2018LEMA1029

THESE DE DOCTORAT DE
LE MANS UNIVERSITE
COMUE UNIVERSITE BRETAGNE LOIRE
ECOLE DOCTORALE N° 596

Matière Molécules et Matériaux
Spécialité : Ondes et Matériaux
Par
Ngoc Minh TRAN
Applications of nonlinear magneto-photonics at the nanoscale
Thèse présentée et soutenue au Mans, le 13 décembre 2018

Unité de recherche : Institut des Molécules et Matériaux du Mans (IMMM) – Le Mans Université
Thèse N° : 2018LEMA1029
Rapporteurs avant soutenance : Composition du Jury :

Michael Farle Prof., Universität Duisburg-Essen Michael Farle Prof., Universität Duisburg-Essen
Peter Gaal Prof., INF - Universität Hamburg Peter Gaal Prof., INF - Universität Hamburg
Ulrike Woggon Prof., IOAP - Technische Universität Berlin
Béatrice Dagens DR, C2N - Université Paris-Saclay
Directeur de thèse
Vasily Temnov CNRS CRCN, IMMM CNRS 6283
Encadrant de thèse
Gwenaëlle Vaudel CNRS IR1, IMMM CNRS 6283
Titre : Application de la magneto-photonique non-linéaire à l'échelle nanométrique
Mots clés : magneto-plasmonique, génération de seconde harmonique, nanostructures hybrides

métal/ferromagnétiques, géométrie Kretschmann, réseaux de diffraction magnétique, modèles phénoménologiques
d’optique non-linéaire.
Résumé : La génération de seconde harmonique signal ont été mesurés par la technique de l’effet Kerr
magnétique (mSHG pour magnetic Second Harmonic magnéto-optique transverse (MOKE) en fonction de
Generation) est un phénomène physique très sensible l’angle d’incidence. Via l’utilisation de sources lasers
apparaissant grâce aux brisures de symétrie aux niveaux femtosecondes émettant dans le proche infrarouge,
des surfaces et interfaces des structures métalliques domaine spectral où les variations de la dispersion des
magnétiques. Elle constitue donc un outil puissant pour SPP et du coefficient d’amortissement sont significatives,
explorer ce type d’interfaces et des nanostructures. Dans nous avons pu distinguer les différentes contributions
ce travail, nous nous intéressons aux couplages et linéaires et non-linéaires aux processus d’excitation. Ce
interactions entre la mSHG et les ondes travail de thèse a ainsi permis de montrer que l’accord de
électromagnétiques pouvant se propager en surface des phase entre la mSHG et les ondes électromagnétiques de
matériaux. Un intérêt spécifique est porté sur l’excitation surface peuvent contribuer très efficacement à
de (i) plasmon polaritons de surfaces (SPP) dans des films l’augmentation des signaux SH et de contraste magnétique
métalliques en structures multicouches, (ii) d’anomalies associé. Ces résultats sont très importants pour le contrôle
de diffraction (dîtes de Wood) dans des nanostructures de la réponse non-linéaire magnéto-optique des matériaux
métalliques périodiques. Pour étudier l’influence de et peuvent trouver leurs applications dans le
l’excitation linéaire et non-linéaire des SPP sur la mSHG, développement de composants pour le stockage
l’intensité du signal réfléchi par génération de seconde magnétique des données, de biocapteurs sans marquage,
harmonique (SH) et le contraste magnétique lié à ce ou de bascules magnéto-optiques non-linéaires.
Title : Applications of nonlinear magneto-photonics at the nanoscale
Keywords : magneto-plasmonics, magnetic second harmonic generation, hybrid metal-ferromagnet nanostructures,

magnetic diffractions gratings, Kretschmann configuration, phenomenological models in nonlinear optics.
Abstract : Owing to surface and interface sensitivity, the angle of incidence. The use of different femtosecond
the magnetic Second Harmonic Generation (mSHG) light sources in the near-infrared optical range, where the
represents a useful tool to probe magnetic interfaces and SPP dispersion and damping exhibit significant variations,
nanostructures. This work investigates the coupling and made it possible to disentangle linear and nonlinear
interaction of the mSHG with electromagnetic waves contributions to the excitation of surface waves. In this
propagating along the surface. Two types of surface thesis, it is proven that phase-matching of the mSHG and
waves have been studied: (i) surface plasmon polaritons surface electromagnetic waves can lead to the
(SPP) at surfaces of metallic thin films and multilayers, enhancement of both the SH yield and the nonlinear
and (ii) the diffraction anomaly at the surface of magneto-optical signal. These results are important for
periodically arranged metallic nanostructures. To study controlling of the nonlinear magneto-optical response and
influence of linear and nonlinear excitation of surface could impact the development of magnetic storage
waves on the mSHG, the reflected second harmonic (SH) devices, label-free biosensors and nonlinear magneto-
intensity and the magnetic SH contrast in the transverse optical switches.
magneto-optical geometry were measured as a function of
Abstract
Owing to surface and interface sensitivity, the magnetic Second Harmonic Genera-
tion (mSHG) represents a useful tool to probe magnetic interfaces and nanostructures.
This work investigates the coupling and interaction of the mSHG with electromagnetic
waves propagating along the surface. Two types of surface waves have been studied:
(i) surface plasmon polaritons (SPP) at surfaces of metallic thin films and multilay-
ers, and (ii) the diffraction anomaly at the surface of periodically arranged metallic
nanostructures. To study influence of linear and nonlinear excitation of surface waves
on the mSHG, the reflected second harmonic (SH) intensity and the magnetic SH con-
trast in the transverse magneto-optical geometry were measured as a function of the
angle of incidence. The use of different femtosecond light sources in the near-infrared
optical range, where the SPP dispersion and damping exhibit significant variations,
made it possible to disentangle linear and nonlinear contributions to the excitation of
surface waves. In this thesis, it is proven that phase-matching of the mSHG and sur-
face electromagnetic waves can lead to the enhancement of both the SH yield and the
nonlinear magneto-optical signal. These results are important for controlling of the
nonlinear magneto-optical response and could impact the development of magnetic
storage devices, label-free biosensors and nonlinear magneto-optical switches.
The coupling of the collective oscillation of electrons at the metallic surface and elec-
tromagnetic wave results in the SPPs, in which the electromagnetic field is highly con-
centrated within a sub-wavelength volume near the surface. Employing Kretschmann
configuration, SPPs can be excited in hybrid metal-ferromagnet multilayers. Here, two
sets of distinct structures have been inspected: Au/Co bilayers and Au/Co/Ag trilay-
ers. In both types of structures, it has been proven that the strong plasmonic field
enhancement at the fundamental frequency is the main reason for the characteristic
peak in the angular dependence of the SH output. In contrast, the large-angle SH dip
originates from the interference of the non-resonant SH field and the radiative SPP at
the doubled SH frequency. Comparison of the experimental data with the phenomeno-
logical model reveals the fingerprints of linear SPP and second-harmonic SPP in the
magnetic SH contrast, where the Lorentzian phase-matching function represents the
basic result in nonlinear magneto-plasmonics. Thus, the angular dependence of the
far-field mSHG response can be utilized as a measure of the excitation efficiency of
SPPs. Also, it has been shown that upon the excitation of the nonlinear SPP, the
mSHG signal can be enhanced up to 60%.
Nonlinear magneto-photonics on periodic structures has been performed in a dif-
ferent regime where grating effects dominate over the plasmonic ones. Periodic arrays
of nickel nanodimers on a silicon substrate were fabricated by electron lithography
in such a way that their low surface filling factor ruled out the SPP excitation. In
this case, surface electromagnetic waves can be generated under conditions of Wood’s
anomaly, when the diffraction order emerging at the surface depletes the energy of
other diffraction orders. The observation of Wood’s anomaly depends on the ratio of
the fundamental optical wavelength λ to the grating period Λ > λ/2. In the case
of nonlinear-optical excitation, multiple harmonics become involved making the phe-
nomenon of such mixed Wood’s anomaly very complicated. However, in the case of
I
SH generation and the grating period falling within the range of λ/4 < Λ < λ/2, the
excitation of the nonlinear Wood’s anomaly is allowed exclusively at the SH wavelength
while the linear one is prohibited. From the comparison of the angular dependencies
of the linear reflectivity and the mSHG response, the redistribution of energy between
different diffractions orders in the nonlinear regime is about one order of magnitude
higher than that in the linear regime. Furthermore, the angular spectrum of the SH
magnetic contrast clearly shows the fingerprints of the nonlinear diffraction anomaly.
II
IV
List of Figures
2.1 Comparison of the dielectric function of silver estimated by Drude model

and measured experimentally . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Three configurations of magneto-optical Kerr effects . . . . . . . . . . . 13
2.3 Comparison of harmonic and anharmonic potential . . . . . . . . . . . 14
2.4 Illustration for the anharmonic oscillation of a free electron . . . . . . . 18
2.5 Phasor diagram of the non-magnetic and magnetic SH fields . . . . . . 20
2.6 Boundary conditions for p-polarized light at the interface between metal
and dielectric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.7 Real and imaginary parts of dielectric functions of silver, gold and cobalt 22
2.8 Vector field of SPP localized at gold/air interface at wavelength 1560 nm 23
2.9 Dispersion relations of SPP at silver/air interface . . . . . . . . . . . . 24
2.10 Different geometries where SPP is excited by prism coupling . . . . . . 26
2.11 Excitation of SPP by a diffraction grating. . . . . . . . . . . . . . . . 27
2.12 Comparison between the angular dependence of linear reflectivity mea-
sured experimentally and simulated using MTM . . . . . . . . . . . . . 28
2.13 Propagation length and skin depth of the SPP field excited at metal/air
surface in different structures . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Kretschmann configuration applied for the metal/ferromagnet bilayers

and trilayers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Scheme of a linear magneto-plasmonic setup employing Kretschmann
configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Top-view of a SH nonlinear magneto-plasmonic setup based on Kretschmann
geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Dependence of magnetic field strength on distance from the pole of a
Neodymium permanent magnet . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Fluence dependence of the signal detected by the PMT . . . . . . . . . 36
3.6 An example of the counting scheme . . . . . . . . . . . . . . . . . . . . 37
3.7 Angular dependence of linear transversal MOKE measured on the bi-
layers at the wavelength 633 nm . . . . . . . . . . . . . . . . . . . . . . 39
3.8 Angular dependences of the linear reflectivity, SH intensity and magnetic
contrast for the bilayer (20 nm)Au/ (30 nm)Co . . . . . . . . . . . . . 40
3.9 Angular dependences of the linear reflectivity, SH intensity and magnetic
contrast for the set of trilayers . . . . . . . . . . . . . . . . . . . . . . . 41
V
3.10 Illustration of nonlinear excitation of SPP by SHG in bilayer (20 nm)Au/
(30 nm)Co employing Kretschmann configuration . . . . . . . . . . . . 42
3.11 The phasor diagram of non-magnetic and magnetic SHG fields in a hy-
brid structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.12 Angular dependence of electric field |E| components at different inter-
faces of the trilayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.13 Illustration for the electric field components in reflection. . . . . . . . . 50
3.14 Forward approach: comparison between experimental and simulated SH
intensity and magnetic contrast . . . . . . . . . . . . . . . . . . . . . . 51
3.15 Reverse approach: fit the newly-defined magnetic contrast φ to estimate
the Lorentzian function . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.16 Dispersion of the Lorentzian function . . . . . . . . . . . . . . . . . . . 54
3.17 Angular dependence of the conventional magnetic contrast ρ at different
wavelengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.18 Angular dependence for the bilayer (20 nm)Au/ (30 nm)Co excited at
the fundamental wavelength 1560 nm . . . . . . . . . . . . . . . . . . . 56
3.19 Angular dependence of the SH intensity and magnetic contrasts of tri-
layer (5 nm)Au/(10 nm)Co/(25 nm)Ag excited at the fundamental wave-
length 1200 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.20 The nonlinear Lorentzian function and SH fields for trilayers with dif-
ferent Ag thicknesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.21 Illustration of a charge at the boundary between two conducting media
Co and Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.22 The trajectory of electrons in the Au and in the Co bulk . . . . . . . . 61
4.1 Illustrations of linear and nonlinear diffraction orders in reflection on a

grating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Schematic for the co-excitation of linear and nonlinear Wood’s anoma-
lies, and the structure of periodic arrays of nickel dimers . . . . . . . . 66
4.3 The top-view of the experimental setup for nonlinear Wood’s anomaly
in the periodic structures. . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.4 Angular dependence of the linear reflectivity, the SH intensity and mag-
netic contrast in the dimer arrays at the fundamental wavelength 820 nm 68
4.5 Normalized linear reflectivity and SH intensity for the dimer arrays at
different wavelengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.6 Dispersion of the Wood’s anomalies excited in the two-dimensional nanoar-
rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 False-color images of the SH magnetic contrast at different grating con-
stants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.1 Angular dependence of the mSHG intensity and the SH magnetic con-
trast in the bilayer and the trilayer at different pump powers . . . . . . 78
5.2 Angular scan of the linear reflectivity for the bilayer and the trilayer at
different pump powers . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.3 A simplified schema for the setup of the magneto-plasmonic auto-correlation 81
VI
5.4 Non-resonant and resonant SH pulses in the Kretschmann configuration 82
5.5 Time-integrated measurements of linear reflectivity, the mSHG outputs
and the SH magnetic contrast for the trilayer (5 nm)Au/ (10 nm)Co/
(15 nm)Ag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.6 Intensity auto-correlation in the layer (5 nm)Au/ (10 nm)Co/ (15 nm)Ag
employing the Kretschmann configuration . . . . . . . . . . . . . . . . 86
5.7 Time-resolved second-harmonic magnetic contrast ρ of the trilayer (5 nm)Au/
(10 nm)Co/ (15 nm)Ag. . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.8 The mSHG cross-correlation on the Kretschmann configuration. . . . . 88
5.9 A simplified scheme for the cross-correlation of mSHG on the Kretschmann
geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.10 Spectrally-resolved linear cross-correlation of a reference wave with the
mSHG signal at different fixed time delays . . . . . . . . . . . . . . . . 90
A.1 Spectra of the fundamental and reference (local oscillator) radiations for
the cross-correlation experiment in Konstanz. . . . . . . . . . . . . . . 98
B.1 Power dependence of the TH output in the trilayer . . . . . . . . . . . 102

B.2 Angular dependence of the TH intensity in the trilayer (5 nm) Au/(10 nm)
Co/(15 nm) Ag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
VII
List of Tables
2.1 Interband transition threshold and plasmon energy of common plas-

monic metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Dielectric tensors, reflectivity matrix and reflected signal for three dif-
ferent MOKE configurations . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Set of samples used in this work . . . . . . . . . . . . . . . . . . . . . . 33

3.2 The non-zero second-order nonlinear susceptibility components for p-
polarized fundamental radiation. . . . . . . . . . . . . . . . . . . . . . . 43
3.3 The roles of different SH susceptibilities in the multilayer structure of
Au/Co/Ag/glass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4 Parameters used for the simulation of mSHG in bilayer Au/Co . . . . . 55
3.5 Dominant SH field components assumed for the mSHG in trilayer Au/
Co/ Ag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.1 Estimation of relative thicknesses between different layers in the hybrid

structure employing the SPP . . . . . . . . . . . . . . . . . . . . . . . . 97
A.2 Techinical specification of Toptica fiber laser for nonlinear magneto-
plasmonic setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
A.3 Technical specifications of the photon counting regime . . . . . . . . . . 98
IX
Contents
Abstract I
List of Figures V
List of Tables IX
Contents XI
1 Introduction and Motivation 1
2 Light matter interaction 5

2.1 Light propagation in matters . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Linear magneto-optical responses . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Drude model and optical properties of metals . . . . . . . . . . 9
2.2.2 Linear magneto-optical effects . . . . . . . . . . . . . . . . . . . 11
2.2.3 Magneto-optical Kerr effect (MOKE) . . . . . . . . . . . . . . . 12
2.3 Nonlinear magneto-optical responses . . . . . . . . . . . . . . . . . . . 14
2.3.1 Model of anharmonic oscillation and SHG . . . . . . . . . . . . 14
2.3.2 Magnetization-induced Second Harmonic Generation . . . . . . 19
2.4 Surface plasmons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.1 Propagating surface plasmon polaritons . . . . . . . . . . . . . . 20
2.4.2 Excitation of SPP . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.3 Simulation of SPPs in multilayer structures . . . . . . . . . . . 27
2.4.4 Magneto-plasmonics . . . . . . . . . . . . . . . . . . . . . . . . 30
3 Nonlinear magneto-plasmonics on Kretschmann configuration 31

3.1 Measurements of Kretschmann magneto-plasmonics . . . . . . . . . . . 31
3.1.1 Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.2 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.3 Light source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Bilayer structure of Au/Co . . . . . . . . . . . . . . . . . . . . . 38
3.2.2 Trilayer structure of Au/Co/Ag . . . . . . . . . . . . . . . . . . 40
3.3 Theoretical description . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Radiative nonlinear surface plasmon polaritons . . . . . . . . . . 43
3.3.2 Calculate SHG magnetic contrast (abcd-model) . . . . . . . . . 45
XI
3.4 Simulation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.4.1 Forward approach . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4.2 Reverse approach . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4.3 Conclusion about two simulation approaches . . . . . . . . . . . 53
3.5 Dispersion of Lorentzian function . . . . . . . . . . . . . . . . . . . . . 53
3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.6.1 Bilayer structures . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.6.2 Trilayer structures . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6.3 SHG due to instantaneous change of damping constant . . . . . 59
3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4 Nonlinear magneto-optics in diffraction anomalies 63

4.1 Nonlinear Wood’s anomaly . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.1 Diffraction on grating . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.2 Wood’s anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.1 Nonlinear Wood’s anomaly and MOKE in nanodimers . . . . . 68
4.3.2 Spectral analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3 Specifying the order of Wood’s anomaly . . . . . . . . . . . . . 71
4.3.4 Impact of Wood’s anomaly on nonlinear MOKE . . . . . . . . . 72
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5 Femtosecond time-resolved nonlinear magneto-plasmonic cross-correlation

measurements 77
5.1 Fluence dependence of the mSHG . . . . . . . . . . . . . . . . . . . . . 77
5.2 Magneto-plasmonic auto-correlation . . . . . . . . . . . . . . . . . . . . 80
5.2.1 Magneto-plasmonics auto-correlation setup . . . . . . . . . . . . 81
5.2.2 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . 85
5.3 Magneto-plasmonic cross-correlation . . . . . . . . . . . . . . . . . . . 88
5.3.1 Magneto-plasmonic cross-correlation setup . . . . . . . . . . . . 88
5.3.2 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . 89
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6 Summary and conclusion 93
Appendix A Technical specifications 97

A.1 Verify film thickness using ATIR . . . . . . . . . . . . . . . . . . . . . . 97
A.2 Technical specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Appendix B Third-harmonic generation in metallic thin film 99

B.1 Estimation of Third-harmonic generation (THG) in Au film . . . . . . 99
B.2 Third-harmonic magneto-plasmonics . . . . . . . . . . . . . . . . . . . 102
Abbreviations 104
XII
List of publications 105
Bibliography 107
XIII
Chapter 1
Introduction and Motivation
Observed for the first time by Wood[1] in 1902, electromagnetic surface waves propagat-
ing on metallic surface, specifically surface plasmon polaritons and Wood’s anomalies,
have been employed in a lot of research topics and applications. They are one of the
main subjects in the field of nano-photonics, where the excitation of these electro-
magnetic surface waves on metallic nanostructure enhances significantly the optical
properties, such as magneto-optics, nonlinear optics, extraordinary transmission, etc.
Surface plasmons (SPs) refer to the oscillations of free electrons at metallic surfaces.
The coupling of an electric field at optical frequencies with the SPs, known as surface
plasmon polaritons (SPP), results in a strong electric field localization in the vicinity
of the metallic surface. SPPs enhance a multitude of optical phenomena in metallic
nanostructures [2], including magneto-optical effects [3], second-harmonic generation
(SHG) [4, 5] and the combination of these two effects, i.e. the magnetization-induced
SHG (mSHG) [6, 7].
Magneto-plasmonics[3] is an emerging concept which attempts to improve the
magneto-optical properties employing SPs and, conversely, to control the SPs using
an external magnetic field. It can be applied to design and improve the performance
of all-optical magnetic data storage, optical switches, label-free sensors [2]. Magneto-
plasmonics structures must couple magnetic properties and SP possible excitation.
Owing to the low optical absorption in the visible spectral range, noble metals (gold,
silver, copper) have been used as efficient media for the excitation of SPs. However,
these metals are non-magnetic and exhibit measurable magneto-optical properties only
under strong external magnetic fields [8–10]. In contrast, ferromagnetic transition
metals (iron, cobalt, nickel) show sizable magneto-optical effects, but have high op-
tical losses. One solution is to combine ferromagnetic and noble metals in a single
hybrid metal/ferromagnet structure. For example, a thin ferromagnetic film sand-
wiched between two noble metals has demonstrated a strong enhancement of both
linear [3, 11–14] and nonlinear [6, 7, 15] magneto-optical effects induced by the prop-
agating surface plasmon polaritons (SPP) in Kretschmann configuration. Apart from
continuous thin films, far-field radiation can be coupled to the SPPs in nanostructured
metallic media, forming a class of magneto-plasmonic crystals. For example, a gold
grating on top of magnetic dielectric [16–18] enables the phase-matched SPP excita-
tion by taking advantages of the periodic corrugations. Another interesting approach
1
CHAPTER 1. INTRODUCTION AND MOTIVATION
makes use of extraordinary optical transmission in subwavelength hole arrays [19–23]

featuring SPP excitation at the surface of a magnetic film [24, 25]. Further, perfo-
rated transition metal layers have shown enhanced magneto-optical effects [26–28] and
SPP-induced phase shifts between the magnetic and the nonmagnetic second harmonic
(SH) fields [29]. Besides, metal/ferromagnet heterostructures have been exploited ex-
tensively in the of ultrafast acoustics [30], exchanged-coupled magnonics [31], spintron-
ics [32, 33] and terahertz emission [34, 35]. Thus, we can envision the consolidation of
plasmonics with the ultrafast effects [7] in a single metallic device.
The mSHG refers to the SH signals that are sensitive to the magnetization. They
are generated from the magnetization-odd second-order nonlinear susceptibility, which
is high in ferromagnetic metals. In most metals, the mSHG in magnetic multilayer
structures occurs exclusively at surfaces and interfaces [36–39], where the spatial in-
version symmetry is broken [40–42]. The measurement of mSHG has been applied to
study magnetic phenomena in hybrid thin films, such as magnetization dynamics [43]
and spin-dependent effects [33].
The combination of the advantages of SPP and mSHG gives rise to the nonlinear
magneto-plasmonics, a sensitive technique to inspect the magnetic interfaces. The exci-
tation of SPP on hybrid thin films of noble-metal (NM)/ ferromagnet (FM) has usually
been performed using conventional Kretschmann configuration due to the setup sim-
plicity. Inherited from the researches in linear regime [3], the trilayer of NM/FM/NM
has usually been employed for the nonlinear magneto-plasmonics [6, 15]. In fact, the
plasmonic enhancement of the mSHG has been observed in a mono-layer of iron [44],
where the coupling to SPP is not supposed to be efficient. To fill in the gap, it is nec-
essary to investigate the bilayer NM/FM as well, even though it has been disregarded
in the field of linear magneto-plasmonics [3]. In this work, both the bilayer NM/FM
and the trilayer NM/FM/NM will be inspected, in order to explain the interplay be-
tween plasmonics and magneto-optics in the nonlinear regime. The experiments will
be interpreted within the framework of simplified theoretical models.
Merging functionalities of plasmonic materials [19] with the optical properties of diffrac-
tion gratings and periodic structures with a sub-wavelength periodicity, i.e. optical
meta-surfaces [45–50], is also an attractive approach. These systems are often discussed
in the context of plasmonics [21–23] and magneto-plasmonics [16–18, 24, 25, 51], if
one of the ingredients displays magneto-optical properties.
In plasmonics, spectrally narrow resonances, commonly referred to as surface lattice
resonances (SLRs) [46, 47, 52, 53], arise from an unusual interplay between the local-
ized surface plasmon resonance (LSPR) and the emerging diffraction orders under the
condition of Wood’s anomaly. The latter describes redistribution of the light energy
between different diffracted beams upon opening of a new diffraction order propagating
along the grating surface [1].
In magneto-plasmonics, the excitation of Wood’s anomalies in two-dimensional ar-
rays of ferromagnetic nanodisks resulted in the enhancement of the magneto-optical
effects [53, 54]. While one can expect a weak LSPR field from a single optically-lossy
nanodisk [55], bringing two disks closer (thus forming a dimer) boosts the nanoantenna
2
effect in the air gap. The MOKE signals show a clear dependence on the dimer size, that
is, distance between the pair of the nickel disks, enabling a ruler at the nanoscale [56].
In these structures, plasmonic effects have been enabled either by a high relative con-
tent of the metal resulting in the excitation of propagating surface plasmon-polaritons
(even despite high optical losses in ferromagnetic transition metals), or LSPRs in the
nanosized metallic objects.
A conventional diffraction grating requires the grating period Λ to be larger than a
half of the optical free space wavelength λ (Λ > λ/2) to ensure the presence of at least
one diffraction order. In contrast, optical meta-surfaces are usually characterized by a
much smaller periodicity Λ ≪ λ [50, 57, 58]. Overcoming this frontier, an intriguing
crossover regime can be explored within the realms of nonlinear optics, where the
wavelength λ of the fundamental radiation is converted to λ/n by means of a nonlinear-
optical process of the order n > 1. In this case, the same periodic structure can serve as
a meta-surface for the fundamental light at λ, and as a regular grating for the outgoing
radiation with the wavelength λ/n. Although the condition for this transition regime
λ ≫ Λ > λ/2n is relatively easy to fulfill in the visible and near-infrared optical range,
the nonlinear-optical properties of such hybrid gratings remain up to date unexplored.
Despite the large amount of research on magneto-photonic crystals in general, the
studies of nonlinear magnetization-induced Wood’s anomalies are very scarce. In par-
ticular, the role and the impact of the Wood’s anomalies in ferromagnetic arrays on
the SH generation remains widely unexplored. To fill this gap, this work studies the
Wood’s anomaly, i.e. the intrinsic property of an optical grating, in the aforementioned
nonlinear transition regime. The angle-dependent SH spectroscopy is performed with
tunable femtosecond laser source on a rectangular array of nickel dimers featuring a
nanoscale gap, similar to that studied in Ref. [46]. Taking advantage of the unequal
periodicity of the array in two orthogonal in-plane directions, one can identify the
frontiers of the novel crossover regime for nonlinear Wood’s anomaly, i.e. when the
structure exhibits grating properties exclusively at the SH wavelength λ/2 but not at
the fundamental wavelength λ. Our experiments represent the first step towards the
application of nonlinear optics for understanding the transition between optical diffrac-
tion gratings and meta-surfaces, extending the concept of Wood’s anomaly [59, 60] to
the nonlinear regime.
This thesis consists of six chapters. Chapter 2 introduces the theoretical basis for the
following chapters. I try to derive the optical phenomena by classical model without
quantum optics. The linear optical response of metals, including the magneto-optical
Kerr effect, will be derived using Drude model. Then the second-harmonic generation
(SHG) will be described by the model of an anharmonic oscillator. The magnetization-
induced SHG will be as well discussed in the section of nonlinear optical response. A
part is also devoted to the SPP, a solution of the Maxwell’s equations, and the meth-
ods to excite it. The formulation of the magneto-plasmonics in the linear regime is
introduced.
Chapter 3 discusses the nonlinear magneto-plasmonics employing the Kretschmann
geometry. The samples, measurements method and preliminary results will be pre-
3
CHAPTER 1. INTRODUCTION AND MOTIVATION
sented in details. Two sets of samples are examined: the trilayer Au/Co/Ag and the
bilayer Au/Co. A simplified theoretical model is then developed, based on which the
mSHG and the SH magnetic contrast in the trilayer and the bilayer can be simulated
in high agreement with the experimental data. The features in the angular dependence
of mSHG and the SH magnetic contrast can be explained by the coupling of the fields
to SPP.
Chapter 4 explores the nonlinear Wood’s anomaly on two-dimensional arrays of
nickel nanodimers. The surface waves in this case is no longer SPP, but the exci-
tation method is similar to that using periodic grating. The chapter will revise the
definition of Wood’s anomaly, a less common term compared to SPP, in the linear and
the nonlinear regime. It will be shown that the excitation of the linear and nonlinear
Wood’s anomalies at different transversal and longitudinal periodicities of the rectan-
gular arrays, which are λ/4 < ΛT < λ/2 and ΛL > λ/2 respectively, demonstrates
experimentally the transition regime speculated for the nonlinear diffraction. The ef-
fects of the nonlinear Wood’s anomaly on the mSHG and the magnetic contrast will
be discussed in details. The possibility to couple the Wood’s anomaly with the LSPR
within the nanodimer gap to form the hybrid SLR will also be considered.
Chapter 5 includes miscellaneous experiments related to the nonlinear magneto-
plasmonics in Kretschmann multilayers. Three main parts in this chapter are (i) the
demagnetization of the SH magnetic contrast at high power, (ii) the influence of SPPs
on the magnetically-controlled auto-correlations and (iii) the spectrally-resolved cross-
correlation of the mSHG induced by SPPs. The thesis is concluded in Chapter 6 with
a brief summary of the work and an overlook.
4
Chapter 2
Light matter interaction
Light has wave-particle dualitiy. But usually the wave properties of light is sufficient
to describe the light-matter interaction. In this chapter, the linear optical responses
of materials, the optical-nonlinear generation and the surface plasmons will be derived
from the classical descriptions of the electromagnetic waves.
2.1 Light propagation in matters

Interaction of electromagnetic waves and matter are described by four Maxwell’s equa-
tions in S.I. units [61, 62]
∂B(r, t)
∇ × E(r, t) = −
∂t
∂D(r, t)
∇ × H(r, t) = + j(r, t) (2.1)
∂t
∇ · D(r, t) = ρ(r, t)
∇ · B(r, t) = 0
where E is the electric field, H is the magnetic field, D is the electric displacement,
B is the magnetic induction, ρ is the charge density and j is the current density. The
bold letters show vector quantities, which depend on position and time, such as E(r, t).
∂ρ
The conservation of charge requires that ∇ · j = − .
∂t
For a monochromatic field, the time-dependent electric field can be presented as

1
E(r, t) = ℜ E(r)e−iωt = E(r)e−iωt + E ∗ (r)eiωt (2.2)

2
where E(r) is the complex spatial component of a harmonic electric field oscillating
in time at a fixed frequency ω. For plane waves, we have E(r) = E0 e±ik·r , where k
√
is the wavevector with |k| = ω/c µε. As a homogeneous solution of the Maxwell’s
equations when ρ = 0 and ∇ · D(r, t) = 0, plane waves also require that k · E0 = 0
and H0 = ωµ1 0 (k × E0 ).
5
CHAPTER 2. LIGHT MATTER INTERACTION
In general, the spectrum of a time-dependent field E(r, t) is defined by the Fourier

transform Z +∞
1
Ê(r, ω) = √ E(r, t)e−iωt dt (2.3)
2π −∞
and the inverse Fourier transform gives time-dependent field from its spectrum
Z +∞
1
E(r, t) = √ Ê(r, ω)eiωt dω (2.4)
2π −∞
Applying the Fourier transform1 , Maxwell’s equations in Eq. 2.1 can be written in
the frequency domain
∇ × Ê(r, ω) = iω B̂(r, ω)
∇ × Ĥ(r, ω) = −iω D̂(r, ω) + ĵ(r, ω)
(2.5)
∇ · D̂(r, ω) = ρ̂(r, ω)
∇ · B̂(r, ω) = 0
Consider a structure consisting of two media, indicated by the indices p and q, where
the material properties in each medium is independent on position r. The subdomains
p and q are homogeneous, but the whole structure, including the boundary pq between
the two media, is inhomogeneous. In this case, instead of solving the wave equations
for an inhomogeneous medium, it is more convenient to find the solutions in each sub-
domain separately, then combine them by boundary conditions. Using separation of
variables, the Helmholtz equations below can be derived from Maxwell’s equations2
and represent for the wave equations in medium p.
∇ρp
∇2 + kp2 Ep = −iωµ0 µp jp +

ǫ0 ǫ p (2.6)
2 2

∇ + kp Hp = ∇ × j p
√
where the wavenumber kp = |kp | = ωc µp ǫp . The Helmholtz equations become homo-
geneous when ρp and jp = 0.
To guarantee the continuity of the electromagnetic field across the boundary pq, it
is necessary to take integral forms of Maxwell’s equations3 , which leads to the following
conditions at the interface [62]
n × (Ep − Eq ) = 0
n × (Hp − Hq ) = K
(2.7)
n · (Dp − Dq ) = σ
n · (Bp − Bq ) = 0
∂ F.T F.T 1
1
R
One can replace the time derivative and integral by a factor: ∂t −−→ iω and dt −−→ iω
2 2
with the identity ∇ × ∇× = −∇ + ∇∇·
3
It means to take surface integral for E(r, t), H(r, t), and volume integral D(r, t), B(r, t) across
the boundary pq.
6
2.1. LIGHT PROPAGATION IN MATTERS
where n is the normal vector at the interface, K is the surface current density, and σ
is the conductivity of the medium. Fresnel reflection and transmission coefficients [61]
are the results of the boundary conditions applied for plane wave incident on a planar
interface. The boundary condition is essential for the description of surface plasmon
field in section 2.4.
In the macroscopic regime, the response of the material to an electromagnetic field

is included in the polarization P and magnetization M , which are the volume density
of electric and magnetic dipole moments, respectively.
D(r, t) = ε0 E(r, t) + P (r, t)
1 (2.8)
H(r, t) = B(r, t) − M (r, t)
µ0
where ε0 = 8.854×10−12 F/m is the permittivity of free space, and µ0 = 4π×10−7 H/m
is the vacuum permeability. These two constants are related by ε0 µ0 = 1/c2 , where
c = 2.998 × 108 m/s is the speed of light. In a conducting medium, the conduction
current density depends on the electric field with j = σE, where σ is the conductivity.
Inserting Eq. 2.8 into Eq. 2.1 gives the wave equations in a medium
1 ∂ 2E ∂ ∂P
∇ × ∇ × E + 2 2 = −µ0 j+ +∇×M
c ∂t ∂t ∂t (2.9)
1 ∂ 2H ∂P ∂M
∇×∇×H + 2 =∇×j+∇× + µ0
c ∂t2 ∂t ∂t
Here E(r, t) is denoted as E for simplicity. This set of equations describes the propaga-
tion of an electromagnetic wave in a medium having polarization P and magnetization
M . Similar to Eq. 2.6, these wave equations become homogeneous in the case of no
free charge (ρ = 0) and non-conducting (j = 0) medium.
In general, P can be expressed as a power series of the electric field E as following
[40] h i
P = ǫ0 χ(1) E + χ(2) E 2 + χ(3) E 3 + . . .
(2.10)
(1) (2) (3)
= P + P + P + ...
where χ(1) is the linear(first-order) susceptibility. In the linear optics, the expansion
stops at the first-order polarization P (1) , which is proportional to the incident electric
field E. In that case, the electric displacement given in Eq. 2.8 is written as
D = ε0 E + P
(2.11)
= ǫ0 1 + χ(1) E = ǫ0 ǫE

where the dielectric function is denoted as ǫ = 1 + χ(1) . When the electric field is
written in the vector form E = Ex , Ey , Ez , the dielectric function and the linear
susceptibility expand into 3-by-3 tensors. Eq. 2.11 can thus be written in the matrix
form     
Dx ǫxx ǫxy ǫxz Ex
Dy  = ǫ0 ǫyx ǫyy ǫyz  Ey  (2.12)
Dz ǫzx ǫzy ǫzz Ez
7
In isotropic homogeneous non-magnetic media, the off-diagonal elements of the ǫ tensor

vanish, and the diagonal components are equal ǫxx = ǫyy = ǫzz = ǫ, making the
dielectric function ǫ a constant. The off-diagonal components become non-zero in
magnetic media, which will be discussed in later section.
Note that the relation between the complex dielectric function ǫ = ǫ′ + iǫ′′ and the
optical refractive index n = n′ + in′′ of the medium is as following
ǫ = n2 = (n′ + in′′ )2
ǫ′ = n′2 − n′′2 (2.13)
ǫ′′ = 2n′ n′′
Back to Eq. 2.10, in nonlinear optics, the power series of P can be extended to
the second-order nonlinear polarization P (2) = ǫ0 χ(2) E 2 , the third-order nonlinear
polarization P (3) = ǫ0 χ(3) E 3 and so on. χ(2) and χ(3) are called the second-order
and third-order nonlinear susceptibility, respectively. It is expected for condensed
matter that the linear susceptibility χ(1) is of the order of unity, χ(2) ∝ χ(1)/Eat and
2
χ(3) ∝ χ(1)/Eat [40], where Eat is the atomic electric field strength4 . As the power of
the electric field increases, the rank of the susceptibility tensor rises accordingly. For
example, χ(1) is a second-rank tensor with 32 = 9 components, χ(2) is a third-rank
tensor with 33 = 27 components. However, in later section, it will be shown that the
symmetry can reduce significantly the number of tensor components.
2.2 Linear magneto-optical responses

As mentioned in the previous section, the dielectric function and susceptibility rep-
resent for the optical responses of materials. Several models have been developed to
characterize these properties, such as the classical Lorentz [63, 64] and Drude [65, 66]
models. The difference between these two models is that Lorentz model concerns the
interaction of bound electrons or lattice with the electromagnetic field, while Drude
model considers only the unbound electrons. Drude model is used to describe the
interaction of light and metals or plasma, where free electrons dictate the optical prop-
erties. Based on this model, the dielectric function ǫ can be derived from the equation
of motion of a free electron driven by an external electromagnetic field. Following this
strategy, the magneto-optical properties of ferromagnetic metals, the nonlinear gener-
ations at metallic interfaces and the plasmonic properties of noble metals can also be
explained, as discussed in the following sections. Drude model can accurately predict
the optical properties of metals in the infrared/terahertz spectral range, where there
is only intraband transition between states within the sp-band and the electrons are
unlocalized. In the visible regime, one has to take into account the interband transi-
tions, where the incident photon has enough energy to excite an electron from occupied
states in d-band to unoccupied states in sp-band, which is higher than Fermi level. For
example, the interband transition threshold5 of gold occurs at around 2.4 eV [5, 67]
4
Eat = 4πǫq0 a2 = 5.14 × 101 1 V/m, where a0 is the Bohr radius of the hydrogen atom.
0
5
Transition threshold is the lowest energy between an occupied state in a lower band and an
unoccupied state in an upper conduction band.
8
2.2. LINEAR MAGNETO-OPTICAL RESPONSES
(or 516 nm, which is the reason for its color in bulk state). In our research, both
the fundamental excitation (0.8 eV) and second harmonic generation (1.6 eV) are far
below interband transition threshold of noble metals (see Table. 2.1). Drude model,
concerning only the low-frequency intraband transition, is thus applicable.
Metal Einter (eV) λinter (nm) ~ωp (eV)

Gold 2.4 516 8.55-8.99
Silver 4.0 310 8.9-9.0
Copper 2.17 571 7.4-7.9
Aluminium 1.55 800 14.7
Table 2.1 – The second and third columns are the interband transition thresholds of
plasmonic metals in term of photon energy Einter (eV) and wavelength λinter (nm)
[5, 67–71]. The fourth column shows plasmon energy ~ωp (eV) of these metals [72–76].
2.2.1 Drude model and optical properties of metals

Imagine that free electron gas is driven by an external electric field E and shifts to one
side of the material. This causes a difference in electron density and thus an induced
electric field to pull the electron sea back to equilibrium. This is a harmonic oscillation
which can be described by solving the equation of motion of an unbound electron in a
time-dependent electric field.
∂ 2r ∂r
m + γm = −qE(t) (2.14)
∂t2 ∂t
where m is the effective mass of electron, r is the electron trajectory, q is the elementary
charge, γ is the damping constant and E(t) is the external electric field6 . Here the
magnetic component of the incident electromagnetic field is disregarded, since its effect
is negligible compared to the electric component.
Driven by the electric field, the free electron travels at Fermi velocity vF until it
collides with a phonon or another electron [77–80]. The time between two collisions
is called scattering time or relaxation time τsc = l/vF , where l is the mean free path
of electron. τsc contributes to Eq. 2.14 through the damping constant γ = 1/τsc .
For plasma, electrons travel without collision, thus there is no scattering γ = 0. For
noble metals, such as gold, silver and copper, the electron mean free path at room
temperature is very long, ≈40-50 nm[81]. The motion of electron is π out-of-phase to
the driving electric field E(t).
The oscillation of an electron gives an electric dipole moment µ(t) = −qr(t). In
macroscopic regime, polarization corresponds to the volume density of dipole moments
P (t) = N µ(t) = −N qr(t) (2.15)

6
This is the difference between the Lorentz and Drude models. In Lorentz model, there is a
restoring force from positively charged lattice site Frestoring = mω02 r(t) acting on the bound electron
with a resonant frequency ω0 .
9
where N is the volume electron density. Combining Eq. 2.14 and 2.15, the equation of
motion can be rewritten to:
∂ 2P ∂P
2
+γ = ǫ0 ωp2 E(t) (2.16)
∂t ∂t
q
q2
where the plasma frequency is introduced ωp = N ǫ0 m
, with ǫ0 the electric permittivity
of free space. Note that ωp is the resonance frequency of the natural oscillation of free
electron gas, so-called plasma, when there is no external driving force7 . The plasma
frequency sets a barrier to incident radiations, known as screening effect. This involves
that electromagnetic waves having frequency shorter than plasma frequency ωp will be
reflected by the electron cloud, only those having higher frequencies can be transmitted
through. The plasma frequency ωp (rad/s) is usually represented via plasmon energy
~ωp (eV), as listed in Table. 2.1 for common plasmonic metals.
In frequency domain, Eq. 2.16 becomes
∂ 2 [P (ω)e−iωt ] ∂[P (ω)e−iωt ]

+ γ = ǫ0 ωp2 E(ω)e−iωt
∂t2 ∂t
−ω 2 P (ω) − iωγP (ω) = ǫ0 ωp2 E(ω)
And finally, we obtain:
ωp2
P (ω) = ǫ0 2
E(ω) = ǫ0 χ(1) E(ω) (2.17)
−ω − iγω
with the frequency-dependent linear susceptibility χ(1) = ωp2 /(−ω 2 − iγω). The dielec-
tric function is thus expressed as
(1)
ωp2 ωp2
ǫ(ω) = 1 + χ =1+ =1+ (2.18)
−ω 2 − iγω GD
where GD = −ω 2 − iγω denotes the Drude oscillation term 8 .

Eq. 2.18 allows to calculate the dielectric function ǫ from the frequency of the inci-
dent field ω. In Fig. 2.1, the dielectric function of silver ǫAg measured by Johnson and
Christy [74] (solid lines) is compared with ǫD , a dielectric function calculated based
on Eq. 2.18 using plasmon energy ~ωp =8.9 eV [72] and scattering time τsc =40 fs [74].
It is seen that the Drude model well estimates the dielectric function of silver in the
infrared spectral range (<1.5 eV). However, the deviation between the model and the
experiment increases in the visible range, especially for the real part of the dielectric
function Re(ǫAg ). The simple Drude model also fails to illustrate the interband transi-
tion in silver, which occurs at 4 eV. Other complex oscillation terms must be included
to simulate the interband transition.
7
Plasma frequency is found by solving the equation of motion of electron without external driving
2
force and scattering: m ∂∂t2x = −qE, where x is the displacement of electron and E = Nǫqx 0
is the
electric field generated by the charged slab N qx.
8
Another difference between Lorentz and Drude models. In the Lorentz model, the oscillation term
becomes ω02 − ω 2 − iγω with the resonant frequency ω0 .
10
50
0
Dielectric function ǫ
-50
-100
-150
Re(ǫD)
-200 Im(ǫD)
Re(ǫAg )
-250
Im(ǫAg )
-300
1 2 3 4 5 6
Energy h̄ω (eV)
Figure 2.1 – Comparison of the dielectric function of silver estimated by Drude model
(dots) and that measured experimentally on a 30 nm silver thin film (solid lines) [74]
2.2.2 Linear magneto-optical effects

Here the Drude model is used to describe linear magneto-optical effects for magnetized
metallic interfaces. While the magnetic field component of the incident electromagnetic
field hardly produces any optical effect, a strong external magnetic field Bext can exert
a Lorentz force on the moving electrons, and hence introduce an additional term to
the equation of motion of a free electron according to the Drude model [82]:
qh dr i d2 r dr
− E(t) + ×B = 2 +γ (2.19)
m dt dt dt
In Cartesian coordinates, this equation extends into:
∂ 2x ∂x q ∂y ∂z q
2
+γ + Bz + By = − Ex e−iωt
∂t ∂t m ∂t ∂t m
∂ 2y ∂y q ∂z ∂x q
+γ + Bx + Bz = − Ey e−iωt (2.20)
∂t 2 ∂t m ∂t ∂t m
∂ 2z ∂z q ∂x ∂y q
+ γ + B y + B x = − Ez e−iωt
∂t2 ∂t m ∂t ∂t m
We look for the solution in the form r(t) = (x0 , y0 , z0 )e−iωt , where x0 = Lx /L0 ,
y0 = Ly /L0 and z0 = Lz /L0 [82]. We have thus:
h ωq 2 i
L0 = GD G2D − (Bx2 + By2 + Bz2 ) (2.21)
m
11
and
q nh 2 ωq 2 i
Lx = GD − Bx Ex
m m
h ωq ωq 2 i
− iGD Bz − Bx By Ey (2.22)
m m
h ωq ωq 2 i o
− iGD By − Bx Bz Ez
m m
where GD is the Drude coefficient defined in section 2.2.1. Ly and Lz can be written
in a similar manner by cyclic permutation.
In a magnetic medium, the off-diagonal components in the linear susceptibility
tensor χ(1) become non-zero and depends on B = (Bx , By , Bz ), with the symmetry
χij (Bx , By , Bz ) = −χji (Bx , By , Bz ) [83]. Thus we have the polarization which can be
written as
    
Px χxx χxy (Bz ) χxz (By ) Ex
Py  = ǫ0 −χxy (Bz ) χyy χyz (Bx )   Ey  (2.23)
Pz −χxz (By ) −χyz (Bx ) χzz Ez
As the polarization is a volume density of electric dipole moments, shown in Eq. 2.15,
the tensor components of the susceptibility χ(1) can be derived from r = (x0 , y0 , z0 ).
For example, when a constant external magnetic field B = (Bx , By , Bz ) is applied, the
three non-zero tensor components in the first row are
N q2 h 2 ωq 2 i
χxx = G − Bx
ǫ0 mL0 D m
N q2 h ωq ωq 2 i
χxy = − iGD Bz − Bx By (2.24)
ǫ0 mL0 m m
N q2 h ωq ωq 2 i
χxz = iGD By − Bz Bx
ǫ0 mL0 m m
The other tensor elements
p can be written in the same manner. We see that besides the
plasma frequency ωp = N q /(ǫ0 m) and the Drude oscillation term GD = −ω − iγω,
2 2
the off-diagonal element χij also depends on the cyclotron frequency ωL = qB/m, which
makes the change of sign (ωL −→ −ωL ) upon the reversal of the external magnetic
field B −→ −B.
2.2.3 Magneto-optical Kerr effect (MOKE)

Eq. 2.24 reveals that by an external magnetic field, the polarization or intensity of light
can be modified. For light reflected on a magnetized surface, the phenomenon refer
to the magneto-optical Kerr effect (MOKE) [84]. The similar magneto-optical effect
observed in transmission is called Faraday effect [85].
In general, there are three configurations of MOKE, according to the mutual orien-
tations of the magnetization M , surface and the plane of incidence. Fig. 2.2 shows the
magneto-optical configurations in Cartesian coordinates: (a) polar MOKE, where Mz
is normal to the magnetized surface (b) longitudinal MOKE, where the in-plane Mx
12
y Mz Mx My
(a) (b) (c)
Figure 2.2 – Three configurations of magneto-optical Kerr effects: (a) polar, (b) longi-
tudinal and (c) transversal. Plane of incidence is the xz plane.
is parallel to the plane of incidence, and (c) transversal MOKE, where My is perpen-
dicular to the plane of incidence. A table for the dielectric tensors ǫ(ω) in each case
is presented in Tab. 2.2. The non-magnetic diagonal component ǫ remains regardless
of the magnetization M . The magnetic off-diagonal components are written as the
first approximation with respect to M , for example ǫxy Mz . In the polar and longitu-
dinal MOKE, the polarization changes and is quantified by the complex Kerr rotation
Φ = Φ′ + iΦ′′ , where Φ′ is the real part of Kerr rotation and Φ′′ is the ellipticity. The
Kerr rotation angle in the polar configuration is about one order of magnitude larger
than that in the longitudinal geometry [86]. The notation of the reflectivity, rps for
example, stands for the reflectivity of a s-polarized light which initially has incident
polarization p. In the polar configuration, rsp = rps . In contrast, rsp = −rps in the
longitudinal MOKE [87]. In the transversal geometry, only rpp depends on M , thus
gives the magnetization-induced change in the intensity of reflectivity. Besides these
three MOKE geometries, hybrid cases occur when M is tilted. In such cases, both the
polarization and intensity of reflected light could change upon the reversal of M .
Polar Longitudinal Transversal

     
ǫ ǫxy Mz 0 ǫ 0 0 ǫ 0 ǫxz My
ǫ(ω) −ǫxy Mz ǫ 0 0 ǫ ǫyz Mx   0 ǫ 0 
0 0 ǫ 0 −ǫyz Mx ǫ −ǫxz My 0 ǫ

rpp rps (Mz ) rpp rps (Mx ) rpp (My ) 0
Mr
rsp (Mz ) rss rsp (Mx ) rss 0 rss
rps (Mz ) rps (Mx ) ∆Rpp (My )

Signal Φ= rpp
Φ= rpp Rpp
Table 2.2 – The dielectric tensors ǫ(ω), reflectivity matrix Mr and reflected signal for
three different configurations of MOKE. The off-diagonal components of the dielectric
tensor ǫ(ω) depend linearly on magnetization M. The table is adapted from Ref. [3]
13
2.3 Nonlinear magneto-optical responses

The previous section has shown that the motion of electrons driven by an external
electromagnetic field determine the optical responses of the materials. Following this
approach, the nonlinear optical responses can be attributed to a nonlinear motion of
electrons. It is analog to an anharmonic oscillation of a swing or a pendulum, where
strong driving force results in nonlinearity.
2.3.1 Model of anharmonic oscillation and SHG

Perturbation of potential
To illustrate for the anharmonic oscillation, let us start with a simple example of a
spring oscillating along z direction. The restoring force in this case is Frestoring = −Kz,
where K is the spring constant. The harmonic binding potential is thus:
1 1
Z
V (z) = − Frestoring dz = Kz 2 = mω02 z 2 (2.25)
2 2
8
2
V=z
7 2 3 4 5 6
V=z +az +bz +cz +dz
6
Potential V(z)
0
−1 −0.5 0 0.5 1
Distance z
Figure 2.3 – Illustrations of the harmonic potential (red dashed line) and the power
expansion of the potential (blue solid line). For large positive z, the blue line deviates
from the parabolic curve and approaches the case of a potential wall.
When the oscillation encounters a potential wall, V (z) deviates from the parabolic
shape and can be approximated by the power series
1 1 1
V (z) = mω02 z 2 + maz 3 + mbz 4 + . . . (2.26)
2 3 4
where a and b are the perturbation factors. Fig. 2.3 illustrates the parabolic potential
V (z) ∝ z 2 of a harmonic oscillation, and the potential of an anharmonic oscillation
14
2.3. NONLINEAR MAGNETO-OPTICAL RESPONSES
approximated by power expansion. The force in this case is

∂V (z)
F (z) = − = −mω02 z − maz 2 − mbz 3 − . . . (2.27)
∂z
We can note that the inversion center in centrosymmetric crystals[40] causes the
spatial symmetry in potential V (z) = V (−z). Eq. 2.26 and 2.27 show that the sym-
metry of potential happens only if
1 1
V (z) = mω02 z 2 + mbz 4 + . . .
2 4 (2.28)
∂V (z)
and F (z) = − = −mω02 z − mbz 3
∂z
meaning that the even-order anharmonic oscillations (z 2 , z 4 , etc.) vanish. In contrast,
non-centrosymmetric structures permit both even and odd harmonics of the funda-
mental frequency.
Anharmonic oscillation of a free electron

The motion of a free electron, driven by an external electric field and stopped by a
surface (potential wall), is written in Eq. 2.29 below [88]. The charge is free, so there is
no resonant frequency ω0 and the restoring force in Eq.2.27 is reduced to perturbation
force Fperturbation = −amz 2 . Thus the motion equation of a free electron gives:
∂ 2z
m = Fdriving + Fdamping + Fperturbation
∂t2 (2.29)
∂z 2
= −qE − γm − amz
∂t
where γ is the damping constant, an inversion of scattering time τsc , and a the
perturbation constant. Thus, Eq. 2.29 can be rewritten as:
q ∂ 2z ∂z
− E = 2 +γ + az 2 (2.30)
m ∂t ∂t
According to Euler’s formula, a monochromatic time-dependent oscillation zω (t) =
z(ω) cos(−ωt) with real z(ω) is written as:
1 1
zω (t) = z(ω)e−iωt + z(ω)eiωt
2 2 (2.31)
1
= z(ω)e−iωt + c.c.
2
where c.c. denotes the complex conjugate. For the case of anharmonic oscillation, what
we look for is a solution z(t) which is a linear combination of oscillations at fundamental
frequency ω and its harmonics 2ω, 3ω, etc. Thus we can write the series of z(t) as
z(t) = Λzω (t) + Λ2 z2ω (t)

h1 i h1 i (2.32)
z(t) = Λ z(ω)e−iωt + c.c. + Λ2 z(2ω)e−i2ωt + c.c.
2 2
15
where Λ is a scaling factor applied on the external electric field E(t) = ΛEω (t). The
scaling factor helps to separate the terms related to the first-order polarization (P (ω) ∝
ΛE(ω)), the second-order polarization (P (2ω) ∝ Λ2 E(ω)), the third-order polarization
(P (3ω) ∝ Λ3 E(ω)) and so on. Putting Eq. 2.32 into Eq. 2.30 gives:
q ∂ 2 zω (t) ∂zω (t)

− ΛEω (t) = Λ 2
+ Λγ
m ∂t ∂t
2
∂ z 2ω (t) 2 ∂z2ω (t) (2.33)
+ Λ2 2
+ Λ γ + aΛ2 zω2 (t)
∂t ∂t
+ Λ3 ... + Λ4 ...
In the next part, we will separate the terms scaling with Λ and Λ2 , which respectively
correspond to the linear and the second-order oscillations.
Scaling with Λ - Linear harmonic oscillation

Based on Eq. 2.33, the following terms are scaled with factor Λ:
q ∂ 2 zω (t) ∂zω (t)

−
Eω (t) = 2
+γ
m ∂t ∂t
similar to the equation of motion Eq. 2.14 in section 2.2.1. In the frequency domain,
this equation becomes:
q
− E(ω) = −ω 2 z(ω) − iωγz(ω)
m (2.34)
= z(ω)[−ω 2 − iγω]
and finally we have the frequency-dependent solution:
q E(ω)
z(ω) = − × (2.35)
m −ω 2 − iγω
From Eq. 2.35 and Eq. 2.15, we can deduce the relation between the linear polar-
ization P (ω) and the oscillation of electron z(ω):
q2 E(ω)
P (ω) = −N qz(ω) = N × (2.36)
m −ω 2 − iγω
where N is the volume density of electrons. Eq. 2.36 presents the same linear polariza-
tion P (ω) as in Eq. 2.17. Recall that the relation between P (ω) and E(ω) is expressed
ωp2
by the Lorentzian complex function of linear optical susceptibility χ(1) = −ω2 −iγω =
ωp2
G(ω)
, where GD (ω) is the frequency-dependent Drude oscillation term.
Scaling with Λ2 - Second-order anharmonic oscillation

The terms scaled with factor Λ2 in Eq. 2.33 include:
∂ 2 z2ω (t) ∂z2ω (t)
− azω2 (t) = 2
+γ (2.37)
∂t ∂t
16
where the term zω2 (t) can be derived from Eq. 2.31
1 1 1
zω2 (t) = z 2 (ω)ei2ωt + z 2 (ω)e−i2ωt + z 2 (ω) (2.38)
4 4 2
For the oscillation at the double-frequency 2ω, we consider only the part with
e−i2ωt
. The other components correspond to other second-order nonlinear processes;
for example, the DC term 1/2z 2 (ω) is related to the rectification effect. Thus, we put
z2ω (t) = 12 z(2ω)e−i2ωt and zω2 (t) ≈ 41 z 2 (ω)e−i2ωt into Eq. 2.37 to have:
a
− z 2 (ω) = −2ω 2 z(2ω) − iγωz(2ω) (2.39)
4
With z(ω) from the linear oscillation in Eq. 2.35, Eq. 2.39 becomes:
a a q2 E 2 (ω)
− z 2 (ω) = − × 2 ×
4 4 m (−ω 2 − iγω)2 (2.40)
1 1
= (−4ω 2 − iγ2ω)z(2ω) = G(2ω)z(2ω)
2 2
where the Drude term at the double frequency is GD (2ω) = −4ω 2 −iγ2ω. The solution
for the second-order oscillation z(2ω) is thus given by
a q2 E 2 (ω)
z(2ω) = − × 2 × 2 (2.41)
2 m GD (ω)GD (2ω)
In the macroscopic regime, the Taylor expansion of the polarization is:
P = ΛP (ω) + Λ2 P (2ω) + . . . (2.42)
The second-order nonlinear polarization P (2ω) relates to the dipole moment at the
double frequency 2ω and is written as:
P (2ω) = −N qz(2ω) = ǫ0 χ(2) E 2 (ω) (2.43)
where the second-order nonlinear susceptibility χ(2) is expressed by
a q ωp2
χ(2) = × × 2 (2.44)
2 m GD (ω)GD (2ω)
Fig. 2.4 illustrates the trajectory of a free electron oscillating anharmonically with
the external electric field. One can see that the second-order harmonics at double
frequency z(2ω) (Eq. 2.41) is smaller than the oscillation at the fundamental frequency
z(ω) (Eq. 2.35) by a factor of 2mGDaqE(ω)
(ω)GD (2ω)
. Thus the nonlinear process is eligible only
when the perturbation factor a and the driving field E(ω) are relatively large. That
is why ultrashort laser pulses with high peak power are required in nonlinear optical
generation processes.
17
10
Re [Λ zω (t)]
Re[Λ 2 z2ω (t)]

5 Re[z(t)]
z (arb. u.)
-5
-10
0 2 4 6 8 10
Time (fs)
Figure 2.4 – Illustration for the anharmonic oscillation of a free electron. The trajectory
of the electron has the form z(t) = Λzω (t) + Λ2 z2ω (t)
SHG at the surface and interfaces

According to the electric dipole approximation above, second-order nonlinear optical
processes (second harmonic generation - SHG, different harmonic generation - DFG,
sum frequency generation - SFG) are prohibited in structures having center of inversion.
Many metals have centrosymmetric cubic crystal structures. For example, the crystal
structure of Au, Ag, Al, Cu, Ni is face-centered cubic (fcc), while Fe crystal is body-
centered cubic (bcc) at room temperature. In these structures, SHG is supposed to
be allowed exclusively where the inversion of spatial symmetry is broken, i.e. at the
metallic surface and interfaces. However, besides the surface-limited SHG, there is a
relatively weak SH emission occurring within the metallic bulks, which can only be
explained by quadrupole mechanism[89–91]. In general, the SH polarization P (2ω) in
a centrosymmetric medium includes both the dipole surface term and the quadrupole
bulk term [29, 92]
P (2ω) = Psurf (2ω) + Pbulk (2ω)

(2) (2) (2.45)
= χsurf E(ω)E(ω) + χbulk E(ω)∇E(ω)
(2) (2)
where χsurf and χbulk correspond to the surface and bulk second-order susceptibilities.
The surface contribution Psurf (2ω) occurs exclusively at the surface and interfaces of
metal. In contrast, the nonlocal bulk source of SHG Pbulk (2ω) is located within the
metallic medium and proportional to the gradient of fundamental field ∇E(ω). The
surface and bulk contributions depend on the shape of structures and the field distri-
18
bution. In thin films of gold, the surface contribution dominates over the bulk term
[93]. In cobalt perforated layers, the strong field gradient caused by the surface plas-
mon polariton results in a significant bulk contribution in the magnetization-induced
SHG [29]. Throughout this work, the dipole-allowed Psurf (2ω) is considered the main
source of SH emission.
2.3.2 Magnetization-induced Second Harmonic Generation

In space, the second-harmonic polarization is
(2)
Pi (2ω) = ǫ0 χijk Ej (ω)Ek (ω) (2.46)
(2) (2)
where i, j, k denote Cartesian coordinates. Since χijk = χikj due to the permutation
symmetry, the susceptibility tensor in this case is a 3 × 6 tensors [86]:
 
Ex Ex
     Ey Ey 
Px χxxx χxyy χxzz χxzy χxzx χxxy   Ez Ez 

Py  = ǫ0 χyxx χyyy χyzz χyzy χyzx χyxy  
2Ey Ez 
 (2.47)
Pz χzxx χzyy χzzz χzzy χzzx χzxy   
2Ex Ez 
2Ex Ey
Due to the geometry, among 18 components of χ(2) tensor, there are 13 components
that are odd to magnetization, i.e. χodd (−M ) = −χodd (M ). The asymmetry leads to
magnetization-induced change in the second-order polarization P (2ω). The odd com-
ponents χodd depend on M [37, 38], and they vanish if the structures are non-magnetic.
The other χ(2) components are even to magnetization, χeven (M ) = χeven (−M ), and
indifferent to the reversal of magnetization vector M . The magnetic (odd) and non-
magnetic (even) components of χ(2) are
 
χxxx (My ) χxyy (My ) χxzz (My ) χxzy (Mz ) χxzx χxxy (Mx )
χyxx (Mx ) χyyy (Mx ) χyzz (Mx ) χyzy χyzx (Mz ) χyxy (My ) (2.48)
χzxx χzyy χzzz χzzy (Mx ) χzzx (My ) χzxy (Mz )
The SH output from a magnetized interface is the interference of non-magnetic and
mag
magnetic SH fields, denoted as E2ω
nm
and E2ω , respectively.
nm mag 2
I2ω (±M ) = E2ω ± E2ω
nm 2 mag 2 nm mag (2.49)
= |E2ω | + |E2ω | ± 2|E2ω · E2ω | cos ϕ
mag
where ϕ is the phase difference between the complex E2ω nm
and E2ω , as shown in the
phasor diagram in Fig. 2.5. To measure the influence of the magnetization on the
nonlinear emission, a nonlinear magnetic contrast [6, 7, 15] is defined as
I2ω (+M ) − I2ω (−M )
ρ=
I2ω (+M ) + I2ω (−M )
nm mag (2.50)
2|E2ω · E2ω | cos ϕ
= nm 2 mag 2
|E2ω | + |E2ω |
19
The formulations of I2ω (±M ) and ρ above show how the far-field magnetization-
induced SH signals depends on the phase ϕ. The SH magnetic contrast ρ becomes
mag
null when the phase difference ϕ = π/2, i.e. when E2ω
nm
and E2ω are orthogonal on
the phase diagram.
even
Enm2 ∝ EE
E2 (+M)
E2 (-M)
odd
E2mag∝ EE
-Emag
Figure 2.5 – Phasor diagram of the non-magnetic and magnetic SH fields, denoted as
mag
nm
E2ω and E2ω respectively, and their interference E2ω (±M ).
2.4 Surface plasmons

As we have seen in sections 2.2 and 2.3, the oscillation of free electrons driven by an
electromagnetic field results in optical properties in metals. After having described
magneto-optic properties, we now concentrate in surface plasmon excitation, and its
coupling with electromagnetic surface waves, namely SPP9 . In this section, the SPP
field will be described as a solution of Maxwell’s equations at the interface of metal
and dielectric.
2.4.1 Propagating surface plasmon polaritons

Let us consider a system consisting of two semi-infinite media: metal (ǫm = ǫ′m + iǫ′′m )
and dielectric (ǫd real) in Fig. 2.6. SPP field is an electromagnetic field localized at the
interface of these media, thus it follows Maxwell’s equations (Eq. 2.1) and the material
responses in (Eq. 2.8). When ρ = 0, we have the following homogeneous equations
∂Hj
∇ × Ej = −µj
∂t
∂Ej
∇ × Hj = ǫj (2.51)
∂t
∇ · (ǫj Ej ) = 0
∇ · Hj = 0
9
The surface plasmon is different from the bulk plasmon, which is a longitudinal oscillation and
can not be coupled to the transversal optical electromagnetic wave.
20
2.4. SURFACE PLASMONS
It is considered that µj = 1 in the optical spectral range.
If the incident wave is p-polarized (TM), the field components in material j 10 is

defined by:
Ej = Ej,x , 0, Ej,z ei(kj,x x+kj,z z−ωt)

(2.52)
Hj = 0, Hj,y , 0 ei(kj,x x+kj,z z−ωt)

2
where kj,x
2 2
+ kj,z = kj2 = ǫj k02 = ǫj ωc .
ϵd Ed,z
z
Ed,x
x ϵd
ϵm
y Em,x
ϵm Em,z
Figure 2.6 – The boundary conditions in the case of p-polarized electromagnetic fields
at the interface between metal (ǫm ) and dielectric (ǫd ).
As there is no source in either medium, we consider the surface current density

K = 0 and the conductivity σ = 0. The boundary conditions in Eq. 2.7 become:
Em,x = Ed,x
Hm,y = Hd,y
(2.53)
ǫm Em,z = ǫd Ed,z
km,x = kd,x = kx
It is seen that while the tangential components Ex and Hy are continuous across the
interface, the normal component Ez experiences a jump by a ratio of the two dielectric
functions.
For plane waves, kj · Ej = 0, or we have
kx Em,x + km,z Em,z = 0

(2.54)
kx Ed,x + kd,z Ed,z = 0
Along with 2.53, we have

kd,z km,z
= (2.55)
ǫd ǫm
10
The subscription j indicates the medium, either m if metal, or d if dielectric.
21
50
0
Dielectric function ǫ
-50
-100 Re(ǫAg )
Im(ǫAg )
-150
Re(ǫCo )
-200 Im(ǫCo )
Re(ǫAu )
-250
Im(ǫAu )
-300
500 1000 1500 2000 2500
Wavelength (nm)
Figure 2.7 – The real and imaginary parts of the dielectric functions of silver (blue),
gold (green) [74] and cobalt (red) [94], the three metals used extensively in this work.
By definition, we have
kx2 + km,z
2
= ǫm k02
kx2 + kd,z
2
= ǫd k02
Subtracting the two equations above and applying the relation in Eq. 2.55, we have
the dispersion relations for kz
ω ǫm
km,z = ·√
c ǫm + ǫd
ω ǫd (2.56)
kd,z = ·√
c ǫm + ǫd
and for kx r
ω ǫm ǫd
kx = (2.57)
c ǫm + ǫd
Since ǫm = ǫ′m + iǫ′′m is complex, we have the complex wavevector components
kx = kx′ + ikx′′ and kz = kz′ + ikz′′ . Fig. 2.7 shows the dielectric functions of silver, gold
and cobalt, the three metals used extensively in this work. The dielectric function
of noble metals [74], such as gold and silver, have large negative real part ǫ′m and
low positive imaginary part ǫ′′m in the visible and near-infrared spectral range, making
kx > ω/c and km,z , kd,z . As a metal, cobalt also has negative real part of dielectric
function [94], but the large imaginary part makes it lossy in the optical region.
One can describe the behaviour of the SPP field from the SPP wavevector compo-
nents in Eq. 2.56 and 2.57. The SPP field propagates along the interface of metal/dielectric
with inplane wavevector kspp = kx = nsp k0 , where nsp is the effective refractive in-
22
dex of SPP. The SPP wavelength is λspp = 2π/kx′ , and the propagation length11 is
Lspp = 1/kx′′ . Along the normal direction, SPP field decays into both media by a factor
exp −|kj,z |z , leading to a “bound” mode, where the field is strongly concentrated
near the interface. The 1/e decay length of the field into metal, called “skin depth”, is
δskin = 1/|km,z |, and the decay length into the dielectric medium is δd = 1/|kd,z |. It is
noteworthy that the penetration depth of power into metals is a half of the skin depth
δp = δskin /2 = 1/2|km,z |. Inserting these notations to Eq. 2.52, we can write out the
SPP fields in metal
kx ′
− x − z
Em = E0 , 0, − E0 eikx x−iωt e Lspp δskin
km,z
ω ǫ (2.58)
− x − z
′
m
Hm = 0, · E0 , 0 eikx x−iωt e Lspp δskin
c km,z
Similarly, the SPP fields in dielectric can be written replacing ǫm by ǫd , km,z by kd,z ,
and δskin by δd . Fig. 2.8 shows the vector field of SPP at gold/air interface at the
wavelength 1560 nm.
−6 −8
x 10 x 10
5 10
Normal to the surface z (m)
Normal to the surface z (m)
5
3
2
0
1
0
−5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Along the surface x (m) −6 Along the surface x (m) −6
x 10 x 10
(a) (b)
Figure 2.8 – The vector field of SPP localized at gold/air interface (z = 0) at the
wavelength 1560 nm. The red vectors illustrate SPP field in the air. The blue vectors
are SPP in gold. (a) The extension of SPP in both media. The skin depth for electric
field is twice higher than the penetration depth of intensity. (b) A zoom-in to the
interface shows the flow of electric field.
If the incident electromagnetic wave is s-polarized (TE), Ey is continuous across the

boundary, and we can no longer derive the dispersion relation in Eq. 2.53. Equivalent
conditions can be achieved for Hx and Hz only if one of the material has negative real
part of µ.
23
Figure 2.9 – Dispersion relations of kspp at silver/air interface (red solid line), nkspp at
silver/substrate interface (red dashed line), light line k0 in the air (green solid line),
light line nk0 in the substrate (green dashed line) and the inplane wavevector kx at the
angle of incidence θ = 47 deg (blue line). Magenta lines mark the coupling between
the incident kx and the kspp at silver/air interface.
2.4.2 Excitation of SPP

The dispersion relation in Eq. 2.53 is plotted in Fig. 2.9. To the right of the light
√
line nd k0 = ǫd k0 , which is simply k0 in the air (green solid line), it is the region
for non-radiative mode. To the left of the light line, the mode is radiative (leaky) [2,
95].
p The dispersion relation of SPP is presented by the red solid curve of kspp =
ǫm ǫd /(ǫm + ǫd )k0 . We note a situation when ǫm + ǫd −→ 0 in Eq. 2.57, kspp = kx
becomes very large. In this case, the replacement of ǫm ≈ √ ǫ′m ≈ −ǫd into ω 2 =
ωp2 /(1 − ǫm ) in the Drude model12 will give ω −→ ωsp = ωp / 1 + ǫd . ωsp is called
surface plasma frequency, and is recognized by a bend backwards of kspp (red solid
line) at ~ωsp = 3.4 eV. ωsp shows the limit of the wavevector kspp achievable in an
experiment13 . Beyond this value, the mode is strongly localized inside metal and the
loss is high. When the frequency increases further, reaching the plasma frequency
ωp , the mode is called Brewster mode. It is no longer a surface mode [62, 96] and
the frequency is not in the optical range. Thus, we will limit the discussion to the
low-energy branch of the kspp .
Within the range of low frequency ω < ωsp , kspp is always larger than the light line
k0 , thus it is impossible to directly couple the incident light to the surface plasmons.
In order to excite surface plasmon by light, it is essential to compensate the difference
between the light line and the kspp . Here I discuss two main approaches that will be
applied in our measurements, prism coupling and grating coupling.
11
The propagation length for the intensity (or power) is 1/2kx′′ .
12
Eq. 2.18 when the damping constant γ = 0
13
In reality, the interband√ transition and non-zero damping constant γ 6= 0 bring down the asymp-
tote of kspp , i.e. ωsp < ωp / 1 + ǫd
24
Prism coupling
Since a prism (or substrate)phas dielectric constant ǫs larger than ǫd in Eq. 2.53, the
√
light line will be tilted by ǫs /ǫd , or simply n = ǫs if ǫd = 1. In Fig. 2.9, the
tilted light line nk0 is marked by the green dashed line. The grey region between the
light line in the air k0 and the light line in prism nk0 shows the range of available
kspp . At a certain angle of incidence θ, the inplane component of incident wavevector
kx = nk0 sin θ (blue solid line) can match with the SPP wavevector kspp , giving us
kx = kspp
(2.59)
r
ǫm ǫd
n sin θ =
ǫm + ǫd
Based on this principle, several geometries have been deployed, such as Kretschmann
configuration [97], Otto configuration [98], near-field objective (modified Kretschmann
configuration) [2, 62, 99], as shown in Fig. 2.10. Regardless of the incident medium,
the inplane wavevector kx is preserved in these geometries, and only matched to the
wavevector of the SPP field localized at the interface of metal/open-space (air) as in
Eq. 2.59. It is impossible to couple to the SPP wave at metal/substrate interface be-
cause its wavevector nkspp (red dashed line) is always to the right of the light line in the
substrate nk0 (green dashed line), as shown in Fig. 2.9. The limit of these geometries
is the distance between the incident surface, where the incident light is reflected, and
the surface where SPP is excited. In both conventional and modified Kretschmann
configurations, this distance is the thickness of the metallic film, which allows efficient
contact between the evanescent fields of the incident light and the SPP within the
metal. Thus Kretschmann geometries cannot excite SPP in thick metallic films. The
optimal thickness of the film is around 50 nm for the case of gold and silver [62, 95].
In contrast, in Otto arrangement, the limiting distance is the open-space gap between
the prism and the metal film, which is within a few µm [62]. Otto configuration can
be used to excite SPP at the surface of thick metallic films, but it requires precise
mechanical adjustments.
The conventional Kretschmann-Raether geometry is used extensively in this work. In

a general structure comprising three different media dielectric/metal/air with a suf-
ficiently thin metallic layer, there are SP fields at both the metal/dieletric and the
metal/air interfaces. The hybridization of these two SP fields gives rise to four SP
modes [100, 101]. In the particular case of Kretschmann configuration, two symmetric
modes [14, 101] are accessible within the metallic film, a bound mode and a leaky
mode. The bound mode is non-radiative, meaning that the energy from substrate and
air flows into the metallic film and gets absorbed. In contrast, the leaky mode can ra-
diate towards the transparent substrate, causing radiative loss. Inspecting the angular
dependence of the light going through the prism and reflected on the studied metallic
layer, a drop of reflectivity is observed near the angle of total internal reflection14 . This
14
The total internal reflection angle here is meant for the transition from the dielectric to air,
regardless of the metallic layer.
25
kspp
kspp
kspp kx
kx θ
θ kx
θ
(a) (b) (c)
Figure 2.10 – Different geometries where SPP is excited by prism coupling: (a) con-
ventional Kretschmann configuration (b) Otto configuration (c) modified Krestchmann
configuration.
configuration is also called “attenuated total internal reflection" (ATIR) measurement.

The dip in the angular spectrum of reflectivity is the result of the destructive inter-
ference of the aforementioned leaky (radiative) mode and the reflected light from the
metal[2, 95].
When a thin film made of plasmonic metals is irradiated at the infrared spectral
range, the minimum in the angular dependence of reflectivity is usually utilized as an
indicator for the excitation of SPP field at the interface of metal/air. However, in
cases of high optical loss, such as non-noble metals at high optical frequencies, the dip
of reflectivity broadens and no longer signifies the SPP field enhancement [101, 102].
If one puts a near-field detector close to the open-surface (air side) of the metallic
film and a far-field detector at the output of the prism, the angle where the near-
field detector sees the highest SPP intensity is different from the angle where the
far-field detector observes the minimal reflected light. This phenomenon can be seen
using electromagnetic field simulations and has been proven experimentally [102]. It is
explained by the discrepancy between the absorptive mode and SPP mode when the
dispersion is significant.
A scan over the angles of incidence, i.e. the inplane wavevector kx , helps to de-
termine the phase-matching conditions of not only the linear SPP (as in Eq. 2.59)
but also the nonlinear SPPs [6, 7, 15, 99, 103]. In the visible/infrared spectral range,
the excitations of these SPPs occur at different angles because of the dispersion of
the substrate. Thus, the SPPs excited in different regimes become distinguishable.
This will be discussed in more details in chapter 3. Besides the applications in nano-
photonics [3, 7, 30], the Kretschamnn geometry has also been employed for label-free
sensing [104, 105], in which a minus change of refractive index of the environment
above the metallic interface nd is detected via a shift of kx , namely the shift in either
the angle of incidence or the incident wavelength.
26
Diffraction grating
Using a diffraction grating, we bridge the gap between the inplane wavevector by a
multiple m of grating wavevector G = 2π/ΛG , where ΛG is the grating periodicity.
Thus we have
kx ± mG = kspp
(2.60)
r
2π ǫm ǫd
k0 n sin θm ± m = k0
ΛG ǫm + ǫd
with m is an integer. In this case, there can be several m and corresponding angles
θm , where the k-matching condition is satisfied. An illustration of the SPP excitation
using grating coupler is shown in Fig. 2.11. Similar principle can be applied to excite
Wood’s anomaly in periodic structures, which will be discussed in chapter 4. Unlike
the prism coupling, grating coupling allows optical excitation of SPP from the air,
which avoids the dispersion of the substrate. On the other hand, the fabrication of
metallic gratings is more complicated than the continuous thin film.
θ kspp
kx G
Figure 2.11 – Excitation of SPP by a diffraction grating.
2.4.3 Simulation of SPPs in multilayer structures

Matrix-Transfer-Method
Matrix-Transfer-Method (MTM) is a numerical method to solve Maxwell’s equations
for one-dimensional structures [61, 106–108], from which the optical responses (modes,
reflectance, transmittance) are derived. In case of multilayer structures, the dielectric
function ǫ(z) varies along z, normal to the film surface, and remains constant along x
and y. The multilayer thin films are thus considered as 1D optical devices and can be
well simulated using the MTM.
For 1D structure along z, kx and ky are continuous, while kz is not. The transfer
matrix Mj of each individual layer j describes the propagation of the (forward and
backward) modes through the thickness Lj , taking into account the boundary condi-
tions. In general, Mj is a 4 × 4 matrix; but for isotropic non-magnetic medium, it can
be reduced to a 2 × 2 matrix as below.

E(z + Lj ) E(z)
= Mj (2.61)
H(z + Lj ) H(z)
27
where " ǫ k #
cos(kzj Lj ) −i kjz 0 sin(kzj Lj )
Mj = kz
j
(2.62)
−i ǫj kj0 sin(kzj Lj ) cos(kzj Lj )
The global transfer matrix for a system consisting of Nz layers is:
Nz
Y
Mglobal = Mj
j=1
and formulated as
M11 M12
Mglobal =
M21 M22
The amplitude transmittance t and reflectance r coefficients are then:
1
t = 2kR
(M11 + M12 kT )kR + M21 + M22 kT
(2.63)
M11 + M12 kT )kR − M21 + M22 kT
r=(
(M11 + M12 kT )kR + M21 + M22 kT
where kR = ǫkrr,z
k0
and kT = ǫktt,z
k0
, with kr,z and kt,z respectively the z-coordinated wave
vectors in reflection region (dielectric constant ǫr ) and transmission region (dielectric
constant ǫt ). For a structure consisting of a single metallic layer, MTM describes
Fabry-Pérot interferometer[109]. Regarding Eq. 2.61 and the boundary conditions in
Eq. 2.53 that Ex , Hy and ǫEz are continuous across the interface, the fields at the
boundary between any two layers can be calculated.
Reflectivity at 1250 nm
1
R mtm
0.9 R exp
0.8
0.7
Reflectivity
0.6
0.5
0.4
0.3
0.2
35 40 45 50 55
Internal angle of incidence (deg)
Figure 2.12 – Comparison between the experimental angular dependence of linear re-
flectivity (blue solid) and the one simulated using MTM (red solid). The structure is
(4nm)Au/(10nm)Co/(15nm)Ag.
28
Simulating the linear reflectivity using MTM

With the MTM described above, the reflectivity R = |r|2 at each angle of incidence
can be obtained since kz depends on the angle. If the refractive index and the thickness
of each individual layer are known, the simulated angular dependence of reflectivity
should be consistent with the experimental data. Employing the known database for
the refractive indices of gold, silver [74] and cobalt [94], one can fit the experimental
and MTM-simulated linear reflectivity, as shown in Fig. 2.12. The method helps to
estimate the relative thickness between layers and can also be used as a quick check for
the quality of the metallic film (see Appendix A). However, this SPP-based approach
only works well for the p-polarized light at the infrared wavelength, where the low
dispersion of SPP gives a sharp dip in the reflectivity. For shorter wavelengths, the
dispersive SPP and lossy metallic layers broaden the angular dip, and the method is
no longer reliable.
1800 250
Ag Ag
along x ( µ m)
1600 Au Au
Penetration depth of Espp along z (nm)
Au/Co Au/Co
1400 200
Au/Co/Ag Au/Co/Ag
1200
spp
150
Propagation length of E
1000
800
100
600
400 50
200
0 0
500 1000 1500 500 1000 1500
Wavelength (nm) Wavelength (nm)
(a) (b)
Figure 2.13 – Propagation length and skin depth of the SPP field excited at metal/air
surface in different structures: monolayer of Ag (blue), monolayer of Au (green), bilayer
of (20 nm)Au/(30 nm)Co, trilayer of (5 nm)Au/(10 nm)Co/(25 nm)Ag.
Effective medium approximation

The effective dielectric function of a metallic multilayer structure can be approximated
by an integral of ǫ(z) within the penetration depth δp of an electromagnetic radiation
into the metal [7, 12]:
1 ∞
Z
ǫef f = ǫ(z)e−z/δp dz (2.64)
δp 0
Besides MTM, effective medium approximation is a useful tool to predict the optical
properties of metallic multilayers. For examples, one can apply Eq. 2.64 to calculate
29
the effective dielectric function of a (20 nm)Au/ (30 nm)Co bilayer and a (5 nm)Au/
(10 nm)Co/ (25 nm)Ag trilayer (Au on top). Then replace ǫm in Eq. 2.57 by ǫef f to
find the wavevector kspp of the SPP field excited at the surface of Au/air. Fig. 2.13
shows the propagation length and skin depth of the SPP fields excited in the hybrid
metal/ferromagnet thin films (the bilayer and the trilayer) and compare them with the
SPPs excited in the monolayer of Ag and Au. It can be seen that the ferromagnetic
layer introduces optical losses to the system and reduces significantly the propagation
length of the SPP. However, the skin depth of the SPP field into the metals remains
around 24 nm at long wavelengths, not much different from the pure monolayer.
2.4.4 Magneto-plasmonics
Magneto-plasmonics refer to the magneto-optical effects in metallic nanostructures
where the coupled mode of surface plasmons and incident light can be activated. Fur-
thermore, in the hybrid metal/ferromagnet thin films, the magnetization in the ferro-
magnetic layer causes non-zero off-diagonal components of the dielectric function, as
we have seen in Table. 2.2. For example, in the transversal MOKE configuration, the
effective dielectric tensor has the form
0 ǫxz
 
ǫef f ef f My
ǫ̂ef f (ω) =  0 ǫef f 0  (2.65)
xz
−ǫef f My 0 ǫef f
The tensor component ǫxz leads to a magnetization-induced modulation ∆kmp in the

wavevector of a SPP excited at metal/air interface [3, 7, 12, 14, 25, 110]
kspp (±M ) = kspp,0 ± ∆kmp

ǫef f iǫxz
ef f M (2.66)
r
= k0 1± √
ǫef f + 1 1 − ǫ2ef f ǫef f
where ǫxz ef f can be well approximated by effective medium approximation in Eq. 2.64
using the known ǫxz (z) of the ferromagnetic layer. The non-magnetic SPP wavevector
kspp,0 is close to that obtained in a single layer of noble metal, except a higher imaginary
part kspp,0
′′
representing for the losses. This model has been used to explain the decay
of magneto-plasmonic signal as the magnetized layer is buried further from the open
surface.
For the nonlinear magneto-plasmonics, the magnetic modulation of nonlinear emis-
sion can no longer be described by the magnetization-induced change of wavevector.
In these cases, the complexity of the model escalates due to the involvement of dif-
ferent sources of SH emission inside the materials. In the simplest scenario, one must
determine a dominant nonlinear susceptibility, i.e. a dominant SH polarization, and
then calculate the corresponding SH fields, their interference with the SPP fields and
the out-coupled Fresnel coefficients. The theoretical description and the simulation
approaches will be present in chapter 3.
30
Chapter 3
Nonlinear magneto-plasmonics on
Kretschmann configuration
While the idea of coupling SPP with the magneto-optical effects have attracted some
attentions in recent years[3, 14], the researches related to the nonlinear magneto-
plasmonics are scarce due to the complexity of the magnetization-induced second-
harmonic emission at the surface and interfaces. While some phenomenological models
have been developed [6, 7, 15], the coupling between SPP field and the SH polariza-
tions is still unclear. This chapter is devoted for our experimental findings in this
subject. Besides, it is demonstrated in this chapter that the nonlinear SPP coupling
in the Kretschmann configuration can be described by classical theories, and the SH
magnetic contrast is actually a measure for the SPP coupling efficiency.
3.1 Measurements of Kretschmann magneto-plasmonics

3.1.1 Samples
Following the discussion in section 2.4, in order to sustain the propagating SPP in the
visible/infrared spectrum, the metal should have dielectric function with small imagi-
nary part within this spectral range, which is the property of noble metals (gold, silver,
copper). SPP mode can carry the information along the film surface with a propaga-
tion length up to several hundreds of microns [111], and then it can be coupled out to
extract the optical signal. Our goal in this work was to control the propagating SPP by
combining it with magneto-optical effect. The propagating SPP field can be tuned by
an external magnetic field, but the magneto-optical effect of diamagnetic noble metal is
rather small and thus impractical [8–10, 30]. So in order to have large magneto-optical
response, ferromagnetic metals (nickel, cobalt, iron) are favorable since the in-plane
magnetization of a thin (sub 30 nm) ferromagnetic layer can be switched with small
external magnetic field (few millitesla) [7]. To combine the functionalities of magneto-
optics and SPP, the simplest structure is a stack of metallic layers.
The simplest form one can think of is the bilayer of ferromagnetic/noble metal with
31
CHAPTER 3. NONLINEAR MAGNETO-PLASMONICS ON KRETSCHMANN
CONFIGURATION
z
3 x Au
Au 2 Co
2
Co Ag
1
By 1
Sapphire
Sapphire Silica
θ θ
SHG(2ω) SHG(2ω)
R(ω) R(ω)
(a) (b)
Figure 3.1 – Kretschmann configuration applied for the metal/ferromagnet multilay-

ers (a) Bilayer structure of Au/Co deposited on sapphire. (b) Trilayer structure of
Au/Co/Ag deposited on fused-silica. The dashed rectangles mark the interfaces where
SH emission is considered in the later simulation.
the noble metal exposing to the air, as depicted in Fig. 3.1a. However, the high optical
loss of ferromagnetic metal (due to high imaginary part of dielectric function in the
visible spectrum) causes the damping of the SPP propagating along Au/air surface, as
shown in Fig. 2.13. Though it is not an effective magneto-plasmonic structure in linear
regime, the simple bilayer is a good playground for fundamental research in nonlinear
regime [32], especially when ultrafast acoustic can be integrated in the same structure.
It has been shown in prior research that it is possible to probe the acoustic modulation
in thin metallic films by the SPP excited in Kretschmann geometry [112] since the SPP
wavevector kspp is highly sensitive to the free-carrier density1 within its skin depth [113].
The bilayer of gold/cobalt is chosen here because its dynamic properties have been well
explored [31], and we can expect to combine acousto-magneto-plasmonics in a single
structure.
The studied bilayer samples consist of hexagonal crystal structure Co(0001) and
face-centered cubic Au(111) consecutively deposited on a 500 µm double-side epipol-
ished sapphire substrate Al2 O3 (0001)2 (Crystec GmbH) using dc-magnetron sputtering
in a high vacuum chamber (base pressure 10−7 mbar) at temperature 250℃. Apart from
the optical transparency needed for reflectivity measurement in the visible/infrared
range, high melting point (2053 ℃) and thermal conductivity (42 W/m·K) of sapphire
single crystal are essential in our measurement, where a large fluence of excitation is
needed to have detectable nonlinear signals. Varying the thickness of gold layer on top
of the 30 nm-thick cobalt, as in Table 3.1, is a way to tune the SPP field. Transversal
linear and nonlinear MOKE will be performed on these samples.
In contrast to the bilayer, a trilayer structure, in which the ferromagnetic metal is
sandwiched between two noble metals and buried within the SPP field skin depth, has
shown superior magneto-plasmonic response in linear regime [3]. Previous systematic
researches have identified that for the trilayer Au/Co/Au, cobalt thickness of approx-
imately 6 nm will give optimum SPP-enhanced linear magneto-optical response [25].
Applying these knowledge for nonlinear magneto-plasmonics, we choose to study the
1
which is proportional to the real part of dielectric function at frequency of probe photons
2
c-axis is perpendicular to the substrate, with the tilt less than 0.3◦
32
3.1. MEASUREMENTS OF KRETSCHMANN MAGNETO-PLASMONICS
Structure Thickness of layers

(20 nm)Au/(30 nm)Co/sapphire
Bilayer (30 nm)Au/(30 nm)Co/sapphire
(40 nm)Au/(30 nm)Co/sapphire
(5 nm)Au/(10 nm)Co/(15 nm)Ag/fused silica
Trilayer (5 nm)Au/(10 nm)Co/(25 nm)Ag/fused silica
(5 nm)Au/(10 nm)Co/(35 nm)Ag/fused silica
Table 3.1 – In this work, the linear magneto-plasmonic measurements were performed
on the set of bilayer. The set of trilayer was used for the nonlinear magneto-plasmonic
measurements.
structure of Ag/Co/Au trilayers, illustrated in Fig. 3.1b, for the reasons below:
i) the active layer for SPP excitation is silver layer, assuring that the photon energy
of SHG (1.6 eV) and THG (2.4 eV) will not exceed the interband transition (4.0 eV)
and nonlinear SPP will not be overdamped.
ii) the inplane magnetization of the 5 nm-thick cobalt layer can be switched by a small
external magnetic field 20 mT.
iii) the magnetic SHG (mSHG) originated from both sides of cobalt layer, i.e. Ag/Co
and Co/Au interfaces, will not destructively interfere, which might happen in Au/Co/Au
structure.
Similar to the set of bilayers, varying the thickness of silver layer (Table. 3.1) will
modify the efficiency of SPP excitation, and we can define the optimal condition for
nonlinear magneto-plasmonic signal. In general, the trilayers have higher SH yield than
the bilayers by one-order of magnitude, and thus are a better candidate for practical
applications.
3.1.2 Setup
The used setup scheme is shown on Fig. 3.2. A right-angle prism and a detector are
mounted on two co-axial motorized rotating stages (Newport PR50PP), in which the
rotation angles of the two devices are θ and 2θ, respectively. The illuminated spot
on the thin film is aligned to the rotation axis of motorized stages. The alignment is
performed at angle of incidence 45◦ at the hypotenuse side of the prism, where the
studied sample is installed. The incident beam is thus perpendicular to the square
sides of the prism. This is the “reference angle”, preset as zero angle of rotation. The
prism can be rotated in the range of -30◦ to +25◦ around zero angle. The rotation
angle of detector is twice as much as that of the prism. The rotation angle of prism is
then recalculated into angle of incidence inside sample substrate applying Snell law3 .
After conversion, the maximum range of incident angle in the air is 20-75◦ , minimum
angular increment 0.05◦ , unidirectional repeatability 0.05◦ . The overall goniometric
motorized platform is compact, enabling multi-parameter measurements and can be
3
n1 sin θ1 = n2 sin θ2 , where n1 , n2 are refractive indices of two transparent media 1 and 2. θ1 , θ2
are the angles of the light beam in these media to the surface normal.
33
CONFIGURATION
Figure 3.2 – Scheme of a linear magneto-plasmonic setup employing Kretschmann

configuration. (a) Measurement of magnetization-induced variations in reflectivity
∆R(±M ). (b) Measurement of the average reflectivity R. The ratio ∆R/R implies
the normalized transversal MOKE signal.
used for various configurations, such as linear and nonlinear magneto-plasmonics in

either Kretschmann geometry or diffraction grating.
Linear magneto-plasmonic measurements

For linear magneto-plasmonic measurements, an external magnetic field is applied. It is
generated by an electromagnet, field switching frequency 440 Hz (resonance frequency
of the registered LC circuit). A piece is cut from the ferrite ring magnet in order to
have straight magnetic field lines between two ends. The prism is held free-standing
between the two poles in such a way that the magnetic field is perpendicular to the
plane of incidence (see By in Fig. 3.2a).
The (TM) p-polarized collimated beam is used to ensure that each angle of inci-
dence corresponds to only one kx and thus one kspp , i.e. k-space measurement. A
silicon photodiode is used for visible spectral range (400-1070 nm) and a germanium
photodiode is used for the near-infrared spectrum (800-1700 nm). In the transversal
MOKE configuration discussed in section 2.2.3, reversal of the external magnetic field
results in the modulation of reflectivity ∆R/R, so-called MOKE signal. Using the lock-
in amplifier SR830, the tiny magnetization-induced change ∆R = R(+M ) − R(−M )
is measured through a reflectivity locked-in at the switching frequency of the magnetic
field 440 Hz (see Fig. 3.2a). The signal is normalized to R, which is the reflectivity
measured when the magnet is removed and the incident light is chopped at 440 Hz, as
shown in Fig. 3.2b.
Nonlinear magneto-plasmonic measurements

In the nonlinear magneto-plasmonic setup, high magnetic field switching frequency is
not necessary because of the long integration time of the photon counter and strong
magnetic nonlinear signal. The external magnetic field is generated by Neodymium
34
Figure 3.3 – Top-view of a SH nonlinear magneto-plasmonic setup based on

Kretschmann geometry.
permanent magnets with field strength of ≈150 mT at the distance of 5 mm from the
pole (see Fig. 3.4). The change of magnet poles is performed with a linear motorized
level (Newport). Distance between the two magnets on the level is 34 mm, ensuring no
stray magnetic field between two poles at the distance of 5 mm underneath the magnets.
Sample holders are made of non-magnetic materials, like aluminium or teflon.
600
Magnetic field strength (mT)
500
400
300
200
100
0
0 5 10 15 20 25 30
Distance from the pole (mm)
Figure 3.4 – The dependence of magnetic field strength on the distance from a pole of
a Neodymium permanent magnet cube (each side 1.5 mm).
In nonlinear measurements (see Fig. 3.3), the p-polarized (TM) fundamental radi-
ation is focused to the thin film by an uncoated plano-convex lens (Thorlabs) with the
focal length f=300 mm. Focusing angle, calculated for the incident 3.5 mm-diameter
collimated beam at wavelength 1560 nm, is 0.67◦ in the air, which becomes 0.45◦ inside
glass prism and 0.35◦ inside sapphire substrate. This angle is less than the angular
width4 0.8◦ of ATIR dip of trilayer structure. Focusing the incident light will increase
4
full-width half maximum FWHM
35
CONFIGURATION
the number of incident wave vectors, causing angular broadening. However, it is nec-
essary to have a high excitation fluence in a small area for the generation of nonlinear
signal. The choice of the focal length is thus a compromise between effective coupling
to SPP modes and SH output.
A color filter is placed after the lens to filter possible radiations from the optical
components, thus ensures pure fundamental radiation. Generated on the metallic thin
films, the SH output is then spectrally separated by a color BG-39 filter (Schott) and
recorded with a photomultiplier tube (PMT). The power dependence of the detected
signal in Fig. 3.5 shows a quadratic behavior I(2ω) ∝ I 2 (ω), assuring that the signal
coming through the low-pass filters is second-harmonic generation.
4
x 10 5
3.5 10
SHG SHG
2
3 1.6x2 1.6x
SHG intensity (counts/s)
2.5 4
10
1.5
3
10
1
0.5
2
0 10 1 2 3
0 50 100 150 10 10 10
Average power (mW) Average power (mW)
(a) (b)
Figure 3.5 – The fluence dependence of the signal detected by the PMT. The quadratic
behavior shown in (a) and the linear dependency in the logarithm plot (b) ensures that
the detected signal is a second-harmonic generation.
To detect the optical nonlinear signals, the PMT (Hamamatsu R955) is utilized
along with a two-channel gated photon counter (SR400). Generated at cathode based
on photoelectric effect, the photocurrent is amplified in 9 dynodes of the PMT be-
fore reaching anode, with a typical gain of 107 at voltage 1000 V. The multialkali
photocathode of the PMT allows wide spectral response 160-900 nm, with quantum
efficiency of 3% and cathode radiance sensitivity 20 mA/W for wavelength 780 nm.
At very low level of illumination, each photoelectron generated at PMT cathode is
represented in the oscilloscope as an electric pulse with typical rise time 2.2 ns and
transit time 22 ns. The pulse rate is proportional to the number of incident photons per
second. To separate the dark current (typically 150 counts/s at supply voltage 900 V),
the photon counter is set at gated regime, i.e. one channel counts the overall signal
while the other channel concerns only the dark counts. It is performed by chopping
36
the excitation light, then use the chopping frequency as the trigger for the counting
gates. High chopping frequency helps to reduce the heat accumulation in sample. The
counting scheme is illustrated in Fig. 3.6. For the chopping frequency of 1.7 kHz, the
trigger period is ≈590 µs. The temporal width of both counting gates is 230 µs (≈
40% of trigger period), much larger than the PMT transition time. The signal is the
difference of count rate between gate A and gate B. Integration time for each data
point is 0.46 s. The detection of weak signals, like angular dependence of nonlinear
magnetic contrast, usually requires high statistics. This detection scheme is limited by
shot noise, which relates to the particle nature of photons, and thermal noise, which
is the cause of dark counts.
590 µs
Trigger
(chopper)
Gate A
(signal+bkg)
230 µs
Gate B
(bkg)
230 µs
Figure 3.6 – An example of the counting scheme. The trigger frequency is set by the
chopper. Gate A counts both the signal and the background (bkg) at the open phase
of the chopper. Gate B counts the background when the chopper is close.
3.1.3 Light source

The choice of the light source is based on the dispersion relation of SPP (Fig. 2.9). In
Kretschmann configuration, the light line is tilted by a factor of n sin θ and thus allows
the variation of incident wave vector within a grey cone for θ = 0 − 90 deg. As it is
clearly depicted in the dispersion relation, at a certain frequency ω only one angle can
excite the corresponding SPP, represented as a sharp dip in the angular dependence of
reflectivity. But in fact, there is always an angular broadening, due to the optical loss
of SPP. The damping of SPP is particularly high when there is spectral overlap with
interband transition (see Fig. 2.13).
The telecommunication wavelengths 1.3 µm (0.95 eV) and 1.56 µm (0.8 eV) are
employed, which give SHG at the wavelengths 650 nm (1.9 eV) and 780 nm (1.55 eV),
respectively. Both the fundamental and SH photon energies are far below the interband
transition thresholds of noble metals (Table. 2.1). Due to the low dispersion, the SPPs
excited at the infrared spectral range give sharp features in the angular spectrum.
Furthermore, as shown in the dispersion relation in Fig. 2.9, the light line inside the
37
CONFIGURATION
prism (substrate) at the fundamental frequency nk0 (green dashed line) is very close to
the SPP wavevector kspp (red solid line) at small photon energy (< 2 eV), which means
that without significant dispersion, a fundamental radiation having photon energy
~ω < 1 eV can excite both the fundamental SPP with wavevector kspp (ω) and the
second-harmonic SPP with wavevector kspp (2ω) ≃ 2kspp (ω). The wavelength 1.55 µm
(0.8 eV) even allows us to inspect the nonlinear magneto-plasmonics at the third-
harmonic generation (THG) at the wavelength 520 nm (see Appendix B), though the
THG would be much more dispersive than the SHG.
A femtosecond Erbium-doped fiber laser (Toptica FemtoFiber pro NIR) installed at
IMMM is used as the light source. The laser has two outputs, fundamental wavelength
at 1560 nm and second-harmonic wavelength (generated by an integrated piece of
periodically poled lithium niobate PPLN) at 780 nm. The pulse duration of both
outputs is sub-120 fs, repetition rate 40 MHz.
Some nonlinear magneto-plasmonic measurements were performed at Department
of Physics (University of Konstanz). There, the light source is a fiber laser operated
at center wavelength 1200 nm with bandwidth of ≈400 nm. The pulse duration is
sub-10 fs, and repetition rate 40 MHz.
3.2 Results
In this section, the priliminary results of the Au/Co bilayer and the Au/Co/Ag trilayer
are shown to give a general look on the properties of magneto-plasmonics in multilayer
structures. The experimental data will be discussed in more details and fitted with the
numerical simulations in the next sections.
3.2.1 Bilayer structure of Au/Co

In the prior research [6, 7, 30], the magneto-plasmonics becomes effective in the metal-
lic multilayer structures when the ferromagnetic layer is buried within the intensity
penetration depth 12 nm from the open surface. However, in this work, I have shown
that as long as the ferromagnetic layer is within the field skin depth, i.e. 24 nm from
the open surface, reasonable magneto-plasmonics behaviors appear in both the linear
and the nonlinear regimes.
Linear MOKE in the bilayers

Cobalt, with its high imaginary part of dielectric function ǫ′′ in the visible spectrum, is
the cause for optical loss in multilayer structures. It results in the significant reduction
of the SPP propagation length along the surface of Au/air, as shown in Fig. 2.13 in
the previous chapter. The bilayer Au/Co thus has been considered as an ineffective
magneto-plasmonic structure, and people in the field prefer to work with the trilayer
Au/Co/Au or Au/Co/Ag, where the SPP-induced MOKE is more pronounced. Despite
the loss, I have observed the SPP-induced transversal MOKE on the bilayer of (20-30-
40 nm)Au/(30 nm)Co at the wavelength 633 nm using HeNe laser source, shown in
Fig. 3.7.
38
3.2. RESULTS
(20 nm)Au/(30 nm)Co

(30 nm)Au/(30 nm)Co
(40 nm)Au/(30 nm)Co
Figure 3.7 – Angular dependence of linear transversal MOKE measured on the bilayer
(20-30-40 nm)Au/(30 nm)Co at the wavelength 633 nm.
As discussed in the section 2.4.4, the angular variation of MOKE signal represents
for the magnetic modulation of the SPP wavevector, specifically ∆kspp = kspp (+M ) −
kspp (−M ), which originates from the variation of off-diagonal components ǫxz ef f My in
the dielectric function of bilayer. The maximal variation of the MOKE response follows
the decay of ∆kspp when the top Au layer is thicker, i.e. when the distance from cobalt
layer to the open surface increases, shown in the inset of Fig. 3.7. This observation
confirms the undeniable effect of SPP on the magneto-optics in bilayer.
Nonlinear magneto-plasmonics in the bilayer
Nonlinear Kretschmann ATR measurement on sample (20 nm)Au/ (30 nm)Co/ sap-
phire, in general, gives low SH output with the peak at only few hundreds counts/s.
However, the effects of linear and nonlinear SPP modes shown on the k-spectrum of
the SH intensity and the nonlinear magnetic contrast in the bilayer are very clear be-
cause the active interfaces (air/Au, Au/Co and Co/sapphire as shown in Fig. 3.1a) are
well separated. Fig. 3.8 shows the angular dependence of the SH intensities (green and
black solid lines) at opposite magnetizations and the corresponding magnetic contrast
(magenta solid line). The linear reflectivity at the fundamental wavelength 1560 nm
(red dashed line) and at the half-wavelength 780 nm (blue dashed line) mark the an-
gles where the SPP at these wavelengths can be excited. In Fig. 3.8a, the SH intensity
(green solid line) has a peak at 35.4◦ and a dip at 36.1◦ , which are in accordance with
the angular minima of linear reflectivity at 1560 nm and 780 nm, where the SPP(ω)
39
CONFIGURATION
Figure 3.8 – (a) Angular dependence of SH output from the bilayer of (20 nm)Au/
(30 nm)Co measured at two different magnetizations (green and black solid lines). (b)
The angular dependence of the SHG magnetic contrast (magenta solid line). The linear
reflectivity (arbitrary unit) at fundamental (1560 nm) and SHG (780 nm) wavelengths
are shown in (a) and (b) as dashed lines in order to mark the angles of incidence where
SPP(ω) and SPP(2ω) are excited in the linear configuration.
and SPP(2ω) are supposed to be excited.

Similarly, the SH magnetic contrast in Fig. 3.8b has a double-dip feature, in which
each dip occurs near the angular minimum of a linear reflectivity. Contradict to the
linear magnetic contrast of <1%, the second-harmonic magnetic contrast can reach
as high as 60%. If the angular dependence of linear contrast ∆R/R shows just one
oscillator, the nonlinear magnetic contrast exhibits two modes resonant at different
angles.
3.2.2 Trilayer structure of Au/Co/Ag

The peak of SH intensity in the case of trilayer can reach few thousands counts/s, an
order-of-magnitude higher than that in the bilayer. Similar to the linear regime [3,
14], the nonlinear magneto-optical signals in metallic multilayers can be enhanced or
damped by SPP field. To demonstrate that possibility, one can vary the thickness
of Ag layer in the trilayer structures (5 nm)Au/ (10 nm)Co/ (15-25-35 nm)Ag. The
linear reflectivity, SH intensity and nonlinear magnetic contrast of these three samples
are shown in Fig. 3.9. Thinner silver layer supports stronger SPP modes, resulting
in deeper reflectivity minima in Fig. 3.9a and stronger SHG output in Fig. 3.9b oc-
curring at the same angle. The peak SH intensity at angle 44.5◦ represents for the
second-harmonic signal generated at the surface Au/air. Its intensity thus follows the
attenuation of an electromagnetic field at 780 nm in silver, plotted as the inset of
40
3.2. RESULTS
(a)
(b)
(c)
Figure 3.9 – Angular dependence of (a) linear reflectivity at 1560 nm, (b) SHG
intensity at 780 nm and (c) transversal nonlinear magnetic contrast ρ in trilayer
(5 nm)Au/(10 nm)Co/(15-25-35 nm)Ag. The SHG peak at 44.5 deg corresponds to
the linear SPP(ω) mode, confirmed by the dip of linear reflectivity at the same angle.
41
CONFIGURATION
Fig. 3.9b.
The magnetic contrast ρ in Fig. 3.9c also shows increasing value for samples with
thinner layer. For the sample (5 nm)Au/ (10 nm)Co/ (15 nm)Ag, the SH contrast (red
solid curve) varies from around -60% at angle 45.2◦ to 40% at angle 46◦ , higher than
the reported data for a nonlinear magneto-plasmonic signal in a trilayer [6]. From the
technical perspective, it means that the SH intensity can be amplified four times by a
weak external magnetic field of a permanent magnet.
3.3 Theoretical description

As we have discussed in the section 2.3, the sources of SHG in metallic multilayer
structures are limited to the surface and interfaces, according to the electric dipole
approximation. The nonlinear polarizations at metal/dielectric interfaces have been
described in Eq. 2.46
Pi (2ω) = ǫ0 χijk Ej (ω)Ek (ω)
where χijk is the second-order susceptibility tensor, and i, j, k denote Cartesian coor-
dinates. These polarizations are sources of light at the double frequency (2ω), which
radiate towards both the dielectric substrate and the air (see A0 in Fig. 3.10). In
Kretschmann configuration, the field of SPP(2ω) can be coupled when the SHG phase-
matching condition 2kx = kspp (2ω) is fulfilled, resulting in a nonlinear leaky (radiative)
mode detectable at the far-field, Aspp in Fig. 3.10.
Figure 3.10 – Illustration of nonlinear excitation of SPP by SHG in bilayer (20 nm)Au/
(30 nm)Co employing Kretschmann configuration. External magnetic field is applied
perpendicular to the plane of incidence. Incident field (red) generates nonlinear sheet
polarizations and excite SPP(ω) at Au/air surface. Nonlinear polarization at each
interface acts as radiative source of SHG (blue). Aspp (2ω) stands for the radiative SPP
at 2ω, while A0 (2ω) denotes the non-resonant SHGs.
42
3.3. THEORETICAL DESCRIPTION
Within our concern, the fundamental excitation is p-polarized with the electric
field components Ex (ω) and Ez (ω). The external magnetic field is perpendicular to
the plane of incidence By , as in the tranversal MOKE configuration. Referring to
section 2.3.2, six non-zero second-harmonic susceptibilities are valid in this case, and
three among them are dependent on the magnetization, as listed in Table. 3.2. The
resulting SH polarizations are either Px (2ω) or Pz (2ω), i.e. p-polarized SHG.
Non-magnetic (even) χ(2) Magnetic (odd) χ(2)

χzzz χxzz
χzxx χxxx
χxzx χzzx
Table 3.2 – The non-zero second-order nonlinear susceptibility components for p-

polarized fundamental radiation.
Now, the model of nonlinear surface plasmon polariton (SPP) excited in Kretschmann
geometry will be introduced in the following. This section includes two parts: (i) the
radiation of nonlinear SPP at 2ω excited in Kretschmann configuration, and (ii) the
derivation of SHG magnetic contrast under the excitation of nonlinear SPP(2ω).
3.3.1 Radiative nonlinear surface plasmon polaritons

−−→
The amplitude Aspp,n of the nonlinear SPP field Espp (2ω) launched by the nonlinear
→
−
sheet polarization P (2ω) at a certain interface n is the scalar (dot) product of those
two:
→
− −−→
Aspp,n = P (2ω) · Espp (2ω)
(3.1)
= Px (2ω)Espp,x (2ω) + Pz (2ω)Espp,z (2ω)
The total amplitude, as a result of the interference of scattered and reflected fields
with amplitudes An , can be written:
ZZ
Aspp ∝ ∗
P (2ω)Espp (2ω)dzdx (3.2)
In this equation, the second-order nonlinear polarization P (2ω) represents the sum
of sources localized at N individual interfaces, each is located at z = zn (z = 0 at the
open surface) and can be thus written
N
X
P (2ω) ∝ Pn δ(z − zn )exp[2ikx (ω)x − 2iωt]
n=1
This nonlinear polarization propagates along the surface with the phase velocity de-
termined by 2kx (ω) = 2k0 (ω)n(ω) sin θ, where k0 (ω) is vacuum wavevector, n(ω) is the
refractive index of the dielectric substrate at the fundamental frequency, and θ denotes
the angle of incidence.
43
CONFIGURATION
The electric field of the co-propagating SPP(2ω), which decays along the normal
(z) direction and propagates along the in-plane x direction, is written as
z
Espp (2ω) ∝ exp[− + ikspp (2ω)x − 2iωt]
δskin
where δskin is the field skin depth (≈ 24 nm5 in metals). As a consequence, Eq. 3.2
becomes:
ZZ
∗
Aspp ∝ P (2ω)Espp (2ω)dzdx =
N
Z X Z
z
−δ
∝ Pn δ(z − zn )e skin dz × e−i[kspp (2ω)−2kx (ω)]x dx (3.3)
n=1
N
!
X − δ zn 1
= i Pn e skin
n=1
kspp (2ω) − 2kx (ω)
kx (ω) is real, but kspp (2ω) is complex and can be written as kspp = kspp
′ ′′
+ ikspp , where
kspp and kspp denote the real and imaginary parts of the SPP wavevector. The second
′ ′′
factor in Eq. 3.3 is then:

1 1
= ′ (3.4)
kspp (2ω) − 2kx (ω) ′′ (2ω)
[kspp (2ω) − 2kx (ω)] + ikspp
which has a shape of a Lorentzian complex function, coherent with the SPP theory.
To work with a normalized and dimensionless function, Eq. 3.5 is multiplied by kspp
′′
:
′′
kspp (2ω)
L(kx ) = (3.5)
′ (2ω) − 2k ] + ik ′′ (2ω)
[kspp x spp
This equation reaches the maximum when kspp

′
(2ω) − 2kx (ω) = 0, and thus one gets
′
kspp (2ω) = 2k0 (ω)n(ω) sin θ (3.6)
Eq. 3.6 describes the nonlinear phase-matching condition for the second-harmonic
SPP(2ω). Linearizing sin θ at the vicinity of the resonant angle θ0 6 , Eq. 3.6 becomes:
′
kspp (2ω)
θ − θ0 = − tan θ0 (3.7)
2k0 (ω)n(ω) cos θ0
The complex Lorentzian function L(kx ) can be then written in term of angle
of incidence θ by dividing both the numerator and the denominator of Eq. 3.5 by
2k0 (ω)n(ω) cos θ0 :
Γ
L(θ) = (3.8)
(θ − θ0 ) + iΓ
5 δskin
The intensity penetration depth is =12 nm
6
2
The linearization of sin(θ) at angle θ0 is: sin θ0 + cos θ0 × (θ − θ0 )
44
3.3. THEORETICAL DESCRIPTION
′′
kspp (2ω)
in which the angular width is described by Γ = and L(θ0 ) = 1.
2k0 (ω)n(ω) cos θ0
The Lorentzian shape is kept consistent with the theory. As such, the resonant angle
θ0 represents the excitation condition of the nonlinear SPP(2ω), and the angular width
Γ corresponds to the optical loss of the field via kspp
′′
(2ω).
The final form of the second-harmonic SPP field amplitude Aspp in Eq. 3.3 reads:
" N #
X
(3.9)

Aspp (θ) ∝ i Pn (θ) exp − zn /2δskin L(θ)
n=1
The approximation in Eq. 3.9 will be employed in our simulations of magnetic SHG
generated in multilayer structures in the later section. Note that exp − zn /2δskin is

a constant factor for each interface n, which is positioned at the depth zn from the
open surface.
To complete the story, the SHG observed at the far-field is the interference of non-
resonant second-harmonic fields (denoted as A0 = P0 ) and the leaky nonlinear SPP
field Aspp evoked by the polarizations located near metal/air surface, as summarized
in Fig. 3.10. The detected SHG intensity is thus:
I2ω = |A0 + Aspp |2 (3.10)
3.3.2 Calculate SHG magnetic contrast (abcd-model)

Based on the coupling with surface plasmons at the double-frequency 2ω, represented
by the function L(θ) described in Eq. 3.8, one can distinguish the sources of SHG into
the SPP non-resonant sources (A0 ∝ Enon−res ) and the SPP resonant ones (Aspp ∝
LEres ). The total SHG intensity is thus:
I2ω = |Enon−res + LEres |2 (3.11)
where each of these components includes magnetic and non-magnetic parts:
Enon−res = ã + b̃m
(3.12)
˜
Eres = c̃ + dm
Here m denotes the in-plane magnetization. ã and c̃ are magnetization-independent,

while the magnetic components b̃ and d˜ are considered to be very small. The total
SHG intensity is:
˜ 2
I2ω = |(ã + b̃m) + L̃(c̃ + dm)| (3.13)
Note that the average SHG is the intensity at m = 0, which is:
I2ω (m = 0) = |ã + L̃c̃|2 (3.14)
ã, b̃, c̃, d˜ and L̃ are, in general, complex. For simplification, the tilde is removed
from here on; for example, the complex component ã is denoted as a, and its complex
45
CONFIGURATION
E(+M)=Enm+Emag
Enm=a+Lc
E(-M)=Enm-Emag
φ Emag=b+Ld
-Emag=-(b+Ld)
Figure 3.11 – The phasor diagram of non-magnetic and magnetic SHG fields, re-
spectively Enm and Emag , taking account the non-resonant (a, b) resonant (c, d) SH
components. Zero magnetic contrast, i.e. E(+M ) = E(−M ), happens when the phase
difference is ϕ = π2
conjugation is a∗ . The intensity I2ω is fully expanded into:
I2ω = [a + bm + L(c + dm)][a∗ + b∗ m + L∗ (c∗ + d∗ m)]

= |a|2 + a∗ bm + ab∗ m + |b|2 m2 + L(a∗ c + a∗ dm + cb∗ m + db∗ m2 ) (3.15)
+ L∗ (ac∗ + c∗ bm + ad∗ m + bd∗ m2 ) + |L|2 (|c|2 + c∗ dm + cd∗ m + |d|2 m2 )
According to this, the change of SHG intensity depending on magnetization, as-

suming that m is very small, is:
∂I2ω
= a∗ b + ab∗ + 2|b|2 m + L(a∗ d + cb∗ + 2db∗ m)
∂m (3.16)
+ L∗ (c∗ b + ad∗ + 2bd∗ m) + |L|2 (c∗ d + cd∗ + 2|d|2 m)
The difference in SHG intensity upon the reversal of magnetization is approximately

its derivative at m = 0, giving:
∂I2ω
∆I2ω (±m) = (m = 0)
∂m
= a∗ b + ab∗ + L(a∗ d + cb∗ ) + L∗ (c∗ b + ad∗ ) + |L|2 (c∗ d + cd∗ ) (3.17)
= 2ℜ{(a∗ + L∗ c∗ )(b + Ld)}
= 2ℜ{|a + Lc|e−iϕ |b + Ld|}
where ϕ is the phase difference between the non-magnetic (a + Lc) and the magnetic
(b + Ld) components, as illustrated in Fig. 3.11, the phasor diagram of the SH fields.
Formulated in Eq. 2.50 in the previous chapter, the magnetic contrast ρ is usually
used as a technical measure for the nonlinear magneto-optical effect. Here ρ can be
written as
46
3.4. SIMULATION METHODS
∆I2ω 2ℜ{|a + Lc|e−iϕ |b + Ld|}

ρ= =
I2ω (m = 0) |a + Lc|2
(3.18)
2ℜ{|b + Ld|e−iϕ }
=
|a + Lc|
Noting that the magnetic components b and d are small, and L can be represented
in term of angle of incidence L(θ), the angular dependence of the SH magnetic contrast
ρ(θ) should imitate the shape of ℜL(θ).
Eq. 3.18 includes both the non-magnetic |a + Lc| and the magnetic |b + Ld| field
components in the magnetic contrast. One can annihilate the non-magnetic parts in
eq. 3.17 dividing the ∆I2ω by I2ω (m = 0):
p
∆I2ω
φ= p = 2ℜ{|b + Ld|e−iϕ } (3.19)
I2ω (m = 0)
This newly-defined magnetic contrast φ in Eq. 3.19 has simpler form than the con-
ventional contrast ρ in Eq. 3.18, but it is not yet normalized. Thus, to quantify and
to compare the nonlinear magneto-optical effects, it is still more convenient to use
the original magnetic contrast ρ. The new contrast φ is adopted particularly in this
chapter just to illustrate the influence of the nonlinear SPP excitation L(θ) on the SH
magneto-optical signals, which is otherwise hidden by the non-magnetic component
|a + Lc| in the description of ρ. Throughout the text, both contrasts will be shown
and compared.
In a structure where the plasmonic-active layer is thick; for example, the trilayer
of (5 nm)Au/(10 nm)Co/(35 nm)Ag, the magnetic component of the non-resonant
SHG, observed at low angle of incidence, is almost zero as shown in blue curve in the
Fig. 3.9c, meaning b ≈ 0. The SH magnetic contrasts in this case can be written as:
∆I2ω 2ℜ{Lde−iϕ }
ρ= =
I2ω (m = 0) |a + Lc|
(3.20)
∆I2ω
φ= p = 2ℜ{Lde−iϕ }
I2ω (m = 0)
The magnetic contrast φ now becomes a direct measure for the nonlinear SPP influence
L(θ) in the magneto-plasmonic structures.
3.4 Simulation methods

This section shows how one can predict the SPP coupling efficiency L(θ). Two ap-
proaches will be presented (i) the forward approach, where one calculates the nonlin-
ear L(θ) and magnetization-induced SHG using effective medium approximation and
numerical tools, and (ii) the reverse approach, where the SPP coupling efficiency is
directly obtained from the experimental magnetic contrast.
47
CONFIGURATION
3.4.1 Forward approach

In order to attain the nonlinear SPP coupling efficiency L(θ) (Eq. 3.5) and the reso-
nant second-harmonic field amplitude Aspp (Eq. 3.3), first we need to derive the SPP
wavevector kspp . It is thus necessary to know the effective dielectric functions of the
plasmonic multilayer structure. This was done using Effective Medium Approxima-
tion [12], knowing that the SPP penetration depth δp at wavelength 1.56 µm is about
12 nm. By using Eq. 2.64, one can estimate the effective dielectric function ǫef f of a
metallic multilayer structure. Then ǫef f can be used to calculate the SPP wavevector
kspp , based on Eq. 2.57
ǫef f
r
kspp (x) = k0
1 + ǫef f
(3.21)
q ǫ
2 (x) = k p ef f
kspp (z) = ǫef f k02 − kspp 0
1 + ǫef f
With the complex kspp (x) = kspp
′ ′′
+ ikspp , one can readily calculate the coupling
efficiency of SPP in the linear regime:
′′
kspp (ω)
Llin (ω) = ′ ′′ (ω)
[kspp (ω) − k0 (ω)n(ω) sin θ] + ikspp
′′ (3.22)
kspp (2ω)
Llin (2ω) = ′ ′′ (2ω)
[kspp (2ω) − k0 (2ω)n(2ω) sin θ] + ikspp
and in nonlinear regime, as in Eq. 3.5
′′
kspp (2ω)
Lnl (2ω) = ′ (2ω) − 2k (ω)n(ω) sin θ] + ik ′′ (2ω)
[kspp 0 spp
Note that the nonlinear Lorentzian function (Lnl (2ω)) is different from the linear
Lorentzian function at the frequency 2ω (Llin (2ω)) because of dispersion. Within the
optical range, the refractive index of the substrate n(2ω) in Llin (2ω), in general, differs
from n(ω) in Lnl (2ω) .
Disentangle χ(2) components

Each interface in the multilayer structure is a source of SHG, which itself includes of
SH fields generated with different χ(2) components (referring to Table. 3.2). To simplify
the model, we assume that there is a dominant χ(2) component at each interface. They
are distinguished based on their magnetic properties and the possibility to couple
with SPP field. For example, in the structure of Au/Co/Ag/substrate mounted on the
Kretschmann configuration, SPP fields cannot be excited at the Ag/substrate interface.
And if the distance from this interface to the open surface is larger than the field skin
depth, i.e. 24 nm, the SH polarization located at the Ag/substrate interface is non-
resonant with the SPP excited in Au/air surface. This SH source is also non-magnetic,
since it is not in contact with the ferromagnetic Co layer.
48
The field amplitude at each interface of the multilayer system can be resolved by
the MTM. Fig. 3.12 shows the field distribution over angle of incidence in the sample
(28 nm)Ag/ (10 nm)Co/ (5 nm)Au. In this simulation, the field at the metallic side
of the interface is taken; for example, Ez (ω) and Ex (ω) at Au side in the interface of
Au/air is taken for calculation of sheet polarization Px (2ω) ∝ χxzx Ez (ω)Ex (ω). For
the interfaces between ferromagnetic and non-magnetic metals, such as Co/Au and
Ag/Co, the fields are taken at the ferromagnetic side. Due to the boundary condition,
Ex is continuous across the interface, but Ez will differ by a ratio of two dielectric
functions.
0.4
|E| (arb. u.)
10 1 |Ez| - Au/air
0.2 |Ex| - Au/air
0
35 40 45 50 55
0.6
|E| (arb. u.)
1
0.4 10 |Ez| - Co/Au
|Ex| - Co/Au
0.2
0
35 40 45 50 55
0.6
|E| (arb. u.)
10 1 |Ez| - Ag/Co
0.4
|Ex| - Ag/Co
0.2
0
35 40 45 50 55
0.3
|E| (arb. u.)
0.2
1
0.1 10 |Ez| - silica/Ag
|Ex| - silica/Ag
0
35 40 45 50 55
Angle of incidence θ (deg)
Figure 3.12 – Angular dependence of electric field |E| components of p-polarized

incident light, calculated at each interface using MTM method for the structure
(25 nm)Ag/ (10 nm)Co/ (5 nm)Au doposited on fused-silica.
As can be seen in Fig. 3.12, the roles of E-fields at the interfaces of Co/Au and
Ag/Co in the trilayer are similar. Thus the model can be further simplified by con-
sidering the Co/Au interface the only source of the magnetization-induced SHG. The
dominant role of each metallic layer in the structure Au/Co/Ag is summarized in
Table. 3.3 below.
In general, three non-magnetic susceptibilities (χzzz , χzxx and χxzx ) at the interface
Co/Au are non-zero, but they do not have as much influence compared to the interfaces
49
CONFIGURATION
Interfaces (index) Dominant χ(2) Magnetic properties SPP-resonance

Au/air (3) χxzx or χzxx no yes
Co/Au (2) χxxx yes yes
glass/Ag (1) χzzz no no
Table 3.3 – The roles of different χ(2) susceptibilities in the multilayer structure of
Au/Co/Ag/glass.
of metal/dielectric, such as interface 1 (glass/Ag) and interface 3 (Au/air).
Fit the SPP-induced magnetic SHG by numerical methods
θ Ez Ep
Ex
kinc kref
Figure 3.13 – Illustration for the electric field components in reflection.
The SPP-induced magnetic SHG in the trilayer of Au/Co/Ag above can be formu-
lated as
I(2ω)(±M ) = χzzz E1z E1z sin θ

+ Lnl (2ω) χxzx E3z E3x cos θ (3.23)
2
± χxxx E2x E2x cos θ
where χzzz is set as 1, and the other χ(2) components, such as χxzx and χxxx , are
complex factors relative to χzzz . sin θ and cos θ are respectively the projection of Ez
and Ex field components to the reflected electric field, as seen in Fig. 3.13.
Including all the sources at different interfaces, the angular dependence of magnetization-
induced SHG in the trilayer can be simulated numerically, as shown in Fig. 3.14. Com-
paring different sets of χ( 2)-components, it is observed that χxxx always gives better
angular consistency with the magnetic contrast, thus it could be the dominant magnetic
(odd) susceptibility component.
3.4.2 Reverse approach

Eq. 3.19 suggests that from the experimental magnetic contrast φ, one can estimate
the nonlinear SPP coupling efficiency Lnl (2ω) without the usage of effective medium
approximation described in the forward approach. Recalling that χxxx is considered
as the dominant magnetic second-order susceptibility, the non-resonant and resonant
magnetic SHG sources can be formulated as b = χxxx,1 Ex Ex and d = χxxx,2 Ex Ex ,
respectively. Thus, the new magnetic contrast φ can be re-written as:
50
20
SHG exp (+M)
SHG intensity (a.u.)
SHG exp (-M)

15
SHG sim (+M)
SHG sim (-M)
10
0
35 40 45 50 55
Angle of incidence (°)
0.15
ρexp
Magnetic contrast ρ (%)
0.1 ρsim
0.05
-0.05
-0.1
-0.15
35 40 45 50 55
Figure 3.14 – Forward approach: (a) Simulated curves (dashed lines) are fitted with
the experimental mSHG(±M ) (solid lines) measured for the structure (5 nm)Au/
(10 nm)Co/ (25 nm)Ag excited at fundamental wavelength 1200 nm. (b) Comparison
of the experimental (black) and simulated (red) magnetic contrasts. In the simulation,
only one dominant χ(2) -component is considered at each interface, referring to Eq. 3.23.
51
CONFIGURATION
∆I2ω
φ= p
I2ω (m = 0)
= 2ℜ{|b + Ld|eiϕ }
(3.24)
b
= 2ℜ{|( + L)d|eiϕ }
d
= 2ℜ{(ξ + L)χxxx,2 Ex Ex eiϕ }
where ξ is the complex ratio between the non-resonant and resonant magnetic suscep-
tibilities ξ = χxxx,1 /χxxx,2 .
1.5
1
Normalized φ =∆ I 2ω/sqrt(I 2ω) (a.u.)
0.5
-0.5
φ
-1
ℜ(L )
2 ω,exp
ℜ(L2 ω,eff )
1 πi
-1.5 F=ℜ[(L 2 ω +0.25e )*Ex E x ]
F under focus: conv(F,G)
G (FWHM=0.46°)
-2
35 40 45 50 55
Angle of incidence (deg)
Figure 3.15 – Reverse approach: fit the magnetic contrast φ to estimate the Lorentzian
function. φ = dI(±M
√
I
)
(black solid) is well fitted with F = ℜ{(b + Ld)} (red solid) in
the structure of (5 nm)Au/ (10 nm)Co/ (25 nm)Ag. The measurement was performed
at the fundamental wavelength 1200 nm.
This magnetic contrast φ is deduced experimentally by

I2ω (+M ) − I2ω (−M )
φ= p (3.25)
I2ω (+M ) + I2ω (−M )
Fig. 3.15 illustrates the magnetic contrast φ normalized to unity for the sample (5 nm)Au/
(10 nm)Co/ (25 nm)Ag. The experimental φ (black solid) is fitted by the numerical
contrast F calculated by Eq. 3.24 (red solid). The convolution of F with a Gaussian
function G (blue curve), which imitates the focusing condition in the experiments,
gives the blue dashed curve, a closer fit to the experimental φ (black solid).
From the experimental φ (black line), one can obtain the Lorentzian function of
the form
Γ
L(θ) =
(θ − θ0 ) + iΓ
52
3.5. DISPERSION OF LORENTZIAN FUNCTION
taking the resonant angle θ0 where the angular derivative δφ/δθ is highest. The slope
of the normalized φ gives us an estimation for the angular broadening of the Lorentzian
function Γ. Γ, estimated by this experimental approach, has the value of Γexp ≈ 1.1 deg,
which is very close to that calculated for the same structure Γcal ≈ 1.02 deg using
effective medium approximation.
The Lorentzian function Lexp (θ) = (θ−461.1◦ )+i1.1◦ can be substituted for Lnl (2ω) in
◦
the Eq. 3.23 in order to fit for the I2ω (±M ). The conventional magnetic contrast ρ can
then be deduced according to Eq. 2.50. The reverse approach is advantageous because
magnetic contrast is a stable quantity, which depends on the relative phase between
magnetic and non-magnetic components, and is not affected by the scattering of SHG.
Ideally, it permits to predict the nonlinear SPP coupling condition when the dielectric
function of a material consisted in the multilayer structure is unknown, an usual case
for ultra-thin layers.
3.4.3 Conclusion about two simulation approaches

In Fig. 3.15, the green solid line shows the real part of the Lorentzian function esti-
mated by the reverse approach, with the resonant angle θexp ≈ 46 deg and the angular
broadening Γexp ≈ 1.1 deg. In contrast, the magenta dashed line shows the real part
of the Lorentzian function calculated by the forward approach using effective medium
approximation, with θcal ≈ 45.3 deg and Γcal ≈ 1.02 deg. Comparing between the two
approaches, one sees that the angular broadening Γ does not differ much, but the res-
onant angle θ0 in the reverse approach experiences a noticeable shift towards larger
angles.
The forward approach helps to guess the dominant susceptibility components at dif-
ferent interfaces, i.e. non-magnetic χzxx or χxzx at Au/air and magnetic χxxx at Co/Au
interface. On the other hand, the reverse approach allows to estimate the nonlinear
SPP coupling efficiency L2ω directly from the experimental magnetic contrast, with-
out knowing the dielectric functions and thickness of individual layer. The forward
approach has been employed in a prior research [6] to simulate the magnetization-
induced SHG, while the reverse approach is introduced exclusively in this work.
3.5 Dispersion of Lorentzian function

From the effective dielectric function ǫef f (λ) estimated using effective medium approx-
imation, one can get the dispersion of the Lorentzian function
Γ(λ)
L(λ) =
θ − θ0 (λ) + iΓ(λ)
for the structure (5nm)Au/ (10nm)Co/ (25nm)Ag, as shown in the graph 3.16. It
is seen that the resonant angle θ0 does not vary a lot at the infrared spectral range.
In contrast, the angular broadening of Lorentzian function Γ increases for shorter
fundamental wavelength and peaks at 650 nm (SHG at 325 nm) due to the interband
transition threshold of silver (at ≈3.8 eV).
53
CONFIGURATION
50
45
40 θ
0
Γ
35
Angle (deg)
30
25
20
15
10
0
400 600 800 1000 1200 1400 1600 1800 2000
Fundamental wavelength (nm)
Figure 3.16 – Dispersion of the Lorentzian function is shown via θ0 (λ) (blue) and Γ(λ)
(red).
Even though the dispersion of magnetic contrast does not depend only on the
dispersion of Lorentzian function L but also on the fundamental field Ex (ω) according
to Eq. 3.24, the dispersion of the Lorentzian function is the main factor that affects
the SH magneto-optical effects. The experimental data of the conventional magnetic
contrast ρ are shown at different fundamental wavelengths in Fig. 3.17. It is clear
that shorter fundamental wavelengths cause angular broadening in the SH magnetic
contrast, in consistence with the increased Γ in Fig. 3.16.
3.6 Discussion
3.6.1 Bilayer structures
Now we return to the nonlinear magneto-plasmonics on sample (20nm)Au/ (30nm)Co/
sapphire, which has been discussed briefly in section 3.2.1. The experimental SH in-
tensities at opposite magnetizations are shown in Fig. 3.18a as green and black dots.
The measured SH magnetic contrast ρ is plotted in Fig. 3.18b as magenta dots. Us-
ing the reverse approach, one can obtain the experiment-extracted Lorentzian Lf it ,
shown as the violet dashed line in Fig. 3.18c. Applying Lf it in Eq. 3.23, the simulated
SHG(±M ) are deduced and plotted as the green and black solid lines in Fig. 3.18a.
The parameters used in the fitting are summarized in Table 3.4. The simulated curves
reproduce all the features of the angular dependence of the mSHG, such as the peak
at 35.4◦ , the dip at 36.1◦ and the background. The magnetic contrast ρ is given by
substituting the simulated SHG(±M ) into Eq. 2.50, and then plotted as the magenta
solid line in Fig. 3.18b. Except the small peak at 35.1◦ , the calculated contrast is in
54
3.6. DISCUSSION
30
Magnetic contrast ρ (%) 20
10
-10
780 nm
800 nm
-20
860 nm
920 nm
-30 1300 nm
1560 nm
-40
38 40 42 44 46 48 50
Figure 3.17 – The angular dependence of the conventional magnetic contrast ρ is broad-
ened at short fundamental wavelengths. The graph combines the data from Ref. [6]
(780-920 nm) and the data measured in this work (1300-1560 nm) for the sample of
(5nm)Au/ (10nm)Co/ (25nm)Ag.
Interface Non-resonant Resonant

Au/air (3) 0 χzzz,3 E3z E3z
Co/Au (2) χzzz,2 E2z E2z ±χxxx,2 E2x E2x
Co/sapphire (1) χxzx,1 E1z E1x 0
±χxxx,1 E1x E1x
Table 3.4 – Parameters used for the simulation of mSHG in bilayer Au/Co. E1 , E2
and E3 represent for the fundamental field E(ω) taken at the interface Co/sapphire
(Co side), Co/Au (Co side) and Au/air (Au side), respectively.
high agreement with the experimental data (magenta dots), espcially the double-dip
and the background contrast of 60%.
Fig. 3.18c shows the Lorentzian functions Llin (ω), Llin (2ω) and Lnl (2ω) calculated
based on the forward approach (Eq. 3.5 and 3.22). Each of them represents for a
different SPP momentum-matching condition. We can see that the experimentally
extracted Lorentzian Lf it (violet dashed line) is highly consistent with the calculated
Lnl (2ω) (yellow solid line). Importantly, the angles of the double zero-dips feature in
the magnetic contrast ρ (Fig. 3.18b) coincide with the resonant angles of the linear
Lorentzian at fundamental wavelength Llin (ω) and the second-harmonic Lorentzian
Lnl (2ω). These are the evidences for the destructive interference of the SPPs and the
magnetic second-harmonic field.
55
CONFIGURATION
(a) exp SHG (+M) sim SHG (+M)

exp SHG (-M) sim SHG (-M)
(b)
sim
exp
(c)
Figure 3.18 – Angular dependence for the bilayer (20 nm)Au/ (30 nm)Co excited
at the fundamental wavelength 1560 nm. (a) SHG intensity upon the reversal of
magnetization ±M (dots) and their corresponding simulations (solid lines) and (b)
transversal SH magnetic contrast ρ in (c) The Lorentzian functions Llin (ω), Llin (2ω)
and Lnl (2ω) are calculated based on the forward approach, and each of them represents
for a different SPP momentum-matching condition. Lf it was estimated based on the
reverse approach.
56
3.6. DISCUSSION
Interface Non-resonant Resonant

Effective 0 χzxx,2 E2x E2x
Co/Au/air (2) ±χxxx,2 E2x E2x
Ag/silica (1) χzzz,1 E1z E1z 0
Table 3.5 – Dominant SH field components assumed for the mSHG in trilayer Au/ Co/
Ag. E1 and E2 correspond to the fundamental field E(ω) taken at the bottom interface
of Ag/ silica (Ag side) and the effective top interface of Co/ Au/ air, respectively.
3.6.2 Trilayer structures

Similar simulation can be performed for the mSHG in the trilayer of Au/Co/Ag. Here
the sample (5 nm)Au/ (10 nm)Co/ (25 nm)Ag is taken as an example. Following the
reverse approach described in the previous section, the experimental Lorentzian Lf it
(blue dashed line) can be extracted from the magnetic contrast φ defined in Eq. 3.19
(black dots), as shown in Fig. 3.19a. Compared to the Lorentzian Lnl (2ω) (green solid
line) calculated based on the effective medium approximation, Lf it has similar angular
width, but its resonant angle shifts to 46◦ . The shift of 0.7◦ of the Lf it might result
from the error in the input refractive indices of metallic layers and substrate. In the
calculation, the natural corrosion of the metallic layers, especially cobalt, was not taken
into account. In fact, the cobalt layer in the trilayer is protected by only 3-5 nm-thick
gold layer, and is more vulnerable to the oxidation than that in the bilayer structure.
Thus, the experimentally extracted nonlinear Lorentzian function Lf it appears to be
a good assumption to fit the data when the quality of the structure is not clear.
Lf it Fig. 3.19. Fig. 3.19b shows the experimental SH emission SHGexp (±M ) (green
and black lines) and the simulated ones (magenta lines). In this simulation, only
two interfaces are considered: the effective top interface including Co/Au/air, and
the bottom interface Ag/silica. Unlike the case of bilayer, there is only one source
of magnetic SH emission considered in the trilayer, which is the top surface. The
dominant SH field components assumed for the simulation of the mSHG in Fig. 3.19b
are summarized in Table. 3.5. The figure shows agreement between the experimental
data and the calculation based on the reverse approach, as good as the fitting based on
the forward approach in Fig. 3.14. The conventional SH magnetic contrast ρ can also
be calculated from the simulated SHGsim (±M ), and is shown as the red solid curve
in Fig. 3.19c. The experimental contrast, presented as black dots, is well fitted by the
simulated one.
The high consistence between the numerical simulation and the experimental data
shows the validity of the reverse approach, which estimates the resonance of the non-
linear SPP(2ω) based on the new magnetic contrast φ = ∆I √ . Furthermore, it confirms
I
the fingerprint of the nonlinear SPP coupling efficiency, represented by the Lorentzian
function Lnl (2ω), on the magnetization-induced SHG.
As shown before in Eq. 3.24, the factors affecting the nonlinear magnetic contrast
in the plasmonic structures include the second-harmonic SPP excitation Lnl (2ω), the
fundamental field at magnetic interface Ex (ω) and the phase difference ϕ between even
and odd second-harmonic fields. In order to see the reason why the magnetic contrast
57
CONFIGURATION
(a)
(b)
(c)
Figure 3.19 – Angular dependence of the SH intensity and magnetic contrasts of trilayer
(5 nm)Au/(10 nm)Co/(25 nm)Ag excited at the fundamental wavelength 1200 nm.
(a) Lf it (blue dashed line) in the nonlinear Lorentzian function estimated based on
the reverse approach. Lnl (2ω) (green solid line) is the Lorentzian function calculated
based on the forward approach. The normalized magnetic contrast φ (black dots) is in
agreement with F applying Lf it to the Eq. 3.24. Similarly, Lf it can be used to fit the
SHG(±M) (b) and the conventional magnetic contrast ρ (c).
58
3.6. DISCUSSION
in the trilayer of Au/Co/Ag is stronger when the Ag layer is thinner, we will inspect
the dependence of the aforementioned factors on the Ag thickness. Fig. 3.20a shows the
Lorentzian function Lnl (2ω) calculated for the set of trilayer (5 nm)Au/ (10 nm)Co/
(15-25-35 nm)Ag, where only the thickness of Ag layer is varied. Fig. 3.20b shows
the product of Ex (ω)Ex (ω) calculated based the MTM for the three samples. One
can see in Fig. 3.20a that the variations in the resonant angle and the angular width
of the Lorentzian Lnl (2ω) in the three samples are negligible. In contrast, Fig. 3.20b
shows a strong dependence of the field amplitude on the thickness of Ag, the SPP
active layer. Thus we can conclude that the main reason behind the high magnetic
contrast of sample (5 nm)Au/ (10 nm)Co/ (15 nm)Ag is the strong concentration of
fundamental field component Ex (ω) at interface Co/Au, i.e. the field enhancement of
linear SPP(ω).
(a) (b)
Figure 3.20 – Real part of (a) the nonlinear Lorentzian Lnl (2ω) and (b) the product
Ex Ex at the magnetic interface Co/Au calculated for trilayer (5 nm)Au/ (10 nm)Co/
(15-25-35 nm)Ag.
3.6.3 SHG due to instantaneous change of damping constant

From the simulations of both the bilayer and the trilayer, one can see that χxxx is the
dominant magnetic SH susceptibility, which corresponds for the magnetic polarization
localized at the Co/Au interface. This cannot be simply explained by the model of
anharmonic oscillator in section 2.3 because the motion of electrons along metallic
surface driven by Ex (ω) should not encounter any potential wall7 . Thus, another
explanation is introduced for this case.
First, we consider the case when an external magnetic field is introduced and not
parallel to the initial velocity of electron B ∦ v. The oscillation of electrons in different
metals experience different damping constants due to the direction of the magnetic field
(±By ). Fig. 3.21 shows the illustration of a charge dragged by Ex along the interface
7
The model of anharmonic oscillator can explain for the case of electrons driven by Ez (ω), leading
to χzzz for example.
59
CONFIGURATION
between two metals. When the magnetic field By is applied, the trajectory r of the
charge is deflected due to the Lorentz force. The direction of the moving change
changes upon the reversal of the magnetic field.
r(+By)
Au z
Ex x
By
Co
r(-By)
Figure 3.21 – Illustration of a charge at the boundary between two conducting media Co
and Au. The trajectory of the moving charge is deflected towards different directions
when the external magnetic field is reversed.
Now we consider the electrons travelling across the interface between the ferro-
magnetic Co and noble metal Au. According to Eq. 2.31 and 2.35, the trajectory of
electrons at the fundamental frequency ω in Au and Co are
1q E(ω)
e−iωt + eiωt

zAu (t) = × 2
2 m ω + iγAu ω
(3.26)
1q E(ω)
e−iωt + eiωt

zCo (t) = × 2
2 m ω + iγCo ω
where γAu = 1/τAu and γCo = 1/τCo are damping contansts in Au and Co, respectively.
The trajectory of electrons in these two media at the excitation wavelength 1560 nm
(ω=0.8 eV) are shown in Fig. 3.22, assuming that the scattering time of electron in
Au τAu = 27 f s and in Co τCo = 17 f s [81]. The difference between the trajectories,
shown by the yellow curve, is very small but it modulates at the doubled frequency. It is
assumed that upon crossing the interface between Co and Au, an instantaneous change
in the damping constant results in a nonlinear optical susceptibility. If the electrons
are driven by the normal component of the incident field Ez , their trajectories should
be affected by the external magnetic field ±By in a similar way. However, at the
metal/ferromagnet interface, the coupled SPP can enhance Ex much stronger than Ez
(see Fig. 3.12). This is most likely the reason for the dominance of the magnetization-
induced SH polarization Px (2ω, ±M ) ∝ χxxx Ex (ω)Ex (ω) in the magneto-plasmonic
structures.
3.7 Conclusion
In this chapter, the SPP-induced second harmonic generation and magneto-optical
effects have been described using physical expressions. Based on that, two simula-
tion methods, forward and reverse approaches, have been proposed, and both have
60
3.7. CONCLUSION
×10 -19
1.5 ℜ[zAu (t)]
ℜ[zCo (t)]
1 10 2 |ℜ[zAu (t)]-ℜ[zCo (t)]|
0.5
z (m)
-0.5
-1
-1.5
0 2 4 6 8 10
Time (fs)
Figure 3.22 – The trajectory of electrons in the Au and in the Co bulk, assuming that
the scattering time of electron in Au τAu =27 fs and in Co τCo =17 fs[81]. The difference
between two trajectories has the doubled frequency.
demonstrated high consistencies with experimental signals. One important part in

this chapter is the discussion about the excitation condition of different SPP modes,
their coupling efficiency L(θ) and dispersion. The models successfully explain the effect
of linear and nonlinear SPP modes on second-harmonic magneto-optical responses in
different multilayer structures, the trilayer Au/Co/Ag and the bilayer Au/Co.
It was also demonstrated experimentally that it is possible to tune the nonlinear
magnetic signals by varying the thickness of the plasmonic layer, which is a way to
tailor the SPP field. It represents the first step towards the new generation of nonlinear
magneto-plasmonic devices for light modulation at the nano-scale. Besides, the new
observation of magneto-plasmonics on bilayer Au/Co opens many opportunities for
our research. Since the bilayer metal/ferromagneti structure has been shown to be a
“perfect” structure for ultrafast acousto-plasmonics [7] and exchange-coupled magnon-
ics [31], the combination of ultrafast magnonics and plasmonics can be expected. Their
outstanding heat conduction properties allow the nonlinear (magnetic SHG and fem-
tosecond time-resolved pump-probe) optical measurements to be performed at high
radiation powers, resulting in the significant increase of spatially transient electron
and lattice temperature in ferromagnetic cobalt.
61
Chapter 4
Nonlinear magneto-optics in
diffraction anomalies
We have discussed in chapter 2 that propagating surface plasmon polaritons can be

excited in metallic gratings. In this geometry, the discrepancy between the incident
wave-vector and the SPP wave-vector is compensated by a multiples of grating wave-
vectors G. The advanced technologies in nano-fabrications now allow to go further.
Periodic structures with sub-wavelength periodicity, so-called optical meta-surfaces,
can be produced with high precision and regularity. Besides the delocalized surface
modes, surrounding the nanosized metallic objects is the localized surface plasmon
resonances (LSPRs) [3, 56, 114, 115]. Under certain conditions, the arrays of nanos-
tructures permit the hybridization of the LSPRs and the Wood’s anomaly across the
surface, causing spectrally narrow resonances, commonly referred to as surface lattice
resonances (SLRs) [46, 47, 52, 53]. This chapter is devoted to present the experimental
observation of the second-harmonic nonlinear Wood’s anomaly in nickel nanodimers
arrays.
4.1 Nonlinear Wood’s anomaly

4.1.1 Diffraction on grating
A conventional diffraction grating requires the grating periodicity Λ to be larger than
a half of the optical free space wavelength λ0 (Λ > λ0 /2) to ensure the presence
of at least one diffraction order. On Fig. 4.1 are illustrated the cases of diffraction
on a grating in the linear and nonlinear-optical regimes. For gratings with opaque
substrates, the diffraction orders can be observed in reflection as in Fig. 4.1. For
transparent substrates, the diffraction can also occur in transmission, but at smaller
diffraction angles due to the dispersion in the substrate. Fig. 4.1a-b show the case of
a regular grating with the periodicity Λ ≥ λ/2 irradiated at an angle of incidence θ in
the free space and reciprocal space, respectively. The phase-matching conditions for
diffraction can be conveniently illustrated in the reciprocal space (Fig. 4.1b).
The emergence and direction of a diffraction order mlin rely on the phase-matching
63
CHAPTER 4. NONLINEAR MAGNETO-OPTICS IN DIFFRACTION
ANOMALIES
I (0) G G G G
I inc
ω
ω
I (0)
2ω (-1)
I (-1) k 2ω (0) (0)
2ω k 2ω k 2ω
I (-1)
ω G
θ (-2)
(-1)
k 2ω
I (-2)
2ω I (+1) k 2ω inc
kω (0)
inc
kω (0)
θd
(+1)
2ω kω kω
(+1)
(-1) k 2ω
kω
-2k 0 -k 0
k 0
2k 0 -2k 0 -k 0
k 0
2k 0
(a) (b) (c)
Figure 4.1 – (a) Linear (red) and nonlinear (SH, blue) diffraction orders in reflection on
a grating with a periodicity Λ > λ/2. When the angle of incidence θ of the fundamental
radiation Iωinc increases in the counter-clockwise direction, all diffraction orders rotate
clock-wise. (b-c) In the reciprocal space, linear and nonlinear diffraction orders are
shifted by a multiple of the grating wavevector G = 2π/Λ. (b) For G ≤ 2k0 or
Λ ≥ λ/2, both linear and second-order nonlinear diffraction orders can be observed.
(c) When 2k0 < G < 4k0 or λ/4 < Λ < λ/2, besides the specular m = 0, only the
second-order nonlinear diffraction order with mnl = −1 appears.
condition between the wavevectors, as following:
kxinc (ω) + mlin G = kxd (ω) (4.1)
where kxinc = k0 sin θ and kxd = k0 sin θd are in-plane (x-) components of the incident
wavevector k inc and the diffracted wavevector k d , respectively. Here, k0 = 2π/λ0 , θ is
the angle of incidence, θd is the diffracted angle of the corresponding diffraction order
(see Fig. 4.1), and G = 2π/Λ is the wavevector of the grating. mlin is an integer,
which represents the linear diffraction order. A similar condition can be written for
the nonlinear (SH) diffraction [116]:
2kxinc (ω) + mnl G = kxd (2ω) (4.2)
where another integer mnl 6= mlin is introduced to indicate the nonlinear diffraction
order.
These phase-matching conditions require the linear wavevector kω (red) and the
SH wavevectors k2ω (blue) to rotate within the semicircles with the radii k0 and 2k0 ,
respectively. The linear and nonlinear diffraction orders are shifted by a multiple of
the grating wavevector G. Comparing Eq. 4.1 and Eq. 4.2, one can see that a linear
diffraction order mlin is always accompanied by a SH diffraction order mnl = 2mlin
once kxd (2ω) = 2kxd (ω). Indeed, the parallel wavevectors of the linear diffraction order
mlin = −1 and the SH diffraction order mnl = −2 in Fig. 4.1b ensure that these
diffraction orders can be excited at the same angle of incidence θ and observed at the
same diffraction angle θd .
In the nonlinear regime, one can extend G to be larger than 2k0 (Λ < λ/2) and still
observe the presence of the diffraction orders, which is the case illustrated in Fig. 4.1c.
Besides the specular m = 0, there is no other linear diffraction order, but the nonlinear
diffraction order mnl = −1 can be observed. Thus it becomes possible to combine the
64
4.2. EXPERIMENTAL METHODS
meta-surfaces, which are characterized by a much smaller periodicity than the free
space wavelength Λ ≪ λ [50, 57, 58], and the nonlinear diffraction. Indeed, the same
periodic structure can serve as a meta-surface for the fundamental light at λ, and as
a regular grating for the harmonic radiations with wavelengths λ/n (n > 1). Overall,
the condition for this transition regime λ ≫ Λ > λ/2n is expected to be easily fulfilled
in the visible and near-infrared optical range.
4.1.2 Wood’s anomaly

“Wood’s anomaly” is defined as the redistribution of the light energy between different
diffracted beams upon the emergence of a new diffraction order propagating along the
grating surface [1]. In other words, it occurs when one of the diffraction orders is paral-
lel to the grating surface, i.e. the in-plane component of the diffraction wavevector kxd
is highest. In the case of transparent substrate, there would be two Wood’s anomalies
of the same order; one travels along the interface between grating and open space (air
for simplicity), and the other occurs at the interface between grating and substrate.
Even though SPP can be excited where there is high relative content of the metal (de-
spite high optical losses in ferromagnetic transition metals), the large distance between
the nanostructures can prohibit the propagation of the mode. Thus, using the terms
“propagating SPP" is not rigorous and besides the historical reason, “Wood’s anomaly"
would be a better term to portray the surface mode excited in the periodic structures,
which depletes the energy from the other diffraction orders upon its emergence.
The excitation of this surface mode should fulfill the following phase-matching
condition:
2π
kx (ω) + mlin = ks (ω) (4.3)
Λ
in the linear regime. Here ks (ω) is the surface mode wavevector, corresponding to the
linear Wood’s anomaly with order mlin . One can recognize that Eq. 4.3 is similar to
the phase-matching condition of SPP on the grating surface described in Eq. 2.60 in
section 2.4.2, even though SPP is not supposed to be excited in this structure. This
comes from the fact that SPP is itself a type of electromagnetic surface waves [22] that
can be excited by the momentum conservation. Following the highly similar formalism
of the nonlinear SPP excitation in chapter 3 and the previous research [6, 99, 101, 116],
we can outline the nonlinear excitation condition for the Wood’s anomaly with order
mnl :
2π
2kx (ω) + mnl = ks (2ω). (4.4)
Λ
In this work, the nonlinear-optical technique of magneto-induced second harmonic
(SH) generation is employed to study the Wood’s anomaly, i.e. the intrinsic property
of an optical grating in the nonlinear transition regime.
4.2 Experimental methods

The angle-dependent SH spectroscopy with tunable femtosecond laser source was per-
formed on a rectangular array of nickel dimers featuring a nanoscale gap, similar to
65
ANOMALIES
that studied in Ref. [46]. The purpose of this study is to observe the nonlinear Wood’s
anomaly (Fig. 4.1c and Fig. 4.2a) and compare its nonlinear-optical properties with
the general case where linear and nonlinear Wood’s anomalies coexist (Fig. 4.1b).
The sample under investigation was designed and built by our collaborator at De-
partement of Physics and Astronomy, Uppsala University, Sweden. It is an array of
nickel nanodimers deposited on a Si substrate, shown in Fig. 4.2. Good heat conduc-
tivity of Si (≈100 W/m·K at 100 ℃ [117]) ensures the sample damage resilience at high
peak optical fluences required for sizeable nonlinear-optical effects. The nickel film was
DC sputtered directly on the Si[100] substrate, and then the structure was patterned
using electron beam lithography. The nanodisks (average diameter 145 nm, height
60 nm) are grouped into dimer cells with the air gap of about 15 − 60 nm (in average
35 nm), as shown in Fig. 4.2c. Nickel structure is covered by a layer of natural oxide
N iO, which is about 1-2 nm in thickness. The two axes of the array are denoted as
pL (longitudinal) and pT (transverse), illustrated in Fig. 4.2 b. Along pL , the average
distance between two adjacent dimer cells is ΛL ≈ 445 nm. Along pT , the periodicity
is ΛT ≈ 265 nm (see Fig. 4.2c).
(a) Ez
E Rω m=0
(c)
Ex ΛL ≈445 nm
I 2ω
z
x θ ΛL
Ni h=60 nm
By
I (-1) I (-
-2) Si ΛT
ω 2ω
≈265 nm
(b)
pT
120 nm 35 nm
d=145 nm
pL
Figure 4.2 – (a) Schematic for the co-excitation of linear and nonlinear Wood’s anoma-
(−1) (−2)
lies, denoted as Iω and I2ω respectively, in an array of nickel nanodimers. An
external magnetic field is applied perpendicular to the plane of incidence. (b) Scan-
ning electron microscopy image of the nickel nanodimers. The scale bar is 600 nm
long. (c) Dimensions of the two-dimensional nanodisks array.
Assuming an isotropic interface and the p-polarized incident fundamental radia-

tion, six nonlinear susceptibility components remain non-zero [37, 41], as summarized
in Table. 3.2 in chapter 3. The SH fields generated through these components give
rise to a far-field magnetization-dependent SH intensity I2ω (±M ). Originating in the
interference of the odd and even SH sources, the variations of the SH intensity upon
reversal of the in-plane magnetization My is conventionally quantified by the nonlinear
magnetic contrast ρ defined in Eq. 2.50, similar to the study of the nonlinear transversal
MOKE in the hybrid thin films (chapter 3).
To study Wood’s anomalies in these nonlinear processes, we investigate the de-
pendence of the far-field magnetization-induced SH intensity I2ω (±M ) on the in-plane
66
4.2. EXPERIMENTAL METHODS
fundamental wavevector kx (ω), which is related to the angle of incidence θ through

kx = k0 sin(θ). According to the definition, Wood’s anomalies are the “uneven distri-
bution of light"[1] between different diffraction orders. Putting the detectors at the
output of the specular (zero diffraction order m = 0), it is expected to observe a drop
in the linear reflectivity Rω when a Wood’s anomaly is excited at the fundamental
frequency, and a drop in the SHG intensity I2ω when the diffraction anomaly is excited
at the double frequency.
Figure 4.3 – The top-view of the experimental setup for nonlinear Wood’s anomaly in
the periodic structures.
The experiment was conducted on a goniometric platform which allows measuring

the reflected signals at different angles of incidence, as illustrated in Fig. 4.3. Basically,
it is similar to the setup utilized in chapter 3, but without the prism. The fundamental
radiation is produced by a Ti:Sa laser (MaiTai HP from Spectra Physics), tunable
in the spectral range between 690 and 1040 nm, with a pulse duration of ∼100 fs,
repetition rate of 80 MHz. The p-polarized (TM) fundamental radiation was loosely
focused on the sample surface (spot diameter 80 µm). The linear reflectivity R was
measured using a biased silicon photodiode (Newport). The SH output was spectrally
separated by a color BG-39 filter (Schott), collected by a lens and registered with
a photomultiplier tube (Hamamatsu) operating in the photon counting regime. The
external transverse magnetic field By ≥100 mT, was applied perpendicular to the plane
of incidence, saturating the magnetization of the nanodimers.
The measurements were performed for two azimuthal angles of the sample (0◦
and 90◦ ) corresponding to the two orientations of the spatial periodicity of the array
with respect to the in-plane fundamental electric field and the magnetization direction
(Ex k pT and Ex k pL ). In Ref. [56], it has been demonstrated that the electric field
is highly localized in the dimer gap when Ex is oriented along longitudinal dimer axis
67
ANOMALIES
(Ex k pL ). When Ex k pT , the plasmon mode in the gap is strongly suppressed.
4.3 Results and discussions

4.3.1 Nonlinear Wood’s anomaly and MOKE in nanodimers
The specular (zero diffraction order m = 0) linear reflectivity Rω , the magnetization-
dependent SH intensity I2ω (±M ), and the nonlinear magnetic contrast ρ were measured
as function of the angle of incidence θ. Fig. 4.4a and Fig. 4.4b show the experimen-
tal data obtained at the fundamental wavelength λ0 = 820 nm for the two different
configurations Ex k pT and Ex k pL , respectively.
(a) (b)
8 6 8 8
I (+M) ρ (%) I (+M) ρ (%)
2ω 2ω
7 I (-M) R (arb. u.) 5 7 I (-M) R (arb. u.) 7
2ω ω 2ω ω
SHG magnetic contrast ρ (%)
SHG magnetic contrast ρ (%)

SHG intensity I2ω (kcounts/s)
SHG intensity I2ω (kcounts/s)
6 4 6 6
5 3 5 5
4 2 4 4
3 1 3 3
2 0 2 2
1 -1 1 1
0 -2 0 0
30 40 50 60 70 30 40 50 60 70
Angle of incidence θ (deg) Angle of incidence θ (deg)
Figure 4.4 – The angular dependence of the linear reflectivity Rω (dashed black), the
SH intensity I2ω (±M ) for the two opposite directions of magnetization (solid red and
blue), and the SHG magnetic contrast ρ (solid green) at the fundamental wavelength
of 820 nm, measured for the case of (a) Ex k pT and (b) Ex k pL .
In these plots, the high values of Rω beyond 60◦ − 70◦ are a result of the normal-
ization, where the raw data were divided by relatively small values near the Brewster’s
angle. Leaving aside this artificial effect, no significant drop of Rω was observed in the
case of Ex k pT (Fig. 4.4a). It means that the linear Wood’s anomalies are not excited
in this configuration, which is consistent with the case illustrated in Fig. 4.1c, where
ΛT < λ0 /2 and no linear diffraction is allowed except the zero order. In contrast, in
the Ex k pL geometry with the spatial periodicity ΛL ≈ 445 nm (ΛL > λ0 /2), a dip
of Rω was observed at 50◦ (Fig. 4.4b) as a fingerprint of the linear Wood’s anomaly
excitation at around this angle.
In the nonlinear regime, the angular dependence of the SH intensity shows a rec-
ognizable minimum in both configurations, Ex k pT and Ex k pL . The SH output is
more than 103 counts/s, which is nearly an order of magnitude higher than the SH
68
4.3. RESULTS AND DISCUSSIONS
signal from a bare silicon substrate (≈ 600 counts/s). In fact, the SH output from
these arrays are comparable to the SPP-enhanced SH yields observed in the thin films
in chapter 3 and in prior research [6, 7]. The pronounced characteristic drops of the
SH signals, appearing at 32◦ for Ex k pT in Fig. 4.4a and 53.5◦ for Ex k pL in Fig. 4.4b,
originate in the destructive interference of the SH fields. In thin continuous metal
films, the SH peak is usually observed at the same angle of incidence as the dip in
the linear reflectivity. And so, the most efficient SPP excitation at the fundamental
frequency ω is accompanied by the most prominent enhancement of the electric field
Eω .
A deviation from this highly intuitive picture, manifesting in the experiment as a
noticeable angular shift of the SH intensity peak, is indicative of a highly lossy plasmon-
polariton mode [100, 102]. However, nonlinear excitation of the Wood’s anomaly in
nanodimer arrays can enable SH enhancement even in the absence of the resonance in
the linear reflectivity. Indeed, when the linear Wood’s anomaly is excited in the Ex k pL
geometry, the first SHG peak appears at 48.5◦ , deviating from the reflectivity minimum
at 50◦ . Furthermore, in the case of Ex k pT the SH enhancement in the vicinity of
θ = 30◦ occurs in the absence of any noticeable features in the linear reflectivity at
all. The discrepancies between the linear reflectivity and the SHG thereby illustrate a
striking difference between the nonlinear properties of the SPP and Wood’s anomaly
modes.
Sizable angular variations of the magnetic contrast ρ in SH are seen for both orien-
tations of the sample. The characteristic magnitude of the resonant variation is around
5 % (Fig. 4.4), one-order of magnitude higher than the usual linear transverse MOKE
signal generated in magneto-plasmonic thin films[3]. A drop in the angular dependence
of the nonlinear magnetic contrast appears near the angle corresponding to the first
SH output peak, namely 30◦ for Ex k pT and 48.5◦ for Ex k pL .
4.3.2 Spectral analysis

Further insights in the nature and nonlinear properties of the Wood’s anomalies can be
obtained from the spectral measurements. To that end, we performed similar angular
measurements of the linear reflectivity and magnetization-dependent SH output at
various fundamental wavelengths. As a result, we observed a pronounced wavelength-
dependence of the angular features in the linear reflectivity as well as SH intensity and
magnetic contrast. The data for the two sample geometries are summarized in Fig. 4.5.
In Fig. 4.5a-b, the linear reflectivities were normalized to the reflectivity in the
bare silicon substrate. Within the wavelength range of the 800 − 900 nm and angles of
incidence 25◦ −75◦ , the angle of the reflectivity minimum increases with the wavelength.
The drop of reflectivity is about 7% for all wavelengths in the case of Ex k pL . In the
Ex k pT configuration, the reduced spatial periodicity ΛT = 265 nm denies linear
access to the surface modes, and thus no resonant variations of the reflectivity were
registered.
A considerably richer picture can be recovered from the SH angular spectra (Fig. 4.5c-
d). The dependence of SH output on the angle of incidence and wavelength in the
Ex k pT geometry exhibits a strong SHG peak at θ ≈ 61.5◦ , independent on the wave-
69
ANOMALIES
1
Normalized R (arb. u.)
Normalized R (arb. u.)

1
0.9
0.9
7%
0.8 800 nm
ΛT = 265 nm 800 nm 0.8
820 nm 820 nm
0.7 840 nm 0.7 ΛL = 445 nm 840 nm
860 nm 860 nm
0.6 Ex 880 nm
880 nm 0.6 Ex
900 nm 900 nm
0.5 0.5
30 40 50 60 70 30 40 50 60 70
(a) (b)
8
8 SHG intensity (kcounts/s)
SHG intensity (kcounts/s)
400 nm 400 nm
410 nm 410 nm
420 nm 6 420 nm
6
430 nm 430 nm
440 nm 440 nm
4 450 nm 4 450 nm
460 nm 460 nm
2 2
0 0
30 40 50 60 70 30 40 50 60 70
(c) (d)
1.2
Normalized SHG (arb. u.)
1.2
Normalized SHG (arb. u.)
400 nm 70%
410 nm
1 1
420 nm
430 nm
0.8 0.8
440 nm
70% 450 nm 400 nm
0.6 0.6 410 nm
460 nm
420 nm
0.4 0.4 430 nm
440 nm
0.2 0.2 450 nm
460 nm
0 0
30 40 50 60 30 40 50 60 70

(e) (f)
Figure 4.5 – The left column (a, c, e) shows the analysis for Ex k pT . The right
column (b, d, f) presents the data for the case of Ex k pL . (a, b) Angular dependence
of the linear reflectivity at different wavelengths. (c, d) Angular dependence of SHG
intensity at different SHG wavelengths. The black dashed line illustrates the empirical
guideline for normalization of SHG in panels e-f. Small arrows mark the SHG dips.
(e, f) Normalization of SHG intensities at different wavelengths shows the distribution
of SHG between different diffraction orders. The empty dots show the Wood’s angles
at each wavelength.
70
length. In contrast to that, the small-angle SH peak shifts towards larger angles of
incidence for longer fundamental wavelengths. A similar dispersive shift of the SH
features is observed in the other geometry (Ex k pL ), in that case for larger angles.
This dissimilar behaviour of the two geometries is again consistent with the unequal
spatial periodicity along the two directions. It is worth mentioning that, in the Ex k pL
geometry, the LSPR can be excited, characterized with an enhancement of the electric
field concentrated within the dimer gap. This LSPR excitation on the SH output can
be speculated to skew the wavelength dependence at a fixed angle of incidence θ. How-
ever, as seen in Fig. 4.5d, it is difficult to disentangle the impact of LSPR from other
contributions. Comparing Fig. 4.5c and d, when the LSPR is respectively disable and
enable, it is unlikely that the qualitative shapes of the curves in Fig. 4.5d are strongly
modified by the LSPR mode. Furthermore, the nickel content (the filling fraction of
the surface) is 29%, sufficiently low so that propagating surface plasmon-polaritons
cannot be efficiently excited.
The transfer of SH intensity between the nonlinear diffracted orders and the spec-
ular m = 0 is illustrated in Figs. 4.5e-f, where the SH outputs were normalized to an
empirical model (∝ sin4 θ) for the angular dependence of SHG at the thin film surface
(the black dashed lines in Fig. 4.5c-d). It shows that the nonlinear Wood’s anomaly
light coupling is extremely efficient. In both geometries, about 70% of the optical en-
ergy at the double frequency is dissipated at large angles of incidence for a multitude
of fundamental wavelengths. Notably, the percentage of the energy pumped into the
second-order mnl = −2 in the configuration Ex k pL is not consistent with the 7% drop
of the reflectivity Rω measured for the linear diffraction anomaly (see Fig. 4.5b). Thus,
our results evidence the“uneven distribution of light" between SH diffraction orders,
extending the Wood’s picture onto the nonlinear-optical domain and demonstrating
the immense capabilities and sensitivity of SH spectroscopy for these effects. It is
worth noticing that the 70% drop of the SHG intensity is almost identical for the case
of purely nonlinear case (Ex k pT in Fig. 4.5e) and the case of mixed linear-nonlinear
Wood’s anomaly (Ex k pL in Fig. 4.5f). Based on this observation we speculate that
in the mixed case the nonlinear Wood’s anomaly dominates.
4.3.3 Specifying the order of Wood’s anomaly

From Eq. 4.3, one can derive:
λ0
sin θ + mlin = nef f (ω) (4.5)
Λ
where the wavevector of the surface mode ks (ω) was represented as ks (ω) = k0 nef f (ω),
with nef f (ω) the effective refractive index of the mode at the fundamental frequency.
nef f (ω) can be understood as the weighted refractive index of the metal and the di-
electric (air) at both sides of the interface, which is sampled by the evanescent field
of the surface waves [22]. The nonlinear phase-matching condition in Eq. 4.4 can be
rewritten as:
mnl λ0
sin θ + = nef f (2ω) (4.6)
2 Λ
71
ANOMALIES
From Eq. 4.5-4.6 above, the sine of the resonant angle of incidence θ is linearly
proportional to the normalized fundamental wavelength λ/Λ, where the slope is the
diffraction order and the offset is the effective refractive index. In Fig. 4.6, the sine of
the angles, where the pronounced intensity dips in the linear reflectivity Rω and the
SHG I2ω are observed at each fundamental wavelength, are marked by empty dots.
In Fig. 4.6, three groups of data can be seen. The dips in the linear reflectivity R
observed in the Ex k pL geometry (red diamond dots) fit with the dispersion curve of
the linear Wood’s anomaly with order mlin = −1 and nef f (ω) ≈ 1.07 (red solid line).
Occurring at the same condition, the SHG minima in the Ex k pL geometry (blue
empty circles) correspond to the second-order nonlinear anomaly with order mnl = −2
and nef f (2ω) ≈ 1.04 (blue solid line). The co-excitation of the linear Wood’s anomaly
with order mlin = −1 and the second-harmonic Wood’s anomaly with order mnl = −2
is the direct result of the nonlinear optical diffraction discussed in section 4.1. The
small deviation in nef f indicates negligibly small dispersion of the surface mode.
In the Ex k pT configuration, there was no pronounced dip in the linear reflectivity,
as expected for the case of ΛT < λ0 /2, with the fundamental wavelength λ0 in the range
of 800-920 nm. The dips in the angular dependence of SHG (magenta empty circles)
conform with the nonlinear Wood’s anomaly with order mnl = −1 and nef f (2ω) ≈ 1.01
(green solid line). The lower nef f (2ω) might come from the fact that there is almost
no highly concentrated LSPR mode involved in this configuration.
Following the Eq. 4.4, it is expected that the Ex k pL configuration can support
another second-order nonlinear Wood’s anomaly with mnl = +1, which is plotted as
the black dashed line in Fig. 4.6a. However, according to our simulation, this mode
can only be excited at small angle of incidence (θ ≤ 8◦ ), out of the angular range in
the registered setup (20 − 75◦ ).
Fig. 4.6b-c summarize the SH data in Fig. 4.5c-d into false color two-dimensional
maps, where the smoothed wavelength dependence takes into account the spectral
width of the fundamental radiation (on the order of 5 − 7 nm). In both configurations,
the minimum of the specular SH output (registered in the zero diffraction order) is well
correlated with the calculated dispersion of the nonlinear Wood’s anomalies based on
Eq. 4.4 (white lines). We can recall that the drastic difference between two geometries is
that linear excitation of Wood’s anomaly and LSPR with the electric field enhancement
in the dimer cell are together active only when Ex k pL . The similarity of the SH
angular spectra obtained in the two orthogonal geometries with unequal linear optical
properties strongly corroborates the dominant role of the nonlinear anomaly excitation.
4.3.4 Impact of Wood’s anomaly on nonlinear MOKE

Having identified the modes observed in the linear and SH spectra as well as their
dispersion, we now turn to the resonant magneto-optical effects in SH. The fingerprint
of the Wood’s anomalies on the nonlinear magneto-optical signals is shown in Fig. 4.7.
There, the drop of the magnetic contrast ρ (dark area) is well correlated with the
calculated dispersion of the nonlinear diffraction anomalies (white solid lines), which
are mnl = −1 for Ex k pT (Fig. 4.7a) and mnl = −2 for Ex k pL (Fig. 4.7b). This
behaviour is similarly observed in both configurations, regardless of the excitation of
72
SHG (kcounts/s)

1 70
5
mnl=-2 60
4
0.8 50 3
mlin=-1
40 mnl =-1 2
(E || p ) 1
0.6 x L 30
mnl =-1 400 410 420 430 440 450 460

sin(θ)
(E x || p T) SHG wavelength (nm)

0.4
(b)
SHG (kcounts/s)
0.2 R minimum (E x || p L) Angle of incidence (deg)
70
5
mnl=+1
SHG minimum (E || p ) 60
0 x L 4
SHG minimum (Ex || p T) 50 3
mnl=-2 2
-0.2 40
1.5 2 2.5 3 3.5 30

1
Wavelength per period λ/Λ 400 410 420 430 440 450 460
SHG wavelength (nm)

(a) (c)
Figure 4.6 – (a) The lines show the theoretical dispersion of the Wood’s anomalies
excited in the two-dimensional nanoarrays. The empty dots show the sine of the angle
where the angular spectrum of linear reflectivity R(ω) and average SH output I2ω
reach minimum in corresponding configuration. (b) In Ex k pT configuration, the first-
ordered nonlinear diffraction anomaly mnl = −1 (white line) can be excited, matched
with the dip (dark region) in the angular dependence of SHG. (c) In the case when
Ex k pL , the SHG minima (dark region) fit in the nonlinear Wood’s anomaly with
mnl = −2 (white line).
73
ANOMALIES
linear anomalies and LSPR field enhancement within the dimer. In particular, since
both linear Wood’s anomaly excitation and LSPR field enhancement are inactive when
Ex k pT , the angular dependence of the SH magnetic contrast is attributed to the
nonlinear Wood’s anomalies.
While the linear excitation of Wood’s anomaly has low or almost no impact on
the nonlinear MOKE, the nonlinear mechanism for the surface mode excitation in fer-
romagnetic nanostructured arrays has significantly larger potential for the variations
of nonlinear magneto-optical signals. Interestingly, this result is inherently consistent
with our previous findings in the case of nonlinear magneto-SPP excitation in thin
metal films [6, 7]. From the nonlinear-optical point of view, the similarity between the
two situations is related to the strong variations of the electric field at the double fre-
quency 2ω, as opposed to the fundamental field E(ω). The generality of this effect has
a potential of becoming a very interesting and relevant question of nonlinear photonics
and requires further investigation.
mnl=-2
mnl =-1
(a) (b)
Figure 4.7 – The false-color images of the SH magnetic contrast for (a) Ex k pT ,
where the white line shows the theoretical dispersion for the second-harmonic anomaly
mnl = −1, and (b) Ex k pL , where the white line marks the dispersion of mnl = −2.
4.4 Conclusion
In this chapter, the impact of Wood’s anomalies on the SH output and nonlinear
MOKE in the two-dimensional arrays of nickel nanodimers have been investigated.
The angular spectra of the linear reflectivity and the magnetic SH intensity I2ω (±M )
were measured for a series of fundamental wavelengths in the two sample orientations.
A detailed comparison of these two sets of data highlighted the relevant linear and
nonlinear mechanisms of the excitation of the Wood’s anomaly.
The results suggest that the linear Wood’s anomaly has little influence on the SH
(0)
yield. The decrease of specular SHG intensity I2ω upon the emergence of the new
mnl = −1 or mnl = −2 diffraction orders in the nonlinear regime is 70%, i.e. an order
of magnitude larger than 7% reflectivity variations in its linear counterpart. The SHG
magnetic contrast also shows a fingerprint of the nonlinear diffraction anomaly, which
74
4.4. CONCLUSION
is the main factor contributing to the ∼ 5% angular variation of the SHG magnetic
contrast ρ around the Wood’s anomaly. Nonlinear magneto-optical effects can thus
be utilized as a measure of the efficiency of diffraction gratings in periodic arrays of
magnetic nanoparticles.
In the investigated array of nickel nanodimers, the large difference in the lattice
period between transversal and longitudinal configurations determines the character of
Wood’s anomalies. In future, possible contributions of LSPRs in nanodimers resulting
in the excitation of SLRs upon interference with Wood’s anomalies might be identi-
fied in similar experiments using structures with identical longitudinal and transverse
lattice periods. In general, depending on the grating constant, one can excite the
diffraction anomalies in the fundamental and (second-order, third-order, etc.) nonlin-
ear regimes, which emerge either in the air or substrate interface. This property allows
to select the preferred diffraction order and combine it with the optical meta-surfaces
by geometrical design.
75
Chapter 5
Femtosecond time-resolved nonlinear

magneto-plasmonic cross-correlation
measurements
Chapter 3 and 4 describe the coupling of the magnetization-induced second-harmonic

generation (mSHG) with the nonlinear surface waves in the static regime, which is
the integration over time of the phenomenon. In this chapter, the dynamics of the
nonlinear magneto-plasmonics will be explored. The field enhancement of SPP makes
it a sensitive probe for the ultrafast phenomena occurring within the optical skin
depth of metallic surfaces[112, 118, 119]. The temporal evolution of the SPP field can
be recorded by interferometry[120, 121]. Taking advantage of the SPP propagation
length along the metallic interface, the plasmonic interferometry can be performed
directly on the metallic surface [30, 113], and is proved to be magnetization-sensitive
when a magnetic layer is introduced within the field skin depth[12, 110]. In this work,
the auto-correlation and cross-correlation of the mSHG pulse are performed on the
Kretschmann geometry, where the SPPs can be coupled in both the linear and the
nonlinear regimes.
5.1 Fluence dependence of the mSHG

The first section is devoted to the experimental observations of the fluence dependence
of the nonlinear magneto-plasmonics on Kretschmann geometry. The setup is the same
as in Fig. 3.3 in the chapter 3, the only variable is the average power of the radiation
coming from a femtosecond laser source (centered wavelength 1560 nm, pulse duration
sub-120 fs, repitition rate 80 MHz). The incident power is tuned by a set consisting
of a half-wave plate and an analyzer. Fig. 5.1 shows the angular dependence of the
mSHG intensity and the magnetic contrast for the bilayer of (20nm)Au/ (35nm)Co
(Fig. 5.1a) and the trilayer of (5nm)Au/ (10nm)Co/ (15nm)Ag (Fig. 5.1b) using the
p-polarized fundamental excitation at wavelength 1560 nm. When the power of the
incident light increases, remaining below the damage threshold of the sample, the in-
tensity of the mSHG increases quadratically with the pump power, as illustrated in
77
CHAPTER 5. FEMTOSECOND TIME-RESOLVED NONLINEAR
MAGNETO-PLASMONIC CROSS-CORRELATION MEASUREMENTS
Fig. 3.5. However, it has been observed that the nonlinear magnetic contrast ρ, de-
fined in Eq. 2.50, reduces gradually as the pump power increases. For the sample of
(20nm)Au/ (35nm)Co bilayer deposited on sapphire, the reduction of the SH magnetic
contrast occurs not only at the angles where linear and nonlinear SPPs are excited
but also at off-resonant angles. In contrast, for the trilayer of (5nm)Au/ (10nm)Co/
(15nm)Ag deposited on fused-silica substrate, the reduction of the SH magnetic con-
trast is only observable around the angle of 46.6◦ , where the nonlinear coupled SPP(2ω)
takes effects (refer to the section 3.6).
p Pol, 35Co/20Au 4 p Pol, 5Au/10Co/15Ag

x 10
5000 3.5
SHG (+M) SHG (+M)
SHG (-M) 3 SHG (-M)
4000
2.5
3000 2
1.5
2000
1
1000
0.5
0 0
34 35 36 37 38 42 44 46 48
0.4 0.3
0.2
0.2
0.1
0
0
0.1
40mW 36mW
0.2 60mW 0.2 60mW
80mW 84mW
0.3
100mW 108mW
0.4 0.4
34 35 36 37 38 42 44 46 48
Angle of incidence (deg) Angle of incidence (deg)
(a) (b)
Figure 5.1 – The angular dependence of the mSHG intensity and the SH magnetic
contrast ρ in (a) the bilayer (20 nm)Au/ (35 nm)Co/ sapphire and (b) the trilayer
(5 nm)Au/ (10 nm)Co/ (15 nm)Ag/ fused-silica at different pump powers, represented
by different colors. The fundamental wavelength in both cases is 1560 nm. In the
top row, the dotted lines illustrate the SHG(+M) while the dashed lines are used for
SHG(-M).
This phenomenon implies that the generated SH field depends on the temperature
of the sample. Since the SH field is the interference of the non-magnetic and magnetic
78
5.1. FLUENCE DEPENDENCE OF THE MSHG
SH fields as in Eq. 2.49, it can be written as:

nm mag
E2ω (±M ) = E2ω ± E2ω ∝ χ(2)
nm ± χ (2)
mag M Eω E ω
(2) (2)
where χnm and χmag are the non-magnetic (even) and the magnetic (odd) SH nonlin-
(2) (2)
ear susceptibilities, respectivily. One considers χmag (M ) = χmag M at the first-order
(linear) approximation. From this formulation, the thermal dependence of the mSHG
might come from the SH susceptibility1 χ(2) , the magnetization M or the fundamental
field Eω .
Keeping in mind the relaxation dynamics of laser-excited electrons in metallic mul-
tilayers [31, 113], the optical properties of the structure can be influenced by a high
power optical pulse below the damage threshold. Hot electrons, which acquired high
energies above the Fermi-level by absorbing photons from a strong pump pulse, trans-
fer heat by electron-electron thermalization and electron-phonon scattering [122–124].
The thermal equilibrium established a few picoseconds after the laser pump [113] might
cause the temperature dependence of the effective dielectric function ǫef f . In order
to see whether the effective dielectric function ǫef f of the metallic structure depends
on temperature, the fluence dependence of the linear reflectivity is verified. Fig. 5.2
shows the angular dependence of the linear reflectivity in (a) the bilayer of (20 nm)Au/
(35 nm)Co and (b) the trilayer of (5 nm)Au/ (10 nm)Co/ (15 nm)Ag at different op-
tical powers. Even though there is slight change in the scale of the linear reflectivity,
the angular position and angular broadening of the peak do not seem to depend on the
incident power. Thus, one can conclude that the linear dielectric function ǫef f does
not vary significantly within the limit of registered setup.
The temperature dependence of the optical properties in the nonlinear regime can
be verified by the fluence dependence. If the second-order nonlinear susceptibilities
χ(2) were sensitive to temperature, the far-field SHG intensity
I2ω ∝ |E2ω |2 ∝ |χ(2) Eω Eω |2
would be no longer proportional to the squared intensity of the fundamental light

Iω2 . Here, the quadratic dependency at a constant magnetization, like in Fig. 3.5,
is preserved at all angles of incidence for both the bilayer Au/Co and the trilayer
Au/Co/Ag. Similar to ǫef f , the pump power within our measurement range does not
seem to have significant effects to the nonlinear susceptibility χ(2) in the near-infrared
spectral range. Even when the strength of the electric field is high, i.e. at the angles
of incidence where the linear SPP(ω) and nonlinear SPP(2ω) are excited, the scaling
I2ω ∝ Iω2 holds.
Ruling out all other factors, the dependence of the magnetic contrast ρ on the pump
power seems to be the result of thermal demagnetization M (T ). Apparently, there are
two noteworthy effects that can cause demagnetization: (i) an increase of the average
temperature or heat accumulation caused by absorption of multiple pump pulses during
the time of open chopper and (ii) ultrafast demagnetization [125–127] induced by
1
here the general notation of χ(2) can be understood as either the magnetic or the non-magnetic
susceptibily.
79
p Pol, 35Co/20Au p Pol, 5Au/10Co/15Ag
0.75 1
17.2 mW
0.7 36.1 mW 0.9
59.5 mW
0.65 84.5 mW 0.8
Reflectivity
Reflectivity
108 mW
0.6 0.7
127.1 mW
0.55 139.8 mW 0.6
144.5 mW
0.5 0.5
0.45 0.4
0.4
34 36 38 40 45 50
Angle of incidence (deg) Angle of incidence (deg)
(a) (b)
Figure 5.2 – The angular dependence of the linear reflectivity in (a) the sample
(20 nm)Au/ (35 nm)Co/ sapphire and (b) the sample of (5 nm)Au/ (10 nm)Co/
(15 nm)Ag/ fused-silica at different pump powers, represented by different colors. The
excitation is centered at wavelength 1560 nm and p-polarized.
single pulse where the first half of the pulse changes the conditions for magnetic SHG
generated by its second half. The heat accumulation, which should scale linearly
with the number of pulses transmitted by the mechanical chopper during its opening
time, can be reduced by increasing the chopper frequency. Variation of the chopping
rate2 up to 4.45 kHz (see specifications in AppendixA) does not seem to reduce the
thermal effect on the SH magnetic contrast suggesting that heat accumulation is not
the primary reason for demagnetization. The reason why the SH magnetic contrast ρ in
the trilayer Au/Co/Ag reduces exclusively when the nonlinear SPP(2ω) is excited can
not be explained either. One would envisage pump-probe experiments in Kretschmann
configuration to clarify the role of SPP-excitation on the dynamics of magnetization.
5.2 Magneto-plasmonic auto-correlation

Sub-picosecond pulses cannot be measured by high speed photo detectors because their
rise and fall time, in picosecond to nanosecond scale, is still longer than the pulse dura-
tion. Thus the ultrashort pulse is used to gate itself in the auto-correlation and cross-
correlation measurements, where the information of the pulse can be obtained [128].
The nonlinear optical effects are employed in the intensity auto-correlation and cross-
correlation because they are instantaneous phenomena, which occur exclusively upon
the temporal overlap of the fundamental pulses. Both the nonlinear auto-correlation
in this section and the spectrally-resolved cross-correlation in the next section were
performed in corporation with LS Leitenstorfer at University of Konstanz, where the
2
considering that the heating of sample is linear with time.
80
5.2. MAGNETO-PLASMONIC AUTO-CORRELATION
ultrashort sub-10 fs laser, center wavelength 1200 nm (bandwidth ≈400 nm) with
repetition rate 40 MHz was employed.
All the measurements in this part were performed on the trilayer of (5 nm)Au/
(5 nm)Co/ (15 nm)Ag deposited directly on UV fused-silica prism, prepared by Dr.
Denys Makarov (HZDR, Dresden). This sample exhibits the largest mSHG intensity
and magnetic contrast, based on the discussion on the trilayers in chapter 3. The
purpose of direct deposition is to avoid the parasitic thermal effect coming from the
heating of phase-matching liquid, which has been used for sample attachment. Com-
pared to those deposited on flat substrate, the sample exhibits stronger SHG signal
and sharper angular dependence.
5.2.1 Magneto-plasmonics auto-correlation setup

The objective of the magnetically-controlled autocorrelation performed in Kretschmann
geometry is to observe the influence of propagaring SPPs on the mSHG pulse. The
linear reflectivity at ω is measured by an integrating sphere (made of GaAs for in-
frared wavelength and Si for visible wavelength). The SHG is detected using PMT
(Hamamatsu R955) coupled with a High Frequency Lock-in Amplifier (Zurich Instru-
ments) working at 40 MHz, the repetition rate of the laser, which permits to reduce the
1/f -noise. Unlike the photon counting regime described in chapter 3, this detection
scheme can keep away the signal from cosmic rays due to much higher statistics, but
is vulnerable to electronic noise.
Figure 5.3 – A simplified schema for the setup of the magneto-plasmonic auto-
correlation. The inset shows the wavevectors of incident beams and the SH emitted
beam.
81
Fig. 5.3 illustrates the intensity auto-correlation of the mSHG performed in Kretschmann
configuration. The external magnetic field is perpendicular to the plane of incidence
as in the transverse MOKE geometry. The magnetic field strength is around 200 mT
at the focused spot on the sample, enough to saturate the in-plane magnetization of
cobalt thin film. The p-polarized fundamental fs beam is splitted into two pulse trains,
denoted as Eω (t − τ ) and Eω (t). The time delay τ = 2Ldelay /c between two pulses is
adjusted by a retro-reflector mounted on a linear translation stage. It depends on the
travelling distance 2Ldelay of the delay stage and the speed of light c. In this setup, τ
is varied in the range from -25 fs to +25 fs, with an increment of 1 fs. The two beams
are directed to a spherical mirror (focal length f=325 mm), symmetric to the plane of
incidence with vertical beam separation of approximately 2 mm. They are focused on
the same spot on the sample surface. Small beam separation, hence small solid angle,
ensures a large kx (and very small kz ) wave vector component for SPP excitation.
Figure 5.4 – In the Kretschmann configuration, two fundmental pulses, delayed by time
τ , generate cross-product SH field Eω (t)Eω (t − τ ) at bottom and top interfaces of the
multilayer. Enon−res is the SH field at the bottom interface. Eres refers to the coupling
of SH polarization at the top interface and the radiative SPP(2ω). The propagation
of SPP along the metal/air interface cause the delayed arrival of the resonant pulse.
The conventional intensity auto-correlation usually considers a single source of

SHG, such as a thin BBO crystal. In Kretschmann configuration, one has to con-
sider the non-resonant SH field generated in the bottom interface and the radiative
SPP at the double-frequency coming from the upper interface, as discussed in chapter
3. Here the auto-correlation measurement has been performed in the non-collinear
fashion, different momentum-matching conditions are valid for the generation of SH
fields, such as 2k1 (ω) = k1 (2ω), 2k2 (ω) = k2 (2ω) and k1 (ω) + k2 (ω) = k(2ω), where
k1 and k2 refer to the wavevectors of two incident beams. The SH fields generated
82
by different paths are separated in space, as shown by the inset in Fig. 5.3. Using
the aperture made of a pair of razor blades, fundamental pulses at frequency ω and
any SH signals created by individual input beams (both non-resonant and resonant SH
fields) are blocked. Only the SH fields generated by two incident pulses, corresponding
to the phase-matching condition k1 (ω) + k2 (ω) = k(2ω), are retained. In quantum-
optical picture, it means that a SH photon is emitted upon the annihilation of two
fundamental photons coming from two different pulses. The cross-product SH fields
E2ω (t, τ ) ∝ χ(2) Eω (t − τ )Eω (t) can be generated both at the bottom and top interfaces
of metallic multilayer. The SH field at the bottom interface, Enon−res , does not interact
with the SPP field at the metal/air interface because the distance to the open surface
is larger than the field skin depth. In contrast, Eres is the result of the SH polarization
in the upper interface coupled to the nonlinear SPP at 2ω, as proven in chapter 3.
The time it takes for the radiative SPP pulse to propagate along the metallic surface
is arbitrary τsp = Lsp /c3 . Besides that, there is a phase difference between Enon−res
and √ Eres due to the difference in the optical path lengths inside the prism, which is
Lsp / 2 for an angle of incidence ≈ 45◦ .
Non-resonant Enon−res and SPP-resonant Eres SH fields along x read:
Enon−res (t, τ ) ∝ Eω (t)Eω (t − τ ) exp (i2kx x − i2ωt)

h Lsp t i (5.1)
Eres (t, τ ) ∝ Eω (t)Eω (t − τ ) exp iksp x − √ − i2ω(t + τsp ) − .
2 τsp
We note that 2kx (ω) = ksp (2ω) due to the nonlinear phase matching condition in
Eq. 3.6, and exp −t/τsp represents for the decay of the SPP field along x during the
time τsp . The total SH intensity is the interference of these two fields, as defined in
Eq. 3.11. Thus one obtains
I2ω (t, τ ) = |Enon−res (t, τ ) + LEres (t, τ )|2

= |Enon−res (t, τ )|2 + |LEres (t, τ )|2 + LEnon−res (t, τ )Eres (t, τ ) cos ϕnr,res ,
(5.2)
where ϕnr,res is the phase difference between the non-resonant Enon−res and resonant
LEres SH fields. Recall that L is the normalized Lorentzian function defined in chap-
ter 3, which depends on the angle of incidence. Referring to Eq. 3.8, L = 1 at the angle
θ0 where SPP(2ω) is excited by the nonlinear mechanism. Deviated from this angle,
L decays within a certain angular width Γ decaying to zero at non-resonant angles.
In the vicinity of the resonant angle of SPP(2ω) and L 6= 0, the signal collected by
3
Considering that the group velocity of SPP is close to its phase velocity c.
83
the detector (PMT in this case) is the time integral of I2ω (t, τ )
Z +∞
S2ω (τ ) = I2ω (t, τ )dt
−∞
Z +∞
= |Enon−res (t, τ )|2 dt
−∞
Z +∞
(5.3)
+ |LEres (t, τ )|2 dt
−∞
Z +∞
+ LEnon−res (t, τ )Eres (t, τ ) cos ϕnr,res dt
−∞
= Snon−res (τ ) + Sres (τ ) + Smix (τ )
In this equation, the non-resonant part is
Z +∞
Snon−res (τ ) = |Enon−res (t, τ )|2 dt
−∞
Z +∞
= |Eω (t)Eω (t − τ )|2 dt (5.4)
−∞
Z +∞
= Iω (t)Iω (t − τ )dt
−∞
Snon−res (τ ) is the same as the intensity auto-correlation function obtained in case of a

single SH source.
The resonant intensity is
Z +∞
Sres (τ ) = |LEres (t, τ )|2 dt
−∞
Z +∞ (5.5)
h Lsp t i2
= LEω (t)Eω (t − τ ) exp iksp x − √ − i2ω(t + τsp ) − dt
−∞ 2 τsp
The last term
Z +∞
Smix (τ ) = LEnon−res (t, τ )Eres (t, τ ) cos ϕnr,res dt
−∞
is equivalent to the spectrum of SH field in the SH interferogram. At the angles of

incidence where SPP(2ω) is not excited, i.e. L = 0, Sres (τ ) and Smix in Eq. 5.2 vanish,
leaving only the non-resonant SH signal Snon−res (τ ).
As an external magnetic field is present in the system, the SH fields generated from
the magnetic (odd) χ(2) components are sensitive to the reversal of the magnetization.
As Eres originates from the upper interface, and the ferromagnetic cobalt layer is buried
within the SPP field skin depth, it is expected that the magnetic properties primarily
modify the Eres . Furthermore, LEres is responsible for the angular dependence of the
SH magnetic contrast ρ, according to the discussion in chapter 3. The magnetization-
induced components in Enon−res , if the evanescent field can leak through the bottom
silver film to reach the cobalt layer, would have a rather small effect on the magnetic
contrast.
84
5.2.2 Results and discussion
Figure 5.5 – The time-integrated (static) measurements of the linear reflectivity (black
solid line), the mSHG outputs at different magnetizations (red solid line and dotted
line) and the corresponding SH magnetic contrast (green solid line) in the Kretschmann
geometry for the trilayer (5 nm)Au/ (10 nm)Co/ (15 nm)Ag deposited directly on a
fused-silica prism.
The angular dependence of the linear reflectivity at the fundamental wavelength,

mSHG(±M ) and the magnetic contrast of the sample (5 nm)Au/ (10 nm)Co/ (15 nm)Ag
are shown in Fig. 5.5. The dip of the linear reflectivity at ω (black solid line) corre-
sponds to the peaks of the mSHG intensities (red solid line and dotted line) at the
angle 44.3◦ , signifying the excitation of the linear SPP(ω) by the p-polarized incident
beams. The magnetic contrast (green solid line) is not in the form of the real part of a
Lorentzian function as in chapter 3. However, based on the reverse approach, one can
estimate that the SPP(2ω) is coupled by the nonlinear mechanism at around 46◦ .
Fig. 5.6a shows the magnetization-dependent auto-correlation in the same sample
at four different angles of incidence: 34.7◦ , 44.3◦ , 46◦ and 55.3◦ . Besides the important
angles 44.3◦ and 46◦ mentioned above, the two others are off-resonance angles, where
no modes of SPPs are involved. Fig.5.6a shows that the intensity auto-correlation at
angle 44.3◦ is the highest, while that at angle 46◦ is the smallest, in agreement with the
peak and dip of the mSHG output in the time-integrated regime (red curves in Fig. 5.5).
One can also see that the pulse has little dependence on the magnetization M , except
at angle 46◦ , where the SH magnetic contrast is maximized and the nonlinear coupling
to the SPP(2ω) is dominant. Fig.5.6b normalizes the intensity auto-correlation to
unity for the same magnetization.
At the angles where the radiative SPP(2ω) is not coupled, such as 34.7◦ , 44.3◦ and
55.3◦ , the Lorentzian L = 0 and Snon−res (τ ) is the only remaining signal in Eq. 5.3. The
85
auto-correlation functions at these angles have the Gaussian shape with FWHM≈10 fs.
It corresponds to a Gaussian pulse with the pulse duration of ≈7 fs4 , as expected with
the incident optical pulse. It is noteworthy that at angle 44.3◦ , where the linear SPP(ω)
is excited, the auto-correlation intensity is very high due to the SPP-enhancement of
the fundamental fields, Eω (t) and Eω (t − τ ). However, there is no obvious influence of
the linear SPP(ω) to the shape of the auto-correlation function.
In contrast, the auto-correlation at the angle 46◦ has different shape from those
at the other angles, with two supplementary wings symmetric at both side of the
main Gaussian function. The smaller Gaussian functions are separated from the main
function by a fixed time, thus both the non-resonant Snon−res (τ ) and the Sres (τ ) related
to the radiative SPP(2ω) are present in this auto-correlation5 . The interference of
the two SH fields in Smix (τ ) should be present in the auto-correlation function when
L 6= 0. However, from the time-integrated measurement in Fig. 5.5, there is dip of the
SH intensity near angle 46◦ , meaning that the Enon−res and Eres interfere destructively
around this angle. Apparently, the phase cos ϕnr,res ≈ 0, and the effect of Smix (τ ) to
the auto-correlation at angle 46◦ is unnoticeable.
SHG (+M) at different angles

20 1
55.3deg SHG(+M) 1
46 deg
55.3 deg
Normalized SHG
+M
Normalized SHG intensity (arb. u.)
55.3deg SHG(-M)
46.0deg SHG(+M) 0.8 46.0 deg
0.5
46.0deg SHG(-M) 44.3 deg
15 -M
34.7 deg
SHG intensity (µV)
44.3deg SHG(+M)
44.3deg SHG(-M) 0.6 0
-20 -10 0 10 20
34.7deg SHG(+M) Time delay (fs)
10 34.7deg SHG(-M)
0.4
0.2
5
0
-0.2
-20 -10 0 10 20 -20 -10 0 10 20
Time delay (fs) Time delay (fs)
(a) (b)
Figure 5.6 – (a) Intensity auto-correlation in the layer (5 nm)Au/ (10 nm)Co/
(15 nm)Ag employing the Kretschmann configuration. Each color represents for a
different angle of incidence. The solid and dashed lines are the signals for +M and
−M , respectively. (b) The intensity auto-correlation of the SHG (+M) pulses at dif-
ferent angles of incidence are normalized to unity. The inset shows the normalized
intensity at angle 46◦ for opposite magnetizations.
From the auto-correlation function at opposite magnetization, the τ -dependent SH

magnetic contrast ρ can be calculated based on Eq. 2.50, and shown in Fig. 5.7a. The
contrast ρ becomes strong exclusively at angle 46◦ where the SPP(2ω) is excited by
For a sech2 pulse, the relation between the width of the intensity auto-correlation
4
√ ∆τA and the
pulse duration ∆τp is ∆τA = 1.54∆τp . For a Gaussian pulse, the ratio is 2.
5
The signals are symmetric about the zero delay time due to the temporal symmetry of the auto-
correlation S(τ ) = S(−τ ).
86
the nonlinear mechanism, and is suppressed at other angles. This observation agrees
with the expectation in the previous section that the SH magnetic properties are at-
tributed to the radiative nonlinear SPP Eres , which is generated at the upper interface.
At non-resonant angles, such as 44.3◦ and 55.3◦ , Snon−res (τ ) does not carry the mag-
netic information, causing zero magnetic contrast as shown in Fig. 5.7a. At angle 46◦ ,
the presence of the magnetization-dependent Sres (τ, ±M ) (and possibly Smix (τ, ±M ))
causes a large SH magnetic contrast exceeding 70% on sub-10 fs timescales. Normal-
ized intensity autocorrelation at angle 46◦ (inset of Fig. 5.6b) reveals that the the
supplementary wings remain unchanged regardless of the magnetization M , but the
central Gaussian function around τ = 0 extinguished completely when the magnetiza-
tion is reversed. To be able to understand the effect, it would be better to enlarge the
auto-correlation time window, which is unfortunately unattainable in the registered
setup.
However, it is clear that the giant nonlinear contrast, ρ(τ ) > 70%, observed at
angle 46◦ on short timescales is suppressed by approximately a factor of 2 in the time-
integrated measurements, ρ ≈ 40% in Fig. 5.5. Such behavior points to the coherent
nature of the mSHG signal and, possibly, to the detrimental role the electron-phonon
scattering plays in reducing the observable magnetic contrast in the time-integrated
measurements. At different time delays, the angular dependence of the SH magnetic
contrast resembles the shape at the time-integrated measurement, as seen in Fig. 5.7b.
The maximal contrast is achieved when the two beams overlaps temporally, τ = 0. It
again confirms the dynamics of the SH magnetic contrast ρ and the magnetization-
dependent radiative SPP Eres (±M ) excited by the nonlinear mechanism.
100 60
-42 fs
Magnetic contrast (%)
55.3 deg
Magnetic contrast (%)
50 -33 fs
46.0 deg
-25 fs
44.3 deg 40 -17 fs
50 -8 fs
30 0 fs
8 fs
20 17 fs
25 fs
0 33 fs
10
-50 -10
-20 -10 0 10 20 42 44 46 48 50
Time delay (fs) Angle of incidence (deg)
(a) (b)
Figure 5.7 – Time-resolved second-harmonic magnetic contrast ρ of the trilayer

(5 nm)Au/ (10 nm)Co/ (15 nm)Ag.
87
5.3 Magneto-plasmonic cross-correlation

5.3.1 Magneto-plasmonic cross-correlation setup
Au
Co
Ag
Figure 5.8 – The mSHG cross-correlation on the Kretschmann configuration.
The idea of a mSHG cross-correlation is sketched in Fig. 5.8. A pump at funda-

mental frequency Eω (t) and a reference signal, so-called local oscillator, E2ω (t − τ )
delayed by a time τ , are focused on the sample in a co-linear geometry. The external
magnetic field is perpendicular to the plane of incidence as in the tranversal MOKE
configuration. A magnetization-dependent SH field E(2ω, t, M ) ∝ χ(2) Eω (t)Eω (t) is
generated in the Kretschmann trilayer structure (5 nm)Au/(10 nm)Co/(15 nm)Ag.
To check for possible spectral phase of the emitted mSHG signals from this structure,
the spectrally-resolved cross-correlation measurements were performed. Toward this
end, the mSHG signal E2ω (t, M ) was linearly mixed with the reference field Eref (t − τ )
I2ω (τ, ±M ) ∝ |E2ω (t, ±M ) + Eref (t − τ )|2

∝ |E2ω (t, ±M )|2 + |Eref (t − τ )|2 (5.6)
+ 2E2ω (t, ±M )Eref (t − τ ) cos ϕ±M
Comparing the phase ϕ±M when the magnetization M is reversed, one expects to re-
cover the phase difference ϕ = ϕ+M −ϕ−M between the magnetic and the non-magnetic
SH fields, illustrated in the phasor diagram in Fig. 3.11 (chapter 3).
A simplified setup for the experiment is shown in Fig. 5.9. Same as the previous
experiment of auto-correlation, the fundamental spectrum is centered at 1200 nm, with
FWHM=275 nm. The local oscillator, generated by a Beta barium borate (BBO) crys-
tal, has the wavelength centered at 600 nm, with FWHM=70 nm. The broad spectral
88
5.3. MAGNETO-PLASMONIC CROSS-CORRELATION
range of the fundamental and the reference pulses (see AppendixA) are necessary for
the spectral-resolved measurements. After the polarizing beam splitter, the fundamen-
tal beam goes through a chopper and then a delay stage for delay time τ . Meanwhile
the reference beam goes through a polarizer, then reflects several times through the
two parallel mirrors (the chirp compensator) before joining the fundamental beam af-
ter upon reflection on the glass plate. The two beams are focused on the thin film in
the Kretschmann geometry using a spherical mirror (f=325 mm). The reference beams
co-propagates through the sample structure ensuring robustness of the phase-resolved
measurement. The dispersion of the reference beam (blue beam) in the fused-silica
prism is pre-compensated by the chirp compensator. The experiment is carried out
at the angle of incidence 46.4◦ , where the SH magnetic contrast is highest. After the
prism, a low-pass filter is used to collect the SH signal. Both the mSHG output and
the reference field are p-polarized. Linear interference of the mSHG signal and the
reference wave is spectrally resolved via a grating-based spectrometer equipped with a
fast CCD camera. Amplitude modulation of the fundamental wave allows for the high
sensitivity measurement of the differential spectra ∆T /T .
Figure 5.9 – A simplified scheme for the cross-correlation of mSHG on the Kretschmann
geometry.
5.3.2 Results and discussion

The interference patterns of the mSHG signal and the reference local oscillator at
different time delay are shown in Fig. 5.10. The interference fringes are clear. Red
89
and blue colors illustrate the two opposite magnetizations of the mSHG E2ω (t, ±M ).
At a fixed delay between the fundamental and local oscillator fields, the observed
spectrally-resolved interference shows no visible dependence on the direction of the
external magnetic field. It can be concluded that there exists no noticeable structure
in the phase spectrum of the mSHG signal. This result refers to a phase-matched
broadband nature of the generation process of the nonlinear signal and suggests no
significant spectral interference arising from the propagation aspects of the SPP within
the multilayer structure. No phase shift when the magnetization is reversed, i.e. ϕ+M =
mag
ϕ−M , means zero phase difference ϕ = 0 between the magnetic SH field E2ω and the
non-magnetic one E2ω nm
. This result is controversial. There is also a possibility that
the noise signal is too high that the real phase shift could not be recovered.
Figure 5.10 – Spectrally-resolved linear cross-correlation of a reference wave with the

mSHG signal for opposite directions of the magnetization M (red and blue lines) at
different fixed time delays.
5.4 Conclusion
In this chapter I have described an explorative part of collaborative research, aiming
at understanding the transition from the static to the dynamic regime in nonlinear
magneto-plasmonics. The power dependence of the SH magnetic contrast has been
discussed, which is believed to be a result of the demagnetization of the structure. The
effect is obvious as it manifests itself as a 50% reduction of the SH magnetic contrast
dropping even further at high pump intensities. Although the microscopic picture of
this power dependence is still unclear, it is an effect worth exploring.
The magnetically controlled femtosecond time-resolved auto-correlation measure-
ments in Kretschmann configuration display the dynamic magnetic contrast exceeding
90
5.4. CONCLUSION
70% at the angle of incidence 46◦ , where the mSHG output at one orientation of mag-
netization is almost suppressed. At this angle we have observed a qualitative change
in shape of the auto-correlation function and a huge magnetic contrast. These effects
are related to the excitation of SPP(2ω) by the nonlinear phase-matching mechanism
discussed in chapter 3. However, further modelling of autocorrelation functions should
be conducted in order to quantify the contribution of SPP(2ω) to the autocorrelation
function. Spectrally-resolved cross-correlation of the mSHG provides measurable inter-
ference fringes in the spectral domain. The difficulties of this experiment were related
to the compensation of dispersion of the reference field in the prism and to obtain spa-
tial and temporal overlap between the generated SH field and reference fields. Though
the experiment was successful in the sense that spectral interference fringes displayed
dynamics on ultrafast time scales, and also resolving the phase of the mSHG from a
hybrid multilayer structures, the expected phase shift between the magnetic (odd) and
non-magnetic (even) SH fields could not be observed. Sophisticated interferometric
phase reconstruction algorithms should be applied in order to discriminate physical
effects from the artifacts of spatio-temporal interferometric measurements in such a
complex experimental geometry.
91
Chapter 6
Summary and conclusion
Surface electromagnetic (EM) waves propagating along metallic interfaces and excited
in various geometries enabling their phase-matched excitation, can enhance the linear
and nonlinear magneto-optical effects in metal-ferromagnet multilayer structures. This
work investigates the phenomena in the nonlinear regime, specifically the influence of
surface EM waves on the magnetization-induced SHG in Kretschmann configuration
combined with the geometry of the transverse magneto-optical Kerr effect. A gonio-
metric magneto-optical setup has been constructed to discriminate between distinct
surface waves excited by different phase-matching conditions. The thesis focuses on
three aspects of the nonlinear magneto-optics mediated by the surface EM waves:
• nonlinear magneto-plasmonics on Kretschmann metal/ferromagnet multilayers
• nonlinear Wood’s anomaly on periodic arrays of nickel nanodimers
• nonlinear magneto-plasmonics in the dynamic regime
Performed on different systems or different measurement setups, these studies help to

identify the role of nonlinear phase-matching conditions for surface EM waves at the
double SHG frequency and to quantify their contributions to the giant magnetic SHG
contrast.
Nonlinear magneto-plasmonics in metal-ferromagnet multilayers

In order to excite SPPs at a free surface of metal-ferromagnet multilayers, the conven-
tional Kretschmann geometry was employed. In various experiments, the excitation
efficiency of different SPP modes has been represented by the Lorentzian function ac-
counting for their phase-matched excitation. In this work, the Lorentzian function
L(θ) has been derived theoretically from the coupling of the SH polarization to SPP’s
electric field at the double-frequency. It is also demonstrated that the SH magnetic
contrast can provide direct experimental access to the parameters of L(θ), especially
the one related to the nonlinear coupling of SPP(2ω). This finding allowed us to pro-
pose a reverse approach, which estimates the nonlinear SPP coupling efficiency by the
experimental method.
93
CHAPTER 6. SUMMARY AND CONCLUSION
The role of linear and nonlinear SPPs was clarified while inspecting angular-dependencies
of the mSHG output and the magnetic contrast. The linear SPP(ω) enhances the
fundamental electric field, and thus results in the characteristic peak in the angular
dependence of the SHG in metallic thin films. The nonlinear SPP(2ω) is excited via
the nonlinear phase-matching condition and is detected through the interference of its
radiative emission into the glass prism with the non-resonant SH sources located at
various interfaces. The pronounced dip in the angular dependence of the SH emis-
sion results from the destructive interference between the non-resonant SH and the
resonant SPP(2ω) fields. A comparative study of two distinct metal-ferromagnet sys-
tems, i.e. the trilayer of (5 nm)Au/ (10 nm)Co/ (15-25-35 nm)Ag and the bilayer of
(20-30-40 nm)Au/ (30 nm)Co, suggests that the SPP-enhanced SH emission in the
trilayer is typically an order-of-magnitude stronger than that in the bilayer. It means
that the field enhancement of the linear SPP in the trilayer Au/Co/Ag is larger. In
terms of magnetic properties, the SH magnetic contrast in the bilayer Au/Co is influ-
enced both by the linear and the nonlinear SPPs. On the other hand, the position of
the maximum SH magnetic contrast in the trilayer Au/Co/Ag is determined by the
nonlinear SPPs, while its magnitude depends on the fundamental field strength, and
is enhanced by the linear SPP(ω). It thus suggests that nonlinear magneto-optical
signals can be tuned by tailoring SPP fields. Indeed, it has been demonstrated in this
work that varying the thickness of the plasmonic (silver) layer in trilayers can boost
the SH magnetic contrast up to 60%, almost twise as large as compared to the previ-
ously reported values. Therefore, the trilayer Au/Co/Ag represents a physical system
for nonlinear-optical proof-of-concept studies such as the modulation of light at the
nano-scale by an external magnetic field.
Nonlinear Wood’s anomaly on nickel nanodimers

In periodic structures where the surface filling factor of metals is low, the phased-
matched excitation of surface EM waves is based on the principle of Wood’s anomaly,
which occur when a diffraction order emerges along the surface of the grating. The
possibility to co-excite linear and nonlinear Wood’s anomalies is governed by the ratio
between the optical wavelength λ and the periodicity of the grating Λ. For example,
the linear and the second-harmonic Wood’s anomaly can be excited simultaneously
when Λ ≥ λ/2, but only the SH diffraction anomaly at the order mnl = −1 is allowed
when λ/4 < Λ < λ/2. In this work, Wood’s anomalies have been investigated in a
rectangular 2D-array of nickel nanodimers with different sizes of a unit cell: one is
greater and the other is less than half of the fundamental wavelength.
The results show that the linear Wood’s anomaly has little influence on the SH yield,
which differs significantly from the linear SPP in metallic multilayers and from Wood’s
anomaly on plasmonic gratings [4]. However, similar to the SPP case, the magnetic
contrast shows a fingerprint of the SH Wood’s anomaly excited through the nonlinear
phase-matching mechanism. The nonlinear-optical framework is further sketched for
the analysis of periodic structures with smaller periodicities and argue that nonlinear-
optical and magneto-optical effects can be utilized as a highly sensitive measure of the
efficiency of diffraction gratings in periodic arrays of magnetic nanoparticles.
94
The nickel nanodimers structures were originally designed to observe any possible
contributions of localized surface plasmon resonance (LSPR), which can be excited in
the dimer gap within the spectral range 500-1000 nm, to the SH yield and magnetic
contrast. Under specific conditions, this mode can be coupled to the Wood’s anomalies
forming a hybrid mode called surface lattice resonance (SLR). At the current stage,
it was difficult to disentangle these effects from the mSHG signal, except for a minor
change in the effective refractive index of the surface EM waves. The idea of hybrid
magneto-plasmonic SLRs might be revised in future in experiments with a different
sample design. Another outlook is to excite nonlinear Wood’s anomaly in different
meta-surfaces, such as periodic arrays of sub-wavelength holes, where one can speculate
about the phenomenon of magnetization-controlled extraordinary optical transmission
in the nonlinear regime.
Nonlinear magneto-plasmonics in the dynamic regime

While previous sections describe the interaction of surface EM waves and their non-
linear magneto-optical effects in the time-integrated regime, this section takes the first
step to explore the dynamics of the phenomena. It is an exploratory part of the thesis,
where open questions could inspire further investigations.
Pump-power dependence of SH magnetic contrast in metal-ferromagnet multilayers
at power levels indicates on the ultrafast demagnetization. This finding urges us to
investigate the dynamics of magneto-plasmonic interactions. Using the Au/Co/Ag tri-
layer system as the second-order nonlinear medium, the auto-correlation of ultrashort
pulses has been measured in Kretschmann configuration. In this case, two sources of
SHG must be considered, which are non-resonant and resonant to the nonlinear SPP
field propagating at metal-air interface. Between them, the resonant source exhibits
the magnetization dependence, due to the ferromagnet layer buried within the SPP
field skin depth. The intensity auto-correlation trace shows a normal Gaussian shape
at the non-resonant angles of incidence. However, the complexity escalates when radia-
tive SPP(2ω) are excited and interfere with the non-resonant SH fields. The presence
of long-living wings in the auto-correlation function is the result of this interference.
Interestingly, the central peak in the auto-correlation can be completely suppressed
upon the reversal of the magnetization. The fact that the nonlinear magnetic con-
trast observed on short timescales is reduced by approximately a factor of two in the
time-integrated measurements implies the coherent nature of the mSHG signal.
Carried out on the same sample, the spectrally resolved cross-correlation measure-
ments attempt to retrieve the phase of mSHG in the dynamic regime. The broadband
nature of the generation process of the nonlinear signal suggests no significant spec-
tral interference arising from the propagation aspects of the SPP within the multilayer
structure, in agreement with preliminary experimental results. However, more detailed
studies and application of sophisticated interferometric phase reconstruction algorithms
are necessary to confirm these premature conclusions.
95
Appendix A
Technical specifications
A.1 Verify film thickness using ATIR

When the precise thickness of each layer is unknown, one can compare the angular de-
pendences of reflectivity calculated by Matrix-Transfer-Method (MTM) and measured
experimentally to estimate the relative thickness between layers. Using this method,
the relative thickness of each sample can be estimated as in Table. A.1.
Samples Estimation using MTM

(15nm)Ag/(5nm)Co/(5nm)Au (15nm)Ag/(10nm)Co/(4nm)Au
Table A.1 – Employing the known database for the refractive indices of gold, silver
(Johnson and Christy 1972) and cobalt (Artem 1968), the relative thicknesses between
different layers in the hybrid structure can be estimated. The right column is the
designed thicknesses of the samples, but they were not confirmed by electron micro-
scope. The left column is the approximation of the relative thickness by fitting the
experimental angular dependence of reflectivity using MTM.
Inversely, knowing the thickness of the layer, one can use this SPP-based ellip-
sometry to estimate the effective dielectric function of the multilayer. The limits of
the method are polarization (p-polarized light) and the wavelength of the incident
radiation (infrared spectral range), where the SPP-dip is distinguishable.
A.2 Technical specifications
97
APPENDIX A. TECHNICAL SPECIFICATIONS
Output Output
Properties
1560 nm 780 nm
Average power at output (mW ) 390 170
Estimated energy per pulse (mJ) 9.75 × 10−6 4.25 × 10−6
Peak power (mW ) 9.75 × 107 4.25 × 107
Number of photons per pulse 7.65 × 1010 1.67 × 1010
Typical beam diameter (mm) 3.5 1.2
Focused by lens f=+300 mm, beam waist (µm) 170 250
Focused by lens f=+300 mm, spot area (mm2 ) 2.28 × 10−2 4.84 × 10−2
Focused by lens f=+300 mm, fluence (µJ/cm2 ) 42.8 8.78
Focused by lens f=+300 mm, Rayleigh length (µm) 170 250
Table A.2 – Techinical specification of Toptica fiber laser for nonlinear magneto-
plasmonic setup
Settings Optimized value

PMT voltage supply 900 V
Trigger frequency 4450 Hz
Gates per counting period 5000
Gate width 80 µs
Gate A delay 45 µs
Gate B delay 155 µs
Channel discrimination level -2 mV
Trigger discrimination level 100 mV
Table A.3 – Technical specifications of the photon counting regime
(a) (b)
Figure A.1 – Spectra of the fundamental and reference (local oscillator) radiations for
the cross-correlation experiment in Konstanz.
98
Appendix B
Third-harmonic generation in metallic

thin film
B.1 Estimation of Third-harmonic generation (THG)

in Au film
Pulse energy of the laser
The laser source at 1560 nm have an average power of 100 mW after the mirrors, half
waveplate and filter. This corresponds to a pulse energy:
Average power Pave 100 × 10−3 W

Ep = = = = 2.5 × 10−9 J (B.1)
Repetition rate frep 4 × 107 Hz
Peak power of a pulse:
Pulse energy Ep 2.5 × 10−9 J

P1 = = = = 2.5 × 104 W (B.2)
Pulse duration τ 100 f s
Note that each photon at 1560 nm has energy of 1.27×10−19 J or 0.795 eV , number
of photons per pulse:
Ep 2.5 × 10−9 J
Np1 = = = 1.96 × 1010 photons/pulse (B.3)
Eω 1.27 × 10−19 J
Focusing condition
Under the focused Gaussian condition, one can estimate the beam waist, or focal spot
size, 2w0 according to:
4λ F
2w0 = . (B.4)
π D
where λ = 1.56 × 10−6 m, focal length of the lens F = 300 mm and beam diameter
before the lens D = 4 mm. The beam waist radius is hence w0 = 74.5 µm
99
APPENDIX B. THIRD-HARMONIC GENERATION IN METALLIC THIN FILM
Field strength of incident pump

Intensity of the incident beam at the focal spot can be calculated from the power:
P1 2.5 × 104 W
I1 = 2
= 2
= 1.43 × 1012 W/m2 (B.5)
πw0 −6
π × (74.5 × 10 m)
The field strength A1 of incident field relates to the intensity via I = 2nǫ0 c|A|2 .
r
I1
|A1 | = = 5 × 106 V /m (B.6)
2n1 ǫ0 c
where refractive index of gold at 1560 nm: n1 = 0.53005 + 10.82i (Johnson & Christy),
vacuum permittivity ǫ0 = 8.854 × 10−12 F/m and speed of light c = 3 × 108 m/s.
Field strength of THG

The field strength of THG estimated according to Eq. 2.10.11 1 :
i3ω (3) 3
A3 = χ A1 J3 (∆k, z0 , z) (B.7)
2n3 c
where ∆k = 3k1 − k3 and J3 (∆k, z0 , z) represents the integration of complex amplitude
A3 (z) over the length of interaction region from z0 to z. In the limiting case where the
2πw02
interaction region is much smaller than the depth of focal region |z0 |, |z| ≪ b =
λ
, J3 reduces to:
z
ei∆kz − ei∆kz0
Z
J3 (∆k, z0 , z) = ei∆kz dz ′ = (B.8)
z0 i∆k
or:
∆kL
|J3 (∆k, z0 , z)|2 = L2 sinc2 ( ) (B.9)
2
where L = z − z0 is length of the interaction region.
∆kL
For the perfect phase matching, ∆k = 0, sinc( ) = 1 and thus J3 (∆k, z0 , z) = L.
2
So Eq. B.7 becomes:
i3ω (3) 3
A3 = χ A1 L (B.10)
2n3 c
Taking the interaction region L as the skin depth of the field upon reflection L ≈
10 nm, refractive index of gold at 520 nm: n3 = 0.63512+2.0721i (Johnson & Christy),
2πc
ω= = 1.21 × 1015 rad/s and χ(3) = 2.45 × 10−19 m2 /V 2 , we have |A3 | = 0.85 V /m
λ1
1
Chapter 2.10, Nonlinear Optics 3rd edition, R. W. Boyd
100
B.1. ESTIMATION OF THIRD-HARMONIC GENERATION (THG) IN AU FILM
Power of THG
Again, intensity of THG:
I3 = 2n3 ǫ0 c|A3 |2 = 2.4 × 10−3 W/m2 (B.11)

√
The beam√waist radius of THG is 3 times smaller than that of the incident beam.
So w3 = w0 / 3 = 43 µm. The THG intensity comes from a region having radius w3 ,
hence the power of THG pulse:
P3 = I3 πw32 = 1.41 × 10−11 W (B.12)
If we consider that the duration of THG pulse is similar to that of the excitation
pulse, i.e. 100 f s, then the THG pulse energy:
Ep3 = P3 × τ = 1.41 × 10−11 W × 10−13 s = 1.41 × 10−24 J (B.13)
This is equivalent to the THG average power:
Pave3 = Ep3 × frep = 1.41 × 10−24 W × 40 M Hz = 5.66 × 10−17 W (B.14)
Or numbers of photons per second (not per pulse):
Pave3 5.66 × 10−17 W

N3 = = = 143 photons/s (B.15)
E3ω 3.82 × 10−19 J
Sensitivity of detector
The photomultiplier tube (PMT) Hamamatsu R955 has quantum efficiency at 520 nm
13%, meaning that the number of photoelectrons per second: 0.13 × N3 = 19 elec-
trons/s.
Besides, the PMT has an indication for the photon-limited signal to noise ratio
(SNR), called equivalent noise input (ENI), which refers to the amount of light in
watts to produce SN R = 1. For PMT R955, EN I = 1.3 × 10−16 W . Compare to
Eq. B.14, power of incident light is less than the noise level and can’t be detected.
However, a paper 2 states that THG signal can be enhanced by 103 − 104 times
when the excitation beam is coupled to SP modes. So around the angle of resonance,
we might detect some THG photons.
Unit of nonlinear susceptibility χ(n)

The Gaussian unit (e.s.u.) can be converted to SI units (MKS) via:
4π
χ(2) (m/V ) = 4
χ(2) (e.s.u.) (B.16)
3 × 10
4π
χ(3) (m2 /V 2 ) = χ(3) (e.s.u.) (B.17)
(3 × 104 )2
2
T. Y. F. Tsang, Optics Letter Vol. 21, No. 4, 1996
101
APPENDIX B. THIRD-HARMONIC GENERATION IN METALLIC THIN FILM
B.2 Third-harmonic magneto-plasmonics
4
8000 10
THG THG
7000 8e−3 x3 8e−3 x
3
3
6000 10
THG count rate (cts/s)
THG count rate (cts/s)

5000
2
4000 10
3000
1
2000 10
1000
0
0 10 0 1 2
0 50 100 10 10 10
Average power (mW) Average power (mW)
Figure B.1 – Power dependence of the THG in the trilayer (5 nm) Au/(10 nm)
Co/(15 nm) Ag at angle of incidence 43.9◦ in the normal scale (left) and in logarithmic
scale (right).
The measurement of third-harmonic magneto-plasmonics was performed on trilayer

(5 nm) Au/(10 nm) Co/(15 nm) Ag, where the SPP-induced SH emission was the
largest among the three samples of trilayer structure. The fundamental radiation was
provided by the MaiTai laser, center wavelength 1560 nm, repetition rate 80 MHz,
pulse duration < 120 fs. The setup is similar to the one described in Fig. 3.3, except
that the low-pass filter in front of the PMT was changed to a band-pass filter FB520
(Thorlabs) in order to select third-harmonic emission at 520 nm.
2000
THG(+M)
THG(−M)
THG intensity (counts/s)
1500
1000
500
0
40 42 44 46 48 50
Incident angle inside prism (deg)
Figure B.2 – The angular dependence of the THG in the trilayer (5 nm) Au/(10 nm)
Co/(15 nm) Ag at opposite magnetizations (±M ).
102
B.2. THIRD-HARMONIC MAGNETO-PLASMONICS
The fluence dependence of the THG is plotted in Fig. B.1. The dependence I3ω ∝ Iω3
was preserved upto the average power of 80 mW, proving that the detected signal is
the TH emission. Next, the scans over the angles of incidence were performed under an
external magnetic field By applied perpendicular to the plane of incidence. The THG
signals for the two opposite magnetizations are presented in Fig. B.2. As expected from
the estimation above, the SPP-enhanced peak at angle 43.9◦ is an order-of-magnitude
higher than the noise level, and is detectable by the registered setup. Similar to the
case of the SHG peak, this characteristic peak is the result of the fundamental field
enhancement of the linear SPP(ω). Recalling that the TH output involves both the
bulk and the surface contributions, the nonlinear SPP(3ω) takes effect within a volumn
skin depth. Thus, describing plasmonic THG would be more complicating than the SH
case. We also note that there is no visible magnetization-induced variations of THG.
As the bulk could be the dominant source of TH emission, the thin cobalt layer might
not generate sufficient magnetic contrast. However, if the cobalt layer is too thick, the
optical loss will increase accordingly, limiting the effect of the radiative SPP.
103
Abbreviations
ATIR attenuated total internal reflection

EM electromagnetic
BBO Beta barium borate
BS beam splitter
LSPR localized surface plasmon resonance
mSHG magnetization-induced second-harmonic generation
MOKE magneto-optical Kerr effect
PMT photomultiplier tube
SH second-harmonic
SHG second-harmonic generation
SLR surface lattice resonance
SP surface plasmon
SPP surface-plasmon polariton
SPR surface plasmon resonance
TE transverse electric
TM transverse magnetic
B.2. THIRD-HARMONIC MAGNETO-PLASMONICS
List of publications
• Tom Seifert, Ngoc Minh Tran, Oliver Gueckstock, Seyed Mohammedreza Rouze-
gar, Lukas Nadvornik, Samridh Jaiswal, Gerhard Jakob, Vasily V. Temnov,
Markus Muenzenberg, Martin Wolf, Mathias Klaeui and Tobias Kampfrath:
Terahertz spectroscopy for all-optical spintronic characterization of
the spin-Hall-effect metals Pt, W and Cu80 Ir20 ; (Special Issue on Spin-
caloritronic) J. Phys. D: Appl. Phys. 51(36), 364003(2018), pp8 ; [DOI: 10.1088/
1361-6463/aad536]; Publication Date (Online/WEB): Jul 23, 2018.
• Ngoc-Minh Tran, Ioan-Augustin Chioar, Aaron Stein, Alexandr Alekhin, Vincent

Juvé, Gwenaëlle Vaudel, Ilya Razdolski, Vassilios Kapaklis and Vasily Temnov:
Observation of the nonlinear Wood’s anomaly on magneto-photonic
crystals of nickel nanodimers; Physics Review B (submitted); [arXiv:1811.
04189]
• Ngoc-Minh Tran, Christian Traum, Alexandr Alekhin, Dmitry A. Kuzmin, Vin-

cent Juvé, Gwenaëlle Vaudel, Ilya Razdolski, Denys Makarov, Denis Seletskiy and
Vasily V. Temnov: Resonantly enhanced nonlinear magneto-plasmonic
interactions in hybrid metal-ferromagnet multilayer structures; Physics
Review Applied (in preparation).
105
Bibliography
[1] R W Wood. On a remarkable case of uneven distribution of light in a diffraction

grating spectrum. Proceedings of the Physical Society of London, 18(1):269, 1902.
[2] Stefan Alexander Maier. Plasmonics: fundamentals and applications. Springer

Science & Business Media, 2007.
[3] Gaspar Armelles, Alfonso Cebollada, Antonio García-Martín, and María Ujué
González. Magnetoplasmonics: combining magnetic and plasmonic functionali-
ties. Advanced Optical Materials, 1(1):10–35, 2013.
[4] H. J. Simon, D. E. Mitchell, and J. G. Watson. Optical second-harmonic gener-

ation with surface plasmons in silver films. Phys. Rev. Lett., 33:1531–1534, Dec
1974.
[5] Markus B Raschke, Samuel Berweger, and Joanna M Atkin. Ultrafast and nonlin-
ear plasmon dynamics. In Plasmonics: Theory and Applications, pages 237–281.
Springer, 2013.
[6] Ilya Razdolski, Denys Makarov, Oliver G. Schmidt, Andrei Kirilyuk, Theo Ras-
ing, and Vasily V. Temnov. Nonlinear surface magnetoplasmonics in kretschmann
multilayers. ACS Photonics, 3(2):179–183, 2016.
[7] Vasily V Temnov, Ilya Razdolski, Thomas Pezeril, Denys Makarov, Denis Se-
letskiy, Alexey Melnikov, and Keith A Nelson. Towards the nonlinear acousto-
magneto-plasmonics. Journal of Optics, 18(9):093002, 2016.
[8] P Haefner, E Luck, and E Mohler. Magnetooptical properties of surface

plasma waves on copper, silver, gold, and aluminum. physica status solidi (b),
185(1):289–299, 1994.
[9] Yakov M. Strelniker and David J. Bergman. Transmittance and transparency of

subwavelength-perforated conducting films in the presence of a magnetic field.
Phys. Rev. B, 77:205113, May 2008.
[10] B. Sepúlveda, J. B. González-Díaz, A. García-Martín, L. M. Lechuga, and

G. Armelles. Plasmon-induced magneto-optical activity in nanosized gold disks.
Phys. Rev. Lett., 104:147401, Apr 2010.
107
BIBLIOGRAPHY
[11] Juan Bautista González-Díaz, Antonio García-Martín, G Armelles, José Miguel

García-Martín, C Clavero, Alfonso Cebollada, RA Lukaszew, JR Skuza,
DP Kumah, and Roy Clarke. Surface-magnetoplasmon nonreciprocity effects in
noble-metal/ferromagnetic heterostructures. Physical Review B, 76(15):153402,
2007.
[12] Vasily V Temnov, Gaspar Armelles, Ulrike Woggon, Dmitry Guzatov, Alfonso
Cebollada, Antonio Garcia-Martin, Jose-Miguel Garcia-Martin, Tim Thomay,
Alfred Leitenstorfer, and Rudolf Bratschitsch. Active magneto-plasmonics in
hybrid metal–ferromagnet structures. Nature Photonics, 4(2):107, 2010.
[13] Diana Martín-Becerra, Juan B González-Díaz, Vasily V Temnov, Alfonso Cebol-

lada, Gaspar Armelles, Tim Thomay, Alfred Leitenstorfer, Rudolf Bratschitsch,
Antonio García-Martín, and María Ujué González. Enhancement of the magnetic
modulation of surface plasmon polaritons in au/co/au films. Applied Physics Let-
ters, 97(18):183114, 2010.
[14] Jorge F Torrado, Juan B González-Díaz, Antonio García-Martín, and Gaspar

Armelles. Unraveling the relationship between electromagnetic field intensity
and the magnetic modulation of the wave vector of coupled surface plasmon
polaritons. New Journal of Physics, 15(7):075025, 2013.
[15] V. V. Pavlov, G. Tessier, C. Malouin, P. Georges, A. Brun, D. Renard, P. Meyer,

J. Ferré, and P. Beauvillain. Observation of magneto-optical second-harmonic
generation with surface plasmon excitation in ultrathin au/co/au films. Applied
Physics Letters, 75(2):190–192, 1999.
[16] VI Belotelov, IA Akimov, M Pohl, VA Kotov, S Kasture, AS Vengurlekar,

Achanta Venu Gopal, DR Yakovlev, AK Zvezdin, and M Bayer. Enhanced
magneto-optical effects in magnetoplasmonic crystals. Nature Nanotechnology,
6(6):370, 2011.
[17] Jessie Yao Chin, Tobias Steinle, Thomas Wehlus, Daniel Dregely, Thomas Weiss,
Vladimir I. Belotelov, Bernd Stritzker, and Harald Giessen. Nonreciprocal plas-
monics enables giant enhancement of thin-film faraday rotation. Nat. Commun.,
4:1599, 2013.
[18] Alexander L Chekhov, Alexander I Stognij, Takuya Satoh, Tatiana V Murzina,

Ilya Razdolski, and Andrzej Stupakiewicz. Surface plasmon-mediated nanoscale
localization of laser-driven sub-terahertz spin dynamics in magnetic dielectrics.
Nano letters, 18(5):2970–2975, 2018.
[19] T W Ebbesen, H J Lezec, H F Ghaemi, T Thio, and P A Wolff. Extraordinary

optical transmission through sub-wavelength hole arrays. Nature, 391:667, feb
1998.
108
BIBLIOGRAPHY
[20] Henri J Lezec, A Degiron, E Devaux, RA Linke, L Martin-Moreno, FJ Garcia-

Vidal, and TW Ebbesen. Beaming light from a subwavelength aperture. Science,
297(5582):820–822, 2002.
[21] William L Barnes, Alain Dereux, and Thomas W Ebbesen. Surface plasmon
subwavelength optics. nature, 424(6950):824, 2003.
[22] Henri J Lezec and Tineke Thio. Diffracted evanescent wave model for enhanced
and suppressed optical transmission through subwavelength hole arrays. Optics
express, 12(16):3629–3651, 2004.
[23] C Genet and TW Ebbesen. Light in tiny holes. In Nanoscience And Technology:
A Collection of Reviews from Nature Journals, pages 205–212. World Scientific,
2010.
[24] GA Wurtz, W Hendren, R Pollard, R Atkinson, L Le Guyader, Andrei Kirilyuk,

Th Rasing, II Smolyaninov, and AV Zayats. Controlling optical transmission
through magneto-plasmonic crystals with an external magnetic field. New Jour-
nal of Physics, 10(10):105012, 2008.
[25] Jorge F Torrado, Juan B González-Díaz, María U González, Antonio García-

Martín, and Gaspar Armelles. Magneto-optical effects in interacting localized
and propagating surface plasmon modes. Optics express, 18(15):15635–15642,
2010.
[26] G Ctistis, E Papaioannou, P Patoka, J Gutek, P Fumagalli, and M Giersig.

Optical and magnetic properties of hexagonal arrays of subwavelength holes in
optically thin cobalt films. Nano letters, 9(1):1–6, 2008.
[27] Evangelos Th Papaioannou, Vassilios Kapaklis, Emil Melander, Björgvin Hjör-

varsson, Spiridon D Pappas, Piotr Patoka, Michael Giersig, Paul Fumagalli, An-
tonio Garcia-Martin, and Georgios Ctistis. Surface plasmons and magneto-optic
activity in hexagonal ni anti-dot arrays. Optics express, 19(24):23867–23877,
2011.
[28] Emil Melander, Erik Östman, Janine Keller, Jan Schmidt, Evangelos Th Pa-
paioannou, Vassilios Kapaklis, Unnar B Arnalds, Blanca Caballero, Antonio
García-Martín, Juan Carlos Cuevas, et al. Influence of the magnetic field on
the plasmonic properties of transparent ni anti-dot arrays. Applied Physics Let-
ters, 101(6):063107, 2012.
[29] I Razdolski, DG Gheorghe, Emil Melander, Björgvin Hjörvarsson, P Patoka,

AV Kimel, A Kirilyuk, E Th Papaioannou, and Th Rasing. Nonlocal nonlinear
magneto-optical response of a magnetoplasmonic crystal. Physical Review B,
88(7):075436, 2013.
[30] Vasily V Temnov. Ultrafast acousto-magneto-plasmonics. Nature Photonics,

6(11):728, 2012.
109
BIBLIOGRAPHY
[31] Viktor Shalagatskyi. Ultrafast acoustics in hybrid and magnetic structures. PhD
thesis, Université du Maine, 2015.
[32] Tobias Kampfrath, Marco Battiato, Pablo Maldonado, G Eilers, J Nötzold, Se-
bastian Mährlein, V Zbarsky, F Freimuth, Y Mokrousov, S Blügel, et al. Tera-
hertz spin current pulses controlled by magnetic heterostructures. Nature nan-
otechnology, 8(4):256, 2013.
[33] Alexandr Alekhin, Ilya Razdolski, Nikita Ilin, Jan P. Meyburg, Detlef Diesing,
Vladimir Roddatis, Ivan Rungger, Maria Stamenova, Stefano Sanvito, Uwe
Bovensiepen, and Alexey Melnikov. Femtosecond spin current pulses generated
by the nonthermal spin-dependent seebeck effect and interacting with ferromag-
nets in spin valves. Phys. Rev. Lett., 119:017202, Jul 2017.
[34] Tom Seifert, S Jaiswal, U Martens, J Hannegan, Lukas Braun, Pablo Maldonado,
F Freimuth, A Kronenberg, J Henrizi, I Radu, et al. Efficient metallic spintronic
emitters of ultrabroadband terahertz radiation. Nature photonics, 10(7):483–488,
2016.
[35] Tom Seifert, Ngoc Minh Tran, Oliver Gueckstock, Seyed Mohammedreza Rouze-
gar, Lukas Nadvornik, Samridh Jaiswal, Gerhard Jakob, Vasily V Temnov,
Markus Muenzenberg, Martin Wolf, et al. Terahertz spectroscopy for all-optical
spintronic characterization of the spin-hall-effect metals pt, w and cu _{80} ir
_{20}. arXiv preprint arXiv:1805.02193, 2018.
[36] H. A. Wierenga, M. W. J. Prins, D. L. Abraham, and Th. Rasing.

Magnetisatization-induced optical second-harmonic generation: A probe for in-
terface magnetism. Phys. Rev. B, 50:1282–1285, Jul 1994.
[37] Ru-Pin Pan, HD Wei, and YR Shen. Optical second-harmonic generation from
magnetized surfaces. Physical Review B, 39(2):1229, 1989.
[38] Andrei Kirilyuk and Theo Rasing. Magnetization-induced-second-harmonic gen-

eration from surfaces and interfaces. JOSA B, 22(1):148–167, 2005.
[39] J Reif, JC Zink, C-M Schneider, and J Kirschner. Effects of surface magnetism on
optical second harmonic generation. Physical review letters, 67(20):2878, 1991.
[40] Robert W Boyd. Nonlinear optics. Elsevier, 2003.
[41] Yuen-Ron Shen. The principles of nonlinear optics. New York, Wiley-
Interscience, 1984, 575 p., 1984.
[42] Tony F Heinz. Second-order nonlinear optical effects at surfaces and interfaces.
In Modern Problems in Condensed Matter Sciences, volume 29, pages 353–416.
Elsevier, 1991.
110
BIBLIOGRAPHY
[43] T. M. Crawford, T. J. Silva, C. W. Teplin, and C. T. Rogers. Subnanosecond

magnetization dynamics measured by the second-harmonic magneto-optic kerr
effect. Applied Physics Letters, 74(22):3386–3388, 1999.
[44] Wei Zheng, Aubrey T Hanbicki, Berry T Jonker, and Gunter Lüpke. Surface
plasmon-enhanced transverse magnetic second-harmonic generation. Optics ex-
press, 21(23):28842–28848, 2013.
[45] Baptiste Auguié and William L. Barnes. Collective resonances in gold nanopar-
ticle arrays. Physical Review Letters, 101(14):1–4, 2008.
[46] V G Kravets, F Schedin, and A N Grigorenko. Extremely narrow plasmon res-

onances based on diffraction coupling of localized plasmons in arrays of metallic
nanoparticles. Physical Review Letters, 101(8):87403, aug 2008.
[47] Wei Zhou and T. W. Odom. Tunable subradiant lattice plasmons by out-of-plane
dipolar interactions. Nature Nanotech, 6:423–427, 2011.
[48] Robert Czaplicki, Antti Kiviniemi, Janne Laukkanen, Joonas Lehtolahti, Markku
Kuittinen, and Martti Kauranen. Surface lattice resonances in second-harmonic
generation from metasurfaces. Optics letters, 41(12):2684–2687, 2016.
[49] Lior Michaeli, Shay Keren-Zur, Ori Avayu, Haim Suchowski, and Tal Ellenbogen.
Nonlinear surface lattice resonance in plasmonic nanoparticle arrays. Physical
review letters, 118(24):243904, 2017.
[50] Dmitry A. Kuzmin, Igor V. Bychkov, Vladimir G. Shavrov, and Vasily V. Tem-
nov. Plasmonics of magnetic and topological graphene-based nanostructures.
Nanophotonics, 7(3):597–611, 2018.
[51] AA Grunin, AG Zhdanov, AA Ezhov, EA Ganshina, and AA Fedyanin. Surface-

plasmon-induced enhancement of magneto-optical kerr effect in all-nickel sub-
wavelength nanogratings. Applied Physics Letters, 97(26):261908, 2010.
[52] Wei Zhou, Yi Hua, Mark D Huntington, and Teri W Odom. Delocalized lat-
tice plasmon resonances show dispersive quality factors. The journal of physical
chemistry letters, 3(10):1381–1385, 2012.
[53] M Kataja, TK Hakala, Aleksi Julku, MJ Huttunen, Sebastiaan van Dijken, and
Päivi Törmä. Surface lattice resonances and magneto-optical response in mag-
netic nanoparticle arrays. Nature communications, 6:7072, 2015.
[54] AV Chetvertukhin, AA Grunin, AV Baryshev, TV Dolgova, H Uchida, M In-

oue, and AA Fedyanin. Magneto-optical kerr effect enhancement at the wood’s
anomaly in magnetoplasmonic crystals. Journal of Magnetism and Magnetic
Materials, 324(21):3516–3518, 2012.
111
BIBLIOGRAPHY
[55] Valentina Bonanni, Stefano Bonetti, Tavakol Pakizeh, Zhaleh Pirzadeh, Jianing
Chen, Josep Nogués, Paolo Vavassori, Rainer Hillenbrand, Johan Åkerman, and
Alexandre Dmitriev. Designer magnetoplasmonics with nickel nanoferromagnets.
Nano letters, 11(12):5333–5338, 2011.
[56] Irina Zubritskaya, Kristof Lodewijks, Nicolò Maccaferri, Addis Mekonnen,
Randy K Dumas, Johan Åkerman, Paolo Vavassori, and Alexandre Dmitriev.
Active magnetoplasmonic ruler. Nano letters, 15(5):3204–3211, 2015.
[57] Nanfang Yu and Federico Capasso. Flat optics with designer metasurfaces. Na-
ture Materials, 13:139–150, 2014.
[58] Alexander V. Kildishev, Alexandra Boltasseva, and Vladimir M. Shalaev. Planar
photonics with metasurfaces. Science, 339(6125), 2013.
[59] Robert W Wood. Anomalous diffraction gratings. Physical Review, 48(12):928,
1935.
[60] Daniel Maystre. Theory of wood’s anomalies. In Plasmonics, pages 39–83.
Springer, 2012.
[61] Max Born and Emil Wolf. Principles of Optics: Electromagnetic Theory of Prop-
agation, Interference and Diffaction of Light. Pergamon Press, 1959.
[62] Lukas Novotny and Bert Hecht. Principles of nano-optics. Cambridge university
press, 2012.
[63] Hendrik Antoon Lorentz. The Theory of Electrons and Its Applications to
the Phenomena of Light and Radiant Heat: A Course of Lectures Delivered in
Columbia University, New York, in March and April, 1906, volume 29. Columbia
University Press, 1909.
[64] IF Almog, MS Bradley, and V Bulovic. The lorentz oscillator and its applications.
Massachusetts Institute of Technology, 2011.
[65] Paul Drude. The theory of optics. Courier Corporation, 1925.
[66] A. Sommerfeld. Zur elektronentheorie der metalle auf grund der fermischen
statistik. Zeitschrift für Physik, 47(1):1–32, Jan 1928.
[67] N. Del Fatti, C. Voisin, M. Achermann, S. Tzortzakis, D. Christofilos, and F. Val-
lée. Nonequilibrium electron dynamics in noble metals. Phys. Rev. B, 61:16956–
16966, Jun 2000.
[68] Davy Gérard and Stephen K Gray. Aluminium plasmonics. Journal of Physics
D: Applied Physics, 48(18):184001, 2014.
[69] C. Guo, G. Rodriguez, A. Lobad, and A. J. Taylor. Structural phase transition
of aluminum induced by electronic excitation. Phys. Rev. Lett., 84:4493–4496,
May 2000.
112
BIBLIOGRAPHY
[70] Nick N. Lepeshkin, Aaron Schweinsberg, Giovanni Piredda, Ryan S. Bennink,

and Robert W. Boyd. Enhanced nonlinear optical response of one-dimensional
metal-dielectric photonic crystals. Phys. Rev. Lett., 93:123902, Sep 2004.
[71] J.-Y. Bigot, V. Halté, J.-C. Merle, and A. Daunois. Electron dynamics in metallic
nanoparticles. Chemical Physics, 251(1):181 – 203, 2000.
[72] Honghua U Yang, Jeffrey D’Archangel, Michael L Sundheimer, Eric Tucker,

Glenn D Boreman, and Markus B Raschke. Optical dielectric function of sil-
ver. Physical Review B, 91(23):235137, 2015.
[73] Zhenlin Wang, Che Ting Chan, Weiyi Zhang, Naiben Ming, and Ping Sheng.
Three-dimensional self-assembly of metal nanoparticles: Possible photonic crys-
tal with a complete gap below the plasma frequency. Physical Review B,
64(11):113108, 2001.
[74] P. B. Johnson and R. W. Christy. Optical constants of the noble metals. Phys.
Rev. B, 6:4370–4379, Dec 1972.
[75] Pablo G Etchegoin, EC Le Ru, and M Meyer. An analytic model for the optical
properties of gold. The Journal of chemical physics, 125(16):164705, 2006.
[76] MA Ordal, LL Long, RJ Bell, SE Bell, RR Bell, RW Alexander, and CA Ward.

Optical properties of the metals al, co, cu, au, fe, pb, ni, pd, pt, ag, ti, and w in
the infrared and far infrared. Applied optics, 22(7):1099–1119, 1983.
[77] J-Y Bigot, J-Y Merle, O Cregut, and A Daunois. Electron dynamics in copper
metallic nanoparticles probed with femtosecond optical pulses. Physical review
letters, 75(25):4702, 1995.
[78] C-K Sun, F Vallée, LH Acioli, EP Ippen, and JG Fujimoto. Femtosecond-tunable

measurement of electron thermalization in gold. Physical Review B, 50(20):15337,
1994.
[79] JG Fujimoto, JM Liu, EP Ippen, and N Bloembergen. Femtosecond laser interac-

tion with metallic tungsten and nonequilibrium electron and lattice temperatures.
Physical Review Letters, 53(19):1837, 1984.
[80] Christophe Voisin, Natalia Del Fatti, Dimitris Christofilos, and Fabrice Vallée.
Ultrafast electron dynamics and optical nonlinearities in metal nanoparticles,
2001.
[81] Daniel Gall. Electron mean free path in elemental metals. Journal of Applied
Physics, 119(8):085101, 2016.
[82] Stefan Visnovsky. Optics in magnetic multilayers and nanostructures. CRC

Press, 2006.
113
BIBLIOGRAPHY
[83] Ml Freiser. A survey of magnetooptic effects. IEEE Transactions on magnetics,

4(2):152–161, 1968.
[84] John Kerr. Xliii. on rotation of the plane of polarization by reflection from the
pole of a magnet. The London, Edinburgh, and Dublin Philosophical Magazine
and Journal of Science, 3(19):321–343, 1877.
[85] Michael Faraday. On the magnetization of light and the illumination of magnetic
lines of force. Philosophical Transactions of the Royal Society of London, 136:1–
20, 1846.
[86] Karl Heinz Bennemann. Non-linear optics in metals. Number 98. Oxford Uni-
versity Press, 1998.
[87] J Zak, ER Moog, C Liu, and SD Bader. Universal approach to magneto-optics.

Journal of Magnetism and Magnetic Materials, 89(1-2):107–123, 1990.
[88] JA Armstrong, N Bloembergen, J Ducuing, and PS Pershan. Interactions be-

tween light waves in a nonlinear dielectric. Physical review, 127(6):1918, 1962.
[89] B Koopmans, AM Janner, HA Wierenga, Th Rasing, GA Sawatzky, and

F van der Woude. Separation of interface and bulk contributions in second-
harmonic generation from magnetic and non-magnetic multilayers. Applied
Physics A, 60(2):103–111, 1995.
[90] P Guyot-Sionnest and YR Shen. Bulk contribution in surface second-harmonic

generation. Physical Review B, 38(12):7985, 1988.
[91] Victor Mizrahi and John E Sipe. Phenomenological treatment of surface second-
harmonic generation. JOSA B, 5(3):660–667, 1988.
[92] Jerry I Dadap, Jie Shan, and Tony F Heinz. Theory of optical second-harmonic
generation from a sphere of centrosymmetric material: small-particle limit. JOSA
B, 21(7):1328–1347, 2004.
[93] Fu Xiang Wang, Francisco J. Rodríguez, Willem M. Albers, Risto Ahorinta, J. E.

Sipe, and Martti Kauranen. Surface and bulk contributions to the second-order
nonlinear optical response of a gold film. Physical Review B, 80(23):233402, 2009.
[94] G. S. G. S. Krinchik and V. A. Artem’ev. Magneto-optical properties of ni, co,

and fe in the ultraviolet visible, and infrared parts of the spectrum. Sov. Phys.
JETP, 26:1080–1085, 1968.
[95] H RAETHER. Surface plasmons on smooth and rough surfaces and on gratings.
Springer Tract in Modern Physics, 111, 1988.
[96] Bo E Sernelius. Surface modes in physics. John Wiley & Sons, 2011.
114
BIBLIOGRAPHY
[97] Erwin Kretschmann and Heinz Raether. Radiative decay of non radiative surface
plasmons excited by light. Zeitschrift für Naturforschung A, 23(12):2135–2136,
1968.
[98] Andreas Otto. Excitation of nonradiative surface plasma waves in silver by the
method of frustrated total reflection. Zeitschrift für Physik A Hadrons and nuclei,
216(4):398–410, 1968.
[99] Stefano Palomba and Lukas Novotny. Nonlinear excitation of surface plasmon
polaritons by four-wave mixing. Phys. Rev. Lett., 101:056802, Aug 2008.
[100] J. J. Burke, G. I. Stegeman, and T. Tamir. Surface-polariton-like waves guided

by thin, lossy metal films. Phys. Rev. B, 33:5186–5201, Apr 1986.
[101] Nicolai B. Grosse, Jan Heckmann, and Ulrike Woggon. Nonlinear plasmon-
photon interaction resolved by k-space spectroscopy. Phys. Rev. Lett.,
108:136802, Mar 2012.
[102] Jonathan J. Foley IV, Hayk Harutyunyan, Ralu Divan Daniel Rosenmann,
Gary P. Wiederrecht, and Stephen K. Gray. When are surface plasmon po-
laritons excited in the kretschmann-raether configuration? Sci. Rep., 5:9929,
2015.
[103] T Y Tsang. Surface-plasmon-enhanced third-harmonic generation in thin silver

films. Optics letters, 21(4):245–247, 1996.
[104] Jiřı Homola, Ivo Koudela, and Sinclair S Yee. Surface plasmon resonance sensors
based on diffraction gratings and prism couplers: sensitivity comparison. Sensors
and Actuators B: Chemical, 54(1-2):16–24, 1999.
[105] Jiří Homola and Marek Piliarik. Surface plasmon resonance (spr) sensors. In
Surface plasmon resonance based sensors, pages 45–67. Springer, 2006.
[106] Pochi Yeh. Optical waves in layered media, volume 61. Wiley-Interscience, 2005.
[107] Pochi Yeh, Amnon Yariv, and Chi-Shain Hong. Electromagnetic propagation in
periodic stratified media. i. general theory. JOSA, 67(4):423–438, 1977.
[108] Zdenek Knittl. Optics of thin films: an optical multilayer theory, volume 1. Wiley
London:, 1976.
[109] Charles Fabry. Theorie et applications d’une nouvelle methods de spectroscopie

intereferentielle. Ann. Chim. Ser. 7, 16:115–144, 1899.
[110] Diana Martín-Becerra, Vasily V Temnov, Tim Thomay, Alfred Leitenstorfer,

Rudolf Bratschitsch, Gaspar Armelles, Antonio García-Martín, and María Ujué
González. Spectral dependence of the magnetic modulation of surface plas-
mon polaritons in noble/ferromagnetic/noble metal films. Physical Review B,
86(3):035118, 2012.
115
BIBLIOGRAPHY
[111] Yu Gong, Alan G. Joly, Dehong Hu, Patrick Z. El-Khoury, and Wayne P. Hess.
Ultrafast imaging of surface plasmons propagating on a gold surface. Nano Let-
ters, 15(5):3472–3478, 2015.
[112] Martin van Exter and Ad Lagendijk. Ultrashort surface-plasmon and phonon
dynamics. Physical review letters, 60(1):49, 1988.
[113] Vasily V Temnov, Keith Nelson, Gaspar Armelles, Alfonso Cebollada, Tim
Thomay, Alfred Leitenstorfer, and Rudolf Bratschitsch. Femtosecond surface
plasmon interferometry. Optics express, 17(10):8423–8432, 2009.
[114] Dmitri K. Gramotnev and Sergey I. Bozhevolnyi. Plasmonics beyond the diffrac-
tion limit. Nature Photonics, 4(2):83–91, 2010.
[115] Martti Kauranen, Anatoly V Zayats, Rm Osgood, L Cao, Nc Panoiu,

W Fan, S Zhang, Kj Malloy, and Srj Brueck. Nonlinear Plasmonics. Focus,
6(October):737–748, 2009.
[116] John C. Quail and H. J. Simon. Second-harmonic generation from a silver grating
with surface plasmons. J. Opt. Soc. Am. B, 5(2):325–329, Feb 1988.
[117] PHILIPPE Leturcq, J-M Dorkel, ANDRZEJ Napieralski, and ERIC Lachiver.
A new approach to thermal analysis of power devices. IEEE Transactions on
Electron Devices, 34(5):1147–1156, 1987.
[118] R. H. M. Groeneveld, R. Sprik, and A. Lagendijk. Ultrafast Relaxation of Elec-

trons Probed by Sureface Plasmons at a Thin Sliver Film, 1990.
[119] Vasily V. Temnov, Christoph Klieber, Keith A. Nelson, Tim Thomay, Vanessa
Knittel, Alfred Leitenstorfer, Denys Makarov, Manfred Albrecht, and Rudolf
Bratschitsch. Femtosecond nonlinear ultrasonics in gold probed with ultrashort
surface plasmons. Nature Communications, 4:1466–1468, 2013.
[120] Atsushi Kubo, Ken Onda, Hrvoje Petek, Zhijun Sun, Yun S Jung, and Hong Koo
Kim. Femtosecond imaging of surface plasmon dynamics in a nanostructured
silver film. Nano letters, 5(6):1123–1127, 2005.
[121] Atsushi Kubo, Niko Pontius, and Hrvoje Petek. Femtosecond microscopy of
surface plasmon polariton wave packet evolution at the silver/vacuum interface.
Nano letters, 7(2):470–475, 2007.
[122] JG Fujimoto, JM Liu, EP Ippen, and N Bloembergen. Femtosecond laser interac-

tion with metallic tungsten and nonequilibrium electron and lattice temperatures.
Physical Review Letters, 53(19):1837, 1984.
[123] S. D. Brorson, J. G. Fujimoto, and E. P. Ippen. Femtosecond electronic heat-

transport dynamics in thin gold films. Physical Review Letters, 59(17):1962–1965,
1987.
116
BIBLIOGRAPHY
[124] H. Petek and S. Ogawa. Femtosecond time-resolved two-photon photoemission

studies of electron dynamics in metals. Progress in Surface Science, 56(4):239–
310, 1997.
[125] E Beaurepaire, J-C Merle, A Daunois, and J-Y Bigot. Ultrafast spin dynamics
in ferromagnetic nickel. Physical review letters, 76(22):4250, 1996.
[126] Jean-Yves Bigot, Mircea Vomir, and Eric Beaurepaire. Coherent ultrafast mag-
netism induced by femtosecond laser pulses. Nature Physics, 5(7):515, 2009.
[127] Andrei Kirilyuk, Alexey V Kimel, and Theo Rasing. Ultrafast optical manipu-
lation of magnetic order. Reviews of Modern Physics, 82(3):2731, 2010.
[128] Jean-Claude Diels, Wolfgang Rudolph, Paul F Liao, and Paul Kelley. Ultra-
short laser pulse phenomena: fundamentals, techniques, and applications on a
femtosecond time scale, optics and photonics. Academic, New York, 1996.
117

2018LEMA1029

Transféré par

Droits d'auteur :

Formats disponibles

2018LEMA1029

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

2018LEMA1029

Transféré par

Droits d'auteur :

Formats disponibles

THESE DE DOCTORAT DE

ECOLE DOCTORALE N° 596

Thèse présentée et soutenue au Mans, le 13 décembre 2018

Rapporteurs avant soutenance : Composition du Jury :

Mots clés : magneto-plasmonique, génération de seconde harmonique, nanostructures hybrides

Title : Applications of nonlinear magneto-photonics at the nanoscale

Keywords : magneto-plasmonics, magnetic second harmonic generation, hybrid metal-ferromagnet nanostructures,

2.1 Comparison of the dielectric function of silver estimated by Drude model

3.1 Kretschmann conﬁguration applied for the metal/ferromagnet bilayers

4.1 Illustrations of linear and nonlinear diﬀraction orders in reﬂection on a

B.1 Power dependence of the TH output in the trilayer . . . . . . . . . . . 102

2.1 Interband transition threshold and plasmon energy of common plas-

3.1 Set of samples used in this work . . . . . . . . . . . . . . . . . . . . . . 33

A.1 Estimation of relative thicknesses between diﬀerent layers in the hybrid

1 Introduction and Motivation 1

2 Light matter interaction 5

3 Nonlinear magneto-plasmonics on Kretschmann configuration 31

4 Nonlinear magneto-optics in diffraction anomalies 63

5 Femtosecond time-resolved nonlinear magneto-plasmonic cross-correlation

6 Summary and conclusion 93

Appendix A Technical specifications 97

Appendix B Third-harmonic generation in metallic thin film 99

Introduction and Motivation

makes use of extraordinary optical transmission in subwavelength hole arrays [19–23]

Light matter interaction

2.1 Light propagation in matters

For a monochromatic ﬁeld, the time-dependent electric ﬁeld can be presented as

In general, the spectrum of a time-dependent ﬁeld E(r, t) is deﬁned by the Fourier

In the macroscopic regime, the response of the material to an electromagnetic ﬁeld

In isotropic homogeneous non-magnetic media, the oﬀ-diagonal elements of the ǫ tensor

2.2 Linear magneto-optical responses

Metal Einter (eV) λinter (nm) ~ωp (eV)

2.2.1 Drude model and optical properties of metals

P (t) = N µ(t) = −N qr(t) (2.15)

∂ 2 [P (ω)e−iωt ] ∂[P (ω)e−iωt ]

And ﬁnally, we obtain:

where GD = −ω 2 − iγω denotes the Drude oscillation term 8 .

2.2.2 Linear magneto-optical effects

2.2.3 Magneto-optical Kerr effect (MOKE)

(a) (b) (c)

Polar Longitudinal Transversal

rps (Mz ) rps (Mx ) ∆Rpp (My )

2.3 Nonlinear magneto-optical responses

2.3.1 Model of anharmonic oscillation and SHG

approximated by power expansion. The force in this case is

Anharmonic oscillation of a free electron

z(t) = Λzω (t) + Λ2 z2ω (t)

q ∂ 2 zω (t) ∂zω (t)

Scaling with Λ - Linear harmonic oscillation

q ∂ 2 zω (t) ∂zω (t)

Scaling with Λ2 - Second-order anharmonic oscillation

In the macroscopic regime, the Taylor expansion of the polarization is:

P = ΛP (ω) + Λ2 P (2ω) + . . . (2.42)

P (2ω) = −N qz(2ω) = ǫ0 χ(2) E 2 (ω) (2.43)

where the second-order nonlinear susceptibility χ(2) is expressed by

Re[Λ 2 z2ω (t)]

SHG at the surface and interfaces

P (2ω) = Psurf (2ω) + Pbulk (2ω)

2.3.2 Magnetization-induced Second Harmonic Generation

2.4 Surface plasmons