French Roadmap for Cosmic Microwave Background Science

Home , Norman Jarosik, QUIJOTE CMB Experiment, Tenerife Experiment, The E and B Experiment

French roadmap for Cosmic Microwave Background science

June 2016 The front cover image illustrates our view of the observable Universe in the millimetre range. The colours indicate the intensity variations, while the lines give the direction of polarization. The inner sphere corresponds to emission from our Galaxy, while the outer sphere corresponds to the Cosmic Microwave Background (CMB) signal, the most remote light we can ever measure. The CMB can be interrogated so as to reveal the physical conditions in the primordial Universe, symbolised here by the white outer shell. The central question mark represents the assessment presented in this document on how the polarized sky will be mined for cosmological content, and whether France will be part of such a quest. The image of the back cover illustrates the intensity part of the observable Universe emissions in the millimetre range as mapped by the Planck mission facing simulations of the evolution of the large scale distribution of matter. These images under creative commons license were designed and realised with data from ESA and the Planck collaboration by the HFI French outreach group and the Canopée company.

ii French roadmap for Cosmic Microwave Background science

Executive summary The highly successful Planck cosmic microwave background (CMB) mission has now accurately measured more than a million harmonic modes of the CMB sky with a signal-to-noise greater than one, and has completely fulfilled its principal goal of extracting most of the cosmological information contained in the primary CMB anisotropies in temperature (at least within the ΛCDM class of models). Planck has also exceeded its goals and measured the first 100 000 polarized E-modes, and contributed to the measurement of the tens of fainter B-modes now known. The basic ΛCDM model fits all the data, with parameters known at the per cent level. The deviation from scale invariance expected from inflation has been established beyond doubt, and the knowledge of many key parameters have been improved about a hundred-fold, providing precise constraints on a host of possible extensions to the minimal model (e.g., spatial curvature, neutrino properties, primordial non-Gaussianities or isocurvature modes). Fundamental questions however remain, in particular on early universe physics and cosmology. Nevertheless, the CMB will continue to offer arguably the cleanest experimental window on these, through the millions of additional, but weaker, modes remaining to be measured. Indeed, the precision with which the cosmological model can be determined scales with the inverse of the square root of the number of relevant modes. Among other things, the poorly-known polarization B-modes offer the exciting prospect of a first detection of primordial gravitational waves, which would be a key experimental manifestation of quantum gravity. Such a detection is likely to remain completely out of reach of direct gravitational detectors for the foreseeable future. On the late-time cosmology side, another very exciting endeavour is the mapping of the dark matter distribution through its lensing effect on the CMB at higher redshifts than possible by any other methods. In addition to constraints on the evolution of dark energy, or the locus of Galaxy formation, it will tell us about the neutrino sector of particle physics. Additionally, the CMB constraining power on extensions to the basic ΛCDM model will be enormously increased, potentially discovering the failure or limitation of the now standard ΛCDM model; this will also offer a large increase in the leverage of other astrophysical probes such as Euclid and LSST. It should also be noted that the best constraints on distortions of the mean CMB spectrum are still those determined by COBE-FIRAS some twenty years ago, yet it should now be possible to search for these, which must exist at a level well within reach of today’s technology, and to extract the additional information encoded therein, e.g., on the end of the dark ages or the integrated effect from hot and warm matter. With the completion of the Planck project, the French community is now at a crossroad, with demonstrated expertise in all relevant aspects of CMB science and the ability to continue to contribute to this cosmological quest with future experiments; this expertise indeed includes instrument design, assembly, calibration, opera- tions, processing and scientific analysis in all relevant scientific areas. In that context, it is particularly important to emphasize its strong expertise in promising and rapidly developing technologies in cryogeny and detectors, as well as data processing. However, if not preserved through a long-term program of CMB measurements, these key human and technological assets may well disappear and will be very hard to rebuild. In order to propose a roadmap for future CMB measurements, this report reviews the CMB scientific potential and the obstacles, surveys the current landscape and proposes an analysis of current projects. The main conclusions are the following. Space provides a unique environment for CMB measurements by offering freedom from atmospheric fluctuations and the lack of transparency caused by the atmosphere at essential frequencies, access to the full sky resulting in no loss of information particularly on large angular scales, long term measurement stability, and exacting control of systematic errors. Given the necessity of space to provide the benign environment from which measurements will allow the extraction of crucial information that cannot be obtained otherwise, the long term priority for the French community is a strong participation in a CORE-like experiment. Such a CMB polarization mission will be proposed by the European community in October 2016 to ESA’s call for opportunity for the M5 slot, as an evolution of the previous proposals in L3 and M4. M5 is a priori targeted for a 2030

iii launch, although it might be launched as early as 2026, if the instruments can be ready in time, perhaps within the context of a collaboration with a non-European agency like NASA and/or JAXA. There are two, potentially earlier, smaller scale CMB projects in space which are known to us, LiteBIRD at JAXA, and PIXIE at NASA. LiteBIRD is less powerful and of narrower scientific scope than CORE, but is still technically quite challenging, especially given the cost cap. It should only be considered for substantial participation in the event that CORE is not selected for a phase A study. PIXIE is proposing an interesting experimental alternative with exciting and unique scientific path-finding capabilities regarding CMB spectral distortions. Since it would involve only a relatively small fraction of the French community, participation was considered as highly commendable as a proposed mission of opportunity. Any space project will give results in more than 10 years from now, even including the LiteBIRD and PIXIE projects which are not firmly selected yet. All new data and progress in this field in the next decade will therefore come exclusively from the ground and balloons. This exclusive phase will then be followed by a synergy phase between space and ground where the data sets will be complementary in order to cover (optimistically) the full sky at the required frequencies at high angular resolution. To preserve and leverage its expertise, as well as to benefit fully from lessons to be learned in the coming decade, the French community must therefore strongly participate in the suborbital effort. Following the classification of our US colleagues, the CMB ground-based experimental path may be described in a series of stages, with stage-three currently beginning to deploy on the ground of the order of 10 000 detectors to map portions of the sky. This will be followed by a stage-four targeting of the order of 500 000 detectors to observe from the ground an increasing fraction of the sky. Perhaps optimistically, and depending on funding, those experiments which are part of stage-four may start to operate after around 2020. These plans set the stage for any French effort on the ground in the coming years, which would best be framed within a joint European effort. Of course European ambitions on the ground will take into account the fate of the M5 proposal to scope it, but, in any case, direct participation is vitally needed and activities in that direction need to be vigorously pursued. Shorter term projects have been analysed within this global M5/S4 framework. In particular, we note the scientific promise of the proposed B-SIDE balloon-borne experiment that is com- petitively positioned to reveal crucial information on the detrimental effects of dust polarization and its complexities on CMB B-mode experiments, provided it can be flown before 2020. In addition to this niche, B-SIDE also offers a proving ground in real conditions for KIDS detectors, a most promising technology, expertise in which constitutes a significant French asset both for the long term highest priority projects M5 and S4. The ground project QUBIC proposes an innovative way to control low-level polarization systematics. The project has to demonstrate rapidly the validity of its instrumental concept on the sky in order to be a possible stepping stone towards participation in the S4 effort. To achieve this, the current schedule has to be strictly adhered to in order to demonstrate nominal sensitivity and systematics control on the sky by the end of 2018. The QUBIC collaboration and funding bodies should therefore make clear decisions urgently. In any case, the suborbital effort in France and Europe has now to change gears. Indeed, while the CMB future is bright, it may very well turn out to be rather bleak in France barring fast actions from the community and its funding agencies.

iv Contents

1 Introduction 1 1.1 Current status...... 3 1.2 The French community...... 4

2 Scientiﬁc potential of CMB measurements5 2.1 The early universe...... 5 2.2 The spectral distortions...... 9 2.3 Constraining the matter content of the universe...... 12 2.4 Summary...... 16

3 Foregrounds obstacle 17 3.1 Emission components...... 17 3.2 Component separation...... 21 3.3 The foregrounds challenge...... 23 3.4 Summary...... 24

4 Science beyond the primary CMB science 24 4.1 Galaxy Cluster and Large-Scale Structure Science...... 24 4.2 Cosmic Infrared Background...... 27 4.3 Interstellar Medium...... 28 4.4 Summary...... 29

5 Instrumental aspects 29 5.1 Diﬀerent classes of instruments...... 30 5.2 Focal Plane Unit...... 32 5.3 Systematic eﬀects...... 34 5.4 Spectrometers...... 36 5.5 Summary...... 37

6 Data processing and analysis aspects 37 6.1 Types of CMB challenges...... 37 6.2 Data analysis. From time streams to model constraints...... 39 6.3 Overcoming the challenges...... 42 6.4 Summary...... 43

7 Landscape today 43 7.1 Current Sub-Orbital CMB Experiments...... 43 7.2 CMB Stage 4...... 48 7.3 Multiple Telescopes versus Large Focal Planes...... 49 7.4 Location...... 50

8 The Future in CMB science 51 8.1 Long-term projects...... 52 8.2 Mid-term projects...... 56

9 Outreach 58

10 Conclusions 59

A French CMB PhD Theses since Planck inception 67

B Elements on pre-S4 funding in the US 71

v C Feuille de route du groupe “Futur des mesures du CMB” 77

vi 1

Preamble The French governing bodies for fundamental and high energy physics and astrophysics (CNES, INSU, IN2P3 and CEA) called for a joint roadmap on the future of CMB measurements from the ground, balloons, and in space. Speciﬁcally, they requested that the “Programme National Cosmologie et Galaxies” (PNCG) assemble an ad hoc committee to propose such a roadmap. The committee was set up in September 2015 and its membership was initially the following: François R. Bouchet (Chair), Francis Bernardeau, Anthony Banday, François-Xavier Désert, Marian Douspis, Kenneth Ganga, Guilaine Lagache, Louis Rodriguez, Mathieu Tristram, with ex oﬃcio members Monique Arnaud (link to PNCG), Pierre Binetruy (link to CNES-Fundamental Physics group), Cecile Renault (link to CNES- Astronomy group). In the course of the work, Nicolas Ponthieu and François Boulanger were added. Its mandate is given in annexC. The committee organised four town hall meetings, all followed by closed sessions of the committee, on 2015-11-26 & 27, 2016-02-4 & 5, 2016-04-5 & 6 and 2016-06-6 & 7. Presentations made at these meetings can be found on the committee wiki at http://prospective.planck.fr/index.php?n=Main.Meetings. The committee also held two additional internal meetings. A preliminary version of this document was sent to the community for reaction early June. The committee thanks the many members of the community who gave feed-back and substantial written contributions which have helped improve this document.

Before proceeding, let us note that the experimental situation in the “post-Planck” era is not yet stable, with most projects not fully determined, and even less decided/funded, which has limited the possibility to consider all possible scenarios. Still, as we shall see, ﬁrm recommendations nevertheless emerge.

1. Introduction CMB science has historically been undertaken with a complementary combination of ground-based, airborne, balloon-borne and space experiments. Penzias & Wilson(1965), of course, ﬁrst detected the CMB using a ground-based instrument. The CMB dipole was seen from the ground (Conklin 1969), balloons (Henry 1971), and aircraft (Smoot et al. 1977), all before it was ever seen from a satellite (Strukov & Skulachev 1984). While COBE received the Nobel Prize for its exquisite measurement of the CMB spectrum (displayed in Fig.1, Mather et al. 1990) and for the detection of CMB anisotropies (Smoot et al. 1992), Gush et al.(1990)

Figure 1: The CMB frequency spectrum as determined by COBE-FIRAS. The ﬁt corresponds to a black body spectrum at T = 2.725 K with an energy density contribution Ω h2 = 2.471 10 5. γ × −

French roadmap for CMB science 30/06/2016 2 1 INTRODUCTION

Angular scale

90◦ 1◦ 0.2◦ 0.1◦ 0.04◦

103 CMB- TT Planck ACT SPT 102 ACTPol SPTpol POLARBEAR BICEP2/Keck/Planck BICEP2/Keck 101 2 K] µ [ `

D 100 CMB- EE

1 10−

2 10− CMB- BB

3 10− 2 180 500 1500 3000 5000 Multipole `

2 Planck (2015) SPT ]

7 Planck (2013) ACT 1.5 10 × [ π

2 1 / φφ L C 2 0.5 + 1)] L

( 0 L [

0.5 − 1 10 100 500 1000 2000 L

Figure 2: Top CMB angular power spectra determinations as of mid-2015 (Modiﬁed from Planck Collaboration et al. (2015g) thanks to E. Calabrese). This corresponds to the determination (with S/N > 1) of 1 114 000 modes measured with TT, 96 000 with EE (60 000 with TE, not shown), and tens of modes in BB (and weak constraints on TB and EB). Bottom Lensing potential power spectrum measurement from Planck (Planck Collaboration et al. 2015f), as well as earlier measurements. The goal for the future is now to measure the million polarisation modes which are still unknown. .

French roadmap for CMB science 30/06/2016 1.1 Current status 3 published a highly significant spectral measurement at almost the same time based on the COBRA rocket-borne interferometer, and the so-called “first acoustic peak” was first observed with ground-based (Miller et al. 2002) and balloons (Mauskopf et al. 2000; de Bernardis et al. 2000a; Hanany et al. 2000) experiments. Finally, while the WMAP satellite was the first to publish a significant temperature-polarization cross-correlation (Kogut et al. 2003), a ground-based instrument was the first to make a “clean” detection of CMB polarization without the aid of temperature cross-correlations (Kovac et al. 2002). Space does provide the ideal platform for many CMB measurements, but sub-orbital observations have advantages of their own; while more limited in scope, they can be performed more quickly, for less money, and with greater angular resolution than can be achieved from space. Sub-orbital observations have played a necessary role in the development of the CMB field, and will continue to do so. Here, we give a summary of the state of experiments today.

1.1. Current status

The two-point angular power spectra of the CMB contain all of the information available if the CMB is statistically isotropic and distributed as a multivariate Gaussian, which we now know is an excellent approximation. The CMB power spectra are in turn uniquely determined by the underlying cosmological model and its parameters. We need to measure four independent power spectra to characterize the temperature (T) anisotropies and their linear polarisation. Because linear polarization is given by both an amplitude and direction, it can in turn be decomposed into two coordinate-independent quantities with different dependence on the cosmology (e.g, Kamionkowski et al. 1997; Zaldarriaga & Seljak 1997). One, the so-called E-mode which is the curl-free part, is determined by much the same physics as the intensity, and therefore allows an independent measurement of the background cosmology, as well as an improved determination of some parameters (e.g., the reionization optical depth). The other polarization observable, the B-mode, is only sourced at early times by tensor modes (gravitational radiation), as produced for example during an inflationary epoch. The E and B components are also conventionally taken to be isotropic Gaussian random fields, with only E expected to be correlated with intensity. Thus we expect to be able to measure four independent power spectra, namely the three auto-spectra TT EE BB TE C` , C` , and C` , along with the cross-spectrum C` . Figure2a shows the state of knowledge of the two-point angular auto-power spectra as of late 2015 (Planck Collaboration et al. 2015g). This corresponds to the determination, with S/N>1, of about 1.1 million of temperature modes, of order 100 000 E-modes, and tens of B-modes (and lensing modes, in addition to more than 60 000 TE modes and weak constraints on TB and EB, which are not shown). The model shown with dashes corresponds to the best fit determined with the temperature data alone. While these temperature measurements are driven by Planck, the polarisation and ground determinations offer comforting verification. One way to look at the difference in the numbers of temperature and polarization modes measured is that the millions of unmeasured polarized modes are exactly what ongoing and future experiments will endeavour to reveal and to exploit. Naming just a few of what these (mostly temperature) measurements by Planck allowed, the knowledge of the relative amplitude of the primordial fluctuations at large and small scales was improved a hundred-fold in about 20 years and reached at last the accuracy needed for testing inflation (n = 0.965 0.006), and another S ± hundred fold improvement was achieved in less than 15 years on possible primordial non-Gaussianity or global curvature ( f Loc < 10 and Ω = 0.000 0.005, both at 95% confidence). NL K ± In 2015 and 2016, the BICEP2/Keck collaboration published polarisation data which, using Planck data to account for foregrounds, established the best limit on the primordial tensor-to-scalar ratio obtained to-date, r < 0.07 (95% confidence limit). This improved the previous best upper limit set by Planck using temperature alone of r < 0.10 (95% confidence limit). Similarly, the Planck collaboration has recently published an improved analysis of the EE spectrum at low multipoles from the HFIinstrument which tightens the determination of the optical depth to reionization to τ = 0.055 0.009. This value is lower than previous CMB estimates and relieves ± the tension between CMB data and models of reionization based on the formation of stars and galaxies (Planck Collaboration et al. 2016a,b). Planck has also made powerful measurements of higher order correlations of the anisotropy fields, in particular the four-point function arising from lensing from which is deduced the lensing potential spectrum of Fig.2b (and which constitutes a 40 σ detection).

French roadmap for CMB science 30/06/2016 4 1 INTRODUCTION

In addition to spatial anisotropy measurements, the CMB frequency spectrum is another key observable. Departures of the CMB black body spectrum, i.e., spectral distortions, encode information about the full thermal history of the Universe from its early stages until today. The µ-type distortions probe the era before the last scattering surface and are sensitive to any process that injects energy into the electromagnetic plasma between redshifts 5 104 and 2 106. The y-type distortions, in turn, probe the stages after the recombination, × × when CMB photons interact with ionized and hot gas. Among the processes that induce spectral distortions are: the reionization of the Universe and structure formation; decay or annihilation of particles; dissipation of primordial density fluctuations (Silk damping); primordial black holes; small-scale magnetic fields; cosmological recombination, etc. Many of these processes are part of our standard cosmological model and are expected to leave observable spectral distortions. It has been known that the average CMB spectrum is extremely close to that of a perfect black body since the COBE-FIRAS (Fig.1) and COBRA measurements in the early 1990’s. Possible deviations from the black 5 4 body spectrum must be of order or less than about 10− to 10− . The Planck data, despite their limited spectral coverage, provide additional limits on the CMB spectral distortions. Using the Planck-HFI data, Khatri & Sunyaev(2015a) produced a tentative µ-type distortion map dominated by the y-type distortion contamination from the hot gas in the low redshift Universe. They set a limit for the amplitude of µ-type distortions below 6.4 10 6 on scales near 10 arc minutes, 14 times tighter than the COBE-FIRAS (95% confidence) limit on the × − mean of µ. After decoupling, structures grow and evolve to form the first stars, which reionized the Universe between redshifts 10 and 6. Galaxies assembled and formed the observed large-scale structure (LSS) – clusters of galaxies, filaments and voids. CMB photon interactions with the ionised gas in the LSS through inverse Compton scattering generates y-type distortions. Using the Planck y-map (Planck Collaboration et al. 2015h) estimated the contribution of the fluctuating part of the y-distortion monopole from resolved clusters of galaxies to be of order 5 10 8, a factor of 6.8 below than the best upper 95% confidence limit from COBE-FIRAS × − (1.5 10 6). The contribution to the Compton parameter from unresolved sources and ionised gas in larger but × − undetected structures is expected to be 1.6 10 6 (Khatri & Sunyaev 2015b). × −

1.2. The French community In addition to scientiﬁc results, the Planck heritage includes a broad, well-trained, French community. Figure3 presents the eleven laboratories with people involved in post-Planck CMB experiments or projects. The skills

Figure 3: Number, location, involvement of French CMB researchers.

French roadmap for CMB science 30/06/2016 5 of these teams cover most aspects of an entire project: from detectors and cryogenics to likelihood and cosmological parameters. The community comprises experts in all types of scientific analysis (CMB, lensing, dust, SZ, CIB, etc.). Note that the number of people involved in the Planck-HFI data analysis, including PhD students (as listed in AppendixA) and post-doctorates, is almost four times larger than the number of “ CORE people” who officially endorsed the M4 proposal, i.e., the 49 persons with permanent positions. Such an experienced community affords real strength to continue mining the CMB data for cosmological information, including its most demanding aspect - the robust detection and characterisation of primordial B-modes.

Without a common long term program centred on major projects, however, the community will disperse and it may turn out to be very diﬃcult to reconvene all the required expertise under a future common framework. In the short term, we recommend the establishment of a structure recognised by the authorities which could extend the action and work of the report committee, and allow the French community to both discuss CMB relevant results and help deﬁne and analyse new possibilities.

2. Scientiﬁc potential of CMB measurements

2.1. The early universe 2.1.1. State of the art The latest observations of the Cosmic Microwave Background (CMB) temperature anisotropies and polarizations strongly support the idea that the early Universe underwent a period of inflation (Planck Collaboration et al. 2015d). By definition, inflation is a phase of accelerated expansion (Starobinsky 1980; Guth 1981; Linde 1982) which is supposed to take place at very high energy, between the Large Hadron Collider (LHC) scale, 104 GeV, and the Grand Unified Theory (GUT) scales, 1016 GeV. Inflation allows us to understand several ' ' puzzles that plagued the pre-inflationary standard model (before 1981) and that could not be understood otherwise. Without inflation, the standard model of cosmology would remain incomplete and highly unsatisfactory. The most spectacular achievement of inflation (and the one which definitively convinced most researchers that this is the correct scenario) is that, combined with quantum mechanics, it provides a convincing mechanism for the origin of the cosmological fluctuations (Mukhanov & Chibisov 1981, 1982; Starobinsky 1982) (the seeds of the galaxies and of the CMB anisotropies) and predicts that their spectrum should be almost scale invariant (i.e., equal power on all spatial scales) which is fully consistent with the observations. Let us notice in passing that this part of the scenario is particularly remarkable since it combines general relativity and quantum mechanics. In fact, inflation is probably the only case in physics where an effect based on general relativity and quantum mechanics leads to predictions that, given our present day technological capabilities, can be tested experimentally. Given all these spectacular successes and given the fact that, despite many efforts, inflation has no real challenger any more, this scenario has gradually become a crucial part of modern cosmology. In order to produce a phase of inflation, one needs a situation where the matter content of the universe is dominated by a fluid with negative pressure. In standard astrophysics, matter is usually modelled by gases which have a positive pressure. At very high energy, the correct description of matter is no longer fluid mechanics but field theory, the prototypical example being a scalar field. Quite remarkably, if the potential of this scalar field is sufficiently flat so that the field moves slowly, then the corresponding pressure is negative. This is why it is believed that inflation is driven by one (or several) scalar field(s). For obvious reasons, this scalar field was given the name “inflaton”. However, the physical nature of the inflaton and its relation with the standard model of particle physics and its extensions remain elusive. This is in fact not so surprising since, as mentioned above, the inflationary mechanism is supposed to take place at energies larger than the LHC scale, in a regime where particle physics is not known and has not been tested in accelerators. po Another crucial aspect of the inflationary scenario is how it ends and how it is connected to the subsequent hot big bang phase. It is believed that, after the slow-roll period, the field reaches the bottom of its potential, oscillates and decays into radiation. In this way, through this so-called reheating epoch, inflation is smoothly connected to the radiation-dominated epoch (Turner 1983; Traschen & Brandenberger 1990; Shtanov et al.

French roadmap for CMB science 30/06/2016 6 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS

1995; Kofman et al. 1997). The temperature at which the latter starts is called the reheating temperature and represents the first temperature ever acquired by the universe in its history. The picture that seems to emerge from the recent high accuracy Planck measurements is that inflation is realised in its simplest version, namely single-field slow-roll with a minimal kinetic term in its Lagrangian. Additional features, such as the presence of several fields or non-minimal kinetic term, which may appear as (natural) consequences of embedding inflation in high energy physics, do not seem to be relevant. If, indeed, inflation is really realised in its vanilla version, an important challenge is to understand, from the high energy point of view, why these extra ingredients are in fact not present. Important questions such as the physical nature of the inflaton field also remains unanswered although interesting pieces of information have been collected. Indeed, the shape of the potential is now constrained and appears to be of the “plateau shape” (Martin et al. 2014b,a), a typical example of this class of scenarios being the Starobinsky model (SI) “R+R2”(Starobinsky 1980). Interestingly enough, this model leads to a potential the shape of which is exactly that of Higgs inflation (HI) (Bezrukov & Shaposhnikov 2008). As the name indicates, Higgs inflation is an inflationary scenario where the inflaton is the Higgs boson recently discovered at CERN. It was realised that, if a non-minimal coupling between the Higgs and gravity is introduced, which seems very generic given that this term is automatically generated by quantum corrections in curved space-time, then the corresponding potential is exactly the one of SI. As a consequence, the power spectra in the two models are identical. The difference between the two models only manifests itself in the coupling to matter, hence at the level of reheating. Let us however notice that, if one goes beyond the tree level calculation described above, then the two scenarios becomes distinguishable not only by their reheating properties but also by their power spectra (Bezrukov & Shaposhnikov 2009; Barvinsky et al. 2008). Clearly, if the inflaton field were nothing but the Higgs field, this would have far-reaching consequences for physics. It should also be added that other “plateau shape” scenarios have survived the Planck data. In these models, the inflaton is usually a field appearing in the extensions of the standard model of particle physics (usually extensions based on super-symmetry). Other important results have also been obtained, for instance the fact that popular models such as monomial potentials (which were among the first models of inflation) are now disfavoured. Interestingly enough, inflationary reheating is also constrained by the Planck data (Martin & Ringeval 2010; Martin et al. 2014d). The constraints are model dependent and correspond to an average reduction of the prior- to-posterior ratio of about 40%. The improvement of the most recent Planck 2015 data over Planck 2013 roughly corresponds to one bit of information (Martin et al. 2016).

2.1.2. Constraining the simplest models of inflation There are two types of inflationary perturbations: scalar perturbations related to the fluctuations of the inflaton field and tensor perturbations related to the fluctuations of the gravitational field or, in other words, (primordial) gravitational waves. Their relative amplitude is characterized by the scalar-to-tensor ratio, r. In the theory of inflation, both types of perturbations are quantized. The next generation of CMB observations opens new observational windows. Indeed, with exquisite determination of the primordial B-modes of the CMB polarization, one can either determine or put quite relevant upper limit on the amplitude of the tensor modes. Either way, this would have drastic implications for our understanding of the physical properties that prevailed in the early universe. At this moment there is only a known upper bound on r, namely

r . 0.07 (95% confidence limit), (1) as set by the latest Planck and Bicep2data analyses (see Fig.4). We are in a situation where there is no natural range for r, in particular there is no relevant lower bound. In fact, a lower bound can be obtained either by imposing that the energy scale during inflation should be large enough to allow baryogenesis to take place – at least at the TeV scale – or that classical rolling should be the dominant evolution of the field compared to quantum fluctuations, but this leads to such tiny numbers that, in practice, no useful lower bound exists. Still, improving the detection capability of tensorial primordial fluctuations at the level accessible to future CMB experiments (r 10 3) would be quite informative. ∼ − First of all, a detection (through CMB B-mode polarization) would be a confirmation of the only prediction of (vanilla) inflation that has not yet been experimentally verified. This would also represent the first detection

French roadmap for CMB science 30/06/2016 2.1 The early universe 7

Figure 2: ExistingFigure 4: Existing and expectedand expected constraints on nS onand nr.S Theand oranger. The and yellow orange contours and yellowshow the contours 68% and 95% show confi- the 68% and 95% confidencedence regions regions expected expected from from the baseline the baseline configuration configuration of a typical next of generationCOrE+. medium The size possibility CMB space to experiment improve the error bars by delensing(specifically is notCORE included+, as was in proposed this forecast. at ESA for The the M4 fiducial call). The model possibility is the to Starobinsky improve the errorR2 barsmodel by delensing [7]. The is blue and not included in this forecast. The fiducial model is the Higgs inflation model (or equivalently Starobinsky R + R2 model, cyan contourssee text).show The the bluePlanck and cyan2013 contours constraints, show the Planck while2013 the constraints, gray contours while the show grey contours the WMAP show the9-year WMAP constraints. 9-year The symbols showconstraints. predictions The symbols of a few show other predictions well of known a few other inflationary well known models. inflationary The models. violet, The yellow, purple, yellow, and red and redregions show vacuum-dominatedregions show convex vacuum-dominated potentials ( convexV 00 > potentials0), convex (Vφφ potentials> 0), convex vanishing potentials vanishing at their at minimum, their minimum, and and concave concave potentials (V 00 < 0; hilltoppotentials or plateau (Vφφ < 0; inflation),hilltop or plateau respectively. inflation), respectively. Taken from Martin et al.(2014b). parity ‘E mode’ and an odd parity ‘B mode’ [9, 10]. The scalar fluctuations produce only E modes, whereas the tensor fluctuationsof a quantum gravitational produce both wave, E clearly and a B breakthrough modes. Thus for quantum B mode gravity polarization (moreover, o the↵ers amplitude a sensitive of these and highly model-independentprimordial gravitationalprobe of tensor waves cannotfluctuations. be seen by experiments such as LIGO or VIRGO, even by eLISA). In fact, inflation is probably the only case in physics where an effect based on general relativity and quantum me- Detection of the long wavelength, nearly scale-invariant tensor fluctuations is considered as an observa- chanics leads to predictions that, given our present day technological capabilities, can be tested experimentally. tional tell-taleAs a consequence,sign that inflation if any experimental occurred signatures at energies of quantum a trillion gravity times is ever higher obtained, than it is verythe likelyones that achieved this by the Large Hadronwill be Collider through the (LHC) study ofat inflation CERN. and At its such cosmological high energies predictions. we Probing may also B-polarization see hints precisely of quantum exem- gravity. Consequently,plifies the the main idea of science using inflation goal as of aCOrE tool towards+ will a better give understanding us a powerful of the clue theoretical concerning and observational how the Universe began and theaspects precise of quantum character gravity. of In the other fundamental words, our ability laws to of see nature through (i.e.,the inflationary how gravity window and has the turned other the forces in nature areearly unified). universe into a laboratory for ultra-high energy physics at energies entirely inaccessible to conventional experimentation. Inflation is thought to be powered by a single energy component called ‘inflaton’. The precise physical nature of theAnother inflaton crucial is unknown aspect related but to it a is detection often assumed of the B-modes to be is thata scalar this would field, lead just to alike determination the Higgs of field the recently energy scale of inflation which is, as recalled above, still presently unknown. More precisely the energy scale discoveredof by inflation the LHC is [11, 12]. The simplest models of inflation are based on a single scalar field with a potential energy density V (). We can easily generalize to models involving more fields. The potential r 1/4 2 energy drives the scale factor of the UniverseV1/4(φ to) evolve1016 GeV as a(t) exp(, Ht)whereH (8⇡G/3)(2)V (). As a ' 0.01 / ⇡ result, the Universe is quickly driven to a spatially flat, Euclidean geometry, and any memory of the initial state of thewhere observableV(φ) is the Universe potential is of e the↵ectively inflaton fielderased,φ. This since determination a patch of of space the energy that scale undergoes is the primary inflation goal becomes exponentiallyof any stretched CMB missions. and smoothed. Determining the value r would undoubtedly be a major discovery, re-enforcing the Accordinginflationary to inflation, paradigm the and large it would patch set the of stage the for Universe any subsequent that theoretical we live in attempts originated to build from global a models tiny region in space thatof was inflation. stretched We would to a know large how size far by from inflation. the Planck The or string original scale region inflation was proceeded. so tiny that quantum mechanics played an important role. Namely, the energy density stored in the inflaton field varied from place to place according to the laws of quantum mechanics. This scalar quantum fluctuation is the seed for all the structures thatFrench we roadmap see in for the CMB Universe science today [6]. This is a remarkable prediction of inflation, 30/06/2016 which agrees with all the observational data we have collected so far [8]. The only missing piece is the existence of tensor quantum fluctuations, which would appear as long-wavelength gravitational waves propagating through our Universe [7]. We wish to detect this using the B mode polarization of CMB. An important prediction of inflation is that the scalar and tensor fluctuations are nearly, but not exactly, scale-invariant—namely that the variance of fluctuations depends only weakly on the spatial length scale. More specifically, the variance of fluctuations decreases slowly toward smaller length scales [6]. This behavior in the scalar fluctuations has now been convincingly detected by WMAP [13, 14] and Planck [8]. While

7 8 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS

Detecting tensor perturbations would also give us a measurement of the inﬂaton ﬁeld excursion since

∆φ r 1/2 Ne . (3) MPl ' 8

In this generic formula (known as the Lyth bound), MPl is the reduced Planck mass and Ne is the number of e-folds probed in the observational window (in practice, N 7). This implies that the field excursion during e ' inflation can easily be of the order of, or even larger than the Planck mass depending on r. In fact, this leads to a “natural” value of r, namely r 10 3, corresponding to a field excursion of the order of the Planck mass. ' − From an effective field theory point of view this means that the higher order operators that are the “remnants” of quantum gravity at the inflationary scale can become crucial and can affect the shape of the inflationary potential. This inflationary Ultra-Violet (UV) sensitivity can be turned to our advantage and used to probe quantum gravity if one can reach the limit r 10 3. ' − Another consequence of a detection would be a measurement of the first derivative of the inflaton potential. Indeed, the tensor-to-scalar ratio can be written as

!2 Vφ r = 8M2 , (4) Pl V and, hence, a detection of the B-polarization would allow us to infer the first derivative of the inflaton potential, Vφ. This is important because, today, we only have a measurement of the second derivative, Vφφ, and no significant constraint of the higher derivatives. The constraint on Vφφ is derived from the measurement of the scalar spectral index !2 d ln ζ Vφ Vφφ n 1 P 3M2 + 2M2 . (5) S − ≡ d ln k ' − Pl V Pl V

Planck has shown for the first time at the 5σ level that nS , 1 (a crucial prediction of inflation) and has obtained n 0.96. Further improving the precision of the determination of n , and possibly a detection of its variation S ' S (the so-called running index), is of key interest for constraining models of inflation. Next generation can extend the lever arm for nS , particularly in the polarization spectrum (EE-modes). It may indeed be possible to extend the primary E-mode spectrum to multipoles of a few thousands because of the very low level of polarized foregrounds at high ` (see §3). It allows a direct determination of the primary metric fluctuation spectrum of wave-modes of about k = 0.35 h/Mpc for an ` of about 5000 (the maximum values of ` and k are proportional). A measurement of r would also significantly impact model building and model selection outlook since precise observations of nS and r can bring constraints on specific models of inflation. In other words, with a detection of B-polarization, our understanding of the shape of the potential would drastically improve, opening the possibility to learn about the physical nature of the inflaton field. Of particular interest, the minimal Higgs inflation (HI) model introduced before predicts r 10 3, see Fig.4, a target already encountered before. As ' − a consequence, checking observationally whether the inflaton field is the Higgs field is within reach of – and therefore an exciting goal for – future CMB experiments. Of course, many other models than HI can also be constrained. This is also illustrated in Fig.4 where the predictions of a small field model, SFI4, have been displayed [The corresponding potential is given by V(φ) = M4[1 (φ/µ)p] where µ and p are two free parameters]. In fact preliminary studies on model selection − indicate that the next experiments should be able to exclude more than 4/5 of the vanilla scenarios (Martin et al. 2014c), as opposed to 1/3 for Planck which gives an idea of the constraining power of those observational projects. It is very important to stress that this conclusion is true if a detection of B-modes is achieved but also in the situation where only an upper bound on r is obtained. Finally, the next generation of experiments will allow us to significantly improve our knowledge of reheating (the phase that concludes inflation). Again, this is illustrated in Fig.4. For a given potential and for fixed values of the free parameters characterizing the shape of the potential, different reheating histories lead to different points in the (nS , r) space. Those points can be inside or outside the experimental contours thus opening the possibility to probe the reheating phase. We have already seen that Planck has obtained model-dependent constraints corresponding to prior-to-posterior reduction of about 40%. Preliminary studies show that an experiment such as CORE could raise this number to 90% (Martin et al. 2014c). Again, this conclusion is true even if only an upper bound on r is obtained. In any case, obtaining relevant constraints on the reheating epoch

French roadmap for CMB science 30/06/2016 2.2 The spectral distortions 9 and on the value of the reheating temperature would really be very important since this will allow us to extend the range of e-folds during which the inflationary scenario is observationally probed. This could even lead to constraints on the coupling between the inflaton field and other fields (fermions, gauge bosons etc. . . ) present in nature (Drewes 2016). We end this section by a few words on the so-called consistency relation. Vanilla inflation predicts that

r = 8n , (6) − T where nT is the tensor spectral index, i.e., the logarithmic slope of the power spectrum of primordial tensorial fluctuations. Verifying this relation would the final proof that inflation, in its simplest realization, occurred.

However, we see that this requires a measurement of nT , a task which is probably already out of reach given the current upper limits on r (due to the magnitude of the irreducible uncertainty due to the ﬁnal number of modes measurable on the sky, usually called “cosmic variance”).

2.1.3. Constraining more complicated models? In the previous section, we have discussed the simplest models of inflation. However, inflation could be more complicated and could involve more degrees of freedom. A detection of B-modes could help to discriminate between “vanilla” inflation and these more complicated scenarios. For instance, in multiple field inflation (a situation where there are many inflatons, not just one), the consistency check (6) is modified and becomes

r = 8n sin2 ∆ < 8n , (7) − T − T where ∆ is a quantity describing the presence and evolutions of (additional) isocurvature modes. Another example is when a Cherns-Simon term is present, as often predicted by string theory. In that case, the primordial gravitational waves become birefringent and the two polarizations propagate differently leading to a different observational imprint (Alexander & Martin 2005). A final instance would be to exclude challengers to inflation. Inflation generically predicts a red spectrum for tensor modes, namely nT < 0. Alternatives, such as string gas cosmology (Brandenberger 2015), may on the contrary predict a blue spectrum nT > 0. A measurement of the tensors could therefore allow us to exclude non-inflationary scenarios (or revolutionize the field). It should also be noted that specific models of the early Universe can, in principle, be explored using a variety of probes. If B-mode CMB observations offer a unique window on the physics of the early Universe, other probes such as direct gravitational wave detections can be used. For example, a specific class of models called axion inflation models (with a non-standard coupling of the dual electromagnetic tensor and a pseudo scalar inflaton field) are expected to produce gravitational waves in ample quantities (Barnaby et al. 2012; Linde et al. 2013). As a consequence they can be constrained from direct gravitational wave experiments (such as e-LISA). This would also be the case in models which generate topological defects such as cosmic strings.

2.2. The spectral distortions A different observational window is emerging through the possibility of vastly improved measurements of the spectral distortions of the CMB frequency spectrum (away from a pure black body shape). Spectral distortions are naturally induced by second order effects or interactions with hot electron gas along the line of sight and are then of the y-type. This corresponds to a non-thermal distortion of the black body spectrum with a conserved number density of photons. Annihilation or dissipations of particles or field modes are bound to induce other spectral distortions. The nature of such spectral distortions depends on the time of the energy injections. Early energy injection can allow partial thermalisation leading to an thermal distribution with a non-zero chemical potential, µ. If the thermalisation is imperfect, intermediate spectral forms (between y and µ distortions can be generated. So far CMB observations were not sensitive enough to allow precise bounds on a large variety of spectral distortions. The current direct constraints on y and µ are still those derived from the FIRAS experiment on board the COBE satellite1, y 1.5 10 5, µ 9. 10 5.(Fixsen et al. 1996; Fixsen & Mather 2002; Mather | | ≤ − | | ≤ − et al. 1994). 1 See Sect. 2.1.1 1 for the best indirect constraint using Planck constraint for µ.

French roadmap for CMB science 30/06/2016 10 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS 3

be computed as [37] 2.2.1. The spectral distortionwhere types... p indicates the average over a period of oscillation andh ⇣iis the primordial curvature perturbation. The 2h⌫3 n(i)(⌫) d(Q/Spectral⇢ ) distortions candi be↵usion due to damping different length specific appearing mechanisms in the (not above including formula, those occurring in the post- I(i)(⌫)= zk z 2 5 recombinationk periods),instead, is given by c 4 10 dz z zk k X ⇥ (2) 3 (i) 2h⌫ nz (⌫) Decaying particles. The spectral distortion+ it induces depends2 on16 the lifetime of the particles. Shorter lifetime k •(i) 1 1+z R + 15 (1 + R) µz . k (z)= dz . (4) µ ⌘ c2 4 10 5 ⇥ k means most energy is releasedD at earlier times so that the distortion2 is closer to a pure -distortion with a z s z HneT 6(1 + R) Xk ⇥ smaller residual distortion. IncreasingZ the lifetime (lowering zX), the overall amplitude of the residual spectral 4 distortion (in other forms) increases. At redshifts z . 1.5 10 also elastic Compton scat- If we consider the ensemble average of µ, we see that tering is not ecient enough:⇥ there is no kinetic equilib- it is equal to the log-integral of the primordial power Annihilating particles, for which the distortion has a fixed shape (but is neither of y- or of µ-types) and only rium and the distortion is of y-type. The• y-type distor- spectrum multiplied by a window function the overall amplitude changes, depending on the annihilation efficiency, fann.. tions is expected to be dominated by astrophysics at low z 2k2/k2 dC redshifts (created when the CMB photons are scattered Wµ(k)=2.3 e D , (5) Dissipation of small-scale acoustic modes. The shapes of the spectralzµ-y distortions, from y to µ types, possibly in the clusters of galaxies by hot electrons,• the tSZ effect). While this signal is very interestingdepend by itself on the as modes a wavelengths as illustrated on Fig.5. Since the tight-coupling approximation is very accurate probe of the matter distribution in the universe [39–41], at redshifts much before recombination we expect this our goal is studying the contribution dueThe to latter dissipation mechanism provides a lower bound on µ-types distortions as they are unavoidably produced in ΛCDM models, being generatedto be a goodby the approximation damping of very for primordial the µ-distortion metric fluctuations. ampli- It leads to an amplitude of acoustic waves, and so we will marginalize over it in tude. This window function and the analogous one for of order µ = (10 8). our analysis (see Sec. VIII B). O − y-distortions are shown in Fig. 1. Additional spectral distortions are the ones created during recombination [35, 42] and reionization [35, 43, 44]: we are not going to include them in our analysis, 1.5 assuming they can be computed with high enough precision to subtract them when looking for the primordial 3 signal. 1.0 In this work we will not consider these intermediate distortions, and take the transition between the µ and 4 y era to be instantaneous at a redshift zµ-y 5 10 [11]: in the case of an energy release that does⇡ not⇥ vary 0.5 abruptly with redshift, we do not expect the inclusion of i-distortions to alter significantly the constraints on the 4 parameters describing d(Q/⇢ )/dz. We leave the analy- 0.0 sis of their e↵ect on forecasts for cosmological parameters -4 -2 0 2 4 6 for future work. While there are many non-standard potential sources of spectral distortions, e.g. decaying orFigure annihilating 5: This cartoonDark plotFIG. shows 1. theThis two cartoon range of plot scales shows in (very) the scales light which grey and are blueprobed which by are respectively probed by Matter particles [12, 34, 36], a sourceµ of-and heatingy-type that spectral is distortions,µ-and usingy-type approximate spectral distortions, window functions using the in “windowwave-modes. function” These 2 numbers µ and y are thus present also in the standard picture issensitive the dissipation to the primordial of approximation power at smaller of scales Eq. (5 then). those probed by direct CMB anisotropies (shown by the left range perturbations in the primordial plasmain due pink). to Silk Figure damp- extracted from Cabass et al.(2016). ing. Even before recombination, when the tight-coupling This simplified picture allows us to obtain a qualitative approximation holds, photons are random-walking within understanding of the possible constraints coming from an 1 the plasma with a mean free path mfp =(In thisneT context) .In the detectionexperiment of a like post PIXIE recombination [2]. µ-signal would further tighten constraints on n and the fluid description, this amounts to anisotropic stresses s its possible variation. OurWe ability also to account do so foris illustrated adiabatic on cooling Fig.5. [34 It, shows45], namely what is the mode range (in k) that induce dissipation. One can compute the (inte- that sources the y and morethe factimportantly that electronsµ distortions and types. baryons As can alone be seen would from cool this plot the k-range extents grated) fractional energy lost by these acoustic waves : down faster than photons. Because of the continuous in- to about 10,000 h/Mpc scale. As a consequence, it allows tighter constraints on n . These are however very in the tight-coupling approximation Eq. (1)reducesto teractions, they e↵ectively extract energy from the pho-s broad band observations,tons unlike to maintain the left hand the band same (from temperature, CMB anisotropies) leading forto an which observations of each 1.4 zdC µ 2(z,x) mode can be made. Suchadditional a detection source would of therefore distortions provide of only the aCMB global spectrum. integrated (monopole) value of the p ⇡ 4 h i zµ-y mode amplitudes. As suchDuring it has the littleµ-era, discriminatory this energy power extraction on the basic results cosmological in a neg- parameters in a standard z (3) dk1dk2 2 inflationary2 2 dC setting. It sets however the bottom value of such spectral distortions.9 ik+ x (k1 +k2 )/kD ative µ-distortion of order µBEC 2.7 10 (for the 2.3 3 e · ⇣k1 ⇣k2 e , ⇡ ⇥ ⇡ (2⇡) zµ-y Planck 2015 best-fit values of cosmological parameters). Z 2.2.2. Non-generic models of inflation While limited, spectral observationsIII. EXPECTATIONS nevertheless open FROM a new LARGE window which SCALES can be used to explore specific 3 This has recently shown to be possible for a standard recombination history [42]. models of inflation, specifically hybrid type inflation with multiple field where it is possible to have a large 4 Since i-distortions are not degenerate with µburst-and ofy-type metric distor- perturbationsAs we at discussed these small in thescales. previous Such models section, make the expectedµ and y more pri- easily detectable, but for tions (see Sec. VIII B), they can be useful forfine probing tuned the models redshift only.mordial This is the spectral case in distortion particularµ foris models a function of multi-field of cosmologi- inflation with sudden turns in dependence of di↵erent energy release histories [37]. cal parameters that play a role during the early universe

French roadmap for CMB science 30/06/2016 23 24 24 0.4 0.4 1."10!25 0.3 0.3 23 " " t t x x ! 0.2 ! f 0.2 " f 1 ! 5. 10!26 Hz " 0.1!25 0.1 1 1."10 ! Sr

2 0.0 ! "

0.0 m 0 10 20 1 5 15 ! 0 10 !26 20 W 5 5. 1015 Hz " ! 0 x 1 t ! I $ x Sr

t 2 ! FIG. 22: Function f(xt). m W FIG. 22: Function ! f(xt). 0 I !5."10!26 $ 5"10!8 1 26 10 100 1000 !5."105 ! 50 500 5"10!8 !8 !GHz" 2"10 1 Ν 5 10 50 10024 500 1000

0.4 Ν !GHz"

Ζ 8 8 1"10!

FIG. 21: Intermediate (dashed) and µP -type (solid) spectral distortion for the eﬀective model of a softly turning trajectory 2"10! Model analysis 0.3 and the same parametersFIG. as in 21: FIG. Intermediate 20. The (dashed) spectra and forµ-type three (solid) of the spectral parameter distortion sets for are the e superimposedﬀective model of and a softly actually turning do trajectory not "

t 9 9 !9x significantly differ from thatand of the a same single-field parameters5 10 trajectory! 0.2 as in FIG. with 20. The constant spectra forns.Theyleadto three of the parameterµ =5 sets.0 are10 superimposed− and y and=5 actually.4 10 do− not. " f Single-field models : Methodology for model selection - 3 criteria

Ζ !8 9 9 1"10 significantly differ from that of a single-field trajectory with constant ns.Theyleadto×µ =5.0 10− and y =5×.4 10− . In theP case of µiso =1(green),whichcorrespondstoamaximalenhancementofpower,0.1 i-type and µ×-type distortions× are respectively about two timesIn the or case three of timesµiso =1(green),whichcorrespondstoamaximalenhancementofpower, larger,3 Effective but neverthelessmodels of multi-field should inflation not reach : the level of detectabilityi-type and µ-type by PIXIE distortions and are 8 respectively about two9! times9 0.0 or three times larger, but nevertheless should not reach the level of detectability by PIXIE and !9 2"10 8 9 PRISM5" (µ10=1.47 10− andPRISMy =5 (µ =1.4 .4719−10−.) and0 y =5.5Softly4 19−10 turning.) 15 trajectory20 NOT DETECTABLE × × × × 2.2 The spectral distortionsxt 11 !70 Suddenly!65 turning!60 trajectory!55 DETECTABLE!50 !45 FIG. 22: Function f(xt). 25 N 2"10!9 non-detectable very sharp feature of the spectrum. However,k as noticed in Refs [46, 47], a transversed heavy field can non-detectable very sharpbe feature excited by of the the sudden spectrum. turn and However, the resulting as noticedhigh-frequency in Refs field [46, oscillations 47], a transversedsubsequently aff heavyect the field inflationary can 6."10!25 FIG. 23: Power spectrum5"10 of!8 curvature perturbations for suddenly turning trajectories, with Nt µ=-type 55 (blue and red curves) and be excited by the suddendynamics. turn and Two the eff resultingects generating high-frequency an2 imprint on field the power oscillations spectrum2 subsequently of curvature perturbations affect theµ = can inflationary1.3 inx 10 general-7 be !70 Nt =65 50 (yellow and!60 green curves), and55 for µisoα =100(blueandyellow)and50 45 µisoα =10(redandgreen).i-type dynamics. Two effects! generatingdistinguished: an the imprint modification! on the of the power! Hubble spectrum parameter! of (called curvature the4.deformation"10!25 perturbations effect) and can the in mixing general between be "

!8 1 adiabatic and isocurvature2"10 modes (called the conversion effect). Interestingly,! in the case of models with canonical Nk distinguished: the modification of the Hubble parameter (called the deformationHz effect) and the mixing between 1 kinetic terms, the parametric resonances induced by the two effects! accidentally2."10!25 cancel out each other [47], such that Sr Ζ !8 1"10 2 adiabatic and isocurvature modes (calledP the conversion effect). Interestingly, in the case of models with canonical examples are the well-known F- and D-term models [31–36]. Moreover, they do! not necessarily require super-Planckian

14 Chluba and Jeong the main feature generated in the power spectrum is a very clear peakm at the turning scale. In the following, we FIG. 23: Power spectrum of curvature perturbations for suddenly turning trajectories, with Nt = 55 (blue and red curves) and W

! 0 kinetic terms, the parametricfieldlook values2 for (contraryresonances a regime to in! large9 which induced field features models), by theof neither2 the two power a efine-tuningffects spectrum accidentally of arise initial on field scales values cancel relevant [27–30] out for (contrary each CMB other distortions, to small [47], field suchwhile prior that to 5"10 I

Nt = 50 (yellow and green curves), and for µisoα =100(blueandyellow)andµisoα =10(redandgreen).z $ models),the turn, and a the long spectrum phase of should inflation remain can occur unaX ff forected. a very For low this energy purpose, density. we use However, the analytical for the original approximation hybrid derived in the main feature generated in the powerx 6 spectrumx 6 x 5 isx a5 very5 x clear4 x 4 peak4 at the turning scale. In the following, we 5 times PIXIE sensitivity model [26,A 54], the scalar4.8 10 spectral2 10 index5 10 takes2 10 values10 larger5 10 than2 10 unity,10 4800 which is now! excluded25 by Planck, whereas for Ref. [47]ζ by using the in-in formalism, !2."10 lookA for x a-8 regime in which featuresn of2" the10!9 power spectrum arise on scales relevant for CMB distortions, while prior to Reference ζ = 5 10 S -1 the simplest versions of F-term and D-term models, it must be larger than-5 ns ! 0.98 which is now observationally 10 2 10 1 2 ] the turn, the spectrumdisfavored should. In remain hybrid models, unaffected. inflationFor is usually this realized purpose, along we a nearly use(sin flatx thet valleyx analyticalt cos of1x thet) potential. approximation10 It ends when an derived100 in 1000 ζ 10 0 2 5 50 500 !70 !65 !60 !55 !50 !45 A ζ (k)= (k) 1+µisoα − , (33) examples are the well-known F- and D-termauxiliary models field [31–36]. develops Moreover, a Higgs-type they tachyonic do not instability, necessarilyζ where require a phase super-Planckian of tachyonic3 preheating is triggeredΝ GHz [56–58]. / ! " Ref. [47] by using the in-in formalism, P Nk P x field values-8 (contrary to large field models),Eventually, neither the a fieldfine-tuning configuration of initial reaches field one values of the [27–30] global! minima (contrary of the to potential. smallt field It is" a common assumption to 10 FIG. 24: Intermediate (dashed) and µ-type (solid) spectral distortion for the effective model of a sudden-turning trajectory, x considerwhichFIG. a23: nearly Power is valid spectrum instantaneous in of the curvature heavy perturbations waterfall mass for regime phase sudden turning (lastingm trajectories, less10H than with, andNt one= 55 where e-fold). (blue and-6 x red But curves)k/k it hasand.Thesuddenturnischaracterizedbythe been shown that inflation can models), and a long phase of inflationFigure can 6: occur Amplitude for of thea very metric fluctuations low energy induced2 density. by a sudden However,iso turn! of the2 field forand trajectory same the parameter originalafter10 horizont2 values hybridthan crossing.t in FIG. The 20. Nt = 50 (yellow and green curves),0 and for µisoα = 100 (blue and yellow)(sin and µisoxα t= 10 (redxt andcos green).xt) ≡ [ 5 continue during the waterfall10 for0 more than 60 e-folds2 [59–64], what consequently2 2 modifies the observable predictions x purple band corresponds to the window of the µ effects. The models are fine tuned in order for the turn to happen at the angle α performed(k)= in the( fieldk) space1+µ and,α as before,−µiso m γ /H , 9/4, as well as the scale kt that becomes(33) model [26, 54], the scalar spectral index takes values largerζ The than peak unity, ζincreases which3 ~ is iso now excluded with by Planck,iso whereas for He / D 3 ≡ ρ − rightof hybrid time so thatmodels. the scales It is are also within interesting the window. to Figure notice from that S. Clesse. topological defects that/ can be formed at the point of instability, n n P [ eV ] P x -2 super-Horizon during the turn, bound which occurs at the e-foldt time Nt.γ the simplest versions of F-termrun = 0 and D-termmodels, = -0.2 it mustX be larger than n 0.98 which is now observationally

z s 10 and thatrunD-term may models have [31–36]. dramatic Moreover theyconsequences do not! require for super-planckian! cosmology field are values convenientlythat (contrary2 have3 been# to large studied stretched field" so models), far (original outside and our F/D-term observable moles), if the patch waterfall of phase lasts for N 60 e-folds the scalar ∆ρ ≫ 2 i-type / > µ-type Fig. 22 shows the function f(xt) (sin xt xt cos xt)spectral/x ,whichismaximalfor index-7 is given by ns =1 4/Nkp [60–62]xt which2.46 is withtoo lowf for(2 being.46) observationally0.43, and viable. On the other hand neither a fine-tuning of initialX field values [27–30] (contrary to small field models), and a longt phase of inflation can disfavored . In hybrid models, inflation is usually realizedf along a nearly flat≡ valley of− the potential. It10 ends when an− ≃ ≃ which is valid in the heavythenbe generated mass exhibits at regime very a low series energy-1 m density. ofiso damped! However10H oscillations.for, the and original where hybrid The modelx amplitudet [26,if N 54],k/k! the60t entropic scalar.Thesuddenturnischaracterizedbythe of the spectral modes oscillations index induce an enhancement in the powerof to the power spectrum spectrum is of therefore curvature perturbations by several auxiliary field develops a Higgs-type tachyonic instability, where10 a phase of tachyonic preheatingorder is of triggered magnitudes, well [56–58]. above the observed amplitude [62]. Relative Error takes values larger than unity, which is now excluded by Planck, whereas for2 the≡ simplest2 versions of F-term and angle α performedn = 0.2 in thecontrolledD-term field models space it by mustµα be, and, larger whereas than asn theirbefore,0.98 which frequency isµ now observationally is a fixedm disfavored prediction/HFor2. In7those hybrid reasons9 of/ models,4, the we asconsider model. inflation well in this Several as section the an examples scaleintermediatek have casethat where been the becomes waterfall plotted lasts on for 20 " N " 60 e-folds Eventually, the field configurationrun reachesfield onetrajectory of the (as illustrated global minima in Fig.6) fors ! of which the one potential. caniso reach µ Itas is highiso a common as 10− or, for assumption waterfall trajectories to t is usually realized along a nearly flat valley of the potential. It ends when an auxiliaryof inflation. fieldµ develops The model a Higgs-type is based on the two-field potential 2 Fig. 23, focusing on the range a detection6 ≡ with CMB distortionµ − experiments like PIXIE or PRISM may be possible. super-Horizon duringin thehybridThe spectral turn, type inflation, index which can for nevertheless which occursµ can at be be lowered as the high e-fold asto 10 acceptable. time valuesN . for the F-term1 model if a soft-SUSY breaking mass term is added consider a nearly instantaneous waterfall phasetachyonic (lasting instability less forcing than the field one trajectories e-fold). to reach But one− it of has the global been minimat shown of thethat potential-8 inflation where a phase can 2 2 2 2 µ2 10 ψ φ 2φ ψ to theNoticeof tachyonic potential that preheating [55]. for This anglesis triggered-2 neverthelessα [56–58]." π requires/ It2, is a one common some requires2 assumption tuning3 # ofm to theiso consider parameters.! a(100) nearly instantaneousH for inducing waterfall anV ( observableφ, ψ)=Λ 1 level+ of+ CMB distortions., (34) µ 2 2 2 continue during-3 theFig. waterfall 22 shows for more the than function 60phase e-folds (lastingf(x lesst [59–64],) than(sin10 one e-fold). whatxt But consequentlyx itt hascos beenx shownt) that/x modifies inflation,whichismaximalfor can the continueO observable during3 the waterfall predictions forx moret 2.46 with−fM(2.46)µ φc M0.43, and 10 The corresponding distortion spectra are plottedt on FIG. 24. Contrary to the single-field!" VHI# model, we$ find that 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.52.2.3. 1.6 1.7 Energythan 1.8 60 1.9 e-folds release 2 [59–64].≡ from This decaying therefore− or modifies annihilating the observable particles predictions of hybrid models. It is also interesting≃ ≃ of hybrid models.then It is exhibits also interesting a series to notice of damped that topological oscillations.6 defects7 The that8 amplitude can be9 formed of10 at the thewhere11 oscillations pointψ = 12M ofis the instability, position in the of the global power minima spectrum of the potential at isφ = therefore0 and where the parameter µ controls the n theto notice intermediate that topologicali defects10-type that distortions10 can be formed10 at are the10 more point of instabilityimportant10 and10 have than dramatic10 those± consequences of the forµ-type, because the oscillatory features in and that may have dramatic consequencesS forcosmology cosmology are conveniently are stretched conveniently outside our observable stretchedt [ patchsec ] of outside the Universe. ourslope However, observable of the for potential the hybrid close patch models to the of critical point of instability φc. The observable modes with CMB anisotropies leave controlled by µα, whereasthe their spectrum frequency are damped is a and fixed thus prediction theX smallest ofscales thethe contribute model. Hubble horizon less Several prior significantly to instability examples point to and the thus have distortion the scalar been power signal. plotted spectrum Note on those on also scales can be calculated that a variation of the turning scale changesz the ratiowith between the standardi-type 1-field and slow-rollµ-type formalism. distortions. If there is no additional term to the potential, one would get ϵ2 = 0, 5 timesFig. PIXIE 23, sensitivity focusing on the rangeA a detection with CMB distortionX experimentsand because a mild like waterfall PIXIE phase requires orϵ PRISM1 1 at the point may of instability, be possible. the scalar spectral index would be very ζ 6 6 5 5 5 4 4 4 ≪ -8 2 n 4.8x10 2x10 5x10 2x10 10 5x10 2x10close10 to unity.4800 This problem can be avoided if the inflaton φ gets a mass term driving ϵ2 to acceptable values. Such Reference A = 5 x 10 and n = 0 The spectral index can nevertheless-12 be lowered to acceptable values for the F-term model if a soft-SUSY breaking mass term is added Noticeζ thatrun for angles α "to thenπS potential/2, one [55]. This10 requires nevertheless requiresm someiso tuning! of the(100) parameters.H for inducingan additional mass an term observable could arise for instance level due to of logarithmic CMB loop distortions. corrections to the potential 3. run O According to Refs. [60–62], two phases can be distinguished during the waterfall. During the phase-1, the transverse 2 The corresponding distortion spectra are plotted on FIG.C. 24. Mild Contrary waterfall to trajectory the single-field VHI model, we find that 1 The spectral index can nevertheless be lowered to acceptable values for the F-term model if a soft-SUSY breakingfield contribution mass term to the slope is added of the potential in the inflaton direction is subdominant compared to the µ− term. It

] -1 ∼ to the potentialζ 10 [55]. This nevertheless requires some tuning of the parameters. is dominant during the phase-2,whichiseﬀectively single ﬁeld. During phase-1, the power spectrum is enhanced due

A the intermediate i-type distortions are more important than those of the µ-type, because the oscillatory features in -13 to the important contribution of entropic modes. In the following we look for the region of the parameter space leading / 10 -8 Absolutethe spectrumerror are dampedHybrid and thus models the are smallest a particularly scales well-motivated contribute class lesstoof an signiﬁcantly increase inﬂation of power models, by more to becausethan the a factor distortion they ten (which can corresponds be signal. embedded approximatively Note in various also to the level of detectability of

10 the resulting CMB distortions) on CMB distortions scales, compared to the amplitude on CMB anisotropy scales. 3 x high-energy frameworks. They areHe / D most commonly studiedAn inanalytical the context approximation of supersymmetry, of the power spectrum of where curvature the perturbation most prominent for modes exiting the Hubble radius that a variation of the turning scale changes the ratiobound between i-type and µ-type distortions.

[ 5 in the phase-1 has been derived in Ref. [62] by using the δN formalism, x -14 2 10 ΛM µφc Relative error [ GeV ] ζ (k) . (35) X 2 6 2 P ≃ 192π M χ2ψ Y Errors pl k vis

E C. Mild waterfall trajectorywhere ξ and χ have been defied as φ = φc exp(ξ) φc(1 + ξ) and ψ = ψ0 exp(χ). The validity of this approximation ≃ -15 have been checked numerically by unsung the δN formalism and by integrating the exact multi-field background 10 µ µ1 -2 Hybrid models are a particularly well-motivatedµ class of inflation models,3 because they can be embedded in various 10 2 Notice however that for F-term and D-term, the mass of φ at the critical point in the mild waterfall regime cannot reconcile with the µ spectral index. Other models should therefore be envisaged 0.5 0.6 high-energy0.7 0.8 0.9 1 1.1 1.2 frameworks. 1.3 1.4 1.5 1.6 1.7 They 1.8 1.9 are 2 most10-16 commonly3 studied in the context of supersymmetry, where the most prominent n S 6 7 8 9 10 11 12 10 10 10 10 10 10 10 1 t Figure 13. Expected uncertainties of A⇣ (k0 = 45 Mpc ), nS, and nrun using X [ sec ] measurements of µ, µ1, and µ2. We assumed 5 times the sensitivity of PIXIE 8 and A⇣ = 5 10 as reference value (other cases can be estimated by simple Figure 14. Detectability of µ, µ , µ and µ . The upper panel shows the lim- ⇥ Figure 7: Constraints on the yield parameter as a function1 2 of the3 lifetime of decaying particles, if they exist. What is rescaling). For the upper panel we also varied nrun as indicated, while in the its for ✏X = fX/zX, while the lower panel uses the standard yield variable, computed is the required value of YX for which a 1σ-detection of the corresponding variable is possible with PIXIE. lower panel it was fixed to nrun = 0. The purple shaded area is excludedEvisYX (cp., by Kawasaki measurements et al. 2005). of the Forprimordial a given 3particle He/D lifetime,abundance we ratio com- (1 σ level, adapted from ✏ Kawasaki et al. 2005). Frompute Chluba the required & Jeong value(2014 of ).X for which a 1 -detection of the corresponding variable is possible with PIXIE. The violet shaded area is excluded by mea- dicted uncertainties for some representative cases using the MCMC surements of the primordial 3He/D abundance ratio (1-level, adapted from method of Chluba (2013a), finding excellent agreement. Overall, Fig. 42 of Kawasaki et al. 2005). our analysis shows that CMB SD measurement provide an unique 8 5 It is important to note that the range µ = 10− to µ = 10− offers a genuine window of discovery. In particular probe of the small-scale power spectrum, whichfor candecaying be utilized particles. This is illustrated on Fig.7 for the PIXIE nominal mission. to directly constraint inflationary models. Especially, if the small- range. To directly constrain tX, at least a measurement of µ1 is scale power spectrum is close to scale-invariant with small running, needed. At PIXIE sensitivity this means that the lifetime of par- 9 10 very robust constraints can be expected from PIXIE and PRISM, if ticles with 2 10 sec . tX . 6 10 sec for ✏X & 0.1 eV and ⇥ 12 ⇥ 1 8 7 French roadmap for CMB3 science108 sec . t . 10 sec for ✏ & 1 eV will be directly measur- 30/06/2016 A⇣ (k0 = 45 Mpc ) 10 10 . ⇥ X X ' able. Most of this parameter space is completely unconstrained [see upper limit from measurements of the primordial 3He/D abundance 5.3.4 Decaying relic particles ratio9 (from Fig. 42 of Kawasaki et al. 2005) in Fig. 14]. Higher sensitivity will allow cutting deeper into the parameter space and The distortion signals for the three decaying particle scenarios presented in Table 2 will all be detectable with a PIXIE-like experiment. More generally, Fig. 14 shows the 1-detection limits for 9 In the particle physics community the abundance yield, YX = NX/S , µ, µ , µ , and µ , as a function of the particle lifetime. CMB 1 2 3 and deposited particle energy, Evis [GeV], are commonly used. Here, NX ⇢ SDs are sensitive to decaying particles with ✏X = fX/zX as low 4 is the particle number density at t tX and S = 3 kT 7 N 2 7 10 3 3 3 ⌧ ' ' as 10 eV for particle lifetimes 10 sec . tX . 10 sec. For 2.9 10 (1 + z) cm denotes the total entropy density. We thus find ' 3 ⇥ 9 19 PRISM the detection limit will be as low as ✏ 10 eV in this ✏ (E Y ) 10 S/[N (1 + z )] 1.5 10 (E Y )/(1 + z ). X ' X ⌘ vis X H X ' ⇥ vis X X

c 0000 RAS, MNRAS 000, 000–000 12 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS

2.3. Constraining the matter content of the universe 2.3.1. From of the shape of the angular power spectrum

The effective number of relativistic species, usually referred to as Neff, is a very general probe of the physics of the early universe. Indeed, the small scale angular power spectrum of CMB anisotropies is sensitive to the entire radiation content of the early universe. In the standard models of cosmology and particle physics, Neff is simply a way to quantify the energy density of the cosmic neutrino background. It should be equal to 3.046 for the standard ΛCDM model when we have 3 families of neutrinos, but in general Neff can receive contributions from all forms of relativistic species that are decoupled from charged particles (that is all relativistic particles besides photons). As a consequence, a measured value of Neff that exceeds the prediction of the standard model would be an indication that there is some additional species present at recombination or that the thermal history of the universe is non-standard. There is a huge number of possible sources for such contributions, including gravitational waves (see, for example, Boyle & Buonanno 2008), sterile neutrinos Abazajian et al.(2012); Boyarsky et al.(2009); Viel et al.(2005), or more exotic objects such as dark photons Ackerman et al.(2009), 2 massless fields etc. The current precision for σNeff from the Planck 2015 results is about σNeff < 0.2 with no indication of departure from the expected standard value. But this limit can be improved by an order of magnitude with future CMB data.

2.3.2. The lensing potential

Figure 8: The signal to noise ratio for the determination of the lensing potential power spectra. a) (left) Planck only measuredFigure a 3: handful Reconstruction of large scale noise modes of the around lensing` deflection30 with a signal power to spectrum noise ratio from comparablePlanck to2015 one (left) (but still and o asffering ∼ a 40forecastσ statistical for COrE detection+ at overall). the goal The and black required curve specifications displays the expected (right). TheΛCDM deflection signal withpower dashes spectrum displaying is plotted the non- linearbased corrections. on the linear b) (right) matter Future power CMB spectrum experiment (black will solid) dramatically and with extend nonlinear the range corrections of modes (black for which dashed). this power spectrum can be accurately measured. The experimental specification here are typical of what can be achieved from space withby a cosmicmedium variance size mission. of the In primary both panels, CMB the fluctuations red curves because correspond their to theamplitude noise level on small in a temperature scales is essentially only analysis (TT),zero and regardless the blue curve of the to value a minimum of r. Such variance polarization-based combination using reconstructions additional polarisation have been information. demonstrated recently from ground-based experiments [24, 44], and also Planck, but are currently very noisy. This situation will be completely transformed with COrE+, which will reconstruct lensing with S/N > 1 per mode up to Lensingmultipoles efflects600 are (see ubiquitous Fig. 3) over in deep nearly astrophysical the full sky. observations. Significantly, COrE In CMB+ can observations, extract essentially those e allffects the can ⇡ beinformation isolated either in the from lensing high deflection order correlation power spectrum function on of scales the where temperature linear theory anisotropies is reliable. map, or from an E Weak gravitational lensing directly probes the clustering of all matter integrated along the line-of-sight 2 back to the source. This makes lensing potentially a very powerful probe of the matter (clustering) power These provide a particularly interesting target value of δNeff = 0.027 for massless fields in thermal equilibrium (with a detailedspectrum, contribution which whichis free dependsfrom astrophysical on the nature uncertainties -spin- of this such field.) as biasing (i.e., the uncertain relation between the clustering of luminous objects and the underlying mass distribution) that complicate the interpretation of galaxy redshift surveys. Lensing by large-scale structure can be probed using optical imaging surveys through the coherent distortion of the shapes of background galaxies (cosmic shear), and also using CMB lensing. FrenchAs discussed roadmap in for Sec. CMB 2.6, science cosmic shear and CMB lensing are highly complementary. Their combination 30/06/ is2016 particularly powerful (see below for Euclid combined with COrE+). We summarize some of the science enabled by the lensing measurements from COrE+. COrE+ will be a powerful probe of neutrino physics through the combination of its lensing measurement (for neutrino masses), and its precision measurements of the damping tail of the polarization power spectra (for the e↵ective number of relativistic degrees of freedom at last scattering)2. Neutrino oscillation data show neutrinos must be massive, but the oscillation data are insensitive to the absolute neutrino mass scale. For a normal hierarchy of masses (m ,m m ), the mass summed over all 1 2 ⌧ 3 eigenstates is at least 0.06 eV, while for an inverted hierarchy (m m ,m ) the minimal summed mass 3 ⌧ 1 2 is 0.1eV. The individual neutrino masses in these hierarchical limits are well below the detection limit of current and future laboratory -decay experiments but can be probed by cosmology. Massive neutrinos suppress gravitational clustering on scales below the horizon size at the non-relativistic transition, thus reducing the lensing power spectrum. Combining the anisotropy and lensing power spectra of COrE+we

2The accuracy of parameter inferences from the temperature power spectrum measured by Planck [42] are now close to being limited by errors in the modelling of extragalactic foregrounds. Fortunately, further progress can be made with the polarization anisotropies on small angular scales [57], since the degree of polarization of the anisotropies is relatively larger there (around 4% by l = 2000) than the foreground emission.

10 2.3 Constraining the matter content of the universe 13

Figure 9: Contour plot of the K`(k) kernels of Eq.8 which relates the lensing (convergence) power spectrum C` to that of the mass distribution. It shows for each ` the values of k contributing to C`. The red area corresponds to the transition from the linear to the non-linear regime where theoretical predictions are subject to large theoretical uncertainties. Theses are thus quite weak for CMB lensing.

polarisation map or from the B polarisation modes they induce. This set of signatures allows the measurement of projected mass maps, and the power spectrum of line-of-sight mass fluctuations. This is illustrated in Fig.8 which shows the signal to noise ratio for the ground breaking measurement of this power spectrum by Planck (shown in Fig.2b) and for a representative next generation CMB mission (specifically CORE+, as was proposed for the M4 selection). The `-dependent power spectrum, C` of the (projected) lensing potential or equivalent quantities (like the convergence), gives access to a line of sight convolution of the three-dimensional mass density power spectrum, P(k). More precisely, Z dk C` = K`(k), (8) k where K`(k) is a kernel function whose detailed expression depends on the cosmological parameters (such as the energy content of the universe), the redshift of the source plane and the redshift evolution of the matter power spectrum, including its non-linear evolution. There exists different ways of probing such line-of-sight density power spectra besides CMB observation. This is in particular the case for weak lensing observations for which specific kernels can be defined. Fig.9 shows the contour plot of such kernels. One clearly see that they are centred at higher k for higher `. 3 This makes an observable such as C` somewhat sensitive to the theoretical uncertainties in the prediction of the actual shape of the non-linear power spectrum (but much less that lower-z probes of large scale structures). The estimated 1% limit is shown here as a red region: on the right hand side, one must rely on non-linear phenomenological models.

2.3.3. The neutrino mass

The measurement of the lensing projected potential along the line of sight oﬀers a unique possibility for constraining the shape and time evolution of the matter power spectrum. It sheds light on the matter content of the universe in a unique way. It can in particular help in determining the sum of the neutrino mass which we know is non-zero.

3 By using nulling techniques and with the help of multiple source planes, it is actually possible to narrow the k dependence of observed C` (Bernardeau et al. 2014; Huterer & White 2005; Joachimi & Schneider 2008; Kitching & Taylor 2011).

French roadmap for CMB science 30/06/2016 14 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS

If sufficiently massive the neutrinos can account from a fraction of the current energy density of the universe4 which is directly proportional to their total mass. It is roughly given by P 2 mν Ωνh (9) ≈ 93 eV And we know, from neutrino flavor oscillation experiments, that the total mass of the neutrinos should be above 50 meV5. spectrumOur ability to realization constrain the corresponding mass of the neutrinos to comes the same from the model. fact that Asthey illustratedbecome non-relativistic in [22], during the two the formation of the large-scale structure of the Universe. The presence of species that are initially relativistic indeedoptions changes lead the to growth the same rate of forecast structures errors, for modes sok forthat simplicity are larger than we the assume free streaming an observed scale of power the neutrinosspectrumkfs. equalThis change to the of growth theoretical is related power to the spectrum mass fraction offν thein the fiducial neutrinos model. species. Thus during the matter dominated era the growth of structure is proportional to

1 3/5 fν -5 M = 0.21 eVδ a − ,-5 M = 0.21 eV (10) 10 ν cdm ∝ 10 ν Mν = 0 Mν = 0 , z) , z) insteadµ of a for the non-relativistic species. It causes the powerµ spectrum to be damped relatively to a pure , shot noise , shot noise ref ref

CDM(k model-6 for modes that are larger than the free streaming(k scale-6 in a way which is directly related to fν, g 10 g 10 P P 2 2 Pmatter(k; fν) Pmatter(k; fν) 1 for k kfs and 1 8 fν for k kfs. (11)

(z)b(z)] Pmatter(k; fν = 0) ≈ Pmatter(z)b(z)] (k; fν = 0) ≈ − A A 10-7 10-7 This double plateau prediction is true however only at linear order and the transition from one plateau value to H(z)/[D H(z)/[D another is rather broad (it spansz=0.5, µ about=0 2 decades). It is also redshift dependent. Fig. z=2.0,10 illustrates µ=0 the impact of massive10-8 neutrinos on the shape of power spectra, at linear order10-8 (dashed lines) and with non-linear corrections (solid lines). The results 0.01 are shown here 0.1 for redshifts of 0.5 and 2. 0.01 0.1 kref (h/Mpc) kref (h/Mpc)

1.05 1.05 z=0.5, µ=0 z=2.0, µ=0 1.04 1.04

1.03 1.03 observational theoretical observational theoretical 1.02 1.02 error error error error 1.01 1.01

1 1

0.99 0.99 relative error relative error 0.98 0.98

0.97 0.97 [Mν = 0.05 eV] / [massless] [Mν = 0.05 eV] / [massless] 0.96 0.96

0.95 0.95 0.01 0.1 0.01 0.1

kref (h/Mpc) kref (h/Mpc)

Figure 10: The plots show the part of the relative error on the power spectrum coming from observational or theoretical errorsFigure only (cosmic 1: Observable variance is spectrumincluded in the (top) observational and relative error) for error a Euclid on this type mission. spectrum In these (bottom), plots, the for individual the first 1redshiftσ error on bin each (left) data point and has last been redshift rescaled binby the (right) square root of a ofEuclid the number-like of galaxy points, in redshift such a way survey. that the The edges quan- of ∆χ2 = thetity error displayed bands correspond in the to top a shift isthe between galaxy theory power and observation spectrum leadingPg(k toref ,µ,z)1, as when a functiononly the observational of the fiducial or theoreticalwavenumber error isk incorporated, for fixed in redshiftthe likelihood and expression. perpendicularly In order to to gauge the theline interesting of sight range (µ = of 0), uncertainties, rescaled by we the also show for comparisonref in the lower part of the plots the ratio between a massless model and a model with the minimum 2 2 totalinverse mass allowedsquared by bias neutrinob(z) experiments, and bym aν = factor0.05 eVH ((Audrenz)/DA et(z al.) 2013b: it is). therefore a dimensionless quantity. The upper plots show a comparison between a model with massless neutrinos and our fiducial model (M⌫ =3m⌫ =0.21 eV). Solid lines are derived from the non-linear matter power spectrum using theNote updated also on thesehalofit plotsversion the level of of ref. statistical [24], while and theoretical dashed lines errors are one derived can reasonably from the assume. linear This power reflect spec- thetrum. uncertainty The lower that a plotsffect our show ability the to part predict of the the shape relative of a error power coming spectrum from shape observational for a given cosmological or theoretical 4errorsIf massless only the (cosmic energy density variance of a givenis included species decays in the similarly observational to that of theerror). radiation In and these is now plots, negligible. the individual 51-The error lower on bound each for data the total point mass has is obtained been rescaled for a normal by mass the squarehierarchy, root that isof assuming the number the largest of points, mass split in is such betweena way thethat two the more edges massive of the species. error For bands an inverted correspond hierarchy, to when a shift the largest between mass theory difference and is observation between the two lead- lightest, the total2 neutrinos mass should be at least twice as large. Note that the matter transfer function is in principle not onlying sensitive to to= the 1, total when mass only but also the to observational each of the neutrino or theoreticalmass. Such eff errorects are is however incorporated too weak into be the detected. likelihood expression. In these lower plots, we also show for comparison the ratio between a massless model and a model with the minimum total mass allowed by neutrino experiments, M⌫ =0.05 eV. French roadmap for CMB science 30/06/2016 We fit the mock and Euclid-like spectra using the MCMC code MontePython [27]. MontePython uses the Metropolis-Hastings algorithm like CosmoMC [28], but is in- terfaced with class [29, 30] instead of camb [31], is written in python, and has extra functionality; it will soon be released publicly, including the Euclid-like likelihood codes

–4– 2.3 Constraining the matter content of the universe 15 model. Such an uncertainty has been incorporated in an analysis such as in (Audren et al. 2013b) for a Planck+ Euclid type mission. The resulting errors on the total mass is X σ mν = 18 meV. (12)

From Planck to a CORE+ type mission one expects however that the accessible `-range to be larger thus improving on such constraints. It can be potentially as good as (as advocated in Kitching et al.(2015)) X σ mν = 3 meV, (13) but in this case no theoretical uncertainties have been taken into account up to k = 5h/Mpc scale. This cannot be realistic. Assessing the real theoretical precisions with which the total mass of the neutrinos can be determined is still an open question6. It remains that it is probably within the 10 meV range. We further note that one can cut back to (quasi-)linear scales using CORE plus Euclid and do about as well as with Euclid alone using all scales, oﬀering precious information and veriﬁcations on the theoretically uncertain non-linear corrections to the matter power spectrum that Euclid will have to apply in order to reveal its full potential. In any case, Euclid may not be able to determine unambiguously the total mass of the neutrinos, even with the help of the Planck data. This should be possible with the help of an advanced CMB mission.

2.3.4. CMB Lensing as a nuisance

θFWHM = 20 ))

r 12

( θFWHM = 50 σ θFWHM = 100 10

(delensing improvement to 2 α

0 100 101 ∆P (µK-arcmin)

Figure 11: Improvement to the tensor to scalar ratio r from delensing, α = σ0(r)/σ(r). In this case, delensing is obtained using a CMB polarization based lensing estimate, iterating on the delensing and estimation step to increase the delensing eﬃciency. Taken from Smith et al.(2012).

We described in Sect. 2.3.2 how the detection of the large scale structure induced lensing effect on the CMB (and in particular its polarization) allows the determination of the matter distribution in the Universe and how this information can be used for inferring constraints on the cosmological parameters, in particular on the neutrino masses. However, CMB lensing can also be seen as a nuisance. Lensing mixes the E and B polarization modes (Lewis & Challinor 2006). Since the primordial B-modes only arise from the tensorial primordial perturbations which are much smaller than the scalar ones (which create E polarization), this effectively amounts to the redistribution of some of the E-mode polarization power into B. The lensing induced B-modes peaks at around ` 1000 and, with r < 0.11, this signal is always larger than the primordial one after the reionization ∼ bump. It is, after the Galactic diffuse emission, the second largest nuisance for the determination of r (see Figs. 14 and 16 for examples of the effect compared to the level of different astrophysical foregrounds). In order to recover the level of the primordial B-mode polarization, it has been proposed to delens the polarization maps (Sherwin & Schmittfull 2015; Simard et al. 2015; Marian & Bernstein 2007; Seljak & Hirata 2004; Kesden et al. 2002; Knox & Song 2002; Smith et al. 2012). Further it has been advocated that, in order to simplify the analysis of the joint CMB lensing and CMB constraints on the neutrino masses, it will be

6 A precise assessment of the theoretical precision on the mass of the neutrinos is diﬃcult to achieve in particular because in such observables the covariance matrix is non-diagonal.

French roadmap for CMB science 30/06/2016 16 2 SCIENTIFIC POTENTIAL OF CMB MEASUREMENTS

necessary in the future to delens the CMB maps (Kaplinghat et al. 2003). Similarly, constraints on Neﬀ have been shown to improve signiﬁcantly after a delensing of the E-modes (see the CMB S4 Science-Book draft at https://cosmo.uchicago.edu/CMB-S4workshops/index.php/File:Cmbs4_scibook_160305.pdf).

Delensing proposes to remove the lensing induced signal from the measured CMB power spectra and as such requires an estimate of the unlensed CMB and an estimate of the matter distribution. The latter can be obtained either from external surveys (cosmic shear, galaxy catalogues, 21 cm radiation, CIB) or from the CMB lensing estimate itself. The former can be obtained at first order from the observed E-modes, since lensing is only a small correction. Note that when delensing is performed using the CMB lensing estimate, its accuracy can be improved by iterating the delensing step, replacing the CMB maps by the delensed one from a previous step. The delensing accuracy depends on the characteristics of the experiments and on the choice for the matter distribution predictor. A detailed analysis of the delensing efficiency and how this limits r constraints is discussed in detail in Smith et al.(2012). Here, we reproduce Fig. 11 from this article showing the improvement on σ(r) due to delensing (using a CMB polarization-based lensing estimate), as a function of instrumental beam and noise level. As a rule of thumb, it is important to note that unobserved E-modes between 100 ` 1000 will significantly degrade the reconstruction of lensed B-modes and therefore their subtraction. ≤ ≤ In addition, for a given CMB map sensitivity level, the reconstruction of the lensing potential will be optimized by looking at rather high-` in all CMB power spectra. A more detailed analysis, including foreground residuals, and investigating different cases for the matter distribution predictor or tracer (e.g., a CIB or lensing potential map), can be found in Errard et al.(2016).

2.4. Summary

The Cosmic Microwave Background continues to oﬀer the cleanest experimental window on the physics of the early Universe. The next generation of CMB experiments with spectral and polarization capabilities can provide:

a genuine possibility to validate the inﬂationary paradigm and to determine the absolute energy scale of • inﬂation with the detection of the primordial B-modes of the CMB polarization and the measurement of r;

a unique discovery potential in high energy physics with the exploration of the spectral distortions of • the CMB;

the ultimate means for precise determination of fundamental cosmological parameters such as the inﬂa- • ton potential shape from the measurement of the polarization E-mode power spectrum;

an unambiguous measurement of the total mass of the neutrinos from lensing reconstruction based on • detailed polarization observations.

For inflation, the natural goal is to be able to measure beyond doubt the tensor-to-scalar ration even for > 3 4 Higgs inflation, i.e., at the r 2 10− at 5σ, that is with a final uncertainty σ 2 10− . If this does × r ∼ × not lead to a detection, this∼ will discard altogether the whole class of “large field” models whose field excursion would be larger than the Planck mass. For neutrinos physics, future CMB data will allow to severely constrain by themselves the neutrinos sector, from measuring the total number of degrees of freedom to the sum of the neutrinos mass. This in turn will increase the constraining power of lower-redshift probes (like BAO), in particular those from the Euclid satellite, to the point of deciding their hierarchy of masses, normal or inverted. With such capabilities, the CMB constraining power on extensions to the standard base ΛCDM model will additionally be enormously increased, offering a minima a large increase in the leverage of other astrophysical probes such as Euclid and LSST, and potentially discovering the failure or limitation of the standard ΛCDM model.

French roadmap for CMB science 30/06/2016 17

3. Foregrounds obstacle After Planck, we know that the primordial B mode polarization of the CMB cannot be measured without − subtracting the Galactic foreground emission, even in the faintest dust-emitting regions at high Galactic latitude (Planck Collaboration Int. XXX 2016; BICEP2/Keck and Planck Collaborations et al. 2015). All CMB projects must have a strategy to perform component separation. They must also assess the conﬁdence with which it can be performed with their data. Any claim for a detection will face a critical assessment by the community against an alternative interpretation involving Galactic emissions residuals. Component separation will also be a challenge for any CMB experiment dedicated to the measurement of spectral distortions. Here we summarize our knowledge on foregrounds (with an emphasis on polarization) and our tools for

foreground subtraction. ThisPlanck is done Collaboration: in the context Di↵use component of future separation: CMB Foreground experiments maps but (as far as possible) independently of the experimental details. uncertainties. Statistical uncertainties are propagated from raw 30 44 70 100 143 217 353 545 857

sky maps to ﬁnal results by means of standard MCMC sampling ) Sum fg RJ 2 3.1.techniques, Emission while components various model errors are assessed by end-to- K µ 10 Thermal dust end simulations. All data products are made publicly available, Figureas summarized 12 provides in Table an5. overview of the main Galactic components in both temperature and polarization,CMB summa-

Three particularly noteworthy highlights from this analysis 1

rizedinclude in the terms following: of the r.m.s. brightness temperature evaluated10 over 93% and 73% of the sky, respectively. Dust and synchrotron emission are the dominant polarized foregrounds to the CMB. Dust dominates at frequencies – We have presented the ﬁrst full-sky polarized thermal map, Free-free higherwhich than is a direct about result 70 GHz of the and exquisite synchrotron sensitivity ofat the lower HFI frequencies0 (Planck CollaborationSpinning dust X 2016). These two instrument. This map will remain a cornerstone of future 10 CO 1-0 components are the main ones to consider, and probably remain the only ones of relevanceSynchrotron in the search for CMB cosmology for the decade or more, as the search for 2 primordial B-modes down to a tensor-to-scalar ratio r = 10− . Beyond this limit, additional polarized com-

primordial gravitational waves enters the next phase in which -1 RMS brightness temperature ( ponentsforegrounds will most are more probably important interfere than instrumental with the noise. separation of10 the CMB from the foreground. We introduce the – We have also presented a full-sky spinning dust intensity relevant components below. 10 30 100 300 1000 Planck Collaboration: Di↵use componentmap. separation: In addition Foreground to its mapsobvious scientiﬁc value, this map is Frequency (GHz) also interesting for algorithmic reasons, as a clear demon- uncertainties. Statistical uncertainties are propagated from raw stration of both30 the 44 importance 70 100 and 143 power 217 353of joint 545 global 857 30 44 70 100 143 217 353 ) sky maps to ﬁnal results by means of standard MCMC sampling analysis:) NeitherSum WMAP fg nor Planck have the statistical Synchrotron RJ RJ 2 2 K techniques, while various model errors are assessed by end-to- powerK to disentangle spinning dust from synchrotron, but to- µ 10 µ 10 Thermal dust end simulations. All data products are made publicly available, gether beautiful new results emerge. We believe that this will be the default approach for virtually all future microwave as summarized in Table 5. CMB surveys, as no experiment will have the power to replace 1

Three particularly noteworthy highlights from this analysis 1 10 include the following: Planck10 and WMAP by themselves. Rather, each new exper- Thermal dust iment will contribute with a new critical piece of informa- Sum fg – We have presented the ﬁrst full-sky polarized thermal map, tion regarding a given phenomenon or frequency range, and

Free-free 0 which is a direct result of the exquisite sensitivity of the HFI 0 Spinning dust

thereby help refining the overall picture. Global Bayesian 10 10 CMB instrument. This map will remain a cornerstone of future CO 1-0 analysis provides a very natural frameworkSynchrotron for this work. CMB cosmology for the decade or more, as the search for – A second useful illustration of the power of global analy- -1 primordial gravitational waves enters the next phase in which sis-1 presented in this paper is the identification of important RMS brightness temperature ( RMS brightness temperature ( 10 foregrounds are more important than instrumental noise. instrumental10 systematic errors. One example is the detec- – We have also presented a full-sky spinning dust intensity tion of, and correction for, systematic errors in the Planck 10 30 100 300 1000 10 30 100 300 1000 map. In addition to its obvious scientific value, this map is bandpass measurements.Frequency More generally, (GHz) the residual maps Frequency (GHz) also interesting for algorithmic reasons, as a clear demon- shown in Figs. 2, 21 and 40 comprise a treasure trove of in- stration of both the importance and power of joint global Figureformation 12: Spectral on instrumental30 (electromagnetic) 44 70systematics 100 143 distribution that 217 should 353 prove of foregrounds very emissions in total intensity (left) and polarization (right). analysis: Neither WMAP nor Planck have the statistical valuable) Synchrotron for improving the raw Planck sky maps before the Fig. 49. Brightness temperature rms as a function of frequency MapsRJ are smoothed at 1 degree and r.m.s. are estimated between 80 and 90% of the sky. Units are r.m.s brightness 2 power to disentangle spinning dust from synchrotron, but to- nextK data release. and astrophysical component for temperature (top) and polar- µ gether beautiful new results emerge. We believe that this will temperature.10 Planck Collaboration X(2016). ization (bottom). For temperature, each component is smoothed be the default approach for virtually all future microwave All things considered, the sky model presented in this paper to an angular resolution of 1 FWHM, and the lower and up- surveys, as no experiment will have the power to replace provides an impressive fit to the current data, with temperature per edges of each line are defined by masks covering 81 and 1

residuals10 at the few microkelvin level at high latitudes across Planck and WMAP by themselves. Rather, each new exper- Thermal dust 93 % of the sky, respectively. For polarization, the correspond- the CMB dominated frequencies, and with median fractional ering smoothing scale is 40 , and the sky fractions are 73 and 93 %. iment will contribute with a new critical piece of informa- 3.1.1. Galactic dust andSum synchrotron fg polarization 0 tion regarding a given phenomenon or frequency range, and rors below 1 % in the Galactic plane across the Planck frequen- Note that foreground rms’s decrease nearly monotonically with cies. For0 polarization, the residuals are statistically consistent sky fraction, whereas the CMB rms is independent of sky frac- thereby help refining the overall picture. Global Bayesian 10 CMB analysis provides a very natural framework for this work. Non-sphericalwith instrumental dust noise grains at high have latitudes, their but spin limited axis, by signif- perpendiculartion, up to random their long variations. axis, statistically aligned with the – A second useful illustration of the power of global analy- localicant temperature-to-polarization orientation of the Galactic leakage magnetic in the Galactic field. This plane. alignment results in dust emission polarized perpendicular sis presented in this paper is the identification of important Overall,-1 this model represents the most accurate and complete RMS brightness temperature (

to the10 magnetic field projection on the plane of the sky. Dust polarized emission has been measured over the instrumental systematic errors. One example is the detec- description currently available of the astrophysical sky between trophysical foregrounds at low frequencies. As emphasized re- tion of, and correction for, systematic errors in the Planck whole20 and 857 sky GHz. by Planck. Much has been learned from thesepeatedly, data. even when combining the Planck and WMAP observa- 10 30 100 300 1000 bandpass measurements. More generally, the residual maps TheFigure intrinsic49 provides dust an polarization overviewFrequency of (GHz) theis now main known components to be in 25%.tions, as This done is in the this value paper, we degeneracies would observe between for synchrotron, a uniform shown in Figs. 2, 21 and 40 comprise a treasure trove of in- both temperature (top panel) and polarization (bottom panel),∼ free-free and spinning dust remain the leading source of uncer- formation on instrumental systematics that should prove very magneticsummarized field in terms in the of plane the brightness of the sky. temperature In the data, rms sucheval- hightainty values on the are low only frequency measured side. over Additional a very observations small fraction be- valuable for improving the raw Planck sky maps before the ofuatedFig. the 49. over skyBrightness 93 (Planck % and temperature 73 Collaboration % of the rms sky, as respectively. Int. a function XIX 2015 of For frequency). polar- The observedtween, say, values 2 and 20 are GHz on are average essential lower to break (the these mean degenera- value next data release. towardsization,and astrophysical this the is southern the component first version Galactic for of temperature cap such is a 12 plot (top that1%)) and is and based polar- highlycies. variable For a more over complete the sky analysis over a of range the low-frequency of scales. These fore- onization observations (bottom). alone. For temperature, For temperature, each component the most± recent is smoothed pre- ground model presented here, we refer the interested reader to All things considered, the sky model presented in this paper resultsviousto an version angular reflect is resolution a figure significant 22 of of 1Bennett depolarizationFWHM, et al. and(2013 the e), lowerffect, summariz- andassociated up- Planck with Collaboration the structure XXV of(2015 the turbulent). component of the provides an impressive fit to the current data, with temperature magneticingper the edges WMAP field of each temperature along line the are foregroundlinedefined of bysight, masksmodel. which covering While varies the 81 two over and theOn sky the (Planck high-frequency Collaboration side, the main Int. XLIV outstanding 2016 issue). are residuals at the few microkelvin level at high latitudes across versions93 % of agree the sky, well respectively. in terms of For total polarization, foreground thepower correspond- and lo- uncertainties in the net 545 and 857 GHz calibration, i.e., the the CMB dominated frequencies, and with median fractional er- cationing smoothing of the foreground scale is 40 minimum,0, and the sky there fractions are a few are 73 subtle and 93 dif- %. product of calibration and bandpass uncertainties. As of today, rors below 1 % in the Galactic plane across the Planck frequen- ferencesNote that as foreground well. The rms’smost importantdecrease nearly of these monotonically is the relative with the 545 GHz calibration is uncertain at least at the 1–2 % level, cies. For polarization, the residuals are statistically consistent Frenchamplitudesky fraction, roadmap of synchrotron whereas for the CMB and CMB spinning science rms is independentdust. Specifically, of sky syn- frac- and this translates into an e↵ective 3–6 % uncertainty 30/06 for/2016 the with instrumental noise at high latitudes, but limited by signif- chrotrontion, up dominates to random over variations. spinning dust at all frequencies in the 857 GHz channel in our fits, in order to maintain a physical ther- icant temperature-to-polarization leakage in the Galactic plane. WMAP model, whereas in our new model spinning dust dom- mal dust frequency scaling. Cross-correlations with H i observa- Overall, this model represents the most accurate and complete inates over synchrotron between 15 and 60 GHz. Such di↵er- tions suggests a total systematic error on the thermal dust tem- description currently available of the astrophysical sky between encestrophysical are not foregrounds surprising, considering at low frequencies. the complexity As emphasized of the as- re- perature at high Galactic latitudes of 1–2 K. Recognizing both 20 and 857 GHz. peatedly, even when combining the Planck and WMAP observa- Figure 49 provides an overview of the main components in tions, as done in this paper, degeneracies between synchrotron, both temperature (top panel) and polarization (bottom panel), 58free-free and spinning dust remain the leading source of uncer- summarized in terms of the brightness temperature rms eval- tainty on the low frequency side. Additional observations be- uated over 93 % and 73 % of the sky, respectively. For polar- tween, say, 2 and 20 GHz are essential to break these degenera- ization, this is the first version of such a plot that is based cies. For a more complete analysis of the low-frequency fore- on observations alone. For temperature, the most recent pre- ground model presented here, we refer the interested reader to vious version is figure 22 of Bennett et al. (2013), summariz- Planck Collaboration XXV (2015). ing the WMAP temperature foreground model. While the two On the high-frequency side, the main outstanding issue are versions agree well in terms of total foreground power and lo- uncertainties in the net 545 and 857 GHz calibration, i.e., the cation of the foreground minimum, there are a few subtle dif- product of calibration and bandpass uncertainties. As of today, ferences as well. The most important of these is the relative the 545 GHz calibration is uncertain at least at the 1–2 % level, amplitude of synchrotron and spinning dust. Specifically, syn- and this translates into an e↵ective 3–6 % uncertainty for the chrotron dominates over spinning dust at all frequencies in the 857 GHz channel in our fits, in order to maintain a physical ther- WMAP model, whereas in our new model spinning dust dom- mal dust frequency scaling. Cross-correlations with H i observa- inates over synchrotron between 15 and 60 GHz. Such di↵er- tions suggests a total systematic error on the thermal dust tem- ences are not surprising, considering the complexity of the as- perature at high Galactic latitudes of 1–2 K. Recognizing both

58 Planck Collaboration: Frequency dependence of thermal emission from Galactic dust in intensity and polarization 18 3 FOREGROUNDS OBSTACLE Table 6. Mean microwave SED for polarization computed using the CC analysis.

α Frequency [GHz] The polarization power spectra of the dust are well described by power laws in multipole, C` ` , with ∝ Experiment the same exponent α 2.4 for both the EE and BB spectra. The amplitudes of the polarization power ' − Quantity WMAP Planck WMAP WMAP Planck WMAP Planck WMAP Planck Planck Planck Planck spectra are observed to scale with the average dust brightness as < I >231.9 (Planck 28.4 Collaboration 33 41 Int. 44.1 XXX 61 70.4 94 100 143 217 353 [˜↵P]1T . . . . . 0.9481 0.4038 0.3351 0.1793 0.1525 0.1179 0.1129 0.1852 0.1900 0.3029 0.5624 1.0000 2016). A systematic difference has been discovered betweenh ⌫ 353 thei amplitudes of the Galactic B- and E-modes, BB EE stat ...... 0.1201 0.0538 0.0402 0.0292 0.0190 0.0198 0.0118 0.0261 0.0050 0.0048 0.0062 0.0068 such that C` /C` = 0.5. Planck Collaboration Int. XXXVIIIc⌫ [%] .(2016 . . . . .) showed 1.0 that 1.0 the correlation 1.0 1.0 between 1.0 the 1.0 0.5 1.0 0.5 0.5 0.5 1.0 filamentary structure of matter and the Galactic Magnetic Fieldcmb . . (GMF) . . . . . orientation 0.0006 0.0006 may account 0.0006 for 0.0006 the E 0.0006and B 0.0005 0.0003 0.0005 0.0002 0.0002 0.0001 0.0000 tot ...... 0.1204 0.0539 0.0403 0.0293 0.0190 0.0199 0.0118 0.0262 0.0051 0.0050 0.0067 0.0114 asymmetry, as well as the TE correlation, reported in theS analysis/N ...... of. . thepower 7.9 spectra 7.5 of 8.3 the Planck 6.1353 8.0 GHz 5.9 9.6 7.1 37.1 60.4 83.6 87.7 polarization maps. P 1T Notes. [˜↵ ] Mean polarization SED in KRJ units, normalized to 1 at 353 GHz, from the correlation with the 353 GHz templates. The values h ⌫ 353i⌘ The power spectra of the synchrotron polarization haveare not been colour measured corrected. withstat Statistical both Planck uncertaintyand onWMAP the meandata polarization SED. c⌫ Uncertainties on the inter-calibration [%] between Planck and WMAP frequencies⌘ (Planck Collaboration I 2014; Bennett et al. 2013). Uncertainty⌘ on the mean polarized SED introduced by (Planck Collaboration X 2016). The associated power-law spectral indices are similar to those measured for cmb ⌘ the CMB-subtraction multiplied by the inter-calibration factor c⌫. tot Total uncertainty on the mean polarized SED. dust polarization.Planck Collaboration: Both emissions Frequency show dependence the same of thermal asymmetry emission from between GalacticE dustand inB intensitymodes, and polarizationwhich is thought⌘ to arise for the synchrotron emission from the alignment of the magnetic field with the orientation of emission 1 15 50 features. The spectral energyPlanck distribution and the main emission10 features (the radio loops) of theDI+AI synchrotron

1 Planck PSI+PDI WMAP maps10 are discussed in Planck Collaboration XXV(2016). In Planck Collaboration Int. XXII(2015), thePlanckPlanck 10 WMAP Planck and WMAP data are combinedThermal dust to characterize (MBB) the frequency dependence of emissionSynchrotron that is(PL) spatially correlatedWMAP WMAP Analytical AME model 25 0 with dust emission at 353 GHz, for both intensity and polarization,5 in a consistentThermal manner. dust (MBB) At ν 100 GHz,

DI+AI 10 ≥ 0

PSI+PDI the mean spectral energy distribution (SED) of the correlated emission is well fitted by a modified black body 10 3T 353 ] 1T 353 0 ] I spectrum for which the mean dust temperature of 19.6 K; this is derived from an SED fit of the dust total inten- 0 P [˜ [˜ sity up to 3000 GHz (100 µm). It is found that the opacity has a spectral index of 1.59 0.02 for polarization, 1 5

1 ± a value very close (slightly lower) to that measured for the total intensity (see Fig. 13). This analysis reveals 10 10 25 100x(Data-Model)/Model 100x(Data-Model)/Model a signiﬁcant correlation between dust and synchrotron polarization. It also shows that the SEDs do not fully 10 constrain the emission mechanisms. For example, the low frequency component in the polarization SED of 2 Fig. 13 could include a non-negligible contribution from spinning2 dust, which the Planck team was not able to 15 50 10 distinguish20 from30 40 dust-correlated50 70 100 synchrotron150 200 polarization.300 400 2010 20 3030 40 50 70 100100 150150 200200 300300 400400 20 30 40 50 70 100 150 200 300 400 [GHz] [GHz][GHz] [GHz] 1 15 50 10 Planck Planck DI+AII PSI+PDI (ﬁxed P) 1 s WMAP 10 WMAP Planck 10 Planck Thermal dust (MBB) Synchrotron (PL) WMAP

25 WMAP

SpDUST WNM 0 Thermal dust (MBB) 5

shifted SpDUST CNM 10 P 0 PSI+PDI (ﬁxed )

10 DI+AII 3T 353 1T 353 ] ] 0 0 I P [˜ [˜ 1 5 1 10 25 10 100x(Data-Model)/Model 100x(Data-Model)/Model 10 2 2 50 15 10 10 20 30 40 50 70 100 150 200 300 400 2020 3030 40 50 70 100100 150150 200200 300300 400400 20 30 40 50 70 100 150 200 300 400 [GHz] [GHz][GHz] [GHz]

Fig. 11. Mean polarized SED in KRJ units, normalized to 1 at 353 GHz correlated with the Stokes Q and U 353 GHz maps. The polarized spectral Fig.Figure 8. Mean 13: dust Mean SED dust in K SEDRJ units, in normalized K units to of 1 at the 353 dust-correlated GHz, with di↵erent emission, spectral fits normalized and the respective to unity residuals. at 353 The GHz, two with parametric different model RJ model with and without the constraint on P match the observed data points. fits are DI+AI (top left), and DI+AII (bottom+ left), as presented in Sect. 7.2. Right: residuals after removing the+ bests fit model listed in Table 4, fromspectral the mean fits: dustleft: SED.for Theintensity two spectral (DI AII), models a provide modified good black-body fit to the data, for with the residuals thermal compatible dust emission with zero.a linear combination of two spinning dust components arising from the typical cold neutral medium (CNM) and warm neutral medium (WNM); right: for polarization (PSI+PDI), superposition of a power-lawThe synchrotron polarization spectrum SED first and decreases a modified with black-body decreasing for fre- the 9.2. Low frequency rise of the polarization SED polarized map at 353 GHz (P ) for all the sky patches. To com- quency,The then mean turns value up below of the 60 dust GHz. spectral This is index the first for timepolarization that thermal dust emission. Figure353 from Planck Collaboration Int.such XXII a( behavior2015). has been observed for polarized emission corre- In this section, we show that a synchrotron component correlated pute P , we use a 1 smoothed map of P derived in Planck is di↵erent from that for intensity, 1.51 0.01 (Sect. 6) over 353 353 lated with dust polarization, though it has been± seen before for with dust is the most likely interpretation for the low frequency P the same sky area. In the next section, we check whether the Collaboration Int. XIX (2015). We derive for each sky the total sky polarization (Bennett et al. 2013). rise of the polarization SED. d,mm di↵erence of spectral indices in intensity and polarization is a patch from RP (353, 217), taking into account the local estimate 100 robust result against systematics present in the polarization data. 3.1.2. AdditionalI Galactic components A107, page 15 of 25 of Td derived from R (3000, 857) (Sect. 5.2). We assume that the temperature of the dust grains contributing to the polarization is 8.2. Uncertainties in P theSeveral same as additional that determined components for the dust may emission contribute, in intensity. at a fainter level than thermald,mm dust and synchrotron emission, Thisto the is not sky necessarily polarization true atif the microwave polarization frequencies. is associated These with includeFor the mean the dipolar polarized emission dust spectral from index, small we spinning use the resultsdust specificgrains, dust molecular grains, e.g., line the emission silicates versus (mainly carbon the dust CO (Martin rotationalfrom lines), full mission magneticPlanck dipolarpolarization emission maps from with the ferromag- two detec- 2007; Draine & Fraisse 2009). This should be kept in mind in netic particles and the zodiacal light from interplanetarytor dust. set maps The as anomalous fixed templates microwave (Sect. 8.1). emission To estimate (AME, the sys- thinking of physical interpretations. Here we use the spectral tematic uncertainty for the mean P , we apply the CC analysis indicesFig. 13)P is thoughtand I toas arise a mathematical mainly from way small to quantify spinning the dust particles but it could also included,mm a significant con- d,mm d,mm on multiple subsets of the Planck data, including the combina- ditribution↵erence between from ferromagnetic the dust SED for particles intensity and and polarization. inclusions intotion silicate of yearly dust maps grains. (YR1 We and briefly YR2), discuss the full mission each of half-ring these P P The scatter on the R100(353, 217) values increases for 353 < maps (HR1 and HR2), the combination of odd surveys (S1+S3) P 20 µK due to data noise. The histogram of d,mm from the 400 sky and even surveys (S2+S4), and the detector set maps (DS1 and P DS2; see Sect. 2.1.2 for more details). We use these subsets of patches is presented in Fig. 9. The distribution of d,mm has a meanFrench value roadmap of 1.592 for (round-o CMB↵ scienceto 1.59), with a 1 dispersion the data as maps and templates at 353 GHz. Table 30/065 /lists2016 the P of 0.174 (round-o↵ to 0.17). This dispersion is the same if we derived mean d,mm, for all the sky patches from each com- P use the mean dust temperature of 19.6 K for all sky patches. The bination of the data subsets. The dispersion of the d,mm val- P statistical uncertainty on the mean d,mm is computed from the ues in Table 5, 0.02, is consistent with the 1 dispersion on 1 dispersion divided by the square root of the number of inde- the mean polarization spectra index from statistical uncertain- pendent sky patches (400/Nvisit) used, which is 0.02. ties estimated in Sect. 8.1, making it dicult to separate the

A107, page 13 of 25 3.1 Emission components 19 components. Little is presently known about them. Sensitive imaging observations for a number of spectral bands, spread over a large frequency range, are required to identify them and deal with them when separating components. The spinning dust emission may be polarized at the level of 1% (Hoang et al. 2013) but this remains a ∼ model prediction open to debate and unconfirmed by observations. The polarization fraction is expected to decrease for increasing frequencies. Observations around 20 GHz of compact sources with, e.g., the Quijote experiment (Génova-Santos et al. 2015) represent the most promising approach to detect the polarization of this emission in the coming years. The CO emission was predicted to be polarized at a level up to 10 % (Goldreich & Kylafis 1982) and this polarization has been measured in different sources (e.g. Li & Henning 2011). The polarization is expected to be the highest for CO emission with moderate optical depth from the diffuse interstellar medium. It may thus be significant over the high Galactic latitude sky area used for CMB studies. The analysis of CO polarization with the Planck data has been so-far hampered by systematics. To measure it, very precise bandpass measurements are needed per detector. Part of the AME may be accounted for by magnetic dipolar emission. This emission, if present, could contribute to the microwave polarization in a direction perpendicular to that of thermal dust (Draine & Hensley 2013; Hoang & Lazarian 2016). The emission from interplanetary dust was detected at sub-mm wavelengths with the Planck data. It contributes large scale emission that is time-variable. This emission may be polarized although we do not know of any quantitative model to give some numbers.

3.1.3. The complexity of Galactic foregrounds

The microwave dust polarized emission B mode amplitude is at a level of r 0.1 at the recombination bump − ∼ (` = 80) in the cleanest sky patches at high Galactic latitude (Planck Collaboration Int. XXX 2016). To reach the accuracy targeted by the next generation of ground-based CMB experiments (detection of primordial B- 2 3 modes down to r = 10− within the next few years and 10− later), the separation between Galactic polarization and the primordial CMB has to be performed with an exquisite precision. To achieve this one will have to face the complexity of Galactic foregrounds. Several physical processes couple the emission properties of dust grains and their alignment with the density structure of matter and that of magnetic fields. Likewise the cosmic-ray energy spectrum, and thereby the synchrotron emission spectrum depend on the magnetic field structure and the distribution of localized energy sources. These physical couplings break the simplest assumption for component separation by which the spectral frequency dependence of the Galactic polarization and its angular structure on the sky are separable. They make polarized foreground intrinsically complex. The complexity arising from the interplay of emission processes with interstellar MHD turbulence is already apparent in the Planck data. Planck maps provide evidence for spatial variations of the spectral behaviour of the polarized dust emission. This possibility was first investigated at large angular scales in Planck Collaboration Int. XXX(2016) by comparing the amplitudes of power-law fits to auto and cross-spectra of dust polarization at 353 and 217 GHz, but no evidence for variability was found. A recent study is extending this work towards the smaller angular scales by computing the angular power spectra correlation ratio between 353 and 217 GHz across all the accessible multipoles (Planck Collaboration Int. L 2016). This work shows significant variations of the dust polarization SED at high Galactic latitude, larger than those measured for dust intensity. These variations are thought to reflect the interplay between dust polarization properties and interstellar MHD turbulence. Correlations along the line of sight, a key but often ignored dimension of Galactic foregrounds, may generate changes of both the polarization fraction and the polarization angle over microwave frequencies. Similar effects are expected for synchrotron polarization. There is an on-going effort in France to model these effects for both dust and synchrotron polarization in order to provide simulated maps which may be used to assess uncertainties in component separation in present and future experiments. The spatial variations of the dust polarization SED measured by Planck can mistakenly be interpreted as a (false) detection of primordial CMB B modes. Data must have the necessary spectral (and spatial) coverage − to properly take them into account. They may be large enough to prevent ground-based experiments to reliably detect primordial B modes down to r = 10 2, without complementary observations (in balloon or in space) − −

French roadmap for CMB science 30/06/2016 20 3 FOREGROUNDS OBSTACLE that will map dust polarization at higher frequencies not accessible from ground. This is even more true for higher precision.

3.1.4. Extragalactic emissions

Cosmic Infrared Background (CIB) The CIB power spectrum can be considered as the sum of two contributions usually called the “1-halo” and “2-halo” terms. The 1-halo term represents the correlation of two different galaxies in the same dark-matter halo (pairs of galaxies inside the same halo); the 2-halo term, capturing the galaxy correlations in different dark-matter haloes, describes the large-scale clustering. In polarization, the 1-halo term is smaller than the Poisson noise of polarized extragalactic sources while the 2-halo term is null provided that there is no correlation of the polarization of galaxies within distinct halos. We consider here that the correlation between the spin of halos and the orientation of cosmic web filaments (Codis et al. 2012) has a negligible impact.

100 100 100 GHz 70 GHz 10-1 143 GHz 10-1 100 GHz 217 GHz 143 GHz ] ]

2 -2 2 -2 217 GHz

K 10 K 10 µ µ [ [ π π 2 2

/ -3 r=0.1 / -3 r=0.1

` 10 ` 10 C C ) ) 1 1

+ -4 r=0.01 + -4 r=0.01

` 10 ` 10 ( ( ` `

10-5 r=0.001 10-5 r=0.001

10-6 10-6 100 101 102 103 100 101 102 103 ` `

Figure 14: Polarization power spectra for the shot noise compared to the CMB B-mode gravitational wave contribution when r = 0.1 (solid), 0.01 (dashed-dot) and 0.001 (dashed). left: dusty star-forming galaxies (considering 0.5% polarization) with a flux detection limit of 340, 250, and 200 mJy at 100, 143 and 217 GHz respectively (typical of a Planck-like experiment); right : radio galaxies, with a flux detection limit of 100 mJy (numbers from Tucci & Toffolatti 2012).

Point sources Point sources may be polarized but their orientation is not correlated on the sky and polarization emission average to zero in the mean. However, we expect a contribution in the 2-point correlation function. The two populations of sources that matters at mm-frequencies are dusty star-forming galaxies and radio galaxies. The polarization amplitude of dusty galaxies is likely to be low. A polarization fraction of 0.4% is measured on M82 at 850 µm (Greaves & Holland 2002). On the contrary, radio galaxies are a more important foreground contaminant for CMB measurements (Battye et al. 2011; Tucci & Toﬀolatti 2012). Figure 14 shows the expected level of polarized shot noise for radio and dusty galaxies.

3.1.5. Spectral distribution of foregrounds in intensity For intensity, many more components than for polarization need to be considered. Figure 15 shows the estimated average (zero-level) of CMB and foreground emissions spectra, together with typical expected y and µ spectral distortions. The zero-level spectra of foregrounds are estimated on the best 50% of sky for each individually. Modelling uncertainties can be at the level of tens of percent to factors of two. Emissions from molecular lines such as CO and isotopologues, HCN, HCO+, H2O, not represented, would also be detectable in many frequency bands. This figure outlines the difficulty of measuring primordial distortions of the CMB spectrum, which will be in each frequency band at least 2 to 3 orders of magnitude below the total foreground emission. Moreover, such measurements will require the identification of each component individually and then to determine their SEDs

French roadmap for CMB science 30/06/2016 3.2 Component separation 21 Courtesy J. Delabrouille

Figure 15: Average (zero-level) of CMB and foreground emissions spectra compared to expected y and µ spectral distortions. with suﬃcient accuracy and their contribution to the monopole. This will be the major astrophysical challenge for an experiment like PIXIE (cf. Sect. 8.1.3).

3.2. Component separation

104 primordial B modes 103 E modes temperature 2 lensing B modes

] 10 2

K 101 µ

[ du st +synchro C` tron ` 0 fsky =90 dust % @ 2 C +synchro 00GH 10 C` tron z

) fsky =90% @ 100GH

π z -1 2 10 / )

1 du st +synchro -2 C` tron + 10 f = sky 1% @ 200 ` GHz (

` -3 r=0.1 ( 10 du st +synchro C` tron fs =1 ky % @ 100G 10-4 r=0.01 Hz

10-5 r=0.001 101 102 103 `

Figure 16: Angular power spectra showing primordial B modes, lensing B modes, total intensity, and E modes, as well as the total contribution of polarized B-mode foregrounds (dust plus synchrotron), expected on the cleanest 1% to 90% of the sky, at 100 and 200 GHz. Note that, as these results are derived from Planck’s Galactic masks and are not therefore optimized for high-resolution, ground-based instruments, there is potential for discovery of small patches of sky (e.g., fsky . 5%) cleaner than those indicated here. From Errard et al.(2016), made with τ = 0.066.

As seen previously in Fig. 12, foreground emission is a challenge for observing the CMB anisotropies in temperature and polarization. Figure 16 shows CMB temperature, E-mode and B-mode angular spectra for diﬀerent values of the tensor-to-scalar ratio r, as well as the expected amplitude of dust and synchrotron B modes, estimated using various portions of the Planck data. It should be apparent that, in the pursuit of

French roadmap for CMB science 30/06/2016 22 3 FOREGROUNDS OBSTACLE primordial B-mode measurements, correcting for the foreground contribution is unavoidable. Indeed, Errard et al.(2016) estimate that on 50% sky coverage, the minimum foreground is reached at an angular scale ` 80 ≈ and frequency ν 74 GHz with an effective amplitude of r 0.1. This does not necessarily mean that all ≈ ∼ experiments have to bracket this frequency, since foregrounds dominant in different frequency ranges vary in complexity and strength, and different cleaning strategies may be envisaged for different scientific objectives. But in any case, reaching the ambitious target of σ 10 3 or lower necessitates highly efficient foreground r ∼ − cleaning (together with a corresponding treatment of lensing B modes). The general process of the separation of the cosmological signal from foreground contributions can be referred to as component separation. A wide range of approaches have been developed in the past two decades, and are briefly summarized here. They generally differ in the assumptions they make about the foregrounds, and whether they are applied in pixel-space, harmonic-space, or via wavelet-type operators. However, the approaches are limited by the number of available frequency channels, since the number of free parameters adopted in the model cannot exceed this.

3.2.1. Template removal

Historically, fits to external templates of the foreground emission, based on observations at higher or lower frequencies where the emission from single foreground components are expected to dominate, have generally been considered to provide an adequate solution to the component separation problem. However, the use of measurements of the foreground emission at wavelengths far-removed from the microwave regime can result in large errors when fitting the resultant template sky emission to the data. In particular, real spectral index variations in the foregrounds can lead to significant changes in the morphology of the foreground emission, and residual signal can remain in the cleaned data. It should also be noted that there is a lack of high quality external templates for polarized emission. However, with suitable multi-frequency observations of the microwave sky, the necessary information can be extracted directly from the observed sky maps. In some cases, specific frequencies are adopted to trace specific components, e.g., within Planck the 353 GHz map is used to trace polarized thermal dust emission. Alternatively, map differences at varying frequencies can be computed, such that the resulting maps then consist only of linear combinations of Galactic foregrounds and noise, with no CMB component. The foreground- cleaned sky is obtained by fitting and subtracting these so-called internal templates from the CMB-dominated channels. Since the templates are now constructed from the microwave data itself, only small errors due to spectral index variations can be introduced into the analysis. This idea has been utilized by the WMAP team by considering the difference of their K and Ka band data as a template for low frequency emission in temperature (Hinshaw et al. 2007). It also forms the basis for the SEVEM approach (Fernández-Cobos et al. 2012) that has been used for both temperature and polarization component separation by Planck (Planck Collaboration et al. 2015c).

3.2.2. Blind algorithms

Other approaches are based on concepts from image processing. These so-called blind or semi-blind algorithms make minimal speciﬁc assumptions about the number or statistical nature/morphology of the foreground components. The advantage of blind methods is their ability to treat unknown or complex foreground contamination. Examples include the Internal Linear Combination (ILC) approach of Bennett et al.(2003), the Independent Component Analysis (ICA) method as implemented by Maino et al.(2002), the Spectral Matching Independent Component Analysis (SMICA) described by Delabrouille et al.(2003), and the Correlated Component Analysis (CCA) due to Bonaldi et al.(2007). The ILC approach has been widely adopted in part due to its simplicity, in that it makes minimal assumptions concerning the foregrounds and seeks only to minimize the variance of the CMB, i.e., the sky component possessing a black body spectrum. One potential problem with the approach is the bias that can arise from cross-correlations between the CMB and foreground components (Hinshaw et al. 2007). A needlet (wavelet on the sphere) version of the ILC referred to as NILC (Delabrouille et al. 2009) was utilized by Planck for both temperature and polarization data, together with SMICA. The latter is a non-parametric method that works in the spherical harmonic domain. Foregrounds are modelled as a small number of templates with arbitrary

French roadmap for CMB science 30/06/2016 3.3 The foregrounds challenge 23 frequency spectra, arbitrary power spectra and arbitrary correlation between the components. On Planck, it usually gave the most precise CMB determinations.

3.2.3. Parametric approaches

The parametric approach makes the maximum use of prior knowledge of foreground emission and fits foreground unknowns along with the CMB. In general, a parametric spectral model is assumed for each signal component, and the parameters fitted on a pixel-by-pixel basis. The advantages of this approach follow. The fitted model may be chosen freely, and the method is therefore completely general, all assumptions are trans- parent, no restrictions on spatial variations of foreground properties are imposed and the results may be rigorously monitored by goodness-of-fit tests. Most importantly, the methodology allows a joint CMB estimation and component separation and therefore reliable error estimates are obtained on all estimated quantities with foreground uncertainties rigorously propagated through to CMB power spectrum and cosmological parameter inference. A specific implementation of a Bayesian parametric method, Commander (Eriksen et al. 2006, 2008), was adopted for Planck temperature and polarization analysis, leading not only to estimates of the CMB but also the foreground emission (Planck Collaboration X 2016).

3.3. The foregrounds challenge

Given our current knowledge of the foreground emissions and complexity, we think we have in-hand tools to reach the level of residuals required for the detection of a tensor-to-scalar ratio at a level down to σ 0.01. r ∼ In fact, several studies, particularly those considering proposed space missions (Katayama & Komatsu 2011; Errard et al. 2016), claim that a significant detection of r at a level of 0.001 is feasible. However, such results are often derived either under simplifying assumptions – ignoring complexities in the Galactic foreground emission due to synchrotron and dust, and neglecting potential contaminants such as AME and point-sources – or by adopting component separation methods, generally parametric in nature, that essentially assume a model well-matched to the simulated foregrounds under study. Remazeilles et al.(2016) test some of these assumptions explicitly, and find biases in the derived value of r at this level due to the presence of a 1% polarized AME component unaccounted for in the foreground model, or by neglecting the curvature of the synchrotron emission law in the parametric fit to the data. Moreover, a real-world example of a false detection due to a bias in the foreground removal procedure was already seen in the initial claims of the Bicep2 team, arising due to the impact of a mismatch between real and assumed foreground spectral behaviour.

Thus for higher precision determination of r (smaller σr needed for a mere detection at lower values of r), the situation is less clear. It will certainly be more difficult to rely only on the current knowledge of Galactic emission, the complexity of which has become apparent with recent Planck studies, and additional information will be required both on its spectral behaviour and spatial distribution. Moreover, the uncertainty coming from the foreground models will become the dominant part in the error budget in the near future. An accurate propagation of those uncertainties must be a priority in order to give realistic limits and eventually claim a detection of the B-mode signal with confidence. This mandates a continuing effort on foreground simulations and astrophysical modelling to be able to deal with more complex spectral dependencies and spatial distributions. Work on Planck data has shown that the combination of both angular resolution and multiple spectral bands over a large frequency range is essential to discriminate foreground components.

Finally, the connection between foreground emission and systematic error contributions to observations of the sky should be considered. A particular example arises when measurements from multiple detectors are required in order to reconstruct the polarization signal adequately. Bandpass mismatch between the detectors can lead to spurious signals arising from the leakage of temperature signal into polarization. Modelling such a contribution can be problematic, particularly at lower frequencies, where the temperature foreground emission certainly has additional signiﬁcant emission components that still remain diﬃcult to disentangle (Planck Collaboration X 2016; Planck Collaboration XXV 2016). Even when single detectors are used to measure

French roadmap for CMB science 30/06/2016 24 4 SCIENCE BEYOND THE PRIMARY CMB SCIENCE polarisation (by rotating them around each sky direction), large bandpasses raise the issue of the accuracy of per-pixel colour-corrections7, which require a high precision bandpass determination for each detector.

3.4. Summary

Every CMB experiment requires a well thought-out foreground mitigation strategy, in order to interpret correctly what is being measured. This leads to the following conclusions and recommendations.

Given our current knowledge of foreground emission, we believe that component separation approaches • can rapidly achieve the level of residuals required for σ 0.01. r ∼ Nevertheless, one must be careful about spatial variations in the SED that may bias the reconstruction • of CMB B-modes, if not accounted for in the analysis. This has implications on the frequency range required to allow adequate foreground modelling. For a more ambitious goal, typically σ 0.001 or lower, additional information will be required • r ∼ both on the spectral behaviour and spatial distribution of Galactic foregrounds. Moreover, other polarized foreground emission will become important (in particular, radio sources at frequencies less than 100 GHz). Since little is known about the zero-levels (monopoles) of individual contributions to the foregrounds, • component separation methods should be developed for this speciﬁc case in order to realise the scientiﬁc potential of CMB spectral distortion measurements. The uncertainty arising from foreground residuals and their potential interplay with instrumental sys- • tematics will become the dominant term in the error budget. Dedicated programs (observational and simulation-based) developed for understanding foregrounds • must be well supported.

As we shall see below, atmospheric limitations make frequencies above 220 GHz hard and expensive to measure from the ground, and frequencies above 280 GHz unattainable at the required sensitivity. ∼ Exploiting the full potential of the CMB window mandates using balloons and, ultimately, a much longer duration space mission.

4. Science beyond the primary CMB science

4.1. Galaxy Cluster and Large-Scale Structure Science Over the past decade, large surveys have signiﬁcantly advanced galaxy cluster (GC) and large-scale structure (LSS) studies for both cosmology and astrophysical science. The surveys have opened new spectral windows and observed large sky fractions to produce multi-wavelength cluster samples with data in the optical/near- infrared (NIR), millimetre and X-ray bands for thousands of objects with redshifts 0 z 1. In particular, the ≤ ≤ surveys by Planck, the Atacama Cosmology Telescope (ACT) and the South Pole Telescope (SPT) pioneered novel techniques now ripe for further exploitation. The discovered clusters are currently targeted by follow-up multi-wavelength campaigns. The millimetre band, giving direct access to the pressure of the hot gas in clusters and LSS will continue playing a key role.

4.1.1. Cluster Astrophysics and Cosmology Finding clusters and proto-clusters The current generation of millimetre experiments have harvested the 0 < z < 0.5 redshift range for cluster studies and begun ﬁnding clusters in the 0.5 < z < 1.5 window. Future experiments will potentially detect objects out to z 3, depending on the exact heating of the gas in early clusters. These systems are of particular ∼ 7 Colour corrections arise from the fact that the integral in a wide band is rather sensitive to the spectral behaviour of each component resulting in, e.g., an eﬀective mean frequency of the band per component; this can only be accounted for if the bandpass of each detector is precisely know in a wide frequency range (due to the steepness of some emissions).

French roadmap for CMB science 30/06/2016 4.1 Galaxy Cluster and Large-Scale Structure Science 25 importance because they reach into the peak period of galaxy formation near z = 2 where stellar mass is rapidly increasing and key processes shape the galaxy population. The quenching of star formation and the emergence of the passive elliptical galaxy population in clusters is one such key process that remains poorly understood. High redshift samples from future millimetre experiments are unique probes of this critical period in galaxy formation. The progressive upgrades of current ground based experiments8 will multiply the number of clusters detected at millimetre wavelengths by a factor of 100 between now and 20259, and the upper redshift limit will move 13 up from z = 1.5 to z 3. They should reach a limiting mass of order of M500 5 10 M but on a limited ∼ ∼ × fraction of the sky. The CORE satellite, with its 6 arcmin resolution (at 150 GHz) in at least 15 frequency 14 bands, will detect massive clusters (M500 > 10 M ) over the full sky, and produce large samples with reliable

flux estimates for statistical analyses thanks to its large frequency coverage. These samples will contain the rarest most massive clusters in the Universe around z 3. ∼ The satellite eROSITA (launch scheduled in 2017) will provide a catalogue of order 100 000 clusters detected in X-rays up to redshift 1.5. In the optical, ground-based surveys such as DES and Pan-STARRS will publish catalogues soon, while LSST will start delivering data in the early 2020’s. From space, WFIRST and Euclid will detect clusters, the latter 100,000 of them, similar to eROSITA, but in the optical/NIR with broader redshift coverage out beyond z = 2. These surveys in X-rays, and optical/NIR will be complementary to millimetre surveys, providing additional information (such as gas and weak lensing masses, and redshifts). All together, they will permit a full understanding of selection effects. The clusters are expected to contain less gas and more dust as the redshift increases. The millimetre experiments will move from the now well established SZ detection of clusters (gravitationally bound objects, dominated by their gas emission) to the detection of proto-clusters. These systems at high redshift should differ notably from their descendants at lower redshift: disturbed and with bright infrared dust emission from active star formation not yet quenched (see Fig. 7 in Planck Collaboration XXIII 2015). Planck already provides a glimpse of these systems and found proto-clusters through their infrared signature (Planck Collaboration Int. XXVII 2015). Future experiments, such as CORE, will use both sources of emission (SZ and infrared signature) to detect proto-clusters in the 2 < z < 3 redshift range. High frequency (ν > 600 GHz) channels will play an invaluable role in detecting and characterizing such objects.

Cosmology and CMB Halo Lensing Masses The well-characterized selection of clusters by the SZ effect makes the millimetre surveys particularly valuable for cosmology. Cluster counts, measurements of TCMB(z) using the SZ effect, and construction of the Hubble diagram, H(z), from SZ and X-ray observations are all profitable cosmological probes. Future SZ cluster samples will improve constraints on cosmological parameters, such as the neutrino mass scale or the effective number of relativistic degrees-of-freedom, which are central goals of the CMB-S4 experiment, together with improved constraints on spatial curvature, on the dark energy equation-of-state, and on departures from General Relativity. These goals require adequate modelling of cluster physics; for instance, preliminary studies for CORE show that a precision of better than 1% on the normalisation of the flux-mass scaling law is required to exploit the full potential of such large cluster samples. The new technique of CMB mass estimation through halo lensing has the potential to satisfy this requirement. Measurement of halo masses is also critical for cluster astrophysics and large-scale structure studies. Ground- based observatories already provide measurements through weak gravitational lensing, but require prohibitive amounts of observing time outside of dedicated surveys, limiting samples to less than 100 objects. The DES will soon greatly increase this number, followed by Euclid, WFIRST and LSST with their large sky coverage. These facilities will nonetheless be limited by the number of background source galaxies that drops rapidly beyond z = 1. The only way to extend mass measurements over a much broader redshift range is with CMB halo lensing, because the primary CMB offers a well-characterized source plane at z = 1100. The CMB-S4

8 See the experimental landscape overview in §7. Naming a few here, from the 1,000 detector scale (ACTPol, SPTPol), to the 10,000 detector scale (Advanced ACT, SPT-3G), and latter to the 100,000 detector scale (CMB- S4, cf. http://indico.cern.ch/event/506272/contributions/2138027/attachments/1274161/1889672/ CMB-S4_CERN-160517-v2.pdf). 9 https://cosmo.uchicago.edu/CMB-S4workshops/index.php/Main_Page

French roadmap for CMB science 30/06/2016 26 4 SCIENCE BEYOND THE PRIMARY CMB SCIENCE and CORE experiments will make full use of this technique to measure masses for both individual clusters and statistically on large samples.

Cluster physics High resolution, sub-arcmin ground-based experiments, like MUSTANG (9” resolution and an 8’x8’ field of view) and NIKA2 (the only dual band instrument with a 18.5” resolution at 150 GHz and a field of view of 6.5’) are complementary to CMB experiments. They will allow detailed study of the gas pressure in clusters, from shocks at large scales to turbulence using pressure fluctuations. These are a direct indication of the cluster dynamical history. In addition, the velocity measurement using the kSZ effect for the brightest objects will be complementary to the Athena measurement of bulk motion and turbulent velocities from X-ray lines. Dedicated follow-ups at other wavelengths (e.g., in X-rays) will allow degeneracies to be broken and the recovery of physical characteristics. For example, combining the high resolution thermal SZ map of a cluster with the density map from X-rays provides its temperature map, free of any X-ray spectroscopic temperature systematics. This will give access to the cluster entropy, which encompasses the thermodynamic history of the gas, and to its total mass. Olimpo will detect for the first time the Sunyaev-Zel’dovich effect at 30 independent frequencies10, and should characterize accurately the spectrum of the various intra-cluster emissions.

4.1.2. Large-Scale Structure Science

Studies of the relation between dark matter and baryons – diﬀuse ionized gas, peculiar velocities and star formation (through dust emission) – in the cosmic web are some of the pioneering results of the present generation of millimetre experiments.

All-sky thermal SZ Current SZ maps remain contaminated by residual Galactic dust at large scales and by unresolved point sources and the cosmic infrared background (CIB) at intermediate and small scales. The higher sensitivity and frequency coverage of future experiments are critical to producing cleaner thermal SZ maps. These maps are of prime importance for studying the distribution of the diﬀuse hot gas in the Universe. For example, PIXIE will be able to constrain tightly the history of gas physics in the Universe using CMB y-spectral distortions (Hill et al. 2015). This is an area where ground-space synergy is essential, the ground for its high sensitivity and angular resolution and space for the high frequencies.

Velocity fields The sensitivity and frequency coverage of future millimetre experiments will enable the unambiguous measurements of individual cluster velocities and bulk flows for large numbers of objects through the kSZ effect. This requires sufficient resolution to disentangle the kSZ effect and primary CMB anisotropies, which have the same frequency signature, as well as broad frequency coverage to separate out the thermal SZ and dust emission – another area for important synergy between ground and space.

Halo content (gas, mass, dust) from small to large scales Cross-correlation and stacking (or bin-averaging) studies were pioneered by Planck, ACT and SPT and have been a notable success. Such studies trace the cosmic web on large scales, including the ﬁlaments, while stacking probes the properties of dark matter halos hosting baryons along the web (e.g., gas in the locally brightest galaxies, Planck Collaboration Int. XI 2013). Low resolution experiments are suﬃcient for the former, but stacking requires an angular resolution of at least a few arc minutes. These new techniques promise important advances in understanding of the cosmic web with future millimetre experiments: PIXIE is well-suited for cross-correlation studies, while CORE and CMB-S4 will employ both cross-correlation and stacking methods. Once again, synergy between the ground and space is essential to this science.

10 http://planck.roma1.infn.it/olimpo/performance.html

French roadmap for CMB science 30/06/2016 4.2 Cosmic Infrared Background 27

4.2. Cosmic Infrared Background How the clumpy Universe that we see today evolved from the smoothly distributed matter that existed during the dark ages is one of the most pressing questions of modern Cosmology. In the last decade, it has become clear that dusty star-forming galaxies (DSFG) are significant participants in this transformation. Indeed, they are critical players in the assembly of stellar mass and the evolution of massive galaxies. Moreover, the formation of dusty galaxies at high-z remains a difficult problem for theorists to address. To date, no model simultaneously matches both the observed number counts of bright DSFG and their inferred physical properties, as well as the z = 0 stellar mass function in an ab initio manner. Observing dusty star-formation at high redshift requires (sub-)mm experiments. But the detection of DSFGs is very difficult as they are so faint and numerous, and, given the angular resolution achievable in the far-IR to mm, confusion plagues observations substantially. As a result, CMB experiments such as Planck, SPT or ACT can only see the brightest objects that represent the tip of the iceberg in terms of galaxy mass halos and star formation rates – indeed, only ALMA can see them all (but over very small areas). Fortunately, CMB experiments are sensitive enough to measure the cumulative IR emission from all galaxies throughout cosmic history, the cosmic IR background (CIB). The study of CIB anisotropies has two main goals:

1. Derive information on the physical processes governing star formation and galaxy evolution: CIB anisotropies are then used to probe the clustering properties of DSFGs, constrain the relationship between star formation and the dark matter distribution, measure the cosmic abundance of dust and the star formation rate density, and derive mean spectral energy distributions of galaxies. Cross-correlation with other (dark) matter tracers allows CIB anisotropy tomography.

2. Probe cosmology: CIB anisotropies are an important tracer of large-scale structures and can be used as an integrated mass tracer. The tight correlation between CMB lensing and CIB anisotropies is a promising route for de-lensing CMB B-mode observations and thus extracting the primordial signal contribution. The cross- correlating of CIB anisotropies with the CMB can also be used to derive constraints on dark energy through the integrated Sachs-Wolfe (ISW) eﬀect.

The last five years have been extremely fruitful for this area of research with significant breakthroughs in CIB anisotropy measurements occurring thanks to Planck and Herschel data (CIB maps on very large area, >2200 deg2; CIB angular power spectrum but also bi-spectrum; CIB CMB lensing measurements; Planck × Collaboration XXX 2014; Viero et al. 2013). Even if small improvements can still be obtained, the analysis of CIB anisotropies has reached a point where the limitations do not arise from the measurements themselves but from their interpretation. Indeed, CIB anisotropies are an integrated measurement over redshift, and this limits their interpretation in terms of the physics of galaxy formation and evolution. Due to severe degeneracies (e.g., between z and luminosities), different models can fit the data very well while giving notably different answers on key parameters (e.g., the mean halo mass which is most efficient at hosting star formation, star formation rate density, see Planck Collaboration XXX 2014). The limitation also arises from the fact that the SEDs of galaxies at high z are very uncertain. Adding cross-correlation measurements or even doing CIB tomography (using, for example, numerous frequency bands or external galaxy catalogues) allows for consistency checks but does not significantly improve the situation. To understand the physics of galaxy formation and evolution, CIB anisotropies will not be a competitive tool against the planned galaxy surveys. ALMA is now routinely probing the (very) high-z Universe, and NIKA2/IRAM will conduct large-area deep surveys that will complement the very deep but small-area ALMA surveys. In addition, while different techniques have been developed to probe into the confusion, it appears today very difficult to go much deeper in such analyses (e.g., Béthermin et al. 2013). In summary, CMB experiments will not play a key role in developing a deeper understanding of dusty star formation at high redshift. However, ground-based high-angular resolution and large-area CMB maps will detect numerous lensed high-z sources that can be used as a laboratory to study their gas content. Furthermore, PIXIE (see Sect. 8.1.3) could measure the CIB mean spectrum with much better accuracy than FIRAS, providing a reference for the global energy budget of dusty star formation. In addition, PIXIE could prove to be an efficient CO and CII-line intensity mapping machine, probing the gas content of galaxies and galaxy clustering on large scales up to z 7. Note that the polarization of the CIB is not a meaningful measurement for ∼

French roadmap for CMB science 30/06/2016 28 4 SCIENCE BEYOND THE PRIMARY CMB SCIENCE the community working on galaxy formation and evolution.

However, the CIB acts as a critical source of information in relation to the delensing of CMB B-mode measurements (see Sect. 2.3.4). The tight correlation observed between CIB anisotropies and CMB lensing can be used to remove the lensing contribution to the CMB B-mode angular power spectrum. Such a correlation has been used to make the first detection of the lensing B-mode signature (Hanson et al. 2013). Sherwin & Schmittfull(2015) show that CIB de-lensing can remove more than 50% of the lensing power and lower the effective noise of a B-mode measurement by a factor of 2.2. For the proposed LiteBIRD satellite, the resolution ∼ would be too low for its internal de-lensing to outperform CIB delensing. Characterising the CIB lensing × correlation and understanding how CIB anisotropies trace the dark-matter gravitational potential is thus of primary importance for all future CMB experiments targeting the B-modes, if anything offering an important cross-validation of the internal delensing of more capable experiments (see Sect.7 below) like CORE from space and S4 from the ground.

4.3. Interstellar Medium CMB experiments image dust polarization from the Galaxy with unprecedented sensitivity to diﬀuse extended emission, and synchrotron polarization at frequencies where Faraday rotation may be neglected. The data hold promises of exciting steps forward in our understanding of the magnetized interstellar medium and interstellar dust including the physics of grain alignment.

4.3.1. Statistical analysis of Galactic magnetic ﬁelds

Polarization observations provide a unique opportunity to study magneto-hydrodynamical (MHD) turbulence and dynamo action in great detail within our Galaxy. What can be learned from CMB experiments on dust and synchrotron polarization will complement advances expected from Faraday tomography measurements with lower frequency telescopes like LOFAR, eVLA, ASKAP, and SKA. The detection potential for relevant plasma processes and their characteristic scales like turbulent energy injection and dissipation would be increased considerably by the sensitivity and statistics expected from upcoming experiments. Dust and synchrotron radiation from the Galaxy provide complementary views of interstellar magnetic fields. Synchrotron radiation traces magnetic fields over the whole volume of the Galaxy, while dust polarization traces them within the disk where interstellar matter is concentrated and stars form. The statistical properties of Galactic magnetic fields are imprinted on these observables. Methods to extract this information from observational data have been developed or are under construction. Quantities highly relevant for an understanding of galactic turbulence and dynamo processes, such as the energy, helicity and tension force spectra, were shown to be encoded in synchrotron intensity, polarization and Faraday rotation measure (e.g. Waelkens et al. 2009; Junklewitz & Enßlin 2011; Lazarian & Pogosyan 2016). The statistical analysis of dust polarization is less advanced due to the lack of a sufficiently large data set prior to the Planck survey. The analysis of Planck data prompted a number of studies, which are relating dust polarization to the magnetic field structure and its interplay with the density structure of matter (e.g., Planck Collaboration Int. XX 2015). Since dust grains are mixed with interstellar gas and dust sub-mm emission is an optically thin tracer of interstellar matter, dust polarization data are best suited to investigate the physical coupling of magnetic fields with the gas dynamics and density structure. Due to our location within the Galaxy, the fluctuations at a given angular scale in observables correspond to magnetic field structures of different physical sizes. Disentangling these in order to identify physical scales such as those related to turbulent injection or dissipation will be a challenge. Highly accurate polarimetric data with an all sky view, expected for a space mission to be proposed for the M5 call of ESA, will be of greatest value for studying interstellar interstellar magnetic fields. The probing of a large range of physical scales and the leap forward in statistics from Planck (a factor of 103 in the number of measured modes) will magnify our opportunity to identify signatures of the processes involved in MHD turbulence and dynamo theory. What will be learned from these data will complement what will be learned from large observatories observing dust and synchrotron polarization (e.g., ALMA, SKA). Together these projects will have a major impact on our understanding of the role magnetic fields play in galaxy and star formation.

French roadmap for CMB science 30/06/2016 4.4 Summary 29

4.3.2. The nature of interstellar dust The combination of spectral and spatial information provided by CMB experiments will open new means to study interstellar dust: its nature, and its evolution within the interstellar medium. CMB experiments will extend to the diffuse interstellar medium the research that will be carried out on dust in star forming molecular clouds including pre-stellar COREs, proto-stellar objects and proto-planetary disks, at higher angular resolution with mm/sub-mm ground-based telescopes and interferometers, in particular ALMA and IRAM. Dust properties (size, temperature, emissivity) are found to vary from one line of sight to another within the diffuse interstellar medium and molecular clouds. These observations indicate that dust grains evolve through the interstellar medium. They can grow through the formation of refractory or ice mantles, or by coagulation into aggregates in dense and quiescent regions. They can also be destroyed by fragmentation and erosion of their mantles under more violent conditions. The composition of interstellar dust reflects the action of interstellar processes, which contribute to break and re-build grains over time scales much shorter than the time scale of injection by stellar ejecta. While there is wide consensus that interstellar dust is not star dust, the relevant interstellar processes are still poorly understood. Understanding interstellar dust evolution is a major challenge in astrophysics underlying key physical and chemical processes in interstellar space. Large dust grains (size > 10 nm) dominate the dust mass. Within the diffuse interstellar medium, the grains are cold ( 10 20 K) and emit over the frequency range observed by CMB experiments. Dipolar emission ∼ − from small, rapidly spinning, dust particles is an additional emission component, which is thought to account for the so-called anomalous microwave emission but possibly not all of it. Magnetic dipole radiation from thermal fluctuations in magnetic nano-particles may also be a significant emission (polarized) component over the frequency range relevant to CMB studies (Hoang & Lazarian 2016). To achieve the objectives of CMB experiments on the primordial B modes, it is necessary to characterize the spectral dependence of the polarized − signal from each of these dust components with high accuracy. This is a challenge but also a unique opportunity for dust studies. The spectral energy distribution of dust emission and the polarization signal will be cross-correlated with tracers of the interstellar medium (e.g. dust reddening, HI and molecular line emission) to characterize the physical processes that rule the composition and evolution of interstellar dust. The same data analysis will also allow us to study the physics of grain alignment (Andersson et al. 2015).

4.4. Summary

Planck has allowed astrophysicists to address diverse ancillary science with spectacular results. The • ancillary science accounts for a large fraction of the publications of this space mission and weights in substantially in its overall success.

For several research ﬁelds in astrophysics, CMB experiments have provided unique data that comple- • ment those obtained from other observatories. This will continue to be true for future experiments.

Much of the ancillary science – SZ from clusters and large scale structures, extragalactic sources, dust • polarization – calls for arc minute resolution, all-sky, imaging in the sub-mm. Such data will boost by very large factors the limited statistics of current studies and open new discovery space.

The previous point calls for a combination of space and ground based CMB experiments to combine • frequency and spatial coverage together with high angular resolution.

5. Instrumental aspects This section addresses the constraints that prevail on the design of integrated instruments to pursue the scientific goals outlined previously. Typically, not all science targets require the same angular resolution or sensitivity, and these impact on the design of the instrument, together with considerations on foreground contamination and systematic effects. We start by recalling the main drivers for the design of a telescope, then focus on detectors and associated electronics and the cryogenics chain. We also discuss systematic effects that are specific to polarization measurements. Spectrometers are a separate class of instruments that are addressed in a dedicated subsection.

French roadmap for CMB science 30/06/2016 30 5 INSTRUMENTAL ASPECTS

Balloon Ground Based Telescope Satellite 25 Ground Based Telescopes 150 GHz 2 weeks 100 500 100 100 1 year 100 100 20 20 5 years 60 40 9 9 10 years 40 30 6 6 280 GHz 2 weeks 100 3000 100 600 1 year 100 600 20 100 5 years 60 300 9 50 10 years 40 200 6 40

Table 1: Comparison of instrumental noise integration with time for ground, balloon, satellite and massively replicated ground telescopes at a typical CMB frequency (150 GHz) and at a minimum frequency for dust constraints (280 GHz). We here assume that the virtual focal planes of these instruments are the same and vary only the atmospheric noise and the integration time. This table takes 100 as a reference number describing sensitivity for a two week balloon experiment. It assumes that noise integrates like 1/ √t and that a balloon can have at most one flight every two years. Relative noise at 150 GHz and 280 GHz are scaled from Fig. 17(Hanany et al. 2013). We note that this does not account for some important experimental effects, such as systematics or differing sky coverage.

This section is complemented by a review of currently operating and forecast instruments in Sect.7.

5.1. Different classes of instruments From a general point of view, telescopes can be located either on the ground, under a stratospheric balloon or on board a satellite. There are many advantages to running observations from the ground. First, the observational planning is more flexible compared to a balloon or a satellite, thus releasing constraints on the impact of a potential subsystem failure. Cutting edge technology can be used on a shorter time scale than that required by satellite standards. It is possible to maintain, repair and upgrade the instrument, which also reduces constraints on design and testing before deployment. Power consumption and cryogenics are also easier to cope with on the ground than on any other platform that rely solely on solar panels and batteries and that are weight and volume limited. The combination of site topology, ground pick-up, atmosphere transparency and stability set constraints on the actual available range in elevation. This, combined with Earth rotation, sets a limit on the observable fraction of the sky. Another restriction on ground-based instruments is their ability to observe and separate foregrounds due to the atmosphere. It is a true challenge above 250 GHz where such foregrounds ∼ must be characterized. Without such data, foregrounds set a strong limit on our ability to measure the CMB, as recently illustrated by BICEP2/Keck and Planck Collaborations et al.(2015). A satellite solves all of the problems related to atmospheric absorption or contamination, therefore permitting observations at virtually any frequency. It offers conditions of incomparable stability, control of optics including far side lobes, and allows for full sky observations. The successes of past and recent satellite missions such as COBE, WMAP and Planck illustrate these facts. This comes at the expense of a more challenging system integration and its associated development, together with limitations on the size, weight and power supply, that typically limit its angular resolution and mission duration. An intermediate step between ground and space telescopes is the class of suborbital a.k.a. balloon experiments. By operating above most of the atmosphere, we gain in sensitivity, stability, and optical constraints. The observable frequency range is extended and complementary to ground observations as illustrated by recent and proposed projects (Chapman et al. 2014; Lazear et al. 2014; Niemack et al. 2015; Monfardini et al. 2016). Flights are, however, still limited to a few days, at least in France at the current time. System complexity and flight scheduling set constraints nearly as stringent as for satellites. Sect.7 gives an overview of the current and forecast balloon near/mid-term projects, in Europe and in the US. The choice between these three locations is driven by many parameters that are mutually constraining. First, we recall that for a given frequency of observation, the required aperture is inversely proportional to the target angular resolution. At 100 GHz, a resolution of less than 5 arc min requires a mirror larger than 2 m in diameter, which is close to the limit of what can be envisaged on board a satellite (at least in a medium size mission)

French roadmap for CMB science 30/06/2016 5.1 Diﬀerent classes of instruments 31

Figure 17: Photon noise from 1 to 1 000 GHz for a single mode of radiation for a 20% bandwidth in frequency, assuming no foregrounds. Noise from the atmosphere (Chile or South Pole, zenith angle of 45◦) for bolometers is presented as a dashed green line. Bolometers on a balloon are presented as a dotted blue line. The atmospheric noise shown is due to thermal emission and does not include contributions from turbulence, changes in column density, or water vapour, which can increase the noise many-fold. or a balloon experiment today. This class of instrument is therefore more naturally designed to operate from the ground. Several past and current experiments belong to this class, e.g., QUAD (Ade et al. 2008), PolarBear (The Polarbear Collaboration: P. A. R. Ade et al. 2014), ACT (van Engelen et al. 2015) and SPT (Keisler et al. 2015). Ground sites can of course also host telescopes of smaller aperture that target larger angular scales. BICEP/Keck (BICEP2 Collaboration et al. 2014) is the leading project in this category. It currently sets the tightest upper limit on primordial B-modes to date, thus showing that a promising strategy to detect primordial BB B-modes is to pursue observations of the so-called reionization bump and measure C` on multipoles larger than ` 10. The small aperture permits compact experiments and simplifies the operational complexity com- ∼ pared to larger telescopes. While resolution is not crucial, cosmic variance sets constraints on the minimal size of the sky patch of around a few percent. However, foreground contamination such as dust emission are more severe on large angular scales. Lensing induced B-modes are also a nuisance in the quest for primordial B-modes and their contribution is harder to estimate with a low resolution experiment (Hu 2002). These lensed modes themselves offer the opportunity to constrain the total mass of neutrinos. As presented in various contributions such as the CORE+ M4 proposal and Allison et al.(2015), setting limits to a few meV requires access to high-` multipoles and thus imposes an instrument resolution not larger than a few arcminutes, at the limit of what can realistically be done from space or balloon. From the sensitivity point of view, detector performances (see Sect. 5.2), atmospheric noise and integration time must all be accounted for. All things being equal, is it more favourable to put a focal plane on ground based telescope at the expense of atmospheric noise, on a balloon at the expense of flight duration or on a satellite, at the expense of project development and time scale? The question is on purpose provocative and cannot be simply answered. To guide intuition though, Table1 compares how noise integrates over time for the same instrument placed on each of these platforms, all other things kept equal. The bottom line is that for a pure CMB frequency (150 GHz), over 10 years, the continuous integration of ground based telescopes hardly competes with 5 flights of a balloon, and that replicating the same telescope 25 times (following the same

French roadmap for CMB science 30/06/2016 32 5 INSTRUMENTAL ASPECTS line of thought as the CMB stage-4 strategy currently envisaged in the US ) compares to a satellite mission. At 280 GHz, where the CMB is weaker and the atmosphere much more problematic, it takes 25 telescopes to compete with 5 balloon ﬂights.

5.2. Focal Plane Unit Three goals must be achieved for the next generation of CMB observatories, which we briefly review in this section: polarization measurements to derive Stokes parameters of the observed sky (Sect. 5.2.1), wavelength diversity to disentangle the different foreground contributions from the CMB radiation (Sect. 5.2.2), and sensitivity to reach a given threshold of intensity in a reasonable amount of time (Sect. 5.2.3). As far as focal planes are concerned, assuming a similar cost for the chosen technological platform, a good parameter to evaluate the cost of producing an array of detectors is the number of technological steps involved. This is strongly related to the number of patterning (lithography) procedures, that is of the order of 1-3 for non- thermal detectors (e.g. KID) and 6-10 for TES and bolometers. We stress that different technological platforms charge very different amounts per technological step. The cost of tests, on the other hand, can be considered as similar for all detector technologies. In the case of “light” technologies, this cost is usually dominant compared to pure fabrication. Concerning the constraints set on the detector chain (including optics), sensitivity, frequencies, number of detectors, multiplexing, time constants, current demonstrated sensitivities on various detector and electronics readout chains, we refer to the report issued by CNES’ working group on mm and sub-mm detectors. The same document gives forecasts for short and mid-term developments, energy budget, manufac- turing cycles and Technological Readiness Level (TRL). The corresponding presentation made to the committee can be found on our public wiki http://prospective.planck.fr/uploads/Main/Y1-151127% 20groupe%20CMB%20Conclusion%20Feuille%20route%20d%E9tection%20mm-submmx.pdf

5.2.1. Measuring Polarization

For incoherent detectors, such as bolometers and pair-breaking devices, modulation of the plane of polarization, or differential power measurements, are the two ways to separate polarized from unpolarized radiation. In the first method, the polarization plane is rotated, either at the end of scans or continuously by a transmission waveplate or a tunable reflecting polarimeter. One direction of polarization, selected by a grid analyser, can be measured at the time, with a loss of half of the incoming radiation, and the other component obtained after the waveplate rotation. In order to recover the totality of the photons and measure both components simultaneously, either polarization-sensitive detectors or 45 degree polarizers might be used. The advantage of the first solution is the compactness, while the second is probably less affected by systematic effects (e.g., beam dependence upon the polarization component, polarization component purity). Nevertheless, a field rotation of 45 is needed ± ◦ to retrieve the polarization Stokes parameters. It can be obtained by the incremental field rotation of the observatory along the orbit, or by the waveplates described above. Today, solutions have been found to produce the three Stokes parameters I, Q and U in a single observation, via suitable pixel arrangement (if systematics are stable in time). If angular redundancy does not allow independent maps of Q and U per detector to be made, but imposes the use of a combination of multiple detectors, this sets stringent constraints on the detector cross-calibration in order not to leak total intensity into polarization. In the opposite case, complexity is shifted to the modulation module (e.g., a rotating Half Wave Plate, see Sect. 5.3) and a constraint is put on the stability of the detector on the timescale of the angular modulation.

5.2.2. Wavelength coverage

Wavelength diversity can be achieved through filters selecting spectral bands for different detector groups. This technology is now well mastered in Europe for mm and sub-mm wavelengths. A large number of bands is critical for discriminating between foreground and background emissions. The simplest way to achieve this is to populate the telescope focal plane with groups of detectors associated with various bandpass filters. All of the detector groups must then cover the full surface of the sky as a consequence of the spatial scan strategy. The drawback of this solution is the poor instantaneous spectral efficiency per square cm in the focal plane and

French roadmap for CMB science 30/06/2016 5.2 Focal Plane Unit 33 the possible optical cross-talk due to reflected light in the instrument. Many developments around multicolour pixels potentially allow an increase in the focal plane overall efficiency (see the detectors roadmap document). The other alternative is spectroscopy. Spectroscopy provides almost continuous spectral coverage of the same field of view. The spectrum can be obtained by high finesse (multiple interferences) methods - gratings or Pérot-Fabry systems - or low finesse devices, as Fourier Transform spectrometers (FTS). The most efficient are dispersing systems such as gratings or grisms at the expense of instrumental complexity. The FTS presents many advantages, but also some technical issues such as the need to stare at the same field of view during optical path difference scanning. Amongst the diverse FTS configurations, the Martin-Puplett solution is “by nature” polarization sensitive. An analysis of this peculiar configuration is proposed in Sect. 5.4.

5.2.3. Sensitivity and optical coupling

Considerable progress has been made in recent years in detection technology and cryogenics to produce photon limited individual detectors cooled to a few tenths of a milli-Kelvin, even under very low background instruments (i.e., cryogenically cooled to few Kelvins). At this level, the only gain in sensitivity can be obtained by the combination of a large number of detectors, covering a significant part of the optical focal plane. However, “tensions” then arise in terms of various budgets (volume, mass, power, etc.), whose limits depend strongly on the final locations of the instruments: ground based, balloon borne or space observatory. The individual assembly of single detectors is no longer a credible solution as it was during Planck preparation. Today, collective array manufacture is mandatory. This raises the individual light concentrator issue. The classical solution is the adoption of the Winston horn to couple the telescope optics to the individual pixels. The main advantage of this solution is the far side-lobe control afforded by a very basic optical design. Optically, this structure is ideal for already identified point sources. Conversely, for diffuse and large structures like the CMB, 60% of the photons are rejected due ∼ to poor contiguous coverage in the focal plane. Even if dithering or scanning retrieves the full image of the sky, the overall efficiency is strongly affected. In the case of a space application with thousands of horns, the qualification of such a huge focal plane can appear problematic. Today, it has become clear that bare arrays of detectors can reach the same stray light rejection as horn coupled pixels. A cold stop is however required in an elaborate optical design. Bare arrays have, in principle, a higher geometrical efficiency and a much lighter suspended mass. A mass production strategy with high yield is credible to allow deployment within the next decade. The multiplexing issue is a consequence of having thousands of pixels per focal plane. It saves power dissipation and wire conduction at the coldest stage but needs cryogenic electronics. Depending on the technology, multiplexing factors of 32:1 were achieved for MIS bolometers, 128:1 for TES bolometers and 1000:1 for ∼ KIDs, in French laboratories.

5.2.4. External sources of perturbations, cosmic rays

Sensitivity also has to deal with external sources of perturbations. Ionizing particles affect all detector types. The cross section of the detectors sensitive area must be reduced and the phonon link to massive areas must be reduced as much as possible. In some cases magnetic deflectors can be used to reduce low energy electrons. The detector recovery time, or time constant, is another important parameter, determining the data loss rate due to cosmic-induced glitches. The superconducting detectors and in some cases their readout (SQUIDs) are sensitive to varying magnetic fields (e.g. electric actuators, drives) and require careful shielding. Long-term thermal stability is also an issue for the 1/ f part of the noise spectrum. In the case of bolometers, the sensitivity level that must be reached imposes a very strong stability requirement on the focal plane temperature variations between calibrator measurements. Corrections can be made off-line if the individual behaviour of the detector with temperature is recorded during calibration phases. KIDS are less prone to these effects as they are not thermally coupled to a reference plate and have sub-millisecond time constants as well.

French roadmap for CMB science 30/06/2016 34 5 INSTRUMENTAL ASPECTS

Figure 18: Angular power spectra of the CMB in temperature and polarization. The TE cross-correlation is not shown for readability. The foreground contribution and its associated uncertainties are taken from the CORE+ M4 proposal and the uncertainty area varies with the observed fraction of the sky (upper limit for 75%, lower limit for 25%). The 5µK.arcmin noise is displayed for reference assuming three typical angular resolutions. The primordial B-modes are in solid red for two possible values of r and the lensing induced B-modes are in red dash line.

5.2.5. Instrumental leakage

Finally, the instrument itself can induce some permanent effects on the polarization measurements. As a result, each reflection in the optical path affects the polarization and must be compensated. Table2 summarizes the leakage requirements that impose constraints on the design.

5.3. Systematic effects

As illustrated in Fig. 18, there are several orders of magnitude between CMB temperature anisotropies, polarized foregrounds (Galactic dust and synchrotron) and primordial B-mode anisotropies. This sets the overall scale for controlling systematic effects and cleaning foregrounds. Roughly speaking, temperature to polarization leakage residuals are induced e.g., by differential gain mismatch between detectors. Another example is that of a residual absolute rotation of the maps that produces E to B leakage. Systematic effects couple to the scanning strategy of the instrument and the inhomogeneity of the sky coverage. This has been studied via simulations in various contexts (Hu et al. 2003; O’Dea et al. 2007; Shimon et al. 2008; Bock et al. 2009; Rosset et al. 2010). Most “differential” effects between detectors can be mitigated by the design of the scanning strategy when it provides sufficient angular redundancy to produce independent maps per detector. Beam asymmetry is an exception to this because it couples to the temperature anisotropy as a spin-2 quantity and is thus “synchronous” with polarization when the instrument rotates (except if modulation is obtained with a half wave plate, see below). It is however possible to correct for this effect during the data analysis as illustrated by BICEP2 Collaboration et al.(2014). This then requires percent accuracy on the knowledge of beam widths and orientation (CORE+, M4 proposal). Bandpass mismatch couples to foregrounds and emission lines in a non trivial way as observed in Planck (Planck Collaboration VIII 2015) and must be controlled as well to great precision, typically 0.1% (CORE+, M4 proposal). To summarize and set the scale, Table2 gives limits on some generic e ffects depending on the target sensitivity. To translate these numbers into instrument specifications requires a precise description of the instrument

French roadmap for CMB science 30/06/2016 5.3 Systematic eﬀects 35

` = 3 ` = 80 2 3 2 3 r = 10− r = 10− r = 10− r = 10− I to Q, U 1.2 10 4 3.7 10 5 1.8 10 4 5.8 10 5 × − × − × − × − Q, U mixing (deg) 0.575 0.182 0.356 0.113 Foregrounds residuals (200GHz) 0.003 0.001 0.022 0.007 Lensing residuals - 0.663 0.201 0.063

BB Table 2: Speciﬁcations to maintain generic spurious components to below a tenth of the primordial C` at ` = 3 and ` = 80 if r = 0.01 or r = 0.001. Except the residual mixing angles in degrees, other numbers are dimensionless and can be seen as multiplicative factors of the nuisance component. The white noise limit is for a 1 arc min FHWM instrument.

[Jy/beam] [Jy/beam] [Jy/beam] 150 GHz Stokes U map 0.10 150 GHz Stokes U map corrected 0.10 150 GHz Stokes Q map 0.25 30″ 40 30″ 40

100

04′ 00″ 04′ 00″ 0.04 0.04 0.14 20 20 50 30″ 30″

Dec. -0.03 Dec. -0.03 0 0.03 0 0 03′ 00″ 03′ 00″

-50 -0.09 -0.09 -0.08 30″ -20 30″ -20

-100 02° 02′ 00″ 02° 02′ 00″

-0.20 -20 0 R. A. 20 -0.15 -20 0 R. A. 20 -0.15 h m s s s s s h m s s s s s -100 -50 0 50 100 12 29 10 08 06 04 02 12 29 10 08 06 04 02 R. A. R. A. arcsec Center: R.A. 12 29 06.56 Dec +02 03 12.5 Center: R.A. 12 29 06.56 Dec +02 03 12.5

Figure 19: Left: Q map of Uranus as observed by NIKA at 2 mm before any further correction. Uranus is unpolarized and can be considered as a point source compared to NIKA’s FWHM. Rather than zero, this map shows a significant signal due to instrumental polarization, form of which is not a simple scaling of the Gaussian total intensity point spread function. Middle: This effect also shows on the U map of the know polarized quasar 3C273. Right: U map of 3C273 after correction of the leakage effect. For more details, see Ritacco(2016). and appropriate simulations (as, for example in Bock et al. 2009; Rosset et al. 2010), because a significant part of the systematics mitigation comes from the data analysis approach and the choice of the observation modes. Instrumental polarization is the generic term for effects that induce a polarized signal when observing an unpolarized source. The simplest example is a reflection on a mirror that will produce Q and U signal directly proportional to the incoming intensity, with an angle fixed in the instrument reference frame. This effect can be mitigated by a rotation of the instrument with respect to the sky. If characterized correctly, the residual can be removed from the polarization maps by the subtraction of a scaled template of the total intensity (with uncertainties on the model and the total intensity template accuracy as limiting factors). Other types of instrumental polarization have been observed, leading to an effective differential pointing (Ade et al. 2008; BICEP2 Collaboration et al. 2014). More recently, NIKA observed a similar effect (Fig. 19). Although its exact cause is still being investigated, the NIKA team has been able to correct this effect at the data processing level. During its flight in September 2015, PILOT has observed opposite variations of the signal in its two focal planes when the HWP is rotated. After comparison between flight and laboratory data, taken under various background conditions, this parasitic signal is found to be consistent with an unpolarized instrumental background signal being polarized through propagation in the optics (private communication). This effect translates into offsets in the time lines when the HWP changes positions, but they are removed during the data analysis along with other classic low frequency drifts. This type of non-trivial instrumental polarization is a strong driver for developments in numerical simulations of the optics of integrated instruments and we thus recommend that a significant effort be put on whole system modelling and the development of efficient simulation and analysis tools. These tools must be available when concepts have to be compared and selected for a future satellite mission. HWP are now commonly used to modulate the incoming polarization. They allow the sky to be rotated w.r.t the instrument at any angle and therefore offer incomparable angular redundancy and mitigation of the aforementioned systematic effects. The HWP rotation can be constant during a sequence of observation, then step rotated (Fissel et al. 2010; Rahlin et al. 2014; The Polarbear Collaboration: P. A. R. Ade et al. 2014;

French roadmap for CMB science 30/06/2016 36 5 INSTRUMENTAL ASPECTS

Misawa et al. 2014) or continuously rotated (Chapman et al. 2014; Ritacco et al. 2016). While the former offers more simplicity in terms of control and cryogenics, when the HWP is rotated at about 1 Hz, the latter has the unique advantage of placing the incoming polarized signal in the white noise part of the instrument noise and thus naturally rejects atmosphere and electronic 1/ f contamination. This is a crucial point for ground-based observations. It also ensures optimal angular coverage, allowing the construction of independent maps per detector, thus averaging down any differential effects such as gain and bandpass mismatches between detectors. On the other hand, the HWP and its rotation mechanism usually introduce a parasitic signal that is synchronous to the rotation and that is much larger than the signal. This parasitic signal can however be removed very efficiently as demonstrated in Chapman et al.(2014) and Ritacco et al.(2016).

5.4. Spectrometers The CMB spectrum is described by black-body radiation at a temperature of 2.726 K, which implies a maximum intensity at a frequency of 160 GHz. It extends from 20 to 500 GHz (15 mm to 600 µm) at a significant level of intensity. In this spectral domain, heterodyne receivers are commonly used for spectroscopy, allowing high-resolution studies on relatively narrow bands. Spectroscopy of the CMB continuum requires only low spectral resolution, but on a broad domain. The only workable broadband spectrometer for this spectral domain is the Fourier transform spectrometer. From its multiplex properties a single bolometer makes it possible to cover the full CMB spectrum. A millimetre FTS is based on the Martin-Puplett interferometer (Martin & Puplett 1970). It takes this name from the two authors of the prototype who developed wire-grid polarisers as beam splitters, to introduce in a dual-input, dual-output interferometer. These beam splitters made of parallel, equally-spaced wires (5 to 10 µm diameter) on a stainless steel frame have the advantage of being achromatic over a large range, with efficiency very close to 50/50. The two output beams are obtained by roof-top mirrors replacing the flat mirrors of the standard Michelson interferometer. Thus, all the entering energy is recovered by the two output detectors. The two interferometric signals are in opposite phase. With two output beams, the interferometer has also two input ports. For the spectroscopic analysis of an astronomical object placed on one entrance, the second input port is generally matched to the sky background next to the source which will be automatically subtracted from the main signal. In the particular case of the CMB, a black-body emitter adjusted to a temperature fitting the CMB emission must fill the second input port. That was the successful design of the FIRAS spectrometer, mounted on the COBE satellite, which made the first full measurement of the CMB spectrum and showed it to be consistent with pure black-body radiation. A new spectrometer for the CMB with the goal of detecting the polarized component of the emission could be built on this legacy. In addition, the full advantage of the FTS can be provided by extending with the same bolometer the spectral coverage towards high frequencies, beyond the CMB spectrum. Access will be provided to the pure spectra of the foregrounds, which may be an important tool to separate them from the cosmological emission. This approach has been chosen by the American space proposal PIXIE (Primordial Inflation Explorer), which represents a unique way to analyse the CMB emission. With the advent of detector arrays at all wavelengths, Imaging FTS have been developed (Maillard et al. 1993) by the coupling of a FTS with bi-dimensional detectors, providing an integral field spectrometer, able to cover a wide field in a single exposure, particularly at low spectral resolution. PIXIE, as FIRAS before it, is a mono-pixel instrument. An imaging version with bolometer arrays could improve the sensitivity by a factor related to the size of the array. To note, a millimetre - submillimeter imaging FTS is in operation on JCMT (Naylor et al. 2006), demonstrating that such an FTS can work in this spectral domain, in addition to the success of SPIre on board Herschel. However, for future CMB studies, to benefit fully from all of the advantages of this spectrometer, in imaging or single pixel mode, a spatial instrument is mandatory. It is only in space that each detector, working in multi-mode conditions, receives the complete spectral domain of the emission. With a high sensitivity, this interferometric spectrometer represents the only instrument that can make possible the mapping of the spectral distortions of the CMB spectrum.

French roadmap for CMB science 30/06/2016 5.5 Summary 37

5.5. Summary

High angular resolution requires a large aperture which is limited to about 3 m in space (Herschel) and • is thus more suited for ground based observations. We note though, that a 6 m dish with deployable technology is expected for JWST.

Roughly speaking, in the race to primordial gravity wave detection and to characterizing the pair (r, n ), • T multiplying ground telescopes “à la S4” is a strategy that improves sensitivity with time at a comparable rate to the development of a satellite over typically 10 years. At higher frequencies, most important for foreground mitigation, ground based experiments are more limited and balloons show similar merit (recall our ﬁducial example of 25 telescopes) while satellites remain out in front.

As far as CMB continuum spectroscopy is concerned, the only workable broadband spectrometer for • this spectral domain is the Fourier transform spectrometer (in space).

In terms of focal plane units, we are now in the era of ﬁlled arrays. This posed important challenges in • detector development and multiplexing that have now been met. Various technologies (TES, bolometers and KIDS) are now available in French laboratories. Among these three, KIDS emerge as simpler, cheaper and faster to manufacture, less prone to temperature instability induced by cosmic rays and with demonstrated capabilities on the sky with NIKA2.

Systematic effects will have to be controlled to an even greater level than before. New instruments and • their associated sensitivity have shown systematic effects such as instrumental background polarization that had not been considered at the time of their design nor in past proposals of polarization satellites. The group recommends that a significant effort be put on whole system modelling and the development of efficient simulations and analysis tools. These tools must be available when concepts have to be compared and selected for a future satellite mission.

6. Data processing and analysis aspects CMB experiments pose substantial (and exciting) data processing and analysis challenges. While none of these are show stoppers in themselves, preparing to address them is essential, if anything because this can, and should, influence the design of the experiments in order to to maximize the scientific potential of the experiment and its data. In fact, the quality of the CMB data analysis is of comparable importance to the data acquisition hardware and associated techniques in determining the final outcome of the experiment. The data processing strategy cannot be left as an afterthought. We start by describing 3 classes of data processing challenges arising from the scientific requirements of future CMB experiments. We then sketch a prototypal data analysis pipeline and describe how these issues manifest themselves in such an environment. Finally, we comment on ways to overcome these difficulties.

6.1. Types of CMB challenges Future CMB data analysis challenges may be grouped into 3 diﬀerent areas: - computational complexity/data volume, - non-linearity/non-Gaussianity/non-stationarity, - modelling uncertainties/limitations, each of which we describe in turn.

6.1.1. Computational complexity and data size The scientific goals of future CMB experiments require a vastly reduced noise level in the datasets compared to contemporary experiments. This will be mainly reached by significantly increasing the number of detectors simultaneously observing the sky, going from the few hundred used in typical instruments today to a few hundred thousand for ground based stage-4 experiments (see Sect. 7.2 below), or a few thousand in the case of satellites. Such a dramatic increase will create practical difficulties for data analysis.

French roadmap for CMB science 30/06/2016 38 6 DATA PROCESSING AND ANALYSIS ASPECTS

If the complexity of this task scaled linearly with the number of detectors (N), then a thousandfold increase might be passively absorbed by the availability of more capable computers (assuming that Moore’s law continues to hold over the next 15 to 20 years). However, some of the data analysis steps are non-local, requiring the study of correlations between data points measured over different time scales. Indeed, the inter-calibration of the detectors will naïvely scale as the square of their number – new methods or models will have to be implemented to avoid the full million-fold increase in complexity, depending on the level of precision of inter- calibration between each detector that is required. Not to mention the higher demand for ground experiments bound to deploy much larger number of detectors to cope with the effect of the atmosphere. In addition, even with a linear scaling, and taking full advantage of new computer hardware, the massive data volume will translate into increased complexity in both its management and subsequent processing. To give an example, the state-of-the-art data processing and simulations for the Planck 2015 analysis required of the order of a peta-byte of storage, whereas the primary data volume was only a few tera-bytes (Planck Collaboration et al. 2015e). It may be difficult to simply scale these numbers by a factor of 1000, yet failing to store data generated by intermediate processing steps and only retaining the final results and products will inevitably pose challenges of their own. This problem will be less severe for space based experiments, with at most a mere hundred-fold increase in primary data volume. However, the level of accuracy to which each detector will need to be calibrated, and inter-calibrated with the others, leads us to expect that the scaling in complexity will probably be closer to N2 than to a linear one.

6.1.2. Non-linearity, Non-Gaussianity, Non-stationary

One reason that the power of the CMB to constrain cosmological and fundamental physics beyond the standard model of particle physics remains unmatched is that it deals with tiny fluctuations whose evolution is linear, or at least can be dealt with perturbatively. The CMB sky itself is very well described by an isotropic Gaussian field emerging from a linear transformation of the primordial perturbations. The scientific goals of future CMB experiments mandate going beyond the linear approximation. Measuring the neutrino masses or searching for primordial B-modes requires correction for the lensing effect, which induces non-Gaussian (or non-stationary) features in the observed CMB (Benoit-Lévy et al. 2012). Lensing is poorly represented by a linear approximation on small scales (Lewis & Challinor 2006). Similarly, most of the modelling of CMB datasets, to date, has relied on linear descriptions of the data. Similarly, the CMB foregrounds have been treated with relatively crude models, ignoring non-linear effects and assuming in most cases a Gaussian statistical representation. For Planck, this approach was sufficient to handle foreground contamination both for the application of blind component separation methods (Planck Collaboration et al. 2015a) or in the construction of a cosmological likelihood (Planck Collaboration et al. 2015g). However, the goals of next generation projects will require that the non-isotropic nature of the Galactic emission is accounted for. In the case of instrument modelling, linear approximations are already insufficient for the processing of the Planck data (Planck Collaboration et al. 2015b; Planck Collaboration VIII 2015) which would have severely limited the measurement of large scale polarization (Planck Collaboration et al. 2016b). It remains unclear whether the dramatic increase in detector number will reduce the magnitude of the problem (by averaging non-linearities) or exacerbate it. Beyond the increased complexity of the data analysis and statistical characterization, the scientific goals of future experiments will also prohibit the use of various null tests. For example, going beyond the standard model and testing models that can create TB or EB correlation prevents the assumption that these correlations are null when testing the efficiency of the data processing and the existence of residual systematic or foreground contamination.

6.1.3. Modelling uncertainty

Another strength of the CMB as a physics probe is the extent to which its theoretical modelling, as well as that of related foregrounds and instrumental behaviour, is well established, and consistent with the level of precision

French roadmap for CMB science 30/06/2016 6.2 Data analysis. From time streams to model constraints 39 of current observations. In order to maintain this situation in the near future, improvements in specific areas will be required. We provide a few examples below. Fig. 10 showed to what extent the constraint on neutrino masses is affected by modelling uncertainties, that then need to be marginalized over or accounted for. On the foreground side, Sect. 3.1.1 describes the current uncertainty in Galactic foreground modelling. Refinements of the description of the various foreground components, specifically their detailed frequency scaling and sky morphologies, are needed to allow the separation of foreground induced polarization from the cosmological signal. On smaller scales, improvements in our understanding of possible extragalactic polarized contributions could be needed. Similarly, ground-based experiments need to model the atmospheric foreground effect in fine detail, if only to marginalize over the extra uncertainty they add to such measurements. Finally, to reach the desired level of sensitivity, instrument models, including prescriptions for their beams, detector cross-talk and leakage between signal components, will have to be included in the analysis. This is a particular area where the increased number of detectors could result in significant complication of data processing. Indeed, individual detectors on common wafers can suffer from important cross-talk, that will have to be taken into account. In this case, instead of being averaged over by the multiplicity of detectors, the correlation between individual channels will impose significant limitations on the modelling of systematics. In summary, modelling uncertainties will be important issues for future CMB experiments. Solving them will require improved modelling approaches. In the case of foregrounds and even more so for instrumental ones the situation can be improved by dedicated studies.

6.2. Data analysis. From time streams to model constraints Very schematically, one can describe the data processing of CMB experiments from detector time streams to cosmological constraints in three steps: map making, foreground mitigation/removal, and the statistical characterization of the data. To these three steps, one must add an extra (parallel) one, which is required for the overall assessment of the accuracy and precision of the analysis: simulations. Note that we have included detector calibration and beam determination in the map-making step.

6.2.1. From Time streams to maps The first step consists of projecting the raw time-ordered data acquired by the detectors into sky maps for known frequency bands. Instrument systematics and noise are generally more easily characterised in the time ordered data streams than once projected on the sky. These are much larger in size than the sky maps, allowing sufficient redundancy for noise and systematic artefact reduction, and (in some detector configurations) a reliable estimate of the polarization information obtained from numerous observations of the same sky coordinate with different orientations. The data from the detectors can be described by the equation

d = A[s] + n, where the data vector d and noise and systematics n live in the time domain, while the signal s is pixelised on the sky. Both are linked through an observation transfer function A which accounts for the response of the detector chain (including electronic and optical effects, and atmospheric effects for ground based experiments) to the signal emerging from the sky (which includes CMB as well as all other components). Since A links the data and sky spaces, it also contains all the information on pointing and scanning strategy. Given that the time streams are much larger in size than a sky map, telemetry constraints will typically prevent the download of all of the data acquired aboard both balloons and satellites. Some on-board data reduction is then required, since lossless data compression can be insufficient. Some pre-pre-processing could be performed on-board, but this could translate into increased data uncertainty, as it assumes an early data model that would otherwise be improved post-factum. Initially, a good mathematical representation of the observation function A has to be constructed, and the noise properties n estimated. This is done by establishing a model for the observation, and calibrating its parameters on tests performed first on the ground, then while data acquisition is under way, and finally and most accurately from the data and housekeeping information directly. Currently, a simple linear model has proved sufficient for A, since the model allows the separation of relatively local (in the data stream) and non-local parts

French roadmap for CMB science 30/06/2016 40 6 DATA PROCESSING AND ANALYSIS ASPECTS of the observation matrix, resulting in various simplifications. However, Planck has now reached a precision where the linear approximation is no longer sufficient to describe the instrument response to the required level of accuracy. This situation will be typical in the future. If the detectors are uncorrelated, it is easy to decouple the large observation matrix into smaller, more manageable parts. The proliferation of correlated detectors prevents the use of this simplification, and therefore will require either new approximations to partially decorrelate the system, the adoption of a slow brute force approach, or, preferably the design and implementation of new algorithms. Simulations are essential to test whether the instrumental model and its parameters are a faithful representation of the data. We will return to this issue in the following sections, but note already that the simulation program will be fundamentally limited by the number of detectors that need to be represented at the highest fidelity. Before projecting the data on the sky, it is necessary to select data that is not affected by glitches, e.g. from cosmic ray hits, and to correct for biases (due to glitch residuals, thermal drifts. . . ). It is also more efficient to deal with the detector temporal response and to filter out high frequency noise as well as, in the case of ground based experiments, atmospheric fluctuations. This step often involves the simultaneous analysis of multiple detectors, as some of the effects under consideration can be correlated among different time streams (glitches, atmospheric fluctuations). The optical response however, since it couples different pointings on the map, is more easily handled at the map level and is not corrected at this step. Effective beams on the map can contain residuals from partially corrected temporal responses. For this reason, the calibration of effective beams is an important step of the process, which requires dedicated observation of known point sources, and must be carefully validated with simulations. The next step is to invert the data equation and obtain the “best” signal estimate. In the case of a linear model, an unbiased estimate can be obtained by using t 1 t m = (A WA)− A Wd, where W is a positive definite matrix. A significant part of this weighting matrix accounts for the filtering of the data stream. The minimal variance solution is obtained when W is chosen to be the inverse noise covariance matrix. This choice is already impossible to implement for Planck, except for a few test cases. A good alternative is to adopt a destriping algorithm, which makes use of the fact that the pointing revisits regularly the same part of the sky, to account for 1/ f noise and perform a dimensional reduction before inverting the matrices (Kurki-Suonio et al. 2009; Planck Collaboration VIII 2015). Ground based experiments use similar approximations, but rely on different representations of the noise (typically using polynomial templates) and use diagonal weighting matrices based on simple noise modelling (see, for example, Chiang et al. 2010). Beam deconvolution can be performed at this step. It is not attempted usually as it is computationally very expensive already on current state of the art data. It is often more efficient to characterise the effective beam and to deal with it in the next steps of the analysis. For future CMB survey, we will need extensions to the destriping algorithm (or its equivalent for ground based data) that can further decrease the dimension of the problem, for example, by co-adding different detectors before the final map making. Otherwise, the computation complexity will increase as the square of the number of detectors. These algorithms will also need to handle the inter-calibration between detectors (including beam shape variations), making the problem non-linear. Note that large focal plane experiments or the co-analysis of data from different experiments will encounter similar problems, the latter having the added difficulty that the observation matrix A and associated instrument model and parameters will vary between the experimental configurations. A similar problem was encountered for Planck, where the non-linear response of the instruments were described at first order as a variable calibration (Planck Collaboration VIII 2015). The solution to this kind of non-linear problem is to iterate between improvements to the parameters describing a linearised approximation for A and a map making step. Null test are crucial at this step to assess the efficiency of the model and guide improvements therein. This is also useful when dealing with large scale systematics which are more easily measured at the map level. An example of such large scale effects is the dipole or foreground leakage variations between detectors with different bandpasses. We see here how uncertainty in the Galactic contamination model and the anisotropic properties of the Galactic emission can affect even the determination of frequency maps. Future CMB data analysis will need to quickly iterate on the instrument calibration/map making loop and more research on the acceleration of the convergence of this kind of algorithm will be needed (see, for example, Planck Collaboration

French roadmap for CMB science 30/06/2016 6.2 Data analysis. From time streams to model constraints 41 et al. 2015e). This is compounded by the fact that the CMB emission will need to include corrections to a pure black body description (e.g. Rayleigh scattering, or other inescapable spectral distortions).

6.2.2. Foreground mitigation/removal Depending on the intended scientiﬁc target of the CMB map analysis, diﬀerent approaches to foreground mitigation may be adopted. For example, the production of lensing maps requires explicit removal of foregrounds, whereas the derivation of cosmological parameter constraints via the power spectrum utilizes foreground emission templates (and associated covariances). Foreground descriptions and cleaning techniques are described in Sect. 3.1 and Sect. 3.2. The central challenge for the future is to improve the description of the Galactic emission. Further improvements are expected from the correct determination of the possibly non-Gaussian structure of the foregrounds covariance. The non-Gaussian distribution of point sources constitutes a bias in the Planck derivation of the constraints on primordial non-Gaussianity or on the measurement of CMB lensing.

6.2.3. From cleaner maps to cosmological parameters The last step in the path from raw data to cosmological constraints is to build the likelihood of the data given a model, allowing the estimation of posteriors on different (cosmological) parameters. Here we will emphasize the situation for two of the most important goals of future CMB experiments, primordial B-modes and neutrino masses. Both studies require an initial delensing of the data. In the case of B-modes, lensing induced B-mode polarization conceals the gravitational wave signature (see, for example, Lewis & Challinor 2006). In the case of estimates of the neutrino masses, delensing minimizes the impact of non-Gaussian, lensing-induced covariance (as in Audren et al. 2013b) and sharpen the acoustic peaks improving the constraint of some parameters. Delensing is discussed in more details Sect. 2.3.4. CMB lensing effects are measured by averaging 4 point functions under a specially designed weighting function in order to select the non-Gaussian lensing induced signal (Lewis & Challinor 2006). The measurement is limited, in part, by the noise level and instrument resolution, together with any non-Gaussian features present in the maps due to either foreground residuals or instrumental artefacts. These translate into biases on the power spectrum that have to be removed. In practice, a lensing estimate relies heavily on massive simulations to determine such biases. The complex shape of masks and the anisotropic structure of the map noise requires the use of a non-diagonal weighting function which is very costly to apply (typically requiring the application of a Wiener filter to the map). This processing has to be applied to each simulation. At Planck precision, of order 1000 simulations were required to estimate and de-bias the lensing power spectrum and its covariance (Planck Collaboration et al. 2015f). For a hundredfold increase in detector numbers or more, some averaging of the maps from a given frequency band and the subsequent propagation of systematics from these detectors to the average map will have to be obtained. The critical issue is then how map-level uncertainties propagate into the overall lensing error budget. Indeed, the process of delensing has never been applied to data. While the theory is well-known (see, for example, Smith et al. 2012), there is no direct experience yet on how the different non-idealities of the data translate into actual limitations of the delensing. CMB temperature and polarization are correlated through lensing. Until recently, it was reasonable to ignore the effect in the covariance matrices, but this will not be the case for future CMB experiments. The correlation is non-Gaussian and its estimate, either by analytical derivation or Monte-Carlo integration, is expensive (Benoit- Lévy et al. 2012). This is the reason why delensing is considered not only for primordial B-mode measurements. Workable estimates of the covariance of the CMB and CMB lensing after delensing are needed (Namikawa & Nagata 2015). Cosmological parameter estimation methodologies will also require improvements for the application to future CMB data. Current likelihood approximations targeted at r are crude, but not unreasonable given the current sensitivity of the measurements (BICEP2/Keck and Planck Collaborations et al. 2015). An improved approach will probably need to describe the statistics of a very large number of cross-detector spectra, taking into account non-Gaussian features on large scales, and both foreground and lensing residual contributions. Given such complexities, analytic forms are very difficult to derive, therefore an alternate possibility is to base the likelihood estimates on simulations. New statistical methods, such as ABC (Akeret et al. 2015; Ishida et al.

French roadmap for CMB science 30/06/2016 42 6 DATA PROCESSING AND ANALYSIS ASPECTS

2015), will need to be implemented so as to minimize the number of (costly) simulations required to obtain a faithful likelihood approximation. This approach can also be applied to CMB lensing. This, and similar methods should be investigated further and implemented in cosmological toolboxes such as COSMOMC (Lewis & Bridle 2002), or Monte Python (Audren et al. 2013a). Regardless of the way in which the likelihood evaluations are obtained, the methodological validation of parameter constraints will still require simulations. At present, Planck results uses O(100) for parameter studies and (in some cases) O(10000) for covariance validation (Planck Collaboration et al. 2015g). Simpliﬁed, quicker, simulations at the required level of ﬁdelity need to be developed for this particular task, both to reach better accuracy and to deal with the multiplication of detectors.

6.2.4. Simulations Simulations play a crucial roles in the data processing pipeline, to evaluate the faithfulness of the model, the domain of validity of any approximations, or to propagate uncertainties, and they are needed at every step of the analysis. At the science exploitation stage, they are mandatory, e.g., for lensing estimation. The main challenge for future simulation work arises from the anticipated, dramatic increase in the number of detectors. Simulating fully any instrument, at the level of precision achieved by Planck, and with a comparable number of realisations will barely be possible (Planck Collaboration et al. 2015e, 2016b, 2015g). At the same time, the simulations will need to improve the accuracy of models for the foregrounds and secondary effects. Some of the science cases possible with future data sets will call for the correlation between secondary effects (SZ, lensing) and other probes of large-scale structures. In order to correctly incorporate such effects, dedicated N-Body simulations may be required. Such requirements have already been acknowledged by e.g., the Euclid consortium, which has proposed an ambitious program of coherent simulations, from which the CMB community could also benefit. Thus to cope with future CMB data sets in all their size and complexity, methodological improvements will need to be developed specifically for simulations. Similarly, in the data processing environment continuous improvements of codes and their implementation will be mandatory. Beyond that, ways to build short-cuts into the modelling, estimate residual templates, or generate quick emulators of the data will be needed (see, for example, Kwan et al.(2013) for an application of such methods to non-linear P(k) emulation). Novel mathematical methods should be investigated, such as machine learning techniques that allow interpolation in complicated, large dimension, non-linear spaces. In the context of astrophysics Graff et al.(2014) provides an example. Alternately, some open source package (such as https://www.tensorflow.org) might be applicable.

6.3. Overcoming the challenges We know from experience that CMB data processing is an iterating process requiring considerable back-and- forth iterations between map making and parameter estimation, which inform improvements on instrumental or foreground modelling issues. A brute force approach, using the methods and algorithms currently at our disposal, but applying them to larger data sets using larger computers, will not scale satisfactorily in such an iterative process. To overcome the challenges described above, new methods and algorithms have to be developed. Continuous improvements in the implementation of data analysis codes on massively parallel computer systems is clearly a requisite. For Planck, the CMB community had to construct dedicated tools to deal with the then massive data volume of 2TB of raw data. Parallelisation of the Planck simulation pipeline was in itself a significant achievement (Planck Collaboration et al. 2015e). For future missions, we need similar tools for a problem at least 3 orders of magnitude larger in terms of data volume, distributed over a 3 orders of magnitude larger number of computers. Such requirements imply that significant efforts also need to be made to train the scientist and engineers in parallel processing. Of course, access to large computer resources will have to be secured. This can be an issue for small ground based experiments. In Planck, the coming of age of Bayesian exploration techniques and statistical methods were instrumental in the derivation of certain CORE results, such as the generation of different foreground component maps (Planck Collaboration et al. 2015a; Planck Collaboration X 2016). We have already highlighted above how techniques similar to ABC (Akeret et al. 2015; Ishida et al. 2015) could help to efficiently evaluate simulation

French roadmap for CMB science 30/06/2016 6.4 Summary 43 based likelihoods, or how machine learning algorithms could be used to construct emulators to reduce the cost of simulations. This approach was actually used with great success in Planck, when estimating the cosmological parameters on diﬀerent data subsets; indeed, for the many tests needed to evaluate the robustness of the parameters estimation, we relied on interpolations between spectra generated by the full power spectrum computation code (Fendt & Wandelt 2007; Planck Collaboration et al. 2015g). It is clear that eﬀorts must be made in order to train astronomers in such methods and their application to CMB analysis.

6.4. Summary

The quality of the CMB data processing and analysis is of comparable importance to the data acquisi- • tion hardware and associated techniques in determining the ﬁnal outcome of the experiment. The data processing strategy cannot be left as an afterthought.

Future datasets will become more massive by at least two orders of magnitude (and much more on the • ground).

At the same time, the physics and the complexity of the instruments will force us to explore cases where • simplifying hypotheses such as the linearity of the detector response of the underlying signal cannot be taken for granted.

When models are less solid, we will need to rely even more on costly simulations for diﬀerent steps • of the data processing. While, of course, improvement in the models of the instrument or physical processes will help, some computations will probably be too expensive to be performed more than a few times.

Simulation results will become an integral part of the data delivery, as the only realistic means to • fully capture the properties of the processed data and allow further analyses (and conﬁdence) by the community.

Technical improvements in our coding techniques and algorithms are mandatory to cope with the in- • crease in data size. There is hope to overcome these challenges by investing more research in new mathematical methods in statistics and possibly machine learning. Above all, a continuous eﬀort in the training of scientist and engineers to these new techniques will be needed. The tasks ahead appear as daunting as those faced in the early days of Planck. But the Planck challenges were met thanks to a long term program which started around 1992. There is no reason to believe that the now much better prepared (and larger) CMB community will not solve these issues given a long term CMB program encompassing them from the start.

7. Landscape today

7.1. Current Sub-Orbital CMB Experiments Until recently, CMB experiments were often divided between experiments which observed at frequencies below and above 100 GHz. This was largely because (1) the technology used was different (heterodyne versus ∼ bolometric detection), and (2) the Galactic foregrounds are different (synchrotron versus thermal dust emission) in the two frequency regimes. In the quest for ever more sensitivity, however, current experiments have been moving towards the region above 100 GHz, both because the bolometric technology used in this regime can be more sensitive, and because the shorter wavelengths involved allow for more detectors to be put in a given amount of focal plane area. Here we will instead divide experiments based on their aperture size. At a given wavelength, of course, the aperture is inversely proportional to the angular resolution of the experiment. All other factors being equal, it is the angular resolution that defines the science which can be done by an experiment, with larger-aperture experiments being able to map arc-minute anisotropies in the CMB, but having more difficulties in measuring the largest angular-scale anisotropies that are targeted by smaller-aperture experiments.

French roadmap for CMB science 30/06/2016 44 7 LANDSCAPE TODAY

7.1.1. Large Aperture

As noted in Sect. 2.3.3 and 4.2, a wealth of astrophysics and cosmology becomes accessible when experiments can measure anisotropies at arc-minute angular scales. In addition, and as discussed in Sect. 2.3.4, if the amplitude of primordial B-Modes are too low, they could be hidden by the B-Modes generated by the lensing of E-Modes by large scale structure in the Universe. Reconstructing this contamination so that it can be removed from larger angular scale data will also require measurements of the CMB at an angular resolution of order a few arc-minutes. To get an angular resolution of order one arc-minute near 150 GHz requires a dish of nearly 7 m diameter. This is difficult to do anywhere but from the ground, as it is difficult to fit something of this size in a rocket faring to send it into space or to put it on a balloon or air plane. There are, however, a few such telescopes operating on the ground already. They are:

ACT: The Atacama Cosmology Telescope, or ACT for short, is a six meter telescope in the Atacama Desert in Chile. It observes in the 100-250 GHz range, using arrays of transition-edge sensors read out with time-domain multiplexing (Swetz et al. 2011). ACT has gone through diﬀerent incarnations where it has also been called Advanced ACT (AdvACT) and AdvACTPol. It has recently joined with the Polarbear team and the Simons Array to form the Simons Observatory (http://simonsobservatory.org/news.html). These experiments are shown in Fig. 20. More information can be found at http://act.princeton.edu/.

Figure 20: CMB experiments in the Chilean Andes. The ACT experiment is hidden within the large ground shields towards the right of the ﬁgure. The Polarbear telescope is labelled and to the left of ACT. The locations of future Simons Array telescopes are also indicated, as is the anticipated site of the CLASS telescope. Taken from Carlstrom(2016).

French roadmap for CMB science 30/06/2016 7.1 Current Sub-Orbital CMB Experiments 45

Polarbear and the Simons Array: The Polarbear project is growing to include multiple telescopes and being re-dubbed the Simons Array. It is also integrating with the ACT team to create the Simons Observatory. Polarbear has a 3.5 arc-minute beam at 150 GHz, with antenna-coupled, polarization sensitive TES bolometers cooled to 250 mK. It is shown near the centre of Fig. 20, just to the left of ACT. The sites of future Simons Array telescopes are also noted in Fig. 20. More information can be found at http://bolo.berkeley.edu/ polarbear/.

SPT: The South Pole Telescope, or SPT for short, is the largest active CMB telescope, with a 10 m primary, though it is conservatively illuminated, for an arc-minute beam at 150 GHz. It uses detectors similar to those of Polarbear, multiplexed in the frequency domain. More information can be found at http://pole.uchicago. edu/ and in Benson et al.(2014). It is the right-most telescope in Fig. 21.

MUSTANG: MUSTANG is a high-resolution instrument placed on the Green Bank telescope, which has been used to make high-angular-resolution Sunyaev-Zel’dovich measurements. More information can be found in Dicker et al.(2014) and the references therein.

NIKA2: The “New Instrument of KIDs Arrays”, or NIKA2, placed on the IRAM 30 m telescope, is the ﬁrst array of Kinetic Inductance Devices (KIDs) to make maps of the sky. It is a general purpose device open to the community, but the team that made the instrument has a large block of time they use for high-resolution SZ studies. More information can be found at http://ipag.osug.fr/nika2/Welcome.html.

7.1.2. Small Aperture

Smaller-aperture experiments are geared towards measuring the CMB on large angular scales. They can often be constructed using simpler optics, and as they have poorer angular resolution than large-aperture systems, they ignore smaller-scale features and can cover the more sky more quickly while preserving larger-scale information. The scientific goals of these experiments are also different from those of large-aperture systems. The primary goals for small-aperture systems are the optical depth to reionization, τ, and the Inflationary Tensor-to-Scalar Ratio, r, which do not require exquisite angular resolution.

BICEP/Keck: The BICEP/Keck project has been taking data longer than any other small-aperture experiment presently running. They have published data at 90 and 150 GHz to set model-dependent limits on r of 0.07 (in combination with Planck, BAO, and other cosmology data). Data at 220 GHz is being taken now. Keck is enshrouded by the wooden ground shields towards the right of Fig. 21, and BICEP is within the metallic ground shield between SPT and Keck.

CLASS: The Cosmology Large Angular Scale Surveyor (CLASS) is a future, funded array of telescopes to measure CMB polarization between 40 and 220 GHz in Chile. In addition to using TES sensors at 45 GHz, the project is also innovative by using variable-delay polarization modulators rather than half-wave plates. It plans to cover 70% of the sky (i.e., the sky fraction available from Chile). CLASS is indicated on the left of Fig. 20. More information can be found at http://cosmos.pha.jhu.edu/bennett/class.html.

QUBIC: QUBIC is perhaps the most innovative of the currently planned CMB experiments, being a bolometric, adding interferometer. Proposers had initially hoped to put it in Dome C in Antarctica. A shift of location to Argentina has recently been decided (06/2016) by the team, after technical demonstrations are performed in Europe. More information can be found at http://qubic.in2p3.fr/QUBIC/Home.html.

QUIJOTE: The Q-U-I JOint Tenerife Experiment (QUIJOTE) is a low-frequency (11-40 GHz) experiment mounted in the Canary Islands. It plans to map of order 5000 square degrees, with an angular resolu-

French roadmap for CMB science 30/06/2016 46 7 LANDSCAPE TODAY

Figure 21: The South Pole Telescope, BICEP and Keck, a few hundred meters from the South Pole. SPT is the telescope on the left of the image “facing” away from the observer. The wooden ground shields on the right of the photo surround the Keck instrument, and the smaller ground shield on the building between SPT and Keck is the BICEP instrument. Taken from Carlstrom(2016). tion of one degree. More information can be found at http://www.iac.es/proyecto/cmb/pages/en/ quijote-cmb-experiment.php.

7.1.3. Balloons

Balloons have been used successfully almost since the discovery of the CMB (Henry 1971). The ability to get above the Earth’s atmosphere has historically meant that single flights could achieve comparable sensitivity to that which ground-based instruments have only achieved over months or years of observations. Over time, they have become increasingly more sophisticated, and the ability to mount longer campaigns has made precision measurements such as that from BOOMERANG possible (de Bernardis et al. 2000b). So-called ultra-long- duration balloons (ULDB) are expected to be possible soon (see, for example, https://science.jpl.nasa. gov/projects/SuperBIT/), though their payloads will, of course, necessarily be more limited than those of shorter flights, at least in the near-team. These increases in flight duration are important, since the sophistication has come a price – the time to mount a balloon campaign has increased, to the point where it is becoming hard for balloons to compete with ground-based experiments which field multiple focal-planes and operate continuously. Still, balloons are considered important not only for higher frequency measurements, where the atmosphere makes ground observations difficult, but also for space qualification and perhaps some large- angular-scale measurements where side lobe pick-up also makes ground measurements difficult. The main characteristics of balloon experiments, current and planned, designed to search for primordial B-modes are listed in Table3. Here are some additional information.

French roadmap for CMB science 30/06/2016 7.1 Current Sub-Orbital CMB Experiments 47

Table 3: Balloon-borne CMB/Dust polarization experiments (current and planned).

Experiment Telescope Frequency Number Sky fraction Dust1 Flight Schedule Status diameter (m) range (GHz) of bands % sensitivity next ﬂight BFORE 1.3 150-350 4 50 - ULDB 2020 Proposed EBEX-IDS 1.5 150-360 7 4 0.2 Antarctica 2020 Proposed LSPE 0.50 44-240 5 20 - Svalbard - Funded PIPER 0.3 200-600 4 85 0.2 Texas/Australia 20162 Funded SPIDER 0.25 95-280 3 10 1 Antarctica 2016 Funded B-SIDE 0.85 400-700 1-2 5 0.2 Australia 2019 Proposed

(1) : Brightness sensitivity to dust polarization after scaling with the dust SED to a common frequency, normalized to 1 for that of Planck at 353 GHz over the southern Galactic cap (4.6 µK deg 1σ). Values for the balloon projects are taken from Errard et al.(2016) and the B-SIDE proposal. B-SIDE, PIPER and EBEX promise to achieve a sensitivity 5 times lower (i.e. better) than Planck. (2) : The ﬁrst ﬂights will map the sky at the two lowest frequencies 200 and 270 GHz.

Spider: Spider is a balloon-borne experiment very similar to the BICEP project – it is occasionally called “BICEP on a rope”. It had a successful long-duration flight at the beginning of 2015, and another is planned for late 2016. The first flight had detectors at 100 and 150 GHz; the second will add 285 GHz detectors.

EBEX: The E and B Experiment, EBEX, has had two ﬂights, with bands covering the 150-450 GHz range. It has recently been re-proposed to ﬂy with a complement of 20,000 detectors in its focal plane, covering seven frequency bands from 150 to 360 GHz. More information can be found at http://groups.physics.umn. edu/cosmology/ebex/.

PIPER: The Primordial Inflation Polarization Explorer is a balloon program designed to measure sky polarization at the four frequencies 200, 270, 350 and 600 GHz. It is planned for 8 conventional flights from both the northern and southern hemisphere to map 85% of the sky. Two frequencies are observed per flight. PIPER benefits from a lower background than other balloon projects because all optics, including the primary mirror, are cooled cryogenically to 1.5 K. More information can be found at http://asd.gsfc.nasa.gov/piper/.

Olimpo: Olimpo is a balloon-borne, 2.6 m telescope designed to measure the SZ eﬀect at 140, 220, 410 and 540 GHz. More information can be found at http://planck.roma1.infn.it/olimpo/.

LSPE: The Large-Scale Polarization Explorer, LSPE, is “spinning”, balloon-borne experiment designed to measure the polarization of the CMB over 20-25% of the sky. It will have ﬁve frequency bands covering the spectral region from 40 to 250 GHz, with an angular resolution of order 1.3◦. More information can be found at http://planck.roma1.infn.it/lspe/.

BFORE: BFORE is a proposal to NASA to ﬁeld a ULDB (ultra-long duration balloon) circum-terrestrial ﬂight with four spectral bands from 150 to 350 GHz, in order to characterize the polarization properties of dust. More information can be found in Niemack et al.(2015).

B-SIDE: B-SIDE is a balloon mission proposal to the French Space Agency, CNES, for high-frequency KIDs measurement of dust polarization in a CMB ﬁeld observed by a complementary ground-based experiment.

All the balloon projects listed in Table3, with the exception of B-SIDE, aim at measuring both CMB and dust. This choice follows from the wish to be able to reach a result on primordial B modes without ancillary − data. Overall, this strategy is not optimal because ground-based observations may reach a better sensitivity than balloon experiments at microwave frequencies, thanks to a much longer observing time. B-SIDE takes this into

French roadmap for CMB science 30/06/2016 48 7 LANDSCAPE TODAY account proposing a distinct strategy where ground-based and balloon-borne data are combined. The full ﬁeld of view of the instrument and all of the observing time are dedicated to observe dust polarization, at one or two frequencies, which may not be observed from the ground and provide dust polarization observations more sensitive than Planck at 353 GHz.

7.2. CMB Stage 4 CMB-S4 is gaining momentum as a community-based proposal for the next generation (or “stage”) of ground- based CMB work. The basic framework would be to continue CMB work at established sites at the South Pole and in Chile, perhaps with an additional northern site, if feasible. It expressly aims to be independent of satellite and balloon work, though endeavours to be complementary with them. A graphical synopsis of CMB-S4 is given in Fig. 22.

Figure 22: A rough timeline for CMB-Stage 4 from Carlstrom(2016). Stage-3 experiments are starting to be deployed now, to take data in the coming seasons. Among these are ACTPol, the BICEP/Keck program, the Simons Array and SPT-3G. At Stage-4, other experiments such as CLASS will join these.

As noted in the diagram, recent experiments (such as BICEP2/Keck, for example), should be considered “CMB-S2”. We are presently entering into stage-3, with some experiments fielding of order 10 000 detectors. The name stage-4 is meant to indicate that total number of detectors fielded by the combination of experiments will be of order half-a-million. The columns labelled σ (r), σ (Neff), σ (Σmν), and “D.E. F.O.M.” indicate the science targeted by S4, and give indications of the sensitivities expected to be attained as a function of time. σ (r) indicates that one of the goals of stage-4 will be to set limits on the Inflationary Tensor-to-Scalar Ratio, r. The second and third goals are to confirm and/or bound the effective numbers of neutrinos (or, in fact, early relativistic species) and a detection of the sum of the masses of neutrinos, rather than an upper limit. Finally,

French roadmap for CMB science 30/06/2016 7.3 Multiple Telescopes versus Large Focal Planes 49

S4 endeavours to improve our understanding of Dark Energy (D.E.), so the right column gives the ﬁgure of merit (F.O.M.) than can be attained in conjunction with other cosmological measurements. There is certainly much other science to be done with S4 surveys, but these are the baseline goals towards which the experiment will be scoped. Some of the implications of this outline for the experiment design are touched upon below.

7.3. Multiple Telescopes versus Large Focal Planes

Perhaps the easiest way to get the numbers of detectors necessary for the next generation of CMB science would be to replicate existing apparatuses. The BICEP project provides an illustration of one of the ways the field is evolving to obtain more sensitivity. The original BICEP instrument had about 100 individual detectors. This was increased in the BICEP2 instrument by more efficiently filling the focal plane with an array of detectors. This was increased again in the Keck instrument by using multiple, identical telescopes on the same mount. And the BICEP3 incarnation again increases the number of detectors looking at the sky by increasing the numbers of mounts. This evolution is shown in Fig. 23.

Figure 23: The evolution of the BICEP and Keck telescopes from Pryke(2016).

This replication is feasible for the small aperture instruments, which are relatively inexpensive to mount. Unfortunately, larger telescopes are more expensive than smaller telescopes (for optical telescopes, for example, the cost varies with diameter to the power of 2.7, NOAO 2002). Thus, while it may be possible to simply replicate small-aperture telescopes such as BICEP, the proposition is more diﬃcult for a large-aperture telescope such as SPT. Thus, the group is exploring ways to put signiﬁcantly more detectors into focal planes with given areas (see, for example, Niemack 2015).

French roadmap for CMB science 30/06/2016 50 7 LANDSCAPE TODAY

7.4. Location

Figure 24: Atmospheric transmission at various sites globally. The colour table indicates that in addition to the South Pole and the Atacama Desert, locations in Greenland and/or Tibet may be appropriate for CMB measurements. This ﬁgure is from Suen et al.(2014).

Most of the data from experiments mentioned above have been taken with telescopes at either the South Pole or the Atacama Desert. This is because there are very few sites with the combination of low optical depth and stability of atmosphere that can be found at these two.

7.4.1. Differences Between the South Pole and Chile Though both have been used successfully for CMB measurements, there are diﬀerences in the atmospheric quality even between the South Pole and the Chilean Atacama.

Atmospheric Stability While both the South Pole and Chile have been successfully used for CMB experiments, they are not the same in terms of atmosphere. Simply put, the South Pole seems to be qualitatively better than Chile in terms of atmospheric noise. This is manifested by the fact that the two existing South Pole CMB programs, BICEP/Keck and SPT, make polarization observations without polarization modulation (for example, from a rotating half-wave plate). In Chile, on the other hand, all experiments to be fielded there consider modulation necessary. This allows one to fit, or lock-in, to any signal with the known modulation and thereby minimize atmospheric contamination, which is negligibly polarized. In principle this should not make a big difference, but it might have implications for how data is compared and correlated. It may not be possible, for example, to take and store the data from the two sites in exactly

French roadmap for CMB science 30/06/2016 51 the same format. The pole appears also to have a larger time fraction of “very good” sky which may allow observing at high frequencies (i.e., 280 GHz).

Sky Coverage The sky visible from Chile is, of course, diﬀerent from that visible from the South Pole, though there is overlap. The combination of the two sites gives access to approximately 70% of the full-sky. The South Pole only has access to 50% of the sky (at most – and typically less than this, as most experiments cannot observe anywhere near the horizon), which is less than that accessible from Chile. However, it should be noted that most of the sky visible from the South Pole is always visible from there, meaning that experiments can “stare” at the same patch continuously. From Chile, on the other hand, more total sky is available, but experiments are forced to split their observations into at least two or three diﬀerent patches, due to the rotation of the Earth.

7.4.2. Other Possible Sites If telescopes are deployed only at the South Pole and the Atacama Desert, unfortunately the entire sky cannot be mapped, since about 30% of the sky is inaccessible even from Chile. This has led people to consider the possibility of opening further sites. As can seen from Fig. 24, there are few places with the atmospheric qualities of the South Pole and the Atacama Desert, but Greenland and Tibet are considered possibilities, and atmospheric testing campaigns that can be used to compare sites for their CMB appropriateness are being started. In addition, sites in the Argentine Atacama Desert (as opposed to Chile) are also being considered as possible CMB telescope sites (see http://www.llamaobservatory.org/site.htm).

8. The Future in CMB science Space offers unique advantages for performing measurements under the exacting requirements imposed by the objective of accurately and precisely measuring the tenuous polarized anisotropies of the CMB. Among other things, it offers an unmatched duration of stable conditions (HFI bolometers remained within 0.1 mK of their operating temperature of 0.1 K over the 900 days of HFI cryogenic operation), full sky coverage (leaving no mode undetected), a unique broad frequency coverage with many frequency bands (needed to tame foreground contributions), and considerable redundancy at widely separated spatial and temporal scales to minimize systematic effect residuals. ESA and the European community have demonstrated with Planck its capability to take full advantage of this privileged environment and set the world standard for this field. The long-term priority for the French community is therefore a strong participation in a CORE-like experiment, i.e., a CMB polarization imager which will be proposed by the European community in October 2016 to ESA’s call for opportunity for the M5 slot11, as an evolution of the previous proposals in L3 and M4. M5 is a priori targeted for a 2030 launch, although (see below the ESA’s CDF study result) it could be as soon as 2026, if the instruments could be ready in time, maybe within the context of a collaboration with a non-European agency like JAXA (but not a LiteBIRD-like experiment) and/or NASA. CORE-like will deliver science results around 2030 at the earliest (for a hypothetical 2026 launch), and more likely within the following 5 years. One should also note that the earliest any space experiment would yield results is after 2025, if the PIXIE proposal to NASA as an explorer class mission were to be decided in 2017 (proposals are due in December 2016), for a potential launch in 2023. The Japanese project LiteBIRD, a low resolution imager in space, is now entering phase A1, and could be slated for a launch in 2025. In both cases, only a small fraction of the French community would be involved, and results are to be expected in about 10 years at best. It follows that all new data and progress in this field in the next decade will come from the ground and balloons, to be followed by a synergy phase between space and ground. To preserve its expertise and benefit fully from lessons learnt in this context, the French community must therefore strongly participate in this ground-based effort (understood broadly, as including balloons). Following the classification of our US colleagues, the CMB ground-based effort may be described in a series of stages (see § 7.2), culminating with stage four (S4) which is supposed to start (if funded as expected) after

11 The selection of the 3 candidate missions for the M5 slot should be announced in June 2017, with their phase 0 during August to October 2017, their phase A kick-off in the first quarter of 2018, and the selection in November 2019. The final mission adoption should be exactly two years later.

French roadmap for CMB science 30/06/2016 52 8 THE FUTURE IN CMB SCIENCE

2020. This sets the bounds for any French eﬀort on the ground. Of course, ambitions on the ground will have to take into account the fate of the M5 proposal, but in any case, some form of participation is vitally needed. Short term projects must therefore be analysed within this global M5/S4 framework. In the following assessment, we thus start with CORE, LiteBIRD, PIXIE and S4, before turning to the analysis of shorter terms projects which can provide the stepping stones to this end. If a CORE-like proposal is not selected for an ESA phase A study, this will only render more imperative a strong participation in ground-based science. What should then be done from space in Europe would probably be within the context of an M6 call, perhaps issued in 2018, for a launch possibly in 2033 and more importantly, outside of the scope of this road map.

8.1. Long-term projects

8.1.1. M5 & CORE The comprehensive study of CMB polarization is a scientific imperative that requires observations by a space mission with exquisite sensitivity, matching control of systematic errors and broad frequency coverage. Any tentative detection of primordial gravitational waves by a suborbital experiment will require a confirmation because of uncertainties in the removal of Galactic foreground contamination. More broadly, space is necessary to best extract the unique information accessible with the CMB, by maximizing the number of modes measurable given the actual foregrounds whose distribution must be mapped in the first place, and by improving the accuracy and precision of the results through its unique calibration and cross-check capabilities. Moreover, the very large scales required for measuring the reionization bump of the B-mode polarized signal (multipoles below about ten) are extremely hard to obtain from the ground, if at all. One should also note that the so-called anomalies (unlikely values within the minimal ΛCDM model) were all found at large scales, and precise large scale polarisation measurements are needed to help in assessing their origin. CORE-like proposals to ESA have been designed to provide the ultimate experiment for probing the physics of the early Universe from CMB polarization observations. Specifically, they have been designed in order to provide the best possible detection and characterization of the primordial B-mode polarization for any value of the tensor-to-scalar ratio, r, larger than r 10 3 (implying σ 10 4), even in the presence of complex ∼ − r ∼ − polarized foregrounds. This sensitivity goal guarantees a 5-σ detection of the gravitational waves in Higgs inflation, the simplest, most natural minimal model (with no further fields than currently known). Conversely, an upper limit at this level would disfavour the complete class of so-called “large field” inflation models (where the field excursion during inflation is larger than the Planck mass). A space mission with this sensitivity and angular resolution will also provide the best precision achievable with the CMB on cosmological parameters in a broad sense, i.e., including extensions to the currently sufficient minimal ΛCDM model (i.e., with S/N > 1 at multipoles ` < 3000 for E-modes which are less sensitive to foregrounds). Such an experiment also probes the distribution of clustered mass in the Universe through the observation of the lensing of CMB polarization due to dark matter structures between our telescopes and the last scattering surface. The reconstruction of the CMB lensing potential will provide high signal-to-noise ratio maps of the distribution of dark matter at redshifts z = 1–3 without recourse to biased baryonic tracers. In addition to providing a map of the dark matter integrated along the line of sight up to high redshift, this measurement, combined with cosmological constraints from Euclid and/or other probes, could in principle constrain the sum of the 3 light neutrino masses with a statistical error of 3 meV (but see comment on eq. 13), 5 times better than any single cosmological probe alone and sufficient to distinguish unambiguously between the standard neutrino hierarchy, with a minimum mass sum of about 60 meV, and the inverted hierarchy, with a minimum mass sum of about 100 meV. In order to mitigate the effect of foregrounds, CORE will have to have a large number of channels over a large frequency range, and this will allow astrophysicists to address diverse ancillary science on clusters and large scale structures through their SZ signature, extragalactic sources, dust polarisation, alone and in synergy with other experiments (e.g., CMB-S4 for the high resolution SZ information, but also surveys in X-ray, and optical/near infrared). At the time of writing, the baseline for the proposed instrument design is not yet finalized, but the French community is fully involved in studying the trade-offs between cost, complexity, science return, etc., under

French roadmap for CMB science 30/06/2016 8.1 Long-term projects 53 various assumptions of synergy with other probes. The sensitivities and resolutions discussed are between 1.5 and 2.5 muK.arcmin for sensitivity and 4 to 6 arc minute in angular resolution. A somewhat “descoped” option with a 1.2m class telescope is also discussed, as well as further reduced angular resolution schemes. The starting point for the ongoing work is the M4 COrE+ proposition which was based on an array of 2410 cryogenically cooled, linearly polarized detectors (KIDs or TES) at the focus of a 1.5 meter aperture Gregorian telescope. The entire sky was to be surveyed with 19 frequency bands spanning the range from 60 to 600 GHz. The spacecraft would be located in a large Lissajous orbit around the Sun-Earth L2 Lagrange point to avoid far side-lobe contamination. The combination of three rotations of the spacecraft at different time scales provides an observation pattern such that each sky pixel is crossed frequently along many different directions. This scan strategy provides for a strong mitigation of systematic effects12 and will thus ensure optimal use of the inherent high sensitivity, especially for extracting large angular scale signals. The CORE+ instrument builds on the success of Planckand Herschel, re-utilizing many of the subsystems and methods developed by the mm/sub-mm community. In early 2016, a CDF study was conducted at ESA with CORE and Japanese scientists in order to investigate the technical feasibility of a potential collaborative project between JAXA and ESA in this field (to be proposed as a candidate for ESA’s M5 opportunity). It was agreed to study a 1.2 meter telescope, no half- wave plate, and various specifications (frequency range, sensitivity, scan parameters) lying between the M4 and LiteBIRD proposals. Details of the study and its outcome can be found at http://sci.esa.int/trs/ 57795-cmb-polarisation-mission-study/. No specific technical show-stoppers have been identified, and cost drivers and cost-saving options were identified; it was noted that a telescope larger than 1.5 m would be hard to accommodate within the proposed concept. These conclusions are a valuable input which is being used by the CORE proposers to finalize the baseline for the M5 proposal which has a 550 millions Euros ESA cost cap at completion. As has been the case for Planck, achieving the CORE cosmological science programme will require accurate separation of the many astrophysical foregrounds as well as exquisite control and assessment of calibration and systematic errors. The resulting CORE ultra-high sensitivity maps of the three Stokes parameters I, Q, and U in 19 frequency bands will establish a long-standing legacy and a reference dataset for microwave and submillimeter emission in both intensity and polarization over the full sky. With its large frequency coverage and its resolution, it will provide astrophysicists with the most detailed view yet of the Galactic magnetic field, unveiling its role in creating the filamentary web-like structures where stars form. This is critical in order to reach the low level of foreground residuals required in the CMB maps after component separation.

This M5 project is a unique opportunity for the European community to remain at the forefront of CMB research, building on the investment in, and successes of, the Planck mission. In addition to its broad science appeal, it oﬀers the possibility for both developing and using cutting edge technologies of wide- spread interest, including millimetre range detectors, cryogeny, sophisticated data analysis methods, and high performance computing. In addition, developments made for space would be a strategic asset in building a participation to the shorter term eﬀort on the ground (see next). For all these reasons, contributing to the M5 selection and securing a leading role in the mission if selected is the current top priority of the French CMB community.

8.1.2. LiteBIRD LiteBIRD is a JAXA-led satellite project, currently undertaking a phase A study in Japan as a candidate strategic large JAXA mission and costing of order 270 million dollars. Its Principal Investigator (PI) is M. Hazumi (IPMU/KEK). A US contingent is funded to participate in the study as a mission of opportunity at NASA, with A.T. Lee (UCB) nominated as PI. The science goals of LiteBIRD are to detect the tensor-to-scalar ratio, r, with a precision better than σ(r) 10 3 (including the effects of foreground and systematic effect cleaning and ∼ − their residuals), to characterize the CMB B-mode and E-mode spectra down to degree scales, and to constrain models of the large scale magnetic field in our Galaxy. The instrument, which consists of two telescopes

12 In particular one can build polarized maps per detector over a large fraction of the sky ( 40%) over a short time scale ∼ ( a week). This avoids combining detectors with diﬀerent band-passes to derive a polarisation measurement, which was ∼ a major diﬃculty in Planck.

French roadmap for CMB science 30/06/2016 54 8 THE FUTURE IN CMB SCIENCE

(at high and low frequency), will feature a total of over 2000 sensitive, multi-chroic detectors operating at 100 mK. They will probe the sky in 15 frequency bands, covering the range from 40 GHz up to 400 GHz, with an instrument resolution of 30 arc minutes at 150 GHz. It will be deployed at the Lagrange L2 point of the Sun- Earth system and, in order to control the instrumental systematics, it will implement an advanced scanning strategy combining spacecraft rotation, precession and spin, in addition to active modulation of the polarized signal via a rotating half-wave plate. The launch is scheduled currently for the spring of 2025. The project is intended to meet a budgetary constraint of about half the cost of CORE, but with essentially the same orbit, mass, and power requirements, due to the comparable target sensitivity and the implied cooling requirements (an important cost driver). Aside from cost (and any diﬀerences in the costing procedures at JAXA and ESA), LiteBIRD diﬀers in the planned signal modulation which severely limits the maximum angular resolution that can be achieved. This has direct implications on the project’s ability to perform adequate component separation and undertake science on its own.

LiteBIRD is less powerful and of narrower scientiﬁc scope than CORE, but is still technically quite challenging, especially given the cost cap. It should only be considered for substantial participation in the event that CORE is not selected for phase A study.

8.1.3. PIXIE PIXIE is a space-based spectrometer which covers a frequency range from 30 GHz to 6 THz over a number of tunable, contiguous bands. It continuously scans the sky both spatially and spectrally, and is based on experience with COBE-FIRAS. PIXIE is a fully differential system, which allows control over many systematic errors to be achieved quite efficiently. As compared to FIRAS, the larger etendue and lower bolometer NEP of PIXIE will provide large gains in signal-to-noise on the CMB, but care will have to be taken with the higher- frequency end of the bands to avoid photon noise. It will be re-proposed to NASA at the end of 2016, and if successful will be launched in 2023, before all other current space projects. PIXIE has a 2.6 degree top-hat beam. It is tailored for measuring large-scale phenomena such as the primordial CMB B-modes. PIXIE sensitivity to r is, in principle, comparable to that of LiteBIRD and not far from that of CORE, with a quoted σ(r) 5 10 4. One should note though that this is a novel instrumental concept for ∼ × − mapping CMB polarization and one lacks hindsight on systematic effects which may affect this experimental set-up. In some sense, this is a pathfinder mission for a promising instrumental approach. With the baseline configuration, the instrument is even more sensitive to sub-millimetre and far infrared frequencies than it is to CMB frequencies. PIXIE is the proposed experiment which will characterize the foregrounds best at large scales. Recent results in the field have shown that foreground mitigation will indeed be crucial for more sensitive future measurements. PIXIE will bring full sky information over a very broad range of frequency, but in a fairly limited number of 2.6 2.6 pixels. We also note that its maps of the Compton y- ◦ × ◦ parameter at large scales will be of great scientific value when analysed in combination with other experiments involving significant French participation, such as LSST and Euclid. This makes it complementary to higher resolution CMB experiments which will also be able to measure large scale structure through CMB lensing, which PIXIE cannot do with its modest angular resolution. In addition to measuring the polarization of the CMB, PIXIE’s spectral sensitivity will allow it to measure the spectral distortions in the CMB and characterize a number of high-frequency (compared to the CMB) emission mechanisms. It will extend FIRAS’s study of line emission in our Galaxy. It will measure the spectrum of the large-scale components of the Cosmic Infrared Background. It will measure the monopole component of the total Sunyaev-Zel’dovich effect. It will detect the y distortion from the epoch of reionization. It may even detect the traces of dark matter annihilation or other exotic scenarios. Given its sensitivity and how tenuous all distortions of primordial origin are as compared to more recent and much larger sources of distortions, PIXIE is guaranteed to teach us crucial information about these sources and reveal how to circumvent them for mining further primordial spectral information. On this front, too, PIXIE is a great pathfinder. On a side note, the committee was surprised to see no mention of the optical depth to reionization, τ, in the proposal. PIXIE should be able to set better limits on τ than Planck, and since this parameter is degenerate with other cosmological parameters, it is needed by other CMB experiments trying, for example, to limit the numbers of species of neutrinos and the sum of their masses. We were also surprised by the apparent

French roadmap for CMB science 30/06/2016 8.1 Long-term projects 55 lack of developments regarding the actual potential of PIXIE regarding component separation at high spectral resolution but low angular resolution, and in particular regarding the speciﬁcities of disentangling the many expected contributions to the zero levels.

Given its particular areas of expertise, French participation in PIXIE may actually help to address some of the questions raised above concerning the concept, in particular regarding systematics and component separation. French participation in PIXIE will however be limited to only a fraction of the community, and PIXIE in any case leaves aside the higher resolution science which is addressed by the Stage-4 and CORE experiments. Still, PIXIE would be the first data available from any proposed projects in space and it is in many respects a pioneering project. It has a very reasonable cost for the French community, which is offered proportionately large participation for the proposed financial involvement. We therefore support strongly the proposed French participation.

8.1.4. S4

As discussed above, all new data and progress in CMB science in the next decade will come exclusively from the ground and balloons. This phase will then be followed by a fruitful period of synergy between space and ground (see more on such synergetic aspects in the next section). To preserve and leverage its expertise, as well as to benefit fully from future lessons learnt during that phase, the French community must therefore strongly participate in this ground-based effort (understood more broadly, as including balloon-based programs). Sect. 7.2 already introduced the community-based effort in the US to develop the stage-4 program of CMB research on the ground. The scientific goals of this stage are very ambitious, although the strategy is not fully developed yet. Part of it stems from the fact that the US CMB community arose from a collection of competitive consortia, each built around different stage-2 and 3 experiments. It is now coming together to try to meet the challenges of stage-4, in particular for rationalising the production of detector arrays which need to be massively produced at a fraction of current costs. On both fronts, Europeans have substantial assets, in so far as Planck told us how to work together, and promising key technologies are well developed in Europe, from filtering techniques to CORE detector technologies. Indeed, NIKA2 or the joint work of the SpaceKIDS initiative demonstrated the potential of KIDS technologies to offer the promise of highly scalable arrays at reasonable costs both in space and on the ground. Given the scale of the effort involved, and the advancement of S3 experiments in the US, French teams cannot realistically be considered to contribute much scientific weight by themselves. This also applies to other CMB groups throughout Europe. It would therefore be best to join forces in Europe and discuss the many scientific, strategic and technological choices to be made to yield the best scientific return from the effort, as viewed from our perspective. We should be aiming at a substantial participation, perhaps of the order of 50 to 100 million Euros (AppendixB gives some details on the magnitude of pre-S4 US funding). One problem, of course, is that there is no European agency in charge of this sub-field of science, and one may then have to consider coordinated proposals to individual national funding agencies and to EU programs. Time is pressing though, since many groups in Europe have already been approached to contribute elements and/or money to the individual US sub-groups. Rather than buying-in with individual contributions, it would be more effective, though not easy to arrange, to negotiate joint participation, designed around a full telescope+instrument thus offering maximum flexibility of choice. Preliminary contacts show that similar thoughts are being pursued elsewhere in Europe, and in particular in the UK and Italy. In any case, existing contacts and collaborations between the French and US communities must also be reinforced.

We recommend helping lead CMB scientists in Europe to urgently set up a scientiﬁc consortium whose charge will be to provide a forum for discussions within Europe, undertake discussions with the US S4 stakeholders, investigate all funding options, and coordinate all necessary European actions with the help of the European APPEC and Astronet networks of funding agencies in Astro-particle Physics and Astronomy in order to become full partners in the CMB-S4 world eﬀort.

French roadmap for CMB science 30/06/2016 56 8 THE FUTURE IN CMB SCIENCE

8.1.5. Long term synergies CORE, PIXIE and CMB-S4 each address important aspects of the future of CMB observations. While each of them alone has the capacity to achieve ground breaking science, the joint analysis of these data sets opens yet additional opportunities. PIXIE is unique for measuring absolutely calibrated intensity maps to look for CMB spectral distortions and obtain the zero-level (monopole) of maps over a large frequency range. CMB-S4 is unique for obtaining high sensitivity maps with an angular resolution at the arc minute scale in CMB channels. CORE is unique for providing high quality, full sky, few-arc minute resolution maps in 15-20 frequency channels covering the full range of frequencies required for component separation. By combining PIXIE and CORE, 15-20 absolutely calibrated, high resolution full sky maps ranging from 60 to 200 GHz will be obtained. Data with increased angular resolution will help to clean foregrounds from PIXIE spectra in order to - subtract bright clusters and infrared sources, - model radio source emission, - model and subtract dust, zodiacal light and CIB emission. Reciprocally, PXIE spectra will help dealing with, e.g., line emission contamination in the CORE photometric bands. In addition, both will measure large scale polarisation, with a very different observation strategy and sets of systematic effects to tame, allowing for cross-validation of large scale observations. By combining the resulting maps with CMB-S4 observations, the (absolute) calibration of the space mission can be propagated to the ground-based observations. The increased sensitivity and angular resolution in the sky area covered by CMB-S4 will allow for both better CMB sensitivity and better lensing maps as well as delensing (and hence tightened constraint on r). The comparison of the lensing reconstruction by both will allow checking the impact of foreground emission and systematic effects on the ground-based data. Full-resolution, high sensitivity cluster science can be done by combining the high resolution CMB-S4 maps in the negative part of the tSZ (below 220 GHz), and the (relatively) high resolution CORE maps in the positive part (above 220 GHz), allowing tSZ to be disentangled from kSZ and relativistic corrections, as well as from contamination by radio sources and dusty galaxies.

8.2. Mid-term projects We now discuss the extent to which the mid-term projects identiﬁed so far (B-SIDE, QUBIC) can be considered or not as stepping stones toward our long term goals M5/S4.

8.2.1. B-SIDE: an experiment dedicated to Galactic dust polarization For frequencies greater than 70 GHz, thermal emission from elongated dust grains aligned with respect to the Galactic magnetic field is the dominant polarized foreground for CMB polarization studies. From Planck, we learned that the primordial B-mode polarization of the CMB cannot be measured without subtracting the Galactic dust emission, even in the faintest dust-emitting regions at high Galactic latitude (Planck Collaboration Int. XXX 2016). Any claim for a detection for primordial B-modes will have to face a critical assessment against an alternative interpretation involving Galactic residuals. The sensitivity goal of the next generation of CMB experiments (Stage-3) corresponds only to 5% of the expected dust B-mode signal at 150 GHz, at the peak of the recombination bump (` = 80) for the cleanest regions of the sky (such as that of the BICEP field). The contrast between the CMB and Galactic signal is even worse in larger fields. It is clear that the required accuracy in component separation cannot be reached without complementing ground-based data with observations measuring dust polarization with a higher sensitivity than Planck. Indeed, incorrect modelling of the spatial variation of the polarized dust SED, within the constraints set by the Planck data, will lead to a bias in the CMB estimation, larger than the sensitivity limit targeted by the next generation of CMB experiments.

The dust B-mode amplitude is such that it cannot be neglected anywhere on the sky when trying to measure a CMB B-mode primordial signal even at a level of r 0.1. To add complexity, the Planck data recently ∼ provided evidence for spatial variations of the spectral behaviour of the polarized dust, which are signiﬁcant

French roadmap for CMB science 30/06/2016 8.2 Mid-term projects 57 across microwave frequencies. Variations of the dust SED along the line of sight, correlated with the structure of interstellar matter and that of the magnetic field, generate changes of both the polarization fraction and the polarization angle over microwave frequencies. Current forecast studies quantifying the dust polarized foreground impact on future CMB experiments are ignoring this de-correlation of the dust polarization map with frequency. Thus, prior to a new space mission (with numerous frequency channels), the required accuracy can only be achieved by observing dust polarization from a balloon at a frequency higher than those of ground- based experiments (thus higher than 300 GHz). This is the purpose of the B-SIDE experiment. ∼ B-SIDE is a balloon experiment, designed to map Galactic dust polarization with high sensitivity, matching that expected for CMB ground-based experiments (observing at 95, 150 and 220 GHz) that will search for primordial B-modes down to r = 0.01 (5 σ). B-SIDE will observe the sky at one (possibly two) frequency bands centred at 600 GHz (and 450 GHz) with 5 7 resolution. The focal plane of B-SIDE will be made 0 − 0 of KID arrays, cooled down well below their critical temperature, and sensitive in the frequency range from 400-700 GHz. For a 20-hour effective integration time on the chosen CMB field, an accuracy of 2 10 3 (1 σ) × − in terms of r on the separation of dust and CMB B-modes is achieved for frequencies larger than 520 GHz for the instrument specification, and 390 GHz for the goal sensitivity. A single frequency instrument at 600 GHz is a secure choice if only one-day flights are possible. This alternative is not critical for B-SIDE science goal, nor for the design of the instrument. In both cases, the B-SIDE data combined with Planck 353 GHz and the ground-based 95, 150 and 220 GHz data will provide enough spectral bands to characterise the dust SED and its spatial variations.

This project capitalises on major strengths of the B-SIDE team: - world-class expertise on dust polarization (thanks to the analysis of the Planck data), - leadership in the reduction of polarized sub-millimetre data - unique expertise on KIDS arrays.

Compared to the PILOT experiment, B-SIDE has a much larger field of view (3 deg. diameter versus 0.8 1 deg), allowing an increase in mapping speed by a factor 10! B-SIDE will re-use some existing equipment × to reduce costs and construction time (including the telescope, gondola, pointing reconstruction and service module from PILOT). B-SIDE has a different strategy compared to concurrent experiments. In B-SIDE, the full field of view of the instrument and all of the observing time is dedicated to observe dust polarization at frequencies that are not observable from the ground. Consequently the B-SIDE collaboration has to collaborate with a ground-based experiment to reach a result on the CMB. Such a collaboration between B-SIDE and ground-based experiments will be mutually beneficial and is quite likely to happen, even though it cannot yet be stated upfront by current ground-based proposers. In addition, a balloon experiment can provide a very good test-bed for space mission technologies. In this context, B-SIDE will boost significantly the technology maturity of key technologies for future space projects at millimetre and sub-millimetre frequencies. Among these key technologies are the KIDS and their associated readout electronics, the closed-cycle dilution cooling (e.g. CORE, SPICA, ATHENA+, LiteBIRD, PIXIE) and the half wave plate modulation system using superconducting magnetic levitation (CORE, LiteBIRD). For the above stated reasons, the group strongly recommends the B-SIDE project. B-SIDE represents an invaluable opportunity in the near term. To be a success though, B-SIDE has to fly as soon as possible, and with the expected sensitivity. In order to have access to the most interesting cosmological fields B-SIDE flights have to be undertaken from the Southern hemisphere. Thus, the exact time-line has to be adjusted to match this constraint. As of today, the first flight could occur in April 2019, from Alice Springs (1 to 3 days), followed by a second flight in April 2020; the first flight has to be no later than 2019. After that, the Bfore US experiment – if selected mid-2016 – could become competitive. Needless to say, CNES must ensure that the flight will be as long as possible, certainly in excess of 20 hours.

8.2.2. QUBIC QUBIC is a unique instrumental concept that combines an imaging architecture and interferometry. The idea is to take advantage of both high sensitivity TES detectors and systematics mitigation of interferometry. The

French roadmap for CMB science 30/06/2016 58 9 OUTREACH scientific target is to detect primordial gravitational waves over a multipole range between ` 30 to ` 300, ∼ ∼ targeting a sky coverage of about 1% at an effective resolution of about 30 arcmin. The first module of QUBIC consists of two arrays of 1024 TES (992 on the sky, the remaining as systematic effect monitors), operating at 150 and 220 GHz. With two years of continuous observations, the first QUBIC module should be able to constrain B modes with σr = 0.01. The project, building on past experience (BRAIN), is a collaboration between mainly France and Italy and was meant to be ultimately installed at Dome C in Antartica. The project has reached a stage when the technological demonstrator is funded and supported for laboratory characterization, which should be complete by the end of 2016. Due to the lack of existing structures at Dome C, the installation of QUBIC in Antartica is compromised in the near future. An alternative opportunity recently appeared with Argentina offering funding and support to install QUBIC on the Llama site, which could host instruments around 2018. Our committee acknowledges the achievements reached by the QUBIC team, currently holding the record for detector multiplexing, and the originality of the concept, addressing a different mode of observation and hence associated systematic effects. Together with its claimed sensitivity, it should complement existing experiments 2 and provide valuable confirmation of B-modes detection in the r = 10− range.

To be successfully continued, QUBIC must now either be competitive with stage-3 experiments (Sect.7) or be a clear stepping stone in a strategy towards a European contribution to the fourth generation of CMB experiments. Both require that QUBIC must rapidly demonstrate the validity of its instrumental concept on the sky. For that, it must be so funded as to be able to stick to its current schedule without further delay and validate the technological demonstrator in the laboratory by the end of 2016, achieve first light on the sky in 2017 with the first module, and show nominal sensitivity and systematics control by the end of 2018. Past experience on other projects has shown how critical the stage of first light on the sky is, and we recommend that QUBIC makes this their absolute priority. Since Antartica is not consistent with this schedule, we recommend that the Llama site in Argentina becomes the first choice, at least as a first step. Should the Llama site construction be delayed, there still is interest in demonstrating the instrumental validity on any other site on the same timescale. In any case, QUBIC is the sole well-advanced CMB project on the ground with strong French participation, and it would be a pity not to leverage it and take advantage of past efforts to learn lessons for stage-4. This is only possible with well focused goals, strict adherence to schedule, and proper funding decisions taken at each step.

9. Outreach Cosmology is an attractive subject for the general public. CMB public outreach has been shown to be a powerful draw for significant attention from rather broad French audiences. This allows the conveyance of the scientific methodology in general, and in particular as it applies to astrophysics and cosmology, as well as fo- cussing attention on the technological prowess demanded by CMB science and specifically realised in French laboratories, CNES, and by industrial partners such as Thales Alenia Space (Cannes) or Air Liquide (Grenoble- Sassenage). Moreover, the many facets and challenges of CMB experiments (instrumentation, theory, analysis, international framework) have the capability to attract and motivate students. The French part of the Planck-HFI collaboration has developed a set of tools with the help of communication professionals which is now quite polished, well-known and widely distributed. In particular, this includes a well referenced web site13, exhibition and pedagogical material, etc. In addition to their Planck-specific part, they provide the scientific and methodological ground work required for CMB research.

For the short/mid-term, we recommend to oﬀer visibility to current CMB projects through the well- established planck.fr website, at least for the next three years, the time needed for current plans to be approved or otherwise, and to re-orient the content accordingly. This will keep people informed about CMB activities in France. This will require some continued funding to keep the site alive.

13 The site is www.planck.fr. In 2014, there had already been more than 100,000 unique consultations, with 23% coming from outside France (Canada, Belgium, USA, Switzerland, Morocco. . . ).

French roadmap for CMB science 30/06/2016 59

10. Conclusions The CMB is the gift that keeps on giving. Using next generation CMB measurements we will be searching for inflationary gravitational waves and rigorously testing single field slow roll inflation, determining the neutrino masses, mapping the dark matter and the baryon distribution, constraining the large-scale velocity field, investigating dark energy, testing general relativity and more still. These measurements will also offer numerous opportunities for further progress in many other fields, in particular with complementary planned experiments whose capabilities will be greatly enhanced once combined with CMB data. The French community has substantial assets in this demanding but fertile scientific area, with leading experience accrued from Planck-HFI and cutting-edge technologies further developed in the laboratories. A CORE-like M5 project focussed on polarization science, is the obvious long-term target for this community. This is the only project able to involve all of the French CMB community. In that respect, the community is well engaged in the preparatory work, and well positioned to take a leading position in the event of mission selection. Of course, one should pay very close attention to the technological readiness level, e.g., of closed cycle cooling or KID-based detectors, if they are to be selectable for an M5 mission, since level 6 TRL will be required before mission adoption. Decisions to fund the enhancement of the TRL need to be timed accordingly. The PIXIE project at NASA is of more reduced scope and can only involve a relatively small part of the community; but it is the scientific pathfinder needed to determine how to fully harness the information contained in the CMB spectral distortions. PIXIE is also a novel and promising approach for mapping the anisotropies at low-resolution but with the advantage of a wide spectral coverage. A quite interesting opportunity would be the provision from France of similar closed cycle cooling technology as envisaged for M5, provided that the TRL level 6 can be attained in time. Irrespective of the outcome of the space project selections, the next decade (at least) will be exclusively advanced from the ground and balloons, and it appears to be a vital necessity that French teams be part of the ambitious US-led world effort (S4), taking full advantage of the existing French balloon program. Given the scale of the effort involved, and the many existing collaborations in Europe which were developed through Planck, it is clearly preferable to favour joint European participation. Time is pressing, and various US teams are negotiating buy-ins with individual teams throughout France, Europe and the world to help fund their post- S3 experiments leading to S4. Without reaction, Europeans will be scattered as sub-system providers with little hope to continue developing CORE technologies, such as detectors, or the ability to independently decide on specific strategies. The difficulty is, of course, that there is no obvious European-wide funding agency for this type of mid-size endeavour. Setting up a European scientific consortium to discuss possibilities and promote joint participation is needed urgently, with the help of the APPEC and Astronet networks of European funding agencies in particle physics and astronomy, and possibly the EU. Indeed, while the CMB future is bright, it may very well turn out to be rather bleak in France barring fast action.

French roadmap for CMB science 30/06/2016 60 10 CONCLUSIONS

French roadmap for CMB science 30/06/2016 REFERENCES 61

References Abazajian, K. N., et al. 2012, Light Sterile Neutrinos: A White Paper, ArXiv e-prints, arXiv:1204.5379 Ackerman, L., et al. 2009, Dark matter and dark radiation, Phys. Rev. D, 79, 023519, arXiv:0810.5126 Ade, P., et al. 2008, First Season QUaD CMB Temperature and Polarization Power Spectra, ApJ, 674, 22, arXiv:0705.2359 Akeret, J., et al. 2015, Approximate Bayesian computation for forward modeling in cosmology, J. Cosmology Astropart. Phys., 8, 043, arXiv:1504.07245 Alexander, S. & Martin, J. 2005, Birefringent gravitational waves and the consistency check of inflation, Phys. Rev., D71, 063526, arXiv:hep-th/0410230 Allison, R., et al. 2015, Towards a cosmological neutrino mass detection, Phys. Rev. D, 92, 123535, arXiv:1509.07471 Andersson, B.-G., Lazarian, A., & Vaillancourt, J. E. 2015, Interstellar Dust Grain Alignment, ARA&A, 53, 501 Audren, B., et al. 2013a, Conservative constraints on early cosmology with MONTE PYTHON, J. Cosmology Astropart. Phys., 2, 001, arXiv:1210.7183 Audren, B., et al. 2013b, Neutrino masses and cosmological parameters from a Euclid-like survey: Markov Chain Monte Carlo forecasts including theoretical errors, J. Cosmology Astropart. Phys., 1, 026, arXiv:1210.2194 Barnaby, N., Pajer, E., & Peloso, M. 2012, Gauge field production in axion inflation: Consequences for monodromy, non-Gaussianity in the CMB, and gravitational waves at interferometers, Phys. Rev. D, 85, 023525, arXiv:1110.3327 Barvinsky, A. O., Kamenshchik, A. Yu., & Starobinsky, A. A. 2008, Inflation scenario via the Standard Model Higgs boson and LHC, JCAP, 0811, 021, arXiv:0809.2104 Battye, R. A., et al. 2011, Statistical properties of polarized radio sources at high frequency and their impact on cosmic microwave background polarization measurements, Monthly Notices of the Royal Astronomical Society, 413, 132 Bennett, C. L., et al. 2003, First-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Foreground Emission, ApJS, 148, 97, arXiv:astro-ph/0302208 Benoit-Lévy, A., Smith, K. M., & Hu, W. 2012, Non-Gaussian structure of the lensed CMB power spectra covariance matrix, Phys. Rev. D, 86, 123008, arXiv:1205.0474 Benson, B. A., et al. 2014, in Proceedings of the SPIE, Vol. 9153, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VII, 91531P Bernardeau, F., Nishimichi, T., & Taruya, A. 2014, Cosmic shear full nulling: sorting out dynamics, geometry and systematics, MNRAS, 445, 1526, arXiv:1312.0430 Béthermin, M., et al. 2013, The redshift evolution of the distribution of star formation among dark matter halos as seen in the infrared, A&A, 557, A66, arXiv:1304.3936 Bezrukov, F. & Shaposhnikov, M. 2009, Standard Model Higgs boson mass from inflation: Two loop analysis, JHEP, 07, 089, arXiv:0904.1537 Bezrukov, F. L. & Shaposhnikov, M. 2008, The Standard Model Higgs boson as the inflaton, Phys. Lett., B659, 703, arXiv:0710.3755 BICEP2 Collaboration, et al. 2014, Detection of B-Mode Polarization at Degree Angular Scales by BICEP2, Physical Review Letters, 112, 241101, arXiv:1403.3985 BICEP2/Keck and Planck Collaborations, et al. 2015, Joint Analysis of BICEP2/Keck Array and Planck Data, Physical Review Letters, 114, 101301, arXiv:1502.00612 Bock, J., et al. 2009, Study of the Experimental Probe of Inflationary Cosmology (EPIC)-Intemediate Mission for NASA’s Einstein Inflation Probe, ArXiv e-prints, arXiv:0906.1188 Bonaldi, A., et al. 2007, WMAP 3-yr data with Correlated Component Analysis: anomalous emission and impact of component separation on the CMB power spectrum, MNRAS, 382, 1791, arXiv:0707.0469 Boyarsky, A., Ruchayskiy, O., & Shaposhnikov, M. 2009, The Role of Sterile Neutrinos in Cosmology and Astrophysics, Annual Review of Nuclear and Particle Science, 59, 191, arXiv:0901.0011 Boyle, L. A. & Buonanno, A. 2008, Relating gravitational wave constraints from primordial nucleosynthesis, pul- sar timing, laser interferometers, and the CMB: Implications for the early universe, Phys. Rev. D, 78, 043531, arXiv:0708.2279 Brandenberger, R. H. 2015, String Gas Cosmology after Planck, Class. Quant. Grav., 32, 234002, arXiv:1505.02381 Cabass, G., Melchiorri, A., & Pajer, E. 2016, µ distortions or running: A guaranteed discovery from CMB spectrometry, Phys. Rev. D, 93, 083515, arXiv:1602.05578 Carlstrom, J. 2016, CMB-S4: The Next Generation CMB Experiment, presentation at the workshop “Towards a next space probe for CMB observations and cosmic origins exploration”, CERN, Geneva, Switzerland [Accessed: 2016 05 20]. Chapman, D., et al. 2014, in American Astronomical Society Meeting Abstracts, Vol. 223, American Astronomical Society Meeting Abstracts #223, 407.03 Chiang, H. C., et al. 2010, Measurement of Cosmic Microwave Background Polarization Power Spectra from Two Years of BICEP Data, ApJ, 711, 1123, arXiv:0906.1181

French roadmap for CMB science 30/06/2016 62 REFERENCES

Chluba, J. & Jeong, D. 2014, Teasing bits of information out of the CMB energy spectrum, MNRAS, 438, 2065, arXiv:1306.5751 Conklin, E. K. 1969, Velocity of the Earth with Respect to the Cosmic Background Radiation, Nature, 222, 971 de Bernardis, P., et al. 2000a, A flat Universe from high-resolution maps of the cosmic microwave background radiation, Nature, 404, 955, arXiv:astro-ph/0004404 de Bernardis, P., et al. 2000b, A flat Universe from high-resolution maps of the cosmic microwave background radiation, Nature, 404, 955, arXiv:astro-ph/0004404 Delabrouille, J., et al. 2009, A full sky, low foreground, high resolution CMB map from WMAP, A&A, 493, 835, arXiv:0807.0773 Delabrouille, J., Cardoso, J.-F., & Patanchon, G. 2003, Multidetector multicomponent spectral matching and applications for cosmic microwave background data analysis, MNRAS, 346, 1089, arXiv:astro-ph/0211504 Dicker, S. R., et al. 2014, MUSTANG 2: A Large Focal Plane Array for the 100 m Green Bank Telescope, Journal of Low Temperature Physics, 176, 808 Draine, B. T. & Hensley, B. 2013, Magnetic Nanoparticles in the Interstellar Medium: Emission Spectrum and Polarization, ApJ, 765, 159, arXiv:1205.7021 Drewes, M. 2016, What can the CMB tell about the microphysics of cosmic reheating?, JCAP, 1603, 013, arXiv:1511.03280 Eriksen, H. K., et al. 2006, Cosmic Microwave Background Component Separation by Parameter Estimation, ApJ, 641, 665, arXiv:astro-ph/0508268 Eriksen, H. K., et al. 2008, Joint Bayesian Component Separation and CMB Power Spectrum Estimation, ApJ, 676, 10, arXiv:0709.1058 Errard, J., et al. 2016, Robust forecasts on fundamental physics from the foreground-obscured, gravitationally-lensed CMB polarization, J. Cosmology Astropart. Phys., 3, 052, arXiv:1509.06770 Fendt, W. A. & Wandelt, B. D. 2007, Computing High Accuracy Power Spectra with Pico, ArXiv e-prints, arXiv:0712.0194 Fernández-Cobos, R., et al. 2012, Multiresolution internal template cleaning: an application to the Wilkinson Microwave Anisotropy Probe 7-yr polarization data, MNRAS, 420, 2162, arXiv:1106.2016 Fissel, L. M., et al. 2010, in Proc. SPIE, Vol. 7741, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy V, 77410E–77410E–14 Fixsen, D. J., et al. 1996, The Cosmic Microwave Background Spectrum from the Full COBE FIRAS Data Set, ApJ, 473, 576, arXiv:astro-ph/9605054 Fixsen, D. J. & Mather, J. C. 2002, The Spectral Results of the Far-Infrared Absolute Spectrophotometer Instrument on COBE, ApJ, 581, 817 Génova-Santos, R., et al. 2015, QUIJOTE scientific results - I. Measurements of the intensity and polarisation of the anomalous microwave emission in the Perseus molecular complex, MNRAS, 452, 4169, arXiv:1501.04491 Goldreich, P. & Kylafis, N. D. 1982, Linear polarization of radio frequency lines in molecular clouds and circumstellar envelopes, ApJ, 253, 606 Graff, P., et al. 2014, SKYNET: an efficient and robust neural network training tool for machine learning in astronomy, MNRAS, 441, 1741, arXiv:1309.0790 Greaves, J. S. & Holland, W. S. 2002, in American Institute of Physics Conference Series, Vol. 609, Astrophysical Polarized Backgrounds, ed. S. Cecchini, S. Cortiglioni, R. Sault, & C. Sbarra, 267–270 Gush, H. P., Halpern, M., & Wishnow, E. H. 1990, Rocket measurement of the cosmic-background-radiation mm-wave spectrum, Physical Review Letters, 65, 537 Guth, A. H. 1981, The Inflationary Universe: A Possible Solution to the Horizon and Flatness Problems, Phys. Rev., D23, 347 Hanany, S., et al. 2000, MAXIMA-1: A Measurement of the Cosmic Microwave Background Anisotropy on Angular Scales of 10’-5deg, ApJ, 545, L5, arXiv:astro-ph/0005123 Hanany, S., Niemack, M. D., & Page, L. 2013, CMB Telescopes and Optical Systems, 431 Hanson, D., et al. 2013, Detection of B-Mode Polarization in the Cosmic Microwave Background with Data from the South Pole Telescope, Physical Review Letters, 111, 141301, arXiv:1307.5830 Henry, P. S. 1971, Isotropy of the 3 K Background, Nature, 231, 516 Hill, J. C., et al. 2015, Taking the Universe’s Temperature with Spectral Distortions of the Cosmic Microwave Background, Physical Review Letters, 115, 261301, arXiv:1507.01583 Hinshaw, G., et al. 2007, Three-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Temperature Analysis, ApJS, 170, 288, arXiv:astro-ph/0603451 Hoang, T. & Lazarian, A. 2016, Polarization of Magnetic Dipole Emission and Spinning Dust Emission from Magnetic Nanoparticles, ApJ, 821, 91, arXiv:1511.03691 Hoang, T., Lazarian, A., & Martin, P. G. 2013, Constraint on the Polarization of Electric Dipole Emission from Spinning Dust, ApJ, 779, 152, arXiv:1305.0276 Hu, W. 2002, Dark synergy: Gravitational lensing and the CMB, Phys. Rev. D, 65, 023003, arXiv:astro-ph/0108090

French roadmap for CMB science 30/06/2016 REFERENCES 63

Hu, W., Hedman, M. M., & Zaldarriaga, M. 2003, Benchmark parameters for CMB polarization experiments, Phys. Rev. D, 67, 043004, arXiv:astro-ph/0210096 Huterer, D. & White, M. 2005, Nulling tomography with weak gravitational lensing, Phys. Rev. D, 72, 043002, arXiv:astro-ph/0501451 Ishida, E. E. O., et al. 2015, COSMOABC: Likelihood-free inference via Population Monte Carlo Approximate Bayesian Computation, Astronomy and Computing, 13, 1, arXiv:1504.06129 Joachimi, B. & Schneider, P. 2008, The removal of shear-ellipticity correlations from the cosmic shear signal via nulling techniques, A&A, 488, 829, arXiv:0804.2292 Junklewitz, H. & Enßlin, T. A. 2011, Imprints of magnetic power and helicity spectra on radio polarimetry statistics, A&A, 530, A88, arXiv:1008.1243 Kamionkowski, M., Kosowsky, A., & Stebbins, A. 1997, Statistics of cosmic microwave background polarization, Phys. Rev. D, 55, 7368, arXiv:astro-ph/9611125 Kaplinghat, M., Knox, L., & Song, Y.-S. 2003, Determining Neutrino Mass from the Cosmic Microwave Background Alone, Physical Review Letters, 91, 241301, arXiv:astro-ph/0303344 Katayama, N. & Komatsu, E. 2011, Simple Foreground Cleaning Algorithm for Detecting Primordial B-mode Polarization of the Cosmic Microwave Background, ApJ, 737, 78, arXiv:1101.5210 Keisler, R., et al. 2015, Measurements of Sub-degree B-mode Polarization in the Cosmic Microwave Background from 100 Square Degrees of SPTpol Data, ApJ, 807, 151, arXiv:1503.02315 Kesden, M., Cooray, A., & Kamionkowski, M. 2002, Separation of Gravitational-Wave and Cosmic-Shear Contributions to Cosmic Microwave Background Polarization, Physical Review Letters, 89, 011304, arXiv:astro-ph/0202434 Khatri, R. & Sunyaev, R. 2015a, Constraints on µ-distortion fluctuations and primordial non-Gaussianity from Planck data, J. Cosmology Astropart. Phys., 9, 026, arXiv:1507.05615 Khatri, R. & Sunyaev, R. 2015b, Limits on the fluctuating part of y-type distortion monopole from Planck and SPT results, J. Cosmology Astropart. Phys., 8, 013, arXiv:1505.00781 Kitching, T. D., Heavens, A. F., & Das, S. 2015, 3D weak gravitational lensing of the CMB and galaxies, MNRAS, 449, 2205, arXiv:1408.7052 Kitching, T. D. & Taylor, A. N. 2011, On mitigation of the uncertainty in non-linear matter clustering for cosmic shear tomography, MNRAS, 416, 1717, arXiv:1012.3479 Knox, L. & Song, Y.-S. 2002, Limit on the Detectability of the Energy Scale of Inflation, Physical Review Letters, 89, 011303, arXiv:astro-ph/0202286 Kofman, L., Linde, A. D., & Starobinsky, A. A. 1997, Towards the theory of reheating after inflation, Phys.Rev., D56, 3258, arXiv:hep-ph/9704452 Kogut, A., et al. 2003, First-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Temperature- Polarization Correlation, Astrophysical Journal Supplement, 148, 161, arXiv:astro-ph/0302213 Kovac, J. M., et al. 2002, Detection of polarization in the cosmic microwave background using DASI, Nature, 420, 772, arXiv:astro-ph/0209478 Kurki-Suonio, H., et al. 2009, Destriping CMB temperature and polarization maps, A&A, 506, 1511, arXiv:0904.3623 Kwan, J., et al. 2013, Cosmic Emulation: Fast Predictions for the Galaxy Power Spectrum, ArXiv e-prints, arXiv:1311.6444 Lazarian, A. & Pogosyan, D. 2016, Spectrum and Anisotropy of Turbulence from Multi-frequency Measurement of Synchrotron Polarization, ApJ, 818, 178, arXiv:1511.01537 Lazear, J., et al. 2014, in Proc. SPIE, Vol.9153, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VII, 91531L Lewis, A. & Bridle, S. 2002, Cosmological parameters from CMB and other data: A Monte Carlo approach, Phys. Rev., D66, 103511, arXiv:astro-ph/0205436 Lewis, A. & Challinor, A. 2006, Weak gravitational lensing of the CMB, Phys. Rep., 429, 1 Li, H.-B. & Henning, T. 2011, The alignment of molecular cloud magnetic fields with the spiral arms in M33, Nature, 479, 499, arXiv:1111.2745 Linde, A., Mooij, S., & Pajer, E. 2013, Gauge field production in supergravity inflation: Local non-Gaussianity and primordial black holes, Phys. Rev. D, 87, 103506, arXiv:1212.1693 Linde, A. D. 1982, A New Inflationary Universe Scenario: A Possible Solution of the Horizon, Flatness, Homogeneity, Isotropy and Primordial Monopole Problems, Phys. Lett., B108, 389 Maillard, J. P., et al. 1993, in Bulletin of the American Astronomical Society, Vol. 25, AAS/Division for Planetary Sciences Meeting Abstracts #25, 1095 Maino, D., et al. 2002, All-sky astrophysical component separation with Fast Independent Component Analysis (FASTICA), MNRAS, 334, 53, arXiv:astro-ph/0108362 Marian, L. & Bernstein, G. M. 2007, Detectability of CMB tensor B modes via delensing with weak lensing galaxy surveys, Phys. Rev. D, 76, 123009, arXiv:0710.2538 Martin, D. H. & Puplett, E. 1970, Polarised interferometric spectrometry for the millimeter and submillimeter spectrum., Infrared Physics, 10, 105

French roadmap for CMB science 30/06/2016 64 REFERENCES

Martin, J. & Ringeval, C. 2010, First CMB Constraints on the Inflationary Reheating Temperature, Phys.Rev., D82, 023511, arXiv:1004.5525 Martin, J., et al. 2014a, The Best Inflationary Models After Planck, JCAP, 1403, 039, arXiv:1312.3529 Martin, J., Ringeval, C., & Vennin, V. 2014b, EncyclopÃ¦dia Inflationaris, Phys.Dark Univ., arXiv:1303.3787 Martin, J., Ringeval, C., & Vennin, V. 2014c, How Well Can Future CMB Missions Constrain Cosmic Inflation?, JCAP, 1410, 038, arXiv:1407.4034 Martin, J., Ringeval, C., & Vennin, V. 2014d, Observing the Inflationary Reheating, arXiv:1410.7958 Martin, J., Ringeval, C., & Vennin, V. 2016, Information Gain on Reheating: the One Bit Milestone, Phys. Rev., D93, 103532, arXiv:1603.02606 Mather, J. C., et al. 1994, Measurement of the cosmic microwave background spectrum by the COBE FIRAS instrument, ApJ, 420, 439 Mather, J. C., et al. 1990, A preliminary measurement of the cosmic microwave background spectrum by the Cosmic Background Explorer (COBE) satellite, Astrophywical Journal Letters, 354, L37 Mauskopf, P. D., et al. 2000, Measurement of a Peak in the Cosmic Microwave Background Power Spectrum from the North American Test Flight of Boomerang, Astrophysical Journal Letters, 536, L59, arXiv:astro-ph/9911444 Miller, A., et al. 2002, The QMAP and MAT/TOCO Experiments for Measuring Anisotropy in the Cosmic Microwave Background, Astrophysical Journal Supplement, 140, 115, arXiv:astro-ph/0108030 Misawa, R., et al. 2014, in Proc. SPIE, Vol. 9153, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VII, 91531H Monfardini, A., Boulanger, F., & Bernard, J. P. e. a. 2016, BSIDE, CNES Mukhanov, V. F. & Chibisov, G. 1981, Quantum Fluctuation and Nonsingular Universe. (In Russian), JETP Lett., 33, 532 Mukhanov, V. F. & Chibisov, G. 1982, The Vacuum energy and large scale structure of the universe, Sov. Phys. JETP, 56, 258 Namikawa, T. & Nagata, R. 2015, Non-Gaussian structure of B-mode polarization after delensing, J. Cosmology Astropart. Phys., 10, 004, arXiv:1506.09209 Naylor, D. A., et al. 2006, in Proc. SPIE, Vol. 6265, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 626530 Niemack, M. D. 2015, Designs for a large-aperture telescope to map the CMB 10X faster, ArXiv e-prints, arXiv:1511.04506 Niemack, M. D., et al. 2015, BFORE: The B-mode Foreground Experiment, Journal of Low Temperature Physics, arXiv:1509.05392 NOAO. 2002, Enabling a Giant Segmented Mirror Telescope for the Astronomical Community, Tech. rep., NOAO O’Dea, D., Challinor, A., & Johnson, B. R. 2007, Systematic errors in cosmic microwave background polarization measurements, MNRAS, 376, 1767, arXiv:astro-ph/0610361 Penzias, A. A. & Wilson, R. W. 1965, A Measurement of Excess Antenna Temperature at 4080 Mc/s., Astrophysical Journal, 142, 419 Planck Collaboration, et al. 2015a, Planck 2015 results. IX. Diffuse component separation: CMB maps, ArXiv e-prints, arXiv:1502.05956 Planck Collaboration, et al. 2015b, Planck 2015 results. VII. HFI TOI and beam processing, ArXiv e-prints, arXiv:1502.01586 Planck Collaboration, et al. 2015c, Planck 2015 results. IX. Diffuse component separation: CMB maps, ArXiv e-prints, arXiv:1502.05956 Planck Collaboration, et al. 2016a, Planck intermediate results. XLVII. Planck constraints on reionization history, ArXiv e-prints, arXiv:1605.03507 Planck Collaboration, et al. 2015d, Planck 2015 results. XX. Constraints on inflation, ArXiv e-prints, arXiv:1502.02114 Planck Collaboration, et al. 2015e, Planck 2015 results. XII. Full Focal Plane simulations, ArXiv e-prints, arXiv:1509.06348 Planck Collaboration, et al. 2015f, Planck 2015 results. XV. Gravitational lensing, ArXiv e-prints, arXiv:1502.01591 Planck Collaboration, et al. 2015g, Planck 2015 results. XI. CMB power spectra, likelihoods, and robustness of parameters, ArXiv e-prints Planck Collaboration, et al. 2015h, Planck 2015 results. XXII. A map of the thermal Sunyaev-Zeldovich effect, ArXiv e-prints, arXiv:1502.01596 Planck Collaboration, et al. 2016b, Planck intermediate results. XLVI. Reduction of large-scale systematic effects in HFI polarization maps and estimation of the reionization optical depth, ArXiv e-prints, arXiv:1605.02985 Planck Collaboration Int. L. 2016, Planck intermediate results. L. Evidence for spatial variation of the polarized thermal dust spectral energy distribution and implications for CMB B-mode analysis, A&A, submitted, arXiv:1606.07335 Planck Collaboration Int. XI. 2013, Planck intermediate results. XI. The gas content of dark matter halos: the Sunyaev- Zeldovich-stellar mass relation for locally brightest galaxies, A&A, 557, A52, arXiv:1212.4131 Planck Collaboration Int. XIX. 2015, Planck intermediate results. XIX. An overview of the polarized thermal emission from Galactic dust, A&A, 576, A104, arXiv:1405.0871

French roadmap for CMB science 30/06/2016 REFERENCES 65

Planck Collaboration Int. XLIV. 2016, Planck intermediate results. XLIV. The structure of the Galactic magnetic field from dust polarization maps of the southern Galactic cap, ArXiv e-prints, arXiv:1604.01029 Planck Collaboration Int. XX. 2015, Planck intermediate results. XX. Comparison of polarized thermal emission from Galactic dust with simulations of MHD turbulence, A&A, 576, A105, arXiv:1405.0872 Planck Collaboration Int. XXII. 2015, Planck intermediate results. XXII. Frequency dependence of thermal emission from Galactic dust in intensity and polarization, A&A, submitted, 576, A107, arXiv:1405.0874 Planck Collaboration Int. XXVII. 2015, Planck intermediate results. XXVII. High-redshift infrared galaxy overden- sity candidates and lensed sources discovered by Planck and confirmed by Herschel-SPIRE, A&A, 582, A30, arXiv:1503.08773 Planck Collaboration Int. XXX. 2016, Planck intermediate results. XXX. The angular power spectrum of polarized dust emission at intermediate and high Galactic latitudes, A&A, 586, A133, arXiv:1409.5738 Planck Collaboration Int. XXXVIII. 2016, Planck intermediate results. XXXVIII. E- and B-modes of dust polarization from the magnetized filamentary structure of the interstellar medium, A&A, 586, A141, arXiv:1505.02779 Planck Collaboration VIII. 2015, Planck 2015 results. VIII. High Frequency Instrument data processing: Calibration and maps, ArXiv e-prints, arXiv:1502.01587 Planck Collaboration X. 2016, Planck 2015 results. X. Diffuse component separation: Foreground maps, A&A, submitted, arXiv:1502.01588 Planck Collaboration XXIII. 2015, Planck 2015 results. XXIII. The thermal Sunyaev-Zeldovich effect–cosmic infrared background correlation, ArXiv e-prints, arXiv:1509.06555 Planck Collaboration XXV. 2016, Planck 2015 results. XXV. Diffuse, low-frequency Galactic foregrounds, A&A, in press, arXiv:1506.06660 Planck Collaboration XXX. 2014, Planck 2013 results. XXX. Cosmic infrared background measurements and implications for star formation, A&A, 571, A30, arXiv:1309.0382 Pryke, C. 2016, The CMB-S3 Landscape (US ground based), presentation at the workshop “Towards a next space probe for CMB observations and cosmic origins exploration”, CERN, Geneva, Switzerland [Accessed: 2016 05 20]. Rahlin, A. S., et al. 2014, in Proc. SPIE, Vol. 9153, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VII, 915313 Remazeilles, M., et al. 2016, Sensitivity and foreground modelling for large-scale cosmic microwave background B-mode polarization satellite missions, MNRAS, 458, 2032, arXiv:1509.04714 Ritacco, A., et al. 2016, NIKA 2: next-generation continuum/polarized camera at the IRAM 30 m telescope and its prototype, ArXiv e-prints, arXiv:1602.01605 Ritacco, A. et al, i. p. 2016, Polarimetry at millimeter wavelength with NIKA: calibration and characterization, ArXiv e-prints Rosset, C., et al. 2010, Planck pre-launch status: High Frequency Instrument polarization calibration, A&A, 520, A13, arXiv:1004.2595 Seljak, U. & Hirata, C. M. 2004, Gravitational lensing as a contaminant of the gravity wave signal in the CMB, Phys. Rev. D, 69, 043005, arXiv:astro-ph/0310163 Sherwin, B. D. & Schmittfull, M. 2015, Delensing the CMB with the cosmic infrared background, Phys. Rev. D, 92, 043005, arXiv:1502.05356 Shimon, M., et al. 2008, CMB polarization systematics due to beam asymmetry: Impact on inflationary science, Phys. Rev. D, 77, 083003, arXiv:0709.1513 Shtanov, Y., Traschen, J. H., & Brandenberger, R. H. 1995, Universe reheating after inflation, Phys.Rev., D51, 5438, arXiv:hep-ph/9407247 Simard, G., Hanson, D., & Holder, G. 2015, Prospects for Delensing the Cosmic Microwave Background for Studying Inflation, ApJ, 807, 166, arXiv:1410.0691 Smith, K. M., et al. 2012, Delensing CMB polarization with external datasets, J. Cosmology Astropart. Phys., 6, 014, arXiv:1010.0048 Smoot, G. F., et al. 1992, Structure in the COBE differential microwave radiometer first-year maps, Astrophysical Journal Letters, 396, L1 Smoot, G. F., Gorenstein, M. V., & Muller, R. A. 1977, Detection of anisotropy in the cosmic blackbody radiation, Physical Review Letters, 39, 898 Starobinsky, A. A. 1980, A new type of isotropic cosmological models without singularity, Phys. Lett., B91, 99 Starobinsky, A. A. 1982, Dynamics of Phase Transition in the New Inflationary Universe Scenario and Generation of Perturbations, Phys. Lett., B117, 175 Strukov, I. A. & Skulachev, D. P. 1984, Deep-Space Measurements of the Microwave Background Anisotropy - First Results of the Relikt Experiment, Soviet Astronomy Letters, 10, 1 Suen, J. Y., Fang, M. T., & Lubin, P. M. 2014, IEEE Proceeding on Terahertz Science and Technology Swetz, D. S., et al. 2011, Overview of the Atacama Cosmology Telescope: Receiver, Instrumentation, and Telescope Systems, Astrophysical Journal Supplement, 194, 41, arXiv:1007.0290

French roadmap for CMB science 30/06/2016 66 REFERENCES

The Polarbear Collaboration: P. A. R. Ade, et al. 2014, A Measurement of the Cosmic Microwave Background B-mode Polarization Power Spectrum at Sub-degree Scales with POLARBEAR, ApJ, 794, 171, arXiv:1403.2369 Traschen, J. H. & Brandenberger, R. H. 1990, PARTICLE PRODUCTION DURING OUT-OF-EQUILIBRIUM PHASE TRANSITIONS, Phys.Rev., D42, 2491 Tucci, M. & Toﬀolatti, L. 2012, The Impact of Polarized Extragalactic Radio Sources on the Detection of CMB Anisotropies in Polarization, Advances in Astronomy, 2012, 624987, arXiv:1204.0427 Turner, M. S. 1983, Coherent Scalar Field Oscillations in an Expanding Universe, Phys. Rev., D28, 1243 van Engelen, A., et al. 2015, The Atacama Cosmology Telescope: Lensing of CMB Temperature and Polarization Derived from Cosmic Infrared Background Cross-correlation, ApJ, 808, 7, arXiv:1412.0626 Viel, M., et al. 2005, Constraining warm dark matter candidates including sterile neutrinos and light gravitinos with WMAP and the Lyman-α forest, Phys. Rev. D, 71, 063534, arXiv:astro-ph/0501562 Viero, M. P., et al. 2013, HerMES: Cosmic Infrared Background Anisotropies and the Clustering of Dusty Star-forming Galaxies, ApJ, 772, 77, arXiv:1208.5049 Waelkens, A. H., Schekochihin, A. A., & Enßlin, T. A. 2009, Probing magnetic turbulence by synchrotron polarimetry: statistics and structure of magnetic ﬁelds from Stokes correlators, MNRAS, 398, 1970, arXiv:0903.3056 Zaldarriaga, M. & Seljak, U. 1997, All-sky analysis of polarization in the microwave background, Phys. Rev. D, 55, 1830, arXiv:astro-ph/9609170

French roadmap for CMB science 30/06/2016 67

Appendix A: French CMB PhD Theses since Planck inception

French roadmap for CMB science 30/06/2016 68 A FRENCH CMB PHD THESES SINCE PLANCK INCEPTION

Name Title Year Institute R. Adam Observation of galaxy clusters via the Sunyaev-Zel’dovich 2015 LPSC effect and the polarization of the cosmic microwave background: from Planck to NIKA Dana Alina Analysis of the interstellar dust polarized emission with the 2015 IRAP, Toulouse Planck Satellite Survey data Lapo Fanciullo New insights on dust properties from Planck intensity and 2015 IAS, Univ Paris- polarization data Sud, Orsay D. Guéry Étude statistique des structures à grand redshift observées 2015 IAS, Univ Paris- par les satellites Planck et Herschel Sud, Orsay Florent Leclercq Bayesian Large Scale Structure inference and cosmic web 2015 IAP Paris analysis A. Miniussi Étude et modélisation de l’interaction des particules cos- 2015 IAS Orsay miques avec les détecteurs de l’instrumentation spatiale sub- millimétrique et X Marta Spinelli Cosmological parameter estimation with the Planck satellite 2015 LAL, Univ. Paris- data: from the construction of a likelihood to neutrino prop- Sud, Orsay erties A. Bracco Statistical properties of the Galactic magnetic field observed 2014 IAS with the Planck satellite Matthieu Roman Amas de galaxies déltectés par Planck avec l’effet Sunyaev- 2014 Paris Zel’dovich thermique: contraintes cosmologiques et spectre angulaire Flavien The blind Bayesian approach to cosmic microwave back- 2014 IAP Paris Vansyngel ground data analysis Guillaume Castex Simulation et modélisation du ciel dans le domaine mil- 2013 Paris limétrique S. Ilic The large scale structures. A window on the dark compo- 2013 IAS, Univ. Paris- nents of the Universe Sud, Orsay Fabien Lacasa Non-gaussianity and extragalactic foregrounds to the Cosmic 2013 IAS, Univ. Paris- Microwave Background Sud, Orsay Loïc Maurin La mesure de la polarisation avec Planck HFI : Calibration, 2013 APC Paris effets systématiques et sources compactes T. Déchelette Mesure de l’effet de lentille gravitationnelle dans le fond cos- 2012 IAP mologique avec le satellite Planck Anne Ducout Non gaussianités dans le fond diffus cosmologique observé 2012 IAP, Sorbonne- par le satellite Planck UPMC, Paris Clément Filliard Etude de la calibration et de la reconstruction de cartes du 2012 LAL, Univ. Paris- ciel pour les données Planck-HFI Sud, Orsay G. Hurier Observations de la toile cosmique par la mesure du rayon- 2012 LPSC nement fossile à 3K avec le satellite PLANCK J. Lanoux Impact de la poussière sur la formation et l’évolution des 2012 Toulouse grandes structures S. Puisieux "Etude des effets Sunyaev-Zel’dovitch cinétique et relativiste 2013 SPP dans les données Planck." L. Sanselme Observation du rayonnement fossile avec Planck : des 2012 LPSC mesures aux contraintes sur la réionisation Alexis Lavabre Détection de l’effet de lentille gravitationnelle dans les don- 2011 LAL, Univ. Paris- nées de Planck-HFI Sud, Orsay Gaël Roudier Contraintes sur la biréfringence cosmique à partir de 2011 APC Paris l’analyse des données polarisées du fond diffus cosmologique de Planck Andrea Catalano Développement de modèles numériques de l’instrument 2010 IAS Paris Haute fréquence (HFI) de Planck nécessaires à son exploitation L. Fauvet Cosmologie observationnelle avec le satellite PLANCK : 2010 LPSC Etude de la polarisation du fond diffus cosmologique et mod- élisation des émissions d’avant-plan polarisées

French roadmap for CMB science 30/06/2016 69

D. Girard Cosmologie observationnelle avec le satellite PLANCK : 2010 LPSC Extraction du signal astrophysique des données brutes et étude de l’énergie sombre N. Taburet L’effet Sunyaev-Zel’dovich dans l’analyse du fond diffus 2010 IAS, Univ. Paris- cosmologique Sud, Orsay Marc Bétoule Analyse des données du fond diffus cosmologique : simula- 2009 Paris tion et séparation de composantes Simona Donzelli Instrument and data-processing signature in CMB experi- 2009 Uni. Milano, ments: issues on non-Gaussianity science Univ. Paris- Diderot Marcella High frequency BOOMERanG-03 analysis and Planck-HFI 2009 APC Paris Veneziani calibration Pierrick Abrial Transformations multi-échelles sur la sphère et applications 2008 Paris Jérôme Bobin Diversité morphologique et analyse de données multivaluées 2008 Paris Frédéric Guilloux Analyse harmonique et estimation spectrale sur la sphère 2008 Paris J. Aumont Etude des différentes composantes de la polarisation du ciel 2007 LPSC en vue de l’observation du Fond Diffus Cosmologique avec le satellite Planck C. Leroy Contribution à l’étude des effets systématiques pour le traite- 2007 IRAP ment des données de l’instrument Planck-HFI Stéphane Bargot Archeops et la préparation de la mission Planck : 2006 LAL, Univ. Paris- Détermination des paramètres cosmologiques Sud, Orsay Ludovic Montier Planck : de l’étalonnage de l’instrument à l’étude des pous- 2005 Toulouse sières galactiques et intergalactiques M. Tristram Mesure des anisotropies primaires et secondaires du fond 2005 LPSC diffus cosmologique avec l’expérience Archeops et prépara- tion de Planck-HFI Alexandre Contributions à l’analyse des données dans l’expérience 2004 LAL, Univ. Paris- Bourrachot Archeops et à la mesure de la masse des neutrinos avec les Sud, Orsay expériences CMB Jean-Baptiste Amas de galaxies et effet Sunyaev-Zel’dovich: Observations 2004 Paris Melin et étude des effets de sélection dans les sondages Svitlana Zinger Interpolation et rééchantillonage de données spatiales et ap- 2004 Paris plications à la cartographie urbaine et à la détermination du fond cosmologique primordial Guillaume Analyse multi-composantes d’observations du fond diffus 2003 Paris Patanchon cosmologique N. Ponthieu Etude de la polarisation du fond diffus cosmologique et de la 2003 LPSC poussiere Galactique par l’expérience Archeops C. Rosset Contribution à la mesure de la polarisation du fond diffus 2003 APC cosmologique dans le cadre des programmes ARCHEOPS et PLANCK A. Amblard Analyse des anisotropies du fond diffus cosmologique dans 2002 APC le cadre de l’expérience Archeops Hichem Snoussi Approche bayésienne en séparation de sources. Applications 2003 Paris en imagerie Julien Brossard Couplage optique du telescope Planck avec l’instrument HFI 2002 IRAP, Toulouse : Modelisation et caracterisation Xavier Dupac Reconstruction optimale d’images bolométriques, applica- 2002 IRAP, Toulouse tions à PRONAOS, Archeops et Planck Ph. Filliatre Etalonnage sol et analyse des données de l’expérience bal- 2002 LPSC lon Archeops mesurant les anisotropies du Fond Diffus Cosmologique. Etude des contraintes sur l’inflation K. Madet Mesure du rayonnement cosmologique : Préparation et 2002 Néel étalonnage des instruments Archeops et Planck O. Doré Etudes autour des anisotropies du corps noir cosmologique 2001 IAP - Sorbonne- et des amas de galaxies UPMC, Paris Michel Piat Contributions à la Définition des Besoins scientifiques et des 2000 IAS, Univ. Paris- Solutions instrumentales du Projet Planck-HFI Sud, Orsay

French roadmap for CMB science 30/06/2016 70 A FRENCH CMB PHD THESES SINCE PLANCK INCEPTION

Benoit Revenu Anisotropies et polarisation du rayonnement fossile: méth- 2000 Paris odes de détection et traitement de données Etienne Contribution à l’étude des propriétés physiques du milieu in- 1999 Toulouse Pointecouteau tergalactique à partir d’observations allant de l’infrarouge au millimétrique Jacques Simulation and analysis of cosmic microwave background 1998 IAS Univ. Paris- Delabrouille anisotropy measurements Sud, Orsay Guilaine Lagache Emission en infrarouge lointain et submillimétrique: du mi- 1998 IAS Univ. Paris- lieu interstellaire aux galaxies lointaines Sud, Orsay Simon Prunet Polarisation du ciel micro-ondes 1998 IAS Univ Paris- Sud, Orsay S. Gaertner Système de lecture des bolomètres optimisé pour la mission 1997 Toulouse spatiale Planck Surveyor Nabila Aghanim Contribution à l’étude des anisotropies secondaires du fond 1996 IAS Univ. Paris- de rayonnement cosmologique Sud, Orsay

French roadmap for CMB science 30/06/2016 71

Appendix B: Elements on pre-S4 funding in the US This appendix gives the result of a search of the public NSF database for grants to BICEP, ACT and SPT. While this is a very partial view, it does give an idea of the order of magnitude of the pre-stage 4 eﬀort in the US. This does not include all ground-based experiments, and does not include balloons at all, which all fall in the budget range of tens of millions of US dollars (with EBEX at 9.4 M$, SPIDER at 10 M$, and PIPER at 12 M$.) It is even ∼ ∼ ∼ harder to get a full account of the very sizable private funding made available to these CMB experiments. Let us just at least mention the 2.3 million dollars for the Keck array announced at http://www.caltech.edu/article/13201, or the 5 million US dollars at http://ucsdnews.ucsd.edu/pressrelease/nsf_awards_simons_array_project_ 5_million_to_study_origins_of_the_universe, not to mention the recently announced 40 million from the Simmons foundation to help join the Simmons and ACT projects (https://simonsobservatory.org/).

ACT grants Collaborative Research with the Atacama Cosmology Telescope (ACT): Probing Fundamental Physics Through Observations of Cosmic Structure Award Number:0408698; Principal Investigator:Lyman Page; Co-Principal Investigator:; Organization:Princeton University;NSF Organization:AST Start Date:01/01/2004; Award Amount:$12,981,621.00; Relevance:83.35;

ACTPol: The Atacama Cosmology Telescope with Polarization Award Number:0965625; Principal Investigator:Lyman Page; Co-Principal Investigator:Suzanne Staggs, Joseph Fowler, David Spergel; Organization:Princeton University;NSF Organization:AST Start Date:09/15/2010; Award Amount:$9,400,000.00; Relevance:81.7;

Gravitational Physics from the Cosmic Microwave Background Award Number:0855887; Principal Investigator:Lyman Page; Co-Principal Investigator:Suzanne Staggs, Norman Jarosik, Joseph Fowler; Organization:Princeton University;NSF Organization:PHY Start Date:05/15/2009; Award Amount:$2,428,739.00; Relevance:81.17;

Gravitational Physics from the Cosmic Microwave Background Award Number:1214379; Principal Investigator:Lyman Page; Co-Principal Investigator:Suzanne Staggs; Organization:Princeton University;NSF Organization:PHY Start Date:06/15/2012; Award Amount:$2,190,000.00; Relevance:80.58;

Statistical Techniques for the Analysis of High Resolution CMB Data Award Number:0707731; Principal Investigator:David Spergel; Co-Principal Investigator:; Organization:Princeton University;NSF Organization:AST Start Date:08/01/2007; Award Amount:$465,142.00; Relevance:80.56;

Cosmology with the kinematic Sunyaev-Zel’dovich eﬀect: Testing fundamental physics and hydrodynamical simulations Award Number:1517049; Principal Investigator:Michael Niemack; Co-Principal Investigator:; Organization:Cornell University;NSF Organization:AST Start Date:09/01/2015; Award Amount:$143,714.00; Relevance:80.23;

Advanced ACTPol Award Number:1440226; Principal Investigator:Suzanne Staggs; Co-Principal Investigator:Lyman Page, David Spergel; Organization:Princeton University;NSF Organization:AST Start Date:09/15/2014; Award Amount:$7,569,979.00; Relevance:79.45;

Combining Thermal SZ and Gravitational Lensing Measurements: A Novel Approach to Measuring the Amplitude of Matter Fluctuations Award Number:1311756; Principal Investigator:David Spergel; Co-Principal Investigator:; Organization:Princeton University;NSF Organization:AST Start Date:09/15/2013; Award Amount:$539,488.00; Relevance:79.42;

New Science with ACTPol Award Number:1312380; Principal Investigator:Arthur Kosowsky; Co-Principal Investigator:; Organization:University of Pittsburgh;NSF Organization:AST

French roadmap for CMB science 30/06/2016 72 B ELEMENTS ON PRE-S4 FUNDING IN THE US

Start Date:09/15/2013; Award Amount:$247,154.00; Relevance:78.62;

Discovering Properties of Neutrinos, Inﬂation, and Dark Energy Using the Cosmic Microwave Background Award Number:1513618; Principal Investigator:Neelima Sehgal; Co-Principal Investigator:; Organization:SUNY at Stony Brook;NSF Organization:AST Start Date:09/01/2015; Award Amount:$560,000.00; Relevance:78.61;

MRI: Acquisition of a High-Performance Computing Cluster for Astrophysics Award Number:0722479; Principal Investigator:James Stone; Co-Principal Investigator:Jeremiah Ostriker, Bruce Draine, David Spergel; Organization:Princeton University;NSF Organization:AST Start Date:08/15/2007; Award Amount:$663,315.00; Relevance:78.61;

Optical Observations of the ACT Sky Region Award Number:0546035; Principal Investigator:Arthur Kosowsky; Co-Principal Investigator:; Organization:University of Pittsburgh;NSF Organization:AST Start Date:08/15/2005; Award Amount:$120,990.00; Relevance:78.61;

Collaborative Research: Polarization Sensitive Multi-Chroic MKIDs Award Number:1509078; Principal Investigator:Philip Mauskopf; Co-Principal Investigator:; Organization:Arizona State University;NSF Organization:AST Start Date:08/01/2015; Award Amount:$194,802.00; Relevance:77.77;

Collaborative Research: Polarization Sensitive Multi-Chroic MKIDs Award Number:1506074; Principal Investigator:Kent Irwin; Co-Principal Investigator:; Organization:Stanford University;NSF Organization:AST Start Date:08/01/2015; Award Amount:$320,816.00; Relevance:77.77; MRI: Development of a Sensitive Broadband Dual-Frequency Millimeter-Wavelength Polarimeter Award Number:1429236; Principal Investigator:Charles Bennett; Co-Principal Investigator:Tobias Marriage; Organization:Johns Hopkins University;NSF Organization:AST Start Date:08/15/2014; Award Amount:$2,030,147.00; Relevance:61.15;

EAPSI: Understanding Cosmic Dust in Lyman-Alpha Emitters Award Number:1515278; Principal Investigator:Alex Hagen; Co-Principal Investigator:; Organization:Hagen Alex R;NSF Organization:OISE Start Date:06/01/2015; Award Amount:$5,070.00; Relevance:60.29;

REU Site: National Radio Astronomy Observatory Summer Research Experience in Astronomy for Undergraduates Award Number:1358169; Principal Investigator:Alison Peck; Co-Principal Investigator:; Organization:Associated Universities Inc/National Radio Astronomy Observatory;NSF Organization:AST Start Date:05/01/2014; Award Amount:$522,107.00; Relevance:60.28;

RUI: Experiments on Cooled Cosmic Dust Analogs to Determine their Optical Properties in the Millimeter/Sub- Millimeter Award Number:1313261; Principal Investigator:Thushara Perera; Co-Principal Investigator:; Organization:Illinois Wesleyan University;NSF Organization:AST Start Date:08/01/2013; Award Amount:$178,261.00; Relevance:60.28;

Galaxies at Redshifts z 2: The Apex of Galaxy Formation Award Number:1009452; Principal Investigator:Desika Narayanan; Co-Principal Investigator:Lars Hernquist; Organization:University of Arizona;NSF Organization:AST Start Date:01/15/2011; Award Amount:$565,216.00; Relevance:57.68;

Galaxies at Redshifts z 2: The Apex of Galaxy Formation Award Number:1445357; Principal Investigator:Desika Narayanan; Co-Principal Investigator:; Organization:Haverford College;NSF Organization:AST

French roadmap for CMB science 30/06/2016 73

Start Date:12/31/2013; Award Amount:$541,076.00; Relevance:57.67;

Management and Operation of the National Radio Astronomy Observatory FY2010-2015 Award Number:0836064; Principal Investigator:Ethan Schreier; Co-Principal Investigator:; Organization:Associated Universities Inc/National Radio Astronomy Observatory;NSF Organization:AST Start Date:10/01/2009; Award Amount:$.00; Relevance:56.96;

SPT grants Award Number:8920223; Principal Investigator:John Carlstrom; Co-Principal Investigator:Richard Kron; Organization:University of Chicago;NSF Organization:PLR Start Date:02/01/1991; Award Amount:$37,710,839.00; Relevance:88.5;

Collaborative Research: Optical Conﬁrmation, Redshifts, and Richness for Galaxy Clusters Discovered with the South Pole Telescope Award Number:1009012; Principal Investigator:Christopher Stubbs; Co-Principal Investigator:; Organization:Harvard University;NSF Organization:AST Start Date:09/15/2010; Award Amount:$997,781.00; Relevance:86.45;

Collaborative Research: Optical Conﬁrmation, Redshifts, and Richness for Galaxy Clusters Discovered with the South Pole Telescope Award Number:1009649; Principal Investigator:Antony Stark; Co-Principal Investigator:; Organization:Smithsonian Institution Astrophysical Observatory;NSF Organization:AST Start Date:09/15/2010; Award Amount:$930,881.00; Relevance:86.45;

Collaborative Research: Building an Event Horizon Telescope: (Sub)millimeter VLBI from the South Pole Telescope Award Number:1207730; Principal Investigator:John Carlstrom; Co-Principal Investigator:; Organization:University of Chicago;NSF Organization:AST Start Date:08/01/2012; Award Amount:$120,057.00; Relevance:85.42;

Cosmological Research with the 10-meter South Pole Telescope Award Number:1248097; Principal Investigator:John Carlstrom; Co-Principal Investigator:John Ruhl, Lloyd Knox, William Holzapfel, Nils Halverson; Organization:University of Chicago;NSF Organization:PLR Start Date:09/01/2013; Award Amount:$9,892,153.00; Relevance:84.16;

Cosmological Research with the 10 meter South Pole Telescope Award Number:0638937; Principal Investigator:John Carlstrom; Co-Principal Investigator:William Holzapfel, Nils Halverson, Joseph Mohr, John Ruhl; Organization:University of Chicago;NSF Organization:PLR Start Date:10/01/2007; Award Amount:$17,449,975.00; Relevance:84.16;

MRI-R2: Development of a Polarimeter for the 10 Meter South Pole Telescope Award Number:0959620; Principal Investigator:John Carlstrom; Co-Principal Investigator:Stephan Meyer, John Ruhl, William Holzapfel, Nils Halverson; Organization:University of Chicago;NSF Organization:PLR Start Date:02/15/2010; Award Amount:$1,259,490.00; Relevance:83.53;

CAREER: Large focal plane array development for CMB polarization studies with the South Pole Telescope Award Number:0956135; Principal Investigator:Nils Halverson; Co-Principal Investigator:; Organization:University of Colorado at Boulder;NSF Organization:AST Start Date:09/15/2010; Award Amount:$875,145.00; Relevance:83.53;

Physics Frontier Center at the Kavli Institute for Cosmological Physics: Pushing Cosmology to the Edge Award Number:1125897; Principal Investigator:Michael Turner; Co-Principal Investigator:; Organization:University of Chicago;NSF Organization:PHY Start Date:09/01/2011; Award Amount:$17,000,000.00; Relevance:81.84;

Collaborative Research: Coordinated Surveys to Study the Nature of Dark Energy Award Number:0507562; Principal Investigator:Joshua Frieman; Co-Principal Investigator:John Carlstrom; Organization:University of Chicago;NSF Organization:AST

French roadmap for CMB science 30/06/2016 74 B ELEMENTS ON PRE-S4 FUNDING IN THE US

Start Date:07/15/2005; Award Amount:$25,000.00; Relevance:81.47;

Collaborative Research: Polarization Sensitive Multi-Chroic MKIDs Award Number:1509211; Principal Investigator:Bradley Johnson; Co-Principal Investigator:; Organization:Columbia University;NSF Organization:AST Start Date:08/01/2015; Award Amount:$298,114.00; Relevance:80.81; Collaborative Research: Polarization Sensitive Multi-Chroic MKIDs Award Number:1509078; Principal Investigator:Philip Mauskopf; Co-Principal Investigator:; Organization:Arizona State University;NSF Organization:AST Start Date:08/01/2015; Award Amount:$194,802.00; Relevance:80.81;

Galaxy Cluster Velocities and Cosmology Award Number:0807790; Principal Investigator:Arthur Kosowsky; Co-Principal Investigator:; Organization:University of Pittsburgh;NSF Organization:AST Start Date:07/01/2008; Award Amount:$174,453.00; Relevance:80.81;

BICEP grants Background Imaging of Cosmic Extragalactic Polarization (BICEP): An Experimental Probe of Inﬂation Award Number:0230438; Principal Investigator:Andrew Lange; Co-Principal Investigator:Brian Keating, William Holzapfel, James Bock; Organization:California Institute of Technology;NSF Organization:PLR Start Date:01/01/2003; Award Amount:$2,636,218.00; Relevance:96.83;

Additional Detectors for QUaD and BICEP Award Number:0634562; Principal Investigator:James Bock; Co-Principal Investigator:; Organization:California Institute of Technology;NSF Organization:AST Start Date:07/01/2006; Award Amount:$80,000.00; Relevance:93.7;

CAREER: The Birth Pangs of the Big Bang: Detecting Primordial Gravitational Waves with Microwave Background Imaging of Cosmic Extragalactic Polarization (BICEP) Award Number:0548262; Principal Investigator:Brian Keating; Co-Principal Investigator:; Organization:University of California-San Diego;NSF Organization:AST Start Date:03/01/2006; Award Amount:$490,593.00; Relevance:93.7;

Collaborative Research: Next Generation CMB Polarization Measurements with the QUEST Experiment on DASI Award Number:0338238; Principal Investigator:Clement Pryke; Co-Principal Investigator:; Organization:University of Chicago;NSF Organization:PLR Start Date:04/15/2004; Award Amount:$918,025.00; Relevance:84.33;

Collaborative Research: Next Generation CMB Polarization Measurements withthe QUEST Experiment on DASI Award Number:0338335; Principal Investigator:Andrew Lange; Co-Principal Investigator:James Bock; Organization:California Institute of Technology;NSF Organization:PLR Start Date:04/15/2004; Award Amount:$380,611.00; Relevance:84.33;

Collaborative Research: Next Generation CMB Polarization Measurements with the QUEST Experiment on DASI Award Number:0338138; Principal Investigator:Sarah Church; Co-Principal Investigator:; Organization:Stanford University;NSF Organization:PLR Start Date:04/15/2004; Award Amount:$659,914.00; Relevance:84.33;

Collaborative Research: Science Observation with BICEP3 CMB Polarization Experiment Award Number:1313062; Principal Investigator:James Bock; Co-Principal Investigator:; Organization:California Institute of Technology;NSF Organization:PLR Start Date:08/15/2013; Award Amount:$366,207.00; Relevance:81.22;

Collaborative Research: Science Observation with BICEP3 CMB Polarization Experiment Award Number:1313158; Principal Investigator:Clement Pryke; Co-Principal Investigator:; Organization:University of

French roadmap for CMB science 30/06/2016 75

Minnesota-Twin Cities;NSF Organization:PLR Start Date:08/15/2013; Award Amount:$303,448.00; Relevance:81.22;

Collaborative Research: BICEP2 and SPUD - A Search for Inﬂation with Degree-Scale Polarimetry from the South Pole Award Number:0742592; Principal Investigator:Clement Pryke; Co-Principal Investigator:John Carlstrom; Organization:University of Chicago;NSF Organization:PLR Start Date:05/15/2008; Award Amount:$874,342.00; Relevance:81.22;

Collaborative Research: Next Generation CMB Polarization Measurements with the QUaD Experiment Award Number:0638532; Principal Investigator:Clement Pryke; Co-Principal Investigator:; Organization:University of Chicago;NSF Organization:PLR Start Date:04/01/2007; Award Amount:$254,807.00; Relevance:81.22;

Collaborative Research: Next Generation CMB Polarization Measurements with the QUaD Experiment Award Number:0638615; Principal Investigator:Sarah Church; Co-Principal Investigator:; Organization:Stanford University;NSF Organization:PLR Start Date:04/01/2007; Award Amount:$149,315.00; Relevance:81.22;

Collaborative Research: Science Observation with BICEP3 CMB Polarization Experiment Award Number:1313010; Principal Investigator:Chao-Lin Kuo; Co-Principal Investigator:; Organization:Stanford University;NSF Organization:PLR Start Date:08/15/2013; Award Amount:$326,000.00; Relevance:81.21;

Collaborative Research: Science Observation with BICEP3 CMB Polarization Experiment Award Number:1313287; Principal Investigator:John Kovac; Co-Principal Investigator:; Organization:Harvard University;NSF Organization:PLR Start Date:08/15/2013; Award Amount:$400,104.00; Relevance:81.21;

Collaborative Research: BICEP2 and SPUD - A Search for Inﬂation with Degree-Scale Polarimetry from the South Pole Award Number:0742818; Principal Investigator:John Kovac; Co-Principal Investigator:Andrew Lange, James Bock; Organization:California Institute of Technology;NSF Organization:PLR Start Date:05/15/2008; Award Amount:$1,259,000.00; Relevance:81.21;

Collaborative Research: BICEP2 and SPUD - A Search for Inﬂation with Degree-Scale Polarimetry from the South Pole Award Number:1044978; Principal Investigator:John Kovac; Co-Principal Investigator:; Organization:Harvard University;NSF Organization:PLR Start Date:11/01/2009; Award Amount:$2,919,786.00; Relevance:81.21;

CAREER: Sharing Deep CMB Maps for Cosmological Discovery Award Number:1255358; Principal Investigator:John Kovac; Co-Principal Investigator:; Organization:Harvard University;NSF Organization:AST Start Date:03/01/2013; Award Amount:$804,614.00; Relevance:81.21;

Collaborative Research: BICEP2 and SPUD - A Search for Inﬂation with Degree-Scale Polarimetry from the South Pole Award Number:1110087; Principal Investigator:Clement Pryke; Co-Principal Investigator:; Organization:University of Minnesota-Twin Cities;NSF Organization:PLR Start Date:09/01/2010; Award Amount:$944,253.00; Relevance:81.21;

Collaborative Research: Next Generation CMB Polarization Measurements with the QUaD Experiment Award Number:0637420; Principal Investigator:Andrew Lange; Co-Principal Investigator:James Bock; Organization:California Institute of Technology;NSF Organization:PLR Start Date:04/01/2007; Award Amount:$100,000.00; Relevance:81.21;

Instrumentation and Laboratory Improvement Award Number:9751127; Principal Investigator:Art Hoﬀman; Co-Principal Investigator:Robijn Bruinsma; Organization:University of California-Los Angeles;NSF Organization:DUE

French roadmap for CMB science 30/06/2016 76 B ELEMENTS ON PRE-S4 FUNDING IN THE US

Start Date:07/01/1997; Award Amount:$53,043.00; Relevance:69.04;

French roadmap for CMB science 30/06/2016 77

Appendix C: Feuille de route du groupe “Futur des mesures du CMB” Suite à la non-sélection de CORE+ par l’ESA et face à la multiplication des propositions de nouvelles expériences soumises aux agences, il devient nécessaire de disposer d’une feuille de route dans ce domaine qui permettrait de définir une stratégie incluant les aspects sol et spatiaux (y compris le cas échéant ballon, la mesure de la polarisation des avant-plans étant un élément important de cette stratégie). Le mandat est donné par le CNES, l’INSU, l’IN2P3 et le CEA au Programme National Cosmologie et Galaxies d’établir les éléments de cette stratégie en s’appuyant sur la feuille de route des détecteurs millimétriques/submillimétriques et en intégrant la réflexion européenne en cours (http://indico.cern.ch/event/376392/overview). Il conviendra de traiter les points suivants: - Etat des lieux, incluant une estimation de la taille de la communauté française concernée, panorama international des moyens existants, des projets décidés et proposés, au sol et spatiaux, ayant pour finalité la mesure du CMB, avec une indication du calendrier pour les projets - Définition d’une stratégie scientifique : objectifs (avant-plans, CMB, échelles spatiales à atteindre, . . . ) avec des priorités et un calendrier cible de réalisation - Identification des projets qui pourraient être portés par la France, ou dans lesquels la participation française pourrait être importante, à court (< 5 ans), moyen (5-10 ans), et long terme, et identification d’un ou plusieurs scénarios possibles - Le cas échéant, actions préparatoires ou complémentaires à mener (R&D, organisation de la communauté, etc.). Compte tenu de la forte compétition internationale et de la perspective d’un call M5 très prochainement, cette feuille de route doit être établie rapidement. Des premiers éléments devraient être disponibles début octobre pour que le contexte dans lequel seront prises les décisions sur la suite du programme PILOT (revol en Australie, modification du plan focal) soit quelque peu éclairci. Une version finale est attendue fin 2015.

French roadmap for CMB science 30/06/2016