LETTER doi:10.1038/nature14166 The formation of a quadruple star system with wide separation Jaime E. Pineda 1 , Stella S. R. Offner 2,3 , Richard J. Parker 4 , He ´ctor G. Arce 5 , Alyssa A. Goodman 6 , Paola Caselli 7 , Gary A. Fuller 8 , Tyler L. Bourke 9,10 & Stuartt A. Corder 11,12 The initial multiplicity of stellar systems is highly uncertain. A num- ber of mechanisms have been proposed to explain the origin of binary and multiple star systems, including core fragmentation, disk frag- mentation and stellar capture 1–3 . Observations show that protostellar and pre-main-sequence multiplicity is higher than the multiplicity found in field stars 4–7 , which suggests that dynamical interactions occur early, splitting up multiple systems and modifying the initial stellar separations 8,9 . Without direct, high-resolution observations of forming systems, however, it is difficult to determine the true initial multiplicity and the dominant binary formation mechanism. Here we report observations of a wide-separation (greater than 1,000astronomical units) quadruple system composed of a young protostar and three gravitationally bound dense gas condensations. These condensations are the result of fragmentation of dense gas filaments, and each condensation is expected to form a star on a time- scale of 40,000 years. We determine that the closest pair will form a bound binary, while the quadruple stellar system itself is bound but unstable on timescales of 500,000 years (comparable to the lifetime of the embedded protostellar phase 10 ). These observations suggest that filament fragmentation on length scales of about 5,000 astronomical units offers a viable pathway to the formation of multiple systems. Barnard 5 (B5) is a dense core in the Perseus star-forming region (at a distance of 250 pc) that hosts at least one young, forming star 11 . Imaging of B5 in the emission of the dense-gas-tracing NH 3 (1,1) line shows it to be an example of a ‘coherent’ dense core 12 , which is a contiguous high-density region with subsonic levels of turbulence 13 . Higher-resolution imaging reveals narrow filamentary structure within the coherent core 14 . We observed the NH 3 (1,1) and (2,2) lines using the Karl G. Jansky Very Large Array (VLA) 15 , which reveals that the filaments in B5 are frag- menting and that they are in the process of forming a wide-separation multiple stellar system. Nearly half of all stars reside in multiple star systems 4,16 . Consequently, a host of phenomena, ranging from supernova rates to planet formation, depend on understanding stellar multiplicity 17 . Because of the observa- tional challenges associated with observing early systems, the dominant ideas for binary formation are based on simulations and analytic argu- ments, which naturally require a variety of assumptions 4,18 . To date, observations have not captured the formation of a binary system at a stage where its origin is unambiguous, and prior observations of core substructure lack the spatial and kinematic resolution to be used in predicting whether observed structures would form protostars and/or produce a bound system 19,20 . The observed kinematics and separation (.1,000 astronomical units, AU) of the B5 system is significant because it demonstrates a clear mechanism for wide binary formation and pro- vides convincing evidence that the observed condensations will become a bound multiple star system. Detailed knowledge of the underlying distribution of dense gas is the key to determining which structures will go on to form stars. Here we identify the dense gas structures that are most likely to form stars using the dendrogram technique 21 . Dendrogram analysis is a hierarchical struc- ture decomposition that uses isocontours to identify individual features, while also determining where these contours merge with adjacent struc- tures to create a new parental structure. We refer to the smallest scale (and brightest) structures in the dendrogram as condensations. These are the most likely places for an individual star to form. Figure 1a shows the B5 region as seen in dense gas (number density of H 2 , n H2 >10 4 cm -3 ), with the protostar and the identified gas condensations shown by a star and circles, respectively. The mass of the well known protostar B5-IRS1 is 0.1 solar masses (M Sun ; ref. 22), while the masses of condensations B5-Cond1, B5-Cond2 and B5-Cond3 are 0.36 6 0.09 M Sun , 0.26 6 0.12 M Sun and 0.30 6 0.13 M Sun , respectively. Uncertainty in these masses is dominated by the uncertainty in the temperature used to convert mea- sured fluxes to masses. The radii of the three condensations are respec- tively 2,800 AU, 2,300 AU and 2,500 AU, while the projected separations between the same three condensations and the protostar are 3,300 AU, 5,100 AU and 11,400 AU (see Methods). The half-mass radii of the con- densations are about half the condensation radii. This, combined with the mass radius relations (Extended Data Fig. 2), suggests that the central regions will collapse faster than the whole condensations and before interactions between condensations can play a major role in the stars’ initial separations. Although these separations are large, they are con- sistent with initial protostellar pair separations predicted for core frag- mentation by numerical simulations 2 . In the simulations, protostellar separations evolve rapidly on timescales of 0.1 Myr, and some systems become unbound while others migrate to closer proximity. Projected proximity on the sky does not necessarily imply that objects are physically related. However, the line-of-sight velocities of the observed condensations are similar and the grouping is likely to be physically asso- ciated. The velocity dispersion, s v , of the dense gas provides another im- portant piece of information, the gas kinetic energy, which is needed to determine whether the condensations are transient structures or grav- itationally bound and likely to form a star. The velocities and velocity dispersions of the condensations are determined by fitting NH 3 (1,1) and (2,2) line profiles 14 . The condensations and protostar display the same centroid velocity to within 0.2 km s 21 and are therefore associated with the same dense core. The level of turbulence in this region is so low, s turb < 0.53–0.66 times the sound speed in the gas 14 , that gravity will over- whelm the combined turbulent and thermal pressure in all the identified condensations, and a star will probably form in each case. The timescale for these condensations to undergo gravitational collapse is approximately the gas free-fall time, which we estimate to be 40,000 years (Methods). This timescale is sufficiently short to ensure that the system’s spatial 1 Institute for Astronomy, ETH Zurich, Wolfgang-Pauli-Strasse 27, CH-8093 Zurich, Switzerland. 2 Department of Astronomy, Yale University, PO Box 208101, New Haven, Connecticut 06520-8101, USA. 3 Department of Astronomy, University of Massachusetts, 710 North Pleasant Street, Amherst, Massachusetts 01003, USA. 4 Astrophysics Research Institute, Liverpool John Moores University, 146 Brownlow Hill, Liverpool L3 5RF, UK. 5 Department of Astronomy, Yale University, PO Box 208101, New Haven, Connecticut 06520-8101, USA. 6 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, USA. 7 Max-Planck-Institut fu ¨ r extraterrestrische Physik (MPE), Gießenbachstrasse 1, D-85741 Garching, Germany. 8 UK ARC Node, Jodrell Bank Centre for Astrophysics, School of Physics and Astronomy, Alan Turing Building, Oxford Road, University of Manchester, Manchester M13 9PL, UK. 9 SKA Organisation, Jodrell Bank Observatory, Lower Withington, Macclesfield, Cheshire SK11 9DL, UK. 10 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, USA. 11 Joint ALMA Observatory, Alonso de Cordova 3107, Vitacura, Santiago, Chile. 12 National Radio Astronomy Observatory, 520 Edgemont Road, Charlottesville, Virginia 22903, USA. 12 FEBRUARY 2015 | VOL 518 | NATURE | 213 Macmillan Publishers Limited. All rights reserved ©2015