THE GEMINI NICI PLANET-FINDING CAMPAIGN: THE FREQUENCY OF GIANT PLANETS AROUND YOUNG B AND A STARS

Because planets and brown dwarfs cool and become fainter throughout their lifetime, determining the age of a substellar companion's host star is essential to determining the physical properties of the companion. Therefore, direct imaging surveys must be able to reliably age-date their target stars to understand both detected companions and overall survey sensitivities. Age-dating methods are typically most successful for solar-type (FGK) stars, where a variety of techniques are available, including lithium absorption, calcium emission, and gyrochronology (e.g., Mamajek 2009). For stars with higher mass the most robust method is to identify stars in young moving groups which share a common age and determine a consistent age for all the stars in the group (e.g., Zuckerman & Song 2004). Alternatively, if an A star is in a binary system with a solar-type star, age-dating the solar-type secondary can yield the age of the higher-mass primary. In cases where a high-mass star is single and does not belong to a kinematic group, however, we require an age-dating method that only utilizes the properties of the star itself.

Previous work has typically addressed the need to age-date B and A stars by turning to the CMD. In order to estimate the mass of the brown dwarf companion HR 7329 B, Jura et al. (1995) and Lowrance et al. (2000) placed its primary star HR 7329 A on an [M_V, B − V] CMD, along with A stars from nearby clusters and in the field (see Figure 3 of Lowrance et al. 2000). They noted that young (≲100 Myr) clusters such as the Pleiades, Alpha Per, and IC 2391 have A stars near the bottom of this CMD (i.e., at fainter absolute V magnitudes) compared to brighter A stars of the older (∼600 Myr) Hyades and Praesepe clusters. They also noted that the famous young A stars β Pic, HD 141569, and HR 4796 all lie at the very bottom of this diagram along with HR 7329 A, thus arguing for a young (≲100 Myr) age for the HR 7329 system. Moór et al. (2006) and Rhee et al. (2007) expanded this method to other high-mass stars with debris disks, noting that many of these stars also lay near the bottom of the CMD. Vigan et al. (2012) use a similar approach by defining the dereddened single-star sequence of high-mass stars in the Pleiades on the same [M_V, B − V] CMD and then assigning the age of the Pleiades (125 Myr; Stauffer et al. 1998) to stars with similar color–magnitude positions as the Pleiades A stars. However, this common approach to flagging young A stars is likely too optimistic in many cases, due to two effects: (1) the degeneracy between the effects of age and metallicity, and (2) the increase of main-sequence lifetime with decreasing stellar mass.

In order to demonstrate the limitations of this approach, we construct a volume-limited sample of Hipparcos stars within 100 pc and with parallax uncertainties below 5%, spectral types earlier than F0, luminosity class IV or V, −0.2 < B − V < 0.4, and purged of known binaries, resulting in 776 stars. All stellar data are taken from the extended Hipparcos compilation of Anderson & Francis (2012). Figure 1 illustrates the difficulty in using CMD position as a youth indicator for this volume-limited sample of B and A stars. The Siess et al. (2000) isochrones show that the "low CMD" stars (defined here as lying below the young cluster fit of Lowrance et al. 2000, with all other B and A stars referred to as "high CMD" stars) can be equally explained as young solar-metallicity stars or older sub-solar-metallicity stars. Therefore a low position on the CMD is not a unique indicator of youth, but rather suggests either youth or low metallicity.

Zoom In Zoom Out Reset image size

The Siess et al. (2000) isochrones themselves provide a good fit to B and A stars, as we demonstrate in Figure 2, where we plot the Siess et al. (2000) 120 Myr isochrones for sub-solar, solar, and supersolar metallicities against Pleiades stars from Stauffer et al. (2007), finding a good match between the solar track and the Pleiades single-star sequence. A more detailed analysis of Pleiades stars and main-sequence binaries by Bell et al. (2012) supports the validity of these models, showing that the Siess et al. (2000) isochrones are a good fit to both main-sequence and pre-main-sequence stars.

Zoom In Zoom Out Reset image size

Metallicities between 0 and −0.3 dex are not particularly rare for B and A stars as illustrated by Figure 3, which shows B and A stars from the Hipparcos sample with literature metallicity measurements (again from the Anderson & Francis 2012 compilation), divided into "low CMD" and "high CMD" samples. No appreciable difference is seen between the two samples of stars, nor between these B and A stars and the Casagrande et al. (2011) metallicity measurements of young (<1 Gyr) F and G dwarfs. The lack of a trend between CMD position and metallicity means that while "low CMD" stars are not uniformly young, neither are they uniformly low metallicity. They instead seem to have a similar metallicity distribution to the solar neighborhood. Therefore there are young solar or supersolar metallicity stars in the sample of "low CMD" stars, mixed in with lower-metallicity, older stars.

Zoom In Zoom Out Reset image size

In addition to metallicity, another important consideration is that most of a main-sequence star's evolution across the CMD occurs during the final two-thirds of a star's main-sequence lifetime (e.g., Soderblom 2010). The Siess et al. (2000) models indicate that this trend holds for stellar masses between 1.5 and 4.5 M_☉ (appropriate to our NICI sample), where stars show little change in B − V or V from the zero-age main sequence (ZAMS) to ∼1/3 of the main-sequence lifetime, and then a brightening in V over the remainder of the star's main-sequence lifetime. This effect can be seen in Figure 1, where at B − V > 0.1, the 30 Myr and 100 Myr isochrones for a given metallicity overlap, while all isochrones between 30 and 300 Myr overlap for B − V > 0.2. Even if we were to (incorrectly) assume that all stars have solar metallicity, while all Pleiades-age stars would lie low on the CMD, not all "low CMD" stars would be Pleiades age. In this uniform solar metallicity scenario, at best we would be constraining the ages of stars to be in the first third of their main-sequence lifetime, which is true for all Pleiades A stars. But the longer lifetimes for later-type A stars would make the default assigned age of 125 Myr increasingly inaccurate. For A9V stars, with an expected lifetime of 1.5 Gyr, "low CMD" stars would have ages up to ∼500 Myr, a factor of four older than the age of the Pleiades. Figure 4 demonstrates this by plotting the fraction of "low CMD" stars as a function of B − V from our Hipparcos volume-limited sample. For comparison, we also plot the expected fraction of stars younger than the Pleiades, which is the quotient of 125 Myr and the main-sequence lifetime for solar-metallicity stars from Siess et al. (2000), assuming a constant star formation rate in the solar neighborhood. While the fraction of "Pleiades-like" stars derived from the CMD appears to be relatively constant for all values of B − V, the expected fraction of Pleiades-aged stars drops by a factor of eight over this range. In other words, the "low CMD" stars are expected to have an increasingly large age spread at lower masses simply from stellar evolution. This effect is missed by the common procedure of assigning Pleiades ages to all "low CMD" stars.

Zoom In Zoom Out Reset image size

For a given location on a CMD, there are multiple combinations of age, mass, and metallicity that can reproduce the color and magnitude of a given star. We turn to Bayesian inference in order to determine the relative likelihoods of these combinations and to incorporate prior knowledge about the distributions of these stellar parameters. Such a technique has been previously used for solar-type stars by Takeda et al. (2007), who determined the ages of target stars for RV planet searches through a Bayesian analysis of their spectroscopically derived properties (T_eff, [Fe/H], and log(g)) combined with their V magnitude.

To determine the ages for B and A stars using photometry, we begin with Bayes' theorem:

$$\begin{equation} P({\rm model}|{\rm data}) \propto P({\rm data}|{\rm model}) P({\rm model}). \end{equation} \tag{ 1 }$$

That is, the probability of a model given a set of data (P(model|data)) is proportional to the product of the probability of the data given the model (P(data|model)) and the probability of the model itself (P(model)). From left to right, the three terms above are referred to as the posterior probability density function (PDF), the likelihood, and the prior.

Our model is the set of three stellar parameters, age (τ), mass (M_*), and metallicity ([Fe/H]), and our data are the measured absolute magnitude M(V) and B − V color for a given star. At any combination of age, mass, and metallicity Siess et al. (2000) predict the M(V) and B − V. It is then straightforward to compute the chi-squared statistic for this model:

$$\begin{eqnarray} && \chi ^2(\tau ,M_*,[{\rm Fe/H}]) = \sum \frac{(O - E)^2}{\sigma ^2} \nonumber\\ && = \frac{((B-V)_O - (B-V)_E)^2}{\sigma _{B-V}^2} + \frac{(M(V)_O - M(V)_E)^2}{\sigma _{M(V)}^2}, \quad \end{eqnarray} \tag{ 2 }$$

where (B − V)_O and M(V)_O are the observed (O) color and absolute magnitude, with observational errors σ_{B − V} and σ_M(V), respectively. (B − V)_E and M(V)_E are the expected (E) color and magnitude, predicted by Siess et al. (2000) for a given age, mass, and metallicity. The relative likelihood of a specific combination of three stellar parameters is then

$$\begin{eqnarray} P({\rm data}|{\rm model}) &=& P(B-V,M(V)|\tau ,M_*,[{\rm Fe}/{\rm H}]) \nonumber\\ & \propto & e^{- \frac{1}{2} \chi ^2(\tau ,M_*,[{\rm Fe}/{\rm H}])}. \end{eqnarray} \tag{ 3 }$$

This formulation assumes Gaussian statistics, which is a reasonable assumption given that our errors are from photometry and parallax measurements.

Siess et al. (2000) provide pre-main-sequence and main-sequence evolutionary tracks, which we interpolate to a regular three-dimensional grid in age, mass, and metallicity. Age is logarithmically gridded between 1 Myr and 10 Gyr, with the predicted photometry becoming undefined when the star leaves the main sequence. Mass is uniformly gridded between 1 and 5 M_☉. Metallicity is uniformly gridded between −0.3 < [Fe/H] < +0.3, since a linear grid in [Fe/H] corresponds to a logarithmic grid in absolute metal abundance.

We adopt priors to incorporate knowledge of the solar neighborhood when deriving probability distributions of stellar parameters.

• 1.  
  For metallicity, we turn to the Casagrande et al. (2011) metallicities for a magnitude-limited sample (V < 8.3 mag) of 16,000 F and G dwarfs in the solar neighborhood. In particular, we use their Figure 16 to provide a metallicity distribution for stars younger than 1 Gyr, an appropriate age range for field B and A stars. We model the Casagrande et al. (2011) metallicity histogram with a normal distribution with mean of −0.05 and σ of 0.11 dex and use this as our metallicity prior. We note that the Casagrande et al. (2011) metallicity distribution of young F and G stars in the solar neighborhood is consistent with the Nieva & Przybilla (2012) metallicity results for 20 nearby (<500 pc) early-B stars ([Fe/H] = 0.0 ± 0.1 dex). This good match to the young F and G dwarf measurements occurs despite the fact that the lifetimes of early-B stars are over an order of magnitude younger than 1 Gyr. Further support for the near solar metallicity of young stars is given by Biazzo et al. (2012), who find [Fe/H] = 0.10 ± 0.03 dex for five solar-type stars in the AB Dor moving group (100 Myr). So all the evidence to date suggests that solar neighborhood B and A stars have a metallicity distribution similar to young F and G stars.
• 2.  
  For our age prior, since the star formation rate is basically constant in the solar neighborhood over at least the last 2 Gyr (Cignoni et al. 2006), we adopt a prior that is uniform in linear age for B and A stars. Our gridding of the Siess et al. (2000) models is logarithmic in age, and so we weight each grid point such that equal linear age intervals have equal probability. The relative probability of age bin i is given by δτ_i/T, where δτ_i is the age range encompassed by the i-th bin, and T is the total age range for all bins. This prior has the effect of pushing the posterior PDF toward older ages, since older logarithmic age bins encompass a greater span of time than younger ones.
• 3.  
  Finally, we apply a Salpeter initial mass function (IMF) of the form dN/dM ∝ M^−2.35 (Salpeter 1955) as our mass prior. Each mass bin is weighted by this IMF, favoring lower masses over higher ones. We note that the effect of this prior is minimal, since each star has a narrow, well-defined mass posterior PDF. As a check, we computed the shift in the median of the mass and age posterior PDFs for all our target stars with and without the IMF prior. When the IMF prior is applied, the median shift is 0.2% toward lower masses and 1.4% to older ages. The standard deviation of the shift is 0.15% in mass and 1.4% in age.

Our final age probability distribution for each star is then given by marginalizing over all metallicities and masses. This is done after determining all likelihoods for each model (given by Equations (2) and (3)), weighted by the metallicity and age priors (Equation (1)). Figure 5 shows an example of our Bayesian age analysis for the A0V star HD 24966. Since the main-sequence lifetime for a 2.1 M_☉ star is 600 Myr, a median age for HD 24966 of 128 Myr (and 68% confidence level between 51–227 Myr) suggests that this star is in the first third of its main-sequence lifetime.

Zoom In Zoom Out Reset image size

Figure 6 shows the combined [Fe/H] posterior PDFs marginalized over age and mass from all 70 target stars compared to our adopted Casagrande et al. (2011) prior. The posterior and prior closely match each other for the "low CMD" stars, while there is a small 0.05 dex offset toward higher metallicities for the "high CMD" stars. This is as expected, since "high CMD" stars are best fit by higher-metallicity isochrones.

Zoom In Zoom Out Reset image size

In addition to age, mass, and metallicity, we can also produce a posterior PDF of main-sequence lifetime for each star from the same likelihood and priors that produce posterior PDFs for age, mass, and metallicity (though main-sequence lifetime is a function only of mass and metallicity and independent of age). We express the combined age PDFs for all stars in Figure 7 as a fraction of their main-sequence lifetime, using the lifetime PDF for each star. Despite allowing metallicity to vary (though following the Casagrande et al. 2011 prior), our Bayesian technique finds "low CMD" stars to be systematically younger than "high CMD" stars as expected.

Zoom In Zoom Out Reset image size

In Figure 8, we display the same combined age PDF as a function of absolute age. The most common ages for our target stars are between 50 and 500 Myr, with low-probability tails extending below 10 Myr and above 1 Gyr. The median age, 68% confidence limit, and 95% confidence limit for each of our 70 target stars are given in Table 1. We also show the properties of our target stars in Figure 9. About a fifth (14 out of 70) of our stars have independent age measurements from moving group membership, as indicated in Table 1, which we adopt as the final ages for these stars. Also, we adopt the age of 5 Myr for HD 141569 from Weinberger et al. (2000), based on their study of its two M-star companions. These independent ages are often significantly younger than the ages derived by our Bayesian analysis, which at best can only place a star within the first third of its main-sequence lifetime.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Stars with debris disks are generally well studied and have literature age measurements that are widely cited (e.g., Moór et al. 2006; Rhee et al. 2007). Nevertheless, for most of the debris disk stars in our B and A star sample, we do not use these literature ages and instead use our Bayesian ages. Of the 28 stars in our sample with debris disks, 10 belong to well-known moving groups, and we adopt that moving group age for these stars. Also, we use the 5 Myr age for HD 141569 from Weinberger et al. (2000). Of the remaining 17 stars, 3 (HIP 85340, γ Oph, and Fomalhaut) have ages taken from isochrones, while 14 (HD 10939, HD 17848, HD 24966, HD 31295, HD 32297, HD 54341, HD 71155, HD 85672, HD 110411, HD 131835, HD 138965, HD 176638, HD 182681, and HD 196544) are flagged as young for their low position on the CMD. For these our Bayesian analysis provides more accurate representations of the ages than isochrones or CMD position alone, so we override the literature ages for these 17 stars with our Bayesian ages.

Figure 10 compares our ages to stars that we have in common with Su et al. (2006). For the ages derived by our Bayesian analysis, we find good agreement for stars which Su et al. (2006) determine to be older than 100 Myr. Stars that Su et al. (2006) find to be younger than 100 Myr show a systematic disagreement, with the median of our Bayesian age distribution significantly older than the single age found by Su et al. (2006). This is as we would expect, since most movement across the CMD happens in the final two-thirds of a star's main-sequence lifetime, so older stars can be age-dated more definitively while younger stars can be anywhere in the first third of their main-sequence lifetimes. Our Bayesian method takes this uncertainty into account, rather than assigning very young ages (≲100 Myr) for "low CMD" stars. Also, stars that belong to a well-known moving group (as well as HD 141569) have adopted ages that are uniformly younger than the Bayesian distributions. This is not surprising, as nearby moving groups are significantly younger than the field population.

Zoom In Zoom Out Reset image size

Finally, we consider the ages derived by Vigan et al. (2012) for their sample of young A stars. We apply our Bayesian analysis to their sample and compare the histogram of our derived ages to theirs in Figure 11. Out of the 39 stars from their sample, 11 belong to moving groups, and we adopt the same moving group ages for these stars. For the remaining 28 stars, we derive ages from our Bayesian analysis and populate each bin of our histogram with the appropriate contribution from the posterior age PDF. As expected, the Bayesian approach results in a significant tail at large ages (>200 Myr). By contrast, almost half of the Vigan et al. (2012) ages (17/39) are incorrectly assumed to be 125 Myr, the age of the Pleiades. For stars not in young moving groups, we compute the offset from the Vigan et al. (2012) ages to the median age from our Bayesian analysis. The median of this offset for these 28 stars is 177 Myr, an increase in age of 67%, which significantly increases the minimum detectable mass around these stars. The typical contrast achieved at 2'' in Vigan et al. (2012) is 14.2 mag, which for 27 of these 28 stars corresponds to planetary masses using their ages and the Baraffe et al. (2003) evolutionary models for companion luminosity. When we instead adopt the median of our Bayesian age distributions for these stars, only 15 stars reach planetary masses at 2'', and the median increase in detectable mass at 2'' for these 15 stars is 13%. For only one star, HIP 41307, is our median age of 249 Myr younger than the age assumed by Vigan et al. (2012), 400 Myr. For the remaining 14 stars, the magnitude of this increase in detectable mass ranges from 0 to 6 M_Jup, with a median of 1.6 M_Jup. Therefore, if our older ages were to be used, the constraints on exoplanets from the Vigan et al. (2012) dataset would become weaker.

Zoom In Zoom Out Reset image size

In Figure 12 we examine how this age uncertainty for "low CMD" stars depends on B − V color. We divide "low CMD" stars in our volume-limited sample of Hipparcos stars into nine color bins and combine the marginalized PDFs for age for the stars in that bin. As expected, the median age in each color bin increases with redder color. Assuming a Pleiades age for "low CMD" stars is a fair representation of stars bluer than B − V of −0.05, but for redder stars such an assumption significantly underestimates the age.

Zoom In Zoom Out Reset image size

To summarize the results of our age analysis, the standard practice of assigning an age of 125 Myr to stars in the Pleiades locus on the CMD produces systematically younger ages than we find with our Bayesian method. Figure 8 shows that these Pleiades-like stars in the NICI Campaign sample have a median age of 200 Myr with a 68% confidence level between 160 and 840 Myr. This is expected since identifying a star as similar to a Pleiades A star only places it within the first third of its main-sequence lifetime, which for late-type A stars can imply ages much larger than 125 Myr.

Given NICI's ability to achieve very high contrasts (median of ∼15 mag—a flux ratio of 10⁶—at 1'') and its 18'' × 18'' field of view, it is expected that our images will contain a large number of background objects in addition to any common proper motion (CPM) companions. The Campaign data were reduced and checked for candidate companions by both an automated procedure and visual inspection of the images as described in Wahhaj et al. (2013a). Stars with candidate companions were flagged, and archival high contrast imaging data (from the Very Large Telescope (VLT), Hubble Space Telescope (HST), Gemini-North, or Keck) were retrieved and analyzed to see if the candidates could be recovered. If archival data were insufficient to determine whether the candidates were background or CPM, an additional NICI observation was typically obtained to measure the astrometry of the companion(s) at a second epoch. In a few cases multiple additional epochs were required to definitively classify each candidate. We followed up all candidates within 400 AU except for most stars with large numbers (≳10) of candidates due to these stars being in the foreground of the galactic bulge or at galactic latitude close to 0°.

A candidate companion observed at multiple epochs with sufficient astrometric precision will be either in the same location as in the first epoch (CPM, with the possibility of a small amount of orbital motion) or displaced from the first epoch position, following the expected motion of a background object. A distant (≳500 pc) background object will have negligible proper motion (≲2 mas yr⁻¹) and thus appear to move with respect to the target star in the opposite direction indicated by the target star's (much larger) proper and parallactic motion. To quantitatively classify stars as background (bg) or CPM, we compute the chi-square statistic for the two scenarios:

$$\begin{equation} \chi ^2_{{\rm bg}} = \sum _{i} \left(\frac{(\rho _{{\rm obs},i} - \rho _{{\rm bg},i})^2}{\sigma _{\rho ,i}^2} + \frac{({\rm P.A.}_{{\rm obs},i} - {\rm P.A.}_{{\rm bg},i})^2}{\sigma _{{\rm P.A.},i}^2} \right) \end{equation} \tag{ 4 }$$

$$\begin{equation} \chi ^2_{{\rm CPM}} = \sum _{i} \left(\frac{(\rho _{{\rm obs},i} - \rho _{0})^2}{\sigma _{\rho ,i}^2} + \frac{({\rm P.A.}_{{\rm obs},i} - {\rm P.A.}_{0})^2}{\sigma _{{\rm P.A.},i}^2} \right), \end{equation} \tag{ 5 }$$

where the summation is over all epochs except the reference epoch; ρ_{obs, i} and P.A._{obs, i} are the observed separation and position angle at the i-th epoch; ρ_{bg, i} and P.A._{bg, i} are the separation and position angle predicted for a background object given the position at the reference epoch; ρ₀ and P.A.₀ are the separation and position angle at the reference epoch; and σ_{ρ, i} and σ_{P.A., i} are the uncertainties in separation and position angle at the i-th epoch. We convert chi-square to reduced chi-square ($\chi ^2_\nu$) using the number of degrees of freedom (dof), which is (N_epochs − 1) × 2, where N_epochs is the total number of epochs. The number of epochs in the dof calculation is reduced by one, since the background and CPM predictions are tied to the reference epoch, and then multiplied by 2, because there is a separation and P.A. measurement at each epoch. The result for each of our candidates is unambiguous, with $\chi ^2_\nu$ for background close to one and $\chi ^2_\nu$ for CPM much larger than one, or vice versa.

Table 3 lists the properties of each candidate companion for which we have more than one epoch. For each of these companions we then give the astrometry at each epoch in Table 4. Errors in the background track are computed using a Monte Carlo method, combining astrometric errors at the reference epoch for the candidate and errors in the proper motion and parallax of the target star. All proper motion and parallax measurements for our targets are from van Leeuwen (2007). We typically choose the reference epoch to be the first NICI epoch for that candidate. We also show the motion of each companion on the sky with respect to the background track in Figures 13–17.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 3. Properties of Candidate Companions

Name	Number	Sep	P.A.	ΔH	Δτ	Epochs	$\chi ^2_{\nu }$(BG)	$\chi ^2_{\nu }$(CPM)	dof	Comp?
		('')	(deg)	(mag)	(yr)
HD 1160	1	0.780	244.3	9.6	9.28	12	11.22	0.47	22	CPMa
HD 1160	2	5.150	349.8	7.5	9.28	19	2.47	0.49	36	CPMa
HIP 14146	1	3.936	296.6	17.3	1.78	3	0.31	57.65	4	BG
HD 31295	1	6.261	271.6	13.3	1.95	2	0.09	69.71	2	BG
HD 31295	2	8.805	316.7	14.5	1.95	2	0.02	187.48	2	BG
HIP 25280	1	5.219	80.4	14.9	1.84	2	1.82	25.00	2	BG
ζ Lep	1	5.295	110.0	18.0	2.24	2	0.44	28.80	2	BG
HD 54341	1	3.080	131.2	12.7	2.03	2	0.40	3.05	2	BG
HD 54341	2	4.372	137.3	11.8	2.03	2	0.91	2.56	2	BG
HIP 36188	1	7.460	205.0	16.1	2.66	2	15.38	182.97	2	BG
HD 71155	1	5.766	86.7	14.9	2.41	3	0.84	65.12	4	BG
HD 71155	2	7.360	357.8	16.8	2.41	2	0.02	19.78	2	BG
HD 71155	3	9.403	246.1	14.2	2.41	2	1.12	95.44	2	BG
HIP 41373	1	3.043	260.9	12.0	2.28	3	2.00	21.35	4	BG
HIP 41373	2	5.916	0.9	11.7	2.28	3	0.39	2.39	4	BG
HIP 43121	1	4.760	350.0	14.9	2.10	2	0.11	39.74	2	BG
HIP 50191	1	5.192	280.4	9.9	1.05	2	0.20	79.56	2	BG
HIP 50191	2	5.144	285.6	13.3	1.05	2	0.43	74.46	2	BG
HIP 57328	1	1.599	268.1	15.4	2.06	2	1.98	88.06	2	BG
HR 4796	1	4.472	322.7	8.6	18.07	3	0.73	57.48	4	BG
HD 110058	1	3.030	252.5	12.9	3.19	2	0.37	50.83	2	BG
HD 110058	2	3.280	94.3	13.2	3.19	2	0.94	24.62	2	BG
HD 110058	3	6.210	160.3	8.7	3.19	2	0.64	14.51	2	BG
HD 110058	4	7.810	231.1	8.2	3.19	2	9.77	108.96	2	BG
HD 110058	5	7.830	125.8	12.2	3.19	2	0.31	15.33	2	BG
HD 110058	6	8.400	142.6	11.3	3.19	2	0.56	10.39	2	BG
HD 118878	1	3.316	181.9	10.7	3.16	2	0.52	21.90	2	BG
HD 118878	2	6.203	213.7	14.2	3.16	2	0.12	35.13	2	BG
HD 118878	3	7.254	120.3	11.1	3.16	2	0.07	9.30	2	BG
HD 118878	4	9.668	217.0	12.5	3.16	2	0.94	52.59	2	BG
HD 136482	1	0.940	241.5	11.4	1.08	2	0.42	2.78	2	BG
HD 136482	2	2.960	156.7	11.3	1.08	2	0.58	2.78	2	BG
HD 136482	3	3.772	181.8	14.1	1.08	2	0.28	3.88	2	BG
HD 136482	4	5.081	146.0	13.5	1.08	2	0.09	1.29	2	BG
HD 136482	5	5.480	251.8	8.9	1.08	2	2.79	13.70	2	BG
HD 136482	6	6.334	51.9	12.3	1.08	2	0.85	9.09	2	BG
HD 138965	1	1.708	61.4	10.6	2.20	2	0.79	75.76	2	BG
HD 138965	2	4.085	244.6	12.2	2.20	2	0.19	81.31	2	BG
HD 138965	3	4.284	269.6	14.6	2.20	2	0.43	35.78	2	BG
HD 138965	4	7.386	157.3	11.4	2.20	2	0.37	35.91	2	BG
HD 138965	5	7.503	102.1	12.8	2.20	2	1.06	36.51	2	BG
HD 138965	6	7.838	236.4	11.4	2.20	2	0.02	75.49	2	BG
HD 138965	7	8.048	326.9	11.3	2.20	2	0.16	17.66	2	BG
HD 138965	8	9.970	4.4	13.6	2.20	2	1.13	41.08	2	BG
HD 141569	1	6.400	13.2	14.6	2.16	2	1.50	5.12	2	BG
HD 141569	2	7.850	209.7	12.8	2.16	2	1.80	24.48	2	BG
HIP 79781	1	4.509	179.2	11.9	1.50	2	0.46	3.78	2	BG
HIP 79781	2	6.525	130.2	8.9	1.50	2	0.03	5.59	2	BG
HIP 79797	1	5.737	276.5	14.4	1.91	3	0.04	16.31	4	BG
HIP 79797	2	5.759	18.8	13.4	1.91	3	0.64	44.62	4	BG
HIP 79797	3	6.213	118.0	11.6	1.91	3	1.64	12.56	4	BG
HIP 79797	4	6.691	293.9	7.4	7.77	5	54.54	0.81	8	CPMb
HIP 79797	5	6.949	197.2	14.9	1.91	3	1.65	88.02	4	BG
HIP 79797	6	6.907	329.4	13.6	0.90	2	0.28	4.92	2	BG
HIP 79797	7	7.737	250.7	12.5	0.90	2	0.07	8.35	2	BG
HIP 79797	8	9.598	6.5	10.7	1.01	2	0.22	29.66	2	BG
HIP 79881	1	1.627	267.7	12.5	0.98	2	0.96	56.83	2	BG
HIP 79881	2	4.551	175.6	10.0	0.98	2	0.09	26.23	2	BG
HIP 79881	3	6.600	233.2	12.6	0.98	2	0.20	16.98	2	BG
HIP 85922	1	7.359	40.3	13.2	1.59	2	4.44	50.95	2	BG
γ Oph	1	4.528	59.2	14.6	1.07	2	0.04	14.21	2	BG
γ Oph	2	5.902	239.4	16.0	1.07	2	0.13	15.85	2	BG
γ Oph	3	6.210	267.3	12.6	1.07	2	0.15	6.82	2	BG
γ Oph	4	6.602	275.0	15.0	1.07	2	0.16	4.42	2	BG
γ Oph	5	6.964	59.1	13.6	1.07	2	0.11	14.71	2	BG
γ Oph	6	8.677	96.3	12.3	1.07	2	1.71	11.78	2	BG
γ Oph	7	9.271	253.6	15.5	1.07	2	0.09	7.01	2	BG
HD 172555	1	7.702	319.0	14.5	3.92	2	0.04	11.29	2	BG
HD 176638	1	3.529	157.8	14.1	0.90	3	0.64	8.34	4	BG
HD 176638	2	4.743	93.7	12.6	0.90	3	1.18	12.96	4	BG
HD 176638	3	5.222	265.6	14.5	0.47	2	0.24	13.04	2	BG
HD 176638	4	9.390	319.7	10.6	0.90	3	0.92	3.33	4	BG
HR 7329	1	4.175	167.6	6.6	10.78	2	133.91	3.79	2	CPMc
HD 196544	1	4.743	90.8	13.1	2.63	5	0.34	12.06	8	BG
HD 196544	2	4.352	93.4	14.5	2.63	5	0.19	9.60	8	BG
HIP 104308	1	1.681	281.9	12.4	0.75	2	0.69	27.31	2	BG
HIP 104308	2	2.895	264.0	15.5	0.75	2	0.05	14.08	2	BG
HIP 107556	1	6.179	169.5	17.7	1.09	3	0.15	368.24	4	BG

Notes. Summary of the candidate companions detected around each target star. For each candidate we list the separation and position angle at the reference epoch, the contrast between candidate and host star, the time baseline for our astrometric data, the number of epochs, the reduced chi-square statistic for the candidate to be background (BG) or common proper motion (CPM), the number of degrees of freedom for the astrometric fit, and the final determination of each candidate: background or common proper motion. Astrometry for individual epochs is in Table 4. ^aNielsen et al. (2012). ^bOriginally identified by Huélamo et al. (2010). ^cOriginally identified by Lowrance et al. (2000).

Download table as:  ASCIITypeset images: 1 2

Of the candidate companions imaged around our target stars, there are five that are CPM companions, found around HD 1160, HR 7329, and HIP 79797. HD 1160 B is a brown dwarf with mass 33$^{+12}_{-9}$ M_Jup, and HD 1160 C is an M3.5 star with mass 0.22$^{+0.03}_{-0.04}$ M_☉ (Nielsen et al. 2012). We also detect the known substellar companion to HR 7329 B (Lowrance et al. 2000), which is a brown dwarf with mass between 20 and 50 M_Jup (Neuhäuser et al. 2011). The companion to HIP 79797 (also known as HR 6037) was independently discovered as a single object by Huélamo et al. (2010), who determined a mass of 62 ± 20 M_Jup for the 67 companion.

NICI observations of HIP 79797 not only further confirm that HIP 79797 B is a CPM companion but reveal that the companion is in fact a binary, with a separation of 006 and a near-unity flux ratio at JHK_S. We have retrieved the archival 2004 and 2006 VLT data used by Huélamo et al. (2010) but are unable to split the binary in these images, given the low signal-to-noise ratio (S/N) of the data. In five NICI epochs (2010–2012), however, we are able to resolve the binary (Figure 18) and determine relative astrometry for the two components (Table 5). We summarize the properties of the HIP 79797 system in Table 6.

Zoom In Zoom Out Reset image size

Table 6. Properties of the HIP 79797 System

	HIP 79797 A	HIP 79797 Ba	HIP 79797 Bb
Age (Myr)	200$^{+170}_{-130}$
Distance (pc)	52.2 ± 1.1a
Spectral type	A4	M9b	M9b
ΔJ (Bab − A, mag)		8.55 ± 0.14
ΔH (Bab − A, mag)		7.37 ± 0.17
ΔKS (Bab − A, mag)		7.31 ± 0.19
ΔJ (Bb − Ba, mag)		0.33 ± 0.07
ΔH (Bb − Ba, mag)		0.12 ± 0.06
ΔKS (Bb − Ba, mag)		0.20 ± 0.06
J (mag)	5.77 ± 0.03c	14.92 ± 0.19	15.26 ± 0.19
H (mag)	5.68 ± 0.05c	13.74 ± 0.15	13.87 ± 0.15
Ks (mag)	5.66 ± 0.02c	13.62 ± 0.17	13.81 ± 0.17
MJ (mag)	2.18 ± 0.06	11.3 ± 0.2	11.7 ± 0.2
MH (mag)	2.10 ± 0.07	10.16 ± 0.16	10.28 ± 0.16
MKs (mag)	2.07 ± 0.05	10.03 ± 0.17	10.23 ± 0.17
Mass	1.76 ± 0.4 M☉	58$^{+21}_{-20}$ MJup	55$^{+20}_{-19}$ MJup
Mass ratio (Bb/Ba)		0.93 ± 0.03
Separation (from A, arcsec)d		6.691 ± 0.009
Position angle (from A, deg)d		293.9 ± 0.2

Notes. ^avan Leeuwen (2007). ^bHuélamo et al. (2010), and assuming that the integrated light spectrum is similar to the spectra of the individual objects, which is consistent with the near-unity flux ratios. ^cCutri et al. (2003). ^dAstrometry is from epoch 2010.3534 and is the separation and position angle of the photocenter of the binary from HIP 79797 A.

Download table as:  ASCIITypeset image

We estimate the masses of the now-resolved brown dwarf binary companions using our measurement of the difference in the K_S magnitudes of the combined two components with respect to the primary (7.31 ± 0.17 mag), the K_S-band flux ratio of the two components (0.20 ± 0.06 mag), and our Bayesian age distribution for HIP 79797 (median 203 Myr, 68% confidence interval between 71 and 372 Myr). We also assume that the spectral type of M9 measured by Huélamo et al. (2010) applies to both objects (which is reasonable given the nearly identical colors of the two components) and use a K-band bolometric correction from Golimowski et al. (2004) to determine the luminosity of each component, which we then convert to mass using the Lyon/DUSTY evolutionary models (Chabrier et al. 2000). We determine uncertainties through a Monte Carlo method, assuming Gaussian photometric and astrometric errors and using our posterior PDF for age. We find similar masses for the two components (as expected given their small flux ratio) of 58$^{+21}_{-20}$ M_Jup and 55$^{+20}_{-19}$ M_Jup, with a mass ratio of 0.93 ± 0.03. These are similar values to the mass estimated by Huélamo et al. (2010) based on their K_S-band photometry assuming that the system was only a single object, meaning our mass estimates are discrepant with theirs. While our median age for the system of 203 Myr is slightly younger than their age of 300 Myr, we also find a brighter K_S-band flux for the combined Bab system of 13.0 ± 0.17 mag, compared to the 13.9 ± 0.1 mag and 14.4 ± 0.2 mag measured by Huélamo et al. (2010) in their two VLT/NACO (NAOS-CONICA) datasets. We note that the NACO data were taken with a neutral density filter to avoid saturation, and so HIP 79797 Ba/Bb is detected at low S/N in these images.

To assess the orbital properties, we have performed a preliminary Markov Chain Monte Carlo (MCMC) analysis on our 2 yr of astrometric data for HIP 79797 Ba and Bb. Any orbital motion over our short time baseline is less than half a NICI pixel, so we do not attempt to measure a dynamical mass for HIP 79797 Ba and Bb. Instead, we fix the combined mass of the brown dwarfs to 113 M_Jup and attempt to constrain the possible orbital parameters. Our MCMC fit favors an edge-on orbit (i = 87° ± 5°) with semi-major axis of 0.48$^{+1.87}_{-0.37}$'' (25$^{+97}_{-19}$ AU). This preference for inclined orbits is understandable given that the change in the separation from the first to the last epoch (17 mas) is only slightly larger than the mean error on a single epoch (12 mas), while the P.A. of the binary does not appear to change (within errors). That is, our data suggest, at low confidence, that the projected sky motion of HIP 79797 Bb is directly away from HIP 79797 Ba, as would be expected in an edge-on orbit. Similarly, the semi-major axis must be large enough for there to be such a small amount of motion over 2 yr. The corresponding orbital period is then 380$^{+3700}_{-330}$ yr. We stress that these are preliminary results, based on a short time baseline with very little on-sky motion, and so the 68% confidence interval for orbital periods spans almost two orders of magnitude. If the orbital period is on the shorter side of the estimated period range, future astrometric monitoring could plausibly yield a precise orbit and dynamical masses, as ∼30% coverage of an astrometric orbit is sufficient to yield these results (e.g., Liu et al. 2008; Dupuy & Liu 2011). In particular, a dynamical mass would be of immense value since the only other brown dwarf binary with a dynamical mass and an independent age estimate is the L4+L4 binary HD 130948 (Dupuy et al. 2009). Such benchmark brown dwarf binaries are key to testing models of brown dwarf atmospheres and evolution.

For 20 of our 70 B and A stars, there are candidate companions that only appear in one epoch of NICI data and have no detections in archival observations. These candidates fall into three categories. (1) The most common case was that follow-up of these companions was given low priority, as we focused our observing time on candidates with the smallest projected physical separations (<400 AU). (2) In the second case follow-up was attempted but the second epoch observation was unable to recover the candidate, because the candidate was especially faint or close to the target star. In some cases with multiple candidates around a single star, we were able to confirm the nature of some candidates but not others. (3) Finally, since the target star is not centered on the NICI detector, the orientation of the detector on the sky depends on the parallactic angle of the observation. For wide-separation candidates (≳63) this can mean a candidate is on the detector in one epoch and off the detector in another. While we typically tried to schedule our second epoch observations to detect these wide candidates, it was not always possible to do this so some remain unvalidated.

For the candidates around these 20 stars, we are unable to confirm whether these are background sources or co-moving companions. Given that only 4 of the 78 candidate companions in Table 3 are not background objects, it is likely that few, if any, of these single-epoch candidates are co-moving companions, especially given that most of them are at wider separations. Nevertheless, we present the astrometry for these candidates in Table 7 as a reference for future observations of these stars.

Table 7. Candidate Companions with One Epoch of Data

Star	Number	Sep	Sep	P.A.	ΔH	Epoch	Action
		('')	(AU)	(deg)	(mag)
HD 31295	1	7.77	277	162.4	15.6	2009.0356	Sep < 777
	2	8.66	309	277.1	17.5	2009.0356
HIP 36188	1	10.64	528	352.7	14.6	2011.8384	Sep < 1064
HIP 40916	1	6.26	421	332.7	12.8	2009.0466	Sep < 626
	2	6.95	467	131.0	12.6	2009.0466
	3	7.65	514	60.9	7.4	2009.0466
	4	7.67	515	90.9	8.2	2009.0466
HD 71155	1	9.93	372	221.0	15.2	2008.9563	Sep < 993
HIP 50191	1	4.11	128	257.9	15.4	2010.0110	Revert to 2008.9563 ASDI
HD 118878	1	9.21	1114	330.7	11.6	2009.1041	Sep < 881
	2	12.66	1532	69.2	14.9	2009.1041
	3	12.84	1553	39.5	14.9	2009.1041
	4	12.99	1571	145.1	13.9	2009.1041
	5	8.81	1066	299.4	10.7	2012.2650
	6	10.88	1317	175.8	15.0	2012.2650
GJ 560 A	1	5.71	95	230.7	15.9	2009.1891	Sep < 588
	2	5.88	97	185.6	15.8	2009.1891
	3	6.15	102	27.5	16.3	2009.1891
	4	6.30	104	52.6	16.1	2009.1891
	5	6.86	114	269.0	14.6	2009.1891
	6	7.34	122	284.4	15.0	2009.1891
HD 131835	1	2.31	280	274.2	14.4	2009.2712	Drop
	2	5.27	638	20.2	12.3	2009.2712
	3	5.70	689	22.7	11.5	2009.2712
	4	6.13	741	303.2	11.4	2009.2712
	5	8.09	979	239.8	13.8	2009.2712
	6	8.29	1003	91.6	14.6	2009.2712
	7	8.93	1080	100.7	13.8	2009.2712
HD 135454	1	3.82	658	135.2	12.1	2009.1918	Drop
	2	4.96	854	205.7	14.4	2009.1918
	3	5.32	916	26.2	14.9	2009.1918
	4	6.01	1033	349.7	10.1	2009.1918
	5	6.83	1174	178.2	13.2	2009.1918
	6	7.30	1255	118.8	10.1	2009.1918
	7	8.21	1412	82.3	12.3	2009.1918
HD 136482	1	6.05	823	157.1	15.2	2009.1891	Sep < 605
	2	8.59	1168	80.1	15.5	2009.1891
	3	8.94	1215	280.1	13.6	2009.1891
	4	9.24	1257	181.1	15.1	2009.1891
	5	9.57	1302	77.7	10.3	2009.1891
	6	10.46	1422	17.4	13.9	2009.1891
HD 138965	1	8.34	651	97.6	13.4	2009.1151	Sep < 816
	2	12.12	946	111.1	13.9	2009.1151
	3	13.19	1029	90.8	10.6	2009.1151
	4	8.16	636	244.6	14.7	2011.3151
	5	10.92	852	199.7	12.4	2011.3151
HIP 77464	1	8.83	477	31.9	12.7	2010.5671	Sep < 883
HIP 79797	1	13.23	691	291.3	13.7	2011.3589	Sep < 13232
HIP 78106	1	3.31	202	280.9	12.4	2009.2739	Drop
	2	5.02	306	104.9	16.3	2009.2739
	3	6.37	389	47.9	16.4	2009.2739
	4	7.14	436	153.9	15.9	2009.2739
	5	7.58	462	114.8	14.2	2009.2739
	6	9.15	558	305.1	12.7	2009.2739
HIP 81650	1	1.44	71	87.1	13.0	2010.3534	Drop
	2	1.83	91	245.7	13.8	2010.3534
	3	2.10	104	253.1	12.5	2010.3534
	4	2.28	112	32.4	13.2	2010.3534
	5	2.83	140	120.2	14.1	2010.3534
	6	2.89	143	218.4	12.0	2010.3534
	7	2.96	146	355.1	11.9	2010.3534
	8	3.41	168	350.0	15.1	2010.3534
	9	3.60	178	64.6	14.1	2010.3534
	10	3.75	185	111.0	11.6	2010.3534
	11	3.92	194	190.3	9.6	2010.3534
	12	3.98	197	227.4	15.0	2010.3534
	13	4.23	209	292.8	14.9	2010.3534
	14	4.29	212	123.2	12.7	2010.3534
	15	4.31	213	219.6	12.3	2010.3534
	16	4.37	216	189.7	14.7	2010.3534
	17	4.44	219	284.8	14.8	2010.3534
	18	4.50	222	211.8	13.8	2010.3534
	19	4.54	224	207.6	13.9	2010.3534
	20	4.65	230	251.1	14.6	2010.3534
	21	5.00	247	5.6	12.5	2010.3534
	22	5.15	254	176.6	14.5	2010.3534
	23	5.49	271	25.0	12.9	2010.3534
	24	5.54	274	108.9	14.8	2010.3534
	25	5.59	276	333.3	12.5	2010.3534
	26	5.60	277	347.2	13.9	2010.3534
	27	5.74	284	133.7	13.6	2010.3534
	28	5.93	293	233.4	10.8	2010.3534
	29	6.21	307	30.4	13.9	2010.3534
	30	6.25	309	353.9	13.8	2010.3534
	31	6.33	313	344.2	14.6	2010.3534
	32	6.44	318	39.7	10.8	2010.3534
	33	6.78	335	13.6	15.1	2010.3534
	34	6.79	335	89.8	14.5	2010.3534
	35	7.10	351	285.4	12.6	2010.3534
	36	7.12	352	255.7	15.0	2010.3534
	37	7.14	353	59.1	14.3	2010.3534
	38	7.37	364	128.0	14.3	2010.3534
	39	7.37	364	94.6	15.5	2010.3534
	40	7.65	378	200.1	14.5	2010.3534
	41	7.80	386	222.8	14.6	2010.3534
	42	7.81	386	28.3	14.9	2010.3534
	43	8.07	398	295.3	14.3	2010.3534
HIP 85038	1	4.26	264	156.1	13.0	2010.3562	Drop
	2	4.30	267	121.2	12.1	2010.3562
	3	5.35	332	202.5	13.9	2010.3562
	4	5.45	338	321.3	13.2	2010.3562
	5	5.70	354	115.0	14.9	2010.3562
	6	5.83	361	216.5	8.5	2010.3562
	7	6.71	416	224.3	9.8	2010.3562
	8	7.91	491	23.5	13.6	2010.3562
γ Oph	1	9.24	291	241.7	14.5	2010.2657	Sep < 924
HD 176638	1	9.76	577	153.6	13.6	2011.7917	Sep < 976
HIP 93805	1	3.45	131	90.9	14.5	2010.2712	Drop
	2	3.89	147	169.9	14.9	2010.2712
	3	4.35	165	275.9	15.3	2010.2712
	4	4.84	183	46.4	14.5	2010.2712
	5	5.24	198	36.6	15.9	2010.2712
	6	6.01	228	80.5	13.1	2010.2712
	7	6.58	249	173.3	15.4	2010.2712
	8	7.24	275	63.2	16.2	2010.2712
HD 182681	1	4.49	314	250.1	11.2	2010.6521	Drop
	2	5.29	370	251.4	15.7	2010.6521
	3	8.59	600	305.1	13.6	2010.6521
	4	8.87	620	289.6	12.0	2010.6521
	5	9.54	667	303.7	14.9	2010.6521

Notes. Target stars with candidate companions for which we only have a single epoch of data. Since we cannot classify these as either CPM or background, we provide the properties of these candidates here and note the changes we make to the contrast curves in the Action column (see Section 5.5.1 for details). Contrast curves are edited by either reverting to a less sensitive contrast curve or restricting the contrast curve to within a given separation. For cases where we have only a single epoch and there are candidates inside the 100% coverage region for position angle, we drop the star from our analysis.

Download table as:  ASCIITypeset images: 1 2