Forcite Ewald Sums/ Exclude 1_3 interact/ Parallel Simulations

Hi,

I have been using the Materials Studio & the Forcite Plus module to build my models and test run my simulations with various versions of the forcefield.  The test runs suggest that this is going to work and I would like to run much longer simulations.  I have three issues about which I would like to ask for some clarity.  Sorry to stack these issues all into one thread.  If it'd be better to break them up into separate issues, I'll gladly parse the post.  The first involves the implementation of the Ewald Sums in the Forcite module; the second concerns the exclusions that result from the toggle buttons in the *.off forcefield gui; and the third concerns running parallel MD simulations with the Forcite Module.

1) The Ewald sum implementation in the Forcite Plus dynamics simulations indicates that the accuracy is calculated differently then it is in the Discover module.  Namely, that Ewald sum accuracies are calculated on a per-atom basis in Forcite but on a per cell basis in Discover.  Is there any articles available comparing the accuracy of these two different means of determining the accuracy?  Or even an article showing the derivation of the algorithm for implementation?  Is there a correlation table that gives comparable parameter settings between cell-based and atom-based accuracies?  Another odd occurrence, if I open the "Calculation" document from and MD run with Forcite and Ewald summations, it has, on occasion, indicated that the summation method is "Atom-based", although the text document for the simulation summary indicates "Ewald Sum" for the summation method.  This isn't every time and it may just be an artifact of some default setting, but it is unsettling.  Am I misinterpreting the nature of an "atom-based" accuracy for a Ewald sum?

2)  There are two versions of the forcefield and different protonation states that I am testing.  In the one model and forcefield version, I have to exclude the 1_3 interactions for atoms whose interaction energies are defined by a harmonic angle term.  The documentation states that non-bond interactions between two atoms separated by two bonds are excluded.  I have a number of functional groups with these harmonic angle energies across an interface.  Are the non-bond interactions (Coulomb and LJ) excluded for all atoms of the forcefield types 1 & 3 defined in the bond (per FFtype) or only for the atoms in that individual angle (probably by atom no.)?  i.e. do adjacent functional groups interact through non-bonded forces?

3) As the test MDs have been successful for 100 ps, I'd like to run simulations on the nanosecond timescale.  I've tried to export the model to other simulation programs (LAMMPS) to run it in parallel, but the model results do not match those in Materials Studio.  The total energy produced by the two different programs (despite identical starting geometries) differs by 100,000 kcal/ mol for a system with ~2400 atoms.  The energy expression is dominated by the non-bond pairwise interactions (~90% nonbond energies).  I have been unable to run Forcite Plus MD simulations on multiple processors on either my local (Windows, core duo) machine or the larger server (Unix, 12 processors).  I am open to alternative solutions to exporting the model to LAMMPS, but right now, I can't get results in either parallel computing environment.

I'd like to ask you all for your advice and suggestions in solving any or all of these issues.  I look forward to your responses.

Thank you in advance,

Aric