this is for holding javascript data
Jacob Sanders edited Molecular Vibrations 2.tex
over 9 years ago
Commit id: 7d41c327c674979cde038d843fed56074a04ea92
deletions | additions
diff --git a/Molecular Vibrations 2.tex b/Molecular Vibrations 2.tex
index aaa9391..427692f 100644
--- a/Molecular Vibrations 2.tex
+++ b/Molecular Vibrations 2.tex
...
As a final illustration, We must discuss one more aspect of computing quantum mechanical Hessian matrices before we
can outline the entire
procedure procedure, namely that they can be computed an entire \emph{column} at a time, rather than element-by-element. The second derivatives required for the Hessian can be calculated analytically via perturbation theory or numerically via finite differences~\cite{Pulay1969,Jo_rgensen_1983,Baroni2001,Kadantsev2005}. We can write the second derivative of the energy as a first derivative of the force
\begin{equation}
\label{eq:force}
\frac{\partial^2E(\vec{R}^1,\ldots,\vec{R}^N)}{\partial R^A_i \partial R^B_j} =
- \frac{\partial F^B_j(\vec{R}^1,\ldots,\vec{R}^N)}{\partial R^A_i} \ .
\end{equation}
By the Hellman-Feynman theorem~\cite{Feynman1939}, a single energy calculation yields the forces acting over \emph{all} atoms, so the evaluation of eq.~\ref{eq:force} by finite differences for fixed \(A\) and
perform it \(i\) yields the derivatives for all values of \(B\) and \(j\), a whole column of the Hessian. An equivalent result holds for analytic derivatives obtained via perturbation theory~\cite{Baroni1994,Kadantsev2005}. Thus, our compressed sensing procedure focuses on
anthracene (\(\textrm{C}_{14}\textrm{H}_{10}\)), measuring random columns of the quantum mechanical Hessian rather than individual random entries.
A key aspect of compressed sensing is that measurements must be taken in a
moderately-sized polyacene consisting of three linearly fused benzene rings, basis that is as incoherent as possible with respect to
show the sparse basis. Thus, we want to compute columns of the quantum mechanical Hessian in
practice what kinds the basis that is the discrete cosine transform (DCT) of
results the molecular mechanics normal modes. In principle, computing each column in the DCT basis is no more expensive than computing columns in the traditional basis of atomic coordinates, and
speed-ups may be obtained. The the computational advantage stems from the fact that substantial undersampling of columns is possible in the DCT basis.
Thus, the full compressed sensing procedure
which must to be implemented is
as follows: the following:
\begin{enumerate}
\item Calculate approximate vibrational modes using molecular mechanics or another relatively cheap method.
...
\item Diagonalize the quantum mechanical Hessian to obtain the vibrational modes and frequencies.
\end{enumerate}
As
discussed above, a
key aspect of final illustration, we outline the
compressed sensing entire procedure
is incoherent sampling. The key step which must be implemented in the quantum mechanical code to enable incoherent sampling is step 4. In particular, columns of the true quantum mechanical Hessian must be computed in the basis that is the discrete cosine transform and perform it on anthracene (\(\textrm{C}_{14}\textrm{H}_{10}\)), a moderately-sized polyacene consisting of
the molecular mechanics normal modes. In principle, computing each column is no more expensive than computing columns three linearly fused benzene rings, to show in
the traditional basis practice what kinds of
atomic coordinates, results and
the computational advantage stems from the fact that substantial undersampling of columns is possible in this discrete cosine transform basis. speed-ups may be obtained.
Figs.~\ref{fig:AnthraceneFreqError} and \ref{fig:AnthraceneNormModeError} illustrate Fig.~\ref{fig:Anthracene} illustrates the results of applying our Hessian recovery procedure to
anthracene. Fig.~\ref{fig:AnthraceneFreqError} anthracene (\(\textrm{C}_{14}\textrm{H}_{10}\)), a moderately-sized polyacene consisting of three linearly fused benzene rings. The middle panel illustrates the error in the vibrational frequencies obtained by our compressed sensing procedure for different extents of undersampling. Sampling only 30\% of the columns gives rise to a maximum frequency error of less than 3 cm\(^{-1}\).
Fig.~\ref{fig:AnthraceneNormModeError} The bottom panel illustrates the error in the vibrational normal modes; this error is calculated as one minus the overlap (dot product) between the exact quantum mechanical normal mode and the normal mode recovered via compressed sensing. Once again, sampling only 30\% of the columns gives accurate recovery of all vibrational normal modes to within 1\%. In short, for the specific case of anthracene, the compressed sensing procedure we have devised reduces the number of expensive quantum mechanical computations by a factor of three, resulting in a 3x speed-up.