M398_11b 129..132

v < va. In the other extreme, when v ˆ 0:4 arcsec, the magnetic ®eld from the magnetar would be negligible, but the inferred ®eld strength can be easily explained as arising from compression of the ambient ®eld by the bow shock. The mean expansion speed of the nebula is 3:5 3 1010d5v4t 2 1 10 cm s 2 1 where t10 is the age of the nebula in units of 10 days. Thus in the limit of v ˆ 0:4 arcsec the nebula would have expanded relativistically. This is possible if previous outbursts have swept up the ambient gas. In this limit, Umin is comparable with the isotropic burst energy of 3 3 10 42d25 erg, as estimated from the ̄uence of the 27 August burst (,10 2 3 erg cm 2 ; M. Feroci, personal communication). In either case, it is clear that nebular expansion accounts for the rapidly decaying radio emission. Thus the above energy estimates (derived from observations 7±10 days after the burst) need to be revised upwards to obtain the initial release of energy. In the future, the availability of broad-band observations and/or direct measurement of the size of the nebula should enable us to estimate directly the energy of the burst in particles (without the uncertainties of beaming that bedevil estimates from g-ray data). Equally important are accurate measurements of the average particle luminosity, as particle-aided spindown can substantially modify estimates for the magnetic ®eld and the characteristic age of a neutron star. These estimates, to our knowledge, are unobtainable in any other fashion. In those cases where the SGR is immersed in a high-pressure region (for example, SGR1806-20), we are able to trace the entire history of energy loss from the magnetar. The sub-arcsecond localization presented here should greatly help the identi®cation of possible stellar counterparts of this SGR (as was done for SGR1806-20; refs 22, 23). M