Theoretical calculations of the potential energy surface (PES) for the [NH, + HCl] system are presented using several standard ab initio methods such as Hartree-Fock (HF), second-order Mdller-Plesset perturbation theory (MP2), coupled cluster (CC), complete active space self-consistent-field (CASSCF), density functional theory (DFT), and less traditional ab initio approaches such as Dirac-Fock four-components and the use of effective Hamiltonian techniques, such as the recently proposed K functional. All calculations predict a single minimum for the complex, corresponding to a hydrogenbonded structure, confirming early studies. The dynamical and nondynamical contributions to the correlation energy are discussed for different cuts of the PES, involving different N-Cl distances. The complex has also been characterized by performing a full geometry optimization within the HF and DFT schemes; with the latter we have performed also the vibrational analysis. The predicted binding energies and infrared (IR) spectrum are compared with other theoretical and experimental results. For the gas phase, we propose a binding energy of -5.3 f 0.5 kcal/mol, thus revising the experimental value of -8.0 k 2.8 kcal/mol; for the minimum, the predicted N-H and * is done with approximate inclusion of solvent effects (Onsager reaction field), the minimum is shifted and it corresponds to the ion pair NH:. C1-structure, similar to Mulliken's outer complex. Since the first ab initio computation for the NH,Cl complex is the pioneer work in 1967 by E. Clementi, the present work provides us with an opportunity to comment on some aspects of the evolution in computational chemistry, particularly for energy determinations. We have concluded our comments with the invitation to use four-components Fock-Dirac for molecules both with high and low Z atoms, rather than the traditional Hartree-Fock and related methods. In other words, we are of the opinion that the time is ready in quantum chemistry to switch from the Schrodinger to the Dirac representation, due to new developments in computer hardware and software. In addition, the use of effective Hamiltonians, like the recently proposed " K functional," seems to deserve attention, because of their computational simplicity and physical reliability in predicting correlation corrections. 0 1996 John Wiley & Sons, Inc.