ULX spectra revisited: Accreting, highly magnetized neutron stars as the engines of ultraluminous X-ray sources

Aims. In light of recent discoveries of pulsating ultraluminous X-ray sources (ULXs) and recently introduced theoretical schemes that propose neutron stars (NSs) as the central engines of ULXs, we revisit the spectra of eighteen well known ULXs, in search of indications that favour this newly emerging hypothesis. Methods. We examine the spectra from high-quality XMM-Newton and NuSTAR observations. We use a combination of elementary black body and multicolour disk black body (MCD) models, to diagnose the predictions of classic and novel theoretical models of accretion onto NSs. We re-interpret the well established spectral characteristics of ULXs in terms of accretion onto lowly or highly magnetised NSs, and explore the resulting parameter space for consistency. Results. We confirm the previously noted presence of the low-energy (less than or similar to 6 keV) spectral rollover and argue that it could be interpreted as due to thermal emission. The spectra are well described by a double thermal model consisting of a hot (greater than or similar to 1 keV) and a cool (less than or similar to 0.7 keV) multicolour black body (MCB). Under the assumption that the cool MCD emission originates in a disk truncated at the neutron star magnetosphere, we find that all ULXs in our sample are consistent with accretion onto a highly magnetised (B greater than or similar to 10(12) G) neutron star. We note a strong correlation between the strength of the magnetic field, the temperature of the hot thermal component and the total unabsorbed luminosity. Examination of the NuSTAR data supports this interpretation and also confirms the presence of a weak, high-energy (greater than or similar to 15 keV) tail, most likely the result of modification of the MCB emission by inverse Compton scattering. We also note that the apparent high-energy tail, may simply be the result of mismodelling of MCB emission with an atypical temperature (T) versus radius (r) gradient, using a standard MCD model with a fixed gradient of T similar to r(-0.75). Conclusions. We have offered a new and robust physical interpretation for the dual-thermal spectra of ULXs. We find that the best-fit derived parameters of our model, are in excellent agreement with recent theoretical predictions that favour super-critically accreting NSs as the engines of a large fraction of ULXs. Nevertheless, the considerable degeneracy between models and the lack of unequivocal evidence cannot rule out other equally plausible interpretations. Deeper broadband observations and time-resolved spectroscopy are warranted to further explore this newly emerging framework.