Any tips on how to get rid of these tiny imaginary frequencies? This structure was optimized at the b3lyp/cc-pvtz (spherical) level. The input geometry was had been optimized to a local minimum (no imaginary frequencies) at the same level in Gaussian 09.