arXiv:1905.11325v1 [astro-ph.SR] 27 May 2019 as hi eeto n dnicto ed spatial needs identification and be- resolution. Sun, detection the energetic their on most only cause the observed are CMEs, phenomena, coronal the atmosphere. in stellar occurring outer phenomena energetic highly sociated eoncin ntecrn.Terlae nryis released The corona. the in reconnections scenario dard ocle tla antcactivity magnetic stellar so-called of cause major loss. a momentum be angular suggesting and indeed levels, mass could activity CMEs higher the that to of extrapolation case the solar in clues provide values op ihvelocity with loop, o lsa ( of plasmas motions downward hot and upward indicate S that in detected shifts distinctly Doppler observed we 9024 Chandra/HETGS, HR star with active the on flare a stellar of time- spectroscopy X-ray Using high-resolution resolved unexplored. observationally main be to significant predicted are environments circumstellar M iei nryof energy kinetic CME higher oteflr.Fo hseiec ewr able of were mass we CME evidence a this From derive to flare. the to ( in with tube. tion, blueshift magnetic O later flaring the a detected in of also we model notably, Most a with ment hwmgei ciiylvl pto up Stars levels Sun. activity the magnetic on show occurring phenomena netic flares with ated ∼ Msaecoeylne oflares to linked closely are CMEs nes tla antcfilsaersosbefrthe for responsible are fields magnetic stellar Intense ooa aseetos(Ms,otnassoci- often (CMEs), ejections mass Coronal 4 tla aecrnlms jcineetrvae by revealed event ejection mass flare-coronal stellar A K,ta eaciet M coupled CME a to ascribe we that MK), 4 n M ffcso tla hsc and physics stellar on effects CME and , viii 1 5,6,7,8,9 nvriyo aem,Dprmn fPyisadChemistry and Physics of Department Palermo, of University v 11 iewihrvasa padmo- upward an reveals which line ae r rvnb musv magnetic impulsive by driven are flares 90 = ∼ .Argiroffi C. 3 2 1,2,3 mtsna srpyia bevtr,M-,6 adnSt Garden 60 MS-3, Observatory, Astrophysical Smithsonian NF-Osraoi srnmc iPlro izadlParl del Piazza Palermo, di Astronomico Osservatorio - INAF 10 oee,selrCE re- CMEs stellar However, . − v r h otpwru mag- powerful most the are , ± xvi ∼ .Bonito R. 25 0k s km 30 Si , 100 K ihnteflaring the within MK) 5 . 2 1 − − +27 xiv , 1 2 3 − . 0 ms km 400 ⋆ . 2 6 .Reale F. , . 1 − +2 n Mg and , 7 10,4 fco plasma cool of , -a lsamotions plasma X-ray 0 2 × . . .Miceli M. , 6 8 n o h as- the for and , 3 10 × ntestan- the In . 34 − 10 1 erg nagree- in , 21 10 ⋆ 1 xii g These . costanza.argiroffi@unipa.it , 4 2 .J Drake J. J. , a 8 2019 28, May n a and times lines 1 , 2 .Orlando S. , 1 opin bevdi h cisn rctcymcbi- precataclysmic ab- eclipsing UV/X-rays the CMEs in transient observed corona. explain sorptions quiescent to constant candidates flar- a were isothermal and an plasma of ing with assumptions power combined oversimplified diagnostic spectroscopy, the limited X-ray spuri- the low-resolution a from of be coming can result absorption ous increased the that sinuate n ol infiatyiflec exoplanets influence significantly could and aetmeaueadeiso measure emission and dur- temperature observed flare absorption flares X-ray ing increased brighten- explain chromospheric events to by evaporation or explained ings be also could eesmtmsatiue oCMEs to attributed lines sometimes chromospheric were of components Blueshifted CMEs. osaonso asadkntceeg loss energy kinetic and enor- to extrapo- mass cause (up This of should CMEs amounts stellar one. mous that solar suggests the coronae lation stellar from or- active profoundly several though differ to even extrap- up magnitude, presumed relation of be ders flare-CME only solar prop- can the the CMEs Currently, olating stellar one. of solar erties the than higher times hi ciiylvl esrdb h -a obolo- to X-ray plasma ratio, the by coronal luminosity measured metric denser level, and activity hotter Their , flare h neligcrmshr,ta xad upward expands heats of that hundreds and at lines chromosphere, field underlying magnetic the the along transported as n iei nryices ihicesn flare increasing with energy. increase energy and energy kinetic mass occurrence, and CME of mass, that amounts demonstrate observations large Solar away CMEs, carry plasma, confined that large- previously also of cause expulsions often drivers scale de- magnetic (flare flare conductively The and cay). radiatively gradu- down plasma this cools Then ally phase). rising (flare structure hsfr hr aebe e liso stellar of claims few a been have there far, Thus ciesashv togrmgei ed,higher fields, magnetic stronger have stars Active izadlPraet ,914 aem,Italy. Palermo, 90134, 1, Parlamento del Piazza , 3 .Ciaravella A. , 1,2 ∼ orbrtn h aeCElink. flare-CME the corroborating , 18 2 et abig,M 23,USA. 02138, MA Cambridge, reet, 10 n .Peres G. and , oee,tesmlaeu aitosof variations simultaneous the However, . mno1 03,Plro Italy. Palermo, 90134, 1, amento − 9 M ms km ⊙ yr − − 1 ligteoeligmagnetic overlying the filling , 1 2 .Testa P. , and L 16,17 X 1 , /L 2 ∼ Mswr invoked were CMEs . bol 0 a eu to up be can , . 1 L 3 , 13,14,15 bol respectively), , 18 5 u they but , ( . EM 6,7,8,9 in- ) 10 12 4 . 21 31 −1 HR 9024 indeed shows coronal (LX ∼ 10 ergs , with T ∼ 1 − 100 MK) and magnetic field properties 22 2 (a dominant poloidal field with Bmax ∼ 10 G) anal- ogous to that of other active stars. Therefore, irre- spective of the origin of some magnetic components, and bearing in mind that active stars may show di- verse magnetic configurations, the coronal phenomena occurring on HR 9024 can be considered as represen- tative of those of active stars. HR 9024 X-ray spectrum was collected during a 98 ks-long Chandra/HETGS observation (Fig. 1a), during which a strong flare (peak luminosity ∼ 1032 ergs−1 and X-ray fluence ∼ 1036 erg) was regis- tered (Fig. 1b). The high energy of this flare maximizes the proba- bility to have an associated CME 1. The flaring loop located near the stellar disk center, as hinted by the Fe fluorescence 25, maximizes the possibility to detect radial velocities of both flaring and CME plasmas. We measured time-resolved individual line positions, considering only strong and isolated lines that probe plasma with T ranging from 2 to 25 MK (Fig. 2 and Methods and Supplementary Table 1). We found: Figure 1: Observed X-ray spectra and curve of HR 9024. a, X-ray spectra collected with the Medium Energy Grating (MEG) - significant blueshifts during the rising phase of and High Energy Grating (HEG) during the 98 ks long Chandra the flare in the S xvi line at 4.73 Å (−400 ± observation, with the strongest emission lines labeled. MEG and 180kms−1) and in the Si xiv line at 6.18 Å HEG bin size are 5 and 2.5mÅ. b, X-ray light curve of registered (−270 ± 120kms−1), with a 99.99% combined sig- during the Chandra observation, obtained from the ±1 order spectra of HEG and MEG. nificance of the two line shifts. - significant in the Si xiv line at 6.18 Å (140±80kms−1) and Mg xii line at 8.42 Å (70±50 nary V471 Tau 19. However, such absorptions can and 90±40kms−1), during the maximum and de- equally be produced by a stellar wind 20. In addition cay phases of the flare, with a 99.997% combined to their ambiguous interpretation, all these candidate significance of the three line shifts. detections never provide the CME physical properties, unless substantial assumptions are made. - a significant blueshift in the O viii line at 18.97 Å We present here strong evidence for the detection (−90 ± 30kms−1), after the flare, significant at and identification of a stellar CME, the estimate of the 99.9% level. its properties, and the simultaneous monitoring of the associated flare energetics. The Sun indicates that a The first two Doppler shifts tell us that hot flaring flare-CME event should produce hot plasma moving plasma moves upward at the beginning of the flare, and upward and downward within the flaring loop, and, then settles back down to the chromosphere, as pre- after the flare onset, cool plasma in the CME mov- dicted 11. We compared the observed velocities with ing upward. Therefore, monitoring the plasma veloc- predictions based on the flaring loop model 21 (Meth- ity at different temperatures during a stellar flare is ods and Supplementary Fig. 1). For each line and each a potentially powerful method of searching for CMEs. time interval, we computed the corre- The unrivalled X-ray spectral resolution of the Chan- sponding to the predicted line centroid, for different dra/HETGS jointly with detailed hydrodynamic (HD) possible inclinations of the flaring loop (identified by modeling allowed us to investigate a very favorable the φ and θ angles defined in Fig. 3a). case: the strong flare 21 observed on the active star The best agreement between observed and predicted HR 9024. velocities (Fig. 3b-f) is obtained assuming a loop ob- 22 ◦ ◦ HR 9024 is a G1 III single giant , with M⋆ ∼ served from above (φ = 0 and θ = 90 , see Meth- 25 2.85 M⊙ and R⋆ ∼ 9.45 R⊙, located at 139.5pc. Its ods), confirming previous independent results . The convective envelope and rotational period 23 (24.2 d) agreement for the Si xiv and Mg xii lines is striking indicate that an efficient dynamo is at work 22, as ex- (Fig. 3c-f). The observed S xvi blueshift, associated pected in single G-type giants 24. Even if some contri- with chromospheric plasma upflows, is of the same or- bution to its magnetic field could have fossil origin 22, der but even more extended in time than expected.

2 Figure 2: Time-resolved line fits. a, Analysis of the S xvi line at 4.73 Å, as registered with the MEG ±1 orders. In all plots vertical bars indicate errors at 1σ. a1, count rate detected in a 0.1 Å interval centered on the line. Letters and colors indicate the different time intervals considered to collect the line profile. a2, Observed line profile, in different time intervals (black), with superimposed (in different colors, following the same color-code used for the different time intervals) the corresponding best fit function (with dashed curves corresponding to the two transitions of the Lyα doublet, solid curve to their sum). Vertical dashed gray lines mark the rest position of the two Lyα components. On the x axis we report the velocity in the stellar reference frame with respect to the bluest Lyα component. a3, Line Doppler shifts, in the different time intervals, computed in the stellar reference frame. Horizontal bars specify the time interval. Filled circles mark velocities different from zero at 1σ level at least, open circles indicate velocities compatible with zero. The other plots analogously show the time-resolved analysis of: Si xiv line at 6.18 Å as registered with MEG (b1-b3) and HEG (c1-c3), Mg xii line at 8.42 Å as registered with MEG (d1-d3) and HEG (e1-e3), O viii line at 18.97 Å as registered with MEG (f1-f3).

3 Figure 3: Comparison between observed and predicted velocities. a, Viewing geometry of the flaring loop with respect to the observer direction (red line) identified by the φ and θ angles. b-f, Comparison between the line Doppler shifts observed (red circles) and those predicted by the flare model, for φ = 0◦ and θ = 90◦, for the lines showing significant shifts. Small dark diamonds indicate velocities predicted in 1 ks intervals. Large blue diamonds mark predicted velocities computed integrating the model spectra over the same time intervals considered for the observed spectra. Vertical bars indicate errors at 1σ. Horizontal bars mark to the time interval duration. g, Schematic illustration of the star/loop/CME configuration.

The agreement obtained for flaring plasma velocities is extrapolation of the magnetic configuration of the Sun an important validation of the standard flare model for to more active stars is not straightforward, a CME re- flare energies up to 1036 erg. mains the most rational and only explanation. Interestingly, we found that the coolest inspected Hypothesizing that the CME moves exactly along line, O viii Lyα, that forms at ∼ 3 MK, is signif- the line of sight, the distance traveled by the CME in −1 icantly blueshifted (with v = −90 ± 30kms ) in the post-flare phase is ∼ 0.8 R⋆. Most likely the CME the post-flare phase, while no shift is observed dur- started its motion simultaneously to the flare onset (so- ing the flare. The low rotational velocity of HR 9024 lar flare and CME onsets differ at most of 1ks, and the (v sin i ≈ 20kms−1) 23 excludes that this motion is due O viii profile during the flare appears broad, with some to structures fixed on the stellar surface. blueshifted excess, Fig. 2f). Assuming a constant veloc- We identify this motion as the signature of a CME ity, the total distance traveled by the CME is ∼ 1.4 R⋆. (Fig. 3g): it involves only cool plasma, it occurs after At this distance from the stellar surface the escape ve- a strong flare located near the stellar disk center, and locity is 220kms−1, larger by a factor of ∼ 2.4 than the observed velocity is within the range of solar CME the CME velocity. The real CME velocity, as well as its velocities 3 (i.e. 20 − 3000kms−1). Solar flares are traveled distance, can be higher because of the possible sometimes followed also by the formation of expanding inclination between the CME trajectory and the line of giant arches 26. However, this expansion is very slow sight. Assuming an inclination of 45◦ (i.e. the max- (∼ 1 − 10kms−1) and possibly only apparent, because imum separation angle for solar flare-CME pairs 1,2) due to the sequential brightening of different loops 27, the ratio between the local escape velocity and the real and hence unrelated to Doppler shifts. Although the CME velocity would reduce to ∼ 1.6.

4 Figure 4: Extrapolation of the solar flare-CME relation. Mass M (a) and kinetic energy Ekin (b) of solar CMEs, as a function of the X-ray fluence of the associated flares 1, with the corresponding power-law relations 7, compared to the analogous properties of the observed flare-CME pair on HR 9024, and that of other candidate stellar CMEs 9.

The initial CME mechanical energy has however a suggests that the correlation with the flare energy, ob- minor importance in determining the CME ending. served for the Sun 6,7, might hold also for stronger Solar CMEs follow non-ballistic motions. Magnetic flares at higher activity levels. Despite their unset- forces and wind interactions often cause strong out- tled identification (as CMEs or chromospheric evap- ward acceleration up to heights of several solar radii 3. oration), unconstrained mass estimate (the reported The detected CME on HR 9024 shows indeed a veloc- values are lower limits 9), and lack of X-ray coverage ity approximately constant during the post-flare phase (X-ray flare fluence was assumed to be comparable to (−100 ± 50 and −80 ± 50kms−1 in the 40 − 70 and that in the U band 9), the previous candidate stellar 70 − 98 ks time intervals, respectively), indicating that CMEs 9 (gray squares in Fig. 4a) also agree with this strong magnetic forces act on the CME balancing the high-energy extrapolation. Conversely, the obtained stellar gravity. Having no data afterwards, we cannot CME kinetic energy is ∼ 104 less than expected from firmly conclude whether the CME does eventually es- solar data extrapolations (Fig. 4b). cape to infinity. Therefore, CMEs of active stars may not be a scaled viii Assuming that the post-flare O line entirely version of solar CMEs. In terms of the amount of comes from the CME, and inferring a CME temper- mass ejected, the CME formation mechanism appears ature of 4 ± 1 MK, we derive an EM of (2.8 ± 1.0) × to scale smoothly from the solar case to higher flare 53 −3 10 cm (see Methods and Supplementary Table 2). energy and higher magnetic activity level. In terms of As in hot plasmas of solar CMEs, we expect that the kinematics, remembering that solar CMEs can expe- CME plasma is optically thin and that, in the observed rience acceleration both in the low corona (< 2 R⊙) 28 interval, there is no significant heating source . The or at large distances 3, the CME acceleration mecha- duration of the observed CME X-ray emission (sta- nism appears instead less efficient, possibly suggesting ble in the post-flare phase) indicates a radiative cool- a different energy partition in the flare-CME pair. ing time τ > 60 ks. Being conservative, we assumed These CME parameters (i.e. a mass compatible with τ ∼ 200 ks and a factor of 10 for its confidence interval solar extrapolations, and a significantly reduced E ) (i.e. 60ks <τ < 600 ks). From that we inferred the kin fit well with magnetohydrodynamic models 29,30, indi- CME density (n ), volume (V ), mass (M), and kinetic e cating that the balance between the magnetic forces energy (E , see Methods and Supplementary Table 2, kin acting on the CME can be different on active stars, obtaining: with a higher efficiency of the inward force due to the magnetic tension of the overlying large-scale field with n = (5.5+11.8) × 108 cm−3, e −3.7 respect to the outward force due to the magnetic pres- +10.3 36 3 sure of the flux rope. V = (1.1−1.0 ) × 10 cm , As integrated effect, assuming that the observed +2.6 × 21 M = (1.2−0.8) 10 g, CME eventually escapes the star, the inferred large +27.7 34 CME mass supports the hypothesis that CMEs can be E = (5.2 ) × 10 erg. kin −3.6 a major cause of mass and angular momentum loss in 6,7,8,9 Computing also the X-ray flare fluence (EflareX ≈ active stars , even if it remains unclear down to 2.8 × 1036 erg), we can compare this flare-CME pair what energy flares can cause eruptions in active stars. with the solar ones (Fig. 4). The inferred CME mass The diminished Ekin to EflareX ratio indicates instead

5 that, at high activity levels, the energy fraction carried into account. The σ of the Gaussian was fixed to the out by CMEs diminishes. That supports that magnetic predicted value 32. Since the Lyα lines are doublets, we activity at most extracts ∼ 10−3 of the stellar bolomet- fit their observed profiles with two Gaussians, with rel- ric luminosity, and excludes the huge magnetic energy ative positions fixed to the predicted differ- budget implied by solar case extrapolations to higher ence, and relative intensity fixed to the predicted value activity levels 7. (i.e. 2:1 in the optically thin emission regime).

Methods The observed line shifts, with respect to the pre- dicted , provide velocity with respect to

Data analysis the Chandra satellite reference frame (vsat). We as- sumed that these velocities are the same as that with HR 9024 was observed on August 2001 for 98ks with respect to the Earth, because of the low satellite ve- Chandra/HETGS (ObsID 1892). This instrument con- locity (at most of ∼ 1 − 2kms−1 with respect to the figuration consists of two transmission gratings, the Earth). We computed the plasma velocity in the stel- HEG and the MEG, used with the ACIS-S detector. lar reference frame as v = vsat + vE − v⋆, where vE is the The two gratings simultaneously collect spectra in the line-of-sight Earth velocity at the epoch of the observa- − − 1.2 15 Å (HEG) and 2.5 31 Å (MEG) intervals, with tion (i.e. v = 19.9kms−1 in the heliocentric reference a spectral resolution (FWHM) of 12 and 23mÅ, respec- E frame), and v⋆ is the radial velocity of HR 9024 (i.e. tively. In this work we inspected the HEG and MEG −1 v⋆ = −1.6kms in the heliocentric reference frame). spectra of HR 9024 separately, each one obtained by Throughout the paper we indicated outward motions − adding +1 and 1 diffraction orders. The data have with respect to us (redshifts) with positive radial veloc- been retrieved from the Grating-Data Archive and Cat- 31 ities, and inward motions with respect to us (blueshifts) alog . with negative radial velocities. The wavelength calibration of the Chandra/HETGS, allows velocity measurements down to a few tens of kms−1 by comparing line positions within and between The lines showed in Fig. 2, and discussed in the paper observations, as confirmed by several studies 32,33,34,35. are the ones for which a significant shift was detected Hence the Chandra/HETGS is well suited to measure in at least one time interval. For lines with S/N ratio the velocity of flaring and CME plasmas. high enough, we analyzed both HEG and MEG spectra. To search for Doppler shifts we selected a sample of Because of the different S/N and spectral resolution strong and isolated emission lines (Supplementary Ta- between corresponding MEG and HEG spectra (with ble 1). Inspecting isolated lines allows us to avoid line MEG providing higher S/N but lower spectral resolu- position uncertainties due to line blending. Moreover tion than HEG), significant line shifts were sometimes monitoring individual lines, instead of inspecting the found only in one grating. In these cases, the shift mea- whole spectrum, allows us to probe separately plasma surements obtained with the two gratings (even if only components at different temperatures. The selected one was significantly different from zero) were anyhow line sample is composed of: Lyman series lines of N vii, compatible among themselves. O viii, Mg xii, Si xiv, and S xvi; He-like ion lines of Mg xi, Si xiii, and O vii; and the strong Fe xvii lines We display in Supplementary Fig. 2 the observed and in the 15 − 17 Å range. We did not inspect the Ne ix predicted shifts for those lines with a S/N high enough and Ne x lines, because of the severe blendings with to allow time-resolved spectral fitting, for which no sig- Fe lines. The maximum formation temperature of the nificant shift was obtained. For the hottest of these inspected line sample ranges from 2 to 25 MK. lines (i.e. the Si xiii at 6.65 Å, and the Mg xii at We determined the position of each selected line 7.11 Å), the expected radial velocities during the flare by least-squares fitting its observed profile in different evolution are high enough to be detectable with the time intervals. The selected duration of the inspected Chandra gratings. However, the low S/N collected time intervals is aimed at: a) separating the phases for these lines (total line counts are listed in Supple- corresponding to significantly different predicted veloc- mentary Table 1) avoids precise shift measurements ities (HD model, described below and in Supplemen- and/or prevents from us exploring time bins short tary Fig. 1, predicts high upward velocity during the enough. For the coolest of these latter lines (i.e. all rising phase, and slower downward motions during the the Fe xvii lines), in addition to the low S/N, signif- maximum and decay phases, Fig. 3b-f); b) collecting icantly smaller shifts only in very short time intervals enough counts to perform the line fit. For each line are expected. Notice that the large redshifted veloci- we performed the fit in a small wavelength interval, ties of ∼ 200kms−1 expected for the very final part of with width ∼ 0.1 − 0.2 Å, around the line rest wave- the flare correspond to phases in which the line fluxes length. As best-fit function we assumed a Gaussian become negligible, because of the vanishing EM of the plus a constant, to take also the continuum emission flaring loop.

6 Flaring loop model and line emission synthesis reproduce that observed. In general, since the coronal part of the loop is most To infer the expected line Doppler shifts due to plasma of the flaring volume, filled with upflowing plasma at motions during the flare, and to constrain the loop ori- temperatures ≥ 50−100 MK, we would expect Doppler entation with respect to the observer, we considered shifts in highly ionized lines, like Ca xx formed at 21 the flare loop model presented by Testa et al. . This ∼ 50 MK, but they are not detected 21. Instead, we de- model assumes that: the stellar magnetic field is so tected Doppler shifts, associated with motions of flar- strong as to confine the plasma inside single closed ing plasma, in lines at T ∼ 10 − 25 MK, hence emitted magnetic tubes (coronal loops); the footpoints of the mostly at the loop footpoints. This happens because flaring loop are anchored to the photosphere; the flar- Doppler shift measurements are best obtained in lines ing plasma moves and transports energy only along detected with high S/N ratio. The spatial distribu- the field lines, and its evolution can be described with tions of T and EM along the flaring loop determine a 1D HD model along the tube. The HD equations that hotter lines, mainly emitted by higher loop por- for a compressible plasma fluid are solved numerically tions where the EM is lower, have lower S/N with to obtain the evolution of the plasma density, temper- respect to cooler lines, that come primarily from lower ature, and velocity along the loop. The flare is trig- loop portions where the EM is higher. gered with a strong heat pulse injected inside an ini- Finally, the agreement obtained between observed tially hydrostatic and relatively cool loop atmosphere and predicted velocities, consdering that this HD flare which includes a tenuous corona linked to a dense chro- model was tuned to match only X-ray flux and plasma mosphere. Tuning the model parameters to reproduce temperature, support the hypothesis of a flare occur- the observed evolution of T and EM during the flare, 11 ring in a single loop, and allows us to confirming the it came out that: the total loop length is 5 × 10 cm; loop geometry, the temporal and spatial distribution the duration and rate of the heat pulse are 15 ks and 33 −1 of heating, and the kinetic energy budget involved in 1.2 × 10 ergs , respectively, for a total injected en- plasma motions. ergy of ∼ 1.7 × 1037 erg; the pulse heats the plasma to a maximum temperature of ∼ 150 MK and makes it expand from the chromosphere at a maximum speed of Cool plasma velocity after the flare −1 −1 ∼ 1800kms , which drops rapidly below 400kms The blueshifted emission, detected at 3σ level in the after a few ks since the heat pulse has started. The evo- post-flare phases in the O viii line, is per se robust be- lution of velocity, temperature, and EM of the flaring cause of the accuracy of the Chandra wavelength cal- loop are shown in Supplementary Fig. 1. ibration: in a sample of active stars 32 the shift dis- The X-ray emission of the flaring loop is assumed played by this line is always smaller than 13kms−1. to be optically thin. Line emissivities were retrieved To further corroborate this detection we anyhow in- from the APED database 36. We computed line emis- spected the N vii Lyα at 24.78 Å that forms mainly at sion from the flaring loop considering both short time 2 MK. This line does not have enough counts to fit its intervals (i.e. 1 ks), to monitor line profiles on time profile. Selecting the post-flare interval, the average scales corresponding to the characteristic time scales position of the photons detected in the ±1000kms−1 of the flaring loop variability, and long time intervals interval around its rest position is −110 ± 80kms−1, (i.e. corresponding to that adopted for the observed thus confirming the cool plasma blueshift at 1.4σ level. line profiles) to perform a proper comparison between These two simultaneous blueshifts further increase the observed and predicted line shifts. significance of the detected plasma motion. Summariz- We computed predicted line profiles for different ing, considering all the inspected lines in the post-flare viewing geometries of the loop, exploring the range phase, plasma at T . 4 MK moves upward, hence it is 0◦ <φ< 90◦ and 0◦ <θ< 90◦ (Fig. 3a). The φ angle located in the CME, possibly representing its hottest determines a global scaling factor in the predicted line component; conversely, plasma with T & 5 MK ap- shifts, with φ ∼ 0◦ corresponding to the largest shifts, pears motionless, and hence situated in stable coronal and φ ∼ 90◦ corresponding to no shift. Taking the pos- structures. sible values of θ into account, and considering that the footpoint portions of the loop are responsible for the CME parameter estimation highest emission and highest velocity, (Supplementary Fig. 1), configurations with θ ∼ 90◦ maximize the line Assuming that the O viii emission is entirely due to a shifts, while configurations with θ ∼ 0◦ minimize the CME, we have a direct measurement of the CME aver- line shifts (causing also some line broadening, because age radial velocity, v, and of the CME total luminosity of the simultaneous redshifted and blueshifted contri- in the O viii line, LOVIII, corrected for the interstellar butions originating in the motions of plasma located absorption 37 of 4×1020 cm−2. To infer the temperature near the loop apex). We found that only for φ = 0◦ T of the CME we considered the observed post-flare and θ = 90◦ the predicted shifts are large enough to line ratio between the O vii resonance line at 21.60 Å

7 36 and the O viii Lyα. This ratio provides a lower limit In the estimation of the uncertainty of ne, V , M, and of 3MK for T . In addition, the Fe xvii lines, that form Ekin, we considered only the uncertainty on τ, because mainly at 5 MK, do not show any blueshift, indicating it is significantly larger than the uncertainties on T that they are produced by coronal loops and not by and EM. Only in the computation of the upper limit the CME. We therefore deduced that the CME tem- of the confidence interval of Ekin we included also a fac- perature T should be 4 ± 1 MK. We do not detect any tor related to the possible flare-CME separation angle, significant decline in the O viii line flux. This suggests in agreement with solar observations 1,2 that indicate that the CME likely moves as a coherent structure, at most separations of 45◦. The observed O viii line without experiencing significant adiabatic expansion, does not show significant broadening in the post-flare in spite of the large distance travelled in the related interval, its width is in fact compatible with the instru- time range. mental width. Therefore, velocity dispersion along the The line luminosity together with the plasma tem- line of sight in the CME plasma is expected to be small −1 perature allow us to determine the CME emission mea- (≤ 100kms ), corroborating further the inferred Ekin sure EM as: value.

We finally notice that both M and Ekin are directly L viii EM = O proportional to τ. Therefore, the strict lower limit of AOG(T ) 60 ks, provided by the stable post-flare emission in the where G(T ) is the line emissivity function (APED O viii, corresponds to the lower limits of the confidence 36 database ), and AO is the abundance (we intervals of M and Ekin. Conversely, the already large adopted the abundances inferred from the post-flare upper limit on M is an a posteriori confirmation on the emission 21). adopted upper limit on τ. Moreover, the reasonable The CME emission is observed for ∼ 60 ks. There- assumptions made for the τ confidence interval is also fore, its radiative cooling time τ must be longer. We supported by the ne value inferred, neatly compatible assumed τ = 200 ks, with a confidence interval of a with the density observed in solar CMEs 28,38,39. factor 10, that corresponds to 60ks <τ < 600 ks. By also assuming that, during the observed emission, there Data Availability is no heating source in the CME, that its emission is optically thin, and that the hydrogen to electron den- The Chandra dataset analyzed in this sity ratio nH/ne is 0.83 (value corresponding to high- work (ObsID 1892) can be accessed from temperature plasma with cosmic abundances), we can http://cxc.harvard.edu/. The data that support estimate the CME electron density from: plots and findings of this study are available from the corresponding author upon reasonable request. 3 Eint = (ne + nH) V kBT 2 References ˙ Erad = nenHV Λ(T ) [1] Yashiro, S. & Gopalswamy, N. Statistical relation- ship between solar flares and coronal mass ejec- Eint 3 ne kBT tions. In Gopalswamy, N. & Webb, D. F. (eds.) τ = ⇒ ne = +1 E˙ rad 2 nH  τΛ(T ) Universal Heliophysical Processes, vol. 257 of IAU Symp., 233–243 (2009). where Eint is the CME internal energy, E˙ rad is the radia- tive loss rate, V is the CME volume, and Λ(T ) is the [2] Aarnio, A. N., Stassun, K. G., Hughes, W. J. & radiative loss function per EM unit in the 1 − 2000 Å McGregor, S. L. Solar Flares and Coronal Mass wavelength interval, computed assuming the plasma Ejections: A Statistically Determined Flare Flux emissivities of the APED database 36. The estimates - CME Mass Correlation. Sol. Phys. 268, 195–212 for EM and ne finally allow us to derive the volume V , (2011). mass M, and kinetic energy E of the CME: kin [3] Webb, D. F. & Howard, T. A. Coronal Mass Ejec- EM tions: Observations. Living Rev. Sol. Phys. 9, 3 V = (2012). nenH EM [4] Wright, N. J., Drake, J. J., Mamajek, E. E. & M = VnH m = m ne Henry, G. W. The Stellar-activity-Rotation Re- lationship and the Evolution of Stellar Dynamos. 1 2 Ekin = Mv Astrophys. J. 743, 48 (2011). 2 where m is the mean mass per hydrogen atom. All the [5] Khodachenko, M. L. et al. Coronal Mass Ejection values of the relevant CME parameters are reported in (CME) Activity of Low Mass M Stars as An Im- Supplementary Table 2. portant Factor for The Habitability of Terrestrial

8 Exoplanets. I. CME Impact on Expected Mag- [18] Moschou, S.-P., Drake, J. J., Cohen, O., Alvarado- netospheres of Earth-Like Exoplanets in Close-In Gomez, J. D. & Garraffo, C. A Monster CME Habitable Zones. Astrobiology 7, 167–184 (2007). Obscuring a Demon Star Flare. Astrophys. J. 850, 191 (2017). [6] Aarnio, A. N., Matt, S. P. & Stassun, K. G. Mass Loss in Pre-main-sequence Stars via Coronal Mass [19] Wheatley, P. J. ROSAT observations of V471 Ejections and Implications for Angular Momen- Tauri, showing that stellar activity is determined tum Loss. Astrophys. J. 760, 9 (2012). by rotation, not age. Mon. Not. R. Astron. Soc. 297, 1145–1150 (1998). [7] Drake, J. J., Cohen, O., Yashiro, S. & Gopal- swamy, N. Implications of Mass and Energy Loss [20] Mullan, D. J., Sion, E. M., Bruhweiler, F. C. & due to Coronal Mass Ejections on Magnetically Carpenter, K. G. Evidence for a cool wind from Active Stars. Astrophys. J. 764, 170 (2013). the K2 dwarf in the detached binary V471 Tauri. [8] Osten, R. A. & Wolk, S. J. Connecting Flares and Astrophys. J. Lett. 339, L33–L36 (1989). Transient Mass-loss Events in Magnetically Active Stars. Astrophys. J. 809, 79 (2015). [21] Testa, P., Reale, F., Garcia-Alvarez, D. & Huen- emoerder, D. P. Detailed Diagnostics of an X-Ray [9] Odert, P., Leitzinger, M., Hanslmeier, A. & Lam- Flare in the Single Giant HR 9024. Astrophys. J. mer, H. Stellar coronal mass ejections - I. Esti- 663, 1232–1243 (2007). mating occurrence frequencies and mass-loss rates. Mon. Not. R. Astron. Soc. 472, 876–890 (2017). [22] Borisova, A. et al. The different origins of mag- netic fields and activity in the Hertzsprung gap [10] Noyes, R. W., Hartmann, L. W., Baliunas, S. L., stars, OU Andromedae and 31 Comae. Astron. Duncan, D. K. & Vaughan, A. H. Rotation, Astrophys. 591, A57 (2016). convection, and magnetic activity in lower main- sequence stars. Astrophys. J. 279, 763–777 (1984). [23] Strassmeier, K. G., Serkowitsch, E. & Granzer, T. Starspot photometry with robotic telescopes. [11] Shibata, K. & Magara, T. Solar Flares: Magneto- UBV(RI)_C and by light curves of 47 active stars hydrodynamic Processes. Living Rev. Sol. Phys. in 1996/97. Astron. Astrophys. Suppl. Ser. 140, 8, 6 (2011). 29–53 (1999). [12] Güdel, M. X-ray of stellar coronae. Annu. Rev. Astron. Astrophys. 12, 71–237 (2004). [24] Pizzolato, N., Maggio, A. & Sciortino, S. Evolu- tion of X-ray activity of 1-3 Msun late-type stars [13] Houdebine, E. R., Foing, B. H. & Rodono, M. Dy- in early post-main-sequence phases. Astron. As- namics of flares on late-type dMe stars. I - Flare trophys. 361, 614–628 (2000). mass ejections and stellar evolution. Astron. As- trophys. 238, 249–255 (1990). [25] Testa, P. et al. Geometry Diagnostics of a Stel- lar Flare from Fluorescent X-Rays. Astrophys. J. [14] Vida, K. et al. Investigating magnetic activity in 675, L97 (2008). very stable stellar magnetic fields. Long-term pho- tometric and spectroscopic study of the fully con- [26] Švestka, Z. Speeds of Rising Post-Flare Struc- vective M4 dwarf V374 Pegasi. Astron. Astrophys. tures. Sol. Phys. 169, 403–413 (1996). 590, A11 (2016). [27] West, M. J. & Seaton, D. B. SWAP Observations [15] Vida, K. et al. The quest for stellar coronal of Post-flare Giant Arches in the Long-Duration mass ejections in late-type stars. I. Investigating 14 October 2014 Solar Eruption. Astrophys. J. Balmer-line asymmetries of single stars in Virtual Lett. 801, L6 (2015). Observatory data. Astron. Astrophys. 623, A49 (2019). [28] Landi, E., Raymond, J. C., Miralles, M. P. & [16] Gunn, A. G., Doyle, J. G., Mathioudakis, M., Hara, H. Physical Conditions in a Coronal Mass Houdebine, E. R. & Avgoloupis, S. High-velocity Ejection from Hinode, Stereo, and SOHO Obser- evaporation during a flare on AT Microscopii. As- vations. Astrophys. J. 711, 75–98 (2010). tron. Astrophys. 285, 489–496 (1994). [29] Drake, J. J., Cohen, O., Garraffo, C. & Kashyap, [17] Berdyugina, S. V., Ilyin, I. & Tuominen, I. The V. Stellar flares and the of CMEs. active RS Canum Venaticorum binary II Pegasi. In Kosovichev, A. G., Hawley, S. L. & Heinzel, P. III. Chromospheric emission and flares in 1994- (eds.) Solar and Stellar Flares and their Effects on 1996. Astron. Astrophys. 349, 863–872 (1999). Planets, vol. 320 of IAU Symp., 196–201 (2016).

9 [30] Alvarado-Gómez, J. D., Drake, J. J., Cohen, O., C.A. and J.J.D. contributed to analysis of observa- Moschou, S. P. & Garraffo, C. Suppression of tional data. F.R. contributed to the HD model de- Coronal Mass Ejections in Active Stars by an velopment, and C.A. to the synthesis of the model line Overlying Large-scale Magnetic Field: A Numeri- profile. cal Study. Astrophys. J. 862, 93 (2018). Competing interests [31] Huenemoerder, D. P. et al. TGCat: The Chandra Transmission Grating Data Catalog and Archive. The authors declare no competing financial interests. Astron. J. 141, 129 (2011).

[32] Argiroffi, C. et al. Redshifted X-rays from the Additional information material accreting onto TW Hydrae: Evidence of a low-latitude accretion spot. Astron. Astrophys. Correspondence and requests for materials should be 607, A14 (2017). addressed to C. Argiroffi.

[33] Brickhouse, N. S., Dupree, A. K. & Young, P. R. X-Ray Doppler Imaging of 44i Bootis with Chan- dra. Astrophys. J. 562, L75–L78 (2001).

[34] Chung, S. M., Drake, J. J., Kashyap, V. L., Lin, L. W. & Ratzlaff, P. W. Doppler Shifts and Broad- ening and the Structure of the X-Ray Emission from Algol. Astrophys. J. 606, 1184–1195 (2004).

[35] Ishibashi, K., Dewey, D., Huenemoerder, D. P. & Testa, P. Chandra/HETGS Observations of the System: The Primary as a Dominating X- Ray Source. Astrophys. J. 644, L117–L120 (2006).

[36] Smith, R. K., Brickhouse, N. S., Liedahl, D. A. & Raymond, J. C. Collisional Plasma Models with APEC/APED: Emission-Line Diagnostics of Hydrogen-like and Helium-like Ions. Astrophys. J. 556, L91–L95 (2001).

[37] Singh, K. P., Drake, S. A., White, N. E. & Simon, T. ROSAT Observations of Five Chromospheri- cally Active Stars. Astron. J. 112, 221 (1996).

[38] Cheng, X., Zhang, J., Saar, S. H. & Ding, M. D. Differential Emission Measure Analysis of Multi- ple Structural Components of Coronal Mass Ejec- tions in the Inner Corona. Astrophys. J. 761, 62 (2012).

[39] Landi, E., Miralles, M. P., Raymond, J. C. & Hara, H. Hot Plasma Associated with a Coronal Mass Ejection. Astrophys. J. 778, 29 (2013).

Acknowledgements

The authors acknowledges modest financial contribu- tion from the agreement ASI-INAF n.2017-14.H.O.

Author contributions

C.A., F.R., J.J.D., A.C., P.T., R.B., M.M., S.O., G.P. contributed to scientific discussion and text writing.

10 Supplementary Information

Supplementary Table 1: Selected emission lines

a b c d e Index λ Elem. Tmax FMEG FHEG (Å) (MK) (cts) (cts) 1a 4.7274 SXVI 25.1 66±15f 21±10 1b 4.7328 SXVI 25.1 ······ 2a 6.1804 SiXIV 15.8 505±31f 231±21f 2b 6.1858 SiXIV 15.8 ······ 3 6.6479 SiXIII 10.0 218±24 83±14 4 6.7403 SiXIII 10.0 185±23 83±14 5a 7.1058 MgXII 10.0 115±23 34±11 5b 7.1069 MgXII 10.0 ······ 6a 8.4192 MgXII 10.0 345±26f 228±20f 6b 8.4246 MgXII 10.0 ······ 7 9.1687 MgXI 6.3 136±21 35±11 8 15.0140 FeXVII 5.0 62±14 12±7 9a 16.0055 OVIII 3.2 52±13 4±6 9b 16.0067 OVIII 3.2 ······ 10 16.7800 FeXVII 5.0 53±12 7±6 11 17.0510 FeXVII 5.0 55±12 5±6 12 17.0960 FeXVII 5.0 56±12 4±6 13a 18.9671 OVIII 3.2 112±14f 4±6 13b 18.9725 OVIII 3.2 ······ 14 21.6015 OVII 2.0 8±8 ··· 15a 24.7792 NVII 2.0 21±9 ··· 15b 24.7846 NVII 2.0 ······ a Lines indicated with the same index number, but different letters, are the two components of H-like resonance doublets. b Predicted wavelength from the APED database. c Temperature of maximum formation of the line. d Line counts observed in the MEG spectrum during the whole observation. e Line counts observed in the HEG spectrum during the whole observation. f Lines for which a significant shift was detected in at least one time interval.

11 Supplementary Figure 1: Flaring loop model. Time evolution of velocity along the loop (a), temperature (b), and emission measure (c) in the flaring loop, as predicted by the HD model. Negative velocities (marked in blue) correspond to upward motion in the loop, positive velocities (marked in red) correspond to downward motions.

Supplementary Table 2: CME properties

parameter units value v (kms−1) −90 ± 30 T (MK) 4.0 ± 1.0 +2.6 21 M (g) (1.2−0.8) × 10 +27.7 34 Ekin (erg) (5.2−3.6 ) × 10 −3 +11.8 8 ne (cm ) (5.5−3.7 ) × 10 +400 τ (ks) 200.0−140 EM (cm−3) (2.8 ± 1.0) × 1053 3 +10.3 36 V (cm ) (1.1−1.0 ) × 10

12 Supplementary Figure 2: Comparison between observed and predicted Doppler shifts for lines showing no significant shifts. Symbols are the same as in Fig. 3.

13