The Galactic Center Cloud G2 -- a Young Low-Mass Star with a Stellar Wind
aa r X i v : . [ a s t r o - ph . H E ] M a r Draft version October 1, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
THE GALACTIC CENTER CLOUD G2 – A YOUNG LOW-MASS STAR WITH A STELLAR WIND
N. Scoville , and A. Burkert (Dated: Accepted ApJ 2/25/13) Draft version October 1, 2018
ABSTRACTWe explore the possibility that the G2 gas cloud falling in towards SgrA ∗ is the mass loss envelopeof a young TTauri star. As the star plunges to smaller radius at 1000 to 6000 km s − , a strong bowshock forms where the stellar wind is impacted by the hot X-ray emitting gas in the vicinity of SgrA ∗ .For a stellar mass loss rate of 4 × − M ⊙ per yr and wind velocity 100 km s − , the bow shockwill have an emission measure ( EM = n vol ) at a distance ∼ cm, similar to that inferred fromthe IR emission lines. The ionization of the dense bow shock gas is potentially provided by collisionalionization at the shock front and cooling radiation (X-ray and UV) from the post shock gas. Theformer would predict a constant line flux as a function of distance from SgrA ∗ , while the latter willhave increasing emission at lesser distances. In this model, the star and its mass loss wind shouldsurvive pericenter passage since the wind is likely launched at 0 . ∼ M ⊙ ). In this model, the emission cloud willprobably survive pericenter passage, discriminating this scenario from others. Subject headings: accretion black hole physics ISM: clouds Galaxy: center INTRODUCTION
The recently discovered G2 cloud which is infallingtoward SgrA ∗ is a most intriguing astronomical discov-ery (Gillessen et al. 2012) – both its origin and natureare unclear as yet. Nevertheless, in the space of a fewyears from the first detection, one will observe its passagewithin ∼ ∗ will increase andover what timescale.G2 was first observed in the HI Br γ and Br δ HI and2.058 µ m HeI emission lines and detected in the near in-frared continuum with an extremely low 550K color tem-perature (Gillessen et al. 2012; Eckart et al. 2013). Overthe period 2004 to 2012 its 3d velocity has increased from1200 to over 2500 km s − (Gillessen et al. 2013). Thelatest orbital determination indicates an eccentricity of0.966 and pericenter passage at 2 × cm from SgrA ∗ ,when the 3d velocity will be 6340 km s − (Gillessen et al.2013). The orbital period is 198 yrs with an apocenterdistance of 1 . × cm and velocity 108 km s − .The observed flux in the Br γ line requires an ionizedgas emission measure EM = R n e dvol ∼ cm − .Surprisingly, the line flux exhibits no change greater than10% over the 4 yr period (Gillessen et al. 2012, 2013).The cloud is resolved along its orbital path but un-resolved in the transverse direction (i.e. ≤ cm);Gillessen et al. (2013) adopt an effective spherical radius California Institute of Technology, MC 249-17, 1200 EastCalifornia Boulevard, Pasadena, CA 91125 University Observatory Munich, Scheinerstrasse 1, D-81679Munich, Germany Max-Planck-Fellow Max-Planck-Institute for Extraterrestrial Physics, Giessen-bachstrasse 1, 85758 Garching, Germany . × cm for the emitting region and thereby de-duce a mean density of 6 × cm − and a total mass ∼ earth (assuming unity volume filling for the ionizedgas). If this is the whole story (i.e. G2 has only the massseen in the ionized gas and it is uniformly distributed),then it is clear that the cloud can not survive pericen-ter passage since the Roche limit for tidal stability is n ∼ . × cm − .Several models have been proposed for the origin andnature of G2. Gillessen et al. (2012), Schartmann et al.(2012) and Burkert et al. (2012) have suggested that itmay have formed as an interstellar cloud from collidingstellar winds in the young stellar ring at 2 × cmradius. Meyer & Meyer-Hofmeister (2012) suggest it isa ring of gas formed by a Nova explosion. These expla-nations might account for the low angular momentumand high eccentricity orbit. Burkert et al. (2012) modelthe subsequent hydrodynamic evolution of the cloud as itfalls toward SgrA ∗ – yielding reasonable agreement withthe observed emissivities and kinematic evolution. Alter-natively, Murray-Clay & Loeb (2011) proposed that G2is a star with a protoplanetary disk, also scattered out ofthe young stellar ring (presumably from a triplet star sys-tem). Then as the system descends towards SgrA ∗ , thedisk is photo-evaporated and tidally disrupted to producethe G2 cloud. With the orbital parameters known atthe time they proposed this model, they argued that theouter protoplanetary disk at 5 - 10 AU would probablysurvive pericenter passage. However, with the most re-cent orbit determination (Gillessen et al. 2013), the diskwill now probably be tidally stripped to within 2 AUradius at pericenter. In both of the above scenarios, itseems we are then extraordinarily fortunate to be ob-serving a one-off event (i.e. a single orbital event) onlynoticed within a few years of its final demise. And yetthere do not appear to be large numbers of similar ob-jects further out.Here we explore a different scenario – G2 being ayoung low mass TTauri star, formed in the young stel- Scoville & Burkertlar ring and subsequently injected into the eccentricorbit. Many TTauri stars have mass-loss winds at200 - 500 km s − with ˙ M ∼ − × − M ⊙ yr − during their first million years. The mass-loss ratesfor those with measured rates ( ∼
50% of the sam-ple) have a very large range 10 − . − M ⊙ yr − (Hartigan et al. 1995) and a median value 2 . × − M ⊙ yr − . White & Hillenbrand (2004) obtainedmedian values 10 − . and 10 − . M ⊙ yr − for samples of8 and 42 class I and II TTauri stars, respectively. Theseoutflows may originate as a centrifugally driven windfrom the inner accretion disk (e.g. Blandford & Payne1982). If so, then the observed velocities imply a launchradius well inside 1 AU radius from the star. Thisscenario for G2 has a superficial similarity to that ofMurray-Clay & Loeb (2011) in having a young stellarobject formed in the young stellar ring and having acircumstellar disk, but is very different in the physicsof the mass-loss material and the possibility of pericen-ter survival. In the case of the TTauri star wind, theoutflow velocities ∼ − are much larger than the10km s − expected for a photo-evaporating disk with ve-locities ∼ − . For the TTauri star wind, a verydense bow shock is formed at radius ∼ cm and it isthis gas which produces the observed emission lines. Thestellar wind plus bow shock model readily reproduce theobserved emission line fluxes using standard TTauri starwind parameters.In the following we analyze the mass-loss wind param-eters and the interaction with the hot X-ray emitting gasin the vicinity of SgrA ∗ , followed by a detailed numeri-cal model for the bow shock at the upstream side of theplunging star. This allows us to track the evolution ofthe bow shock structures and their emissivity as a func-tion of distance from SgrA ∗ . Throughout most of theorbit, tidal stripping is not very significant since the bowshock on the front side of the star is pushed to smallerradii from the star as the orbit approaches the centralblack hole (due to the higher stellar velocity and higherdensity of the ambient hot X-ray gas at small galacticradii).Using the derived density structure, we then ana-lyze the ionization of the gas, concluding that the mostlikely source of the observed emission lines is the densebow shock where the outflowing stellar wind meets the10 − K shocked layer of ambient gas. The ionizationmay be provided by photons in the free-free continuumand line emission of the gas cooling behind the shock. Inaddition, there will be collisional ionization as the windmaterial passes through the shock at 200 - 500 km s − .The expected Lyman continuum from young stars in thecentral parsec does not appear to be adequate. If theionization is collisional it would account for the apparentconstancy of the emission line fluxes, since the mass-lossrate is probably constant; however, unless the mass-lossrate is > − M ⊙ yr − this ionization is insufficient toaccount for the inferred emission measures. SPHERICAL MASS-LOSS WIND AND INTERACTIONWITH CORONAL X-RAY EMITTING GAS
As the low mass star with a stellar wind descends to-wards SgrA ∗ , the outer envelope will interact with theambient hot X-ray emitting gas (Baganoff et al. 2003;Muno et al. 2004; Shcherbakov & Baganoff 2010). For the stellar wind, we adopt the following fiducial numbers: ˙ M = 4 × − M ⊙ yr − ≡ ˙M ∗ (1)and V W = 100 km s − . (2)which implies a density distribution in the mass-envelope, ρ W = ˙ M πr V W = 9 . × − ˙ M ∗ r AU V W gr cm − (3) n W ≃ × H cm − at 1 AU . (4)For the stellar wind we adopt an inner launch radiuswhich scales with the wind velocity, assuming the out-flow velocity is equal to the escape velocity for materialinitially in circular orbit at the launch radius. This is agood approximation for outflow where the radial accel-eration is gradual (rather than explosive). Figure 1.
The stellar 3d velocity as a function of distance fromSgrA ∗ . The adopted density, temperature and ram pressure of thehot X-ray emitting gas are also shown. For the hot ambient medium, we use the radial densityand temperature distributions given by Burkert et al.(2012), which are similar but not identical to the X-ray model of Yuan et al. (2003). The temperature dis-tribution was assumed appropriate to hydrostatic equi-librium of the gas in the potential of the 4 . × M ⊙ SgrA ∗ black hole. Specifically, we adopt : ρ hot = 9 . × − (cid:18) cm r (cid:19) gr cm − (5)and T hot = 2 × (cid:18) cm r (cid:19) K . (6)The above ignores the possibility that some of the X-ray emission is from stellar sources (Sazonov et al. 2012).These distributions are shown in Fig 1 together with the2 33d stellar velocity and the resultant ram pressure fromthe hot X-ray emitting gas. Figure 2.
The stagnation radius in front of the star (Eq. 8) andthe Roche radius for tidal stripping are shown as a function ofdistance from SgrA ∗ . The Roche radius was calculated assuming astar mass of 2 M ⊙ . Inside 10 cm radius from SgrA ∗ , the emissionregion is stable against tidal disruption, although tides will shearthe tail out behind the star. The outflowing stellar wind will terminate on the up-stream side of the star in a shock front at the point whereits ram pressure equals the thermal pressure of the hotambient gas or the ram pressure of the hot gas. In theportion of the orbit where G2 is currently observed, itis plunging with velocity significantly greater than thelocal circular, virial velocity; thus the ram pressure islikely to dominate the hot gas thermal pressure (sincethe thermal temperature of this gas was estimated as-suming virial equilibrium in the central potential, whichis dominated by the black hole at these radii).The bow shock of the stellar wind against the hotmedium will have a stagnation or standoff radius in frontof the star (balancing ram pressures) at : R s = ˙ M ∗ V W πρ H V ∗ (7)where ρ H ∼ − gr cm − is the mass density in the hotmedium at an orbital distance ∼ cm, correspondingto the position of G2 in mid 2012. For mid 2012, V ∗ =2000 km s − and therefore R s = 2 . × cm " ˙M ∗ V W100 ρ H − V ∗ / ∼
14 AU . (8)At this radius the wind density is n W = 3 × H cm − .For the mass-loss star moving supersonically throughthe hot medium near SgrA ∗ , the bow shock on the up-stream side will occur at R s and a conical compressed gaslayer will extend downstream from the star. In fact therewill be two shocked layers, the first where the ambienthot gas meets the stellar wind bow shock, and the second,an interior bow shock, where the outflowing stellar wind meets the compressed gas at the stagnation point on theupstream side of the star. We will refer to these as the hot bow shock and the cold bow shock respectively(Fig. 3). The immediate post shock gas temperatures aregiven by T = 1 . × ( V shock /
100 km s − ) K for anionized gas (McKee & Hollenbach 1980), implying tem-peratures of 10 − and ∼ × K, respectively, behindthe two bow shocks.The first shock front will be adiabatic since this veryhot gas cools slowly. The density in this shock is shownin Fig. 4 and is ∼ − . The second shock fronthas a very high post shock density ( ∼ cm − ), allow-ing it to cool in just 10 − yrs (comparable to the soundcrossing time for the cold bow shock). This shock will bemodeled very approximately as isothermal. Interior tothis second, isothermal shock which is a very thin sheetof gas, lies the free streaming stellar wind.In the context of this model, there are clearly several lo-cations from which observed ionized emission lines mightarise: 1) the high density, interior, cold bow shock; 2)the stellar mass-loss envelope below the bow shock and3) the outer, hot bow shock. The latter is not a signifi-cant source given the very high temperatures (and hencelow recombination rate) and its relatively small emissionmeasure ( n L ). To evaluate the expected emissivitiesfrom the mass-loss envelope and the cold bow shock, wedevelop a detailed model for the bow shock structure in § § BOW SHOCK MODEL
Figure 3.
The computed density structure of the mass-loss enve-lope and bow shocks are shown when the star is 10 cm from SgrA ∗ . The density is shown with logarithmic scaling from n = 100to 10 cm − in grayscale. The stellar velocity at this radius is3278 km/s and the ambient medium density of the X-ray emittinggas is 400 cm − . Two bow shocks are formed – a thick one in theshocked hot upstream medium and a very thin, high density coldlayer in the shocked stellar wind. For this model, V wind = 100km s − and ˙ M = 4 × − M ⊙ yr − To model the bow shock region, we make use of theanalytic treatment by Dyson (1975) modeling a fast stel-lar wind ablating an expanding dust globule. Dyson de-veloped two limiting cases : 1) with mixing of the twoshocked layers (hot gas and cold gas) and 2) withoutmixing of the shocked layers, resulting in a full tangen-tial discontinuity between the layers. In the following we Scoville & Burkertuse the latter approximation. The 3d velocity of the starrelative to the ambient medium is taken from the latestorbit of G2 given by Gillessen et al. (2013) and the am-bient medium density and temperature as a function ofdistance from SgrA ∗ were taken from Eq. 2 and all areshown in Fig. 1.Inside ∼ × cm distance from SgrA ∗ , the mass-lossstar is plunging supersonically toward SgrA ∗ . At theseradii the star is essentially in free-fall, whereas the hotgas is assumed to be in hydrostatic equilibrium (hencehaving thermal sound speed similar to the circular veloc-ity). The supersonic motion of the star through the hotgas will result in the two bow shocks mentioned earlier:the cold bow shock in the stellar wind material just insidethe stagnation radius (see Fig 2), and the hot bow shockin the hot ambient medium just outside the stagnationradius. Figure 3 shows these shocks and the stellar enve-lope structure as computed for a distance of 10 cm fromSgrA ∗ when the star is moving at 3300 km s − through theambient medium of density ∼
440 cm − . The cold bowshock has a thickness of only ∼ × at this pointand so it is hardly visible in Fig. 3 but its density is 10 cm − , so its potential emission measure ( n L ) is large.We say ’potential’ since it is also required that the gas beionized if it is to account for the observed line emission. Figure 4.
The density of the cold and hot bow shocks are shownas a function of distance from SgrA ∗ . [The density in the coldshock is independent of V W and ˙ M .] The structure and physical conditions in the shockedenvelope were computed over the full stellar orbit as-suming stellar wind parameters of V wind = 100 km s − ,˙ M = 4 × − M ⊙ yr − . Figures 4 and 5 show the de-rived densities and thicknesses (perpendicular to the bowshock) for the two shocks as a function of the distancefrom SgrA ∗ . Over most of the orbit the cold shock den-sity is 4 orders of magnitude higher than that of the hotshock, while the thickness of the hot shock is only ∼ EM ∝ n vol ), provided the gas is ionized. Figure6 shows the tangential flow velocities in the cold and hotshow regions. (These velocities are parallel to the bow Figure 5.
The thickness of the cold and hot bow shocks are shownas a function of distance from SgrA ∗ . [The shock thickness ofthe cold bow shock scales approximately as V − . W ˙ M . and as V / W ˙ M / for the hot shock.] Figure 6.
The tangential velocities in the cold and hot bow shocksare shown as a function of distance from SgrA ∗ . [The cold shocktangent velocity will scale linearly with V W and is independent of˙ M . The velocity in the hot shock is independent of both parame-ters.] shock.)The gas in the hot shock layer flows around the coldshock layer tangentially at velocities ∼ − .The mass-flux of hot shock material intercepting the coldshock provides an upper limit to the cold shock ablationrate. For hot gas densities ∼ cm − and an effectiveradius of 10 cm for the cold shock, this yields a maxi-mum ablation rate of 10 − M ⊙ yr − which is two ordersof magnitude less than the wind mass-loss rate. Ablationis therefore probably not significant.The derived emission measures (EM) for the cold shockand the mass-loss envelope integrated down to the launchradius ( ∼ .
18 AU for V wind = 100km s − ) are shown inFig. 7 as a function of distance from SgrA ∗ . For thebow shock, the EM was calculated only out to radius ∼ × , i.e. the area shown in Fig. 3. This clearly is a2 5 Figure 7.
The integrated emission measure ( n vol ) is shown forthe cold bow shock and the stellar envelope as a function of distancefrom Sgr A ∗ . For the mass-loss envelope n vol is constant with avalue determined by the adopted mass-loss, and very importantly,by the inner radius adopted for the outflow (see discussion followingEq. 2). At distance 10 cm from Sgr A ∗ in mid 2012, the totalemission measure for the ionized gas is ∼ cm − . [The emissionmeasure from the cold shock scales as V − / W ˙ M / .] lower limit since there will be significant additional EM inthe extended downstream tail of the bow shock. The EMvalues at ∼ cm (or 0.1 ′′ ) from SgrA ∗ are in reasonableagreement with the Gillessen et al. (2012) value of ∼ cm − for the ionized gas, considering that we have notincluded the downstream tail. The EM of the bow shockis also ∼ × − M ⊙ yr − . In the foregoing discussion we have ignoredthe issue of whether the gas in the 3 zones will actuallybe ionized and clearly that is a critical consideration. IONIZATION
Ionization of the material may potentially be providedby: UV photons from hot stars in the central few parsecs,UV/X-ray photons from the hot plasma in the centralparsec or the hot gas (10 − and 10 − K) in the two bowshocks, or collisions at the inner bow shock where theenvelope material moving at 100 - 500 km s − is shocked.In the case of photoionization by extended UV sourcessuch as the hot stars and the X-ray emitting plasma, itis important to recognize that it is not simply a matterof counting photons; one must also estimate the flux intothe emission region. If the region of high EM is compact,only a small fraction of the available photons will actuallybe intercepted.In Fig. 8 we show the n > ∗ ) and collisional ionization due to mass-lossmaterial passing through the bow shock . In the case ofcooling radiation from the hot plasma regions we have in-cluded only the free-free continuum, not the line cooling Figure 8.
The black dashed line shows the maximum hydrogenrecombination rate (to n >
1) for the cold bow shock (= n e volα B ,i.e. assuming complete ionization of the shell) as a function ofdistance from SgrA ∗ , together with curves showing the possiblesources of ionization: collisional ionization of stellar wind materialat the bow shock; free-free photons at energy greater than 13.6 evfrom the hot X-ray emitting ambient medium, and from the hotand cold bow shock layers; and Lyman continuum photons fromhot stars in the inner parsec. which dominates by a factor of a few at 10 − K, and forwhich some of the photons can ionize H (Shull & McKee1979). Although these ionization estimates are quite ap-proximate, it is clear that the only viable sources arethe collisional ionization at the bow shock and the gascooling within the ’cold’ bow stock where T ∼ − K.The Lyman continuum from the young stars in thegalactic nucleus is not well constrained but we adoptedthe equivalent of a O5 star within the central 1 pc ra-dius, i.e. a Lyman continuum production rate of Q =10 Lyc s − . This production rate implies an averagephoton flux of ∼ × Lyc cm − s − and integrat-ing over the bow shock area we arrive at the estimategiven in Fig. 8. Although the production rate of Lymancontinuum photons from hot stars is large, the area ofthe bow shock is small, so very few of them will be in-tercepted. Similarly, although there is a potentially largeEM at the base of the mass-loss envelope, the interceptedarea of that region at radius 0.18 AU is so small that thematerial would not be ionized, except by UV from thestellar accretion shock. The latter may of course be sub-stantial since the TTauri star winds exhibit ionized gasemission lines and some of the gas arriving at the bowshock may already be ionized. Constancy of the Emission Line Fluxes
Gillessen et al. (2013) report that the observed linefluxes for the Br γ emission are constant to within10% over the 4 year period 2008.5 to 2012.5. This isvery surprising, given the fact that its distance fromSgrA ∗ changed from 3.6 to 1.3 × cm and its 3d ve-locity increased from 1500 to 2900 km s − . In almostany model, one would expect the gas mass and/or theexcitation of the emission in G2 to correlate with the dis- Scoville & Burkerttance from SgrA ∗ and/or infall velocity. In the context ofthe model proposed here, the approximate constancy ofthe flux might be understood if the dominant source ofionization is collisional, as stellar wind material from theinside passes through the cold bow shock at 100 to 500km s − . In this case, the total number of ionizations persecond (hence the line emission flux) will be constant andsimply a few times the number of atoms passing throughthe shock front. This number flux is ∼ . × s − for ˙ M = 4 × − M ⊙ yr − . Since the HI ionizationenergy of 13.6 eV corresponds to an HI particle veloc-ity of 50km s − , this estimate should be multiplied bya factor ∼ ( V w / − ) to account for the energyavailable in the shock compared to what is needed toionize HI. In Fig. 8, we have taken this factor to be ∼ > . ∼ sec − ), similar to what is needed to maintain theobserved emission. If the ionization is not due to colli-sions at the bow shock, it is unclear how to explain theconstancy of the emission line fluxes since all the othersources of ionization vary with distance from SgrA ∗ .It is worthwhile noting that some variation in the emis-sion line fluxes is a desirable feature of any model inwhich G2 comes in from larger radii. If the fluxes arerelatively constant with radius, one should expect to seea large number of similar emission regions at the largerradii – so far these have not been seen. A drop off inthe expected emissivity out beyond several 10 cm (asshown in Fig. 7) is therefore a desirable feature, in thatit reduces the visibility of such precursors. THE STAR AND ITS ORBIT
Eckart et al. (2013) report detection of an object theycall DSO in the K and L band continua with position andproper motion similar to G2. The source has a very lowcolor temperature, ∼
500 K, and the K-band magnitudeis ∼ . M ⊙ TTauri stars, M K ∼ − ∼ m K ∼
20 mag. Thus,the star itself, if it is a TTauri star, would be very diffi-cult to directly detect unless it is in a period of enhancedactivity. In fact, Eckart et al. (2013) have suggested thatG2/DSO is a dust enshrouded star. One would also ex-pect there to be mass-loss red giant stars in the galacticnucleus; however, such stars can probably be ruled outfor G2 since their brightness would be higher than theobserved L-band flux.The origin of the high eccentricity orbit of G2 remainspoorly understood. Although one might posit stellar in-teractions in a triplet system, these would generally bedisruptive of any circumstellar material. The injection toa highly eccentric orbit from circular orbit requires a ve-locity kick of ∼
500 km s − . To provide such an impulsewith a single star-star scattering would require a closeapproach well inside 1 AU. Such close encounters wouldcertainly disrupt any large protoplanetary disk such asthat invoked by Murray-Clay & Loeb (2011) – and possi-bly also the inner disk from which the TTauri star windsare launched (as discussed here). In the face of such difficulties, it is attractive to con-sider the possibility that the young stars would haveto be formed in eccentric orbits through collisions ofgas clumps with cancelation of angular momentum (Alig et al. ◦ to the Galactic plane(Jackson et al. 1993; Christopher et al. 2005) – both ofwhich indicate substantial non-planar, non-circular dy-namics for the ISM there.Alternatively, the increase in eccentricity might bebuilt up by a series of many smaller amplitude scatter-ings (e.g. Murray-Clay & Loeb 2011). Is it possible thata large increase in the eccentricity from an initial circularorbit could be induced similar to the Kozai oscillationsin exo-planetary systems? If the star was formed in theyoung stellar ring at 2 × cm radius, the orbital pe-riod is ∼
200 yr. The star and any companions couldhave then orbited the galactic center ∼ times. Massclumps associated with both the circumnuclear gas diskand stars might possibly provide perturbations to initiatethe process. IMPLICATIONS
Figure 9.
The mass deposition from the stellar mass loss is shownas a function of distance from SgrA ∗ , obtained by integrating themass-loss along the orbit. The usually assumed mass for G2 of 3 M ⊕ was derivedby Gillessen et al. (2012) assuming the gas is distributedhomogeneously with density ∼ × cm − . For ourmodel the emission measure is produced by gas at density ∼ cm − , resulting in a decrease in the required massof emitting gas by a factor ∼ ∗ . Approximately 0.1 M ⊕ is depositedin the vicinity of SgrA ∗ .It would be a shame if this object, which so intriguesus now, were to disappear this September at pericenter.The model proposed here looks to a brighter future. Atpericenter, the tidal radius is reduced to ∼ −3 AU –this major disruption in the mid-radii of the disk will2 7result in stripping to the exterior and some depositionto the interior disk – inside 1 AU. The latter could re-sult in greatly enhanced stellar mass-loss rates – hencemuch brighter emission at and after pericenter passagefor several years.We would like to thank Lynne Hillenbrand for helpfuldiscussions on the properties of TTauri stars and AndreasEckart for suggestions. AB thanks Caltech for the hos-pitality of a visit which stimulated this project. We ac-knowledge discussions with Alessandro Ballone and MarcSchartmann who are further developing this model withdetailed numerical simulation.