Monc As I Luque 2022 Diffusion

Structure–property relationship in (La,Sr)MnO3
memristive devices
Carlos Moncasi Luque
To cite this version:

Carlos Moncasi Luque. Structure–property relationship in (La,Sr)MnO3 memristive devices. Cristal-
lography. Université Grenoble Alpes [2020-..], 2022. English. �NNT : 2022GRALI048�. �tel-04621403�
HAL Id: tel-04621403

https://theses.hal.science/tel-04621403
Submitted on 24 Jun 2024
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
THÈSE
Pour obtenir le grade de
DOCTEUR DE L’UNIVERSITÉ GRENOBLE ALPES

École doctorale : I-MEP2 - Ingénierie - Matériaux, Mécanique, Environnement, Energétique,
Procédés, Production
Spécialité : 2MGE : Matériaux, Mécanique, Génie civil, Electrochimie
Unité de recherche : Laboratoire des matériaux et du génie physique (LMGP)
Relation entre la structure et les propriétés dans les mémoires

résistives à base de films (La,Sr)MnO3
Structure-property relationship in (La,Sr)MnO3 memristive devices
Présentée par :
Carlos MONCASI LUQUE
Direction de thèse :
Mónica BURRIEL LOPEZ Directrice de thèse
Chargé de Recherche HDR, CNRS délégation Alpes
Maria del Carmen JIMENEZ AREVALO Co-encadrante de
Ingénieur de recherche, CNRS thèse
Rapporteurs :
Olivier THOMAS
PROFESSEUR DES UNIVERSITES, Aix-Marseille Université
Guilhem DEZANNEAU
DIRECTEUR DE RECHERCHE, CNRS délégation Ile-de-France Sud
Thèse soutenue publiquement le 27 juin 2022, devant le jury composé de :

Liliana BUDA-PREJBEANU Présidente du jury et
PROFESSEUR DES UNIVERSITES, Université Grenoble Alpes Examinatrice
Olivier THOMAS Rapporteur
PROFESSEUR DES UNIVERSITES, Aix-Marseille Université
Guilhem DEZANNEAU Rapporteur
DIRECTEUR DE RECHERCHE, CNRS délégation Ile-de-France Sud
Gabriel MOLAS Examinateur
DOCTEUR EN SCIENCES, Weebit Nano Ltd.
Mónica BURRIEL LOPEZ Directrice de thèse
CHARGE DE RECHERCHE HDR, CNRS délégation Alpes
Abstract
Structure–property relationship in (La,Sr)MnO3 memristive devices
This PhD thesis focuses on the study of the relationship between (La,Sr)MnO3 (LSM) thin film’s
microstructure and their functional properties in memristive devices. Valence Change Memories
(VCMs), which rely on resistive switching phenomena, are exciting candidates for the next generation
of non-volatile memories as they offer fast writing speed, high storage density and low power
consumption. These memories store data in the form of a tunable resistance, which can be varied by an
electrical stimuli. Typically, oxygen drift within the system provokes a redox reaction, which triggers
the change of resistance. However, the VCMs performance is strongly related to the elements that
compose the memory, i.e. active material, electrode and their interfaces.
In this thesis, the memristive response of La1-xSrxMnO3 (LSM) perovskite oxide was studied in detail.
LSM presents flexible oxygen stoichiometry accommodated through the change of Mn valence, which
in turn modifies its electrical conduction properties. LSM thin films were successfully grown by pulsed-
injection metal-organic chemical vapor deposition on three substrates to induce different
microstructures: epitaxial films with low and high concentration of extended defects were grown on
SrTiO3 (STO) and LaAlO3 (LAO) single crystals, respectively, and polycrystalline films with a large
density of grain boundaries were grown on Si3N4 substrates. The memristive response was studied in
micro-fabricated Ti/LSM\Pt planar devices. The formation of an oxygen deficient TiOx interlayer at the
Ti/LSM interface in pristine devices was observed by STEM. Moreover, X-Ray Absorption Near-Edge
Spectroscopy measurements in a pristine device indicate a strong reduction of the Mn under the Ti
electrode (Mn valence ca. +2.6) while the rest of the film was oxidized (valence ca. +3.5). The reduction
of LSM suggests that, during the fabrication process the TiOx is formed at the interface by scavenging
oxygen from LSM.
First, the memristive performance was measured in the epitaxial samples, for which it was proven that
the performance of the nanoionic devices can be boosted by microstructure engineering. Bipolar
interface-type resistive switching was observed, where the transition to the high resistance state (HRS)
or RESET was assigned to the drift of oxygen from LSM towards the Ti electrode (oxidizing it), while
the reverse redox reaction leads to a low resistance state (LRS) via a SET process. The gradual
oxidation/reduction can account for the multilevel resistance states measured in these devices. While
the RESET takes place under similar conditions in LSM/LAO and LSM/STO, the SET process greatly
depends on the microstructure of the LSM film. The high crystal quality of the LSM/STO epitaxial films
seems to hinder the reincorporation of oxygen into LSM during the SET transition, and thus, asymmetric
voltage operation is required, reaching a HRS/LRS ratio up to 7. On the other hand, in LSM/LAO
i
devices where the extended defects act as fast migration paths to reincorporate the oxygen into LSM,
symmetric voltage operation is possible up to ±20 V with HRS/LRS up to 23.
Finally, polycrystalline Ti/LSM\Pt devices built on LSM/Si3N4 show similar behavior to that of the
LSM/LAO devices, where the multilevel capacities are assigned to fast oxygen drift along the grain
boundaries in both directions, i.e. during RESET and SET operation. Hence, these devices present larger
operation windows, even at smaller writing voltage, i.e. HRS/LRS>10 for Vwrite=±14 V. The results
obtained for polycrystalline devices are very exciting for memory applications as they are compatible
with the semiconductor industry. The miniaturization of these devices is expected to lead to larger
operating window and multilevel storage, while operating at lower voltages and thus, reducing power
consumption.
ii
Contents
Abstract ........................................................................................................................................ i
Contents ...................................................................................................................................... ⅲ
List of Figures ..................................................................................................................................... ⅶ
List of Tables .................................................................................................................................... ⅹⅲ
Nomenclature ..................................................................................................................................... ⅹⅴ
Acknowledgments............................................................................................................................... ⅹⅸ
Résumé de la thèse en français ........................................................................................................ ⅹⅹⅲ
Chapter 1:
Introduction ....................................................................................................................................... 1
Motivation ................................................................................................................................. 1
Resistive switching phenomena ............................................................................................... 2
Device phenomenology and first observations .............................................................. 2
Classification ................................................................................................................. 3
Strontium-substituted lanthanum manganite ........................................................................ 6
Bulk: crystal structure, electronic properties and phase diagram .................................. 6
LSM thin films: electrical properties ........................................................................... 10
Oxygen mass transport ................................................................................................ 13
Memristive response of La1-xSrxMnO3 .................................................................................. 15
Role of the electrodes .................................................................................................. 19
Scope and organization of the thesis ..................................................................................... 22
Chapter 2:
Experimental procedure ..................................................................................................................... 23
Pulsed Injection Metal Organic Chemical Vapor Deposition ............................................ 23
Device micro-fabrication........................................................................................................ 25
Device electrical characterization ......................................................................................... 28
iii
Chapter 3:
Optimization of the growth conditions of La1-xSrxMnO3-δ thin films.............................................. 33
Geometrical optimization of the reactor............................................................................... 33
Deposition optimization of LSM thin films with different composition ............................ 36
Correlation between the precursor solution and the film composition ........................ 36
Effect of the LSM composition on the structural properties........................................ 41
Morphology and microstructure .................................................................................. 48
Raman of the LSM20 and LSM50 films ..................................................................... 51
Effect of the composition on the LSM films electrical properties ............................... 53
Conclusions ............................................................................................................................. 55
Chapter 4:
Effect of the nanostructure in La0.8Sr0.2MnO3-δ thin films in memristive devices ......................... 57
Epitaxial LSM20 as memristive material ............................................................................. 58
Structural characterization of epitaxial films ............................................................... 58
Electrical characterization of epitaxial LSM20 devices .............................................. 63
Discussion: oxygen migration modelling .................................................................... 73
Summary on epitaxial LSM20 as memristive material ............................................... 75
Epitaxial and polycrystalline texture in thicker LSM20 films ........................................... 76
Description of the samples........................................................................................... 77
Comparison between epitaxial and polycrystalline LSM20 memristive devices ........ 78
Cross-comparison: the role of thickness ............................................................................... 99
Conclusions ........................................................................................................................... 103
Chapter 5:
Memristive response of epitaxial La0.5Sr0.5MnO3 - based devices ................................................. 107
Growth of epitaxial La0.5Sr0.5MnO3-δ films on LAO .......................................................... 108
Structural characterization of epitaxial LSM50 ......................................................... 108
Electrical characterization of LSM50/LAO-based devices ............................................... 111
Initial resistance (IR) of the LSM50/LAO devices .................................................... 112
Memristive response of the LSM50/LAO devices .................................................... 113
Reproducibility and stability of the LSM50/LAO devices ........................................ 116
Scaling of LSM50/LAO devices ............................................................................... 119
Role of the strontium content on LSM/LAO-based devices ............................................. 121
Conclusions ........................................................................................................................... 124
iv
Contents
Chapter 6:
Effect of oxygen stoichiometry in LSM films structural and electrical properties ..................... 127
Sensing oxygen by Raman spectroscopy ............................................................................ 128
Raman spectroscopy on LSM20 annealed samples ................................................... 128
Raman spectroscopy on LSM20 and LSM50 annealed samples ............................... 130
In-situ Raman during annealing of LSM20 sample ................................................... 131
PAS: Probing point defects .................................................................................................. 133
Resistivity as function of temperature and pO2 ................................................................. 136
XANES: assessing oxygen content through Mn oxidation state ....................................... 139
Conclusions ........................................................................................................................... 145
Chapter 7:
Concluding summary and prospects ............................................................................................... 147
References ................................................................................................................................... 153
Annex A:
Techniques and equipment ............................................................................................................... 169
A.1. High energy radiation-based characterization techniques .................................................... 169
A.1.1. Rutherford Backscattering Spectrometry and Nuclear Reaction Analysis .................. 169
A.1.2. X-ray Absorption Spectroscopy .................................................................................. 171
A.2. Techniques for thin film characterization ............................................................................... 173
A.2.1. X-ray diffraction .......................................................................................................... 173
A.2.2.Transmission Electron Microscopy .............................................................................. 173
A.2.3. Scanning electron microscopy ..................................................................................... 174
A.2.4. Electron Probe Micro Analysis and Wavelength dispersive spectroscopy .................. 174
A.2.5. Atomic Force Microscopy ........................................................................................... 174
A.2.6. Raman Spectroscopy ................................................................................................... 175
A.2.7. Electrical Characterization of thin films and devices .................................................. 175
v
List of Figures
Figure 1.1. a) Sketch of a RRAM cell with MIM structure. Typical current-voltage characteristics of b)
unipolar and c) bipolar RRAM devices ................................................................................................... 3
Figure 1.2. Mechanisms responsible of resistive switching phenomena and their classification based on
the key element in the system (blue) and working polarity (green). ....................................................... 4
Figure 1.3. Schematic of the a) filamentary and b) interface-type models of resistive switching.......... 5
Figure 1.4. Representation of the cubic perovskite structure with chemical formula ABO3.................. 7
Figure 1.5. a) Electron occupancy of the Mn 3d levels for different oxidation states. b) Sketch of the
Double Exchange mechanism. ................................................................................................................ 8
Figure 1.6. Bulk electrical resistivity as a function of the temperature in a) LSM with different
composition, x, and b) LSM with x=0.33 and different oxygen content, δ. ............................................ 8
Figure 1.7. Effect of Sr substitution on the oxygen stoichiometry in LSM at 1073 K. .......................... 9
Figure 1.8. Phase diagram of La1-xSrxMnO3 in the whole range of Sr substitution, x. ......................... 10
Figure 1.9. a) HRTEM Fourier filtered image of LSM20 grain boundary formed by an array of
periodically arranged dislocation. b) Colored enlargement of the red square in a) where the dislocation
are marked as “⊤”.................................................................................................................................. 11
Figure 1.10. Effect of the epitaxial strain in electrical properties of La0.7Sr0.3MnO3 films. ................. 12
Figure 1.11. Oxygen mass transport in LSM20 and LSM50.. .............................................................. 14
Figure 1.12. Schematic of the band diagrams and charge distribution of a p-type semiconductor ...... 20
Figure 1.13. Temperature dependence of the free energy a) formation of some noble and transition metal
oxides, and b) formation of sub-oxides in TiO2. ................................................................................... 21
Figure 2.1. Schematic of the PI-MOCVD JIPELEC reactor and the main steps involved in the growth
of films by CVD. ................................................................................................................................... 24
Figure 2.2. Workflow of the micro-fabrication of memristive device using a positive photo-resin .... 26
Figure 2.3. a) Lithography mask used for the fabrication of memristive devices displaying the
arrangement of the electrodes on a sample of 10×10 mm2. b) Sketch of a memristive device. ............ 27
Figure 2.4. a) Schematic of a sweep voltage cycle. b) Scheme of the initialization process. .............. 29
Figure 2.5. a) Scheme of contact array for TLM and b) cross-section ................................................. 30
vii
Figure 3.1. Depositions on 4 inches silicon wafers for geometrical optimization of the reactor ......... 35
Figure 3.2. Correlation of the a), b) strontium substitution; and c), d) A/B ratio of the LSM film with
the atomic ratio of the precursor solution. ............................................................................................. 38
Figure 3.3. XRD patterns of LSM films with different composition grown on a) Si (Bragg-Brentano
geometry) and b) Si3N4 (grazing incidence XRD).. .............................................................................. 42
Figure 3.4. XRD patterns in Bragg-Brentano configuration of highly-oriented La1-xSrxMnO3-δ (LSM)
films with different composition grown on a) MgO and b) LaAlO3 substrates.. .................................. 43
Figure 3.5. Pseudo-cubic cell parameter as function of Sr substitution of the LSM films grown on a) Si
and Si3N4, and b) MgO and LAO substrates. ........................................................................................ 44
Figure 3.6. XRD pattern of La0.8Sr0.2MnO3 (LSM20) films grown using solution 4 at different
temperatures on a) Si, b) MgO, c) Si3N4, and d) LAO substrates. ........................................................ 46
Figure 3.7. Pseudo-cubic lattice parameter of the LSM20 films grown on different substrate as function
of the temperature of deposition. ........................................................................................................... 47
Figure 3.8. SEM surface morphology observed by secondary electrons for La 1-xSrxMnO3 (LSM) thin
films with different stoichiometry grown on Si substrates. ................................................................... 49
Figure 3.9. Surface morphology observed by secondary electrons in SEM for La1-xSrxMnO3 (LSM) thin
films with different stoichiometry grown on MgO substrates. .............................................................. 50
Figure 3.10. Morphology of the LSM20 and LSM50 films grown on Si, LAO and MgO substrates.. 51
Figure 3.11. Raman spectra of LSM20 and LSM50 films grown on a) MgO, b) LAO and c) Si 3N4
substrates.. ............................................................................................................................................. 52
Figure 3.12. Electrical resistivity as function of Sr substitution, x, in La1-xSrxMnO3 films grown at
700 °C on MgO, LAO and Si3N4 substrates. ......................................................................................... 54
Figure 4.1. Structural characterization of the LSM20 films. XRD patterns of LSM20 films grown on
a) STO and b) LAO... ............................................................................................................................ 59
Figure 4.2. TEM cross-section characterization of LSM20/STO and LSM20/LAO films.. ................ 60
Figure 4.3. STEM-EDS chemical mapping and profile of pristine Ti/LSM20/STO and Ti/LSM20/LAO
stacks that suggest the presence of a TiOx interlayer. ........................................................................... 62
Figure 4.4. I(V) characteristics of symmetrical devices fabricated on a) LSM20/STO and b)
LSM20/LAO. ........................................................................................................................................ 64
Figure 4.5. Initial resistance (IR) of the Ti/LSM20\|Pt pristine memristive devices. ........................... 65
Figure 4.6. a), b) Resistance-voltage and c), d) current-voltage characteristics of the initialisation
process carried out at different voltage amplitudes in Ti/LSM20\Pt standard devices on a), c)
LSM20/STO and b), d) LSM20/LAO samples. .................................................................................... 66
viii
List of Figures
Figure 4.7. Resistance-voltage (-electric field) characteristics of initialized Ti/LSM20\Pt standard

devices. .................................................................................................................................................. 67
Figure 4.8. Evolution of the HRS and LRS resistance states measured after each cycle (bottom axis)
and as a function of writing voltage (top axis) for standard devices ..................................................... 68
Figure 4.9. HRS and LRS obtained at Vwrite=±20 V (solid data points) and ±10 V (open data points) at
different times for LSM20/LAO. .......................................................................................................... 70
Figure 4.10. Pulse operation of the Pt/Ti/LSM20\Pt memristive devices using ±10 V/60 ms writing
voltage pulses.. ...................................................................................................................................... 70
Figure 4.11. a) HRS and LRS evolution with time of Pt/Ti/LSM20\Pt memristive devices programmed
at ±20 V on LSM20/STO and LSM20/LAO. b) Multi-states resistance evolution with time. .............. 71
Figure 4.12. Area dependence resistance of the HRS (blue shade, top) and LRS (red shade, bottom) at
different writing voltages for Ti/LSM20\Pt devices in LSM20/LAO ................................................... 72
Figure 4.13. Sketch of the resistive switching phenomenology in epitaxial devices. .......................... 74
Figure 4.14. XRR scan of a LSM20-8/ Si3N4 sample and the fit result. The fitting indicates a thickness
of 7.90 nm. ............................................................................................................................................ 78
Figure 4.15. SEM images and XRD diffractograms for LSM20-8 films grown on a) STO, b) LAO, and
c) Si3N4. ................................................................................................................................................. 80
Figure 4.16. TEM cross-section of LSM20-8/Si3N4 and the stack Pt/LSM20-8/Si3N4. ....................... 81
Figure 4.17. Electrical resistivity of LSM20 thin films of different thicknes. ..................................... 82
Figure 4.18. Electrical device characterization of LSM20 thin films with different thickness. ........... 83
Figure 4.19. R(V) curves at increasing voltage amplitude for Ti/LSM20-8\Pt standard devices ......... 84
Figure 4.20. Evolution of HRS, LRS and HRS/LRS with cycles and writing voltage amplitude........ 85
Figure 4.21. XRR scan of a sample of LSM20-21/ STO and the fit result........................................... 89
Figure 4.22. SEM images and XRD patterns for LSM20-21 films. ..................................................... 91
Figure 4.23. HRTEM cross-section images of a), b) LSM20-21/STO, c), d) LSM20-21/LAO, and e),
f) LSM20-21/Si3N4.. .............................................................................................................................. 92
Figure 4.24. Reciprocal space maps on LSM20-21/LAO around a) symmetrical LSM peak (024) and
b) asymmetrical LSM peak (042). ......................................................................................................... 93
Figure 4.25. R(V) characteristics of Ti/LSM20-21\Pt devices for films grown on a) STO, b) LAO, and
c) Si3N4 substrates. ................................................................................................................................ 95
Figure 4.26. Proof of damage in a Ti/LSM20\Pt device....................................................................... 96
Figure 4.27. Asymmetric voltage operation of standard LSM20-21/STO-based memristive devices. 97
Figure 4.28. Characterization summary of LSM20 thin films with different thickness and grown on
different substrates. ............................................................................................................................. 100
Figure 4.29. Resistance-voltage characteristics of the Ti/LSM20\Pt devices in films with different
thickness grown on a) STO, b) LAO, and c) Si3N4 substrates. ........................................................... 101
ix
Figure 5.1. XRR pattern of the LSM50/LAO films. .......................................................................... 109

Figure 5.2. a) XRD patterns of the LSM50/LAO film. b) Schematic of the epitaxial growth of LSM50
on LAO. ............................................................................................................................................... 110
Figure 5.3. Surface morphology of the LSM50/LAO films observed by a) SEM and b) AFM. ........ 110
Figure 5.4. TEM cross-section images of 21 nm thick LSM50/LAO film. and selective area electron
diffraction at the LSM50/LAO interface ............................................................................................. 111
Figure 5.5. Initial resistance (IR) values of the Ti/LSM50\Pt pristine memristive devices. .............. 113
Figure 5.6. I(V) curves of symmetrical M/LSM50\M devices ........................................................... 114
Figure 5.7. a), b) I(V) and c), d) R(V) characteristics of the initialization procedure of a Ti/LSM50\Pt
memristive device................................................................................................................................ 115
Figure 5.8. Evolution of the resistance states reached after each cycle and as function of the writing
voltage amplitude in a standard Ti/LSM50/LAO\Pt memristive device. ............................................ 116
Figure 5.9. Evolution of the resistance states during the initialization and during multiple cycling.. 117
Figure 5.10. Evolution of HRS and LRS values with time for Ti/LSM50/LAO\Pt devices programmed
at Vamp=10 V. ....................................................................................................................................... 119
Figure 5.11. Initialization curves of Ti/LSM50\Pt memristive devices with different electrode area 120
Figure 5.12. HRS and LRS area scaling of Ti/LSM50\Pt memristive devices. ................................. 121
Figure 5.13. Comparison of the memristive response in standard (A=2002 µm2) Ti/LSM\Pt devices built
on a), c) LSM20 (7 nm)/LAO and b), d) LSM50 (21 nm)/LAO. ........................................................ 123
Figure 6.1. Comparison of LSM20/LAO pristine and samples annealed in different atmospheres. .. 130
Figure 6.2. Raman spectra of a) LSM20 and b) LSM50 grown at 700 °C on Si3N4 and LAO substrates
annealed at 500 °C for 1 h in Ar and O2 atmosphere. ......................................................................... 131
Figure 6.3. In situ Raman spectroscopy on LSM20 (100 nm)/LAO films. ........................................ 132
Figure 6.4. Annihilation line parameters S and W (low and high electron momentum fraction) as a
function of positron implantation energy Ep in LSM20 films. ............................................................ 134
Figure 6.5. a) Average positron lifetime and b) positron lifetime components τi, for the pristine and
thermally treated LSM20 films. .......................................................................................................... 135
Figure 6.6. Electronic conductivity of 8 nm thick LSM20 films grown on LAO and STO substrates in
O2 atmosphere as function of temperature.. ........................................................................................ 137
Figure 6.7. Electronic resistivity of the 8 nm thick a) LSM20/STO and b) LSM20/LAO films as
function of temperature and atmosphere ............................................................................................. 139
x
List of Figures
Figure 6.8. Normalized XANES spectra at the Mn K edge position for Mn references with different
oxidation state. .................................................................................................................................... 140
Figure 6.9. XANES spectra and first derivative at the Mn K-edge of LSM20/Si3N4 and LSM50/Si3N4
films annealed in oxidizing and reducing conditions. ......................................................................... 141
Figure 6.10. XANES spectra and first derivative of the regions of interest in the Ti/LSM20\Pt
memristive device on Si3N4. ................................................................................................................ 142
Figure 6.11. a) Mn K-edge position, calculated b) Mn valence and c) oxygen off-stoichiometry, δ, of
the annealed LSM20 films and the three ROI within a pristine LSM20-based memristive device. ... 143
Figure 6.12. a) Mn K-edge XANES spectra under the Ti contact for pristine devices and devices
programmed ex-situ at different resistance states. b) Comparison of the R(V) characteristics in the same
device before and after being exposed to the beam. ............................................................................ 144
Figure A.1. RBS-NRA experimental spectra of a LSM20/MgO sample, the simulated profile and the
elemental deconvolution...................................................................................................................... 170
Figure A.2. Setup for operando XANES.. ......................................................................................... 172
Figure A.3. Scheme of the two van der Pauw configurations used to measure the resistivity as function
of temperature. .................................................................................................................................... 175
xi
List of Tables
Table 1.1. Collection of relevant resistive switching studies using La1-xSrxMnO3-δ, for x=20-33 and
50 %, highlighting the type of switching and the mechanism proposed ............................................... 17
Table 3.1. Deposition parameters used for the growth of LSM after the geometrical optimization of the
reactor .................................................................................................................................................... 34
Table 3.2. Relative precursor concentration in the solutions prepared to optimize the composition. .. 37
Table 3.3. Atomic ratio measured by EDS, WDS and RBS-NRA on the LSM films with composition
close to stoichiometric ABO3.. .............................................................................................................. 39
Table 3.4. LSM formula derived from RBS-NRA weighted average composition.............................. 41
Table 4.1. Results of Transmission line measurements (TLM) using Pt contacts. ............................... 63
Table 4.2. Description of LSM20 films grown on, STO, LAO and Si3N4 substrates ........................... 77
Table 4.3. Operating conditions of the optimized memristive switching devices. ............................... 99
Table 5.1. Parameters obtained from the TLM on the LSM50/LAO film. ......................................... 112
Table 5.2. Resistance value of the resistance states and HRS/LRS ratio in LSM50-based devices ... 117
Table 5.3. Main structural properties in LSM20 and LSM50, and operating conditions and performance
of the Ti/LSM\Pt memristive devices for x=0.2 and 0.5, studied in the thesis.................................... 122
Table 6.1. Summary of the main samples discussed in the Chapter ................................................... 128
xiii
Nomenclature
Acronyms and Abbreviations
8W Eight-wise, referred to the cycling sense of a device

AFM Atomic Force Microscopy
BRS Bipolar Resistive Switching
C8W Counter-eight-wise, referred to the cycling sense of a device
C-AFM Conductive Atomic Force Microscopy
CC Current Compliance
CCW Counter-clockwise, referred to the cycling sense of a device
CMOS Complementary Metal Oxide Semiconductor
CPU Central Processing Unit
CVD Chemical Vapor Deposition
CW Clockwise, referred to the cycling sense of a device
DB-VEPAS Doppler broadening variable energy positron annihilation spectroscopy
DFT Density Functional Theory
DRAM Dynamic Random Access Memory
ECM(s) Electrochemical Memory(ies)
EDS Energy Dispersive Spectroscopy
EIS Electrochemical Impedance Spectroscopy
ELBE Electron Linac for beams with high Brilliance and low Emittance (HZDR, Germany)
EPIR Electric-Pulse-Induced Resistance
EPMA Electron Probe MicroAnalysis
ESRF European Synchrotron Radiation Facility (Grenoble, France)
EXAFS Extended X-ray Absorption Fine Structure
FEM Finite Element Method
FT Fourier Transform
GB(s) Grain boundary(ies)
GIXRD Grazing-incidence X-Ray Diffraction
HAXPES Hard X-ray Photoelectron Spectroscopy
HRS High Resistance State
HRTEM High Resolution Transmission Electron Microscopy
xv
HZDR Helmholtz-Zentrum Dresden - Rossendorf (Germany)

ICDD-PDF International Centre for Diffraction Data - Powder Diffraction File
IEDP-SIMS Isotope Exchange Depth Profile - Secondary Ion Mass Spectrometry
IPA IsoPropyl Alcohol
IR Initial Resistance
JT Jahn-Teller
LAO Lanthanum aluminate, LaAlO3
LMO Lanthanum manganite, undoped lanthanum manganite, LaMnO3-δ
LRS Low Resistance State
LSM Strontium substituted lanthanum manganite, La1-xSrxMnO3-δ
LSM20 La0.8Sr0.2MnO3
MePS Mono-energetic Positron Source beamline
MIEC Mixed Ionic Electronic Conductor
MIM Metal-Insulator-Metal
MIT Metal-Insulator Transition
MOS Metal Oxide Semiconductor
MU Memory Unit
NDR Negative Differential Resistance
NRA Nuclear Reaction Analysis
NVM(s) Non-Volatile Memory(ies)
PAS Positron Annihilation Spectroscopy
PCMO Praseodymium calcium manganite, Pr1-xCaxMnO3
PCMO30 Pr0.7Ca0.3MnO3
PiMOCVD Pulsed-injection Metal-Organic Chemical Vapor Deposition
PTA Plateform Technologique Amont
RBS Rutherford Backscattering Spectrometry
ReRAM(s) Redox Random Access Memory(ies)
ROI Region of Interest
RRAM(s) Resistive Random Access Memory(ies)
RS Resistive Switching
xvi
Nomenclature
RSM Reciprocal Space Mapping

RT Room Temperature
SAED Selective Area Electron Diffraction
SEM Scanning Electron Microscopy
SOFC Solid Oxide Fuel Cell
SPONSOR Slow positron System of Rossendorf (Germany)
STEM Scanning Transmission Electron Microscopy
STO Strontium titanate, SrTiO3
TCM(s) Thermochemical Memory(ies)
TEM Transmission Electron Microscopy
TF Tolerance Factor
TLM Transmission Line Measurement
URS Unipolar Resistive Switching
VCM(s) Valence Change Memory(ies)
VEPALS Variable Energy Positron Annihilation Lifetime Spectroscopy
WDS Wavelength Dispersive Spectroscopy
XANES X-ray Absorption Near Edge Structure
XRD X-Ray Diffraction
XRR X-Ray Reflectometry
xvii
Acknowledgments
Like river flows, doing a Ph.D. was the most natural way right after university to keep learning
while merging for the first time into the professional world. Even if the decision was not trivial, I was
looking forward to new experiences that could break the comfort shell I was wearing at the time. So,
this took me to a foreign country whose language I did not know. Indeed, once calm waters suddenly
turned into troubled waters, which required balance and effort, before I learned that they were navigable
too. The Ph.D. boat met plenty of people who joined the journey, some of them without knowing it, and,
in one way or another, for better or for worse, shaped it into what it became. After many turns and water
risings, this trip has finally reached the sea. I am not sure this is the sea I expected when I first jumped
into the boat, if I was expecting one at all. Yet I am glad of the journey on my back and excited about
the new horizons. Open sea, settling on the shore or back to land? I am not sure which path I will take,
but I do know with certainty that each of them hides beauty and mystery, and they will provide new
opportunities to make this great journey a unique and worthwhile living experience.
We cannot reach any higher

if we can’t deal with ordinary love.
– U2
In this increasingly competitive world, kindness and gratitude are endangered species that
sometimes are faked and used as a tool for selfish interests. Today, I would like to bring some hope and
share some of this threatened gratitude with all the people that helped me through the years of Ph.D. in
Grenoble; either in the laboratory or outside - and for those who did both…my most sincere thanks.
If you want to go fast, go alone.

If you want to go far, go together
– African proverb
First of all, I would like to express my deepest gratitude to my supervisors for giving me the
opportunity of the Ph.D. and to evolve as scientist in a foreign country while doing science in my mother
tongue, gracias! Thank you for your guidance, the scientific discussion with your respective chemical
and physical approach of the subject. Mónica, thank you for availability, informality, and yet correctness
and rigor. Carmen, thank you for your relentless work in all scales that makes the lab move and
especially, thanks for seeing beyond what is visible.
Next, I would like to thank the members of the jury for having accepted evaluating my work and
for your diverse input before, during and after the defense. Stop learning is not an option.
xix
This work would have not been possible without financial support. Therefore, I would like to
acknowledge the ANR - PCRI Mangaswitch project and the other members involved: Prof. Dr. Regina
Dittmann, Prof. Dr. Roger de Souza and Dr Jacqueline M. Börgers. Even though, the situation did not
allow a laboratory exchange, I am glad we shared project meetings and I got to know your personal
scientific approach.
This work was carried out in an excellent and precious lab environment. I want to acknowledge
the direction of LMGP, Franz Bruckert and Carmen Jiménez, for their efficient management of the
laboratory even when times were difficult.
I would like to thank Odette Chaix, Laetitia Rapenne, Hervé Roussel and Isabelle Gélard for their
support with the characterization techniques and also for helping on the interpretation of the results.
I would like to acknowledge the work of the technical and HR staff, in particular, Nadjib Medjaldi,
Michèle San Martin, Annie Ducher and Josiane Viboud, who make the lab run smooth but more
important, thank you for your kindness, patience and your cheerfulness throughout the days.
Nanoionics, former and current members, thank you for making me feel so comfortable since the
beginning and for teaching me the fundamentals of the work so I could become independent. We were
few members when I started and now we almost have problems to sit down on the seminar room. I hope
that, regardless the number of people, a good bonding can be established between the members to do
better research and most importantly make it bearable. Until next time… potatoes!
My most sincere gratitude goes to my LMGP colleagues, former and present. You were the boat
that allowed me to reach the sea and without you this would have not been the same. Thank you for all
the outdoors activities and for all the sorts of talks and discussions we had. I would like to start by the
people who was always there: Raquel, Clém, Klaasjan, Antalya, Dorina, João, Adrià, Mónica, Ragnar,
and Max. You already know how valuable you have been to me so I will keep it brief: thank you for
your constant support and the time spent together either hiking, having drinks, at the movie nights,
having brunch or having a coffee on a Sunday morning. Thanks to all the office mates I had, from the
loft, the “post-docs” office and the cold office, for your availability to discuss and chat when things were
getting blurry. In particular, danke, Alex, for bringing such cheerful mood into the office and the team
(most of the times), your help with research, your vermut, and also for talking clear about the worries
and problems that we face (it feels we are less alone).
José and Chiara, La Choza, you became more than colleagues. We started this adventure together
and soon we understood it was going to be nicer together. Thank you for making from an apartment
more than just walls. Thanks for the great time we had together and for all the anecdotes in and outdoors.
xx
Acknowledgments
I would also like to thank my friends back in Catalonia as they always take great care of me when
I go back home and make of the holidays a great and truly disconnection time. Gràcies.
Finally I would like to thank my family who always encourage me to pursue new adventures:
moltes gràcies pel support que m’heu demostrat sempre i per animar-me a seguir créixent, fins i tot quan
això implicava marxar encara més lluny de casa. Des de la distancia però sempre a prop.
xxi
Résumé de la thèse en français
Introduction
L'objectif de cette thèse était de contribuer à la compréhension des mécanismes physico-chimiques sous-
jacents de la commutation résistive dans les mémoires à changement de valence, et d'élucider le rôle de
la micro- et nanostructure dans la performance des dispositifs. En particulier, deux compositions
d'oxydes pérovskites La1-xSrxMnO3 (LSM), avec x=0.2 (LSM20) et x=0.5 (LSM50) ont été étudiées dans
cette thèse et les principales propriétés sont incluses dans le Chapitre 1. La procédure expérimentale et
la méthodologie sont présentées dans le Chapitre 2. Le LSM a été étudié sous forme de couche mince à
l'aide d'un réacteur MOCVD (metal organic chemical vapor deposition) pour lequel l'optimisation du
réacteur et l'étude pour atteindre la composition cible sont discutées dans le Chapitre 3. Le Chapitre 4
se concentre sur la caractérisation des couches minces de LSM20 avec différentes microstructures et
épaisseurs, et sur les performances des dispositifs memristifs fabriqués sur ces couches. Le Chapitre 5
se concentre sur les couches épitaxiales de LSM50 et rapporte pour la première fois une commutation
résistive fiable et reproductible pour ce matériau. Ce chapitre se termine par une comparaison des
dispositifs en LSM20 et LSM50. Le Chapitre 6 comprend une caractérisation complémentaire visant à
élucider le rôle de l'oxygène dans les propriétés des couches. Il comprend des mesures par spectroscopie
Raman, Positrons, ainsi que des mesures de conductivité à différentes températures et sous différentes
atmosphères. Les conclusions et les perspectives de ce travail sont présentées dans le Chapitre 7. Les
équipements utilisés sont décrits dans l'Annexe A.
Le phénomène de commutation résistive (Resistive Switching en anglais, RS) a suscité beaucoup

d'attention pour le développement d'un nouveau type de mémoires (mémoires résistives à accès
aléatoire, RRAM), car elles peuvent offrir un fonctionnement rapide, une endurance élevée, une longue
durée de rétention et une faible consommation d'énergie, tout en surmontant les limites d'échelle des
mémoires à base de silicium actuels.
La RS est exploitée dans les dispositifs dits "memristifs", qui consistent en un matériau intercalé entre
deux électrodes (Figure 1a). Lors de l'application d'un stimulus électrique, ces dispositifs peuvent
changer de façon réversible entre au moins deux états de résistance différents qui sont conservés dans
des conditions de champ nul. En général, ces dispositifs commutent entre un état de résistance élevé
(HRS) et un état de résistance faible (LRS), bien que certains d'entre eux aient montré des propriétés
multiniveaux où plusieurs états de résistance intermédiaires peuvent être programmés, augmentant ainsi
la densité de stockage par dispositif.
Les premières observations de RS datent des années 1960 sur des oxydes binaires tels qu’Al2O3, Ta2O5,
ZrO2. Après une période avec peu de recherches, la RS a refait surface dans les années 1990 avec
xxiii
l'apparition des oxydes pérovskites et de leurs propriétés exotiques, et a gagné un intérêt supplémentaire
lorsque les résultats expérimentaux ont été liés à la théorie des memristors de L. Chua.
Dans cette thèse, nous nous sommes concentrés sur les mémoires à changement de valence (VCMs), où
le changement de résistance est le résultat simultané de la migration d’ions oxygène et du changement
local de valence du sous-réseau de cations sous l'application d'un champ électrique. De plus, les espèces
actives impliquées dans la commutation dans les VCM peuvent être accumulées soit dans une région
localisée, correspondant à une RS de type dit filamentaire (Figure 1b), ou à l'interface
électrode/matériau, connu sous le nom de RS de type interface (Figure 1c). La RS filamentaire est
généralement observée après une étape d'électroformage à haute tension et les caractéristiques
électriques montrent des changements abrupts de courant (ou de résistance) liés à la
connexion/déconnexion du filament, ce qui se traduit par un comportement numérique ou binaire (ON
et OFF, ou 1 et 0). Néanmoins, l'électroformage peut produire plusieurs filaments, ce qui compromet la
fiabilité et la reproductibilité des dispositifs. En revanche, la RS de type interface ne nécessite pas d'étape
d'électroformage, et les caractéristiques I(V) sont continues (type analogique) grâce à l'accumulation
progressive d'espèces à l'interface. En outre, plusieurs états de résistance intermédiaires peuvent être
obtenus en modifiant la tension de fonctionnement.
Figure 1. a) Schéma d'un dispositif memristif. Mémoire à changement de valence avec commutation résistive de
type b) filamentaire et c) interface.
La RS a été rapportée dans de nombreux matériaux. Cependant, le mécanisme sous-jacent de la RS, et

donc des performances du dispositif, est influencé par la composition exacte du matériau, la topographie
et la morphologie, le matériau de l'électrode, leurs interfaces, la dimension de chaque élément du
dispositif et l'histoire électrique. Dans cette thèse, nous avons choisi l'oxyde perovskite fortement corrélé
La1-xSrxMnO3-δ (LSM), dont la flexibilité chimique (appauvrissement en cations et/ou anions), les
propriétés électriques et les propriétés de transport d'oxygène en font un candidat approprié pour les
VCMs, et en particulier, nous nous sommes concentrés sur l'influence de la microstructure sur la
performance du dispositif.
xxiv
En bref, les propriétés du LSM, et en particulier la résistivité, dépendent du rapport Mn3+/Mn4+ qui peut
être modifié par la substitution en Sr, x, et/ou par la teneur en oxygène, δ, (formation-annihilation de
lacunes d'oxygène). Les LSM présentent également un transport d'oxygène qui nécessite la présence de
lacunes d'oxygène, et il est donc renforcé le long des défauts structurels étendus, tels que les dislocations
et les joints de grains, qui sont riches en ces lacunes d'oxygène.
Le LSM a déjà été étudié pour des applications de RS et la plupart des publications s'accordent sur le
rôle clé des lacunes d’oxygène (présence et migration) pour déclencher une réponse memristive.
Néanmoins, le rôle de la microstructure dans la RS n'a jamais été directement évalué et étant donné le
grand nombre de variables liées au dispositif présentes dans la littérature, une évaluation généralisée à
partir de celles-ci est compliquée. Par conséquent, cette thèse vise à combler le vide existant dans la
littérature.
Croissance des couches de LSM
Les couches minces de LSM sont développées dans un réacteur de dépôt chimique en phase vapeur de
type métal-organique à injection pulsée (PI-MOCVD). La solution de précurseurs est injectée par une
électrovalve à grande vitesse pour laquelle, dans des conditions de dépôt fixes, le nombre de pulsations
détermine l'épaisseur de la couche. Afin d'obtenir des couches LSM adaptées aux applications
memristives, deux études sont nécessaires : l'optimisation du réacteur et l'optimisation de la composition
de la couche. Dans la première, les conditions de dépôt (débit de gaz, pression et emplacement du porte-
échantillon) sont modifiées jusqu'à l'obtention d'une zone de dépôt de 2 pouces de diamètre homogène
en épaisseur et en composition, ce qui permet la croissance simultanée sur plusieurs substrats de 10x10
mm2. Dans la deuxième optimisation, plusieurs couches sont développées sur différents substrats, à
différentes températures, et en utilisant une solution de précurseurs dont les concentrations relatives ont
été systématiquement modifiées, afin de découvrir : (i) la relation entre la composition de la solution de
précurseurs et la composition de la couche, (ii) l'effet de la température et (iii) l'effet du substrat sur les
couches.
La composition des couches a été mesurée par EPMA et RBS-NRA (Rutherford Backscattered
Spectrometry and Nuclear Reaction Analysis), dont les résultats sont présentées Figure 2. L'analyse de
la microstructure et de la morphologie a été effectuée par DRX et MEB, respectivement. De ces
caractérisations nous concluons ce qui suit :
 Des compositions similaires sont obtenues dans la gamme de températures de dépôt étudiée.
 L'incorporation du Sr dans la couche est moins favorable que celle du La et de grandes quantités
dans la solution sont nécessaires pour la croissance des LSM20 et LSM50.
 En raison du décalage du réseau cristallin des LSM avec celui le substrat :
o Des couches de LSM polycristallines sont obtenues sur les substrats Si et Si3N4.
xxv
o Des couches hautement orientées (ou épitaxiées) sont obtenues sur des substrats LaAlO3
et MgO.
 Les couches synthétisées à 675 °C présentent une plus grande densité de surface (continuité) et
c'est pourquoi cette température est choisie pour la croissance de couches destinées à fabriquer
des dispositifs memristifs.
Figure 2. Composition de la couche déposée à différentes températures en fonction de la composition de la solution

de précurseurs. Les compositions qui ont donné les LSM20 et LSM50 sont mises en évidence.
Rôle de la nanostructure dans le LSM20
La microstructure va déterminer les propriétés physiques des couches minces et donc les performances
du dispositif. La microstructure peut être modifiée en choisissant les substrats adéquats. Des couches
épitaxiales très fines (7 nm) de LSM20 ont été synthétisées sur des substrats de SrTiO3 (STO) et de
LaAlO3 (LAO), afin d'obtenir des couches déformées en traction et en compression dans le plan,
respectivement. En outre, la couche LSM20/LAO présentait quelques dislocations qui pourraient être
liées à la plus grande discordance des réseaux cristallins. Des dispositifs memristifs planaires, où les
électrodes se trouvent sur le dessus de la couche, ont été fabriqués en salle blanche en utilisant du Pt
comme électrode inerte et du Ti (recouvert de Pt pour éviter l'oxydation atmosphérique) comme
électrode active (Figure 3a). Les dispositifs vierges présentent une couche intermédiaire en TiOx
déficiente en oxygène à l'interface Ti/LSM20, qui se forme lorsque le Ti pompe l'oxygène du LSM20
lors de la fabrication, comme le montrent les techniques spectroscopiques, c'est-à-dire les mesures EDS-
STEM et XANES (réalisé dans les installations du synchrotron ESRF).
xxvi
Figure 3. a) Schéma d'un dispositif memristif planaire Ti/LSM\Pt. Caractéristiques résistance-tension d'un
dispositif memristif sur des couches b) de LSM20/STO et c) de LSM20/LAO. d) Schéma du modèle
phénoménologique de l'oxygène de la section transversale à l'interface Ti/LSM dans les états de HRS et LRS.
Les dispositifs memristifs Ti/LSM20\Pt fabriqués sur LSM20/STO (Figure 3b) et LSM20/LAO (Figure
3c) montrent des caractéristiques R(V) dans le sens inverse des aiguilles d'une montre, où le HRS et le
LRS sont programmés à une polarité positive et négative, respectivement. Ces dispositifs présentent
également des états de résistance multiniveaux qui, ajoutés au changement progressif de la résistance,
suggèrent une RS de type interface. Les dispositifs sur LSM20/LAO présentent une plus grande fenêtre
de fonctionnement, c'est-à-dire un rapport HRS/LRS, et une meilleure reproductibilité que ceux sur
LSM20/STO, pour lesquels le LRS peut difficilement être restauré à des tensions élevées et entraîne une
augmentation de la résistance pendant le fonctionnement. Afin d'expliquer les performances observées,
nous avons proposé un modèle phénoménologique (Figure 3d) qui peut être résumé par les points
suivants :
 L'état de résistance est déterminé par l'échange d'oxygène entre le LSM20 et l'électrode de Ti :
o Le HRS est déterminé par la réduction du LSM20 et l'oxydation simultanée du Ti en
TiOx. Les réactions redox inverses se produisent dans le LRS.
o Les multi-états peuvent être expliqués par des degrés intermédiaires de ces réactions
redox.
 La présence de dislocations dans les dispositifs à base de LSM20/LAO permet une migration
rapide de l'oxygène, ce qui permet d'obtenir des HRS/LRS plus importants et de faire passer le
dispositif d'un HRS à un LRS de manière réversible.
 La haute qualité cristalline du LSM20/STO (pas de dislocation), empêche la réincorporation de
l'oxygène du TiOx dans le LSM20, ce qui augmente la résistance lors du fonctionnement.
Les mêmes performances ont été obtenues pour des couches légèrement plus épaisses (8 nm) de LSM20
développées sur STO et LAO. En outre, du LSM20 polycristallin (8 nm) a été développé sur des
substrats Si3N4. Il est intéressant de noter que ces dispositifs polycristallins présentent un rapport
xxvii
HRS/LRS encore plus élevé que les dispositifs en LSM20/LAO, ce qui a été attribué à la migration
améliorée de l'oxygène à travers les joints de grain présents dans la couche. Les résultats obtenus sur le
LSM20/Si3N4 polycristallin sont très intéressants car la couche est formé sur un substrat compatible
avec l'industrie des microtechnologies.
Des couches en LSM20 plus épaisses (21 nm) ont été développées sur les trois types de substrats
précédents. Les dispositifs memristifs fabriqués sur le LSM20/STO montrent des performances
reproductibles lorsque la tension négative SET est plus grande que la tension positive RESET : ce qui
empêche l'augmentation de la résistance observée dans les dispositifs plus fins et permet de "restaurer"
le LRS. Cela suggère que l'oxygène peut être réincorporé dans le LSM20 sous des tensions SET plus
élevées par rapport à la tension RESET.
En conclusion, ces résultats suggèrent que les défauts structurels étendus améliorent les performances
des dispositifs memristifs à base de LSM20.
Commutation résistive dans le LSM50 épitaxial
Des couches épitaxiales de LSM50 de 21 nm ont été développées sur des substrats LAO. Les couches
présentent deux types de domaines cristallins presque identiques et peu de dislocations (Figure 4a). Les
caractéristiques électriques des dispositifs memristifs planaires Ti/LSM50\Pt montrent le même sens de
commutation que les dispositifs à base de LSM20 : passage au HRS à une tension positive, et au LRS à
une tension négative (Figure 4b). Cependant, la transition vers le LRS présente de petits changements
abrupts de résistance qui suggèrent la présence de multifilaments. Les dispositifs à base de LSM50
présentent des performances reproductibles, fonctionnent à une tension plus faible, et le rapport
HRS/LRS mesuré est plus important que celui du LSM20. La différence observée est attribuée à la plus
grande concentration intrinsèque de lacunes d'oxygène dans le LSM50 par rapport au LSM20, et à la
présence de dislocations qui permettent des changements plus importants de la résistance. Cependant,
des recherches supplémentaires sont en cours pour comprendre ces dispositifs et proposer une image
cohérente de la phénoménologie.
xxviii
Figure 4. a) Coupe transversale MET de LSM50/LAO et image de la transformée de Fourier inverse où une
dislocation est mise en évidence ("⊤"). b) Courbe résistance-tension d'un dispositif memristif planaire
Ti/LSM50\Pt.
Effet de la stœchiométrie de l'oxygène dans les couches en LSM
En complément, d'autres expériences ont été menées pour élucider le rôle de l'oxygène dans les
propriétés des LSM. La teneur en oxygène des couches en LSM a été modifiée par des traitements
thermiques sous différentes atmosphères.
Les spectres Raman de ces couches montrent des différences dans les bandes Jahn Teller (JT) sensibles
à la teneur en oxygène : les couches réduites présentent une bande JT plus large et pointue tandis que
les couches oxydées présentent des bandes JT plus basses et plus larges. Cette technique permet de
"suivre" la teneur en oxygène dans les dispositifs memristifs VCM dans des conditions operando.
Les mesures de spectroscopie d'annihilation de positrons (PAS) ont suggéré la présence de lacunes
d'oxygène dans le VCM, dont la concentration et la "taille" augmentent dans les couches recuites sous
atmosphère réductrice.
La spectroscopie d'absorption des rayons X (XAS) au niveau du K-edge du Mn a révélé que le Mn est
extrêmement réduit sous l'électrode de Ti dans un dispositif vierge, alors que la valence du Mn du reste
du dispositif suggère un état d'oxydation plus grand. Ces résultats indiquent une oxydation spontanée
du Ti en TiOx pendant la fabrication du dispositif par pompage de l'oxygène du LSM. Comme discuté
dans le manuscrit principal, cette technique offre également la possibilité de suivre le changement de la
teneur en oxygène dans les dispositifs memristifs dans des conditions operando.
xxix
Enfin, la résistivité électrique des couches en LSM20/LAO et en LSM20/STO en fonction de la

température et de la pO2 a également été mesurée pour évaluer le rôle de la microstructure/substrat. Bien
que la résistivité du LSM20/LAO soit légèrement supérieure à celle du LSM20/STO, les deux couches
subissent une diminution de résistivité sous atmosphère d'oxygène. À température ambiante, la couche
en LSM20/STO présente une conduction métallique avec une transition métal-isolant (MIT) à 62 °C.
D'autre part, le LSM20/LAO présente une conduction de type isolant sur une large gamme de
température avec une MIT autour de -160 °C. La différence de résistivité électrique dans les couches en
LSM20/STO et en LSM20/LAO a été attribuée à la déformation induite par le substrat, où le
chevauchement des orbitales est modifié, ce qui entraîne une modification des propriétés des LSM
fortement corrélés.
Conclusion et perspectives
Dans ce travail, nous avons démontré que dans les dispositifs memristifs, la microstructure est aussi
importante que le matériau. En effet, ces dispositifs électrochimiques présentent différentes régions
actives qui peuvent être exploitées pour améliorer les performances. De nombreux résultats stimulants
ont été obtenus, mais de nombreuses expériences ont aussi échoué pour lesquelles nous proposons les
conditions qui devraient être remplies pour réussir. Néanmoins, une compréhension plus profonde
pourrait être poursuivie, par exemple :
 Différentes électrodes actives pourraient être étudiées.

 Étude des conditions environnementales pendant le fonctionnement
 Effectuer des caractérisations operando (XANES, TEM, Raman...)
 Mesurer des dispositifs intégrés verticalement (configuration proche de l'application finale).
 Caractérisations des dispositifs plus poussées par fonctionnement par pulsation (comme dans
l'application finale).
xxx
Introduction
Motivation ................................................................................................................................. 1
Resistive switching phenomena ............................................................................................... 2
Device phenomenology and first observations .............................................................. 2
Classification ................................................................................................................. 3
Strontium-substituted lanthanum manganite ........................................................................ 6
Bulk: crystal structure, electronic properties and phase diagram .................................. 6
LSM thin films: electrical properties ........................................................................... 10
Oxygen mass transport ................................................................................................ 13
Memristive response of La1-xSrxMnO3 .................................................................................. 15
Role of the electrodes .................................................................................................. 19
Scope and organization of the thesis ..................................................................................... 22
Motivation
The last years have proven that our society evolves towards digitalization where many professional and
non-professional interactions are made through a screen. These interactions involve a large amount of
data and power consumption that keeps raising as society asks for more content, better quality, faster
operation and better energy-efficiency. On top of that, industry and domestic life evolve towards
digitalization. For example, users can control the heating of their house remotely while keeping track of
the energy usage every hour. All in all, the proliferation of data scales fast and so does the power
demand. While energy production and storage are an interesting field of research, this thesis targets one
of the core elements of the energy consumption in electronic devices: the memory.
The current memory market offers two main types of technologies: Dynamic Random Access Memories
(DRAM) and Flash memories. Both types rely on CMOS (Complementary Metal-Oxide-
Semiconductor) technology, which internal architecture yields different characteristics. DRAMs exhibit
fast operating time (write, erase and read time down to 10 ns) and large endurance (>106 cycles).
However, they exhibit very poor retention times (60 ms) and have to be constantly powered. On the
other hand, Flash technology offers slower operation times (0.1-1 ms), shorter endurance (103-
106 cycles), but larger retention time (10- 20 years), and so they are used as non-volatile memory
1
Chapter 1: Introduction
(NVM). As a result of the different characteristics, DRAMs are used for data processing and Flash
memories are used for data storage, typically forming part of two physically separated units in the
devices with von Neumann architecture: the Central Processing Unit (CPU) and Memory Unit (MU).
The physical separation of the units and their different operating speeds limit the overall performance
of the devices by creating a data flow bottleneck. Increasing the number of cores (“CPU+MU”) allows
a larger amount data traffic to flow in parallel, yet the bottleneck remains. Additionally, since 1965, the
number of transistors on a chip has roughly doubled every two years, observation which is known as
Moore’s law.[1] Increasing the number of transistors while decreasing their size allows a larger data
traffic, however, further downscaling leads to fundamental physical limits that cannot be overcome even
by improving the manufacturing techniques. The ‘International Roadmap for Devices and Systems’ has
addressed this issue defining the research focus named Beyond CMOS.[2] This group investigates
variation of MOS devices beyond purely electronic conduction for the next generation of memories.
Some of the new memories proposed rely on spintronics, magnetics, photonics, molecular electronics,
ferroelectricity, and memristive phenomena, among others. In particular, memristive phenomena are
exploited in the so-called Resistive Random Access Memories (RRAM). RRAM are very promising
candidates to overcome the scaling limitations while combining the best properties of the two main
memory technologies, i.e. high speed, high endurance and large retention times,[3]–[5] and hence, they
satisfy the requirements of an ideal and unified memory.[6]
Resistive switching phenomena
Device phenomenology and first observations
The high interest for RRAM lies in its simple two-terminal metal-insulator-metal (MIM) structure where
the “M” is a good electronic conductor, and “I” is an electronic insulator (or semiconductor) and often
ionic-conducting[7] (Figure 1.1). RRAM relies on Resistive Switching (RS) phenomena where, by
applying an electrical stimuli, the resistance can be reversibly switched between at least two resistance
states which are retained under zero-field conditions: the high resistance state (HRS), and the low
resistance state (LRS).[8] In the literature, the HRSLRS transition is called SET, while the LRSHRS
transition is referred to as RESET.
The first observations of RS date back to the 1960s when hysteretic behavior of the current-voltage
curves was reported in MIM structures using binary oxides: Al2O3, Ta2O5, ZrO2, TiO2, NiO, and SiO.[9]–
[12]
In the 1990s, with the outbreak of perovskites oxides and their exotic properties, RS became a hot
topic again, being reported for example in Pr1-xCaxMnO3 (PCMO)[13] and La1-xSrxMnO3 (LSM).[14] RS
2
1.2 Resistive switching phenomena
Figure 1.1. a) Sketch of a RRAM cell with MIM structure. Typical current-voltage characteristics of b) unipolar
and c) bipolar RRAM devices. The resistance states in b) and c) are named ON (LRS) and OFF (HRS). Figure b)
and c) reproduced with permission from [7].
blossomed in 2008 when S. Williams’ group at Hewlett-Packard Labs linked the experimental electrical
characteristics of the switching materials with the memristor theory,[15] first established in 1971[16] and
revisited in 1976[17] by Chua et al.
Classification
Resistive switching has been observed in many materials and stacks. However, a clear understanding of
the mechanism that triggers the change of resistance is still missing and several models have been
proposed.[18] These models, and so RS, can be classified based on different parameters.
Regarding to the switching voltage polarity, Unipolar or Bipolar Resistive Switching can be defined.
In Unipolar Resistive Switching (URS), the SET and RESET transitions depend only on the amplitude
of the applied voltage but not on the polarity. Typical current-voltage characteristics are sketched in
Figure 1.1b, where the resistance states are named ON (LRS) and OFF (HRS). The positive polarity is
considered in the following description. A pristine cell in HRS will switch to a non-volatile LRS when
a threshold voltage is reached (VSET>0). Then, the cell will switch from LRS back to HRS when voltage
is applied (VRESET>0). During the SET process a current compliance (CC) is imposed to prevent
damaging the cell whereas the RESET process does not require it. Typically the threshold voltage of the
RESET is smaller than the SET (|VRESET|<| VSET|).
In Bipolar Resistive Switching (BRS), the SET and RESET transitions occur at opposite voltage
polarities. Figure 1.1c shows a typical I(V) curve of a BRS device. BRS is further classified depending
3
Figure 1.2. Mechanisms responsible of resistive switching phenomena and their classification based on the key
element in the system (blue) and working polarity (green). Special emphasis is given to redox based related
mechanism (red). Reproduced with permission from [23].
on the switching sense observed in the I(V) curves as: (i) counter-eight-wise (C8W), when the RESET
transition (LRS to HRS) takes place at positive voltage, as shown in Figure 1.1c; or (ⅱ) eight-wise (8W),
for a SET transition (HRS to LRS) at positive voltage. While these terms refer to linear I(V) plots, they
are frequently replaced by (i) clockwise (CW), and (ⅱ) counter-clockwise (CCW) when using semi-
logarithmic plots.
In all cases, the states can be measured by a non-destructive readout voltage (Vread) under the condition
|Vread |<< |V(RE)SET|.
Resistive switching can be triggered by different physical and/or chemical phenomena: magnetoresistive
phenomena,[13], [14], [19], phase changes,[20] electronic effects,[21], [22], or ionic effects.[23] In the latter, the
ionic migration at the nanoscale induces local redox reactions which affect the conductivity of the
material. Resistive memories based on redox reactions are known as ReRAM (Redox based Resistive
Random Access Memories). Figure 1.2 shows a concise classification of the mechanisms responsible
for RS, emphasizing those in ReRAM. For the mechanisms not mentioned in this thesis, the reader is
referred to the source reference.[23] ReRAM can be split based on the two main mechanisms responsible
of ionic migration, temperature and/or electric field, and the species involved.
4
1.2 Resistive switching phenomena
Thermochemical Memories (TCMs) are inherently URS systems where the raise of temperature induced
by current, i.e. Joule heating, changes the local stoichiometry and the conducting properties in a
filament-shaped region.[24]
Electrochemical Metallization Memories (ECMs) are BRS systems, in which the cations from an active
electrode, such as Ag or Cu, migrate in an insulating matrix causing an electrochemical metallization
and hence, a decrease of resistance when they reach the opposite electrode. The migration can be
reversed by applying voltage of opposite polarity. The mechanism in ECMs is associated to conducting
filaments.
Valence Change Memories (VCMs) are BRS systems, in which the active species are the anions,
typically oxygen ions that migrate back and forward under an electric field inducing a valence change
in the cation sublattice.[25] The first spectroscopic evidence on the change of valence in oxides was
reported by Janousch et al. in 2007.[26]
ReRAMs can be also classified regarding the area involved in the switching event: (i) filamentary, where
the change occurs in a very localized region interconnecting the electrodes, (ⅱ) interfacial, where the
event takes place at the interface of the electrode and the active material, and (ⅲ) volume-type, which
can be understood as an interface-type expanding towards the whole volume of the active material. A
schematic of the filamentary and interface-type models is presented in Figure 1.3. The three types of
ReRAMs (TCM, ECM and VCM) can present filamentary-like switching after a high voltage
electroforming step.[7] This step can be understood as a controlled dielectric breakdown between the two
terminals to form the conducting filament in the insulating matrix. However, multiple filaments might
form affecting the reproducibility and reliability of the devices. These filaments are constituted by
oxygen vacancies in TiO2 after an electroforming step, as observed by high resolution transmission
electron microscopy (HRTEM) in 2010.[27] On the other hand, VCMs can also present interface-type
switching in which RS is triggered by a gradual modification of the chemical environment next to the
Figure 1.3. Schematic of the a) filamentary and b) interface-type models of resistive switching. Reproduced with
permission from [18].
5
interface and thus, changing the conducting properties.[28] Interface-type devices have gained late
interest as they could offer more reliable performance and multilevel storage, i.e. several resistance
states.[28] The underlying mechanism behind RS can be altered by changing the nature of the oxide
material and its morphology and topography, as well as the materials used as electrodes.[18], [23]
This thesis is focused on the strong correlated perovskite oxide La1-xSrxMnO3-δ (LSM), which chemical
flexibility, electrical properties and oxygen mass transport properties make it a suitable candidate for
VCMs. A description of the bulk and thin film properties of LSM as function of the composition (Sr and
O), as well as how they are modified by defects, is presented in the following section.
Strontium-substituted lanthanum manganite
Bulk: crystal structure, electronic properties and phase diagram
Perovskite compounds have gained a lot of attention in the last decades due to the fact that the physical
properties are strongly correlated with the interaction between the different degrees of freedom of the
system, e.g. spin, charge, lattice and orbital.[29] In addition, most of the elements of the periodic table
can stabilize in a perovskite structure.[30] Some of the properties shown by perovskites are:
ferroelectricity, ferromagnetism, metal-insulator transition, colossal magneto-resistance, high
temperature superconductivity, piezoelectricity, catalytic properties and high ion conductivity.
Perovskite compounds are described by the chemical formula ABO3, where A and B are cations, being
B a 3d transition metal. The prototypical perovskite adopts a cubic structure where the metal A is in the
center of the unit cell, and B in the vertices in octahedral coordination with oxygen anions, forming a
BO6 octahedron, as shown in Figure 1.4. However, different A and B atoms may cause the system to
present other structures. This distortion can be predicted based on the relative cation-oxygen bond
distance, according to the Goldsmith tolerance factor (TF), defined as
𝑇𝐹 = (𝑟𝐴 + 𝑟𝑂 )⁄[√2 (𝑟𝐵 + 𝑟𝑂 )]

Equation 1.1
where rA, rB and rO are the radii of the A cation, B cation and oxygen anion, respectively. The ideal cubic
perovskite structure with no distortions corresponds to a TF=1. In most cases, the TF deviates from 1,
and lower symmetry structures are accommodated in the range of 0.80<TF<1.1.[30] A polar distortion
occurs for TF>1, where the relative displacement of the cation and the oxygen anion creates an electrical
dipole. When TF<1, the oxygen octahedra around the B cations rotate and hence, the B-O-B angle is
6
1.3 Strontium-substituted lanthanum manganite
A
B
O
Figure 1.4. Equivalent representations of the cubic perovskite structure with chemical formula ABO3.
reduced.[31] This angle is important in materials where the magnetic and/or conduction properties arise
from the overlap between the 2p orbitals of the oxygen anion and 3d orbitals of the B cation, such as the
manganites studied in this thesis.
Manganites belong to the perovskite family where the B-site is a Mn cation and the A-site is partially
occupied by a trivalent (typically a rare earth metal) and a divalent ion (typically an alkaline metal). The
partial substitution in A is usually accommodated by a change in Mn valence. For this reason, they are
frequently referred to as mixed-valence manganites.[32] In strontium-substituted lanthanum manganites,
with chemical formula La1-xSrxMnO3 (LSM), the A-site is occupied by La3+ and Sr2+ and manganese
presents an oxidation state between +3 and +4 to maintain the charge neutrality. The manganese
oxidation state increases with strontium substitution, quantified as x, and due to the different ionic radii
of La3+ and Sr2+, the Mn-O-Mn bond angle is modified, which ultimately changes the physical properties
(a more detailed description is given below).
The crystal field potential generated by the oxygen octahedron breaks the degeneracy of the Mn 3d
electron multiplet into three low energy t2g degenerate states and two high energy eg degenerate states.
The electron occupancy of these states is presented for Mn2+, Mn3+ and Mn4+ in Figure 1.5a. On top of
that, another distortion and break of degeneracy takes place due to the half-filled eg band in Mn3+: the
Jahn-Teller (JT) distortion which results in an uniaxial elongation or compression of the MnO6
octahedron.[31], [33], [34] However, the JT distortion does not occur for Mn+2 and Mn+4, where the eg band
is fully occupied and empty, respectively. In LSM, the Mn t2g electrons are considered localized due to
a lower hybridization with the O 2p orbitals, while the eg electrons, largely hybridized with the oxygen
orbitals, are considered the charge carriers.[34] Nonetheless, the eg electrons may become localized due
to the JT distortion, increasing the electrical resistivity.[34] The electrical conduction takes place by
hopping of an eg electron from a Mn3+ ion to a neighbor Mn4+ hole through the oxygen orbital with a
probability proportional to cos(θ/2), where θ is the angle between the Mn spins.[32], [35] This mechanism
is known as the Double Exchange (DE) mechanism[36], [37] and requires of empty states or holes, i.e.
Mn4+, as depicted in Figure 1.5b.
7
a) b)
1.
dx2-y2 Mn3+ O2- Mn4+
eg eg eg
dz2
2.
t2g t2g dxy t2g
dxz, dyz
Mn4+ O2- Mn3+
Mn2+ Mn3+ Mn4+
Figure 1.5. a) Electron occupancy of the Mn 3d levels for different oxidation states. b) Sketch of the Double
Exchange mechanism. Redrawn from [172].
The electrical properties of LSM are strongly related the average Mn valence. [38] On the one hand, the
substitution of La3+ cations by Sr2+ increases the Mn4+/Mn3+ ratio and the electronic conductivity,
reaching a maximum around x=0.5.[38]–[43] The temperature dependence of the electrical resistivity, ρ, in
LSM with x=0-0.4 is presented in Figure 1.6a. While the parent compound LaMnO3 (LMO, x=0)
presents an insulating (or semiconducting, dρ/dT<0) behavior in a broad range of temperatures, the
introduction of Sr triggers a metallic behavior (dρ/dT>0) at low temperatures. As the Sr content
increases, the metal to insulator transition (MIT) temperature increases (indicated with arrows in
Figure 1.6a). On the other hand, oxygen deficiency in LSM, decreases the Mn4+/Mn3+ ratio and induces
a similar trend in electrical resistivity. The electrical resistivity of La0.67Sr0.33MnO3-δ with different
Figure 1.6. Bulk electrical resistivity as a function of the temperature in a) LSM with different composition, x,
and b) LSM with x=0.33 and different oxygen content, δ. Reprinted with permission from [14] and [142], respectively.
8
3.2
La1-xSrxMnO3+δ at 1073 K
3.1
Sr, x
3.0
3+δ
2.9
x=0
x = 0.2
2.8 x = 0.4
x = 0.5
2.7
-20 -15 -10 -5 0
log [pO2 (bar)]
Figure 1.7. Effect of Sr substitution on the oxygen stoichiometry in LSM at 1073 K. Redrawn from [38].
oxygen stoichiometry is shown in Figure 1.6b. Similar variations of the electrical properties with
oxygen content have been reported for other manganites, such as La0.67Ba0.33MnOα (2.80<α<2.99).[44]
Furthermore, the presence of oxygen vacancies is also related to the Sr substitution. As Sr 2+ cations
replace La3+ cations, the system requires of extra positive charges to maintain electro-neutrality. Hence,
Mn4+ is formed at expenses of Mn3+ (as discussed before), and positively charged oxygen vacancies (𝑉𝑂∙∙)
are generated.[45], [46] Figure 1.7 presents the oxygen stoichiometry (3+δ) of LSM with different Sr
substitution as function of pO2 at 1073 K. While LSM with x=0 can accommodate oxygen excess, i.e.
δ>0 (by formation of cation vacancies), the system becomes more reducible (less oxygen content) with
the incorporation of Sr2+.[38], [47] In numbers, for LSM with x=0.2, oxygen stoichiometry (δ=0) is retained
down to pO2=10-14 bar, while for x=0.5 down to pO2=10-10 bar, which suggests easier oxygen vacancy
formation for the latter. This hypothesis was further confirmed by Lee et al.[48] by DFT calculations in
which the oxygen vacancy formation energy decreases with strontium substitution: 4.3 eV (x=0), 4.0 eV
(x=0.25), 3.5 eV (x=0.5), and 1.92 eV (x=1).
Being LSM a semiconductor, the majority carriers depend on the average Mn valence by self-doping
mechanism. LSM is a p-type (n-type) semiconductor when z≤0.5 (z≥0.5), being z=x+2δ.[38]
LSM is a strongly correlated material, which properties are defined by the interaction of the degrees of
freedom of the system. While strontium substitution has an impact on the electrical properties, the
magnetic and crystal structure are also modified, originating a rich phase diagram, as presented in
Figure 1.8. Briefly, the two end-compounds of LSM, i.e. x=0 and x=1, show paramagnetic and
insulating behavior at room temperature. LSM is metallic in the range 0.17<x<0.93, range in which it
adopts different crystal structures: rhombohedral (0.17<x<0.47) and tetragonal (0.47<x<0.93), and a
9
Figure 1.8. Phase diagram of La1-xSrxMnO3 in the whole range of Sr substitution, x, indicating the crystal structure
[JT distorted orthorhombic: O’, orthorhombic: O, orbital-ordered orthorhombic: O”, rhombohedral: R, tetragonal:
T, monoclinic: Mc, and hexagonal: H], the magnetic structures [paramagnetic: PM (green), short-range order (SR),
canted (CA), A-type antiferromagnetic structure: AFM (yellow), ferromagnetic: FM (blue), phase separated (PS)
and AFM C-type structure] and the electronic state [insulating: I (dark), metallic: M (light)]. Reprinted with
permission from [49].
ferromagnetic phase is stabilized for 0.2<x<0.6. For further details the reader is referred to references
[49], [50]
.
Therefore, the two LSM compositions studied in this thesis are:
 LSM with x=0.2 (LSM20): rhombohedral structure, metallic conduction (p-type) and
ferromagnetic behavior.
 LSM with x=0.5 (LSM50): tetragonal structure, metallic conduction (p-type) and ferromagnetic
behavior.
Despite LSM50 may present n-type conduction, according to Mizusaki et al.[38], the unavoidable
presence of oxygen vacancies will turn it into a p-type conductor.
LSM thin films: electrical properties
The electrical properties of LSM thin films show a similar dependence with strontium substitution [51]
and oxygen deficiency[52] to those previously described for bulk LSM and shown in Figure 1.6.
Nonetheless, as the dimensionality decreases, i.e. from bulk (3D) to thin films (2D), the contribution of
interfaces and defects on the measured properties become non-negligible. Therefore, the impact of the
10
substrate on the film should be taken into account, which might result in polycrystalline or epitaxial film
under a substrate-induced strain, and which may induce formation of defects.
Polycrystalline films and grain boundaries
Polycrystalline films are composed of grains of the same composition oriented in different directions.
The interface between the grains is referred to as a grain boundary (GB). GBs can be classified based
on the misorientation angle, θ, of the neighboring grains: where θ is defined as the rotation angle required
to align the grains in the same orientation. GBs with small θ accommodate the different atomic
arrangement on both sides of the interface by a periodic arrangement of dislocations. Figure 1.9 presents
a) a Fourier filtered HRTEM image of a grain boundary in LSM20 with the array of dislocations pointed
by arrows, and b) a colored enlargement of the previous.[53] As the angle θ increases, and thus, the atomic
coherence across the interface decreases, the dislocation density at the GB increases. Therefore, GBs in
LSM can be considered extended defects where the structural disorder alters the chemical composition
resulting in an oxygen-vacancy enrichment[53]–[55] and different cationic composition than the grain (also
referred to as bulk).[53] As result of the chemical disorder at the core of the GBs, a rearrangement of the
chemical species around the core occurs to preserve charge neutrality, forming the so-called space
charge region. Therefore, the GBs in LSM exhibit physical properties different than the bulk[56], [57] and
in particular they present an electronic insulating behavior.[53]
Epitaxial films and strain
On the other hand, strain-free epitaxial films, typically grown on substrates with comparable cell
parameter, present long range atomic order with absence of GBs, and with properties comparable to
those of the single crystals.[58] Typically, the film’s lattice will elastically deform to adopt the atomic
Figure 1.9. a) HRTEM Fourier filtered image of LSM20 grain boundary formed by an array of periodically
arranged dislocation (arrows). b) Colored enlargement of the red square in a) where the dislocation are marked as
“⊤”. Reprinted with permission from [53].
11
Figure 1.10. Effect of the epitaxial strain in electrical properties of La0.7Sr0.3MnO3 films. a) Ratio between the out-
of-plane (azz) and in-plane (axx) lattice parameter of epitaxial LSM30 films as function of the pseudo-cubic in-
plane substrate lattice parameter. b) Electrical resistivity as function of temperature in strained epitaxial LSM30
thin films. The films grown on SmScO3, GdScO3, and NdScO3 present a resistivity of 1 Ω·cm at room temperature.
Reprinted from [58], with the permission of AIP Publishing.
arrangement imposed by the substrate with different cell parameter. This substrate-induced deformation
is referred to as biaxial epitaxial strain. When the cell parameter of the substrate is larger (smaller) than
the cell parameter of the film, the film presents in-plane tensile (compressive) strain and concomitant
shrinkage (elongation) in the out-of-plane direction. Nonetheless, if the difference between film and
substrate cell parameters (misfit or mismatch) is large, nucleation and growth of defects will occur to
release the strain, causing plastic deformation. This phenomenon is called strain relaxation. Strain
relaxation can also occur in thick films with small mismatch, where the elastic energy (imposed by the
substrate) builds up with thickness and eventually overcomes the nucleation and growth energy of
defects, such as dislocations. As it occurs for the JT distortion in LSM films, the substrate-induced
elastic deformation of the film changes the hybridization of the Mn eg electrons and O 2p orbitals, and
thus, the electrical properties.[58], [59] Figure 1.10 presents a) the cell deformation of La0.7Sr0.3MnO3
(LSM30) thin films, expressed as the film‘s cell parameter out-of-plane/in-plane (azz/axx) ratio, grown
on different substrates, and b) the electrical resistivity as function of temperature of those films.
Interestingly, the magnitude of strain (either compressive or tensile) induces a similar effect on the
electrical properties to the increase of oxygen vacancies in LSM or the decrease of Sr content (for x=0-
0.5) (see Figure 1.6 for comparison). At room temperature, the resistivity of slightly strained
(εxx=+0.6%) LSM30/SrTiO3 is 5 mΩ·cm, with metallic conduction, while for the largely strained
(εxx=- 2.3%) LSM30/LaAlO3 is 20 mΩ·cm, with insulating conduction.[58] Similar results with smaller
changes have been reported by Chaluvadi et al.[59] in epitaxial La0.67Sr0.33MnO3 (LSM33), where the
effect of strain relaxation (increasing thickness) was also assessed: the larger the strain (mismatch), the
larger the relaxation with thickness and hence, larger decrease of resistivity.
12
Oxygen mass transport
As introduced earlier, oxygen mobility is a key factor to trigger the reversible redox reaction at the origin
of the switching phenomena in VCMs. The properties of LSM at high temperatures (500-1000 °C) have
been widely studied for its implementation as cathode material in Solid Oxide Fuel Cells (SOFC). [45],
[60], [61]
While LSM is usually considered a good electronic conductor with poor ionic conductivity,
several studies have succeed to enhance its ionic conductivity by microstructure engineering, thus
turning it into a mixed ionic electronic conductor (MIEC) which improves SOFC performance. [53], [62]
As discussed by Börgers et al.[63], bulk LSM is characterized by a slow oxygen diffusion (very low
oxygen vacancy concentration with low diffusivity), while faster oxygen diffusion along extended
defects is observed. The oxygen mass transport properties, i.e. oxygen diffusion (𝐷∗𝑂 ) and oxygen
exchange coefficient (k*), are typically measured by 18O isotope exchange and depth profiling analysis
by secondary ion mas spectroscopy (IEDP-SIMS).[64]–[68] Other techniques, such as electrochemical
impedance spectroscopy (EIS)[66], [69] or 18O tracer Raman spectroscopy [67], [70]–[72]
are also used. The
results are frequently cross-checked by numerical methods such as Finite Element Modelling (FEM).[66],
[67], [73]
Figure 1.11a presents the oxygen diffusion coefficient of ceramic LSM with x=0.2 and 0.5, and
polycrystalline and epitaxial LSM20 thin films as a function of temperature. To the best of my
knowledge, there is no experimental data available for oxygen diffusion in LSM50 thin films and
thereby, nor along extended defects. In LSM20, the oxygen diffusion component associated to the bulk,
(shaded area in Figure 1.11a) is similar for both ceramic and films, with 𝐷∗𝑂≈10-17 (10-16) cm2·s-1 at
600 °C (700 °C). However, polycrystalline thin films grown on Y:ZrO2 (YSZ) present an extra
component related to the GBs (DGB), while epitaxial films grown on LaAlO3 (LAO) and SrTiO3 (STO)
substrates show a component linked to dislocations (Dd). The oxygen mass transport along extended
defects in LSM20 is enhanced by dislocations (Dd≈10-15-10-14 cm2·s-1 at 600 °C), and particularly by
grain boundaries (DGB ≈ 10 – 14 -10 - 10 cm2·s-1 at 600 °C). The bulk diffusion coefficient increases with
Sr substitution and slightly larger values are reported for LSM with x=0.5 (𝐷∗𝑂 =2×10 - 15 cm2·s-1 at
[64], [65], [74], [75]
700 °C) . The faster migration has been associated to a larger concentration of oxygen
vacancies,[45] while the oxygen-vacancy migration barrier has been shown to be rather constant (0.6-
0.8 eV) with Sr content, as determined by DFT and Molecular Dynamics (MD) simulations.[48], [55], [76],
[77]
Nonetheless, the presence of point defects (antisite and cation vacancies) increases the migration
barrier to values comparable to the experimental data (1.2-2.4 eV).[55] Moreover, cationic ordering in
LSM50 has been predicted to enhance oxygen diffusivity. MD simulations showed an increase by factor
13
Figure 1.11. Oxygen mass transport in LSM20 and LSM50. a) Oxygen diffusion coefficient (𝐷𝑂∗ ) of the bulk,
dislocation (Dd) and grain boundaries (DGB) reported in ceramic LSM20 and thin films as function of temperature.
Thin films on LAO and STO are epitaxial while those on YSZ are polycrystalline. The oxygen mass transport
properties were obtained by Finite Elements analysis (FEM), Electrochemical Impedance Spectroscopy (EIS) or
by IEDP-SIMS (no additional text). b) Sketch of charge carrier concentration in log scale around a dislocation
core (negatively charged) forming a space charge tube, proposed by Börgers et al. [63] (redrawn from [63]).
1.5 in oxygen diffusivity for A-site ordered LSM50 with the La and Sr ions alternatively arranged
compared to a random arrangement.[76] The influence of cationic ordering in oxygen diffusion was
anticipated, experimentally and by simulations, for other good ionic conductive oxides, such as
Gd0.5Ba0.5MnO3 and Gd0.5Ba0.5CoO3.[78]–[81] An ordered cation lattice would prevent oxygen-vacancy
trapping by point defects (cation vacancies).[55], [79], [82] According to Norby et al.,[82] the enhancement in
oxygen transport properties is limited to those materials that can exhibit cationic ordering while
preserving a deficient and disordered anion sublattice. The cation order and anion disorder is more
favorable as the number of different cationic sites in the material increases as they can adopt more
complex perovskite structures.
Börgers et al.[63] offered a condensed picture of the oxygen transport along extended defects and
proposed a simple model based on the space charge phenomenon. The sketch of the model is reproduced
in Figure 1.11b. They suggest that the cation vacancies are responsible of the negatively charged
dislocation core. Congruently, the accumulation of positively charged oxygen vacancies and the
′′′ ′
depletion of negatively charged species (cation vacancies, 𝑉𝑐𝑎𝑡 , and Sr in a La site, 𝑆𝑟𝐿𝑎 ) take place in
14
1.4 Memristive response of La1-xSrxMnO3
the surroundings of the core. The presence of cation vacancies is known to hinder the oxygen vacancy
diffusion and therefore, oxygen vacancy diffusivity within the space charge tube increases towards the
core.
Some studies have addressed the impact of the A-site over B-site ratio (A/B = [(La+Sr)/Mn]) in LSM20
on its oxygen transport properties. On the one hand, in manganese deficient LSM20 with A/B=1.1-1.25,
DGB increases with Mn-deficiency up to one order of magnitude compared to stoichiometric LSM20
(A/B=1, x=0.2).[53] This enhancement was ascribed to the change of local chemical environment of the
GBs, in particular, to the increase of oxygen-vacancy concentration with Mn-deficiency. On the other
hand, manganese-rich polycrystalline LSM20 films with A/B=0.95 show similar bulk diffusion
coefficients to those of the films with A/B=1.[73] However, the grain boundary diffusion coefficient for
manganese-rich LSM20 with A/B=0.95 is enhanced by one order of magnitude at low temperatures
(600 °C) compared to LSM with A/B=1, while similar values are obtained for both compositions at
higher temperatures (1000 °C).
Memristive response of La1-xSrxMnO3

As seen above, the physical properties of LSM are strongly correlated to its composition and structure,
while the role of defects becomes more important as the dimensionality is reduced. Therefore, finding a
good balance between all the properties is crucial for the application of LSM into a device. As introduced
earlier, the exotic properties of LSM were studied in the 1990s, when a change of resistance was
obtained by a magnetic field.[14] Since then, electrically triggered RS has been reported for different
LSM compositions,[22], [83], [84] being La0.7Sr0.3MnO3 the most studied. Different device configurations
(vertical or planar) have been studied, as well as several electrode/LSM (or LMO) combinations, with a
large variety of electrode sizes used and film thicknesses. In the first case, a vertical device is built across
the film, where the film is sandwiched between a conductive substrate and a top electrode (or probe). In
the second case, a planar device is built along the film, typically grown on an insulating substrate, and
hence, two top electrodes are required to make the terminals of the device. The large amount of
combinations gave rise to diverse RS curves and a diversity of RS mechanism were proposed.
Interestingly, before Waser’s group demonstrated in 2006 the relation between the RS in VCMs, the
oxygen content and the change of valence,[25] many groups explained the RS phenomena in terms of a
space-charge-limited current (SCLC) transport. After 2006, oxygen ions (or vacancies) are progressively
incorporated into the phenomenological explanation of RS. In the following, a summary of the state-of-
the-art in RS on LSM is provided, with emphasis on the different electrode/LSM combinations and the
mechanisms proposed.
15
Let us start by the beginning, the parent compound LMO, which presents an orthorhombic phase (see
Section 1.3.1) and a large resistivity. The first RS studies on LMO were conducted by Xu et al. in 2012
using Pt/LMO/Nb:STO[85] and by Jin et al. in 2013 in Au/LMO/Nb:STO[86] devices. In both cases, when
using noble electrodes, the change of resistance was ascribed to the modification of the Schottky barrier
at the metal/LMO interface due to an increase of oxygen deficiency in LMO. Our research group
(Nanoionics, LMGP) also studied RS in LMO.[83], [87], [88] Using Au electrodes, Rodriguez-Lamas et
al.[83], showed promising filamentary-type RS in semiconductor-industry-compatible LMO films grown
on Pt/Si substrate. Meunier et al.[87], [88] demonstrated that RS in LMO was due to the change of its
oxygen content. Operando HAXPES (hard X-ray photoelectron spectroscopy) was carried out in
TiN/LMO/Pt cell, determining the key role of oxygen in RS phenomena: during RESET, reduction
(oxidation) of LMO (TiN electrode) occurs, while the reverse redox reaction takes place for the SET.
On another work from our group done using conductive atomic force microscopy (C-AFM) on planar
LMO/Au devices, the oxygen was suggested to remain adsorbed on the film surface after RESET, while
the reincorporation of it would occur during SET.
Concerning Sr-substituted manganites (LSM), most of the research has been done on 30% and 33%
LSM. There are fewer studies for 20% LSM and, to the best of my knowledge, there is only one
published work on the RS behavior of LSM 50%, even though it is not exhaustive. A collection of
relevant works of RS using LSM with x=0.2 – 0.33 and 0.5, and the RS mechanism proposed is presented
Table 1.1. For LSM20, interface-type RS is typically observed as a result of (i) its larger conductivity
compared to LMO (prevents dielectric breakdown typical in filamentary-RS), and (ⅱ) the planar devices
(separation between electrodes in the scale of microns). The first hints of interface- and volume-type RS
were reported by Chen et al. in 2006.[89] Using a C-AFM on LSM20/Ir/Si stack, the authors observed
that, after applying a local electric pulse with a 30 nm tip, the change of resistance was extended around
100 nm away from the tip/film contact point. They further validated RS (they referred to it as electrical
pulse induced resistance-change, EPIR) by using a top Ag electrode and suggested that the drift of
oxygen vacancies induced a reversible change of carrier concentration, particularly at the interfaces.
Oxygen vacancies have also been suggested to be responsible of RS in other works on LSM20, in which
Puig’s group observed volume-type RS.[90]–[93] Using epitaxial CeO2-x/LSM20/STO devices, they
suggested the oxygen exchange between LSM20 and the top layer of CeO2-x acting as an oxygen
reservoir: the reduction (oxidation) of the films occurs during RESET (SET) and is not influenced by
pO2 in the atmosphere neither in air or vacuum.[90], [94] They also studied LSM20 without CeO2-x oxygen
reservoir layer. In this case, they suggested that the oxygen exchange occurs with the surrounding
atmosphere, as the HRS/LRS window decreases in air and a larger shrink is observed at low pO 2. The
16
Table 1.1. Collection of relevant resistive switching studies using La1-xSrxMnO3-δ (LSMx, being x the strontium
substitution percentage), for x=20-33 and 50 %, highlighting the type of switching and the mechanism proposed.
The term ‘probe’ corresponds to electrically biased probe in direct contact with the film, whereas the term ‘tip’ is
restrained to conductive-AFM (C-AFM) in which the cantilever is biased. VCM stands for Valence Change
Mechanism; and SCLC stands for space-charge-limited-current. Reproduced from Moncasi et al.[171].
Device Type of Mechanism

Thin film Texture Substrate Device Refs.
configuration switching proposed
LSM20 Poly.a) Irb) Vertical Ag/LSM20/Ir Interface VCM: change in  [173]
LSM20 Poly.a) Irb) Vertical Tip/LSM20/Ir Interface VCM: change in  [173]
CeO2-x/ [90],
LSM20 Epitaxial STO Planar Volume VCM: MIT
LSM20\CeO2-x [174]
[84],
Probe/
LSM20 Epitaxial STO Planar Volume VCM: MIT [90],
LSM20\Probe
[174]
Tip/LSM20 [84],
(grounded) [93]
LSM20 Epitaxial STO Planar Pt/LSM20\Pt Interface VCM [95]
Tip/LSM30 [93],
(grounded) [97]
Conducting
LSM30 Epitaxial LAO Planar Au/LSM30\Au Filamentary multifilaments [96]
along dislocations
Ti/SCMOd)/ Modification of
LSM30 Epitaxial SROc)/STO Vertical Interface [21]
LSM30/SROc) Schottky barrier
Interface-
LSM30 Poly.a) Pt/Ti/SiO2/Si Vertical Probe/LSM30/Pt VCM [104]
filamentary
LSM30 Poly.a) Pt/Ti/SiO2/Si Vertical Ag/LSM30/Pt Interface SCLC [102]
LSM30 Poly.a) Pt/Ti/SiO2/Si Vertical Probe/LSM30/Pt Interface VCM: change in  [103]
LSM33 Epitaxial STO Planar Ag/LSM33\Ag Interface SCLC [22]

Probe/ VCM: topotactic
LSM33 Epitaxial Nb:STO Vertical Volume [20]
LSM33/Nb:STO phase transition
LSM50 Poly.a) Pt/Ti/SiO2/Si Vertical Probe/LSM50/Pt Interface VCM [104]
a) b) c) d)
Polycrystalline; Ir: Ir/TiN/SiO2/Si; SRO: SrRuO3; SCMO: Sm0.7Ca0.3MnO3.
modification of the oxygen content in LSM20 depends on the magnitude of the applied voltage. At large
enough applied voltage, oxygen vacancies can be created across the film thickness, leading to a volume-
type RS. The authors showed that the formation of oxygen vacancies in LSM implies a strong
decrease in carrier concentration and an increase of resistivity (ρHRS/ρpristine=100 at RT). Oxygen
𝐻𝑅𝑆
vacancies are also responsible for the decrease of MIT temperature of the HRS (𝑇𝑀𝐼𝑇 =200 K, and
𝑝𝑟𝑖𝑠𝑡𝑖𝑛𝑒
𝑇𝑀𝐼𝑇 =315 K), proving that their RS relies on the intrinsic properties of LSM.
17
Luo et al.[95] highlighted the crucial role of oxygen vacancies in the pristine LSM20 device to trigger RS
in planar devices. LSM20 films grown on STO were annealed at different oxygen partial pressure to
induce oxygen deficiency. While stoichiometric Pt/LSM20\Pt (δ=0) devices exhibit ohmic
characteristics, a pinched hysteresis loop was observed in the oxygen-deficient (δ<0) films, for which
the change of resistance increases with ambient temperature. The authors associated this phenomenon
to electro-thermal enhanced oxygen migration that results in the accumulation of oxygen
vacancies at the Pt/LSM20 interface.
LSM with x=0.3-0.33 has been largely studied for RS as it also presents colossal magnetoresistance
properties.[22], [93], [96]–[103] As previously mentioned, many studies date before 2006 and the proposed
mechanism is space-charge-limited-current (SCLC) transport. In 2006, Xie et al.[22] observed that RS
was easily triggered in rough LSM33/LAO films (roughness Sq=0.5 µm) using pairs of Ag electrodes.
They suggested a SCLC mechanism involving the formation of an insulating layer at the Ag/LSM33
due to the presence of surface states. In 2008, Lau et al.[96] reported multilevel states in LSM30/LAO
films using two Ag electrodes after a forming process to increase the resistance. They proposed the
formation of multifilaments, even if no abrupt increase of current, characteristic of filamentary type,
was observed during SET. In 2008, Wu et al.[102] measured polycrystalline Ag/LSM30/Pt vertical
devices. The I(V) curves show a gradual transition between resistance states and they attributed it to a
trap-controlled SCLC mechanism. In the three exposed cases, the use of Ag electrode may hinder the
correct interpretation of RS, as Ag is an active electrode that can diffuse into the oxide matrix, and hence
is typically used in ECMs.
In 2020, Gonzalez-Rosillo et al.[93] carried out a comparative study between LSM30 and LSM20 films
grown on STO, where both films exhibited volume-type RS related to oxygen vacancies (as explained
above). They concluded that due to the strong correlation effects in manganites, a better memristive
performance can be obtained in materials with a MIT temperature lying close to operating temperature;
in their case, LSM20 showed smaller operating voltage working at RT.
It has also been shown that RS switching may involve the nucleation of new phases. Yao et al.[100]
studied RS in LSM33/Nb:STO by in situ TEM (transmission electron microscopy) in vacuum
conditions. The authors observed the nucleation and growth of an oxygen vacancy-ordered
brownmillerite phase, corresponding to La0.67Sr0.33MnO2.5, due to the applied voltage. While the
brownmillerite phase can be achieved by thermal treatment in reducing atmospheres, they suggest that
its nucleation in devices is thanks to the electro-thermal oxygen vacancy migration. The brownmillerite
phase, with a high resistivity, can be reversed back to a perovskite structure with lower resistivity.
Following studies by the same group demonstrated the change of resistance by nanomechanical
stress.[101]
18
Finally, Choi et al.[104] studied polycrystalline Au probe tip/LSM/Pt vertical devices for x=0.1, 0.3 and
0.5. The devices exhibit the same RS cycling sense and was attributed to the formation of an oxygen-
deficient layer that increases the resistivity. The operating window was increased from x=0.1 to 0.3
thanks to the concomitant formation of oxygen vacancies upon Sr2+ substitution. However, the
unexpected narrowing of the operation window observed for x=0.5 was ascribed to the poor crystallinity
of the film.
Noteworthy, the HRS and LRS in LSM (and other ReRAMs) can be non-ohmic and ohmic, respectively,
or both show the same behavior depending on the mechanism responsible of the RS. Many RS
mechanisms in LSM have been proposed in the literature. However, many of the mechanisms rely on
the electrically-assisted change of oxygen content to modify the resistivity properties of the material,
and so the resistance state of the device. Hence, oxygen content and mobility within the LSM lattice are
likely to play a major role in the memristive properties.
Role of the electrodes
Furthermore, the memristive response in VCMs can also be modified by changing the electrodes.[105] In
particular, the work function and the oxygen affinity play an important role. Current-voltage
characteristics with ohmic behavior are expected when a p-type material, such as LSM (x<0.5), is put in
contact with a large work function metal as Pt (5.9-6.2 eV[106]). On the other hand, if the p-type material
is in contact with a low work function metal as Ti (4.3 eV[107]), the band will bend forming a Schottky
barrier with rectifying behavior. A schematic picture of the band diagrams was provided by Maas,[108]
and reproduced in Figure 1.12. In the case of a Ti/LSM interface, the depletion or decrease of carriers
at the interface of the oxide, expanding for a width Wdep, will localize the voltage drop at the interface.
The Schottky barrier can be further modified by formation of interface defects as reported by Sawa et
al.[109] The authors studied RS in p-type PCMO with x=0.3 (PCMO30) using two metals with similar
work function, Al and Ti. The Ti/PCMO30 interface showed a larger rectifying behavior than when Ag
was used. This difference was ascribed to the larger oxygen affinity of Ti, which partially oxidizes it
into TiOx, reduces PCMO30 at the interface, and hence, increases the width and/or height of the Schottky
barrier. Oxidizable electrodes can accommodate different oxygen content and thus act as an oxygen
reservoir.[23], [90], [109] Asanuma et al.[110] studied Ti/PCMO(x=0-0.8)/SrRuO3 memristive devices and they
corroborated, using microscopy and spectroscopic techniques, that the Ti electrode uptakes oxygen from
the oxide. Indeed, pristine devices presented an amorphous oxygen deficient TiOx interlayer that grows
upon voltage operation. The largest HRS/LRS ratio was obtained in devices with x=0.4, suggesting that
a moderate hole-carrier concentration is required to trigger RS while heavy-doping suppress it. As just
19
a) High work function metal (ex: Pt, Au) b) Low work function metal (ex: Ti, Ag, Al)
Figure 1.12. Schematic of the band diagrams and charge distribution of a p-type semiconductor oxide (such as
LSM, x<0.5) before and after being put in contact with a) a high and b) low work function metal, φM. When put in
contact with a high (low) φM, the bands of the semiconductor bend upwards (downwards) where the accumulation
(depletion) of charge carriers at the interface results in current-voltage characteristics with ohmic (rectifying)
behavior, as shown in the insets. Φj =q·φj is the energy required to extract a charge from the semiconductor (j=SC)
or from the metal (j=M); χSC and ISC are the electron affinity and ionization potential of the semiconductor,
respectively; Egap is the difference energy difference between the minimum of the conduction and maximum of
the valence band (EC and EV, respectively) of the semiconductor; Ef,j (with j=SC, M) is the Fermi energy; Ei is the
intrinsic Fermi level of the undoped semiconductor; ΦB =q·φB is the Schottky barrier height; VBI is the built-in
potential; Wacc and Wdep are the accumulation and depletion width, respectively. Adapted from [108].
seen, the oxygen affinity of the electrodes also plays an important role where new layers can be created
in the system during fabrication. The oxygen affinity can be quantified through the free energy formation
of the metal oxides, also called Gibbs energy (ΔG) where the more negative the value, the more stable
20
Figure 1.13. Temperature dependence of the free energy a) formation of some noble and transition metal oxides,
and b) formation of sub-oxides in TiO2. Figure 1.13a reprinted from [175], with the permission of AIP Publishing.
Figure 1.13b reprinted with permission from [176].
the oxide is. On the contrary, positive values indicate that the oxide is unstable. Figure 1.13 presents
temperature dependence of the free energy formation of a) some noble and transition oxides and
b) formation of sub-oxides in Ti. Metallic Ti present a large oxygen affinity (ΔG= -860 kJ/mol at 298 K)
which suggest that spontaneous oxidation will occur under the presence of oxygen. On the other hand,
the Gibbs energy of MnO (II) (-730 kJ/mol at 298 K) is slightly larger than that of Ti. Competitive
reactions can occur when oxides and metals are put in contact.
Even though there are many publications on RS in LSM, the role of the microstructure in RS has never
been directly assessed. Given the large number of variables present in literature (choice of electrode, its
size, film thickness, deposition techniques, electrical history…) a generalized evaluation from them is
complicated. Therefore, this thesis aims to fill the gap existent in the literature.
21
Scope and organization of the thesis

The goal of this PhD thesis was to contribute to the understanding of the underlying physico-chemical
mechanisms of resistive switching in valence change memories, and to elucidate the role of the micro-
and nanostructure in the performance of the devices. In particular, two compositions of the strongly-
correlated perovskite oxides La1-xSrxMnO3 (LSM), with x=0.2 (LSM20) and x=0.5 (LSM50) were
studied in this thesis, and the main properties are presented in Section 1.3. The experimental procedure
for the growth of thin films, the microfabrication of devices and their electrical characterization is
presented in Chapter 2. The material was implemented as a thin film using a MOCVD (metal organic
chemical vapor deposition) reactor for which first, the deposition conditions were optimized to obtain a
large homogeneous deposition area, and next, the precursor solution composition was adjusted to grow
LSM20 and LSM50 films (Chapter 3). In order to establish a relation between structure in LSM20 and
the device performance (Chapter 4), the following strategy was followed. First, epitaxial LSM20 films
were grown on two different single crystal substrates; one with a small mismatch, and another one with
larger mismatch to induce strain and structural defects. Based on the experimental results, comprising
film characterization and electrical characterization of planar Ti/LSM20\Pt devices, a phenomenological
mechanism is proposed. The mechanism relies on a reversible oxygen exchange between the film and
the electrode, which would act as an oxygen reservoir. The oxygen exchange would be enhanced by the
presence of extended structural defects, i.e. dislocations, and therefore, the device exhibit better
performance in terms of multiple resistance states, cyclability, reproducibility and retention. Finally,
polycrystalline LSM20 films, with a large density of grain boundaries (another type of structural
extended defect) were grown on Si-based substrates. The devices built on these films could be described
with the same resistive switching phenomenological model proposed before. On the other hand, the
performance of LSM50 as a memristive material was studied (Chapter 5). For the first time,
reproducible and reliable resistive switching is reported in epitaxial LSM50 films with very low density
of structural defects. A similar electrical characterization protocol than the one used for LSM20 was
followed, revealing different yet very promising memristive characteristics that were related to the
nature of the LSM50 oxide. In parallel to the device characterization, the effect of oxygen on LSM thin
films structural and electrical properties was also studied for a better comprehension of the performance
of the devices (Chapter 6). Finally, the main highlights and conclusions of this work are summarized,
and next steps and prospects are suggested (Chapter 7). The description and details of the main
techniques and equipment used during this thesis is presented in Annex A.
These results highlight the importance of extended defects in the functional properties of the materials
for memristive devices, for which microstructure engineering arises as an alternative route (or
complementary) to classical cationic substitution.
22
Experimental procedure
Pulsed Injection Metal Organic Chemical Vapor Deposition ............................................ 23

Device micro-fabrication........................................................................................................ 25
Device electrical characterization ......................................................................................... 28
This chapter is dedicated to the experimental methodology and is divided into the three main steps
carried out along the thesis: the growth of thin films, the fabrication of devices and their electrical
characterization. For each of them a brief introduction is given followed by the experimental details.
The characterization equipment used in the thesis is described in Annex A.
Pulsed Injection Metal Organic Chemical Vapor Deposition

Metal Organic Chemical Vapor Deposition (MOCVD) is a chemical process, involving vapor phase that
uses metal organic precursors for the growth of films on substrates. Technological effort focused on the
delivery system of precursors to allow the growth of functional complex oxides for which the precursors
present thermal stability problems and/or low vapor pressures.[111] Among them, Pulsed Injection Metal
Organic Chemical Vapor Deposition (PI-MOCVD) allows using liquid precursors that are flash
evaporated when injected into the evaporation chamber.[112] The growth of complex oxides with fine
variation of the composition has been achieved using MOCVD and PI-MOCVD.[83], [111], [113]–[115]
In this thesis, La1-xSrxMnO3-δ (LSM) films were grown using a PI-MOCVD JIPELEC reactor. As for
other CVD systems, this reactor allows the growth of films over a large area with homogeneous
composition and thickness. Whereas the flow rate of the precursor in conventional sublimation CVD
systems is determined by the source temperature and the gas flow, in PI-MOCVD, it only depends on
the feeding rate of precursor into the chamber. The injection of precursors is done through a high speed
electrovalve controlled by a software that allows setting the number of pulses, pulse width (opening
time) and frequency. Hence, under fix depositions conditions, the thickness is determined by the number
of pulses. This system allows the growth of multicomponent films, such as LSM films (with three
different cations), using one single liquid source kept in inert atmosphere and provided that all the
precursors are soluble and compatible with the same solvent. In case of chemical incompatibility,
23
Chapter 2: Experimental procedure
different sources can be used for different precursors as long as they possess similar temperature
window.[116]
In order to grow LSM, the solution was constituted of La(thd) 3, Sr(thd)2-tetraglyme, and Mn(thd)3,
[thd=2,2,6,6,-tetramethyl-3,5-heptanedionato] precursors, purchased from Strem Chemicals, and
dissolved in m-xylene from Alfa Aesar. The precursor cationic concentration was 22.5 mM while the
relative concentration of the cations permit the growth of LSM films with different strontium
substitution, as discussed in Section 3.2. The solution was prepared right before each deposition and
stirred for 1 h at 40 °C before being transferred into the flask compatible with the electrovalve/reactor.
Figure 2.1 presents a sketch of the reactor used in this work and the main steps involved in the growth
by CVD. The solution of precursor is contained in a glass flask at room temperature and in inert
atmosphere. The injection into the evaporation chamber is done through a high speed electrovalve thanks
to the overpressure created in the solution flask and the vacuum in the evaporator, held between 230 °C
and 250 °C. Then, the microdoplets of the liquid solution experience a flash evaporation. The newly
formed precursor gas is transported by a controlled Ar flow along the heated line as the temperature
increases. On the bottom of the reactor, the precursor and gas flow are mixed with a controlled flow of
oxygen. Finally, they reach the reaction chamber which is heated at the deposition temperature, i.e. at
650-750 °C. The substrates on which the growth of the films will occur are fixed to the sample holder
in the center of the reaction chamber. In this particular reactor, the substrates are up-side down facing
the flow of gas precursor. As depicted in Figure 2.1, the growth involve several steps occurring at the
Figure 2.1. Schematic of the PI-MOCVD JIPELEC reactor and the main steps involved in the growth of films by
CVD. The reactor is constituted of the evaporation chamber, where the precursor is injected, a transport line and
the reaction chamber where the growth takes place.
24
2.2 Device micro-fabrication
boundary region: adsorption of precursors, gas-solid reaction, nucleation and growth of LSM films. The
by-products of the reaction, gas solvent and precursor that did not react are carried away from the
chamber by the gas flow and they are collected in a liquid N2 cooled trapped for later proper disposal.
The reaction chamber is composed of a quartz tube surrounded by a furnace, i.e. hot wall reactor. The
pressure within the system is controlled by a butterfly valve placed between the reactor and the vacuum
pump (not shown). This reactor offers two gas flow configurations: (i) flow from bottom to top (as just
described and depicted in Figure 2.1), and (ⅱ) flow from top to bottom. The first, is used during the
growth of the films when precursor is injected, as the substrates are directly exposed to the gas front.
The second one is used before and after the growth when no precursor is injected to prevent projection
of particle on the substrates or films. To switch between the two configurations the inlet of gases is
changed with a valve (inlet bottom or top for (i) and (ⅱ), respectively), and with a pneumatic valve for
the location of the vacuum exhaust (top or bottom for (i) and (ⅱ), respectively). Further details are
provided in Section 3.1 and Section 3.2, along with the discussion and characterization of the films
deposited. The growth parameters are presented in Table 3.1.
Device micro-fabrication
The microfabrication of the memristive devices was carried out in the cleanroom from PTA (Plateforme
Technologique Amont) in Grenoble. The main steps are summarized in Figure 2.2 and described below:
(1) First of all, the samples are cleaned by immersion into three beakers containing acetone,
isopropanol and deionized water, respectively, while doing a gentle agitation. Then, some
isopropanol is poured on top of the film to help remove the excess of water, and any solvent left
on the surface is dried with a N2 gun.
(2) Next, the photoresist is spin coated. In our case and for 10×10 mm2 substrates, we use 200 µL of
a positive resist (AZ1512) which is spread over the surface at 4000 rpm for 60 s (acceleration:
2000 rpm/s). The photoresist is cured on a hot-plate at 100 °C for 90 s to remove the excess of
solvent and harden the resist. The approximate thickness of the photoresist layer is 1 µm.
(3) The motifs of the mask are transferred to the sample by laser assisted lithography (µPG 101 from
HEIDELBERG Instruments) at a power of 15 mW. The irradiated areas of the resist become
soluble to the resist developer.
(4) In the development step, the film is immersed and agitated in the appropriate resist developer (a
diluted base in our case) for 15 s, followed by immersion and agitation in deionized water for 45 s
to stop the reaction. After this step, the resist that was exposed to the laser is removed and the
LSM film underneath is revealed.
25
Figure 2.2. Workflow of the micro-fabrication of memristive device using a positive photo-resin. Acetone,
isopropanol (IPA) and deionized water are used to clean the film. In order to fabricate memristive planar devices
with two different electrodes facing each other, the cycle 1 to 6 has to be repeated twice.
(5) During the metal evaporation a directional layer of metal is deposited on top of the sample: on
top of the resist where there is resin and on top of the film for the regions that were exposed to
light and then developed. The system used is a MEB550 from Plassys. Prior the metal
evaporation, the surface of the sample is gently etched with Ar+ at 250-300 V to remove organic
residuals in the developed areas. The metallic sources are evaporated by an electron gun at 10 kW
and the thickness is controlled using a quartz microbalance.
(6) Finally, the resin is removed completely in the lift-off step to reveal the desired pattern of contacts.
The exact procedure is highly dependent on the resist as well as the material/film adherence. In
particular, the resist AZ1512 is dissolved in acetone. For Ti and Pt electrodes, the sample is fully
immersed in a beaker with acetone under soft sonication for 3-4 min and further hand-agitation
may be required to remove completely the resist and potential metallic interconnected regions.
As basis, gentle agitation to produce a “slow” dissolution of the resin is required to ensure the
integrity of the metallic contact, especially when positive resists are used. At the end, the sample
is cleaned as described in step 1.
A cycle is completed after doing the steps 1 to 6 for which the desired motifs are fabricated on the
samples. These six steps would allow for example the fabrication of vertical devices provided an
electronic conducting substrate. However, more complicated device architectures require multiple
26
2.2 Device micro-fabrication
cycles. In this work, we use insulating substrates, so two different electrode materials are deposited on
top of the film to build planar memristive devices. Hence, two microfabrication cycles are required, one
for each electrode. A 100 nm thick Pt layer is used as inert electrode while 10 nm thick layer of Ti is
used as active electrode. A Pt layer is deposited on top of the Ti electrodes during same evaporation
process (change of target) to prevent the atmospheric oxidation of Ti. The thickness of this layer is
around 90 nm although thinner layers have been evaporated in the devices measured in the synchrotron
(Section 6.2).
The lithography mask used for fabricating the memristive devices on samples of 10×10 mm2 is presented
in Figure 2.3a. The mask display the arrangement of the electrodes on the sample where three main
regions are defined. (1) An array of rectangular contacts of Pt (2×0.2 mm2) separated at distances of
1000, 500, 200, 100, 50, 20, and 10 µm for Transmission line measurements (TLM) (see Section 2.3 for
more information about the measurements). The other two regions concern the planar Ti/LSM\Pt
memristive devices. Region (2) allows to study the effect of the length of the devices, L, being 50, 100
and 200 µm, where square electrodes with an area A=2002 µm2 are used. On the other hand, region (3)
is designed to study the effect of the area, being A=502, 1002, 2002, 3002, and 4002 µm2 while L=50 µm.
Figure 2.3b displays a sketch of a Ti/LSM\Pt planar memristive device where the two electrode
geometric variables are emphasized.
Figure 2.3. a) Lithography mask used for the fabrication of memristive devices displaying the arrangement of the
electrodes on a sample of 10×10 mm2. b) Sketch of a memristive device. The two metals that build the memristive
devices (Pt in grey and Ti in yellow) are split into two digital “layers”. The mask is divided in three regions:
(1) Array of rectangular (2×0.2 mm2) Pt contacts for TLM measurements. (2) Memristive devices of different
length, L, with square contacts of area A=50×50 µm2, and (3) Memristive devices with distance between contact
L=50 µm and different contact size.
27
Device electrical characterization

The electrical characterization of the devices was conducted using two biased tungsten probes mounted
on micromanipulators, and a source measure unit (Keithley 4200) in a voltage-controlled mode where
the bias was applied to the Ti electrode and the Pt electrode was grounded.
The majority of the measurements were standardized to allow the comparison of the devices with
different contact area (A) and length (L). Let us begin by defining the basic unit of the measurements:
the cycle. A writing cycle comprises the change of applied voltage according to the following sequence:
0 V +Vamp  0 V  -Vamp  0 V, where Vamp is the voltage amplitude of the writing process. First,
the device is biased in positive polarity by a dual voltage sweep: the voltage is increased to a maximum
and decreased back to 0 V. After this, the resistance is measured by a read positive voltage sweep (Vread).
Then, a dual voltage sweep is carried out in negative polarity followed by a read positive voltage sweep
to measure the resistance of the device. The voltage sweep rate is defined by power line cycles and is
approximately 0.1 V/s. Figure 2.4a presents a schematic of a cycle composed by the writing and read
sweep. The change from a low resistance state (LRS) to a high resistance state (HRS) is referred as
RESET process, while the reverse switch (HRSLRS) is known as SET. In the devices presented in
the thesis, the RESET occur at positive voltage, and at negative voltage for SET process. In addition,
most of these devices show a gradual transition between the resistance states and no threshold switching
voltage can be defined. Therefore, and for simplicity, the transition voltage of the RESET (VRESET) and
SET (VSET) processes are referred to as the voltage amplitude taking into consideration the polarity. The
procedure described here correspond to a symmetric writing process where the |VRESET|=|VSET|, yet if
asymmetrical writing (|VRESET|≠|VSET|) has been also studied. After one cycle, the device has switched
from LRS to HRS, and again to LRS if the working voltage is adequate. In order to read the resistance
and not alter it, the following condition is required: Vread << Vamp. As presented in Chapter 4 and
Chapter 5, the writing voltages are comprised between 2-20 V and the reading voltage was set to
100 mV.
The second main standardized measurements is the initialization protocol. This protocol has a double
role: study the evolution of the resistance states, i.e. HRS and LRS, as function of the writing voltage
amplitude, and homogenize the active materials to obtain reproducible performance. In the initialization
protocol the voltage amplitude is increased every 5 cycles up to a maximum amplitude, and then is
decreased in a symmetrical manner. A scheme of the initialization protocol is provided in Figure 2.4b.
Yet if we refer to the initialization protocol as the whole procedure described above, the device is
considered initialized after the cycles carried out at largest voltage amplitude. After this, reproducible
performance is obtained operating the device at lower voltages.
28
2.3 Device electrical characterization
a) b)
V V
+Vreset Vamp << |Vread | 5 cycles
+Vread
time time
-Vset
Figure 2.4. a) Schematic of a sweep voltage cycle where the amplitude of the writing voltage (Vamp), i.e. RESET
voltage (VRESET) and SET voltage (VSET), are smaller than the reading voltage. In symmetrical voltage operation:
Vamp =|VSET |=|VRESET|. b) Scheme of the initialization process where the voltage amplitude is first increased and
then decreased every 5 sweep cycles.
Transmission line measurements (TLM)
TLM allows to obtain electrical properties of the film and the electrode/film interface by measuring the
resistance between pairs of electrodes separated at different distances. For example, TLM measurements
are used in Section 4.1. Figure 2.5 shows a) a typical array for TLM measurements and b) a cross-
section of the film with two electrodes where the current lines are drawn in red. The mathematical
derivation to extract the electrical properties is given below.
The total resistance of the metal/semiconductor\metal device can be expressed as the sum of serial
resistances, as in Equation 2.1:
𝑅𝑇 = 2𝑅𝑀 + 2𝑅𝑐 + 𝑅𝑠𝑒𝑚𝑖𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟
Equation 2.1
where RM, RC, and Rsemiconductor are the metal (Pt in our case), contact and semiconductor (LSM film)
resistance. The resistance of the metallic contacts is negligible compared to that of the semiconductor.
Expressing the resistance of the film in geometrical parameters, Equation 2.1 turns into Equation 2.2:
29
a) b)
LT L LT
l = 2 mm
L (variable) LSM t W=200 µm LSM W

substrate
Figure 2.5. a) Scheme of contact array for TLM and b) cross-section where the current lines (red) are drawn to
show the current crowding effect.
𝐿
𝑅𝑇 = 2𝑅𝑐 + 𝑅𝑆
𝑊
Equation 2.2
where RS is the sheet resistance of the film, L is the length between the two metallic contacts, and W is
the width of the contacts. The contact resistance can be expressed in terms of specific contact resistivity,
ρc (Ω·cm2), according to Equation 2.3.
𝜌𝐶
𝑅𝐶 =
𝐿𝑇 𝑊
Equation 2.3
where LT is the transfer length defined as the average distance that a charge carriers travels underneath
the metal contact before being injected in configurations experiencing current-crowding effect. The
current-crowding effect is the non-homogeneous distribution of current lines which accumulate in the
closer edges of the two facing electrodes. The transfer length can be expressed as a function of the
contact resistivity and sheet resistance (Equation 2.4)[117]:
30
2.3 Device electrical characterization
𝜌𝐶
𝐿𝑇 = √
𝑅𝑆
Equation 2.4
Combining Equation 2.3 and 2.4, the total resistance can be expressed as in Equation 2.5.
𝑅𝑆 𝑅𝑆
𝑅𝑇 = (𝐿 + 2𝑅𝐶 ) = (𝐿 + 2𝐿 𝑇 )
𝑊 𝑊
Equation 2.5
Hence, when the resistance measured between two pairs of electrodes of the same width is plotted as a
function of the distance between electrodes (RT vs L), it is possible to trace a linear regression where:
 the slope, m=RS/W,

 the y-axis intercept, a=2×RC,
 the x-axis intercept, b=2×LT.
31
Optimization of the growth conditions of
La1-xSrxMnO3-δ thin films
Geometrical optimization of the reactor............................................................................... 33
Deposition optimization of LSM thin films with different composition ............................ 36
Correlation between the precursor solution and the film composition ........................ 36
Effect of the LSM composition on the structural properties........................................ 41
Morphology and microstructure .................................................................................. 48
Raman of the LSM20 and LSM50 films ..................................................................... 51
Effect of the composition on the LSM films electrical properties ............................... 53
Conclusions ............................................................................................................................. 55
In order to obtain suitable La1-xSrxMnO3-δ (LSM) thin films for memristive applications by chemical
vapor deposition (CVD), two studies are required: optimization of the reactor and optimization of the
film’s composition. Hence, in the first section of this chapter, the optimization of the reactor to achieve
a homogeneous deposition over a large area is presented. The second section tackles the La1-xSrxMnO3
layers, in which several films are grown on diverse substrates to determine the relation between the
composition of the film and the solution of precursors used. These films are then characterized and
further relations are established between composition, growth conditions (e.g. temperature),
microstructure and electrical properties. From these results, the optimal precursors’ solution
composition and deposition temperature are selected to grow the thin films presented in the following
chapters of this thesis.
Geometrical optimization of the reactor

The goal of the first optimization is to obtain a homogeneous deposition area large enough to
simultaneously deposit on several functional (10×10 mm2) substrates. The homogeneity is defined in
terms of thickness variation and density (the films should be continuous and dense). The pulsed-injection
metal-organic chemical vapor deposition (PI-MOCVD) reactor used in this thesis was previously
optimized to grow undoped lanthanum manganite (LaMnO3, LMO).[83], [118] The deposition parameters
used for the growth of LMO were taken as starting point for the optimization of the deposition of
33
Chapter 3: Optimization of the growth conditions of La1-xSrxMnO3-δ thin films
Table 3.1. Deposition parameters used for the growth of LSM after the geometrical optimization of the reactor.
The Ar flow (400 sccm) is exclusively added from the evaporator.
Parameter Optimal value, growth of LSM
Substrate temperature 700 °C a)
Evaporation 210 °C - 250 °C

temperature
57% Ar (400 sccm) +
Carrier gas 43% O2 (300 sccm)
Total pressure 5 Torr
Solution concentration 22.5 M
Number of pulses 15.000 a)
Pulse frequency 2.5 Hz

a)
Note the growth of LSM films for functional devices is done at 675 °C, as discussed in Section 3.2, and the
number of pulses is adjusted to fit the requirement of the devices (typically 1800 to 5000 pulses).
strontium substituted lanthanum manganite (La1-xSrxMnO3-δ, LSM). Table 3.1 summarizes the optimal
parameters for the growth of LSM. In this first optimization, the relative precursors cationic
concentration was [(La+Sr)/Mn]=2.0 and [Sr/(La+Sr)]=0.50, which was estimated to yield slightly Sr-
substituted LSM films based on the previous work done on LMO.[118] As discussed later on this chapter,
this solution composition is close to LSM films with a composition of x=0.2.
The parameters allowing for the optimization of the deposition area are: (i) the position of the substrate,
as referred to the reactor walls; (ⅱ) the pressure; and (ⅲ) the gas flows. Several depositions were carried
out until a homogeneous deposition over a 2 inches diameter area was successfully obtained. A scheme
of the reactor highlighting the parameters that were modified is presented in Figure 3.1a. A comparison
by pairs of the films grown after modifying the main parameters is shown in Figure 3.1b. Figure 3.1c
presents a comparison between the first and last deposition of the geometrical optimization, and
highlights the 2 inches diameter homogeneous area.
Thin films may present different colors as a result of light interference related to thickness. The smallest
thickness corresponds to light brownish colors, which evolve towards dark brown, purple, blue, green
and yellow as the thickness increases. After this, the sequence of colors repeats with increasing the
thickness.
In order to obtain a homogeneous deposition, two types of diffusers were tested to homogenize the flow
of precursors, namely a mesh and a solid mask (see inset in Figure 3.1a). Nonetheless, both diffusors
34
3.1 Geometrical optimization of the reactor
a) Bare b)
Flow diffuser X-Y displacement Double Pressure in chamber
Mask
• Holder position (XYZ)
• Temperature
• Pressure Double gas flow - Z displacement
Ar flow Mesh
• Number of pulses
• Opening time
• Frequency c)
• Flow diffuser
• Flow
O2 flow • Gas
Ar flow
composition
First deposition on Largest homogeneous Highlight of 2” homogeneous
a 4” Si wafer area achieved area achieved
Figure 3.1. Depositions on 4 inches silicon wafers for geometrical optimization of the reactor. a) Scheme of the
reactor highlighting the main parameters that affect the deposited film. b) Comparison of the films deposited on
silicon wafer after modifying each one of the previous parameters. c) Comparison between the first deposition and
the one that has shown the largest coverage, considered as optimized. Using these deposition parameters the
obtained film is practically homogeneous over 2 inches.
have metallic components, which heat up and capture part of the precursors, limiting the deposition on
the wafer itself. Using a diffusor, in the best of cases, the wafer showed a larger film coverage but greater
thickness inhomogeneity, as can be observed by the large range of colors of the film (Figure 3.1b).
Therefore, the use of a diffusor was discarded. Another approach to obtain a large area consisted in
exposing the wafer to a larger quantity of precursor. This was done by increasing the pressure inside the
reactor chamber, from 5 to 10 Torr. However, the deposition yield decreased, with practically no film
deposited, and thus the pressure was set back to its previous value, i.e. 5 Torr. The flow dynamics are
expected to be different by modifying the gas flow. Indeed, reducing both oxygen and argon flow to half
of its original value resulted in a visually thicker film with larger coverage. Eventually, the oxygen flow
was increased to 75% of the initial value. Yet, the film was not as symmetrical nor homogeneous as
required for the application.
Finally, the inlet of gases into the reactor chamber was brought closer to the wafer by adding a quartz
tube extruding out between 5 and 8 centimeters (corresponding to -z displacement in Figure 3.1b). After
this modification, the wafer was fully covered with a thicker film (completely blue) but two different
gas flow fronts seemed to be present. The two fronts were suggested to arise from a non-homogenous
mixture of the precursor in the gas mixture. Originally, the microdroplets of precursors are carried by a
100 sccm Ar flow from the evaporator and then, they mix with a 800 sccm (50:50 Ar:O2) gas flow in
the bottom of the reactor prior traveling up towards the reactor chamber. In order to have a more
homogeneous deposition on the wafer, the flow of argon was injected exclusively into the evaporator
chamber (400 sccm carrying the precursor), and only oxygen (300 sccm) was added from the bottom of
35
the reactor. As a result, a large coverage and homogeneous deposition is obtained. The last depositions
of the optimization consisted in the readjustment of the holder position until a 2 inches diameter
deposition area was located at the center of the wafer (Figure 3.1c).
From the whole study it can be concluded that the key parameters to obtain a large area were (i) the
relative position of the gas inlet and the holder, (ⅱ) the gas flow, and (ⅲ) the position at which the gas
is injected. The adjustment of these three parameters combined with the other deposition parameters
previously optimized for the growth of LMO (temperature profile of the heating line, approximate
deposition temperature, and pressure in the chamber) allowed to obtain a large area to deposit
simultaneously on several 10×10 mm2 substrates.
Deposition optimization of LSM thin films with different

composition
To obtain the two compositions selected for this thesis, namely La0.8Sr0.2MnO3 (LSM20) and
La0.5Sr0.5MnO3 (LSM50), the composition of the film has to be calibrated as a function of the
composition of the solution. The cationic ratio in the film was determined by electron probe
microanalysis (EPMA) and wavelength dispersive X-ray spectroscopy (WDS), and Rutherford
backscattering spectrometry (RBS) and nuclear reaction analysis (NRA). The description of the
techniques and the data analysis can be found in the Annex A.
Four different substrates were selected to allow for the identification of the elements in the film and
assess the film growth on substrates with different cell parameter: silicon (Si), silicon nitride (Si3N4),
lanthanum aluminate (LaAlO3, LAO) and magnesium oxide (MgO).
Correlation between the precursor solution and the film

composition
Using the deposition parameters described in Table 3.1, six precursor solutions with different cationic
ratio were prepared, as summarized in Table 3.2. The concentration of A-site ions (La and Sr) over B-
site cations (Mn), i.e. A/B ratio, used was chosen based on the optimized solution used to grow LMO
films deposited at 650-700 °C in a previous work.[118] The A/B ratio was fixed at 2.0 and the relative
concentration of the A-site atoms was systematically varied from 0.10 to 0.55 in steps of 0.15 (solutions
1 to 4 in Table 3.2). These films were grown at three different temperatures: 675 °C, 700 °C and 750 °C.
The cationic composition of the films (A-site substitution and A/B ratio) grown on Si and MgO
substrates were measured by EPMA-WDS. LSM20 is obtained using solution 4, but for a solution A/B
ratio of 2 the composition of LSM50 was not obtained from any of the prepared solutions. Hence, two
36
3.2 Deposition optimization of LSM thin films with different composition
Table 3.2. Relative precursor concentration in the solutions prepared to optimize the composition. The solutions
are indexed for simplicity reasons. The nearly stoichiometric LSM films are included in the last row, where the
number designates the Sr percentage in the film.
Cationic concentration of the solution Value
Solution number 1 2 3 4 5 6
A/B ratio [(La+Sr)/Mn] 2.0 2.0 2.0 2.0 2.3 2.4
A-site substitution [Sr/(La+Sr)] 0.10 0.25 0.40 0.55 0.80 0.90

Stoichiometric
LSM20 LSM50 LSM75
composition obtained
solutions richer in Sr and with a larger A/B ratio, i.e. 0.8, 2.3 and 0.9, 2.4, respectively, were prepared
(these correspond to solution 5 and 6 in Table 3.2). The cationic composition of the LSM thin films
grown at 700 °C on Si and MgO substrates using these solutions was characterized by RBS-NRA and
WDS. The results are summarized in Figure 3.2, where the solid symbols correspond to EPMA-WDS
and the open symbols to the RBS-NRA analysis. In a), b) the Sr-substitution is represented, and c), d)
corresponds to the A/B ratio of the films grown on Si and MgO, respectively. For cross-validation of
the techniques, the samples grown at 675 °C and 700 °C using the solution 4, were also measured by
RBS-NRA.
The Sr substitution in the films grown on Si and MgO (Figure 3.2a and b) increases with the strontium
content in the solution (solution 1-4) following approximately a second order polynomial relationship.
The thin film grown with solution 1, using the lowest strontium content, barely exhibits traces this
element. However, the film grown with solution 4 (Sr substitution in solution of 0.55), presents a
strontium substitution close to LSM20. Noteworthy, in this temperature range, the Sr content of the
films grown on both substrates does not depend on the temperature of deposition. The A/B ratio in the
film, shown in Figure 3.2c and d, decreases linearly with increasing Sr content in solution for both
substrates, and a value close to stoichiometric A/B=1 is achieved for the films grown using solution 4.
Therefore, solution 4 is chosen to grow films with stoichiometry close to La0.8Sr0.2MnO3-δ (LSM20).
The decreasing A/B ratio with increasing Sr content in solution suggests a synergic effect of the cations,
where preferential incorporation of equal parts of A (La+Sr) and B-site (Mn) cations is favored as the
Sr content in the solution increases. Surprisingly, the solution with the lowest strontium content
(solution 1) yields an A/B ratio larger than 1.5, whereas stoichiometric LMO was obtained using a very
similar solution, but without the strontium precursor.[118] The incorporation of Sr, and probably, the
changes introduced in the geometrical optimization, result in a large La enrichment in the film. No
influence of the temperature is observed on the A/B ratio in the films grown using solution 4; however,
37
Tdep 675°C Tdep 700°C Tdep 750°C Trend line from 700°C
Solid: WDS data Open: RBS-NRA data
a) 0.8 b) 0.8
LSM / Si LSM / MgO
0.7 0.7
Sr/(La+Sr) in the film

Sr/(La+Sr) in the film
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Sr/(La+Sr) in solution Sr/(La+Sr) in solution
c) 2.2 d) 2.2
LSM / Si LSM / MgO
2.0 2.0
(La+Sr)/Mn in the film
(La+Sr)/Mn in the film
1.8 1.8
1.6 1.6
1.4 1.4
1.2 1.2
1.0 1.0
0.8 0.8
0.6 0.6
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Sr/(La+Sr) in solution Sr/(La+Sr) in solution
Figure 3.2. Correlation of the a), b) strontium substitution; and c), d) A/B ratio of the LSM film with the atomic
ratio of the precursor solution for films grown on a), c) Si and b), d) MgO substrates at different deposition
temperatures, i.e. 675 °C, 700 °C, 750 °C. Values measured by EPMA-WDS are solid symbols, and by RBS-NRA
are open symbols. The A/B ratio in the solution was 2.0 (no background pattern), 2.3 (diagonal pattern) and 2.4
(horizontal pattern). The green-dashed line corresponds to the trend line from the data at Tdep=700 °C obtained by
WDS. The target ratios are indicated by horizontal lines: (a, b) strontium substitution 20 and 50%, and (c, d)
A/B=1.
the slopes of A/B|film vs Sr-substitution|solution (Figure 3.2c and d) are smaller for the films grown at
750 °C, i.e. the incorporation of Mn at high temperature is favored in Sr-poor films.
The trend line of the films grown at 700 °C and measured by WDS in Figure 3.2a and b, indicates that
50% strontium in the film would be obtained employing a solution with Sr/(La+Sr) of 0.8. However,
such solution would result in an A-site deficient film, around 0.75, (see trend line in Figure 3.2c and d).
Therefore, as mentioned earlier, manganite films were grown at 700 °C using two solutions with larger
A/B ratio and Sr-substitution (solution 5 and 6 in Table 3.2) and analyzed by RBS-NRA together with
the LSM20 films (open symbols in Figure 3.2). The cationic ratio obtained by WDS and RBS-NRA are
38
Table 3.3. Atomic ratio measured by EDS, WDS and RBS-NRA on the LSM films with composition close to
stoichiometric ABO3. The oxygen ratio in a stoichiometric perovskite is equal to 0.6, and thus the films are slightly
hyper-stoichiometric. The values in parenthesis are estimated from the ratio between WDS and EDS in the films
on the same substrate.
Composition
Film Solution Film
Tdep Parameter
(solution #) LSM / Si LSM / MgO
WDS RBS EDS WDS RBS
Sr/(La+Sr) 0.55 0.22 0.23 n/a 0.20 0.22
LSM20 (4) 675 °C A/B 2.0 1.05 1.09 n/a 1.07 1.18
O/ions -- 0.63 0.67
Sr/(La+Sr) 0.55 0.23 0.23 0.24 0.21 0.23
LSM20 (4) 700 °C A/B 2.0 1.08 1.12 0.97 1.09 1.16
O/ions -- 0.61 0.65
Sr/(La+Sr) 0.80 n/a 0.55 0.56 (0.51) 0.52
LSM50 (5) 700 °C A/B 2.3 n/a 0.97 0.91 (1.06) 0.96
O/ions -- 0.61 0.65
Sr/(La+Sr) 0.90 n/a 0.74 0.77 (0.69) 0.75
LSM75 (6) 700 °C A/B 2.4 n/a 0.88 0.82 (0.95) 0.93
O/ions -- 0.62 0.61
in good agreement thus, validating the results obtained by both techniques. The strontium substitution
shows a similar trend as the one obtained from solutions with lower A/B ratio (Figure 3.2a and b). The
films grown with solution 5 exhibit a Sr-substitution close to the target 50% and an A/B ratio close to 1.
Therefore, solution 5 allows obtaining films with stoichiometry close to La0.5Sr0.5MnO3-δ (LSM50).
The films grown using solution 6 present 75% strontium substitution and are slightly A-site deficient,
A/B=0.9, (LSM75).
Table 3.3 summarizes the Sr substitution, A/B ratio and oxygen ratio measured by EDS, WDS and RBS-
NRA of the nearly stoichiometric films, i.e. LSM20, LSM50 and LSM75, grown on Si and on MgO
substrates. The results obtained by the three techniques are consistent, especially for the polycrystalline
LSM/Si films. EDS was only performed on the LSM/MgO films as the semi-quantification of Sr in
LSM/Si films may introduce some error as the most intense lines of Si (Kα=1.73968 keV[119]) and Sr
(Lα=1.80565 keV[119]) used for quantification are close. The strontium rich films, i.e. LSM50 and
LSM75, were not measured by WDS, hence the values reported in parenthesis in the table correspond
to an estimation using the proportionality factor found between EDS and WDS in the LSM20/MgO
films (Table 3.3). Regarding the LSM/MgO films, the strontium substitution measured by EDS and
WDS is similar to that obtained by RBS-NRA. However, the A/B ratio for the LSM20 films is slightly
higher by RBS-NRA than by EDS and WDS, whereas for the LSM50 and LSM75 films it lays between
the values obtained by EDS and the estimated values of WDS. As explained in Annex A.1.1, RBS-NRA
is a non-destructive technique that allows depth profiling and quantification of light elements such as
39
oxygen. The oxygen ratio is expressed as the amount of oxygen over the total number of ions in the
perovskite (Equation 3.1), where each element represents the atomic percentage found by RBS-NRA.
𝑂
𝑂𝑟𝑎𝑡𝑖𝑜 =
𝐿𝑎 + 𝑆𝑟 + 𝑀𝑛 + 𝑂
Equation 3.1
In the stoichiometric perovskite with formula ABO3, Oratio is 0.6 (3/5). The LSM chemical formula
derived from RBS-NRA measurements is presented in Table 3.4 for both substrates. The oxygen
stoichiometry was calculated as 3+δ according to Equation 3.2. We note that small changes in Oratio
implies drastic changes in the oxygen stoichiometry, δ. Given that we could not retrieve error bars from
the estimation of the composition, the stoichiometric formulas presented in Table 3.4 are meant to
illustrate the trend rather than being an absolute truth.
3+ 𝛿
𝑂𝑟𝑎𝑡𝑖𝑜 =
5
Equation 3.2
As discussed by Töpfer and Goodenough,[120] due to the close packaging of the perovskite lattice, oxygen
excess (O > 3) is accommodated by the formation of cation vacancies rather than by oxygen interstitials.
Therefore, the perovskite formula commonly expressed as La1-xSrxMnO3±δ, where δ represents the
oxygen off-stoichiometry (excess or depletion), would correspond in reality to (La1-xSrx)1-εMn1-εO3 with
ε= δ/(3+ δ). Nonetheless, for simplicity we will use the first expression.
As shown in Table 3.4, all the films present an oxygen excess with δ ranging from +0.04 to +0.33. A
certain presence of carbon was measured by RBS-NRA in the outermost layers of LSM (5-7%), which
decreases in the inner layers (<1%). This carbon is likely related to organic surface contamination and
traces of the metal-organic precursors trapped within the film. However, the amount of oxygen of
organic traces is smaller than the oxygen from the manganite, and hence, the oxygen ratio and
stoichiometry are not deeply influenced by the carbon contamination. In addition to the expected cation
vacancies derived from oxygen “excess”, additional cation off-stoichiometry exists in the films. On the
one hand, the LSM20 films are Mn-deficient, with a B-site occupancy around 0.9 for polycrystalline
LSM20/Si and 0.85 for highly-oriented LSM20/MgO. On the other hand, A-site vacancies are present
in the LSM50 films (0.97 and 0.96 A-site occupancy on Si and MgO, respectively), whereas a larger
amount of A-site vacancies are found in LSM75 (0.88 and 0.93 on Si and MgO, respectively).
40
Table 3.4. LSM formula derived from RBS-NRA weighted average composition (Table 3.3). Note that oxygen
excess is actually accommodate it by cation vacancies (see main text).
Film
Tdep LSM / Si LSM / MgO
(solution #)
Mn
LSM20 (4) 675 °C La0.77Sr0.23Mn0.92O3.15 La0.78Sr0.22Mn0.85O3.33
deficient
Mn
LSM20 (4) 700 °C La0.77Sr0.23Mn0.89O3.07 La0.77Sr0.23Mn0.86O3.26
deficient
LSM50 (5) 700 °C (La0.45Sr0.55)0.97MnO3.06 (La0.48Sr0.52)0.96MnO3.24 Mn rich
LSM75 (6) 700 °C (La0.26Sr0.74)0.88MnO3.11 (La0.25Sr0.75)0.93MnO3.04 Mn rich
Effect of the LSM composition on the structural properties
Strontium substitution effect
In this subsection, we focus on the LSM films grown on different substrates at 700 °C using solution 1
to 6 from Table 3.2. The XRD patterns of the polycrystalline films (LSM/Si and LSM/Si 3N4) and the
highly oriented films (LSM/MgO and LSM/LAO) are presented in Figure 3.3 and Figure 3.4,
respectively, where LSM and LAO are indexed in pseudo-cubic notation. The cationic composition
given in each figure corresponds to that measured by WDS for solution 1 to 3 and by RBS-NRA for
solution 4 to 6.
By comparing the position of the diffraction peaks in the XRD patterns of the LSM/Si and of LSM/Si3N4
films with the database references, it can be observed that they match well to rhombohedral (R-3c) or
tetragonal (I4/mcm) LSM perovskite (bottom panel in Figure 3.3) For LSM/Si (Figure 3.3a, acquired
in Bragg-Brentano configuration), the relative intensity of the diffraction peaks matches that of the
patterns from the database. Therefore, we conclude that the LSM films grown on the silicon-based
substrates are polycrystalline. A silver contamination diffraction peak, marked as an open circle, coming
from the silver paste used to glue the substrates during the deposition is observed in some of these
patterns. This indicates that silver migrates through the substrate at the deposition temperature. As a
consequence, the use of silver glue was avoided in posterior depositions on silicon-based substrates.
The evolution of the pseudo-cubic lattice parameter derived from the XRD patterns with the strontium
content is presented in Figure 3.5. As reported for bulk samples,[14], [121] the cell parameter of LSM
decreases when the Sr content increases. As Sr2+ replaces La3+ cations, the Mn4+/Mn3+ ratio increases to
maintain charge neutrality and, given that the ionic radius of Mn 4+ (67 pm[122]) is smaller than that of
41
Figure 3.3. XRD patterns of LSM films with different composition grown on a) Si (Bragg-Brentano geometry)
and b) Si3N4 (grazing incidence XRD). The films are polycrystalline. LSM is indexed in pseudo-cubic notation.
All patterns presented in a) correspond to films grown at 700 °C, whereas in b) the temperature of growth is
indicated for each pattern (675 or 700 °C). The bottom panels display the database references for La 0.8Sr0.2MnO3
(R-3c, ICDD-00-053-0058) and La0.5Sr0.5MnO3 (I4/mcm, ICDD-01-084-6879) for comparison. The table provides
the composition obtained by EPMA-WDS (solution 1-3) and RBS-NRA (solution 4-6) in LSM/Si samples.
Mn3+ (72 pm[122]), the LSM unit cell shrinks. This can be observed in the XRD pattern, where larger
angles correspond to smaller interplanar distances. A shift of the peaks towards larger angles can be
clearly observed in Figure 3.3a, especially for the family planes (002) and (112). Such shift is also
observed in the LSM/Si3N4 patterns. As shown in Figure 3.5a, the lattice parameter decreases with
strontium substitution, where the LSM20/Si3N4 film presents a cell parameter of apc=3.904±0.001 Å,
close to the value of 3.89 Å reported in literature for La0.8Sr0.2MnO3±δ.[123]–[125] The pseudo-cubic cell
parameter of LSM50/Si3N4, apc=3.861±0.002 Å, is in agreement with the values reported for bulk
42
Figure 3.4. XRD patterns in Bragg-Brentano configuration of highly-oriented La1-xSrxMnO3-δ (LSM) films with
different composition grown on a) MgO and b) LaAlO3 substrates. LSM and LAO are indexed in pseudo-cubic
notation. The patterns presented in a) and b) correspond to films grown at 700 °C. The bottom panels display the
database references for La0.8Sr0.2MnO3 (R-3c, ICDD-00-053-0058) and La0.5Sr0.5MnO3 (I4/mcm, ICDD-01-084-
6879) for comparison. The table provides the composition obtained by EPMA-WDS (solution 1 to 3) and RBS-
NRA (solution 4 to 6) in LSM/Si samples. The peak shift with the composition is shown in the insets for 2θ in the
range of 44-49°.
La0.5Sr0.5MnO3±δ, 3.86 Å.[49], [121], [124], [125]

Several LSM/Si3N4 samples were analyzed by GI-XRD
(Figure 3.3b) to avoid the high intensity peak of the silicon substrate and allow a better resolution and
clearer study of the phases present in the film. No secondary phase was detected even for the A-site rich
off-stoichiometric films (2>A/B >1.3). However, the peaks in these films presents a wider FWHM and
larger asymmetry than the nearly stoichiometric films (1.2>A/B>0.8). This suggests the presence of a
mixture of phases of LSM with different strontium substitution and Mn occupancy. The LSM/Si films
present a similar behavior with strontium content than the one described for LSM/Si3N4. The differences
observed in the lattice parameter are due to a poor peak fitting quality for the LSM/Si films.
43
On the other hand, LSM films grown on cubic (100)-oriented MgO single crystal (Figure 3.4a) with
cell parameter aMgO=4.211 Å (Fm-3m, ICDD 00-045-0946) present a different peak intensity than the
polycrystalline films previously discussed (Figure 3.3b). Due to the similar lattice parameter between
the substrate and the LSM (apc=3.86-3.92 Å, as summarized in Figure 3.5a), the film can partially adapt
the substrate’s atomic arrangement and thus, it grows with preferential orientation along the pseudo-
cubic [00l] direction. Nonetheless, the lattice mismatch (or misfit) between substrate and film, defined
as Equation 3.3, is very large (εmf=+8.0% and +9.1% for LSM20 and LSM50, respectively), and thus
other orientations are observed in the out-of-plane direction.
𝑎𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒 − 𝑎𝑓𝑖𝑙𝑚
𝜀𝑚𝑓 =
𝑎𝑓𝑖𝑙𝑚
Equation 3.3
In the case of LSM75 with apc=3.83 Å, the lattice mismatch, εmf=9.9%, is too large for the film to
accommodate the strain induced by the substrate and thus, all orientations are observed as in the
polycrystalline films. This observation is in agreement with the work from Chaluvadi et al.[59] where
La0.67Sr0.33MnO3 (abulk,pc=3.873 Å) films grown on MgO are fully relaxed due to the large mismatch with
the substrate (εmf=+8.7%). The LSM/MgO films exhibit, apart from the (00l) diffraction peaks, low
intensity diffraction peaks indexed as (hh0) which indicate small degree of polycrystallinity. This is in
agreement with the cell parameter derived from the XRD pattern and shown in Figure 3.5b, where the
values are close to the relaxed component of the LSM/LAO films (discussion below).
a) b)
4.00 4.00
Pseudo-cubic lattice parameter (Å)
3.98 Tdep = 700°C LSM / Si 3.98 Tdep = 700°C LSM / MgO

LSM / Si3N4 LSM / LAO
3.96 3.96 LSM / LAO - Strained
3.94 3.94
3.92 3.92
bulk La0.8Sr0.2MnO3 bulk La0.8Sr0.2MnO3
3.90 3.90
3.88 3.88
bulk La0.5Sr0.5MnO3 bulk La0.5Sr0.5MnO3
3.86 3.86
3.84 3.84
3.82 3.82
3.80 3.80
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Sr substitution, x (La1-xSrxMnO3) Sr substitution, x (La1-xSrxMnO3)
Figure 3.5. Pseudo-cubic cell parameter as function of Sr substitution of the LSM films grown on a) Si and Si 3N4,
[60], [123], [125]
and b) MgO and LAO substrates. Reported bulk values are given for LSM20 and LSM50 [49], [60], [121],
[125]
. The values were obtained using TOPAS in a), and using Fityk in b). See Annex A.2 for more details.
44
The LAO substrate, with a smaller lattice parameter of apc=3.794 Å (R-3c, ICDD 04-007-2676), allows
the LSM films to accommodate the strain thanks to a smaller misfit (εmf=-2.8% and -1.8% for LSM20
and LSM50, respectively). Hence, the LSM films grown on (100)pc-oriented LAO single crystal show a
larger degree of preferential orientation along [00l] direction, compared to LSM/MgO films
(Figure 3.4a-b). The shape of the LSM diffraction peaks, with a shoulder on the left side (see inset in
Figure 3.4b), suggests the existence of a gradient in the (00l) interplanar distances. Whereas in the non-
stoichiometric films (A/B>1.4) such gradient may be related to different cationic content, in the nearly
stoichiometric films (0.9<A/B<1.2) it is likely related to the deformation of the lattice as a result of the
strain induced by the substrate. In other words, while the in-plane lattice parameters of LSM shrink to
match those of the LAO substrate, the LSM cell experiences an elongation along the out-of-plane axis
that translates into a shift toward smaller angles in the XRD pattern. In the case of LSM20/LAO (green
pattern: (4)-700 °C), the (00l) peaks can be deconvolved into two main peaks. For the (002) diffraction
peak, the largest contribution is at 2θ=46.58° and the second contribution is at 2θ=45.9°. The lattice
parameters derived from these two contributions are apc=3.89±0.01 Å and 3.94±0.01 Å, respectively.
Thus, the main contribution corresponds to a strain-free component while the second one shows a slight
degree of strain (out-of-plane strain, εzz=+1.21%). The evolution from a strained state towards a strain-
free state is commonly known as strain relaxation and it occurs gradually as the film thickness increases,
involving the formation of defects (see Section 1.3.2). The gradual relaxation is displayed in the XRD
as a prolonged shoulder at the left side of the peaks, where the strain increases as the angle diminishes.
Eventually, as the thickness increases and strain relaxation takes place, the appearance of other crystal
orientations is more favorable. That is the case of the less intense (hh0) diffraction peaks (Figure 3.4b).
The cell parameters of LSM/LAO are presented in Figure 3.5b, where the components have been
separated into relaxed and strained for the applicable cases. The strain-free lattice parameter found for
LSM20 is in good agreement with values reported for bulk La0.8Sr0.2MnO3.[123]–[125] As a result of the
small substrate mismatch (εmf=-1.8%), the LSM50/LAO presents a single component with an average
lattice parameter cpc=3.843±0.002 Å, which is slightly smaller than the reported values for
La0.5Sr0.5MnO3 (3.86 Å).[49], [121], [124], [125] This small difference is expected considering the film has a
strontium content slightly larger, x=0.52, Mn excess, A/B=0.96, and it is oxygen-rich, as discussed
before.
Even though, no secondary phases were found by XRD (Bragg Brentano or GI-XRD), the presence of
impurity phases with small diffraction intensity or non-crystallized phases might be present, especially
for the off-stoichiometric samples.
45
Deposition temperature effect
In this section, the effect of the temperature in the thin film growth using solution 4 will be presented.
Figure 3.6 presents the XRD pattern of the LSM20 films deposited at 675 °C, 700 °C, and 750 °C, on
a) Si, b) MgO, c) Si3N4 and d) LAO substrates. For the polycrystalline samples (Figure 3.6a, c), the
Figure 3.6. XRD pattern of La0.8Sr0.2MnO3 (LSM20) films grown using solution 4 at different temperatures on
a) Si, b) MgO, c) Si3N4, and d) LAO substrates. LSM and LAO are indexed in pseudo-cubic notation.
46
3.98

LSM 20
3.96
3.94
3.92
3.90 bulk
La0.8Sr0.2MnO3
3.88
LSM / Si
3.86 LSM / MgO
bulk
LSM / Si3N4
La0.5Sr0.5MnO3
3.84 LSM / LAO - Relaxed
LSM / LAO - Strained
3.82
675 700 725 750
Tdep (°C)
Figure 3.7. Pseudo-cubic lattice parameter of the LSM20 films grown on different substrate as function of the
temperature of deposition.
diffraction peaks were assigned to Si and to LSM20 (ICDD 00-053-0058). As mentioned earlier, silver
is present in these samples and its diffraction peak intensity increases with the deposition temperature,
suggesting the diffusion of a larger amount of silver through the silicon-based substrates. As shown in
Figure 3.7, while the pseudo-cubic lattice parameter remains constant for the LSM20/Si3N4 films grown
at different temperature (apc≈3.90 Å), it increases linearly with temperature for the LSM20/Si films,
from 3.87 to 3.90 Å. The LSM20/Si films present a larger amount of silver (larger intensity of Ag peaks
in Figure 3.6a, c) than the LSM20/Si3N4 films, which may alter the crystallinity and induce distortions
in the lattice that affect the calculated lattice parameter.
The LSM20 films grown on MgO and LAO (Figure 3.6b, d) exhibit preferential orientation along [00l]
direction where the relative intensity of other orientations such as (hh0) increases with temperature. The
LSM20/MgO films present a single component peak with a cell parameter related to a relaxed state
around apc=3.89±0.03 Å for Tdep=675 and 700 °C, whereas it increases to apc=3.91±0.01 Å at 750 °C.
On the other hand, in LSM20/LAO films, the relaxed component of the cell parameter does not vary
with the deposition temperature, 3.90±0.01 Å (675 °C) and 3.90±0.01 Å (750 °C) (Figure 3.7). The
strained component shows a small shrinkage at low temperatures, from 3.95±0.02 Å (675 °C) to
3.94±0.01 Å (700 °C), but it increases for Tdep=750 °C up to 3.97±0.04 Å which corresponds to an out-
of-plane strain of εzz=+1.7%. Indeed, the diffraction peak assigned to the strained component appears
clearer and more separated at high temperatures (Figure 3.6d).
47
While the LSM/Si films are largely affected by the presence of silver, the films on the other substrates
present a stoichiometry and lattice parameter close to the reported bulk values of La0.8Sr0.2MnO3 within
the studied range of temperatures. The small variations of cell parameter observed can be attributed to
the experimental error coming from small differences in the precursors’ solution concentration. In this
thesis, highly oriented and thinner films (7-21 nm) grown on single crystal substrates, i.e. LAO and
SrTiO3 (STO), are used to establish the relation between structure (with some defects and with virtually
no defects) and memristive performance. In order to reduce the appearance of different orientations, as
observed for the highly oriented films grown on MgO and LAO at 750 °C, and reduce the risk of
formation of secondary phases, the temperature of deposition chosen for the growth of functional
films to fabricate memristive devices is 675 °C.
Morphology and microstructure
The LSM films grown from different solutions for the compositional study were characterized by SEM
to assess their morphology and their density. Figure 3.8 presents the SEM images acquired on LSM/Si
films grown at a)-d) 675 °C and e)-h) 700 °C with a strontium content x<0.25 (solution 1 to 4, where
solution 4 corresponds to LSM20 composition). Despite the small different composition measured by
WDS, no major morphological differences are found between the film grown at 675 °C and 700 °C
using the same precursor solution. The films present randomly distributed grains, in agreement with the
polycrystalline texture observed by XRD. The average grain size is around tens of nanometers and
decreases with strontium content in films with x≤0.1 (Figure 3.8a-c and e-g). However, for x=0.2, i.e.
LSM20, the grains coalesce forming larger grains with a tetrahedral shape (Figure 3.8d and h).
The morphology of LSM/MgO films with x=0-0.2 grown at 700 and 750 °C is presented in Figure 3.9.
Whereas the average grain size is around a few tens of nanometers for the films with x≤0.1 at
Tdep=700 °C (Figure 3.9a-c), the grains show a more ordered arrangement as stoichiometry is reached
(larger Sr-substitution and decrease of A/B ratio towards 1). The LSM20/MgO grown at 700 °C is
composed of a bottom layer of squared-shaped ordered grains aligned in two perpendicular directions
and some tetrahedral-shaped grains on top. The latter resemble the grains found in LSM20/Si films, and
hence, they might be related to the low intensity (hh0)pc diffraction peaks found by XRD (Figure 3.6b),
causing the polycrystallinity of the LSM20/MgO films. Then, the square grains can be assigned to the
(00l)pc crystallographic orientation and prove the cube on cube growth of the LSM20 on MgO substrate.
However, the LSM20/MgO films grown at 750 °C exhibit a larger number of grains on the surface
associated to the polycrystalline nature, in agreement with the larger contribution found by XRD.
Nonetheless, a bottom layer of ordered grains could also be present next to the substrate interface, based
on the preferential orientation found by XRD. The LSM/MgO films with a lower strontium content show
similar morphology, in which the outermost layers present grains related to the polycrystalline nature.
48
Figure 3.8. SEM surface morphology observed by secondary electrons for La 1-xSrxMnO3 (LSM) thin films with
different stoichiometry grown on Si substrates. The thin films were grown at a)-d) 675 °C and at e)-h) 700 °C,
using precursor solution with an A/B ratio of 2.0, and Sr substitution of a), e) 0.10, b), f) 0.25, c), g) 0.40, and d),
f) 0.55. The film in d) and h) correspond to LSM20 compositions. The cationic composition ratios in the films
measured by WDS are given below each image.
The morphology of the nearly stoichiometric films, i.e. LSM20 and LSM50, grown on Si, LAO and
MgO substrates, grown with 15000 pulses is presented in Figure 3.10. The highly oriented LSM20/LAO
(Figure 3.10b, e) and LSM20/MgO (Figure 3.10c, f) films are composed of square-shaped and
tetrahedral grains, that might be related to the preferential (00l)pc orientation and polycrystalline
component, respectively. In agreement with the XRD pattern, the LSM20/LAO films exhibit a smaller
49
Figure 3.9. Surface morphology observed by secondary electrons in SEM for La 1-xSrxMnO3 (LSM) thin films with
different stoichiometry grown on MgO substrates. The thin films were grown at a)-d) 700 °C and at e)-h) 750 °C,
using precursor’s solution with an A/B ratio of 2.0, and Sr substitution of a), e) 0.10, b), f) 0.25, c), g) 0.40, and
d), f) 0.55. The cationic composition ratios in the films measured by WDS are given below each image.
amount of the polycrystalline grains. The differences observed in the texture are derived from the
mismatch between LSM and the substrate, where the smaller it is the more favorable is to retain the
epitaxial texture inherited from the substrate. As described in Figure 3.9, the grains evolve towards a
tetrahedral shape with strontium content and are clearly defined in the LSM50 films (Figure 3.10g, h, i).
Whereas the presence of tetrahedral grains is expected in polycrystalline LSM/Si films
50
Figure 3.10. Morphology of the LSM20 and LSM50 films grown on Si, LAO and MgO substrates. The thickness
was measured by SEM cross-section in different locations.
(Figure 3.10a, d, g), their appearance on the thick LSM/LAO and LSM/MgO might be the result of
strain relaxation due to the large thickness of the films presented here (90-110 nm). Another
consequence of the large thickness is the appearance of space between the grains, increasing the surface
area and porosity in the films. Nonetheless, the layers underneath are expected to be continuous. These
morphologies are very interesting for solid oxide cells electrodes, where due to their large surface area
the upper layer would increase the number of reaction sites and the bottom dense layer would allow for
continuous oxygen migration and electronic paths.[69] Whereas the thicknesses measured in cross-section
by SEM for LSM/LAO and LSM/MgO are comparable, the equivalent films on Si substrates are larger.
The difference can be attributed to the appearance of porosity in the films.
Raman of the LSM20 and LSM50 films
The Raman spectra of the optimized LSM20 and LSM50 samples grown on a) MgO, b) LAO and
c) Si3N4 substrates is presented in Figure 3.11. MgO substrate does not present any Raman signal and
hence, allows to study the Raman scattering of the LSM films. Therefore the LSM/MgO films are used
here to present the Raman spectra of LSM (Figure 3.11a). The Raman spectra of rhombohedral LSM20
51
a) LSM20/MgO-Tdep 675°C b) LSM20/LAO-Tdep 675°C

LSM20/MgO-Tdep 700°C LSM20/LAO-Tdep 700°C
LSM50/MgO-Tdep 700°C
JT ⨂ LSM50/LAO-Tdep 700°C
JT ⨂
Intensity (arb. un.)

JT JT
▼
A1g Eg Eg A1g Eg Eg ▼
100 200 300 400 500 600 700 800 900 100 200 300 400 500 600 700 800 900
Wavenumber (cm-1) Wavenumber (cm-1)
c) LSM20/Si3N4-Tdep 675°C ▼
LSM20/Si3N4-Tdep 700°C
LSM50/Si3N4-Tdep 700°C
JT ⨂
A1g Eg Eg
JT
100 200 300 400 500 600 700 800 900
Wavenumber (cm-1)
Figure 3.11. Raman spectra of LSM20 and LSM50 films grown on a) MgO, b) LAO and c) Si 3N4 substrates. The
approximate positions of the LSM modes according to literature (see main text) are given by dashed lines. The
signal arising from the substrates is marked with an inversed triangle.
consists of two broad bands around 520 cm-1 and 630 cm-1 which are assigned to the electron-phonon
interaction: Jahn-Teller (JT) distortion.[126]–[128] Other three peaks of lower intensity are found at
wavenumbers 220, 350 and 420 cm-1, corresponding to the A-site (La, Sr) vibrations (A1g), tilt of the
Mn-O-Mn angle of the octahedra (Eg), and the bending of (La, Sr)-O bonds (Eg).[126]–[129] The exact
position of the modes, especially the non-JT modes, are found to be sample sensitive and may vary in a
range of 15-20 cm-1.[126] Nonetheless, a monotonic decrease in the wavenumber position of the modes
has been observed as Sr content increases and it can be used to estimate the Sr content in films of
unknown composition.[130] On the other hand, the JT modes are sensitive to the charge concentration
that causes the JT distortion. As Sr concentration increases in LSM films, the JT modes show a decay
of intensity as result of the loss of the electrons (oxidation of Mn 3+ to Mn4+) that cause the distortion.
These modes are not visible in the tetragonal LSM50 sample, which does not present JT distortion.
52
Similar trends have been reported for other manganites with A-site substitution.[131] A description of JT
distortion and its origin is given in Section 1.3. The last feature that can be observed in the Raman
spectra is a sharp peak at around 660 cm-1, marked as ⨂ in Figure 3.11. The assignment of this peak,
typically observed in other manganites, is matter of controversy as it is not predicted by lattice dynamics
calculations and many authors related to traces of a precursor binary oxide, such as Mn3O4.[130], [132], [133]
Similar spectra are found for the LSM/LAO films (Figure 3.11b), in which the LAO substrates present
three additional peaks in the studied range. The study of LSM grown on silicon-based substrates by
Raman spectroscopy is hindered due to the large intensity peak of silicon at 520.7 cm-1 which overlaps
with one of the JT modes (Figure 3.11c). The LSM20 films grown at 675 °C and 700 °C present similar
spectra and the differences found can be assigned to small phase inhomogeneities. The evolution of
Raman spectra with oxygen content in LSM20 is discussed in Section 6.1.
Effect of the composition on the LSM films electrical properties
The electrical properties of the films were measured by an in-line four-point probe station. Figure 3.12
shows the electrical resistivity of the LSM films grown at 700 °C on MgO, LAO and Si3N4 substrates
as a function of the strontium content in the film (measured by WDS and/or RBS-NRA). As expected
from literature, the resistivity decreases with strontium content up to x=0.5, and then, increases again.[14],
[51], [134]
The resistivity values found for the off-stoichiometric films with low strontium content
(>10 Ω·cm for x≤0.1) are larger than those reported in literature (0.24 Ω·cm for LMO/MgO (x=0); and
0.05 Ω·cm LSM/MgO (x=0.1))[134] which might be related to the presence of amorphous and/or impurity
phases.
As described in Section 3.2.1, the A/B ratio approaches 1 in the LSM films studied as the Sr content
increases (x=0-0.2). This means that less Mn vacancies are present and therefore Mn-O-Mn paths
responsible of electronic conduction are restored. The resistivity measured in close-to-stoichiometry
LSM20 films (0.06-1.24 Ω·cm) is in agreement with reported values (bulk: 0.01-0.05 Ω·cm,[14], [51] thin
film: 0.01-0.35 Ω·cm,[53], [93], [134]) although the value for each film greatly depends on the cationic
occupancy, oxygen content, texture and strain (see Section 1.3 for the description of conduction
mechanism and influence of defects).
As can be clearly observed in Figure 3.12 the resistivity of the close-to-stoichiometry films with
strontium substitution x=0.2, 0.5 and 0.75 (LSM20, LSM50, LSM75) depends on the substrate. As
expected, the polycrystalline films, LSM/Si3N4, show the largest resistivity as a result of the oxygen
deficient and insulating nature of the grain boundaries.[54] The larger resistivity measured on the
polycrystalline films might be also due to their porous morphology. On the other hand, the preferentially
oriented LSM/LAO and LSM/MgO films present less grain boundaries and therefore, the resistivity is
53
106
LSM / MgO
105
LSM / LAO
104 LSM / Si3N4
Resistivity (Ω·cm-1) 103
102
101
100
10-1
10-2
10-3
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Sr substitution, x (La1-xSrxMnO3)
Figure 3.12. Electrical resistivity as function of Sr substitution, x, in La1-xSrxMnO3 films grown at 700 °C on
MgO, LAO and Si3N4 substrates.
smaller. The larger resistivity found for LSM/MgO films might be related to their more porous
morphology in comparison to that of LSM/LAO films. For the thick films discussed here, strain effects
can be neglected.[59] Finally, the resistivity of the LSM75 films increases as expected, as the composition
gets closer to the insulating SrMnO3 parent compound.[38]
Remarks on the section
From this study, after optimization, the 20, 50 and 75% Sr substitution compositions were obtained
using solution 4, 5 and 6, respectively. The deposition conditions selected to grow samples of 20 and
50 % used in memristive devices are summarized in Table 3.1, where the deposition temperature is
675 °C and the number of pulses used is between 1800 and 3600 (exact number given in each
corresponding section) to obtain thin (7 to 21 nm) and more homogeneous films.
54
3.3 Conclusions
Conclusions
The conclusions withdrawn from this chapter are organized according to its sections.
From Section 3.1 Geometrical optimization of the reactor:
 The key parameters to obtain a large homogeneous deposition area are (i) the relative position
of the gas inlet and the holder, (ⅱ) the gas flow, and (ⅲ) the position at which the gas is injected.
 A 2 inches diameter area of homogeneous deposition was achieved and it allows the growth
of LSM films on several 10×10 mm2 substrates.
From Section 3.2 Deposition of LSM thin films with different compositions:
Regarding the optimal LSM growth conditions:
 Using the deposition conditions summarized in Table 3.1, the optimal deposition temperature
is 675 °C. At this temperature, the films show larger apparent density, are as well-crystallized
as those grown at 700 °C, and the potential appearance of impurity phases is minimized. In
addition, the deposition at this temperature is faster and less energy-consuming.
 Thin films with stoichiometry close to La0.8Sr0.2MnO3-δ (LSM20) are obtained using a
precursor solution concentration of 22.5 mM, with [(La+Sr)/Mn] ratio of 2.0 and [Sr/(La+Sr)]
ratio of 0.55.
 Thin films with stoichiometry close to La0.5Sr0.5MnO3-δ (LSM50) are obtained using a
precursor solution concentration of 22.5 mM, with [(La+Sr)/Mn] ratio of 2.3 and [Sr/(La+Sr)]
ratio of 0.80.
Regarding the composition and characterization of the optimized films:
 The LSM20 and LSM50 films are polycrystalline when grown on Si and Si3N4 substrates, and
present preferential orientation along (00l)pc when grown on (001)pc-oriented LaAlO3 (LAO)
and (001)-oriented MgO single crystals substrates. A small degree of polycrystallinity is found
in the latter films as result of the large thickness (ca. 100 nm) of the films deposited for the
study of the composition.
 The LSM20 films grown on MgO are relaxed (cpc=3.89±0.01 Å) whereas those grown on LAO
present an in-plane compressive strained component, with concomitant out-of-plane
elongation (cpc=3.94±0.01 Å), and a relaxed component (cpc=3.89±0.01 Å).
 The LSM50 films grown on MgO and LAO present one single component with an average
cell parameter, cpc=3.843±0.002 Å.
 No temperature dependence of the cell parameter is observed for the LSM20 films grown on
Si3N4 and LAO (relaxed component) substrates for Tdep=675-750 °C. On the contrary, the cell
55
parameter increases for the strained component of LSM20/LAO, and for LSM/20MgO at
750 °C.
 The silver paste used to glue the substrates during the deposition migrates through the silicon-
based substrates, as detected by XRD, and its use has been avoided for the deposition of
functional LSM layers on silicon-based substrates.
 As found by RBS-NRA, the cationic composition of the film does not depend on the substrate.
The LSM films present nominal oxygen excess when grown on Si and MgO substrates, which
is expected to be accommodated by the formation of cation vacancies. Further off-
stoichiometry is found (A/B≠1), where LSM20 films are Mn-deficient (A/B=1.1-1.2) and
LSM50 films are A-site deficient (A/B=0.9). These results can be extrapolated to other
polycrystalline and preferentially oriented (epitaxial) films, respectively.
 Polycrystalline LSM20 films present randomly oriented grains which cluster into rough
tetrahedral-shaped grains. Polycrystalline LSM50 films present randomly oriented grains with
well-defined faces and tetrahedral shape.
 Highly oriented LSM20 and LSM50 films on LAO present a bottom layer of square-shaped
ordered grains and scattered tetrahedral-shaped grains, similar to the polycrystalline samples.
The number of the latter increases with deposition temperature (from 675 to 700 °C).
 Highly oriented LSM20/MgO film presents a bottom layer of square-shaped grains, similar to
that of LSM20/LAO, but larger concentration of tetrahedral-shaped grains. The latter
contribution covers all the surface in LSM50/MgO, hiding the presumed bottom layer.
 The polycrystalline LSM/Si3N4 films present larger resistivity than the highly oriented films,
in agreement with the insulating nature of the oxygen-depleted grain boundaries.
Regarding the off-stoichiometric films:
 No secondary phase is found by XRD in any of the films with A/B=2-0.9. Nonetheless, binary
oxides, or other impurities with low crystallinity, below the XRD detection limit, could be
present. Ag from the silver paste used for deposition is found in silicon-based substrates.
 The experimentally measured cell parameter decreases with the incorporation of strontium
into the film, similar to the shift observed for stoichiometric lanthanum manganite with
different strontium substitution levels. However, the diffraction peaks in these films are wider
and less intense than in stoichiometric films, which suggest a poorer crystallization.
 The apparent density of the off-stoichiometric film decreases with A/B ratio composition.
 The electrical resistivity of the films increases as A/B increases, in agreement with the
apparent density and the potential presence of insulating impurity phases.
56
Effect of the nanostructure in
La0.8Sr0.2MnO3-δ thin films in memristive
devices
Epitaxial LSM20 as memristive material ............................................................................. 58
Structural characterization of epitaxial films ............................................................... 58
Electrical characterization of epitaxial LSM20 devices .............................................. 63
Discussion: oxygen migration modelling .................................................................... 73
Summary on epitaxial LSM20 as memristive material ............................................... 75
Epitaxial and polycrystalline texture in thicker LSM20 films ........................................... 76
Description of the samples........................................................................................... 77
Comparison between epitaxial and polycrystalline LSM20 memristive devices ........ 78
Cross-comparison: the role of thickness ............................................................................... 99
Conclusions ........................................................................................................................... 103
The following chapter focuses on the integration as planar memristive devices of thin films with 20%
strontium substitution, i.e. La0.8Sr0.2MnO3-δ (LSM20). The chapter is divided in three sections. In the
first two sections, we present LSM20 films with different nanostructure and same thickness (several
thickness are discussed). In particular, in the first section we focus on 7 nm thick epitaxial LSM20 films
grown on single crystal substrates. In the second section, we assess the reproducibility of memristive
devices on epitaxial films and we introduce polycrystalline LSM20 films. Epitaxial and polycrystalline
films with two thicknesses, ca. 8 nm and ca.21 nm, are presented. In each of these two sections, we
present the nanostructure characterization, the memristive characteristics, followed by a discussion of
the influence of the nanostructure on the memristive response, and few remarks. The third section, which
serves as summary, is a cross-comparison of each nanostructure as a function of the thickness.
57
Chapter 4: Effect of the nanostructure in La0.8Sr0.2MnO3-δ thin films in memristive devices
Epitaxial LSM20 as memristive material1

As mentioned in the introduction, the properties of the surface and interface become more important
than bulk properties when scaling down any system. In the case of thin films with nanoscale thickness,
the microstructure has a direct influence in the functional properties of the devices, grain boundaries,
strain in the lattice and defects acquire an important role. In this section, we aim to evaluate the influence
of the nanostructure on the memristive characteristics of two model systems with different strain.
Structural characterization of epitaxial films
The orientation, strain and extended structural defects of a thin film can be easily modified by using
different perovskite single crystal substrates with lattice cell parameters close to those of the layer. In
order to compare the resistive switching response of LSM20 memristive devices with two different
microstructures at the nanoscale, 7 nm thick LSM20 thin films were grown simultaneously by PI-
MOCVD on (001)-oriented SrTiO3 (STO) and on (001)pc-oriented LaAlO3 (LAO) (pc=pseudo-cubic)
substrates, hereafter named as LSM20/STO and LSM20/LAO, respectively. The pseudo-cubic bulk cell
parameter of LSM20 is aLSM20, pc=3.898 Å.[60], [123] On one hand, thin LSM20 films grown on STO are
expected to grow epitaxial as a result of the similar cell parameter (aSTO=3.905 Å), with a very small in-
plane tensile strain (ε= +0.18%, Equation 4.1). On the other hand, films grown on LAO, with a smaller
cell parameter (aLAO,pc=3.79 Å), would exhibit in-plane compressive strain (ε=-2.77%, Equation 4.1)
and concomitant expansion in the out-of-plane direction, as reported elsewhere.[135] The precursor
solution cationic concentration that allows the growth of La0.8Sr0.2MnO3-δ was used to grow the epitaxial
films presented here (solution 4 in Section 3.2). The composition is assumed to be reproducible as energy
dispersive x-ray spectroscopy (EDS) cannot be used to resolve their relative composition due to the
simultaneous presence of La or Sr on the film and the substrate.
𝑎𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒 − 𝑎𝑓𝑖𝑙𝑚
𝜀𝑚𝑓 =
𝑎𝑓𝑖𝑙𝑚
Equation 4.1
1
Part of this section is adapted from the paper “Structural Defects Improve the Memristive Characteristics of
Epitaxial La0.8Sr0.2MnO3-based Devices”, Carlos Moncasi, Gauthier Lefèvre, Quentin Villeger, Laetitia Rapenne,
Hervé Roussel, Ahmad Bsiesy, Carmen Jiménez, Mónica Burriel, Adv. Mater. Interfaces, vol. 9, no. 23, p.
2200498, 2022.[171]
58
4.1 Epitaxial LSM20 as memristive material
Figure 4.1. Structural characterization of the LSM20 films. XRD patterns of LSM20 films grown on a) STO and
b) LAO. LSM and LAO peak are indexed in pseudo-cubic (pc) system, and STO peaks in cubic system. The films
exhibit epitaxial growth. Black drop lines correspond to La0.8Sr0.2MnO3 bulk positions according to pattern ICDD-
00-053-0058. Surface imaging by SEM (left) and AFM height profile (right) of the c) LSM20/STO and
d) LSM20/LAO films.
The XRD patterns of the LSM20/STO and LSM20/LAO films are shown in Figure 4.1a and b,
respectively. These show high intensity peaks assigned to the reflections of the (00l) planes of the single
crystal substrates, in addition to which only the characteristic (00l)pc reflections of LSM20 can be
observed next to the substrate peaks. For the LSM20/STO sample, the film reflections overlap with the
high intensity peaks of STO, while for the LSM20/LAO sample, the peaks assigned to the film appear
at smaller angles than for the bulk (drop lines in Figure 4.1a and b, ICDD-PDF 00-053-0058), in
agreement with the expected in-plane compressive strain induced by the LAO substrate. The
experimentally measured out-of-plane cell parameter obtained for the LSM20 film on LAO is 3.95 Å,
which corresponds to an out-of-plane elongation of ε=+1.33%, related to a partial relaxation of film’s
strain, as discussed below.
The surface morphology of the films was studied by SEM and by AFM in tapping mode for LSM20/STO
and LSM20/LAO as shown in Figure 4.1b and c, respectively. The as-grown films are homogeneous,
59
Figure 4.2. TEM cross-section characterization of LSM20/STO and LSM20/LAO films. HRTEM image of the
7 nm thick LSM20 film on a) STO and c) LAO. b), d) Fourier transform showing the reciprocal space of the
images presented in a), b) respectively, and inverse Fourier transform image obtained applying a mask on the
(010)LSM20 and (010)STO and LAO spatial frequencies on LSM20/STO and LSM20/LAO, highlighting the planes
perpendicular to the surface. The dislocations in d) LSM20/LAO are marked with “⊥” and “⊤” symbols.
dense and flat, with a root mean square roughness, Sq, of around 0.6 nm. The film is composed of small
grains which may appear as the result of slight misorientation between adjacent growth nuclei in an
island growth type. The grains possess the same crystallographic orientation with no clearly defined
boundary (seen by HRTEM in Figure 4.2) and strictly speaking they cannot be referred to as
crystallographic grains. The average grain size is 29±2 and 21±2 nm for LSM20/STO and LSM20/LAO,
respectively.
The crystal quality of the film was assessed by HRTEM after the fabrication of the devices. The epitaxial
nature of the film is observed by the HRTEM cross-section images of LSM20[001]/STO[001] and
60
LSM20[001]/LAO[001] in Figure 4.2a and b, respectively. Figure 4.2a shows long-range domains of
ordered atoms in the LSM20 film when grown on STO. However, the LSM20 film presents some
extended structural defects, when grown on LAO (Figure 4.2c), probably caused by the larger cell
mismatch, compared to that on STO. The Fourier transform of the HRTEM images presented in
Figure 4.2a and c (LSM20/STO and LSM20/LAO, respectively) is shown in the inset of Figure 4.2b
and d, respectively. When a mask is applied to the diffraction spots corresponding to (010)LSM20 and
(010)LAO or STO, which are superimposed, we obtain the inverse Fourier transform filtered image presented
in Figure 4.2b and d, where only the diffraction planes perpendicular to the interface are visible. In
LSM20/STO (Figure 4.2b), the planes are straight and extend from the bottom interface to the top, thus
proving a good epitaxy with absence of extended defects. On the contrary, the inverse Fourier transform
image in LSM20/LAO (Figure 4.2d), suggests the presence of extended structural defects in the LSM20
film appearing as discontinuity of the planes, extra planes (zoomed-in in Figure 4.2d), and deflection
of planes, which may cause partial relaxation of the strain. Considering a fully relaxed LSM film the
|𝑏|
theoretical dislocation spacing can be calculated by 𝑆 = 𝜀
. Assuming a burgers vector of one unit cell
of the substrate, for example in the direction of [100], b = aLAO[100] (|𝑏|= 3.79 Å), and a misfit
ε=- 2.77%, the dislocation spacing for a full relaxation is S=13.7 nm, which corresponds to a linear
dislocation density of 7.3×105 cm−1 and means one dislocation every ~ 36 unit cell of the LAO structure.
In the image shown in Figure 4.2d a first dislocation is found at the left of the imaged (marked with a
“ꓕ” symbol, enlarged region inside a white square). In addition, a second complex defect, which could
be formed by two dislocations of opposite sign cancelling each other (marked with “⊥” and “⊤”
symbols), is found at the right side of the image, at a distance of 17-20 nm from the first one. We thus
confirm that the number of dislocations observed in this region correspond well to the expected for a
partially relaxed LSM20/LAO film. The presence of extended defects on LSM20/LAO was also
reported by Navickas et al.[135] who assigned them to dislocations across and along the film. Other
regions of LSM20, both with and without top electrodes, were observed and exhibit similar film crystal
quality as those presented for LSM20/STO and LSM20/LAO in Figure 4.2a and c, respectively. The
average thickness of the films is 7 nm. Hence, we confirm that both films are epitaxial, with the cube-
on-cube growth of the LSM20 perovskite on top of the substrate perovskite lattice. However, while
LSM20/STO is epitaxial, presents no extended defects and a very small in-plane tensile strain, the
LSM20/LAO film shows indeed the presence of dislocations and in-plane compressive strain.
Memristive devices in planar configuration (electrodes on top of the film) were fabricated using square
10 nm thick Ti and 100 nm thick Pt as active and ohmic electrode, respectively. The Ti electrode was
capped with an additional 90 nm thick Pt layer to prevent atmospheric oxidation. The microfabrication
process and the arrangement of devices on the film is described in Section 2.2. The metals were chosen
61
Figure 4.3. STEM-EDS chemical mapping of pristine a) Ti/LSM20/STO and b) Ti/LSM20/LAO stack that
suggest the presence of a TiOx interlayer. Chemical profile of the previous c) Ti/LSM20/STO and
d) Ti/LSM20/LAO stacks. Shaded areas in d) correspond to the interfaces with LSM20.
based on their band alignment and, in the case of Ti, on its potential to act as oxygen reservoir. More
information is provided in Section 1.4.1.
The presence of a TiOx interlayer between the LSM20 film and the Ti top electrode was observed by
EDS mapping on TEM lamellas of the Pt/Ti/LSM20/STO (Figure 4.3a) and Pt/Ti/LSM20/LAO
(Figure 4.3b) stacks using pristine devices. Figure 4.3c, d presents the chemical profile across the
Pt/Ti/LSM20/STO and Pt/Ti/LSM20/LAO stacks, respectively. It should be noted that the slight Al
concentration increase in the Pt layer (Figure 4.3d) most probably corresponds to a parasite signal
coming from backscattered and high angle diffused electrons interacting with the aluminum washers
used to fix the sample holder. Focusing first on the LSM20 film, we detect a slight accumulation of Mn
at the bottom LSM20/LAO interface (Figure 4.3d), while the La, Sr and O profiles are quite
homogeneous across the LSM20 film. Within the given resolution, no Sr segregation is observed at the
top LSM20 surface (LSM20/Ti interface), as often observed for Sr perovskites.[136]–[138] Oxygen clearly
62
accumulates on the Ti/LSM20 interface, forming a TiOx interlayer of ∼2-4 nm. On the LSM20 side of
the Ti/LSM20 interface (green shaded area in Figure 4.3d), the O/Ti ratio is roughly 2 and decreases to
1 after ∼1 nm. Similar characteristics were found in the EDS maps and profile obtained on
Ti/LSM20/STO stack (Figure 4.3a and c).
Electrical characterization of epitaxial LSM20 devices
Transmission line measurement and symmetric devices
The Pt/LSM20 contact resistance and LSM20 resistivity were obtained by transmission line
measurements (TLM) for both samples. These values are of the same order of magnitude for both
samples, as summarized in Table 4.1. We note that the resistivity is larger than other values reported in
literature, 5-220 mΩ·cm [54], [93]
although they are reasonable considering the differences in
stoichiometry, oxygen content and other parameters that determine the exact value (see Section 1.3).
More information about the TLM measurements and the analytical resolution is provided in Section 2.3.
As shown in Figure 4.4, ohmic and rectifying responses were obtained in both samples for symmetric
Pt/LSM\Pt and Ti/LSM\Ti planar devices. For the Pt/LSM\Pt devices ohmic behavior (a linear response)
was obtained even up to ±20 V (not shown), with a resistance value of 150 kΩ and 260 kΩ for LSM/STO
and LSM/LAO, respectively. On the other hand, Ti/LSM\Ti devices present rectifying behavior with a
resistance around 106 Ω (at +0.1 V), and a hysteretic behavior in both polarities for large voltage ranges
(not shown). These responses are in agreement with the band alignment, confirming the passive role of
the inert Pt electrode and the active role of the Ti electrode, inducing the memristive behavior. The
different resistance values are coherent with the formation of a Schottky barrier, and the concomitant
reduction of LSM and oxidation of Ti into an oxygen deficient TiOx layer, as observed by STEM-EDS.
Initial resistance (IR) of asymmetric devices
The memristive behavior was studied on asymmetric planar Pt/Ti/LSM20\Pt devices, in which the
voltage bias was applied to the Pt/Ti active electrode while the Pt electrode was grounded. Previously,
the initial resistance (IR) of the pristine devices was measured using a unipolar DC sweep at +0.1 V in
two sets of devices: (i) devices with electrodes of different area, A, and separated by 50 μm and,
(ⅱ) devices with different separation between electrodes, L, and with a fixed electrode
Table 4.1. Results of Transmission line measurements (TLM) using Pt contacts.
LSM-7/STO LSM-7/LAO
Contact resistance, RC 2.27 ∙ 104 Ω 2.12 ∙ 104 Ω
Resistivity, ρ 0.38 Ω∙cm 0.52 Ω∙cm
63
a) b)
3x10 -5 LSM20 / STO 3x10-5 LSM20 / LAO
2x10-5 Pt/LSM\Pt 2x10-5 Pt/LSM\Pt

Ti/LSM\Ti Ti/LSM\Ti
Current (A)
Current (A)
1x10-5 1x10-5
0 0
-5 -5
-1x10 -1x10
-2x10 -5
-2x10-5
-3x10-5 -3x10-5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage (V) Voltage (V)
Figure 4.4. I(V) characteristics of symmetrical devices fabricated on a) LSM20/STO and b) LSM20/LAO.
Symmetrical devices of Ti/LSM20\Ti (solid line) show rectifying behavior, and Pt/LSM20\Pt (dashed line) show
ohmic behavior.
A of 200x200 μm2. The values of IR obtained for the two sets of devices for LSM20/STO and
LSM20/LAO are presented in boxplot in Figure 4.5. On one hand, Figure 4.5a and b show that the IR
is independent of the substrate and of the distance between contacts, L=50, 100 and 200 μm, suggesting
a negligible contribution of the LSM20 sheet resistance to the total IR measured. On the other hand,
Figure 4.5c and d show the dependence of IR with contact area, which varies in a larger way (between
5 MΩ and 20 MΩ in LSM20/STO, and 5 MΩ and 30 MΩ in LSM20/LAO). In both samples, the
measured resistance increases as the contact area scales down. The measured resistance of the devices
is expected to decrease with the area with a negative unitary slope in a double logarithmic plot (log(R)
vs. log(A)). The slope found for both samples deviates from the negative unitary slope which suggests
that the resistance does not strictly depend of the geometry, and other resistive elements might be present
and/or show a nonlinear change of resistance with the applied voltage (e.g. different degree of oxidation
in the TiOx interlayer). As a first approach, the total resistance (RT) can be calculated as the sum of serial
resistances of each material and interface, according to Equation 4.2.
𝑅𝑇 = 𝑅𝑃𝑡 + 𝑅𝐶1 + 𝑅𝐿𝑆𝑀 + 𝑅𝐶2 + 𝑅𝑇𝑖
Equation 4.2
𝐿
being 𝑅𝐿𝑆𝑀 = 𝑅𝑆 𝑊 , where RS is the sheet resistance of the LSM film, L is the length between the
electrodes, and W is the width of the electrodes. Using the resistivity (sheet resistance) and contact
resistance between Pt/LSM20 (RC1) values obtained by TLM, it can be inferred that the Ti/LSM20
64
interface (RC2) governs the total resistance, being responsible of 90-94 % and 77-95 % of the total
resistance for LSM20/STO and LSM20/LAO, respectively. The large contribution of the RC2 component
to the IR is in good agreement with the formation of a Schottky barrier and a TiOx layer, as shown by
STEM-EDS (Figure 4.3), as well as the depletion of oxygen in the LSM20 film under Ti. Both these
phenomena are expected to increase the RC2 resistance contribution. The larger IR distribution measured
for LSM20/LAO compared to LSM20/STO could come from a non-homogeneous structural defect
Figure 4.5. Initial resistance (IR) of the Ti/LSM20\|Pt pristine memristive devices. a), b) Distance and c), d) area
dependence study on LSM20/STO and LSM20/LAO, respectively. The area of the electrode (A) and distance
between electrodes (L) in the distance study is A=2002 µm2 and L=50, 100 and 200 µm2, while for the area study
is A=1002, 2002, 3002, and 4002 µm2 and L=50 µm. The insets in a) and c) show a schematic top-view of the
electrode arrangement as function of L and A, respectively. Box range: 25th to 75th percentile. Whiskers: to 10th
(lower limit) and to 90th (upper limit) percentile. Median value: horizontal line in box. Mean: open data point.
Outliers: solid data points. For plots c) and d), the dashed gray lines correspond to slope -1. In LSM20/STO
(LSM20/LAO), the data of the distance dependence study was obtained from 15, 15, 30 (16, 16, 32) devices for
L=50, 100 and 200 µm2, respectively; and for the area dependence study it was obtained from 9 (12) devices for
each size.
65
distribution within the LSM20 film, leading to a certain inhomogeneity of the TiOx layer in terms of
oxidation level and thickness, as well as different gradient in the concentration of oxygen vacancies
inside LSM20.
Memristive characteristics
Memristive behavior could be obtained in Pt/Ti/LSM20\Pt planar devices of all sizes. Here we focus on
devices with the contacts separated by L=50 μm and A=2002 µm2, hereafter referred to as standard
devices. All the devices exhibit clockwise log I(V) (or counter-clockwise, CCW, R(V)) characteristics
where the HRS RESET and LRS SET are programmed at positive (Vreset) and negative voltage (Vset),
respectively. An initialization procedure was carried out to evaluate the operating voltage amplitude
range of the LSM20-based devices. The procedure consists of symmetric bipolar writing voltage sweeps,
0 V  +Vamp 0 V  -Vamp 0 V (defined as a cycle), where the voltage amplitude Vamp was increased
Electric field (V/m) Electric field (V/m)

a) -4x105 -2x105 0 2x105 4x105 b) -4x105 -2x105 0 2x105 4x105
LSM20 / STO LSM20 / LAO

Resistance (Ω)
Resistance (Ω)
107 107
106
106
-20 -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20


5
c) -4x105 -2x105 0 2x105 4x105 d) 10-4-4x10 -2x105 0 2x105 4x105
-4
10
103 103
|Current density| (A/m2)
10-5 10-5
|Current| (A)
|Current| (A)
10 2
102
10-6 10-6
10 1
101
10-7 10-7
100 100
10-8 10-8
-20 -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20
Figure 4.6. a), b) Resistance-voltage and c), d) current-voltage characteristics of the initialisation process carried
out at different voltage amplitudes in Ti/LSM20\Pt standard devices on a), c) LSM20/STO and b), d) LSM20/LAO
samples.
66
every five cycles in steps of 2 V up to 20 V, and then decreased back to 2 V. For these devices the
highest applied voltage of 20 V corresponds to an electric field of 4·105 V/m. However, as previously
explained, the largest contribution to the device’s total resistance comes from the RC2 component, which
corresponds to the depleted LSM (under the electrode) and the LSM/Ti (TiOx) interface, implying that
the electric field drop will not be uniformly distributed between the electrodes, but concentrated at the
region close to the Ti electrode. The resistance state is measured at +0.1 V after each half cycle. All the
measured devices exhibit hysteretic behavior at large voltage, but not always at small voltage. Here we
have considered that the devices either do not present memristive behavior or that their memristive
response is unreliable when the HRS/LRS ratio at small voltage (+0.1 V) is smaller than 1.3 and/or the
ratio is very variable, with no voltage/cycle dependence, usually due to noisy measurements. For
LSM20/STO, more than 75% of the devices of all sizes (1002 to 4002 μm2) showed memristive behavior.
In the case of LSM20/LAO, reliable switching was found to occur more likely in small devices
(A=1002 and 2002 μm2), with more than 75% of success. On the other hand, reliable switching in large
devices was obtained for 20% (3002 μm2) and 50% (4002 μm2) of the cases.
Figure 4.6 presents the R(V) and I(V) characteristics of the initialization process at increasing writing
voltage amplitude for LSM20/STO- and LSM20/LAO-based standard devices. For simplicity, the R(V)
characteristics after initialization, at the largest voltage range, are shown in Figure 4.7a and b for
LSM20/STO and LSM20/LAO, respectively. The curves are CCW in both cases but with different
shape, especially in the positive region. The R-I(V) curves in LSM20/STO are asymmetric in both
polarities with a negative differential resistance (NDR) region in the LRS between +5 and +10 V
(Figure 4.6a, c), similar to those reported for La0.8Sr0.2MnO3/STO by C-AFM.[93] The R-I(V)
characteristics of the Pt/Ti/LSM20\Pt planar devices on LSM20/LAO show a similar shape in both
a) Electric field (V/m) b) Electric field (V/m)

-4x105 -2x105 0 2x105 4x105 -4x105 -2x105 0 2x105 4x105

Resistance (Ω)
Resistance (Ω)
107 107
106
106
-20 -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20

Figure 4.7. Resistance-voltage (-electric field) characteristics of initialized Ti/LSM20\Pt standard devices
fabricated on the films a) LSM20/STO and b) LSM20/ LAO films showing gradual SET and RESET transitions.
67
polarities with a large increase of resistance at small applied voltage for the HRS and the LRS
(Figure 4.7b). Similar curves have been reported for GdBaCo2O5/LaNiO3 devices using Ag and Pt
electrodes,[139] and for La0.67Sr0.33MnO3/LAO using two Ag electrodes,[22] where the formation of an
interfacial layer due to surface defects acts as trapping/detrapping center modifying the Schottky barrier
and the resistance.
Figure 4.8a and b presents for LSM20/STO and LSM20/LAO, respectively, the evolution of the HRS
and the LRS as a function of the total number of cycles (bottom axis) and the writing voltage amplitude
(top axis) during the initialization process (open symbols), and applying the same voltage protocol a
second time (solid symbols).
Devices on LSM20/STO: During the initialization process represented by open symbol in Figure 4.8a,
the HRS and the LRS states are distinguishable only at Vwrite=12 V (cycles 1 to 25), after an abrupt
resistance increase occurring during the RESET process. At this voltage, the HRS/LRS ratio is 1.9.
Then, the evolution of the resistance states with increasing writing voltage is mainly dominated by an
HRS increase until a Vwrite of 20 V, where the HRS/LRS ratio is around 3. Next, as the voltage amplitude
Write voltage amplitude (V)

2 4 6 8 10 12 14 16 18 20 20 18 16 14 12 10 8 6 4 2
108
a)
Read resistance (Ω) Read resistance (Ω)
LSM20 / STO
HRS
107
LRS
106
108
b) LSM20 / LAO
HRS
107
LRS
106
0 10 20 30 40 50 60 70 80 90 100
Number of cycles
Figure 4.8. Evolution of the HRS and LRS resistance states measured after each cycle (bottom axis) and as a
function of writing voltage (top axis) for standard devices on a) LSM20/STO and b) LSM20/LAO. The resistance
was read at + 0.1V after each half cycle. The devices are initialized during the first experiment at which the writing
voltage amplitude is increased from 2 V to 20 V and decreased back (open data points). The experiment was
carried out a second time (solid data points) to assess reproducibility.
68
is decreased (cycles 51 to 100 in Figure 4.8a), the HRS shows a staggered decrease, as expected for a
multilevel capacity memory, but the LRS remains practically constant at Vwrite smaller than 16 V, with
HRS/LRS ratio of 2.2. The same procedure was carried out a second time (solid data points in
Figure 4.8a). The resistance evolution is similar, however the final resistance states measured during
the decreasing voltage steps are higher than during the first experiment. A similar increasing trend was
observed in a 3rd and 4th repetition of the complete cycling procedure (not shown). It can be thus
concluded that the Pt/Ti/LSM20\Pt planar devices on LSM20/STO do not show reproducible behavior
at high operating voltages (e.g. 20V) as the resistance states programmed at the same voltage amplitude
constantly increase by cycling. However, as will be later shown through pulsed measurements, these
devices are able to work at lower amplitude voltages (e.g. 10V).
Devices on LSM20/LAO: For LSM20/LAO-based devices, the value of the HRS drastically increases
at low voltages during the initialization procedure, with two well differentiated states (HRS/LRS=1.6)
already at 10 V (Figure 4.8b, open symbols). The HRS/LRS ratio increases with voltage amplitude
reaching a value of HRS/LRS=5 at Vwrite=20 V, for the device here presented. The increase in HRS/LRS
ratio is due to a simultaneous increase in HRS and a decrease in LRS with voltage amplitude. More
interestingly, both states show a staggered evolution towards an intermediate resistance value when the
voltage amplitude is decreased. Repeating the same voltage cycle procedure, the resistances states (solid
symbols in b) are perfectly reproducible.
The good reproducibility and multilevel capacity of the devices on LSM20/LAO was further confirmed
by cycling the same device at varying voltage amplitude: 10 cycles at 20 V, followed by 35 cycles at
10 V, and finally 10 cycles at 20 V. The HRS and the LRS obtained in this biasing experiment are
plotted in Figure 4.9 where the patterned area corresponds to the states obtained at Vwrite=±20 V and
±10 V in the experiment described in Figure 4.8b. We can conclude that, once initialized, the
LSM20/LAO-based device leads to HRS and LRS values that are function of Vwrite, stable over large
number of cycles, and are reproducible regardless the history. In the device presented, the average HRS
and LRS at Vwrite=10 V are 23±1 MΩ and 8.4±0.1 MΩ (HRS/LRS=2.7), respectively, while at
Vwrite=20 V they are 28±4 MΩ and 6±1 MΩ, respectively (HRS/LRS=4.6). The devices on
LSM20/LAO show reproducible performance as long as the applied voltage is equal or smaller than the
largest voltage ever applied. In the opposite case, the HRS and the LRS values written at certain voltage
will be redefined and expected to be reproducible. To illustrate that, consider the case of a device
initialized at 10 V: the HRS at 10 V (open symbols cycle 20-25 in Figure 4.8b) is different than that of
the device initialized at 20 V (solid symbols cycle 20-25 in Figure 4.8b). Hence, it can be concluded
that devices on LSM20/LAO show multilevel capacity and reproducible performance at high and
low amplitude voltages.
69
Write voltage amplitude (V)

20 10 20 10 20
4x107
HRS
Resistance (Ω)
107
LRS
4x106
0 5 5 5 10 15 20 25 30 35 40 45 50
Number of cycle
Figure 4.9. HRS and LRS obtained at Vwrite=±20 V (solid data points) and ±10 V (open data points) at different
times for LSM20/LAO. Patterned region correspond to the states achieved at 20 V and 10 V during the second run
of the initialization experiment (Figure 4.8b solid symbols). Non shaded area correspond to consecutive cycles at
20 V, 10 V and 20 V. The breaks in x-axis and reset of number of cycle indicate non-consecutive cycles.
In addition, pulsed measurements were also carried out on both LSM20/STO and LSM20/LAO
memristive standard devices. The writing procedure consisted of 60 ms pulses at Vwrite=±10 V, followed
by a read measurement, as schematized in Figure 4.10a. This voltage amplitude was selected from the
previous I(V) cycling measurements (see Figure 4.7), so that the LSM20/STO would also operate
reversibly and in a reproducible manner. Figure 4.10b and c show the resistance states read for
Figure 4.10. Pulse operation of the Pt/Ti/LSM20\Pt memristive devices using ±10 V/60 ms writing voltage pulses.
Evolution of HRS and LRS at +0.1 V over 100 consecutive bipolar pulses for devices on a) LSM20/STO and
b) LSM20/LAO.
70
LSM20/STO and LSM20/LAO, respectively, for 100 bipolar consecutive pulses. Both devices show a
stable HRS with cycling, whereas the LRS increases during the first cycles and remains quite stable
after around 10 repetitions. The HRS/LRS ratio for LSM20/STO and LSM20/LAO at the end of the 100
bipolar pulses (Vwrite=±10 V) is around 1.3 and 3, respectively. These ratios are almost two times smaller
than the values obtained by performing I(V) voltage sweeps at the same voltage amplitude (i.e. 0 V 
+10 V  0 V  -10 V  0 V). In both cases, the smaller ratios are induced by larger LRS during pulse
operation than in voltage sweep, 9.9 MΩ and 5.8 MΩ, respectively for LSM20/STO, and 45 MΩ and
35 MΩ, respectively for LSM20/LAO, while the HRS is similar in both operating conditions and
devices. Hence, asymmetric voltage pulses with a larger |Vset| to achieve a complete LRS could provide
similar results to those obtained by voltage sweep. Asymmetric sweep voltage operation for
LSM20/STO-based devices is discussed in Section 4.2.2.5.
A key parameter for non-volatile memories is the retention time, defined as the time that a state remains
in its resistance value, with no applied voltage, so that the ‘data’ (bit) stored is not lost. [140] Next, the
retention was assessed for two standard pristine devices on LSM20/STO and on LSM20/LAO. The
devices were first initialized, and then cycled 5 times at ±20 V before programming each desired
resistance state by voltage sweep. The R(V) characteristics and HRS/LRS ratio values are representative
of each sample. The HRS/LRS ratio prior retention measurements is 3.1 for HRS and 2.4 for LRS in
LSM20/STO devices, and 23 and 18, respectively, in LSM20/LAO devices. Next, the evolution of the
resistance with time, measured at +0.1 V for 5 hours, is presented in Figure 4.11a. For both devices, the
Figure 4.11. a) HRS and LRS evolution with time of Pt/Ti/LSM20\Pt memristive devices programmed at ±20 V
on LSM20/STO (solid line) and LSM20/LAO (dashed line). b) Multi-states resistance evolution with time, for the
HRS programmed at Vreset=+10 and +20 V, and the LRS at Vset=-10 and -20 V, in Ti/LSM20\Pt fabricated in
LSM20/LAO. The resistance is measured at +0.1 V under constant bias.
71
Contact side (μm)

100 200 300 400
109
LSM20 / LAO HRS - Vwrite 6V LRS - Vwrite 6V
HRS - Vwrite 10V LRS - Vwrite 10V
Read resistance (Ω)

108
107
slope = -1
104 105
Area (μm2)
Figure 4.12. Area dependence resistance of the HRS (blue shade, top) and LRS (red shade, bottom) at different
writing voltages for Ti/LSM20\Pt devices in LSM20/LAO. The data corresponds to one device for each size.
resistance states present certain volatility before stabilizing. The resistance state relaxation curves were
fitted to a two-phase exponential decay. For the LRS relaxation, similar characteristic times were
obtained for both samples, i.e. τ1 ∼ 10 min and τ2 ∼ 3 h, which suggest that the same phenomena could
be taking place in both devices. However, the HRS resistance decay in the LSM20/STO device is
somewhat faster than in the LSM20/LAO device, with characteristic times of τ1 ∼ 8 min, τ2 ∼ 1 h for
LSM20/STO devices, and τ1 ∼ 20 min, τ2 ∼ 6 h for LSM20/LAO devices. Furthermore, as shown in
Figure 4.11b, the stability of four different multi-states has been confirmed for the LSM20/LAO
optimized devices, corresponding to two HRS states (programmed at +10 V and at +20V) and to two
LRS (programmed at -10 V and at - 20V).
Finally, in order to evaluate the area dependence of the switching of the multilevel and reproducible
Ti/LSM20\Pt devices in LSM20/LAO, the HRS and the LRS was measured for devices with different
contact area ranging from 1002 to 4002 μm2. Both states were programmed at different writing voltages,
as presented in Figure 4.12, showing a clear resistance decrease by increasing the pad size, as expected
for interface-type switching. Whereas the HRS shows a linear dependence with area and a slope of -1,
the LRS shows a weaker dependence, which might be related to a non-homogeneous tunneling effect in
the case of the larger size devices. The electroforming-free nature of the devices, together with the HRS
and LRS dependence with device area, and the planar geometry of the device strongly support the
existence of interface-type resistive switching (non-filamentary).[18] No clear trend is observed in the
devices on LSM20/STO because of the low reproducibility.
72
Discussion: oxygen migration modelling
The Pt/Ti/LSM20\Pt memristive devices fabricated on LSM20/STO and LSM20/LAO showed

clockwise log I(V) characteristics with a gradual change in resistance, showing that the HRS and LRS
can be tuned by the voltage amplitude of the writing process. The use of Ti as oxygen reservoir and the
ability of LSM20 to accommodate different oxygen content through the Mn redox reaction are at the
origin of the change in resistance observed in the resistive switching. On one hand, the role of Ti
electrodes as an oxygen reservoir, via the formation of an insulating oxygen deficient TiO x interlayer
when put in contact to an oxide, has been reported to trigger the switching response in Pr 1-xCaxMnO3
(PCMO)–based devices.[109], [110] The n-type Ti/TiOx interlayer acts as a Schottky barrier when put in
contact with the p-type LSM20, where the width and height of the barrier are determined by the oxygen
content, and hence, changing the electronic conducting properties of the interface. This interlayer
(Figure 4.3) would facilitate oxygen exchange from (to) the LSM20 films with the concomitant
formation (annihilation) of oxygen vacancies in the perovskite. On the other hand, as introduced in
Section 1.3, the conductivity of LSM20 is determined by the Mn3+/Mn4+ ratio which can be modified by
the oxygen content in LSM20. Then, the presence of an oxygen-depleted LSM20 region decreases the
hole-carrier concentration, increases the band-gap, and thus, the measured resistance rises, as for
PCMO.[109], [110] This hypothesis is in agreement with the drift of negatively charged oxygen ions towards
(away from) the positively (negatively) biased Ti electrode during HRS RESET (LRS SET) operation
(see sketch in Figure 4.13).
This model can explain the memristive behavior of the LSM20/LAO devices (Figure 4.13d-f). The
resistance states are very reproducible and pivot around an intermediate resistance value (Figure 4.8b)
with HRS/LRS ratios between 3 and 23. Therefore, the multilevel resistances could be related with the
gradual oxidation (reduction) of TiOx and increasing (decreasing) oxygen vacancy concentration in
LSM20 for the HRS (LRS). This mechanism is fully reversible under the symmetric operating voltage
used in this work (up to 20 V). However, it can be observed that the LSM20/STO devices only present
resistive switching behavior after a relatively high RESET process (at + 12V). We suggest that the large
increase in resistance occurring at this stage is related to the formation of oxygen vacancies in LSM
where the oxygen ions would be incorporated in the TiOx interlayer (with increase of thickness or further
oxidation). The presence of oxygen vacancies in LSM would allow for the resistive switching behavior
with two well differentiated HRS and LRS states, which occurs thereafter. This observation highlights
the importance of the initialization step in creating oxygen vacancies to trigger the switching. This
hypothesis is supported by the work of Luo et al.[95] where resistive switching in Pt/LSM20\Pt devices
was only obtained for reduced films (with presence of oxygen vacancies), while no switching and ohmic
73
a) b) LRS HRS c) HRS LRS

+ + -
- Pt Pt
Ti / TiOx Ti / TiOx
O2- O2-
𝐿𝑎 𝑆𝑟 2 𝑀𝑛𝑂
LSM20 LSM20
STO STO
𝑇𝑂
d) e) LRS HRS f) HRS LRS
𝑂2
+ Pt
- Pt
𝑉𝑂
Ti / TiOx Ti / TiOx
O2- O2-
LSM20
LSM20 LSM20
LAO LAO
Figure 4.13. a) Sketch of a Pt/Ti/LSM20\Pt memristive device for LSM20/STO films, and d) sketch of the half-
[135]
loops dislocations present in LSM20/LAO films, in agreement to reference . Cross-section scheme of the
oxygen drift at the Ti/LSM20 interface in the memristive devices during operation at large voltage (i.e. ±20 V).
Scheme of oxygen drift during HRS and LRS programming b), c) in LSM20/STO and d), e) LSM/LAO-based
devices. In all cases the bias is applied on the Pt/Ti electrode while the Pt electrode is grounded. We note that the
extended defects may present different lengths and orientations, and do not necessarily always connect both
electrodes as it may be derived from Figure 4.13d.
behavior was obtained for oxygen-stoichiometric LSM20-based devices. In addition, smaller ratios are
measured for the LSM20/STO devices, where the resistance states are only reproducible at intermediate
voltage ranges (up to 10 V), while at higher voltages both states increase with operation (Figure 4.8a)
due to the continuous increase of LRS. This suggests that for the LSM20/STO devices the HRS reached
at large operating voltage cannot be fully reversed by applying the same voltage of opposite polarity to
obtain the LRS (Figure 4.13b and c).
The differences observed between the two kinds of samples can be explained by differences in oxygen
mobility. As a first approximation, we compare the high-temperature oxygen diffusion coefficients
reported in the literature for LSM20 although, for our devices, the voltage (field) would be the driving
force resulting in an oxygen drift instead of thermal diffusion. Similar bulk oxygen diffusion coefficients
have been measured at 600 °C for LSM20 films around 130 nm thick grown on LAO and STO.[63], [135]
Nonetheless, the presence of space-charge tubes around dislocations has been reported to be the cause
of the enhanced oxygen diffusion along these defects,[63] as experimenally measured by other groups in
epitaxial thin films,[135] and along grain boudaries.[62], [73] In terms of oxygen migration, a device in the
LRS would reset to the HRS by the drift of oxygen towards the positively biased Ti electrode as a result
of the electrical bias.[20] This operation is favored in the case of LSM20/LAO due to the presence of
74
many extended defects, such as dislocations, with a high concentration of oxygen vacancies, that would
act as fast oxygen diffusion paths (sketched in purple in Figure 4.13e-f), leading to a larger HRS/LRS
window. On the other hand, during the LRS set process, the re-incorporation of oxygen and drift within
LSM20 might be limited for the LSM20/STO devices due to the high quality of the epitaxial LSM20
film, which does not present many extended defects (Figure 4.13c).[62], [63], [135]
Eventually, the
concentration of oxygen in TiOx and oxygen vacancies in LSM20/STO will increase with cycling raising
the total resistance of the device.
As for the relaxation process, it was measured at +0.1 V and thus an oxygen drift is not expected.
However, a short-scale transient fickian diffusion and/or a redox reaction at the interface may occur as
a result of the chemical gradient induced during operation. The increase of resistance of the LRS can be
considered as either a small reincorporation of oxygen from the LSM back to the Ti via a redox reaction,
or a gradient re-equilibration within the layers. These phenomena would be similar for both types of
devices, LSM20/STO and LSM20/LAO, as similar characteristic times were found. Similarly, the
relaxation of the HRS shows a decay of resistance, in this case faster in LSM20/STO devices, that could
be related to the back-exchange of oxygen from TiOx into the LSM film or a re-equilibration within the
layers. The differences between sutbstrates in this case might be related to local differences in
stoichiometry at close proximity of the interface after the applicaton of the RESET voltage.
Summary on epitaxial LSM20 as memristive material
Epitaxial thin films of La0.8Sr0.2MnO3-δ with different nanostructural details have been grown by pulsed
injection metal-organic chemical vapor deposition on SrTiO3 and LaAlO3 perovskite substrates. The
negligible strain present in the LSM20/STO films and the substrate-induced strain retained in
LSM20/LAO, leading to an out-of-plane cell enlargement of +1.33% was corroborated by X-ray
diffraction and HRTEM. LSM20/STO showed perfect epitaxial growth with long-range atomic order
and practically no defects, while LSM20/LAO showed the presence of extended defects, such as
dislocations. By microscopy and spectroscopy techniques, the spontaneous formation of TiO x at the
Ti/LSM interface was measured in pristine devices, likely resulting from oxygen-scavenging of the Ti
electrode and the concomitant oxygen vacancy formation in the LSM film, which introduces a
significant contribution to the initial resistance value.
The gradual clockwise log I(V) characteristics, and the HRS and LRS electrode size dependence support
the existence of interface-type switching, where the tunable resistance states programmable at different
writing voltages would be related to different levels of oxidation and reduction of LSM20 and Ti that
modifies the Schottky-like pn interface in these VCM memristive devices. Reproducible performance
is achieved in both types of devices by operating at intermediate voltages, i.e. ±10 V, as shown by pulsed
measurements over at least 100 cycles. HRS RESET occurs when a positive voltage is applied to the Ti
75
electrode, and hence oxygen drifts from LSM20 toward the Ti electrode, which is oxidized. LRS is
obtained under negative polarization, where oxygen drifts back into “bulk” LSM20. While the films
resistivity, topography and morphology were similar in both samples, the Ti/LSM20\Pt planar devices
showed substrate-dependent memristive behavior, particularly for high voltage operation ranges
( 20 V). Whereas LSM20/STO shows an increase of the LRS with continuous cycling under bipolar
symmetrical operation, LSM20/LAO exhibits reproducible performance and larger operating windows.
The presence of extended defects, such as dislocations, observed by HRTEM in LSM20/LAO, would
allow faster oxygen drift through LSM20. Instead, the absence of extended defects in LSM20/STO
would hinder the drift of oxygen into the LSM20, resulting in an incomplete SET to intermediate
resistance states, and leading to a gradual increase of the LRS. As for the retention, both states
experience a small relaxation that might be related to an oxygen back-diffusion process at the Ti/LSM20
interface as a consequence of the chemical gradient induced during operation. The implications of these
results could also be extended to other perovskite systems and technologies, such as micro solid oxide
fuel cells and electrolysers. Hence, engineering the nanostructure by introducing extended structural
defects which enhance the oxygen mobility can result in an effective way of improving the performance
of different nanoionic electrochemical devices.
Epitaxial and polycrystalline texture in thicker LSM20 films

As presented in the previous section, the slightly different cell parameters between the materials of the
film and the substrate (mismatch) is at the origin of the strain generated in the epitaxial thin films. The
strain is typically accommodated by the distortion of the lattice, known as biaxial strain. However, the
energy required to hold the strain imposed by the substrate can exceed the formation energy of defects
and lead to their nucleation and growth at expenses of retaining the elastic deformation. This event is
referred to as strain relaxation and it is more likely to occur in thicker films where the total energy of
the strained lattice is large.[135] The relaxation usually takes place by formation of point defects, and
dislocations (edge, screw and misfit dislocations) at an extended scale. On the other hand, when the
lattice mismatch of the film and the substrate is very large (or there is no translation symmetry along
their surfaces), no orientation relationship is expected between the film and the substrates. [141] In this
case, the layers are formed by randomly oriented grains, between which grain boundaries are formed,
leading to polycrystalline films. As discussed in Section 1.3, the change in microstructure may alter the
properties of the material and eventually the functional properties of the devices. In this section, we
investigate the role of thickness and texture in La0.8Sr0.2MnO3-δ (LSM20) thin films on the memristive
characteristics. Epitaxial LSM20 thin films were grown on STO and LAO substrates while
polycrystalline thin films of different thickness were grown on Si3N4. The structure, microstructure and
76
4.2 Epitaxial and polycrystalline texture in thicker LSM20 films
Table 4.2. Description of LSM20 films grown on, STO and LAO single crystal substrates and on Si3N4. The films
are grouped by their thickness (grown in the same deposition). Film properties are given when available: resistivity
measured by TLM, roughness obtained by 5×5 μm2 scans in AFM, out-of-plane parameter calculated from XRD,
and experimental out-of-plane strain, εzz. Most of the samples exhibit one single contribution of the cell parameter
while two contributions are observed for thick films, i.e. strained (S) and relaxed (R). The strain is calculated using
Equation 4.3, where cbulk is the bulk cell parameter of LSM20, 3.898 Å.[123] The samples are further grouped by
the time when they were grown: before or after a technical maintenance of the reactor.
Section
Thickness Roughness, Out-of-plane Experimental Date of
in Sample Pulses Substrate
(nm) Sq (nm) parameter (A) strain, εzz, (%) growth
chapter
STO 0.57 n/a n/a Before reactor
4.1 LSM20-7 7 3600
LAO 0.58 3.95 +1.33 maintenance
STO 0.56 3.890 ± 0.004 -0.21
4.2.2.
LSM20-8 8 1800 LAO 0.42 4.00 ± 0.03 +2.62
1-3
Si3N4 0.67 3.89 ± 0.01 n/a
STO 0.56 3.888 ± 0.003 -0.26
4.2.2.
LSM20-21 21 3600 LAO 0.48 3.993 ± 0.006 +2.44
4-6
Si3N4 1.56 3.894 ± 0.002 n/a After reactor
maintenance
STO n/a 3.889 ± 0.003 -0.23
R: 3.896 ±
0.001
4.3 LSM20-88 88 15000 LAO n/a S: -1.26
S: 3.947 ±
0.001
Si3N4 n/a 3.90 ± 0.02 n/a
memristive characteristics observed in these films are discussed in this section. As will be shown, the
polycrystalline LSM20 films presented here arise as promising candidates for memristive devices as
they are grown on Si-based substrates and show good switching performance.
Description of the samples
The films presented in this section were grown on three types of substrates; epitaxial films were grown
on single crystal substrates, namely STO and LAO, whereas polycrystalline films were obtained on
Si3N4–coated Si substrates (300 nm Si3N4/ 100 nm SiO2/ Si). These samples were grown using the same
deposition conditions as those presented in the previous section. However, it should be noted that a
technical maintenance of the reactor was necessary between the depositions of the thinnest samples
(presented in Section 4.1) and the thicker ones (presented in the current section), and as a result the
deposition yield was modified. Nonetheless, the resistivity measured on the thicker LSM20 films is
closer to the values reported in literature.[51], [52], [93], [134], [142]
Table 4.2 summarizes the main
characteristics of the samples discussed in this section. In order to facilitate comprehension, the samples
77
are grouped and presented by the number of pulses used for the growth of the film and hence, the
thickness. As the growth on each of the selected substrates was performed simultaneously during the
same deposition, we infer that the film thickness is similar for all substrates. In this section, the samples
will be referred by the measured thickness and the corresponding substrate. The samples discussed in
the previous section, i.e. 7 nm thick LSM20 films (LSM20-7), are included in the table for the sake of
comparison, together with samples LSM20-8, LSM20-21 and LSM20-88, corresponding to a thickness
of 8, 21, and 88 nm, respectively.
Comparison between epitaxial and polycrystalline LSM20

memristive devices
4.2.2.1. Structural characterization of epitaxial and polycrystalline LSM20-8

thin films
Thin films were deposited simultaneously on Si3N4 substrates (for polycrystalline films), and on STO
and LAO substrates (for epitaxial films). The thickness of the batch of LSM20 films is around 8 nm
(LSM20-8), as measured X-ray reflectometry (XRR) on a polycrystalline sample equivalent to LSM20-
8/Si3N4 sample. The experimental XRR data and the fit result are presented in Figure 4.14. The sample
used for fitting consisted of (i) a LSM20 layer which thickness and roughness were defined as fitting
variables, and (ii) the substrate comprising ∼300 nm thick Si3N4, ∼100 nm thick thermal SiO2 and
500 μm thick Si, for which only the thickness was fitted. The thickness and roughness, root mean square
1010
Experimental
109 Fit
Critical angle
108
Intensity (counts)
107
106
105
104
103
102
0 1 2 3 4 5 6
2 theta (°)
Figure 4.14. XRR scan of a LSM20-8/ Si3N4 sample and the fit result. The fitting indicates a thickness of 7.90 nm.
78
height, Sq, of the layer obtained after the fitting is 7.90 nm and 1.04 nm, respectively, as schematized in
the inset of Figure 4.14. As the maxima and minima of the fitted curves match those of the experimental
data, we consider that the obtained thickness obtained is reliable. However, the lower intensity of the
fitted curve next to the critical angle indicates a larger experimental density than the one used for the
fitting (6.521 g.cm-3 from ICDD-PDF 00-53-0058).
The LSM20-8 films grown on the different substrates, i.e. LAO, STO and Si3N4, were characterized by
means of SEM and XRD, as presented in Figure 4.15. From the SEM images, it can be observed that
all the films are dense. The LSM20-8/STO and LSM20-8/LAO films exhibit small grains of 20 ± 3 nm
and 28 ± 4 nm diameter, respectively, in agreement with the results found for the LSM20-7 film (21 nm
for LSM20-7/STO, and 29 nm for LSM20-7/LAO). On the other hand, the grains found in LSM20-
8/Si3N4 are indeed related to different crystallographic orientations, and hence, this film presents grain
boundaries, as shown by TEM in Figure 4.16. The average grain size is 22 ± 8 nm, where the large
dispersion of sizes arises from the irregular shape of the grains. Regardless the polycrystalline nature,
this film is flat with a roughness of Sq=0.67 nm, measured by AFM. Slightly smaller roughness values
are found for the epitaxial films with Sq equal to 0.56 and 0.42 nm for LSM20-8/STO and LSM20-
8/LAO, respectively.
The XRD (Figure 4.15a-b) patterns reveal preferential growth along LSM20 (00l)pc for the films grown
on STO and LAO. The out-of-plane lattice parameter, c, obtained from the XRD indicates a larger
degree of strain in these films in comparison to LSM20-7 ones. The out-of-plane strain, εzz, was
calculated using Equation 4.3 where cbulk is the bulk lattice parameter of La0.8Sr0.2MnO3 (3.898 Å[123]).
The value obtained for LSM20-8/STO shows slight in-plane tensile strain (cpc=3.89 Å, εzz=- 0.21%)
whereas that on LSM20-8/LAO proves larger in-plane compressive strain (cpc=4.00 Å, εzz=+2.62%).
𝑐𝑓𝑖𝑙𝑚 − 𝑐𝑏𝑢𝑙𝑘
𝜀𝑧𝑧 =
𝑐𝑏𝑢𝑙𝑘
Equation 4.3
The polycrystalline nature of the LSM20 films grown on Si3N4 substrates is confirmed by the presence
of all family planes peaks using XRD in Bragg-Brentano configuration (not shown). Grazing incidence
XRD (GI-XRD, Figure 4.15c) was used to minimize the high intensity peaks of the silicon substrate for
the sample LSM20-8/Si3N4. All the family planes could be identified in the diffractogram and no
secondary phase was found. The pattern of the polycrystalline sample was fitted using TOPAS code to
obtain the rhombohedral lattice parameters, a=5.51 ± 0.01 Å, and c=13.47 ± 0.05 Å; which are
comparable to the reference used in this thesis (ICDD-PDF 00-053-0058: a=5.529 Å and c=13.367 Å).
79
We attribute the difference in cell parameters to a poor fitting of the XRD (TOPAS is a complete and
Figure 4.15. SEM images and XRD diffractograms for LSM20-8 films grown on a) STO, b) LAO, and c) Si 3N4.
Morphology and film texture can be observed in the SEM images and XRD, respectively. The diffraction pattern
of c) was obtained in grazing-incident (GI-XRD) configuration to reduce the signal from the substrate. The signal
arising from the substrate is denoted by the inverse triangle (▼).
80
Figure 4.16. TEM cross-section of a) LSM20-8/Si3N4 and b) of the counter electrode stack Pt/LSM20-8/Si3N4.
complex software, and some fitting parameters we not taken into account). The calculated pseudo-cubic
lattice parameter of the film, apc=3.89 ± 0.01 Å, is in good agreement with the bulk values reported for
stoichiometric La0.8Sr0.2MnO3-δ, 3.898 Å[123] and hence demonstrating that there is no strain induced by
the substrate. We note that the TEM analysis in the polycrystalline sample were carried out after the
XANES measurements (see Section 6.4), so we attribute the large roughness observed in Figure 4.16a
to some mechanical damage during the transport and/or the experiment. Indeed, the apparent roughness
of the film protected by the Pt electrode layer (Figure 4.16b) is in agreement to the one obtained by
XRR (Sq=1 nm).
The electrical properties of the films were measured by four-point probe in-line measurements, as
summarized in Figure 4.17a. The resistivity values obtained for the LSM20-8 films depend on the
substrate and cover a range of three orders of magnitude. As expected, LSM20-8/STO film shows the
smallest value (∼12 mΩ·cm), followed by the film grown on LAO (∼100 mΩ·cm), whereas the
polycrystalline film, LSM20-8/Si3N4, exhibits the largest resistivity (∼2 Ω·cm). The value of resistivity
is greatly influenced by the presence of defects, such as oxygen vacancies, and slight deviations of
composition.[14], [51], [52], [93], [142] As reported for epitaxial La1-xSrxMnO3-δ films with x=0.30 and 22 nm
thick [58] and x=0.33 with 50 nm thick,[59] the resistivity increases with net biaxial strain. The discussion
on the role of defects and strain on electrical properties is presented in Section 1.3. In agreement with
the biaxial strain, the resistivity of LSM20-8/LAO is one order of magnitude larger than that on LSM20-
8/STO even though smaller values of resistivity, 7.7 mΩ·cm, have been reported in literature for 10 nm
thick La0.8Sr0.2MnO3 films on STO.[93] On the other hand, the large resistivity observed on LSM20-
81
-21
-88
LSM-7
8
M-
M
LS
LS
LS
101
LSM20 / STO
LSM20 / LAO
Resistivity (Ω.cm) LSM20 / Si3N4
100
10-1
10-2
10 100
Film thickness (nm)
Figure 4.17. Electrical resistivity of LSM20 thin films of different thickness (LSM20-7, LSM20-8, LSM20-21,
and LSM20-88) grown on STO, LAO and Si3N4 substrates.
8/Si3N4 could arise as a consequence of the polycrystalline texture, in which the grain boundaries can
act as scattering centers and high resistivity barriers.[54], [143]
4.2.2.2. Memristive behavior in LSM20-8 films
The memristive response of the films was measured on Ti/LSM20\Pt standard devices, similar to those
described in the previous section for LSM20-7. In addition, these films were chosen for X-ray absorption
near edge spectroscopy (XANES) measurements in synchrotron facilities to assess the Mn oxidation
state in different locations and for different resistance states. Hence, thinner electrodes were used to
increase the signal of the underlying LSM20 layer. The 10 nm thick Ti contact was capped with a thin
layer of 10 nm of Pt, as schematized in Figure 4.19d. More details about the sample, XANES
experiment, results and discussion are presented in the Section 6.4.
Initial resistance (IR) of the LSM20-8-based devices
The average initial resistance (IR) of standard devices built on LSM20-8 films was 1.0·106, 0.6·106, and
7·106 Ω for LSM20-8/STO, LSM20-8/LAO, LSM20-8/Si3N4 respectively, and is largely dominated by
the Ti/LSM20 interface component, accounting for more than 80% of the IR (Figure 4.18). As in
LSM20-7, the Ti/LSM20 interface is likely composed by a spontaneously formed TiO x interlayer and
reduced LSM20, which increases the width and/or height of the Schottky-like barrier. The value of IR
seems to be related to the resistivity of the film and the TiOx interlayer (and reduction of LSM20, as
82
21
21
LS 7
8
LS 7
8
M-
M-
M-
M-
M-
M-
a) b)
LS
LS
LS
LS
100
LSM20 / STO
LSM20 / LAO 90
107
RTi/LSM / IR (%)
LSM20 / Si3N4
80
IR (Ω)
6
10 70
60
105 LSM20 / STO
50 LSM20 / LAO
LSM20 / Si3N4
104 40
5 10 15 20 5 10 15 20
Film thickness (nm) Film thickness (nm)
Figure 4.18. Electrical device characterization of LSM20 thin films with different thickness (LSM20-7, LSM20-
8, LSM20-21, and LSM20-88) grown on STO, LAO and Si3N4 substrates. a) Average initial resistance (IR) of
Ti/LSM20\Pt standard devices (A=200x200 μm2, L=50 μm) and b) contribution of the Ti/LSM20 interface to the
IR obtained using Equation 4.2.
found by XANES, see Section 6.4). This interlayer can account for the difference observed in IR, as its
formation may depend on several parameters, for instance the surface roughness, surface orientation,
oxygen stoichiometry, presence of defects, microfabrication process and other uncontrolled parameters
during the evaporation (see more details in Section 2.2).
Epitaxial devices: LSM20-8/STO and LSM20-8/LAO
The R(V) characteristics at different voltage amplitude for standard devices (A=200x200 μm2, L=50 μm)
fabricated on the three samples are presented in Figure 4.19a-c for comparison. The HRS and LRS
resistance states, obtained after every cycle and as a function of writing voltage during the initialization
procedure are presented in Figure 4.20 for LSM20-8/STO and LSM20-8/Si3N4.
Bipolar memristive behavior with CCW log R(V) characteristics was observed in epitaxial LSM20-8
films at operating voltages, with no previous electroforming step required (Figure 4.19a and b). The
R(V) characteristics obtained in LSM20-8 are similar to their analogous (same substrate) in LSM20-7,
but shifted around one order of magnitude towards smaller resistance. LSM20-8/STO devices show a
NDR in the LRS state at around +5 V. As for LSM20-7/LAO, devices on LSM20-8/LAO present a sharp
increase of resistance at small voltages. Slightly smaller HRS/LRS ratios are found on LSM20-8/LAO
devices (2.0 and 2.5 at Vwrite 10 and 16 V), in comparison to those in LSM20-7/LAO (2.6 and 4.5 at Vwrite
10 and 16 V).
83
Figure 4.19. R(V) curves at increasing voltage amplitude for Ti/LSM20-8\Pt standard devices for the films grown
on a) STO, b) LAO and c) Si3N4 substrates. d) Schematic of the devices measured in these samples.
The initialization procedure described in Section 4.1.2 was carried out on LSM20-8/STO devices,
(Figure 4.20a). Two different resistance states can be observed from Vwrite= ±4 V. While the HRS
increases with increasing voltage amplitude, the LRS remains quite constant up to 18 V, where the
largest HRS/LRS ratio of 5 is obtained. At 20 V, the LRS increases its resistance and keeps rising even
when the voltage amplitude is decreased (cycle 51 to 100 in Figure 4.20a). At 18 V, the HRS decreases,
as expected from a multilevel capacity memory, but the HRS/LRS ratio drastically drops to a value of
2. The resistance states cannot be clearly distinguished at smaller voltages. This behavior matches that
presented for LSM20-7/STO, where LRS cannot be fully set at high voltages and leads to narrowing of
the HRS/LRS window. The increase of resistance could be related to the irreversible incorporation of
oxygen into the Pt contact,[144] decreasing the amount of charge carriers by the reduction of Mn4+ to
Mn+3. Overall, and regardless the differences pointed out in the films properties and in IR, we can
conclude that memristive performance in very thin epitaxial LSM20 is comparable for thicknesses of 7
and 8 nm. Hence, we move the focus to polycrystalline LSM20.
84
Figure 4.20. Evolution of HRS and LRS (left axis) and HRS/LRS (right axis) with cycles (bottom axis) and writing
voltage amplitude (top axis) for devices on a) LSM20-8/STO and b) LSM20-8/Si3N4. The resistance was measured
at Vread=+0.1 V. Resistance is plotted in different range for a) and b). The HRS/LRS is displayed to ensure clarity
in each plot.
Polycrystalline devices: LSM20-8/Si3N4
The same initialization procedure applied on LSM20-8/STO devices was carried out on devices
fabricated on the polycrystalline film. The R(V) curves, at increasing voltage amplitude, and the
measured resistance states of the initialization are presented in Figure 4.19c. The standard devices on
LSM20-8/Si3N4 show hysteretic response, CCW and gradual R(V) characteristics at all operating voltage
tested without any previous electro-forming step required, similar to the epitaxial LSM20 films. These
characteristics suggest an interface-type resistive switching. The R-I(V) characteristics show NDR for
LRS at voltages close to positive voltage amplitude, +Vamp. The characteristics also present a sharp
increase of resistance at small voltage, similar to the one observed in LSM20/LAO (see Figure 4.6b and
Figure 4.19b). Two different resistance states are observed at Vwrite as low as 4 V (HRS/LRS ≃ 1.5).
The HRS increases with cycles at constant voltage, particularly for 2 V < Vreset < 10 V, and with voltage
amplitude (cycles 1 to 50 in Figure 4.20b), reaching values of 3.9·108 Ω at Vwrite=±18 V. On the other
85
hand, the LRS becomes smaller at increasing voltage amplitude up to 14 V. A slight increase of LRS
with cycles at same voltage amplitude is also observed, and it becomes dominant at |Vwrite|> 14 V. The
LRS resistance value keeps increasing until 3.3·107 Ω, where it stabilizes after a few cycles at 20 V
(cycles 51 to 55 Figure 4.20b). Given the sequence of the sweeps, where at certain voltage amplitude
the HRS is programmed first and then switched to the LRS, the increase of resistance observed for the
LRS might be related to the increase previously induced in the HRS. In the second half of the
initialization process, cycles 51 to 100, the HRS decreases and the LRS increases with decreasing
voltage amplitude towards an intermediate resistance value. The HRS/LRS ratio shows a staggered
decrease from 8 to 1.4, for Vwrite from 20 to 2 V. A similar behavior was described for the LSM20-
7/LAO devices, and hence, both systems show potential application as multilevel resistive random
access memories. Given the similarities between the LSM20/Si3N4 and the LSM20/LAO devices, we
would expect the system to show a reproducible response similar to the last half (cycles 51 to 100) if no
larger voltage amplitude is applied. At the light of the results obtained, we would suggest to limit the
operation voltage for LSM20-8/Si3N4 to 12-14 V. At these voltages, the HRS/LRS ratio lays around 12
– 17, and both resistance states are stable with cycles. Operating in these conditions would allow a large
operating window (> 10), possibility of multilevel storage, and reproducible performance with stable
HRS and LRS.
Comparison between epitaxial and polycrystalline LSM20-8
The standard devices fabricated on the three LSM20-8 films show a memristive response with CCW
and gradual R(V) characteristics without any previous forming step required. The same features were
described for epitaxial LSM20-7-based devices. The R(V) curves of the devices in LSM20-8/STO and
in LSM20-8/Si3N4 show a NDR for the LRS at small (ca. +5 V) and large (close to +Vamp) voltages,
respectively, which could be related to the onset of LRS to HRS transition. However, in LSM20-8/STO,
the resistance curve of the HRS flattens when it approaches 0 V from both polarities, while in LSM20-
8/Si3N4 devices the HRS curve rises, as found in LSM20-7/LAO and LSM20-8/LAO (Figure 4.6b and
Figure 4.19b).
The HRS and the LRS in epitaxial and polycrystalline devices depend on the writing voltage applied,
except for the LRS in LSM20-8/STO, which is rather constant (Figure 4.20). The HRS in LSM20-
8/Si3N4 shows a larger variation with reset voltage than in LSM20-8/STO and hence, the HRS/LRS
ratios are larger in the former (e.g. at Vwrite=±12 V, HRS/LRS=12 and 3, for LSM20-8/Si3N4 and
LSM20-8/STO, respectively). On the other hand, the LRS increases continuously with cycling at large
writing voltage in LSM20-8/STO (Vwrite=±20) and LSM20-8/Si3N4 (±16 V). In both cases, this could be
related to a loss of electron hole carriers (and increase of resistivity) and non-fully reversible resistance
switching, which is more severe in LSM20-8/STO (with very small HRS/LRS ratio after initialization,
cycles 51 to 100 in Figure 4.20a) than in LSM20-8/Si3N4 (which still shows multilevel capacity). The
86
loss of electron hole carriers by formation of oxygen vacancies (which act as electron donor), might be
related to the formation of a stable TiOx layer. More details about the resistive switching mechanism
and oxygen model of these systems are given below.
Resistive switching mechanism model
The memristive response of the systems can be explained by an oxygen migration model, as the one
applied for LSM20-7 (Section 4.1.3). In agreement with the CCW log R(V) characteristics and the bias
configuration (Ti electrode biased and Pt electrode grounded), positive writing voltage triggers HRS
RESET by the drift of the oxygen from the LSM20 towards the Ti: the Ti will partially oxidize into an
oxygen deficient TiOx interlayer (increasing the width and/or height of the Schottky barrier, formed by
n-type Ti/TiOx and p-type LSM20) and the LSM20 will reduce (decreasing the hole carrier
concentration).[109], [110] As a result of the redox reaction, the measured resistance of the system increases.
On the other hand, the LRS would be reached under negative writing voltage by the reverse redox
reaction, where the LSM20 is re-oxidized and TiOx reduced. We can infer that interface-type resistive
switching is triggered by the modification of the Ti/LSM20 interface, where the thickness and degree of
oxidation and reduction of TiOx and LSM20 is gradually tuned with voltage.
LSM20-8/STO: As seen by XRD, we can infer a high epitaxial quality LSM20-8/STO layer where the
main path for oxygen to migrate is the bulk of the film. Due to the larger oxygen affinity of Ti over Mn
(LSM20),[23], [109] the redox reaction occurring during the HRS RESET might be more favorable than the
one of the LRS SET. In agreement with this, continuous switching cycles at large voltage may result in
a net accumulation of oxygen in the TiOx interlayer, and hence, the resistance will continuously increase,
as observed for the LRS in LSM20-8/STO devices (Figure 4.20a). Consequently, in an extreme scenario
where the oxygen remains trapped in the TiOx interlayer, the HRS and the LRS would converge towards
the same resistance, as seen in cycles 51 to 100 in Figure 4.20a.
LSM20-8/Si3N4: The polycrystalline film, LSM20-8/Si3N4, presents a high density of grain boundaries,
as detected by SEM (Figure 4.15) and TEM (Figure 4.16). Grain boundaries have been reported to
enhance oxygen diffusion in manganites at high temperature.[62], [73]
Hence, the oxygen ions in
polycrystalline LSM20-8/Si3N4 are expected to migrate both through the bulk and along the grain
boundaries. Indeed, a similar model to the one used for LSM20-7/LAO, which exhibits extended
structural defects such as dislocations, can be used for polycrystalline LSM20-8/Si3N4. At large
operating voltage, the HRS resistance in LSM20-8/Si3N4 (>108 Ω) is larger than in epitaxial LSM20-8
(HRS=106-107 Ω), and larger than in epitaxial LSM20-7 (107-108 Ω), which has a similar IR and
resistivity than the former. This suggests a larger amount of “moveable” oxygen which may come from
the interconnected network of grain boundaries while the structural extended defects in LSM20-7/LAO
87
are less and not interconnected (otherwise the film would be completely relaxed as explained by
Navickas et al. for the same system[135]). Therefore, the grain boundaries would allow the migration of
oxygen through the Ti/LSM20 interface in both directions, and consequently, the reversible change of
resistance. Nonetheless, the increase of HRS with voltage amplitude (cycles 1 to 50 in Figure 4.20b)
and with cycles (particularly for 2 V < Vreset < 10 V), suggests a more favorable drift of oxygen towards
the Ti electrode. On the other hand, the value of the LRS increases with the number of cycles for |Vset|≥
16 V, behavior previously observed in LSM20-8(-7)/STO devices, which involves the loss of electron
hole carriers in LSM20 due to an incomplete SET (oxygen progressively accumulates at the TiO x
interlayer). Another phenomenon that could explain the rise of resistance is the trapping of oxygen in
the Pt electrode, in agreement with Zurhelley et al.[144] This scenario can take place under large negative
voltage (LRS RESET), and it will increase the resistance by formation of an oxygen depleted LSM20
region at the Pt/LSM20 interface and the decrease of the electron hole carrier concentration.
The Ti/LSM20\Pt standard devices on polycrystalline LSM20-8/Si3N4 have shown HRS/LRS ratios that
would satisfy industry requirements (>10) while the HRS and the LRS are stable upon cycling. However,
the operating voltage amplitudes (±12 to ±14 V) are too high to raise industrial interest. Nonetheless,
the operating voltage is imposed by the geometry used (hundreds of microns distance between
electrodes) and the planar configuration. Smaller voltages with similar operating window could
potentially be achieved by using vertical configuration (distance between electrodes of a few
nanometers). This approach was tested within the framework of this thesis and LSM films were grown
on platinized silicon substrates (not included in this manuscript). Nonetheless, the temperatures required
for the growth of LSM (630-675 °C) provoke pin holes in the LSM20 layer as a result of dewetting of
platinum layer (bottom electrode). Thus, the LSM layers prepared were not suitable to build memristive
devices. A dedicated study would have to be done in the future with the aim of either changing to a more
stable bottom electrode, or optimizing the LSM deposition conditions to reduce the temperature.
4.2.2.3. Summary on epitaxial and polycrystalline LSM20 memristive devices
LSM20 memristive devices La0.8Sr0.2MnO3-δ films with a thickness of 8 nm (LSM20-8) were grown
single crystal substrates, namely STO and LAO, and on Si3N4. The epitaxial LSM20-8 films exhibit an
out-of-plane shrinkage (εzz=-0.21%) in LSM20-8/STO, and elongation (εzz=+ 2.62%) in LSM20-
8/LAO, related to the biaxial strain induced by the substrate. On the other side, LSM20-8/Si3N4 is
polycrystalline with a cell parameter similar to that of bulk LSM20 (apc=3.89 ± 0.01 Å[123]). In
agreement with literature,[58], [59] the resistivity measured increases around one order of magnitude with
net biaxial strain, ca.12 mΩ·cm and 100 mΩ·cm for LSM20-8/STO and LSM20-8/LAO. The less
electronic conductive nature of the grain boundaries[54] is revealed on polycrystalline LSM20-8/Si3N4
with a film resistivity of 2 Ω·cm.
88
The IR of the Ti/LSM20-8\Pt memristive devices on the three samples are dominated by the Ti/LSM20
interface, composed of TiOx and reduced LSM20, accounting for more than 80% of the IR, and the R(V)
characteristics suggest that Ti/TiOx acts as an oxygen reservoir. The memristive response of epitaxial
LSM20-8/STO and LSM20-8/LAO show similar CCW R(V) characteristics and HRS/LRS ratio than
their analogous in LSM20-7. The role of polycrystalline texture was assessed on memristive devices on
LSM20-8/Si3N4 films. The R-I(V) characteristics of LSM20-8/Si3N4 devices resemble those of
LSM20/LAO with a HRS/LRS ratio larger than 10. The grain boundaries provide fast oxygen migration
paths in the polycrystalline LSM20, allowing symmetric voltage operation with particularly large HRS,
and consequently larger HRS/LRS ratio (17 at Vamp=±14 V) than in LSM20-7/LAO (4 at Vamp=±14 V).
However, large operating voltage (≥16 V) leads to an increase of the LRS, similar to that found on
LSM20/STO in symmetric voltage operation and thus, the operating voltage range is ±6 V to ±14 V. In
both systems, the increase of the LRS can be related to a loss of electron hole carriers due to the redox
reaction at the Ti/LSM20 interface.
4.2.2.4. Structural characterization of epitaxial and polycrystalline LSM20-21

films
LSM20 films of 21 nm thickness (LSM20-21) were grown on the three substrates used so far, i.e. STO,
LAO and Si3N4, to assess the impact of strain relaxation and microstructure on resistive switching. The
thickness was measured by XRR as shown in Figure 4.21 for LSM20-21/STO, and corroborated by
TEM cross-section images. The thickness and roughness of the film were fitted to 20.51 nm and
1010
Experimental
109 Fit
108
Intensity (counts)
107
106
105
104
103
102
0.0 0.5 1.0 1.5 2.0 2.5 3.0
2 theta (°)
Figure 4.21. XRR scan of a sample of LSM20-21/ STO and the fit result. The thickness and roughness obtained
from the fitting are 20.51 nm, 1.06 nm, respectively. The density was set to 6.521 g.cm-3 (ICDD 00-053-0058).
89
1.06 nm, respectively, while the density was fixed at 6.521 g·cm-3, corresponding to the theoretical value
(ICDD 00-53-0058).
Figure 4.22 presents the SEM surface images and XRD patterns of LSM20-21 films grown on the three
substrates. The SEM images show that the LSM20-21 films are dense. The topography of the films
grown on STO and LAO is quite similar (Figure 4.22a and b), which exhibit square-shaped grains with
average side of 28 ± 6 and 33 ± 9 nm, respectively. These grains are composed of smaller square crystals
(easier to see in LSM20-21/LAO in Figure 4.22b). The small grains were only observed for the epitaxial
films of LSM20-21, thanks to a better image quality (less interaction with the insulating substrate),
although thinner films are likely to present the same morphology and structure.
Similar roughness, with Sq values of 0.56 nm and 0.48 nm for LSM20-21/STO and LSM20-21/LAO,
respectively, were measured by AFM. The epitaxial nature of these samples was verified by XRD and
TEM. A preferential orientation was found in the diffraction pattern, which shows exclusively the peaks
of LSM20 (00l)pc next to the (00l) STO and (00l)pc LAO. The calculated out-of-plane cell parameter in
LSM20-21/STO (cpc=3.888 ± 0.003 Å) indicates slight out-of-plane shrink (εzz=-0.26 %) while LSM20-
21/LAO (cpc=3.993 ± 0.006 Å) is elongated (εzz=+2.44 %), in agreement with the substrate induced in-
plane tensile and compressive strain, respectively. The strain was calculated using Equation 4.3.
Furthermore, the presence of Kiessig fringes in LSM20-21/LAO (peaks around the main LSM20 peak
in Figure 4.22 and Figure 4.28) reveals a very good crystal quality. Figure 4.23 shows the cross-section
of the three LSM20-21 films observed by HRTEM at different magnification. HRTEM cross-section of
LSM20-21/STO and LSM20-21/LAO shows good film thickness homogeneity (Figure 4.23a, c) and
good crystal quality (Figure 4.23b, d). For LSM20-21/STO (Figure 4.23a, b), the atoms in LSM20
(observed along [001]pc zone axis and substrate [001] zone axis) show long range order and very few
defects. On the other hand, LSM20-21/LAO shows mid-range order but larger concentration of defects,
as expected from the lattice mismatch (Figure 4.23c, d).
The larger thickness of these films allow complementary characterization of the structural properties.
Reciprocal space mapping (RSM) was carried out on LSM20-21/LAO (Figure 4.24) to assess the in-
plane orientation of the film. Two diffraction planes were studied: a) rhombohedral, R-3c, symmetrical
(024), and b) asymmetrical (042), which correspond to (001)pc and (131)pc, respectively. In both cases,
the peaks of the substrate and the film are aligned in the same Qx, which indicates epitaxy with the same
in-plane cell parameter. On the other side, the LSM20 peak is located at a smaller Qz (0.4980 for (024))
than the LAO peak (0.5277 for (024)) in good agreement with the out-of-plane expansion observed from
XRD. Whereas the peak assigned to the substrate is intense and defined, the film of LSM20 shows a
less intense contribution with a larger dispersion. This suggests a broader distribution of the in-plane
and out-of-plane lattice parameters, probably caused by the relaxation of the strain in the outermost
layers of the film.
90
The LSM20-21/Si3N4 films shows polycrystalline texture, as observed by GI-XRD (Figure 4.22c). The
rhombohedral cell parameters, a=5.507 ± 0.003 Å and c=13.41 ± 0.02 Å, (apc=3.894 ± 0.002 Å) are in
Figure 4.22. SEM images and XRD patterns for LSM20-21 films grown on a) STO, b) LAO, and c) Si3N4. Surface
morphology and film texture can be observed in the SEM images and XRD, respectively. The diffraction pattern
of c) was obtained in GI-XRD configuration to reduce the signal from the substrate. The signal arising from the
substrate is denoted by the inverse triangle (▼).
91
good agreement with that reported on the literature for La0.8Sr0.2MnO3±δ, (a=5.512 Å and c=13.356 Å;
apc =3.898 Å) [123]. The surface observed by SEM (Figure 4.22c) presents grains of different shapes that
a) b) Pt
Glue Ti
Pt
LSM[001]
Ti
LSM[001]
STO[001] STO[001]
c) d) Glue
Glue
LSM[001]
LSM[001]
LAO[001] LAO[001]
e) f)
Glue
Pt
Pt
LSM
LSM
Si3N4 Si3N4
Figure 4.23. HRTEM cross-section images of a), b) LSM20-21/STO, c), d) LSM20-21/LAO, and e), f) LSM20-
21/Si3N4. The LSM20 films are observed along [001] substrate zones axis for a)-d) (pseudo-cubic notation for
LAO).
92
Figure 4.24. Reciprocal space maps on LSM20-21/LAO around a) symmetrical LSM peak (024) and
b) asymmetrical LSM peak (042). The indexation used correspond to rhomobhedral R-3c, and they are equivalent
to pseudocubic (002) and (131), respectively.
correspond to randomly oriented grains, similar to those observed for polycrystalline 100 nm thick
LSM20 films (Section 3.2.3). The average grain size is 23 ± 7 nm, where the large dispersion arises
from the irregular shape of the grains. As a consequence of the grains extruding out of the films (in light
contrast in Figure 4.22c), larger roughness (Sq=1.54 nm) was measured by AFM. The TEM cross-
section images of the LSM20-21/Si3N4 film, presented in Figure 4.23c, d, show the protruding grains
and roughness, as well as grains with different crystallographic orientation.
The electrical properties of the film were assessed by four-point probe resistivity measurements. The
resistivity values obtained for LSM20-21 films grown on STO, LAO and Si3N4 are 0.010 ± 0.002,
0.11 ± 0.04, and 1.05 ± 0.08 Ω·cm, respectively (see Figure 4.17a). As previously mentioned, the
values of resistivity are affected by the magnitude of strain[59] and congruently, the LSM20-21/STO
film, with a strain of |-0.26%|, is less resistive than the largely strained film on LAO, |+ 2.44%|
(Table 4.2). In strained films, the formation of defects leads to relaxation of strain, and hence, decrease
of resistivity for thicker films, which is more noticeable in largely strained films.[59] The values found
for epitaxial LSM20-21 films are similar to those found in LSM20-8 (Figure 4.17a). So, we can argue
that the LSM20-8 and LSM20-21 films are similarly strained and the thickness for the complete
relaxation of strain is beyond 21 nm. The polycrystalline film, LSM20-21/Si3N4, shows a large
resistivity (1 Ω·cm), comparable to the polycrystalline LSM20-8 (2 Ω·cm), governed by the less
conducting behavior of the grain boundaries.[54], [143] The role of composition, defects and strain is
discussed in Section 1.3.
93
4.2.2.5. Memristive response in LSM20-21 films
The standard microfabrication procedure, using laser lithography, was carried out on the samples
LSM20-21/LAO and LSM20-21/Si3N4. Due to technical issues, the resin exposure for the devices on
LSM20-21/STO was carried out using Smart Print UV system. In the three LSM20-21 samples, the
metal contacts are Pt (100 nm)/Ti (10 nm) and Pt (100 nm).
Initial resistance (IR)
The IR of standard devices and the weight of the interface resistance to IR is presented in Figure 4.18.
The devices on epitaxial films, i.e. LSM20-21/STO and LSM20-21/LAO, exhibit an IR around 30 kΩ,
while the devices on LSM20-21/Si3N4 are around one order of magnitude larger, 200 kΩ. The IR of the
three LSM20-21-based devices do not scale with the distance between the contacts, but they show a
weak trend with the electrode area size (slope between - 0.3 and - 0.5 in log(IR)-log(area)). The
contribution of the interface resistance to IR was obtained using Equation 4.1. While LSM20-21/STO
is clearly dominated by the interface, Rinterface=89% of IR, only 46% of IR can be attributed to the
interface in LSM20-21/LAO. In the case of LSM20-21/Si3N4, the model used for estimation of the
interface resistance cannot be used, as the resistivity of LSM20 (and RLSM20) are comparable to the IR
measured.
Memristive response: symmetric voltage operation
The R(V) characteristics of Ti/LSM20-21\Pt standard devices are shown in Figure 4.25a-c where the
curves correspond to devices biased at increasing voltage amplitude.
LSM20-21/LAO and LSM20-21/Si3N4: The electrical responses of the devices fabricated on LSM20-
21/LAO and LSM20-21/Si3N4 for different sweeping cycles are similar (Figure 4.25b and c,
respectively). Their R(V) curves show small CCW hysteresis at large voltage, but no permanent change
of resistance close to 0 V for writing voltages up to ±20 V. The curves are almost symmetrical in both
polarities for both samples. The devices on LSM20-21/Si3N4 show an increase of resistance at small
voltages while in those on LSM20-21/LAO the increase is smaller, as seen for their analogous in
LSM20-8. The largest HRS/LRS ratio measured, read at +0.1 V, is 1.2 and 1.4, for Ti/LSM20\Pt devices
in LSM20-21/LAO and LSM20-21/Si3N4, respectively. Irreversible and large increase of resistance, first
under positive and then negative bias, has been observed in some devices of LSM20-21/LAO and
LSM20-21/Si3N4 when cycling at large voltages (±20 V) but no memristive response was observed
neither. In these cases, we conclude that the devices have experienced irreversible damage and are no
longer functional.
LSM20-21/STO: Memristive response is found in the devices on LSM20-21/STO (Figure 4.25a).

During the initialization, for Vamp≤6 V, a very narrow hysteresis is observed in the whole range of
94
Figure 4.25. R(V) characteristics of Ti/LSM20-21\Pt devices for films grown on a) STO, b) LAO, and c) Si 3N4
substrates. The devices were biased with a Vamp equal in both polarities and increased with time.
positive bias while a large hysteresis is found for negative bias smaller than -1 V. The HRS/LRS ratio
measured at +0.1 V remains below 1.1. Carrying out cycles at ±8 V induces and abrupt increase of
resistance (transition not shown): first, a slight increase of resistance during HRS reset of about 70 kΩ
at +0.1 V (+45% IR) takes place, followed by a rise of about two orders of magnitude during LRS set,
from 140 kΩ to 17 MΩ. The hysteresis is also observed at positive bias, with a NDR region as observed
in previous LSM20/STO-based devices, and a gradual RESET to the HRS. However, the SET from the
HRS to the LRS is filamentary-like with a sharp decrease of resistance around -6 V. At Vamp=8 V, the
HRS/LRS ratio is around 3.5 and 2.3, where the HRS decreases and the LRS increases with the number
of cycles. When the voltage sweeps are carried out at ±10 V, the sharp SET transition to the LRS shifts
to -7.5 V and the NDR feature vanishes with cycling. The HRS and the LRS are not so stable with cycles
and the HRS/LRS ratio ranges 1.6-4.6. Larger voltage amplitudes (≥12 V) induced irreversible damage
in the device: the curves present a lot of noise and eventually the resistance increases beyond 10 GΩ (at
|V| < 5 V). The damage induced in a LSM20-21/STO-based device can be observed at naked eye, as
shown in Figure 4.26. We note that similar features have been observed after failure in other type of
95
Figure 4.26. Proof of damage in a Ti/LSM20\Pt device. Optical microscopy image of a LSM20-21/STO-based
device a) in pristine state, and b) after applying large voltage (12-20 V). c) Detail of the damaged region circled
in b) observed by secondary electron SEM.
samples and devices. A bright contour around the whole Ti electrode can be observed after applying
large voltage, i.e. 12-20 V (Figure 4.26b). The SEM inspection of that area reveals a different
morphology of the LSM20 film where some of the grains have grown and other ones have coalesced
forming stripes (Figure 4.26c). These new features are slightly orientated pointing away from the Ti
electrode, parallel to the diverging (converging) lines of the electric field on the Ti/LSM20 interface
when Ti is positively (negatively) biased. Yao et al.[20] performed electro-thermal simulations on
La0.67Sr0.33MnO3/Nb:STO memristive devices and found that the electric field lines diverge (converge)
from the positively (negatively) biased tip (in direct contact with the film). Considering that the whole
Ti electrode is at the same positive potential, the electric field lines will diverge away from the Ti contact
and into the LSM20 before they tilt towards the grounded Pt counter-electrode. The features of damage
follow this evolution:
 at intermediate operating voltage they might appear mainly on the Ti side facing the Pt contact
(circle in Figure 4.26b),
 as the voltage increases, the features extend and emerge on the lateral sides of the Ti contact,
and eventually on the back-side.
In order to explain the damage observed we consider the oxygen exchange mechanism reported for
resistive switching in LMO[88] and LSM20,[94] in which oxygen is released as an adsorbed species in the
HRS and reincorporated back in the LRS; and in LMO[87], where the exchange occurred with a TiOxNy
electrode. In the present case, an oxygen pocket could be formed at the Ti/LSM20 interface leading to
loss of contact between Ti and LSM20. The CCW log R(V) characteristics of the LSM20-21-based
devices suggest that the sense of oxygen migration described in the model above (Section 4.1.3) is the
same. At small operating voltages, the emergence of the damage and loss of contact could be related to
the noisy transition from LRS to HRS (LRS branch at positive voltage for Vwrite=+8 V in Figure 4.25a).
Although the damage and oxygen release might start at positive and/or smaller voltages, the large
96
increase of resistance related to the damage occurs under negative bias on LSM20-21/STO devices. The
contact between Ti electrode and LSM20 films is not completely lost in any of the three samples.
Memristive response in LSM20-21/STO: asymmetric voltage operation
The operating window (or HRS/LRS ratio) of the LSM20-21/STO-based devices at small operating
voltages is too narrow for reliable operation (Figure 4.25a). In order to obtain larger window and to
prevent the failure of the device, asymmetrical voltage sweeps have been tested. As mentioned above,
the final failure occurs during SET operation (negative bias) and may involve oxygen from the TiO x
interlayer. However, the oxygen in the interlayer originates from the LSM20 film during HRS RESET
operation. Therefore, here, the positive voltage was limited to prevent the formation of oxygen pocket,
and the LRS was programmed at different voltages. A standard device was first initialized by cycling
three times at ±6 V. The device was further cycled at fixed Vreset = +2 V and different Vset = -6, -8 and
- 20 V. The device was cycled three times at each voltage range, except for the +2/-20 V range, for
which was cycled four times. Figure 4.27a, b summarizes the R(V) characteristics of the last cycle at
each Vreset / Vset, and the resistance states (at +0.1 V) and HRS/LRS ratio of all the cycles.
The R(V) curves (Figure 4.27a) present a gradual RESET and SET process at the applied voltage studied
with no abrupt resistance change, as previously observed (see Figure 4.25a). The curves and the
HRS/LRS ratio of the current device cycled at ±6 V are similar to those shown in Figure 4.25a.
Moreover, the R(V) characteristics of the device cycled at Vamp=6 V are similar to those with Vreset=+2 V
and Vset=-6 to -8 V, except for the smaller hysteresis found at negative voltages in the asymmetric cycles
(Figure 4.27a). The shape of the curves at negative polarity is not altered for the cycles +2 V/-20 V
(Vreset / Vset), while at positive polarity, a NDR region appears, similar to that observed in symmetrical
Figure 4.27. Asymmetric voltage operation of standard LSM20-21/STO-based memristive devices. a) R(V)
characteristics at different Vreset and Vset, and b) the resistance states HRS and LRS (left y-axis), and HRS/LRS
ratio (right y-axis) of every cycle. The cycles were carried out in the same order as they appear in the legend. The
inset in a) shows the whole voltage range.
97
operation in LSM20-7(-8)/STO devices. The resistance states measured in symmetric, Vamp= ±6 V, and
asymmetric operation, Vreset=+2 V and Vset=-6 to -8 V, are very similar with HRS=890±90 kΩ and
LRS=641±60 kΩ (HRS/LRS=1.4±0.1). However, a larger operating window (8±3) is measured when
the device is stressed at +2/-20V. Under these conditions the HRS decreases (970 to 590 kΩ) and the
LRS increases (820 to 980 kΩ) with cycles and seem to stabilize at the fourth cycle.
The most remarkable result in the asymmetric operation is that the HRS seems to be determined by the
largest reset voltage applied (initialization voltage), in this case Vreset=+6 V, and does not vary with
smaller Vreset (=+2 V) nor by changing the Vset (=-6 to -20 V). On the other hand, the LRS depends on
the set voltage, particularly for Vset=-8 to -20 V. Hence, LSM20-21/STO-based memristive devices
exhibit potential for multilevel storage capacity with tunable LRS (at different Vset) while the HRS would
remain constant (at fixed Vreset) and determined by the initialization voltage amplitude.
4.2.2.6. Summary on LSM20-21 films
On the one hand, the strain is retained for 21 nm thick epitaxial La0.8Sr0.2MnO3-δ (LSM20-21) films
grown on STO (εzz=- 0.26%) and LAO (εzz=+ 2.44%) substrates. On the other hand, as expected, no
strain induced by the substrate exists on polycrystalline LSM20-21 grown on Si3N4 which presents a cell
parameter close to bulk LSM20. The LSM20-21 film show long range order of atoms when grown on
STO while the range is shorter for the film grown on LAO, as the larger mismatch between film and
substrate induces nucleation and growth of defects. As expected, the electrical resistivity increases with
net biaxial strain (0.010, 0.11 Ω·cm, for LSM20-21/STO and LSM20-21/LAO, respectively) and
ultimately with the presence of grain boundaries (1.05 Ω·cm for LSM20-8/Si3N4), similar to the results
found in LSM20-8 films.
Memristive response, with different resistance close to zero-field conditions, was only achieved in
Ti/LSM20\Pt devices on LSM20-21/STO. These devices show a large contribution of interface
resistance to the initial resistance value (89%). The difference in resistance after positive and negative
voltage sweeps was practically null for devices on LSM20-21/LAO and LSM20-21/Si3N4. The exact
origin of the different behavior between LSM20-21/STO, and LSM20-21/LAO and LSM20-21/Si3N4
remains unknown although it could be related to the absence (LSM20-21/STO) and presence of extended
defects (dislocations and grain boundaries). The surface roughness, and a difference in oxygen content
in the LSM below the evaporated Ti electrode, compared to thinner films with smaller LSM thickness,
could also be at the source of the lack or small memristance observed in these devices. For LSM20-
21/STO-based devices, distinguishable HRS and LRS, and the characteristic shape of R-I(V) curve with
NDR appear at Vwrite>6 V. However, the HRS to LRS transition seems filamentary. Eventually, these
devices break down at voltages larger than 10 V probably related to the release of molecular oxygen.
Finally, asymmetric voltage operation of LSM20-21/STO devices, with small constant Vreset and large
98
4.3 Cross-comparison: the role of thickness
variable Vset, have proved reproducible performance, multilevel storage capabilities (by tuneable LRS)
and prevent failure.
Cross-comparison: the role of thickness

In this section, we review the results presented grouped by substrates (and micro/nanostructure), as a
function of the thickness. Figure 4.28 presents the SEM top-view images and XRD of the LSM20 film
with thickness of 7, 8, 21 and 88 nm, grown on substrate of STO, LAO and Si3N4. The samples of 88 nm
thick LSM20 (LSM20-88) have been included to highlight the evolution of structure with thickness,
although no devices were microfabricated on them. The properties and characteristics of these samples
have been summarized in the previous sections: roughness and out-of-plane cell parameter in Table 4.2;
film resistivity in Figure 4.17; initial resistance (IR) and weight of the Ti/LSM20 interface in IR in
Figure 4.18. In addition, the operating conditions and resistive switching figures of the optimized
LSM20-based memristive devices are summarized in Table 4.3, and the R(V) curves of the standard
devices seen in this chapter are collected in Figure 4.29 .
Table 4.3. Operating conditions of the optimized LSM20-based memristive switching devices discussed in the
chapter.
Operation Multilevel
Sample Vreset (V) Vset (V) HRS/LRS
mode states
LSM20-21 / STO Asymmetric +2 -2 to -20 LRS 1.4 to 8
LSM20-7 / STO Symmetric +10 -10 None 1.2-3
LSM20-7 / LAO Symmetric +6 to +20 -6 to -20 HRS-LRS 1.5 to 5-23
LSM20-8 / Si3N4 Symmetric +4 to +14 -4 to -14 HRS-LRS 1.5 to 17
LSM20/STO
The XRD pattern of the LSM20/STO films reveal highly oriented films for all thickness studied
(Figure 4.28b). The presence of Kiessig fringes in the pattern of LSM20-21/STO suggests good crystal
quality, i.e. epitaxy, as found by TEM. The evolution of the out-of-plane lattice parameter of
LSM20/STO films suggest the presence of small in-plane tensile strain (Table 4.2), which might be
induced by the substrate, as expected from the misfit (+0.18%, Equation 4.3). The morphology of the
7, 8 and 21 nm thick LSM20/STO films is similar: small square-shaped crystals, which size grows with
99
Figure 4.28. Characterization summary of LSM20 thin films with different thickness and grown on different
substrates. a), c), e) Secondary electron SEM surface imaging and b), d), f) XRD pattern of the LSM films grown
on STO, LAO, and Si3N4 substrates. The SEM image are presented in the same scale for comparison. The XRD
pattern of LSM/Si3N4 (f) was obtained in grazing incidence configuration (GI-XRD) to minimize substrate
interaction.
thickness, and root-mean-square height roughness, Sq, around 0.6 nm. We hypothesized that these grains
are related to the [00l]pc orientation found by XRD. On the other hand, the thick LSM20-88/STO film
presents grains with other crystallographic orientations, assigned to [hh0]pc (detected by XRD, angle
range not shown) which are related to the triangular-grains observed by SEM (Figure 4.28a) (similar
discussion in ca.100 nm thick films is presented in Section 3.2.2). In agreement with the constant out-
of-plane cell parameter, the resistivity of LSM20 lies around 10 mΩ·cm for LSM20-8, LSM20-21 and
100
4.3 Cross-comparison: the role of thickness
a) 108
LSM20 / STO 7 nm
Resistance (Ω)
107
106 8 nm
105
21 nm
104
103
-20 -15 -10 -5 0 5 10 15 20
Voltage (V)
b) 108
LSM20 / LAO
Resistance (Ω)
10 7 7 nm
106
8 nm
105
21 nm
4
10
-20 -15 -10 -5 0 5 10 15 20
Voltage (V)
c)
1010
LSM20 / Si3N4
109
Resistance (Ω)
108 8 nm
107
106
21 nm
105
104
-20 -15 -10 -5 0 5 10 15 20
Voltage (V)
Figure 4.29. Resistance-voltage characteristics of the Ti/LSM20\Pt devices in films with different thickness
grown on a) STO, b) LAO, and c) Si3N4 substrates. The curves correspond to standard memristive devices
operated at the maximum voltage amplitude. For LSM20-21/STO (a), two curves are presented: before (±6 V) and
after (±8 V) the appearance of filamentary-like switching.
LSM20-88, (Figure 4.17) comparable to those reported in literature, 8 mΩ·cm.[93] The sample LSM20-
7/STO, grown before the maintenance of the reactor, shows a resistivity about two orders of magnitude
larger than that of LSM20-8/STO, for which the reason behind this difference remains unknown.
101
Ti/LSM20\Pt devices fabricated on LSM20-7,-8 and -21 films show forming-free resistive switching
where the IR is governed by the less conducting Ti/LSM interface. As shown in Figure 4.29a, the
devices on LSM20-7 and LSM20-8 show gradual SET and RESET which suggests interface type RS,
which can be related to the redox reaction at the Ti/LSM20 interface. However, the thicker LSM20-
21/STO-based devices, presents a mixed interface and filamentary-type RS. More studies are required
to elucidate the origin of the filamentary behavior, which could be related to short-circuit in local regions
in the reduced LSM or the TiOx interlayer.
LSM20/LAO
The LSM20/LAO films show an out-of-plane preferential orientation along [00l]pc, as shown in the XRD
in Figure 4.28d. The calculated out-of-plane lattice parameter (Table 4.2) of the LSM20 films suggests
the existence of a substrate-induced in-plane compressive strain and out-of-plane tensile strain. Similar
to LSM20/STO, the morphology of the 7, 8, 21 nm thick LSM20/LAO films observed by SEM shows
the presence of square-shaped grains, while the thicker LSM20-88/LAO also shows triangular-shaped
grains, which can be related to the polycrystalline component observed in the XRD (angle range not
shown). The TEM cross-section of the LSM20/LAO films used for devices reveals the presence of
extended structural defects, for which no trend with thickness could be derived. Indeed the presence of
Kiessig fringes in the XRD pattern of LSM20-21/LAO indicates a good crystal quality. Similar
resistivity is found for the strained LSM20-8 and LSM20-21 films grown on LAO, ca. 100 mΩ·cm,
while, as expected, the partially relaxed LSM20-88/LAO films exhibit a lower resistivity of 30 mΩ·cm
(Figure 4.17). On the contrary, the LSM20-7/LAO films has a resistivity of 1 Ω·cm, which follows the
same trend as LSM20/STO films.
The memristive devices fabricated on LSM20-7 and LSM20-8 films show forming-free and interface-
type resistive switching (Figure 4.29a), for which the IR is governed by the Ti/LSM interface. After the
initialization, these devices show multi-states in the HRS and the LRS programmable up to 20 V (larger
voltages have not been tested). However, the devices on LSM20-21/LAO films did not show resistive
switching and irreversible damage was induced at the operating voltages used in the other samples. No
explanation has been found for the absence of switching in LSM20-21/LAO devices.
LSM20/Si3N4
The LSM20/Si3N4 films show polycrystalline texture as found by XRD (Figure 4.28f), from which a
constant pseudo-cubic lattice parameter of 3.89 Å was estimated for the different thickness. The surface
of the films present grains of different size and randomly oriented, forming grain boundaries, as
observed by SEM (Figure 4.28e). The thinnest film presented, 8 nm thick (LSM20-8), shows a smooth
surface (Sq=0.67 nm) comparable to the thin epitaxial samples. The 21 nm thick LSM20/Si3N4 films
exhibits triangular-shaped grains and a larger roughness, Sq=1.56 nm, as measured by AFM. As
102
4.4 Conclusions
thickness increases, the majority of the grains adopt the triangular shape and the film becomes rougher,
as seen in LSM20-88/Si3N4 (Figure 4.28e). The resistivity of the polycrystalline films is larger than the
epitaxial samples which can be ascribed to the electronic less conducting nature of the grain
boundaries.[56], [143]
Planar Ti/LSM20\Pt memristive devices were fabricated on the polycrystalline LSM20-8 and LSM20-
21 films (Figure 4.29c). Memristive response was observed only in the thinner film, which shows
gradual SET and RESET, suggesting an interface-type RS. Multi-states in the HRS and the LRS can be
obtained at different writing voltage amplitude. However, no RS was observed in devices on LSM20-
21/Si3N4, similar to those in LSM20-21/LAO films. The reason remains unknown it but could be related
to the presence of extended structural defects.
Conclusions
In this chapter we established a relation between the nanostructure of La0.8Sr0.2MnO3 (LSM20) thin films
and the memristive response of planar Ti/LSM20\Pt devices. The main points are summarized next:
 The nanostructure and oxygen migration in LSM20:

o Three different nanostructures are obtained due to the mismatch between the film and
substrate: epitaxial slight in-plane tensile strained, epitaxial large in-plane compressive
strained and polycrystalline films are obtained when LSM20 is grown on SrTiO3 (STO),
LaAlO3 (LAO) and Si3N4 substrates, respectively.
o The epitaxial LSM20/STO films show long range order and very few defects. Thus, the
oxygen drift is expected to occur predominantly through the bulk.
o The epitaxial LSM20/LAO films shows extended structural defects, i.e. dislocations,
which are expected to increase with thickness as a result of strain relaxation. In this
case, the oxygen drift would occur through the bulk and along the dislocations, for
which it is enhanced.
o The LSM20/Si3N4 films are polycrystalline with a large presence of grain boundaries,
which also are expected to behave as fast oxygen migration pathways.
 General remarks on the memristive response:
o The presence of a TiOx interlayer formed at the Ti/LSM20 interface of pristine devices
has been observed by STEM-EDS. The interlayer may form during Ti evaporation by
scavenging the oxygen from LSM20, creating a reduced LSM20 interlayer at the
interface. The resistance states of the Ti/LSM20\Pt devices are defined by the
103
modification of the conducting properties of the two interlayer by reversible redox

reactions.
o The Ti/LSM20\Pt devices show rectifying behavior in all the LSM nanostructures.
However, memristive response was only observed in very thin LSM20 films, i.e. 7 and
8 nm thick films for all the nanostructures, and on 21 nm thick LSM20/STO. The exact
reason for the absence of memristive behavior in 21 nm thick LSM20/LAO and
LSM20/Si3N4 remains unknown. Yet, it could be related to differences in roughness,
and oxygen content (both pristine and during operation), and differences in current
density and current lines distribution, due to differences in film thickness.
o The R-I(V) characteristic show a gradual transition between high and low resistance
states, i.e. HRS and LRS, respectively, (and vice versa), in agreement with a gradual
change of resistance and modification of the interface. The multilevel storage capacity
and the area-dependent resistance suggest a memristive switching dominated by the
interface.
o Due to the larger oxygen affinity of Ti over Mn, i.e. LSM20, the HRS is more favored
(oxidation of Ti and reduction of LSM20) than the LRS.
o The small relaxation of the resistance observed after removing the applied voltage could
be related to an oxygen back-diffusion process, which is triggered by a chemical
gradient induced during device operation. While the relaxation is similar for the
different devices at different programming voltages, the HRS/LRS ratio before
removing the bias determines if the states are distinguishable over time.
 Nanostructure and memristive response (values for standard devices):
o The LSM20/STO-based devices show (i) reproducible performance at intermediate
symmetric voltage (±10 V) and by pulses, with HRS/LRS ratio around 3,
(ii) reproducible performance in asymmetric bias operation (with multilevel LRS, and
HRS/LRS up to 6-7).
o The LRS increases in defect-poor LSM20/STO films, under large symmetric bipolar
biasing (|Vreset|=|Vset|=12-20 V), which causes non-reproducible performance of the
device at these voltages.
o The LSM20/LAO-based devices can operate at symmetric voltage up to ±20 V with
(i) multilevel storage capacity (for both the HRS and the LRS), (ⅱ) HRS/LRS ratio
between 3 and 23, (ⅲ) good retention and (ⅳ) by pulses.
o The LSM20/Si3N4-based devices can operate at symmetric voltages up to ±14 V with
(i) multilevel storage capacity (for both the HRS and the LRS), (ⅱ) HRS/LRS ratio
larger than 10, and (iii) good retention.
104
4.4 Conclusions
o Enhanced oxygen migration through extended defects, such as dislocations and grain
boundaries in LSM20/LAO and LSM20/Si3N4, respectively, would allow larger oxygen
drift from the TiOx interlayer into the LSM20 and facilitate the LRS SET.
The results presented here represent a fundamental study to shed light to the role of the nanostructure in
LSM20-based memristive devices where we take advantage of the simpler crystal structure of the
epitaxial films to later understand the polycrystalline–based devices. Polycrystalline films are more
likely to be implemented by the micro-technology industry, especially if they are based on silicon
compatible technologies. Therefore, the polycrystalline-film-based devices are the proof-of-concept to
show the potential of nanostructured LSM20-based as memristive devices. Miniaturization of the
devices and change of geometry (to vertical configuration, using conducting substrates such as
platinized silicon) would allow decreasing the operating voltage, ideally below 2 V, in agreement with
industry requirements, while preserving the characteristics found, large operating window and
multilevel resistance states.
105
Memristive response of epitaxial
La0.5Sr0.5MnO3-δ - based devices
Growth of epitaxial La0.5Sr0.5MnO3-δ films on LAO .......................................................... 108
Structural characterization of epitaxial LSM50 ......................................................... 108
Electrical characterization of LSM50/LAO-based devices ............................................... 111
Initial resistance (IR) of the LSM50/LAO devices .................................................... 112
Memristive response of the LSM50/LAO devices .................................................... 113
Reproducibility and stability of the LSM50/LAO devices ........................................ 116
Scaling of LSM50/LAO devices ............................................................................... 119
Role of the strontium content on LSM/LAO-based devices ............................................. 121
Conclusions ........................................................................................................................... 124
This chapter focuses on the application of La0.5Sr0.5MnO3 (LSM50) as a material for memory
applications. The first section focuses on the structural characterization, and the second section tackles
the memristive response of Ti/LSM50\Pt planar memristive devices. Finally, the third section presents
a comparison and discussion of the memristive response obtained in LSM50-based devices and in
La0.8Sr0.2MnO3 (LSM20)-based devices, discussed in Chapter 4.
To the best of my knowledge, this chapter presents for the first time reliable and reproducible resistive
switching in LSM50. The growth of LSM was simultaneously carried out on LaAlO3 (LAO) and SrTiO3
(STO) substrates. However, due to time constraints, the memristive properties of the 50%-substituted
LSM were only studied in depth for the devices built on LSM50/LAO.
107
Chapter 5: Memristive response of epitaxial La0.5Sr0.5MnO3-δ - based devices
Growth of epitaxial La0.5Sr0.5MnO3-δ films on LAO

To contextualize this chapter, a brief description of the properties of LSM50 is given below. More
information can be found in Section 1.3. In stoichiometric LSM50 (δ=0), the A-site is equally occupied
by La and Sr ions and the Mn3+/Mn4+ ratio is equal to 1. The La3+ substitution by Sr2+ has a double effect:
it oxidizes Mn3+ to Mn4+, creating a hole (losing one electron), and it creates oxygen vacancies. [38], [45],
[64]
According to Mizusaki et al.[38] and supported by computational works,[48], [76] the formation of
oxygen vacancies in LSM50 is more favored than in any other Sr substituted La1-xSrxMnO3 (LSM)
composition, in agreement with the differences in tracer bulk diffusion coefficient (10 - 16 cm2·s- 1 and
2·10 - 15 cm2·s-1 for LSM20 and LSM50 at 700 °C, respectively).[57], [65] As reported by Mizusaki et al.,[38]
oxygen-stoichiometric LSM50 (δ=0) lays on the edge between p-type and n-type conductivity, but the
presence of oxygen vacancies will turn it into p-type conductor. Due to the large electron hole
concentration, LSM50 presents a higher conductivity compared to any other Sr-substituted
compositions,[38], [41], [49] for which the homogeneous distribution of electric field within the material is
expected to favor an interface-type memristive response, preventing the dielectric breakdown typical of
filamentary-type RS. As Choi et al.[104] suggested, who studied polycrystalline LSM vertical devices
with x=0.1, 0.3 and 0.5, larger operating windows are expected as the Sr content increases (in the studied
range of x=0.1-0.5). Nonetheless, the authors observed a narrowing of the HRS/LRS window from x=0.3
to x=0.5 which was ascribed to a poor crystallization of the LSM x=0.5 film. Asanuma et al.[110] studied
the effect of cation substitution Pr1-xCaxMnO3 (PCMO) with x=0-1 (similar system to LSM) in the
memristive properties of vertical Ti/PCMO/SrRuO3 devices. The authors observed the presence of
amorphous TiOx interlayer, for which the thickness increases during operation, and reduced PCMO at
the interface, which increases the band gap. The change in the band diagram induced by Ca substitution
(which increases the work function, and has the smallest band gap around x=0.5) and oxygen vacancies
(which increases the band gap) was ascribed to the RS performance observed, for which the largest
HRS/LRS ratio was measured around x=0.4. Therefore, according to these previous works, large
operating window are expected for LSM50.
Structural characterization of epitaxial LSM50
A 21 nm thick LSM50 film was grown on (001)pc-oriented LAO substrate by PI-MOCVD using 5000
pulses of the precursor solution which had been optimized to obtain La0.5Sr0.5MnO3 (solution 5 in
Section 3.2). The thickness was measured by XRR, as shown in Figure 5.1, where the density was fixed
to the theoretical value of 6.294 g·cm-3 (ICDD-PDF 01-084-6879). Similar values of thickness
(21.3 nm) were obtained when the density was also fitted, even though the values were slightly larger
(6.77 g·cm-3).
108
5.1 Growth of epitaxial La0.5Sr0.5MnO3-δ films on LAO
1010
Experimental
109 Fit
108
Intensity (counts)
107
106
105
104
103
102
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
2 theta (°)
Figure 5.1. XRR pattern of the LSM50/LAO films. The thickness and roughness obtained by fitting are indicated
in the schematic.
The XRD pattern obtained in Bragg-Brentano configuration (Figure 5.2a) for the LSM50/LAO film
showed the preferential growth along the tetragonal [hh0] and [00l] directions, indicating the presence
of two domains. For the (00l)-oriented domain, the c axis of the tetragonal LSM lattice is oriented out-
of-plane, whereas the a and b axis are in-plane and rotated 45° respect the apc and bpc axis of LAO (see
Figure 5.2b). A similar stacking is found for the (hh0)-oriented domain after the appropriate translation
and rotation of the LSM50 lattice. Using the cell parameters of LSM50 (a=b=5.451 Å, and c=7.675 Å,
from ICDD-PDF 01-084-6879), the theoretical in-plane distance between two consecutive Mn in the
(00l)-oriented domain is d[001]LSM50=a/√2=3.854 Å, and in the domain (hh0) it is d[110]LSM50=c/2=3.838 Å.
Therefore, the (00l)- and (hh0)-oriented domains in LSM50/LAO present different compressive in-plane
strain, ε[001]=- 1.67 % and ε[110]=- 1.24 %, respectively. Considering the pseudocubic cell parameters of
LSM50 (apc=3.854 Å and cpc=3.838 Å) and LAO (apc=3.794 Å and cpc=3.785 Å), the average cell
parameter of the two domains measured by XRD (cpc=3.93±0.01 Å) indicates an elongation of the cell
along the out-of-plane direction (εzz=+2 %) due to the in-plane strain induced by the substrate.
The surface morphology of the LSM50/LAO films was characterized by SEM and AFM, and show a
dense layer with no pinholes, as presented in Figure 5.3a-b. Similar to epitaxial LSM20 films, the
LSM50/LAO film is composed of “grains”, possible different crystal domains, of 27±3 nm with no
particular arrangement over the surface. The roughness of the LSM50 film is Sq=0.5 nm, in the same
order than XRR results, and similar to the roughness measured in 21 nm thick epitaxial LSM20.
The crystalline quality of the film was evaluated by TEM observations of a cross-section. The film is
homogeneous and with long range atomic order, as observed in the TEM cross-section image of
109
a) b)
(001)pc, LAO
(002)pc, LAO
(003)pc, LAO
(110)LSM
(002)LSM
(220)LSM
(004)LSM
Intensity (log scale) (cps)
(006)LSM
(330)LSM
La0.5Sr0.5MnO3 (I4/mcm) - (01-084-6879) LaAlO3 (R3c) - (04-007-2676)
20 30 40 50 60 70 80
2q (deg.)
Figure 5.2. a) XRD patterns of the LSM50/LAO film. The LSM50 and LAO diffraction peaks are indexed in
tetragonal and pseudo-cubic (pc) notation, respectively. The bottom panel displays the database reference for
LSM50 and LAO. b) Schematic of the epitaxial growth of LSM50 on LAO viewed along the out-of-plane, c, axis
(green: La and Sr, blue: Al; purple: Mn; red: O). The tetragonal lattice of LSM50 is rotated 45° around the c axis
respect to the pseudo-cubic lattice of LAO.
LSM50[001]/LAO[001] (Figure 5.4a). The overlapping diffraction spots of LSM50 and LAO in the
selective area electron diffraction (SAED) carried out at the LSM50/LAO interface (Figure 5.4b)
Figure 5.3. Surface morphology of the LSM50/LAO films observed by a) SEM and b) AFM. The average grain
size is 27±3 nm and the roughness Sq=0.5 nm.
110
5.2 Electrical characterization of LSM50/LAO-based devices
a) b) c) FT
SAED FT-1
mask
LSM50 [001]
LAO [001]
Figure 5.4. a) TEM cross-section image of 21 nm thick LSM50/LAO film. b) Selective area electron diffraction
at the LSM50/LAO interface. c) Fourier transform of a) and inverse Fourier transformed image obtained by
applying the mask on (010)LSM, LAO. Dislocations are marked as “⊤”.
confirms the perovskite on perovskite (cube-on-cube) growth of the LSM50 film on the LAO substrate.
The inset in Figure 5.4c presents the Fourier transform of the image in Figure 5.4a. The inversed
Fourier transform image (FT-1), after applying a mask on the (010)LSM, LAO of the FT is shown in
Figure 5.4c, where only the diffraction planes perpendicular to the interface are visible. These planes
extend from the substrate interface to the film surface, demonstrating the very good crystal quality of
the film. Nonetheless, as presented in the zoomed area in Figure 5.4c, some defects are observed: (i) the
planes deflect due to the strain induced by the substrate, and (ⅱ) extra planes appear, i.e. edge
dislocations marked as “⊤”, at thickness larger than ∼10 nm. This type of defects were also observed in
the LSM20/LAO films presented in Chapter 4. Due to the smaller lattice mismatch between LSM50 and
LAO (-1.24 to -1.67 %) compared to LSM20/LAO (-2.77 %), the 21 nm thick LSM50/LAO film
presents a qualitatively lower defect density than in the 7 nm thick LSM20/LAO. Therefore, the
morphology and microstructure of LSM50/LAO film is suitable for device micro-fabrication and to
assess the memristive properties of epitaxial LSM50 thin films for the first time.
Electrical characterization of LSM50/LAO-based devices

Memristive planar devices were micro-fabricated using square contacts composed by a 10 nm thick Ti
layer capped with 90 nm thick Pt as an active electrode, and 100 nm thick Pt layer as inert electrode.
The electrodes were arranged according to the design presented in the Section 2.2, and similar to that
used for the LSM20 films, consisting of two main types of devices: (i) devices with contact area
A=2002 µm2 for which the electrodes are spaced at different distances L=50, 100, 200 µm, and
(ⅱ) devices with variable A=1002, 2002, 3002, and 4002 µm2 separated by constant L=50 µm. The
resistivity of LSM50 and the contact resistivity of Pt/LSM50, summarized in Table 5.1, were obtained
111
Table 5.1. Parameters obtained from the Transmission line measurements (TLM) on the LSM50/LAO film using
Pt contacts.
Parameter Value
Contact resistivity (Ω·cm2) 3.0·10-2
Resistivity (Ω·cm) 1.6·10-3
by transmission line measurements (TLM) using Pt contacts. The electrical characterization of the
Ti/LSM50\Pt devices was carried out using tungsten probes, where the Pt electrode was grounded and
the Ti electrode was biased.
Whereas the bulk resistivity of LSM50 lays between 7.9·10-4 and 8.5·10-3 Ω·cm,[49], [145], [146] the value
in thin films widely depends on the processing technique and microstructure (0.1·10-3 -
17·10- 3 Ω·cm)[145], [147], for which the oxygen vacancy-rich grain boundaries increase the resistivity. As
introduced in Section 1.3, the electrical properties of La1-xSrxMnO3 are dominated by the mean valence
of the Mn, which can be tuned either by the cation and/or the oxygen content.[38] Therefore, the resistivity
value found in epitaxial LSM50/LAO lies in the upper half of the values reported in literature, which
could be related to the presence of oxygen vacancies.[38]
Initial resistance (IR) of the LSM50/LAO devices
The initial resistance (IR) of the devices was measured at +0.1 V on the two types of planar Ti/LSM50\Pt
memristive devices mentioned above: (i) as function of distance, L, and (ⅱ) as function of contact area,
A. As shown in Figure 5.5a, no clear trend of the IR with the distance between contacts is observed
(value around ∼1.5·103 Ω) which suggests a small contribution of the LSM50 film sheet resistance to
the total resistance measured. On the other hand, Figure 5.5b shows that the resistance is inversely
proportional to the contact area (3.0·103 Ω and 0.8·103 Ω for A=1002 and 4002 µm2, respectively) with
a slope larger than -1 (around -0.5). As a first approach, the IR can be expressed as the sum of series
resistances of the contacts, LSM50 and the two interfaces, as explained in Section 4.1. The resistance of
the metallic 100 nm thin contacts and the Pt/LSM contact resistance (with a maximum calculated value
of 300 Ω for devices with A=1002 µm2, using contact resistivity in Table 5.1) is smaller than the IR
values measured. The calculated resistance of the LSM50 film in the device is 9.6·102 Ω (which
corresponds to 67 % of the total IR measured), while for the Ti/LSM50 interface it is 4.6·102 Ω (33 %
112
a) b) Contact side (μm)

100 200 300 400
3 4
1.8x10 10
1.6x103
Resistance (Ω)
Resistance (Ω)
1.4x103 103
1.2x103
1.0x103 102
50 100 200 104 105
Distance between contacts, L (μm) Area (μm2)
Figure 5.5. Initial resistance (IR) values of the Ti/LSM50\Pt pristine memristive devices as a function a) of the
distance between metallic contacts for an area of 2002 µm2, and b) of the area of the contact for the same
distance between electrodes of 50µm. The dashed lines in b) correspond to slope -1.
of the IR). Furthermore, symmetrical Pt/LSM50\Pt planar devices show ohmic (linear) I(V)
characteristics with resistance values around 200 Ω, while Ti/LSM50\Ti devices exhibit rectifying
behavior with resistance values in the range of 800-4000 Ω (up to ±5 V), as shown in Figure 5.6. A
similar I(V) response and IR dependence was observed for the LSM20-based devices, which showed a
high resistive Ti/LSM interface composed by an oxygen deficient TiOx interlayer (observed by
HRTEM) and reduced LSM film (measured by XANES) spontaneously formed during device
fabrication. Therefore, given the resemblances between both systems, the Ti/LSM50\Pt devices might
present an oxygen deficient TiOx interlayer and concomitant reduced LSM50 film. We note that the IR
and its dispersion in LSM50-based devices is smaller compared to the LSM20-based devices. This might
be related to a thinner TiOx layer and/or less reduced LSM50, which has not been experimentally
validated yet, although the interlayer is expected to grow and homogenize after initialization. The
deviation of the IR from the negative unitary slope in Figure 5.5b may be also due to inhomogeneous
thickness and oxidation/reduction level of the two interlayers.
Memristive response of the LSM50/LAO devices
The Ti/LSM50/LAO\Pt devices exhibit memristive response without an electroforming step required.
Resistive switching phenomena was observed in 94% (17/18) of the cases for devices with A=2002 µm2
and distance L= 50 µm, while the lowest success corresponds to those with A=3002 µm2 with 67% (2/3).
Therefore, in the following we focus on the devices with A=2002 µm2 and L=50 µm (hereafter referred
to as standard devices, as did for LSM20). An initialization procedure consisting in five symmetrical
voltage sweeps, 0 +Vamp0 -Vamp 0, where Vamp was systematically changed (similar to that
applied for LSM20-based devices), was carried out on LSM50-based standard devices. Figure 5.7
113
Electric field (V/m)

-1x105 -5x104 0 5x104 1x105
6x105
Pt / LSM50 \ Pt
2.0x10-2 Ti / LSM50 \ Ti
4x105
Current density (A/m2)

1.0x10-2
2x105
Current (A)
0.0 0
-2x105
-1.0x10-2
-4x105
-2
-2.0x10
-6x105
-5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage (V)
Figure 5.6. I(V) curves of symmetrical M/LSM50\M devices with A=2002 µm2 and L=50 µm for M = Pt (dashed
line) and Ti (solid line).
presents I(V) and R(V) characteristics of the initialization procedure for a standard device. First, Vamp
was increased from 2 to 12 V (Figure 5.7a, c) and then decreased to 2 V (Figure 5.7b, d) in steps of
2 V (cycles were also conducted at Vamp=11 V at increasing writing voltage amplitude, see
Figure 5.7a, c). The device presents clockwise (CW) log I(V) characteristics (counter-clockwise, CCW,
log R(V)) in all the voltage range studied. The RESET to the high resistance state (HRS) occurs in a
gradual way in positive polarity. However, the SET to low resistance state (LRS) takes place in a very
small voltage range at negative polarity presenting multiple jumps in current (and resistance). It should
be noticed that abrupt jumps in resistance are usually associated to filamentary connections. As the
writing voltage amplitude increases (Figure 5.7a, c), the device switches to a larger HRS, while the LRS
value remains practically constant and the transition occurs at larger |Vset|. During the RESET operation
at Vamp=10 V, the I(V) curve shows a small negative differential resistance (NDR) around voltages of
+2 V (Figure 5.7a) and it becomes flatter at larger writing voltages. When the Vamp is decreased
(Figure 5.7b, d), the NDR is not observed anymore. Similarly to the first half of the initialization, the
|Vset| increases with voltage amplitude. The HRS value remains constant while the LRS increases, which
suggests that the HRS is the stable state. It is also important to notice that a few standard devices
experienced an irreversible increase of resistance at a voltage larger than 12 V and therefore, the
operating voltage has been limited at 12 V.
The resistance of the device was measured by a voltage sweep up to +0.1 V after each half cycle (when
crossing 0 V). Figure 5.8 presents the HRS and the LRS of the initialization procedure (open symbols)
described in Figure 5.7. From a writing voltage amplitude of Vamp=8 V, the HRS (10.4±2 kΩ) and the
LRS (6.9±4 kΩ) can be distinguished (HRS/LRS=1.5±0.1). At Vamp=10 V, the operating window largely
114

-2x105 -1x105 0 1x105 2x105 -2x105 -1x105 0 1x105 2x105
10-1 10-1
Vamp= 2 → 12 V 106 Vamp= 12 → 2 V 106

-2 -2
10 10
SET 105 SET 105
|Current| (A)
|Current| (A)
10-3 10-3
104 104
10-4 RESET 10-4 RESET

103 103
-5 -5
10 10
102 102
-6 -6
10 10
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
c) Electric field (V/m) d) Electric field (V/m)

-2x105 -1x105 0 1x105 2x105 -2x105 -1x105 0 1x105 2x105
Vamp= 2 → 12 V Vamp= 12 → 2 V
105 105
Resistance (Ω)
Resistance (Ω)
104 104
RESET RESET
103 103 SET
SET
102 102
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Figure 5.7. a), b) I(V) and c), d) R(V) characteristics of the initialization procedure of a Ti/LSM50\Pt memristive
device. The device was cycled five times at each Vamp. The fifth cycle is shown when the voltage amplitude is a),
c) increased from 2V to 12V, and b), d) decreased from 12V to 2V.
increases due to the increase of the HRS value (∼100 kΩ). The window further increases at Vamp=12 V,
with HRS/LRS=36±5, where the HRS and the LRS reach 200±30 kΩ and 5.5±0.2 kΩ, respectively. In
the second half of the initialization experiment (Vamp=122 V), while the HRS does not show any
dependence with writing voltage and it remains rather constant (143±14 kΩ), the LRS shows multilevel
properties with values of 5.5±0.3 kΩ, 13±1 kΩ, 58±5 kΩ, and 101±4 kΩ at Vamp=12, 10, 8 and 6 V,
respectively (HRS/LRS=36±5, 10±1, 2.3±0.1, and 1.48±0.03, respectively). A small increase of the
resistance states with cycling at same voltage amplitude is observed. This increase seems to reach
stabilization after 5 cycles. In order to assess reproducibility, the initialization biasing protocol is carried
out a second time, for which the device is also cycled at Vamp=11 V at increasing and decreasing voltage
amplitude (solid symbols in Figure 5.8). The HRS and the LRS obtained during the second biasing
experiment are comparable to those programmed during the second half of the initialization, and thus
we conclude that the device exhibits reproducible memristive performance.
115
|Vwrite| (V)
2 4 6 8 10 11 12 12 11 10 8 6 4 2
106
HRS
Resistance (Ω)
105
104
Open: First run (initialization)
LRS Solid: Second run
103
0 10 20 30 40 50 60 70
Number of cycle
Figure 5.8. Evolution of the resistance states, HRS and LRS, reached after each cycle (bottom axis) and as function
of the writing voltage amplitude (top axis) in a standard Ti/LSM50/LAO\Pt memristive device. The device is
initialized during the first run at which voltage amplitude is increased from 2 V up to 12 V and decreased back
(open data points). The voltage experiment is carried out a second time (solid data points) to assess reproducibility.
Reproducibility and stability of the LSM50/LAO devices
Another standard device was initialized following the protocol described above (Vamp=2 12 2 V)
and further cycled at 10 V and 12 V to assess the cycle-to-cycle and device-to-device reproducibility.
Figure 5.9 presents the HRS and the LRS measured for the device as function of the number of cycles
(bottom axis) and the writing voltage amplitude (top axis). The patterned area correspond to the
initialization procedure, where the cycles carried out at 12 V (solid symbols) and 10 V (open symbols)
are emphasized. The shaded areas numbered with roman numerals (i-ⅳ) correspond to the retention
measurements that are explained below. For clarity, the average resistance values and HRS/LRS ratios
measured in the experiment shown in Figure 5.9 are summarized in Table 5.2. The evolution of the
HRS and LRS during the initialization (cycles 1 to 70 in Figure 5.9) is similar to that described for the
previous device. The HRS first increases the resistance with Vamp (cycles 1 to 40) and then, it remains
stable as the writing voltage amplitude is decreased (cycles 41 to 70). On the other hand, the LRS show
the same multilevel properties previously observed for the other device. The HRS and the LRS obtained
in the second half of the initialization at Vamp=12 V (10 V) are 56±2 kΩ (64±4 kΩ) and 5.9±0.3 kΩ
(19±2 kΩ), respectively, with a HRS/LRS ratio of 9.5±0.5 (3.43±0.06). Then, the device is cycled 10
times at ±10 V (cycles 71 to 80), prior to the LRS retention test of 29 h ((i) in Figure 5.9). The values
during the retention measurement are not included in this graph but in Figure 5.10 (discussion below).
116
|Vwrite| (V)
2 12 10 2 10 10 12 10 10 12
Initialization HRS
105
(iv) HRS retention

(iii) LRS retention
(ii) HRS retention
(i) LRS retention
Resistance (Ω)
104
LRS
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280
Number of cycle
Figure 5.9. Evolution of the resistance states, HRS and LRS, during the initialization (cycle 1-70) and during
multiple cycling (cycles 71 to 270) at Vamp 10 V (open data points) and 12 V (solid data points). The stability of
each resistance state, LRS (i, ⅲ) and HRS (ⅱ, ⅳ), programmed at Vamp 10 V are indicated by the shaded areas and
is presented in Figure 5.10.
Table 5.2. Resistance value of the resistance states and HRS/LRS ratio obtained at writing voltage amplitude 10 V
and 12 V from Figure 5.9. The values correspond to the average and standard deviation of the cycles.
Before (i), Before (ⅱ), Before (ⅲ), Before (ⅳ),

Vamp Parameter Initialization
cycles 71-80 cycles 91-100 cycles 111-185 cycles 196-270
HRS (kΩ) 64±4 72±1 76±2 81±2 70±2
10 V LRS (kΩ) 19±2 21±1 23±1 26±1 27.3±0.3
HRS/LRS 3.4±0.1 3.4±0.1 3.3±0.1 3.1±0.1 2.6±0.1
HRS (kΩ) 56±2 n/a n/a 66±3 68±1
12 V LRS (kΩ) 5.9±0.3 n/a n/a 9.4±0.4 9.9±0.3
HRS/LRS 9.5±0.5 n/a n/a 7.1±0.3 6.8±0.3
After the LRS retention, the device is cycled 10 more times (cycles 91 to 100) prior to the HRS retention
test of 29 h ((ⅱ) in Figure 5.9).
The resistance states programmed at Vamp=±10 V after the initialization (cycles 71 to 100 in Figure 5.10)
are slightly larger than those measured during initialization (cycles 51 to 60 in Figure 5.10). The
resistance states written at 10 V before the retention test (i) [(ⅱ)] increase with cycles in a similar fashion,
117
being the average HRS=72±1 kΩ [76±1 kΩ] and the LRS=21±1 kΩ [23±1 kΩ], which results in a
practically constant HRS/LRS ratio of 3.4±0.1 [3.3±0.1].
In a second step, the device is cycled 25 times at Vamp=±12 V, followed by 50 cycles at Vamp=±10 V
(cycles 111 to 185 in Figure 5.9) before assessing again the retention of the LRS ((ⅲ) in Figure 5.9).
The device presents a larger HRS in the first cycle at 12 V (cycle 111) but then, after being programmed
to LRS, the HRS in the second cycle is smaller (66±3 kΩ) and similar to those of the initialization. On
the other hand, the LRS (9.4±0.4 kΩ) is larger than in the initialization and consequently, the operating
window is reduced, i.e. HRS/LRS=7.1±0.3. Both resistance states exhibit an increase of resistance with
cycling at Vamp=12 V. When the device is operated at Vamp=10 V, the HRS and the LRS experience an
increase of resistance for the first 10 and 5 cycles, respectively, to reach a stable value of 81±2 kΩ and
26±1 kΩ, respectively, and, as seen before, the HRS/LRS ratio is systematically diminished to 3.1±0.1.
Then, after the LRS retention test (ⅲ), the device is cycled 25 times at Vamp=12 V and 50 cycles at
Vamp=10 V (cycles 196 to 270 in Figure 5.9) to assess again the retention of the HRS. The resistance
states programmed in this voltage sequence are similar to the ones just described (cycles 111 to 185).
The HRS and LRS for Vamp=12 V (10 V) are more stable with cycles, 68±1 kΩ (70±2 kΩ) and
9.9±0.3 kΩ (27±1 kΩ), respectively, although the HRS/LRS ratio decreases to 6.8±0.3 (2.5±0.1).
In conclusion, after 170 writing-erasing cycles at different voltages, the standard device shows a well-
defined HRS, independent of writing voltage, while the LRS can be tuned between at least two different
values which are programmed at Vamp=10 and 12 V. Nonetheless, a small decrease of the HRS and the
LRS are observed upon cycling at different voltages, which decreases the operating window from
9.5±0.5 (3.43±0.06) to 6.8±0.3 (2.6±0.1) at Vamp=12 V (10 V).
As mentioned above, the retention measurements, or evolution of the resistance over time, was measured
for devices programmed at Vamp=10 V with different history, i.e. just cycles at ±10 V or cycles at ±12 V
followed by cycles at ±10 V. In a first step, the device is cycled 10 times at Vamp=10 V before assessing
the retention of each resistance state, the LRS and the HRS, which correspond to (i) and (ⅱ) in
Figure 5.9. In a second step, the device was cycled 25 times at Vamp=12 V, followed by 50 cycles at
Vamp=10 V, and then the evolution of the resistance, for LRS and HRS was measured, which corresponds
to (ⅲ) and (ⅳ) in Figure 5.9, respectively. We note a complete cycle, 0 +Vamp0 -Vamp0, leaves
the device in the LRS and hence, an extra sweep at +10 V is required to program the HRS before the
retention measurements. Figure 5.10 presents the evolution of the LRS (i, ⅲ) and the HRS (ⅱ, ⅳ). The
resistance over time was measured by intermittent voltage sweeps at +0.1 V. In all the cases, the
resistance states (HRS and LRS) experience relaxation towards larger values. The relaxation curves
were fitted to a double exponential decay equation to evaluate the kinetics. The two LRS relaxation
curves (i, ⅲ) present similar characteristic times, τ1=9-10 min and τ2=4-5 h, while the evolution of the
118
9x104
(ii)
8x104 HRS
(iv) LRS HRS
Resistance (Ω)
τ1 9-10 min 3-6 min

7x104 τ2 4-5 h 3-4 h
5x104 (iii)
4x104
3x104 LRS
(i)
2x104
0 6 12 18 24 30
Time (h)
Figure 5.10. Evolution of HRS and LRS values (in linear scale) with time for Ti/LSM50/LAO\Pt devices
programmed at Vamp=10 V. The devices was biased for 10 cycles at Vamp=10 V for curves (i) and (ii) whereas 25
cycles at 10 V followed by 50 cycles at 12 V were carried out before acquisition of the curves (iii) and (iv), as
depicted in Figure 5.9.
two HRS relaxation curves (ii, iv) is somewhat faster, τ1=3-6 min and τ2=3-4 h, even though, comparable
to the characteristic times of the LRS relaxation curves.
Scaling of LSM50/LAO devices
Devices with non-standard geometry, i.e. L=50 µm and A=1002, 3002 and 4002 µm2, were also measured
and they all showed forming-free and CCW log R(V) characteristics. The devices with A=1002 µm2
exhibit R-I(V) curves and writing voltage dependence of the resistance states similar to the one described
in the standard devices (A=2002 µm2). Nonetheless, as the contact area increases, the steps of the abrupt
change in resistance during the LRS SET decrease in size, and the change becomes more gradual.
Figure 5.11 presents the typical R(V) curves of memristive devices with contact area a) A=3002 µm2
and b) 4002 µm2. Furthermore, contrary to the 2002 µm2 devices in which only one HRS can be obtained,
in devices with A=3002 and 4002 µm2, both resistance states, HRS and LRS, are tunable with writing
voltage and hence, show multilevel capacities. The complete initialization procedure, with 20 cycles per
Vamp, is shown in Figure 5.11c and d for devices with A=3002 and 4002 µm2, respectively.
Figure 5.12 presents the area-scaling dependence of the HRS and the LRS programmed at Vamp=10 and
12 V, after initialization. The value given corresponds to the average of 4, 10, 3 and 4 devices with
A=1002, 2002, 3002 and 4002 µm2, respectively. For the larger devices with A=3002 and 4002 µm2, the
119

a) -2x105 -1x105 0 1x105 2x105 b) -2x105 -1x105 0 1x105 2x105
105 105
A = 3002 µm2 A = 4002 µm2
Vamp= 12 → 2 V Vamp= 12 → 2 V
Resistance (Ω)
Resistance (Ω)
104 104
103 103
SET SET
RESET
102 102
RESET
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Voltage applied (V) Voltage applied (V)
|Vwrite| (V) |Vwrite| (V)

c) 2 4 6 8 10 11 12 12 11 10 8 6 4 2
d) 2 4 6 8 10 11 12 12 11 10 8 6 4 2
105 105
A = 3002 µm2 A = 4002 µm2

HRS HRS
Rread (Ω)
Rread (Ω)
104 104
LRS
LRS
103 103
0 40 80 120 160 200 240 280 0 40 80 120 160 200 240 280
Cycle number (accumulate) Cycle number (accumulate)
Figure 5.11. Initialization cycles of Ti/LSM50\Pt memristive devices with different electrode area, A. R(V) curves
of the second half of the initialization procedure at voltage amplitude 12 V, 10 V and 8 V for memristive devices
with a) A=300 µm2 and b) A=400 µm2. Read resistance of the HRS and the LRS during the complete initialization
procedure for devices with c) A=300 µm2 and d) A=400 µm2.
slope of the resistance states programmed at 10 and 12 V is close to the expected value of -1 for
interfacial-type resistive switching, while the HRS and LRS values are quite similar between the 1002
and 2002 µm2, not showing an electrode-size dependence in the small size range.[18] Furthermore, the
1002 and 2002 µm2 show a similar behavior, where the HRS is constant even at different reset voltages,
Vreset, and does not depend on Vamp, but the LRS present multi-states at different set voltage, Vset. Hence,
the smaller devices (A=1002 and 2002 μm2) suggest a mixed filamentary-interfacial-type resistive
switching (mainly filamentary), while the devices larger than 2002 μm2 would present an interfacial-
type switching (they might also present filamentary switching).
120
5.3 Role of the strontium content on LSM/LAO-based devices
Contact side (μm)

100 200 300 400
106
HRS 10V HRS 12V
LRS 10V LRS 12V
Read resistance (Ω) 105
104 slope -1
103
104 105
Area (μm2)
Figure 5.12. HRS and LRS area scaling of Ti/LSM50\Pt memristive devices. The resistance states were
programmed at Vamp=10 V and 12 V. The data was collected on 4, 10, 3 and 4 devices with A=1002, 3002 and
4002 µm2, respectively.
Role of the strontium content on LSM/LAO-based devices

In this section we present a comparison of the memristive performance observed for the LSM with x=0.2
and 0.5 grown on LAO, to explain the role of Sr2+ content. A detailed explanation of the effect of Sr2+
substitution is presented in Section 1.3. Table 5.3 summarizes the main structural and electrical
properties for each composition as well as the operating conditions and performance of the LSM-based
memristive devices. For standard Ti/LSM\Pt devices with x=0.2 and 0.5, Figure 5.13 shows the a),
b) R(V) curves at the largest operating voltage, and c), d) the HRS and LRS writing voltage dependence
for the initialized devices, respectively.
For Ti/LSM\Pt devices with x=0.2 and x=0.5, the absence of (or small) initial resistance (IR)-distance
dependence suggests in both cases the presence of barrier at the Ti/LSM interface in the pristine state.
In the case of LSM50, the TiOx interlayer in the pristine state would be thinner and/or less oxidized
(small Rinterface=4.6·102 Ω) compared to LSM20 (larger Rinterface=107 Ω). The existence of a TiOx has been
proven for LSM20 and is currently under analysis by STEM-EDS for LSM50. Nevertheless, the
thickness and oxidation of TiOx interlayer is expected to increase during initialization. HRTEM analysis
combined with EDS corroborated the existence of the TiOx layer in a pristine Ti/LSM20/LAO stack as
well as the presence of defects for the two composition layers, with a larger number in the LSM20 films.
121
Table 5.3. Main structural properties in LSM20 and LSM50, and operating conditions and performance of the
Ti/LSM\Pt memristive devices for x=0.2 and 0.5, studied in the thesis. The resistance of the Ti/LSM interface is
estimated from Equation 4.1. For LSM50 devices, the operation was limited to ±12 V, however, larger voltage
and HRS/LRS are expected for the devices with area A=3002 and 4002 μm2.
Parameter LSM20 / LAO LSM50 / LAO

apc, literaure 3.898 Å 3.854 Å
domain (00l):-1.66 %
Misfit -2.77 %
(hh0):-1.24%
Defect density ↑↑ ↑
Thickness 7 nm 21 nm
Resistivity 5.2·10 Ω·cm
-1
3.0·10-3 Ω·cm
Average IRa) 1.9·107 Ω 1.4·103 Ω
RTi/LSM (% IR) a) 1.8·107 Ω (95 % of IR) 4.6·102 Ω (33 % of IR)
IR-area dependence YES YES
IR-distance dependence NO NO
Device area 100 to 400 μm
2 2 2
100 - 200 μm
2 2 2
3002 - 4002 μm2
Switching proposed Interfacial Interfacial-filamentary Interfacial
Vreset / +6 to +20 / +8 to +12 / +10 to +12 /
Vset -6 to -20 -8 to -12 -10 to -12
Multi-states Both LRS Both
HRS/LRS ratio 1.5 to 5-23a) 2 to 20-36 3 to 12
a)
Values for devices with A=2002 μm2
The Ti/LSM\Pt devices show CCW log R(V) characteristics in both systems with a gradual HRS RESET
and LRS SET at positive and negative voltage, respectively (Figure 5.13a, b). The I(V) characteristics
and the resistance area dependence of the devices suggest an interface-type resistive switching. For both
compositions, the resistive switching observed in all the devices can be described in terms of oxygen
drift where the Ti electrode acts as an oxygen reservoir by oxidation into TiOx and concomitant reduction
of the LSM. In both cases, the presence of extended defects is considered of great importance in the
performance in the LSM/LAO-based devices, particularly during the less thermodynamically favorable
LRS SET redox reactions. In the case of LSM50, the study of “defect-free” LSM50 devices, i.e.
LSM50/STO, would allow to validate the influence of defects in this composition. According to the
phenomenological model proposed, the resistance measured in the devices is determined by the degree
of oxidation and thickness of the two interlayers which can be further tuned by the applied voltage.
122
5.3 Role of the strontium content on LSM/LAO-based devices

-4x105 -2x105 0 2x105 4x105 -2x105 -1x105 0 1x105 2x105
LSM20 / LAO LSM50 / LAO

105
Resistance (Ω)
Resistance (Ω)
107
104
106 103
102
-20 -15 -10 -5 0 5 10 15 20 -12 -8 -4 0 4 8 12
c) |Vwrite| (V) d) |Vwrite| (V)
10
11
12
12
11
10
10
12
14
16
18
20
20
18
16
14
12
10
2
4
6
8
8
6
4
2
2
4
6
8
8
6
4
2
8
10
LSM20 / LAO LSM50 / LAO
HRS
105 HRS
Rread (Ω)
Rread (Ω)
107
LRS LRS
104
106
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70
Cycle number Cycle number
Figure 5.13. Comparison of the memristive response in standard (A=2002 µm2) Ti/LSM\Pt devices built on a), c)
LSM20 (7 nm)/LAO and b), d) LSM50 (21 nm)/LAO. a), b) R(V) cycle at the largest operating voltage, i.e. Vamp=
20 V and 12 V for LSM20 and LSM50, respectively. c), d) Resistance states-writing voltage dependence of an
initialized device.
Indeed, the LSM20-based devices with different contact area (A) present multilevel states for the HRS
and the LRS which are programmable at different Vreset and Vset amplitude, respectively (Figure 5.13c).
Comparable response is observed in the larger devices in LSM50 (A=3002 and 4002 µm2), which would
suggest a similar mechanism, although further studies are required to build and validate a
phenomenological model for this composition. On the contrary, only the small devices (A=1002 and
2002 µm2) in LSM50/LAO show constant HRS, and equal to that measured at the largest Vreset amplitude,
i.e. initialization voltage (Figure 5.13d), which suggests a mixed of interface-type and filamentary-type
resistive switching. Both compositions show relatively large HRS/LRS ratio; up to 36 at Vamp=12 V for
LSM50; and up to 23 at Vamp=20 V (around 5 at Vamp=12 V) for LSM20. These findings are congruent
with the work of Choi et al.[104] where the memristive performance is expected to improve as the content
123
of Sr2+ in LSM (x=0.1- 0.5) increases; and in agreement with the work of Asanuma et al.[110] where the
largest HRS/LRS ratio was found for PCMO memristive devices with x=0.4.
In all the cases (composition and size of device), the initialization step is fundamental to create (more)
oxygen vacancies required to trigger resistive switching and at the same time, homogenize their
distribution to obtain a good reproducibility. In agreement with the easy oxygen vacancy formation in
LSM50, the devices fabricated on that film present larger operating window at smaller writing voltages.
Conclusions
In this chapter, we have presented for the first time reliable and reproducible resistive switching in
La0.5Sr0.5MnO3 (LSM50) thin films. The LSM50 layers were grown on LaAlO3 single crystal substrates
and planar memristive devices were micro-fabricated using Ti and Pt as active and inert electrodes,
respectively. Although a phenomenological mechanism of RS in LSM50 was not presented, additional
characterization to propose one model is currently on going. The main points are summarized:
The nanostructure and oxygen migration in LSM50:
 The LSM50/LAO film presents epitaxial growth with two crystal domains, i.e. (00l)- and (hh0)-
out-of-plane-oriented, both under in-plane compressive strain.
 The LSM50 film presents long range atomic order with the presence of extended defects,
probably related to strain relaxation.
Remarks on the LSM50 memristive response:
 The Ti/LSM50\Pt planar memristive devices fabricated on LSM50/LAO have shown to be

promising candidates for VCMs.
 The pristine devices might present a thin oxygen deficient TiOx interlayer spontaneously formed
during the fabrication of devices where the oxygen is uptaken from the LSM50, and thus
creating an oxygen-depleted LSM50 interlayer. Both interlayers are expected to increase the
resistance.
 The reversible oxidation and reduction of the TiOx and reduced LSM50 interlayer can account
for the multilevel capabilities observed in Ti/LSM50\Pt devices in agreement with the R(V)
characteristics where a gradual transition between the high and low resistance states are
observed.
 The facile oxygen vacancy formation in LSM50 combined with the enhanced oxygen transport
properties along dislocations and Ti acting as oxygen reservoir, might be related to the large
124
5.4 Conclusions
operating window at reasonably low writing voltages, i.e. HRS/LRS up to 36 at Vamp=±12 V in

devices with contact area A=2002 µm2.
 The area-scaling resistance suggest an interface-type resistive switching in large devices
(A=3002 and 4002 µm2) for which the HRS and the LRS show multi-states.
 A mixed interface-type and filamentary-like switching (mainly filamentary) is observed for the
small devices (A=1002 and 2002 µm2) for which only the LRS show multi-states while the HRS
does not depend on Vreset and it is defined by the initialization voltage.
 The initialization step is fundamental as it is expected to create more oxygen vacancies, required
to trigger RS, and in the case of the small devices, it will determine the HRS value.
Remarks on the LSM/LAO-based devices with x=0.2 and 0.5:
 The easier reduction of LSM50, i.e. formation of oxygen vacancies, compared to LSM20, has
been momentarily ascribed to the larger HRS/LRS ratios, even at smaller operating voltage.
 In both systems, the initialization step is required to generate oxygen vacancies, homogenize
them in the LSM, and obtain reproducible device performance.
125
Effect of oxygen stoichiometry in LSM films
structural and electrical properties
Sensing oxygen by Raman spectroscopy ............................................................................ 128

Raman spectroscopy on LSM20 annealed samples ................................................... 128
Raman spectroscopy on LSM20 and LSM50 annealed samples ............................... 130
In-situ Raman during annealing of LSM20 sample ................................................... 131
PAS: Probing point defects .................................................................................................. 133
Resistivity as function of temperature and pO2 ................................................................. 136
XANES: assessing oxygen content through Mn oxidation state ....................................... 139
Conclusions ........................................................................................................................... 145
As discussed in the introduction (Section 1.3), the properties of LSM depend on its oxygen
stoichiometry, which can be modified by thermal treatments.[38], [148]–[151] In VCMs, where the active
material is an oxide such as LSM, oxygen ions are considered the active species, at the core of the
resistive switching phenomena.[23] Therefore, understanding the influence of oxygen content on the
materials properties is of great importance to elucidate the switching mechanism and to optimize the
performance of LSM-based memristive devices. The following section summarizes the experiments and
characterization of LSM films thermally treated in different atmospheres. In order to help the reader to
relate the results from this chapter with those presented in others, Table 6.1 summarizes the samples
discussed in the chapter and their equivalent (or same) sample and the section where it was discussed.
127
Chapter 6: Effect of oxygen stoichiometry in LSM films structural and electrical properties
Table 6.1. Summary of the main samples discussed in the Chapter, their characteristics and equivalency to the
samples discussed in other chapters, where (E) denotes the samples is equivalent and (S) denotes it is the same
sample.
Equivalent to
Section in Sample Thickness Substrate
Chapter Sample in Section
6.1.1 LSM20 ∼100 nm LAO (S) LSM20/LAO 3.2
6.1.2 LSM20 ∼100 nm Si3N4 (S) LSM20/ Si3N4 3.2
6.1.2 LSM50 ∼100 nm Si3N4 (S) LSM50/ Si3N4 3.2
6.2 LSM20 ∼20 nm LAO (E) LSM20/LAO 4.2.2
6.3 LSM20 ∼8 nm STO (E) LSM20/STO 4.2.1
6.3 LSM20 ∼8 nm LAO (E) LSM20/LAO 4.2.1
6.4 LSM20 ∼8 nm Si3N4 (S) LSM20/ Si3N4 4.2.1
Sensing oxygen by Raman spectroscopy

Raman spectroscopy is a non-destructive technique capable of providing local insight of the
microstructure and composition. As described in Section 1.2, the oxygen migration and content is crucial
to define resistance states in VCMs. The experiments described in this subsection aim to observe the
change of the Raman spectra with the oxygen content and to establish a ground for operando resistive
switching-Raman measurements. A description of the Raman spectra in LSM is given in Section 3.2.4.
Raman spectroscopy on LSM20 annealed samples
Given the limited number of samples available, this experiment was carried out using the ca. 100 nm
thick LSM20/LAO films grown 750 °C, during the optimization of the composition study (Section 3.2).
For the sake of comparison, the sample was cut in four pieces and the pieces were annealed at 500 °C
128
6.1 Sensing oxygen by Raman spectroscopy
for 1 h in different atmospheres, namely O2, Ar, and 5% H2/Ar. Figure 6.1 summarizes the a-d) SEM,
e) XRD and f) Raman spectroscopy carried out on the pristine and annealed samples.
No signal from secondary phases is found by XRD after any thermal treatment (Figure 6.1e). The out-
of-plane cell parameter of the most relaxed contribution (peak ca 46.5°) has slightly decreased respect
the pristine sample (3.902±0.002 Å) for the oxygen (3.893±0.001 Å), argon (3.899±0.002 Å) and
hydrogen (3.898±0.002 Å) annealed samples. However, the shift is very small and within the
experimental error to extract further conclusions than some microstructural processes, such as
crystallization or partial strain relaxation may occur, not strictly related to the atmosphere. Thus, longer
and/or higher temperature anneal might be required to induce a larger change in the lattice parameter.
The relative shift between the annealed samples is very small where the oxidized sample is expected to
have the smallest lattice parameter as a result of Mn oxidation, and the reduced samples (Ar and 5% H2)
would exhibit a larger one.[152], [153] As shown in Figure 6.1a-d, only the oxidized samples b) presents
some larger grains which can be related to the growth of grains as result of oxygen incorporation.
The Raman spectra (Figure 6.1f) of the annealed samples reveals differences based on the annealing
atmosphere (the Raman spectra of the pristine sample was acquired on another day than for the annealed
samples). The oxidized sample has lower signal of the broad Jahn-Teller bands (JT, especially that
around 600 cm-1) as result of the oxidation of Mn3+ to Mn4+. As described in Section 1.3, JT distortion
is a tetragonal distortion of the unit cell originated by the loss of degeneracy of the half-filled eg band in
Mn3+ ions (Figure 1.5). The other oxidation states of manganese (Mn2+ and Mn4+) present degenerated
states for the t2g and eg orbitals and therefore no distortion. Congruently, the samples annealed in
reducing atmospheres, i.e. Ar and 5% H2, show larger intensity and narrowing of the JT bands indicating
a larger concentration of Mn3+ as result of charge compensation due to the formation of oxygen
vacancies. The fitting of the JT bands is not trivial as the position found from the fitting depends on the
relative intensity of the other peaks, particularly the peak at 660 cm-1 (labeled as ⨂ in Figure 6.1); thus,
the wavenumber maxima of each mode are given to illustrate the trend rather than the exact position.
Furthermore, the peak around 660 cm-1 is centered at smaller Raman shift for the samples annealed in
reducing atmospheres. Unfortunately, this trend cannot be explained as the peak ⨂ is matter of
controversy as it is suggested to be either impurities of binary oxides, [130] or symmetrical stretching of
the MnO6 octahedra.[128]
129
Figure 6.1. Comparison of LSM20/LAO pristine and samples annealed in different atmospheres. SEM of the
a) pristine, b) oxygen, c) argon and d) 5% hydrogen annealed samples. e) XRD pattern and f) Raman spectra of
the pristine and annealed samples. The Raman acquisition time for the pristine sample was different than that of
the annealed films.
Raman spectroscopy on LSM20 and LSM50 annealed samples
The previous thermal treatment (500 °C for 1 h in Ar and O2) was carried out on another batch of
ca. 100 nm thick films of LSM20 and LSM50 grown at 700 °C on LAO and Si3N4 substrates. The
objective was to assess the change of the Raman spectra in LSM films to evaluate the possibility to use
this technique for mapping the oxygen concentration of RS devices.
130
6.1 Sensing oxygen by Raman spectroscopy
Figure 6.2. Raman spectra of a) LSM20 and b) LSM50 grown at 700 °C on Si3N4 (green and blue on top) and
LAO (red and black on bottom) substrates annealed at 500 °C for 1 h in Ar and O2 atmosphere.
The Raman spectra of the annealed LSM20 and LSM50 films are presented in Figure 6.2a, b,
respectively. The trend previously described remains valid for LSM20/LAO grown at 700 °C: Ar-
annealed sample shows JT modes of larger intensity and smaller width compared to the film annealed
in O2. The same trend is found on the polycrystalline samples although the silicon peak at 520.7 cm-1
disturbs the interpretation of the JT band at ∼500 cm-1 (Figure 6.2a).
The Raman spectra of the annealed LSM50 do not show any difference with the annealing atmosphere
(nor with the pristine spectra, not shown) (Figure 6.2b). The characteristic JT distortion of LSM20 does
not take place for LSM50 as a result of the tetragonal symmetry (phase transition at x=∼0.47).[49]
Nonetheless, if the rhombohedral phase was present as a result of different stoichiometry, the large Mn4+
content induced by Sr substitution would decrease the JT Raman bands similar to La0.5Ca0.5MnO3.[126]
In-situ Raman during annealing of LSM20 sample
Once the relation between annealing atmosphere and the changes in Raman spectra were assessed, the
following experiment aimed to track the evolution with temperature. Therefore, the Raman spectra of
LSM20 (100 nm)/LAO pristine film grown at 675 °C were acquired at different temperature in Ar and
O2 atmospheres as schematized in Figure 6.3a. The Raman spectra measured at different temperature
during the cool down in O2 and Ar atmosphere is presented in Figure 6.3c, d, respectively. The intensity
of the Raman modes decreases with temperature and only the modes at 200-210 cm-1 (A1g, MnO6
131
Figure 6.3. In situ Raman spectroscopy on LSM20 (100 nm)/LAO films. a) Scheme of the temperature profile
and acquisitions. b) Raman spectra obtained at room temperature after the annealing experiments. Raman spectra
obtained at different temperature during the cooling in c) O2 and d) Ar atmosphere.
octahedra rotation[126]) and 660 cm-1, which shift to smaller wavenumbers due to thermal expansion, are
visible above 100 °C. The Raman spectra acquired at room temperature after the annealing
(Figure 6.3b) show the same trend as mentioned above: the oxidized sample show smaller signal of the
JT bands, especially at 620 cm-1, and the band at 660 cm-1 shifts to the right. Although the onset of the
oxygen incorporation could not be found, the following statements are drawn:
 The intensity of the JT bands decreases with temperature and the remaining peaks at 210 cm-1
and 660 cm-1 shift to smaller wavenumbers.
 JT bands are sensitive to the atmosphere and hence oxygen content: oxidized samples show
smaller and broader signal than reduced samples.
132
6.2 PAS: Probing point defects
These two conclusions are interesting for in-situ memristive Raman in LSM20 films as they allow
distinguishing between Joule heating and local oxidation (related to oxygen migration during device
operation). Unfortunately, such experiment could not be carried out due to the very low intensity Raman
spectra of the thinner films used for the devices (8-21 nm).
PAS: Probing point defects

As mentioned in the introduction, the presence of defects alters the functional properties of the materials
and therefore, the device performance. In order to gain understanding about the presence/absence,
concentration and type of defects, Positron Annihilation Spectroscopy (PAS) measurements were
conducted at SPONSOR and MePS beamlines of ELBE at HZDR (Dresden-Rossendorf, Germany). DB-
VEPAS (Doppler broadening variable energy positron annihilation spectroscopy) is more sensible to
defect concentration and VEPALS (Variable energy positron annihilation lifetime spectroscopy) senses
the size of defects. Nonetheless, the results from DB-VEPAS and VEPALS are influenced by both the
size and concentration of defects. For a detailed explanation of the principles the reader is referred to
references [154]–[156]. The experiments, results and interpretation presented below were carried out by Dr.
Maciej Oskar Liedke and colleagues from HZDR.
Several 20 nm thick LSM20/LAO films with different conditions were studied: (i) pristine sample,
(ⅱ) sample annealed in Ar (600 °C 1h), (ⅲ) sample annealed in O2 (600 °C 1h), and (ⅳ) pristine sample
capped with 10 nm thick Ti layer and 10 nm thick Pt layer to prevent atmospheric oxidation of Ti. In
the following, a qualitative analysis is presented. A quantitative analysis requires DFT simulations and
fitting with specific software (VEPFit code and PALSfit). Moreover, the small thickness of the films
studied hinders proper evaluation as result of surface and substrate interference, especially for the Ti-
capped sample which interpretation is not included.
DB-VEPAS: defect concentration
The positrons implanted into a solid lose their kinetic energy due to thermalization and eventually
annihilate in delocalized lattice sites or localized vacancy-like defects emitting two anti-collinear
511 keV gamma photons when recombine with electrons. Due to the small momentum of positrons, the
gamma photons experience an energy broadening which can be characterized by two parameters, S and
W, defined as the fraction of the annihilation line in the middle (511±0.93 keV) and the outer regions
(508.33±0.35 keV and 513.67±0.35 keV), respectively. The S-parameter corresponds to the positrons
annihilating with low momentum valence electrons and it is related to vacancy-type defects and their
concentration. The W-parameter is characteristic of the electronic structure and surroundings of the
annihilation site.
133
Figure 6.4. Annihilation line parameters S and W (low and high electron momentum fraction) as a function of
positron implantation energy Ep and mean positron implantation depth <z> LSM20 films treated at different
thermal and ambient conditions.
Figure 6.4 presents low and high electron momentum fraction (S and W, respectively) as function of the
positron implantation energy (and thickness) of the samples. In Figure 6.4, the LSM pristine and
annealed films start at z=0 nm whereas the LSM film for the Ti-capped samples starts around 20 nm.
The relative concentration of defects can be estimated from the slope of the S(Ep) curve for the pristine
and annealed samples. The smaller slope of O2-annealed samples suggests a lower concentration of
defects compared to Ar-annealed sample, which is similar to the pristine sample. Therefore, O2 anneal
reduces the concentration of defects compared to the pristine samples. Nonetheless, the Ar anneal may
have induced positively charged defects, such as positively charge oxygen vacancies, which are not
visible by positrons (coulombic repulsion).
134
6.2 PAS: Probing point defects
Figure 6.5. a) Average positron lifetime and b) positron lifetime components τi, 1st represented by solid symbols
and 2nd represented by open symbols, and c) the relative intensity of the two components I 1=I1/I2 as a function of
positron implantation energy Ep and mean positron implantation depth <z> for the pristine and thermally treated
LSM20 films.
VEPALS: size of defect
The positron lifetime is usually directly proportional to defects size, where the larger the open volume,
the smaller the probability of positrons annihilation and the longer the positrons lifetime. [157]–[159] The
positron lifetime and its intensity has been probed as function of implantation energy (and thickness).
Figure 6.5 presents a) the average positrons lifetime depth profile, b) the two components of positrons
lifetime and c) their relative intensity for the pristine, annealed and Ti capped LSM films. As a result of
the superimposition of annihilation in the surface and substrate with film signal, the interpretation of
results is complex. The film is located around an implantation energy Ep=2 keV for the non-capped films.
At this energy, the change in average positron lifetime of the two components, τav, (Figure 6.5a) shows
a large variation of average defect size. Nonetheless, it may be influenced by increasing contribution of
surface states (longer lifetimes) due to lower trapping in the film (less trapping centers). As shown in
Figure 6.5b, the lifetimes decay down to the values expected for oxygen vacancy (V O) and bulk. As in
DB-VEPAS, the oxygen vacancies are weak trapping centers and can only be detected if they are neutral.
The component τ1 of the pristine sample lays between a mono-vacancy and bulk, which can be assigned
to dislocations. This type of defect is the most common, as it shows the largest intensity (Figure 6.5c),
and decreases its size with thickness (constant intensity and variable τ1). Whereas the Ar- and O2-
annealed samples also present this type of defect, the intensity is lower suggesting a reduction of
trapping cross-section or defect density. Most likely, the size is the same and hence, Ar leads to a change
135
in the trapping yield probably related to the formation of neutral oxygen vacancies upon Ar annealing
(larger intensity compared to O2 anneal). The second component, τ2, represents the surface states (due
to small thickness) or vacancy clusters at the grain boundaries (if exist at all).
Resistivity as function of temperature and pO2

In order to evaluate the dependence of the resistivity on the temperature, two experiments were designed:
(i) at high temperature (up to 600 °C) and (ⅱ) at low temperature (down to -188 °C). The temperature
profiles of each experiment (high and low temperature) are schematized in Figure 6.6a and b, where the
heating and cooling rates were set at 10 °C/min. At the end of each temperature plateau, the resistivity
was measured by van der Pauw method using silver contacts painted on the corners of the 10×10 mm2
samples of 8 nm thick LSM20 grown on SrTiO3 (STO) and LAO substrates (these samples are
equivalent to those used to build memristive devices, which are presented in Section 4.2). A current of
10 μA (50 μA) was applied on two consecutive electrodes of the LSM20/LAO (LSM20/STO) sample
and the voltage drop was continuously measured on the other two electrodes during the experiments.
More information about the van der Pauw method and the measurement configuration is given in
Annex A.2.7.
At room temperature, the resistivity of strained LSM20/LAO (74 mΩ·cm) is larger than in the relaxed
LSM20/STO (20 mΩ·cm). The values are in agreement with the effect of epitaxial strain for which the
resistivity increases with the magnitude of biaxial strain.[59] Similar values were found in equivalent
samples, such as those used for memristive devices in Section 4.2 (12 mΩ·cm for LSM20/STO; and
97 mΩ·cm for LSM20/LAO).
In the high temperature experiment (Figure 6.6a), the samples were heated up to different maximum
temperatures, i.e. 50, 300, 400, 500 and 600 °C, temperature which was maintained during 1 h (and 2 h
at 600 °C). Before increasing the temperature, the sample was cooled down to 50 °C and re-measured
in van der Pauw. This experiment was carried out first under a flow of N2 (pO2< 0.003 mbar) and then
under O2 (1 bar). The evolution of resistivity during the high temperature experiment in N2 and O2
atmosphere in the LSM20/STO and LSM20/LAO films is presented in Figure 6.7a and b, respectively.
In the low temperature experiment (Figure 6.6b), the temperature was first raised up to 250 °C to assess
reproducibility with the previous experiment, followed by a cooling to 25 °C and 0 °C, temperature at
which the resistivity was measured. Then, the samples were cooled down from 25 to – 190 °C in steps
of 10 °C. Finally, the samples were heated up to 25 °C, 250 °C, and back to 25 °C to assess the impact
of the low temperature experiment on the electrical properties at room and mid-temperature. This
136
6.3 Resistivity as function of temperature and pO2
a) b)
600 Temperature profile Temperature profile
Van der Pauw acquisition 200 Van der Pauw acquisition
Temperature (°C)
Temperature (°C)
500 Rate: 10°C/min Rate: 10°C/min

100
400
300 0
200
-100
100
-200
0
0 2 4 6 8 10 12 14 16 0 1 2 3 4 5 6
Time (h) Time (h)
Temperature (°C) Temperature (°C)
c) -200 -100 0 100 200 300 400 500 600 d) 700 300 100 0 -100 -150 -187
1x100
@ O2 (1 bar) 5.5 @ O2 (1 bar)
5.0
log(σ·T) (S cm-1 K)
Resistivity (Ω.cm)
1x10-1 4.5
LSM20 (8nm) / LAO LSM20 (8nm) / STO
4.0
3.5
1x10-2 3.0
LSM20 (8nm) / STO LSM20 (8nm) / LAO
2.5
2.0
1x10-3
0 100 200 300 400 500 600 700 800 900 1 2 3 4 5 6 7 8 9 10 11 12
Temperature (K) 1000/T (K-1)
Figure 6.6. Electronic conductivity of 8 nm thick LSM20 films grown on LAO and STO substrates in O 2
atmosphere as function of temperature. Temperature profile a) at high temperature (600 °C) and b) low
temperature (-188 °C) where the voltage drop was continuously measured (solid line), and the resistivity was
measured by Van der Pauw method at the end of each plateau (sphere). c) Resistivity of the LSM20/LAO and
LSM20/STO films under O2 atmosphere as function of temperature where the arrow indicates the MIT transition.
d) Van der Pauw electronic conductivity of LSM20/LAO and LSM20/STO films under O 2 atmosphere as function
of temperature, plotted according to small polaron model.
experiment was carried out under a flow of O2 (1 bar) and the temperature was maintained for 10 min
at each set-point, i.e. plateau, before performing the van der Pauw measurement.
Figure 6.6c presents the evolution of resistivity, calculated from the voltage drop during continuous
acquisition, for the epitaxial LSM20/LAO and LSM20/STO samples in the whole range of temperature
studied, i.e. -188 °C to 600 °C (85-873 K). The resistivity of LSM20/STO presents a transition from
insulating (or semiconducting) state at low temperature to a metallic state at TMIT=324 K, in agreement
with 10 nm thick LSM20/STO (TMIT=335 K) films reported in literature.[93] On the other hand, the
LSM20/LAO sample is semiconducting in a much larger range of temperature with TMIT=113 K. After
137
the heating experiment first in N2 and then in O2, the resistivity at room temperature of LSM20/LAO
(40 mΩ·cm) is remains larger than in LSM20/STO (17 mΩ·cm). These values are smaller than in the
pristine state by a factor 1.2 and 1.9 for LSM20/STO and LSM20/LAO, respectively. As shown in
Figure 6.7a and b, the resistivity of the two films decreases over time (smaller value for the cooling
than during the heating) in N2 and also in O2 atmosphere for which the decrease is larger. The smaller
resistivity found after the high temperature experiment suggests a change in the structure in both
atmosphere, and/or increase of oxygen content in O2 atmosphere. To the best of our knowledge, the
change in slope in the resistivity curve observed in both samples at T≈750-800 K has not been reported
in literature. The change might be related to oxygen incorporation, as smaller changes of slope are found
in N2 atmosphere. Another explanation to the change of slope is film recrystallization at high
temperatures rather than to a phase transition.
The conductivity of the LSM20/LAO and LSM20/STO films measured in O2 atmosphere is presented
as an Arrhenius plot in Figure 6.6d. At high temperature (>RT), the conduction mechanism of both
films is in agreement with the small polaron hopping where the polarons move to the Nearest Neighbor
Hopping site with the assistance of phonons.[38], [54] The conductivity of this mechanism in the adiabatic
limit is expressed as in Equation 6.1,
𝜎 𝐸𝑎
𝜎(𝑇) = ∙ exp ( )
𝑇 𝑘𝑏 𝑇
Equation 6.1
where Ea is the activation energy of the hopping mechanism, kb is the Boltzmann constant and σ0 is the
pre-exponential factor. The activation energy measured from Figure 6.6d in the range 300-600 °C in
LSM20/LAO (0.049 eV) and in LSM20/STO (0.063 eV) is slightly smaller than the values reported for
epitaxial La0.8Sr0.2MnO3 (0.075 eV) and polycrystalline films (0.18-0.21 eV).[54] Moreover, during the
change of atmosphere from nitrogen to oxygen at room temperature, the resistance measured decreases
which suggests p-type carriers (electron holes) involved in the electronic transport, as expected.
MIT transition temperatures close to operating temperature of volume-type resistive switching devices
have been correlated to better performances, i.e. larger HRS/LRS ratios and smaller transition voltage
with narrower distribution.[93] Therefore, our LSM20/STO-based devices are expected to exhibit larger
resistance ratios than devices based on LSM20/LAO. However, interface-type switching is observed
and LSM/STO devices show a narrower operating window than those in LSM20/LAO. Whereas the
intrinsic properties of the material, in this case assessed by the MIT, are important for the performance
of the devices, one should consider all the elements of the device. We ascribe the performance observed
to the use of different electrodes where the interface and oxygen exchange between film and electrode
138
6.4 XANES: assessing oxygen content through Mn oxidation state
a) 1x10-1 LSM20 (8nm) / STO

b) 1x10-1 LSM20 (8nm) / LAO
Resistivity (Ω.cm)
Resistivity (Ω.cm)
1) N2 atmosphere 1) N2 atmosphere
time
1x10-2 time
2) O2 atmosphere 1x10-2 2) O2 atmosphere
1x10-3
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Temperature (°C) Temperature (°C)
Figure 6.7. Electronic resistivity of the 8 nm thick a) LSM20/STO and b) LSM20/LAO films as function of
temperature and atmosphere (N2 in blue, O2 in red). The temperature profile is shown in Figure 6.6a, where the
samples were first measured in N2 and then in O2 atmosphere. The resistivity decreases with time, related to the
temperature profile, under both atmospheres.
are intrinsically different, and more significantly to the different microstructure. The memristive
performance of epitaxial LSM20/STO and LSM20/LAO is discussed in Chapter 4.
XANES: assessing oxygen content through Mn oxidation state

In order to get direct insight into the mechanism involved in RS phenomenon in La1-xSrxMnO3 films, X-
ray absorption near-edge spectroscopy (XANES) was carried out in beamline ID12 at ESRF (Grenoble),
under the experiment number MA-4802. As described in Annex A.1.2, the edge position of elements is
related to their oxidations state. Therefore, the determination of the Mn K-edge gives insight of the
oxidation state which is expected to change as result of the variable local oxygen content of the
memristive devices. These experiments are strongly related with the devices presented in Chapter 4 and
gives the clue to the mechanisms proposed for RS behavior.
The experiment comprises two parts: (i) the measurement of LSM films annealed in different conditions
and (ⅱ) the measurement of the devices in different locations, resistance states and in operando
conditions. The edge positions were corrected respect to the manganese metal K-edge at 6539.05 eV,
measured on Mn88Ni12 alloy. Three manganese oxides, namely MnO, Mn2O3, and MnO2 with nominal
valence +2, +3 and +4 respectively, were measured previously and are used to determine the formal
valence of the LSM films. The spectra are corrected to the Mn K-edge of Mn alloy measured at the same
time. Figure 6.8 presents the normalized spectra at Mn K-edge of the Mn alloy and the Mn oxides where
139
2.0
1.8 Mn88Ni12
Normalized intensity (arb. u.)

MnO
1.6
Mn2O3
1.4 MnO2
1.2
1.0
0.8
0.6
0.4
0.2
0.0
6535 6540 6545 6550 6555 6560 6565

Energy (eV)
Figure 6.8. Normalized XANES spectra at the Mn K edge position for Mn references with different oxidation
state.
the edge position is indicated by a vertical line. The edge position of the Mn alloy corresponds to the
maximum of the first derivative whereas for the Mn oxides it was assigned to half maximum of the edge
step. This adjustment returns an increase of 0.31 valence per 1 eV, close to the commonly reported
change of 0.24-0.28,[160]–[162] even if larger changes of 0.58 have been also reported.[163]
Annealed samples
According to the stoichiometric formula, the formal valence of Mn in La0.8Sr0.2MnO3 and La0.5Sr0.5MnO3
is +3.2 and +3.5, respectively, which can be further modified by changing the oxygen content.
Polycrystalline ca. 20 nm thick films of LSM20 and LSM50 grown on Si3N4 were annealed at 600 °C
(10 °C/min) for 1 h on O2 and Ar atmosphere to serve as internal reference for the LSM-based devices.
The XANES spectra at the Mn K-edge, the first derivative and the edge position of the annealed samples
is presented in Figure 6.9. The spectra of the samples share common characteristics in the pre-edge and
EXAFS regions that indicate a similar atomic environment (a detailed analysis of the EXAFS to
determine precisely the environment is out of the scope of the experiment). However, the edge position
is shifted towards larger energies for the samples annealed in O2 atmosphere, which suggests larger Mn
oxidation state than their Ar-annealed counterpart. The LSM20 films present a smaller edge position
than the LSM50 annealed in the same atmosphere, which indicates a lower Mn valence as expected from
the lower Sr content.
140
1.4 XANES
1.2
1.0
0.8
0.6
Normalized intensity (arb.u.)
LSM20/Si3N4 - Oxidized 6554.281 eV
0.4 LSM20/Si3N4 - Reduced 6553.477 eV
0.2 LSM50/Si3N4 - Oxidized 6554.683 eV
0.0 LSM50/Si3N4 - Reduced 6554.281 eV

0.3
0.25 1st derivative
0.2
0.20
0.1
0.15
0.0
0.10
-0.1
0.05 6545 6550 6555 6560
0.00
-0.05
-0.10
-0.15
6520 6540 6560 6580 6600 6620
Energy (eV)
Figure 6.9. XANES spectra (top) and first derivative (bottom) at the Mn K-edge of LSM20/Si3N4 and
LSM50/Si3N4 films annealed in oxidizing and reducing conditions.
LSM20-based pristine memristive devices
The Mn oxidation state was assessed on Ti/LSM20\Pt devices fabricated on 8 nm thick LSM20/Si3N4
(detailed explanation of the memristive characteristics of this sample is given in Section 4.2.2) using a
beam spot size of 5 μm. Three regions of interest (ROI) were defined on the memristive devices: ROI_1
under the Ti contact, ROI_2 the inter-contact region, and ROI_3 under the Pt contact. In first two ROI,
two different regions were further defined to investigate the presence of a gradient. Hence, five locations
were measured, as schematized in Figure 6.10a. The XANES spectra and the first derivative acquired
in the five locations of a Ti/LSM20\Pt pristine device is shown in Figure 6.10b. No significant change
in the edge position is found for the locations in the same ROI. However, a shift is observed between
the different ROIs. It can be clearly observed that the Mn K-edge under the Ti contact is shifted to
smaller energies (E0=6550.81 eV) compared to the position under the Pt contact (E0=6553.68 eV) and
the inter-contact region (E0=6553.82 eV). This suggests that the Mn under the Ti electrode is reduced
compared to the rest of the film.
141
a) b)
2.5
XANES
2.0
(5) ROI_3
Pt Pt 1.5

Ti / TiOx
1.0
(1) (2) (3)(4) (5) (3)
(1) ROI_2
LSM20 0.5 ROI_1 (4)
(2)
Si3N4 0.0
0.4 0.4 ROI_3

1st derivative
ROI_1 ROI_2
0.3 0.2
ROI_1: (1) Center of Ti contact
0.0
(2) Edge of Ti contact 0.2
-0.2
0.1
6545 6550 6555 6560
ROI_2: (3) LSM, next to Ti 0.0
(4) LSM center
-0.1
-0.2
ROI_3: (5) Edge of Pt contact
6520 6540 6560 6580 6600 6620
Energy (eV)
Figure 6.10. a) Cross-section scheme of the regions of interest (ROI) in the Ti/LSM20\Pt memristive device on
Si3N4. Points 1-2, ROI_1 under Ti electrode; 3-4, ROI_2 inter-contact region; 5 ROI_3 under Pt electrode.
b) XANES spectra and first derivative of the regions of interest.
Discussion: Mn valence and oxygen stoichiometry from XANES
Figure 6.11 summarizes a) the Mn K-edge position, b) the calculated Mn formal valence and c) oxygen
off-stoichiometry, δ, of the annealed LSM20 films together with the three ROI in the pristine device.
The Mn valence was estimated from the edge position and formal valence of the binary Mn oxides
(Figure 6.8) while δ was calculated assuming the principle of charge neutrality in La0.8Sr0.2MnO3+δ.
As shown in Figure 6.11b, the average Mn valence of LSM20 films annealed in O2 and Ar are +3.44
and +3.69, respectively, which are larger than the valence of stoichiometric LSM20 (+3.2). In agreement
with this, the valence value found for the pristine device in the ROI_2, inter-contact region is +3.55.
These valence values indicate the presence of oxygen excess, as shown in Figure 6.11c. These results
fit well with those reported for LaMnO3, La0.7Sr0.3MnO3 and La0.5Sr0.5MnO3, with Mn average valence
of 3.4, 3.7 and 3.8, respectively.[162]
142
a)
ROI_2: Inter-contact
ROI_ 1: Under Ti
LSM20 Reduced
LSM20 Oxidized
ROI_3: Under Pt
.81
65 .48
.68
.82
.28
50
53
53
53
54
65
65
65
65
Mn K-edge (eV)
b) c)
ROI_ 1: Under Ti
ROI_ 1: Under Ti
LSM20 Reduced
LSM20 Reduced
LSM20 Oxidized
LSM20 Oxidized
ROI_3: Under Pt
ROI_3: Under Pt
2.6
2.8
3.0
3.2
3.4
3.6
3.8
0.0
0.1
0.2
0.3
3
1
-0.
-0.
-0.
2.63 3.44 3.50 3.55 3.69 -0.28 0.12 0.15 0.18 0.25
Mn valence δ in La0.8Sr0.2MnO3+δ
Figure 6.11. a) Mn K-edge position, calculated b) Mn valence and c) oxygen off-stoichiometry, δ, of the annealed
LSM20 films and the three ROI within a pristine LSM20-based memristive device.
We compare now the valence value in the device. The Mn valences found for the ROI_3, under the Pt
contact (+3.50) indicates that the Mn under the Pt contact is more reduced than the uncovered LSM film,
with a value of +3.55. Whereas these two valence values lay between those of the annealed samples in
reducing and oxidizing atmospheres, the Mn in ROI_1, under the Ti contact, presents an average valence
of +2.63. This valence value corresponds to δ= - 0.28 indicating the presence of oxygen vacancies under
the Ti contact.
The oxygen vacancies are consequence of the Ti high oxygen affinity, as suggested by Sawa et al. for
Ti/Pr0.7Ca0.3MnO3 stack.[109] This observation is in agreement with the oxygen-deficient TiOx interlayer
observed upon Ti evaporation (see Section 4.1 and Figure 4.3) and indicates that oxygen is “pumped”
from the LSM underneath the Ti electrode. Hence, and as discussed in Chapter 4 and Chapter 5, the
presence of oxygen vacancies in Ti/LSM\Pt devices triggers the memristive response within the LSM.
While the LSM under the Ti contact is strongly reduced (Mn+2.63), the LSM next to the metal contact
(labeled as (3) in Figure 6.10a) is slightly oxidized (Mn+3.44). Noteworthy, the apparent absence of an
oxygen gradient between these regions is due to the volume sensed (across the 8 nm thick LSM film
and part of the substrate). Therefore, it cannot be excluded that the Ti uptake oxygen from the outermost
143
layers next to the contact. The latter consequence of the volume sensed, implies that the Mn valence is
an average across the film and hence, the LSM in contact with the Ti could be further locally reduced,
which results in an increase of the LSM resistivity[52], [142] crucial for the memristive devices.
Several devices were programmed at high and low resistance states (HRS and LRS, respectively) to
assess the change of Mn oxidation state as function of the resistance state, before performing the XANES
Mn K-edge measurements. However, no difference was observed by XANES in any of the ROI.
Figure 6.12a shows the results for the ROI_1, under the Ti contact in the pristine, HRS and LRS states.
When checking the R-I(V) characteristics of the device after beam exposition, it was found that they are
different than those before exposure and the cycling sense does not allow programming the LRS
(Figure 6.12b). At this stage, we propose that the beam is injecting current into the device via
photoelectron generation of silicon substrate, probably related to the insulating substrate, and the
memristive response cannot be triggered as in laboratory conditions. Several solutions were tested,
among them removing the electric cables and tips from the contacts, diluting the photon density of the
beam and grounding the insulating Si3N4 substrate (scratched to reach the Si layer and bridge them with
silver paste to a mass). Only a low photon density beam with a spot size of 200 μm (density divided by
factor 40) allowed obtaining similar R(V) characteristic although no difference was observed in XANES
and the operando measurements remained unachievable. Therefore, using a conducting substrate seems
to be crucial to perform operando XANES, as in the case of TiN/La2NiO4/Pt vertical devices
(manuscript in preparation by Khanh Khuu from Nanoionics, LMGP).
a) 1.6 b)
1.4 ROI_1 HRS Before beam

Measured Resistance (Ω)
LRS 108 After beam

1.2 3
Pristine
1.0 4 2
0.8 1.5
107
0.6
1.0 2
0.4 3 1
0.5
0.2
106 1
0.0 0.0
6540 6545 6550 6555
4
-0.2
6520 6540 6560 6580 6600 6620 -10 -8 -6 -4 -2 0 2 4 6 8 10
Energy (eV) Applied Voltage (V)
Figure 6.12. a) Mn K-edge XANES spectra in ROI_1, under the Ti contact, for pristine devices and devices
programmed ex-situ at different resistance states. b) Comparison of the R(V) characteristics in the same device
before and after being exposed to the beam showing the change in the sense of rotation of the resistance states and
volatility of the HRS/LRS values.
144
6.5 Conclusions
Conclusions
The conclusions withdrawn from this chapter are detailed below.
Regarding Raman spectroscopy:
 Raman spectroscopy can be used to observed differences in oxygen content in LSM20 films.
 The most oxidized sample shows lower intensity of the Jahn-Teller (JT) modes at 500-
530 cm- 1 and 600-620 cm-1 compared to the pristine sample.
 The most reduced sample exhibit larger intensity of JT modes.
 No onset of oxygen incorporation/excorporation could be found by in situ Raman, as the
Raman signal decreases with temperature.
Regarding the positrons measurements conducted at ELBE and HZDR:
 From DB-VEPAS: the O2-annealed LSM20/LAO samples present a lower concentration of

defects compared to pristine and Ar annealed films. However, the later film might present
defects non-visible to positrons for instance, positively charge oxygen vacancies.
 From VEPALS: the presence of oxygen vacancies point defects related to dislocations are the
most common in pristine LSM20 films, and their size decreases with thickness.
Regarding the conductivity as function of temperature and pO2:
 The LSM20/STO film exhibit metallic behavior at room temperature where the MIT occurs at
324 K.
 The LSM20/LAO film show semiconducting behavior over a large range of temperature where
the MIT takes place at 113 K.
 Both types of epitaxial films, LSM20/STO and LSM20/LAO, show an increase of conductivity
with pO2, in agreement with a p-type conductivity type.
 The increase of conductivity at high temperature has been observed in both films and at high
and low pO2. No further change of conductivity has been observed with the pO2.
Regarding XANES conducted at ESRF ID12 beamline:
 The Mn K-edge in LSM/Si3N4 shifts towards larger (lower) energies for the film annealed in
O2 (Ar) atmosphere, in agreement with the incorporation (release) of oxygen and concomitant
Mn oxidation (reduction).
 The Mn valence of O2- and Ar-annealed LSM20 films, +3.44 and +3.69, respectively, suggest
and oxygen excess.
145
 XANES was acquired in different regions of interest of a pristine planar Ti/LSM20\Pt

memristive device. The Mn valence of LSM under the Pt contact and inter-contact region are
+3.50 and +3.55, respectively. The values lay within the boundaries defined by the annealed
samples. However, the LSM film under the Ti contact is extremely reduced with an average
Mn valence of +2.63.
 The strongly reduced LSM under Ti is in agreement with the oxygen scavenging that occurs
during Ti evaporation and spontaneous formation of an oxygen-deficient interlayer of TiOx
(discussed in Section 4.1).
 The XANES of devices previously programmed at different resistance states are identical. The
memristive response was altered after beam exposure: the resistance states were volatile and
the cycling sense did not allow programming the low resistance states. Hence, the change of
Mn oxidation state during memristive operation could not be assessed. The injection of
electrons photogenerated in the silicon-based substrate (Si3N4/SiO2/Si) is the most plausible
explanation. Hence, using conducting substrate might be a limiting factor to perform electrical-
related operando XANES.
146
Concluding summary and prospects
Valence change memories (VCMs) have been proposed as alternative memories with interesting
features, as they show the fast operation characteristic of the RAM combined with the non-volatility of
flash memories. Their performance, however, is greatly influenced by the elements building up the
device, being the oxygen transport within and through the elements a key piece to tune the resistive state
of the memory. In this thesis, we focused on the La1-xSrxMnO3-δ perovskite as an active material for
VCMs, and in particular, on understanding the relation between its microstructure and its functional
properties. LSM can accommodate different oxygen stoichiometry by the valence change in Mn
(Mn3+/Mn4+) which in turn modifies its electronic conductivity making it an interesting candidate for
VCMs.
LSM was synthetized as a thin film by PI-MOCVD after optimization of the deposition parameters. The
first reactor optimization led to a homogenous deposition area of roughly 2 inches, and was achieved by
modifying the gas flow, its injection point and the relative position of the holder and the precursor. The
second optimization consisted in the calibration of the composition of the LSM films as a function of
precursor solution composition. The two selected compositions, LSM with 20% and 50% Sr substitution
were obtained:
 La0.8Sr0.2MnO3-δ (LSM20): using a solution with [(La+Sr)/Mn]=2.0 and [Sr/(La+Sr)]=0.55

 La0.5Sr0.5MnO3-δ (LSM50): using a solution with [(La+Sr)/Mn]=2.3 and [Sr/(La+Sr)]=0.80
The fabrication of planar memristive devices (electrodes on top of the film) was carried out in cleanroom
facilities using laser lithography and metal evaporation, in which, based on the band alignment, Pt and
Ti were selected as inert and active electrode, respectively (Ti/LSM\Pt devices).
LSM20 films
Thin LSM20 films (1800-5000 pulses corresponding to thicknesses of 7 to 20 nm) were grown at 675 °C
on STO and LAO single crystals of STO and LAO, in order to alter the microstructure of LSM by
changing the film-substrate mismatch. TEM cross-section analysis of the LSM20 films showed perfectly
oriented and long-range ordered LSM/STO films with a small lattice mismatch, while the presence of
extended defects was found on the LSM/LAO films, most likely related to a much larger mismatch. The
147
Chapter 7: Concluding summary and prospects
electrical properties of the LSM20 films are also influenced by the nature of the substrate, as seen in the
resistivity temperature dependence measurements. On the one hand, the slightly strained LSM20/STO
films show a smaller resistivity and metallic-type conduction at room temperature, with a MIT around
324 K. On the other hand, the largely strained LSM20/LAO presents a larger resistivity and
semiconducting-type conduction in a wide range of temperature, with MIT at 113 K.
The Ti/LSM20\Pt memristive planar devices fabricated show, in agreement with the band alignment,
that Pt/LSM forms a low resistance contact with ohmic characteristics, while the Ti/LSM contact shows
larger resistance and Schottky-like characteristics. The presence of an amorphous and oxygen-deficient
TiOx interlayer at the Ti/LSM20 interface was observed by STEM-EDS in pristine devices. In agreement
with the high oxygen affinity of Ti, we suggested that this interlayer forms spontaneously during the
evaporation of Ti by scavenging the oxygen ions from the LSM and thus, creating a reduced LSM layer.
The initial resistance of the Ti/LSM20\Pt devices showed scaling with the electrode contact area while
no trend was found with the distance between contacts, compatible with the presence of the two resistive
interlayers.
The devices built on LSM20/STO and LSM20/LAO show CCW R(V), in agreement with the oxygen
drift that would result in the oxidation of Ti and reduction of LSM during the HRS RESET, and the
reverse reaction during the LRS SET. The LSM20/LAO-based standard devices showed multi-states
for the HRS and the LRS which were ascribed to the gradual redox reaction of the Ti/LSM interface.
The different states could be reached by tuning the Vreset and Vset, respectively, being |Vreset|=|Vset|. The
presence of extended defects was suggested to highly increase the drift of oxygen ions, particularly in
the less favorable LRS SET (reduction of Ti and oxidation of LSM), accounting for HRS/LRS operating
windows between 1.5 and 23 at operating voltage Vamp=±6 to ±20 V. In addition, the LSM20/LAO-
based devices showed good retention, cyclability and area-scaling which points towards an interface-
type resistive switching. In contrast, the LSM20/STO-based devices experienced an increase of
resistance at large operating voltage (Vamp >16 V). This was related to the lower concentration of defects
in the LSM20/STO films, in which the SET process is less favored, and would imply a monotonous
oxidation of Ti and loss of carriers upon cycling. Reproducible performance over 100 cycles was
achieved in LSM20/STO and LSM20/LAO by pulsed operation at low voltages (Vamp =±10 V).
Additionally, Ti/LSM20\Pt devices built on polycrystalline LSM20/Si3N4 films showed similar

memristive response to those in LSM20/LAO, i.e. CCW R(V) characteristics, multi-states for the HRS
and the LRS, and large operating window (1.5-17 at Vamp =±4 - ±14 V). The oxygen migration in the
polycrystalline devices would be enhanced along the grain boundaries in both voltage polarities.
However, these devices experience an increase of resistance at large voltage (Vamp >14 V), similar to
that observed in LSM20/STO. The existence of a strongly reduced LSM20 volume under the Ti
electrode was assessed by XANES in pristine devices, for which an average Mn valence of +2.63 was
148
estimated, while in the electrode-free regions of the devices the LSM film has a formal Mn valence of
+3.5 to +3.55.
LSM50 films
Furthermore, and to the best of our knowledge, we obtained for the first time reliable and reproducible
memristive behavior in LSM50-based devices. The highly oriented LSM50/LAO film contains few
extended defects (between LSM20/STO and LSM20/LAO), as observed by TEM. The IR area
dependence (very weak or null dependence with distance) in Ti/LSM50\Pt devices hints towards the
presence of the TiOx interlayer. The memristive device showed CCW R(V) curves and large HRS/LRS
ratio up to 36 even at relatively small voltages (Vamp =±12 V). This performance was linked to the
results in literature, which show a fast oxygen migration along defects and an easier formation of oxygen
vacancies in LSM50 compared to other LSM compositions. The memristive devices showed two
different responses based on the contact area. On the one hand, large devices (A=3002 and 4002 µm2)
showed multilevel properties for the HRS and the LRS. On the other hand, only the LRS in the small
devices (A=1002 and 2002 µm2) showed multi-states while the HRS did not show voltage dependence
and its value was determined by the largest Vreset, i.e. the initialization voltage. Further investigations of
LSM50-based devices are currently ongoing to present a phenomenological description of the
memristive response observed.
Effect of oxygen content in LSM thin films
In all the devices studied in this thesis, the appearance of RS seems to be related to the presence of
oxygen vacancies. These vacancies can be present in pristine devices, and/or be formed by the
application of an intermediate voltage, and their concentration is expected to increase with the amplitude
of the applied voltage. Therefore, the initialization procedure, carried out at operating voltages, is
vital for the formation of new oxygen vacancies and their homogenization, and hence, for the
performance of the LSM-based memristive devices.
Complementary to the characterization of devices, thick LSM20 films (ca. 100 nm) were annealed in
oxidizing and reducing atmosphere to modify their oxygen content. While no clear change was observed
in the lattice parameter, Raman spectroscopy showed to be very sensitive to the modification of oxygen
content. The Raman JT bands of LSM broadened and decreased their intensity for the film annealed in
oxidizing atmosphere, in agreement with the loss of electrons in the Mn 3d orbitals which cause the JT
distortion.
149
In conclusion, the results presented here represent a fundamental study to understand the role of the
nanostructure in LSM-based memristive devices. The characterization and understanding of epitaxial
films also allowed the comprehension of grain-boundary-rich-polycrystalline-film-based devices, which
are more likely to be implemented by the micro-technology industry, especially if they are based on
silicon compatible technologies. Therefore, the polycrystalline-film-based devices are the proof-of-
concept to show the potential of LSM-based as memristive devices. Miniaturization of the devices and
change of geometry (to vertical configuration) would allow decreasing the operating voltage, ideally
below 2 V, in agreement with industry requirements, while preserving the memristive characteristics,
i.e. large operating window and multilevel resistance states. The results could also apply to similar
manganite and perovskite systems, and other mixed ionic-electronic-based technologies, such as micro
solid oxide fuel cells. Hence, engineering the nanostructure by controlling the defects can result in an
effective way of improving the performance of nanoionic devices.
The prospects of this thesis are the following.
Regarding LSM20 memristive devices:
 The performance of LSM20/STO devices seem to improve by asymmetric voltage operation.

Further device characterization would allow understanding the RS mechanism behind
asymmetric voltage operation.
 Similarly, LSM20/LAO may benefit of asymmetric voltage operation where the favorable
RESET process could occur at smaller voltage.
 The polycrystalline LSM20/Si3N4-based devices showed deleterious performance at large
voltage. There is still room for improvement in the operating conditions of this very promising
candidate.
Regarding memristive LSM50 devices:
 Deeper understanding of the LSM50/LAO-based devices is required to build a RS mechanism

model for the different electrode sizes studied.
 Electrical characterization of the readily available LSM50/STO samples, which have already
shown larger operating window than LSM20/STO, but smaller than LSM50/LAO devices.
 As for LSM20, polycrystalline LSM50-based devices may exhibit better performance than the
epitaxial devices.
 To the best of my knowledge, there is a lack of literature reporting the oxygen transport
properties along grain boundaries and dislocations in LSM50 thin films. Therefore, IEDP-SIMS
150
(Ion Exchange Depth Profiling – Secondary Ions Mass Spectroscopy) in LSM50 films with
different microstructures would be of great importance for understanding the material
properties, and by extension the memristive response observed.
From a general point of view, regarding LSM-based memristive devices:
 The phenomenological models proposed are built from the experimental memristive
observations and the experimentally measured and/or reported properties of the materials.
However, further electrical measurements should be carried out to complete and validate the
model, among them:
o Change of active electrode, or removing the top electrode.
o A better comprehension of the influence of thickness (for which the lack of switching
in Ti/LSM20(21 nm)\Pt devices on LAO and Si3N4 still remains unclear).
o Change of the environmental conditions (pO2, humidity, temperature).
o A larger statistical sample.
 In addition, operando characterization would be of great value to elucidate the mechanism
responsible of the resistive switching phenomena. For example:
o Operando XANES to assess the Mn oxidation state (by measuring the Mn K-edge) of
the LSM film and the Ti oxidation state (by measuring the Ti K-edge) during switching
may be possible by using a conducting bottom electrode, such as platinum.
o Operando TEM combined with spectroscopic tools such as EELS would allow to assess
changes in structure, microstructure, oxidation state and chemical composition in
localized regions. However, we emphasize the ultra-high vacuum required for this
experiment which could alter the switching characteristics if oxygen is lost in the
chamber and not reincorporated back in the film.
o Operando Raman spectroscopy is sensible to the change in oxygen content in LSM
films. Even though it was one of the goals in the framework of the PhD, operando
Raman in memristive devices could not be carried out as a result of the low Raman
signal of the very thin films. Some approaches to enhance the signal are the following:
(i) study of thicker films if the device is operable, (ii) the study of rhombohedral LSM
with x=0.17, for which the JT is expected to be larger than for LSM for any other Sr
richer compositions, and/or (iii) operando Raman at low temperatures where the signal
is increased (the memristive response observed might change at low temperature).
 Keeping the end-application in mind, the performance of LSM-memristive devices operated by
pulses with a width down to tens of nanosecond should be addressed. At such low time-scales,
the performance may be limited by kinetic factors, such as oxygen migration, for which the
151
presence of extended defects might be crucial, being the polycrystalline films the most
promising ones.
 Furthermore, the miniaturization (from the microscale to the nanoscale) should be envisioned
and the fabrication of vertical (top-bottom configuration) devices is of great interest. We note
that this change may imply a diverse performance of the devices, for instance they may present
filamentary-type resistive switching or a mix with interface-type.
 For future application in real devices the viability of the vertical LSM devices could be assessed
by: pulsed operation, retention measurements (of multistates) at 85 °C, endurance studies of
>105 cycles, change in environmental humidity and pO2…
 The fabrication of vertical devices, requires of a bottom conducting electrode, such as Pt. In this
work we did not succeed at growing LSM on Pt/Si substrates as the Pt layer suffered dewetting
even at the deposition temperature of 630 °C. However, previous studies have suggested
different strategies to overcome this problem as for example, a two steps deposition where a
partially-crystallized thin layer of perovskite is deposited at low temperature to” stabilize” the
Pt, followed by deposition of a thicker layer at higher temperature to allow the crystallization
of the “two” oxide layers. Nonetheless, this procedure is not straightforward and many new
challenges may appear on the way to obtain a dense and crystallized film.
 Ultimately, the fabrication of a cross-bar array of vertical devices, combining lithography,
etching (for instance RIE), and evaporation techniques could be carried out, which, combined
with the appropriate electrical measurements, would serve as a proof-of-concept for
neuromorphic computing.
152
References
[1] G. E. Moore, “Cramming more components onto integrated circuits,” Proc. IEEE, vol. 86, no. 1,
pp. 82–85, 1998, doi: 10.1109/JPROC.1998.658762.
[2] “International Roadmap for Devices and Systems: Beyond CMOS,” 2021.
https://irds.ieee.org/editions/2021 (accessed Apr. 06, 2022).
[3] S. Y. Wang, D. Y. Lee, T. Y. Huang, J. W. Wu, and T. Y. Tseng, “Controllable oxygen vacancies
to enhance resistive switching performance in a ZrO2-based RRAM with embedded Mo layer,”
Nanotechnology, vol. 21, no. 49, 2010, doi: 10.1088/0957-4484/21/49/495201.
[4] C. Yoshida, K. Tsunoda, H. Noshiro, and Y. Sugiyama, “High speed resistive switching in PtTi
O2 TiN film for nonvolatile memory application,” Appl. Phys. Lett., vol. 91, no. 22, 2007, doi:
10.1063/1.2818691.
[5] Y. Chen et al., “Forming-free HfO2 Bipolar RRAM Device with Improved Endurance and High
Speed Operation,” vol. 29, pp. 37–38, 2009.
[6] J. S. Meena, S. M. Sze, U. Chand, and T. Y. Tseng, “Overview of emerging nonvolatile memory
technologies,” Nanoscale Res. Lett., vol. 9, no. 1, pp. 1–33, 2014, doi: 10.1186/1556-276X-9-
526.
[7] R. Waser and M. Aono, “Nanoionics-based Resistive Switching Memories,” Nat. Mater., vol. 6,
no. 11, pp. 833–840, 2007, doi: 10.1038/nmat2023.
[8] R. Waser et al., “Introduction to Nanoionic Elements for Information Technology: From
Fundamentals of Nanoionic Redox Processes to Memristive Device Applications,” 2016, pp. 1–
30. doi: 10.1002/9783527680870.ch1.
[9] S. Telecommunication and L. Limited, “New conduction and reversible memory phenomena in
thin insulating films,” Proc. R. Soc. London. Ser. A. Math. Phys. Sci., vol. 301, no. 1464, pp. 77–
102, 1967, doi: 10.1098/rspa.1967.0191.
[10] T. W. Hickmott, “Low-frequency negative resistance in thin anodic oxide films,” J. Appl. Phys.,
vol. 33, no. 9, pp. 2669–2682, 1962, doi: 10.1063/1.1702530.
[11] J. F. Gibbons and W. E. Beadle, “Switching properties of thin Nio films,” Solid State Electron.,
vol. 7, no. 11, pp. 785–790, 1964, doi: 10.1016/0038-1101(64)90131-5.
[12] H. Biederman, “Metal-insulator-metal sandwich structures with anomalous properties,” Vacuum,

vol. 26, no. 12, pp. 513–523, 1976, doi: 10.1016/S0042-207X(76)81130-X.
153
[13] A. Asamitsu, Y. Tomioka, H. Kuwahara, and Y. Tokura, “Current switching of resistive states
in magnetoresistive manganites,” Nature, vol. 388, no. 6637, pp. 50–52, 1997, doi:
10.1038/40363.
[14] A. Urushibara, Y. Moritomo, T. Arima, and A. Asamitsu, “Insulator-metal transition and giant
magnetoresistance in La1-xSrxMnO3,” Phys. Rev. B, vol. 51, no. 20, pp. 103–109, 1995, [Online].
Available: http://prb.aps.org/abstract/PRB/v51/i20/p14103_1
[15] D. B. Strukov, G. S. Snider, D. R. Stewart, and R. S. Williams, “The missing memristor found,”
Nature, vol. 453, no. 7191, pp. 80–83, 2008, doi: 10.1038/nature06932.
[16] L. Chua, “Memristor - The Missing Circuit Element,” IEEE Trans. Circuit Theory, vol. CT-18,
no. 5, pp. 507–519, 1971, doi: 10.1109/TCT.1971.1083337.
[17] L. O. Chua and S. M. Kang, “Memristive Devices and System,” Proc. IEEE, vol. 64, no. 2, pp.
209–223, 1976, doi: 10.1109/PROC.1976.10092.
[18] A. Sawa, “Resistive switching in transition metal oxides,” Mater. Today, vol. 11, no. 6, pp. 28–
36, 2008, doi: 10.1016/S1369-7021(08)70119-6.
[19] X. J. Chen et al., “Metal-insulator transition above room temperature in maximum colossal
magnetoresistance manganite thin films,” Phys. Rev. B - Condens. Matter Mater. Phys., vol. 72,
no. 10, pp. 1–7, 2005, doi: 10.1103/PhysRevB.72.104403.
[20] L. Yao, S. Inkinen, and S. Van Dijken, “Direct observation of oxygen vacancy-driven structural
and resistive phase transitions in La2/3Sr1/3MnO3,” Nat. Commun., vol. 8, p. 14544, 2017, doi:
10.1038/ncomms14544.
[21] A. Sawa, T. Fujii, M. Kawasaki, and Y. Tokura, “Interface resistance switching at a few
nanometer thick perovskite manganite active layers,” Appl. Phys. Lett., vol. 88, no. 23, pp. 86–
89, 2006, doi: 10.1063/1.2211147.
[22] Y. W. Xie, J. R. Sun, D. J. Wang, S. Liang, and B. G. Shen, “Reversible electroresistance at the
Ag/La0.67Sr 0.33MnO3 interface,” J. Appl. Phys., vol. 100, no. 3, p. 033704, 2006, doi:
10.1063/1.2222069.
[23] R. Waser, R. Dittmann, C. Staikov, and K. Szot, “Redox-based resistive switching memories
nanoionic mechanisms, prospects, and challenges,” Adv. Mater., vol. 21, no. 25–26, pp. 2632–
2663, 2009, doi: 10.1002/adma.200900375.
[24] D. Ielmini, R. Bruchhaus, and R. Waser, “Thermochemical resistive switching: Materials,

mechanisms, and scaling projections,” Phase Transitions, vol. 84, no. 7, pp. 570–602, 2011, doi:
154
References
10.1080/01411594.2011.561478.
[25] K. Szot, W. Speier, G. Bihlmayer, and R. Waser, “Switching the electrical resistance of
individual dislocations in single-crystalline SrTiO3,” Nat. Mater., vol. 5, no. 4, pp. 312–320,
2006, doi: 10.1038/nmat1614.
[26] M. Janousch, G. I. Meijer, U. Staub, B. Delley, S. E. Karg, and B. P. Andreasson, “Role of

oxygen vacancies in cr-doped SrTiO3 for resistance-change memory,” Adv. Mater., vol. 19, no.
17, pp. 2232–2235, 2007, doi: 10.1002/adma.200602915.
[27] D. H. Kwon et al., “Atomic structure of conducting nanofilaments in TiO2 resistive switching
memory,” Nat. Nanotechnol., vol. 5, no. 2, pp. 148–153, 2010, doi: 10.1038/nnano.2009.456.
[28] S. Bagdzevicius, K. Maas, M. Boudard, and M. Burriel, “Interface-type resistive switching in

perovskite materials,” J. Electroceramics, vol. 39, no. 1–4, pp. 157–184, 2017, doi:
10.1007/s10832-017-0087-9.
[29] E. Dagotto, “Complexity in strongly correlated electronic systems,” Science (80-. )., vol. 309,
no. 5732, pp. 257–262, 2005, doi: 10.1126/science.1107559.
[30] I. Tatsumi, Perovskite Oxide for Solid Oxide Fuel Cells. Springer Science & Business Media,
2009. doi: 10.1007/978-0-387-77708-5.
[31] J. B. Goodenough and J. S. Zhou, “Localized to Itinerant Electronic Transitions in Transition-

Metal Oxides with the Perovskite Structure,” Chem. Mater., vol. 10, no. 10, pp. 2980–2993,
1998, doi: 10.1021/cm980276u.
[32] J. M. D. Coey, M. Viret, and S. Von Molnár, “Mixed-valence manganites,” Adv. Phys., vol. 48,
no. 2, pp. 167–293, 1999, doi: 10.1080/000187399243455.
[33] H.A. Jahn and E. Teller, “Stability of Polyatom ic M olecules in D egenerate E lectronic States
I-O rbital D egeneracy,” Proc. R. Soc. London, vol. 161, pp. 220–235, 1937.
[34] Y. Tokura, “Critical features of colossal magnetoresistive manganites,” Reports Prog. Phys., vol.
69, no. 3, pp. 797–851, 2006, doi: 10.1088/0034-4885/69/3/R06.
[35] P. W. Anderson and H. Hasegawa, “Considerations on double exchange,” Phys. Rev., vol. 100,
no. 2, pp. 675–681, 1955, doi: 10.1103/PhysRev.100.675.
[36] C. Zener, “Interaction between the d-shells in the transition metals. II. Ferromagnetic compounds
of manganese with Perovskite structure,” Phys. Rev., vol. 82, no. 3, pp. 403–405, 1951, doi:
10.1103/PhysRev.82.403.
[37] A. J. Millis, P. B. Littlewood, and B. I. Shraiman, “Double exchange alone does not explain the
155
resistivity of La1-xSrxMnO3,” Phys. Rev. Lett., vol. 74, no. 25, pp. 5144–5147, 1995, doi:
10.1103/PhysRevLett.74.5144.
[38] J. Mizusaki et al., “Electronic conductivity, Seebeck coefficient, defect and electronic structure
of nonstoichiometric La1-xSrxMnO3,” Solid State Ionics, vol. 132, no. 3, pp. 167–180, 2000,
doi: 10.1016/s0167-2738(00)00662-7.
[39] Z. Li, “Crystalline Structure and Electrical Conductivity of Bulk-Sintered and Plasma-Sprayed
La1−xSrxMnO3−δ with 0 ≤ x ≤ 0.9,” ECS Proc. Vol., vol. 1993–4, no. 1, pp. 171–179, 1993,
doi: 10.1149/199304.0171pv.
[40] M. Pajot, V. Duffort, E. Capoen, A. S. Mamede, and R. N. Vannier, “Influence of the strontium
content on the performance of La1-xSrxMnO3/Bi1.5Er0.5O3 composite electrodes for low
temperature Solid Oxide Fuel Cells,” J. Power Sources, vol. 450, no. December 2019, pp. 1–8,
2020, doi: 10.1016/j.jpowsour.2019.227649.
[41] H. K. Lee, “Electrochemical characteristics of La 1 − x Sr x MnO 3 for solid oxide fuel cell,”
Mater. Chem. Phys., vol. 77, no. July 2001, pp. 639–646, 2002.
[42] A. Hammouche, E. J. L. Schouler, and M. Henault, “Electrical and thermal properties of Sr-
doped lanthanum manganites,” Solid State Ionics, vol. 28–30, no. PART 2, pp. 1205–1207, 1988,
doi: 10.1016/0167-2738(88)90358-X.
[43] and O. Y. Y. Takeda, R. Kanno, M. Noda, Y. Tomida, “Cathodic Polarization Phenomena of

Perovskite Oxide Electrodes,” J. Electrochem. Soc. Electrochem. Sci. Technol., vol. Vol. 134,
pp. 2656–2661, 1987.
[44] H. L. Ju et al., “Dependence of giant magnetoresistance on oxygen stoichiometry and

magnetization in polycrystalline La0.67Ba0.33MnOz,” Phys. Rev. B, vol. 51, no. 9, pp. 6143–
6146, 1995, [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevB.51.6143fFn
[45] S. P. Jiang, “Development of lanthanum strontium manganite perovskite cathode materials of

solid oxide fuel cells: A review,” J. Mater. Sci., vol. 43, no. 21, pp. 6799–6833, 2008, doi:
10.1007/s10853-008-2966-6.
[46] A. N. Grundy, B. Hallstedt, and L. J. Gauckler, “Assessment of the La – Sr – Mn – O system,”

vol. 28, pp. 191–201, 2004, doi: 10.1016/j.calphad.2004.07.001.
[47] J. Mizusaki et al., “Oxygen nonstoichiometry and defect equilibrium in the perovskite-type
oxides La1-xSrxMnO3+d,” Solid State Ionics, vol. 129, no. 1, pp. 163–177, 2000, doi:
10.1016/S0167-2738(99)00323-9.
156
References
[48] N. Lee, Y. Lansac, H. Hwang, and Y. H. Jang, “Switching mechanism of Al/La1-xSrxMnO3

resistance random access memory. I. Oxygen vacancy formation in perovskites,” RSC Adv., vol.
5, no. 124, pp. 102772–102779, 2015, doi: 10.1039/c5ra21982e.
[49] J. Hemberger et al., “Structural, magnetic, and electrical properties of single-crystalline La1-
xSrxMnO3 (0.4<x<0.85),” Phys. Rev. B - Condens. Matter Mater. Phys., vol. 66, no. 9, pp. 1–8,
2002, doi: 10.1103/PhysRevB.66.094410.
[50] S. M. Dunaevskiǐ, “Magnetic Phase Diagrams of Manganites in the Electron Doping Region,”
Phys. Solid State, vol. 46, no. 2, pp. 193–212, 2004, doi: 10.1134/1.1649411.
[51] G. Tang, Y. Yu, W. Chen, and Y. Cao, “The electrical resistivity and thermal infrared properties
of La1-xSrxMnO3 compounds,” J. Alloys Compd., vol. 461, no. 1–2, pp. 486–489, 2008, doi:
10.1016/j.jallcom.2007.07.021.
[52] J. R. Sun, H. W. Yeung, H. K. Wong, T. Zhu, and B. G. Shen, “Effects of vacuum annealing on
the transport property of la 0.67 Sr0.33 MnO3-δfilms,” Eur. Phys. J. B, vol. 35, no. 4, pp. 481–
491, 2003, doi: 10.1140/epjb/e2003-00301-6.
[53] F. Chiabrera et al., “Engineering Transport in Manganites by Tuning Local Nonstoichiometry in

Grain Boundaries,” Adv. Mater., vol. 31, no. 4, p. 1805360, 2018, doi: 10.1002/adma.201805360.
[54] F. Chiabrera et al., “Unraveling bulk and grain boundary electrical properties in La 0.8 Sr 0.2
Mn 1-y O 3± δ thin films,” APL Mater., vol. 7, no. 1, pp. 0–8, 2019, doi: 10.1063/1.5054576.
[55] J. M. Börgers and R. A. De Souza, “The surprisingly high activation barrier for oxygen-vacancy
migration in oxygen-excess manganite perovskites,” Phys. Chem. Chem. Phys., vol. 22, no. 25,
pp. 14329–14339, 2020, doi: 10.1039/d0cp01281e.
[56] R. Gross et al., “Physics of grain boundaries in the colossal magnetoresistance manganites,” J.
Magn. Magn. Mater., vol. 211, no. 1, pp. 150–159, 2000, doi: 10.1016/S0304-8853(99)00727-
1.
[57] R. A. De Souza, J. A. Kilner, and J. F. Walker, “A SIMS study of oxygen tracer diffusion and
surface exchange in La 0.8 Sr 0.2 MnO,” Mater. Lett., no. March, pp. 43–52, 2000.
[58] C. Adamo et al., “Effect of biaxial strain on the electrical and magnetic properties of (001) La0.7
Sr0.3 MnO3 thin films,” Appl. Phys. Lett., vol. 95, no. 11, pp. 1–4, 2009, doi:
10.1063/1.3213346.
[59] S. K. Chaluvadi et al., “Epitaxial strain and thickness dependent structural, electrical and
magnetic properties of La0.67Sr0.33MnO3films,” J. Phys. D. Appl. Phys., vol. 53, no. 37, 2020,
157
doi: 10.1088/1361-6463/ab8e7b.
[60] M. Gaudon, C. Laberty-Robert, F. Ansart, P. Stevens, and A. Rousset, “Preparation and

characterization of La1-xSrxMnO3+δ (0 ≤ x ≤ 0.6) powder by sol-gel processing,” Solid State
Sci., vol. 4, no. 1, pp. 125–133, 2002, doi: 10.1016/S1293-2558(01)01208-0.
[61] T. M. Huber, M. Kubicek, A. K. Opitz, and J. Fleig, “ The Relevance of Different Oxygen
Reduction Pathways of La 0.8 Sr 0.2 MnO 3 (LSM) Thin Film Model Electrodes ,” J.
Electrochem. Soc., vol. 162, no. 3, pp. F229–F242, 2015, doi: 10.1149/2.0061503jes.
[62] A. M. Saranya et al., “Engineering mixed ionic electronic conduction in La0.8Sr0.2MnO3+δ

nanostructures through fast grain boundary oxygen diffusivity,” Adv. Energy Mater., vol. 5, no.
11, p. 1500377, 2015, doi: 10.1002/aenm.201500377.
[63] J. M. Börgers et al., “Faster Diffusion of Oxygen Along Dislocations in (La,Sr)MnO3+δ Is a

Space-Charge Phenomenon,” Adv. Funct. Mater., vol. 31, no. 51, pp. 2105647–2105647, 2021,
doi: 10.1002/adfm.202105647.
[64] S. Carter, A. Selcuk, R. J. Chater, J. Kajda, J. A. Kilner, and B. C. H. Steele, “Oxygen transport
in selected nonstoichiometric perovskite-structure oxides,” Solid State Ionics, vol. 53–56, no.
PART 1, pp. 597–605, 1992, doi: 10.1016/0167-2738(92)90435-R.
[65] R. A. De Souza and J. A. Kilner, “Oxygen transport in La1-xSrxMn1-yCoyO 3±δ perovskites

Part I. Oxygen tracer diffusion,” Solid State Ionics, vol. 106, no. 3–4, pp. 175–187, 1998, doi:
10.1016/s0167-2738(97)00499-2.
[66] A. M. Saranya et al., “Unveiling the Outstanding Oxygen Mass Transport Properties of Mn-Rich
Perovskites in Grain Boundary-Dominated La0.8Sr0.2(Mn1- xCox)0.85O3±δ Nanostructures,”
Chem. Mater., vol. 30, no. 16, pp. 5621–5629, 2018, doi: 10.1021/acs.chemmater.8b01771.
[67] R. Rodriguez-Lamas et al., “Epitaxial LaMnO3 films with remarkably fast oxygen transport
properties at low temperature,” J. Mater. Chem. A, vol. 9, no. 21, pp. 12721–12733, 2021, doi:
10.1039/d0ta12253j.
[68] S. Fearn, J. C. H. Rossiny, J. A. Kilner, and J. R. G. Evans, “Measurement of oxygen transport

in La 0.8Sr 0.2MnO 3 perovskite grains,” Solid State Ionics, vol. 211, pp. 51–57, 2012, doi:
10.1016/j.ssi.2012.01.029.
[69] A. Stangl et al., “Tailored nano-columnar La 2 NiO 4 cathodes for improved electrode
performance,” J. Mater. Chem. A, vol. 10, no. 5, pp. 2528–2540, 2022, doi: 10.1039/d1ta09110g.
[70] D. Stender, S. Heiroth, T. Lippert, and A. Wokaun, “A comparison between micro-Raman
158
References
spectroscopy and SIMS of beveled surfaces for isotope depth profiling,” Solid State Ionics, vol.
253, pp. 185–188, 2013, doi: 10.1016/j.ssi.2013.09.053.
[71] D. Pla et al., “In-situ Raman Spectroscopy: a tool for oxygen exchange properties in oxide
conductors,” 2019. [Online]. Available: http://hdl.handle.net/20.500.12210/13671
[72] M. Burriel et al., “In Situ Raman Spectroscopy – A novel approach to study oxygen defect
dynamics in LaMnO3 perovskites,” 2017. [Online]. Available:
http://hdl.handle.net/20.500.12210/13251
[73] E. Navickas et al., “Fast oxygen exchange and diffusion kinetics of grain boundaries in Sr-doped
LaMnO3 thin films,” Phys. Chem. Chem. Phys., vol. 17, no. 12, pp. 7659–7669, 2015, doi:
10.1039/c4cp05421k.
[74] M. S. Islam, M. Cherry, and C. R. A. Catlow, “Oxygen Diffusion in LaMnO 3 and LaCoO 3
Perovskite-Type Oxides :,” J. Solid State Chem., vol. 237, pp. 230–237, 1996.
[75] M. Cherry, M. S. Islam, and C. R. A. Catlow, “Oxygen Ion Migration in Perovskite-Type

Oxides,” J. Solid State Chem., vol. 118, pp. 125–132, 1995, doi: 10.1006/jssc.1995.1320.
[76] O. Tripathy and P. P. Kumar, “Molecular dynamics investigation of oxide ion transport in Sr-
doped LaMnO3,” J. Mater. Sci., vol. 52, no. 11, pp. 6542–6553, 2017, doi: 10.1007/s10853-017-
0889-9.
[77] J. M. Polfus, B. Yildiz, and H. L. Tuller, “Origin of fast oxide ion diffusion along grain
boundaries in Sr-doped LaMnO3,” Phys. Chem. Chem. Phys., vol. 20, no. 28, pp. 19142–19150,
2018, doi: 10.1039/c8cp02443j.
[78] S. Roy and P. P. Kumar, “Influence of Cationic ordering on ion transport in NASICONs:
Molecular dynamics study,” Solid State Ionics, vol. 253, pp. 217–222, 2013, doi:
10.1016/j.ssi.2013.09.030.
[79] D. Parfitt, A. Chroneos, A. Tarancón, and J. A. Kilner, “Oxygen ion diffusion in cation
ordered/disordered GdBaCo2O 5+δ,” J. Mater. Chem., vol. 21, no. 7, pp. 2183–2186, 2011, doi:
10.1039/c0jm02924f.
[80] A. A. Taskin, A. N. Lavrov, and Y. Ando, “Achieving fast oxygen diffusion in perovskites by
cation ordering,” Appl. Phys. Lett., vol. 86, no. 9, pp. 1–3, 2005, doi: 10.1063/1.1864244.
[81] A. A. Taskin, A. N. Lavrov, and Y. Ando, “Fast oxygen diffusion in A-site ordered perovskites,”
Prog. Solid State Chem., vol. 35, no. 2-4 SPEC. ISS., pp. 481–490, 2007, doi:
10.1016/j.progsolidstchem.2007.01.014.
159
[82] T. Norby, “Fast oxygen ion conductors - From doped to ordered systems,” J. Mater. Chem., vol.
11, no. 1, pp. 11–18, 2001, doi: 10.1039/b003463k.
[83] R. Rodriguez-Lamas et al., “Integration of LaMnO 3+δ films on platinized silicon substrates for
resistive switching applications by PI-MOCVD,” Beilstein J. Nanotechnol., vol. 10, pp. 389–
398, 2019, doi: 10.3762/bjnano.10.38.
[84] J. C. Gonzalez-Rosillo et al., “Volume Resistive Switching in metallic perovskite oxides driven
by the Metal-Insulator Transition,” J. Electroceramics, vol. 39, no. 1–4, pp. 185–196, 2017, doi:
10.1007/s10832-017-0101-2.
[85] Z. T. Xu et al., “Evidence for a crucial role played by oxygen vacancies in LaMnO 3 resistive
switching memories,” Small, vol. 8, no. 8, pp. 1279–1284, 2012, doi: 10.1002/smll.201101796.
[86] Y. L. Jin, Z. T. Xu, K. J. Jin, C. Ge, H. Bin Lu, and G. Z. Yang, “Mechanism study on oxygen
vacancy induced resistance switching in Au/LaMnO3/SrNb0.01Ti0.99O3,” Mod. Phys. Lett. B,
vol. 27, no. 11, pp. 8–10, 2013, doi: 10.1142/S0217984913500747.
[87] B. Meunier et al., “Resistive switching in a LaMnO3 + δ/TiN memory cell investigated by
operando hard X-ray photoelectron spectroscopy,” J. Appl. Phys., vol. 126, no. 22, 2019, doi:
10.1063/1.5125420.
[88] B. Meunier et al., “ Microscopic Mechanisms of Local Interfacial Resistive Switching in LaMnO
3+δ ,” ACS Appl. Electron. Mater., vol. 1, no. 5, pp. 675–683, 2019, doi:
10.1021/acsaelm.9b00030.
[89] X. Chen, N. J. Wu, J. Strozier, and A. Ignatiev, “Spatially extended nature of resistive switching
in perovskite oxide thin films,” Appl. Phys. Lett., vol. 89, no. 6, pp. 10–13, 2006, doi:
10.1063/1.2236213.
[90] R. Ortega-Hernandez et al., “Resistive switching in CeO2/La0.8Sr0.2MnO3 bilayer for non-

volatile memory applications,” Microelectron. Eng., vol. 147, pp. 37–40, 2015, doi:
10.1016/j.mee.2015.04.042.
[91] J. C. Gonzalez-Rosillo, “Volume Resistive Switching in metallic perovskite oxides driven by the
Metal-Insulator Transition,” 2017. doi: 10.1007/s10832-017-0101-2.
[92] J. C. Gonzalez-Rosillo et al., “SI. Engineering Oxygen Migration for Homogeneous Volume
Resistive Switching in 3-Terminal Devices,” Adv. Electron. Mater., vol. 5, no. 9, 2019, doi:
10.1002/aelm.201800629.
[93] J. C. Gonzalez-Rosillo et al., “Nanoscale Correlations between Metal–Insulator Transition and
160
References
Resistive Switching Effect in Metallic Perovskite Oxides,” Small, vol. 16, no. 23, pp. 1–10, 2020,
doi: 10.1002/smll.202001307.
[94] J. C. Gonzalez-Rosillo et al., “Engineering Oxygen Migration for Homogeneous Volume

Resistive Switching in 3-Terminal Devices,” Adv. Electron. Mater., vol. 5, no. 9, pp. 1–8, 2019,
doi: 10.1002/aelm.201800629.
[95] X. D. Luo et al., “Resistance switching mechanism of La0.8Sr0.2MnO3-δ thin films,” Phys. B
Condens. Matter, vol. 483, pp. 99–102, 2016, doi: 10.1016/j.physb.2015.12.031.
[96] H. K. Lau and C. W. Leung, “Nonvolatile multilevel memory effect by resistive switching in
manganite thin films,” J. Appl. Phys., vol. 104, no. 12, 2008, doi: 10.1063/1.3043801.
[97] C. Moreno, C. Munuera, S. Valencia, F. Kronast, X. Obradors, and C. Ocal, “Reversible resistive
switching and multilevel recording in La 0.7Sr0.3MnO3 thin films for low cost nonvolatile
memories,” Nano Lett., vol. 10, no. 10, pp. 3828–3835, 2010, doi: 10.1021/nl1008162.
[98] C. Kalkert et al., “Resistive switching at manganite/manganite interfaces,” Appl. Phys. Lett., vol.
99, no. 13, pp. 10–13, 2011, doi: 10.1063/1.3643425.
[99] M. Coll, A. Palau, J. C. Gonzalez-Rosillo, J. Gazquez, X. Obradors, and T. Puig, “Integration of

atomic layer deposition CeO2 thin films with functional complex oxides and 3D patterns,” Thin
Solid Films, vol. 553, pp. 7–12, 2014, doi: 10.1016/j.tsf.2013.08.131.
[100] L. Yao, S. Inkinen, and S. Van Dijken, “Direct observation of oxygen vacancy-driven structural
and resistive phase transitions in La2/3Sr1/3MnO3,” Nat. Commun., vol. 8, pp. 1–9, 2017, doi:
10.1038/ncomms14544.
[101] L. Yao, S. Inkinen, H. P. Komsa, and S. van Dijken, “Structural Phase Transitions to 2D and 3D
Oxygen Vacancy Patterns in a Perovskite Film Induced by Electrical and Mechanical
Nanoprobing,” Small, vol. 17, no. 10, pp. 1–9, 2021, doi: 10.1002/smll.202006273.
[102] F. Wu, X. Li, W. Yu, Y. Zhang, and X. Cao, “Correlation between crystalline qualities and
resistive switching effects of La 0.7 Sr 0.3 MnO 3 films,” Sixth Int. Conf. Thin Film Phys. Appl.,
vol. 6984, no. March 2008, p. 69841J, 2008, doi: 10.1117/12.792624.
[103] H. S. Lee et al., “A study on the resistive switching of La0.7Sr0.3MnO3 film using
spectromicroscopy,” Appl. Mech. Mater., vol. 597, pp. 184–187, 2014, doi:
10.4028/www.scientific.net/AMM.597.184.
[104] S. G. Choi, H. S. Lee, H. Choi, S. W. Chung, and H. H. Park, “The effect of Sr concentration on
resistive switching properties of La 1-xSrxMnO3 films,” Thin Solid Films, vol. 529, pp. 352–
161
355, 2013, doi: 10.1016/j.tsf.2012.07.069.
[105] R. Waser, “Redox-based resistive switching memories,” J. Nanosci. Nanotechnol., vol. 12, no.
10, pp. 7628–7640, 2012, doi: 10.1166/jnn.2012.6652.
[106] G. N. Derry and Z. Ji-zhong, “Work function of Pt ( 111 ),” Phys. Rev. B, vol. 39, no. 3, pp.
1940–1941, 1989, [Online]. Available:
https://journals.aps.org/prb/pdf/10.1103/PhysRevB.39.1940
[107] F. Borgatti et al., “Chemical insight into electroforming of resistive switching manganite
heterostructures,” Nanoscale, vol. 5, no. 9, pp. 3954–3960, 2013, doi: 10.1039/c3nr00106g.
[108] K. Maas, “La2NiO4, a Mixed Ionic-Electronic Conductor for Interface-Type Valence Change
Memories,” Université Grenoble Alpes (UGA), 2018.
[109] A. Sawa, T. Fujii, M. Kawasaki, and Y. Tokura, “Hysteretic current-voltage characteristics and
resistance switching at a rectifying Ti/Pr0.7Ca0.3MnO3 interface,” Appl. Phys. Lett., vol. 85, no.
18, pp. 4073–4075, 2004, doi: 10.1063/1.1812580.
[110] S. Asanuma, H. Akoh, H. Yamada, and A. Sawa, “Relationship between resistive switching
characteristics and band diagrams of Ti/Pr1-x CaxMnO3 junctions,” Phys. Rev. B - Condens.
Matter Mater. Phys., vol. 80, no. 23, pp. 1–8, 2009, doi: 10.1103/PhysRevB.80.235113.
[111] J. Sénateur, C. Dubourdieu, F. Weiss, M. Rosina, and A. Abrutis, “Pulsed injection MOCVD of
functional electronic oxides,” Adv. Mater. Opt. Electron., vol. 10, no. 35, pp. 155–161, 2000,
doi: 10.1002/1099-0712(200005/10)10:3/5<155::aid-amo406>3.3.co;2-u.
[112] J.-P. Senateur, R. Madar, F. Weiss, O. Thomas, and A. Adulfas, “Method and device for feeding
precursors into a chemical vapour deposition chamber,” EP0730671B1, 1993
[113] J. P. Sénateur et al., “Synthesis and characterisation of YBCO thin films grown by injection-
MOCVD,” J. Alloys Compd., vol. 251, no. 1–2, pp. 288–291, 1997, doi: 10.1016/S0925-
8388(96)02699-0.
[114] A. M. Haghiri-Gosnet and J. P. Renard, “CMR manganites: Physics, thin films and devices,” J.
Phys. D. Appl. Phys., vol. 36, no. 8, 2003, doi: 10.1088/0022-3727/36/8/201.
[115] M. Vagner et al., “PI-MOCVD technology of (La, Sr)(Mn, Co)O3: From epitaxial to
nanostructured films,” Surf. Coatings Technol., vol. 385, no. December 2019, p. 125287, 2020,
doi: 10.1016/j.surfcoat.2019.125287.
[116] A. Devi, “‘Old chemistries’ for new applications: Perspectives for development of precursors for
MOCVD and ALD applications,” Coord. Chem. Rev., vol. 257, no. 23–24, pp. 3332–3384, 2013,
162
References
doi: 10.1016/j.ccr.2013.07.025.
[117] D. K. Schroder, Semiconductor Material and Device Characterization, Third. New Jersey: John
Wiley & Sons, Inc., 2006.
[118] R. Rodríguez-Lamas, “Resistive switching in LaMnO3 + d based heterostructures for valence

change memories (VCMs),” 2020.
[119] J. A. Bearden, “X-Ray Wavelengths,” Rev. Mod. Phys., vol. 39, no. 1, pp. 78–124, 1967.
[120] J. Töpfer and J. B. Goodenough, “LaMnO3+δRevisited,” J. Solid State Chem., vol. 130, no. 1,
pp. 117–128, 1997, doi: 10.1006/jssc.1997.7287.
[121] R. Bindu, “Structural studies of La1-xSrxMnO3+δ (x = 0.1-1.0),” Eur. Phys. J. B, vol. 37, no. 3,
pp. 321–327, 2004, doi: 10.1140/epjb/e2004-00062-8.
[122] R. D. Shannon, “Revised Effective Ionic Radii and Systematic Studies of Interatomie Distances
in Halides and Chaleogenides Central Research and Development Department , Experimental
Station , E . L Du Pont de Nemours The effective ionic radii of Shannon & Prewitt [ Acta,” Acta
Crystallogr. Sect. A, vol. 32, pp. 751–767, 1976, doi: 10.1107/S0567739476001551.
[123] T. Grande, J. R. Tolchard, and S. M. Selbach, “Anisotropic thermal and chemical expansion in
Sr-substituted LaMnO 3+δ: Implications for chemical strain relaxation,” Chem. Mater., vol. 24,
no. 2, pp. 338–345, 2012, doi: 10.1021/cm2030608.
[124] K. . Gnanasekar et al., “Nanocrystalline bulk and thin films of La1−xSrxMnO3 (0≤x≤0.3),” Solid
State Ionics, vol. 148, pp. 575–581, 2002.
[125] J. Mizusaki, “Relationship of Crystal Structure, Lattice Parameters and Thermal Expansion to
Nonstoichiometry in La1−xMnO3+d and La1−xSrxMnO3+d Determined by High-Temperature
X-Ray Diffraction Under Controlled Oxygen Partial Pressures,” ECS Proc. Vol., vol. 1995–1,
no. 1, pp. 444–453, 1995, doi: 10.1149/199501.0444pv.
[126] M. N. Iliev and M. V. Abrashev, “Raman phonons and Raman Jahn-Teller bands in perovskite-
like manganites,” Journal of Raman Spectroscopy, vol. 32, no. 10. pp. 805–811, 2001. doi:
10.1002/jrs.770.
[127] L. Martín-Carrón, A. De Andrés, M. J. Martínez-Lope, M. T. Casais, and J. A. Alonso, “Raman

phonons as a probe of disorder, fluctuations, and local structure in doped and undoped
orthorhombic and rhombohedral manganites,” Phys. Rev. B - Condens. Matter Mater. Phys., vol.
66, no. 17, pp. 1–8, 2002, doi: 10.1103/PhysRevB.66.174303.
[128] N. Van Minh, “Raman scattering study of perovskite manganites,” J. Phys. Conf. Ser., vol. 187,
163
2009, doi: 10.1088/1742-6596/187/1/012011.
[129] A. Dubroka, J. Humlíček, M. V. Abrashev, Z. V. Popović, F. Sapiña, and A. Cantarero, “Raman

and infrared studies of La1-y Sry Mn1-x Mx O3 (M=Cr, Co, Cu, Zn, Sc or Ga): Oxygen disorder
and local vibrational modes,” Phys. Rev. B - Condens. Matter Mater. Phys., vol. 73, no. 22, pp.
1–10, 2006, doi: 10.1103/PhysRevB.73.224401.
[130] V. B. Podobedov, A. Weber, D. B. Romero, J. P. Rice, and H. D. Drew, “Effect of structural and
magnetic transitions in La1-xMxMnO3 (M=Sr, Ca) single crystals in Raman scattering,” Phys.
Rev. B, vol. 58, no. 1, pp. 43–46, 1998, doi: 10.1103/PhysRevB.58.43.
[131] M. N. Iliev, M. V. Abrashev, V. N. Popov, and V. G. Hadjiev, “Role of Jahn-Teller disorder in

Raman scattering of mixed-valence manganites,” Phys. Rev. B - Condens. Matter Mater. Phys.,
vol. 67, no. 21, pp. 3–6, 2003, doi: 10.1103/PhysRevB.67.212301.
[132] B. Güttler, V. A. Amelitchev, O. Y. Gorbenko, and A. R. Kaul, “Static and dynamic Jahn-Teller
distortions in CMR manganites: A Raman spectrometric study,” Phase Transitions, vol. 76, no.
1–2, pp. 63–72, 2003, doi: 10.1080/0141159031000076057.
[133] V. B. Podobedov et al., “Diagnostics of ‘colossal’ magnetoresistance manganite films by Raman

spectroscopy,” Appl. Phys. Lett., vol. 73, no. 22, pp. 3217–3219, 1998, doi: 10.1063/1.122723.
[134] Y. Shen, C. Wang, L. Li, Q. Shen, and L. Zhang, “Effect of Sr doping on structure and emittance
of La1-xSr xMnO3 smart thermal control films,” Key Eng. Mater., vol. 602–603, pp. 902–905,
2014, doi: 10.4028/www.scientific.net/KEM.602-603.902.
[135] E. Navickas et al., “Dislocations Accelerate Oxygen Ion Diffusion in La0.8Sr0.2MnO3 Epitaxial
Thin Films,” ACS Nano, vol. 11, no. 11, pp. 11475–11487, 2017, doi: 10.1021/acsnano.7b06228.
[136] B. Koo, K. Kim, J. K. Kim, H. Kwon, J. W. Han, and W. C. Jung, “Sr Segregation in Perovskite
Oxides: Why It Happens and How It Exists,” Joule, vol. 2, no. 8, pp. 1476–1499, 2018, doi:
10.1016/j.joule.2018.07.016.
[137] H. Jalili, J. W. Han, Y. Kuru, Z. Cai, and B. Yildiz, “New insights into the strain coupling to
surface chemistry, electronic structure, and reactivity of La0.7Sr0.3MnO3,” J. Phys. Chem. Lett.,
vol. 2, no. 7, pp. 801–807, 2011, doi: 10.1021/jz200160b.
[138] H.-S. Lee, C.-S. Park, and H.-H. Park, “Effect of La 3+ substitution with Gd 3+ on the resistive
switching properties of La 0.7 Sr 0.3 MnO 3 thin films,” Appl. Phys. Lett., vol. 104, no. 19, p.
191604, 2014, doi: 10.1063/1.4876115.
[139] S. Bagdzevicius et al., “Superposition of interface and volume type resistive switching in
164
References
perovskite nanoionic devices,” J. Mater. Chem. C, vol. 7, no. 25, pp. 7580–7592, 2019, doi:
10.1039/c9tc00609e.
[140] R. Waser et al., Introduction to Nanoionic Elements for Information Technology. 2016. doi:
10.1002/9783527680870.ch1.
[141] I. V Markov, Crystal Growth for Beginners. WORLD SCIENTIFIC, 2003. doi:
doi:10.1142/5172.
[142] Y. Baikov, E. I. Nikulin, B. T. Melekh, and V. M. Egorov, “Conductivity, Magnetoresistance ,

and Specific Heat of Oxygen-deficient La 0.67 Sr 0.33 MnO 3 - a (0 ≤ a ≤ 0.16),” Phys. Solid
State, vol. 46, no. 11, pp. 2086–2093, 2004.
[143] T. M. Huber et al., “Interplay of Grain Size Dependent Electronic and Ionic Conductivity in
Electrochemical Polarization Studies on Sr-Doped LaMnO 3 (LSM) Thin Film Cathodes,” J.
Electrochem. Soc., vol. 165, no. 9, pp. F702–F709, 2018, doi: 10.1149/2.1081809jes.
[144] A. F. Zurhelle, W. Stehling, R. Waser, R. A. De Souza, and S. Menzel, “Oxygen Diffusion in

Platinum Electrodes : A Molecular Dynamics Study of the Role of Extended Defects,” vol.
2101257, 2021, doi: 10.1002/admi.202101257.
[145] B. Gharbage, F. Mandier, H. Lauret, C. Roux, and T. Pagnier, “Electrical properties of

La0.5Sr0.5MnO3 thin films,” Solid State Ionics, vol. 82, no. 1–2, pp. 85–94, 1995, doi:
10.1016/0167-2738(95)00206-L.
[146] O. Chmaissem, B. Dabrowski, S. Kolesnik, J. Mais, L. Suescun, and J. D. Jorgensen, “Effects of

internal structural parameters on the properties of Ba-substituted La0.5 Sr0.5 Mn O3,” Phys. Rev.
B - Condens. Matter Mater. Phys., vol. 74, no. 14, pp. 1–10, 2006, doi:
10.1103/PhysRevB.74.144415.
[147] H. C. Pan, C. C. Chou, and H. L. Tsai, “Low-temperature processing of sol-gel derived La0.5Sr
0.5MnO3 buffer electrode and PbZr0.52Ti 0.48O3 films using CO2 laser annealing,” Appl. Phys.
Lett., vol. 83, no. 15, pp. 3156–3158, 2003, doi: 10.1063/1.1619214.
[148] L. Cao et al., “Reversible Control of Physical Properties via an Oxygen-Vacancy-Driven

Topotactic Transition in Epitaxial La 0.7 Sr 0.3 MnO 3−δ Thin Films,” Adv. Mater., vol. 31, no.
7, pp. 1–7, 2019, doi: 10.1002/adma.201806183.
[149] J. H. Kuo, H. U. Anderson, and D. M. Sparlin, “Oxidation-reduction behavior of undoped and

Sr-doped LaMnO3 nonstoichiometry and defect structure,” J. Solid State Chem., vol. 83, no. 1,
pp. 52–60, 1989, doi: 10.1016/0022-4596(89)90053-4.
165
[150] J. H. Kuo, H. U. Anderson, and D. M. Sparlin, “Oxidation-reduction behavior of undoped and

Sr-doped LaMnO3: Defect structure, electrical conductivity, and thermoelectric power,” J. Solid
State Chem., vol. 87, no. 1, pp. 55–63, 1990, doi: 10.1016/0022-4596(90)90064-5.
[151] D. Rasic, R. Sachan, J. Prater, and J. Narayan, “Structure-property correlations in thermally

processed epitaxial LSMO films,” Acta Mater., vol. 163, pp. 189–198, 2019, doi:
10.1016/j.actamat.2018.10.023.
[152] S. Miyoshi et al., “In situ XRD study on oxygen-excess LaMnO 3,” vol. 175, pp. 383–386, 2004,
doi: 10.1016/j.ssi.2004.01.082.
[153] N. Abdelmoula, A. Cheikhrouhou, K. Guidara, and J. C. Joubert, “Oxygen Deficiency Effects

upon Structural , Magnetic and Electrical Properties of Doped Manganite Oxides,” vol. 533, pp.
533–540, 2000.
[154] C. D. Garner, “X-ray absorption spectroscopy,” Nature, vol. 277, no. 5692, pp. 89–90, 1979, doi:
10.1038/277089a0.
[155] J. Yano and V. K. Yachandra, “X-ray absorption spectroscopy,” Photosynth. Res., vol. 102, no.
2, pp. 241–254, 2009, doi: 10.1007/s11120-009-9473-8.
[156] P. Horodek, L. H. Khiem, K. Siemek, L. A. Tuyen, and A. G. Kobets, “Positron Annihilation

Spectroscopy in Material studies,” Commun. Phys., vol. 29, no. 4, p. 501, 2019, doi:
10.15625/0868-3166/29/4/14282.
[157] R. Krause-Rehberg and H. Leipner, Positron annihilation in semiconductors: defect studies.

Springer Science and Business Media, 1999.
[158] F. Tuomisto and I. Makkonen, “Defect identification in semiconductors with positron

annihilation: Experiment and theory,” Rev. Mod. Phys., vol. 85, no. 4, pp. 1583–1631, 2013, doi:
10.1103/RevModPhys.85.1583.
[159] E. H. Khan, M. H. Weber, and M. D. McCluskey, “Formation of isolated Zn vacancies in ZnO

single crystals by absorption of ultraviolet radiation: A combined study using positron
annihilation, photoluminescence, and mass spectroscopy,” Phys. Rev. Lett., vol. 111, no. 1, pp.
1–5, 2013, doi: 10.1103/PhysRevLett.111.017401.
[160] G. Subías, J. García, M. Proietti, and J. Blasco, “X-ray-absorption near-edge spectroscopy and
circular magnetic x-ray dichroism at the Mn edge of magnetoresistive manganites,” Phys. Rev.
B - Condens. Matter Mater. Phys., vol. 56, no. 13, pp. 8183–8191, 1997, doi:
10.1103/PhysRevB.56.8183.
166
References
[161] V. Cuartero et al., “X-ray absorption and emission spectroscopy study of Mn and Co valence
and spin states in TbM n 1-x C ox O 3,” Phys. Rev. B, vol. 94, no. 15, pp. 1–10, 2016, doi:
10.1103/PhysRevB.94.155117.
[162] V. R. Mastelaro, D. P. F. De Souza, and R. A. Mesquita, “X-ray absorption spectroscopic studies

of Mn atoms in La1-xSrxMnO3+δ compounds,” X-Ray Spectrom., vol. 31, no. 2, pp. 154–157,
2002, doi: 10.1002/xrs.543.
[163] B. Gilbert et al., “Multiple scattering calculations of bonding and X-ray absorption spectroscopy
of manganese oxides,” J. Phys. Chem. A, vol. 107, no. 16, pp. 2839–2847, 2003, doi:
10.1021/jp021493s.
[164] Mayer Matej (Max-Planck-Institut für Plasmaphysik), “SIMNRA - Software for simulation of
RBS, ERDA, NRA, MEIS and PIGE.” https://home.mpcdf.mpg.de/~mam/
[165] G. S. Henderson, F. M. F. de Groot, and B. J. A. Moulton, “X-ray absorption near-edge structure

(XANES) spectroscopy,” in Spectroscopic Methods in Mineralogy and Materials Sciences, vol.
78, no. 1, 2014, pp. 75–138. doi: 10.2138/rmg.2014.78.3.
[166] “XRD DIFFRAC.SUITE EVA by Bruker.” https://www.bruker.com/en/products-and-

solutions/diffractometers-and-scattering-systems/x-ray-diffractometers/diffrac-suite-
software/diffrac-eva.html (accessed Apr. 10, 2022).
[167] “XRD DIFFRAC.SUITE TOPAS by Bruker.” https://www.bruker.com/en/products-and-

software/diffrac-topas.html (accessed Apr. 10, 2022).
[168] “Fityk — curve fitting and peak fitting software.” https://fityk.nieto.pl/ (accessed Apr. 10, 2022).
[169] “XRD DIFFRAC.SUITE LEPTOS by Bruker.” https://www.bruker.com/en/products-and-

software/diffrac-eva.html (accessed Apr. 10, 2022).
[170] L. J. Van Der PAUW, “A method of measuring the resistivity and Hall coefficient on lamellae
of arbitrary shape,” Philips Technical Review, vol. 20, no. 8. pp. 220–224, 1958. [Online].
Available: http://www.mendeley.com/research/method-measuring-resistivity-hall-coefficient-
lamellae-arbitrary-shape/
[171] C. Moncasi et al., “Structural Defects Improve the Memristive Characteristics of Epitaxial La
0.8 Sr 0.2 MnO 3 ‐Based Devices,” Adv. Mater. Interfaces, vol. 9, no. 23, p. 2200498, 2022, doi:
10.1002/admi.202200498.
167
[172] E. Dagotto, T. Hotta, and A. Moreo, “Colossal magnetoresistant materials: The key role of phase
separation,” Phys. Rep., vol. 344, no. 1–3, pp. 1–153, 2001, doi: 10.1016/S0370-1573(00)00121-
6.
[173] X. Chen, N. Wu, J. Strozier, and A. Ignatiev, “Spatially extended nature of resistive switching
in perovskite oxide thin films,” Appl. Phys. Lett., vol. 89, no. 6, pp. 2004–2007, 2006, doi:
10.1063/1.2236213.
[174] J. C. Gonzalez‐Rosillo et al., “Engineering Oxygen Migration for Homogeneous Volume

Resistive Switching in 3‐Terminal Devices,” Adv. Electron. Mater., vol. 5, no. 9, p. 1800629,
Sep. 2019, doi: 10.1002/aelm.201800629.
[175] T. Yamamoto, R. Yasuhara, I. Ohkubo, H. Kumigashira, and M. Oshima, “Formation of

transition layers at metalperovskite oxide interfaces showing resistive switching behaviors,” J.
Appl. Phys., vol. 110, no. 5, 2011, doi: 10.1063/1.3631821.
[176] J. J. Yang et al., “Metal/TiO2 interfaces for memristive switches,” Appl. Phys. A Mater. Sci.
Process., vol. 102, no. 4, pp. 785–789, 2011, doi: 10.1007/s00339-011-6265-8.
168
Annex A
Techniques and equipment

A.1. High energy radiation-based characterization techniques .................................................... 169
A.1.1. Rutherford Backscattering Spectrometry and Nuclear Reaction Analysis .................. 169
A.1.2. X-ray Absorption Spectroscopy .................................................................................. 171
A.2. Techniques for thin film characterization ............................................................................... 173
A.2.1. X-ray diffraction .......................................................................................................... 173
A.2.2.Transmission Electron Microscopy .............................................................................. 173
A.2.3. Scanning electron microscopy ..................................................................................... 174
A.2.4. Electron Probe Micro Analysis and Wavelength dispersive spectroscopy .................. 174
A.2.5. Atomic Force Microscopy ........................................................................................... 174
A.2.6. Raman Spectroscopy ................................................................................................... 175
A.2.7. Electrical Characterization of thin films and devices .................................................. 175
This Annex is divided into two sections. The first one details the characterization techniques in high-
energy radiation source facilities and the operating conditions. The second section describes other
characterization techniques and equipment used in this work.
A.1. High energy radiation-based characterization techniques
A.1.1. Rutherford Backscattering Spectrometry and Nuclear Reaction

Analysis
In order to have a larger certainty of the elemental chemical composition, selected films grown on Si
and MgO substrates (Section 3.2) were analyzed by Rutherford Backscattering Spectrometry (RBS) and
Nuclear Reaction Analysis (NRA). The measurements were performed in Centro de Micro-Análisis de
Materiales in Universidad Autónoma de Madrid (Spain) using a 5 MeV tandetron accelerator.
RBS is a powerful technique for the determination of composition and depth profiling which measures
the backscattered high energy ions that have elastically interacted with the target sample. The energy of
the backscattered ions is proportional to the atomic number of the element of the film they have
interacted with. The energy of the backscattered ions decreases across the sample (before and after
interacting with the scattering species) which allows for nanometric depth resolution.
169
Annex: Experimental procedure, techniques and equipment
Figure A.1. RBS-NRA experimental spectra of a LSM20/MgO sample, the simulated profile and the elemental
deconvolution. a) RBS spectrum acquired with a beam incident energy of 2 MeV. NRA-RBS spectra of the sample
acquired with a beam energy of b) 3.035 MeV, and c) 4 MeV, corresponding to the resonant energy of oxygen and
carbon respectively. d) Scheme of the sample used for the simulation where the LSM film was subdivided in three
regions. The normalized cationic composition is given for each sub-layer.
RBS is very sensible to heavy elements which increase the backscattering yield. The cations of the film
were analyzed with a 2 MeV He+ beam and silicon surface barrier detectors place at a scattering angle
of 170°. The sensitivity of light elements can be enhanced by performing RBS at energies of nuclear
reactions. The analysis of O and C was carried out by 4He+ ion beam at the oxygen and carbon resonant
frequency, 3.035 MeV and 4 MeV, respectively. The spectra of each sample were acquired using an
implantation charge of 12µC. The spectra were treated using SIMNRA code[164] where the composition
of each sample was refined until it converged to reproduce the spectra at the three different energies.
Figure A.1 presents the RBS-NRA spectra acquired at a) 2 MeV, b) 3.035 MeV and c) 4 MeV in a
ca. 90 nm thick La0.8Sr0.2MnO3 (LSM20) film grown on MgO substrate, and d) the sample used for the
simulation with the normalized elemental concentration. The experimental spectra correspond to the red
dots while the simulated spectra using the sample presented in Figure A.1d is shown as a blue line. The
deconvolution of each element from the simulated profile is also included in the spectra. The signal from
170
heavy elements appears at high energy and decreases with the atomic number. The width of the peak is
related with the thickness. The intensity of the peaks depends of the concentration and nature of the
element (effective cross-section). In Figure A.1a at 2 MeV, from right to left we find the heavy cations
La, Sr, Mn, which are present on the surface of the sample; followed by the Mg from the substrate for
which the signal extends until low energies (infinite thickness compared to the film). The light elements,
O and C, appear at low energies and superimposed with the signal of the substrate. The signal of O and
C is enhanced in the NRA-RBS spectra presented in Figure A.1b and c, respectively.
A.1.2. X-ray Absorption Spectroscopy
X-ray absorption spectroscopy (XAS) allows to study the oxidation state, coordination chemistry and
atomic distance of an element in a material upon the absorption of X-rays. This absorption causes the
promotion of a core electron to an empty state or to the continuum state. The energy onset where the
binding electrons of an atom are excited to the continuous states is called absorption edge. The
absorption edge depends on the core-level electron excited of the atom (i.e. K edge for the 1s electrons,
L edge for 2s and 2p electrons, M edge for 3s, 3p and 3d electrons) and hence, they are unique for
element. On top of that, the edge position depends on the oxidation state of the element sensed. X-ray
near edge spectroscopy (XANES) studies the absorption spectra at energies around the edge of an
element. Therefore, XANES allows the determination of the oxidation state of an element by comparing
it to the XANES spectra of compounds with known oxidation state. The study of the absorption spectra
at energies larger than the edge is known as extended X-ray absorption fine structure (EXAFS). In
EXAFS, the excited photoelectrons generate waves that interact with the chemical environment
originating interference pattern in the spectra. The detailed analysis of the spectra provide information
about the structural and chemical environment around the absorbing element.
When the X-ray energy is closed to the edge position of an element, more radiation is absorbed. The
absorption can be measured either directly, by measuring the transmitted intensity across the sample, or
indirectly, by measuring the fluorescence signal created when the core-hole state is occupied. For a more
detailed description of the theory, applications, advantages and limitations of XAS, the reader is referred
to references [154], [155], [165].
XANES spectra at the Mn-K edge in LSM thin films were acquired at the ESRF ID12 beamline
(Grenoble, France) to estimate the oxidation state of Mn of the films under different conditions
(described in Section 6.4). XANES was acquired in an energy range of 6521.9 eV to 6622.9 eV. Two
setups and conditions were used.
171
(i) For the annealed samples used as references, the XANES spectra of the whole sample
(10×10 mm2) was collected under vacuum at 298 K using a nearly constant beam current of
196 mA. The intensity detector used was a silicon photodiode and the total fluorescence
yield was collected in backscattered geometry (detector plane at 90° respect to the beam
vector). The Mn K-edge was measured in grazing incidence configuration.
(ii) For the film with memristive devices, the XANES spectra was acquired at atmospheric
pressure, in open air and at 298 K. Two silicon drift detectors were used to collect the partial
fluorescence yield in backscattering configuration. A 10 µm thick Cr foil was used to reduce
elastic scattering. A beam spot size of 5 µm was used for precise XANES mapping, and a
200 µm spot size was sued for low-photon density XANES over large areas. A Keithley
2410 sourcemeter was used to bias the device (voltage applied on Ti contact, and Pt contact
grounded). Figure A.2 shows pictures of a) the setup in place and b) the stage for the
operando measurements.
The edge position was determined by the maximum of the first derivative unless specified otherwise.
The spectra were corrected to metallic Mn reference (Mn88Ni12 alloy). The oxidation state of the LSM
films was estimated comparing the spectra Mn binary oxides measured.
Figure A.2. Setup for operando XANES. a) General view of the set-up in acquisition conditions. b) Picture of the
stage where the sample is electrically biased with two tungsten probe connected to a source measure unit.
172
A.2. Techniques for thin film characterization
A.2.1. X-ray diffraction
X-ray diffraction (XRD) in Bragg Brentano configuration θ-2θ was used to identify phase and
orientation of the LSM films, particularly those grown on single crystal substrates, i.e. STO and LAO.
The equipment used was a Bruker D8 Advance serie II diffractometer using monochromatic Cu Kα
radiation (λ=1.5406 Å) and 1D detector (Bruker Lynxeye) with a resolution of 2θ of 0.06°.
Diffractograms were acquired on a spinning sample typically using the following parameters: 2θ range
20-80°, with a step of 0.011° and 3 s per step.
The offset and displacement in the diffractograms were corrected using the signal of the substrate as
references, in the DIFFRAC.SUITE EVA software from Bruker.[166]
The polycrystalline samples have been measure in a 5-circle RIGAKU Smartlab diffractometer in
grazing incidence XRD (GIXRD) configuration to increase the signal of the films and minimize the
intensity of substrate (silicon based substrates).
The lattice parameters of the polycrystalline films obtained using the fitting DIFFRAC.SUITE TOPAS
software from Bruker.[167] The corrected XRD patterns of the oriented films grown on single crystal
substrates were fitted using Fityk code[168] where the signal from pseudo-cubic (00l) was used to
calculate the out-of-plane lattice parameter.
The thickness of the films was measured by X-ray reflectivity (XRR) in a RIGAKU Smartlab
diffractometer. Data treatment and fitting simulation were carried out using DIFFRACT.SUITE
LEPTOS software from Bruker.[169]
A.2.2.Transmission Electron Microscopy
Transmission electron microscopy (TEM) was used to study the cross-section structure of the films.
Different equipment were used depending on the state of the sample and the complementary
characterization required.
For devices in films grown on single crystals:

Lamellas were prepared by Dr Gauthier Lefèvre using Focused Ion Beam (FIB) milling in a FEI dual-
beam Helios 450S. A thin layer of platinum was sputtered to prevent damage form the ion beam. The
milling consisted of two main step: rough milling at 30 kV and fine milling at 2-8 kV. The observation
of the lamellas was carried out at 200 kV in a Tecnai Osiris FEI microscope equipped with a Super-X
detector for Energy Dispersive-X-ray Spectroscopy (EDS) chemical compositions analysis.
Other films and after device electrical characterization was completed:

173
Lamellas were prepared by Dr Laetitia Rapenne using automated polishing, MultiPrepTM system
(Allied High Tech Products Inc.) The final polishing was performed using a felt-covered disc
impregnated with a silica solution and the thickness was finally reduced by Ar-milling. These lamellas
were observed at 200 kV in a JEOL JEM 2010 LaB6 TEM with a resolution of ∼0.19 nm.
Complementary, STEM EDS (Scanning Transmission Electron Microscopy EDS) was carried out in a
JEOL 2100F FEG microscope at 200 kV with a 0.2 nm resolution in scanning mode.
A.2.3. Scanning electron microscopy
Scanning electron microscopy (SEM) was used to study the surface and cross-section morphology and
compositional contrast of the films by detecting secondary electrons and backscattered electrons,
respectively. The majority of SEM analysis were conducted in a FEG ZEISS GeminiSEM 300 while a
Quanta250 environmental SEM FEG form FEI was used for the samples presented in Chapter 3. Semi-
quantitative chemical composition analysis of the films were done by EDS using an Oxford Inca Energy
detector coupled to the GeminiSEM 300.
In ZEISS GeminiSEM 300, images were acquired in high vacuum at an operating voltage of 2-4 kV
working at a distance of 3 mm. EDS analysis were typically done at 15 kV and 8 mm where a Cu
reference was used as calibration for the semi-quantitative chemical analysis.
In SEM Quanta 250, the images were acquired in high vacuum at 5-10 kV at a working distance of
11 mm.
A.2.4. Electron Probe Micro Analysis and Wavelength dispersive

spectroscopy
The cationic composition of the films grown on Si and MgO substrates presented in Section 3.2 was
analyzed by Electron Probe Micro Analysis (EPMA) using a CAMECA SX50 equipped with
Wavelength dispersive spectroscopy (WDS) spectrometer. Different areas per sample were measured at
energies in the range of 10-16 keV. The data was analyzed using STRATAGem software.
A.2.5. Atomic Force Microscopy
The topography and roughness of the films was measured by Atomic Force Microscopy (AFM) using a
Bruker ICON AFM in tapping mode. Height and phase maps were simultaneously acquired to validate
the homogeneity of the films. Gwyddion software was used to treat the data and obtained the roughness
(root mean square, rms or Sq). The resonance frequency was in the range of 270±10 kHz.
174
A.2.6. Raman Spectroscopy
Raman spectra was acquired using a Jobin Yvon/Horiba Labram spectrometer equipped with a liquid
N2-cooled CCD detector. The measurements were done in micro-Raman and backscattered
configuration using a blue laser (λ=488 nm). The laser with a spot size of 1 µm2 was focused on the
sample surface using an objective ×100. The spectra acquired were corrected using a silicon reference
with a theoretical position of 520.7 cm-1 at room temperature.
A.2.7. Electrical Characterization of thin films and devices
The resistivity of the samples was obtained averaging five locations using in-line four-point probes setup
from Lucas Lab (S302 four point probe manual stand).
The resistance measurements as function of temperature and in controlled atmospheres described in

Section 6.3 were carried out using two different temperature cells equipped with electrical microprobes:
Nextron chamber (for high temperature) and Linkam chamber (below room temperature). The
temperature, pO2, flow and electrical measurement are automatically controlled through LabView using
[69]
the equipment described in reference . The electrical properties were measured in van der Pauw
configuration, with silver contacts painted on the corners of the samples, applying small excitation
current (10-50 µA). The resistivity was measured in two Van der Pauw configurations, as presented in
Figure A.3, where the current was applied in both directions. In the resistance-temperature experiment
described in Section 6.3, the current was applied during the whole experiment in one single
configuration to track changes with time and temperature. At the end of each temperature plateau, the
current was applied in the two configuration presented in Figure A.3 and the resistivity was calculated
according to the van der Pauw method.[170]
Figure A.3. Scheme of the two van der Pauw configurations used to measure the resistivity as function of
temperature. The current was applied forward and reversed for each configuration.
175
The electrical characterization of the devices, i.e. current-voltage (I(V)) characteristics and chrono-
amperometry (I(t)), was carried out using a source measure unit (Keithley 4200) with two tungsten
probes mounted on micromanipulators and operated in voltage-controlled mode. All the setup was
placed on a marble table to reduce vibration which could affect the contact between the tips and the
electrodes.
176

Monc As I Luque 2022 Diffusion

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Monc As I Luque 2022 Diffusion

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Monc As I Luque 2022 Diffusion

Uploaded by

Copyright:

Available Formats

Structure–property relationship in (La,Sr)MnO3

To cite this version:

HAL Id: tel-04621403

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

DOCTEUR DE L’UNIVERSITÉ GRENOBLE ALPES

Relation entre la structure et les propriétés dans les mémoires

Thèse soutenue publiquement le 27 juin 2022, devant le jury composé de :

Résumé de la thèse en français ........................................................................................................ ⅹⅹⅲ

References ................................................................................................................................... 153

Figure 4.7. Resistance-voltage (-electric field) characteristics of initialized Ti/LSM20\Pt standard

Figure 5.1. XRR pattern of the LSM50/LAO films. .......................................................................... 109

8W Eight-wise, referred to the cycling sense of a device

HZDR Helmholtz-Zentrum Dresden - Rossendorf (Germany)

RSM Reciprocal Space Mapping

We cannot reach any higher

If you want to go fast, go alone.

Le phénomène de commutation résistive (Resistive Switching en anglais, RS) a suscité beaucoup

La RS a été rapportée dans de nombreux matériaux. Cependant, le mécanisme sous-jacent de la RS, et

Croissance des couches de LSM

Figure 2. Composition de la couche déposée à différentes températures en fonction de la composition de la solution

Rôle de la nanostructure dans le LSM20

Commutation résistive dans le LSM50 épitaxial

Effet de la stœchiométrie de l'oxygène dans les couches en LSM

Enfin, la résistivité électrique des couches en LSM20/LAO et en LSM20/STO en fonction de la

 Différentes électrodes actives pourraient être étudiées.

Resistive switching phenomena

Device phenomenology and first observations

Strontium-substituted lanthanum manganite

Bulk: crystal structure, electronic properties and phase diagram

𝑇𝐹 = (𝑟𝐴 + 𝑟𝑂 )⁄[√2 (𝑟𝐵 + 𝑟𝑂 )]

Therefore, the two LSM compositions studied in this thesis are:

LSM thin films: electrical properties

Polycrystalline films and grain boundaries

Epitaxial films and strain

Oxygen mass transport

Memristive response of La1-xSrxMnO3

Device Type of Mechanism

LSM20 Poly.a) Irb) Vertical Ag/LSM20/Ir Interface VCM: change in  [173]

LSM20 Poly.a) Irb) Vertical Tip/LSM20/Ir Interface VCM: change in  [173]

LSM20 Epitaxial STO Planar Pt/LSM20\Pt Interface VCM [95]

LSM30 Poly.a) Pt/Ti/SiO2/Si Vertical Probe/LSM30/Pt Interface VCM: change in  [103]

LSM33 Epitaxial STO Planar Ag/LSM33\Ag Interface SCLC [22]

Role of the electrodes

Scope and organization of the thesis

Pulsed Injection Metal Organic Chemical Vapor Deposition ............................................ 23

Pulsed Injection Metal Organic Chemical Vapor Deposition

Device electrical characterization

+Vreset Vamp << |Vread | 5 cycles

Transmission line measurements (TLM)

𝑅𝑇 = 2𝑅𝑀 + 2𝑅𝑐 + 𝑅𝑠𝑒𝑚𝑖𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟

L (variable) LSM t W=200 µm LSM W

 the slope, m=RS/W,

Geometrical optimization of the reactor

Parameter Optimal value, growth of LSM

Substrate temperature 700 °C a)

Evaporation 210 °C - 250 °C

Total pressure 5 Torr