submitted Sept 12, 2006 to ApJ; accepted December 20, 2006 The Late-T Dwarf Companion to the Exoplanet Host Star HD 3651: A New Benchmark for Gravity and Metallicity Effects in Ultracool Spectra Michael C. Liu,1,2,3 S. K. Leggett,4 Kuenley Chiu5 ABSTRACT 7 0 0 We present near-IR (1.0–2.5 µm) photometry and spectroscopy of HD 3651B, the 2 low-luminosity, wide-separation (480 AU) companion to the K0V exoplanet host star n a HD 3651A. We measure a spectral type of T7.5±0.5 for HD 3651B, confirming both its J substellar nature and the fact that wide-separation brown dwarfs and giant planets can 4 co-exist around the same star. We estimate an age of 3–12 Gyr for the primary star 1 HD3651Aandfindthatitis≈3×olderthantheK4VstarGl570A(≈1–5Gyr),thehost v 1 star of the T7.5 dwarf Gl 570D. We derive a bolometric luminosity of log(Lbol/L⊙) = 1 −5.58±0.05 for HD 3651B and infer an effective temperature of 780–840 K and a mass 1 of 40–72 M ; the luminosity and temperature are among the lowest measured for any 1 Jup 0 browndwarf. Furthermore,HD 3651Bbelongstotherareclassofsubstellarobjectsthat 7 arecompanionstomain-sequencestarsandthusprovidesanewbenchmarkforstudying 0 / very low-temperature objects. Given their similar temperatures (∆T ≈ 30 K) and h eff p metallicities (∆[Fe/H] ≈ 0.1 dex) but different ages, a comparison of HD 3651B and - o Gl 570D allows ustoexamine gravity-sensitive diagnostics inultracool spectra. We find r t that the expected signature of HD 3651B’s higher surface gravity due to its older age, s a namely a suppressed K-band flux relative to Gl 570D, is not seen. Instead, the K-band : v flux of HD 3651B is enhanced compared to Gl 570D, indicative of a younger age. Thus, i X the relative ages derived from interpretation of the T dwarf spectra and from stellar r activity indicators appear to be in discord. One likely explanation is that the K-band a fluxes are also very sensitive to metallicity differences. Metallicity variations may be as important as surface gravity variations in causing spectral differences among field late-T dwarfs. Subject headings: stars: brown dwarfs — stars: atmospheres — infrared: stars 1Institutefor Astronomy,University of Hawai‘i, 2680 Woodlawn Drive, Honolulu, HI 96822; [email protected] 2Alfred P. Sloan Research Fellow 3Visiting Astronomer at the Infrared Telescope Facility, which is operated by the University of Hawaii under Cooperative Agreement no. NCC 5-538 with the National Aeronautics and Space Administration, Science Mission Directorate, Planetary Astronomy Program. 4Gemini Observatory Northern Operations Center, 670 North A’ohoku Place, Hilo, HI 96720 5Astrophysics Group, School of Physics, Universityof Exeter, StockerRoad, Exeter, EX4 4QL, UK – 2 – 1. Introduction Discovery and scrutiny of brown dwarfs have been fertile avenues for understanding self- luminous objects, extending traditional stellar astrophysics into new domains of mass and effective temperature. The coolest brown dwarfs, the T dwarfs, are characterized by very red optical colors from pressure-broadened alkali resonance lines and very blue near-IR colors from strong methane absorption and collision-induced H absorption (e.g. Oppenheimer et al. 1995; Geballe et al. 2002; 2 Kirkpatrick 2005). T dwarfs are the lowest luminosity and coolest objects directly detected outside of our solar system, with bolometric luminosities (Lbol) of . 10−4.5L⊙ and effective temperatures (T ) of ≈ 700−1300 K (e.g. Vrba et al. 2004; Golimowski et al. 2004; Burgasser et al. 2006a). As eff such, analyzing their spectra to infer temperature, gravity, metallicity, and mass is a key pathway to understanding the properties of gas-giant extrasolar planets. The first T dwarf, Gl 229B, was discovered as a 45-AU companion around a nearby M dwarf (Nakajima et al. 1995). However, since then the vast majority have been found as free-floating field objects by the 2MASS and SDSS wide-field surveys (e.g. Burgasser et al. 1999; Leggett et al. 2000; Chiu et al. 2006), with about 100 T dwarfs identified to date. Such a populous sample has been a boon for probing substellar astrophysics. But this sample is also inevitably hindered by the unknown ages and metallicities of the field population. Binary brown dwarfs provide a partial solution to this challenge, as they constitute systems of common age and metallicity; however, the absolute values of these quantities are unknown for binaries. In this regard, brown dwarfs that are resolved companions to stars are highly prized, as their distances, ages and metallicities can be established from their primary stars, given the conservative assumption that the systems formed coevally and from material of the same composition. These brown dwarf companions serve as “benchmarks” for studying substellar atmospheres and evolution. Such objects are rare, as numerouspublished(and unpublished)imaging surveys havesearched for brown dwarf companions to nearby main-sequence stars with very limited success (e.g. Lloyd et al. 2004; Oppenheimer et al. 2001; McCarthy & Zuckerman 2004; Carson et al. 2006). Mugrauer et al. (2006)have recently identified a low-luminosity companion to thenearby K0V star HD 3651 (GJ 27, 54 Psc, HR 166, HIP 3093, SAO 74175). The primary star is also notable as it possesses a sub-Saturn mass planet (M sini = 0.20 M ) with an orbital semi-major axis Jup of 0.3 AU (Fischer et al. 2003). HD 3651B has a projected separation of 43′′ (480 AU) and is confirmed to be physically associated by its common proper motion with HD 3651A. Given the distance of 11.11±0.09 pc to the primary (Perryman et al. 1997), the very faint absolute H-band magnitudeandtheabsenceofanoptical counterpartinphotographicplatessuggestthatHD3651B is a very cool brown dwarf. In this paper, we present near-IR photometry and spectroscopy to characterize HD 3651B and compare it to Gl 570D, the coolest known companion around a main-sequence star. After our paper was submitted, Luhman et al. (2006) reported identification and characterization of HD 3651B based on Spitzer/IRAC and ground-based near-IR measurements, and Burgasser (2006) – 3 – reported near-IR spectroscopy and analysis. 2. Observations 2.1. Photometry Weobtainednear-infrared(IR)imagingofHD 3651Bon2006September3UTfromtheUnited Kingdom Infrared Telescope (UKIRT) located on Mauna Kea, Hawaii. We used the facility IR camera UFTI (Roche et al. 2003) and the J (1.25 µm), H (1.64 µm), and K (2.20 µm) filters from theMaunaKeaObservatories(MKO)filterconsortium(Simons & Tokunaga2002;Tokunaga et al. 2002). Wereadouta512×512pixelregionof UFTI’sdetector centered onHD 3651B, leadingtoa 47′′ field of view. The primary star HD 3651A was placed off the array. UKIRT has a fast-steering secondary mirror that provides tip-tilt correction, producing an image quality of 0.55′′ FWHM during our observations. HD 3651B appeared as a single object in the images. Sky conditions were photometric. We obtained a series of 9 dithered images in each filter, for a total on-source integration time of 9 minutes in J, H, and K each. The images were reduced in a standard fashion using the facility data pipeline. The flux calibration was established from observations of the UKIRT standard stars FS 154 and FS 102 (Leggett et al. 2006), observed immediately before and after HD 3651B and at similar airmass. Table 1 presents our final UKIRT/UFTI photometry. 2.2. Spectroscopy We obtained spectroscopy of HD 3651B on 2006 September 4 and 5 UT using the UKIRT facility spectrograph CGS4 (Wright et al. 1993). CGS4 has a Santa Barbara Research Center (SBRC) 256×256 pixel InSb detector. The two-pixel (1.2′′) slit was used for all observations. HD 3651B was observed at three grating settings to span the J-, H- and K-bands, covering 1.03– 1.35, 1.38–2.02 and 1.88–2.52 µm with spectral resolutions of 21 ˚A at J (R = λ/∆λ ≈ 600) and 50 ˚A at H and K (R ≈ 330 and 440). Individual exposure times were 120 seconds for the J setting, 60 seconds for H and 40 seconds for K, with total on-source exposure times of 48, 32, and 59 minutes at J, H, and K, respectively. The target was nodded along the slit by 60′′ in order to avoid scattered light from the very bright primary star. The spectrum of the target at λ< 1.2 µm was still difficult to extract due to the light of the primary, and thus we do not use it. CGS4 has a calibration unit with lamps that provide accurate flat-fielding and wavelength calibration. The F6V star HD 615 was used as a calibrator to remove the effects of the terrestrial atmosphere, with H I recombination lines in its spectrum removed artificially prior to ratioing. The three separate spectraobtainedforHD 3651B weremergedtogether basedonourUKIRT/UFTIphotometryfrom the night before. After scaling, the 1.94–2.01 µm regions that overlapped in the H and K-band spectra agreed to about 5×10−18 W/m2/µm (cf., the K and H-band emission peaks of 2.1×10−16 and 8.2×10−16 W/m2/µm, respectively), thereby indicating therobustnessof themerging process. – 4 – We also obtained low-resolution (R ≈150) spectra of HD 3651B on 2006 September 13 UT from NASA’s Infrared Telescope Facility (IRTF) located on Mauna Kea, Hawaii. Conditions were photometric with excellent seeing conditions, around 0.5′′ FWHM at K-band near zenith. We used the facility near-IR spectrograph Spex (Rayner et al. 1998) in prism mode, obtaining 0.8–2.5 µm spectra in a single order. We used the 0.5′′ wide slit, oriented at the parallactic angle to minimize the effect of atmospheric dispersion. HD 3651B was nodded along the slit in an ABBA pattern, with individual exposure times of 200 sec, and observed over an airmass range of 1.00–1.08. The telescopewasguidedusingimagesofthenearbystar2MASSJ00391738+2115104 obtainedwiththe near-IR slit-viewing camera. Thetotal on-source exposuretime was 73 min. We observed theA0 V star HD 7215 contemporaneously with HD 3651B for flux and telluric calibration. All spectra were reducedusingversion3.3oftheSpeXtoolsoftwarepackage (Vacca et al.2003;Cushing et al.2004). ThereducedIRTF/SpexspectrumisplottedinFigure1,ingoodagreementwiththeUKIRT/CGS4 spectrum. Also, the IRTF/Spex data provide the λ . 1.2 µm region that was unobtainable with the UKIRT data. 3. Results 3.1. Near-IR Spectral Types and Photometry Figure 1 presents our near-IR spectra of HD 3651B, showing the strong water and methane absorption bands that are the hallmarks of the T spectral class. Spectra of other late-T dwarfs are shown for comparison. We classified HD 3651B from the system of five spectral indices established by Burgasser et al. (2006b), as measured independently from our UKIRT/CGS4 and IRTF/Spex data (Table 2). The UKIRT/CGS4 data indicate a spectral type of T7.5, while the IRTF/Spex data indicate T8. (The difference does not arise from the differing spectral resolution of the two datasets, as we verified by smoothing the CGS4 data.) Figure 1 shows that the IRTF/Spex data slightly suggest a later type, based on the very slightly deeper H O and CH absorption. However, 2 4 in both datasets, HD 3651B does not appear as late as the T8 dwarf 2MASS 0415−0935 based on thedepthof the1.15 µmH Oabsorptionband. We alsovisually classified HD 3651B by comparing 2 with UKIRT/CGS4 and IRTF/Spex spectra of late-T dwarfs classified by Burgasser et al., which have the same resolutions and instrumental setups as our data. For data from each instrument, the depth of the H O and CH absorption bands were examined, normalizing the spectra of HD 3651B 2 4 and the comparison objects to their peak fluxes in the J, H, and K-bands. This process confirmed that HD 3651B is later than T7 but earlier than T8. In fact, the depth of the absorption bands for HD 3651B as judged by the indices and by eye are nearly identical to the T7.5 dwarf Gl 570D. Therefore, we assign a spectral type of T7.5 for HD 3651B, with the nominal error of ±0.5 subclasses from the Burgasser et al. system. Our typing is in agreement with other results. Mugrauer et al. (2006) estimate T7–T8 based solely on the H-band absolute magnitude. Using independent sets of IRTF/Spex prism spectra, Luhman et al. (2006) determine T7.5±0.5 based – 5 – on visual examination, and Burgasser (2006) find T8±0.5 based on spectral index measurements. With such a late-type spectrum, HD 3651B is unambiguously a substellar object. Figure 2 shows our near-IR photometry of HD 3651B compared to other ultracool dwarfs; again, HD 3651B appears to be very similar to Gl 570D. Our UKIRT/UFTI H and K-band photometry for HD 3651B agrees within the errors with photometry obtained by Luhman et al. (2006) using the IRTF/Spex slit-viewing camera. However, the J-band results differ by 0.15 mag, inthatourUKIRT/UFTImeasurementtwo monthslater is fainter. Toexplorethisdiscrepancy, we used our single-order IRTF/Spex spectrum to synthesize the near-IR colors and found conflicting results: ourSpex-synthesized J−H agrees with ourUKIRT/UFTI photometry butthesynthesized H − K is redder by 0.08 mag. It is likely that the discrepancy with the Luhman et al. J-band photometry is not significant at these levels. The IRTF/Spex slit-viewing camera is used primarily for acquisition and guiding, and it has not been rigorously tested for precision photometry, e.g. the linearity of the detector response has not been well-characterized (J. Rayner, priv. comm.). However, it is also possible that HD 3651B is variable at the ≈10% level — further monitoring could be valuable. Similarly, the H − K discrepancy between our IRTF/Spex and UKIRT datasets is within the plausible errors of the overall flux calibration for the Spex data. Burgasser et al. (2006a) examined the consistency of broad-band photometric colors compared to colors synthesized from low-resolution IRTF/Spex spectra of 16 late-T dwarfs. They found typical deviations between the observed and synthesized colors of 5% or less, with afew sources having differences as large as 15%. 3.2. Age of the HD 3651AB System We consider several methods for establishing the age of the primary star HD 3651A. Age determination for main-sequence field stars is challenging and imperfect. For solar-type stars, methods for estimating ages largely rely on the increase in stellar rotation period as stars grow older. Stars spin down as they age because stellar winds carry away angular momentum; the increased rotation periods then lead to a decline in stellar activity due to the underlying stellar dynamo. For solar-type stars, chromospheric activity as traced by CaII HK emission provides an age estimate. Donahue (1993, 1998) provide an age calibration for this index: log(t)= 10.725−1.334R +0.4085R2 −0.0522R3 (1) 5 5 5 whereR = 105R′ ,validforlog(R′ )=−4.25to−5.2. Gray et al.(2003)measurelog(R′ ) = 5 HK HK HK −5.09 for HD 3651A with an uncertainty of about 0.05 (R. Gray, priv. comm.) and describe it as an inactive star. Wright et al. (2004) find log(R′ ) = −5.02, with a full range of about 10% HK over 7 years of measurements. (Duncan et al. 1991 report a 40% change in CaII HK emission over 16 years of monitoring.) These measurements lead to an age estimate of about 5–9 Gyr. An error – 6 – estimate is not available for this technique. However, out of a sample of 21 binaries studied by Donahue (1998), most had age estimates differing by .2 Gyr. X-ray emission of solar-type stars also declines with age. Hempelmann et al. (1995) measure L /L = −5.69 for HD 3651A with a 30% uncertainty. Gaidos (1998) provides an age calibration X bol based on scaling relations for stellar activity: log(L /L ) = −6.38−2.6αlog(t /4.6)+log[1+0.4(1−t /4.6)] (2) X bol 9 9 where t is the age in Gyr and α is the coefficient that relates rotation period to stellar age, either 9 α = 0.5 (Skumanich 1972) or α = 1/e (Walter & Berry 1991). Following Wilson et al. (2001), we adopt the zeropoint of −6.38 based on the X-ray luminosity of the Sun from Maggio et al. (1987). Includingtheuncertainties,weestimateanageof0.9–2.2GyrforHD3651A.Asapointofreference, the X-ray luminosity of HD 3651A is log(L ) = 27.6 (Hempelmann et al. 1995). This is about X 6 times fainter than Hyades stars of similar spectral type (Stern et al. 1995). Preibisch & Feigelson (2005) estimate L ∝ t−0.77 for solar-type stars, implying that HD 3651A is about 4× older than X the Hyades (650 Myr) or around 3 Gyr. Due to angular momentum carried away by stellar winds, the rotation periods of solar-type stars increase as they age, believed to follow a power-law relation of P ∼ tα (e.g. Skumanich rot 1972), where α is the same as in Equation 2. Baliunas et al. (1983) report a period of 48 d for HD 3651A’s chromospheric activity, which we adopt as the stellar rotation period. Using the Sun as a reference point (P = 26 d and t = 4.6 Gyr), the scaling relation gives an implausibly large rot age of 16–24 Gyr. Lachaume et al. (1999) have provided an age calibration for main-sequence stars based on a sample from the Hipparcos catalog: log(t )= 2.667log(P)−0.944(B −V)−0.309[Fe/H]+6.530. (3) 9 where t is the age in Gyr and P is the period in days. With B −V = 0.85 mag (Perryman et al. 9 1997) and [Fe/H]=0.09–0.16 (Gray et al. 2003; Santos et al. 2004; Valenti & Fischer 2005), this gives an age of 15 Gyr. Lachaume et al. (1999) caution that their relation is less accurate for stars older than 3 Gyr. Finally, from high resolution spectroscopic analysis combined with bolometric magnitudes and theoretical stellar evolutionary isochrones, Valenti & Fischer (2005) derive an age estimate of 8.2 Gyr with a possible range of 3–12 Gyr, and Takeda et al. (2006) infer a minimum age of 11.8 Gyr. The quoted age range from Valenti & Fischer spans the aforementioned activity-based estimates from the X-ray data (≈1–3 Gyr) and CaII HK data (≈5–9 Gyr), and thus we adopt an age of 3–12 for HD 3651A, with a geometric mean of 6 Gyr. This old age is supported by the star’s slow rotation and the Takeda et al. estimate. – 7 – 3.3. Luminosity, Mass, and Effective Temperature To detemine the bolometric luminosity of HD 3651B, we use a K-band bolometric correction (BC )of2.07±0.13 mag, basedontheGolimowski et al.(2004)polynomialrelationofBC versus K K near-IR spectral type. This gives log(Lbol/L⊙) = −5.58 ± 0.05, with the uncertainty coming from the quadrature sum of the uncertainties in the K-band absolute magnitude, the distance to the system, the BC due to the spectral typing uncertainty, and the scatter about the fitted K polynomial relation. HD 3651B has the second smallest L measured among all brown dwarfs bol with trigonometric distances, comparable to that of the T7.5 dwarf Gl 570D and exceeded only by the T8 dwarf 2MASS 0415−0935, which have log(Lbol/L⊙) = −5.53±0.05 and −5.73±0.07, respectively (Golimowski et al. 2004).1 Instead of the BC from the polynomial relation as a function of spectral type, we could also K have usedthe individualBC values determined for thefourlate-T dwarfs inthe Golimowski et al. K (2004) sample: 2MASS 0727+1710 (T7; BC = 2.24±0.13 mag), Gl 570D (T7.5; 1.90±0.13 mag); K 2MASS 1217−0311 (T7.5; 2.28±0.13 mag); and 2MASS 0415−0935 (T8; 2.03±0.13 mag), where the spectral types are those assigned by Burgasser et al. (2006b). The average of the two T7.5 dwarfs is 2.09 mag, consistent with our adopted value. If we adopt the Gl 570D value based on its close spectral resemblance to HD 3651B, we would obtain an L value 0.08 dex brighter, which bol would not have a significant impact on our conclusions. Finally, Luhman et al. (2006) used near- IR spectra, Spitzer/IRAC thermal IR (3.2–9.2 µm) photometry, and an assumed long wavelength Rayleigh-Jeans distribution to derive a bolometric luminosity of log(Lbol/L⊙) = −5.60±0.05, in excellent agreement with our assessment. Toestimate theT andmassof HD 3651B, weusetheobservational constraints ofthederived eff L and estimated age of 6+6 Gyr combined with the solar-metallicity evolutionary models of bol −3 Burrows et al. (1997). For a nominal age of 6 Gyr, we find an effective temperature of 810 K and a massof56M . Therangesofthesevaluesare780–840 Kandmassesof40–72 M ,withyounger Jup Jup assumed ages leading to cooler temperatures and lower masses. The resulting derived properties are given in Table 1.2 1Analysis of the near-IR spectra of the T8 dwarf 2MASS 0939−2448 and the T7.5 dwarf 2MASS 1114−2618 by Burgasser et al. (2006a) suggest that these two objects may be even cooler and lower luminosity than 2MASS 0415−0935, though they do not yet haveparallaxes measurements. 2Inprinciple,oneshoulduseevolutionarymodelscomputedforthesamemetallicityastheparentstars. However, the slightly super-solar metallicity of HD 3631A (discussed in § 4) should not lead to a significant difference in the derivedproperties. Tables2and3ofSaumon et al.(2006)showthatforlate-Tdwarfs,changingthemetallicityfrom 0.0 dex to 0.3 dex changes the model-derived mass by.5%, Teff by.1%, and log(g) by.0.05 dex for a given age. – 8 – 4. Discussion 4.1. Relative Ages, Masses and Temperatures of HD 3651B and Gl 570D HD3651BisremarkablysimilartotheT7.5dwarfGl 570D,awide-separationcompaniontothe triple system Gl 570ABC (Burgasser et al. 2000; Geballe et al. 2001), separated by 1525 AU from theK4Vprimarystar. Thesetwobrowndwarfshaveverycomparableproperties,namelytheirnear- IR spectra (both spectral type T7.5), JHK absolute magnitudes (HD 3651B is .0.1 mag fainter), and JHK colors (HD 3651B is redder by .0.15 mag). Their primary stars both have about solar metallicity, [Fe/H] = 0.01–0.10 for Gl 570A (Cayrel de Strobel et al. 1997; Feltzing & Gustafsson 1998;Santos et al.2005;Valenti & Fischer2005)and[Fe/H]=0.09–0.16 forHD3651A(Gray et al. 2003;Santos et al. 2004;Valenti & Fischer 2005). Hence, wewould expect that differences between these two brown dwarfs arise primarily from their differing masses and ages, whose observational manifestation is surface gravity. Geballe et al. (2001) estimated an age of 2–5 Gyr for Gl 570A, younger than our estimate of 3–12 Gyr for HD 3651A, but the two estimates derive from different methods. Here we re-examine the relative ages of these two primary stars and in the next section consider the implications in interpreting the spectra of their late-T dwarf companions. Gl 570A is more chromospherically active as judged by its log(R′ ) values −4.49 and −4.75 HK (Henry et al. 1996; Soderblom et al. 1991), compared to log(R′ ) ≈ −5.05 for HD 3651A (§ 3.2). HK Equation 1 gives corresponding ages of 0.8 and 2.2 Gyr for Gl 570A, compared to ≈7 Gyr for HD 3651A.3 Geballe et al. (2001) suggested that the young age inferred from the CaII HK data maybeduetothefactthatGl570Awasobservedduringaperiodofenhancedactivity. Asapointof reference, the Sun varies from log(R′ ) = −5.10 to −4.75, which would lead to chromospherically HK inferred ages of 3–8 Gyr (e.g. Baliunas et al. 1998). Assuming the CaII HK variability of Gl 570A and HD 3651A is comparable to the solar cycle, the two stars would have to had been observed at nearly the opposite extremes of their activity cycles to still have the same age. We discount this possibility and conclude that the chromospheric data indicate that Gl 570A is younger than HD 3651A. Similarly, the other age indicators we have considered support a younger age for Gl 570A compared to HD 3651A. Based on log(L ) = 27.7 from Schmitt & Liefke (2004) and bolometric X corrections from Kenyon & Hartmann (1995), we find Gl 570A has log(L /L ) = −5.27. Using X bol Equation 2, this gives an age of 0.4–0.8 Gyr, compared to 1–3 Gyr inferred for HD 3651A from the same approach. As pointed out by Geballe et al. (2001), Gl 570A has a lower X-ray luminosity compared Hyades stars of similar spectral type, setting a lower age limit of 650 Myr. Also, Gl 570A has a somewhat shorter rotation period of 40 days, compared to 48 days for HD 3651A, implying a ≈50–60% younger age for Gl 570A based on Equation 3 (§3.2). Finally, isochrone analyses 3Strassmeier et al. (2000) report log(R′HK) = −4.21 for Gl 570A, which is higher level of activity than covered by the Donahue (1998) calibration, suggesting an age of .10 Myr. However, the absence of Li I 6708 ˚A absorption indicates that Gl 570A is at least a zero-age main sequencestar (Rocha-Pintoet al. 2002). – 9 – for Gl 570A by Valenti & Fischer (2005) and Takeda et al. (2006) give ages of 3.3+8.3 Gyr and −3.1 <0.6 Gyr, respectively. Table 3 summarizes the age estimates for HD 3651A and Gl 570A. While there is considerable scatter in the absolute ages, the data all suggest that Gl 570A is ≈3× younger than HD 3651A. In the analysis that follows, we adopt a conservative age range of 1–5 Gyr for Gl 570D, with a geometric mean of 2 Gyr. The upper limit of 5 Gyr is the same as that of Geballe et al. (2001) and derives from the scatter of the Donahue (1998) calibration sample. Our adopted lower age limit of 1 Gyr is younger than the 2 Gyr used by Burgasser et al. (2000) and Geballe et al. (2001). Their value is based on the kinematical age of the system as derived from its (U,V,W) space motion and the absence of Hα emission from the Gl 570BC M-type binary. The former is a statistical measure and therefore is not a very strong constraint for an individualstar. Thelatter is also not a strong constraint, since the age dependence of Hα emission for field M dwarfs is also largely based on kinematic analyses (Hawley et al. 1996). The ≈1 Gyr lower age limit discussed here based on stellaractivity indicatorsfortheK-typeprimaryGl570Aisamoreconservativeestimate, especially since these indicators are calibrated with data from open clusters.4 Assuming that the brown dwarfs are coeval with their primary stars, Figure 3 illustrates the model-derived masses, surface gravities, and effective temperatures for HD 3651B and Gl 570D, based on the Burrows et al. (1997) evolutionary models and the observational constraints in Ta- ble 1. The L measurements more strongly constrain T , while the age of the primary star sets bol eff log(g). For Gl 570D and a nominal age of 2 Gyr, the bolometric luminosity of 10−5.53±0.05 L⊙ (Golimowski et al. 2004) gives an effective temperature of 780 K and a mass of 33 M , with Jup ranges of 750–825 K and 24–51 M for ages of 1–5 Gyr. (Note that the apparent overlap of the Jup uncertainties in the Figure 3 is misleading, since the uncertainties represent the plausible spread in the absolute ages. The relative ages of the primary stars are known to better accuracy, as just discussed.) 4.2. Surface Gravity and Metallicity Effects in Late-T Dwarf Spectra Given the age and metallicity determinations for the primary stars, HD 3651B and Gl 570D provide two benchmarks for examining the differential effects of surface gravity and metallicity on the spectra of late-T dwarfs. Figure 3 shows that the older age of HD 3651B leads to a ≈0.3 dex higher inferred surface gravity compared to Gl 570D, with nearly the same T for the two objects. eff For late-T dwarfs, this difference would be most prominently manifested in the K-band emission, which is heavily influenced by opacity from collision-induced H absorption (e.g. Linsky 1969). 2 Higher surface gravity objects will have higher photospheric pressures, and thus the H opacity 2 4Saumon et al.(2006)suggestanagerangeof3–5GyrfortheGl570ABCDsystem,basedonmodelingtheoptical to mid-IRspectrum of the T-dwarf companion Gl 570D. – 10 – will be stronger and the K-band flux more heavily depressed. Indeed, variations in the near-IR propertiesoflate-Tdwarfsaretypically attributedtovariationsinsurfacegravity (e.g.Knapp et al. 2004; Burgasser et al. 2006a; Burrows et al. 2006), which translates into variations in mass because the radii of field brown dwarfs are very similar. However, the inferred surface gravity difference does not seem to be in accord with the ap- pearance of the near-IR spectra of these two objects. Since HD 3651B and Gl 570D have almost identical metallicities (∆[Fe/H] ≈ 0.1 dex) and effective temperatures (∆T ≈ 30 K), the higher eff surface gravity of HD 3651B should lead to bluer near-infrared colors compared to Gl 570D due to stronger H opacity. But instead the observed colors of HD 3651B are redder by 0.12±0.08 and 2 0.15±0.08 mag at H−K and J−K, respectively. Thus, there appears to bea discrepancy between the expected behavior of H opacity and the actual HD 3651B/Gl 570D comparison. Similarly, 2 Burgasser et al. (2006a) use the pressure sensitivity of H opacity to constrain the surface gravity 2 of T dwarfs through spectral indices that ratios the K and H-band peak fluxes. Their Figure 5 implies that increasing gravity by 0.3 dex at T ≈ 800 K should produce a decrease in the K/H eff ratios of ≈20% — instead we see an increase of 17±8% (Table 4). In short, comparing the near-IR colors and spectra of HD 3651B and Gl 570D suggest that HD 3651B has a lower surface gravity and thus a younger age, in contradiction to the relative ages inferred for their primary stars in § 4.1. Wesuggestthatthisapparentdiscrepancyarisesfromthesmallmetallicitydifference(≈0.1dex) betweenthetwosystems. Collision-inducedH opacity isalsoexpectedtobeaffected bymetallicity 2 variations (Borysow et al. 1997). Higher metallicities lead to more opaque atmospheres; therefore, the photosphere resides at lower pressures and collision-induced H opacity is reduced. Models 2 of T dwarf spectra by Burrows et al. (2006) as a function of metallicity ([Fe/H] = −0.5 to +0.5) and surface gravity (log(g)= 4.5 to 5.5) show that both variables can have strong effects on the emergent near-IR SED. For instance, Figure 20 of Burrows et al. shows that at fixed T , the J−K eff model colors vary greatly with metallicity and surface gravity and that the two quantities act in opposite senses, as expected: the higher surface gravities that produce bluer near-IR colors can be counteracted by higher metallicities leading to redder colors. Likewise, models by M. Marley et al. (in prep.) suggest that the ≈0.1 dex metallicity difference between HD 3651B and Gl 570D could counteract and even outweigh the effect of the 0.3 dex difference in log(g) on their relative K-band fluxes. To quantify the sensitivity of near-IR spectra to changes in metallicity and surface grav- ity, we examine the condensate-free atmospheric models from the Tucson group, as described in Burrows et al. (2002) and Burrows et al. (2006). Burgasser et al. (2006a) define a set of 5 spectral indices to characterize the emission peaks and H O absorption in late-T dwarf spectra; our mea- 2 surements for HD 3651B and Gl 570D are given in Table 4. For a particular spectral index, one can envision that its model-predicted values constitute a 3-dimensional surface in the parameter space of {T , log(g), Z}. We compute the local slope of the surface to quantify how the index eff varies with effective temperature, surface gravity, and metallicity about nominal reference values