Abstract The routing and spectrum assignment problem in modern optical networks is an NP‐hard problem that has received increasing attention during the last years. The majority of existing integer linear programming models for the problem uses edge‐path formulations where variables are associated with all possible routing paths so that the number of variables grows exponentially with the size of the instance. To bypass this difficulty, precomputed subsets of all possible paths per demand are typically used, which cannot guarantee optimality of the solutions in general. Our contribution is to provide a framework for the use of edge‐path formulations to minimize the spectrum width of a solution. For that, we select an appropriate subset of paths to operate on with the help of combinatorial properties in such a way that optimality of the solution can be guaranteed. Computational results indicate that our approach is indeed promising to solve the routing and spectrum assignment problem.