1
Computational Nanoscience: An Emerging Tool for Exploring and Exploiting the Nanoscale
Peter T. Cummings
Department of Chemical Engineering, Vanderbilt Universityand
Nanomaterials Theory Institute and Chemical Sciences DivisionOak Ridge National Laboratory
Fall Creek Falls Workshop on High-End Computing in Science and Engineering Fall Creek Falls, TNOctober 26-28, 2003
2
Acknowledgments Research support:
Division of Chemical Sciences, Office of Basic Energy Sciences, U.S. Department of Energy
National Science Foundation: Interfacial, Transport and Thermodynamics, NANO and Division of Materials Research
National Energy Research Supercomputing Center (NERSC) Center for Computational Sciences, Oak Ridge National Laboratory
Collaborators: Clare McCabe (CSM), Milan Predota (Czech Academy of Sciences),
Sharon Glotzer and John Keiffer (UMich), Matt Neurock (Virginia), David Keffer (U. Tenn), Shengting Cui (U. Tenn.)
ORNL: David Dean, Predrag Krstic, Jack Wells, Xiaoguang Zhang, Hank Cochran
Yongsheng Leng (VU) Collin Wick (DOE Comp Sci Fellow from U. Minn.), Jose Rivera (U. Tenn),
T. Ionescu (VU)
3
Outline of Talk
Introduction Theory, Modeling and Simulation in Nanoscience ORNL Center for Nanophase Materials Science and
Nanomaterials Theory Institute Algorithmic aside
Molecular electronics Multiscale modeling
Nanoconfined fluid rheology and structure Rheology of nanoconfined alkanes
Classical Simulations of Carbon Nanotubes Sliding behavior of multiwall nanotubes
Organic-inorganic nanocomposite materials Multiscale modeling
Conclusions
4
*M. Roco, FY 2002 National Nanotechnology Investment Budget Request
Introduction
Nanoscale science and engineering Ability to work at molecular level, atom by atom, to
create large structures with fundamentally new properties and functions* At least one dimension is of the order of nanometers Functionality is critically dependent on nanoscale size
Promise of unprecedented understanding and control over basic building blocks and properties of natural and man-made objects*
National Nanotechnology Initiative http://www.nano.gov
$710 million approved by Congress for FY 2003. FY 2004-2006 request: ~$2B
5
Introduction
Theory, modeling and simulation (TMS) Expected to play key role in nanoscale science and technology
“Nanotechnology Research Directions: IWGN Workshop Report. Vision for Nanotechnology Research and Development in the Next Decade,” edited by M.C. Roco, S. Williams, P. Alivisatos, Kluwer Academic Publisher, 2000
– Also available on-line at http://www.nano.gov– Chapter 2, Investigative Tools: Theory, Modeling, and Simulation, by
D. Dixon, P. T. Cummings, and K. Hess
• Discusses issues and examples McCurdy, C. W., Stechel, E., Cummings, P. T., Hendrickson, B., and
Keyes, D., "Theory and Modeling in Nanoscience: Report of the May 10-11, 2002, Workshop Conducted by the Basic Energy Sciences and Advanced Scientific Computing Advisory Committees of the Office of Science, Department of Energy
– Published by DOE– Also available on the web at
http://www.sc.doe.gov/bes/Theory_and_Modeling_in_Nanoscience.pdf
6
Introduction
Heirarchy of methods relevant to nanoscale science and technology Connection to
macroscale
7
Introduction
TMS advances over past 15 years relevant to nanoscience Moore’s Law
Gordon Bell Prize: 1Gflop/s in 1988 vs. 27 Tflop/s in 2002– More than four order of magnitude increase in 14 years
Explosion in application and utility of density functional theory Molecular dynamics on as many as billions of atoms Revolution in Monte Carlo methods (Gibbs ensemble, continuum
configurational bias, tempering, etc) Extraordinarily fast equilibration of systems with long relaxation times
New mesoscale methods (including dissipative particle dynamics and field-theoretic polymer simulation) Applicable to systems with long relaxation times and large spatial scales
Quantum Monte Carlo methods for nearly exact descriptions of the electronic structures of molecules
Car-Parrinello and related methods for ab initio dynamics Reactions, complex interfaces,…
8
Why is TMS so crucial to NSE?
Interpretation of experiment Many experiments at the nanoscale require TMS to
understand what is being measured
QuickTime™ and aYUV420 codec decompressorare needed to see this picture.
Unlike bulk systems, large-scale manufacturability of nanoscale systems will require deep theoretical understanding
Inherent strong dependence on spatial dimensions
9Center for Nanophase Materials SciencesOak Ridge National Laboratory
Neutron Science Opportunity to assume world leadership using unique capabilities of
neutron scattering to understand nanoscale materials and processes
Synthesis Science Science-driven synthesis will be the enabler of new generations of
advanced materials
Theory/Modeling/Simulation The Nanomaterials Theory Institute
Scientific thrusts in 10 multidisciplinary research focus areas
Access to other major ORNL facilities Spallation Neutron Source High-Flux Isotope Reactor
Began Sept, 2002 $60M for building and equipment; $18M/yr ongoing
Specializing in neutron science, synthesis science, and theory/modeling/simulation
Center for Nanophase Materials Sciences
HFIR
SNS
34333231302928272625242322212019181716151413
R
4
3
2
1
R
4
3
2
1 1
2
3
4
R
80'40'0' 80'40'0'
Nanofabrication Research Lab
Multistory Lab/Office Building
http://www.cnms.ornl.gov
10
Examples Nanorheology [NSF NANO]
Nanoconfined fluid rheology and structure Sliding double-walled carbon nanotubes
Molecular electronics [ORNL LDRD; DOE Comp Nanoscience] Structure of self-assembled monolayers
Hybrid organic-inorganic nanocomposite materials [NSF NIRT] Polyhedral oligomeric silsequioxane (POSS) molecules
Nanoscale complexity at metal oxide surfaces [DOE NSET, NSF NIRT] Planar interfaces and nanoparticles
Fluids in nanopores [ORNL LDRD] Phase equilibria of water and water/carbon dioxide mixtures in
single-walled carbon nanotubes Self-Assembly of Polyelectrolyte Structures in Solution: From Atomic
Interactions to Nanoscale Assemblies [DOE NSET]
11
Algorithmic Aside
Solve Newton’s equations (or variation) for positions and velocities of atoms
Solve 6N first-order non-linear differential equations numerically, where N is typically 103-106 (as high as 109)
Time step ~10-15 s 1 ns ~ 106 time steps, 1s ~ 109 time steps Numerically intensive Complexity measure = #atoms x #time-steps
– Extreme values: 1012-1014
– Protein folding for 1s : 104 x 109
dridt
=pi
mi
, dpi
dt=Fi
ri =positionpi =momentum
Fi =forcemi =mass
12
Algorithmic Aside Massively parallel molecular dynamics
Domain decomposition - large systems Replicated data - long simulation times
Easily implementedEqually effective for short- and long-ranged forcesDoes not scale well for large numbers of molecules
Scales well for homogeneous systemsDifficult to program/implementBest suited to short-ranged forces
13
Algorithmic Aside
Time-complexity bottleneck
Number of Atomic Units
Increasing Massively Parallel Compute Power
Replicated Data
Domain Decomposition
Global communications
limit
Challenging problemsin physics, chemistry and biology
€
1 sec simulated ≅1015 MD steps
1 year = 32 ×106sec ≅107sec
message passing time ≅10−5sec
Max #time steps/year ≅1012
Size is easyTime is hard
14
Algorithmic Aside Algorithm/theory vs hardware
David Landau, U. of Georgia
•1970 •1975 •1980 •1985 •1990 •1995 •2000•1
•10
•100
•1000
•10000
•100000
•1000000
•1E7
•1E8
•1E9
•1E10
•relative performance
•computer speed
15
Outline of Talk
Introduction Theory, Modeling and Simulation in Nanoscience ORNL Center for Nanophase Materials Science and
Nanomaterials Theory Institute Molecular electronics
Multiscale modeling Nanoconfined fluid rheology and structure
Rheology of nanoconfined alkanes Classical Simulations of Carbon Nanotubes
Sliding behavior of multiwall nanotubes Organic-inorganic nanocomposite materials
Multiscale modeling Conclusions
16
Molecular Electronics
Motivation Lithographic fabrication of solid-state silicon-
based electronic devices that obey Moore’s law is reaching fundamental and physical limitations and becoming extremely expensive
Self-assembled molecular-based electronic systems composed of many single-molecule devices may provide a way to construct future computers with ultra dense, ultra fast molecular-sized components
Possibility of fabricating single-molecule devices proposed theoretically by Ratner Aviram and Ratner, "Molecular rectifiers," Chem. Phys. Letts., 29, 283 (1974)
Landmark experiment Experimental measurement of conductance of 1,4-benzenedithiol between Au
contacts: Reed et al., "Conductance of a Molecular Junction," Science, 278, 252-254 (1997)
17
Molecular Electronics
Experimental uncertainties ~30 papers by young experimental star at Bell Labs found
to be fraudulent Experiments are difficult, usually not reproducible by other groups
Established results are being questioned
18
19
Molecular Electronics Experiment vs. theory
3 orders of magnitude difference
Di Ventra, et al., Phys. Rev. Letts., 84, 979-982 (2000).
Au
Au
S
S
C
Reed et al., Science, 278, 252-254 (1997)
20
Molecular Electronics
Closing the gap Cui et al. experiment
Ensures bonded contact at each end of octanethiol molecule
Conductance measured in integer increments corresponding to 1-5 contacts
– Single-molecule conductance inferred
– Reduces difference between theory and experiment to less then one order of magnitude
Cui, et al., Science, 294, 571-574 (2001)
Kipps, Science, 294, 536-537 (2001)
21
Molecular Electronics
Multi-level approach Ab initio calculations to characterize gold-BDT interaction Structure of self-assembled monolayer (SAM) of BDT on
Au(111) surface Ab initio calculation of conductance using structure within
SAM
Parameterization of the molecular potentials (structure calculation)
SAM structure determination (molecular dynamics)
Conductance and effects beyond DFT (electronic structure, correlation methods, finite bias, ...)
22
GEMC results
Condensed phase of benzenethiol molecules adsorbed
on 111 Au surface (T=298K)
High-resolution STM image of ordered structure of benzenethiol SAM. From Wan et al., J. Phys. Chem. B, 104 (2000) 3563-3569
Structure of benzenethiol SAM by GEMC: Probabil-ity density of sulfur atom around sulfur atom
Structure of benzenethiol SAM by GEMC: Lines connecting sulfur-occup-ied sites show same structure as STM
23
Molecular Dynamics Molecular configuration and BDT
dynamics behavior Fully occupied Au(111) surface Structure on surface
Partly influenced by details of Au-BDT interaction
– Dependent on basis function Largely determined by
intermolecular excluded volume interactions
– Average tilt (), azimuthal () and twist () angles quite independent of basis function
Well-ordered herringbone structure in 120° direction
S
S
Krstíc et al., “Computational Chemistry for Molecular Electronics,” Comp. Mat. Res., in press (2003); Leng et al., “Structure and dynamics of benzenedithiol monolayer on Au (111) surface,” J. Phys. Chem. B, in press (2003)
24
Electronic Structure Calculations
Significant advance in theory Closed form expression for
transmission with semi-infinite leads Krstíc et al., “Generalized
conductance formula for multiband tight-binding model,” PRB, 66, 205319 (2002)
Place pseudo-molecules between leads in calculation
– Consistency shows correctness
3-44-3
1-31-1
3-3
3-3
Total
3-610-3
10-2
10-1
Transmission
-0.250 -0.200 -0.150 -0.100
Band energy (a.u.)
“molecule 1” “molecule 2”
25
Outline of Talk
Introduction Theory, Modeling and Simulation in Nanoscience ORNL Center for Nanophase Materials Science and
Nanomaterials Theory Institute Molecular electronics
Multiscale modeling Nanoconfined fluid rheology and structure
Rheology of nanoconfined alkanes Classical Simulations of Carbon Nanotubes
Sliding behavior of multiwall nanotubes Organic-inorganic nanocomposite materials
Multiscale modeling Conclusions
26
0.1
1
10
100
103
104
105
106
1 100 104 106 108 1010 1012
Granick Experiment
Viscosity/ cP
/s-1γ
Experimentally see an increase in viscosity of the confined fluid of several orders of magnitude above the bulk value Dodecane confined between mica sheets at ambient conditions
Rheology of Confined Dodecane
Hu et al., Phys. Rev.Letts., 66, 2758-2761 (1991)
Experimental bulk value
.
27
Hard Disk Drive Lubrication
http://alme1.almaden.ibm.com/sst/storage/hdi/interaction.shtml
28
Intramolecular interactions
Intermolecular interaction
Bond stretching Bond bending
Torsional potential
r
Lennard-Jones
ubs =12kbs r −req( )
2
ubb =12kbbθ−θeq( )
2
uts =c0 +c1 1+cosφ( )
+c2 1−cos2φ( )+c3 1+cos3φ( )
uLJ =4εσr
⎛ ⎝ ⎜
⎞ ⎠ ⎟
12
−σr
⎛ ⎝ ⎜
⎞ ⎠ ⎟
6⎡
⎣ ⎢ ⎢
⎤
⎦ ⎥ ⎥
Models and Methodology
29
-100
-50
0
50
100
150
200
250
0 200 400 600 800
Lower wall displacementUpper wall displacementNet wall displacement
Displacement
Time/ ps
Constant applied shear stress
Displacement under Constant Shear Stress
Dodecane confined between mica sheets Six-layer gap (2.36 nm) Shear stress 2.8 x107 N/m2
30
Sliding Behavior Dodecane confined between mica sheets
Six-layer gap (2.36 nm) Shear stress 2.8 x107 N/m2
QuickTime™ and aCompact Video decompressorare needed to see this picture.
31
Stick-Slip Behavior Dodecane confined between mica sheets
Six-layer gap (2.36 nm) Shear stress 2.8 x106 N/m2
QuickTime™ and aCompact Video decompressorare needed to see this picture.
32
Stick-Slip Behavior Dodecane confined between mica sheets
Six-layer gap (2.36 nm) Shear stress 2.8 x106 N/m2
270 ps
33
0.1
1
10
100
103
104
105
106
1 100 104 106 108 1010 1012
Granick Experiment
Viscosity/ cP
/s-1γ
0.1
1
10
100
103
104
105
106
1 100 104 106 108 1010 1012
Granick ExperimentSimulation - constant velocitySimulation - constant stress
Viscosity/ cP
/s-1γ
Comparison of Simulation and Experiment
Strain rate dependent viscosity for confined dodecane
S. T. Cui, C. McCabe, P. T. Cummings, H. D. Cochran, J. Chem. Phys., 118 (2003) 8941-8944
0.1
1
10
100
103
104
105
106
1 100 104 106 108 1010 1012
Granick ExperimentSimulation - constant velocitySimulation - constant stressSimulation - bulk viscosity
Viscosity/ cP
/s-1γ
34
Rheology of Confined Alkane Liquids
Klein and Kumacheva, Science 269, 816 (1995); J. Chem. Phys. 108, 6996 (1998); ibid. 108, 7010 (1998) OMCTS and cyclohexane confined between mica sheets in
surface force apparatus When film thickness is six layers or less
Dramatic increase in viscosity of six to seven orders of magnitude Ability to sustain a non-zero yield stress (solid-like behavior)
When film thickness seven layers or more No yield stress (fluid-like behavior)
35
Solidification Behavior
Cui et al., J. Chem. Phys., 114 (2001) 7189–7195
Dodecane confined between mica sheets Liquid-like behavior for 7 layers
of fluid or more (no yield stress) Solid-like behavior (yield stress ~
106 N/m2) at 6 or less layers of confined fluid
First simulation studies consistent with experiments of Klein
Completely consistent with corresponding states theory for shift in freezing temperature under confinement Radhakrishnan et al., JCP, 112,
11048 (2000)
36
Origin of Solidification Behavior
Dodecane Dodecane
Mica
Mica
Dodecane
P = 0.1 MPabulk
P < 0.1 MPa
Mica
Mica
DodecaneP = 0.1 MPaconfined
compress
Dodecane
Dodecane
37
εwf
εCH
2
No. ofLayers
ρconfined
(g/cc)1.00 6 0.714 ρconfined < ρbulk-liquid
4.47 8 0.785 ρbulk-liquid < ρconfined < ρbulk-solid
4.47 7 0.829 ρbulk-liquid < ρconfined < ρbulk-solid
4.47 6 0.850 ρconfined > ρbulk-solid
4.47 3 0.881 ρconfined > ρbulk-solid
Origin of Solidification Behavior
At 0.1 MPa: bulk-liquid = 0.75 g/ccbulk-solid = 0.84 g/cc
38
Outline of Talk
Introduction Theory, Modeling and Simulation in Nanoscience ORNL Center for Nanophase Materials Science and
Nanomaterials Theory Institute Molecular electronics
Multiscale modeling Nanoconfined fluid rheology and structure
Rheology of nanoconfined alkanes Classical Simulations of Carbon Nanotubes
Sliding behavior of multiwall nanotubes Organic-inorganic nanocomposite materials
Multiscale modeling Conclusions
39
Experimental
Separating inner shells Cumings and Zetl, Science 289, 602
(2000).
MultiwalledLength 330 nmDynamic Strength0.43 MPa
40
Experimental Antecedents
Separating inner shells Yu, Yakobson, and Ruoff, J. Phys. Chem B, 104, 8764 (2000).
MultiwalledLength 7,000 nmDynamic Strength0.08 MPa
41
Theoretical predictions Oscillatory Behavior
Zheng and Jiang, Phys. Rev. Lett. 88, 045503 (2002).
<-force
released
freq length-1€
f =14
c1ΠΔ
1L
42
This Work
Molecular dynamics NVT ensemble Multiple time step algorithm (r-RESPA)
Slow Time step of 2.2 fs
T = 298.15 K, vacuum Spherically truncated Lennard-Jones potential
Cutoff radius of 13.6 Å
Potential: DRIEDING model Guo, Karasawa, and Goddard, Nature 351, 464 (1991). Fully flexible model composed of Lennard-Jones sites. Used to predict packing structures in fullerenes, pure and doped
carbon nanotubes
Rivera, J. L., McCabe, C., and Cummings, P. T., Nano Letters, 3 (2003) 1001-1005 [Highlighted by Phillip Ballin Nature Materials, June 12, 2003]
43
This Work
Systems Studied Double-wall carbon nanotube Chiral conformation: (7,0) / (9,9)
Incommensurate system Superlubricity
Interlayer distance: 0.34 nm Outer diameter: 1.22 nm Lengths: 12.21 - 98.24 nm
44
Damped Oscillations - 24.56 nm
QuickTime™ and aYUV420 codec decompressorare needed to see this picture.QuickTime™ and aYUV420 codec decompressorare needed to see this picture.
45
Damped Oscillations - 24.56 nm
QuickTime™ and aYUV420 codec decompressorare needed to see this picture.QuickTime™ and aYUV420 codec decompressorare needed to see this picture.
46
Model Predictions
12.21 nm
24.56 nm
47
Model Predictions
49.27 nm
36.92 nm
48
Outline of Talk
Introduction Theory, Modeling and Simulation in Nanoscience ORNL Center for Nanophase Materials Science and
Nanomaterials Theory Institute Molecular electronics
Multiscale modeling Nanoconfined fluid rheology and structure
Rheology of nanoconfined alkanes Classical Simulations of Carbon Nanotubes
Sliding behavior of multiwall nanotubes Organic-inorganic nanocomposite materials
Multiscale modeling Conclusions
49
H
O
Si
Multiscale Modeling of Hybrid Nanostructured Materials
Polyhedral oligomeric silsesquioxanes (POSS) Initial focus on cubic POSS as basic nano building block
(HSiO1.5)8
Most experimental data Extremely versatile
Functionalized in many ways– Functionalization affects solubility,
diffusivity, rheology,… Cross-linked to create network structures “Alloyed” with polymer
– Nanocomposites Can be synthesized on large scale
Hybrid Plastics
[(RSiO1.5)8]
Si
Si
Si
Si
Si
Si Si
Si
O
O
O
O
O
O
O
O
O
O
O
O
RR
R
R
R
R R
R
50
Ab initioForce field
developmentAtomistic simulation
Coarse-grained
simulation
Molecular theory
Cummings(Vanderbilt)
Glotzer(Michigan)
Kieffer(Michigan)
Neurock(Virginia)
McCabe(CSM)
Multiscale Modeling of Hybrid Nanostructured Materials
NSF NIRT project DMR-0103399
51
Ab initio studies of POSS
Structure of POSS(H)8 cube
Exp.Plane Wave(VASP)
AtomicOrbital(DMOL)
RHF(cc-pVdz)
(GAUSSIAN 98)
Molecular Mechanics
UFF Cerius2
CompassKeiffer
RFF
Si-O (Å) 1.619 1.630 1.654 1.650 1.592 1.624 1.64
Si-H (Å) 1.48 1.463 1.481 1.462 1.470 1.473 1.47
Si-O-Si 147.5˚ 146.7˚ 145.9˚ 148.7˚ 146.8˚ 146.9˚ 146.1˚
O-Si-O 109.6˚ 109.6˚ 109.6˚ 109.1˚ 110.0˚ 110.2˚ 109.6˚
52
Ab initio studies of POSS
Low energy (4eV) collision of atomic O with a POSS Atomic O inserted in POSS structure:
Insertion in O-H bond (yielding a silanol)
Torsional energy profile along dihedral angle Si-C-C-C in propyl- and butyl-POSS
QuickTime™ and aCinepak decompressorare needed to see this picture.
Dihedral Energy Profile
0
1
2
3
4
5
6
7
0 30 60 90 120 150 180
Rotation Angle
Energy (kcal/mol)
T-propyl
n-butane
T-butyl
53
Conclusions
Theory, modeling and simulation (TMS) play vital role in nanoscale science and engineering Interpretation of experiments Design of experiments Characterization and design of nanostructured materials Design and control of manufacture
TMS in nanoscale science and engineering Typically requires many different techniques Future advances in field will result from development of
additional methods Multiscale methods, electron transport dynamics, optical
properties, self-validating forcefields,…
54
QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.
http://www.cnms.ornl.gov
55
http://www.mines.edu/Academic/chemeng/fomms
FOMMS 2003Keystone Resort, CO
July 6-11, 2003 Second international conference in series
Applications and theory of computational quantum chemistry and molecular simulation integrated with process and product design
Topics of special interest include: Nanoscale systems Molecular Materials Design Conceptual Chemical Process Design Molecular Scale Reaction Engineering Molecular Rheology Multiple Time Scale and Mesoscopic
Simulation Techniques Future Trends in Molecular Modeling,
Simulation and Design
