AN EXTREME HIGH-VELOCITY BIPOLAR OUTFLOW IN THE PRE-PLANETARY NEBULA IRAS 08005-2356

I 08005 shows an hourglass morphology in the HST F439W, F555W, and F675W images (Figure 1)—the lobes, separated by a narrow dark lane, flare out from the central star's location and attain a cylindrical shape. The SE lobe is significantly brighter than the NW one. A narrow, slightly curved feature of linear extent 038 is seen extending from the central star in the SE lobe (see the inset). The feature is oriented at a PA = 120°, different from the PA of the nebula symmetry axis, PA = 143°. The total linear extent of the nebula along its long axis, as seen in the F555W image, is ∼26 (1.2 × 10¹⁷ cm).

Zoom In Zoom Out Reset image size

We detected the CO J = 2–1, 3–2 and SiO J = 5–4 lines with the SMA. The CO J = 3–2 and SiO J = 5–4 have significantly lower signal-to-noise, and the ¹³CO J = 2–1 is tentatively detected. The (spatially) integrated SMA CO J = 2–1 line profile (using a 15'' diameter circular aperture) shows a very broad profile (Figure 2) covering a total velocity extent of about 350 km s⁻¹, i.e., from −150 ≲ V_lsr (km s⁻¹) ≲ 200. The SMT CO J = 2–1 line profile (Figure 2) is similar to the SMA one (a narrow emission feature at V_lsr = 36 km s⁻¹due to the presence of an unrelated line-of-sight interstellar cloud has been removed from the profile by interpolation). A Gaussian fit to the profile gives a central velocity V_lsr ∼ 25 km s⁻¹ and an FWHM of 196 km s⁻¹.

Zoom In Zoom Out Reset image size

We have divided the velocity range spanned by the CO J = 2–1 line into three intervals on each side of the central velocity—the extreme high-velocity (EV) component (blue: −150 < V_lsr (km s⁻¹) < −85, red: 130 < V_lsr (km s⁻¹) < 200), a medium-velocity (MV) component (blue: −85 < V_lsr (km s⁻¹) < −30, red: 70 < V_lsr (km s⁻¹) < 130), and a low-velocity (LV) component (blue: −30 < V_lsr (km s⁻¹) < 25, red: 25 < V_lsr (km s⁻¹) < 75). The SMA CO J = 2–1 emission in the blue- and redshifted parts of the EV, MV, and LV components show separations of about 21, 11, and 05, respectively, along the nebular symmetry axis (Figure 3(a)).

Zoom In Zoom Out Reset image size

These results imply the presence of a fast, bipolar outflow with a (roughly) linear velocity-gradient, directed along the nebular symmetry axis, and a linear extent of ∼21. Since AGB mass loss is typically spherical, with outflow velocities of 10–15 km s⁻¹, we conclude that the outflow observed in CO J = 2–1 emission is a fast, collimated post-AGB outflow, typical of the PPN evolutionary phase. We made similar plots from the SMA data cubes for CO J = 3–2 (Figure 3(b)) and SiO J = 5–4 (not shown), dividing the emission velocity-range into two halves only (blue: −90 < V_lsr (km s⁻¹) < 20, red: 30 < V_lsr (km s⁻¹) < 130) due to the lower S/N in these lines and found that these show a separation of 085 roughly along the nebular axis. This separation is equal to that derived from our CO J = 2–1 data for the same red and blue outflow-velocity ranges, implying that the SiO outflow's linear extent is likely comparable to that of CO.

We also detected the CO J = 4–3, J = 3–2, and ¹³CO J = 2–1 as well as the SiO (v = 0) J = 5–4 and 6–5 lines with the SMT, but with lower S/N (Figure 2). We have compared the total CO J = 2–1 line fluxes from our SMA and SMT data, as these have the highest S/N, and find that these are consistent within 15%, i.e., within typical calibration uncertainties—we conclude that the SMA data do not suffer from any significant flux losses.

Continuum images were obtained after removing spectral regions containing line emission using the Miriad task uvlin. No continuum emission was detected from I 08005. Using line-free channels, we find a 1σ noise of 1.05 (1.2) mJy beam⁻¹ in the USB (LSB) at 1.3 mm and 11 (18) mJy beam⁻¹ in the USB (LSB) at 0.87 mm.

We now determine the physical properties of the high-velocity bipolar outflow (e.g., scalar momentum, mass, and age) from an analysis of the molecular-line data. The observed ratios of the CO J = 2–1, 3–2, and J = 4–3 fluxes (in K km s⁻¹) are 1:1.27:1.2, which imply corresponding non-beam diluted flux ratios of 1:R_32/21:R_43/21 = 1:0.6:0.3 since the source is unresolved by the SMT beams of 32'', 22'', and 16'' in these lines.

We first make a simple emission model assuming a spherical outflow at a uniform excitation temperature equal to the kinetic temperature, T_kin, and carry out a least squares fit to the line fluxes,⁶ varying T_kin, the outer radius (R_out), and the mass of the emitting region (M_g). We find R_out = 14 and T_kin = 13.6 K. However, the derived T_kin is rather low compared to energy of the CO J = 4 level (55.4 K), and modeling with the non-LTE RADEX code (van der Tak et al. 2007) shows that the J = 4–3 excitation temperature, T_ex(43), is lower than T_kin by about (20–25)% due to the relatively low CO line optical depths in our model (<0.01) and the average density, n_av ∼ 3.5 × 10⁴ cm⁻³, implied by the emitting region's mass and volume. Thus, the resulting, corrected model R_43/21 ∼ 0.11 is too low to fit the data.

Using RADEX, we find that the average density required to bring T_ex(43) closer to T_kin in order to fit the observed source brightness temperature ratios is n_av > 7.5 × 10⁴ cm⁻³; the required T_kin = 15.5 K, giving a source size of 24, comparable to the CO source size estimated from the SMA data. The model source size is similar to the size derived from the HST image, implying that the CO emission likely comes from the dense walls of the bipolar lobes seen in scattered light. The CO column density is N_CO > 5 × 10¹⁷ cm⁻² (and the J = 2–1, 3–2, and 4–3 lines have optical depths of 0.48, 0.44, and 0.20). Assuming that the average emitting column is equal to the source radius, we find a CO-to-H₂ abundance ratio of, f_CO = 1.3 × 10⁻⁴, in reasonable agreement with the value typically assumed for PPNe, 2 × 10⁻⁴ (e.g., Bujarrabal et al. 2001). Assuming a spherical emitting volume, the mass is M_g = 0.076 M_⊙. The ¹²CO/¹³CO abundance ratio is f(¹²C/¹³C) = 9.6 from fitting the ¹²CO/¹³CO J = 2–1 line flux ratio corrected for the different beam dilutions (8.1).

The above values of M_g and f(¹²C/¹³C) are lower limits since this is our "minimum-mass" model—models with higher values of n_av are allowed. However, the total mass increases more slowly than the average density since in models with higher values of n_av, the emitting region's size is smaller. For example, if we use the dust mass M_d = 0.0019 L_⊙ (D/2.85 kpc)² derived by OBS05 from a detailed 2D radiative transfer model of I 08005's SED, scale it to D = 3 kpc, and adopt a typical gas-to-dust ratio for oxygen-rich AGB stars M_g/M_d = 200, we get M_g = 0.42 M_⊙. For this value of ${M}_{g}$, we need a CO model with ${n}_{\mathrm{av}}\sim 8\times {10}^{5}$ cm⁻³, a factor of ∼10 higher than in our minimum-mass model; T_kin = 11.5 K, and N_CO = 1.6 × 10¹⁸ cm⁻². Since the CO J = 2–1 optical depth is higher (τ₂₁ = 1.95), the abundance ratio is higher, f(¹²C/¹³C) = 17.

We calculate the scalar momentum using the formulation described in Bujarrabal et al. (2001). Using our minimum-mass model, we find P_sc ∼ 2.8 × 10³⁹ g cm s⁻¹ for an inclination angle of the nebular axis to the sky plane, i = 30° (OBS05). The kinetic energy in the outflow is ${E}_{\mathrm{kin}}\sim 2.6\times {10}^{45}$ erg. These values of P_sc and E_kin lie near the upper end of the range for PPNe, 10³⁷⁻⁴⁰ g cm s⁻¹ and 10⁴²⁻⁴⁶ erg (Bujarrabal et al. 2001). This outflow cannot be driven by radiation pressure because the lobes' dynamical (expansion) timescale t_dyn = 190 years (from dividing the model CO shell size by the CO J = 2−1 FWHM line width) is much smaller than that required by radiation pressure to accelerate the observed bipolar outflow to its current speed, ${t}_{\mathrm{rad}}={P}_{\mathrm{sc}}/(L/c)\sim 6.6\times {10}^{4}$ years, given I 08005's luminosity of 6980 L_⊙ at D = 3 kpc (using the value derived by OBS05, $6300{(D/2.85\;\mathrm{kpc})}^{2}$ L_⊙.) The mass-loss rate in the outflow is >5.8 × 10⁻⁴ M_⊙ yr⁻¹.

The observed SMT SiO J = 6–5 to 5–4 line flux ratio is 0.94 (with about ±15% uncertainty), implying an intrinsic flux ratio (i.e., corrected for beam dilution) of R(SiO)_65/54 = 0.66 since the source is unresolved by the SMT beams of 34'' and 284 in these lines. Using RADEX, we find that for T_kin = 15.5 K, the average H₂ density in the SiO emission region is n_av > 2 × 10⁵ cm⁻³ in order to fit the observed value of ${\rm{R}}{(\mathrm{SiO})}_{65/54}$. The fractional SiO abundance is $\sim 4\times {10}^{-5}$, close to the maximum value assuming cosmic abundances ($6\times {10}^{-5}$), and comparable to the observed maximum circumstellar SiO abundance in oxygen-rich AGB stars (${few}\times {10}^{-5}$; Sahai & Bieging 1993). If the SiO emission comes from shocked gas at a higher kinetic temperature than that for CO, the minimum density required is higher.

Evidence of very fast outflows in I 08005 comes from three independent probes: (i) CO data (presented here), (ii) Hα spectrum in SCetal08, and (iii) OH maser emission (VLA mapping) reported in Zijlstra et al. (2001). The OH maser features cover the velocity range of $0\lt {V}_{\mathrm{lsr}}$ (km s⁻¹) < 100, i.e., roughly the range covered by the red half of the CO J = 2–1 profile. In contrast to OH and CO emission, the Hα absorption probes atomic gas. Since the Hα absorption feature is not spatially resolved in the ground-based long-slit spectra, it is difficult to directly establish its relationship to the fast molecular outflow; however, the close agreement between the terminal outflow velocities derived for these indicates that they are closely associated. The atomic gas may be located inside the lobes and may constitute unshocked material of an underlying jet and/or the interface between the latter and the lobe walls, as proposed for IRAS 22036+5306 (Sahai et al. 2006). The jet-like feature seen in the HST image (Figure 1, inset) may represent the precessing jet's signature close to its launch site.

The high-speed bipolar outflow in I 08005 may be driven by an accretion disk. Bakker et al. (1997) find numerous narrow, double-peaked, chromospheric emission lines from neutral and singly ionized metals and propose that these might arise in an accretion disk. We conjecture that the broad Hα wings seen toward this object might arise as a result of Raman-scattering of Lyβ emission generated by such a disk—a process that produces an Hα profile with a width that is a factor 6.4 larger than the Lyβ width and a ${\lambda }^{-2}$ wing profile. The very wide ($\mathrm{FWZI}\sim 2400$ km s⁻¹) Hα line wings in I 08005 show this shape; furthermore, a weak emission feature around 6830 Å with $\mathrm{FWZI}\sim 150$ km s⁻¹, corresponding to Raman-scattering of the 1032 Å component of the O vi doublet at $\lambda \lambda 1032,1038$ is also seen (SCetal08). The 7088 Å feature that corresponds to the 1038 Å component is a factor of four weaker and too weak to be visible.

Other line-broadening mechanisms include electron scattering and emission from a rotating disk. Arrieta & Torres-Peimbert (2003) find that electron scattering requires extreme densities (e.g., ${n}_{e}\gt {10}^{12}$ cm⁻³ in M 2−9), making that an implausible mechanism. Keplerian rotation in a disk around the central star is too low to directly account for the broad line-width in I 08005 (and PPNe in general; Sahai et al. 2011b)—taking the radius for the central post-AGB star of I 08005 to be $R\sim 50$R_⊙ (Slijkhuis et al. 1991) and a nominal stellar mass of 1M_⊙, we find the maximum rotation speed is ${V}_{\mathrm{mx}}\lt 56.5$ km s⁻¹, too low for generating the extreme line wings of the Hα profile, even with the factor of 6.4 increase provided by Raman-scattering. We therefore conclude that the accretion disk in I 08005 must be around a much smaller star, e.g., a main-sequence companion.

The Raman-scattering likely occurs in the innermost regions of the fast neutral wind seen via its blueshifted absorption feature signature in the Hα profile. In this case, following Sahai et al. (2011b), we scale from the minimum scattering column density, ${N}_{s}={10}^{19-20}$ cm⁻², needed to achieve the Raman conversion efficiency necessary for producing the very broad observed line-widths (see Figure 1 of Lee & Hyung 2000) to find that ${{dM}}_{s}/{dt}\gt (0.35-3.5)\times {10}^{-8}$M_⊙ yr⁻¹(${r}_{w}/30$ AU) (D/3 kpc) (V_exp/100 km s⁻¹) (${N}_{s}/{10}^{19}$ cm⁻²), where r_w is the radius of the neutral wind where the Raman-scattering occurs. V_exp is set equal to $0.5\times \mathrm{FWHM}$ of the broad CO J = 2–1 line profile and similar to the neutral outflow's average outflow-velocity derived by SCetal08. The mass-loss rate requirement is easily met in I 08005 since OBS05 derive a mass-loss rate of $4.4\times {10}^{-6}$M_⊙ yr⁻¹ for a collimated fast wind in the 24−780 AU region around the central star,