1. Scientific report: auto-evaluation of the ADAMIS team a. - - PDF document

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

1

• 1. Scientific report: auto-evaluation of the ADAMIS team
• a. Activities and results

ADAMIS is an interdisciplinary research group aiming at more effective, advanced, and robust scientific exploitation and interpretation of existing and anticipated astrophysical and cosmological data sets in particular in the view of their rapidly growing sizes and complexity. ADAMIS approach to reaching such goals is to perform research at the interface of physics, statistics, computer science, signal processing, and applied mathematics with an aim of not only capitalizing on the most recent developments in those areas but of engaging directly in interdisciplinary research to find novel, more comprehensive and robust solutions. This often requires simultaneous advances in multiple science areas and consequently interdisciplinary collaborations involving researchers from diverse fields and which the group members strive to instigate and coordinate. Two main themes of the group research are a development of novel data analysis techniques as driven by needs of actual experiments and their application to some of the most exciting current and forthcoming data, and advanced numerical simulations of complex astrophysical phenomena, e.g. compact sources. These two themes are closely intertwined, as simulated, high-quality mock data sets are necessary to validate and tests data analysis algorithms. They also share a similar technical background. The group’s focus is on 3 science areas: CMB data analysis, GW data analysis, and simulations. CMB data analysis PLANCK ADAMIS has played and continues on playing an important role within the Planck HFI collaboration. Permanent members of ADAMIS (Cardoso, Delabrouille, LeJeune, Stompor) together with students and postdocs have worked over the years on a number of aspects of the mission often in leading and coordinating roles as outlined below. They are HFI Core Team members and have Planck Scientists status within the collaboration. ADAMIS has been a major player on French and international arena in the foreground component separation effort. Two of the ADAMIS conceived and developed numerical codes, SMICA and NILC (Cardoso, Delabrouille, LeJeune), have been retained in a final selection of 4 codes (out of more than dozen in the running) to be used for final Planck

• analysis. Cardoso and LeJeune lead the effort of applying these codes to the actual Planck data.

The group researchers (Stompor) have also contributed in a major way to the development of the Planck HFI map- making codes, developing, together with the group at LAL, a polarized destriper code, polkaPIX, (Tristram et al, 2011, [in2p3-00604975]), which is principal map-making work horse software of the HFI Data Processing Center used to produce official Planck products as well as internal releases. Complementary, Stompor has been one of the co-authors

• f the maximum likelihood map-making code, MADmap, (Cantalupo et al, 2011, [in2p3-00517907]), which is the

leading code of this type used for Planck at the US Data Center. (Incidentally, and notably, the code has also become the backbone of the on-board map-maker of the Hershel SPIRES instrument.)

• Fig. 1. An example demonstrating the performance of the component separation code, SMICA, on a simulated PLANCK data set. Left panel shows the recovered total intensity

CMB map. Right panel depicts the corresponding power spectrum (in yellow) as compared to the input one (black line) as well as residual noise and foreground contamination

• levels. The low level of contamination of the SMICA CMB map is one of the criteria that have led to the selection of SMICA for the final analysis of the Planck data.

Delabrouille has led the Planck-wide effort of modeling the sky in the microwave band, so-called Planck Sky Model, initially for the purpose of the Planck data analysis and later to become one of the Planck products. The preprint detailing the work has been just made public (Delabrouille et al, 2012, preprint) and the software has been released. Delabrouille also coordinated the work of the Planck Working Group 2 devoted to the component separation problem. Cardoso co-led the effort of a challenge-like comparison of Planck component separation codes (Leach et al, 2008, [in2p3-00294140], including Delabrouille and Stompor) - a work, which has become since a reference in the field. Cardoso and Stompor have contributed to a development of the power spectrum estimators for Planck HFI, such as the Maximum Likelihood techniques, CLICK (with IAP), and pseudo-spectral ones, XPOL/XPURE (with LAL). SMALL-SCALE CMB POLARIZATION EXPERIMENTS Stompor is a senior investigator on two cutting-edge, US-led small-scale experiments: balloon-borne EBEX and ground-based POLARBEAR. The experiments aim at the detection and later characterization of the elusive CMB B- mode polarization signal and thus providing a tantalizing hint in favor of inflation. They also promise to provide an exquisite constraint on the total mass scale of neutrinos potentially deciding the nature of their mass hierarchy. Stompor has instigated and coordinated a French involvement in those, securing funding from INSU and IN2P3 PICS

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

2

programs, France-Berkeley Fonds and Marie Curie Re-Integration program (FP6), to permit for exchanges of young

• researchers. ADAMIS (co)organized multiple internal data analysis workshops and collaboration meetings. The ADAMIS

EBEX (Stompor) and POLARBEAR collaborators (Stompor and PhD students Errard and Fabbian) contributed essential

• Fig. 2. Examples of preliminary data from the POLARBEAR experiment: left panel - a total intensity map with superimposed polarization field of the supernova remnant TauA.

Located at the heart of the Crab Nebula and polarized due to synchrotron emission, it is observed regularly by the POLARBEAR experiment and used as a calibrator for the polarization angles of its focal plane ; right panel - a map of the emission of a H II region of our Galaxy, where young stars are forming, as observed by the POLARBEAR Telescope during its commissioning phase in March 2012 and corresponding to roughly 30 mins of observation time.

parts of the data analysis pipeline for these experiments including component separation techniques (Stompor et al, 2009, [in2p3-00517908]) and power spectrum estimators (Grain et al, 2009, 2012, [in2p3-00434062]). They have also performed first realistic estimation of performance of such experiments as far as detecting the primordial, inflationary signal is concerned, in the presence of the foregrounds and the E-to-B leakage effect (Stivoli et al 2010 [in2p3-00535987], Fantaye et al 2011 [in2p3-00618398]). ANALYSIS OF PUBLICLY AVAILABE DATA SETS Delabrouille, Cardoso, and collaborators produced a superior, high resolution, low foreground contamination, full- sky CMB map from the WMAP 3-year data and studied its properties (Delabrouille et al 2009, [in2p3-00709909]). The map has become a basis for many WMAP re-analysis and follow-up investigations. More recently, Delabrouille and a postdoc (Basak), benefiting from the tools developed as part of the ADAMIS work, have preformed a unique re-analysis of the WMAP 7-year data improving in multiple ways over that performed by the

• riginal team and resulting in an improved, high quality data product in a form of a cleaner more reliable CMB map of

the sky (Basak & Delabrouille 2012, [in2p3-00709880], in2p3-00709880]). Delabrouille and Cardoso studied the foregrounds in the WMAP data (Ghosh etal 2011, [in2p3-00709898]).

• Fig. 3. This figure illustrates component separation performed on the WMAP data (the K-band map, at 22GHz, is displayed in the top left panel). A method developed in the

ADAMIS group is used to first estimate in a near-optimal way the Cosmic Microwave Background signal present in the map. This method makes use of constrained linear combinations of all WMAP observations (5 frequency channels), using a decomposition of the data on a frame of spherical needlets (wavelets on the sphere). The CMB map (top right) shows no sign of contamination by foreground emission. It is subtracted from the K-band map to yield a CMB-free map of residuals (bottom-left) which is then filtered to reduce the contamination by noise, yielding a low-noise map of foreground emission in which both compact sources and diffuse emission from galactic synchrotron are clearly seen (bottom right).

NOVEL TECHNIQUES AND ALGORITHMS Faÿ et al (2008), [in2p3-00709912] have proposed a novel needlet based pseudo power spectrum estimator for total intensity data. This technique is particularly suitable for experiments with partial sky coverage. Stompor et al (2009) [in2p3-00517908] rephrased the standard parametric Maximum Likelihood component separation approach to allow it to be performed into two separate and computationally feasible steps. The approach, permitting statistically sound error propagation, has become a stepping stone for a number of performance studies evaluating the impact of the foregrounds on the output of the CMB polarization experiments.

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

3

Grain et al (2009)[in2p3-00434062] (with Stompor) presented and validated a fully consistent implementation of the pseudo-power spectrum technique applicable to the polarized data and controlling the so-called E-to-B leakage. Cardoso and Delabrouille with a postdoc (Remazeilles) have generalized standard internal linear combination component separation algorithm by allowing to introduce constraints on frequency scaling of some of the components included in the mixture and lossless compression of the information, if the number of components is different than the number of available frequency channels (Remazeilles et al, 2011, [in2p3-00709894], [in2p3-00709890]). Delabrouille with a student (Castex) have proposed a new power spectrum estimator applicable to ILC cleaned CMB maps (Dick et al, 2012, preprint). Grigori et al. 2012, (accepted for a publication), including Stompor, presented a new iterative algorithm to solve maximum likelihood map-making problem. It offers a speed-up of a factor of 5-6 over the standard solvers. The paper, a result of the ANR-MIDAS’09 project, has been selected for a presentation at the very prestigious Supercomputing’12 meeting this fall with a paper acceptance rate of 21%. It was also a forerunner (as one of roughly dozens of papers out

• f more than 500) for the best paper nomination.

The project ANR-MIDAS’09 (coordinated by Stompor and LeJeune and involving Rogier, Cargemel, Dauvergne) released two open-source numerical libraries: (1) S2HAT for massively parallel spherical harmonic transforms on diverse types of supercomputer architectures and (2) MIDAPACK for CMB data analysis toolbox on parallel computers. FUTURE CMB POLARIZATION SATELLITES Delabrouille and Stompor have been involved in preparations of two CMB polarization satellite mission proposals submitted in answer to the ESA Cosmic Vision calls and referred to as B-Pol (2008) and COrE (2010). Delabrouille led the COrE study related to the effects of the foregrounds on the scientific performance of the mission, while Stompor was responsible for the overall data analysis overview for both B-Pol and COrE. Delabrouille and Stompor were invited to participate in US-led CMBpol mission study and to participate in the CMBpol Workshop at Fermilab in 2009. They are members of an Inflationary Physics Scientific Advisory Group, an advisory body to NASA decisions concerning future CMB polarization satellites in the US. Delabrouille was a co-organizers of the ‘Beyond COrE’ workshop (Paris, June 2012) devoted to the planning of future CMB polarization satellite mission. Delabrouille and Stompor (with students Betoule and Errard respectively) have performed one of the most in-depth studies of the impact of the polarized foregrounds on the performance of the CMB measurements (Betoule et al 2009, [hal-00350898] Errard & Stompor 2012, [in2p3-00692321]). Stompor together with a student (Errard) and an INRIA post-doc (Stivoli) have proposed a unique framework for optimizing the hardware of such experiments to ensure the best achievable performance, while conforming with hardware constraints (Errard et al, 2011, [in2p3-00622812]). GW data analysis The ADAMIS group has been involved in a number of activities related to gravitational wave (GW) astronomy with an direct involvement in both the ground-based detectors Virgo and LIGO, and in the space-based mission eLISA/NGO. We also contribute to future projects such as the Einstein Telescope and participate to the analysis of Pulsar Timing data. GROUND-BASED DETECTORS The first generation of ground-based interferometric GW detectors including the French-Italian Virgo and the American LIGO has just completed recently their third and last science data taking. A large fraction of the volume of a bit less than 2 years of data has been searched for a wide range of expected signals. The results are presented in about 20 published journal articles, with another 20 still up-coming. The ADAMIS group has contributed to this achievement at various levels. We took part to the joint analysis of Virgo and LIGO data with that of the high-energy neutrino telescope ANTARES (PhD thesis under way on this subject). We developed a specific analysis pipeline able to

• Fig. 4. This figure compares the sensitivity of chirplet-based detection algorithm (b) with standard wavelet-based methods (a). The sensitivity is expressed in terms of the

maximum distance (horizon) at which a coalescing binary of compact objects of mass $M_1$ and $M_2$ can be detected in LIGO/Virgo data (at a fixed false-alarm rate). The diagram in (c) shows the ratio of the two horizon distances, which demonstrates an potential improvement of 30% over the whole parameter space.

process the GW data in coincidence in time and direction with neutrino events. We also took part to the first optical follow-up program consisting in sending alerts to robotic telescopes after low-latency pipelines have identified an interesting GW candidate event. We contributed to the global implementation of the alert system and are currently finalizing the search for optical transient in the images that have been collected. Finally, thanks to the introduction

• f frequency-modulated wavelets (chirplets) into one of the standard burst search algorithms, we demonstrated the

improved performances of this algorithm for binary merger waveforms.

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

4

We are also involved in the external review of Advanced Virgo detector through the participation (Porter) to the Virgo Science and Technology Advisory Committee (STAC) since 2010, and the EGO External Computing Committee since 2012, which are charged with ensuring the technological, scientific and computational readiness of the project. Both Chassande-Mottin and Porter are members of the Einstein Telescope Science Team and contributed to the ET Concept Design document submitted to FP7 in 2011. SPACE-BASED DETECTOR eLISA/NGO During 2011, Petiteau and Porter coordinated a science investigation taskforce, put together by ESA to study the scientific impact of different possible configurations for the eLISA/NGO mission (new European lead mission based on the old LISA concept). The taskforce involved scientists from approximately 15 different institutions on both sides of the Atlantic. The results of this taskforce can be found in the eLISA/NGO Yellow Book and in the paper Amaro-Seoane, et al, (2012) [hal-00730227]). Petiteau and Porter have also been members of the Mock LISA Data Challenge (MLDC) (e.g., Babak et al, 2010, [in2p3-00706985]) taskforce since 2008 and 2006 respectively. For eLISA, it is assumed that there will be approximately 60 million GW sources in the data-set at any one time. This large number of sources, and the finite mission duration, requires the existence of source search and resolution algorithms well before the mission launches. We currently have algorithms that search for supermassive black hole (SMBH) binaries, extreme mass ratio inspirals (EMRIs), galactic binaries and cosmic strings. These algorithms range from the simple (time-frequency) to extremely sophisticated (Metropolis-Hastings Markov Chains, Parallel Tempering Markov Chains, Hybrid Evolutionary Algorithms and Genetic Algorithms). The efficiency and accuracy of these algorithms have been demonstrated in a number of international blind data challenges for the LISA detector (MLDC). In the future, it is expected that these blind challenges will recommence for the eLISA mission, where the ADAMIS group will continue to be involved at the highest level. ADAMIS has collaborations with AEI Germany, Cambridge University UK, IEEC Spain and Montana State University US on the development of search algorithms.

• Fig. 5. This figure shows convergence of a Hybrid Evolutionary Algorithm to the sky position solution for a pair of supermassive black hole binaries. The squares represent the

true and antipodal sky positions, while the stars represent the different organisms (solution sets). We see that not only does the algorithm quickly converge to the primary and secondary solutions, but also provides us with the positions of the next nearest solutions in this multi-modal problem.

The ADAMIS group is involved in the modeling and the effects of spin, eccentricity etc on GW sources. The ADAMIS group has current collaborations with AEI-Potsdam, Cambridge University, MIT US, Montana State University and University of Zurich, Switzerland on this subject. Collaboration also exists with IAP on the inclusion of physical effects such as spin, eccentricity and amplitude corrections into the waveforms. The ADAMIS group is also involved in the modeling of EMRIs. These are complex objects that spend most of their eLISA lifetime in the strong field regime close to the SMBH and are important to fundamental physics. The ADAMIS group is also involved in the field of the use of hybrid waveforms for SMBHs, which merges analytical waveforms (inspiral) and Numerical Relativity (the actual merger and ringdown). Finally, ADAMIS is involved in astrophysical modeling of GWs, in particular the role of residual gasous disks in the GW emission from a binary black hole system (collaboration with the University of Milan). Since 2005, the LISAFrance collaboration has developed the LISACode simulator mainly at APC with the help of ARTEMIS-Nice, IAP-Paris and LUTH-Meudon). LISACode is a simulator that models a space-based interferometer for GW

• bservation. It has been developed in the context of the LISA mission and is used now for eLISA. Its ambition is to

achieve a new degree of sophistication allowing to map, as closely as possible, the impact of the different subsystems

• n the measurements. It is also a useful tool for generating realistic data including several kinds of sources and

preparing for their analysis. It generated data for the MLDC and was a key tool in the design of the eLISA where it was used for testing different possible configurations. LISACode is still evolving with the participation of European eLISA collaboration, and in particular, with the Albert-Einstein-Institute (Hannover and Potsdam, Germany) with aiming at a full realistic scientific simulator of the mission. A key mission in the field of space based GW observatories is the mission LISAPathfinder which is a dedicated technology demonstrator. An APC team, including members of ADAMIS, is involved in the data analysis of LISAPathfinder with the complementary data center hosted at the FACe.

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

5

More recently we started to apply a method based on genetic algorithms to detect low frequency GW (nHz) with the system Pulsar Timing Array (precise timing of several stable pulsars observed by radiotelescopes) in collaboration with AEI-Potsdam/Hannover-Germany and LCP2E-Orleans. Simulations The simulation thematic of the ADAMIS group is an entity composed of two permanent CNRS researcher (P. Varniere at 100% and E. Porter at 10% since 2011) and one “enseignant-chercheur” (F. Casse). M. Tagger left at the end of 2007, to become the director of LPC2E. They closely collaborate on the software development with F. Dodu from the algorithmic group of the Laboratory. The activity of the team is articulated around the study of instabilities within astrophysical plasmas through the use

• f high performance global numerical simulations. The active members of this thematic are part of the ADAMIS group,

recognizing simulation as a transverse activity throughout the laboratory. Indeed, the high-performance simulations carried out within the ADAMIS group aimed to interact with the different groups and experiments developed at APC. In particular, we all have a second affiliation to a different thematic group and are linked with several missions such as INTEGRAL (Varniere), LOFT (Casse, Varniere), SVOM (Casse), Virgo and LISA (Porter). Simulations are at the heart of the early development of such missions, being used to establish the science case for the yellow book and later on for the calibration. This numerical description of the sources is based on constraints inferred from archival observations. It is important to note that this interaction functions inversely as well, and the future data will be used to more tightly constrain the physical processes studied. The members of the simulation thematic also pursue algorithmic development in order to construct innovative computational methods and federate around a single advanced numerical tool that could be applied to a wide variety of problems (AMRVAC).

Fig 6. Left panel shows 3D structure of a vortex in a proto-planetary disk (Méheut, et al 2012, [hal-00705321]). Right panel shows Image of a 2D optically thick disk emitting blackbody radiation with an inclination of 85◦ and in the presence of the RWI. The first order image (the distorted complete ring) clearly shows the spiral shape of the emitting

• region. The beaming effect makes the emission brighter on the approaching side of the disk (here, on the left of the image). The second order image is the portion of ring at the

bottom of the image. It is due to photon swirling around the black hole before reaching the observer. The third order image is the thin ring of illuminated pixels at the center of the image, due to photons orbiting around the black hole very close to the event horizon before reaching the observer. This ring is the so-called black hole silhouette. It is truncated here due to optical thickness.

The scientific production of the simulation thematic since 2007 is of 16 publications in refereed journals dispatched in our three main research axes. Indeed, our activity alternates according to the degree of advancement between archival data analysis (4 publications), theoretical modelling of phenomena (6) and the numerical simulation

• f this processes (6). In terms of astrophysical sources, our activity can be regrouped around three families of sources

(with a small selection of key results presented for each of them) Study of compact object environment: We integrated our results from analytical work, numerical simulation and

• bservational data to form a coherent model for the behaviour of black hole binaries. This led us to identify four basic

states based on the theorized physical processes associated with quasi-periodic oscillations (QPO) observed (Varnière, et al 2011, [in2p3-00691891]). This result is the basis of our most recent collaboration with R. Remillard (MIT) to prove, starting from observations this time, the existence of these four basic states. Description acceleration processes in astrophysical shocks: We developed a multi-scale numerical description of the transport and acceleration of electrons and cosmic rays in the vicinity of astrophysical shocks occurring in young supernova remnants. We were able to describe the key role of the relaxation of the magnetic turbulence in the creation of X-ray rings (Marcowith & Casse 2010, [hal-00447523]) while identifying its mechanisms. Planet core formation in protoplanetary disks: First study of the formation of 3D vortices in protoplanetary disks through the Rossby Wave Instability (RWI) and unveiling of their complex, fully 3D, structure which plays a central role in dust aggregation which in turn can lead to planet core formation (Méheut, et al 2010, [hal-00684062] ; Méheut, et al 2012, [hal-00705321]). This domain of research, started in the simulation group with the arrival of P. Varniere and the PhD thesis of H. Meheut, is emergent and one of the three unifying theme of the LabeX “UnivEarths” as it also touches on the activity of AIM and the “Institut de Physique du Globe”. In the context of this rapidly expanding discipline we are organizing in the fall a workshop on protoplanetary disks, subsidized by the “campus spatial”, to regroup and federate the French specialists of the domain. The work accomplished in the simulation thematic is often done with external collaborators with whom we share techniques, expertise or a more transverse collaboration such as those with observers who need our modeling of their

• data. In that respect our small group has a wealth of national and international collaborations across several
• disciplines. Our main collaborators include: Rodriguez (AIM), Ménard, Ferreira (IPAG), Meliani, Sauty (LUTh), Marco-

with (LUPM), Tagger (LPC2E), Keppens (Leuven, Belgium), Cadolle-Bel (ESAC, Netherlands), Benz (Bern, Switzerland), Remillard (MIT, US), Blackman (Rochester, US), Eikenberry (Florida, US), Melia (Arizona, US), Hartigan (Rice, US).

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

6

Academic recognition and standing in the field. ADAMIS researchers participate in a number of international high- profile projects as listed in Annex 2. In those they often play important and leading roles. The members of the group are leaders in the component separation work for the Planck mission as well as other CMB experiments. They coordinate the development of the Planck Sky Model. They have contributed to preparations of the future CMB mission concepts, B-Pol and COrE, where they led the work on the component separation and end-to-end data analysis pipeline issues specific to those missions. ADAMIS researchers have coordinated the work on re-scoping of the LISA mission and are involved in building the science case for the Einstein Telescope and the LOFT satellite mission. The group has gained a unique involvement in some US-led small-scale CMB experiments. The expertise of the group is broadly recognized as emphasized by numerous invited presentations given by the group researches as well as their participations in supervisory or experts’ committees and evaluation or advisory

• panels. ADAMIS researchers have also participated in organization of international workshops and conferences.

The group has instigated a number of interdisciplinary projects and is active, and successful, in soliciting financial support from interdisciplinary programs including Conception et Simulation program of ANR, ‘campus spatial’ of Universite Paris Diderot, interdisciplinary University projects, and many others (Annex 2). In particular, the group is principal coordinator of the ANR project MIDAS’09. ADAMIS members have been successful in obtaining and maintaining an access to supercomputing resources (CCRT, IDRIS, CINES) in France and in the US (DOE NERSC). Interactions with social, economic, and cultural environnement. ADAMIS has been actively involved in outreach

• activities. Members of the group have written popular books, popular book chapters, have organized or contributed to

festivals of astronomy and popular science manifestation, giving popular talks or lectures, e.g., directed at the high school teachers, and performing demonstrations. Involvement in professional training. The ADAMIS group is very active in providing training by research. Its members currently supervise 5 PhD students in addition to 5, who obtained their degree in the past 5 years. Two of the past students were international co-supervisions. All the ADAMIS PhD students have completed their studies within a nominal period of 3 years and all have successfully continued their career in the research. The group actively provides training-by-research also on the master level and has trained 11 France-based and 4 foreign (from the US and Italy) students. The group members contribute to undergraduate, master, and doctoral level courses by either giving or coordinating the university courses or summer school lectures (Annex 4).

• b. Analysis of the team resources

Human resources. The group currently consists of 7 permanent researchers: 5 CNRS (Sections 1 (4), 17 (1)); 2 UPD. The background of one of them (Chassande-Mottin) is mainly in signal processing, while others (Casse, Delabrouille, Petiteau, Porter, Stompor, Varniere) though predominantly (astro)physicists have strong and demonstrated history of interest in, and contributions to, interdisciplinary aspects of astrophysics and/or cosmology. 3 of the current members have joined the team within the last 5 years either via a direct recruitment (Petiteau, Porter) or transfers from the other laboratories (Varniere). One permanent member, Tagger, has left in that period to take up a directorship of the LPCE Laboratory in Orleans. The group also includes 2 permanent software engineers (Le Jeune, Dodu) collaborating closely on algorithmic and numerical aspects (Planck, ANR-MIDAS’09, simulations). The team has a long-term associate, Jean-Francois Cardoso (Telecom) (50%, since 2005), an expert in the signal

• processing. Gilles Faÿ, a statistician from Universite-Lille-1, spent 3 years with the group and is now a full professor at

Ecole Centrale. The group hosted a mathematician, Dominque Picard (P7), for a yearlong sabbatical leave in 2009. ADAMIS has had 4 post-doctoral researchers: J. Grain (2006-2008, FP6 grant), O.Rabaste (2007-2009, VSF grant), M. Remazeilles (2009-2011, CNES grant), S. Basak (2009-2011, P2I grant) and 1 ATER – F. Vincent (2011-2012), all of whom successfully continue their careers in research either as permanent researchers (Grain –CNRS, Rabaste – ONERA)

• r postdocs (Remazeilles – IAS, Basak – CEA, Vincent – CAMK, Poland).

The group has been also joined by 3 fixed term software engineers supported by the ANR MIDAS’09 project: A. Rogier (2010-2011), P. Cargemel (2011-2012), and F. Dauvergne (2011-2012).

• Funding. ADAMIS has been supported by a number of grants obtained directly by its members and also by funds of the

projects to which ADAMIS has been contributing and which have been allocated to the projects directly. The latter include: Planck, Virgo, and LISA project. The development of novel techniques and methods for CMB data analysis have been supported predominantly by ANR- MIDAS’09 grant (150k€, 2009-2012) but also some smaller grants, such as ‘Astro-needlet’ grant obtained from the Campus Spatiale program (9k€, 2012). The Planck work was supported by the Planck APC funds but also by the P2I grant awarded directly to ADAMIS (100k€, 2009-2011). Virgo data analysis was supported by the project money and also a grant awarded directly to E. Chassande-Mottin by the Virgo Collaboration (90k€, 2007-2009). The LISA, and now eLISA/NGO, effort has been suport by the project. The involvement in the small CMB experiments have been supported by the FP6 MCurie grant (80k€, 2006-2008), France-Berkeley Fundation (10k$, 2010) and by the PICS program (7k€/year, from 2012). The multi-messanger activity has received funds from the GDR PCHE. The simulations work has been supported by grants from Campus Spatiale, GDR PCHE, and Egide program (Tournesol). Hébergement. The members of the ADAMIS team have offices either on the 3rd floor of the Condorcet building on the Paris-Diderot University campus or in Francois Arago Center. The team bi-localization, though not ideal, has not shown to be a problem for the team to date. The computer resources are provided by the Francois Arago Center (Adamis cluster), CC-IN2P3, and supercomputing centers in France (CCRT, CINES, IDRIS) and the US (NERSC).

• Resume. ADAMIS is an active and dynamic research group strongly contributing to the laboratory goals as far as both

research and teaching is concerned. The group has matured over the last 5 years and plays a unique and distinguishable role within the APC lab with its emphasis on interdisciplinary aspects and novel techniques in data analysis and simulations, reaching out beyond the standard physicists’ toolbox. The strength of ADAMIS lies

Unités de recherche Vague D : campagne d’évaluation 2012-2013 Février 2012

7

predominantly in the national and international visibility of its members, who play important roles in their respective collaborations - Planck, Virgo, eLISA/NGO, simulations - and in the quality of their research. The group is instrumental in providing a lively and stimulating environment within which such research can be conducted. Nevertheless inter- actions between the group members, though frequent and fruitful, remain typically informal and have not resulted yet in common official collaborations or papers on a level one could ideally wish for, given the existing synergies between different science areas represented within the group. We hope that a recruitment of a simulation person, whom we currently actively pursue, would further expand the simulation effort within the group but also integrate it better with the data analysis work, could at least partially amend this situation. We also explore other opportunities to stimulate such inter-group collaborations. On the lab arena, ADAMIS works closely and productively with the Cosmology and Gravitation group. The collaborations with the other groups have been initiated, in particular with the High Energy Astrophysics group, e.g., multi-messenger astrophysics – E. Chassande-Mottin; compact objects simulations - F. Casse, P. Varniere, but the full realization of their potential is still a matter of the future. The other future challenges for ADAMIS involve the need to diversify its cosmological research beyond the CMB area.

• 2. Organisation

The ADAMIS group is involved in three major projects, Planck, Virgo, Lisa, all of which are shared with the Cosmology and Gravitation group of the lab. Consequently, there exists particularly strong working partnership between these two groups with both groups sharing leadership on common tasks and working together to reach common goals as well as to determine the future directions and evolution of the projects. Close interactions exist also outside of the current projects, for instance as part of preparatory work for future initiatives or scientific manifestations. On occasions those involve also other groups as in the case of collaboration between ADAMIS, Cosmology & Gravitation, and Theory groups aiming at building the science case for the future CMB satellite. Collaborations with the other groups are however more occasional, though there are quickly developing interactions between the ADAMIS and High Energy Astrophysics group in two potentially very fruitful directions: multi-messenger astronomy involving GW observations, and compact objects simulations, as well as in some other projects, e.g., exploring the connection between the Galactic magnetic field, cosmic rays, and Galactic foregrounds. This last topic also involves the Theory group. ADAMIS has established very good and fruitful connections with the algorithmic group of ‘service informatique’ of the lab, manifested by very successful common projects leading to a development of successful high performance codes and numerical libraries, such as component separation codes: SMICA and NILC used by Planck, public CMB data analysis software libraries, S2HAT and MIDAPACK, or high performance magneto-hydrodynamic code ARMVAC. ADAMIS has also played a pivotal role in purchasing, and is a main user of, a medium-size computer cluster, now at Arago Center (FACe), which alongside the supercomputing centers is the major computational resource of the group. The group is strongly involved in the lab’s life with two of its members (Varniere, Errard) serving on the APC ‘Conseil du Laboratoire’ and one (Varniere) - on the informatics advisory group of the lab. The group strives to animate interdisciplinary research on the level of laboratory by organizing regular interdisciplinary seminars, occasional interdisciplinary conferences, and by maintaining the presence at the lab of researchers from different walks of science, e.g., by hosting interdisciplinary postdoctoral researchers and co- supervises students in areas of computer science, applied math, and statistics. ADAMIS is coordinated by a group lead (Delabrouille till 2010, Stompor since then) and a deputy group lead (Tagger till 2008, Chassande-Mottin since then). It holds regular, bi-weekly group meetings interleaved with the group

• seminars. The group meetings facilitate an information flow and provide a venue for collective decision-making.