Since the precessional frequencies of fat and water are different, water precesses faster; they do not move in phase. So from the Larmor relationship, we see that the fat and water spins will respond differently to the refocusing gradient. This difference means that phase differences will accumulate between the two spectral components in the signal (water and fat). Thus there will be echo times (TEs) when fat and water spins (at a similar location, i.e. within the same voxel) will be totally in phase and totally out of phase. If the selected TE is such that the spins are out of phase, then a black boundary will be seen around organs surrounded by fat, such as the kidneys and muscles (signal cancellation occurs in voxels where there is both fat and water). This sharp delineation of the muscle-fat boundary is sometimes visually appealing but is not an anatomical structure. At 1.5 T, the 3.5 ppm difference in frequency between water and saturated fat results in cancellation of spins at 4.6 ms multiples, starting at about 2.3ms; for example at 6.9ms, 11.5ms, and 16.1ms. To avoid this artefact, in-phase TEs that are multiples of 4.6ms (for 1.5T) should be chosen (4.6, 9.2, 13.8 ms, etc). Similarly at 3T, in-phase TEs that are multiples of 2.3ms should be chosen (2.3, 4.6, 9.2 ms, etc). Unlike chemical shift of the first kind, this phase cancellation artefact can appear in both encoding directions: at any interface at which voxels contain both fat and water. Black boundary artefact does not appear in conventional SE imaging because of the refocusing 180° pulse, not present in GE imaging.