S T U D Y A N D C H A R A C T E R I Z AT I O N O F M A G N E T I C

A N D M U LT I F E R R O I C M AT E R I A L S B Y
F I R S T - P R I N C I P L E S C A L C U L AT I O N S

andrés camilo garcía castro

—-
Faculté des Sciences
Université de Liège
&
Centro de Investigación y de Estudios Avanzados IPN

Abril ����
S T U D Y A N D C H A R A C T E R I Z AT I O N O F M A G N E T I C
A N D M U LT I F E R R O I C M AT E R I A L S B Y
F I R S T - P R I N C I P L E S C A L C U L AT I O N S

andrés camilo garcía castro

Supervisors:
Prof. Aldo Humberto Romero Castro, CINVESTAV-Querétaro
Dr. Eric Bousquet, Université de Liège (Secretary)
Jury Members:
Prof. Nicola Spaldin, ETH-Zürich
Prof. Philippe Ghosez, Université de Liège (President)
Prof. Sergio Jiménez Sandoval, CINVESTAV-Querétaro
Prof. Rebeca Castanedo Pérez, CINVESTAV-Querétaro
Prof. Arturo Mendoza Galván, CINVESTAV-Querétaro

—-
Ph.D-Candidate
Physics Department
Université de Liège
&
Centro de Investigación y de Estudios Avanzados IPN

Abril ����
Andrés Camilo García Castro: Study and Characterization of Magnetic and Mul-
tiferroic Materials by First-principles Calculations, —-, c Abril ����
Dedicated to my father and my grandfather memory that always encouraged me
to follow my dreams and be consistent with it.

Dedicada a mi padre y a la memoria de mi abuelo que siempre me motivaron a

seguir mis sueños y actuar consistentemente con ellos.
ABSTRACT

In the last fifteen years, multifunctional materials, and more specifically, multi-
functional oxides have been widely studied due to its wide range of properties.
Properties that go from superconductivity to ferroelectricity passing through mag-
netism and multiferroism have been reported. Nonetheless, the fluoride family was
left aside and little information is known about its possible ferroelectricity or mul-
tiferroism. In this Ph.D thesis, we explored the electronic, vibrational, structural
and magnetic properties of fluoride perovskite-based compounds. To such pur-
poses, We performed ab-initio calculations based in the density-functional theory
(DFT) as implemented in VASP and CRYSTAL codes.

Our first step was to perform vibrational analyses in a large set of fluoroper-
ovskites ABF3 . Based on the results, we proposed a model that establishes an
A-site geometrically driven ferroelectric vibrational instability in fluorides. Our
studies reveal a different behavior as a function of isotropic pressure for NaBF3
with respect to oxides (e.g. BaTiO3 ) with B = Ca, V, Mn, and Zn. For these
compounds we found an increase of the ferroelectric instability as a function of
hydrostatic pressure. This probably due to the “transformation” of eigendisplace-
ments responsible for the mode that creates the corresponding instability. In
particular, an increase of ionic A-site radii present a strong influence in FE-polar
instability.

We also have shown, based on our first-principles calculations and symmetry

theory analysis that all post-perovskites ABX3 with an active magnetic B-site
cation can exhibit a noncollinear magnetic configuration, which happens to be
allowed by symmetry. With these findings, the magnetic properties found exper-
imentally were clarified for this particular high-pressure phase perovskite found
at the Earth’s mantle. Additionally, We have predicted that NaMnF3 suffers a
structural phase transition under pressure to a post-perovskite phase, where non-
collinear ferromagnetism and large magnetic moment components are obtained
within this high-pressure phase..

Going beyond, We have shown that it is possible to achieve multiferroic-induced

state in NaMnF3 under epitaxial strain at compressive or tensile strain. We found
a nonlinear behavior of the ferroelectric instability as well as a non-linear piezo-
electric response as a function of epitaxial strain. The later completely different
as the one found in oxide perovskites. Similarly, an out-of-plane polarization was
observed, a property that has not been observed in oxides. We observed a Na
+ Mn sites cooperative ferroelectric ordering for compressive strain against a
pure A-site geometrically driven ferroelectricity at tensile values of the ac-strain.

vii
Magnetic ordering reveals a non-collinear ground state with the Gz Ax Fy repre-
sentation. Even more interesting, and non-linear magnetoelectric coupling was
found under the strained Pna�1 ground state becoming the first known multifer-
roic/magnetoelectric perovskite fluoride.

Later, in order to go further, We studied the electronic and structural proper-

ties of novel heterostructures based on oxyfluorides (KTaO3 )n /(KBF3 )l B = Zn
and Ni interfaces. We found that the orbital levels splitting at the interfaces is
strongly modified by the O–B–F coordination. The polar catastrophe phenomena
also takes place in the oxyfluoride interfaces similarly to oxide heterostructure,
however, we found that less number of layers are needed in order to achieve
the insulator-to-metal transition when comparing to SrTiO3 /LaAlO3 superlat-
tices. We observed that the magnetism in the KTaO3 /KNiF3 exhibits a moment
magnitude modulations. Nevertheless, the magnetic structure keeps the G-type
antiferromagnetism such as in the bulk former compound. Surprisingly, we ob-
served a large k3 -Rashba type splitting in at the oxyfluoride interfaces, at least
four times larger than the one reported in SrTiO3 /LaAlO3 interface and twice
of the KTaO3 -based transistor.

In conclusion, we observed that fluorides-perovskites are good prototypes for

multifunctional properties as oxides. Therefore, based on the results reported in
this thesis, we expect that experimentalist and theoreticians can be motivated in
characterization of fluorides, which can lead to a new set of unexplored materials
with potential novel applications in electronics.

viii
RÉSUMÉ

Durant les quinze dernières années, les matériaux multifonctionnels et plus par-
ticulièrement les matériaux multifonctionnels à base d’oxyde ont été extrême-
ment étudiés du fait de leurs propriétés très diverses. Ces propriétés vont de la
supraconductivité à la ferroélectricité en passant par le magnétisme ou le multi-
ferroïsme. Dans cet engouement pour les oxydes, la famille des fluorites a cepen-
dant été délaissée de telle manière que peu d’informations sont connues quant
à leur propension à être ferroélectrique ou multiférroïque. Dans cette thèse de
doctorat, nous nous sommes proposés d’explorer les propriétés électroniques, vi-
brationnelles, structurales et magnétiques des matériaux perovskites à base de
fluor. Cette étude repose principalement sur des calculs ab-initio basés sur la
théorie de la fonctionnelle de la densité. Les codes de calcul utilisés sont princi-
palement VASP et CRYSTAL.

Dans un premier temps nous avons analysé les propriétés vibrationnelles d’une
large gamme de fluoro-perovskites ABF3 . Cette première étude nous a permis
de montrer que ces cristaux présentent une instabilité ferroélectrique dans leur
phase cubique qui est d’origine géométrique et qui est dominée par le cation A.
Cette origine géométrique rend la dépendance de l’instabilité ferroélectrique à
une pression hydrostatique très différente de celle observée dans les oxydes. Dans
les fluorites la pression a tendance à augmenter l’instabilité ferroélectrique tandis
que dans les oxydes elle a tendance à la supprimer.

En nous basant sur une analyse des symétries et sur des calculs ab-initio, nous
avons aussi montré que les phases perovskites Pnma, observées à basse pression,
et les phases post-pérovskites, observées à haute pression, contenant un cation
B magnétique ont toutes une structure magnétique non-colinéaire. Cette étude a
permis de clarifier les propriétés magnétiques observées expérimentalement et de
prédire une transition Pnma/post-perovskite dans le NaMnF3 sous haute pres-
sion avec un ferromagnétisme léger bien plus important que dans tous les autres
systèmes reportés jusqu’à présent.

Au delà d’une simple caractérisation de matériaux, nous avons aussi prédit la

possibilité d’induire une phase multiferroïque dans le NaMnF3 en jouant avec
des contraintes d’épitaxie. Contrairement à ce qui est observé dans les oxydes, la
polarisation est induite dans la direction perpendiculaire à la contrainte bi-axiale
et le couplage entre la polarisation et la contrainte est dominé par une réponse
piézoélectrique non-linéaire. Cette dominante non-linéaire permet d’obtenir une
réponse originale de la polarisation où cette dernière est augmentée pour des
contraintes d’épitaxie aussi bien positives que négatives. Nous expliquons cette

ix
réponse inhabituelle à une modification du vecteur de déplacement ferroélectrique
des atomes où sous tension la polarisation est dominée par un déplacement
géométrique du cation A et sous compression par le déplacement des cations
A et Mn. L’analyse du magnétisme montre un ordre antiferromagnétique domi-
nant de type G associé à des ordres non-colinéaires de type A et F (Gz Ax Fy ).
L’association de l’ordre ferroélectrique et du ferromagnétisme léger (Fy ) fait que
NaMnF3 sous contrainte d’épitaxie est un nouveau multiferroïque. L’étude de
la réponse magnétoélectrique sous champ magnétique montre qu’elle s’effectue
également à l’ordre non-linéaire.

Enfin, dans un dernier temps nous avons aussi étudié la structure électronique
et les propriétés structurales d’heterostructures oxyfluorites (KTaO3 )n /(KBF3 )l
où B = Zn et Ni. Dans ces interfaces entre un oxyde et une fluorite nous avons
observé que la levée de dégénérescence des orbitales atomiques est fortement mod-
ifiée par la liaison O–B–F. Un phénomène de catastrophe polaire est aussi observé
où l”epaisseur critique pour l’apparition d’un gaz bidimensionnel est plus petite
que celle reportée dans les super-réseaux SrTiO3 /LaAlO3 . Dans le cas B = Ni, le
magnétisme dans la couche de KNiF3 reste globalement du type G mais présente
une modulation en amplitude des moments magnétiques près de l’interface. De
manière surprenante, nous avons aussi observé une levée de dégénérescence de
type k3 -Rashba à l’interface de ces heterostructures oxyfluorites et d’amplitude
au moins quatre fois supérieure à celle rapportée dans l’interface SrTiO3 /LaAlO3
et deux fois supérieur à celle des transistors à base de KTaO3 .

En résumé, notre étude a permis de conclure que les cristaux perovskites à base
de fluorites sont de bons candidats à considérer dans la recherche de propriétés
multifonctionnelles. Nous espérons que nos résultats motiveront des expérimenta-
teurs et d’autres théoriciens dans l’étude et la caractérisation des fluorites dans
la recherche de nouveaux matériaux aux propriétés inédites et à haut potentiel
pour des applications technologiques et industrielles.

x
RESÚMEN

En los últimos quince años, los materiales multifuncionales, y más específicamente,

los óxidos multifuncionales han sido ampliamente estudiados debido a su basta
gama de propiedades. Se han reportado propiedades que van desde la supercon-
ductividad y ferroelectricidad pasando por el magnetismo y multiferroismo. Sin
embargo, la familia de fluoruros perovskita a sido dejada de lado y poca informa-
ción se conoce sobre su posible ferroelectricidad y/o multiferroismo. En mi tesis
doctoral, he explorado las propiedades electrónicas, vibracionales, estructurales y
magnéticas de los compuestos base fluoruro perovskita. Para tal fin, he realizado
ab-initio cálculos basados en la teoría del funcional de la densidad (DFT) tal
como esta implementada en los códigos VASP y CRYSTAL.

Los análisis vibracionales se llevaron a cabo en fluoroperovskitas tipo ABF3 .

En base a los resultados, hemos propuesto un modelo que muestra la presencia de
una inestabilidad ferroeléctrica geométricamente dominada por el sitio A en estos
fluoruros. Mis estudios revelan un comportamiento diferente para los compuestos
NaBF3 con B = Ca, V, Mn y Zn como función de la presión isotrópica, esto
en comparación con los óxidos (por ejemplo BaTiO3 ). Para estos compuestos se
encontró un aumento de la inestabilidad ferroeléctrico como función de la tensión
hidrostática. Lo anterior posiblemente debido a la “transformación” observada de
eigendisplacements responsables de este modo. Así mismo,El incremento del ra-
dio iónicos del sitio A presentan una fuerte influencia en la inestabilidad FE-polar.

También se ha demostrado, que todos las post-perovskitas ABX3 con una sitio
B magnéticamente activo exhiben la posibilidad de presentar un comportamiento
magnético no colineal permitido por la simetría. Lo anterior con base en los cálcu-
los de primeros principios y el análisis de teoría de grupos. Con estos resultados
fueron aclaradas las propiedades magnéticas encontrados experimentalmente en
esta fase perovksita de alta presión encontrada principalmente en el manto de la
Tierra. Además, fue posible predecir la existencia de esta transición a fase post-
perovskita en el compuesto NaMnF3 . Con esto, este fluoruro se convertirse en un
nuevo miembro de esta familia en el cual, las propiedades magnéticas, y sobre
todo, su ferromagnetismo no colineal muestra un alta componente de momento
magnético.

Yendo más allá, se ha demostrado que es posible alcanzar el estado multiferroico

en NaMnF3 inducido bajo tensión epitaxial para esfuerzos de compresión o expan-
sión. Se encontró un comportamiento no lineal de la inestabilidad ferroeléctrico
seguida por una respuesta piezoeléctrica no lineal como una función de la ten-
sión epitaxial. Este último completamente diferente a los hallazgos en óxidos tipo

xi
perovskita. Además, se observó una polarización fuera de plano en la tendencia
opuesta a lo encontrado en óxidos. Se encontró ademas un ordenamiento ferroeléc-
trico cooperativo entre los sitios de Na y Mn para el esfuerzo de compresión. Lo
anterior en contraste con una ferroelectricidad geométricamente dominada por el
sitio A a valores de tensión en el plano ortorrómbico xz. El ordenamiento mag-
nético revela un estado fundamental no colineal bajo la representación Gz Ax Fy .
Aún más interesante, se observó un acople magnetoeléctrico no lineal en este ma-
terial bajo la simetría Pna�1 de estado fundamental, convirtiéndose con esto en
el primer multiferroico/magnetoeléctrico conocido en estructura perovskita base
fluoruro.

Con el fin de ir más lejos aún, se estudiaron las propiedades electrónicas y

estructurales de nuevas heteroestructuras basadas en interfaces de oxifluoruros
tipo (KTaO3 )n /(KBF3 )l con B = Zn y Ni. Se encontró que la simetría de los
niveles orbitales en las interfaces está fuertemente modificado por la coordinación
del enlace O–B–F. El fenómeno conocido como “polar-catastrophe” también tiene
lugar en las interfaces de oxifluoruro, sin embargo, se encontró que se necesita
menos cantidad de capas con el fin de lograr la transición de aislante a metal
cuando se compara con superredes como SrTiO3 /LaAlO3 . Hemos observado que
el magnetismo en el sistema KTaO3 /KNiF3 exhibe modulaciones el la magni-
tud de momento magnético. Sin embargo, la estructura magnética mantiene el
ordenamiento G-antiferromagnetico tal y como en el compuesto en bulto. Sorpren-
dentemente, se encontró en las interfaces tipo oxifluoruro un gran desdoblamiento
de espín tipo k3 -Rashba, el cual es al menos cuatro veces mayor que el reportada
en la interfaz SrTiO3 /LaAlO3 y el doble del encontrando en el transistor basado
en KTaO3 .

En conclusión, se observó que los fluoruros-perovskita son buenos materiales

prototipos para propiedades multifuncionales tal y como los óxidos. Por ende, es-
peramos que nuestros resultados puedan motivar a experimentales y teóricos en la
búsqueda de las propiedades de estos sistemas, que son, a nuestro entender, inex-
plorados y representan una gran promesa para el descubrimiento de propiedades
interesantes.

xii
P U B L I C AT I O N S

Some ideas and figures have appeared previously in the following publications:

[�] A. C. Garcia-Castro, A. H. Romero, E. Bousquet, Strain-engineered

multiferroicity in Pnma NaMnF3 fluoroperovskite, Physical Review Letters, ���,
������ (����).

[�] A. C. Garcia-Castro, Nicola A. Spaldin, A. H. Romero, E. Bousquet, Ge-

ometric ferroelectricity in fluoroperovskites, Physical Review B. ��, ������ (����)
editor’s suggestion.

[�] A. C. Garcia-Castro, A. H. Romero E. Bousquet, First-principles study

of vibrational and noncollinear magnetic properties of the perovskite to postper-
ovskite pressure transition of NaMnF3 , Physical Review B. ��, ������ (����).

[�] A. C. Garcia-Castro, A. H. Romero, E. Bousquet, Non-collinear Mag-

netism in Post-perovskites from First-principles: comparing CaRhO3 and NaNiF3 ,
Physica Status Solidi B. ���, ��� (����) Front cover.

[�] A. C. Garcia-Castro, M. Vergniory, E. Bousquet, A. H. Romero, Spin-

texture induced by oxygen vacancies is strontium perovskites surfaces: A theoreti-
cal comparison between SrTiO3 and SrHfO3 (���), Physical Review B. ��, ������
(����).

[�] P. Borisov, Trent A. Johnson, A. C. Garcia-Castro, K. Amit, D. Schre-

congost, C. Cen, A. H. Romero, D. Lederman, Multiferroic BaCoF4 in thin film
form: ferroelectricity, magnetic ordering, and strain, ACS Applied Materials and
Interfaces, �, ���� (����).

[�] Weitao Dai, Sanjay Adhikari, A. C. Garcia-Castro, Aldo H. Romero,

Hyungwoo Lee, Jung-Woo Lee, Sangwoo Ryu, Chang-Beom Eom, Cheng Cen,
Tailoring LaAlO3 /SrTiO3 interface metallicity by oxygen surface adsorbates, Nano
Letters, ��, ���� (����).

[�] Sanjay Adhikari, A. C. Garcia-Castro, Aldo H. Romero, Sangwoo Park,

Chang-Beom Eom, Cheng Cen, The importance of water adsorption in shaping
LaAlO3 /SrTiO3 interface properties, Advanced Functional Materials. (���� - Un-
der consideration).

xiii
 [�] Sobhit Singh, A. C. Garcia-Castro, Irais Valencia-Jaime, Francisco Munoz,
and Aldo H. Romero, Prediction of a controllable Weyl semi-metallic phase in
inversion-asymmetryc BiSb, Advanced Materials, (���� - To be submitted).

[��] P. Borisov, Trent A. Johnson, W. Ibarra-Hernandez, A. C. Garcia-Castro,

E. Bousquet, K. Amit, D. Schrecongost, C. Cen, A. H. Romero, D. Lederman,
Muliferroic behavior of NaMnF3 thin films growth by molecular beam epitaxy,
Journal of Applied Physics. (���� - To be submitted).

xiv
“Because success is stumbling from failure to failure with no loss of enthusiasm”
—Wiston S. Churchill.

ACKNOWLEDGEMENTS

Sometimes, you face the difficult decision to leave your country and your home in
the search for better opportunities and to explore new horizons. This path that
I traced in my life � years ago now is rewarded with this amazing achievement
not just for my professional life, but also for my personal formation as well. Thus,
the culmination of the last � year of this journey, full of nice and not that good
experiences, is condensed in the present manuscript. Therefore, I’m taking this
chance, and a few words, to acknowledge all this people that were and have been
there for me and support me along this journey, because, I have always believed
that when you achieve something worthy in your life, this is the sum of all the
efforts of the people around you.

First, I would like to thank to prof. Aldo H. Romero for his great supervision
and friendship that always extended to me kindly, and for bring me the oppor-
tunity to join his research group. I also want to thank him for encouraging and
support my scientific travels used to interact with the scientific community.

I would also equally thank to Dr. Eric Bousquet for all his help, guidance and
friendly patience to explain me a lot of scientific aspects, but more important
than that, for being there to help me even in small things such as finding an
apartment to rent.

I am very grateful to my fiancée, Sara Orozco, who was always there for me
listening my scientific explanations and ideas, even knowing that she is not a
scientist. Who has been there for me and had understood when I have to travel
for a long time and leave her alone.

Especial thanks to prof. Beatriz Cruz in Colombia that introduced me in the

research world and in the meanwhile bring me her unconditional support which
make me take the change lo look beyond the colombian horizons

I would like to thanks to all my colleagues at the department of physics at

the Université de Liège: Henu, Julien, Nicholas, Antoine, Dr. Eric, prof. Philippe,
prof. Matthieu, Dra. Zeila, Begum, Denis, Safari, Naihua, Marco, Hanen, Sebas-
tian, Alain, Brahim, and Fabio.

To the Euraxess fellowship and its stuff for their friendly help and support that
allowed me to be in Liège for more than a year and enjoy the delicious chocolates,

xv
nice beers and tasty food.

I thank to prof. Nicola A. Spaldin and her research group at the ETH-Zürich
for the chance to work with them and have the opportunity to be in the beautiful
Zürich for several months.

To CONACyT for the invaluable opportunity that bring me to México as a

starting point of my professional career that opened the world for me.

To my friend and colleague Irais Valencia for her support and patience to hear
my ideas even when they did not take me to a successful results.

It is worth to mention that all the computational work presented here was pos-
sible thanks to support —for more than � millions of computational hours— from
the Extreme Science and Engineering Discovery Environment, XSEDE at United
States with the supercomputers Stampede and Ranger (project TG-DMR������).
Thanks also to the computational time provided by the Belgian CECI projects
and specially the Zenobe stuff.

At last, but not least, I would like to thanks to my parents, Alberto and Martha,
for teaching me the value of the honesty, hard work, and respect. Thanks to my
sister Cindy Johana, that is constantly in touch with me giving me the courage
to go forward. I thank to my friends Luis Alberto, Sebastian, and Catalina that
were always worried about my thesis progress and were continuously asking: “how
is the thesis going?”. Finally and specially, thanks to Wilfredo and his wife Rosa
Elisa, that opened their home for me and were a friendly face every time that I
traveled from México to Belgium.

xvi
CONTENTS
Acronyms xx
Introduction xxi

i density functional theory and electronic structure

�
� theoretical background �
�.� Many-Body Problem . . . . . . . . . . . . . . . . . . . . . . . . . . �
�.� Density-Functional Theory . . . . . . . . . . . . . . . . . . . . . . �
�.�.� The Hohenberg-Kohn Theorems . . . . . . . . . . . . . . . �
�.�.� Kohn-Sham Equation . . . . . . . . . . . . . . . . . . . . . �
�.� Exchange–Correlation Energy and Functionals . . . . . . . . . . . �
�.�.� LDA and GGA . . . . . . . . . . . . . . . . . . . . . . . . . �
�.�.� DFT+(U,J) . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.�.� Hybrid Functionals . . . . . . . . . . . . . . . . . . . . . . ��
�.� Density Functional Perturbation Theory . . . . . . . . . . . . . . ��
�.� From the theory to the implementation . . . . . . . . . . . . . . . ��
�.�.� Plane-waves expansions . . . . . . . . . . . . . . . . . . . . ��
�.�.� Periodic Boundary Conditions . . . . . . . . . . . . . . . . ��
�.�.� Pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . ��
�.�.� Projector Augmented-Waves (PAW) method . . . . . . . . ��
�.�.� Zeeman applied magnetic field . . . . . . . . . . . . . . . . ��
�.�.� Used codes: VASP and CRYSTAL . . . . . . . . . . . . . . ��

ii structural properties of perovskites abf 3 ��

� vibrational behavior in fluoroperovskites abf 3 ��
�.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Ionic A - and B -site Radii Size Effects . . . . . . . . . . . . . . . . ��
�.� Ferroelectricity Model Proposed for ABF 3 Fluorides . . . . . . . . ��
�.� Ferroelectric Instability Enhancement in NaBF 3 Fluorides . . . . ��
�.�.� Modes Contribution in Pm 3̄ m to Pnma Transition in NaBF3
Fluorides . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.�.� Special cases: LiNiF 3 and NaNiF 3 . . . . . . . . . . . . . . ��
�.� Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��

iii strain and pressure effect in fluoroperovskites ��

� prediction of the p p v transition in n a m n f 3 ��
�.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Structural and vibrational properties of the Pnma phase of NaMnF 3 ��
�.� Study of NaMnF 3 under hydrostatic pressure . . . . . . . . . . . ��
�.� Magnetic properties of the pPv phase . . . . . . . . . . . . . . . . ��

xvii
xviii contents

�.� Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
� non-collinear magnetism in c a r h o 3 and n a n i f 3 post-
perovskites ��
�.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Structural Characterization of CaRhO 3 and NaNiF 3 . . . . . . ��
�.� Non-collinear Ordering . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Exploring the Magnetic Interactions . . . . . . . . . . . . . . . . . ��
�.� Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . ��
� induced multiferroic and magnetoelectric behav-
ior in n a m n f 3 ��
�.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Epitaxial strain in Pnma NaBF 3 Fluorides . . . . . . . . . . . . . ��
�.� Strain-Induced Ferroelectricity in NaMnF 3 . . . . . . . . . . . . ��
�.�.� Second-order piezoelectricity in Pna�1 NaMnF3 . . . . . . ��
�.� Non-Collinear magnetism and ME coupling in Pna�1 NaMnF 3 . ��
�.�.� Field-induced distortions as the source of the ME coupling ��
�.�.� Effect of the U and J into the ME coupling in Pna�1 NaMnF3 ��
�.�.� Special remark: some ferroelectrics under strain . . . . . . ��
�.� Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��

iv oxyfluorides abo 3 /abf 3 interfaces ��

� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3
interfaces ��
�.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ��
�.� Oxyfluorides KNiF 2.5 O 0.5 and KT aO 2.5 F 0.5 . . . . . . . . . . . ��
�.� KT aO 3 / KBF 3 ( B = Zn and Ni ) superlattices . . . . . . . . . . . ���
�.�.� �DEG and �DHG in KTaO3 /KBF3 superlattices: . . . . . ���
�.�.� Magnetic ordering in KTaO3 /KNiF3 superlattice: . . . . . ���
�.�.� Isolated p- and n- interfaces: . . . . . . . . . . . . . . . . . ���
�.� Cubic-Rashba spin-splitting at oxyfluoride interface . . . . . . . . ���
�.� Charge-ordering, magnetism, and ferroelectricity in LaScO 3 / NaMnF 3
superlattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ���
�.� Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ���
� conclusions and perspectives ���

v appendix ���
a overview of the magnetic interactions in fluoroper-
ovskites ���
b group theory analysis to the non-collinear mag-
netism in ppv ���
c b a c o f 4 thin films ���
c.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ���
c.� Strain Effect in B a C o F4 Fluoride: . . . . . . . . . . . . . . . . . . ���
c.� Noncollinear Magnetism and Multiferroic Behavior . . . . . . . . . ���
 contents xix

c.� Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ���

d spin-texture in s r bo 3 (b) = t i and h f (���) surfaces ���
d.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ���
d.� ST O : T iO 2 terminated surface . . . . . . . . . . . . . . . . . . . ���
d.� ST O : T iO 2- surface with oxygen vacancies . . . . . . . . . . . ���
d.� Pbnm SrHfO 3 : HfO 2 surface with oxygen vacancies . . . . . . . ���
d.� Summary and Discussions . . . . . . . . . . . . . . . . . . . . . . . ���
e role of surface water adsorptions in s r t i o 3 /l a a l o 3
interface ���
e.� Computational Details . . . . . . . . . . . . . . . . . . . . . . . . . ���
e.� Water dissociation in the surface . . . . . . . . . . . . . . . . . . . ���
List of Figures ���
List of Tables ���

references ���
ACRONYMS

AFD Antiferrodistortive
AFM Antiferromagnetism
A-AFM A-type Antiferromagnetism
BZ Brillouin Zone
C-AFM C-type Antiferromagnetism
DFPT Density-Functional Perturbation Theory
DFT Density-Functional Theory
FE Ferroelectric
FM Ferromagnetism
G-AFM G-type Antiferromagnetism
GGA Generalized Gradient Approximation
HSE�� Heyd-Scuseria-Ernzerhof
IMT Insulator-to-Metal Transition
LDA Local Density Approximation
LDOS Local-Density Of States
LSDA Local Spin Density Approximation
ME Magnetoelectric
MF Multiferroic
PAW Projected Augmented-Wave
PBE Perdew Burke Erznzerhof
PBEsol Perdew Burke Erznzerhof Revised for Solids
Pv Perovskite
pPv Post-perovskite
SOC Spin-Orbit Coupling
VASP Vienna Ab-initio Simulation Package

xx
INTRODUCTION

Since its discovery in BaTiO3 , ferroelectricity in perovskite oxides has attracted

tremendous interest, ranging from fundamental studies to technological applica-
tions [��]. Indeed, the transition-metal/oxygen bond, with its large polarizability,
is particularly favourable for promoting the transition-metal off-centering that can
result in a ferroelectric ground state [��, ��]. Ferroelectrics also exist, of course, in
many material chemistries that do not contain oxygen, with a particularly exten-
sive range of fluorine-based examples, including both polymers [��] and ceramics
in many crystal classes (for a review see Ref. [��]). Perhaps not surprisingly given
the low polarizability of bonds with fluorine, the mechanism for ferroelectricity in
fluorine-based ferroelectrics are distinct from that in oxides, ranging from molec-
ular reorientation in polymers [��], to geometric reconstructions in ceramics [��].
These alternative mechanisms are of particular interest because, again unlike the
oxides, they are not contra-indicated by transition metal d-electrons, and so al-
low simultaneous ferroelectricity and magnetic ordering (multiferroism) as we are
going to discuss in this thesis..

Fluoride materials tend to exhibit multiferroism —defined by the coexistence of

ferroic order parameters— and magnetoelectricity —known as the induced mag-
netic (electric) polarization by means of an applied external electric (magnetic)
field— as mentioned in the work by Scott and Blinc [��]. However, there is no
mention of such behavior in ternary perovskite fluorides and even when a few of
them have been identified as possible ferroelectrics —with a R�c polar symmetry
group—. Therefore, there is no mention in the literature related to this topic
of a single multiferroic fluoroperovskite as possible multiferroic nor magnetoelec-
tric, while the perovskite family is an extremely important class of materials for
multifunctional applications due to their simple cubic block structure easily man-
ageable for growing in thin films, superlattices, etc [��, ��, ��, ��, ��].

In the field of multiferroics and magnetoelectrics, there are still some open ques-
tions and challenges to be addressed that concerns the achievement of a room
temperature multiferroic or a magnetic-field switchable polarization. The break-
through results we present in this thesis stand for the realization of a strain engi-
neering multiferroism/magnetoelectric in a class of materials that nobody has an-
ticipated. Moreover, in the community of ferroelectrics no one would have looked
at ionic compounds with very low polarizability. Besides, due to the low polariz-
ability and large electronegativity of the fluorine atom, novel and unexpected non-
linear strain/polarization properties can be found, which to our knowledge, has
never been reported before in any ferroelectric perovskite-based systems. Thus,
the unexplored behavior in perovskite fluorides is very appealing at the funda-

xxi
xxii introduction

mental level —that it has never been considered— and at the experimental level
—that it has never been observed— and it opens totally new and unexpected
perspectives in the hot field of ferroelectrics and magnetoelectrics, which is in a
constant search for new materials that can be addressed experimentally and drive
novel responses and applications.

From the industrial point of view, fluorides have proven to be of high inter-
est for numerous long term applications, such as fluorides-based glasses with a
large thermal expansion, low refractive and non-linear index [��, ��], strong mag-
nets with an optically transparency in the visible light [��, ��], electrochemical
devices, solid-state batteries, gas sensors, and electrochromic systems [��] or cata-
lyst surfaces based on metal fluorides such as AlF3 [��]. Thus, adding multiferroic
properties in fluoroperovskites would open an exciting opportunity to their use
in novel and extended applications.

Therefore, in this thesis we propose to analyze and understand the microscopic

origin of the fluoroperovskites through first-principles simulations and to high-
light new and unexpected multiferroic properties to offer some understanding on
the properties of fluoride perovskites. We analyze their vibrational behavior, fer-
roelectricity, magnetic canting and non-collinear magnetism, pressure and strain
effects, mixed interfaces and electronic structure. These results are organized as
follows:

An introduction to DFT is presented in Chapter � where it is presented a

basic theoretical background focused around the many-body problem, density-
functional theory as well as its fundaments and principal implementation consid-
erations.

The vibrational and structural properties of ABF3 will be addressed in the

Chapter �. This analysis will give us the fundamental background to the under-
standing of the hidden properties that lie beneath the fluoroperovskites.

Taking advantage of the explored properties of fluorides, we analyze the hy-

drostatic pressure effects on the structural and magnetic properties of NaMnF3
where we predict a transition toward a post-perovskite phase at relatively low
pressure —in Chapter �—. We also present in this chapter the symmetry analysis
of the possible non-collinear magnetic orders we can found in the post-perovskite
space group and compare it with our calculations and other pPv systems such as
CaRhO3 and NaNiF3 —at Chapter �—.

Moving forward, in Chapter � we present the study of the epitaxial strain-

engineered multiferroic/magnetoelectric state in NaMnF3 where a deep discus-
sion on the non-collinear magnetism, ferroelectricity, magnetoelectric-coupling
 introduction xxiii

and non-linear piezoelectric effect is carried out.

Finally, the electronic structure at oxide/fluoride perovskite (oxyfluoride) inter-

faces are analyzed in Chapter �. Surprisingly, the results suggest a larger Rashba
effect at these interfaces, in comparison to oxide/oxide interfaces, calling for a
deeper understanding of this mixed type of heterostructures.

At the end of this document several appendixes are presented. The magnetic
properties —exchange constants, group theoretical details, etc— of fluoride per-
ovskites and post-perovskites is explained in Appendixes A and B. We also present
in Appendices C, D, and E additional side studies we performed during our thesis
such as strained BaCoF4 thin films, the role of water adsorptions in STO/LAO
interfaces and the spin texture at the SrBO3 (B = Ti and Hf) surfaces in the
presence of an oxygen vacancy respectively.
 Part I

D E N S I T Y F U N C T I O N A L T H E O RY A N D
ELECTRONIC STRUCTURE

Density-Functional Theory is nowadays, one of the most successful

theoretical developments for electronic structure calculations in solid-
state physics. Within this theory, the approximations used have shown
to provide an excellent balance between computational accuracy and
cost ideal for the study of the physics behind many materials. In
next section is described the theoretical foundations for the research
and the developed work presented in this thesis. Then, a short but
condensed review of the theory, implementation, and computational
methods is described in order to give to the reader enough background
to address the thesis content.
 THEORETICAL BACKGROUND
1
In condensed matter physics, several theoretical methods have been developed
and consequently used to explain various phenomena and properties in materi-
als. Within these methods, first-principles approaches and specifically, density-
functional theory (DFT) have gained a lot of attention and recognition. DFT
has been applied successfully to a wide range of materials providing a deeper
understanding of their properties and microscopic origins. All the research shown
in the development of this PhD thesis was performed within the framework of
DFT. Then, in this chapter, the reader can find a brief, but focused description of
the theoretical and mathematical background needed in the development of the
results condensed in this thesis. This theoretical background address a density-
functional theory description starting from the well-known problem in the de-
scription of many-body systems. Later, the discussion of the Hohenberg-Kohn
theorems and the Kohn-Shan ansatz that makes possible the practical solution
and application of the theory is carried out. After, a small description of the func-
tionals, pseudopotentials and more practical solid-state theory will be addressed.
The later in order to fill the gap between the theory and the implementation
leading with this, to a powerful use of its advantages. All of this with the aim
to investigate and predict the origin of many novel and exciting materials and
properties.

�.� Many-Body Problem

Taking the quantum mechanics as the basis to describe the ions and interacting
electrons, we know that a system is represented by the wavefunctions and their
interactions expressed by the Schrödinger’s equation in the time-independent�
approach as [��]:

Ĥ | i = E | i , (�.�)
where the Hamiltonian operator (Ĥ) acting over the wavefunction ket (| i) has
as a result, the energy eigenvalues (E) times the wavefunction. The Ĥ operator
for the ions composed by the nuclei and electrons is described by the kinetic and
potential energy interactions taking the expression:

Ĥ(R, r) = T̂N (R) + T̂e (r) + V̂NN (R) + V̂ee (r) + V̂eN (r, R), (�.�)
� The time-dependent Schrödinger’s equation present the form: Ĥ(t) | i=ih( @t
@
)| i

�
� theoretical background

here, the operators for the kinetic energy of the nuclei (T̂N ) and the electrons
(T̂e ) are as follows:

h2 X 1 2
T̂N (R) = - r , (�.�)
2 MI I
I
h2 X 2
T̂e (r) = - ri , (�.�)
2me
i

by notation, the coordinates related to the ions and electrons are given in upper
and lower cases subscripts respectively (i.e. I, J, etc. and i, j, etc.). MI and me
are the masses for the ions and the electrons respectively.
The electrostatic potentials for the ion-ion, electron-electron, and electron-ion
interactions are given by:

e2 X ZI ZJ
V̂NN (R, R) = , (�.�)
2 kRI - RJ k
I6=J
e2 X 1
V̂ee (r, r) = , (�.�)
2 kri - rj k
i6=j
X ZI
V̂eN (r, R) = -e2 , (�.�)
kri - RI k
i,I

In order to reduce the number of terms in the Hamiltonian presented in Eq.

�.� it can be assumed that the relation between the ions and electrons masses is
MI >>me . This can be understood in terms of the motion velocity of the electrons
with respect to the ionic motion, which turns out to be almost instantaneously
(adiabatically) with respect to the nucleus. Therefore, the kinetic energy of the
ions can be neglected and the wavefunction can be treated only based in the
electronic degrees of freedom. The later statement is called Born–Oppenheimer
approximation and have the effect on the Hamiltonian as shown in the Eq. �.�:

Ĥe (R, r) = T̂e (r) + V̂NN (R) + V̂ee (r) + V̂eN (r, R). (�.�)

The ion-ion potential, V̂NN is recognized such as the Madelung energy, which
can be easily computed in periodic systems. The exact calculation of this term
can be reviewed in the appendix F by Kaxiras [��]. The V̂eN potential is now
independent of the ion RI coordinates and can be represented only as a function of
the electron positions (ri ). After these considerations, the electronic Hamiltonian,
Ĥe , assuming a fixed nuclei, is described by the sum shown in Eq �.�:

X
Ĥe = T̂e + V̂ee + V̂ext = T̂e + V̂ee + v(ri ). (�.�)
i
 �.� density-functional theory �

Thus, the electronic energy can be computed such as:

Ee = h e | Ĥ | e i =h e | T̂e + V̂ee + V̂ext | ei , (�.��)

where e = e (r1 , r2 , ... , rN ) for the N electrons. As we know from quantum

mechanics, this wavefunction must be normalized (i.e h | i=1) and antisymmet-
ric under the exchange of two particles (i.e. e (r1 , r2 , ... , rN ) = - e (r2 , r1 , ... ,
rN )).
At this point, even when several approximations have been taken into account,
the problem of solving the Eq. �.�� and finding the | e i wavefunctions in a crystal
is by far, a very complicated task due to the large number of electrons (of the
order of ��23 ) and degrees of freedom in a crystal. Some methods were developed
to find the solution to this problem, being the density functional theory (DFT)
one of the most successful.

�.� Density-Functional Theory

In DFT, we will find one of the most successful methods for electronic structure
calculations in solid-state physics. The later because the approximations used
have shown to provide an excellent balance between computational accuracy and
cost. Then, much larger systems can be studied than with others ab initio meth-
ods. Additionally, DFT is a completely rigorous and formally methodology to
model a many-body interacting system (as shown previously). This is achieved
by “switching” the point of view to a much simpler and easier route to solve the
non-interacting problem. For those reasons, the DFT methodology has been, is
and it will be one of the most applied methodologies to understand and unveil
the physics behind materials in a large variety of fields.
Lets start by saying that we can define the number of electrons in a system by
integrating the electronic density (n = n(r)) over a volumetric space as follows:

Z
N = n(r)dr. (�.��)

Now, lets take the electronic density as a function of such as:

Z Z
n(r) = N . . . | (r1 , r2 , . . . , rN )|2 dr1 , dr2 , . . . , drN , (�.��)

here, if we think into invert the dependence of terms in the expression in Eq.
�.��, we will see that the wavefunction will become = [n(r)], which is indeed
� theoretical background

a functional of the density� . Even if the later reasoning looks quite simple, it is
indeed the foundation of the density-functional theory as shown next.

�.�.� The Hohenberg-Kohn Theorems

As previously discussed, the wavefunctions can be expressed in terms of the elec-

tronic density. This argument is very powerful because it means that now we can
think in solving the Schrödinger’s equation just by defining a functional of the
electronic density with only three spatial variables instead of the 3N -variables in
the wavefunctions. This statement was condensed and proved by the Hohenberg-
Kohn theorems [��]:

Theorem �: The ground-state energy from Schrödinger’s equation is a unique

functional of the electron density [��].

Theorem �: The electron density that minimizes the energy of the overall
functional is the true electron density corresponding to the full solution of the
Schrödinger’s equation [��].

The proofs for these theorems are beyond the aim of this chapter and it can be
find in several electronic structure textbooks such as Martin [��]. The previous
theorems establish that there is a one-to-one relationship between the ground
state energy and the electronic density n(r) determined by a unique external
potential. Besides, if the “true” functional is known, then, we can perform a min-
imization of the energy in order to find the related electron density.
Thus, just after considering these theorems, we can redefine the energy (pre-
sented in Eq. �.��) as:

E[n(r)] = h e | T̂e + V̂ee | e i + h e | V̂ext | e i (�.��)

Z
= F[n(r)] + v(r)n(r)dr, (�.��)

being n(r) the electronic density for the ground state and F[n(r)] a functional of
the density that regroups the kinetic and potential energies added to the external
potential energy. Nevertheless, in all the development shown by Hohenberg and
Kohn, the F[n(r)] is unknown and to find a solution to the Schrödinger’s equation
it is still a problem at this stage of the framework. Then, it was only after the
seminal work by Kohn and Sham [��] that some light was shed onto this problem.

� As a simple example of a functional lets take first the function f(x) = x2 + 1, then a func-
R1
tional F[f(x)] of the function f(x) can be written as: F[f] = f(x)dx. After evaluating the later
-1
expression we have that F[f] = 38 .
 �.� density-functional theory �

�.�.� Kohn-Sham Equation

The main idea of Kohn and Sham was to map the many-electron system into
a single-particle description [��]. For that purpose, the F[n(r)] functional was
redefined as:

F[n(r)] = Tks [n(r)] + VH [n(r)] + Exc [n(r)], (�.��)

here, Tks is the kinetic energy of a system with density n but, excluding the
electron-electron interactions. VH is known as the Hartree potential derived from
the classic Coulomb potential for the electrons. Finally, Exc defines the exchange-
correlation energy missing in Tks and VH . It is important to note at this point
that Tks is not the real kinetic energy, nonetheless, its magnitude is comparable
and can be computed without any approximation. Then, all the terms in the Eq.
�.�� can be evaluated exactly, except for the exchange-correlation energy, which
as we will see, it is the main concern in the theory. As a next step, by considering
the Eqs. �.�, �.�, �.��, and �.�� we have the Schrödinger’s equation with the
form:

" Z #
h2 2 n(r 0 ) 0
- r + dr + vxc (r) + v(r) | ii = ✏i | ii , (�.��)
2me |r - r 0 |

where the exchange-correlation potential is defined as a functional derivative

of the exchange energy Exc :

Exc [n(r)]
vxc = . (�.��)
n(r)
As we can see in the Eq. �.��, if we regroup the terms with a potential character,
we have the non-interacting one-electron Schrödinger’s equation expressed in the
Eq. �.��:


h2 2
- r + veff | ii = ✏i | ii . (�.��)
2me

The Eq. �.�� will be renamed from now on as the Kohn-Sham equation. Based
in the new Konh-Sham orbitals, | i i, the electronic density can be computed as:

X
nKS (r) = | 2
i (r)| . (�.��)
i

Then, the advantage of solving the Kohn-Sham equations is evidently based in

the knowledge of the solution of the single-particle Schrödinger’s equation with
� theoretical background

a local effective potential, assuming that the core-valence and valence-valence

electrostatic interactions can be directly evaluated. Nevertheless, Exc is different
because cannot be calculated explicitly and some approximations must be used
in order to obtain it.
Another small trouble will become visible if you carefully check the equations,
thus you will realize that in order to solve the Kohn–Sham equations (Eq. �.��),
you need to define the Hartree potential (Eq. �.��), and to define the Hartree
potential, you need to have the knowledge of the electron density. However, to
find the electron density, you must get the single-electron wave functions (Eq.
�.��), and to know these wavefunctions we must solve the Kohn–Sham equations,
going back again to the Eq. �.��. To break this loop and obtain a real solution,
the proposed process is treated as an iterative method as outlined in the following
algorithm:

�. Define an initial, trial electron density, n(r).

�. Compute the effective potential (Eq. �.��) using the electron density.
�. Solve the Kohn–Sham equations defined (Eq. �.��) by using the trial elec-
tron density and the effective potential to find the single-particle wavefunc-
tions, i (r).
�. Calculate the electron density nKS (r) using the Eq. �.�� defined by the
Kohn–Sham single–particle wavefunctions as obtained from the previous
step.
�. Compare the calculated electron density, nKS (r), with the electron density
used to solve the Kohn–Sham equations, n(r). If the two densities are the
same, then it means that this is the electron density for the ground state,
and it can be used to compute the total energy. If the two densities are
different, then the trial electron density must be updated to the nKS (r).
Once this is done, the process begins again going back to the step �.

�.� Exchange–Correlation Energy and Function-

als
As commented previously, the density functional theory is exact with the excep-
tion of the exchange–correlation energy, which needs to be computed by means
of numerous approximations. Here we present some of the most used ones, which
show great accuracy in explaining and predicting material’s properties.

�.�.� LDA and GGA

The most basic form, but at the same time, the most used methodology to ap-
proximate the exchange–correlation energy is by means of the local-density ap-
 �.� exchange–correlation energy and functionals �

proximation (LDA). Within this approximation, the exchange-correlation energy

is defined as:
Z
ELDA
xc [n(r)] = n(r)✏hom
xc [n(r)]dr, (�.��)

where ✏hom
xc is the exchange–correlation energy per particle in a homogeneous
�
electron gas . Here, it is assumed that the energy per particle at the point r is
only a function of this point, besides, the exchange energy of a homogeneous
electron is known having the form as follows:

3 1
✏hom
x [n] = - (3⇡2 n) 3 . (�.��)
4⇡
The missing correlation part can be computed using the quantum Monte Carlo
approach as shown in Ref. [��]. Surprisingly, even when this approximation looks
quite rudimentary, nowadays, it is one of the most used and gives a good agree-
ment with experimental results (e.g. ferroelectric properties in materials [��]).
Nonetheless, one of its disadvantages is the over estimation of the cohesive en-
ergy that it can be appreciated in smaller lattice cell parameters in comparison
to experimental data. The wrong computation of the electronic band gap with an
underestimation of its value is also a known disadvantage of the approximation.
The updated representation of the exchange–correlation energy to overcome
some misdescribed issues is based onto the generalized gradient approximation
(GGA). In this approximation, the variation of the electron density is taken into
account in the ✏xc term as:
Z
EGGA
xc [n(r)] = n(r)✏xc [n(r), |rn(r)|]dr. (�.��)

There are hundreds of exchange–correlation functionals used in condensed

mater and chemistry. Nonetheless, the most used ones are the Ceperley-Alder
(CA) [��], Perdew Burke Ernkzerhof (PBE) [��], AM�� [��, ��], PBE revised for
solids (PBEsol) [��], and the Perdew-Wang �� (PW��) [��].
Even when several of the mentioned functionals give a large amount of precision
in the computed properties of materials; some physical and chemical phenomena
related to strongly correlated electrons [��] shows a lack in the description when
are obtained using the same functionals. Then, more advanced approaches (with
respect to LDA and GGA) have been developed in order to overcome this problem.
Approaches such as DFT+U and hybrid functionals have been successfully ap-
plied to explain the electronic properties in systems such as A2+ TiO3 /R3+ TiO3
superlattices [��] and other interfaces [��]. These methodologies are described in
the next sub-sections.
� Spin-polarized systems can be taken into account just by having the energy such as: E[n(r)] )
E[(n" (r), n# (r))]
�� theoretical background

�.�.� DFT+(U,J)

In strongly correlated systems such as transition metal oxides, rare-earth based-

materials, and interfaces, DFT calculations show a wrong description of the elec-
tronic properties due to the lack in the definition of the exchange–correlation
energy. This is due to the non-systematic development of the XC functionals
and the lost of information of the highly localized –and then highly interacting–
electronic systems. A correction of the on-site Coulomb interaction (U ) and the
on-site exchange (J ) has been implemented by means of the DFT+U approach.
For that purpose, there are two main strategies developed to solve the DFT+U
Hamiltonian: a first one in which the U and J parameters enter as independent
corrections in the calculations [��]; and a second one in which only a single ef-
fective Ueff = U –J parameter accounts for the Coulomb interaction, neglecting
thereby any higher multi-polar terms [��]. In the DFT+U approach, an addi-
tional orbital-dependent interaction is included. Then, the total energy of the
system can be expressed in the most general form as [��]:

EDFT +U [n(r)] = EDFT [n(r)] + EHub [⇢lmm 0 ] - Edc [⇢l ], (�.��)

here, the correction term EHub is known as Hubbard Hamiltonian for corre-
lated states. Edc is the correction term for the double-counting of the states. This
problem arises because the Kohn-Sham energies already include the interactions
effects within the Hartree and exchange correlations terms. Then, adding the
Hubbard Hamiltonian leads to a double-counting of some interactions that needs
to be corrected. Note that in the EHub and Edc the atomic site index, the state,
and the spin are taken into account with the l, m, and lower and upper indices
respectively. Finally, the ⇢lmm 0 are the occupation numbers of the localized or-
bitals. In order to solve the Eq. �.��, the EHub and Edc terms can be computed
as follows:

X  Ul X 0 Ul l l
EHub
[⇢lmm 0 ] - Edc [⇢l ]= ⇢lm ⇢lm 0 - ⇢ (⇢ - 1) , (�.��)
2 2
l j

with j = m, 6= m 0 , 0 and occupation numbers computed from the projection

of
P
the Kohn–Sham orbitals onto the states of a localized basis set as: ⇢lmm 0 =
k,v fk,v h k,v | m 0 i h m | k,v i, where the fk,v coefficients are the occupations of
l l

the KS states.
One of the problems within this approach is that there is no exact determination
of the U and J values, and then, they must be adjusted based on experiment
data such as magnetic moment, magnetic anisotropy, and electronic band gap.
An alternative method to overcome this issue was proposed by Cococcioni and
de Gironcoli [��]. In this method, a linear response approach is used in order
to calculate the interaction parameters entering to the LDA+U functional. This
 �.� exchange–correlation energy and functionals ��

calculation is performed internally and self-consistently to avoid the ill-defined

parameters into the Hubbard Hamiltonian. Even when this method has been
tested and show great accuracy in describing the electronic properties of highly-
correlated systems, it is not commonly implemented in most of the available DFT
codes.
For a better understanding of the theory involved in the development of the
E Hub and Edc the reader can have a look into the review by Himmetoglu et al.
[��] and the original papers of Liechtenstein et al. [��] and Dudarev et al. [��].

�.�.� Hybrid Functionals

The hybrid functionals own their name to the mixture character of the exchange–
correlation energy. This energy is defined in terms of an amount of an orbital-
dependent Hartree-Fock (HF) energy plus and explicit DFT part. As far as we
know, those are among the most accurate methods up to day —in addition to
Dynamical Mean-Field Theory, DMFT [��] and GW [��] methods for example—.
The exchange-correlation functionals can be expressed in their most general form
as:

Exc = ↵EHF DFT

x + (1 - ↵)Exc , (�.��)

where, the EDFT

x denotes any GGA or LDA functional. Thus, a non- and fully-
interacting electronic system will be equivalent to ↵ = � and ↵ = � respectively.
Several “mixing” functionals have been developed and here we will comment about
some of them. The B�LYP [��] is among the most used ones and it is expressed
as follows:

EB3LYP
xc = ELDA
xc + a0 (EHF DFT
x - Ex ) + ax EB88
x + ac EPW91
c , (�.��)

here the a0 , ax , and ac coefficients are fixed to �.��, �.�� and �.�� values
respectively adjusted to fit atomic and molecular data. EB88 x is the exchange
energy term taken from the Becke’s ���� gradient correlation within LSDA func-
tional [��]. Finally, the correlation term EPW91
c is taken from the Perdew-Wang
�� functional [��].
Another example is the functional developed by J. Perdew, M. Ernzerhof, and
K. Burke (PBE�) [��]:

1
Exc = ELDA
xc + (EHF - EDFT
xc ), (�.��)
4 x
in this functional, �/� of mixing is employed based in several test in which the
EDFT
xc is set to the PBE functional.
�� theoretical background

The B�WC [��] hybrid functional was developed especially to be used in per-
ovskite oxides. In this functional the GGA approach is taken within the Wu-Cohen
[��] representation and it is employing an ��% of mixing as shown in the Eq. �.��:

EB1 GGA
xc = Exc + 0.16(EHF GGA
x - Exc ). (�.��)

The hybrid Heyd-Scuseria-Ernzerhof exchange–correlation functional, HSE06

[��, ��] is differentiated from the previous ones basically by employing a screened
Coulomb potential. Starting from the exchange part of the PBE0 representation
(Eq. �.��) and splitting the terms into long- and short-range components (denoted
as SR and LR respectively) we obtain:

EPBE0
x = a[EHF,SR
x (!) + EHF,LR
x (!)[+(1 - a)[EPBE,SR
x (!) + EPBE,LR
x (!)], (�.��)

where ! is an adjustable parameter governing the interplay between the SR

and LR interactions. For this parameter, several numerical tests show that if we
take ! = �.��, the interaction up to the second neighbors are well reproduced.
Besides, it has been established that the HF and PBE long-range interactions are
rather small and can be neglected. Then, we obtain a screened Coulomb potential
hybrid functional as follows:

E!PBEh
xc = aEHF,SR
x (!) + (1 - a)EPBE,SR
x (!) + EPBE,LR
x (!) + EPBE
c . (�.��)

In the Eq. �.�� the !PBEh term converges to PBE0 and PBE for ! ! � and
! ! 1 respectively. For more detailed information about the derivation of the
SR and LR terms, please refer to the reference of Heyd et al. [��].
It is important to note that some of these hybrid functionals will be used in
the analysis of oxide/fluoride interfaces in the chapter �.

�.� Density Functional Perturbation Theory

At this stage of the theory, a quite wide description of the electronic and structural
landscape of materials is possible within DFT. But to improve the properties
characterization, analysis, and consequently to move forward with the theory, the
density functional perturbation theory (DFPT) [��] was successfully developed
and implemented. With this advance, the study on phonons, dielectric response
functions, nonlinear optical susceptibilities, electro-optic coefficients, and electron-
phonon interactions became more accessible. To achieve that, the dependence of
the energy and its derivatives with respect to several parameters of interest need
to be computed (see Eq. �.��). Then, the main idea is to expand the energy into a
power serie around the ground state based in the perturbation parameter . The
later can be equivalent to the atomic displacements ⌧, electric field E, and/or
 �.� density functional perturbation theory ��

strain ⌘ depending the case. Thus, taking the expansion up to the first-order we
can access to the forces (F ⌘ @E/@⌧), polarization (P) and strain constants ( ) as
shown in the Table �.�. When the expansion is carried out up to second-order,
information related to phonons (i.e Interatomic force constants, C), Born effective
charges (Z⇤ ⌘ @2 E/@⌧@E ⌘ @P/@⌧) and the internal strain coupling parameters
( ) can be obtained (see the Table �.�). Other quantities like the clamped-ion
elastic constant (c0 ), the clamped-ion piezoelectric tensor (e0 ), and the optical
dielectric tensor (✏1 ) can be also computed as presented in the Table �.�. The
third order derivatives give information about nonlinear response such as Raman
tensors, optical susceptibilities and electro-optic properties.

X @E 1 X @2 E
E( ) = E(0) + i + i j+
@ i 0 2 @ i@ j 0
i i,j
(�.��)
1X @3 E
+ i j k +...
6 @ i@ j@ k 0
i,j,k

The calculations of the derivatives shown in Eq. �.�� are based on a variational
principle involving a minimization procedure giving also access to non-linear re-
sponses based in the "�n+�" theorem [��]. In this theorem is established that the
energy of the �n+� state can be obtained with the knowledge of the states from
0 through n.

Table �.�: Available physical quantities through the first- and second-order energy
derivatives with = ⌧, ⌘, and E, (see the Eq. �.��). Table adapted from
Ref. [��].
�st -order �nd -order
@/@⌧ @/@⌘ @/@E
@/@⌧ F C Z⇤
@/@⌘ c0 e0
@/@E P Z⇤ e0 ✏1

An alternative method to obtain the energy derivatives used in the dynamical

analysis is based in the so-called, frozen-phonon technique [��]. In this approach,
higher-order derivatives are determined by a finite amplitude perturbation of the
structure. Then, the energy of both, distorted and undistorted structures are
computed and related to the displacements from one to another having with this
an estimation of the energy derivative. This process can be performed for several
distortions and then, the dynamical matrix can be obtained accessing with this
to phonon-dispersion curves (widely used in this thesis), dielectric tensors and
elastic constants.
�� theoretical background

�.� From the theory to the implementation

Even when all the theory has been developed and the KS equations have set a tan-
gible solution to numerically solve the many-body problem, the implementation
still requires a lot of computational efforts in order to make it affordable. Then,
another simplification has been made by taking advantage of the periodicity of
the crystals as shown in the next subsections.

�.�.� Plane-waves expansions

In a periodic crystal with periodic boundary conditions, an electronic wavefunc-

tion can be written in terms of the plane-waves basis set and using the Bloch
theorem as follows:

1
n,k (r)= p un (k, r)eikr (�.��)
⌦
where k is a lattice wave-vector into the reciprocal space, ⌦ represents the cell
volume and n stands for the band number. un is the periodic function that holds
all the information related to the crystal periodicity. Later, for a more practical
view of the Bloch wavefunctions, those are decomposed into a Fourier expansion
of the form:

1 X
n,k (r) =p Cn,k (G)ei(k+G)r (�.��)
⌦ G
Here, Cn,k are the Fourier coefficients and G is a vector in the crystal recip-
rocal space. Therefore, for any reciprocal lattice vector we have that n,k (r) =
n,k+G (r). Nonetheless, from the computational point of view this expansion
must be limited up to a finite number of plane-waves. For this purposes, a cut-off
energy Ecut condition is established as:

h2
|k + G|2 6 Ecut (�.��)
2m
The Ecut value is independent for each crystal, symmetry and atomic species
and then, a correct selection of an energy value should be set in each particular
case which is commonly known as the energy cutoff convergence.
One of the disadvantages of this methodology is that a large number of plane-
waves is required to describe the electronic wavefunctions close to the nucleus.
This due to the faster change of the valence electrons wavefunctions in the region
close to the atomic nucleus. Thus, a very large Ecut —and thus, a large number
of plane-waves— is necessary to describe these wavefunctions correctly in com-
parison to the energy needed for the electrons close the valence states. In order
to overcome this issue, another approximation has been developed known as the
pseudopotential approach as it will be described later.
 �.� from the theory to the implementation ��

�.�.� Periodic Boundary Conditions

The Eq. �.�� take advantage of the Born-von Karman periodic boundary condi-
tions, which allow us to represent an infinite periodic crystal into a representative
building block of construction known as unit cell. At the same time, thanks to
the periodic boundary conditions, the wavefunctions can be presented in the form
of the Bloch functions as commented previously. In order to calculate the wave-
funtions from the Eq. �.�� —and consequently, the electronic density as shown
previously—, an integration over the reciprocal space involving all the bands and
the Bloch functions over an infinite k-points number is required. In practice, this
is impossible and needs to be truncated into a finite number of k-points in a
similar way as done with the Ecut . Therefore, the convergence of the calculations
require enough k-points in order to reach a sufficient precision within a reasonable
computation time. The later process is usually known as the k-points sampling
and determine the accuracy level in the calculations (i.e. bigger is the number
of k-points, higher is the calculation’s precision but larger is the computational
time).
Several methods to sample the reciprocal space and then, to select the k-points
mesh have been developed. This is the case for example, of the Monkhorst-Pack
[��] mesh in which essentially the k-points are distributed homogeneously in the
Brillouin zone. In this technique, rows and/or columns of k-points are running
parallel with respect to the reciprocal lattice vectors in the specific Brillouin zone.
The previous considerations within the periodic boundary conditions are very
convenient when bulk materials and/or superlattices are represented. Neverthe-
less, the modeling of molecules, isolated surfaces, thin films and/or material’s
defects require the definition of special computational unit cells. For more details
in these special systems, please check the reference of Sholl and Steckel [��].

�.�.� Pseudopotentials

The pseudopotential approach [��] is based on the idea of replacing the strong
Coulomb potential of the nucleus by an effective ionic potential (i.e. shielding
potential) acting on the valence electrons only. This approximation can be done
because most of the electronic properties of materials are mainly determined by
the valence electrons while the core electrons can be considered as frozen. The
later statement has a few exceptions in which, deeper electronic levels in compar-
ison to the valence states have an influence on the properties (e.g. ferroelectric
materials under pressure [��, ��]). Thus, the pseudo wavefunction have no radial
nodes in the core region and the pseudopotential is smoother than the all-electron
potential near to the nucleus but identical beyond the cut-off radius, rc . The key
feature of a pseudopotential lies in its transferable and accurate character, then,
even if the pseudopotential was constructed in one environment (the atom itself),
it should faithfully reproduces the properties of a material when it is placed in
different environments such as molecules, solids and ions.
�� theoretical background

Finally, we can remark that the pseudopotential approximation allows us to

decrease the number of plane-waves needed in the Bloch expansion (shown in
the previous subsection). Besides, the number of electrons considered explicitly
in the Kohn-Sham equations is reduced to only the interacting valence electronic
configuration, achieving with this, a considerable saving of the calculation time.

�.�.� Projector Augmented-Waves (PAW) method

The PAW method was proposed by Blöchl [��] in ����. This method deal with
the main shortcoming of the pseudopotential approach, in which, most of the
information regarding to the wavefuntion is lost in the region close to the nuclei.
Here is important to recall that in the bonding region the wavefunction is smooth,
but close to the nuclei it displays rapid oscillations that are very demanding on
the numerical representation (i.e. large number of plane-waves in the pseudopo-
tential approach). Then, to address this problem the PAW method proposed the
following mathematical formalism:

Lets define a linear transformation T̂ , which transform the true all-electron

↵ KS
˜
single particle wavefunction (i.e. | n i) to an auxiliary smooth one ( n ), then:
↵
| n i = T̂ ˜ n . However, it is known that the true wavefunctions are already
smooth at a certain minimum distance from the core. Thus, T̂ should only modify
the wavefunction close to the nuclei. Based on the later statement, we define the
operator for that region only as:

X
T̂ = 1 + T̂ a , (�.��)
a

where the a index is an atom, and T̂ a is the atom-centered transformation that

has no effect outside a certain atom-specific augmentation region defined by |r -
Ra | < ra
c . Here, the cut-off radii rc should be chosen to avoid the overlap of the
a

augmentation spheres. Now, for convenience,

↵ we choose a complete basis inside
the sphere known as partial waves, i . In the same way as for the wavefunctions,
a

the T̂ acts on the smooth partial waves as:

↵
| a
ni = (1 + Tˆa ) ˜ a ˆa ˜ a
i ,T | ii=|
a ˜a
i i-| i i. (�.��)

Here is important↵ to remark that the T̂ operator is only acting inside the sphere,
then, | a
n i = ˜ a when r > ra . Under the assumption that the smooth partial
n c
waves form a complete set inside the augmentation sphere, we can expand the
smooth all-electron wavefunctions as:

↵ X a ↵
˜n = Pni ˜ a
i , (�.��)
i
 �.� from the theory to the implementation ��

where Pni
a are the expansion coefficients. If we multiply by T̂ in both sides of the

Eq. �.��Pwe find that the same is also true for the non-smooth wavefunctions (i.e.
| in = i Pni a | a i). Therefore, the P a expansion coefficients are the same for
n ni
both, smooth and non-smooth wavefunctions. ↵In order to find such coefficients,
we choose a smooth projector functions, p̃a i that satisfies the orthogonality
condition and must be complete inside the augmentation spheres, then:

hp̃a ˜a
i| ji = ij . (�.��)
X
|˜a a
i i hp̃i | = 1. (�.��)
i
⌦
Then, multiplying the Eq. �.�� by p̃a
i and using the Eq. �.�� condition we
have that:

X X
hp̃a ˜
i | ni =
a
Pnj hp̃a ˜
i | ji =
a
Pnj ij
a ,
=Pni (�.��)
j j

now, we have the expansion coefficients as a function of the projectors and

the smooth wavefunctions. Finally, we have to write down the operator T̂ as a
function of the projectors and the partial waves using the Eq. �.�� and the Eq.
�.�� as follows:

X X
T̂ a = T̂ a · 1 = T̂ a |˜a a
j i hp̃i | = (T̂ a | ˜ a a
j i) hp̃i |
i i
X (�.��)
= (| a ˜ a a
i i - | i i) hp̃i | ,
i

Then, the full operator (based in the Eq. �.��) can be expressed as:

X
T̂ = 1 + (| a ˜a a
i i - | i i) hp̃i | , (�.��)
i,a

In conclusion, the KS wavefunctions can be obtained

↵ in terms of the projectors
and the partial waves (knowing that | n i = T̂ ˜ n ) as follows:

↵ X X ↵ a
| ni = ˜n + | a a
i i Pni - ˜a
i Pni . (�.��)
i,a i,a

The relation �.�� can be schematically represented as in the Fig. �.�. In the
figure it can be noted that the full wavefunction is expressed as the sum of the
pseudo plane waves (first term in Eq. �.�) plus the all-electron “on-site” interaction
(close to the nuclei and inside de augmentation spheres, second term in Eq. �.�)
minus the “on-site” part from the pseudo wavefunctions (third term in Eq. �.�).
�� theoretical background

= + -

Full wavefunction Plane waves All-electron “on-site” Pseudo “on-site”

Figure �.�: Schematic representation of the main PAW method that exemplifies the Eq
�.��.

In this way, we are keeping the information about the wavefunction around the
core region without expending large computational time. The later because the
number of plane waves is considerably reduced in comparison to the regular pseu-
dopotential approach discussed in the previous section. With this, the accuracy
is significantly increased and the computational efforts are reduced.

For a deeper understanding of the pseudopotentials generation, its basis, and

the PAW method the reader can refer to the chapter �� in the electronic structure
book of Martin [��].

�.�.� Zeeman applied magnetic field

In order to study the electronic properties as a function of the magnetic applied

field, we used the implementation of Bousquet et al. [��]. In this implementation,
the magnetic energy is added to the Hohenberg-Kohn energy functional. Then,
the Legendre transformation of this functional becomes:

⇤[n 0 (r), H] = EHK [n 0 (r)] - µ0 H · mtot , (�.��)

where n 0 (r) is the spin-density matrix, which takes into account a non-
collinear framework in the �⇥� matrix allowing to take into account the mag-
netization spin-density variations (i.e. its magnitude and direction). The EHK
term is the zero-field energy functional as shown in the previous sections. H is
the applied magnetic field, µ0 is the magnetic permeability of free space, and
mtot is the total spin magnetic moment of the system obtained from the spatial
integration of the m(r). In the same way as shown previously, the Eq. �.�� can be
solved by the Kohn-Sham approach into the single-particle spinor orbitals schema.
Then, in addition to the usual terms, an external magnetic field acting on the
spinor orbitals is obtained and the related potential shift have the form:

!
g Hz Hx + iHy
V 0 = - µB µ0 , (�.��)
2 Hx - iHy -Hz
 �.� from the theory to the implementation ��

here, the g term is the Landé factor, Hj (with j= x, y, and z) are the magnetic
field vector components and µB is the Bohr magneton. This implementation allow
us the computation of several properties of interest such as the magnetoelectric
coupling as shown in the Chapter � for the NaMnF3 compound. This, opening the
possibility to unveil the properties in multifunctional compounds with a related
magnetic behavior.

�.�.� Used codes: VASP and CRYSTAL

VASP— The Vienna ab-initio simulation package (VASP) is the main code used
in the development of this thesis [��, ��]. Projector Augmented-Waves (PAW)
[��] pseudo-potentials were used to represent the valence and core electrons. The
exchange correlation was represented within the GGA - PBEsol parametrization
[��]. The magnetic character of these systems was include and d -electron localiza-
tion corrected by means of the GGA+U within the Liechtenstein formalism [��],
for which, the U value was found by fitting the gap and the magnetic moment as
a function of this parameter. The periodic solutions of these crystalline structures
were represented by using Bloch states with fully converged k-points and energy
cut-off values. The spin-orbit coupling was include into non-collinear calculations
[��]. Born effective charges and phonon calculations were performed within the
density functional perturbation theory (DFPT) [��]. The dynamical matrix was
then unfolded using the Phonopy code [��]. The ferroelectric spontaneous polar-
ization computed through the Berry phase approach [��].
CRYSTAL��— For big cells and large number of atoms, hybrid calculations
have been performed using the linear combination of atomic orbitals (LCAO) as
implemented in the CRYSTAL code [��]. This code is based in a Gaussian-type
basis sets including polarization orbitals with the advantage of decrease consid-
erably the computation time comparing to plane-wave basis set as in VASP.

In conclusion, a basic landscape of all the theory involved in the development

of this thesis has been described, Nonetheless, for more and deeper information
the reader can check also the references of Martin [��], Kaxiras [��], Dobson et al.
[��], and Sholl and Steckel [��, ��].
 Part II

S T R U C T U R A L P R O P E RT I E S O F
PEROVSKITES ABF3

Interestingly, none of the known perovskite fluorides have been re-

ported to be ferroelectric and no explanation has been addressed in
the literature about the absence of ferroelectricity in this crystal class.
In this section we use first-principles density functional calculations
to investigate computationally the tendency towards ferroelectricity
in perovskite fluorides. Our goals are two-fold: First, by comparing
to the behavior of ferroelectric perovskite-structure oxides, we further
understanding of the driving forces for, and competition between, po-
lar and non-polar structural distortions in both systems. Second, we
aim to identify conditions, for example of strain or chemistry, under
which ferroelectricity, and perhaps multiferroism, could be stabilized
in the fluoride perovskites.
 V I B R AT I O N A L B E H AV I O R I N
2
FLUOROPEROVSKITES ABF3

Ferroelectricity in perovskite oxides has attracted a huge interests going from

fundamental studies to technological and industrial applications. Nowadays, this
interest is even stronger because of the great challenge of finding materials that ex-
hibits both ferroelectricity and magnetism —AFM or FM orderings—, although
the highly electronic correlated systems, ABO3 perovskites are considered as the
most attractive and promising candidates [��]. Interestingly, Scott and Blink have
recently pointed out that multiferroism and magnetoelectricity can also be widely
found in the leaved aside fluoride-based materials [��]. Examples of these systems
recently studied are the multiferroics based on barium-fluorides BaBF4 (B =
Mn, Fe, Co, Ni) [��] that present a quasi-�D layer-by-layer structural octahedral
arrangement. This arrangement leads to octahedral rotations with an unstable
phonon mode yielding to a net spontaneous polarization that generates its ferro-
electricity. This fluoride family also presents an antiferromagnetic behavior due to
transition magnetic metal ions located at fluorine-octahedra center. These ions
possess partially filled d-levels that give rise to magnetic order with magnetic
moments that tend to align in the bc-plane making an angle of �� with respect
to the b-axis [��]. Other examples are KMnFeF6 and Ba2 Ni7 F18 with a magne-
toelectric behavior predicted by phenomenological Landau analysis [��]. KCrF4
with Pnma structure and quasi-�D antiferromagnetic ordering with infinite Cr-F6
octahedra columns along b-axis [��]. This material possess a spontaneous polar-
ization components of Py (or Pz ) as a function of applied magnetic field along z
(or y) crystalline direction. The family Ba6 Bn F12+2n (BMF) [��] also make part
of this kind of materials with a magnetic frustration due to corner-sharing octa-
hedra phenomena and spontaneous polarization induced in all three directions
by the application of a magnetic field. Then, as shown previously, ferroelectric-
ity has been analyzed only in a few number of fluoride compounds with a lack
in the understanding of the microscopic origin of the polarization within. Thus,
as one of the aims of our work is to explore this question by scrutiny analyze
the origin of the polar ferroelectric instabilities in magnetic perovskite fluorides.
Interestingly, we found that the ferroelectric instability in ABF3 has a totally dif-
ferent origin and behavior than in oxides even if their structural and vibrational
properties are very similar. Ferroelectricity in ABO3 has been first described by
the softening and the condensation of a soft� polar mode when cooling down
the temperature going from the cubic Pm3̄m high temperature structure to po-
� A soft mode is a phonon mode with a frequency that falls to zero, as a function of temperature,
at the phase transition.

��
�� vibrational behavior in fluoroperovskites abf 3

lar ferroelectric ground state [��]. This was first attributed by a subtle balance
between the short-range (SR) forces and the long-range (LR) Coulomb interac-
tions in which the later destabilizes the crystal toward a ferroelectric state [��].
Later on, first-principles calculations evidenced that the microscopic key ingredi-
ent for ferroelectricity is a dynamical charge transfer between the oxygen anions
and the B-cation due to hybridization between O-�p and empty B-site d-orbitals
[��]. This effect is highlighted by the presence of anomalously large Born effective
charges (Z⇤ ) on these atoms [��] and by an unstable polar phonon mode at the
center of the Brillouin zone of the cubic high temperature phase [��]. Thus, to
our knowledge, almost all experimental and theoretical researches in the field of
magnetoelectrics and multiferroics have been dedicated to oxide systems, with
special focus to perovskite structures and only a few works and reports —in com-
parison to oxides— have been focused in analogous systems such as fluorides.
Then, the understanding of physical phenomena is particular to each material
and cannot be generalized. Here, as an alternative, emerge some structures based
on fluoride systems with a similar crystal structure than the already discussed
oxides, such as KBF3 [��], and NaBF3 [��, ��]. In these systems, the physical
phenomena such as superexchange interaction, stabilization of octahedral-tilting,
and possible multiferroic or magnetoelectric behavior is still unclear. According
to our previous discussion, we carried out a research on ABF3 fluoroperovskites
in order to explore the possibility to include this material as a new multiferroic
in this selected group.

In terms of the tolerance factor (t), perovskite fluorides follows pretty well the
trend of the phase as a function of the t factor. For example, for values below t =
�.�� the ABF3 adopts the R�c structure, the Pnma symmetry can be found for t
between �.�� to �.��. The Pm3̄m in �.�� to �.�� range and P�3 /mcm from �.��
to �.�� respectively. There are a few exceptions due to the Jahn-Teller distortion
in KCuF3 , KCrF3 with a I�/mcm structure [��]. Besides, taking into account the
large change on the ionic radii of the A-site going from Li to Cs, it is expected
to observe a large change in the expected structure and then, the vibrational
landscape that suppress the ferroelectricity. A systematic study of the structural
instabilities in the high symmetry cubic halide perovskites was carried out in the
��’s using interionic potentials [��]. Phonon frequencies were calculated at the ,
X, M and R high symmetry points of the cubic Brillouin zone for ABX3 , with A
= Li, Na, K, Rb or Cs, B = Be, Mg or Ca, and X a halide (F, Cl, Br or I). While
none of these perovskites has a ferroelectric ground state —although some, mostly
Li members of the series have the non-perovskite ferroelectric LiNbO3 structure
as their ground state—, �� of the �� cases were found to show a ferroelectric
(FE) instability. The instability was usually absent in compounds with a larger
A cations, which anyway usually form in a non-perovskite hexagonal structure,
and so it was concluded that the Na compounds should be the most promising
candidates for ferroelectricity. It was found, however, that in all cases a competing
antiferrodistortive (AFD) zone boundary instability corresponding to rotations
 �.� computational details ��

and tilts of the X octahedra around the B cation, dominated over the weaker
polar instability, resulting in a centrosymmetric AFD ground state [��]. Such
competition between AFD and FE instabilities has been widely discussed in the
perovskite oxides literature [��, ��, ��, ��] and recently shown, in the case of
Pnma perovskites to result from a mutual coupling to an X+ 5 mode involving
antipolar displacements of the A-cations and X-sites along the [���] axis. Indeed
artificially removing the X+5 mode in the calculations for perovskite oxides was
shown to induce a ferroelectric ground state [��]. Thus, in the next sections we
are going to go further in this vibrational analysis in order to obtain a deeper
understanding of the structural properties in fluoroperovskites to be later used
in the “tuning” and study of their properties.

�.� Computational Details

Our density functional calculations of the electronic and structural properties
were performed using the VASP code [��, ��]. PAW [��] pseudo-potentials were
used to represent the valence and core electrons. The electronic configurations —
for valence and semicore electrons— taken into account in the pseudo-potentials
are presented in Table �.�. The exchange correlation was represented within the
GGA - PBEsol parametrization [��]. Due to the large d-electron localization, we
corrected the high correlation by means of the DFT+U method [��] (GGA+U,
U = � eV), with a U value converged for all magnetic compounds. Due to the
magnetic character of these systems, we have included the spin in the calculation
and make the proper use of the exchange correlation (LSDA+U ). The periodic
solution of these crystalline structures was represented by using Bloch states
with a Monkhorst-Pack k-point mesh of (�⇥�⇥�) and ��� eV energy cut-off,
which has been tested already to give forces convergence to less than �.���� eV/Å
and an error in total energy around ± �.� meV in all of the tested compound.
Additionally, vibrational modes were fully converged with respect to energy and
k-point to obtain values in error less than � cm-1 . Full Phonon-dispersion curves
were computed by calculating the dynamical matrix from the �⇥�⇥� supercell
and mapped to the unit cell. The Dynamical matrices and Born effective charges
were obtained through the DFPT formalism [��] as implemented in VASP and
post-processed in the Phonopy code [��]. Additionally, the phonon dispersion
curves were also interpolated from the interatomic force constants (IFCs) using
the anaddb module provided with the Abinit software [��].

�.� Ionic A- and B-site Radii Size Effects

Vibrational properties and stability were analyzed by DFT in fluoroperovskites
ABF3 . The analysis were carried out first in the family NaBF3 for B = Mn, V, Ni.
Additionally to that, some non-magnetic cases were also taken into account (i.e.
B = Ca, Zn) in order to study the ferroelectric behavior among some elements in
�� vibrational behavior in fluoroperovskites abf 3

Table �.�: Electronic configuration of pseudopotentials. Valences and semicore electrons

taken into account for calculations.
A-site B-site X-site
Li (�s2 �s1 ) Ca (�s2 �p6 �s2 ) O (�s2 �p4 )
Na (�p6 �s1 ) V (�p6 �d3 �s2 ) F (�s2 �p5 )
K (�p6 �s1 ) Mn (�p6 �d5 �s2 )
Rb (�p6 �s1 ) Ni (�p6 �d8 �s2 )
Cs (�s2 �p6 �s1 ) Zn (�d10 �s2 )

period IV. This with the aim to understand the possible influence of ionic radii
size and magnetic ordering in the structure and the vibrational behavior. The
later will help us to understand the ground state at the fluoroperovskite familie
by studying and knowing the nature of the displacive transitions posible in the
perovskite fluorides. In this type of transition, the modes that drives the trans-
formation are the same phonon modes driving the lattice dynamics. Therefore,
it is of high importance to understand the vibrational landscape of the materials
investigated on this thesis. For a better understanding of the nature of the dis-
placive transitions and the theory beneath it, the reader can check the references
from Dove [��] and Lines and Glass [��].
First at all, we start by defining the phonon frequencies and eigenvectors within
the harmonic approximation as:

X
0 0
D̃k↵,k 0 (q) mq (k ) = !2mq mq (k ↵) (�.�)
k 0,

where ↵ and are the cartesian coordinates of displacement,  and ’ are the
atoms of interest, !mq is the phonon frequency of a m mode at a wave vector q,
mq is the associated
p phonon eigenvector corresponding to the eigendisplacement
defined as: ⌘mq = M mq and finally, D̃k↵,k 0 is the dynamical matrix, formed
by the Fourier transform of the interatomic force constants matrix (IFC, C̃k↵,k 0 )
as:

X p
D̃k↵,k 0 (q) = C̃k↵,k 0 (q)/ Mk Mk 0 (�.�)
k 0,

In the real space the IFC matrix is defined in terms of the energy of the system
as follows:

@2 E
C↵, 0 (l, l 0 ) = 0)
, (�.�)
@⌧↵ (l)@⌧  0 (l
 �.� ionic A - and B -site radi i size effects ��

where E is the total energy, l and l’ are the cells of respective atoms and ⌧ is
the displacement. In Fig. �.� are presented the fully relaxed phonon-dispersion
bands for NaMnF3 computed in the high temperature symmetric cubic Pm3̄m
perovskite [��]. Structural instabilities (i.e. imaginary modes represented as neg-
ative frequency values) can be seen from the phonon-dispersion curve through
the entire Brillouin zone. A polar ferroelectric mode is observed at zone-center
-point that has been also found in all ferroelectric systems based on oxides such
as BaTiO3 , PbTiO3 , and antiferroelectric PbZrO3 [��].

X5+ M3+ R4+

Figure �.�: (Above) Phonon-dispersion curves for NaMnF3 computed at the Pm 3̄ m

phase. The Brillouin zone is presented in the inset. The imaginary modes
are plotted as negative values . Atomic displacement for each unstable mode.
(Below) Antipolar displacements, octahedral rotations and tilts (depicted by
arrows) can be observed for the X+ 5 , M 3 , and R 4 . modes.
+ +

Additionally, two instabilities with high intensity in frequency value were also
observed at R-point showing a flat branch along the R–M path characteristic of
the AFD modes reported in some oxides too. It is believed that this geometrical
instability competes with FE mode and, in some cases can completely suppress
�� vibrational behavior in fluoroperovskites abf 3

it in the ground state. Thus, all these instabilities combined leads most of the
NaBF3 compounds to an orthorhombic Pnma (��) ground state, in which, fer-
roelectricity is totally suppressed by symmetry. As an exception to the rule, the
NaCaF3 case has been reported in a rhombohedral R�c structure in ground state
symmetry with possible ferroelectric ordering [��]. Nevertheless, up to now there
is no experimental results corroborating the FE behavior in this material. The
phonon-dispersion bands for B = Ca, V, Ni, and Zn are similar to B = Mn and
because of that, are not presented here.
According to these vibrational results, we have found for all structures a clear
dependence of FE-instability as a function of the ionic B-site radii for B2+ state,
of course, taken into account the number of coordination (i.e. B-F6 bonding) [��].
Then, as the ionic B-site radii decrease, the FE-mode decrease as well as a function
of this parameter (see Table �.�). Additionally, we observed the same behavior
for one of two instabilities at R-point, meanwhile the second one remains almost
constant. In the same table we also presented the computed values for BaTiO3
for comparison with oxide systems.

Figure �.�: Polar 4- mode, into the Bethe’s notation [��], as a function of the A- and
B-site in ABF3 perovskites, the imaginary modes are presented as negative
values.

Now, we are going to talk about the A-site effect, for this proposes the B-site
was fixed to Ni case, this in order to keep small lattice parameters. From Table
�.� it can be appreciated that FE polar instability present a more critical depen-
dence with respect to the ionic A-site radii size than B- one. The later findings
are presented graphically in the Fig. �.�. Therefore, an increase of this parameter
can totally suppress the FE imaginary mode and AFD instabilities, which leads
the structure to a totally different crystal ground state symmetries as listed in
Table �.�. This behavior can be explained in terms of eigendisplacements, ⌘ (see
Table �.�) for this mode, in which the dominant contribution in the FE-mode is
precisely the A-site displacements. Increasing the A-site radii, the displacements
 �.� ionic A - and B -site radi i size effects ��

decreases as well and with that a strong reduction of the frequency it is appreci-
ated, meanwhile, the imaginary mode is suppressed for A = K, Rb, and Cs (see
Table �.�).
 ��

Table �.�: Summary of most unstable high symmetry point phonon eigenvalues, relaxed lattice constants for cubic Pm 3̄ m symmetry,
ionic A and B sites radii and reported ground state symmetry for all compounds [��]. HS: High Spin, LS: Low Spin. All modes
+ + + -1 .
in ( 4- , R+
4 , R 5 , X 5 and M 3 ) in cm
-
Compound ao [Å] B-radiia [pm] Godschmidt, t FE - 4 R+
4 R+
5 X+
5 M+
3 Ground State [��]

BaTiO3 �.���� ��.� �.�� ���i Non Non ���i ���i P�mm [���]

NaCaF3 �.���� ��� �.�� ���i ���i ���i ���i ���i R�c [��]
NaMnF3 �.���� �� (HS) �.�� ���i ���i ��i ��i ���i Pnma
NaVF3 �.���� �� �.�� ��i ���i ��i ��i ���i Pnma
NaZnF3 �.���� �� �.�� ��i ���i ��i ��i ���i Pnma
NaNiF3 �.���� �� (LS) �.�� ��i ���i �� ��i ���i Pnma
A-radii [pm]
CsNiF3 �.���� ��� �.�� ��� ��� ��� ��� ��� P�3 /mmc, z =�
RbNiF3 �.���� ��� �.�� ��� ��� ��� ��� ��� P�3 /mmc, z =�
KNiF3 �.���� ��� �.�� ��� �� ��� ��� �� Pm3̄m
NaNiF3 �.���� ��� �.�� ��i ���i ��i ��i ���i Pnma
LiNiF3 �.���� �� �.�� ���i ���i ���i ���i ���i R�c b
a Ions are �-coordinate [��]
vibrational behavior in fluoroperovskites abf 3

b Expected ground state

 �.� ionic A - and B -site radi i size effects ��

Table �.�: Eigendisplacements ( ⌘ ) for all compounds at relaxed cubic structure. X ?

and X k are the sites located perpendicular and parallel to the M-F bonding
respectively
⌘, (AB F3 ) [a.u.]
Compound A-site B -site X? Xk
BaTiO3 �.��� �.��� -�.��� -�.���
NaMnF3 �.��� -�.��� -�.��� -�.���
NaVF3 �.��� -�.��� -�.��� -�.���
NaZnF3 �.��� -�.��� -�.��� -�.���
NaNiF3 �.��� -�.��� -�.��� -�.���
CsNiF3 �.��� -�.��� -�.��� -�.���
RbNiF3 �.��� -�.��� -�.��� -�.���
KNiF3 �.��� -�.��� -�.��� -�.���
NaNiF3 �.��� -�.��� -�.��� -�.���
LiNiF3 �.��� �.��� -�.��� -�.���

As previously discussed, the �-fold degenerated eigendisplacements in FE-instabilities,

condensed in Table �.�, taken at the relaxed volume for each compound, reveal a
strong influence of the A-site in the FE-mode. The ionic B-site radii size have a
weaker influence in the FE-instability than the A-site. Additionally, the eigendis-
placements of F ions along the atomic movement (perpendicular to B-F bonding
direction) has an important contribution that decreases as ionic B-site radii does.
From the analysis of values obtained for different A-sites, we can also noted an
induced “transition”, from negative to positive, in FE-polar mode eigendisplace-
ments of the B-site (Table �.�), going from positive displacements in Li to a
negative ones at K, this means an opposite movement of Li against the B-sites
in comparison with the other compounds. The later change can also contribute
to the totally different structural stabilization symmetries for ground states in
A-site = Li, Na, K as listed in Table �.�.
Phonon-dispersion curves computed for A = K, Rb, and Cs do not exhibit
any instabilities at any point of Brillouin zone, suggesting a cubic symmetry as
a ground state for this compounds. Here, it is important to remark that the
previous conclusion is only true for KNiF3 compound which present a Pm3̄m
symmetry. For A = Cs and Rb cases the experimentally reported ground state
belongs to a hexagonal symmetry. As it is well known, there is no direct structural
transition path from cubic to hexagonal symmetries through octahedral tilting.
Then, a phase transition from cubic to hexagonal structure requires the break
of bondings at octahedral corner-shearing site. This fact explains the vibrational
behavior of these compounds at the assumed cubic structure as it has been taken
into account in our studies. We can now analyze the dependence of distorted
�� vibrational behavior in fluoroperovskites abf 3

structure perovskites as a function of the ionic radii ratio. We start from the
respectively tolerance factor (i.e. t, Table �.�) defined by Goldschmidt as follows:

rA + rX
t= p , (�.�)
2(rB + rX )

where rA , rB , and rX are the ionic radii for the A-, B- and X-sites respectively.
It can be noted that the Eq. �.� shows a close relationship between the tolerance
factor and ionic sites radii, which at the same time, it can be related to the ground
state and the FE/AFD instabilities ratio. Then, for smaller t values (i.e. LiNiF3
and NaCaF3 ) we observed that the FE frequency is higher in comparison to
AFD one. Here, this partial conclusion suggest that the FE freezing is sufficient
enough to lead these systems from the cubic high temperature structure to a
rhombohedral R�c ground state as previously reported for NaCaF3 but, to out
knowledge, found for a first time in LiNiF3 . This opens the possibility for a novel
G-type AFM multiferroic fluoride-based perovskite. For other compounds, an
orthorhombic Pnma structure is expected at tolerance factors that lies between
�.�� to �.��. From �.�� to �.�� a cubic Pm3̄m ground state is expected. At higher
values, above t = �.�� are anticipated different hexagonal symmetries. A few
exceptions such as KCuF3 and KCrF3 have been reported due to Jahn-Teller
distortions [��]. All these results are in full agreement with the our computed
vibrational properties listed in Table �.�.
In Fig. �.� are presented the results for the ferroelectric instability as a function
of the tolerance factor. Here, the LiBF3 and NaBF3 series are shown, in which, it
can be observed that the values of the frequency for the Li family are well above
��� icm-1 . These values are close to the ones observed in oxides perovskites. For
the Na family we found frequency values between �� to ��� icm-1 showing and
stronger dependence on the B-site radii.

Figure �.�: Ferroelectric instability in LiBF3 and NaBF3 (B = Ca, Mn, V, Zn, and Ni)
as a function of the tolerance factor, t.
 �.� ionic A - and B -site radi i size effects ��

Another useful parameter used to characterize the ferroelectricity in oxides, is

the calculation and analysis of the Born effective charges (Z⇤ ) tensor. In oxides,
the Z⇤ components exhibit anomalous values far from the nominal charges. The
later due to the covalent character of the B–O bonding and the dynamical charge
transfer widely studied and reported [���]. The Z⇤ tensor is defined as:

@P
Z⇤,↵ = ⌦ | , (�.�)
@⌧↵, E=0
where ↵ and are the cartesian coordinates, P is the polarization and E is the
macroscopic electric field. Our computed Z⇤ values for all fluorides in question
and BaTiO3 compounds are presented in Table �.�. For both, A- and B-sites
we found that Z⇤ s are close to nominal charge (i.e. A1+ , B2+ ). However, for
F-ions the Z⇤ are �.� e- away from its valence, which could be possibly due to
a weak covalent interaction in the F-(B-site)-F bonding. It is clear that these
values are lower than in the oxides case, but are strong enough to contribute
to the superexchange magnetic interaction. The later interaction responsible for
the antiferromagnetic ordering observed experimentally in most of perovskite
fluorides —with a magnetically active B-site— at their ground state structure
[��]. In the same table, the Z⇤ components for barium titanate are also presented.
Anomalous Z⇤ for Ti and O sites parallel to the bonding are observed as expected
due to the strong covalent contribution that allows the stabilization of the FE
ground state in this oxide. These Z⇤ close to nominal charges stand for another
substantial difference of the fluorides in comparison to oxides perovskites that
can be exploited.
In order to achieve a better understanding of the vibrational properties in
fluorides, we analyzed the interatomic force constants (IFC) calculated from Eq.
�.�. The analysis of IFCs reveals information about interatomic interactions that
are predominant in the vibrational instabilities and lead to a stable structure.
We present our calculated on-site force constants (on-site IFC) in Table �.�. The
on-site IFC are positive for all the elements in the ABX3 compounds, showing
that they are all stable under an isolated atomic displacement. Then, the phonon
instabilities observed in phonon dispersions and discussed above, only can be
removed or compensated by a cooperative movement of all atoms in order to
gain energy and lower the total enthalpy of the system. One exception inside the
analyzed compounds is the negative on-site IFC for Li atoms in LiNiF3 perovskite
suggesting a necessary collective movement of Li ions in lattice to reach a stable
state. This fact can explain the R�c crystalline group expected as ground state for
LiNiF3 in which a huge displacement and distortion of cubic structure is required
to achieve the structural transition. From the Table �.� can be also noted that B
= V and Ni presents the highest values of on-site IFC values for A, B, and F sites,
this fact suggest a low movement of self-site ions from its position in comparison
with the other compounds. The remarkable low value in the on-site IFC for Na in
NaCaF3 suggest a possible A-site distortion needed to stabilize the structure and
then, suppress the unstable phonon branch exhibited at zone-center. Additionally,
�� vibrational behavior in fluoroperovskites abf 3

Table �.�: Born Effective Charges (Z⇤ , e- units) in for all compounds computed at
relaxed cubic structure computed according to the Eq. �.�. The X? and Xk
are the components perpendicular and parallel to the B–X bonding.
Z⇤ (ABX3 ) [e- ]
Compound A-site B-site X? Xk
BaTiO3 �.��� �.��� -�.��� -�.���
NaCaF3 �.��� �.��� -�.��� -�.���
NaMnF3 �.��� �.��� -�.��� -�.���
NaVF3 �.��� �.��� -�.��� -�.���
NaZnF3 �.��� �.��� -�.��� -�.���
NaNiF3 �.��� �.��� -�.��� -�.���
CsNiF3 �.��� �.��� -�.��� -�.���
RbNiF3 �.��� �.��� -�.��� -�.���
KNiF3 �.��� �.��� -�.��� -�.���
NaNiF3 �.��� �.��� -�.��� -�.���
LiNiF3 �.��� �.��� -�.��� -�.���

it is observed a clear increase of on-site IFC for A-site in NaBF3 compounds as a

function of B-site radii, making more difficult the movement of this ion within the
structure due to the reduced volume. This conclusion is also consistent with the
trend of the FE instabilities at the zone center as a function of ionic radii presented
in Table �.�. For comparison, we also presented the values for two well known
ferroelectric oxides, BaTiO3 and PbTiO3 . For the first one, a high value of on-site
IFC in Ba reveal a stable position of this site making difficult its movement. This
is in full agreement with the ferroelectric mechanism in this compound which
strongly depend on the Ti–O–Ti chain displacement [���]. Instead of that, in
PbTiO3 the on-site IFC for Pb is low and it is related to the strong role of Pb
site in the ferroelectricity in lead titanate, in which Pb–O movements presents a
strong participation in the phenomena added to Ti–O–Ti chain displacements [��].
On-site IFC for fluorine sites presents a remarkable difference between Fx=y sites
(i.e. that lies in A-F plane) and Fz (i.e. that lies along B–F bonding) values. This
difference suggests a stable position for Fz interaction, however, the low value for
Fx=y indicate a possible movement of this sites inside of A–F planes which could
reinforce the BF6 tilting expected for the compounds. The later in comparison
with oxides also presented in Table �.�, in which no oxygen octahedral tilting is
expected.
 �.� ionic A - and B -site radi i size effects ��

Table �.�: On-site interatomic constants (on-site IFC, in eV/Å2 units) for the ABX3
flouroperovskites. The on-site IFC for the BaTiO3 and PbTiO3 are also pre-
sented for comparison. The Fx=y sites lies in A-F plane meanwhile the Fz lies
along B–F bonding.
On-site IFC for ABX3 [eV/Å2 ]
Compound A-site B-site X X
x=y=z x=y=z x=y z
BaTiO3 �.�� ��.�� +�.�� +��.��
PbTiO3 [��] �.�� ��.�� +�.�� +��.��
NaCaF3 �.�� ��.�� �.�� ��.��
NaMnF3 �.�� ��.�� �.�� ��.��
NaVF3 �.�� ��.�� �.�� ��.��
NaZnF3 �.�� ��.�� �.�� ��.��
NaNiF3 �.�� ��.�� �.�� ��.��
CsNiF3 �.�� �.�� �.�� �.��
RbNiF3 �.�� ��.�� �.�� �.��
KNiF3 �.�� ��.�� �.�� ��.��
NaNiF3 �.�� ��.�� �.�� ��.��
LiNiF3 -�.�� ��.�� �.�� ��.��

Figure �.�: Selected IFC interactions in the � ⇥ � ⇥ � supercell. A - B - and F-sites are
depicted in green, violet and grey respectively. Additionally, the atoms are
labeled to correlate the IFCs shown in the Table �.�.
 ��

Table �.�: Selected IFC according to interactions illustrated in the Fig.�.� for the A -, B -, and F-sites respectively.
IFC for ABF3 compounds one quarter of unit cell [eV/Å2 ]
Compound A1 -A3 A1 -A2 A1 -B1 B1 -B2 A1 -F1 B1 -F1
xx=yy zz k ? xy xx k ? xx=yy xy zz k ?
NaCaF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
NaMnF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� -�.�� +�.�� +�.�� -�.�� -�.�� -�.��
NaVF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
NaZnF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
NaNiF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
CsNiF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� -�.�� -�.�� -�.�� -�.�� -�.��
RbNiF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� -�.�� -�.�� -�.�� -�.�� -�.��
KNiF3 -�.�� -�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� -�.�� -�.�� -�.�� -�.�� -�.��
NaNiF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
LiNiF3 -�.�� +�.�� -�.�� +�.�� -�.�� -�.�� -�.�� +�.�� +�.�� +�.�� -�.�� -�.�� -�.��
vibrational behavior in fluoroperovskites abf 3
 �.� ionic A - and B -site radi i size effects ��

Figure �.�: Double-well energy obtained at the FE-mode freezing in NaBF3 compounds.
The gain in energy is in agreement with previous results that shows that the
bigger is the B-site, bigger is the FE instability and with this, the possibility
to be ferroelectric. ± denotes the full amplitude of the freezing.

Additionally to on-site IFC, some selected IFC for interactions were computed
within a quarter of a �⇥�⇥� supercell as labeled in Fig.�.� and presented in
Table �.�. The analysis were taken into account for components parallel and per-
pendicular to the direction of line that connects the related atoms. This analysis
was performed with the aim to study some selected interactions and analyze the
relationship between the ions that present the highest contribution in the total
energy. First, it is noted the significantly high repulsion force between (B1 –F1 )k
sites, which induce the movement of all B-sites with the fluorine ions instead of
against them, like in barium and lead titanates. The IFCs values as a function of
A-site radii shows that B–F repulsion interaction decrease when A-radii increase
as well, and then, a freezing of octahedra tilting and stabilization of cubic struc-
ture is achieved in the KNiF3 case. Besides, the B1 –B2 k IFCs values and the
high on-site IFC obtained for the center-sites suggest that B-ions cannot move
easily in comparison with the center-site displacement observed in some oxides
(e.g. barium titanate). Nevertheless, these B-sites are still presenting a small con-
tribution in the FE-mode that cannot be neglected as discussed above based on
eigendisplacements values. The contributions from another interactions such as
A–A and A–F listed in the Table �.� are smaller than those discussed previously
and do not present a considerable contribution.
Moving forward, we analyze the energy landscape of the ferroelectric mode. In
Fig.�.� are presented the double-well energy curves obtained through the freezing
of ferroelectric instabilities along z -axis. From this figure it is observed a clear
dependence of energy minimum as a function B-site, in which the higher value
of FE frequency corresponds to a lowest energy value. These results are in full
agreement with our previous analysis and are summarized in the next section.
In NaCaF3 case, the double-well analysis suggests a prominent predisposition of
�� vibrational behavior in fluoroperovskites abf 3

these materials to freeze the FE ground state over the AFD one due to the high
energy gain.

�.� Ferroelectricity Model Proposed for ABF3 Flu-

orides
Double wells in BaTiO3 and NaMnF3
Double wells in BaTiO3 and NaMnF3

BaTiO3
Geometric%FE%ins
(a) BaTiO3 because%no%ch
Geometric%FE%inst
(a) BaTiO3 BaTiO3
(a) BaTiO
(a) 3
transfer%and%smal
because%no%cha
in%a%big%cell%(4
(a) BaTiO3
transfer%and%small
in%a%big%cell%(4.1

NaMnF3
(b) NaMnF3
(b) NaMnF3
Figure 3.4: Double-Well energy curves(b)
Figure 3.4:Andrés
Double-Well energy
Camilo García curves
Castro,
and
NaMnF
and
(b)
NaMnF
schematic representation of atomic
3 3
schematic representation of atomic
NaMnF
[email protected]
displace-
displace-
ments in different models of ferroelectricity for oxides 3 a) BaTiO3 and b) NaMnF3 WVU-Physics
ments3.4:
Figure in different models
Double-Well of ferroelectricity
energy for oxides
curves and schematic a) BaTiO3 and
representation b) NaMnF
of atomic displace-
3
Figure 3.4:Andrés
Double-Well energy
Camilo García curves
Castro, and (b)
schematic representation of atomic displace-
[email protected]
ments in different models of ferroelectricity for oxides a) BaTiO3 and b) NaMnF3 WVU-Physics D
mentsFerroelectric
3.3 in different models of ferroelectricity
Instability for oxides a) BaTiO
Enhancement 3 and b)3NaMnF
in NaMF Fluorides
3
3.3 �.�:Ferroelectric
Figure Double-Well energy Instability Enhancement
curves and in NaMF3ofFluorides
schematic representation atomic displace-
An
3.3 interesting behavior
Ferroelectric of ferroelectricity
Instability and then,
Enhancement of the FE-instability
in NaMF as a function
Fluorides
mentsbehavior
An interesting in different models of
of ferroelectricity ferroelectricity
and then, of thein forNaMF
oxides 3a)Fluorides
3
FE-instability asBaTiO 3 and b)
a function
of 3.3 Ferroelectric
isotropic external strainInstability
has been found Enhancement
in some oxide based compounds like BaTiO
Anof interesting
isotropic external
NaMnF 3 . strain
behavior has been foundand
of ferroelectricity in some
then, oxide
of thebased compounds
FE-instability aslike BaTiO33
a function
and
AnPbTiO 3 [23,behavior
interesting 26], in which the FE ordering can be totally suppressedasbya function pressure.
ofand PbTiO
isotropic 3 [23, 26],
external strainin of ferroelectricity
which
has been the found
FE ordering and then,
in somein can of
oxidebethe FE-instability
totally
based suppressedlike
compounds by pressure.
BaTiO3is
Then,
Then,
this
of isotropic
this
physical
physical
phenomena
externalphenomena
strain has been widely
widely founddiscussedin someinoxide
discussed
oxides
oxides
but
based
but
less
compounds
less
known
known
in fluorides
like
in BaTiO3is
fluorides
As
and
and allPbTiO
PbTiO
proposed the
for previous
[23,
3 study
[23,
26],
in
26],
in
the discussion
in
which
framework
which
the
the and
FE
of
FE this analysis
ordering
research.
ordering
can
can suggest,
be totally
First,
be we
totally the
have mechanism
suppressed
studied
suppressed
by
the
by responsible
pressure.
effects
pressure. of
proposed
Then, this for studyphenomena
3
physical in the framework widely ofdiscussed
this research. in First,but
oxides weless
haveknown studied in the effectsisof
fluorides
for the
pressure ferroelectric
Then, this
pressure ininvibrational
physical
vibrational
mode
properties
phenomena is
properties
based
widely on
ofoffluorides-likethe
discussed in
fluorides-like
A -site movement,
perovskites.
oxides but less
perovskites. In
In figure
which
known3.5
figure in is
3.5
is strongly
presented
fluorides
is presented is de-
proposed for study in the framework of this research. First, we have studied the effects of
pendent
the square
proposed
the of of
square vacuum
foroffrequency
study inspace
frequency thefor or
FE-mode
framework
for free space
FE-mode instability
of for A
this research.
instability as -site
as displacements
aafunction
First,
function we ofof strain.
have studied
strain. related
From
Fromtheexpansionthen,
effects
expansion of to a
pressure
strain zone
pressure ininvibrational
can be
vibrational properties
observed that
properties ofthe
offluorides-like
square
fluorides-like of perovskites.
frequency
perovskites. for Inall
In figure 3.5
compounds
figure 3.5 is
is presented
converge
presented
steric
the effect.
strain
square zone Thus,
ofoffrequency a
can be observed decrease
for of
that the
FE-mode B -site radii,
square of
instability as reduce
frequency
aafunction the
for
of lattice
all compounds
strain. volume
From convergethen,
expansionand
almost
the to
square the same value.
frequency This
for could
FE-mode be a consequence
instability as of
function structural
of strain. distortions
From induced
expansion
reduce
almost
strain the
zoneto free
the
can space
same
be value.
observed for This
A -site
that could
the movement.
be a consequence
square of Also,ofthis
frequency factcompounds
structural
for all explains the
distortionsconverge strongly
induced
bystrain
bymixed
mixedzone can
phenomena be
phenomena observed
between
between that
ionic the
ionic square
charge
charge of
effectsfrequency
effects and
and for
ionic
ionic all
B-sitecompounds
B-site radii
radii converge
that
that needs
needs
correlated
almost
almostto
further to dependence
the
studies.thesame
same
For value.
value. ofThis
FE-mode
This could
could be beisas a function
aaobserved
consequence
consequence ofstructural
of
of A-site
structural radii, thus,induced
distortions
distortions a complete
induced
byfurther
mixed studies.
phenomena Forcompression
compression
between
strain
strain
ionic chargeis observed
effects
aanon-linear
non-linear
and ionic
increase
increase
B-site
of
of instability
radii instability
that needs
in
in
by mixed
opposite
opposite phenomena
direction
direction as as between
observed in
observed instrain ionic
BaTiO
BaTiO charge
3 [23] ineffects
which
[23] in which and
a ionic
suppresion B-site by
a suppresion by pressure radii
pressure that has needs
been
has been
further
furtherstudies.
studies. For compression
Forwas compression isis3observed aanon-linear
non-linear increase of
of instability in
demostrated.
demostrated. Also,
Also, was observed
observed astrain
a“transition”
“transition” observed ofofeigendisplacements
eigendisplacements increase instability
responsible
responsible of
of in
this
this
opposite
opposite direction
direction asas observed
observed in BaTiO
inNaNiF
BaTiO 3 [23]
3 [23]
in which
in which a suppresion
a suppresion by pressure
by pressure has been
has been
mode
mode (figure
(figure 3.6).
3.6).Inwas
Influorides,
fluorides, NaNiF 3 3case
casepresent
present aatotal
total suppression
suppression of
of FE-instability
FE-instability
demostrated.
demostrated. Also,
Also, was observed
observed aa“transition”
“transition” ofofeigendisplacements
eigendisplacements responsible
responsible of
of this
this
atat
external
external strain above
strainIn 2% as
above 2% asNaNiF observed for
observed for present BaTiO in the
BaTiO33 in the same same figure 3.5. This
This different
figure 3.5.FE-instability different
mode
mode (figure
(figure3.6).
3.6). Influorides,
fluorides, NaNiF3 3case case presentaatotal totalsuppression
suppression of of FE-instability
atat
external
externalstrain
strainabove
above2% 2%asasobserved
observedfor forBaTiO
BaTiO3 in inthe
thesame
same figure
figure 3.5. 3.5. This
This different
different
 �.� ferroelectric instability enhancement in NaBF 3 fluorides ��

suppression of the FE mode is observed for ionic radii beyond Potassium. Addi-
tionally, it is noted that F-ions present a strong participation in the FE-instability,
and also exhibit the same trend as a function of ionic B-site radii. Besides that,
the ferroelectric polar mode is less dependent on the covalent interaction present
along the B–F bonding, making this phenomena more stable than in oxides. In
order to corroborate this idea, double-well energy curves were computed for the
NaBF3 compounds (As discussed before and shown in Fig.�.�).
To compute this curves, the polar FE phonon-branches were frozen in order to
obtain the energy barrier for this mode. Additionally, we computed the curves
for the independent freezing of each A and B-sites eigendisplacements in fluo-
rides and BaTiO3 oxide. In Fig.�.�a it can be appreciated the energy curves for
which in the oxide case, the lowering in energy is induced by the B-site movement
as widely studied and expected. Any contribution of the A-site (i.e. Ba) was
observed according to the curves. In contrast, for the fluoride case —here choos-
ing the NaMnF3 as example— the largest contribution comes from the A-site
displacements (see Fig.�.�b), and no contribution of B-site was observed. The
later is in agreement with all the previous results that explains the A-site ge-
ometrically driven FE mode. In the same figures are also schematically shown
—with arrows— the atomic displacements for FE-modes in each case. As dis-
cussed above, we found different contributions from the A- and B- sites in the
fluoride and oxide cases respectively. This substantial difference is making this
phenomena completely characteristic of the fluoroperovskites in comparison to
oxides. In conclusion, all the results demonstrates an A-site geometrically-driven
FE-instability mechanism in fluorides rather than a B-site charge-transfer one as
in most of the oxides.

�.� Ferroelectric Instability Enhancement in NaBF3

Fluorides
Interestingly, in some oxide-based compounds —such as BaTiO3 and PbTiO3 —
has been found a particular behavior of ferroelectricity and then, of the FE-
instability as a function of isotropic external strain or pressure [���, ���]. In
these compounds the FE ordering can be totally suppressed by pressure. The
study of the later physical phenomena widely discussed in oxides, but less known
in fluorides, has been considered in the framework of this research. First, we have
studied the isotropic pressure effects in the vibrational properties of fluorides
perovskites. Fig.�.� shows the squared frequency for the FE-mode instability as
a function of isotropic pressure, here shown as expansion and compression cell
percentage with respect to the relaxed volume. From the expansion zone (i.e. neg-
ative pressure) it can be observed that the squared frequency for all compounds
converge almost to the same value, -��⇥��3 cm-2 . This could be a consequence of
structural distortions induced by mixed phenomena between ionic charge effects
and ionic B-site radii that needs further studies. For compression values (i.e. posi-
�� vibrational behavior in fluoroperovskites abf 3

tive pressure) it is observed a non-linear increase of the FE instability in opposite

direction to the values obtained for BaTiO3 [���] in which, a FE suppression by
pressure has been demonstrated. Our results also show a “transition” of eigendis-
placements responsible of this mode (Fig.�.�). We found an exception to the main
trend in the NaNiF3 case. This fluoroperovskite shows a total suppression of FE-
instability at positive pressure equivalent to -�%. The suppression of the FE is
in the same way as observed for BaTiO3 (see Fig.�.�). The different trend for
nickel-based compound will be discussed in more detail later. From Fig.�.� it is
also observed that the increase of the instability as a function of pressure also
present a clear dependence of the ionic B-site radii size, which is higher for Ca
and lower for Zn. It is important to remark that close to the relaxed cubic lattice
parameter value (i.e. P ⇡ �), the slope in the !2 vs. P line is small in comparison
with the one observed BaTiO3 . This facts indicates a strong stability of the ferro-
electric instability as a function of pressure in the ABF3 compounds. The analysis
of the electronic structure as a function of pressure do not show an influence of
unoccupied or occupied d-levels in the B-site and/or an electronic hybridization
between F (�s2 ) and Na (�p6 ) as demonstrated in some oxide perovskite systems
at high pressure [��].

Figure �.�: Unstable FE mode at the cubic Pm 3̄ m structure in ABX3 compounds.

In order to go deeper in the understanding of the FE-instability behavior for

all compounds, except NaNiF3 discussed later, we carried out an analysis of the
eigendisplacements and eigenvectors for the polar ferroelectric modes. In Fig.�.�
we presented the eigendisplacements as a function of lattice parameter variation
(i.e. hydrostatic pressure). Most of the fluoroperovskites exhibit the same qualita-
tive behavior, then, only the NaCaF3 case is presented. We found a predominant
 �.� ferroelectric instability enhancement in NaBF 3 fluorides ��

influence of the A-site for pressure regions located close to the relaxed lattice and
expansion strain. In the region of constrain values, below -�% with respect to the
relaxed volume in cubic Pm3̄m structure, the influence of the A-site decreases
and the mode is dominated by B-site displacement. We found that the eigendis-
placements behave similarly in all of the NaBF3 materials, with the exception
of B = Ni. To reinforce this change in A- and B-sites displacements, the F-sites
also present a ⌘z transformation. For ⌘z parallel to B–F bonding, a “transition” of
eigendisplacements from negative, at expansion strain values, to positive values
are observed beyond strain pressures of -�%. The eigendisplacements for F-site
perpendicular to B–F bonding remains negative but goes from ⌘z = -� to ⌘z =
-�� a.u. indicating an increase on their movement amplitude. The eigendisplace-
ments are schematically represented in the Fig.�.� for two regions of strain. The
transformation on the nature of the eigendisplacements is possibly the cause for
the frequency variation as a function of pressure observed in BaTiO3 (Fig.�.�).
In this oxide, the same change in ⌘ is observed for the ferroelectric mode at the
FE-mode suppression when the hydrostatic pressure is applied.

Strain < 0%

Strain > 0%

Figure �.�: Eigendisplacements for NaCaF3 compound showing the ⌘ transition also
found in B -site: V, Mn, and Zn (left). Schematic representation of eigendis-
placement in the fluorides compounds for strain values above and below of
the relaxed structure (right).

In Table �.� we presented the on-site IFC values for all NaBF3 compounds
computed at a fixed volume that is equivalent to an expansion pressure in all
of them. The later analysis in order to understand the B-site radii effect on the
IFC. It can be noted that the A-site IFC constants are negative and close to zero,
this indicating a possible uniform displacement of these sites required to lower
the total energy. The on-site IFC values for F-sites increase as the B-site radii
decrease, going from Ca to Ni. Therefore, the cubic lattice volume reduction leads
to a lower F free space. The Fz values decrease with B-site radii, which implies a
reduction of the attractive interaction force allowing a larger Fz -site displacement.
The close value observed for on-site IFC site of A-ions could be the reason of the
�� vibrational behavior in fluoroperovskites abf 3

Figure �.�: Eigendisplacements for BaTiO3 showing the change of the mode as cause
of suppression of ferroelectric instability as a function of isotropic pressure.

observed concurrence of the FE-mode frequency at expansion strain, taken into

a count that this site dominates the ferroelectric mode in ABF3 structures.

Table �.�: On-site IFC for all NaMF3 compounds at same lattice parameter equivalent
to expansion.
On-site IFC [eV/Å2 ], a = �.�� Å
Compound A-site B-site F F
x=y=z x=y=z x=y z
NaCaF3 -�.�� +��.�� +�.�� +��.��
NaMnF3 -�.�� +�.�� +�.�� +�.��
NaVF3 -�.�� +��.�� +�.�� +�.��
NaZnF3 -�.�� +�.�� +�.�� +�.��
NaNiF3 -�.�� +�.�� +�.�� +�.��

On-site IFCs as a function of pressure are presented in Fig. �.�� for each site.
It can be seen that for a certain value in each case, between -�% to -�%, each
of the Fx=y constants exhibit a “transition” from positive to negative valued,
which implies a change of interaction from attraction to repulsion respectively for
these atoms. This transformation in the nature of interaction possibly induces
some instabilities responsible for the increment of the FE-mode frequency as a
function of pressure observed in the Fig.�.�. It can be also noted that on-site
IFC values for NaNiF3 remain positive, this could explain the different behavior
of this compound in comparison with the rest of the members of this family
discussed above. Additionally, A-, B- and Fz constants are quite larger for B =
Ni in comparison with the Ca, Mn, V, and Zn.
 �.� ferroelectric instability enhancement in NaBF 3 fluorides ��

Figure �.��: On-site IFC for A -, B - and F-sites (x=y, and z ) as a function of strain for
each of sites in Na B F3 with B = Ca, Mn, V, Zn, and Ni.

�.�.� Modes Contribution in Pm3̄m to Pnma Transition in NaBF3

Fluorides

The analysis and study of vibrational properties allow us to understand the nature
of the ferroelectric behavior in perovskite compounds. Based on that, a deeper
analysis was carried out in order to understand the coexistence of ferroelectric,
antipolar, rotational and tilting distortions in the NaBF3 family of fluorides. First
of all, it can be noted that in fluorides the contribution of the antipolar X+ 5
mode is higher than in oxides (compare Fig.�.�� with the values reported in
Ref. [��]). This mode is related to antipolar distortions dominated by the A-site
eigendisplacements. The R+ 5 mode contribution can be appreciated as well which
is almost missing in oxides [��]. R+
5 mode is also related to the A-site motion. M2
+

mode present a smaller effect and, in the range of scale it cannot be observed.
The R+ 4 mode and M3 are also present as in oxides. This fact clearly modifies
+

the balance picture between the contribution of modes to distortion when going
from Pm3̄m to Pnma.
Here, the picture of the competition of modes is more interesting because the
polar ferroelectric instability —present at -point—, the antipolar mode —at
X-point— and the antiferroelectric mode —at R-point— depend as well to the A-
site displacements. Despite the picture in oxides, where the ferroelectric instability
is leaded by B-site displacement meanwhile the A-site dominated antipolar X+ 5
mode suppresses the ferroelectric one in the Pnma symmetry as demonstrated
by Benedek and Fennie [��].
�� vibrational behavior in fluoroperovskites abf 3

Figure �.��: Contribution of modes in the phase transition from Pm 3̄ m to Pnma in the
Na B F3 fluoroperovskites.

�.�.� Special cases: LiNiF3 and NaNiF3

For NaNiF3 , we found a total suppression of FE instability as a function of exter-

nal pressure as shown in Fig.�.�. In Fig.�.�� is presented the squared frequency
for all the unstable modes as a function of external pressure. A total suppression
of FE-instability at is appreciated for strains above -�% with respect to the
cubic relaxed lattice parameter. Additionally, the same behavior was observed
for one of the instabilities located at R-point. The suppression of the instabilities
is clearly different in contrast to the behavior observed for the others NaBF3
compounds previously mentioned. We noticed that for this case, Ni posses the
smallest ionic radii in comparison with all others compounds that belong to the
fluoroperovskites family considered in this work.
Fig.�.�� depicts the eigendisplacements as a function of hydrostatic pressure
for three of polar modes at the zone-center -point including the FE-instability.
For this mode, no transition or considerable change in eigendisplacements were
observed. While for the others two modes, a quasi-constant dependence was ob-
served. Then, no anomalous trend responsible for transition of FE-mode was
appreciated.
We also analyzed the Born effective charges as a function of strain. A decrease
of Z⇤ for Ni and F ions was observed at high expansion strain values going from
�.�� e- (in expansion) to �.�� e- (in compression) and -�.�� e- to �.�� e- for
Ni and F respectively. A small fluctuation in values can be appreciated close to
the eigendisplacements transition pressure.
Then, up to this point, some questions remain open around this compound:
Why this compound is the only one to present this behavior? Is there a different
 �.� ferroelectric instability enhancement in NaBF 3 fluorides ��

Figure �.��: Behavior of FE and R-point instabilities as a function of hydrostatic pres-

sure for NaNiF3 .

Figure �.��: Eigendisplacements for all polar modes in NaNiF3 .

covalence bonding nature evidenced by the deviations in Z⇤ ? Further studies are

needed to find the origin of this out-of-trend behavior.

Finally, to conclude the ABF3 analysis, we stress that the ionic A and B site
radii size have a direct influence on FE-instability. From the theoretical analysis
carried out and discussed above, the LiNiF3 compound was chosen due to its
higher FE-instability. This compound presents a relaxed cubic lattice parameter
of �.�� Šand its ferroelectric instability is higher than AFD-modes. This fact
could lead to a stable ferroelectric ordering at the relaxed ground state struc-
ture, in which a R�c crystal symmetry type —like in LiNbO3 — is expected as
suggested by Claeyssens et al. [���]. In Fig.�.�� are presented the analysis of in-
stabilities as a function of strain at the Brillouin zone-center and boundary zone.
For strain values around �% of the relaxed volume, the FE-instability shows the
highest frequency value but below -�% the AFD-instability becomes dominant
in the vibrational landscape for this compound. The eigendisplacements for the
�� vibrational behavior in fluoroperovskites abf 3

FE-mode as a function of strain (see Fig.�.��) show an A-site dominated mode at

the relaxed structure. For hydrostatic pressure below the unconstrained volume,
the influence of A-site decreases non-linearly and mode begin to be dominated
by B-site movement. A sign change of the eigendisplacements was found at high
strain around -�%. This behavior is similar such as the one observed previously in
NaBF3 . It is important to remark that for all compounds a high stability of ⌘ as
a function of external expansion and compression strain around relaxed volume
has been found.

Figure �.��: Squared frequency at the vibrational instabilities. We can observe the FE-
mode at and AFD-modes at R -point as a function of hydrostatic strain
for LiNiF3 (left). Eigendisplacements of FE-mode instability in LiNiF3 are
also shown (right).

As future work an analysis is necessary in order to understand the freezing of

the FE-modes, and then, to corroborate the ground state of the system and its
possible ferroelectricity.

�.� Conclusions
Vibrational analysis were carried out in fluoroperovskites ABF3 . Different behav-
ior for in NaBF3 in comparison with oxides (e.g. BaTiO3 ) was found for B =
Ca, V, Mn, and Zn. For these compounds we found an increase of ferroelectric
instability as a function of hydrostatic strain, this possible due to the “transition”
of eigendisplacements responsible for polar mode. The increase of ionic A-site
radii present a strong influence in FE-polar instability, generating totally differ-
ent ground states symmetries for all compounds going from Li to K. The com-
pound NaNiF3 present a quite different trend in comparison with fluoride family,
for this case, we observed a suppression of FE instability. However, it was not
found change in mode eigendisplacements. The compound LiNiF3 reveal a appar-
ently predisposition to ferroelectric behavior with a strong FE-mode instability
in comparison with the others instabilities observed along M–R branch in the
phonon-dispersion curve. A model for an A-site geometrically driven ferroelec-
tricity in fluorides is proposed according to results and analysis. Some questions
 �.� conclusions ��

remain unclear such as: what is the mechanism for cancellation of FE mode at
the ground state? How can be tuned the polar mode in a fluoride structure to en-
gineer a ferroelectric fluoroperovskite? Some of these questions will be addressed
in the next chapters.
 Part III

STRAIN AND PRESSURE EFFECT IN

FLUOROPEROVSKITES

Taking advantage of the vibrational analysis performed in fluoroper-

ovskites, the pressure and biaxial strain was tested and the results
are accordingly presented in the following chapters. The perovskite
to post-perovskite (pPv ) transition in fluoroperovskites will be dis-
cussed in the following two chapters including a detailed analysis on
the non-collinear magnetic behavior of the pPv. Afterwards, the epi-
taxial strain effect on the multifunctional properties is discussed and
the successful engineering of a multiferroic/magnetoelectric in a fluo-
roperovsite is described.
PREDICTION OF THE pPv
3
TRANSITION IN NaMnF3

Perovskite materials with stoichiometry ABX3 are well known in material sci-
ences due to their numerous properties and their high interest for technological
applications. These properties are very diverse such as ferroelectricity [��], ferro-
magnetism [���], piezoelectricity [���], pyroelectricity [���] or superconductivity
[���]. Going further, recent researches focused on the achievement of multifunc-
tionalities in a single material that would open the door to new types of applica-
tions such as computer memories or spintronics [���, ���].
One of the main features of these compounds is that their properties can be con-
veniently tuned by external control parameters such as electric and/or magnetic
fields, temperature, strain or pressure [���, ���]. Hydrostatic pressure appeared to
be an interesting external parameter to be considered to enhance our understand-
ing of the perovskite physics, since they exhibit profuse structural, electronic and
magnetic phase transitions [���] under this parameter. A particularly appealing
transition is the perovskite to post-perovskite (pPv ) structural phase transition
(see Fig. �.�a) that has been observed in MgSiO3 crystal in ���� [���, ���]. The
discovery of this pPv phase transition have important consequences in the physics
of the earth mantle where MgSiO3 is abundant, under very high pressure con-
ditions [���, ���]. The pPv transition has then attracted great interest and it
has been observed in other oxide and fluoride perovskites such as MgGeO3 [���],
CaSnO3 [���], NaMgF3 [���] or NaCoF3 [���, ���] (see Table �.� for more details).
Interestingly, the fluoroperovskite crystals have a lower pressure and tempera-
ture pPv phase transition than the oxides and thus making them attractive for
the study of the pPv phase. Furthermore, it has been reported the possibility
of keeping the pPv phase at atmospheric pressure after releasing the pressure
in fluroroperovskites [���], while all the oxides return to their Pnma distorted
perovskite ground state [���].
Besides the interest for the pPv transition in earth mantle studies, we have
recently highlighted the existence of a ferroelectric instability in the cubic phase
of ABF3 perovskites that originates from a geometric effect rather than a dynam-
ical charge transfer, as is it the case in oxides —for more details see Chapter �
and Ref. [�]—. It was demonstrated that ferroelectric polarization in NaMnF3 is
induced by applying an epitaxial strain, an observation that leads to new possibil-
ities of engineering multiferroic materials within the family of fluoroperovskites.
This makes NaMnF3 a good candidate for multifunctional properties based on its
particular behavior as a function of external constraints. Additionally, in the fam-
ily of �d metal transition pPv fluorides, NaMnF3 remains unexplored along the

��
�� prediction of the p p v transition in n a m n f 3

pPv transition. The study of NaMnF3 under pressure is then of high interest for
scrutinizing the origins of its unique properties and its potential multifunctional
character.

Table �.�: Lattice parameters of the pPv phase (a, b, and c) and related temperature
(T) and pressure (P) transition parameters from the Pnma phase to the pPv
phase of fluoride and oxide perovskites. In the pPv phase, the A and B cations
occupy the 4a and 4c Wyckoff positions and the O/F anions occupy the 4c
and 8d positions. Most of the oxides have higher pressure and temperature
transitions than the fluorides.
Crystal a/b/c [Å] P [GPa] T [K] Ref.
Oxides ABO3
MgSiO3 �.���/�.���/�.��� ��� ��� [���]
MgGeO3 �.���/�.���/�.��� �� ��� [���]
MnGeO3 �.���/�.���/�.��� �� - �� ���� - ���� [���]
CaSnO3 �.���/�.���/�.��� �� ��� [���]
CaIrO3 �.���/�.���/�.��� �-� ���� - ���� [���, ���]
CaRuO3 �.���/�.���/�.��� �� - �� ���� - ���� [���, ���, ���]
CaPtO3 �.���/�.���/�.��� � ���� [���, ���]
CaRhO3 �.���/�.���/�.��� �� - �� ���� - ���� [���, ���]
NaIrO3 �.���/��.���/�.��� �-� ���� [���]
Fluorides ABF3
NaMgF3 �.���/�.���/�.��� �� - �� ��� - [���]
NaZnF3 �.���/��.���/�.��� �� - �� ��� - [���]
NaFeF3 �.���/��.���/�.��� � ��� - [���]
NaCoF3 �.���/��.���/�.��� �� - �� ��� - [���, ���]
NaNiF3 �.���/��.���/�.��� �� - �� ���� - ���� [���, ���]
NaMnF3 �.���/�.���/�.��� � � [�]

In this chapter we present the results on the structural, electronic and magnetic
properties of NaMnF3 as function of applied hydrostatic pressure. The results are
presented as follows. We first determine the structural ground state structures
over a wide range of pressure conditions and we predict a transition from the
Pnma ground state to a pPv phase that has not been previously explored exper-
imentally. We analyse the pPv phase transition through a vibrational character-
ization and a structural study. We then focus on the analysis of the magnetic
behaviour with a special interest on the non-collinear magnetic ground states
allowed in the pPv. We found a high-spin ground state associated with a large
ferromagnetic canting that has not been reported in any of the previous pPv
 �.� computational details ��

transition observed in other compounds. The validity of the results are discussed
with respect to the values of the Coulomb, U and exchange, J parameters within
the DFT+U approach.

�.� Computational Details

Total energy calculations were performed with the VASP code [��, ��] and within
PAW method to describe the valence and core electrons.[��] The electronic config-
urations taken into account in the PAW pseudopotentials are presented in Table
�.�. We used the GGA exchange correlation functional within its PBEsol vari-
ant [��] and corrected it to increase the d-electron localization by means of the
DFT+U approximation, with U = �.� eV and J = �.� eV within the Liechten-
stein formalism [��]. Due to the magnetic character of NaMnF3 , the spin degree
of freedom have been included in the calculations. Non-collinear magnetism calcu-
lations have been carried out by including the spin-orbit coupling as implemented
in VASP by Hobbs et al.[��]. Different values of the U and J parameters were
systematically varied to study the influence of these parameters on the results.
The reciprocal space has been discretized by a Monkhorst-Pack k-point mesh of
(�⇥�⇥�) and the plane wave expansion has been limited by an energy cut-off
of ��� eV. These convergence parameters were necessary in order to have a res-
olution on the force of less than �⇥��-4 eV/Å and a resolution on the energy
of �.�� meV. Analysis of the optimal structure as a function of pressure was
carried out according to the third-order Birch-Murnaghan equation of state (BM-
EOS) [���] by fitting the total energy dependence of the primitive cell volume per
molecule. Vibrational properties were computed through the formalism of density
functional perturbation theory (DFPT) [���, ���] as implemented in VASP and
post-processed with the Phonopy code [��].

�.� Structural and vibrational properties of the

Pnma phase of NaMnF3
At room temperature and atmospheric pressure, the sodium manganese fluoride
compound NaMnF3 crystallizes in the Pnma structure (space group number ��),
which is composed by tilted Mn–F6 octahedra along the three cubic directions and
antipolar motions of the Na cations [���]. This is the result of the condensation of
zone-boundary antiferrodistortive unstable modes observed in the cubic reference
structure phonon-dispersion curves [�]. By performing the structural relaxation
of the atomic positions and of the cell parameters of NaMnF3 in the Pnma
phase, we obtain the lattice parameters a = �.��� Å, b = �.��� Å and c = �.���
Å. This corresponds to a difference of �.�% with respect to the experimental
measurements reported by Daniel et al. [���].
�� prediction of the p p v transition in n a m n f 3

In Tables �.� and �.� we report the calculated silent, Raman and infra-red
(IR) phonon modes as well as the mode Grüneisen parameters ( i ) of the relaxed
Pnma ground state. The irreducible representation of the Pnma perovskite phase
is =�Au ��B1u �B2u ��B3u �Ag �B1g �B2g �B3g . Both the
Raman (Ag , B1g , B2g and B3g labels) and IR modes (B1u , B2u and B3u labels)
are in good agreement with the available experimental reports [���, ���]. We
note that a very soft polar B2u mode is present with a frequency of �� cm-1 .
This indicates that the Pnma phase is close to a ferroelectric phase transition
and this mode is responsible for the epitaxial strain induced ferroelectricity of
NaMnF3 as we reported in the Refs. [�, �] and explain in detail in Chapter
�. From the Grüneisen parameters reported in Table �.�, we observe that two
phonon modes are particularly sensitive to a change in volume: the soft B2u
mode at �� cm-1 with = �.��� and the Ag mode at �� cm-1 with =
�.���, both can be easily detected from Raman experiments. Phonon modes,
corresponding to a Au symmetry (at �� cm-1 and ��� cm-1 ), have negative
and correspond to the modes that become soft as the volume decreases, which
indicates that probably these two phonon modes are close to a nearest neighbour
force instability. Besides, these modes can contribute to a competition between
positive and negative thermal expansion of the material, depending on how the
phonon modes are populated [���].
In regards to the magnetic ordering, the total energy differences between the
ferromagnetic (FM), C- and A-type antiferromagnetic (AFM) with respect to
G-type AFM ordering in the Pnma structure are respectively ��.��, ��.�� and
��.�� meV/u.c. The lowest energy is thus given by the G-type AFM ordering,
in agreement with the experimental measurements that establish a G-type AFM
ordering for the Pnma phase of NaMnF3 [���]. All of that confirms that our
choice of parameters for the DFT calculations (GGA PBEsol and U = �.� eV) of
NaMnF3 gives good results when comparing to the experimental measurements.

�.� Study of NaMnF3 under hydrostatic pressure

In this section we study the pressure phase transitions of NaMnF3 . In order to
identify the enthalpy ground state under hydrostatic pressure, we fitted the total
energy versus the volume through the Birch-Murnaghan equation of state (BM-
EOS) [���] for the following common structures found in perovskites: Cmcm (pPv
phase), I�/m, I�/mcm, P�/mbm, P�1 /c, P�1 /m, P�3 cm, P�3 /mmc, P1̄, Pm3̄m,
and Pnma. For all of them, the G-type AFM magnetic ordering was considered
in our calculations. In Fig. �.�a we report the total energy values per atom as
a function of the volume change with respect to the equilibrium volume. As we
would expect, the Pnma and Cmcm-pPv structures exhibit the lowest energy
values at low volumes. In order to identify the lowest energy phase, we report the
enthalpy (H), as obtained from the equation of state, versus the pressure in Fig.
�.�b.
 �.� study of NaMnF 3 under hydrostatic pressure ��

Table �.�: Calculated Raman, Infra-red (IR) and silent modes frequencies (cm-1 ) of
the Pnma phase of NaMnF3 (at zero pressure). In brackets are presented the
experimental values of the Raman [���] and the IR [���] modes.
Raman modes
Ag B1g B2g B3g
�� (��) �� (—–) ��� (���) �� (��)
��� (���) ��� (���) ��� (���) ��� (—–)
��� (���) ��� (���) ��� (���) ��� (—–)
��� (���) ��� (���) ��� (���) ��� (���)
��� (���) ��� (���) ��� (—–) ��� (—–)
��� (���) — ��� (���) —
��� (���) — ��� (—–) —
Silent IR modes
Au B1u B2u B3u
�� (—–) ��� (���) �� (—–) �� (���)
��� (—–) ��� (���) ���(���) ��� (���)
��� (—–) ��� (���) ��� (���) ��� (���)
��� (—–) ��� (���) ��� (—–) ��� (���)
��� (—–) ��� (���) ��� (—–) ��� (���)
��� (—–) ��� (���) ��� (—–) ��� (���)
��� (—–) ��� (���) ��� (—–) ��� (���)
��� (—–) ��� (—–) — ��� (—–)
— ��� (—–) — ��� (—–)

Going from the lowest to the highest pressure, we found the following low en-
thalpy phases. Between -�� GPa and -�.� GPa the hexagonal P�3 /mmc phase is
the most stable structure. Between -�.� GPa and -�.� GPa we found the P�1 /c
structure to be the most stable. Then between -�.� GPa and �.� GPa the or-
thorhombic Pnma structure is the ground state and thus passing through the
atmospheric pressure as expected from the experimental results. At higher pres-
sures (above � GPa, up to �� GPa) the orthorhombic Cmcm pPv structure is the
lowest energy phase. Then, we found that NaMnF3 presents a phase transition
at relatively low pressure (� GPa) from the Pnma perovskite phase to the Cmcm
pPv phase.
In Fig. �.�a we represent a schematic view of the Pnma and the pPv phases.
It is clear that the structural transition between the Pnma and the pPv phase is
reconstructive since a breaking of Mn–F–Mn bonds is needed to connect the two
�� prediction of the p p v transition in n a m n f 3

Table �.�: Calculated mode Grüneisen parameters ( i ) for the NaMnF3 Pnma phase
at and around atmospheric pressure.
Ag B1g B2g B3g Au B1u B2u B3u
�.��� �.��� �.��� �.��� -�.��� �.��� �.��� �.���
�.��� �.��� �.��� �.��� -�.��� �.��� �.��� �.���
�.��� �.��� �.��� �.��� �.��� �.��� -�.��� �.���
�.��� �.��� �.��� �.��� �.��� �.��� �.��� �.���
�.��� �.��� �.��� �.��� �.��� �.��� �.��� �.���
�.��� — �.��� — �.��� �.��� �.��� �.���
�.��� — �.��� — �.��� �.��� �.��� �.���
— — — — �.��� �.��� — �.���
— — — — — �.��� — �.���

structures. This corresponds to a transition between corner-shared octahedra in

the Pnma phase to edge-shared octahedra in the pPv phase. The pPv phase can
be viewed as layers of edge-shared octahedra separated by layers of Na cations
along the y-direction. In Fig. �.�b we report the calculated X-ray diffraction
patterns (XRD) for the Pnma and the post-perovskite Cmcm phases with a
Cu-K↵ = �.���� Å X-ray wavelength. The simulated XRD pattern of the room
temperature Pnma is in good agreement with the experimental reports shown as
a triangles in Fig. �.�b [���, ���, ���]. No experimental data exists for the pPv
phase of NaMnF3 but the calculated XRD pattern fit the pPv symmetry found
in other compounds such as MgGeO3 [���].
An interesting parameter to be analysed in the Pnma to pPv phase transition is
the dependence of the octahedral tilting angle of the Pnma structure as function
of pressure. It has been reported that the transition from the Pnma to the pPv
structure takes place when the octahedra tilting angle reaches �� in A2+ B4+ O3
[���] and Na1+ B2+ F3 [���] compounds, where the octahedral angle has been
calculated according to O’Keeffe et al. [���]. This criteria is structurally based
on the relationship between AX12 and BX6 polyhedral volumes, (VA and VB
respectively), which establishes that the VA /VB ratio decreases with pressure due
to the smaller compressibility of the polyhedral VB with respect to VA . Thus, a
critical ratio is reached where the Pnma perovskite becomes structurally unstable
and a transformation to a post-perovskite structure is observed. This critical
volumes ratio is reached when the octahedral BX6 tilting angle is equal to ��
[���]. Additionally, this condition is achieved in materials for which the tilt angle
is at least �� at ambient pressure in the Pnma phase [���, ���]. For this purpose,
we computed the pressure dependence of the octahedra titling angle of the
Pnma structurepfollowing to the relationship deduced by O’Keeffe et al. [���] such
as = cos-1 ( 2c2 /ab) where a, b and c are the optimized cell parameters for
 �.� study of NaMnF 3 under hydrostatic pressure ��

(a)

(b)

Figure �.�: (a) Total energy of NaMnF3 as a function of the volume per atom for all
different crystal structures. Cmcm and Pnma structures show the lowest
energy values for lower volumes. (b) H/atom as a function of external
pressure in NaMnF3 structures. Pnma to Cmcm structural transition can
be seen at the critical pressure of � GPa.

an specific pressure. We report our results as a function of pressure for NaMnF3

and the experimental data for NaBF3 compounds with B = Zn, Ni, Co, Mg,
and Mn in Fig. �.�c. The experimental results for B = Zn, Ni, Co and Mg are
taken from Yusa et al. [���]. Experimental data for lattice parameters used in
the calculation of were extracted from the report of Katrusiak et al. [��]. As it
can be appreciated, all the experimental measurements follow the tilting criterion
proposed by O’Keeffe, i.e. a pPv phase transition at =�� . The experimental
results for NaMnF3 (filled squares) are in good agreement with our calculations
(open squares) up to �.� GPa (maximum pressure explored experimentally). Our
calculations predict that the tilt angle reaches �� at � GPa, where the transition
�� prediction of the p p v transition in n a m n f 3

(a)
P

(b)

Intensity [a.u.]

Pnma

Cmcm

10 20 30 40 50 60 70 80
!-2! [deg]
(c)

Figure �.�: (a) Schematic view of the phase transition from Pnma to Cmcm. Na, Mn and
F atoms are represented in green, violet and grey colors respectively. The
breaking of bondings for the transition from corner-shared to edge-shared
octahedra can be appreciated. (b) Simulated XRD patterns for Pnma and
Cmcm structures. The room temperature Pnma phase is in good agreement
with the experimental reports (open triangles) [���, ���, ���]. (c) Octahe-
dra tilt angle as a function of pressure in NaBF3 . For B = Zn, Ni, Co and
Mg experimental results are taken from Yusa et al. [���]. All of these com-
pounds present the transition from the Pnma to the Cmcm structures at a
tilting angle close to �� . Experimental results for NaMnF3 at low pressure
were taken from Katrusiak et al. [��] Full agreement between experimen-
tal (filled squares) and our theoretical calculations (empty squares) can be
appreciated.
 �.� study of NaMnF 3 under hydrostatic pressure ��

should occur according to the O’Keeffe rule, which is indeed in agreement with
our BM-EOS prediction of � GPa as the critical pressure.
Our theoretical findings for the transition between the Pnma and the Cmcm
pPv phase do not follow the recent experimental report of Akaogi et al. [���]. In
their pressure measurements at high temperature (T = ���� K), Akaogi et al. do
not observe a transition from the Pnma to the pPv structure up to �� GPa but
found, instead, a MnF2 + Na3 Mn2 F7 phase decomposition around � GPa. We
note that Katrusiak et al. also report a transition from the Pnma to the cubic
phase Pm3̄m at Tc = ��� K and ambient pressure [��] for this compound, which is
not observed by Akaogi et al. at very high temperature (higher than Tc = ���K).
We suggest that at the very high temperature where Akaogi et al. performed their
measurements, it is indeed possible that a phase decomposition appears as they
do not observe the pPv transition. We also remark that at such high temperature,
the O’Keeffe rule is not fulfilled since the crystal is cubic or with titling angle
smaller than the required critical value of �� to observe the Pnma to pPv phase
transition under pressure, therefore, our calculations are still valid at low temper-
atures. Following the suggested experimental analysis and in order to understand
the decomposition phases, we performed calculations for Na3 Mn2 F7 and MnF2
compounds. The first stoichiometry is the one of the Ruddlesen-Popper (RP)
phases [���] with An-1 A20 Bn X3n+1 formula, which in our particular case corre-
sponds to n = � and A 0 = A = Na, B = Mn and X = F. To our knowledge, there
is no previous report of the sodium-manganese based compound in the RP phase.
Therefore, all reported phases for similar compounds (oxides [���] and fluorides
[���]) were tested and the energy minimum per structure was analysed. Neverthe-
less, none of our simulated XRD patterns of the considered RP phases match the
one reported by Akaogi et al. [���], while for the pristine MnF2 compound, we
obtain good agreement with the experiments [���]. We are thus confident that
the crystal phase transition of NaMnF3 from the perovskite to the pPv phase
occurs at the predicted � GPa pressure. In any case, more experimental measure-
ments would be required in order to clarify the pressure phase transition that
takes place in NaMnF3 at lower (room) temperature.
We now focus on the vibrational characterization of the pPv phase at � GPa.
In Table �.� we report the computed phonon mode frequencies and the mode
Grüneisen parameters ( i ) in the Cmcm pPv structure with the lattice parame-
ters relaxed at � GPa (also reported in Table �.�). The irreducible representation
at the zone center point is: �Au �B1u �B2u �B3u �Ag �B1g B2g
�B3g . Two silent modes (Au label), sixteen IR active modes (B1u , B2u and
B3u ) and twelve Raman active modes (B1g , B2g and B3g ) are thus expected. As
we can see in Table �.�, all the calculated modes at the point of the BZ are
positive, indicating a vibrational local stability of the pPv phase at �GPa. The
lowest frequency mode is a silent mode Au at �� cm-1 and the lowest polar mode
is a B1u mode with a frequency of �� cm-1 . The obtained Grüneisen parameters
are important in the vibrational identification of the transition related to Raman
experiments under pressure. In general, they are smaller than the one obtained
�� prediction of the p p v transition in n a m n f 3

Table �.�: Computed Raman, Infra-red and silent modes of NaMnF3 in post-perovskite
Cmcm phase as well as the mode Grüneisen parameters ( i ) at around ��
GPa.
Raman, IR and silent modes [cm-1 ]
Ag B1g B2g B3g Au B1u B2u B3u
��� ��� ��� ��� �� �� ��� ���
��� ��� — ��� ��� ��� ��� ���
��� ��� — ��� — ��� ��� ���
��� — — ��� — ��� ��� —
— — — — — ��� ��� —
Mode Grüneisen parameters ( i )
�.��� �.��� �.��� �.��� �.��� -�.��� -�.��� �.���
�.��� �.��� — �.��� �.��� -�.��� �.��� �.���
�.��� �.��� — �.��� — �.��� �.��� �.���
�.��� — — �.��� — �.��� �.��� —
— — — — — �.��� �.��� —

in the Pnma ground state structure. This can be explained from an interatomic
force stiffness larger in the pPv phase than in the Pnma phase. Even though
there are some phonon modes with negative Grüneisen parameter (B1u modes
at �� cm-1 and ��� cm-1 and the B2u mode at ��� cm-1 ), they are not large
enough to have an effect on the thermal expansion such as the one observed in
↵-ZrW2 O8 [���].
We also report in Fig. �.� the evolution of the Raman and IR modes as a
function of pressure from � GPa to �� GPa. After the transition, the system goes
to a reduced crystallographic unit cell such as the number of modes is two times
smaller in the Cmcm pPv phase than it is in the Pnma phase. We note that
for both Pnma and pPv structure, all the phonon mode frequencies harden when
increasing the pressure. This is also valid for the soft B2u polar mode of the Pnma
structure that hardens as the pressure increases and thus in agreement with the
usual behaviour of perovskites in the presence of the AFD distortions, which are
inimical to the ferroelectric distortions and are enhanced with pressure.

�.� Magnetic properties of the pPv phase

Before performing the non-collinear magnetic calculations, it is useful to make a
preliminary symmetry analysis in order to determine the magnetic orderings that
are allowed in the pPv structure. The Cmcm space group belongs to the D2h point
 �.� magnetic properties of the ppv phase ��

Figure �.�: Vibrational behavior as a function of isotropic pressure for Pnma and Cmcm
pPv. The open and filled symbols are the Raman and the Infra-red modes
respectively.

group [���, ���] and the Mn atoms are placed at the 4c Wyckoff position with
coordinates (�,�,�) for Mn1 , (�,�, 12 ) for Mn2 , ( 12 , 12 ,�) for Mn3 and ( 12 , 12 , 12 ) for
Mn4 . From these � magnetic cations, we can define � collinear magnetic orderings
depending on the sign of the magnetic moments of (Mn1 , Mn2 , Mn3 , Mn4 ): F
(+, +, +, +) for the FM ordering, A (+, +, -, -) for the A-type AFM ordering,
C (+, -, +, -) for the C-type AFM ordering and G (+, -, -, +) for the G-
type AFM ordering, where + and - correspond to spin up and down respectively.
Of course, the ferrimagnetic ordering (+, +, +, -) could be considered but it
is unlikely to appear in real conditions because of the symmetry of the system
and the superexchange interaction [���, ���]. For the non-collinearity we have
to consider the three Cartesian directions x, y and z for each magnetic moment
which gives �⇥� possible states: Fx , Fy , Fz , Ax , Ay , Az , Cx , Cy , Cz , Gx , Gy and
Gz . Then, taking into account the fact that the spins are axial vectors, one has
to apply the symmetry operations of the D2h point group on each of the non-
collinear magnetic orders defined above and see how they transform. The result
of this approach applied on the pPv phase of NaMnF3 is summarised in Table
�.�.
�� prediction of the p p v transition in n a m n f 3

Table �.�: Magnetic ordering allowed in the Cmcm structure according to the D2h
point group [���, ���]. The transformation of each magnetic ordering under
each symmetry operation is labeled as +� and -�, which indicates when the
ordering is invariant under the related transformation (i.e. +�) or when it is
reversed under the application of the symmetry operation (i.e. -�).
Magnetic Orderinga I �z �y �x -� mz my mx
Fx , Ax � -� -� � � -� -� � B3g
Fy , Ay , Cz , Gz � -� � -� � -� � -� B2g
Fz , Az , Cy , Gy � � -� -� � � -� -� B1g
Cx , Gx , � � � � � � � � Ag

a More detailed information can be found in the Appendix B.

As we can see from Table �.�, the Fx and Ax orders have to the same char-
acter. This means that in an energy expansion with respect to these two order
parameters the second order term Fx · Ax is invariant by symmetry. This implies
that if one of the two order parameters condenses in the structure, the second
one can develop as well but as a secondary or a slave order (pseudo-proper �
[���]). This secondary order parameter usually appears as a spin canting in the
real system and its intensity will depend on the amplitude of the Fx · Ax coupling.
Going through the other magnetic orders, in Table �.� we also observe that Fy ,
Ay , Cz and Gz belong to the same B2g character, Fz , Az , Cy and Gy belong
to the B1g character and Cx and Gx belong to the Ag character. Interestingly,
whichever is the main magnetic order we observe in the NaMnF3 pPv phase, a
canting of the spins is always possible. Additionally, we also remark that three
of the four possible characters to which belong all the possible magnetic orders,
three of them allow for ferromagnetism and so have the potential to exhibit weak
FM.
As discussed above, NaMnF3 compound in the Pnma phase has a G-type AFM
ordering [��, ���, ���]. In Table �.� we report the computed Mn–F–Mn distances
and angles of the Pnma phase at room pressure and of the pPv phase at � GPa.
We note that in the Pnma structure, the Mn–Mn bonding in the three directions
have similar Mn–F–Mn angles (about ��� ) and distances (about � Å) while
their are much more anisotropic in the pPv phase (angles of �� and ��� and
one non-existent angle due to the octahedral edge sharing of the structure with
distances of �.�� Šand �.�� Šin the x - z plane and �.�� Šin the y direction).
Having angles closer to ��� in the Pnma explains the G-type AFM ordering

� A pseudo-proper transition appears when a secondary order parameter (Y ) has the same sym-
metry as the main transition order parameter (X) such as they coupled linearly (first order
coupling) in a free energy expansion: F=aX2 +bY 2 +cXY 2 . In the present case the ferromagnetic
(secondary) order parameter transforms as the main antiferromagnetic order parameter, giving
rise to the so-called weak-ferromagnetism.
 �.� magnetic properties of the ppv phase ��

Table �.�: Comparison of the distances and the bond angles along the Mn–F–Mn bonds
between the Pnma and the Cmcm structures of NaMnF3 .
Distance and Angle Pnma Cmcm
[���] �.���� [���] �.����
dMn-Mn [Å] [��-�] �.���� [���] �.����
[���] �.���� [���] �.����
[���] ���.�� [���] ��.��
Mn-F-Mn [deg] [��-�] ���.�� [���] ���.��
[���] ���.�� [���] —

since with such bond angle the superexchange interaction between the Mn-d5
spins favours an AFM alignment [���, ���, ���]. In the pPv case, we can see that
there is no direct bonding between the Mn atoms along the y direction such as
we can expect a weak spin coordination between the spins along the y direction
and one would expect a quasi-�D magnetic behaviour in the pPv structure. In
the plane perpendicular to the y direction we have two types of bonding giving
rise to an angle close to ��� along the z direction and to an angle close to ��
along the x direction. On the top of that, we also note that the �� bonding
along the x direction is given through two bonds since it is the direction where
the octahedra share their edge. The superexchange interaction should thus drives
strong FM ordering along the x axis and AFM ordering along the z axis and a
weak interaction along the y axis. Interestingly, the magnetic ordering that fulfils
these superexchange rules is the C-type AFM and it is indeed the lowest energy
collinear magnetic ordering that we found in our calculations (see Fig. �.�b).
Therefore, we can only consider the set of non-collinear calculations containing
the C-type AFM: (Cz , Fy , Gz , Ay ), (Cy , Fz , Az , Gy ) and (Cx , Gz ) as predicted
by group theory. We can also expect the Cx case to be unlikely observed since
it forces the spins to be antiferromagnetically aligned along the x direction while
a strong FM interaction is expected from the double �� bond angle connecting
the Mn atoms in this direction.
Relaxing the spin structure within the non-collinear regime and within the
C-type AFM ordering, we found that the lowest energy is obtained for the Cy
orientation with a large Fz component along the z direction (no canting along
the x axis is observed). This magnetic ground state is however dependent on
the U and J values of the GGA+U exchange correlation functional. In Fig. �.�a
we report the energy difference between the Cy Fz ordering and the other lowest
energy cases Cz Fy , Cz Ay , and Gy Az versus the J value (fixing J = � eV). Here
we remark that the Cy Fz , Cz Fy , Cz Ay , and Gy Az states are very close in energy
such as we can expect an easy magnetic phase transitions with small perturbations
such as an applied magnetic field. Such a transition under a magnetic field has
been observed experimentally in the pPv phase of NaNiF3 where a field parallel
�� prediction of the p p v transition in n a m n f 3

CyFz

CzAy
(a) (b)

Figure �.�: (a) Energy differences between the different magnetic orderings of the pPv
phase of NaMnF3 versus the U parameter. (b) Sketch of the C y F z and the
C z A y magnetic orderings in the pPv phase of NaMnF3 . For U < �.� eV
the system exhibits a AFM+FM magnetic state with a high ferromagnetic
canting. For U > �.� eV, the system transit toward a canted AFM state
with no weak FM.

to the y-axis induces a transition to a AFM+weak-FM ground state [���]. We

can see from Fig. �.�a that the Cy Fz corresponds to the lowest energy state from
U = �.� eV to U = �.� eV while from U = �.� eV to U = �.� eV the Cz Ay is
observed to be the most stable. This shows that a magnetic spin flop transition
appears in the pPv phase of NaMnF3 as function of the electronic Coulomb direct
exchange U. In Fig. �.� we report the evolution of the spin canting with respect to
U. The amplitude of the Fz canting goes from �.�� to �.� µB /atom when varying
U from �.� to �.� eV. We note that the Fz component strongly increase from U =
�.� to �.� eV and reach a saturation value beyond U =�.� eV. Unfortunately, no
experimental data of the non-collinearity of NaMnF3 in its pPv phase has been
reported in the literature such as it is not possible to validate which U value
would be the best to reproduce the spin canting amplitude. In previous studies
with d5 half filled orbitals the range of U values that have been found to treat
correctly the property of the systems is around U = �.� eV. It is important to
note that the pressure calculations performed with U = �.� eV in the Pnma and
pPv phase give the best agreement with experiments such as we will use U =�.�
eV in the rest of our analysis.
Beside the U parameter, for non-collinear magnetism within the DFT+U ap-
proach it is also necessary to check the dependence with respect to the J param-
eter [���]. In Fig. �.� we present the evolution of the spin canting versus the J
parameter in the Cy Fz ground state of the pPv phase of NaMnF3 with U = �.�
eV. As we can see, the FM component Fz is strongly affected by J, going from
�.�� µB to �.�� µB when J goes from �.� to �.� eV respectively. Here again, it
 �.� magnetic properties of the ppv phase ��

Figure �.�: Magnetic canting dependence as a function of U and J parameters for

NaMnF3 . Canting angle (') varies from �.�� to ��.�� when U goes from
�.� to �.� eV. The canting angle goes from ��.�� to �.�� when J varies
from �.� to �.� eV.

is difficult to predict what would be the best J value for NaMnF3 without any
experimental feedback. Previous studies about the DFT+U,J parameter report
that low J values for d5 orbital filling are usually the best values [���] such as
the large values of the Fz canting we obtain at low J (close to � µB /atom) in the
pPv phase of NaMnF3 would be the optimal value.
In order to estimate the error bar of our calculations, we report in Table �.�
a list of known magnetic pPv phase in oxide and fluoride perovskites with their
related magnetic orderings, d-orbital filling and S quantum number. We note
that in all oxide cases as well as in NaNiF3 , the ground state spin configuration
always corresponds to the low spin state. According to our calculations, only
NaMnF3 has the high spin configuration as ground state. In order to verify if this
is due to the DFT parameters used in our calculations or if this has a physical
meaning, we performed calculations of the pPv phase of NaNiF3 and CaRhO3
with U =�.� eV. We found a low spin ground state in both NaNiF3 (S=�) and
CaRhO3 (S= 12 ) and so in good agreement with the experimental results. Going
further, we also estimated the non-collinear spin ground state of NaNiF3 and
CaRhO3 in their pPv phase. As expected from the symmetry analysis, we observe
canted magnetic ground state in the two systems. The ground state magnetic
ordering of NaNiF3 is found to be Az with a Gy canting. Then, we did not find
any weak ferromagnetic spin canting in NaNiF3 in agreement with experimental
data. We note that experimentally a transition towards an AFM state with a
ferromagnetic canting is observed in the pPv phase of NaNiF3 when applying
a small magnetic field. The direction of the field induced canted FM spins in
NaNiF3 are the same than the one we report in the pPv phase of NaMnF3 without
magnetic field. This shows that the high-spin large FM canting we report in
NaMnF3 might be metastable in NaNiF3 such as a small magnetic perturbation
can switch to this state. More theoretical and experimental analysis would be
welcome to understand this transition. In the case of CaRhO3 we found Cy with
a Fz canting as magnetic ground state.
�� prediction of the p p v transition in n a m n f 3

Table �.�: Magnetic orderings allowed in the Cmcm structure for ABX3 systems with U
= �.� eV and J = � eV. The large magnetic canting for fluorides compounds
can be noted. The canting angle (') is measured with respect to the ab-plane
on the pPv phase. For NaNiF3 , the two possible magnetic states Cz Fy and
Cy Fz were computed.
Compound Fi [µB /atom] ' [deg] Magnetic Ground-state
CaRhO3 �.�� �.� Cy Fz
NaNiF3 �.�� (�.��) ��.� (�.�) Cz Fy (Cy Fz )
NaMnF3 �.�� ��.� Cy Fz

We found that when considering the cases with ferromagnetic spin canting, the
canting angle is much smaller in CaRhO3 (�.� ) and in NaNiF3 (�.� ) than it
is in NaMnF3 (��.� , see Table �.�). If we look at the amplitude of the canting
per atom we observe a very large amplitude for NaMnF3 (�.�� µB /atom) due to
the high spin configuration while it is much smaller in the case of NaNiF3 (�.��
µB /atom) and of CaRhO3 (�.�� µB /atom). This large canting can be the result
of a rather large Dzyaloshinsky-Moriya interaction [���, ���] or a large single-ion
anisotropy [���]. A more detailed analysis in order to identify the exact origin of
this large spin canting can be found in Chapter �. These previous calculations
allow us to be more confident about the unique high spin ground state of NaMnF3
with a large ferromagnetic spin canting in its pPv phase.
We also report in Fig. �.� the local density of states (LDOS) of the pPv phase
of NaMnF3 . The LDOS shows an overlap between the Mn-3d and F-2p states,
suggesting a covalent contribution between the Mn and F atoms due to the p - d
hybridization, which is a fundamental requirement for the superexchange interac-
tion. Moreover, we observe that the fluorine states are lower in energy then the
Mn-d orbitals, which corresponds to the expected bonding-antibonding picture
of covalency between magnetic cations and anions. Close to the Fermi level, no
contribution of Na states to the DOS is observed such as mostly Na size effects
are expected within the structure, a key-point parameter that is at the origin
of the geometric ferroelectric instabilities observed in cubic fluoroperovskites [�].
The dxy , dxz , dyz , dz2 and dx2 -y2 orbital occupancy gives the � orbitals filled
below the Fermi level for a spin channel and the � orbitals empty above of the
Fermi level for the other spin channel, which confirms the high spin electronic
configuration of Mn2+ in the pPv phase of NaMnF3 .

�.� Conclusions
We have studied, by means of first-principles calculations, the high pressure phase
diagram of NaMnF3 . We predicted a phase transition from the Pnma distorted
perovskite to a Cmcm post-perovskite phase at a critical pressure of � GPa, which
 �.� conclusions ��

Table �.�: Magnetic ordering in fluorides and oxides post-perovskite systems, AFM: an-
tiferromagnetic, DM: diamagnetic, FM: ferromagnetic, CWP: Curie-Weiss
paramagnetic. All pPv compounds exhibit a low-spin magnetic configuration
but NaMnF3 that according to our calculations is found to be in the high-spin
state.
Compound Magnetic ordering Ref.
pPv Oxides ABO3
CaIrO3 quasi-�D AFM S= 12 Ir4+ : �d5 (LS) [���]
CaRuO3 quasi-�D AFM S=� Ru4+ : �d4 (LS) [���, ���]
CaPtO3 DM S=� Pt4+ : �d6 (LS) [���, ���]
CaRhO3 AFM+weak-FM S= 12 Rh4+ : �d5 (LS) [���]
NaIrO3 CWP S=� Ir5+ : �d4 (LS) [���]
pPv Fluorides ABF3
NaFeF3 AFM S=� Fe2+ : �d6 (LS) [���]
NaCoF3 AFM — Co2+ : �d7 (LS) [���]
NaNiF3 AFMa S=� Ni2+ : �d8 (LS) [���, �]
NaMnF3 AFM+FM S= 52 Mn2+ : �d5 (HS) [�]

a NaNiF3 compound present a magnetic transition from AFM to AFM+FM ordering at applied
external magnetic field [���].

Figure �.�: Projected local density of states (LDOS) of the Cmcm phase of NaMnF3 .
A regular F-p bonding Mn-d antibonding is observed with small Mn:3d –
F:2p overlaping DOS is responsible for the small superexchange interaction
present in NaMnF3 .
�� prediction of the p p v transition in n a m n f 3

is relatively low with respect to the similar phase transitions observed in other
compounds. As expected from the tilting criteria[���, ���], this phase transition
fits very well with the octahedra tilting angle requirements where the transition
to the pPv is observed when the octahedra tilting angle reach the critical value
of �� with an initial ambient pressure tilting angle of �� . Unfortunately high
pressure experiments have been performed at very high temperature by Akaogi
et al. et al.[���] where the tilting criteria requirement is not fulfilled and also by
Katrusiak et al. [��] who unfortunately did not go beyond � GPa such as none
of them observed the pPv transition. Because of the partially filled d orbitals of
Mn2+ cation we also determined the magnetic ground state of NaMnF3 in its pPv
phase. Our calculations predicted a high spin ground state of the d5 spins with
a magnetic moment of �µB on each Mn. This high spin structure is particularly
interesting and unique since all the other known magnetic compounds presenting
the pPv phase shows a low spin ground state. We predicted an AFM magnetic
ordering ground state of the C-type with the spins lying along the y-direction
and with a surprisingly large ferromagnetic spin canting along the z-direction.
This makes NaMnF3 of particular interest for fundamental and geophysical stud-
ies since it has this unique property of high spin magnetic ground state with
potentially large ferromagnetic spin canting and because it can be synthesized
at relatively small pressure. The fact that the fluoride post-perovskites can be
maintained when the pressure is released makes NaMnF3 even more interesting
to study.
It is important to remark that very recently, another transition at higher pres-
sures was reported and named as a post-pPv [���]. In this phase, the structure
transforms to a face-shared octahedra with a substantial compression along the
crystallographic a-axis into a ppPv-Pnma phase —see Fig. � in Ref [���]—. More-
over, it has been found that some oxide compounds transforms directly from the
room conditions Pnma to the post-pPv under isotropic pressures above �� GPa.
Interestingly, the magnetic ordering in the ppPv materials, with magnetically
active A- and B-sites, shows magnetic “double” one-dimensional channels in ad-
dition to a triangular-like coordination of the B-site generating a geometrically
frustrated magnetic system.
Therefore, we hope that our results on the pPv phase and the reported ppPv
transition will motivate the experimentalist to carry out more studies of NaMnF3
in order to confirm the pPv phase transition at low pressure and low tempera-
ture and to check its potential for high spin structure with a sizeable canted
ferromagnetism.
NON-COLLINEAR MAGNETISM IN
4
CaRhO3 AND NaNiF3
POST-PEROVSKITES

Perovskite (Pv ) oxides are materials that have demonstrated to be a crucial

family in the search for multifunctionality and controllable properties in simple
stoichiometries where the keystone is to have several physical properties in a sin-
gle crystal [���, ���]. The pressure have demonstrated to be a very important
external parameter to understand the physics of these crystals because it is possi-
ble to modify and tune the structural parameters and the electronic and magnetic
properties. For example, in perovskites, ferroelectricity can be suppressed or in-
duced by pressure and strain [��]. Although, new and interesting phases can be
synthetised such as the post-perovskite (pPv ) phase [���]. In the case of the
(pPv ) phase, which can be found under high pressure in some ABX3 compounds,
several interesting properties have been reported such as �D and �D magnetism
[���, ���, ���] related to the layered-like structure of the pPv phase or a Mott
insulator state [���]. The interest in the magnetic ordering in pPv ’s is mostly
found through experimental studies such as the magnetic field induced antifer-
romagnetic (AFM) to ferromagnetic (FM) transition in NaNiF3 [���] and the �-
and �-dimensional magnetic ordering with a clear non-collinear magnetic contri-
bution in some of these compounds [���, ���, ���]. The magnetic ordering in pPv
compounds is induced by a magnetically active B-site cation where the CaIrO3
crystal is one of the most studied system [���] because it presents a quasi-�D
AFM ordering similar to CaRuO3 [���]. Interestingly, it has also been observed
that CaRhO3 [���], NaCoF3 [���] and NaNiF3 [���] behave as a quasi-�D AFM
system. Nonetheless, even though the magnetic properties have been measured
in some of these pPv ’s compounds, the source of the observed canted magnetic
structure is still unclear and, to our knowledge, there are only few theoretical
efforts to present a clear picture of its physical origin. Additionally, a wide range
of properties remains unexplored in this phase such as the competition between
the structural distortions and the non-collinear magnetism, suggesting promising
and unexpected phenomena to be understood.

In this chapter, we present a study of the non-collinear magnetism in the pro-

totype pPv phase by means of first-principles calculations. We have selected two
different post-perovskite systems, one oxide CaRhO3 and one fluoride NaNiF3 .
The possible non-collinear ordering are explored through group theory symmetry
analysis and simulated through density functional theory calculations. We then

��
�� non-collinear magnetism in c a r h o 3 and n a n i f 3 post-perovskites

scrutinised the origin of the non-collinear magnetism by computing the exchange

magnetic couplings (Jij ), the anti-symmetric Dzyaloshinsky-Moriya (Dij ) inter-
action and the single-ion anisotropy (SIA) parameters to determine the origin of
the non-collinear magnetism.

�.� Computational Details

First-principles calculations were performed within the VASP code [��, ��] and
the PAW formalism to describe the valence and core electrons [��]. The electronic
configuration of pseudopotentials as valences and semicore electrons taken into
account for calculations are Na (�p6 �s1 ), Ni (�p6 �d8 �s2 ), and F (�s2 �p5 ) for
NaNiF3 and Ca (�s2 �p6 �s2 ), Rh (�p6 �d8 �s1 ), and O (�s2 �p4 ) for CaRhO3 . The
exchange correlation was represented within the GGA approximation, with the
PBEsol pseudopotential [��] and corrected by means of the GGA+U formalism,
with U = �.� eV and J = � within the Liechtenstein approach [��]. The magnetic
character was taken into account in the calculations according to the spin polar-
ized DFT approximation used for the exchange correlation functional [��, ��]. The
periodic solution was represented through Bloch states with a Monkhorst-Pack
k -point mesh of (�⇥�⇥�) and ��� eV of energy cut-off, which has been tested to
achieve forces convergence less than �⇥��-3 eV/Å and an accuracy of �⇥��-6 eV
in energies. Vibrational properties were computed through the formalism of den-
sity functional perturbation theory (DFPT) [���, ���] and the post-processing
analysis was carried out with the Phonopy code [��]. For the non-collinear mag-
netic calculations, the spin-orbit coupling has been included as implemented in
VASP [��].

�.� Structural Characterization of CaRhO3 and NaNiF3

At atmospheric pressure and room conditions, CaRhO3 and NaNiF3 exhibit an
orthorhombic Pnma structure (Space group No. ��). However, under high hy-
drostatic pressure, the CaRhO3 and NaNiF3 compounds crystalizes in the pPv
phase (Cmcm space group No. ��) for pressures above � GPa and �� GPa re-
spectively [���, ���]. Based on the advantage of these two materials to conserve
the pPv structure at ambient conditions once the pressure is released, we carried
out calculations at atmospheric pressure in order to compare and validate our
calculations with the available experimental findings. After the structural relax-
ation within the C-type AFM collinear magnetic ordering, we obtained lattice
parameters a = �.��� Å, b = �.��� Å, c = �.��� Å and a = �.��� Å , b =
�.��� Å, c = �.��� Å for CaRhO3 and NaNiF3 respectively. These results are
in agreement with the experimental measurements with an error of about �.�%
[���, ���]. The corresponding XRD patterns (see Fig. �.�) have the characteristic
peaks observed in the pPv phase of CaRhO3 [���, ���] and NaNiF3 [���, ���],
which demonstrates a good agreement between our calculations and the experi-
 �.� non-collinear ordering ��

Figure �.�: Simulated XRD patterns with K ↵ :Cu for CaRhO3 (above) and NaNiF3
(below) in the Cmcm phase at atmospheric pressure are observed.

ments. In Table �.� we present the calculated Raman, Infra-Red (IR) and silent
mode frequencies of the pPv phase of CaRhO3 and NaNiF3 at ambient pressure.
Even if the pPv is not the ground state structure at room pressure (The Pnma
phase is the ground state), we do not observe any phonon instabilities confirm-
ing that this phase is a metastable phase (The Pnma phase has a lower energy
than the pPv phase) and as observed experimentally when releasing the pressure
[���, ���]. The stability of the pPv phase at room conditions is of great interest
due to the possibility to explore the properties of the pPv structure and its use
for technological applications.

�.� Non-collinear Ordering

Having a correct description of the structure, we now focus on the magnetic
ground state of the pPv phase. The magnetic behavior in perovskites and post-
perovskites is strongly dependent on the coordination of the B-site magnetic
cation in addition to the orbital symmetry and its electron filling [���]. In pPv
oxides, the magnetism comes from the �d and �d electrons (i.e. Rh [���], Ru [���]
and Ir [���]) while in fluorides it is due to the �d elements (for example Co [���,
���] and Ni [���, ���]). This can induce different properties and behaviour when
comparing the magnetism present in oxides and the one in fluorides pPv phases
due to the change in the chemical environment provided by the fluorine compared
to the oxygen. The octahedral tilting and rotation are known to affect the non-
collinear arrangement of the magnetic moments as demonstrated by Weingart
et al. in the AFeO3 systems [���]. Thus, due to the strong modification of the
octahedra coordinations in the pPv phase with respect to the Pnma phase (the
octahedra share their edge in the pPv phase while they are corner-shared in the
Pnma phase) we can expect a strong modification of the magnetic interactions in
�� non-collinear magnetism in c a r h o 3 and n a n i f 3 post-perovskites

Table �.�: Vibrational characterization modes of CaRhO3 and NaNiF 3 Cmcm phase at
atmospheric pressure with the representation: = �Au �B1u �B2u
�B3u �Ag �B1g B2g �B3g .
Raman , IR and silent (Au ) modes [cm-1 ]
CaRhO3
Ag B1g B2g B3g Au B1u B2u B3u
��� ��� ��� ��� ��� ��� ��� ���
��� ��� — ��� ��� ��� ��� ���
��� ��� — ��� — ��� ��� ���
��� — — ��� — ��� ��� —
— — — — — ��� ��� —
NaNiF3
��� ��� ��� ��� �� ��� ��� ���
��� ��� — ��� ��� ��� ��� ���
��� ��� — ��� — ��� ��� ���
��� — — ��� — ��� ��� —
— — — — — ��� ��� —

the pPv phase when comparing with the regular distorted perovskite structures
as the Pnma phase.

CzAy CzFy GyAz CyFz

Figure �.�: The non-collinear magnetic orderings in the pPv Cmcm symmetry labeled as
Cz Fy , Cz Ay , Cy Fz and Gy Az are presented in the figure. Two of them with
pure AFM (Cz Ay and Gy Az ) behavior and two AFM+weak-FM (Cz Fy and
Cy Fz ) can be observed (Fig. elaborated with the VESTA software [���]).

Before performing the non-collinear magnetic calculations, it is important to

first check the magnetic states that are allowed by symmetry. In Table �.� we
present the main possible magnetic orderings and couplings in the ABF3 pPv
phase (Cmcm space group), where the magnetic B-cation are located at the
Wyckoff positions �c (see details in Ref. [�] and Appendix B). Including addi-
tional criteria such as the superexchange interaction related to the bondings, we
 �.� exploring the magnetic interactions ��

can neglect unlikely states and thus reducing the possible magnetic states to
Cz Fy , Cz Ay , Cy Fz and Gy Az , as summarized in Table �.� and depicted in Fig.
�.�.

Table �.�: Magnetic orderings allowed in the Cmcm pPv structure according to the D2h
point group [�]. Here F is ferromagnetic ordering and A - C - and G - are the
different AFM magnetic orderings where the subindex indicate the direction
along the particular Cartesian axis.
Magnetic Ordering Character
Fx , Ax B3g
Fy , Ay , Cz , Gz B2g
Fz , Az , Cy , Gy B1g
Cx , Gx , Ag

According to our total energy calculations, the ground state corresponds to the
Cz Fy and Gy Az magnetic states for CaRhO3 and NaNiF3 respectively (see Table
�.�). It can be noted that these magnetic ground states as well as their canting
amplitude are in good agreement with available experimental data. In Table �.�,
we also report the experimental values of the magnetic state of NaNiF3 under
an applied magnetic field and we compare it with a calculation made with a
higher energy configuration. The later corresponding to the magnetic ground
state observed experimentally where we get a good agreement. We remark that
the atomic magnetic moment in the NaNiF3 case is almost three times larger
that in the oxide CaRhO3 . This is due to the occupation of the B-site magnetic
cation that drives a S = � state in NaNiF3 and S = 12 in CaRhO3 .

Table �.�: Non-collinear ordering and moments in CaRhO3 and NaNiF3 . Experimental
and our theoretical calculations are in good agreement. The magnetic mo-
ments are in µB /atom for each magnetic B-site cation. The induced magnetic
ground state by applied magnetic field (Hext ) in NaNiF3 is also presented.
Compound Ordering C, G F, A Fi exp.
CaRhO3 Cz Fy �.�� �.�� �.�� [���]
NaNiF3 Gy Az �.�� �.�� — [���]
NaNiF3 (Hext > � kOe) Cz Fy �.�� �.�� ⇡ �.� [���]

�.� Exploring the Magnetic Interactions

In the Pnma Pv systems, in which the canting is allowed, the source of the non-
collinear ordering lies in the DM vectors (see a detailed analysis in Ref. [���] for
�� non-collinear magnetism in c a r h o 3 and n a n i f 3 post-perovskites

the AFeO3 systems). Here, we propose to analyse the origin of the non-collinear
magnetism in the pPv phase of NaNiF3 and CaRhO3 by decomposing the mag-
netic interactions into its isotropic exchange (J ), DM and SIA components. To
that end, it is convenient to start by defining the magnetic Heisenberg Hamilto-
nian (H) as follows:

H = -2[Jij S~i · S~j + D~ij · (S~i ⇥ S~j ) + S~i · A

~i · S~i ] (�.�)
Here Jij are the magnetic exchange constants between the S~i and S~j magnetic
moments of the i and J interacting atoms. Dij are the components of the anti-
symmetric exchange coupling known as the DM interaction [���, ���] and Ai are
the SIA components. The Jij , Dij and Ai interactions were computed according
to the methodology proposed by Xiang et al. [���] from the total energy of partic-
ular magnetic configurations summarized in Eqs. �.�, �.�, and �.� respectively.
For this purpose, �⇥�⇥�, �⇥�⇥� and �⇥�⇥� supercells were used in order to
completely isolate the interactions in each direction. Based on the symmetry of
the pPv phase, the magnetic interlayer exchange interactions are defined such as:
J12 (along x-axis) ⌘ Ja and J13 (along c-axis) ⌘ Jc as depicted in Fig. �.�. The
interaction between octahedral layers, J14 (in the xy-plane) ⌘ Jab is also included.
Using the same notation than for the exchange constants, the antisymmetric DM
vectors are defined as: D12 ⌘ Da and D13 ⌘ Dc (see Fig. �.�). Due to the fact
that there is no direct bonding of the B-sites between the layers, it is expected
to have a quasi-null contribution from the DM interaction for that direction.

Da Dc

Jab

Ja
Jc

Figure �.�: Schematic representation of the atomic sites and magnetic interactions in
ABX3 (A, B and X -sites in blue, yellow and red respectively). Exchange
constants labeled as Ja , Jc and Jab and DM interactions represented as Da
and Dc equivalent to D12 and D13 respectively are shown. (Fig. elaborated
with the VESTA software [���]).

E"" + E## - E"# - E#"

Jij = , (�.�)
4S2
 �.� exploring the magnetic interactions ��

Eij ij ij ij
"! + E# - E" - E#!
Dk
i 0j 0 = , (�.�)
4S2

Ej" + Ej# - Ek k
" - E#
Ai = , (�.�)
2S2
The values for the Jij constants (see Table �.�) demonstrate a ferromagnetic
interaction along the x-axis with Ja > �. Additionally, we can observe a strong
AFM coupling along the z-axis and a weak interlayer interaction, in agreement
with the non-collinear magnetism results reported in Table �.�. In the CaRhO3
compound, Jab >�, which corresponds to a FM interlayer interaction as expected
from the Cz Fy magnetic ground state. In NaNiF3 Jab < 0 which confirms the
Gy Az ordering and in agreement with the group theory analysis. It is important
to remark that the magnetic constants in CaRhO3 are at least one order of
magnitude larger that in NaNiF3 . This can be explained by the lower covalence
in the B–F–B bonding, when compared to the B–O–B and related also to the
expected JT -distortion with Rh4+ cation as well as the fact that �d and �d
electrons are known to have higher exchange constants than the �d electrons
[���].
As demonstrated by Kim and Min [���] in d3 , d5 and d8 , B-sites, the DM
vector in perovskites can be related to the inter ionic vectors as Dij / (x̂i ⇥
x̂j ) [���], where x̂i and x̂j are unitary vectors along the B–X–B bonds. Thus,
according to the previous relation and the symmetry of the pPv compounds, the
Da must be zero and the Dc has to be along the x-axis, then: Da x = Dy = Dz =
a a

� and Dy = Dz = �. However, if a Jahn-Teller distortion is present (i.e. such as

c c

in Mn3+ (d4 ), Cu2+ (d9 ) or Rh4+ with t52g occupation), there is a change in the
orbital symmetry, breaking the mirror symmetry and creating a deviation from
the Kim and Min rule (as it happens in LaMnO3 [���]). In Table �.� we report
the components of the DM vectors for the two considered compounds. Analyzing
the DM components, we conclude that NaNiF3 , with Ni2+ , perfectly fulfill the
Kim and Min rule with a non-zero Dcx = �.��� meV. However, in the CaRhO3
case, a deviation is observed where Da x = -�.��� meV. This is probably due to the
t52g state that induces a JT -distortion, which in turn changes the orbital symmetry
and then, change the DM vectors.
According to the relation between the exchange constants and the magnitude
of the DM vector, the magnetic canting angle ' can be defined as [���]:

!
1 ~ ij |
|D
' = tan-1 (�.�)
2 Jij

if there is no SIA in the direction of the canting [���, ���]. We can estimate
the canting angle from this simple formulae and compare the results with the
�� non-collinear magnetism in c a r h o 3 and n a n i f 3 post-perovskites

relaxed values obtained through the DFT calculations. For NaNiF3 the canting
angle coming from the relaxed DFT calculations are �.�� and �.�� for Gy Az
and Cz Fy respectively, and we obtain ' = �.�� according to the calculated Dij
and Jij parameters and the use of Eq.�.�. In CaRhO3 the canting angle coming
from the fully relaxed calculations is �.�� while it is �.�� from Eq.�.�. The large
discrepancy between the relaxed and estimated canting angle in CaRhO3 can be
related to the competition between JT -distortions, orbital symmetry [���] and
electronic Coulomb localization [���] through the SIA. The contribution of SIA
to the canting angle has been first reported by Moriya in NiF2 where the DM
interaction is forbidden by symmetry [���, ���].

Table �.�: Computed magnetic exchange constants for pPv CaRhO3 and NaNiF3 . Units
in meV.
Ja Jc Jab
CaRhO3 ��.�� -��.�� �.��
NaNiF3 �.�� -�.�� -�.��

Table �.�: Dzyaloshinsky-Moriya (Dij ) vectors computed for pPv CaRhO3 and NaNiF3 .
Here, D12 ⌘ Da and D13 ⌘ Dc as shown in Fig. �.�. Units in meV.
Da Dc
Da
x Da
y Da
z Dcx Dcy Dcz
CaRhO3 -�.��� �.��� �.��� -�.��� �.��� �.���
NaNiF3 �.��� �.��� �.��� �.��� �.��� �.���

To elucidate this problem, we report in Table �.� our calculated amplitude of

the SIA components obtained using the Eq. �.�. The SIA amplitude of NaNiF3
is one order of magnitude smaller than its DM vector components which suggest
a DM driven magnetic canting in the fluoride compound. However, in CaRhO3 ,
the Ax and Ay parameters are two times larger than the Dcx component of Dc ,
which suggests a strong interaction between the DM and SIA interactions. This
can explain why the simple tangent formula Eq. �.� that takes into account only J
and DM interactions gives a bad estimation of the canting angle in the CaRhO3
case while it works pretty well in the case of NaNiF3 . According to the larger
canting observed in the fully relaxed calculations with respect to one obtained
with Eq. �.�, we can conclude that the SIA is cooperative with the DM vector
in developing a canting of the magnetic moments.
 �.� summary and conclusions ��

Table �.�: Computed SIA parameters for pPv CaRhO3 and NaNiF3 . Units in meV.
Ax Ay Az
CaRhO3 -�.��� �.��� �.���
NaNiF3 �.��� �.��� -�.���

�.� Summary and Conclusions

By means of first-principles calculations we have studied the non-collinear mag-
netism in the high pressure post-perovskite phases of CaRhO3 and NaNiF3 . We
have found that according to group theory the magnetic canting is allowed by
symmetry in all of the pPv ABX3 with a magnetically active B-site cation. The
exchange constants, the single-ion anisotropy parameters and the Dzyaloshinsky-
Moriya vectors components were computed and we found a ferromagnetic ex-
change interaction along the x-axis and an AFM interaction along the z-direction.
The exchange coupling Jab between layers along the y-axis is much smaller than
the interlayer Ja and Jc constants by at least two orders of magnitude, which
confirms the quasi-�D magnetic behaviour observed in the magnetic pPv crystals.
Interestingly, we found that the DM interaction is responsible for the spin canting
in NaNiF3 while the single-ion anisotropy is found to play an important role in
the canting of CaRhO3 and cannot be neglected if one wants to understand the
non-collinear magnetism in this system. The origin of this difference can be due
to the presence of a Jahn-Teller active cation in CaRhO3 that strongly affects
the magnetism of the Rh cation. With this work, we hope to motivate the exper-
imentalist to further analyze the magnetic properties of the quasi-�D magnetic
post-perovskites materials.
I N D U C E D M U LT I F E R R O I C A N D
5
M A G N E T O E L E C T R I C B E H AV I O R I N
NaMnF3

In the search for new and innovative materials with exciting multifunctional prop-
erties, multiferroics and magnetoelectrics has been a keystone in condensed mat-
ter physics in the last decade [���, ���]. Several systems and compounds based
on oxide perovskites have been reported as ideal candidates. In these systems, the
presence of the ferroic orders can be induced and/or tuned by different physical
constrains such as the case of the biaxial epitaxial strain [���] in thin films and
superlattices form [���]. In this search, one of the main challenges is to induce a
ferroelectric order in magnetic and non-magnetic perovskites. Then, taking the
biaxial epitaxial strain as a control parameter, the condensation of a stable fer-
roelectric state has been successfully reported in CaTiO3 [���], SrTiO3 [��] and
CaMnO3 [���, ���] the first two of them paramagnetic materials at their ground
state and the later antiferromagnetic (AFM) with a weak ferromagnetism (weak-
FM) due to a magnetic canted structure [���]. Here, the strain couples with the
softest mode and thus, induce a polar displacement achieving with this a phase
transition to a polar state in a already AFM structure [���] and thus exceeding the
so-called d0 -ness rule that is expected to prevent the formation of a FE phase in
magnetic perovskites [���]. However, the possibility for new materials with these
multifunctional properties in new stoichiometries is still evasive. Recently, fluo-
rides have been reported as possible candidates for a multiferroic behavior with
promising properties [��]. Nonetheless, none of the reported fluoride candidates
belong to the most claimed perovskite family.

Lately we have shown that even if none of the fluoroperovskites is reported

with a FE ground state (except CsPbF3 [���, ���], and the theoretically pre-
dicted NaCaF3 [��], and NaCdF3 [���]) they have nevertheless the propensity
to have a FE instability in their high symmetry cubic reference structure [�]. We
have identified that the FE instability of the fluoroperovskites is related to a
steric geometric effect when small cations lie at the A-site. The later contrary
to the charge transfer or A-site stereochemically active lone pair origin observed
in the oxides and to the phonon mode coupling origin present in improper and
hybrid-improper FE [���]. Then and unfortunately, these fluoroperovskites keep
the “undesired” competition between the FE and antiferrodistortive instabilities
such as the AFD dominate in the bulk ground states. Thus, the possible tuning of
the ferroelectricity taken advantage of this hidden polar mode in the paraelectric

��
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

room temperature ground state becomes interesting and promising. Additionally,

in oxides, the FE strain-induced state can be only obtained through tensile strain,
instead compressive one, in which the frequency of soft polar mode goes to higher
values. Besides, a linear behavior of the FE–frequency as a function of strain has
been observed [���]. Then, based in the different origin of the polar instability
in fluorides, a different strain-polarization coupling should be expected. More-
over, it is worth to mention that one of the key factors in the search for new
magnetoelectrics with a strong ME coupling is the requirement for a large fer-
romagnetic canting [���, ���] (induced by the Dzyaloshinsky-Moriya interaction
[���, ���]) which in the fluoroperovskites tend to be higher (see Chapters � and
�). This need for a large FM canting is based in the weaker effect of electric field
in magnetism than the effect of an applied magnetic field on the ferroelectric
polarization [���, ���].
Though the sections of this chapter the reader can find the study on the mul-
tiferroic state of NaMnF3 under epitaxial strain and its non-linear ferroelectric
behavior as a function of epitaxial strain. It is shown that in this material a mul-
tiferroic state can be realized for tensile or compressive strain, this completely
different in comparison with oxides [���, ��, ���]. Additionally, out-of-plane po-
larization is observed instead of the ac Pnma in-plane spontaneous polarization
reported in oxides. An B-site contribution is observed for negative strain values,
therefore an A- and B-sites leaded FE was found in the Pna�1 strained structure.
Magnetoelectric coupling allowed by symmetry is studied in the AFM+FM+FE
Pna�1 NaMnF3 . Then, it can be concluded that according to our knowledge,
the NaMnF3 is the only reported multiferroic/magnetoelectric fluoroperovskite
material obtained so far.

�.� Computational Details

Here again, we used DFT as implemented in VASP code [��, ��] as the frame-
work of this work. PAW [��] pseudo-potentials were used to represent the valence
and core electrons. The electronic configurations taken into account in pseudo-
potential as valence electrons are presented in Table �.�. The exchange corre-
lation was represented within the GGA - PBEsol parametrization [��]. Due to
the magnetic character of these systems, we included the electron-spin in the
calculations. The d-electron localization is corrected by means of the GGA+U,
with U = �.� eV and J = �.� eV within the Liechtenstein formalism [��]. The
periodic solution of these crystalline structures was represented by using Bloch
states with a Monkhorst-Pack k-point mesh of �⇥�⇥� and ��� eV energy cut-off,
which has been tested already to give forces convergence to less than �.���� eV/Å.
Spin-Orbit coupling (SOC) was include into non-collinear calculations [��]. Born
effective charges and phonon calculations were performed within the DFPT for-
malism [��]. The dynamical matrix was then unfolded using the Phonopy code
[��]. The ferroelectric spontaneous polarization was computed through the Berry
 �.� epitaxial strain in pnma NaBF 3 fluorides ��

phase approach [��] as implemented in the VASP code. The ME coupling was
obtained by computing the spontaneous polarization as a function of the applied
Zeeman magnetic field as implemented by Bousquet et al. [��] within the LDA
approach —for more details see Chapter �—. Besides, we have evaluated the ME
response against the U and J parameters of the DFT+U as shown later.

�.� Epitaxial strain in Pnma NaBF3 Fluorides

Up to this point, with all the acquired knowledge about the vibrational properties
of the fluoroperovskites (Chapter �) and the effect of the isotropic pressure in
these compounds (Chapters � and �) we now continue by analyzing the biaxial or
epitaxial strain (") effect into their properties. For that purpose we selected the
NaBF3 family based in the results discussed in Chapter �. In Fig. �.� we present
the frequency value of the FE instability (T1u ) at the cubic high temperature
symmetry as well as the softest B2u mode in the Pnma ground state for B =
Mn, V, Zn, and Ni (in their ionic radii order). It can be easily observed that
both modes, T1u and B2u follow the ionic radii size trend. Then, it is obviously
expected that as bigger is the instability, softest is the polar B2u mode and then,
easier is the destabilization of the mode to induce ferroelectricity as the ionic
B-site radii increases.

Figure �.�: Ferroelectric instability T1u at the cubic Pm3̄m as well as the frequency
values for the softest B2u mode in the Pnma NaBF3 (B = Mn, V, Zn, and
Ni) compounds.

Based on an experimental
p point of view, we suppose a cubic perovskite sub-
strate where a = c = 2ao . Thus, the �% strain value corresponds to ao =
ai +ci
2 where ai and ci are the unstrained relaxed cell parameters of the Pnma
phase defining " = a-aao 100%. After the analysis and calculations, the only flu-
o

oroperovskite that exhibit a ferroelectric instability under epitaxial strain is the

NaMnF3 compound which is indeed, the compound with the largest ferroelectric
instability (around ��� icm-1 ) and the lowest soft mode at the ground state (see
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

Fig. �.�). Thus, a detailed research in the later was carried out and all the results
are presented in the following sections.

�.� Strain-Induced Ferroelectricity in NaMnF3

As commented previously, sodium-manganese fluoride, NaMnF3 crystallizes in
the Pnma structure at room conditions. For this phase, lattice parameters ob-
tained by us within DFT methodology are ai = �.��� Å, bi = �.��� Å and ci =
�.��� Šin well agreement with experimental reports by Daniel et al. [���]. At high
temperature cubic Pm3̄m structure, NaMnF3 presents antiferrodistortive (AFD)
modes related to octahedral Mn–F6 tilting and rotations at the M–R branch as
well as antipolar displacements at the X-point according to the phonon-dispersion
curves (see Chapter �). The FE-mode amplitude is comparable with the rotations,
AFD distortions and antipolar displacements modes with values of �� icm-1 ,
��� icm-1 , ��� icm-1 and �� icm-1 , for R+ 5 , M3 , R4 and X5 respectively (see
+ + +

Chapter � and Ref. [�]). Thus, the competition between ferroelectric, antipolar
displacements and AFD distortions at the ground state are present in this fluoride
and the balance between those is really delicate. Despite the presence of the FE
mode, as discussed above, the ground state belongs to a non-polar orthorhombic
Pnma structure. However, as we found, the ferroelectric instability in fluorides
present a totally different origin, which is related to A-site geometrically driven
polar displacements [�].
In FE oxides it is well known that in some particular cases, the epitaxial strain
can induce ferroelectricity in the Pnma phase [���] or enhance it in compounds
with different ground states and symmetries such as PbTiO3 or BiFeO3 [���]. In
NaMnF3 the lowest FE mode has the B2u symmetry and it is polarized perpen-
dicular to the biaxial constraint (the orthorhombic b-axis). Unexpectedly, this
mode becomes unstable whatever the value of the epitaxial strain in compression
or expansion regime as presented in Fig. �.�a. By freezing the unstable mode at
and performing the structural relaxation, the new ferroelectric ground state is the
polar Pna�1 space group as it can be observed from the double well energy curve
presented in Fig. �.�c. Then, the ground state of the epitaxially constrained struc-
ture is always ferroelectric and thus multiferroic due to the G-type AFM NaMnF3 .
This finding is completely different from oxides in which some of them, originally
paraelectric, can only reach the ferroelectric state for strain values above �.�%
such as in CaTiO3 [���], �.�% in SrTiO3 [��] and �.�% in CaMnO3 [���]. Ad-
ditionally, for negative or positive strain, the polarization is enhanced up to a
value of �� µC·cm-2 at +�% and � µC·cm-2 at -�% respectively, and here again,
a strong difference with previous observations and reports in ferroelectric oxides
[���]. We note also that the polarization develops in the direction where the anti-
polar motions of the X+ 5 mode are absent and thus, where there is still possibility
for the A-site to move in the remaining flat energy landscape direction obtaining
a FE stable state along b-axis with an anti-polar character in the ac-plane.
 �.� strain-induced ferroelectricity in NaMnF 3 ��

Figure �.�: a) Epitaxially strain induced B2u instability in Pnma NaMnF3 in which
a nonlinear coupling is observed, imaginary frequencies are represented as
negative values by notation. b) Eigendisplacements of the polar B2u along
b-axis (⌘b ) for each site (Na in green squares, Mn in violet circles and
F? , Fk in upper and down triangles respectively) in NaMnF3 in Pna�1 . c)
Ferroelectric mode freezing in which the transition from the paraelectric
Pnma to a polar Pna�1 phase is observed at equivalent �% in the ac-plane.

In order to understand the origin and physical phenomena within the transi-
tion and the stable FE ground state, we analyzed the eigendisplacements of the
unstable mode B2u under epitaxial strain (see Fig. �.�b). From this figure it can
be concluded first that as in oxides, the Na alkali-metal ion (in this case the
A-site) tries to annihilate the ferroelectricity when the strain goes from �%, to
-�% with eigendisplacements of �.��� and �.��� for Na respectively. However, the
antipolar motion of B-site (Mn in this case) decrease and the eigendisplacements
of this site begins to present a polar contribution to the b-axis that compensates
the decrease in contribution of the Na site (see Table �.�). This can explain the
anomalous behavior of FE-instability at compression epitaxial strain. The fluo-
rine displacements remain almost invariant for expansion or compression strain
with no significant effect on the instability behavior. The analysis of the Born
effective charges (Z⇤ in the Table �.�) reveal that values of this parameter are
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

Table �.�: Born effective charges (Z⇤ , in e- ) of selected strain values ("). F? and Fk rep-
resent the Born effective charge of the F atom when it is displaced respectively
against the Mn cation and along the perpendicular direction. Eigendisplace-
ments (⌘b ) of the FE unstable mode along the b-axis in NaMnF3 . F? and
Fk have the same sense as given in Born effective charges.
Z⇤ [e- ] ⌘b [a.u.]
" [%] Na Mn F? Fk Na Mn F? Fk
Nominal � � -� -� — — — —
+�% �.�� �.�� -�.�� -�.�� �.��� �.��� -�.��� -�.���
�% �.�� �.�� -�.�� -�.�� �.��� �.��� -�.��� -�.���
–�% �.�� �.�� -�.�� -�.�� �.��� �.��� -�.��� -�.���

close to nominal ones (+�, +� and -� for Na, Mn, and F respectively) which is
related to a geometric origin of the ferroelectricity in NaMnF3 as expected [�]
instead of a charge-transfer phenomena found in most of the ferroelectric oxide
perovskites [��]. A x
Ax

Gz

Py

Fy

Ax
NaMnF3
G z Ax Fy + Py

a) b)

Figure �.�: a) Polarization along b -axis in NaMnF3 in Pna�1 symmetry. The second-
order dependence of the polarization as a function of ac -plane epitaxial
strain is observed as a result of the non-linear piezoelectricity in this per-
ovskite fluoride. b) NaMnF3 , Pna�1 structure is also presented where Na,
Mn and F ions are depicted in green, violet and grey respectively. Addi-
tionally, the non-collinear magnetic ground state as well as the polarization
vector along the b -axis are schematically depicted in the figure showing the
multiferroic character of this fluoroperovskite.

�.�.� Second-order piezoelectricity in Pna�1 NaMnF3

We found a non-linear ferroelectricity-strain coupling in this compound becom-

ing another different property in the engineered multiferroicity in Pna�1 when
 �.� strain-induced ferroelectricity in NaMnF 3 ��

comparing to oxides. This fact is demonstrated by the quadratic dependence of

the spontaneous polarization as a function of epitaxial strain as observed in Fig.
�.� and expressed in terms of the strain, piezoelectric constants and polarization
in Eq. �.�. To our knowledge, this piezoelectric property has been found only in
very few compounds, where some happen to be zinc-blende semiconductors [���]
and now this fluoride perovskite. The behavior of the polarization as a function
of the strain is expressed as follows:

X X
Pµ = eµj "j + Bµjk "j "k , (�.�)
j j,k

where Pµ is the spontaneous polarization with µ = �, �, and � for the Cartesian

components x, y, and z respectively. The strain tensor is represented by "j with
j = �, �,...., � as the components of the tensor in the Voigt notation. Finally, eµj
and Bµjk are the linear and quadratic piezoelectric coefficients respectively. If we
oriented the plane of epitaxial strain into the xz-plane, we have that "4 = "5 = "6
= � and "1 ⇡ "3 � . Additionally, the in- and out-of plane strain relationship can
be related to the Poisson ratio such as n = –("1 /"2 ). If we expand up to second
order the polarization as a function of strain in order to obtain the piezoelectric
constants allowed in the NaMnF3 we find the expression shown in Eq. �.� which
can now be expressed as follows:

⇣ e ⌘
Py = e21 + e23 - 22 "1 + ...
✓ n ◆
1 B221 + B223 B222 (�.�)
+ B211 + B233 + 2B213 - 2 + 2 "21
2 n n

The computed linear piezoelectric coefficients e21 , e23 , e22 , e16 , and e34 at
�% strain are equal to �.���, -�.���, �.���, -�.���, and �.��� C·m-2 respectively
and are not far from the ones observed in BaTiO3 [���, ���]. Nonetheless, the
computation of quadratic order components is beyond of the reach of this chap-
ter and the description and analysis of those parameters needs further studies.
Nevertheless, from the quadratic fitting observed in Fig. �.� we can extract that
the non-linear contribution related to the sum of the Bµjk component is at least
the ��% of the linear piezoelectric constants which points out the necessity for a
further study of this property in this material.

� In the Pnma structure that a and c parameters are not exactly the same and an average between
them is taken in order to apply the bi-axial strain. Then, an small plus and minus strain is
intrinsically applied to the a and c parameters respectively.
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

�.� Non-Collinear magnetism and ME coupling in

Pna�1 NaMnF3
With the idea to understand better the magnetic ordering in NaMnF3 , we per-
formed a theoretical characterization of Pna�1 phase. Based on group theory
analysis, the possible magnetic orderings and couplings were obtained. Character
table and possible magnetic couplings and orderings are presented in Table �.�.
In the Pna�1 independent of the possible magnetic states, the magnetoelectric
coupling is allowed by symmetry (Table �.�). For the Pnma phase it is expected a
predominant G-type AFM behavior, with a canted structure and a weak magne-
tization along c-axis as reported by Shane [���], however, no more experimental
or theoretical reports has been carried out to generate a deeper understanding of
the magnetic structure. Despite of the experimentally reported ground state, an
Gx Cy Az magnetic ordering structure was found over a wide range of U and J
values (from � to � eV and � to �.� eV for U and J respectively) within the PBEsol
exchange-correlation functional. The later in complete disagreement with the ex-
perimental findings [���]. Within the LDA approximation, the magnetic ground
state is fully converged and it is in complete agreement with the experimental re-
ports showing a Gz Ax Fy magnetic ordering. In the Pna�1 phase, the same ground
state remains after the structural transition from the Pnma structure and none
magnetic state transformation was observed. It is important to remark that as
explained in the computational details section, the Zeeman effective magnetic
field is only implemented within the LDA approach, thus, for the non-collinear
magnetic ordering and ME-coupling analysis, the LDA formalism is taken from
now on.
The magnetic ground state in the Pna�1 was analyzed according to the or-
derings proposed in Table �.�. Gz Ax Fy was found as the ground state and it
is depicted in Fig. �.�b. Analyzing the noncollinear magnetic structure we noted
the ferromagnetic component Fy is my = �.�� µB /atom. This FM component lies
at the same order of magnitude as expected from some weak ferromagnets such
as CaMnO3 perovskite where the FM component is my = �.�� µB /atom [���].
Nonetheless, a larger magnetic canting should be expected according to its trend
as a function of the B-site ionic radii and octahedral rotation shown in Table A.�.
Based on the magnetic point symmetry group (m’m’�) found for the magnetic
Gz Ax Fy ground state, all the magnetic, electric and magnetoelectric couplings
are permitted by symmetry within the thermodynamic potential expansion, as
shown in Ref. [���]. Thus, a second order ME can be expected for the NaMnF3 .
In agreement with the later statement, all the couplings and interactions that
affects the electric polarization for this fluoroperovskite are included and defined
in the Eq. �.� as follows:

1
Pi = ↵ik Hk + ijk Hj Hk + jik Hj Ek , (�.�)
2
 �.� non-collinear magnetism and me coupling in pna� 1 NaMnF 3 ��

where Pi is the spontaneous ferroelectric polarization along each particular

Cartesian axis (i = x, y and z). Hj is the applied magnetic field along the j axis.
↵ik is the linear magnetoelectric matrix, and ijk , jik are the second order ME
coupling tensors.

Table �.�: Allowed magnetic orderings in the D2h point symmetry group [���, ���]
for Pnma [���] and epitaxially-induced phase Pna�1 . Besides, the magneto-
electric couplings [���] for Pna�1 are shown depending to the non-collinear
ordering and the magnetic group.
Magnetic Pna�1
Ordering Character ME-linear ME-second order
Cx , Gy , Fz A2 ↵yz , ↵zy ijk , jik
Ax , Fy , Gz B1 ↵xx , ↵yy , ↵zz ijk , jik
Fx , Ay , Cz B2 ↵xz , ↵zx ijk , jik
Gx , Cy , Az A1 ↵xy , ↵yx ijk , jik

The magnetoelectric coupling was computed for several strain values according
to the implementation of finite magnetic field given in Ref. [��] and explained in
Chapter �. The results are condensed in Fig. �.� and in Table �.�. In Fig. �.� we
report the evolution of the electric polarization with respect to the amplitude of
the applied magnetic field along different directions. In Table �.� we report the
extracted ME coefficients at three different epitaxial strains. Our results reveal
that a sizeable non-linear ME coupling is present when the magnetic field is
applied along the y-axis (parallel to the weak-FM moment and the spontaneous
polarization). When the field is applied along the x-axis, we observe a linear ME
response along the same direction (↵xx ) and a non-linear one along the y-axis
( yxx ).
In comparison to Cr2 O3 that has been widely studied and in which the exper-
imental reports show a ME coupling in the range between �.�� – �.�� ps·m-1
[���, ���, ���, ���], the linear magnetoelectric response in NaMnF3 is in the
range, when the strain is close to +�%. Additionally, the ME response in this
fluoroperovskite is of the same order as in CaMnO3 [���].
Additionally, it can be noted that all the ME coefficients increase when the
strain goes from positive to negative values (for example the ↵yy coefficient
varies from -�.��� to -�.��� ps·m-1 for +�% to -�% respectively), which is
consistent with the fact that the polarization contribution of the magnetic Mn
atoms is larger for compressive strains and thus we can expect a stronger cou-
pling between electric polarization and magnetism. The global form of the ME
responses in this fluoroperovskite is very similar to the one predicted for CaMnO3
[���], which shows that this might be a general rule for strain-induced FE in the
Pnma structure. Finally, we also note that the ME response is observed in spite
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

Figure �.�: Magnetoelectric coupling in NaMnF3 in Pna�1 structure where Py vs By

(upper panel), Px vs Bx (center panel) and Py vs Bx (lower panel) are
presented. In dashed lines are presented the linear and quadratic fitting
according to the Eq. �.�. The second-order ME coupling can be appreciated
from this figure. Additionally, an increase of the non-linear behavior as a
function of the epitaxial strain is observed. The computed first and second
order ME tensor are presented in Table �.�.

Table �.�: ME coefficients of strained NaMnF3 Pna�1 expressed as in the Eq. �.�, ↵ik
in [ps·m-1 ] and jik in [⇥��-8 ps·A-1 ].
Strain, " [%] ↵yy ↵xx yyy xxx yxx
+�% -�.��� �.��� -�.��� �.��� -�.���
�% -�.��� �.��� -�.��� �.��� -�.���
–�% -�.��� �.��� -�.��� �.��� -�.���
–�% -�.��� �.��� -�.��� �.��� -�.���

of a non Mn-driven polar distortion which claims for a study of the ME-coupling
source as discussed next.
 �.� non-collinear magnetism and me coupling in pna� 1 NaMnF 3 ��

�.�.� Field-induced distortions as the source of the ME coupling

A phonon projection was carried out, then, the effect of the magnetic field and
the related modes activated as a function of it were analyzed. In Fig. �.� we plot
the overlap between the magnetic B-field induced distortions in the Pna�1 phase
(at �� T amplitude of the field along the y-axis, here denoted as | i in Fig. �.�)
and the Pnma phonon eigendisplacements basis set (|⌘i in Fig. �.�) for epitaxial
strain values of -�%, �%, and +�%. The eigendisplacements are normalized as
follows h⌘i |M|⌘i i = 1 where M is the mass matrix of the atoms.

Figure �.�: Projection of the magnetic B-field induced distortions (at �� T in the Pna�1
FE phase) against the phonon basis set of the Pnma phase for -�%, �%, and
+�% values of the epitaxial strain. We note that the B 2u FE unstable mode
(the one with negative frequencies) is strongly dominant at compressive
strain while the A g modes are more dominant at tensile strains. This fulfills
the fact that at compressive strain the FE distortions are much more Mn-
driven (and thus much more sensitive to magnetic field) than at tensile
strains and this also explains why under compressive strain the ME response
is larger than under tensile strains.

Our results suggest that the mechanism driving the magnetoelectric coupling in
the NaMnF3 is related to the activation of the Ag modes (Na+F sites) induced
by applied magnetic field (see Fig. �.�). Thus, the activation of those modes
modify the eigendisplacements of the polar B2u phonon and then, tunes the
FE polarization. This type of ME effect in which, a field-induced distortions
are dominated by different phonon modes others than the responsible for the
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

ferroelectric polarization has been previously demonstrated in the LiNiPO4 [���].

Additionally, the increase of the B2u activation under the magnetic applied field
as a function of the strain, going from +�% to -�% explains the increase in the
↵xx and ↵yy ME components.

�.�.� Effect of the U and J into the ME coupling in Pna�1 NaMnF3

In order to analyze the effect of the Coulomb repulsion and exchange interaction
on the magnetoelectric coupling, calculations for different U and J constants were
performed. In Table �.� presentes the computed constants for the ↵ik and jik
terms. It can be observed that the linear ME terms decreases by approximately �.�
ps·m-1 for ↵xx and ↵yy . However, the second-order coupling term jik increases
by one order of magnitude. Additionally, it can be appreciated that all the tensor
components decreases when the J term increase from �.� to �.� eV. In conclusion,
even when the values for the magnetoelectric coupling changes, the deviation is
rather small and the effect in the ME coupling is not very large when LDA+U
approach is used. Besides, the presence of the ME-coupling is strongly supported
by the symmetry analysis.

Table �.�: ME coefficients (↵ik and jik in ps·m-1 and ⇥��-8 ps·A-1 respectively) of
strained NaMnF3 as a function of U and J parameters [in eV] for the �%
value.
U, J [eV] ↵yy ↵xx yyy yxx
�, �.� -�.��� �.��� -�.��� -�.���
�, �.� -�.��� �.��� -��.��� -��.���
�, �.� -�.��� �.��� -��.��� -��.���
�, �.� -�.��� �.��� -��.��� -��.���

�.�.� Special remark: some ferroelectrics under strain

The B term in the Eq. �.� reflects the non-linear character of the strain-polarization
coupling of a FE crystal. In order to check whether this nonlinearity is really spe-
cific to materials presenting a steric geometric origin of the FE, we have analyzed
the polarization-strain coupling in different cases. We first looked at this response
in BaCoF4 (see Fig. �.�), which is known to be a geometric FE without anomalous
Z⇤ . Interestingly, we found the same parabolic polarization vs. strain behavior as
for NaMnF3 where the B-term is approximately ��% of the linear one.
We also report in the figure the polarization vs. strain (P vs. ") of cubic BaTiO3 ,
cubic CaTiO3 and cubic NaMnF3 (i.e. in the absence of the AFD motions in the
last two of them). We can see that in BaTiO3 the P vs. " coupling is linear while
 �.� non-collinear magnetism and me coupling in pna� 1 NaMnF 3 ��

Figure �.�: Polarization (out-of-plane) vs. epitaxial strain in the ab-plane for the geo-
metric FE BaCoF4 , for cubic BaTiO3 , cubic NaMnF3 and cubic CaTiO3
case in which we simply relax the polar distortion without the AFD motions,
and for improper-FE PbTiO3 /SrTiO3 and CaTiO3 /SrTiO3 superlattices.
We note that the strong non-linear polarization/strain coupling is present
mostly in steric geometric FE crystals such as NaMnF3 and BaCoF4 .

it is strongly non-linear for NaMnF3 (which means that the octahedral motion
do not play a determinant role for this mechanism). We note that for CaTiO3 ,
the P vs. " is linear but it also exhibits some non-linearity. We would thus con-
clude that BaTiO3 is driven by a charge transfer mechanism while in CaTiO3 a
mixed charge transfer and steric geometric effect are present. To go beyond, we
also performed the same analysis in improper-FE superlattices PbTiO3 /SrTiO3
(PT/ST) and CaTiO3 /SrTiO3 (CT/ST) and we found that in PT/ST the P vs.
" is linear and slightly non-linear in CT/ST. Improper-FE is thus not necessar-
ily related to a geometric effect in the sense of our steric definition and thus
NaMnF3 has a unique response in the perovskite family. This reinforces the idea
that the linear vs. non-linear polarization-strain coupling can be a key ingredient
to distinguish between a charge transfer FE and a steric geometric FE but further
studies would be welcome to confirm this hypothesis. Nonetheless, it seams that
this large non-linear piezoelectric response is very unique to the steric geometric
origin of the ferroelectricity.
�� induced multiferroic and magnetoelectric behavior in n a m n f 3

�.� Conclusions
Using density functional theory we studied the effect of the epitaxial strain on the
structural, electronic, and magnetic properties in the fluoride perovskite NaMnF3 .
A multiferroic-induced state was predicted for NaMnF3 under epitaxial strain at
compressive or tensile strain. A non-linear behavior of ferroelectric instability,
and then, a non-linear piezoelectric response as a function of epitaxial strain was
found. The later completely different as it has been reported in perovskite oxides.
Additionally, an out-of-plane polarization, with respect to the epitaxial strain,
was observed again in opposite trend when comparing to oxides. A cooperative
FE ordering between Na and Mn sites was found for compressive strain. The
later against a pure A-site geometrically driven ferroelectricity at tensile values of
orthorhombic ac-strain. Magnetic ordering reveals a non-collinear ground state
under the Gz Ax Fy representation. And even more interesting, and non-linear
magnetoelectric coupling was found in this material under the strained Pna�1
ground state. We hope that our results can motivate to experimentalist in the
search of this material which is, to our knowledge, the first perovskite-fluoride in
which non-linear piezoelectric and magnetoelectric response co-exist.
 Part IV

OXYFLUORIDES ABO3 /ABF3

I N T E R FA C E S

Perovskite interfaces have demonstrated to be a huge source of novel

and exciting properties that are completely absent in their bulk coun-
terparts. This has been recognized by the large number of research
performed in many different oxide interfaces. In the following we will
present a study of the unexplored fluoride/oxide interfaces.
�DEG, �DHG, FE, AND MAGNETISM
6
IN OXYFLUORIDES ABO3/ABF3
I N T E R FA C E S

Novel electronic properties in oxide/oxide interfaces can be accomplished by tun-

ing the ferroic and spin properties of the chosen component materials and guided
by the complex electron-electron interactions known in strongly correlated mate-
rial interfaces, which gives rise to a large set of intriguing phenomena (see Fig.
�.�). In oxide/oxide interfaces these phenomena include two-dimensional electron
gases (�DEG) at oxide insulator/insulator interfaces [���, ���] exotic electronic
states due to strong spin-orbit coupling with an intriguing spin-texture struc-
ture [���], metal-insulator transitions tunable by electric fields [���], interface
magnetic phases [���], improper ferroelectricity [��, ���, ���], and multiferroic
interactions [���]. While studies of these phenomena have focused only on com-
plex oxide materials, there are strong indications that complex fluorides may have
analogous, or even enhanced properties with respect to the oxide materials. In
particular, our recent computational work on fluoride perovskites, (ABF3 ) indi-
cates that interfaces between ABF3 and other fluorides, or hybrid systems with
isostructural oxide perovskites ABO3 , have unique advantages in terms of their
magnetic, magneto-transport, and multiferroic properties [��]. To date little ex-
perimental work has been devoted to fluorides, either in bulk [���, ���] or in
thin film forms [���] and therefore there is very limited knowledge about their
surfaces, and interface properties. Then, the aim of our work is to overcome this
lack of knowledge by performing a first-principles investigation in fluoride-based
interfaces and superlattices. This, will lead to a novel strategies for searching new
and complex perovskite fluoride-based magneto-electronic materials.

In addition to the studies in bulk fluorides, much less is known about the mag-
netoelectric properties of interfaces and surfaces of complex fluorides. There are
a few reported examples of complex fluorides thin films and the corresponding
interface characterization. These include Fe/KFeF3 bilayers [���] and the trilay-
ered system Fe/KNiF3 /FeF2 [���] that has a dynamic and rotatable exchange
anisotropy. These systems exhibit exchange bias phenomena, which results in a
center shift of the hysteresis loop of the ferromagnetic Fe layer away from H =
0, as a result of the magnetic interface interaction with the antiferromagnetic
material.
Recently, several efforts have been dedicated to study and to understand mixed
systems by considering the effect of fluorine replacement by oxygen ions in ox-

��
�� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces
Introduction: Electronic phenomena at interfaces

Charge transfer
KNiF3

Strain
KTaO3 Charge
Ch
harge
–
Orb
Orbital
Frustration +

Annu. Rev. Condens. Matter Phys. 2011.2:141-165. Downloaded from www.annualreviews.org

KNiF3 Spin
Lattice

Symmetry

by 190.71.56.244 on 12/29/12. For personal use only.

breaking
Electrostatic
KTaO3 coupling

Figure 2
The rich behavior of transition metal oxides arises from the complex interactions between their charge,
orbital, spin, and lattice degrees of freedom. These interactions can be modified at interfaces between
different oxides through the effects of local symmetry breaking, charge transfer, electrostatic coupling,
strain, and frustration, leading to fascinating new phenomena.

Figure �.�: Some of the physical phenomena that have lead to several multifunctional LaAlO3/SrTiO3 system. The close proximity in energy of different structural phases and

properties found in oxides/oxides interfaces and expected in the mixed ox-

the typically large polarization-strain coupling of perovskite compounds make them highly
sensitive to the epitaxial strain generated at the interface between two materials with
different lattice parameters. Application of biaxial strain through epitaxy has become a

ide/fluoride (KTaO3 and KNiF3 in the figure) interfaces. The diagram prop- popular method for controlling the lattice distortions, which in turn influence the elec-
tronic degrees of freedom, providing a powerful tool for adjusting materials’ properties
and stabilizing alternative ground states. The advances in the field of strain engineering of

erties has been taken from Zubko et al. [���]. oxide heterostructures are outside the scope of this review, and the interested reader is
referred to References 34–37.
When polar materials are involved, electrostatic interactions also come into play, and the
P. Zubko et. al, Annu.
boundary Rev. Condens.
conditions Matter
at the interfaces lead toPhys. 2, 141–165
an electrostatic coupling(2011).
between the layers,
which is discussed in some detail in Section 4.5. In addition, competition between different
interactions at interfaces and frustration, for example of spins, can also result in qualitatively
Previous reports in oxyfluorides: Moon, E. J. et al.. J. Am. Chem. Soc. 136, 2224–2227 new(2014).
structural and electronic phases. These concepts of symmetry breaking, charge transfer,

ide perovskites with different stoichiometries. Thus, a process called “fluorina-

Andrés Camilo García Castro, [email protected]
electrostatic coupling, strain, and frustration are referred to throughout the following discus-
Predoctoral, Querétaro-México, Nov. 2015
sion of the fascinating phenomena observed or predicted at interfaces between complex oxides
or even within a single material (see sidebar on domain boundary engineering).

tion” has been successfully applied to epitaxial growth of SrFeO3-x Fx (x from An important issue to consider is that of length scales. Although a particular phenomenon
may have its origin at the interface, it is not necessarily confined to it. Unscreened Coulomb

0 to 1) thin films [���] where substantial changes in the electric conductivity www.annualreviews.org ! Complex Oxide Interfaces 147

and transport properties has been demonstrated, mainly due to an induced

change in the nominal oxidation state of the Fe cation. Then, an even more sur-
prisingly, multifunctional properties such as superconductivity in Sr2 CuO2 F2+
[���], ionic conductivity at BaFeFx O3- [���], and robust antiferromagnetism in
Ba0.8 Sr0.2 FeFx O3- [���] can be tuned or modified in fluorinated oxides —also
called oxyfluorides—.
Even though, there are only a few publications in relation to fluoride-based
interfaces meanwhile, the reported results on bulk systems shows that a research
in this topic could open an opportunity to find interesting and appealing prop-
erties. Therefore, we propose a series of important questions that we will try to
address: Is there a presence of another interesting electronic phenomena such as
�DEG, or polar catastrophe in fluoride/oxide interfaces in analogy with some
oxide/oxide structures? if that is the case, is the mechanism the same? What
is the influence and effect of the O–B–F bonding on the electronic properties,
the crystal-field splitting and the magnetic ordering? Is there a possibility to
change the confinement of the �DEG at the interface due to the fluoride layer
and its lower covalence? Is there an occurrence or enhancement of ferroelectricity
in fluoride-based systems at the interface? Is there a presence of improper ferro-
electricity due to interfaces in superlattice structures as it has been reported in
some oxide systems? What is the influence of magnetic ordering in the possible
ferroelectric behavior at interfaces? What is the multiferroic (or magnetoelectric
if exist) behavior at the interface and then at the superlattices? Then having all
these unanswered questions and more to come on the road, we present in the next
sections our results that aim to shed light onto these novel heterostructures.
 �.� computational details ��

�.� Computational Details

In this chapter, two different codes were used based in their particular abilities
to compute several properties.
VASP [��, ��]— We used PAW [��] pseudo-potentials to represent the va-
lence and core electrons. The electronic configurations taken into account in the
pseudopotentials as valence electrons are presented in Table �.�. The exchange
correlation was represented within the GGA - PBEsol parametrization [��]. The
magnetic character of these system was include and d-electron localization cor-
rected by means of the GGA+U within the Liechtenstein formalism [��], for
which, the U value was found by fitting the electronic gap and magnetic moment.
The periodic solution of these crystalline structures was represented by using
Bloch states with a centered gamma k -point mesh of �⇥�⇥x (x dependent of the
c parameter longitude of the computational cell) and ��� eV energy cut-off, which
has been tested already to give forces convergence to less than �.��� eV/Å. Spin-
Orbit coupling was include such that non collinear magnetic configurations can
be attained [��]. Born effective charges and phonon calculations were performed
within the density functional perturbation theory (DFPT) [��].

Table �.�: Electronic configuration of pseudopotentials used in the analysis of this chap-
ter. Valences and semicore electrons taken into account for calculations.
A-site B-site X-site
K (�p6 �s1 ) Ta (�p6 �d5 �s2 ) O (�s2 �p4 )
Na (�p6 �s1 ) Zn (�d10 �s2 ) F (�s2 �p5 )
La (�s2 �p6 �d1 �s2 ) Ni (�p6 �d8 �s2 )
Mn (�p6 �d5 �s2 )
Sc (�p6 �d1 �s2 )

CRYSTAL��— Due to the fact that some of the interesting interfaces count
with a large number of atoms and in order to describe correctly the strongly cor-
related nature of the electrons, hybrid calculations have been performed using the
linear combination of atomic orbitals (LCAO) as implemented in the CRYSTAL
code [��]. This code uses a Gaussian-type basis set to expand the Kohn-Sham or-
bitals, including polarization effects, with the advantage that less number of the
basis is required, which impact into the requirements of the computational time
used to calculate strongly correlated materials when compared to plane wave ba-
sis set as it is the case of VASP. Additionally, all the electrons have been included
in K, Ni, Zn, O, and F while a pseudopotential method has been used for the
case of Ta. For the exchange-correlations term, the hybrid functionals taken into
account are the B�WC [��], B�LYP [���, ���], and PBEsol0 (derived from the
PBE0 functional but based in the PBEsol approximation) [��] all of them with a
�� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

��%, ��%, and ��% of exact exchange-correlation term respectively (see Chapter
�).

�.� Oxyfluorides KNiF2.5O0.5 and KT aO2.5F0.5

As commented previously, oxyfluorides have been used in several technological
applications based on the tuning of their properties by means of the control of the
fluorine/oxygen ratio. Here, as a first approach to the mixed X-site systems with
oxide/fluoride interfaces, we analyzed the electronic properties of the KTa(O/F)3
and KNi(F/O)3 oxyfluorides. The precursor compounds (i.e. KTaO3 and KNiO3
respectively) exhibit a cubic ground state at room conditions. Thus, just one atom
of the X-site is replaced, which is equivalent to a composition of KNiF2.5 O0.5
and KTaO2.5 F0.5 respectively. This composition is graphically understood as the
replacement of one X-site in the MX6 octahedra as shown in Fig. �.� and �.�.
Both oxyfluorides were fully relaxed by means of the PBEsol approach. It can be
observed that the KTaO2.5 F0.5 oxyfluoride present a tetragonal expansion with
lattice parameter a = b = �.��� Å and c = �.��� Å. In contrast to this compound,
the KNiF2.5 O0.5 present a tetragonal compression with a = b = �.��� Šand c
= �.��� Å. These tetragonal distortions can be understood based in terms of the
bonding strength and length, which, in the oxide case is stronger and shorter
in comparison to the metal–fluoride bonding. This is also responsible for the
polarization induced in the cell due to the off-centering of the B-site.

KTaO2.5F0.5 KNiF2.5O0.5

Figure �.�: Localized-DOS for the �d-Ni and �d-Ta at the KNiF2.5 O0.5 and
KTaO2.5 F0.5 oxyfluorides respectively. This DOS was computed by means
of the hybrid functional HSE06.

As expected, the different octahedral coordination has an effect in the electronic

structure due to the change in the oxidation states of the central B-site metal
cations. For example, the Ta-site is going from a +5 oxidation to a +(5 - ) state
—equivalent to a d-orbital occupation d(0+ ) —. Accordingly, the Ni-site is going
from +2 to a +(2 + ) oxidation state —equivalent to an occupation d(8- ) —
as observed in Fig. �.� and �.�. This in agreement with the experimental results
 �.� oxyfluorides KNiF 2.5 O 0.5 and KT aO 2.5 F 0.5 ��

in SrFeO3-x Fx thin films [���]. Then, the tetragonal distortion and the change
in the electronic occupation of the metal cations can induce a tetragonal (T )
splitting of the d-orbitals.
In order to understand in more detail, we computed the localized DOS by means
of the hybrid functional HSE06 (Heyd-Scuseria-Ernzerhof) [��]. This functional
is known to be more precise in the electronic structure calculations correcting the
problems due to the a lack in the description of the exchange-correlation energy
(see Fig. �.�). In KTaO2.5 F0.5 the lowest orbitals are the dxz/yz followed by dxy .
These orbitals exhibit a partial occupation that is in agreement with the expected
Ta oxidation state as discussed previously. The electronic configuration of the d-
orbitals for the Ta are depicted in Fig �.�. In the KNiF2.5 O0.5 oxyfluoride we
can see that the lowest occupation belongs to the dxy orbital, followed by the
dxz/yz orbitals. It is important to note that the dxz/yz is slightly occupied,
which is induced by the d(8- ) Ni state. Next, we can observe a half occupied
dx2 -y2 orbital and right after, around -� eV, the dz2 orbital. This electronic
configuration corresponds to the T -compression configuration as depicted in Fig.
�.�. A magnetic moment is observed in the Ni case and the ordering remains
G-type AFM as expected from the former KNiF3 . Nonetheless, in the Ta-based
case, we do not observe a magnetic moment for the partial �d occupation.
Electronic Density of States

Tetragonal z-expansion: Ta(5-d)+ Density (states/eV)

(Ta-O)xy: 1.995 Å 7.0

6.5
6.0

KTaO2.5F0.5 dx2-y2 (Ta-O)z: 1.944 Å 5.5

5.0
4.5

5d:eg (Ta-F)z: 2.183 Å 4.0

3.5

dx2-y2 dz2 dz2 a = b = 3.984 Å

3.0
2.5
2.0

c = 4.127 Å
1.5
1.0 Electronic Density of States

Tetragonal z-expansion: Ta(5-d)+

0.5
Density (states/eV)
0.0

dxy (Ta-O)xy: 1.995 Å −0.5

7.0
−1.0
6.5
−1.5
6.0

KTaO2.5F0.5 5d:t2g dx2-y2 (Ta-O)z: 1.944 Å −2.0

5.5
−2.5
5.0

dyz , dxz
−3.0
4.5

5d:eg (Ta-F)z: 2.183 Å −3.5

4.0
−4.0
3.5

dxydx2-yd2yz ddz2xz dz2 a = b = 3.984 Å

−4.5
3.0
−5.0
2.5
−5.5
2.0

KTaO3 KTaO2.5F0.5
−6.0

c = 4.127 Å
1.5
−6.5
1.0
0.5
−10.0 −8.0 −6.0 −4.0 −2.0 0.0
0.0 ElectronicEnergy
Density
(eV)
of States
−0.5

d(2+d)+
Density (states/eV)

Tetragonal z-compression: Ni xyand crystal-field splitting of the

−1.0

Figure �.�: Electronic configuration corresponding

−1.5
80.0

5d:t2g
−2.0
75.0

KNiF O (Ni-F)xy: 2.007 Å

−2.5
70.0

KTaO 3 oxide and the KTaO2.5 F0.5 oxyfluoride.

2.5 0.5 dyz , dxz
−3.0
65.0
−3.5
60.0

dxy dyz dxz 2

dz (a1g) (Ni-F)z: 2.024 Å −4.0
55.0
50.0
−4.5
45.0
−5.0

3d:eg (Ni-O)z: 1.813 Å

40.0
−5.5
35.0

KTaO3 KTaO2.5dF
−6.0
30.0
20.52
x -y (b1g)
−6.5

a = b = 4.005 Å
25.0

dx2-y2 dz2
20.0 −10.0 −8.0 −6.0 −4.0 −2.0 0.0
15.0 ElectronicEnergy
Density
(eV)
of States
10.0

Tetragonal z-compression: Ni(2+d)+ c = 3.837 Å Density (states/eV)

5.0
0.0
−5.0
80.0

dyz , dxz (eg)

−10.0
75.0

KNiF2.5O0.5 (Ni-F)xy: 2.007 Å −15.0

70.0
−20.0
65.0

3d:t2g (Ni-F)z: 2.024 Å

−25.0
60.0
−30.0
55.0

ddz2xy(a(b1g2g) ) −35.0
50.0
−40.0
45.0

(Ni-O)z: 1.813 Å
−45.0

3d:eg dxy dyz dxz

40.0
−50.0
35.0
−55.0
30.0

dx2-y2 (b1g)
−60.0

a = b = 4.005 Å
25.0
−65.0

dx2-y2 dz2
20.0
−70.0
15.0

KNiF3 KNiF2.5O0.5
−75.0
10.0

c = 3.837 Å 5.0 −7.0 −6.5 −6.0 −5.5 −5.0 −4.5 −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.
0.0 Energy (eV)
−5.0

dyz , dxz (eg)

−10.0
−15.0

dxy3d:t2g dxz dyz dx2 dz2

−20.0
−25.0
−30.0
Tokura, N. Nagaosa, Y. Science (80-. ). 28
−35.0

dxy (b2g) −40.0

−45.0

dxy dyz dxz −50.0

−55.0

z
−60.0
−65.0
−70.0

KNiF3 KNiF2.5O0.5
−75.0
−7.0 −6.5 −6.0 −5.5 −5.0 −4.5 −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0.0 0
Energy (eV)

Figure �.�: Electronic dand dxz configuration

xy magnetic dyz xdz2
dx2 corresponding to the crystal-field Tokura, N. Nagaosa, Y. Science (80-. ). 28
y
splitting for the KNiF3 fluoride and the KNiF2.5 O0.5 oxyfluoride.
z

x
y
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

Therefore, we should expect these type of distortions and alterations of the

electronic structure close to the interface when oxides and fluorides are displaced
for mixed systems which could induce an unobserved electronic behavior.

�.� KT aO3 /KBF3 (B = Zn and Ni) superlattices

As a first fluoride/oxide superlattices, we considered two different pristine materi-
als: KTaO3 /KBF3 with B = Zn and Ni. These structures are formed by a polar lay-
ered oxide with (TaO2 )+1 and (KO)-1 layers and a neutral fluoride with (MF2 )0
and (KF)0 layers. Superlattices with ratios m/l were used where m=l= �, �, . . . ,
� with alternating p- (Zn+2 F-1 2 /K
+1 O-2 ) and n-type (Ta+5 O-2 /K+1 F-1 ) in-
2
terfaces (see Fig. �.�). It is observed a considerable increase of the polar distortion
of the B-site with respect to the X-site apical planes for all the superlattices. The
later can be concluded from Fig. �.� for the �/� case taken as example. This dis-
tortion is larger at the TaO2 /KF and ZnF2 /KO interfaces and then, it is almost
constant up to the next interface. The off-centering shows opposite directions for
the Ta and Zn respectively, suggesting opposite polarization directions within the
KTaO3 and KZnO3 layers. KTaO /KZnF
3 U = 0eV
3 (6/6)

(KZnF3)m/(KTaO3)n
Insulator
m/n: 1/1

Metal m/n
Ta+5O2-2/K+1F-1 Zn+2F2-1/K+1O-2
m/n: 7/7 n-type p-type

B
Figure �.�: �/� and �/� superlattices at which the insulator-to-metal transitions takes
place. The TaO2 /KF and ZnF2 /KO interfaces can be observed.

Additionally, the structural characterization of the superlattices shows a large

c-lattice parameter deviation in addition to the large displacements of the B-sites
for each oxide and fluoride cases as commented before. For example, the K–K
lattice that contains the TaO2 /KF interface (see Figs. �.� and �.�) exhibits a
large expansion from �.�� Što �.�� Šas shown in Fig. �.�. At the same time,
the ZnF2 /KO interface induces a strong compression of the lattice parameter
from �.� Å to �.� Å, this can caused by changes in the electronic environment
and coordination as discussed in previous oxyfluoride section.
 �.� KT aO 3 / KBF 3 ( B = Zn and Ni ) superlattices ���

KZnF3

ZnF2 /KO
TaO2/KF

ZnF2/KO

TaO2 /KF
KTaO3

KTaO3

KZnF3

Figure �.�: Atomic displacements of the B-sites along the z-axis (left) with respect to
the oxygen apical plane for the �/� superlattice. K–K c-lattice parameter in
which, a large expansion and compression can be observed at the TaO2 /KF
and ZnF2 /KO interfaces respectively (right).

In order to understand better the effect of the correlation in this superlattice

system, we analyzed the electronic structure of the �/� superlattice ratio by means
of the HSE06 hybrid functional implemented in VASP [��]. We obtained the
localized density of states for KTaO3 /KZnF3 and KTaO3 /KNiF3 as shown in
Fig. �.�. For the non-magnetic structure (B= Zn), we noticed that the electronic
gap of �.� eV is close to the experimental value for the KTaO3 bulk compound
(�.�� - �.�� eV [���]). It can be observed that the valence bands close to the
Fermi energy (here at 0 eV by notation) are dominated by oxygen p-orbitals and
the fully occupied Zn states lies below -� eV. The conduction bands are mainly
composed by the �d:Ta orbitals and a small contribution of the oxygen sites due
to the p–d hybridization. In the magnetic case (i.e. B = Ni) the electronic band
gap is still close to the KTaO3 value, besides, the valences states are composed
by the O and half occupied Ni states. The empty states are formed by �d–Ta in
addition to the �d–Ni (eg ) levels.

(KZnF3)1/(KTaO3)1 (KNiF3)1/(KTaO3)1

Figure �.�: Localized density of states for M = Zn and Ni in the �/� superlattice ratio
obtained considering the HSE06 hybrid functional.

As a next step, we analyzed the magnetic moment per Ni-atom as a function of

U (within the GGA+U approach) and we correlated the results with the findings
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

obtained based in the hybrid functional calculation. We found that the correct
electronic structure and magnetic moment is expected for U values above �.� eV.
Besides, without the use of an U value within the GGA calculations, the unoccu-
pied Ni states are located between the valence oxygen states and the conduction
Ta bands, which represents a wrong description of the electronic structure when
is compared to the one obtained from the use of the hybrid functional.
Within the hybrid functional calculations, we also found a weak magnetic mo-
ment of m = �.��� µB /atom localized on the Ta cations. These moments are
following a FM interaction along z with respect to the Ni sites ordered antifer-
romagnetically with m = �.��� µB /atom in agreement with a Ni+2 . This is ex-
plained by the strongest FM interaction between half-occupied �d–Ni eg and the
empty �d–Ta [���] rather that an AFM interaction driven by the super-exchange
interaction between the Ni–O–Ta or Ta–F–Ni.
The dynamical and vibrational stability of the �/� superlattice was investi-
gated by computing the phonon-dispersion curves. We found no unstable mode
in the full Brillouin zone, which suggests a stable structure of the superlattice
configuration as shown in Fig �.�.

�.�.� �DEG and �DHG in KTaO3 /KBF3 superlattices:

The basic superlattice structure is shown in Fig. �.� for B = Zn where alternating
n- and p-type interfaces can be appreciated. The insulator-to-metal transition
(IMT) is a function of the number of layers taken into account in the calcula-
tions. This is in full agreement with the “polar catastrophe” scenario expected
from unbalanced charge at the Zn+2 F-1 2 /K
+1 O-2 and Ta+5 O-2 /K+1 F-1 inter-
2
faces [���]. Then, after several number of layers we found the presence of two-
dimensional electron gases (�DEG) and a two-dimensional holes gases (�DHG)
in KTaO3 /KBF3 superlattices induced by the polar catastrophe and the local
electronic density reconstruction which, is discussed in detail below.
First, we found a decreasing band gap with an increasing of the number of layers
in each compound with a stoichiometry (KTaO3 )n /(KBF3 )l , where n and l are
the number of atomic layers in each compound. This occurred mainly due to the
electrons and holes donated by each layer at the interface and the compensation
of the electrostatic potential generated by the additional electrons in the system
due to the O–B–F bondings. The insulator-to-metal transition was observed for
the n = l = � in both cases. It is important to note that the transition is achieved
with less polar/non-polar layers than in oxides systems such as LaAlO3 /SrTiO3
in which, the IMT is observed for values above ��/�� ratio [���]. Another point
that is important to remark is the localization of the �DEG and �DHG (Fig. �.�)
that, according to our results comes from the dxy –Ta and O p-orbital respectively.
Thus, the fluoride film is not actively having an effect close to the Fermi energy in
the electronic structure and the electrons and holes are only present at the oxide
film. Then, the fluoroperovkite film is acting as a “condensating” layer retaining
the �DEG and �DHG at the oxide layers. Additionally, from the electrostatic point
 �.� KT aO 3 / KBF 3 ( B = Zn and Ni ) superlattices ���

(KZnF3)7/(KTaO3)7 (KNiF3)7/(KTaO3)7
Zn+2F2-1/K+1O-2 Ni+2F2-1/K+1O-2
p-type p-type

KZnF3 KNiF3
G-type AFM

TaO2-2/K+1F-1 Ta+5O2-2/K+1F-1
n-type n-type

KTaO3 KTaO3

~ 2eV ~ 2eV

Figure �.�: Layered revolved DOS for the KTaO3 /KZnF3 and KTaO3 /KNiF3 (�/�)
superlattices where a �DEG and �DHG are observed at the TaO2 /KF and
MF2 /KO.

of view a larger electric field across the entire system was found in both cases
in comparison to the LAO/STO oxide/oxide counterpart. For example, when B
= Zn: E = ��� mV/Šagainst E = �� mV/Š[���]. At the metallic interface we
observed that the effective masses in both cases (B = Zn and Ni) are close to
�.�me for the electrons at the �DEG that lies in the lowest �dxy :Ta band. In
contrast, heavier hole carriers with m⇤ of �.�me at the �DHG localized at the
p-orbital of the oxygens at the KO layer. Then, in the �DEG we have a high
mobile carriers at the interfaces, lighter that in the SrTiO3 /LaAlO3 [���]. In
order to explain the electrostatic behavior and the formation of the �DEG and
�DHG we presented the Band-edge diagram built for this system in Fig. �.�. We
can see that the conduction fluoride band-edge lies well above the Fermi level,
besides, the valence band is almost ��� meV below the Fermi energy. Then, it is
again observed that the �DEG and the �DHG are localized at the oxide layer.

�.�.� Magnetic ordering in KTaO3 /KNiF3 superlattice:

Now, we study the magnetic ordering as another degree of freedom in the oxyflu-
orides system for the case of B = Ni. Our results suggest a non-collinear mag-
netic ordering at NiF2 /KO interface, which resulted in a large net magnetization
around �.� µB /atom near the interface for the �/� superlattices at low U values.
Nonetheless, for U values above the fitted one (U > � eV) the magnetic ordering
converged to a pure G-type AFM.
The O–B–F bonding is one of the main points and differences with respect
to other oxides-based superlattices that completely introduces a new degree of
freedom to the oxyfluoride interfaces never explored before. This bonding induces
strong differences in the electronic structure and orbital distribution at the in-
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

(ZnF2/KO)-1 (TaO2/KF)+1 (ZnF2/KO)-1

CB

E VB
2DHG 2DEG
EF

KTaO3 KZnF3

KTaO3/KZnF3 U=0eV
Figure �.�: Band-edge
IMPORTANT:diagram for the electrostatic behavior of the system where the
electric field profile and the �DEG (�DHG) formation are presented at the
Even with an U = 6eV in the KTaO inducing an increase of 1eV in the gap, the system is still metal.
3
Then, the polar catastrophe is still taking place and the electric field goes from 105 to 140 meV/Å. (in
oxyfluoride
comparison to TaO
57 meV/Å /KF
2 in LaAlO(ZnF
3 /SrTiO ). /KO) interface. This was obtained based on the
3
2
layered resolved
Can we say DOS
that the fluorides (see
is still Fig
insulator but �.�).
the oxideCB, VBmaterial?
is the metal an EF denotes the conduction-
band, valence-band and the Fermi energy respectively.

terfaces as it can be observed in Fig. �.�. In this figure we can see the breaking
of the degeneracy of the d-orbitals due to the tetragonal (T ) like distortions. Be-
sides, the d-orbital configurations for both types of interfaces can be the source of
the particular magnetic behavior. Due to the strong electronic correlation in the
partially occupied �d Ta orbitals and the T Ni states at the interfaces as shown
previously, we can see that the magnetic ordering at the interface is changing
with respect to the U value. We can observe, for the �/� superlattice, that the
magnetic state goes from a pure G-type AFM at U =� eV to a C-type AFM plus a
FM layer at the interface for U = �.� eV. To get a closer look in the disagreement,
calculations based on hybrid functionals were carried out.
In order to corroborate the results related to the magnetic ordering and because
of the limitations in the computational efficiency of the calculation with hybrid
functionals within the plane-wave expansion, we used the CRYSTAL code. Using
this code, we selected three different approaches within the hybrid approximation:
B�WC, B�LYP, and PBEsol0. Several magnetic orderings were tested, G-, C-, A-
AFM respectively and mixed G-AFM+FM layer (at the NiF2 immediately after
the TaO2 /KF interface and at the NiF2 /KO layer). Nevertheless, according to
the calculations the magnetic ordering always converges to the G-type AFM as
ground state in the KNiF3 film. These results are in full agreement with those
obtained with the GGA+U approximation for U values above � eV. However,
it is important to mention that the mixed G-AFM+FM-layer (at the NiF2 /KO)
state is very close in energy, E = �.�� meV/atom, closer than other magnetic
ground state (i.e. A- and C-AFM).
Interestingly, there is a magnetic-moment modulation across de entire thickness
of the magnetic counterpart in all the superlattices taken into account. This
modulation can be observed in Fig �.�� where variations between a range of �.��
µB /atom are expected.
Even if a charge-transfer, charge-ordering, and orbital ordering is not achieved
in these interfaces due to the small valence deviations at the interfaces (as dis-
 �.� KT aO 3 / KBF 3 ( B = Zn and Ni ) superlattices ���

Figure �.��: Magnetic moment per layer at the KTaO3 /KNiF3 superlattices. A moment
modulation can be appreciated when n/l ratio is varied from �/� to �/�.

cussed in the oxyfluoride section) we have to keep in mind that fluorides can be
better candidates for a charge-ordering state where a larger charge-disproportion
cation is observed [���] in comparison to their oxide counterparts.

�.�.� Isolated p- and n- interfaces:

With the main idea to understand the isolated p- and n-type interfaces, we have
studied both terminations. A KO/ZnF2 and TaO2 /KF terminated computational
cells were used respectively where a KO plane is set at the center as a mirror
symmetry plane of the full heterostructure. Then a �� u.c. was used, �–KTaO3
and �–KZnF3 respectively. After full relaxation we have found magnetism and
interestingly, this magnetism is coming from two different sources. In the p-type
interface, p-orbital magnetism has been found in agreement with the electron
vacancies and the �DHG formed at the interface (see Fig �.��). The magnetic
structure shows a AFM-type ordering with mT a = �.���µB /atom and mO =
�.���µB /atom. On the other hand, in the n-type interface a �DEG is clearly
formed as shown in Fig �.��. This in addition to the d-orbital magnetism coming
from the Ta:dxy (see inset in Fig �.��) orbital in a A-type AFM with a magnetic
moment mT a of �.���µB /atom at the surface and a �.���µB /atom at the layer
right below. From these interfaces it can be noted as well that the effective electron
masses for hole and electron carriers are �.�me and �.�me respectively. Thus, the
n-carriers are lighter and with a high mobility perfect for electronic applications
[���]. It is important to note that the fluoride layer (i.e. in this case KZnF3 )
is not actively participating in the phenomena, however, is acting as insulating-
confining layer.
We presented in Fig. �.�� the bands for the case of the magnetic fluoride layer.
We can observe that the carriers in the n-type interface are slightly heavier in
comparison to the non-magnetic case (i.e. m⇤ = �.�me against m⇤ = �.�me ).
For the p-type case, the hole carrier present almost the same behavior with m⇤ =
�.�me . Additionally and surprisingly, in the n-type interface we observed a charge-
ordering at the third lower layer with respect to the surface. This is concluded
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

Figure �.��: Bands structure for the p- (left) and n-type (right) interfaces along the
X–M– –X path in the Brillouin zone for the KTaO3 /KZnF3 system. The
formation of the �DHG and �DEG at the p- and n- interfaces can be
appreciated.

from a change in the magnetic moment from -�.��� µB /atom to �.��� µB /atom
suggesting an d8 – d6 occupation. This could explain the deeper penetration of
the �DEG below the Fermi level, indicating a larger carrier density in the B =
Ni n-type interface than in the B = Zn one.
In the p-type interface Fig. �.��, we do not observe a charge-ordering. We
found that the magnetism at the KNiF3 remains in its G-type AFM ordering in
addition to the p-orbital magnetism. The later is also observed in the oxyfluoride
interface explained previously with a non-magnetically active B-site.
In comparison, even when the carriers in the n-interface are heavier for B = Ni,
it is important to mention that the �DEG is more pronounced, which in this case
lies at -�.� eV against -�.�� eV for B = Zn. Knowing that the in-plane lattices
parameters and polarization across the interface are quite the same, it can be
suggested that the charge missing at the KNiF3 layer is transferred directly to
the �DEG increasing the amount of density carriers at the �DEG in the interface.
However, further studies and analysis are needed to explain and exploit this
particular behavior in the KTaO3 /KNiF3 n-interface oxyfluoride.

�.� Cubic-Rashba spin-splitting at oxyfluoride in-

terface
Recently it was claimed that in oxide interfaces such as SrTiO3 /LaAlO3 [���,
���, ���, ���] the appearance of spin-splitting and then, a complex spin-texture
related to a Rashba-effect [���] is achievable. Besides, some reports have appeared
 �.� cubic-rashba spin-splitting at oxyfluoride interface ���

Figure �.��: Bands structure for the p- (left) and n-type (right) interfaces along the
X–M– –X path in the Brillouin zone for the KTaO3 /KNiF3 system. The
�DHG and �DEG at the p- and n- interfaces can be appreciated. Besides,
and empty Ni:eg band is observed between �.� to �.� eV respectively.

showing a similar behavior in SrTiO3 - and KTaO3 -based transistors [���, ���].
Then, for these single perovskite-based devices, a k-cubic dependence of the split-
ting was found (see Eq �.�) in contrast to the linear plus cubic dependence of the
Rashba splitting at the LaAlO3 /SrTiO3 interface due to the dxy – dxz/yz multior-
bital nature of the lowest bands [���]. Additionally, in the case of Au(���) metal
surface, it has been shown that even when the surface states exhibit a mainly
p–orbital character, the d-orbitals drives the Rashba splitting and dictates the
direction of the spin by means of the orbital-angular momentum (OAM) [���, ���].
Therefore, the transition metal d-orbitals at perovskite structures can be the key
parameter for a large and controllable electron band splitting in future spintronic
and spin-orbitronic devices [���, ���]. This later property gives an additional de-
gree of freedom to the already rich field of multifunctional perovskites. Then, in
order to look for the spin-texture and spin-orbit coupling effects in the oxyfluoride
interfaces, the analysis of the electronic structure is done by taking into account
the SOC. Due to the large computational cost, the study of the bands was per-
formed in the �/� superlattice. Notice that these results can be traced up to the
�/� system where, the �DEG and �DHG is formed and the IMT takes place as
shown previously. Thus, the results presented here can be extrapolated to larger
n/l superlattices. in Fig �.�� it can be observed the band structure along the
kx path in which, the sx , sy , and sz spin components are presented separately
and a colors notation of the up (red) and down (blue) spin was adopted. First, it
can be noted that the electronic structure and the spin-texture at the conduction
bands above to Fermi level (and thus, close to the TaO2 /KF interface at the �/�)
are entirely �D spin-polarized at the kx –ky plane. Therefore, no band-dispersion
is observed for the sz spin component. Surprisingly, a large spin-splitting of the
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

bands is appreciated for an energy around E = �.� eV. Additionally, the SOC
effect is observed at the heavy dxz/yz crossing with the dxy lighter bands (see
the upper inset in Fig �.��).

Δk = 0.08 π/a
ΔE = 61 meV

sx sy sz
2π/a 2π/a 2π/a

Figure �.��: Band structure for the k x path in which, s x , s y , and s z spin components
are taken into account. The up and down orientation of the spin are repre-
sented by red and blue colors respectively.

The observed Rashba splitting is of the cubic-type character as observed in

oxide surfaces and interfaces and is expressed by the Hamiltonian presented in
Eq. �.�:

HR3 = ↵3 Ez i(k3- + - k3+ - ), (�.�)

where ↵3 is the cubic-Rashba coupling term which measures the strength of the
splitting (in units of eV·Å3 ), ± = �/�( x ± i y ) and x,y are the Pauli matrices.
The Ez is the electric field perpendicular to the electron plane movement denoted
by k± = (kx ± iky ) [���]. Here, the E vs. k profile for the bands is represented
by the relationship:

h2 k2
E± (k) = ± ↵3 k3 . (�.�)
2m⇤
This type of spin-splitting can be understood like an opposite rotation of the
lowest spin–up band with respect of the spin-down around an axis fixed at the
vertices of the parabolic d-bands. This is in contrast with the linear horizontal dis-
placement (or offset) of the spin–up and spin–down bands respectively observed
in the most common linear-Rashba effect� present in materials such as BiTeI
[���].
� The linear-Rashba effect has the E vs. k form: E± (k) = (h2 k2 /2m⇤ ) ± ↵|k| coming from the
Hamiltonian of the form HR = ↵Ez i(k- + - k+ - )
 �.� cubic-rashba spin-splitting at oxyfluoride interface ���

In order to get an idea of the order of magnitude of the spin-splitting, we

compare it with similar systems as follows. At the energy value of �.� eV we found
that k = �.��� Å-1 which is more than twice of the one found at the KTaO3 -
based transistor (i.e. Au/parylene-Al/KTaO3 ) [���]. Besides, the vertical energy
difference between the up and down spins is E = �� meV at the same bands
as shown in Fig �.��. In comparison with an oxide/oxide interface we computed
the spin-texture of the SrTiO3 /LaAlO3 system under the same conditions in a
�/� superlattice. We found that k = �.��� Å-1 and a E = �� meV. Then, this
spin-splitting is four-times smaller in comparison to �.��� Å-1 and �� meV at
the oxyfluoride interface.
Going further and with the aim to understand these results we first computed
the spontaneous polarization, which is directly related to the electric field Ez in
Eq. �.�. The layer-by-layer polarization is presented in the Table �.�. It is observed
that the polarization is in average close to -�.� µC·cm-2 per KTO layer. For
this case, n = l = �, the total polarization is -��.�� µC·cm-2 at the oxide side.
Moreover, the polarization is in the opposite direction of the fluoride layer with a
value of +�.�� µC·cm-2 . Then, at the end, a total polarization of -�.�� µC·cm-2
is still present along the entire superlattice structure. This big difference suggests
that the depolarizing field is not enough to compensate the polarization along the
entire supercell. Besides, the opposite directions of the polarizations is explained
in the breaking of symmetry due to the O–M–F bonding along z. Therefore, by
increasing the number layer (i.e. the n/l ratio), the KTaO3 polarization will
increase and then, is expected that the spin-splitting will be incremented directly
it is in agreement with the Eq �.�. Ultimately, it is important to note that the
SOC coupling in the Ta is at least �� times larger that the one in the Ti cation.
This can be taken as advantage in the mixed oxide/fluoride interfaces due to the
large lattice parameters of the fluorides perovskites that allows the incorporation
of bigger perovskite oxides with �d orbitals in novel superlattices and interfaces
with mixed properties.

Table �.�: Layer-by-layer spontaneous polarization along the z-axis (Pz in µC·cm-2 ) in
the �/� superlattice is shown. Besides, the oxide and fluoride polarizations
are PKT aO3 = -��.�� µC·cm-2 and PKZnF3 = �.�� µC·cm-2 respectively.
KTaO3 KZnF3
Layer � � � � � � � �
Pz -�.�� -�.�� -�.�� -�.�� �.�� �.�� �.�� �.��

From the magnetic superlattice, KTaO3 /KNiF3 we noticed that the same spin-
texture remains at the TaO2 /KO. Nonetheless, the spin-texture is becoming more
complex due to the overlapping of the �d:Ta band with the unoccupied �d(eg ):Ni
states above the Fermi level.
It is important to take advantage of the fact that the magnetic properties
can be conserved in the KNiF3 . Then, a mixed multifunctional properties of
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

�DEG+�DHG+k3 -Rashba-splitting+G-AFM are condensed in the same heterostruc-

ture where a possible A-site driven ferroelectricity can be engineered in systems
such as KTaO3 /NaMnF3 .
It is also important to remark that the relevance of the k3 -Rashba dependence
has been highlighted in the field of multifunctional properties based in the signif-
icant difference in the effective field symmetry between the k-linear and the k3
Rashba. Thus, the k3 symmetry influences all of the SOC related phenomena in
materials, which, is not the case of the the k-linear Rashba term. For example,
in case of the spin Hall effect, it has been predicted the k-cubic Rashba term can
give rise to a larger spin Hall conductivity [���, ���].
The possible tuning of the energy position of the �d:Ta bands with respect
of the Fermi energy by controlling the n/l ratio stand as an additional feature
of these systems. Therefore, the amount of splitting and consequently, the spin-
transport properties can be tuned as well. The later making even more interesting
and appealing this type of interfaces.

�.� Charge-ordering, magnetism, and ferroelectric-

ity in LaScO3/NaMnF3 superlattice
As demonstrated for the multiferroic properties in NaMnF3 and the electronic
properties in superlattices and interfaces of KTaO3 /KBF3 , in fluoride-based sys-
tems, the tuning of properties can be achieved independently. For example, the
FE order in the NaMnF3 is tuned by epitaxial strain and driven by the A-site,
meanwhile the magnetism is raised from the B-site avoiding the d0 -ness rule. An-
other example of this behavior is the magnetic properties that remains in the
KNiF3 and the �DEG and �DHG localized at the KTaO3 in the KTaO3 /KNiF3
heterostructure. Then, an additional effort was carried out in order to tune sev-
eral properties into a single superlattice. This is the case of the LaScO3 /NaMnF3
system where MF/ME and �DEG can be expected coming from the fluoride and
oxide layer respectively.
In this heterostructure, we used at the beginning the ground state at room
conditions of both perovskites. Both belongs to the orthorhombic Pnma (space
group No. ��) for which, experimental lattice parameters of a = �.��� Š(�.���
Å), b = �.��� Å (�.��� Å), and c = �.��� Å (�.��� Å) were used for the LaScO3
(NaMnF3 ) in the calculations. Then, an equivalent strain of �.�% and �.�% for the
a- and c-axis is expected respectively at the interface. This due to the approach
in superlattices for which the ac-plane is fixed to the average between the lattice
parameters of the two compounds for each axis.
As a starting point, we begin the analysis with the �/� superlattice. For this
superlattice ratio after full electronic relaxation (with U = � eV at the Mn cation),
an insulator state remains with an electronic band gap close to ��� meV was
observed. We expect an IMT above the �/� superlattice, in which, according
Electronic properties in LaScO3/NaMnF3
�.� charge-ordering, magnetism, and ferroelectricity in LaScO 3 / NaMnF 3 super

to the previous results an interesting spin-texture and spin-splitting should be

observed at the Sc d-orbitals.

LaScO3 NaMnF3

-2.6 μB/Mn
-4.6 μB/Mn

4.6 μB/Mn

-4.6 μB/Mn
4.6 μB/Mn

4.6 μB/Mn

4.6 μB/Mn
-4.6 μB/Mn

4.6 μB/Mn

-4.6 μB/Mn

2.6 μB/Mn

-4.6 μB/Mn

4.6 μB/Mn

-4.6 μB/Mn
A. H. Romero, [email protected] PACIFICHEM, December 15-20, Hawaii
Figure �.��: LaScO3 /NaMnF3 (�/�) superlattice. The AFD distortions and rotations
can be observed in the upper figure in which the KO/MnF2 and ScO2 /KF
interfaces can identified. In the lower figure it is shown the spin-polarized
charge density, red and blue colors denote the up and down magnetic mo-
ment directions. The G-type AFM ordering is observed as well as the CO
state.

The results shows a charge-ordering (CO) state found at the middle Mn layer
in the fluoride NaMnF3 part (see Fig. �.��). The CO state is demonstrated by a
change in the magnetic moment from �.� µB /atom to �.� µB /atom as shown in
Fig. �.��. This moment difference suggests an d5 – d3 Mn occupation. Nonethe-
less, for a U = � eV (value normally used in Mn at oxides and fluorides [�, �, �]),
the CO-state remains but now, the magnetic moment at the middle of the layer
goes from �.� µB /atom to �.� µB /atom. The later claims for a deeper study and
understanding of the exchange-correlation in this particular oxyfluoride. We ob-
served that this CO state remains after several U values (from � to � eV). Besides,
the main magnetic ordering exhibit a G-type AFM after testing several possible
states (i.e. G-, C-, and A-type AFM and FM).
The analysis of the spontaneous polarization reveal values of Pz = ��.�� µC·cm-2
and Pxy = �.�� µC·cm-2 . Here, the Pz is induced by the symmetry breaking at
the KO/MnF2 and ScO2 /KF interfaces and the F–Mn–O and O–Sc–F bondings
respectively as can be seen in Fig. �.��. This polarization cannot be reversed,
nonetheless, can be used to enhance the polar catastrophe phenomena and tune
the �DEG as demonstrated in SrTiO3 /PbTiO3 [���], in this interface the polar
catastrophe is neglected due to the neutral character of the atomic layers. How-
��� �deg, �dhg, fe, and magnetism in oxyfluorides abo 3 /abf 3 interfaces

ever, the ferroelectric polarization along the PTO layer induce a �DEG. It is
important to note that possibly this polarization can be coupled with the strain-
induced Pz in the NaMnF3 as shown in the Chapter �. The Pxy component is
explained in terms of the hybrid-improper FE phenomena where different A-sites
at the orthorhombic interfaces brakes the anti-polar displacements and a net po-
larization can be obtained in the xy-plane [���]. This type of displacements can
be observed in the Table �.� in which are presented the atomic displacements per
layer at the xy-plane

Table �.�: A -site displacements per layer in the xy -plane taking the high-symmetry
position in the cubic symmetry as reference. Displacements in Å units.
LaScO3
-�.�� +�.�� -�.�� +�.�� -�.�� +�.�� -�.��

NaMnF3
+�.�� -�.�� +�.�� -�.�� +�.�� -�.�� +�.��

This preliminary results are very promising but further studies are needed in
order to unveil the geometric+hybrid-improper ferroelectricity in these superlat-
tices. Besides, the possible coupling of the FE with the CO-state and the magnetic
degrees of freedom needs to be investigated.
The n/l ratio effect for several n and l values and the study of the �DEG above
the �/� superlattice ratio is on their way.

�.� Conclusions
We have analyzed by means of first-principles calculations the electronic and struc-
tural properties of oxyfluorides (KTaO3 )n /(KBF3 )l B = Zn and Ni interfaces. We
found that the orbital levels symmetry at the interfaces is strongly modified as a
function of the O–B–F coordination. The polar catastrophe phenomena also takes
place in the oxyfluoride interfaces, however, we found that less number of layers
are needed in order to achieve the insulator-to-metal transition. We observed that
the magnetism in the KTaO3 /KNiF3 exhibits a moment magnitude modulation.
Nevertheless, the magnetic structure keeps the G-type AFM such as in the bulk
former compound. Surprisingly, we observed a large k3 -Rashba type splitting in
at the oxyfluoride interfaces, at least four times larger than the one reported in
SrTiO3 /LaAlO3 interface and twice of the KTaO3 -based transistor.

Looking for novel oxyfluoride multiferroics, the LaScO3 /NaMnF3 was studied.
This interface show promising preliminary results where charge-ordering, G-type
AFM, hybrid-improper ferroelectricity and possible tuning of the electronic band
 �.� conclusions ���

gap by means of the polar-catastrophe can be achieved.

CONCLUSIONS AND
7
PERSPECTIVES

In this thesis we have explored several aspects of fluoride-perovskites, for which

we have employed state of the art first-principles calculations. The main achieve-
ments of this work were: i) To understand the vibrational properties of fluoride
perovskites. ii) To analyze the effect of external agents as the pressure and biaxial
strain. iii) To study novel oxide/fluoride interfaces and understand their proper-
ties. Therefore, in this investigation we have particularly found that:

The fluoride-perovskites ABF3 -type materials (A = Li and Na, B = Ca, Zn,

Mn, Ni, and V) exhibit a huge potential to be ferroelectric with a large polar
instability. This ferroelectricity, at first glance potentially similar to that of oxide
perovskites, revealed to have a different origin, being based on a A-site geometric
instability. On the top of ferroelectricity, this is suggesting the possibility to find
new multiferroics within the fluoride family. We then predicted that this could
indeed be realized in NaMnF3 thin films under appropriate epitaxial strain. In
these films, we discovered an unprecedented non-linear polarization/strain cou-
pling, related to the geometric origin of ferroelectricity. We have predicted that
the strained NaMnF3 can exhibit a unique multiferroic state driving large and un-
common second-order piezoelectric and non-linear magnetoelectric responses. In
a close collaboration with experimentalists, this fluoroperovskite has been success-
fully grown and its multiferroic properties are under experimental confirmation.
Interestingly, we observed that the strong nonlinear polarization-strain coupling is
only observed for the steric geometric FE crystals, which confirms that this prop-
erty is unique to geometric FE systems. The latter conclusion extracted when
comparing the geometric FE BaCoF4 , the regular FE BaTiO3 , and improper FE
PbTiO3 /SrTiO3 superlattice.

Independently, based on DFT calculations and symmetry analysis, we have also

shown that ABX3 post-perovskite phases, with an active magnetic B-site cation,
have the possibility to present a non-collinear magnetic behavior. With these
findings, we clarified some of the magnetic properties experimentally observed
in this particular high-pressure form of perovskites found in the Earth’s mantle.
Additionally, we identified a similar perovskite to post-perovskite’s transition in
NaMnF3 under isotropic pressure, yielding a phase in which the non-collinear
ferromagnetism shows a large canted magnetization.

���
��� conclusions and perspectives

We have also analyzed from first-principles the electronic and structural proper-
ties of oxyfluorides (KTaO3 )n /(KBF3 )l superlattices (B = Zn and Ni). We found
that the orbital splitting at the interfaces is strongly dependent of the O–B–F co-
ordination. Moreover, a polar catastrophe phenomena, similar to that reported at
the LaAlO3 /SrTiO3 interface, also takes place in the oxyfluoride polar interfaces.
We observed that the thickness required to achieve the insulator-to-metal transi-
tion is however smaller than in the oxides. We observed that the magnetism in the
KTaO3 /KNiF3 system exhibits a modulation of the magnetic moment amplitude
across the entire superlattices. Surprisingly, we reported a large k3 -Rashba-type
splitting at the oxyfluoride interfaces, at least four times larger than that reported
in SrTiO3 /LaAlO3 interface and twice that of the KTaO3 -based transistor.

As demonstrated for the multiferroics properties in the NaMnF3 and the elec-
tronic properties in superlattices and interfaces of KTaO3 /KBF3 , in fluoride-
based systems, the tuning of properties can be achieved independently. For ex-
ample, the FE in the NaMnF3 is tuned by the A-site and the magnetism from
the B-site avoiding the d0 -ness rule. Another example of this behavior is the mag-
netic properties that remains in the KNiF3 and the �DEG and �DHG localized at
the KTaO3 in the KTaO3 /KNiF3 system. Then, an additional effort was carried
out in order to tune the magnetic and electric properties. This is the case of the
LaScO3 /NaMnF3 systems in which, MF/ME and �DEG can be expected from
the fluoride and oxide layer respectively.

On the side of what we presented in the main manuscript we also performed side
projects that are presented in the Appendices. In Appendix C, we present the
study of the multiferroic behavior of BaCoF4 thin films under epitaxial strain
where we found a non-linear dependence of the spontaneous polarization as a
function of the external constrain. Besides, several magnetic transitions were ex-
plained in terms of the non-collinear magnetic ordering. In Appendix D we present
a study of the electronic and particularly, spin-texture properties of the SrBO3
(���) with B = Ti and Hf surfaces. We found a tangible Rashba-type splitting
for B = Hf, besides, despite the reported giant-splitting for B = Ti, we did not
found such spin-splitting. Finally, in Appendix E we report the insulator-to-metal
transition control and tuning at the LaAlO3 /SrTiO3 interface by means of OH
and H molecules implantation at the LaAlO3 surface.

Additional perspectives of this work cover also the field of superconductivity

based in the mixed valence of the oxyfluorides [���]. For example, this is the case
of the superconductivity in oxyfluorides such as Sr2 CuO2 F2+ [���, ���] where
the Cu present a valence close to +(�- ) and has a mixed octahedral coordination
in a Cu-O4 F2 octahedra. Then, we believe that it is possible to achieve a high-
Tc superconductor in systems such as KTaO3 /KCuF3 and/or LaCuO3 /KCuF3 .
In these systems, higher Tc than the one found in KTaO3 (Tc = �� mK [���])
could be obtained. Additionally to superlattices, thin films can be also analyzed
 conclusions and perspectives ���

in order to study the surface effects and the isolated interfaces that are formed
depending on the chosen substrate. For example future studies can be focused on
SrTiO3 (���)/NaMnF3 (���) and LaMnO3 (���)/NaMnF3 (���) thin films.

Finally, we believe that the discovery of new multiferroic materials could lead
to new possible solutions for the challenges in the field, even though the present
result does not fully achieve the “ideal” target of room temperature magneto-
electrics. Moreover, we hope that the results we have presented in this thesis will
motivate further theoretical and experimental studies in the search of new and
exotic multifunctional properties in fluoride and oxyfluoride-based perovskites.
CONFERENCES AND MEETINGS

In the next section are listed the conferences and meeting in which, the thesis
results were presented:

– XIII Meeting on Ferroelectricity, Talk: Novel magnetoelectric and non-linear

ferroelectricity-strain coupling in NaMnF3 fluoroperovskite, Porto, Portugal, from
�� June to � of July, ����.

– VI Workshop Chile-México about Magnetism, Nanoscience and their Applica-

tions, Talk: Unexpected Large Canting and Non-collinear Magnetism in Fluoride-
Perovskites ABF3 by First-principles, Cancún, from �� to ��th of May, México,
����.

– ���� Workshop on Fundamental Physics of Ferroelectrics, Talk: Geometric

Ferroelectricity in Fluoro-Perovskites, Washington DC, from �� to ��th of Jan-
uary, USA, ����.

– Invited talk: Multiferroic Phenomena with Geometry Ferroelectricity in Fluo-

roperovskites, Physics Department, West Virginia University, USA, February �th ,
����.

– Response Treatment for Dynamical Properties of Materials with the ABINIT

Package, Zürich, from �� to ��th of October, Switzerland, ����.

– IV Workshop Chile-México about Magnetism, Nanoscience and their Appli-

cations, México-City, from �� to ��th of March, México, ����.

���
 Part V

APPENDIX

The next appendix list condenses the work related to scientific col-
laborations that lead to an important learning in my PhD formation
but, that are not directly related with the main topic of the thesis.
Nonetheless, the study keep its focus on the electronic structure of
fluorides and some perovskites.
 O V E RV I E W O F T H E M A G N E T I C
A
INTERACTIONS IN
FLUOROPEROVSKITES

Now that we have explored the vibrational landscape of the fluoroperovskites,

we move forward by analyzing the magnetic behavior in these materials. As it
is well known, most of the fluorides perovskites present a magnetic ordering due
to their �d B-site magnetically active cation. We focus in this section on the
non-collinear magnetism, weak-FM interaction due to a magnetic canting. First,
in the Table A.� we present the weak magnetic moment for several oxides and
fluoride perovskites most of them in the orthorhombic Pnma structure, except for
the RbFeF3 . In this space group, the magnetic ordering exhibit a non collinear
Gz Ax Fy configuration, which corresponds to a G-type AFM along the z-axis, and
A-type AFM along x and a weak-FM along y.

Table A.�: Magnetic canted moment Fy (µB /atom) at low temperature in Pnma ABF3
systems. The large magnetic canting for fluorides compounds can be noted
in the Gz Ax Fy configuration. The ionic radius of the magnetic B-site (rB )
was taken into account the valence state and the bonding coordination [��].
Compound Fy Structural parameters
ABX3 U =� eV Exp. [���] d-Occ. rB [pm] a [deg]

KMnF3 �.��� �.��� d5 �� �.��

NaNiF3 �.��� �.��� d8 �� ��.��
NaCoF3 — — d7 ��.� ��.��
NaFeF3 �.��� �.��� d6 �� ��.��
NaMnF3 �.��� — d5 �� ��.��
RbFeF3 b — �.��� d6 �� —
CaMnO3 �.��� �.��� d3 �� ��.��
LaFeO3 �.��� �.��� d5 ��.� ��.��
YCrO3 �.��� �.��� d3 ��.� ��.��

a Computed form experimental lattice parameters.

b RbFeF3 in hexagonal structure [��].

In Table A.� is presented a summary of the weak-FM component, experimen-

tal and computed in oxides and fluorides. Here are also included the d-orbital

���
��� overview of the magnetic interactions in fluoroperovskites

Table A.�: Experimentally reported exchange interaction (J) constants in ABF3 com-
pounds [���]. The components of the DM vector are in meV.
ABX3 J1 J2 Structure
KMnF3 �.��� �.��� Pnma
KFeF3 �.��� — Pm3̄m
KNiF3 �.��� — Pm3̄m
KCuF3 ��.��� �.��� I�/mcm
NaFeF3 �.��� �.��� Pnma
NaMnF3 �.��� �.��� Pnma
LaFeO3 �.�� �.�� Pnma
BiFeO3 �.�� �.�� R�c

occupation, B-site radii, and the tilt angle, . For fluorides it can be noted that
the rotation angle are almost twice of the ones found in the oxides (see Fig. A.�).
Additionally, the weak-FM follows a trend with the octahedral angle: higher is the
angle, larger is the Fy magnetic moment. At the same time, the octahedral rota-
tion increases when the B radii does it is shown in Table A.�. The later magnetic
behavior is in agreement with the Dzyaloshinskii-Moriya interaction [���, ���],
then: Dij / x̂i ⇥ x̂j where x̂i,j are the unitary vector along the M–X–M bondings.
We can observe the reported weak-FM component for the NaFeF3 is rather large,
around �.��� µB /Fe, following the rotation and radii trend. Therefore, the ex-
pected Fy component in the NaMnF3 should be around �.� µB /Fe, which is one
order of magnitude large that the ones found in perovskite oxides. Unfortunately,
the computed values are not in well agreement with the reported experimental
moments. Then, a careful research within the DFT+U +J needs to be carried
out in future studies. The RbFeF3 compound shows the largest Fy component,
nonetheless, its ground state belongs to an hexagonal structure and the structural
differences needs to be taken into account in the analysis.
As expected from the weak covalent bonding in fluorides, comparing with ox-
ides, the exchange-constants, Ji are smaller by at least one order of magnitude
as it is presented in Table A.�. The KCuF3 compound exhibit a quite large J1
component due to the Jahn-Teller distortion presented in the partially occupied
d9 -orbital. Surprisingly, the KNiF3 shows large exchange interaction. These low
exchange interactions can explain the low magnetic transition temperatures ob-
served in fluoride perovskites. Therefore, even when the wake ferromagnetism
tend to be larger than in oxides, the temperature regime for this particular mag-
netic behavior seems to be very low. The later statement needs to be taken into
account in the application of these type on materials in further technological appli-
cations. Nonetheless, compound with several phenomena, such as the Jahn-Teller
 overview of the magnetic interactions in fluoroperovskites ���

Figure A.�: Octahedral rotation for several oxides and fluorides compounds. The
fluoroperovskites are ordered according to the B-site radii.

distortion can be used in order to increase the magnetic exchange interactions,

and then, the magnetic transition temperatures.
G R O U P T H E O RY A N A LY S I S T O T H E
B
NON-COLLINEAR MAGNETISM IN
PPV

The first step into the analysis is to identify the space group and its point group.
In the case of the post-perovskites we have a Cmcm space group (No. ��) with a
D2h point group, this point group have the symmetry operations: I, -I, 2z , 2y ,2x ,
mz , my , and mx extracted from the Bilbao Crystallographic Server [���, ���].
Then, we have:

h i
I = 1 1 1
h i
-I = -1 -1 -1
h i
2z = -1 -1 1 + ⌧(0, 0, 1/2)
h i
2y = -1 1 -1 + ⌧(0, 0, 1/2)
h i
2x = 1 -1 -1
h i
mz = 1 1 -1 + ⌧(0, 0, 1/2)
h i
my = 1 -1 1 + ⌧(0, 0, 1/2)
h i
mx = -1 1 1

Additionally to all the previous operations, a translational operation of [⌧(1/2,

1/2, 0)] needs to be applied in order to obtain the total of �� allowed symmetry
operations. Now, we note that in our unit cell we have four different magnetic
sites at the crystallographic positions as follows:

Table B.�: Crystallographic positions of the magnetic cations, M at the post-perovskite

Cmcm.
Atom M1 M2 M3 M4
Coordinate (0, 0, 0) (0, 0, 1/2) (1/2, 1/2, 0) (1/2, 1/2, 1/2)

���
��� group theory analysis to the non-collinear magnetism in ppv

Now, we analyze how the magnetic sites transform under the allowed symmetry
operations it can be observed that, for example, the M1 cation transforms to the
M2 under the 2z operation as follows:

2 3 2 3 2 3 2 3
-1 0 0 0 0 0
6 7 6 7 6 7 6 7
2z · M1 = 4 0 -1 05 · 405 + 4 0 5 = 4 0 5 ⌘ M2 (B.�)
0 0 1 0 1/2 1/2

Therefore we have that all the magnetic sites transforms under the symmetry
operations as follows obtaining the Table B.�:

Table B.�: Transformation of the magnetic sites under the symmetry operations. The
numbers in the table are the i index of the Mi magnetic cation.
I 2z 2y 2x -I mz mx mx
M1 � � � � � � � �
M2 � � � � � � � �
M3 � � � � � � � �
M4 � � � � � � � �
+ ⌧(�/�, �/�, �)
M1 � � � � � � � �
M2 � � � � � � � �
M3 � � � � � � � �
M4 � � � � � � � �

Here, at this point, we know how the magnetic sites transform under the sym-
metry operations. We now proceed with the analysis of the magnetic moments
transformation. For this purpose, we have to define the possible magnetic config-
urations into the collinear level. The selected notation is F for ferromagnetic, and
A, C, and G for A-, B- and G-type antiferromagnetic orderings. Then, we have:
In Table B.� the + and - signs are for up and down spin orientation respectively.
We proceed by calculating how the spin transforms under the same symmetry
operations such as the magnetic sites does. Therefore, we have to include the
non-collinear orderings by defining the spin pseudo-vector such as s = (sx , sy ,
sz ). Due to the pseudo-vector character of the spin and in order to understand
how it transforms, we need to be defined such pseudo-vector as the product of A
(where A = (x, y, z)) and B (with B = (x 0 ,y 0 ,z 0 )) as follows: s = A⇥B. Then, s
= (yz 0 - zy 0 , zx 0 - xz 0 , xy 0 - yx 0 ).
 group theory analysis to the non-collinear magnetism in ppv ���

Table B.�: Definition of the different magnetic orderings. The G, A-, and C-type anti-
ferromagnetic orderings and the ferromagnetic F state are defined for each
magnetic cation where + and - denoted up and down spin orientations.
M1 M2 M3 M4
F + + + +
A + + – –
C + – + –
G + – – +

To exemplify how proceed, we show how the spin will transform under the 2z
operation as follows:

2 3 2 3 2 3 2 3
-1 0 0 yz 0 - zy 0 -yz 0 + zy 0 -1
6 7 6 0 7 6 7 6 7
2z · s = 4 0 -1 05 · 4 zx - xz 0 5 = 4 -zx 0 + xz 0 5 ⌘ 4-15 · s (B.�)
0 0 1 xy 0 - yx 0 xy 0 - yx 0 +1

Then, after applying to all the operations we have that the spin transforms as
condensed in Table B.�. It is important to note that the s transforms under the
-I as -s = B⇥A. Nonetheless, the spin must be exchanged and undistinguible,
therefore, it transforms in the same way for the -I as for the I symmetry operation.
The later statement is also valid for the mz , my , and mx .

Table B.�: Transformation sx , sy , and sz spin orientations under the symmetry opera-
tions. The transformation of each magnetic ordering under each symmetry
operation is labeled as +� and -�, which indicates when the ordering is invari-
ant under the related transformation (i.e. +�) or when it is reversed under
the application of the symmetry operation (i.e. -�).
I 2z 2y 2x I mz my mx
sx +� –� –� +� +� –� –� +�
sy +� –� +� –� +� –� +� –�
sz +� +� –� –� +� +� –� –�

Finally, to obtain how the non-collinear magnetic orderings transforms we take

the Tables B.�, B.� and B.�. The symmetry characters are obtained as follows:
For example, for the Cx component under the 2z operation, originally –� from
Table B.�, we see, from Table B.�, that the M1 transforms into M2 , and then
looking at Table B.� we observed that changes from + to -. Therefore, the Cx
under the 2z symmetry operation is reversed and +� is obtained. The characters
for each ordering along each Cartesian axis are presented in the Table B.�.
��� group theory analysis to the non-collinear magnetism in ppv

Table B.�: Transformation of the magnetic ordering under the allowed symmetry oper-
ations.
I 2z 2x 2x -I mz my mx
Fx +� –� –� +� +� –� –� +� B3g
Fy +� –� +� –� +� –� +� –� B2g
Fz +� +� –� –� +� +� –� –� B1g
Ax +� –� –� +� +� –� –� +� B3g
Ay +� –� +� –� +� –� +� –� B2g
Az +� +� –� –� +� +� –� –� B1g
Cx +� +� +� +� +� +� +� +� Ag
Cy +� +� –� –� +� +� –� –� B1g
Cz +� –� +� –� +� –� +� –� B2g
Gx +� +� +� +� +� +� +� +� Ag
Gy +� +� –� –� +� +� –� –� B1g
Gz +� –� +� –� +� –� +� –� B2g

By re-grouping the orderings into the same symmetry character we obtain the
Table B.�.

Table B.�: Magnetic ordering allowed in the Cmcm structure according to the D2h point
group [���, ���].
Magnetic Ordering I 2y 2y 2x -I mz my mx
Fx , Ax +� –� –� +� +� –� –� +� B3g
Fy , Ay , Cz , Gz +� –� +� –� +� –� +� –� B2g
Fz , Az , Cy , Gy +� +� –� –� +� +� –� –� B1g
Cx , Gx , +� +� +� +� +� +� +� +� Ag

To obtain the possible non-collinear states we take into account the superex-
change interaction in order to avoid unprovable orderings. Besides, we neglect the
ferrimagnetic orderings. Then, finally we have that the allowed states are Cz Fy ,
Cz Ay , Cy Fz and Gy Az .

For more details in the group theory analysis applied to the non-collinear mag-
netism the reader can check the Ref. from Bertaut [���].
BaCoF4 THIN FILMS
C
The orthorhombic fluoride family BaMF4 , with M = Mn, Fe, Co, Ni, and Zn
[���, ���, ���] was intensively studied four decades ago [��] due to their interest-
ing multifunctional properties and physics discovered. These fluorides exhibit si-
multaneously antiferromagnetic and ferroelectric long-range orders. While the an-
tiferromagnetic order has a pronounced two-dimensional character, the ferroelec-
tricity mechanisms are fundamentally different from the “classical” charge-transfer
phenomena widely known in ferroelectric perovskites oxides such as BaTiO3 and
PbTiO3 [��, ��]. Recently, new theoretical analysis [��, ���] that includes a pre-
dicted magnetoelectric effects present in some of those materials (for example in
BaNiF4 ) motivated further experimental studies on those multiferroic systems.
As commented before, the BaMF4 compounds crystallize in orthorhombic lattice
with the space group Cmc�1 No. �� [��]. In this structure, a shared-corners dis-
torted fluorine octahedra M–F6 in the ac-plane form puckered sheets, which are
separated by Ba layers along the crystallographic b-axis [��, ���]. A few years ago
was theoretically described the ferroelectric long range order for the family mem-
bers M = Mn, Co, Fe and Ni which has its origin in a combination of octahedral
rotations in M–F6 and polar displacements of Ba cations, in a so-called geomet-
ric proper ferroelectricity [���] making it very appealing to tune multifunctional
properties. Previous experimental reports have revealed ferroelectric switching in
two members of the family, BaCoF4 and BaNiF4 with ferroelectric saturation
polarization along the a-axis and Ps = �.� µC·cm-2 and �.� µC·cm-2 , respec-
tively [���]. However, no switching has been observed for BaFeF4 and BaMnF4 ,
nonetheless, the analysis by Ederer and Spaldin [��] proposed that the ferroelec-
tric switching in those materials is likely possible by means of higher electric
fields. From the magnetic point of view, magnetic moments of M2+ have their
strongest antiferromagnetic interactions in the ac-plane, with nearly linear M–F–
M–F chains along the c-axis, and a zig-zag M–F–M–F configuration along the
a-axis [���, ���]. Since the interplanar spacing along the b-axis is much larger
(i.e. about �� Å) than the intraplanar ones (about �-� Å), the antiferromagnetism
in BaMF4 has a pronounced two-dimensional character. Spins are oriented along
the b-axis [���] except in BaCoF4 , where they lie along the a-axis. It is important
to remark that two antiferromagnetic phases have been found in case of BaCoF4
[���]: phase A and phase B being distinguished by the sign of the interplanar cou-
pling along the b-axis, negative and positive, respectively (see Fig. C.�). Then,
in this appendix we present the calculations and analysis that support the exper-
imental synthesis of multiferroic BaCoF4 fluoride thin films which are for first

���
��� b a c o f 4 thin films

time achieved. For more details in this work please check the Ref. of Borisov et al.
[�].

c.� Computational Details

First-principles calculations were performed within DFT as implemented in VASP
[��, ��]. PAW [��] pseudo-potentials were used to represent the valence and core
electrons. The electronic configurations taken into account in pseudo-potential
as valence electrons were Ba (�s2 �p6 �s2 ), Co (�p6 �d7 �s2 ) and F (�s2 �p5 ). The
exchange correlation was represented within the GGA approach with PBEsol
parametrization [��]. The periodic solution of these crystalline structures was rep-
resented by Bloch states with a Monkhorst-Pack with a k-point mesh of (�⇥�⇥�)
and ��� eV energy cut-off, which has been previously tested to give converged
forces to less than �.���� eV/Å. Additionally, the correct description of the d-
electron localization was achieved through the DFT+U approximation with U =
�.� eV [��]. Spin-orbit coupling (SOC) in the non-collinear analysis was included
according to the implementation by Hobbs et al. [��]. In order to take into ac-
count the AFM magnetic interactions along the a-axis, a (�⇥�⇥�) (�� atoms)
cell was used for all the calculations. Finally, the spontaneous polarization was
computed within the Berry phase approach as implemented in VASP [��, ��].
Isotropic strain calculations were carried out for strain values ranging between
-�% and +�%, calculated as (d - d0 )/d0 , where d and d0 are the strained and
optimized theoretical Cmc�1 bulk lattice parameters, respectively. The strain was
applied with the same magnitude and sign to a and c unit cell directions, allowing
relaxation along the b-axis as well as for the inner coordinates, until forces and
the stress along the b-axis vanished.

In the next sections we will discuss the results regarding the effect on the
xz-plane due to an applied epitaxial strain in BaCoF4 thin films. Thus the multi-
ferroic state is detailed analyzed by means of the employed theory. In particular,
the spontaneous polarization as the non-collinear magnetism as a function of the
strain is studied.

c.� Strain Effect in BaCoF4 Fluoride:

In order to understand the ferroelectric and magnetic behavior of BaCoF4 thin
films, the density functional theory calculations were performed. According to
the experimental findings, when BaCoF4 grows on Al2 O3 substrates [���] ori-
entation is the prefer one, which defines the specific crystallographic directions
where the theoretical part should focused. Such as it is well known, the strain
can induce phase transitions in materials when deposited into a substrate en-
hancing their properties with a considerable mismatch, even less than �%. This
is the case of SrTiO3 [��], CaTiO3 [���, ���] and NaMnF3 [�, �], the first two ox-
 C.� strain effect in b a c o f 4 fluoride: ���

ide perovskites and originally paraelectrics, which under epitaxial strain becomes
ferroelectric. The fluoride case is rather different due to its larger sensitivity to
strain in which a paraelectric to ferroelectric transition is achieves for values of
strain less than �.�%. Then, keeping fixed the ac-lattice plane to the experimen-
tal values, internal coordinates and z-axis are allowed to fully relax. After this
structural relaxation a new phase for the strained BaCoF4 was found, going from
the Cmc�1 (symmetry group No. ��) to Pca�1 (symmetry group No. ��). In this
new phase, the ferroelectric behavior remains unchanged due to the polar nature
of the structure as in the Cmc�1 phase. This transition in driven by atomic dis-
tortions (represented as red arrows) according to the 1 mode as shown in Fig.
C.�.

Figure C.�: BaCoF4 [���] supercell in which the red arrows represent the distortion to
go from the Cmc�1 to Pca�1 space groups. Atoms in green, blue, and grey
colors for Ba, Co, and F species respectively. This figure was elaborated
with the support of the VESTA code [���].

As a next step and with the goal to test the thermal stability of the new BaCoF4
phase, phonon dispersion calculations were performed in the relaxed strained
structure with lattice parameters such as found in thin films (approx. +�.�%).
No imaginary modes traduced in unstable modes were found, which suggest a
stable Pca�1 phase structure for this value of strain. After the later analysis, we
proceeded to compute the ferroelectric and magnetic properties in the relaxed
BaCoF4 structures for strains between -�% to +�% to complete the landscape
of the multiferroic properties as a function of the epitaxial strain such as in
experimentally growth thin films.
Fig. C.�, we present the behavior of spontaneous polarization as a function of
the ac-plane strain. It can be observed the non-linear coupling of the polarization
as a function of the strain, which suggest a non-linear piezoelectric effect in this
fluoride compound, which is quite rare in these materials [�]. As expected, the
polarization develops along the crystallographic b-axis such as has been found
in bulk compounds [���]. Nonetheless, small components of polarization along
x and z are observed but are not shown here (with an order of �.� µC/cm2 ).
��� b a c o f 4 thin films

Figure C.�: (Top) Spontaneous polarization along the b-axis. a, b, and c lattice param-
eters (Middle) and behavior of the octahedral angle bonding (Co–F–Co)
along the x-axis (bottom) as a function of the ac epitaxial strain respec-
tively.

This can be due to the additional distortion of the CoF6 octahedra induced
by the epitaxial strain. For values of an epitaxial strain close to those observed
in the BaCoF4 /Al2 O3 (����) thin films, which is close to �.�%, we compute a
polarization value of Ps = �.�� µC·cm-2 . Additionally, Fig. C.� also shows that
the b lattice parameter decrease as a and c increase as can be expected from the
Poisson ratio.
Regarding the magnetic ordering in the collinear approach, three different con-
figurations have been proposed. These configurations are labeled as A, B and C
respectively as depicted in the Fig. C.�,. We found that the B-phase is the most
stable, however, the A and C are at �.�� and �.�� µeV/f.u. larger in energy. This
fact suggests an easy switching of the ordering achieving by small fields and even
temperature with that takes an AFM to AFM magnetic transition.
 C.� noncollinear magnetism and multiferroic behavior ���

Figure C.�: From left to right, the A, B and C collinear magnetic orderings. In green,
blue navy and grey are Ba, Co and F ions respectively. Magnetic moments
in each atom are located along the y-axis. In the figures can be observed
the unit cell in black lines as well as the computation magnetic cell that
takes into account �� atoms.

c.� Noncollinear Magnetism and Multiferroic Be-

havior
The calculations presented in this section were obtained taking the B-phase type
magnetic ordering as the main magnetic structure. Based on this structure and
the octahedral tilting in the BaCoF4 , a weak-AFM is allowed as a slave ordering
of the main AFM with a z -easy axis. Nonetheless, the strain induces an increase
of the Co–F–Co bonding angle along the x-axis. Additionally, within the Pca�1
space group, several magnetic point groups such as m�m, m�m�, m’�m’, m�’m’,
m’�’m are allowed and possible some of them can exhibit weak-FM (see note �).
Taking this into account, a weak-FM ordering can be also observed in agreement
with the experimental results that show a weak-FM behavior at low temperatures
below �� K. Then, three different orderings can be found of each of the Cartesian
axis, a G-type AFM as the main AFM ordering with easy axis along z, a weak-
AFM along y-axis as a first slave ordering (called C-type) and a second weak-FM
slave along x-axis (Fx ). Thus, three magnetic states were tested: Gz , Gz Cy and
Gz Cy Fx .
From Table C.� we can note that independently of compression or expansion
strain, the ground state in the BaCoF4 is pure AFM with a weak-AFM as slave
(Gz Cy ). However, for small strain values (close to �% as observed in the experi-
mentally synthesized thin films), the ground state corresponds to an AFM (Gz )
configuration coupled to the weak-AFM (Cy ) and weak-FM (Fx ) that is in agree-
��� b a c o f 4 thin films

ment with the experimental findings (Gz Cy Fx ). It is important to remark that the
weak-AFM ordering exhibit two different magnetic moment components which
are in agreement with the different octahedral tilting of the CoF6 in the xz-layer
such as can be observed in the Fig. C.� and in the values presented in the Table
C.�.

Table C.�: Magnetic moment per atom (or Co atom) along each axis (x, y and z) for
each magnetic ordering where the Gz is the main collinear AFM order and
Cy and Fx are the weak-AFM and weak-FM spin orders, respectively. All
strains are isotropic in the ac-plane, except for the �.�% strain result which
was calculated with strain applied only along the z-axis, per experimental
results.
Strain mi [ µB /atom]
ac-plain Gz Cy Fx
-�% �.��� �.���,�.��� �.���
�% �.��� �.���,�.��� �.���
+�.�% (exp.) �.��� �.���,�.��� �.���
+�% �.��� �.���,�.��� �.���

Surprisingly, the mx magnetic moment induced by the non-collinear magnetism

remains almost constant with small deviations within the error regime, thus
no observable trend as a function of the epitaxial strain was found even to a
largest stress values (around �%). This can be explained in terms of the competi-
tion between the Dyzaloshinskii-Moriya interaction and the single-ion anisotropy
(SIA), such as it has been found in the oxide compound CaRhO3 [�]. Besides,
the highly distorted CoF6 octahedral can induce structural contributions to the
non-collinear ordering.
If we analyze the complete collinear arrangement at the �.�% strain (which is
the equivalent to the one found in the thin films), we observed a total energy of
-�.���������� eV/atom for the B-phase (Gz ) ground state. Thus, based in the
energy difference landscape between the magnetic phases, taken the total energy
of the Gz Cy and Gz Cy Fx orderings from the table �, we can expect a possible
magnetic transitions sequence such as Gz Cy Fx ) ( E = �.�� meV/atom) Gz Cy
) ( E = ��.�� meV/atom) Gz respectively as the temperature goes from � K to
TN . These results can explain the magnetic transitions experimentally found in
the M vs. T measurements.
It is important to note that according to the experimental reports, a possible
magnetic transition can be achieved by pressure applied along the y-axis making
the octahedral layers closer and changing the weak interlayer interaction [���].
 C.� conclusions ���

c.� Conclusions
We have studied by means of first-principles calculations, the behavior of the
magnetic and ferroelectric properties of the BaCoF4 as a function of the ac-plane
epitaxial strain. We found a phase transition from the Cmc�1 to Pca�1 phase
as a function of the epitaxial strain was found. A non-linear polarization-strain
coupling was observed suggesting a second order dielectric behavior allowed in
the BaCoF4 fluoride. An enhancement of the polarization is observed for values
of strain > �%. For the value of epitaxial strain correlated to the experimental
thin films, a non-collinear ground state was found. Besides, a possible path of
magnetic transitions such as Gz Cy Fx ) Gz Cy ) Gz is observed according to
the total energy differences. Non trend in the weak-FM as a function of the
strain was observed even when the Co–F–Co bonding exhibit a change from �
to �� going from -�% to +�% strain respectively. The full understanding of the
non-collinear magnetism and its origin is still unclear and needs further studies.
SPIN-TEXTURE IN SrBO3 (B) = Ti
D
A N D H f ( � � � ) S U R FA C E S

Oxide perovskites exhibit a very large variety of physical properties, which allow
their use in a wide range of applications such as spintronics or thermoelectricity
[���, ���, ���]. In particular, during the last decade it has been observed new and
exotic functional properties at the surface of oxide crystals with no parallel to
their bulk counterpart [���, ���]. In this field of multifunctional oxide materials,
SrTiO3 (STO) has attracted a lot of attention due to its wide range of physical
and chemical properties that goes from superconductivity [���] to engineered fer-
roelectricity [��]. In various studies a two-dimensional electron gas (�DEG) has
been reported at the (���) surface of STO crystal, both theoretically[���] and ex-
perimentally by using angle-resolved photoemission spectroscopy (ARPES)[���,
���], which appeared to be induced by the presence of oxygen vacancies and ex-
plained by an electronic charge redistribution at the surface. The dynamics of the
Ti–d electron quasiparticle on this surface have been experimentally character-
ized and the appearance of cooperative effects of the electrons has been reported,
giving rise to an orbital ordering itself leading to a Rashba-like splitting at the
dxy - dxz/yz bands crossing close to the Fermi level[���]. Recently, a mixed char-
acter of the confined electrons of the (���) surface of STO displaying a complex
spin texture at the Fermi surface has been reported. This unexpected result was
ascribed to a Zeeman interaction plus Rashba-like splitting of the lowest dxy owed
as well to oxygen vacancies[���].
As commented previously, similar Rashba-effect[���] in oxides has been ob-
served in the LaAlO3 /SrTiO3 interface [���, ���, ���, ���] or in SrTiO3 - and
KTaO3 -based transistors [���, ���]. For these single perovskite-based devices, a
k-cubic dependence of the splitting was found in contrast to the linear+cubic
dependence of the Rashba splitting at the LaAlO3 /SrTiO3 interface due to the
dxy – dxz/yz multiorbital nature of the lowest bands [���]. As in the case of
Au(���) surface, where it has been shown that even when the surface states ex-
hibit a mainly p–orbital character, the d–orbitals drives the Rashba splitting
and dictates the direction of the spin by means of the orbital-angular momen-
tum (OAM) [���, ���]. Therefore, the d-orbitals of the transition metal atom of
perovskite oxides can be the key parameter for a large and controllable electron
band splitting in future spintronic devices [���, ���]. This later property gives an
additional degree of freedom to the rich field of multifunctional perovskites. How-
ever, the recent findings of a giant spin-splitting at the STO surface, opens some
questions about the nature of the spin-splitting in this type of materials, such as:
What is the origin of this giant-splitting? Could it be manipulated or observed

���
��� spin-texture in s r bo 3 (b) = t i and h f (���) surfaces

in similar compounds? What is the role of the spin-orbit coupling strength in

the Rashba splitting of d-orbitals in oxide perovskites? Since magnetism at the
surface was reported, how does it couple with the splitting?. From the theoretical
point of view, it is well known that density-functional theory formalism provides
invaluable information related to the structural, magnetic and electronic struc-
ture of oxide interfaces, superlattices and surfaces [���, ���, ���, ���]. This has
been particularly true for the SrTiO3 /LaAlO3 interface [���, ���] or the STO
surface [���] where DFT helped in understanding the origin of the �DEG through
the polar catastrophe scenario. In the present appendix we address the previous
questions by performing such DFT calculations on the (���) STO surface with
oxygen vacancies in order to (i) shed light onto this novel experimental finding
and (ii) to scrutinize the microscopic origin of the giant Rashba spin-splitting
and the Ti d-orbitals spin-texture.

12Å Vacuum
O
Ti
OV Sr

Octahedral
AFD-distortions

y
x’
y’

x
z z
y x

Figure D.�: Schematic view of the SrTiO3 : TiO2- (���) surface used in the calcula-
tions with an oxygen vacancy (OV) at the yz- (left) and xz-plane (right)
respectively. The OV is located at the ( 14 , 14 , z) position with respect to the
computational x 0 - and, y 0 - coordinates). The amplification of the AFD dis-
tortions due to the oxygen vacancy can be clearly appreciated. Additionally,
a surface induced off-centering of the Ti along the z direction is observed
with respect to the oxygens (rumpling), inducing an electric dipole. In the
right bottom part we present a top view of the unit cell in the xy-plane.
This figure was elaborated with the support of the VESTA code [���].
 D.� computational details ���

d.� Computational Details

The first-principles calculations have been performed within the DFT as imple-
mented in VASP [��, ��]. The interaction between ion cores and valence electrons
was treated by the projector augmented wave method [��]. The orbitals consid-
ered as valence electrons in the atomic pseudo-potentials were: Sr (�s2 �p6 �s2 ), Ti
(�s2 �p6 �s2 �d2 ), Hf (�s2 �p6 �d2 �s2 ), and O (�s2 �p4 ). To describe the exchange-
correlation energy, we used the General Gradient Approximation (GGA) with
the PBEsol parameterization [��]. In order to account for the magnetic character
of the system we performed spin polarized calculations. The Hamiltonian con-
tained the scalar relativistic corrections and the spin-orbit coupling was taken
into account by the second variation method [��]. A Monkhorst-Pack k-point
grid of (�⇥�⇥�) for reciprocal space integration and ��� eV energy cut-off of the
plane wave expansion have been used to get a residual error on the forces below
� meV/Å and a fully converged electronic structure including SOC. To correctly
describe the (���) surface and avoid interactions between periodic slabs, we used
a supercell with a vacuum space of �� Å. Additionally, in order p toptake into ac-
count the octahedra rotations and tilting, we considered a ( 2 ⇥ 2⇥�) cell as
shown in Fig. D.�, enforcing the cubic symmetry at the cell center. Aiming to
prevent a fake polarization along the z-axis, the central plane is used as mz mirror
symmetry such as the slab is repeated along the z direction.
In what follows, the electronic characterization of the Rashba-like splitting
for SrBO3 with B = Ti and Hf, including surface oxygen vacancies at the BO2
termination according to experiments [���] is discussed. Great emphasis is done
on the size of the splitting and the spin structure.

d.� ST O : T iO2 terminated surface

Fully relaxed STO surface reveals a change in the interatomic distances close
to the surface due to the missing Ti–O surface bonds (rumpling). The Ti–Ti
distance calculated between the Ti surface layer and the next just below is �.���
Å, which is much smaller than the STO bulk distance of �.��� Å. Additionally,
the out-of-plane Ti–O bond length at the surface changes its distance from �.���
to �.��� Å, leading to a local polar like displacement. We also found a substantial
lowering of the calculated electronic gap, which goes from �.� eV at the bulk level
to �.� eV in the slab geometry. We computed the electronic band structure with
and without SOC and we did not observe substantial differences, which implies
that SOC has a small contribution to the changes in the electron band structure
compared with the geometrical changes. Furthermore, no octahedral rotations or
tilting were observed in the fully relaxed surface when the cubic symmetry is
imposed at the central slab plane.
��� spin-texture in s r bo 3 (b) = t i and h f (���) surfaces

SrTiO3 SrHfO3

sx sx

Energy [eV]
Energy [eV]

sy sy

sz sz

kx ky kx ky

Figure D.�: Band structures around -point for SrTiO3 and SrHfO3 with an oxygen
vacancy on the BO2 terminated surface. Red and blue denote positive and
negative projection of spin vector ~s on Cartesian axes, units of the k-vector
in units of ⇡/a. The non spin-polarized bands are depicted in dark grey.
The effect of the SOC and a Rashba-type splitting is observed in the dxy
and dxz/yz bands crossing (noted with the dashed circles) as well as the
vertical energy shift ⌘ EZeeman + El . For SHO a spin-splitting of the
lower bands is observed for the x and y components around E = –�.� eV.
Figures were created by using the PyProcar script [���].

d.� ST O : T iO2- surface with oxygen vacancies

Experimentally, STO surface samples become conducting in the presence of oxy-
gen vacancies [���, ���]. In order to analyze their effect, we performed full struc-
tural relaxation of the STO:TiO2 surface by removing one oxygen in the last
TiO2 termination layer ( = �.�), imposing a bulk cubic symmetry to allow us a
direct comparison with the pristine case. As we illustrate it in Fig. D.�, octahe-
dral tilting and rotations are now observed in the presence of oxygen vacancies
(the rotation angle is ��.� with respect to the z-axis for the tetrahedral closest
to the vacancy), in contrast to the surface without oxygen vacancies, where no
rotation was observed.
In what follows we study the electronic and spin structure of the (���) STO:TiO2-
surface. In Fig. D.�, we report the calculated electronic band structure around
the –point including non-collinear magnetism and SOC. We found a conducting
behavior very similar to what has been observed at the LaAlO3 /SrTiO3 interface
[���], where the lowest energy bands are the Ti–�d orbitals. At the bulk level, the
 D.� ST O : T iO 2- surface with oxygen vacancies ���

Ti–�d orbitals present in the TiO6 octahedra are expected to split into t2g (dxy ,
dxz , dyz ) and eg (dz2 , dx2 -y2 ) due to the crystal-field splitting. Here, we remark
that small distortions in the octahedral geometry, as the one observed at the sur-
face, generate a small shifting in the energy-values of the electron levels ( El ) and
a possible mixing of the t2g and eg states, mainly due to the symmetry breaking
and the uncompensated TiO6 octahedra at the surface [���]. In the oxygen va-
cancy STO surface we found that the lower d–levels bands corresponds to the dxy
orbitals in the paths that runs from -�.�kx – – +�.�kx and -�.�ky – – +�.�ky .
In addition, we also found a magnetic moment on the Ti atoms close to the sur-
face. The amplitude of these magnetic moments is around �.��� µB /Ti-atom in
the surface and �.��� µB /Ti-atom in the sub-surface layer. These results are in
agreement with the experimental observation of magnetism in STO with oxygen
vacancies [���] and other theoretical works of the same system [���, ���, ���].
The spin polarization is analyzed by plotting the sx , sy and sz components of the
magnetization for each path, where up and down spin orientations are depicted
by red and blue colors respectively. We can appreciate that lower light bands close
to the Fermi level are separated vertically due to an induced Zeeman splitting
( EZeeman ) as expected from the magnetism found at the surface and the en-
ergy splitting El . This magnetic moment is induced by the effect of the electron
surface density rearrangement as a response of the uncompensated orbitals, thus
oxygen vacancy acts as a magnetic impurity [���, ���].
From Fig. D.� we can also see the parabolic-like shape of the d-bands, which is
a characteristic of a �DEG, as reported in previous studies [���, ���, ���]. Here,
it is important to note that this �DEG is induced by the electronic reconstruction
related to the confinement of the electrons at the surface [���, ���], rather than
a polar catastrophe origin as mentioned for the SrTiO3 /LaAlO3 interface [���].
Thus, we can expect that the behavior of the conducting electrons in STO differs
from the one found in the oxide/oxide interfaces.
The effect of the SOC can be clearly appreciated in the Rashba-like spin-
splitting at the dxy and dxz/yz crossing (pointed out by the dashed circles in
Fig. D.�), similarly to what has been previously reported in this surface [���].
For the dxy state, we obtain an effective mass of �.� me for the orbital located at
the second layer of TiO2 . This calculated effective mass is two times larger than
the one reported in Ref.[���] (�.�� me ), which is too small for d like states [���].
However, a spin-splitting such as in a classical Rashba picture is not observed (i.e.
crossing of the spin up and down bands) between the light branches or a spin
inversion within the dxy bands for any of the spin-orientations along the kx and
ky paths. This result is in contrast with the interpretation of the experimental
measurements of Santander-Syro et al. [���] who reported a giant spin-splitting
in the lowest d-bands of the (���) STO surface. One possible explanation of this
splitting is that it can be a consequence of the surface ionic vibrations. In that
respect, we have performed ab–initio molecular dynamics calculations at constant
temperature of ��� K for the slab geometry. Selected uncorrelated frames were
taken from the simulation and electronic band structures were calculated for each
��� spin-texture in s r bo 3 (b) = t i and h f (���) surfaces

case. As for the � K ground state, we did not observe any Rashba splitting like
the one reported in Ref [���], which indicates that at least for low temperatures
there is no coupling between the ionic vibrations and the electronic response.
Another possibility could be that it comes from anharmonic vibrations at much
higher temperatures, a possibility we do not explore in this work. The absence
of a large spin-splitting can be understood from the small SOC present in the
�d–Ti orbitals. To test this hypothesis, we performed a computer experiment
by replacing all the Ti atoms by a heavier cation with much larger SOC such
as Hafnium. Taking the advantage that the Hf atom has the same electronic
configuration as the Ti but with �d orbitals � . In these calculations cell lattice
parameters and internal coordinates of the optimized geometry of the STO surface
were fixed in order to keep the same structure. The �DEG generated at the
HfO2 surface with oxygen vacancies presents the same main features as in the
STO case. However, a large change in the spin-splitting state is observed in the
electron dispersion as shown in Fig. D.�, and somehow closer to the experimental
suggestion of A. F. Santander-Syro et al. [���], who claimed for a spin-inversion
direction within the lowest d-band. In both, kx and ky paths in Fig. D.� and at
the Fermi level the bands are mainly composed by sz spin direction. The spin
switching is observed for the light dxy Hf band at E = –�.� eV and reciprocal
points close to –�.� and �.� for sx and sy spin components respectively. This
suggests that a strong surface polarization is not enough to induce a Rashba-like
spin-splitting and the presence of high-Z cations in the structure is also necessary
to achieve this spin state as previously reported for Ag, Cu, and Au surfaces [���].
The Rashba-like splitting in the SrHfO3 (SHO) is only observed for the sx and
sy spin components of the xy-plane perpendicular to z-axis as expected from the
Rashba Hamiltonian HR = ↵R (~ez ⇥ ~k) · ~s [���].
In addition to the band analysis, we have also extracted the in-plane spin-
texture as shown in the Fig. D.�. In Fig. D.� we can see two concentric rings for
up and down spin polarization respectively for the STO (at E = � eV) and one
single ring for SHO (E = –�.� eV) surface. The rings in the STO appears as a
result of the vertical splitting due to EZeeman + El , where El is induced by
the lowering of the dxy energy bands in comparison to dyz and dzx orbitals as
discussed before. In the case of SHO (Fig. D.�), spin precession in the lower ring
is observed in addition to a spin switching of the dxy band in agreement with the
bands in Fig. D.�) as expected for the Rashba-type splitting.
The absence of the spin-splitting in the STO surface could be also linked to
magnetism, as observed at the the Gd(����) surface [���] where the presence of
magnetism and a Rashba-type splitting has been reported. For this system the
splitting is lower than �.� Å-1 even with the large SOC present in Gd (around
� times larger than in Ti)[���]. Moreover, due to the presence of EZeeman ,
there is no crossing of the bands and the direction of the magnetization vector
(M
~ ) determine the presence of the spin-splitting. This is due to the fact that the

� The spin-orbit coupling in Hf is around �� times larger than in Ti based in the (Ze2 /hc)2
expression
 D.� ST O : T iO 2- surface with oxygen vacancies ���

sx sy
SrTiO3 sz
ky ky ky
kx kx kx

SrHfO3
sx sy sz
ky ky ky
kx kx kx

Figure D.�: Band structure at the kx -ky plane (in ⇡/a) where the spin-texture is cal-
culated for each spin component along the cartesian axis (i.e. sx , sy and
sz ). For SrTiO3 compound (above, E = � eV) simple concentric rings for
up (red) and down (blue) spin polarizations are observed. In contrast a
Rashba-like splitting is observed by replacing the Ti cation by the heavier
Hf in the SrHfO3 (bottom, E = – �.� eV). Figures were created by using
the script PyProcar [���].

Rashba-splitting vanishes for ~k k M~ and only EZeeman remains [���]. The later
statement needs to be carefully taken into account in the STO case for which we
find a magnetic moment components at the surface mx =�.���, my =�.��� and
mz =�.��� µB /Ti-atom respectively and thus after considering several possible
magnetic orientations. Different starting magnetic configurations, along x- y- and
z-axis respectively were taken into account and full structural and energy mini-
mizations were applied. Nonetheless, the system always relaxed toward the same
magnetic ground state in which a diagonal direction of the magnetic moment is
observed with respect to the surface. No traces of magnetization were observed
at the oxygen sites confirming the Ti–localized magnetic moments. Thus, this
rather large in- and out-of-plane magnetic moments have the tendency to pre-
vent the possibility to have a giant Rashba splitting, in addition to the small
SOC discussed previously.
In the case of SHO, we obtain that the Hf magnetic moments are strongly re-
duced to mx =�.���, my =�.��� and mz =�.��� µB /Hf–atom respectively close
to the surface, which strongly reduced the EZeeman . Fig. D.� shows that the
lowest d-band in SHO only exhibit spin components along the xy-plane. Nonethe-
less for higher energy values, the bands are mainly spin-polarized along the z-axis,
demonstrating a very complex spin structure with a unique spin-texture. There-
��� spin-texture in s r bo 3 (b) = t i and h f (���) surfaces

fore, only a deep understanding of this subtle competition between the polar
distortions, the symmetry conditions, SOC, and magnetism would allow us to
tune the Rashba-splitting at the (���) BO2 surface.
In order to test the validity of our results against the exchange-correlation
functional, we also carried out the same calculations with both the Local Den-
sity Approximation [���] (LDA) and the GGA PBEsol+U functionals. Using the
LDA approximation, we did not find substantial differences with the GGA PBEsol
calculations. Due to the strong correlation of the d-orbitals that drives the mag-
netism, we might expect the DFT+U formalism [���] to enhance the electron cor-
relation and change the physical picture discussed previously. Indeed, above U =
� eV, we observe an insulating state of the STO surface with the oxygen vacancy,
which do not agree with the experimental measurements [���, ���, ���]. Then,
for whatever U value below � eV, we do not observe a large Rashba+Zeeman
(RZ) type spin-splitting. In order to test the oxygen vacancy position effects, we
performed a second set of calculations considering the oxygen vacancy in the sec-
ond inner layer from the surface instead of the first layer and, as before, we did
not observe a Rashba spin-splitting in the STO surface within this configuration.

d.� Pbnm SrHfO3 : HfO2 surface with oxygen va-

cancies
We completed the analysis of the Rashba splitting mentioned above by analysing
the spin textures in the SrHfO3 ground state, i.e. in the Pbnm phase with a =
�.���� Å, b = �.���� Å, and c = �.���� Å [���]. To analyze the new SHO surface
in its Pnma phase, we fully relaxed the surface including an oxygen vacancy in
the HfO2- termination slab. Our results reveal an increase of octahedral tilting
and rotations as a result of the free surface plus the oxygen vacancy, which induce
a structural reconstruction with respect to the bulk (see Fig D.�).
From the electronic structure analysis we observe a strong reduction of the
band gap, from �.�� eV in the bulk to �.�� eV in the surface with oxygen vacancy
and thus an absence of the �DEG. Additionally, from the band structure plotted
for each spin orientation (see Fig. D.�, bands around the –point along kx and
ky paths) we can appreciate that the lowest Hf bands corresponds to �d-levels,
and more precisely to the �dxy states. According to Fig. D.�, we note that the
major spin components for the �d bands are along the z-axis and a Rashba-
type splitting is also observed for this orientation in most of the branches, which
happen to be out-of-plane and perpendicular to the surface according to our
results. This behavior can be explained in terms of the polarization close to the
surface and the Rashba Hamiltonian expressed before. Here, the bands present a
spin-splitting for a dispersion along a ~k-vector perpendicular to the electric field
~exy . To confirm this hypothesis, we computed the spontaneous polarization by
means of the Berry phase approach [���]. We obtained that the Pz component is
zero by construction and the in-plane components are equal to �.�� µC·cm-2 and
 D.� pbnm SrHfO 3 : HfO 2 surface with oxygen vacancies ���

sx OV
Energy [eV]

sy

sz

ky
kx ky

Figure D.�: Band structures around the -point along kx and ky for SrHfO3 surface
with an oxygen vacancy on the top of the HfO2 layer for each direction of
the spin. Red and blue colors represent the up and down spin directions
respectively. A larger Rashba-type splitting of the bands is observed for
the z component as appreciated for the ky – sz extracted image. The SHO
slab structure is also shown in which the large octahedral tilting can be
observed. Figures were created by using the script PyProcar [���].

�.�� µC·cm-2 found along the x-axis (Px ) and y-axis (Py ) respectively. This in-
plane polarization in orthorhombic perovskites is quite well known and is related
to the presence of anti-polar motion of the A-site in the xy-plane, which, in the
case of the surface symmetry breaking, do not cancel out from layer to layer
and thus resulting in an effective spontaneous polarization along the xy-plane
[���, ���]. Interestingly, and in contrast to the STO case, we do not find any
presence of magnetism in the Pbnm SHO surface with oxygen vacancies and at
the same time the Rashba-splitting is considerably enhanced, confiming the rule
discussed above. Interestingly, when replacing the Hf by Ti in the relaxed Pbnm
SrHfO3 surface, the magnetism appears again and the Rashba splitting dissapear.
��� spin-texture in s r bo 3 (b) = t i and h f (���) surfaces

All of that confirms the influence of the magnetic ordering and the SOC strength
of the B-site at the spin-texture of SrBO3 surfaces with oxygen vacancies.

d.� Summary and Discussions

We have analyzed from first principles calculations the (���) SrBO3 , (B = Ti and
Hf) surface with oxygen vacancies. We found an insulator to metal transition
behavior when an oxygen vacancy is included in STO at the TiO2 surface, which
produces an electronic charge rearrangement that results into a �DEG located
at the surface and magnetization of the Ti–d orbitals close to the vacancy. In
the limits of our DFT calculations, it is unlikely that a large Rashba-type split-
ting in the light lowest dxy bands exists in the presence of an oxygen vacancy at
the STO:TiO2 surfaces. This is in contrast with the giant spin-splitting of about
�.� Å-1 reported by Santander-Syro et al. [���] in their experiments. This dis-
crepancy can come from different origins. On the experimental side, due to the
methodology of the spin-ARPES technique used by Santander-Syro et al., excited
states could have been induced at the surface, with a different spin-polarization
of the photoelectrons. Then, these states can become accessible and be measured
by this technique as already reported before [���, ���, ���]. There is also a pos-
sibility of high temperature gradients at the surface due to the high energy from
the laser field. This could create large ionic vibrations, which can create large
ionic vibrations, even if the adiabatic approximation still holds [���]. On the cal-
culation side, it is possible that the configuration chosen for the oxygen vacancy
does not correspond to what is present experimentally. In that respect imaging
techniques such as STM would be of use in order to check the surface structure.
It has been shown from photoemission spectroscopy that the Rashba spin-
splitting in these �DEG systems is extremely sensitive to the charge and orbital
ordering, the SOC or the lattice degrees of freedom [���]. We observed such a
sensitivity when replacing Ti by Hf atom with a much larger SOC, which induces
a larger change in the spin-splitting landscape at fixed lattice degrees of free-
dom. Indeed, when relaxing toward the SHO ground state, we observe a totally
different magnetic and Rashba spin-splitting states, all of that confirming the
strong spin structure sensitivity of these surfaces. However, in all the configura-
tions we considered, we never obtained a Rashba spin-splitting as large as the one
reported by Santander-Syro et al. and the fact that it is much larger than the one
observed in the LaAlO3 /SrTiO3 interfaces (<�.�� Å-1 [���, ���] and supported
by theoretical analysis [���, ���]) call for additional studies (theoretical and ex-
perimental) in order to understand this unexpected large splitting. Going beyond
this discrepancy, we proved that sizeable Rashba-type splitting can be achieved in
multifunctional perovskite oxides with large SOC, such as SrHfO3 not observed
before. These systems thus opens the possibility to tune exotic properties for new
technological applications such as spin-orbitronics [���, ���].
R O L E O F S U R FA C E WAT E R
E
ADSORPTIONS IN SrTiO3/LaAlO3
I N T E R FA C E

The surface functionalization properties can significantly affect the performance

of semiconducting electronics [���, ���]. On one hand, surface states can induce
undesired carrier trapping [���]. On the other hand, surface assisted doping is also
actively investigated for many material systems where bulk doping is restrained.
Such systems include wide band gap semiconductors [���], two dimensional ma-
terials [���], nanostructures [���], and chemically fragile organic semiconductors
[���]. LaAlO3 /SrTiO3 heterostructure is another system where the surface re-
lated effects are vital [���]. The formation of interfacial two dimensional electron
gas (�DEG) as well as the �DEG’s unique characters in metal-insulator transi-
tion [���], superconductivity [���], magnetism [���] and optical responses [���]
have evoked widespread interest. A mechanism involving the charge transfer be-
tween the surface and the interface (polar catastrophe) were proposed to explain
the formation of �DEG [���, ���]. Possible carrier depletion by oxygen dangling
bonds at the surface was suggested [���]. Different capping layers were explored
to allow further tuning of the interface properties [���]. Effects of surface adsor-
bates formed in air can also be significant. Among the various airborne molecules,
the dissociation and migration of water molecules on oxide surfaces bear partic-
ular technological importance [���]. In LaAlO3 /SrTiO3 heterostructures, effects
of liquid water on the interface electrical properties have been experimentally
observed [���, ���]. In addition, reversible interface metal-insulator transitions
controlled by conducting atomic force microscope (c-AFM) were previously re-
ported [���, ���, ���]. Experiments performed in controlled environments later
links the effects directly to the surface water [���]. Despite of these interesting
experimental findings, clear understanding of the underlying mechanism is still
lacking. Several critical questions remains open: How are water molecules ad-
sorbed on LaAlO3 surfaces and is the adsorption energetically stable? How do
the surface dangling bonds and water adsorptions in air affect the metal-insulator
transitions observed in heterostructures with different LaAlO3 layer thicknesses?
What are the dynamic processes of water adsorption and desorption? How do
the adsorbed water respond to the external field or environment changes? In
this work, we seek to shed light onto these open questions by systematic the-
oretical and experimental investigations. A partially dissociated water layer is
found to strongly bond to the polar LaAlO3 surface. The densities and distri-
butions of adsorbed protons and hydroxides strongly affect the structural and

���
��� role of surface water adsorptions in s r t i o 3 /l a a l o 3 interface

electrical properties of the LaAlO3 /SrTiO3 heterostructures. Intrinsic interface

carrier doping depending on the LaAlO3 thickness and extrinsic contributions to
the interface metallicity from dissociated water are both observed. Considering
the weakly adsorbed molecular water layers also turned out fruitful in that the
associated hydrogen bond network enables active surface proton transfer at room
temperature. The tunable thermally activated proton hopping processes provide
an effective pathway for engineering the interface properties from the surface. The
multifaceted role of water revealed by this work is undoubtedly critical in under-
standing the unique characters at LaAlO3 /SrTiO3 interfaces observed so far. Our
results are also expected to provide valuable guidelines for the future development
of oxides based surface engineering methodologies and device applications. For
more information please check the Refs. [�, �].

e.� Computational Details

DFT calculations were performed using the VASP. PAW [��, ��] pseudo-potentials
were used to represent the valence and core electrons. The electronic config-
urations taken into account in pseudo-potential as valence electrons were Sr
(�s2 �p6 �s2 ), Ti (�s2 �p6 �s2 �d2 ), La (�s2 �p6 �d1 �s2 ), Al (�s2 �p1 ) and O (�s2 �p4 ).The
exchange correlation was represented within the GGA and PBEsol parametriza-
tion [��]. The periodic solution of these crystalline structures was represented
by Bloch states with a Monkhorst-Pack with a (�⇥�⇥�) k-point mesh and an
energy cut-off of ��� eV tested to give converged forces to less than �.��� eV/Å.
To correctly describe the system, �u.c of SrTiO3 (STO) as a substrate was taken
into account. Then, �u.c. and �u.c of LaAlO3 (LAO) were used as thin films for
the analysis. Additionally, in order to avoid periodic interactions between peri-
odic slabs, we used a supercell with a vacuum space of �� Šalong the z-axis. The
dipole correction was applied along the out-of-plane z-axis to cancel dipole-dipole
interactions and correct the errors caused by periodic boundary conditions [���].
In addition and in p order
p to take into account the octahedral rotations and tilting,
we considered a 2⇥ 2 cell in the xy-plane with lattice parameter fixed to the
PBEsol relaxed STO a = �.���� Å.

e.� Water dissociation in the surface

Different theoretical calculations have demonstrated that the electronic recon-
struction induced by the polar catastrophe phenomena is the main responsible
for the insulator-to-metal transition (IMT) in LAO/STO and the appearance of
a �DEG at the polar-nonpolar perovskite oxide interfaces (see for example Ref
[���]). Therefore, modifications of the surface formal charges at the AlO2 top
layer is expected to change the interfacial properties. Then, a clear understand-
ing of this phenomenon is crucial in order to explain the macroscopic electronic
transport properties discussed in this work.
 E.� water dissociation in the surface ���

Figure E.�:
Figure 5. Simulated lattice distortions
Relaxed geometries in 3uc LaAlO
for the pristine, 3/SrTiO
and the heterostructureand
OH, 3 �.�H+�.�OH, with
H
different surface H and OH densities. Tiltation of oxygen octahedra becomes more
groups. It can be observed the octahedral tilting and rotation increase when
significant asthe
themolecules
surface proton density
are placed atincreases.
the surface.

We have noticed that the octahedral rotations and tilts across the LAO layer
—dissipating up to the STO substrate— increases when H and OH are placed
at the LAO surface (see Fig E.�). These AFD distortions have a direct effect on
the interfacial properties and specifically, at the carriers density at the interface
as we have observed in a computational experiment where, the carried density
was computed with and without full atomic relaxations. Here, it is important to
remark that those distortions are present even in the pristine bulk structure due
to the rhombohedral R�c ground state of the LaAlO3 [���].
In the �uc pristine case, the electric field of �� mV/Š(computed from the layer
resolved DOS) is clearly not enough to overcome the existing electron gap in the
STO/LAO system, leading to the obtained insulated behavior (Eg = �.� eV).
The water dissociation and the later removal of the (OH)- groups with the AFM
tip leaves only absorbed hydrogen atoms on the surface which, according to our
results, act as electron-donors. These additional electrons, under the effect of the
electric field existing in the LAO film, are transferred to the interface, which then
generate a �DEG with dxy character from the Ti orbitals. This electron transfer
explains the induced IMT in the �uc LAO, as depicted in Fig. E.�. Consequently,
the absorbed hydrogen act as electron-donor and provide additional charge, which
is enough to compensate the surface charge responsible for the polar discontinuity.
The later clearly alters the out-of-plane electrostatic potential. For example, when
��� role of surface water adsorptions in s r t i o 3 /l a a l o 3 interface

the surface is fully hydrogen saturated, the field takes a value of -�� mV/Šthat
is in opposite direction to the potential in the (AlO2 )- pristine surface. This field
crosses through � mV/Å for �.�H (see Fig. E.�), which demonstrates the direct
effect of the electron-donors in the electrostatic response, as was expected. On the
other hand, if we consider �uc of LaAlO3 on SrTiO3 (���), the electric field is now
close to ��� mV/Šacross the entire LAO film as shown in Fig. E.�. This value is
sufficient to overcome the present electronic band gap of the STO material and
the system becomes metallic, as it is widely known. Additionally, a �DEG and a
�DHG are created at the interface and surface respectively. Now, with the same
process of water absorption and later (OH)- removal, as in the previous �uc case,
the hydrogen electron-donor alters the amount of electrons in the system. These
additional electrons are transferred from the surface to the interface, similarly to
the case of �uc of LaAlO3 , and the amount of carriers density increase drastically
as observed in Fig. E.�, as it is also supported from experimental results.

Figure E.�: Layer-by-layer density of states for the �uc in which the electric field profile
can be observed. A insulator-to-metal transition can be observed when going
from the pristine to the H-doped structures.

Figure E.�: Layer-by-layer density of states for �uc for the pristine structure and the
water dissociated at the surface.
 E.� water dissociation in the surface ���

Our theoretical analyses also reveal the appearance of magnetic moments at the
interface for the �uc case, as expected from the unpaired �d-Ti orbitals occupation
as it has been already reported [���]. In the �uc case, magnetic moments are only
observed at the �d-Ti when the interface becomes metallic due to the bonded
hydrogen. When only (OH)-groups are in the LAO surface, the magnetism is
only present at the top layer. This is due to the electron holes at the localized
oxygen p-orbitals as it has been shown in some d0 semiconductor [���].
The fact we are considering oxides in which, strong electron correlations takes
place, makes important to investigate the effect in the results of the exchange
correlation functional by enhancing the electron-electron repulsion in the so called
DFT+U [��]. This method allows to increase the electron correlation lacking in
the d-orbitals and then improving the electron band gap, in closer agreement with
experimental predictions. This correction can also affect negatively the searched
properties because it changes at the same time some other orbitals. Thus, we have
considered initially the �uc case as a prototypical example. All results here were
re-computed with a U = �.� and �.� eV for Ti and La respectively and J = � eV
[���]. We found that the general description holds but the valence carrier density
increases by more than one order of magnitude, which is not in agreement with
the reported experiments. Therefore, our results reported here only correspond
to the case of U = J = � eV.

Figure E.�: Computed carrier density for �uc and �uc respectively. Here we considered
several molecules as a dopant agent at the surface.

Now, if we think in the idea to reverse the voltage in the tip, and then, the
H+ is removed from the surface, the effect can be seen in the reduction of the
electric field in the LAO film. This field goes from ��� mV/Što �� mV/Šfor
the pristine structure to the OH saturated surface. Nonetheless, according to our
results, this field reduction is not having a major effect in the carrier density as
can be appreciated in Fig. E.�. This can be explained by the fact that in the
�uc, the charge is already confined at the interface and there is no charge transfer
going back to the surface, which is traduced in an increase of the holes at the
��� role of surface water adsorptions in s r t i o 3 /l a a l o 3 interface

surface. In the �uc case, the electric field is also reduced to �� mV/Šand the
system remains insulator.
We considered the possibility of reversing the voltage on the AFM tip, H+ is
removed and OH- are kept on the surface, such that there is a reduction of the
electric field in the LAO film. This field goes from ��� mV/Što �� mV/Šfor the
pristine structure to the OH saturated surface respectively. However, according
to our results, this field reduction is not having a major effect in the carrier
density as can be appreciated in Fig. E.�. This can be explained by the fact that
in the �uc, the charge is already confined at the interface and there is no charge
transfer going back to the surface, which is already saturated with OH. The later
is traduced in an increase of the holes at the surface. In the �uc case, where the
heterostructure is already an insulator, the electric field is also reduced to ��
mV/Å and the system kept a finite electronic band gap.
LIST OF FIGURES

Figure �.� Schematic representation of the main PAW method that

exemplifies the Eq �.��. . . . . . . . . . . . . . . . . . . . ��
Figure �.� (Above) Phonon-dispersion curves for NaMnF3 computed
at the Pm3̄m phase. The Brillouin zone is presented in the
inset. The imaginary modes are plotted as negative values
. Atomic displacement for each unstable mode. (Below)
Antipolar displacements, octahedral rotations and tilts
(depicted by arrows) can be observed for the X+ 5 , M3 ,
+

and R+ 4 . modes. . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Polar 4- mode, into the Bethe’s notation [��], as a func-
tion of the A- and B-site in ABF3 perovskites, the imagi-
nary modes are presented as negative values. . . . . . . . ��
Figure �.� Ferroelectric instability in LiBF3 and NaBF3 (B = Ca,
Mn, V, Zn, and Ni) as a function of the tolerance factor, t. ��
Figure �.� Selected IFC interactions in the �⇥�⇥� supercell. A- B-
and F-sites are depicted in green, violet and grey respec-
tively. Additionally, the atoms are labeled to correlate the
IFCs shown in the Table �.�. . . . . . . . . . . . . . . . . ��
Figure �.� Double-well energy obtained at the FE-mode freezing in
NaBF3 compounds. The gain in energy is in agreement
with previous results that shows that the bigger is the
B-site, bigger is the FE instability and with this, the pos-
sibility to be ferroelectric. ± denotes the full amplitude
of the freezing. . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Double-Well energy curves and schematic representation
of atomic displacements in different models of ferroelec-
tricity for oxides a) BaTiO3 and b) NaMnF3 . . . . . . . . ��
Figure �.� Unstable FE mode at the cubic Pm3̄m structure in ABX3
compounds. . . . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Eigendisplacements for NaCaF3 compound showing the
⌘ transition also found in B-site: V, Mn, and Zn (left).
Schematic representation of eigendisplacement in the flu-
orides compounds for strain values above and below of the
relaxed structure (right). . . . . . . . . . . . . . . . . . . . ��
Figure �.� Eigendisplacements for BaTiO3 showing the change of the
mode as cause of suppression of ferroelectric instability as
a function of isotropic pressure. . . . . . . . . . . . . . . . ��

���
��� List of Figures

Figure �.�� On-site IFC for A-, B- and F-sites (x=y, and z ) as a func-
tion of strain for each of sites in NaBF3 with B = Ca, Mn,
V, Zn, and Ni. . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.�� Contribution of modes in the phase transition from Pm3̄m
to Pnma in the NaBF3 fluoroperovskites. . . . . . . . . . ��
Figure �.�� Behavior of FE and R-point instabilities as a function of
hydrostatic pressure for NaNiF3 . . . . . . . . . . . . . . . ��
Figure �.�� Eigendisplacements for all polar modes in NaNiF3 . . . . . ��
Figure �.�� Squared frequency at the vibrational instabilities. We can
observe the FE-mode at and AFD-modes at R-point as
a function of hydrostatic strain for LiNiF3 (left). Eigendis-
placements of FE-mode instability in LiNiF3 are also shown
(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� (a) Total energy of NaMnF3 as a function of the volume
per atom for all different crystal structures. Cmcm and
Pnma structures show the lowest energy values for lower
volumes. (b) H/atom as a function of external pressure in
NaMnF3 structures. Pnma to Cmcm structural transition
can be seen at the critical pressure of � GPa. . . . . . . . ��
Figure �.� (a) Schematic view of the phase transition from Pnma to
Cmcm. Na, Mn and F atoms are represented in green, vio-
let and grey colors respectively. The breaking of bondings
for the transition from corner-shared to edge-shared oc-
tahedra can be appreciated. (b) Simulated XRD patterns
for Pnma and Cmcm structures. The room temperature
Pnma phase is in good agreement with the experimental
reports (open triangles) [���, ���, ���]. (c) Octahedra tilt
angle as a function of pressure in NaBF3 . For B = Zn,
Ni, Co and Mg experimental results are taken from Yusa
et al. [���]. All of these compounds present the transi-
tion from the Pnma to the Cmcm structures at a tilting
angle close to �� . Experimental results for NaMnF3 at
low pressure were taken from Katrusiak et al. [��] Full
agreement between experimental (filled squares) and our
theoretical calculations (empty squares) can be appreciated. ��
Figure �.� Vibrational behavior as a function of isotropic pressure for
Pnma and Cmcm pPv. The open and filled symbols are
the Raman and the Infra-red modes respectively. . . . . ��
 List of Figures ���

Figure �.� (a) Energy differences between the different magnetic or-
derings of the pPv phase of NaMnF3 versus the U pa-
rameter. (b) Sketch of the Cy Fz and the Cz Ay magnetic
orderings in the pPv phase of NaMnF3 . For U < �.� eV
the system exhibits a AFM+FM magnetic state with a
high ferromagnetic canting. For U > �.� eV, the system
transit toward a canted AFM state with no weak FM. . . ��
Figure �.� Magnetic canting dependence as a function of U and J pa-
rameters for NaMnF3 . Canting angle (') varies from �.��
to ��.�� when U goes from �.� to �.� eV. The canting
angle goes from ��.�� to �.�� when J varies from �.� to
�.� eV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Projected local density of states (LDOS) of the Cmcm
phase of NaMnF3 . A regular F-p bonding Mn-d anti-
bonding is observed with small Mn:3d – F:2p overlaping
DOS is responsible for the small superexchange interac-
tion present in NaMnF3 . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Simulated XRD patterns with K↵:Cu for CaRhO3 (above)
and NaNiF3 (below) in the Cmcm phase at atmospheric
pressure are observed. . . . . . . . . . . . . . . . . . . . . ��
Figure �.� The non-collinear magnetic orderings in the pPv Cmcm
symmetry labeled as Cz Fy , Cz Ay , Cy Fz and Gy Az are
presented in the figure. Two of them with pure AFM
(Cz Ay and Gy Az ) behavior and two AFM+weak-FM (Cz Fy
and Cy Fz ) can be observed (Fig. elaborated with the VESTA
software [���]). . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Schematic representation of the atomic sites and magnetic
interactions in ABX3 (A, B and X -sites in blue, yellow
and red respectively). Exchange constants labeled as Ja ,
Jc and Jab and DM interactions represented as Da and Dc
equivalent to D12 and D13 respectively are shown. (Fig.
elaborated with the VESTA software [���]). . . . . . . . . ��
Figure �.� Ferroelectric instability T1u at the cubic Pm3̄m as well as
the frequency values for the softest B2u mode in the Pnma
NaBF3 (B = Mn, V, Zn, and Ni) compounds. . . . . . . . ��
Figure �.� a) Epitaxially strain induced B2u instability in Pnma NaMnF3
in which a nonlinear coupling is observed, imaginary fre-
quencies are represented as negative values by notation. b)
Eigendisplacements of the polar B2u along b-axis (⌘b ) for
each site (Na in green squares, Mn in violet circles and F? ,
Fk in upper and down triangles respectively) in NaMnF3
in Pna�1 . c) Ferroelectric mode freezing in which the tran-
sition from the paraelectric Pnma to a polar Pna�1 phase
is observed at equivalent �% in the ac-plane. . . . . . . . ��
��� List of Figures

Figure �.� a) Polarization along b-axis in NaMnF3 in Pna�1 sym-

metry. The second-order dependence of the polarization
as a function of ac-plane epitaxial strain is observed as a
result of the non-linear piezoelectricity in this perovskite
fluoride. b) NaMnF3 , Pna�1 structure is also presented
where Na, Mn and F ions are depicted in green, violet and
grey respectively. Additionally, the non-collinear magnetic
ground state as well as the polarization vector along the
b-axis are schematically depicted in the figure showing the
multiferroic character of this fluoroperovskite. . . . . . . . ��
Figure �.� Magnetoelectric coupling in NaMnF3 in Pna�1 structure
where Py vs By (upper panel), Px vs Bx (center panel)
and Py vs Bx (lower panel) are presented. In dashed lines
are presented the linear and quadratic fitting according to
the Eq. �.�. The second-order ME coupling can be ap-
preciated from this figure. Additionally, an increase of the
non-linear behavior as a function of the epitaxial strain is
observed. The computed first and second order ME tensor
are presented in Table �.�. . . . . . . . . . . . . . . . . . . ��
Figure �.� Projection of the magnetic B-field induced distortions (at
�� T in the Pna�1 FE phase) against the phonon basis
set of the Pnma phase for -�%, �%, and +�% values of
the epitaxial strain. We note that the B2u FE unstable
mode (the one with negative frequencies) is strongly dom-
inant at compressive strain while the Ag modes are more
dominant at tensile strains. This fulfills the fact that at
compressive strain the FE distortions are much more Mn-
driven (and thus much more sensitive to magnetic field)
than at tensile strains and this also explains why under
compressive strain the ME response is larger than under
tensile strains. . . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Polarization (out-of-plane) vs. epitaxial strain in the ab-
plane for the geometric FE BaCoF4 , for cubic BaTiO3 ,
cubic NaMnF3 and cubic CaTiO3 case in which we simply
relax the polar distortion without the AFD motions, and
for improper-FE PbTiO3 /SrTiO3 and CaTiO3 /SrTiO3
superlattices. We note that the strong non-linear polariza-
tion/strain coupling is present mostly in steric geometric
FE crystals such as NaMnF3 and BaCoF4 . . . . . . . . . ��
Figure �.� Some of the physical phenomena that have lead to several
multifunctional properties found in oxides/oxides inter-
faces and expected in the mixed oxide/fluoride (KTaO3
and KNiF3 in the figure) interfaces. The diagram proper-
ties has been taken from Zubko et al. [���]. . . . . . . . . ��
 List of Figures ���

Figure �.� Localized-DOS for the �d-Ni and �d-Ta at the KNiF2.5 O0.5
and KTaO2.5 F0.5 oxyfluorides respectively. This DOS was
computed by means of the hybrid functional HSE06. . . . ��
Figure �.� Electronic configuration corresponding and crystal-field
splitting of the KTaO3 oxide and the KTaO2.5 F0.5 oxyflu-
oride. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
Figure �.� Electronic and magnetic configuration corresponding to
the crystal-field splitting for the KNiF3 fluoride and the
KNiF2.5 O0.5 oxyfluoride. . . . . . . . . . . . . . . . . . . ��
Figure �.� �/� and �/� superlattices at which the insulator-to-metal
transitions takes place. The TaO2 /KF and ZnF2 /KO in-
terfaces can be observed. . . . . . . . . . . . . . . . . . . . ���
Figure �.� Atomic displacements of the B-sites along the z-axis (left)
with respect to the oxygen apical plane for the �/� super-
lattice. K–K c-lattice parameter in which, a large expan-
sion and compression can be observed at the TaO2 /KF
and ZnF2 /KO interfaces respectively (right). . . . . . . . ���
Figure �.� Localized density of states for M = Zn and Ni in the �/�
superlattice ratio obtained considering the HSE06 hybrid
functional. . . . . . . . . . . . . . . . . . . . . . . . . . . . ���
Figure �.� Layered revolved DOS for the KTaO3 /KZnF3 and KTaO3 /KNiF3
(�/�) superlattices where a �DEG and �DHG are observed
at the TaO2 /KF and MF2 /KO. . . . . . . . . . . . . . . . ���
Figure �.� Band-edge diagram for the electrostatic behavior of the
system where the electric field profile and the �DEG (�DHG)
formation are presented at the oxyfluoride TaO2 /KF (ZnF2 /KO)
interface. This was obtained based on the layered resolved
DOS (see Fig �.�). CB, VB an EF denotes the conduction-
band, valence-band and the Fermi energy respectively. . . ���
Figure �.�� Magnetic moment per layer at the KTaO3 /KNiF3 super-
lattices. A moment modulation can be appreciated when
n/l ratio is varied from �/� to �/�. . . . . . . . . . . . . . ���
Figure �.�� Bands structure for the p- (left) and n-type (right) inter-
faces along the X–M– –X path in the Brillouin zone for
the KTaO3 /KZnF3 system. The formation of the �DHG
and �DEG at the p- and n- interfaces can be appreciated. ���
Figure �.�� Bands structure for the p- (left) and n-type (right) in-
terfaces along the X–M– –X path in the Brillouin zone
for the KTaO3 /KNiF3 system. The �DHG and �DEG at
the p- and n- interfaces can be appreciated. Besides, and
empty Ni:eg band is observed between �.� to �.� eV re-
spectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . ���
��� List of Figures

Figure �.�� Band structure for the kx path in which, sx , sy , and sz

spin components are taken into account. The up and down
orientation of the spin are represented by red and blue
colors respectively. . . . . . . . . . . . . . . . . . . . . . . ���
Figure �.�� LaScO3 /NaMnF3 (�/�) superlattice. The AFD distortions
and rotations can be observed in the upper figure in which
the KO/MnF2 and ScO2 /KF interfaces can identified. In
the lower figure it is shown the spin-polarized charge den-
sity, red and blue colors denote the up and down magnetic
moment directions. The G-type AFM ordering is observed
as well as the CO state. . . . . . . . . . . . . . . . . . . . ���
Figure A.� Octahedral rotation for several oxides and fluorides com-
pounds. The fluoroperovskites are ordered according to
the B-site radii. . . . . . . . . . . . . . . . . . . . . . . . . ���
Figure C.� BaCoF4 [���] supercell in which the red arrows repre-
sent the distortion to go from the Cmc�1 to Pca�1 space
groups. Atoms in green, blue, and grey colors for Ba, Co,
and F species respectively. This figure was elaborated with
the support of the VESTA code [���]. . . . . . . . . . . . ���
Figure C.� (Top) Spontaneous polarization along the b-axis. a, b, and
c lattice parameters (Middle) and behavior of the octahe-
dral angle bonding (Co–F–Co) along the x-axis (bottom)
as a function of the ac epitaxial strain respectively. . . . ���
Figure C.� From left to right, the A, B and C collinear magnetic or-
derings. In green, blue navy and grey are Ba, Co and F
ions respectively. Magnetic moments in each atom are lo-
cated along the y-axis. In the figures can be observed the
unit cell in black lines as well as the computation magnetic
cell that takes into account �� atoms. . . . . . . . . . . . ���
Figure D.� Schematic view of the SrTiO3 : TiO2- (���) surface used
in the calculations with an oxygen vacancy (OV) at the yz-
(left) and xz-plane (right) respectively. The OV is located
at the ( 14 , 14 , z) position with respect to the computa-
tional x 0 - and, y 0 - coordinates). The amplification of the
AFD distortions due to the oxygen vacancy can be clearly
appreciated. Additionally, a surface induced off-centering
of the Ti along the z direction is observed with respect
to the oxygens (rumpling), inducing an electric dipole. In
the right bottom part we present a top view of the unit
cell in the xy-plane. This figure was elaborated with the
support of the VESTA code [���]. . . . . . . . . . . . . . ���
 List of Figures ���

Figure D.� Band structures around -point for SrTiO3 and SrHfO3
with an oxygen vacancy on the BO2 terminated surface.
Red and blue denote positive and negative projection of
spin vector ~s on Cartesian axes, units of the k-vector in
units of ⇡/a. The non spin-polarized bands are depicted
in dark grey. The effect of the SOC and a Rashba-type
splitting is observed in the dxy and dxz/yz bands crossing
(noted with the dashed circles) as well as the vertical en-
ergy shift ⌘ EZeeman + El . For SHO a spin-splitting
of the lower bands is observed for the x and y components
around E = –�.� eV. Figures were created by using the
PyProcar script [���]. . . . . . . . . . . . . . . . . . . . . ���
Figure D.� Band structure at the kx -ky plane (in ⇡/a) where the spin-
texture is calculated for each spin component along the
cartesian axis (i.e. sx , sy and sz ). For SrTiO3 compound
(above, E = � eV) simple concentric rings for up (red) and
down (blue) spin polarizations are observed. In contrast a
Rashba-like splitting is observed by replacing the Ti cation
by the heavier Hf in the SrHfO3 (bottom, E = – �.� eV).
Figures were created by using the script PyProcar [���]. . ���
Figure D.� Band structures around the -point along kx and ky for
SrHfO3 surface with an oxygen vacancy on the top of the
HfO2 layer for each direction of the spin. Red and blue col-
ors represent the up and down spin directions respectively.
A larger Rashba-type splitting of the bands is observed for
the z component as appreciated for the ky – sz extracted
image. The SHO slab structure is also shown in which
the large octahedral tilting can be observed. Figures were
created by using the script PyProcar [���]. . . . . . . . . ���
Figure E.� Relaxed geometries for the pristine, and the OH, �.�H+�.�OH,
and H groups. It can be observed the octahedral tilting
and rotation increase when the molecules are placed at
the surface. . . . . . . . . . . . . . . . . . . . . . . . . . . ���
Figure E.� Layer-by-layer density of states for the �uc in which the
electric field profile can be observed. A insulator-to-metal
transition can be observed when going from the pristine
to the H-doped structures. . . . . . . . . . . . . . . . . . . ���
Figure E.� Layer-by-layer density of states for �uc for the pristine
structure and the water dissociated at the surface. . . . . ���
��� List of Figures

Figure E.� Computed carrier density for �uc and �uc respectively.
Here we considered several molecules as a dopant agent at
the surface. . . . . . . . . . . . . . . . . . . . . . . . . . . ���
L I S T O F TA B L E S

Table �.� Available physical quantities through the first- and second-
order energy derivatives with = ⌧, ⌘, and E, (see the Eq.
�.��). Table adapted from Ref. [��]. . . . . . . . . . . . . ��
Table �.� Electronic configuration of pseudopotentials. Valences and
semicore electrons taken into account for calculations. . . ��
Table �.� Summary of most unstable high symmetry point phonon
eigenvalues, relaxed lattice constants for cubic Pm3̄m sym-
metry, ionic A and B sites radii and reported ground state
symmetry for all compounds [��]. HS: High Spin, LS: Low
Spin. All modes in ( 4- , R+4 , R5 , X5 and M3 ) in cm
+ + + -1 . . ��
Table �.� Eigendisplacements (⌘) for all compounds at relaxed cubic
structure. X? and Xk are the sites located perpendicular
and parallel to the M-F bonding respectively . . . . . . . ��
Table �.� Born Effective Charges (Z⇤ , e- units) in for all compounds
computed at relaxed cubic structure computed according
to the Eq. �.�. The X? and Xk are the components per-
pendicular and parallel to the B–X bonding. . . . . . . . . ��
Table �.� On-site interatomic constants (on-site IFC, in eV/Å2 units)
for the ABX3 flouroperovskites. The on-site IFC for the
BaTiO3 and PbTiO3 are also presented for comparison.
The Fx=y sites lies in A-F plane meanwhile the Fz lies
along B–F bonding. . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Selected IFC according to interactions illustrated in the
Fig.�.� for the A-, B-, and F-sites respectively. . . . . . . ��
Table �.� On-site IFC for all NaMF3 compounds at same lattice
parameter equivalent to expansion. . . . . . . . . . . . . . ��
Table �.� Lattice parameters of the pPv phase (a, b, and c) and
related temperature (T) and pressure (P) transition pa-
rameters from the Pnma phase to the pPv phase of fluo-
ride and oxide perovskites. In the pPv phase, the A and
B cations occupy the 4a and 4c Wyckoff positions and the
O/F anions occupy the 4c and 8d positions. Most of the
oxides have higher pressure and temperature transitions
than the fluorides. . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Calculated Raman, Infra-red (IR) and silent modes fre-
quencies (cm-1 ) of the Pnma phase of NaMnF3 (at zero
pressure). In brackets are presented the experimental val-
ues of the Raman [���] and the IR [���] modes. . . . . . ��

���
��� List of Tables

Table �.� Calculated mode Grüneisen parameters ( i ) for the NaMnF3

Pnma phase at and around atmospheric pressure. . . . ��
Table �.� Computed Raman, Infra-red and silent modes of NaMnF3
in post-perovskite Cmcm phase as well as the mode Grüneisen
parameters ( i ) at around �� GPa. . . . . . . . . . . . . ��
Table �.� Magnetic ordering allowed in the Cmcm structure accord-
ing to the D2h point group [���, ���]. The transformation
of each magnetic ordering under each symmetry operation
is labeled as +� and -�, which indicates when the ordering
is invariant under the related transformation (i.e. +�) or
when it is reversed under the application of the symmetry
operation (i.e. -�). . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Comparison of the distances and the bond angles along
the Mn–F–Mn bonds between the Pnma and the Cmcm
structures of NaMnF3 . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Magnetic orderings allowed in the Cmcm structure for
ABX3 systems with U = �.� eV and J = � eV. The large
magnetic canting for fluorides compounds can be noted.
The canting angle (') is measured with respect to the
ab-plane on the pPv phase. For NaNiF3 , the two possible
magnetic states Cz Fy and Cy Fz were computed. . . . . . ��
Table �.� Magnetic ordering in fluorides and oxides post-perovskite
systems, AFM: antiferromagnetic, DM: diamagnetic, FM:
ferromagnetic, CWP: Curie-Weiss paramagnetic. All pPv
compounds exhibit a low-spin magnetic configuration but
NaMnF3 that according to our calculations is found to be
in the high-spin state. . . . . . . . . . . . . . . . . . . . . ��
Table �.� Vibrational characterization modes of CaRhO3 and NaNiF3
Cmcm phase at atmospheric pressure with the represen-
tation: = �Au �B1u �B2u �B3u �Ag �B1g
B2g �B3g . . . . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Magnetic orderings allowed in the Cmcm pPv structure
according to the D2h point group [�]. Here F is ferromag-
netic ordering and A- C- and G- are the different AFM
magnetic orderings where the subindex indicate the direc-
tion along the particular Cartesian axis. . . . . . . . . . . ��
Table �.� Non-collinear ordering and moments in CaRhO3 and NaNiF3 .
Experimental and our theoretical calculations are in good
agreement. The magnetic moments are in µB /atom for
each magnetic B-site cation. The induced magnetic ground
state by applied magnetic field (Hext ) in NaNiF3 is also
presented. . . . . . . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Computed magnetic exchange constants for pPv CaRhO3
and NaNiF3 . Units in meV. . . . . . . . . . . . . . . . . . ��
 List of Tables ���

Table �.� Dzyaloshinsky-Moriya (Dij ) vectors computed for pPv CaRhO3

and NaNiF3 . Here, D12 ⌘ Da and D13 ⌘ Dc as shown
in Fig. �.�. Units in meV. . . . . . . . . . . . . . . . . . . ��
Table �.� Computed SIA parameters for pPv CaRhO3 and NaNiF3 .
Units in meV. . . . . . . . . . . . . . . . . . . . . . . . . . ��
Table �.� Born effective charges (Z⇤ , in e- ) of selected strain values
("). F? and Fk represent the Born effective charge of the
F atom when it is displaced respectively against the Mn
cation and along the perpendicular direction. Eigendis-
placements (⌘b ) of the FE unstable mode along the b-axis
in NaMnF3 . F? and Fk have the same sense as given in
Born effective charges. . . . . . . . . . . . . . . . . . . . . ��
Table �.� Allowed magnetic orderings in the D2h point symmetry
group [���, ���] for Pnma [���] and epitaxially-induced
phase Pna�1 . Besides, the magnetoelectric couplings [���]
for Pna�1 are shown depending to the non-collinear or-
dering and the magnetic group. . . . . . . . . . . . . . . . ��
Table �.� ME coefficients of strained NaMnF3 Pna�1 expressed as in
the Eq. �.�, ↵ik in [ps·m-1 ] and jik in [⇥��-8 ps·A-1 ]. ��
Table �.� ME coefficients (↵ik and jik in ps·m-1 and ⇥��-8 ps·A-1
respectively) of strained NaMnF3 as a function of U and
J parameters [in eV] for the �% value. . . . . . . . . . . . ��
Table �.� Electronic configuration of pseudopotentials used in the
analysis of this chapter. Valences and semicore electrons
taken into account for calculations. . . . . . . . . . . . . . ��
Table �.� Layer-by-layer spontaneous polarization along the z-axis
(Pz in µC·cm-2 ) in the �/� superlattice is shown. Besides,
the oxide and fluoride polarizations are PKT aO3 = -��.��
µC·cm-2 and PKZnF3 = �.�� µC·cm-2 respectively. . . . ���
Table �.� A-site displacements per layer in the xy-plane taking the
high-symmetry position in the cubic symmetry as refer-
ence. Displacements in Šunits. . . . . . . . . . . . . . . . ���
Table A.� Magnetic canted moment Fy (µB /atom) at low tempera-
ture in Pnma ABF3 systems. The large magnetic canting
for fluorides compounds can be noted in the Gz Ax Fy con-
figuration. The ionic radius of the magnetic B-site (rB )
was taken into account the valence state and the bonding
coordination [��]. . . . . . . . . . . . . . . . . . . . . . . . ���
Table A.� Experimentally reported exchange interaction (J) constants
in ABF3 compounds [���]. The components of the DM
vector are in meV. . . . . . . . . . . . . . . . . . . . . . . ���
Table B.� Crystallographic positions of the magnetic cations, M at
the post-perovskite Cmcm. . . . . . . . . . . . . . . . . . ���
��� List of Tables

Table B.� Transformation of the magnetic sites under the symmetry

operations. The numbers in the table are the i index of
the Mi magnetic cation. . . . . . . . . . . . . . . . . . . . ���
Table B.� Definition of the different magnetic orderings. The G, A-,
and C-type antiferromagnetic orderings and the ferromag-
netic F state are defined for each magnetic cation where
+ and - denoted up and down spin orientations. . . . . . ���
Table B.� Transformation sx , sy , and sz spin orientations under the
symmetry operations. The transformation of each mag-
netic ordering under each symmetry operation is labeled
as +� and -�, which indicates when the ordering is invari-
ant under the related transformation (i.e. +�) or when it is
reversed under the application of the symmetry operation
(i.e. -�). . . . . . . . . . . . . . . . . . . . . . . . . . . . . ���
Table B.� Transformation of the magnetic ordering under the al-
lowed symmetry operations. . . . . . . . . . . . . . . . . . ���
Table B.� Magnetic ordering allowed in the Cmcm structure accord-
ing to the D2h point group [���, ���]. . . . . . . . . . . . ���
Table C.� Magnetic moment per atom (or Co atom) along each axis
(x, y and z) for each magnetic ordering where the Gz is the
main collinear AFM order and Cy and Fx are the weak-
AFM and weak-FM spin orders, respectively. All strains
are isotropic in the ac-plane, except for the �.�% strain
result which was calculated with strain applied only along
the z-axis, per experimental results. . . . . . . . . . . . . ���
REFERENCES

[�] A. C. Garcia-Castro, A. H. Romero, and E. Bousquet, Phys. Rev. Lett.

���, ������ (����).
[�] A. C. Garcia-Castro, N. A. Spaldin, A. H. Romero, and E. Bousquet, Phys.
Rev. B ��, ������ (����).
[�] A. C. Garcia-Castro, A. H. Romero, and E. Bousquet, Phys. Rev. B ��,
������ (����).
[�] A. C. Garcia-Castro, A. H. Romero, and E. Bousquet, Physica Status Solidi
(B) ���, ��� (����).
[�] A. C. Garcia-Castro, M. G. Vergniory, E. Bousquet, and A. H. Romero,
Phys. Rev. B ��, ������ (����).
[�] P. Borisov, T. A. Johnson, A. C. Garcia-Castro, A. KC, D. Schrecongost,
C. Cen, H. Romero, Aldo, and D. Lederman, ACS Applied Materials and
Interfaces �, ���� (����).
[�] W. Dai, S. Adhikari, A. C. Garcia-Castro, A. H. Romero, H. Lee, J.-W. Lee,
S. Ryu, C.-B. Eom, and C. Cen, Nano Letters ��, ���� (����).
[�] S. Adhikari, A. C. Garcia-Castro, A. H. Romero, S.-w. Park, C.-B. Eom,
and C. Cen, Adv. Func. Mater. Submitted, (����).
[�] S. Singh, A. C. Garcia-Castro, I. Valencia-Jaime, F. Munoz, and A. H.
Romero, Adv. Mater. To be submitted, (����).
[��] P. Borisov, T. A. Johnson, A. C. Garcia-Castro, E. Bousquet, K. Amit,
D. Schrecongost, C. Cen, A. H. Romero, and D. Lederman, Journal of
Applied Physics To be submitted, (����).
[��] M. E. Lines and A. M. Glass, Principles and applications of ferroelectrics
and related materials, clarendon press ed. (Oxford, ����).
[��] G. Sághi-Szabó, R. E. Cohen, and H. Krakauer, Phys. Rev. Lett. ��, ����
(����).
[��] N. A. Hill, J. Phys. Chem. B ���, ���� (����).
[��] A. J. Lovinger, Science ���, ���� (����).
[��] J. F. Scott and R. Blinc, Journal of Physics: Condensed Matter ��, ������
(����).

���
��� references

[��] D. Guo and N. Setter, Macromolecules ��, ���� (����).

[��] C. Ederer and N. A. Spaldin, Phys. Rev. B ��, ������ (����).
[��] J. F. Scott and R. Blinc, Journal of physics. Condensed matter : an Institute
of Physics journal ��, ������ (����).
[��] E. Bousquet, M. Dawber, N. Stucki, C. Lichtensteiger, P. Hermet,
S. Gariglio, J.-M. Triscone, and P. Ghosez, Nature ���, ��� (����).
[��] J. Haeni, P. Irvin, W. Chang, R. Uecker, and P. Reiche, Nature ���, ���
(����).
[��] H. Lee, H. Christen, and M. Chisholm, Nature ���, ��� (����).
[��] G. Hammerl and N. Spaldin, Science (New York, N.Y.) ���, ��� (����).
[��] M. J. Pitcher, P. Mandal, M. S. Dyer, J. Alaria, P. Borisov, H. Niu, J. B.
Claridge, and M. J. Rosseinsky, Science ���, ��� (����).
[��] J. J. Videau and J. Portier, in Inorganic Solid Fluorides, edited by P. Ha-
genmuller (Academic Press, ����) pp. ��� – ���.
[��] J. Ravez, in Inorganic Solid Fluorides, edited by P. Hagenmuller (Academic
Press, ����) pp. ��� – ���.
[��] R. Wolfe, A. J. Kurtzig, and R. C. LeCraw, Journal of Applied Physics ��,
���� (����).
[��] J.-M. Dance and A. Tressaud, in Inorganic Solid Fluorides, edited by P. Ha-
genmuller (Academic Press, ����) pp. ��� – ���.
[��] J.-M. Reau and J. Grannec, in Inorganic Solid Fluorides, edited by P. Ha-
genmuller (Academic Press, ����) pp. ��� – ���.
[��] E. Kemnitz and S. Rudiger, “High surface area metal fluorides as catalysts,”
in Functionalized Inorganic Fluorides (John Wiley and Sons, Ltd, ����) pp.
��–��.
[��] R. Martin, Electronic Structure: Basic Theory and Practical Methods, mar-
tin, richard ed. (Cambridge University Press, Cambridge, ����).
[��] E. Kaxiras, Atomic and Electronic Structure of Solids, kaxiras, e ed. (Cam-
bridge University Press, Cambridge, ����).
[��] P. Hohenberg and W. Kohn, Phys. Rev. ���, B��� (����).
[��] W. Kohn and L. J. Sham, Phys. Rev. ���, A���� (����).
[��] D. M. Ceperley and B. J. Alder, Phys. Rev. Lett. ��, ��� (����).
 references ���

[��] P. Ghosez and J. Junquera, “First-principles modeling of ferroelectric oxide

nanostructure,” in Handbook of Theoretical and Computational Nanotech-
nology, edited by M. Rieth and W. Schommers (American Scientific Pub-
lishers, ����) pp. ���–���.
[��] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. ��, ����
(����).
[��] R. Armiento and A. E. Mattsson, Phys. Rev. B ��, ������ (����).
[��] A. E. Mattsson and R. Armiento, Phys. Rev. B ��, ������ (����).
[��] J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria,
L. A. Constantin, X. Zhou, and K. Burke, Phys. Rev. Lett. ���, ������
(����).
[��] J. P. Perdew and Y. Wang, Phys. Rev. B ��, ����� (����).
[��] R. Pentcheva and W. E. Pickett, Journal of Physics: Condensed Matter ��,
������ (����).
[��] N. C. Bristowe, J. Varignon, D. Fontaine, E. Bousquet, and P. Ghosez, Nat
Commun � (����), article.
[��] A. I. Liechtenstein, V. I. Anisimov, and J. Zaanen, Phys. Rev. B ��, R����
(����).
[��] S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P.
Sutton, Phys. Rev. B ��, ���� (����).
[��] B. Himmetoglu, A. Floris, S. de Gironcoli, and M. Cococcioni, International
Journal of Quantum Chemistry ���, �� (����).
[��] M. Cococcioni and S. de Gironcoli, Phys. Rev. B ��, ������ (����).
[��] A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg, Rev. Mod. Phys.
��, �� (����).
[��] L. Hedin, Phys. Rev. ���, A��� (����).
[��] A. D. Becke, The Journal of Chemical Physics ��, ���� (����).
[��] A. D. Becke, Phys. Rev. A ��, ���� (����).
[��] J. P. Perdew, M. Ernzerhof, and K. Burke, The Journal of Chemical Physics
���, ���� (����).
[��] D. I. Bilc, R. Orlando, R. Shaltaf, G.-M. Rignanese, J. Íñiguez, and
P. Ghosez, Phys. Rev. B ��, ������ (����).
[��] Z. Wu and R. E. Cohen, Phys. Rev. B ��, ������ (����).
��� references

[��] J. Heyd, G. E. Scuseria, and M. Ernzerhof, The Journal of Chemical

Physics ���, ���� (����).
[��] A. V. Krukau, O. A. Vydrov, A. F. Izmaylov, and G. E. Scuseria, The
Journal of Chemical Physics ���, ������ (����).
[��] S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Rev. Mod.
Phys. ��, ��� (����).
[��] X. Gonze, Phys. Rev. A ��, ���� (����).
[��] K. Kunc and R. M. Martin, Phys. Rev. B ��, ���� (����).
[��] H. J. Monkhorst and J. D. Pack, Phys. Rev. B ��, ���� (����).
[��] D. S. Sholl and J. A. Steckel, “Dft calculations for surfaces of solids,” in
Density Functional Theory (John Wiley and Sons, Inc., ����) pp. ��–���.
[��] D. R. Hamann, M. Schlüter, and C. Chiang, Phys. Rev. Lett. ��, ����
(����).
[��] I. A. Kornev, L. Bellaiche, P. Bouvier, P.-E. Janolin, B. Dkhil, and
J. Kreisel, Phys. Rev. Lett. ��, ������ (����).
[��] I. a. Kornev and L. Bellaiche, Phase Transitions ��, ��� (����).
[��] P. E. Blöchl, Phys. Rev. B ��, ����� (����).
[��] E. Bousquet, N. A. Spaldin, and K. T. Delaney, Phys. Rev. Lett. ���,
������ (����).
[��] G. Kresse and J. Furthmüller, Phys. Rev. B ��, ����� (����).
[��] G. Kresse and D. Joubert, Phys. Rev. B ��, ���� (����).
[��] D. Hobbs, G. Kresse, and J. Hafner, Phys. Rev. B ��, ����� (����).
[��] X. Gonze and C. Lee, Phys. Rev. B ��, ����� (����).
[��] A. Togo, F. Oba, and I. Tanaka, Phys. Rev. B ��, ������ (����).
[��] D. Vanderbilt, Journal of Physics and Chemistry of Solids ��, ��� (����).
[��] D. Roberto, O. Roberto, C. Bartolomeo, R. Carla, S. V. R., and Z.-W. C.
M., Zeitschrift f¸r Kristallographie - Crystalline Materials ���, ��� (����).
[��] J. E. Dobson, G. Vignale, and M. P. Das, International Journal , Vol. ���
(Springer, Science+Business, New York, NY, ����) pp. ����–����.
[��] D. S. Sholl and J. A. Steckel, “What is density functional theory?” in Density
Functional Theory (John Wiley and Sons, Inc., ����) pp. �–��.
 references ���

[��] D. S. Sholl and J. A. Steckel, “Dft calculations for simple solids,” in Density
Functional Theory (John Wiley and Sons, Inc., ����) pp. ��–��.
[��] M. Fiebig, Journal of Physics D: Applied Physics ��, R��� (����).
[��] A. Poole, B. Roessli, O. Zaharko, and K. W. Kramer, Journal of Physics:
Condensed Matter ��, ������ (����).
[��] G. Nenert, Journal of Physics: Condensed Matter ��, ������ (����).
[��] G. Nenert and T. T. M. Palstra, Journal of Physics: Condensed Matter ��,
������ (����).
[��] W. Cochran, Phys. Rev. Lett. �, ��� (����).
[��] P. Ghosez, X. Gonze, and J.-P. Michenaud, EPL (Europhysics Letters) ��,
��� (����).
[��] R. Cohen, Nature ���, ��� (����).
[��] P. Ghosez, J.-P. Michenaud, and X. Gonze, Phys. Rev. B ��, ���� (����).
[��] P. Ghosez, E. Cockayne, U. V. Waghmare, and K. M. Rabe, Phys. Rev. B
��, ��� (����).
[��] M. Guennou, P. Bouvier, G. Garbarino, J. Kreisel, and E. K. H. Salje,
Journal of Physics: Condensed Matter ��, ������ (����).
[��] P. Edwardson, L. Boyer, R. Newman, D. Fox, J. Hardy, J. Flocken, R. Guen-
ther, and W. Mei, Physical Review B ��, ���� (����).
[��] Z. Friedman, M. Melamud, J. Makovsky, and H. Shaked, Phys. Rev. B �,
��� (����).
[��] D. Babel, in Structure and Bonding (����).
[��] J. W. Flocken, R. A. Guenther, J. R. Hardy, and L. L. Boyer, Phys. Rev.
B ��, ���� (����).
[��] W. Zhong, D. Vanderbilt, and K. M. Rabe, Phys. Rev. B ��, ���� (����).
[��] D. Vanderbilt and W. Zhong, Ferroelectrics ���, ��� (����).
[��] N. A. Benedek and C. J. Fennie, The Journal of Physical Chemistry C ���,
����� (����).
[��] S. Amisi, E. Bousquet, K. Katcho, and P. Ghosez, Phys. Rev. B ��, ������
(����).
[��] A. Togo, F. Oba, and I. Tanaka, Phys. Rev. B ��, ������ (����).
��� references

[��] X. Gonze, J.-M. Beuken, R. Caracas, F. Detraux, M. Fuchs, G.-M. Rig-

nanese, L. Sindic, M. Verstraete, G. Zerah, F. Jollet, M. Torrent, A. Roy,
M. Mikami, P. Ghosez, J.-Y. Raty, and D. Allan, Comp. Mater. Sci. ��,
��� (����).
[��] M. T. Dove, American Mineralogist ��, ��� (����).
[��] A. Katrusiak and A. Ratuszna, Solid State Communications ��, ��� (����).
[��] R. Shannon, Acta Crystallographica Section A: Crystal Physics A��, ���
(����).
[��] H. Bethe, Annalen der Physik ���, ��� (����).
[���] E. Bousquet and P. Ghosez, Phys. Rev. B ��, ������ (����).
[���] K. M. Rabe, C. Ahn, and J.-M. Triscone, Springer (����) p. ���.
[���] P. Ghosez, X. Gonze, and J.-P. Michenaud, Ferroelectrics ���, �� (����).
[���] I. a. Kornev and L. Bellaiche, Phase Transitions ��, ��� (����).
[���] F. Claeyssens, J. M. Oliva, D. Sanchez-Portal, and N. L. Allan, Chem.
Commun. , ���� (����).
[���] D. Khomskii, in Spin Electronics, Lecture Notes in Physics, Vol. ���, edited
by M. Ziese and M. Thornton (Springer Berlin Heidelberg, ����) pp. ��–
���.
[���] G. Sághi-Szabó, R. E. Cohen, and H. Krakauer, Phys. Rev. Lett. ��, ����
(����).
[���] A. N. Morozovska, E. A. Eliseev, M. D. Glinchuk, L.-Q. Chen, and
V. Gopalan, Phys. Rev. B ��, ������ (����).
[���] C. N. R. Rao, Ferroelectrics ���, ��� (����).
[���] I. Žutić, J. Fabian, and S. Sarma, Rev. Mod. Phys. ��, ��� (����).
[���] H. Bea, M. Gajek, M. Bibes, and A. Barthemy, J. Phys. Cond. Matter ��,
������ (����).
[���] P. Zubko, S. Gariglio, M. Gabay, P. Ghosez, and J.-M. Triscone, Annual
Review of Condensed Matter Physics �, ��� (����).
[���] Y.-H. Chu, L. W. Martin, M. B. Holcomb, and R. Ramesh, Materials Today
��, �� (����).
[���] R. J. Angel, J. Zhao, and N. L. Ross, Phys. Rev. Lett. ��, ������ (����).
[���] M. Murakami, K. Hirose, K. Kawamura, N. Sata, and Y. Ohishi, Science
(New York, N.Y.) ���, ��� (����).
 references ���

[���] T. Tsuchiya, J. Tsuchiya, K. Umemoto, and R. M. Wentzcovitch, Earth

and Planetary Science Letters ���, ��� (����).
[���] K. Hirose, American Mineralogist ��, ��� (����).
[���] S. Tateno, K. Hirose, N. Sata, and Y. Ohishi, Phys. Chem. Minerals ��,
��� (����).
[���] C. D. Martin, W. a. Crichton, H. Liu, V. Prakapenka, J. Chen, and J. B.
Parise, American Mineralogist ��, ���� (����).
[���] H. Yusa, Y. Shirako, M. Akaogi, H. Kojitani, N. Hirao, Y. Ohishi, and
T. Kikegawa, Inorganic Chemistry ��, ���� (����).
[���] D. P. Dobson, S. a. Hunt, A. Lindsay-Scott, and I. G. Wood, Physics of
the Earth and Planetary Interiors ���, ��� (����).
[���] Y. Shirako and H. Kojitani, American Mineralogist ��, ��� (����).
[���] N. A. Bogdanov, V. M. Katukuri, H. Stoll, J. van den Brink, and L. Hozoi,
Phys. Rev. B ��, ������ (����).
[���] T. B. Ballaran, R. G. Tronnes, and D. J. Frost, American Mineralogist ��,
���� (����).
[���] Y. Shirako, H. Satsukawa, H. Kojitani, T. Katsumata, M. Yoshida, Y. In-
aguma, K. Hiraki, T. Takahashi, K. Yamaura, E. Takayama-Muromachi,
and M. Akaogi, J. Phys.: Conference Series ���, ������ (����).
[���] H. Kojitani, Y. Shirako, and M. Akaogi, Physics of the Earth and Planetary
Interiors ���, ��� (����).
[���] Y. Shirako, H. Satsukawa, X. X. Wang, J. J. Li, Y. F. Guo, M. Arai, K. Ya-
maura, M. Yoshida, H. Kojitani, T. Katsumata, Y. Inaguma, K. Hiraki,
T. Takahashi, and M. Akaogi, Phys. Rev. B ��, ������ (����).
[���] K. Ohgushi, Y. Matsushita, N. Miyajima, Y. Katsuya, M. Tanaka, F. Izumi,
H. Gotou, Y. Ueda, and T. Yagi, Phys. Chem. Minerals ��, ��� (����).
[���] Y. Inaguma, K.-i. Hasumi, M. Yoshida, T. Ohba, and T. Katsumata, Inorg.
Chem. ��, ���� (����).
[���] Y. Shirako, H. Kojitani, M. Akaogi, K. Yamaura, and E. Takayama-
Muromachi, Physics and Chemistry of Minerals ��, ��� (����).
[���] K. Yamaura, Y. Shirako, H. Kojitani, M. Arai, D. P. Young, M. Akaogi,
M. Nakashima, T. Katsumata, Y. Inaguma, and E. Takayama-Muromachi,
J. Am. Chem. Soc. ���, ���� (����).
[���] M. Bremholm, S. E. Dutton, P. W. Stephens, and R. J. Cava, J. Solid
State Chem. ���, ��� (����).
��� references

[���] S. Yakovlev, M. Avdeev, and M. Mezouar, J. Solid State Chem. ���, ����
(����).
[���] F. L. Bernal, K. V. Yusenko, J. Sottmann, C. Drathen, J. Guignard, O. M.
Lovvik, W. A. Crichton, and S. Margadonna, Inorganic Chemistry ��,
����� (����), pMID: ��������, http://dx.doi.org/��.����/ic������y .
[���] F. Birch, Phys. Rev. ��, ��� (����).
[���] X. Gonze, Phys. Rev. A ��, ���� (����).
[���] X. Gonze, Phys. Rev. A ��, ���� (����).
[���] P. Daniel, M. Rousseau, A. Desert, A. Ratuszna, and F. Ganot, Phys. Rev.
B ��, ����� (����).
[���] A. Lane, D. Sharp, and J. Barraclough, J. Chem. Soc. ��, �� (����).
[���] V. Gava, A. L. Martinotto, and C. A. Perottoni, Phys. Rev. Lett. ���,
������ (����).
[���] J. R. Shane, J. Appl. Phys. ��, ���� (����).
[���] M. Akaogi, Y. Shirako, H. Kojitani, T. Nagakari, H. Yusa, and K. Yamaura,
Physics of the Earth and Planetary Interiors ���, ��� (����).
[���] S. Tateno, K. Hirose, N. Sata, and Y. Ohishi, Phys. Earth and Planetary
Interiors ���, �� (����).
[���] M. O’keeffe, B. Hyde, and J. Bovin, Phys. Chem. Minerals ���, ��� (����).
[���] C. D. Martin and J. B. Parise, Earth and Planetary Science Letters ���,
��� (����).
[���] B. Beznosikov and K. Aleksandrov, Crystallography Reports ��, ���
(����).
[���] D. Babel and A. Tressaud, in Inorganic Solid Fluorides, edited by P. Ha-
genmuller (Academic Press, ����) pp. �� – ���.
[���] M. Aroyo, L, J. M. Perez-Mato, C. Capillas, E. Kroumova, S. Ivantchev,
G. Madariaga, A. Kirov, and H. Wondratschek, Zeitschrift fur Kristallo-
graphie - Crystalline Materials ���, �� (����).
[���] M. I. Aroyo, A. Kirov, C. Capillas, J. M. Perez-Mato, and H. Wondratschek,
Acta Cryst ��, ��� (����).
[���] P. Anderson, Physical Review �� (����).
[���] J. Kanamori, Journal of Physics and Chemistry of Solids ��, �� (����).
 references ���

[���] J. C. Toledano and P. Toledano, The Landau Theory of Phase Transitions

(World Scientific, Singapore, ����).
[���] J. Kanamori, Prog. Theor. Phys. ��, ��� (����).
[���] J. Kanamori, Prog. Theor. Phys. ��, ��� (����).
[���] J. B. Goodenough, J. Phys. and Chem. Solid. �, ��� (����).
[���] E. Bousquet and N. Spaldin, Phys. Rev. B ��, ������ (����).
[���] I. Dzyaloshinsky, J. Phys. and Chem. Solid. �, ��� (����).
[���] T. Moriya, Phys. Rev. ���, �� (����).
[���] G. Maekawa, S., Tohyama, T., Barnes, S.E., Ishihara, S., Koshibae, W.,
Khaliullin, Physics of Transition Metal Oxides (Springer, Berlin, ����).
[���] C. Xu, B. Xu, Y. Yang, H. Dong, A. R. Oganov, S. Wang, W. Duan, B. Gu,
and L. Bellaiche, Phys. Rev. B ��, ������ (����).
[���] L. Martin, Y.-H. Chu, and R. Ramesh, Materials Science and Engineering:
R: Reports ��, �� (����).
[���] M. Murakami, K. Hirose, K. Kawamura, N. Sata, and Y. Ohishi, Science
���, ��� (����).
[���] Y. Shirako, Y. Shi, A. Aimi, D. Mori, H. Kojitani, K. Yamaura, Y. Inaguma,
and M. Akaogi, Journal of Solid State Chemistry ���, ��� (����).
[���] Y. Shirako, H. Satsukawa, X. X. Wang, J. J. Li, Y. F. Guo, M. Arai, K. Ya-
maura, M. Yoshida, H. Kojitani, T. Katsumata, Y. Inaguma, K. Hiraki,
T. Takahashi, and M. Akaogi, Phys. Rev. B ��, ������ (����).
[���] L. Du, X. Sheng, H. Weng, and X. Dai, EPL (Europhysics Letters) ���,
����� (����).
[���] K. Yamaura, Y. Shirako, H. Kojitani, M. Arai, D. P. Young, M. Akaogi,
M. Nakashima, T. Katsumata, Y. Inaguma, and E. Takayama-Muromachi,
Journal of the American Chemical Society ���, ���� (����).
[���] D. P. Dobson, S. A. Hunt, A. Lindsay-Scott, and I. G. Wood, Physics of
the Earth and Planetary Interiors ���, ��� (����).
[���] C. Weingart, N. Spaldin, and E. Bousquet, Phys. Rev. B ��, ������ (����).
[���] K. Momma and F. Izumi, Journal of Applied Crystallography ��, ����
(����).
[���] H. J. Xiang, E. J. Kan, S.-H. Wei, M.-H. Whangbo, and X. G. Gong, Phys.
Rev. B ��, ������ (����).
��� references

[���] E. E. Rodriguez, F. Poineau, A. Llobet, B. J. Kennedy, M. Avdeev, G. J.

Thorogood, M. L. Carter, R. Seshadri, D. J. Singh, and A. K. Cheetham,
Phys. Rev. Lett. ���, ������ (����).
[���] B. H. Kim and B. I. Min, New Journal of Physics ��, ������ (����).
[���] W. Eerenstein, N. D. Mathur, and J. F. Scott, Nature ���, ��� (����).
[���] L. Martin, Y.-H. Chu, and R. Ramesh, Materials Science and Engineering:
R: Reports ��, �� (����).
[���] O. Diéguez, K. M. Rabe, and D. Vanderbilt, Phys. Rev. B ��, ������
(����).
[���] E. Bousquet and N. Spaldin, Phys. Rev. Lett. ���, ������ (����).
[���] C.-J. Eklund, C. Fennie, and K. Rabe, Phys. Rev. B ��, ������ (����).
[���] T. Günter, E. Bousquet, a. David, P. Boullay, P. Ghosez, W. Prellier, and
M. Fiebig, Phys. Rev. B ��, ������ (����).
[���] C. Ederer and N. A. Spaldin, Phys. Rev. Lett. ��, ������ (����).
[���] N. A. Hill, J. Phys. Chem. B , ���� (����).
[���] P. Berastegui, S. Hull, and S.-G. Eriksson, Journal of Physics: Condensed
Matter ��, ���� (����).
[���] E. H. Smith, N. A. Benedek, and C. J. Fennie, Inorganic Chemistry ��,
���� (����).
[���] C.-g. Duan, W. Mei, J. Liu, W.-G. Yin, J. Hardy, R. Smith, M. Mehl, and
L. Boyer, Phys. Rev. B ��, � (����).
[���] J. Young, A. Stroppa, S. Picozzi, and J. M. Rondinelli, Journal of Physics:
Condensed Matter ��, ������ (����).
[���] Y. J. Choi, C. L. Zhang, N. Lee, and S.-W. Cheong, Phys. Rev. Lett. ���,
������ (����).
[���] S.-W. Cheong and M. Mostovoy, Nature Mater. �, �� (����).
[���] D. N. Argyriou, Physics �, �� (����).
[���] G. Bester, X. Wu, D. Vanderbilt, and A. Zunger, Phys. Rev. Lett. ��,
������ (����).
[���] S. Sanna, C. Thierfelder, S. Wippermann, T. P. Sinha, and W. G. Schmidt,
Phys. Rev. B ��, ������ (����).
[���] T. Furuta and K. Miura, Solid State Communications ���, ���� (����).
 references ���

[���] J. R. Shane, Journal of Applied Physics ��, ���� (����).

[���] A. S. Borovik-Romanov and H. Grimmer, International Tables for Crystal-
lography D (����).
[���] V. J. Folen, G. T. Rado, and E. W. Stalder, Phys. Rev. Lett. �, ��� (����).
[���] E. Kita, K. Siratori, and A. Tasaki, Journal of Applied Physics ��, ����
(����).
[���] H. Wiegelmann, A. G. M. Jansen, P. Wyder, J.-P. Rivera, and H. Schmid,
Ferroelectrics ���, ��� (����).
[���] J. Íñiguez, Phys. Rev. Lett. ���, ������ (����).
[���] E. Bousquet, N. A. Spaldin, and K. T. Delaney, Phys. Rev. Lett. ���,
������ (����).
[���] J. Mannhart, D. Blank, H. Hwang, A. Millis, and J.-M. Triscone, MRS
Bulletin ��, ���� (����).
[���] A. Joshua, S. Pecker, J. Ruhman, E. Altman, and S. Ilani, Nature commu-
nications �, ���� (����).
[���] S. Caprara, F. Peronaci, and M. Grilli, Phys. Rev. Lett. ���, ������ (����).
[���] B. R. K. Nanda and S. Satpathy, Phys. Rev. B ��, ������ (����).
[���] N. Benedek and C. Fennie, Phys. Rev. Lett. ���, ������ (����).
[���] A. T. Mulder, N. A. Benedek, J. M. Rondinelli, and C. J. Fennie, Advanced
Functional Materials ��, ���� (����).
[���] Z. Zanolli, J. C. Wojdeł, J. Íñiguez, and P. Ghosez, Phys. Rev. B ��,
������ (����).
[���] M. A. Carpenter, E. K. H. Salje, and C. J. Howard, Phys. Rev. B ��,
������ (����).
[���] W. Pang, R. L. Stamps, L. Malkinski, Z. Celinski, and D. Skrzypek, Journal
of Applied Physics ��, ���� (����).
[���] S. Widuch, R. L. Stamps, D. Skrzypek, and Z. Celinski, Journal of Physics
D: Applied Physics ��, ������ (����).
[���] E. J. Moon, Y. Xie, E. D. Laird, D. J. Keavney, C. Y. Li, and S. J. May,
Journal of the American Chemical Society ���, ���� (����).
[���] M. AI-Mamouri, P. P. Edwards, C. Greaves, and M. Slaski, Nature ���,
��� (����).
��� references

[���] M. Sturza, S. Daviero-Minaud, H. Kabbour, O. Gardoll, and O. Mentre,

Chemistry of Materials ��, ���� (����).
[���] M. Sturza, H. Kabbour, S. Daviero-Minaud, D. Filimonov, K. Pokholok,
N. Tiercelin, F. Porcher, L. Aldon, and O. Mentre, Journal of the American
Chemical Society ���, ����� (����).
[���] A. D. Becke, The Journal of Chemical Physics ��, ���� (����).
[���] A. D. Becke, The Journal of Chemical Physics ��, ���� (����).
[���] in Ternary compounds, organic semiconductors, edited by O. Madelung,
U. Rassler, and M. Schulz (Springer Berlin Heidelberg, ����) pp. � – �.
[���] P. Yu, Y. Chu, and R. Ramesh, Materials Today ��, ��� (����).
[���] N. Bristowe, E. Artacho, and P. Littlewood, Phys. Rev. B ��, ������
(����).
[���] K. Yamauchi and S. Picozzi, Phys. Rev. Lett. ���, ������ (����).
[���] P. D. C. King, R. H. He, T. Eknapakul, P. Buaphet, S.-K. Mo, Y. Kaneko,
S. Harashima, Y. Hikita, M. S. Bahramy, C. Bell, Z. Hussain, Y. Tokura,
Z.-X. Shen, H. Y. Hwang, F. Baumberger, and W. Meevasana, Phys. Rev.
Lett. ���, ������ (����).
[���] A. D. Caviglia, M. Gabay, S. Gariglio, N. Reyren, C. Cancellieri, and J.-M.
Triscone, Phys. Rev. Lett. ���, ������ (����).
[���] Z. Zhong, A. Tóth, and K. Held, Phys. Rev. B ��, ������ (����).
[���] S. Hurand, A. Jouan, C. Feuillet-Palma, G. Singh, J. Biscaras, E. Lesne,
N. Reyren, A. Barthélémy, M. Bibes, J. E. Villegas, C. Ulysse, X. Lafosse,
M. Pannetier-Lecoeur, S. Caprara, M. Grilli, J. Lesueur, and N. Bergeal,
Scientific Reports �, ����� EP (����).
[���] Y. A. Bychkov and E. I. Rashba, Journal of Physics C: Solid State Physics
��, ���� (����).
[���] H. Nakamura, T. Koga, and T. Kimura, Phys. Rev. Lett. ���, ������
(����).
[���] H. Nakamura and T. Kimura, Phys. Rev. B ��, ������ (����).
[���] H. Lee and H. J. Choi, Phys. Rev. B ��, ������ (����).
[���] S. R. Park and C. Kim, Journal of Electron Spectroscopy and Related
Phenomena ���, � (����), special issue on electron spectroscopy for Rashba
spin-orbit interaction.
 references ���

[���] K. Shanavas, Journal of Electron Spectroscopy and Related Phenomena

���, ��� (����).
[���] K. Ishizaka, M. S. Bahramy, H. Murakawa, M. Sakano, T. Shimo-
jima, T. Sonobe, K. Koizumi, S. Shin, H. Miyahara, A. Kimura,
K. Miyamoto, T. Okuda, H. Namatame, M. Taniguchi, R. Arita, N. Na-
gaosa, K. Kobayashi, Y. Murakami, R. Kumai, Y. Kaneko, Y. Onose, and
Y. Tokura, Nat Mater ��, ��� (����).
[���] O. Bleibaum and S. Wachsmuth, Phys. Rev. B ��, ������ (����).
[���] R. Moriya, K. Sawano, Y. Hoshi, S. Masubuchi, Y. Shiraki, A. Wild, C. Neu-
mann, G. Abstreiter, D. Bougeard, T. Koga, and T. Machida, Phys. Rev.
Lett. ���, ������ (����).
[���] B. Yin, P. Aguado-Puente, S. Qu, and E. Artacho, Phys. Rev. B ��, ������
(����).
[���] J. T. Zhang, J. Kim, M. Huefner, C. Ye, S. Kim, P. C. Canfield, R. Prozorov,
O. M. Auslaender, and J. E. Hoffman, Phys. Rev. B ��, ������ (����).
[���] B. Morosin, E. Venturini, J. Schirber, R. Dunn, and P. Newcomer, Physica
C: Superconductivity ���, ��� (����).
[���] K. Ueno, S. Nakamura, H. Shimotani, H. T., Yuan, N. Kimura, T. Nojima,
H. Aoki, Y. Iwasa, and M. Kawasaki, Nat Nano �, ��� (����).
[���] J. B. Goodenough and M. Longo, “� crystallographic and magnetic prop-
erties of perovskite and perovskite-related compounds,” in Group III Con-
densed Matter, Part A, Vol. �A, edited by K.-H. Hellwege and A. M. Hell-
wege (Springer, Landolt-Bornstein, ����) pp. ���–.
[���] E. F. Bertaut (Spin Configurations of Ionic Structures: Theory and Practice,
New York, vol �, (����)).
[���] M. DiDomenico, M. Eibschütz, H. Guggenheim, and I. Camlibel, Solid
State Communications �, ���� (����).
[���] E. T. Keve, S. C. Abrahams, and J. L. Bernstein, The Journal of Chemical
Physics ��, ���� (����).
[���] D. E. Cox, M. Eibschütz, H. J. Guggenheim, and L. Holmes, Journal of
Applied Physics ��, ��� (����).
[���] C. Ederer and N. A. Spaldin, Phys. Rev. B ��, ������ (����).
[���] M. Eibschütz, H. Guggenheim, S. Wemple, I. Camlibel, and
M. DiDomenico, Physics Letters A ��, ��� (����).
[���] M. Eibschütz, L. Holmes, H. J. Guggenheim, and D. E. Cox, Phys. Rev. B
�, ���� (����).
��� references

[���] R. Zorin, B. Al’shin, and D. Astrov, Soviet Physics JETP ��, ��� (����).
[���] M. Bibes and A. Barthélémy, Electron Devices, IEEE Transactions on ��,
���� (����).
[���] S. Fusil, V. Garcia, A. Barthélémy, and M. Bibes, Annual Review of Ma-
terials Research ��, �� (����).
[���] L. Martin, Y.-H. Chu, and R. Ramesh, Materials Science and Engineering:
R: Reports ��, �� (����).
[���] N. Reyren, S. Thiel, A. D. Caviglia, L. F. Kourkoutis, G. Hammerl,
C. Richter, C. W. Schneider, T. Kopp, A.-S. Rüetschi, D. Jaccard,
M. Gabay, D. A. Muller, J.-M. Triscone, and J. Mannhart, Science ���,
���� (����).
[���] J. Shen, H. Lee, R. Valentí, and H. O. Jeschke, Phys. Rev. B ��, ������
(����).
[���] A. F. Santander-Syro, O. Copie, T. Kondo, F. Fortuna, S. Pailhès, R. Weht,
X. G. Qiu, F. Bertran, A. Nicolaou, a. Taleb-Ibrahimi, P. Le Fèvre, G. Her-
ranz, M. Bibes, N. Reyren, Y. Apertet, P. Lecoeur, a. Barthélémy, and
M. J. Rozenberg, Nature ���, ��� (����).
[���] P. D. C. King, S. McKeown Walker, A. Tamai, A. de la Torre, T. Eknapakul,
P. Buaphet, S.-K. Mo, W. Meevasana, M. S. Bahramy, and F. Baumberger,
Nature Communications �, � (����).
[���] A. F. Santander-Syro, F. Fortuna, C. Bareille, T. C. Rödel, G. Landolt,
N. C. Plumb, J. H. Dil, and M. Radović, Nat. Mater ��, ���� (����).
[���] M. Dawber and E. Bousquet, MRS Bulletin ��, ���� (����).
[���] N. C. Bristowe, P. Ghosez, P. B. Littlewood, and E. Artacho, Journal of
Physics: Condensed Matter ��, ������ (����).
[���] Z. Q. Liu, C. J. Li, W. M. Lü, X. H. Huang, Z. Huang, S. W. Zeng, X. P.
Qiu, L. S. Huang, A. Annadi, J. S. Chen, J. M. D. Coey, T. Venkatesan,
and Ariando, Phys. Rev. X �, ������ (����).
[���] A. H. Romero and F. Munoz, “Pyprocar,” http://sourceforge.net/p/
pyprocar/PyPROCAR/HEAD/tree/ (����).

[���] S. McKeown Walker, A. de la Torre, F. Y. Bruno, A. Tamai, T. K. Kim,

M. Hoesch, M. Shi, M. S. Bahramy, P. D. C. King, and F. Baumberger,
Phys. Rev. Lett. ���, ������ (����).
[���] N. C. Plumb, M. Salluzzo, E. Razzoli, M. Månsson, M. Falub, J. Krempasky,
C. E. Matt, J. Chang, M. Schulte, J. Braun, H. Ebert, J. Minár, B. Delley,
K.-J. Zhou, T. Schmitt, M. Shi, J. Mesot, L. Patthey, and M. Radović,
Phys. Rev. Lett. ���, ������ (����).
 references ���

[���] G. Khalsa, B. Lee, and A. H. MacDonald, Phys. Rev. B ��, ������ (����).
[���] W. D. Rice, P. Ambwani, M. Bombeck, J. D. Thompson, G. Haugstad,
C. Leighton, and S. a. Crooker, Nature materials ��, ��� (����).
[���] H. Choi, J. D. Song, K.-R. Lee, and S. Kim, Inorganic Chemistry ��, ����
(����).
[���] L. P. Gor’kov, Journal of Physics: Condensed Matter ��, ������ (����).
[���] C. Lin and A. A. Demkov, Phys. Rev. Lett. ���, ������ (����).
[���] W. Meevasana, P. D. C. King, R. H. He, S.-K. Mo, M. Hashimoto, a. Tamai,
P. Songsiriritthigul, F. Baumberger, and Z.-X. Shen, Nature materials ��,
��� (����).
[���] G. Hautier, A. Miglio, D. Waroquiers, G.-M. Rignanese, and X. Gonze,
Chemistry of Materials ��, ���� (����).
[���] F. Reinert, G. Nicolay, S. Schmidt, D. Ehm, and S. Hüfner, Phys. Rev. B
��, ������ (����).
[���] O. Krupin, G. Bihlmayer, K. Starke, S. Gorovikov, J. E. Prieto, K. Döbrich,
S. Blügel, and G. Kaindl, Phys. Rev. B ��, ������ (����).
[���] J. P. Perdew and A. Zunger, Phys. Rev. B ��, ���� (����).
[���] A. I. Liechtenstein, V. I. Anisimov, and J. Zaanen, Phys. Rev. B ��, R����
(����).
[���] B. J. Kennedy, C. J. Howard, and B. C. Chakoumakos, Phys. Rev. B ��,
���� (����).
[���] D. Vanderbilt, Journal of Physics and Chemistry of Solids ��, ��� (����).
[���] N. A. Benedek, J. M. Rondinelli, H. Djani, P. Ghosez, and P. Lightfoot,
Dalton Trans. ��, ����� (����).
[���] J. Sánchez-Barriga, A. Varykhalov, J. Braun, S.-Y. Xu, N. Alidoust, O. Ko-
rnilov, J. Minár, K. Hummer, G. Springholz, G. Bauer, R. Schumann, L. V.
Yashina, H. Ebert, M. Z. Hasan, and O. Rader, Phys. Rev. X �, ������
(����).
[���] C.-H. Park and S. G. Louie, Phys. Rev. Lett. ���, ������ (����).
[���] C. Jozwiak, C.-H. Park, K. Gotlieb, C. Hwang, D.-H. Lee, S. G. Louie, J. D.
Denlinger, C. R. Rotundu, R. J. Birgeneau, Z. Hussain, and A. Lanzara,
Nature Physics �, ��� (����).
[���] J. M. Rondinelli and N. A. Spaldin, Phys. Rev. B ��, ������ (����).
��� references

[���] T. Kuschel and G. Reiss, Nature Nanotechnology ��, �� (����).

[���] W. Zhang, M. B. Jungfleisch, W. Jiang, J. Sklenar, F. Y. Fradin, J. E.
Pearson, J. B. Ketterson, and A. Hoffmann, Journal of Applied Physics
���, ������ (����).
[���] J. C. Ho, R. Yerushalmi, Z. A. Jacobson, Z. Fan, R. L. Alley, and A. Javey,
Nat Mater �, �� (����).
[���] P. Zhang, E. Tevaarwerk, B.-N. Park, D. E. Savage, G. K. Celler, I. Kneze-
vic, P. G. Evans, M. A. Eriksson, and M. G. Lagally, Nature ���, ���
(����).
[���] C. Di Valentin, G. Pacchioni, and A. Selloni, Phys. Rev. Lett. ��, ������
(����).
[���] P. Strobel, M. Riedel, J. Ristein, and L. Ley, Nature ���, ��� (����).
[���] C. Coletti, C. Riedl, D. S. Lee, B. Krauss, L. Patthey, K. von Klitzing, J. H.
Smet, and U. Starke, Phys. Rev. B ��, ������ (����).
[���] S. Kobayashi, T. Nishikawa, T. Takenobu, S. Mori, T. Shimoda, T. Mitani,
H. Shimotani, N. Yoshimoto, S. Ogawa, and Y. Iwasa, Nat. Mater. �, ���
(����).
[���] J. Mannhart and D. G. Schlom, Science ���, ���� (����).
[���] S. Thiel, G. Hammerl, A. Schmehl, C. W. Schneider, and J. Mannhart,
Science ���, ���� (����).
[���] B. Kalisky, J. A. Bert, B. B. Klopfer, C. Bell, H. K. Sato, M. Hosoda,
Y. Hikita, H. Y. Hwang, and K. A. Moler, Nat Commun �, ��� (����).
[���] P. Irvin, Y. Ma, D. F. Bogorin, C. Cen, C. W. Bark, C. M. Folkman, C.-B.
Eom, and J. Levy, Nat Photon �, ��� (����).
[���] N. Nakagawa, H. Y. Hwang, and D. A. Muller, Nat Mater �, ��� (����).
[���] A. Janotti, L. Bjaalie, L. Gordon, and C. G. Van de Walle, Phys. Rev. B
��, ������ (����).
[���] C. Cen, S. Thiel, J. Mannhart, and J. Levy, Science ���, ���� (����).
[���] F. Bi, D. F. Bogorin, C. Cen, C. W. Bark, J.-W. Park, C.-B. Eom, and
J. Levy, Applied Physics Letters ��, ������ (����).
[���] C. Cen, S. Thiel, G. Hammerl, C. W. Schneider, K. E. Andersen, C. S.
Hellberg, J. Mannhart, and J. Levy, Nat Mater �, ��� (����).
[���] J. Neugebauer and M. Scheffler, Phys. Rev. B ��, ����� (����).
 references ���

[���] C. L. Jia, S. B. Mi, M. Faley, U. Poppe, J. Schubert, and K. Urban, Phys.

Rev. B ��, ������ (����).
[���] M. Gabay and J.-M. Triscone, Nat Phys �, ��� (����), news and Views.
[���] H. Peng, H. J. Xiang, S.-H. Wei, S.-S. Li, J.-B. Xia, and J. Li, Phys. Rev.
Lett. ���, ������ (����).
[���] S. Nazir, C. Bernal, and K. Yang, ACS Applied Materials & Interfaces �,
���� (����).
D E C L A R AT I O N

This thesis is a presentation of my original research work. Wherever contributions

of others are involved, every effort is made to indicate this clearly, with due
reference to the literature, and acknowledgement of collaborative research and
discussions. The work was done under the guidance of Professor Aldo Humberto
Romero Castro at Cinvestav-Querétaro and Dr. Eric Bousquet at Université de
Liège.

Querétaro, Abril ����

Andrés Camilo García

Castro, Abril ����
EL JURADO DESIGNADO POR LA UNIDAD QUERÉTARO DEL CENTRO DE
INVESTIGACIÓN Y DE ESTUDIOS AVANZADOS DEL INSTITUTO POLITÉC-
NICO NACIONAL Y LA UNIVERSITÉ DE LIÈGE, APROBÓ LA TESIS DOC-
TORAL DE ANDRÉS CAMILO GARCÍA CASTRO TITULADA: “ESTUDIO Y
CARACTERIZACIÓN DE MATERIALES MULTIFERROICOS Y MAGNETOELEC-
TRICOS MEDIANTE CALCULOS DE PRIMEROS PRINCIPIOS”, FIRMAN AL
CALCE DE COMÚN ACUERDO LOS INTEGRANTES DE DICHO JURADO, EN
LA CIUDAD DE QUERÉTARO, QRO., A LOS DOCE DÍAS DEL MES DE ABRIL
DEL DOS MIL DIECISÉIS.

Dr. Aldo Humberto Romero Castro, CINVESTAV-Querétaro

Dr. Eric Bousquet, Université de Liège

Dra. Nicola Spaldin, ETH-Zürich

Dr. Philippe Ghosez, Université de Liège

Dra. Rebeca Castanedo Pérez, CINVESTAV-Querétaro

Dr. Sergio Jiménez Sandoval, CINVESTAV-Querétaro

Dr. Arturo Mendoza Galván, CINVESTAV-Querétaro

THE THEIS JURY DESIGNATED BY THE CENTRO DE INVESTIGACIÓN Y DE
ESTUDIOS AVANZADOS DEL INSTITUTO POLITÉCNICO NACIONAL QUERE-
TARO UNIT AND THE UNIVERSITÉ DE LIÈGE, APPROVED THE PH.D THESIS
OF ANDRÉS CAMILO GARCÍA CASTRO ENTITLED “STUDY AND CHARAC-
TERIZATION OF MAGNETIC AND MULTIFERROIC MATERIALS BY FIRST-
PRINCIPLES CALCULATIONS”. BELOW THE MEMBERS OF THIS JURY SIGN
IN COMMUN AGREEMENT IN THE CITY OF QUERETARO, QRO., AT THE
TWELVE DAYS OF APRIL OF TWO THOUSAND AND SIXTEEN.

Dr. Aldo Humberto Romero Castro, CINVESTAV-Querétaro

Dr. Eric Bousquet, Université de Liège

Dra. Nicola Spaldin, ETH-Zürich

Dr. Philippe Ghosez, Université de Liège

Dra. Rebeca Castanedo Pérez, CINVESTAV-Querétaro

Dr. Sergio Jiménez Sandoval, CINVESTAV-Querétaro

Dr. Arturo Mendoza Galván, CINVESTAV-Querétaro