rotating a tensor where the tensor elements are functions of a third index (conceptually)

I have a 3x3 tensor "A" where the elements are a functions of wavelength (e.g. 1000 deep). I need to rotate the tensor, R*A*R', then do other operations. Here R is the usual 3x3 rotation matrix. It would be most convenient to have the tensor look like:

A = [ axx(k),   0(k),    0(k);
      0(k),    ayy(k),   0(k);
      0(k),   0(k),    azz(k) ];

I can do all this my looping over k and treating each k-layer as a 3x3, but is there a better way?

There are really 2 questions.

1) Is there a more efficient way to load the 3x3xk array from 1D vectors of axx(k), ayy(k), and azz(k), compared to constructing each k-layer one at a time in a for loop?

2) Is there a way to do the rotation of A (the 3x3xk matrix) without doing the rotation one 3x3 layer at a time in a for loop?