THE DISCOVERY OF A COMPANION TO THE LOWEST MASS WHITE DWARF1Mukremin Kilic,2Warren R. Brown,3Carlos Allende Prieto,4M. H. Pinsonneault,2and S. J. Kenyon3Received 2007 April 3; accepted 2007 April 8ABSTRACTWe report the detection of a radial velocity companion to SDSS J091709.55+463821.8, the lowest mass white dwarfcurrently known, with M  0:17 M. The radial velocity of the white dwarf shows variations with a semi-amplitudeof 148:8 6:9kms1and a period of 7:5936 0:0024 hr, which implies a companion mass of M  0:28 M. Thelack of evidence of a companion in the optical photometry forces any main-sequence companion to be smaller than0.1 M, hence a low-mass main-sequence star companion is ruled out for this system. The companion is most likelyanother white dwarf, and we present tentative evidence for an evolutionary scenario that could have produced it.However, a neutron star companion cannot be ruled out, and follow-up radio observations are required to search for apulsar companion.Subject headinggs: stars: individual (SDSS J091709.55+463821.8) — stars: low-mass, brown dwarfs — white dwarfs1. INTRODUCTIONRecent discoveries of several extremely low mass white dwarfs(WDs) in the field (Kilic et al. 2007; E isenstein et al. 2006;Kawka et al. 2006) and around pulsars (Bassa et al. 2006; vanKerkwijk et al. 1996) show that WDs with mass as low as 0.17 Mare formed in the Galaxy. No galaxy is old enough to produce suchextremely low mass WDs through single star evolution. The oldestglobular clusters in our Galaxy are currently producing 0.5 MWDs ( Hansen et al. 2007); therefore, lower mass WDs must ex-perience significant mass loss. The most likely explanation is aclose binar y companion. If a WD forms in a close binary, it canlose its outer envelope without reaching the asymptotic giant branchand without ever igniting helium, ending up as a low-mass, he-lium core WD. Confirmation of the binary nature of several low-mass WDs by Marsh et al. (1995) supports this binary formationscenario.White dwarf binaries provide an important tool for testing bi-nary evolution, specifically the efficiency of the mass-loss processand the common-envelope phase. Since WDs can be created onlyat the cores of giant stars, their properties can be used to recon-struct the properties of the progenitor binar y systems. Using asimple core mass-radius relation for giants and the known orbitalperiod, the initial orbital parameters of the binary system can bedetermined. For a review of binary evolutio n involving WDs,see e.g., Sarna et al. (1996), Iben et al. (1997), Sandquist et al.(2000), Yungelson et al. (2000) , Nelemans & Tout (2005), andBenvenuto & De Vito (2005).Known companions to low-mass WDs include late-type main-sequence stars ( Farihi et al. 2005; Maxted et al. 2007), heliumor carbon /oxygen core WDs ( Marsh et al. 1995; Marsh 2000;Napiwotzki et al. 2001), and in some cases neutron stars (Niceet al. 2005). Late-type stellar companions to low-mass WDs havea distribution of masses with median 0.27 M(Nelemans & Tout2005). This median companion mass is nearly identical to thepeak companion mass of 0.3 M(spectral type M3.5) observedin the field population of late-type main-sequence stars within20 pc of the Earth (Farihi et al. 2005). Low-mass WD-WD bina-ries, on the other hand, tend to have equal mass WDs. The medianmass for both the brighter and dimmer components of the knownlow-mass WD binary systems is 0.44 M(Nelemans & Tout2005).The discovery of extremely low mass WDs around pulsars sug-gests that neutron star companions may be responsible for creatingwhite dwarfs with masses of about 0.2 M. PSR J04374715,J0751+1807, J1012+5307, J1713+0747, B1855+09, and J19093744 are pulsars in pulsar-He WD binary systems with circularorbits and orbital periods of 0.2–100 days ( Nice et al. 2005).So far, only two of these companions are spectroscopically con-firmed to be helium core WDs. Van Leeuwen et al. (2007) searchedfor radio pulsars around eight low mass WDs, but did not find anycompanions. They concluded that the fraction of low mass heliumcore WDs with neutron star companions is less than 18% 5%.Kilic et al. (2007) reported the discovery of the lowest massWD currently known: SDSS J091709.55+463821.8 (hereafterJ0917+46). With an estimated mass of 0.17 M, J0917+46 pro-vides a unique opportunity to search for a binary companion andto test our understanding of the formation scenarios for extremelylow-mass WDs. Do extremely low-mass WDs form in binarieswith neutron stars, WDs, or late-type stars? If the companion is aneutron star, the mass of the neutron star can be used to constrainthe neutron star equation of state. In case of a WD or a late-typestar companion, the orbital parameters can be used to constrainthe common-envelope phase of binary evolution in these systems.In this paper, we present new optical spectroscopy and radial ve-locity measurements for J0917+46. Our observations are dis-cussed in x 2, while an analysis of the spectroscopic data and thediscovery of a companion are discussed in x 3. The nature of thecompanion is discussed in x 4.2. OBSERVATIONSWe used the 6.5 m MMT telescope equipped with the BlueChannel Spectrograph to obtain moderate resolution spectros-copy of SDSS J0917+46 nine times over the course of five nightsbetween UT 2006 December 22–27 and five times on UT 2007March 19. The spectrograph was operated with the 832 line mm11Observations reported here were obtained at the MMT Observatory, a jointfacility of the Smithsonian Institution and the University of Arizona.2The Ohio State University, Department of Astronomy, Columbus, OH 43210.3Smithsonian Astrophysical Observatory, 60 Garden Street, Cambridge,MA 02138.4McDonald Observatory and Department of Astronomy, University of Texas,Austin, TX 78712.1088The Astrophysical Journal, 664:1088 – 1092, 2007 August 1# 2007. The American Astronomical Society. All rights reserved. Printed in U.S.A.grating in second order, providing a wavelength coverage of 3650–4500 8. Most spectra were obtained with a 1.000slit, yielding aresolving power of R ¼ 4300; however, a 1.2500slit was used on2006 December 24, which resulted in a resolving power of 3500.Exposure times ranged from 15 to 22 minutes and yielded signal-to-noise ratio S/N > 20 in the continuum at 4000 8. All spectrawere obtained at the parallactic angle, and comparison lamp ex-posures were obtained after every exposure. The spectra