1 / 43

Nanomaterials Informatics & Materials QSPR Curt Breneman, Linda Schadler and Cate Brinson

Nanomaterials Informatics & Materials QSPR Curt Breneman, Linda Schadler and Cate Brinson June 2009. Quantitative Structure-Property Relationship (QSPR) Modeling. Find "hidden" information in chemical data Retrieve structures with a well-defined property range from a database

obert
Download Presentation

Nanomaterials Informatics & Materials QSPR Curt Breneman, Linda Schadler and Cate Brinson

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Nanomaterials Informatics & Materials QSPR Curt Breneman, Linda Schadler and Cate Brinson June 2009

  2. Quantitative Structure-Property Relationship (QSPR) Modeling • Find "hidden" information in chemical data • Retrieve structures with a well-defined property range from a database • Identify of appropriate descriptor types for each situation • Assess model applicability domain • Validate modeling results

  3. General Perceptions of Predictive Modeling

  4. Materials Informatics:Significance and Context • Unified vision of Cheminformatics and Materials Informatics • In terms of fundamental physics of molecular interactions • Builds upon expertise and capabilities of RECCR • QSPR • Simulations • Machine Learning • Software tools • RECON • PEST/PESD • NWCHEM • Computational resources (CCNI)

  5. What is Materials Informatics? Wikipedia strikes again: • Based on information, but… • …not just databases. Brinson Advanced Materials Lab

  6. Polymer Nanocomposites Design:A Materials Informatics Example Why nanocomposites? Nanoparticle properties Multifunctional capabilities Chemical functionalization Huge interphase zone Large design space Nanoparticle / polymer combos…etc nanotubes nanographite nanoclay nanosilica

  7. Microcomposites vs Nanocomposites 1% 10m carbon fiberNo impact on bulk polymer response 1% SWNT Huge impact on bulk polymer response 60% 10m carbon fiber Huge impact on bulk material response Brinson Advanced Materials Lab

  8. Design Options Nanoparticles All nanotube varieties All graphite plate varieties All nano-clay varieties C60, alumina, silica, …. Polymer matrices Thermoplastics - amorphous, crystalline Thermosets Functionalization Natural chemical interactions between particles and matrix Designed functionalization - group chemistry, group length Processing Solvent based Melt mixing Solid state processing Layer by layer, extruding, fiber spinning, … “Go make a nanocomposite for our new project.” How many choices does Jane Graduate Student have? YIKES! Brinson Advanced Materials Lab

  9. Challenges and Opportunities inNanocomposite Informatics • -What role do surface interactions play in CNT/matrix adhesion? • Which chemical descriptors/modeling methods are important? • How can clustering and percolation be quantified? • Characterization of the interphase? • - How can we relate these to macroscopic properties? • Establish relationships through modeling • Utilize intrinsically multi-scale character of models • Involves interdisciplinary effort • Impact of processing conditions • - Materials Informatics: the bridge between chemistry, physics, materials science and engineering • - Materials Quantitative Structure Property Relationship (MQSPRs) are the link between scales.

  10. Structure-Property Relationships X PROPERTY STRUCTURE Statistical or Pattern Recognition Methods Computational Chemistry MOLECULAR DESCRIPTOR REPRESENTATION

  11. Simple models are better Interpretable models are better Reality: need to balance predictive ability and interpretability Model Parsimony Rules:

  12. Descriptor Choices No particular class of descriptors address all problems Must be chosen to be problem- and scale-specific May be chosen to be best match for learning methods For nanoparticles, we need to choose descriptors that address interfacial interactions!

  13. Interphase – a crucial issue Surface area increases 103x Altered mobility & chain conformation Altered crystallinity Decreased interparticle distance Crosslink density Chemistry Surface Area/Volume PMMA-MWNT fractograph rf Interaction Zone Particle Interaction zones will overlap at low volume fractions (< 2 vol%)

  14. Interphase properties Success depends upon understanding and designing the interphase • Unlike particle and matrix, interphase properties unknown • Contributing factors: • Geometry of nanoparticle • Functional groups on nanoparticle • Polymer chains on nanoparticle • Polymer characteristics • Percolated interphases affect bulk properties • Attractive interactions, wettability of interface: Tg • Repulsive interactions, dewetted interface: Tg • Stiffness, strength, toughness impacted Brinson Advanced Materials Lab

  15. Multi-Scale Nanocomposite Descriptors

  16. Representing Nanostructures Structural Descriptors Physiochemical Descriptors Topological Descriptors Geometrical Descriptors Molecular Structures Descriptors Model Property

  17. RECON/TAE Electronic Descriptors

  18. Surface Electronic Descriptors

  19. Modeling and Validation DATASET Training set Test set Real Y-scrambling model validation! Bootstrap sample k Predictive Model Training Validation Scrambled Learning Model Tuning / Prediction Prediction

  20. Dataset Truncation testing: Do 3D Descriptors improve model robustness? – regression example

  21. Materials Informatics:Nanocomposites Fundamentally we have: Dispersion Interphase Morphology Macroscopic properties Nanoparticle + surface chemistry Polymer and its chemistry interact interact Use informatics to begin to correlate across these domains with goal to achieve intelligent design • Multiple interacting constituents • Multiple length scales of concern • Multiple mechanisms affecting properties • (atomistic and mesoscale)

  22. Nanocomposite Informatics Correlate results to indicate ability for design (model validation) • Take experimental data on nanocomposites from lab and literature: • We know constituents • We know morphology (dispersion) • We know macroscopic properties (Tg) • Perform chemical modeling • Obtain parameters describing nanoparticle surface • Obtain parameters describing polymer (“solubility” descriptors) • Perform meso scale modeling • Correlate degree of dispersion to interphase properties and connectivity • Predict macroscale properties Brinson Advanced Materials Lab

  23. Materials Informatics:Sample Modeling Applications • Nanocomposite Glass Transitions • Dynamic Behavior of Composites

  24. Nanocomposite Informatics:Sample Modeling Applications • Nanocomposite Glass Transitions • Dynamic Behavior of Composites

  25. Experimental data F F F F F F F F F F Silica filler loading (wt%) DTg -Ddescriptor bare silica TEOSH2ONH3EtOH fluorination SiO2 SiO2 15 or 150 nm Done for: 3 particles, 3 polymers, 5 functionalizations 15 nm particles Significant agglomeration - 75 nm used as particle size for calculation.

  26. Meso-scale modeling: clustering & percolation! Use Finite Element approach to analyze the clustering effect and percolation (2D example) Randomly generated locations of fillers enclosed with interphase Investigate mechanical and rheological percolation Interphase properties: freq. shift by 102 Matrix properties (E’(), E’’()) Meshed 2D Model Sketch of 2D Model

  27. Clustering: Pattern Generation An Example: 90 particles, 15 cluster regions, xcl=30% The centers of cluster region are randomly generated Each cluster has the same diameter and contains six particles Particles are randomly distributed inside of the cluster region Periodic geometry Define: Inhomogeneity xcl is the particle volume fraction in a cluster region xcl Increases, the agglomeration of particles becomes more severe

  28. Micromechanical Modeling:Inhomogeneous Structure x=10% x=20% Random Dlog P(o/w) small Dlog P(o/w) large x=30% x=25% critical clustering

  29. Attractive- Repulsive Interaction Dlog P(o/w) small large Relation between Tg in interphase and clustering degree Critical Clustering is 25% For random, Xi=0

  30. Towards Multiscale & Informatics • Train: Determine appropriate molecular descriptors of interface • Correlate: experimental data for property (Tg) • Predict: Micromechanics parameter relates interface descriptor to relaxation time change and propensity to aggregate Tg increase Tg increase compatibility increase compatibility increase

  31. Towards Multiscale & Informatics Particle Surface Matrix Chemistry Predict: Micromechanics parameter relates interface descriptor to relaxation time change and propensity to aggregate Molecular Descriptor: Dlog P(o/w) • What’s the Big Deal? • 10 degree change in Tg • 2 order of magnitude change in average relaxation time • AND percolation Interphase Incompatibility

  32. Nanocomposite Informatics:Sample Modeling Applications • Nanocomposite Glass Transitions • Dynamic Behavior of Composites

  33. CNT / Epoxy Composite:Experimental Observation

  34. Finite Element Model • CNTs: • Straight , elastic and isotropic • Random orientation • Random dispersion • Beam elements • Two aspect ratio: L/d=20 & 200 L/d=200 L/d=20 • Matrix: • Polycarbonate • Master curves measured by DMA • Approximated by 28 term prony series

  35. Interphase • Interphase regions • Capsule shape surrounding the CNTs • Two layers with equal thickness • Same thickness for both aspect ratio (five times the tube diameter) • Interphase properties • Attractive interaction • Properties determined by the subtraction of normalized E” curves • Average relaxation time is two decades longer than the bulk PC • One decade less mobile for outer layer • Three decades less for inner layer

  36. Broadening Effect Experiment Simulation The composite with short nanotubes shows broader loss modulus which could be attribute the more interphase volume In FE models, interphase vol% of L/d=20 is 25% more than that of L/d=200

  37. FSP vs Broadness Experiment Simulation Free space parameter (FSP) to characterize particle-free regions in matrix The value of FSP depends on dimensionality and image quality But the trends are same!

  38. First, we need to define "acceptable" or “useful” predictive modeling results… Those that come from the best theory and most computing possible? Those that yield “Experimental accuracy”? Those that are stable to small changes of input values? Those that get the “Rank Order” correct? How about those that permit good decisions to be made? Is more computing always better? (Hint: NO!) What is the level of diminishing return? Are there any other downsides? (YES!) Temptations of Our Time: Big Hardware leads to Over-Computation Do we sometimes use more computer resources simply because they are there? What are the real costs of over-computation? How Much Modeling is Enough?

  39. Next step: Prospective Predictions • Can move from this to design. • Given particle + polymer • Given functionality • Calculate interaction parameters • Can then predict with meso model Tg!!! • More work to do: • Utilize molar refractivity, polarizability descriptors • Utilize more sophisticated model than single coefficient • Link delta log P*SA to clustering parameter • Appropriately modify clustering parameter and interphase properties • …

  40. Intelligent Design of Polymer Nanocomposites via Informatics The future is near! Andrew Murray of Harvard University: the bioinformatics approach "saves us from the era of one graduate student, one gene, one PhD". Cate Brinson of Northwestern: Similarly, the goal of materials informatics is to save us from “one graduate student, one (nano)composite, one PhD”.

  41. Future Directions and Challenges1) Advances in data collection ontologies and utilization2) Data Fusion methodologies3) Context-based multi-scale descriptors4) Relating modeling results at each length scale with the next higher length scale

  42. The RECCR Community http://reccr.chem.rpi.edu

  43. ACKNOWLEDGMENTS • Current and Former members of the RECCR/DDASSL group • Breneman Research Group (RPI Chemistry) • N. Sukumar • M. Sundling • Min Li • Long Han • Jed Zaretski • Theresa Hepburn • Mike Krein • Steve Mulick • Shiina Akasaka • Hongmei Zhang • C. Whitehead (Pfizer Global Research) • L. Shen (BNPI) • L. Lockwood (Syracuse Research Corporation) • M. Song (Synta Pharmaceuticals) • D. Zhuang (Simulations Plus) • W. Katt (Yale University chemistry graduate program) • Q. Luo (J & J) • Embrechts Research Group (RPI DSES) • Tropsha Research Group (UNC Chapel Hill) • Bennett Research Group (RPI Mathematics) • Collaborators: • Brinson Advanced Materials Lab – (Northwestern) • Schadler Research Group (RPI Materials Engineering) • Funding • Office of Naval Research (ONR) • Lockheed-Martin (LMCO) • NIH (GM047372-07) • NIH (1P20HG003899-01) • NSF (BES-0214183, BES-0079436, IIS-9979860) • GE Corporate R&D Center • Ford-Boeing Alliance • NSF NIRT Program • NASA • Chemical Computing Group (CCG)

More Related