LECTURE 11 DFT Supercell Calculations of Surfaces Fall Semester 2007 W. F. Schneider CBE 547 1
LECTURE 11
DFT Supercell Calculations of Surfaces
Fall Semester 2007W. F. Schneider CBE 547 1
2
Alkaline Earth Oxide Bulk Structures2
-250
LDA VASP
experiment
MgOBulk MgO
-350
-300
ergy
(eV
)
LDA VASPGGA VASPLDA CPMD
Bulk MgO• Rocksalt structure
Represent as• Simple cubic unit cell
-400
EneSimple cubic unit cell
• 8 atom basis
• Optimize lattice constant-450
3.5 4 4.5 5
lattice constant (Ang.)
Exp’t 4.212 Å
Optimize lattice constant
LDA Vasp 4.182
GGA Vasp 4.255
LDA CPMD 4.240a
Fall Semester 2007W. F. Schneider CBE 547
3
Supercell Slab Models of Surfaces3
(MgO bulk) (MgO slab)Mg
c
g
O
ab
c
ab
f fSelection of a surface model• Surface cleavage plane
• “Thickness” of slab
• Lateral cell dimensions
• Size of vacuum spacing
Side view of 3 slab images
Fall Semester 2007W. F. Schneider CBE 547
Top view of 4 “2x2” unit cells
4
Miller Index Naming for Surface Planes4
Specify cleavage plane in integer lattice vector coordinates of a vector orthogonal to the planeDescription thus depends on lattice vector choice
Complete definition also must include vertical location of planeS f ti hi hl d d t l lSurface properties highly dependent on cleavage plane
(001) (013)
z
y
x
z
Supercell model of a stepped surface“Basal” plane
Fall Semester 2007W. F. Schneider CBE 547
y
x(Lattice vectors become more messy to determine)
5
More Miller Planes5
(111)
z y xPolar surface
Opposite faces have different compositions
Typically high energy and unstable to “reconstruction”
Fall Semester 2007W. F. Schneider CBE 547
6
Supercell Dimensions6
Slab thickness typically a few “formula unit” layersVacuum spacing large enough to isolate slab from periodic image ( 10 15 Å)(~10-15 Å)Lateral dimensions based on need to isolate chemistry within a cellAlways test to be sure properties of interest are robust to model
a
2 atom
Nominally equivalent surface supercells
Different effective separation between
i hb
8 atom
4 atom
16 atom neighbors16 atom
Fall Semester 2007W. F. Schneider CBE 547
7
Surface Relaxation7
Exposed surfaces will often relax in response to loss of bonding at surfaceTypical to “freeze” one side of slab to simulate constraints of the bulkIdeally converge results with respect to slab thickness
0.01 Å 0.03 Å
0.004 Å0.01 Å
4.212 Å
FixedMgO
0.02 J/m2 relaxation MgO (103) step relaxations
Fall Semester 2007W. F. Schneider CBE 547
0.02 J/m relaxation MgO (103) step relaxations
8
Surface Energy8
(MgO bulk)
(MgO slab)slab bulkE nE−
ab
c
cslab bulk
2Aγ =
Energy cost to expose a surface by cleaving a bulk material
Always a positive quantityUseful for valdiation, experimental comparison
ab
Distance between steps
Comparison of energies between supercells of different shapes, requires
k-point convergenceEnergy cutoff convergenceEnergy cutoff convergenceEquivalent computational parameters
Slab k-pointsm × n × 1
Fall Semester 2007W. F. Schneider CBE 547
m, n inversely proportional to lateral dimensions
9
Molecular mechanism of trapping on catalyzed oxides9
NO + ½ O2 NO2
CO, H2, HC+
NO2
N2, CO2, H2O
SO2 + ½ O2SO3
Pt O
NO2
Pt O
NO2
2 2
O
SO3
O
SO3
Lean ConditionsNOx and SOx oxidized & trapped
Trapping oxide (e.g. BaO)
Rich ConditionsNOx decomposed and reduced, SOx recalcitrant
Trapping oxide (e.g. BaO)
Key Questions:How does NOx get oxidized and adsorbed on a trapping material?How does the SO chemistry compete with NOx?How does the SOx chemistry compete with NOx?How can we select trap materials to optimize selectivity for NOx over SOx?
This work: Compare SOx and NOx adsorption mechanisms on MgO(001)
Fall Semester 2007W. F. Schneider CBE 547
x x
10
Sulfur Oxides10
SO2
O2-
∠OSO = 106°S
S O ∠OSO
SO 2
OOS-O ∠OSO
Exp’t 1.43 118.9
LDA 1.449 119.4
½ O2
SO32-
(sulfite)
SO3 ∠OSO = 109 5°3 ∠OSO = 109.5
S-O
Exp’t 1.42
LDA 1.439
O2-
Fall Semester 2007W. F. Schneider CBE 547
SO42-
(sulfate)
11
MgO (001) + SO211
1.5021.47
117 0
1.5051.536
1.655
2.1321.755 2.0622.17117.0
2.211
2.132
Physisorbed SO2-16 kcal mol-1 LDA
Terrace chemisorbed SO2-42 kcal mol-1 LDA
2.448
Step chemisorbed SO2-62 kcal mol-1 LDA
Metastable dipolar physisorptionOnly structure found in cluster simulationsConverts to chemisorbed at ~200 K in MD simulations (Ea ≈ 0.5 kcal mol-1)
Strong “sulfite”-like chemisorption (MgO + SO2 MgO·SO2)Specific S–Os and O–Mgs interactions; significant charge transfer to adsorbatePronounced local lattice distortion, dies off over short range
SO2 chemisorbs preferentially on coordinatively unsaturated edge ionsSimilar sulfite-like adsorption geometry
Fall Semester 2007W. F. Schneider CBE 547
SO2 oxygen fill vacant lattice sitesIntroduces pronounced lattice distortions along step edge
11
12
SO2 Chemisorption on MgO12
S O O LDA GGA1 2 3 4
BE, kcal mol-1
S O1 O2 LDA GGA (100) -37 -25 (103) 4 1 3 -62 -46 1 1 3 -60
3 3 2 59
1 2 3 4
1
3 3 2 -59 1 1 2 -52 2 1 2 -46 2 3 2 -34 S
O
O
(104) 4 1 3 -62 corner -92 -81
Multiple SO2 chemisorption sitesSO2 fills “virtual” lattice sitesAdsorption energy tracks with lattice ion coordinative
Fall Semester 2007W. F. Schneider CBE 547
p gyunsaturation
13
MgO + SO313
1.470
1.454 1.4401.578
1.498
1.657 2.0662.603
2.420 Å
“Sulfate”-like chemisorption (MgO + SO3 MgO·SO3)
MgO (001) + SO3-69 kcal mol-1 LDA
MgO (103) + SO3-95 kcal mol-1 LDA
Primary S–Os bonding → approximately tetrahedral SO4 structures
Large, local relaxation of Os, charge transfer to SO3
Secondary O–Ms interactions fix orientation
Chemisorption energy ~30 kcal mol-1 greater than SO2
Fall Semester 2007W. F. Schneider CBE 547
Chemisorption energy 30 kcal mol greater than SO2
Strong preference for coordinatively unsaturated ions
14
MgO + SOx Vibrational Spectroscopy14
Fall Semester 2007W. F. Schneider CBE 547
15
NO•/NO2• Physisorption on MgO Terrace15
Physisorbed NO2
1.19
Physisorbed NO1.24
125º1 24
128º -0.23 e-0.26 e-0.19 e
2.27110º
2.252.25
1.24
MgO(001) + NO MgO(001) +NO2 O-downMgO(001) +NO2 N-down
NO and NO2 physisorb on MgO terraces
-7 kcal mol-1 -10 kcal mol-1-4 kcal mol-1
Low NOx Lewis basicity produces long, weak adsorbate-surface bondsCharge-dipole interaction, limited charge transfer to adsorbate
Not consistent with observations of “nitrite” and “nitrate” upon exposure of MgO to NOx!
Fall Semester 2007W. F. Schneider CBE 547
g x
1616
Fall Semester 2007W. F. Schneider CBE 547
17
Cooperative NO2 Adsorption on MgO(001)17
“nitrate”NO2
+ + Os2-
“nitrite”NO2
- + Mgs2+
1.28 1.28
e-
-0 23 e -0 73 e
1.41
1.22
2.322.09 2.11
1170.23 e -0.26 e
-1.22 e
-0.78 e0.73 e
-15 kcal mol-1
+
Isolated physisorbed NO2
Charge transfer generates cooperatively adsorbed NO2 pairStructural modifications consistent with chemisorptionMulliken populations reflect redistribution of charge between adsorbatesBinding energy enhanced by 15 kcal mol-1 (100%) over two isolated NO2!
Mixed nitrite/nitrate consistent with experimental observation for NO2-exposed M
Fall Semester 2007W. F. Schneider CBE 547
MgO
18
Distance Dependence of Cooperative Chemisorption18
“Lewis Basic” NO2
Unique “Lewis Acidic” NO2 Sites
OMg Mg O
MgO Mg O
2
253035
nerg
y
OMg
O Mg
Mg O Mg O
MgOA Mg O10152025
Ads
orpt
ion
en(k
cal/
mol
)
Mg O
O Mg
05
10
Tota
l A B C D E F oo
Adsorption site
Mg OB Mg O
MgOD Mg OE
Mg OC
OF Mg
Cooperative effect decreases slowly with adsorbate separatione- transfer/pairing can occur over several Angstroms
“3x3” MgO supercellAdsorption site
Fall Semester 2007W. F. Schneider CBE 547
/p g gForced charge separation yields physisorption
19
Bulk and Surface Structures of FCC Metals19
2.95
3
-5.6
-5.5
Lattice ConstantTotal energyÅ)
Bulk FCC Pt
2.7
2.75
2.8
2.85
2.9
-6.1
-6
-5.9
-5.8
-5.7Total energy
attic
e C
onst
ant (
Å
2.5
2.55
2.6
2.65
0 50 100 150 200-6.5
-6.4
-6.3
-6.2
k points
La
a aa
k-points
Theory Expt’
Lattice Constant Lattice Constant (Å)
3.986 3.923
Surface energy (J/m2)
1.48 --
Fall Semester 2007W. F. Schneider CBE 547
20
FCC (111) Surface Supercell Models20
11.271 Å8.453 Åa
3 Pt × 3 Pt 4 Pt × 4 Pt
1×1
2×2
20.9
21 Å
√3×√3
b a
c
1×2
-120
-100
-80
y (k
cal m
ol-1
)
NO
(a)4×4
-60
-40
-20Ad
sorp
tion
Ene
rgNO2
NO
Fall Semester 2007W. F. Schneider CBE 547
03 layer 4 layer 5 layer
A
21
Adsorption Sites on FCC (111) Surfaces21
Adsorption sites on the (111) facet:
bridge – 2-fold
hcp – 3-fold coordination
fcc – 3-fold coordination
atop – 1-foldcoordination
atop 1 fold coordination
Hexagonal arrays of sites
o DFT calculations performed with Vasp
o 3×3, 4×4, 5×5 Pt(111) supercell models
o 8×8×1, 6×6×1, and 4×4×1 Monkhorst‐Pack k‐point grids
o “PAW” electron core treatment
91 li d di i i
Fall Semester 2007W. F. Schneider CBE 547
o PW91 generalized gradient approximation
22
Oxygen adsorption and interactions on Pt(111)22
Low coverage (1/16 ML) O preference for FCC adsorption
fcc: -1.29 eV atop: +0.18 eVfcc-hcp bridge (TS)hcp: -0.89 eV
Pt(111) + ½ O2 < < <
Lateral interaction effects on O adsorption (4 x 4 supercell)
2.4 Å
fcc-fcc: 1NN fcc-fcc: 2NN fcc-fcc: 3NN fcc-hcp: 1NN fcc-hcp: 2NN
+0 20 eV +0 10 eV +0 01 eV +0 82 eV +0 32 eV
Fall Semester 2007W. F. Schneider CBE 547
+0.20 eV +0.10 eV +0.01 eV +0.82 eV +0.32 eV
Characterizing Pt(111)-O Surface Coverage
+Adsorbates
O atom
Pt atom
Clean Pt(111) surface p(1 × 1)—O orderingAdsorbatesinteract with each other to produce different stable configurations at different 60
80Atop onlyHCP onlydifferent
coverages
rm(m
eVÅ‐2)
0
20
40y
Mixed
E for
-60
-40
-20
0.000 0.125 0.250 0.375 0.500 0.625 0.750 0.875 1.000O Coverage (ML)
W. F. SchneiderACS ‐ 21 August 2007
Surface free energies
( , , ) OF T V N E TS= − (1) ( , , )O O OF T V E TS Nμ μ= − −
Canonical ensembleHelmholtz free energyConstant atom number
Grand canonical ensembleTransformed Helmholtz energyConstant oxygen potential
(1) (1)F F (1) (1)F F(1)(1) (1)
surf bulk(111) Pt
F F
Aγ
−=
Bulk Pt Pt(111)
(1)(1) (1)
Pt(111) F F
Aθ
θγ−
Δ =
Pt(111)‐θ(O)
( ) ( )
( , )
, ,
DFT DFTSurf Bulk
ZP ZPSurf Bulk
Vib VibSurf Bulk
E T V E E
E E
E T V E T V
Δ = −
+ −
+ −
( )
( ) ( )
Pt (111)
Pt (111)
Pt (111)
, ,
, ,
DFT DFT
ZP ZP
Vib Vib
E T V E E
E E
E T V E T V
θ
θ
θ
θΔ = −
+ −
+ −( ) ( )f
( ) ( ) ( ), , ,Vib VibSurf BulkS T V S T V S T VΔ = − ( ) ( ) ( )Pt(111), , , ,Vib VibS T V S T V S T VθθΔ = −
Supercell calculations on thick slabsPhonon calculations for bulk and surface
4‐layer supercellsO atom vibrations
W. F. Schneider ACS ‐ 21 August 2007
Phonon calculations for bulk and surfaceEinstein vibrational model
O atom vibrationsEinstein vibrational model
Pt(111)/O2 Phase Diagram
2
2 2 2O1 ( ) ln2 1
ODFT ZPO O O B
PE E G T k T
barμ
⎛ ⎞⎛ ⎞= + + Δ +⎜ ⎟⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠
O2 (T,P)
O coverage limited by accessible O2 T
O2 (g) + 2* 2O*12001200
O O O O
Pt(111) surface
g y 2and PXPS performed to determine O coverage at various O2 exposures
Ribeiro et al.
W ll k l θ = 0 25 800
900
1000
1100
800
1000
re (K
)
clean surface
p(4×4)–O(2 4) O
O atom
Pt atom
Well known low-pressure θ = 0.25 coverage used to calibrate XPS
Consistent with p(2×2)-O ordering
High-pressure θ = 0.5 coverage consistent with p(2×1)-O ordering 500
600
700
800
600
800
Tem
pera
tur
θ = 0.39θ = 0.40
θ = 0.52
p(2×2)–O
p(2×4)–Oc(4×4)–3O
Likely highest coverage accessible with O2 at ambient conditions
Kinetics limit rate of O2 loss from surface 200
300
400
-13 -11 -9 -7 -5 -3 -1 1
-1-3-5-7-9-11 1
400
-13200
θ = 0.25 θ = 0.49
θ = 0.48
c(4×4)–7O p(2×1)–Oc(2×4)–3O
1357911log10(PO2/1 bar)
113
W. F. SchneiderACS ‐ 21 August 2007
Coverage-dependent O binding energies
1.5
2
NO→ NO
¼ ML ½ ML
V/O)
OxygenformdE
BEdθ
∝
0.5
1
p(2×1)‐O
p(√3×√3)‐2ONO + O* → NO2½ O2 → O*
+ NO → + NO2
ΔE2 (θ)
Apparent window of
favorable ΔEng ene
rgy (eV
-1
-0.5
0
+ ½ O2 →
ΔE1 (θ)
favorable ΔE1and ΔE2
xygen bind
in
S f hibit t l t t di ti t bi di i
-1.50 0.2 0.4 0.6 0.8 1
ΔE1 (θ)
Oxygen coverage (ML)
Ox
Surface exhibits at least two distinct binding energy regimesLow coverage BE in middle of experimentally available values
O2 dissociation exothermic only up to a limiting coverage near θ = 2/3NO oxidation exothermic only at coverages above about θ = ½
W. F. SchneiderACS ‐ 21 August 2007
Based on experimental NO oxidation energy
O2 adsorption on Pt(111)-O(X)
• TBT and TFB isomers energetically
θO = 0.25θO = 0 θO = 0.5
equivalent
• Exothermically bound at low θO
• Adsorption energy diminishes greatly with coverage– Strong electrostatic
l i
ΔE = ‐0.80 eV ΔE = ‐0.23 eV ΔE = +0.57 eV
repulsion
• Endothermicallybound at catalytic coverage ΔE = ‐1.56 eV ΔE = ‐1.03 eV ΔE = ‐0.41 eVg– O2* unlikely precursor to
dissociation
• How does O2participate in the
ti ??
W. F. SchneiderACS ‐ 21 August 2007
reaction??
O2 dissociation on Pt(111)-O(X)28
“Nudged l b d”
28
a) O2 (g) + 2* → 2O*
elastic band” (NEB) calculations of reaction
h 2
3
2.14 eVpathways
Reaction barrier depends on
0
1
2
O2 (g)
ergy
(eV
)
-0.05 eV
surface coverage
-3
-2
-1
2O (g)transTS
Ene
p(2×1)-O Pt(111)
Fall Semester 2007W. F. Schneider CBE 547
p( ) ( )