Interaction

 with  matter  and  electronic  

spectroscopies  (STM  and  Infrared)

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
 Scanning  tunnelling  microscope  (STM)

•An  instrument  for  imaging  surfaces  invented  in  1981  by  Gerd  Binnig  and  
Heinrich  Rohrer  (at  IBM  Zürich):  Nobel  prize  in  physics  1986  

•more  than  30  years  later,  the  most  advanced  nanotechnology  center  @  IBM  
Rüschlikon  was  dedicated  to  their  name
 Tunnelling  effect  and  STM

•A  very  simple  illustration  of  the  tunnelling  effect.

 Principles  of  operation

Basic components of STM:

•Piezoelectric transducers
•Vb applied and I measured
The scanner

•distance from the surface few Å can be

mounted with
Five
the tip or the

• Constant current mode by using feedback control sample stage. 1.

2.
• Stable and sharp metal tip (single atom) 3.

4.

5.

Images: L. Zang, Univ. Utah

M. Iannuzzi, UZH

• Tunneling current from tip to sample or vice-versa depending on bias;

• Current is exponentially dependent on distance;
 STM  tips

•STM  tip  should  be  conductiong  (metals,  like  Pt)  

•STM  plays  with  the  very  top  atom  at  the  tip  and  the  surface  of  the  sample  
•Tips  with  single  atoms  on  the  top  can  be  obtained  by  cutting  a  thin  metal  wire  using  a  wire  cutter  
•An  alternative  is  chemical  etching  (Millunchick  research  group)
 Tunnelling  current

Metal  electrons  Sill  levels   Specimen  and  tip  are   A  voltage  is  applied  and  
until  Fermi  energy,  and   brought  close  separated   electrons  can  tunnel  from  
the  minimum  energy  to   by  vacuum the  specimen  to  the  tip  or  
extract  an  electron  is  
viceversa
called  work  function
 Tunnelling  and  available  states

•Electrons  that  may  tunnel  through  the  

barrier  would  have  no  empty  states  
available  on  the  other  side

•When  a  voltage  is  applied,  we  effectively  

“raise”  the  Fermi  energy  of  the  sample.  
States  are  available  for  tunnelling  into.

Images: L. Zang, Univ. Utah

 Tunnelling  current

•The  tunnelling  current  decays  exponentially  with  the  distance  from  the  sample  
(note:  𝚽  is  the  work  function  of  the  sample)  

•Since  the  current  decays  typically  by  one  order  of  magnitude  every  angstrom,  this  
makes  STM  very  sensitive  to  surface  corrugation  at  the  atomic  level  

Maintaining  a  constant  tunneling  current  by  adjusting  the  height  with  a  piezo-­‐
electric  crystal,  and  monitoring  the  piezo  voltage  while  scanning,  allows  one  to  
image  a  surface,  under  ideal  conditions,  to  atomic  resolution.    
 How  to  compute  the  tunnelling  current

First-­‐order  perturbation  theory  for  the  transition  rate,  counting  all  possible  initial  
and  Sinal  states
 Corrugation  and  electron  density

One  of  the  factors  affecting  resolution  is  corrugation,  i.e.  how  much  the  electron  
density  of  surface  atoms  varies  in  height  above  the  surface.    
Graphite  has  a  large  corrugation,  and  is  very  planar,  and  thus  is  one  of  the  easiest  
materials  to  image  with  atomic  resolution.  (see  next  slide  for  example)    
STM  does  NOT  probe  the  nuclear  position  directly,  but  rather  it  is  a  probe  of  the  
electron  density,  so  STM  images  do  not  always  show  the  position  of  the  atoms.  
STM  imaging  depends  on  the  nature  of  the  surface  and  the  magnitude  and  sign  of  
the  tunneling  current.  For  example,  if  you  have  Cu  and  Si  on  the  same  surface,  under  
the  same  condition,  the  current  with  Cu  is  much  higher  .    
Since  STM  images  the  outermost  atom  on  sample  surface,  ultra  high  vacuum  (UHV)  
is  normally  required  to  assure  no  surface  contamination  (e.g.,  coverage  of  air  
molecules  or  water)  so  as  to  image  single  atoms  or  at  atomic  resolution)
 Examples of STM images
•Allows to obtain images of surfaces up to atomistic resolution.
•Huge importance in the development of nanoscience over the past 30 years
(Nobel prize award 1986).
Ruffieux et al., Empa laboratories (2010)

Complex metal surface patterns:

Au(111) herringbone reconstruction Molecular networks on
surfaces: C60 over Ag/
Pt(111)

Atomically resolved STM

images of individual
carbon nanotubes 3 nm

J.V. Barth et al., Phys Rev B 42, 9307 (1990)

L.C. Venema et al., Appl. Phys. A 66, S153–S155 (1998)
 Graphite  and  honeycomb  networks

Polyphenylene  honeycomb  networks  on  different  metal  surfaces  

(Bieri  et  al,  Empa,  2010)
“Watching”  molecular  orbitals!

A layer of h-BN is deposited on Rh(111), forming a periodic nano mesh with wires and pores.
Molecules show preferential adsorption and orientation.
Tersoff-­‐Hamann  and  density  of  states
 Limitations  of  simple  theory

•Undistorted  electronic  states  

•No  interaction  between  tip  and  sample  
•Often  the  tip  is  closer  than  few  Angstrom  
•tip  is  a  s-­‐like  wavefunction,  though  TM  are  used  
•No  control  on  chemical  nature  of  outermost  atoms
 Corrugated samples
•In the case of corrugated samples more tip atoms can be equivalently distant from
the sample.

Corrugated
Flat surface: sample:
Single s-orbital Single s-orbital
is ok description might
be not sufficient

•Going beyond a single s-orbital description poses some problems :

The actual geometry of the tip used for scanning is usually not known.
Models explicitly including tip geometries should not be too computationally
demanding for applications to large systems.
 Modeling finite size tips
•Using a two atoms tip the tunneling current is proportional to the sum of the sample LDOS
at the position of the two atoms R1,R2 *:

•The same reasoning holds for an N-atoms tip:

W

•By choosing an effective tip volume W we average over the

compatible atomic configurations C :

Constant current images reflect contour levels

of the convolution between the sample LDOS
and the shape function s(r)

*Mizes H.A. et al., Phys. Rev. B 36, 4491 (1987)

 Application to Au(677)*
•Stepped Au(111) surfaces display interaction between
steps and surface reconstruction. Step lines alternatively face
regions with hcp or fcc stacking. At low coverage molecular
adsorption takes place at fcc regions. Why?
•Large scale DFT calculations on stepped Au(111) surfaces.

Sample LDOS are calculated from Kohn-Sham states

and convoluted ( hemispherical tip R = 7 Angstroms).

•Experimental STM line profiles across the two regions are

different.
•DFT calculations suggest a huge structural difference between
hcp and fcc step lines, which explains the substrate selectivity.
•Our model suggests that this difference can only be probed with
sharp tips. For large tips convolution effects, rather than the
sample LDOS, determine the image.

*R.Gaspari, C.A. Pignedoli, R. Fasel, M. Treier, D.Passerone, Phys. Rev. B 82, 041408(R) (2010)
 Application to HBC on Cu(111)
•Large scale DFT calculations on stepped HBC/Cu(111)
surfaces

Sample LDOS are calculated from Kohn-Sham states

and convoluted ( hemispherical tip R = 7 Angstroms)

•Experimental STM line profiles across the two regions are

different and show that the apparent height of the molecule is
smaller than the physical one.
•Tersoff-Hamann profiles display discrepancies with respect to
the experimental lines.
•Convoluted LDOS isosurfaces display correct height and width.

The height of the molecule depends on the size of the tip

 Finite  ribbons  and  “termini”  of  GNR
full  credits  to:  Leopold  Talirz  &  Carlo  Pignedoli

21
22
23
24
25
26
27
28
29
30
31
Infrared  spectroscopy
 Spectroscopies

•Methods  to  study  the  properties  of  matter  (atoms,  molecules  and  solids)  by  
investigating  its  interactions  with  particles  (photos,  electrons,  neutrons,  ions).  

•We  consider  here  today  and  next  time  in  particular  infrared  and  ultraviolet  
(UV)  spectroscopy:  interaction  of  photons  with  matter.
Infrared  spectrum
Examples:  IR

(diatomic molecule)
 Atomic  structure  vs.  molecular  vibrations

•The  presence  of  certain  functional  groups  can  easily  be  tested  through  
vibrational  spectroscopy  

•Vibrations  in  a  molecule  are  not  directly  visible  

•Different  techniques  to  investigate  them  
•Two  primary  examples:  IR  (infrared  radiation)  and  RAMAN  (inelastic  
scattering  of  usually  visible  light)
 Theoretical  determination  of  vibrational  spectrum

•Calculation  of  vibrational  normal  modes  in  harmonic  approximation  

(diagonalization  of  the  dynamical  matrix)  

•It  neglets  anharmonicities  and  needs  correction  

•One  selection  rule  that  inSluences  the  intensity  of  infrared  absorptions,  is  that  a  
change  in  dipole  moment  should  occur  for  a  vibration  to  absorb  infrared  energy  

•Intensities  for  IR  (and  Raman)  are  obtained  from  the  derivatives  of  the  dipole  
moment  and  polarizability,  along  the  normal  mode  vectors  in  such  static  
calculation    

•Problem  in  condensed  phase:  Static  calculations  need  to  be  started  from  an  
optimized  minimum  of  the  PES:  molecules  in  the  gas    phase  

•ALTERNATIVE:  Averages  from  (ab  initio)  molecular  dynamics  

 Vibrational  analysis  in  standard  quantum  chemistry

•Example  of  normal  modes:  carbon  dioxide

•In  general  linear  molecules  have  3N-­‐5  normal  modes  (three  for  the  center  of  mass  
translation,  two  rotational  angles);  non-­‐linear  molecules:  3N-­‐6  normal  modes

•Harmonic  oscillator:  we  need  to  obtain  the  unknown  force  constant  from  the  
potential.  Hooke’s  law:

dV 1 2
F = = kx ! V = kx
dx 2
•Newton’s  law:
d2 x
m 2 = kx ! x(t) = A sin(2⇡⌫t)
dt

•Double  derivation  with  respect  to  t  gives:

4⇡ 2 ⌫ 2 mx = kx
 Normal  mode  analysis

•Specify  coordinates  of  each  atom  and  the  differences  to  the  equilibrium  positions:

•Use  classical  or  ab  initio  is  used  to  compute  V  as  a  function  of  the  coordinates…  and  
thus  second  derivatives  as  a  function  of  the  displacement

V (x1 , y1 , z1 , x2 , y2 , z2 , . . . , xN , yN , zN )

•We  sample  the  change  in  the  forces  as  we  move  the  different  atoms  in  different  
directions:
 Hessian  diagonalization

•The  complete  list  of  the  force  constants  is  called  the  Hessian:

•Let  us  consider  a  linear  molecule  and  only  vibrations  along  the  axis  (symmetric  
and  antisymmetric  stretching  modes):
 Hessian  diagonalization

•Setting  mass-­‐weighted  coordinates

•Matrix  form:

•The  eigenvalues  are  the  negative  of  the  squared  normal  mode  frequencies  
•The  eigenvectors  are  the  mass-­‐weighted  coordinated  displacements
 In  quantum  mechanics…
•Vibrations  are  quantised  along  each  mode.  The  fundamental  frequency  is  the  one  
obtained  by  the  classical  analysis…  but  the  energy  levels  and  eigenvectors  for  
each  mode  are  given  by  the  Schrödinger  equation  (Hn  are  Hermite  polynomials)  
 Note  on  energy

•If  a  certain  normal  mode  has  a  certain  energy,  then  E~k  A^2  with  A  maximum  
amplitude  of  an  oscillation.    

•If  all  modes  have  the  same  energy  (energy  equipartition)  it  follows  that  the  
amplitude  is  inversely  proportional  to  the  eigenfrequency  and  inversely  
proportional  to  the  square  root  of  the  reduced  mass  m.  

•So  in  order  to  obtain  a  spectrum  with  the  same  intensity  in  the  each  mode,  we  
need  to  rescale  the  spectrum  accordingly.
 IR  spectroscopy
•In  an  IR  spectroscopy  experiment  we  introduce  light,  an  electromagnetic  wave,  
which  interacts  with  a  molecule  by  disturbing  the  local  electric  Sield  EF.  The  energy  
of  the  photon  must  match  the  difference  in  energy  between  two  vibrational  states.  

•The  strength  of  the  absorption  depends  on  the  initial  and  Sinal  state,  and  on  the  
dipole  moment  of  the  molecule

•The  dipole  moment  is  modiSied  during  the  vibration:  Taylor  expanding  around  
equilibrium…

•For  small  vibrations,  we  keep  the  Sirst  order  and  we  note  that  the  transition  
between  states  will  depend  on  the  derivative  of  the  dipole  moment  (the  Sirst  term  
is  zero  for  state  orthonormality)
 Autocorrelation  functions

•All  vibrational  spectra  calculated  from  molecular  dynamics  simulation  are  based  
on  the  Fourier  transform  of  certain  autocorrelation  functions:  

•power  spectra  -­‐-­‐>  particle  velocities  

•IR  spectra  -­‐-­‐>  molecular  dipole  moments  
•Raman  spectra  -­‐-­‐>  molecular  polarizabilities  
•Wiener-­‐Khintchine  theorem:

connecting  the  autocorrelation  in  time  domain  with  the  Fourier  transform.
 Intensity  of  the  peaks  of  the  infrared  spectrum

•In  static  calculations,  the  intensities  of  a  mode  are  given  by  the  square  of  the  
derivative  of  the  dipole  moment  along  the  normal  coordinate  of  that  mode

•In  molecular  dynamics,  we  should  keep  in  mind  that  in  harmonic  approximation  if  
the  run  is  well  equilibrated,  each  oscillator  contains  the  same  amount  of  energy  
(equipartition)  

•The  amplitude  of  a  mode  is  inversely  proportional  to  the  eigenvalue  frequency  ω  
and  to  the  square  root  of  the  mass,  so  to  obtain  the  same  intensity  for  each  mode  
one  should  rescale  the  spectrum  by  ω2,  which  is  formally  equivalent  (and  
numerically  more  convenient)  to  correlating  time  derivatives  (Fourier  transform  
properties).

•In  the  particular  case  of  the  infrared,  we  autocorrelate  the  derivative  of  the  dipole  
moment
Examples:  methanol  and  acetone
 For  faster  MD  (less  accurate):  
Density  functional  tight  binding:  DFTB
•DFTB is based on Density Functional Theory (DFT) but uses empirical approximations to improve the performance
whilst maintaining accuracy.

•The main approximation in the tight- binding approach is the replacement of the exact many-body Hamiltonian in DFT
with a parameterized Hamiltonian matrix. In the DFTB description pseudo atomic wave functions, written in terms of
Slater-type orbitals and spherical harmonics, are used to model the electron density.

•The orbital basis is fitted against DFT theory and then used to calculate the Hamiltonian and overlap matrix elements. The
DFTB approach uses approximations to expand the total energy of the DFT representation to the second order in charge
density and spin density fluctuations.

•The  matrix  elements  do  not  fully  describe  the  total  energy  of  the  system,  the  remaining  part  of  the  total  energy  is  
included  as  a  short-­‐range  repulsive  term  which  can  be  described  in  terms  of  pair  wise  potentials  between  atoms.  
The  pair  wise  potentials  are  acquired  through  a  Sitting  process.  

•For  systems  where  the  balance  of  charges  between  pairs  of  atoms  is  very  small,  DFTB+  also  uses  self  consistent  
charge  (SCC-­‐DFTB)

to  improve  the  description  of  the  bond.  In  order  to  simplify  the  calculations,  the  SCC  calculation  is  performed  on  
Mulliken  charges  rather  than  the  potential  and  charge  density