Numerical methods

 

 

 
  • "Node patch" boundary element method (NBEM):

    The boundary element patches constructed around nodes and supposed to be constant, we call this a constant "node patch" BEM. This BEM considerably improves the speed performance compared with the constant element and linear element methods.

    A simple way to construct these new patches is illustrated in right figure in which an example ``node patch" at the i-th node that has five neighboring elements is constructed. All the centroids {O_l, l=1,...5} of the five adjacent triangles and the midpoints {C_l, l=1,...5} of the attached five edges are listed, and then the area formed by {O_1,C_1,O_2,C_2,... O_5,C_5,O_1} is the new patch that we want. Therefore, there is one third of the area of each neighboring triangle occupied by the ``node patch" $\Delta S_i$. The other patches are similarly constructed. All the node patches connect and cover the whole surface. Now, the far-field integrals on the new patch Si become



    where, supposing all neighboring elements of ith node form a set {L},



    where nl is the unit normal vector of the lth neighboring element, Sl is the area of the lth adjacent triangular element, and &Delta Sib here should also be considered as a vector. As in the constant element treatment, for near-patch integration, a normal quadrature method is used. Similar treatments apply to the integrations for the kernel u and its derivative, as well as for the other second-derivative terms if the dBIEs are used.

    There are three main advantages of this ``node patch" treatment in the BEM. First, as aforementioned, the unknowns are reduced by almost half relative to the constant element method. Second, straightforward storage of matrix coefficients, because it is easy to locate the ``near points" (local list) of any node for integration. This, though not explicit, but maybe the main advantage of NBEM. Third, convenient to use interpolation method as a postprocess to compute the potential and its gradient (not only the normal derivative) at any position near or on the molecular surface, e.g., for stress and force calculation (see the reference).


    Reference:
    Lu, B. Z. and McCammon, J. A. Improved boundary element methods for Poisson-Boltzmann electrostatic potential and force calculations. J. Chem. Theory. Comput. 3(3):1134–1142, 2007. link .

 
     

Home |Education  Research | Publications  Resume| Useful Links  Personal Link

Questions or comments about me, email me at blu@mccammon.ucsd.edu

Copyright 2006-