Free Access
Issue
A&A
Volume 630, October 2019
Article Number L6
Number of page(s) 9
Section Letters to the Editor
DOI https://doi.org/10.1051/0004-6361/201936277
Published online 30 September 2019

© ESO 2019

1. Introduction

η Car is the archetype of unstable massive stars, eruptive mass loss, and supernova (SN) impostors, and it is a reference for understanding the precursor eruptions that lead to superluminous SNe. The star has a current mass of about 100 M and is in an eccentric binary system (Damineli et al. 1997; Davidson & Humphreys 2012 and references therein). During an eruptive event in the nineteenth century (see Frew 2004 for the historical lightcurve), the system ejected ≥45 M of material (Morris et al. 2017; see also Smith et al. 2003) that forms the bipolar Homunculus nebula.

η Car is often as an extreme case of a luminous blue variable (LBV). LBVs are evolved massive stars with initial masses ≥20 M that exhibit instabilities that are not understood, resulting in enhanced mass loss (Conti 1984, 1997; Humphreys & Davidson 1994; Nota & Lamers 1997). They have been considered to be stars in transition to the Wolf-Rayet stage (e.g., Maeder 1983; Langer et al. 1994), but more recent observational and theoretical work suggests that some LBVs might be the immediate progenitors of SNe (e.g., Kotak & Vink 2006; Trundle et al. 2008; Gal-Yam & Leonard 2009; Smith et al. 2007; Groh et al. 2013).

A crucial parameter for understanding η Car is the luminosity because this provides a mass estimate through the Eddington limit. Most of the visible and UV light of η Car is absorbed by circumstellar dust and is reradiated in the IR (Davidson 1971; Davidson & Ostriker 1972; Pagel 1969; Westphal & Neugebauer 1969; Robinson et al. 1973; Sutton et al. 1974). The stellar luminosity can thus be obtained from the IR spectral energy distribution (SED). η Car is unparalleled as an IR radiation source, and its brightness makes observations with sensitive IR satellite instrumentation impossible. Most of the 0.25−0.44 M of cool dust is located in the central 5″ region in a “disrupted” equatorial structure (Morris et al. 1999, 2017) that is also referred to as the Butterfly nebula (Chesneau et al. 2005). Early estimates of the total luminosity of η Car are on the order of L = 5 × 106L (Neugebauer & Westphal 1968; Westphal & Neugebauer 1969; Pagel 1969; Robinson et al. 1973; Sutton et al. 1974; Cox et al. 1995; Davidson & Humphreys 1997). Morris et al. (2017) concluded that the total luminosity of η Car in the 1970s was L = 4.1 × 106L for the commonly adopted distance of 2.3 kpc. A distance revision to 2.6 kpc could imply an increase in luminosity of 25−30% (Davidson et al. 2018a). The luminosity of the secondary star is a factor of 10 lower (Mehner et al. 2010).

Some authors have suggested a change in the mid-IR flux since the 1970s. Russell et al. (1987) reported a 30% decrease in the 10.5 μm silicate emission compared to values for the 1970s. Smith et al. (1995) found a significant decrease in mid-IR flux in the 1990s compared to the 1970s. Morris et al. (2017) derived a total luminosity of L = 3.0 × 106L for the 1990s, which is a decline of 25% in 20 yr. They argued that the decline in mid-IR flux indicates a reduction in circumstellar extinction either by the expansion of dust shells or dust destruction. This could cause the simultaneous brightening at UV and optical wavelengths (Davidson et al. 1999, 2005; Martin et al. 2006).

We present newly obtained and previously unpublished mid-IR observations, which are the basis of a reevaluation of the mid-IR evolution of η Car from 1968 to 2018. In Sect. 2 we describe the new mid-IR images. We present our findings and reanalyze the SED of η Car over a period of 50 years in Sect. 3. In Sect. 4 we conclude that η Car has not undergone any appreciable long-term luminosity changes during this period.

2. Mid-IR observations in 2005 and 2018 with VISIR and in 2003 with TIMMI2

η Car was observed in 2018 with the upgraded VLT imager and spectrometer for the mid-IR (VISIR; Lagage et al. 2004; Käufl et al. 2015). The VISIR AQUARIUS detector provides a field of view of 38.0″ × 38.0″ with a pixel scale of 0.045″. We sampled the mid-IR wavelength range with several filters from 7.8−19.5 μm, see Tables 1 and B.1, and Fig. A.1. The observations were carried out at an airmass below 1.3 and at a precipitable water vapor between 1.2 and 3.5 mm, ensuring good image quality. The burst read-out mode was used, which results in a large number of short-exposure images with instantaneous point spread functions, corresponding to the momentary atmospheric turbulence. High spatial resolution images can be obtained by recentering and adding the individual short-exposure images. We reach spatial resolutions close to the diffraction limits in the given filters: 0.25−0.30″ in N band and ∼0.50″ in Q band. For the flux calibration, the science observations were either preceded or followed by observations of a mid-IR standard star (HD89682; Cohen et al. 1999) that were obtained with the same setup as the science observations.

Table 1.

Mid-IR flux densities of the spatially integrated Homunculus nebula (2003–2018) around η Car.

The data reduction was performed with a custom-made PYTHON pipeline. The photometry was determined using classical aperture photometry with an aperture radius of 11″, which encompasses the entire Homunculus nebula. VISIR absolute flux calibration is accurate to 10% in the N band and 20% in the Q band. The dominating source of uncertainty is the mid-IR sky variability. The count levels in the central bright core of the Homunculus nebula (1″ in diameter, see Fig. 1) are outside the linear regime for most filters, which implies that we may be underestimating the total flux by a few percent. We did not attempt to correct for this because longward of 8 μm less than 10% and longward of 10 μm less than 5% of the flux is produced by this central core. The resulting mid-IR spectral flux densities and statistical uncertainties are listed in Table 1.

thumbnail Fig. 1.

VISIR image of the Homunculus nebula at 12.5 μm in 2018, tracing the thermal emission from heated dust and the H I 7−6 emission. The field of view is 25″ × 25″, and the spatial resolution is 0.3″. The flux density in Jy per detector pixel (0.045″ pixel−1) is shown as well as the integrated flux along the image axes. The brightest knot in the center of the nebula is due to a shell that surrounds the star, to an inner torus or disk, or to a pin-wheel structure, and it includes the Weigelt complex (Weigelt & Ebersberger 1986). The two bright loops form the Butterfly nebula from which 50% of the integrated flux originates.

Open with DEXTER

In what follows, we reconstruct the mid-IR SED with our 2018 observations and literature data (references are provided in the caption of Fig. 2). In order to be as complete as possible, we also make use of mid-IR images in the ESO Science Archive Facility with unpublished photometry1. η Car was observed with VISIR in 2005 in the filters PAH1, PAH2, NeII, and Q2 with the former Boeing detector. Standard star observations were only obtained in the filters PAH2 and NeII, see Table B.1. The pixel scale is 0.075″ and the images have spatial resolutions of ∼0.35″. Detector artifacts, such as strong striping and ghosts, result in poor flux measurements. In January, March, and May 2003, η Car was observed with the Thermal infrared Multi-Mode instrument (TIMMI2, Reimann et al. 2000) at the La Silla 3.6 m telescope, see Table B.2. The TIMMI2 array shows smear and stripe patterns for bright sources, which lead to large errors in the flux calibration. We have averaged the flux values of all TIMMI2 observations and present the standard deviation as errors in Table 1. The TIMMI2 data should be interpreted with caution.

thumbnail Fig. 2.

Mid-IR photometry of the integrated Homunculus nebula from 1968 to 2018. Colored symbols represent our new and historic mid-IR measurements performed in the 1960s (Westphal & Neugebauer 1969), 1970s (Gehrz et al. 1973; Robinson et al. 1973; Aitken & Jones 1975; Sutton et al. 1974; Harvey et al. 1978), 1980s (Hackwell et al. 1986; Russell et al. 1987), 1990s (Smith et al. 1995; Polomski et al. 1999; Morris et al. 1999, 2017), and 2000s (Smith et al. 2003). Calibration uncertainties may be larger than reported because of unaccounted for systematic uncertainties, sky variability, detector artifacts, and partial saturation. Open symbols represent values estimated from isophotal contour maps (Hackwell et al. 1986; Smith et al. 2003).

Open with DEXTER

3. Results

3.1. Flux evolution in the mid-IR

Figure 2 compares the 2018 mid-IR photometry from our VISIR images with previous observations in the period 1968–2005. Figure 3 displays the time evolution of four mid-IR wavelength regions over 50 years and their averages. With a few exceptions, individual values are consistent with each other within 1σ (22% for M band, 13% for N band, and 19% for Q band) of the average. The figures demonstrate long-term stability of the mid-IR flux between 5 and 20 μm over 50 years within the uncertainty of mid-IR flux calibration.

thumbnail Fig. 3.

Time evolution of the mid-IR flux of the Homunculus nebula around η Car from 1968 to 2018 in four wavelength regions, chosen for best temporal coverage. Vertical dashed lines indicate periastron passages. The 1σ region of the average flux for each wavelength region is shown (averages exclude the lower ISO flux values). There is no evidence for a long-term change, but variations with the orbital period cannot be ruled out.

Open with DEXTER

The timescales for the condensation and destruction of dust grains depend on the chemistry and shielding conditions, which are complex in the Homunculus nebula around η Car. Ionizing UV and X-ray radiation that escapes from the central binary varies with orbital phase, but we cannot confirm periodic changes in the mid-IR because the observations were obtained at a mix of orbital phases, the orbital coverage is sparse, and the calibration uncertainties are large. The observations in 1971/72 indicate a trend to higher flux values in mid-cycle, similar to what is observed in the radio (White et al. 2005), but Robinson et al. (1973) stated that there is no evidence for variations from October 1971 to July 1972 larger than the measurement uncertainties. Polomski et al. (1999) found that the N band and 18 μm flux density increased between March 1997 and November 1998, that is, around periastron. The emission of the central bright core (1″ in diameter) contributes to less than 10% at 8 μm, 5% at 10 μm, 4% at 13 μm, and 2% at 19 μm. Any variations due to recent dust formation or destruction close to the star is hidden in the uncertainties of the nebula-encompassing photometry.

Morris et al. (2017) reported that the mid-IR flux of η Car decreased by 25% over the past decades2. Their result was based on data obtained in 1996 with the spectrometers on board the Infrared Space Observatory (ISO; Kessler et al. 1996). While the absolute spectrophotometric calibration of the ISO spectra was hampered by different detector materials with their own signal-dependent susceptibilities to nonlinearities and memory effects (e.g., Van Malderen et al. 2004; Morris et al. 2017), the reported decline may at least partly be due to intrinsic orbital variations. A detailed color-temperature analysis with orbital phase will reveal the thermal characteristics of short-term variations, but such a study will require a more homogeneous set of spatial and temporal monitoring observations.

In contrast to Russell et al. (1987), Smith et al. (1995), and Morris et al. (2017), our SED re-construction and investigation of the time evolution does not reveal a noticeable long-term decrease in the mid-IR flux from 8 to 20 μm since the first available mid-IR photometry in the late 1960s (see Figs. 2 and 3). The data support a rather stable mid-IR flux over 50 years, within the large uncertainties of ground-based mid-IR photometry and excluding variations within ∼25% of the mean levels. As a consequence, the UV and optical brightening since the late 1990s does not correlate with the integrated mid-IR flux. This is an important finding because the lack of any systematic mid-IR flux change rules out large changes in the luminosity of η Car. We derive a luminosity of about 4.6 × 106L for η Car, using the 2.3 μm, 3.4 μm, and 4.9 μm photometry of Gehrz et al. (1973), the 2005 and 2018 VISIR mid-IR photometry, and the flux values at 35−175 μm by Harvey et al. (1978) and at 450 μm and 850 μm by Gomez et al. (2010).

3.2. Dust extinction

Ever since the 1940s, η Car has brightened steadily at optical wavelengths (Frew 2004), and an accelerated brightening has been observed since the late 1990s (Davidson et al. 1999). Today, η Car at UV and visual wavelengths is more than 2 mag brighter than in the 1990s and the change is remarkably gray (Davidson et al. 2018b; Damineli et al. 2019). Because η Car is close to the Eddington limit, this can in principle not constitute an increase in bolometric luminosity. Davidson et al. (1999, 2005) and Martin et al. (2006) attributed this brightening to a rapid decrease in circumstellar extinction. The material around η Car is not distributed spherically. The central region is seen through a dust condensation, which strongly attenuates the central source (Hillier & Allen 1992; Weigelt et al. 1995). Davidson et al. (1999) argue that movement of a dusty condensation that intercepts our line-of-sight is unlikely, leaving dust destruction or a decrease of the dust formation rate close to the star as possible explanations. A possible cause may also be small changes in the stellar parameters, which alter the shape of the wind-wind shock cone such that we no longer look through the newly formed dust. Damineli et al. (2019) proposed the dissipation of a dusty clump (“coronagraph”) in our line of sight.

The variability in the near-IR from 1972 to 2013 was studied by Whitelock et al. (2004) and Mehner et al. (2014). The star shows a long-term brightening in JHK, whereas the L-band emission remains basically constant. The 4−8 μm flux shows no long-term variation either (Russell et al. 1987; Fig. 3). Because longward of 4 μm thermal radiation dominates the SED rather than extinction, we would expect a significant decrease in flux in the 10−20 μm range only for a net destruction of warm (∼200 K) dust grains. Nonetheless, Fig. 3 indicates a short-term increase in emission levels at mid-cycle in the 1970s and during periastron in 1998.

The extinction of the UV and optical light is probably caused predominantly by dust within the inner core. In the VISIR J7.9 filter image, the central core is resolved in at least three knots (Fig. A.1). In the PAH2 and NeII filters, for which we have comparison VISIR images in 2005, this core is unresolved. For an aperture of 0.5″ radius around the central bright source, we derive Fν,  PAH2,  2005 = 3130 ± 50 Jy and Fν,  NeII,  2005 = 2710 ± 180 Jy, compared with Fν,  PAH2,  2018 = 3140 ± 10 Jy and Fν,  NeII,  2018 = 3200 ± 30 Jy. Additional systematic errors are about 10%. There appears to be no detectable change in mid-IR flux of the inner core in the last 13 yr. This implies either an equilibrium or very weak dust formation and destruction close to the star. If the geometry of this central knot would simply expand, then the optical depth τ(UV, optical) would decline, depending on grain properties (chemistry, shape, and size). In this case, and if τ(IR) < 1, we would expect a small effect (< 2−10% depending on wavelength and hidden in the uncertainties) in the integrated mid-IR flux because the IR emission is determined by the larger nebula. Any stochastic increase or decrease in line-of-sight dust instantly changes the UV and optical extinction but not the IR emission because the dust particles are heated and will radiate.

3.3. Resolved dusty circumstellar structures

Several authors have suggested a disk-, ring-, or torus-like structure of 5−6″ in diameter around the central star, which absorbs, scatters, and reradiates the stellar radiation in the IR (Gehrz et al. 1973; Sutton et al. 1974; Hyland et al. 1979; Warren-Smith et al. 1978; Rigaut & Gehring 1995; Smith et al. 1995; Polomski et al. 1999). Morris et al. (1999) showed that most of the dust in the equatorial region is located in toroidal structures, probably created in shock-heated gas of the nineteenth century eruption. Some dust may also be located in an inner unresolved torus or disk and/or pinwheel-like structures created by colliding winds in the orbital plane (Weigelt et al. 2016). Smith et al. (2002) presented mid-IR images from 4.8 to 24.5 μm that support the hypothesis of a circumstellar torus or disk. Artigau et al. (2011) argued that this Butterfly nebula (Chesneau et al. 2005) is not a coherent physical structure or equatorial torus, but consists of spatially separate clumps and filaments that were ejected at different times. Radial velocity information from ALMA observations supports the general picture that the loops are a pinched torus in the orbital plane, perpendicular to the Homunculus lobes (Smith et al. 2018). The direction of the pinched material matches periastron/apastron orientation (Madura et al. 2013) and the companion may have played a role in disrupting the torus soon after its ejection.

The geometry of the dust surrounding η Car is quite complex, see Fig. 1 (also Smith et al. 2002). In the center, two bright ring-like structures form the Butterfly nebula. Most of the dust is located in this central region (50% of the flux originates within a 3″ radius from the central source, and 80% within a 5″ radius). In the mid-IR, we can identify strong density or temperature gradients and knots along these loops, but also spatially coherent structures.

Comparison between the 2005 and 2018 VISIR images show that the two prominent loops are expanding at a rate of up to 0.01″ yr−1 projected on sky. This corresponds to projected velocities of ∼100 km s−1 at a distance of 2.3 kpc, which is consistent with velocities in the equatorial plane (Davidson et al. 2001). The rings seem to expand without losing their overall appearance, which supports the idea that these are physically coherent structures. The farthest extent of the Butterfly nebula is about 2″ from the central source, which points toward an ejection during or in the decades before the nineteenth century eruption. We do not find any brightness changes of the inner loop-like structures between observations separated by 13 years. For an aperture of 3″ radius, we derive Fν,  PAH2,  2005 = 34170 ± 520 Jy and Fν,  NeII,  2005 = 30100 ± 2010 Jy, compared with Fν,  PAH2,  2018 = 34500 ± 150 Jy and Fν,  NeII,  2018 = 32020 ± 280 Jy. The response time for the dust formation or destruction at these distances from the star is not clear. However, the constant flux and overall appearance of these loops may imply that there is no recent dust formation or destruction and the material is simply expanding because these structures are likely optically thin.

Clues to the history and origin of mass ejection phases of η Car can be found in its circumstellar material. Hydrodynamical simulations show that a spherical mass ejection into a massive pre-existing torus of gas and dust could result in the present-day bipolar geometry of the Homunculus nebula (Frank et al. 1995; Dwarkadas & Balick 1998). The massive torus may have been created shortly before the nineteenth century eruption, for instance, through nonconservative mass transfer, leaving an unstable core that then erupted.

4. Conclusions

The mid-IR flux densities of the Homunculus nebula around η Car from 8 to 20 μm show no long-term decline since the first available mid-IR photometry in 1968. The luminosity of η Car has thus probably been stable over the past five decades (∼4.6 × 106L, adopting a distance of 2.3 kpc). Mid-IR observations were obtained irregularly and at different orbital phases. The large uncertainties of the mid-IR photometry (10% in N band and 20% in Q band) prevents the confirmation of smaller short-term fluctuations or variations with the orbital period.

Contrary to previous publications, we find no long-term decline of the mid-IR photometric levels, which would have indicated a reduction in circumstellar extinction and could have explained the increase in UV and optical brightness. The most likely scenario to explain the brightening of η Car at UV and optical wavelengths and its stability in the mid-IR is that the extinction caused by circumstellar dust declines only in our line of sight.


2

The values displayed in Fig. 7 in Morris et al. (2017) are too high for the 11 μm photometry by Gehrz et al. (1973) because of a transcription error. The flux values at 11 μm reported in Table 1 and Fig. 1 of Gehrz et al. (1973) are inconsistent by a factor of about 2. We use the values from their Table 1.

Acknowledgments

We thank the anonymous referee for the constructive feedback. MJB acknowledges support from ERC grant SNDUST 694520. We have used SketchAndCalc to calculate the areas in isophotal contour maps (E. M. Dobbs; www.SketchAndCalc.com).

References

Appendix A: Mid-IR images with VISIR in 2018

thumbnail Fig. A.1.

VISIR 2018 images of the Homunculus nebula around η Car. The flux density in Jy per detector pixel (0.045″ pixel−1) is shown on a logarithmic scale. The N-band images have a field of view of 25″ × 25″, and that of the Q-band images (last row) is 22.5″ × 22.5″.

Open with DEXTER

Appendix B: Journal of VISIR and TIMMI2 observations

Table B.1.

Journal of VISIR observations.

Table B.2.

Journal of TIMMI2 observations.

All Tables

Table 1.

Mid-IR flux densities of the spatially integrated Homunculus nebula (2003–2018) around η Car.

Table B.1.

Journal of VISIR observations.

Table B.2.

Journal of TIMMI2 observations.

All Figures

thumbnail Fig. 1.

VISIR image of the Homunculus nebula at 12.5 μm in 2018, tracing the thermal emission from heated dust and the H I 7−6 emission. The field of view is 25″ × 25″, and the spatial resolution is 0.3″. The flux density in Jy per detector pixel (0.045″ pixel−1) is shown as well as the integrated flux along the image axes. The brightest knot in the center of the nebula is due to a shell that surrounds the star, to an inner torus or disk, or to a pin-wheel structure, and it includes the Weigelt complex (Weigelt & Ebersberger 1986). The two bright loops form the Butterfly nebula from which 50% of the integrated flux originates.

Open with DEXTER
In the text
thumbnail Fig. 2.

Mid-IR photometry of the integrated Homunculus nebula from 1968 to 2018. Colored symbols represent our new and historic mid-IR measurements performed in the 1960s (Westphal & Neugebauer 1969), 1970s (Gehrz et al. 1973; Robinson et al. 1973; Aitken & Jones 1975; Sutton et al. 1974; Harvey et al. 1978), 1980s (Hackwell et al. 1986; Russell et al. 1987), 1990s (Smith et al. 1995; Polomski et al. 1999; Morris et al. 1999, 2017), and 2000s (Smith et al. 2003). Calibration uncertainties may be larger than reported because of unaccounted for systematic uncertainties, sky variability, detector artifacts, and partial saturation. Open symbols represent values estimated from isophotal contour maps (Hackwell et al. 1986; Smith et al. 2003).

Open with DEXTER
In the text
thumbnail Fig. 3.

Time evolution of the mid-IR flux of the Homunculus nebula around η Car from 1968 to 2018 in four wavelength regions, chosen for best temporal coverage. Vertical dashed lines indicate periastron passages. The 1σ region of the average flux for each wavelength region is shown (averages exclude the lower ISO flux values). There is no evidence for a long-term change, but variations with the orbital period cannot be ruled out.

Open with DEXTER
In the text
thumbnail Fig. A.1.

VISIR 2018 images of the Homunculus nebula around η Car. The flux density in Jy per detector pixel (0.045″ pixel−1) is shown on a logarithmic scale. The N-band images have a field of view of 25″ × 25″, and that of the Q-band images (last row) is 22.5″ × 22.5″.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.