J. Phys. Chem. zyxwvuts 1995,99, zyxwvu 4935-4940 4935 Density Functional Study of Iron Bound to Ammonia Joice Terra and Diana Guenzburger* Centro Brasileiro de Pesquisas Fisicas, Rua Dr. Xavier Sigaud, zyxwvu 150, 22290-180 Rio de Janeiro, RJ Brazil Received: April 29, 1994; In zyxwvutsrqp Final Form: January 17, 1999 Density functional calculations were performed to investigate the species formed by the interaction of an Fe atom and ammonia. The discrete variational method was employed, and total energy calculations were performed for several configurations. It was found that the ground state is a 5E, with Fe configuration -3d6.6 4sI.l as obtained in a Mulliken-type population analysis; the Fe-N interatomic distance was determined to be 1.98 A. The hyperfine parameters isomer shift, quadrupole splitting and magnetic hyperfine field were also calculated and compared to reported experimental values obtained by Mossbauer spectroscopy in frozen ammonia. I. Introduction Transition metals may absorb ammonia strongly on their surfaces, even at room temperature. The heterogeneous catalytic synthesis and decomposition of ammonia is a subject of great technological importance.’ Thus an understanding of the bonding mechanism between transition metals and ammonia is desirable. On the other hand, the technique of isolation of atoms and small molecules in frozen gases allows the use of solid- state experiments such as Mossbauer spectroscopy to probe charge and spin distributions. Accordingly, an investigation of Fe isolated in solid ammonia has been reported; the reaction product FeNH3 was identified, and Mossbauer hyperfine parameters were measured.* However, it became evident that quantum chemical calculations would be needed to better understand the origin of the values obtained. In this work, we report density functional theory (DFT) calculations3 for the species FeNH3. We used the discrete variational method (DVM).4 We determine the ground state by performing total energy calculations for several electronic configurations. The charge and spin distributions are analyzed. Finally, the Mossbauer hyperfine parameters isomer shift (d), quadrupole splitting (AEQ), and components of the magnetic hyperfine field (HF) are calculated and compared to experiment. This paper is organized as follows: in section 2 we briefly describe the theoretical method, in section 3 we discuss the electronic structure, in section 4 we report results for the hyperfine parameters, and in section zyxwvuts 5 we summarize our conclusions. 11. Theoretical Method The DVM method employed has been described in the original literat~re,~ here we give a summary and some details of the calculations. The Kohn-Sham equations of DFT3 are solved in a three-dimensional grid of points. The local exchange-correlation potential employed was that derived by von Barth and HedinG5 In these spin-polarized calculations, appropriate for an open-shell molecule, the electron density of spin up is allowed to be different from spin down. The molecular density for each spin is obtained from the sum of the squared amplitudes of the molecular orbitals, which in turn are expansions on a basis of numerical atomic orbitals. According to the DVM scheme, secular equations are obtained, * Author to zyxwvutsrqp whom correspondence should be addressed. @ Abstract published in Advcince ACS Ahsrracrs, March 15, 1995 0022-365419512099-4935$09.00/0 0 which are solved self-consistently until a desired criterion is met. In the present calculations, convergence was carried out to < low4 in the charge and spin densities. To calculate the Coulomb potential by one-dimensional integrations, the molecular charge density eO is fitted to a multicenter multipolar expansion:6 I j v m The summation is over a set I of atoms equivalent by symmetry, RN are piecewise parabolic radial functions centered at atoms Y and I distinguishes different basis functions of a given 1 (j = I, 1, zyxwvuts A, N). This expansion may be carried out to any degree of accuracy in the fit with the “true” density; in the present calculations, partial waves up to 1 = 2 were employed for Fe and N, and 1 = 1 for H; the least-squares error of the fit of e was -0.04. The total energy E is defined as the expectation vaJue (sum over integration mesh) of the energy density eC;,{R,}). To control numerical errors, the actual computation of E is made by point-by-point subtraction of a reference system of nonin- - teracting (NI) atoms, as in the basis, located at the nuclear sites R,: Here we employed the algorithms of Delley and Ellis’ to calculate E. The variational basis set included all inner orbitals, e.g., no “frozen core” approximation was made. The valence functions included 3d, 4s, and 4p on Fe, 2s, 2p, and 3d on N, and Is, 2s, and 2p on H. The self-consistent process was initiated with basis functions from the neutral atoms. After convergence, the basis was improved by generating atomic functions for atoms with configurations similar to those of the molecule. These are obtained by a Mulliken-type population analysis, in which the overlap population is divided proportionally to the coefficients of the atoms.* This basis optimization process is repeated several times, until the configuration of the atoms in the molecule is approximately the same as in the basis. The three-dimensional grid of points was divided into two regions: the volume inside spheres placed around each nucleus, where precise polynomial integrations are performed on a regular grid,9and elsewhere in space, where the pseudorandom Diophantine point generator is used.4 The spheres had radii 1995 American Chemical Society