensen(2020) found no evidence of a spin period in his 23.3 scadence photometry. For each night’s time series, wecomputed the Lomb-Scargle power spectrum up to theNyquist frequency of the sampling, but we also didnot detect any significant candidate spin periods in thepower spectra. Because the Palomar photometry wasobtained on consecutive nights, we combined them into From the APASS catalog (Henden et al. 2015) https://github.com/caltech-chimera/PyChimera Figure 1. left:
Trailed spectra of J0240 around H α and He I 6675 ˚A. A full binary orbit is shown and the sequence isrepeated. Several absorption lines from the secondary are visible in the continuum. The full orbit is constructed by combiningthe September and October runs resulting in discontinuities in the emission profiles at phases 0.50 and 0.84. right: The R -bandlight curve synthesized from the individual spectra. The star was brighter and flaring more active during the October run. one dataset and performed the Lomb-Scargle analysis,but as before, there was no sign of coherent variability.To estimate a limit on the amplitude of any spin sig-nal, we removed the low-frequency variations in theP200 data and randomized the magnitude measure-ments before calculating a power spectrum. We re-peated this process 500 times, recording the maximumpower between frequencies 0.7 < f < . σ level in the combined g -band P200 data. We concludethat any spin modulation with a period between 6.3 sand 85 s must have an amplitude below 4 mmag duringour observation.Because AE Aqr flares can display relatively high-amplitude quasi-periodic oscillations (QPOs) close tothe spin period (Patterson 1979), we also created two-dimensional power spectra of each light curve in orderto search for intermittent periodic variations during the flares (Figure 2). However, we do not see any evidenceof a QPO. 3.2. Spectroscopy
Flares
While collecting the data it was obvious that the starwas varying rapidly in brightness, color, emission lineflux and line width. The synthetic R -band magnitudevariations shows the flaring pointed out by Thorstensen(2020) that is consistent with that seen in AE Aqr.The H α emission profile over a full binary orbit isshown in Figure 1, although the orbit was constructedfrom two runs separated by a month. Over the Septem-ber run, the system slowly brightens from inferior con-junction (orbital phase φ = 0 . φ = 0 .
40. The width of the H α emission line narrows asthe flare fades. The continuum reaches a deep minimum Table 1.
Photometric ObservationsDate Start Time Length Cadence a Filter Telescope(UT) (UT) (hours) (s)2020-11-11 04:01 6.87 16.5 B MDM2020-11-13 07:09 0.93 3.0 g (cid:48) , i (cid:48) P2002020-11-14 03:40 1.60 3.0 g (cid:48) , i (cid:48) P2002021-01-07 03:51 1.96 20.4 B MDM2021-01-08 01:50 2.00 20.4 B MDM2021-01-10 05:03 2.15 20.4 B MDM2021-01-11 03:15 2.02 20.4 B MDM a Cadence is the average time between consecutive exposures. r e l . f l u x c y c l e s / m i n hours UTC Figure 2.
The MDM and Palomar light curves, with their two-dimensional power spectra. The smooth red curves weresubtracted from the observed data (black points) prior to computation of the power spectra. There is no convincing evidenceof a stable period in any of these light curves. around superior conjunction ( φ = 0 .
5) that is likely dueto a lack of flaring combined with the ellipsoidal shapeof the secondary. The September
LBT sequence transi-tions to the October run at φ = 0 .
5. Flaring in J0240was more active in October than in September, display-ing many overlapping events and a bright continuum.In AE Aqr, the highest emission velocities duringflares reach ± − (Welsh et al. 1998). In com-parison, emission at the peak of the flares for J0240extends to ± − . These high velocities lastonly a few spectra, or approximately 10 min. In iso-lated flares (see Figure 1), the continuum flux fades morequickly than the emission line flux, a behavior also seenin AE Aqr (Welsh et al. 1998). Balmer emission is visible throughout the orbit, evenduring periods of relatively low activity. During quies-cent periods between flares, H α is seen with a substantialvelocity width of 1200 km s − HWZF. Around inferiorconjunction the H α emission width reaches its minimumvalue of 980 km s − HWZF. From this behavior we inferthat some level of flaring activity is visible at all times,except when the flaring region is obscured by the sec-ondary star at inferior conjunction.During the September sequence, a fairly isolated flareoccurred that allows us to examine its evolution. Wesubtracted the H α profile obtained immediately beforethe onset of the flare that began at orbital phase 0.40and the results are displayed in Figure 3). As notedabove, the high-velocity wings are seen early in the flare Figure 3.
The evolution of the H α emission profile duringa September flare. The average spectrum just before theinitiation of the flare was subtracted from each subsequentspectrum. The evolution is divided into three panels to showthe changes clearly. The top panel displays the rapid rise ofthe flare and the development of the high-velocity wings. Redand blue-shifted emission is seen out to 3000 km s − duringthe rise, but the redshifted emission fades as the blue sidecontinues to strengthen. Over the next 8 minutes (centerpanel), the H α profile remains fairly constant except for theloss of the highest velocity wings. A narrow absorption isseen over the flare. Its blueshift increases from 300 km s − to 400 km s − over 25 minutes. The lower panel shows theemission narrowing while the flux remains well above thepre-flare level. and fade quickly, on the time-scale of the continuum de-cline. The emission profile is always asymmetric withthe red side twice the flux of the blue. This asymmetrysuggests that the blue-shifted emission is self-absorbedin an expanding wind. Some asymmetry in the lineshape is seen in AE Aqr (Welsh et al. 1998), but thehigh inclination of J0240 may increase this effect.Once the flare reaches its peak emission flux the lineshape remains fairly constant except for the loss of thehigh-velocity wings. An unresolved absorption featureis seen throughout the flare. Its blue-shifted velocity in- creases from 300 to 400 km s − over the 25 min. Thisabsorption may correspond to the P-Cyg features seenduring the second half of the orbit as its velocity isconsistent with an extrapolation from the later orbitalphases.In Figure 4, we show representative spectra of J0240 inthree different intervals: at the maximum of a flare, be-fore the flare, and during eclipse. Before a flare, Balmer,Paschen, He I, and Ca II emission lines are visible. Dur-ing the flares, the emission lines broaden and intensify.The strengths of He I lines are enhanced, weak He II λ λ D , = F ( H α ) /F ( H β ), is5.4 ± . D , = 3 . ± . D , = 2 . ± . α optical depth (e.g. Netzer 1975)caused by viewing the circumbinary gas close to the or-bital plane. 3.2.2. Color Variations
The photometry synthesized from the spectra andcompared with the variations in the emission line fluxesare shown in Figure 5. Flaring is seen around infe-rior conjunction and at orbital phase 0.4, as well asnearly continuous activity in the October run that cov-ered phases 0.50 to 0.85.The Balmer emission line flux and continuum bright-ness rise quickly, on time scales of a few minutes, thenfade more slowly. The emission lines clearly take longerto fade after a flare when compared with the continuumflux. The U -band variations are more similar to theemission line changes than to the B -band fluctuations. Figure 4.
Example spectra of J0240 obtained during theinferior conjunction (eclipse; red), before a flare (orange),and at the peak of a flare (blue). The flare spectra are shownwith the eclipse spectrum subtracted. The eclipse spectrumis dominated by the early M-type secondary, narrow Balmerand CaII emission (note the eclipse spectrum flux has beendivided by 10).
For example, the H β flux and U -band brightness beginto rise at the time of inferior conjunction (orbital phase0.0), while the B -band and B − V color are relativelyflat around conjunction.The B − V color index is reddest near inferior con-junction suggesting that the eclipse is blocking nearlyall of the flaring activity. The color index at eclipseis B − V = 1 . ± . . > B − V > . Orbital Motion of the Secondary
Absorption features from the M-star secondary areclearly seem in all the
LBT spectra. This allows us tomeasure the orbital motion of the secondary star withthe goal of precisely measuring the time of inferior con-junction relative to the eclipse ephemeris (Littlefield &Garnavich 2020). Thorstensen (2020) classified the sec-ondary star as as an M1.5 dwarf, so we cross-correlatedsections of the
LBT spectra with a late stellar spectrum
Figure 5.
Light curves derived from the September andOctober spectroscopy. In all the panels the dotted verticalline indicates the transition between the two runs and thesolid line is the time of inferior conjunction. top:
Relativeflux of H β versus orbital phase. middle: Synthesized U and B magnitudes versus orbital phase. bottom: Colorindex of the synthesized B and V magnitudes versus orbitalphase. The color index reaches 0.8 to 0.6 mag during flares.Around eclipse the color is 1.6 mag, consistent with the Mtype secondary star. synthesized from the MILES stellar library (Vazdekis etal. 2010) .Cross-correlations between the spectra and the tem-plate were performed using the fxcor routine imple-mented in IRAF. Only the red arm of M ODS − . A heliocentric cor-rection was applied to each spectrum and the result- http://research.iac.es/proyecto/miles/ Figure 6. Top:
The radial velocity estimates for thesecondary star as a function of orbital phase. The pointsare measurements from individual MODS2 spectra obtainedon three separate nights. The blue line is the best fit sinu-soid assuming a fixed orbital period from Littlefield & Gar-navich (2020). The data show that inferior conjunction oc-curs 3.0 ± . Bottom:
Residuals between the data and the best fit si-nusoid. The large flare seen at orbital phase 0.4 impactedthe cross-correlation velocities and generated a 10 km s − amplitude wave in the radial velocity curve. ing radial velocity curve for the J0240 secondary star isshown in Figure 6.We converted the barycentric Julian day (BJD) at thecenter of each exposure to photometric phase using theLittlefield & Garnavich (2020) ephemeris and fit a si-nusoid function to the data using a least-squares algo-rithm. The half amplitude of the radial velocity curve is241 ± − , which is consistent with the Thorstensen(2020) estimate of 250 ± − . Inferior conjunctionwas found to occur at phase φ = − . ± . ± . T with the Littlefield & Garnavich (2020) orbital period, we obtain a new ephemeris of T conj = 2459105 . . × E, (1)where T conj is the time of inferior conjunction expressedas a Barycentric Julian Date in Barycentric DynamicalTime (TDB) and E is the integer cycle count. The pre-cise time of inferior conjunction defines the radius vectorbetween the stars and this is critical in narrowing thelocation of the flaring emission in the binary system.The radial velocity curve shown in Figure 6 shows asmall deformation at orbital phase 0.40. The amplitudeof the wave is about 10 km s − . This was a phase cov-ered in the 2020 September run and is coincident witha large flare seen in the continuum and in the emissionlines. While the cross-correlation segments avoided ma-jor emission features, some lines such as the Si II 6340 ˚Aare seen in emission only during flares. These weakeremission features appear to have a mild impact on theabsorption line cross-correlation velocity estimates. Figure 7.
The flux-weighted average H α velocity versus or-bital phase (repeated for two orbits). The November runoverlaps with the October data so it is shown in magenta.The red line is a best-fit sinuosoid with an amplitude of145 km s − and the opposite phase of the secondary star.The zero point used in plotting the velocities is the sys-temic radial velocity, found from the mean velocity of thesecondary’s absorption lines. Orbital Motion of the Emission Line Region
A close inspection of the trailed H α spectrum in Fig-ure 1 suggests a pattern in the velocity of the emissioncentroid with orbital phase. The emission appears mostredshifted near φ ≈ .
75 and mildly blueshifted around φ ≈ .
25, although the high velocity flares and P-Cygniabsorption makes this difficult to detect. In disk sys-tems, a “double Gaussian” is often used to detect theorbital motion of the inner disk by focusing on the high-velocity wings of emission lines. For AE Aqr and J0240,the high velocity emission likely comes from shocked gasthat would not be a good tracer of orbital motion. In-stead, we measure the flux-weighted mean wavelengthof the H α emission feature. The results are shown inFigure 7.The shifting centroid of the H α emission moves inanti-phase to that of the secondary star’s motion. Thevelocity centroid is redshifted over the orbital phases0 . < φ < . − , but this shiftmay result from the asymmetry of the emission causedby the P-Cyg absorption and other optical depth effectsthat have a major impact on the blue side of the line.We propose that the H α line emitting region is nearlyfixed in the binary frame and orbiting about the sys-tem’s center of mass. As the H α emission has aboutthe same phasing as expected for the WD, the emis-sion would be originating within 20 ◦ of the radius vec-tor connecting the two stars and on the WD side of thesystem. Indeed, the velocity amplitude is close to whatis expected for the WD given the orbital period of thissystem. To illustrate this point, from the velocity ampli-tude of the secondary, V , and its orbital angular velocity(Ω − =4203.5 sec rad − ), we can estimate the physicaldistance of the secondary from the center-of-mass, r ,as r = V / (Ω sin( i )) (2)and for a mass ratio q , the WD separation from thecenter-of-mass is simply r = q r . The parameter ofinterest is the location of the emitting region relative tothe WD, or, r Ha /r = V Ha / ( qV ) (3)where r Ha /r is the ratio of the center-of-mass distancesof the emitting region and WD and V Ha is the orbitalamplitude of the emission region. The inclination de-pendence has cancelled, but it would be small anywaygiven that J0240 is eclipsing. The velocities have beenestimated as V = 241 km s − and V Ha = 145 km s − ,so r Ha /r = 0 . /q . If the velocity amplitude of the H α emission is a good representation of its orbital motion,then the emitting region is close to the WD for reason-able mass ratios. For example, for q ≈ .
5, the emitting region is 20% further from the center-of-mass than theWD. From this analysis, it is unlikely that ejected blobscolliding at several stellar separations from the center ofmass would produce such a low velocity amplitude andbe in phase with the WD orbit.
Figure 8. bottom:
Trailed spectra centered on the H α emission from LBT data taken 2020 October (left) and 2020November (right). The narrow blueshifted absorption fea-ture is seen continuously over the second half of the orbit. top:
Individual spectra from the two data sets. On the left isa spectrum showing a pair of narrow absorption lines at − − − . Displayed on the right is a spectrum ob-tained near inferior conjunction where the absorption is wellbelow the continuum of the secondary star. The deep ab-sorption at this orbital phase suggests that the interveninggas is on the opposite side of the system from the WD andin circumbinary orbit after acceleration by the propeller. Blue-shifted Absorption Feature
Thorstensen (2020) noted a strong P-Cygni featurethat he associated with periods of flaring. Our spectrashow that this absorption is detected over a wide rangeof orbital phases and is seen even outside of flares. Simi-lar absorption features have not been noted in spectra ofAE Aqr. The trailed spectra of J0240 displayed in Fig-ure 8 clearly shows the blue-shifted absorption slowlyvarying in velocity between orbital phases 0.5 < φ < −
380 km s − at φ ∼ . −
600 or −
700 km s − approaching inferior conjunction.A quadratic function, shown as a gray spiral in Figure 9,fits the velocity variation well.The equivalent widths (EW) of the absorption canbe as high as 4 ˚A, but typically the EW are between1 and 3 ˚A over the second half of the orbit. The P-Cygni absorption is generally seen on the blue wing ofthe Balmer emission lines. However, when the system isnot flaring, the absorption falls below the continuum ofthe secondary. This is clearly seen in the spectrum takennear inferior conjunction and shown in Figure 8. Theline deep into the red star’s continuum demonstratesthat the absorption is coming from gas on the oppositeside of the secondary from the WD. This is likely thestream ejected by the magnetic propeller and placed incircumbinary orbit (Wynn et al. 1997).A few spectra showed multiple blue-shifted absorptioncomponents with the velocity of the additional compo-nents between 1000 to 1200 km s − . An example ofa double absorption feature is shown in Figure 8 at φ = 0 .
86. The high-velocity component was detectablefor approximately 15 minutes.The width of the absorption feature is difficult to mea-sure precisely because of the steep continuum slope onthe line wings. The absorption full-width at half min-imum (FWHM) corresponds to 230 ±
30 km s − , whichis similar to the FWHM of the night sky lines implyingthat the absorption feature is unresolved.3.3. Eclipse Photometry
Queue scheduling at MDM allowed us to specificallytarget times around inferior conjunction to study theeclipse of the flaring region. We obtained photometry offive eclipses in the B -band filter and the resulting lightcurves are shown in Figure 10. Three of the eclipsesoccurred during periods of flaring and the outline of theeclipse is well-defined. Two of the runs displayed littleor no flaring and the eclipse is not evident.As a guide, we have generated a model eclipse lightcurve for a WD and a Roche lobe-filling secondarythat assumes a mass ratio of q = 0 . Figure 9.
The blue-shifted velocity (in km s − ) of the nar-row P-Cygni feature as a function of orbital phase depictedon a polar projection. The spectra obtained in November(blue points) shows a slightly larger velocity over the sameorbital phases when compared with the September (orange)and October (red) data. The size of each point is propor-tional to the equivalent width of the absorption. The P-Cygni absorption feature is most clearly seen over the sec-ond half of the orbit. The solid red line is a second-orderpolynomial fit to the P-Cygni velocities. justed the Gaussian parameters and found that a widthof 0.15 a (FWHM), or 15% of the stellar separation ap-proximated the ingress length. Finally we adjusted theorbital inclination to i = 81 ◦ so that the model approx-imately matches the length of the eclipse. Certainly,other combinations of parameters could also approxi-mate the observed eclipse profile, but the goal for thismodel is simply to use it as a consistent reference tocompare the various photometric and spectroscopic ob-servations.The brightness of the system at inferior conjunctionwas extremely consistent in all five light curves withthe average measured at B = 19 . ± .
02 mag. Dur-ing strong flares the brightness at mid-eclipse is withina mmag of the mid-eclipse brightness measured in theruns without significant flaring. This suggests that theregion producing the continuum flares is totally blockedby the secondary. Some of the B -band light curves showa slight rise at the 0.1 mag level immediately after in-ferior conjunction. Further, the eclipse egress duringflaring is more shallow than the ingress. Both of thesecharacteristics imply that the intensity distribution ofthe continuum flaring region may be asymmetric, yetsufficiently compact to be nearly totally obscured by thesecondary star.The photometry obtained by Thorstensen (2020) nearinferior conjunction hints at some variability in the0eclipse width when compared with the very consistenteclipse profile seen in our data. Any comparisons arecomplicated by the rapid brightness changes in the flaresand differences in the central wavelengths between thetwo sets of observations.Our LBT spectra obtained during an eclipse indicatethat the H α emission line flux does not match the pro-file of the continuum eclipse. The line flux rises rapidlybeginning at inferior conjunction, suggesting that theemission lines may be slightly offset from the locationof continuum flares, or extend further from the flare lo-cation than the continuum emission. As seen in bothAE Aqr and J0240 flares, the emission lines fade moreslowly than the continuum flux, implying that the gasproducing the emission lines may flow out of the locationof the continuum flare. Figure 10.
Light curves around the time of inferior con-junction. During periods of active flaring the outline of theeclipse is clearly seen. The red lines are a model eclipse of aconstant source located at the position of a WD as describedin the text. left:
Five nights of MDM B -band photometry.The three runs where flaring was active is plotted in the toppanel, while two quiescent eclipses are shown in the lowerpanel. right: The blue flux density and H α emission lineflux from the September LBT spectra are displayed with aneclipse model. A small rise in the continuum flux accompa-nied by a larger brightening in the emission line flux beginsprecisely at inferior conjunction.4.
DISCUSSION4.1.
Location of the Flares
The presence of eclipses in J0240 provides an oppor-tunity to estimate the location of the flaring activity inthe binary system. One of the longstanding debates withAE Aqr is whether its flares are produced when blobsare shocked in the WD’s magnetosphere (Eracleous &Horne 1996) or when they collide in the propeller’s out- flow, well past the WD (Welsh et al. 1998). The majordifference between these two scenarios is that in the lat-ter, the flares are produced at a large distance from theWD.We confirm the inference by Littlefield & Garnavich(2020) that there is no visible flaring in optical broad-band photometry during the eclipses. Furthermore, ourspectra verify that mid-eclipse coincides with the sec-ondary’s inferior conjunction, implying that the flaresare constrained to occur near the line passing throughthe two stars. This strongly favors the Eracleous &Horne (1996) model where the flares are predicted tobe generated in the WD magnetosphere.As pointed out in section 3.2.4, the apparent orbitalvelocity amplitude of the H α emission region is similarto that expected for the WD. In addition, the radialvelocity of the line emitting region is anti-phased to thatof the secondary. Combining these observations suggeststhat the emission line emitting region is near the WD.In the Welsh et al. (1998) blob-blob collision model,the expected flaring location would result in eclipseswell before inferior conjunction. The large extent of thepredicted interaction region would generate long par-tial eclipses that would be difficult to identify giventhe flaring variability. Furthermore, in the Welsh etal. (1998) scenario, the location of the collision regionshould vary because blobs follow different trajectories,depending on their size and density. We would thereforeexpect eclipses to vary in phase, rather than producingthe sharply defined eclipse seen in survey photometry(Littlefield & Garnavich 2020).That said, the emission line and ultraviolet (UV) lightcurves generated from our spectra (see Figure 5), differfrom the broad-band photometric eclipses shown in Fig-ure 10. The UV and line emission start to rise at zeroorbital phase, while the continuum flux is relatively flaton either side of inferior conjunction. The flat-bottomedcontinuum eclipses suggest that most of the continuum-emitting region is blocked by the secondary. The rise ofH α emission and UV flux during the continuum eclipseimplies that the line-emitting region may be more ex-tended and peeking around the secondary at the mo-ment of inferior conjunction.This geometry is consistent with the Eracleous &Horne (1996) model where blobs traveling ballisticallyfrom the secondary are shocked when encountering theWD magnetosphere. This occurs near the WD and theshocks generate the continuum flares. The acceleratedblobs continue past the WD while expanding and cool-ing, resulting in Balmer emission on the leading side ofthe WD. This process is likely to be messy judging bythe range of observed absorption velocities, but with the1data we have in hand, the continuum flares appear tobe confined to within 0 . a of the WD, where a is binaryseparation.4.2. Flare High Velocity Emission
We interpret the extreme velocity dispersion observedduring the flares (HWZI ∼ − ) as the expan-sion velocity of the blobs when they are shocked, as op-posed to the initial velocity of the blobs immediatelyfollowing their acceleration by the propeller. We basethis inference on the presence of the extreme, nearly-simultaneous redshifts and blueshifts observed aroundinferior conjunction, when the trajectory of the expelledblobs is roughly perpendicular to our line of sight. If wetreat the blobs as simple particles following trajectoriesout of the system, their velocity vectors at certain or-bital phases would be perpendicular to our line of sight,so we would observe high-velocity flares only at certainorbital phases. Conversely, if the blobs rapidly expandafter being shocked, the resulting Doppler shifts wouldbe observable even if the blobs were moving in the trans-verse direction. Eracleous & Horne (1996) argued thatshocked blobs could expand at several thousand km s − because they would be opaque to the X-rays producedinternally by the shock-heated plasma.The P Cyg absorption component provides a secondargument against interpreting the HWZI as the initialvelocity of the ejecta. The hyperbolic excess velocity v term of an object is given by v term = (cid:113) v − v escape ,and we find that if v ∼ − , the outflow willdecelerate only slightly. Since the outflow expands inthe radial direction (see Sec. 4.4), the projected veloc-ity of the outflow’s P Cyg absorption component wouldtherefore be much larger than what is observed.4.3. The Lack of Coherent Oscillations
In most characteristics, J0240 and AE Aqr are ex-tremely similar. A 33 s oscillation is seen in AE Aqrassociated with the spin period of its WD; however, ourfast photometry does not detect a coherent oscillationfrom J0240. For AE Aqr, Patterson (1979) found thespin and its harmonic with amplitudes ranging between1 and 10 mmag, but generally the signals were 2 to3 mmag, just below our limit of detection. Observationsby Bruch et al. (1994) found that the spin oscillations inAE Aqr fell below detectability for a month, suggestingthat we might simply be unlucky in the timing for oursearch.From UV observations, Eracleous et al. (1994) mod-eled the spin modulations from AE Aqr as hotspots onits WD. Their best fit placed the spots at mid-latitudeson the star in order to match the very large amplitude variations seen in the UV. In J0240, any hotspots may beclose to the spin axis, resulting in a smaller amplitude.It may even be that J0240 is a more efficient propellerthan AE Aqr, and no accretion heating is occurring atthe magnetic poles.More sensitive searches for the WD spin signature areneeded. In AE Aqr, the amplitude of the spin modu-lation increases toward shorter wavelengths, so fast ca-dence photometry in the UV might be the best way todetect the WD spin period in J0240.4.4.
Application of Wynn et al. (1997) to J0240
Our results are strongly consistent with the model ofAE Aqr from Wynn et al. (1997), who argued that theaccretion flow is clumpy, with discrete, heterogeneousblobs that each experience a drag force as they encounterthe WD’s magnetosphere. Near their point of closestapproach to the WD, the blobs are accelerated above thebinary escape velocity and expelled. As they leave thesystem, the blobs trace a spiral pattern in the binary restframe. We refer to the blob velocity after ejection as the“terminal” velocity, but this is just an approximation asthe blobs continue to slowly decelerate in the potentialwell of the binary.Although the modeling in Wynn et al. (1997) is tai-lored to AE Aqr specifically, nearly all of our obser-vations can be understood within its framework. Forexample, in the Wynn et al. (1997) model, there is noaccretion disk, and we see no evidence of one in our ob-servations. Even more importantly, the persistent P Cygprofile seen during the second half of the orbit is a nat-ural consequence of the outflow trajectories computedby Wynn et al. (1997). Their Figure 7 predicts thatshortly after its interaction with the magnetosphere, thepropeller’s outflow will begin to spiral outward from thebinary (as seen in the binary rest frame), and we wouldexpect to see line absorption when this outflow lines upwith our view of the secondary and flaring region. Thevelocity vectors in their Figure 7 are pointed outward ra-dially, so the resulting absorption would be blueshiftedby several hundred km s − , as observed. Unless theorbital inclination is exactly 90 ◦ , the outflow will even-tually be too far from the binary to intersect our line ofsight to the secondary, and the absorption will thereforecease.At present, J0240 cannot be modeled in the same wayas AE Aqr because its spin period is unknown. There-fore, to provide a quantitative basis for our interpreta-tion, we shall consider how AE Aqr would appear if itwere viewed at a high orbital inclination. Blobs are ex-pected to expand out of the orbital plane, and they maybecome comparable in size to the WD’s Roche lobe af-2 orbital phase Spatial orbital phase Doppler map orbital phase Absorption velocity km s Figure 11. Left:
Schematic diagram of an outflow that produces line absorption at high inclination, presented in the binaryrest frame. The magnitude of the velocity vector is represented in color, and the numbers around the edge of the figure indicateorbital phases. The parameters used to create this model are discussed in the text.
Center:
The predicted radial velocity of anabsorption feature resulting from the outflow model displayed in the left panel. This can be directly compared to the observedabsorption velocities in J0240 displayed in Figure 9.
Right:
The motion of the outflow in velocity space. This panel describesemission from the outflow (not absorption), so the requirement of a backlight from the center panel is lifted. ter being expelled (Eracleous & Horne 1996). In thisscenario, ejected blobs could intercept the line betweenthe observer and a source of light in the system, such asthe secondary or flaring region. For example, at orbitalphase 0.75, the propeller’s outflow would be ∼ a fromthe secondary in the modelling by Wynn et al. (1997). Ifthe secondary and one of the expelled blobs have radiiof 0 . a and 0 . a , respectively, the blob could occludethe secondary if i ≥ tan − (4 a/ (0 . a + 0 . a )) = 80 ◦ .A quarter-orbit later, when the outflow is ∼ a fromthe secondary, the inclination required for an occlu-sion increases to i ≥ ◦ . These simple geometric con-siderations suggest that i dictates the range of orbitalphases that the P Cyg absorption component remainsvisible, because as the outflow becomes increasingly dis-tant from the secondary, an increasingly high value of i would be needed to produce an absorption feature.Additionally, we note that a blob’s path and its termi-nal velocity, can be tweaked by adjusting the drag coef-ficient, k , from Wynn et al. (1997). Increasing this valuecauses the accretion stream to begin interacting furtherfrom the WD and results in a higher terminal velocity.We illustrate this for AE Aqr using the parameters fromWynn et al. (1997) in Figure 11, which shows a single-particle trajectory for AE Aqr for k = 2 × − s − . Ofparticular note is the center panel of Figure 11, whichshows the predicted radial velocity of a blob backlit byan emitting source at the binary center of mass. Thecenter of mass was chosen for this calculation because itis conveniently situated between the two likely sourcesof light in the system: the secondary star and the flaringregion near the WD. The predictions of the absorption velocities from gas close to the interaction region areunreliable in this approximation, but the absorption ve-locities from distant blobs do not strongly depend onthe location of the light source.Figure 11 displays a model for AE Aqr, but its over-all similarity to the P Cyg velocities and orbital phasesobserved in J0240 (Figure 9) is quite striking. Thereare certainly differences in the details between the ob-served and modelled absorption features. The terminalvelocities in the AE Aqr model are nearly constant be-tween phases 0 . < φ < . − . Incontrast, the observed J0240 absorption velocities areseen to increase from 400 km s − near φ = 0 .
5, to 600to 700 km s − at inferior conjunction. If the spin pe-riod were known, the k parameter could be adjusted toachieve a better fit with the observed velocities. Never-theless, with the parameters of AE Aqr, we cannot finda value of k that causes the predicted absorption veloc-ity to increase at later orbital phases; indeed, this gas isfar from the influence of the propeller and conservationof energy suggests that the expelled blobs would slowlydecelerate as they move outward through the binary’sgravitational potential.We speculate that the drag coefficient in a real pro-peller system may systematically vary on a time scale ofminutes, hours, and days. This could result from varia-tions in the blob properties as they leave the secondary,or from active flares modifying blobs approaching theWD. We see kinks with amplitudes of 50 km s − in theabsorption line velocities (Figure 9) over a single observ-ing run, as well as differences of ∼
100 km s − betweenruns, suggesting that the ejection velocities are varying3at both short and long time scales. Further high-qualityJ0240 spectroscopy is needed to determine if the absorp-tion velocities consistently increase with orbital phase asseen in two of our observing runs.The ease of detecting the outflow in J0240 offers astark but refreshing contrast with AE Aqr, whose out-flow is challenging to detect at that system’s much lowerinclination. Many previous studies have attempted touse Doppler tomography to discern the outflow (e.g.Wynn et al. 1997; Welsh et al. 1998; Horne 1999),but the resulting tomograms have tended to show lit-tle more than a featureless blob of emission centered inthe − V x , − V y quadrant. In contrast, the P Cyg absorp-tion in J0240 offers a direct and unambiguous means ofstudying the outflow. In particular, if a future studycan identify the spin period as well as the orbital incli-nation i from eclipse modeling, the exact trajectory ofthe outflow could be rigorously mapped. Likewise, the-oretically predicted outflow velocities, such as those inFig. 11, can be tested against the radial velocity of theP Cyg absorption (see Figure 9) once the spin period isknown. CONCLUSIONOur observations and analysis establish that J0240 isa magnetic CV in a propeller state and the first eclipsingAE Aqr type star. Our major conclusions are as follows.1. We confirm that the flaring region undergoeseclipses by the secondary star in J0240 as notedby Littlefield & Garnavich (2020).2. The optical flares noted by Thorstensen (2020)coincide with transient emission-line flares whosewings extend to ± − in the Balmer andHe I lines. This unique high-velocity flaring is seenin AE Aqr and points to a strong similarity be-tween the two systems.3. We identify a persistent narrow Balmer absorp-tion feature between orbital phases 0.5-1.0. Its blue-shifted velocity is seen to increase with or-bital phase. We argue that this P Cyg absorptionresults from gas ejected by the propeller into cir-cumbinary orbit.4. The narrow absorption is seen below the level ofthe continuum from the secondary around infe-rior conjunction, showing that the absorbing gasis consistent with the outflow models of Wynn etal. (1997) and our own simulations.5. The emission lines are formed primarily in two dis-tinct regions. One of them, which we identify asthe magnetic-propeller region close to the WD, isthe source of the higher-order Balmer, He I, andhigh-velocity H α and H β emission. The other isan extended outflow that produces mostly low-velocity H α emission, and its blue edge is definedwith the presence of P Cyg absorption.6. We unsuccessfully searched for a photometric sig-nature of the spinning WD as is seen in AE Aqr.We placed a 4 mmag upper limit on the g -bandspin amplitude for periods between 6.3 s and 85 s.Because of its high orbital inclination, J0240 enablesobservational tests that are impossible with AE Aqr,which has a much lower orbital inclination. It is there-fore a compelling candidate for theoretical modeling.ACKNOWLEDGMENTSWe thank R. Pogge and O. Kuhn for their help inobtaining the LBT observations. This work is partlybased on observations obtained at the MDM Observa-tory, operated by Dartmouth College, Columbia Uni-versity, Ohio State University, Ohio University, and theUniversity of Michigan.REFERENCES Bruch, A., Beskrovnaya, N., Ikhsanov, N., et al. 1994,Information Bulletin on Variable Stars, 3996, 1Campana, S., Stella, L., Mereghetti, S., et al. 2018, A&A,610, A46. doi:10.1051/0004-6361/201730769Echevarr´ıa, J., Smith, R. C., Costero, R., et al. 2008,MNRAS, 387, 1563. doi:10.1111/j.1365-2966.2008.13248.xEggleton, P. P. 1983, ApJ, 268, 368. doi:10.1086/160960Eracleous, M., Horne, K., Robinson, E. L., et al. 1994, ApJ,433, 313 Eracleous, M. & Horne, K. 1996, ApJ, 471, 427.doi:10.1086/177979Harding, L. K., Hallinan, G., Milburn, J., et al. 2016,MNRAS, 457, 3036. doi:10.1093/mnras/stw094Henden, A. A., Levine, S., Terrell, D., et al. 2015, AmericanAstronomical Society Meeting Abstracts 4