3 Radial velocity results

By analysing the complete data set of the CES planet search program until April 1998 with the Austral code we obtained precise differential radial velocities for all 37 survey stars. Table 2 summarizes the RV results by giving the total RV rms-scatter, the average internal error for each star, the mean S/N-ratio of the CES spectra and the duration of monitoring by the CES survey. The internal RV measurement error is the uncertainty of the mean value of the RV distribution along one CES spectrum of the typically 90-pixel long spectral segments, for which the modeling is performed independently (see Endl et al. 2000 for a detailed description). A histogram of the RV scatter is shown in Fig. 3, with the exclusion of binaries and the 3 fainter M-dwarfs.

The average RV rms-scatter of the complete target sample (37 stars) is $24.1~{\rm m~s}^{-1}$ (in the cases of $\iota$ Hor (see next section), $\kappa$ For, HR 2400, HR 3677 (three new binaries, see Sect. 3.2) and $\alpha$ Cen A & B (see Endl et al. 2001a) we take the RV residuals after subtraction of either the planetary or stellar secondary signal). The dependence of the RV scatter on spectral type is demonstrated in Fig. 4. The average RV scatter for F-type stars is $29.8~{\rm m~s}^{-1}$ (7 stars), for G-type stars $20.7~{\rm m~s}^{-1}$ (21 stars), for K-type stars $12.3~{\rm m~s}^{-1}$ (7 stars) and for the M-dwarfs $64.4~{\rm m~s}^{-1}$ (3 stars). The scatter declines from spectral type F to K, which can be explained as the functional dependence of the measurement precision on the spectral line density (velocity information content) in the CES bandpass. In the case of the short CES spectra the RV precision is clearly depending on the total number of spectral lines within this bandpass. Since the line density is higher for stars with later spectral type, one can expect the highest achievable RV precision for K or M stars. This is the case for K-type stars as demonstrated in Fig. 4. The strong increase of scatter and internal error for the 3 M-type stars is caused by the low S/N-ratio of the obtained spectra (they are all fainter than V>9.5), which degrades the measurement precision despite their higher line density.

 

 

Table 2: Radial velocity results of all survey stars. N is the total number of analysed spectra, rms is the total scatter of the RVs, while m.int.err. gives the mean internal measurement error, the mean S/N-ratio of the spectra and T denotes the duration of monitoring (i.e. the timespan from first to last observation of this star).

Star	N	rms	m.int.err.	S/N	T
		[ ${\rm m~s}^{-1}$]	[ ${\rm m~s}^{-1}$]		[days]
$\zeta$ Tuc	51	21.6	16.7	257	1889
$\beta$ Hyi	157	23.3	20.5	161	1888
HR 209	35	23.1	19.6	151	1573
$\nu$ Phe	58	17.9	15.9	212	1927
HR 448	24	17.1	20.5	129	439
HR 506	23	23.9	23.3	173	1574
$\tau$ Cet	116	11.3	14.1	196	1889
$\kappa$ For	40	780.9	14.8	199	1890
HR 753	6	10.1	18.7	118	64
$\iota$ Hor	95	52.5	17.4	163	1976
$\alpha$ For	65	55.2	36.8	197	1890
$\zeta ^{1}$Ret	14	17.7	15.9	109	185
$\zeta ^{2}$ Ret	58	21.8	16.9	180	1977
$\epsilon$ Eri	66	13.7	9.7	174	1890
$\delta$ Eri	48	15.5	12.9	189	1889
$\alpha$ Men	41	9.8	11.3	170	1853
HR 2400	53	254.9	25.3	150	1925
HR 2667	66	16.5	21.4	144	1935
HR 3259	35	16.2	14.2	124	1852
HR 3677	34	486.1	16.7	145	1925
HR 4523	27	15.0	14.5	210	1925
HR 4979	52	14.0	12.5	185	1934
$\alpha$ Cen A	205	165.3	11.9	225	1853
$\alpha$ Cen B	291	205.1	9.9	206	1853
HR 5568	40	7.7	12.9	114	384
HR 6416	57	25.6	15.0	154	1845
HR 6998	51	19.6	22.9	137	1789
HR 7373	8	8.2	8.9	209	266
HR 7703	30	13.3	14.1	162	1042
$\phi ^{2}$ Pav	90	35.4	31.3	184	1969
HR 8323	20	19.8	17.3	147	1068
$\epsilon$ Ind	73	13.5	9.9	203	1889
HR 8501	66	34.0	17.5	184	1890
HR 8883	31	65.2	38.6	137	1259
Barnard	24	37.2	46.5	31	1414
GJ 433	15	49.9	61.0	26	337
Prox Cen	65	106.1	88.0	18	1728

Figure 3: Histogram of the RV scatter of all stars with rms < $100~{\rm m~s}^{-1}$ and without the 3 faint M-dwarfs. The distribution peaks at $14~{\rm m~s}^{-1}$, the stars with higher scatter are: HR 8883 ( $65.2~{\rm m~s}^{-1}$), $\alpha$ For ( $55.2~{\rm m~s}^{-1}$), $\iota$ Hor ( $52.5~{\rm m~s}^{-1}$), $\phi ^{2}$ Pav ( $35.3~{\rm m~s}^{-1}$) and HR 8501 ( $34.0~{\rm m~s}^{-1}$).

Figure 4: RV scatter (full circles) of the 37 target stars as a function of spectral type. Open circles represent the mean internal measurement error for the stars in each bin. Minimum for both distributions are K-type stars, which can be explained by their higher intrinsic line density, while the increase at faint M-type stars is due to the weak signal.

Appendix A (Figs. .1-.10) presents the RV results for all stars, plotted for comparison in the same time frame (JD 2 448 800 to JD 2 451 000). The near sinusoidal RV variation caused by the orbiting planet around $\iota$ Hor clearly stands out of the rest of the sample (see Fig. .3).

For the faint M dwarf Prox Cen (V=11.05) the larger rms-scatter is caused by the insufficient S/N-ratio of the CES spectra obtained with the 1.4 m CAT telescope (the average S/N-ratio of the Prox Cen spectra is only 18). The results for the inner binary (components A & B) of the $\alpha$ Centauri system were already presented in Endl et al. (2001a). The large scatter seen in the RV results for $\kappa$ For, HR 2400 and HR 3677 is caused by apparent binary orbital motion and will be discussed in detail.

3.1 The planet orbiting $\iota$ Horologii

The G0V star $\iota$ Hor (HR 810, V = 5.4) has been earlier identified as an RV variable star and thus as a "hot candidate'' in the CES survey for having a planetary companion (Kürster et al. 1998; Kürster et al. 1999a). A possible eccentric Keplerian signal with a period of 600 days was found, but with a low confidence level.

After the analysis of all 95 spectra of $\iota$ Hor using the Austral code the resulting RVs have a total rms scatter of  $52.5~{\rm m~s}^{-1}$, an average internal error of $17.4~{\rm m~s}^{-1}$ and reveal a near sinusoidal variation which is apparent during the last 2 years of monitoring (see Fig. .3). The 95 spectra were taken between November 1992 and April 1998 and have an average S/N-ratio of 163. A period search within this time series using the Lomb-Scargle periodogram (Lomb 1976; Scargle 1982) detected a highly significant signal with a period of 320 days and a very low False Alarm Probability (FAP) of <10^-11. It was possible to find a Keplerian orbital solution for these RV data and thus successfully detect an orbiting extrasolar planet. We presented this discovery already in Kürster et al. (2000) and we refer the reader to this earlier paper for a more detailed description. Here we want to summarize the orbital, planetary and stellar properties. Figure 5 displays the found Keplerian orbital solution and Table 3 lists the parameters of the planet and its orbit (note that in Kürster et al. 2000 the time of maximum RV was given wrong by one day due to a typo).

Figure 5: Keplerian orbital solution for $\iota$ Hor (solid line) plotted along the RV data. The RV semi-amplitude K is $67.0~{\rm m~s}^{-1}$, the orbital period 320.1 days and the eccentricity e = 0.16. The planet has an $m\sin i$ value of $2.26~{M}_{\rm Jup}$. The residual rms scatter around this best-fit orbit is $27.0~{\rm m~s}^{-1}$.

 

 

Table 3: Parameters of the planet and its orbit around $\iota$ Hor.

Minimum planet mass	$m\sin i=2.26\pm 0.18~{M}_{\rm Jup}$
Orbital period	$P=320.1\pm 2.1~{\rm d}$
Orbital semi-major axis	$a=0.925\pm 0.104~{\rm AU}$
Orbital eccentricity	$e=0.161\pm 0.069$
RV semi-amplitude	$K=67.0\pm 5.1~{\rm m~s}^{-1}$
Time of maximum RV	$T_{\circ }={\rm BJD} 2,450,306.0\pm 3.0$
Periastron angle	$\omega = 83^{\circ }\pm 11^{\circ }$

$\iota$ Hor b was the first planet to be detected residing entirely within the so-called "habitable zone'' (as defined in Kasting et al. 1993) of its parent star. The residual rms scatter around the orbit is $27.0~{\rm m~s}^{-1}$, larger than the error expected from the RV precision tests in Endl et al. (2000). A lot of this excess scatter is probably caused by stellar activity as it turned out that $\iota$ Hor is a quite young (ZAMS) and active star. Both the RV variation caused by the planet as well as the excess scatter have been confirmed in the meantime by Butler et al. (2001) and Naef et al. (2001).

There are indications that the $\iota$ Hor system might host additional planetary companions: the periodogram of the RV residuals (after subtraction of the orbit) reveals a peak at $P\approx620$ days. This looks intriguing especially after the detections of extrasolar planets moving in near-resonance orbits, e.g. the two companions of HD 83443 in a 10:1 resonance (Mayor et al. 2000), the planetary pair around GJ 876 in a 2:1 resonance (Marcy et al. 2001), and the two planets orbiting 47 UMa in a 5:2 resonance (Fischer et al. 2001). We demonstrate in Kürster et al. (2000) that the $P\approx620$ days peak is not due to spectral leakage from the P=320 days signal (see panel d of Fig. 1 in Kürster et al. 2000). This could indicate the presence of a second planet located close to the 2:1 resonance. However, the FAP of this peak is still above $0.1\%$ and we cannot confirm yet the presence of a second companion. After the replacement of the Long Camera at the CES with the Very Long Camera we continued to monitor $\iota$ Hor using the same I₂-cell for self-calibration. The analysis of the new data and merging it with the Long Camera data set might allow us in the near future to verify the existence of the second planet.

The CES Long Camera results also contributed to another extrasolar planet detection: our RV data for the nearby (3.22 pc) K2V star $\epsilon$ Eri add to the evidence for a long-period ( $P\approx 6.9$ yrs) planet, as presented in Hatzes et al. (2000).

3.2 Three new spectroscopic binaries: $\kappa$ For, HR 2400, and HR 3677

$\kappa$ For, HR 2400 and HR 3677 were found to be single-lined spectroscopic binaries, their large RV scatter (see Table 2) is the direct result of huge RV trends induced by high mass (stellar) companions. These trends were already discovered by an earlier analysis of a fraction of the data of these 3 stars (Hatzes et al. 1996). Now the analysis of the entire Long Camera data of $\kappa$ For and HR 3677 exhibits a curved shape of the RV trends and - in the case of $\kappa$ For - allows us to find a preliminary Keplerian orbital solution, while the very long period for HR 3677 and the linearity of the RV trend for HR 2400 prohibits this.

Figure 6: Preliminary Keplerian orbital solution for $\kappa$ For, the best-fit orbit (dashed line, $\chi ^{2}_{\rm best}=38.3$, $\chi ^{2}_{\rm red}=1.13$) is plotted along with the 40 RV measurements (diamonds).

Figure 7: Residual RV scatter of $\kappa$ For after subtraction of the best-fit Keplerian orbit (Fig. 6). The rms scatter of $14.24~{\rm m~s}^{-1}$ agrees well with the mean internal error of $14.8~{\rm m~s}^{-1}$.

The G0V star $\kappa$ For has the largest RV scatter (rms $=780.9~{\rm m~s}^{-1}$) of all stars in the CES sample. We find a preliminary Keplerian orbital solution (see Fig. 6) with the following parameters: orbital period P=7700 days, time of periastron passage T=2 454 466 JD, a low eccentricity e=0.0576, an RV semi-amplitude $K=2302~{\rm m~s}^{-1}$ and periastron angle $\omega=269.07^{\circ}$. This fit to the 40 RV measurements gives a $\chi ^{2}_{\rm best}=38.3$, and a reduced $\chi ^{2}_{\rm red}=1.13$ (with 34 degrees of freedom) and $P_{\chi}(\chi^{2})=0.28$. In other words the found preliminary Keplerian orbit represents a good fit to the RV data. By changing the value of P (and letting the remaining orbital parameters vary until $\chi ^{2}=\chi ^{2}_{\rm best}+1$) we determined the uncertainty of the period to be $\pm295$ days. Since our RV data cover only a fraction of one orbital cycle and do not constrain the orbit well enough, it was not possible to find a simultaneous solution for all orbital parameters and derive the error-range for the remaining 5 parameters. The mass function is $f(m)=(9.67\pm3.17)\times10^{-3}~{M}_\odot$ and the orbital period transforms to $a\approx 8.5$ AU. The scatter around this orbit is $14.24~{\rm m~s}^{-1}$ (Fig. 7) and consistent with the mean internal error of $14.8~{\rm m~s}^{-1}$. In the Hipparcos catalogue $\kappa$ For was given a double/multiple systems annex flag G, meaning that higher-order terms were necessary to find an adequate astrometric solution. This is an indication that $\kappa$ For is a long-term (P>10 yrs) astrometric binary, consistent with our results. The RV variabilty of $\kappa$ For was also noted by Nidever et al. (2002) who find a linear RV slope of $-1.73~{\rm m~s}^{-1}$ per day for their 7 measurements of this star.

HR 2400 (F8V) reveals a linear trend in its RV data indicating a high mass companion in a long-period orbit which does not allow us to find a Keplerian orbital solution. Figure 8 shows the best-fit linear function with a slope of $-0.42 \pm 0.005~{\rm m~s}^{-1}~{\rm d}^{-1}$ (the error range of the slope is determined by varying the value of the slope, whereby for each slope the zero-point is always fitted, until $\chi ^{2}=\chi ^{2}_{\rm best}+1$). The residual rms-scatter around this slope is $24.9~{\rm m~s}^{-1}$ which is of the same order as the average internal error of $25.3~{\rm m~s}^{-1}$ (see Fig. 9). The linearity of the trend does not allow an estimate of the mass or period of the secondary. Moreover, HR 2400 does not possess a double/multiple systems annex flag in the Hipparcos catalogue indicating that the period is indeed very long compared to the monitoring time spans of both programs (Hipparcos: 3.2 yrs, CES Long Camera: 5.2 yrs).

Figure 8: Linear fit (dashed line) to the RV data of HR 2400 (diamonds with errorbars), the residuals are shown in Fig. 9. This best fit linear trend has a slope of $-0.42 \pm 0.005~{\rm m~s}^{-1}~{\rm d}^{-1}$.

Figure 9: RV residuals of HR 2400 after subtraction of the linear slope (Fig. 8). The residual rms scatter of $24.9~{\rm m~s}^{-1}$ is consistent with the mean internal error of $25.3~{\rm m~s}^{-1}$.

Figure 10: Best parabolic fit for HR 3677 (G0III) indicating an orbital period much longer than the monitoring time ( Hipparcos astrometry gives a period of $\approx$75 years). The best-fit curved trend is plotted as dashed line along with our RV data (diamonds). The $\chi ^{2}_{\rm red}$ of this fit is 0.99, indicating a good fit. See Fig. 11 for the residuals.

The giant HR 3677 (G0III) is - with a distance of 192.31 pc - by far the most distant star in the CES sample. A parabolic fit to the RV results is shown in Fig. 10. This fit gives an acceptable description of the data with a $\chi ^{2}_{\rm red}$ of 0.99 and $P_{\chi}(\chi^{2})=0.48$. Figure 11 shows the residuals after subtraction of this best-fit curved trend. The residual rms-scatter around this orbit is $20.0~{\rm m~s}^{-1}$, slightly larger than the average internal error of  $16.7~{\rm m~s}^{-1}$. From the Hipparcos measurements of HR 3677 a two-component astrometric solution was derived. The angular separation of the components is given as $0.131\pm$0.010 arcsec which corresponds at the distance of 192.31 pc to a minimum orbital separation of $\approx$$25\pm2$ AU. The orbital period would be around 75 years, too long to determine a Keplerian solution, but it seems to be consistent with the RV-variation we find for HR 3677.

Figure 11: Residual RVs of HR 3677 after subtraction of the best-fit curved trend (shown in Fig. 10). The rms-scatter is $20.0~{\rm m~s}^{-1}$ slightly larger than the mean internal error of $16.7~{\rm m~s}^{-1}$. This larger residual scatter is primarily caused by the one outlier at JD 2 450 350; without this data-point the scatter is reduced to $14.7~{\rm m~s}^{-1}$ and is consistent with our measurement errors.