Computational study on NiAl: ground state, structure, and spectroscopic constants using densityfunctional theory T. Cundari1, S.S. Janardan1 1University of North Texas, Department of Chemistry Abstract The least computational expensive Hamiltonian (methods) and wavefunction (basis sets) for diatomic nickel-aluminum (NiAl), was calculated, along with the ground state, bond length, frequency, and binding energy. Researchers believe nickel-aluminum can be used in high performance turbine jet engines, in place of aluminum-coated nickel. The effects of basis functions under increasing levels of theory were compared to methods under increasing levels of theory. The free energy at two possible ground states, doublet or quartet, of nickel-aluminum was compared. The results consistently show the ground-state of nickel-aluminum is the doublet, and therefore has a bond between two valence electrons, one from each nickel and aluminum, and one unpaired electron. The data converged when using large number of basis sets, but did not produce precise results when compared with different methods. Results suggest the density-functional theory (DFT) and an augmented correlation consistent basis set are needed, at minimum, to properly optimize nickel-aluminum. The least computationally expensive, most precise basis set/method combination for diatomic NiAl allows for further research in the least computationally expensive combination, of method and basis set, for large microclusters of NiAl, which could be used in engines as well. Keywords: NiAl; level of theory; Hamiltonian; wavefunction; ground-state I. Introduction Nickel-based superalloys are widely used in high-performance jet engines, and thus constitute an important area of research in material science, among both the experimental and modeling communities. These materials display extreme durability at high temperatures and possess the ability to precisely control their mechanical properties via the addition of many alloying metals. The jet engine application requires that alloys be both heat and corrosion resistant, and able to withstand creep and friction [1]. Nickel has traditionally been used in aerospace alloys [2], but it has a low oxidation resistance and requires a thermal barrier coating (TBC) [3]. Recent studies have combined nickel and aluminum (along with other minor components) using high-temperature synthesis [1]. The alloy NiAl has a heat resistance of 1640 ºC - 1677 ºC [1, 35] before the alloy starts to melt, an increase over traditional Ni alloys [1, 2]. Nickel-aluminum alloys display extremely high resistance to corrosion and oxidation compared to nickel [1, 3-5]. Recent calculations on molecular NiAl materials include a report by Oymak and Erkoc [6] using density functional theory (DFT) and effective core potential (ECP) methods. The geometries for AlkTilNim clusters were calculated for (k + l + m = 2, 3), examining geometries, vibrational spectra, dissociation, and electronic states. The binding energy of the diatomic NiAl, equilibrium interatomic separation, and stretching frequency were calculated to be 3.35 eV, 2.533 Å, and 263.5 cm-1, respectively [6]. NiAl was found to have greater binding energy as compared to diatomic AlTi and diatomic NiTi, and a smaller bond length and stretching frequency as compared to these same molecules. When NiAl was compared to all AlkTilNim clusters studied by Oymak and Erkoc, NiAl was found to have the largest highest occupied molecular orbital and lowest unoccupied molecular orbital (HUMO-LUMO) gap, typically used as an indicator of kinetic stability, as compared to AlTi and NiAl. More recent calculations by Fengyou et al. [7] found the bond length between nickel and aluminum in NiAl to be 2.19 Å and the binding energy to be 1.23 eV. Fengyou et al. used the B3PW91 density functional, as did Oymak and Erkoc [6], but Fengyou et al. used a relatively small LANL2DZ valence basis set, possibly accounting for the large difference in numbers between the two calculations. Out of all NiAl microclusters (combinations of 5, 13, and 52 nickel and aluminum molecules) analyzed, NiAl was found by Fengyou et al. to have the shortest bond length between the nickel and aluminum. The frequency of diatomic NiAl was calculated to be 374 cm-1, a large difference versus the calculations of Oymak et al. However, the frequency of NiAl was the highest among all nickelaluminum clusters tested by these researchers [7]. In a report published at approximately the same time, Belcher et al. [8] found the bond length of diatomic NiAl to be 2.20 Å, quite similar to the results of Fengyou et al. [7]. Belcher et al. [8] used plane wave DFTs and pseudopotential methods within the Vienna Ab-initio Simulation Package (VASP) code; AlX (X = transition metal) diatomics were compared, and NiAl was found to have the shortest bond length as compared with aluminum-gold (AlAu) and aluminum-copper (AlCu), but NiAl still had a smaller HUMOLUMO gap than AlAu and AlCu. However, the calculation for the binding energy of NiAl came to be -3.30 eV, showing a large gap in data when compared to other experiments [8]. The most recent calculations by Ouyang et al. [9] did use DFT, more specifically a hybrid density functional method B3LYP. The preferred basis set used here was the 6-311+G(2d) and the Lan12DZ basis set. The calculated bond length is similar to the calculation done by Oymak et al. [6] at 2.39 Å, but the basis set used is the same as Fengyou et al. [7]. The frequency was calculated to be 294 cm-1, similar to Oymak et al. [6] by have a large difference in numbers comparatively. Calculation of the binding energy yielded 2.24 eV, half way between previous calculations of Oymak and Erkoc [6] and Fengyou et al. [7]. The ground state of NiAl was also examined, and compared o aluminum-platinum (AlPt) and aluminum-palladium (AlPd). The ground state of NiAl was found to be very similar to AlPt and AlPd, although few specifics were discussed [9]. The data from previous research indicate that the wavefunctions/basis set used in NiAl calculations influences the results greatly. The effect of different basis set has yet to be carefully assessed for NiAl materials to produce the “best” basis set with respect to computational accuracy and cost. This is quite surprising in terms of the great interest in Ni-Al intermetallics, particularly given the popularity of utilizing first-principles calculations on small cluster to derive interatomic potentials that may be used to perform largerscale atomistic simulations. This paper is divided into three main parts. The first part discusses the different Hamiltonians/method (Ĥ), wavefunctions/basis sets (ψ), and computing methods used to III. Results and Discussion a. Ground State of NiAl Surprisingly, one issue that has not been fully addressed in the prior literature is the ground state of NiAl. Nickel has a triplet (3F) ground state and aluminum is a doublet (2P), so that the two most plausible 224/aug-cc-pVQZ 163/cc-pVQZ 143/aug-cc-pVTZ 102/cc-pVTZ 60/VTZ 47/TZV 40/3-21G 42/pVDZ 35 30 25 20 15 10 5 0 34/VDZ II. Computational Methods Calculations employed the Gaussian 03 package. For each basis set and method, four calculations were performed: Ni (triplet ground state), Al (doublet ground state), NiAl doublet, and NiAl quartet. The main methodology used in the present calculations was density functional theory (DFT). However, wavefunction-based 2nd order Møller–Plesset perturbation theory (MP2), Hartree–Fock (HF), and coupled cluster (CC) techniques were also used to compare to and contrast with DFT. Multiple basis sets and Hamiltonians were used in assessing the “best” level of theory, looking for convergence of properties for diatomic NiAl. The HF method provides a starting point for the postHartree-Fock methods, viz CC and MP2 methods. Two CC methods were evaluated for the calculations on NiAl: coupled cluster with single and double excitations (CCSD) and coupled cluster with single, double and perturbative triple excitations (CCSD(T)). The latter proved to be too expensive using the extended basis sets needed to describe NiAl and hence this work focuses on the CCSD methodology. Two DFT methods were extensively evaluated with different basis sets used for the calculations of NiAl: the Becke, three-parameter, Lee-Yang-Parr exchange-correlation functional (B3LYP) and B971, the latter being the most popular in molecular calculations on transition metal compounds and the former being evaluated as superior with respect to thermodynamic predictions in a previous survey of 3d transition metal complexes and 4? functionals. As with increasing sophistication in the Hamiltonians, the basis sets may be enhanced in complexity until convergence in properties is seen. The quality of atom-centered basis sets for molecular calculations are, in general, directly correlated with the number of contracted Gaussian-type orbitals (cGTOS). For example, the 3-21G split-valence basis set is one of the smallest basis set evaluated here, comprising only 40 cGTOS for NiAl. The VDZ[10] and VTZ[10] basis sets possess 34 and 60 cGTOS, respectively. To add more flexibility, a polarized double-zeta basis set (pVDZ)[10], 42 cGTOS was evaluated. The last of the computational inexpensive basis sets utilized in this research is the TZV[11], which is a split-valence basis set and has 47 cGTOS for NiAl. One issue that has traditionally plagued transition metal computational chemistry is the absence of basis sets that have been methodically improved to allow one to more smoothly approach the basis set limits for the calculation of various molecular properties. Correlation consistent (cc) basis sets were originally developed by Dunning and coworkers for main group elements and recently extended to transition metals by Peterson et al. The present calculations used cc-pVTZ, cc-pVQZ, cc-pV5Z basis sets, with 102, 163, and 248 cGTOS, respectively, for NiAl. Augmenting these further by the addition of diffuse functions to the correlation consistent basis sets creates the largest basis sets used in the present calculations: aug-cc-pVTZ, aug-cc-pVQZ, aug-cc-pV5Z), with 143, 224, and 333 cGTOS, respectively for the diatomic. Diatomic NiAl was geometry optimized, and its equilibrium molecular properties calculated at the aforementioned levels of theory and using both (B3LYP and B971) DFT methods. The NiAl diatomic was studied both in the quartet and doublet spin states for each level of theory, along with nickel (triplet) and aluminum (doublet) separately. ground state multiplicities for NiAl arises from high-spin (4NiAl) and low-spin (2NiAl) coupling of the constituent atoms. We note that in these calculations that no symmetry restrictions were placed on the wavefunction and all simulations were spin-polarized (unrestricted). No evidence of significant spin contamination was observed in the calculations. Doublet-Quartet Gap (Kcal/mol) calculate the properties of NiAl. The second part analyzes the results of the calculations in relation to the difference in the lowest energy doublet and quartet states (GDQ), equilibrium bond lengths (re), equilibrium stretching frequency (e), and binding energy (De) for different levels of theory. The last part contains the conclusions. Number of Basis Functions Figure 1. Doublet vs. quartet free energy difference (a.u.) in diatomic NiAl. The number of contracted Gaussians is denoted along with the basis set name. The B3LYP functional is used for these simulations. The calculation of the relative energy of the most stable doublet and quartet spin states states of NiAl produced consistent results for nearly all levels of theory, apart from the small 3-21G basis sets, as seen in Figure 1. Doublet NiAl consistently yielded a lower free energy than quartet NiAl. When GDQ is analyzed for the B3LYP method and all basis sets, convergence to a free energy difference of ca. 25½ kcal/mol (~1.1 eV) is seen, with the doublet state always being lower in free energy. The aug-cc-pVQZ basis set had a value of 25.6187 kcal/mol for the doublet quartet split. The DFT calculations thus point to the most probable spin state of NiAl as a doublet and this suggests a significant degree of Ni-Al covalent bonding via spin pairing, Scheme 1. Ni Ni (3F) + 4 NiAl Al Al (2P) Ni Al 2 NiAl Scheme 1 b. Equilibrium Bond Length Structural data remains one of the most fundamental parameters that is either incorporated into empirical potentials or which is required to be faithfully reproduced by said potentials. The NiAl equilibrium bond length (re) was calculated for each basis set and both the B3LYP and B97-1 functionals. Similar to the convergence patterns for basis sets and methods, re was seen to converge at higher theory level basis sets for both DFT methods, Figure 2. However, while re convergence required a large basis set, Figure 2, it did not require the largest basis sets evaluated. For consistency, however, the calculated re with the largest basis set - aug-cc-pVQZ – and the B3LYP functional = 2.3744Å. This result is comparable with other 224/AUG-CC-pVQZ 163/CC-pVQZ 143/AUG-CC-pVTZ 102/CC-pVTZ 60/VTZ 47/TZV 42/pVDZ 40/3-21g 500 450 400 350 300 250 200 150 100 50 0 34/VDZ Fundamental Frequency (cm-1) Fig. 2 Bond length (Å) versus Basis Functions in B3LYP Number of Basis Functions Fig. 3 Fundamental Frequency versus Basis Functions in B3LYP c. Equilibrium Stretching Frequency The energetic costs of distorting atoms from their equilibrium positions, whether it is represented by the elastic constants or phonon modes of a solid material or the vibrational stretching frequencies of a molecular entity, is one of the most important quantities in developing and utilizing empirical potentials. Such fundamental quantities necessarily impact and inform material properties such as bulk, shear and Young’s moduli. The equilibrium stretching frequency (e), was calculated for all basis sets and both DFT methods. The basis set convergence patterns of the frequencies matched all properties discussed previously, Fig. 3. The calculated e, akin to re, was less basis set dependent than the free energy. Using the largest aug-cc-pVQZ basis set and B3LYP, the frequency was calculated to be 283 cm-1. This theoretical value is not comparable to previous research seen, since the other values consisted of 263.5 cm-1 [6], 374 cm-1 [7], and 294 cm-1 [9]. d. Binding Energy 224/AUG-CC-pVQZ 163/CC-pVQZ 143/AUG-CC-pVTZ 102/CC-pVTZ 60/VTZ 47/TZV 42/pVDZ 34/VDZ Binding Energy (Kcal/mol) 224/AUG-CC-pVQZ 163/CC-pVQZ 143/AUG-CC-pVTZ 102/CC-pVTZ 60/VTZ 47/TZV 42/pVDZ 40/3-21g Number of Basis Functions 40/3-21g Number of Basis Functions 2.45 2.4 2.35 2.3 2.25 2.2 2.15 2.1 2.05 2 1.95 1.9 34/VDZ Bond Length (Å) reports, which suggested re = 2.533 Å[6] and 2.3875 Å [9]. 0 -10 -20 -30 -40 -50 -60 -70 -80 -90 Fig. 6 Binding Energy versus Basis Functions in B3LYP Each method and basis set was calculated four times, once each with Ni, Al, NiAl-quartet and Ni-doublet. Based on the equation Ni + Al NiAl, the binding energy can be calculated from the free energy results. Due to errors in the computing process very few basis sets had all four calculations run, and therefore produce the calculation for binding energy. As with bond energy and frequency, the convergence pattern remained the same when comparing binding energy across basis sets and methods, fig. 6. Similar to the bond length and frequency, the convergence of the binging energy comes to around 102 basis sets. Although, not enough data was found to accurately claim this. The final result for binding energy was calculated using the AUG-cc-pVQZ basis set and DFT method, producing -44.2 (kcal/mol). IV. Conclusion The most efficient basis set was determined to be AUG-CCpVQZ because of the high accuracy to computational cost ratio. The best method was determined to be density-functional theory, because the other methods did not have accurate enough results. The ground state of NiAl was determined to be a doublet, with one electron from nickel and aluminum forming a magnetic bond, leaving one electron free. Further exploration of the doublet NiAl revealed the bond length, frequency, and binding energy to be 2.37Å, 283 cm-1, and 44.2 (kcal/mol) respectively. The most efficient basis set and method were only determined for diatomic NiAl. A common basis set between all variations of NiAl still needs further research. Small microclusters in ratios of 3:1 nickel to aluminum can be built up in multiples and evaluated to find a common basis set and method. The common basis set for diatomic NiAl can also be used when researching further into the properties of NiAl (e.g. ductile-to-brittle transition temperatures). V. Acknowledgements The authors thank the Center for Advanced Scientific Computing and Modeling (CASCaM) for providing computing resources (NSF grant CHE-??????), and Dr. Nathan J. DeYonker, formerly of the UNT Chemistry Department, for his assistance. The authors acknowledge financial support from the Air Force Research Laboratory (Contract No. FA8650-08-C-5226). The authors also thank the Texas Academy of Mathematics and Science for opportunities and scholarships. VI. References [1] Pascal, C., Marin-Ayral, R. M., Tédenac, J.C., & Merlet, C. (2003). Combustion synthesis: a new route for repair of gas turbine components—principles and metallurgical structure in the NiAl/RBD61/superalloy junction. Materials Science and Engineering A, 341, 144-151. [2] Johnson, B. J., Kennedy, F. E., & Baker, I. (1996). Dry sliding wear of NiAl. Wear. 192, 241-247. [3] Guo, H., Sun, L., Li, H., & Gong, S. (2008). High temperature oxidation behavior of hafnium modified NiAl bond coat in EB-PVD thermal barrier coating system. Thin Solid Films, 516, 5732–5735. [4] Darolia, R. (2000).Ductility and fracture toughness issues related to implementation of NiAl for gas turbine applications. Intermetallics. 8, 1321-1327. [5] Pomeroy, M. J. (2005).Coatings for gas turbine materials and long term stability issues. Materials and Design. 26, 223-231. [6] Oymak, H., & Erkoc, S (2002). Structural and electronic properties of Alk Til Nim microclusters: Densityfunctional-theory calculations. Physical Review A. 66, 033202. [7] Fengyou, H., Yongfang, Z., Xinying,, L., & Fengli, L. (2007). A density-functional study of nickel/aluminum microclusters. Journal of Molecular Structure: Theochem, 807, 153–158. [8] Belcher, D. R., Radny, M. W., & King, B. V. (2007). Structure and stability of small bimetallic Al-based clusters: An ab initio DFT study. School of Mathematical and Physical Sciences, The University of Newcastle, Callaghan, Australia. Materials Transactions, 48(4), 689-692. [0] Rey, C., Garcıa-Rodeja, J., & Gallego, L. J. (1996). Computer simulation of the ground-state atomic configurations of Ni- Al clusters using the embedded-atom model. Physical Review B. 54, 2942-2948. [9] Ouyang, Y., Wang, J., Hou, Y., Zhong, X., Du, Y., & Feng, Y. (2008). First principle study of AlX (X=3d, 4d, 5d elements and Lu) dimer. Journal of chemical physics. 128, 074305. [10] A. Schafer, H. Horn and R. Ahlrichs, J. Chem. Phys. 97, 2571 (1992). [11] A. Schafer, C. Huber and R. Ahlrichs,J. Chem. Phys. 100, 5829 (1994). Bailey, M. S., Wilson, N. T., Wilson, C., & Johnston, R. L. (2003). Structures, stabilities and ordering in Ni-Al nanoalloy clusters. Eur. Phys. J. D, 25, 41-55. Krajci, M., & Hafner, J (2002). Covalent bonding and bandgap formation in transition-metal aluminides: di-aluminides of group VIII transition metals. Journal of Physics: Condensed Matter. 14, 5755–5783. Reddy, B. V., Deevi, S. C., Lilly, A. C., & Jena, P (2001). Electronic structure of sub-stoichiometric iron aluminide clusters. Journal of Physics: Condensed Matter, 13, 8363–8373. Reddy, B. V., Sastry, D. H., Deevi, S. C., & Khanna, S. N. (2001). Magnetic coupling and site occupancy of impurities in Fe3Al. Physical Review B. 64, 224419. Rexer, E. F., Jellinek, J., Krissinel, E. B., Parks, E. K., & Riley, S. J. (2002). Theoretical and experimental studies of the structures of 12-, 13-, and 14-atom bimetallic nickel/aluminum clusters. Journal of Chemical Physics. 117, 82-94