Comment on "Crystal Structure and Pair Potentials: A Molecular-Dynamics Study"
Extracted source text ยท arXiv cond-mat/0306040
\documentclass[aps,twocolumn]{revtex4}
%\documentstyle[aps,twocolumn]{revtex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\addtolength{\topmargin}{0.4in}
\begin{document}
\title{Comment on ``Crystal Structure and Pair Potentials: A Molecular-Dynamics
Study''}
\author{Gang Liu}
\address{{\small High Performance Computing Virtual Laboratory}\\
{\small Queen's University, Kingston, ON K7L 3N6, Canada}}
\date{March 12, 2004}
\pacs{61.50.Ah, 62.20.-x, 64.10.+h.}
\maketitle
Parrinello~and Rahman\cite{pr} proposed a molecular dynamics (PRMD) for
crystal structure under external constant pressure in 1980. Many
simulations of structure phase transitions induced by external forces have
been carried out with it even till recently\cite{Mart,Rou,Mor,Tra}, especially after
it was combined with the well-known Car-Parrinello MD.
It uses the scaled particle position vectors ${{\bf s}_i}^T$ $%
=\left( \xi _i{\bf ,}\eta _i{\bf ,}\zeta _i\right) $ as generalized coordinates, defined by
\begin{equation}
{\bf r}_i=\xi _i{\bf a+}\eta _i{\bf b+}\zeta _i{\bf c=Hs}_i\text{, }
\label{c1}
\end{equation}
where ${\bf r}_i$ is the $i$th particle position vector, $\ {\bf a}${\bf , }$%
{\bf b}$ and{\bf \ }${\bf c}$ are the cell edge vectors, the matrix ${\bf H}%
=\left\{ {\bf a,b,c}\right\} $, and $T$ is the transpose.
It is obvious that using either ${\bf s}_i$ or ${\bf r}_i$ will
produce the same results in physics. It further uses
\begin{equation}
K^{\prime }=\text{ }\sum_i\frac 12m_i \dot {{\bf s}_i}^T{\bf H}^T{\bf H%
} \dot {{\bf s}_i} \label{c2}
\end{equation}
as the particles' kinetic energy, while the true kinetic energy for
particles is
\begin{eqnarray}
K &=&\text{ }\sum_i\frac 12m_i\dot {{\bf r}_i}^2 \nonumber
\\
&=&\text{ }\sum_i\frac 12m_i\left( \dot {{\bf s}_i}^T{\bf H}^T+{\bf s}%
_i^T\dot {\bf H}^T\right) \left( {\bf H}\dot {{\bf s}_i}+%
\dot {\bf H}{\bf s}_i\right) , \label{c3}
\end{eqnarray}
where $m_i$ is the $i$th particle's mass. The PRMD starts
from the Lagrangian $L$ introduced as
\begin{equation}
L=K^{\prime }-\Phi +L_{{\bf H}}, \label{c6}
\end{equation}
where $\Phi $ is the potential energy among particles and $L_{{\bf H}}$ is
the Lagrangian associated with $\dot {\bf H}$, ${\bf H}$, and
external pressure and $L_{{\bf H}}$ has no contribution to the particles'
dynamical equations. The aim of this comment is to check whether the chosen
kinetic energy is appropriate by comparing the dynamical equations for
particles with the Newton's Second Law
\begin{equation}
m_i\ddot {{\bf r}_i}={\bf F}_i\text{,} \label{c5}
\end{equation}
with the net force ${\bf F}_i$ acting on particle $i$. In order to interpret
the Eq.~(2) of the PRMD's original paper\cite{pr} accurately, especially the force
term, let us deduce it again in the following three case steps.
Case 1, from the Lagrangian dynamical equation, we can easily reproduce
Eq.~(\ref{c5}) by using ${\bf r}_i$ as ``generalized coordinates'' and using the
complete kinetic energy Eq.~(\ref{c3}) to replace $K^{\prime }$ in Eq.~(\ref
{c6}), with
\begin{equation}
{\bf F}_i=-\frac {\partial \Phi} {\partial {\bf r}_i} \text{.} \label{c7}
\end{equation}
Case 2, let us use ${\bf s}_i$ as generalized coordinates and still use the
complete kinetic energy of Eq.~(\ref{c3}). Combining Eq.~(\ref{c1}) and (\ref{c7}%
), we obtain $-\frac \partial {\partial {\bf s}_i}\Phi ={\bf H}^T{\bf F}_i$%
. By also using of $\frac \partial {\partial {\bf s}_i}K=m_i\dot {\bf %
H}^T\left( {\bf H}\dot {{\bf s}_i}+\dot {\bf H}{\bf s}%
_i\right) $ and $\frac \partial {\partial \dot {{\bf s}_i}}K=m_i{\bf H%
}^T\left( {\bf H}\dot {{\bf s}_i}+\dot{\bf H}{\bf s}_i\right)
$, the Lagrangian dynamical equation, $\frac d{dt}\frac \partial {\partial
\dot {{\bf s}_i}}L=\frac \partial {\partial {\bf s}_i}L$, becomes
\begin{equation}
m_i{\bf H}^T\left( {\bf H}\ddot {{\bf s}_i}+2\dot {\bf H}%
\dot {{\bf s}_i}+\ddot{\bf H}{\bf s}_i\right) ={\bf H}^T{\bf F%
}_i, \label{c13}
\end{equation}
which is equivalent to Eq.~(\ref{c5}).
Case 3, as in the PRMD, let us use ${\bf s}_i$ as generalized coordinates
and use the incomplete kinetic energy of Eq.~(\ref{c2}). Similar to the
above case, with $\frac \partial {\partial \dot {{\bf s}_i}}K^{\prime
}=m_i{\bf H}^T{\bf H}\dot {{\bf s}_i}$, the particle's dynamical
equation becomes
\begin{equation}
m_i\left( {\bf H}^T{\bf H}\ddot {{\bf s}_i}+{\bf H}^T\dot {\bf %
H}\dot {{\bf s}_i}+\dot{\bf H}^T{\bf H}\dot {{\bf s}_i}%
\right) ={\bf H}^T{\bf F}_i, \label{c15}
\end{equation}
which is the same as the Eq.~(2) of the PRMD's original paper\cite{pr}, but
different from Eq.~(\ref{c5}). The discrepancy can be found in between $%
\frac d{dt}\frac \partial {\partial \dot {{\bf s}_i}}\left( K-K^{\prime }\right)
$ and $\frac \partial {\partial {\bf s}_i}\left( K-K^{\prime }\right) $. So
the incomplete kinetic energy of Eq.~(\ref{c2}) is in-appropriate.
If the complete kinetic energy of Eq.~(\ref{c3}) is used, the first term on
the right-hand side of Eq.~(4) of the PRMD's original paper\cite{pr} will disappear
in physics. Anyhow, the
PRMD is correct for crystal structure when all accelerations and
velocities are zero.
\begin{references}
\bibitem{pr} M. Parrinello,~and A. Rahman, Phys. Rev. Lett. {\bf 45,} 1196
(1980).
\bibitem{Mart} R. Martonak, {\it et al.} , Phys. Rev. Lett. {\bf 84,} 682 (2000).
\bibitem{Rou} R. Rousseau, {\it et al.} , Phys. Rev. Lett. {\bf 85,} 1254 (2000).
\bibitem{Mor} T. Morishita, Phys. Rev. Lett. {\bf 87,} 105701 (2001).
\bibitem{Tra} A. Trave, {\it et al.}, Phys. Rev. Lett. {\bf 89,} 245504 (2002).
\end{references}
\end{document}