Adri van DuinProfessor in Mechanical Engineering

Director, Materials Computation Center

Dept. of Mechanical and Nuclear Engineering

Dept. of Chemical Engineering

Dept. of Engineering Science and Mechanics

Penn State University, 240 Research East Building

phone: 814-8636277;

E-mail acv13@psu.edu

The ReaxFF method and its applications to atomistic-scale

simulations on atomic-layer deposition and chemical vapor

deposition in complex 2D-materials

2DCC Webinar

Sept 25, 2018

1Work partially funded through NSF award DMR-1539916

mailto:acv13@psu.edu

Current Penn State group members and projects

Postdoctoral staff

Dr. Yun-Kyung Shin Metal alloys, Sulfur-embrittlement, Proteins

Dr. Nadire Nayir 2D-materials

Dr. Weiwei Zhang Fuel cells, proton transfer

Dr. Chen Chen Fuel cells, industry projects

Dr. Dundar Yilmaz Ferroelectrics, polymers

Dr. Malgorzata Kowalik Carbon fibers

Dr. Chowdhury Ashraf Combustion, carbon materials

PhD-students

Jamil Hossain Battery interface simulations

Dooman Akbarian Ferroelectrics, polymerization

Abhishek Jain Atomistic and continuum scale combustion

Seung Ho Hahn Silica based glasses, treibochemistry, leaching

Kate Penrod Electron transfer reactions in water

Behzad Damirchi Carbon materials

Sharmin Shabnam Combustion

Nabankur Dasgupta Polymer hydrolysis,, supercritical water

Mert Sengul Machine learning, cold sintering

Siavash Rajabpour Carbon fibers, chemical vapor deposition

Karthik Ganeshan MXene/water interfaces

2

Outline

- The ReaxFF reactive force field

- Overview of ReaxFF applications

- Applications to MXenes

- Applications to MoS2 CVD growth

- Summary

ReaxFF MD simulation

of S2 gas reacting with a

MoO3 slab at T=1000K

ReaxFF MD simulation of the

indentation of a Ni-slab with a

diamond fragment (Tavazza et al.,

J.Phys.Chem C 2016, 119, 13580ReaxFF MD simulation

of char combustion at

T=2500K

Material Computation Center (MCC) within Penn State

Material Research Institute (MRI)

MCC Four Lab Solution: Theory, Synthesis,

Fabrication, Characterization

Evaluate

material

candidates

Nanofab

Material Characterization Lab

Fabricate

materials

Characterize

materials

- MCC allows

Nanofab/MCL/2DCC to focus

on high-probability materials

- Simulation is relatively

inexpensive – allows testing

of out-of-the-box concepts

2DCC

Synthesize

materials

MCC has responded to over 1500 requests for software and/or parameter sets

through the MCC-website query form – started July 2015.

MCC Software/parameter distribution

Quantum mechanics

(1-1000 atoms)

Empirical force

fields (1-108 atoms)

Grain/Grain

boundaries

Multigrain

Design

Phase field,

CALPHAD

CFD, Finite

element

Length scales in Material Computation

Reactive force fields

Reactive

force fields

-To get a smooth transition from nonbonded to single, double and

triple bonded systems ReaxFF employs a bond length/bond order

relationship [1-3]. Bond orders are updated every iteration.

-All connectivity-dependent interactions (i.e. valence and torsion

angles, H-bond) are made bond-order dependent, ensuring that their

energy contributions disappear upon bond dissociation.

- Nonbonded interactions (van der Waals, Coulomb) are calculated between

every atom pair, irrespective of connectivity. Excessive close-range

nonbonded interactions are avoided by shielding.

- ReaxFF uses EEM, a geometry-dependent charge calculation

scheme that accounts for polarization effects [4].

Key features of ReaxFF

1. Brenner, D. W., (1990) Physical Review B 42, 9458-9471

2. Tersoff, J., (1988) Physical Review Letters 61, 2879-2882.

3. Abell, G. C., (1985) Physical Review B 31.

4. Mortier, W. J., Ghosh, S. K., and Shankar, S. (1986) JACS 108, 4315-4320.

Calculation of bond orders from interatomic distances

Introduction of bond orders

6,

4,

2,

5,

3,

1,

'

exp

exp

exp

bo

bo

bo

p

o

ij

bo

p

o

ij

bo

p

o

ij

boij

r

rp

r

rp

r

rpBO

Sigma bond

Pi bond

Double pi bond

0

1

2

3

1 1.5 2 2.5 3

Interatomic distance (Å)

Bond o

rder

Bond order (uncorrected)

Sigma bond

Pi bond

Double pi bond

6,

4,

2,

5,

3,

1,

'

exp

exp

exp

bo

bo

bo

p

o

ij

bo

p

o

ij

bo

p

o

ij

boij

r

rp

r

rp

r

rpBO

Sigma bond

Pi bond

Double pi bond

6,

4,

2,

5,

3,

1,

'

exp

exp

exp

bo

bo

bo

p

o

ij

bo

p

o

ij

bo

p

o

ij

boij

r

rp

r

rp

r

rpBO

Sigma bond

Pi bond

Double pi bond

0

1

2

3

1 1.5 2 2.5 3

Interatomic distance (Å)

Bond o

rder

Bond order (uncorrected)

Sigma bond

Pi bond

Double pi bond

Reaction barriers for concerted reactions

0

10

20

30

40

50

60

70

Reaction coordinate

En

erg

y (

kca

l/m

ol)

water 2

water 3

water 4

water 5

water 6

0

10

20

30

40

50

60

70

Reaction coordinate

En

erg

y (

kca

l/m

ol)

water 2

water 3

water 4

water 5

water 6

QM ReaxFF

Neutral

ReaxFF barrier for Grob

fragmentation (collaboration with

John Daily, Boulder). QM barrier:

65 kcal/mol (Nimlos et al., JPC-A

2006)

General rules for ReaxFF

- MD-force field; no discontinuities in energy or forces even during

reactions.

- User should not have to pre-define reactive sites or reaction

pathways; potential functions should be able to automatically handle

coordination changes associated with reactions.

- Each element is represented by only 1 atom type in the force field;

force field should be able to determine equilibrium bond lengths,

valence angles etc. from chemical environment.

0.01

0.1

1

10

100

1000

10000

100000

1000000

0 100 200 300 400

ReaxFF

QM (DFT)

Nr. of atoms

Tim

e/ite

ration (

seconds)

ReaxFF Computational expense

x 1000,000

-ReaxFF allows for

reactive MD-simulations

on systems containing

more than 1000 atoms

- ReaxFF is 10-50 times

slower than non-reactive

force fields

- Better scaling than QM-

methods (NlogN for

ReaxFF, N3 (at best) for

QM

- ReaxFF combines covalent, metallic and ionic elements allowing

applications all across the periodic table

- All ReaxFF descriptions use the same potential functions, enabling

application to interfaces between different material types

- Code has been distributed to over 750 research groups

- Parallel ReaxFF (LAMMPS/ReaxFF) available as open-source

- Incorporated into the ADF/BAND graphical user interface

not currently

described by

ReaxFF

Current development status of ReaxFF

ReaxFF transferability

High energy materials

Batteries

2D-materials

Vp = 3 km/s

RDX dissociation channels

(Strachan et al. JCP 2005)Comparison with experiment – shock velocity and carbon

clustering (Strachan et al. PRL 2013; Zhang et al. JPC-A 2009) Void effects on HE-response

(Nomura et al. PRL 2007)

Li-migration around S4(Islam et al. PCCP 2015)

Li-etching of a defected, strained carbon

nanotube (Huang et al APL 2013)

Li-migration in a carbon onion anode

(Raju et al. JCTC 2015)

Stability of various MoS2 defects

(Ostadhossein et al. in progress)

Comparison of c-lattice expansion for MXenes with

DFT and experiment (Osti et al. ACS-AMI 2016)

High-speed collision of a silica

nanoparticle on graphene (Yoon et al,

Carbon 2016)

Applications to catalytic carbon-growth on Ni-surfaces

Development of Ni/hydrocarbon ReaxFF [1]

Integration of force-

biased Monte Carlo [2]

Electric field effects on

CNT-growth [3]

Ar-bombardment

enhanced surface

catalysis[4]

(1) Mueller, J. E.; van Duin, A. C. T.; Goddard, W. A., III.

Journal of Physical Chemistry C 2010, 114, 4939.

(2) Neyts, E. C.; Shibuta, Y.; van Duin, A. C. T.; Bogaerts,

A. ACS Nano 2010, 4, 6665.

(3) Neyts, E. C.; van Duin, A. C. T.; Bogaerts, A. Journal of

the American Chemical Society 2012, 134, 1256.

(4) Neyts, E. C.; Ostrikov, K.; Han, Z. J.; Kumar, S.; van

Duin, A. C. T.; Bogaerts, A. Physical Review Letters 2013,

110, 065501.

Collaborations with Jonathan Mueller (Caltech, currently U.Ulm) and Erik Neyts (U. Antwerp)

(a) Structures of MAX and MXene phases during MXene synthesis1

(b) Accordion-like structure of Ti3C2Tx MXene2

16 of 41

Introduction (5/7)

Next: Effect of cation

MXenesThe novel material for electrodes

1M. Naguib, V. N. Mochalin, M. W. Barsoum, and Y. Gogotsi, Adv. Mater. 26, 992 (2014). 2F. Wang, C. H. Yang, M. Duan, Y. Tang, and J. F. Zhu, Biosens. Bioelectron. 74, 1022 (2015).

M – early transition metal

A – group 13 or 14 metal (e.g. Al)

X – carbon or nitrogen

Introduction to MXenes

Qualitative comparison of properties with carbon-based materials1

17 of 41

Introduction (6/7)

Next: Effect of cation

MXenesQualitative comparison to carbon-only materials

1P. Simon, Y. Gogotsi, and B. Dunn, Science (80-. ). 343, 1210 (2014).

Introduction to MXenes

Experiment

DFT ReaxFF

Applications of ReaxFF to MXenes

Ti/C MXene defect structure formation and evolution

- MXenes defects show ‘cannibalistic’ growth patterns – material moves

away from the defect sites to deposit additional TixCy patterns

2D-MXene reactions with Urea

- ReaxFF simulations explain high activity of Ti/C/OMXene surfaces to urea conversion

– significantly faster than urea conversion in the gas phase or in water.

- Overbury et al. J.Am.Chem. Soc. 2018, published online, DOI: 10.1021/jacs.8b05913

ReaxFF Experiment

MXene conversion to TiO2/graphene during oxidation

- ReaxFF simulations enable us to follow the conversion of MXenes into

thermodynamically more stable graphene/TiO2 materials

- Lotfi, R., Naguib, M., Yilmaz, D., Nanda, J. and van Duin, A.C.T. (2018) Journal of

Materials Chemistry A 6, 12733-12743.

Development and applications to MoS2

Alireza Ostadhossein, Ali Rahnamoun, Peng Zhao, Sulin Zhang, Yuanxi Wang and Vin Crespi

Bending of pristine and vacancy/mismatched MoS2

Ostadhossein, A., Rahnamoun, A., Wang, Y., Zhao, P., Zhang, S., Crespi, V. H., and van Duin, A. C. T., 2015. ReaxFF

Reactive Force-Field Study of Molybdenum Disulfide (MoS2). Journal of Physical Chemistry Letters manuscript in

preparation.

Comparison of ReaxFF and

DFT Vacancy energies for

MoS2

Extension to MoS2/graphene interfaces

Chowdhury Ashraf and Sungwook Hong (currently USC)

MoS2-Graphene project: Binding energy of S on graphene sheet

Graphene with

monovacancies (C196)

C196S8: Nucleation of S8 on

graphene sheet (11.75 ps)

C195S6: CS2 (in circle) leaving the

graphene sheet (112.25 ps)

S Growth on Graphene Sheet with Vacancies

C195S10 (316.75 ps)C195S30 : Largest S cluster (413 ps)

C195S20 (425 ps)C195S14 (922 ps)

MD vs. fbMC/MD • Alternative MD/fbMC provides better crystallinity, compared

to pure MD

25

Pure MD MD/fbMC

top

side

16S8+ 32Mo in 80×80×80 Å3 box

MoS2 crystallite expansion using the fbMC/MD method

with S/Mo atom addition

Simulations were started by running

10000 MD steps and 10000 FBMC

steps after adding one Mo atom and

three S atom (Atom addition

section).

After adding all components, MD

Simulation was run for 500000

iterations. Then FBMC Simulation

was run for 500000 iterations

(Equilibrium section).

Simulations were run at three

temperatures of 800, 1000,1200 K.

Roghayyeh Lotfi

Side view of growth (atom

addition section only)

MoS2 Growth from S/Mo by FBMC/MD method at 1000 K

Top view of growth (atom addition and

equilibrium sections)

Defect Design and Functionalization in 2D-materials

Dundar Yilmaz, Roghayyeh Lotfi, Chowdhury Ashraf, Sungwook Hong and Adri van

Duin, Journal of Physical Chemistry C 2018, 122, 11911

A technique similar to ”Potato Stamp” can be used to create sulfur vacancy

defects on the MoS2 surfaces. Later these defects can be functionalized

with – for example - small epoxy molecules.

4,000 atoms LAMMPS/ReaxFF simulations (4 cores)

a

c

90%

83%

b

500nm5 μm

Synthesis of MoS2 on hBN with full orientation control. (a) Schematic of PVT system. (b) SEM image of triangular MoS2flakes epitaxially grown on mechanically exfoliated hBN, on a Si/SiO2 substrate. A step edge separates two regions, each

with 83% or 90% of the flakes at the same orientation. Inset shows the same image color-coded by orientation. (c) TEM

image of triangular MoS2 flakes grown on freestanding ME-hBN where crystallinity and alignment with the hBN substrate

are verified by the annular dark field (ADF-) STEM image of a Mo-terminated MoS2 edge and the selected area electron

diffraction from the circled area.

MoS2 growth on Boron Nitride (hBN)with Wei Zhang, Yuaxi Wang and Vin Crespi

Pair binding energy (eV)

Heterostack

Frenkel pairs

Adatom in layer 1

Vacancy in layer 2

Pair binding energy (eV)

Among all defect pair binding energies, a Moad+VB complex is the strongest.

Stable defect pairings in neighboring layers of 2D materials are likely

Frenkel pairs – an adatom in one layer binds strongly to a vacancy in

the other layer.

H

BN

Mo on MoS2

2.90Å

Mo on B vacancy in hBN

2.15Å

Mo interstitial

5.05Å

Periodic MoS2 on hBN

Mo

S

B

N

A Mo interstitial atom sandwiched between pristine MoS2 and hBN, with Mo above a boron vacancy, equilibrates to 5.05

Å interlayer spacing, close to the 4.96 Å of pristine MoS2 on pristine hBN. The individual separations of Mo from each of these sheets in isolation also sum to essentially the same value. Thus Mo+VB on hBN can nucleate the growth of an

MoS2 overlayer with no significant deformation of the surrounding ideal bilayer structure.

Angle θ

Triangle centered on:

Energies of finite MoS2 flakes on monolayer h-BN with boron vacancy and Mo interstitial (black) and without (colored, scattered plots).

h1 h2

doped Mo on hollow site

doped Mo on metal site

m1m2

DFT: 12.39kcal/mol

ReaxFF: 10.41kcal/mol

DFT: 0.00kcal/mol

ReaxFF: 0.00kcal/mol

DFT: 20.33kcal/mol

ReaxFF: 7.63kcal/mol

DFT: 0.10kcal/mol

ReaxFF: -2.23kcal/mol

Comparison between DFT and ReaxFF

Simulation strategy:

Simulations temprature:1200 K, Mo:S ratio=1:2

Before adding, simulations were done by

running 10000 MD steps and 10000 fbMC steps

(5 cycles MD/fbMC).

After one Mo and two S atoms adding, 10000

steps of MD simulation and then 10000 steps of

fbMC simulation (5 cycles MD/fbMC).

Note: h-BN with high density of Mo-doping,

high frequency adding Mo/S atoms randomly on

top of h-BN sheet (fix).

MoS2 growth from adding Mo/S by ReaxFF MD/fbMC simulation

Periodic h-BN sheet with Mo(S)-doping

<

DFT shows the right structure is more stable (top view).

※Left two pictures show the final 2D-structure.

※ The key point for the growth is the formation

of Mo-Mo-Mo triangle structure. Once it formed,

the rotation of MoS2 becomes difficult.

MoS2 growth from adding Mo/S by ReaxFF MD/fbMC simulation

1st 2nd

DFT (side view)

Proposed growth process to build a layer of 2D-MoS2 on h-BN in CVD

3-Mo(triangle) 4-Mo atoms 5-Mo 6-Mo 7-Mo

7-Mo relax (hexagon)2D-MoS2 island formed

MoS2 on h-BN with

Mo-doping (top view)

ReaxFF simulation shows Mo/S species are likely to first aggregate around Mo-sites of BN sheet,

and then grow up to get large area 2-dimentional MoS2(without rotation).

expansion

ReaxFF development for W(CO)6/H2Se CVD mixtures

Roghayyeh Lotfi, Yuanxi Wang and Dundar Yilmaz

- Fairly complex chemistry

– not all CO ligands

dissociate directly, first

several Se-groups have

to bind to the metal center

- After H2Se binding, H2release is exothermic –

likely end product

WSe4H2 or similar

species

ReaxFF connection to CFD (Yuan Xuan-group)

CFD Reacting Simulation resultsAbhishek Jain and Yuan Xuan

W(CO)6

5x10-4

0

W(CO)4(Se)2

W(CO)2(SeH)2(Se)2

W(SeH)2(Se)2

4x10-6

0

4x10-6

0

3x10-6

0

- ReaxFF trained CFD can give detailed gas-phase composition predictions,

which can feed back into experimental CVD chamber design and operation

ReaxFF CFD

W(CO)6 in H2Se

Reaction barriers, pre-exponential factorsW(CO)6

W(SeH)2(Se)2

Sample stage

UQ, weights - go beyond DFT?

Experiment

ReaxFF MD/MC

Hyperdynamics

Crystal seeds, grain

structures vs. time

Morphology

recognition

Machine

learning

Long time behavior

DFT

Multi-scale simulation concept for predicting chalcogenide growth

- ReaxFF has proven to be transferable to a wide range of materials

and can handle both complex chemistry and chemical diversity.

Specifically, ReaxFF can describe covalent, metallic and ionic materials

and interactions between these material types.

- The low computational cost of ReaxFF (compared to QM) makes the

method suitable for simulating reaction dynamics for large (>> 1000

atoms) systems (single processor). ReaxFF has now been parallelized,

allowing reactive simulations on >>1000,000 atoms.

: not currently

described by

ReaxFF

Summary

Collaborators:

- Vin Crespi, Sulin Zhang, Susan Sinnott, Yuanxi Wang (Penn State)

- Kimberley Chenoweth, Vyacheslav Bryantsev and Bill Goddard (Caltech)

- Aidan Thompson, Steve Plimpton (Sandia), Ananth Grama (Purdue),

Metin Aktulga (Purdue) (parallel MD)

Funding: - PSU/KISK startup grant #C000032472

- Illinois Coal grant ICCI 10/7B-3

- NSF (TiO2/water, PdO/Ceria, 2DCC)

- NETL/RUA (Fuel catalysis)

- DoE/NETL (Refractory materials)

- AFRL/SBIR (Hydrocarbon cracking)

- AFOSR/MURI (O-resistant materials)

- DoE/EFRC FIRST-center

- Exxon (Software development, catalysis)

- British Royal Society (initial ReaxFF funding)

Acknowledgments

Websites: http://www.engr.psu.edu/adrihttp://www.rxffconsulting.com

Office: 240 Research East

Phone: 814-863-6277

E-mail: acv13@psu.edu

More

information:

Parallel ReaxFF simulation of

hydrocarbon cracking (4800 atoms, 4

processors)

http://www.engr.psu.edu/adri