Reactive Molecular Dynamics Andrei Smirnov Rolando A. Carreno-Chavez Jaggu Nanduri http://nift.wvu.edu/remody

1
Reactive Molecular Dynamics Andrei
SmirnovRolando A. Carreno-ChavezJaggu Nanduri
http//nift.wvu.edu/remody
2
Reactive Molecular Dynamics

1. Problem and objective, relation to the long term
   goal
2. Methodology
3. Problems, issues, and solutions
4. Accomplishments and results
5. Future work and direction

3
Problem and Objective

• Bridging the Nano and Macro Scales
• Atomistic Modeling lt1nm 1nm
• Molecular Dynamics 1nm 1mc
• Continuum Modeling 1mc - 1cm
• Reducing the number of empirical constants
• Predicting kinetic reaction rates
• Porous medium diffusion constants
• Capturing new effects
• In-pore kinetic reactions at microscale
• Transient concentration polarization effects

4
Methodology
Kinetic/Collision Theory

• http//en.wikipedia.org/wiki/Collision_theory

Nwall 0.25 vavg N/V 0.25den/m(8kT/pim)(1/
2)?
Reactions in the bulk
Reactions on the surface
YSZ
Ni, O--
5
Input Parameters
Species

• Atomic mass
• Size
• Heat capacity DOF

Reactions

• Activation energy
• Enthalpy

Species and reactions data above can be specified
both on the boundaries and in the bulk.
6
Problems and Solutions
Memory Efficient Implementation
Advanced data structures to accommodate several
million molecules on a single processor vs.
several hundred in QM calculations.
Time-efficient Execution Algorithm
Space segmentation algorithm to exploit the
local nature of the interaction potential.
Adaptive time-stepping.
7
Looped Lined Lists
Enable to avoid memory allocations and
deallocations. Instead.
8
Looped Lined Lists
Enable to avoid memory allocations and
deallocations. Instead.
9
Looped Lined Lists
Enable to avoid memory allocations and
deallocations. Instead new links are created.
10
Looped Lined Lists
Enable to avoid memory allocations and
deallocations. Instead new links are created and
old links are reassigned.
11
Looped Lined Lists
Enable to avoid memory allocations and
deallocations. Instead new links are created and
old links are reassigned.
12
Looped Lined Lists
Enables more efficient nested looping over
neighboring particles.
13
Looped Lined Lists
Enables more efficient nested looping over
neighboring particles.
14
Looped Lined Lists
Enables more efficient nested looping over
neighboring particles.
15
Interaction Acceleration
Space segmentation scheme Enables to achieve
near linear dependence of execution time on the
number of molecules.
16
Species/Reactions
OOP Approach Implementing classes of Atoms,
Molecules, Species, and Reactions in a
object-oriented framework enabled flexible data
input and problem setup for hundreds of species
and reactions.
17
Syngas Gas Phase Reactions
H2CH3CH4H CH4OOHCH3 OHCH3CH4O CH4O2HO2
CH3 HO2CH3CH4O2 CH4OHH2OCH3 H2OCH3CH4OH C
O2HOHCO OHCOCO2H CO2OO2CO COOCO2 COO2
CO2O COHO2CO2OH
OHOHH2OO OHOHHO2H OHO2HO2O OHOO2H OHH
2H2OH OHHH2O OHHH2O OHHO2H2OO2 O2H2OOH
HO2 O2H2OHOH O2H2HO2H O2HOHO O2HHO2
OH2OHH OHOH OHO2OHO2 OOO2 HHH2 H2OHH
2OH H2OOHO2H H2OOOHOH HO2HH2O2 HO2HH2O
O HO2HOHOH CH4HH2CH3
18
Syngas Surface Reactions
H2-gtOHH H2O-gtOHCH O-gtO2 H-gtH2 H2O-gtH2OH H2O-gtHO
2H H2O-gtOHOH HO2-gtH2O2 HO2-gtH2OO HO2-gtOHOH CH
4-gtH2CH3 H2-gtCH4H CH4-gtOHCH3 OH-gtCH4O
CH4-gtHO2CH3 HO2-gtCH4O2 CH4-gtH2OCH3 H2O-gtCH4OH
CO2-gtOHCO OH-gtCO2H CO2-gtO2CO CO-gtCO2 O2-gtCO2O
HO2-gtCO2OH O-gtCO CO2-gtCOCO H2O-gtOHCH
OH-gtH2OO OH-gtHO2H O2-gtHO2O OH-gtO2H H2-gtH2OH O
H-gtH2O OH-gtH2O OH-gtH2OO2 O2-gtOHHO2 O2-gtOHOH O2
-gtHO2H O2-gtOHO O2-gtHO2
19
Accomplishments Results
10 mil/GB molecules on a single processor
Simulations in 1mc3 pore
1000 molecules H2O2 reaction.
15 mil H2 O(s) H2O
20
VALIDATION

• First validation of the Remody program
  (histogram) with Maxwell Boltzmann Velocity
  distribution for 10000 molecules of hydrogen at
  850 K.

Validation of the Remody program (histogram) with
Maxwell Boltzmann Velocity distribution for 10000
molecules of helium at 3000 K.
21
1 million - molecules
22
2 millions - molecules
23
3 millions - molecules
24
4 millions - molecules
25
5 millions - molecules
26
6 millions - molecules
27
7 millions - molecules
28
8 millions - molecules
29
9 millions - molecules
30
10 millions - molecules
31
11 millions - molecules
32
12 millions - molecules
33
13 millions - molecules
34
14 millions - molecules
35
15 millions - molecules
36
16 millions - molecules
37
17 millions - molecules
38
Concentrations
39
Syngas
1.5 million molecules in one cubic micron
40
Syngas
3.0 million molecules in one cubic micron
41
Syngas
4.0 million molecules in one cubic micron
42
Syngas
5.0 million molecules in one cubic micron
43
Syngas
4.0 mil molecules
44
Syngas
5.0 mil molecules
45
Accomplishments

1. The capability was developed to simulate tens of
   millions of reacting molecules on a single
   workstation.
2. Developed techniques enable to conduct
   simulations in nanometer-to-micron range,
   bridging the gap between ab-initio QM and
   continuum mechanics paradigms.
3. Hundreds of bulk gas and surface reactions can be
   easily incorporated.
4. H2 anode reactions inside one cubic micron pore
   were simulated.
5. Simulation of anode-Syngas reaction inside 1mc3
   pore, including 41 surface and 38 bulk reactions
   is continuing.

46
Future Work

1. Investigate transient effects in Syngas
   simulation in micron size pores.
2. Investigate transient effects of polarization.
3. Extend surface reaction model with surface
   species kinetics.
4. Extend simulations to larger size pores using a
   workstation cluster.
5. Investigate the effects of low ppm impurities on
   surface degradation.
6. Predict kinetic reaction rates.