EELS in the TEM

Contents

- Introduction
- Generation of the energy loss signal
- Instrumentation
- The spectrum
- Quantitative analysis
- Plasmons
- Fine structure
- Channeling
Introduction

EELS
• Energy resolution
0.1-1eV
• Absorption edges
• Background high
• Fine structure
observed
• More L and M edges
than K peaks
observed
 First EELS measurements

Rudberg: electron spectrum depends on concentration (1929)

First Transmission Spectrum

Ruthemann: first
electron
spectrum in
transmission
(5keV) (1941)
First EELS - TEM
 Spectrometer Performances

First experiments
Hillier,Baker: 20nm probe for EELS
-> first inner shell losses recorded
Möllenstedt: spectral resolution 1/50000 (i.e.
2eV at 100keV primary electrons)
Today:
Spectral resolutions: 0.1-1eV
Spatial resolutions: 0.1-1nm
A typical spectrum

Typical events:
• ZLP
• Plasmons
• Innershell
ionisation edges
• Electron loss near
edge structure
• Extended electron
loss fine structure
Spectrum TiO

ELNES
The EELS spectrometer

Spectrometer
tasks :
• Disperse
energy in x
direction
• Focus in
dispersion
plane
Coupling electrons to spectrometer
STEM TEM
• α semi-
convergence
angle
• β semi-
collection
angle
• L β entrance
aperture

Best spectral resolution:

Electron diffraction pattern on spectrometer entrance slit
 Different coupling modes
Dedicated STEM TEM: diffraction coupling mode
Parallel detection

Components:
• Entrance slit
• 90° prism
• Qx-Sy:
Correctiors
• Q1-4:
Magnifiers
• Photodiode:
Parallel
acquisiton
of spectrum
 Post-column Image Filter

Image formation
and correction

Dispersion
Energy
selection

Image filter:
• Energy filtered image formation
• Post column-> can be coupled to all TEMs
• Spatial resolution: typically 0.5-1nm on conventional TEM
Generation of the energy loss signal
• Elastic
Inelastic and elastic
scattering Ko=Kf
• Inelastic
Ko≠Kf
• θE Characteristic angle
(θE=Eav/(γm0v2) )
• Conserv. Momentum:
q= Ko2 (θ2+ θE2)
 Elastic cross section I
Born approximation

a: Bohr radius
f: atomic scattering factor
Nuclear screening (Yukawa or Wentzel potential)

r0: screening radius

 Elastic cross section II
Differential cross section including screening

θ0= 1/(k0r0)

-> dσ/dΩ Lorentian curve

-> scattering cutoff angle typically 100mrad at 100keV electrons
 Elastic cross section III
Inelastic differential cross section

E: average energy loss

θ0= E/(γm0v2)
Elastic / inelastic scattering
• Inelast. scattering
at smaller angles
than elastic
• Differential cross
section: dσ/dΩ
• θE: dσ/dΩ starts to
decrease
(θE = 0.2mrad in C for
100keV electrons)
• θ0 = 20mrad in C
for 100keV
electrons
Total inelastic cross section

-> EELS good

for Alkali
metals, only
low signal
for rare gases
 Why large difference between atomic σ
and solid state σ
dσ/dΩ ~ E-1 * df/dE
-> df/dE weak dependence of oscillator strength
!! atomic σ dominated by valence band
ionisation
!! solid state σ dominated by plasmon losses
AND: EP>Eion in many cases (i.e.: Al)
-> peak in solid state shifted to higher Energy
-> due to 1/E evolution lower σ in solid state
 Plasmon losses

• In a solid: most interaction of fast electrons

with outer shells E<100eV
• Plasmons: collective (quasi) free electron
excitations in the solid
-> described by Plasmon energy Ep, width dE
and mean free path MFP
Plasmon MFP and acceleration voltage

MFP of plasmons êm
MFP/m MFP for GaAs -> increase of
2× 10-7
acceleration voltage
200nm leads to increase of
MFP
1.5× 10-7
a) Plasmon scattering
main source of
1× 10
-7
100nm multiple scattering
b) Multiple scattering
degrades core loss
5× 10-8
edges
150kV 300kV -> work at sample
Kinetic Energie êeV
50000 100000 150000 200000 250000 300000
thickness < MFP
Acceleration voltage (eV)
 Typical mean free paths

Data for 100keV electrons

• Good agreement between calc. and exp.Ep
• MFP typically between 100-200nm
• Cutoff angle smaller than 10mrad
Multiple scattering

• Electrons are scattered

due to Poisson
statistics
• Probability of n-fold
scattering:
Pn= (In/I) =
(1/n!) (t/λ)n exp(- t/λ)
λ: mean free path
t: sample thickness
 Calculation of sample thickness

P0= (I0/I)= exp(- t/λ)

-> t= λ ln(- It/ I0)
-> measure sample thickness from low loss region
 Background modelling

Background contributions:
- core loss
- valence electron scattering
- plural scattering
-> Powerlaw for background fit function:

I= A*E-r
 Background Extrapolation

Bckgrd
1) Fit background in
pre-edge region
2) Extrapolate
background model to
edge region
3) Subtract background
from total intensity in
∆IQuant intensity
ITi ∆IQ -> edge intensity ITi
Plural scattering in core loss edges

Effect of multiple scattering

To prevent core loss scattering -> t<0.2MFP

Ideal edge follows theoretical cross sections
Deconvolution

Deconvolution
-> removal of
multiple
scattering
 Quantification
Different edge indices k and j

∆ quantification interval same offset and width with respect

to edge onset

For deconvolved edges edge indices k and j

∆ quantification interval different offset and width with

respect to edge onset -> more flexible for choice of ∆
Example AlxGa1-xAs
The edges
Fitting Hartree Slater cross sections

Best fit at low sample

thickness

-> Use thin sample for

analysis
Characterisation of sample surface

Problem:
1.5nm thick non
stoichiometric
film at sample
surface
-> use thick sample
Here compromise:
Sample thickness
60-120nm
 Quantification: Precision

1.1

c(bck)
1 c(deconv)
c(nom)
Precision of
0.9
AlxGa1-xAs
c(bck)
0.8
quantification:
0.7

0.6
∆x=0.02
0.5

0.4
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
c(nom)
Signal to Noise Ratio – Trace Elements

Which method is more sensitive: EELS or EDS?

•up to Z=25:
-> EELS more
sensitive
•For Z>25
-> use EDS
(Leapman, 1991)
And for non trace elements: EELS or EDS

Even for Z= 50
-> EELS is more
than 20 times
more sensitive
-> Acquisition
time
(EELS/EDS)=
1/400 for same
statistical
precision
ELNES

• Graphite: π
orbitals
• Diamond:
only σ orbitals
-> π* peak
appears
 Conduction Band Shifts in
Strained Quantum Wells (Batson
Probe Size 2Å, energy et al. 1995)
Si

resolution 200meV, 3 n m G e 0 .4 S i0 . 6
resolution of energy
Si
shifts: ±20meV

-> strain induced peak

splitting of 280meV
Observed!!
Observation of GaN bandgap

Brockt 1995:
Bandgap in
AlN and GaN
observed
Plasmon shifts: phase identification

Williams, Egerton
1976
Al13%Li
Plasmon shift
analysis: sharp
interface
Filtered Images

Filtered images with at

energy loss E and with
energy window width
∆ are acquired
• Three window
technique leads to
background subtracted
elemental map
 Example Filtered Images
BF image Ni L

Hofer (Micron 97)

Light element map
OK Cr L
Nanometer
resolution
 Spectrum images

EELS map containing 100 colums and 30 lines

-> different segments in vertical quantum well identified
 Channeling
E beam
Sample surface
Bloch wave calculation:
AlAs Electrons redistribute
4 nm around atomic colums
-> Channeling pattern
-> excitation of elemental
8nm edges depends on
channel pattern (and
not only on cross
12nm section)
Orientation dependence of channeling in AlAs
Reduction of channeling by increase of convergence angle

But high convergence

angle reduces
spatial resolution
Using small electron probes

Pennycook
->Monolayer
EELS
resolution
High energy resolution: Monochromator

Microsc. Microanalysis 9, 2003

From TiO2 to
TiO distortion
decreases
->
Monochromated
TEM (0.2meV)
shows additional
fine structure
 Oscillator strength and double
differential cross section
• Formula 3.22
• Formula 3.24
• Formula 3.25
• Formula 3.29
• -> in EELS consider angular and energy
distribution
Differential inelastic cross section

• Inokuti p 144

Potential 145

• Inelastic event only with electrons not

nucleus
Choice of experimental conditions

• Convergence collection angle

• Beam energy
• Typical acquisition times