Universite D'aix-Marseille

UNIVERSITE D’AIX-MARSEILLE ED 352 PHYSIQUE ET SCIENCES DE LA MATIÈRE FACULTÉ DES SCIENCES Laboratoire d'accueil: UMR7345 - Physique des Interactions Ioniques et Moléculaires

Thèse présentée pour obtenir le grade universitaire de docteur

Discipline : Energie, Rayonnement et Plasma

Kostiantyn ACHKASOV

Study of negative ion surface production in cesium-free

H2 and D2 plasmas: application to neutral beam injectors for ITER and DEMO

Soutenue le 9 décembre 2015 devant le jury composé de :

Ane AANESLAND LPP, Ecole Polytechnique CR, Rapporteur Freddy GABORIAU LAPLACE, Université de Toulouse MC, Rapporteur Ivo FURNO EPFL, Suisse MER, Examinateur Alix GICQUEL LSPM, Université de Paris Nord Professeur, Examinateur Piergiorgio SONATO Consorzio RFX, Université de Padova Professeur, Examinateur Jean-Marc LAYET PIIM, Aix-Marseille Université Professeur, Examinateur Gilles CARTRY PIIM, Aix-Marseille Université Professeur, Directeur de thèse Alain SIMONIN CEA Cadarache Ingénieur-chercheur, Co-directeur

Contents

Résumé français i English abstract iii Résumé français élargi v List of acronyms xxiv Introduction and motivation 1 1. Negative ion surface production measurements 11 1.1 General principle of measurements 11 1.2. Experimental set-up 12 1.2.1. PHISIS 12 1.2.2. Materials 15 1.2.3. Mass spectrometer 16 1.3. Experimental protocol 19 1.3.1. Representation of spectra 19 1.3.2. RF plasma 20 1.3.3. ECR plasma 23 1.4. Conclusion 27 2. Modeling and reconstruction of NIEDF 28 2.1. Modeling of NIEDF for sample normal to the MS axis 29 2.1.1. Description of the model 29 2.1.2. SRIM software 30 2.1.3. Calculation of NI trajectories 31 2.1.4. Change of surface coverage 33 2.1.5. Effects of surface roughness 36 2.2. Modeling of NIEDF for different tilt angles of the sample 39 2.3. Conclusions on the NIEDF modeling 44 2.4. Reconstruction of surface-produced NIEDF 44 2.4.1. Determination of adjacent angular ranges 45 2.4.2. Collection efficiency 45 2.5. Results and discussion 50 2.5.1. Experimental conditions 50 2.5.2. HOPG 56 2.5.3. Gadolinium 64 2.6. Conclusion 72 3. NI production on different materials 73

Contents

3.1. Molybdenum: background measurements 74 3.2. Carbon materials 76 3.2.1. Graphitic materials 76 3.2.2. Diamond materials 79 3.3. Metals 85 3.3.1. Gadolinium 85 3.3.2. Heated tungsten 90 3.3. Conclusion 95 4. Surface state characterization of NI enhancers 98 4.1. Raman spectroscopy 99 4.1.1. Basics of Raman spectroscopy 99 4.1.2. Raman experimental set-up 101 4.1.3. Raman micro-spectrometer calibration 103 4.1.4. Raman spectroscopy of carbons 105 4.1.5. Raman spectroscopy of carbons exposed to plasma 108 4.2. Temperature programmed desorption 113 4.2.1. TPD principle and experimental conditions 113 4.2.2. Mass spectrometer calibration 116 4.2.3. In situ measurements on HOPG 118 4.2.4. Ex situ measurements on HOPG 120 4.2.5. Correlation of Raman spectroscopy with TPD for HOPG 128 4.2.6. Measurements on diamond films 132 4.2.7. Correlation of Raman spectroscopy with TPD for MCBDD 143 4.2.8. Comparison of total desorption amount 147 4.3. Conclusion 150 5. Pulsed-bias approach 152 5.1. General principle of measurements 152 5.2. Experimental results 157 5.2.1. Results for a conductive sample: HOPG 157 5.2.2. Results for an insulating material: MCD 162 5.2.3. Results for heated materials 172 5.3. Conclusion and perspectives 178 General conclusion 180 Bibliography 184 Publications 193

Résumé français

L'objectif de cette thèse était trouver des solutions pour produire de hauts rendements d’ions négatifs (IN) H–/D– sur des surfaces dans des plasmas de H2/D2 sans Cs pour des applications en fusion thermonucléaire. Ce travail a été effectué en utilisant une source de plasma à haute densité et basse pression PHISIS. Les IN sont produits sur un échantillon polarisé négativement dans le plasma. Une partie des ions positifs attirés par l'échantillon capture des électrons sur la surface et produit les IN qui sont ensuite détectés en fonction de leur énergie par un spectromètre de masse. La forme des fonctions de distribution en énergie des ions négatifs (FDEIN) mesurés dépend d'abord du mécanisme de production en surface, puis des trajectoires des ions dans le plasma et dans le spectromètre de masse. La modélisation des FDEIN a montré un accord remarquable avec l'expérience pour les matériaux carbonés. Une méthode de reconstruction mis au point dans le cadre de cette thèse a permis de déterminer les distributions en énergie et en angle des IN émis de la surface. La méthode a été validée par un bon accord des calculs SRIM avec les distributions reconstituées à partir des données expérimentales pour le graphite. Après vérification, la méthode de reconstruction a été utilisée pour caractériser la production des IN sur la surface d’un métal à faible travail de sortie (Gd) qui a donné un bon accord avec les calculs SRIM. L'algorithme de reconstruction ne dépend pas du mécanisme de production des IN en surface ni du matériel utilisé. Par conséquent, la méthode de reconstruction peut être appliquée à tout type de surface et/ou d’IN. Une étude de la production des IN en surface a été réalisée sur une grande variété de matériaux (des différents types de graphite, couches de diamant et métaux). L'influence sur le rendement des IN de la température de surface, de la tension de polarisation et du temps d'exposition au plasma a été étudiée. Une méthode de polarisation pulsée a été développée pour permettre l'étude de production des IN sur les surfaces de matériaux isolants tels que le diamant microcristallin non dopé. Nous avons montré que pour toutes les couches diamant le rendement passait par le maximum autour de 400°C – 500°C. Nous avons également montré qu’en mode pulsé le rendement était 2-3 fois supérieur à celui du régime continu. L'utilisation de diagnostics de surface ex situ tels que la désorption programmée en température (DPT) et la spectroscopie Raman ont permis de caractériser l'état de surface des matériaux carbonés. Les analyses de DPT et de spectroscopie Raman ont été effectuées sur les échantillons exposés au plasma à différentes températures de surface, afin de mettre en corrélation l'évolution du rendement des IN avec la température et le changement d'état de surface.

Résumé français

L’ensemble des études a permis de montrer que pour optimiser le rendement des IN sur le diamant, il faut travailler avec une surface moins dégradée. Celle ci peut être obtenu en augmentant la température de surface jusqu’à 400°C – 500°C ce qui permet de restaurer les propriétés intrinsèques des diamants. L'état de surface moins dégradé peut également être obtenu en appliquant une polarisation pulsée qui donne la possibilité de diminuer les défauts induits par l'exposition au plasma.

Mots clés : interactions plasma-surface, ions négatifs, injecteurs de neutres, spectrométrie de masse, plasmas à basse pression, désorption programmée en température, spectroscopie Raman

English abstract

The objective of this thesis was to find solutions to produce high yields of H–/D– negative ions (NI) on surfaces in Cs-free H2/D2 plasmas for thermonuclear fusion applications. This work was conducted using a low-pressure high-density plasma source PHISIS. NI are produced on a sample negatively biased in the plasma. Some of the positive ions attracted by the sample capture electrons at the surface and produce NI detected by a mass spectrometer according to their energy. The shape of the measured negative-ion energy distribution functions (NIEDF) depends first on the surface production mechanism and then on the trajectories of the ions in the plasma and the mass spectrometer. Modeling of NIEDF has shown remarkable agreement with experiment for carbon materials. The reconstruction method developed in the course of this thesis has allowed to determine the distribution in energy and angle of NI emitted from the surface. The method was validated by the good agreement of SRIM calculations with the distributions reconstructed from experimental data for highly oriented pyrolitic graphite (HOPG). After the verification, the reconstruction method was used to characterize NI production on the surface of a low work-function metal (Gd) which has given a good agreement with SRIM calculations. The algorithm of reconstruction does not depend either on the NI surface production mechanism or on the used material. Therefore, the reconstruction method can be applied to any type of surface and/or NI. A study was performed on a large variety of materials: different types of graphite, diamond films and metals. The influence of surface temperature, bias and plasma exposure time on NI yield was investigated. The method of pulsed bias was developed to enable the study of NI production on surfaces of insulating materials such as microcrystalline non-doped diamond (MCD). It was shown that for all diamond films the NI yield had a maximum at around 400°C–500°C. We have also shown that in pulse bias mode the yield was 2-3 times higher than in continuous bias mode. The use of ex situ surface diagnostics such as temperature programmed desorption (TPD) and Raman spectroscopy has allowed to characterize the surface state of carbon materials. TPD and Raman analyses were performed on the samples exposed to plasma at different surface temperatures in order to correlate the NI yield evolution with temperature to the surface state changes. Basing on the performed studies, we demonstrated that to optimize the NI yield on diamond one has to work with a less degraded surface. This can be obtained rising the surface temperature to 400°C–500°C which allows restoring intrinsic properties of diamond. The less degraded surface state can also be obtained by applying the pulsed iii

English abstract

bias which gives the possibility to increase the H2/D2 surface coverage and diminish the defects induced by plasma exposure.

Keywords : plasma-surface interactions, negative ions, neutral beam injectors, mass spectrometry, low-pressure plasmas, temperature programmed desorption, Raman spectroscopy

Résumé français élargi

Introduction et motivation

1) Principes de la fusion par confinement magnétique

La réaction de fusion nécessitant la plus basse température ionique Ti qui a été choisi pour être utilisée dans le réacteur thermonucléaire expérimental international (ITER) se lit comme suit [1]:

1D2 + 1T3 → 2He4 + 0n1 (17.6 MeV) Eq. (0.1) L’énergie totale acquise dans le processus réactionnel est réparti entre les produits: 20% de l’énergie est porté par les He+ de 3.5 MeV (que l’on appelle particules α), tandis que les neutrons de 14.1 MeV obtient les 80% restants. La particule α reste confinée par le champ magnétique de l’appareil de fusion et contribue par conséquent au chauffage du plasma par des collisions avec des ions et des électrons. Le neutron n’est pas chargé, de sorte qu’il s’échappe du plasma et atteint la paroi de l’enceinte à vide où il transfère son énergie cinétique au fluide circulant à l’intérieur de la couverture. Ce fluide qui contient du lithium comme agent de refroidissement sert aussi à produire le tritium par les réactions suivantes:

0n1 + 3Li6 → 1T3 + 2He4 (4.8 MeV) Eq. (0.2)

0n1 + 3Li7 → 1T3 + 2He4 + 0n1 (–2.5 MeV) Eq. (0.3) La première réaction a une section efficace plus grande pour les neutrons thermiques, et la seconde – pour les neutrons rapides. Donc, comme les neutrons à haute énergie seront thermalisés par les collisions avec des atomes de la couverture, ils peuvent subir des réactions qui augmentent la quantité de produits de réaction. De cette manière, le réacteur de fusion va créer son propre combustible avec une fiabilité élevée [2]. En fin de compte, le liquide de refroidissement chaud produit par l’absorption des neutrons est utilisé pour produire de l’électricité. Afin d’atteindre l’ignition (soit une réaction de fusion autonome sans contribution d’alimentation externe) le gain de puissance de fusion doit dépasser les pertes. Le produit triple de densité d’électrons ne, température ionique Ti et temps de confinement

τE définit la valeur minimale nécessaire pour initier la réaction de fusion [3]:

ne · τE · Ti ≥ 3 · 1021 [keV · m−3 · s] Eq. (0.4)

Résumé français élargi

qui par exemple serait atteinte par ne = 1020 m-3, Ti = 10 keV et τE = 3 s. Le temps de confinement τE définit la vitesse à laquelle le système perd de l’énergie vers son environnement (par exemple par rayonnement de freinage) [4]. Comme les plasmas de fusion ont besoin de températures de l’ordre de centaines de millions de kelvin (~ 10 keV) pour atteindre l’allumage, aucun matériau serait capable de résister à de telles températures. Pour cette raison un champ magnétique est utilisé pour confiner le plasma pendant un temps suffisamment long τE et pour le séparer des murs. Les forts champs magnétiques sont créés par des bobines supraconductrices. Afin de confiner le plasma de manière efficace, il faut des lignes de champ magnétique qui se ferment sur elles-mêmes au sein de l’enceinte à vide, de sorte que les particules chargées ne frappent pas le mur. On pourrait également utiliser l’effet de miroir magnétique avec des lignes de champ ne se fermant pas sur elles-mêmes au sein de la chambre au plasma, mais cette méthode a montré une moindre efficacité (faible temps de confinement par rapport aux machines actuelles de fusion). Donc, la géométrie d’un tore a été choisie ce qui a abouti à l’apparition des concepts tels que les tokamaks et les stellarators. Le mot tokamak est une translittération du russe "токамак" - chambre toroïdale avec bobines magnétiques (тороидальная камера с магнитными катушками). Le dispositif a été inventé dans les années 1950 par les physiciens soviétiques Igor Tamm et Andreï Sakharov qui ont été inspirés par l’idée originale de Oleg Lavrentiev. Le champ magnétique d’un tokamak est constitué de deux composantes: toroïdale et poloïdale. La composante de champ toroïdal est produite par des courants dans les bobines entourant le tore comme représenté sur la Figure I. Le champ est produit en utilisant de nombreuses bobines de champ toroïdal, pour produire un champ aussi uniforme que possible [5]. Ainsi, les particules suivant les lignes de champ toroïdales fermées doivent rester dans l’enceinte de confinement. Cependant, avec une telle géométrie de la bobine, le champ magnétique sur la face intérieure du tore serait plus forte que sur le côté

Figure I. Les bobines entourant le tore qui produisent le champ toroïdal [5]. vi

Résumé français élargi

extérieur (Btor ~ 1 / r), parce que les bobines sont situées plus près les unes des autres au centre du tore. En conséquence, les particules vont subir une dérive dans le sens opposé pour les électrons et les ions. Cela signifie une séparation de charge qui va créer un champ électrique à l’intérieur de l’enceinte à vide et pousser l’ensemble du plasma vers la paroi extérieure à cause de la force E × B. Afin de compenser cette dérive, il faut réaliser une torsion des lignes du champ. Elle est obtenue en ajoutant une composante du champ magnétique poloïdal Bθ qui se superpose à la composante de champ toroïdal

Bφ pour créer un champ magnétique hélicoïdal entièrement contenu dans la chambre de confinement [2]. Ce champ magnétique entourant est capable de tenir le plasma en place et à assurer une balance des forces. Une ligne de champ magnétique après avoir fait un grand nombre de tours définit effectivement une surface fermée qui est appelée une surface de flux magnétique. Les surfaces de flux magnétique représentant différentes lignes de champ hélicoïdaux sont imbriquées les unes dans les autres formant des surfaces concentriques. Les lignes de champ magnétique ne peuvent pas traverser les surfaces de flux. Par conséquent, les particules qui sont piégées en suivant les lignes de champ restent piégées sur une surface de flux magnétique donnée [6]. La dernière surface de flux magnétique fermée (i.e. la surface qui ne touche aucun des composants de l’enceinte de plasma) est appelée la séparatrice. Dans les tokamaks le champ magnétique poloïdal est généré par le courant toroïdal circulant dans le plasma. Dans les expériences actuelles le courant de plasma est créé par un champ électrique toroïdal qui en induit par un enroulement primaire (ou solénoïde central) qui agit comme le primaire d’un transformateur, le secondaire étant le plasma (voir la Figure II) [3]. C’est généralement le noyau de fer du transformateur qui génère le changement de flux à travers le tore car il nécessite une puissance

Figure II. (a) Un changement du flux magnétique à travers le tore induit un champ électrique toroïdal qui génére le courant toroïdal. (b) La variation de flux est produit par l'enroulement primaire souvent l'aide d'un transformateur [3]. vii

Résumé français élargi

relativement faible et a des champs magnétiques parasites réduits [7]. La composante du champ magnétique toroïdal est généralement un ordre de grandeur plus grande que la composante poloïdale, ce qui donne aux tokamaks des caractéristiques de stabilité de plasma favorables. En outre, des bobines supplémentaires sont généralement installées pour contrôler la position verticale du plasma, sa forme et sa diagonalité. Le plasma dans un tokamak est naturellement chauffé par la puissance de chauffage ohmique POH = Ip2R, où Ip est le courant poloïdal et R est la résistance électrique de plasma associée. Le chauffage ohmique devient inefficace pour Te > 2 keV (parce que résistivité du plasma diminue), de sorte qu’un chauffage additionnel est nécessaire pour atteindre l’ignition.

2) Chauffage du plasma

Il existe deux types de chauffage additionnel du plasma: chauffage par les radiofréquences (RF) et injection de faisceau de neutres. Le chauffage RF implique des hautes puissances électromagnétiques injectées dans le plasma qui transfèrent l’énergie par des interactions résonantes avec les particules du plasma. La fréquence de résonance est choisie pour correspondre à la fréquence cyclotron des ions ou des électrons: le Ion Cyclotron Resonance Heating (ICRH) fonctionnera dans ITER à une fréquence allant de 40 MHz à 55 MHz, tandis que le Electron Cyclotron Resonance Heating (ECRH) va fonctionner à 170 GHz [8]. L’injection des neutres (IdN) consiste à injecter un faisceau de haute énergie d’atomes neutres, généralement du deutérium, dans le cœur du plasma de fusion. Le faisceau neutre pénètre dans le plasma du tokamak sans aucune déviation lié au champ magnétique jusqu’au centre du tore, où il interagit avec des ions et des électrons du plasma. Une fois que l’atome est ionisé, il est piégé par le champ magnétique de tokamak et il est ralenti par de multiples interactions avec des électrons et des ions du plasma. L’énergie de faisceau choisie va dépendre de la densité et du volume de plasma; pour un chauffage efficace du plasma, la puissance additionnelle doit être absorbée au cœur du plasma. En plus du chauffage, l’IdN ravitaille en "carburant" le plasma par l’injection d’atomes. Il fournit également la dynamique toroïdal angulaire (rotation du plasma), qui est bénéfique car elle permet d’améliorer le confinement du plasma [9], contribue à la génération de courant non inductif et supprime certaines instabilités du plasma (modes de déchirement, etc.). La génération de courant non inductif se réfère à la production d’un courant toroïdal en plus du courant d’induction, afin de permettre au réacteur de fusion de fonctionner de façon continue (sans courant inductif). L’IdN permet la production d’un faisceau d’ions positifs chargés qui circule autour du tore et transfère une quantité de mouvement toroïdal par collisions résultant en un courant toroïdal net. Pour une énergie du faisceau très au-dessus de l’énergie critique (de 1 à 2 MeV) le

viii

Résumé français élargi

transfert de quantité de mouvement se produit principalement sur les électrons du plasma ce qui augmente considérablement l’efficacité de génération du courant dans le plasma [8]. Jusqu’à présent, les systèmes d’IdN ont été utilisés avec succès sur les grands tokamaks en exploitation dans le monde entier. Les systèmes d’IdN ont apporté des hautes températures et des performances de fusion élevées, et les scénarios de plasmas avancés ont été atteints. La ligne de faisceau neutre pour ITER (représenté sur la Figure III) sera composée de deux systèmes d’IdN. Chaque système a une longueur de 15 m, une largeur de 4.7 m et une hauteur de 5.3 m (avec la traversée haute tension, la hauteur totale atteinte est 9 m). Le composant principal d’un injecteur typique est une source d’ions qui est constituée d’un plasma froid d’hydrogène ou de deutérium (Te ~ 2 eV à 20 eV, ne ~ 1018 – 1019 m-3). Les ions extraits de la source sont accélérés à une énergie élevée (~ 1 MeV pour ITER) par un accélérateur électrostatique grâce à la polarisation des électrodes d’accélération (appelés grilles) par des tensions continues élevées. Dans les systèmes d’IdN classiques, le faisceau ionique passe à travers d’une cellule de gaz

(appelé agent neutralisant) où les collisions inélastiques avec du gaz injecté (D2) se produisent, ce qui conduit à une neutralisation partielle du faisceau (~55% de neutralisation pour les ions D– à 1 MeV en cas d’ITER). Les ions restants (D+ et D–) sont déviés magnétiquement ou électrostatiquement sur une surface de cuivre refroidi à l’eau (dépotoir d’ions résiduels), laissant le faisceau neutre continuer vers le tore [8]. Comme la taille du futur réacteur de fusion ITER est plus grande que celle des tokamaks modernes, l’énergie du faisceau d’IdN doit être plus élevée afin de déposer la puissance du faisceau dans le cœur du plasma: 250 à 500 keV / uma (D0 de 0.5 MeV à 1 MeV). Avec une source d’ions positifs (IP) de deutérium, l’efficacité de neutralisation diminue fortement au-dessus de 100 keV, et aucune neutralisation ne peut se produire au-dessus de 500 keV, comme on peut le voir sur la Figure IV. Dans l’IdN installé sur la plupart des machines de fusion modernes, tel que JET, les faisceaux neutres ont une énergie de 100 keV.

2a) L’injection de neutres d’ITER

Pour ITER nécessitante des faisceaux de 1 MeV de D0, l’utilisation du système d’IdN basé sur l’accélération et la neutralisation d’IP est donc impossible. Par contre, il est facile de détacher un électron supplémentaire de H–/D– dans une cellule de gaz (neutralisateur), avec un rendement de neutralisation de ~ 55%, même à haute énergie de faisceau (voir la Figure IV) [10]. L’utilisation d’IN représente ainsi un grand intérêt pour la production des faisceaux neutres. Pourtant, la faible énergie de liaison de cet électron (0.75 eV pour H–) qui permet un détachement facile crée une énorme difficulté pour la production d’IN et ralentit l’exploitation de ces systèmes. Le succès de la future génération de machines de fusion repose sur le chauffage efficace fourni par le système ix

Résumé français élargi

Figure III. Schéma de la ligne de faisceau neutre pour ITER avec la description des principaux composants et les réactions qui ont lieu.

100 90 80

70 - D 60 50 40 (%) Fraction 30 + 20 D

0 0 200 400 600 800 1000 Energy of D+ or D- (keV)

Figure IV. L'efficacité maximale de neutralisation (fraction en%) des ions de deutérium en fonction d’énergie [8].

Résumé français élargi

d’IdN de haute énergie nécessitant des courants d’IN intenses. Il a été calculé que pour le chauffage efficace des plasmas de fusion, le système d’IdN d’ITER aura besoin d’un courant stable de H–/D– à 1 MeV avec 69 A en hydrogène et 57 A en deutérium [11–13]. Par conséquent, le développement d’une source d’IN efficace devient particulièrement important. Les IN dans les plasmas à basse pression sont généralement créés par l’attachement dissociatif des électrons du plasma froid (< 1 eV) sur les molécules excitées vibrationnellement [14, 15]. Ce processus est appelé la production d’IN en volume, et il est utilisé dans de nombreuses applications allant de la microélectronique [16–18] jusqu’à fusion par confinement magnétique [19, 20], y compris la propulsion spatiale [21] et les sources d’accélérateurs de particules [22–25]. Toutefois, les IN peuvent également être produits par ionisation de surface sur les matériaux en contact avec le plasma. Dans la plupart des études expérimentales et de modélisation des plasmas à basse pression, la production des IN en surface est généralement un mécanisme d’importance mineure. Cependant, dans certaines circonstances, elle peut être très efficace. Lorsque les surfaces de métaux alcalins tels que le césium (Cs) sont en contact avec le plasma, une énorme production des IN en surface par conversion des IP ou des atomes est observée [26]. Cet effet est la base des sources les plus intenses des ions négatifs H–/D– développées pour les applications de fusion ou des accélérateurs de particules [23, 27, 28]. La source conçue pour ITER est composée de huit générateurs de plasma à couplage inductif (ICP), appelés "drivers", chacun alimenté avec environ 100 kW de puissance RF. La source doit générer un flux uniforme d’IN (avec une uniformité de ± 5%) à la matrice de 1280 ouvertures d’extraction sur la face avant de l’accélérateur (grille plasma). Étant créé dans les "drivers" avec Te ~ 15 eV, le plasma se propage dans la région de diffusion. Un champ magnétique transversal de ~ 50 Gauss est appliquée entre la région de diffusion et la région d’extraction. Le champ filtrant agit comme une barrière magnétique qui refroidit les électrons chauds générés dans les "drivers" et empêche un taux de destruction élevé des IN formés sur la grille plasma. Dans cette région, la densité d’électrons est réduite et Te ~ 1 eV. Le césium est évaporé à l’intérieur de l’enceinte et est déposé sur les surfaces, en particulier sur la grille plasma [29]. Cela réduit le travail de sortie du matériau de grille jusqu’à ~ 2.3 eV et permet à certains neutres (D0) de capturer un électron au niveau de Fermi par effet tunnel en quittant la surface. Actuellement, cette méthode est considérée comme le seul moyen de répondre aux besoins d’ITER dans des délais courts. Une source d’IN RF à l’échelle1/2 de celle d’ITER (ELISE) est en étude en Allemagne (IPP Garching) [30]. Des expériences dédiées au transport du plasma à travers le champ magnétique sont menées au laboratoire LAPLACE (Toulouse, France) [31]. Il est également prévu de construire une installation IdN d’essai, PRIMA (Padova Research on ITER Megavolt Accelerator), en Italie. Elle comprendra une source d’ions négatifs en pleine échelle, SPIDER, et un prototype complet d’injecteur d’ITER, MITICA. Le but est de développer les injecteurs d’ITER [32, 33]. En parallèle, des travaux xi

Résumé français élargi

de modélisation sur les sources d’ions utilisant le Cs ont été réalisées dans plusieurs centres de recherche européens tels que IPP Garching, laboratoire LAPLACE (Toulouse, France) [34–37], l’Université de Sofia et l’Université Technique de Sofia (Bulgarie) [38], la plus part étant en collaboration avec le CEA Cadarache.

2b) IdN pour DEMO

Depuis 2006, un nouveau programme de recherche cible l’émergence d’une nouvelle génération de systèmes d’IdN pour le futur réacteur de fusion DEMO (DEMOnstration Power Plant). Des travaux sont en cours entre plusieurs laboratoires en Europe. Les spécifications requises pour faire fonctionner un système d’IdN sur DEMO sont très exigeantes: le système doit fournir le chauffage du plasma, la génération du courant et le contrôle du plasma à très haut niveau de puissance (jusqu’à 150 MW) et d’énergie (1 ou 2 MeV). Le système doit en outre avoir des performances élevées en termes de rendement (η > 60%), une haute disponibilité et une grande fiabilité.

2c) Projet Siphore

Compte tenu de cet objectif, un nouveau concept d’IdN nommé Siphore a été proposé par Alain Simonin du CEA Cadarache [39]. Il est basé sur le photodétachement du faisceau énergétique d’ions négatifs. La clé de voûte de ce nouveau concept est le développement d’un photoneutraliseur dans lequel un flux de photons de haute puissance (~ 3 MW) généré dans une cavité de Fabry Perot se croise et photo-détache partiellement le faisceau intense d’IN accéléré aux hautes énergies. Il a eté montré qu’un tel système d’IdN basé sur la photoneutralisation aurait la capacité de fournir plusieurs dizaines de MW de D0 par ligne de faisceau avec un rendement supérieure à 60%. Une étude de faisabilité du concept a été lancée entre différents laboratoires (y compris le PIIM) pour aborder les différents aspects de la physique, i.e., la production d’ions négatifs, la modélisation du plasma, la simulation de l’accélérateur d’ions, la photoneutralisation et la tenue haute tension sous vide [39]. La source d’ions Cybèle [39–41] est une source d’ions haute et étroite avec un rapport d’aspect rectangulaire qui est prévu pour être utilisé dans Siphore. Elle est actuellement une source de plasma filamentaire dans laquelle des ensembles de 3 à 5 filaments de tungstène sont utilisés comme cathodes le long de l’axe vertical de la source. Les filaments fournissent au cœur du plasma des électrons primaires le long de l’axe vertical. Un champ magnétique uniforme parallèle à l’axe vertical de la source est généré par deux bobines latérales situées sur des côtés opposés d’un cadre rectangulaire de fer qui entoure la source. Les deux bobines génèrent des champs magnétiques dans les directions opposées à l’intérieur de la structure en fer. C’est le champ de fuite entre les deux bobines qui remplit uniformément le volume de la source de plasma. L’intensité du xii

Résumé français élargi

champ magnétique dans le volume entier de plasma peut être réglée entre 0 mT et 7 mT en réglant le courant électrique continu dans les bobines. Cybèle devra répondre aux besoins de la source d’ions d’ITER en termes de densité du courant d’IN (250 A/m²), pression de la source (<0,3 Pa), l’uniformité, etc. Pour obtenir une haute densité de plasma avec un taux d’ionisation élevée et une efficacité énergétique élevée il est nécessaire de se tourner vers des solutions autres que les filaments ou les solutions ICP. La source Helicon est un candidat intéressant pour alimenter Cybèle avec un plasma dense et chaud le long de son axe central. A cet effet, un générateur de plasma helicon de 10 kW et 13.56 MHz est en cours de développement au CRPP-EPFL Lausanne, avec une source helicon basée sur le concept d’un reseau d’antennes résonantes en forme de cage [42–44]. Bien qu’un seul générateur helicon de 10 kW ne puisse probablement pas atteindre la densité du plasma nécessaire pour la source d’IdN de DEMO, la source helicon de 10 kW est une étape intermédiaire vers des puissances plus grandes. Il est prévu de tester des solutions alternatives pour produire les IN dans Cybèle. En effet, des inconvénients graves à l’utilisation de Cs dans les sources d’IN actuelles ont été identifiés. Tout d’abord, afin d’obtenir des courants d’ions négatifs stables sur de pulses longs, une injection de césium continue est nécessaire, ce qui conduit à une forte consommation de césium (~ 3 pg/s [27, 45]). Deuxièmement, la diffusion de césium et la pollution de l’accélérateur peuvent causer des faisceaux parasites et/ou des claquages de tension et impliquent un entretien régulier et restrictif dans un environnement nucléaire. La feuille de route d’Eurofusion a inclus dans la liste des priorités la réduction de la consommation de Cs ou son élimination complète dans les sources d’IN de la génération future. La réduction de la consommation de Cs pourrait être, par exemple, réalisée par l’implantation de Cs dans du molybdène où un progrès récent a été démontré par Schiesko et al [12]. Une autre possibilité est l’utilisation de matériaux alternatifs pour la conversion d’IN avec une efficacité comparable à celle de sources utilisant le Cs. Ainsi, le développement de sources d’IN sans Cs de haute intensité serait très précieux pour la fusion thermonucléaire. Le diamant est un des matériaux qui sont prévus pour être testés dans Cybèle comme convertisseur d’IN. Il est bien connu pour sa capacité à émettre des électrons à haute température et même aux faibles champs électriques [46]. Des expériences de faisceau sur le diamant ont montré une production d’ions H– en surface avec des rendements élevés jusqu’à 5.5%. En outre, il a été observé dans une expérience de plasma que le rendement de production d’IN sur le diamant dopé au bore peut être augmenté d’un facteur 5 en augmentant la température à 400°C [47]. Il a été montré par de nombreux auteurs que le diamant terminé par l’hydrogène présente une affinité électronique négative (sa bande de conduction est au dessus du niveau du vide) [48, 49]. Par conséquent, les électrons peuvent être efficacement émis dans le vide. Toutes ces indices font de diamant un candidat intéressant pour la production d’IN en surface et un des candidats principaux pour être utilisé dans Cybèle. xiii

Résumé français élargi

Cette thèse présente une étude de la production d’IN en surface dans plasmas d’hydrogène et deutérium sans Cs sur différents matériaux (en particulier, le diamant), au sein d’une collaboration sur la R&D du système d’IdN de DEMO. Le travail de recherche est composé de cinq chapitres. Le chapitre 1 "Mesures d’ions négatifs produits en surface" explique le principe général des mesures de production des ions négatifs en surface et donne les détails du dispositif expérimental employé (PHISIS). Le protocole expérimental utilisé pour plusieurs types de mesures y est également précisé. Le chapitre 2 "Modélisation et reconstruction des FDEIN" est dédié à la modélisation et à la reconstruction de la fonction de distribution d’IN produits en surface en énergie et en angle d’émission. Une étude réalisée sur une grande variété de matériaux tels que les différents types de graphite, les couches de diamant et les métaux est présentée dans le chapitre 3 "Production des IN sur matériaux différents". La caractérisation de l’état de surface de convertisseurs d’IN par des diagnostics externes tels que la désorption programmée en température (TPD) et la spectroscopie Raman est décrite dans le chapitre 4 "Caractérisation d’état de surface des convertisseurs d’IN". Le dernier chapitre intitulé "Méthode de polarisation pulsée" fournit des informations sur la méthode de polarisation pulsée développée récemment qu’a permis l’étude de la production d’IN sur des surfaces de matériaux isolants.

Ce projet est mené dans le cadre d’un Programme Européen (Eurofusion) et du projet national français (ANR) "H Index Tripled". Le projet ANR réunit quatre centres de recherche différents: PIIM (Marseille), LSPM (Paris), (Eindhoven, Pays-Bas) et IRFM (CEA Cadarache). Nous avons également reçu le soutien de la région Provence-Alpes- Côte d’Azur (projet PACA GING). Cette thèse de doctorat a été cofinancée par le CEA Cadarache et la région PACA.

1. Mesures d’ions négatifs produits en surface

L’idée principale de la thèse est d’étudier la production d’IN en surface dans des plasmas d’hydrogène et de deutérium à basse pression. L’échantillon est placé à l’intérieur de la chambre à plasma et polarisé négativement par rapport au potentiel du plasma, de sorte que les ions positifs sont accélérés vers la surface de l’échantillon par le champ électrique formé dans la gaine. Le bombardement ionique conduit à la production d’ions négatifs sur la surface. Ces ions négatifs sont accélérés dans la direction opposée et traversent la gaine en face de l’échantillon, la région de plasma et la gaine en face du nez du spectromètre de masse (MS) avant d’arriver au détecteur du MS. Ainsi, la fonction de distribution en énergie des ions négatifs (FDEIN) produits en surface est mesurée. xiv

Résumé français élargi

Les conditions expérimentales choisies dans les plasmas RF sont : 2.0 Pa de pression

H2/D2 avec une puissance RF de 20 W injecté et un écran mis à la terre placé au-dessus de l’échantillon pour minimiser les fluctuations RF. Pour les plasmas ECR les conditions sont les suivantes : 1 Pa H2/D2 avec 60 W de puissance micro-ondes effectivement absorbée (la puissance réfléchie était typiquement 5–10%).

La polarisation négative de la surface de l’échantillon a été choisie à Vs = –130 V pour la plupart des mesures, assurant ainsi une bonne efficacité d’auto-extraction des ions négatifs produits en surface. La population d’ions positifs dominante dans ces conditions

était H3+/D3+ ce qui donne une énergie des ions de 45 eV par nucléon à la surface. Un état stationnaire de la surface sous le bombardement ionique est obtenu après environ 10 minutes d’exposition au plasma pour les échantillons de diamant et de graphite (estimé grâce aux mesures d’évolution temporelle des FDEIN). Les principaux échantillons étudiés dans ce travail ont été le graphite pyrolytique fortement orienté (HOPG), le diamant microcristallin dopé au bore (MCBDD) et le diamant microcristallin non dopé (MCD). 2. Modélisation et reconstruction des FDEIN

Une étude de la production d’IN sur HOPG a été réalisée par Loïc Schiesko au cours de sa thèse [55–57]. Il a prouvé que les IN sont créés sur les surfaces de matériaux de carbone dans nos conditions expérimentales par deux mécanismes: a) rétrodiffusion des ions positifs incidents comme IN; b) pulvérisation d’un atome H/D adsorbé sous forme d’IN par un ion positif incident. Le pic de basse énergie des FDEIN résulte de l’interaction de ces deux mécanismes alors que la queue énergétique ne contient que la contribution de la rétrodiffusion. L’énergie maximale de la FDEIN est définie par l’énergie acquise par l’ion positif dominant dans la gaine entre l’échantillon et le plasma.

Dans le cas de l’ion moléculaire (par exemple H3+), cette énergie doit être divisée par le nombre de fragments (E0/3), parce que tous les ions subissent une dissociation et une neutralisation à l’impact. Ces conclusions ont été prouvées expérimentalement. Ahmad Ahmad au cours de sa thèse a développé un modèle qui peut expliquer la forme des FDEIN mesurées expérimentalement pour un échantillon placé perpendiculairement à l’axe du MS [52, 57]. Le choix initial d’une distribution d’IN sur la surface de l’échantillon (en énergie E et en angle d’émission θ) a été réalisé en utilisant le logiciel SRIM. En se basant sur les E et θ donnés par SRIM, les trajectoires des ions ont ensuite été calculées à l’intérieur de la gaine et dans le plasma, et aussi à l’intérieur de MS. Après avoir subi ces modifications, la distribution d’IN a été comparée aux données expérimentales et a abouti à un accord excellent. L’accord remarquable du modèle avec l’expérience pour HOPG a confirmé que la probabilité d’ionisation d’IN Piz (E, θ) est constante indépendamment de l’énergie des particules neutres et leur angle d’émission. Cela a également vérifié le choix de SRIM pour fournir la bonne proposition initiale

Résumé français élargi

f SRIM(E, θ) pour les matériaux de carbone, à fin que les paramètres d’entrée pour le calcul SRIM sur les couches a-C:H sont bien connus dans la communauté scientifique. Dans cette thèse, l’objectif était d’utiliser les FDEIN expérimentales (mesurées à différents angles d’inclinaison de l’échantillon par rapport à l’axe de MS) afin de reconstruire la distribution initiale d’IN émise par la surface. La méthode de reconstruction permet de s’effranchir de l’utilisation du logiciel SRIM et déterminer la distribution en énergie et en angle des IN émis par la surface. Cette technique peut être utilisée dans tous les cas, en particulier dans le cas où la probabilité de formation des IN n’est pas constante avec E et θ, ou lorsque les paramètres SRIM sont inconnus. La méthode de reconstruction a été validée par le bon accord des calculs SRIM sur HOPG avec les distributions reconstruites à partir de données expérimentales. Après vérification, la méthode de reconstruction a été utilisée pour caractériser la production d’IN sur la surface d’un métal à faible travail de sortie : gadolinium (Gd). Comparaison avec les calculs SRIM, même si elle doit être faite avec beaucoup de soin, a renforcé notre confiance dans la méthode de reconstruction. Il faut mentionner que l’algorithme de reconstruction ne dépend pas du mécanisme de production d’IN en surface. Les seules entrées qui sont nécessaires pour calculer les trajectoires des ions sont les paramètres du plasma et des gaines. Par conséquent, le procédé de reconstruction peut être appliqué à tout type de surface et/ou d’ions négatifs. Un autre point important est que le HOPG a montré un résultat comparable avec Gd en termes de production d’IN en surface, malgré son travail de sortie plus élevé. Cela confirme que les matériaux carbonés sont intéressants pour être utilisés comme des convertisseurs d’IN.

3. Production des IN sur différents matériaux

L’étude présentée dans ce chapitre a été exécutée sur une grande variété de matériaux carbonés. L’influence de la température de surface pendant l’exposition au plasma sur les rendements de H–/D– a été étudiée. Les matériaux carbonés non-diamant de l’étude ont été le graphite pyrolytique fortement orienté (HOPG), les composites de fibre de carbone (CFC) et le carbone amorphe tétraédrique (ta-C). Le HOPG a été choisi comme matériau de référence pour les études de production d’IN en surface à cause de ses propriétés uniques telles que son clivage facile, son rendement d’IN élevé, etc. L’étude des autres matériaux carbonés non-diamant (CFC et ta-C) a été conduite pour comparaison. Le diamant est un des matériaux qui est prévu pour être testé dans une véritable source d’IN (Cybèle) en tant que convertisseur d’IN. La largeur de la bande interdite présente dans le cas d’isolants tels que le diamant réduit la probabilité de capture d’électrons, mais la perte d’électrons par un IN devient également moins probable. Si l’IN xvi

Résumé français élargi

est formé sur la surface isolante sous l’impact d’un IP, il sera déjà éloigné de la surface lorsque son niveau d’énergie entrera en résonance avec les états inoccupés dans la bande de conduction. Par conséquent la probabilité de perte d'électron réduite sur le diamant pourra conduire aux hautes rendements d’IN. Dans ce chapitre, l’étude des couches de diamant polycristallin dopé au bore et de diamant polycristallin non dopé a été effectuée. La taille des cristaux dans les échantillons étudiés a varié de ~ 5 μm (diamants microcristallins: MCBDD, MCD) à 50–100 nm (diamant nanocristallin : NCD). Le gadolinium (Gd) a servi de référence pour les métaux au faible travail de sortie. Son travail de sortie est de seulement 2.9 eV en comparaison avec 4.6 eV pour HOPG et il a une capacité de stockage d’hydrogène élevée [83, 84]. En outre, il est assez facile à manipuler et ne réagit pas rapidement à l’air contrairement au barium, par exemple. Le tungstène (W) est hautement résistant au plasma d’hydrogène et est actuellement le candidat principal en tant que composant face au plasma (PFM) pour le divertor d’ITER. Si il s’avérait efficace en termes de production d’IN, il pourrait être facilement mis en œuvre dans la source d’IN pour les applications de la fusion. Des mesures sur le molybdène (Mo) ont été réalisées afin d’évaluer le signal de fond d’IN, étant donné que le porte-échantillon est réalisé en Mo et pourrait donner une certaine contribution à la FDEIN mesurée. L’évaluation de Mo a révélé qu’il donne des niveaux de signal d’IN suffisamment faibles pour ne pas perturber les mesures. Le comportement de tous les matériaux carbonés non-diamant avec la température de surface est très semblable et démontre une décroissance exponentielle du rendement d’IN à partir de 200°C. Le changement dans la couverture de deutérium en surface ne peut pas expliquer seul la diminution du signal d’IN avec la température ce qui signifie qu’il doit y avoir un changement dans la probabilité d’ionisation. Actuellement, les couches de diamant chauffées sont les meilleurs matériaux pour la production d’IN en surface. Contrairement au comportement des matériaux graphitiques, la production d’IN sur les couches de diamant augmente avec la température de surface jusqu’à 400–500°C et puis commence à diminuer. Cette augmentation d’un facteur ~5 a été observé en H2 et D2. L’augmentation du rendement d’IN de température ambiante à 400°C n’est pas liée à la variation de la couverture de deutérium en surface. Elle peut être expliquée par l’augmentation de la probabilité d’ionisation, car l’on considère que le rendement de pulvérisation et de rétrodiffusion est constant avec la température. On peut également conclure que les différences entre les rendements d’IN sur les matériaux au sein du groupe (graphite ou diamant) provenants de la différence de morphologie initiale de l’échantillon (la rugosité, la taille et l’orientation des grains) sont mineures. Par conséquent, l’état de surface modifié de l’échantillon exposé au plasma et ses propriétés électroniques doivent être les paramètres clés pour expliquer l’évolution du rendement d’IN avec la température. Afin d’obtenir une idée sur l’état de surface des diamants en cours d’exposition au plasma, des mesures d’évolution temporelle ont été effectuées sur l’échantillon MCBDD avec Vs = –130 V (Ei = ~ 45 eV/nucléon) pour des températures de surface différentes. On xvii

Résumé français élargi

a observé que le rendement d’IN à température ambiante est réduit d’un facteur 3 avec le temps, ce qui peut être expliqué par la création des défauts sur la surface. D’autre part, le chauffage de MCBDD à 400°C a augmenté le rendement d’IN presque jusqu’au niveau initial (mesurée sur une surface de MCBDD non modifiée à température ambiante). Nous suggérons que le chauffage de l’échantillon empêche la création de défauts et maintient la surface dans un état favorable pour la production d’IN. Comme il a été précédemment observé par [91], la gravure des défauts sp2 de surface est renforcée par l’augmentation de la température. Par conséquent, l’état de surface, en particulier le nombre de défauts, peut être considéré comme résultant d’un équilibre entre la création de défauts par le bombardement d’ions et leur destruction via la gravure chimique assistée par la présence d’ions. Les mesures sur la surface de Gd ont été réalisées de manière analogue à celles sur des échantillons de carbone. Le signal d’IN sur Gd reste constant jusqu’à une température de surface de 200°C et diminue ensuite, de manière similaire aux matériaux graphitiques. Le rendement d’IN à température ambiante est près de 2 fois plus petit que dans le cas d’HOPG en raison d’une contribution de pulvérisation moins importante. D’autre part, la baisse de rendement avec la température de surface pour Gd est beaucoup moins prononcée que pour HOPG. L’évolution de signaux d’IN avec la polarisation de surface montre que l’augmentation de la tension de polarisation conduit à l’augmentation à la fois du pic principal de FDEIN et de la queue énergétique. Après avoir effectué le cycle de chauffage, les mesures d’évolution en fonction de la polarisation ont été répétées. Il a été remarqué que les intensités des pics des FDEIN à toutes les valeurs de polarisation ont augmenté, montrant l’augmentation de la création d’IN probablement en raison d’une couverture d’hydrogène en surface plus élevée. On peut imaginer que pendant les mesures à haute température, les espèces d’hydrogène diffusent à l’intérieur du matériau créant un réservoir. Afin de suivre l’évolution avec la température des IN pulvérisés, des mesures dans le plasma d’argon ont été effectuées. On peut conclure que le réservoir d’hydrogène à l’intérieur de Gd, une fois qu’il a été créé, ne pouvait pas être complètement éliminé ni par bombardement d’ions Ar+ énergétiques, ni par le chauffage. En comparant notre étude à des investigations précédentes, on pourrait penser que l’augmentation du rendement d’IN observé sur la surface de Gd après l’adsorption d’hydrogène (renforcée par le cycle de chauffage) peut

être reliée à la formation de complexes de GdHx. La conclusion principale de l’étude de la production d’IN sur le filament de tungstène chauffé est qu’il démontre un très faible rendement d’IN par rapport aux matériaux carbonés. On n’a observé aucune augmentation de la production d’IN sur la surface de W lorsque la température a été augmentée de 300 K à plus de 3000 K. Le maximum de production d’IN en surface a été réalisé à environ 1000 ÷ 2000 K. Ces résultats ne sont pas encourageants et semblent indiquer que l’augmentation du nombre d’électrons émis par la surface ne suffit pas à promouvoir la production d’IN.

xviii

Résumé français élargi

4. Caractérisation d’état de surface des convertisseurs d’IN

Les résultats présentés dans les chapitres précédents ont prouvé l’efficacité du système expérimental pour étudier la production d’IN sur des surfaces de différents matériaux. On peut obtenir des informations sur les distributions en énergie et angle des IN émis de la surface, sur la dépendance du rendement d’IN avec la température de surface, avec le temps d’exposition au plasma, etc. Cependant, l’état de surface de l’échantillon dans les diverses conditions d’exposition au plasma n’a pas été bien défini. Quand un échantillon est immergé dans le plasma avec une polarisation négative, il interagit avec les IP incidents et les espèces neutres qui sont présents dans la chambre à plasma. Les premièrs nanomètres de la surface deviennent désordonnés et représentent un arrangement différent de la surface vierge. Le degré de désordre peut dépendre, par exemple, de la vitesse de la gravure assistée par plasma qui, à son tour, dépend de la température de surface de l’échantillon et des conditions du plasma. Tous ces effets ont une influence sur l’efficacité de la production d’IN en surface. Par conséquent, il est apparu nécessaire de caractériser le changement de l’état de surface après exposition au plasma. Comme les matériaux carbonés ont montré le plus grand potentiel pour la production d’IN en surface, il a été décidé d’étudier HOPG et les couches de diamant avec des techniques d’analyse de surface ex situ telles que la spectroscopie Raman et la désorption programmée en température (TPD). L’objectif est d’expliquer la variation du rendement d’IN en fonction de la température de surface et d’obtenir une idée sur l’état des échantillons exposés au plasma à différentes températures de surface. Ce chapitre apporte une contribution à la compréhension de la production d’IN en surface dans des plasmas d’hydrogène et de deutérium grâce à l’analyse de thermodésorption (TPD) et la spectroscopie Raman. Les surfaces de graphite pyrolytique hautement orienté (HOPG), diamant microcristallin dopé au bore (MCBDD) et diamant microcristallin non dopé (MCD) ont été exposées au plasma à des températures de surface différentes et les rendements d’IN ont été mesurés à l’aide de la spectrométrie de masse. L’analyse TPD a été effectuée sur les échantillons exposés afin de corréler l’évolution des rendements d’IN avec la température et le changement de l’état de surface. Les mesures de spectroscopie Raman ont permis la caractérisation d’HOPG et des couches de diamant après l’exposition au plasma. En outre, des mesures de spectroscopie Raman en fonction de la température de surface ont été réalisées pour HOPG et MCBDD précédemment exposés au plasma à la température ambiante ("room temperature" : RT) et à 400°C. Cela a été fait pour corréler avec la TPD et attribuer les pics de désorption TPD à un certain type de défaut ou à un arrangement de surface (phases sp2 ou sp3).

xix

Résumé français élargi

Une recherche bibliographique a permis de caractériser l’état de surface des matériaux pour différentes températures d’exposition au plasma. L’état de surface d’HOPG RT montre les caractéristiques de graphite nanocristallin avec une certaine quantité de défauts sp2 dans le plan et des défauts sp3 hors du plan, avec une teneur en deutérium semblable aux "hard films" a-C:H. Le spectre Raman pour HOPG 400°C a montré une diminution des bandes liées aux défauts dans le plan et hors du plan qui confirme le rétablissement des phases sp2 et l’augmentation de la taille des domaines sp2, et qui corrèle bien avec l’évolution de la teneur en deutérium. Comme prouvé par les mesures de spectroscopie Raman, à 800°C seulement la bande G attribuée au graphite orienté est visible dans le spectre d’HOPG. Ce résultat pourrait être lié à la reconstruction des phases graphitiques sp2 et à l’élimination de presque tous les défauts. La présence de bandes liées aux défauts dans les spectres Raman de MCBDD RT indique la formation de doubles liaisons C=C. Ce fait prouve que l’état de surface de MCBDD RT est désordonné, avec des phases sp2 et sp3 non diamant créés en raison de l’exposition au plasma. Le chauffage du diamant avec l’exposition au plasma permet la reconstruction de la surface originale de diamant grâce à la gravure renforcée des phases sp2 [91]. Quand l’exposition au plasma a été réalisé pendant 5 sec au lieu de 30 min habituelles, le spectre TPD de MCBDD RT a montré les pics de désorption aux mêmes endroits que dans le spectre de MCBDD RT 400°C. Ces résultats suggèrent qu’à 400°C la structure originale de diamant est partiellement reconstruite. Les spectres TPD des couches de diamant chauffées démontrent la désorption de grandes quantités de deutérium sous forme de D2. Pour MCBDD et MCD chauffées à 400°C et 800°C la désorption de deutérium entre 1000 K et 1200 K correspond aux pics observés par divers auteurs [127–129] sur la surface (100) partiellement dégradée du diamant monocristallin. Ceci suggère que le chauffage du diamant avec un ravitaillement constant de D à partir du plasma crée un certain arrangement de D en surface qui permet l’apparition de l’affinité électronique négative (AEN). L’augmentation du rendement d’IN pour les surfaces de diamant chauffées peut être expliquée par la présence d’AEN via la formation de complexes d’hydrures en surface (monohydrure ou dihydrure). La diminution du signal d’IN de 400°C à 800°C est probablement liée à la désorption de deutérium depuis la surface (passage d’une haute à une faible couverture de deutérium). Après la comparaison de spectres TPD des échantillons MCBDD et MCD exposés au plasma à 400°C, on conclut que le rôle du bore dans la production d’IN est insignifiant, même si il permet la conductivité du matériau à température ambiante. La calibration du spectromètre de masse a permis le calcul du montant total des espèces désorbées et leur dynamique. En calculant la concentration atomique de deutérium désorbée sous formes de D2 et CD4 pour des températures de surface différentes, nous avons vu que les propriétés de la couche et les mécanismes de formation des IN sont très différents pour HOPG et les couches de diamant. L’évolution du rendement d’IN est corrélée qualitativement avec la concentration totale du xx

Résumé français élargi

deutérium dans la couche d’interaction, ce qui est en accord avec la modélisation développée précédemment.

5. Méthode de polarisation pulsée

La méthode de polarisation en courant continu pulsé a été développé pour permettre l’étude de la production d’IN sur des surfaces de matériaux isolants (tels que MCD). Une méthode similaire a été utilisée par Samara et al [139] pour mesurer le courant de saturation ionique, ou par Kudlacek et al [140] afin de contrôler l’énergie d’ions dans les procédés de plasma industriels concernant les matériaux isolants. Pour surmonter les limitations de cette technique, Wang et Wendt [141] ont introduit une méthode de pulse modulé, où la tension pendant le pulse est croissante et compense exactement la chute de tension due à la charge du substrat en cours de traitement. Dans cette thèse, seulement des pulses de forme rectangulaire ont été utilisés et les techniques de diagnostic ont été les mêmes que celles utilisées pour étudier la production d’IN sur des surfaces conductrices. Comme il a été mentionné précédemment dans le chapitre 3 "Production des IN sur matériaux différents", la surface de MCD redevient conductrice qu’à partir de 300°C. En dessous de cette température, le signal d’IN ne pouvait pas être mesuré. La présente technique a permis d’étudier la production d’IN sur la surface de MCD pour toute la gamme de températures à partir de température ambiante. De plus, cette technique étend la mesure de la production d’IN sur les surfaces de matériaux isolants potentiellement intéressants. En effet, la capture d’électrons sur des isolants est difficile, mais la perte d’électrons est limitée, ce qui entraîne une probabilité d’ionisation potentiellement élevée. Les tests de polarisation pulsée ont été effectués sur HOPG pour démontrer la faisabilité de la méthode. En modifiant la fréquence des pulses de polarisation (et le rapport cyclique), il a été possible d’obtenir le matériau HOPG avec une couverture d’hydrogène en surface différente et donc un état de surface différent. Ce nouvel état de surface avec la couverture supérieure de H2/D2 en surface réalisé en mode de polarisation pulsée a résulté en une augmentation du rendement d’IN entre 30% et 50% selon les conditions expérimentales. Le changement de la couverture en surface a été également confirmé par la modélisation et expliqué par la présence de la période OFF (sans polarisation) qui permet aux atomes neutres de re-couvrir la surface de l’échantillon. Après avoir prouvé la faisabilité de la méthode de polarisation pulsée sur l’HOPG, l’optimisation des paramètres expérimentaux a été réalisée sur MCD en prenant en compte les effets de charge. Ces paramètres ont été: durée d’impulsion de polarisation Tpulse, durée d’acquisition de spectromètre de masse Tacq, délai entre le début du pulse et l’acquisition du spectromètre Tdelay, la fréquence de polarisation f, la polarisation appliquée Va et la polarisation de surface résultante Vs.

xxi

Résumé français élargi

L’accumulation de charge sur la surface pendant le pulse de polarisation a été explorée en faisant un pulse long et en mesurant les FDEIN avec de courtes Tacq à différents Tdelay (à partir du début jusqu’à la fin du pulse). On en a conclut que la charge de MCD se passe d’une manière linéaire, avec un taux de ~1V/10 μs. Afin de définir les paramètres expérimentaux optimaux, l’influence de la durée de pulse de polarisation

Tpulse sur la forme des FDEIN a été étudiée. Le temps d’acquisition de Tacq = 10 μs a été choisi pour permettre l’acquisition relativement rapide et fournir un décalage de Vs de

~1V. À 1 kHz, la durée de pulse Tpulse = 15 μs (Toff = 985 μs) permet à la surface de se décharger complètement de la charge d’IP accumulée. La polarisation de surface Vs en début de pulse a montré une dépendance non linéaire avec la durée de pulse de polarisation Tpulse. Ce comportement a été expliqué par la perturbation du potentiel de surface exercée par la charge accumulée sur la surface elle même. Deux régimes de décharge pendant la période de pulse OFF ont été identifiés (rapide et lent). Ce raisonnement a été confirmé par des mesures d’IP résolues en temps pour deux durées de pulse de polarisation: 50 μs et 500 μs. L’effet de chargement de la surface visible comme le décalage des FDEIN pourrait apparaître lorsque la durée de polarisation est trop longue, mais aussi lorsque la fréquence de polarisation est trop élevée. L’influence de la fréquence de polarisation a donc été étudiée dans la gamme de 1 à 50 kHz. On peut noter que pour les hautes fréquences l’échantillon de MCD ne peut pas se décharger complètement pendant Toff, car la polarisation arrive trop vite. Pour la durée de polarisation utilisé dans l’expérience

(Tpulse = 10 μs), la fréquence ne doit pas dépasser 10 kHz afin d’éviter le décalage de polarisation de surface. Si une compensation est utilisée pour contrebalancer le changement de Vs induite par la fréquence de polarisation élevée (augmentation de la valeur Va), le problème de polarisation insuffisante peut être résolu, mais seulement à une fréquence inférieure à 10 kHz. Au-dessus de 10 kHz la forme de FDEIN change beaucoup et l’intensité des FDEIN diminue fortement. L’origine de ceci n’a pas été identifiée. Par conséquent, le fonctionnement à f > 10 kHz doit être évité, si possible. Après avoir effectué l’optimisation des paramètres de polarisation et d’acquisition sur le MCD, la méthode choisie a dû être adaptée pour répondre aux mesures d’évolution avec la température. Un protocole expérimental spécial a été créé, qu’est spécifié dans la section 5.2.3. "Résultats pour les matériaux chauffés". Les paramètres de polarisation et d’acquisition choisis sont: 1 balayage, V = [–80 V, 0 V], f = 10 kHz, Tpulse = 15 μs, Tacq = 10

μs, Tdelay = 17 μs, ΔTstep = 100°C. La comparaison des FDEIN pour polarisation constante et pulsée à température ambiante a révélé que le rendement d’IN augmente à peu près d’un facteur 2 pour tout les matériaux : HOPG et la paire MCBDD/MCD. À haute température, un rendement d’IN ~ 4 fois plus élevé a été mesuré sur la surface de MCD dans le cas de la polarisation pulsée par rapport à la polarisation constante. Ceci est le plus haut rendement d’IN qui ait jamais été mesuré sur le dispositif PHISIS, ce qui nous conduit à la conclusion qu’il y a encore de la place pour l’optimisation du rendement d’IN sur les matériaux carbonés. xxii

Résumé français élargi

Les résultats des mesures d’évolution temporelle du rendement d’IN pour HOPG et MCD sont similaires à ceux obtenus dans le cas de la polarisation constante (pour HOPG et MCBDD). La seule différence majeure fut le temps 2–3 fois plus long qui est nécessaire pour atteindre l’état stable pour les échantillons avec la polarisation pulsée (~15 minutes). Les défauts produits sur la surface du au plasma ont un effet favorable sur la conversion d’IN sur l’échantillon d’HOPG, mais ont un effet opposé sur des couches de diamant. Lorsqu’on polarise l’échantillon en mode pulsé, les défauts peuvent être induits sur la surface de MCD seulement pendant la période ON. Par conséquent, l’état de surface est plus proche de l’état de diamant vierge. Les résultats des études sur l’évolution temporelle nous amènent à la conclusion que, dans le cas de la polarisation pulsée, la surface du diamant est moins dégradée et plus hydrogénée, ce qui est favorable pour la production d’IN en surface. Considérant le résultat global des mesures en polarisation pulsée, on peut conclure que pour optimiser le rendement d’IN sur le diamant, il faut travailler avec une surface moins dégradée. Celle-ci peut être obtenu en augmentant la température de surface à 400°C–500°C ce qui permet de restaurer les propriétés intrinsèques des diamants probablement due à la gravure renforcée des phases sp2 [50, 91]. L’état de surface moins dégradé peut également être obtenu en appliquant la polarisation pulsée qui donne la possibilité d’augmenter la couverture de H2/D2 en surface et de diminuer les défauts induits par l’exposition au plasma. Cependant, la polarisation pulsée est une solution uniquement pour les études fondamentales. Ce n’est pas une solution qui pourrait être utilisée dans une véritable source d’IN puisque le rendement absolu d’IN est faible (il est égal au rendement d’IN mesuré pendant la durée de la période de pulse fois le rapport cyclique). Au contraire, la réduction des défauts pourrait ainsi être effectuée en appliquant une polarisation de surface plus petite ce qui également permettrait de s’approcher des conditions de la véritable source d’IN conçue pour ITER. Comme l’extraction des IN dépend également de la valeur de la polarisation de surface, une étude complète doit être effectuée dans le futur.

Mots clés : interactions plasma-surface, ions négatifs, injecteurs de neutres, spectrométrie de masse, plasmas à basse pression, désorption programmée en température, spectroscopie Raman

xxiii

List of acronyms

2D: Two Dimensional 3D : Three Dimensional a-C : amorphous carbon a-C:H : hydrogenated amorphous carbon C : Carbon CCP : Capacitively Coupled Plasma CEA : Commissariat à l’Energie Atomique et aux énergies alternatives, France CFC : Carbon Fiber Composite CX : Charge eXchange D : Deuterium d : distance between the MS and the surface in mm DEMO : DEMOonstration Power Plant DLC : diamond-like carbon

E1 : l’énergie d’entrée des ions dans le plasma

Ei : initial energy of a NI at the surface in eV

EMS : energy of NI at the entrance of the mass spectrometer ECR : Electron Cyclotron Resonance ECRH : Electron Cyclotron Resonance Heating EDS : Energy-dispersive X-ray spectroscopy EELS : Electron Energy Loss Spectroscopy EFDA : European Fusion Development Agreement

Eimpact : impact energy of PI in SRIM calculations

f(θ, E) : NIEDF emitted at the surface calculated by SRIM with a given Eimpact f '(E) : NIEDF collected by the MS calculated by the model f''(E) : NIEDF detected by the MS calculated by the model Gd : gadolinium H : Hydrogen H-mode : High confinement mode He : Helium HOPG : high oriented pyrolitic graphite HREELS : High Resolution Electron Energy Loss Spectroscopy ICP : Inductively Coupled Plasma ICR : Ion Cyclotron Resonance ICRH : Ion Cyclotron Resonance Heating IPP : Max-Planck Institute of Plasma Physics, Garching, Germany IRFM : Institute de Recherche sur la Fusion par confinement Magnétique, CEA xxiv

List of acronyms

Cadarache, France ITER : International Thermonuclear Experimental Reactor JET : Joint European Torus tokamak, Culham, UK L-mode : Low confinement mode LAPLACE : LAboratoire PLAsma et Conversion d’Energie, Université de Toulouse, France LASER : light amplification by stimulated emission of radiation LEED : Low Energy Electron Diffraction LHD : Large Helical Device, National Institute for Fusion Science, Toki, Japan

LN2 : Liquid Nitrogen (at temperature of 77 K) LP : Langmuir Probe MASsoft: a software for measurement and control of the MS MATLAB : MATrix LABoratory: a numerical computing software MC : Monte-Carlo MCBDD : micro-crystalline boron-doped diamond MCD : micro-crystalline diamond (non-doped) MS : mass spectrometer NBI : Neutral Beam Injection NCD : nano-cristalline diamond ne : electronic density in m-3 NI : negative ions NIEDF : negative ion energy distribution function NRA : Nuclear Reaction Analysis O : Oxygen PECVD : Plasma Enhanced Chemical Vapor Deposition PFC : Plasma Facing Component PI : positive ions PWI : Plasma Wall Interactions QMS : Quadrupole Mass Spectrometer RF : Radio Frequency RGA : Residual Gas Analysis SIMION : a software for calculation of ion transport inside the MS SOL : Scrape-Off Layer SRIM : Stopping and Range of Ions in Matter; a group of programs designed to calculate the stopping of ions (10 eV – 2 GeV/a.m.u.) in matter T : Tritium ta-C : tetrahedral amorphous carbon TDS : Thermal Desorption Spectroscopy TPD : Temperature Programmed Desorption

Te : electronic temperature in eV

TMS : transmission inside the MS xxv

List of acronyms

TS Tore Supra : superconducting tokamak in IRFM, Cadarache, France

Vext : extraction potential in V

VMS : MS nozzle potential

Vp : plasma potential in V

Vs : surface potential in V W : Tungsten α : angle between the normal to the sample and the MS axis in °

θi : initial angle of NI emitted from the surface

θ1 : angle of entrance of NI in the plasma θMS : angle of arrival of NI to the MS entrance

xxvi

Introduction and motivation

1) Principles of magnetically confined fusion

The fusion reaction which requires the lowest ion temperature Ti and which was chosen to be used in International Thermonuclear Experimental Reactor (ITER) reads as follows [1]:

1D2 + 1T3 → 2He4 + 0n1 (17.6 MeV) Eq. (0.1) The total energy gained in the reaction process is distributed between the products: 20% of energy is carried by 3.5 MeV He+ (so-called α-particle), while 14.1 MeV neutron gets the remaining 80%. The α-particle stays confined by magnetic field of the fusion machine and hence contributes to plasma heating through collisions with plasma ions and electrons. Neutron is not charged, so it escapes directly to the wall of the vacuum vessel and transfers its kinetic energy to the fluid circulating within the blanket. This fluid which contains lithium serves as coolant as well as tritium breeder via the following reactions: 0n1 + 3Li6 → 1T3 + 2He4 (4.8 MeV) Eq. (0.2)

0n1 + 3Li7 → 1T3 + 2He4 + 0n1 (–2.5 MeV) Eq. (0.3)

The first reaction has larger cross-section for thermal neutrons, and the second one – for fast neutrons. So, as neutron with high energy will be thermalized by colliding with blanket atoms, it may undergo both reactions which increases the amount of reaction products. In such a way, the fusion reactor will create its own fuel with high reliability [2]. In the end, the hot coolant produced by the neutron absorption is used to generate electricity. In order to reach ignition (i.e. a self-sustaining fusion reaction without external power input) the fusion power gain must exceed the losses. The triple product of electron density ne, ion temperature Ti and confinement time τE defines the minimum required value necessary to ignite the fusion reaction [3]:

ne · τE · Ti ≥ 3 · 1021 [keV · m−3 · s] Eq. (0.4) which for example would be reached by ne = 1020 m−3, Ti = 10 keV and τE = 3 s.

Confinement time τE measures the rate at which system loses energy to its environment (e.g. by Bremsstrahlung radiation) [4]. As fusion plasmas need the temperatures of around hundred million of kelvin (~10 keV) to reach the ignition, no materials would be able to withstand such amounts of heat. That’s why a magnetic field is used to confine the plasma for a sufficiently long τE and to separate it from the walls. If the electric fields are used, it would result in charge separation which is undesirable. So, plasmas are confined with a strong magnetic fields

Introduction and motivation

exerted by the superconducting coils around the vacuum vessel. In order to confine the plasma efficiently, one needs magnetic field lines that close on themselves within the vacuum vessel, so that charged particles would not hit the wall. One may also use the magnetic mirror effect which assumes that the field lines do not close on themselves within the plasma chamber, but this approach has proved to be less efficient (low confinement times as compared to the present fusion machines). So, the geometry of a torus was chosen which has resulted in the appearance in fusion science of such concepts as tokamaks and stellarators. The word tokamak is a transliteration from Russian “токамак” – toroidal chamber with magentic coils (тороидальная камера с магнитными катушками). The device was invented in 1950s by Soviet physicists Igor Tamm and Andrei Sakharov who were inspired by the original idea of Oleg Lavrentiev. The magnetic field of a tokamak is made up of two components: toroidal and poloidal. The toroidal field component is produced by currents in coils linking the torus, as shown in Figure 0.1. The field is produced using many toroidal field coils, to produce as uniform a field as possible [5]. So, particles following the closed toroidal field lines should remain within the confinement chamber. However, with such coil geometry, the magnetic field on the inner side of the torus would be stronger than on the outer side (Btor ~ 1/r), because the coils are located closer to each other as they pass through the central core of the torus. As a consequence, particles will undergo a drift opposite in direction for electrons and ions. This means an eventual charge separation that will create an electric field inside the vacuum vessel and push the whole plasma towards the outer wall due to the E×B force. In order to compensate this drift, there must be a twist to the field lines. It is achieved by adding a poloidal magnetic field component Bθ which is superimposed on the toroidal field component Bφ to create a helical magnetic field entirely contained within the confinement chamber [2]. This encircling magnetic field is able to hold the plasma in place and to provide an equilibrium force balance.

Figure 0.1. Coils linking the torus which produce the toroidal magnetic field [5]. 2

Introduction and motivation

A magnetic field line after making a large number of toroidal transits effectively defines a closed surface which is called a magnetic flux surface. Magnetic flux surfaces representing different helical field lines are nested inside one another forming concentric surfaces. Magnetic field lines are not allowed to cross the flux surfaces. Therefore, particles that are trapped following field lines also remain trapped on a given magnetic flux surface [6]. The last closed magnetic flux surface (i.e. the surface that does not touch any components of the plasma vessel) is called the separatrix. In tokamaks poloidal magnetic field is generated by the toroidal current flowing in the plasma itself. In present experiments the plasma current is driven by a toroidal electric field which is induced by a primary winding (or central solenoid) of the transformer, as depicted in Figure 0.2 [3]. It’s usually an iron transformer core which generates the flux change through the torus because it requires relatively low power and has reduced stray magnetic fields [7]. The toroidal magnetic field component is usually an order of magnitude larger than the poloidal one, which gives tokamaks favorable plasma stability characteristics. Moreover, additional coils are usually installed to control the vertical position of the plasma, its shape and diagonality.

A plasma in a tokamak is naturally heated by the Ohmic heating power POH = Ip2R, where Ip is the poloidal current flowing in the plasma and R is the associated plasma electrical resistance. As plasma is heated, Ohmic heating becomes inefficient for Te >2keV (because plasma resistivity decreases), so an additional heating is required to achieve the ignition.

Figure 0.2. (a) A change of the magnetic flux through the torus induces a toroidal electric field which drives the toroidal current. (b) The flux change is produced by primary winding often using a transformer core [3].

Introduction and motivation

2) Plasma heating There are two types of additional plasma heating: radio-frequency (RF) heating and neutral beam injection. RF heating implies high power electromagnetic waves launched into the plasma which transfers energy by resonant interactions with plasma particles. The resonant frequency is chosen to match the cyclotron frequency of ions or electrons: Ion Cyclotron Resonance Heating (ICRH) will operate in ITER at a frequency ranging from 40 MHz to 55 MHz, while Electron Cyclotron Resonance Heating (ECRH) will operate at 170 GHz [8]. Neutral Beam Injection (NBI) involves injecting a high-energy beam of neutral atoms, typically deuterium, into the core of the fusion plasma. The neutral beam penetrates the tokamak plasma without any deflection due to the magnetic field until the center of the torus tube, where it will interact with the plasma ions and electrons. A neutral atom can be ionized at once or in multiple steps via intermediate excited states. Once the atom is ionized, it is trapped by the tokamak magnetic field and it is slowed down by multiple interactions with the plasma electrons and ions. The beam energy is dependent on the plasma density and volume; for efficient plasma heating, the additional power has to be absorbed near the plasma core. In addition to heating, NBI refuels the plasma by the injection of atoms, it also provides toroidal angular momentum (plasma rotation), which is beneficial as it improves the plasma confinement [9], contributes to non-inductive current drive and suppresses some plasma instabilities (tearing modes, etc.). Non-inductive current drive refers to the production of a toroidal current in addition to the inductive current, in order to enable the future fusion reactor to operate continuously (without any inductive current). NBI results in generation of a positive charged ion beam which circulates around the torus and transfers a toroidal momentum via collisions resulting in a net toroidal current. For beam energy much above the critical energy (1 to 2 MeV) the momentum transfer occurs mainly on the plasma electrons that significantly increases the current drive efficiency [8]. Up to now, NBI systems have been successfully used on large operating tokamaks worldwide. NBI systems have yielded high temperature and high fusion performances, and advanced plasma scenarios have been achieved. The neutral beam line for ITER is represented in Figure 0.3 and one of the two complete NBI systems in Figure 0.4. Each system has a length of 15 m, a width of 4.7 m and a height of 5.3 m (together with the high voltage bushing the total height reaches 9 m). The principal component of a typical injector (number 1 in Figure 0.3) is an ion source which generates cold hydrogen (or deuterium) plasma (Te ~ 2 eV to 20 eV, ne ~ 1018 − 1019 m-3). The ions extracted from the source are accelerated to a high energy (~ 1 MeV for ITER) by an electrostatic accelerator (number 2 in the figure) through the polarization of accelerating electrodes (called grids) by high DC voltages. In conventional NBI systems, the ion beam passes through a gas cell (called a neutralizer, number 3 in the figure) where inelastic collisions

Introduction and motivation

with the injected gas (D2) occur, leading to a partial neutralisation of the beam (~55% neutralization for ITER). The remaining ions (D+ and D−) are magnetically or electrostatically deflected onto a water-cooled copper surface (Residual Ion Dump, number 4 in the figure), leaving the neutral beam to continue towards the torus through the NB duct (number 5) [8]. As the size of the future fusion reactor ITER is bigger than nowadays tokamaks, the beam energy of the NBI needs to be higher in order to deposit the beam power in the core of the plasma: 250 to 500 keV/amu (0.5 MeV to 1 MeV D0). With a deuterium positive ion (PI) source, the neutralization efficiency drops dramatically above 100 keV, and no neutralization can occur above 500 keV, as can be seen from Figure 0.5. In the NBI installed on most of the modern fusion machines like JET, the neutral beams have the energy of 100 keV.

2a) ITER NBI For ITER requiring 1 MeV of D0, the use of the NBI system based on the acceleration and neutralization of PI is impossible. On the contrary, it is easy to detach a supplementary electron from H−/D− in a gas cell (neutralizer) on the way of the ion beam, with a neutralization yield of ~55%, even at high beam energies (see Figure 0.5) [10]. The use of NI represents thereby a wide interest for the production of the neutral beams. Yet, the weak binding energy of this electron (0.75 eV for H−) which enables an easy detachment creates a huge difficulty for NI production and slows down the exploitation of such systems. The success of the future generation of fusion machines relies on effective heating provided by the high energy NBI system requiring intense NI currents. It has been calculated that for efficient heating of fusion plasma, the NBI system of ITER will need a stable current of 1 MeV H−/D− with 69 A in hydrogen and 57 A in deuterium [11–13]. Therefore, the development of an efficient NI source becomes particularly important. NI in low-pressure plasmas are usually created by dissociative attachment of cold plasma electrons (< 1 eV) on vibrationally excited molecules [14, 15]. This process is called NI volume production, and it is employed in many applications ranging from microelectronics [16–18] to magnetically confined fusion [19, 20] including space propulsion [21] and sources for particle accelerators [22–25]. However, NI can also be produced by surface ionization on materials in contact with the plasma, with higher energies as compared to volume-produced NI. In most experimental and modeling studies of low-pressure plasmas, NI surface production is usually a mechanism of minor importance. However, in certain circumstances it can be very efficient. When alkali metal surfaces such as cesium (Cs) are in contact with plasma, a huge surface production of NI by conversion of PI or atoms is observed [26]. This effect is the basis of the most intense H−/D− negative ion sources developed for fusion or particle accelerator applications [23, 27, 28]. 5

Introduction and motivation

Figure 0.3. Schematics of the beam line for ITER with the description of main components and the reactions taking place.

Figure 0.4. Neutral beam injection (NBI) system for ITER with the description of main components and physical dimensions. 6

Introduction and motivation

100

90 80 - 70 D 60

40 Fraction (%) Fraction 30 + 20 D 10 0

0 200 400 600 800 1000

+ - Energy of D or D (keV)

Figure 0.5. Maximum neutralization efficiency (fraction in %) of deuterium ions as a function of energy [8].

The source designed for ITER (see Figure 0.6) is composed of eight Inductively Coupled Plasma (ICP) generators, called Drivers, each supplied with about 100 kW of RF power. The source has to generate a uniform flux of NI (with ±5% uniformity) to the array of 1280 apertures on the accelerator front face, the plasma grid. Being created in the drivers with Te ~ 15 eV, the plasma propagates in the diffusion region. A transverse magnetic field ~ 50 Gauss is applied between the diffusion region and the extraction region. The filter field acts as a magnetic barrier which cools down the hot electrons generated within the drivers and prevents a high destruction rate of the NI formed on the plasma grid. In this region, the electron density is reduced and Te ~ 1 eV. Cesium is being evaporated inside the extraction region and deposited on the surface, in particular on the plasma grid [29]. This lowers the material work function down to ~ 2.3 eV and allows some of the neutrals leaving the surface to capture an electron at the Fermi level via the tunnel effect. Currently, this approach is considered as the only way to meet ITER requirements within short timescales. A half ITER-size RF driven ion source (ELISE) is under study in Germany (IPP Garching) [30]. Dedicated laboratory experiments on on plasma transport across the magnetic field are conducted in LAPLACE laboratory (Toulouse, France) [31]. As a next step, it was decided to build the NBI test facility, PRIMA (Padova Research on ITER Megavolt Accelerator), in Italy, including a full- size negative ion source, SPIDER, and a prototype of the whole ITER injector, MITICA, aiming to develop the heating injectors to be installed in ITER [32, 33]. In parallel, the modeling work on Cs-seeded ion sources has been carried out in several Europe research centers in Europe such as IPP Garching, LAPLACE laboratory (Toulouse, France) [34–37], Sofia University and Technical University-Sofia (Bulgaria) [38], many of them in collaboration with CEA Cadarache. 7

Introduction and motivation

2b) DEMO NBI Since 2006, a new research program targeting the emergence of a new generation of NBI systems for the future fusion reactor DEMO (DEMOnstration Power Plant) has been underway between several laboratories in Europe. The specifications required to operate a NBI system on DEMO are very demanding: the system has to provide plasma heating, current drive and plasma control at a very high level of power (up to 150 MW) and energy (1 or 2 MeV), including high performances in term of wall-plug efficiency (η > 60%), high availability and reliability. 2c) Siphore project In view of this objective, a novel NBI concept called Siphore has been proposed by Alain Simonin from CEA Cadarache [39]. It is based on the photodetachment of the energetic negative ion beam. The keystone of this new concept is the development of a photoneutralizer where a high power photon flux (~3 MW) generated within a Fabry Perot cavity will overlap, cross and partially photodetach the intense NI beam accelerated to high energies. It is shown that such a photoneutralization based NBI system would have the capability to provide several tens of MW of D0 per beam line with a wall-plug efficiency higher than 60%. A feasibility study of the concept has been launched between different laboratories (including PIIM) to address the different physics aspects, i.e., negative ion production, plasma modelling, ion accelerator simulation, photoneutralization and high voltage holding under vacuum [39].

Figure 0.6. The ITER RF driven ion source: horizontal cross-section of an individual section (composed of two ICP drivers) on the left; the front side of the complete NI source on the right. Dimensions and description of the main components are given. 8

Introduction and motivation

The Cybele ion source [39–41] is a tall and narrow ion source with a rectangular aspect ratio that is planned to be used in Siphore. It is presently a filamented plasma source in which 5 sets of 3 tungsten filaments are used as cathodes along the source vertical axis. The filaments supply the plasma core with primary electrons along the vertical axis. A uniform magnetic field parallel to the source vertical axis is generated by two lateral coils located on opposite sides of an iron rectangular frame which surrounds the source. The two coils generate magnetic fields in the opposite direction inside the iron structure. It is the leakage field between the two coils that fills uniformly the plasma source volume. The magnetic field intensity within the whole plasma volume can be tuned between 0 mT and 7 mT by adjusting the DC electric current in the coils. Cybele will meet the ITER ion source requirements in terms of NI current density (250 A/m²), source pressure (<0.3 Pa), uniformity, etc. The ability to obtain high plasma density with high ionisation rate and much higher power efficiency than ICP generators makes Helicon source an interesting candidate to feed Cybele with a dense and hot plasma along its central axis. For this purpose, a 10 kW, 13.56 MHz helicon plasma generator is under development at the CRPP-EPFL Lausanne, with a helicon source based on the concept of a resonant birdcage network antenna [42−44]. Although a single 10kW helicon generator will probably not achieve the relevant plasma density required for DEMO NBI source, the 10 kW helicon source is an intermediate step towards larger powers. It is planned to test in Cybele alternative solutions to produce NI. Indeed, severe drawbacks to the use of Cs in the present NI sources have been identified. First of all, to obtain stable negative-ion currents over long shots, a continuous injection of cesium is required, leading to high cesium consumption (~3 µg/s [27, 45]). Secondly, cesium diffusion and pollution of the accelerator stage might cause parasitic beams and/or voltage breakdowns and imply a regular and restrictive maintenance in a nuclear environment. The Eurofusion roadmap has included in the list of priorities the reduction of Cs consumption or its complete elimination in the NI sources of the future generation. The reduction of Cs consumption could be, for instance, achieved by implanting molybdenum with Cs where a recent progress has been shown by Schiesko et al [12]. Another possibility is the use of alternative materials for NI conversion with efficiency comparable to Cs seeded sources. Thus, development of high-intensity Cs-free NI sources would be highly valuable for many applications within thermonuclear fusion. Diamond is one of materials which are planned to be tested in Cybele as NI enhancers. It is well known for its ability to emit electrons at high temperature and even at low electric fields [46]. Beam experiments on diamond showed surface production of H− ions with high yields up to 5.5%. Moreover, it has been observed in plasma experiment that NI production yield on boron-doped-diamond can be increased by a factor 5 when increasing the temperature to 400°C [47]. It has been shown by various authors that hydrogen-terminated diamond shows negative electron affinity (its conduction band lies above the vacuum level) [48, 49]. Therefore, electrons can be efficiently supplied to the 9

Introduction and motivation

vacuum, since the valence band is located at higher position. All these evidences make diamond an interesting candidate for NI surface production and one of the main candidates to be used in Cybele.

This thesis presents a study of NI surface production in Cs-free H2/D2 plasmas on different materials (in particular diamond), within the collaboration for R&D around the DEMO-relevant NBI system. The research work is summarized in five chapters. Chapter 1. Negative ion surface production measurements explains the general principle of the negative ion surface production measurements and gives the details of the employed experimental set-up PHISIS. The experimental protocol used for several types of measurements is also specified therein. Chapter 2. Modeling and reconstruction of NIEDF is dedicated to the modeling and reconstruction of surface-produced NI energy and angular distribution functions. A study performed on a large variety of materials such as different types of graphite, diamond films and metals is presented in Chapter 3. NI production on different materials. The surface state characterization of NI enhancers by external diagnostics such as temperature programmed desorption (TPD) and Raman spectroscopy is described in Chapter 4. Surface state characterization of NI enhancers. The last chapter entitled 5. Pulsed-bias approach provides information about the newly developed method of pulsed bias which has enabled the study of NI production on surfaces of insulating materials.

This project is conducted within the framework of European Program (Eurofusion) and a French national project (ANR) “H Index Tripled”. The ANR project unites four research centers: PIIM (Marseille), LSPM (Paris), DIFFER (Eindhoven, Netherlands) and IRFM (CEA Cadarache). We have also received support from the region Provence-Alpes- Côte d’Azur (project PACA GING). This doctoral thesis was co-financed by CEA Cadarache and PACA region.

1. Negative ion surface production measurements

1.1 General principle of measurements The main idea of the thesis is to study NI surface production in low pressure hydrogen and deuterium plasmas. The sample is placed inside the plasma chamber and negatively biased with respect to the plasma potential, so the positive ions are accelerated towards the sample surface by the electric field formed in the sheath. The ion bombardment leads to the production of negative ions on the surface. These negative ions are accelerated in backward direction and cross the sheath in front of the sample, the plasma region and the sheath in front of the mass spectrometer (MS) nozzle before arriving to the MS detector. Hence, the negative ion energy distribution function (NIEDF) of the surface-produced NI can be measured (see Figure 1.1).

Figure 1.1. Principle of negative-ion measurements on PHISIS set-up with potential profile between the sample and the MS. Vp – plasma potential, Vs – surface bias, VMS – mass spectrometer potential. Output: negative ion energy distribution function (NIEDF). 11

1. Negative ion surface production measurements

The extraction of NI from plasma could be more or less efficient depending on experimental conditions such as sample surface bias Vs, MS potential VMS, etc. The modeling for several conditions was performed (see Section 1.2.3. Mass spectrometer for details), and the case of Vs = -130 V and VMS = 0 V has proven to provide plane sheath in front of the sample and efficient self-extraction of NI from the plasma. This condition was chosen for all the realized studies, if not specified otherwise. 1.2. Experimental set-up

1.2.1. PHISIS Plasma exposure was performed in a helicon reactor PHISIS (Plasma Helicon to Irradiate Surfaces In Situ) [50] which consists of a an upper cylindrical source chamber (360mm long and 150mm diameter) and a lower spherical diffusion chamber (radius 100 mm). The plasma is initiated by a Huttinger PFG 1600 RF generator (13.56 MHz) followed by a Huttinger matchbox connected to a Boswell antenna (helicon antenna) [51] surrounding a pyrex tube (150mm diameter, 200mm length) installed in the source chamber [52]. Plasma is then allowed to diffuse vertically down into the diffusion chamber, where plasma-surface interactions take place. With the help of a special sample holder, negatively biased samples are placed into the diffusion chamber. The sample surface can be biased up to 1 kV, heated (up to

800°C) and cooled (at LN2 temperature). The quadrupole mass spectrometer with an energy analyzer is placed at a distance of 37 mm in front of the sample to detect negative ions produced at the sample surface. The set-up is equipped with a load-lock system with a base pressure of 10−3 mbar (provided by a primary pump) for a quick change of samples. The base pressure of 10−7 mbar is achieved in the reactor using a 400 l s−1 turbo molecular pump (Alcatel ATP400) installed below the diffusion chamber. The detailed scheme of the experimental set-up is represented in Figure 1.2 and Figure 1.3. There is a provision to use a magnetic field (vertical direction) in the source chamber of the plasma reactor. However, no magnetic field was applied in the plasma reactor so that only capacitive and inductive plasma coupling could be used. All experiments have been performed in the capacitive mode (E-mode). To minimize RF fluctuations of plasma potential arising due to capacitive coupling, a mechanical grounded screen was placed horizontally approximately 5 cm above the sample (in between the source and the diffusion chamber) [47] and low injected power of 20W was used at 2.0 Pa gas pressure. The hydrogen/deuterium gas injection into the reactor was realized by BROOKS Mass Flow Controller 5850TR of 20 sccm (with 5.2 sccm for hydrogen and 7.6 sccm for deuterium). The pressure in the chamber was monitored by a Baratron gauge (MKS). To maintain the desired pressure in the chamber and to reduce hydrogen consumption, a 150mm inner diameter Riber gate valve installed just before the turbo molecular pump was adjusted accordingly [52]. 12

1. Negative ion surface production measurements

The plasma density and temperature for the usual experimental conditions were measured by a Langmuir probe inserted in the diffusion chamber (Scientific System smart probe): ne = 2·1013 m-3 and Te = 3.5 eV. At this pressure the plasma is mainly populated with H3+ ions (~80%) for hydrogen and D3+ ions (~90%) for deuterium because of the fast conversion: D2+ + D2 → D3+ + D

The domination of D3+ ions at 2.0 Pa persists for different biases applied as well as for different distances between the MS and the sample surface: see Figure 1.4 (a). Typically, the separation between the MS and the sample is 37 mm which is approximately equal to the mean free path for electron detachment by collisions of H– with H2 at 2 Pa (this is a dominant mechanism in our conditions [50]).

For most of the measurements the negative bias of the sample was Vs = -130 V resulting in the incoming positive ion energy of 45 eV per nucleon, as the plasma potential is typically ~ 5 V when the surface bias is applied (measured by the MS). The ion flux in these conditions was approximately 3·1012 H3+ or D3+/cm2·s, as given by the Langmuir probe.

Figure 1.2. (a) Experimental set-up PHISIS, (b) details of the sample holder, (c) inside the diffusion chamber showing the sample holder, Langmuir probe and MS. 13

1. Negative ion surface production measurements

Isolating valve

Load-lock

Transfer rod

Figure 1.3. Experimental set-up PHISIS with outgoing transfer rod.

6 10 + + D 2 Pa, D D 2 Pa, D 2 2 a) + (b) D + 2 25 W d = 37 mm D 20 W 2 + V = -130 V D + s 3 D 105 3 104

PI flux, arb.u. PI flux, PI flux, arb.u. PI flux, 103

102 103 -140 -120 -100 -80 -60 -40 -20 0 36 37 38 39 40 41 42 43 44 45 46 Vs, V d, mm

Figure 1.4. Positive ion flux (defined as an integral over PI distribution function) for different deuterium ions as measured by the MS versus (a) applied surface bias, (b) distance between the MS and the sample surface.

1. Negative ion surface production measurements

The samples were held in place on the sample holder by a 2 mm thick molybdenum clamp as shown in Figure 1.2 (b) which was later replaced by a thinner (0.1 mm thick) clamp. The radii of the clamp hole defining the sample area exposed to plasma were 4 mm, 2 mm and 1 mm depending on the size of the sample in use. By default, the radius of 4 mm is assumed if not indicated otherwise. For thin samples, a washer with a hole having a radius of 4 mm was put in front of the sample in order to establish a good thermal contact between the sample and the sample holder.

1.2.2. Materials

The most important samples studied within this thesis were highly oriented pyrolitic graphite (HOPG), microcrystalline boron-doped diamond (MCBDD) and microcrystalline non-doped diamond (MCD). The HOPG material was of ZYB type purchased from MaTeck GmbH Company. The density and electrical resistivity of HOPG were 2.265 g·cm−3 and 3.5 × 10−5 Ω·cm, respectively. Polycrystalline MCBDD and MCD films were deposited at LSPM laboratory by using plasma-enhanced chemical vapor deposition (PECVD). MCBDD was deposited in a bell jar reactor (PLASSYS BJS 150) operating with a mixture of H2, CH4 and B2H6 gases [53]. The boron to carbon ratio in the gas phase (defined as 2×[B2H6]/[CH4]) and the methane concentration were set to 1000 ppm and 4%, respectively. The reactor was operated at 200 mbar pressure and the injected microwave power was set to 3kW. The films were deposited on a <100> oriented 1 mm thick silicon substrate. The deposited diamond layer for MCBDD had a thickness of 3.2 ± ± 0.1 μm (determined by the weight gain of the substrate). For MCD films, the mixture of H2 and CH4 was used during PECVD. The exact layer thickness is unknown, but it was estimated to be several micrometers. The grain size for both samples is around 5 μm in the horizontal plane, as seen by scanning electron microscopy. The charge carrier density for MCBDD was estimated to be 1019 ÷ 1020 cm−3, which leads to good electrical conductivity for biasing of the diamond layer. Before performing any experiments with the samples, a certain preparation procedure was followed. HOPG was cleaved by using a scotch tape which would remove several monolayers from the top surface, thereby providing a “pristine” sample. The diamond films were originally deposited on a circular Si wafer of 5 cm in diameter. By using a diamond pencil and several glass inking slabs (in order to break the wafer precisely along the symmetry direction), the square samples of 1 cm by 1 cm were produced. The samples were installed into the sample holder via dedicated window in the load-lock chamber. They were treated by a dry air spray in order to remove the dust from the surface (and macroscopic delaminations in case of HOPG) before being introduced into the diffusion chamber.

1. Negative ion surface production measurements

1.2.3. Mass spectrometer The main diagnostic tool employed to study NI surface production on PHISIS set-up is the mass spectrometer Hiden EQP 300 (Electrostatic Quadrupole Plasma). It combines the electrostatic analysis sector for energy with a quadrupole mass filter of high performance. The MS Hiden EQP 300 is designed for energy analysis between 0 and 1100 eV for a mass up to 300 a.m.u. of ions and neutrals generated in the plasma. A maximum pressure of 10-6 mbar inside the MS is necessary to perform the measurements in order to limit the collisions between the analyzed species and the gas present in the device. The base pressure inside the MS which is differentially pumped is typically 5·10−8 mbar, the measurement being performed by a Bayer-Alper type gauge. The opening orifice of the MS relaying it to the diffusion chamber is 100 μm which allows maintaining the pressure below 10-6 mbar even in the presence of 2 Pa of H2 in the assembly. The MS is controlled by a PC via MASsoft software produced by Hiden Analytical. It allows to acquire, store and manipulate the mass spectra and the energy distributions of neutrals, as well as positive and negative ions. 3D schematics of the Hiden EQP MS is represented in Figure 1.5 with a brief description of the main components.

Figure 1.5. 3D schematics of the HIDEN EQP - the mass/energy analyser for plasma diagnostics and characterisation. 16

1. Negative ion surface production measurements

Figure 1.6. Scheme representing the HIDEN EQP MS with its main components and electrodes. For complex components like mass and energy filters the individual electrodes are specified in the brackets.

40 mm Grounded orifice

5 mm

Extractor hole 100 μm Extractor is biased

Figure 1.7. Schematic representation of the used MS nozzle configuration.

1. Negative ion surface production measurements

The MS could function in two modes: Residual Gas Analysis (RGA) and Secondary Ion Mass Spectrometry (SIMS). The RGA mode is used to analyze the mass and energy of the neutral species (molecules, radicals, atoms) in order to determine the composition of the gas or the plasma. The neutrals are extracted from the plasma and then ionized inside the internal ionizer by electron impact. The energy of ionizing electrons could be controlled in order to allow the detection of radicals. The ions emitted from the ionizer are right away directed into the center of the energy filter. In the SIMS mode the ions are extracted from the plasma via the opening orifice of 100 μm perforated in the polarized extractor. The electrostatic lens called “Lens 1” is located just behind the extractor and serves to focus the ion beam into the energy filter in order to maximize the signal. The MS is composed of four main parts: extractor, energy analyzer, mass filter and the detector (see Figure 1.6 for details). The extraction of ions is realized by a polarizable orifice of 100 μm surmounted by a 5 mm hole. The extractor potential Vext could be defined directly by MASsoft and is adjustable between –100 V and +360 V, it allows to choose depending on its sign the positive or negative ions from the plasma. Extractor and all the lenses are connected to a reference potential Vref which could be varied between –1000 V and +1000 V. The voltage of the MS nozzle sets itself to the potential equal to the sum of the voltages applied to the extractor and to the reference: VMS = Vext +

+ Vref. The nozzle potential has a big influence on the trajectories of the measured NI and on the shape of the sheath in front of the MS. In order to ensure that the sheath in front of the MS is planar which simplifies the modeling presented in Chapter 2. Modeling and reconstruction of NIEDF, the collaboration with LPGP laboratory in Paris Sud University was established. T. Minea and S. Mochalskyy have used the code named ONIX (Orsay Negative Ion eXtraction) [34– 37] to study the transport electrons as well as positive ions in the vicinity of the MS nozzle (see the scheme represented in Figure 1.7). The code has included the geometry of the MS with the grounded orifice in front and the extractor hole of 100 μm, and the extraction potential VMS was used as the input parameter, as well as ne and Te measured by the Langmuir probe. The simulations have demonstrated that for VMS = 0 V at 2 Pa of gas pressure the shape of the sheath is nearly planar. Therefore, this setting has been chosen for all the experiments performed during this thesis. Before performing the measurements for the first time for a given ion mass and/or ion energy, the MS should be tuned to maximize the measured signal. It is done with a help of an autotune file which is provided by MASsoft and could be customized or generated manually. In the SIMS mode, the recommended tuning order is as follows: energy, extractor, lens1, lens2, focus2. Typically, when the surface bias Vs of the sample is changed significantly (by more than 20 V) the autotune procedure should be redone to improve the collection of NI.

1. Negative ion surface production measurements

1.3. Experimental protocol

1.3.1. Representation of spectra

Negative ions formed on the negatively biased sample surface upon PI bombardment are accelerated by the sheath toward the plasma. Under the low-pressure conditions considered here, most of the NI cross the plasma without any collision [47] and reach the mass spectrometer where they are detected according to their energies. A NI created at the surface has the initial total energy:

ETi = Ekini + Epoti = Ekini – eVs Eq. (1.1) where Ekini is its initial kinetic energy and Epoti is initial potential energy at the surface. This ion arrives to the MS detector with the final total energy:

ETf = Ekinf + Epotf = Ekinf – e(Venergy +Vaxis +Vref) Eq. (1.2) where eVenergy is the ion energy registered by the software, Vref is reference potential of the MS and Vaxis is axis potential to which the ion is accelerated/decelerated to be able to pass through the energy filter. Therefore, for our MS settings the collected ion must satisfy the condition: – eVaxis = Ekinf = 40 eV Eq. (1.3)

According to the energy conservation principle ETi = ETf, so we obtain:

Ekini – eVs = Ekinf – e(Venergy +Vaxis +Vref)

Ekini – eVs = – e(Venergy +Vref)

Ekini = e(Vs –Venergy –Vref) Eq. (1.4) The expression above allows calculating the initial kinetic energy of the NI for known surface bias, reference potential and the NI energy registered by MASsoft. This procedure was performed for all measured NIEDF presented in this thesis in order to represent the NI signal as function of the energy at which the ions were created at the surface. So, the NI with a more negative value of eVenergy in the measured distribution would appear with a bigger kinetic energy in the treated NIEDF. If a NIEDF shown on some graph in this thesis starts at negative energy, according to Eq. (1.4) it can be connected either to the decrease of the surface bias (due to insulating properties of the sample) or to the origin of NI other than the sample surface (which imposes that they don’t gain Epoti = Ekini – eVs when crossing the sheath in front of the sample). The reference potential was kept Vref = –Vs – 20V for all the measurements within this thesis for the sake of convenience, so that all registered distributions would be starting at –20 V independently of the applied surface bias.

1. Negative ion surface production measurements

1.3.2. RF plasma

The base pressure in PHISIS set-up was typically maintained below 5·10−7 mbar. Before introducing the sample inside the reactor, the load-lock chamber was pumped down to 10-3 mbar, and then the valve separating the load-lock system and the diffusion chamber was opened. Once a pressure of 10-3 mbar has been reached, plasma was ignited at least for 10-15 minutes to condition the reactor with a sample still in load-lock (the plasma was not diffusing inside the narrow arm). Consequently, the sample was introduced inside the reactor with a help of the transfer rod and the measurements were performed. If the sample was heated, the interior of the rod had to be cooled by compressed air during the whole heating time. As a part of experimental protocol estimation, it was necessary to define the time needed for the surface exposed to plasma to reach a stationary state before performing the measurements. As could be observed in Figure 1.8, the NIEDF measured on HOPG evolve as a function of time. In order to trace the time evolution of the signal, no averaging procedure was applied to the NIEDF, so the signal to noise ratio is low. Figure 1.9 shows the NI yield for HOPG and MCBDD as a function of time (yield is defined as integral over the whole NIEDF). It could be seen that for this experimental condition the interaction of HOPG surface with plasma enhances the NI production most probably via creation of defects and hydrogenation, as will be discussed in more detail in the next chapters. For MCBDD surface the effect is the opposite. However, it could be noted that for both samples a stationary state is achieved after approximately 10 minutes of plasma exposure under the given bias Vs = -130 V. Before these experiments, samples have been pre-heated under vacuum for 5 minutes at 500°C which was enough to degas the impurities (as shown in Figure 1.10). This procedure was performed only for NIEDF time evolution measurements as one had to be sure that the signal change during the first minutes is not connected to the removal of impurities by plasma. For the usual measurements, the samples were not pre-heated under vacuum, but just exposed to plasma under the chosen bias for at least 10 minutes until the stationary state was reached (with similar number of counts as for the pre-heated samples). The number of counts on HOPG has varied from one experiment to another, as it was influenced by the sample cleavage quality, plane orientation, etc. As for the diamond films, homogeneity of the film deposition, grain size and orientation had a certain influence on the NI signal level for different samples. The main parameters which define the energy scan were set using MASsoft. The scan range can be varied from –100 V to +100 V, the energy interval used was typically 0.2 V. The averaging was usually turned on in order to increase the signal-to-noise ratio of the measured spectra, unless performing the time evolution measurements. The delta-m parameter (which gives control over mass resolving power for low masses) was set to discriminator which defines the signal threshold of the channeltron detector (electron

1. Negative ion surface production measurements

0 min HOPG D 2 1 min 2 Pa 20 W 2 min V = -130 V 4 s 3 min 10 14 min

103 Intensity, cts/s Intensity,

102 0 5 10 15 20 25 30 35 40 45 Energy, eV

Figure 1.8. NIEDF for HOPG as a function of time from the beginning of plasma exposure at 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

4.0x105 HOPG MCBDD

3.5x105 Samples pre-heated under vacuum 5 3.0x10 to 500°C for 5 min

2.5x105

2.0x105

5 NI Yield, arb.u. Yield, NI 1.5x10

5 1.0x10

4 5.0x10 0 2 4 6 8 10 12 14 Time, min

Figure 1.9. NI yield for HOPG and MCBDD as a function of time from the beginning of plasma exposure at 2.0 Pa D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

1. Negative ion surface production measurements

7.0x10-6 400°C

6.0x10-6 Degasing the sample pressure in the reactor 300°C 5.0x10-6 200°C -6

4.0x10

3.0x10-6 p, mbar p,

2.0x10-6

1.0x10-6 100°C 400°C 0.0 0 2 4 6 8 10 12 14 Time, min

Figure 1.10. Pressure in the diffusion chamber as a function of time during the degassing of MCBDD sample under vacuum. The instantaneous temperatures of the sample are also indicated in the graph.

Figure 1.11. Photo of the ECR plasma source inserted in the diffusion chamber of PHISIS creating the H2 plasma. The sample holder can be also seen. The grounded screen is placed on the bottom as it is unnecessary.

1. Negative ion surface production measurements

multiplier) was set at –10% as given by default. The dwell time which defines the time used to acquire a single point in the scan was set to 100 % of the recommended value [54]. The settle time which defines the time to allow the electronics to settle before the scan is started was typically 1000% of the recommended value. 1.3.3. ECR plasma In order to avoid the RF fluctuations which broaden the recorded NIEDF in hydrogen plasma, some of the experiments have been performed in microwave plasma created by an ECR source (Electron Cyclotron Resonance) from “Boreal Plasmas”. The frequency of the RF generator is 13.56 MHz, which is close to the ion plasma frequency in our conditions. In case of the ECR source, the electric field is sent at frequency fE = q·B/m, where q and m are respectively elementary charge and electron mass, and B is the magnetic field. The frequency is 2.45 GHz for the magnetic field of B = 845 Gauss, which is much larger than the ion plasma frequency. The consequence is visible on the shape of the NIEDF: no broadening is observed. Therefore, instead of the RF source used in capacitive mode to generate the plasma in the source chamber, the ECR source was inserted into the diffusion chamber, ~50 mm from the sample (see Figure 1.11). The magnetic field in the proximity of the sample was B ≈ 50 Gauss as given by a Gaussmeter, and the MS measurements demonstrated no influence of B on the shape of NIEDF. With the plasma created by the ECR source, the electronic density was increased by approximately a factor of hundred. At usual conditions for capacitively coupled RF plasma ne = 2·1013 m-3 and Te = 3.5 eV whereas with ECR source ne = 2.5·1015 m-3 and

Te = 0.95 eV. The values of ne and Te were measured by the Langmuir probe close to the sample with the ECR source located as mentioned above. The plasma potential was measured by both Langmuir probe and the MS when applying the bias of Vs = –130 V, and the results were 5 V and 7 V in case of the RF and the microwave plasma correspondingly. The generator used to ignite the microwave plasma was MPG-4 from

Opthos Instruments Inc., using the frequency of 2.45 GHz. The working pressure in H2 plasma was 1 Pa with 60 W of effectively absorbed microwave power (the reflected power was typically 5–10%), and the applied surface bias was Vs = –130 V. The minimum reflected power was achieved by tuning the position of two impedance matchers located inside the waveguide of the ECR source.

Variation of ne and Te with time (connected to the heating up of the ECR source and the whole reactor assembly) could cause the change of the NIEDF shape which is undesirable. As low signal-to-noise ratio measurements of NIEDF for 35 values of tilt angle α are quite long (4–5 hours), it was necessary to ensure that that NIEDF shape does not change as a function of time. For this purpose, we measured at 1Pa and 100W injected power one NIEDF per 2 minutes, no averaging applied. As can be seen from Figure 1.12, the intensity of the NI signal decreases with time in overall.

1. Negative ion surface production measurements

1 Pa, ECR 100 W 5 10 H plasma 2 2 min 7 min 11 min

104 22 min 33 min

55 min

Intensity, cts/s Intensity, 10

102 0 5 10 15 20 25 30 35 Energy, eV

Figure 1.12. Experimentally measured NIEDF (1Pa, 100W ECR H2 plasma) on HOPG sample as a function of time. The sample is located at normal incidence: α = 0°.

1 Pa, wave 100 W 3.5x105 H2 plasma

HOPG usual

HOPG pre-heated

3.0x105

5 2.5x10

NI yield, arb.u.

2.0x105

0 20 40 60 80 100 120 140 160 180

Time, min

Figure 1.13. NI yield collected experimentally by the MS from HOPG surface as a function of plasma exposure time in ECR hydrogen plasma (1Pa, 100W): virgin sample (black curve, after 53 minutes plasma was stopped for 2h and restarted again); virgin pre- heated/degassed sample (red curve). 24

1. Negative ion surface production measurements

To follow the evolution of the signal with time, a freshly-cleaved HOPG sample was used, and the NIEDF were integrated to give NI yield. The signal evolution was followed for about 1 hour (see left part of the black curve in Figure 1.13). One could notice a signal increase during the first 5–10 minutes and then a decrease. This increase is probably connected to the establishment of equilibrium between the exposed sample surface and the plasma. Then, after 53 minutes of the sample exposure, the plasma was stopped for 2h and restarted again without cleaving the sample (see continuation of the black curve in Figure 1.13). No effect comparable to the previous slow signal increase has been observed, but the NI yield has grown. This could be explained by thermal effect (cooling down of the ECR source and of the whole reactor assembly). The red curve in Figure 1.13 shows a similar experiment, but with the HOPG sample heated to 500°C under vacuum for 15 minutes to remove the impurities before the plasma exposure. Since the curve form is the same, there is no effect of the impurities on NI yield change which confirms our previous conclusions. To study the influence of the incident PI flux on the NI signal decrease with time, a separate experiment was performed: three PIEDF (for H3+, H2+ and H+) were measured versus time, with no averaging applied. As seen in Figure 1.14 at first we have a fast decrease (most probably due to ECR source magnet heating) and then a slow decrease of the PI flux with time. Some fluctuations are seen, but overall the signal decreases slowly.

At these conditions (1Pa, 100W and Vs=-145 V) H3+ was dominating and H2+ was the smallest ion population.

5 7x10 + H 5 + 6x10 H2 + H3 5x105

4x105

5 3x10 yield,arb.u. PI 2x105

1x105

0 10 20 30 40 50 60 70 80 90 100 110 120 Time, min Figure 1.14. PI yield for 3 ion populations collected by the MS from HOPG surface as a function of plasma exposure time in ECR hydrogen plasma (1Pa, 100W). 25

1. Negative ion surface production measurements

The PI yield has decreased in first 20 minutes by 20% (fast decrease) and in total after 2 hours by 30% (slow decrease). The H+ signal had bigger fluctuations and a less pronounced slow decrease. Most probably, the PI flux decrease can be explained by the heating up of the ECR source and of the whole reactor assembly which changes the plasma equilibrium and/or the microwave coupling. The temperature of the sample was rising up to 70–80°C during the long term operation in ECR plasma which should have no important influence on the NI signal level (see Chapter 3. NI production on different materials).

The conclusion drawn from the time evolution experiments in ECR microwave plasma are the following:

 first 10 minutes are needed for HOPG sample to establish an equilibrated surface state under the plasma exposure. The surface temperature during the long term operation in ECR plasma is typically 70–80°C which has no important influence on the NI signal level

 during the first 20-30 minutes of operation in microwave plasma for high input power values (~100 W) the heating of the ECR source and of the whole reactor assembly changes the plasma equilibrium and/or the microwave coupling. Therefore, the incoming PI flux decreases which explains, at least partially, the decrease of the NI yield

 smaller input powers (~60W) should be used for further measurements; the measurements should be started after at least 30 min of plasma exposure; the slow signal decrease should be monitored by periodically checking the signal at normal sample orientation (α=0°)

1. Negative ion surface production measurements

1.4. Conclusion

The general principle of the negative ion surface production measurements and the details of the experimental set-up employed were given in this chapter. The experimental protocol used for most of the measurements was specified. The chosen experimental conditions in RF plasmas were 2.0 Pa H2/D2 pressure with the injected RF power of 20W and a grounded screen placed above the sample to minimize the RF fluctuations. For ECR plasmas the conditions were the following: 1 Pa H2/D2 pressure with 60 W of effectively absorbed microwave power (the reflected power was typically

5–10%). The flow rates were 5.2 sccm for H2 and 7.6 sccm for D2.

The negative bias of the sample surface was chosen at Vs = –130V for most of the measurements ensuring the efficient self-extraction of negative ions from the plasma.

The dominant positive ion population in this condition was H3+/D3+ giving the incoming ion energy of 45 eV per nucleon at the surface. The surface stationary state under the given bias is achieved after approximately 10 minutes of plasma exposure for diamond and graphite samples (estimated thanks to NIEDF time evolution measurements). The radius of the circle exposing the sample to plasma was typically 4 mm if not specified otherwise. The main samples studied within this work and their properties were presented. These were highly oriented pyrolitic graphite (HOPG), microcrystalline boron-doped diamond (MCBDD) and microcrystalline non-doped diamond (MCD).

2. Modeling and reconstruction of NIEDF

The study of NI surface production on HOPG has been performed by Loïc Schiesko during his thesis [55–57]. He has proven that NI are created on surfaces of carbon materials in our experimental conditions via two mechanisms: a) backscattering of incoming positive ion as a NI; b) sputtering of an adsorbed H/D atom as a NI by incoming positive ion. The low-energy peak of NIEDF results from interplay of the two mechanisms whereas the energetic tail contains only the backscattering NI contribution. The maximum energy of NIEDF is defined by the potential acquired by the dominant positive ion in the sheath between the sample and the plasma. In the case of molecular ion (for example H3+), this energy has to be divided by the number of fragments (E0/3), because all the ions undergo dissociation and neutralization at impact. These conclusions have been proven experimentally. Ahmad Ahmad during his thesis has developed a model which could explain the shape of experimentally measured NIEDF for the sample placed perpendicularly to the MS axis [52, 57]. The initial guess of NI distribution on the sample surface (in energy and angle of emission) was taken from SRIM; the ion trajectories were calculated inside the sheaths and in the plasma, and also inside the MS. The details of this model will be given in the first section of the chapter. After undergoing these modifications, NI distribution was compared to the experimental data and yielded an excellent agreement. In the present thesis, the goal was to use experimental NIEDF (measured at different tilt angles of the sample with respect to the MS axis) in order to reconstruct the initial distribution of NI emitted from the surface.

2. Modeling and reconstruction of NIEDF

2.1. Modeling of NIEDF for sample normal to the MS axis

2.1.1. Description of the model

Analysis of NIEDF contains information about the mechanisms of NI creation on surfaces of different materials. A NI is created on the surface of the sample with a certain energy E and emission angle θ with respect to the normal to the MS axis. The distribution of emitted NI is described by the function f(E,θ). After their creation at the sample surface, NI are being repelled by the sample potential and self-extracted from the plasma towards the MS. Their distribution function is modified by the acceleration and deflection in the sheaths and in the plasma and becomes f ’(E,θ). After that, NI cross the MS and reach the channeltron detector. As a result, the distribution function is transformed into f ”(E,θ) = = f ’(E,θ)·TMS(E,θ), where TMS(E,θ) is the MS transmission. The initial distribution in energy and in emission angle f(E,θ) can be obtained by SRIM code (Stopping and Range of Ions in Matter) [59]. SRIM models the distribution of backscattered and sputtered neutral particles leaving the surface. Under the assumption that NI ionization probability is independent of the energy and angle of emission

(Piz = const ∀ E and θ), the f(E,θ) given by SRIM is also valid for NI. This distribution serves as an input for the model developed by Ahmad Ahmad during his thesis [52]. Within the model, ion trajectories are calculated inside the sheaths and in the plasma in order to obtain f ’(E,θ). The transmission through the MS was modeled using SIMION software by Thimotée Pasquet [52] and thereby f ’(E,θ) can be transformed into f ”(E,θ) which represents the distribution function at the MS detector and can be compared to the experimental data. However, the MS transmission did not show a significant influence on the NIEDF form. For this reason, f ”(E,θ) will not be considered within the present thesis, and experimental NIEDF will be compared to f ’(E,θ). In the experiment, the signal registered by the MS is an integration of f ’(E,θ) over all NI emission angles: fexp(E) = ∫ f ’(E,θ) dθ. Therefore, a comparison between experimentally measured NIEDF and modeled f ’(E,θ) reveals the lost information about the NI emission origin. For carbon materials the modeled distributions compare quite well to the experimental data in most of the cases (when Piz = const). In such a way, the initial choice of f (E,θ) given by SRIM can be validated by a good agreement between modeling and experiment.

2. Modeling and reconstruction of NIEDF

2.1.2. SRIM software

SRIM software (Stopping and Range of Ions in Matter) [60, 61] is a group of programs designed to calculate the stopping of ions (10 eV – 2 GeV/a.m.u.) in matter. It is intended to calculate the interactions of energetic ions with amorphous targets. SRIM is based on Monte Carlo approach [62, 63] and treats ion-atom collisions according to the laws of quantum mechanics. SRIM accepts complex surfaces; it is possible to construct a surface out of eight different components each one composed of 12 different sub-components. With SRIM one could calculate the final distribution of ions in 3D with all the kinetic phenomena associated with ion energy loss: target damage, sputtering and ionization. The target material is considered to be amorphous with atoms occupying random positions and a mean distance which corresponds to interatomic distance of given material. Calculations performed during this thesis have contained 106 or 107 hydrogen/deuterium ions with incident energy Eimpact = 45 eV impinging on 10 nm of target material. The choice of energy was derived from the experimental conditions. For the applied bias Vs = -130 V and plasma potential Vp = 5 V the ion energy before contact with the surface is E0 = e(Vp- Vs) = 135 eV. Since the dominating ion population is H3+or

D3+, after the dissociation of the molecular ion, energy is equally divided between three fragments: Eimpact = E0/3= 45 eV. The target material density was taken as ρ = 2.2 g/cm3 for HOPG and as ρ = 3.5 g/cm3 for diamond films. The composition of the surface corresponded to the hydrogen saturation level in hydrogenated amorphous carbon films (a:C-H) and contained 30% of H/D atoms and 70% of carbon atoms [64 – 69]. The chosen type of calculation was “Surface Sputtering/Monolayer Collision steps”, since it is more adapted to treat backscattered and sputtered particles. The other input parameters were:

 Displacement Energy - the energy necessary for a recoil to overcome the lattice forces and to move more than one atomic spacing away from its original site. It was fixed at 25 eV for carbon and 2.5 eV for H/D [70 – 75].  Surface Binding Energy - the energy that target atoms have to overcome in order to leave the surface of the target. It is approximated by the sublimation energy for mono-element material and is unknown for composite materials. Surface binding energy is a key parameter to estimate the sputtering yield (number of target atoms leaving the target surface). The value of 4.5 eV was taken for carbon and 3 eV for H/D [70, 76, 77].  Lattice Binding Energy - the energy lost by every recoiling target atom when it leaves its lattice site and recoils in the target. For both carbon and H/D the value of 3 eV was chosen [70, 77].

2. Modeling and reconstruction of NIEDF

The incidence angle of the incoming ions was fixed at 0° for all the calculations in which the roughness of the layer is not taken into account, since the ions are considered to impact the surface perpendicularly. SRIM provides calculation results in the form of text file which contains all the emitted particles with the origin of their emission (backscattering or sputtering), atomic number, initial energy Ei, depth (x), lateral position (y, z) and the direction of their movement: cos (x), cos (y), cos (z). From these parameters, the initial emission angle θ  i and the initial velocity  i can be calculated for each ion with the help of the routine written in Scilab [78]. Thereby, the distribution function of emitted ions in energy and angle f(E,θ) can be deduced from the file given by SRIM. The distribution f(E,θ) is the first output of the model. 2.1.3. Calculation of NI trajectories Each ion emitted from the surface passes through three different regions before arriving to the MS (see Figure 2.1):

 Sheath in front of the sample (ionic sheath)  Plasma  Sheath in front of the MS (ionic sheath)

Since the sample surface is negatively biased, the first sheath is always ionic which attracts positive ions towards the surface and repels the electrons. The electric field established between the plasma and the sample also provides for the self-extraction of the NI created on the surface. With the information given by SRIM, it is possible to calculate the trajectories of NI in  the sheaths and in the plasma, if the electric field E ( x , y , z ) is known. For a planar sheath, the electric filed is perpendicular to the sample surface, so the potential can be simply described by the Child-Langmuir law in one dimension. In this work, the experimental arrangement and the experimental conditions lead to planar sheath in front of the sample and in front of the MS (see Section 1.2.3. Mass spectrometer). Therefore, potential distribution inside the sheaths φ(x) is known from the Child-Langmuir law and the measured plasma parameters: Vp, Te, ne. The weak electric field in the plasma is neglected.

For NI to be detected by the MS, it should fulfill two requirements:

 its deviation should be inferior to dmax = 4 mm (sample radius) assuming that the sample is centered with respect to the MS axis

 its arrival angle θMS should be smaller than the MS acceptance angle: θMS < θaa. The acceptance angle is known from the simulation of the MS performed with SIMION software [52]

2. Modeling and reconstruction of NIEDF

For each ion emitted from the surface the model calculates its arrival angle θMS, arrival energy EMS and deviation d (see Figure 2.1) from the known Ei, υix, υiy, υiz and φ(x) (potential distribution in the sheath given by the Child-Langmuir law). The arrival angle θMS is calculated using the energy conservation law and the deviation d is calculated with the Newton's second law [52]. All the ions with θMS > θaa and/or d > dmax are eliminated. The remaining ions are described by the distribution f ’(E,θ) which after integration over all emission angles becomes f '(E): NIEDF at the entrance of the MS. This is the second output of the model. Note that these calculations assume that the sample surface is perpendicular to the MS axis. The ions which have entered the MS orifice are described by f ’(E,θ). They pass through the MS components (electrostatic lenses, energy and mass filters, etc.) and arrive at the MS detector with distribution f ”(E,θ) = f ’(E,θ)·TMS(E,θ), where TMS(E,θ) is the MS transmission calculated by SIMION. The integration of f ”(E,θ) over all θ gives NIEDF at the MS detector f ”(E): third output of the model. The results of the model compared to the actual experimental data (measured for HOPG at RT) are presented in the Figure 2.2. The agreement is excellent. The RF- fluctuations of the plasma potential were suppressed by using the ECR source. When using the RF source, the broadening of the peak can be observed for high input power,

X X θMS

υMS

VMS MS

Sheath 2

Vp Plasma

θi θi Sheath 1

Z υi VS Z deviation Surface

Y Y

Figure 2.1. Sketch of NI trajectories between the sample surface and the MS for the sample perpendicular to the MS axis. The black trajectory demonstrates the case of non- collection and the red one – the case of collection of the emitted NI. 32

2. Modeling and reconstruction of NIEDF

SRIM f(E,) with 30% of H f '(E) 0.1 f "(E)

experiment RT

HOPG, ECR 0.01 1 Pa H , 60 W 2 V =5V,V =-130V p s

Normalizedintensity d=37mm,  = 0° + H dominant 3 1E-3 0 10 20 30 40 Energy (eV)

Figure 2.2. Comparison between experimental NIEDF (purple) and calculated NIEDF for HOPG in hydrogen: f(E,θ) emitted from the surface as given by SRIM (black squares), f '(E) at the entrance of the MS given by the model (red with dots), f “(E) at the MS detector given by the model (blue with dots). Eimpact = E0/3 = e (Vp-Vs)/3  45 eV. Experimental conditions: 1.0 Pa of H2 plasma, Q = 5.2 sccm, PECR = 60 W, Vs = -130 V, VMS = 0 V, without screen. but for PRF = 20 W the fluctuations are minimized efficiently by using the grounded screen. As one can see in Figure 2.3, the modeling compares quite well with the experiment also in the case of RF plasma. 2.1.4. Change of surface coverage

One of the important SRIM input parameters influencing the shape of the modeled NIEDF is H/D surface coverage Θ. The typical value used in the calculations corresponded to the hydrogen saturation level in a:C-H films (30% of H/D atoms and 70% of carbon atoms). However, this percentage could vary as a function of the surface state which depends on the material type, surface temperature, etc. The modeling performed for different values of hydrogen surface coverage Θ ranging from 40% to 0% with the step of 10% is shown in Figure 2.4. The value of surface binding energy of hydrogen was kept Ub = 3 eV for all calculations. Note that the curves are normalized by the maximum value. As could be seen from the graph, the tail of NIEDF grows as the coverage diminishes which means that sputtering is being reduced and backscattering makes a bigger contribution to the NI signal (see Figure 2.4). 33

2. Modeling and reconstruction of NIEDF

SRIM f(E,) with 30% of H SRIM f '(E)

0.1 SRIM f "(E)

experiment RT

HOPG, RF 1 Pa H , 20 W 0.01 2 V =5V,V =-130V p s

Normalizedintensity d=37mm,  = 0°

H + dominant 3 1E-3 0 10 20 30 40 Energy (eV) Figure 2.3. Comparison between experimental NIEDF (cyan) and calculated NIEDF for HOPG in hydrogen: f(E,θ) emitted from the surface as given by SRIM (black squares), f '(E) at the entrance of the MS given by the model (red with dots), f “(E) at the MS detector given by the model (blue with dots). Eimpact = E0/3 = e (Vp-Vs)/3  45 eV. Experimental conditions: 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

Another parameter having a similar effect on the NIEDF shape is the surface binding energy. Figure 2.5 shows calculations performed for different values of Ub specified in the inlet. On can notice that changing Ub in the range of 1-3 eV results in the biggest changes of NIEDF shape whereas starting from Ub ≈ 10 eV the NIEDF shape is not modified anymore. These facts can be explained by the influence of the surface binding energy on sputtering probability which approaches zero for the case of Ub > 10 eV.

This leads to a conclusion that change in one of the two parameters (Θ and Ub) results in sputtering/backscattering contribution change which in return alters the shape of NIEDF. Figure 2.6 clearly demonstrates that the NIEDF shape could be changed in equivalent manner by modifying either Θ or Ub in SRIM input parameters. For instance, the black NIEDF curve demonstrates a lower tail which could correspond to the H/D coverage Θ = 40% with Ub = 3 eV or to Θ = 30% with Ub = 2 eV. The other pairs of Θ and Ub were found correspondingly to give the same NIEDF shape (see red and blue curves).

In this work Ub is fixed at 3 eV [79] and Θ is varied. One must keep in mind that deduced Θ values are correct for the condition that Ub does not strongly differ from 3 eV.

2. Modeling and reconstruction of NIEDF 1   HOPG: U = 3 eV b  E = 50 eV, H i 2  0.1 

0.01 Sputtering Backscattering

Normalized intensity Normalized

1E-3

0 5 10 15 20 25 30 35 40 Energy (eV)

Figure 2.4. Comparison between calculated f “(E) at the MS detector given by the model for HOPG in hydrogen for different values of hydrogen surface coverage Θ. The contributions of sputtering and backscattering are shown for Θ = 40%.

U = 1 eV 1 HOPG: = 30% b E = 50 eV, H Ub= 2 eV i 2

Ub= 3 eV

Ub= 4 eV

U = 6 eV 0.1 b U = 10 eV b U = 20 eV

0.01

intensity Normalized

1E-3 0 5 10 15 20 25 30 35 40

Energy (eV) Figure 2.5. Comparison between calculated f “(E) at the MS detector given by the model for HOPG in hydrogen for different values of hydrogen surface binding energy Ub. 35

2. Modeling and reconstruction of NIEDF

HOPG: E = 50 eV, H  1 i 2 U = 3 eV b   U = 2 eV b  = 30% Ub= 6 eV 0.1 Ub= 10 eV

0.01

Normalized intensity Normalized

1E-3 0 5 10 15 20 25 30 35 40

Energy (eV) Figure 2.6. Modeled f “(E) at the MS detector for HOPG in hydrogen for different pairs of surface coverage Θ and surface binding energy Ub resulting in the same NIEDF shape.

2.1.5. Effects of surface roughness In the previous calculations it was assumed that the incidence angle of the incoming ions is 0° (normal incidence). Yet, in case of a rough sample the effective incidence angle of PI could be changed which might influence the sputtering yield of NI [80, 81, 82]. Ahmad Ahmad in his thesis has performed SRIM calculations for HOPG in hydrogen with a certain distribution of PI as a function of impinging angle. The choice of the distribution, however, was not based on any precise physical observations. As a result, a slight (practically negligible) increase in the collection of high energy NI was observed. These calculations were not performed for any other material or ion type. In case of diamond films (MCBDD and MCD) the size of the crystals was in the range of several micrometers which results in PI incidence angle change. The thickness of the sheath in front of the sample is typically in the order of several centimeters, so the crystal orientation does not influence the sheath form. The idea was to test the influence of the surface roughness of a diamond sample (MCD) in D2 plasma on the form of NIEDF. Since MCD is an insulator, it has to be heated to 500°C during the plasma exposure in order to provide the surface conductance and hence the NI signal (see Chapter 3. NI production on different materials). After the plasma exposure, the sample was analyzed by Atomic Force Microscopy (AFM) which gave information about the surface topography (see Figure 2.7). The acquisition and data analysis software Nova NT-MDT

2. Modeling and reconstruction of NIEDF

has allowed to estimate the slopes for all the structures visible on MCD surface. These slope values were basically giving the PI incidence angles. Based on this topographic information, the impinging ions were divided into 4 categories depending on the incidence angle χ: from 0° – 9°, 10° – 19°, 20° – 29°, >30°. The value of the mean angle for each category (6°, 14°, 24° and 35°) and the relative percentages of ions (21%, 32%, 34% and 13% correspondingly) were estimated. This data was used as the input for the SRIM calculation with 45 eV D+ ions impinging on the diamond surface (target material density of ρ = 3.5 g/cm3, Θ = 30% and Ub = 3 eV). Additionally, the mean angle of all impinging PI was estimated based on AFM data. It was found to be 17° and has served as an input for another SRIM calculation. The results of all above mentioned calculations are presented in Figure 2.8 with comparison to the experimental data. One can notice that all the modeled f “(E) correspond quite well to the experiment. The only difference between the modeled curves is the higher collection of high energy NI for the incidence different from normal which fits slightly better to the experimental data. In any case, this result is very similar to that of Ahmad Ahmad and indicates that the surface roughness has a minor influence on the NI yield for the materials under study (HOPG and diamond films).

Figure 2.7. Topographic image measured by AFM on MCD surface previously exposed to D2 plasma while heated to 500°C. The color bar gives the height in micrometers.

2. Modeling and reconstruction of NIEDF

0° 1 MCD: T = 500 °C 17° E = 50 eV, D i 2 6°, 14°, 24°, 35° U = 3 eV b experiment MCD P = 20 W; V = -130 V RF s 0.1

0.01

intensity Normalized

1E-3 0 5 10 15 20 25 30 35 40

Energy (eV)

Figure 2.8. Comparison between the experimental NIEDF of MCD heated to 500°C (green solid curve) and modeled f “(E) at the MS detector for HOPG in deuterium plasma at different PI incidence angles χ: normal incidence (black curve with dots), mean incidence angle of 17° (red), four categories of impinging ions (6° – 21%, 14° – 32%, 24° – 34% and 35° – 13%) as based on AFM data (blue).

However, the study performed in this work is not complete and will have to be pursued. Indeed, the angle of incidence of PI has been taken into account according to the surface roughness, but the angle of NI emission has not been modified. This will be done in the future.

2. Modeling and reconstruction of NIEDF

2.2. Modeling of NIEDF for different tilt angles of the sample

Until now we have considered the case when the sample surface was positioned perpendicularly to the MS axis. In this situation out of all NI emitted from the surface (f(E,θ) provided by SRIM) we collect only a certain part of the distribution (f ”(E,θ) given by the model). This can be observed in Figure 2.9 which represents a polar plot of emitted and collected NI. Each grey dot on the plot is an emitted ion and each blue dot – a collected one. The radial position of NI on the plot represents the energy of a NI normalized by the impact energy Eimpact = 45 eV of impinging PI. The angular position gives the emission angle of a NI with respect to the MS axis. In case of the normal incidence (α = 0°), most of the collected NI have an emission angle θi < 4°, but the biggest possible collection angle is 27°. This result can be explained by the fact that the ions emitted from the surface with small angles and energies have trajectories which can be easily rectified by the electric field present in the sheaths.

Therefore, they arrive to the MS entrance with the angle θMS < θaa and are collected. The

Figure 2.9. Polar plot of NI emitted from HOPG surface (f (E,θ), grey dots) and NI collected by the MS detector (f ”(E,θ), blue dots) as given by the model for V = -130 V. The radial position s gives the energy of a NI normalized by the impinging PI impact energy. The angular position gives the emission angle of a NI with respect to the MS axis. The sketches represent the orientation of the sample with respect to the MS characterized by the tilt angle α. 39

2. Modeling and reconstruction of NIEDF

trajectories of the ions emitted at high energies could not be rectified efficiently enough unless their emission angle is close the normal. In such a way, experimentally one favors the collection of NI emitted with small angles (close to the normal) and/or small energies. When the sample is tilted by an angle α with respect to the MS axis, the sheath in front of the sample would also turn by α (see Figure 2.10). Furthermore, the emission angle

φi (azimuthal angle in spherical coordinates) now must be taken into account (see Figure 2.11), since the cylindrical symmetry around the MS axis is lost (as shown in

Figure 2.12). Only θi (polar angle in spherical coordinates) and Ei had to be taken into account for α = 0°. The 3D calculation is quite similar to the 2D calculation described in [52] and does not present any particular difficulty. It has been realized by Jérôme Dubois during his Master thesis internship.

As a result of the sample tilt, the collected ions (which fulfill the requirement θMS <

θaa) would come from a different part of the initial distribution f(E,θ). This could be observed in the Figure 2.13, where the polar plots for α = 0°, 5°, 10° and 15° are represented in case of HOPG. This leads us to a conclusion that by tilting the sample it is possible to collect NI having any emission angles and energies. In the Figure 2.14 the NIEDF from the experiment and the model are compared for different tilt angles of the sample from 0° to 20° in steps of 5°. When α grows, the minimum energy of the distribution increases. The shift of those minimum energies between the NIEDFs is very well reproduced by the model. The ending energies of the distributions also correspond quite well, a small discrepancy in the distribution tails between the model and the experiment could emerge from the fact that H2+ and H+ populations were not taken into account in the model. The global intensities are decreasing with α and each new curve fits to the end of the previous one. This is also

Figure 2.10. Scheme of the sample rotated by angle α with respect to the MS.

2. Modeling and reconstruction of NIEDF x v θ

z φ y Figure 2.11. Scheme of the speed vector of an ion emitted from the sample.

a) S θ Plasma

u S

r h e φ f a S a t h c h e a e θ t h

 Figure 2.12. Scheme of the sample with respect to the MS for normal incidence (α=0°) and for rotation by angle α; example of two trajectories: φ=0° and φ=180°, same E and θ. a) α=0°, for a given θ, all trajectories towards the MS are equivalent, independently of φ.

b) θ Plasma S u φ S r h f e a a t S h h c e e a θ t

≠ b) α≠0°, for a given θ, all trajectories towards the MS depend on φ. 41

2. Modeling and reconstruction of NIEDF

well reproduced by the model. However, the decrease of intensity of the low-energy peak between α = 0° and α ≠ 0° given by the model does not match the experiment. This issue is not yet understood. Let us note that the important drop of intensity in the NIEDF low-energy peak between α = 0° and α = 2° predicted by the model is due to the loss of symmetry in the ion collection (at α = 0° all ions are collected independently of their azimuthal angle of emission φi).

Figure 2.13. The polar plots of NI emitted from HOPG surface (f (E,θ), grey dots) and NI collected by the mass-spectrometer detector (f ”(E,θ), blue dots) as given by the model for different tilt angles α of the sample. The surface bias is Vs = -130 V. The radial position gives the NI emission energy normalized to the impact energy of the PI Eimpact = 45 eV. Black line is the normalized angular distribution of the emitted NI. Red line is the normalized angular distribution of the collected NI. 42

2. Modeling and reconstruction of NIEDF

1 Lines: experiment 0 ECR H 1Pa 60W 2 2 5 n =2.5x1015, T =0.95 eV e e 10 15 Symbols: model + 20 0.1 35%, H only,  =1° 3 aa

0.01

1E-3 Normalizedintensity

1E-4 0 10 20 30 40 50 Energy (eV) Figure 2.14. Comparison of the experimental NIEDF (solid lines) with the modeled ones (points) for different tilts of the sample (from α = 0° to 20°, with a step of 5°). All the modeled distributions are divided by the number corresponding to the maximum intensity of the experimental NIEDF at α = 2°.

Figure 2.15. Maps showing the spatial distributions of the origin on the sample surface of NI collected by the mass spectrometer. The maps are drawn for tilts of the sample varying from α = 0° to 20°, in steps of 5°. Uniform emission distribution in E, θ, φ is assumed. Plasma parameters (ne = 2·1015 m−3, Te = 1eV, Vp = 3V, Vs = −130V) used in the model correspond to ECR plasma conditions (1 Pa, 60 W). 43

2. Modeling and reconstruction of NIEDF

Figure 2.15 demonstrates that the NI collected at α = 0° originate from a circle on the sample. When the symmetry is broken (α ≠ 0°), only a part of the distribution in φ is collected and the signal strongly decreases. Moreover, the spot of origin of the ions transforms into an ellipse and shifts perpendicularly to the rotation axis. The spot remains small in dimensions (~2 mm) compared to the total sample surface (8 mm in diameter). Despite the shift, it can be seen that even at the tilt angle α = 30°, NI never originate from the edge of the sample. Thereby we prove that the measured intensity is not limited by the sample dimensions. It is only limited by the acceptance angle of the MS. Let us note that this calculation has been done for a uniform distribution (i.e. f UNI = 1 ∀ θ, E, φ). Similar results are obtained using the SRIM distribution for HOPG. 2.3. Conclusions on the NIEDF modeling The remarkable agreement of the model with experiment for HOPG confirms that the proper description of the diagnostic technique included in the model leads to correct interpretation of the measured NIEDF. It also verifies the choice of SRIM to provide the correct initial distribution f(E,θ) for carbon materials. At last, it proves that ionization probability Piz = const ∀ E and θ (independently of the neutral particle energy and angle of emission). However, this strong assumption may not be always valid, particularly for metals for which Piz depends on the perpendicular velocity of the outgoing particle. Furthermore, the use of this modelling method supposes that SRIM parameters for the given material and experimental conditions are known. This is the case for a-C:H films because of several decades of fusion research on carbon, but this may not be the case for other materials. Therefore, it would be interesting to determine the distribution in energy and angle of NI emitted from the surface in case when one cannot obtain an initial guess of f(E,θ), whether because Piz ≠ const or because SRIM input parameters are unknown. This would allow characterizing NI production on the surfaces of other materials like low work-function metals or insulators. 2.4. Reconstruction of surface-produced NIEDF

As it could be seen from Figure 2.13, for a given tilt α of the sample the collected ions come from a certain angular range of emission. Therefore, it seems possible to reconstruct the whole distribution of NI leaving the surface based on the NIEDF measurements at different tilt angles of the sample. For a given tilt angle α, NI measured at certain energy E originate from a certain angular range [θα a θα b] (see Figure 2.11 and Figure 2.17). For a different tilt angle α ’,

NI measured at E originate from a different angular range [θα ’ a θα ‘ b] (see Figure 2.11 and Figure 2.17). By choosing tilt angles in a way so that the angular ranges of NI origin are adjacent (i.e. θα b = θα ’ a…), one can obtain information about ions emitted at energy E 44

2. Modeling and reconstruction of NIEDF

with any angle. However, the corresponding NIEDF experimental signals Iα(E) have to be corrected to take into account the collection efficiency which depends on α (i.e. ions at energy E are collected less efficiently at α ’ than at α). The “collection efficiency” calculation is detailed later. Identifying adjacent angular ranges allows to determine the angular distribution function for each energy E from a set of NIEDFs measured at different α. After the correction, by summing up the matrix elements for all energies one obtains the angular distribution of NI produced on the surface, and in case of the sum for all emission angles – the energy distribution of surface-produced NI. The details of the model are given in the following sections. 2.4.1. Determination of adjacent angular ranges To be able to determine surface-produced NI distribution from the measured NIEDF, one has to collect NI originating from the entire range of emission angles θ. This is done by doing measurements at different tilt angles α. In order to prevent overlaps, one has to determine the tilt angles giving adjacent θ intervals of the collected NI. The step in energy in the model was set to 1 eV and the maximum energy recorded was 64 eV. There are thus 64 energy values for which the values of α giving adjacent angular ranges were determined with a dedicated Scilab routine. The results are illustrated in Figure 2.16 which gives for 6 different energies the adjacent θ angular intervals (colors in the figure) and the corresponding α values (numbers below the figure). This calculation relies only on ion trajectories between the surface and the MS, and is independent of the NI emission distribution function f(E,θ). Therefore, it can be done with the model presented before without any assumption on f(E,θ). From these values of α, one can select the corresponding NIEDF which will be used to reconstruct the distribution function on the surface at this energy (as shown in Figure 2.17). One can note that α values are not integer numbers while α resolution in the experiment is 1°. As a consequence, α is rounded to the closest integer and the corresponding NIEDF fα(E) is used. For instance, NIEDF measured at α = 0° – 2° – 4° – 6° will be used to obtain the angular distribution of ions emitted from the surface at energy 1 eV (see Figure 2.16). NIEDF were measured from α=0° to 35°, with a step of 1°. 2.4.2. Collection efficiency One has to note that NIEDF experimental intensities at each energy for given α are ions reaching the detector of the MS, which represent only a part of ions emitted from the sample. In order to obtain the distribution at the surface of the sample, one should take into account the ions which were not collected, meaning that the collection efficiency versus α and E must be calculated. Since the transmission inside the MS doesn’t significantly change the shape of the NIEDF (see Section 2.5. Results and discussion for details), it will not be taken into account in the calculation of the collection efficiency. In the following paragraphs, the correction of the distributions fα(E) 45

2. Modeling and reconstruction of NIEDF 0  deg i 30 10

6 60

eV i

E 4

0 0 90

E = 8 eV E = 10 eV E = 1 eV E = 2 eV E = 4 eV E = 6 eV 2 0 0 0 0 0 0 1.89 1,91 1,91 1,88 1,88 1,88 3.78 3,81 3,78 3,75 3,75 3,75 4 5.67 5,72 5,66 5,63 5,63 5,63 7,61 7,53 7,5 7,56 7,5 9,41 9,41 9,44 9,38 6 11,28 11,31 11,25 12,43 13,19 13,13 15,02 8 15,64 Figure 2.16. Adjacent intervals of the emission angle θ (colors in figure) for the collected NI of 6 different energies and the corresponding tilt angle α values (numbers 10 below the figure). Conditions used: ne = 4·1015 m−3, Te = 1 eV, Vp = 7V, Vs = −130V, VMS = 0 V,

Eimpact = E0/3 = e(Vp − Vs)/3 ≈ 48 eV correspond to ECR plasma (1 Pa, 60 W).

Ei = 20 eV

HOPG 1Pa 60W ECR with  = 10000 0° 5° 16° 22° 1000

100

H-intensity (a.u.)

0 10 20 30 40 50 Energy (eV) Figure 2.17. Illustration of the model principle. For a given energy E (example for Ei = 20 eV) and a given tilt angle α, the NIEDF experimental signal (left) results from the ions emitted from the surface with different angles θ (see θ intervals on the right). 46

2. Modeling and reconstruction of NIEDF

collected by the MS (taking into account the collection efficiency at a given α and E) will be described in detail.  Each ion has a speed vector v defined by three parameters in spherical coordinates: E, θ and φ, as shown in Figure 2.11. In the case of normal incidence (α=0°), for a given θ angle and a given energy E, the collected ions originate from a circle on the sample, taking into account the radial symmetry of the system: φ ∈ [0°:360°]. At α≠0°, the system is not symmetrical anymore along the spectrometer axis. Therefore, not all the ions at a given θ and E are collected by the MS as can be seen from Figure 2.12(b). In order to illustrate the collection efficiency dependence on α and E, the following calculation was performed. A uniform distribution E, θ and φ of ions emitted from the surface was created. Then, transmission of these ions inside the sheaths and inside the plasma for α ∈ [0°:35°] with a step of 1 was calculated by the model presented in Section 2.1. Modeling of NIEDF for sample normal to the MS axis. As a result, one can represent for each energy E and each tilt angle α the values of the emission angles θ and φ of the collected ions (see Figure 2.18). Each dot in Figure 2.18 symbolizes a collected ion with a unique combination of angles θ and φ, and the color indicates the tilt angle α. One can see from Figure 2.18 that for α=0° ions are collected from a given θ interval (~ 0°–10° at 2 eV) without any restriction on φ angle (from 0° to 360°). On the contrary, for α=2° there is a strong reduction in the φ angles that lead to collection of the ions (φ range is approximately from 60° to 120°). One can also see in this graph that the θ intervals leading to collection of ions are broader at low energy than at high energy. Therefore, the calculation will have a better angular resolution in θ for ions with high energy. Finally, the most important point is that the number of collected ions (number of dots of the same color in Figure 2.18) is strongly dependent on E and α. The collection efficiency is varying with E and α.

90 90  = 0 E = 2 eV  = 0 E = 17 eV 80  = 2 80  = 2  = 4  = 4 70  = 6 70  = 6  = 8 60 60  = 10  = 12 50 50  = 14 40 40  = 16  = 18

 = 20 Theta, deg Theta, Theta, deg Theta, 30 30 20 20 10 10 0 0 40 50 60 70 80 90 100 110 120 130 140 40 50 60 70 80 90 100 110 120 130 140

Phi, deg Phi, deg

Figure 2.18. Distribution f ’(θ, φ) of the NI collected by the MS obtained from different sample tilt angles α for a given energy: Ei = 2 eV (left) and Ei = 17 eV (right). The initial distribution f UNI(E, θ) has been taken uniform in E, θ and φ. 47

2. Modeling and reconstruction of NIEDF

We can therefore write the expression for the ion signal measured for a given tilt angle α and a given energy E as follows (transmission inside the MS is neglected):

I (E)  f (E,)sin d d Eq. (2.1)     ( ) In this formula, Δθ is the interval of polar angles leading to a possible collection of the ion emitted at energy E, for a tilt angle α. Outside of this interval Δθ, there is absolutely no collection of ions emitted at energy E. Inside this interval Δθ, only ions emitted over a certain azimuthal angle interval Δφ(θ) are collected by the MS. The ions outside this interval either have the exceeding deviation or the arrival angle which is higher than the acceptance angle. Iα (E) is given by the experiment and f (E, θ) is the distribution of ions emitted by the surface that we intend to determine. The simplest way to do the calculation is to assume that f is constant over the interval Δθ (f is constant over any Δφ interval since there is no ion emission dependence on the azimuthal angle). Under this assumption, the formula becomes:

I (E)  f (E,) 1sin d d Eq. (2.2)     ( )

The integral represents the signal that would be collected at E and α for a uniform distribution of surface-produced ions (f UNI = 1 whatever θ and φ). This is the collection efficiency at E and α. The integral is the sum of all ions collected for a given E and α, it is the sum of all dots of the same color in Figure 2.18. To compute the integral, we have first created a uniform fictive surface distribution with an equal repartition in E, θ and φ (f UNI = 1 for any E, θ and φ). The model presented in Section 2.1. Modeling of NIEDF for sample normal to the MS axis is then used to compute f UNI’ (E) for each α. It leads to a matrix in E and α where the elements are the collection efficiencies (the sum of all ions collected at these given E and α values). Then, f (E,θ) is obtained through: I (E) I (E) f (E,  )     Eq. (2.3) f UNI' (E) f UNI sin d d     ( )

where f α UNI’ (E) is computed using the model. Before concluding, let us raise several points.

 The collection efficiency is basically a calculation of ions trajectories across the sheath and the plasma. Therefore, it depends on E and α, but also on the ions flux and surface bias which determine the potential distribution in the sheaths (Child- Langmuir law). The collection efficiency matrix has to be calculated for all experimental conditions investigated. 48

2. Modeling and reconstruction of NIEDF

 The calculation allows to determine f (E,θ) with a step of 1 eV in E (chosen by us) and with a step in θ (Δθ) which depends on E and which is not constant even for a given E (Δθ is varying with θ as shown in Figure 2.16). Δθ is the θ interval range which leads to ion collection for a given E and α. For a given experimental condition the only way to decrease the Δθ intervals is to decrease the acceptance angle of the MS (lower θaa leads to a lower θ interval resulting in ion collection for a given E and α). However, this decreases the measured NI signal and hence, the method sensitivity.  One possible improvement of the model would be to model the distribution f inside the interval Δθ by a linear function f = aθ + b rather than by a constant value. This possibility has not been explored during this thesis. This method allows to compute f (E,θ) without any a priori assumption on f and any assumption on the ionization probability. The method can therefore be applied to any type of material and NI. The f (E) distribution for various θ and f (θ) for various E are represented in Figure 2.19 and Figure 2.20.

8 2.0x10 Energies 8 1 eV 1.8x10 2 eV 8 1.6x10 3 eV

8 5 eV 1.4x10 10 eV 1.2x108 20 eV 30 eV 1.0x108 40 eV 8.0x107 64 eV

7 arb.u. Intensity, 6.0x10 4.0x107

7 2.0x10 0.0

0 10 20 30 40 50 60 70 80 90  emission, deg

Figure 2.19. Individual angular distribution functions for each energy E ∈ [1:64] eV with a step of 1 after the correction by the collection efficiency.

2. Modeling and reconstruction of NIEDF

0.0

108 3.0 6.0 9.0 20.1 30.0 107 60.0 89.7

106 arb.u. Intensity,

105

0 5 10 15 20 25 30 35 40 45 50 55 60 65 Energy, eV

Figure 2.20. Individual energy distribution functions for each emission angle θ ∈ [0°:90°] with a step of 0.3 after the correction by the collection efficiency.

2.5. Results and discussion

2.5.1. Experimental conditions The experiments were carried out on PHISIS set-up in hydrogen plasma, but with conditions different from usual. In order to get the best angular resolution possible, the acceptance angle of the MS was reduced by decreasing the voltage of the lens 1 (L1), whose function is to focus the ions coming from the entrance to the energy filter. The reduction of θaa results in the signal intensity drop: a compromise was found to set the L1 to the lowest value allowing to keep an acceptable signal-to-noise ratio, even at large angles. SIMION modeling shows that with this setting θaa = 1° (L1 = 125 V), whereas in usual conditions θaa = 2° (L1 = 252 V). Reduced L1 value strongly decreased the intensity of the recorded NIEDF due to less efficient focusing of ions, and the influence on low energy ions was more noticeable than on high energy ones. However, this has resulted only in slight change of the NIEDF shape: the tail in normalized distributions f “ (E,θ) was a bit higher as compared to f ‘ (E,θ). Therefore, it was decided not to take into account the transport of NI inside the MS for the conditions with reduced L1 for the sake of simplicity.

2. Modeling and reconstruction of NIEDF

Since the rotation and reduced L1 value have strongly decreased the intensity of the recorded NIEDF, high density plasma was necessary to compensate for the signal decrease. Instead of the RF source creating a capacitively coupled plasma in the source chamber, the ECR source from “Boreal Plasmas” was inserted into the diffusion chamber from the top right flange shown in Figure 1.2 (a) and Figure 1.11. Following the experimental protocol established in Chapter 1. Negative ion surface production measurements, reliable angular-resolved measurements were performed in H2 ECR plasma for a pressure of 1 Pa and 60 W of effectively absorbed power. The rotation of the sample could be controlled within 1° thanks to an external dial measuring the tilt angle α. The measurements were realized in one set for α ∈ [0°:35°] periodically checking the signal at α=0° (see Section 1.3.3. ECR plasma). No NI signal decrease was noticed during the whole series of measurements, but only some slight fluctuations, so no signal correction had to be applied. The result of the measurements is represented in Figure 2.21 and demonstrates a clear shift of the NIEDF onset with angle α. Figure 2.22 shows the comparison of several experimental NIEDF measured in opposite rotation directions with respect to the normal incidence (α = 0°). A slight discrepancy in the NIEDF shift could be explained by a limited precision in the estimation of the tilt angle α (which plays an important role for large α). The difference in signal levels also becomes evident with the increase of α: anticlockwise rotation demonstrates a higher signal than clockwise rotation (see illustration in Figure 2.23). The differences in NIEDF observed while rotating the sample in the opposite directions could originate from spatially inhomogeneous electron density created by the ECR source. In order to check the positive ion flux incident on the sample surface for both directions of rotation, the following procedure has been realized. HOPG sample was replaced with a copper (Cu) plate partially covered by insulating capton scotch, so it could be biased independently from the sample holder via an insulated wire. The surface exposed to plasma represented a square inscribed in 4 mm radius circle and had an area of 32 mm2 (see Figure 2.24). In this way, sample and sample holder acted as a planar probe with a guard ring allowing to measure the ion saturation current. Before the measurements, the sample surface was cleaned during 30 min under

Vs = –200 V bias in Ar plasma created by the ECR source. It was done in order to remove the impurities from the Cu surface which could not be removed by acetone, ethanol or mechanically. After the cleaning, the PI saturation current was measured in usual ECR plasma conditions as a function of the sample tilt angle α. It has been checked by scanning Vs that the PI current is indeed saturating when the sample is negatively biased in the range from Vs = –20 V to –100 V (see Figure 2.25). For the sample bias going below –100 V, the PI current starts to grow gradually. However, for our usual experimental conditions with Vs = –130 V the current has increased by less than 10% as compared to the PI saturation current, as can be seen from Figure 2.25. Therefore, we can consider the PI current at Vs = –130 V as nearly PI saturation current and the sheath in front of the sample holder as planar. 51

2. Modeling and reconstruction of NIEDF

HOPG 1Pa 60W 100000 ECR with  = 0° 11° 22° 33° 1° 12° 23° 34° 2° 13° 24° 35° 3° 14° 25° 10000 4° 15° 26° 5° 16° 27° 6° 17° 28° 7° 18° 29° 1000 8° 19° 30° 9° 20° 31° 10° 21° 32°

NI intensity, cts/s 100

10 0 10 20 30 40 50 60 70 80

Energy (eV)

Figure 2.21. Experimentally measured NIEDF for different tilts of HOPG sample: from α = 0° to 35°, with a step of 1°. Rotation direction: anticlockwise.

HOPG 1Pa 60W 100000 wave with  = 0° 1° 5° 10° 10000 15° 20° 25° 30° 1000 Bold - anticlockwise Thin - clockwise

NI intensity, cts/s 100

10 0 10 20 30 40 50 60 70 80 Energy (eV)

Figure 2.22. Comparison of experimental NIEDF measured on HOPG in opposite rotation directions with respect to the normal incidence (α = 0°): anticlockwise and clockwise. 52

2. Modeling and reconstruction of NIEDF

(a) Source chamber (b) Source chamber

Plasma clockwise Plasma

anti - clockwise

< >

Figure 2.23. Illustration of the sample rotation direction: (a) anticlockwise – sample is rotating towards the source chamber; (b) clockwise – sample is rotating towards the bottom of the diffusion chamber.

capton

Cu Mo

Figure 2.24. Capton insulated copper sample placed inside the sample holder. The area of the square exposed to plasma is 32 mm2.

2. Modeling and reconstruction of NIEDF

0 I  = 0° sample P = 65 W, H -10 ECR 2

A p = 1 Pa  -20

-30

Cusample current, -40

-50

-200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0 20

Surface bias V , V s Figure 2.25. I-V characteristics measured by the Cu sample (planar probe with a guard ring). The signal saturates between –20V and –100V, only slightly increasing below –100V.

1.25 P = 65 W, H ECR 2 p = 1 Pa 1.20 I V = -130V 1.15 sample s I V = -130V holder s 1.10 linear fit

1.05

1.00

0.95 Equation y = a + b*x Weight No Weighting 0.90 Residual Sum of 0.00105 Squares

Normalizedarb.u. current, Pearson's r -0.99819 0.85 Adj. R-Square 0.99613 Value Standard Error ?$OP:A=1 Intercept 1.00845 0.00203 0.80 ?$OP:A=1 Slope -0.00531 8.27641E-5

-40 -30 -20 -10 0 10 20 30 40 Tilt angle , deg

Figure 2.26. PI currents measured in ECR hydrogen plasma: on the Cu sample individually with total applied bias of Vs = 0 V and Vs = –130 V; on the sample holder with Vs = –130 V as a function of the tilt angle α. The negative values of α correspond to anticlockwise rotation, and positive – to clockwise rotation. 54

2. Modeling and reconstruction of NIEDF

Figure 2.26 shows PI saturation currents collected by the sample individually with

Vs = –130 V, as well as PI current collected by the whole sample holder at Vs = –130 V. The sample holder collects the current from both front and back sides. The sample holder current is therefore constant and does not represent a proper measurement of the PI flux incident on the sample as a function of the tilt angle α. On the other hand, inhomogeneous electron density in the plasma results in increase or decrease of the PI flux depending on the direction of rotation. For the usual conditions (Vs = –130 V), the dependence of the PI current on angle α is nearly linear, as can be concluded from the fit (see Figure 2.26). In order to compensate for the incident PI flux difference in both rotation directions, the PI current was fitted by a line and the correction was applied to the experimental NIEDF according to the tilt angle at which they were measured. The difference between the current at normal incidence α = 0° and at αmax = ±35° was found to be ~15-20%. The result of the correction can be better seen for large values of α, as demonstrated for NIEDF at α = 25° and α = 30° measured for both rotation directions (Figure 2.27). After the correction, the signal levels approached each other (see red and blue curves: solid for anticlockwise rotation and dotted for clockwise rotation), but the difference remains non-negligible. It could be connected to the axis of sample holder rotation which is not

10000

HOPG 1Pa 60W ECR with  = 25° before 25° after 30° before

1000 30° after calibration by incident PI flux Direction: Bold - left dot-dash - right

100

NI intensity, cts/s

10 10 20 30 40 50 60 70 80 Energy (eV)

Figure 2.27. Comparison of experimental NIEDF measured on HOPG at α=25° and α = 30° in opposite rotation directions: anticlockwise (solid curves) and clockwise (dotted curves). Black and green curves – NIEDF before correction, red and blue curves – after correction by the incident PI flux.

2. Modeling and reconstruction of NIEDF

perfectly aligned with the Y axis (see Figure 2.1, Y axis is perpendicular to the MS axis X). This problem can be solved only by purchasing a more accurate sample holder device. Hereafter, the NIEDF measured while rotating anticlockwise will be used in the reconstruction of the surface-produced distributions, since the signal/noise ratio is bigger. The correction by the incident PI flux was omitted in the further data manipulations due to its minor influence. 2.5.2. HOPG The surface-produced energy distribution function for HOPG reconstructed from the experimental NIEDF by the above described method is represented in the Figure 2.28 by green curve. The initial f SRIM(E) given by SRIM for 45 eV impinging ion energy is shown in black for comparison. This distribution f SRIM(E) is considered to represent quite accurately the NI emitted by the surface except maybe for this high energy. Indeed, the modeling of NIEDF based on this distribution fits quite well to the experiment with the exception of high energy (see Figure 2.14). The agreement with the reconstructed f EXP(E) distribution (green in Figure 2.28) is not excellent, especially at high energy. The disagreement can have several origins such as experimental artifacts (we have pointed out one of them previously), incorrectness of the SRIM distribution (a slight change of f SRIM(E) could lead to a better global agreement in Figure 2.14 and in Figure 2.28, in particular for the high energy part of the distribution), or an improper definition of the reconstruction problem. To get rid of this last possibility, we have used the following method. The f SRIM(E,θ) has been used to compute the NIEDF versus α (such as in Figure 2.14). These NIEDF has been used as fictive “experimental” NIEDF and the reconstruction method has been applied using these functions. The reconstruction method should give in return f REC(E,θ) = f SRIM(E,θ) since there is no possibility of experimental artifact, or of bad choice of the initial f(E,θ) distribution. The result of the reconstruction is shown in red demonstrating the validity of the reconstruction procedure. Indeed, apart from a slight overestimation of the low energy ions (E < 2 eV) the overall agreement with the initial f SRIM(E) distribution is good. Another type of f (E,θ) representation is a contour plot which shows by a corresponding color the number of NI emitted from the surface having a unique pair of values E and θ. The representation of f SRIM(E,θ) is shown in Figure 2.33 for 107 H3+ ions and the reconstructed distribution f REC(E,θ) in Figure 2.34. By comparing these two figures, one can observe that the reconstruction has a reduced statistics and works less accurately for low energies and high emission angles. This fact can be explained by looking at Figure 2.19 showing f EXP(θ) for various energies. It demonstrates the disadvantage of the reconstruction method: the adjacent intervals for low energies are much wider than for high energies which reduces the precision of the reconstruction. Moreover, Figure 2.20 showing f EXP(E) for various values of emission angle θ demonstrates a reduced signal-to-noise ratio of the distributions with high θ values.

2. Modeling and reconstruction of NIEDF

Despite the disadvantages of the method for low energies and high emission angles, the reconstruction can be regarded as successful. The disagreement between f SRIM(E) and the experimental f EXP(E) distribution is noticeable. One can notice in particular that the distribution reconstructed from experiment has a high energy tail not present in f SRIM(E). Combining this with the fact that our model never fits correctly the high energy tail of NIEDF (see Figure 2.14), one can conclude that the reason for disagreement may be the approximation used for the

SRIM calculation. In the SRIM simulations, we considered only H3+ at 45 eV impinging energy (Vs = -130 V). In reality, our plasma is populated with H3+ and H2+ which constitute 88% and 12% of the impinging PI flux correspondingly (for 1 Pa and

PECR = 60 W). The population of H+ makes a very small contribution in terms of ion flux and can be neglected. Therefore, SRIM simulation should rather include two groups of ions: 88% of ions impinging at 45 eV and 12% of ions at 70 eV. Furthermore, each ion population is distributed in energy because of collisions and/or plasma potential variation inside the plasma. Therefore, impinging H3+ and H2+ have a distribution in energy around the average value of 45 eV and 70 eV. Figure 2.29 shows the decomposition of these two populations by several groups of ions with different energies (the percentages are also given).

0.1 SRIM HOPG + 7 H 100% 45 eV 10 ions 3

Normalizedintensity reconstructed from SRIM HOPG

reconstructed

from exp HOPG

0.01 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Energy, eV

Figure 2.28. NI energy distributions emitted from HOPG surface: f SRIM(E) given by SRIM for 100% of PI at 45 eV (only H3+ population, total ion number 107) shown in black; f REC(E) reconstructed from SRIM shown in red and f EXP(E) reconstructed from experiment performed in ECR H2 plasma shown in green. 57

2. Modeling and reconstruction of NIEDF

4.0x10 5 1x105 29.4 % 34.8 % + 5 + 5 H experiment 1x10 H experiment 3.5x10 3 2 45 eV 4 24.8 % Histogram 9x10 Histogram 5 3.0x10 8x104 70 eV 4 5 45.3 eV 7x10 2.5x10 17.9 % 16.5 % 6x104 22 % 5 19.8 % 2.0x10 44.7 eV 70.5 eV 5x104 69.5 eV 45.7 eV 15.5 %

1.5x105 4x104 Intensity, cts/s Intensity, 4 71 eV 7.9 % 3x10 1.0x105 46 eV 2x104 4 4 3.4 % 4 % 5.0x10 2.2 % 1x10 1.3 % 0.0 0 44.3 eV 46.3 eV 69 eV 71.5 eV 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 Energy, eV Energy, eV Figure 2.29. Experimentally measured PI distributions of H3+ (left) and H2+ (right) approximated by several groups of ions with different energies (represented by histograms). The energies of the ion groups when impinging on the sample and their percentages within the given ion population are specified.

6 SRIM HOPG 10 ions 0.1 H + 88% 44-46 eV 3 H + 12% 69-72 eV 2 Normalizedintensity reconstructed from SRIM HOPG reconstructed from exp HOPG 0.01 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Energy, eV

Figure 2.30. NI energy distributions emitted from HOPG surface: f SRIM(E) given by SRIM for 88% of H3+ and 12% of H2+ accounting their energy distributions (total ion number 106) shown in black; f REC(E) reconstructed from SRIM shown in red and f EXP(E) reconstructed from experiment performed in ECR H2 plasma shown in green.

2. Modeling and reconstruction of NIEDF

All this information was included in the SRIM calculation and yielded the result shown in Figure 2.30. The number of ions used in the SRIM calculation was diminished (106 instead of 107) in order to reduce the computational time, because 13 groups of ions had to be introduced in the software manually one after another. As could be seen from Figure 2.30, the experimentally reconstructed distribution f EXP(E) fits fNEW SRIM(E) calculated by SRIM much better than previously, thanks to the appearance of high energy tail resulting from the contribution of H2+.

The reconstruction fNEW REC(E,θ) of the improved distribution fNEW SRIM(E) is shown in Figure 2.30 in red. The curve is noisy due to the reduced number of ions used in the SRIM calculation (106). Nevertheless, it is visible that the reconstruction method is valid for even more complex initial SRIM distribution. The improved distribution fNEW SRIM(E) was also used for modeling of NIEDF as a function of the tilt angle α (analogously to

Figure 2.14). The agreement between calculated f ‘ NEW SRIM(E) and experimental NIEDF improved considerably for high NI energies, as demonstrated in Figure 2.31. One should note that both ion populations during the SRIM calculation which has produced initial fNEW SRIM(E) for Figure 2.31 were considered as monoenergetic (i.e. H3+ energy is

45 eV only and H2+ energy is 70 eV only) in order to increase the statistics (107 ions). In general, taking into account the relative PI populations and their energy distributions significantly improves the agreement between SRIM and the reconstructed distribution. The agreement is still not perfect, but the absolute measurement of relative ion populations (H3+, H2+ and H+) can be quite difficult due to the different transmission of the considered PI inside the MS. Taking into account all experimental and computational assumptions and/or difficulties, the agreement between f EXP(E) reconstructed from the experiment and fNEW SRIM(E) can be considered as good and the reconstruction method can be considered as valid. The result of the reconstruction method for the angular distributions produced on the sample surface is shown in Figure 2.32 for 100% of PI at 45 eV. By comparing black and red curves, one could notice that the method is in agreement with SRIM calculations for the emission angle range θ ∈ [5°:70°]. The reconstructed distribution is quite close to the one predicted by SRIM, but the maximum of emitted NI is located at 45° comparing to 40° given by SRIM. The result of f EXP(E,θ) reconstruction is shown in Figure 2.35in a contour plot. By comparing it to the prediction of SRIM (see Figure 2.33), we see that the f (E,θ) maxima positions are in agreement: the energy is 3–5 eV, and the emission angle range is θ ∈ [20°:45°]. However, the plateau predicted by SRIM for E ∈ [5eV : 30eV] and θ ∈ [20°:60°] shown in red (corresponding to a big number of particles) is absent in case of the distribution reconstructed from experiment. The intensity decrease with energy for f EXP(E,θ) is rather continuous starting from E ≈ 7 eV rather than abrupt like for f SRIM(E,θ) at E >30 eV.

2. Modeling and reconstruction of NIEDF

Lines: experiment 1 0 ECR H 1Pa 60W 2 2 5 n =2.5x1015, T =0.95 eV e e 10 Symbols: model 15 20 0.1 30%,  =1° aa H + 88%, H + 12% 3 2 0.01

1E-3 Normalizedintensity

1E-4 0 10 20 30 40 50

Energy (eV) Figure 2.31. Comparison of the experimental NIEDF (solid lines) with the modeled ones (points) for different tilts of the sample (from α = 0° to 20°, with a step of 5°). SRIM distribution fNEW SRIM(E) used in the model takes into account H3+ and H2+ ion populations. All the modeled distributions are divided by the number corresponding to the maximum intensity of the experimental NIEDF at α = 2°.

1.0

0.9

0.8 0.7

0.6

0.5 0.4 SRIM HOPG H + 100% 45 eV 107 ions 0.3 3 Normalizedintensity reconstructed 0.2 from SRIM HOPG 0.1 reconstructed from exp HOPG 0.0

0 10 20 30 40 50 60 70 80 90  emission, deg Figure 2.32. NI angular distributions emitted from HOPG surface: f SRIM(θ) given by SRIM for 100% of PI at 45 eV (only H3+ population, total ion number 107) shown in black; f REC(θ) reconstructed from SRIM shown in red and f EXP(θ) reconstructed from experiment performed in ECR H2 plasma shown in green. 60

2. Modeling and reconstruction of NIEDF 90 0.000

80 7.909

70 15.82

60 23.73

, deg ,  50 31.64

39.55 40 47.45 30

52.20 Emissionangle 20

0 10 20 30 40 50 60

Energy, eV

Figure 2.33. Contour plot representing f SRIM(E,θ) as predicted by SRIM for HOPG, 100% of PI at 45 eV (only H3+ population, total ion number 107). The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ. 90 0.000

80 7.545

70 15.09

60 22.64

deg ,  30.18 50

37.73 40

45.27

30 49.80

Emissionangle 20

0 10 20 30 40 50 60

Energy, eV Figure 2.34. Contour plot representing f REC(E,θ) reconstructed from SRIM data for HOPG, 100% of PI at 45 eV (only H3+ population, total ion number 107). The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ. 61

2. Modeling and reconstruction of NIEDF 90 0.000

80 2.909E+07

70 5.818E+07

60 8.727E+07

, deg ,

 1.164E+08 50

1.455E+08 40 1.745E+08 30

1.920E+08 Emissionangle 20

0 10 20 30 40 50 60

Energy, eV

Figure 2.35. Contour plot representing f EXP(E,θ) reconstructed from the experiment on HOPG in ECR H2 plasma. The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ.

Another way to check the agreement of the surface-produced NI distribution predicted by SRIM and the experimental reconstruction is to represent the number of emitted particles as a function of their perpendicular velocity. The υ┴ value can be calculated from the known υix, υiy, υiz given by the TRIMOUT.txt file or for known E and θ of the emitted NI by the following expression:

2E   cos Eq. (2.4) m Then, the emitted NI are summed up according to the value of their perpendicular velocity for both cases which is shown in Figure 2.36. One can see that the agreement between the distributions is less good for big perpendicular velocities. Such NI usually appear in the tail of the NIEDF for which SRIM has proven to be less accurate. Considering that reconstructed f EXP(E,θ) represents a real distribution of the NI emitted from the surface and is more precise than SRIM, it has been used as input for NIEDF modeling described in Section 2.1. Modeling of NIEDF for sample normal to the MS axis (in case of α = 0°). The original TRIMOUT.txt file produced by SRIM was replaced by a matrix containing f EXP(E,θ). The result is shown in Figure 2.37 (f EXP(E,θ) and f EXP‘(E,θ) by black and red points correspondingly) and compared to the experimental data for HOPG in ECR plasma (purple curve). The agreement is absolutely excellent which confirms the coherence of both: NIEDF modeling and the reconstruction method. 62

2. Modeling and reconstruction of NIEDF

0.1

f(V ) for HOPG perp 6 0.01 SRIM 10 ions + H 88% 44-46 eV intensity Normalized 3 H + 12% 69-72 eV 2 reconstructed from experiment 1E-3 4 4 4 4 5 5 0.0 2.0x10 4.0x10 6.0x10 8.0x10 1.0x10 1.2x10

Vperp, m/s SRIM + + Figure 2.36. f (υ┴) as predicted by SRIM for HOPG with 88% of H3 and 12% of H2 6 EXP accounting their energy distributions for 10 incident ions (black curve) and f (υ┴) reconstructed from the experimental data (red curve). 1 HOPG, ECR 1 Pa H , 60 W 2 V =-130V,  = 0° s 0.1

0.01

EXP 1E-3 f (E,) reconstructed EXP

f '(E,) reconstructed Normalizedintensity experiment HOPG

1E-4 0 10 20 30 40 50 60 Energy (eV) Figure 2.37. Comparison between experimental NIEDF (purple) and calculated NIEDF for HOPG in hydrogen (the reconstructed distribution f EXP(E,θ) was used as input). Experimental conditions: 1.0 Pa of H2 plasma, Q = 5.2 sccm, PECR =60W, Vs=–130 V, VMS = 0V, without screen. f EXP(E) emitted from the surface given by reconstruction method is shown by black dots, f EXP‘(E) at the entrance of the MS given by the model is shown by red dots. 63

2. Modeling and reconstruction of NIEDF

In conclusion, one can state that agreement for surface-produced energy and angular distributions f (E,θ) with the predictions of SRIM is quite good and validates the reconstruction method on the example of HOPG for which the initial f SRIM(E,θ) is thought to be reliable. Moreover, excellent agreement between f EXP‘ (E,θ) calculated by the model from reconstructed f EXP(E,θ) and the experimental NIEDF demonstrates the coherence of NIEDF modeling and the reconstruction method. It also shows that to the limit of accuracy of the experiments, f EXP(E,θ) is a good approximation of the NI distribution emitted by the surface. Next step is to test the reconstruction method for other materials for which the initial distribution function f SRIM(E,θ) is a priori unknown. 2.5.3. Gadolinium The material chosen to apply the reconstruction method was Gadolinium (Gd) – a low work function metal purchased from NEYCO Company. Its work function is only 2.9 eV comparing to 4.6 eV for HOPG and it has an enhanced hydrogen storage capacity [83, 84]. The dimensions of the original circular sample were 20 mm × 1 mm (it was cut in four pieces afterwards), and the purity was 99.9 %. The result of the angle-resolved measurements performed on Gd is represented in Figure 2.38. Comparing to HOPG, one can see an increased backscattering contribution to the signal (the tail is much higher than for NIEDF on HOPG). The onset of NIEDF shifts with angle α and the tails superpose as for HOPG. The reconstruction of the surface-produced energy and angular distributions is shown in Figure 2.39 and Figure 2.40 respectively.

0° 11° 22° 33° 10000 Gd 1Pa 60W ECR with  = 1° 12° 23° 34° 2° 13° 24° 35° 3° 14° 25° 4° 15° 26° 5° 16° 27° 6° 17° 28° 1000 7° 18° 29° 8° 19° 30° 9° 20° 31° 10° 21° 32°

100

NI intensity, cts/s

10 0 10 20 30 40 50 60 70 80 Energy (eV)

Figure 2.38. Experimentally measured NIEDF for different tilts of Gd sample: from α = 0° to 35°, with a step of 1°. Rotation direction: anticlockwise. 64

2. Modeling and reconstruction of NIEDF

0.1

6 0.01 SRIM Gd 10 ions H + 88% 44-46 eV 3 H + 12% 69-72 eV 2 Normalizedintensity 1E-3 reconstructed from SRIM Gd reconstructed from exp Gd 1E-4 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 Energy, eV Figure 2.39. NI energy distributions emitted from Gd surface: f SRIM(E) given by SRIM for 88% of H3+ and 12% of H2+ accounting their energy distributions (total ion number 106) shown in black; f REC(E) from SRIM reconstructed by the method shown in red and total f EXP(E) reconstructed from experiment performed in ECR H2 plasma shown in green.

1.0

0.9 0.8

0.7

0.6 0.5

0.4 SRIM Gd

H + 100% 45 eV 107 ions 0.3 3

Normalizedintensity reconstructed

0.2 from SRIM Gd

0.1 reconstructed from exp Gd 0.0 0 10 20 30 40 50 60 70 80 90  emission, deg

Figure 2.40. NI angular distributions emitted from Gd surface: f SRIM(θ) given by SRIM for 100% of PI at 45 eV (only H3+ population, total number 107) shown in black; f REC(θ) from SRIM reconstructed by the method shown in red and total f EXP(θ) reconstructed from experiment performed in ECR H2 plasma shown in green. 65

2. Modeling and reconstruction of NIEDF

For Gd we have no information on hydrogen coverage, hydrogen surface binding energy and other surface parameters. However, the shape of the measured NIEDF suggests that there is much less NI created by sputtering on Gd than on HOPG (unless if

Gd was previously heated in H2 plasma, see Chapter 3. NI production on different materials for details). Indeed, for any material, sputtered particles have a lower average energy than the backscattered ones and should give an enhanced low energy contribution to the NIEDF (a pronounced peak), since low energy NI are more efficiently collected by the MS, as can be seen in Figure 2.9. Therefore, we have used SRIM to compute the backscattered distribution on pure Gd with the surface parameters of Gd given by the software: 25 eV for displacement energy, 3.57 eV for surface binding energy and 3 eV for lattice binding energy. This f SRIM(E) distribution can be compared to the reconstructed NI distribution (green curve in Figure 2.39) only if ionization probability

Piz = const, which is probably not the case for Gd. However, the comparison is still interesting. The result of the SRIM calculation for Gd is shown as black curve in Figure 2.39 and its reconstruction – as red curve. Note that both ion populations H3+ and H2+ with their energy distributions around 45 eV and 70 eV correspondingly were taken into account in the calculation. The agreement between the SRIM calculation and the distribution reconstructed from the experiment is unexpectedly good. The comparison must be done with care, because of the ionization probability Piz issue. However, one can see that the f SRIM(E) distribution for Gd is really different from the one obtained for carbon. Between 0 eV and 35-40 eV the distribution demonstrates an increase and then a decrease, while it is constant for carbon until 35-40 eV. This tendency (increase and then decrease) is observed also in f EXP(E) reconstructed from the experiment which increases our confidence in the reconstruction method. The NIEDF has a maximum at 35 eV, only 10 eV less than the average incident energy of the dominating ion population H3+. This can be explained by poor energy transfer between incident H particles and Gd atoms due to bigger mass difference than for carbon atoms. The absolute values between f EXP(E) and f SRIM(E) do not compare well, especially at low energy. However, one must keep in mind that f SRIM(E) has to be convoluted with the change of Piz, in order to be directly compared to f EXP(E). The result of the reconstruction for the surface-produced angular distributions is shown in Figure 2.40. The reconstructed distribution is slightly shifted compared to that of SRIM for the emission angle range θ ∈ [5°:75°], and then deviates from SRIM for θ > 75°. In any case, f EXP(E) and f SRIM(E) demonstrate similar tendencies despite the fact that the change of Piz was not taken into account.

The contour plot representing f SRIM(E,θ) as predicted by SRIM for Gd, 88% of H3+ and

12% of H2+ accounting their energy distributions for the total ion number 106 is shown in Figure 2.41 and its reconstruction in Figure 2.42. The reconstructed distribution demonstrates some noise for high values of θ, but it can be considered as successful overall, taking into account the reduced number of ions used in the calculation. 66

2. Modeling and reconstruction of NIEDF 90 0.000

80 17.88

70 35.76

60 53.64

deg ,

 71.52 50 89.39 40

107.3

30 118.0

Emissionangle 20

0 10 20 30 40 50 60 Energy, eV

Figure 2.41. Contour plot representing f SRIM(E,θ) as predicted by SRIM for Gd, 88% of H3+ and 12% of H2+ accounting their energy distributions (total ion number 106). The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ. 90 0.000

80 18.33

70 36.67

60 55.00

, deg ,

 73.33 50

91.67 40 110.0 30

121.0 Emissionangle 20

0 10 20 30 40 50 60

Energy, eV

Figure 2.42. Contour plot representing f REC(E,θ) reconstructed from SRIM data for Gd,

88% of H3+ and 12% of H2+ accounting their energy distributions (106 ions). The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ. 67

2. Modeling and reconstruction of NIEDF 90 0.000

80 1.750E+07

70 3.500E+07

60 5.250E+07

deg ,

 7.000E+07 50 8.750E+07 40

1.050E+08 30 1.155E+08

Emissionangle 20

0 10 20 30 40 50 60 Energy, eV Figure 2.43. Contour plot representing f EXP(E,θ) reconstructed from the experiment on Gd in ECR H2 plasma. The color bar indicates the number of NI emitted from the surface having a unique pair of values E and θ.

The distribution f EXP(E,θ) reconstructed from the experiment on Gd in ECR H2 plasma shown in Figure 2.43 demonstrates a good agreement with the SRIM prediction. The differences are observed for small energy (E<10 eV) and for high emission angles θ > 75°. In the latter case, the contribution comes from NI of all energies and could be clearly seen as plateau in Figure 2.40. The origin of this signal is yet unknown, however one can argue that it does not originate from the reconstruction method algorithm, but from the experiment itself. If one compares HOPG and Gd surfaces in terms of NI production efficiency, HOPG is ahead. Figure 2.44 which demonstrates experimentally collected NI yield as a function of the tilt angle α in ECR hydrogen plasma shows that up to α = 15° more NI are collected from HOPG surface. This result could be explained by a bigger sputtering contribution to the NI production on HOPG, as sputtered NI have a smaller average energy comparing to the backscattered ones and therefore are mainly collected at smaller tilt angles. Starting from α = 15°, more NI are collected from Gd since the backscattered NI have a bigger average energy for Gd comparing to HOPG (as could be concluded from momentum transfer). However, it is more important to compare the yield of NI emitted from the surface rather than collected ones. When comparing the intensities in cts/s of f EXP(E) and f EXP(θ) reconstructed from the experimental data for HOPG and Gd, the difference is ~6% in favor of HOPG (see Figure 2.45 and Figure 2.46). This result is good for HOPG and demonstrates that despite its high work function it is an efficient NI surface production material. 68

2. Modeling and reconstruction of NIEDF

1100000

1000000 1Pa 60W 900000 wave HOPG 800000 Gd 700000 600000

500000

400000 NI Yield,a.u. NI 300000 200000

100000

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 Alpha, deg

Figure 2.44. NI yield measured by the MS from HOPG (black) and Gd (red) surfaces as a function of the tilt angle α in ECR hydrogen plasma.

experiment HOPG experiment Gd

1010

cts/s Intensity, 109

0 10 20 30 40 50 60 70 Energy, eV

Figure 2.45. NI energy distributions emitted from the surface f EXP(E), reconstructed from experiment performed in ECR H2 plasma for HOPG (green solid) and Gd (blue dashed) with their relative intensities (in cts/s). 69

2. Modeling and reconstruction of NIEDF

3.5x109 experiment HOPG experiment Gd 3.0x109

2.5x109

2.0x109

1.5x109

cts/s Intensity,

1.0x109

5.0x108

0.0 0 10 20 30 40 50 60 70 80 90  emission, deg

Figure 2.46. NI angular distributions emitted from the surface f EXP(θ), reconstructed from experiment performed in ECR H2 plasma for HOPG (green solid) and Gd (blue dashed) with their relative intensities (in cts/s).

0.1

f(V ) for Gd perp

0.01 SRIM 106 ions

H + 88% 44-46 eV intensity Normalized 3 H + 12% 69-72 eV 2 reconstructed from experiment 1E-3 0.0 20.0k 40.0k 60.0k 80.0k 100.0k 120.0k

Vperp, m/s SRIM + + Figure 2.47. f (υ┴) as predicted by SRIM for Gd with 88% of H3 and 12% of H2 6 EXP accounting their energy distributions for 10 incident ions (black curve) and f (υ┴) reconstructed from the experimental data (red curve).

2. Modeling and reconstruction of NIEDF

However, in Chapter 3. NI production on different materials it will be described that NI production on Gd can become more efficient after performing the heating of the sample under plasma exposure. This creates a hydrogen reservoir inside the material and increases the amount of NI surface production by sputtering. The surface-produced NI distributions as a function of perpendicular velocity of emitted NI are shown in Figure 2.47. Contrary to the case of HOPG, the agreement between the distributions for Gd is less good for small perpendicular velocities. A reduced ionization probability Piz for particles with small υ┴ might explain this disagreement. The comparison between experimental NIEDF and calculated f EXP ’(E,θ) for Gd in H2 plasma in case of normal incidence (α = 0°), for which the reconstructed distribution f EXP(E,θ) was used as input, is shown in Figure 2.48. The excellent agreement proves again the consistency between the NIEDF modeling and the reconstruction method. From a more general point of view, the results on Gd prove that the reconstruction method can be used successfully for any material, even in case when the ionization probability Piz is not constant with E and θ of the emitted particle. Moreover, the reconstruction method does not depend on SRIM, so it can be successfully applied in case when the initial distribution function f SRIM(E,θ) is unknown.

0.1

Gd, ECR 1 Pa H , 60 W 2

V =-130V,  = 0° 0.01 s EXP f (E,) reconstructed EXP  Normalizedintensity f '(E, ) reconstructed experiment Gd 1E-3 0 10 20 30 40 50 60

Energy (eV) Figure 2.48. Comparison between experimental NIEDF (dark yellow) and calculated NIEDF for Gd in hydrogen (the reconstructed distribution f EXP(E,θ) was used as input). Experimental conditions: 1.0 Pa of H2 plasma, Q = 5.2 sccm, PECR= 60W, Vs=–130 V, VMS = 0V, without screen. f EXP(E) emitted from the surface given by reconstruction method is shown by black dots, f EXP‘(E) at the entrance of the MS given by the model is shown by red dots. 71

2. Modeling and reconstruction of NIEDF

2.6. Conclusion The model described in this chapter included a proper description of the diagnostic technique and of the ion transport in plasma and the sheaths which has resulted in better understanding of measured NIEDF. The remarkable agreement of the model with experiment for HOPG confirmed that the NI ionization probability Piz(E,θ) is constant independently of the neutral particle energy and angle of emission. This has also verified the choice of SRIM to provide the correct initial distribution f SRIM(E,θ) for carbon materials, since the input parameters for SRIM calculation on a-C:H layers are well known in the scientific community. The reconstruction method has allowed to determine the distribution in energy and angle of NI emitted from the surface in case when the NI formation probability is not constant with E and θ of the emitted particle or when the SRIM parameters are unknown, so one cannot obtain f (E,θ). The reconstruction method was validated by the good agreement of SRIM calculations on carbon with the distributions reconstructed from experimental data for HOPG on several levels. After the verification, the reconstruction method was used to characterize NI production on the surface of low work-function metals (on the example of Gd) which has given an unexpectedly good agreement with SRIM calculations. The comparison with SRIM calculations, even though it has to be done with much care, reinforces our confidence in the reconstruction method. Moreover, the comparison between experimental NIEDF and f EXP ’(E,θ) calculated from the direct model (for which the reconstructed distribution f EXP(E,θ) was used as input) has yielded an excellent agreement for both HOPG and Gd which demonstrates the coherence of NIEDF modeling and the reconstruction method. One should mention that the reconstruction algorithm does not depend on the NI surface production mechanism. The only input which is necessary to calculate the ion trajectories are the parameters of the plasma and the sheaths. Therefore, the reconstruction method can be applied to any type of surface and/or NI. Another important point that cannot be left out is that HOPG showed a comparable result with Gd in terms of NI surface production despite its high work function. This confirms that carbon materials are interesting to be used as NI enhancers.

3. NI production on different materials

A study performed on a large variety of materials such as different types of graphite, diamond films and metals is presented in this chapter. The measurements on molybdenum (Mo) were performed in order to evaluate the NI background signal, since the sample holder, the clamp and the screws are made of Mo and could give a certain contribution to the measured NIEDF. HOPG has been chosen as a reference material for NI surface production studies due to its unique properties such as easy cleavage, high NI yield, etc. The interest to study other non-diamond carbon materials (CFC and ta-C) arose for the sake of comparison. Gadolinium (Gd) has served as reference for low work function metals. Its work function is only 2.9 eV comparing to 4.6 eV for HOPG and it has an enhanced hydrogen storage capacity [83–84]. Furthermore, it is quite easy to manipulate and does not react quickly at air contrary to Ba, for instance. Tungsten (W) material has proven to be highly resistant to hydrogen plasma and is presently the main candidate as a plasma-facing material (PFM) for the divertor in ITER. In case of efficiency in terms of NI production, it could be easily implemented in the NI source for fusion applications. Diamond is one of materials which are planned to be tested in a real NI source (Cybele) as NI enhancers. It is well known for its ability to emit electrons at high temperature and even at low electric fields [46]. Beam experiments on diamond showed surface production of H− ions with high yields up to 5.5%. Moreover, it has been observed in plasma experiment that NI production yield on boron-doped-diamond can be increased by a factor 5 when increasing the temperature to 400°C [47]. The band gap present in the case of insulators such as diamond hinders the electron capture probability, but the loss of electron by a NI also becomes less probable. If NI is formed on the insulating surface upon a PI impact, it would already be far away from the surface when its electron energy level enters in resonance with unoccupied states in the conduction band. Moreover, hydrogen-terminated diamond shows negative electron affinity (its conduction band lies above the vacuum level) [48 – 49]. Therefore, electrons can be efficiently supplied to the vacuum, since the valence band is located at higher position. All these evidences make diamond an interesting candidate for NI surface production and one of the main candidates to be used in Cybele. Within this chapter, the NI production study of polycrystalline boron-doped and non-doped diamond films was performed. The size of crystals in the studied samples was ranging from ~ 5 μm (microcrystalline diamonds: MCBDD, MCD) to 50 – 100 nm (nanocrystalline diamond: NCD).

3. NI production on different materials

3.1. Molybdenum: background measurements

The sample holder used to install different materials in the diffusion chamber was already presented in Figure 1.2 (b). It is composed of molybdenum (Mo), so it is important to see the contribution of Mo to the measured NI signal. For this reason a piece of Mo similar to the sample clamp used (0.5 mm thick plate) was installed in the sample holder. This has allowed to measure the NI signal as a function of the applied bias voltage which is shown in Figure 3.1. One can conclude that the NI surface production on Mo is negligible comparing to the typical signal levels on the samples under study. The peak intensity of NIEDF in case of Mo varies between 250 cts/s and 1250 cts/s whereas for the studied samples the variation is usually between 20·103 cts/s and 160·103 cts/s. Moreover, a comparison between polished and non- polished Mo issued by different metal providers has been performed. Despite the smaller background signal given by the polished sample, no critical difference between the signal levels of polished and non-polished samples could be pointed out.

1400 Mo clamp I_peak 1200 Mo non-polished sample

I_peak 1000 Mo polished sample 800 I_peak

600

400

Peak intensity, cts/s

200

0 0 -20 -40 -60 -80 -100 -120 -140 -160 -180 Bias voltage, V

Figure 3.1. Intensity of the NI signal on Mo samples: sample clamp material (black), non-polished (red) and polished (green) Mo samples as a function of bias in H2 RF plasma. 74

3. NI production on different materials

The time evolution studies performed by Master student Rouba Moussaoui on the same Mo samples at Vs = -10 V have demonstrated a decrease of NI yield by 5-7 times down to ~ 1000 cts/s within 20 min (see Figure 3.2). The increased amount of NI produced on Mo samples within first 5 min of plasma exposure is probably related to the impurities sitting on the surface which are gradually removed by the plasma. As a conclusion, the evaluation of Mo as a background material has revealed that it gives sufficiently low NI signal levels even for non-polished samples and for samples which contain some impurities on the surface visible by naked eye (sample clamp material).

8000

time evolution at V =-10V 7000 s Mo clamp: Yield

6000 Mo non-polished sample: Yield 5000 Mo polished sample: Yield

4000

3000

arb.u. yield, NI 2000

1000

0 00:00 00:05 00:10 00:15 00:20 Time, min

Figure 3.2. Time evolution of NI yields on Mo samples: sample clamp material (black), non-polished (red) and polished (green) Mo samples; measured at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -10 V, VMS = 0 V, with screen.

3. NI production on different materials

3.2. Carbon materials

3.2.1. Graphitic materials

Due to the unique properties of HOPG (easy cleavage, high NI yield), it has been chosen as a reference material for NI surface production studies. The main parameters influencing NI surface production are surface temperature of the sample, plasma exposure time, bias voltage amplitude (which defines impinging ion energy) and bias duration (see Chapter 5. Pulsed-bias approach). For the studies described in this chapter, the continuous bias was applied during the whole plasma exposure time. The change of the NIEDF with the surface temperature on HOPG measured at usual

RF plasma conditions in D2 is shown in Figure 3.3: color distribution from dark blue to red corresponds to the surface temperature of the sample. One could notice the decrease of the low-energy part of the distribution with temperature. This is partially explained by the decrease of sputtering contribution due to atomic deuterium leaving the surface as temperature rises. The change of the deuterium surface coverage was included in SRIM calculations, and the result of the modeling was compared to the experiment: see Figure 3.4. It can be seen that a constant coverage of 35% can fit more or less measurements at RT and 200°C while a lower coverage of 20% is required to fit the NIEDF at 300°C. For higher temperatures, the modeling does not compare well to the experiment, even if one puts the surface coverage at 0%. This could mean that the reduction of NI signal is related to the decrease of ionization probability Piz(E,θ).

Moreover, the assumption Piz(E,θ)= const might no longer be valid. Another way to represent the change of NI surface production with temperature is the NI yield. The incoming PI flux is considered to be constant and independent of the surface temperature of the sample. The yield variation of HOPG with temperature is shown in black symbols in Figure 3.5 and demonstrates an exponential decay starting from 200°C (note that the y-scale is logarithmic). The interest to study other non- diamond carbon materials arose for the sake of comparison. The temperature evolution measurements were also performed for Carbon Fiber Composite (CFC) [85] and tetrahedral amorphous carbon (ta-C) samples. The CFC sample represented a 3D texture of bundles of fibers (diameter of ~ 7 μm) embedded in a pyrolitic carbon matrix. It was produced by SNECMA Propulsion Solide and denominated Sepcarb® N11. Within the fibers and the matrix, the interplane distance was evaluated to be 3.36 Å which demonstrates that the structure is close to that of graphite, but with different spatial arrangement. The fibers in CFC are composed of carbon polymer chains bound together in ribbons. These ribbons are more or less aligned parallel to the long axis of the fiber. While being based on hexagonal graphitic structure, the ribbons do not have the long range ordering, so the resulting carbon fibers are amorphous. To understand the structure of CFC, one could refer to Figure 3.6. 76

3. NI production on different materials

5 10 Vs = - 130 V RT HOPG 200°C 2 Pa D 20 W 2 300°C

4 400°C 10 500°C 600°C 700°C 800°C 103

cts/s Intensity, 2 10

1 10 0 5 10 15 20 25 30 35 40 45 50 55 Energy, eV

Figure 3.3. NIEDF measured on HOPG as a function of sample surface temperature at 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

1 Vs = - 130 V Experiment HOPG RT 2 Pa D 20 W 200°C 2 300°C Modeling

= 35% 0.1  = 20%

0.01

intensity Normalized

1E-3 0 5 10 15 20 25 30 35 40 45 50 55 Energy (eV)

Figure 3.4. NIEDF measured on HOPG (2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen) at RT, 200°C and 300°C surface temperatures compared to the modeling with different D atom surface coverage. 77

3. NI production on different materials

105

D 2 Pa 20 W 2

NI Yield, arb.u. HOPG

104 CFC ta-C

0 100 200 300 400 500 600 700 800 900 Temperature of plasma exposure, °C Figure 3.5. NI yield dependence on the surface temperature for carbon samples (HOPG, CFC and ta-C) at 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

Figure 3.6. SEM images of a surface CFC tile F27T10 extracted from an erosion zone of the Tore Supra Toroidal Pump Limiter. The B-field direction on the surface is indicated by the arrow [86]. 78

3. NI production on different materials

It shows a Scanning Electron Microscopy (SEM) image of the CFC sample which was extracted from an erosion zone of the Tore Supra tokamak’s limiter and has suffered a high-flux plasma bombardment [86]. The ta-C sample is a hard and dense carbon material with a large fraction of sp3- bonded carbon (this fraction was unfortunately unknown for our sample). The sample was produced using Pulsed Filtered Cathodic Vacuum Arc (PFCVA) deposition process by Manresa company (Spain). The bias applied during the deposition was -400 V with frequency of 100 kHz and the duty cycle of 20%. The deposition time was 2 hours and has resulted in ta-C layer with a thickness of up to 1 μm. The average roughness of the surface was estimated to be ~ 320-360 nm. The behavior of both CFC and ta-C samples with temperature resembles much that of HOPG, as could be seen from Figure 3.5. Raman spectroscopy measurements performed on HOPG prior to and after plasma exposure indicate that heating provides for the elimination of out-of-plane defects (consisting of sp3-bonded carbon) and reconstruction of pristine sp2-phase graphitic surface. These results will be discussed in more detail in Chapter 4. Surface state characterization of NI enhancers. It should be noted that the present results in deuterium are very similar to the ones obtained previously in hydrogen on HOPG [47]. 3.2.2. Diamond materials At the moment, the diamond films at elevated surface temperatures have shown to be the best materials for NI surface production. It has been observed in plasma experiment that NI yield on boron-doped-diamond can be increased by a factor 5 when increasing the temperature to 400°C in H2 plasma [47]. Due to its low reactivity with hydrogen, the chemical erosion rate of diamond by thermal hydrogen atoms is about two orders of magnitude lower than that of graphite [87 – 89], although the physical sputtering yields of diamond and graphite are similar for 1 keV hydrogen particles. These facts could be in favor for the choice of diamond as NI enhancer material for fusion Cs-free NI sources of future generation. Polycrystalline diamond films were deposited at the LSPM laboratory by PECVD in a bell jar reactor (PLASSYS BJS 150) operating with a mixture of H2, CH4 and B2H6 gases for MCBDD and the mixture of H2 and CH4 for MCD. Nanocrystalline Diamond (NCD) films were deposited at the LSPM laboratory using the following gas mixtures: 1% CO2 /

1% CH4 / 98% H2 for NCD 1% and 5% CO2 / 5% CH4 / 90% H2 for NCD 5%. They were characterized at PIIM laboratory by Scanning Electron Microscopy (SEM) prior to plasma exposure. The measurements were performed on ESEM Philips XL 30 device with a resolution of 5 nm in the Microscopy Center of Aix-Marseille University. Figure 3.7 shows SEM images of unexposed surfaces of diamond materials. By looking at the images of microcrystalline diamond films (a and b), one could see crystals of ~ 5 μm in size. The height of these crystals lies in the range of μm, as could be concluded from AFM image of MCD shown in Figure 2.7. On the contrary, the 79

3. NI production on different materials

(a) (b)

Figure 3.7. SEM images of unexposed surfaces of diamond materials: (a) MCBDD, scale 5 μm; (b) MCD, scale 20 μm; (c) NCD 1%, scale 2 μm; (d) NCD 5%, scale 2 μm.

RT 5 Vs = - 130 V 10 MCBDD 200°C 2 Pa D 20 W 400°C 2 600°C

800°C

104

3 10

c/s Intensity,

102

0 5 10 15 20 25 30 35 40 45 50 55 Energy (eV) Figure 3.8. NIEDF measured on MCBDD as a function of sample surface temperature at 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen. 80

3. NI production on different materials

nanocrystalline diamonds (parts c and d) demonstrate a much finer structure. In case of NCD 1%, one could observe small grains ~ 50 nm in size forming a homogeneous arrangement on the top surface. There are also big grains of 300-500 nm in diameter which could produce agglomerations sizing in several μm. NCD 5%, on the other hand, exhibits the grains of a larger size (~ 100-200 nm) which form cauliflower type structures on the top surface. The size of these structures covering completely the surface of the film varies from 0.2 to 0.6 μm. The composition of all the grains is the same, as verified by Energy-dispersive X-ray spectroscopy (EDS). The EDS spectrum shows an intense C peak and a weak Si peak (coming from the substrate), and the peak ratio is constant for all the grains. The evolution of NIEDF with surface temperature was studied on all of the above- mentioned samples. The results of the measurements on MCBDD at usual RF plasma conditions in D2 are shown in Figure 3.8, with color corresponding to the surface temperature of the sample. Contrary to the behavior of graphitic materials, the NI production increases with temperature up to 400-500°C and then starts to decrease. The behavior observed here in deuterium has already been observed in hydrogen [47]. The change of the deuterium surface coverage with temperature was included in SRIM calculations, and the comparison of modeling and experiment is shown in Figure 3.9 and Figure 3.10. The coverage has very slightly increased while heating to 400°C: the tail of NIEDF for 400°C (shown in green in Figure 3.9) is a bit lower than that of NIEDF for RT (shown in blue). One can see that for diamond films the NI yield increases for this temperature range, but the H surface coverage does not. It may be explained by the increase of ionization probability, if one considers the sputtering and backscattering yield to be constant with temperature. We consider the latter approximation to be true within the given temperature range (from RT to 400°C). When reaching 800°C (see Figure 3.10), the modeling predicts the coverage of 20%, which indicates the decrease of H surface coverage for T > 600°C. However, this result could be regarded only qualitatively. The modeling assumes that ionization probability is constant with energy and angle of outgoing particle, which may no longer be true at such high temperatures. The change of NI yield with surface temperature for the samples under study is represented in Figure 3.11. One can see that on all diamond layers NI yield increases starting from 150°C, reaches a maximum at around 500°C and afterwards starts to decrease. It should be mentioned that it is not possible to bias MCD below 300°C due to insulating properties of the modified layer which is created on the surface upon the bombardment by ions and neutral coming from the plasma. Since both the modified layer and the pristine layer underneath are probably insulating [90], it is impossible to bias the sample and produce NI for T < 300°C. The results for graphitic materials are shown in Figure 3.11 for comparison. One can conclude that the differences between the NI yields of materials within the group (graphite or diamond) that could originate from the difference in initial sample morphologies (roughness, grain size and

3. NI production on different materials

Experiment 1 Vs = - 130 V RT MCBDD 400°C 2 Pa D 20 W 2 Modeling = 30% 0.1

0.01

Intensity (arb. (arb. units) Intensity

1E-3 0 5 10 15 20 25 30 35 40 45 50 55

Energy (eV)

Figure 3.9. NIEDF measured on MCBDD in D2 RF plasma at RT and 400°C surface temperatures compared to the modeling with D atom surface coverage of 30%.

1 Vs = - 130 V Experiment 800°C MCBDD 2 Pa D 20 W Modeling 2  = 20%

0.1

0.01

(arb. units) Intensity

1E-3 0 5 10 15 20 25 30 35 40 45 50 55 Energy (eV)

Figure 3.10. NIEDF measured on MCBDD in D2 RF plasma at 800°C surface temperature compared to the modeling with D atom surface coverage of 20%. 82

3. NI production on different materials

orientation) are minor. Therefore, the modified surface state of the exposed sample and its electronic properties must be the key parameters to explain NI yield temperature evolution. In order to get an insight into the surface state evolution of diamond materials, time evolution measurements were performed on MCBDD sample at Vs = -130 V for different surface temperatures as shown in Figure 3.12. The empty symbols represent the NI yield from the first series of measurement on a virgin MCBDD 1 sample in D2 plasma. One can observe a major NI yield decrease during the first 5 minutes (see black empty symbols) probably connected to the degradation of the sample surface. When the surface state is stabilized, the NI yield stays constant. Then, the heating of the sample was performed in plasma (duration ~ 1 min) at surface bias Vs = -130 V still applied. The red empty symbols show the increase of NI yield after the heating to 400°C which also happens within 5 minutes. The NI production becomes more efficient and the yield risesa bit below the initial level. Based on previous studies [57] and the present results, it can be assumed that the surface state has partially recovered from the defects inducedby the plasma exposure at RT. The etching of the defects is enhanced by the increase of temperature [91]. Therefore, the surface state, in particular the number of defects, can be seen as resulting from a balance between the creation of defects by the ion bombardment and their destruction through ion-assisted chemical etching.

6 10

5 10 D 2 Pa 20 W 2 HOPG MCBDD MCD NI Yield, arb.u. CFC NCD 1% 104 NCD 5% ta-C

0 100 200 300 400 500 600 700 800 900

Temperature of plasma exposure, °C Figure 3.11. NI yield dependence on the surface temperature for diamond films (MCBDD, MCD, NCD 1% and NCD 5%) and graphitic samples (HOPG, CFC and ta-C) at 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen. 83

3. NI production on different materials

In order to check if heating to 400°C reconstructs the surface completely, a second series of measurements on MCBDD 2 was carried out (see solid symbols in Figure 3.12). The sample was immersed in the plasma being already heated to 400°C, and the decrease of the NI yield was still observed. However, the yield went down by a factor of 2 instead of factor of 3 seen previously, and the decrease was happening within ~ 20 min. In such a way, heating of MCBDD has hindered the creation of defects by the plasma exposure and has kept the surface in a state which is more favorable for NI production. The red solid symbols show the continuation of experiment with MCBDD 2 after letting it cool down overnight in vacuum. One can notice that the heated surface presents initially the same NI yield as an unexposed sample, showing that the surface state at 400°C is close to an undisturbed surface state. For the second half of the experiment (see solid red and green symbols in Figure 3.12), MCBDD 2 has acted as the sample MCBDD 1. Again, the heating to 400°C (solid green symbols) was performed in plasma with applied surface bias. The differences in the NI yields between MCBDD 1 and MCBDD 2 in the same conditions (15% of difference at the end of the experiments) are thought to be due to the experimental uncertainty comprising different durations of plasma exposure, uncertainty on the surface temperature, alignment of the normal to the sample surface with respect to the mass spectrometry axis, etc.

5 8.0x10 400°C MCBDD 1 MCBDD 2 7.5x105 virgin RT virgin 400°C 7.0x105 RT heating to 400°C Cooled overnight 6.5x105 RT RT

6.0x105 heating to 400°C 5 5.5x10 5.0x105 400°C 5 4.5x10 400°C 400°C 4.0x105 3.5x105 5

Yield,arb.u. NI 3.0x10 2.5x105 RT 5 RT 2.0x10 400°C 1.5x105 5 400°C 1.0x10 5.0x104 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 Time, min

Figure 3.12. Time evolution of NI yield on MCBDD at different surface temperatures. Empty symbols correspond to MCBDD 1 sample and solid symbols to MCBDD 2. The color of symbols goes with the chronological order of the experiments: black, red, green. The surface temperature of the sample is indicated next to the curves with a corresponding color. 84

3. NI production on different materials

3.3. Metals

3.3.1. Gadolinium As explained in Chapter 2. Modeling and reconstruction of NIEDF, Gadolinium (Gd) was chosen as a model of a low work-function metal. The measurements on Gd surface were performed analogously to the ones on carbon samples with the only difference being hydrogen plasma instead of deuterium. The evolution of NI production with surface temperature (see Figure 3.13) has been measured. One can notice that the NI signal stays constant up to the surface temperature of 200°C and then decreases, similar to graphitic materials. The comparison of NI yield evolution with temperature in

H2 plasma for HOPG, MCBDD, MCD and Gd is shown in Figure 3.14. The behavior for carbon materials is very similar to that in deuterium. Concerning Gd (see green symbols), the NI yield at RT is ≥ 2 times smaller than in case of HOPG due to less important sputtering contribution. On the other hand, the yield decrease with surface temperature for Gd is much less pronounced than for HOPG: at 800°C Gd gives 3 times more NI. However, one should bear in mind that the comparison of NI yield evolution with temperature at α = 0° is only relevant, if the angular emissions for these materials are similar. This is the case for carbons, but the angular emission for metals is different. In order to make a correct comparison, reconstruction of the NI energy distribution on the surface from measurements at various tilt angles α is necessary (as presented in Chapter 2. Modeling and reconstruction of NIEDF). When comparing the NI yields of the distributions emitted from the surface which were reconstructed from the experimental data for HOPG and Gd, the difference is 6% in favor of HOPG. Therefore, the reconstruction should be done at various surface temperatures, so that the proper comparison between NI yields on carbons and metals as a function of temperature can be made. However, these tedious measurements and the reconstruction procedure have been performed so far only at RT. After performing the heating cycle, it was noticed that the peak intensities of the

NIEDF at Vs = -130 V have increased showing increase of the NI creation probably due to higher hydrogen surface coverage. One can imagine that during high temperature measurements, hydrogen species diffuse inside the material creating a reservoir. When coming back to room temperature, the hydrogen coverage is increased due to this reservoir of hydrogen species which is then being gradually exhausted. The time evolution of NIEDF at Vs = -130 V is shown in Figure 3.15 which demonstrates that stable state is reached after approximately 1 hour. The study of NIEDF as a function of applied surface bias was performed on previously unexposed Gd sample and on the same sample after performing a heating cycle, once the stable state has been reached (see Figure 3.16). The peak intensities of the NIEDF have been increased for all biases which could confirm the increase of hydrogen surface coverage. A bigger quantity of H on the surface causes a bigger sputtering contribution 85

3. NI production on different materials

10000

Gd H 2 Pa 20 W 2

1000

c/s Intensity, 100

RT 500°C

200°C 600°C 300°C 700°C 400°C 800°C 10 0 10 20 30 40 50 60 Energy, eV

Figure 3.13. NIEDF measured on Gd as a function of sample surface temperature at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen.

2 Pa 20W H 6 2 10 V = - 130 V s  = 0°

105

NI Yield, arb.u. Yield, NI HOPG MCBDD MCD 104 Gd

0 100 200 300 400 500 600 700 800 900 Temperature, °C Figure 3.14. NI yield dependence on the surface temperature for HOPG, MCBDD, MCD and Gd samples at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen. 86

3. NI production on different materials

1.4x105 Gd H 2 Pa 20 W 2 after T cycle: loaded with H 5 1.4x10

1.3x105

5 1.3x10

NI Yield, arb.u. 1.2x105

5 1.1x10

0 10 20 30 40 50 60 Time, min

Figure 3.15. NI yield time evolution for Gd sample at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen, after performing the heating cycle.

(a) Before heating cycle (b) After heating cycle 4 Gd H2 2 Pa 20 W 4 Gd H2 2 Pa 20 W 10 -20 V 10 Vs= Vs= -20 V -60 V -60 V -100 V -100 V -130 V -130 V -170 V -170 V

103 103

Intensity,cts/s Intensity,cts/s 10 2 102

101 101 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Energy, eV Energy, eV

Figure 3.16. NIEDF for different bias voltages Vs applied in increasing order on Gd surface at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, with screen, at RT before (a) and after (b) the heating cycle.

3. NI production on different materials

(the main peak grows). The increase of the maximum energy in NIEDF tail (compare the measurements at Vs = –20 V and –60 V in Figure 3.16 (a) and (b)) and also as the main peak increase for NIEDF at Vs =–20 V are not easy to understand, since the incmoing energy for the dominant ion (Ei = ~13 eV/nucleon) is probably below the sputtering threshhold in this case. Note, that the measurements used for reconstruction of the NI energy and angular distributions on the surface of Gd presented in Chapter 2. Modeling and reconstruction of NIEDF were performed on previously unexposed Gd sample. Therefore, the H surface coverage must have been low, so the NI creation by sputtering was not as significant as on Gd sample after the heating cycle. Aiming to remove hydrogen from the sample, the sample was cleaned in Ar plasma during ~ 10 min under the following conditions: 1 Pa, PRF = 50 W, Vs = – 300 V, Vpl = 21 V;

Ei = ~ 320 eV/nucleon, without screen, sample facing up to maximize the incoming particle flux. After that, the NI measurements were performed in Ar plasma at Vs = – 130V (giving Ei = ~ 150 eV/nucleon), but in order to decrease the RF-fluctuations the input power was brought down to PRF = 4 W and the grounded screen was put back in place. The results are shown in Figure 3.17 in red, compared to the NIEDF measured after the heating cycle in H2 plasma at Vs = – 130 V (giving Ei = ~ 45 eV/nucleon) shown in black. Contrary to the NIEDF in H2 plasma which contains both sputtering and backscattering contributions to the measured NI signal, the NIEDF measured in Ar plasma is originated only from sputtered H atoms. However, the incoming particle energy Ei and mass mi are not the same in these conditions, so no direct comparison can be made. In order to follow the evolution of sputtered NI with temperature, measurements in Ar plasma have been performed (as shown in Figure 3.17 and Figure 3.18). NIEDF shown by bold lines were measured while heating the sample, and the ones shown by thin lines – while cooling down. As can be observed from the figure, the contribution of sputtered NI decreases starting from 200°C, but stays important up to 400°C. For higher temperatures the NIEDF maximum is ~ 200 cts and can originate from H impurity present in Ar plasma. The cooling of the sample has resulted in a partial recovery of the signal (~ 2 times less than the original signal level). This could mean that hydrogen which has been diffused deeper inside the material by heating, comes back to the surface. One can conclude that the reservoir of H inside Gd, once it has been created, could not be eliminated completely neither by energetic Ar+ ion bombardment nor by ion bombardment accompanied by heating. Some further investigations with ex situ surface diagnostics are needed to explain better this phenomenon. Note that measurements in Ar plasma on carbon surfaces have shown some NI signal, but for a very short time, since the hydrogenation of the surface was eliminated in the absence of hydrogen reservoir. The phenomenon of hydrogen trapping by Gd surface has been described by various authors. It was seen that hydrogen adsorption on gadolinium is a dissociative chemisorption process that occurs at the surface. Hydrogen atoms tend to form 1 × 1 88

3. NI production on different materials

104 Gd after T cycle: loaded with H H2 2 Pa 20 W: BS+SP Ar 1 Pa 4 W: only SP

103

10 Intensity, cts/s Intensity,

101 0 10 20 30 40 50 60

Energy, eV Figure 3.17. NIEDF measured at RT with screen on Gd surface after loading it with H2. Black curve: right after the heating cycle, at 2.0 Pa of H2 plasma, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, VMS = 0 V, Ei = ~ 45 eV/nucleon. Red curve: at 1.0 Pa of Ar plasma, PRF = 4 W, Vs = -130 V, VMS = 0 V, Ei = ~ 150 eV/nucleon.

4 Bold - heating; thin - cooling down Gd Ar 1 Pa 4 W 10 RT 200°C 300°C 400°C 500°C 3 600°C 10 700°C 800°C

2 c/s Intensity, 10

101 0 5 10 15 20

Energy (eV) Figure 3.18. NIEDF measured on Gd surface after loading it with H2. Conditions: 1.0 Pa of Ar plasma, PRF = 4 W, Vs = -130 V, VMS = 0 V, Ei = ~150 eV/nucleon, with screen, as a function of sample surface temperature. Bold lines: heating, thin lines: cooling down. 89

3. NI production on different materials

islands and affect the Gd (0001) surface in a "local" fashion [83]. In another study, thin Gd films were deposited under UHV conditions on glass and converted ‘‘in situ’’ at RT into GdHx (0 < x < 3) under H2 pressure of the order of 100 Pa. It was found that in the process of GdHx formation, light-reflecting metallic thin Gd film is transferred into transparent trihydride while its electrical resistance increases by several orders of magnitude. At low coverage, positively polarized hydrogen adspecies arise, penetrating quickly into the bulk. When the average H/Gd concentration approaches 1, negatively charged hydrogen adspecies appear directly on the surface, slowly penetrating into the bulk [84]. Comparing to our study, one could suggest that the NI yield increase observed on Gd surface after H adsorption (enhanced by the heating cycle) could be connected to the formation of GdHx complexes and H– on the sample surface and in the bulk.

3.3.2. Heated tungsten

Tungsten (symbol W in the periodic table) has proven to be highly resistant to hydrogen plasma. It is presently the main candidate as a plasma-facing material (PFM) for the divertor in the International Thermonuclear Experimental Reactor (ITER), primarily due to its high melting temperature, high energy threshold for physical sputtering, low physical sputtering yield and resistance to chemical sputtering [92 – 94]. By heating a W filament, the emission of electrons, governed by the Richardson law (thermal electron emission) is increased. It was decided to study the influence of this effect on NI creation on the surface of the W filament. The test of NI surface production was performed on W filament from OSRAM halogen lamp HLX 64655-250W-24V. The filament could be biased and heated at the same time. The area of the filament facing the MS was comparable to the area of a usual flat sample (53 mm2 for the filament vs. 50 mm2 for flat samples). The filament was positioned perpendicularly to the MS, so the experimental conditions were comparable to the usual ones. The actual photo of the tungsten filament is represented in Figure 3.19 and the electric circuit used in the experiment in Figure 3.20. The measurements have been performed in D2 plasma at usual operating conditions: 20W, 2Pa, grounded screen touching the MS, d = 37 mm distance between sample and the MS. The sample was biased with Vs = -130 V with increasing amount of the heating current applied. The temperature of the filament was estimated from the resistivity vs. temperature dependence for tungsten [95] shown in Figure 3.21 by black symbols. The resistivity was calculated with the following formula: V heat  A R  A I    heat Eq. (3.1) L L d where A   ( )2 is the filament cross-section and L is its total length. 2

3. NI production on different materials

Figure 3.19. Photo of the W filament mounted on a support, with two wires (covered by insulating layer) dedicated to bias the filament and pass the heating current through the circuit.

1÷10 Ω Wsample R A R

V Vheat Iheat Z plasma

−

Vs +

Figure 3.20. Electric circuit used in the experiment with heated W filament.

3. NI production on different materials 130 120 W resistivity vs T [72] 110 W resistivity measurement

100 W resistivity measurement

90 (including correction term)

2 80

cm 70

 , ,

 50 40 30 20 10 0 500 1000 1500 2000 2500 3000 3500 4000 Temperature, K

Figure 3.21. Dependence of the resistivity of W filament on its temperature as given by theory [95] (black symbols); measurement of the resistivity during the present experiment fitted to the theoretical dependence without any correction (green symbols) and including the loss term in the calculations (red symbols). The green symbols represent the values of filament temperature from this experiment fitted to the theoretical dependence [95]. It was not possible to make a measurement of resistivity at 300 K, because driving the minimum measurable current

Iheat through the circuit would heat the filament to unknown temperature. The first measurement was made at weak orange glow of the filament and corresponded to the temperature of 1650 K according to the formula given earlier (see green symbols in Figure 3.21). To a common knowledge, the glow of tungsten filament starts at around 1000 K. By subtracting a loss of ~ 0.25Ω (that probably originated from imperfect electric contact) from the resistance value in the formula, the resistivity values were corrected (red symbols in Figure 3.21). Figure 3.22 presents the NIEDF for heated tungsten filament: 300 K (just biasing the filament), 1060 K (weak orange glow), 1507 K (bright orange glow), 1896 K (bright yellow glow), 2270 K (very bright yellow glow), 2622 K and above (bright white glow). In Figure 3.22 (a) one can see that at 300 K there are only surface produced NI in a very small quantity (at around 0 eV which means they were produced with zero initial kinetic energy at the filament surface). When heating the filament, their amount somewhat increases, but they are accompanied by NI produced at the MS nozzle (see Figure 3.22 (b)). These NI appear at –100 eV when the raw data are processed in a usual way (i.e. as if the NI were produced on the sample surface, see Section 1.3.1. Representation of spectra for details). At 2622 K there were no more W surface-produced NI (see Figure 3.22 (c)). From Figure 3.22 (d) it can be seen that the quantity of nozzle-produced NI

3. NI production on different materials

increases with filament temperature. The total NI yield which originates from the NI created at the filament surface is presented in Figure 3.23 as a function of filament temperature. The yield of surface-produced NI reaches a plateau at 1000÷2000 K and then decreases. The whole NIEDF of NI produced on MS nozzle could not be measured completely

with Vext = 110V, since it was outside of the scanning range. When the extractor potential

Vext was changed from 110 V to 70 V (always keeping Vref = - Vext to hold the total MS potential at 0 V) the distribution shows the usual form: main peak + tail (see black and red curves in Figure 3.23). With increasing the RF input power from 20 W to 300 W the quantity of nozzle-produced NI was enlarged (see red and green curves in Figure 3.23). Note that the NIEDF appears at negative energy with a tail extending towards positive energies only because the MS raw data were treated as though the NI were coming from the sample surface. When data are processed considering that the NI come from the MS

nozzle biased at (Vext + Vref), the NIEDF distribution rises at 0 eV with a tail extending towards positive energies.

30 (a) NI at the W filament surface 30 (b) NI at the W filament surface Tfilament, K Tfilament, K 300 300 25 1060 25 NI at the MS nozzle 1060 2270 2270 20 20

2622 2622

15 15

Intensity, cts/s

Intensity, cts/s 10 10

5 5

0 0 -120 -100 -80 -60 -40 -20 0 20 40 60 80 -120 -100 -80 -60 -40 -20 0 20 40 60 80 Energy, eV Energy, eV

NI at the MS nozzle 30 (c) NI at the W filament surface T , K 1000 (d) filament V = 110 V NI at the MS nozzle extractor 300 T , K 25 NI at the MS nozzle 1060 filament 800 2622 2270 3150 20 2622 3200 600

3232

15 3345 400

Intensity, cts/s 10 Intensity, cts/s

200 5

0 0 -120 -100 -80 -60 -40 -20 0 20 40 60 80 -110 -100 -90 -80 -70 -60 -50 -40 Energy, eV Energy, eV

Figure 3.22. NIEDF measured when heating the W filament: (a) RT (300 K); (b) from 300 K to 2270 K; (c) from 300 K to 2622 K; (d) from 2622 K to 3345 K. The origin of the NI (either the W filament surface or MS nozzle) is indicated. 93

3. NI production on different materials 150

NI at the W filament surface

125

100

NI Yield, arb.u.

0 0 500 1000 1500 2000 2500 Temperature, K Figure 3.23. NI Yields measured on W filament as a function of heating temperature. The measured NI originate from the filament surface. Experimental conditions: 2.0 Pa of D2 plasma, Q = 7.6 sccm, PRF = 20 W, Vs = -130 V, with screen. 3000 2750 NI at the MS nozzle 2500 T = 3207 K V = 2250 extractor 110 V, 20 W 2000 70 V, 20 W 1750 70 V, 300 W

1500

1250 1000

Intensity, cts/s 750

500

250 0 -110 -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0 Energy, eV Figure 3.24. NIEDF measured on W filament heated to 3207 K with different values of extractor potential Vext and RF input power (see the legend). NI originate from the MS nozzle. Experimental conditions: 2.0 Pa of D2 plasma, Q = 7.6 sccm, Vs = -130 V, with screen. 94

3. NI production on different materials

For T > 2500 K there are only NI produced at the MS nozzle, their quantity increases with filament temperature. The reason for this increase is not known. The MS nozzle was heated during the experiment due to the simultaneous electron and photon bombardment. The increase of the temperature might have influenced the NI surface production through cleaning of the stainless steel surface, for instance. The NI yields obtained on the W filament cannot be compared to the yields obtained on carbon surfaces, since the extraction of the NI from the MS nozzle is completely different from the extraction of sample produced NI. Controlled experiment with a stainless steel sample would be necessary to achieve a better understanding of these results. The main conclusion of this study is that we were not able to observe any enhancement of NI production on a W surface when the temperature was increased from 300K to more than 3000K. The maximum of surface NI production lay at around 1000÷2000 K and was very low compared to NI created on carbon materials. Despite the particular shape of the sample (it was not a perfectly planar surface for instance), these results are not encouraging and seem to indicate that increasing the number of emitted electrons from the surface is not enough to promote NI surface production.

3.3. Conclusion The evaluation of Mo as a background material has revealed that it gives sufficiently low NI signal levels even for non-polished samples and after having suffered numerous plasma exposures (sample clamp material). A study presented in this chapter was performed on a large variety of carbon materials. The non-diamond carbon materials under study were Highly Oriented Pyrolitic Graphite (HOPG), carbon fiber composite (CFC) and tetrahedral amorphous carbon (ta-C). The diamond materials were microcrystalline boron-doped diamond (MCBDD), microcrystalline non-doped diamond (MCD), nanocrystalline diamonds (NCD 1% and NCD 5%). The influence of the surface temperature during the plasma exposure and exposure time on H-/D- yields was investigated. The behavior of all non-diamond carbon materials with surface temperature is very similar and demonstrates an exponential decay starting from 200°C. The change in deuterium surface coverage cannot explain alone the decrease of the NI signal with temperature which means that there must be a change in the ionization probability involved. At the moment, the heated diamond films have shown to be the best materials for NI surface production. Prior to plasma exposure, the diamond films were characterized by Scanning Electron Microscopy (SEM) to estimate their surface characteristics. Contrary to the behavior of graphitic materials, the NI production on diamond films increases with surface temperature up to 400-500°C and then starts to decrease. The signal increase by almost a factor 5 previously seen in hydrogen was observed also in

3. NI production on different materials

deuterium. The NI yield increase from RT to 400°C is not related to the change in deuterium surface coverage, as it stays constant according to the modeling. It may be explained by the increase of the ionization probability, if one considers the sputtering and backscattering yield to be constant with temperature. One can also conclude that the differences between the NI yields of materials within the group (graphite or diamond) that could originate from the difference in initial sample morphologies (roughness, grain size and orientation) are minor. Therefore, the modified surface state of the exposed sample and its electronic properties must be the key parameters to explain NI yield temperature evolution. In order to get an insight into the surface state of diamond materials during the plasma exposure, time evolution measurements were performed on MCBDD sample at Vs = -130 V (Ei = ~45 eV/nucleon) for different surface temperatures. It was observed that the NI Yield at RT has been decreased by a factor of 3 with time, which may be explained by the creation of defects on the surface. On the other hand, heating of MCBDD to 400°C has increased the NI Yield almost to the initial level (measured on non-modified MCBDD surface at RT). We suggest that heating of the sample hinders the creation of defects by the plasma exposure and keeps the surface in the state favorable for NI production. As it has been previously observed by [91], etching of the sp2 surface defects is enhanced by the increase of temperature. Therefore, the surface state, in particular the number of defects, can be seen as resulting from a balance between the creation of defects by the ion bombardment and their destruction through ion-assisted chemical etching. The measurements on Gd surface were performed analogously to the ones on carbon samples. It was possible to characterize the evolution of NI production with surface temperature. NI signal on Gd stays constant up to the surface temperature of 200°C and then decreases, similar to graphitic materials. The NI yield at RT is nearly 2 times smaller than in case of HOPG due to less important sputtering contribution. On the other hand, the yield decrease with surface temperature for Gd is much less pronounced than for HOPG. However, the reconstruction of the NI energy distribution on the surface should be done at various surface temperatures, so that the proper comparison between NI yields on carbons and metals as a function of temperature can be performed. The evolution of NI signal with surface bias shows that increase of bias voltage results in the increase of both main peak of NIEDF and the tail. After performing the heating cycle, the bias evolution measurements were repeated. It was noticed that the peak intensities of the NIEDF at all bias values have increased, showing increase of the NI creation probably due to higher hydrogen surface coverage. One can imagine that during high temperature measurements, hydrogen species diffuse inside the material creating a reservoir. In order to follow the evolution of sputtered NI with temperature, measurements in Ar plasma have been performed. One can conclude that the reservoir of H inside Gd, once it has been created, could not be eliminated completely neither by energetic Ar+ ion bombardment nor by ion bombardment accompanied by heating. Comparing our study to previous investigations, one could suggest that the NI yield increase observed on Gd 96

3. NI production on different materials

surface after H adsorption (enhanced by the heating cycle) could be connected to the formation of GdHx complexes and H– on the sample surface and in the bulk. The main conclusion of the study of NI production on heated tungsten filament is that it demonstrates a very low NI Yield compared to carbon materials. One has not observe any enhancement of NI production on a W surface when the temperature was increased from 300 K to more than 3000 K. The maximum of surface NI production has been achieved at around 1000÷2000 K. These results are not encouraging and seem to indicate that increasing the number of emitted electrons from the surface is not enough to promote NI surface production.

4. Surface state characterization of NI enhancers

The results presented in previous chapters have proven the efficiency of the experimental arrangement to study NI production on surfaces of different materials. One can obtain information about the energy and angular distributions of emitted NI, about the dependence of NI yield on the surface temperature, plasma exposure time, etc. However, the sample surface state under various plasma exposure conditions was not well defined. When a sample is immersed into the plasma at a negative bias, it interacts with the incoming PI and the neutral species that are present in the plasma chamber. The first several nanometers of the surface become disordered and represent an arrangement different from the pristine surface. The degree of disorder can depend, for instance, on the rate of the plasma-assisted etching which, in its turn, depends on the sample surface temperature and on plasma conditions. All these effects have an influence on the NI surface production efficiency. Therefore, it appeared necessary to characterize the change of the surface state after the plasma exposure. As carbon materials have shown the biggest potential in the NI surface production, it was decided to study HOPG and diamond films with ex situ surface analysis techniques such as Raman spectroscopy and temperature programmed desorption. The goal was to explain the variation of the NI yield as a function of the surface temperature and obtain an idea about the surface state of the samples exposed to plasma at different temperatures. These investigations are described in the present chapter. The introduction of the chosen surface diagnostics is given in the first place, then the experimental protocol is specified and the obtained results are discussed.

4. Surface state characterization of NI enhancers

4.1. Raman spectroscopy

4.1.1. Basics of Raman spectroscopy

The phenomenon of inelastic light emission has been studied by the Indian scientist Sir C.V. Raman in 1928. He has noticed that the radiation emitted by the molecules which are being irradiated by the photons contains the photons of the same frequency as the incident radiation, but also the photons of different frequencies. This is a very weak effect: approximately 1 photon for a million (0.0001%) is emitted with a wavelength slightly different from the incident wavelength. This process was later associated with the name of its discoverer, and the photon frequency change was called “Raman effect”. In the end of 1930s, the method became widely used by scientific community. Raman spectroscopy is a non-destructive technique for structural and chemical analysis of materials based on the study of their vibrational and/or rotational- vibrational properties. It is considered as supplementary to infrared (IR) spectroscopy (which deals with the IR light absorbance by molecules at specific frequencies) even though being based on a different physical origin. With the help of Raman spectroscopy it is possible, for instance, to prove the coexistence of several phases (amorphous/crystalline) of a solid sample. The given technique can be used for all types of samples: solid, liquid or gaseous. The list of domains where Raman spectroscopy is employed could be non-exhaustive: research (characterization and study of the fundamental mechanisms relied to the properties of nanotubes, etc.), medicine (detection of the breast cancer cells [96]), analysis of the cultural heritage objects [97], scientific police (drug identification, structural and chemical analysis of the Martian soil in the nearest future [98]). In the same way as IR absorption spectroscopy, the Raman spectroscopy makes use of dipolar electric interaction with the matter. Yet, one deals with the two-photon mechanism which combines absorption and emission. If a molecule is in periodic vibrational motion (with frequency νvib) and possesses a dipole moment in equilibrium, this molecule could irradiate at the frequency of its motion. Likewise, a molecule which does not possess a dipole moment at equilibrium is able to acquire an “induced dipole moment” under the influence of a periodic light radiation. This molecule could thereby absorb energy corresponding to the same frequency.  Being more precise: in Raman diffusion, the electric field E of the monochromatic  radiation (with frequency ν0) induces in the studied molecule a dipole moment p , oscillating at frequency ν0. The dipole moment is given by:   p  []E Eq. (4.1) where [α] is the polarizability tensor.

4. Surface state characterization of NI enhancers

  The electric field oscillates in time: E  E0 cos( 2 0t) Eq. (4.2) On the other hand, the polarizability tensor is as well modulated in time, because the molecule is animated by a motion with characteristic vibration at frequency νvib:

[]  []0 [']cos( 2 vibt) Eq. (4.3) where [α]0 is the polarizability of the molecule at rest and [α’] is the amplitude of the variable part. Therefore, the induced moment becomes doubly modulated, as one could see after performing the substitution and trigonometrical transformations:    1 1  p  E0[]0 cos(20t)  [']cos(2[ 0  vib ]t)  [']cos(2[ 0  vib ]t)  2 2  Eq. (4.4) Thus, taking into account the induced dipole moment modulation naturally reveals the frequencies of Rayleigh (ν0), Anti-Stokes (νvib+ν0) and Stokes (νvib–ν0). These frequencies represent the three possible light diffusion mechanisms, shown in Figure

4.1 with En and Em being the energies of two stationary states of the molecule. The virtual state, to which the molecule is temporarily excited during the light scattering, is designated as a dashed line in the scheme. It does not correspond to a real energy level of the molecule.

h0

h0 h0

E n hvib Em

Stokes Rayleigh Anti-Stokes

Figure 4.1. Representation of different light diffusion processes. n and m stand for stationary states of a molecule, whereas the virtual state is designated as a dashed line.

100

4. Surface state characterization of NI enhancers

In the IR spectroscopy the variation of the dipole moment during the vibration imposes the presence of a ray associated with a transition in the spectrum. Contrary to that, in case of Raman spectroscopy the variation of polarizability tensor during the vibration is responsible for a transition activity. In such a way, these two techniques could be complementary to each other.

4.1.2. Raman experimental set-up

Raman spectra were recorded using a Horiba-Jobin-Yvon HR800 LabRAM apparatus in the backscattering geometry, the external appearance of which is shown in Figure

4.2. The source of monochromatic light (typically green laser with λL=514.5) is focalized on the surface of any sample with the help of a microscope. The components of the sample interact with the light which results in the creation of photons characterized by the wavelength different from that of the laser. The signal of interest is superimposed with the Rayleigh diffusion signal which is considered as inconveniency since it is much more intense than the Raman diffusion signal. Therefore, the inelastically scattered photons are collected by the microscope objective (typically ×100 for a better collection) to be sent to a grating after passing through the notch filter. The notch filter serves to diminish the undesirable Rayleigh diffusion signal. At the exit of the grating the Raman diffusion signal is dispersed and arrives to a CCD camera which records the spectrum.

There is a possibility of switching between green, red (λL=633.8 nm) and UV

(λL=325 nm) lasers for multi-wavelength Raman spectroscopy. One could also choose between several objectives such as ×10, ×35 (UV only), ×50 and ×100. The laser power was kept below ~1mW/μm2 to prevent damages. Spectra were recorded with various exposure times to check that samples do not evolve under laser irradiation. In addition, it is possible to heat a sample in the gas cell (equipped with a glass window) under a constant Ar flow. In such a way, evolution of the Raman signal with temperature could be followed. With the help of experimental set-up, it is possible to visualize a sample zone which could be defined by using a little camera coupled to the microscope (see Figure 4.3). A shutter allows for switching between “spectrum recording” mode and “sample visualization” mode. The latter mode relies on the use of a white light source placed on the optical table and connected to the microscope with an optical fiber. Therefore, it is possible to locate the laser spot in the right place on the sample surface (which is especially useful for cartographic Raman measurements) and focus the laser source to maximize the signal intensity. Both spectrum recording and sample visualization are realized via the software “LabSpec 5”.

101

4. Surface state characterization of NI enhancers

LASER 325 nm

LASER 514 nm

Grating trap

Entry trap

White light of microscope Platen CCD sensor Optical fiber Shutter Raman/Microscopy Microscope optique Figure 4.2. External view of the Horiba-Jobin-Yvon HR800 LabRAM apparatus for Raman spectroscopy placed on the optical table, as seen by the user. Raman micro-spectrometer

LASER Source (514.5 nm) Grating

M2 M1 M5

NF CH FW

M3 M4

Microscope optique

CCD sensor ~1 m

~ 1μm

CH: Confocal Hole FW: Filter Wheel NF: Notch Filter M: Mirror

Figure 4.3. Schematic representation of the experimental set-up used for Raman spectroscopy including optics, sample holder and hardware components. 102

4. Surface state characterization of NI enhancers

4.1.3. Raman micro-spectrometer calibration

Before performing any measurements, one should verify that the set-up is well calibrated. The goal of the calibration of the grating is to associate a correct wavelength value to a given pixel of the CCD camera. It is achieved by positioning the grating correctly with respect to the detector. A simple test with a a Si sample may be performed to evaluate the necessity of the calibration. The position of the principal Si Raman band should be located at ~520 cm-1 as predicted by theory. If the experimentally registered peak significantly deviates from the predicted position, the grating must be calibrated. The definitive calibration is done by centering the Rayleigh diffusion signal around the theoretically predicted position of the laser wavelength λL = 514.532 nm. This procedure is carried out by adjusting two parameters which define the grating position: zero and koef. Another important procedure to perform is the temperature calibration of the gas cell which is used for sample heating. Raman spectroscopy can be used to monitor the local temperature on the Si sample surface, so that it can be associated with the temperature in the gas cell (reading of the temperature controller). The changes in the Raman band at 520 cm−1 with temperature have been measured many times with a high accuracy [99]. Balkanski et al. [100] have shown that, taking into account cubic and quartic anharmonic terms, the silicon band position could be written as:

2 3 3  (T)  0  C(1 )  D(1  ) Eq. (4.5) ex 1 e y 1 (e y 1)2 with x  hc  0 / 2 k B T , y  hc  0 / 3 k B T , and  0 , C and D being constants. Good fits −1 with experimental data were obtained between 0 and 1200 K with  0 = 528 cm , C = −2.96 cm−1, and D = −0.174 cm−1 [101]. The Raman spectra measured on the Si sample placed in the gas cell are shown in Figure 4.4 for various temperatures from RT to 950°C with a step of 100°C. The gradual shift of the Si band towards the shorter wavelengths and the Raman signal decrease with the temperature rising can be observed. The variations in the signal intensity (visible for T ≥ 450°C) could be connected to slight defocusing of the laser. Figure 4.5 demonstrates the position of the Si Raman peak as a function of gas cell temperature (black points) fitted by the equation from [101] shown in red. The exact peak position was determined by performing a Gaussian fit of the band for each temperature, and the error bars originate from the uncertainty of the position estimation given by the fit. Generally, one can conclude that the gas cell temperature read on the controller corresponds well to the actual temperature on the Si sample surface within the chosen temperature range. It is supposed that the calibration should hold for any other sample studied under the same experimental conditions.

103

4. Surface state characterization of NI enhancers

1200 Raman peak of Si RT 1000 150°C

250°C 800 350°C 450°C 600

550°C

650°C 400 750°C

Intensity, cts/s Intensity, 850°C 200 950°C

400 420 440 460 480 500 520 540 560 580 600

-1 Raman shift, cm Figure 4.4. Raman spectra measured on the Si sample placed in the gas cell for various temperatures from RT to 950°C with a step of 100°C.

525

Measured in cell 520 predicted by Helwig et al

-1 515

510

505

500

Si bandposition, cm

495

0 100 200 300 400 500 600 700 800 900 1000 Temperature, °C Figure 4.5. The position of the Si Raman peak as a function of the gas cell temperature (black points, reading on T controller) fitted by the equation from [101] shown in red. 104

4. Surface state characterization of NI enhancers

In the present thesis, the Raman spectroscopy measurements as a function of temperature were performed as follows. The sample was heated up to a chosen temperature with a linear ramp β = 1 K/s set by the controller. Then the heating was stopped and the sample was cooled down to RT under a constant Ar flow. It was supposed that the state of the sample did not evolve significantly during the cooling phase and was characteristic of the surface at a given heating temperature. In such a way, the measurements could be performed for a longer time providing a bigger signal- to-noise ratio. At each temperature, the laser was refocused in order to keep a comparable signal level throughout the whole measurement series.

4.1.4. Raman spectroscopy of carbons

Raman spectroscopy is commonly used to characterize the carbon-based materials such as graphene, graphite, nanocrystalline graphite (nc-G), amorphous carbon (a-C, a-C:H, ta-C, ta-C:H, etc.). In the case of graphite several atomic vibration modes exist, but only two of them are Raman-active:

1) Atomic vibrations perpendicular to the planes of graphene. Raman diffusion due to this mode (ν = 50 cm-1) is very close to Rayleigh diffusion and is not of interest in the present analysis. 2) Stretching of aromatic cycles and aliphatic bonds of C=C pairs parallel to the planes of graphene which is called “stretching mode” (see Figure 4.7a).

A typical Raman spectrum on different carbon surfaces in shown in Figure 4.6 on the example of (a) graphite, (b) nanocrystalline graphite nc-G, (c) amorphous carbon a-C and (d) diamond. The G band situated at around 1600 cm-1 is the principal band for the Raman spectrum of graphite and is due to the above-mentioned stretching mode (see Figure 4.7a). For a prefect graphitic structure (without defects) it has a narrow full width at half-maximum (FWHM) and its position is well-defined, as can be seen in Figure 4.6a. For the less structured carbon (like nc-G) other bands appear in the spectrum. The D band situated at 1350 cm-1 is Raman active in the presence of defects created in the graphite plane (aromatic domains become of finite size) [102]. This band is called the breathing mode of aromatic cycles (see Figure 4.7b). Sometimes the D’ band situated around 1620 cm-1, near the G band, is also associated with it. Both bands D and G result from carbon sp2 bonds (see below). Please note that in general case the origin of G band is the bond-stretching of all pairs of sp2 atoms, in both rings and chains [103]. Therefore, its FWHM and position are modified in the presence of C=C chains, as can be seen in Figure 4.6c.

105

4. Surface state characterization of NI enhancers

(a)

(b)

(c)

(d)

Figure 4.6. Typical Raman scattering spectra of different carbon surfaces: (a) non- modified graphite; (b) nano-crystalline graphite; (c) amorphous carbon; (d) diamond.

Figure 4.7. Carbon motions in the (a) G and (b) D modes. Note that the G mode is just due to the relative motion of sp2 carbon atoms and can be found in chains as well [102].

106

4. Surface state characterization of NI enhancers

sp2 is a type of orbital hybridization typical for graphitic materials. It is characterized by a double C=C bond (one of which is σ-bond and the other one is π-bond) and the angle of approximately 120° between the orbitals lying within the same plane (as shown in Table 4.1 [104]). Another type of orbital hybridization is sp3 which is characteristic for diamonds and tetrahedral amorphous carbons (ta-C) and represents four orbitals located in different planes in tetrahedral geometry, with an angle of ~109.5° between each other, connected by σ-bonds (see Table 4.1). For amorphous materials like ta-C [102] the D-band becomes less noticeable in the Raman spectrum due to its broadening and intensity decrease. If one compares the spectra of nc-G and amorphous carbons, the difference is quite remarkable [103]: the D and G-bands are much broader for amorphous materials. In order to characterize the degree of surface amorphization, such parameter as the coherence length of aromatic domains La is introduced (shown as red dashed circles in Figure 4.6). Basically, it represents an average distance between two defects in the aromatic plane. La gives an indication of the material crystallinity and is related to the intensity ratio of D and G-bands, which are in turn connected to the presence of defects in the aromatic planes:

C() La  ID IG Eq. (4.6) where C(λ) is a parameter depending on the excitation wavelength (C = 4.4 nm for

λL = 514 nm) [102]. For a virgin HOPG surface La is approaching infinity (no defects in the material) whereas for nc-G La is confined between 2 nm and 1 μm and becomes < 2 nm for amorphous carbons [58].

Table 4.1. 3 simplest types of orbital hybridization with the corresponding sketches [104]. 107

4. Surface state characterization of NI enhancers

4.1.5. Raman spectroscopy of carbons exposed to plasma

When a sample is inserted in H2/D2 plasma under negative bias in PHISIS, it is subjected to the flux of energetic positive ions (with energy depending on the applied surface voltage Vs and on the ion type) and neutral particles (with energy ~ 26 meV as corresponding to room temperature). As was already demonstrated by time evolution measurements in the previous chapters, the surface state of the samples plays an important role in NI surface production. The first several nanometers of the sample surface are modified upon plasma bombardment and do not represent the initial arrangement anymore. The surface state strongly depends on the plasma exposure conditions, so it was decided to characterize the samples by Raman spectroscopy after exposure in the typical conditions used for NI surface production studies. The conditions of plasma exposure for these experiments were as follows: low injected RF power of

20 W, at 2.0 Pa of H2/D2, with a grounded screen placed horizontally approximately 5 cm above the sample to suppress the RF fluctuations of the plasma potential. The plasma density and temperature measured by the Langmuir probe inserted in the diffusion chamber were: ne = 2·1013 m-3 and Te = 3.5 eV. At this pressure the plasma is mainly populated with H3+/D3+ ions (~90%). The negative bias of the sample was set at –130 V resulting in the incoming positive ion energy of 45 eV per nucleon (plasma potential was

~5 V). The ion flux was approximately 3·1012 (H3+/D3+)/(cm2·s) and the exposure time was 30 minutes.

The Raman spectroscopy measurements performed on HOPG sample in H2 plasma at RT shown in Figure 4.8 demonstrate a typical nanocrystalline graphite (nc-G) structure rather than an amorphous carbon situation [47]. The nc-G represents a structure with separate graphitic clusters which are immersed into the amorphous matrix. The bands that constitute the HOPG RT spectrum are the following: GHOPG (G band of the underlying graphite); GIL (G band from the interaction layer “IL”); D2-IL and D’2-IL (defect-related bands from the IL); D3-IL (sp3 defect-related band from the IL); D4-IL (amorphous carbon band) – see Figure 4.8(a) with a fit. The bands GIL, D2-IL and D’2-IL are attributed to in- plane defects in the graphite structure, which imply the finite size of graphitic domains; bands D3-IL and D4-IL – to out-of-plane defects, as described by Pardanaud et al [85]. The exact origin of D3 band is the vibration of the C=C double bond whose frequency is modified by the presence of sp3-bonded carbon which can happen in the area between the graphitic clusters (matrix). For D2-IL band this is the breathing of aromatic cycles which becomes Raman-active in the presence of defects on the graphitic cluster boundaries (in-plane defects), as already discussed previously. The orientation of defects is shown by different colors in Figure 4.8 (blue for in-plane and orange for out- of-plane defects). The graphite G-band (seen from undisturbed under-layer) resulting from the stretching mode of C=C double bond is shown by green color. The sum of the individual bands (total fit) is shown as a red dotted curve.

108

4. Surface state characterization of NI enhancers

1.0 HOPG H2 RT (a) experiment 300 K out-of-plane defects 0.8 in-plane defects graphite G-band

total fit spectrum

0.6

0.4

GHOPG

Normalized intensity Normalized D2-IL 0.2

G D'2-IL D4-IL IL D 3-IL 0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1

1.0 HOPG H2 400°C (b) experiment 300 K out-of-plane defects 0.8 in-plane defects graphite G-band

total fit spectrum

0.6

0.4

GHOPG

Normalized intensity Normalized GIL D'2-IL 0.2 D 2-IL

D3-IL

0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1

Figure 4.8. Raman spectra of HOPG exposed to H2 plasma (2 Pa, 20 W, Vs = –130V) at different temperatures: (a) RT; (b) 400°C. Experimental data is shown in black solid with the fit of individual bands (see caption) and their sum (total fit shown by red dotted curve). The wavelength of the laser λL = 514 nm. Spectra are normalized by the maximum value. 109

4. Surface state characterization of NI enhancers

The Raman spectrum for HOPG exposed to H2 plasma at surface temperature of 400°C is plotted in Figure 4.8(b). The total integral of the bands related to in-plane defects has decreased by nearly 2 times as compared to HOPG RT spectrum (mostly due to GIL diminishing). The integral of the out-of-plane defect-related bands was reduced by 3 times: D4-IL disappeared completely and D3-IL was strongly reduced (please note that the G-band of the underlying graphite is normalized to one in both figures). This shows the re-establishment of sp2 phases and increase of the sp2 cluster size. The coherence length calculated using Eq. (4.6) for ID2-IL/IG-IL, gives La ≈ 4 nm for HOPG RT and La ≈ 5 nm for

HOPG 400°C. However, referring to the dependence of La on the full width at half- maximum (FWHM) of the GIL band [105], the passage from 60 to 50 cm-1 for HOPG RT and 400°C accordingly, corresponds to the increase of the coherence length from La ≈ 5 nm to 8 nm.

The position of GIL-band for the given excitation wavelength of 514 nm corresponds to the position characteristic for nc-G [105]. However, the presence of the bands in the spectra related to the out-of-plane defects suggests a certain amount of sp3 content in the material. Unfortunately, the exact content of sp3 phase in the material cannot be estimated with the help of Raman spectroscopy for two main reasons. First of all, this technique is more sensitive to sp2 phase than to sp3 phase by several orders of magnitude. Secondly, the out-of-plane defect-related bands in our HOPG spectra result from the modification of sp2 bond vibrations due to the presence of sp3 phase in the material, but not directly to sp3 bond vibrations. The content of hydrogen cannot be either estimated by Raman spectroscopy, as only carbon-carbon bonds can be probed and not C-H bonds. According to the three-stage model described in [105], as we consider structural changes starting from ordered graphite, following to nanocrystalline graphite, following to amorphous carbon and finally to sp3 bonded ta-C, the sp2 groups become first smaller, then topologically disordered, and finally change from ring to chain configurations. This process is called amorphization trajectory, consisting of 3 stages from graphite to ta-C:

(1) graphite → nc-G; (2) nc-G → sp2 a-C (with ~20% of sp3 content); (3) a-C → ta-C (with up to 85% of sp3 content).

The presence of sp3-related bands in our Raman spectra suggests that HOPG RT sample exposed in PHISIS was undergoing the second stage of amorphization. When comparing our spectra to the spectra of HOPG subjected to low energy (90 eV) Ar+ bombardment by an ion gun [106], one reveals many similarities in case of high ion doses (1013 and 1014 Ar+/cm²). The effect of the sample heating to 400°C in our case was similar to the reduction of the ion dose from 1014 to 1013 Ar+/cm² in [106] which is in agreement with the removal of the out-of-plane defects and increase of the coherence length. The Raman spectroscopy measurements for HOPG at 800°C (not shown here), 110

4. Surface state characterization of NI enhancers

demonstrate that only G band attributed to sp2 bonded carbon is visible [47]. This result could be related to reconstruction of graphitic sp2 phases and almost complete removal of all types of defects. In case of MCBDD, exposure to plasma also results in creation of defects on the surface, as can be observed in Figure 4.9. The Raman spectra of the virgin MCBDD sample shown by the black curve demonstrates a peak at 1323 cm−1 attributed to diamond with an asymmetric profile at ~1350 cm−1 (Fano profile of zone centre optical phonons [107, 108]). Let us note that this vibration is due to sp3 carbon atoms and can be seen, despite the low sp3 Raman cross section, because of the very high sp3 phase content in the material. After plasma exposure at RT (blue curve), appearance of a broad band in the interval from 1450 cm−1 to 1700 cm−1 (similar to bombarded HOPG) was seen. It can be explained by the presence of sp2 C=C bonds [103, 109] giving rise to the

GIL band, and also the presence of the out-of-plane defects resulting in D3-IL band. The disappearance of the Fano profile at ~1350 cm−1 suggested the presence of an underlying

D2-IL band. These surface modifications indicate a loss of crystallinity and creation of defects accompanied by sp3 to sp2 conversion [110, 111]. Exposure to plasma at high temperatures has resulted in decrease of the broad band

(GIL + D3-IL) centered at 1600 cm−1 in the Raman spectra: for 400°C the decrease of the band intensity maximum is almost factor 2 and for 800°C – factor 5 as compared to exposure at RT. The re-appearance of the Fano profile at ~1350 cm−1 indicates the decrease of the underlying D2-IL-band. These facts suggest that removal of defects by plasma assisted etching becomes more efficient at high temperatures and the diamond surface is partially restored. Indeed, except for the presence of the GIL band at 1600 cm−1, the spectra of MCBDD 800°C and virgin MCBDD are very similar. The partial recovery of the pristine diamond state together with sufficient hydrogen coverage may contribute to the increase of the NI yield for diamond films heated to 400°C by nearly 5 times as compared to room temperature [47, 58].

111

4. Surface state characterization of NI enhancers

MCBDD boron virgin exposed to plasma at RT 400°C 800°C 34°C

400°C

800°C

Intensity, arb.u. Intensity, virgin

1000 1200 1400 1600 1800 -1 Raman shift (cm )

Figure 4.9. Raman spectra of MCBDD exposed to H2 RF plasma (2 Pa, 20 W, Vs = –130V) at room temperature (blue curve), 400°C (red), 800°C (green) and non-exposed (black). The wavelength of the laser λL = 514 nm. Spectra are shifted vertically for better visibility.

112

4. Surface state characterization of NI enhancers

4.2. Temperature programmed desorption

4.2.1. TPD principle and experimental conditions

In order to understand the nature of the NI yield evolution with temperature which is related to the surface state change, the thermodesorption method was involved as an additional surface analysis tool. This technique is called temperature programmed desorption (TPD); it is based on the measurement of the molecular desorption flux from a surface while its temperature is increased. The temperature should be increased linearly with a certain heating rate β by means of a dedicated oven or a filament. When heating the surface, the energy transferred to the species adsorbed on the sample surface will force them to desorb. The desorption rate from unit surface area can be described by the following equation: n E / RT a N(t)  d / dt  n  e Eq. (4.7) where n is the order of the desorption reaction, σ is the surface coverage

(molecules/cm2), vn is the rate constant, and Ea the activation energy of desorption (cal/mole), R is the universal gas constant, and T is the temperature [112]. The desorbed species are then detected by the MS according to their masses. A TPD spectrum shows the number of desorbed species of a certain mass as a function of desorption temperature. Thus TPD can provide information on the binding energy, the mass of the species and their quantity (by integrating the spectra). The scheme illustrating the principle of TPD and the assembly used in the present thesis is shown in Figure 4.10. The thermodesorption measurements were carried out in the UHV chamber with background pressure ~5·10-10 mbar, equipped with a differentially pumped mass spectrometer system and oven for heating the sample [113], shown in Figure 4.10. During a TPD experiment it is crucial to discriminate molecules originating from the heated sample and molecules coming from the heating assembly. For this purpose, a differentially pumped quadrupole mass spectrometer (QMS) system connected to the sample chamber by a circular tube with a diameter of 2 mm has been used. In such configuration, the collection of molecules from the sample surroundings is minimized [113]. During the TPD measurements the platen holding the sample was placed on top of the oven regulated with a PID controller (Eurotherm) and heated linearly from 300 K to 1370 K with a heating rate β = 1 K/s. Because the samples were transferred under vacuum, the sample temperature could not be measured directly. Instead, the temperature on the oven in the vicinity of the sample assembly was measured with a K- type thermocouple. To obtain a linear temperature ramp on the sample, a preliminary calibration was performed and a non-linear temperature ramp was programmed on the oven [113]. The uncertainty in the peak temperature estimation was around 30 K within

113

4. Surface state characterization of NI enhancers

the same experimental campaign. However, the difference up to 50 K was seen between the spectra from two separate campaigns. The reasons for that is the slight position change of the oven thermocouple due to numerous sample transfers and the difference in thermal contact between the sample and the sample holder. By chance, we had a possibility to perform two plasma exposures (of HOPG and MCBDD) at RT in the plasma reactor connected to the same assembly as the ultra high vacuum TPD chamber. These will be denoted as “in situ measurements”. The plasma reactor is composed of a RF source and a diffusion chamber. RF power (13.56 MHz), delivered by the CESAR 1320 RF generator from “Advanced Energy”, is applied to the antenna to create plasma in the chamber. Impedance matching is achieved by a matching network. The values of the electron density were measured by the Langmuir probe. At low RF power the plasma is in a capacitive mode, the electron density is around 1014 m-3. As the RF power increases, there is a sudden change of the plasma brightness, indicating the transition from a capacitive mode to an inductive mode with the electron density of about 1015 m-3. A moveable sample holder is installed in the diffusion chamber, the distance between the RF antenna and the sample holder was ~14 cm. The working pressure in the chamber is controlled by a capacitive gauge; the base pressure is typically ~10-8 mbar.

Figure 4.10. Scheme illustrating the principle of thermodesorption and the assembly used in the TPD experiment. 114

4. Surface state characterization of NI enhancers

Before performing the plasma exposure, samples were mounted on a transfer platen made of molybdenum (Mo, purity >99.9 wt. %, ultrasonically cleaned in acetone and ethanol). The platen is drilled with a 9 mm diameter hole in its center and the sample sits on the hole. It was done to improve the thermal transfer between the oven and the sample. The exposure was carried out under the following conditions: 2.0 Pa, D2 plasma in capacitive mode, PRF = 60 W, Vs = -130 V which is similar to the conditions in PHISIS.

The ion flux was approximately 5·1013 D3+/(cm2·s) (as measured by the LP in similar conditions) and the exposure time was 30 min. Right after the exposure, the sample was transferred under vacuum (p ~ 10-9 mbar) into the TPD chamber to perform the desorption experiment. Most of the TPD measurements were performed on the samples exposed to plasma in PHISIS set-up and then transferred in air to the TPD chamber (“ex situ measurements”). A set of temperatures and materials has been defined for TPD measurements as given in Table 4.2. The conditions of plasma exposure for these experiments were already detailed in Section 4.1.5. Raman spectroscopy of carbons exposed to plasma. Plasma exposure time for each sample was 30 minutes, except for several samples which were prepared with the exposure of 5 seconds (to see the deuterium retention in the surfaces with a small amount of modifications). Right after the plasma exposure, the samples were taken out in air and mounted on the Mo transfer platen with a help of a mask and four screws (also made of Mo) which was immediately introduced into the TPD chamber. The typical duration of the transfer for samples exposed to plasma at RT was around 20 minutes. If the plasma exposure was performed at high temperatures, the sample was cooled down under vacuum in PHISIS. This could take from 45 to 60 minutes depending on the sample temperature during the plasma exposure. It should be mentioned that contrary to the ex situ plasma exposure (for which only the sample was exposed to plasma), during the in situ exposure the Mo transfer platen, sample clamp and the screws have been also in contact with the plasma. The duration of the sample transfer after the in situ plasma exposure was around 20 minutes as well.

Exposure T / material HOPG MCBDD MCD 35°C 30 min / 5 sec 30 min / 5 sec – 400°C 30 min 30 min 30 min / 5 sec 750°C 30 min 30 min 30 min

Table 4.2. The list of all samples studied by thermodesorption technique with the characteristics of preceding plasma exposure: temperature and duration.

115

4. Surface state characterization of NI enhancers

4.2.2. Mass spectrometer calibration

The estimation of the global particle transmission (through the entrance hole and the mass spectrometer) was performed in order to obtain the absolute amount of desorbed species from the studied samples. The MS calibration was done by introducing a gas in the TPD chamber and measuring the MS signal as a function of the pressure in the chamber. This procedure was performed for H2, D2, He, CH4, Ne, N2 and Ar gases. Gas pressure was measured by Granville-Phillips Stabil-Ion Gauge and corrected using corresponding sensitivity factors. The MS signal [cts/s] versus pressure [Torr] dependence for each gas was fitted by a linear curve with statistical weighting. The slope extracted from each calibration curve was then divided by the cross section of electron-impact ionization of the most probable reaction taking place inside the MS. The cross sections were taken for the electron energy of 50 eV from [114–117] and NIST database [118] resulting in Figure 4.11. The error bars in the figure originate from the uncertainty in estimation of electron-impact ionization cross-section for a given reaction and from the imperfection of the fit. The allometric fit of the points has been performed yielding 1 m dependence on ion mass. This is an indication that the MS operates in the flux-dependent regime (signal is proportional to the entrance particle flux) rather than in pressure-dependent regime (signal is proportional to the average pressure in the MS). Indeed, the average flux of particles in the UHV chamber in a chosen direction is obtained by integrating the Maxwell-Boltzmann distribution, and gives as the result:

1 1 P 8kT P 2 1 MB  nv   [m s ] Eq. (4.8) 4 4 kT m 2mkT where n is particle density obtained from the ideal gas law and v is the average thermal velocity of particles. Therefore, the particle flux entering the MS in our experiment reads as follows: 2 P 1 Q    r [s ] Eq. (4.9) 2mkT with r being the radius of the MS entrance hole. In such a way, one can observe the proportionality of the particle flux Q entering the MS (and hence the measured signal intensity) to the mass of incoming particle as . which implies low collisionality and flux-dependent regime. In order to account for aging of the MS detector, the calibration procedure should be performed each time before the series of TPD measurements. However, the full calibration is tedious to perform, so it is done only for deuterium. In such a way, the MS sensitivity factor is taken as the slope of the D2 calibration curve.

116

4. Surface state characterization of NI enhancers 30 H Global transmission 2 Allometric fit ~m- 0.5 25

mbarn) * 20

torr He

* D 15 2

, counts/(s , CH 4  10 Ne N 2

Slope/ Ar 5

0 5 10 15 20 25 30 35 40

m/z, a.m.u. Figure 4.11. Global transmission (entrance hole + MS) as a function of ion mass fitted by allometric function. The error bars originate from the uncertainty in estimation of electron-impact ionization cross-section for a given reaction and from the imperfection of the fit. 7

6 Radius of 4.5 mm 4.0 mm 5

Detected from total flux (%)

1 0 1 2 3 4 5 6 7 8

Distance (mm)

Figure 4.12. The result of the Monte-Carlo simulations with 106 particles as a function of the distance between the sample and the MS head from 0 to 8 mm for two chosen radii: 4 mm and 4.5 mm. 117

4. Surface state characterization of NI enhancers

The absolute desorption rate Γdes for a compound X with the mass different from that of D2, is given by the formula: counts I[ ] N atoms at  [ ]  s des 2 Eq. (4.10) s  m counts  M D F[ ] A[m2 ] G  X 2 particles  D2 M X

where I is the measured signal, Nat is the number of atoms in the molecule, F is the MS sensitivity factor for D2, A = 0.503 · 10-4 m² is the area of the sample surface with a radius of 4 mm facing the MS, G is the geometrical factor stating the percentage of the desorbed flux detected by the MS, σ is the electron-impact ionization cross section for 50 eV and M is the mass to charge ratio in a.m.u. The geometrical factor G is estimated by a Monte-Carlo simulation with 106 particles created with random position and momentum on the exposed surface of the sample of radius R. The emission angle of a particle is defined by the cosine law [113]. Figure 4.12 shows the result of these simulations as a function of the distance between the sample and the MS head from 0 to 8 mm for two chosen radii: 4 mm and 4.5 mm. As could be seen for the graph, for our experimental conditions (radius of 4 mm and distance of 2 mm) the geometrical factor G = 5%. Having accounted for all the possible effects, one could obtain the real amount of desorbed components from which D2 and CDx are of major interest. However, the signal of CD3 coincides with that of H2O which largely dominates in the spectra. The signals of

CD2 and CD coincide with CH4 and CH2 correspondingly whereas hydrocarbons are present in all TPD spectra (including those of non-exposed samples) as an impurity. For instance, elevated amounts of desorbed CH4 have been observed for all the samples. The heavier components like C2D4 and C2D6 were almost undistinguishable. Therefore, the only hydrocarbon signal that could be followed without any restrictions is that of CD4 at

20 a.m.u., as it coincides only with D2O which should not be present. For all these reasons and also in order to minimize the influence of the sample transfer in air, D2 and not H2 was chosen for the plasma exposure.

4.2.3. In situ measurements on HOPG

The result of the in situ measurement on HOPG at RT in absolute units at/(s·m²) calculated by Eq. (4.10) is shown in Figure 4.13. The black solid curve in the graph shows several characteristic D2 desorption peaks. Low temperature peaks are located at 350 K and 410 K; high temperature peaks are seen at 515 K, 580 K (with a shoulder at

630 K), and 810 K. The D2 desorption continues up to 1280 K with maximas at ~1010 K and ~1200 K. CD4 (shown by red solid curve) demonstrates similar desorption peaks as

D2 at low temperature (shoulder between 340 K and 400 K, and a peak at 420 K), as well

118

4. Surface state characterization of NI enhancers

as the high temperature peaks at 515 K and 580 K (with a shoulder at 630 K). The inset in Figure 4.13 shows the zoom for better visibility of the high temperature D2 desorption peaks on HOPG. The blue dashed curve represents the D2 desorption signal (multiplied by 10) from a molybdenum platen bombarded by an ion gun in the TPD chamber. It should be noted that for the latter experiment the full Mo platen was used instead of the standard transfer platen with a 9 mm diameter hole employed for all other measurements. The bombardment was done by D2+ ions with a smallest possible energy of 130 eV which gave 65 eV per deuteron, under the angle of 45° with respect to the sample surface. The ion gun was collimated to minimize the bombardment outside of a circle with a radius of 4 mm. The ion flux was 1.25 · 1011 D2+/(cm2·s) which has resulted in the total ion dose of 6.38 · 1014 D2+/cm2 after 85 minutes of bombardment. In such a way, the influence of the background signal coming from the Mo transfer platen, clamp and the screws that were exposed to plasma together with the sample was studied.

However, only the peak position and not the absolute amount of D2 desorption on Mo could be considered for two reasons:

1) conditions of plasma exposure and ion gun bombardment were different in terms of ion dose, particle type, impinging angle, etc.

17 8x10 HOPG RT in-situ signal from D 17 2 7x10 signal from CD 4 6x1017 Mo holder (signal from D ) x 10 2 5x1017 HOPG RT in-situ signal from D 17 1.5x1017 2 signal from CD 4x10 4 Mo holder 17 signal from D *10 17 1.0x10 2 3x10 D flux, at/s*m² 16

D flux, at/s*m² flux, D 5.0x10 2x1017 0.0 17 400 600 800 1000 1200 1x10 Temperature, K

0 300 400 500 600 700 800 900 1000 1100 1200 1300

Temperature, K

Figure 4.13. TPD spectra of D2 (black solid curve) and CD4 (red solid curve) measured in situ on HOPG exposed to plasma in situ at RT (60 W, 2 Pa, Vs = – 130 V). The D2 spectrum of Mo platen bombarded by the ion gun (130 eV/D2+, magnified by 10 times) is shown as blue dashed curve for comparison. 119

4. Surface state characterization of NI enhancers

2) TPD assembly is designed to minimize the collection from outside the sample area (circle with r = 4 mm). In plasma experiment only the outside area is made of Mo. In ion gun experiment all the area is made of Mo.

One can notice that Mo presents a D2 desorption peak at 440 K (see Figure 4.13) very close to the HOPG peak at 410 K. The retained deuterium dose in Mo platen after the ion gun bombardment was ~5% according to the calculations. Thus we conclude that the D2 desorption from the surrounding area made of Mo could give a non-negligible contribution to the D2 signal at 410 K during the in situ measurements on HOPG.

However, it cannot explain the appearance of the intense D2 desorption peak at 350 K and of CD4 desorption peaks, since there was no CD4 desorption seen on Mo. Looking at HOPG spectra in Figure 4.13, one can notice that low temperature peaks at 350 K and 410 K dominate the spectra, whereas they are negligible in ex situ spectra

(as will be shown later). The desorption peak at 350 K in case of D2 is accompanied by the release of impurities such as H2 and H2O; the shoulder between 340 K and 400 K seen for CD4 coincides with the desorption of HD. Desorption peak at 410-420 K is also seen for H2, H2O, CO and HD. Such a low temperature of desorption (corresponding to 50 – 150°C), big content of stored deuterium and simultaneous desorption of impurities suggest that these peaks are not characteristic of the sample surface state during the plasma exposure. It is possible that the products of chemical erosion during the plasma exposure (such as D2 and CD4) were trapped in gaseous form inside the interaction layer and blocked by the impurities which covered the surface after the end of the exposure. Moreover, these desorption peaks have not been observed by any other authors during the TPD measurement on HOPG which reinforces our previous conclusions. Additional in situ measurements are necessary to understand the exact origin of the low temperature peaks at 350 K and 410 K. Therefore, we will not discuss these peaks in the following sections, but include them in the discussion about the total deuterium content in the plasma-exposed materials (see Section 4.2.8. Comparison of total desorption amount).

4.2.4. Ex situ measurements on HOPG

The comparison of D2 desorption in situ and ex situ measured on HOPG at RT is shown in Figure 4.14 in absolute units (at/s·m²). One can notice that the low temperature peaks at 340 K and 410 K from in situ spectrum are largely reduced by exposing the sample to the ambient atmosphere. Analogously to the in situ measurements, in the ex situ spectrum these peaks are accompanied by the release of impurities such as H2, H2O, CO,

CO2, CH4 and C which were also present in the spectrum of non-exposed HOPG sample (not shown here). These impurities could adsorb on the surface during the sample transfer. The reduction of D2 desorption as compared to the in situ spectrum can be explained mostly by the contact of the sample with ambient atmosphere since the 120

4. Surface state characterization of NI enhancers

waiting time was nearly the same as for in situ measurements. The peak 515 K from the in situ spectrum has become a shoulder in the ex situ spectrum; the shoulder at 630 K has been dramatically decreased. On the other hand, the peaks at 580 K, 810 K, 1010 K and 1200 K have not been influenced considerably by the exposure to air.

The spectra of D2 and CD4 desorption from HOPG performed at different surface temperatures during the ex situ plasma exposure are shown in Figure 4.15 in absolute units (at/s·m²). Figure 4.15a demonstrates the spectrum of HOPG RT, with the black solid curve showing D2 desorption and the red dashed curve – CD4 desorption. Figure 4.15b shows the spectrum of HOPG 400°C and Figure 4.15c the spectrum of HOPG 800°C. The spectra from the Figure 4.15a and b were measured during the first experimental campaign (denoted later as “series 1”), whereas the spectrum from Figure

4.15c was measured during the second experimental campaign (“series 2”). The CD4 signal from the series 1 and 2 was measured with a reduced dwell time (leading to a low signal-to-noise ratio), so the adjacent averaging with 10 points has been applied. The baseline was removed for all spectra presented in Figure 4.15. The plasma exposure of HOPG at 400°C reveals that the amount of released deuterium in D2 form does not decrease due to heating (similar area below the peaks, see later). The characteristic D2 desorption peaks are located at 820 K, 1000 K and ~1200 K (see black solid curve in Figure 4.15b). The first peak of HOPG 400°C coincides

8x1017

HOPG RT 17 7x10 D desorption 2 6x1017 in situ

17 ex situ 5x10

17 4x10 3x1017

D flux, at/s*m² 2x1017

17 1x10

300 400 500 600 700 800 900 1000 1100 1200 1300 Temperature, K

Figure 4.14. TPD spectra of D2 measured on HOPG exposed to plasma at RT in situ (black solid curve, 60 W, 2 Pa, Vs = – 130 V) and ex situ (blue dashed curve, 20 W, 2 Pa, Vs = – 130 V). 121

4. Surface state characterization of NI enhancers

with the peak of HOPG RT at ~820 K and represents the main desorption site. The peaks at 1000 K and ~1200 K have considerably grown as compared to their intensity in the spectrum of HOPG RT. CD4 (see red dashed curve in Figure 4.15b) is desorbed at the same temperature as the main D2 peak: 820 K. The amount of desorbed CD4 has been significantly reduced as compared to HOPG RT. Note that there is no desorption visible up to ~700 K (400°C) which corresponds to the sample surface temperature during the plasma exposure. The black solid curve in Figure 4.15c shows the spectrum of HOPG 800°C with only one peak at 1150 K which within the experimental uncertainty corresponds to the peak at 1200 K for HOPG RT and HOPG 400°C. Since all the other adsorption sites are eliminated by heating, this is the only one left at such an elevated temperature. The third experimental campaign was conducted in order to check the reliability of measured TPD spectra. These results are presented in Figure 4.16: the first two spectra (a and b) from the series 1 were replaced by the spectra from the series 3; the last spectrum (c) was kept for comparison. For better visibility, the spectra between the series 1 (blue dashed curve) and 3 (black solid curve) were compared for HOPG RT in Figure 4.17 and for HOPG 400°C in Figure 4.18. The temperature ramp programmed on the controller and the thermal contact were slightly different between the two series, leading to the difference in desorption peak positions, especially for lower temperatures. The main desorption peak of HOPG RT has moved from 650 K in series 1 to 580 K in series 3. Desorption peaks at 800 K and 1050 K are more pronounced in the spectrum from series 3, but the difference in D2 content is not dramatic: only ~ 21% more D2 desorption for series 3. If considering the total deuterium content in the sample (D2 +

+ CD4), the difference increases to 30% in favor of series 3. For HOPG 400°C series 1 has

~20% more particles released in D2 form than series 3, but calculating the total deuterium content in the sample (D2 + CD4) gives only 13% in favor of series 1. The spectra of HOPG 400°C differ considerably between these series; this should be discussed in more detail. The peaks at 1000 K – 1050 K and 1150 K – 1200 K between the two series correspond to the same positions within experimental uncertainty and have similar shapes. The main desorption peak has shifted towards lower temperatures from 820 K (series 1) to 770 K (series 3) which can be explained by the differences in the temperature ramp and the thermal contact. However, desorption of both D2 and CD4 in series 3 starts already at 500 K which corresponds to the surface temperature of 200°C. Since the sample was cooled down in vacuum (~10-7 mbar), it is suggested that there has been a process of deuterium re-distribution into other trapping sites during the sample cooling phase. This peak shape has been observed for two measurements performed within the series 3. It remains unclear, why this process did not take place during the measurement from series 1. The only difference was the specific HOPG sample used for the measurements, but both of the samples were of the same ZYB type and were purchased from the same supplier.

122

4. Surface state characterization of NI enhancers

a. Series 1 17 signal from D 2x10 HOPG RT 2

signal from CD4

) m²

* s

(

1x1017

D flux, at/

400 600 800 1000 1200 1400

Temperature, K

b. Series 1 17 HOPG 400°C signal from D 2x10 2 signal from CD

) 4

m²

* s

(

17 1x10

D flux, at/

0 400 600 800 1000 1200 1400 Temperature, K

c. Series 2 17 2x10 HOPG 800°C signal from D2 signal from CD

) 4 m²

* s ( 1x1017

D flux, at/

400 600 800 1000 1200 1400

Temperature, K Figure 4.15. TPD spectra of D2 (black solid curve) and CD4 (red dashed curve) measured ex situ on HOPG exposed to plasma (20 W, 2 Pa, Vs = – 130 V) at different temperatures: a. RT; b. 400°C; c. 800°C. The CD4 signal is smoothed by 10 points adjacent averaging. 123

4. Surface state characterization of NI enhancers

a. Series 3 HOPG RT 17 signal from D 2x10 2 signal from CD4

) m²

* s

(

1x1017

D flux, at/

0 400 600 800 1000 1200 1400

Temperature, K

b. Series 3 HOPG 400°C 17 2x10 signal from D2 signal from CD

) 4

m²

* s

(

17 1x10

D flux, at/

0 400 600 800 1000 1200 1400 Temperature, K

c. Series 2 17 HOPG 800°C signal from D 2x10 2 signal from CD

) 4 m²

* s ( 1x1017

D flux, at/

0 400 600 800 1000 1200 1400

Temperature, K

Figure 4.16. TPD spectra of D2 (black solid curve) and CD4 (red dashed curve) measured ex situ on HOPG exposed to plasma (20 W, 2 Pa, Vs = – 130 V) at different temperatures: a. RT; b. 400°C; c. 800°C. The CD4 signal in c is smoothed by 10 points adjacent averaging. 124

4. Surface state characterization of NI enhancers

2.5x1017

HOPG RT ex situ signal from D 2.0x1017 2 3rd series

1st series 17 1.5x10

1.0x1017

D flux, at/s*m² 5.0x1016

0.0 300 400 500 600 700 800 900 1000 1100 1200 1300

Temperature, K

Figure 4.17. TPD spectra of D2 measured ex situ within series 3 (black solid curve) and series 1 (blue dashed curve) on HOPG exposed to plasma (20 W, 2 Pa, Vs = –130 V) at RT.

17 2.5x10 HOPG 400°C ex situ

signal from D2 2.0x1017 3rd series

1st series

1.5x1017

17 1.0x10

D flux, at/s*m² 5.0x1016

0.0

300 400 500 600 700 800 900 1000 1100 1200 1300 Temperature, K

Figure 4.18. TPD spectra of D2 measured ex situ within series 3 (black solid curve) and series 1 (blue dashed curve) on HOPG exposed to plasma (20 W, 2 Pa, Vs = –130 V) at 400°C. 125

4. Surface state characterization of NI enhancers

Research in the literature revealed that the desorption peaks at 490 K and 590 K have been observed by Hornekær et al after performing adsorption of hot deuterium atoms (2200 K) on the HOPG surface by means of atomic beam source [119]. Their sample was annealed prior to deuterium adsorption. The energy of the impinging atoms at this temperature was not more than 0.2 eV, which suggests that adsorption occurred only on the topmost surface. The peaks at 490 K and 590 K in their study have been attributed to

D2 desorption from different dimer structure types on HOPG top surface as can be seen in Figure 4.19 [119]. These peaks within experimental uncertainty coincide with the high temperature peaks at 515 K and 580 K in our spectrum. A peak at 510 K (with a small shoulder at 580 K) has been seen by Güttler [121] and Zecho [122] after admission of 10 to 12 ML of thermal D atoms on different HOPG-ZYH surfaces (pristine or annealed at high temperatures in air). This peak was attributed to the C(0001)-D bond breaking and D2 desorption from (0001) terraces on the sample surface which correlates with the attribution of Hornekær et al., since terraces can be understood as flat graphitic surfaces placed at different heights, because of HOPG sample irregularities. Therefore, our peaks at 515 K and 580 K can probably originate from D atoms bonded on terraces. The peak at 820K, 950 K and 1230 K has been observed by Zecho after annealing the HOPG surface at 1100 K in air (which created the surface modifications of ~ 20 μm). The peaks were attributed to decomposition of CD3, CD2 and CD groups (correspondingly) formed at (0001) terrace edges. Their quantity has been largely increased by annealing in air, because the number of terrace edges was elevated thanks to the emerging defects. Both peaks at 820 K and 1010 K were also seen by Liu et al. in the simulations of H+ irradiated graphite [120]. In these simulations hydrogen ions were assumed to be pre- implanted into the porous graphite with ion energy of 1.0 keV at room temperature with the fluence around 5 · 1013 H+/cm2. The authors suggest that appearance of the peak at around 820 K stems from detrapping of H atoms from CH2 complex belonging to sp3 configuration, and a peak at around 1000 K results from C-H bond breaking (sp2 configuration).

D2 desorption peak at 850 K as well as smaller peaks at ~1130 K and 1220 K were seen by Pisarev et al. for hard C-D films deposited in CD4 plasma on W substrate with the potential on the sample of -100 V [123]. In the study of Salançon et al [124] about hard (high density ~ 2.2 g/cm3) and soft (low density ~ 1.3 g/cm3) a:C-D films deposited on Si substrate, the same desorption maxima positions are observed for hard films. In the study of Küppers of C:H films (deposited from ethane at 350 K), he has found broad hydrogen desorption peaks at 920 K and 1120 K and comparing to HREELS spectra has correlated them with the elimination of sp3 CHx group and sp2 CH group respectively [125]. HOPG 400°C spectrum is in agreement with Pisarev et al and Salançon et al what concerns desorption maxima location [123, 124]. However, our desorption peaks in TPD spectra are much better resolved which suggests that our sample surface state looks more like HOPG with small induced defects (small doses of 0.5 keV Ar+, Güttler et al)

126

4. Surface state characterization of NI enhancers

Figure 4.19. D2, TPD spectrum from the HOPG surface after a 2 min D atom dose (ramp rate: 2 K/s below 450 K, 1 K/s above). (a) Dimer A structure of hydrogen atoms adsorbed on neighbor carbon atoms. (b) Dimer B structures of hydrogen adsorbed on carbon atoms at opposite sides of the graphite hexagon [119].

rather than strongly modified HOPG (big doses of 1 keV D2+, Kimura et al [126]) or typical a-C:H layers (Pisarev et al, Salançon et al). The performed bibliographical research leads us to the following attribution of HOPG RT desorption peaks:

 with a confidence, the peaks at and 1010 K and 1200 K can be attributed to the elimination of sp2 CD groups, which is in agreement with the attribution by

other authors. The fact which reinforces this conclusion is that solely D2 desorption is seen at these temperatures.  the peaks at 515 K and 580 K are attributed to the dimer formation on the top surface; the other high temperature peaks in our case stem from the defects induced by the plasma exposure.

In order to go further, TPD is correlated with Raman spectroscopy in the next section. 127

4. Surface state characterization of NI enhancers

4.2.5. Correlation of Raman spectroscopy with TPD for HOPG

The Raman measurements versus surface temperature were performed on two samples (HOPG exposed at RT and HOPG exposed at 400°C); heating was done with a procedure described in Section 4.1.3. Raman micro-spectrometer calibration. The linear heating rate was set to β = 1 K/s in order to match the TPD conditions. The spectra of HOPG exposed to plasma at RT are shown in Figure 4.20a for different temperatures in the Raman heating cell. These temperatures were chosen to match the crucial points of the corresponding TPD spectrum, but one should bear in mind that the difference in the peak temperature between Raman spectroscopy and TPD can be up to 50 K. As can be observed in Figure 4.20a, heating of the sample results in elimination of defects and reconstruction of the initial surface arrangement. For temperatures up to 460 K, the surface state does not change according to the measured spectra. Then, at 660 K, a decrease of the non-zero background between 1400 K and 1550 K (stemming mostly from D3-IL band) is seen, whereas the peak at 1350 K (D2-IL band) does not lose the intensity, but demonstrates a reduced FWHM. When heating further, the intensities of both bands decrease, so that at 1100 K the spectrum looks like a pristine HOPG surface, but with a small D2-IL band remaining. The evolution of the above-mentioned Raman bands with temperature in HOPG RT spectra can be better seen in Figure 4.21a where the normalized TPD spectrum is also shown for comparison. The symbols of different colors denote the normalized integrated areas of D2-IL and D3-IL bands taken from the Raman spectra fit. Note that the band areas are normalized by the maximum value of the bigger D2-IL band. The most important conclusion from this graph can be drawn by looking at 660 K: the area of the D2-IL band has decreased by 20% (due to diminishing of FWHM), whereas the D3-IL band has decreased by 2 times. This allows us to attribute the main TPD peak of HOPG RT to out- of-plane defects induced by plasma exposure (sp3 phase located in the amorphous matrix between the graphitic clusters). At this temperature, according to Küppers, elimination of sp3-bonded H/D does not allow the re-formation of graphitic cycles, but leaves the radicals with dangling, non-saturated bonds [125]. This is in agreement with our observations: D3-IL band area decreases while D2-IL band stays nearly constant.

For 800 K the situation is reversed: D2-IL band decreases by about two times while

D3-IL band only diminishes by 20%. Küppers suggests that in the temperature range of 700–900 K the thermal decomposition (under UHV conditions) of a-C:H films occurs, with the change from sp3 to sp2 dominance at ~750 K (480°C). Indeed, for T>600 K the radicals are not stable anymore, so carbon atoms readily form C=C double bonds, thereby re-increasing the size of graphitic domains (coherence length La increases)

[125]. This reasoning can explain the visible diminishing of D2-IL band in the Raman spectrum between 660 K and 800 K, since the intensity of this band is correlated to the 128

4. Surface state characterization of NI enhancers

coherence length (see Eq. (4.10)). The presence of CD4 desorption in the TPD spectrum

(see Figure 4.15a and Figure 4.16a) and the decrease of the D3-IL band in the Raman spectra indicates the elimination of out-of-plane sp3 defects from the amorphous matrix.

The decrease of the D2-IL band indicates the increase of the aromatic domain size due to elimination of sp3 defects, creation of carbon radicals and thermally activated re- formation of sp2 aromatic domains.

Between 800 K and 1100 K there is still a strong decrease of D2-IL band, while D3-IL band has already a very small intensity (see Figure 4.20a). According to Küppers, starting from 900 K (630°C) the films become purely graphitic or aromatic sp2 (with the sp2 bonding type depending on the residual amount of hydrogen retained in the film) [125]. This confirms that the peak at 1050 K originates from elimination of sp2 in-plane defects (deuterium bonded to aromatic cycles). In case of HOPG heated to high surface temperatures under plasma exposure, there must be similar processes taking place to the ones described by Küppers for a-C:H films heated under UHV conditions. This is confirmed by the reduction of out-of-plane defects in the initial Raman spectrum of HOPG 400°C as compared to HOPG RT (as already discussed in Section 4.1.5. Raman spectroscopy of carbons exposed to plasma). The spectra of HOPG 400°C shown in Figure 4.20b behave similarly to HOPG RT as a function of temperature in the Raman heating cell: the D2-IL band decreases with the surface temperature and so does a much smaller D3-IL band. By looking at Figure 4.21b, it can be seen that after heating to 660 K, the areas of both bands are slightly reduced. At

820 K the area of D2-IL band is significantly reduced, but the D3-IL band drops sharply practically to its minimal level. This leads us to a conclusion that the TPD peaks below 820 K originate from the sp3 phases present in the amorphous matrix. The decrease of the D2-IL band happens due to the formation of unstable carbon radicals during the desorption of D2 and CD4 from sp3-bonded carbon. These unstable radicals form C=C double bonds leading to the decrease of defects in the aromatic domains and hence the increase of the coherence length La, as already discussed above.

However, the D2 desorption peaks at higher temperatures are not related to the out- of-plane defects since the D3-IL band is practically removed at 820 K. Taking into account the change from sp3 to sp2 dominance at ~750 K [125], the D2 desorption peaks at 1000 K and 1150 K can be attributed to the surface state with sp2 arrangement. At such high temperatures (T>900 K), the remaining deuterium is bonded to the boundaries of the aromatic domains (graphitic clusters). Its desorption increases the size of these domains, hence D2-IL band diminishes further. Another important observation to be made is that GIL band in HOPG 400°C Raman spectra disappears almost completely starting from 950 K. This suggests that carbon chains are removed from the IL and the arrangement is purely aromatic (rings), even though defects at the graphitic cluster boundaries are still present.

129

4. Surface state characterization of NI enhancers

1.0 HOPG H RT (a) 2 initial

0.8 370 K 460 K

660 K 800 K 0.6 1100 K

0.4

Normalized intensity Normalized

0.2

0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1

1.0 HOPG H 400°C (b) 2 initial 660 K 0.8 820 K 950 K 1110 K 0.6 1370 K

0.4

intensity Normalized 0.2

0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1

Figure 4.20. Raman spectra of HOPG at different temperatures inside the heating cell (see legend). The samples were exposed to H2 RF plasma in PHISIS (20 W, 2 Pa, Vs = – 130 V) at: (a) room temperature; (b) 400°C. The wavelength of the laser λL = 514nm. Spectra are normalized by the maximum value. 130

4. Surface state characterization of NI enhancers

3 sp 1.0 HOPG RT (a)

integral D2-IL-band

integral D3-IL-band 0.8 TPD spectrum

0.6

3 2 sp +sp 0.4

sp2 Normalized intensity Normalized 0.2

0.0

300 400 500 600 700 800 900 1000 1100 1200 1300 1400

Temperature, K

3 2 1.0 sp +sp HOPG 400°C (b)

integral D2-IL-band

0.8 integral D3-IL-band sp3 TPD spectrum sp2 0.6

0.4

sp2 Normalized intensity Normalized 0.2

0.0

300 400 500 600 700 800 900 1000 1100 1200 1300 1400 Temperature, K

Figure 4.21. Area of the Raman bands D2-IL and D3-IL (normalized by the max value of D2-IL-band) extracted from the fit for HOPG exposed to H2 RF plasma in PHISIS (20 W, 2 Pa, Vs = – 130 V) at: (a) room temperature; (b) 400°C. The corresponding TPD spectra are shown for comparison with a probable peak origin as concluded from Raman analysis. 131

4. Surface state characterization of NI enhancers

4.2.6. Measurements on diamond films

The result of the in situ measurement on MCBDD at RT in absolute units at/(s·m²) calculated by Eq. (4.10) is shown in Figure 4.22 and the comparison of D2 desorption in situ and ex situ measured on MCBDD RT is shown in Figure 4.23. One can notice that the situation is very similar to HOPG at low desorption temperatures: we observe four D2 desorption peaks at nearly the same positions (345 K, 390 K, 490 K and 590 K).The high temperature D2 desorption peaks are seen at 750 K, 920 K and 1070 K. The red solid curve in Figure 4.22 shows desorption of CD4 in the range from 300 K to 900 K with the peaks at the same positions as for D2. The inset in Figure 4.22 shows the zoom for better visibility of the high temperature D2 desorption peaks on MCBDD. The blue dashed curve represents the D2 desorption signal (multiplied by 10) from a molybdenum platen bombarded by an ion gun in the TPD chamber. The low temperature peaks at 345 K and 390 K are dominant in the in situ spectrum, but they almost disappear after exposing the sample to the ambient atmosphere (see Figure 4.23), as for HOPG. The peak at 490 K from the in situ spectrum has been also somewhat reduced in the ex situ spectrum. Desorption of D2 for T > 590 K is very similar for both in situ and ex situ spectrum of MCBDD.

The spectra of D2 and CD4 desorption from MCBDD performed at different surface temperatures during the ex situ plasma exposure are shown in Figure 4.24 in absolute units (at/s·m2). Figure 4.24a demonstrates the spectrum of MCBDD RT from series 3, with the black solid curve showing D2 desorption and the red dashed curve – CD4 desorption. The following D2 desorption features are seen: peaks at 515K, 580 K and

900 K, shoulders at 750 K and 1100 K. Desorption of D2 at 515K and 580 K is accompanied by various impurities such as H2, H2O, CO2, CO and CH4. The red dashed curve in Figure 4.24a shows desorption of CD4 in the range of 400 K – 1000 K, with a main peak at 580 K, shoulders at 515K and 750 K, and a minor high temperature peak at ~1150 K. The MCBDD RT spectra between the series 1 (blue dotted curve) and 3 (black solid curve) are compared in Figure 4.25. The main desorption peak of MCBDD RT has moved from 930 K in series 1 to 900 K in series 3 which lays within experimental uncertainty. Desorption peak at 560 K in series 1 correlates with the peak at 580 K in series 3, but there are no peaks at lower temperature and no desorption whatsoever below 470 K for series 1. For the rest, the shape of the spectra seems very similar.

Considering the total deuterium content in the sample (D2 + CD4), the difference between the spectra is 36% in favor of series 3.

In the spectrum of MCBDD 400°C (black solid curve in Figure 4.24b) D2 desorption peaks at 750 K and 1110 K are seen, and a shoulder in the range 950 – 1000 K is visible. Both of the peaks are also present in the “virgin-like” situation for 5 sec plasma exposure (black solid curve in Figure 4.26) which suggests that at this temperature a slightly

132

4. Surface state characterization of NI enhancers

17 8x10 17 1.5x10 17 7x10 1.0x1017

6x10 5.0x1016 D flux, at/s*m² flux, D

5x1017 0.0 400 600 800 1000 1200 MCBDD RT in-situ Temperature, K 17 4x10 signal from D 2 17 signal from CD 3x10 4 D flux, at/s*m² Mo holder 17 2x10 (signal from D ) x 10 2 1x1017

0 300 400 500 600 700 800 900 1000 1100 1200 1300

Temperature, K

Figure 4.22. TPD spectra of D2 (black solid curve) and CD4 (red solid curve) measured in situ on MCBDD exposed to plasma in situ at RT (60 W, 2 Pa, Vs = – 130 V). The D2 spectrum of Mo platen bombarded by the ion gun (130 eV/D2+, magnified by 10 times) is shown as blue dashed curve for comparison.

8x1017 MCBDD RT 17 D desorption 7x10 2

17 6x10 in situ ex situ 17 5x10

4x1017

3x1017 D flux, at/s*m² 2x1017

1x1017

0 300 400 500 600 700 800 900 1000 1100 1200 1300

Temperature, K Figure 4.23. TPD spectra of D2 measured on MCBDD exposed to plasma at RT in situ (black solid curve, 60 W, 2 Pa, Vs = – 130 V) and ex situ (blue dashed curve, 20 W, 2 Pa, Vs = – 130 V). 133

4. Surface state characterization of NI enhancers

a. Series 3

signal from D2 17 MCBDD RT 3x10 signal from CD 4

2x1017

D flux, at/s*m² 17 1x10

400 600 800 1000 1200 1400 Temperature, K

b. MCBDD 400°C 3x1017 Series 1 signal from D 2 17 signal from CD 2x10 4

D flux, at/s*m² 1x1017

0 400 600 800 1000 1200 1400

Temperature, K

c. Series 2 MCBDD 800°C signal from D 3x1017 2 signal from CD 4

2x1017

D flux, at/s*m² 1x1017

0 400 600 800 1000 1200 1400

Temperature, K Figure 4.24. TPD spectra of D2 (black solid curve) and CD4 (red dotted curve) measured ex situ on MCBDD exposed to plasma (20 W, 2 Pa, Vs = – 130 V) at different temperatures: a.RT; b. 400°C; c. 800°C. CD4 signal in b and c is smoothed by 10 points adjacent averaging. 134

4. Surface state characterization of NI enhancers

3.0x1017 MCBDD RT ex situ

17 signal from D 2.5x10 2 3rd series st 2.0x1017 1 series

1.5x1017

1.0x1017 D flux, at/s*m²

5.0x1016

0.0 300 400 500 600 700 800 900 1000 1100 1200 1300

Temperature, K Figure 4.25. TPD spectra of D2 measured ex situ within series 3 (black solid curve) and series 1 (blue dashed curve) on MCBDD exposed to plasma (20 W, 2 Pa, Vs = – 130 V) at RT.

defective film structure is rebuilt. The shoulder of MCBDD 400°C may come from the same adsorption site as the peak at 900 K of MCBDD RT. The form of MCD 400°C spectrum (shown in solid black in Figure 4.27) is very similar to that of MCBDD (shown in solid grey for comparison) except for the peak at 750 K. There is some D2 desorbed in the range 550 K – 850 K, but it is much less pronounced. What concerns the peak origin in our TPD spectra, the low temperature desorption peaks which dominate the in situ MCBDD spectra are supposed to originate from the products of chemical erosion during the plasma exposure trapped in gaseous form inside the IL and blocked by the impurities, analogously to HOPG. Therefore, no corresponding correction will be made in the total desorption amount calculation for the samples analyzed ex situ. It has been noticed that all MCD 400°C samples reproducibly demonstrated an intense narrow D2 peak appearing in the range 1000 – 1200 K accompanied by the abrupt desorption of a great quantity of CO and a minor quantity of HD and CO2 (see Figure 4.24). These features were also seen by Kimura et al [126]. The sample in their study was a polycrystalline diamond exposed to plasma at RT. They have seen a narrow maximum present in the spectrum at 1200 K which could be induced by structure annealing such as rupture of bubbles.

135

4. Surface state characterization of NI enhancers

MCBDD RT 5sec

17 signal from D 3x10 2 signal from CD4 MCBDD 400°C

signal from D 2x1017 2

D flux, at/s*m² flux, D 17 1x10

0 400 600 800 1000 1200 1400

Temperature, K Figure 4.26. TPD spectra of D2 (black solid curve) and CD4 (red dashed curve) measured on MCBDD exposed to plasma (20 W, 2 Pa, Vs = – 130 V) for 5 sec at RT. CD4 signal is smoothed by 10 points adjacent averaging. The spectrum of MCBDD 400°C (exposed to plasma for 30 min) is shown in grey solid for comparison.

MCD 400°C

signal from D2 17 3x10 signal from CD 4 MCBDD 400°C

signal from D 2x1017 2

D flux, at/s*m² flux, D 17 1x10

400 600 800 1000 1200 1400 Temperature, K

Figure 4.27. TPD spectra of D2 (black solid curve) and CD4 (red dashed curve) measured on MCD exposed to plasma (20 W, 2 Pa, Vs = – 130 V) at 400°C. The CD4 signal is smoothed by 10 points adjacent averaging. MCBDD 400°C spectrum is shown in grey solid for comparison. 136

4. Surface state characterization of NI enhancers

The black solid curve in Figure 4.24c shows the spectrum of MCBDD exposed at 800°C which may be put in correlation with the spectrum at 850°C in Bobrov et al [127]. Deuterium adsorption-desorption from a diamond (100) single crystal surface was investigated by TPD in their study. Exposures of the sample to deuterium were carried out via an effusive leak source (hot tungsten filament was creating atomic D from the molecules). Repeated deuterium adsorption-thermal desorption from the diamond (100) surface has resulted in surface degradation, i.e. formation of disordered carbon on the surface with enhanced sp2 character (confirmed by electron energy loss spectroscopy: EELS).

The following adsorption sites have been identified in their study:

 α1 – peak at 1350 K characteristic for clean diamond (100) surface. It is also seen for partially degraded surface and disappears when surface becomes fully degraded

 α2 – peak at 1250 K seen on partially degraded surface

 α3 – peak around 1100 K, typical for fully degraded surface. It shifts towards lower temperature with surface degradation

Su and Lin performed a LEED (low electron energy diffraction) and TPD studies on type IIa natural diamond (100) single crystal [128]. The procedure was following: polishing → microwave H-plasma treatment → exposure to D-atoms created by passing

D2 over a W filament. As a result, H-terminated surface was becoming fully deuterated. Then, subsequent LEED/TPD studies were undertaken. After that, 15 cycles of H adsorption-annealing treatment were performed on the sample, and LEED/TPD studies were repeated.

As a result, two desorption sites were identified:

 α-sites: 1250 K, ~ 80 kcal/mol, 1st order kinetics, well-ordered (2×1) LEED pattern → H-desorption from (2×1) domains  β-sites: 1150 K, ~ 75 kcal/mol, 2nd order kinetics, (2x1) LEED pattern with diluted half-order spots (from underlying layer) → H-desorption from (2×1) domain boundaries, smaller (2×1) domains due to surface degradation

Performing TPD and infrared spectroscopy studies on type Ia (100) diamond, Yang et al have also identified the same desorption sites [129]. Note that all the above- mentioned studies concern only the top surface of diamond, since there was no energetic bombardment. Concerning the TPD spectra in the present study, the peaks at around 1100 K and 1250 K from Bobrov et al (1200 K in our study) are seen for MCBDD 800°C and MCD 800°C (not shown here). This suggests that the diamond film at this temperature demonstrates a behavior similar to that of a single crystal with a partial amount of surface degradation (H stored on small 2×1 domains and at domain boundaries), 137

4. Surface state characterization of NI enhancers

because α2 and α3 sites are seen (analogous correspondingly to α and β-sites of Su and Lin). It is also visible that some deuterium still remains on the sample surface for T > 1370 K. The peak at 1110 K in MCBDD/MCD 400°C spectra corresponds well to the

α3 site from Bobrov et al (β-site from Yang et al and Su and Lin) characteristic for a partially degraded diamond (100) surface. Bobrov et al has also performed deuterium adsorption at different surface temperatures followed by TPD measurements. The diamond TPD spectrum at 450°C looks exactly alike with MCD 400°C spectrum: peak at 1150 K and a shoulder at 1000 K are seen [127]. The attribution of the peak at 750 K on MCBDD 400°C can be confirmed by comparing the MCBDD and MCD spectra at 400°C (see Figure 4.27), where its absence in case of MCD suggests the origin related to boron. Annen et al in the study of amorphous hydrogenated boron (a-B:H) thin films have seen two H2 desorption maxima: at 600 K and 820 K [130]. They were attributed to B-H-B 3-center bond and B=H terminal bond correspondingly and may explain the origin of our D2 desorption peak centered at 750 K. However, assuming one D-atom per boron atom, we would need 8% or boron doping to explain the desorption peak at 750 K while in our samples the doping percentage is around 1–2%. The form of BDD RT spectrum is very similar to the spectra seen by Küppers for C(B):H films with 1–2% of boron doping which corresponds to the case of the present study [125]. Küppers speculates that boron present in the material is unable to participate in the growth of graphitic segments in the film. In such a way, a B atom in the C-B-H network serves as a dead end for further graphitic growth which leads to the increase of the sp3 content in the material at the expense of sp2 component. Therefore, in our case the presence of boron together with plasma exposure may lead to the formation of non-diamond sp3 phase to which deuterium will be bound thereby explaining the D2 desorption peak at 750 K. As one can see from Figure 3.11, the negative ion yields for MCBDD and MCD at 400°C are comparable which means that boron does not play any role in NI formation, but only in surface conductivity of the material. One can conclude that for MCBDD 400°C and 800°C desorption from 1000 K to 1200 K corresponds to the peaks observed by several authors [127 – 129] on degraded diamond (100) single crystal surface. At 400°C the number of defects corresponding to these peaks has increased compared to 800°C. Heating the diamond under plasma exposure thus allowed reconstructing the original diamond surface through enhanced etching of sp2 phases [91]. The D release at 400°C is observed mostly in molecular form, whereas the origin of CD4 desorption for a higher temperature ~ 1200 K is not clear yet. At RT the diamond layer is more defective; from general considerations and the Raman spectra shown in Figure 4.9 one can conclude that the main D2 desorption peak at 900 K corresponds to sp2 defects. The fact that CD4 desorption does not accompany the D2 peak reinforces this conclusion. The peaks at 490 K and 590 K in TPD spectrum of MCBDD RT approximately coincide with desorption peaks of HOPG RT. They could 138

4. Surface state characterization of NI enhancers

originate from the amorphous sp3 phase on the top surface resulting from the plasma exposure. The performed bibliographical research leads us to the following attribution of MCBDD desorption peaks:

 The peaks at and 1010 K and 1200 K can be attributed to the α2 and α3 sites from Bobrov et al (analogous correspondingly to α and β-sites of Su and Lin) characteristic of a diamond single crystal (100) with a partial amount of surface degradation. These are probably top surface sites.  The peak at 750 K on MCBDD 400°C can be attributed to the presence of boron which together with plasma exposure may lead to the formation of non- diamond sp3 phase to which deuterium will be bound.

In order to identify the origin of the other desorption peaks, TPD will be correlated with Raman spectroscopy in the next section. Several authors performing TPD on diamond surfaces proposed similar attribution of their desorption sites. Bobrov et al attributed site α1 to D desorption from monohydride

C-D on a well-defined diamond (100) surface. Sites α2 and α3 are attributed to D desorption from a mixture of CDx species formed on disordered diamond (100) degraded surface [127]. Su and Lin suggested monohydride formation (α-sites) for low surface H-coverage (see Figure 4.28b and Figure 4.29b) and dihydride phase formation for high surface H-coverage (Figure 4.28c and Figure 4.29c) [128]. The latter phase has never been observed experimentally. Energetic predictions of the calculations performed by Yang et al revealed that full dihydride is thermodynamically unstable as the system may reduce its energy by desorption of H2, producing the monohydride [129]. The other possibility is an alternating sequence of one and two H atoms attached to each carbon – “1.33hydride” which would be thermodynamically stable (Figure 4.28d). The infrared spectroscopy evidence was seen for the monohydride surface structure (with one hydrogen atom per surface carbon atom) by Yang et al for the first time on diamond (100) [129]. Several other authors explain their experimental results by the formation of monohydride complexes on polycrystalline diamond surfaces: Thoms et al (HREELS (high resolution EELS) study of thermal H and D adsorption at polycrystalline diamond surfaces with dominating (111) oriented facets and synthetic diamond (100) surfaces [131,132]), Koleske et al (TPD study of polycrystalline diamond surfaces [133]). Hamza et al measured TPD spectra at H exposed (100) and polycrystalline diamond [134]. The surfaces exhibited molecular hydrogen desorption in the range 800 K (high coverage) to 1250 K (low coverage). He has concluded that C-H bond-breaking is not the rate limiting step for desorption, and the release of molecular hydrogen from the hydride complexes with simultaneous formation of the C=C double bond (see Figure 4.28a and Figure 4.29a) provides a low-lying transition state for desorption. Similar conclusions were made by Koleske et al. 139

4. Surface state characterization of NI enhancers

Influence of surface states, defects and adsorbates on the electronic properties of diamond surfaces was investigated by Ristein et al [135]. They have combined photoelectron spectroscopy and in situ conductivity measurements to elucidate the surface conductivity of diamond. They found that H passivation leads to a negative electron affinity (NEA) of diamond surfaces due to a dipole layer which is induced by the heteropolar C-H bonds of the surface atoms. It was concluded that the surface conductivity of diamond is a consequence of H-termination of the surface together with coverage by physisorbed adsorbates (impurities, for instance).

Figure 4.28. Top and side views of atomic structures of clean and hydrogenated diamond (100) surfaces: (a) (2 × 1); (b) (2 × 1):H; (c) (1 × 1):2H; (d) (3 × 1):1.33H. Bond lengths are given in ångstroms [129]. Hollow circles of different sizes represent carbon atoms in top, second, and third layers respectively; black circles represent hydrogen atoms. 140

4. Surface state characterization of NI enhancers

Figure 4.29. Schematic 3D illustration of hydrogenated diamond (100) surfaces: (a) clean diamond (100)–(2×1), (b) (100)–(2×1):H monohydride with one hydrogen atom per surface carbon atom, and (c) (100)–(1×1):2H full dihydride with two hydrogen atoms per surface carbon atom [128].

J. van der Weide et al has used ultraviolet photoelectron spectroscopy (UPS) and LEED to study natural diamond wavers with (100) surface orientation annealed at various temperatures [48]. They have noticed the lowering of the work function with annealing temperature and the appearance of negative electron affinity (NEA) after annealing at 1070°C. Referring to literature, J. van der Weide et al state that NEA has been 141

4. Surface state characterization of NI enhancers

demonstrated on diamond (111) surface and associated with presence of H bonded to the surface. Density functional theory (DFT) ab initio theoretical study performed by J. van der Weide et al has confirmed that the electron affinity is positive on hydrogen-free diamond C(100)–2×1 surface (see Figure 4.28a and Figure 4.29a) and negative on monohydride C(100)–2×1:H surface (Figure 4.28b and Figure 4.29b) [48]. Similar results have been also obtained by Hui Ying Hoh et al performing advanced DFT calculations [49]. On diamond with (111) orientation which is present in our polycrystalline samples the situation can be different. The clean (111) diamond surface undergoes (2×1) reconstruction with a π-bonded chain structure, C(111)–(2×1), as shown in Figure 4.30a. On the other hand, at complete or near complete hydrogen coverage diamond (111) surface has a bulk-terminated (1×1) structure, with the dangling bonds or radicals terminated by hydrogen atoms, C(111)–1×1:H as shown in Figure 4.30b [136]. Thermal treatment of the hydrogen-free (111) diamond surfaces can lead to graphitization of a surface region. It is known that even a small amount of hydrogen can stabilize the diamond surface against graphitization and the presence of hydrogen plays an important part in CVD diamond growth. It should be mentioned that the structure of the C(111) surface at intermediate H coverages includes a formation of a metastable structure, as confirmed by molecular dynamics simulations [136] and sum-frequency generation (SFG) spectroscopy experiments [137]. As could be inferred from the study of Ristein et al and other various authors, the H/D-termination is a necessary, but not the only requirement for diamond surface conductivity. This suggests that heating of diamond with a constant D supply from the plasma creates a certain D surface arrangement which enables this mechanism. The enhancement of the NI yield for heated diamond surfaces could be explained by achievement of NEA via the formation for surface hydride complexes (monohydride or dihydride). The following NI signal decrease starting from 400°C up to 800°C is probably connected to the desorption of D from the surface (passage from high to low D- coverage).

(a) (b)

Figure 4.30. Side views of diamond (111) surfaces. The larger blue and red spheres represent C atoms, and the smaller yellow spheres represent H atoms. (a) C(111)–(2×1) – hydrogen-free diamond C(111) surface with (2×1) structure; (b) C(111)–1×1:H – bulk- terminated (1×1) structure obtained on diamond C(111) in the presence of hydrogen [138]. 142

4. Surface state characterization of NI enhancers

4.2.7. Correlation of Raman spectroscopy with TPD for MCBDD

The Raman measurements for MCBDD RT and MCBDD 400°C heated to different temperatures were performed analogously to HOPG. The initial spectrum of MCBDD RT in the wavenumber range corresponding to the defect-related band (at 1450 cm-1 – 1700 cm-1) is shown in Figure 4.31. As can be concluded from the band fit, there are two components which contribute to the measured Raman signal: D3-IL-band centered at 1515 cm-1 related to the sp2 defects whose frequency is modified in the presence of sp3 phase and surface amorphization, and GIL-band centered at 1585 cm-1 also suggesting the presence of sp2 phase defects. The presence of these bands proves that the surface state of MCBDD RT is disordered and sp2 phases are created owing to plasma exposure. Knowing the initial crystalline sp3 structure, we can expect the formation of aliphatic double C=C bonds rather than aromatic cycles.

0.5

MCBDD RT

0.4 data

band fit 0.3 total fit

0.2

0.1

Normalized intensity

0.0

1350 1400 1450 1500 1550 1600 1650 1700 1750

Wavenumber, cm-1

Figure 4.31. Raman spectrum in the range corresponding to the defect-related band of MCBDD exposed to H2 RF plasma in PHISIS at RT. Experimental data is shown by the black curve with the fit of individual bands (red solid curves) and their sum (total fit shown by red dashed curve). The wavelength of the laser λL = 514 nm.

143

4. Surface state characterization of NI enhancers 1.6 (a) initial MCBDD RT 514 nm 1.4 600K NORM by Di peak 800K

1.2 900K 1000K 1100K 1.0

0.8

0.6

Normalized intensity Normalized 0.4

0.2

0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1 1.6 (b) MCBDD 400°C 514 nm 1.4 NORM by Di peak initial 1.2 850K 1000K 1125K 1.0 1200K

0.8

0.6

Normalized intensity Normalized 0.4

0.2

0.0

1000 1100 1200 1300 1400 1500 1600 1700 Wavenumber, cm-1

Figure 4.32. Raman spectra of MCBDD at different temperatures inside the heating cell (see legend). The samples were exposed to H2 RF plasma in PHISIS (20 W, 2 Pa, Vs = – 130 V) at: (a) room temperature; (b) 400°C. The wavelength of the laser λL = 514 nm. Spectra are normalized by the intensity of the diamond peak at 1302 cm-1. 144

4. Surface state characterization of NI enhancers

MCBDD RT 1.0 (a) integral GIL-band integral D3-IL-band TPD spectrum 0.8

0.6

0.4

Normalized intensity Normalized

0.2

0.0 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500

Temperature, K

MCBDD 400°C 1.0 (b) integral GIL-band integral D3-IL-band

TPD spectrum 0.8

0.6

0.4

intensity Normalized

0.2

0.0 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 Temperature, K

Figure 4.33. Area of the Raman bands GIL and D3-IL (normalized by the max value of GIL- band) extracted from the fit for MCBDD exposed to H2 RF plasma in PHISIS (20 W, 2 Pa, Vs = – 130 V) at: (a) room temperature; (b) 400°C. The corresponding TPD spectra are shown for comparison. 145

4. Surface state characterization of NI enhancers

When heating MCBDD RT (see Figure 4.32a), an unexpected behavior is observed: the defect-related band starts to grow, reaches a maximum at 800 K and drops down for higher temperatures. This effect has been verified by several experimental series and for MCBDD with two percentages of boron doping, and it was always reproducible. For the temperatures above 1200 K, the graphitization of the samples was observed from the Raman spectra (not shown here) and even by looking at the sample with a naked eye. The behavior of the individual bands can be observed much better in Figure 4.33a where the band areas are normalized by the maximum value of the more intense GIL- band and the TPD spectrum of MCBDD RT is shown. The D3-IL band denoted by red symbols stays nearly constant up to 900 K and then disappears. This behavior may be correlated with CD4 desorption visible up to 900 K on MCBDD RT, as can be seen in Figure 4.24a. This suggests that deuterium is bound to sp3 non-diamond phase originating from the plasma exposure which is eliminated between RT and 900 K. On the other hand, GIL band behavior resembles the form of the TPD spectrum with a slightly shifted maximum. It may indicate that heating MCBDD in vacuum (during Raman spectroscopy or TPD experiments) rearranges the surface in a way that leads to the formation of sp2 phase. One can speculate that similarly to HOPG, desorption of deuterium from sp3-bonded carbon creates unstable radicals that convert into sp2 arrangement for T > 650 K [125]. Therefore, desorption peaks below 800 K for MCBDD RT is due to the deuterium bound with sp3 carbon (probably in amorphous form). Desorption at higher temperatures (peak at 900 K and shoulder at 1100 K) originates from deuterium bound to sp2 carbon and to the degraded top surface (probably in form of monohydride).

The effect of the GIL band increase is as well present in MCBDD 400°C spectra depicted in Figure 4.32b, but at higher temperatures (above 1000 K). Figure 4.33b demonstrates that at lower temperatures the sp2 defect-related band is absent in the

Raman spectra. The GIL band has a maximum at 1000 K and then abruptly decreases; the

D3-IL band follows the same trend. We can expect a similar mechanism as for MCBDD RT. The elimination of sp3 defects below 1000 K leads to creation of unstable radicals that convert into sp2 carbons with the assistance of high temperature, as can be seen from the increase of GIL band. Deuterium bound to sp2 carbon desorbs for T>1000 K accompanied to deuterium bound to the top surface in form of monohydride and/or dihydride. The fact that D3-IL band is very weak suggests that amorphization of the surface is minimized at 400°C. This may give a hint for the explanation of the NI yield increase on MCBDD 400°C. Monohydride/dihydride complexes on the diamond surface result in surface conductivity and the appearance of NEA (based on the studies of other researchers). At RT the surface amorphization does not allow the appearance of NEA.

146

4. Surface state characterization of NI enhancers

4.2.8. Comparison of total desorption amount

A crucial point of the TPD measurements was to estimate deuterium surface coverage for our samples. The desorbed deuterium during the measurements comes from the whole implantation range (from thickness 0 to Xmax) in D2 and CDx forms. Therefore, we could calculate the mean abundance of D-atoms with respect to the total number of atoms (atomic concentration) in the whole interaction layer.

The value of Xmax was obtained from SRIM simulations assuming 10 nm of amorphous carbon layer with two different densities (2.2 g/cm3 to model HOPG and 3.5 g/cm3 for diamond films) which was bombarded by 45 eV H+/ D+ ions. The value of deuterium surface coverage all samples were taken as 0%. We have defined Xmax as depth containing 99% of the implanted ions. As a result, the atomic concentration of deuterium in the interaction layer could be estimated for all temperatures of plasma exposure and all materials. Then, the calculated concentrations of deuterium for each condition were included in the next SRIM calculation as deuterium surface coverage parameter. In such a way, more precise values of Xmax and therefore of deuterium concentrations were estimated (see Figure 4.34) for deuterium desorbed in D2 and CD4 forms. Various points on the graphs for the same temperature stem from the TPD spectra measured within different experimental series. The error bars in the figures emerge from imperfection of the calibration fit for D2 and CD4, and additionally for CD4 from the uncertainty in σ estimation, because according to the Eq. (4.10) it was necessary to take into account the ratio σCD4/σD2 during the calculations. Moreover, the MS sensitivity was not exactly dependent on the compound mass as 1 m , so it has given an additional error in case of CD4 in the calculation of absolute desorption rate and hence the deuterium concentration in the IL. The total deuterium concentration in the modified layers is represented in Figure 4.35 and demonstrates a certain correlation with NI production yields (shown in Figure 3.11) for given experimental conditions. Ahmad et al have demonstrated by means of modeling that with a sample oriented normally with respect to the MS on PHISIS setup, there is a preferential collection of sputtered negative ions [52]. The contribution of sputtered NI signal at the detector could be as big as 44% at impact energy of 45 eV. The amount of sputtered D– directly depends on the amount of D on the sample surface, and the modeling predicts the change of deuterium surface coverage with plasma exposure temperature in qualitative agreement with TPD results for studied materials, as can be seen in Figure 4.36. Deuterium coverage on carbon materials was obtained by comparing modeling with experimentally measured NIEDF on samples exposed to plasma at different surface temperatures.

Integrating the area below HOPG RT ex situ TPD spectra for D2 + CD4 (in average for series 1 and 3) gives 22% of D-atoms in the interaction layer (IL) and when heating to 400°C, the total amount of deuterium in the interaction layer lowers down to 14% (in average for series 1 and 3). The in situ TPD spectrum demonstrates the half of all 147

4. Surface state characterization of NI enhancers

25 12 a. b. D signal 2 CD signal 10 4 20 HOPG MCBDD 8 MCD 15

10 4

5 2

D concentration, % in the modifiedconcentration,inD the %layer D concentration, % in the modifiedconcentration,inD the %layer 0 0 0 200 400 600 800 0 200 400 600 800 T during plasma exposure, °C T during plasma exposure, °C

Figure 4.34. Deuterium concentration in the modified layers after their exposure to plasma at different surface temperatures: a. accounting desorption in D2 form; b. in CD4 form. The lines serve as a guideline for the eye.

35 HOPG 30 MCBDD MCD

D + CD signal 25 2 4

modifiedconcentration,inD the % layer 0 0 100 200 300 400 500 600 700 800

T during plasma exposure, °C

Figure 4.35. Total deuterium concentration in the modified layers after their exposure to plasma at different surface temperatures for both D2 and CD4. 148

4. Surface state characterization of NI enhancers

40 Model SRIM 50 eV H2

HOPG 35 MCBDD

30 MCD

15 10

% coverage, surface H 5

0 0 100 200 300 400 500 600 700 800

Temperature, °C

Figure 4.36. Deuterium coverage on carbon materials (in %) obtained by comparing modeling with experimentally measured NIEDF on samples exposed to plasma at different surface temperatures.

deuterium content contained in the low temperature peaks that almost disappear in the ex situ spectra. Taking them into account for the ex situ spectra would double the deuterium content in the IL, giving 44% of D-atoms at RT and 28% at 400°C. This is similar to the H/D content in soft (up to 50%) and hard (~30%) a-C:H/D films [124], but seems over-estimated for our nc-G case. The modeling predicts the coverage of 35% for HOPG RT and the decrease to 20% already at 300°C, which is in qualitative agreement with TPD measurements (22% and 14% correspondingly). For T > 400°C the modeling does not match the experimental NIEDF for HOPG. The Raman spectrum for HOPG 400°C (see Figure 4.7b) shows a decrease of bands related to in-plane defects by nearly 2 times (mostly due to GIL diminishing) and decrease of out-of-plane defect-related bands by 3 times. These facts confirm the re- establishment of sp2 phases and increase of the sp2 cluster size and correlate well with the deuterium content evolution. For HOPG 800°C the total deuterium content in the IL is very low (2% for D2 + CD4 within the series 2). As proven by Raman spectroscopy measurements, at such a high temperature only G band attributed to sp2 carbon is visible [47]. This result could be related to reconstruction of graphitic sp2 phases and almost complete removal of defects. For MCBDD RT analyzed ex situ, the average deuterium concentration in the IL (for series 1 and 3) gives ~26%. When accounting that in the in situ MCBDD spectrum the

149

4. Surface state characterization of NI enhancers

half of all deuterium content is also contained within the low temperature peaks, this results in the deuterium concentration of 52% which also seems over-estimated for our case. The modeling predicts the coverage of 30% which is in better agreement with the ex situ measurements. At 400°C the TPD measurements give the deuterium concentration of 25% for MCD and 27% for MCBDD, and the modeling predicts a slight coverage increase (to 35%). At 800°C the result of the TPD measurements is ~5% of deuterium in the IL for both diamond layers versus 20% predicted by the model. One can conclude that only a qualitative agreement for the deuterium coverage evolution with temperature between the TPD measurements and modeling is preserved (same trends). However, we must remember that the model assumes that NI surface production is independent of E and θ (NI energy and angle of emission). This is a strong assumption that has been validated at RT, but not at high temperature. The in situ measurements indicate that the real deuterium content might be higher due to the contribution of the low temperature desorption sites. However, because of the difference in plasma exposure conditions and unclear origin of the low temperature desorption peaks, more studies are required to conclude on the importance of the in situ results.

4.3. Conclusion

This chapter makes a contribution in understanding of the NI surface production in

H2/D2 plasmas thanks to the thermodesorption (TPD) analysis and Raman spectroscopy. The surfaces of highly oriented pyrolitic graphite (HOPG), microcrystalline boron-doped diamond (MCBDD) and microcrystalline non-doped diamond (MCD) were exposed to plasma at different surface temperatures and NI yields were measured by means of mass spectrometry. The TPD analysis was performed on the exposed samples in order to correlate the NI yield evolution with temperature to the surface state changes. The Raman spectroscopy measurements have enabled characterization of HOPG and diamond films after plasma exposure. Additionally, Raman spectroscopy measurements as a function of surface temperature were performed for HOPG and MCBDD previously exposed to plasma at RT and 400°C. This was done on order to cross check with TPD and attribute the desorption peaks to a certain type of defect or surface arrangement (sp2 or sp3 phase). A bibliographic research has allowed to characterize the surface state of the materials for different plasma exposure temperatures. The HOPG RT surface state shows the features of nanocrystalline graphite with a certain amount of in-plane sp2 and out-of- plane sp3 defects and with deuterium content similar to hard a-C:H films. The Raman spectrum for HOPG 400°C has shown a decrease of bands related to in-plane and out-of- plane defects which confirms the re-establishment of sp2 phases and increase of the sp2 cluster size and correlates well with the deuterium content evolution. As proven by 150

4. Surface state characterization of NI enhancers

Raman spectroscopy measurements, at 800°C only G band attributed to sp2 carbon is visible in HOPG spectrum. This result could be related to reconstruction of graphitic sp2 phases and almost complete removal of defects. The presence of the defect-related bands in MCBDD RT Raman spectra indicates the formation of double C=C bonds. This fact proves that the surface state of MCBDD RT is disordered, with sp2 and non-diamond sp3 phases created due to plasma exposure. Heating the diamond under plasma exposure allows reconstructing the original diamond surface through enhanced etching of sp2 phases [91]. The TPD spectra of heated diamond films demonstrate desorption of big quantities of deuterium in D2 form which could stem from the surface. The results of 5 sec plasma exposure suggest that at 400°C the original diamond structure is partially rebuilt. From the bibliographical research one can conclude that for diamond films heated to 400°C and 800°C desorption from 1000 K to 1200 K corresponds to the peaks observed by several authors [127 – 129] on degraded diamond (100) single crystal surface. This suggests that heating of diamond with a constant D supply from the plasma creates a certain D surface arrangement which enables the appearance of negative electron affinity (NEA). The enhancement of the NI yield for heated diamond surfaces could be explained by achievement of NEA via the formation for surface hydride complexes (monohydride or dihydride). The following NI signal decrease starting from 400°C up to 800°C is probably connected to the desorption of D from the surface (passage from high to low D- coverage). It is concluded that the role of boron in NI production is insignificant, though it provides for the conductivity of the material at room temperature. The calibration of the mass spectrometer allowed the calculation of the full amount of desorbed species and their dynamics. By calculating deuterium atomic concentration in the modified layers in D2 and CD4 forms at different surface temperatures, it was seen that the film properties and the NI formation mechanisms are very different for HOPG and diamond films. The NI yield evolution is shown to correlate qualitatively with the total deuterium concentration in the interaction layer, which is in agreement with the modeling developed previously.

151

5. Pulsed-bias approach

The method of pulsed DC bias was developed to enable the study of NI production on surfaces of insulating materials (such as MCD). A similar method have been used by Samara et al [139] to measure the ion saturation current, or by Kudlacek et al [140] in order to control the ion energy in industrial plasma processes dealing with insulators. To overcome the limitations of the method, Wang and Wendt [141] introduced a modulated pulsed approach, where the voltage during the pulse is sloped and thus exactly compensates for the drop of voltage due to the charging of the substrate being processed. Within this thesis, only rectangular shape pulses were used and the diagnostic techniques were the same as employed to study NI surface production on conducting surfaces. As it was previously mentioned in Chapter 3. NI production on different materials, MCD surface was conducting only starting from 300°C. Below this temperature, the NI signal could not be measured. The present technique has enabled to study NI surface production on MCD for the whole temperature range starting from RT. Moreover, this technique extends the measurement of NI surface production to potentially interesting materials. Indeed, electron capture on insulators is difficult, but electron loss is limited, resulting in a potentially high ionization probability. 5.1. General principle of measurements An insulator immersed into plasma acts as a capacitor. When being negatively DC biased, it accumulates positive ions on the surface. If a periodic short-time negative DC- bias is applied to the insulator, and for the rest of the time it is left at floating potential, it allows unloading of the positive charge by electrons during the unbiased period. The NI emitted from the sample can be measured during the time of the bias pulse, so the measured signal can be compared directly to NIEDF obtained with continuous DC-bias. Figure 5.1 illustrates the situation when the bias pulse is ON: sketch on the left shows the change of the voltage with time, and scheme on the right represents the equivalent electric circuit of the sample immersed into plasma. The voltage applied by the bias source is denoted as Va, and the resulting potential on the surface as Vs. The sample itself represents a dielectric separating these two plates, and could be regarded as a planar capacitor. In such a way, the surface bias reads:

Vs = Q/C + Va = d·Q/(ε0·εr·S) + Va Eq. (5.1) where Q is the charge accumulated by the capacitor (in Coulomb) and C its capacitance which can be calculated from the known vacuum permittivity ε0 = 8.85·10-12 (F/m), 152

5. Pulsed-bias approach

relative permittivity of the insulating material εr, surface of the planar insulator S and its thickness d.

Figure 5.1. Situation when the bias pulse is ON. Left part – change of the voltage with time: voltage applied by the bias source Va (red), resulting voltage on the surface Vs (blue) during the charging period of the capacitor. Right part – equivalent electric circuit: plasma (PI current source), insulator (plane capacitor) and a source of pulsed bias.

Figure 5.2. Situation when the bias pulse is OFF. Left part – change of the voltage with time. Right part – equivalent electric circuit of the sample immersed into plasma.

153

5. Pulsed-bias approach

In the beginning of the pulse (at t = 0) the capacitor charge is zero, therefore Vs = Va. Then, the surface bias can be obtained by integrating its time variation: dVs 1 dQ 1 dQ Iisat   Vs  t Va Vf  t Va Vf Eq. (5.2) dt C dt C dt C

Here we consider that the applied bias Va is sufficiently negative so that the ion current saturates. The ion saturation current Iisat of the PI attracted towards the sample reads: kTe Iisat  0.6eni  S Eq. (5.3) mi where ni is plasma ion density, mi is ion mass and Te is electron temperature. Since the ion saturation current is constant provided that Vs during the pulse stays much smaller than Vp, the charging of the capacitor is linear with time as shown by the blue curve on the left part of Figure 5.1.

When the pulse is over, the applied voltage is zero Va = 0, so the surface bias becomes positive: Vs = Q/C. If the charge Q is high enough, the surface bias becomes higher than the plasma potential. In such a way, a huge electron flux is attracted towards the sample during the first instants due to high electron mobility. This causes a perturbation in the plasma, resulting in a local increase of the plasma potential which limits the electron loss towards the sample surface. Then, the sample is discharged slowly by the electrons at a surface potential close to floating potential (as demonstrated by the blue curve in Figure 5.2). The same situation arises if the initial charge Q is low and the surface bias is not higher than the plasma potential:

dVs 1 dQ 1 dQ Q0 1   Vs  t   (Itot t  Q0 ) Eq. (5.4) dt C dt C dt C C In the expression above, the total current towards the sample surface contains both ion and electron components Itot = Iisat + Ie in the case Vs ≤ Vp. Whereas the ion current could be approximated by Iisat, the electron current is governed by the exponential law:  e[V V ]  e[V V ] p s 8kTe p s Ie (Vs )  Iesat exp( )  ene  S exp( ) kTe  me kTe Eq. (5.5) where Iesat is electron saturation current, ne is plasma electron density, me electron mass and the thermal electron velocity is given by:

8kT e    m e,th Eq. (5.6) e 154

5. Pulsed-bias approach

In the other situation, when Vs > Vp, the ion contribution to the total current is negligibly small, so that: Itot = Iesat All the considerations above suppose that the ion and electron currents saturate which implies that edge effects (sheath expansion at the edge of the sample) are negligible which is true for moderate variations of the surface voltage.

Since Vs & Vp change during the bias OFF period (Vp abruptly increase and then slowly decreases), the capacitor discharge happens in one or two regimes depending on the sign of (Vp - Vs). In any case the discharge ends up at low rate with the surface bias close to the floating potential. NI need some time (~12 μs for H–) to arrive to the MS detector. They travel the distance d = 37 mm (crossing the sheaths and the plasma) with a maximum energy of

E = e (Vp – Vs) = 150 eV, and enter the extractor. After crossing the extractor (~ 6 cm) the

NI are accelerated/decelerated to the axis potential Vaxis = 40 V before entering the 27 cm long energy analyzer. Then the ions arrive to the 18 cm long mass filter which they cross at 3 eV (it takes most of the time) and finally to the detector which is 4 cm long and is crossed by the NI accelerated to the dynode potential Vdyn = 1000 V. The measurement technique relies on the use of three different signals as demonstrated in Figure 5.3 The bias voltage Va (shown in red) whose value is determined by the constant DC-bias source is applied during Tpulse (defined by the pulse generator). The period shown in Figure 5.3 in green corresponds to the arrival of NI to the MS detector. The signal drawn in blue shows the start of the MS acquisition which lasts for time Tacq. The time separating the beginning of the bias pulse and the start of the MS acquisition is denoted Tdelay. Both of the latter periods are defined by the delay generator (see Figure 5.5 for all hardware components used in the pulsed-bias experiment). The arriving NI could be observed on the oscilloscope (in the persistence mode) by connecting the output of the MS detector which produced a pulse each time when a NI was detected. In such a way, all the previous pulses from detected NI were kept on the screen for better visibility and tuning of Tdelay (see Figure 5.4 which demonstrates the applied bias pulse and the output of the detector on the oscilloscope screen).

Tdelay = 13 μs was applied between the beginning of the pulse and start of the acquisition in order to get rid of NI created during the first microsecond of the pulse when the applied bias is establishing. In the same way the acquisition duration Tacq was usually set shorter than the applied bias in order prevent the collection of NI created during the fall down of the sheath in front of the sample at the end of the pulse. The main parameters that could be controlled by the primary pulse generator are the pulse frequency f and the pulse duration Tpulse which are linked by so-called “duty cycle”:

Tpulse · f · 100(%) = Tpulse / T · 100(%), where T is the total period of the signal. Basically, it represents the duration of the bias pulse (in percents) with respect to the total period.

155

5. Pulsed-bias approach

In order to find optimal experimental conditions and to test the feasibility of the method at RT, first tests were performed on HOPG as a reference material. Thereafter, the optimization of the bias and acquisition parameters was performed on MCD by taking into account the charging effects. Finally, the results for the whole temperature range were obtained by slightly modifying the chosen parameters to adjust them for the heating experiment.

Figure 5.3. Measurement technique used in the pulsed-bias experiment.

Figure 5.4. The photo of oscilloscope screen demonstrating the applied bias pulse of Va = –130 V during Tpulse = 40 μs (in pink) and the output of the MS detector in persistence mode (in yellow). Each yellow peak corresponds to a NI arriving to the detector. 156

5. Pulsed-bias approach

(d) (a)

(b) (e)

(c)

Figure 5.5. Actual photo of the system used for the pulsed bias experiment: (a) constant DC-bias source defining Va; (b) primary pulse generator; (c) converter of constant DC-bias Va into the pulsed bias by using the inputs from the pulse generator and the constant DC- bias source; (d) delay generator producing “MS input” for acquisition; (e) oscilloscope. The outputs produced by components (b), (c) and (d) are sketched on the right.

5.2. Experimental results

5.2.1. Results for a conductive sample: HOPG

The test of pulsed bias has been performed in H2 plasma at RT under usual experimental conditions: 2 Pa, RF power of 20 W, screen touching the MS, Vs = -130 V distance between the MS and the sample holder d = 37 mm. However, since the usual sample holder was sent for reparation, a provisional sample holder was used for the measurements. It was mounted onto the diffusion chamber on the flange usually dedicated to the Langmuir probe (see Figure 1.2). As a result, the PI distribution in plasma was slightly different from the usual case which gave rise to longer tails in measured NIEDF. The measurements could be performed only at RT as the heating mechanism was not included in the design of the provisional sample holder. First of all, it was confirmed that no charging of HOPG surface was happening during the applied bias pulse. The NIEDF were measured in following conditions: f = 100 Hz,

Tpulse = 1 ms, Tacq = 50 μs. In order to trace the possible variation of the surface bias, Tdelay was varied from 13 μs to 953 μs. The form of NIEDFs was unchanged for all the measurements (not shown here), which means that no charging effects are present. 157

5. Pulsed-bias approach 3.0x105

2.8x105

5 2.6x10

5 2.4x10

2.2x105

2.0x105

Yield,arb.u. NI pulsed bias pulsed acqusition 1 5 1.8x10 pulsed bias pulsed acqusition 2

5 DC-bias pulsed acqusition 1.6x10 DC-bias const acqusition

5 1.4x10 1 10 100 f, kHz

Figure 5.6. NI yield dependence for HOPG in H2 RF plasma as a function of the pulsed- bias frequency (varied from 0.25 to 90 kHz). Parameters: Tpulse = 10 μs, Tacq = 8 μs and Tdelay = 13. Black and red symbols correspond to two different series of measurements. Blue symbols represent the results for constant DC-bias, but pulsed MS acquisition. The orange star gives the value of NI yield for constant bias and constant acquisition. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, with screen. 1 DC-bias V = -130 V s 1 kHz P = 20 W 5 kHz 2 Pa H 30 kHz 2 0.1 90 kHz Bias: 10 s Acq: 8 s Delay: 13 s

0.01

Normalizedintensity 1E-3

0 10 20 30 40 50 60 70 Energy, eV

Figure 5.7. Normalized NIEDF measured on HOPG in H2 RF plasma at RT for constant DC-bias (black curve) and for pulsed-bias at various frequencies (in color). Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Vs = -130 V, with screen. 158

5. Pulsed-bias approach

However, the change of the bias frequency has proven to influence the hydrogen surface dynamics. Figure 5.6 demonstrates the NI yields measured with Tpulse = 10 μs,

Tacq = 8 μs and Tdelay = 13 μs as a function of frequency varied from 0.25 to 90 kHz (duty cycle from 0.25% to 90%). It can be seen that NI signal decreases with bias frequency starting from 1 kHz (see black symbols representing the first series of measurements). However, the red symbols showing the second series of measurements clearly demonstrate the signal saturation for low frequencies (in the range from 0.25 to 1 kHz). These two series were slightly different in terms of signal levels (by approximately 10%), so the NI yield of the second series was put in correspondence with the value at 1 kHz of the first series. The blue symbols in Figure 5.6 represent the results for constant DC-bias, but pulsed MS acquisition. As one can see, the NI yield stays nearly constant for different acquisition frequencies (decrease after 1 kHz happens due to the time evolution of the NI signal). Therefore, it can be concluded that the NI yield decrease with frequency during the pulsed bias is connected to the surface state change of the material and not to the acquisition problems. The orange star in Figure 5.6 gives the value of NI yield for constant bias and constant acquisition and coincides with the pulsed acquisition results as expected. Another important point of the graph is that pulsed bias for f <20 kHz results in the enhanced NI production as compared to constant DC-bias. If we bias HOPG in pulsed mode, an increase of the NI yield up to 25% is observed (as seen from Figure 5.6). By regarding the normalized NIEDF, one could get a better idea of what is happening with HOPG surface under different pulsed bias conditions. As could be seen from Figure 5.7, the NIEDF for 1 kHz has the lowest tail and it grows and approaches the situation of constant DC-bias with increasing pulse frequency. It was verified that for low frequencies (from 0.25 to 1 kHz) the shape of the NIEDF was exactly the same. From the modeling of Chapter 2. Modeling and reconstruction of NIEDF, we know that an increase of the tail with respect to the peak of the NIEDF corresponds to an increase the contribution of backscattering to the total NI yield and a decrease of the sputtering contribution (decrease of the hydrogen coverage θ). To support this reasoning, modeling was done for several values of θ on HOPG surface and compared to the experimental data for various pulsed-bias frequencies. As one could see from the Figure 5.8, Figure 5.9 and Figure 5.10, the hydrogen coverage on HOPG surface pulse-biased with 5 kHz is around 40%, for 30 kHz – around 35% and for f > 70 kHz approaches 30%. The value of θ for DC-biased HOPG was taken as 30%. Therefore, we conclude that the increase of the yield with the decrease of the frequency is due to an increase of the hydrogen coverage on the surface which increases the number of NI created by sputtering. The increased surface coverage can be understood as follows. Surface coverage is increased by atoms and ions impinging on the surface and decreased by the self-sputtering (sputtering of hydrogen atoms by hydrogen ions) and by physico-chemical etching. The self-sputtering is not continuous in this case (and most probably the physico-chemical etching is time modulated), so the time for atomic hydrogen to fill the surface up to the saturation if the 159

5. Pulsed-bias approach 1 V = -130 V s DC-bias P = 20 W 90 kHz 2 Pa H 30% 2

0.1

0.01

Normalizedintensity

1E-3

0 10 20 30 40 50 Energy, eV

Figure 5.8. NIEDF measured on HOPG in H2 RF plasma at RT for constant DC-bias (black curve) and 90 kHz pulsed-bias (dark blue curve) compared to the modeling with H atom surface coverage of 30% (black symbols).

1 V = -130 V s 30 kHz P = 20 W 35% 2 Pa H 2

0.1

0.01

Normalizedintensity

1E-3

0 10 20 30 40 50 Energy, eV

Figure 5.9. NIEDF measured on HOPG in H2 RF plasma at RT for 30 kHz pulsed-bias (solid curve) compared to the modeling with H atom surface coverage of 35% (symbols). 160

5. Pulsed-bias approach

bias OFF period is sufficiently long. That is probably why the NI signal is frequency- independent until 1 kHz. On the other hand, as we approach to the continuous bias situation while increasing the frequency, we leave less time for H atoms to come back to the surface. So the coverage gets smaller until it reaches the continuous DC-bias value. As a summary, the pulsed-bias tests performed on HOPG have proven that:

 as expected, no charging effects were observed on the conductive surface of HOPG. Surprisingly, the yields in pulsed mode and continuous mode were different  by changing the pulsed-bias frequency (and the duty cycle) one can obtain HOPG material with different hydrogen surface coverage and hence a different surface state. At frequency below 1kHz, the hydrogen coverage is higher than in continuous mode, leading to a NI yield increase by 25%

However, let us note that the pulsed bias method does not represent a way to increase the production of NI since we measure the NI yield only during the pulse ON period. The total production yield in pulsed mode is the product of the yield during the pulse ON period by the duty cycle (assuming no production during the OFF period), which is well below the yield in continuous mode.

1 V = -130 V s 5 kHz P = 20 W 40% 2 Pa H 2

0.1

0.01

Normalizedintensity

1E-3

0 10 20 30 40 50 Energy, eV

Figure 5.10. NIEDF measured on HOPG in H2 RF plasma at RT for 5 kHz pulsed-bias (solid curve) compared to the modeling with H atom surface coverage of 40% (symbols).

161

5. Pulsed-bias approach

In the end, it remains unclear why the NI yield for f > 20 kHz is smaller than for constant DC-bias. As could be seen in Figure 5.6, it is not connected to the MS acquisition issues. Figure 5.7 demonstrates that for f > 30 kHz the hydrogen coverage is nearly the same as for constant DC-bias. Therefore, there must be another explanation, probably based on the HOPG surface state. Further studies are required to conclude on that. 5.2.2. Results for an insulating material: MCD

a) Charge accumulation on the surface After proving the feasibility of the pulsed-bias approach on HOPG, the tests on the insulating surface of MCD could be performed. As charging effects had to be taken into account, the charge accumulation on the surface during the bias pulse was explored. The idea was to make a long pulse (100 μs) and measure NIEDF with short acquisition time

(Tacq = 10 μs) at different delay (Tdelay). The scheme demonstrating the MS acquisition periods with respect to the bias pulse is depicted in Figure 5.11. The color of each acquisition window corresponds to the color of measured NIEDF shown in Figure 5.12. A low bias frequency of 1 kHz has been chosen for the measurements. All the raw data from the MS have been processed as if the surface bias was equal to

Vs = –130 V. Typically, Va = –140 V was applied in order to obtain Vs = –130 V on the insulating sample during the bias pulse. The onset of the first NIEDF (Tdelay = 13 µs) is at 0 on the energy scale indicating that for this measurement the surface bias was indeed equal to –130 V. For the other measurements, one can observe a shift of the NIEDF onset indicating a decrease of the surface bias (refer to Eq. (1.4) for details). This shift is equal to the difference between the actual surface bias and Vs = –130 V initially present on the surface. Therefore, the rising slope in the beginning of NIEDF served as an indication of bias voltage Vs present on the surface at a given moment. In such a way, the charging of the sample was followed in real time as demonstrated in Figure 5.13.

Figure 5.11. Scheme demonstrating the MS acquisition periods at a given moment of the bias pulse. The pulse duration is Tpulse = 100 μs and the acquisition duration is Tacq = 10 μs. The color of each acquisition window corresponds to the color of measured NIEDF on next figure. A schematic representation of the applied voltage Va and actual bias on the MCD surface Vs is shown as red and blue lines.

162

5. Pulsed-bias approach

4 T of: 10 Bias for 100 s delay acquisition for 10 s 13 s at 1 kHz rate 23 s 33 s 43 s 53 s 103 63 s 73 s

83 s

93 s

c/s Intensity, 103 s

102

-15 -10 -5 0 5 10 15 20 25 30 35 40 45 50 55 Energy, eV

Figure 5.12. NIEDF measured at a given moment of the bias pulse. The pulse duration is Tpulse = 100 μs and the acquisition duration is Tacq = 10 μs. Tdelay is varied with a step of 10 μs from the beginning until the end of the pulse. Bias frequency is 1 kHz. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen.

-119 Bias for: -120 100 s -121 linear fit -122 Acquisition for 10 s at 1 kHz rate -123

-124

-125

-126 Equation y = a + b*x Weight No Weighting

Bias voltage, V Residual Sum of 1.16533 -127 Squares Pearson's r 0.99292 Adj. R-Square 0.98412 -128 Value Standard Error ?$OP:A=1 Intercept -130.55133 0.27196 -129 ?$OP:A=1 Slope 0.09933 0.0042 -130 0 10 20 30 40 50 60 70 80 90 100 110 Tdelay, s

Figure 5.13. Bias voltage Vs on MCD surface at each moment of 100 μs bias pulse (defined by Tdelay) sown in red symbols with a linear curve fit (blue line). Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen. 163

5. Pulsed-bias approach

One could conclude that charging of MCD is happening as expected in a linear manner with a rate of ~1V/10 μs: see Eq. (5.4). The same measurements were performed for two other cases: with a very long bias pulse (Tpulse = 300 μs, Tacq = 10 μs and f = 1 kHz) and a very short one (Tpulse = 20 μs, Tacq = 2 μs and f = 10 kHz). The charging rate has shown to be linear and practically independent of the bias duration, as it was expected.

The definition of the optimal Tpulse and Tacq is useful for the following reasons. First of all, the charging of the sample during the pulse decreases the value of bias voltage Vs present on the surface. Therefore, Tacq should be sufficiently short to have undistorted NIEDF form which could be compared to the measurements for constant DC-bias case.

The acquisition time of Tacq = 10 μs seems reasonable to allow for relatively fast acquisition and provide a quasi-constant surface bias Vs (Vs decreases only by 1 V over

130 V during the measurement). Secondly, Tpulse should be short enough to allow the surface to discharge during the pulse-off period.

b) Influence of pulse duration at constant frequency

In order to get information about the surface discharge, the study of the bias duration influence on the shape of NIEDF was performed and is presented in Figure 5.14. The duration of acquisition was kept constant Tacq = 5 μs with the same delay from the beginning of the pulse Tdelay = 11.8 μs. From the NIEDF onset shift of Figure 5.14 a, one can deduce the surface bias which is plotted in three different manners (Figure 5.14 b, c, d), as a function of Tpulse (Figure 5.14 b), duty cycle (Tpulse/T) (Figure 5.14 c) and Toff

(T-Tpulse/T) (Figure 5.14 d). One must note here that the variations of the NIEDF intensity with the bias duration are not of interest. Indeed, the MS is set to optimize the transmission of ions created at Vs = –130 V. However, when the bias duration increases, the surface bias is no more –130 V (see hereafter) and the transmission of ions is no more optimized. As one can notice from Figure 5.14 b, the surface voltage is almost constant with

Tpulse up to the bias duration of 100 μs (duty cycle of 10%) or for pulse-off period Toff

>900 μs. After that, the surface voltage varies fast with Tpulse, indicating that the charge on the surface has not been eliminated between two pulses. Indeed, if the OFF period is not long enough, there is a remaining charge on the surface at the beginning of next pulse and the surface bias is no more equal to the applied bias plus the undisturbed floating potential:

Vs = Va + Q/C + Ii/C Eq. (5.6)

To explain the behavior of Vs shift with bias pulse duration, one can refer to the scheme shown in Figure 5.15. In the first case (corresponding to Tpulse = 50 μs) the charge accumulated during the bias pulse leads to a surface bias of ~ 5V when the applied bias is switched off. This value is close to the floating potential, therefore, no significant perturbation of the plasma is induced at the end of the pulse and the

164

5. Pulsed-bias approach

Bias duration: Frequency 1 kHz -80 5 s (b) Acquisition 5 s 4 10 s 10 Delay 11.8 s 50 s -90 100 s (a) 250 s 500 s -100 750 s 3 10 800 s

-110

Intensity, c/s Bias voltage, V -120

2 10 -130

-30 -20 -10 0 10 20 30 40 50 0 100 200 300 400 500 600 700 800 900 T , s Energy, eV pulse -80 (c) -80 (d)

-90 -90

-100 -100

-110 -110

Bias voltage, V -120 -120

-130 -130

0 10 20 30 40 50 60 70 80 90 0 100 200 300 400 500 600 700 800 900 1000 Toff, s Duty cycle, %

Figure 5.14. Influence of bias pulse duration on NI production for MCD surface at RT in H2 plasma: f = 1 kHz, Tacq = 5 μs and Tdelay = 11.8 μs (a) NIEDF measured for Tpulse from 5 μs to 800 μs; (b) Vs shift as a function of Tpulse; (c) Vs shift as a function of duty cycle; (d) Vs shift as a function of Toff. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = –140 V, with screen.

Figure 5.15 Scheme representing the charging and discharging of an insulating sample for two cases: Tpulse = 50 μs (slow discharge) and 500 μs (quick + slow discharge). The applied voltage is Va = –140 V; the change of Vs from the first to the second pulse is indicated for Tpulse = 500 μs. 165

5. Pulsed-bias approach

discharge of the surface is slow. In another case (for Tpulse = 500 μs) the accumulated charge leads to a surface bias of ~ 50V when the applied bias is switched off. When the pulse is over, the surface appears to have a large positive bias Vs > Vp. Electrons are attracted resulting in a high loss of electrons and a consecutive increase of the plasma potential. This results in a quick initial unload of the surface first (electrons are attracted) and then (for Vs ≤ Vp) the unloading happens in a usual slow regime, as illustrated in Figure 5.15. This reasoning is supported by positive ion (PI) energy distribution function measurements for the two indicated bias pulse durations. Figure 5.16 shows PI energy distribution for the case of Tpulse = 50 μs with Tacq = 5 μs and various values of Tdelay. PIEDF for no applied bias is shown in black symbols for the dominant ion population in these conditions (H3+). In order to arrive to the MS detector, H3+ need ~ 19 μs. Therefore, the red curve corresponding to Tdelay = 0 μs is the PIEDF just before the bias pulse is applied. As expected, it coincides well with the curve for no bias, meaning that no effect from the previous pulse is present in the plasma. The blue curve represents the situation for Tdelay = 70 μs and demonstrates that immediately after the bias pulse, the PIEDF is identical to the one before the pulse or in the absence of bias. It shows no disturbance of the plasma potential by the accumulated charge on the surface.

The situation for the long bias pulse is shown in Figure 5.17 for Tpulse = 500 μs. In this case, the charge accumulated on the insulating surface is ~ 50 V, so Toff = 500 μs appears to be too short to unload it completely. PIEDF immediately before the pulse (Tdelay = 0 µs) is not strictly identical to the one with no bias. The measurements after the pulse (Tdelay

> 520 µs) show that some high-energy H3+ are present in the PIEDF, demonstrating that the distribution of plasma potential has been disturbed. Most probably, the plasma potential close to the sample is increased in order to shield the positive charge and limit the loss of electrons. This high energy ion component of the PIEDF slowly decreases in the OFF period, demonstrating a slow decrease of the surface load accompanied by a slow decrease of the plasma potential. This study has demonstrated the complex surface charge dynamics during the OFF period, and possible plasma potential disturbances due to the accumulated charge on the surface. However, these effects have very limited influence on NIEDF measurements during the pulse ON period. Indeed, for NIEDF measurements, a short pulse ON period is chosen in order to have an almost constant surface bias during the acquisition. Then, even if the OFF period is not long enough to eliminate the accumulated charge on the surface, one can increase the applied bias Va to obtain the desired bias on the surface at the beginning of the pulse (Vsurf = Va + Q/C at t = 0, with │Va│>│Vdesired│). The disturbance of the plasma potential during the OFF period is not an issue as long as the PI flux during the bias ON period is equal to the one in the DC bias situation.

166

5. Pulsed-bias approach

Frequency 1 kHz no bias 5 Bias 50 s Delay: 10 Acquisition 5 s 0 s 70 s

104

3 10 Intensity, cts/s

2 10

0 5 10 15 20 25 30 35 40 45 Energy, eV Figure 5.16. PIEDF of H3+ with Tpulse = 50 μs and Tacq = 5 μs for various values of Tdelay during the bias pulse (see inset). PI distribution for no bias applied is shown in black symbols. Plasma parameters: 2.0Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen.

Frequency 1 kHz no bias Bias 500 s Delay: 5 10 Acquisition 5 s 0 s

560 s

615 s

715 s 4 10 815 s 915 s

950 s

10 Intensity, cts/s

2 10

0 5 10 15 20 25 30 35 40 45 Energy, eV

Figure 5.17. PIEDF of H3+ with Tpulse = 500 μs and Tacq = 5 μs for various values of Tdelay during the bias pulse (see inset). Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen. 167

5. Pulsed-bias approach

Figure 5.18 illustrates the comparison between PIEDF for constant and pulsed bias techniques. When biasing the sample with constant DC voltage at Vs = –130 V, PI are attracted towards the sample holder and a depleted region of plasma is created in the vicinity of the MS. This results in a decrease of plasma potential from 13 V to 3 V, and the quantity of detected PI also decreases (compare blue and orange solid curves in Figure 5.18). The same situation is observed with pulsed bias technique for equivalent conditions (pulse duration Tpulse = 500 μs was chosen for comparison). Just before the pulse (Tdelay = 0 μs, black dotted curve in Figure 5.18), PIEDF resembles the one with no bias applied, if one neglects the certain number of high energy PI. During the pulse ON period (Tdelay = 510 μs, red dotted curve in Figure 5.18), attraction of PI towards the sample holder has a similar effect on PIEDF as for constant bias case (compare red and orange curves). Note that the signal-to-noise ratio for pulsed bias measurements is lower due to a smaller number of averaging. One can conclude that PI energy and flux during the bias ON period in pulsed bias technique are approximately equal to the ones in the DC bias situation, so the comparison between NIEDF in pulsed and constant DC bias modes can be done without any restrictions.

no bias

5 const bias 10 at V = -130 V s

Pulsed bias 4 f = 1 kHz; V = -130 V 10 s

T = 500 s; T = 5 s bias acq T = 0 s delay (just before the pulse) 103 T = 510 s

delay Intensity, cts/s (during the pulse)

102

0 5 10 15 20 25 30 35 40 45

Energy, eV

Figure 5.18. PIEDF of H3+. Solid curves: for no bias (blue), constant DC bias at Vs = –130 V (orange). Dotted curves: pulsed bias at Vs = –130 V with Tpulse = 500 μs and Tacq = 5 μs for two values of Tdelay: 0 μs (just before the bias pulse, black curve); 510 μs (during the bias pulse, red curve).

168

5. Pulsed-bias approach

c) Influence of pulse frequency at constant bias duration

The surface charging is visible when the bias duration is too long, but also when the bias frequency is too high. Figure 5.19 shows the measurements with an applied bias

Va = -140 V and for Tpulse = 10 μs at different frequencies from 1 to 50 kHz which gives the duty cycle from 1% to 50% and Toff range from 991 to 11 μs. The acquisition was done for

Tacq = 9 μs and the delay was Tdelay = 12.5 μs. The value of the surface bias Vs was extracted from the graph and plotted versus the bias frequency, duty cycle and Toff. The lower is the bias frequency (and the smaller is the duty cycle), less is the Vs shift. The variation of surface bias with frequency and the duty cycle is linear, whereas the dependence on Toff is not. It varies fast at short Toff and then hardly varies for Toff higher than 200 µs. For high frequencies the MCD sample cannot discharge completely during

Toff because the bias arrives too quickly. For Toff > 100–200 μs the bias voltage drop is not significant. It means that for the bias duration used in the experiment (Tpulse =10 μs), the frequency should not be higher than 5–10 kHz to get a surface bias of –130 V with an applied bias of –140 V. For higher frequency, the applied bias has to be increased to get the appropriate surface bias. One can note in Figure 5.19 that the NI intensity decreases as the pulse bias frequency grows. It cannot be interpreted easily, because the MS autotune is not optimized anymore for the correct NI energy. In order to prevent the signal loss resulting from unsuitable autotune, the following experiment was performed as depicted in Figure 5.20: the pulse bias frequency was varied from 250 Hz to 90 kHz while increasing the applied bias Va in order to have Vs = -130 V on the sample surface. In such a way, the situation was comparable to the use of a conducting sample as the surface bias remained at desired value. As can be seen from Figure 5.20, the tails of measured NIEDF superimpose up to f = 50 kHz and only the distribution maximum decreases. This indicates the decrease of the sputtering contribution to the NI production resulting from the hydrogen surface coverage decrease. As the coverage depends on the bias OFF time (which allows hydrogen atoms to re-cover the surface) the higher bias frequency shortens Toff and decreases the coverage down to the value for constant DC bias. However, it was observed for HOPG that the NI yield for f > 20 kHz is smaller than for constant DC bias. It is suggested that coverage decreases below the DC-bias value due to the changes of the surface state induced by the pulsed bias at such high frequencies. Moreover, for f > 50 kHz the NIEDF shape for MCD clearly deviates from the usual, the tail ends at lower energies and there is no more peak at low energy, reinforcing the idea of the low hydrogen coverage. However, one must keep in mind that the incoming PI flux and energy might be affected by the change of the surface state for short Toff. Therefore, further investigations are needed in order to understand the behavior of MCD at high frequency.

169

5. Pulsed-bias approach

As can be seen in Figure 5.21, the NI yield for MCD surface (shown in blue triangles) decreases with pulsed bias frequency. The results for HOPG (previously demonstrated in Figure 5.6) are plotted on the same graph in black squares normalized to correspond to the NI yield for MCD at f = 1 kHz. One could notice that starting from f = 10 kHz the signal decrease for MCD is more important than for HOPG. On the other hand, the NI yield saturation observed on HOPG for low frequencies (f = 0.25 kHz to 1 kHz) as shown in Figure 5.6 is not present in the case of MCD for which the NI yield always grows when bias frequency decreases.

The orange stars depicted in Figure 5.21 show the applied surface bias Va which was

adjusted for each measurement to maintain Vs = -130 V on the sample surface. Below f =

10 kHz no adjustment was necessary, Va = -140 V was applied. For f > 10 kHz the applied voltage had to be increased additionally and the NI yield diminution for MCD became

more important than for HOPG. This suggests that Va compensation leads to experimental complication for high frequencies, so the operation at f > 10 kHz should be

-108 Bias 10 s 104 -110 (b) Acquisition 9 s -112 Delay 12.5 s -114 (a) -116

3 10 -118

-120

-122

-124 Bias voltage, V Intensity, c/s 102 -126 -128 -130 -132 -20 -10 0 10 20 30 40 50 0 10k 20k 30k 40k 50k Energy, eV Frequency, Hz -108 -108 -110 1 kHz - 1% - 991 s off -110 (d) 5 kHz - 5% - 191 s off -112 -112 10 kHz - 10% - 91 s off -114 20 kHz - 20% - 41 s off -114 -116 30 kHz - 30% - 24 s off -116 40 kHz - 40% - 16 s off -118 -118 50 kHz - 50% - 11 s off

-120 -120

Bias voltage, V -124 -124 -126 -126 -128 -128 -130 -130 -132 -132 0 5 10 15 20 25 30 35 40 45 50 0 100 200 300 400 500 600 700 800 900 1000 Duty cycle, % Toff, s

Figure 5.19. Influence of bias pulse frequency on NI production for MCD surface at RT in H2 plasma: Tpulse = 10 μs, Tacq = 9 μs and Tdelay = 12.5 μs (a) NIEDF measured for f = 1 kHz to 50 kHz; (b) shift of bias voltage as a function of f; (c) Vs shift as a function of duty cycle; (d) Vs shift as a function of Toff. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen. 170

5. Pulsed-bias approach

different Va to get Vs = -130 V 5 250 Hz 10 P = 20 W 1 kHz 2 Pa H2 5 kHz 10 kHz Bias: 10 s 4 30 kHz 10 Acq: 8 s 50 kHz Delay: 13 s 60 kHz 70 kHz 103 80 kHz Intensity, c/s Intensity, 90 kHz

2 10

0 10 20 30 40 50 60 Energy, eV

Figure 5.20. NIEDF measured for f = 0.25 kHz to 90 kHz (see inset) on MCD surface at RT in H2 plasma: Tpulse = 10 μs, Tacq = 8 μs and Tdelay = 13 μs. The applied surface bias Va is adjusted to maintain Vs = -130 V on the sample surface. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen. -130 1.2x106 -140

6 1.0x10 -150

5 -160 8.0x10 -170 6.0x105 -180 4.0x105

NI Yield,arb.u. NI -190

AppliedV voltage, 5 -200 2.0x10 HOPG pulsed

MCD pulsed -210 0.0 V to have V =-130V for pulsed MCD a s -220

0.1 1 f, kHz 10 100 Figure 5.21. Left Y-scale: NI yield dependence for HOPG (black squares) and MCD (blue triangles) normalized by the value for MCD at f = 1 kHz, in H2 RF plasma, as a function of the pulsed bias frequency (varied from 0.25 to 90 kHz). Pulsed bias parameters: Tpulse = 10 μs, Tacq = 8 μs, Tdelay = 13 μs. Plasma parameters: 2.0 Pa of H2, Q = 5.2 sccm, PRF = 20 W, Va = -140 V, with screen. Right Y-scale: the applied surface bias Va necessary to maintain Vs = -130 V on the surface (orange stars). 171

5. Pulsed-bias approach

avoided if possible in order to maintain the efficiency of the pulsed bias method. As the goal of this work was to prove the feasibility of pulsed bias for insulating materials, we did not investigate the situation at high frequency which is not of immediate interest. This should be done in the future, taking into account the complex dynamics of the plasma potential which probably has an influence at high frequency. 5.2.3. Results for heated materials

After performing the optimization of the bias and acquisition parameters on MCD, the chosen approach had to be tailored to suit the temperature evolution measurements. The experiment was performed with a conventional sample holder which typically provides heating up to 800°C with constant DC bias at the same time. Nevertheless, the simultaneous pulsed biasing and heating of the sample was not possible, because the temperature controller operation was perturbed by the pulsed bias. Therefore, the following protocol has been used for heating experiments:

Heating without bias to the temperature slightly higher than the desired value Tdes.

For Tdes = 100°C the initial temperature was Tinit = 105°C, for Tdes = 800°C the initial temperature was equal to the maximum possible heating temperature (~840°C). For the rest of Tdes, the initial temperature was defined as follows: Tinit = Tdes + Tdes · 10% –

10°C. Refer to Figure 5.22 for the dependence of Tinit on Tdes for this experiment (left Y- scale).

1. Waiting for ~5 min to stabilize the surface temperature and the surface state at

temperature Tinit. 2. Turn off the temperature controller and start the pulsed bias and the MS acquisition immediately. The bias and acquisition parameters were: 1 sweep,

V = [–80 V, 0 V], f = 10 kHz, Tpulse = 15 μs, Tacq = 10 μs and Tdelay = 17μs. During

this acquisition sweep, temperature will decrease by ΔT = (Tinit – Tfinal) in the

absence of heating. The value of Tdes should fall approximately in the middle of

ΔT interval; see Figure 5.22 for the dependence of ΔT on Tdes for this experiment (right Y-scale). 3. Observation of the position of the NIEDF onset. Indeed, it has been observed that the NIEDF onset was shifted due to high frequency applied and due to the

surface conductivity change induced by the heating. Adjust Va to maintain

Vs = –130 V on the surface.

4. Repeat steps 1–3 to produce the correct NIEDF for Tdes. The step in

temperature ΔTstep = 100°C was typically used during the whole series of measurements.

As a starting point, NIEDF measured at RT were compared for both materials (see

Figure 5.23) in D2 plasma. Being twice heavier as comparing to hydrogen, D– needed

172

5. Pulsed-bias approach 900 T 250 800 init

700 200 600

150

500

T, °C T,  °C ,

init 400 T 100 300 50 200

100 T = T - T init final 0 0 0 100 200 300 400 500 600 700 800 900 T , °C des Figure 5.22. Left Y-scale: dependence of initial heating temperature Tinit as a function of the desired one Tdes for the pulsed bias heating measurements (black squares). Right Y-scale: dependence of temperature variation during the pulsed bias heating measurements ΔT = (Tinit – Tfinal) on Tdes (red dots).

constant DC bias pulsed-bias 5 HOPG f =10 kHz; T = 15 s 10 pulse MCBDD T = 10 s; T = 17 s acq delay HOPG MCD 4 10

D plasma 2 V = -130V s 103

Intensity, cts/s

2 10

0 5 10 15 20 25 30 35 40 45 Energy, eV

Figure 5.23. Comparison of NIEDF measured for constant and pulsed bias (solid and dotted curves correspondingly) on HOPG, MCBDD and MCD at RT. Plasma parameters: 2.0 Pa of D2 RF plasma, Q = 7.6 sccm, PRF = 20 W, with screen. Pulsed bias parameters: Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, f = 10 kHz, Vs = –130 V, VMS = 0 V. 173

5. Pulsed-bias approach

~16 μs to arrive to the MS detector, so Tdelay = 17 μs was used during the pulsed bias measurements in D2 plasma. Conventional sample holder had a better alignment as compared to the provisional one and the signal levels have noticeably increased. In any case, the comparison of NIEDF for constant and pulsed bias at RT shown in Figure 5.23 (see solid and dotted curves correspondingly) reveals that the distribution maximum has increased by nearly 2 times for both HOPG and pair MCBDD/MCD. As it was noticed that the NI signal evolves with time, the samples were left under bias for 20- 30 minutes in order to reach the stable surface state before the measurements.

The change of the NIEDF with the surface temperature in D2 RF plasma on pulse- biased HOPG is shown in Figure 5.24: color distribution from dark blue to red corresponds to the surface temperature of the sample. One could notice the decrease of the NI signal with temperature which is similar to the constant DC-bias case previously demonstrated in Figure 3.3. The signal-to-noise ratio is quite low for T>200°C, because all the measurements were performed with 1 sweep (no averaging). This problem was not present in the case of MCD (see Figure 5.25) as the measured NI signal levels were very high and already 1 sweep provided relatively good statistics. The behavior of pulsed bias MCD resembles much that of MCBDD with constant DC-bias: the NI production increases with temperature starting from 200°C, reaches the maximum at 400°C and then starts to decrease. To compare the NI signal increase for the diamond films, one should refer to Figure

5.26 where the change of NI yield with surface temperature in D2 RF plasma is shown for both constant (solid symbols) and pulsed bias (empty symbols). The x-error bars are drawn for the pulsed bias case resulting from the uncertainty in temperature estimation introduced by the employed measurement technique. As could be seen from the graph, the NI yield for HOPG and pair MCBDD/MCD has increased by ~2 times in the case of pulsed bias. At high temperature, a ~4 times higher NI yield was measured on MCD surface in the case of pulsed bias as compared to constant DC bias. This is the highest NI yield to be ever measured on PHISIS set-up which leads us to a conclusion that there is still room for optimization of the NI yield on carbon materials in terms of finding a suitable surface state. The signal increase by a factor 5 as previously seen on MCBDD in

H2 RF plasma [47] with constant DC bias has been extended to ~10 times when comparing constant biased MCBDD at RT with pulse-biased MCD at 400°C in D2 plasma. Some information about the surface state of the materials for pulsed bias conditions was given by time evolution measurements presented in Figure 5.27 for HOPG and MCD. Similarly to the results for constant bias, the signal on freshly cleaved HOPG increases within ~10 min of plasma exposure under pulsed bias of Vs = –130 V (see black hollow squares). As one could conclude, the defects produced by plasma on the sample surface act favorably on the NI conversion on HOPG, as already observed for constant bias.

174

5. Pulsed-bias approach

V = - 130 V RT 105 s HOPG pulsed 100°C 2 Pa D 20 W 200°C 2 300°C 400°C

500°C 4 10 800°C

3 Intensity, cts/s Intensity, 10

102 0 5 10 15 20 25 30 35 40

Energy (eV) Figure 5.24. NIEDF measured on HOPG as a function of sample surface temperature for pulsed bias. Plasma parameters: 2.0 Pa of D2 RF plasma, Q = 7.6 sccm, PRF = 20 W, with screen. Pulsed bias parameters: Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, f = 10 kHz, Vs = –130 V, VMS = 0 V. 6 10 RT Vs = - 130 V MCD pulsed 100°C 200°C 5 2 Pa D 20 W 10 2 300°C

400°C 700°C 800°C 104

Intensity, cts/s Intensity, 3 10

102 0 5 10 15 20 25 30 35 40 Energy (eV) Figure 5.25. NIEDF measured on HOPG as a function of sample surface temperature for pulsed bias. Plasma parameters: 2.0 Pa of D2 RF plasma, Q = 7.6 sccm, PRF = 20 W, with screen. Pulsed bias parameters: Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, f = 10 kHz, Vs = –130 V, VMS = 0 V. 175

5. Pulsed-bias approach

6 10

5 10 Continuous bias Vs: HOPG Yield,arb.u. NI MCBDD MCD Pulsed bias V : s 104 HOPG MCD 0 100 200 300 400 500 600 700 800

Temperature, °C Figure 5.26. NI yield dependence on the surface temperature for HOPG, MCBDD and MCD for constant bias (solid symbols) and pulsed bias (empty symbols). Plasma parameters: 2.0 Pa of D2 RF plasma, Q = 7.6 sccm, PRF = 20 W, with screen. Pulsed bias parameters: Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, f = 10 kHz, Vs = –130 V. 1.6x106

D 2 Pa 20 W HOPG virgin 2 6 MCD after heating cycle 1.4x10 Vs = -130 V Pulsed bias at RT 1.2x106 f = 10 kHz Tbias = 15 s 6 1.0x10 Tacq = 10 s T = 17 s

delay 8.0x105

5 Intensity, cts/s 6.0x10

5 4.0x10

5 2.0x10 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34

Time, min Figure 5.27. Time evolution of NI yield on pulse-biased HOPG and MCD at RT. Plasma parameters: 2.0 Pa of D2 RF plasma, Q = 7.6 sccm, PRF = 20 W, with screen. Pulsed bias parameters: Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, f = 10 kHz, Vs = –130 V. 176

5. Pulsed-bias approach

The measurements demonstrated in Figure 5.27 by blue hollow triangles were performed on MCD at RT the next morning after the series of temperature measurements. Note that the sample was cooled down and stayed overnight at vacuum. Therefore, the optimized surface state has been preserved, and the degradation started only during the plasma exposure. As could be concluded from the graph, it takes MCD ~15 min to reach the stable surface state under plasma exposure at pulsed bias. The results for both materials resemble much the ones presented in Figure 1.9 for constant bias, with only major difference being a 2-3 times longer time which it takes to reach the stable state for pulse-biased samples. However, taking into account the duty cycle of 15% for pulsed bias and 100% for constant DC bias, the stationary surface state is reached faster in the pulsed bias case. It is another indication that a different surface state is reached in the pulsed bias mode. The defects produced by plasma on the sample surface act favorably on the NI conversion on HOPG, but have an opposite effect on diamond films. When biasing the sample in pulsed mode, the defects could be induced on the MCD surface only during the bias ON period. Therefore, the surface state is closer to the pristine diamond state. The results of time evolution studies bring us to the conclusion that in pulsed bias case the diamond surface is less degraded and more hydrogenated, which is favorable for NI surface production. This situation is similar to biasing a fresh diamond sample with constant DC bias for a very short exposure time. Most probably, under ion bombardment, diamond samples lose their attractive electronic properties. With a very short exposure time, or using the pulsed bias technique, it is possible to maintain electronic properties close to the pristine diamond ones, and obtain higher NI yields. As already mentioned, it indicates that there is still room for NI yield optimization on carbon materials. Future studies will have to focus on less degraded surfaces by using, for instance, a lower bias voltage for both constant and pulsed bias techniques. It would be also interesting to trace the evolution of the total NI energy and angular distributions with time and temperature to check if the results are the same as for normal incidence (α = 0°). These investigations should be pursued in the future.

177

5. Pulsed-bias approach

5.3. Conclusion and perspectives The method of pulsed bias was developed to enable the study of NI production on surfaces of insulating materials such as MCD. The present technique has enabled to study NI surface production on MCD for the whole temperature range starting from RT. The pulsed-bias tests were performed on HOPG to demonstrate the feasibility of the method. By changing the pulsed-bias frequency (and the duty cycle) it was possible to obtain HOPG material with different hydrogen surface coverage and hence a different surface state. This new surface state with higher H2/D2 surface coverage achieved in pulsed bias mode has resulted in the NI yield increase up to 30-50% depending on experimental conditions. The change of the surface coverage was also confirmed by the modeling and explained by the presence of the bias-off period which has allowed neutral atoms to re-cover the sample surface. After proving the feasibility of the pulsed bias approach on HOPG, the optimization of the experimental parameters was performed on MCD by taking into account the charging effects. These parameters were: bias pulse duration Tpulse, MS acquisition duration Tacq, delay between the start of the pulse and MS acquisition Tdelay, bias frequency f, applied bias Va and the resulting surface bias Vs. The charge accumulation on the surface during the bias pulse was explored by making a long pulse and measuring NIEDF with short Tacq at different Tdelay (from the beginning until the end of the pulse). One could conclude that charging of our MCD is happening in a linear manner with a rate of ~1V/10 μs. In order to define the optimal experimental parameters, the influence of the bias pulse duration Tpulse on the shape of

NIEDF was studied. The acquisition time of Tacq = 10 μs was chosen to allow for relatively fast acquisition and provide the Vs shift of ~1V. At 1 kHz, the pulse duration Tpulse = 15 μs

(Toff = 985 μs) has proven to let the surface discharge completely from the accumulated

PI charge. The surface bias Vs has shown to depend non-linearly on the bias pulse duration Tpulse. This behavior was explained by the perturbation exerted by the accumulated surface charge on the surface potential. Two discharge regimes during the pulse-off period were identified (quick and slow). This reasoning has been confirmed by the time-resolved PI measurements for two bias pulse durations: 50 μs and 500 μs. The effect of surface charging visible as a shift of NIEDF could appear when the bias duration is too long, but also when the bias frequency is too high. The influence of the bias frequency has been studied in the range from 1 to 50 kHz. It could be noted that for high frequencies the MCD sample cannot discharge completely during Toff, because the bias arrives too quickly. For the bias duration used in the experiment (Tpulse = 10 μs), the frequency must not exceed 10 kHz to avoid the surface bias shift. If compensation is used to counteract the Vs shift induced by high bias frequency (applying an increased Va value), the problem of insufficient bias could be solved, but only at frequency below 10 kHz. Above 10 kHz NIEDF shape changes a lot and NIEDF intensity strongly decreases. The origin of this has not been identified. Therefore, the operation at f > 10 kHz should be avoided, if possible. 178

5. Pulsed-bias approach

After performing the optimization of the bias and acquisition parameters on MCD, the chosen approach had to be tailored to suit the temperature evolution measurements. A special experimental protocol has been established, specified in Section 5.2.3. Results for heated materials. The chosen bias and acquisition parameters are: 1 sweep,

V = [–80 V, 0 V], f = 10 kHz, Tpulse = 15 μs, Tacq = 10 μs, Tdelay = 17 μs, ΔTstep = 100°C. The comparison of NIEDF for constant and pulsed bias at RT revealed that the NI yield increased by nearly 2 times for both HOPG and pair MCBDD/MCD. At high temperature, a ~ 4 times higher NI yield was measured on MCD surface in the case of pulsed bias as compared to constant DC bias. This is the highest NI yield to be ever measured on PHISIS set-up which leads us to a conclusion that there is still room for optimization of the NI yield on carbon materials in terms of finding a suitable surface state. The results of NI yield time evolution measurements for HOPG and MCD are similar to the ones for constant DC bias case (for HOPG and MCBDD). The only major difference was a 2-3 times longer time which it takes to reach the stable state for pulse- biased samples (~15 min). The defects produced by plasma on the sample surface act favorably on the NI conversion on HOPG, but have an opposite effect on diamond films. When biasing the sample in pulsed mode, the defects could be induced on the MCD surface only during the bias ON period. Therefore, the surface state is closer to the pristine diamond state. The results of time evolution studies bring us to the conclusion that in pulsed bias case the diamond surface is less degraded and more hydrogenated, which is favorable for NI surface production. Considering the overall result of the pulsed bias measurements, one can conclude that to optimize the NI yield on diamond, one has to work with a less degraded surface. This can be obtained rising the surface temperature to 400°C–500°C which allows restoring intrinsic properties of diamond probably due to the enhanced etching of sp2 phases [50, 91]. The less degraded surface state can also be obtained by applying the pulsed bias which gives the possibility to increase the H2/D2 surface coverage and diminish the defects induced by plasma exposure. However, pulsing the bias is a solution only for fundamental studies. It is not a solution that could be used in a real NI source since the absolute NI yield is low (it is equal to the NI yield measured during the pulse period times the duty cycle). On the contrary, the reduction of defects could as well be done by applying a smaller surface bias which is also approaching the conditions of the real NI source designed for ITER. As the extraction of the NI also depends on the surface bias value, a complete study has to be performed in the future.

179

General conclusion

This thesis makes an important contribution in understanding of NI surface production in H2/D2 plasmas. The research work could be summarized in several sections: modeling and reconstruction of surface-produced NI energy and angular distribution functions (Chapter 2. Modeling and reconstruction of NIEDF); study of NI surface production by means of mass spectrometry on various groups of materials: carbons, metals, insulators (Chapters 1. Negative ion surface production measurements, 3. NI production on different materials and 5. Pulsed-bias approach); surface state characterization of NI enhancers by external diagnostics (Chapter 4. Surface state characterization of NI enhancers). Modeling of NIEDF performed previously has shown remarkable agreement with experiment for HOPG confirmed that the NI ionization probability Piz(E,θ) is constant independently of the neutral particle energy and angle of emission. This has also verified the choice of SRIM to provide the correct initial distribution f (E,θ) for carbon materials. The reconstruction method developed in the course of this thesis has allowed to determine the distribution in energy and angle of NI emitted from the surface in case when the NI formation probability is not constant or when the SRIM parameters are unknown, so one cannot obtain an initial guess of f (E,θ). The reconstruction method was validated by the good agreement of SRIM calculations on carbon with the distributions reconstructed from experimental data for HOPG on several levels. After the verification, the reconstruction method was used to characterize NI production on the surface of low work-function metals (on the example of Gd) which has given an unexpectedly good agreement with SRIM calculations. The comparison with SRIM calculations, even though it has to be done with much care, reinforces our confidence in the reconstruction method. One should mention that the reconstruction algorithm does not depend on the NI surface production mechanism. The only input which is necessary to calculate the ion trajectories are the parameters of the plasma and the sheaths. Therefore, the reconstruction method can be applied to any type of surface and/or NI. It is planned to test it on other low work-function metals and insulators. The experimental protocol for NI measurements by means of mass spectrometry on PHISIS set-up was established for various types of experiments. The chosen experimental conditions in RF plasmas were: 2.0 Pa H2/D2 pressure with the injected RF power of 20W and a grounded screen placed above the sample to minimize the RF fluctuations. The dominant positive ion population in this condition was H3+/D3+ giving the incoming ion energy of 45 eV per nucleon at the surface for the sample surface bias of Vs = –130 V. The surface stationary state under the given bias is achieved after 180

General conclusion

approximately 10 minutes of plasma exposure for diamond and graphite samples (estimated thanks to NIEDF time evolution measurements). A study presented in this thesis was performed on a large variety of materials. The graphitic materials under study were Highly Oriented Pyrolitic Graphite (HOPG), carbon fiber composite (CFC) and tetrahedral amorphous carbon (ta-C). The diamond materials were microcrystalline boron-doped diamond (MCBDD), microcrystalline non-doped diamond (MCD), nanocrystalline diamonds (NCD 1% and NCD 5%). The metals were: molybdenum (Mo), gadolinium (Gd) and tungsten (W). The influence of plasma exposure temperature, surface bias and exposure time on H–/D– yields was investigated. The behavior of all graphitic materials with surface temperature is very similar and demonstrates an exponential decay starting from 200°C. The heated diamond films have shown to be the best materials for NI surface production. NI production on diamond films increases with surface temperature up to 400-500°C and then starts to decrease. The signal increase by a factor 5 was observed both in hydrogen and deuterium. Time evolution measurements were performed on MCBDD for different surface temperatures. It was observed that the creation of defects on the surface has decreased the NI Yield by a factor of 3. On the other hand, heating of MCBDD to 400°C has hindered the creation of defects by the plasma exposure and kept the surface in the state favorable for NI production. The NI yield evolution with temperature for all carbon materials could not be explained by the change in H/D surface coverage, but by the change of ionization probability Piz(E,θ) and the modified surface state of the exposed samples with its electronic properties. The evaluation of Mo as a background material has revealed that it gives sufficiently low NI signal levels. The evolution of NI production on Gd sample and W filament with surface temperature was performed. NI signal on Gd stays constant up to the surface temperature of 200°C and then decreases, similar to graphitic materials. After performing the heating cycle, Gd demonstrated an increase of NI production probably due to higher hydrogen surface coverage. One can imagine that during high temperature measurements, hydrogen species diffuse inside the material creating a reservoir. In order to follow the evolution of sputtered NI with temperature, measurements in Ar plasma have been performed. One can conclude that the reservoir of H inside Gd, once it has been created, could not be eliminated completely neither by energetic Ar+ ion bombardment nor by ion bombardment accompanied by heating. Comparing our study to the literature, one could suggest that the NI yield increase observed on Gd surface after H adsorption (enhanced by the heating cycle) could be connected to the formation of GdHx complexes and H– on the sample surface and in the bulk. The main conclusion of the study of NI production on heated W filament is that one does not observe any enhancement of NI production on a W surface when the temperature is increased from 300 K to more than 3000 K. The maximum of surface NI production lay at around 1000÷2000 K and was very low compared to NI created on carbon materials. These results are not encouraging and seem to indicate that increasing 181

General conclusion

the number of emitted electrons from the surface is not enough to promote NI surface production. The method of pulsed bias was developed to enable the study of NI production on surfaces of insulating materials such as MCD. The present technique has enabled to study NI surface production on MCD for the whole temperature range starting from RT. The pulsed-bias tests were performed on HOPG to demonstrate the feasibility of the method. By changing the pulsed-bias frequency (and the duty cycle) it was possible to obtain HOPG material with different hydrogen surface coverage and hence a different surface state (with the NI yield increase up to 30-50%). After proving the feasibility of the pulsed bias approach on HOPG, the optimization of the experimental parameters was performed on MCD by taking into account the charging effects. The comparison of NIEDF for constant and pulsed bias at RT revealed that the NI yield increased by nearly 2 times for both HOPG and pair MCBDD/MCD. At high temperature, a 3.5 times higher NI yield was measured on MCD surface in the case of pulsed bias as compared to constant DC bias. This is the highest NI yield to be ever measured on PHISIS set-up which leads us to a conclusion that there is still room for optimization of the NI yield on carbon materials. The results of time evolution studies for HOPG and MCD bring us to the conclusion that in pulsed bias case the diamond surface is less degraded and more hydrogenated, which is favorable for NI surface production. Considering the overall result of the pulsed bias measurements, one can conclude that to optimize the NI yield on diamond, one has to work with a less degraded surface. This can be obtained rising the surface temperature to 400°C–500°C which allows restoring intrinsic properties of diamond. The less degraded surface state can also be obtained by applying the pulsed bias which gives the possibility to increase the H2/D2 surface coverage and diminish the defects induced by plasma exposure. However, pulsing the bias is a solution only for fundamental studies. It is not a solution that could be used in a real NI source since the absolute NI yield is low. The use of ex situ surface diagnostics such as temperature programmed desorption (TPD) and Raman spectroscopy has allowed to characterize the surface state of carbon materials: HOPG, MCBDD and MCD. The TPD analysis was performed on the samples exposed to plasma at different surface temperatures in order to correlate the NI yield evolution with temperature to the surface state changes. The calibration of the mass spectrometer allowed the calculation of the full amount of desorbed species and their dynamics. By calculating deuterium atomic concentration in the modified layers in D2 and CD4 forms at different surface temperatures, it was seen that the film properties and the NI formation mechanisms are very different for HOPG and diamond films. The NI yield evolution is shown to correlate qualitatively with the total deuterium concentration in the interaction layer, which is in agreement with the modeling developed previously. Bibliographic research allows to characterize the surface states of the materials for different plasma exposure temperatures. The HOPG RT surface state shows the features 182

General conclusion

of nanocrystalline graphite with a certain amount of in-plane sp2 and out-of-plane sp3 defects and with deuterium content similar to hard films. On the other hand, HOPG 400°C shows the decrease of both types of defects (especially the out-of-plane ones) and the reduction of deuterium content. These facts could be explained by the re- establishment of sp2 phases and increase of the sp2 cluster size. At 800°C the HOPG sample undergoes the reconstruction of graphitic sp2 phases and almost complete removal of the defects. The MCBDD RT sample shows deuterium bound with carbon mostly in sp3 configuration with some defect-induced sp2 bonds and resembled much the spectra of C:H films for which the desorbed species come from the whole thickness of the film. The heated diamond films (MCBDD and MCD) demonstrate desorption of big quantities of deuterium in D2 form which could stem from the surface. The results of 5 sec plasma exposure suggest that at 400°C the original diamond structure is partially rebuilt. The enhancement of the NI yield for heated diamond surfaces could be explained by the achievement of NEA via the formation of the surface hydride complexes. This hypothesis, however, was not confirmed experimentally. The following NI signal decrease starting from 400°C up to 800°C is most probably connected to the desorption of D from the surface (passage from high to low D-coverage). It is concluded that the role of boron in NI production is insignificant, though it provides for the surface conductivity of the material at room temperature.

183

Bibliography

1. T.J.M. Boyd, J.J. Sanderson, The Physics of Plasmas (Cambridge University Press, New York, USA, 2003). 2. Achkasov, K. Evaluation of Beam Emission Spectroscopy as a Diagnostic Tool for the TJ-II Stellarator. Master thesis, Erasmus Mundus FUSION-EP Program, 2012. 3. John Wesson, Tokamaks (Clarendon Press, Oxford, England, 2004). 4. Kogut, D. Study of wall conditioning in tokamaks with application to ITER. Doctoral thesis, Aix-Marseille University, 2014. 5. Wesson, J. (2006) The Science of Jet. 2nd Edition, JET Joint Undertaking, Abingdon. 6. Peterson, Joshua. “Vacuum Magnetic Flux Surface Measurements Made on the Compact Toroidal Hybrid,” August 15, 2008. https://etd.auburn.edu/handle/10415/1237 7. Francis F. Chen, An Indispensable Truth. How Fusion Power Can Save the Planet (Springer, New York, USA, 2011). 8. Simonin, A., “Towards an efficient Neutral Beam Injection system for the future thermonuclear fusion reactors”. Memoire de HDR, 2013. 9. Asai, T., Y. Suzuki, T. Yoneda, F. Kodera, M. Okubo, S. Okada, and S. Goto. “Experimental Evidence of Improved Confinement in a High-Beta Field-Reversed Configuration Plasma by Neutral Beam Injection.” Physics of Plasmas (1994- Present) 7, no. 6 (June 1, 2000): 2294–97. 10. K. Watanabe, T. Amemiya, M. Hanada, T. Iga, T. Imai, T. Inoue, M. Kashiwagi, M. Kuriyama, T. Morishita, Y. Okumura, T. Takayanagi, and T. Yamamoto, Review of scientific instruments, 73, 2002. 11. Hemsworth R S, Tanga A and Antoni V 2008 Status of the ITER neutral beam injection system (invited) Rev. Sci. Instrum. 79 02C109. 12. L. Schiesko et al First experiments for Cs doped Mo as surface converter for negative hydrogen ion sources for fusion: determination of the H- surface generation mechanisms and long term stability. 13. Franzen P et al 2011 Magnetic filter field dependence of the performance of the rf driven IPP prototype source for negative hydrogen ions Plasma Phys. Control. Fusion 53 115006 14. Bacal M 2006 Physics aspects of negative ion sources Nucl. Fusion 46 S250 15. Béchu, S., A. Soum-Glaude, A. Bès, A. Lacoste, P. Svarnas, S. Aleiferis, A. A. Ivanov Jr, and M. Bacal. “Multi-Dipolar Microwave Plasmas and Their Application to

184

Bibliography

Negative Ion Production.” Physics of Plasmas (1994-Present) 20, no. 10 (October 1, 2013): 101601. 16. Samukawa S, Sakamoto K and Ichiki K 2001 High-efficiency low energy neutral beam generation using negative ions in pulsed plasma Japan. J. Appl.Phys. Part 2 40 L997–9 17. Draghici M and Stamate E 2010 Properties and etching rates of negative ions in inductively coupled plasmas and dc discharges produced in Ar/SF6 J. Appl. Phys. 107 123304 18. Vozniy O V and Yeom G Y 2009 High-energy negative ion beam obtained from pulsed inductively coupled plasma for charge-free etching process Appl. Phys. Lett. 94 231502 19. Fantz U, Franzen P and Wünderlich D 2012 Development of negative hydrogen ion sources for fusion: experiments and modelling Chem. Phys. 398 7–16 20. McNeely P and Schiesko L 2010 Investigation of fringe plasma parameters on a high power rf driven ion source Rev. Sci. Instrum. 81 02B111 21. Aanesland, A., D. Rafalskyi, J. Bredin, P. Grondein, N. Oudini, P. Chabert, D. Levko, L. Garrigues, and G. Hagelaar. “The PEGASES Gridded Ion-Ion Thruster Performance and Predictions.” IEEE Transactions on Plasma Science 43, no. 1 (January 2015): 321–26. 22. Ueno A et al 2010 Interesting experimental results in japan proton accelerator research complex H ion-source development (invited) Rev. Sci. Instrum. 81 02A720 23. Moehs D P, Peters J and Sherman J 2005 Negative hydrogen ion sources for accelerators IEEE Trans. Plasma Sci. 33 1786–98 24. Lettry, J., D. Aguglia, P. Andersson, S. Bertolo, A. Butterworth, Y. Coutron, A. Dallocchio, et al. “Status and Operation of the Linac4 Ion Source Prototypes).” Review of Scientific Instruments 85, no. 2 (February 1, 2014): 02B122. 25. Stockli, Martin P., K. D. Ewald, B. X. Han, S. N. Murray Jr, T. R. Pennisi, C. Piller, M. Santana, J. Tang, and R. Welton. “Recent Performance of the SNS H− Ion Source and Low-Energy Beam Transport Systema).” Review of Scientific Instruments 85, no. 2 (February 1, 2014): 02B137. 26. Belchenko Yu.I., Dimov G.I. and Dudnikov V.G. 1974 Nucl. Fusion 14 113. 27. Schiesko L et al 2011 Cesium influence on plasma parameters and source performance during conditioning of the prototype ITER neutral beam injector negative ion source Plasma Phys. Control. Fusion 53 085029 28. Dudnikov, V., and R. P. Johnson. “Cesiation in Highly Efficient Surface Plasma Sources.” Physical Review Special Topics - Accelerators and Beams 14, no. 5 (May 25, 2011): 054801. 29. C. F. A. van Os, P. W. van Amersfoort, and J. Los, J. Appl. Phys. 64 (1988) 3863.

185

Bibliography

30. P. McNeely, H.-D. Falter, U. Fantz, P. Franzen, M. Fröschle, B. Heinemann, W. Kraus, Ch. Martens, R. Riedl, and E. Speth; Development of a rf negative-ion source for ITER neutral beam injection; Rev. Sci. Instrum. 77, 03A519 (2006). 31. Gaboriau, F., R. Baude, and G. J. M. Hagelaar. “Experimental Evidence of the Increased Transport due to the Wall Bounded Magnetic Drift in Low Temperature Plasma.” Applied Physics Letters 104, no. 21 (May 26, 2014): 214107. 32. Sonato, P., V. Antoni, M. Bigi, G. Chitarin, A. Luchetta, D. Marcuzzi, R. Pasqualotto, et al. “Status of PRIMA, the Test Facility for ITER Neutral Beam Injectors.” In AIP Conference Proceedings, 1515:549–58. AIP Publishing, 2013. 33. Pasqualotto, R., G. Serianni, P. Sonato, M. Agostini, M. Brombin, G. Croci, M. Dalla Palma, et al. “Diagnostics of the ITER Neutral Beam Test Facilitya).” Review of Scientific Instruments 83, no. 2 (February 1, 2012): 02B103. 34. S. Mochalskyy, A. F. Lifschitz, T. Minea, Extracted current saturation in negative ion sources, J. Appl. Phys. 111(11) (2012) 113303. 35. S. Mochalskyy, A. F. Lifschitz, T. Minea, 3D modeling of negative ion extraction from a negative ion source, Nucl. Fusion 50 (2010) 105011. 36. S. Mochalskyy, A. F. Lifschitz, T. Minea, Origin of extracted negative ions by 3D PIC-MCC modeling. Surface vs Volume comparison, NIBS 2010, Japan, November 14-19, conference proceeding, 2010, pp 30-39. 37. S. Mochalskyy, A. Lifschitz, T. Minea, Modeling of the negative ions extraction from a hydrogen plasma source. Application to ITER Neutral Beam Injector, 19th International Symposium on Plasma Chemistry (ISPC). Bochum, Germany, July 26-31, Conference proceeding, 2009. 38. Todorov, D., Kh Tarnev, Ts Paunska, St Lishev, and A. Shivarova. “Spatial Distribution of the Plasma Parameters in a Radio-Frequency Driven Negative Ion Source.” The Review of Scientific Instruments 85, no. 2 (February 2014): 02B104. 39. A. Simonin, Jocelyn Achard, K. Achkasov, S. Bechu, C. Baudouin, O. Baulaigue, C. Blondel, JP. Boeuf, D. Bresteau, G. Cartry, W. Chaibi, C. Drag, H.P.L. de Esch, D. Fiorucci, G. Fubiani, I. Furno, R. Futtersack, P. Garibaldi, A. Gicquel, C. Grand, Ph. Guittienne, G. Hagelaar, A. Howling, R. Jacquier, M.J. Kirkpatrick, D. Lemoine, B. Lepetit, T. Minea, E. Odic, A. Revel, B.A. Soliman, P. Teste. “R&D around a photoneutralizer-based NBI system in view of a DEMO steady state fusion reactor”. Nuclear Fusion, accepted. 40. Simonin, A., L. Christin, H. P. L. De-Esch, R. Futtersack, P. Garibaldi, and F. Villecroze. “Mirror-like Plasma Confinement for a Uniform Large Negative Ion Source.” Nuclear Fusion 52, no. 6 (2012): 063003. 41. A. Simonin, P. Garibaldi; Review of Scientific Instruments 77, 03A525, 2006. 42. Ph. Guittienne et al., Journal of Applied Physics 98, 083304, 2005. 43. Ch. Hollenstein et al., Plasma Sources Sci. Technol. 22, 055021, 2013. 44. Ph. Guittienne et al., Phys. Rev. Letters, 111, 125005, 2013. 45. U. Fantz et al.; Rev. Sci. Instrum. 81 02B102, 2010.

186

Bibliography

46. Belhaj M, Tondu T, Inguimbert V, Barroy P, Silva F and Gicquel A 2010 J. Phys. D: Appl. Phys. 43 135303P. 47. Kumar P et al 2011 Enhanced negative ion yields on diamond surfaces at elevated temperatures J. Phys. D: Appl. Phys. 44 372002. 48. Van der Weide, J., Z. Zhang, P. K. Baumann, M. G. Wensell, J. Bernholc, and R. J. Nemanich. “Negative-Electron-Affinity Effects on the Diamond (100) Surface.” Physical Review B 50, no. 8 (August 15, 1994): 5803–6. 49. Hoh, Hui Ying, Kian Ping Loh, Michael B. Sullivan, and Ping Wu. “Spatial Effect of C–H Dipoles on the Electron Affinity of Diamond (100)-2×1 Adsorbed with Organic Molecules.” ChemPhysChem 9, no. 9 (June 23, 2008): 1338–44. 50. Ahmad Ahmad, Cédric Pardanaud, Marcel Carrère, Jean-Marc Layet, Alix Gicquel, Pravin Kumar, David Eon, Cédric Jaoul, Richard Engeln, and Gilles Cartry. “Negative-Ion Production on Carbon Materials in Hydrogen Plasma: Influence of the Carbon Hybridization State and the Hydrogen Content on H− Yield.” Journal of Physics D: Applied Physics 47, no. 8 (February 26, 2014). 51. Boswell, R.W., Plasma Phys. and Controlled Fusion 26: 1147 (1984). 52. Ahmad A et al 2013 Negative-ion surface production in hydrogen plasmas: modeling of negative-ion energy distribution functions and comparison with experiments Plasma Sources Sci. Technol. 22 025006. 53. Silva F, Hassouni K, Bonnin X and Gicquel A 2009 J. Phys.: Condens. Matter 21 364202. 54. http://www.hidenanalytical.com/en/ 55. Schiesko L et al 2010 A comparative study of H− and D− Production on graphite

surfaces in H2 and D2 plasmas Plasma Sources Sci. Technol. 19 (4) 045016. 56. Schiesko L et al 2009 Negative ion surface production through sputtering in hydrogen plasma Appl. Phys. Lett. 95 (19) 191502. 57. Schiesko L et al 2008 H− production on a graphite surface in a hydrogen plasma Plasma Sources Sci. Technol. 17 (3) 035023 58. Ahmad A., Etude de la production d’ions négatifs sur des surfaces de carbone dans un plasma d’hydrogène sans Cs à basse pression. Doctoral thesis, Aix- Marseille University, 2012. 59. http://www.srim.org/ 60. Ziegler, J. Biersack, U. Littmark, The stopping and range of ions in solids, Pergamon, New York, 1985. 61. James F. Ziegler, M.D. Ziegler, J.P. Biersack, Nuclear Instruments and Methods in Physics Research B 268 (2010) 1818–1823. 62. J. P. Biersack, L. G. Haggmark: Nucl. Instrum. Methods 174, 257 (1980). 63. J. P. Biersack, W. Eckstein: Appl. Phys. A 34, 73 (1984). 64. Jacob W, Thin Solid Films 326 (1998) 1–42. 65. A. von Keudell, W. Möller, R. Hytry, Diamond Rel. Mater. 2 (1993) 251. 66. A. von Keudell, W. Möller, R. Hytry, Appl. Phys. Lett. 62 (1993) 937. 187

Bibliography

67. W. Jacob, M. Unger, Appl. Phys. Lett. 68 (1996) 475. 68. A. von Keudell, W. Jacob, J. Vac. Sci. Technol. A 15 (1997) 402. 69. A. Annen, W. Jacob, Appl. Phys. Lett. 71 (1997) 1326. 70. R.M.A. Heeren, M.J. de Graaf, D. Ćirić, H.J. Hopman, A.W. Kleyn, Applied Surface Science 70-71 (1993) 332-336. 71. B.T. Kelly, Physics of Graphite, Applied Science Publishers, London, 1981. 72. W. Möller, in: R.E. Clausing, L.L. Horton, J.C. Angus, P. Koidl (Eds.), Diamond and diamond-Like Films and Coatings, NATOASI Series B: Physics, Vol. 266, Plenum Press, New York, 1991, p.229. 73. W. Möller, Thin Solid Films 228 (1993) 319. 74. P. Reinke, W. Jacob, W. Möller, J. Appl. Phys. 74 (1993) 1354. 75. D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 75th ed., CRC Press, Boca Raton, FL, 1994. 76. Van os C. F. A., Heeren R. M. A., Vanamersfoort P. W., Applied Physics Letters 51 (1987) 1495-1497. 77. R Middleton, Nuclear Instruments and Methods, 144 (1977) 373. 78. http://www.scilab.org/

79. Cartry G et al 2012 Production of negative ions on graphite surface in H2/D2 plasmas: experiments and SRIM calculations Phys. Plasmas 19 063503. 80. M. Kustner, W. Eckstein, V. Dose, J. Roth, Nuclear Instruments and Methods in Physics Research B 145 (1998) 320. 81. M. Mayer, W. Eckstein, and B. M. U. Scherzer , J. Appl. Phys. 77, 6609 (1995). 82. Wada, M., T. Kasuya, T. Kenmotsu, and M. Sasao. “Angular Distributions of Surface Produced H− Ions for Reflection and Desorption Processesa).” Review of Scientific Instruments 85, no. 2 (February 1, 2014): 02B114. 83. Li, Dongqi, Jiandi Zhang, P. A. Dowben, and M. Onellion. “Altering the Gd(0001) Surface Electronic Structure with Hydrogen Adsorption.” Physical Review B 48, no. 8 (August 15, 1993): 5612–20. 84. Nowicka, E., R. Nowakowski, and R. Duś. “Surface and Bulk Phenomena in the Process of GdHx (0.01 < X < 3) Formation in Thin Gd Films: Transition from Metal to Transparent Semiconductor.” Applied Surface Science, Proceedings of the 3rd International Workshop on Surface Physics “Nanostructures on Surface”IWSP- 2007, 254, no. 14 (May 15, 2008): 4346–51. 85. Pardanaud, Cedric, Celine Martin, Gilles Cartry, Ahmad Ahmad, Loic Schiesko, Gregory Giacometti, Marcel Carrere, and Pascale Roubin. “In-Plane and out-of- Plane Defects of Graphite Bombarded by H, D and He Investigated by Atomic Force and Raman Microscopies.” Journal of Raman Spectroscopy, 2014. 86. Mellet, N., C. Martin, B. Pégourié, G. Giacometti, J. P. Gunn, G. Cartry, P. Languille, et al. “Modelling of the Micrometric Erosion Pattern Observed on the Tore Supra Limiter Tiles.” Nuclear Fusion 54, no. 12 (December 1, 2014): 123006.

188

Bibliography

87. Donnelly, C. M., R. W. McCullough, and J. Geddes. “Etching of Graphite and Diamond by Thermal Energy Hydrogen Atoms.” Diamond and Related Materials 6, no. 5–7 (April 1997): 787–90. 88. De Temmerman, G, R P Doerner, P John, S Lisgo, A Litnovsky, L Marot, S Porro, et al. “Interactions of Diamond Surfaces with Fusion Relevant Plasmas.” Physica Scripta T138 (December 2009): 014013. 89. Porro, S., G. De Temmerman, S. Lisgo, D. L. Rudakov, A. Litnovsky, P. Petersson, P. John, and J. I. B. Wilson. “Diamond Coatings Exposure to Fusion-Relevant Plasma Conditions.” Journal of Nuclear Materials, Proceedings of the 19th International Conference on Plasma-Surface Interactions in Controlled Fusion, 415, no. 1, Supplement (August 1, 2011): S161–64. 90. Philipp, P., L. Bischoff, U. Treske, B. Schmidt, J. Fiedler, R. Hübner, F. Klein, A. Koitzsch, and T. Mühl. “The Origin of Conductivity in Ion-Irradiated Diamond-like Carbon – Phase Transformation and Atomic Ordering.” Carbon 80 (December 2014): 677–90. 91. Villalpando I, John P, Porro S andWilson J I B 2011 Hydrogen plasma etching of diamond films deposited on graphite Diamond Relat. Mater. 20 711–6. 92. Federici, G., C. H. Skinner, J. N. Brooks, J. P. Coad, C. Grisolia, A. A. Haasz, A. Hassanein, et al. “Plasma-Material Interactions in Current Tokamaks and Their Implications for next Step Fusion Reactors.” Nuclear Fusion 41, no. 12 (December 1, 2001): 1967. 93. Tian, Z., J. W. Davis, and A. A. Haasz. “Deuterium Retention in Tungsten at Fluences of up to 1026 D+/m2 Using D+ Ion Beams.” Journal of Nuclear Materials 399, no. 1 (April 1, 2010): 101–7. 94. Liu, Fangshu, Yong Zhang, Wenjia Han, Jiangang Yu, Guanghong Lu, and Kaigui Zhu. “Investigation of Hydrogen Behavior in Tungsten Exposed to High Energy Hydrogen Plasma.” Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, The 18th International Conference on Ion Beam Modifications of Materials (IBMM2012), 307 (July 15, 2013): 320–23. 95. Jones, H. A. “A Temperature Scale for Tungsten.” Physical Review 28, no. 1 (July 1, 1926): 202–7. 96. A.S. Haka et al. PNAS 102 (2005), 1237 97. Website of the Research and Restoration Center of French Museums: http://c2rmf.fr/ 98. Fernando Rull Pérez and Jesus Martinez-Frias. “Raman spectroscopy goes to Mars”. Spectroscopy Europe, vol. 18, no. 1, (2006). 99. P. Mishra, K.P. Jain, Temperature-dependent Raman scattering studies in nanocrystalline silicon and finite-size effects, Phys. Rev. B 62 (2000), 14790– 14795.

189

Bibliography

100. M. Balkanski, R.F. Wallis, E. Haro, Anharmonic effects in light scattering due to optical phonons in silicon, Phys. Rev. B 28 (1983) 1928–1934. 101. Helwig, Andreas, Jan Spannhake, Gerhard Müller, Noel Rosman, and Thierry Pagnier. “Temperature Characterization of Silicon Substrates for Gas Sensors by Raman Spectroscopy.” Sensors and Actuators B: Chemical, Functional Materials for Micro and NanosystemsEMRSContaining Selected papers from the European Materials research Society (E-MRS 2006) Symposium G:, 126, no. 1 (September 20, 2007): 240–44. 102. Ferrari AC, Robertson J. Interpretation of Raman spectra of disordered and amorphous carbon. Phys Rev B. 2000:61: 14095. 103. Ferrari, A. C., and J. Robertson. “Resonant Raman Spectroscopy of Disordered, Amorphous, and Diamondlike Carbon.” Physical Review B 64, no. 7 (July 26, 2001): 075414. 104. http://www.mhhe.com/physsci/chemistry/essentialchemistry/flash/hyb rv18.swf 105. Andrea Carlo Ferrari, John Robertson. “Raman Spectroscopy of Amorphous, Nanostructured, Diamond-like Carbon, and Nanodiamond. Philos Trans Roy Soc Lond Ser A-Math Phys Eng Sci.” Philosophical Transactions. Series A, Mathematical, Physical, and Engineering Sciences 362, no. 1824 (2004): 2477– 2512. 106. Lucchese, M. M., F. Stavale, E. H. Martins Ferreira, C. Vilani, M. V. O. Moutinho, Rodrigo B. Capaz, C. A. Achete, and A. Jorio. “Quantifying Ion-Induced Defects and Raman Relaxation Length in Graphene.” Carbon 48, no. 5 (April 2010): 1592–97. 107. Pruvost F, Bustarret E and Deneuville A 2000 Characteristics of homoepitaxial heavily boron-doped diamond films from their Raman spectra Diamond Relat. Mater. 9 295–9 108. Pruvost F and Deneuville A 2001 Analysis of the Fano in diamond Diamond Relat. Mater. 10 531–5 109. Ferrari A C and Robertson J 2001 Origin of the 1150 cm−1 Raman mode in nanocrystalline diamond Phys. Rev. B 63 121405 110. Brunetto R, Baratta G A and Strazzulla G 2004 Raman Spectroscopy of ion irradiated diamond J. Appl. Phys. 96 380–6. 111. Chen H-C, Palnitkar U, Sundaravel B, Lin I-N, Singh A P and Ranade R K 2009 Enhancement of field emission properties in nanocrystalline diamond films upon 100 MeV silver ion irradiation Diamond Relat. Mater. 18 164–8. 112. P. A. Redhead, “Thermal desorption of gases,” Vacuum, vol. 12, no. 4, pp. 203–211, 1962. 113. Bisson R, Markelj S, Mourey O, Ghiorghiu F, Achkasov K, Layet J-M, Roubin P, Cartry G, Grisolia C and Angot T. “Dynamic fuel retention in tokamak wall

190

Bibliography

materials: a laboratory study of deuterium release from polycrystalline tungsten at room temperature”, Journal of Nuclear Materials. 114. Tawara, H., Y. Itikawa, H. Nishimura, and M. Yoshino. “Cross Sections and Related Data for Electron Collisions with Hydrogen Molecules and Molecular Ions.” Journal of Physical and Chemical Reference Data 19, no. 3 (May 1, 1990): 617–36. 115. Tawara, H., M. Kato. “Electron Impact Ionization Data for Atoms and Ions - up-dated in 1998.” NIFS-DATA-51 (Feb. 1999). 116. Janev R.K., J.G. Wang, I. Murakami and T. Kato. “Cross Sections and Rate Coefficients for Electron-Impact Ionization of Hydrocarbon Molecules”. NIFS- DATA-51 (Oct. 2001). 117. Itikawa, Yukikazu. “Cross Sections for Electron Collisions with Nitrogen Molecules.” Journal of Physical and Chemical Reference Data 35, no. 1 (December 8, 2005): 31–53. 118. NIST Chemistry WebBook: http://webbook.nist.gov/chemistry/ 119. Hornekær L et al “Metastable Structures and Recombination Pathways for Atomic Hydrogen on the Graphite (0001) Surface”. Physical Review Letters 96, 156104 (2006). 120. Liu, Shengguang, Jizhong Sun, Shuyu Dai, and Dezhen Wang. “Dynamic Monte-Carlo Modeling of Thermal Desorption of H2 from H+ Irradiated Graphite.” Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms 269, no. 4 (February 15, 2011): 431–36. 121. Güttler, A., T. Zecho, and J. Küppers. “Adsorption of D(H) Atoms on Ar Ion Bombarded (0 0 0 1) Graphite Surfaces.” Surface Science 570, no. 3 (October 20, 2004): 218–26. 122. Zecho, Thomas, Andreas Güttler, and Jürgen Küppers. “A TDS Study of D Adsorption on Terraces and Terrace Edges of Graphite (0 0 0 1) Surfaces.” Carbon 42, no. 3 (2004): 609–17. 123. Pisarev, A., Yu. Gasparyan, A. Rusinov, N. Trifonov, V. Kurnaev, A. Spitsyn, B. Khripunov, T. Schwarz-Selinger, M. Rasinski, and K. Sugiyama. “Deuterium Thermal Desorption from Carbon Based Materials: A Comparison of Plasma Exposure, Ion Implantation, Gas Loading, and C–D Codeposition.” Journal of Nuclear Materials 415, no. 1, Supplement (August 1, 2011): S785–S788. 124. Salançon, E., T. Dürbeck, T. Schwarz-Selinger, F. Genoese, and W. Jacob. “Redeposition of Amorphous Hydrogenated Carbon Films during Thermal Decomposition.” Journal of Nuclear Materials 376, no. 2 (May 31, 2008): 160–68. 125. J. Küppers. “The hydrogen surface chemistry of carbon as a plasma facing material”. Surface Science Reports 22 (1995) 249-321. 126. Kimura, H., M. Sasaki, Y. Morimoto, T. Takeda, H. Kodama, A. Yoshikawa, M. Oyaidzu, et al. “Thermal Desorption Behavior of Deuterium Implanted into

191

Bibliography

Polycrystalline Diamond.” Journal of Nuclear Materials 337–339 (March 1, 2005): 614–618. 127. Bobrov, K., H. Shechter, M. Folman, and A. Hoffman. “Deuterium Adsorption-Desorption from Diamond (100) Single Crystal Surfaces Studied by TPD.” Diamond and Related Materials 7, no. 2–5 (February 1998): 170–76. 128. Su, C, and J. -C Lin. “Thermal Desorption of Hydrogen from the Diamond C(100) Surface.” Surface Science 406, no. 1–3 (May 31, 1998): 149–66. 129. Yang, Yuemei L., Lisa M. Struck, Leyla F. Sutcu, and Mark P. D’Evelyn. “Chemistry of Hydrogen on Diamond (100).” Thin Solid Films 225, no. 1–2 (March 25, 1993): 203–11. 130. Annen, A, R Beckmann, and W Jacob. “Deposition and Characterization of Dense and Stable Amorphous Hydrogenated Boron Films at Low Substrate Temperatures.” Journal of Non-Crystalline Solids 209, no. 3 (February 1997): 240– 46. 131. B.D. Thoms, P.E. Pehrsson and J.E. Butler, J. Appl. Phys. 75 (1994) 1804. 132. B.D. Thoms and J.E. Butler, Surf. Sci. 328 (1995) 291. 133. D.D. Koleske, S.M. Gates, B.D. Thoms, J.N. Russell, Jr. and J.E. Butler, J. Chem. Phys. 102 (1995) 992. 134. A.V. Hamza, G.D. Kubiak and R.H. Stulen, Surf. Sci. 237 (1990) 35. 135. Ristein, J., F. Maier, M. Riedel, J.b. Cui, and L. Ley. “Surface Electronic Properties of Diamond.” Physica Status Solidi (a) 181, no. 1 (September 1, 2000): 65–76. 136. Zhigilei, Leonid V., Deepak Srivastava, and Barbara J. Garrison. “Intermediate Metastable Structure of the C{111}/(1×1)H–C{111}/(2×1) Surface Phase Transition.” Physical Review B 55, no. 3 (January 15, 1997): 1838–43. 137. Chin, R. P., J. Y. Huang, Y. R. Shen, T. J. Chuang, and H. Seki. “Interaction of Atomic Hydrogen with the Diamond C(111) Surface Studied by Infrared-Visible Sum-Frequency-Generation Spectroscopy.” Physical Review B 52, no. 8 (August 15, 1995): 5985–95. 138. http://people.virginia.edu/~lz2n/Diamond.html 139. Samara, V., J.-P. Booth, J.-F. de Marneffe, A. P. Milenin, M. Brouri, and W. Boullart. “A Dc-Pulsed Capacitively Coupled Planar Langmuir Probe for Plasma Process Diagnostics and Monitoring.” Plasma Sources Science and Technology 21, no. 6 (2012): 065004. 140. Kudlacek, P., R. F. Rumphorst, and M. C. M. van de Sanden. “Accurate Control of Ion Bombardment in Remote Plasmas Using Pulse-Shaped Biasing.” Journal of Applied Physics 106, no. 7 (October 1, 2009): 073303. 141. S.-B. Wang and A. E. Wendt, J. Appl. Phys. 88, 643, (2000).

192

Publications

[1] Dikii, N. P., A. N. Dovbnya, Yu V. Lyashko, D. V. Medvedev, E. P. Medvedeva, V. L. Uvarov, and K. V. Achkasov. “Diffusion of Sodium, Potassium, Calcium, Manganese, and Radon in Tuff and Clinoptilolite under Leaching.” Technical Physics 56, no. 7 (July 15, 2011): 1018–22. doi:10.1134/S1063784211070103.

[2] Schiesko, L., G. Cartry, C. Hopf, T. Höschen, G. Meisl, O. Encke, P. Franzen, B. Heinemann, K. Achkasov, C. Hopf, and U. Fantz. “Cs-Doped Mo as Surface Converter for H−/D− Generation in Negative Ion Sources: First Steps and Proof of Principle.” In AIP Conference Proceedings, 1655:020003. AIP Publishing, 2015. doi:10.1063/1.4916412.

[3] L. Schiesko, G. Cartry, C. Hopf, T. Höschen , G. Meisl, O. Encke, B. Heinemann, K. Achkasov, P. Amsalem and U. Fantz. “First experiments with Cs doped Mo as surface converter for negative hydrogen ion sources.” J. Appl. Phys. 118, 073303 (2015). doi:10.1063/1.4928861.

[4] Bisson R, Markelj S, Mourey O, Ghiorghiu F, Achkasov K, Layet J-M, Roubin P, Cartry G, Grisolia C and Angot T. “Dynamic fuel retention in tokamak wall materials: a laboratory study of deuterium release from polycrystalline tungsten at room temperature”. Journal of Nuclear Materials. Volume 467, part 1, pp. 432–438 (December 2015). doi:10.1016/j.jnucmat.2015.07.028

[5] A. Simonin, Jocelyn Achard, K. Achkasov, S. Bechu, C. Baudouin, O. Baulaigue, C. Blondel, JP. Boeuf, D. Bresteau, G. Cartry, W. Chaibi, C. Drag, H.P.L. de Esch, D. Fiorucci, G. Fubiani, I. Furno, R. Futtersack, P. Garibaldi, A. Gicquel, C. Grand, Ph. Guittienne, G. Hagelaar, A. Howling, R. Jacquier, M.J. Kirkpatrick, D. Lemoine, B. Lepetit, T. Minea, E. Odic, A. Revel, B.A. Soliman, P. Teste. “R&D around a photoneutralizer-based NBI system in view of a DEMO steady state fusion reactor”. Nuclear Fusion, Volume 55, Number 12, pp. 123020- 123038(19), (November 2015). http://dx.doi.org/10.1088/0029-5515/55/12/123020

[6] Y. Addab, C. Martin, C. Pardanaud, J. Khayadjian, K. Achkasov, D. Kogut, G. Cartry, G. Giacometti, M. Cabié, J.L Gardarein, and P. Roubin. “Formation of thin tungsten oxide layers: characterisation and exposure to deuterium”. Physica Scripta, Volume 2016, Number T167. http://dx.doi.org/10.1088/0031-8949/T167/1/014036

193

Publications

[7] D. Jerome, K. Achkasov, D. Kogut, A. Ahmad, J-M Layet, A. Simonin and G. Cartry. “Negative-ion surface production in hydrogen plasmas: determination of the negative- ion energy and angle distribution function using mass spectrometry”. Journal of Physics D: Applied Physics, submitted.

194