At what particle energy do extragalactic cosmic rays start to predominate?

We have argued [J. Szabelski et al. (2002)] that the well-known `ankle' in the cosmic ray energy spectrum, at logE (eV) ~ 18.7-19.0, marks the transition from mainly Galactic sources at lower energies to mainly extragalactic above. Recently, however, there have been claims for lower transitional energies, specifically from logE (eV) ~ 17.0 [G. Thompson et al. (2004)] via 17.2-17.8 [V.S. Berezinsky et al. (2004)] to 18.0 [A.M. Hillas (2004)]. In our model the ankle arises naturally from the sum of simple power law-spectra with slopes differing by ~ 1.8; from differential slope -3.8 for Galactic particles (near logE = 19) to ~ -2.0 for extragalactic sources. In the other models, on the other hand, the ankle is intrinsic to the extragalactic component alone, and arises from the shape of the rate of energy loss versus energy for the (assumed) protons interacting with the cosmic microwave background (CMB). Our detailed analysis of the world's data on the ultra-high energy spectrum shows that taken together, or separately, the resulting mean sharpness of the ankle (second difference of the log(intensity*E^3) with respect to logE) is consistent with our `mixed' model. For explanation in terms of extragalactic particles alone, however, the ankle will be at the wrong energy - for reasonable production models and of insufficient magnitude if, as seems likely, there is still a significant fraction of heavy nuclei at the ankle energy.