Görling Andreas, Hesselmann Andreas, Jones Martin, Levy Mel
Lehrstuhl für Theoretische Chemie, Universität Erlangen-Nürnberg, Egerlandstr 3, Erlangen, Germany.
J Chem Phys. 2008 Mar 14;128(10):104104. doi: 10.1063/1.2826366.
Recently, Staroverov, Scuseria, and Davidson [J. Chem. Phys. 124, 141103 (2006)] presented examples of exchange-only optimized effective potential (xOEP) calculations that yield exactly the Hartree-Fock (HF) total energy. Here, building on their work, arguments showing under which conditions xOEP methods, with finite basis sets, do or do not yield the HF ground state energy but a higher one, are given. While the orbital products of a complete basis are linearly dependent, the HF ground state energy can only be obtained via a finite basis set xOEP scheme in the case that all products of occupied and unoccupied orbitals emerging from the employed orbital basis set are linearly independent of each other. Further, exchange potentials leading to the HF ground state energy likely exhibit unphysical oscillations and do not represent a Kohn-Sham (KS) exchange potential as a functional derivative of the exchange energy. These findings appear to explain the seemingly paradoxical results of Staroverov et al. that certain finite basis set xOEP calculations lead to the HF ground state energy despite the fact that within a real space (or complete basis) representation, the xOEP ground state energy is always higher than the HF energy. Moreover, independent of whether or not the occupied and unoccupied orbital products are linearly dependent, it is shown that finite basis set xOEP methods only represent exact exchange-only (EXX) KS methods, i.e., proper density-functional methods, if the orbital basis set and the auxiliary basis set representing the exchange potential are balanced to each other, i.e., if the orbital basis is comprehensive enough for a given auxiliary basis. Otherwise xOEP methods do not represent EXX KS methods and yield unphysical exchange potentials. The question whether a xOEP method properly represents a KS method with an exchange potential that is a functional derivative of the exchange energy is related to the problem of the definition of local multiplicative operators in finite basis representations. Plane wave calculations for bulk silicon illustrate the findings of this work.
最近,斯塔罗韦罗夫、斯库塞里亚和戴维森[《化学物理杂志》124, 141103 (2006)]给出了仅交换优化有效势(xOEP)计算的示例,这些计算能精确得出哈特里 - 福克(HF)总能量。在此,基于他们的工作,给出了一些论据,表明在有限基组的情况下,xOEP方法在何种条件下能得出或不能得出HF基态能量,而是得出更高的能量。虽然完备基的轨道乘积是线性相关的,但只有在所使用的轨道基组中出现的占据轨道和未占据轨道的所有乘积彼此线性独立的情况下,才能通过有限基组xOEP方案获得HF基态能量。此外,导致HF基态能量的交换势可能表现出非物理的振荡,并且不代表作为交换能泛函导数的科恩 - 沈(KS)交换势。这些发现似乎解释了斯塔罗韦罗夫等人看似矛盾的结果,即某些有限基组xOEP计算能得出HF基态能量,尽管在实空间(或完备基)表示中,xOEP基态能量总是高于HF能量。此外,无论占据轨道和未占据轨道的乘积是否线性相关,研究表明,如果表示交换势的轨道基组和辅助基组相互平衡,即如果对于给定的辅助基组,轨道基组足够完备,那么有限基组xOEP方法才仅代表精确的仅交换(EXX)KS方法,即恰当的密度泛函方法。否则,xOEP方法不代表EXX KS方法,并且会产生非物理的交换势。xOEP方法是否能恰当地代表具有作为交换能泛函导数的交换势的KS方法这一问题,与有限基表示中局部乘法算子的定义问题相关。体硅的平面波计算说明了这项工作的发现。