More generally one must use the local order-parameter traceless tensor ${\mathsf{S}}_{ij}\left(\mathbf{r}\right)$. This can be particularly important in the smallest droplets, where significant order parameter depressions can be induced by the strong director gradients.

The divergence of the stress tensor actually does not provide the complete electromagnetic force in electrodynamical systems. An additional force of the form $-(\partial \mathbf{S}∕\partial t)∕c^2$, where $\mathbf{S}$ is the Poynting vector of the optical wave, should be included (this is part of the so-called Abraham force), as discussed in Ref. [13]. However, after optical-cycle averaging this last term is negligible whenever the matter dynamics is very slow with respect to the droplet crossing time by light, as in our case (see the Appendix for a fuller explanation).

In Eq. (13), a contribution to the angular momentum density associated with the electromagnetic field itself, of the form $\mathbf{r}\times \mathbf{S}∕c^2$, has been neglected for the same reasons as those discussed in [16].

When doing experiments with 5CB droplets, it was relatively easy to heat them above the clearing temperature (nematic to isotropic transition), either by laser-induced heating of the single droplet or by heating the whole sample. In any case, radial droplets usually transformed into axial after being heated above the clearing point and then left relaxing back to the nematic phase. This showed that axial alignment is probably the most stable one.