Polynomial or cross-product effects are polynomial regressor effects, defining a single parameter times the product of the values of the corresponding continuous factors. Using polynomial effects, you can construct models for the response which are not strictly linear in the factors, e.g., quadratic, cubic, etc.