A&A 464, 1045-1047 (2007)
DOI: 10.1051/0004-6361:20066554

Interferometric observations of $\eta $ Carinae with VINCI/VLTI[*]
(Research Note)

P. Kervella

LESIA, UMR 8109, Observatoire de Paris-Meudon, 5 place Jules Janssen, 92195 Meudon Cedex, France

Received 12 October 2006 / Accepted 28 November 2006

Abstract
Context. The bright star $\eta $ Carinae is the most massive and luminous star in our region of the Milky Way. Though it has been extensively studied using many different techniques, its physical nature and the mechanism that led to the creation of the Homunculus nebula are still debated.
Aims. We aimed at resolving the central engine of the $\eta $ Carinae complex in the near-infrared on angular scales of a few milliarcseconds.
Methods. We used the VINCI instrument of the VLTI to recombine coherently the light from two telescopes in the K band.
Results. We report a total of 142 visibility measurements of $\eta $ Car, part of which were analyzed by Van Boekel et al. (2003, A&A, 410, L37). These observations were carried out on projected baselines ranging from 8 to 112 m in length, using either two 0.35 m siderostats or two 8-m Unit Telescopes. These observations cover the November 2001-January 2004 period.
Conclusions. The reported visibility data are in satisfactory agreement with the recent results obtained with AMBER/VLTI by Weigelt et al. (2006), asuming that the flux of $\eta $ Car encircled within 70 mas reaches 56% of the total flux within 1400 mas, in the K band. We also confirm that the squared visibility curve of $\eta $ Car as a function of spatial frequency follows closely an exponential model.

Key words: stars: individual: $\eta $ Carinae - stars: circumstellar matter - technique: interferometric - stars: early-type

1 Introduction

$\eta $ Carinae, the brightest example of the S Doradus class of stars, is the most massive, most luminous star in our region of the Milky Way. Over the last two hundred years $\eta $ Car has shown many signs of violent activity, with in particular a spectacular eruption in the 1840s that created the Homunculus nebula. The study of $\eta $ Car raises important questions about how the most massive stars may end their lives. The central object was studied by Weigelt & Ebersberger (1986) and Falcke et al. (1996) using speckle interferometry at an angular resolution of the order of 30 milliarcsec (mas). This revealed a complex structure with several equatorial blobs at distances of 0.1 to 2 arcsec from the star, but the central engine remained unresolved. Long baseline interferometry, currently the only technique allowing the mas resolution necessary to resolve $\eta $ Car, was recently applied to this star in the near- and mid-infrared domains by Van Boekel et al. (2003), Chesneau et al. (2005) and Weigelt et al. (2007). At the estimated distance of $\eta $ Car of 2.3 kpc (Davidson & Humphreys 1997; Davidson et al. 2001; Smith 2006), one mas corresponds to 2.3 AU. We report in this Research Note the complete corpus of VINCI observations of $\eta $ Car in the K band, including those discussed by Van Boekel et al. (2003).

  
2 Observations

The Very Large Telescope Interferometer (VLTI, Glindemann et al. 2003) has been operated by the European Southern Observatory on top of the Cerro Paranal, in Northern Chile since March 2001. For the present work, the light from $\eta $ Car and its calibrators was collected either by two 0.35 m VLTI Test Siderostats or two 8 m Unit Telescopes (UTs) without adaptive optics. It was subsequently recombined coherently in the VINCI instrument using a K band filter ( $\lambda=2.0{-}2.4~\mu$m).

We have observed $\eta $ Car repeatedly over the period November 2001 to January 2004. This resulted in a total of 71 000 interferograms on this target, out of which 50% (35 639) were selected automatically by the pipeline. Approximately the same quantity of data were obtained on the calibrators. We used the standard VINCI data reduction pipeline (Kervella et al. 2004, version 3.1) to derive instrumental visibilities. The calibration of $\eta $ Car's visibilities was done using well-known reference stars selected in the Bordé et al. (2002) and Cohen et al. (1999) catalogues, except $\beta$ Car. The diameter of $\beta$ Car was computed from an interferometric measurement obtained with the Intensity Interferometer (Hanbury Brown et al. 1974). The original V band uniform disk (UD) angular diameter was converted into a K band uniform disk angular diameter ( $\theta_{\rm UD} = 1.54 \pm 0.10$ mas) using linear limb darkening coefficients from Claret et al. (2000). Thanks to the relatively low values of $\eta $ Car's visibilities, the systematic uncertainty due to the calibrators is in general a small fraction of the total error bars.

The calibrated visibility values obtained on $\eta $ Car are listed in Table 1. Thanks to the use of several different telescope configurations and to the supersynthesis effect, we were able to cover a broad range of baseline lengths and azimuth. The (u,v) coverage of our observations is presented in Fig. 1.


  \begin{figure} \par\includegraphics[width=8cm,clip]{aa6554f1.eps} \end{figure} Figure 1: (u,v) coverage of the $\eta $ Car observations. Large spots represent the UT observations and small spots the siderostat observations (North is up and East is to the right).
Open with DEXTER

   
3 Effective wavelength

The VINCI instrument has no spectral resolution and its bandpass corresponds to the K band filter (2.0-2.4 $\mu$m). It is thus important to compute the precise effective wavelength of the instrument in order to determine the spatial frequency of the observation. The true effective wavelength differs from the filter mean wavelength mainly because of the object spectrum shape, the detector quantum efficiency, and the fiber beam combiner transmission. To derive the effective wavelength of our observations, we computed a model taking into account $\eta $ Car's spectrum. The instrumental transmission of VINCI and the VLTI was measured on bright reference stars with the UTs (see Kervella et al. 2003, for details). Due to the extraordinarily dense, opaque stellar wind, the shape of the $\eta $ Car spectrum in the infrared is different from the curve of a black body at the effective temperature of the central object. In particular, the flux is increasing by about 20% from 2.0 to 2.5 $\mu$m (Smith 2002). In our model, no spectral line either in emission or absorption has been taken into account, considering the relatively limited contribution of these spectral features to the total flux in the K band. Taking the average wavelength of this model spectrum gives an effective wavelength of $\lambda_{{\rm eff}} = 2.196~\mu$m for our $\eta $ Car observations, slightly longer than the typical 2.179 $\mu$m value for solar-type stars. We estimate the uncertainty on this effective wavelength to less than $\pm$0.5%, or $\pm$$0.01~\mu$m.

  
4 Interferometric field of view

When injecting the light from an extended astronomical object into a single-mode fiber, the wavefront corrugation by the atmosphere (loss of coherence) is converted into photometric fluctuations. They are easily corrected during the data processing using the dedicated photometric channels of VINCI. Unfortunately, the restoration of coherence by spatial filtering comes at the expense of a very small field of view (FOV). It is well approximated by the diffraction pattern of a telescope whose size is the geometric mean of the apertures of the two telescopes. In the case of our homogeneous two-UT observations, the FOV is thus 70 mas in the K band. Considering the extension of the $\eta $ Car complex, this limited FOV has an impact on the measured visibilities.

Guyon (2002) studied in detail this limitation for the interferometric observation of extended objects. One important conclusion is that the effective FOV depends on the seeing, and so does the visibility. This is particularly true when large telescopes are used without adaptive optics, as this was the case for our observations. While all the UTs are now equipped with MACAO adaptive optics systems (Arsenault et al. 2004), the early observations reported here were all obtained with atmosphere limited point spread functions. The atmospheric turbulence creates a large cloud of speckles on the fiber head, and incoherent light coming from separate parts of the object is coupled into the fiber, therefore reducing the contrast of the fringes. As a second order effect, different local seeing conditions for the two UTs could also slightly degrade the visibilities. In the case of small objects such as single stars, this effect is negligible, but $\eta $ Car is surrounded by a large and bright envelope that is resolved by the UTs and contributes significantly to the light distribution within the FOV.

Practically, this means that the visibility measurements obtained with the UTs should be debiased from the seeing fluctuations. Unfortunately, this is not an easy task because the relationship between the speckle cloud size (defined by the seeing) and the flux coupled into the optical fiber is unknown. Tentatively, we mention as a first estimation of the UTs FOV the observatory seeing in the K band at the time of the observations. The seeing values from the Paranal DIMM, obtained at $\lambda = 0.5~\mu$m have been converted to the K band assuming a classical $\lambda^{-6/5}$ dependance. Future comparisons of the visibility measurements reported in the present Note with results from other instruments should take into account their relative interferometric FOV.

On the other hand, the observations obtained with the 0.35 m siderostats are in principle not affected by this bias because most of the $\eta $ Car flux is coming from an area on the sky that is contained into the Airy pattern of these telescopes. Therefore, the obtained visibility is expected to be a faithful measurement of $\eta $ Car's intrinsic visibility in the 1.40 arcsec FOV of the siderostats. For the E0-G1 baseline, many visibility points have been obtained on different nights, with a broad range of seeing conditions. The fact that they give very consistent visibility values is a confirmation that the FOV variation is negligible for the siderostats.

5 Discussion

Figure 2 shows a comparison of the VINCI squared visibilities with the AMBER model fitting result of Weigelt et al. (2007), represented as a thick curve. The VINCI squared visibilities show a strong decrease with increasing spatial frequencies, clearly indicating that the central source is resolved by the interferometer. The measurements obtained with the UTs, though in principle affected by an uncertainty due to the variation of the FOV with the seeing, are roughly consistent with the siderostat data obtained on comparable baselines. The simple model developed by Hillier et al. (2001, 2006), was adjusted by Weigelt et al. (2007) to the AMBER observations of $\eta $ Car in the continuum at $\lambda=2.174~\mu$m. This model is well reproduced by an exponential curve following the expression:

\begin{displaymath}V^2 = 1.008~\exp~(-0.016\ s), \end{displaymath} (1)

where $s=B/\lambda$ is the spatial frequency. Our wavelength reference is $\lambda=2.196~\mu$m (Sect. 3). On the same figure, the dashed curve is an exponential fit to the VINCI data:

\begin{displaymath}V^2 = 0.322~\exp~(-0.016\ s). \end{displaymath} (2)

The slopes (in logarithmic scale) of the VINCI data fit and the model representing the AMBER measurements are in excellent agreement. However, the ratio of the two (VINCI/AMBER) is $\rho^2 = 32$% in squared visibilities, translating into a factor $\rho = 56$% in visibilities. This ratio is constant with the spatial frequency, the signature of a fully resolved component.
  \begin{figure} \par\includegraphics[width=8cm,clip]{aa6554f2.eps}\end{figure} Figure 2: Squared visibilities obtained on $\eta $ Car with VINCI, compared to the model fitting result of Weigelt et al. (2007), represented as a solid curve. The UT data points are represented with open squares.
Open with DEXTER

To estimate the contribution of this extended component, we can consider the FOV of the two instruments. While the AMBER observations were obtained with the MACAO adaptive optics system in function (the FOV was thus $\approx$70 mas), the FOV of the VINCI siderostat observations was much larger, $\approx$1400 mas. From the observed ratio $\rho$ between the visibilities measured by VINCI and AMBER, we can infer that 56% of the 1400 mas encircled K band flux of $\eta $ Car comes from within the 70 mas point spread function of a single UT. This value is nicely consistent with the independent measurement by Van Boekel et al. (2003), based on adaptive optics observations with the NACOinstrument, that gives an encircled energy of 57% within 70 mas. When corrected for the contribution of the extended emission, the visibilities measured by AMBER and VINCI are in excellent agreement.

A discussion of the shape of the dense stellar wind of $\eta $ Car can be found in Smith et al. (2003) and Van Boekel et al. (2003). To improve the currently simplified spherical models, this observable appears highly desirable. The operating VLTI instruments are now routinely providing spectro-interferometric datasets on $\eta $ Car (Weigelt et al. 2007; Chesneau et al. 2005), and the planned second generation will combine at least four telescopes, allowing to obtain rich data cubes at mas scales. This is an essential effort to follow the extremely fast evolution of $\eta $ Car (Martin et al. 2006). In this context, the simple, two-telescopes, broadband VINCI data provide an interesting fiducial.

Acknowledgements
Based on observations made with ESO's VLT Interferometer at Cerro Paranal, Chile. The VINCI data were retrieved from the ESO/ST-ECF Archive. This research made use of the SIMBAD and VIZIER databases at the CDS, Strasbourg (France), and of NASA's Astrophysics Data System Bibliographic Services.

References

 

  
Online Material

Table 1: Squared visibilities measured with VINCI on $\eta $ Car. The seeing in the K band at the time of the observation is given as the FOV with the UTs (see Sect. 4). N is the number of processsed interferograms, B the baseline length, and Az. the azimuth angle of the projected baseline (North = 0$^\circ $, East = 90$^\circ $). The squared visibility values and error bars are expressed in percents. The statistical and systematic (from the calibrator star estimated angular size) error contributions are given separately.


Copyright ESO 2007