An electron repulsion integral compression algorithm

A data compression algorithm for packing/unpacking floating point numbers is presented. The method has been used to compress large volumes of data commonly generated in ab initio quantum mechanical calculations. To retain an accuracy of 10−6 Hartree on the final energy, the required file space needed is approximately half its original size whereas the CPU time required to solve the Hartree–Fock self‐consistent field equations increases with 30–60%. © 1993 John Wiley & Sons, Inc.