GSOLVER has a crossed grating feature, what are the limitations here?
Since GSOLVER incorporates a full vector, rigorous diffraction efficiency solver it seemed natural to include the crossed grating case for completeness. In this case, the number of orders is quadratic in the Orders retained (there are two spatial degrees of freedom). Thus all the memory, and time requirements grow quadratically (very fast) with orders retained. This can be seen by setting up a crossed grating structure and looking at the estimate of the required memory as the orders parameter is increased. 7 orders retained already require more than 100MB of RAM. This suggest that crossed gratings with no more than 1 or 2 real propagating orders might be analyzed, depending on the strength of the evanescent fields supported by the grating interface. Even so, this represents the upper limit. I am aware of some who have used GSOLVER to analyze crossed grating structures that require much more RAM than this, and they have reported that theory and measurement were in good agreement. Howeve
