Surface Evolver Newsletter no. 10

                      Surface Evolver Newsletter Number 10
                              December 16, 1994

                    Editor: Ken Brakke, brakke@geom.umn.edu

Contents:
  Query on curvature.
  Version 1.97


Query on curvature.
  From Christian Burger <burger@pc1442.phys-chemie.uni-marburg.de>:

     I am interested in the investigation of physical systems which are
     usually highly symmetric two-phase systems separated by an interface
     which should be some kind of minimal surface. Very often, the systems
     are bicontinuous, i.e., two unconnected 3D network domains live in a
     common matrix of surrounding matter (I'm speaking in physical terms
     here, not mathematical). Naturally, I'm using the evolver with the
     torus model. The domains have defined volumes and are incompressible.
     Physical examples for such systems are micro-emulsions,
     polyelectrolyte-surfactant complexes, block copolymers, etc. It is
     likely that curvature plays an important role for these systems. If the
     structure unit building up the interface is not symmetric this leads to
     a `spontaneous curvature', i.e., a finite H_ZERO in terms of evolver
     nomenclature.  If you are aware of any work in this direction people 
     have done so far using the evolver, I'd like to get some references on it.

Version 1.97:

 Version 1.97 is now available for ftp (except Mac version hasn't
 been updated yet.)

 Changes from 1.96 are mostly a lot of obscure bug fixes.  But
 there are some new features of general interest.

 New Hessian commands: "eigenprobe value" will print the number of
 eigenvalues of the energy Hessian that are less than, equal to, and
 greater than the given value.  It is OK to use an exact eigenvalue
 (like 0, often) for the value.   Useful for probing stability.
 "lanczos value" will do a little Lanczos algorithm and report
 the 15 nearest approximate eigenvalues to the given value.  Do not
 use an eigenvalue for the value! Not real polished yet.

 The default way of factoring the Hessian is with routines from the
 Yale Sparse Matrix Package (YSMP), patched a bit to cope with zeroes
 on the diagonal.  I've been tinkering around with my own version
 of minimal degree factoring that is a little more aware of the
 surface structure.  It can be activated by "ysmp off" and deactivated
 by "ysmp on".  My version seems to have 10-20% less fill, and so
 be a little faster.  It also seems to not have as many spurious
 zeroes on the diagonal.  Further, in an attempt to cope with
 indefinite Hessians, there is a toggle "bunch_kauffman" that enables
 a version of Bunch-Kauffman factoring in my version.

 Certain named quantity attributes can now be referred to in expressions.
 For quantity qqq, qqq.value is the current value, qqq.target is 
 the constrained value, qqq.modulus is the same as the modulus in the
 datafile definition, and qqq.pressure is the Lagrange multiplier
 for a constrained quantity.  You can assign values to qqq.modulus
 and qqq.target.

 The dump command 'd' now saves almost all toggle states (except some
 debugging toggles) if not the default values.