2 The Coudé Echelle Spectrometer planet search program at La Silla

The search for extrasolar planets in the southern hemisphere using the Coudé Echelle Spectrometer at ESO La Silla was started in November 1992 (Kürster et al. 1994; Hatzes et al. 1996). At the beginning, the CES survey was a "classical'' RV planet search program in the sense that at a time when no extrasolar planet had been found, the common expectation was to discover planets similar to Jupiter. Thus the observing strategy was tailored for long-period and low amplitude signals. Observations were performed on an irregular temporal basis, starting with 2-night runs performed every other month. The sampling density was later increased, after the discovery of the short-period planet around 51 Peg by Mayor & Queloz (1995), to assure also detection capability for planets of this type.

2.1 The target sample

At the beginning of the survey targets were selected according to the following criteria: stars with V < 6 (with few exceptions) to attain a sufficient S/N-ratio, spectral type F8V-M5V, stars with declination < $10^{\circ}$ to avoid overlap with surveys in the northern hemisphere (again with some exceptions), known (at that time) close binaries were rejected (with the $\alpha$ Centauri system being another exception) and known active stars were neglected.

The final target list consists of 37 bright late-type stars (mostly) in the southern hemisphere: 6 F-, 21 G-, 7 K-, and 3 M-type stars. Table 1 summarizes all targets with their HR, HD and GJ catalogue number, spectral classification, V-magnitudes, distance in parsecs, chromospheric emission indices (log  $R^{/}_{\rm HK}$) and their X-ray luminosity ($L_{\rm X}$). Distances are based on the Hipparcos parallaxes (ESA (1997)), the log  $R^{/}_{\rm HK}$-values are taken from the survey of Ca II H&K emission in southern solar-type stars by Henry et al. (1996) and the $L_{\rm X}$ values are coming from the RASS (ROSAT All Sky Survey) results (Hünsch et al. 1998, 1999). Chromospheric emission in the cores of the Ca II H&K lines and the X-ray luminosities serve as stellar activity indicators since increased chromospheric Ca II H&K and coronal X-ray emission is a common sign of active stars. Stellar activity can produce intrinsic RV variability and adds an additional noise source into the RV measurement (e.g. Saar & Donahue 1997). It can even mimic the RV signature of a short-period planet like in the case of HD 166435 (Queloz et al. 2001). In 1992 the Ca II and X-ray results were not known, which would have probably led to the exclusion of stars like $\kappa$ For, $\iota$ Hor, $\alpha$ For and HR 8883. Figures 1 and 2 show the V-magnitude and distance histograms of the CES target sample.

Not all targets were monitored since November 1992: three targets, HR 753, HR 5568 and GJ 433, were added to the sample in May 1997 as Hipparcos candidates for having short-period substellar companions (H.-H. Bernstein & U. Bastian priv. comm.), while HR 7373 was only observed for a short time in 1996 and 1997 searching for a short-period planetary companion.

Although listed as a target, $\tau$ Cet was mainly included as a reference object since it is a known long-term RV-constant star (Campbell et al. 1988; Walker et al. 1995) and two other stars with known extrasolar planets, 51 Peg (Mayor & Queloz 1995) and 70 Vir (Marcy & Butler 1996), were also observed after 1995 to serve as RV precision check stars.

 

 

Table 1: Target list of the CES planet search program. HR, HD and GJ catalogue numbers, spectral type, visual magnitude V, distance in parsec (based on Hipparcos parallaxes), chromospheric emission index from Henry et al. (1996) and X-ray luminosity $L_{\rm X}$ from Hünsch et al. (1998, 1999) are given.

HR	HD	GJ	Name	Sp.type	V	d	log $R^{/}_{\rm HK}$	$L_{\rm X}$
					[mag]	[pc]		$10^{27}~{\rm [erg/s]}$
77	1581	17	$\zeta$ Tuc	F9V	4.2	8.59	-4.85
98	2151	19	$\beta$ Hyi	G2IV	2.8	7.47	-4.99	6.4
209	4391	1021		G1V	5.8	14.94	-4.55	55.8
370	7570	55	$\nu$ Phe	F8V	4.96	15.05	-4.95	14.3
448	9562	59.2		G2IV	5.76	29.66	-5.10
506	10647	3109		F9V	5.52	17.35
509	10700	71	$\tau$ Cet	G8V	3.5	3.65	-4.96	1.1
695	14802	97	$\kappa$ For	G0V	5.19	21.93		397.1
753	16160	105A		K3V	5.82	7.21	-4.85	1.8
810	17051	108	$\iota$ Hor	G0V	5.41	17.24	-4.65	68.3
963	20010		$\alpha$ For	F8V	3.87	14.11		524.6
1006	20766	136	$\zeta ^{1}$Ret	G2.5V	5.54	12.12	-4.65	5.8
1010	20807	138	$\zeta ^{2}$ Ret	G1V	5.24	12.08	-4.79
1084	22049	144	$\epsilon$ Eri	K2V	3.73	3.22	-4.47	20.9
1136	23249	150	$\delta$ Eri	K0IV	3.51	9.04	-5.22	0.9
2261	43834	231	$\alpha$ Men	G6V	5.1	10.15	-4.94	2.9
2400	46569	1089		F8V	5.58	37.22
2667	53705	9223A		G3V	5.54	16.25	-4.93
3259	69830	302		G7.5V	5.95	12.58		3.0
3677	79807			G0III	5.86	192.31
4523	102365	442A		G3V	4.91	9.24	-4.95
4979	114613	9432		G3V	4.85	20.48	-5.05	23.0
5459	128620	559A	$\alpha$ Cen A	G2V	-0.01	1.347	-5.00	2.2
5460	128621	559B	$\alpha$ Cen B	K1V	1.33	1.347	-4.92	2.2
5568	131977	570A		K4V	5.74	5.91	-4.48	3.5
6416	156274	666A		G8V	5.47	8.79	-4.94	1.9
6998	172051	722		G4V	5.86	12.98	-4.89	4.3
7373	182572	759		G8IV	5.16	15.15		3.9
7703	191408	783A		K3V	5.31	6.05	-4.99
7875	196378	794.2	$\phi ^{2}$ Pav	F8V	5.12	24.19
8323	207129	838		G0V	5.58	15.64	-4.80	11.2
8387	209100	845	$\epsilon$ Ind	K4.5V	4.69	3.63	-4.56	1.6
8501	211415	853A		G3V	5.33	13.61	-4.86	12.2
8883	220096			G4III	5.66	100.81		34175
		699	Barnard	M4V	9.54	1.82		0.1
		433		M2V	9.79	9.04
		551	Prox Cen	M5Ve	11.05	1.29		1.7

Figure 1: Histogram of V-magnitudes of the CES targets. The distribution peaks in the magnitude range of 5-6 mag while the 3 M-dwarfs form the faint "tail'' on the right side.

Figure 2: Histogram of the distances in the CES sample. Not shown are the two stars with distances of more than 100 pc: HR 3677 (192.3 pc) & HR 8883 (100.8 pc). The rest of the targets are located within 50 pc, with the bulk lying closer than 20 pc. Distances are based on Hipparcos parallaxes (ESA 1997).

2.2 Observations

All stars were observed with the 1.4 m Coudé Auxiliary Telescope (CAT) on La Silla which fed the CES via a direct beam from the telescope. All spectra were taken in a single echelle order centered at a wavelength of $5389~{\rm\AA}$. The Long Camera yields a resolving power of R=100 000 and a small spectral range of 48.5 Å. After the installation of the Very Long Camera in April 1998 the resolving power of the CES was raised to R=230 000 but the spectral coverage was even reduced (depending on which CCD was used). The results presented in this work all refer to the Long Camera configuration (R=100 000) prior to this modification, which thus form a homogeneous data set.

In order to assure the necessary high long-term precision for RV measurements all CES spectra are self-calibrated by a superimposed absorption spectrum of molecular iodine (I₂) vapor. This is achieved by passing the starlight through a temperature controlled cell filled with I₂ (see also Kürster et al. 1994).

Typical exposure times of the CES survey were 10 to 15 min, and the S/N-ratios of the obtained spectra were in the range of 100 to 250. For the brightest targets exposure times were much shorter (in the order of 10 to 30 s), while for the faint M-dwarfs we set a maximum exposure time of 30 min, in order to minimize timing uncertainties and subsequent systematic errors in the barycentric velocity correction.

2.3 Data analysis

To extract the RV information from I₂ self-calibrated spectra it is necessary to perform a full spectral modeling. For the analysis of the CES planet search data we employ the Austral code which establishes a model of the observation based on high resolution templates of the stellar and the I₂ spectrum. For a detailed description of this analysis technique we refer the reader to Paper I of this series (Endl et al. 2000). The modeling process includes the reconstruction of the shape and asymmetry of the spectrograph instrumental profile (IP) as well as Maximum Entropy Method deconvolution to obtain a higher resolved stellar template spectrum. All computations are carried out on an oversampled sub-pixel grid and a multi-parameter $\chi ^{2}$-optimization is performed to achieve a best-fit model. The algorithm follows in general the modeling idea first outlined by Butler et al. (1996) and IP reconstruction techniques by Valenti et al. (1995).

The main limiting factor for the achievable RV precision with the CES is the small spectral bandwidth of $48.5~{\rm\AA}$. Using different test scenarios we demonstrated in Endl et al. (2000) that a long-term RV precision of 8- $15~{\rm m~s}^{-1}$ was attained.