| 1 |
\documentclass[10pt]{article} |
| 2 |
|
| 3 |
\usepackage{amsmath} |
| 4 |
\usepackage{amssymb} |
| 5 |
|
| 6 |
\title{Manual for nemo and mp2 programs in MADNESS} |
| 7 |
|
| 8 |
\begin{document} |
| 9 |
|
| 10 |
\maketitle{} |
| 11 |
|
| 12 |
\section{Intro} |
| 13 |
The nemo program computes molecular HF and DFT energies and gradients using a nuclear correlation factor (ncf) as well as vibrational frequencies.\cite{Bischoff:2014du} |
| 14 |
In contrast, the moldft program does the same without using an ncf. |
| 15 |
|
| 16 |
The mp2 program computes the MP2 energies on top of either the nemo or the moldft HF reference wave function.\cite{Bischoff:2014ky,Bischoff:2013cx} |
| 17 |
|
| 18 |
\section{Input sections} |
| 19 |
\subsection{nemo} |
| 20 |
The nemo program uses the same input parameters as moldft, with only a few restrictions and extensions. |
| 21 |
In addition to the regular moldft input parameters there are the following options (all are optional) |
| 22 |
\begin{description} |
| 23 |
\item[nuclear\_corrfac:] [\textit{slater x, gaussslater, polynomialN, none}]\\ |
| 24 |
with $x>1$ recommended $x=2.0$, and $3<N<11$, recommended $N=8$. Default is "{}none"{}. |
| 25 |
\item[hessian] if present computes the hessian matrix and vibrational frequencies |
| 26 |
\item[read\_cphf] if present read in the solutions to the CPHF equations from a prior hessian calculation |
| 27 |
\end{description} |
| 28 |
There is no open-shell/spin-polarized version of nemo available. |
| 29 |
|
| 30 |
Note there is a inconsistency with the \texttt{eprec} keyword in the geometry section. |
| 31 |
Because there is no singular potential in the nemo calculation, the \texttt{eprec} keyword is used to smoothen the derivatives of the nuclear correlation factor (It is a step function). |
| 32 |
Typically a factor of 1e-3 to 1.e-4 will give you micro-Hartree accuracy with respect to the energy of the (regularized) nuclear potential, and about 1 cm-1 accuracy with respect to vibrational frequencies. |
| 33 |
|
| 34 |
\subsection{mp2} |
| 35 |
On top of the HF reference wave function (from nemo or moldft) a MP2 calculation may be performed. |
| 36 |
The relevant input is contained in the mp2 input section with the following keywords: |
| 37 |
\begin{description} |
| 38 |
\item[econv] \textit{val} energy convergence threshold for the correlation energy. Recommended is 1.e-3 or 1.e-4. |
| 39 |
\item[dconv] \textit{val} [optional] residual for the convergence of the MP1 wave function. Default is sqrt(econv)*0.1; |
| 40 |
\item[freeze] \textit{val} number of frozen core electrons. Default is 0. |
| 41 |
\item[maxsub] \textit{val} size of the subspace of the KAIN solver for the MP1 wave function. Default is 3. |
| 42 |
\end{description} |
| 43 |
Other options like k, L, or the nuclear correlation factor are taken from the reference input section. |
| 44 |
For 6D calculations the polynomial order k must be chosen small to keep calculations reasonably fast, i.e. k=5,6. |
| 45 |
The \texttt{econv} keyword of the 3D reference will be overriden by mp2 to be 100 times tighter than the 6D correlation. |
| 46 |
Local orbitals may be used. |
| 47 |
|
| 48 |
\section{Example input} |
| 49 |
dft\\ |
| 50 |
xc hf\\ |
| 51 |
canon\\ |
| 52 |
k 5\\ |
| 53 |
econv 1.00e-05\\ |
| 54 |
dconv 1.00e-04\\ |
| 55 |
maxiter 20\\ |
| 56 |
L 20\\ |
| 57 |
nuclear\_corrfac slater 2.0\\ |
| 58 |
end\\ |
| 59 |
\\ |
| 60 |
mp2\\ |
| 61 |
econv 1.00e-03\\ |
| 62 |
dconv 3.16e-03\\ |
| 63 |
maxsub 3\\ |
| 64 |
freeze 1\\ |
| 65 |
end\\ |
| 66 |
\\ |
| 67 |
geometry\\ |
| 68 |
eprec 1.e-3\\ |
| 69 |
O 0.0 0.0 0.0\\ |
| 70 |
H 1.4375 0.0 1.15\\ |
| 71 |
H -1.4375 0.0 1.15\\ |
| 72 |
end\\ |
| 73 |
|
| 74 |
\bibliographystyle{aip} |
| 75 |
\bibliography{references} |
| 76 |
|
| 77 |
|
| 78 |
\end{document} |
| 79 |
|