Electrical and chemical mapping of silicon pn junctions using energy-filtered X-ray PhotoElectron Emission Microscopy

HAL Id: tel-00765630

https://tel.archives-ouvertes.fr/tel-00765630

Submitted on 15 Dec 2012

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

using energy-filtered X-ray PhotoElectron Emission

Microscopy

Maylis Lavayssiere

To cite this version:

Maylis Lavayssiere. Electrical and chemical mapping of silicon pn junctions using energy-filtered X-ray

PhotoElectron Emission Microscopy. Instrumentation and Detectors [physics.ins-det]. Université de

Grenoble, 2011. English. �tel-00765630�

THÈSE

Pour obtenir le grade de

DOCTEUR DE L’UNIVERSITÉ DE GRENOBLE

Spécialité : Micro et Nano Electronique

Arrêté ministériel : 7 août 2006

Présentée par

Maylis LAVAYSSIÈRE

Thèse dirigée par Nicholas BARRETT

préparée au sein du Laboratoire d'Electronique et des

Technologies de l'Information (LETI) du CEA Grenoble

dans l'École Doctorale Electronique, Electrotechnique,

Automatisme et Traitement du Signal

Electrical and chemical mapping

of silicon pn junctions using

energy-filtered X-ray

PhotoElectron Emission

Microscopy

Thèse soutenue publiquement le 02 mars 2011,

devant le jury composé de :

M. Roland MADAR

LMGP PHELMA (Grenoble) - Président

M. Jacques CAZAUX

GRESPI/Matériaux Fonctionnels (Reims) - Rapporteur

M. Gerd SCHÖNHENSE

Johannes Gutenberg University (Mainz) - Rapporteur

M. Jean-Charles JOUD

SIMAP PHELMA (Grenoble) - Examinateur

M. Pierre MÜLLER

CINaM-CNRS (Marseille) - Examinateur

M. François ROCHET

Université Paris VI (Paris) - Examinateur

M. Nicholas BARRETT

CEA DSM IRAMIS (Saclay) - Directeur de thèse

M. Olivier RENAULT

Les albums ! et non in assables, mes joujoux !"

Afterthreeyearsanda oupleofmonthslearning,performingexperiments,thinkinganda

lotofwriting,thisPhDworkends. Ithasbeenafruitfulexperien ebothataprofessional

and a personal level whi h has aorded me intera ting with lotsof interesting people. I

wish to express my deepest gratitude to allthose who have ontributed tothis work, as

wellastothosewhohave ontributedtomakingthisperiodapleasantandenjoyabletime.

My gratitude goesto Jean-Claude Royerand Frédéri Laugier,who gave me the

op-portunity to work with them in the DPTS department and in the LCPO laboratory. I

alsowouldliketothankNar isoGamba orti,AmalChabli,FrançoisBertin, forproviding

knowledgeable dis ussions.

I am grateful to Ni k Barrett and Olivier Renault, who planned and laun hed the

XPEEM proje t,for the opportunity they giveme todo my PhD thesis in agreat work

environment. Without them, this thesis would not have been a omplished.

ForthisPhD,Iwouldliketothankmyreading ommitteemembers: Ja ques Cazaux,

andGerdS hönhensefortheirinterestinmystudies,andtheirhelpful omments. Iwould

also like to thank the other members of my oral defense ommittee: Roland Madar,

Jean-Charles Joud,PierreMüller andFrançoisRo het, fortheirtimeand theirinsightful

questions.

I amthankful to the OMICRON and FOCUS teams, spe ially Konrad Winkler,

Di-etmar Funnemann,Burkhard Kroemker, Matthias Es her, and NilsWeber fortheir kind

guidan e and ooperation during the entire period of my study, and for the help they

gave meon te hni al problems aswell assimulation ones.

I am deeply indebted to Denis Mariolle, for his expert guidan e, his patien e and

supportat alllevels; he helps me tomove forward inmy resear h.

I am also pleased to re ord my gratitude to Denis Renaud and Jean-Mi hel

Hart-mann, for the time they spent helping me at ea h pro ess step of the epitaxial sili on

doped samplefabri ation.

I further would like to mention Véronique Robert and Thomas Ernst for all those

s ienti and friendlydis ussions, and spe iallyVéronique for helping mein dis overing

Rivallin and Brigitte Florin whi h ontributions, detailed omments and insights, have

been of great value to me.

During this work I have ollaborated with many olleagues for whom I have great

regard, and I wish to extend my warmest thanks to those who have helped me with

my hemistry work, spe ially for the leaning steps and the numerous tests performed:

Jean-Mar Fabbri atGrenoble and Jo elyne Leroy at Sa lay and SOLEIL.

I'm highly obliged to Bruno Delomez, Julien Rault, Yanyu Mi, Giovanni Vana ore,

IngoKrug for allthe ni e momentsspent together on syn hrotronfa ility.

My gratitude also goes to all the people who helped performing my experiments in

theSOLEILsyn hrotron: Fausto Sirotti,MathieuSilly,and StéphanieBlan handin. I'm

also in debt of Claudine Chaleil, who gave me the opportunity to perform valuable

ex-periments.

I amgratefultoAlainFaure for hisguidan eand knowledgeable dis ussions not only

aboutultrava uum butalsoabout foodand ooking. It has been areal pleasure towork

withyou!

I wish to thank the whole laboratory for all the extra-work a tivities, spe ially the

oee breaks, and parti ularly all the olleagues with whom I have spent more time.

Thankyouallforthoseenjoyable momentsgathered inthe anteenfor thelun hbreaks:

ChristopheLi itra,Anne-MariePapon,RobertTru he,DominiqueLafond,Cé ileProvost,

Eri De Vito, Eugénie Martinez, Névine Ro hat and Ni olas Chevalier. I also want to

thankspe iallythe bestse retaryIhaveevermet: Marie-Andrée Lesbre,forherpatien e

and kindnessto assist mein many dierent ways anytime I needed.

My sin erethanks gotothe olddream-team olleagues whi h have be ometrue friends:

KhaledKaja,ClémentGaumer,CyrilAilliot,Aude Bailly,Mi haëlJublot,and Madeline

Lambert. Thank you for all the ni e and funny moments we shared together! Other

fellows have also reinfor ed this team whi h now I would a knowledge: Edouard

De-s haseaux,OlivierDesplats,MatthiasKühne, Hélène Rotella,AdelineGrenier,Matthieu

Py, Sylvain Pierre, Vi tor Malgras, Yann Pitis, Charles Bourin, Pauline Calka, Ra hid

Boujamaa,LinYouandLukasz Borowik. I owespe iallyClaireMathieu sin eI have

ap-pre iated our inspirational dis ussions, with lots of great moments engraved in memory.

Thank you for your warm support!

My sin ere gratitude goes to spe ial people whi h have guided me, helping me in

my hoi es: my physi s tea hers Daniel and Marie-Christine Labernède, and my

as-tronomer friends Philippe Dupouy, Jean Le a heux, François Colas, Guillaume

Blan- hard, FrançoiseColas, Martine Castets, Valérie Desnoux and Christian Buil. I amalso

thankfultomyoldfriendsfor beingagreat ompany alwayspresent, andsopatientwith

and allmy familyforalltheirlove anden ouragement. I dedi ate thiswork tothemand

toPapyLouis whi hisstillpresent withus. I wouldliketonish thesea knowledgments

with two people who I onsider belonging to my family: Pauline Rose, thank you for

your support and your understanding at the most happy, but also the heavy moments

during these years. Jean-Mar Navasthank youfor your onstant en ouragements! And

nally, my greatest regards and sin ere thanks for my loving, supportive, and patient

1 Imaging of sili on pn jun tions 3

1.1 Needs formultis ale hara terization of doping indevi es . . . 4

1.2 Stru ture of pn jun tionat thermodynami equilibrium . . . 4

1.2.1 The pn jun tion . . . 4

1.2.2 Band bending and depletion zone . . . 6

1.2.3 The built-in potential. . . 7

1.2.4 Lateral ele tri eld a ross pn jun tion . . . 7

1.3 2D dopantmappingof pn jun tions: state of the artof existing te hniques 7 1.3.1 S anning ion probe te hniques . . . 8

1.3.2 S anning Probe Mi ros opy Te hniques . . . 10

1.3.3 Ion mi ros opy-based te hniques . . . 15

1.3.4 Ele tron mi ros opy-basedte hniques . . . 16

1.3.5 Photoemission-basedmi ros opy. . . 20

1.3.6 Summary . . . 22

1.4 Imaging of pn jun tions with ele tron mi ros opies . . . 25

1.4.1 Introdu tionto ontrast . . . 25

1.4.2 State-of-the-art of the interpretation of ontrast when imaging pn jun tions with anele tron mi ros ope. . . 25

1.4.3 Positioningof our study inPEEM imaging . . . 30

1.5 Surfa e Photovoltage investigation: impa ton 2D dopant te hniques. . . 34

1.5.1 Carrier generation and re ombination . . . 34

1.5.2 Surfa e Photovoltage . . . 34

1.6 Con lusion . . . 36

2 Energy-Filtered XPEEM 39 2.1 Prin iples . . . 40

2.1.1 Photoele tron Spe tros opy . . . 40

2.1.2 PhotoEmission Ele tron Mi ros opy (PEEM) . . . 51

2.2 Energy-FilteredXPEEM . . . 53

2.2.1 Present trendsin energy-lteredPEEM . . . 53

2.2.2 The NanoESCA . . . 58

2.2.3 The laboratoryFo usedX-ray Sour e(FXS) . . . 62

2.2.4 Syn hrotron sour es . . . 63

2.3 Information available fromEnergy-Filtered PEEMimaging . . . 67

2.3.1 Data format . . . 67

2.3.3 Core-Levelimaging . . . 69

2.4 Improvement of state-of-the-art spatial resolution with ore-level labora-tory XPEEM . . . 72

2.4.1 Introdu tion . . . 72

2.4.2 Lateral resolution measurement . . . 73

2.4.3 Improving the lateralresolution . . . 74

2.4.4 Results a hieved with anAlK

α

laboratory X-ray sour e . . . 76

2.5 Con lusion . . . 77

3 Patterned doped sili on samples by lo alized epitaxy 79 3.1 Framework for XPEEM studies: photoemissionin sili on . . . 80

3.1.1 Flat band onditions and band bending . . . 80

3.1.2 Work fun tionand ore-level studies . . . 83

3.1.3 Obje tive . . . 84

3.2 Sample fabri ationand hara terization . . . 84

3.2.1 Requirements . . . 84

3.2.2 Chara terization . . . 91

3.3 Sample passivation . . . 96

3.3.1 Obje tive . . . 96

3.3.2 Literature review onsili on passivation . . . 97

3.3.3 The three-steps passivation proto ol. . . 99

3.3.4 XPS analysis of the passivation pro ess . . . 101

3.4 Summary and on lusions . . . 110

4 Energy-ltered XPEEM of passivated patterned sili on samples 113 4.1 Se ondary ele trons: workfun tion mapping . . . 114

4.1.1 Experimental onditions and work fun tion analysis proto ol . . . 115

4.1.2 Work fun tionmapping with laboratoryex itation . . . 116

4.1.3 Work fun tionmapping with syn hrotron radiation . . . 119

4.1.4 Summary . . . 122

4.2 Si 2p ore-levelmi rospe traand imaging. . . 123

4.2.1 Laboratory study . . . 123

4.2.2 Syn hrotron radiationstudy . . . 128

4.2.3 Summary . . . 137

4.3 Valen e bandmi rospe tros opy . . . 137

4.3.1 Laboratory He I ex itation . . . 138

4.3.2 Syn hrotron ex itation . . . 140

4.3.3 Summary . . . 142

4.4 Con lusion . . . 143

5 Dis ussion of the energy-ltered XPEEM imaging of pn jun tions 145 5.1 Contrast of threshold XPEEM images. . . 146

5.1.1 Denition of the ontrast . . . 146

5.1.2 Fa tors inuen ing the ontrast . . . 147

5.2.1 XPEEM imaging atthreshold . . . 152 5.2.2 Core-levelspe tromi ros opy . . . 156 5.2.3 Complementaryanalyses . . . 157 5.2.4 Dis ussion . . . 165 5.3 SIMION simulations . . . 166 5.3.1 Methodology . . . 166

5.3.2 Inuen e of ele tri eld onPEEM imaging . . . 172

5.3.3 Dark eldPEEM imaging . . . 186

5.3.4 Con lusion . . . 193

5.4 Con lusion . . . 194

6 Résumé 201 6.1 Introdu tion . . . 202

6.2 Positionnement de la te hnique XPEEM dans la ara térisation 2D de dopants . . . 202

6.3 XPEEM ltré en énergie: leNanoESCA . . . 204

6.3.1 Prin ipe . . . 204

6.3.2 Optimisation des onditions d'imagerie et mesure de résolution latérale. . . 206

6.4 Etude de motifs de sili ium dopés . . . 207

6.4.1 Fabri ation des é hantillons . . . 207

6.4.2 Préparation des surfa es : passivation des é hantillons. . . 208

6.4.3 Analyse spe tromi ros opique des é hantillons . . . 211

6.4.4 Contraste en XPEEM en imagerieniveau de ÷ur sur pi Si2p . 215 6.5 Imagerie XPEEM ltrée en énergiede jon tions pn . . . 217

6.5.1 Contraste en XPEEM au seuil de photoémission . . . 217

6.5.2 Contraste auseuil de photoémission. . . 218

6.5.3 Triple ontraste en imageriePEEM . . . 219

6.6 SimulationsSIMION . . . 223

6.6.1 Méthodologie . . . 223

6.6.2 Inuen e de divers paramètres sur les onditions d'imagerie PEEM 225 6.6.3 Imagerie PEEM en hamp sombre . . . 227

6.7 Con lusion . . . 228

AC Alternative Current

AFM Atomi For e Mi ros opy

ARXPS Angle-Resolved XPS

ATP Atom Probe Tomography

BSE Ba ks attered Ele trons

CA Contrast Aperture

CAE Constant Analyser Energy

CCD Charge Coupled Devi e

CEA Commissariat àl'Energie Atomique

CITS Current Imaging TunnellingSpe tros opy

CRR Constant Retard Ratio

d Dire t Current

DNA DeoxyriboNu lei A id

DOS Density of States

DPTS Departement Plate-formeTe hnologique Sili ium

DRT Dire tion de laRe her he Te hnologique

EA Energy Aperture

ESCA Ele tron Spe tros opy for Chemi al Analysis

FE SEM Field-Emission S anning Ele tronMi ros opy

FIB Fo used IonBeam

FIT Flexible Image Transport

FoV Field of View

FTIR FourierTransform InfraRed spe tros opy

FXS Fo used X-ray Sour e

HDA Hemispheri al Double Analyzer

HF HydroFluori a id

HFET Heterostru ture FieldEe t Transistor

HSA Hemispheri al Single Analyzer

IDEA Imaging Double Energy Analyzer

IRAMIS Institut Rayonnement Matière de Sa lay

KFM Kelvin For e Mi ros opy

LCPO Laboratoirede Cara terisation Physique O-line

LETI Laboratoired'Ele troniqueet de Te hnologiesde l'Information

LINAC LINear AC elerator

LURE Laboratoirepour l'Utilisationdu Rayonnement Ele tromagnetique

MCP Multi Channel Plate

MIR Multiple InternalRee tion

MOS Metal-Oxide-Semi ondu tor

MOSFET Metal-Oxide-Semi ondu tor Field-Ee t Transistor

MXPS Multi-ProbeXPS

NEMS Nanome hani alsystems

NEXAFS Near Edge X-ray Absorption FineStru ture

NanoESCA Nano Ele tronSpe tros opy for Chemi al Analysis

PEEM PhotoEle tron Emission Mi ros opy

PES Photoele tron Spe tros opy

PNA PeptideNu lei A id

RP Resolving Power

SCM S anning Capa itan e Mi ros opy

SCPIO Servi e de laCara terisation Physique In-line/O-line

SCR Spa e Charge Region

SE Se ondary Ele tron

SEM S anning Ele tron Mi ros opy

SIMS Se ondary IonMass Spe trometry

SLEEM S anning Low-Energy Ele tronMi ros ope

SMART Spe tro-Mi ros ope with Aberration orre tion for many Relevant Te hniques

SNR Signal-to-Noise Ratio

SOLEIL Sour e Optimisée de Lumière d'Energie Intermédiairede LURE

SPEM S anning PhotoEle tron Mi ros opy

SPHINX Spe tromi ros ope for PHotoele tron Imaging of Nanostru tures with X-rays

SPV Surfa e Photovoltage

SRAM Stati Random A ess Memory

SSRM S anning Spreading Resistan e Mi ros opy

STM S anning TunellingMi ros opy

TEM Transmission Ele tronMi ros opy

TEMPO Time resolved Experiments onMaterials with Photoele tron Spe tros opy

ToF Time-of-Flight

UHV Ultra HighVa uum

UPS UltravioletPhotoele tron Spe tros opy

UV Ultraviolet

VBM Valen e BandMaximum

XAS X-ray AbsorptionSpe tros opy

XMCD X-ray Magneti Cir ularDi hroism

XMLD X-ray Magneti Linear Di hroism

XPEEM X-ray PhotoEle tronEmission Mi ros opy

XPS X-ray Photoele tron Spe tros opy

1D one-dimensional

2D two-dimensional

Avogadro's number N

A

6.022

×

10

23

atoms permole

Boltzmann's onstant

k

1.38

×

10

−

23

J.K

−

1

8.62

×

10

−

5

eV.K

−

1

Ele troni harge

q

1.602

×

10

−

19

C Va uum permittivity

ǫ

0

8.854

×

10

−12

F.m

−1

8.854

×

10

−

14

F. m

−

1

Sili on permittivity

ǫ

0

1.04

×

10

−12

F.m

−1

1.04

×

10

−

14

F. m

−

1

Sili on gap

E

g,Si

1.12 eV

Sili on ele tron anity

χ

Si

4.05 eV Intrinsi arrier on entration for sili on

n

i

1.45

×

10

10

atoms. m

This thesis reports on the appli ation of energy ltered X-ray PhotoEle tron Emission

Mi ros opy (XPEEM) to analyze doped sili on patterns. It fo uses on two important

aspe ts:

•

the ele tri al and hemi al mapping of pn jun tions as a fun tion of the surfa e state,

•

the inuen e of the jun tions onPEEM imaging.

For the hara terization of these materials a new on ept of energy ltered PEEM

was used, and experiments were performed in both laboratory and syn hrotron

envi-ronments. We have obtained omplementary results to highlight our understanding of

the pn jun tions in terms of ele tri al and hemi al hara terization. Thanks to full

eld, energy-ltered ele tron imaging, this instrument dire tly measures spatial hanges

in work fun tion, ore-level and valen e band depending on the sili on doping type and

leveland the surfa e state.

Wehavedesigned,produ edand hara terizedoptimisedpatternedsili onsurfa esin

order toprobethe bandbendinginthedepletion zoneduetothe pnjun tions. However,

be auseof the unavoidable presen e of a native oxide onthe surfa e, wehave developed

a dedi ated passivation proto olto attainat band onditions onditions.

These ondissue isdedi atedtothe imaging onditionsthemselves, sin e,when

imag-ingpnjun tions,onehas totakeintoa ounttheinherentpropertiesofthejun tion. We

fo us onthelo allateralele tri aleld existinga rosspn jun tions,whi hinuen esthe

photoele trontraje toriesontheirextra tionfromthe surfa e. Thiseld analterlateral

resolution inPEEM imaging. Tobetterappre iate itsinuen e,we ompare simulations

with experiments, varying dierent parameters.

ThisthesisworkwasdoneattheCEALETIMINATECinGrenoble,in ollaboration

with the IRAMIS instituteatthe CEAof Sa lay and theSOLEILsyn hrotronfa ilityin

Saint-Aubin.

The organisationof this manus ript isdivided in ve hapters:

•

In hapter1,abriefpresentationofthephysi sgoverningsili onpnjun tionsisrst made. We position then our study in the 2D dopant hara terization te hniques

by reviewing the most important advantages and drawba ks of ea h of them. We

give then the various interpretations in terms of se ondary ele tron ontrast in

ele tron emissionmi ros opy. Wenishbyreviewingthe studiesalreadyperformed

•

Chapter 2 introdu es the physi al prin iples of energy lteredPEEM. We present the twomainlaboratoryandsyn hrotronX-raysour esusedduringthis thesis. We

then detail the performan es of our instrument in terms of energy resolution, and

the state-of-the-artlateral resolution a hieved in laboratory.

•

In hapter 3, the fabri ation pro ess of the dierent sili on doped samples is de-s ribed. It is ompleted by an extensive hara terization study of these patterned

samples. We detail then the optimisation of a passivation proto ol established to

remove the native oxidefrom the sample surfa e, and thus, rea h at band

ondi-tions. We show the advantages and limitations of su h a pro ess thanks to XPS

analysis of the Si 2p ore-leveland ompare the results withthe existing literature

on the Si 2p spe tra of passivated surfa es.

•

Chapter 4is dedi ated to the presentation of the results a quired with this energy ltered PEEM, by mi rospe tros opy and spe tromi ros opy at dierent energy

range: se ondary ele trons, ore-level and valen e band photoele trons. The

har-a terizationofthesamplesisdoneintermsofworkfun tionmapping,de onvolution

ofhigh-resolved ore-levelSi2p spe tra,andvalen ebandmaximumdetermination.

We present ore-level imagingintegrated over the whole Si2p peak.

•

Chapter 5 addresses rst the ontrast observed inphotoemissionthreshold PEEM at the jun tion. Se ondly, we study inmore detailthe triple ontrast observed for

losed mi ron s ale patterns. Modelling the losed regions as a diode provides a

oherent explanation of the triple ontrast. The hapter presents numeri al

sim-ulations of modi ations in the PEEM ontrast due to the presen e of a lateral

ele tri eld a ross the pn jun tion. Ele tri al and physi al topography inuen e

theapparentpatternsizeandjun tionpositionasobservedinPEEM.Aqualitative

omparisonwithexperimentprovidesabetterunderstandingofthephenomena.

Fi-nally,thepossibilityofobservingindire tlythelo alele tri aleldsusingdarkeld

Imaging of sili on pn jun tions

Contents

1.1 Needs for multis ale hara terization of doping in devi es . . 4

1.2 Stru ture of pn jun tion at thermodynami equilibrium . . . 4

1.2.1 The pnjun tion . . . 4

1.2.2 Bandbendingand depletion zone . . . 6

1.2.3 The built-inpotential . . . 7

1.2.4 Lateral ele tri elda rosspnjun tion . . . 7

1.3 2D dopant mapping of pn jun tions: state of the art of ex-isting te hniques . . . 7

1.3.1 S anning ionprobe te hniques . . . 8

1.3.2 S anning Probe Mi ros opyTe hniques . . . 10

1.3.3 Ion mi ros opy-basedte hniques . . . 15

1.3.4 Ele tron mi ros opy-based te hniques . . . 16

1.3.5 Photoemission-based mi ros opy . . . 20

1.3.6 Summary . . . 22

1.4 Imaging of pn jun tionswith ele tron mi ros opies . . . 25

1.4.1 Introdu tion to ontrast . . . 25

1.4.2 State-of-the-art of the interpretation of ontrast when imaging pn jun tionswithan ele tronmi ros ope . . . 25

1.4.3 Positioningof our studyinPEEM imaging . . . 30

1.5 Surfa e Photovoltage investigation: impa t on 2D dopant te hniques . . . 34

1.5.1 Carrier generation andre ombination . . . 34

1.5.2 Surfa e Photovoltage . . . 34

1.1 Needs for multis ale hara terization of doping in

devi es

The s aling of integrated ir uits for te hnologi al breakthrough for es improvements

in terms of design and fabri ation. To observe su h stru tures requires the

develop-ment apabilities of hara terization at dierent s ales, from few mi rons to a tenth of

a nanometer. The fabri ation pro esses must be fully understood sin e, for example,

defe ts an reate dopant diusion. In 2010, pro ess simulations are alsostillperformed

to des ribe in detail the intera tions between implantation doping and defe ts [1℄. The

devi es omplexity interms ofte hnologi alsteps grows with the needs of better

perfor-man es. To improve indeed the fabri ation of high-performan e devi es, it is ne essary

tounderstandthe surfa eproperties and dopant distribution,whateverthe devi e s ale.

In most ases, semi ondu ting materials are doped with impurity atoms to hange

their properties and reate jun tions. These jun tions an be ombined and exploited

to form devi es having a key role in integrated ir uits. One of the most ommon use

whi h an be itedistheMetal OxideSemi ondu tor FieldEe tTransistor(MOSFET)

devi e: smaller and smaller transistor devi es at the nano-s ale are ne essary for logi

appli ations, with urrent resear h performed on the 22 nm node. However, the

semi- ondu tor devi e lands ape also on erns larger s ales: this is the ase for transistors

dedi ated to power appli ations (Heterojun tion eld ee t transistors HFETs) and for

nano-andmi ro-ele trome hani alsystems (NEMSandMEMS). HFET devi espresent

gatelengthsrangedfromapproximately0.25to5

µ

mwithgatewidthsbetween50and800

µ

m to be high-e ien y and ompa t power ampliers. Sili on nanowires are expe ted to have appli ations in eld ee t transistors, sensors, resonators and thermoele tri

systems. Another example are mi roele trome hani al systems (MEMS) whi h refer to

mi ros opi devi es whi h have a hara teristi length of less than 1 mmbut more than

100nm, and nanoele trome hani alsystems(NEMS) refer tonanos opi devi es smaller

than100 nm, down to 10nm. Both ombine ele tri al and me hani al omponents.

The measurement of 2D arrier or dopant distribution a ross the jun tion is be oming

moreandmoreimportant,forexample,lo atingdopantswithever-in reasingpre isionat

dierents ales. A hievingtwodimensions(2D) hara terizationofele tri aland hemi al

statesrequires multidis iplinary te hniques atthe mi ro-to nano- s ales.

1.2 Stru ture of pn jun tion at thermodynami

equi-librium

1.2.1 The pn jun tion

When n-type and p-type sili on are joined together under equilibrium onditions, the

Fermilevelisat throughthe entire stru ture,and orrespondstothe straighthorizontal

dashedline ingure 1.1 a), apn jun tion is formed. This is not possible by joiningtwo

materials me hani ally, but by hemi al pro esses allowing the existen e of these two

dierent regionswithin a single rystal.

are majority arriers, while the region on the right is n-type with a donor density

N

d

determined by the ele trons on entration. The physi al lo alisationwhere the dierent

doping typesare in onta t is alledthe metalli jun tion. It isimportanttodistinguish

this notionfromthe ele tri aljun tion sin e,their positionsdonot always oin ide. The

latter parameter is dened as the zone inside the depletion region, where the ele tron

on entration is equalto the hole one.

Figure 1.1: a)Energybanddiagramof apnjun tionatequilibrium. b)Distributionof harges and )

representationof the potentialinthe spa e hargeregion atthermal equilibrium. Denition ofthe

built-inpotential

V

bi

. d)Representationof the ele tri al eldin thespa e hargeregion atthermal equilibrium.

1.2.2 Band bending and depletion zone

A physi alsystem inequilibrium doesnot mean that everything is stati : ele trons and

holesareindeedowinginbothdire tionsa rossthejun tion. Theele tronsleavebehind

themionizeddonors witha positive hargeinthe n-regionof the jun tion,and the holes

leaveanegative hargeinthep-one,given thepresen eofionizeda eptors. Thispro ess

reatesinthe pnjun tion, aregionaroundthe jun tionwherethe free arriers density is

negligible: the depletionregion orspa e harge zone.

Even asemi ondu tor whi h possesses lo alized ele troni surfa e states usually indu es

aperturbationof thelo al harge, reatingadepletionzone. Wewillalsostudythis kind

of depletion zone o urring, for example, at the buried interfa e with an oxide and the

samplesurfa e whi hresults in band bending.

The theoryof the depletionzone atthe surfa e ofa semi ondu torhas beengiven by

S hottky et al. [2℄ and Mott [3℄. Garrett et al. [4℄ also onsidered the properties of the

spa e hargeregionand therole ofsurfa estates atasemi ondu tor surfa e. The harge

redistributiondepends onthe surfa e dopant typeand onthe position of theFermilevel

atthe surfa e. Thesesurfa e states arry harge,whi hs reens anopposite hargeinside

the semi ondu tor material. The higher the ele tron density, the shorter the range over

whi h ele trons have torearrange inorder to establishanee tive shielding.

This depletion zone, is represented in gure 1.1. We onsider two frontiers on the

abs issa

x

p

and

x

n

respe tively for the p-doped zone and for the n-doped zone. The harge density in the pn jun tion an be expressed as:

ρ(x) =







0

for

x < x

p

and

x > x

n

−qN

a

for

x

p

< x < 0

qN

d

for

0 < x < x

n

where

q

isthe ele tri harge.

Its width depends on the doping on entrations of the n-doped and p-doped layers.

However, around

x

n

some ele trons an penetrate into the depletion zone, and the same phenomenonappearsfortheholesaround

x

p

. Thisleadstoaperturbationofthepotential ona ertain distan e alledthe Debye length,respe tively

L

Dn

inn-typeregion and

L

Dp

inp-type region.

L

Dn

=

s

kT ǫ

2q

2

_N

d

.

(1.1)

L

Dp

=

s

kT ǫ

2q

2

_N

a

.

(1.2)

Todetermine the spa e hargewidthin then-doped region,the use of the expression

1.1of the built-inpotentialyields:

W

n

= 2L

Dn



1

1 + N

d

/N

a

ln

 N

d

N

a

n

2

i



1

/2

.

(1.3)

Bythe sameway,using expression 1.2, the spa e harge widthin the p-doped region is:

W

p

= 2L

Dp



1

1 + N

a

/N

d

ln

 N

d

N

a

n

2

i



1

/2

,

(1.4)

The depletion width is given by the sum of the spa e harge extension of the two

regions:

W = W

n

+ W

p

. If the jun tion is strongly asymmetri , the spa e harge zone spreads over the less doped one.

1.2.3 The built-in potential

When, in thermalequilibrium,no external voltage is applied between the n-type and

p-typeregions,there isaninternalpotentialofthe ondu tionele tron whi hvaries froma

value

V

p

inthep-dopedregion,toavalue

V

n

inthe n-typeone. Sin ethermalequilibrium implies that the Fermi energy is onstant through the entire pn stru ture, the dieren e

betweenthesetworegions,is alledthebuilt-inpotential

V

bi

. Itisdenedasthedieren e inenergybetween then-typeandthep-typesemi ondu tor. Itsvalueis learlypresented

in gure 1.1 ) and isequal tothe full band bending inequilibrium.

As a onsequen e, the built-inpotentialis given by:

V

bi

=

kT

q

ln

 N

d

N

a

n

2

i



.

(1.5)

1.2.4 Lateral ele tri eld a ross pn jun tion

A ording to Poisson's equation, the harge distributiono urring in the depletion zone

results in an ele tri eld whi h is a ompanied by a band bending that an be

inter-preted as anenergy barrier.

The presen e of a spa e harge zone implies the existen e of a potential variationa ross

the barrier,andthus, thepresen eofanele tri aleld: itishighinthe majorpartofthe

depletion zone, as presented in gure 1.1 d). The width of the depletion zone depends

on the doping on entrationsof the n-doped and p-doped layers.

The behaviorof the ele tri aleld onthe x axisisgiven by

E = −dV/dx

forthe two doped regions:

E = −

qN

a

ǫ

(x − x

p

)

for

x

p

< x < 0 ,

(1.6)

E =

qN

d

ǫ

(x − x

n

)

for

0 < x < x

n

,

(1.7)

where

ǫ

is the va uumdiele tri permittivity.

1.3 2D dopant mapping of pn jun tions: state of the

art of existing te hniques

dopant proling te hniques the most relevant for sili on dopant imaging are presented

withtheirpresentadvantagesanddrawba ks. Te hniquesintrodu edhavebeen lassied

inve ategories:

•

S anning ion probe te hniques, su h as Se ondary Ion Mass Spe trometry (SIMS) and Time of Flight(ToF) SIMS.

•

S anning probe near-eld mi ros opy-based te hniques, su h as Kelvin For e Mi- ros opy(KFM),S anningCapa itan eMi ros opy(SCM)andS anningSpreading

Resistan e Mi ros opy (SSRM).

•

Ele tronmi ros opy-basedte hniques,su hasS anningEle tronMi ros opy(SEM) and ele tron holography.

•

Ionmi ros opy-basedte hniques whi h orrespondstotheAtomProbeTomography (APT) te hnique.

•

Photoemission-based mi ros opy te hniques, su h as S anning PhotoEle tron Mi- ros opy (SPEM) and PhotoEle tron EmissionMi ros opy (PEEM).

Theyare assessed onthebasis ofsensitivity, spatialresolution,natureof theinformation

provided by the te hnique (whi h an be either hemi al or ele tri al), and apa ity to

donon-destru tive analysis of real devi es [5℄.

1.3.1 S anning ion probe te hniques

1.3.1.1 2D Se ondary Ion Mass Spe trometry and Imaging SIMS

SIMS anprovidein-depth ompositionalinformationfrommaterials. It anbeextended

toinvestigate two- and three-dimensions (3D) elementaldistribution for tra e elements.

SIMS an beoperated either instati or indynami mode.

An energeti primary ion beam of 1 keV to 20 keV sputters the sample, removing

material from the surfa e. These primary parti les are either rea tive or inert. It is

a destru tive te hnique sin e the sample is damaged during the analysis [6℄. Only a

smallfra tionofthesputteredmaterial orrespondstoatomsormole ulesionized,either

positively or negatively and are alled se ondary ions, the rest being eje ted as neutral

atomsor mole ules. The se ondary ions are olle ted and separated in energy and mass

thankstothe ombinationofanenergy analyzerandamagneti analyzer,whi hformsa

se ondaryionmassspe trometer[7,8℄. Theenergyanalyzer anbease tor,aquadrupole

ora time-of-ight. During the ablation,the surfa e isremoved layer by layer, providing

ahighlysensitive hemi alprole on entration of the material.

The sample studied an be either bulk, thin layer or powder. The rater size an vary

from100 nm to 500

µ

m. Sputter rates an be adjusted to analyse depths ranging from many mi ronsto few nanometers. All elements and isotopes an be re orded, in luding

hydrogen, with a high surfa e sensitivity from ppm down to ppb. The quanti ation

requires referen e samples to alibrate the instrument: for ea h atomi spe ies present,

on entration alibration with respe t to the bulk matrix is established. This te hnique

isoneof the mostused fordopantdepthproling,thanks toitshighdynami sensitivity

Stati imaging SIMS Stati SIMSinvestigatesthe hemi al ompositionofthe

outer-most atomi monolayers, providing rapid hara terization of both organi and inorgani

spe iespresent onthe surfa ewith ahighsensitivity. In this onguration, a small

num-berofprimaryions are usedtoanalyse thesample. Ithelpsinstudying phenomenasu h

as orrosion or adsorption. It an be applied by using the Time of Flight ToF-SIMS

onguration, wherethese ondary ionseje tedare analyzedandseparatedthankstothe

fa t that ions lighter have higher velo ities than the heavier ones, and then, rea h the

dete tor earlier. Figure 1.2 shows some negative se ondary ion images obtained from a

Peptide Nu lei A id (PNA) biosensor hip whi h was hybridized with omplementary

deoxyribonu lei a id(DNA)[9℄. Thespatialdistributionofspe i hemi alspe ies an

be learly identied. By s anning a nely fo used primary ion beam over the sample

surfa e, and a quiringa mass spe trum at ea h pixel of the CCD dete tor, it ispossible

to arry out the hemi almapping. The lateralresolution an be down to100 nm but it

is limited by the ion beam diameter.

Figure 1.2: NegativeToF-SIMSimages ofaSiwafer PNA/DNAbiosensor hip. (25 keVBi

3+

,eld

of view(1500

×

1500) mm

2

. Light gray orrespondstohigh intensity[9℄.

Dynami 2D SIMS Dynami SIMSenables2D depth-prolinganalysis ofthin layers

and interfa es[10,11℄. Itismoredestru tivethanstati SIMS,sin eituses highintensity

primary ionbeamsand prolongedexposure. Su h onditionsallowtoremovemu hmore

material and toa quire a series of mass spe tra at dierent depths.

In this mode, material removal and mass spe trometry are arried out either

simulta-neously or in analternating mode, produ ing mass spe trometri images. In ToF-SIMS

onguration, the entire mass spe trum an bestored at ea h y le of the depth prole.

Then, the 2D dopant distribution is obtained by re ombination of these depth proles

thanks toalgorithms. The major s ienti impa tofdynami SIMSisthe in-depth

anal-ysis of semi ondu tor materials,where lowlevel dopants are analyzed in su essive thin

sli es of materialsdown to1 or 2nm thi k [12℄.

The 2D SIMS method was developed by Hill et al. but the te hnique remains

om-plex [13℄. Ukraintsev et al. demonstrated a simplest version where the SIMS lateral

resolution an be de reased down to the photomaskpixel size, i.e. 10 nm, for a dopant

on entration sensitivity of 10

17

atoms. m

−3

apabilitiesof the primary ionbeam to penetrate inside the sample. Moreover, the

di- ultiesto ombine high resolution with sensitivity in lowdoped areas limitsthe samples

to be studied in su h onditions. Se ondly, spatial resolution is limited by the beam

width,whi h an not be de reased under 50 nm. Finally,the stoi hiometri and matrix

properties of the sample must beknown tointerpret properlythe data.

1.3.2 S anning Probe Mi ros opy Te hniques

Two reviews present the Atomi For e Mi ros opy (AFM)-based ele tri al

hara teriza-tionte hniques[14,15℄. Table1.1summarizesthevariouste hniqueswhi h anbe

onsid-ered,displayingthe type ofprobe used andthe measured physi al quantity. Duhayonet

al. ompared S anning Capa itan eMi ros opy (SCM), S anning Spreading Resistan e

Mi ros opy (SSRM) and Kelvin probe For e Mi ros opy (KFM) [16℄ and De Wolf et al.

addedS anningTunnelingMi ros opy(STM)tothesethreemethods[17℄for omparison.

The basi hara teristi sof these four main te hniques are nowintrodu ed.

Te hnique Mode Probe Measuredquantity

STM STM Metalli needle

No. Dopingatoms I-Vspe tra Sele tiveet hing

NC-AFM UltrasharpSi

Topographyafter

+AFM hemi alet hing

SCM C-AFM

Metal- oatedSi Depletion apa itan e ormetalli C-Vspe tra

SSRM C-AFM

Diamond- oatedSi Ele tri alresistan e ormetalli I-Vspe tra

KFM NC-AFM

Metal- oatedSi Ele trostati potential ormetalli (ele tri eld)

SSHM STM

Metalli needle

Depletion apa itan e withmi rowave avity

Table 1.1: Summaryofthe dierent s anningprobe mi ros opy te hniqueswhi h an beutilizedfor

2D arrierproling. Column moderee tsthe s anning mode (NC:non onta t; C: onta t)[17℄.

SSHM orrespondstoS anningSurfa e Harmoni Mi ros opy.

S anning Transmission Mi ros opy This te hnique probes the surfa e

topograph-i aland ele tri al stru ture with high lateral resolution of ondu tive samples, down to

1 Å. It is able to image atom arrangements on surfa es. STM is at the origin of the

s anning probe mi ros opies. It was developed by Binnigand Rohrer who were awarded

withthe Nobelprize for physi s in1986.

The prin iple of STM onsists in tunneling of ele trons between two ele trodes,

orre-spondingrespe tively tothe outer atoms ofthe samplesurfa e and those of atip, under

anele tri eld. Tokeepa urrent onstant,apiezoele tri feedba ksystemenablessmall

movements by applying a voltage

V

to the probe when the latter is s anned above the sample. Its movement isre orded and displayed asan imageof the samplelo aldensity

of states,as presented in gure 1.3a).

STM anbeusedtoimageadepletionzone atasili onpn jun tionsurfa e[18,19℄. Clear

Figure 1.3: a)S hemati illustrationof STM measurement onananos alepn jun tion andresulting

image. b) ross-se tion of aMOSFETtransistorwith agate lengthof38 nmandSTM2D (CITS

mode) arrier proleextra ted[21,22℄.

This te hnique has some drawba ks: depending on the ele tri al ondu tivity in the

sample,thefor ebetween thetipandthislatter anvary, reatingme hani altip-sample

intera tions during the observations, whi hrenders the ability to position the tunneling

tip reliably and reprodu ibly more di ult[23℄.

Moreover,variationsofthedepletionlayerobserved byFukutome[21℄dependingonSTM

bias voltage is onsidered to be due to the lo al band bending aused partially by the

STM tip. This very surfa e sensitive te hnique must be performed under Ultra High

Va uum (UHV) to avoid ontamination and native oxide on the samples surfa e sin e

it requires a ondu tive surfa e. To measure bulk ele troni stru ture, it is important

to eliminate as mu h surfa e states or defe ts in the band gap as possible, sin e they

ause the surfa eFermilevel pinning[24℄. A possibilitytoavoidthemis topassivate the

surfa e, inorder toobtain near at band onditions [25℄.

Combiningthespe tros opi apabilitiesandthes anningabilityofanSTMgivesrise

to the Current Imaging Tunneling Spe tros opy (CITS) te hnique. This latter enables

to obtain 2D spatially resolved spe tros opi information from the tunneling urrent

hara teristi swhi hareusedto onstru timagesthatrevealatomi -tonanometer-s ale

variationsin ele troni stru ture onthe sample surfa e.

Su hasystemworksasfollows: stabilizingthe urrentataxedvalueforagivenvoltage,

a onstant- urrenttopographi s anisperformedoverthesamplesurfa e. Forea hpoint,

a urrent-voltage spe trum is measured, whose variations orrespond to variations in

ele troni stru ture a ross the sample surfa e. Plotting the urrent measured atspe i

bias voltages is alled a urrent image. In 2010, Fukutome [22℄ has realized a 2D view

arrier prole as a tunneling urrent image with UHV-STM in CITS mode of p-type

Metal-Oxide-Semi ondu tor Field-Ee t Transistor (MOSFET), with a gate length of

S anning Capa itan e Mi ros opy This te hnique measures the variation of the

ele tri apa ity

dC

of a MOS stru ture formed arti ially between a tip and an oxide- overed surfa e, when applying an ele tri al modulation

dV

. It measures the doping on entration,andthe arriertypeof ross-se tionorplanarsamples[26℄. Forpnjun tion

imaging,thete hnique isoftenutilizedat onstantvoltagemode: the alternating urrent

(a ) voltage is applied to the sample and the hange in apa itan e under the tip is

re orded. SCM has a lateral resolution of 10-20nm and the jun tion position isdire tly

obtained [2729℄.

The main drawba k is the sample preparation: ross-se tion samples are di ult to

obtainsin e itis very hard to ontrolthe atta kspeed and tostop atthe exa t required

lo alization. One solutionis touse Fo usedIon Beam(FIB) preparation ombinedwith

plasmaet hing [30℄. Moreover, the natureof the diele tri oxide overage on the sample

surfa eis of primeimportan e: high-quality surfa e oxidewith onstant thi kness inthe

s anned region is required, otherwise artifa ts an appear [31℄. The 2 nm native oxide

growing onthesili onistoothinwhereasathi keruniformoxideof3nmormore,grown

by dryor wetoxidationenables SCM measurements. The ombinationof this oxideand

diamond- oatedprobeissuitableforstable,reliableand reprodu iblemeasurements,but

remainsdi ult hallenging [32℄.

Figure1.4: a)S hemati ross-se tionof the wafer: AFMandSCM images ofthis waferb) with

mask misalignmentand )with orre tmaskalignment [33℄.

The breakthrough in sample preparation has avoided the ontrast reversal ee t,

hara terizedby the fa t that the SCM output isnot always amonotoni ally in reasing

signal with de reasing dopant on entration [34℄. This drawba k, whi h has prevented

quanti ationformanyyears anbeduetointerfa estates,semi ondu torwork-fun tion

extra tion have improved the measurements [36℄. Figure 1.4 shows that SCM visualizes

and identiesthe positionof thedopantareasand an helpforexample, inthedete tion

of wrongmask alignmentduring fabri ationpro ess: omparing b) and )images of the

n welladja ent totheP

+

areas represented bythe twored arrows onasili on wafer,one

observe that this latter is toonarrowin b)and that urrent leakage o urs [33℄.

SCMissensitiveto10

15

to10

20

atoms. m

−3

,andisalmostnondestru tive. Itsspatial

resolutionislimitedbythesensitivityofthe apa itan esensorandthetip-sample onta t

area, the tip geometry, and the dopant level and the topography of the sample surfa e.

Forthislatter ase, Buzzoet al.[37℄ ompareda2Ddopantanalysisof4Hsili on arbide

pn jun tionswithSCM toquantify the inuen eofthe surfa eroughnessonthejun tion

position.

S anning Spreading Resistan e Mi ros opy This te hnique onsists in s anning

a hard ondu tive diamond- oated sili on probe, in onta t mode a ross the sample. It

measures the ele tri resistan e between a onta t tip and the onta t on the sample

ba kside when applying a potential dieren e. The resistan e depends on the doping

levelbut itisnot possibletodetermine the arriertype. On ultrashallowpn jun tionfor

the delineation of arriers within sili on devi es, Zhang et al. [38℄ have a hieved a 1 nm

spatial resolution, as presented in gure1.5 a).

Figure 1.5: a)SSRMimages ofultrashallowjun tion p-typeMOSFETs, gate lengthsof a)135nm, b)

60nm, and )40 nm[39℄.

SSRM employs astrongtippressure and thus, both the samplesurfa e and the

s an-ning tip are damaged [40,41℄. The te hnique is less dependent than SCM to surfa e

preparation: it providesa better spatial resolution, sensitivity, quanti ationand

repro-du ibility [42,43℄. The dopant gradient resolution ishigher: 1-2nanometers per de ade,

thanks to the resistan e in rease for small arrier on entrations [44℄. SSRM also gives

some omplementaryinformationabout highlydoped region, and enables, asseen in

g-ure 1.5 b) and ) to distinguish an arseni -halo orresponding to impurities within the

p-type MOSFET devi e.

are omplementary ambient te hniques for dopant hara terization. Both are sensitive

tothe whole dynami rangebetween 5

×

10

14

atoms. m

−

3

and 2

×

10

21

atoms. m

−

3

.

KelvinFor eMi ros opy Thisisanon onta tandnondestru tiveele trostati for e

mi ros opyte hnique. Itsprin iple onsistsins anningasamplewithanAFM antilever

ex itedele tri allyby applyinga voltage (a +d ) tothe tip. Theamplitude of vibration

is proportional to the ele trostati restoring for e between the tip and the ondu tive

sample. This potentialdieren ebetween the tipand the sample isobtained by varying

thedire t urrentvoltageuntilthealternating urrentvibrationofthetipatthefrequen y

near the antilever resonan e be omes nil. The ele tro hemi al potential of the sample

surfa e is measured with respe t to the tip one. This potential depends on the

work-fun tiondieren e, fromwhi hone an dedu e the lo al arrier on entration. However,

quantitativepotentialprolingislimitedbysurfa e hargesandadsorbedmole uleswhi h

indu ebandbending. Sarafetal.[45℄havemeasuredthe2D-potentialdistributioninside

asymmetri pn jun tions. Combining with a 3D analysis of the tip-sample ele trostati

intera tion, they are able to dedu e the dire t lo al measurement of surfa e harge and

bandbendingin semi ondu tors.

Figure 1.6: a)Surfa epotentialimage ofapn jun tionarraymeasuredbyKFM[46℄. b)The

topography1-and surfa epotential2- images ofStati Random A essMemory (SRAM).The

ross-se tion prolesalong the dire tionsindi atedwith redarrowsin1- and2-areshownin the top

andbottom partsof 3-[47℄.

KFMissensitivetovariationsindopant on entrationbetween10

15

to10

20

atoms. m

−

3

.

Tsuietal. haveestablished orrelationsbetween surfa epotentialdieren eofapn

jun -tionand arrier-dopant on entration. Besides,the ross-se tionalproleofapnjun tion

array extra ted from its surfa e potential mapping has been su essfully demonstrated

byTsuietal.[46℄,andispresented ingure1.6a). Figure1.6b)shows the lear ontrast

whi h an beobtained dependingonthe dopanttypeforStati RandomA ess Memory

(SRAM)studies. The spatial resolution an attain 30nm using a lowhumidity

environ-ment, su h as nitrogengas ushing.

Ahigherimagingresolution anbea hievedbyusingasmallerprobetip,whenmodifying

theAtomi For eMi ros ope antileverwithaMulti-WalledCarbonNanotube(MWNT):

it helps resolving the dopant distribution to within 10 nm in air [48℄. The sample

1.3.3 Ion mi ros opy-based te hniques

1.3.3.1 Atom Probe Tomography

Atom ProbeTomography (APT) provides real analyti alatomi s alemappingof

hem-i al spe ies, in luding ultra-shallow and extremely highly-doped pn jun tions, as well

as MOS stru tures [49℄. It utilizes a high voltage to evaporate atoms as harged ions

one-by-one using, either a pulsed eld, or a mass spe trometer, and is ompleted by a

delay-lineposition-sensitivedete tor[50℄. Theoriginalpositionfromwhi htheatomeld

evaporated an be determined, and data are then rearranged in a 3D re onstru tion of

the volume using omputer al ulation. An introdu tiontothe history of this te hnique

and the underlyingphysi s is given by Millerand Forbes[51℄.

Thiste hniqueissensitivetoeldevaporatedionspe iesinaeldofviewoftypi ally

50 nm to 100 nm in diameter where it is possible to distinguish the physi al variations

in dopant distribution down to 0.5nm in lateralresolution [52℄. Figure 1.7presents two

atom maps of sili on devi es: a) is a Si:SiGeB:Si multilayer test stru ture. Here APT

spatial sensitivity for interfa e analysis is learly seen and also helps inidentifyingareas

of lo al Ge a umulation. Figure 1.7 b) shows the APT apabilities to provide atomi

s ale maps of hemi al spe ies, here when hara terizing a metal-oxide-semi ondu tor

(MOS) stru ture.

Figure 1.7: a)Si:SiGeB:Si multilayer teststru ture[53℄. b) Spatial distributionofelements withina

polySi/Hafnia high kdiele tri sta k. Ea h pointrepresents anatom [54℄.

The major limitation of this te hnique is the sample preparation, whi h requires

milling of sharpneedle shaped spe imensof several nanometers indiameter. Me hani al

ele tri-is olle ted in 3D, the data a quisition is long and the zone probed stays small, around

10

6

nm

3

. The lateral resolution is degraded for larger volumes. The dete tion level is

10

19

atoms. m

−3

for arseni and 5

×

10

18

atoms. m

−3

forboron [52℄.

Laser pulsing to evaporate atoms have drasti ally improved the apabilities of the

te h-nique in terms of mass resolution and the Signal-to-Noise Ratio (SNR) in rease, thus

improving the spatial resolving power of the instrument [49,57℄. This leads to a better

sensitivity tolow on entrations, and anultimate dete tionlimitof tens of ppm.

Anotherlimitationof the te hniqueisthe in lusionofmaterialswithdierentkineti

of eld evaporation in the same sample volume: when onsidering spe ies with higher

evaporationelds, they evaporate more slowly, ompli atingthe depth re onstru tion of

the volume.

1.3.4 Ele tron mi ros opy-based te hniques

1.3.4.1 Ele tron Holography

Ele tronholographyis ahigh resolution oherent interferometri te hnique. It maps the

lo al on entrationof a tivedopantsviaele trostati potentialdistribution[58,59℄. This

te hnique an be alsoused on reverse biased pn jun tions or deep submi ron transistor

stru tures[60℄. Re ently,Yooetal.[61℄reportedthequantitativeanalysisofapnjun tion

and estimated the built-in potential a ross the jun tion. Ele tron holography provides

a ess to the phase of the ele tron wavefront whi h has rossed the sample. It is a

te hnique allowing the ombination of nanometer s ale spatial resolution and su ient

sensitivity to dete t the implanted dopants [62,63℄ with doping levels from 10

16

to 10

21

atoms. m

−3

[16,64℄and aresolution of 10 nm [65℄.

The ele tron opti al geometry for o-axis ele tron holography in Transmission

Ele -tron Mi ros ope (TEM) is presented by Midgley[67℄: highly oherent ele trons emitted

froma eldemission gun are divided intoa referen ewave, traveling inva uum, and an

obje t wave, passing through the sample. Dueto its inner potentialwhi h in reases the

ele tron kineti energy, the ele tron wavelength be omes shorter in the spe imen than

in va uum. An ele trostati biprism pla ed after the sample bends the two waves so

that they form aninterferen e pattern alled hologramre orded by a2D dete tor. This

hologram ontains the ele tron wave, as presented in gure 1.8a): amplitude and phase

informations. The phase shifts relative to the referen e wave sin e the inner potential

hanges as a result of the built-in potential generated at the jun tion. Part b) of the

same gurepresents the re onstru ted amplitude and phase imageof the transistor, the

phaseimages being proportionalto the inner ele trostati potential distribution.

The sample preparationis di ult sin e this methodrequires spe imen whose thi kness

should be in the range of 200 nm to 400 nm, in order to optimize the signal-to-noise

ratio [68℄. Several methods exist, su h as tripod polishing, Ar ion beam milling and

FIB milling but they indu e ele tri ally dead layers near surfa e regions. Damage

in-du ed by FIB an be of two types: amorphisation of the sili on surfa e and gallium

implantation [69℄. Their presen e reate intera tions with the ele tron beam and ae t

the holography measurements [7072℄. Figure 1.8 ) illustrates this for FIB prepared

samples,the theoreti al phaseprole being ompared tothe experimentalone for

Figure 1.8: a)Hologramofa0.18

µ

m p-MOStransistorstru ture. b)Re onstru tedamplitude and phaseimage of thetransistor[60℄. )Topimages represent phaseimages of400 nm-thi k-spe imenfor

ea hof thepn jun tionswithdierent doping,andprepared by onventional FIBmilling. Bottom

images show the experimentalphaseproleextra tedfrom the whitere tangles, omparedtothe

theoreti al stepinphasefor perfe tbulk-like spe imens[66℄.

proles t the theoreti al one. Cooper et al. [66℄ dedu ed that the ele tri ally ina tive

thi kness is stronglydependent on the dopant on entration inthe spe imens.

Charging an also have aserious ee t on ele tron holographi measurementsof

ele tro-stati potential,espe iallyinsample regions lose to the interfa e with va uum. For low

doped samples, the measurements are dependent on the strength of the ele tron beam

due to the number of ele tron-hole pairs reated in the sample whi h an be losed to

the arrier on entration in the pn jun tion. Therefore, the relative ee ts of harging

an bestronger forlowdoped samples,whi himpliesthiste hniquenot tobeused below

1.3.4.2 Ele tron Mi ros opy

S anning Ele tron Mi ros opy This te hnique is the best known and the most

widely used to study the surfa e of materials. The ontrast in the Se ondary Ele tron

(SE) image reveals information about the sample omposition and its surfa e

topogra-phy and potential. Images in SEM are a quired by s anning a highly fo used primary

ele tronbeam onthe samplesurfa e. Itsenergy variesinthe range5keV to30keV.The

SEemittedare then olle ted by a dete tor.

Theimplementationof eld-emissionele tronguns helps insu h studies. Field-Emission

SEM(FE-SEM)is parti ularlywellsuited forthe delineation of ele tri aljun tions[74℄:

itimproves resolution and redu es surfa e damages due to the radiation. It enables the

dire tobservation of ele tri allya tive dopantdistribution. It dete ts ontrastwith

sen-sitivity to dopant on entration of 10

15

to 10

20

atoms. m

−

3

, and a sub-10 nm spatial

resolution with an a ura y of

±

10% [75,76℄, as an be seen in gure 1.9. Su h resolu-tion is possible thanks to the primary ele tron beam whi h an be highlyfo used down

to very small spots. Contrary to onventional instrumentation, this te hnique has the

advantage to provide narrower probe diameters with low voltages. It generates mu h

smallerintera tion volumes, i.e. lowerpenetration depthsaround 5 nm to 50nm, hen e

givingrise to higher resolution imaging apabilities. The spe imen is usually immersed

in a relatively strong magneti eld in order to obtain high resolution of 1 nm or even

less with aprimary beam energy of 15keV.

Figure1.9: SEimage of leavedsili onandthe orresponding SEMintensityprole. Thep-doped

layers labeled A,B,C,D, E, and Fare B-dopedto2

×10

15

,

1 × 10

19

,1

×10

15

,1

×10

18

,1

×10

15

,and 1

×10

17

atoms. m

−3

respe tively. Layer Gisanintrinsi layer, andlayerH isann-dopedlayer(Sb

doped)with adoping on entrationof 5

×10

18

atoms. m

−3

when interpreting the images, keepingin mind the samplesurfa e quality inuen es the

observations: ontamination layer deposited on the surfa e ae ts indeed the intensity

signal. Moreover, dopantprolingofdevi es animplytoprepare a rossse tionthrough

the areaof interest. PreparingsampleswithFIB, orbypolishing,due totopography an

alsogenerateathi kdamagedsurfa elayerwhi hbringsdi ultiestoobserveagood

se -ondary ele tron dopant ontrast with se ondary ele trons [30,78℄. It is better toobserve

leavedsamples,orFIB hoosing onditionstoredu ethedamagedlayer[79℄. Theimage

quality is also degraded in presen e of native oxide. Hydrouori a id (HF) leaning is

often used to minimizethis problem.

Though onventional SEMis sensitive tothe doping level and itslogarithmi

depen-den e of ontrast regarding the p-type arrier on entration being well known, it is not

possible to distinguish dierently n-doped zones. S hönjahn et al. [80℄ have got rid of

this problem by using an energy-ltering dete tor. Conventional SEM remains

qualita-tive whereas reliable quantitative analysis of the dopant on entration an be derived

from the energy distribution of se ondary ele trons emittedperdoped regions with this

energy-ltering system.

SEMs an be operated with dierent kinds of lters, some of whom are presented in

hapter2,se tion2.2. Liuet al.[81℄showed thatSEMisanex ellenttoolto hara terize

the leakage me hanisms in SRAM jun tions. Kazemian et al. [82℄ measured the

poten-tial dieren e a ross pn jun tions of dierent doping levels. They observed that ltered

se ondary ele tron imagingreveals and quanties the dopant distributions. First

appli- ation of the SE ontraston sili on arbidewas realized,showing the possibilitiesof this

te hnique to quantitatively delineate the ele tri al jun tion on both hetero- and

homo-jun tions [83,84℄. Further investigations are still in progress to establish a quantitative

onversionofthis ontrastintothelo al arriersdensityandthen,intothedopantprole.

The growing use of se ondary ele tron imagingto map dopant distributions has

mo-tivated investigationof the me hanisms that give rise to dopant ontrast, a point whi h

will be detailedinse tion 5.2.

S anning Low-EnergyEle tronMi ros ope ASEM an easilybe onvertedintoa

SLEEM. These a elerationvoltagesin the olumnare lowered, usually below50 eV, by

using a retarding-eldopti al elementsu h as a athode lens. The sample isbiased at a

high negativepotential inorder toretard the primaryele trons beforetheir impa t[85℄.

This a hieves nearly onstant spatialresolutionthrough the energy s ale. A goodreview

on this te hnique is given by Müllerováet al.[86℄.

SLEEM is a valuable addition to the standard SEM be ause it is sensitive to stru ture

and orientation in rystalline materials. Moreover, sin e the in ident ele trons diuse

intoa smallerintera tion volume, itworks atlower samplingdepth and isable to image

surfa es with high sensitivity, as an be seen in gures 1.10 a) and b). Regarding the

lateral resolution, nearly the same as the SEM one an be a hieved with an ele tron

beamat1keV and aretardingeld opti s. Resolutionofabout 4nm at100 eVwas thus

obtained experimentally insu h onditions.

Figure1.10: SLEEM images: a)patterned80 nmPt layeronSiandb)P

+

typedopedareaonNtype

Si(111)for several ele tron energies [87℄.

tion of adsorbed hydro arbons under ele tron beam impa t. Low primary energy beam,

in the range 50 eV to 100 eV, orresponds for the se ondary ele trons to the minimum

of the inelasti mean free path. This implies not taking into a ount the photoemitted

ele trons and the fast ba ks attered ele trons [88℄. The role of the dierent types of

se ondary ele tron SE

2

and SE

3

inLowVoltageSEM has been studiedby Cazaux [89℄.

1.3.5 Photoemission-based mi ros opy

Photoemission ele tron mi ros opy is a powerful surfa e-sensitive te hnique suitablefor

imagingofdoping-indu ed ontrastinsemi ondu tors. Twokindsofimagingmethodsare

presented: the mi rospe tros opy, or S anning PhotoEle tron Mi ros opy (SPEM) and

thespe tromi ros opy orPhotoEle tronEmissionMi ros opy(PEEM). Thesete hniques

are inreality spatially resolved photoele tron spe tros opy. The PEEM te hnique being

the heart of this thesis, a more omplete overview will be realized in hapter 2 se tion

2.1.2.

SPEM The SPEM requires a well fo used photon beam whi h s ans the sample

sur-fa easdepi tedingure1.11a). Thephotoemittedele trons are olle ted byadete tor.

By s anning the sample with the photon beam, a 2D image of this surfa e is a quired.

One an obtain a hemi al map but this pro edure an ause artifa ts when following

dynami alpro esses [90℄.

SPEM spatial resolution is only limited by the photon spot size. In laboratory

on-ditions, the resolution is not better than several tens of mi rons due to the size of the

Figure1.11: Dieren ebetweena)SPEM andb)PEEMte hniques.

Rea hing lateralresolution betterthan50nmremainsdi ultdue tosignalandworking

distan e onstraints[93℄.

Figure1.12: a)S annedSi 2pSPEM intensitymaps showing energy dependent ontrasta rossa

sili onp-stripe, atele tron energies of b)397.22 eV, )396.98 eV,andd)396.82 eV.Field of view

(FoV)is(6.4

±

12.8)

µ

m [94℄.

PEEM In ontrast toSPEM, PEEM does not use a s anned fo used probe beam but

thesampleisuniformlyilluminated,generallyredu ingbeamindu eddamagesondeli ate

surfa es. The data a quisitionis full-eld, as an be seen in gure 1.11 b). The surfa e

magnied image anbeobserved dire tlyandinreal-timeontoauores ents reen. This

te hniquegivesthe ele troni and hemi alsurfa epropertieswithspatialresolutiononly

limited by the hromati aberrations of the obje tive lens. It varies between 10 nm to

30 nm when onsidering SE imaging, these parti les having thus a kineti energy lose

to zero [92,95,96℄. Spatial resolution down to 5 nm has been measured on biologi al

spe imen with a PEEM aberration orre ted in syn hrotron environment [97℄. PEEM

an perform real-timeimaging,and iswell adapted for real time observations. The time

resolution an be de reased down to 180 ps using a PEEM ombined with an imaging

retarding eld analyser [98℄.

The advantages of full-eld mi ros opy with PEEM versus those of s anning

artifa ts if the sample is not at or ondu tive. Photoemission ele tron mi ros opy is a

surfa e-sensitive probeprovidingdopant ontrast in imaging.

Figure1.13: Top: sele tedXPEEM imagesfrom theenergy lteredimage seriesa rossthe Si2p ore

level onsample N

+

/P. Bottom: lo al Si2p orelevel spe traandbest least squarests: the Si

4+

omponent duetothenative SiO 2

oxide isinlightgray, the sub-oxide omponentsgrayandthe Si

0

substrate omponentbla k. FoVis25

µ

m [101℄.

The ombinationof anenergylter andanele tron mi ros opeenablestoobtainfull

spe tralinformationavailableto hara terizeindetailthe hemi alandele tri al

proper-tiesoftheimagedsurfa e. Imagingthe ontrastvariationsa rossasili onpnjun tion an

be a hieved in SPEM or inPEEM. Figures 1.12 and 1.13 respe tively show ontrast on

ore-levelimagesof lateralsili ondopedpn-jun tion, thus, variationinthe energy bands

a ross adevi e an be imaged dire tly. Core-level imagingwith a sub-mi ron resolution

is possible, and less than 500 nm have been rea hed with an X-ray laboratory sour e.

Lateral resolution of 100 nm has been rea hed by Bailly et al. [102℄ using syn hrotron

radiation.

Combined with anele tron analyser, the XPEEM an indire tly determine dopant

on- entrations fromband lineup due to variationsof the Fermi level. Ballarotto et al. have

observed that dieren es in relative PEEM intensities show a systemati variation with

p-type dopant on entration in the range of 1

×10

17

atoms. m

−3

to 2

×10

20

atoms. m

−3

,

asseen ingure 1.14 a)[103℄. Figure1.14 b) presents the intensity on the PEEM image

in rease with doping on entration, at a rate of approximately 2 per de ade, in good

agreement with al ulations basedon photoemissionfromthe valen eband fora photon

energies up to 0.18 eV above threshold.

1.3.6 Summary

The visualisation of 2D dopant mapping is still of importan e nowadays, as the study

of pn jun tion for devi es at dierent s ales. A non exhaustive list of the most used

Figure 1.14: a) PEEMimageof lateral array ofpnjun tions. The doping on entrationsare given in

the image. b) Cal ulatedthreshold photoyields fromSi(001)versus doping level atthe sample surfa e.

The reddiamonds showthe measuredrelative values ofthe PEEMintensities.

δ

E orrespondstothe positionofthe pinningstatesrelative tothe surfa e valen eband[103℄.

A omparison of dierent dopant proling te hniques using identi al samples has

been done by some laboratories [16,17℄. They provide a state-of-the-art preview of the

te hniques. Table 1.2 isextra ted fromreferen e[17℄, resuming well the lastpart of this

hapter. Notethat the improvementssin e thesepubli ations insensitivity orresolution

Ima ging of sili on pn jun tions (nm)

cm

−3

resolution SCM 10 10

15

-10

20

Power C Limited Un ertaintiesatjun tions, poorquanti ationpro edure

SSHM 5 NA Power C No Noquanti ationpro edure

STM-atom ounting atomi 10

18

-10

20

Linear D Yes OnlyGaAs,notonSi

STM-STS/CITS atomi NA Log. C Limited

Onlyjun tiondelineationand type(norp)identi ation

STM-STP 10 NA Limited C Limited Onlyjun tiondelineation

KPM 100 10

15

-10

20

Limited C Limited

Poorquanti ationpro edure, stray-eldslimittheresolution

SSRM 20 10

15

-10

20

Linear C Yes Availabilitydiamondprobes

Chemi alet h 1 10

17

-10

20

Limited C Limited Di ulttoquantify,

+AFM/STM poorreprodu ibility

ImagingSIMS 100 NA Linear D Yes Sensitivitylimitedbytargetvolume

2D-SIMS 30-50 10

16

-10

21

Linear D Yes Spe ialstru turesrequired

2D-TomographySIMS 50 NA Linear D Yes

Spe ialstru turesrequired. Complexsamplepreparation

LateralSIMS 5-10 Done Linear D Yes

Onlythelateraldose Distributionismeasured

2D-SRP 100 10

15

-10

21

Linear C Yes Spe ialstru turesrequired

Chemi alet h

20 10

17

-10

21

- C Limited Onlyqualitative +SEM/STM FE-SEM 10-20 4.10

16

-10

20

Limited Limited

Robustmodelforquanti ation isnotavailable Ele tronHolography 1-10 10

17

-10

20

Limited C Limited Inversemodeling NA NA Inversemodeling C Yes

Resolutionanda ura yareunknown, with

C − V

te hniques long al ulationtimes