Electron and photon spectrometries in (S)TEMs for extended characterizationfrom the nanoscale to the single atom

Christian ColliexLPS, CNRS & UPS, Orsay, France

colliex@lps.u-psud.fr

Pre-congress IFSM Advanced School(Sunday, September 19th, 2010)

Four EM imagesWhite and blackSame sizeBut no scaleAll show dislocations!!

What do you learn from an image?

Courtesy Ph. Buffat

What do you learn from an analytical TEM ?

from D. Williams & C. B. Carter

Our major sources of references outside…

… the contributions from my group at Orsay !!

1. O. Stéphan; 2. M. Kociak; 3. A. Gloter; 4. M. Walls

13 2 4

What can we learn from Electron Energy Loss Spectroscopy ?

A few selected examples

to learn more, attend symposium I 5

To identify atoms (one by one!) in a nanostructure

3nm

Peapods :

Gd@C82@SWCNT

A : HREM image

B : schematics

C : Superimposed chemicalmaps of both Gd N45 and C K signals caluclated from a 32x128 spectre-image

Suenaga K. et al. Science 290 (2000) 2281

To identify column of atoms (one by one!) in a nanostructure

La Mn Ti

1 nm

Atomic-resolved EELS mapping across an interface

To probe the electronic states (extremely) locally

SrTiO3

La2/3Sr1/3MnO3

La2/3Sr1/3MnO3

SrTiO3(substrate)

<100>

LSMO

STO

Manganite tunnel jonctions with giant magneto-resistivity

Tunnel barrier STO from 1,5 nm to 5,5 nm

From L. Samet et al. EPJB 34 (2003) 179

?

These changes can be correlated to changes in the valence state of Mn:

From 4+ Mn gets a 3+ contribution at the interface

modification of the signal at the interfaces

20 nm

Mn

635 640 645 650 655 660

Rel

ativ

e EE

LS in

tens

ity (e

lect

rons

)

Energy Loss (eV)

0 nm

- 0.3 nm

- 0.6 nm

- 0.9 nm

0.3 nm

0.6 nm

0.9 nm

5 nm4.7 nm4.4 nm

6.2 nm

5.3 nm

5.6 nm

5.9 nm

50000

LSMO

STOLSM

O

L3 L2

Mn-2p

EELS map (Mn)

Evolution of Mn spectroscopic signature acrossthe LSMO/STO/LSMO interface

For estimating the optical gap (and its nature) and extracting optical constants of semi-conductor materials

Gap measurements for different momentumtranfers in GaN (selection of electronictransitions associated to various highsymmetry points in the Brillouin zone)

S. Lazar et al. Ultramicroscopy 73 (2006) 035312

Optical constants (refractive index and extinction coefficients) in NiSi deduced froma local EELS spectrum in a bi-layer of NiSi/Ni2Si2Comparison with optical techniques

M.C. Cheynet et al. Micron 37 (2006) 377

GaN

NiSi

Optical Optical measurementsmeasurementsEELSEELS

Gap measurements from surface energy losses

Aloof geometry:

BN below 10 eV

R. Arenal, O.Stephan, M. Kociak, D. Taverna, C. Colliex, A. Loiseau . (PRL 2005)Value independent of the number of walls and diameter

P(ω) ~ Im(ε⊥ - 1/ε// )

e-

Inte

nsity

(arb

. uni

ts)

876543Energy Loss (eV)

Three walls surfaceSW 2.3 nmSW 1.8 nm SW 1.4 nm

5.8 - 5.9 eV

For knowing « the color » of a nanoparticle

From the PhD thesis work of J. Nelayah (Orsay University, 2007)

Local measurement of optical (1.5 eV~850 nm, 3.8 ~ 330 nm) properties!

1

2

1

2

3

3

U.V.I.R.

About the physics of EELS

Which signal?

Photons- visible light, X-rays - continuum radiation- Cerenkov radiation

Electrons- secondary- Auger - backscattered

Incident electrons

Elastically scattered electronsBragg Scattering

InelasticallyInelastically scatteredscatteredelectronselectrons

An EELS experimentmeasures the energylost by primaryelectrons TransmittedTransmitted electronselectrons

Principle of an EELS experiment

EE-ΔEHow to measure E?

dΩ

θ

e electron (fast particle)

k’

k

Target = solid

E0 , k

E’, k’

θ

q

k’

k

Measured quantities:

d2σ(E,q)dEdΩ

Momentum transfer q

Energy loss E

E’-E0 = E

EELS spectroscopy : spectral domains

0 100 200 300 400 5000

0.5

1.0

1.5

2.0

2.5

Energy loss (eV)

Inte

nsity

(cou

nts

num

ber x

10

6 )

600 700

IR

Phonons

x50

Low losses

visibleUV

Plasmons

Plasmons, dielectric function, joint density of states, optical gap

x106

Core lossesCK

MnL2,3

RX

Absorption edges

Absorption edges:Intensity: elemental quantificationShape: structure and local bonding

EELS: Involved electron populations and associated transitions

Low losses

0 10 20 30 40Energy loss (eV)

Plasmon modes Unoccupied conduction states

Core lossesL edge)

690630 650 670Energy loss (eV)

MnL2,3

Mn 2p core states

250 300 400350

CK

Energy loss (eV)

C 1s core states

Core losses (K edge)

Occupied valence states

EF

EELS spectroscopy : spectral domains

250 300 400350

CK

250 300 400350

CK

Energy loss (eV)690630 650 670

Energy loss (eV)

MnL2,3

Map with high accuracy the nature,the position and bonding

of the atoms responsible for thestructural properties

of real materials(defects, interfaces, nanomaterials)

Requires instruments withbest spatial and energy resolutions

(0.1 nm, 0.1 eV)

Core energy-loss domain

Map different physical parameters, electronic,

optical or magnetic, which are especially important

for electronic industries

Requires instruments adapted to measure the properties of interest

at the relevant scale

Towards the nanolaboratory

Low energy-loss domain

In all cases, develop the theory for interpretingspectroscopical data, i.e. a physics of excited states

0 10 20 30 40Energy loss (eV)

Plasmon modes

Core electron energy loss spectroscopy

Edge classification and dipolar selection rules

Basic shapes (ionisation of an isolated atom, transitions to continuum states)

Spectroscopic notations based on the principal quantum number n of the core level(initial state), i.e. K, L, M, N, OSpin-orbit coupling causes splittingL, M, N, O 2,3M, N, O, 4,5N, O 6,7(based on the angular momentum quantum number) l

Transitions occur to unoccupied electronicstate (continuum state) whenΔl = ± 1 (« dipole » selection rules; thisapplies in the limit q 0, moderated energylosses and small scattering angles)

e-

atom ionisation

Shape of the edges depend on the Coulombicpotential that the ejected electron « feels » whenleaving the atom (centrifugal barrier)

Absorption edges domain : three types of information

Identification of elements

Elementary quantification

Study of the unoccupied electron states distribution

250 300 400350

CK

Energy loss (eV)

Quantitative elemental analysis

200 300 400Energy loss (eV)

BK

CK

NKIn

tensit

y

S

σ

Characteristic signal : proportional to the number of atoms per unit area for the element detected in the analysed area

S = ct. I N σ

Atomic concentration ratios:

NA

NB=

SA

SB

σB

σA

Local thickness measurement

Poissonian distribution of scattering events

Measurement of the local thickness t

Estimation of the total inelastic mean free path λ

From Egerton (1996) section 5.1.1

t/λ mapping

sample: Ag nanoparticle (50 nm edge length)

deposited on clived mica

I0=Ite-t/ λ

I0

It

almost no difference between both signals (the inelastic signal is very weak as compared to the non scattered signal)

t/λ

EELS fine structures : measuring a local density of states

One probes final states

∂ 2σ∂E∂Ω

∝ φ fr q .

r r j

j∑ φi

2

δ(εi −ε f + h ω) = M(h ω)D(h ω)

sp2

sp3

2p states

- of given symmetry

Energy Loss (eV)

NK

BK

σ∗

π∗

0 20 40 60

Hexagonal boron nitride

- on a given atomic site

Electronic structure and theoretical approaches

Multiple scattering

Band structures

Molecularorbitals

Multiplet theory

Influence of core hole

Periodicalstructure

Cluster of 0.5 nm radius

(7 layers)

Single atom+

Crystal field

LMTO FLAPW (Wien)Pseudo-Pot (Vasp, Castep,(DFT + LDA or GGA)Ab-init, …)LKKR

FEFFICXANESCONTINUUM

SCF LCAOSCF Xα

Atomicmultiplet

MetalSemiconductor

InsulatorCorrelated systems. L 2,3 for 3d and 4d TM. M4,5 for for rare earths

Cluster of 0.2 nm radius(1 or 2 layers)

courtesy Virginie Serin

EELS in (S)TEM(combining imaging and spectroscopy)

• The «datacube » and the different modes of acquisition

• Application to filtered images and spatially-resolved EELS

The basic component :the magnetic spectrometer

In the TEM column : Castaing-Henry, Ω..

At the end of the TEM column : Gatan PEELS

R =γmveB

The magneticspectrometer or filter

The in-column filter

Zeiss

Zeiss Sesam microscope at StuttgartThe Ω filter design

FEI Titan G2 60-300kV

Gatan GIF imaging filter

The post-column filter

Combining spatial coordinates (images) and energy-loss data (spectrum)

The elemental bit of information in a 3D space

2 ways to record itspectrum imagingimaging spectrum (filtered images)

y

E

x

Spectrum-image3D data cube

One parallel (EELS) spectrum for one probe position

A

B Scanning the probe (with a STEM) over a

specimen area

y

x

E

E δxδy

δE

Jeanguillaume & Colliex, UM (1989)

I I I I250 300 350 400

0-

40-

(nm)

Energy Loss (eV)

SPECTRE LIGNE

A

B

SPECTRUM LINEHADF image

20 nm

450400350300Energy Loss (eV)

EELS spectrum

AB

Specimen

Magnetic spectrometer

Field emission gun

E

E -ΔE

o

o

CameraCCD

HADF detectors

Spectrum

Probe• 0.1 to 1nA• in 0.5 to 1 nm

Scanning coils

The spectrumimaging mode

100 keV

0.5 to 0.8 eV1 ms to 5 s

HADF

100 200 300 400 500 6000

1

2

3

4

5

6

Energy Loss (eV) x

100

0

Bore

B K

Carbon

C K

Calcium

Ca L

Azote

N K

Oxygen

O K

Elemental mapping of complex nanostructures

x

E

Image-spectrum3D data cube

y

E1 E2

One energy filteredimage at energy loss E

Pile-up of energy filtered images from E1 to E2

Energy filteredmode

E

E-ΔEFiltered Image :ΔE = energy loss

ω = slit width

ΔE

ω

Image on TEM screen

P. Bayle-Guillemaud DRFMC CEA Grenoble

Comparison of the two techniquesfor elemental mapping

Parallel fixed illumination (EFTEM)* Advantage: parallel acquired filtered image (106 pixels within a few seconds)* Drawback: loss of efficiency in data acquisition as only one energywindow is imaged at once,the remaining part of the spectrum being lost

The 3D data cube is built while varying in energy the selection slit

Convergent scanned illumination (STEM/PEELS)* Advantage: minimum dose , all energy losses with high resolution beingacquired in parallel (100ms/pixel), filtered images can be calculated a posteriori + simultaneous multidetection (HADF)* Drawback: the image is acquired while scanning the probe step per step (total recording time is proportional to number of pixels)

The 3D data cube is built while scanning the probe

20 nm

Beyond elemental mapping : mapping bonding states

180 200 220 240 260Energy loss (eV)

Inte

nsity

180 190 200 210Energy loss (eV)

Inte

nsity

BK

Reconstructed spectrum

Exp.

180 190 200 210Energy Loss (eV)

220

BKBN

B2O3

metallic B

ReferencesNNLS Fit

4

4

1

1

2

2

5

5

3

3

B2O3

BN

B

NNLSBN

B2O3

B

Reconstructed images for the different Boron K edges

Metal/oxide particles on a array of BN nanotubes

Bmet@BN@B2O3nanoparticles multilayers

10 nm

10 nm

10 nm

20 nm

20 nm

20 nm

Analysis of ferritin molecules

Courtesy R. Leapman (NIH)

Courtesy R. Leapman (NIH)

Ultimate detection limits

Courtesy K. Suenaga (AIST, Tsukuba, 2010)

EELS identification with atomic spatial resolution(role of Cs corrected probe forming lenses)

Best values reported end 2008

Titan XFEG (300kV) Ip = 600 pA, d = 0.1 nm

--------Ip = 40 pAd = 77pmΔE = 0.14 eV

Hitachi HD 2700C (200kV) Ip = 300 pA, d = 0.14nm

Nion C5 USTEM (100kV) Ip = 700 pA, d = 0.14 nm

To be updated during IMC 17

From D. Muller et al. Science 319 (2008) 1073)

Elemental mapping at the STO-LCMO interfaces

HAADF images (left and centre) and EELS elemental maps at the outer LCMO-STO interface (right) in the 100 (top) and 110 (bottom) substrate orientations. Notice the increased interdiffusion level in 110 case.

<100>

<110>

Mn

Ti

MnLa Ti

<100>

<110>

Mn La

Ti

Courtesy M. Walls, coll. Univ. Barcelona

0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0

0,00

0,25

0,50

0,75

1,00

Rel

ativ

e EE

LS In

tens

ity(m

axim

a no

rmal

ised

to 1

)

Distance (nm)

HA

AD

F intensity signal (a. u.)

0.5 nm

BTO Fea)

b)

a) atomically-resolved STEM-HAADF image of the rastered BTO/Fe area, and b) the correspondingelemental profiles across the studied BTO/Fe interface combined with the HAADF profile. The chemicalprofiles were extracted after a power-law backrgound substraction.

Courtesy L. Bocher (2010)

Blue : Ti L23Red : Ba M45Green : Fe L23Black : HAADF

1

1

0.2

0.2

2

2

∆x (nm)

0.1 0.1

0.3 0.3

1 1

∆E (e

V)

EFTEM70s

Monochromators80s-00s

Orsay STEM90s

IBM STEM90s

0.1

Cs correctors

U-STEM2008

Trends of the accessible performance in terms of spatial and

spectral resolution (updated in 2009)

Must be accompanied with a parallel development in data

processing and modelization tools(propagation of a sub-angström

electron probe across a thinspecimen, physics of the inelasticscattering, calculation of electron

density of states…)

Where are we now ?

Valence electron energy loss spectroscopy

Plasmon excitation in a Si thin foil as a function of foil thickness : it is a quantized excitation

The first detailed treatment of the importance of plasma effects in solids, and theirconsequence on the theory of metals wasmade in a series of papers by Bohm and Pines (1951-1953).

It has then been extended in a series of papersby Nozières and Pines(1958) dealing with the general question of electron-electroninteractions in solids, and their effects on a number of properties, such as the absorption of the electromagnetic radiation or the slowingdown of high energy electrons.

The discreteness of the spectrum of excitations of the electrons was recognized by coining the term« plasmons » to represent the elementaryexcitation of energy ħωp

Excitations in the low-loss region

Collective excitations of the valence electrons and inter-band transitions.

EF

Conduction BandUnoccupied States

2p core states

1s core states

Valence BandOccupied States

What is plasmon excitation ?

Plasmons described as free electron gas oscillations (Drude model)

case of a metal, non interband transitions

� Knowing the plasmon frequency gives insight in the electron density N

ωp (the plasmon energy= characteristic frequency of the oscillation) is the energy at which the Re(ε) is 0 and Im (ε) is small: resonance condition

Plasmonvolume polarisation

Plasmons are directly observable by EELS

Metallic (drude model)

ωp is the energy at which the Re(ε) is 0, and almost where Im(-1/ε) is maximum

Im(ε)

Re(ε)

Im(-1/ε)

Alloys composition determination: discontinuous reaction front in an aged AlLi alloys

Williams and Edington, JoM (1976)

55

Band gap measurements(renewal in experiments with the

development of monochromators)

Im (∑)

Im (-1/∑)(E-Eb)1/2

Indirect gap?Surface states?Instrumental effect?

GaN

S. Lazar et al.

S. Schamm et al.Ultramicroscopy 96 (2003) 535

1( ) Im( )

P ωε ω

⎛ ⎞−∝ ⎜ ⎟

⎝ ⎠

ε (ω) ε (ω)

e-e-e-

LowLow--lossloss EELS and EELS and plasmonplasmon modesmodesAnalysis of the bulk and surface optical properties of nanostructures

Extended medium Nano-objects

Bulk plasmons Surface Plasmons

Response function

Excitations probed

Dielectric constant

ε (ω)

e-

Surface Plasmons

ε (ω)

Plasmon energy(Drude model)

2 1Im1/ 1

llε

+⎛ ⎞⎜ ⎟+ +⎝ ⎠

pω

1Im1

εε

−⎛ ⎞⎜ ⎟+⎝ ⎠

p / 2 1lω +p / 2ω

Surface Plasmons+ Bulk plasmons

?

?

Global polarisability

B

A BC

D

MappingMapping surface surface plasmonplasmon resonancesresonances

of of triangulartriangular silversilver nanoprismsnanoprisms

( Sample L.M. Liz-Marzan et coll., Vigo, Spain)

Energy map of the “tip” mode

J. Nelayah et al. Nature Physics, 3, 348-353 (2007)

78 nm edge long 78 nm edge long nanoprismnanoprism

0.00.20.40.60.81.0

ω = 1.9 eV ω = 2.9 eV

ω = 3.4 eV

EELS simulations of EELS simulations of triangulartriangular Ag Ag nanoprismsnanoprisms

• 100 Kv electrons

• 78 nm long and 10 nm thick Ag prism

Courtesy J. Garcia de Abajo, Madrid

Modes in Split Ring Resonators(coll. S. Linden, N. Feth & M. Wegener, Karlsruhe)

The resonator

Data acquisition

Courtesy G. Boudarham & M. Kociak

Courtesy G. Boudarham (2010)

Photon Spectroscopiesfrom IR (1-2 eV) to X (a few keVs) spectral domains

Physics of signal generation for electron energy loss and photon emission

Core loss EELS and Xray emissionValence loss EELS and « visible »

Photon emission (cathodoluminescence)elemental analysis

electronic and optical properties

Collecting the X rays emittedfrom the specimen within a TEM column

Spurious X ray emissions

The X ray detector facingthe thin specimen

The conventional EDX detector

Improving energy resolutionwith WDS detectorswith bolometers

Improving energy resolutionwith a WDX spectrometercoupled to a TEM column

EELS EDX

energy resolution

spatial resolution

detection efficiency

elements easily detected

quantification

states probed

~0.5eV ~150eV

~0.2nm ~0.5nm (better now?)

high low

low Z, transition elements,lanthanides etc

Z>10(Z=5 possible)

often difficult more straightforward, but...

empty full

EELS and EDX chemical mapping: brief comparison

sample constraints thickness < 100nm (ideal 10-20nm)

can be bulk

signal-to-background ratio low high

Recent improvements in EDX spatial resolution

How to collect and analyse emitted photon wavelengths?

The strategy and device implementedon a VG STEM in Orsay

Courtesy L. Zagonel & M; Kociak

A new dedicated instrument for nano-cathodoluminescence

courtesy L. Zagonel (10/2009)

Courtesy L. Zagonel & M. Kociak

Spectrum-imaging of electron beam induced photon emission

Thank you very much for your attention

and let us departto the Windsor Barra Convention Centre

for the opening ceremony of IMC 17

More to listen to and to learnduring the whole congress