Search

International audience; The absolute frequencies of the hyperfine components of the R(26) 62-0 transition in molecular iodine at 501.7 nm are measured for the first time with an optical clockwork based on a femtosecond laser frequency comb generator. The set-up is composed of an Ar+ laser locked to a hyperfine component of the R(26) 62-0 transition detected in a continuously pumped low-pressure cell (0.33 Pa). The detected resonances show a linewidth of 45 kHz (half-width at half-maximum). The uncertainty of the frequency measurement is estimated to be 250 Hz.

The paper presents an ultra-stable fibre optic link which allows to transfer frequency reference signal between two research laboratories. The link is composed of two 43 km twin optical fibres of the Paris telecommunications network that rely the LPL to LNE-SYRTE. The stability of the transferred signal is degraded by mechanical vibration and temperature variation along the fibre. To correct these effects a compensation system has been developed who acts directly on the fibre length, with a variable optical delay line

International audience; We present in this article results on a new dissemination system of an ultrastable reference signal at 100 MHz on a standard fiber network. The 100 MHz signal is simply transferred by amplitude modulation of an optical carrier. Two different approaches for compensating the noise introduced by the link have been implemented. The limits of the two systems are analyzed and several solutions are suggested in order to improve the frequency stability and to further extend the distribution distance. Nevertheless, our system is a good tool for the best cold atom fountains comparison between laboratories, up to 100 km, with a relative frequency resolution of 10 −14 at 1 s integration time and 10 −17 for 1 day of measurement. The distribution system may be upgraded to fulfill the stringent distribution requirements for the future optical clocks.

Some prescriptions for in-medium complex particle production in nuclear reactions are proposed. They have been implemented in two models to simulate nucleon-nucleus (nIPSE) and nucleus-nucleus (HIPSE) reactions around the Fermi energy \cite{Lac04,Lac05}. Our work emphasizes the effect of randomness in cluster formation, the importance of the nucleonic Fermi motion as well as the role of conservation laws. The key role of the phase-space exploration before and after secondary decay is underlined. This is illustrated in the case of two debated issues: the memory loss of the entrance channel in central collisions and the $(N,Z)$ partitions after the pre-equilibrium stage., Comment: Proceedings of the IWM2005 workshop, Catane (Italy), Nov. 2005. DOWNLOAD HIPSE program at: http://caeinfo.in2p3.fr/theorie/theory_lacroix.html

We use a new technique to disseminate microwave reference signals along ordinary optical fiber. The fractional frequency resolution of a link of 86 km in length is ${10}^{\ensuremath{-}17}$ for a one day integration time, a resolution higher than the stability of the best microwave or optical clocks. We use the link to compare the microwave reference and a ${\mathrm{CO}}_{2}/{\mathrm{OsO}}_{4}$ frequency standard that stabilizes a femtosecond laser frequency comb. This demonstrates a resolution of $3\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}14}$ at 1 s. An upper value of the instability introduced by the femtosecond laser-based synthesizer is estimated as $1\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}14}$ at 1 s.

A series of controlled abiotic syntheses of magnetite nanocrystals were carried out to explore the possibility of using morphological criteria, crystal size distributions, and shape ratio as tools for identifying nanocrystals that would be speciÞ cally produced by magnetotactic bacteria. High-quality magnetite crystals synthesized with various controlled total iron concentrations were shown to have cubo-octahedral shapes and sizes varying from 4 to 24 nm. The mean particle size of the population was found to be 10.5 ± 0.7 nm and no signiÞ cant effect of the total iron concentration on the particle size was observed. Systematical analyses of size and morphology also allowed for the determination of crystal size and shape ratio distributions. Crystal sizes were observed to follow log-normal distribu- tions. Shape factors are bounded by one, with maxima between 0.80 to 1.00. Their distributions are asymmetric, with a cut off toward the high values. Crystal morphologies and shape factors appear not to be a powerful diagnostic tool for the differentiation of abiotic vs. biotic particles. However, crystal size distributions of abiotic crystals are signiÞ cantly different from those of biotic populations. Indeed, opposite asymmetry of the size distributions from biogenic and non biogenic crystals was observed, with cut off toward larger sizes for biogenic nanocrystals and with cut off toward smaller sizes for abiogenic nanocrystals. This therefore constitutes a potential diagnostic tool for deciphering magnetite origin.

We have demonstrated the feasibility of a free-space ultra-stable optical link on a 3 meters test bench, operating at 100 MHz. With this type of link, it is possible to transfer a 100 MHz signal with a relative frequency stability of a few 10/sup -14/ at one second integration time, 10/sup -16/ at one day and a phase stability of a few picoseconds per day in presence of moderate mechanical vibrations and thermal fluctuations. The comparisons of modern clocks of distant (

We developed a stabilized frequency distribution system to transfer a reference through a 43 km standard telecommunication fiber optic. This low noise link has been used to perform frequency comparisons between LPL and SYRTE clocks

We report here experimental results obtained in saturated absorption spectroscopy of iodine in a low pressure cell at 501.7 nm. We obtained lines of 30 kHz wide (HWHM) for the R(26)62-0 transition, which are, to our knowledge, the narrowest resonances ever observed in iodine spectroscopy in cell

We report the first absolute frequency measurement of 127 I2 hyperfine components of the R(26)62-0 transition at 501.7 nm with an optical clockwork based on a femtosecond laser frequency comb generator

In this paper, we demonstrate the possibility to transfer both short and long-term frequency stability of frequency standards through a standard telecom network with compensation of phase fluctuations introduced by environmental perturbations.

Recent advances concerning two-photon Ramsey fringes in SF 6 at 30 THz obtained with a 1 m separation between the two interaction zones are presented. The experiment is referenced to a primary standard, the Hydrogen maser / Cs fountain located at BNM-SYRTE, Observatoire de Paris, via a femtosecond laser frequency comb generator. This results in an absolute frequency measurement of the central fringe to plusmn1.4 Hz which corresponds to 5times10 -14 in relative frequency.

The paper presents an ultra-stable fiber optic frequency link used to transfer reference frequency signal through a standard telecommunication optical fiber between SYRTE (Observatoire de Paris) and LPL (Villetaneuse). The characterization of new ultra-stable frequency standards under progress at LPL , such as the CO2/SF6 two-photon frequency standard in the 30-THz spectral region or Ar/I2 requires such ultra stable frequency distribution system that surpasses the phase stability of all systems previously developed.

The distortion induced by the residual amplitude modulation in FM spectroscopy is studied and a narrowband rejection technique for the suppression of the amplitude modulation is presented.

An ultra-stable fiber optic frequency link used to transfer a reference frequency signal through a standard telecommunication optical fiber from a laboratory at some 20-km distance is presented. This optical fiber is 42.5 km long and propagates a carrier at 1.55 /spl mu/m from SYRTE to LPL. The carrier is modulated by a 100 MHz signal, itself referenced to a H-maser at SYRTE.

An effective narrow-band controller based on a field programmable gate array (FPGA) is developed for the stabilization of a laser beam intensity in the megahertz range. Its principle is derived from the adaptive noise canceling techniques developed in the low-frequency region. The beam intensity is monitored by a photodiode which provides the error signal at the controller input. The canceling signal at the controller output drives the RF input power of an acousto-optic modulator (AOM) working in the linear part of its characteristic, which controls the beam intensity. The close loop system is equivalent to a notch filter tuned to a reference frequency applied to the controller. Some applications of this device in the field of high-resolution spectroscopy of iodine are presented.

Fragment production in central collisions of Xe+Sn has been systematically studied with the INDRA multidetector from 25 to 150 AMeV. The predominant role of collision dynamics is evidenced in multiple intermediate mass fragment production even at the lowest energies, around the so-called "multifragmentation threshold". For beam energies 50 AMeV and above, a promising agreement with suitably modified Antisymmetrised Molecular Dynamics calculations has been achieved. Intriguingly the same reactions have recently been interpreted as evidence for a liquid-gas phase transition in thermodynamically equilibrated systems. The universal fluctuation theory, thanks to its lack of any equilibrium hypothesis, shows clearly that in all but a tiny minority of carefully-selected central collisions fragment production is incompatible with either critical or phase coexistence behaviour. On the other hand, it does not exclude some similarity with aggregation scenarios such as the lattice-gas or Fisher droplet models., Comment: 13 pages, 9 figures; contribution to the XLth Int. Wint. Meet. on Nuclear Physics, Bormio (Italy), January 21-25 2002

Nous avons mis au point une nouvelle technique de mesure de frequence dans l'IR (autour de 30THz) utilisant le peigne d'un laser femtoseconde. La frequence optique est directement comparee a l'ecart entre 2 modes du laser femtoseconde, lequel depend uniquement du taux de repetition f r du laser femtoseconde. La frequence absolue de notre laser a CO 2 asservie sur une raie a deux photons de SF 6 a ete mesuree avec cette methode. Les premiers resultats montrent une deviation d'Allan de 3 × 10 -14 sur 1 s.

We demonstrate a new, simple technique for measuring IR frequencies near 30 THz by using a femtosecond (fs) laser optical comb and sum-frequency generation. The optical frequency is directly compared with the distance between two modes of the fs laser, and the resultant beat note is used to control this distance, which depends only on the repetition rate of the fs laser. The absolute frequency of a CO2CO2 laser stabilized onto a SF6SF6 two-photon line has been measured for the first time to the authors’ knowledge. This line is an attractive alternative to the usual saturated absorption OsO4OsO4 resonances used for the stabilization of CO2CO2 lasers. First results demonstrate a fractional Allan deviation of 3×10–143×10–14 at 1 s.

A narrow-band controller is described, based on an FPGA device for the stabilisation of laser beam intensity at 2.5 MHz. 70 dB rejection of the amplitude modulation of a laser beam is obtained. The noise rejection near 2.5 MHz is also studied.

We present petrophysical data derived from pore-network modelling of CO2-brine pore systems from the Krechba CO2 Storage Site (part of the In Salah Gas Joint Venture project operated by BP, Sonatrach and Statoil). The Carboniferous sandstone reservoir formation has relatively low permeability (c. 10 mD) and is characterized by abundant and variable cementation–mainly quartz, patchy carbonates, grain-coating chlorites and pyrite. These petrographic characteristics make obtaining measurements and estimation of single and multiphase flow properties challenging. Pore-scale modelling is an important new tool which can supplement special core analysis measurements by providing single and two phase flow functions for a range of rock and pore types. CO2/water relative permeability measurements have been carried out on four composite core plugs at reservoir conditions (95°C and 180 bars pore pressure). We have reviewed these experimental data and compared them to new predictions from several pore-scale reconstructions of the matching rock samples. First porosity, absolute permeability and formation factor were calculated and compared experimental data. Pore-networks were then extracted from the rock models and used as inputs to the simulation of CO2/water displacements. Primary drainage and waterflooding sequences were simulated to establish end-point saturations (i.e. Swi and trapped CO2 saturation), capillary pressure and relative permeability curves. Very good agreement was found between the experimental results and those derived from calculations of petrophysical parameters on rock models and multiphase flow simulations through their respective pore-networks. Calculated permeability and porosity match the values estimated from the available logs, and the calculated average cementation exponent (m) for the three reconstructed samples is 2.05, comparable with the experimental value of 1.98. Swi values obtained from the simulations range from 0.29 to 0.34, similar but slightly lower than those obtained from the steady-state experimental study–0.39 to 0.44. The simulated residual CO2 saturation ranges from 36% to 44%. The capillary trapping ensures that part of the injected CO2 will stay disconnected as isolated CO2 clusters in the pore space. These values are comparable to the residual gas saturation estimated from the experiments (from 15 to 40 %). Differences between experiments and models can be related to differences in pore types which are better defined in the pore-network models. We conclude that pore-scale modelling is able to reproduce and supplement special core analysis experiments, even when the simulations are based on relatively simple assumptions, such as non-reactive and immiscible fluids. In addition, pore-scale modelling allows the correlation of end-points with the geometry, topology and morphology of the pore space of the rock, allowing us to improve the basic understanding of CO2 trapping mechanisms in heterogeneous formations.

The transfer of ultra-stable frequencies between distant laboratories is a fundamental issue for a large number of high-sensitivity experiments in advanced metrology and fundamental physics, e.g. test of fundamental constants stability. With the recent development of single ion traps and cold atoms lattices, optical clocks are indeed expected to reach fractional frequency instability level of 10−17 or better at one day measurement time. The transmission of optical frequency standards over optical fibre has been investigated for a few years and several links of more than 100 km with instability in the 10−17 to 10−19 range have already been reported [1–4]. All of them used dedicated dark fibres in urban environment.

Une nouvelle génération d&\#8217;horloges atomiques, basées sur les transitions optiques de diverses espèces (strontium, mercure &\#8230;), permet de réaliser des références de fréquence avec une incertitude relative de quelques 10-18, surpassant ainsi déjà l&\#8217;incertitude de la définition actuelle (césium) d&\#8217;un facteur 100. L'extraordinaire résolution statistique de ces horloges optiques ouvre pour différentes disciplines des perspectives inédites : recherche d&\#8217;une variation des constantes fondamentales, tests d&\#8217;invariance de Lorentz, recherche de matière noire, géodésie chronométrique &\#8230; Pour comparer ces dispositifs, voire synchroniser leurs cycles d&\#8217;opération, les peignes de fréquence sont un outil essentiel : ils mesurent les rapports de fréquences entre horloges et sont à l&\#8217;interface avec les moyens de comparaison, comme le lien optique fibré Européen qui distribue une référence unique à 1542 nm sur le continent. Nous présenterons la capacité de mesure à base de peignes de fréquence et de liens optiques compensés mise en place à l&\#8217;Observatoire de Paris, ainsi que les résultats des comparaisons locales et internationales. Nous verrons notamment quels sont les contraintes obtenues sur une possible dérive des constantes fondamentales ou sur la matière noire. Enfin, nous discuterons des perspectives futures, comme la redéfinition de la seconde et la mission spatiale {PHARAO}-{ACES}.

International audience; The transfer of ultra-stable frequencies between distant laboratories is required by many applications from fundamental metrology to high-precision measurements. For that purpose, optical fibre links have been intensively studied over the last decade. They have demonstrated impressive results far beyond the GPS capabilities on distances up to 2000 km, thanks to an active compensation of the fiber propagation noise.Up to now, optical links has been mainly implemented between two labs, when future applications will require the development of metrological fiber networks. Towards this goal, we have developed a few techniques which allow us to distribute the ultrastable optical signal to many users simultaneously. We have designed repeater laser stations, which can be used to build a cascaded link or to distribute the ultrastable signal among two optical links [1]. These stations allow us also to compare the output end signals of two optical links. These versatile stations will be a key-component of the national metrological network currently being developed in France within the Refimeve+ project. To complete the network, we have demonstrated the extraction of the ultrastable signal from a main link and its seeding to a secondary link [2].Optical links are already been successfully used for clocks comparison. Recently we were able to compare the atomic clocks of the French and German National Metrology Institutes through two connected optical links of total length 1415-km [3]. We found that the two fully independent Sr lattice optical clocks at the two sites agreed to better than 5x10-17 limited only by the clocks uncertainties. A fractional statistical uncertainty of 3x10-17 was reached after only 1000 s averaging time, which is 10 times better, and more than four orders of magnitude faster than with any other existing frequency transfer method. The atomic fountain primary frequency standards from the two institutes were also compared and they were found to agree well within their combined uncertainties of a few 10-16 [4].

We investigated the transfer of an ultra-stable optical frequency via an optical fiber link. We achieved an instability below 6x10-18 for a distance of 86 km, and

Accurate molecular spectroscopy in the mid-infrared (mid-IR) region allows precision measurements with applications in fundamental physics. We present our on-going work towards measuring absolute vibrational frequencies of various polyatomic species — in particular methanol — around 10 μm, at an unprecedented level of accuracy, using a both ultra-stable and widely tuneable near-infrared frequency comb. We have recently been able to lock mid-IR radiation to a frequency comb stabilized to a 1.54 μm near-IR reference [1, 2]. This reference, generated at the French national metrology institute (LNE-SYRTE), is monitored against atomic frequency standards [3] and transferred to LPL via a 43-km long optical fibre [4] (see Fig. 1). This provides the ultimate frequency accuracy (potentially the 3x10−16 of the Cs fountain clock) and stability (∼10−15 after 1s of integration).

Summary form only given. The transfer of ultra-stable frequencies between distant laboratories is a fundamental issue for a wide range of high-sensitivity experiments in advanced metrology and fundamental physics. Since a few years, several experiments of optical frequency transfer were reported over dedicated fiber, with a record distance of 920 km [1] recently demonstrated in Germany. The challenge is now to build a wide frequency dissemination network connecting laboratories and research agencies in Europe.For that purpose we exploit the existing fine meshing of the Internet fiber network, connecting laboratories via the National Research Networks (NRENs). In that scheme, the ultra-stable frequency signal is copropagating simultaneously with the Internet traffic in the same fiber using a dedicated channel in a dense wavelength division multiplexing (DWDM) approach. Optical add-drop multiplexers enable to insert and extract the ultra-stable frequency signal in and from the network.We have demonstrated an ultra-stable optical link over 540 km, including 470 km of public fiber network of the French NREN (RENATER) simultaneously carrying Internet data traffic [2]. The signal bypasses four amplification stages and two Data Center Facilities using multiplexers and custom-made bidirectional erbium-doped fiber amplifiers. One specific point of this experiment is that with Internet fibers, we have a very limited control on the attenuation and noise. For instance the Erbium fiber doped bidirectional amplifiers gain is drastically limited to avoid lasing oscillations due to parasitic reflections due to connectors, splices and more fundamental Rayleigh backscattering. Nevertheless the results are comparable with those obtained with dedicated fibers. Figure 1 shows the measured fractional frequency stability, we obtain 6×10-15 at 1 s and below 10-18 after 10000 second integration time. Such optical link will enable a broad range of high-sensitivity measurements, including the search for fundamental constants variation and gravitational mapping. This concept of fiber link has been very recently used to implement a giant fiber gyroscope [3]

Ultra-high-resolution spectroscopy enables to test modern theories of fundamental physics with molecules as for instance the non conservation of parity or the stability of the electron-to-proton mass ratio. However many of these tests rely on the availability of ultra-stable and accurate laser sources emitting in the mid-infrared (MIR) where molecules exhibit rovibrational transitions. It is thus very challenging to develop a frequency stabilization scheme in the MIR which does not depend on quite rare secondary frequency references.

The dynamics and thermodynamics of phase transition in hot nuclei are studied through experimental results on multifragmentation of heavy systems (A>200) formed in central heavy ion collisions. Different signals indicative of a phase transition studied in the INDRA collaboration are presented and their consistency is stressed., 10 pages, 10 figures. Invited talk to IWM2001 (International Workshop on Multifragmentation and related topics), Catania, Italy - 28 nov-1 dec 2001

The experimental set-up for high-resolution spectroscopy of iodine at 501.7 nm, near the dissociation limit, is described. Line shapes of about 30 kHz wide are obtained with hyperfine components of the R(26)62-0 transition. Preliminary results on the absolute frequency measurements of the Ar + laser locked onto these transitions are given.

The Basel agreements and the associated European directives have made banking prudential capital contingent on their risk profile rather than on their size or turnover. The Solvency 2 Directive (hereinafter the "Directive") repeats this process for European insurers and reinsurers. It constitutes a total paradigm shift for the majority of European insurers. It defines the main regulatory principles aimed at regulating their activity and in particular determining the amount of prudential capital associated with the risks inherent to their activity.In accordance with the directive, the prudential capital corresponds in principle to an insurer with a 99.5% percentile of the change in its basic own funds over the coming year. Such a prospective risk measure requires for an insurer the ability to address two problems: a valuation problem and a simulation problem. In practice, the 99.5% percentile of the change in basic own funds is estimated using the MonteCarlo method. It is particularly sensitive to the one-year law retained for the risk factors vector. Its Monte Carlo valuation would ideally require the simulation of one year risk factor vector x and the valuation of the associated equity values. Given the significant calculation time required for numerical evaluation, this approach is in practice unsuitable. In order to circumvent this problem, the insurers have developed many approximate methods or "proxies" which make it possible to approximate the basic own funds value instantaneously. Today, these methods are rarely accompanied by error controls that would measure the simulation quality. More precisely, the methods currently used by the insurers do not make it possible to control naturally the approximation error generated by the use of the proxy model instead. The proposed error checks are therefore always empirical and too approximate.In order to solve this problematic, we propose, in a first part of this thesis, a new method of constructing the proxy that is both resource-efficient and offers rigorous error control. The second part of this thesis aims at applying the extreme value theory to the prudential capital estimate when information on the covariate is available. In particular, when the covariate is high dimensional, we are confronted with the problem of the curse of dimensionality, which translates into a decrease in the fastest possible convergence rates of estimators of the regression function to their target curve. This problem refers to the phenomenon where the volume of the covariate increases so rapidly that available data become sparse. To obtain a statistically reliable result, the amount of data needed to support the result often increases exponentially with dimensionality, which is generally problematic in many practical applications. To overcome this estimation problem, we propose a new efficient evaluation methodology by combining the generalized additive model and the sparse group lasso method.; Les accords de Bale et les directives européennes associées ont conduit à conditionner les capitaux prudentiels des banques à leur profil de risqué plutôt qu’à leur taille ou chiffre d’affaire. La directive Solvabilité 2 (ci-après la "directive") répète ce processus pour les assureurs et réassureurs européens. Elle constitue un changement total de paradigme pour la majorité des assureurs européens. Elle définit les grands principes réglementaires visant à encadrer leur activité et en particulier à déterminer le montant des capitaux prudentiels associés aux risques inhérents à leur activité.Conformément à la directive, le capital prudentiel correspond en principe pour un assureur au quantile à 99.5% de la variation de ses fonds propres sur l’année à venir. Une telle mesure de risque prospective requiert pour un assureur la capacité d’adresser deux problèmes : un problème de valorisation et un problème de simulation. En pratique, le quantile à 99.5% de la variation de ses fonds propres est estimépar la méthode de Monte-Carlo. Il est particulièrement sensible à la loi jointe à un an retenue pour le vecteur de facteurs de risque x. Son évaluation par méthode de Monte-Carlo nécessiterait idéalement de simuler m réalisations du facteur de risque x à un an et d’évaluer les valeurs des fonds propres associées. Compte tenu du temps de calcul important nécessaire à l’évaluation numérique, cette approche s’avère en pratique inadaptée. De manière à contourner ce problème, les opérationnels ont mis au point de nombreuses méthodes d’approximation ou « proxys » qui permettent d’en approximer la valeur de manière instantanée. Aujourd’hui, ces méthodes sont rarement accompagnées de contrôles d’erreur qui permettraient d’en mesurer la qualité. Plus précisément, les méthodes actuellement utilisées par les opérationnels ne permettent pas de contrôler naturellement l’erreur d’approximation engendrée par l’utilisation du modèle proxy en lieu. Les contrôles d’erreur proposés sont donc toujours empiriques et trop approximatifs.Afin de résoudre cette problématique, nous proposons, dans une première partie de cette thèse, une nouvelle méthode de construction du proxy à la fois économe en ressources informatiques et offrant un contrôle d’erreur rigoureux. La deuxième partie de cette thèse a pour l’objectif d’appliquer la théorie de la valeur extrême à l’estimation du capital prudentiel lorsque l’information sur la covariable est disponible. En particulier, lorsque la covariable est de grande dimension, nous sommes confrontés au problème de la "curse of dimensionality", qui se traduit par une diminution des taux de convergence les plus rapides possibles des estimateurs de la fonction de régression vers leur courbe cible. Ce problème fait référence au phénomène où le volume de la covariable augmente si rapidement que les données disponibles deviennent rares. Pour obtenir un résultat statistiquement fiable, la quantité de données nécessaire à l’appui du résultat augmente souvent de manière exponentielle avec la dimensionnalité, ce qui est généralement problématique dans de nombreuses applications pratiques. Pour surmonter ce problème d’estimation, nous proposons une nouvelle méthodologie d’évaluation efficace en combinant le modèle additif généralisé et la méthode de sparse group lasso.

In this thesis, we study duration models in the context of the analysis of contract termination time in health insurance. Identified from the 17th century and the original work of Graunt J. (1662) on mortality, the bias induced by the censoring of duration data observed in this context must be corrected by the statistical models used. Through the problem of the measure of dependence between successives durations, and the problem of the prediction of contract termination time in insurance, we study the theoretical and practical properties of different estimators that rely on a proper weighting of the observations (the so called IPCW method) designed to compensate this bias. The application of these methods to customer value estimation is also carefully discussed.; Dans cette thèse, nous nous intéressons aux modèles de durée dans le contexte de la modélisation des durées de résiliation de contrats d’assurance santé. Identifié dès le 17ème siècle et les études de Graunt J. (1662) sur la mortalité, le biais induit par la censure des données de durée observées dans ce contexte doit être corrigé par les modèles statistiques utilisés. À travers la problématique de la mesure de la dépendance entre deux durées successives, et la problématique de la prédiction de la durée de résiliation d’un contrat d'assurance, nous étudions les propriétés théoriques et pratiques de différents estimateurs basés sur une méthode de pondération des observations (méthode dite IPCW) visant à corriger ce biais. L'application de ces méthodes à l'estimation de la valeur client en assurance est également détaillée.

In this thesis, we study duration models in the context of the analysis of contract termination time in health insurance. Identified from the 17th century and the original work of Graunt J. (1662) on mortality, the bias induced by the censoring of duration data observed in this context must be corrected by the statistical models used. Through the problem of the measure of dependence between successives durations, and the problem of the prediction of contract termination time in insurance, we study the theoretical and practical properties of different estimators that rely on a proper weighting of the observations (the so called IPCW method) designed to compensate this bias. The application of these methods to customer value estimation is also carefully discussed.; Dans cette thèse, nous nous intéressons aux modèles de durée dans le contexte de la modélisation des durées de résiliation de contrats d’assurance santé. Identifié dès le 17ème siècle et les études de Graunt J. (1662) sur la mortalité, le biais induit par la censure des données de durée observées dans ce contexte doit être corrigé par les modèles statistiques utilisés. À travers la problématique de la mesure de la dépendance entre deux durées successives, et la problématique de la prédiction de la durée de résiliation d’un contrat d'assurance, nous étudions les propriétés théoriques et pratiques de différents estimateurs basés sur une méthode de pondération des observations (méthode dite IPCW) visant à corriger ce biais. L'application de ces méthodes à l'estimation de la valeur client en assurance est également détaillée.

The continuous enhancement of computer performances during the last decades has favored a widespread application of statistical learning theory in multiple domains. Actuaries, longstanding statistical experts, notably turn more and more frequently to these new algorithms for the evaluation of risks they are facing. Thus, in this thesis, we examine how the integration of methods derived from statistical learning can contribute to the development of actuarial science and risk management through the study of three independent problematics, preliminarily presented in a general introduction.The first two chapters propose new mortality forecasting models within the longevity risk evaluation framework that insurance companies and pension funds encounter. Chapter 1 focuses on the singe population case, while Chapter 2 extends the study to the multi-population. In both situations, high-dimensionality appears to be a major concern. We tackle this issue thanks to a penalized vector autoregressive (VAR). This model is directly applied to the mortality improvement rates in the first chapter, and to the time series derived from Lee-Carter's model fits in the second one. The elastic-net penalization preserves the large freedom in the spatio-temporal dependence structure offered by the VAR, while remaining sparse in terms of estimated parameters, thereby avoiding overfitting.In Chapter 3, we analyze lapse risk in life insurance policies through the use of supervised classification algorithms. We apply, among others, support vector machine (SVM) and extreme gradient boosting (XGBoost). In order to compare performances from one classifier to another, we adopt an economic point of view derived from the marketing literature and based on potential profits of a retention campaign. We insist on the importance of the loss function retained in the statistical learning algorithms according to the pursued objective: the use of a loss function linked to the performance measure leads to a significant enhancement in the application of the XGBoost in our study.In the last Chapter, in the financial risks' management framework, we study agricultural commodity price dynamics throughout specific trading sessions where governmental reports, that contain valuable information for the agents, are released. We examine the potential of open data, in particular the satellite data on vegetation index made available thanks to the NASA, for market reactions forecast. We then suggest some improvements to be considered before an operational implementation of this forecasting method., L'augmentation continuelle des performances informatiques des dernières décennies a permis une application répandue de la théorie de l'apprentissage statistique dans de multiples domaines. Les actuaires, experts historiques des statistiques, se tournent en particulier de plus en plus fréquemment vers ces algorithmes novateurs pour l'évaluation des risques auxquels ils sont confrontés. Ainsi, dans cette thèse, nous examinons comment l'intégration de méthodologies issues de l'apprentissage statistique peut contribuer au développement des sciences actuarielles et de la gestion des risques au travers de l'étude de trois problématiques indépendantes, présentées au préalable dans une introduction générale.Les deux premiers chapitres proposent de nouveaux modèles de projection de mortalité dans le cadre de l'évaluation du risque de longévité porté par les compagnies d'assurances ou les fonds de pensions. Le Chapitre 1 s'attarde sur le cas où une seule population est étudiée, alors que le Chapitre 2 étend l'analyse à la multi-population. Dans les deux situations, la problématique de la grande dimension apparait centrale et nous l'abordons à l'aide d'un vecteur autorégressif (VAR) pénalisé. Ce modèle est appliqué directement sur les taux d'amélioration de mortalité dans le premier chapitre, et sur les séries temporelles issues de l'estimation d'un modèle Lee-Carter pour le second. La pénalisation elastic-net permet de garder la grande liberté de structure de dépendance spatio-temporelle qu'offre le VAR tout en restant parcimonieux dans le nombre de paramètres, et donc éviter le surapprentissage.Dans le Chapitre 3 nous analysons le risque de rachat des contrats d'assurance vie par l'utilisation d'algorithmes de classification supervisée. Nous y appliquons entre autres le séparateur à vaste marge (SVM) et l'extreme gradient boosting (XGBoost). Afin de comparer les performances des différents classificateurs nous adoptons une vision économique issue de la littérature du marketing et basée sur les profits potentiels d'une campagne de rétention. Nous insistons sur l'importance de la fonction de perte retenue dans les algorithmes d'apprentissage statistique suivant l'objectif recherché : l'utilisation d'une fonction de perte en lien avec la mesure de performance amène une amélioration significative dans l'application de l'XGBoost dans notre étude.Enfin, dans le cadre de la gestion des risques financiers, nous étudions les dynamiques des prix agricoles lors de sessions de bourse particulières où des rapports gouvernementaux, contenant des informations précieuses pour les agents, sont publiés. Nous examinons le potentiel des données en libre accès, en particulier les images satellites d'indice de végétation rendues disponibles par la NASA, pour la prédiction des réactions des marchés. Nous proposons alors des pistes d'amélioration à considérer pour une mise en ÷uvre pratique de cette méthodologie d'enrichissement des données dans la gestion des risques.

The nuclear matter equation of the state is an essential tool in the description of heavy ion collisions,but also in the description of the formation of astrophysical objects or phenomena (neutron star structure,neutron stars fusion). Establishing the nuclear matter equation of state requires a proper definition of thethermodynamic conditions (density, temperature, proton/neutron asymmetry) in which the system evolves.In this work, we address the issue of equilibration reached in heavy ion collisions, in terms of energy andisospin. To do this, we use the experimental database of the INDRA array built by the collaboration over thepast 25 years, focusing on central collisions in the Fermi energy domain, between 10 and 100 MeV/nucleon.In this document, we present how, with the help of dedicated simulations, it has been possible to link thestopping power of nuclear matter to the in-medium nucleon-nucleon cross-section. We also provide someanswers regarding isospin transport in central collisions using the isobaric ratios A = 3 based on the tritonsand helium-3 particles. These different results allow us to highlight the new experimental apparatus devel-loped by the INDRA and FAZIA collaborations : the FAZIA array. The latter is the result of a ten-yearperiod of research and development, resulting in an array embedded its digital electronic under vacuum, withincreased identification performance (measurement of the Z charge and A mass up to Z = 25) compared tothe previous generations arrays., L’équation d’état de la matière nucléaire est un outil primordial dans la description des collisions entreions lourds, mais également dans la description de la formation d’objets ou de phénomènes astrophysiques(structure des étoiles à neutrons, fusion d’étoiles à neutron). Établir l’équation d’état de la matière nucléairerequiert de définir de manière précise les conditions thermodynamiques (densité, température, asymétrie pro-ton/neutron) dans lesquelles le systèmes évolue. Dans ce travail, nous abordons la problématique de l’étatd’équilibration maximal qui est atteint dans les collisions entre ions lourds, en terme d’énergie et d’isospin.Pour cela, nous utilisons la base de données expérimentale du multi-détecteur INDRA construite par lacollaboration au cours de ces 25 dernières années, en nous intéressant plus particulièrement aux collisionscentrales dans le domaine de l’énergie de Fermi, entre 10 et 100 MeV/nucléon. Nous présentons ainsi dansce document, comment à l’aide de simulation dédiées, il nous a été possible de relier le pouvoir d’arrêt de lamatière nucléaire à la section efficace de collision nucléon-nucléon dans la matière nucléaire. Nous apportonségalement des éléments de réponse au regard du transport de l’isospin dans les collisions centrales à l’aidedes rapports isobariques A = 3 construits à partir des tritons et des hélium-3. Ces différents résultats nouspermettent de mettre en avant le nouveau dispositif expérimental mis en place par les collaborations INDRAet FAZIA : le multi-détecteur FAZIA. Ce dernier est le résultat d’une période de recherche et développementde dix ans, ayant abouti à un multi-détecteur embarquant son électronique numérique sous vide, avec desperformances d’identification accrues (mesure de la charge Z et de la masse A jusqu’à Z = 25) par rapportaux multi-détecteurs des générations précédentes.

The nuclear matter equation of the state is an essential tool in the description of heavy ion collisions,but also in the description of the formation of astrophysical objects or phenomena (neutron star structure,neutron stars fusion). Establishing the nuclear matter equation of state requires a proper definition of thethermodynamic conditions (density, temperature, proton/neutron asymmetry) in which the system evolves.In this work, we address the issue of equilibration reached in heavy ion collisions, in terms of energy andisospin. To do this, we use the experimental database of the INDRA array built by the collaboration over thepast 25 years, focusing on central collisions in the Fermi energy domain, between 10 and 100 MeV/nucleon.In this document, we present how, with the help of dedicated simulations, it has been possible to link thestopping power of nuclear matter to the in-medium nucleon-nucleon cross-section. We also provide someanswers regarding isospin transport in central collisions using the isobaric ratios A = 3 based on the tritonsand helium-3 particles. These different results allow us to highlight the new experimental apparatus devel-loped by the INDRA and FAZIA collaborations : the FAZIA array. The latter is the result of a ten-yearperiod of research and development, resulting in an array embedded its digital electronic under vacuum, withincreased identification performance (measurement of the Z charge and A mass up to Z = 25) compared tothe previous generations arrays.; L’équation d’état de la matière nucléaire est un outil primordial dans la description des collisions entreions lourds, mais également dans la description de la formation d’objets ou de phénomènes astrophysiques(structure des étoiles à neutrons, fusion d’étoiles à neutron). Établir l’équation d’état de la matière nucléairerequiert de définir de manière précise les conditions thermodynamiques (densité, température, asymétrie pro-ton/neutron) dans lesquelles le systèmes évolue. Dans ce travail, nous abordons la problématique de l’étatd’équilibration maximal qui est atteint dans les collisions entre ions lourds, en terme d’énergie et d’isospin.Pour cela, nous utilisons la base de données expérimentale du multi-détecteur INDRA construite par lacollaboration au cours de ces 25 dernières années, en nous intéressant plus particulièrement aux collisionscentrales dans le domaine de l’énergie de Fermi, entre 10 et 100 MeV/nucléon. Nous présentons ainsi dansce document, comment à l’aide de simulation dédiées, il nous a été possible de relier le pouvoir d’arrêt de lamatière nucléaire à la section efficace de collision nucléon-nucléon dans la matière nucléaire. Nous apportonségalement des éléments de réponse au regard du transport de l’isospin dans les collisions centrales à l’aidedes rapports isobariques A = 3 construits à partir des tritons et des hélium-3. Ces différents résultats nouspermettent de mettre en avant le nouveau dispositif expérimental mis en place par les collaborations INDRAet FAZIA : le multi-détecteur FAZIA. Ce dernier est le résultat d’une période de recherche et développementde dix ans, ayant abouti à un multi-détecteur embarquant son électronique numérique sous vide, avec desperformances d’identification accrues (mesure de la charge Z et de la masse A jusqu’à Z = 25) par rapportaux multi-détecteurs des générations précédentes.

The main purpose of this thesis is to explore several approaches for studying multivariate censored data: nonparametric estimation of the joint distribution function, modeling dependence with copulas and k-clustering for the exploratory analysis. Chapter 1 presents the general framework and the contributions of this thesis. Chapter 2 deals with the estimation of the joint distribution function of two censored variables in a simplified survival model in which the difference between two censoring variables is observed. We provide a new nonparametric estimator of the joint distribution function and we establish the asymptotic normality of the integrals with respect to its associated measure. Chapter 3 is devoted to nonparametric copula estimation under bivariate censoring. We provide a discrete and two smooth copula estimators along with two estimators of its density. The discrete estimator can be seen as an extension of the empirical copula under censoring. Chapter 4 provides a new exploratory approach for censored data analysis. We consider a multivariate configuration with one variable subjected to censoring and the others completely observed. We extend the probabilistic k-quantization method in the case of random vector with one censored component. The definitions of the empirical distortion and of empirically optimal quantizer are generalized in presence of one-dimensional censoring. We study the asymptotic properties of the distortion of the empirically optimal quantizer and we provide a non-asymptotic exponential bound for the rate of convergence. Our results are then applied to construct a new two-step clustering algorithm for censored data.; L'objectif de cette thèse est d'explorer plusieurs approches pour l'étude des données censurées multivariées, à savoir l'estimation non paramétrique de la fonction de répartition jointe, la modélisation de dépendance par les modèles de copules et l'étude exploratoire par des méthodes de clustering. Le Chapitre 1 introduit le contexte général de cette thèse ainsi que ses contributions. Le Chapitre 2 est consacré à l'estimation de la distribution jointe des deux variables censurées dans le cadre d'un modèle de durée simplifié où la différence entre deux variables de censure est observée. Un nouvel estimateur non paramétrique de la fonction de répartition jointe y est introduit. La normalité asymptotique a été démontrée, pour les intégrales par rapport à la mesure définie par cet estimateur. Le Chapitre 3 est dédié à la problématique de l'estimation non paramétrique de la copule bivariée, à partir d'un échantillon de données censurées. La copule est d'abord estimée par une fonction discrète qui peut être interprétée comme une extension de la copule empirique en présence de censure, puis par ses versions lisses. Les propriétés asymptotiques et des applications de des estimateurs ont été considérées. Le Chapitre 4 présente une approche exploratoire pour l'étude de données censurées. Plus précisément, une configuration multivariée est considérée où une variable est une durée sujette à la censure, et toutes les autres variables sont observées. Sous ces conditions, une nouvelle méthode de quantification de la loi jointe est introduite. La méthode est étudiée théoriquement et appliquée à la construction d'un algorithme de clustering pour des observations censurées.

The main purpose of this thesis is to explore several approaches for studying multivariate censored data: nonparametric estimation of the joint distribution function, modeling dependence with copulas and k-clustering for the exploratory analysis. Chapter 1 presents the general framework and the contributions of this thesis. Chapter 2 deals with the estimation of the joint distribution function of two censored variables in a simplified survival model in which the difference between two censoring variables is observed. We provide a new nonparametric estimator of the joint distribution function and we establish the asymptotic normality of the integrals with respect to its associated measure. Chapter 3 is devoted to nonparametric copula estimation under bivariate censoring. We provide a discrete and two smooth copula estimators along with two estimators of its density. The discrete estimator can be seen as an extension of the empirical copula under censoring. Chapter 4 provides a new exploratory approach for censored data analysis. We consider a multivariate configuration with one variable subjected to censoring and the others completely observed. We extend the probabilistic k-quantization method in the case of random vector with one censored component. The definitions of the empirical distortion and of empirically optimal quantizer are generalized in presence of one-dimensional censoring. We study the asymptotic properties of the distortion of the empirically optimal quantizer and we provide a non-asymptotic exponential bound for the rate of convergence. Our results are then applied to construct a new two-step clustering algorithm for censored data.; L'objectif de cette thèse est d'explorer plusieurs approches pour l'étude des données censurées multivariées, à savoir l'estimation non paramétrique de la fonction de répartition jointe, la modélisation de dépendance par les modèles de copules et l'étude exploratoire par des méthodes de clustering. Le Chapitre 1 introduit le contexte général de cette thèse ainsi que ses contributions. Le Chapitre 2 est consacré à l'estimation de la distribution jointe des deux variables censurées dans le cadre d'un modèle de durée simplifié où la différence entre deux variables de censure est observée. Un nouvel estimateur non paramétrique de la fonction de répartition jointe y est introduit. La normalité asymptotique a été démontrée, pour les intégrales par rapport à la mesure définie par cet estimateur. Le Chapitre 3 est dédié à la problématique de l'estimation non paramétrique de la copule bivariée, à partir d'un échantillon de données censurées. La copule est d'abord estimée par une fonction discrète qui peut être interprétée comme une extension de la copule empirique en présence de censure, puis par ses versions lisses. Les propriétés asymptotiques et des applications de des estimateurs ont été considérées. Le Chapitre 4 présente une approche exploratoire pour l'étude de données censurées. Plus précisément, une configuration multivariée est considérée où une variable est une durée sujette à la censure, et toutes les autres variables sont observées. Sous ces conditions, une nouvelle méthode de quantification de la loi jointe est introduite. La méthode est étudiée théoriquement et appliquée à la construction d'un algorithme de clustering pour des observations censurées.

Heavy ion collisions are one of tools for studying nuclear system far away from itsequilibrium state. This work concerns the most violent collisions in the Ni + Au system for incidentenergies ranging from 32 up to 90 AMeV. These events were detected with the multidetector INDRAand selected by the Principal Component Analysis (multidimensional analysis). This method classifiesthe events according their detection features and their degree of dissipation. We observed twodesexcitation mechanisms: a fusion/fission - evaporation process and a multifragmentation process.Those two coexist from 32 to 52 AMeV whereas only one subsists at 90 AMeV. For those twomechanisms, an component was observed which seems to be linked to the initial phase of the reaction.The energy fluctuations of this component leads to variations in the energy deposit which determinesthe desexcitation of the system.The experimental multifragmentation data of the Ni + Au system (52 and 90 AMeV) were compared tothe predictions of a statistical model and to the experimental data of the system Xe +Sn at 50 AMeV(also detected with INDRA). These comparisons show the lack of collective radial energy forfragments (Z>10) in the Ni + Au system, and show that the degree of multifragmentation depends ofthe thermal excitation energy. Mean kinetic energies of particles and lights fragments (Z, Les collisions d'ions lourds sont un des moyens de perturber fortement un système nucléaire etde l'amener loin de son état d'équilibre. Ce mémoire concerne les réactions les plus violentes dusystème Ni + Au de 32 à 90 AMéV. Ces événements sont détectés par le multidétecteur INDRA etisolés avec l'Analyse en Composantes Principales (analyse multidimensionnelle). Elle permet declasser les événements selon leurs caractéristiques de détection (sans critère de complétude aupréalable) et de leur degré de dissipation. Deux mécanismes de désexcitation ont été mis en évidence :l'un de type fusion/fission suivi d'une evaporation (type basse énergie) et l'autre de typemultifragmentation. Ces deux mécanismes sont coexistants à 32 et 52 AMeV alors qu'un seul subsisteà 90 AMeV. Les deux mécanismes de désexcitation sont accompagnés d'une composante reliée à laphase initiale de la réaction. Les fluctuations en énergie liées à cette composante entraînent desvariations du dépôt en énergie dans le système et déterminent l'un ou l'autre des mécanismesobservés.Les données expérimentales de multifragmentation du système Ni + Au (52 et 90 AMeV) ont étéconfrontées aux prédictions d'un modèle statistique et aux données de multifragmentation issues dusystème symétrique Xe + Sn à 50 AMeV (mesuré avec INDRA). Ces comparaisons ont mis enévidence l'absence d'énergie collective radiale pour les fragments (Z>10) pour le système Ni + Au, etont montré que le degré de fragmentation dépend principalement de l'énergie d'excitation thermique.Les énergies cinétiques moyennes des particules et des fragments légers (Z