44723

Convergence of the transfer-integral technique in the framework of Peyrard-Bishop model

Favorite this paper

The Peyrard-Bishop model is a simplified physical statistics method proposed in 1989 for calculating key properties of DNA and RNA molecules.
Currently, it is widely used with several theoretical techniques such as molecular dynamics and path integrals, but its main
application still lies with its original method the transfer-integral technique.
It also used to predict hydrogen bonds and stacking interaction in oligonucleotides from melting temperature experiments, complementing experimental
techniques such as NRM and X-ray diffraction.
In 1997 Zhang et al [Phys. Rev. E 56, 7100] showed that due to the form of the DNA stacking interaction
potential the integral equation,
which is central to the method, suffers from a numerical convergence problem which leads to a divergence in the partition function.
The workaround to this divergence problem is either to add a fixed upper limit to the integral equation or to add specific terms to
the model Hamiltonian. These additional terms could be an angle dependence or a stress term as suggested by some authors.
These workarounds have allowed the development of the Peyrard-Bishop model into a highly successful method, still a clean solution to the divergence problem
would be highly desirable.
Recently Estrada et al [J. Comp. Appl. Math. 236, 3561, 2012] developed
an analytical technique where the 1D Hamiltonian is mapped into 3D which they claim would be able to resolve the divergence problem. But a
numerical implementation of this technique was not attempted and it is still unclear if the problem has really been solved. Here we analyse in detail the influence of the various approaches to
the divergence problem on the eigenvalue spectrum. We implement a numerical algorithm for the method proposed by Estrada et al and compare them with the proposed
workarounds to evaluate the effects on the eigenvalue spectrum.

Funding: CPNq, Capes and Fapemig