Discovery of a Little Homunculus within the Homunculus Nebula of η Carinae

We report long-slit spectroscopic mapping of the η Carinae nebula obtained using the Space Telescope Imaging Spectrograph. The observations reveal the presence of a previously unknown bipolar emission nebula (roughly ±2″ along its major axis) embedded within the well-known and larger Homunculus Nebula. A preliminary analysis suggests that this embedded nebula may have originated from a minor eruption event circa 1890, 50 years after the formation of the larger Homunculus.


INTRODUCTION
Observed proper motions confirm that the major eruption of Carinae seen in the 1840s produced the two main lobes of the Homunculus Nebula (Gaviola 1950;Ringuelet 1958;Morse et al. 1998;Currie & Dowling 1999). Some features outside the Homunculus, however, were very likely ejected earlier (Walborn, Blanco, & Thackeray 1978). Apparently, Car may have undergone a number of episodic outbursts before the major eruption (see Davidson & Humphreys 1997 and references therein).
At least two questions arise: Do all of the star's outbursts eventually produce recognizable nebular features? Do all such features tend to have a similar bipolar (quasi-axial) geometry?
The historical light curve indicates a secondary eruption during the 1890s (Humphreys, Davidson, & Smith 1999). Some detectable equatorial material probably originated in that later event Davidson et al. 2001), but cannot easily be identified in continuum Hubble Space Telescope (HST) WFPC2 images (Morse et al. 2001). Here we report the detection and preliminary analysis of a relatively small bipolar structure inside the central regions of the familiar Homunculus lobes, and most likely ejected in the 1890 event. This mini-homunculus became evident in spectroscopic data obtained with the Space Telescope Imaging Spectrograph (STIS).

DISCOVERY OF THE NEW BIPOLAR NEBULA
We first noticed a previously unknown velocity structure near Car in HST STIS spectroscopy obtained on 1998 March 19. Initially the geometric details were not clear, and we referred to it as the '' integral-sign nebula ( H ) '' . Figure 1 illustrates the 52 00 Â 0>1 aperture orientation for those observations, position angle 152=1 332=1. In programs to monitor the star's 5.5 yr spectroscopic cycle and to explore the Homunculus, we obtained similar data on later occasions, listed in Table 1. Because of HST scheduling constraints, the slit orientation differed from the bipolar axis direction by about 20 . Figure 2 shows the appearance of emission lines within 2 00 of the star in the 1998 data; the right-hand part is a median-filter superposition of more than 30 lines, mostly [Fe ii], in the 4000-5000 Å spectral interval.
The velocity structure has two obvious components extending from À2 00 to +2 00 across the star: a narrow straight part at V % À32 km s À1 (heliocentric), and a narrow, curved component with peak velocities V $ AE200 km s À1 . Hereafter we call these the bar and curve components, respectively. The latter appears to be approximately symmetric around the stellar position, suggestive of a bipolar geometry. The bar component, on the other hand, may represent expanding gas located outside the Homunculus lobes (see x 4.3). Figure 3 shows [Ni ii] 7380 (in vacuum) velocity structures along the slit, observed 2000 March 13 with a slit orientation closer to the major axis. Some of these were sketched earlier in Figure 2d of Davidson et al. (2001). 17 Components 1 and 2 represent intrinsic emission in the southeast and northwest Homunculus lobes, while 3 is light reflected by dust grains in the polar region of the southeast lobe. Segments 4 and 5 are equatorial emission structures with ages of roughly 160 and 100 yr, respectively (see also Zethson et al. 1999). In this paper, however, we are concerned with the additional components 6 and 7, the curve component of the central structure, shown with a very different aspect ratio in Figure 2. The bar component is not conspicuous in Figure 3. -Orientation, size, and mapping coverage of the STIS slit used for these observations. The HST WFPC2 image shown here was obtained in 1995, but its scale has been adjusted slightly to allow for expansion to 1998.3 epoch. Its brightness has also been adjusted to bring out the nebular ejecta.  Gull et al. (1999) and  initially speculated that the small internal nebula was a narrow, collimated bipolar flow. This hypothesis was based on the early 1998 data, with one particular slit position angle. In order to check or improve the original model, it was imperative to obtain a spectral map of the region with additional slit positions. The STIS Instrument Development Team allocated four HST orbits to this mapping activity as a key GTO project. The mapping visit was executed in late March 2000 (visit 7 in Table 1). Results are described in the following section.

MAPPING OBSERVATIONS
Spectroscopic mapping near Car was achieved on 2000 March 21 with the HST STIS CCD, using STIS gratings G430M and G750M at standard tilts that sampled wavelength ranges 4820-5100 Å (blue) and 6490-7050 Å (red) (Kimble et al. 1998;Woodgate et al. 1998). The 52 00 Â 0>1 slit, oriented at position angle 145=1 $ 325=1, was placed at various locations, and exposure times were typically 80 s each. Figure 1 shows one of the slit positions and also the total spatial coverage at each central wavelength. The blue mapping extended from 2>9 NE to 1>9 SW (measured perpendicular to the slit) at 0>1 intervals. The red mapping spanned a width of 12 00 at 0>05-0>25 intervals. Several positions close to Car were resampled at shorter exposures because of severe saturation of H and H. Spectral resolutions were approximately 32 and 47 km s À1 for the red and blue mappings, respectively.
Our scientific goals required high-precision astrometry and spectroscopy with the data set. As we were pushing the diffraction limit of the telescope in our extractions, routine pipeline calibration was inadequate. We performed our own recalibration with special tools developed by one of us (K. I.) and by the STIS IDT group.
The key steps were (1) spatial alignment of all spectral images with an accuracy of about 0>013, a quarter of the raw CCD pixel size, including corrections for thermal effects in the STIS instrument (see Gull et al. 1997); (2) improved spectral image rectification in order to correct for a tilt in each spectral image; and (3) correction for wavelength offsets to an accuracy of $AE7 km s À1 . These correction steps were applied to derive input parameters used in the STIS IDT data reduction program CALSTIS (version 6.6; see Lindler 1999 18 ).
We did not use hydrogen and helium lines in the analysis of the velocity structure, since their narrow intrinsic components are blended with broader scattered components. We therefore focused our attention on the [Fe ii] lines, which are ubiquitous and narrow in the spectrum of the Homunculus. We selected three strong, unblended [Fe ii] lines at 4891, 4907, and 4975 Å (vacuum) for the following analysis in order to avoid possible spatial variations in different atomic species. At each slit position, the spectral images of these lines were then combined via a median-filter superposition to enhance faint velocity structures, as shown in the right panel of Figure 2. Furthermore, to emphasize the changes in intrinsic emission features, each composite spectral image was processed to suppress the broad stellar continuum scattered by dust in the Homunculus Nebula. Then apparent Doppler velocities were measured from line profiles extracted at numerous positions along the STIS slit (with a sampling width of five CCD rows, or 0>25). Some spatial sampling may overlap slightly. Figure 4 shows an example of our velocity measurements taken at an offset 0>4 in the southwest direction. Velocity widths may give some indication of the geometrical thickness of a nebular wall. In this image, a typical velocity width in the curve component is %91 km s À1 , although the value varies by 20 km s À1 both along the slit and at each pointing. The intrinsic velocity width is roughly 85 km s À1 after correction for instrumental resolution. These scans were done for the composite frames taken at 1>5 away from the star in both the southwest and northeast directions. Beyond that distance, the emission structure became too faint to be detected with confidence.
We then combined the spectral mapping images to synthesize a velocity data cube. Since spatial coverage was not complete for the red (6490-7050 Å ) data, a simple bilinear interpolation scheme filled the gaps (for more details see . Each data cube allows us to visualize an extremely '' narrowband '' image, tunable in velocity to its pixel resolution, DV % 10-20 km s À1 . Our spectral resolution with the 52 00 Â 0>1 slit was roughly 32 and 47 km s À1 for the blue and red data, respectively, with corresponding spatial samplings roughly 0>05 Â 0>1 (blue) and 0>25 Â 0>1 (red). Velocity cubes are plotted in Figures 5 and 6. The [Fe ii] 4891 maps shown in Figure 5 demonstrate that a narrowband image with subarcsecond spatial resolution can be synthesized from data obtained with a narrow-slit spectrograph on the HST. Each data set allows us to investigate spatial structures of emission in a single atomic line. Each of the [Fe ii] lines mentioned above individually gives practically identical results. The removal of scattered light originating from the extended nebula also simplifies the examination of faint nebular condensations. Figure 6 shows 40 panels of narrowband maps covering the range from À955 to +1022 km s À1 for the H emission line, just to demonstrate the amount of kinetic information that can be gathered by the mapping observation. In this velocity range, the neighboring [N ii] emission lines contribute significantly to the intensity near À670 and +940 km s À1 . These narrowband maps may be useful for future studies of the scattered emission structure of the Homunculus and intrinsic emission from shells and strings in the outer halo. To our knowledge, no previous instrument has ever achieved these levels of angular and spectral resolution simultaneously (see Davidson et al. 1997 for a comparable instance using an earlier HST instrument). (3) scattered emission from the southeast lobe (its counterpart for the northwest lobe is not seen in this element); (4) and (5) equatorial ejecta; and (6) and (7) the caps of the internal bipolar nebula.

RECONSTRUCTION OF THE INTERNAL NEBULAR STRUCTURE
Our goal here is to derive the geometrical structure of the inner bipolar emission nebula: the curve component in Figure 2, or features 6 and 7 in Figure 3. Results can then be compared to the larger bipolar shapes and equatorial structures assessed by Davidson et al. (2001). As usual for studies of the Homunculus, we assume (1) that all these ejecta originated from the central star, and (2) that they are moving ballistically without significant deceleration. We measured apparent heliocentric Doppler velocities and then, consistent with Davidson et al., we added +10 km s À1 as a crude but adequate correction for the systemic velocity of Car. From the resulting velocity V z parallel to our line of sight we calculate the '' distance from the plane of the sky '' z ¼ V z t, i.e., a relative linear coordinate along the line of sight, where t is the age of the ejecta. approximate axial symmetry makes the outer polar caps (segments 1 and 2 in Fig. 3) parallel to the main equatorial debris (segment 4). The inner structures 6 and 7, however, have a very different slope in Figure 3, more or less parallel to segment 5, which probably represents younger equatorial ejecta from the 1890 outburst or later. Thus, we have reason to suspect that the inner mini-Homunculus was ejected later than the well-known outer lobes.

Proper Motion and Age Determination
Proper motions give a more specific estimate of t. For this purpose we compared the 1998 March 19 and 2000 March 20 data sets (Table 1). Careful cross-correlations along the slit direction showed a small but measurable spatial drift of the northwest part of the curve component in Figure 2. In order to minimize contamination by the large-scale Homunculus and the bar component, we used small 0>51 and 0>63 square sample sizes for the correlation. The procedure addressed, separately, 28 [Fe ii] lines and the bright [Ni ii] 7379.86 (vacuum) feature. The southeast segment of the curve component could not be used, as the selected lines were generally too weak there. For each emission line, the expansion age was determined. This analysis showed ejection years 1904 AE 10 and 1897 AE 10 using the 0>51 and 0>63 sampling boxes, respectively (see Fig. 7). Here the quoted 1 errors are dominated by a general AE10 km s À1 velocity uncertainty and also include the formal dispersion of the individual line measurements. It is not surprising that the larger sampling size gave a larger age, since it presumably has more contamination by ejecta from the major circa 1843 event. These results are quite independent of the deduction expressed in the preceding paragraph, and are consistent with it. Therefore (rounding somewhat for simplicity), we adopt t % 100 yr for the inner structure.     Figure 8 (from ) are based on an age of 157 yr; however, for the inner structure marked with dots, we assume t ¼ 100 yr, as explained above. Like the familiar Homunculus, the embedded bipolar nebula appears to be '' approximately '' hollow, i.e., some lower density material exists inside it but the shell surface is well defined Smith, Gehrz, & Krautter 1998;Smith et al. 1995). Its polar caps are expanding outward at about 300 km s À1 , about half as fast as the large Homunculus lobes. We refer to this quasi-hollow inner structure as the Little Homunculus. 19 Figure 9 illustrates a preliminary result of three-dimensional reconstruction of [Fe ii] 4891 (vacuum) emission structure in the Little Homunculus nebula, based on the mapping data set. A bipolar-like nozzle segment of the internal nebula is clearly present in this viewing angle, although the presented three-dimensional data cube suffers from the low signal-to-noise ratio in these data.

The Bar Component
The physical nature of the bar component in Figure 2 is not clearly understood. We speculate that it is spatially detached from both the large Homunculus and the little one. A similar radiative pumping mechanism probably excites the spectra of the Weigelt blobs, the curve component, and the bar component, since all three structures share the same, somewhat unusual set of strongest emission lines. Furthermore, from its constant, extremely narrow velocity structure across the slit position, our first impression of its geometry is a thin shell embedded within an outer diffuse gas in the equatorial region of the system (equatorial halo may be a valid description). If the bar component were physically located within the Homunculus or coincided with the Little Homunculus, then changes in velocity would have been expected across the position of the central star. However, such changes were not detected ). Thus, it seems more appropriate to consider the presence of outer or extended diffuse gas that is radiatively excited by scattered UV photons in the equatorial region. We attempt to illustrate both the curve and bar components in a cartoon model of the Car nebula (see Fig. 10). Figure 11 shows a spectrum of the Little Homunculus sampled 1>53 from the central star at P.A. 332=1, with a 0>1 Â 0>13 rectangular aperture. It exhibits a forest of narrow emission lines like other nebular condensations around the star. In general, the excitation mechanisms of these lines are not well understood; high levels of Fe ii are populated by radiative pumping, while the lower levels may be collisionally excited at temperatures of a few thousand K . Fortunately, however, the gas densities are sufficiently high that many lower, metastable level populations may be close to a quasi-Boltzmann distribution. Therefore, we employ a simple technique first applied to Car by Pagel (1969) and more recently used by Davidson et al. (1995) and Hamann et al. (1999). We consider only forbidden [Fe ii] lines that have fairly low critical densities ($10 6 cm À3 ; see Hamann et al. 1999), and we suppose that their upper levels are populated according to an idealized Boltzmann distribution at some fixed temperature; then the observed relative line intensities depend on temperature and on reddening by dust. The validity of these assumptions can be judged by the mutual consistency of many different line intensities.

PHYSICAL CONDITIONS IN THE LITTLE HOMUNCULUS
Step 1.-We selected 38 unblended [Fe ii] lines in the wavelength range 4100-9000 Å and measured their apparent intensities. These measurements and the atomic parameters are tabulated in Table 2. Decay rates for [Fe ii] lines are quoted from the relativistic Hartree-Fock calculation by Quinet, Le Dourneuf, & Zeippen (1996). We exclude 4 of the 38 measured lines from the following analysis because  their decay rates are doubtful (marked as 1 in Flag column in Table 2; see '' Note added to the proof '' in Quinet et al.). Hence, we use 34 [Fe ii] measurements in the following steps.
Step 2.-We have where I obs is an observed line intensity, C 1 is a constant, ij and A ji are the photon energy and decay rate for the transition from levels j to i, j and g j are the energy and statistical weight of the upper level j, and is a wavelength-dependent effective optical depth due to absorption and scattering by dust grains. Since extinction and reddening near Car are too abnormal to represent with a standard curve (see below), we simply assume that is linear in 1= or in photon energy ; this approximation is adequate for our purposes with most likely reddening curves. In order to give the correct reddening parameter E BÀV expressed in astronomical magnitudes, the linear relation must be ðÞ % C 2 þ 0:58 eV E BÀV : Thus, we can rewrite the earlier equation in a practical form: which can be further simplified as since the ratio ji = j in Table 2 is always 1 or slightly less than The angular dimension of the box is roughly 7 00 Â 10 00 Â 5 00 (width Â height Â depth). The arrow indicates the direction toward the observer. Bright features are shown in white. A bipolar nozzle-like structure corresponds to the Little Homunculus. Because of the poor signal-to-noise ratio in the data set, only a part (primarily the nearside only) of the inner lobe structure is seen. Volume rendering is performed with the AnimaBob graphic package.   Evidently this gives a relation between consistent values of gas temperature T and reddening E BÀV . Figure 13 shows the relation between E BÀV and T based on this analysis. The extinction-to-reddening ratio A V =E BÀV is not known, but available studies find values around 4 or 5 for interstellar extinction of stars near Car, and much greater for dust within the Homunculus itself (see many references cited in Davidson & Humphreys 1997). Interstellar reddening to the Carina Nebula is about 0.45 mag, and the corresponding extinction is roughly 1.7 mag (Walborn 1995). Earlier studies suggest that the reddening of the Weigelt blobs close to the star may be as small as 0.55 mag (Davidson et al. 1995). Therefore, it seems likely that the reddening for the Little Homunculus is in the neighborhood of E BÀV % 0:5 mag. Two vertical lines in Figure 13 indicate values of 0.45 and 0.55. This result suggests that the gas temperature of the Little Homunculus Nebula is likely to be found near 9000 K.
An order-of-magnitude particle density estimate is possible. At T $ 9000 K, we assume an ionization fraction of Fe of approximately 30% and a reddening of E BÀV ¼ 0:50. The partition function varies with particle density (Traving, Baschek, & Holweger 1966): 68,000, 12,000, and 2200 for densities of 10 5 , 10 6 , and 10 7 cm À3 , respectively. In each case the observed brightness of the emission lines requires roughly 2 Â 10 50 , 3 Â 10 49 , and 6 Â 10 48 Fe atoms in the volume of 0>1 Â 0>13 Â 0>5 ($3 Â 10 47 cm 3 ), respectively. Comparing the derived densities with the assumed ones, the acceptable range of density n H is roughly 10 6 < n H < 10 7 cm 3 if the Fe/H abundance is similar to the solar abundance. Suppose that the Little Homunculus consists of a bipolar cap within a radius R ¼ 1>5 and thickness ¼ 0>5 (imagine a '' disklike '' cap). Further assume its filling factor to be unity. Then the total mass is estimated as 0.1 M for n H ¼ 10 6 cm À3 and 1 M for n H ¼ 10 7 cm À3 . At the distance of $2 00 ($4500 AU) from the star, the radiation-density temperature of the nebula is roughly 280 K, assuming that the effective temperature of Car's atmosphere at 60 R is $35,000 K ). This temperature is slightly warmer than that of the Homunculus ($200 K). If the Little Homunculus contains a mass of 1 M of gas, it may contain roughly 0.01 M of dust. Such a mass of dust in the nebula would give rise to a bright thermal IR source with a color temperature of 280 K, and therefore the nebula should be detected clearly in earlier IR images. However, this is not known to be the case . So intuitively, the total mass of the Little Homunculus (at least the cap part) should be much less than 1 M , implying that either the density n H is lower than 10 7 cm À3 or its filling factor is much less than unity. Other factors (a supersolar Fe/H ratio, etc.) may effectively lower the total mass of the Little Homunculus as well.

Temporal Variability of the Little Homunculus
Using three data sets obtained at the same position angle but at different epochs, we also examine temporal variability of the Little Homunculus. Spectral samples were taken for 18 emission lines of [Fe ii], Fe ii, and [Ni ii] and integrated line fluxes were measured for each epoch at the star, Weigelt blobs B and D, and the cap of the Little Homunculus for comparison. None of the selected lines were high-excitation species (e.g., [Fe iii]) that require a very intense, ionizing UV radiation field. High-excitation lines are also known to be susceptible to Car's 5.5 yr periodic cycle. The flux ratios measured for these low-excitation lines are plotted in Figure 14.
Two physically independent variations were present during the period monitored with STIS: the changes due to the 5.5 yr periodic variability (Damineli 1996) and the recently discovered brightening of the stellar source at visual wavelengths . The effect of the recent brightening is illustrated well in the top panel of Figure 14, showing that the stellar flux has increased by an appreciable factor. However, both Weigelt blobs and the Little Homunculus remained fairly constant in their line emission strengths during the event and afterward.
On the other hand, high-excitation line species in the nebula exhibit strong variability throughout our monitoring. High-excitation lines at the cap during the last event were all weaker, following a trend similar to that seen in the Weigelt blobs. Figure 15 illustrates the change in Fig. 12.-Excitation energy j vs. lnð ij A ji g j Þ À lnðI obs Þ. The dotted line indicates the best linear fit through good data points ( plus signs). Four data points, not included in the best fit, are marked with diamonds. If these four data points are included, then the slope of the fit through the lowest levels (dashed line) is much higher, 8:67 AE 1:73 eV À1 . This value is so large that it requires very high values of E BÀV , which are not plausible. Fig. 13.-Reddening E BÀV vs. gas temperature T. The solid curve is derived based on the slope of the best fit given in Fig. 12. The other two dotdashed curves define the upper and lower boundary conditions given by the uncertainty of the fit. The vertical dotted lines mark the range of E BÀV that has been generally accepted in previous studies. [Fe iii] and H lines. 20 The presence of H in emission from the Little Homunculus indicates that hydrogen atoms are present in the ejecta. Presumably, UV photons may drive the enhanced strong high-excitation lines via UV pumping and two-photon excitation mechanism in the Weigelt blobs and other parts of the nebula, according to recent studies on atomic physics (Johansson & Hamann 1993). Assuming that this is the driving mechanism for high-excitation emission lines, our data suggest that changes in the UV ionizing radiation field may occur in both polar and equatorial directions. Naively, it may be an indication of the isotropic change in the ionizing UV radiation density field, although it certainly does not preclude any other possibilities (cf. Smith 2002). and H emission structures are marked with two vertical ticks on the spectral images. The spectral extraction width used to obtain the spectra are also marked. These high-excitation species require the presence of strong ionizing UV radiation in the field, and therefore are very sensitive to the periodic variation associated with the 5.5 yr event of Car. Note that the bar component in [Fe ii] appears to shift redward slightly as it extends outward, whereas the neighboring [Fe ii] lines appear to show a near-constant velocity structure.
In turn, this implies that the 5.5 yr periodic event of Car is driven by gross attenuation of the UV light in the same isotropic manner.

FINAL REMARKS
We have identified a Little Homunculus embedded within the Homunculus Nebula around Car. The existence of this nebula perhaps could have been detected with the narrowband HST WFPC2 imaging of Car, as the blue glow shown in Figure 4 of Morse et al. (1998) suggested the presence of the peculiar hazy structure. However, twodimensional imaging could not have resolved its complex three-dimensional structure, nor could it easily bring out a fainter emission structure over the bright scattered continuum emission from the Homunculus Nebula.
As explained in x 4 above, the age of the Little Homunculus appears to be about 100 yr. This is not a trivial finding.  first argued that the equatorial ejecta came from the 1890 eruption, but Morse et al. (2001) were unable to detect any ejecta coming from the 1890 eruption event based on the HST WFPC2 data sets. STIS spectra studied by Davidson et al. (2001) showed that the equatorial region may contain material from both the 1890 eruption and the Great Eruption in the 1840s. Our paper does not settle the debate on the age of the equatorial ejecta, but we deduce that some additional ejecta inside the Homunculus came from the minor eruption 50 yr after the Great Eruption.
The other controversy in Car to date is whether it has a binary companion. This study does not settle the issue, as the large-scale nebula would see more or less a pointlike ionizing UV source at the heart of the stellar system regardless of whether there is a single star or a close binary system. However, from temporal variations seen in the Weigelt blobs (conceivably on the equatorial plane; see Davidson et al. 1995Davidson et al. , 1997Falcke et al. 1996) and the Little Homunculus, the changes in ionizing UV radiation flux may possibly occur in all directions (one of us [N. S.] has recently found strong evidence that changes in the UV flux are not isotropic; see . In the single-star case, this is what one would naturally expect. In the binary case, however, a simple hypothesis of '' line-of-sight '' eclipsing does not predict isotropic changes (cf. Ishibashi et al. 1999). However, a companion star may be submerged within the changing wind of Car. This type of binary model has been discussed and is plausible Damineli, Lopes, & Conti 1999 and references therein).
Whatever the true nature of Car may be, we are still in the process of determining what we understand and what we do not. We are taking the next step by sorting out emission from several spatial structures and by examining the radiation mechanisms of the nebular condensations surrounding this very massive star.