Predicting X-ray emission from wind-blown bubbles — limitations of fits to ROSAT spectra

Wind-blown bubbles, from those around massive O and Wolf–Rayet stars to superbubbles around OB associations and galactic winds in starburst galaxies, have a dominant role in determining the structure of the interstellar medium (ISM). X-ray observations of these bubbles are particularly important as most of their volume is taken up with hot gas, 105 ≲ T (K) ≲ 108.  However, it is difficult to compare these X-ray observations, usually analysed in terms of single- or two-temperature spectral model fits, with theoretical models, as real bubbles do not have such simple temperature distributions. Spectral fits, and the properties inferred from them, will depend in a complex way on the true temperature distribution and the characteristics and limitations of the X-ray observatory used.  In this introduction to a series of papers detailing the observable X-ray properties of wind-blown bubbles, we describe the method with which we intend to solve this problem, analysing a simulation of a wind-blown bubble around a massive star.  Our model is of a wind of constant mass and energy injection rate, blowing into a uniform ISM, from which we calculate X-ray spectra as they would be seen by the ROSAT PSPC. Analysing these spectra in the same way as a real observation, we compare the properties of the bubble as would be inferred from the ROSAT data with the true properties of the bubble in the simulation.  We find standard spectral models yield inferred properties that deviate significantly from the true properties, even though the spectral fits are statistically acceptable, and give no indication that they do not represent the true spectral distribution. For example, single-temperature spectral fits give best-fitting metal abundances that are only 4 per cent of the true value. A cool bubble has best-fitting temperatures significantly higher than a bubble twice as hot. These results suggest that in any case in which the true source spectrum does not come from a simple single- or two-temperature distribution, the ‘observed’ properties cannot naively be used to infer the true properties. In this situation, to compare X-ray observations with theory it is necessary to calculate the observable X-ray properties of the model.