BABICH

THESE DE DOCTORAT DE
L'UNIVERSITE DE NANTES
ECOLE DOCTORALE N° 596

Matière, Molécules, Matériaux
Spécialité : Physique
Par
Danylo BABICH
Electron-lattice coupling at the Mott transition driven by electric
and/or light pulse
_____________________________________________________________________
Couplage électron-réseau à la transition de Mott induite par impulsion électrique
et/ou lumineuse
Thèse présentée et soutenue à Nantes, le 07/12/2020
Unité de recherche : Institut des matériaux Jean Rouxel (IMN) – UMR 6502
Rapporteurs avant soutenance :

Silke Biermann Professeure des Université ; CPHT — Ecole Polytechnique
Sergiy Lysenko Professor ; University of Puerto Rico
Composition du Jury :
Président : Hervé Cailleau Professeur émérite, IPR — CNRS — Université de Rennes 1

Examinateurs : Claire Laulhé Maître de conférences, SOLEIL — Université Paris-Saclay
Serhiy Kondratenko Professor ; Taras Shevchenko National University of Kyiv
Benoit Corraze Maître de conférences - HDR; IMN-CNRS – Université de Nantes
Dir. de thèse : Etienne Janod Directeur de Recherche; IMN-CNRS – Université de Nantes
Co-dir. de thèse : Laurent Cario Directeur de Recherche; IMN-CNRS – Université de Nantes
Invité(s)
Shinichiro Iwai Professor ; Tohoku University
ACKNOWLEDGMENTS
This manuscript summarizes three and a half years of my stay at the Jean Rouxel Institute
of Materials of Nantes (IMN). It was a great pleasure to become a part of the Physics of materials
and nanostructures (PMN) group, especially of the "BEL" subgroup. This time was rich in a happy
and fruitful experience. My Ph.D. project has become possible thanks to funding from the University
of Nantes within International French-Japan Associated Laboratory IM-LED, Impacting Materials
by Light and Electric field, and watching real-time Dynamics.
I firstly acknowledge members of my defense committee who spent time reading and
discussing my thesis. The discussion during my defense was interesting and pleasant.
I would like to pay special regards to my "BEL" supervisors: Etienne Janod, Laurent Cario,
and Benoit Corraze. It was a pleasure to be their student and work under their leadership. I
appreciate a lot their human and scientific support. Work with them kept my interest in the research
and motivated me to further scientific and self-development. I'm incredibly thankful for your
patience in teaching me during my Ph.D. In addition, I indebted to Etienne for the overnight
corrections of this manuscript and every weekend-long discussion about the thesis subjects with
pauses for running. I will miss Laurent's fresh view on the scientific problem, which guide me a lot
in the scientific labyrinth, and Benoit’s supervising at experiments in the lab. I was very fortunate to
have three supervisors simultaneously, and from each of them, I have learned a lot! Besides, I would
like to thank "BEL" team members, Ph. D. students and interns Julien Tranchant, Marie-Paule
Besland, Shunsuke Sasaki, Coline Adda, Michael Rodriguez, Louis Beni Mvele Eyea, Léo Laborie,
Corentin Logé, Zohra Khaldi, Musa Mezud, and Abdel Mesoudy for extraordinary teamwork and
aids. I would like to recognize IMN Researchers Patricia Bertoncini, Eric Faulques, Bernard
Humbert, Jean-Yves Mevellec, and others for fruitful discussions and help during experiments. In
addition, I am very grateful to all my friends at IMN Anthony Impellizzeri, Angélique Marie,
Thomas Le Neel, Thibaud Guillemin, Hélène Roberge, Youssef El Moussaoui, Dominik
Winterauer, Daniel Funes, and other IMN students and researchers. Finally, I would like to thank
IMN director Florent Boucher and all IMN staff for the excellent working conditions at the institute.
The part of this Ph.D. involved our collaborators from the Institute of Physics in Rennes
(IPR). I want to say "Dziękuję bardzo" to Maciej Lorenc, my unofficial supervisor at IPR, with his
wonderful sense of humor and answering my emails after midnight. I want to thank IPR Ph.D.
students Alix Volte and Yevheniia Chernukha for their collaborative and intensive work on different
joint projects of our Ph.D. projects. I am very thankful to Céline Mariette, Elzbieta Trzop, Roman
Bertoni, Marco Cammarata, Marina Servol, Éric Coller, and Hervé Cailleau for enjoyful
collaboration within several projects of my Ph.D.
An exciting and exotic part of my project was related to Japan and, in particular, for
collaboration with the Shinichiro Iwai group at Tohoku University in Sendai and Keiki Fukumoto
group at the KEK Synchrotron in Tsukuba. I have enjoyed my stay in their labs. I am very grateful
to Prof. Iwai for the opportunity to perform ultrafast measurements in his lab. This measurement
becomes possible thanks to Prof. Yohei Kawakami and Tatsuya Amano help during my work in the
lab. I also appreciate discussions and aids from all S. Iwai team members. The second part of my
research in Japan was done at KEK Synchrotron in Tsukuba with Keiki Fukumoto's help. It was a
pleasure to work with Keiki, and I learned a lot at that time.
3
My Ph.D. also included work at big research facilities such as SLS Synrotron and SOLEIL
Synchrotron, and I am very grateful to the beamline scientists Peter Warnicke and Felisa Berenguer
for the measurements there. I also thank a lot to our collaborators Daniel Bedau, Tyler Hennen,
Jonathan Rupp.
This work also includes many other collaborations. I appreciate work and discussions with
Vinh Ta-Phuoc and Rodolphe Sopracase from the University of Tours, Elsa Abreu and her team at
ETH, Laura Bocher and her group at Université Paris-Sud, and others whom I forget to mention.
I warmly thank my Master’s program advisor at the University of Maine, Thomas Pezeril
(now at the University of Rennes), for the possibility to study at this University and for his
organizational help and support during my education there. It opened a door for my study in France.
Also, I thank Thomas for the scientific collaboration during my Ph.D.
The most important period of my life was the education at the Chair of Optics, Faculty of
Physics of the Taras Shevchenko National University of Kyiv. I am very thankful to my supervisors
and lecturers there: Leonid Poperenko, Andrii Yakunov, Serhii Kondratenko, Vasyl Yashchuk,
Serhii Yablochkov, Kateryna Yablochkova, Vadym Prokopets, Serhii Zelenskyi, Oleksandr
Kopyshynskyi, Igor Shaykevych, Oleksii Makarenko, Viktoriia Stukalenko, Liudmyla Melenevska,
Vasyl Stashchuk, dean of Faculty of Physics Mykola Makarets and all other administration.
I decided to become a physicist thanks to my advisers during my high school education
Oleksandr Zhmailo, Prof. Oleksii Nazarenko, Prof. Andrii Yakunov. Also, I am grateful to my
lyceum teachers Olena Kostiuchenko, Nataliia Fedorenko, Luidmyla Zinchenko, Sofiia
Nechyporenko, Liudmyla Nedbailo, Alla Omelianenko, Iryna Matviichuk, Tamara Tsarkova, Larysa
Bondarenko, lyceum director Oleksandr Nedbailo and others.
I succeeded with the Ph.D. thanks to the continuous support of my friends Rostyslav
Danylo, Yevheniia Chernukha, Andrii Shcherbakov, Serhii Derenko, Mykyta Redkin, Lukas
Nadvornik, Andrii Maksymov, Artem Levchuk, Tymur Papiev, Oleg Stetsuk, Denys Grekov, Artem
Husiev, Anatolii Makhort, Valentyna Pobiedina, Andrii Yampolskyi, Artem Sribnyi, Ruslan
Ryskulov, Vitalii Holovin, Oleg Stetsuk, Oksana Reznichenko, Ievgeniia Chaban, Nazar
Oleksiyevets, Marjan Pokhylyy and many more. I appreciate a lot their help and communication.
Finally, I would like to thank my parents: Pavlo Babich and Vira Bilorus, my grandparents
Antonina Matvienko, Mykola Bilorus, Lubov Babich, my brothers and sisters: Taras Babich, Mariya
Babich, Olena Babich, Mykola Babich, and other families.
4
CONTENTS
CONTENTS ..................................................................................................................................................... 5
INTRODUCTION ......................................................................................................................................... 7
Chapter 1 INSULATOR TO METAL TRANSITION IN MOTT MATERIALS ................. 9

1.1 Basic physical concepts about correlated systems and Mott insulators .................................. 10
1.1.1 Mott insulators: definition ..................................................................................................... 10
1.1.2 Classical control parameters of Mott Insulator to Metal Transitions ........................... 14
1.2 Typical examples of 'canonical' Mott insulators: the AM4Q8 and V2O3 systems .................. 16
1.2.1 The vanadium oxide (V1-xCrx)2O3......................................................................................... 16
1.2.2 AM4Q8 compounds ................................................................................................................ 23
1.3 Mott transitions under Electric field.............................................................................................. 26
1.3.1 Trivial types of transitions (Joule heating, Electromigration)......................................... 26
1.3.2 Electric Mott transition (dielectric breakdown phenomena) .......................................... 27
1.3.3 Microelectronics applications of Electric Mott transition............................................... 37
1.4 Insulator to metal transition driven by light in correlated materials ........................................ 45
1.4.1 Generality about the photoinduced phase transition explored by time-resolved
experimental methods ..................................................................................................................... 45
1.4.2 Light-driven Mott transition ................................................................................................. 54
1.5 Open questions.................................................................................................................................. 66
Chapter 2 NATURE OF METALLIC PHASES CREATED AT AND OUT-OF

EQUILIBRIUM IN MOTT INSULATORS........................................................................................... 67
2.1 Anomaly of standard laws linking pressure and volume at the Mott transition in (V1 - xCrx)2O3:
rationalization by the compressible Hubbard model ........................................................................ 69
2.1.1 Main prediction of the compressible Hubbard model .................................................... 70
2.1.2 Consequences on Raman spectrum .................................................................................... 73
2.1.3 Raman Spectroscopy study of optical phonon modes at the Mott insulator to metal
transition in the (V1-xCrx)2O3 system .............................................................................................. 74
2.1.4 Raman study of the Mott insulator to metal transition in (V1 - xCrx)2O3 as a function of
Cr content .......................................................................................................................................... 78
2.1.5 Discussion ................................................................................................................................ 80
2.2 Lattice Contraction at the Electric-Field-Controlled Mott Transition in (V1-xCrx)2O3 in the PI
phase .......................................................................................................................................................... 81
2.2.1 Study on (V0.975Cr0.025)2O3 crystal .......................................................................................... 82
2.2.2 Study on (V0.95Cr0.055)2O3 thin-film device........................................................................... 85
2.2.3 Discussion ................................................................................................................................ 88
2.3 Shock wave-induced non-volatile IMT on Cr-doped V2O3 thin films ................................... 90
Chapter 3 OUT-OF-EQUILIBRIUM MOTT INSULATOR TO METAL TRANSITIONS:

CARRIERS AND LATTICE DYNAMICS ............................................................................................ 92
3.1 Light control of Electric Mott transition and carrier dynamics................................................ 94
3.1.1 Electro-optically induced Mott transition .......................................................................... 94
5
3.1.2 Concept of an artificial electro-optical Mott neuron .....................................................100
3.1.3 Carrier lifetime in Mott insulators .....................................................................................103
3.2 Ultrafast THz-pulse-driven dynamics in Mott insulator GaTa4Se8 .......................................107
3.3 Photoinduced insulator to metal transition in (V0.95Cr0.05)2O3-PI and V2O3-AFI ................111
3.3.1 Evidence of a non-thermal insulator to metal transition induced by a femtosecond laser
pulse in (V0.95Cr0.05)2O3-PI .............................................................................................................114
3.3.2 Photoinduced insulator to metal transition in V2O3-AFI: evidence for a non-thermal
strain wave pathway .......................................................................................................................118
3.3.3 Physical picture of the photo-induced insulator to metal transition in (V0.95Cr0.05)2O3-PI
and V2O3-AFI..................................................................................................................................130
CONCLUSIONS .........................................................................................................................................132
ANNEX 1 Sample preparation..................................................................................................................134
ANNEX 2 Electro-optic pump / electric probe study on a GaTa4Se8 single crystal ......................138
ANNEX 3 Raman measurements .............................................................................................................141
ANNEX 4 Complementary characterization of conductive filamentary path after non-volatile

Resistive Switching of (V1-xCrx)2O3 ............................................................................................................143
ANNEX 5 Ultrafast measurements on (V1-xCrx)2O3 System ................................................................148
REFERENCES ............................................................................................................................................151
6
Introduction
INTRODUCTION
Mott insulators are a broad class of strongly correlated materials with emergent properties
important both from a fundamental viewpoint but also for modern electronics applications.
According to conventional band theories, such systems should be metallic. Their insulating ground
state, resulting from the on-site electron-electron Coulombian repulsion, which favors the
localization of electrons on atomic sites, is not accounted for in conventional band theories. At
thermodynamic equilibrium, a Mott insulator to metal transitions induced by charge doping or
pressure leads to spectacular properties, such as high-temperature superconductivity in iron- and
copper-based compounds. In this field, one of the current frontiers of research is the study of such
Mott insulator to metal transitions in out-of-equilibrium conditions, driven by external stimuli such
as short electric field or ultrafast laser pulses. In this context, the recently discovered Mott insulator
to metal transition driven by electric field (hereafter called Electric Mott Transition, EMT) is of
particular interest.
On the one hand, this property holds promises of breakthrough applications in information
technologies. This is particularly timely and relevant, since the context of continuous and spectacular
developments of computing technologies initiated in the 60’s is today at a crossroads. The driving-
force behind this progress, i.e. the downscaling of semiconductors-based transistors, is now about to
end due to obvious physical limits. As a consequence, a huge worldwide research effort is dedicated
to find alternatives to the current semiconductor-based computers based on the old Von Neumann
architecture. Mottronics, a field of application aiming at using the properties of Mott insulator in
microelectronics, is one of the promising alternatives currently considered. The recent discovery of
two applications of the Electric Mott Transition, the non-volatile Mott memory and the first
implementation of a single component Mott artificial neuron, could be the first promising examples
of a long series.
E-field Light
Mott Metal
Insulator
On the other hand, the fundamental physical mechanisms explaining the out-of-equilibrium
EMT are not clearly established to date. The role of Mottness (i.e. of the basic ingredients explaining
the Mott insulator ground state) in the EMT, the nature of the metallic phase generated under electric
field and the dynamics of creation of the metallic phase clearly need to be clarified. Interestingly, a
major proposition was that the creation of hot carriers by the electric field is the key parameter
driving the EMT. It implies that similar effects might be generated by other stimuli prone to promote
excited carriers, such as the application of light pulses. Moreover, the fantastic developments of
femtosecond laser during the last two decades opens the possibility to probe the dynamics of Mott
insulator brought out-of-equilibrium with ultimate temporal resolutions.
7
Introduction
In this context, my Ph.D. thesis work aims at deepening the understanding of Mott insulators
brought out-of-equilibrium, more specifically to clarify the dynamics of creation and the nature of
the electric-field- or light-induced metallic state, in order to unveil the mechanisms at work in the
EMT. In the longer term, the objective is to contribute to the development of Mottronics, thanks to
a better control of the out-of-equilibrium Mott insulator to metal transition induced both by electric
field and/or light pulses.
8
CHAPTER 1 – Insulator to metal transition in Mott materials
Chapter 1
INSULATOR TO METAL TRANSITION IN MOTT MATERIALS
***
This chapter introduces the basic concepts explaining Mott insulators' properties and
gives a state of the art about insulator to metal transitions induced by electric field and
ultrashort light pulses in this broad class of compounds. In the first part, it presents
the general concept defining what a Mott insulator is. It presents then two emblematic
families of Mott compounds, which are central in this work, namely the V2O3 and
AM4Q8 systems. The second part presents the current understanding of the Electric
Mott Transition and its envisioned applications in new components for
microelectronics (a field called Mottronics). Finally, the last part introduces the
domains of photoinduced phase transitions and reviews the few studies of photo-induced
phenomena in Mott insulators.
***
Mott insulators represent a broad class of strongly correlated materials with promising
properties. During the last decades, many types of research dealt with Mott phenomena in transition
metal oxides. It led to the discovery of emergent properties like high Tc superconductivity, colossal
magnetoresistance, and insulator to metal transition (IMT). The IMT, also called Mott transition, can
be induced under different external stimuli such as pressure or chemical doping [1]. Nevertheless,
microelectronic application of Mott materials became possible only with the recent discovery of
voltage-driven IMT, i.e. Electric Mott Transition (EMT), which enables easily controllable creation
of simple electronic devices [2–5].
Electric Mott Transition manifests itself in resistive switching [2] resulting from the
promotion of the system into a non-equilibrium state. Intense debates have been launched about the
underlying mechanisms of the EMT. The mechanism proposed by our group, for the initial spark
leading to metallicity creation, is a massive carrier excitation due to an electronic avalanche
phenomena under electric field [6]. At this stage, the full pathway of the Electric Mott Transition,
including the nature of the phase created after resistivity switching, remains unclear. Nevertheless,
several experimental results indirectly suggest that the filamentary path created within the transited
crystal is made of compressed metallic domains [7–9]. It suggests an important interplay between
electronic and lattice degrees of freedom, which is challenging to confirm.
On the other hand, the avalanche breakdown scenario suggests that light could be another
relevant stimulus to break the Mott insulating state. The similarity between electric and light-driven
phenomena and its connection to the Mott physics remains today a completely open area of research.
But if the two phenomena share some similarities, the ultrafast femtosecond laser pump-probe
technique would give access to the real-time dynamics in Mott insulator just after promotion into
the non-equilibrium state and, might therefore give some news clues to understand the Electric Mott
transition.
This thesis aims to better understand the insulator to metal transitions induced by electric
and/or light pulses in Mott materials. The next section introduces a few basic concepts allowing to
understand what is a Mott insulator.
9
1.1 Basic physical concepts about correlated systems and Mott

insulators
1.1.1 Mott insulators: definition
The Mott insulators are a class of materials with an integer number of electron per site, which
should be metallic according to classical band theory. However, many of such compounds, like for
example NiO, demonstrate insulating behavior [10]. In the middle of the 20th century, Sir Nevil
Mott proposed an explanation of this phenomenon. His description was related to on-site electron-
electron repulsion, not accounted in conventional band theories [11,12].
1.1.1.1 The simplest case: the one -band Mott insulator
When the on-site Coulomb repulsion (Hubbard) energy U becomes higher than the
bandwidth W, a band initially half-filled is split into two sub-bands, the Lower (LHB) and Upper
(UHB) Hubbard Bands. This situation is shown in Figure 1.1 in the case of a single band system
with one unpaired electron per site (half-filled system). When U is larger than the bandwidth W, a
Mott-Hubbard gap EG opens up between the LHB and the UHB, roughly equal to Mott ≈ U - W.
t (gain)
e-
e-
U (cost)
EF EF
U
DOS
DOS
LHB UHB
Energy W Energy
W= 2 z t EG=U-W
Figure 1.1: Density of states represented for a one-band half-filled system in the case of a typical metal (left
part) and a Mott insulator (right). When U is smaller than the bandwidth W, the system is metallic.
Theoretically, the Mott insulator state can be well accounted for by the Hubbard model,
which corresponds to the following Hamiltonian:
(1.1)
10
where <ij> represents summation over nearest-neighbor lattice sites, t the hopping integral is equal
to 𝑊/2𝑧 (𝑧 is the coordination number) and 𝑎𝑖𝜎 +
the creation operator of site i with spin direction σ.
The Hubbard Hamiltonian describes the competition between the energy gain due to electron
delocalization (gain = t, hopping integral) and to energy cost due to occupancy of a single site by two
electrons (cost = U). Solving the Hubbard Hamiltonian for the many-electron system has been a
long-standing problem [13,14]. Fortunately, the Dynamical Mean-Field Theory (DMFT) established
during the 90’s [15] has succeeded in giving a proper theoretical description of Mott insulators and
their phase diagram.
(a) W
U=0
(b)
1
0.5 0.05
0
1 U /W = 0.5
0.5
0
U /W = 1.2
1 0.025 Critical point
0.5 Paramagnetic
Fermi liquid insulator
0
1 U /W = 2
U U C1 U C2
0.5 Or der ed
phase
0
EF - U
EF
EF + U 0 1 2
2 ENERGY 2
U /W
Figure 1.2: (a) Evolution of the electronic Density of States (DOS) of a half-filled system as predicted from
DMFT when the correlation strength U/W is decreased from 2 to 0. A Mott insulator to metal transition
occurs slightly above (U/W)c  1. (b) The schematic phase diagram displayed as kBT/W vs. U/W predicted
from DMFT for half-filled compounds. T, U and W are the temperature, the Hubbard electron-electron
repulsion term and the bandwidth, respectively. (a-b) [14].
From the above discussion, it is clear that the important parameter governing the properties
of Mott insulators is the U/W ratio, also called the correlation strength. Starting from the metallic
state at U/W = 0 with a quasiparticle peak (QP) including all the electrons, a narrowing of the QP
is observed upon increasing U/W and the lost spectral weight is transferred to new incoherent peaks,
the Lower and Upper Hubbard Bands, separated by the energy U (see Figure 1.2.a and Refs. [15–
18]). Immediately before the insulator to metal transition, the QP becomes very narrow and a large
fraction of the spectral weight is located in the LHB and UHB. Figure 1.2 b shows the phase diagram
of Mott insulators predicted by DMFT. The first-order (Mott) transition line around U/W  1
separates the paramagnetic metallic (PM) domain at low U/W from the Mott Paramagnetic Insulator
(PI) domain at high U/W. The high-temperature region of the phase diagram is universal, while low-
temperature behavior is material-dependent and corresponds to various long-range (for example,
magnetic or orbital) orders. We will see below that this theoretical prediction is in agreement with
the experimental P - T phase diagram (Figure 1.4), since the physical pressure directly affects the
bandwidth and therefore the U/W term.
11
1.1.1.2 Multiband Mott insulators
The concept of Mott insulator is actually very general and is not restricted to the trivial case
of one electron placed on a single orbital in each site. It is also valid for multiple active orbitals and
N electrons per site (providing that N is an integer), a situation very common in 3d transition metal
compounds, like e.g. V2O3. In such “multiband Mott insulators”, electrons are placed on different
orbitals in each site: the Coulomb repulsion between two electrons located on two distinct orbitals,
denoted U’, is then lower than the repulsion U between two electrons placed on an identical orbital.
Theoretically, we have U’ = U - 2J, where J represents the Hund energy, which tends to align the
spin of electrons placed on the same site along the same direction [19].
Figure 1.3 compares the simple “one-band” case, i.e. one orbital (M = 1) populated by exactly
one electron (N = 1) per site with two more complex multiband situations, two electrons (N = 2)
distributed over M = 2 or 3 orbitals at each site. The critical correlation strength (U/W)c above which
a Mott insulator state appears and the Mott gap Mott can be estimated in the three situations
considered in Figure 1.3. Evaluating the Mott gap value simply requires to calculate the energy of the
lowest electronic excitation: the transfer of an electron from a site to a neighboring site. The energy
of the lowest electronic excitation can be obtained by calculating the energy gain on the site, losing
one electron (N ➔ N - 1), and the cost on the one which receives it (N ➔ N + 1). This analysis
reveals several interesting features:
• the Mott gap increases going from the half-filled one band to the half-filled two-bands (two
electrons on two orbitals) situation, from Mott  U –W to Mott  U + J –W ,
• in the multiband case, the critical correlation strength (U/W)c is roughly three time larger in
the half-filled (N = 2 and M = 2) case compared to the non-half-filled (N = 2 and M = 3)
situation. Accordingly, for given U and W values, the Mott gap in the N = M case is much
larger than for the N  M situation. More generally, larger gap values are expected in all half-
filled (N = M) cases, where Mott = U + (M - 1) J – W, compared to non half-filled (N 
M) ones, where Mott = U – 3J – W [20].
This analysis of the conditions required to obtain the Mott insulators state in a multiband
context allows rationalizing many observations within the strongly correlated transition metal
compounds. For example, it explains why Cr2O3 is a large gap (Mott  3.2 eV) Mott insulator and
V2O3 is a correlated metal. Both compounds display the same corundum structure, have probably
quite similar bandwidth W and Hubbard repulsion U, and contains 3 t2g active orbitals (M = 3). Their
main difference is actually the electronic filling: due to its 3d3 configuration (N = 3), Cr2O3 falls into
the half-filled cases (N = M = 3) and is hence in the Mott insulator state with a large gap Mott  3.2
eV. Conversely, V2O3 has a 3d2 configuration (N = 2) corresponds to a non-half-filled situation
(N = 2 and M = 3) : in this case, the critical correlation strength (U/W)c is required to open a Mott
gap is very large (see Figure 1.3) and the actual U/W in this compound is lower than the critical
value, yielding a correlated metallic state.
12
(a)
1 orbital (M = 1), 2 orbitals (M = 2), 3 orbitals (M = 3),
1 electron (N = 1) / site 2 electrons (N = 2) / site 2 electrons (N = 2) / site
Ground state
N+1 site N-1 site N+1 site N-1 site

Lowest excited state
U’ U’-J
U U U’-J
Excitation : energy gain (N ➔ N-1 site) 0 U’ - J U’ - J
Excitation : energy cost (N ➔ N+1 site) U U’ +U 2U’ - 2J
Gap atom (atomic limit) = Ecost - Egain atom = U atom = U + J atom= U – 3J
Mott Gap Mott (extended solid limit) Mott  U – W Mott  U + J – W Mott  U – 3J – W
Critical U/W for Mott gap opening (U/W)c > 1 (rough) (U/W)c > 0.9 (rough) (U/W)c > 1.8 (rough)
(for J / U = 0.15) > 1.2 (DMFT) > 1.1 (DMFT) > 3.0 (DMFT)
(b) (c)
Figure 1.3: (a) Simplified view describing one-band and multiband Mott insulators starting from the atomic
limit, which explains how the Mott gap (= energy of the smallest charge excitation) can be deduced in both
cases. It allows estimating the Mott gap Mott as a function of the basic parameters U (Hubbard energy), W
(bandwidth), and J (Hund’s energy). In general, we have Mott = U + (M - 1) J - W for the half-filled case
(N = M) and Mott = U – 3J – W for N  M [20]. In the last line of the Table, the critical correlation
strength (U/W)c above which the metal to Mott insulator transition occurs is estimated for the typical value
of the Hund energy J = 0.2 U. (U/W)c is estimated in two ways, first within a rough approximation based
on setting Mott = 0 in the equations above, and also thanks to more precise DMFT calculations [20].
(b) Phase diagram obtained by DMFT for multiband Hubbard Hamiltonian with M = 2 orbitals, for
J / U = 0.15, as a function of the interaction strength U, the half-bandwidth D (D = W/2) and the number
of electrons—from empty (N = 0) to full (N = 4). Darker regions correspond to good metals and lighter
regions to bad metals. The black bars signal the Mott-insulating phases. (c) Phase diagram similar to (b) for
a three orbitals situation, from empty (N = 0) to full (N = 6) electrons. Both phase diagrams are extracted
from Ref. [20].
13
1.1.2 Classical control parameters of Mott Insulator to Metal Transitions
The conventional ways to break the Mott state and induce a Mott insulator to metal transition
are to change pressure, temperature, or doping. [1].
(1) Bandwidth controlled IMT. The process appears due to an increase of orbitals overlap,
enlarging the bandwidth W, and reducing the correlation strength (U/W). The easiest way to enhance
orbitals overlap is to apply external pressure [1,21]. It corresponds to crossing the Mott transition
line along the red arrow in the Pressure-Temperature phase diagram (see Figure 1.4).
Temperature
DOS
DOS
E E
Paramagnetic Paramagnetic
Mott Insulator (PI) Metal (PM)
Insulating magnetic
ordered phases
U/W ≈ 1.15 0
Figure 1.4: Simplified temperature-pressure phase diagram for Mott insulator systems shown as
temperature versus U/W and pressure P along with typical electronic Density of States (DOS) for relevant
regions of the diagram. In canonical Mott insulators such as (V1-xCrx)2O3, the crystallographic structure is
similar for both the Paramagnetic Mott insulator and Paramagnetic Metal phases, while different for low
temperature ordered phases. Red and blue arrows illustrate pressure and temperature controlled
isostructural IMT. The green arrow indicates the pathway for the temperature-induced phase transition
between the ordered low T phase, which is usually insulating, and the paramagnetic metal. This transition
is usually accompanied by a crystallographic change and consequently does not corresponds to a true Mott
transition.
(2) Temperature controlled IMT’s. This type of Mott transition becomes possible only
near the Mott line and corresponds to the blue arrow (vertical axis) in the P-T phase diagram (see
Figure 1.4). The IMT occurs between the low-temperature PM and the high-temperature PI
states [1,22]. This transition differs strikingly from the usual transitions from a low-T insulator to a
high-T metal. An example of such a more traditional IMT is indicated in Figure 1.4 by the green
arrow: this temperature-controlled IMT can occur with the crystallographic change, and therefore
cannot be considered as a Mott transition.
(3) Filling controlled IMT. This IMT occurs when the band filling deviates from an exact
integer number of unpaired electron per site (Figure 1.5) [1,2]. This may be achieved by tuning the
14
electronic filling thanks to chemical doping. Doping a Mott insulator creates a quasi-particle (QP)
peak in the vicinity of the Lower (hole doping) or upper (electron doping) Hubbard Band, with the
Fermi energy in the middle of the QP peak. This causes the appearance of low energy excitations,
which can be easily observed e.g. in optical conductivity (see Figure 1.6 a and Ref. [24]). A similar
situation also occurs in the case of photodoped Mott insulators (Figure 1.6 b). This is quite expected,
since photodoping corresponds to the creation of a hole in the valence band and of an electron in
the conduction band.
(b)
Electron doping
Electron-doped
DOS
Mott insulator (metallic)
DOS
E
E
Mott-FC
W
W
DOS
E Mott Insulator Paramagnetic Metal 0
Mott-FC
U/W W
DOS
E
Hole-doped
W Mott Insulator (metallic)
Figure 1.5 : (a) Diagram U/W versus filling for the doped Mott insulators, (b) other representation of
doped Mott insulators, with schematic DOS for electron and hole doping cases. Violet arrows indicate the
filling-controlled Mott transition departing from the Paramagnetic Mott Insulator state, while the red arrow
indicates the bandwidth-controlled IMT is already described in Figure 1.4.
For Mott insulators located close to the critical U/W value (point “UC” in Figure 1.5.a),
another mechanism might come into play. In undoped Mott insulators, the effective U (Ueff) and
hence the Mott-Hubbard values might be reduced by electron-electron screening effect (= inter-site
e- - e- interaction) [23]. Doping such Mott insulators close to UC generates carriers that further
reinforce screening, leading to a possible slight reduction of Ueff [23]. However, this small reduction
of Ueff upon doping is effective only for Mott insulators close to UC, and quickly becomes negligible
by increasing U/W.
(a)
(b)
Figure 1.6 : (a) Optical conductivity of doped Mott insulator calculated at various temperatures.
(b) Comparison of the calculated optical conductivity of chemically doped and photodoped Mott
insulators [24].
15
1.2 Typical examples of 'canonical' Mott insulators: the AM 4 Q 8

and V 2 O 3 systems
The Mott insulator state is found in many molecular and inorganic materials. Representative
inorganic examples of this broad class of compounds are the oxide (V1-xCrx)2O3 and the chalcogenide
AM4Q8 (A= Ga, Ge; M=V, Nb, Ta, Mo; Q=S, Se) Mott insulators. Both families are possible to
synthesize in our Institute of Materials in Nantes in the thin film or single crystal forms. In this
section, we will describe the main properties of these materials.
1.2.1 The vanadium oxide (V1-xCrx)2O3
Among the prosperous class of strongly correlated compounds, one of the most studied
system for half a century is undoubtedly the vanadium sesquioxide (V1-xCrx)2O3. This long-standing
and continuous interest partly comes from the existence of several insulator to metal transitions in
this system. Thanks to the pioneering work of the Bell labs group in the late 60’s – early 70’s [25–
29], it was early understood that some of these IMT’s could be one of the first experimental
realizations of the long-sought-after “Mott transition”, i.e. of an IMT primarily driven by electron-
electron on-site repulsion [12,30]. At room temperature, V2O3 adopts the corundum structure with
a space group 𝑅3̅𝑐 (Figure 1.7a) and is a paramagnetic metal [31]. A proper isovalent substitution of
vanadium per chromium (V1-xCrx)2O3 turns this system into a paramagnetic Mott insulating (PI)
phase. According to the early studies of this system, this substitution does not affect the band filling
nor the structure type but induces a tensile strain, in other words a negative chemical pressure effect.
Conversely, Figure 1.7b indicates that the application of positive physical pressure on (V1- xCrx)2O3
in the PI phase restores a metallic state through a first-order transition [26].
Overall, the two phase diagrams temperature – xCr [25] and temperature – pressure P [26]
are strikingly similar and contain the essential features of the theoretical phase diagram of a one-band
Hubbard model shown in Figure 1.2b. This good correspondence with theory explains partly why
the (V1- xMx)2O3 has become an emblematic correlated system. On the other hand, the experimentally
observed x – P equivalence leads the Bell labs group to propose the famous “unified” phase diagram
shown in Figure 1.7.c [27,29]. It is based on the idea that the V / Cr and V / Ti substitutions induce
an internal chemical pressure, which has the same effect as the application of external physical
pressure. In this description, a 1 % increase of x in (V1-xMx)2O3 is equivalent to applying a positive
(M = Ti) or negative pressure (M = Cr) of  4 kbar. This phase diagram exhibits a Mott IMT line
ending at a critical endpoint temperature Tendpoint ( 440 K for x = 1.35 %), separating a Mott insulating
phase from a correlated metallic phase.
16
Figure 1.7: (a) crystallographic structure of V2O3 highlighting the presence of V-V dimers along the c
direction. (b) Resistance vs. pressure across the Mott insulator to metal transition in (V0.9625Cr0.0375)2O3 [26].
(c) Unified phase diagram of (V1-xMx)2O3, with M = Cr and Ti proposed by the Bell lab group, based on a
hypothesis of equivalence between the application of pressure and V/M substitution. In this representation,
changing the V/M ratio by 1% is equivalent to applying an external pressure of ≈ 4 kbar [27,29]. (d)
Temperature – composition – pressure 3D phase diagram according to Ref. [27], where the shaded
surfaces represent the PM – PI and PM – AFI transitions. The hatched surface corresponds to the PI –
AFI transition. (e) Evolution of the critical Mott line separating the PI and PM phase in (V1-xCrx)2O3 in the
temperature T – pressure P diagram, according to Ref. [26]. (f) Evolution of Tendpoint, temperature of the
critical endpoint of the PI-PM transition in (V1-xCrx)2O3, as a function of the chromium content x, deduced
from data presented in (e).
The application of a physical pressure allows inducing an Mott IMT between the PI and PM
phases and XRD studies confirm that both phases share the same 𝑅3̅𝑐 space group [26]. This
transition is first-order with a unit cell volume drop (~1 %) at the IMT [25]. For a narrow range of
composition (V1-xCrx)2O3 with 0.7 % < x < 1.77 %, this Mott first-order transition line can also be
crossed by tuning the temperature. This thermally controlled IMT occurs at temperatures between
17
190K and  400 K depending on the composition (Figure 1.7 c). At low temperature, pure and Cr-
substituted V2O3 display a structural and magnetic transition towards a monoclinic (space group
I2/a) and antiferromagnetic insulating (AFI) phase. In (V1- x Crx)2O3 with x > 1.77 %, the transition
occurs between two insulating PI and AFI phases, while the transition occurs between the insulating
AFI phase and the paramagnetic metallic phase (PM) in pure V2O3 [23]. The AFI-PM transition is
not considered as the classical Mott IMT as it involves a crystallographic symmetry breaking and an
additional magnetic ordering. Figure 1.8 compares the primitive cells of AFI and PM phases
inscribed in the hexagonal unit cell of the α-corundum phase.
Figure 1.8 Comparison of trigonal (a) and monoclinic primitive (b) cells of high and low-temperature V2O3
phases, respectively inscribed in the hexagonal unit cell of the a-corundum phase. Vanadium ions only are
shown. For (b) hatched and open circles represent ions of opposite spins [32].
Historically, the above description of the “unified” phase diagram of (V1-xMx)2O3 relating
substitution rate x and pressure P has prevailed thanks to its elegant and synthetic character.
However, early studies of the Bell labs group have identified a clear difference between the
temperature T – x and T - pressure phase diagrams: the temperature Tendpoint of the critical point ending
the Mott line above room temperature is not composition-independent, as incorrectly suggests the
unified phase diagram. Figure 1.7d, which shows the full 3D T –x – P phase diagram [27], and Figure
1.7e [26] indeed unveil a significant decrease of Tendpoint when xCr increases, from 440 K for x = 1.35 %
to 390K for x = 3.75 %. The extrapolation of this trend, shown in Figure 1.7f, indicates that Tendpoint
should be around room temperature for xCr  0.08. Even though it remained unnoticed so far, this
evolution of Tendpoint supports the assumption of a P - x equivalence contained in the unified phase
diagram shown in Figure 1.5c, at least in the PI phase for x > 1.3 %. The P – x equivalence indeed
rests on the idea that the chromium substitution x induces a negative chemical pressure, which,
18
similarly to external physical pressure, impacts a fundamental parameter driving the Mott physics,
the bandwidth W (see discussion in Section 1.1). Actually, a major prediction of the Dynamical Mean
Field Theory is that Tendpoint is proportional to the bandwidth (kB Tendpoint  0.025W in the one-band
Hubbard model, see Figure 1.2). The observed decrease of Tendpoint with x in (V1- xCrx)2O3 hence
strongly suggests that the chromium substitution indeed induces a decrease of the bandwidth W.
Figure 1.9: (a) compared optical conductivity of (V0.989Cr0.011)2O3 (x = 1.1 %) measured under 6 kbar and
of pure V2O3 in ambient conditions, according to Refs. [19,33]. (b) spatially-resolved photoemission
spectra of (V0.989Cr0.011)2O3 measured at 220 K, highlighting the coexistence of metallic (in red) and
insulating (in blue) phases at the microscale. The image size is 50 x 50 µm.
However, the simple assumption of an x - P equivalence in (V1- xCrx)2O3 has recently been
questioned, especially at low chromium content x < 1.3 %, based on experimental and theoretical
studies of its electronic structure [19,34]; more specifically, it was argued that the x- and P- controlled
Insulator to Metal transitions operate according to two different pathways. The studies described in
Refs. [19,34] use two main arguments to question the x - P equivalence in (V1- xCrx)2O3 :
• First, the optical conductivity of (V0.989Cr0.011)2O3 (x = 1.1 %) measured under a pressure of 6
kbar is not metallic-like and is hence different from the one of pure metallic V2O3, (see Figure
1.9a), whereas both optical conductivities should be equivalent in the “x - P equivalence”
scenario. Moreover, spectacular evidence of a metallic and Mott insulator phase coexistence
at the microscale exists for such low Cr composition (see Figure 1.9b). According to Ref. [19],
this phenomenon does not simply correspond to the coexistence expected at any first-order
transition. It may actually result from a distribution of lattice strain within the sample caused
by the presence of Cr-impurities. This view agrees with a detailed EXAFS study of
(V1- xCrx)2O3 [35], according to which Cr atoms play the role of strain defects, and being the
smaller atoms, contract locally the V2O3 host lattice. As a result, adjacent V atoms relax
towards Cr defects, and V-V distances expand relative to pure V2O3. Due to the 3D slow
decay of the strain field, even a small concentration of Cr may be sufficient to form large PI
region. However, these PI domains are most probably not percolating for Cr concentration
smaller or close to 1 %, leading to the phase coexistence shown in Figure 1.9 b. A direct
consequence is that the metallic state observed in (V1- xCrx)2O3 for xCr < 0.7 % using
techniques such as resistivity is actually misleading: at such composition, a strong spatial
disorder should exist, with local domains under tensile strain in the insulating state.
19
(a) (b)
Trigonal
xCr dV-V dimer
Phase Pressure distortion rate trigonal dimer
(%)
(%)
(Å) (= trigonal - dimer)
dV-V c/6 PM ambient 0 15.6 small 2.697 strong Small

PI ambient 3 18.4 strong 2.748 small Large
strong
PI ambient 5 18.4 2.747 small Large
PI ambient 4 18.4 ** strong 2.747 ** small Large

PM 1.3 GPa 4 * * * * *
(c) (d)
Hypothetical Experimental distorted
undistorted VO6 octahedron
VO6 octahedron (trigonal distorsion, a1g
(Oh crystal field) D3d crystal field) a1g  eg-a1g
eg eg
+ trigonal + V-V
distortion dimerization smaller Larger
➔ Larger a1g occupation ➔ smaller a1g occupation
a1g* (antibonding) V2O3 (PM) (V1-xCrx)2O3 (PI)

a1g
dimer. (e) PM PI AFI
trigonal
V2O3 (V1-xCrx)2O3 V2O3
 eg-a1g a1g (bonding)
e g
eg : : :
occupation
(f)
PM PI
Figure 1.10 crystallographic structure of the V2O3 system in the PI and PM phases, and zoom on the V - V
dimer and its oxygen environment.
(b) Different structural parameters of V2O3 and (V1-xCrx)2O3, allowing the estimate of the degree of V-V
dimerization and the trigonal distortion rate. We define the trigonal distortion rate as the relative difference
between the vanadium within the V-V dimer and the distance between successive oxygen plane along the
c-axis (dV-V – c/6) / (c/6). This rate would be equal to zero in a hypothetical structure with undistorted
VO6 octahedra. The structural parameters were obtained from Refs. [27,36,37] . The “*” symbol indicates
either an estimated or undefined values, especially for the PM phase under pressure.
(c) description of the electronic structure of the V2O3 system starting from the atomic limit. The splitting
between the eg and a1g levels by the trigonal distortion is partly balanced by the vanadium-vanadium
dimerization, which lowers the a1g level energy.
(d) schematic view of the small (PM phase, V2O3) and large (PI phase, (V1-xCrx)2O3) energy splitting
∆𝑒𝑔𝜋−𝑎1𝑔 . A direct consequence is that undoped V2O3 is closer to the limit of degenerated multiband Mott
20
insulator with N = 2 electrons spread over M = 3 orbitals (see Figure 1.11).

(e) proposed occupation of eg and a1g orbitals in the three main parts of the phase diagram of the V2O3
system, according to Ref. [38] Where eg1 and eg2 correspond to two eg orbitals.
(f) Spectral function of the V2O3 system in the PM and PI phases calculated by LDA + DMFT [19,39],
evidencing the lower occupation of the a1g band with respect to the eg ones.
• The second argument to refute the x - P equivalence is based on the X-ray absorption
spectroscopy study at the vanadium pre-K edge as a function of x and P in
(V1- xCrx)2O3 [19,34]. The authors essentially argue that the x-induced IMT is driven by a
“crystal field” effect, whereas the P-induced IMT is controlled by the bandwidth. A closer
look at the electronic structure of (V1- xCrx)2O3 is necessary to understand this argument.
Figure 1.10 focuses on the relevant electronic levels of (V1- xCrx)2O3 in the atomic limit, mainly
derived from the vanadium 3d orbitals. In the corundum structure adopted by the
(V1- xCrx)2O3 system (see Figure 1.10a), the transition metal M = V or Cr is placed in an
octahedral MO6 environment: the five 3d levels of the M atoms are hence split by the crystal
field into threefold degenerated t2g and twofold degenerated eg levels. However, this classical
picture is complicated by two additional features related to the corundum structure. First, a
trigonal distortion of the VO6 octahedron splits the t2g levels into an a1g and double degenerate
eg levels (see Figure 1.10c) by an energy trigonal. Moreover, the existence of V-V vanadium
dimers, oriented along the c-axis and located in two face-sharing octahedra, further split the
a1g levels, essentially to form a bonding level a1g and an anti-bonding level a1g*. The dimer
energy stabilizes the a1g level, as shown in Figure 1.10c. Historically, a primary debate
regarding the electronic structure of V2O3 concerned the relative position of the bonding
level a1g and the eg levels. At variance with an initial proposal by Castellani et al. [40,41] in
1978, it is now well established both experimentally [38] and theoretically [42,43] that the a1g
bonding level lies slightly above the eg one (Figure 1.10 c-d). Consequently, the two 3d
electrons per vanadium are spread between the eg and the a1g level in the V2O3 system. A
direct consequence is that the orbital occupations in the PM and PI phases can be described
as the linear combination of eg eg states (two electrons in the eg orbitals) and eg a1g states
(one electron in the eg orbital, the other on the a1g orbital, see Figure 1.11 e). In works refuting
the x – P equivalence [19,34], the authors proposed that the most significant evolution in the
Cr-induced metal to insulator transition is a reduction of the a1g population. This lower a1g
occupation in the PI phase results from an enhanced energy difference ∆𝑒𝑔𝜋−𝑎1𝑔 between the
eg and a1g level (see Figure 1.10 e). Interestingly, a simple structural analysis, summarized in
Figure 1.10 b-d, confirms this statement initially proposed in Ref. [38]. The energy ∆𝑒𝑔𝜋−𝑎1𝑔
is indeed the difference between the trigonal splitting trigonal and the stabilization energy dimer
of the a1g level due to the V-V dimerization. Compared to (V1 - xCrx)2O3 in the PI state, trigonal
is smaller and dimer is larger in V2O3 in the PM state because of the smaller distance dV-V (2.70
Å for PM against 2.75 Å for PI phase), leading overall to a smaller ∆𝑒𝑔𝜋−𝑎1𝑔 in the PM phase
and hence to a larger a1g occupation. Finally, we note that modern DMFT calculations of the
21
electronic structure of V2O3-PM and (V1-xCrx)2O3-PI confirm the lower occupation of the a1g
bands compared to the eg ones.
At this stage, an interesting remark can be made: V2O3 in the PM phase is clearly close to the
situation of 2 electrons (N = 2) spread over 3 active orbitals (M = 3) described earlier in part 1.1.1.2,
which tends to favor a metallic ground state. Conversely, as the a1g level becomes higher in energy
and less “electronically-active” in (V1 - xCrx)2O3 in the PI state, one moves away from this (N = 2, M
= 3) situation towards the half-filled case (N = 2, M = 2). As discussed above, this will favor the
opening of a Mott gap, even with a U/W value unchanged with respect to V2O3 PM (see Figure
1.11). Thus, the metal-insulator transition induced by the chromium substitution differs from the
simple view of the bandwidth-controlled Mott transition. In contrast, according to the authors of
Refs. [19,34], the pressure-induced Mott insulator to metal transition in Cr-substituted V2O3 is much
closer to the bandwidth-controlled scenario considered by DMFT for a canonical Mott insulator
since pressure application modifies only the bandwidth without noticeable change of the a1g band
filling [34]. This "revisited" overview of the different pathways taken in x- and P - controlled IMT’s
in (V1 - xCrx)2O3 is summarized in the phase diagram shown in Figure 1.11. The phase diagram
interpolates between two cases of multiband Mott insulators relevant for the V2O3 systems, i.e. the
(N = 2 electrons on M = 3 orbitals) and the (N = 2, M = 2) situations as already discussed in Figure
1.10. It illustrates clearly the x - P inequivalence: inducing an MIT by Cr substitution and then
applying a pressure to induce an IMT does NOT bring back the system to its initial point.
Figure 1.11: phase diagram highlighting the Mott insulator (PI) and the metallic domains (PM) in the
multiband three orbitals / two electrons situation, plotted as a function of the correlation strength U/W
(horizontal axis). The Mott transition line between the PI and PM phases corresponds to DMFT
calculations with a Hund J = 0.15 U (upper and lower limits of the phase diagram [20] and J = U / 6 (red
∆𝑒𝜋−𝑎
𝑔 1𝑔
circle for 𝑊
= 0.25 , see Ref. [44]. The vertical axis corresponds to the energy difference
∆𝑒𝑔𝜋−𝑎1𝑔 between the low-lying eg levels and the a1g level located at higher energy, normalized to the
bandwidth W. For high ∆𝑒𝑔𝜋 −𝑎1𝑔 /𝑊 values, the a1g orbitals are no longer populated, and the situation
becomes similar to the half-filled two electrons (N = 2) / two orbitals (M = 2) case. As discussed in the
text, the phase diagram parameters are relevant for V2O3 and (V1-xCrx)2O3, which can therefore be
qualitatively positioned on the phase diagram.
22
1.2.2 AM4Q8 compounds
Compared to the V2O3 system, the AM4Q8 compounds (A=Ga, Ge; M= V, Nb, Ta, Mo;
Q=S, Se, Te) represent a more recent family of canonical Mott insulators. These materials exhibit a
lacunar spinel structure, in which the electronic sites correspond to the tetrahedral transition metal
clusters M4 shown in the inset of Figure 1.12a [45].
Figure 1.12b represents the molecular orbital scheme diagram of the clusters. The t2g levels
of the transition metals interact to form a1 level, doubly degenerated e level, and triply degenerated t2
level [46]. For A = Ga, each M4 cluster contains one unpaired electron among seven when M = V,
Nb, Ta. In the case M=Mo, the clusters hold eleven d electrons. These compounds have a narrow
gap of 0.1-0.3 eV, which can be tuned by chemical substitution on the chalcogenide site [47].
LDA+DMFT calculations on GaTa4Se8 confirms that the gap opens thanks to on-site
correlation [48]. The band structure and density of states calculated by LDA+DMFT are represented
around the Fermi level in Figure 1.12c, while a DOS scheme is shown in Figure 1.12d. Resonant
inelastic X-ray scattering measurements and LDA+U calculations have moreover revealed that the
Spin-Orbit Coupling (SOC) is an important ingredient to take into account in this family of
compounds [49,50]. GaTa4Se8 is indeed a rare example of the SOC+U Mott insulator (see Figure
1.12b and e) [49,50].
23
Figure 1.12 : (a) crystallographic structure of AM4Q8 (A = Ga, Ge; M = V, Nb, Ta, Mo; Q = S, Se)
compounds, highlighting the M4 tetrahedral clusters in blue. (b) Representation of the molecular orbital
diagram of the M4 clusters [49]. (c) Result of the LDA + DMFT calculation showing the formation of the
gap between the Lower (LHB) and Upper (UHB) Hubbard bands due to the Coulomb energy U [48]. The
formation of the gap is explained in (d) that sketches the density of state with or without taking into account
the effect of U. (e) sketch of the density of states explaining the formation of the gap in AM 4Q8 by taking
into account the effect of the Spin-Orbit Coupling (SOC) [49].
Besides their electronic structure, the AM4Q8 compounds display two important
characteristics of canonical Mott insulators: they are paramagnetic insulators above 55K [46,51,52]
and do not exhibit any temperature-controlled IMT up to 800 K at ambient pressure [2]. Moreover,
as represented in Figure 1.13a, GaTa4Se8 and GaNb4Q8 (Q=S, Se) undergo a bandwidth-controlled
insulator to metal transition under pressure, with superconductivity at TC ≈ 2-7 K in the pressurized
metallic state above 11 GPa [51,53]. Transport property measurements under pressure in GaTa4Se8
have proven that this pressure-induced bandwidth-controlled IMT is of first order with a hysteresis
(see Figure 1.13a) [48]. Moreover, optical conductivity measurements, displayed in Figure 1.13b,
24
confirms the formation of a quasi-particle peak and, therefore, of a correlated metal under
pressure [54]. Several studies have also demonstrated that the AM4Q8 undergo a filling-controlled
IMT when doped on the A site or the M site [55–57]. The AM4Q8 compounds display, therefore,
the filling and bandwidth control IMT expected in canonical Mott insulators. The AM4Q8 family also
exhibits astonishing electronic properties like negative colossal magnetoresistance, ferromagnetic
half metallicity, multiferroïcity, and Néel type skyrmion lattice, that are encountered in other families
of Mott insulators [55–60].
Figure 1.13: (a) Resistivity vs. temperature at different pressures in GaNb4Se8 and GaTa4Se8, showing the
bandwidth control insulator to metal transition. These compounds also become superconducting under
pressure at low temperatures [53]. (c) Optical conductivity vs. wave number measured for GaTa4Se8 in the
insulating and the metallic state beyond the Mott line under pressure (10.7 GPa). The sketches explain the
low energy excitations (red, blue, and green arrows) expected in the Mott insulator and the correlated metal.
In the correlated metal, the low energy contribution (in green) corresponds to the quasi-particle peak, a
typical signature of electronic correlation [54].
25
1.3 Mott transitions under Electric field
Classical types of Mott insulator to metal transitions driven by pressure or doping can hardly
be used in modern microelectronics applications. However, recent researches demonstrate the
possibility to induce an IMT under electric field. Several studies show nonlinear electrical behaviors
and volatile or non-volatile resistive switching in Mott insulators. Depending on the materials,
different mechanisms are at play to explain this out-of-equilibrium behaviors. This section deals with
Electric transitions in Mott materials and their mechanisms.
1.3.1 Trivial types of transitions (Joule heating, Electromigration)
1.3.1.1 Joule heating and ther mally driven IMT
Temperature can be used as a tuning parameter to trigger a Mott IMT (see Figure 1.4).
Applying an electric field at T < TIMT can also lead to an IMT if, due to Joule effect, the sample
temperature exceeds TIMT and cross the Mott line. Such a thermal mechanism is at play in correlated
metal in the close vicinity of the Mott line, as recently confirmed by a DMFT theoretical study [61].
The main signature of this thermally-induced IMT is a sudden increase in the resistance, which is
volatile as it is maintained only under electric field. This phenomenon was observed in Cr substituted
V2O3 system with low Cr content ((V0.99Cr0.01)2O3) [62] and other Mott insulators [63,64]. It involves
the transition from the low T metallic phase to the high T paramagnetic Mott insulator phase (Figure
1.4). Similar behavior was also observed on GaTa4Se8 under pressure in the vicinity of the Mott IMT
line [48].
A thermally driven mechanism involving the sudden change of resistance was observed
under voltage in numerous correlated compounds that exhibit a non-isostructural phase transition.
It is the case in VO2 [65,66], Ca2RuO4 [67], in pure V2O3 below the AFI - metal transition
temperature [62,68,69], and in magnetite Fe3O4 [70,71]. Depending on the working temperature,
this phenomenon can be volatile (temporally induced low resistance state) or non-volatile
(permanent change) [72,73]. Correlated materials showing a Joule heating induced thermally driven
IMT were mostly considered as possible selectors in ReRAM crossbar arrays [74] in order to suppress
the undesired sneak path currents [75].
1.3.1.2 Electromig ration and filling controlled IMT
In oxide Mott insulators, a filling-controlled Insulator to metal transition is easily achieved

by tuning the oxygen content [1]. Alternatively, the oxygen content may also be changed at the local
scale by applying a voltage pulse. In non-stoichiometric transition metal oxides, the electromigration
of oxygen vacancies was observed under electric field. Voltage induces a modification of the
metal/oxygen ratio, and a valence change of the cations appears in the vicinity of grain boundaries
or dislocations. At the local scale, a filling-controlled IMT may occur and lead to a non-volatile but
reversible bipolar resistive switching by the formation/destruction of a metallic filamentary path
between the electrodes. A bipolar resistive switching depends on the applied pulse polarity: some
26
filament or interfacial states are created with one polarity and destroyed by the opposite one.
Conversely, both polarities have similar effects for a unipolar resistive switching. Alternatively, the
electromigration phenomenon can also occur close to the metallic electrode/ insulator oxide
interface and lead to a bipolar resistive switching by modifying a Schottky barrier. This type of
resistive switching mechanism called Valence Change Mechanism (VCM) was intensively studied in
band insulators and led to the realization of a new type of Resistive Random Access Memory
(ReRAM) [76,77]. This process is also present in various Mott or correlated transition metal oxides.
For example, interfacial VCM type resistive switching was reported in systems like La2CuO4 [77],
Pr0.7Ca0.3MnO3 [78,79], and YBa2Cu3O7−x [80,81]. Alternatively, filamentary VCM type resistive
switching was observed in many transition metal oxide Mott-Hubbard and charge-transfer
insulators [77], such as NiO [82,83], CuO [84,85], and CoO [86], Fe2O3 [87], and MnOx. [88]. The
VCM type of resistive switching may involve the formation of oxidized or reduced phases. For
example, resistive switching in CoO films was proposed to be related to the formation of an oxidized
phase Co3O4 [88]. Conversely, in CuO films, resistive switching was associated with the formation
and destruction of conducting filaments made of a reduced phase, namely Cu2O and possibly metallic
Cu [85]. In NiO, resistive switching is also related to the creation of a reduced Ni metallic phase. In
this case, the Ni filament is formed by a thermally assisted ionic migration process while its
destruction occurs due to Joule heating. Consequently, unipolar resistive switching was mainly
reported for NiO [77,83,89].
The electromigration mechanism is limited in terms of applications. It indeed requires a
forming step, i.e. the application of a specific electrical pulse, usually of high power/energy, to
generate the initial filament or interface.
1.3.2 Electric Mott transition (dielectric breakdown phenomena)
Several groups have recently demonstrated the existence of IMT in a Mott insulator
subjected to an electric field exceeding a threshold value. The physical process behind cannot be
explained by Joule heating and electromigration phenomena. At first, this type of resistive switching
was observed by Prof. Tokura group in the one-dimensional charge transfer insulators Sr2CuO3 and
SrCuO2 [90] and in the insulating charge-ordered compound La2-xSrxNiO4 [91]. So-called dielectric
breakdown occurs in these compounds above a threshold field of the order of 0.1 to 10 kV/cm. The
same phenomenon was reported afterward by our group for the canonical Mott insulators
(V1 - xCrx)2O3 [92,93], NiS2-xSex [92], and AM4Q8 (A = Ga, Ge; M = V, Nb, Ta, Mo; Q = S, Se) [56,94].
Also, similar behavior was found in molecular Mott insulators K-TCNQ [95], and
 - (BEDT- TTF)2Cu[N(CN)2]Br [96], and [Au(Et-thiazdt)2] [97].
The initial theoretical explanations of the dielectric breakdown in Mott insulators subjected
to the electric field dealt with the Zener breakdown phenomenon [84–88]. For instance, calculations
were performed in 1D Hubbard chains using exact diagonalization [98,99], and time-dependent
density matrix renormalization group [100], or in the limit of large dimensions using dynamical
mean-field theory [101]. All these theoretical studies have predicted nonlinear behavior in the
current-voltage characteristics and the existence of a threshold field (Eth) beyond which a field-
induced metal appears. Zener type breakdown should occur when the electric field is such that it
27
bends the Hubbard bands by the gap energy EG within the length ξ of the order of the unit cell.
Hence, the Zener breakdown is predicted to occur for the strength of the electric field Eth ∼ EG / ξ
of the order of 106-107 V/cm [98,99]. It cannot explain the experimental values of the threshold field
that are at least two orders of magnitude lower [90,91,94,96].
As a consequence, volatile resistive switching in Mott insulators cannot be explained by a
Zener breakdown scenario. Alternatively, recent studies on the AM4Q8 Mott insulators support that
the dielectric breakdown originates from an electric field induced electronic avalanche
phenomenon [6,103]. This breakdown may induce a non-volatile resistive switching resulting from
the creation of conducting filamentary paths. A complete description of this mechanism will be given
below.
1.3.2.1 Phenomenolog y of the dielectric breakdown
The most prominent signature of the Electric Mott Transition observed in canonical Mott
insulators is the volatile resistance drop, also called resistive switching, that occurs under an electric
field. Figure 1.14 displays typical resistive switching observed on canonical Mott insulators GaTa4Se8
and (V1-xCrx)2O3 (x=0.15) when voltage pulses are applied to the circuit made of the Mott insulator
crystal in series with a load resistance (Figure 1.14). Each voltage pulse exceeding the threshold value
(of the order of a few kV/cm) induces a volatile resistive switching after a delay time. Fig. 1.8e, f
shows that the resistive switching occurs only above a threshold electric field Eth of a few kV/cm
(see red dotted line) and after a time tdelay, decreasing as the voltage across the sample increases. The
sample field ES after the resistive switching always lies on the same value Eth that also corresponds
to the lower voltage that can induce a resistive switch in DC measurements.
28
Figure 1.14 (a) Experimental setup. Universal dielectric breakdown I-V characteristics (top panels b, c) and
time dependence of the sample voltage VS(t) (bottom panels e, f) are displayed for three different types of
narrow gap Mott insulators. Blue dots correspond to the region below Eth, where no breakdown is
observed. Black symbols correspond to the I-V characteristic in the resistive switching region, above Eth.
The black dots show the initial I-V, before the breakdown, and the black squares indicate the final state.
The open symbols highlight a particular breakdown transition for easier visualization. Measurements on
GaTa4Se8 were performed at 77 K, on V2-xCrxO3 (x=0.3) at 164 K [92].
As a consequence, the Mott insulator compounds exhibit a particular current-voltage

characteristic with two branches. The first one corresponds to the non-transited state, which deviates
only slightly from Ohm's law. The second branch, which is almost vertical and lies at the threshold
field, corresponds to the "transited" state (Figure 1.14.b-c)
1.3.2.2 Microscopic behavior : the creation of hot electrons and

electronic avalanche
All Mott insulator compounds exhibit the same type of I(V) characteristic with a threshold
electric field in the 1-10 kV/cm range [92]. The magnitude of the threshold field in AM4Q8 Mott
insulators, as well as their I(V) characteristics, compare well with the threshold field values and I(V)
characteristics observed for avalanche breakdowns in narrow gap semiconductors [104]. For this
reason, it was proposed that the resistive switching observed in the Mott Insulators AM4Q8 originates
from an avalanche breakdown phenomenon [103]. In semiconductors, the avalanche threshold field
varies as a power law of the bandgap and follows the universal law Eth  EG2.5 [105,106]. Figure
1.15.f reveals that AM4Q8 compounds have a similar variation of the threshold field as a function of
the Mott-Hubbard gap [103]. This power-law behavior provides the first evidence that supports the
avalanche breakdown scenario in these Mott insulators.
29
Figure 1.15 (a - e) Time dependence of the voltage and intensity measured at 77 K across different AM4Q8
single crystals when they are subjected to voltages pulses. All the resistive switchings observed during the
pulses are volatile, i.e. the resistance is the same before and after the electric pulse. The current-voltage
characteristics are plotted from the values of I and V measured during the pulses, before (blue circles) and
after (open squares) the volatile transition (see corresponding symbols in I-V plot). f) Dependence of Eth
in Mott insulators and semiconductors. Threshold electric field as a function of the Mott gap EG for various
AM4Q8 compounds. The solid blue curve corresponds to a power-law dependence Eth  EG2.5. Inset:
comparison of the threshold fields versus gap dependence for the AM4Q8 compounds and classical
semiconductors. The solid blue line displays the universal law Eth[kV/cm] = 173 (EG[eV])2.5 observed for
semiconductors [103].
30
In their pioneering works, Fröhlich and Seitz [107–111] have established the dielectric
breakdown theory in semiconductors. They propose the existence of two different dielectric
breakdown regimes. A first regime, corresponding to a "clean" limit, occurs preferentially at low
temperature in ultrapure semiconductors with long mean free paths. In this case, the avalanche
process is initiated by the tiny number of electrons available, which gain energy independently to
each other. In classical semiconductors such as Si, Ge, or GaAs, avalanche breakdown is the
consequence of an impact ionization: electrons in the conduction band, accelerated by an electric
field, can gain enough energy to induce an impact ionization. This process generates electron-holes
pairs and promotes new electrons in the conduction band. The repetition of this process leads to a
free-carriers multiplication if the average ionization impact rate exceeds the electron-hole
recombination rate. In this regime, the electric-field-induced energy gain in the conduction band is
only limited by the electron-phonon scattering. As e- - ph scattering raises with temperature, the
threshold electric field Eth also increases with T in this regime. Alternatively, a second regime may
appear when e- -e- scattering is not negligible anymore. It may occur in a realistic situation where
defects induce discrete energy levels in the gap distributed on a typical energy width Δε. This specific
density of states is depicted in Figure 1.16.a and may be found in real Mott insulators, as shown in
Figure 1.16.b-c. Unlike the electrons in the conduction band, the trapped electrons cannot be
accelerated by an electric field because these levels are not continuously distributed. However, the
trapped electron gives rise to an efficient scattering with the conduction electrons and the phonon.
As a consequence, the thermal balance of the electronic system consists of the energy
transfer rates from the electric field to the conduction electrons (gain Pin) and from the electrons
localized in shallow levels to the lattice (loss Pout), as shown Figure 1.16d. Under electric field, the
electronic temperature TE rises above the lattice-bath temperature T0 until the two energy transfer
rates (gain and loss) equilibrate (see Figure 1.16e). But above a threshold field Eth, the system became
unstable as the energy rates gained and lost by the electrons can no longer be equilibrated.
Consequently, the electronic temperature rises drastically, which induces the electrical breakdown.
The model leads to two main predictions [107,108]. First, at low field, the resistance of the sample
should deviate from the Ohm's law predictably without external tunable parameter. Second, the
threshold electric field should exhibit a thermally activated behavior at high-temperature and a
crossover to the low-temperature regime Figure 1.17a). These predictions were tested in the series
of AM4Q8 compounds [6]. A nonlinear electrical behavior was indeed found at low field.
Moreover, Figure 1.17b shows the temperature dependence of the threshold field Eth
observed for several AM4Q8 single crystals. A remarkable agreement is found with the theory: the
existence of a threshold electric field, of two temperature regimes, and of an activated dependence
of Eth at high temperature [6]. These experimental observations support, therefore, that the resistive
switching in canonical Mott insulators originates from the creation of hot electrons under the electric
field.
31
Figure 1.16: Schematic description of the main ingredients involved in the dielectric breakdown in the dirty
limit according to the Fröhlich model. (a) The density of states displays a gap εG between the fundamental
impurity level and the conduction band, with discrete impurity levels spread on the energy width Δε. The
blue and red points indicate some electrons in the impurity levels (Ni) and in the conduction band (Nc). (b)
typical DOS in a slightly n-doped Mott-Hubbard insulator, according to [15,17]. An additional peak related
to localized n-type defects level appears contiguous to the Upper Hubbard Band. (c) Experimental evidence
of such defects level unraveled by STM/STS experiments performed close (red curve) and far (black curve)
from a missing Cl defect in Ca2CuO2Cl2 [112]. (d) Thermal balance of the electronic system. Electrons
have a temperature distinct from the lattice temperature. Pin and Pout are the heating and cooling powers
sustained by the electrons. The thermalization between electrons, the heating by the electric field, and the
cooling through the lattice are described. (e) Time evolution of the electronic temperature under several
values of the electric field. The dielectric breakdown regime appears above a threshold electric field Eth [6].
32
Figure 1.17: Evolution of the threshold electric field Eth with temperature. (a) Theoretical predictions of
Fröhlich in the dirty (high temperature) and clean (low temperature) regimes. (b) Temperature dependence
of the threshold field Eth for three compounds of the AM4Q8 family. In the case of GaMo4S8, four samples
(S1-S4) have been measured. The black lines are the fit with an activation law [6].
1.3.2.3 Macroscopic phenomena: the creation of a conducting

filamentar y path
In order to capture the main features of the Electric Mott Transition, a phenomenological
model was proposed in collaboration with LPS [92,113]. Under electric field, hot carriers are
produced, which breaks locally the Mott insulating state to generate a correlated metal phase. To
account for the competition between the two phases under electric field, a two-minima energy
landscape, where the Mott insulator (MI) state is at the lowest energy and the correlated metal (CM)
state at higher energy, was considered and implemented in a resistor network. The probability of
transition from the Mott insulating phase to the correlated metallic phase depends on the applied
voltage, while the probability of the transition back is a thermal activation process over a fixed energy
barrier. This modeling work shows that a volatile resistance drop is obtained thanks to the apparition
of the metallic filamentary path between the electrodes. Figure 1.18a and Figure 1.18b display the
dynamics of the creation / relaxation of metallic sites and of the filamentary path. At the beginning
of the pulse, all sites are in the insulating state, but under electric field, some insulating sites switch
into metallic sites. During the pulse, the production rate of metallic sites largely overcome their
relaxation rate. The fraction of metallic sites hence increases up to a critical threshold above which
a runaway process starts resulting in the formation of a metallic filamentary path across the sample,
bridging the two electrodes [92,113]. This process is illustrated in Figure 1.18c. Calculations are
combining the energy landscape model with a thermal model and support that the onset of the
resistive transition is solely driven by a purely electronic transition. At the same time, Joule heating
occurs once the metallic filament is created, and the current starts to rise in the circuit [113].
33
Figure 1.18: Modeling of the behavior of a Mott insulator under electric field (a) Resistive switching
observed after a delay time decreasing with the applied voltage. This applied voltage is higher for the blue
curve than for the red one [38]. (b) A concomitant increase of the fraction of metallic sites in the resistor
network. The orange area corresponds to the increase of metallic cells before the creation of the filamentary
percolating path (threshold value is ~0.00375). The red area corresponds to the increase of metallic cells
number during and after the creation of this filament. (c) Resistor network evolution as a function of time
showing the filament formation. The red lines at the top and the bottom stand for the electrodes where
the voltage is applied. The local electric field is represented by the color scale in snapshot 1: from zero in
dark blue color to red in the higher. Time of snapshot 1, 4, 6, and 8 are indicated in (a) [114].
Finally, we can summarize the electric Mott transition and concomitant resistive switching
during an electric pulse in Figure 1.19. Above a threshold electric field, free carriers are accelerated
within the Mott insulators, which enhances the electronic temperature and promotes the massive
creation of new carriers [6,92]. This phenomenon leads to an electronic avalanche breakdown, which
induces the formation of a metallic filamentary path within the material after a time delay (tdelay), which
varies as the inverse of the applied voltage (Figure 1.19a and b). The resistive switching occurs when
the filamentary path bridges the opposite electrodes (Figure 1.19c, d, e, f). Once this filamentary path
is created, two situations may be encountered after the end of the voltage pulse, as depicted in Figure
1.19 h and g. The filamentary path can either fully or only partially dissolve. In the last situation, the
resistance does not relax back to its initial value, if the remaining filamentary path percolates between
electrodes. This type of resistive switching is called non-volatile and is usually obtained when the
applied voltage is large compared to the threshold voltage. Conversely when the filament self
dissolves, the resistance relaxes back to its initial value and for this reason this type of resistive
switching is called volatile. This volatile resistive switching is usually observed for voltages slightly
higher than the threshold field. Recently, it was proposed that the stability of the filamentary path
could be related to elastic constraints and that below a critical size, the filamentary path could not be
stable by itself and would completely dissolve. In contrast, larger filaments would be maintained
during a longer time after the end of the electric pulse in a metastable state [115].
34
Figure 1.19: Schematic representation of the electric Mott transition. a) Voltage pulse applied to the Mott
insulator. The blue and red pulse correspond to different voltages above the threshold voltage, b) Variation
of the sample resistance before, during, and after the voltage pulses. Before the voltage pulse, the sample
resistance is in the high resistive state. During the voltage pulse, resistive switching to low resistive states is
observed after a delay time, which decreases with voltage increase. After the pulse, the sample resistance
either returns to its initial value (blue curve), meaning that the resistive switching is volatile, or the sample
resistance remains in the low resistance state, leading hence to a non-volatile resistive switching. c, d, e, f,
g, h) snapshots correspond to the time evolution of the formation of the metallic filamentary path within
the sample associated with the resistive switching. The brown hatched rectangles represent the
electrodes [116].
1.3.2.4 Open question about the nature of the metallic

filamentar y path cr eated at non-Volatile EMT
As discussed above, narrow gap Mott Insulators exhibit a non-volatile resistive switching for
electric fields well above the avalanche threshold field. The non-volatile transition has been
successfully demonstrated in the AM4Q8, (V0.85Cr0.15)2O3 and NiS2 compounds [2]. In all cases, the
transited states present a significant resistance drop after the non-volatile transition. Figure 1.
20 shows the resistance vs. temperature curves measured at a low bias level in the pristine and
transited states of a GaTa4Se8 crystal. Detailed investigation of the low-temperature part of the
transited sample transport shows a conductivity increase, which is completely cancelable by an
external magnetic field (Figure 1. 20 d, e). It would suggest a superconducting like behavior.
35
According to the GaTa4Se8 phase diagram (Figure 1. 20 g), a superconductivity state is possible only
under pressure [8].
In parallel, Scanning Tunneling Microscopy (STM) measurements have revealed that the
non-volatile resistive switching is related to an electronic phase separation at the nanoscale in
GaTa4Se8. While the surface topography of pristine GaTa4Se8 crystals is structureless, granular
filamentary structures qualitatively oriented along the direction of the electric pulses are observed
after resistive switching inside the crystal after cleaving. Spectroscopic (STS) experiments have
confirmed the existence of electric-field-induced metallic nanodomains of about 30-70 nm within
these granular filamentary paths. These nanodomains were carefully investigated by Energy
Dispersive X-Ray spectroscopy and by Transmission Electron Microscopy. No chemical
composition change nor any crystallographic symmetry breaking or amorphization between the
electrodes were detected at the nanometric scale (Figure 1. 20 c). This excludes the formation of
conducting bridge-like filaments, or of amorphous-crystalline transition as observed in phase change
materials [117].
Figure 1. 20 (a) Experimental setup for electric pulse application and image of GaTa4Se8 crystal connected
by gold electrodes [7]. (b) Resistivity change upon non-volatile resistive switching on GaTa4Se8 crystal
obtained by 2 point measurements [7]. (c) SEM image of transited GaV4S8 crystal after ion focused beam
preparation shows no structural modification at the nm scale [8]. (d) and (e) represent critical current and
critical magnetic field dependency of the resistivity on transited GaTa4Se8 crystal [7]. (f) Scanning tunneling
microscopy/spectroscopy of the local tunnel conductance at zero bias on cleaved GaTa4Se8 crystal,
transited by electric field. The green regions correspond to the insulating pristine-like state. Red and blue
areas represent high conductance (metallic-like) and strong insulating (zero conductance) states,
respectively [8]. (g) Schematic phase diagram of GaTa4Se8 compound with superconducting (SC) zone
under pressure [8]. h) Simulation of the filamentary path, assuming conservation of the total area of the
simulated sample and showing a coupling between red (compressed) and violet (expanded) cells [8].
36
The presence of granular superconductivity at low temperature in a transited crystal of

GaTa4Se8 is reminiscent of the superconducting transition observed under pressure below 2-8 K in
this compound [53]. It indirectly suggests the presence of compressed metallic nanodomains in
transited crystals of GaTa4Se8. In other words, electronic avalanche breakdown can probably induce
the collapse at the local scale of the Mott insulating state into a correlated metallic state and leads to
the formation of a conductive filament formed of compressed domains. Such a strong response of
the lattice, i.e. a volume contraction, is also observed at the bandwidth controlled IMT achieved
under pressure [118]. This interesting point will be addressed in Chapter 2.
1.3.3 Microelectronics applications of Electric Mott transition
At least two confirmed solutions for modern electronics based on Electric IMT in Mott
insulators popped up during the last decade. The Electric Mott Transition (EMT) could allow on the
one hand the creation of highly efficient memory cells and on the other hand the realization of a
single component artificial Mott neuron. In the next section, we introduce both applications.
1.3.3.1 Mott Random Access Memories
Flash memory is currently one of the leading technology for non-volatile storage. It is used
in Flash SD cards and Solid State Drives. However, the downscaling limit of this technology will
hinder its development soon [119]. Several emerging Random Access Memories (RAM) [120], like
Phase-Change RAM (PCRAM) [117], Magnetic RAMs (MRAM) [121], and Resistive RAM
(ReRAM) [4], are currently considered as interesting candidates to overcome the shortcomings of
Flash memories. ReRAMs appear as a very appealing solution among these potential candidates,
thanks to a very simple architecture and promising memory performances. In that context, the
EMT's observed in Mott insulators were investigated to build a Mott based ReRAM called the Mott-
memory. In that purpose, the transition needs to be non-volatile and reversible.
The EMT’s discovered in Mott insulator compounds like AM4Q8 or (V0.95Cr0.05)2O3 leads to
non-volatile resistive switching, which makes them potential candidates for ReRAM
applications [2,73,94,122]. Studies on this type of ReRAM was mainly focused on devices made of
(V0.95Cr0.05)2O3 and GaV4S8. The following sections present therefore the realization of MIM devices
using these narrow gap Mott insulators and describe the performances obtained in the context of
ReRAM applications.
The EMT’s observed in single crystals of the AM4Q8 family of compounds, as described
above, could be transposed to thin films of GaV4S8 [123]. Polycrystalline stoichiometric layers of this
compound were obtained either by non-reactive sputtering [124] or by Ar/H2S reactive
sputtering [125]. As presented in Figure 1.21, the application of electric pulses to Au/GaV4S8/Au
symmetric MIM structures with 2x2 µm² bottom electrode size enabled to obtain resistive switching
cycles with 4 < ROFF / RON < 6, as displayed in Figure 1.21. ROFF and RON are the resistance values of
high (OFF) and low (ON) resistance states, respectively. In this case, the pulse protocol consists of
an alternation of 3.2V / 7x500 ns multi-pulses for SET transitions with long pulses of lower voltage
for RESET (1.6V / 500 µs) [126]. Importantly the RS cycles could be obtained without the forming
37
step and with pulse durations as low as 100 ns. In terms of performances, more than 65000 cycles
were obtained with a 0.01% error rate, and data retention could be extrapolated to 10 years at room
temperature without resistance change [127].
Figure 1.21 (a) Top-view TEM picture of the 50×50 μm² Au/GaV4S8/Au MIM structure. The platinum
layer on gold top electrode is due to the sample preparation by the Focus Ion Beam technique.
(b) Schematic drawing of a 2×2 µm² Au / GaV4S8 / Au MIM structure cross-section and SEM pictures of
the corresponding substrate surface, before deposition of the GaV4S8 thin film and the top gold electrode.
(c) Switching performance of the 150nm thick GaV4S8 layer by applying successive 3.2 V / 7 x 500 ns
pulses separated by 3.5 μs off period, and 1.6 V / 500µs pulses [127].
The EMT property was also transferred from single crystals to polycrystalline stoichiometric
thin films of (V0.95Cr0.05)2O3 deposited by Ar/O2 reactive magnetron sputtering of V and Cr
targets [93]. Electrical performances were evaluated on symmetric Metal/Insulator/Metal (MIM)
TiN / (V0.95Cr0.05)2O3 / TiN devices with via diameters in the 330 - 1600 nm range (see Figure
1.22) [115]. The main results, such as programming current, endurance, scalability, and retention
times, are displayed in Figure 1.23 [128]. The application of unipolar 3.5 V / 100 ns pulses to a device
with 330 nm vias and a 100 nm thick (V0.95Cr0.05)2O3 film triggers resistive switching cycles with an
average ROFF / RON ratio of 50, without any forming step before cycling (see Figure 1.23a). In this
case, a programming current of 48 µA with an energy per pulse of 8 pJ was sufficient to induce the
RS (Figure 1.23). Endurance measured on another (V0.95Cr0.05)2O3 based device exceeds 5000
successive cycles, as shown in Figure 1.23c.
The downscaling properties were also investigated on MIM structures by reducing the
electrode via diameter from 1.6 µm down to 330 nm. As displayed in Figure 1.23d, the ROFF / RON
ratio strongly increases with decreasing pad area and reaches values up to 100 for via diameter of
330 nm. This scaling of ROFF / RON ratio with 1/S is consistent with previous observations on GaV4S8
38
thin films [127]. This result can be easily explained considering a filamentary model. As long as the
cycling involves the creation / full dissolution of a single filament, ROFF is indeed expected to scale
with the inverse of the pad area 1/S while RON is expected to depend only on the resistance of the
filamentary conducting path expected to be independent of the pad area. As a consequence,
ROFF/RON should increase as 1/S for small pads area, which is observed experimentally for areas
below 2 µm² (Figure 1.23d) and should keep increasing as long as pad sizes remain larger than the
filamentary conducting paths. ROFF/RON ratios larger than the 103-104 current values can thus be
expected with further pad size downscaling.
Figure 1.22: (a) Top view of a substrate used for deposition of (V0.95Cr0.05)2O3 thin films and (b) SEM
picture of a bottom electrode with via of 330 nm in diameter. (c) Cross-sectional schematic view of the
device based on (V0.95Cr0.05)2O3 and (d) corresponding backscattered electron SEM picture of the multi-
layer [128].
39
Figure 1.23: Typical electrical characteristics obtained for TiN/(V0.95Cr0.05)2O3/TiN MIM structure
exhibiting pad size in the 1.6 µm to 330 nm range. (a) Resistive switching cycles obtained on devices with
100 nm thick films of (V0.95Cr0.05)2O3 and 330 nm vias, by applying 3.5V / 100 ns pulses.(b) Time evolution
of current (Imax = 48µA) and energy (E = 8 pJ) during a 3.5V / 100 ns pulse inducing one of the OFF→ON
transitions in Fig. 4 (a). (c) Cycling endurance obtained on a device with a 1.6 µm via and a 880 nm thick
film demonstrates more than 5000 RS cycles. After this endurance test, RS cycles with a similar ROFF/RON
ratio than Fig. 4 (c) were obtained on this device. (d) Evolution of ROFF/RON ratio with the electrode surface
for thin films of the Mott insulators GaV4S8 and (V0.95Cr0.05)2O3. (e) Data retention at room temperature,
extrapolated to 10 years, for OFF (red circles) and ON (blue squares) resistance states on an 880 nm thick
film of (V0.95Cr0.05)2O3 [128].
The stability of high and low resistive states obtained on MIM structures has been
investigated at room temperature. Extrapolation of ROFF and RON to 10 years shows, respectively, a
slight increase and stagnation of these resistance levels (Figure 1.23e). Both states exhibit therefore
good retention times, which is promising for data storage.
An exciting feature of the resistive switching based on Electric Mott Transition is that it
results from an avalanche breakdown, which is triggered above a threshold electric field (a few
kV/cm). Consequently, the SET voltages measured on single crystals (typically 30-50 V for 10-30 μm
inter-electrode distance) are much larger than those observed on thin films (down to 3.5 V for
100 nm). It enables a straightforward way to tune the SET voltage by changing the thickness of the
device. A SET voltage lower than 1 V is expected for sub-30 nm thick, thin films targeted in future
devices.
To summarize, the endurance of Mott-RAM devices is very promising compared to values
ranging from 103 to 107 cycles currently obtained in Flash technology [136]. The writing time and the
40
erasing time of 100 ns are favorable compared to characteristics achieved in Flash technology, i.e.
writing time of 1 μs and even much better than the typical erasing times of 10 ms. Besides, the
writing/erasing voltage envisioned in future devices in the 1 V range stands as a considerable
advantage when compared to the 12 V reported for Flash memories [119]. Among other ReRAM
emerging technologies, Mott insulator based ReRAM devices could be thus considered as up-and-
coming candidates to take over the Flash technology.
1.3.3.2 Neuromor phic applications : Mott neuron
During the last half-century, the tremendous development of computers has led to the
revolution of information technology. Nevertheless, the way computers store and process the
information has scarcely changed since their inception and relies on the concepts proposed by von
Neumann in the '40s. It is based on a clear separation between the memory and the processing unit,
and is extremely powerful in many cases, such as high-speed processing of massive data streams.
However, von Neumann computers are outperformed by the mammal brain in numerous data-
processing applications such as pattern recognition and data mining [129–133]. The brain is
organized with a very different architecture, based on a network of closely connected neurons and
synapses. Neuromorphic engineering aims to mimic brain-like behavior through the implementation
of artificial neural networks based on the combination of a large number of artificial neurons
massively interconnected by an even larger number of artificial synapses [130,131]. In most cases,
artificial neural networks are software-implemented in conventional hardware; they are programmed
in computers with standard architectures, i.e. on von Neumann architectures. In order to effectively
implement artificial neural networks directly in hardware, it is mandatory to develop artificial neurons
and synapses as downscalable components [132,133]. In that context, Mott insulators were recently
investigated to build up these new types of artificial Mott neurons.
Neurons in the mammal brain fulfill three main functions called Leaky Integrate and Fire.
The neurons receive electric pulses coming from other neurons, which contribute to enhance their
membrane potential (integrate function). The membrane potential relaxes between incoming pulses,
a feature called "Leaky". Above a threshold value of the membrane potential, the neuron emits an
action potential towards other neurons. The event is called "Fire". Lapicque already realized in 1907
that the neuron behaves like a parallel RC circuit before the Fire event. His model, called Leaky
Integrate and Fire (LIF), states that incoming spikes will charge the capacitor with electrical charges,
which will then "leak" through the resistor in parallel [134,135]. When the membrane potential
reaches a given threshold, the neuron fires an output electric spike.
Interestingly, artificial neural networks (ANN) using discrete spikes to compute and transmit
information (Spiking Neural Networks) are technologically appealing thanks to their high energy
efficiency. The building block of such ANNs is usually a Leaky Integrate and Fire artificial
neuron [34]. Due to their unconventional behavior under electric pulse, Mott insulators are
considered to realize LIF type of artificial neurons.
Recently it was demonstrated that the narrow gap Mott insulator behaves, under a train of
electric pulses, like a leaky integrator system [116,136]. Figure 1.24 shows that the response of the
Mott insulator GaTa4Se8 to trains of short pulses that varies in duration tON or separation tOFF. The
Mott insulator appears as an imperfect integrator of the signal. Indeed for a given VPULSE, raising tON
41
or tOFF induces opposite effects on the leaky integration and leads respectively to a decrease or an
increase of NFIRE (Figure 1.24b-c). NFIRE is the number of electric pulses required to induce a FIRE
event.
(a) (b)
tON ↗
tOFF tON tON tOFF
30s 15s 20s 30s
NFIRE = 6 NFIRE = 4
time (s) time (s)

(c)
tOFF ↗ ≈
tON = 15s
tOFF = 170s
≈
NFIRE = 8
≈
time (s)
Figure 1.24: Experimental resistive switching obtained by applying trains of short pulses of various tON and
tOFF on a GaTa4Se8 sample at 78 K. Using tON = 15 µs and tOFF = 30 µs leads to NFIRE =6 (a). Increasing
tON to 20 µs leads to a decrease of NFIRE to 4 (b). Increasing tOFF to 170 µs leads to an increase of NFIRE to
8 (c) [114].
42
Figure 1.25: Schematic representation of a biological neuron receiving input spikes from other neurons and
triggering an output action potential when the membrane potential reaches the threshold value (a). The
LIF artificial neurons based on Lapicke's model reproduces the evolution of the membrane potential thanks
to an RC circuit accumulating electrical charges (b). The Mott artificial neuron sketched in (c) reproduces
the LIF behavior thanks to the accumulation of correlated metallic sites and triggers an output signal when
a filamentary path is created within the sample [136].
Based on a modeling work, it was proposed that the accumulation of correlated metal sites
plays, in Mott insulators, a similar role as the accumulation of charges in the cellular membrane of
the neuron [116,136]. Figure 1.25 compares schematically the behavior of a biological neuron and
the LIF artificial neuron made with a Mott insulator. The Electric Mott transition opens a unique
opportunity to implement the Leaky Integrate and Fire (LIF) functionalities and build up a LIF
artificial neuron in a very simple manner. This is of great interest in the context of neurocomputing,
and first miniaturized devices were therefore realized using either the Mott insulators GaV4S8
(polished crystals) or (V0.95Cr0.05)2O3 (thin films) [114]. Figure 1.26 shows the device prepared from
the (V0.95Cr0.05)2O3 thin films. The electrical measurements performed on this sample with a train of
pulses are similar to those obtained on crystals and demonstrates that miniaturized LIF artificial
neurons may be downscalable and used in artificial neural networks.
43
Figure 1.26: (a) Picture of the miniaturized neuron made of a (V0.95Cr0.05)2O3. Three thin-film bands with
various width (80, 40, and 20 µm) were deposited on Si/SiO2 substrates holding TiN electrodes with various
inter-electrodes distances (2, 3, 5, 20, 40, 80, and 160 µm). (b) The schematic cross-sectional view of the
stacking of the device. (c) LIF behavior was observed on the (V0.95Cr0.05)2O3 planar device [114].
44
1.4 Insulator to metal transition driven by light in correlated

materials
The mechanism described in the previous sections for the Electric Mott Transition highlights
the critical role of hot carriers generated under electric fields in the Mott insulating state's breakdown.
As hot carriers may be easily produced by light, exploring the behavior of Mott insulators under
femtosecond (fs) laser pulses is of great interest in the framework of this thesis. This part will
therefore present general considerations about photoinduced phase transitions and will describe the
few pump-probe experiments done so far in Mott insulators.
1.4.1 Generality about the photoinduced phase transition explored by time-

resolved experimental methods
1.4.1.1 Pump-probe technique
The recent and impressive development of ultrafast laser technology allows probing
fundamental interaction and control of ultrashort lifetime states in the subpicosecond range.
Combined with new femtosecond x-ray sources, this gives access to intrinsic physical processes such
as fast phonon dynamics, charge motion, electron correlation, etc… . Almost all ultrafast laser
experimental methods are based on the pump-probe principle.
Figure 1.27: (a) Summary of pump-probe techniques, (b) Basic optical setup based on the pump-probe
principle [137].
45
The pump-probe concept explores the possibility to study dynamical processes in solids. The
pump pulse excites a process in the material, while the probe, similarly to a stroboscope, takes
snapshots of the processes dynamics. By changing the delay between pump and probe pulses, we
catch different shots of photoexcited phenomena (Figure 1.27b). The pump is usually a
subpicosecond ultrafast light pulse, which induces a new out-of-equilibrium state. However, as
summarized in Figure 1.27a, the variety of probe technics is vast. It includes in particular time-
resolved reflectivity or photoemission to probe electronic phenomena, and time-resolved x-ray or
electron diffraction to watch lattice dynamics [137].
1.4.1.2 Photoinduced phase transitions
The light control of phase transitions, i.e. photoinduced phase transition (PIPT), attracts
much attention from researchers [138–141]. Ultrafast laser excitation drives PIPT causing dramatic
changes in the properties of materials. Figure 1.28 schematically shows a general view of the
photoinduced phase transition. The optical pump can trigger excited states, which can lead to a
transition into a new metastable phase with a state different from the initial one. Usually, such a state
is volatile and disappearing after a finite time, however long enough to be detectable by ultrafast laser
techniques [138,141].
Figure 1.28: Schematic nature of photoinduced phase transition in terms of potential energy as the function
of the structure. Light excitation turns the system to a new phase [138,141].
Among various mechanisms that are at play to explain the PIPT, we will highlight a classical one:
photothermal related to the conversion of light energy into heat. More recently, a mechanism based
on the elastic strain wave was also described. These two mechanisms will be addressed in the
following section.
46
1.4.1.2.1 Thermal mechanism
The most common and easily accessible way to break an insulator state by light is simple
heating due to the energy transfer of photoexcited electrons to the lattice. Femtosecond laser
photoexcitation corresponds to an ultrafast temperature quench. The process dynamics involves two
TE (electronic) and TL (lattice) temperatures model. Before excitation, at equilibrium TE = TL (Figure
1.29a). Considering that the lattice specific heat CL far exceeds the electronic one CE, it is possible to
drive the electrons out-of-equilibrium with respect to the lattice. Ultrafast light pulse does this rapidly
enough so that the energy deposited in the electrons has not yet been transferred to the lattice [142].
In other words, the laser pulse preferentially excites the electrons, which suddenly changes electronic
temperature and takes the system into nonequilibrium (Figure 1.29 b). Then a fast process of
electron-electron thermalization will follow, leading within a few 10’s of fs to a homogeneous
electronic state with a temperature TE larger than the lattice temperature TL (Figure 1.29c). Through
the electron-phonon coupling mechanism, an energy exchange can occur between the electronic and
lattice sub-systems, and the material reaches equilibrium within a few picoseconds (TE = TL).
However, it remains hotter compared to the initial nonexcited state. Finally, within nanoseconds, the
system flows back to the initial temperature by thermal diffusion within the sample or into the
substrate [141].
Figure 1.29: Schematic description of photoinduced dynamics in the condensed matter a) Initially, the
system is at equilibrium (TE = TL). (b) Femtosecond laser excitation creates nonequilibrium electronic
distribution pulse. (c) Photoelectron thermalization through rapid electron-electron scattering results in
TE > TL and finally system equilibrates by transferring energy to the lattice through the electron-phonon
coupling mechanism [142].
47
The thermal phase transformation which follows may occur thanks to a nucleation and
growth mechanism (NGM). It has been claimed that photoinduced IMT’s in VO2 and V2O3 Mott
materials correspond to this mechanism, especially when the material is placed at a base temperature
only a few K below the thermally-driven transition temperature [143,144].
In some materials, a diffusional phase transformation happens in the case of temperature or
other external factors change. This phenomenon is not instantaneous and requires time in order to
reach a full phase transition. Among all mechanisms explaining these dynamics, the most frequently
invoked is related to the nucleation and growth of a new β phase within an initial α phase. In the
end, the transformation rate is affected by the physical size and proximity of neighboring parts of
the material. After a certain size, the nuclei would occupy contiguous volumes, and their growth will
slow down (Figure 1.30) [145].
Figure 1.30: Germination and growth process a) nucleation of β phase within α phase; b) Growth of β
phase domains; c) expanded volume of nuclei β: their growth is sterically hindered [145];.
NGM kinetics depends on different factors such as material, temperature, defects, etc.
Several models have been proposed to describe the transition dynamics analytically, such as the
Avrami equation:
(1.2)
where 𝑓𝛽 is a fraction of the transformed phase, K is constant related to temperature and growth
geometry; the exponent n depends on nucleation nature and geometry as well [141]. However, the
Avrami model is limited by several assumptions, such as a homogeneous growth or a random nuclei
distribution, which is not always valid in solid-state diffusional phase transformations [146].
1.4.1.2.2 Phase transition based on strain wave mechanism
In parallel to classical photothermal mechanisms discussed in the previous section, several

studies report other possible scenarios where optical pump power is insufficient to overcome the
system transition temperature. A recently proposed mechanism introduces the concept of light-
induced acoustical strain wave propagation, which transforms the system into a new phase, especially
in the favorable case where the volume of the new phase is very different from the pristine phase.
In this subsection, we will review the basic principles of the strain wave pathway of PIPT.
• the Thomsen model of coherent acoustical strain wave generation
48
Thanks to the pioneering work of Thomsen in the 80’s [147,148], it is now well established
that when a pump laser is sent to the surface of a sample, a depth-dependent elastic stress is set up.
With a light pulse of energy Q and of beam area A, the total energy W(z) deposited per unit volume
at a distance z from the surface is
(1.3)
where R is the reflectivity at the surface and  the light penetration depth. Then several mechanisms,
discussed a bit later, convert the absorbed energy into photoinduced mechanical stress proportional
to W(z)
𝜎𝑝ℎ𝑜𝑡𝑜𝑖𝑛𝑑𝑢𝑐𝑒𝑑 = −3𝐵𝛼𝑊(𝑧) (1.4)
where B is the bulk elastic modulus and  a coefficient quantifying the transformation of energy into
a stress. An explicit example of the physical meaning of  will be given below. As shown in Figure
1.31.a, this conversion of light energy into mechanical energy induces first stress proportional to the
light absorption and hence extending over the pump penetration depth . However, it is noteworthy
that the strain (deformation) profile in z is strictly zero just after the absorption, as depicted in Figure
1.31.e. For a sample elastically isotropic, the stress-strain equation writes :
1−𝜈
𝜎33 = 3𝐵 ( 𝜂 − 𝛼𝑊(𝑧)) (1.5)
1 + 𝜈 33
where  is the Poisson coefficient and 33 is the z-component of the elastic strain tensor. Solving the
time and z-dependent equation of elasticity yields [148,149].
1+𝜈 𝑄 1 1
𝜂33 (𝑧, 𝑡) = (1 − 𝑅) 𝛼 [𝑒 −𝑧⁄𝜉 (1 − 𝑒 −𝑣𝑠 𝑡⁄𝜉 ) − 𝑠𝑔𝑛(𝑧 − 𝑣𝑠 𝑡). 𝑒 −|𝑧−𝑣𝑠 𝑡|⁄𝜉 ] (1.6)
1−𝜈 𝐴𝜉 2 2
3𝐵 𝑄
𝜎33 (𝑧, 𝑡) = − (1 − 𝑅) 𝛼[𝑒 −(𝑧+𝑣𝑠 𝑡)⁄𝜉 + 𝑠𝑔𝑛(𝑧 − 𝑣𝑠 𝑡). 𝑒 −|𝑧−𝑣𝑠 𝑡|⁄𝜉 ] (1.7)
2 𝐴𝜉
1−𝜈 𝐵
where vS is the longitudinal sound velocity given by 𝑣𝑠 = √3 1+𝜈 𝜌 .  is the density.
As a response to the initial z-dependent stress profile, a strain wave is launched at the speed
of sound from the surface to the interior of the sample. The z and time profiles of the strain wave
are shown in Figure 1.31.e-h. Its overall shape corresponds to an exponential decay of the strain with
z, on top of which a transient bipolar strain travels with the wavefront (see Figure 1.31.g). At a long
time where the front wave is located well beyond the penetration depth , the photoinduced stress
is fully relaxed (see Figure 1.31.d), and the strain profile corresponds to a simple exponential decay
maximum at the surface and decreasing with the depth z (see Figure 1.31.f).
49
(a) stress(z,t) / stress(z=0, t=0) (e) Strain

( light absorption)
1
t=0 pump pump

z z
0 0
0 Distance from surface 0 Distance from surface
stress(z,t) / stress(z=0, t=0) Strain

(b) (f)
1
t < /vs 0
z
0
z
(c) stress(z,t) / stress(z=0, t=0) (g)

1
0.5
t > /vs 0
-0.5 Strain wave propagation

z at vsound
0 4ξ 8ξ
0 Distance from surface
stress(z,t) / stress(z=0, t=0) Strain

(d) (h)
1
( light absorption)
z z
t >> /vs 0
Figure 1.31 (a-d) Evolution of the stress profile 33 along the z-direction (z is the direction perpendicular
to the surface and directed towards the interior of the media ; z = 0 is the surface) and at different times
for a semi-infinite media which surface has been impacted by an ultrafast laser. Just after the absorption of
50
the laser light by the media, a stress is set up with a profile that follows the energy deposited at a distance
z from the surface.  represent the penetration depth of the laser and vS the sound velocity. The conversion
of light into mechanic elastic energy occurs according to a mechanism proposed by Thomsen (see text and
Refs. [147,148]. (e-h) Corresponding evolution of the strain profile along z at different times, propagating
at the longitudinal sound velocity. [150]. At the initial time (t=0), the elastic strain is zero [148].
In the equation presented above, the sign of the parameter  plays a key role. The nature of
the “static” strain let after the passage of the front wave (see Figure 1.31.h) can indeed be either a
tensile (if  > 0) or a compressive (if  < 0) strain. At the microscopic level, there are several
microscopic mechanisms involved in the light generation of acoustical waves in solids. We will
discuss here the most important of these precursor effects, namely the thermoelastic effect, related
to the heating of the lattice, and the deformation potential effect, related to the electronic distribution
change.
The thermoelastic mechanism occurs when the energy W(z) deposited in the sample is
transferred to the lattice and hence into heat, yielding a temperature rise ∆𝑇(𝑧) = 𝑊(𝑧)/𝐶 and
isotropic thermal stress given by
(1.8)
Here, C is the volumic specific heat and  is the linear thermal expansion coefficient. Such a
lattice heating comes from the energy transfer from excited carriers to incoherent phonons during
the nonradiative relaxation of excited electrons [150]. Such a situation typically appears for metals.
Within this mechanism, the coefficient  introduced above is simply the ratio  / C between the
linear thermal expansion and the specific heat. In most of the solids,  is positive, leading to a positive
 value and hence to negative photoinduced stress (see Equation (1.8)). If the thermoelastic
mechanism is dominant, the strain wave launched from the surface will hence let behind it a tensile
strain, i.e. a solid with an expanded volume.
The second important mechanism is the deformation potential (DP), which relates the
energy modification of the electronic distribution to the strain in the material. Changes in electronic
distribution indeed modify interatomic forces, leading to modifications of equilibrium lattice
positions, so that, after some time, the solids will be deformed. Similarly, when electron travel in a
crystal or electron-hole pairs is created, then local carrier density at a particular energy level is
modified (Figure 1.32.a). Consequently, it changes the strength of atomic bonds in a particular region,
which leads to atom displacement and hence to the creation or annihilation of a phonon [148,150].
51
Electronic energy modification

(a) electron
hole
Acoustic phonon creation (or annihilation)
Compression (for example)
(b)
carriers photo-excitation
Figure 1.32 Schematic description of deformation potential mechanism: (a) band structure description.
After photoexcitation, the electronic distribution would be disturbed due to electron-hole pairs creation,
for example, which modify interatomic interaction and change atomic positions. It leads an to elastic stress
generation and acoustic phonon creation or annihilation. (b) molecular (orbital configuration) description.
Light excitation promotes electrons (red circles) to bonding or antibonding levels of orbitals. It leads to
expansive or compressive stress [150].
In a more chemistry-oriented description, the deformation potential mechanism can be

expressed as follows: photoexcited carriers are promoted to empty electronic levels, which can
strengthen or weaken an existing chemical bond depending on the bonding or antibonding character
of the electronic level. This will tend to decrease or increase the corresponding interatomic distances,
respectively (Figure 1.32b) [150]. The nature of the strain induced by the deformation potential
mechanism can hence be tensile or compressive. This is typically the situation encountered in
materials with a bandgap (semiconductors, oxide materials, …). In this situation, the DP mechanism
will be important as soon as some electrons are photoexcited from the valence to the conduction
band. In classical semiconductors, the nature of electronic levels to which electrons are promoted
defines the type of lattice stress, compressive or tensile. For example, in the case of silicon, the
deformation potential is negative, and photoexcitation contracts the lattice. In contrast, for GaAs, it
52
leads to tensile deformation [150]. An interesting way to write the photoinduced stress
𝐷𝑃
𝑝ℎ𝑜𝑡𝑜𝑖𝑛𝑑𝑢𝑐𝑒𝑑 is:
𝜕𝐸𝐺 𝜕𝐸𝐺 (1.9)

𝐷𝑃
𝜎𝑝ℎ𝑜𝑡𝑜𝑖𝑛𝑑𝑢𝑐𝑒𝑑 = −𝛿𝑛𝑒 = −𝛿𝑛𝑒 𝐵
𝜕𝜂 𝜕𝑃
where ne is the concentration of excited electrons and B is the bulk modulus. Interestingly, this
equation relates to the evolution of the bandgap EG under pressure with the sign of the photo-
induced strain. In this mechanism, the coefficient  discussed above is proportional and of the same
sign as EG/P.
In the case of strongly correlated materials and particularly of Mott insulators, the nature of
the precursor effect, i.e. the signs of the photoinduced stress and of the coefficient , remains a fully
open question. The naïve use of Eq. (1.9) valid for classical semiconductors is indeed most probably
irrelevant. Photodoping in classical semiconductors indeed leads only to small shifts of the essentially
rigid valence and conduction bands. Conversely, promoting new carriers in a Mott insulator might
reorganize the electronic structure deeply. In a simple one-band picture, these modifications occur
over a very broad energy range as large as the Hubbard repulsion U [151]. As we will see at the end
of this chapter, this nature of the lattice response (compressive versus tensile) after a massive
excitation of carriers (by light, but also by an electric field) in Mott insulators is a central question of
this Ph.D. work.
• From coherent acoustical phonon generation to a new pathway to induce
phase transitions
The domain of coherent acoustical strain generation is now mature enough, and its
description for classes of materials such as metals or conventional semiconductor is now well
established. However, there are still widely opened questions in this field. In particular, it is tempting
to use the strain waves not only to probe the properties of the material close to its steady-state, but
to really impact and modify its nature and properties. An exciting idea along this line is to use the
propagation of the strain field to drive a true macroscopic phase transition. The systems where such
a mechanism could be efficient enclose the broad class of materials at the vicinity of a phase transition
driven by a moderate pressure in the GPa range. Recently, such an example of strain-induced
Insulator to Metal transition has been observed in the correlated Ti3O5 compound [152]. This
material undergoes an isostructural phase transition between the insulating β phase (see Figure 1.33.a)
and the metallic  phase (Figure 1.33.b) at 460K upon heating. The huge volume change from β to
 is +6.4%. This transition is related to large intracell structural reorganization. Interestingly, a
bistability domain of the β and  phases exists at room temperature, with an extremely low pressure
for converting the  into the β phase of 600 bar (0.06 GPa). C. Mariette and co-workers have
performed a time-resolved optical pump-XRD probe experiment on nanocrystalline samples with
72.5% β phase and 27.5% residual metastable  phase. It was shown the photoexcitation launches a
tensile strain wave, which transforms the β into the  phase (Figure 1.33.d,f,h). The dynamics of
transformation is entirely consistent with the Thomsen model of strain propagation [152].
53
Figure 1.33: Time-resolved XRD study of phase transformation dynamics under photoinduced strain
propagation. (a) and (b) Atomic structures of β and  phases. (d) time profiles of  phase unit cell volume
(dark orange plain circles) and microstrain parameter (light orange plain squares) (f) time profiles of β phase
unit cell volume (blue) and the fraction of  phase. (e) and (g) simulations based on Thomsen model of (d)
and (f), respectively. (h) schematically shows phase front propagation related to strain wave
propagation [152].
1.4.2 Light-driven Mott transition
Inducing insulator to metal transitions by light in Mott insulators is an emerging field of

research within the broadest domain of PIPT’s. Due to the numerous relevant degrees of freedom
governing the properties of strongly correlated materials (electronic, spin, orbital, and lattice), one
can expect that a very rich physics can emerge after light excitation. Numerous theoretical and
experimental papers report several photoinduced IMT in Mott insulators, with and without
crystallography symmetry breaking. Depending on experimental conditions and materials, different
mechanisms are suggested, including classical thermal control [140,153,143,154,155], strain
generation [156–158] and Zener tunneling phenomena [159], as well as other more exotic processes
like inverse IMT related to dynamical localization [160–162] or nonlinear phonons
excitation [163,164]. In this section, we will review the most illustrative studies of mechanisms of
PIPT in Mott materials.
54
1.4.2.1 Electronic mechanisms at play in photoinduced Mott

Insulator to Metal Transition
The classical thermal mechanism of PIPT, discussed in Section 1.4.1.2.1, relies on a rapid
energy transfer from the photoexcited electrons to the lattice, inducing heating effect with a long
characteristic time. Alternatively, there are many examples where the photoexcitation process (=
photodoping) alone can drive a material into a new state in a much shorter (~1ps) time scale [165–
167]. In Mott insulators, photodoping is expected to induce a metallic state very rapidly since two
types of free carriers (holes and electrons) are created. We have indeed seen in part 1.1.2 that doping,
usually with only one type of carriers, electrons or holes, is a classical perturbation allowing to induce
an insulator to correlated metal transition at equilibrium. Depending on photon energy with respect
to the gap value, ultrafast laser pulses can yield an efficient photodoping either by direct electron
excitation from LHB to UHB, or through a Zener tunneling or an electronic avalanche process. In
this section, we will review experimental studies of electronic light-driven Mott IMT.
1.4.2.1.1 Photodoping in Mott insulators
An example of a photoinduced Mott transition from a charge-transfer (CT) insulator to a

metallic state was reported in [Ni(chxn)2Br]Br2 by S. Iwai et al. [168]. CT insulators are conceptually
close to Mott insulators since Coulomb repulsion is also responsible for the insulating gap. However,
unlike in Mott insulators, a filled anionic band is inserted between the Lower and Upper Hubbard
Bands (UHB), and the charge transfer gap appears here between the anionic band and the UHB. [1].
Figure 1.34a shows the crystal structure of halogen-bridged Ni-chain compound
[Ni(chxn)2Br]Br2, which is a prototypical 1D CT insulator. The schematic electronic structure of
[Ni(chxn)2Br]Br2 typical of a CT insulator is sketched in Figure 1.34b: the Coulomb repulsion (U)
splits the Ni 3d eg level into UHB and LHB bands, with the LHB positioned below the anionic Br
4p band.
In order to induce IMT in the system, Prof. Iwai group performed an ultrafast pump-probe
experiment with a pump at 800 nm (1.5 eV) 130 fs laser pulse. The probe pulse was tunable within
a spectral region 0.1-2.5 eV in order to record the transient spectrum. Figure 1.34c shows several
transient reflectivity R spectra for pump power 0.6mJ/cm2 (0.5 photon per Ni site). An ultrafast
reflectivity increase after photodoping (td=0.1ps) was observed in the mid-IR region. The dielectric
constant ε2, calculated by Kramers-Kronig (KK) transformation of R spectrum, monotonously
increase with lowering photon energy in the mid-IR region (Figure 1.34c-bottom). It suggests the
closing of the optical gap and metallicity generation.
In order to check the validity of the KK transformation, S. Iwai and coworkers have
extracted the photoinduced change of the effective number of electrons Neff() by integration of ε2
(). Figure 1.34d show the increase of Neff() in the innergap region with photodoping fluence. It
appears that temporal profiles of Neff() and ΔR for two numbers of photons per nickel xph =0.012
and 0.5 are strikingly different. A fast 0.4 ps decay is observed for the highest fluence (xph=0.5), which
is absent at low fluence. The author suggests that this fast decay is characteristic of the laser-induced
55
metallic states. Overall, this work was the first observation of light-induced Mott transition in a
correlated insulator.
(a) (b) (c)
(d) (e)
Figure 1.34: (a) Crystal structure band (b) electronic structure of [Ni(chxn)2Br]Br2; (c) Top: Transient
Reflectivity spectra observed prior to photoexcitation (black dotted line) and at different time delays (td)
after laser pump at 300K. The excitation density xph was 0.5 photon per Ni site. Bottom: Spectrums of the
calculated imaginary part of the dielectric constant (ε2) from the the top part. (d) The effective number of
electrons change (ΔNeff) at for 1 eV and 0.2eV as a function of photodoping concentration xph at td=0.1ps;
(e) Time evolution of ΔNeff (0.5eV) for xph=0.5 (open circles) and xph=0.5 (clouse circles). Solid lines also
shof Reflectivity change for 0.15eV probe.(from Ref. [168])
Another interesting study involving photodoping of Mott insulators was reported by G.

Lantz et al. [169]. Thanks to a multi-techniques approach, including time-resolved photoelectron
spectroscopy, they observed collapsing of the Mott gap in a (V0.972Cr0.028)2O3-PI crystal after
photodoping Figure 1.34b. The authors proposed a photodoping mechanism specific to the V2O3
system. According to the band structure of (V1- xCrx)2O3 system discussed in Section 1.2.1, the energy
spacing between the eg and a1g levels is smaller for the paramagnetic metal (PM) than for the
paramagnetic Mott insulator (PI). As discussed in Section 1.2.1, the PM phase is therefore closer to
the degenerated multiband Mott insulator with N = 2 electrons spread over M = 3 orbitals, which
tends to favor a metallic ground state (see Figure 1. 35.a). Indeed, DOS calculations for different a1g
occupancies (Figure 1.34d) indicate that a metallic state is restored upon increasing the a1g population.
In this context, the authors propose that a photodoping increasing the occupancy of the a1g level
would lower its energy and hence favor IMT (Figure 1.34c). This mechanism is illustrated in Figure
1.34.a, highlighting the crossing of the IMT line pathway induced by a lowering of Δ𝑒𝑔𝜋−𝑎1𝑔 .
56
(a) (b)
PI Mott IMT line PM X4 X4 50 fs

X4 400 fs
(V1-xCrx)2O3 1
2 ps
∆T=20 K eq
0.25
0
V2 O 3
0 –0.4 –0.2 0.0 0.2
U/W 3 2 1 0
E–EF(eV)
60 na1g = 0.50
(c) (d)
50 na1g = 0.57
na1g = 0.67
3 UHB 3 3 40
DOS (a.u.)
E–EF(eV)
30
20
0 0 0
10
LHB 0
–0.5 0.0 0.5 1.0 1.5 2.0
–3 –3 –3
∆t= 0 fs ∆t= 50 fs ∆t= 400 fs E–EF (eV)
Figure 1. 35 (a) Generalized phase diagram of V2O3 system from Figure 1.11 highlighting photodoping
scenario of PI-PM transition; (b) Transient photoemission difference spectra of V2O3:Cr 2.8% at different
delays after pump excitation. Dashed line shows the spectral difference expected at equilibrium upon a
+20 K heating (see details in Ref. [169]) (c) Schematic view of the proposed mechanism involved in
photoexcitation of V2O3:Cr 2.8% and leading to a lowering of the a1g band [169]. (d) Calculated DOS for
different of a1g occupancies [169].
1.4.2.1.2 THz - light pulse driven Mott Insulators to metal transition
Before this PhD work in 2017, the possibility to induce an insulator to metal transition using
the electric field of a THz pulse was not established. Since 2017, two studies demonstrating THz
pulse-induced Mott insulator to metal transition were published in two canonical Mott insulators, a
strained thin film of -(ET)2X (X=Cu[N(CN)2]Br) [159] and V2O3 in its AFI phase [134].
-(ET)2Cu[N(CN)2]Br is metallic at ambient pressure and becomes superconducting below
10K. In order to open the Mott insulating gap in this material, a thin single crystal (0.6 µm thick) was
strained (negative pressure) by deposition on a diamond substrate. Figure 1.36 shows the schematic
phase diagram of this compound where the metal to insulator transition is caused by negative
pressure, as confirmed by optical density (OD) data. The strain-induced Mott gap was around
0.03 eV [159].
57
Figure 1.36 Phase diagram of the -(ET)2Cu[N(CN)2]Br system with paramagnetic metal (PM),
superconductor (SC), Paramagnetic Insulator (PI), and Antiferromagnetic Insulator (AFI) regions. (right)
Optical Density spectra in the unstrained metallic phase (ODM) and in the Mott insulator state (ODI)
achieved by negative pressure from the substrate. [159]
H. Yamakawa and coworkers performed time-resolved measurements on

- (ET)2Cu[N(CN)2]Br in the Mott insulator state by near IR (0.93 eV) and THz electric field (1 THz
 0.004 eV) pulses. They used a laser probe in the far-infrared region (0.12 eV). After a delay of 0.7
ps after the THz excitation, two main features appear in the transient differential ODTHz spectrum :
(i) the OD decreases at the energy of the broad absorption peak around 0.3 eV and (ii) a monotonous
increase of OD appears with energy decreasing below 0.3eV (Figure 1.37 a). According to Ref. [159],
this is the signature of a THz electric field-induced insulator to metal transition. The group of H.
Yamakava proposed a Zener quantum tunneling between the lower and upper Hubbard bands as a
mechanism of IMT. In the Zener mechanism, electric field bends the bands, which allow electron
tunneling between UHB and LHB and hence the doublon-holon pairs creation (Figure 1.37 g), which
probably led to a collapse of the Mott gap. A fit of the OD evolution versus THz electric field curve
indicates that the IMT occurs above a threshold electric field of 64 kV/cm. A similar dynamic was
also observed for the near-IR pump (Figure 1.37.b). However, a comparison of initial rise time
ODTHz (THz excitation) and ODph (near IR excitation) shows a 0.3ps delay at the near-IR pump
(Figure 1.37 c). The delay was attributed to additional carrier generation via an impact ionization
process, which might occur during the early de-excitation of electrons from their very high initial
energy (pump photons of 0.93 eV) towards the bottom of the conduction band. Finally, in addition
to the THz-driven insulator to metal transition (IMT), this work also confirms the possibility to
induce a purely electronic Mott IMT by a direct infrared photodoping (Figure 1.37 h).
58
Figure 1.37 THz and IR time-resolved pump-probe study of κ-(ET)2Cu[N(CN)]Br in Mott insulator state
(probe energy was 0.124eV, THz pump < 10meV, IR pump 0.93 eV): a) Transient absorption change
spectrum (ΔOD) 0.7ps after THz (ΔODTHz circles) and IR (ΔODph triangles) at 10K. The green ligne shows
scaled steady-state (ΔOD=ODM-ODI); b) ΔOD time dynamics for THz and IR excitations at different
fluences; d) and e) transient ΔOD dependence on THz electric field excitation at 0 ps and 0.7ps after
excitation; f) Comparison between ΔODTHz and ΔODph as a function of excitation fluence; g) schematics
of impulsive dielectric breakdown via Zener tunneling process following by doublon-holon pair creation;
c) ΔOD dynamics induced by THz (170 kV/cm-1) and IR (16 uJ/cm-2). (h) Schematic view of photocarrier
generations and band-structure changes induced by the near-IR pulse. The excess energy of photocarriers
initially produced is used to generate additional carriers, resulting in the collapse of the Mott gap. Extracted
from Ref. [159]
59
A recent study of Flavio Giorgianni et al. on the emblematic V2O3 system also evidenced the
possibility to induce a non-thermal THz field-induced IMT [157]. They performed a time-resolved
THz pump - short wave infrared (SWIR) probe experiment in a transmission geometry (Figure
1.38.a) on a V2O3 thin films starting from the low-temperature AFI phase. At T = 4K, the differential
transmission (-ΔT/T) shows a sub-picosecond switching followed by a slower transient step (Figure
1.38 b). The rapid rise is an experimental indication of purely electronic transition [134].
Figure 1.38 Time-resolved THz pump-Optical probe study of V2O3 canonical Mott insulator: a) schematic
overview of the experiment. The measurement was performed in transmission geometry. THz pump was
below 20 meV, and the optical probe was at 0.68 eV. b) Temporal evolution of the differential transmission
(-ΔT/T) at 4K for different =pump field. c) the same as (b) on a longer time scale at 8 MV cm-1. d) -ΔT/T
at zero delay time as a function of THz electric field strength; f) V2O3 bands distortion under THz field
allows interband electron tunneling resulting in a sub-ps Mott transition e) and g) Temperature evolution
of - ΔT/T dynamics. h) characteristic time tmax of the maximum of the derivative of ΔT/T as a function of
temperature. Depending on the temperature, Mott transition can arise according to two different regimes:
nonthermal tunneling breakdown and a thermal nucleation/growth mechanism. [157]
As in the previous case, the pump photon energy is much smaller than the bandgap, and the
proposed mechanism of transition could be quantum tunneling between 𝑒𝑔𝜋 and 𝑎1𝑔 bands due to its
bending under the THz field (Figure 1.38 f). Moreover, the fit of ΔT/T versus THz excitation field
shows the typical trend of the tunneling process under the electric field with the threshold value at
60
6.7 MV cm-1. However, depending on the base temperature, the dynamic of IMT is largely different
(Figure 1.38 e). Below 100K, the time dynamics remain constant, and in this range, switching remains
purely electronic (Figure 1.38.e-h). With increasing temperature, the system is driven into the
dissipative regime, and dynamics might be dominated by thermal nucleation and percolation to the
metallic phase. At the AFI – PM (175K) transition temperature and above, THz pump accelerates
the delocalized electrons. It increases the electronic temperature, which is associated with the
transmission increase followed by the heating process (Figure 1.38 e,g,h) [157].
1.4.2.2 Photother mal mechanism: examples in the cor related

system VO 2 and V 2 O 3
Several studies report photothermal insulator to metal transition in famous strongly

correlated vanadates materials VO2 and V2O3. Green arrows on VO2 and V2O3 phase diagrams
(Figure 1.39) indicate directions of transitions. In both cases, the transitions are followed by a
crystallographic symmetry breaking.
Figure 1.39: Phase diagrams and of VO2 (a) and (V1-xMx)2O3 (M = Cr, Ti). (b) systems. Green arrows
indicate the classical ways of thermal IMT.
Figure 1.40 summarized several studies demonstrating the creation of a metallic phase in
VO2 [170–172] and V2O3 [143] after optical pump photoexcitation either by reflectivity or optical
conductivity change [170]. According to the authors, the mechanism is a photo-thermal phase
transformation with nucleation and growth mechanism.
61
Figure 1.40: Time-resolved studies on correlated vanadates VO2 and V2O3. (a) optical pump- optical probe
experiment on VO2 for different laser fluences [170]. (b) optical pump - THz probe experiment on VO2 at
different temperatures at constant pump energy 3.8mJ/cm2 [171]. (c) optical pump - THz probe
experiment on VO2 [172]. (d) Optical pump (800 nm) - THz probe experiments on V2O3 thin films at
different temperatures. (e) Normalized curves of (d) with an indication of the initial (Ti) and final (Tf)
temperatures. (f) schematic view on 2D nucleation and growth process according to the authors, with a
radial metallic phase velocity proportional to the sound velocity of the material. (g) zoom on the initial time
of e) [143].
Both compounds undergo thermal IMT between the low temperature insulating and high-
temperature metallic phases. The proposed mechanism involves nucleation and growth
processes [171,173,174]. At equilibrium, metallic domains start to form upon heating as the transition
temperature is reached, and nucleation droplets might appear sporadically (Figure 1.41.a - 1). With
further temperature increase, the domains grow larger and start to form clusters (Figure 1.41a-2),
leading finally to percolating conducting paths throughout the film (Figure 1.41a-3) [174].
62
Figure 1.41: Schematic illustration of nucleation and growth thermal phase transition in vanadates [174] a)
and of IMT induced by a light pulse b).
According to the authors, the light-induced thermal increase might similarly initiate the
mesoscopic nucleation and growth of the metallic phase, leading to macroscopic stabilization of the
metal. The dynamics of photoinduced IMT in vanadates (Figure 1.40) implies different time scales.
In both compounds, the dynamics of the macroscopic phenomena depend on pump fluence (Figure
1.40 a-c) and initial sample temperature (Figure 1.40d,e,g). A 2D nucleation and growth mechanism
was also proposed in V2O3 film, with a radial growth velocity of the metallic patches close to the
sound velocity (Figure 1.40f) [143].
Remarkably the results reported for V2O3 were obtained in experimental conditions favoring
thermal effects. The pump photon energy (1.55eV) was much larger than the bandgap EG = 0.65 eV
(see Ref. [175]), which might lead to an energy transfer to the lattice and hence conversion into heat
during the early electron thermalization towards the bottom of the conduction band. In addition, the
authors' energetic calculations suggest a final temperature after photoexcitation above the transition
one. However, it cannot be excluded that other mechanisms are at play, such as the strain-wave
pathway discussed above. An obvious suggestion would be a change of experimental conditions in
order to limit thermal effects, such as (1) decreasing pump photon energy close to the gap in the
system and (2) choosing a low base temperature so that even a full conversion of the deposited laser
energy into heat will not be sufficient to reach the transition temperature thermally.
1.4.2.3 Elastic strain effect in photoexcited Mott Insulator
In part 1.4.1.2.2 of this chapter, the concept of light-induced strain wave generation and
propagation was introduced. The strongly correlated materials offer an interesting playground to test
the possibility to induce transitions driven by a light-induced strain wave. Their rich phase diagrams
often allows the possibilities to drive phase transitions by the application of pressure or equivalently
by inducing a compressive strain. This is particularly true for the emblematic V2O3 system, where
pressure allows inducing PI to PM as well as AFI to PM transitions. Interestingly, it was shown by
Liu et al. [156] that a light-induced strain wave was able to drive phase transitions in this system.
Starting from V2O3 in its PM phase, their optical pump- THz probe measurements have revealed a
tensile strain-induced modulation of spectral weight towards the PI phase (see Figure 1.42). Additional
63
experiments demonstrate an even more complex pathway in the phase diagram : starting from the
AFI phase, the system first crosses the AFI to PM transition due to the initial laser heating and then,
due to thermoelastic tensile strain, moves towards the PI phase [156].
Figure 1.42 Results of the optical pump- THz probe spectroscopy on V2O3 departing from low-
temperature AFI insulating phase at 143 K (a) phase diagram of the V2O3 system. Red arrows indicate the
thermal component (rapid electron-phonon relaxations) of the laser effect, which switches the system to
the paramagnetic metallic (PM) phase. Then elastic strain modulation dynamically drives V2O3 across the
PM-PI transition (blue arrows). (b) transient THz conductivity at 143K with optical pump 4mJ/cm2 (blue
arrows indicates THz conductivity reduction, which is purely related to tensile strain). (c) Peak conductivity
reduction, at 50 ps after the pump pulse, for different laser fluences versus final temperature.
In this example, the tensile character of the strain starting from the metallic PM state is not
a surprise. As discussed in Refs. [148–150], the laser energy transferred to electrons in a metal is
rapidly converted into heat and, if the coefficient of thermal expansion is positive as for the vast
majority of materials, into a tensile strain wave where the final volume is expanded. The concept of
“strain wave pathway” as a new tool to promote phase transitions is quite recent. In the scarce in-
depth studies published so far in a molecular spin crossover system [176] and in the charge ordered
insulator Ti3O5 [152], the light-induced strain wave always let behind it a tensile state, i.e. an expanded
volume. In this context, the idea to apply the strain wave pathway to Mott insulators as a novel
strategy to induce insulator to metal transition (IMT) is obviously appealing. However, as the Mott
IMT is accompanied by a volume drop, this idea will be relevant only if the “deformation potential”
mechanism (see part 1.4.1.2.2) in Mott insulators yields a compressive stress and finally after a delay a
compressive strain with a contracted unit cell volume.
This introductory part about light-induced coherent acoustical phonons and strain wave
generation might seem at first glance quite far from the central question of this thesis, which is to
understand the mechanism of the Electric Mott transition. There is however, a deep relation, which
is the nature of the precursor stress (tensile or compressive) appearing in Mott insulator after an
electronic excitation caused either by light or electric field. This is up to now still an open question.
Evidencing the propagation of a light-induced compressive strain wave in a Mott insulator would
directly prove that a lattice compression occurs in response to an electronic excitation, whatever the
64
origin of the excitation, light or electric field. This would be a major step in the understanding of the
mechanism of the Electric Mott Transition.
65
1.5 Open questions
The basic concepts about classical Mott insulator to metal transitions induced at equilibrium
by chemical doping and physical pressure are now quite well established. In this field, the current
frontiers of knowledge is now shifted towards Mott insulator to metal transitions occurring under
out-of-equilibrium conditions. From a theoretical point of view, this is a true challenge since it
requires solving a strongly correlated many-body problem out-of-equilibrium. Recent advances have
been achieved in this field using non-equilibrium Dynamical Mean-Field Theory but it is commonly
admitted that only a small fraction of the out-of-equilibrium physics in strongly correlated systems
is currently understood [102].
This is in this context that an out-of-equilibrium Electric Mott transition, i.e. an insulator to
metal transition occurring under electric field in Mott insulators, was discovered a decade ago.
Interestingly, this discovery of this property paves the way to innovative applications in
microelectronics and in artificial intelligence, such as a Mott memory and a Mott artificial neuron.
The physical model proposed by our group for this transition suggests that the electric field launches
an avalanche carrier multiplication phenomena leading to the creation of a conducting filamentary
path. However, many issues remain unanswered concerning the exact nature of the created metallic
phase and the reason why the metallic state can be stabilized permanently. On the other hand, the
carrier multiplication phenomena scenario leading to metallicity suggests that, beyond electric field,
photo-doping might also lead to an insulator to metal transition. A few experimental studies
performed on Mott insulators support this hypothesis. It suggests that electric and light pulses might
lead to similar phenomena. This Ph.D. thesis aims to achieve a better understanding of the Electric
Mott transition mechanisms by addressing the following questions:
• What is the nature of the metallic phase created after a non-volatile Electric Mott
transition (EMT) and how does it compare with that induced by the bandwidth-
controlled insulator to metal transition?
• Can we learn about the EMT mechanism by coupling electric and light pulses and is
it possible to realize an optoelectronic device?
• Can we learn about the role of the lattice response in the EMT by driving the
insulator to metal transition by light and following it by fs pump-probe experiments?
To answer these questions, the next chapter is devoted to the characterization at the local
scale of the metallic phase generated thanks to an Electric Mott transition. The third chapter will be
devoted to an electro-optic neuron's implementation and to pump-probe experiments in Mott
insulators.
66
CHAPTER 2 - Nature of metallic phases created at and out-of equilibrium in Mott insulators
Chapter 2
NATURE OF METALLIC PHASES CREATED AT AND OUT-OF
EQUILIBRIUM IN MOTT INSULATORS
***
This chapter addresses the issue of the nature of the metallic filamentary path created
after a non-volatile Electric Mott Transition (EMT). It also brings up the question
of the response of the lattice to application of an the electric field in a Mott insulator.
The first part of the chapter reports on a Raman study performed under pressure in
the Cr doped V2O3 Mott insulator to enable a comparison with the phase created
under electric field. This study demonstrates a sharp lattice softening at the pressure-
induced IMT. A rationalization of this behavior is proposed in the framework of the
compressible Hubbard model, an approach capable of accounting for the observed
violation at the Mott transition of the usual laws linking pressure, volume and lattice
stiffness. The second part of the chapter presents the Raman and X-ray study of the
conductive filamentary structure created at the EMT on Cr doped V2O3 single-crystal
and thin-film devices. µ-XRD and µ-Raman mapping demonstrate that the
conducting filamentary path consists of an isostructural compressed metallic
(V0.95Cr0.05)2O3 phase. This one shows a strong analogy with the metallic phase
induced under pressure.
***
In the previous chapter, we have discussed different aspects of Electric Mott Transitions
that enabled the realization of breakthrough microelectronic devices, such as a new class of non-
volatile memories. However, despite some significant advances which suggest that the EMT is
triggered by an electronic avalanche effect, some controversies exist concerning the nature of the
metallic phase, which persists after the application of the electric pulse and the exact role of the
electronic correlation in the transition [177,178].
Recent researches of the IMN group on a GaTa4Se8 crystal permanently switched due to
non-volatile Electric Mott Transition indirectly suggest the creation of compressed nanodomains. In
other words, the created metallic phase could share some similarities with the pressure-induced
correlated metal and indirectly involve a bandwidth-controlled IMT (see Chapter 1) [9]. A first-order
volume contraction is indeed theoretically expected at thebandwidth-controlled Mott insulator to
metal transition, as a consequence of electron-lattice coupling. The response of the lattice at the Mott
IMT, compression [176] but also softening [177], will be discussed in this Chapter.
A part of this Ph.D. work aimed to clarify the nature of the phase created during EMT. The
canonical V2O3 system is a good candaidate fo such a study, since the nature of the metallic phase
creatd after a non-volatile EMT can be easily compared with that induces under pressure. Moreover,
we have the possibility to synthesize in the laboratory both single crystals and thin films of
(V1 - xCrx)2O3. We will see further that is is a strong asset since it allows choosing the type of sample,
crystal or film, better suited for a given experimental technique. This chapter summarizes several
67
experiments aiming at clarifying the nature of the metallic phase induced by EMT. We performed
first Raman spectroscopy on Cr doped V2O3 Mott insulators as a function of pressure and Cr
content, in order to clarify the signature of the metallic phase induced by an IMT at equilibrium.
Then we carried out a comprehensive study of metallic induced out-of-equilibrium in Cr doped V2O3
by electronic transport, conducting-atomic force microscopy, Raman mapping and µ-X-Ray
diffraction measurements. Finally, in collaboration with the Institute of Physics of Rennes within the
Ph.D. work of Yevheniia Chernukha [179], we shortly mention our contribution to the discovery of
a new pathway to induce a non-volatile MIT, based on a shock wave technique [180].
68
2.1 Anomaly of standard laws linking pressure and volume a t

the Mott transition in (V 1 - x Cr x ) 2 O 3 : rationalization by the
compressible Hubbard model
Three-quarters of a century ago, F. D. Murnaghan and A.F. Birch proposed an equation of

states reflecting a seemingly universal behavior of solid matter under pressure [181,182] which can
be formulated as follows : the more you compress a solid, the harder it is to compress it further.
𝑑𝑃
Experimentally an increase of bulk elastic modulus B (defined as 𝐵 = −𝑉 𝑑𝑉, where P is the applied
pressure and V the substance volume) under pressure is indeed widely observed. This behavior
originates in the spatial dependence of the interatomic potential, which is the sum of two terms, one
attractive and the other repulsive, leading overall to a dependence B  d-4 [183]. Applying a pressure
tends to decrease the inter-atomic distance d and hence strengthens the bulk elastic modulus B.
Such a predicted lattice hardening under pressure strongly impacts many properties of solids.
To name a few, this bulk modulus strengthening, or equivalently the weakening of the bulk
compressibility K = 1/B, leads to an enhancement of the sound velocity, a hardening of the atomic
vibration (phonons) modes, which lead to strong changes of electronic properties. However, the
prediction of a continuous increase of the bulk elastic modulus B with pressure rests on a crucial
assumption: the two main contributions to B, i.e. the lattice one related to the general evolution of
interatomic potential discussed above, but also the electronic one associated to active electrons close
to the Fermi energy, should not display any singularity under pressure. In this context, strongly
correlated materials are good candidates to break the Birch-Murnaghan equation of states since many
electronic singularities, such as insulator to metal transitions, occur in such systems under pressure.
The bandwidth-controlled Mott Insulator to Metal Transition (IMT) should be a prominent
illustration of such a situation, since the IMT corresponds primarily to an electronic singularity.
As discussed in Section 1.1.1, most of Mott insulator properties, including the IMT, are
essentially described by a purely electronic Hamiltonian, the Hubbard model, which essentially
involve competition between on-site electron-electron repulsion and electronic delocalization. In the
ideal implementation of the Hubbard Hamiltonian, existing e.g. in the repulsively interacting gas of
ultracold fermionic atoms (i.e. atoms with a half-integer nuclear spin) forming a Mott insulator state,
this purely electronic description is fully appropriate. However, an important question is whether
this simplified picture can adequately describe the properties of a solid where the electronic system
is coupled to a deformable lattice. To date, the impact of the elastic degrees of freedom on the Mott
transition has been loosely studied experimentally [184,185]. This is somehow surprising, since the
control of the Mott transition by pressure, through a lattice contraction promoting wave-function
overlap and hence metallicity, directly prove the relevance of the lattice for the Mott transition.
Conversely, one can expect that the electronic system exerts an internal pressure on the elastic system
to which the crystal lattice responds [185].
The emblematic (V1-xCrx)2O3 system represents a well-studied example of a strongly
correlated Mott insulator. It phase diagram follows closely that predicted by DMFT for Mott
insulators (see Section 1.2.1). In the (V1-xCrx)2O3 system, a first-order insulator to metal transition
line, ended by a second-order critical endpoint at TC, separates the paramagnetic insulating (PI) and
metallic (PM) states. So far, the experimental studies of lattice effects on the IMT have mainly
69
focused on temperatures at or above the critical endpoint TC = 440 K [184]. The first-order part of
the Mott line, extending over the broad temperature range 180 – 440 K, has been essentially
overlooked.
In this section, we will show that the compressible Hubbard model is a theoretical framework
well suited to capture the consequences of electron-phonon coupling at the Mott insulator to metal
transition. In addition, we will show that Raman spectroscopy is an appropriate experimental
technique to check the prediction of the compressible Hubbard model. Our Raman study under
pressure of the emblematic Mott insulator (V1 - xCrx)2O3 indeed indicates that a lattice softening
(softening of the bulk modulus B) occurs both:
(1) within the Mott insulator state upon approaching the first-order Mott line where the
frequency of phonons modes slowly decreases as pressure increases;
(2) at the isostructural first-order Mott insulator to metal transition where a sudden drop of
optical phonons frequency is observed.
This lattice softening effect at the vicinity of the IMT can be rationalized based on the
prediction of the compressive Hubbard model. Overall, this work unveils a clear breakdown of the
Birch-Murnaghan equation of states since the observed behavior could be summarized as follows
“the more you compress a Mott insulator close to the IMT, the easier it is to compress it further”.
2.1.1 Main prediction of the compressible Hubbard model
According to the pioneering theoretical works of Majumdar et al. [186,187] and of the group
of A. Georges [21], the simplest possible approach to describe the consequences of the electron-
lattice coupling across the Mott Insulator to Metal Transition (IMT) is the compressible Hubbard
model. We will see that this strategy provides the conceptual framework allowing to understand why
the unit cell volume drops and the lattice softens at the Mott IMT. In their work, they assume that
the dependence of the free energy with the unit cell volume V (or equivalently with the strain
(V - V0) / V0) writes as :
1 (𝑉 − 𝑉0 )2
𝐹 = 𝐹0 − 𝑃0 (𝑉 − 𝑉0 ) + 𝐵0 + 𝐹𝑒𝑙 [𝐷(𝑉)] (2.1)
2 𝑉0
Here, the last term Fel is the contribution of the electronic degrees of freedom that are active
through the transition. To simplify the approach, Fel is taken as the free energy of a single band, half-
filled Hubbard model with a half-bandwidth D(V) depending on the unit-cell volume V. The first
three terms arise from expanding the free energy due to other degrees of freedom about a reference
cell volume V0. P0 and B0 are the corresponding (reference) pressure and the bulk elastic modulus.
As the volume variation under pressure is rather small, the conventional exponential parametrization
of D(V) can be linearized: . Three important quantities can
now be estimated, the kinetic energy of delocalized electrons 𝒯𝑒𝑙 , the “electronic response function”
el and the total bulk modulus Btot :
70
𝜕𝐹𝑒𝑙
𝒯𝑒𝑙 (𝑇, 𝐷(𝑉), 𝑈) = (2.2)
𝜕𝐷
𝜕 2 𝐹𝑒𝑙 𝜕𝒯𝑒𝑙
𝜒𝑒𝑙 (𝑇, 𝐷(𝑉), 𝑈) = − = − (2.3)
𝜕𝐷2 𝜕𝐷
𝐵𝑡𝑜𝑡 𝐵0 𝛾𝐷0 2
= −( ) 𝜒𝑒𝑙 (2.4)
𝑉 𝑉0 𝑉0
We note here that Eq.(2.4) predicts that the bulk modulus B is the sum of two terms, one
linked to the lattice and the other of electronic origin. Calculating such quantities by the Dynamical
Mean-Field Theory allows predicting several physical quantities directly comparable with
experiments:
• in the purely electronic Hubbard model, the critical endpoint 𝑇𝑐𝑒𝑙 of the Mott line, already
introduced in Chapter 1, is defined as the temperature for which the “electronic response
function” el (shown in Figure 2.1.c) diverges. The critical endpoint definition is different in
the compressible Hubbard model. It corresponds to the temperature at which the electronic
𝛾𝐷 2
contribution − ( 𝑉 0) 𝜒𝑒𝑙 to the total bulk modulus counterbalances exactly the lattice one
0
𝐵0
, leading to a bulk modulus Btot equal to zero. A quick look at Eq.(2.4) indicates that this
𝑉0
occurs when el is large enough but not divergent. As illustrated in Figure 2.1.a, a
consequence is that the critical endpoint in the compressible Hubbard model TC will be larger
than its purely electronic counterpart .
• a direct consequence is that below the critical endpoint TC and close to 𝑇𝑐𝑒𝑙 , the bulk modulus
Btot might become negative due to the divergence of el. It is an unstable condition which the
physical system avoids by generating a first-order IMT accompanied by a discontinuous
volume change covering the unphysical region of the phase diagram where Btot < 0, as shown
in the inset of Figure 2.1.a [186]. The volume change at the IMT for T well below TC is [21]:
𝛾𝐷0 (2.5)
Δ𝑉 ≈ (𝒯 − 𝒯𝑒𝑙,𝑖𝑛𝑠 )
2𝐵0 𝑒𝑙,𝑚𝑒𝑡
Interestingly, V mainly depends on the gain of electronic kinetic energy between the PI
(𝒯𝑒𝑙,𝑖𝑛𝑠 ) and the PM phase (𝒯𝑒𝑙,𝑚𝑒𝑡 ) shown in Figure 2.1.b. Physically, it means that the unit
cell volume drop at the PI ➔ PM transition originates from a non-negligible participation
of the conducting electrons in the cohesive energy in the PM phase.
71
(a) (b) (c)
Electronic response function

Electronic kinetic energy
-0.1
PM phase
-0.2 1
(1) metastable metal
(2) metastable insulator
PI phase
(3) actual transition
. . 1 1.1 1.
(d) (e) (P - P ) / B
1 c o
-0,013 -0,011 -0,009 -0,007
265 1
mode (cm )
-1
Prediction T = 0.87 Tc
Normalized frequency ( / )

. 5
Normalized square root of the
260
. 5
total bulk modulus
. 1g
Inflexion point A
255
. 5 0,95
. 5 250
. . 1
0
245
0 1 2
. 5
. . 15 . 1 . 5 . 5 . 1 . 15 .
Pressure (GPa)
Figure 2.1 : (a) Temperature T – pressure P phase diagram computed by Majumdar et al. [186]. Inset:
discontinuity in bandwidth across the transition. The solid line indicates the limits of metastability in the
purely electronic scenario, while the dashed line denotes the boundary of coexistence in the presence of
lattice coupling [186]. (b) Evolution of the electronic kinetic energy 𝒯𝑒𝑙 [186]. (c) Variation of the
electronic response function el with the bandwidth D for various temperatures. The parameters are
U/D0 = 2.492 and T/D0 = 0.060, 0.055, 0.050 from right to left [21]. (d) Evolution of the normalized
square root of the total bulk elastic modulus √𝐵𝑡𝑜𝑡 ⁄𝐵0, with pressure normalized the reference bulk eleastic
modulus P/B0 at various temperatures. √𝐵𝑡𝑜𝑡 ⁄𝐵0 is proportional to /0, i.e. to Raman frequencies 
normalized to values 0 far from the IMT [21]. (e) Comparison of experimental (blue curve) evolutions of
the A1g Raman mode under pressure with the prediction of the compressible Hubbard model shown in (d)
for T = 0.87 TC ~ 297 K (TC for V2O3:Cr 3.8% is 380 K, see Figure 1.7) .
• finally, it is possible, from Eq.(2.4), to calculate the evolution of the square root of the
normalized total bulk modulus √𝐵𝑡𝑜𝑡 ⁄𝐵0 with pressure. Indeed, the electronic part of the
bulk modulus B corresponds to the gain of electronic kinetic energy Tel upon increasing
bandwidth or, almost similarly, upon increasing pressure. Figure 2.1.b clearly shows that
applying pressure enhances efficiently the electronic kinetic energy Tel in the PM state,
whereas it let Tel essentially unchanged in the PI state as long as a Mott gap persists. As a
consequence, the predicted evolution of the elastic constants with pressure P shown in
Figure 2.1.d is strongly asymmetric. At temperature below the critical endpoint TC (see curves
at T = 0.57, 0.65 and 0.87TC in Figure 2.1.d) where the transition is first-order, a small
decrease of √𝐵𝑡𝑜𝑡 ⁄𝐵0 with pressure is predicted on the Mott insulator side, before a sudden
72
drop at the critical pressure leading to the IMT. At higher pressure in the metallic side, the
recovery of √𝐵𝑡𝑜𝑡 ⁄𝐵0 towards its unperturbed value is slow and loosely pressure-dependent.
Overall, this behavior reflects the asymmetry in the electronic response function shown in
Figure 2.1.c.
2.1.2 Consequences on Raman spectrum
Interestingly, the prediction of softening of the bulk elastic modulus at the vicinity of the
IMT Mott line can affect other physical quantities than just the sound velocity. On might indeed
expect that optical phonons, usually detected through Raman and IR spectroscopies, will be
impacted. The frequencies of these atomic vibration modes in extended solids are indeed tightly
connected, as the speed of sound, to the elastic constants. The simplest example is the classical
diatomic chain (2 atoms of mass m1 and m2 with an interatomic distance a, linked by a spring constant
C) where the Raman frequency is 𝑀 and the speed of sound is
) [185]. Here M is defined as . These two expressions lead to an

interesting consequence: in an experiment where the spring constant is modified by an external
stimulus (in our case, the application of an external pressure P), there is an equality between the ratios
of Raman frequency, sound velocity, and the square root of the spring constant:
𝜔𝑅𝑎𝑚𝑎𝑛 (𝑃) 𝑣𝑠𝑜𝑢𝑛𝑑 (𝑃) 𝐶(𝑃) (2.6)

= =√
𝜔𝑅𝑎𝑚𝑎𝑛 (𝑃 = 0) 𝑣𝑠𝑜𝑢𝑛𝑑 (𝑃 = 0) 𝐶(𝑃 = 0)
In this illustrative example, the simple knowledge of the variation of the spring constant C
with P allows predicting the evolution of Raman mode frequencies.
In a three-dimensional solid, this relation still holds for the sound velocity providing that one
replaces the spring constant by the bulk elastic modulus B. For the frequencies of Raman modes,
this equality is no longer generally true for V2O3, since the frequencies of the seven Raman modes
are not determined uniquely by the bulk modulus B. However, if a global phenomenon, such as an
insulator to metal transition, impacts all the elastic constants and hence the bulk modulus B, one
might expect that the evolution of B with pressure will affect more or less all the frequencies of
Raman modes and hence that the relation:
𝜔𝑅𝑎𝑚𝑎𝑛 (𝑃) 𝐵(𝑃)

=√ (2.7)
𝜔𝑅𝑎𝑚𝑎𝑛 (𝑃 = 0) 𝐵(𝑃 = 0)
will be still valid.

Overall, the compressible Hubbard model predicts, for temperature T well below TC, a
lowering of the Raman modes frequencies with pressure upon approaching the IMT followed by a
drop at the PI ➔ PM transition. This is an important result since it predicts, at the vicinity of the
Mott insulator to metal transition, a violation of the Birch-Murnaghan equation of states.
73
2.1.3 Raman Spectroscopy study of optical phonon modes at the Mott

insulator to metal transition in the (V1-xCrx)2O3 system
The previous discussion about the impact of the predicted elastic anomaly of the optical
phonon modes at the vicinity of Mott IMT line requires experimental confirmation. This section will
discuss our Raman under pressure experiment on the (V1-xCrx)2O3 system aiming to clarify this point.
2.1.3.1 Generalities about Raman spectra in V 2 O 3 (PM) and (V 1 -

x Cr x ) 2 O 3
Before going into the study of the Mott insulator to metal transition of (V1-xCrx)2O3 under
pressure by Raman spectroscopy, a brief introduction about the main features of the Raman spectra
of the V2O3 system is necessary. Both (V1-xCrx)2O3 in the PI phase (hereafter referred to as
(V1- xCrx)2O3-PI) and V2O3 in the PM phase (V2O3-PM) crystallize in the 𝑅3̅𝑐 corundum
crystallographic structure [32] with the D3d point group. Seven Raman-active modes are expected,
two A1g modes and five Eg modes. Figure 2.2 shows that five Raman modes can be clearly identified
in (V0.95Cr0.05)2O3–PI as well as in V2O3-PM, which can be assigned to 2 A1g and 3 Eg modes thanks
to the modeling work of Yang and Sladek [188]. As expected, the displacement vectors of the low
energy modes ( < 400 cm-1) mainly involve the heavy vanadium atoms, whereas the higher
frequency ones ( > 400 cm-1) implicate the lighter oxygen atoms. A quick comparison between the
Raman spectra of (V0.95Cr0.05)2O3-PI and V2O3-PM indicates that the presence of the same Raman
modes in both phases. The main differences come from a noticeable change of the Eg mode intensity
around 210 cm-1, drastically enhanced in metallic V2O3. In addition, the Raman modes frequencies
are systematically lower in V2O3-PM compared to those of (V1 - xCrx)2O3-PI. We will address this issue
later in this work.
74
intensity (arb. unit)

Raman scattering
A1g
Eg
A1g Eg
Eg
V2O3
(V0.95Cr0.05)2O3
200 300 400 500 600

 (cm-1)
Figure 2.2 Typical Raman scattering spectra of V2O3 (PM) and (V0.95Cr0.05)2O3 (PI). The main displacement
vectors of vanadium and oxygen atoms (green arrows) are represented within a relevant fragment of the
structure, the two face-sharing VO6 octahedra, containing the V-V dimer oriented along the c-axis.
2.1.3.2 Raman study of the Mott insul ator to metal transition in

(V 0 . 9 6 2 Cr 0 . 0 3 8 ) 2 O 3 under pressure
Despite more than fifty years of active studies on the emblematic Mott insulator
(V1 - xCrx)2O3, the Mott insulator to metal transition under pressure has never been studied by Raman
spectroscopy in this compound. We have hence initiated a Raman study under pressure and at room
temperature on a (V0.962Cr0.038)2O3-PI single crystal, using a Diamond Anvil Cell (see inset of Figure
2.3.a) with methanol as the pressure transmitting medium (see details in ANNEX 3-A).
75
(a) (b)
V2O3:Cr3.8%,
Raman scattering intensity (a.u.)

T = 300 K
Raman scattering intensity (a.u.)
0 GPa
0.50 GPa 0 GPa
0.80 GPa V2O3:Cr3.8%,
0.83 GPa
0.97 GPa 1.78 GPa
1.16 GPa V2O3
1.22 GPa
1.39 GPa T = 300 K
1.50 GPa
1.78 GPa
(V0.962Cr0.038)2O3
150 200 250 300 350
200 250 300
 (cm )
-1
 (cm )
-1
(c) (P - P ) / B (d)
C 0
-0,004 0 0,004
265
, inflexion point (cm )
-1
308
260
_E (cm-1)
304
255
g
300
A1g mode
250
296

245
0 0,5 1 1,5 2 0 0,5 1 1,5 2
Pressure (GPa) P (GPa)
(e) (f)
299 1 . 5
(V0.972Cr0.028)2O3
3
V (Å )
298
1 . 5
V = - 3.2 Å
297 3
=> V/V = - 1.07 %

V (Å3)
296 1 . 5
5.
295
294 .
293
.
0 0.5 1 1.5 2 2.5
5 1 15
P (GPa)
Figure 2.3 (a) Evolution of Raman spectra with pressure of a (V0.962Cr0.038)2O3-PI single crystal at room
temperature below 350 cm-1. Inset: scanning electronic microscopy image of the (V0.962Cr0.038)2O3–PI single
crystal used for the Raman study under pressure. (b) Comparison of Raman spectra in the metallic phases
of (V0.968Cr0.032)2O3 under a pressure of 1.78 GPa and of pure V2O3 at room temperature. (c) Evolution of
the frequency of the low energy A1g mode – more precisely of inflection point located on the high energy
side of the Raman peak – with pressure. (d) evolution of the Eg mode around 300 cm-1 with pressure. (e)
Evolution of the unit cell volume under pressure calculated form (f). (f) Unit cell parameters a and c versus
pressure of a powder of (V1-xCrx)2O3 (x  0.028), extracted from Ref. [189].
76
Due to the additional contribution of the Diamond Anvil Cell and of methanol at high
energy, reliable measurements of Raman spectra under pressure were possible only below 400 cm-1
(see ANNEX 3-A and Figure A 3-1.d). In this energy range, the main effect of pressure on
(V0.962Cr0.038)2O3 shown in Figure 2.3-a is twofold: a substantial increase of intensity for the Eg mode
around 210 cm-1 and a shift of the two other Raman modes (A1g near 250 cm-1 and Eg around 300
cm-1) towards lower energy. A comparison between (V0.962Cr0.038)2O3 at the highest pressure achieved
in this study (1.78 GPa) and V2O3–PM at room pressure and temperature indicates that the
frequencies of the Raman mode are very similar (Figure 2.3 b). This is a first hint suggesting that a
metallic state is reached at 1.78 GPa. A closer look at the evolution of the A1g mode around 250 cm- 1
shows that the frequency decrease of this mode with pressure is monotonous but displays a sudden
drop of  13 cm-1 around 1.2 GPa (Figure 2.3 c). Figure 2.3 d shows that the same trend occurs for
the Eg mode around 300 cm-1: even if the bad signal over noise ratio does not allow to conclude
about the softening of this mode between 0 and 1.2 GPa, a clear frequency drop also appears around
1.2 GPa. According to the pioneering studies of the Bell lab group, this value corresponds exactly to
the critical pressure necessary to induce the Mott insulator to metal transition for a crystal
(V1 - xCrx)2O3 of composition x = 0.038 [26].
Before going further in the analysis, it is instructive to consider the evolution of the unit cell
volume under pressure. Figure 2.3.e-f shows the evolution of the unit cell parameters a and c, as well
as of the unit cell volume V, of a (V1 - xCrx)2O3 powder with x = 0.028, extracted from Ref. [189]. As
expected, the unit cell volume decreases monotonously under pressure and presents a sudden -1.1 %
drop at the first-order Mott insulator to metal transition. Due to the smaller chromium content
(x = 0.028 against 0.038 in our study), the transition occurs at a slightly lower pressure (0.75 GPa
against 1.2 GPa) than in our work. Interestingly, the compared evolution of the unit cell volume and
the A1g and Eg Raman modes is clearly unconventional. The compression of the unit cell is indeed
usually accompanied by a strengthening of the elastic constants and a hardening of the phonon
modes. In most of the solids, this behavior results in a positive value of the Grüneisen parameter 
linking the relative change of unit cell volume and of phonons frequency ,
i.e. [190]. More generally, this classical behavior is also captured by various laws
relating to the volume of a solid body and the pressure to which it is subjected. The most famous is
the Birch-Murnaghan equation of state, which states that the bulk elastic modulus B, defined as
, increases when the volume V of the body is compressed by pressure [182]. Remarkably,
the behavior of (V0.962Cr0.038)2O3 under pressure violates the usual behaviors: the evolutions shown in
Figure 2.3 reveal a negative Grüneisen parameter. It also strongly suggests a softening of the bulk elastic
modulus with increasing pressure. Moreover, this behavior occurs not only at the first-order
transition but also in the compressed Mott insulator state for pressure lower than the one inducing
the insulator to metal transition.
These behaviors are fully consistent with the predictions of the compressible Hubbard for
the frequencies of the Raman modes discussed in previous sections Figure 2.1.e compares the
prediction of Hassan et al. [21] for T = 0.87 TC ~ 297.5K (TC for V2O3-Cr5% is 380K, see Figure
1.7.f) with the experimental evolution of the A1g Raman mode under pressure. The unconventional
decrease of 𝜔𝐴1𝑔 under pressure in the PI state and then the drop at the critical pressure are
qualitatively perfectly reproduced by the model. Quantitatively, the amplitude of the predicted effect
77
is however smaller than the experimentally observed one. Interestingly, the authors of the prediction
based on the compressible Hubbard model [22] clearly mention that their approach underestimates
the amplitude of the softening effect, because of (i) a too simplified evaluation of the screening
effects of the effective Hubbard term U and (ii) the use of a single band model while V2O3 is clearly
a multiband system (see section 2.1.1).
Overall and despite shortcomings of secondary importance, it appears that the compressible
Hubbard model is an appropriate approach to understand the consequence of electron-phonon
coupling on the bandwidth-controlled Mott insulator to metal transition in (V1-xCrx)2O3.
2.1.4 Raman study of the Mott insulator to metal transition in (V1 - xCrx)2O3
as a function of Cr content
In Section 1.2.1, we have proposed a simple rationalization of the difference between the
pressure-driven Mott insulator to metal arising in (V1-xCrx)2O3 and the IMT occurring by tuning the
chromium content. The former corresponds to a bandwidth-controlled IMT, while the latter is
driven between the “crystal field” splitting Δ𝑒𝑔𝜋−𝑎1𝑔 between the vanadium 𝑒𝑔𝜋 and a1g electronic levels.
These ideas were simply summarized in a generalized 2D phase diagram U/W versus Δ𝑒𝑔𝜋−𝑎1𝑔 shown
in Figure 1.11, where both PI and PM phases appear. An important remark is that the concept of
Mott line ending a finite critical temperature TC and separating a PI phase from its isostructural PM
counterpart should be still valid in the generalized phase diagram. This is illustrated in the 3D phase
diagram shown in Figure 2.4.a, where a vertical temperature axis is added. Here the Mott lines
becomes a Mott surface separating the PI and PM phase. An important outcome of this
representation is that the transition point of the xCr-induced and pressure-induced transition in (V1-
xCrx)2O3 are adiabatically connected. From this viewpoint, one might also expect unconventional
evolutions of the Raman modes frequencies across the xCr-induced IMT.
78
(a) Mott IMT T/W (b) 1 1

surface 0%
0,8 1.2%
2%
0,5 3%
Intencity
0,6
PI PM 0,4
5%
10%
(V1-xCrx)2O3 Pressure 0
200 220 240 260 280
0,2
V2O3 0
0
U/W 3 2 1 0
200 300 400 500 600
Raman shift (cm-1)
(c) (d) (e)
260
P pos. (cm )
-1
520
301
240
Å 3)
507
unit cell volume (•

Peak position (cm-1) Eg
A (Peak position) 494 14
1g
A1g
220 A1g(500nm) 5

481 299 13.96

4.98
a (Å)
c (•
IP (cm )
260
-1
Å)
305 13.92
4.96
A (Inflection point) 300

A1g
240 1g 4.94
0.08 0.04 0
13.88
297

295 x-Cr
12 10 8 6 4 2 0 -2
10 8 6 4 2 0 0.08 0.04 0
Cr%
Cr %
xCr
Figure 2.4: (a) Generalized phase diagram of (V1 - xCrx)2O3 highlighting the Mott insulator (PI) and the
metallic domains (PM) in the multiband three orbitals / two electrons situation, plotted as a function of
the correlation strength U/W and normalized temperature T/W (see details in Section 1.2.1). (b) Raman
spectra at different Cr content measured on polished (V1 - xCrx)2O3 pellets at room temperature. The spectra
are shown after a background subtraction procedure (see raw data and details of treatment in ANNEX 3-
B). (c) Evolution of the low energy A1g Raman mode frequency around 240 cm-1 with xCr (top part = peak
position and bottom part = inflection point of the peak on the high energy side). (d) Same as (c) for the Eg
mode around 300 cm-1 and the high energy A1g mode around 500 cm-1. (e) Unit cell volume versus Cr
content. Inset : lattice parameters versus Cr content [27].
Figure 2.4.b shows the evolution of the Raman spectra measured on polished pellets of
(V1 - xCrx)2O3 for several Cr contents 0  x  0.10. Unlike for measurements under pressure, it is
possible to track the evolution of the most of the Raman modes, including the high energy A1g modes
at 500 cm-1. Remarkably, the evolution of all optical modes is similar, with a clear softening of all
Raman modes between x=0.10 and x=0.02, i.e. upon approaching the Cr-driven IMT. At the IMT,
i.e. between x = 0.02 and x= 0, we observe a clear drop of Raman frequency of all modes. As for the
pressure-induced IMT, the softening at the xCr-induced IMT occurs along with a drop of the unit-
cell volume, as shown in Figure 2.4.e [27]. Overall, this Raman study of the (V1 - xCrx)2O3 solid
solution also suggests the softening of all elastic constant at the xCr-induced insulator to metal
transition. It means that the violation of standard laws relating pressure, volume, elastic constants
and vibration modes frequencies (negative Grüneisen parameter and violation of the Birch-
Murnaghan equation of states) arises not only at the pure bandwidth-controlled IMT. It seems to be
also relevant in the closely related “crystal-field-splitting-driven” (= xCr-driven) IMT case, a transition
also driven by Mott physics.
79
2.1.5 Discussion
To summarize, we have seen in this part that the compressible Hubbard model seems to be
a relevant starting point to describe the evolution of Raman modes around the Mott IMT in the
V2O3 system, either for a purely bandwidth-controlled IMT driven by pressure or for “crystal-field-
splitting-driven” IMT. Interestingly, the most interesting predictions of the compressible Hubbard
model at steady state are that:
(1) the discontinuous contraction of the unit cell volume occurs in order to avoid an
unphysical situation of negative bulk modulus; 1
(2) the amplitude of volume contraction is proportional to the gain of the free carriers kinetic
energy in the PM state.
Beyond the specific case of the V2O3 system, we notice that there are only a few published
studies where a crossing of the first-order Mott line is imposed. In the work of Fournier et al. [192]
on -(BEDT-TTF)2X system, a softening of sound velocity is clearly detected in the vicinity of the
isostructural IMT Mott line. Unfortunately, most of the other Raman or sound velocity studies in
Mott insulators were performed above the critical endpoint of Mott transition line [184,193].
Overall, it seems that our study and the one of Fournier et al. [192] confirm the relevance of the
compressible Hubbard model to capture the response of the lattice at the Mott insulator to metal
transition, for both single-band (case of -(BEDT-TTF)2X) and the multibands (case of the V2O3
system) Mott insulators.
Another important question raised by our study performed at thermodynamic equilibrium is
to evaluate if the same mechanism could be at work in out-of-equilibrium conditions where free
electrons and metallicity are generated thanks to ultrafast laser pulses. Under such conditions, even
though electrons are initially promoted to an excited state (which should result in an electronic energy
cost), they might quickly delocalize to nearby sites, thereby gaining electronic kinetic energy. An
obvious question arises here: what will be the lattice response in this situation, and will it result in a
volume contraction, as it occurs in steady-state conditions above the critical pressure leading to an
IMT? From the discussion above, the relevant approach to solving this problem is clearly the
theoretical study of the compressible Hubbard model out-of-equilibrium. To our knowledge, this
theoretical problem has never been addressed until now. In the rest of this manuscript, we will study
by an experimental approach the question of a possible compressive response of the lattice following
a massive electronic excitation of Mott insulators.
1 A similar situation occurs at the first order liquid – gas phase transition. The volume (which is the inverse of the density) change at
the transition also occur in order to avoid the unphysical situation of negative bulk modulus. The equivalence between the Mott
IMT and the liquid-gas transition is discussed in more details in Ref [191].
80
2.2 Lattice Contraction at the Electric -Field-Controlled Mott

Transition in (V 1 - x Cr x ) 2 O 3 in the PI phase
Modern band structure theories of Mott insulators predicts the occurrence of two types of
insulator to metal transitions (IMT), induced either through a change of the bandwidth or of the
electronic filing, with practical near-equilibrium implementations with charge doping and application
of a physical pressure [1,15]. Discovery of out-of-equilibrium Electric Mott transition in numerous
narrow gap Mott insulators [92], including the prototypal Mott insulator systems (V1-xCrx)2O3
launched intense debates about the underlying physical mechanism. Several models have been
proposed based on theoretical or experimental studies to explain the destabilization of the Mott state
by an electric field [61,90,194], including Zener [98] or electronic avalanche effect [6,103]. Besides,
numerous experimental studies demonstrate that this resistive switching is related to the formation
of a metastable filamentary conducting path [2]. The time stability of this filament depends on the
electric pulse applied to the system. For pulses just above the threshold electric field, the filamentary
path self dissolves after the pulse, which leads to a volatile transition. In contrast, for a higher electric
field, the filamentary path is stable for a more extended time and thus corresponds to a non-volatile
transition.
Nevertheless, important questions remain unsettled. In particular, what is the nature of the
metastable metallic phase induced by the Electric Mott Transition? Is the filamentary metallic phase
qualitatively different from the one induced by the classical mechanisms, i.e. the bandwidth- and
filling-controlled IMT’s? Addressing these issues is of great fundamental interest — it could unravel
an as yet unsolved basic property of highly correlated compounds placed out of equilibrium - and it
could also facilitate the use of Electric Mott transitions in Mottronics devices. Knowing the exact
mechanism would allow downscaling Mott memory devices and optimizing its geometry.
Experimentally, clarifying the nature of the filamentary metallic phase requires determining both its
macroscopic electronic response and its concomitant lattice response. In particular, establishing the
possible presence of symmetry breaking and/or volume change is crucial.
Here we study the electric Mott transition in the emblematic Mott insulator (V1-xCrx)2O3. We
performed several experiments on transited (V0.975Cr0.025)2O3 monocrystal and polycrystalline thin-
film (V0.95Cr0.05)2O3 device using two spatially resolved techniques able to unravel possible symmetry
breaking and volume changes, μ-Raman spectroscopy and μ -X-Rays Diffraction within the filament
(Figure 2.5).
Our results show that the main difference between the filamentary metallic phase and the
pristine insulating material is a lattice contraction. Combining electronic transport, conducting-
atomic force microscopy, Raman spectroscopy, and µX-ray diffraction measurements, we bring out
the strong analogy between the compressed phase within the filament and the metallic phase stable
under pressure beyond the Mott insulator to metal transition [27]. Therefore, the Electric Mott
transition mechanism deeply relates to the physics of the Mott insulator to metal transition.
81
(a) (V1-xCrx)2O3 (d) Lattice change in the filament ?

c
c
b a
Symmetry ? Volume ?
µ-Raman (vibration modes) µ-XRD (cell parameters)
b a
(b)
Intensity
Intensity
Light
(c) 2
I V X-Rays
Figure 2.5. Summary of experimental techniques used to probe Cr-doped V2O3. (a) Structure of
(V1 - xCrx)2O3-PI. (b-c) Schematic drawings of the samples (either a single crystal or a 150 nm thick thin
film) patterned with electrodes (in yellow) before (b) and after (c) a non-volatile Electric Mott Transition.
(d) the electric-field-induced new metallic filamentary phase is studied by μ-Raman and μ-X-Ray diffraction
mapping in order to clarify its evolution (symmetry-breaking, volume change?) with respect to the pristine
phase.
2.2.1 Study on (V0.975Cr0.025)2O3 crystal
2.2.1.1 Electric Mott Transition and transport me asurements
As discussed in Section 1.2.1, the rich phase diagram of the prototypal Mott insulator
(V1 - xCrx)2O3 reproduced in Figure 2.6 was established by the pioneering work of McWhan et al. [27].
Depending on chromium content (x), pressure (P) and temperature (T), three phases are expected,
the Paramagnetic Mott Insulator (PI), the Anti-Ferromagnetic Mott Insulator (AFI) and the
Paramagnetic Correlated Metal (PM) phases. For a crystal with a composition (V0.975Cr0.025)2O3, a PI
state is expected at room temperature, with a transition to an AFI state below 180 K.
Figure 2.6 presents the main results of an electronic transport study conducted on a single
crystal with this composition. The resistance R vs. temperature T curve measured on the
(V0.975Cr0.025)2O3 single crystal is in perfect agreement with the phase diagram, with a semiconducting-
82
like behavior (dR/dT < 0) at all temperatures and a tenfold increase of resistance at the transition
temperature TPI→AFI = 180 K. This sample was then submitted to an electric pulse (75 V 
18 kV.cm- 1/ 5µs) at 160 K, leading to a non-volatile Electric Mott Transition with a substantial drop
of electrical resistance. The subsequent R(T) curve measured upon cooling (see curve #2 in Figure
2.6b) reveals a substantial drop of electrical resistance and an unexpected feature at low temperature:
a broad but well-defined increase of resistance around 40 K.
(a) (b)
1
Mott
Insulator Correlated
Metal (PM) 107
(PI)
R ()
2
5
10
Antiferromagnetic
Insulator (AFI)
103
0 40 80 120 160 200
T (K)
(c) 1 2
State (1): State (2) : PI/AFI matrix

homogeneous PI/AFI state + conducting filament
Figure 2.6 : (a) temperature – composition x – pressure phase diagram of (V1-xCrx)2O3. The pressure scale
is positioned for the composition x = 0.025, using the equivalence of 0.4 GPa per Cr percent proposed in
Ref. [195]. (b) Electrical resistance vs temperature of a (V0.975Cr0.025)2O3 single crystal in its initial insulating
state (1) and after application of an electric pulse inducing the creation of a conducting filamentary path
(2), as schematically represented in (c).
The resistance drop may be easily explained by a simple model proposed for other Mott
insulators [56,92] considering the creation of a conducting filament embedded in the pristine
material and that percolates between the contact electrodes (see Figure 2.6c). In this model, the
conductance of the filament becomes predominant at low temperatures [25,196], and the additional
resistance jump at 40 K in (V1-xCrx)2O3 (see curve #2) hence originates from a transition within the
conductive filament. This gives some interesting clues about the nature of the metallic phase created
83
after the electric pulse. According to the phase diagram shown in Figure 2.6.a, a resistivity jump at
40 K is indeed expected at the PM – AFI transition of (V0.975Cr0.025)2O3 put under a pressure of  3
GPa. This macroscopic transport study presented more in detail in ANNEX 4-A, suggests therefore
that the electric pulse has created a filamentary path made of the same phase, but under a
compressive strain strong enough to push it in the correlated metal phase (PM) at room temperature.
2.2.1.2 Local Raman Spectroscopy Characterization
In order to confirm the lattice contraction, a direct visualization and study at the local scale
of the crystallographic structure within the filament are required. A second crystal of composition
(V0.975Cr0.025)2O3 issued from the same batch was hence subjected to a non-volatile Electric Mott
Transition.
Electric Field Pressure

(a) (d)
200 µm
(b) (e)
20 µm
(c) Filament Pristine

(f) 1,8 GPa 0 GPa
(PM) (PI)
Intensity (A.U.)
Intensity (A.U.)
180 210 240 270 180 210 240 270

 (cm-1)  (cm-1)
Figure 2.7: comparison of Raman spectra of the metallic states obtained in (V1-xCrx)2O3 by application of
an electric field (left, x = 0.025) and of hydrostatic pressure (right, x = 0.038). (a) Image of the
(V0.975Cr0.025)2O3 crystal, with four electrodes made of gold wires and carbon paste. Electric pulses were
applied between neighboring electrodes, resulting in the creation of conducting filamentary paths. (b)
Zoom on the region close to a filamentary path. (c) Raman spectra of the (V0.975Cr0.025)2O3 crystal obtained
inside and outside the 4 µm wide conducting filament shown in (b). (d) Scanning Electron Microscope
image of the (V0.962Cr0.038)2O3 single crystal used for Raman scattering experiments under pressure. (e)
Schematics of the Diamond Anvil Cell used for Raman scattering under pressure. (f) Raman spectra of a
(V0.962Cr0.038)2O3 crystal measured at room temperature in the Diamond Anvil Cell close to ambient
pressure in the Mott insulator states (PI) and at 1.8 GPa in the metallic state (PM, see Figure 2.8 ). Details
about the fit of Raman spectra (dotted lines in (c) and (f)) are presented in ANNEX 4.
84
Figure 2.7.a-b shows that this transition leads to the formation of a 4 µm wide filamentary
path bridging the two 25 µm apart conductive electrodes. The size of this filamentary structure allows
to use Raman scattering to probe its crystallographic structure, thanks to the micrometric spatial
resolution of this technique. The Raman spectra measured with a laser beam size of 1-2 µm both
inside and outside the filament are shown in Figure 2.7c (see also Figure A 4-2, Figure A 4-3 in
ANNEX 4-B). Both spectra show only the usual signatures of the corundum structure of the
(V1 - xCrx)2O3 system [32] [197], i.e. a broad structure located between 180 and 280 cm-1 and two
peaks around 300 and 500 cm- 1. No additional peaks are detected within the filament that could be
associated either with a symmetry-breaking or with the formation of other VOx phases. However,
the A1g Raman peak around 240-250 cm-1 measured in the filament is shifted by more than -20 cm-1
compared to the pristine Raman spectrum (see Figure 2.7.c). Moreover, a significant increase of the
Raman signal is observed below 200 cm-1. These modifications of Raman spectra, discussed in more
detail in ANNEX 4-B, support the existence of electronic and structural changes within the filament.
To check if they could be assigned to a lattice contraction, a comparison with the Raman study under
pressure performed at room temperature on a (V0.962Cr0.038)2O3 single crystal is relevant (see details in
Section 2.1.3.2). Figure 2.7.f discloses that this pressure-induced Mott transition causes both a shift
of the A1g Raman vibration mode and an increase of Raman intensity below 200 cm-1. Both changes
are similar to the one observed between the pristine material and the core of the filamentary path
created by the Electric Mott transition. Overall, this micro-Raman study provides additional evidence
that the Electric Mott Transition yields a compressed, isostructural, and metallic (V1-xCrx)2O3
filamentary phase. It strongly supports the lattice contraction scenario presented above based on
resistivity measurements.
2.2.2 Study on (V0.95Cr0.055)2O3 thin-film device
To further confirm the contraction of the lattice within the filament a spatially resolved X-
ray diffraction experiment was engaged using a 1 µm beam size at the micro-XAS beamline of the
Swiss Light Source synchrotron facility. The penetration depth of X-ray being far bigger than the
thickness of the filamentary path, we worked on polycrystalline thin films to maximize the signal
coming from the filament over that of the matrix.
85
(a) Raman conducting-AFM

(b) (d)
15
20 µm 1
I (pA/µm)
(c) (e)
0.94 0
20
µm 20 µm
Figure 2.9: Characterization of pristine and transited thin-film device by μ-Raman and conducting-AFM.
(a) General image of 150 nm thick polycrystalline (V0.95Cr0.05)2O3 thin film with a rectangular 170 x 20 μm
shape deposited on a SiO2/Si substrate patterned with TiN metallic electrodes. A schematic illustration of
electrodes connected to the electric pulse generator is shown. (b-c) Mapping of Raman intensities ratio
measured at 210 and 230 cm-1 in the interelectrode domains in the pristine (b) and transited (c) cases. (d - e)
conducting-AFM mapping of the same regions shown in (b-c).
In that purpose, a 150 nm thick polycrystalline (V0.95Cr0.05)2O3 thin film with a rectangular
170 x 20 μm shape was deposited on a SiO2/Si substrate patterned with TiN metallic electrodes (see
Figure 2.9). Details about the preparation method of this device are given in ANNEX 1-D.
A 1 µs / 70 V electric pulse ( electric field E  35kV/cm) was applied at 160 K between two TiN
electrodes to induce a non-volatile Electric Mott Transition. The sample was then characterized by
conducting-atomic force microscopy (c-AFM) and by Raman spectroscopy. Atomic Force
Microscopy (AFM) images shown in Figure 2.9.d-e reveal that the inter-electrode domain of the
(V0.95Cr0.05)2O3 film is affected by the Electric Mott Transition does not display any morphological
change. Figure 2.9.d-e also shows the distribution of current measured by applying a 25 mV bias
between the conductive AFM tip and the neighboring electrode, integrated over similar areas in
regions affected by the EMT (Figure 2.9.e) and in pristine regions (Figure 2.9.d). Interestingly, the
map shown in Figure 2.9.e unveils a clear  2 µm wide zone of higher conductivity close to the left
edge of the (V0.95Cr0.05)2O3 stripe. Figure 2.9c shows that the Raman mapping measured in the highly
conducting region uncovered by c-AFM present a tendency towards metallicity (ratio of Raman
intensities between 210 and 230 cm-1 > 1) compared to spectra measured outside of this zone. The
c-AFM and Raman measurements confirm therefore the presence of a metallic filamentary path
between the two inner electrodes of the polycrystalline device with similar features as the ones
observed on the crystal.
86
µ-X Rays footprint (110)

(104)
(1 x 5 µm²)
Intensity (a.u.)
(a) (b)
20 µm filament
2 pristine
44 46 48
2 (deg)
a (Å) c (Å) c/a V (Å3)
4.990 2.796
(c) 301
13.95
4.985 2.800
13.96
2.804 300
4.980
13.97
2.808
Figure 2.10: Summary of μ-XRD experiment on (V0.95Cr0.05)2O3 thin-film device (see details in ANNEX 4-
C). (a) The orientation of the sample and the XRD beam. Due to the fixed 11° angle between the X-Ray
beam and the thin film, the footprint of each pixel is 1 µm wide and 5 µm long. The X-Ray was performed
by moving the sample holder. The step between two successive pixels is of 0.5 and 2.5 µm along with these
two directions. (b) Comparison of XRD diffractograms at the filament and pristine regions of the device.
(c) Maps of lattice parameters extracted by XRD refinement of a series of diffractogram like in (b)
A µ-XRD experiment was then engaged by mapping a 200 x 45 µm zone containing the
whole patterned (V0.95Cr0.05)2O3 stripe of the device (170 x 20 µm). Details about the µ-XRD
experiments performed at the Swiss Light Source Synchrotron are provided in ANNEX 4-C. The
map consists of 9191 (101 x 91) "pixels" of 5 x 1 µm, each of them corresponding to a 2D powder
X-Ray diffractogram (see Figure 2.10-a). An azimuthal integration of the Debye rings was performed
for each 2D powder X-Ray diffractogram to extract the intensity vs 2 plots. Representative
examples of the integrated diffraction patterns outside and inside the filament are shown in Figure
2.10.b. Outside the filament the diffraction pattern exhibits the expected Bragg peaks (104), (110)
and (103) of the corundum structure. Inside the filament the same peaks are observed, but slightly
shifted. Moreover, the absence of new diffraction peak discards the change of crystallographic
symmetry or the formation of a chemical phase with different composition.
The mechanism proposed in Ref. [198], which suggests the creation of a V5O9 phase to
explain the resistivity jumps in the low temperature phase, can therefore be excluded in our case.
This crude comparison of XRD patterns in Figure 2.10.b already confirms that the (V0.95Cr0.05)2O3
phase within the filament keeps the corundum structure but with a different cell volume. The a and
c unit cell parameters were then refined for each pixel based on the position of the (104), (110) and
(103) Bragg peaks and were represented as maps in Figure 2.10.c. These maps reveal a remarkable
87
feature: in the filamentary region bridging the two central electrodes an increase of the c parameter
(from 13.95 to 13.97 Å,  + 0.15%) with respect to the rest of the sample is clearly observed.
Conversely, the unit cell parameter a is contracted by about -0.4% compared to the normal zones.
Overall the unit cell volume within this filament undergoes a contraction of -0.6% compared
to the pristine material. This very specific behavior (c increase and a decrease) is strikingly similar
with changes observed at the isostructural Mott transition driven by hydrostatic pressure [27,189].
A +0.3 % increase of c parameter and a -0.8 % decrease of a parameter is indeed observed at the
pressure induced Mott transition at room temperature for a (V0.972Cr0.028)2O3 single crystal [189].
Conversely, our data do not point towards the creation, within the filament, of a non-stoichiometric
phase with a (V+Cr) : O ratio departing from 2:3. Indeed such non-stoichiometry induces a
concomitant decrease of both a and c parameters [199,200], which contrasts with the observed
increase of the c parameter measured in the filamentary path shown in Figure 2.10.c.
2.2.3 Discussion
The electronic transport and c-AFM measurements, as well as the μ-Raman and µ-XRD
studies, presented in this work are consistent with the creation by electric field of a metallic
filamentary phase whose unit cell volume is contracted compared to that of pristine (V1-xCrx)2O3
Mott insulator. Despite stimuli of seemingly very different nature, the electric-field-induced
contraction resembles the one observed under pressure at the transition between the Mott insulator
and paramagnetic metal. Interestingly, and as discussed in Section 2.1, the pressure-induced volume
contraction at the Mott transition can be rationalized within the compressible Hubbard
model [21,201]. In this approach, the lattice degrees of freedom react to the softening of electron
degrees of freedom close to the Mott transition, resulting in a volume change of the lattice in order
to avoid the unphysical situation of negative bulk compressibility.2 The results presented in this work
strongly suggest that a similar mechanism is at play at the Electric Mott transition, i.e. that the creation
of excited and delocalized electrons during the electric pulses leads to a compressive lattice response.
Overall, this work provides strong indication that Mott physics plays a pivotal role in the insulator
to metal transition observed under electric field in (V1-xCrx)2O3. Our study therefore suggests that the
relevant theoretical approach to fully capture the Electric Mott Transition is the out-of-equilibrium
compressible Hubbard model, which, to the best of our knowledge, has never been studied until
now. Finally, the direct evidence of a lattice compression at the Electric Mott Transition in
(V1 - xCrx)2O3 shown in this work rationalizes a posteriori the presence of electric-field-induced
superconductivity in the canonical Mott insulator GaTa4Se8 [203], [204], which is also
superconducting under pressure [205]. This promotes the idea that the lattice compression at the
Electric Mott Transition is a universal property of Mott insulators. Overall, the advances in
2 A similar situation occurs at the first order liquid – gas phase transition, where the volume change at the transition also occurs to
avoid an unphysical negative bulk modulus. The equivalence between the Mott IMT and the liquid-gas transition is discussed in
more details in Ref. [202].
88
understanding the mechanism of the Electric Mott Transition gained in this study will greatly
facilitate the implementation of advanced Mottronics applications such as Resistive Mott memories
and neuromorphic components in the near future [116,136],.
89
2.3 Shock wave-induced non-volatile IMT on Cr-doped V 2 O 3

thin films
In Section 2.2, we demonstrated that the metastable compressed phase created in out-of-
equilibrium conditions during an EMT shares similar characteristics with that obtained at equilibrium
under pressure in Section 2.1. However, in the equilibrium case, the metallic state is restored when
pressure is released. From this simple observation, one can wonder why the system can remain
compressed after an EMT and not after the release of the pressure. A possible explanation might
arise from the difference between pressure and electric pulse induced Mott transitions. Obviously,
the main difference is that EMT happens when the system is out of equilibrium. However, another
factor that can play a role would be time. While pressure application and release is very progressive,
the EMT occurs probably over a 106 to 109 shorter period of time. In a first order transition induced
by an external stimulus (such as temperature or pressure), it is well known that the hysteresis window
widens when the application rates of the external stimulus increases. One might then argue that the
hysteresis windows in the EMT could simply be large enough to allow a metastable compressed
phase down to ambient pressure. In order to evaluate this scenario, a study aiming at applying a laser-
driven pressure shock wave was engaged in a (V0.95Cr0.05)2O3 thin films. This research involved
primarily the group of T. Pezeril (at IMMM in le Mans and then at MIT in Boston), the group of
M. Lorenc in Rennes and our group. The experiments were performed within the Ph.D. project of
Yevheniia Chernukha [179] and the master internship of Ritwika Mandal [206], using
(V0.95Cr0.05)2O3 250 nm thin films deposited sapphire substrates that I prepared at IMN (see details
about sample preparation in ANNEX 1).
In this study, the shock-wave generation was operated by sending a ring-shape laser at the
surface of the thin film. Figure 2.11.a-b schematically shows experimental setup for shock wave
generation. The ring-shaped light impact triggers inward- and outward-propagating acoustical waves
on the samples surface, with pressure exceeding the 1 GPa range at the center of the ring.
Figure 2.11.d compares µ-Raman spectra measured after the shock-wave experiment in
different points inside and outside of ring. Interestingly, we can see a strong Raman shift of the low
energy A1g mode at the center of the ring, where wave amplitude and hence pressure was maximal.
This shift is strongly reminiscent of that obtained under pressure and also after the EMT. Notably,
the created state is permanent for a long time after shock-wave pressure was applied.
In order to check the lattice parameters of the phase identified in µ-Raman data, we
performed μ-XRD measurements at the CRISTAL beamline of the synchrotron SOLEIL. Figure
2.11.e presents the evolution of the unit cell volume across the ring extracted from XRD data.
Clearly, there is a substantial volume drop at the ring center, precisely where the Raman shift has
been observed. Overall, these experiments show that a non-volatile transition in (V0.95Cr0.05)2O3 can
occur without any direct electronic excitation. As discussed above, this could be a clue to understand
the mechanism of the non-volatile EMT. It could result from the very fast release of the electronically-
induced compression, opening sufficiently the hysteresis window around the first-order transition to
maintain a metastable compressed and metallic phase at ambient conditions. This hypothesis is
highlighted since it suggests that the key parameter to stabilize a non-volatile EMT is to impose a
release of the stimulus (here the electric-field) as abrupt as possible, similarly to a supercooling like
phenomena.
90
(a) Sample
(c) (d)
Lens Axicon 0.9 mJ/P
Laser
x z
Diverging
wave
Converging
wave
y
Laser ring (e) 0,5
x
Relative V (A )
3
0
Laser circle 800nm -0,5

(b)
-1
Mott insulator V2O3:Cr 5% 250nm
-1,5
Sapphire (500 µm)
-2
0 50 100 150 200 250 300 350 400
Position (m)
Figure 2.11 (a) Sketch of the experimental setup for laser-driven shock wave generation. Axicon creates a
ring-like shape of a laser beam, which is focalized on the sample surface. The pump wavelength was 800nm,
and pulse duration 300ps. Absorption of light within the ring launches a surface acoustical wave
propagating inside and outside of the ring due to the thermoelastic mechanism. (b) Schematic illustration
of (V0.95Cr0.05)2O3 thin film deposited on a sapphire substrate with a laser ring focused on the top. The
diameter of the excitation ring was 200 μm width 10 μm. (c) Image of the shocked region of (V0.95Cr0.05)2O3
thin film. The sample was shocked by Yevheniia Chernukha at the University of Rennes [179]
(d) Comparison of the Raman spectrum of compressed center (violet curve) with comparison to the
nontranslated zones inside and outside of the ring. (e) Unit cell volume evolution across the ring extracted
from μ-XRD mapping data performed at the CRISTAL beamline of SOLEIL Synchrotron. Ritwika Mandal
from IPR performed the XRD refinement within her Master internship [206].
91
CHAPTER 3 Out-of-equilibrium Mott insulator to metal transitions: Carriers and lattice dynamics
Chapter 3
OUT-OF-EQUILIBRIUM MOTT INSULATOR TO METAL
TRANSITIONS: CARRIERS AND LATTICE DYNAMICS
***
This chapter focuses on the properties of Mott insulator placed out of equilibrium and
on the dynamics of the insulator to metal transition (IMT) induced by laser and/or
electrical pulses. The first part of the chapter reports on pump-pump-probe
experiments, for which a crystal of the Mott insulator GaTa4Se8 is simultaneously
excited by electric and laser pulses while an electric probe monitors its conductivity.
This study shows that the resistive switching is affected by the number of generated
photocarriers rather than by the accumulation of energy deposited by the femtosecond
laser. This result, as well as time-resolved photoemissions experiments support, an
electric Mott transition mechanism driven by accumulation of hot carriers, which opens
the possibility to build up an artificial “leaky integrate and fire” electro-optical Mott
neuron. The second part of the chapter presents and discusses the results of several
ultrafast IR pump-probe experiments performed on (V0.95Cr0.05)2O3 in the PI phase
and on V2O3 in the AFI phase. In the chosen experimental conditions optimized to
avoid thermal effects, we show that the non-thermal photoinduced insulator to metal
transitions can be attained. In V2O3 in the AFI phase, convincing evidence point to
a mechanism based to a light-induced launching of a compressive strain wave leading
to an insulator to metal phase transformation.
***
In the previous chapters, we have seen that Mott insulators subjected to an electric field can
undergo an insulator to metal transition (the Electric Mott Transition, EMT) due to an electronic
avalanche process. In reaction to this transient excited metallic state, the lattice responds with a
compression of the unit cell volume, similarly to what happens at the pressure-induced
( bandwidth-controlled) Mott insulator to metal transition.
The studies reported in Chapter 2 were performed in “static” conditions, on the one hand
at equilibrium during the pressure-induced insulator to metal transition and on the other hand by the
analysis of a metastable metallic state obtained after a non-volatile transition. These studies give no
clues about the dynamical properties of Mott insulators placed out-of-equilibrium, both for the
electron and for the lattice. In particular, important basic information, such as the lifetime of excited
electrons or the dynamics response of the lattice after electronic excitation, are still lacking at this
point. The objective of this chapter is to address these questions. To achieve this goal, it is quite clear
that the stimulus used so far to bring Mott insulator out-of-equilibrium, i.e. the application of electric
pulses in the μs range, is no longer sufficient. As the proposed EMT mechanism involves the massive
creation of excited electrons, the application of short laser pulses is obviously a relevant stimulus.
The application of femtosecond laser in the infrared range is indeed able to promote excited electrons
in the conduction band. In this chapter, we will employ such ultrafast IR laser, alone or in
92
combination with electric pulses, to probe the dynamics of out-of-equilibrium Mott insulators. We
will see that the sole application of femtosecond IR light pulse not only impact the electronic
sub - system, but also induces a strong response of the lattice. In complement, we will see that the
electric field carried by a THz laser pulse (h << Mott-Hubbard gap) can bring Mott insulators into
a metallic state through an original and ultrafast pathway, without any direct photoexcitation of
electrons.
93
3.1 Light control of Electric Mott transition and carrier

dynamics
The main signature of an Electric Mott Transition is a resistive switching occurring under
electric field and which has been explained by an electronic avalanche phenomenon and the creation
of hot carriers [6]. Here we use a pump-pump-probe experiment with electric and laser pulses
simultaneously exciting a crystal of the Mott insulator GaTa4Se8 while probing its electrical
conductivity in order to deepen the understanding of the resistive switching mechanism in this
system. Our study shows that photodoping impacts the dynamics of the EMT. By triggering the
number of photogenerated carriers and the energy provided to the system, we shed light on the
physical quantity that is integrated into this Mott material [126] and establish the proof-of-concept
of an artificial electro-optical "Mott" neuron. Another question addressed in this section deals with
the ability to accumulate (integrate) electrons during the avalanche phenomena. Such an
accumulation is possible for a long enough (> μs) carrier lifetime, which is probably a fundamental
property allowing the realization of artificial Mott neurons. We will explore the scarcely addressed
question of the carrier lifetime by two different approaches, including the recently developed time-
resolved Photo-Electron Emission Microscopy technique (tr-PEEM).
3.1.1 Electro-optically induced Mott transition
The Electric Mott Transition (EMT) was observed in numerous Mott insulators. In this
section, we focus on a member of the family of narrow gap chalcogenide Mott insulators AM4Q8
(A=Ga, Ge; M = V, Nb, Ta, Mo; Q= S, Se) that exhibit a clustered lacunar spinel structure. Figure
3.1 presents the thermal dependence of the resistivity and a typical resistive switching phenomenon
observed on a crystal of GaTa4Se8.
94
Figure 3.1: (a) Resistivity versus temperature measured on a single crystal of the Mott insulator GaTa4Se8.
The inset shows the crystallographic structure of GaTa4Se8 and a picture of the GaTa4Se8 single-crystal
connected with four gold wires. (b-d) A typical example of a volatile Electric Mott Transition induced at
85 K by the application of a 120 µs / 90 V electric pulse on a circuit made of the GaTa4Se8 crystal in series
with a 1 k load resistance. The temporal evolutions of the current Isample (b), of the sample voltage Vsample
(c), and of the sample resistance Rsample = Vsample / Isample (d) clearly indicate a resistive switching occurring
after a time delay of  80 µs.
3.1.1.1 Electronic avalanche and integ ration of car rier

multiplication rate
Details about the avalanche mechanism during EMT and its application to the AM4Q8 Mott
insulators are described in Chapter 1. The model considers the existence of small numbers of
electrons either trapped in localized levels below the conduction band or delocalized in the
conduction band. The latter pre-existing free carriers can be accelerated under electric field. The
95
kinetic energy gained under electric field by individual electrons will be rapidly shared with the other
electrons and the lattice thanks to, respectively, electron-electron (e-e) and electron-phonon (e-ph)
interactions [6]. The average excess of energy of electrons corresponds to an effective temperature
increase of the electronic sub-system Telectronic. Due to the increase of Telectronic, new hot carriers are
generated under electric field until a new equilibrium situation is reached between the power given
to (electric power) and the power taken out from (e-ph) the electronic system. This actually
corresponds to a two temperatures model with the electronic temperature stabilizing slightly over
that of the lattice. But once the applied electric field reaches a threshold value Eth, the equilibrium
between the lattice and electronic sub-systems cannot be maintained, and the system enters into a
nonequilibrium state with the number of hot carriers diverging and inducing an IMT (i.e. the resistive
switching). The theory predicts a temperature-independent deviation from Ohm's law at low field
(E ≪ Eth) with the multiplication rate of new carriers in the conduction band 𝑛̃ = 𝑛⁄𝑛0 depending
only on the square of the electric field if we assume a constant electronic mobility:
𝑛 𝜎(𝐸) 𝐸 2 𝜀𝐺
𝑛̃ = = = 𝑒𝑥𝑝 ( 2 ) (3.1)
𝑛0 𝜎0 𝐸𝑡ℎ 𝑒𝛥𝜀
where 𝑛0 and 𝜎0 are the number of carriers and conductivity in the limit of zero field. 𝜀𝐺 and 𝛥𝜀
represent the width of the gap and of the in gap state levels, respectively, and 𝑒 is Euler's constant.
In the other boundary case of the prediction (E = Eth), the threshold value of the carrier
multiplication rate 𝑛̃𝑡ℎ becomes:
𝜀𝐺
𝑛̃𝑡ℎ = 𝑒𝑥𝑝 ( ) (3.2)
𝛥𝜀
Previous low field conductivity measurements of the narrow gap Mott insulators AM4Q8,
done by the IMN group, are entirely consistent with Fröhlich's prediction [7]. As displayed in Figure
3.2.a, the measurements, performed at different temperatures on a GaTa4Se8 crystal, lie on the
predictable master curve for the carrier multiplication rate 𝑛̃ (see blue line). It fully supports the
proposed model suggesting that, below the threshold field, 𝑛̃ may stabilize in time, while above the
threshold electric field ( when the system exceeds the threshold carrier multiplication rate 𝑛̃𝑡ℎ ), 𝑛̃ can not
stabilize anymore and may diverge in time.
Figure 3.2.b presents the experimental conductivity ratio dynamics 𝜎̃(𝑡) = 𝜎(𝑡)/𝜎0
measured under electric field for a GaTa4Se8 crystal. Below Eth, the conductivity ratio stabilizes over
time and the system remains insulating, which means that the electronic temperature reaches an
equilibrium. But when electric field exceeds the threshold value 𝐸𝑡ℎ , a sudden increase of the
conductivity ratio appears (i.e. a volatile resistive switching occurs) as the carrier multiplication rate
approaches 𝑛̃𝑡ℎ and the system enters into an avalanche carrier multiplication regime (Figure 3.2.a).
96
Figure 3.2: (a) Transport measurements performed at different temperatures and low electric field for a
GaTa4Se8 crystal. All measurements fall on a master curve following the predicted evolution of the carrier
multiplication rate 𝑛̃ (blue dotted line). (b) Conductivity ratio dynamics 𝜎̃ = 𝜎/𝜎0 measured under electric field for
a GaTa4Se8 crystal for 𝐸 < 𝐸𝑡ℎ and 𝐸 > 𝐸𝑡ℎ .
According to the model, new hot carriers are accumulated in the AM4Q8 compounds under
electric field. This model is therefore of great interest in the framework of neurocomputing as it
suggests that the AM4Q8 compounds behave as hot carriers integrators. Similar to biological neurons
that accumulate charges on their membranes until the membrane potential reaches a threshold, the
Mott insulators would accumulate hot carriers until the carrier multiplication rate reaches a threshold.
3.1.1.2 Electro-optical pump-probe experiment
In order to test the electronic avalanche scenario, we have compared the carrier
multiplication rate dynamics obtained under electric pulse alone or when the electric pulse is coupled
to a laser pulse to modify the carrier number. The setup sketched in Figure 3.3 was used to implement
this electro-optical pump and electric probe experiment. A freshly cleaved surface of a GaTa4Se8
97
single crystal was connected by two gold electrodes separated by about 200 µm to allow the use of
ultrashort laser pulses with a fixed spot size slightly smaller than the inter-electrodes distance. We
used a femtosecond Ti:Sapphire regenerative amplifier and an Optical Parametric Amplifier (OPA)
to tune the wavelength of the optical pump between visible and infrared. The electrodes are used
both to apply the electric field thanks to a pulse generator and to monitor the temporal evolution of
the electrical conductivity thanks to an oscilloscope connected in parallel. The setup worked in the
single-shot mode with a synchronous application of the electric and laser pulses. We performed all
experiments under nitrogen cryostream cooling at 85 (± 5) K in order to have optimal stability
conditions to induce the resistive switching experiments (see details on sample preparation and on
experimental conditions in ANNEX 2)
In the first series of experiments, we simultaneously applied electrical voltage pulses of 60V
(in the chosen experimental conditions, a value exceeding the threshold value) and 100 fs / 800 nm
(hν = 1.55 eV) femtosecond pulses, while varying the laser fluence. As depicted in Figure 3.3.a, the
main idea was to tune the number of photoexcited carriers at the beginning of the electric pulse.
Figure 3.3.a shows that the concomitant application of an optical excitation drastically
reduces the time delay necessary to observe the resistive switching. Moreover, this time delay
decreases when the excitation density increases. These experiments already evidence a clear impact
of photoexcited carriers on the resistive switching. They suggest a similar role of hot carriers
generated by light or electric field. This result supports that the photoexcited carriers are added to
the hot electrons accumulated under electric field to reach the threshold carrier multiplication rate
and promote the avalanche phenomenon. However, the possibility that a significant part of the laser
pulses energy is converted into simple thermal heating rather than to the sole creation of new free
carriers cannot be completely discarded based on this experiment alone.
98
Laser pulse
(a)
hν = const nex  const
E UHB E UHB E UHB E UHB E UHB
E-field
ñ
hν=2.3eV
hν=0.5eV
Rload
LHB LHB LHB LHB LHB
(b) tlaser = 0 tlaser = 0

(c) 85
60
Vpulse (V)
Vpulse (V)
0 0
hν = tdelay tdelay
1.55eV ñth ñth
5 5
hν = 2.3eV
hν = 0.5eV
No laser No laser
1 1
0 20 40 60 80 100 0 20 40 60 80 100
Time (µs) Time (µs)
Figure 3.3: Temporal evolution of the normalized conductivity 𝜎̃ =  / 0 during a voltage pulse applied
to different circuits containing a GaTa4Se8 crystal and which is synchronized or not with a single-shot 100
fs laser pulse applied at the beginning of the pulse. (a) Schematic representation of the setup and impact of
laser fluence or wavelength. (b) Evolution of 𝜎̃(𝑡) measured under a 60V voltage pulse and when fs laser
pulses of constant energy (800 nm, h = 1.55 eV) and of increasing fluences are applied (from 1µJ/Pulse
to 10 µJ/P). (c) Evolution of 𝜎̃(𝑡) measured on a different GaTa4Se8 crystal under a 85 V voltage pulse
and when fs laser pulses leading to the same number of induced photocarriers are applied with two different
energies (h = 0.50 and 2.4 eV).
One way to evaluate the role of thermal effect is to tune the excitation density and the pulse
photon energy using the OPA. As sketched in Figure 3.3c, the main objective of these experiments
is to modify the photon wavelength and hence energy, while keeping a constant number of photons.
This series of measurements were performed using another GaTa4Se8 crystal in a configuration similar
to that depicted in Figure 3.3a. The boundary values of the photon energies used were hν = 0.5 eV
and 2.4 eV. The fluence was set to 2.5 µJ/pulse for hν = 0.5 eV and to 8.8 uJ/pulse for hν = 2.4 eV.
These values correspond to 7.1016 and 5,5.1016 absorbed photons per cm2, respectively, and hence to
almost similar densities of photodoped carriers (see ANNEX 2).
99
Electrical measurements presented in Figure 3.3.c demonstrate no change in carrier

multiplication rate dynamics under electric field independently of the used photon energies (0.5 eV
and 2.4 eV, respectively). In particular, the time delays required to reach 𝑛̃𝑡ℎ (i.e. the resistive
switching) are remarkably similar while the energy provided by the light pulses changes by a
factor  3.5 (8.8 µJ against 2.5 µJ). It demonstrates that the key parameter of the optical excitation
that controls the resistive switching is not a trivial photothermal effect. Conversely, it suggests that
the relevant parameter driving the Electric Mott Transition is rather the number of hot carriers
generated. Mott insulators appear therefore as hot carriers integrators that undergo a resistive
switching when the carrier multiplication rate (𝑛̃ ) reaches a threshold value depending only on
intrinsic material features/properties.
We will see in the next part that the above clarification of the physical parameter that is
accumulated during the Electric Mott Transition (the carrier multiplication rate 𝑛̃ ) is of crucial
importance for a possible application of the EMT property, the artificial “Mott” neuron.
3.1.2 Concept of an artificial electro-optical Mott neuron
The classical biological neural circuit consists of neurons interconnected by synapses that
enable the inter-neuron signal transfer. The main building block of the brain, i.e., the biological
neuron, accumulates incoming signals during the "integration time" and "fires" (generates output)
once the integration reaches a threshold. The creation of artificial neurons, therefore, requires
materials possessing a physical property able to implement both Integrate and Fire functionalities
with a clear identification of a physical quantity that is integrated [129]. In Chapter 1, we described
the successful use of the EMT in Mott insulators in a new kind of electronic device, the single
component artificial Mott neuron. The results of the electro-optical pump-probe experiment
presented above open a new possibility, namely the implementation of an electro-optical Mott
neuron for artificial neuromorphic devices [121,129] and neuromorphic photonics [182,183].
As depicted in Figure 3.4.a, when a biological neuron receives input spikes from other
neurons, it is integrated into the membrane potential. In between the incoming spikes, the membrane
potential relaxes, which is known as a leaky behavior. Finally, when the membrane potential reaches
a threshold value, the neuron triggers an output action potential (Firing event). The biological neuron
implements, therefore, three important functionalities that are at the basis of the most used model
of an artificial neuron, namely the Leaky, Integrate and Fire (LIF) artificial neuron. As discussed in
Chapter 1, the IMN team previously demonstrated that the AM4Q8 compounds, as well as other
Mott insulators like (V1-xCrx)2O3 or [Au(iPr-thiazdt)2], behave as LIF artificial neurons [184]. This
behavior was observed when a crystal of Mott insulator was subjected to a train of electrical pulses.
100
(a)
Membrane potential +
Bio-neuron
+ C
+
+
+ Integrate
+
+
+ R Leaky
Input
Input
Output
threshold
Membrane
potential
Leaky Leaky Integrate FIRE
Integrate and Fire (LIF) Leaky Integrate
time
(b) “ tt” E ct -optic neuron

E UHB
Input
Output
Laser
LHB pulse ñth
Output
Rload
ñth
time
Figure 3.4: (a) Schematic picture of a biological neuron receiving input spikes from other neurons and
triggering an output action potential when the membrane potential reaches the threshold value (b) Sketched
electro-optic Mott neuron setup and graphical representation of the electronic dynamics with and without
laser pulse.
Figure 3.5.a-c shows the results of other experiments intended to control the Mott neuron
also by light, and for which a 100 femtosecond (fs) laser pulse was synchronized with the first
electrical pulse of the train. These experiments were conducted by applying the train of electrical
pulses and laser pulses on a GaTa4Se8 crystal using a similar setup, as shown in Figure 3.4.b. The
green curves in Figure 3.5.b, represent the conductivity ratio ( carrier multiplication rate, see above)
measured without laser. When electric pulses of 63V are applied, the system fires ( a resistive
switching is observed) during the fifth pulse, while no firing event is observed when pulses of 52V
101
are used. We will now see that applying a laser pulse (10μJ/pulse, hν = 1.55 eV) at the beginning of
the train of electric pulses has a strong impact on the behavior of the Mott neuron. Indeed, the firing
event occurs much sooner compared to the case without optical excitation. For electrical pulses of
63V, it occurs after only two electrical pulses. In the 52 V case, the application of a single laser pulse
leads to a particularly striking modification since a firing event occurs during the fifth pulse (see violet
curves, Figure 3.5), while no firing event is recorded with electrical stimulus only. These experiments
demonstrate a clear impact of an ultrashort light pulse on the behavior of the Mott artificial neuron.
It reveals that ultrashort light pulses could be used as a new external parameter to tune the firing
time of the device. These experiments provide, therefore, the proof-of-concept for the
implementation of artificial electro-optical Mott neurons.
(a)
50 tOn tOff
VEP (V)
63V 25s 52V 75s

0
(b)
10
without laser without laser
0
(c)
15 Laser
Laser
0
0 100 200 300 400 0 100 200 300 400
time (µs) time (µs)
Figure 3.5: Experimental resistive switching obtained by applying trains of short pulses synchronized with
the laser at two different voltage levels. a) Profile of applied voltage pulses, b) and c) sample conductivity
time profiles without and with laser pulse, respectively. The measurements were performed in similar
conditions to the ones of Figure 3.3.a
To conclude, the electro-optical pump / electrical probe measurements presented in this

work demonstrate the possibility to control the dynamics of the Electric Mott transition by ultrashort
laser pulses in the Mott insulator GaTa4Se8. It further confirms that light and electric field can act
synergistically to induce an insulator to metal transition in a Mott insulator. These results are
compatible with an Electric Mott transition mechanism based on a carrier multiplication
phenomenon whose dynamics can be modified thanks to the generation of extra photo-carriers by
102
optical excitation. The obtained results demonstrate that Mott insulators behave under electric field
as leaky integrators of hot carriers, more precisely of carrier multiplication rate. When the carrier
multiplication rate reaches a threshold value, the firing event (i.e. the resistive switching) occurs. Hot carriers
and carrier multiplication rates are, therefore, the physical quantities of interest to control the Mott
Neuron. Beyond this fundamental output, we have also shown that both light and electric field
stimuli can control the dynamics of the Mott artificial neuron. Our work demonstrates a new concept
of electro-optical artificial Mott neuron and opens a new way in the emerging field of neuromorphic
photonics [207,210].
3.1.3 Carrier lifetime in Mott insulators
At this stage, it is important to recall that the mechanism proposed to explain the Electric
Mott Transition (the electronic avalanche scenario) can be relevant only if the carrier lifetime is
sufficiently long. Too short carrier lifetime, due e.g. to a fast electron-hole recombination associated
to defects, might indeed prevent the electronic temperature rise and hence the resistive switching to
occur. It is therefore important to evaluate if sufficiently long carrier lifetime appears in
representative Mott insulators. In classical semiconductors like silicon, the carrier lifetime varies from
the millisecond when Si is ultrapure to the nanosecond in heavily doped silicon [211,212]. However,
purity levels of Mott insulators are far from one of the ultrapure classical semiconductors, and one
might naively expect very small carrier lifetime in the nanosecond range. Experimentally, the carrier
lifetime in Mott insulators has been scarcely explored, partly due to the lack of appropriate technique
to measure it.
Interestingly, photocarrier lifetime can be estimated indirectly from the previous electro-
optical pump-probe experiment. Besides, direct measurement is now possible with the advent of
time-resolved photoemission electron microscopy (tr-PEEM), a technique well-suited to determine
carriers lifetime [213]. In this section, we discuss our results obtained by both techniques on
GaTa4Se8 and GaV4S8 compounds. Both experiments indicate unusually long carrier lifetimes.
3.1.3.1 Evaluation of photo -car rier lifetime from the electro-

optical pump-probe experiment
An appealing feature of an electro-optical pump-probe setup described above is the

possibility to apply laser pulses before the beginning of the electric pulse. Exploring this situation of
“negative laser time lag” and observing its influence on transition dynamics may give interesting clues
on photo-carriers lifetime. Figure 3.6.a shows an example where a 800 nm / 100 fs / 10 µJ laser
pulse applied 400 µs before the beginning of the electric pulse induces a tdelay decrease of 21 µs. This
is a clear indication that photocarriers might still exist after several hundreds of µs.
103
Laser time lag (µs)

(a) -400 -200 0 200 400 (b) (c)
No laser 4
4
10 10 No laser 0
Laser time
laser signal
(s)
lag : - 400 µs
R (Ohm)
R (Ohm)
-20
delay
tdelay =
21 µs 2 = 770  50 µs
t
3
10
1000 -40
1 = 6  2 µs
-60
-100 -50 0 50 100 -4000 -2000 0
0 20 40 60 80 100
Time after start of electric pulse (µs) laser time lag (µs)
Time after start of the electric pulse (µs)
Figure 3.6: (a) Evolution of sample resistance during a volatile Electric Mott Transition, without (grey
curve) and with (black curve) a 100 fs / 1.55 eV / 10 µJ laser pulse applied 400 µs before the beginning of
the electric pulse. The orange curve corresponds to the signal of the photodiode signaling the application
of the laser. (b) Influence of the laser pulse on the time delay tdelay at which the resistive switching occurs,
for a laser applied between -2 µs and -4000 µs before the beginning of the electric pulses. (c) Fit of the
experimental variation of tdelay with the laser time lag using a model including two relaxation times 1
and 2.
Figure 3.6.b shows the results of a more comprehensive study where we tune the laser time
lag before the beginning of the electric pulse. These curves reveal interesting features, such as a rapid
evolution at short laser time lag (tlaser lag< 50 µs), followed by a slower dependence at higher tlaser lag
values. Finally, the laser-induced shift of tdelay completely disappears for tlaser laglarger than 4 ms.
Overall, this dependence can be adequately fitted as the sum of two exponential decays
𝑡𝑙𝑎𝑠𝑒𝑟 𝑙𝑎𝑔 𝑡𝑙𝑎𝑠𝑒𝑟 𝑙𝑎𝑔
( ∆𝑡𝑑𝑒𝑙𝑎𝑦 = ∆𝑡1 . 𝑒 𝜏1 + ∆𝑡2 . 𝑒 𝜏2 ). It yields relaxation times in the µs-ms range at 85 K in

GaTa4Se8, more precisely 1 = 6  2 µs and 2 = 770  50 µs. We will see now that more direct
determination seems to converge towards similar values of carrier lifetimes.
3.1.3.2 Car rier lifetime in a Mott insulator revealed by time -

resolved Photoemission Electron Microscopy
As discussed in the first chapter, GaV4S8 is a Mott insulator isostructural with GaTa4Se8, i.e.
exhibiting a lacunar spinel structure (Figure 3.7.a) built from a NaCl-like arrangement of GaS4
tetrahedra and V4S4 cubane-like units. In each V4S4 cluster, one out of the seven V 3d electrons is
unpaired and lies on the threefold degenerate molecular t2 level. In the solid, the three narrow bands
deriving from these t2 molecular levels are split by the on-site electron-electron repulsion, leading to
an insulating state with a Mott-Hubbard gap of 0.3  0.1 eV [103].
During my stay at KEK Synchrotron (K. Fukumoto group), we performed time-resolved
Photoemission Electron Microscopy (tr-PEEM) experiments at room temperature at KEK
experimental setup on a freshly cleaved GaV4S8 single crystal. Figure 3.7.b provides a schematic
description of the electron pathway during a pump-probe PEEM experiment. A first pump laser,
either 1.2 eV / 180 fs / repetition rate 500 Hz or 1.1 eV / 10 ns / 200 Hz, promotes electrons in the
104
conduction band. A second laser, the probe, with a tunable delay with respect to the pump, expels
the excited electrons from the conduction band toward the vacuum level, where electrons are
accelerated and detected. A typical tr-PEEM experiment presents the ratio of photoelectrons
number detected with and without pump laser versus the time delay between the pump and the
probe. Figure 3.7.c shows the results of a first experiment carried out without any pump aiming at
establishing the work function  of GaV4S8 and performed by varying the probe energy between 4
and 4.8 eV. This curve displays a clear threshold effect consistent with a work function close to
 = 4.6  0.1 eV. For the tr-PEEM experiments, we chose a probe energy of 4.47 eV slightly below
the  value, in order to detect only the excited electrons in the Upper Hubbard Band without any
pollution by those in Lower Hubbard Band. A first tr-PEEM experiment performed with the
1.2 eV / 180 fs pump laser reveals a striking feature: there is no detectable relaxation of the signal
shown in Figure 3.7d in the 0-100 ps time windows after the pump laser.
1.2 105
(a) (b) PEEM detection (c)
PEEM signal (probe only)

Delay 1 105
Ga Evac 8 104
Work function
  4.6 eV
Work function 
V
S Probe
CB 6 104
4.47 eV / 600 fs
4 104
Pump
Egap
EF 2 104
1.2 eV / 180 fs
VB or 1.1 eV / 10 ns 0 100
4 4.2 4.4 4.6 4.8
Probe energy (eV)
Pump : 1.2 ev, 180 fs laser Pump : 1.1 eV, 10 ns laser

(d) 1.02 (e) 1.1 (f) 1.04
1.05
 >> 100 ps 1 = 43 ns
2 = 270 ns
PEEM signal (a.u.)
PEEM signal (a.u.)

PEEM signal (a.u.)
2 = 270 ns 1.03
1.01 3 = 4 µs
1.05 3 >> 1µs 1.02
1.01
1 1
1
0.99
-20 0 20 40 60 80 100 0 200 400 600 800 1000 0 2 4 6 8 10
Delay (ps) Delay (ns) Delay (µs)
Figure 3.7: (a) crystallographic structure of GaV4S8 and molecular orbital diagram associated with the V4
tetrahedral cluster. (b) schematic description of the electron pathway in a tr-PEEM experiment. (c) PEEM
energy scan of the probe laser without any pump on a GaV4S8 single crystal. (d-f) tr-PEEM signal vs.
various ranges of the time delay between the pump and the probe in GaV4S8 at room temperature.
In order to confirm this result, the second series of experiments were carried out using the
1.1 eV / 10 ns pump laser. Figure 3.7.e-f shows that the relaxation indeed occurs over a much longer
timescale, with a fair description of the relaxation as the sum of three exponential decay terms with1
= 43 ns, 2 = 270 ns and 3 = 4 µs. By comparison with classical semiconductors [211,212], the long
carriers lifetime found here might look surprising in a Mott insulator possessing likely a high level of
native defects. However, a comparison with other strongly correlated insulators might be more
instructive to clarify its origin. Among the scarce studies of carriers lifetime in such compounds,
extremely different values are observed, going from the sub-picosecond range in the Mott insulator
1T-TaS2 [214] to 1-20 µs in the strongly correlated insulator VO2 [215]. According to [215], a possible
clue is a difference between the nature and symmetry of the valence and conduction bands states. In
105
a “one-band” Mott insulator like 1T-TaS2, the Lower and Upper Hubbard Bands derive from the
same band and therefore display the same symmetries. In this case, the electron-hole recombination
is permitted and hence very fast. Conversely, in a “three-band” Mott insulator like GaV4S8, the top
of the LHB and the bottom of the UHB are most probably made of different and orthogonal V3d
bands (see Suppl. Mater. of [48] and Figure 1.12b). It would result in a mostly forbidden interband
transition and hence in slow electron-hole recombination. This scenario is consistent with the
behavior observed in the Mott insulator (V1-xCrx)2O3, another multiband Mott insulator (see Chapter
1). In this compound, long carriers lifetime indeed appear, allowing the accumulation of electrons in
experiments with electric pulses train with  100µs between each pulse [116]. Conversely, the
scenario proposed above would predict that a single band Mott insulator such as 1T-TaS2 or LaTiO3
would present much lower carrier lifetime. Clearly, further theoretical and experimental work is
required to confirm this conjecture.
Overall, the two techniques used here to evaluate the carrier lifetime in GaV4S8 and GaTa4Se8
(the indirect electro-optical pump / electric probe method and the direct tr-PEEM method) give
rather similar results and show that long carriers lifetime, larger then 1 µs, exist in these Mott
insulators. On the theoretical side, this clearly supports the proposed electronic avalanche scenario
for the Electric Mott Transition. From a more application-oriented viewpoint, such long carriers
lifetime confirm the possibility of using artificial neurons based on Mott insulators [116] with typical
operating frequencies above 10 kHz (i.e. with train of pulses with a typical period between each
exciting pulse lower than 100 µs).
106
3.2 Ultrafast THz-pulse-driven dynamics in Mott insulator

GaTa 4 Se 8
We have seen in Chapter 1 that the basic mechanism explaining the Electric Mott Transition
was the acceleration of free carriers by “DC-like” electric fields. It was early realized that an
interesting consequence of this mechanism is that the electric field carried by intense light pulses
should also induce a carrier multiplication and, finally, a transient metallic state. Thanks to amazing
progress during the last two decades, such ultrafast lasers carrying electric fields larger than 1 MV/cm
are now available. Ultrafast THz lasers are the ideal tool to avoid the direct photo-generation of
electrons and to promote the other mechanisms that will act on them. The typical photon energy of
a THz light pulse (1 THz  33 cm-1  4 meV) is indeed well below the gap of Mott insulators
considered in this study. For all of them, the Mott-Hubbard gap Mott-Hubbard exceeds 100 meV. A
collaboration was hence engaged with E. Abreu and S. Johnson at ETH Zurich on time-resolved
THz pump-probe experiments on several Mott insulator considered in this thesis. Even though I
was mostly involved in the sample preparation (see Figure 3.8.a) and in the discussion of the results
of this work (I did not participate in the THz experiments), we report here the main results of this
THz study. A part of this work was published in Ref. [216].
In this work, THz pulses generated via optical rectification in organic DSTMS crystals are
used, with peak field values of  1 MV/cm for the pump and < 100 kV/cm for the probe. We
choose to study GaTa4Se8, since its low Mott-Hubbard gap (Mott-Hubbard  100-150 meV) might
facilitate the THz-field-driven insulator to metal transition phenomenon. As shown in Figure 3.8.b,
we observe a sub-picosecond THz induced decrease of the THz transmission, consistent with an
ultrafast decrease in the resistivity of the material. This response persists for a few picoseconds,
beyond the duration of the THz pump pulse, as seen by comparing the nonlinear signal dynamics
with the time trace of the pump shown at the bottom of Figure 3.8.b. Also, Figure 3.8.c shows that
the amplitude of the THz transmittance drop increases with the amplitude of the pump electric field.
It demonstrates that a sub-picosecond THz pump pulse is capable of inducing a resistivity drop in
GaTa4Se8, similar to what was observed in static measurements using few tens of microseconds long
electric fields, enabling the study of the dynamics of the electric Mott transition mechanism.
107
(a) (b)
1 mm
300 µm hole for

THz transmittance
reference
GaTa4Se8 crystal
over a 300 µm hole
(c) (d)
THz transm. change (%)

550
Threshold field (kV/cm)

450
T/T0 =
350
250
0 50 100 150 200 250
Temperature (K)
T = 105 K 0.5 1
Epump(MV/cm)
(e) (f) Extrapolation of the 1/tdelay vs E4 dependence
tdelay (ps)
1013 0.1
1.4 60K 1
75K GaTa 4Se1.5Te 6.5
90K 1011 10
GaTa 4Se8
1.3 105K
GaNb4Se8
1 / tdelay (s-1)
150K 109
220K GaV4S8
/0
1.2 GaMo4S8
107
1.1
105
1 103
0 0.2 0.4 0.6 0.8 1 1 10 100 1000
E2 (MV2/cm 2) E (kV / cm)
Figure 3.8. (a) Image of the 150 µm thick GaTa4Se8 crystal, which was polished to get two parallel faces,
and then pasted on top of a 300 µm hole drilled in a 5 x 5 x 0.2 mm stainless steel plate. The second hole,
highlighted in green, is used to establish the total amplitude of the THz pulse. (b) THz induced change in
the transmission of the maximum of the THz probe pulse through the GaTa4Se8 sample cooled down to
150 K, following excitation by a THz pump with ~1 MV/cm peak field. Bottom: temporal trace of the
108
square of the THz pump pulse. [216]. (c) THz induced change in the THz transmission for different
amplitudes of the THz pump at 105 K. Black lines are fits to a double exponential recovery. (d) THz
transmission change T/T0 as a function of pump field and temperature. Inset: attempt to fit the field
dependencies with a model based on the Zener breakdown mechanism. (e) Evolution of the conductivity
ratio /0 versus the square of the THz electric field. (f) Inverse of the delay time tdelay-1 required to induce
a resistive switching plotted versus the applied electric field in standard Electric Mott Transition experiment
in various Mott insulators of the AM4Q8 family (from Supplementary Materials of Ref. [6]). The
extrapolation with the tdelay-1  E4 dependence is indicated with dashed line. It leads to a predicted tdelay of
between 0.1 and 1 ps at 1 MV/cm for the Mott insulator GaTa4Se8.
Figure 3.8.d shows the dependence of the nonlinear signal on pump field and on temperature
in the 40 K – 295 K range. Each point in Figure 3.8.d corresponds to the integral of the nonlinear
signal in 0.5 – 2.5 ps time range. The nonlinear signal increases with pump field and with temperature,
with the exception of the room temperature data. A possible reason is that GaTa4Se8 at 295 K is no
longer in the pure Mott insulator phase, but is far in the crossover region of the phase diagram where
the Mott-Hubbard gap is partially filled due to thermal excitations [217]. Overall it is clear that the
relative transmittance change increases faster than linearly both with the electric field and
temperature. The temperature and field dependence of the signal enable us to discuss the origin of
the carriers generation process that leads to population of the Fermi level in this Mott insulator, and
consequently, to the observed increase in transient conductivity, following the application of a strong
THz electric field. The THz photon energy is < 10 meV, much smaller than the  150 meV bandgap
of GaTa4Se8 . Direct photoexcitation of the electrons into the conduction band is hence negligible,
even if multiphoton absorptions processes are taken into account. Two broad classes of carriers
generation processes must therefore be considered: Zener tunneling and electronic avalanche
processes. As discussed in Chapter 1 (see Figure 1.37.g), the Zener mechanism occurs when the
electric field is strong enough to bend the conduction band, so that tunneling of an electron from
the valence to the conduction band becomes possible. The second mechanism is the electronic
avalanche, a process requiring the presence of a few free carriers in the conduction band. The rare
free electrons can be accelerated by the electric field, and then the carriers multiplication can happen
through two distinct regimes. At low temperatures, when the mean free path is large, an impact
ionization process can occur, involving independent electrons. This mechanism explains the
extraordinary carrier multiplication driven by a THz picosecond electric field pulse in the classical
semiconductor GaAs [218]. At higher temperatures, a collective electronic process, the electronic
avalanche described in Chapter 1 [107] can lead to a divergence of the electronic temperature and
hence to the massive creation of new carriers. In both cases, a threshold electric field appears.
For Zener tunneling, the conductance  which is also proportional to the carrier density,
𝐸
−𝜋 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑
should behave as 𝑒 𝐸𝑝𝑢𝑚𝑝
[219]. The inset of Figure 3.8.d shows the threshold field values Eth
extracted at different temperatures from a fit of the conductivity  with such dependence. A striking
result is that the fitted Eth values vary strongly with temperature, from 500 kV/cm at 50 K to 250
kV/cm at 220 K. This is unexpected for the Zener mechanism: tunneling is indeed a quantum effect
which is inherently independent of temperature.
109
To evaluate the relevance of the electronic avalanche mechanism to explain the tr-pump
probe THz data, it is instructive to check if the logarithm of the conductivity change /0 varies
linearly with the square of the field E², as predicted by the Fröhlich model (see Section 3.1.1.1 of this
manuscript and Ref. [6,107]. Figure 3.8.e, shows that a linear behavior is indeed followed at all
temperatures for THz field below 0.5 MV/cm. However, deviations from the ln(/0)  E² law
seems to appear at lower E field when T decreases. This is at odds with the Fröhlich model in the
dirty limit, since it suggests lower threshold fields at low temperature. Nevertheless, let us keep in
mind that the Fröhlich prediction ln(/0)  E² applies once an equilibrium state is reached between
the electronic and lattice sub-systems under continuous stimulation. It is possible that the duration
of the THz pulse (< 0.3 ps, see Figure 3.8.b) is too short to reach such an equilibrium. For standard
DC experiments, a relation tdelay-1  E4 between the delay time necessary to induce the Electric Mott
transition and the applied DC electric field was proposed in Ref. [6], based on both theoretical and
experimental arguments. Figure 3.8.f shows that the extrapolation of the data observed in DC-like
experiments towards 1MV/cm leads to a predicted tdelay between 0.1 and 1 ps. As the duration of the
1 MV/cm pulse is not larger than 0.1 - 0.2 ps, there is hence a possibility that the THz pulse duration
is not large enough to bring GaTa4Se8 in the true avalanche regime predicted by Fröhlich. However,
we cannot discard the existence of the beginning of the avalanche process, especially due to some
non-linearity observed for the highest fluence in Figure 3.2.e.
The above discussion trying to evaluate the mechanism describing the THz-driven
transmittance (Zener versus Fröhlich-like electronic avalanche) can not be considered as definitive.
One of the possibility suggested in Ref. [216] is that electrons are first promoted in the UHB band,
and then accelerated by the electric field, leading either to an impact ionization or to a Fröhlich
mechanism in the dirty limit. Regardless the question of the mechanism, we note that a variation of
the transmittance of 5 - 10 % driven by strong THz pulses was also observed in Ref. [159] (in the
molecular Mott insulator -(BEDT-TTF)2X) and in Ref. [157] (in V2O3 in the AFI phase). Both
articles claim that such a transmittance change is the signature of a Mott transition. In the present
work on GaTa4Se8, the variation of the transmittance also reaches 5 %, as shown in Figure 3.8.d.
Hence, the firm conclusion that can be drawn from this work is that a non-thermal and purely
electronic effect is driven by a sub-picosecond THz electric field of  1 MV/cm and is able to
transiently promote new hot carriers in the Mott insulator GaTa4Se8. This behavior is strikingly
similar to the phenomenology of the Electric Mott Transition driven by the application of 107-109
times longer (1-100 µs) and 102 times smaller fields ( 10 kV/cm).
110
3.3 Photoinduced insulator to metal transition in

(V 0 . 9 5 Cr 0 . 0 5 ) 2 O 3 -PI and V 2 O 3 -AFI
This section is dedicated to the electrons and lattice dynamics in Mott insulators brought
out-of-equilibrium. However, the previous parts have only addressed the electronic part. Studying
the second face of the problem, i.e. lattice dynamics would ideally require time-resolved X-Ray
diffraction experiments with a sub-picosecond resolution. Such experiments became recently
available but in a very limited number of X-ray sources based on Free Electron Lasers (LCLS at
Stanford, USA; SwissFEL in Switzerland; Spring-8 in Japan). A more accessible alternative is to turn
to optical reflectivity time-resolved pump-probe experiments. Even though optical reflectivity
primarily depends on the electronic degree of freedom, we will see that it is also affected by the lattice
response, which can be unambiguously identified thanks to a temporal response slower than the
electronic one.
a) b)
1
V O (PM-300K)
2 3
0,8 V O :Cr2.8% (PI-300K)
2 3
V O :Cr5%(PI-300K)
2 3
V O (AFI-100K)
Reflectivity
0,6 2 3
pump
0,4
0,2
probe
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6
E(eV)
Figure 3.9: (a) Crystallographic structure and phase diagram of (V1-xCrx)2O3. (b) Reflectivity spectra of
(V1 - xCrx)2O3 in different regions of the phase diagram. The reflectivity of V2O3 PM and AFI, as well as of
(V0.972Cr0.028)2O3-PI are extracted from Ref [220]. The spectrum of (V0.95Cr0.05)2O3-PI was obtained in
collaboration with V. Ta Phuoc at the University of Tours. The open black arrows indicate the energy range
(Epump = Eprobe = 1.55 eV) of most of the time-resolved reflectivity studies on the V2O3 system published so
far. The filled black arrows show the pump and probe energies chosen in works presented in this
manuscript (Epump = 0.89 eV and Eprobe in the 0.1 - 1.45 eV range).
We have hence engaged a time-resolved reflectivity (optical pump – optical probe, OPOP)
study on the (V1-xCrx)2O3 system. This choice was motivated by different arguments:
— first of all, (V1-xCrx)2O3 is a prototypical system that contains the true Mott insulator phase
(PI), but also the magnetically ordered AFI phase, as recalled in the phase diagram shown in Figure
3.9.a,
— despite an increasing interest for the out-of-equilibrium properties of the V2O3 and
(V1 - xCrx)2O3 compounds [143,154–157,169,171,221,222], the tr-OPOP studies in the insulating
phases published so far are very limited both for the PI (Ref. [169,221]) and the AFI phase
(Ref. [155]] and Suppl. Mater. of Ref. [171]), and were performed in experimental conditions
111
favoring thermal effects. All the published tr-OPOP studies indeed used a pump energy of 1.55 eV
(800 nm). This energy, much larger than the gap (gap = 0.2  0.1 eV for (V0.95Cr0.05)2O3- PI and
0.65 eV for V2O3-AFI [ [175]) can potentially lead to a huge thermal heating since the fraction of the
pump energy exceeding the gap value can easily generate incoherent phonons and hence heating.
For example, at least 60% of the pump energy (= (1.55-0.65)/1.55) for the AFI phase (87% for the
PI phase) are potentially converted into heat using the 1.55 eV pump. On the other hand, a single
energy probe, usually also at 1.55 eV, is very often used [155,156,169,221]. A simple view on the
reflectivities displayed in Figure 3.9.b shows that this energy range is not fully appropriate to the
problem: only small R variations are expected at 1.55 eV at the insulator to metal transition, while
much larger changes occur at lower energy, below the gap values. It is hence quite clear that the best
conditions to study a possible non-thermal photo-induced insulator to metal transition starting from
the PI and AFI phases of the V2O3 system require lower excitation energy, ideally just above the gap
value. In parallel, probe energy covering a broad spectral range where large changes of reflectivity
are expected (i.e. between 0.1 and 1 eV, see Figure 3.9.b) is fully necessary in order to clarify if a
photo-induced insulator to metal transition has occurred or not.
— thanks to an ongoing collaboration with Prof. S. Iwai at Tohoku University in Sendai
within an associated international laboratory 3 , we had the possibility to perform a tr-OPOP
reflectivity study in the ideal conditions described above. The Tohoku experiment is indeed based
on a pump at 0.89 eV just above the gap of the AFI phase and a probe in the 0.15 to 1.45 eV spectral
range, with a time resolution of 100 fs. Additional tr-OPOP experiments with an ultimate 6 fs
resolution were also carried out by the Sendai group, but we will not discuss them in this manuscript.
Before describing the studies on (V0.95Cr0.05)2O3-PI and on V2O3-AFI performed in Sendai,
let us first briefly recall the tr-OPOP results already published in the PI and AFI phases on V2O3. In
the tr-reflectivity of (V0.972Cr0.028)2O3-PI [221], B. Mansart et al. have observed strong oscillations of
Δ𝑅 ⁄𝑅 |𝑡𝑟 = ( R 𝑝ℎ𝑜𝑡𝑜𝑒𝑥𝑐𝑖𝑡𝑒𝑑 – 𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 )⁄𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 at the probe energy of 1.55 eV, as shown in Figure
3.10.a. It corresponds to a Brillouin scattering effect: as explained in Section 1.4.1.2.2 of this
manuscript, the pump laser has launched at strain wave propagating at the speed of sound and
starting from the surface of the materials impacted by the pump laser. As there is an abrupt change
of optical index at the position of the strain wavefront, the light of the probe can be reflected and
make constructive or destructive interferences with light reflected at the sample surface. This explains
the damped oscillations in the time response of the reflectivity. According to B. Mansart and co-
workers [221], the simple relation linking the temporal period T of the oscillations, the wavelength
𝜆𝑝𝑟𝑜𝑏𝑒
of the probe laser probe, the speed of sound in the material vs and the refractive index n is (𝑇 = )
2𝑣𝑠 𝑛
gives a vs value in agreement with published results (7.5 km/s along the (100) direction, see
Ref. [223]). Beyond these coherent phonon oscillation effect, we note that the non-oscillating part
of the signal displays a very small positive shift (R/R  + 0.07 %), despite very large pump laser
fluences above 30 mJ/cm². This increase is much lower than that expected for a photo-induced
transition to a full metallic state. Figure 3.10.b indeed shows that the relative difference of reflectivity
between the PI and PM states Δ𝑅 ⁄𝑅|𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 = (𝑅𝑃𝑀 – 𝑅𝑃𝐼 )⁄𝑅𝑃𝐼 at 300 K and 1.55 eV amounts to
3 International Associated Laboratory IM-LED, Impacting Materials by Light and Electric field, and watching real-time Dynamics,
http://www.chem.s.u-tokyo.ac.jp/users/lia_im-led/index-e.html
112
+ 15 %, 200 times more than the observed photoinduced value. However, it is hazardous to
conclude that a PI to metal transition can not be photoinduced in (V1-xCrx)2O3: due to the choice of
pump energy (1.55 eV) almost one order of magnitude larger than the gap (~0.2 eV), huge thermal
effects are indeed expected in such an experiment.
In the AFI phase of V2O3, the only published results (Suppl. Materials of Ref. [171]) also
evidence a very small photoinduced R/R at hprobe = 1.55 eV (+0.3 %, see Figure 3.10.d). Even if
this far from the value expected for a full AFI-PM transformation, we note that the pump fluence
(1 mJ/cm²) was very low in these experiments.
Overall, the conditions used in the published tr-OPOP reflectivities are not ideal for catching
potential non-thermal effects. We have hence engaged a study of both phases using more optimized
experimental conditions, in order to examine the possibility to photo-induce a non-thermal insulator-
to-metal transition in the PI and AFI phases.
(a) (b) (c)

R/R V2O2:Cr2.8% & pure V2O3 R
1
hνpump = hνprobe 0,5 (RPM-RPI)/RPI VO
2 3
= 1.55 eV
0,4 hνpump = hνprobe 0,8 V O :Cr2.8%
2 3
= 1.55 eV 300K
0,3 0,6
0,2
PI 0,1
0,4
0 0,2
0 0,4 0,8 1,2 1,6 0 0,4 0,8 1,2 1,6
E(eV) E(eV)
(d) (e) (f)

R/R pure V2O3 R
1
4 V2O3 (PM-300K)
(R -R )/R
PM-300K AFI-100K AFI-100K
0,8 V2O3 (AFI-100K)
3
hνpump = hνprobe= 1.55 eV 0,6
➔ΔR/R = + 0.3 %
AFI 2
hνpump = hνprobe 0,4
= 1.55 eV
1 E= 0.2 eV 0,2
pump 1 mJ/cm2 ➔ ΔR/R = + 260%
0 0
0 0,4 0,8 1,2 1,6 0 0,4 0,8 1,2 1,6
E(eV) E(eV)
Figure 3.10: (a) Time-resolved optical pump-probe temporal profile of reflectivity change on a
(V0.72Cr0.28)2O3 crystal, measured by B. Mansart et al., from Ref. [192]. Both pump and probe used an 800nm
(1.55 eV) laser wavelength. (b) Steady-State R change between PI (V0.72Cr0.28)2O3) and PM (V2O3),
calculated from (c). (c) Comparison of reflectivity for PI (V0.72Cr0.28)2O3) and PM (V2O3) at room
temperature (from Ref. [196]). (d) Time-resolved optical pump-probe temporal profile of reflectivity
change on V2O3 crystal in the low-temperature AFI phase, measured by E. Abreu et al. (see supplementary
materials of Ref [171]). (e) Steady-State R change of V2O3 crystal between AFI at 100K and PM at 300K,
calculated from (f). (f) Comparison of reflectivity for V2O3 crystal between AFI at 100K and PM at 300K
(from Ref. [196])
113
3.3.1 Evidence of a non-thermal insulator to metal transition induced by a

femtosecond laser pulse in (V 0.95Cr0.05)2O3-PI
At first, we have recorded the time-resolved reflectivity spectrum of a polished

(V0.95Cr0.05)2O3-PI crystal (c-axis perpendicular to the polished surface, see details in Annex 1-E) at
room temperature with a 100 fs time resolution. This work was performed during my stay at S. Iwai
laboratory at Tohoku University. The laser pump photon energy was 0.89 eV, while the probe was
tunable in the 0.16 - 1.45 eV range thanks to an Optical Parametric Amplification and Difference
Frequency Generation (see details in ANNEX 5).
a) T (K) d)
500
crossover crossover
Critical endpoint Critical endpoint
400 Paramagnetic
Paramagnetic Mott Insulator (PI) Paramagnetic
Metal (PM) Metal (PM)
300
Paramagnetic
200 Mott Insulator (PI)
Antiferromagnetic Antiferromagnetic
Insulator (AFI) Insulator (AFI)
0
0.05 0 0 2 4
XCr Pressure (GPa)
b) of (V0.95Cr0.05)2O3-PI at 300 K e) of (V0.95Cr0.05)2O3-PI at 300 K versus
versus
at  T at P = 1 and 1.6 GPa
0,25 0,4
tr-R/R (300K) tr-R/R (300K)
0,2 R/R (350K) 0,3
R/R (400K) R/R GPa
R/R (450K)
0,15 R/R (500K) 0,2 R/R GPa
R/R
R/R
0,1 0,1
0,05
hpump = 0.89 eV 0
0 -0,1
-0,05 -0,2
0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1
E (eV) E (eV)
of (V0.95Cr0.05)2O3-PI at 300 K
c) versus f) (V0.95Cr0.05)2O3 : optical conductivity vs P↑ ( 3 K)
at  T
1000
1,4 tr-R/R (300K)
1,2 R(PM -PI )R(PI )

300K 300K 300K 800
R(PM -PI )/R(PI )
1 600K 300K 300K
 ( cm )
−
600
0,8
R/R
−
0,6
400 0 GPa
0,4  GPa
 GPa
0,2 200  GPa
3.48 GPa
0
0
0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1
E (eV) E (eV)
Figure 3.11: tr-OPOP spectroscopy Δ𝑅 ⁄𝑅 |𝑡𝑟 = ( R 𝑝ℎ𝑜𝑡𝑜𝑒𝑥𝑐𝑖𝑡𝑒𝑑 – 𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 )⁄𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 measured 10 ps after
the photoexcitation (pump energy = 0.89eV, laser fluence =14 mJ/cm2), compared to steady-state
114
measurements Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 = (R (𝑇, 𝑃, 𝑥)– 𝑅𝑃𝐼 )⁄𝑅𝑃𝐼 . (a) and (d) Phase diagrams of (V1-xCrx)2O3
emphasizing the different T, P, x points used for the comparison. Filled diamonds indicated conditions of
time-resolved measurements (x = 0.05, T = 300K and ambient pressure). The choice of the delay of 10 ps
is not crucial for this comparison since Δ𝑅 ⁄𝑅|𝑡𝑟 are weakly time-dependent in the 1-1000 ps range (see
Figure A 5-2 in ANNEX 3.3.2.1).
(b) test of the simple heating scenario: comparison with reflectivity change measured upon heating the
(V0.95Cr0.05)2O3-PI crystal up to 500 K. (this work).
(c) test of a photoinduced transition to a state similar to the V2O3-PM phase (x = 0), both at 300 K and
600 K (from this work and Ref. [196]).
(e) test of photoinduced transition to a state similar to a compressed PI phase (P = 1 GPa) and to a metallic
phase just beyond the Mott IMT line (P = 1.6 GPa) at T = 300 K in (V0.95Cr0.05)2O3 (this work).
(f) Optical conductivity of (V0.95Cr0.05)2O3 at different pressures. The extrapolation signaled by the dotted
line indicates that the critical pressure for the insulator to metal transition of (V0.95Cr0.05)2O3 is located
between 1 GPa (presence of a finite gap) and 1.6 GPa (the gap is closed and a finite conductivity exists at
zero energy). Besides, a gap value of EG = 0.16  0.05 eV can be deduced from optical conductivity data
at ambient pressure in (V0.95Cr0.05)2O3.
We measured the reflectivity under pressure and in the 300- 500 K temperature range in the laboratory of
Vinh Ta-Phuoc at the University of Tours.
Figure 3.11.(b-c) display the time-resolved normalized reflectivity change spectra

Δ ( 𝑝 𝑜𝑡𝑜 𝑐𝑖𝑡𝑒𝑑 ) measured 10 ps after the pump excitation (black
curve). We note first that is positive below 1 eV and reaches +20 % below 0.25 eV, a value
roughly 300 times larger than the published value of at 1.55 eV ( +0.07 %, see Figure
3.10.a). Moreover, clearly increases when energy is lowered. This is the behavior expected
when an insulator to metal transition is photo-induced: by definition, the reflectivity of metal indeed
increases at low energy to reach 1 in the limit E → 0, while that of an insulator tends to a constant
value. Therefore, any differential reflectivity ( 𝑚𝑒𝑡𝑎𝑙 𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 )⁄ 𝑖𝑛𝑠𝑢𝑙𝑎𝑡𝑜𝑟 will necessarily increase
at low E. In order to test this hypothesis, we tried to compare the time-resolved signal with
several possible evolutions of the reflectivity of the PI state in steady-state conditions.
We first tried to evaluate the possibility of simple heating of (V0.95Cr0.05)2O3-PI by the pump
laser. Figure 3.11.b shows the relative variation of the reflectivity spectra upon heating with respect
to ambient temperature Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 = (R 𝑇 − 𝑅300𝐾 )⁄𝑅300𝐾 (the phase diagram on Figure 3.11.a
highlights the measured points at different temperatures). The comparison of Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 with
the time-resolved signal Δ𝑅 ⁄𝑅 |𝑡𝑟 shows that heating the sample up to 500 K yields at maximum R
increase of 5% above 0.15 eV, against more than +20% for Δ𝑅 ⁄𝑅 |𝑡𝑟 . Moreover Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒
becomes negative above  0.5 eV, while Δ𝑅 ⁄𝑅 |𝑡𝑟 remains positive between 0.15 and  1 eV. These
differences seem to exclude a purely thermal nature of the photoinduced state.
On the other hand, Figure 3.11.c shows that Δ𝑅 ⁄𝑅 |𝑡𝑟 is much lower than the reflectivity
difference between (V0.95Cr0.05)2O3-PI and V2O3-PM at 300 K in the true PM phase. It is also smaller
than the reflectivity difference between (V0.95Cr0.05)2O3-PI and V2O3 in the high-temperature
crossover regime at 600 K.
Overall, the most convincing comparison is obtained with the reflectivity change under
pressure Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 (𝑃) = (𝑅𝑃 − 𝑅𝑃=0 )⁄𝑅𝑃=0, as shown in Figure 3.11.e. We indeed found that
115
the main features of Δ𝑅 ⁄𝑅 |𝑡𝑟 , i.e. an increase at low frequency and an amplitude of 20-25 % at 0.2 eV,
also exists for Δ𝑅 ⁄𝑅 |𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 (𝑃) with pressures lying between 1 and 1.65 GPa. Interestingly, these
two pressure values are in the vicinity of the P – induced insulator to metal transition. Optical
conductivity calculated from reflectivity data under pressure indeed indicates a small but finite
residual gap at 1 GPa, and a closed gap with a finite optical conductivity at zero energy at 1.65 GPa
(see Figure 3.11.f). Figure 3.11.d highlight point measured at different pressure on the phase diagram.
To summarize this series of comparisons, it appears that photoexciting (V0.95Cr0.05)2O3-PI results in a
transient reflectivity comparable to the state at the border of the PI-PM transition induced under
external pressure.
(a) Pump (0.89eV)- Probe (0.25eV)

0,25
I (mJ/cm )
2 14 7.88 5.09 3.13
ex 10.15 6.67 4.61 1.06
0,2
0,15
R/R
0,1
0,05
0
0 200 400 600 800 1000
Time (ps)
(b) (c)
Normalized R/R
0,25 1,2
0.1 ps 1
0,2
10 ps
R/R (normalized)
100 ps 0,8
0,15 600 ps
1000 ps
R/R
0,6
0,1
0,4 P (mJ/cm2)
14 5,09
0,05 10,15 4,61
0,2 7,88 3,13
6,67 1,16
0
0
0 5 10 15 -1 0 1 2 3
Time (ps)
P (mJ/cm2)
Figure 3.12: (a) Temporal profiles of transient reflectivity change (Δ𝑅 ⁄𝑅 |𝑡𝑟 at different pump laser fluence
Iex. with a pump at 0.89 eV and a probe at 0.25eV. (b) Fluence dependencies at different times, extracted
from (a). (c) R change dynamics at different pump excitation intensity for the first several picoseconds.
Preliminary measurements on a (V0.95Cr0.05)2O3-PI thin film deposited on a sapphire substrate reveal a
similar overall behavior (see Figure A 5-5 in ANNEX 5).
In order to see the dynamics of the new high reflectivity phase creation, we performed a
pump-probe reflectivity time scan in the 0-1 ns range with 100 fs resolution at different pump
116
fluences (see ANNEX 5). In order to get the largest signal, we chose a probe energy of 0.25 eV,
comparable to the gap of (V0.95Cr0.05)2O3-PI, while the pump was identical to the previous experiment
(0.89 eV). Quite surprisingly, Figure 3.12.a shows that the temporal profile of transient R change
(Δ𝑅 ⁄𝑅 |𝑡𝑟 ) at different fluences can be roughly described beyond 1 ps as a simple plateau without
noticeable relaxation over a very extended time scale 1-1000 ps. Also, we observe a linear evolution
of Δ𝑅 ⁄𝑅 |𝑡𝑟 with the fluence (see Figure 3.12.b), without indication of a threshold effect.
One of the deviations from the simple description of Δ𝑅 ⁄𝑅 |𝑡𝑟 as a step-like plateau is a fast
relaxation of Δ𝑅 ⁄𝑅 |𝑡𝑟 within less than 500 fs for fluences below 6 mJ/cm² (see Figure 3.12.c). Due
to this short time scale, this behavior can be attributed to a purely electronic relaxation process
following the initial excitation by the pump. The second deviation from a simple step-like plateau of
Δ𝑅 ⁄𝑅 |𝑡𝑟 is a small-amplitude oscillation that can be detected at the highest fluence during the first
200 ps (Figure 3.12.a). By comparison with the work of Ref. [221] shown in Figure 3.10.a, these
oscillations might be ascribed to a Brillouin scattering effect. The basic mechanisms inducing
Brillouin oscillations (i.e. the launching of a coherent acoustical wave from the surface induced by
the laser pump and the interferences between the probe light reflected by the front wave propagating
at the speed of sound and the surface sample) are recalled in Section 1.4.1.2.2 of this manuscript.
a) b)55
0,3 Pump Probe
0.88eV 0.26eV (4.8m) 0.70eV (1.8m) 50 T
exp
(1.44m) 0.55eV (2.3m) 1.07eV (1.2m)
0.60 eV (2.1m) 1.43eV (0.87m) T
calc
45
0,2
40
T (ps)
-0,025
R/R
0,1 35
30
R/R
0 25 1.43 eV
-0,05
20 0 Time (ps) 40
-0,1 15
0 20 40 60 80 100 0,8 1 1,2 1,4 1,6 1,8 2 2,2
Time (ps)  (m)
probe
Figure 3.13 (a) Time profiles of transient differential reflectivity Δ𝑅 ⁄𝑅 |𝑡𝑟 at different probe wavelengths.
The pump laser fluence was 14mJ/cm2. (b) Comparison of oscillations period T extracted by fitting the
2𝜋
curves shown in (a) with a dependence Δ𝑅 ⁄𝑅 |𝑡𝑟 = 𝐴. sin ( 𝑇 𝑡 + 𝜑) 𝑒 −𝑡⁄𝐵 + 𝐶 (red curve) and estimated by
formula where vs = 8nm/ps (from Ref. [183])and n = 2.5 (average estimated value for the spectral
region) (purple curve).
A usual way to check this hypothesis consists of performing transient reflectivity

measurements at different wavelengths. If a Brillouin scattering mechanism is at play, then the
relation 𝑇 = 𝜆𝑝𝑟𝑜𝑏𝑒 ⁄2𝑣𝑠 𝑛 should exist between the oscillation period T, the wavelength of the probe
laser probe, the speed of sound in the material vs and the refractive index n [193]. We have hence
measured the time profiles at different wavelengths (see Figure 3.13a). We have extracted the period
T by a simple fitting procedure (see inset of Figure 3.13.b and corresponding caption). Figure 3.13.b
represents the extracted T value as a function of the probe wavelength (red curve) and compares it
117
with a simulation (blue curve) supposing a sound velocity vs = 8 nm/ps [188] and n = 2.5 (average
estimated value for the 0.9 – 2 µm spectral region). The predicted and measured oscillation periods
are consistent, supporting the scenario of acoustic phonon propagation after photoexcitation of the
system.
At this stage, it is interesting to discuss the mechanisms that may explain the surprising
plateau-like behavior of Δ𝑅 ⁄𝑅 |𝑡𝑟 in (V0.95Cr0.05)2O3-PI. First of all, we have seen that we can exclude
the scenario of the simple trivial thermal mechanism (see Figure 3.11.a and associated discussion).
A possible explanation in that photo-excitation by the pump creates a “metallic-like“ state within less
than 1 ps (before a full response of the lattice) and that the photocarriers have a very long lifetime
(in the 10 ps to 1 ns range). This explanation is consistent with the long carrier lifetime observed in
the other canonical Mott insulators GaTa4Se8 and GaV4S8 (see Section 3.1.3 of this manuscript). In
this scenario, the response of the lattice in the Δ𝑅 ⁄𝑅 |𝑡𝑟 signal is essentially hidden by the strong
electronic response. It is hence difficult to argue if the strain wave pathway mechanism recently
observed in Ti3O5 [152] plays a decisive role in the generation of the Mott insulator to metal
transition evidenced in the tr-OPOP data shown in Figure 3.11, Figure 3.12 and Figure 3.13. Finally,
we note that pump energy h = 0.89eV used in this study was not fully optimal since it exceeds by
far the gap EG = 0.16  0.05 eV (see Figure 3.11.d) of (V0.95Cr0.05)2O3-PI. This might favor thermal
effects on top of a possible strain wave mechanism.
Interestingly, the problem of the large energy difference between the pump and gap is absent
in the case of V2O3 in the AFI phase, where the gap of 0.65 eV is much closer to the pump energy
of 0.89 eV. We will hence present in the next section the results of a tr-OPOP study carried out on
V2O3 in the AFI phase.
3.3.2 Photoinduced insulator to metal transition in V2O3-AFI: evidence for

a non-thermal strain wave pathway
As discussed previously, V2O3 undergoes a complex phase transition at 160 K involving:

• an order-disorder magnetic transition, from an antiferromagnetic state at low T to a
paramagnetic state above the transition,
• a structural transition from a monoclinic phase (space group I2/a) at low T to the corundum
phase (𝑅3̅𝑐) at high T, accompanied by a drastic volume change (Δ𝑉 ⁄𝑉 |𝐴𝐹𝐼→𝑃𝑀 = − 1.4 %,
• an insulator to metal transition.
The tr-OPOP study of V2O3-AFI was performed during T. Amano's [224] master thesis in
S. Iwai laboratory at Tohoku University in Japan. The time-resolved measurements were carried out
on V2O3 crystals and thin films that we previously prepared at IMN (see details in ANNEX 1). The
analysis of data presented below corresponds to a joint reflection between the group of S. Iwai in
Sendai, the group of the Institute of Physics of Rennes, and the IMN group, all three involved in the
IMLED International Associated Laboratory3. We recall here the main results of this study.
In contrast to the PI-PM transition in Cr-doped V2O3, the AFI-PM transition in pure V2O3 can be
achieved by a simple photothermal mechanism. In order to avoid this trivial effect, appropriate
118
experimental conditions, such as the base temperature and pump photon energy, can be chosen. As
discussed above, the pump energy was chosen at 0.89eV, just above the gap EG = 0.65 (see
Ref. [175]) in order to limit the creation of incoherent phonons and hence of lattice heating. Also,
most of the measurements were performed at 10K, far below the AFI → PM transition temperature
TAFI→PM = 160 K. In such a way, even if the absorbed light is fully converted into heat, the total
energy will not be sufficient to cross the AFI-PM transition thermally, at least for laser fluences below
6-8 mJ/cm² (see discussions below around Figure 3.16 ). The experimental setup was the same as in
the previous Section 3.3.1 (for details of the experimental setup, see ANNEX 5).
3.3.2.1 Transient reflectivity on a V 2 O 3 -AFI single cr ystal
A first study of the time-resolved reflectivity was engaged on a polished V2O3 single crystal.
Its a-axis direction makes an   25° angle with the direction perpendicular to the surface. Figure
3.14.a-c compares the time-resolved spectra Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝐸) at different delays after photoexcitation
phases with the steady-state reflectivity change Δ𝑅 ⁄𝑅 |𝑃𝑀 𝐴𝐹𝐼 = (R 𝑃𝑀 𝑅𝐴𝐹𝐼(100𝐾) )⁄𝑅𝐴𝐹𝐼(100𝐾) across
the AFI→PM transition. Here is either the reflectivity at 300 K at ambient pressure or the
reflectivity at 40K under the pressure of 3.56GPa (both conditions are in the PM state; see phase
diagram on Figure 3.14.d). The reflectivity spectra are similar in both cases (see Figure 3.14.b).
While Δ𝑅 ⁄𝑅 |𝑃𝑀 𝐴𝐹𝐼 is larger than +200 % below 0.6 eV (Figure 3.14.c) and tends to increase
at low energy, Δ𝑅 ⁄𝑅 |𝑡𝑟 remains lower than + 60 % and presents a broad maximum at 0.5 eV before
decreasing at lower energy. The transient spectra in the AFI state are hence very different from those
in the PI state: while Δ𝑅 ⁄𝑅 |𝑡𝑟 (PI) are almost similar to steady-state Δ𝑅 ⁄𝑅 |𝑃𝐼−𝑃𝑀 obtained under
pressure (see Section 3.3.1), the too small amplitude and the spectral response of Δ𝑅 ⁄𝑅 |𝑡𝑟 (AFI)
clearly indicates that a metallic state is not reached. Conversely, we note that the Δ𝑅 ⁄𝑅 |𝑡𝑟 spectra are
not very different from that obtained at steady-state under a pressure of 1.7 GPa, i.e. in a compressed
AFI state below the critical pressure for AFI→PM transition around 2.3 GPa (see Figure 3.14 c).
This might suggest that the transient crystal state is related to one of V2O3 under pressure but still in
the AFI state.
Another striking difference between the time-resolved reflectivities in the PI and AFI states
is the temporal evolution. First, no clear signature of Brillouin oscillations seems to appear, in
contrast to the PI phase. A possible explanation is that their amplitude, always smaller than 2% in
the PI phase here simply hidden due to the higher Δ𝑅 ⁄𝑅 |𝑡𝑟 values observed here. Second, unlike in
(V0.95Cr0.05)2O3-PI, the initial subpicosecond reflectivity rise due to electron photoexcitation is
followed in V2O3-AFI by a second slow R increase of larger amplitude with a typical time rise of
10 – 30 ps, as shown in Figure 3.14.c. Interestingly, this slow rise of Δ𝑅 ⁄𝑅 |𝑡𝑟 varies almost linearly
with time (or deviates slightly in a sublinear way) during the first 10 ps, as shown in Figure 3.14.a. A
linear behaviour is expected for a phase conversion is driven by the propagation of a strain wave
starting from the surface. Indeed, as a first approximation, the amount of phase transformed
increases linearly with time in this scenario, and so will the time-resolved reflectivity. Also, a simple
but striking observation is that the slow component of Δ𝑅 ⁄𝑅 |𝑡𝑟 is increasing toward positive values with
time. In the “strain wave phase transformation” scenario, this can be explained only by a compressive
strain let behind the front wave. A tensile strain yielding a larger unit cell volume would lead to a
119
smaller bandwidth, an increase of the gap, and hence to the evolution of Δ𝑅 ⁄𝑅 |𝑡𝑟 towards negative
values.
(a) (b) 1
d)
PM-300K T(K)
V2O3 crystal
0,8
3.5GPa (PM-40K)
1.7GPa (AFI-40K) PI PM
Reflectivity
AFI-100K
0,6
300
0,4
0,2
0 200
(c)
4 (R -R
AFI(1.7GPa:40K) AFI:100K
)/R
AFI:100K V2O3
3
(R
PM(3.5GPa:40K)
-R
AFI:100K
)/R
AFI:100K
AFI
R/R
100
2
1
0
0 0,2 0,4 0,6 0,8 1
E (eV) 0 2 4
Pressure (Gpa)
(e) (f)
0.1 ps
0,5
10 ps
100 ps
0,4 300 ps
1000 ps
0,3
R/R
0,2
0,1
0 2 4 6
2
I (mJ/cm )
ex
Figure 3.14: (a) Transient spectrum of Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑅𝑝ℎ𝑜𝑡𝑜𝑒𝑥𝑐𝑖𝑡𝑒𝑑 )⁄ measured on a

polished V2O3–AFI crystal (see ANNEX 1-E) at 10K for several delays after pump photoexcitation
(Iex = 4.1 mJ/cm², hνpump=0.89 eV). The experimental conditions are detailed in ANNEX 5. (b) Steady-state
reflectivity spectra of pure V2O3 at 300K (PM phase), at 100K (AFI state, from Ref [220]) and at 40K
under a 1.7 GPa (3.5 GPa) hydrostatic pressure in the compressed AFI (PM) phases. The pressure-induced
PM spectrum at 3.5 GPa is similar to the PM spectrum at 300 K and ambient pressure. (d) Schematic phase
diagram of V2O3 highlighting steady-state measurements under pressure (c) Reflectivity change for
Δ 𝑃𝑀 𝑃𝑀 3.5𝐺𝑃𝑎/ 𝐴𝐹𝐼 100𝐾 𝐴𝐹𝐼 100𝐾 and
Δ 𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑒𝑑 𝐴𝐹𝐼 𝐴𝐹𝐼 1.7𝐺𝑃𝑎/ 𝐴𝐹𝐼 100𝐾 𝐴𝐹𝐼 100𝐾 from (b). (e) Temporal profiles of
transient reflectivity Δ𝑅 ⁄𝑅|𝑡𝑟 at 10 K for various fluence Iex (pump energy = 0.89 eV, probe energy of
0.16eV). (f) Evolution of Δ𝑅 ⁄𝑅 |𝑡𝑟 as a function of the pump fluence, extracted from at 0.16eV probe.
(tr. measurements perf. by T. Amano [224])
In this context, two possible scenarios rationalizing the tr-reflectivity of the V2O3-AFI crystal
can be proposed:
120
(i) the amplitude of the strain is too small at the laser fluences used in the experiment
(< 7mJ/cm2), and the crystal remains in a “compressed AFI state”,
(ii) overall, the mechanical energy, converted from the light energy of the pump laser
thanks to the mechanism described by Thomsen (see page 48 of this manuscript),
would be sufficient to induce an AFI to PM transition. However, this AFI → PM
transition involves a drastic volume change, which could be forbidden by a
“clamping effect”. In the Thomsen mechanism, the initial stress is indeed purely 1D
and oriented along the z-direction. In a solid, the strain that will follow after a delay
in time (see Section 1.4.1.2.2) along the z-axis should be accompanied by a strain in
the perpendicular (x,y) direction due to the Poisson coefficient. Since the laser pump
beam size is smaller than the crystal lateral (x,y) size, this degree of freedom is
blocked in the single crystal by the non-irradiated parts surrounding the laser
footprint.
There is a very simple possibility to test the scenario of a strain-wave-driven AFI-PM
transition prohibited by a clamping effect in the V2O3-AFI single crystal: turning to polycrystalline
thin film of V2O3 not too dense, so that the (x,y) lateral expansion is allowed.
3.3.2.2 Transient optical ref lectivity on polycr ystalline thin films

of V 2 O 3 -AFI
3.3.2.2.1 Experimental results on V2O3-AFI thin films
A series of measurements was hence performed on two polycrystalline V2O3 thin films of
thickness 100 nm and 1 µm deposited on a sapphire substrate. The film preparation is detailed in
ANNEX 1-C.
Interestingly, the time-resolved spectra Δ𝑅 ⁄𝑅|𝑡𝑟 measured at 10 K in both films is very similar
to the (𝑅𝑃𝑀 − 𝑅𝐴𝐹𝐼 )⁄𝑅𝐴𝐹𝐼 at steady-state (see Figure 3.15(a-3) and Figure 3.15(a-2) for the 100 mn
film, and Figure 3.15(d-3) and Figure 3.15(d-2) for the 1 µm one). Unlike in the single crystal, a state
with a reflectivity close to PM phase at 200-300K was hence successfully induced in both films
100 nm and 1 µm V2O3 films. Let’s recall here that the reflectivity under pressure of V2O3 after the
AFI - PM transition is similar to the PM reflectivity at 200-300 K at ambient pressure (see curve at
P = 3.5 GPa in Figure 3.14.b). It means that the transient reflectivity Δ𝑅 ⁄𝑅 |𝑡𝑟 is also similar to that
of the PM state induced under pressure in V2O3-AFI.
Moreover, a new phenomenon appears in the fluence dependence of Δ𝑅 ⁄𝑅|𝑡𝑟 . Figure 3.15.b
and .e indeed unveil a threshold fluence around 1.5-2.0 mJ/cm2. Below the threshold, Δ𝑅 ⁄𝑅 |𝑡𝑟 (t)
essentially relaxes after the initial sub - picosecond electronic excitation. Conversely, a slow rise of
Δ𝑅 ⁄𝑅 |𝑡𝑟 of large amplitude and with a long characteristic time (10-40 ps) appears for fluences above
the threshold, reminiscent of that observed in a single crystal (see Figure 3.14.c). As for the single
crystal of V2O3-AFI, this increase of Δ𝑅 ⁄𝑅|𝑡𝑟 toward positive values with time is compatible with the
propagation of a compressive strain with the front wave in a “strain wave phase transformation”
scenario.
121
(a) d)
V2O3 thin film, 100 nm (1) V2O3 thin film, 1 µm
(1)
R/R (0.25 eV) = 1.60 (2)

(2)
(3) (3)
(b) (e)
1,6
R/R (0.25 eV) = 1.6
expected full metallic state
1,2
R/R
0.1 ps
0,8 1 ps
10 ps
0,4 100 ps
300 ps
1000 ps
0
0 2 4 6 8 10 12
Iex (mJ/cm 2)
(c) (f)
Figure 3.15: Transient differential reflectivity measured at 10 K in the AFI state on V2O3 thin films of
thickness 100 nm (left part) and 1 µm (right part). (a-1) Steady-state reflectivity of a 100 nm V2O3 thin film
on a Al2O3 substrate measured at 66K in the AFI phase and at 200K in the PM state at the University of
Nagoya. (a-2) Steady-state differential spectrum Δ𝑅⁄𝑅|𝐴𝐹𝐼−𝑃𝑀 = (R 200𝐾 – 𝑅66𝐾 )⁄𝑅66𝐾 calculated from 1; (a-3)
Transient Δ𝑅⁄𝑅|𝑡𝑟 = (𝑅𝑒𝑥𝑐 – 𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 )⁄𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 spectra at 80K in the AFI phase at different delays after
pump photoexcitation (Iex = 7.4 mJ/cm²). (b) Evolution of Δ𝑅⁄𝑅|𝑡𝑟 versus pump fluence at different delays
at 10K. The red line indicates the value of an expected conversion to the full metallic state deduced from
the steady-state Δ𝑅⁄𝑅|𝐴𝐹𝐼−𝑃𝑀 . A clear threshold for the second rise is observed at fluences higher than 1.5
to 2 mJ/cm². (c) Temporal profiles of the transient differential reflectivity Δ𝑅⁄𝑅|𝑡𝑟 at different pump laser
fluence Iex. (hpump = 0.89 eV) measured at a probe energy of 0.25eV. (d) - (f) : similar to (a) - (c), but for
the 1 µm film. The oscillations in the Δ𝑅⁄𝑅|𝑡𝑟 spectra results from interference effects between light
reflected at the film surface and at the film-substrate interface (tr. measurements perf. by T. Amano [224])
122
3.3.2.2.2 Discussion about the photo-induced

insulator to metal transition in V2O3-AFI thin film
• Mechanism of the AFI to PM transition: energy aspects

As already mentioned in Eq.(1.3) in Chapter 1, there is a simple link between the fluence of
the pump laser Iex and the volumic energy E(z) deposited at a given depth z from the surface.
𝐼𝑒𝑥𝑐. [𝐽⁄𝑐𝑚2 ] −𝑧⁄𝜉
𝐸 [𝐽⁄𝑐𝑚3 ] = (1 − 𝑅) 𝑒 (3.3)
𝜉
where R is the reflectivity at the pump energy 0.89 eV (R = 0.21, see Figure 3.9.b and  the laser penetration
depth at 0.89 eV ( = 500 nm). Eq.(3.3) indicates that the volumic energy deposited at the surface (z = 0)
for the experimentally observed threshold fluence Iex = 1.5 mJ/cm² is around 23 J/cm3 (see Figure 3.16.e).
This value is much lower than the 100 J/cm3 required to heat the film from 10 K to the AFI-PM transition
temperature (see Figure 3.16.c), but remarkably close to the 14 J/cm3 required to compress V2O3 up to the
AFI-PM transition line at 10 K (see Figure 3.16.d). This analysis of the energy required to induce an
AFI-PM transition using thermal and compression pathways is interesting since it rules out an
Avrami-like scenario for the observed insulator to metal transition in our experimental conditions.
This mechanism indeed requires enough thermal energy to heat the system above the transition
temperature, where the growth of the new phase begins.
As an Avrami-like mechanism is not able to explain the observed photoinduced insulator to
metal transition in V2O3-AFI films at 10 K, it is worthwhile to consider the strain wave mechanism
from the viewpoint of energy required to induce the transition. First, we note that a threshold fluence
is expected in the strain wave scenario, but specifically in the case of a compressive strain let behind the
front wave. A compressive strain indeed allows moving towards the right side of the phase diagram
shown in Figure 3.9.a and Figure 3.14.d. The threshold simply corresponds to a situation where the
energy deposited by the laser, converted into a mechanical elastic energy thanks to the mechanism
described by Thomsen (see Section 1.4.1.2.2), is large enough to compress V2O3 up to the AFI-PM
transition line, as indicated by red arrows in Figure 3.16.a-b. As mentioned above, this energy
corresponds to 14 J/cm3, which is very close to the 23 J/cm3 of total energy deposited at the sample
surface at the experimentally observed threshold fluence of 1.5 mJ/cm² (see Figure 3.16.e). This
close proximity between the observed threshold fluence and the predicted one in the strain wave
scenario is all the more impressive that the pump laser energy is almost 30 % larger than the gap of
V2O3 (0.89 eV against 0.65 eV). Therefore, it seems reasonable to consider that 30 % of the laser
energy will hence be dissipated into heat, while 70 % will be converted into mechanical energy by
the Thomsen mechanism. Under this hypothesis, the predicted threshold for the total deposited
energy becomes 19 J/cm3 (or equivalently a threshold fluence Iex = 1.2 mJ/cm²), very close to the
experimental value of 23 J/cm3 (Iex =1.5 mJ/cm²). We note here that this process is rather energy-
efficient: at the threshold value, the maximum density of absorbed photons, or equivalently of
excited electrons, corresponds to only  3 photons per 1000 vanadium atoms.
123
(a ) (b)
0 2.2 GPa
(c) Thermal scenario (d) Compression scenario

4
2.2 𝐺𝑃𝑎
𝑃𝑑𝑉 = 14 𝐽/𝑐𝑚3
3 0
Pressure (GPa) 2
P.V =
40 J/cm 3
1
coex.
PM PM+AFI AFI
0
0.96 0.97 0.980.988 1
Normalized unit cell volume V/V 0
(e)
Edepmax = 23 J/cm3
Energy deposited (J/cm3)
20.00
𝐼𝑒 . 𝑚2
𝐸 3 = 1−𝑅 e-z/ 𝜉
𝑐𝑚 𝜉
10.00
 = 500 nm
R = 0.21
Iex = 1.5 mJ/cm²

0.00
0 200 400 600 800 1000
z (depth, in nm)
Figure 3.16: temperature – pressure (a) and temperature – volume (b) phase diagrams of V2O3, according
to Ref. [27]. (c) Evaluation of the energy necessary to heat V2O3 from 10 K in the AFI phase to the PM
phase above 160 K (thermal pathway indicated by the blue arrows in (a-b)). According to published specific
heat data on V2O3 [225], 100 J/cm3 are required to heat the sample from 10 K to transition temperature.
Then the latent heat of the transition is 70 J/cm3. (d) Evaluation of the energy required to compress V2O3
at 10 K across the AFI-PM transition line (compression pathway indicated by the red arrows in (a-b)).
Based on the volume at ambient pressure and at 2.2 GPa just before and just after the pressure-induced
AFI-PM transition shown in (b), it is easy to determine the energy required to compress V 2O3 up to the
transition line (14 J/cm3) and the one to cross the transition (40 J/cm3). (e) Simulation of the total energy
124
E absorbed in V2O3 as a function of the depth z in the sample after a laser pulse at various fluence Iex. The
surface hit by the laser is located at z = 0. The z-dependence of E is given by:
2
𝐼𝑒 . [ ⁄𝑐𝑚 ]
𝐸 [𝐽⁄𝑐𝑚3 ] = (1 − 𝑅) 𝑒 −𝑧⁄𝜉 . R is the reflectivity at the pump energy 0.89 eV (R = 0.21, see
𝜉
Figure 3.9.b) and  the laser penetration depth at 0.89 eV ( = 500 nm).
Beyond the volumic energy deposited at the surface by the laser ( represented by Eq.(3.3))
the evolution of E(z) versus the depth z inside of the sample, is quite illuminating to understand and
predict the fluence dependence of the transient Δ𝑅 ⁄𝑅 |𝑡𝑟 . Under the hypothesis that the cooperative
elastic effects play a minor role, there is indeed a one to one correspondence between the initial
energy deposited at a given depth at t = 0 by the pump laser (more precisely, the fraction of laser
energy converted into “mechanical” energy Em(z) at depth z) and the final state of the material at
depth z sufficiently long after the passage of the strain wave. Figure 3.17 explores this idea by plotting
the predicted final state for different pump fluences. It is easy to understand from the Em(z) plot that
we expect not only one threshold fluence but two. The first threshold corresponds to the energy
necessary to compress V2O3-AFI up to the AFI-PM line (14 J/cm3, see red arrows in Figure 3.16.a.b)
and the second one to transform it completely into the PM phase (a supplementary energy of 40
J/cm3 is necessary, yielding a total of 54 J/cm3). Therefore three different regimes can be expected
depending on the pump laser fluence, as shown in Figure 3.17:
✓ regime 1: for deposited mechanical energies below the low energy threshold (Emth1
= 14 J/cm3, see Figure 3.17.a), the state let behind the front wave is only a
compressed AFI phase. In this approach, one predicts the relaxation of the unit-cell
volume with z shown schematically in Figure 3.17.b. Experimentally this situation
occurs below the threshold fluence Iex = 1.5 mJ/cm².
✓ regime 2 (see Figure 3.17.c): for deposited mechanical energies Em(z) at z = 0
intermediate between the low and high energy thresholds 14 J/cm3 and 54 J/cm3,
two domains in z can be defined long after the passage of the front wave. From the
surface up to the depth z2 where Em(z2, t=0) = 14 J/cm3, there is a coexistence
between the PM and the compressed AFI (c-AFI) phases. The unit cell volume of
both phases remains constant between the surface (z=0) and z2. However, the
fraction of PM phase decreases with z to vanish at z=z2. The fraction of AFI phase
evolves in the opposite way and becomes 100 % at z = z2. Beyond z = z2, only the
compressed AFI phase is present, with a unit cell volume evolving with z similarly
to regime 1.
125
80.00
"Mechanical" energy depostied Em (J/cm3)

70.00
PM
60.00
54 J/cm3
50.00
AFI+PM
(a) 40.00
30.00
20.00
14 J/cm3
10.00
AFI
0.00
0 200 400 600 800 1000
V/V0 z (nm)
1
0.988 AFI PM
Regime 1 :
(b) 100
Em < compressed-AFI (c-AFI)
14 J/cm3 0.970
Phase fraction (%)

V/V0
1
0.988
Regime 2:
(c) 14 < Em < PM + c-AFI c-AFI
54 J/cm3 0.970
V/V0
z2
1
Regime 3 :
(d) Em >
0.988
c- 0
c-PM PM + c-AFI
54 J/cm3
0.970 AFI
z3a z3b
Fluence Iex
0
z
Figure 3.17: (a) Evolution of the energy deposited by the pump laser with depth z and converted to an
efficient mechanical elastic energy Em for different fluences Iex. The two threshold values for Em are
indicated, 14 J/cm3 to compress V2O3 up to the AFI-PM transition line and 54 J/cm3 (i.e., an additional
40 J/cm3 of enthalpy change at the transition, see Figure 3.16.d) to convert it into the PM phase. The laser
fluences indicated in the plot make the reasonable hypothesis that 63 % of the deposited light energy is
converted into mechanical elastic energy.
(b-d) Prediction of phases presents long after the passage of the strain wave, based on the energy deposited
at t = 0 and converted into compressive mechanical energy Em. The y-axis and the schematic curves
correspond to unit cell volume of the compressed AFI and PM phases normalized to its value at ambient
pressure in the AFI state. (b) In regime 1, Em(z=0, t=0) is below the threshold energy 14 J/cm3 , resulting
in a compressed AFI phase. The unit cell volume relaxes towards its zero pressure value when z increases.
(c) In regime 2, Em(z=0, t=0) is between 14 and 54 J/cm3. A first domain starting from the surface contains
a coexistence of PM and compressed AFI phases with z-independent unit cell volumes. At large z, a single-
phase region made of compressed AFI phase appears, similar to that of regime1. (d) In regime 3, Em(z=0,
t=0) is above 54 J/cm3. Compared to regime 2, an additional domain appears close to the surface, made of
compressed PM phase.
126
✓ regime 3: for deposited mechanical energy Em(z=0, t=0) above the high energy
threshold 54 J/cm3, the phase let behind the front wave close to the surface is no
longer a phase coexistence, but a pure and compressed PM phase. In this domain,
the unit cell volume varies with z with the strongest strain at the surface z=0.
However, a second domain appears beyond a depth z3a corresponding to Em(z3a,
t=0) = 54 J/cm3. A domain of phase coexistence PM - compressed AFI, similar to
that of regime 2, appears for z > z3a. Finally, the third domain of pure compressed
AFI phase will appear at larger z, for z > z3b defined as Em(z3b, t=0) = 14 J/cm3
similar to regime 1.
Theoretically, the high energy threshold is predicted at a fluence  4 times larger the low
energy one, since it is simply the ratio between the total energy to bring V2O3 from ambient pressure
to the PM state (54 J/cm3) and the one to bring it just below the AFI-PM line (14 J/cm3), as shown
in Figure 3.16.d. In our case, this would correspond to an experimental fluence of
4  1.5 = 6 mJ/cm². We will see below that deviations to a common fluence-independent behavior
of the slow rise components of Δ𝑅 ⁄𝑅 |𝑡𝑟 start to appear for fluence larger than 6 mJ/cm².
• Mechanism of the AFI to PM transition: dynamics aspects
All of the above discussion attempts to make a link between the scenario "strain wave as the
mechanism driving the AFI - PM transition in V2O3" and the energy required to induce this
transition. The good adequacy between the predicted and experimentally observed amount of energy
to reach the threshold energies is quite striking. However, the discussion so far has mostly focused
on the fluence dependence of Δ𝑅 ⁄𝑅 |𝑡𝑟 and not much on the dynamics.
Figure 3.18 shows the transient reflectivity renormalized between 0 and 1 for the slow rise
part for fluences above the threshold value: Δ𝑅 ⁄𝑅 |𝑡𝑟 is set to zero after  0.5 ps (beginning of the
slow rise) and to 1 at a long time (t  50 ps). As discussed for the crystal, Δ𝑅 ⁄𝑅|𝑡𝑟 varies linearly (or
in a slightly sublinear way leading to a negative curvature of Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑡)) during the first picoseconds.
Once again, this behavior is at odds with an Avrami-like mechanism which predicts that the dynamic
𝑛
evolution of the volume fraction f(t) of the new phase evolves as 𝑓(𝑡) = 1 − 𝑒 −𝐾𝑡 . The exponent n
depends on different factors (dimensionality, growth mechanism, see Ref [145]), but n = 2 values
are often invoked [143]. For small values of f(t), i.e. at short time after the pump, Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑡) is roughly
proportional to f(t). As a consequence, the slow rise time of Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑡) should present, at a short
time, a positive curvature and should typically vary as R/R(t)  t² (prediction for n = 2). As
mentioned above, the experiment is in contradiction with the Avrami prediction for n = 2.
Conversely, the linear or slightly sublinear variations observed in Δ𝑅 ⁄𝑅|𝑡𝑟 at a short time are
fully expected in a strain wave model. The amount of new phase indeed grows linearly with time
thanks to the propagation of the planar front of the strain wave at the speed of sound. For the same
reason and unlike in the Avrami scenario, the characteristic time of growth in the strain wave scenario
should not depend much on fluence. Figure 3.18.a (100 nm film) and Figure 3.18.b (1 µm) show that
this prediction is essentially verified. All the transient Δ𝑅 ⁄𝑅 |𝑡𝑟 (t) at fluences above the threshold fall,
after renormalization at the beginning and at the end of the slow rise process, into a single master
plot. The small deviation at the highest fluences (5.9 and 7.7 mJ/cm²) on the 100 nm film might
simply result from a probe penetration depth becoming small enough when a large fraction of PM
127
phase is transformed behind the strain wave. This deviation above 6 mJ/cm² might result from the
creation of a pure PM phase close to the surface (instead of a coexistence PM +c-AFI at
Iex < 6 mJ/cm², as discussed above.
Another approach to evaluate the scenario of strain wave driving the AFI –PM transition is
to simulate the transient evolution of Δ𝑅 ⁄𝑅 |𝑡𝑟 . However, a quick look at Figure 3.17.b-d makes clear
that it is not an easy task: depending on depth z and time, multiple states are expected, including the
coexistence of phases. In this context, we must be cautious about the results of simulations since a
rather large number of hypotheses are necessary to make the simulation possible. Nevertheless, we
tried to launch several simulations of transient Δ𝑅 ⁄𝑅 |𝑡𝑟 of (V1-xCrx)2O3 in AFI and PI phase using the
Transfer-Matrix-Method (TMM) [226]. This simulation work has also involved Alix Volte from the
Institute of Physics of Rennes [227] and Corentin Logé [228] during its two-month internship in
our group.
(a) (b) 1
1 film 100 nm, T = 10 K film 1 µm, T = 10 K
(slow component, renormalized)
(slow component, renormalized)

0.25 eV 0.8 0.25 eV
0.8
7.5 mJ/cm²
7.7 mJ/cm² 0.6 5.9 mJ/cm²
R/R
R/R
0.6
5.9 mJ/cm² 4.4 mJ/cm²
4.4 mJ/cm² 0.4 3.9 mJ/cm²
0.4 3.3 mJ/cm²
3.8 mJ/cm²
3.2 mJ/cm² 2.5 mJ/cm²
0.2 0.2 2.0 mJ/cm²
2.7 mJ/cm²
simulation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (ps) t (ps)
Figure 3.18: Transient reflectivities renormalized between 0 and 1 for the slow rise part: Δ𝑅⁄𝑅|𝑡𝑟 is set to
zero after  0.2 ps (beginning of the slow rise) and to 1 at a long time (t  40 ps), both for the 100 nm (a)
and 1 µm (b) thin films.
As explained in [229], the TMM consists in considering the sample as a multilayer, where
each layer (i) is parallel to the surface impacted by the laser and (ii) is very thin (typically 1 nm), and
has its own optical indices (n, k) value. To do this, we have first obtained the complex refractive
index ñ=n+ik of the films by fitting their steady-state reflectivity spectra in the PM at T = 200 K
(ñPM) and AFI phases at 66 K (ñAFI) by a sum of Drude Lorentz oscillators within a “film on a
substrate” model. The use of the PM reflectivity at 200 K is justified since the R spectra close to
room temperature PM and low T compressed PM phases are similar (see Figure 3.14 b). The
simulation on the 1 μm film displayed in Figure 3.18.b was done assuming that the strain wave
propagates at the speed of sound (8 km/s) and let behind it a mixture of PM and compressed AFI
phase with 100% of PM phase at the surface. The (n, k) values of each layer of the domain with
phase coexistence were estimated by the mixing law between AFI and PM phases proposed in
Ref. [229], ñ(𝑧) = ñ𝑃𝑀 𝑥0 𝑒 −𝑖𝑧/𝜉 + ñ𝐴𝐹𝐼 (1 − 𝑥0 𝑒 −𝑖𝑧/𝜉 ). Here 𝑥0 is the fraction of metallic phase at the
surface and 𝜉 is penetration depth. Figure 3.18.b (black line) shows a successful example of
Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑡) simulation in the 1 µm film compared with experimental data, obtained for 𝑥0 = 1 and
𝜉 =500 nm. We must be careful in interpreting this good agreement: some of the hypotheses
described above, such as the estimation of the optical (n,k) values in the coexistence region, are
128
indeed questionable. According to Figure 3.17, the choice of 𝑥0 = 1 in the simulation should indeed
correspond to regime #2, where the AFI phase is compressed in the coexistence regime.
Unfortunately, estimating reliably the optical indices ñcompressed AFI in the film is not very easy.
Nevertheless, it is striking that the experimentally observed slow rise time seems to be quite well
described in the simulation based on the launching of a compressive wave propagating at the speed
of sound.
• A possible final confirmation of the strain wave model: time-resolved X-ray
diffraction experiments
In order to check the scenario of strain wave propagation, a time-resolved optical pump
(0.89eV) – X-ray Diffraction probe measurements were engaged in our V2O3 thin films during the
Ph.D. work of Alix Volte (Institute of Physics of Rennes). Data were recorded at 95 K with a 100 ps
time resolution at the ESRF synchrotron [227]. Figure 3.19 shows that the unit cell volume drops
above a threshold fluence of 1.5-2 mJ/cm2. At large fluence, a hump indicated by the black arrows
in Figure 3.19 appears at the end of the rise time. As explained in Ref. [152] and in the Ph.D. of Alix
Volte, this key signature of the strain wave mechanism is expected in films thinner than the
penetration depth. In this case, two strain waves are launched, the first one from the surface toward
the substrate and the second one at the film-substrate interface toward the surface. During a short
time interval, both strains will add up, leading to a transient maximum of strain. Overall, these tr-
XRD results demonstrate that the lattice response of the strongly correlated insulator V2O3-AFI after
photoexcitation of the electronic subsystem is a compression of the unit cell volume.
Figure 3.19 a) evolution of the unit cell volume extracted from Rietvelt refinement of the tr-XRD pattern
shown in b). b) tr-XRD differential powder pattern measured on beamline ID09 at ESRF (measurement
at T = 100 K, time resolution 100 ps, zoom on two diffraction peaks) on a 250 nm thick V2O3 film at 95 K.
129
3.3.3 Physical picture of the photo-induced insulator to metal transition in

(V0.95Cr0.05)2O3-PI and V2O3-AFI
In the previous section, we have described the tr-reflectivity studies of three series of
samples, a single crystal of (V0.95Cr0.05)2O3-PI, another of V2O3-AFI, and thin films of V2O3-AFI. The
thin films of V2O3-AFI display the most spectacular behavior, since it was possible to induce a non-
thermal insulator to metal transition for a large fraction of the sample, probably between 50 % and
100 % at the surface impacted by the pump laser. An appropriate choice of experimental conditions
(low base temperature of 10 K and photon energy slightly above the gap) allows to discard a trivial
thermal scenario. Conversely, a scenario of a non-thermal and multistep photo-induced transition is
able to explain most of the observed evolutions of time-resolved reflectivity and X-Ray diffraction.
The first step is the absorption of the pump laser over a characteristic penetration depth. The
resulting photo-excitation will promote electrons mostly in the vanadium a1g band. As the a1g band
has a bonding character with respect to the dimer V-V bond (see Figure 1.10 in chapter 1), this will
tend to lower the energy of the bonding a1g level and stabilize a transient metallic state. At this step,
a compressive mechanical stress should set up in response to the electronic delocalization and to the
metallic state, but without any macroscopic strain so far (t < 1ps). This idea of a deformation potential
due to transient metallicity leading to a compressive stress in Mott insulators is somewhat speculative,
since the out-of-equilibrium compressible Hubbard model has never been treated theoretically to
our knowledge. We recall however that, according to the compressible Hubbard model at
equilibrium (see part 2.1 of this manuscript), this is the electronic delocalization which causes the
volume contraction at the pressure-induced Mott insulator to metal transition. As discussed in part
1.4.1.2.2, the z-profile of the stress is similar to energy deposited in the sample (see e.g. Figure 3.16.e).
Following the argument of Thomsen [148], it is impossible to induce a strain starting inside of a
bulk semi-infinite medium for obvious reasons of elastic cost. Consequently, a coherent strain wave
starting at the free surface will be launched at the speed of sound toward the interior of the sample.
This mechanism is established since the 80’s, but the possibility for a strain wave to drive with and
behind its front wave a true macroscopic phase transition has been scarcely evoked so far [152,176].
The study presented here seems to indicate that this original mechanism is the driving force for the
macroscopic photoinduced insulator to metal transition in V2O3-AFI.
Before closing this section, let us end by mentioning a few open questions regarding ultrafast
photo-induced effects in the V2O3 system. One of them concerns the choice of experimental
conditions for (V1-xCrx)2O3-PI. In this work, we show that a pump laser energy of 0.89 eV is able to
drive (V0.95Cr0.05)2O3 at the vicinity of the PI-PM transition, with a transient Δ𝑅 ⁄𝑅 |𝑡𝑟 quite similar to
the one observed at equilibrium under a pressure close to the critical value PPI→PM. However, due to
the very small gap value of this compound (EG = 0.16  0.05 eV, see Figure 3.11.d), more than 80 %
of the pump photon energy might be lost and transform into heat. This undesired heating effect
most likely modifies the dynamics of the photo-induced transition. Further studies with pump
photon energy closer to the gap of (V0.95Cr0.05)2O3- PI will hence be necessary in the future. This will
allow to capture the impact of the sole non-thermal strain wave mechanism in the photoinduced
insulator to metal transition starting from a pure paramagnetic Mott insulator state. A full clarification
avoiding the undesired heating effect would require pump photon energy around 0.2 eV
130
A second open question concerns the absence of insulator to metal transition in the
V2O3 - AFI single crystal. As proposed in part 3.3.2.1, this could result from a clamping mechanism,
i.e. the impossibility to induce a strain along the z-direction because the associated strain in the
perpendicular (x,y) direction is prohibited by an elastic cost that is too high. This question has to be
deepened since the invoked presence of a “compressed-AFI” phase to explain the slow rise of
Δ𝑅 ⁄𝑅 |𝑡𝑟 (𝑡) is not fully satisfactory: the scenario of a compressed AFI phase still involves a strain
along z, even if it is smaller than that of a full AFI to PM transition. Other co-existing mechanisms
might hence be at play to fully describe the behavior photo-induced effects in the crystal.
Finally, the last issue is at the junction of the two open questions raised above. It is indeed
necessary to clarify why it seems possible to bring a single crystal of (V0.95Cr0.05)2O3- PI across the
Mott insulator to metal transition, while the clamping effect seems to prohibit it for V2O3-AFI. Time-
resolved XRD studies on (V0.95Cr0.05)2O3- PI are clearly required to solve this issue.
131
CONCLUSIONS
CONCLUSIONS
Mott insulators form a large class of materials of potential interest for modern
microelectronics. The Mott insulator to metal transition induced by electric pulses, called Electric
Mott transition (EMT), may indeed lead to neuromorphic and memory applications. However, the
knowledge about the EMT mechanism is far from complete and especially the nature of the metallic
phase created under electric field is barely known. In that framework, this Ph.D. thesis aimed to
achieve a better understanding of the EMT by studying the metallic phase created after a non-volatile
(EMT), and by exploring the possibility to induce a similar Mott insulator to metal transition by light
pulses.
Our study on transited (V1-xCrx)2O3 Mott insulator devices enabled to visualize the
conductive filamentary path created after the EMT. The characterization of the filament by c-AFM
mapping confirmed its conducting nature while a µ-XRD mapping revealed a local unit cell volume
drop compared to that of pristine (V1-xCrx)2O3 Mott insulator. This results were further confirmed
by a µ-Raman characterization which revealed a softening of the Raman modes within the filament
similar to the one observed in the pressure-induced metallic phase in the (V1 - xCrx)2O3 Mott insulator.
This results demonstrate the creation of a filamentary path made of locally compressed (V1-xCrx)2O3
metallic phase. Our Raman studies strongly suggest that the creation of excited and delocalized
electrons at the EMT leads to a compressive lattice response. Overall, this Ph.D. work provides a
strong indication that Mott physics plays a pivotal role in the insulator to metal transition observed
under the electric field in (V1 - xCrx)2O3.
Our study coupling electric and laser pulses using a laser pump - electrical pump - electrical
probe experiment on GaTa4Se8 showed that the resistive switching (EMT) is affected by the number
of generated photocarriers rather than by the accumulation of energy deposited by the femtosecond
laser. This result supports that the EMT originates from an electronic avalanche mechanism.
Moreover, our time-resolved photoemission study on GaTa4Se8, showed an unusually long carrier
lifetime in the system which explains why electron accumulation (i.e. integration property) is possible
in the Mott insulator. This enabled the creation of an artificial electro-optical neuron.
As the EMT is triggered by the creation of hot carriers by electric pulses, this PhD have also
explored the possibility to induce a non-thermal insulator to metal transition thanks to the direct
creation of hot carriers by ultrafast light pulses. Ultrafast pump-probe measurements were
performed on (V1-xCrx)2O3 thin films in insulating AFI and PI phases to promote them into a non-
equilibrium state. These experiments have demonstrated a non-thermal pathway of Photoinduced
Mott insulator to metal transition. As in the case of EMT, the process involves the consequences of
strong electron-lattice coupling in the material resulting in a compressive lattice response after
excitation of the electronic subsystem. Absorption of photons of the laser pulse creates likely
compressive mechanical stress in response to the delocalization of photoexcited carriers. At the
sample surface, this phenomenon launches a coherent compressive strain wave transforming the
material into a metallic phase.
Both Electric Mott Transition (EMT) and Photoinduced Mott Transition (PMT) show
strong signatures of an electron-lattice coupling mechanism, i.e. a compressive lattice response to
electronic excitations. Moreover, in both cases, the strong electronic excitation leads to the
generation of a macroscopic metallic phase similar to the one created at equilibrium under pressure
132
CONCLUSIONS
by the bandwidth controlled IMT mechanism. This work highlights therefore the strong similarity
between EMT and PMT. It supports that driving the Mott insulator out of equilibrium by creation
of hot carriers either by electric or light pulses might have similar consequences especially on the
lattice.
Overall, the fundamental output of this work gives an understanding of the mechanism of
out-of-equilibrium Mott Insulator to Metal Transition driven by light or/and electric field pulses.
Alongside this work establishes the proof of concept of a new electro-optic artificial Mott neuron.
It opens new perspectives of memory and neuromorphic applications based on Mott IMT driven
only by ultrafast laser pulse.
133
ANNEXES
ANNEX 1 SAMPLE PREPARATION

1-A SYNTHESIS OF AM4Q8 SINGLE CRYSTALS
Single crystals of GaM4Q8 (M = V or Ta) were prepared using stoichiometric mixtures of
elemental gallium, metal M ( = V, Ta) and chalcogen Q (= S or Se), all with purities >99.5%. These
mixtures were loaded in evacuated sealed quartz ampoules of about 10 cm and heated at 300°C/h
to 1000°C and hold at this temperature during 160 h. The quartz tubes were subsequently cooled,
first slowly to 800°C at 1°C/h and then faster (300°C/h) to room temperature. These syntheses
yielded black powders containing a high yield of GaM4Q8 metallic-gray cubic and tetrahedral crystals
along with a small amount of impurity phases. The maximal crystal sizes vary from one batch to
another and range from 30 to 300 μm.
1-B SYNTHESIS OF (V1-XCRX)2O3 SINGLE CRYSTALS
(V1-xCrx)2O3 single crystals were prepared using a sulfur-assisted chemical-vapour transport
method. The synthesis consisted first in preparing an homogeneous powder of (V1-xCrx)2O3. Two
methods for powder synthesis were used during this PhD thesis :
- Method 1 (“dry strategy”): after a first drying at 400°C, V2O5 powder (Aldrich, >99.6%)
was mixed with Cr2O3 (Prolabo, 99.9%) in the appropriate ratio,
- Method 2 (“liquid strategy”): ammonium metavanadate NH4VO3 was mixed with hydrated
chromium nitrate Cr(NO3)3, 9 H2O in a aqueous solution, along with citric acid C6H8O7
and ammonium nitrate (NH4)NO3, and according to the following molar ratio:
(1-x) NH4VO3 + x Cr(NO3)3, 9H2O + 2.5 C6H8O7 + 7.5 NH4NO3
where x is the targeted composition of the final (V1-xCrx)2O3 compound. The amount of
ammonium nitrate is adjusted to get a pH close to 8. Once the vanadium and chromium
precursors are fully dissolved, the mixture is heated up to 60-100°C in air in order to dry it.
A pyrolysis reaction might occur at this stage, which starts burning the organic part of the
mixture. In order to burn and hence remove all traces of organic part, an annealing
treatment is performed under oxygen flow at 550°C during 70 h.
The obtained powders was then submitted to a two-step treatment under a 95% Ar–5% H2
gas flow, first at 550°C during 20h in order to transform the V2O5 phase formed during the oxygen
annealing into a more reduced phase (and to avoid the melting of V2O5 around 690°C), and then at
900°C for 10-24 h. Half a gram of the pure (V1-xCrx)2O3 obtained powder was then introduced in a
6 to 10 cm long silica tube (inner diameter 10 mm), with typically 40 mg of sulfur as a vapor phase
transport agent. The tube was vacuum sealed, heated up to Tmax (950-1050°C) in a furnace with a
temperature gradient (≈ 10°C/cm), slowly cooled down to 900°C (from - 0.5 to - 2°C/h) and finally
fast cooled (−300°C/h) to room temperature. Such treatment allows obtaining small single crystals
(typical size <300 μm) within the preparation. The chromium substitution rate on each individual
crystals used in this work was determined by Energy-dispersive X-ray spectroscopy (EDXS) analyses
carried out using a scanning electron microscope JEOL 5800.
1-C SYNTHESIS OF (V1-XCRX)2O3 THIN FILMS
Thin films of (V1-xCrx)2O3 were deposited by reactive magnetron co-sputtering of Vand Cr
targets in Ar/O2 discharge [93]. Post-deposition annealing at 500°C in a reducing atmosphere
yielded crystallized and single-phased (V1-xCrx)2O3 layers as checked by X-ray diffraction
(see Ref. [93]).
134
ANNEXES
108
100 nm (V2O3)
1 m V2O3
Resistivity (Ohm)
106
104
102
100
100 150 200 250 300
T(K)
Figure A 1-1 Transport measurements on typical 100nm and 1um V2O3 thin films.
1-D SYNTHESIS OF DEVICES BASED ON (V1-XCRX)2O3 THIN FILMS

The patterned planar devices used in this work (see Chapter 2) were fabricated during the
PhD thesis of Coline Adda, within a collaboration between the Jean Rouxel Institute of Materials
(France) and the CIC nanoGUNE (Spain). The main steps of the fabrication process, depicted
in [230], consist in :
- step 1 : deposition of TiN (electrode material),
- step 2 : photolithography to design the electrode pattern and etching of TiN,
- step 3 : photolithography to design the active material pattern,
- step 4 : deposition, lift-off and annealing of active material (V0.95Cr0.05)2O3 thin film.
The 150 nm thick (V0.95Cr0.05)2O3 films were deposited by a co-sputtering technique at room
temperature, and were then annealed under a controlled reducing atmosphere at 500° C for 10 h, as
described above. Chemical analyses performed by EDXS confirm the presence of three elements V,
Cr, and O in the films and give a Cr/(V + Cr) ratio of 0.053(4) in good agreement with the targeted
composition.
135
ANNEXES
Figure A 1-2:fabrication process of the patterned planar devices made of titanium nitride electrodes and
(V1.95Cr0.05)2O3 thin films as active materials.
After the application of the electric pulses leading to the resistive switching (see Chapter 2),
the samples used in this study were characterized by Scanning Electron Microscopy (SEM) and
conducting - Atomic Force Microscopy (c-AFM). It clearly appears both in Figure A 1-3.a-b (SEM)
and Figure A 1-3.c (c-AFM) that the resistive switching does not result in any visible microstructural
changes compared to a pristine state.
Figure A 1-3: Scanning Electron Microscope (a-b) and Atomic Force Microscope (AFM) images (c) of a
device based on (V0.95Cr0.05)2O3 thin films. The images were obtained after the application of electric pulses
yielding the Electric Mott Transition (see Figure 2.9 and Figure 2.10 in Chapter 2). The z-profile of the
170 µm long / 20 µm wide polycrystalline (V0.95Cr0.05)2O3 stripe is represented in the lower part of (c), and
indicates a film thickness of 165 nm. AFM images were analyzed using the WSxM software package.
136
ANNEXES
1-E SAMPLE PREPARATION FOR TIME-RESOLVED AND STEADY-STATE EXPERIMENTS
(d) (f)
V2O3 100 nm, V2O3 on sapphire 1 µm, V2O3 on SiO2/Si
(a)
(c) (V1-xCrx)2O3
500μm thin films
(e) (g)
(V0.95Cr0.05)2O3 1mm
(b) X=0; 0.05
500 μm
Figure A 1-4: Samples used for ultrafast optical pump-probe measurements on (V1-xCrx)2O3 system:
(a) Polished V2O3 crystal. (b) Polished (V0.95Cr0.05)2O3 crystal. (c) Typical top view of (V1-xCrx)2O3 thin-film
(x = 0 or 0.05) of 100 nm and 1 μm deposited on c-cut Al2O3 substrate.
Typical top and side views SEM images on 100 nm (d-e) and on 1 μm (f-g) V2O3 film.
137
ANNEXES
ANNEX 2 ELECTRO-OPTIC PUMP / ELECTRIC PROBE STUDY ON A

GaTa 4 Se 8 SINGLE CRYSTAL
2-A SAMPLE PREPARATION AND CHARACTERIZATION
GaTa4Se8 crystals were synthesized by a method described elsewhere [203] , and cleaved into
200 - 300 µm pieces. Сrystals used for transport measurements were contacted using 10μm gold
wires and carbon paste (Electrodag PR-406), and then annealed under vacuum at 150 °C for 30 min.
The low-bias resistance is measured using a Keithley 236 source-measure unit by a standard two- or
four-probe technique. For the application of electric pulses, we have connected two gold wires along
the crystal surface on which light should be applied. The inter-electrode distance was around 200
µm.
Parameter Size (µm)
Inter-electrode distance (d) 180  10
Width W 210  10
Thickness t 90  10
Figure A 2-1: GaTa4Se8 single crystal with gold wires (electrodes) connection and crystal
geometrical parameters
We have checked the overall quality of the crystal and of electrical contacts by measuring the
temperature dependence of its electrical resistance. Data were obtained in a cryostat by using a
modified four-terminal sensing method (Figure A 2-2.b).
(a) (b)
104
GaTa4Se8 crystal, true 4 points
current crystal, "modified" 4 points
103
Resistivity  (Ohm.cm)
102
10
0.1
50 100 150 200 250 300
T (K)
Figure A 2-2: (a) Resistivity versus temperature measured on the current GaTa4Se8 crystal with a “modified”
4-points geometry, compared with another GaTa4Se8 crystal measured with a true 4-points technique. (b)
The schematically depicted experimental setup.
138
ANNEXES
Clearly, the resistivity measured here on the GaTa4Se8 crystal with a modified 4- points
technique is extremely close to the intrinsic resistivity of this phase (see Figure A 2-2.a) measured
with a true 4-points geometry. This result is important since it demonstrates that the contact
resistance of the sample used for this study can be safely neglected.
2-B PUMP – PUMP – PROBE EXPERIMENTS
We have applied two “pumps”, a  100 µs electric pulse and a 100 fs laser pulse in a single-
shot mode, on the GaTa4Se8 crystal, while recording the evolution of the electric resistance (“probe”)
(Figure A 2-3). We used a Ti: Sapphire femtosecond laser synchronized with an electric pulse
generator (Keithley 8114A) so that the laser output was triggered precisely with the first nanoseconds
of the electrical pulse. During the pulse, the voltage and current across the sample were measured
with a Tektronix DPO3034 oscilloscope associated with IeS-ISSD210 differential probes. We used
an optical parametric amplifier (OPA) to tune the laser wavelength from the visible to mid-IR region.
In order to control crystal temperature, we have used a nitrogen Cryostream.
Shutter
100 fms OPA

BS OD M
laser
800 nm Cryostream
Rl S
DP DP
Osc
Figure A 2-3: Experimental optical pump-electrical-pump-electrical-probe setup. The 100 fs laser pulse
with an 800 nm wavelength goes through an OPA for wavelength conversion and shutter. Afterward, the
laser beam is separated by a beam splitter (BS) for the photodiode (Pd) reference signal. The main part of
the laser beam goes through the Round Continuously Variable Metallic Neutral Density Filter (OD) to the
mirror (M). Finally, the laser beam was focused on the interelectrode sample surface by a 50 mm short-
focus lens. Electrical pulse was applied to the crystal in series with an additional resistor. The additional resistor RLOAD
acts as a current limiter and also allows to measure the current flowing through the circuit. Voltages on the resistor
and on the crystal were measured with an oscilloscope using differential probes (DP). The differential probes allow
voltage measurement between two test points where none of the test point is at the ground. Thanks to a
synchronization module, the setup was optimized for single laser shot experiments and time-resolved
measurements. The synchronization module was triggered by the laser pulse repetition rate frequency
signal; then this device sends two control signals to the shutter and electrical pulse generator. This
configuration allowed us to have a laser signal before and during electrical pulse
139
ANNEXES
We carried out the experiments with the same number of absorbed photons by taking care
to keep the laser spot size similar at the different wavelengths. We used a short focus (5 cm) optical
lens. After the experiment on the GaTa4Se8 single crystal, we have performed knife-edge
measurements to recover the spatial profile of the laser beam. It allows a reliable estimate of the
beam size, which were similar for both laser energies (0.5 and 2.5 eV) and covered more than 90 %
of the inter-electrode area. In these experiments, the level of photodoping depends on both the laser
fluence and the impacted volume. The latter is defined by the size of the beam hitting the sample
surface and a characteristic length over which the photocarrier spread inside the materials. Since
photodoped carriers are subjected to applied DC electric fields E close to 3 kV / cm, this
characteristic length corresponds to the diffusion length of the carriers rather than the initial
absorption length (i.e., the light penetration depth). The diffusion length is indeed defined as
ldiff = µ.E., where µ and  are the carriers mobility and lifetime, respectively. Considering a lifetime
of 1 µs typical of AM4Q8 compounds [29] and a mobility representative of Mott insulators of 1 cm².V-
1 -1 [30]
.s leads to a lower limit of 30 µm for the diffusion length. This value exceeds the typical light
penetration depth by more than two orders of magnitude. Consequently, experiments conducted
with similar beam sizes and density of absorbed photons, but with laser pulses at different
wavelengths, will yield analogous levels of photodoping.
140
ANNEXES
ANNEX 3 RAMAN MEASUREMENTS

3-A RAMAN UNDER PRESSURE
Raman spectra were recorded using a µRaman Invia apparatus (Renishaw) using an
excitation at 785 nm. The spectral resolution achieved with the use of gratings of 1200 grooves per
millimeter was 3 cm-1. The microscope was equipped with an automated XYZ table allowing
mapping with a spatial resolution of 1µm.
Figure A 3-1:(a) schematic drawing of the Diamond Anvil Cell (DAC) used for Raman under pressure
experiments. (b) Image of the gasket drilled with a 300 µm hole to define a pressure chamber, taken from
the top of the DAC. (c) Optical image of the single (V0.962Cr0.038)2O3 crystal inside of the pressure chamber.
Inset : Scanning Electron Microscope Image of the crystal. (d) Typical Raman spectra of (V0.962Cr0.038)2O3
measured inside of the DAC, at room pressure in the PI state and at 1.8 GPa in the PM state.
Raman scattering experiments under pressure were performed in the Membrane Diamond
Anvil Cell (MDAC) schematized in Figure A 3-1-a. The (V0.962Cr0.038)2O3 single crystal and a few ruby
micro-crystals were placed in the pressure chamber prepared by drilling a 300 µm hole in a stainless-
steel gasket (see Figure A 3-1-b-c). Pure methanol was used as pressure transmitting medium. The
pressure was estimated by monitoring the fluorescence lines of ruby excited by a 514 nm laser. Two
Raman spectra, measured on the (V0.962Cr0.038)2O3 crystal using a 785 nm laser, are shown in Figure
A 3-1-d. The broad contribution between 350 and 700 cm -1 corresponds to the Raman signal of
(2)
methanol, and complicates the measurement and interpretation of the 𝐴1𝑔 phonon of
(V0.962Cr0.038)2O3 around 510 cm-1.
141
ANNEXES
3-B RAMAN SPECTROSCOPY OF (V1-XCRX)2O3 SAMPLES AS A FUNCTION OF CR CONTENT
(a) 1,2
(V Cr ) O crystals 300 K
1-x x 2 3
0%
2%
0,8 3.5%
5%
10%
0,4
1,2
(b) (V Cr ) O 100nm thin films on Al2O3
1-x x 2 3
0,8 0%
3%
5%
0,4
200 300 400 500 600

-1
Raman shift (cm )
Figure A 3-2: Ramans spectra of (V1-xCrx)2O3 samples at different Cr concentrations measured in a

setup similar to that described in the previous section: (a) single crystals and (b) 100nm thin films
142
ANNEXES
ANNEX 4 COMPLEMENTARY CHARACTERIZATION OF

CONDUCTIVE FILAMENTARY PATH AFTER NON -VOLATILE
RESISTIVE SWITCHING OF (V 1 - X CR X ) 2 O 3
4-A TRANSPORT STUDY ON THE (V0.975CR0.025)2O3 SINGLE CRYSTAL
Figure 2.6 in the Section 2.2 discloses a part of a broader transport study carried out on a
(V0.975Cr0.025)2O3 single crystal before and after an Electric Mott Transition. The results of an extended
study is depicted in Figure A 4-1. After measuring the resistance vs temperature curve in the pristine
insulating state and just after the application of electric pulses, the (V0.975Cr0.025)2O3 was then
submitted to several cooling - warming cycles between 40 and 300K.
(V1-xCrx)2O3, 1
x = 0.025
PM
0
AFI
2.5
PI
x (%)
2
1
3
double-crossing
(heating + cooling)
of the first order
2 AFI→PI transition
Figure A 4-1: results of extended transport study on a (V0.975Cr0.025)2O3 single-crystal submitted to Electric
Mott Transition at 160 K, in complement to Figure 2.6 in the Section 2.2.1.1.
In the first temperature cycle, a resistance jump appears around 40 K (red open circles, curve
#2 in Figure 2.6 and in Figure A 4-1). Upon heating (red filled circles, curve #2 in Figure A 4-1), the
jump in the R(T) curve around 40 K is essentially unmodified, with only a small hysteresis of  1K.
143
ANNEXES
At higher temperature, an attenuated R(T) drop persists at the AFI → PI transition and the resistance
almost superimposes that of the pristine compound above 180 K. Surprisingly, after less than fifteen
minutes spent at 290 K, the R(T) curve measured subsequently by cooling again from room to low
temperature (see curve #3 in Figure A 4-1) unveils a huge shift of the low-temperature transition
from 40 K to 105 K. Subsequent “warming to 290 K – cooling down to 40 K” cycles did no longer
modify the resistance curve, with a low-temperature transition maintained at 105  2 K.
As mentioned in the in the Section 2.2, these observations can be explained by the creation
of a conducting filament which percolates between the contact electrodes. This explanation was
already proposed for other Mott insulators [231], but it has a striking consequence in (V1-xCrx)2O3.
Above the transition temperature TPI→AFI, the conductance of the filament is indeed low compared
to that of the pristine part, letting the macroscopic resistivity almost unchanged. Conversely, the
conductance of the filament becomes predominant below TPI → AFI = 180K, since the conductance
of the pristine part (curve #1 in Figure A 4-1) decreases sharply in the AFI phase, [27,196] As a
consequence, the additional resistivity jumps observed at 40 K in curves #2 and then at 105 K in
curve #3 are attributed to transitions within the conductive filament. These transitions give, therefore
some clues about the nature of the metallic phase created after the electric pulse. According to the
phase diagram of the (V1 - xCrx)2O3 system, the resistivity jumps observed at 40 K and 105 K could
correspond to the PM – AFI transitions that exist at these temperatures in (V0.975Cr0.025)2O3 put under
pressure at  3 GPa and  2 GPa, respectively. Therefore, these measurements strongly suggest that
the electric pulse has created a filamentary path made of the same phase, but under a compressive
strain corresponding to a pressure of 2 - 3 GPa, i.e. strong enough to make it metallic. Within this
picture, the shift of the transition temperature from 40 K to 105 K is simply explained by the strain
relaxation caused by the double-crossing (i.e. by heating and then cooling) of the AFI → PI transition
in the pristine zone surrounding the filament. The abrupt changes of the a (+0.24 %) and c (-0.57 %)
parameters of the pristine zone at the transition [27] might indeed relax the stress state of the
filament and explain the transport properties shown in Figure A 4-1. Finally, the shift of the transition
temperature observed after bringing back the sample to room temperature discards a possible change
of stoichiometry within the filamentary path. At this temperature the ionic mobility is indeed far too
low to induce a stoichiometry change in a measurable time.
144
ANNEXES
4-B COMPLEMENTARY RAMAN SPECTROSCOPY DATA ON (V1-XCRX)2O3 SINGLE CRYSTALS

In complement with the Raman study of the conducting filament shown in Figure 2.7
(Section 2.2.1.2), we present here additional Raman data obtained on the same (V0.975Cr0.025)2O3 crystal
in the PI phase and in the PM phase under pressure, as well as on a Cr-free V2O3 crystal in the PM
phase. Both crystals crystallize in the 𝑅3̅𝑐 corundum crystallographic structure with the D3d point
group [232]. Seven Raman-active modes are expected, two A1g and five Eg modes. Figure A 4-2
shows that three intense Raman modes can be clearly identified in (V1-xCrx)2O3 (x = 0.025), which
can be assigned to two A1g and one Eg modes thanks to the modelling work of Yang and
Sladek [233].
In Section 2.2.1.2, Figure 2.7c represents the changes in Raman spectra between the pristine
state and the metallic state in the conducting filament in the energy window 170-290 cm-1. Figure A
4-2 a shows the same data on a larger energy window (170 - 850 cm-1). This global view shows that
essentially the same Raman vibration modes appear inside and outside of the conducting filament:
(1) (2)
two A1g modes around 240 cm-1 (𝐴1𝑔 ) and 510 cm-1 (𝐴1𝑔 ), as well as an Eg modes of weak intensity
(1) (2)
around 210 cm-1 (𝐸𝑔 ) and another more intense one at 305 cm-1 (𝐸𝑔 ). These measurements, in
perfect agreement with the studies published in the system (V1 - xCrx)2O3 [32], [197], support the
scenario of a filament made up exclusively of the phase (V1 - xCrx)2O3 phase, without any other phase
of different chemical composition.
(V0.975Cr0.025)2O3 (PI)
(V0.975Cr0.025)2O3, inside and outside the conducting filament
vs. V2O3 (PM)
(a) (b) (V0.975Cr0.025)2O3

Mott insulator,
(c) (V0.975Cr0.025)2O3
Mott insulator,
pristine pristine
Raman intensity
Raman intensity
Raman intensity
0, , q = 0, , q =
512, 12, -4 cm-1 512, 12, -4 cm-1
V2O3 metallic,
Filament pristine
0, , q =
511, 18.1, -1.9 cm-1
0, , q =
505, 16.4, -3.3 cm-1
200 300 400 500 600 700 800 400 450 500 550 400 450 500 550
 (cm )  (cm )
-1 -1
 (cm-1)
Figure A 4-2: (a) Full Raman spectra measured inside (red curve) the conducting filament and outside in
the pristine state (blue curve). The atomic displacements associated with the three most intense Raman
modes are schematically represented [233]. (b) Same Raman spectra as in (a), zoomed on the mode
(see text). (c) Comparison of the mode measured in the PI phase on a (V0.975Cr0.025)2O3 crystal and in
the PM phase on a V2O3 crystal at room temperature. The strongly asymmetric shape results from a Fano
resonance, i.e. a coupling between the phonons and electronic transitions within the a1g – bonding
band [234]. Accordingly, the Raman modes were fitted according to a model appropriate to describe the
Fano effect: ² , with , where 0 is the resonance frequency,  is
a measure of the line width, and q-1 is the Fano coupling coefficient. Clearly, q-1 is enhanced (or equivalently
q is lowered) in the metallic phases with respect to the PI value, both in the standard PM phase of pure
V2O3 and in the conducting filament.
145
ANNEXES
(1)
In Section 2.2.1.2, Figure 2.7 shows that a softening of the 𝐴1𝑔 mode occurs in the filament,
very similar to that appearing at the pressure-induced insulator transition. Figure A 4-2b shows that
(2)
the 𝐴1𝑔 mode around 510 cm-1 presents a similar shift, but also a striking change in the peak shape
(2)
of the filament compared to the pristine state. The 𝐴1𝑔 peak in the filament indeed becomes broader
and becomes more asymmetric with an increasing weight on the low energy side. Interestingly, Figure
(2)
A 4-2c shows that these three characteristic evolutions of the 𝐴1𝑔 mode, namely softening,
broadening and increasing asymmetry, are also observed at the composition-induced insulating (PI)
to metal (PM) transition between (V0.975Cr0.025)2O3 and V2O3. On the theoretical side, the anomalous
asymmetric line shape has been assigned to a Fano resonance effect between the A1g phonons and
electronic transitions within the a1g – bonding band. [234] According to this approach, the smaller
asymmetry observed in the insulating phases results from a lower population of the a1g electronic
band [38].
Interestingly, the study carried out in Ref. [234] has demonstrated that the asymmetry effect
(1)
is also very strong for the 𝐴1𝑔 phonon located around 240 cm-1. Figure A 4-3 a-b shows that the
Raman spectrum between 160 and 290 cm-1 can be decomposed as the sum of two phonons modes,
(1) (1)
a symmetrical 𝐸𝑔 mode and an asymmetrical 𝐴1𝑔 mode.
Insulator to metal Insulator to metal

transition by Electric Field transition by pressure
(a) (c)
20 µm
(b) 0, , q =
Filament (d) 1,8 GPa
(PM)
0, , q =
236, 39, -1.6 cm-1 242, 28, -1.6 cm-1
0 GPa
Pristine (PI)
0, , q = 0, , q =
245, 20, -2.8 cm-1 253, 16, -3.1 cm-1
160 200 240 280 160 200 240 280

 (cm-1)  (cm-1)
Figure A 4-3: (a-b) same Raman data as in Figure A 4-2-a, but zoomed on the low energy part of the spectra.
146
ANNEXES
The same decomposition is also able to describe the Raman spectra on both sides of the
pressure-induced insulator to metal transition, as shown in Figure A 4-3-d. The comparison between
the electric-field-induced and pressure-induced transitions clearly shows a similar evolution of the
asymmetry and linewidth parameters. The only difference between these two samples is the
(1)
frequency of the 𝐴1𝑔 phonon mode, 253 cm-1 in the crystal used for pressure experiments against
245 cm-1 for the one used under electric field. Figure A 4-3-e shows that this small difference results
from slightly different composition of these two crystals, xCr = 3.8 % and 2.5 %, respectively.
Overall, these analyses of Raman spectroscopy in (V1-xCrx)2O3 single crystals demonstrate
that the electric-pulse-induced conducting filament (1) is composed only of (V1-xCrx)2O3, without any
other phase with different chemical composition, and (2) presents vibrational signature similar with
those observed at the pressure- and composition-induced insulator to metal transitions.
4-C MICRO-X-RAYS DIFFRACTION MAPPING EXPERIMENTS
µ-X-rays diffraction experiments were carried out at the µ-XAS beamline (SLS Synchrotron,
Switzerland). We chose a rather low X-Rays energy of 6.1 keV ( = 2.0325 Å) in order to keep a
beam diameter as small as 1 µm. The calibration of the experiment geometry (sample-detector
distance, orientation of the Eiger 4M 2D detector with respect to the beam) was performed with the
standard compound Na2Ca3Al2F14. An XRD experiment was carried out on this material inserted in
a 100 µm inner diameter glass capillary, placed at the same position as the sample and oriented along
the y direction (see Figure 2.10, Section 2.2.2). The calibration procedure was done using the Fast
Azimuthal Integration pyFAI python package [235].
147
ANNEXES
ANNEX 5 ULTRAFAST MEASUREMENTS ON (V 1 - X CR X ) 2 O 3 SYSTEM

5-A EXPERIMENTAL SETUP
Regenerative amplifier
Chopper
Detectors
Delay line
Boxcar
Monoch averager
romator
Sample
LPF
OAP mirror
Cryostat
Figure A 5-1 Optical system for pump-probe spectroscopy (adapted from T. Amano Master thesis [224]):
a beamsplitter separates a 100 fs pulse with a center wavelength of 800 nm oscillated from a femtosecond
titanium sapphire regeneration amplifier (Hurricane manufactured by Spectra-Physics: repetition frequency
1 kHz, output value 800 μJ / pulse) into two Optical Parametric Amplifiers (OPA). The wavelength is
converted independently by the OPA and used for measurement as pump light and probe light. OPA 2 is
also coupled with a Different Frequency Generator (DFG). Both the pump and probe beam irradiates a
sample with a time difference imposed by the delay line. Probe signal focalized on the sample by Off-Axis
Parabolic (OAP) mirror and after reflection from the material go through a monochromator to the detector
connected to the boxcar averager. The second channel of boxcar averager is connected to the detector,
capturing pump signal reflected from the sample. The sample holder is located in the free- He cryostat with
a possible temperature range of 10-350 K.
148
ANNEXES
5-B COMPLEMENTARY TIME-RESOLVED MEASUREMENTS ON (V0.95CR0.5)2O3 SINGLE

CRYSTAL
— Transient spectra:
(V Cr ) O crystal
0.95 0.05 2 3
0,3
Time delay
Pump 0.89eV, 14mJ/P
0 ps 100 ps
1 ps 150 ps
0,2 5 ps 400 ps
10 ps 600 ps
50 ps 1000 ps
0,1
R/R
-0,1
0 0,5 1 1,5
E (eV)
Figure A 5-2: tr-OPOP spectrum Δ𝑅 ⁄𝑅 |𝑡𝑟 = ( R 𝑝ℎ𝑜𝑡𝑜𝑒𝑥𝑐𝑖𝑡𝑒𝑑 – 𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 )⁄𝑅𝑛𝑜 𝑝𝑢𝑚𝑝 measured on a
(V0.95Cr0.05)2O3–PI single crystal at 300 K, in the range 0-1000 ps time delay after the photoexcitation (pump
energy = 0.89eV, laser fluence =14 mJ/cm2).
(V0.95Cr0.05)2O3 crystal at 10K in AFI phase

0.6
Pump (0.89eV) - Probe (0.25 eV) (a)
0.5 2 2
14 mJ/cm 2.6mJ/cm
8 mJ/cm2 1.1mJ/cm2
0.4
4,8 mJ/cm2 0.12 mJ/cm2
R/R
0.3
0.2
0.1
0
0 200 400 600 800 1000
Time (ps)
0.6 0.6
(b) (c)
0.5 0.5
0.4 0.4
R/R
R/R
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 2 4 6 8 10 0 20 40 60 80 100
Time (ps) Time (ps)
Figure A 5-3: Temporal profiles of transient reflectivity change (Δ𝑅 ⁄𝑅 |𝑡𝑟 at different pump laser fluence
Iex. with a pump at 0.89 eV and a probe at 0.25eV on a (V0.95Cr0.05)2O3 crystal at 10K in the AFI phase at
different time scales
149
ANNEXES
0.2
14mJ/cm2 4.8 mJ/cm2 1.1 mJ/cm2
8mJ/cm2 2.6 mJ/cm2 0.12 mJ/cm2
R/R-(R/R)t=0 0.15
0.1
0.05
0 100 200 300 400 500

Time (ps)
Figure A 5-4: Temporal profiles of transient reflectivity change (Δ𝑅 ⁄𝑅 |𝑡𝑟 with substructed value at zero
time from Figure A 5-3.
5-C PRELIMINARY PUMP-PROBE EXPERIMENT ON V2O3 THIN FILMS

(V0.95Cr 0.05)2O3 100 nm film on Al 2O3
Pump (0.89eV) - Probe (0.25 eV) 1.10 mJ/cm2

0.3 5.33 mJ/cm2
8.00 mJ/cm2
14.00 mJ/cm2
0.2
R/R
0.1
0
0 200 400 600 800 1000
Time (ps)
Figure A 5-5 Temporal profiles of transient reflectivity change (Δ𝑅 ⁄𝑅 |𝑡𝑟 at different pump laser fluence
Iex. with a pump at 0.89 eV and a probe at 0.25eV on a (V0.95Cr0.05)2O3 polycrystalline 100 nm thin film
deposited on a c-cut sapphire substrate.
150
REFERENCES
REFERENCES
[1] M. Imada, A. Fujimori, and Y. Tokura, Metal-Insulator Transitions, Rev. Mod. Phys. 70, 1039
(1998).
[2] E. Janod, J. Tranchant, B. Corraze, M. Querré, P. Stoliar, M. Rozenberg, T. Cren, D.
Roditchev, V. T. Phuoc, M.-P. Besland, and L. Cario, Resistive Switching in Mott Insulators and
Correlated Systems, Advanced Functional Materials 25, 6287 (2015).
[3] Y. Wang, K.-M. Kang, M. Kim, H.-S. Lee, R. Waser, D. Wouters, R. Dittmann, J. J. Yang, and
H.-H. Park, Mott-Transition-Based RRAM, Materials Today 28, 63 (2019).
[4] F. Pan, S. Gao, C. Chen, C. Song, and F. Zeng, Recent Progress in Resistive Random Access Memories:
Materials, Switching Mechanisms, and Performance, Materials Science and Engineering: R: Reports 83,
1 (2014).
[5] You Zhou and S. Ramanathan, Mott Memory and Neuromorphic Devices, Proc. IEEE 103, 1289
(2015).
[6] P. Diener, E. Janod, B. Corraze, M. Querré, C. Adda, M. Guilloux-Viry, S. Cordier, A. Camjayi,
M. Rozenberg, M. P. Besland, and L. Cario, How a Dc Electric Field Drives Mott Insulators Out of
Equilibrium, Phys. Rev. Lett. 121, 016601 (2018).
[7] C. Vaju, L. Cario, B. Corraze, E. Janod, V. Dubost, T. Cren, D. Roditchev, D. Braithwaite, and
O. Chauvet, Electric-Pulse-Driven Electronic Phase Separation, Insulator-Metal Transition, and Possible
Superconductivity in a Mott Insulator, Adv. Mater. 20, 2760 (2008).
[8] B. Corraze, E. Janod, L. Cario, P. Moreau, L. Lajaunie, P. Stoliar, V. Guiot, V. Dubost, J.
Tranchant, S. Salmon, M.-P. Besland, V. T. Phuoc, T. Cren, D. Roditchev, N. Stéphant, D.
Troadec, and M. Rozenberg, Electric Field Induced Avalanche Breakdown and Non-Volatile Resistive
Switching in the Mott Insulators AM4Q8, Eur. Phys. J. Spec. Top. 222, 1046 (2013).
[9] V. Dubost, T. Cren, C. Vaju, L. Cario, B. Corraze, E. Janod, F. Debontridder, and D.
Roditchev, Resistive Switching at the Nanoscale in the Mott Insulator Compound GaTa 4 Se 8, Nano Lett.
13, 3648 (2013).
[10] J. H. de Boer and E. J. W. Verwey, Semi-Conductors with Partially and with Completely Filled 3 d -
Lattice Bands, Proc. Phys. Soc. 49, 59 (1937).
[11] N. F. Mott and R. Peierls, Discussion of the Paper by de Boer and Verwey, Proc. Phys. Soc. 49, 72
(1937).
[12] N. F. Mott, The Basis of the Electron Theory of Metals, with Special Reference to the Transition Metals,
Proc. Phys. Soc. A 62, 416 (1949).
[13] J. Hubbard and B. H. Flowers, Electron Correlations in Narrow Energy Bands. II. The Degenerate Band
Case, Proceedings of the Royal Society of London. Series A. Mathematical and Physical
Sciences 277, 237 (1964).
[14] J. Zaanen, G. A. Sawatzky, and J. W. Allen, Band Gaps and Electronic Structure of Transition-Metal
Compounds, Phys. Rev. Lett. 55, 418 (1985).
[15] A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg, Dynamical Mean-Field Theory of Strongly
Correlated Fermion Systems and the Limit of Infinite Dimensions, Rev. Mod. Phys. 68, 13 (1996).
[16] G. Kotliar and D. Vollhardt, Strongly Correlated Materials: Insights From Dynamical Mean-Field
Theory, Physics Today 57, 53 (2004).
[17] G. Kotliar, S. Y. Savrasov, K. Haule, V. S. Oudovenko, O. Parcollet, and C. A. Marianetti,
Electronic Structure Calculations with Dynamical Mean-Field Theory, Rev. Mod. Phys. 78, 865 (2006).
[18] R. Bulla, T. A. Costi, and D. Vollhardt, Finite-Temperature Numerical Renormalization Group Study of
the Mott Transition, Phys. Rev. B 64, 045103 (2001).
[19] P. Hansmann, A. Toschi, G. Sangiovanni, T. Saha-Dasgupta, S. Lupi, M. Marsi, and K. Held,
Mott-Hubbard Transition in V 2 O 3 Revisited, Phys. Status Solidi B 250, 1251 (2013).
[20] L. de’ Medici, J. Mravlje, and A. Georges, Janus-Faced Influence of Hund’s Rule Coupling in Strongly
Correlated Materials, Phys. Rev. Lett. 107, 256401 (2011).
151
REFERENCES
[21] S. R. Hassan, A. Georges, and H. R. Krishnamurthy, Sound Velocity Anomaly at the Mott
Transition: Application to Organic Conductors and V 2 O 3, Phys. Rev. Lett. 94, 036402 (2005).
[22] H. Kuwamoto, J. M. Honig, and J. Appel, Electrical Properties of the ( V 1 − x Cr x ) 2 O 3 System,
Phys. Rev. B 22, 2626 (1980).
[23] D. Golež, M. Eckstein, and P. Werner, Dynamics of Screening in Photodoped Mott Insulators, Phys.
Rev. B 92, 195123 (2015).
[24] M. Eckstein and P. Werner, Photoinduced States in a Mott Insulator, Phys. Rev. Lett. 110, 126401
(2013).
[25] D. B. McWhan, T. M. Rice, and J. P. Remeika, Mott Transition in Cr-Doped V 2 O 3, Phys. Rev.
Lett. 23, 1384 (1969).
[26] A. Jayaraman, D. B. McWhan, J. P. Remeika, and P. D. Dernier, Critical Behavior of the Mott
Transition in Cr-Doped V 2 O 3, Phys. Rev. B 2, 3751 (1970).
[27] D. B. McWhan and J. P. Remeika, Metal-Insulator Transition in (V1−xCrx)2O3, Phys. Rev. B 2, 3734
(1970).
[28] A. Menth and J. P. Remeika, Magnetic Properties of (V1-xCr2)2O3, Phys. Rev. B 2, 3756 (1970).
[29] D. B. McWhan, A. Menth, J. P. Remeika, W. F. Brinkman, and T. M. Rice, Metal-Insulator
Transitions in Pure and Doped V2O3, Phys. Rev. B 7, 1920 (1973).
[30] N. F. Mott, Metal-Insulator Transition, Rev. Mod. Phys. 40, 677 (1968).
[31] F. Rodolakis, B. Mansart, E. Papalazarou, S. Gorovikov, P. Vilmercati, L. Petaccia, A. Goldoni,
J. P. Rueff, S. Lupi, P. Metcalf, and M. Marsi, Quasiparticles at the Mott Transition in V 2 O 3 :
Wave Vector Dependence and Surface Attenuation, Phys. Rev. Lett. 102, 066805 (2009).
[32] N. Kuroda and H. Y. Fan, Raman Scattering and Phase Transitions of V2O3, Phys. Rev. B 16, 5003
(1977).
[33] S. Lupi, L. Baldassarre, B. Mansart, A. Perucchi, A. Barinov, P. Dudin, E. Papalazarou, F.
Rodolakis, J.-P. Rueff, J.-P. Itié, S. Ravy, D. Nicoletti, P. Postorino, P. Hansmann, N. Parragh,
A. Toschi, T. Saha-Dasgupta, O. K. Andersen, G. Sangiovanni, K. Held, and M. Marsi, A
Microscopic View on the Mott Transition in Chromium-Doped V2O3, Nature Communications 1, 1
(2010).
[34] F. Rodolakis, P. Hansmann, J.-P. Rueff, A. Toschi, M. W. Haverkort, G. Sangiovanni, A.
Tanaka, T. Saha-Dasgupta, O. K. Andersen, K. Held, M. Sikora, I. Alliot, J.-P. Itié, F. Baudelet,
P. Wzietek, P. Metcalf, and M. Marsi, Inequivalent Routes across the Mott Transition in V 2 O 3
Explored by X-Ray Absorption, Phys. Rev. Lett. 104, 047401 (2010).
[35] A. I. Frenkel, D. M. Pease, J. I. Budnick, P. Metcalf, E. A. Stern, P. Shanthakumar, and T.
Huang, Strain-Induced Bond Buckling and Its Role in Insulating Properties of Cr-Doped V 2 O 3, Phys.
Rev. Lett. 97, 195502 (2006).
[36] W. R. Robinson, High-Temperature Crystal Chemistry of V2O3 and 1% Chromium-Doped V2O3, Acta
Cryst B 31, 4 (1975).
[37] S. Chen, J. E. Hahn, C. E. Rice, and W. R. Robinson, The Effects of Titanium or Chromium Doping
on the Crystal Structure of V2O3, Journal of Solid State Chemistry 44, 192 (1982).
[38] J.-H. Park, L. H. Tjeng, A. Tanaka, J. W. Allen, C. T. Chen, P. Metcalf, J. M. Honig, F. M. F. de
Groot, and G. A. Sawatzky, Spin and Orbital Occupation and Phase Transitions in V 2 O 3, Phys.
Rev. B 61, 11506 (2000).
[39] A. I. Poteryaev, J. M. Tomczak, S. Biermann, A. Georges, A. I. Lichtenstein, A. N. Rubtsov, T.
Saha-Dasgupta, and O. K. Andersen, Enhanced Crystal-Field Splitting and Orbital-Selective Coherence
Induced by Strong Correlations in V 2 O 3, Phys. Rev. B 76, 085127 (2007).
[40] C. Castellani, C. R. Natoli, and J. Ranninger, Metal-Insulator Transition in Pure and Cr-Doped
V2O3, Phys. Rev. B 18, 5001 (1978).
[41] C. Castellani, C. R. Natoli, and J. Ranninger, Magnetic Structure of V2O3 in the Insulating Phase,
Phys. Rev. B 18, 4945 (1978).
152
REFERENCES
[42] K. Held, G. Keller, V. Eyert, D. Vollhardt, and V. I. Anisimov, Mott-Hubbard Metal-Insulator

Transition in Paramagnetic V 2 O 3 : An L D A + D M F T ( QMC ) Study, Phys. Rev. Lett. 86,
5345 (2001).
[43] G. Keller, K. Held, V. Eyert, D. Vollhardt, and V. I. Anisimov, Electronic Structure of Paramagnetic
V2O3: Strongly Correlated Metallic and Mott Insulating Phase, Phys. Rev. B 70, 205116 (2004).
[44] P. Werner, E. Gull, and A. J. Millis, Metal-Insulator Phase Diagram and Orbital Selectivity in Three-
Orbital Models with Rotationally Invariant Hund Coupling, Phys. Rev. B 79, 115119 (2009).
[45] H. B. Yaich, J. C. Jegaden, M. Potel, M. Sergent, A. K. Rastogi, and R. Tournier, Nouveaux
chalcogénures et chalcohalogénures à clusters tétraédriques Nb4 ou Ta4, Journal of the Less Common
Metals 102, 9 (1984).
[46] R. Pocha, D. Johrendt, and R. Pöttgen, Electronic and Structural Instabilities in GaV4S8 and
GaMo4S8, Chem. Mater. 12, 2882 (2000).
[47] V. Guiot, E. Janod, B. Corraze, and L. Cario, Control of the Electronic Properties and Resistive
Switching in the New Series of Mott Insulators GaTa 4 Se 8– y Te y (0 ≤ y ≤ 6.5), Chem. Mater. 23,
2611 (2011).
[48] A. Camjayi, C. Acha, R. Weht, M. G. Rodríguez, B. Corraze, E. Janod, L. Cario, and M. J.
Rozenberg, First-Order Insulator-to-Metal Mott Transition in the Paramagnetic 3D System GaTa 4 Se 8,
Phys. Rev. Lett. 113, 086404 (2014).
[49] M. Y. Jeong, S. H. Chang, B. H. Kim, J.-H. Sim, A. Said, D. Casa, T. Gog, E. Janod, L. Cario,
S. Yunoki, M. J. Han, and J. Kim, Direct Experimental Observation of the Molecular J Eff = 3/2
Ground State in the Lacunar Spinel GaTa4Se8, Nat Commun 8, 782 (2017).
[50] H. Lee, M. Y. Jeong, J.-H. Sim, H. Yoon, S. Ryee, and M. J. Han, Charge Density Functional plus U
Calculation of Lacunar Spinel GaM 4 Se 8 (M = Nb, Mo, Ta, and W), EPL 125, 47005 (2019).
[51] R. Pocha, D. Johrendt, B. Ni, and M. M. Abd-Elmeguid, Crystal Structures, Electronic Properties,
and Pressure-Induced Superconductivity of the Tetrahedral Cluster Compounds GaNb4S8, GaNb4Se8, and
GaTa4Se8, J. Am. Chem. Soc. 127, 8732 (2005).
[52] H. Müller, W. Kockelmann, and D. Johrendt, The Magnetic Structure and Electronic Ground States of
Mott Insulators GeV4S8 and GaV4S8, Chem. Mater. 18, 2174 (2006).
[53] M. M. Abd-Elmeguid, B. Ni, D. I. Khomskii, R. Pocha, D. Johrendt, X. Wang, and K. Syassen,
Transition from Mott Insulator to Superconductor in GaNb4Se8 and GaTa4Se8 under High Pressure, Phys.
Rev. Lett. 93, 126403 (2004).
[54] V. Ta Phuoc, C. Vaju, B. Corraze, R. Sopracase, A. Perucchi, C. Marini, P. Postorino, M.
Chligui, S. Lupi, E. Janod, and L. Cario, Optical Conductivity Measurements of GaTa 4 Se 8 Under
High Pressure: Evidence of a Bandwidth-Controlled Insulator-to-Metal Mott Transition, Phys. Rev. Lett.
110, 037401 (2013).
[55] E. Dorolti, L. Cario, B. Corraze, E. Janod, C. Vaju, H.-J. Koo, E. Kan, and M.-H. Whangbo,
Half-Metallic Ferromagnetism and Large Negative Magnetoresistance in the New Lacunar Spinel
GaTi3VS8, J. Am. Chem. Soc. 132, 5704 (2010).
[56] C. Vaju, J. Martial, E. Janod, B. Corraze, V. Fernandez, and L. Cario, Metal−Metal Bonding and
Correlated Metallic Behavior in the New Deficient Spinel Ga0.87Ti4S8, Chem. Mater. 20, 2382 (2008).
[57] E. Janod, E. Dorolti, B. Corraze, V. Guiot, S. Salmon, V. Pop, F. Christien, and L. Cario,
Negative Colossal Magnetoresistance Driven by Carrier Type in the Ferromagnetic Mott Insulator GaV4S8,
Chem. Mater. 27, 4398 (2015).
[58] K. Singh, C. Simon, E. Cannuccia, M.-B. Lepetit, B. Corraze, E. Janod, and L. Cario, Orbital-
Ordering-Driven Multiferroicity and Magnetoelectric Coupling in GeV 4 S 8, Phys. Rev. Lett. 113,
137602 (2014).
[59] I. Kézsmárki, S. Bordács, P. Milde, E. Neuber, L. M. Eng, J. S. White, H. M. Rønnow, C. D.
Dewhurst, M. Mochizuki, K. Yanai, H. Nakamura, D. Ehlers, V. Tsurkan, and A. Loidl, Néel-
Type Skyrmion Lattice with Confined Orientation in the Polar Magnetic Semiconductor GaV 4 S 8, Nature
Materials 14, 11 (2015).
153
REFERENCES
[60] Á. Butykai, S. Bordács, I. Kézsmárki, V. Tsurkan, A. Loidl, J. Döring, E. Neuber, P. Milde, S.

C. Kehr, and L. M. Eng, Characteristics of Ferroelectric-Ferroelastic Domains in Néel-Type Skyrmion Host
GaV 4 S 8, Scientific Reports 7, 1 (2017).
[61] J. Li, C. Aron, G. Kotliar, and J. E. Han, Electric-Field-Driven Resistive Switching in the Dissipative
Hubbard Model, Phys. Rev. Lett. 114, 226403 (2015).
[62] F. A. Chudnovskii, A. L. Pergament, G. B. Stefanovich, P. A. Metcalf, and J. M. Honig,
Switching Phenomena in Chromium-Doped Vanadium Sesquioxide, Journal of Applied Physics 84,
2643 (1998).
[63] F. A. Chudnovskii, A. L. Pergament, G. B. Stefanovich, P. Somasundaram, and J. M. Honig,
N-Type Negative Resistance in M/NiS2—XSex/M Structures, Physica Status Solidi (a) 161, 577
(1997).
[64] F. A. Chudnovskii, A. L. Pergament, P. Somasundaram, and J. M. Honig, Delay Time
Measurements of NiS2—XSex-Based Switches, Physica Status Solidi (a) 172, 131 (1999).
[65] J. Kim, C. Ko, A. Frenzel, S. Ramanathan, and J. E. Hoffman, Nanoscale Imaging and Control of
Resistance Switching in VO2 at Room Temperature, Appl. Phys. Lett. 96, 213106 (2010).
[66] A. Zimmers, L. Aigouy, M. Mortier, A. Sharoni, S. Wang, K. G. West, J. G. Ramirez, and I. K.
Schuller, Role of Thermal Heating on the Voltage Induced Insulator-Metal Transition in VO 2, Phys.
Rev. Lett. 110, 056601 (2013).
[67] F. Nakamura, M. Sakaki, Y. Yamanaka, S. Tamaru, T. Suzuki, and Y. Maeno, Electric-Field-
Induced Metal Maintained by Current of the Mott Insulator Ca 2 RuO 4, Scientific Reports 3, 1 (2013).
[68] J. S. Brockman, L. Gao, B. Hughes, C. T. Rettner, M. G. Samant, K. P. Roche, and S. S. P.
Parkin, Subnanosecond Incubation Times for Electric-Field-Induced Metallization of a Correlated Electron
Oxide, Nat Nano 9, 453 (2014).
[69] S. Guénon, S. Scharinger, S. Wang, J. G. Ramírez, D. Koelle, R. Kleiner, and I. K. Schuller,
Electrical Breakdown in a V 2 O 3 Device at the Insulator-to-Metal Transition, EPL 101, 57003 (2013).
[70] T. Burch, P. P. Craig, C. Hedrick, T. A. Kitchens, J. I. Budnick, J. A. Cannon, M. Lipsicas, and
D. Mattis, Switching in Magnetite: A Thermally Driven Magnetic Phase Transition, Phys. Rev. Lett. 23,
1444 (1969).
[71] A. A. Fursina, R. G. S. Sofin, I. V. Shvets, and D. Natelson, Origin of Hysteresis in Resistive
Switching in Magnetite Is Joule Heating, Phys. Rev. B 79, 245131 (2009).
[72] T. Driscoll, H.-T. Kim, B.-G. Chae, M. Di Ventra, and D. N. Basov, Phase-Transition Driven
Memristive System, Appl. Phys. Lett. 95, 043503 (2009).
[73] S.-H. Bae, S. Lee, H. Koo, L. Lin, B. H. Jo, C. Park, and Z. L. Wang, The Memristive Properties of
a Single VO2 Nanowire with Switching Controlled by Self-Heating, Adv. Mater. 25, 5098 (2013).
[74] M. D. Pickett and R. Stanley Williams, Sub-100 FJ and Sub-Nanosecond Thermally Driven Threshold
Switching in Niobium Oxide Crosspoint Nanodevices, Nanotechnology 23, 215202 (2012).
[75] M.-J. Lee, Y. Park, D.-S. Suh, E.-H. Lee, S. Seo, D.-C. Kim, R. Jung, B.-S. Kang, S.-E. Ahn, C.
B. Lee, D. H. Seo, Y.-K. Cha, I.-K. Yoo, J.-S. Kim, and B. H. Park, Two Series Oxide Resistors
Applicable to High Speed and High Density Nonvolatile Memory, Adv. Mater. 19, 3919 (2007).
[76] R. Waser and M. Aono, Nanoionics-Based Resistive Switching Memories, Nature Materials 6, 11
(2007).
[77] A. Sawa, Resistive Switching in Transition Metal Oxides, Materials Today 11, 28 (2008).
[78] S. Q. Liu, N. J. Wu, and A. Ignatiev, Electric-Pulse-Induced Reversible Resistance Change Effect in
Magnetoresistive Films, Appl. Phys. Lett. 76, 2749 (2000).
[79] H. S. Lee, S. G. Choi, H.-H. Park, and M. J. Rozenberg, A New Route to the Mott-Hubbard Metal-
Insulator Transition: Strong Correlations Effects in Pr0.7Ca0.3MnO3, Sci Rep 3, 1704 (2013).
[80] M. J. Rozenberg, M. J. Sánchez, R. Weht, C. Acha, F. Gomez-Marlasca, and P. Levy, Mechanism
for Bipolar Resistive Switching in Transition-Metal Oxides, Phys. Rev. B 81, 115101 (2010).
[81] C. Acha and M. J. Rozenberg, Non-Volatile Resistive Switching in the Dielectric Superconductor
YBa2Cu3O7−δ, J. Phys.: Condens. Matter 21, 045702 (2009).
154
REFERENCES
[82] D. C. Kim, S. Seo, S. E. Ahn, D.-S. Suh, M. J. Lee, B.-H. Park, I. K. Yoo, I. G. Baek, H.-J.
Kim, E. K. Yim, J. E. Lee, S. O. Park, H. S. Kim, U.-I. Chung, J. T. Moon, and B. I. Ryu,
Electrical Observations of Filamentary Conductions for the Resistive Memory Switching in NiO Films, Appl.
Phys. Lett. 88, 202102 (2006).
[83] K. Kinoshtia, T. Okutani, H. Tanaka, T. Hinoki, K. Yazawa, K. Ohmi, and S. Kishida, Opposite
Bias Polarity Dependence of Resistive Switching in N-Type Ga-Doped-ZnO and p-Type NiO Thin Films,
Appl. Phys. Lett. 96, 143505 (2010).
[84] K. Fujiwara, T. Nemoto, M. J. Rozenberg, Y. Nakamura, and H. Takagi, Resistance Switching and
Formation of a Conductive Bridge in Metal/Binary Oxide/Metal Structure for Memory Devices, Jpn. J.
Appl. Phys. 47, 6266 (2008).
[85] T. Yajima, K. Fujiwara, A. Nakao, T. Kobayashi, T. Tanaka, K. Sunouchi, Y. Suzuki, M.
Takeda, K. Kojima, Y. Nakamura, K. Taniguchi, and H. Takagi, Spatial Redistribution of Oxygen
Ions in Oxide Resistance Switching Device after Forming Process, Jpn. J. Appl. Phys. 49, 060215 (2010).
[86] H. Shima, F. Takano, Y. Tamai, H. Akinaga, and I. H. Inoue, Synthesis and Characterization of
Pt/Co–O/Pt Trilayer Exhibiting Large Reproducible Resistance Switching, Jpn. J. Appl. Phys. 46, L57
(2007).
[87] S. B. Lee, S. H. Chang, H. K. Yoo, and B. S. Kang, Stabilizing the Forming Process in Unipolar
Resistance Switching Using an Improved Compliance Current Limiter, J. Phys. D: Appl. Phys. 43,
485103 (2010).
[88] S. Zhang, S. Long, W. Guan, Q. Liu, Q. Wang, and M. Liu, Resistive Switching Characteristics of
MnOx -Based ReRAM, J. Phys. D: Appl. Phys. 42, 055112 (2009).
[89] K. M. Kim, D. S. Jeong, and C. S. Hwang, Nanofilamentary Resistive Switching in Binary Oxide
System; a Review on the Present Status and Outlook, Nanotechnology 22, 254002 (2011).
[90] Y. Taguchi, T. Matsumoto, and Y. Tokura, Dielectric Breakdown of One-Dimensional Mott Insulators
Sr 2 CuO 3 and SrCuO 2, Phys. Rev. B 62, 7015 (2000).
[91] S. Yamanouchi, Y. Taguchi, and Y. Tokura, Dielectric Breakdown of the Insulating Charge-Ordered
State in La 2 − x Sr x NiO 4, Phys. Rev. Lett. 83, 5555 (1999).
[92] P. Stoliar, L. Cario, E. Janod, B. Corraze, C. Guillot-Deudon, S. Salmon-Bourmand, V. Guiot,
J. Tranchant, and M. Rozenberg, Universal Electric-Field-Driven Resistive Transition in Narrow-Gap
Mott Insulators, Adv. Mater. 25, 3222 (2013).
[93] M. Querré, E. Janod, L. Cario, J. Tranchant, B. Corraze, V. Bouquet, S. Deputier, S. Cordier,
M. Guilloux-Viry, and M.-P. Besland, Metal–Insulator Transitions in (V1-XCrx)2O3 Thin Films
Deposited by Reactive Direct Current Magnetron Co-Sputtering, Thin Solid Films 617, 56 (2016).
[94] L. Cario, C. Vaju, B. Corraze, V. Guiot, and E. Janod, Electric-Field-Induced Resistive Switching in a
Family of Mott Insulators: Towards a New Class of RRAM Memories, Adv. Mater. 22, 5193 (2010).
[95] R. Kumai, Y. Okimoto, and Y. Tokura, Current-Induced Insulator-Metal Transition and Pattern
Formation in an Organic Charge-Transfer Complex, Science 284, 1645 (1999).
[96] F. Sabeth, T. Iimori, and N. Ohta, Insulator–Metal Transitions Induced by Electric Field and
Photoirradiation in Organic Mott Insulator Deuterated κ-(BEDT-TTF) 2 Cu[N(CN) 2 ]Br, J. Am. Chem.
Soc. 134, 6984 (2012).
[97] P. Stoliar, P. Diener, J. Tranchant, B. Corraze, B. Brière, V. Ta-Phuoc, N. Bellec, M.
Fourmigué, D. Lorcy, E. Janod, and L. Cario, Resistive Switching Induced by Electric Pulses in a
Single-Component Molecular Mott Insulator, J. Phys. Chem. C 119, 2983 (2015).
[98] T. Oka, R. Arita, and H. Aoki, Breakdown of a Mott Insulator: A Nonadiabatic Tunneling Mechanism,
Phys. Rev. Lett. 91, 066406 (2003).
[99] T. Oka and H. Aoki, Ground-State Decay Rate for the Zener Breakdown in Band and Mott Insulators,
Phys. Rev. Lett. 95, 137601 (2005).
[100]F. Heidrich-Meisner, I. González, K. A. Al-Hassanieh, A. E. Feiguin, M. J. Rozenberg, and E.
Dagotto, Nonequilibrium Electronic Transport in a One-Dimensional Mott Insulator, Phys. Rev. B 82,
205110 (2010).
155
REFERENCES
[101]M. Eckstein, T. Oka, and P. Werner, Dielectric Breakdown of Mott Insulators in Dynamical Mean-
Field Theory, Phys. Rev. Lett. 105, 146404 (2010).
[102]H. Aoki, N. Tsuji, M. Eckstein, M. Kollar, T. Oka, and P. Werner, Nonequilibrium Dynamical
Mean-Field Theory and Its Applications, Rev. Mod. Phys. 86, 779 (2014).
[103]V. Guiot, L. Cario, E. Janod, B. Corraze, V. Ta Phuoc, M. Rozenberg, P. Stoliar, T. Cren, and
D. Roditchev, Avalanche Breakdown in GaTa4Se8−xTex Narrow-Gap Mott Insulators, Nat Commun
4, 1722 (2013).
[104]M. E. Levinshtein, J. Kostamovaara, and S. Vainshtein, Breakdown Phenomena in Semiconductors
and Semiconductor Devices, Int. J. Hi. Spe. Ele. Syst. 14, 921 (2004).
[105] J. L. Hudgins, Wide and Narrow Bandgap Semiconductors for Power Electronics: A New Valuation,
Journal of Elec Materi 32, 471 (2003).
[106] J. L. Hudgins, G. S. Simin, E. Santi, and M. A. Khan, An Assessment of Wide Bandgap
Semiconductors for Power Devices, IEEE Trans. Power Electron. 18, 907 (2003).
[107]H. Fröhlich and N. F. Mott, On the Theory of Dielectric Breakdown in Solids, Proceedings of the
Royal Society of London. Series A. Mathematical and Physical Sciences 188, 521 (1947).
[108] S. Whitehead, Dielectric Breakdown of Solids (Clarendon Press, 1951).
[109]H. Fröhlich and N. F. Mott, Theory of Electrical Breakdown in Ionic Crystals, Proceedings of the
Royal Society of London. Series A - Mathematical and Physical Sciences 160, 230 (1937).
[110] F. Seitz, On the Theory of Electron Multiplication in Crystals, Phys. Rev. 76, 1376 (1949).
[111]H. Fröhlich and F. Seitz, Notes on the Theory of Dielectric Breakdown in Ionic Crystals, Phys. Rev. 79,
526 (1950).
[112]C. Ye, P. Cai, R. Yu, X. Zhou, W. Ruan, Q. Liu, C. Jin, and Y. Wang, Visualizing the Atomic-
Scale Electronic Structure of the Ca2CuO2Cl2 Mott Insulator, Nat Commun 4, 1365 (2013).
[113] P. Stoliar, M. Rozenberg, E. Janod, B. Corraze, J. Tranchant, and L. Cario, Nonthermal and
Purely Electronic Resistive Switching in a Mott Memory, Phys. Rev. B 90, 045146 (2014).
[114] C. Adda, L. Cario, J. Tranchant, E. Janod, M.-P. Besland, M. Rozenberg, P. Stoliar, and B.
Corraze, First Demonstration of “Leaky Integrate and Fire” Artificial Neuron Behavior on
(V0.95Cr0.05)2O3 Thin Film, MRC 8, 835 (2018).
[115]F. Tesler, C. Adda, J. Tranchant, B. Corraze, E. Janod, L. Cario, P. Stoliar, and M. Rozenberg,
Relaxation of a Spiking Mott Artificial Neuron, Phys. Rev. Applied 10, 054001 (2018).
[116]C. Adda, B. Corraze, P. Stoliar, P. Diener, J. Tranchant, A. Filatre-Furcate, M. Fourmigué, D.
Lorcy, M.-P. Besland, E. Janod, and L. Cario, Mott Insulators: A Large Class of Materials for Leaky
Integrate and Fire (LIF) Artificial Neuron, Journal of Applied Physics 124, 152124 (2018).
[117]H.-S. P. Wong, S. Raoux, S. Kim, J. Liang, J. P. Reifenberg, B. Rajendran, M. Asheghi, and K.
E. Goodson, Phase Change Memory, Proceedings of the IEEE 98, 2201 (2010).
[118] D. B. McWhan, J. P. Remeika, T. M. Rice, W. F. Brinkman, J. P. Maita, and A. Menth,
Electronic Specific Heat of Metallic Ti-Doped V203, PHYSICAL REVIEW LETTERS 27, 3 (1971).
[119] ITRS 2.0 Home Page, http://www.itrs2.net/.
[120] Y. Fujisaki, Review of Emerging New Solid-State Non-Volatile Memories, Jpn. J. Appl. Phys. 52,
040001 (2013).
[121] J.-G. Zhu, Magnetoresistive Random Access Memory: The Path to Competitiveness and Scalability,
Proceedings of the IEEE 96, 1786 (2008).
[122](Invited) Control of Resistive Switching in Mott Memories Based on TiN/AM4Q8/TiN MIM Devices -
IOPscience, https://iopscience.iop.org/article/10.1149/07532.0003ecst.
[123]E. Souchier, L. Cario, B. Corraze, C. Estournes, V. Fernandez, T. Skotnicki, P. Mazoyer, E.
Janod, and M.-P. Besland, Thin Layers Obtained by Plasma Process for Emerging Non-Volatile Memory
(RRAM) Applications, in 2009 IEEE International Memory Workshop (IEEE, Monterey, CA, USA,
2009), pp. 1–2.
[124]E. Souchier, M.-P. Besland, J. Tranchant, B. Corraze, P. Moreau, R. Retoux, C. Estournès, P.
Mazoyer, L. Cario, and E. Janod, Deposition by Radio Frequency Magnetron Sputtering of GaV4S8
Thin Films for Resistive Random Access Memory Application, Thin Solid Films 533, 54 (2013).
156
REFERENCES
[125]J. Tranchant, A. Pellaroque, E. Janod, B. Angleraud, B. Corraze, L. Cario, and M.-P. Besland,
Deposition of GaV4S8 Thin Films by H2S/Ar Reactive Sputtering for ReRAM Applications, J. Phys. D:
Appl. Phys. 47, 065309 (2014).
[126] J. Tranchant, E. Janod, B. Corraze, M.-P. Besland, and L. Cario, From Resistive Switching
Mechanisms in AM4Q8 Mott Insulators to Mott Memories, in 2015 IEEE International Memory
Workshop (IMW) (2015), pp. 1–4.
[127] J. Tranchant, E. Janod, L. Cario, B. Corraze, E. Souchier, J.-L. Leclercq, P. Cremillieu, P.
Moreau, and M.-P. Besland, Electrical Characterizations of Resistive Random Access Memory Devices
Based on GaV4S8 Thin Layers, Thin Solid Films 533, 61 (2013).
[128] J. Tranchant, M. Querre, E. Janod, M.-P. Besland, B. Corraze, and L. Cario, Mott Memory
Devices Based on the Mott Insulator (V1-XCrx)2O3, in 2018 IEEE International Memory Workshop
(IMW) (2018), pp. 1–4.
[129]G. Indiveri and T. K. Horiuchi, Frontiers in Neuromorphic Engineering, Front. Neurosci. 5, (2011).
[130] C. Mead, Neuromorphic Electronic Systems, Proceedings of the IEEE 78, 1629 (1990).
[131] A. K. Jain, Jianchang Mao, and K. M. Mohiuddin, Artificial Neural Networks: A Tutorial,
Computer 29, 31 (1996).
[132]K. Boahen, Compact, Efficient Electronics Based on the Brain’s Neural System Could Yield Implantable
Silicon Retinas to Restore Vision, as Well as Robotic Eyes and Other Smart Sensors, 8 (2005).
[133]J. Misra and I. Saha, Artificial Neural Networks in Hardware: A Survey of Two Decades of Progress,
Neurocomputing 74, 239 (2010).
[134] L. Lapicque, Recherches Quantitatives Sur l’excitation Électrique Des Nerfs Traitée Comme Une
Polarisation, J. Physiol. Pathol. Gen. (1907).
[135] N. Brunel and M. C. W. van Rossum, Quantitative Investigations of Electrical Nerve Excitation
Treated as Polarization: Louis Lapicque 1907 · Translated By:, Biol Cybern 97, 341 (2007).
[136]P. Stoliar, J. Tranchant, B. Corraze, E. Janod, M.-P. Besland, F. Tesler, M. Rozenberg, and L.
Cario, A Leaky-Integrate-and-Fire Neuron Analog Realized with a Mott Insulator, Adv. Funct. Mater.
27, 1604740 (2017).
[137]C. Giannetti, M. Capone, D. Fausti, M. Fabrizio, F. Parmigiani, and D. Mihailovic, Ultrafast
Optical Spectroscopy of Strongly Correlated Materials and High-Temperature Superconductors: A Non-
Equilibrium Approach, Advances in Physics 65, 58 (2016).
[138] K. Nasu, editor , Photoinduced Phase Transitions (World Scientific, Hackensack, N.J, 2004).
[139]K. Yonemitsu and K. Nasu, Theory of Photoinduced Phase Transitions in Itinerant Electron Systems,
Physics Reports 465, 1 (2008).
[140] J. Zhang and R. D. Averitt, Dynamics and Control in Complex Transition Metal Oxides, Annual
Review of Materials Research 44, 1 (2014).
[141] D. N. Basov, R. D. Averitt, D. van der Marel, M. Dressel, and K. Haule, Electrodynamics of
Correlated Electron Materials, Rev. Mod. Phys. 83, 471 (2011).
[142]R. D. Averitt and A. J. Taylor, Ultrafast Optical and Far-Infrared Quasiparticle Dynamics in Correlated
Electron Materials, J. Phys.: Condens. Matter 14, R1357 (2002).
[143] E. Abreu, S. Wang, J. G. Ramírez, M. Liu, J. Zhang, K. Geng, I. K. Schuller, and R. D.
Averitt, Dynamic Conductivity Scaling in Photoexcited V2O3 Thin Films, Phys. Rev. B 92, 085130
(2015).
[144]S. Lysenko, V. Vikhnin, A. Rúa, F. Fernández, and H. Liu, Critical Behavior and Size Effects in
Light-Induced Transition of Nanostructured VO 2 Films, Phys. Rev. B 82, 205425 (2010).
[145] P. Papon, J. Leblond, and P. H. E. Meijer, The Physics of Phase Transitions: Concepts and
Applications, 2nd ed. (Springer-Verlag, Berlin Heidelberg, 2006).
[146]X. Liu, H. Li, and M. Zhan, A Review on the Modeling and Simulations of Solid-State Diffusional Phase
Transformations in Metals and Alloys, Manufacturing Rev. 5, 10 (2018).
[147] C. Thomsen, J. Strait, Z. Vardeny, H. J. Maris, J. Tauc, and J. J. Hauser, Coherent Phonon
Generation and Detection by Picosecond Light Pulses, Phys. Rev. Lett. 53, 989 (1984).
157
REFERENCES
[148]C. Thomsen, H. T. Grahn, H. J. Maris, and J. Tauc, Surface Generation and Detection of Phonons by
Picosecond Light Pulses, Physical Review B 34, 4129 (1986).
[149] O. Matsuda, M. C. Larciprete, R. Li Voti, and O. B. Wright, Fundamentals of Picosecond Laser
Ultrasonics, Ultrasonics 56, 3 (2015).
[150]P. Ruello and V. E. Gusev, Physical Mechanisms of Coherent Acoustic Phonons Generation by Ultrafast
Laser Action, Ultrasonics 56, 21 (2015).
[151]P. Werner and M. Eckstein, Effective Doublon and Hole Temperatures in the Photo-Doped Dynamic
Hubbard Model, Struct. Dyn. 3, 023603 (2016).
[152]C. Mariette, M. Lorenc, H. Cailleau, E. Collet, L. Guérin, A. Volte, E. Trzop, R. Bertoni, X.
Dong, B. Lépine, O. Hernandez, E. Janod, L. Cario, V. T. Phuoc, S. Ohkoshi, H. Tokoro, L.
Patthey, I. Usov, D. Ozerov, L. Sala, S. Ebner, P. Böhler, A. Keller, A. Oggenfuss, T. Zmofing,
S. Vetter, R. Follath, P. Juranic, A. Schreiber, P. Beaud, V. Esposito, Y. Deng, G. Ingold, M.
Chergui, G. F. Mancini, R. Mankowsky, C. Svetina, S. Zerdane, A. Mozzanica, M. Wulff, H.
Lemke, and M. Cammarata, Strain Wave Pathway to Semiconductor-to-Metal Transition Revealed by
Time Resolved X-Ray Powder Diffraction, Submited to Nature Commun. 30 (n.d.).
[153]H.-T. Kim, Y. W. Lee, B.-J. Kim, B.-G. Chae, S. J. Yun, K.-Y. Kang, K.-J. Han, K.-J. Yee, and
Y.-S. Lim, Monoclinic and Correlated Metal Phase in VO 2 as Evidence of the Mott Transition: Coherent
Phonon Analysis, Phys. Rev. Lett. 97, 266401 (2006).
[154] A. Singer, J. G. Ramirez, I. Valmianski, D. Cela, N. Hua, R. Kukreja, J. Wingert, O.
Kovalchuk, J. M. Glownia, M. Sikorski, M. Chollet, M. Holt, I. K. Schuller, and O. G. Shpyrko,
Nonequilibrium Phase Precursors during a Photoexcited Insulator-to-Metal Transition in V2O3, Phys. Rev.
Lett. 120, 207601 (2018).
[155] O. V. Misochko, M. Tani, K. Sakai, K. Kisoda, S. Nakashima, V. N. Andreev, and F. A.
Chudnovsky, Optical Study of the Mott Transition in V2O3 Comparison of Time- and Frequency-Domain
Results, Phys. Rev. B 58, 12789 (1998).
[156]M. K. Liu, B. Pardo, J. Zhang, M. M. Qazilbash, S. J. Yun, Z. Fei, J.-H. Shin, H.-T. Kim, D.
N. Basov, and R. D. Averitt, Photoinduced Phase Transitions by Time-Resolved Far-Infrared Spectroscopy
in V2O3, Phys. Rev. Lett. 107, 066403 (2011).
[157]F. Giorgianni, J. Sakai, and S. Lupi, Overcoming the Thermal Regime for the Electric-Field Driven Mott
Transition in Vanadium Sesquioxide, Nature Communications 10, 1 (2019).
[158] I. A. Mogunov, S. Lysenko, A. E. Fedianin, F. E. Fernández, A. Rúa, A. J. Kent, A. V.
Akimov, and A. M. Kalashnikova, Large Non-Thermal Contribution to Picosecond Strain Pulse
Generation Using the Photo-Induced Phase Transition in VO2, Nat Commun 11, 1690 (2020).
[159]H. Yamakawa, T. Miyamoto, T. Morimoto, T. Terashige, H. Yada, N. Kida, M. Suda, H. M.
Yamamoto, R. Kato, K. Miyagawa, K. Kanoda, and H. Okamoto, Mott Transition by an Impulsive
Dielectric Breakdown, Nature Mater 16, 1100 (2017).
[160]J. H. Mentink and M. Eckstein, Ultrafast Quenching of the Exchange Interaction in a Mott Insulator,
Phys. Rev. Lett. 113, 057201 (2014).
[161] J. H. Mentink, K. Balzer, and M. Eckstein, Ultrafast and Reversible Control of the Exchange
Interaction in Mott Insulators, Nat Commun 6, 6708 (2015).
[162] T. Ishikawa, Y. Sagae, Y. Naitoh, Y. Kawakami, H. Itoh, K. Yamamoto, K. Yakushi, H.
Kishida, T. Sasaki, S. Ishihara, Y. Tanaka, K. Yonemitsu, and S. Iwai, Optical Freezing of Charge
Motion in an Organic Conductor, Nature Communications 5, 1 (2014).
[163] F. Grandi, J. Li, and M. Eckstein, Ultrafast Mott Transition Driven by Nonlinear Phonons,
ArXiv:2005.14100 [Cond-Mat] (2020).
[164]M. Först, C. Manzoni, S. Kaiser, Y. Tomioka, Y. Tokura, R. Merlin, and A. Cavalleri, Nonlinear
Phononics as an Ultrafast Route to Lattice Control, Nature Physics 7, 11 (2011).
[165] A. Rousse, C. Rischel, S. Fourmaux, I. Uschmann, S. Sebban, G. Grillon, Ph. Balcou, E.
Förster, J. P. Geindre, P. Audebert, J. C. Gauthier, and D. Hulin, Non-Thermal Melting in
Semiconductors Measured at Femtosecond Resolution, Nature 410, 65 (2001).
158
REFERENCES
[166] P. B. Hillyard, K. J. Gaffney, A. M. Lindenberg, S. Engemann, R. A. Akre, J. Arthur, C.

Blome, P. H. Bucksbaum, A. L. Cavalieri, A. Deb, R. W. Falcone, D. M. Fritz, P. H. Fuoss, J.
Hajdu, P. Krejcik, J. Larsson, S. H. Lee, D. A. Meyer, A. J. Nelson, R. Pahl, D. A. Reis, J.
Rudati, D. P. Siddons, K. Sokolowski-Tinten, D. von der Linde, and J. B. Hastings, Carrier-
Density-Dependent Lattice Stability in InSb, Phys. Rev. Lett. 98, 125501 (2007).
[167]S. L. Johnson, P. Beaud, C. J. Milne, F. S. Krasniqi, E. S. Zijlstra, M. E. Garcia, M. Kaiser, D.
Grolimund, R. Abela, and G. Ingold, Nanoscale Depth-Resolved Coherent Femtosecond Motion in
Laser-Excited Bismuth, Phys. Rev. Lett. 100, 155501 (2008).
[168]S. Iwai, M. Ono, A. Maeda, H. Matsuzaki, H. Kishida, H. Okamoto, and Y. Tokura, Ultrafast
Optical Switching to a Metallic State by Photoinduced Mott Transition in a Halogen-Bridged Nickel-Chain
Compound, Phys. Rev. Lett. 91, 057401 (2003).
[169]G. Lantz, B. Mansart, D. Grieger, D. Boschetto, N. Nilforoushan, E. Papalazarou, N. Moisan,
L. Perfetti, V. L. R. Jacques, D. L. Bolloc’h, C. Laulhé, S. Ravy, J.-P. Rueff, T. E. Glover, M. P.
Hertlein, Z. Hussain, S. Song, M. Chollet, M. Fabrizio, and M. Marsi, Ultrafast Evolution and
Transient Phases of the Prototype Out-of-Equilibrium Mott-Hubbard Material V2O3, Nature
Communications 8, 13917 (2017).
[170] A. Cavalleri, Cs. Tóth, C. W. Siders, J. A. Squier, F. Ráksi, P. Forget, and J. C. Kieffer,
Femtosecond Structural Dynamics in VO 2 during an Ultrafast Solid-Solid Phase Transition, Phys. Rev.
Lett. 87, 237401 (2001).
[171] E. Abreu, S. N. Gilbert Corder, S. J. Yun, S. Wang, J. G. Ramírez, K. West, J. Zhang, S.
Kittiwatanakul, I. K. Schuller, J. Lu, S. A. Wolf, H.-T. Kim, M. Liu, and R. D. Averitt, Ultrafast
Electron-Lattice Coupling Dynamics in VO2 and V2O3 Thin Films, Phys. Rev. B 96, 094309 (2017).
[172]D. J. Hilton, R. P. Prasankumar, S. Fourmaux, A. Cavalleri, D. Brassard, M. A. El Khakani, J.
C. Kieffer, A. J. Taylor, and R. D. Averitt, Enhanced Photosusceptibility near T c for the Light-Induced
Insulator-to-Metal Phase Transition in Vanadium Dioxide, Phys. Rev. Lett. 99, 226401 (2007).
[173]B. Mansart, A. Barinov, P. Dudin, L. Baldassarre, A. Perucchi, E. Papalazarou, P. Metcalf, S.
Lupi, and M. Marsi, Photoemission Microscopy Study of the Two Metal-Insulator Transitions in Cr-Doped
V 2 O 3, Appl. Phys. Lett. 100, 014108 (2012).
[174]H. S. Choi, J. S. Ahn, J. H. Jung, T. W. Noh, and D. H. Kim, Mid-Infrared Properties of a VO2
Film near the Metal-Insulator Transition, Phys. Rev. B 54, 4621 (1996).
[175]G. A. Thomas, D. H. Rapkine, S. A. Carter, A. J. Millis, T. F. Rosenbaum, P. Metcalf, and J.
M. Honig, Observation of the Gap and Kinetic Energy in a Correlated Insulator, Phys. Rev. Lett. 73,
1529 (1994).
[176]R. Bertoni, M. Lorenc, H. Cailleau, A. Tissot, J. Laisney, M.-L. Boillot, L. Stoleriu, A. Stancu,
C. Enachescu, and E. Collet, Elastically Driven Cooperative Response of a Molecular Material Impacted
by a Laser Pulse, Nature Mater 15, 606 (2016).
[177]J. A. J. Rupp, M. Querré, A. Kindsmüller, M.-P. Besland, E. Janod, R. Dittmann, R. Waser,
and D. J. Wouters, Different Threshold and Bipolar Resistive Switching Mechanisms in Reactively Sputtered
Amorphous Undoped and Cr-Doped Vanadium Oxide Thin Films, Journal of Applied Physics 123,
044502 (2018).
[178]D. J. Wouters, S. Menzel, J. A. J. Rupp, T. Hennen, and R. Waser, On the Universality of the I –
V Switching Characteristics in Non-Volatile and Volatile Resistive Switching Oxides, Faraday Discuss.
213, 183 (2019).
[179] Y. Chernukha, Investigation of Phase Transitions Triggered by Laser-Induced Focusing
Shock Waves, PhD thesis, Université du Maine, 2019.
[180] T. Pezeril, G. Saini, D. Veysset, S. Kooi, P. Fidkowski, R. Radovitzky, and K. A. Nelson,
Direct Visualization of Laser-Driven Focusing Shock Waves, Phys. Rev. Lett. 106, 214503 (2011).
[181] F. D. Murnaghan, The Compressibility of Media under Extreme Pressures, PNAS 30, 244 (1944).
[182] F. Birch, Finite Elastic Strain of Cubic Crystals, Phys. Rev. 71, 809 (1947).
[183] J. J. Gilman, Bulk Stiffnesses of Metals, Materials Science and Engineering 7, 357 (1971).
159
REFERENCES
[184] S. Populoh, P. Wzietek, R. Gohier, and P. Metcalf, Lattice Softening Effects at the Mott Critical
Point of Cr-Doped V 2 O 3, Phys. Rev. B 84, 075158 (2011).
[185]E. Gati, M. Garst, R. S. Manna, U. Tutsch, B. Wolf, L. Bartosch, H. Schubert, T. Sasaki, J. A.
Schlueter, and M. Lang, Breakdown of Hooke’s Law of Elasticity at the Mott Critical Endpoint in an
Organic Conductor, Science Advances 2, e1601646 (2016).
[186]P. Majumdar and H. R. Krishnamurthy, Lattice Contraction Driven Insulator-Metal Transition in the
d = ∞ Local Approximation, Phys. Rev. Lett. 73, 1525 (1994).
d = ∞ Local Approximation, Phys. Rev. Lett. 74, 3303 (1995).
[188]H. Yang and R. J. Sladek, Lattice-Dynamical Model for the Elastic Constants and Raman Frequencies in
( V 1 − x Cr x ) 2 O 3, Phys. Rev. B 32, 6634 (1985).
[189]F. Rodolakis, Spectroscopies à l’aide du rayonnement synchrotron appliquées aux systèmes
fortement corrélés: Transition métal-isolant dans les oxydes de vanadium, Université Paris Sud
- Paris XI, 2009.
[190] M. T. Dove, Structure and Dynamics: An Atomic View of Materials, Reprinted (Oxford Univ.
Press, Oxford, 2008).
[191]A. Georges, S. Florens, and T. A. Costi, The Mott Transition: Unconventional Transport, Spectral
Weight Transfers, and Critical Behaviour, J. Phys. IV France 114, 165 (2004).
[192] D. Fournier, M. Poirier, M. Castonguay, and K. D. Truong, Mott Transition, Compressibility
Divergence, and the P − T Phase Diagram of Layered Organic Superconductors: An Ultrasonic Investigation,
Phys. Rev. Lett. 90, 127002 (2003).
[193]C. Marini, A. Perucchi, D. Chermisi, P. Dore, M. Valentini, D. Topwal, D. D. Sarma, S. Lupi,
and P. Postorino, Combined Raman and Infrared Investigation of the Insulator-to-Metal Transition in NiS
2 − x Se x Compounds, Phys. Rev. B 84, 235134 (2011).
[194]G. Mazza, A. Amaricci, M. Capone, and M. Fabrizio, Field-Driven Mott Gap Collapse and Resistive
Switch in Correlated Insulators, Phys. Rev. Lett. 117, 176401 (2016).
[195] D. B. McWhan, A. Menth, J. P. Remeika, W. F. Brinkman, and T. M. Rice, Metal-Insulator
Transitions in Pure and Doped V2O3, Phys. Rev. B 7, 1920 (1973).
[196]I. Lo Vecchio, L. Baldassarre, F. D’Apuzzo, O. Limaj, D. Nicoletti, A. Perucchi, L. Fan, P.
Metcalf, M. Marsi, and S. Lupi, Optical Properties of V2O3 in Its Whole Phase Diagram, Phys. Rev. B
91, 155133 (2015).
[197] C. Tatsuyama and H. Y. Fan, Raman Scattering and Phase Transitions in V2O3 and (V1-
XCrx)2O3, Physical Review B 21, 2977 (1980).
[198]J. del Valle, Y. Kalcheim, J. Trastoy, A. Charnukha, D. N. Basov, and I. K. Schuller, Electrically
Induced Multiple Metal-Insulator Transitions in Oxide Nanodevices, Phys. Rev. Applied 8, 054041
(2017).
[199] H. Kuwamoto and J. M. Honig, Electrical Properties and Structure of Cr-Doped Nonstoichiometric
V2O3, Journal of Solid State Chemistry 32, 335 (1980).
[200]Y. Ueda, K. Kosuge, and S. Kachi, Phase Diagram and Some Physical Properties of V2O3+x∗ (0 ⩽
x ⩽ 0.080), Journal of Solid State Chemistry 31, 171 (1980).
$d=\ensuremath{\infty}$ Local Approximation, Phys. Rev. Lett. 73, 1525 (1994).
[202]A. Georges, S. Florens, and T. A. Costi, The Mott Transition: Unconventional Transport, Spectral
Weight Transfers, and Critical Behaviour, J. Phys. IV France 114, 165 (2004).
[203]C. Vaju, L. Cario, B. Corraze, E. Janod, V. Dubost, T. Cren, D. Roditchev, D. Braithwaite,
and O. Chauvet, Electric-Pulse-Driven Electronic Phase Separation, Insulator–Metal Transition, and
Possible Superconductivity in a Mott Insulator, Advanced Materials 20, 2760 (2008).
[204] V. Ta Phuoc, C. Vaju, B. Corraze, R. Sopracase, A. Perucchi, C. Marini, P. Postorino, M.
Chligui, S. Lupi, E. Janod, and L. Cario, Optical Conductivity Measurements of GaTa4Se8 Under High
Pressure: Evidence of a Bandwidth-Controlled Insulator-to-Metal Mott Transition, Phys. Rev. Lett. 110,
037401 (2013).
160
REFERENCES
[205] M. M. Abd-Elmeguid, B. Ni, D. I. Khomskii, R. Pocha, D. Johrendt, X. Wang, and K.

Syassen, Transition from Mott Insulator to Superconductor in G a N b 4 S e 8 and G a T a 4 S e 8 under
High Pressure, Physical Review Letters 93, (2004).
[206] R. Mandal, Elastic Effects in Ultrafast Photoinduced Phase Transition, Master thesis,
Université de Rennes 1, 2020.
[207]E. Goi, Q. Zhang, X. Chen, H. Luan, and M. Gu, Perspective on Photonic Memristive Neuromorphic
Computing, PhotoniX 1, 3 (2020).
[208] P. R. Prucnal and B. J. Shastri, Neuromorphic Photonics (CRC Press, 2017).
[209]C. Adda, B. Corraze, P. Stoliar, P. Diener, J. Tranchant, A. Filatre-Furcate, M. Fourmigué, D.
Lorcy, M.-P. Besland, E. Janod, and L. Cario, Mott Insulators: A Large Class of Materials for Leaky
Integrate and Fire (LIF) Artificial Neuron, Journal of Applied Physics 124, 152124 (2018).
[210] P. R. Prucnal and B. J. Shastri, Neuromorphic Photonics (CRC Press, 2017).
[211] M. S. Tyagi and R. Van Overstraeten, Minority Carrier Recombination in Heavily-Doped Silicon,
Solid-State Electronics 26, 577 (1983).
[212]J. A. del Alamo and R. M. Swanson, Modelling of Minority-Carrier Transport in Heavily Doped Silicon
Emitters, Solid-State Electronics 30, 1127 (1987).
[213] K. Fukumoto, K. Onda, Y. Yamada, T. Matsuki, T. Mukuta, S. Tanaka, and S. Koshihara,
Femtosecond Time-Resolved Photoemission Electron Microscopy for Spatiotemporal Imaging of Photogenerated
Carrier Dynamics in Semiconductors, Review of Scientific Instruments 85, 083705 (2014).
[214] L. Perfetti, P. A. Loukakos, M. Lisowski, U. Bovensiepen, H. Berger, S. Biermann, P. S.
Cornaglia, A. Georges, and M. Wolf, Time Evolution of the Electronic Structure of 1 T − TaS 2
through the Insulator-Metal Transition, Phys. Rev. Lett. 97, 067402 (2006).
[215]C. Miller, M. Triplett, J. Lammatao, J. Suh, D. Fu, J. Wu, and D. Yu, Unusually Long Free Carrier
Lifetime and Metal-Insulator Band Offset in Vanadium Dioxide, Phys. Rev. B 85, 085111 (2012).
[216]E. Abreu, D. Babich, E. Janod, S. Houver, B. Corraze, L. Cario, and S. Johnson, THz Driven
Dynamics in Mott Insulator GaTa 4 Se 8, in 2019 44th International Conference on Infrared, Millimeter, and
Terahertz Waves (IRMMW-THz) (IEEE, Paris, France, 2019), pp. 1–2.
[217] S.-K. Mo, H.-D. Kim, J. Allen, G.-H. Gweon, J. Denlinger, J.-H. Park, A. Sekiyama, A.
Yamasaki, S. Suga, P. Metcalf, and K. Held, Filling of the Mott-Hubbard Gap in the High Temperature
Photoemission Spectrum of (V0.972Cr0.028)2O3, Phys. Rev. Lett. 93, 076404 (2004).
[218]H. Hirori, K. Shinokita, M. Shirai, S. Tani, Y. Kadoya, and K. Tanaka, Extraordinary Carrier
Multiplication Gated by a Picosecond Electric Field Pulse, Nat Commun 2, 594 (2011).
[219] T. Oka, Nonlinear Doublon Production in a Mott Insulator: Landau-Dykhne Method Applied to an
Integrable Model, Phys. Rev. B 86, 075148 (2012).
[220]I. Lo Vecchio, L. Baldassarre, F. D’Apuzzo, O. Limaj, D. Nicoletti, A. Perucchi, L. Fan, P.
Metcalf, M. Marsi, and S. Lupi, Optical Properties of V 2 O 3 in Its Whole Phase Diagram, Physical
Review B 91, (2015).
[221]B. Mansart, D. Boschetto, S. Sauvage, A. Rousse, and M. Marsi, Mott Transition in Cr-Doped
V2O3 Studied by Ultrafast Reflectivity: Electron Correlation Effects on the Transient Response, EPL
(Europhysics Letters) 92, 3 (2010).
[222]N. Kumar, A. Rúa, R. Díaz, I. Castillo, B. Ayala, S. Cita, F. Fernández, and S. Lysenko, Time-
Resolved Light Scattering by Photoexcited V 2 O 3, MRS Adv. 2, 1231 (2017).
[223] W. B. Yelon and J. E. Keem, The Elastic Constants of V2O3 in the Insulating Phase, Solid State
Communications 29, 775 (1979).
[224]Amano A., Orbital / spin degree of freedom is important in Ultrafast electron dynamics in
Mott insulator, Master thesis, Tohoku University, 2019.
[225]H. V. Keer, D. L. Dickerson, H. Kuwamoto, H. L. C. Barros, and J. M. Honig, Heat Capacity of
Pure and Doped V2O3 Single Crystals, Journal of Solid State Chemistry 19, 95 (1976).
[226] S. J. Byrnes, Multilayer Optical Calculations, ArXiv:1603.02720 [Physics] (2019).
[227]A. Volte, Photo-Induced Cooperativity in Bistable Volume-Changing Materials, PhD thesis,
Université de Rennes 1, 2020.
161
REFERENCES
[228] C. Loge, Simulation de la réflectivité de l’isolant de Mott V2O3 sous impulsion lumineuse
ultrarapide, Master thesis, École normale supérieure Paris-Saclay, 2020.
[229] H. Okamoto, Y. Ishige, S. Tanaka, H. Kishida, S. Iwai, and Y. Tokura, Photoinduced Phase
Transition in Tetrathiafulvalene- p -Chloranil Observed in Femtosecond Reflection Spectroscopy, Phys. Rev. B
70, 165202 (2004).
[230] C. Adda, Study, Fabrication and Characterization of « Leaky Integrate and Fire » Single
Component Artificial Neurons Based on Mott Insulators, PhD thesis, Université de Nantes,
2018.
[231]C. Vaju, L. Cario, B. Corraze, E. Janod, V. Dubost, T. Cren, D. Roditchev, D. Braithwaite,
and O. Chauvet, Electric-Pulse-Induced Resistive Switching and Possible Superconductivity in the Mott
Insulator GaTa4Se8, Microelectronic Engineering 85, 2430 (2008).
[232]P. D. Dernier, The Crystal Structure of V2O3 and (V0.962Cr0.0382)2O3 near the Metal-Insulator
Transition, Journal of Physics and Chemistry of Solids 31, 2569 (1970).
[233]H. Yang and R. J. Sladek, Lattice-Dynamical Model for the Elastic Constants and Raman Frequencies in
(V1-XCrx)2O3, Physical Review B 32, 6634 (1985).
[234] A. Okamoto, Y. Fujita, and C. Tatsuyama, Raman Study on the High Temperature Transition in
V2O3, Journal of the Physical Society of Japan 52, 312 (1983).
[235]J. Kieffer and D. Karkoulis, PyFAI, a Versatile Library for Azimuthal Regrouping, J. Phys.: Conf.
Ser. 425, 202012 (2013).
162
Titre : Couplage électron-réseau à la transition de Mott induite par impulsion électrique et/ou lumineuse
Mots clés : Isolant de Mott hors équilibre, transition de phase photoinduite, Transition résistive,
dynamique ultra-rapide, transition isolant – métal, couplage électron-réseau
Résumé : Les isolants de Mott forment une vaste Notre étude de l’isolant de Mott emblématique
classe de matériaux corrélés aux nombreuses V2O3: Cr clarifie la nature de la phase créée sous
propriétés fascinantes. Ces composés subissent champ électrique. Nous montrons en effet que les
divers types de transitions isolant-métal (IMT). La porteurs chauds générés à la TME induisent une
découverte récente d'un TIM pilotée en tension, compression locale du réseau selon des chemins
appelé Transition de Mott Electrique (EMT), a filamentaires. La persistance d'une phase
suscité beaucoup d'intérêt. Cette TME, qui peut compressée métastable explique la non-volatilité
être volatile ou non-volatile, ouvre la voie à des de l'EMT. D'autre part, notre expérience de
applications de rupture, comme les neurones pompe-sonde IR ultra-rapide indique que la
artificiels ou les mémoires de Mott. photoexcitation de V2O3: Cr lance une onde de
Une frontière actuelle de la recherche consiste déformation compressive conduisant à une IMT
donc à étudier et comprendre les TIM dans les macroscopique pilotée par un effet élastique.
isolants de Mott placés hors équilibre par des Enfin, l'application simultanée d'impulsions
stimuli externes tels que des impulsions de champ lumineuses et électriques dans l'isolant de Mott
électrique ou laser ultra-rapides. Dans ce contexte, GaTa4Se8 confirme que la TME est basée sur
ces travaux de thèse portent sur le mécanisme de l'accumulation de porteurs chauds, et permet
l'IMT pilotée par un champ électrique et / ou des ainsi la réalisation d'un neurone électro-optique
impulsions lumineuses dans les isolants de Mott. artificiel de Mott.
Title : Electron-lattice coupling at the Mott transition driven by electric and/or light pulse
Keywords : Mott Insulators Out of Equilibrium, Photoinduced Phase Transition, Resistive

Switching, Ultrafast dynamics, Insulator to metal transition, Electron-lattice coupling
Abstract : Mott insulators are a broad class of This work unravels the pivotal roles of hot carriers
strongly correlated materials with numerous and electron-lattice coupling at the IMT driven by
fascinating properties. These compounds undergo electric or light pulses. Our study of the iconic
various types of Insulator to metal transitions Mott system V2O3:Cr reveals the nature of the
(IMT). The recent discovery of a voltage-driven phase created under electric field. We show that
IMT, called Electric Mott Transition (EMT), has hot carriers generated at the EMT induces a
attracted much attention. This EMT can be volatile lattice compression in a filamentary path. The
or non-volatile. Both properties lead to persistence of a metastable compressed phase is
breakthrough applications, such as Mott artificial responsible for the EMT's non-volatility. On the
neurons or Mott memories. other hand, our ultrafast IR pump-probe
Therefore, a current frontier of research is to study experiment suggests that the photoexcitation of
and understand Insulator to metal transitions in V2O3:Cr launches a compressive strain wave
Mott insulators driven out-of-equilibrium by promoting an Insulator to metal phase
external stimuli such as a short electric field or transformation. Finally, the concomitant
ultrafast laser pulses. In this context, the Ph.D. application of light and electric pulses in the Mott
work focuses on the mechanism of the IMT driven insulator GaTa4Se8 confirms that the EMT is
by electric field and/or light pulses in Mott based on the accumulation of hot carriers, and
insulators. enables the implementation of an artificial electro-
optic Mott neuron.

BABICH

Uploaded by

Document Information

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

BABICH

Uploaded by

Copyright:

Available Formats

THESE DE DOCTORAT DE

ECOLE DOCTORALE N° 596

Rapporteurs avant soutenance :

Président : Hervé Cailleau Professeur émérite, IPR — CNRS — Université de Rennes 1

Chapter 1 INSULATOR TO METAL TRANSITION IN MOTT MATERIALS ................. 9

Chapter 2 NATURE OF METALLIC PHASES CREATED AT AND OUT-OF

Chapter 3 OUT-OF-EQUILIBRIUM MOTT INSULATOR TO METAL TRANSITIONS:

ANNEX 1 Sample preparation..................................................................................................................134

ANNEX 3 Raman measurements .............................................................................................................141

ANNEX 4 Complementary characterization of conductive filamentary path after non-volatile

ANNEX 5 Ultrafast measurements on (V1-xCrx)2O3 System ................................................................148

1.1 Basic physical concepts about correlated systems and Mott

1.1.1 Mott insulators: definition

1.1.1.1 The simplest case: the one -band Mott insulator

1.1.1.2 Multiband Mott insulators

N+1 site N-1 site N+1 site N-1 site

Excitation : energy gain (N ➔ N-1 site) 0 U’ - J U’ - J

Excitation : energy cost (N ➔ N+1 site) U U’ +U 2U’ - 2J

Gap atom (atomic limit) = Ecost - Egain atom = U atom = U + J atom= U – 3J

Mott Gap Mott (extended solid limit) Mott  U – W Mott  U + J – W Mott  U – 3J – W

1.1.2 Classical control parameters of Mott Insulator to Metal Transitions

1.2 Typical examples of 'canonical' Mott insulators: the AM 4 Q 8

1.2.1 The vanadium oxide (V1-xCrx)2O3

dV-V c/6 PM ambient 0 15.6 small 2.697 strong Small

PI ambient 4 18.4 ** strong 2.747 ** small Large

a1g* (antibonding) V2O3 (PM) (V1-xCrx)2O3 (PI)

insulator with N = 2 electrons spread over M = 3 orbitals (see Figure 1.11).

1.2.2 AM4Q8 compounds

1.3 Mott transitions under Electric field

1.3.1 Trivial types of transitions (Joule heating, Electromigration)

1.3.1.1 Joule heating and ther mally driven IMT

1.3.1.2 Electromig ration and filling controlled IMT

In oxide Mott insulators, a filling-controlled Insulator to metal transition is easily achieved

1.3.2 Electric Mott transition (dielectric breakdown phenomena)

1.3.2.1 Phenomenolog y of the dielectric breakdown

As a consequence, the Mott insulator compounds exhibit a particular current-voltage

1.3.2.2 Microscopic behavior : the creation of hot electrons and

1.3.2.3 Macroscopic phenomena: the creation of a conducting

1.3.2.4 Open question about the nature of the metallic

The presence of granular superconductivity at low temperature in a transited crystal of

1.3.3 Microelectronics applications of Electric Mott transition

1.3.3.1 Mott Random Access Memories

1.3.3.2 Neuromor phic applications : Mott neuron

30s 15s 20s 30s

time (s) time (s)

1.4 Insulator to metal transition driven by light in correlated

1.4.1 Generality about the photoinduced phase transition explored by time-

1.4.1.1 Pump-probe technique

1.4.1.2 Photoinduced phase transitions

1.4.1.2.1 Thermal mechanism

In parallel to classical photothermal mechanisms discussed in the previous section, several

(a) stress(z,t) / stress(z=0, t=0) (e) Strain

t=0 pump pump

0 Distance from surface 0 Distance from surface

stress(z,t) / stress(z=0, t=0) Strain

(c) stress(z,t) / stress(z=0, t=0) (g)

-0.5 Strain wave propagation

stress(z,t) / stress(z=0, t=0) Strain

0 Distance from surface 0 Distance from surface

Electronic energy modification

Acoustic phonon creation (or annihilation)

Compression (for example)

In a more chemistry-oriented description, the deformation potential mechanism can be

𝜕𝐸𝐺 𝜕𝐸𝐺 (1.9)

PI ambient 4 18.4 strong 2.747 small Large