LETTER doi:10.1038/nature13791 An ultraluminous X-ray source powered by an accreting neutron star M. Bachetti1,2, F. A. Harrison3, D. J. Walton3, B. W. Grefenstette3, D. Chakrabarty4,F.Fu¨rst3, D. Barret1,2, A. Beloborodov5, S. E. Boggs6, F. E. Christensen7,W.W.Craig8, A. C. Fabian9, C. J. Hailey10, A. Hornschemeier11, V. Kaspi12, S. R. Kulkarni3, T. Maccarone13, J. M. Miller14, V. Rana3, D. Stern15, S. P. Tendulkar3, J. Tomsick6, N. A. Webb1,2 & W. W. Zhang11 12 The majority of ultraluminous X-ray sources are point sources that ULXs, the most luminous being M82 X-1 , which can reach LX(0.3– are spatially offset from the nuclei of nearby galaxies and whose X-ray 10 keV) < 1041 erg s21, and the second brightest being a transient, M82 luminosities exceed the theoretical maximum for spherical infall (the X-2 (also referred to as X42.315913), which has been observed to reach4,14 1,2 40 21 Eddington limit) onto stellar-mass black holes . Their X-ray lumi- LX(0.3–10 keV) < 1.8 3 10 erg s . The two sources are separated by nosities in the 0.5–10 kiloelectronvolt energy band range from 1039 50, and so can only be clearly resolved by the Chandra X-ray telescope. to 1041 ergs per second3. Because higher masses imply less extreme During the M82 monitoring campaign, NuSTAR observed bright emis- ratios of the luminosity to the isotropic Eddington limit, theoretical sion fromthe nuclear region containing the two ULXs. Theregion shows models have focused on black hole rather than neutron star systems1,2. moderate flux variability at the 20% level during the first 22 days of obser- The most challenging sources to explain are those at the luminous end vation. The flux was then found to have decreased by 60% by the time 40 of the range (more than 10 ergs per second), which require black hole of the final observation ,20 days later. The peak X-ray flux, FX(3– masses of 50–100 times the solar value or significant departures from 30 keV) 5 (2.33 6 0.01) 3 10211 erg cm22 s21 (90% confidence; Fig. 1) the standard thin disk accretion that powers bright Galactic X-ray corresponds to a total 3–30 keV luminosity assuming isotropic emis- z0:01| 40 21 binaries, or both. Here we report broadband X-ray observations of sion of 3:7{0:02 10 erg s . the nuclear region of the galaxy M82 that reveal pulsations with an A timing analysis revealed a narrow peak just above the noise at a average period of 1.37 seconds and a 2.5-day sinusoidal modulation. frequency of ,0.7 Hz in the power density spectrum. An accelerated The pulsations result from the rotation of a magnetized neutron star, epoch folding search15 on overlapping 30-ks intervals of data found coher- and the modulation arises from its binary orbit. The pulsed flux alone ent pulsations with a mean period P of 1.37 s modulated with a sinusoidal corresponds to an X-ray luminosity in the 3–30 kiloelectronvolt range period of 2.53 days throughout the 10-day interval starting at modified of 4.9 3 1039 ergs per second. The pulsating source is spatially coin- Julian day (MJD) 56691 (2014 February 03), and also in the last obser- cident with a variable source4 that can reach an X-ray luminosity in vation at MJD 56720 (Fig. 1). The statistical significance of the pulsa- the 0.3–10 kiloelectronvolt range of 1.8 3 1040 ergs per second1.This tions is ,13s during the most significant 30-ks segments, and .30s for association implies a luminosity of about 100 times the Eddington the entire observation. A refined analysis (see Methods) subsequently limit for a 1.4-solar-mass object, or more than ten times brighter enabled the detection of pulsations over a longer interval beginning on than any known accreting pulsar. This implies that neutron stars MJD 56686. The pulsed flux is variable, ranging from 5% to 13% in the may not be rare in the ultraluminous X-ray population, and it chal- 3–30 keV range, and from 8% to 23% in the 10–30 keV range; see Fig. 1c. lenges physical models for the accretion of matter onto magnetized Although the pulsed flux increases with energy, this may result from a compact objects. reduction in the contamination from other sources in the point-spread The brightest accretion-powered X-ray pulsars, A0538-665,SMCX-16 function rather than from a true increase in pulsed fraction. The maximum and GRO J1744-287, are variable, with reported X-ray luminosities up pulsed luminosity of the periodic source, which we will hereafter refer 39 21 z0:02| 39 21 to LX(2–20 keV) < 10 erg s , which are at the low end of the range to as NuSTAR J09555116940.8, is 4:9{0:03 10 erg s (3–30 keV). that defines ultraluminous X-ray sources (ULXs). Such luminosities, Analysis of the period modulation yields a near-circular orbit (upper exceeding the Eddington limit for a neutron star by a factor of about six, eccentricity limit of 0.003; see Methods) with a projected semi-major can be understood8 as resulting from accretionof material at moderately axis of 22.225(4) light s (1s error). In addition to the orbital modula- super-Eddington rates through a disk that couples to the neutron star’s tion, a linear spin-up of the pulsar is evident, with a period derivative strong dipolar magnetic field (surface fields of B < 1012 G). At the mag- P_ < 22 3 10210 ss21 over the interval from MJD 56696 to 56701 when netospheric (Alfve´n) radius, material is funnelled along the magnetic the pulse detection is most significant. Phase-connecting the observa- axis, radiation escapes from the column’s side, and radiation pressure tions enables detection of a changing spin-up rate over a longer timespan is ineffective at arresting mass transfer9,10. Explaining ULXs that have (Fig. 2a) as well as erratic variations likely to be related to local changes 39 21 16 LX . 10 erg s with a ,1M[ compact object using typical accreting in the torque on the neutron star applied by accreting matter . pulsar models is however extremely challenging8. Chandra observed M82 from MJD 56690.8396 to 56692.618, during The NuSTAR (Nuclear Spectroscopic Telescope Array) high energy an epoch when pulsations are detected. Only two sources in the Chandra X-ray mission11 observed the galaxy M82 (at distance d < 3.6 Mpc) seven image, M82 X-1 and M82 X-2, are sufficiently bright to be the counter- times between 2014 January 23 and 2014 March 06 as part of a follow-up part of NuSTAR J09555116940.8 (see Methods). We find the centroid of campaign of the supernova SN 2014J. The galaxy’s disk contains several the pulsed emission (right ascension a 5 09 h 55 min 51.05 s, declination 1Universite´ de Toulouse, UPS-OMP, Institut de Recherche en Astrophysique et Plane´tologie, 9, Avenue du Colonel Roche, BP 44346, 31028 Toulouse Cedex 4, France. 2CNRS, Institut de Recherche en Astrophysique et Plane´tologie, 9, Avenue du Colonel Roche, BP 44346, 31028 Toulouse Cedex 4, France. 3Cahill Center for Astrophysics, 1216 East California Boulevard, California Institute of Technology, Pasadena, California 91125, USA. 4MIT Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA. 5Physics Department, Columbia University, 538 West 120th Street, New York, New York 10027, USA. 6Space Sciences Laboratory, University of California, Berkeley, California 94720, USA. 7DTU Space, National Space Institute, Technical University of Denmark, Elektrovej 327, DK-2800 Lyngby, Denmark. 8Lawrence Livermore National Laboratory, Livermore, California 94550, USA. 9Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK. 10Columbia Astrophysics Laboratory, Columbia University, New York, New York 10027, USA. 11NASA Goddard Space Flight Center, Greenbelt, Maryland 20771, USA. 12Department of Physics, McGill University, Montreal, Quebec H3A 2T8, Canada. 13Department of Physics, Texas Tech University, Lubbock, Texas 79409, USA. 14Department of Astronomy, University of Michigan, 500 Church Street, Ann Arbor, Michigan 48109-1042, USA. 15Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California 91109, USA. 202|NATURE|VOL514|9OCTOBER2014 ©2014 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH a 0.3 a 0.1 Best-fit model period 0.05 0.25 Chandra 0 0.2 –0.05 006 007 008 009 Period – 1372.5 (ms) –0.1 b 0.15 6 Orbital variations removed Counts per s module FPMA 4 FPMB 2 ObsID 002 004 006 007 008–009 011 0.1 0 30 ks Ephemeris: b 1 –2 40 ks PEPOCH (MJD) = 56696 F0 (Hz) = 0.72857393(5) TOA residual (s) TOA residual 50 ks –4 F (Hz s–1) = 5.47(2) × 10–11 0.5 1 c 200 0 Secular trends removed 100 –0.5 0 Period – 1372.5 (ms) –1 c –100 30 5 (ms) TOA residual –200 χ 0 Δ 56685 56687.5 56690 56692.5 56695 56697.5 56700 56702.5 25 –5 MJD 012 20 Phase Figure 2 | The spin-up behaviour of NuSTAR J09555116940.8. a, The residual period after correcting for the sinusoidal orbital modulation given in 15 Extended Data Table 2. The period, displayed through the best-fit in Extended Pulsed flux (%) Data Table 3, decreases consistently, but the spin-up rate is changing. b, Time 10 03–10 keV of arrival (TOA) residuals after removing the best-fit sinusoidal orbital 03–30 keV modulation and a constant period derivative (the parameters are shown in 5 10–30 keV 26 commonly used units ).