Hi,
Try this file.
By default, the input field is 1000 V/m, you’ll see that the light propagates across the crystal almost unperturbed. Now if you change the input field to 1 V/m (switch off the pump), then the light is steered to the other output port.
This is the expected behavior, however, I suspect a sign problem with the nonlinearities somewhere as the behavior is the inverse from the the paper!
You can try and experiment with various coefficients values and power levels.
Best regards