Workshop on “Tools for Condensed Phase Computational Chemistry”

Clayton Hall, University of Delaware, 27-30 May 2003


Format of Workshop


Our workshop is not a usual meeting in which experts talk about their latest work.  Instead, this meeting is about tools.  On the first day, DoD computational chemists will describe their goals, tools, and the problems they encounter. Then, experts in the field will speakwithin the context of the DoD presentationabout the special tools developed or used by them.  Each expert will have about an hour – this includes discussion and questions and we will encourage continual interaction between speakers and audience (we will keep the meeting strictly on schedule, so please be prepared). Near the end of the workshop, we will separate into groups to develop specific answers to questions presented in the first part of the meeting and recommendations.  The last hour of the workshop will be the groups reporting back to everyone.  As you know, the success of this workshop depends on everyone’s participating throughout. The workshop will produce a report, but the on-site interactions are most important.


General Questions to Address


  • What computational tools are best suited to current Army computational problems?
  • What kinds of systems can be treated by these tools and what are computational costs?
  • What kinds of physical properties can be computed?
  • What is the accuracy of predictions?
  • How the methods can be further developed in near future?
  • What investment (time, money) is required to use these tools?
  • How can the tools output be validated (what experiments)?
  • Are the methods predictive enough to use theory to screen notional materials (molecules that may not have been synthesized but which, based on their expected structures, appear to have desired properties). 


Specific Issues to Address


We need reliable computational methods to predict properties of condensed phase materials of relevance to the Army, chiefly energetic materials.  In most cases these are molecular crystals but some ionic crystals are also of interest.  The first step is a determination of interactions between molecules, i.e., electronic structure calculations for fixed nuclear positions.  One approach is to develop reliable potentials (force fields) including effects of many-body forces for molecules involved.  Another approach is to perform electronic structure calculations with periodic boundary conditions for molecules of interest placed in a unit cell.  Since often one has to include a large number of molecules in the unit cell, computational requirements of such calculations can be large.  How do the two approaches compare?

The next step is to determine low-energy crystal structures of materials.  If the relevant potentials are known, one can do it using molecular packing programs.  How feasible is it to go from using empirical atom-atom potentials to potentials computed using modern electronic structure techniques?  Are optimizations of crystal structure feasible, for molecules of interest, within periodic conditions programs calculating the interactions on-the-fly?

Majority of the codes developed or used by the speakers use some flavors of the DFT method.  Comparisons of these implementations and of capabilities of various codes will be in order.  Furthermore, we should discuss whether DFT is a proper computational method for molecular crystals.  DFT is known to often fail to predict reasonable interaction energies, in particular for systems with a significant dispersion component.  Interactions of large organic molecules, the main interest for the workshop, always include a large dispersion component and for some configurations are dispersion dominated (e.g., stacked configurations of DNA bases).

The minimum energy structure immediately gives the density of a material, which is one of the most important properties to know.  Densities have been so far mostly predicted with molecular packing programs based on empirical potentials.  Even this approach is time consuming.  One must generate a variety of crystal structures corresponding to the various crystalline space groups from the coordinates of a single molecule.  Each crystal structure must be optimized, i.e., the crystallographic parameters minimized with respect to energy and all energies compared.  The current criterion for selection of the most probable structure is lattice energy; i.e., the structure that has the lowest lattice energy is assumed to be the “correct” crystal structure.  This criterion has proven to be insufficient and additional criteria, perhaps the evaluation of DG for each crystal structure, should be included in the selection procedure.

The molecular packing calculations would be more predictive if empirical potentials were replaced by ab initio potentials.  However, two-body (pair) potentials would not be sufficient in most cases.  How can we compute many-body potentials?  Can polarization model of many-body forces be sufficient?  Ab initio potentials are typically more complicated than empirical potentials and will make molecular packing calculations more time consuming.

Other important physical effects determining the structure and properties of molecular crystals are intramolecular degrees of freedom. Most of work in the field assumes rigid molecules. However, partial deformations of the molecules compared to their gas-phase structures are always present in crystals.  In some cases, torsional deformations can be large.  For example, molecules like RDX may be adequately modeled as rigid structures, but allowance for molecular deformations will likely be important for floppy molecules like PETN,  as shown by Sorescu, Rice, and Thompson.  Inclusion of a large number of intramolecular degrees of freedom is next to impossible in electronic structure calculations of potentials.  Are on-the-fly approaches better suited to investigate monomer nonrigidity effects?

In addition to density, one needs to know several other properties of materials such as heats of formation, burning rates, etc. Some of them can be obtained from molecular dynamic calculations for crystals.  How reliably can we predict such properties and at what computational cost?  Can the predictions be trusted for notional materials?


Finally, are we able to computationally model chemical reactions in the condensed phase?  What approaches are possible?  Can periodic boundary programs be used for this purpose taking into account that chemical reactions locally break crystal order and therefore unit cells have to become huge?




May 27:           Arrive, dinner (at 19:00 in Clayton Hall), evening orientation and introductions. Registration starts at 18:00.

May 28:

                      DOD Programs and Needs for Computational Chemistry Packages



Betsy Rice, Cary Chaba- ­lowski, William Mattson


Army Needs for Crystal Structure and other Property Predictions


Betsy Rice


Parallelizing Molpak and Planewave






David Singh


Pseudopotential and Planewave Codes


                     Presentations by Experts



Krzysztof Szalewicz

U. Delaware

Intermolecular Interaction Potentials


Herman Ammon

U. Maryland

Crystal Structure Prediction for Energetic Materials






Sally Price

U. Coll. London

Anisotropic Atom-Atom Intermolecular Potentials in Organic Crystal Structure and Property Prediction


Emilio Artacho

Cambridge U.

Material Properties with SIESTA: Strengths, Weaknesses, and Prospects






Julian Gale

Imperial Coll.

General Utility Lattice Program


Anne Chaka


Validation of Computational Methods








Possible Implementations of New Approaches by Army



May 29:

                     Presentations by Experts (continued)



Eric Bylaska


Materials Properties with NWChem


Richard Martin

U. Illinois

Identifying the Key Problems in Present Density Functionals






Gustavo Scuseria

Rice U.

Condensed Phase Simulations using Gaussian Orbitals and Periodic Boundary Conditions






Emily Carter


Modeling Chemistry and Physics in Bulk Crystals, Surfaces, and Interfaces:  What's Possible and What Can We Trust?


                       Group discussion: Best Ways to Proceed on Army Problems



Krzysztof Szalewicz and Bob Shaw


Summary remarks and guidelines for discussion groups


Discussion group I


Chabalowski, Byrd, Mattson, Martin, Carter, Bylaska, Scuseria, Doren, Tchoukova, Murdachaew, Sandler, Shaw


Discussion group II


Rice, Kim, Ammon, Price, Gale, Artacho, Chaka, Singh, Bukowski, Akin-Ojo, Wang, Rowe, Szalewicz








Groups Report back with their Conclusions






May 30:



Van schedule:


                                    Depart from                 Depart from

                                    Embassy Suites           Clayton Hall

  Tuesday   5/27/03:      18:00                            21:00

  Wednesday 5/28/03:   08:00                            21:00

  Thursday  5/29/03:      08:10                            20:00