A method for efficient localization and description of stationary points on the potential energy surface of extended systems is presented. It is based on Warshel's empirical valence bond approach, for which we propose a modification, and combines the potential function description of the total system with a quantum mechanical description of the reaction site ͑QM-Pot͒. We describe the implementation of the method in the QMPOT program, which is basically an optimizer for minima and saddle points and has interfaces to existing quantum mechanical ͑e.g., TURBOMOLE, GAUSSIAN94͒ and interatomic potential function codes ͑e.g., GULP, DISCOVER͒. The power of the method is demonstrated for proton transfer reactions in zeolite catalysts, which may have as many as 289 atoms in the unit cell. As a test case the zeolite chabazite is considered in this study. Its limited unit cell size ͑37 atoms͒ makes comparison with the full periodic ab initio limit possible. The inclusion of long-range effects due to the periodic crystal structure by the QM-Pot method proves crucial in obtaining reliable results. The combined quantum mechanics-interatomic potential function calculations yield reaction barriers within 6 kJ/mol and reaction energies within 3.5 kJ/mol of the periodic ab initio limit. The zero-point vibrational energy corrected reaction barriers are between 58 and 97 kJ/mol for the six different proton jump paths. These are density functional results employing the B3LYP functional.