A Planet with a 3.1 Day Period around a Solar Twin¹

Of the 10 planetary companions found from precision Doppler surveys, seven have semimajor axes of less than 0.25 AU (Mayor et al. 1998; Marcy & Butler 1998; Noyes et al. 1997; Cochran et al. 1997). Four of these are "51 Peg" cases with semimajor axes of ∼0.05 AU. While precision Doppler surveys are most sensitive to planets in small orbits, this "piling‐up" of planets near host stars may be a real effect, rather than purely due to observational bias. In particular, the ongoing 11 year survey of 107 F–M dwarf stars at Lick Observatory would have easily detected Jupiter‐mass planets orbiting between 0.5 and 1.5 AU, yet none has been found. In contrast, four stars from the Lick survey have yielded Jupiter‐mass companions within 0.25 AU (Marcy et al. 1998; Butler et al. 1997).

This high occurrence of planets at small orbital distances has led to a variety of models in which Jupiter‐like planets form beyond the ice boundary in the proto–planetary disk, then suffer orbital migration. Early on, protoplanets entrained in gaps within their disks can be dragged toward the host star along with the disk (Lin, Bodenheimer, & Richardson 1996; Trilling et al. 1998). In the post‐gas era, gravitational interactions between planets can result in orbits with small semimajor axes and large eccentricities (Rasio & Ford 1996; Levison, Lissauer, & Duncan 1998).

Our understanding of the distribution of planets within 1.5 AU (P ∼ 2 yr) will improve within the next 2 years as new Doppler surveys approach a 4 year duration. This paper reports the detection of a Keplerian Doppler variation of the solar‐type star HD 187123 (G3 V, V = 7.83) with a 3.1 day period. This is the first planet found solely from the Keck Planet Survey. Observations and the orbital solution are reported in § 2. Alternative astrophysical sources for the observed periodicity are considered in § 3. A discussion follows in § 4.

Photometry for HD 187123 (=HIP 97336) from Hipparcos yields V = 7.83 and B - V = 0.646 ± 0.01 (transformed from Perryman et al. 1997). For comparison, the Sun has B - V = 0.648 ± 0.006 (Gray 1995), identical to that of HD 187123 within errors. Its parallax is 20.87 mas, giving an absolute visual magnitude of M_V = 4.43 mag. Applying a bolometric correction of BC = -0.06 gives an absolute bolometric magnitude of M_BOL = 4.37, or a luminosity of L = 1.35 L_⊙. Spectral synthesis yields [Fe/H] = +0.16 and T_eff = 5830 ± 40 (Gonzalez 1998).

Our optical spectrum shows HD 187123 to be identical to the Sun within 1% (rms), with no enhancement of the iron line depths. Gonzalez (1998) finds that the equivalent widths of the lines in HD 187123 are 4–5 mÅ greater than those in the Sun, which combined with the slightly higher derived temperature yields the metallicity estimate. Clearly, HD 187123 has properties remarkably similar to those of the Sun, albeit 35% brighter, perhaps caused by metallicity enhancement. Based on the close spectral match with the Sun, we estimate that the mass of HD 187123 is M = 1.0 ± 0.1 M_⊙.

A total of 20 Doppler velocity measurements of HD 187123 have been obtained on Keck I with the HIRES echelle spectrometer (Vogt et al. 1994). The resolution for these spectra was R = 87,000, and they span wavelengths from 3900 to 6200 Å. The calibration of wavelength and the measurement of the spectrometer point‐spread function was determined for each exposure and for each 2 Å chunk of spectrum by using iodine absorption lines superimposed on the stellar spectrum (Butler et al. 1996).

The first two velocities of HD 187123 were taken 6 months apart (1997 December 23 and 1998 June 18) and differed by 84 m s⁻¹. Thus alerted, we observed the star during 5 consecutive nights, 1998 July 15–19 (UT). The first 2 nights revealed a velocity change of 115 m s⁻¹, prompting us to obtain multiple exposures during the remaining 3 nights. The 10 observations from the 5 night campaign are shown in Figure 1.

Zoom In Zoom Out Reset image size

A 3 day periodicity is immediately evident. A phased version of the entire data set, spanning 1997 December–1998 August, is shown in Figure 2. Periodogram analysis and string length minimization codes find a predominant periodicity between 3.096 and 3.100 days. The best‐fit Keplerian has a period of 3.097 days, a semiamplitude K = 73.0 m s⁻¹, and an eccentricity e = 0.03. The rms to the Keplerian fit is 7.50 m s⁻¹. A sinusoidal fit yields the same period with a slightly reduced semiamplitude of 71.6 m s⁻¹. The rms of the sinusoidal fit is 7.62 m s⁻¹, essentially indistinguishable from the Keplerian fit. The best‐fit sinusoid is shown in Figures 1 and 2.

Zoom In Zoom Out Reset image size

The velocities have typical internal uncertainties of 6 m s⁻¹. The errors stem from the Poisson statistics of the photons acquired (∼3 m s⁻¹) and long‐term systematic errors (∼4 m s⁻¹). The former contribution is estimated for every exposure from the uncertainty in the mean of the velocities (σ/√N_chunk) from the many 2 Å wide chunks into which the spectrum was divided.

The best‐fit orbital parameters (shown in Table 1) imply a companion mass of Msin i = 0.52 M_JUP and a semimajor axis of a = 0.0415 ± 0.003 AU. Assuming unbiased orbital inclinations, the likely mass of the companion is between 0.6 and 1 M_JUP.

In addition to being the easiest to detect, the planets with 3–5 day orbits have proven to be the most controversial. Astrophysical effects that can mimic a short‐period Keplerian Doppler velocity signal include radial and nonradial pulsations and rotational modulation due to surface features (i.e., spots). For example, Mayor et al. (1998) report that the solar‐type star HD 166435 (G0 V) exhibits Doppler variations with a 3.8 day period and a semiamplitude of about 80 m s⁻¹, consistent with a planet. Indeed, phase stability is maintained for 100 periods. Photometric monitoring has found the same 3.8 day period and a semiamplitude of 0.06 mag, indicating spots as the cause (Queloz & Henry 1998).

Thus, the question arises regarding spots on HD 187123. The 155 photometric measurements of HD 187123 by Hipparcos exhibit an rms scatter of only 0.015 mag, consistent with noise (Perryman et al. 1997). A search for periodicities in the photometric data near 3.10 days yielded no semiamplitudes above the 0.01 mag limit, in contrast to the 0.06 mag photometric variations of HD 166435. Thus, if HD 187123 has spots, they create at most 1 / 4 of the photometric signal that was induced in HD 166435.

In addition, HD 187123 exhibits much lower chromospheric activity than does HD 166435. Figure 3 shows a spectroscopic comparison of HD 187123 and HD 166435 in the core of the Ca ii H line. The solar spectrum (Kurucz et al. 1984) is shown for comparison. The emission in HD 166435 indicates strong chromospheric activity, in marked contrast to HD 187123, which shows virtually no emission. We find a chromospheric emission measure for HD 166435 of log R^'(HK) = -4.19, which indicates youth, rapid rotation, and strong magnetic activity (Noyes et al. 1984). Our derived chromospheric measure for HD 187123 is log R^'(HK) = -4.93, which is nearly identical to the Sun and typical for chromospherically (and magnetically) quiet stars.

Zoom In Zoom Out Reset image size

The rapid rotation of HD 166435 is also indicated by the broadened absorption lines, from which we derive Vsin i = 8 ± 2 km s^-1. Similar stars in the Lick Observatory planet survey show velocity variability of 20 to 50 m s⁻¹ (Saar, Butler, & Marcy 1998). In contrast, we find Vsin i<2 km s^-1 for HD 187123. This is consistent with the rotation period of HD 187123, as estimated from Ca ii H and K,P_ROT ≈ 30 days (Shirts & Marcy 1998; Noyes et al. 1984).

Apparently, the Doppler variations in HD 166435 are caused by spots, as manifested in the photometric variations, high rotation, and chromospheric emission. Such activity diagnostics are virtually absent in HD 187123; hence its Doppler variations are probably not related to magnetic activity.

Gray (1997) has proposed that slowly rotating solar‐type stars may undergo nonradial pulsations that mimic the Doppler velocity signal of orbiting planets. Such pulsations have now been ruled out for the cases of 51 Peg and τ Boo (Brown et al. 1998; Gray 1998; Hatzes, Cochran, & Bakker 1998), but similar studies have not been carried out for HD 187123.

While Keplerian orbital motion appears to be the most probable explanation for the observed Doppler variations in HD 187123, line bisector studies and precision photometric monitoring will be required to rule out other possibilities. In addition, photometric monitoring will be useful in searching for transits. Assuming random orbital inclination, the probability is 11% that the planet will transit, in which case a Jupiter‐sized planet will block about 1% of the light.

HD 187123 appears to have a planet with the shortest orbital period yet found, 3.097 days. Its mass is Msin i = 0.52 M_JUP, implying a likely mass under 1.5 M_JUP. The previously discovered planets in 3–5 day orbits, 51 Peg, υ And, and τ Boo, have nearly circular orbits. The small eccentricity for HD 187123 is thus not surprising. It is expected that tidal effects would circularize a gas giant planet within 1 Gyr and a solid rocky planet in less than 1000 years (Rasio & Ford 1996; Marcy et al. 1997; Ford, Rasio, & Sills 1998). On the other hand, the tidal torques on HD 187123 have not spun up the star to the orbital period of the planet (30 days vs. 3.1 days). This lack of spin‐up constrains the mass of the planet to be less than 15 M_JUP, similar to the lack of spin‐up for the other close jupiters (Ford et al. 1998). The importance of tidal spin‐up for τ Boo remains unclear.

The Keck Planet Survey began in 1996 July and is currently surveying 420 main‐sequence stars with spectral types between late F and mid M. After 2 years this survey is now sensitive to Jovian‐ (and Saturn‐) mass planets orbiting within 1 AU. Given the obvious selection effects, it is not surprising that the first planet announced from this survey is of the "51 Peg" variety.

The Lick Planet Survey of 107 stars has been taking data with precision similar to the Keck project for 4 years and is thus sensitive to Jovian‐mass planets orbiting out to 2 AU. A histogram of the distribution of the orbital radii for the planets found in the combined Lick and Keck surveys is shown in Figure 4.

Zoom In Zoom Out Reset image size

The apparent piling‐up of planets within 0.3 AU is probably more than a selection effect. The Lick survey would have detected nearly all Jovian‐mass companions out to 2 AU. There is a suggestion that nature manufactures more Jovian‐mass companions in very small orbits (<0.25 AU) relative to orbits between 0.5 and 1.5 AU.

Standard formation theories require gas giant planets to initially form at several AU, beyond the ice boundary in the protodisk (Lissauer 1995; Boss 1995). Protoplanets may then suffer orbital migration through interactions with the disk (Lin et al. 1996; Ward 1997; Trilling et al. 1998). Planets that move into very small orbits (a ∼0.05 AU) may be stabilized against further loss of angular momentum either by tidal interactions with the spin of the star or by the clearing of the inner disk by the stellar magnetosphere (Lin et al. 1996). These mechanisms are incapable of stopping the migration of planets beyond 0.1 AU; thus they might explain the relative dearth of planets between 0.5 and 1.5 AU, but they cannot explain the existence of planets orbiting between 0.1 and 0.3 AU.

Alternatively, planetary orbits can evolve in the disk, after the gas has left, via gravitational interactions between multiple giant planets (Weidenschilling & Marzari 1996; Rasio & Ford 1996; Lin & Ida 1997; Levison et al. 1998). The extant Doppler measurements are insufficient to reveal the presence of any other planets. If the companion to HD 187123 was gravitationally scattered into its present location, other giant planets presumably once existed in the system and perhaps still do exist. This scattering mechanism requires some mechanism by which the planet lost orbital energy within 0.04 AU of the star. It does not seem likely that this mechanism would favor the creation of Jovian‐mass planets orbiting inside of 0.3 AU relative to orbits near 1 AU.

Rayleigh scattering is expected to increase the planet's reflected flux by several orders of magnitude shortward of 4200 Å, while light in the optical is expected to be absorbed by thermalization (Seager & Sasselov 1998). Thus direct detection at optical wavelengths will be difficult.

The effective temperature of a planet in a 3.097 day orbit about HD 187123 would be similar to the companions found 51 Peg and τ Boo, about 1400 K (Burrows et al. 1997). Like the previous stars found to harbor closely orbiting planets, HD 187123 ([Fe/H] = 0.16) appears to be of at least solar metallicity. 51 Peg–like companions have not yet been found around metal‐poor stars.