Surfaces and potentials for metals

Molecular and Materials Modelling (MMM)


HCI D4

Daniele Passerone
Empa, Swiss Federal Laboratories for Materials Science and Technology

Ueberlandstrasse 129, 8600 Dübendorf
daniele.passerone@empa.ch
 Previous  lecture.
The  classical  Hamiltonian:  AMBER  force  7ield
Miller  indexes
 High-­‐symmetry  faces  in  a  cubic  system

They  have  usually  a  low  surface  free  energy:  predominant  in  the  Wulff  construction  

(dominate  the  envelope,  see  next  slides)
De7inition  of  lattice  plane
Example:  fcc  (100)

fcc=face  centered  cubic  

bcc=body  centered  cubic
 Example:  bcc  (111)

fcc:  Cu,  Ag,  Au,  Ni,  Pd,  Pt,  Rh,  Ir,  Pb,  Al  (maximum  packing)  
bcc:  Fe,  W
 Surface:  the  Gibbs  de7inition

•The  7irst  to  introduce  a  mathematical  surface  was  Sir  Josiah  Willard  
Gibbs  (1839-­‐1903)  inventor  of  statistical  mechanics  with  Maxwell  
and  Boltzmann.  

•To  this  mathematical  surface  is  attached  the  concept  of  “surface  
excess”  of  a  given  extensive  quantity
The  mathematical  surface  with  zero  thickness  is  de7ined  as  follows:  

“Let  us  consider  such  a  surface  of  discontinuity  in  a  5luid  mass  which  is  in  equilibrium  and  unin5luenced  by  gravity.  For  the  precise  measurement  of  the  quantities  with  which  we  
have  to  do,  it  will  be  convenient  to  be  able  to  refer  to  a  geometrical  surface,  which  shall  be  sensibly  coincident  with  the  physical  surface  of  discontinuity,  but  shall  have  a  precisely  
determined  position.  For  this  end,  let  us  take  some  point  in  or  very  near  to  the  physical  surface  of  discontinuity,  and  imagine  a  geometrical  surface  to  pass  through  this  point  and  all  
other  points  which  are  similarly  situated  with  respect  to  the  condition  of  the  adjacent  matter.  Let  this  geometrical  surface  be  called  the  dividing  surface,  and  designated  by  the  
symbol  S.  It  will  be  observed  that  the  position  of  this  surface  is  as  yet  to  a  certain  extent  arbitrary,  but  that  the  directions  of  its  normals  are  already  everywhere  determined,  since  
all  the  surfaces  which  can  be  formed  in  the  manner  described  are  evidently  parallel  to  one  another.  Let  us  also  imagine  a  closed  surface  cutting  the  surface  S  and  including  a  part  of  
the  homogeneous  mass  on  each  side.  We  will  so  far  limit  the  form  of  this  closed  surface  as  to  suppose  that  on  each  side  of  S,  as  far  as  there  is  any  want  of  perfect  homogeneity  in  the  
5luid  masses,  the  closed  surface  is  such  as  may  be  generated  by  a  moving  normal  to  S.”    

From  “On  the  equilibrium  of  heterogeneous  substances”,  1878

Surface  excess  is  the  difference  between  the  amount  of  a  component  
actually  present  in  the  system,  and  that  which  would  be  present  in  a  
reference  system  if  the  bulk  concentration  in  the  adjoining  phases  were  
maintained  up  to  a  chosen  geometrical  dividing  surface;  i.e.,  as  though  
the  interface  had  no  effect.
 From  thermodynamics  to  surface  energy

•In  equilibrium,  a  bulk  one-­‐component  system  is  characterized  completely  by  its  internal  energy,  U,  which  is  a  
unique  function  of  the  entropy,  volume  and  particle  number  (three  extensive  properties)  of  the  system:

U = U (S, V, N ),
@U @U @U
dU = dS + dV + dN,
@S V,N @V S,N @N S,V
dU = T dS P dV + µdN.

•The  fact  that  U  is  an  extensive  property U ( S, V, N ) = U (S, V, N ),

leads  to  the  Euler  equation
 U = TS P V + µN

and  the  Gibbs-­‐Duhem  relation:
SdT V dP + N dµ = 0.

•When  we  cleave  the  bulk  and  create  a  surface  of  area  A  from  an  in7inite  solid,  the  total  energy  
increases  by  an  amount  proportional  to  the  area  A.  The  constant  of  proportionality  γ  is  called  
“surface  tension”  (surface  speci>ic  free  energy):

U = TS P V + µN + A.
•In  case  we  stretch  a  SOLID  SURFACE  by  small  amounts,  

we  can  de7ine  the  surface  stress  as  the  matrix  of  the  derivatives  of  U  w.r.t.  the  strains:

X @U @
dU =
@U
dS +
@U
dV +
@U
dN + A d✏ij , ij = ij +
@S V,N,A @V S,N,A @N S,V,A @✏ij S,V,N
@✏ij T
i,j
X
dU = T dS P dV + µdN + A ij d✏ij
Surface  stress  vs.  surface  tension

i,j  (in  a  liquid  they  are  the  same)
Estimates  of  the  surface  free  energy
Units
 Modeling  surfaces:  slab  model

•A  slab  with  two  surfaces  is  modeled  

•Lateral  periodicity  
•Thick  vacuum  orthogonal  to  the  surface  
•Thick  slab  to  mimic  the  semi-­‐in7inite  solid  
•Beware  of  asymmetries  along  the  vertical  
direction  (dipole  moments  to  be  corrected  
in  the  long  range  limit)
 Computing  surface  energy  density  from  slab  calculations

•Compute  the  energy  of  a  bulk  system  with  N bulk  atoms  and  extract  the  bulk  energy  per  atom:

Ebulk Lx
✏bulk =
Nbulk Ly surface

•Compute  the  energy  of  a  slab  with  two  equivalent  surfaces  with  N surf_slab  atoms

Lz
•Subtract  the  bulk  energy  times  the  number  of  surface  slab  atoms  
 surface
from  the  surface  slab  energy,  and  divide  by  the  surface  areas:

Esurf slab Nsurf slab ✏bulk

=
2Lx Ly

•This  procedure  has  convergence  problems,  it  is  better  to  repeat  the  calculation  for  slabs  of  increasing

thickness  and  then  extrapolate  the  result
 Stepped  surfaces

•In  the  7igure  2.4,  a  vicinal  surface  is  depicted:  it  has  “almost”  
an  high  symmetry  orientation  (slightly  tilted)  

•With  a  simple  model,  the  “cusp”  in  the  Wulff  construction  

can  be  explained  (board)
 Creating  a  vicinal  surface  in  the  computer

(n11)
(100)

•computing  the  angle  of  rotation  

•rotating  the  slab  
•adapting  the  periodicity  
•cleaving  the  sample
 Example  with  vmd:  explaining  vicinals,  Cu(711)

1)create  slab

2)tilt  and  adapt  periodicity 3)cleave

STM  experiments:  gold  vicinals  and  nanotemplates

(Empa,  R.  Fasel)

 Surface  reconstructions

•Changes  in  the  positions  of  atoms  at  the  surface  such  that  the  periodicity  
differs  from  that  of  the  bulk  

•Closed-­‐packed  metals:  surface  atoms  try  to  recover  the  number  of  
neighbours  

•Semiconductors  with  covalent/directional  bonding:  reducing  the  half-­‐

occupied  dangling  bonds  formed  upon  cleavage  

•Ionic  surfaces  (insulators):  poor  surfaces  highly  unstable    

•Adsorbed  atoms  or  molecules  form  ordered  structures                                                  
 Example:  missing-­‐row  reconstruction  in  Au(110)

•The  (110)  surface  has  space  to  increase  the  level  of  packing.  The  most  packed  
faces  are  the  (111)  ones
Au  surface  reconstructions
 The  shape  of  a  crystal:  Wulff  construction

The  internal  envelope  of  the  planes  

perpendicular  to  the  polar  plot  of  γ as  a  
function  of  orientation  is  geometrically  
similar  to  the  equilibrium  shape  of  the  
crystal  (Wulff  1901)  
Proof:  next  slides

crystal shape

•An  advanced  example  for  the  construction  of  a  Wulff  solid  is  found  here:  http://
www.ctcms.nist.gov/wulffman/features.html  

•A  simpler  program,  but  more  versatile  for  surface  islands,  will  be  used  today  in  
the  exercise  session  (SOWOS,  Daniele  Scopece)
 Lx
The  “cusp”  in  Wulff’s  plot:  energy  
Ly dependence  on  the  surface  tilting

n  steps

𝛽
𝛾0
a
𝜗
Lx/n
Lx

F = Lx L y = 0 Lx /n cos ✓Ly n + Ly n F:  free  energy

Lx
= 0 cos ✓ +
n
= cos ✓ + | sin ✓| 𝛽:  step  cost  per  unit  
a= | sin ✓| 0
n Lx a length
Interaction  between  steps  (entropy+elastic):  additional  term  involving
 g(T ) n
a  L-­‐2  repulsion  and  density  of  steps. Einterac =
(Lx /n)2 Lx
g(T ) 3
' 0+ |✓| + 3
|✓|
a a
Proof  of  Wulff’s  theorem
End  of  the  proof
 Roughening  transition,  crystal  shape

•The  cusp  blurs  with  temperature  (T melt=327.5  °C)

Phenomenology:  described  on  the  

blackboard.

•NaCl  and  Gold  clusters

Potentials for metals
 Outline

•Failures  of  the  two-­‐body  potentials  for  metals  

•The  embedded  atom  method  (EAM)  
•The  glue  potential  for  gold  
•An  example  of  7itting:  AlPdMn
 Elastic  constants  

(E.  Y.  Tsymbal,  Nebraska)

•In  a  1D-­‐solid,  we  can  determine  elastic  constant  k  from  the  curvature  of  the  
potential  at  the  minimum.  

•In  a  crystal…  The  elastic  properties  are  described  by  considering  a  crystal  as  a  
homogeneous  continuum  medium  rather  than  a  periodic  array  of  atoms.  In  a  
general  case  the  problem  is  formulated  as  follows:
Bulk modulus: resistance to uniform compression
 Two  body  potentials  failures

•In  a  two-­‐body  system  C =C holds.  In  a  metal  the  ratio  can  reach  3.7!  
12 44  

•Vacancy  formation  energy  in  a  two-­‐body  system  ~  equal  to  the  

cohesive  energy  (breaking  pairwise  bonds).  In  a  metal,  very  small!  

•Melting  temperature  T m~0.1  Ec/kB    for  a  two-­‐body  system.  In  gold,  

Tm=0.03  Ec/kB      

•In  a  two-­‐body  system,  E /T

c m~10  :  high  evaporation  rate  at  the  surface  
near  the  melting  point.
 Many-­‐body  effects  in  metals

•We  de7initely  need  many  body  terms!

Embedded  atom  method

•Friedel  suggested  in  1952  that  the  dissolution  of  H  in  Cu  was  related  by  the  
dissolution  of  hydrogen  in  jellium  (impurity  induced  oscillations)  

•Impurity-­‐in-­‐jellium:  embedding  energy    as  function  of  background  gas  density  -­‐
>  Noerskov  and  Lang  (Effective  medium):  heat  of  solution  of  light  impurity  in  a  
metal  (substituted  by  jellium)    

•EAM  basis:  cohesive  energy  of  a  metallic  system  can  be  expressed  in  terms  of  
embedding  energies  +  electrostatic  interaction  

•Each  atom  in  the  metal  is  embedded  into  the  electron  gas  created  by  the  other  
atoms  

•Near  a  defect  -­‐>  atoms  are  embedded  into  different  density  pro7ile  (surface,  
vacancy...)  

•Ansatz  for  a  functional  form  and  7itting  the  parameters  to  bulk  metals
 EAM:  some  details

•The  deviation  from  pairwise  bonding  is  accounted  for  by  a  term  G(embedding  
energy)  dependent  on  the  local  electronic  density;  it  is  the  interaction  of  an  
atom  with  the  background  electron  gas  created  by  the  neighboring  atoms.  

•In  addition,  we  have  U,  an  electrostatic,  two-­‐body  interaction  

•Effective  pair  interactions  (bonding)  are  affected  by  the  density

 Successes

•Bulk  phonons  of  Cu  

•Structure  factor  for  Ag  

•Gibbs  free  energies

Point  defects  and  grain  boundaries
Alloys
 Another  approach:  the  glue  potential  for  gold
(Ercolessi,  Parrinello  and  Tosatti  1988)

•Electronic  cohesion  in  noble  and  near  noble  metals  is  largely  caused  by  d  
electrons  -­‐>  broad  7illed  d  bands  

•Hybridization  and  directional  bonding  are  hindered  

•The  number  of  neighbors  of  a  given  ion  corresponds  to  its  local  electronic  
density:  coordination  -­‐>  local  density  

•The  difference  between  the  vacancy  formation  energy  and  the  cohesive  energy  
is  an  expression  of  the  nonlinearity  of  the  energy  as  a  function  of  the  
coordination  

•Example:  vacancy  formation  leads  the  coordination  from  12  to  11  in  an  fcc  bulk  
•But  extracting  an  atom  from  the  system  leads  its  coordination  from  12  to  zero
 Mathematically
•Pair  potential  +  “glue”  

•Choose  units  so  that  n =  12  in  the  fcc  bulk  

0  

•Compute  the  forces  

•The  central  point  is  that  all  three  functions  are  left  free  initially:  they  are  available  to  the  >it  (EAM:  
n(r)  is  the  sum  of  atomic  contributions)  

•The  physical  meanings  of  U(n)  and  Φ(r)  are  interchangeable:

 Fitting  to  gold

•A  set  of  properties  may  be  chosen  for  7itting:  this  allows  many  different  triplets  
of  ρ,U  and  Φ all  of  which  7it  the  same  quantities,  yet  describe  different  systems  

•Thermal  test  at  high  temperatures  

•Reject  Hamiltonians  leading  to  bad  structures  at  T=0  
•Surface  properties:  coordination  is  lower,  a  different  part  of  U(n)  becomes  
relevant    

•First  neighbor  interactions;  lattice  parameter;  cohesive  energy;  bulk  modulus;  

surface  energy;  tranverse  phonon  frequency  at  X;  vacancy  formation  (~),  thermal  
expansion  (~),  melting  T  (~),  Au(100)  unstable
Analytical  forms
Physical  properties
 Success

DNA  bases  on  Au(111)  (Zerbetto)                                                                          Reconstructions  (Passerone)  

Lead  Clusters  (Doye)

 A  modern  example  for  EAM:  AlPdMn

•AlPdMn  forms  a  high  number  of  CMAs  

•Ξ  phases:  approximants  of  a  decagonal  quasicrystal  (L=1.6  nm)  
•They  show  “metadislocations”  (involving  more  than  10000  atoms)    
•ab  initio  impossible  
•Classical  MD  challenging:  potential?
 EAM  potential  and  analytical  forms

embedding  function  based  on  equation  of  state

•Oscillations  in  the  density  or  in  the  pair  potential:  two  different  potentials,  
about  40-­‐70  parameters

non-­‐oscillatory  Morse  potential

empirical  pair  potential  with  repulsion  +  damped  oscillations

non-­‐oscillatory  density

oscillatory  density
Model  III  (all  oscillatory)
 Fitting  procedure

•POTFIT  package  (Brommer  and  Gähler)  

•ab  initio  reference  database  with    
•relaxed  structures  
•ab  initio  high  T  MD  snapshots  
•strained  samples  
•Minimizing  the  objective  function  (simulated  annealing)  

•Energy  weight  >>  other  weights

 Results

•Transferability:  testing  the  potential  against  ab  initio  data  not  included  in  the  
set
 Tests  on  icosahedral  phases

•Energies  reproduced  with  high  accuracy  

•Comparisons  are  done  both  at  T=0  and  at  T=1200  (classical  and  ab  initio  MD)  
•Compare  the  energies  quenched  from  MD  snapshot  with  the  underlying  
“convex  hull”  with  vertices  in  AlPdMn,  Al12Mn,  Al21Pd8,  and  Al3Pd2  structures.    

•The  number  of  Al  atoms  in  the  7irst  shell  is  counted  as  an  important  structural  
parameter:  Model  III  wins!
Additional  slides:  high  symmetry  slabs
 fcc  (100)

dz

d=a/√2
Lz

a
 dz

a
Lz

d
fcc  (110)
 fcc  (111)

Ly

dz

Lx
Lz