Cyber Activities in the NSF Chemistry Division Celeste M. Rohlfing March 9, 2006.

Cyber Activities in the NSF Chemistry Division

Celeste M. RohlfingMarch 9, 2006

Context of CHE Cyber Investments

• Community developed under NSF’s Information Technology Research (ITR) program (FY00-04)

• Cyber activities already nascent in CHE individual investigator and group grants

• Community guidance sought in 4/04 MPS Cyberscience Workshop and 10/04 CHE Cyber-Enabled Chemistry Workshop

ITR: Computational Science & Engineering On-Line (T. Troung, Utah)

Individual Investigator: Worldwide Grid Computing (V. Pande, Stanford)

• Folding@Home People donate their

idle computer time They visit the

website, download, & run the software

• Very powerful over 100,000 CPUs more powerful than

all NSF supercomputer centers combined

http://folding.stanford.edu

CHE Workshop Charge 10/04

• Identify chemical science problems (“cyber-science” drivers) that will be most impacted by investments in cyberinfrastructure (e.g., fast/cheap computers, global/grid networking, digital databases/libraries)

• Specify the infrastructure needs (hardware, software, middleware, personnel, long-term support, etc.)

• Inform CHE about challenges, both near-term opportunities and long-range transformative activities

Science Drivers

•Multiscale modeling including high-dimension, chemical-accuracy potential energy surfaces

•Real-time feedback to control of reacting systems monitored by sensor technology

•Prediction of optimal experiments (lower cost of discovery and process design)

•Validation of computational models vs. experimental data, and vice versa

Cyber-Enabled Challenges

•New paradigm for information flow (transparent resource sharing such as data grids rather than centrally stored data bases; workflow management tools)

•New paradigm for shared instrumentation (“remote chemistry”) including broadening participation

•Interfacing data and software across disciplines (“interoperability”), and development of cyber collaboration tools (e.g., e-notebooks)

Workshop Feedback

•Strong support for continuation of capacity computing, via “NSF Supercomputer Centers” and follow-on’s

•Explicitly include cyber in new & updated solicitations

•Budgetary support for programmers/software engineers, and code maintenance

•5-year award duration (theoretical model computational algorithms community application code, e.g.)

CHE Response to Workshop

•Solicitations: NSF 05-555; 06-512; FY07 & beyond

•Cyber incorporated into existing solicitations as they are revised

•C&E News editorial in 3/05

•Chemistry Division cyber website launched 4/05 (www.nsf.gov/chem/cyber)

CHE Response to Workshop

•Dear Colleague letter in 4/05

•Active role in MPS & NSF cyber plans and activities

•Cyber-Enabled Chemistry symposium at Fall 2006 ACS National Meeting in San Francisco

CHE FY05 & FY06 Solicitations

FY05 competition: 4 awards, $10M; details in C&EN feature article 10/17/05

FY06 competition: proposals due 03/01/06

Cyberinfrastructure for Next-Generation Biomolecular Modeling

Tools for Ab Initio Molecular Dynamics and Simulation Analysis

Developing Collaboratory Tools to Facilitate Multi-Disciplinary, Multi-Scale Research in Environmental Molecular Sciences

Process Informatics for Chemical Reaction Systems

PrIMe Data Library

X (cm)

R(c

m)

0 2 4 6 8 100

1

2

3

4 1

2

3

5

68

9

1004

27

11

Trial Mechanism

Sensitivity Analysis

Statistical Surrogates

OptimizationData Collaboration

Validation

Thermochemicaland rate data

Trainingtargets

Best current model

Validationtargets

PrIMe Data Depository

•New paradigm for generation of predictive kinetic models, i.e., models with quantified uncertainty

• Analysis of the entire data available• Assures dataset consistency• Dynamic generation of predictive

models based on the entire dataset• Determine which data have the largest

impact on the model accuracy and consistency

• Identify which experiment/theory is to perform that will have the highest impact

Community Review

Process Informatics Model (PrIMe)

Community Submission

Community Use

CHE-0535542 (M. Frenklach et al., UC-Berkeley)

Computer simulation can now predict chemical reaction dynamics, even in challenging cases where chemistry is initiated by light absorption. So far, this capability has been reserved for experts in the field. This project will accelerate widespread application of this technology by creating user-friendly tools aimed at non-experts and experimentalists. This will enable rapid design and interpretation of new experiments and computational design of molecular systems. These tools will also make applications by experts easier through standardization and automation of simulation management and analysis.

Ab Initio Molecular Dynamics & Simulation Analysis Tools

CHE-0535640 (T. Martínez et al., U. Illinois)

Images created by H. Hudock & T. J. Martínez; Experimental data from A. Stolow & S. Ullrich

Comparison of theoretical and experimental transient signals following photoexcitation of thymine (a component of DNA). Experiments provide information about the relevant timescales, but the simulations add detailed information about the molecular mechanisms of DNA photodamage. Simulations like these will become routine with the tools being developed in this project.

Expt

AIMD

Cyberinfrastructure for Next-Generation Biomolecular

ModelingCHE-0535710 (T. Head-Gordon et al., UC-Berkeley)

Proposed efforts:

improved theoretical model infrastructure for the biomolecular simulation community

extensive validation of a recently introduced polarizable protein & water force field, AMOEBA, that is formulated in the molecular mechanics and software package, TINKER

hardened implementation of AMOEBA in software packages that write critical software components in assembly for Pentium systems (e.g., GROMACS)

development of similar strategies for fast simulation codes in TINKER, which has 30,000 users

Number of downloads of the complete TINKER software distribution from public introduction in April 1996 through April 2005, and filtered to remove Washington U. access and multiple access from individual Internetsubdomains during a calendar month

Collaboratory Tools to Facilitate Multi-Disciplinary, Multi-Scale Research in Environmental

Molecular SciencesCHE-0535656 (K. Mueller et al., Penn. State U.)

Cyberinfrastructure tools will be designed to enable and promote multi-disciplinary, multi-scale research in environmental molecular sciences. Tool development by information scientists will be validated and augmented with design feedback from chemists and geochemists. Intelligent data management and use will streamline the process of discovery as information is provided in context while maximizing connectivity among methodologies and disciplines. Ultimately, we will implement a system to provide seamless access to multiple databases and digital libraries for the end user.