Ab-initio study of work functions of

element metal surface

Xiang Ma
Materals Process Design and Control laboratory

May 7th, 2007

Instructor: Professor Zabaras

MAE 715 – Atomistic Modeling of Materials 1

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Outline

 Definition of Work function

 Slab model and super cell
Computation Methods (Density functional theory)
 Change of work function due to the orientation of
clean surface
 Change of work function due to absorption of H
atom
 Conclusion

MAE 715 – Atomistic Modeling of Materials 2

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 From Solid to Surface

A image of surface for

fcc(111)

 In this course, most of the problems we deal with are bulk properties.

 In nature, crystals are not infinite but finite macroscopic three-dimensional

objects terminated by surfaces.

 Many phenomena and processes occur at the interface between a condensed

phase and the environment.

 Modeling surfaces is then of great theoretical and practical interest.

MAE 715 – Atomistic Modeling of Materials 3

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 From Solid to Surface

Atomic resolution on Pt(100)

 The key-ingredient to surface science experiments is ultra-high vacuum (UHV).

 To main a low pressure to assure that a surface stays clean for a time long
enough to do some experiments.

 With the development of density functional theory, we can also explore the
surface properties through the ab-initio study.

 A very good surface science tutorial.

MAE 715 – Atomistic Modeling of Materials 4

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 From Solid to Surface

 A lot of phenomenon associated with surface can be studied by first-principal

calculation

--- surface reconstruction and surface relaxation

--- surface energy
--- adsorption on surfaces
--- interface
--- work function
 With the adsorption of atoms or molecules, the surface electronic structure is
modified and the work function can change by several eV.

 The measurement of the work function changes can give valuable insight in to
the condition of a given surface.

 Nowadays, the work function can be calculated by ab-initio methods in the

framework of density functional theory.

MAE 715 – Atomistic Modeling of Materials 5

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Work function definition
• Work function W is defined as the minimum energy necessary to extract an
electron from the metal at 0K.
• For a crystal with N electrons, if EN is the initial energy of the metal and EN 1
that of the metal with one electron removed to a region of electrostatic potential
Ve , we define

W  ( EN 1  Ve )  EN

Note: The removed electron is assumed to be at rest, and therefore possesses

only potential energy.

MAE 715 – Atomistic Modeling of Materials 6

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Work function definition

At 0K, the chemical potential  is by definition

  EN  EN 1
In the limit of large systems, all polarisation effect
can be neglected after removing the electron. Then
chemical potential is then shown to coincide with
the Fermi energy EF

The work function, finally, is the difference

between the Fermi level and the vacuum level.

Schematic energy diagram

of a metal W  Ve  EF

MAE 715 – Atomistic Modeling of Materials 7

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Work function definition

W  Ve  EF
 The calculation of work function is then divided into two parts.

 First to perform a self-consistent calculation to find the Fermi energy of the slab.

 Second, we need to find the electrostatic potential in the vacuum level.

Macroscopic average

MAE 715 – Atomistic Modeling of Materials 8

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Macroscopic average

 The electronic density n(r ) is the basic variable calculated by DFT.

 Introduce the plane-averaged electronic density:

1
n ( z) 
S S
n(r )dxdy
where z axis is perpendicular to the slab surface S

 The macroscopic-average electronic density n ( z ) is then defined from the

integration over the interplanar distance d of the slab:

1 d /2
n ( z )   n ( z  z )dz 
d d / 2
 The potential is related to the charge density via the Poisson equation. So we
can get a similar relation between plane-averaged potential Ve ( z ) and macroscopic
average Ve ( z )
1 d /2
Ve ( z )   Ve ( z  z )dz 
d d / 2

MAE 715 – Atomistic Modeling of Materials 9

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Macroscopic average

 By plotting the macroscopic average over the z axis, the vacuum level is found.
 Because the curve of the average is nearly flat in the vacuum provided the
vacuum is large enough.
 Subtracting this vacuum level from the Fermi level get the work function for the
metal surface.
Work Function
6

2 Ef
Energy(eV)

-2

-4

-6

-8

-10
0 10 20 30 40 50
Lattice Units (A)

MAE 715 – Atomistic Modeling of Materials 10

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Slab model and supercell approach
Slab model is the most popular way to model
the surface.
The slab model consists of a film formed by a
few atomic layers parallel to the crystalline plane
of interest.
Using plane waves needs to force a 3-D
periodicity. The thin slabs needs to repeat in one
direction.
vacuum
layer To perform a supercell calculation, one
defines a unit cell oriented with one axis
perpendicular to the surface of interest, containing
the inequivalent atoms of a crystalline thin film and
thin super some vacuum layers.
slabs cell
Ideally, the thickness of the vacuum layer and
of the slab must be large enough for two
successive metal surfaces not to interact
significantly.

MAE 715 – Atomistic Modeling of Materials 11

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Slab model and supercell approach

o It is not trivial to construct the slab model at first. You need to visualize them.
o A nice web tool Surface Explorer is used for this purpose.

fcc(100) fcc(110)

fcc(111)

MAE 715 – Atomistic Modeling of Materials 12

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Slab model and supercell approach
 XCrysDen is an application for visualizing crystalline and molecular structures.
 All of the slab models studied were viewed using XCrysDen to ensure that their
geometries were described correctly.
 Surface primitive cell is two-dimensional, which is different from conventional
bulk primitive cell.

MAE 715 – Atomistic Modeling of Materials 13

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 DFT calculations
1. As a preliminary step towards the study of surface, we have to find the
equilibrium lattice constant;

2. It is well known that the equilibrium atomic positions in a crystal surface are
generally different from those in the ideal bulk-terminated surface. We need to
perform a relaxation calculation to find the equilibrium geometry of the surface.

3. The relaxed coordinates are put into another input file to perform a self-
consistent calculation to find the Fermi energy in the slab

4. Using post-process code to extract the electrostatic potential from the output
file.

5. Calculate the macroscopic average potential to determine the vacuum level

6. Put the two values into the definition of the work function to determine the final
solution.

MAE 715 – Atomistic Modeling of Materials 14

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 DFT calculations
• Basis sets --- plane wave
 Cut-off energy of 16 Ry for the plane wave expansion
 ultrasoft pesudopotentials
 The Fermi level is positioned using the Methfessel-Paxton (MP) scheme, with
the smearing parameter set to 0.01 Ry.
 8x8x1 special Monkhorst-Pack special k-points
• slab models
 A surface unit cell with a slab of 8 atom layers and 8 equivalent vacuum
layers was chosen to model the metal surface. H atom coverage is a full
monolayer.
• Exchange-Correlation approximation
 LDA(Perdew-Zunger form)

• Software: Quantum Espresso (opEn-Source Package for Research in

Electronic Structure, Simulation, and Optimization), version 3.2

MAE 715 – Atomistic Modeling of Materials 15

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Summarize of the computation procedure

Fit E Vs V curve to find the theoretical self-consistent calculation to find the

lattice constant (pw.x) Fermi energy (pw.x)

Set up the appropriate thickness of Extract the electrostatic potential form

slabs and vacuums the self-consistent calculation (pp.x)

Calculate the macroscopic average

relax the geometry of the slab (pw.x)
(average.x)

MAE 715 – Atomistic Modeling of Materials 16

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Numerical examples

1. Change of work function depends on the surface

orientation

1. Change of work function due to the H atom adsorption

MAE 715 – Atomistic Modeling of Materials 17

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Lattice constant

Energy Vs. Volume

-4.174

-4.176 Ecut = 20
Ecut = 40
-4.178
Ecut = 60
Energy (Ryd) -4.18

-4.182

-4.184

-4.186
-4.188

-4.19

-4.192
7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 8
Lattice constant (a0)

Theoretic: 7.50 bohr LDA underestimate

Experimental : 7.66 bohr

MAE 715 – Atomistic Modeling of Materials 18

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al and Al(100)

Al unit
• fcc structure cell
• a = 7.50 a.u

a  b  c,       900

(11)
3rd layer

1st layer
Al(100) side view Al(100) top view

MAE 715 – Atomistic Modeling of Materials 19

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al(100)
0.035

Work Function
6
0.03
4

0.025 Ef
2
Electronic Density

Energy(eV)
0.02 0

-2
0.015

-4
0.01
-6

0.005
-8

0 -10
0 10 20 30 40 50 0 10 20 30 40 50
Lattice Units Lattice Units (A)

Plane-averaged electronic charge density (dashed line) Plane-averaged electrostatic potential (dashed line)
Macroscopic average (solid line) Macroscopic average (solid line)

MAE 715 – Atomistic Modeling of Materials 20

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al(110)

(2 1)
3rd layer

1st layer

Al(110) side view Al(110) top view

unit
cell

a  b  c,       900

MAE 715 – Atomistic Modeling of Materials 21

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al(110)

function
0.03

Work
4

Ef
Electronic Density

Energy(eV)
0.02

-2
0.01

-4

-6
0 0 10 20 30
0 10 20 30
Lattice Unit (A) Lattice Unit (A)

Plane-averaged electronic charge density (dashed line) Plane-averaged electrostatic potential (dashed line)
Macroscopic average (solid line) Macroscopic average (solid line)

MAE 715 – Atomistic Modeling of Materials 22

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al(111)

(11)
C
B
A Al(111) top view
Al(111) side view

unit
cell

a  b  c,     90o ,   120o
MAE 715 – Atomistic Modeling of Materials 23
CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Al(111)
0.03

Work Function
6

0.025
3
Ef
0.02
Electronic Density

Energy (eV)
0.015
-3

0.01 -6

0.005 -9

-12
0
0 10 20 30 40 50 0 10 20 30 40 50
Lattice Unit (A) Lattice Unit (A)

Plane-averaged electronic charge density (dashed line) Plane-averaged electrostatic potential (dashed line)
Macroscopic average (solid line) Macroscopic average (solid line)

MAE 715 – Atomistic Modeling of Materials 24

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Result

Calculations of Work function of Al

Al Fermi Level Vacuum (eV) Work Experimental

(eV) Function (eV) (eV)
(100) 2.364 6.782 4.418 4.41  0.02
(110) 2.488 6.768 4.28 4.28  0.02
(111) 2.634 6.869 4.235 4.24  0.03

The results are in a good agreement with the experimental values.

MAE 715 – Atomistic Modeling of Materials 25

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Result

Calculations of Work function of Copper

Cu Fermi Level Vacuum (eV) Work Experimental

(eV) Function (eV) (eV)
(100) 5.551 10.391 4.84 4.59  0.03
(110) 2.390 7.105 4.715 4.48  0.03
(111) 5.581 10.780 5.199 4.94  0.03

The results shows a little deviation from the experimental values. It may be due
to the experiment is performed at room temperature, while the calculation is at
0K. Overall, it shows good accuracy using this method since the error is within
the computational range.

MAE 715 – Atomistic Modeling of Materials 26

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Anisotropy of the work function
 From the Cu, we see that it shows the trend (110), (100),(111) of increasing work
function.
 This is best explained by the Smoluchowski[1] smoothing.

 This smoothing leads to a dipole moment which opposes the dipole created by
the spreading of electron and thus reducing the work function
 Surface orientations of high density experience small smoothing, inducing a
small reverse dipole, and thus a high work function.

[1] R Smoluchowski, Phy. Rev. 60, 1941

MAE 715 – Atomistic Modeling of Materials 27

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Anomaly anisotropy of Al work function

 However, from the calculation, it is seen that the Al doesn’t obey this increasing
ordering.
 In the paper [1], the author investigated this phenomenon and concluded that
the trend of the work function Al can be explained by a charge transfer the atomic-
like p orbitals of the surface ions perpendicular to the surface plane to those
parallel to the surface, when compared to the bulk charge density.
 Thus it results from a dominant p-atomic-like character of the density of states
near the Fermi energy.
 Overall, our methods recovered both the normal and abnormal anisotropy of
then work function of the fcc metals.

[1] C.J.Fall, N.Binggeli and A. Baldereschi, Phy. Rev. B, 58,1998

MAE 715 – Atomistic Modeling of Materials 28

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Adsorption of H on the Al(111) surface

 There are four inequivalent adsorption sites on an fcc (111) surface.

 We consider a monolayer of H atom adsorpted on one Al (111) surface.

MAE 715 – Atomistic Modeling of Materials 29

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 H on the Al(111) surface (top view)

ontop bridge

hcp hollow fcc hollow

MAE 715 – Atomistic Modeling of Materials 30

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 H on the Al(111) surface (side view)

ontop bridge fcc hollow hcp hollow

H/Al(111) Fermi Level Vacuum (eV) Work

(eV) Function (eV)
ontop 1.050 6.106 5.056 Clean surface:

bridge 0.4527 4.791 4.338 4.235 eV

fcc hollow 0.5922 4.807 4.215

hcp hollow 0.6391 4.803 4.164

MAE 715 – Atomistic Modeling of Materials 31

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 H on the Al(111) surface

Calculations of Work function of H/Al(111) ontop site

H coverage Fermi Level Vacuum (eV) Work

(eV) Function (eV)
0.25 1.299 5.772 4.473
0.50 1.174 5.887 4.713
1.00 1.050 6.106 5.056

MAE 715 – Atomistic Modeling of Materials 32

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 H on the Al(111) surface

 Adsorption at the ontop and bridge site increase the work function while at the
hollow sites decrease the work function.

 This is due to the dipole induced by H-adsorption: when the H atom at the ontop
and bridge site, it pulls away the electron from the surface. However, when the
induced dipole opposes the spill-out of the electrons, it reduces the work function.

 The work function increases with the increase of the H coverage. This is
because at the low coverage, the dipole-dipole interaction will keep the atoms apart
, while at high coverage, the same interaction will cause a depolarization of the
dipoles and increase the work function.

MAE 715 – Atomistic Modeling of Materials 33

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Conclusion

 The method can be used to calculate the accurate work function.

 The change of work function depends on the surface orientation, adsorption

sites and the adsorption coverage.

 work function is the fundamental properties of the electronic structure of the

surface. Its measurement can give valuable insight into the condition of a given
surface.

 This method can also be extended to semiconductor.

MAE 715 – Atomistic Modeling of Materials 34

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)
 Acknowledgement

 Prof. Zabaras
 MPDCC cluster for the computation
 Software: Quantum Espressor

Thank you!

MAE 715 – Atomistic Modeling of Materials 35

CORNELL
U N I V E R S I T Y
N. Zabaras (5/7/2007)