this is for holding javascript data
Jacob Sanders edited Molecular Vibrations 2.tex
over 9 years ago
Commit id: 9c3fd1d6f643a919657c1d462b624b6072c46b0b
deletions | additions
diff --git a/Molecular Vibrations 2.tex b/Molecular Vibrations 2.tex
index 6264e9c..7f1a39a 100644
--- a/Molecular Vibrations 2.tex
+++ b/Molecular Vibrations 2.tex
...
Although for large systems the Hessian is expected to be sparse in While the
recovery of molecular
coordinate basis, for more moderately-sized molecules all mechanics Hessians provides a clear illustration of the
atoms scaling of our compressed sensing procedure, molecular mechanics Hessians are
relatively close already quite cheap to
each other so we cannot rely on the
Hessian being sparse. Instead, we pursue an alternative strategy for
sparsifying compute in the
Hessian, namely guessing first place. Since the
basis interactions between pairs of atoms are approximated via a set of
eigenvectors empirically-derived pair potentials, the energy derivatives in
which the Hessian
is diagonal. We believe this procedure may be directly computed as analytical derivatives of
estimating the eigenvectors and then recovering these pair potentials. Hence, from a computational standpoint, the
sparse matrix in
this approximate eigenbasis via real challenge is to apply our compressed sensing
will generalize procedure to
a wide range the computational of
scientific applications beyond molecular vibrations. quantum mechanical Hessians, and it is to this problem which we now turn.
As Fig.~\ref{fig:HessSparsity} shows, for moderately-sized molecules, the sparsity \(S\) of a quantum mechanical Hessian does not scale linearly with the number of atoms \(N\) in the molecule, although for large molecules this is once again expected to be the case~\cite{}. Instead, we must pursue an alternative strategy for sparsifying the Hessian, namely guessing the basis of eigenvectors in which the Hessian is diagonal. We believe this procedure of estimating the eigenvectors and then recovering the sparse matrix in this approximate eigenbasis via compressed sensing will generalize to a wide range of scientific applications beyond molecular vibrations.
For a quantum mechanical Hessian, our method for cheaply estimating the eigenvectors
involves is simply to use a
simple molecular mechanics
computation computation, which is essentially instantaneous for moderately-sized organic molecules.
Molecular mechanics computations approximate Diagonalizing the
interactions between pairs of atoms via a force field, which is a set
of empirically-derived pair potentials between all pairs of atoms in
the molecule. The energy derivatives which appear in the Hessian may
be computed as analytical derivatives of these pair potentials, and
diagonalizing this molecular mechanics Hessian yields a cheap approximation to eigenvectors of the true quantum mechanical
Hessian. Hessian, and hence the quantum mechanical Hessian is expected to be sparse in the basis of molecular mechanics normal modes.
Fig.~\ref{fig:HessianScheme} illustrates the viability of this procedure for
anthracene (\(\textrm{C}_{14}\textrm{H}_{10}\)), an aromatic molecule consisting the simple case of
three fused benzene
rings. (\(\textrm{C}_{6}\textrm{H}_{6}\)). The left side depicts the quantum mechanical Hessian
of anthracene in the basis of
molecular coordinates (\(H\)), atomic Cartesian coordinates, while the right side shows the same matrix in the approximate eigenbasis obtained via an auxiliary molecular mechanics
computation (\(A\)). The computation. As the figure shows, the matrix
\(A\) on the right in the basis of molecular mechanics normal modes is
nearly diagonal and quite sparse, much sparser than the matrix on the left, and is therefore well-suited to
recovery via our compressed sensing procedure. In particular, we expect compressed
sensing. sensing to recover the matrix on the right with far less sampling than would be required to recover the matrix on the left.