1\documentclass{article} 2\usepackage{doi} 3\usepackage{natbib} 4 5\title{Notes on density matrix functional theory} 6\author{Hubertus J. J. van Dam} 7 8\begin{document} 9\maketitle 10 11Nguyen-Dang~\cite{Nguyen_Dang_1985} device a scheme to optimize the 12ensemble and pure-state density matrices. Lude\tilde{n}a uses this paper to 13attack the work by Gilbert claiming that his "proper use of N-representability 14conditions" does not lead to degeneracies of natural orbitals. However, in the 15paper they insist that the pure-state density matrix is idempotent which in 16my view is incorrect. Hence the results are most likely invalid. 17 18Yasuda~\cite{Yasuda_2001} discussed an approach involving density and hole 19density matrices. It is shown that the correlation energy for the hole density 20matrix and the density matrix have to be the same. He proposes a functional in 21Eq.(22). However when examining practical results he uses full-CI 1-electron 22density matrices and other full-CI derived quantities. It is not clear how 23much the results depend on using data from full-CI but this does not seem like 24a practical approach. The paper also points to the importance of Levy's 25homogeneous scaling condition. 26 27Nagy~\cite{Nagy_2002} discussed an approach based on disjoint electron pairs. 28The paper is purely theoretical without any results. Hence the quality of the 29approach is impossible to judge. 30 31Kollmar~\cite{Kollmar_2003} discussed an approach with disjoint electron pairs. 32This will not work in general as the real wavefunction should incorporate all 33electron pairs. It does contain the normalization condition of the electron 34pair functions. 35 36Ayers~\cite{Ayers_2007} discusses N-representability conditions of density 37functionals. The definition is essentially that an N-representable density 38functional cannot produce an energy that is below the true variational 39minimum. This is obviously true but I am not sure how this can be enforced 40for that I would need to understand this work in more detail. 41 42Canc\`{e}s~\cite{Cance_s_2008} proposed a linesearch based algorithm to 43optimize the solutions for DMFT. This is pretty similar to the linesearch 44approach I tried except they don't use occupation functions of course. 45They also claim that most of the Baerends density matrix functionals are not 46proper 1-matrix functionals as they are formulated in terms of occupation 47numbers and natural orbitals but not explicitly in terms of density matrices. 48I am a bit puzzled about this. If one puts the Hartree-Fock wavefunction into 49the Schr\"{o}dinger equation the exchange terms also does not have a density 50matrix in it. However we clearly manage to rearrange the definition to get it 51into such a form (although this may only be true for idempotent density 52matrices). 53 54Rohr~\cite{Rohr_2010} discussed a scheme where a density and a density matrix 55functional are combined. They use range separation whereby the density 56functional can be applied for the short range part and the density matrix 57functional for the long range part. The reason for this approach is to avoid 58double counting of correlation effects and to account for the fact that 59the dynamic correlation is mostly short range and the static correlation long 60range. Nevertheless the accuracy does not seem to be great. Also they compare 61against semi-empirical results from 1974, how accurate can those be? 62 63Mazziotti~\cite{Mazziotti_2012,Mazziotti_2012_a} claims to have solved the 64density matrix N-representability conditions. He does so by building a 65hierarchy of N-representability conditions which couple different orders 66of density matrices. The downside is that fullCI equivalent results can be 67obtained only at a fullCI equivalent cost. Polynomial cost results can be 68obtained only by introducing approximations that break the hierarchy. 69 70Baldsiefen~\cite{Baldsiefen_2013} deviced an optimization scheme where he 71computes a matrix to diagonalize for the natural orbitals. However for the 72occupation numbers he uses a Fermi-smearing approach. This requires a 73temperature tensor which needs a sensible (but automated choice) of 74temperature to get it to work. This is weird and unnecesarrily complicated. 75 76Lude\tilde{n}a~\cite{Lude_a_2013} discusses at great length the 77N-representability of density functionals (which seems a bit of a strange 78concept). He also reiterates his objections against Gilbert's theorem although 79it still isn't clear to me why he gets different results. 80 81Lathiotakis~\cite{Lathiotakis_2014} claims that the orbital optimization does 82not reduce to an iterative eigenvalue problem (as in DFT) and hence is 83expensive. This is a direct contradiction to my experience! Otherwise they 84implement some OEP like approach. However, this introduces division by 85orbital energy differences which seems very dangerous to me as there are bound 86to be degeneracies. 87 88\section{Ideas} 89 90The one thing that still baffles me about density matrices is that somehow they 91have no knowledge of fundamental particles. Yes, there are occupation numbers 92and even limits on occupation numbers, but the density matrix gives no 93information about how those occupation numbers are added up from contributions 94of single particles. For the 1-matrix this is not such a problem as the 95occupation numbers are limited to \$0 \le n \le 1\$ but for 2-matrices the 96admissible range is (much) larger. I.e. a pair occupation number of \$1.2\$ 97might consist of \$1\$ pair plus \$0.2\$ of another pair, or it might consist 98of two \$0.6\$ contributions from two pairs. Within a density matrix 99formalism there is no way of knowing which one is right. 100 101In wavefunctions the number of elementary particles is obvious as each particle 102introduces a new set of coordinates. Hence I am thinking that maybe we need 103N-representability conditions based on orbitals and gemininals (orbitals have 104to be orthonormal due to the anti-symmetry, geminals have to be normalized). 105The orbitals and geminals should consist of occupation functions and natural 106orbitals/geminals to be able to generate physically sensible electron 107distributions while at the same time being able to impose conditions on 108individual 1-electron and electron pair states. Such an approach would 109eliminate the N-representability conditions that refer to occupation numbers 110and replace them with conditions on the wavefunctions. 111 112\bibliographystyle{plain} 113\bibliography{paper} 114\end{document} 115