Academia.eduAcademia.edu

A Hohenberg-Kohn theorem for non-local potentials

1977, Physics Letters A

It is shown that within any class of commuting one-body potentials a Hohenberg-Kohn type theorem is satisfied with respect to an appropriately defined density. The Hohenberg-Kohn theorem for local potentials follows as a special case.

Volume 61A, number 1 PHYSICS LETTERS 4 Apr11 1977 A HOHENBERG-KOHN THEOREM FOR NON-LOCAL POTENTIALS Ehud MERON and Jacob KATRIEL Department of Chemistry, Tech nion-Israel Institute of Technology, Haifa 32000, Israel Received 18 February 1977 It is shown that within any class of commuting one-body potentials a Hohenberg—Kohn type theorem is satisfied with respect to an appropriately defined density. The Hohenberg—Kohn theorem for local potentials follows as a special case. The Hohenberg—Kohn theorem [1] was originally formulated for a Hamiltonian of the form ~ j=1 ‘ j<~j r~1 i1 V(r1), (1) V(r) is a local one-body potential. The theorem asserts that the local potential V(r) is a unique functional of the ground-state density, p(r), provided that the ground state is non-degenerate. The generalization to any Hamiltonian of the form where H=H0 (2) + ~ where H0 is any Hermitian operator common to all Hamiltonians of interest, presents no difficulties whatsoever. Inclusion of one-particle spin operators in V(r) requires a minor modification of the theorem, so as to refer to the density p (x), where x (r, s) contains the spin in addition to the spatial coordinates. However, the local nature of the one-body potential appears as a crucial requirement in the original derivation of the theorem. The following properties characterize the class of systems discussed: (a) All local potentials mutually commute, an immediate implication being that they have a common set of eigenstates 1r0), which, in coordinate space, are represented by the Dirac-delta functions ~(r r0). (b) The ground state density operator 3 can be defined as follows [2] ~Ø(r)=fp(r,r’)Ø(r’)dr’, (3) where the kernel p (r, r’) is the ground state first-order density matrix. The ground-state density p(r) is the diagonal element of the density operator in the s-function basis, (rI~lr). The properties (a) and (b) motivate a generalization of the Hohenberg—Kohn theorem to non-local potentials, along the following lines: (a’) Consider a class X of Hamiltonians H(s) = H0 + ~ V(~)(i), such that [V(S), V(t)] = 0 for any s and t belonging to the class. The ground state wavefunction and energy will be denoted by i,1,(~)and E(s), respectively. Let {Xk} be the set of common eigenfunctions of all these commuting potentials. Any potential in the class is specified by the set of eigenvalues X~defined by 19 Volume 61A, number 1 PHYSICS LETTERS 4 April 1977 (5) (b’) We define the ground-state density in the basis ~Xk}as p(~)(k)= (xkI~~~Ixk). (6) Theorem: Any potential V(s) which belongs to the class X is a unique functional of the corresponding ground state density p(~)(k),provided that the ground state is non-degenerate. The demonstration of this theorem depends on the fact that any N-particle wavefunction i~(x1 XN) can be written in terms of products ofN functions out of the set {Xk~[3], symmetrized or antisymmetrized according to the appropriate Statistics, (7) K where K = (k1, k2, kN) and 0(K) = fl~iXk~(Xj). Specifically, ~,(s) = Here, and in the remainder of the present communication, the symbol ~k designates summation over the discrete spectrum and integration over the continuum. The ground state density matrix can be written in terms of the set as ... p(s)(x’,x)=Nfo(s)*(x’,x2, dxN ...XN)l/1~~)(X,X2,...XN)dX2... =N~ C(s)*(K).C(s)(L)xr(x~)xl1(x)n<Xk~Ixl~). (8) The density p(~)(k)is, therefore, 2 (9) . p(~)(k) (xkLP(5)!xk)Nk~k c(~)(k,k2, k3, ...k~)! Furthermore, N N ~II~ V(~)(i)I,Li) i=1 = = ~I~C*(K)C(L) (Ø(K)I ~ V(~)(i)lO(L) KL ~ C*(K)C(L)( ~ i=1 (Ø(K)~(L))= ~ IC(L)l~~ xv). (10) From the indistinguishability of the particles it follows that (01 V~(i)I0)=N~XilC(L)l2= ~X 1p(l), (11) which reduces for local potentials to (0l~V1kD)=fV(x)p(x)dx. The generalized theorem can now be proved, following the derivation of the original theorem. We write 20 (12) Volume 61A, number 1 E(t) .~ 0(5)IH(t)IiP(5)) = Et~~+ PHYSICS LETTERS [V(t)(j) V(s)(j)] I — = E(S) + 4 April 1977 ~ [Xv) x~~~Jp(~)(l) — (13) and E(~)<E(t)+ ~ [X,,~—X~t)].p(t)(i). Now, assuming that p(t)(i) = p(5)(l), we obtain Ett) + E(~)<E(t) + E(~),a reductio ad absurdum which proves the theorem. The generalization of the Hohenberg—Kohn theorem here proposed may have applications of one kind or another. Its immediate attraction, we feel, is due to the deeper insight it provides into the nature of the original theorem. The two-fold role of the spatial coordinates in the local-potential context, i.e. their being both dynamical variables and quantum numbers specifying the eigenfunctions of the potentials, is explicitly avoided in the generalized context. This is a major contributing factor to the transparency of the generalized formulation of the Hohenberg— Kohn theorem. One of the authors (J.K.) is grateful to Prof. R.G. Parr for having inspired his interest in the Hohenberg—Kohn theorem, as well as for his kind hospitality during a visit to Chapel—Hill. That visit and the present study are parts of a research project supported by the U.S. Israel Bi-national Science Foundation. References [11 P. Hohenberg and W. Kohn, Phys. Rev. B136 (1964) 864. [2] R. McWeeny and B.T. Sutcliffe, Methods of molecular quantum mechanics (Academic Press, London, 1969). [31 P.O. Ldwdin, Phys. Rev. 97 (1955) 1474. 21