In my limited experience, heating a relatively fine anatase powder (average diameter 0.5 micron by gas permeability) for 1 hour from 900 to 1000 degC resulted in just the appearance of traces of rutile, based on XRD. At the same time the powder coarsened somewhat (to average diameter of 1.1 micron). I relied for this on Iida, Y. and Ozaki, Sh., DOI: 10.1111/j.1151-2916.1961.tb13725.x. The purpose was to eliminate water but with minimum conversion. The atmosphere was almost certainly nitrogen (didn't specifically write it down).