You are on page 1of 69

MFlores dfi-FCFM-UChile

Microscopia de efecto tnel

Marcos Flores
DFI-FCFM-UChile
MFlores dfi-FCFM-UChile

Outline

Short description
Theory of 1-D tunneling
Actual 3D barriers
o tip modeling
o atomic resolution
Hardware
Examples
MFlores dfi-FCFM-UChile

Scanning Tunneling Microscopy (STM)

Fundamental process:

Electron tunneling
MFlores dfi-FCFM-UChile

Electron tunneling

Typical quantum phenomenon

Tunneling definition
Wave-particle impinging on barrier
Probability of finding the particle beyond the barrier
The particle have tunneled through it

Role of tunneling in physics and knowledge development:


Field emission from metals in high E field ( Fowler-Nordheim 1928)
Interband tunneling in solids (Zener 1934)
Field emission microscope (Mller 1937)
Tunneling in degenerate p-n junctions (Esaki 1958)
Perturbation theory of tunneling (Bardeen 1961)
Inelastic tunneling spectroscopy (Jaklevic, Lambe 1966)
Vacuum tunneling (Young 1971)
Scanning Tunneling Microscopy (Binnig and Rohrer 1982)
MFlores dfi-FCFM-UChile

Electron tunneling

Elastic Inelastic
Energy conservation during the Energy loss during the process
process. Initial and final states Interaction with elementary
have same energy excitations (phonons, plasmons)

1D 3D
Planar Metal-Oxide-Metal Scanning Tunneling Microscopy
junctions

Rectangular barriers
Planar Metal-Oxide-Metal
junctions

Time independent Time-dependent


Matching solutions of TI TD perturbation approach:
Schroedinger eq. (t) + first order pert. theory
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Time independent
!2 d 2
+V0 = E
2m dz 2

Region 1 1 = eikz + R eikz

Region 2 2 = Ae xz + Be xz

Region 3 3 =Te ikz


MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Continuity conditions on and d/dz give

Reflection probability:

2 2
R =
2 (k 2
+x
) sinh (xs ) 2

2 2
4k x 2 2
+ (k + x ) sinh (xs )
2 2

Transmission probability:

2 4k 2x 2
T = 2
4k 2x 2 + (k 2 + x 2 ) sinh 2 (xs )
2 2
R +T =1
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

A square integrable (normalized) wave


function has to remain normalized in time

d + 2
(z ,t ) dz = 0
dt

In a finite space region this conditions


becomes
b
d b 2 i! * *

dt a dz = 2m z z = 0
a

i! # * * & Probability current


j ( z, t ) = % (
2m $ z z '
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

For the potential barrier

!k
Region 1 j1 ( z, t ) =
2m
( 2
) (
1 R = v 1 R
2
)
ji (z ,t ) = v

!k 2 2
Region 3 jT (z ,t ) = T = vT
2m

ji 2 1
Transmission coefficient T= = T = 2 2 2
jT 2 (k + x )
1+ sinh (xs)
(4k 2 x 2 )
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

For strongly attenuating barriers xs >> 1


Large barrier height (i.e. small )
Exponential is leading contribution

16k 2x 2
T 2 e 2xs

(k + x 2 )2

Barrier width s = 0.5 nm, V0 = 4 eV T ~ 10-5


Barrier width s = 0.4 nm, V0 = 4 eV T ~ 10-4

The transmission coefficient


->Extreme sensitivity to z
depends exponentially on barrier width
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Exponential dependence of tunneling current


MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Ideal situation: Incident state from left has


some probability to appear on right, we can
calculate it.
Real situation: At the surface the
wavefunction is very complicated to calculate

Different approach
If barrier transmission is small, use
perturbation theory
We must to write a perturbed Hamiltonian
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Different approach
If barrier transmission is small, use
perturbation theory
We must to write a perturbed Hamiltonian

Approximate solutions of exact Hamiltonian


within the barrier region

l (z ) = ae kz for z 0
r (z ) = be kz for z s

l has to be matched with the


correct solution of H for z 0 r has to be matched with the
correct solution of H for z 0
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

With the exact Hamiltonian on left and right,


we add a term HT representing the transition
rate from l to r.

d (t)
H (t) = i!
dt
Where the Hamiltonian is
H =(H l + H r ) + H T = H 0 + H T
The left and right solutions

H 0l = Ell
H 0r = Err
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

We choose the wavefunction


iEl t iErt
= c l e ! + d r e !

Put into Hamiltonian

d (t )
H (t ) = i!
dt

iEl t iErt

HT c l e ! + d r e ! =0


MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

The total probability over the space is


*
H T dz = 1

* *
l T r r H Tl dz = 1
H dz +
M lr + M rl = 1

2 2 dN
jt = Mrl
! dEr
In general, the tunneling current contains
information on the density of states of one of
the electrodes, weighted by M
But, each case has to calculated separately
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Introduce a more real potential V(z), which is slowly varying


potential, how to represent it?

!2 d 2
+V (z ) = E
2m dz 2

d 2 (z ) 2m (E V (z ))
= x 2 (z ) with x2 =
dz 2 !2

z
ix (z ')dz '
Try a solution (z ) = 0e
0
particle moving to the right with
continuously varying wave-number (x)

d 2 (z ) dx (z )
= i (z ) x 2 (z ) (z )
dz 2 dz
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

This is true only if the first term is negligible, i.e.

dx (z )
<< x 2
dz

WKB approximation: Wenzel- Kramer Brillouin


variation length-scale of x(z) (approximately the same as the
variation length-scale of V(z)) must be much greater than the
particle's de Broglie wavelength

2 2
For E > V(z), x is real and the probability density is constant (z ) = 0
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Suppose the particle encounters a barrier between 0 < z1 < z2,


so E < V(z) and x is imaginary. Inside the barrier

z1 z z
ix (z ')dz ' z x (z ') dz ' z1 x (z ')dz '
(z ) = 0e 0
e 1
= 1e

The probability density inside the barrier is


z
2 2 z1 2 x (z ') dz '
(z ) = 1 e
2
The probability density at z1 is 1 z2
2 2 z1 2 x (z ') dz '
The probability density at z2 is 2 = 1 e
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

So the transmission coefficient becomes

2 z2
2 z1 2 x(z') dz'
T= 2
=e
1
2 z2

!
z1 2m(EV ( z') ) dz'
=e
The tunneling probability is very small.
The wavenumber is continuosly varying due to the potential: more real.
Reasonable approximation for the tunneling probability if the incident
<< z (width of the potential barrier)
MFlores dfi-FCFM-UChile

Electron tunneling across 1D potential barrier

Square barrier 16k 2 x 2 2 xs Exponential dependence


plane wave T 2 2 2 e of the transmission coefficient
(k + x )

Current depends on transfer


Square barrier 2 2 dN matrix elements (containing exp.
electron states jt = Mrl Dependence) and on DOS
! dEr
True barrier representation if
Real barrier 2 z2 <<z. Varying exponential
z1 2m (E V (z '))dz '
T=e
dependence of the transmission
Plane waves !
coefficient
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Planar tunnel junctions.


insulator = vacuum.
The insulator defines the barrier.
Similar free electron like electrodes.
At equilibrium there is no net
tunneling current and the Fermi level
is aligned.

V (z ) = EF + (z )
1

U=Bias voltage
What is the net current if we apply a bias voltage?
We must consider the Fermi distribution of electrons
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Flux from electrode 1 to electrode 2

Emax
1
N1 =
m
n(vz )T(Ez )dEz
0

Where
vz = electron speed along z. v r2 = v x2 + v y2
m
Er =
2v r2
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Flux from electrode 1 to electrode 2

Emax
1
N1 =
m
n(vz )T(Ez )dEz
0

Where
vz = electron speed along z.
n(vz)dvz = number of electrons/volume with vz

m4 m3
n (v )dv x dv y dvz = f (E )dv x dv y dvz n ( vz ) =

4 !
3 3
2 2 !3
0
f (E)dEr
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Flux from electrode 1 to electrode 2

Emax
1
N1 =
m
n(vz )T(Ez )dEz
0

Where
vz = electron speed along z.
n(vz)dvz = number of electrons/volume with vz
T(Ez) = transmission coefficient of e- tunneling through V(z) with
energy Ez
2 2m s2
s1 E F1 + ( z ) E z dz
T(E z ) = e !
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Flux from electrode 1 to electrode 2

Emax
1
N1 =
m
n(vz )T(Ez )dEz
0

Where
vz = electron speed along z.
n(vz)dvz = number of electrons/volume with vz
T(Ez) = transmission coefficient of e- tunneling through V(z) with
energy Ez.
f(E) is the fermi distribution 1
f (E ) = E EF
1+e KT
MFlores dfi-FCFM-UChile

Electron tunneling across 1D metal electrode

Flux from electrode 1 to electrode 2


E max
m2
N1 = 2 3 T(E z )dEz f (E )dE r
2 ! 0
0

Flux from electrode 2 at positive potential


U to electrode 1
E max
m 2

N2 = 2 3 T(E z )dEz f (E + eU )dE r
2 ! 0
0

Total number of electrons tunneling across junction


Emax
+ m2 % (.
N = N1 N 2 = T(Ez )dEz ,
- 2 !
3'
& 0
f (E)dEr 0
f (E + eU)dEr */
)0
0
MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes

since V (z ) = EF + (z )
1

2 2m s2
s1 E F1 + ( z ) E z dz
T(E z ) = e !

E max
J = T(E z )( 1 2 )dE z
0

T is small when EF-Ez is large

e- close to the Fermi level of the negatively biased electrode


contribute more effectively to the tunneling current

For positive U 2 is negligible so the net current flows from 1 to 2


MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
Applications of tunnel equation

To perform the integration over the barrier


1 s1
define =
s s (z )dz
1

2s 2m
A=
!
By integration it can be shown that
A EF + E z
T(E z ) e 1

em em
At 0 K 1 =
2 2!3
(EF E z
1
) 2 =
2 !
2 3
(
EF Ez eU 1
)
eV 0 < E z < E F1 eU
em
= 2 3 E F1 E z E F eU < E z < E F
2 ! 1 1

0 Ez > EF 1
hence
em E F eU A E F1 + E z E F1 A E F1 + E z
J 2 3
2 !
eU 0
1
e dE z +
E F1 eU
(E F1 )
Ez e dE z

MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
integrating

e
J
4 2!s 2
{ e A
( + eU )e A +eU
} =
1
s
s1
s (z )dz
1

2s 2m
A=
!
Current density flowing from electrode 1 to electrode 2 and vice versa

If V = 0 dynamic equilibrium: current density flowing in either direction


e
J1 e A
e
4 !s
2 2
J2 ( + eU )e A +eU

4 !s
2 2

For positive U 2 is negligible so


the net current flows from 1 to 2
MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
Low biases
e
eU << J
4 !s
2 2
{ e A
( + eU )e A +eU
}
k
k b b
(a + b ) = a 1 + a k 1 + k
k

a a

eU eU eU
A 1+ A 1+

A
e A +eU
=e
e 2
= e A e 2

eU
e A
J 2 ( + eU e
) 2
e A

4 !s 2

MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
Low biases eV <<

Neglect second order contributions in U

e
)1 A eU
A
J ( + eU e
=
4 !s
2 2
2

e eU A
+ A eU e
=
4 !s
2 2
2
e A
eU 1e A
4 !s
2 2 2

e A
A J eU e A
since >> 1 4 2!s 2
2
2
At low biases the current
2s 2m e 2 2m varies linearly with
A= J U e A
applied voltage, i.e.
!
4 2! 2 s Ohmic behavior
MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
High biases eU >
0 s0 U
=
2 s = F =
eU s
Electric field strength

Put into general eq.

e
J
4 !s
2 2
{ e A
( + eU )e A +eU
}
evaluating a numerical factor (not included in eq)

3
3
4 2eU
2.2e F 3 24
2m 0
2 2eU
2.96 !eF
2m 0
2 1+

J e e

2.96 !eF
1 + 0

16 2 !0 0

For this condition eU > Second term of eq is negligible


MFlores dfi-FCFM-UChile
Electron tunneling across 1-D metal electrodes
High biases
3

2.2e F 3 2 4
2m0
2

J e

2.96!eF
16 !0
2

const
J U 2e

U

EF2 lies below the bottom of CB1


Hence e- cannot tunnel from 2 to 1
there are no levels available

The situation is reversed for e- tunneling from 1 to 2: all available levels are empty
analogous to field emission from a metal electrode: Fowler-Nordheim regime
MFlores dfi-FCFM-UChile

Electron tunneling across 1-D potential barrier


16k 2x 2
Square barrier, T 2 e 2xs
Exponential dependence
plane wave (k + x 2 )2 of the transmission coefficient

Square barrier 2 2 dN current depends on transfer matrix


electron states jt = Mrl elements (containing exp. dependence)
! dEr and on DOS

Real barrier 2 z2 Varying exponential dependence


z1 2m (E V (z '))dz '
T=e
Plane waves ! of the transmission coefficient

E max
m2
Real barrier N = N1 N2 = T(E z )dE z
0 2!
3
[ f (E )dE

0
z

0
]
f (E + eU )dE r

Metal electrodes

Tunneling is most effective for e- close to Fermi level

Current flows from to + electrode Low biases: Ohmic behavior


High biases: Fowler-Nordheim
MFlores dfi-FCFM-UChile
3-D potential barrier

Square barrier 2 2 dN
jt = Mrl
electron states ! dEr
E max
m2
Real barrier
Metal electrodes
N = N1 N2 = T(E z )dE z
0 2!
3
[ f (E )dE

0
z

0
]

f (E + eU )dE r

Join and extend the expression to have the equation for the tunneling current
between a tip and a metal surface

2
1) Matrix element
Mrl = l*HT r dz
Consider two many particle states of the sytem 0,
0
= state with e- from state in left to state in right side of barrier

0, are eigenstates given by the WKB approximation


z
ix (z ')dz '
(z ) = 0e 0


Trick: both are good on one side only and inside
the barrier but not on the other side of the barrier
MFlores dfi-FCFM-UChile
3-D potential barrier
is linear combination of one intial state 0 and numerous final states

= a 0e iE t + b e iEt
0

Put into Schroedinger equation and get a matrix with elements like

!2
M = *
H(
H 0 dzdS
*
)
2m
0

Applyng a step function along z that is 1 only over barrier region

!2
M = dS *
( *
)
2m

the tunneling matrix element can be evaluated by integrating


a current-like operator over a plane lying in the insulator slab

The tunneling current depends on the electronic states of tip and surface

Problem: calculation of the surface AND tip wavefunctions


MFlores dfi-FCFM-UChile
3-D potential barrier

Square barrier 2 2 dN
jt = Mrl
electron states ! dEr
E max
m2
Real barrier
Metal electrodes
N = N1 N2 = T(E z )dE z
0 2!
3
[ f (E )dE

0
z

0
]
f (E + eU )dE r

Join the expression to have the equation for the tunneling current
between a tip and a metal surface

2) Current density
2e 2
I = {f (E ) f (E
+ eU )}M (E E )
!

f(E) = Fermi function


U = bias voltage applied to the sample
M = tunneling matrix element
= unperturbed electronic states of the surface Not the many particle states
= unperturbed electronic states of the tip
E (E) = energy of the state () in the absence of tunneling
, are not eigenfunctions of the same H
MFlores dfi-FCFM-UChile
3-D potential barrier

2e 2
I = {f (E ) f (E + eU )} M (E E )
!

f (E )f (E + eU )

2e 2
I = {f (E )[1 f (E + eU )] f (E + eU )[1 f (E )]} M (E E )
!

At low T one can consider only one directional tunneling

2e 2
I = {f (E )[1 f (E + eU )]} M (E E )
!
MFlores dfi-FCFM-UChile
2e 2
I = {f (E )[1 f (E + eU )]} M (E E ) 1
!

f (E ) = E EF
1+e KT
Low T + small (10 meV) applied bias voltage (U)

f (E )
f (E + eU ) f (E ) + eU =
E E f (E )
For the Fermi function = (E )
E
= f (E ) + eU (E EF )

2
{f (E )[1 f (E + eU )]} M (E E ) =

2
{f (E )[1 f (E ) + eU (E E )]} M (E E ) =
F

{f (E )[1 f (E )]M (E E ) + f (E )eU M (E



2

2
}
E F ) (E E F )

Low T + small (10 meV) applied bias voltage (U), E <EF f (E ) 1

2e 2 2
I = U M (E E F ) (E E F )
!
MFlores dfi-FCFM-UChile
Tip modeling
Low T + small applied bias voltage (U)
2e 2 2
I = U M (E E F ) (E E F )
!

Point like tip (unphysical)

2
I (r0 ) (E E F )

The matrix element is proportional to the


probability density of surface states measured at r0
i.e. the local density of states at the Fermi level

The image represents a contour map of


the surface DOS at the Fermi level
MFlores dfi-FCFM-UChile
Tip modeling
Low T + small applied bias voltage (U)
2e 2 2
I = U M (E E F ) (E E F )
!

tip with radius R


s-type only (quantum numbers l 0 neglected) wave functions
with spherical symmetry to calculate the matrix element

2
I U nt (E F ) e 2xR (r0 ) (E EF ) EF = Fermi energy
r0 = center of curvature of the tip

x = (2m)1/2/ = decay rate
= effective potential barrier height
nt(EF) = density of states
at the Fermi level for the tip

2
(r0 , E F ) = (r0 ) (E E F )

Surface local density of states (LDOS) at EF


measured at r0
MFlores dfi-FCFM-UChile
Tip modeling
2
I U nt (E F ) e 2xR (r0 ) (E EF )

The matrix element is integrated in a point of the barrier region s
So the value of at r0 is no physically relevant, but it represents the lateral
averaging due to finite tip size

I e 2xR (r0 , E F )
The exponential dependence comes from the matrix element

2
(r0 ) e 2x (s +R )

I e 2xs (r0 , E F )

STM is imaging the LDOS at the tip position


Multiplied by the tip DOS
MFlores dfi-FCFM-UChile
Tip modeling
Sample wavefunctions have
exponential decay in the z direction
so little corrugation at s from surface

Tip center
position

Surface local density


of states (LDOS) at EF
measured at r0

2
Au lattice
parameter (r0 ) e 2x (s +R )
Calculated LDOS for Au(111)

STM is imaging the LDOS at the tip position


Multiplied by the tip DOS Low T + small applied bias voltage (U)
MFlores dfi-FCFM-UChile
STM: atomic resolution

1.0

We observe features with a spatial resolution better than 0.1 nm


much lower of the tip curvature radius
Smaller than spherical approximation of the tip wavefunctions (0.8 nm)

Model failing to explain the most important feature


of the STM: atomic resolution
MFlores dfi-FCFM-UChile
STM: atomic resolution
Why?
Accuracy of perturbation theory:
depends critically on the choice of the unperturbed wave
functions, or the unperturbed Hamiltonians. 1.0

For 3D tunneling the choice of unperturbed Hamiltonians is


not unique. This is especially true for higher biases, in
which the potential in the tunneling gap is not flat.

Solution
the unperturbed wave functions
of sample and tip has
to be different in the gap region

!2 2
+ US E = 0
2me
!2 2
+ UT E = 0
2me

This unperturbed Hamiltonian minimizes the error introduced by


neglecting the higher terms in the perturbation series.
The tip states are invariant as the bias changes, simplify calculations.
Easier estimation of bias distortion because the bias only affects
the sample wave function, thus can be treated perturbatively
MFlores dfi-FCFM-UChile
STM: atomic resolution
To calculate I, the of the acting atom is expanded in
terms of a complete set of eigenfunctions.

Two choices:
spherical coordinates
parabolic coordinates

Spherical coordinates are appropriate


for describing atom loosely bonded on the tip

Parabolic coordinates are appropriate


for describing atom tightly bonded to the tip body.

, ,
x = cos
y = sin
( )
z =
2

!2 * h h
M = =0 h dd
*

2m
Calculated on the paraboloid
MFlores dfi-FCFM-UChile
STM: atomic resolution

!2 * h h
M = =0 h dd
*

2m

Differences to Bardeen expression

the wave functions are the eigenfunctions


of tip and sample unperturbed Hamiltonians
which are different in the gap region.

It is valid only on the paraboloid that is the boundary


of the tip body, not in the entire barrier region

what is needed for calculating the tunneling matrix elements


is the wave functions on the boundary of the tip
MFlores dfi-FCFM-UChile
STM: atomic resolution
On and outside boundary the tip satisfies the free electron
Schroedinger equation decaying exponentially
2mE
( 2
) = 0
2
2 =
!2
expand in term of the parabolic eigenfunctions
with boundary conditions to be regular at r unperturbed wave functions
of sample and tip different
= lmil (r )lm ( , ) in the gap region

l ,m
The contribution of the tip wave function is determined only by its asymptotic values.

The details of the tip wave functions near the center of the acting atom are not important

On and inside boundary the sample satisfies the free electron ( 2


2 ) = 0
Schroedinger equation decaying exponentially
2mE
expand in term of the parabolic eigenfunctions 2 =
with boundary conditions to be regular at center of the acting atom !2
= lmkl (r )lm ( , )
l ,m
The contribution of the sample wave function is determined only by the values of the sample
wave function in the vicinity of the center of the acting atom.

The details of the sample wave functions outside the tip body are not important
MFlores dfi-FCFM-UChile
STM: atomic resolution
!2 * h h
M = =0 h dd
*

2m

So M has to be integrated using orthonormal wavefunctions


lm ( , ) Spherical harmonics

il (r ), kl (r ) Bessel functions

!2
M = lm lm
2m l m
,

That leads to determine only the coefficients of the tip and


sample expansion on orthonormal wavefunctions.

The coefficients are determined by calculating the derivatives


of the at the center of the acting atom

= lmil (r )lm ( , )
l ,m

M gives the correspondence between tip and sample wavefunctions


MFlores dfi-FCFM-UChile
STM: atomic resolution
!2
M = lm lm = lmil (r )lm ( , )
2m l m
, l ,m
For a choosen tip state, M changes and defines
the relation to the coeffiecients of the surface

Tip states = e r M =
e r + (r0 )
s
2r
xe r x x + x r0
x
4r 2
p
ye r
y y y + y
4r 2 r0

( x 2 y 2 )e r 2 2
2
) (x y 2 )+
(x y ) 2
2 2
8r 3 (x 2
y 2 x 2
y r0
d
xye r
xy 2
4r 3 2 xy xy xy r0

The tunneling matrix elements are related to the sample wavefunction derivatives
MFlores dfi-FCFM-UChile
STM: atomic resolution

The approximation on s state only is wrong


the surface state of a real W tip
extends into vacuum more than s and d states

So the atomic resolution is given by the l 0 wave functions

It is the most protruding electronic states that provides the J


Not only the electron states at the Fermi level
MFlores dfi-FCFM-UChile
STM: atomic resolution
Reciprocity principle

Is a basic microscopic symmetry ofSTM

If the "acting" electronic state of the tip and the sample


state are interchanged, the image should be the same.

An image of microscopic scale may be interpreted either as by


probing the sample state with a tip state or by probing the tip
state with a sample state
MFlores dfi-FCFM-UChile
Band structure effects
eU
I t ( eU E ) s (E T) (E , eU )dE
0
2m t +s eU
2( s +R ) + E z
T (E , eU ) = e !2 2 2

The electron energy in a solid depends on the band structure

E = E (k) k is such that k+G=k

The surface and tip define the direction z

E = E (k ) = E (kz ) + E k ( )
This may results in tunneling from surface or bulk states depending on their spatial extension

Also T is changing as a function of E

2m t + s eU
x =
!2 2
+
2
E (kz ) E k

( )
Electrons in states with large parallel wavevector tunnel less effectively
MFlores dfi-FCFM-UChile
Constant current imaging

Unchanged
Tunneling Unchanged
Current (nA) z Tunneling
Current (nA)

Typical working mode

Constant height imaging

Higher Lower
Tunneling Tunneling
Current (nA) Current (nA)

Applied only on very flat regions


MFlores dfi-FCFM-UChile
Constant current imaging

Imaging: spatial configuration and energy dependence of electron states (LDOS)


need not to correspond in any simple way to the atomic positions

Example: linear lattice Si and Ge (111) cleaved surfaces

i
x
i
x
x
=e
+ a +e a = 2 cos
a
i
x
i
x
x
=e
a e a = 2i sin
a

At the Bragg reflection the potential


gives rise to a forbidden energy region
The band gap
MFlores dfi-FCFM-UChile
Constant current imaging
Imaging: spatial configuration and energy dependence of electron states (LDOS)
need not to correspond in any simple way to the atomic positions

2 x
+ = + cos 2 Charge density ON atomic positions
a
2 x Charge density BETWEEN atomic positions

= sin2
a

In the image always topographic AND electronic features


MFlores dfi-FCFM-UChile

Finite bias
2
eU = 0.01 eV I U nt (E F ) e 2xR (r0 ) (E EF )

But for eU about 1 eV?

The sum has to be done on many different states


Larger distortion of tip and sample wavefunctions

Approximation

Use undistorted tip and sample wavefunctions also at finite bias


2
I U nt (E F ) e 2xR (r0 ) (E EF )

eU
I t ( eU E ) s (E , r )dE 0 Integral over all e- states
0 up to eU from Fermi level
at the tip position
DOStip DOSsample
MFlores dfi-FCFM-UChile
eU
Finite bias
I t ( eU E ) s (E , r0 )dE
0
But DOS sample decays into vacuum depending on barrier
defined by the tip-sample distance so use WKB approximation
2m t +s eU
2( s +R ) 2
+ E
s (E , r0 ) s (E )e ! 2 2
= s (E T
) (E , eU )

The M now appears as DOS but


the effects of finite biases are
included as modified x
eU
I t ( eU E ) s (E T) (E , eU )dE
0

Integral over all electronic states


up to eU from Fermi level
Imaging occupied or unoccupied states
MFlores dfi-FCFM-UChile
Finite bias
eU
I t ( eU E ) s (E T) (E , eU )dE
0
What does it means imaging occupied or unoccupied electronic states?
Occupied Unoccupied

At constant current means tunneling At constant current means tunneling


from all sample occupied states into all from all tip occupied states into all
tip unoccupied states sample unoccupied states
All is defined by bias voltage All is defined by bias voltage
MFlores dfi-FCFM-UChile
Finite bias
eU
I t ( eU E ) s (E T) (E , eU )dE
0
Integral over all electronic states from Fermi level up to eU
The information is geometric and electronic and is convoluted

To separate the two one can collect images at different biases

The two states give different TOTAL intensity in the image


MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
eU
Integral over all electronic states
I t ( eU E ) s (E T) (E , eU )dE up to eU from Fermi level
0

The current is proportional to the occupied or unoccupied integral DOS

For metals the dI/dU is proportional to DOS at a given energy (low eU)

eU
dI dT (E , eU )
t (0 ) s (eU T
) (eU , eU ) + t ( eU E ) s (E ) dE
dU 0
dU
background

However this cannot be measured at constant current with feedback loop on

Large voltage dependent background due to T

Solution: dI/dU at constant separation (feedback loop off)


MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
eU
dI
) (eU , eU ) + t ( eU E ) s (E )dT (E , eU )dE
t (0 ) s (eU T
dU 0
dU

For e- injection into semiconductor unoccupied state


The e- come mainly from EF so the I is mainly due to sample DOS

For e- injection into tip unoccupied state


The e- come mainly from lowest lying levels of semiconductor so:
problem: the I is mainly due to tip DOS?

For now, consider the tip DOS as constant so

eU
dI
) (eU , eU ) + s (E )dT (E , eU )dE
s (eU T
dU 0
dU
MFlores dfi-FCFM-UChile
Tunneling Spectroscopy

DOS background
eU
dI s (E ) dT (E , eU )
s (eU ) + dE
dU 0
T (eU , eU ) dU
I 1
eU
T (E , eU )
(E ) dE
U eU s
T (eU , eU ) Normalization term
0

For semiconductors no low voltage approximation: I needs to be normalized


2m t +t eU
2( s +R ) + E
T (E , eU ) = e !2 2 2

For U > 0 T(E,eU) < T(eU,eU) and maximum transmission occurs at E = eU


The terms have same order of magnitude

For U < 0 T(E,eU) > T(eU,eU) and maximum transmission occurs at E = 0

The background and denominator terms have same order of magnitude


Larger than sample DOS
MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
Acquiring STS spectra
Si(111)-(2x1)
Sample and hold technique

Stop the tip on a location


Disable feedback
Scan V and monitor I

Taken at different initial


measuring conditions, i.e.
different tip-sample distances
MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
Acquiring STS spectra

Measuring at the same time the dI/dV one obtains


the normalized conductance, independent of
Tip-sample distance

Data show that


the normalized conductance
does not depend on tip-sample distance
-bonded chain

Bulk DOS

occupied empty
MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
Band structure effects
2m 2meU 2
x = + k
!2 !2
2m 1
Measured voltage dependence of x x = 2
= 2.2 A Minimum value
!
The data allow to get
(about 4.2 eV)
and gives x = 22 nm-1

2m 2meU 2 2
x = + kz k
But what about the increase below 1 eV? !2 !2
1
Using this with the data one gets k = 1.1 A
Close to the maximum wavevector at the edge of SBZ
At low bias the current is dominated by states at the edge of SBZ
MFlores dfi-FCFM-UChile
Tunneling Spectroscopy
Obtaining STS images

Current-imaging Voltage-dependent imaging


dI/dV with lock in tunneling spectroscopy
Apply modulation (CITS)
Collect dI/dV while scanning Feedback on only 30% of the time Integrate over an energy
simultaneously at each point Collect dI/dV at fixed separation interval at state onset

-0.35 V +0.7 V

-0.7 V

-0.8 V

DOS at the set point


of imaging condition
Emphasize one state
Possible only in stable -1.7 V
tunneling conditions
(not in band gap) Need to be done at V following Spatial relationship between
topography of nuclei occupied and unoccupied states
MFlores dfi-FCFM-UChile
Scanning Tunneling Microscopy (STM)
Design and instrumentation
Approach mechanism

Enables the STM tip to be positioned


within tunneling distance of the sample

High precision scanning mechanism

Enables the tip to be rastered above the surface

Control electronics

Control tip-surface separation


Drive the scanning elements
Facilitate data acquisition.

Vibration isolation

The microscope must be designed to be insensitive


or isolated from ambient noise and vibrations.

Review of Scientific Instruments 60 (1989) 165


Surface Science Reports 26 (1996) 61

You might also like