Dissertations / Theses on the topic 'Semiconductor metrology'

Cette thèse consiste à analyser les différentes propriétés des ateliers de métrologie, proposer de nouvelles approches pour optimiser les taux d'échantillonnage et développer de nouvelles stratégies dynamiques de réduction des risques en fabrication des semi-conducteurs.Une analyse approfondie des ateliers de métrologie sur le site de Rousset de STMicroelectronics a été réalisée. Leurs propriétés physiques ainsi que leurs caractéristiques, comme la qualification des mesures, les stratégies d'échantillonnage et d’ordonnancement des lots et les niveaux de risque, sont prises en compte. De plus, une nouvelle procédure a été développée pour aider à déterminer quelle stratégie d'échantillonnage convient le mieux aux caractéristiques de l’atelier de métrologie et aux valeurs de risque.De nouvelles approches sont ensuite proposées pour optimiser les taux d'échantillonnage des différents types de machines de métrologie en respectant la capacité de métrologie et en prenant en compte des paramètres tels que les débits des machines de production et de métrologie et les probabilités de défaillance des machines de production. Les résultats montrent que la capacité de métrologie est mieux utilisée et que les machines de production sont contrôlées de manière efficace, en fonction de leurs caractéristiques, avec une priorité plus forte sur les machines critiques.Dans la dernière partie de la thèse, des modèles de simulation de plusieurs ateliers de métrologie sont développés. Ces modèles reproduisent le comportement des ateliers pour mieux les comprendre et évaluer l'impact d’améliorations qui sont proposées
This thesis consists in analyzing the different properties of metrology workshops, proposing novel approaches to optimize sampling rates and developing new dynamic strategies for risk reduction in semiconductor manufacturing.A thorough analysis of metrology workshops in the site of Rousset of STMicroelectronics has been carried out. Their physical properties and also their characteristics, such as measure qualification, lot sampling and dispatching strategy and risk levels, are considered. Also, a new procedure is developed that helps to determine which sampling strategy fits better according to the metrology workshop characteristics and risk values.New approaches are then proposed to optimize the sampling rates for different types of metrology tools respecting the metrology capacity and taking into account parameters such as throughput rates of process machines and metrology tools, and the failure probabilities of process machines. The numerical experiments show that the metrology capacity is better used and process machines are efficiently controlled, depending on their characteristics, paying more attention the critical machines.In the final part of the thesis, simulation models of several metrology workshops are developed. These models reproduce the behaviour of the workshops to better understand them and to evaluate the impact of proposed improvements

Double crystal diffractometry and topography are now routinely used in many laboratories for the inspection of epitaxially grown devices. However the trend towards thinner layers and more complex structures requires the continued development of novel approaches using these techniques. This thesis is concerned with the development of these approaches to study the structural uniformity of semiconductor materials. The uniformity of large single crystals of lithium niobate has been studied using synchrotron radiation and double crystal X-ray topography. This study has shown a variety of contrast features including low angle grain boundaries and non-uniform dislocation densities. The abruptness of an interface between a layer and the underlying substrate has been studied using glancing incidence asymmetric reflections. Comparisons to simulated structures revealed that a closer match was achieved by the inclusion of a highly mismatched interfacial layer. This study illustrates the need for careful comparison between experimental and simulated rocking curves as different structures may produce very similar rocking curves. A double crystal topographic study of a AlGaAs laser structure revealed X-ray interference fringes. These are shown to be produced from the interaction of two simultaneously diffracting layers separated by a thin layer. Possible formation mechanisms have been discussed showing that these fringes are capable of revealing changes in the active layer at the atomic level. A novel approach has also been developed using synchrotron radiation to study the non-stoichiometry of GaAs. This approach uses the quasi-forbidden reflections which are present in III-V semiconductors due to the differences in the atomic scattering factors. This study has also discussed the behaviour of strong and weak reflections in the region of absorption edges and modelled their behaviour using the anomalous dispersion corrections of Cromer and Liberman.

The development of state-of-the-art AlGaN/GaN high electron mobility transistors (HEMTs) has shown much promise for advancing future RF and microwave communication systems. These revolutionary devices demonstrate great potential and superior performance and many commercial companies have demonstrated excellent reliability results based on multiple temperature accelerated stress testing. However, a physical understanding of the various reliability limiting mechanisms is lacking and the role and relative contribution of the various intrinsic material factors, such as physical stress and strain has not been clearly explained in the literature. Part of issues that impact device reliability are the mechanical stresses induced in the devices as well as the self-heating that also limit device performance. Thus, quantification of stress and temperature in AlGaN/GaN HEMTs is of great importance. To address some of the needs for metrology to quantify stress in AlGaN/GaN HEMTs, micro-Raman spectroscopy and micro-photoluminescence (micro-PL) were utilized to quantify the residual stress in these devices. Through the use of micro-Raman and micro-PL optical characterization methods, mapping of the vertical and lateral stress distributions in the device channels was performed. Results show that stress can be influenced by the substrate material as well as patterned structures including metal electrodes and passivation layers. Previously developed and reported micro-Raman thermometry methods require an extensive calibration process for each device investigated. To improve the implementation of micro-Raman thermometry, a method was developed which offers both experimental simplicity and high accuracy in temperature results utilizing a universal calibration method that can be applied to a broad range of GaN based devices. This eliminates the need for performing calibration on different devices. By utilizing this technique, it was revealed that under identical power dissipation levels, the bias conditions (combination of Vgs and Vds) alter the heat generation profile across the conductive channel and thus influence the degree of device peak temperature. The role of stress in the degradation of AlGaN/GaN HEMTs was also explored. A combined analysis using micro-Raman spectroscopy, coupled electro-thermo-mechanical simulation, and electrical step stress tests was conducted to investigate the link between performance degradation and the evolution of total stress in devices. It was found that in addition to stresses arising from the inverse piezoelectric effect, the substrate induced residual stress and the operational themo-elastic stress in the AlGaN layer play a major role in determining the onset of mechanically driven device degradation. Overall, these experiments were the first to suggest that a critical level of stress may exist at which point device degradation will start to occur. The optical characterization methods developed in this study show the ability to reveal unprecedented relationships between temperature/stress and device performance/reliability. They can be used as effective tools for facilitating improvement of the reliability of future AlGaN/GaN HEMTs.

Les nouveaux matériaux de type chalcogénures (à base de S, Se, Te) font l’objet d’un intérêt croissant, non seulement pour les applications mémoires avancées, photonique et photovoltaïque, mais également autour des matériaux dichalcogénures innovants à base de métaux de transition (MoS₂, WS₂, ..). Les propriétés de ces matériaux, réalisés sous forme d’alliages binaires ou ternaires, avec ou sans dopage, dépendent fortement de leur composition, du profil de composition dans ces couches très fines, ainsi que des conditions de surface et d’interface (préparation, passivation). La maîtrise des propriétés de ces couches fines, déposées par voie chimique (CVD) ou par co-pulvérisation cathodique magnétron, doit s’appuyer sur des nouveaux protocoles de caractérisation aux incertitudes optimisées et compatibles avec un contrôle de fabrication en ligne. Dans cette thèse, nous présentons les performances de protocoles de métrologie spécifiquement développés pour l’analyse de couches minces de chalcogénures. Ces protocoles, qui s’appuient essentiellement sur les techniques non destructives de spectroscopie de photoélectrons (XPS) et de fluorescence X (XRF), ont été optimisés pour la caractérisation surfacique des couches ultrafines, l’analyse quantitative de la composition des matériaux complexes à base de tellure ou de soufre, et la mesure du profil de composition dans des couches et empilements < 50 nm. Dans un premier temps, nous présentons l’étude par XPS quasi in situ des propriétés de surface des matériaux Ge, Sb, Te ainsi que de leurs composés binaires et ternaires. Nous mettons en évidence l’évolution de la surface après remise à l’air puis vieillissement, et nous comparons l’efficacité de stratégies d’encapsulation in situ de couches minces à base de Te et Se. Nous démontrons ensuite les performances de protocoles d’analyses par XRF à dispersion de longueur d’onde (WDXRF) et XPS pour la quantification précise de la composition chimique de composés Ge-Sb-Te (de 1 à 200 nm) et de couches ultrafines de dichalcogénures à base de métaux de transition (MoS₂, WS₂). L’analyse combinée WDXRF/XPS permet de mesurer l’évolution avec la composition des facteurs de sensibilité relative des composantes Ge3d, Te4d et Sb4d, et par conséquent d’améliorer la précision de mesure par XPS de la composition des matériaux à changement de phase de type GexSbyTez. Nous soulignons également l’influence des effets de matrice sur la capacité de la WDXRF à l’analyse quantitative de l’azote dans des matériaux Ge-Sb-Te. Nous évaluons la possibilité d’un étalonnage de la WDXRF fondé sur des analyses par faisceaux d’ions spécifiques, ce qui permet in fine un suivi en ligne de couches GeSbTeN dans une fenêtre procédé donnée. Enfin, nous présentons deux stratégies de caractérisation non destructive du profil de composition dans des couches minces de chalcogénures. D’une part, nous démontrons que la combinaison des techniques de XRF en géométrie d'incidence rasante (GIXRF) et de réflectométrie X (XRR) permet une mise en évidence non ambiguë de faibles variations dans les procédés de dépôts, voire de phénomènes de diffusion dans des empilements de 10 nm d'épaisseur. L'utilisation de substrats multicouches en lieu et place du silicium permet d’optimiser la distribution en profondeur du champ d'ondes stationnaires, ce qui conduit à une amélioration nette de la sensibilité des stratégies XRR / GIXRF. D’autre part, nous montrons l’adéquation de protocoles fondés sur l’analyse XPS résolue en angle pour la caractérisation du profil de composition dans des couches nanométriques de GeTe et Ge₂Sb₂Te₅, ce qui permet une étude fine des premières étapes de dépôt de ces matériaux
Chalcogenide materials are compounds based on S, Se, and Te elements from group VI of the periodic table. They are receiving an extensive interest not only for applications in resistive memories (PCRAM and CBRAM), photonics and photovoltaics but also in the development of new 2-D materials (e.g. spintronics applications). Chalcogenide materials are already present in the semiconductor roadmaps and it is already replacing flash memories (e.g. phase change material and ovonic threshold switch in new random access memory). For the next technology nodes, chalcogenide properties can be scaled by tuning the chemical composition or by reducing the film thickness. Nonetheless, it also means that their properties become more tightly influenced by the chemical composition, the surface/interface effects and the depth-profile composition. Hence, dedicated metrology protocols must be developed, first to assist the optimization of chalcogenide materials processes in cleanroom environment, then to allow non-destructive process monitoring with industry-driven uncertainties. In this PhD thesis, we developed metrology protocols based on X-ray techniques, dedicated to thin chalcogenides materials and fully compatible with inline monitoring. First, we used quasi in-situ X-ray Photoelectron Spectroscopy (XPS) to characterize the surface of Ge, Sb, Te thin materials and compounds, and to study the composition-dependent evolution of the surface after air break and ageing. The efficiency of in situ capping strategies to protect Te-based and Se-based thin layered materials from ageing was also investigated. Secondly, we demonstrated the ability of improved metrology strategies based on in-line Wavelength Dispersive X-ray Fluorescence (WDXRF) and XPS to accurately quantify the chemical composition of Ge-Sb-Te compounds (from 1 to 200 nm) and ultrathin 2D transition metal dichalcogenides (MoS₂, WS₂). Combined WDXRF/XPS analysis was used to determine refined values of composition-dependent relative sensitivity factors for Te4d, Sb4d and Ge3d that allow for XPS-based metrology of PCRAM materials with mastered accuracy. We pointed the need for in-depth study of the significant matrix effects that alter the ability of WDXRF to quantify Nitrogen in Ge-Sb-Te materials: ion beam analysis was carefully investigated as possible input for WDXRF calibration, and a WDXRF protocol was established for inline monitoring of N-doped Ge-Sb-Te films in a specific process window. Finally, we investigated two ways to non-destructively characterize the in-depth chemical distribution in thin chalcogenide films: we demonstrated that the combination of XRF in grazing incidence geometry (GIXRF) and X-ray reflectometry (XRR) was able to unambiguously reveal small process differences along with process-induced diffusion in 10 nm-thick stackings. We showed that the use of multilayered substrate instead of silicon allowed fine-tuning of the depth-dependent X-ray standing wave field, resulting in improved sensitivity of XRR/GIXRF strategies. We also developed an angle-resolved XPS protocol for the evaluation of the first deposition steps of GeTe and Ge₂Sb₂Te₅ films, revealing the process-dependent elemental distribution as a function of the film growth. Therefore, in this work we not only elaborated advanced metrology protocols for the development of new chalcogenide films but also metrological solutions for the next technology nodes (28 nm and below), since current in-line metrology tools reach their detection limits

A major challenge in the manufacturing of micro and nano devices is the cleaning, rinsing, and drying of very small structures. Without a technology for in situ and real-time monitoring and controlling, the rinse processes that account for a significant fraction of the total processing steps use a large amount of water and energy perhaps unnecessarily. This "blind" processing approach leads to waste that can have significant economic and environmental impacts. An electrochemical residue sensor (ECRS) has been developed and is aimed at in situ and real-time measurement of residual contamination inside the micro and nano structures. Using this technology, the mechanisms and bottlenecks of cleaning, rinsing, and drying can be investigated and the processes can be monitored and controlled.An equivalent circuit model was developed to assist the design of the sensor; its validity was proved by the first prototype. The simulation results and the experimental data predicted a good sensitivity in a wide range of operational frequency. To use the sensor in a practical rinse tank setup, the sensor-on-wafer prototype was designed and fabricated. Both the fab-scale and the lab-scale tests were performed and results illustrated many successes. The sensor is the first and the only available technology that provides the in situ and real-time cleanness information in the microstructures during the rinse processes. The sensor results distinguished four different types of rinse processes and showed high sensitivity to the ionic concentration change in the microstructures. The impacts of cleaning and rinsing parameters such as flow rate, temperature, cleaning solution concentrations, and process time on the sulfuric acid rinsing efficiency were investigated by using the sensor. The investigation discovered that sulfuric acid rinsing is a two-stage process: a flow-control stage and a desorption-control stage. A comprehensive rinse model was developed to correlate the transport process and the trench impedance that is the sensor's signal. This model combined with the experimental data proved that increasing flow rate in the overflow rinse has a low efficiency for the rinse processes controlled by the surface reactions. The model, for the first time, shows the dynamics of the charging of the silicon dioxide surface and the dynamics of the potential build-up in the solution. It also discovered that the cation rinsing is a challenge if the cation adsorbs on or reacts with the surface.

Hintergrund: Die Entwicklung von Nanotechnologien und insbesondere integrierten Schaltkreisen beruht auf dem Verständnis von Struktur und Funktion auf der Nanoskala, wofür exakte Messungen erforderlich sind. Kleinwinkel-Röntgenstreuung unter streifendem Einfall (GISAXS) ist eine Methode zur schnellen, berührungs- und zerstörungsfreien dimensionellen Messung von nanostrukturierten Oberflächen. Ziele: Es soll die Möglichkeit untersucht werden, die zunehmend komplexeren Proben aus Wissenschaft und Industrie mit Hilfe von GISAXS präzise zu vermessen. Ein weiteres Ziel ist es, Messtargets aus der Halbleiter-Qualitätskontrolle mit einer Größe von ca. 40x40 µm² zu messen, deren Signal typischerweise nicht zugänglich ist, weil ein Bereich von ca. 1x20 mm² auf einmal beleuchtet wird. Methoden: Synchrotron-basierte GISAXS-Messungen verschiedener Proben werden mit Hilfe einer Fourier-Konstruktion, der "distorted wave Born approximation" und einem Maxwell-Gleichungs-Löser basierend auf finiten Elementen analysiert. Ergebnisse: Aus GISAXS-Messungen kann die Linienform von Gittern mit einer Periode von 32 nm rekonstruiert werden und sie weicht weniger als 2 nm von Referenzmessungen ab. Eine sorgfältige Bayes'sche Unsicherheitsanalyse zeigt jedoch, dass wichtige dimensionelle Parameter innerhalb der Unsicherheiten nicht übereinstimmen. Für die Messung von kleinen Gittertargets entwerfe ich ein neuartiges Probendesign, bei dem das Target in Bezug auf die umgebenden Strukturen gedreht wird, und stelle fest, dass dadurch parasitäre Streuung effizient unterdrückt wird. Fazit: GISAXS-Messungen von komplexen Nanostrukturen und kleinen Targets sind möglich, jedoch würde GISAXS enorm von effizienteren Simulationsmethoden profitieren, die alle relevanten Effekte wie Rauhigkeit und Randeffekte einbeziehen. Hier gibt es vielversprechende theoretische Ansätze, so dass GISAXS eine zusätzliche Methode für die Halbleiter-Qualitätskontrolle werden könnte.
Background. The development of nanotechnology such as integrated circuits relies on an understanding of structure and function at the nanoscale, for which reliable and exact measurements are needed. Grazing-incidence small angle X-ray scattering (GISAXS) is a versatile method for the fast, contactless and destruction-free measurement of sizes and shapes of nanostructures on surfaces. Aims. A goal of this work is to investigate the possibility of precisely measuring the increasingly complex samples produced in science and industry using GISAXS. A second objective is to measure targets used in semiconductor quality control with a size of approx. 40x40 µm², whose signal is typically not accessible because an area of approx. 1x20 mm² is illuminated at once. Methods. I take synchrotron-based GISAXS measurements and analyze them using reciprocal space construction, the distorted wave born approximation, and a solver for Maxwell's equations based on finite elements. Results. I find that the line shape of gratings with a period of 32 nm can be reconstructed from GISAXS measurements and the results deviate less than 2 nm from reference measurements; however, a careful Bayesian uncertainty analysis shows that key dimensional parameters do not agree within the uncertainties. For the measurement of small grating targets, I create a novel sample design where the target is rotated with respect to the surrounding structures and find that this efficiently suppresses parasitic scattering. Conclusions. I show that GISAXS measurements of complex nanostructures and small targets are possible, and I highlight that further development of GISAXS would benefit tremendously from efficient simulation methods which describe all relevant effects such as roughness and edge effects. Promising theoretical approaches exist, so that GISAXS has the potential to become an additional method in the toolkit of semiconductor quality control.

Dans les noeuds technologiques avancés ainsi que les technologies dérivées, des règles de dessin de plus en plus aggressives sont nécessaires. Cela conduit à une complexification des structures dans les circuits intégrés actuels. De telles structures posent un défi important aux procédés de fabrication, notamment les étapes dites de patterning que sont la lithographie et la gravure. Afin d'améliorer et d'optimiser ces structures, les designers se basent sur les règles et connaissances qu’ont les ingénieurs de leurs procédés. Ces règles ont besoin d'être alimentées par des informations dimensionnelles et structurelles de plus en plus complexes : configurations de type bord arrondi, distance entre deux bouts de lignes, rétrecissement de ligne, etc. La métrologie doit évoluer afin que les ingénieurs soient capables de mesurer et quantifier les dimensions des structures les plus complexes dans le but d'estimer la variabilité de leur procédé. Actuellement la variabilité est principalement estimée à partir de données issues du suivi en ligne de structures simples car elles sont les seules à garantir une mesure robuste et reproductible. Mais, elles peuvent difficilement être considérées comme représentatives du procédé ou du circuit. Utiliser la métrologie par CD-SEM pour mesurer des structures complexes de manière robuste est un défi technique. La création de recettes de mesures est complexe, nécessite un temps non négligeable et ne garantit pas une mesure stable. Cependant, une quantité importante d'informations est contenue dans l'image SEM. Les outils d'analyses fournis par les équipementiers permettent aujourd'hui d'extraire les contours SEM d'une structure présente dans l’image. Ainsi, le CD-SEM prend des images et la partie métrologie est réalisée hors ligne afin d'estimer la variabilité. Cette thèse vise à proposer aux ingénieurs de nouvelles possibilités de métrologie dimensionnelle afin de l’appliquer pour le contrôle des structures les plus complexes. Les contours SEM sont utilisés comme source d’information et exploités pour générer de nouvelles métriques
In advanced technological nodes as well as derived technologies, aggressive design rules are needed. This leads to a complexity of structures in the current integrated circuits. Such structures pose a significant challenge to chip manufacturing processes, in particular patterning steps of lithography and etching. In order to improve and optimize these structures, designers need to rely on the rules and knowledge that engineers have about their processes. These rules need to be fed by complex dimensional and structural information: corner rounding, tip to tip distances, line end shortening, etc. Metrology must evolve so that engineers are able to measure and quantify the dimensions of the most complex structures in order to assess the process variability. Currently the variability is mainly quantified using data from the inline monitoring of simple structures as they are the only ones to guarantee a robust and reproducible measurement. But, they can hardly be considered as representative of the process or the circuit. Using CD-SEM metrology to measure complex structures in a robust way is a technical challenge. The creation of measurement recipes is complex, time consuming and does not guarantee a stable measurement. However, a significant amount of information is contained in the SEM image. The analysis tools provided by the equipment manufacturers allow to extract the SEM contours of a structure present in the image. Thus, the CD-SEM takes images and the metrology part is performed offline to estimate the variability.This thesis offers engineers new possibilities of dimensional metrology in order to apply it for process control of complex structures. SEM contours are used as a source of information and used to generate new metrics

In the modern semiconductor manufacturing processes, chemical mechanical planarization (CMP) has attained important processing step because of its ability to provide global planarization. CMP is the planarization technique which is used for the removal of excess material, as left over from the previous processing steps. In addition, CMP offers a uniform surface that is essential for subsequent processing steps, especially for the high resolution photolithography processes. In simpler notation, CMP is a process where a chemical reaction enhances in obtaining a planar surface through removal of the mechanical materials from a wafer. In this study, CMP performance of three electronic materials was investigated. Chemical vapor deposited (CVD) diamond films, as a first materials, was fabricated using hot-filament chemical vapor deposition technique (HFCVD). The synthesized microcrystalline diamond (MCD) films were characterized using Raman Spectroscopy, Scanning Electron Microscopy (SEM), Atomic Force Microscopy (AFM), and X-ray Diffraction (XRD). The CMP performance of the MCD and nanocrystalline diamond (NCD) synthesized in Nano Materials Research Laboratory (NMRL) were investigated by using commercial slurry procured by Logitech Inc. U.K. The post-CMP characterizations of diamond films were performed by AFM in order to investigate surface roughness. The result showed the significant reduction the surface roughness of MCD films (37 nm to 15 nm) and NCD films (18 nm to 12 nm). In addition, the CMP performance of the silicon dioxide was investigated in this research work. The novel nanodiamond-polymer based slurry was also developed by co-polymerization of N-isopropylacrylamide (NIPAM) and N,N'-methylenebisacrylamide, 3-(trimethoxysilyl) propyl methacrylate (MPS). The synthesized slurry was characterized by Transmission Electron Microscopy (TEM) for observing the dispersion of diamond particles in the polymer matrix. The investigation of silicon dioxide was carried out using conventional ceria based slurry and novel nanodiamond-polymer based slurry. The results showed excellent surface finish at the minor expense of material removal rate with nanodiamond-polymer based slurry. Also, the coefficient of friction of friction was significantly reduced by using novel nanodiamond polymer based slurry. Lastly, CMP behavior of copper wafer was examined under different polishing conditions. The polishing was carried out using the commercial slurry procured from Cabot Microelectronics Inc., U.S. The copper wafers were characterized by AFM in order to analyze surface roughness. The results showed the reduction in average surface roughness occurred from 4.7 nm to 1.7 nm. This range of average surface roughness meets the demands of modern semiconductor industries.

The construction and characterization of ultra-low noise semiconductor-based mode-locked lasers as frequency comb sources with multi-gigahertz combline-to-combline spacing is studied in this dissertation. Several different systems were built and characterized. The first of these systems includes a novel mode-locking mechanism based on phase modulation and periodic spectral filtering. This mode-locked laser design uses the same intra-cavity elements for both mode-locking and frequency stabilization to an intra-cavity, 1,000 Finesse, Fabry-P&"233;rot Etalon (FPE). On a separate effort, a mode-locked laser based on a Slab-Coupled Optical Waveguide Amplifier (SCOWA) was built. This system generates a pulse-train with residual timing jitter of <2 fs and pulses compressible to <1 ps. Amplification of these pulse-trains with an external SCOWA lead to 390 mW of average optical power without evident degradation in phase noise and pulses that are compressible to the sub-picosecond regime. Finally, a new laser is built using a 10,000 Finesse Fabry-P&"233;rot Etalon held in a vacuum chamber. The fluctuations in the optical frequency of the individual comb-lines over time periods longer than 12 minutes are shown to be significantly reduced to <100 kHz in a measurement that is limited by the linewidth of the reference source. The use of these comb sources as local oscillators in multi-heterodyne detection of arbitrary optical waveforms is explored in three different cases. 1) Sampling of mode-locked pulses, 2) sampling of phase modulated continuous wave light and 3) periodically filtered white light. The last experiment achieves spectral interferometry with unprecedented resolution.
Ph.D.
Doctorate
Optics and Photonics
Optics and Photonics
Optics

Dans toutes les industries manufacturières, les chambres mises en parallèles sur une même opération de production sont censées donner un résultat similaire et offrir des produits de qualité identique. Ceci n'est toutefois pas toujours le cas dans les unités de production.Le maintien d'une performance stable des chambres parallèles dans l'industrie des semi-conducteurs est un véritable défi car les machines traitent simultanément un grand nombre de produits dans le but de maximiser le rendement et optimiser l'utilisation des machines. Le travail de thèse consiste à proposer une méthodologie permettant de détecter et de corriger en temps réel ces différences de performance en exploitant toutes les données disponibles, utilisées habituellement de façon séparée, pour identifier les causes racines de toute différence significative entre les chambres traitant des produits identiques.L'approche proposée consiste d'abord à détecter les écarts existants entre les chambres parallèles en se référant aux mesures des paramètres physiques. Les données des capteurs sont ensuite analysées pour mettre en évidence les indicateurs causant ces écarts. Ces indicateurs sont ajustés grâce à un mécanisme de contrôle efficace composé de deux parties : La métrologie virtuelle et la régulation. Tout d'abord, l'impact du réglage des paramètres d'entrée des chambres sur la qualité des produits est modélisé. Les modèles créés sont ensuite utilisés pour mettre en œuvre des boucles de régulation dont le but est de faire correspondre les indicateurs source de variabilité, et compenser ainsi l'erreur de sortie
In the manufacturing industries, the machines/chambers placed in parallel on the same production operation are expected to have similar capabilities and, most importantly, to yield identical product quality. However, this is usually not the case in real practice due to the systematic variations accumulated in time. Maintaining stable performance of parallel machines/chambers in the semiconductor industry is a critical challenge given the fact that, in the large-scale production environment, machines/chambers can process a large number of products simultaneously to maximize throughput and optimize machine utilization. Un- surprisingly, after processing very different settings, called recipes, the conditions of parallel machines/chambers will be no longer the same. This thesis proposes a methodology to detect and correct the performance differences in real-time by using all the available data, namely: measurements of physical parameters, data from sensors installed on machines, data from the control loops, and maintenance data. The core idea is to integrate the different sources of data, which are usually used separately, to identify the root causes of any significant differences among the machines/chambers that process identical recipes.The proposed approach starts by detecting existing gaps between parallel machines/ chambers by referring to the measurements of physical parameters since they reflect the quality of manufactured products. The sensor data are then analyzed to highlight the in- dicators that cause these discrepancies. These indicators are adjusted through an effective control mechanism composed of two parts: 1) virtual metrology and 2) process regulation. First, the impact of recipe changes on product quality is quantified by modeling the link between the inputs and outputs of the mismatched machines/chambers. The constructed models are then used to implement the revised control loops to match as much as possible the controllable input factors and compensate for the output errors

Submitted by Pedro Silva Filho (pfsilva@ipen.br) on 2017-11-21T11:20:13Z No. of bitstreams: 0
Made available in DSpace on 2017-11-21T11:20:13Z (GMT). No. of bitstreams: 0
A calibração, por definição, é o processo pelo qual se estabelece uma relação entre valores de medição de um padrão, com as suas respectivas incertezas, e as indicações com as incertezas associadas do instrumento de medição a ser calibrado. Um protocolo de calibração descreve a metodologia a ser aplicada em um processo de calibração. O método escolhido para a obtenção deste protocolo foi o da espectrometria de feixe de raios X associada à simulação pelo método de Monte Carlo, fundamentado no fato de que ambos são considerados métodos absolutos na determinação de parâmetros de feixes de radiação. Neste trabalho foi utilizado o método de Monte Carlo utilizado para obter a função resposta do detector utilizada para a correção dos espectros obtidos do feixe primário de radiação X; deste modo foram calculadas as taxas de kerma destes feixes e comparadas aos valores obtidos com as câmaras de ionização padrão secundário do Laboratório de Calibração de Instrumentos do IPEN (LCI/IPEN). Foram obtidos os coeficientes de calibração para o sistema padrão com diferenças em relação ao fornecido pelo laboratório primário entre 1,3% e 15,3%. Os resultados obtidos indicaram a viabilidade do estabelecimento deste protocolo de calibração utilizando a espectrometria como padrão de referência, com incertezas relativas de 0,62% para k=1. As incertezas associadas ao método proposto foram satisfatórias, para um laboratório padrão secundário e comparáveis a um laboratório primário.
Tese (Doutorado em Tecnologia Nuclear)
IPEN/T
Instituto de Pesquisas Energéticas e Nucleares - IPEN-CNEN/SP

The full dielectric tensor of monoclinic Ga2O3 (β-phase) was determined by generalized spectroscopic ellipsometry in the spectral range from 1.0 eV up to 8.5 eV and temperatures in the range from 10K up to 300K. By using the oriented dipole approach, the energies and broadenings of the excitonic transitions are determined as a function of the temperature, and the exciton-phonon coupling properties are deduced.

碩士
中原大學
機械工程研究所
94
Abstract The semiconductor of our country regards manufacturing as the subject. Under global keen competition, strict process control already become the indispensable demand. It is most important factor of international competitiveness for Panel factory and Liquid Crystal Display(LCD) factory how does it want effective reducing production cost, increasing equipment efficiency, reducing process risk, increasing process yield. Present most of Panel factory and LCD factory use control technology for statistics process control technique to monitor important process parameters. However the result can’t be measure right away, this monitor maybe cause a large number of base plate loss if it happen process unusual. It will influence the production cost and yield seriously. Because of the flourishing development of information technology and internet network technology in recent years, produced the idea about e-Diagnostics and e-Maintenance. Pretext of internet network and information technology to achieve the goal of virtual metrology, diagnose-self and predict maintain. E-Diagnostics technology can diagnose and repair the equipment, shorten repair time and reduce service cost. Idea of virtual metrology is combine e-Diagnostics and advanced process control to achieve the goal of quality predict. The theory combines advanced process control technology and idea of virtual metrology to filter noise by wavelet theory, and using fuzzy neural networks predict trend and recipe influence. Pretext of predict metrology to improve the quality of process, promote equipment efficiency. Absolute error is under 25 and absolute error rate is under 1.5% in simulate data and absolute error is about 0.05 and absolute error rate is about 3.84% in actual data.

Virtual Metrology (VM) predicts end-of-batch properties (metrology data) from measurable input data composed of pre-process metrology and fault detection and classi cation (FDC) system outputs. This dissertation aims at moving a step closer to the realization of VM in semiconductor manufacturing by providing solutions to the challenges that present VM technology faces. First, various VM methods are introduced and compared in terms of prediction accuracy using four industrial datasets collected from a plasma etch system at Texas Instruments, Inc.. Kalman lter estimation is employed in a novel way to serve as a VM model for predicting outputs of a static process. Recursive PLS regression (R-PLSR) and Kalman filter show the best prediction results as they update the model whenever new measurements are available. Next, two PLS variants (PLS with EWMA mean update and recursive PLS) are proposed as robust VM algorithms that can predict process outputs fairly accurately in the presence of unexpected process drifts and noise. The obtained results reinforce VM technology by suggesting appropriate prediction methods when unexpected process changes occur. For a successful implementation of VM, the data entering the VM model needs to be free from faults. Fault-free (reconstructed) data are obtained by performing fault detection, fault identi cation, and fault reconstruction. A novel fault detection method based on statistics pattern analysis (SPA) is presented. The SPA method provides better fault detection performance for diff erent types of faults as compared to the MPCA-based methods. Next, three well-known fault identi cation methods present in literature are implemented. An equation that relates the RBC with the SVI is derived. The contribution plot method identi es a smaller number of faults correctly as compared to the RBC and the SVI methods. Fairly good estimates of the fault magnitude are obtained when the faults are identi ed correctly. An approach that combines physical measurements with the VM estimates to develop a more robust approach than using VM alone is presented. EWMA-R2R control is implemented using three well-known sampling methods in order to demonstrate the superior performance of two novel control schemes: B-EWMA R2R control and VM-assisted EWMA-R2R control. A new reliance index, which is attractive from a mathematical and practical point of view, is proposed. The VM-assisted EWMA-R2R control yields the best control results among the control schemes employed in this study. The simulation results demonstrate that VM has the potential to reduce measurement costs signi cantly while promising better process control.
text

碩士
國立臺灣大學
工程科學及海洋工程學研究所
97
Photolithography is the key technology driving the advancement of the semiconductor industry and directly influences the critical dimension (CD) of Ultra-Large Scale Integration chips. The reduction of CDs creates more strictly overlay control requirements. In order to control and minimize overlay metrology errors, we have to deal with a number of design parameters both on the metrology tools and on the overlay targets. For speeding the rate of performance improvement, optical simulation can be used to model the effects of target designs on the ultimate metrology performance. Optical simulation on the computer can aid R&D efforts to improve metrology methods. In this thesis, a overlay metrology simulation platform will be presented, developed in-house. The main idea of this platform is to simulate the overlay metrology which integrating with finite difference time domain and non-sequential ray tracing software package. Finite difference time domain method program is used to calculate the overlay target as an amplitude/phase-object in the near-field. And non-sequential ray tracing method program is used to calculate the power of the overlay target in the far-field. The simulation validation test with simulation standard overlay mark BiB（Bar-in-Bar）will also be detailed in this thesis..

This thesis presents the modeling of tool data produced during ion implantation for the prediction of wafer sheet resistance. In this work, we will use various statistical techniques to address challenges due to the nature of equipment data: high dimensionality, colinearity, parameter interactions, and non-linearities. The emphasis will be data integrity, variable selection, and model building methods. Different variable selection and modeling techniques will be evaluated using an industrial data set. Ion implant processes are fast and depending on the monitoring frequency of the equipment, late detection of a process shift could lead to the loss of a significant amount of product. The main objective of the research presented in this thesis is to identify any ion implant parameters that can be used to formulate a virtual metrology model. The virtual metrology model would then be used for process monitoring to ensure stable processing conditions and consequent yield guarantees.
text

碩士
中華大學
資訊工程學系碩士班
94
Semiconductor enters the age of NANO as time goes by; the concept of Advanced Process Control（APC）has gradually been used into semiconductor manufactories. Nowadays the Run-to-Run（R2R）and the Fault Detection and Classification（FDC）systems of Advanced Process Control have become essential in semiconductor manufacture. In terms of DRAM processes which more than 300mm; the trench width shrinks from 110nm to 90nm and may develop to under 60nm in the future by the renewal of process technology. This means the value of a wafer will show multiple growths. Thus taking advantage of APC to stabilize processes, to reduce the production loss and to improve the yield rate will be one of the key factors of profit-making in the new generation of semiconductor manufactories. But how to decrease the bad production rate, in the past, QC examination and Statistical Process Control（SPC）were often used to monitor and control the production quality. However, each process or equipment may not produce in the same fixed pre-determined process procedures in every production; therefore when collecting the data of R2R system, we just can infer a conclusion of complete data through partial data by using statistic methods. Even so, man-made or the natural suddenly changed process parameter and few sampling of data measurement may result in errors. This causes the inaccuracy of R2R system. For this reason, how to improve the accuracy of R2R data feedback becomes one of the important tasks in semiconductor industry. This research will combine Probabilistic Neural Network, clustering analysis in Data Mining and Principal Component Analysis to develop a system to forecast production results of process parameters by using measuring data and FDC production data to serve as the R2R system control basis of process parameter adjustment. Take semiconductor thickness measurement process for example to obtain 200 samplings of real production data to train, 10 samplings to examine. After tests and verification, the research finds this forecast function can infer the result up to 80%, with the Mean Absolute Percentage Error（MAPE） less than 20%. It shows parameter combination result inferred by parameter adjustment function has great accuracy. Combining this forecast system with FDC and R2R systems will reach overall quality management of thickness measurement, shorten the production time and improve the yield rate.

碩士
國立中興大學
資訊科學與工程學系所
100
The key factors in the production processes of semi-conductor industry are stability and production yield. To insure the stability of the process, wafer-to-wafer （W2W） advanced process control（APC）becomes more essential for the critical stages of production. However, one of the main problems in W2W control is the availability of timely metrology data at the wafer level. The cost of W2W testing is high in terms of money and time. To overcome this problem, an approach, called Virtual Metrology（VM）, is defined as the prediction of metrology variables（either measurable or non measurable）using process and wafer state information. The data in the semi-conductor manufacturing process contain many parameters. It is very difficult for professionals to select key parameters. The goal of this thesis is to present a two-phase variable selection method for VM module for semiconductor manufacturing process. The experimental results show that the proposed method outperforms other algorithms.

碩士
國立清華大學
工業工程與工程管理學系
93
Abstract A number of inspection and measurement stations are set in the fabrication process to ensure that the quality of wafer meets the requirement. Because of the limited capacities and costs for in-line wafer inspections, only certain wafers are inspected among a specific number of lots. However, conventional semiconductor wafer fabs meet a variety of economic challenge. The combination of shrinking devices geometries and increasing interconnect levels rapidly increase process complexity, which leads to higher manufacturing costs and longer cycle times. Although there are many existing studies for IC sampling strategy in defect inspection, little research has been done the issue of metrology sampling. In-line metrology was real time to inspect the WIP. Currently, the sampling metrology numbers and sampling frequency are decided via the engineers’ experience. Thus, different engineers may build various sampling strategies. This study aims to determine the optimal sampling strategy by developing a risk-based heuristic for statistically determining the sampling strategy for in-line inspection in wafer fabrication. For general defect inspection in wafer fabrication, Nurani et al. (1996) defined five parameters of sampling strategy including layers to be monitored, frequency for lots, and number of inspection wafers per lot, percentage area of the wafer and pixel size. However, for metrology inspection, we combine forth and fifth parameter to another parameter called number of dies in a wafer. Except layers to be monitor, our sampling strategy considers acceptance sampling plan in a wafer and a lot and sampling frequency that tradeoff the various risk (i.e., the aggregation of cost and probability) under different lot size. Keywords: Bayesian decision analysis, sampling strategy, metrology, inspection, quality control

碩士
國立交通大學
光電工程系所
92
We characterize the phase locking between CW semiconductor laser and mode-locked pulse laser. Through beating signal between CW diode laser and mode-locked pulse laser, semiconductor laser could be locked to the harmonic frequency of repetition rate of mode locked laser. In this thesis, we use the harmonic frequency of mode-locked laser as the standard of dual-wavelength CW semiconductor laser and stabilize the frequency of laser diode .We depress the frequency fluctuation of laser diode from 100MHz to 1~5MHz and lock the frequency of CW diode laser to 358660.800GHz.

碩士
國立交通大學
光電工程所
91
We characterize the phase-locking between dual-wavelength cw semiconductor laser and mode-locked pulse laser. Through biting signal from dipole antenna generated by two various wavelength cw lasers, semiconductor laser could be locked to high order harmonic frequency of repetition rate of mode locked laser was demonstrated. In this thesis, our experiment is be to divided into two parts, part one : we success to depressed timing jitter of mode-locked laser from 7.5ps to 880fs. Part two : We use the harmonic frequency of mode-locked laser as the standard of dual-wavelength cw semiconductor laser and to stabilize the frequency of laser diode .We depress the frequency fluctuation of laser diode from 100MHz to 4~10MHz.

碩士
中原大學
工業與系統工程研究所
103
Under Industry 4.0 concept and the management of big data issues in dealing for rapid decision making for improved productivity, the application development of Virtual Metrology in semiconductor industry for the metrology delay problem was investigated to make the system that becomes more immediacy and accuracy that increase quality and productivity for product/process. The predicted model which is a combination of the methodologies of Principal Component Analysis and Regression Analysis was proposed to filter the key factors of process. Addition to, the Artificial Neural Network are carried out to identify specific problems which are evaluated by Mean Absolute Percentage Error. Further on, a case study in Low pressure chemical vapor deposition equipment of process of semiconductor based on the 18 parameters and 264 data is discussed. The result of this system was proven to have good performance that the MAPE less than 10 percent.

Manufacturing equipments in semiconductor factories (fabs) provide abundant data and opportunities for data-driven process monitoring and modeling. In particular, virtual metrology (VM) is an active area of research. Traditional monitoring techniques using univariate statistical process control charts do not provide immediate feedback to quality excursions, hindering the implementation of fab-wide advanced process control initiatives. VM models or inferential sensors aim to bridge this gap by predicting of quality measurements instantaneously using tool fault detection and classification (FDC) sensor measurements. The existing research in the field of inferential sensor and VM has focused on comparing regressions algorithms to demonstrate their feasibility in various applications. However, two important areas, data pretreatment and post-deployment model maintenance, are usually neglected in these discussions. Since it is well known that the industrial data collected is of poor quality, and that the semiconductor processes undergo drifts and periodic disturbances, these two issues are the roadblocks in furthering the adoption of inferential sensors and VM models. In data pretreatment, batch data collected from FDC systems usually contain inconsistent trajectories of various durations. Most analysis techniques requires the data from all batches to be of same duration with similar trajectory patterns. These inconsistencies, if unresolved, will propagate into the developed model and cause challenges in interpreting the modeling results and degrade model performance. To address this issue, a Constrained selective Derivative Dynamic Time Warping (CsDTW) method was developed to perform automatic alignment of trajectories. CsDTW is designed to preserve the key features that characterizes each batch and can be solved efficiently in polynomial time. Variable selection after trajectory alignment is another topic that requires improvement. To this end, the proposed Moving Window Variable Importance in Projection (MW-VIP) method yields a more robust set of variables with demonstrably more long-term correlation with the predicted output. In model maintenance, model adaptation has been the standard solution for dealing with drifting processes. However, most case studies have already preprocessed the model update data offline. This is an implicit assumption that the adaptation data is free of faults and outliers, which is often not true for practical implementations. To this end, a moving window scheme using Total Projection to Latent Structure (T-PLS) decomposition screens incoming updates to separate the harmless process noise from the outliers that negatively affects the model. The integrated approach was demonstrated to be more robust. In addition, model adaptation is very inefficient when there are multiplicities in the process, multiplicities could occur due to process nonlinearity, switches in product grade, or different operating conditions. A growing structure multiple model system using local PLS and PCA models have been proposed to improve model performance around process conditions with multiplicity. The use of local PLS and PCA models allows the method to handle a much larger set of inputs and overcome several challenges in mixture model systems. In addition, fault detection sensitivities are also improved by using the multivariate monitoring statistics of these local PLS/PCA models. These proposed methods are tested on two plasma etch data sets provided by Texas Instruments. In addition, a proof of concept using virtual metrology in a controller performance assessment application was also tested.