I'd like to find IRC path of some reaction on metal oxide catalyst. Interatomic distances in catalyst are known from EXAFS and XANES studies and I've obtained the simpliest cluster with structure that has good correlation with those studies. But when I tried to model molecular or dissociative adsorption, the cluster strongly distorted. Then I had to freeze several atoms ($ZMAT DLC=.t. AUTO=.t. IFRZAT(1)=... ) and I've obtained the minimum as well as saddle point of possible reaction being studied. But when I started IRC, it came to a minimum (another one will be obtained soon, I hope), but IFRZAT didn't work, so cluster was really distorted! So I'd like to know is there a way to freeze coordinates or atoms in $IRC or not?