A person asked me this and I'm just a lowly physical chemist.
I used a classical analogy (how good or bad is this and how to fix?) Basically, light has a net angular momentum of zero, insofar as it is not polarized into its left and right plane polarized forms until it hits a crystalline structure.
However, once it does hit such a structure, we have left and right plane polarized light--that is left and right photon beams.
Since the original light was not polarized, this polarization (left right) be inherent in the light. The original light is a superposition of left and right polarized light, each with a total angular momentum of -1 and 1, so that they result in the total zero polarized incident light.
Thuse they are entangled to lower their spin (quantum angular momentum). Once we measure one of the particles in the superposition, we know the other by conservation of angular momentum.
Is this close?