HIGH-SPEED BULLET EJECTIONS DURING THE AGB-TO-PLANETARY NEBULA TRANSITION: HST OBSERVATIONS OF THE CARBON STAR, V HYDRAE

We examine the emergence and early movement of bullet 3 by an inspection of the on-source blob emission during epochs 4–6 (Figures 5(b), 6(b) and 7(b)). During these epochs, we see a change in the blob position–velocity structure in the central slits, from one with no velocity gradient in epoch 4 to one with a clear gradient of about 200 $\mathrm{km}\,{{\rm{s}}}^{-1}$ arcsec⁻¹ in epoch 6 (such that the more distant regions of the blob are blueshifted relative to the less distant ones, implying differential acceleration across the bullet). The radial velocity of the peak emission in the blob also varies, increasing (in magnitude) from about −160 $\mathrm{km}\,{{\rm{s}}}^{-1}$ in epoch 4 to −200 $\mathrm{km}\,{{\rm{s}}}^{-1}$ in epoch 5, then decreasing to −180 $\mathrm{km}\,{{\rm{s}}}^{-1}$ in epoch 6 (Table 2).

Since there is no proper motion during epochs 4–5, we infer that bullet 3 moved roughly toward us along the line-of-sight during most of the time between epochs 4 and 5, accelerating at a rate of about 40 $\mathrm{km}\,{{\rm{s}}}^{-1}$ yr⁻¹. Combining the proper motion between epochs 5–6 (which implies a tangential velocity, V_t = 76 $\mathrm{km}\,{{\rm{s}}}^{-1}$), with the radial velocity⁷ in epoch 6, ${V}_{r}=-173$ $\mathrm{km}\,{{\rm{s}}}^{-1}$, we find that (a) the bullet's motion tilted away from the line-of-sight (and thus toward the overall symmetry-axis of V Hya's extended high-velocity bipolar outflow, see Figure 4 of Hirano et al. 2004) by about 23°, sometime between epochs 5 and 6, and that (b) the 3D outflow velocity in epoch 6 was −189 $\mathrm{km}\,{{\rm{s}}}^{-1}$, which is almost the same as that⁷ measured during epoch 5 (−193 $\mathrm{km}\,{{\rm{s}}}^{-1}$). Thus there is little or no change in the speed of the bullet during epochs 5 and 6. A schematic of bullet 3's movement is shown in Figure 10. We defer the investigation of possible mechanisms for the dynamical evolution of the bullet (e.g., a latitudinal density gradient in the ambient circumstellar medium in which the bullet is propagating, or a magnetic field) to a following paper (S. Scibelli et al. 2016, in preparation).

Zoom In Zoom Out Reset image size

For bullet 2, we find that its radial velocity in epoch 3 (−139 $\mathrm{km}\,{{\rm{s}}}^{-1}$) is significantly lower (in magnitude) than when it is next seen as a detached blob in epoch 4 (−180 $\mathrm{km}\,{{\rm{s}}}^{-1}$). Although we cannot reconstruct the 3D motion of bullet 2 soon after ejection as we have done above for bullet 3, if we assume these motions are similar, then the change in radial velocity for bullet 2 from epoch 3–4 could partly be due to a period of acceleration after ejection, and partly due to a change in inclination angle.

The size of the on-source blob (bullet 3) changes during epochs 4–6; during epoch 4, in addition to the emission seen in the central slit, weak [S ii] emission is seen in slit ${S}_{+2b}$, but not in ${S}_{-2b}$ from this blob. During epoch 5, the blob is seen in all 3 slit locations (Figure 6), indicating that its size has increased and its extent has become resolved. Assuming a Gaussian cross-section for it in the N–S direction, the ratio of the average of the peak intensities seen in the ${S}_{+2b}$ and ${S}_{-2b}$ slits to that in the center slit (0.025), implies that its FWHM size is $0\buildrel{\prime\prime}\over{.} 17$. During epoch 6, no [S ii] emission can be detected from the on-source blob in the off-center slits—the ratio of the 3σ upper limit for the off-center intensity to the intensity in the central slit is about 0.007. This implies that the (projected) emitting extent of bullet 3, in the N–S direction, decreases from epoch 5–6. Along the slit direction, i.e., E–W, the half-power size is $0\buildrel{\prime\prime}\over{.} 12$, $0\buildrel{\prime\prime}\over{.} 13$, and $0\buildrel{\prime\prime}\over{.} 11$, in epochs 4, 5 and 6, which is very similar to the FWHM size of the continuum source measured in the on-source slits ($0\buildrel{\prime\prime}\over{.} 13$). Making the reasonable assumption that the continuum source is unresolved, we conclude that emission from bullet 3 is also unresolved along the E–W direction. Its deconvolved size along the N–S direction in epoch 5 is about $0\buildrel{\prime\prime}\over{.} 11$, or 44 au at D = 400 pc.

There are qualitative and quantitative differences between the bullets in periods 1 and 2. Comparing the detached and distant blobs in these periods, we see an interesting pattern related to their locations—(a) in Period 1, the detached blob (bullet 1) is seen predominantly in the slit immediately south of center (i.e., ${S}_{-2b}$ or ${S}_{-1t}$), whereas in Period 2, the detached blob (bullet 2) is seen predominantly in the slit north of center (i.e., ${S}_{+2b}$), and (b) in Period 1, the distant blob (bullet 0) is seen dominantly in slit ${S}_{+2b}$, whereas in Period 2, the distant blob (bullet 1) is seen dominantly in slit ${S}_{-2b}$. This implies that the bullet-ejection axis flips periodically between the east-southeast (ESE) and east-northeast (ENE) directions, in the plane of the sky. This flip–flop pattern is represented schematically in Figure 9.

In addition, the observed radial-velocities of the blobs show that the bullet-ejection axis also flip–flops in a plane that is perpendicular to the sky-plane. First, the radial velocities of the on-source blobs in Period 2 (bullet 3) are significantly more blueshifted than that of the on-source blob in Period 1 (bullet 2). Second, the radial velocities of the detached blobs in Period 2 (bullet 2) are significantly less blueshifted compared to those of the detached blobs in Period 1 (bullet 1). Third, the radial velocity of the distant blobs in Period 2 (bullet 1) appears to be more blueshifted than that of the distant blob in Period 1 (bullet 0). All of these observational features associated with the flip–flop phenomenon are summarized in Table 5.

Interestingly, ground-based observations of the [S ii]$\lambda 4069.7$ line obtained during 1986 December–1990 May by Lloyd Evans (1991) show a radial velocity of −166 $\mathrm{km}\,{{\rm{s}}}^{-1}$, consistent with the less blueshifted radial velocity of bullet 0 (ejected around 1986). Due to the typical $\sim 1^{\prime\prime}$ ground-based seeing, these data would also have included emission from the the higher blueshifted radial-velocity bullet expected to have been ejected in 1977.5, but its contribution would be relatively minor as it would be much fainter than the more recently ejected bullet, given that the blob emission intensity decreases rapidly once the bullet has moved far away from the central source.

Other systematic differences between blobs in different periods include the following: (a) the on-source blob in Period 1 is clearly seen only in one epoch (epoch 3), whereas in Period 2 it is seen in all three epochs (epochs 4, 5, and 6), (b) the blueshift of the detached blob is about 25% higher in Period 1 compared to Period 2, and (c) the average brightness of the detached blob is about a factor 5 higher in Period 1 compared to Period 2.

We also find that the on-source blob in epoch 3 shows a velocity-gradient across its length, with a blueshift that decreases with spatial offset from the center. In contrast, the on-source blob in Period 2 does not show a velocity gradient in epoch 4, and two years later, when it shows a clear velocity gradient, it is in the opposite sense to that in Period 1, i.e., the blueshift increases with spatial offset from the center.

We find that in order to reproduce the observed [Fe ii]λ4245/[S ii]λ4069.7 or [Fe ii]λ4288/[S ii]λ4069.7 line ratios, the kinetic temperature, ${T}_{\mathrm{kin}}\sim 1.1\times {10}^{4}$K; for higher (lower) temperatures the model ratios are significantly lower (higher) than observed. The temperature sensitivity of these line ratios results from the relatively large changes that occur in the fractional ionization of S at ∼10⁴ K—at such temperatures, most of the S is in atomic form and small changes in T_kin make large changes in the fraction of S⁺, whereas the fractional ionization of Fe is much less affected.

The relative intensity of the [S ii] doublet, λ4077.5/λ4069.7 (hereafter [S ii] doublet ratio) is rather insensitive to temperature. In Figure 11(a), we show a plot of the observed and model line-ratios for a range of temperatures (at a density of $3.2\times {10}^{5}$ cm⁻³); the data are reasonably well fitted for T_kin ∼ 11,000 K. The two [Fe ii] lines at 4245 and 4288 Å vary similarly with temperature over a wide range. Hence, we have used the 0.5 × ([Fe ii]λ4245 + λ4285)/S ii λ4069.7 line ratio (hereafter [Fe ii]/[S ii] line ratio) for comparison in these plots, since the [Fe ii] lines are relatively weak.

Zoom In Zoom Out Reset image size

Both S and Fe may be partially incorporated into dust grains, thus affecting the [Fe ii]/[S ii] line ratio discussed above, but it is likely that much of the dust is destroyed in the violent interaction of the high-speed bullet with a more slowly moving ambient circumstellar medium. But if some fraction of each of these elements does remain locked up in dust grains, the net effect is to decrease the [Fe ii]/[S ii] line ratio, since Fe is less volatile than S. Under these circumstances, the observed [Fe ii]/[S ii] line ratio represents a lower limit for the models (which use the cosmic S and Fe abundances), and ${T}_{\mathrm{kin}}\lt {\rm{11,000}}$ K.

We note that the observed [S ii] doublet ratio appears somewhat lower than the model ones at all temperatures in Figure 11(a). We therefore investigate the dependence of the above line ratios on density for a temperature of 11,000 K, and find that the [S ii] doublet ratio decreases with increasing density (Figure 11(b)). The [Fe ii]/[S ii] ratio goes through a minimum at a density of ∼10⁶ cm⁻³, and then increases again. A reasonable fit to both of the ratios implies a density of ∼5 × 10⁶ cm⁻³. A rough estimate of the mass of bullet 1, M_b, can then be made assuming it to be a sphere with uniform density and physical size corresponding to the deconvolved half-intensity angular extent of the detached blob as measured along the slit, in slit ${S}_{-2b}$ in epoch 1 (about $0\buildrel{\prime\prime}\over{.} 1$ or 40 au)⁸ ; we find ${M}_{b}\sim {10}^{27}$ g. For bullet 2, we estimate a mass of about 0.7 times that of bullet 1, using the fractional [S ii]λ4069.7 flux (0.2) and deconvolved size (1.35) of the detached blob in Period 2 (epoch 4), relative to that in Period 1 (epoch 1), and assuming that the line emissivity varies as density-squared.

Hence the average mass-loss rate via bullet ejection in V Hya is about $4.7\times {10}^{-8}$ ${M}_{\odot }\,{\mathrm{yr}}^{-1}$. The momentum flux (mass-loss rate × outflow-speed) in the bullets is roughly comparable to that in the complex high-velocity molecular outflow seen toward V Hya using 4.6 μm CO spectroscopy (Sahai et al. 2009)—the latter have a mass-loss⁹ rate (∼10⁻⁷ ${M}_{\odot }\,{\mathrm{yr}}^{-1}$) about a factor 2 higher, and an average outflow velocity (∼100 $\mathrm{km}\,{{\rm{s}}}^{-1}$) a factor 2 lower than the bullets. This it seems likely that these high-speed molecular outflows seen in the IR in V Hya are driven by the bullet ejections, via a momentum-conserving hydrodynamic interaction of the latter with V Hya's slowly expanding AGB wind.

The observed relative line intensities for the on-source blob are shown in Figure 12, together with model ratios. Due to the presence of the strong stellar continuum superimposed on the on-source blob emission, the line intensities and ratios are more uncertain than in the case of the detached blob (discussed above). On average, both the [S ii] doublet and [Fe ii]/[S ii] line ratios for the on-source blob in Period 2 appear to be somewhat higher than for the detached blob in Period 1. We are not able to simultaneously fit both these line ratios; the biggest discrepancy is for epoch 5, where the S/N is highest—here the [S ii] doublet ([Fe ii]/[S ii] line) ratio is signficantly lower (higher) than the model value. A plausible physical reason for this discrepancy may be the blob material being far from collisional equilibrium during its early ejection history.

Zoom In Zoom Out Reset image size

V Hya's optical spectrum shows anomalously broad lines with a width that has a mean value of 13.5 $\mathrm{km}\,{{\rm{s}}}^{-1}$ and shows a total variation of 9 $\mathrm{km}\,{{\rm{s}}}^{-1}$, in concert with the photometric phase of the primary's 529 day pulsation period, consistent with the expectation for a rapidly rotating star that conserves angular momentum as it rotates (Barnbaum et al. 1995). These authors propose that the inferred fast rotation of the outer layers of V Hya's stellar atmosphere is due to a binary companion in a common-envelope configuration with the primary (although alternative broadening mechanisms are possible, these authors consider them less likely). In our model, the dominant transfer of angular momentum to V Hya's outer layers would occur during periastron passage.

V Hya also shows a very long period, (6160 ± 400 days) in its light curve (Knapp et al. 1999, hereafter K99) when the star undergoes deep minima. K99 suggest that V Hya is an eclipsing binary with an orbital period of 6160 d, and the periodic deep minima that occur result from enhanced extinction by a dust-cloud associated with the companion. The 6160 ± 400 days period, at the lower end of the range, is 16.8 years, very similar to the flip–flop period, estimated to be twice the bullet ejection period, i.e., ∼17 years. So it is plausible that these periods are really the same, as it would be rather unlikely to have two such similar periods associated with different physical mechanisms. If so, then Period 1 (Period 2) of our STIS observations coincides with phase 0 (0.5), i.e., the maxima (minima) of the long-period light-curve.

We propose that the deep minima in the long-period light-curve are associated with extinction associated with the bullets, and that this happens every alternate 8.5 years orbital cycle as follows. We have shown earlier (Section 4.1.2) that in addition to the flip–flop of the bullet-ejection axis in the sky-plane (causing the ejection direction to move between ESE and ENE), there is also a corresponding flip–flop in a plane perpendicular to the sky-plane (causing the pronounced jumps in radial velocity between successive bullets). We suggest that it is this line-of-sight component of the flip–flop that causes the bullets to be ejected in such a way that as they expand and move across our line-of-sight to V Hya, they alternate between being either in front of the primary star or behind it. In this scenario, the bullet ejected in Period 1 (i.e., bullet 2) lies behind the primary star during epochs 1–2, and only emerges in epoch 3; the bullet-ejection occurs most likely somewhere between epoch 1 and epoch 3.

As noted earlier, V Hya lacks X-ray emission, even though its relatively high FUV/NUV flux ratio (a strong indicator of X-ray emission in fuvAGB stars) indicated an expected X-ray luminosity more than two orders of magnitude higher than the observed $3\sigma$ upper limit (see Figure 4(b), Sahai et al. 2015). Our eccentric-orbit model, in which the accretion rate decreases steeply away from periastron, provides a natural explanation for V Hya's lack of X-ray emission, if we make the reasonable assumption that the latter is related to the accretion rate. The X-ray observations were made on 2013 December 18 and thus at least 2.5 years after periastron passage, when the accretion rate must have dropped sharply from its value at periastron.

The map of CO J = 2–1 emission from V Hya (Figure 3 of Hirano et al. 2004) shows high-speed emission blobs that are modestly offset to the N and S from an overall E–W axis in a staggered pattern that is qualitatively consistent with the flip–flop phenomenon seen in our data. The expansion timescale of these blobs is ∼300 years, derived by dividing their radio distance from the center, $\sim 10^{\prime\prime} \,\tan (90^\circ -i)$ at 400 pc or 7000 au, taking Hirano et al.'s value for the angle between the outflow direction and the line-of-sight ($i=30^\circ$) by their mean radial velocity (110 $\mathrm{km}\,{{\rm{s}}}^{-1}$). Thus, it appears that the flip-flop phenomenon in V Hya, and therefore the precessional-resonance of the accretion disk that is responsible for this phenomenon, has lasted several hundred years.