Cosmic ORigins Explorer (COrE )
A satellite mission for probing
Cosmic origins, neutrinos masses and
the Origin of Stars and Magnetic Fields
through a high-sensitivity survey of
the microwave polarization of the entire sky
This proposal involves the cosmic microwave background observation communities from France, Italy,
Spain, and the United Kingdom, as well as several other European countries (presently Denmark, Germany,
Ireland, the Netherlands, Norway, Portugal, Sweden, Switzerland. The US community as represented by
the PPPT groups and his chairman (S. Hanany) has expressed strong interest in an extensive collaboration
if this proposal is successful. A full list of the people currently involved with COrE may be found at our
website: http://www.core-missison.org as well as a list of prominent scientists who support our proposal.
Below we list those members of the COrE community who have contributed most actively in preparing this
proposal, under the working group where they have mainly contributed.
National coordinators:
F. R. Bouchet (FR), P. de Bernardis (IT), B. Maffei (UK), E. Martinez-Gonzalez (SP)
Mission & programmatics working group
F. R. Bouchet, P. de Bernardis, B. Maffei, M. Piat, N. Ponthieu, R. Stompor
Instrument working group:
B. Maffei, M. Bersanelli, P. Bielewicz, P. Camus, P. de Bernardis, M. De Petris, S. Masi, F. Nati, F. Piacentini, L. Piccirillo, M. Piat, G. Pisano, M. Salatino, R. Stompor, S. Withington, T. Peacocke
Science working group:
M. Bucher, D. Barbosa, N. Bartolo, R. Battye, J.-P. Bernard, F. Boulanger, A. Challinor, S. Chongchitnan,
S. Colafrancesco, T. Ensslin, J. Fergusson, P. Ferreira, K. Ferriere, F. Finelli, J. Garcia-Bellido, S. Galli,
M. Haverkorn, M. Hindmarsh, A. Jaffe, M. Kunz, J. Lesgourgues, A. Liddle, M. Liguori, S. Matarrese, A.
Melchiorri, P. Mukherjee, L. Pagano, D. Paoletti, H. Peiris, L. Perroto, C. Rath, J. Rubino Martin, C. Rath,
P. Shellard, J. Urrestilla, B. Van Tent, L. Verde, B. Wandelt
Foregrounds working group:
C. Burigana, J. Delabrouille, C. Armitage-Caplan, A. Banday, A. Bonaldi, S. Basak, D. Clements, G. De
Zotti, C. Dickinson, J. Dunkley, M. Lopez-Caniego, E. Martinez-Gonzalez, M. Negrello, S. Ricciardi, L.
Toffolatti
Contact for proposal:
François R. Bouchet, bouchet@iap.fr, +33 1 44 32 8095, and
Paolo de Bernardis, paolo.debernardis@roma1.infn.it, +39 06 4991 4271.
Executive Summary
We propose COrE , the Cosmic Origins Explorer, a Class M space mission whose aim is to deliver
high precision, reference-quality, full-sky maps of the polarized microwave and sub-mm sky in 15
bands ranging from 45 GHz to 795 GHz. Polarization maps in this frequency range will enable an
unprecedented exploration of Cosmic Origins—from the origins of stars and the origin of cosmic
structure to the origin of the Universe itself.
Owing to COrE ’s exquisite polarization sensitivity, systematics control, and foreground separation
capabilities, COrE will: improve sensitivity to the B-mode signal of primordial gravitational
waves by two orders of magnitude; vastly increase the number of resolved modes available to probe
the cosmological initial conditions, allow an unprecedented probe of non-Gaussianity and
thus either detect or rule out the presence of non-linear physics during inflation or in a pre-Big
Bang epoch; and measure neutrino masses to sufficient precision to distinguish the direct from
the inverted neutrino hierarchies.
At the higher frequencies COrE will observe all-sky polarization maps for the first time. These
maps will uncover the role played by magnetic fields in star formation; revolutionize our picture
of the three-dimensional galactic magnetic field and of the properties of interstellar dust;
determine directly the initial conditions for star formation in the diffuse ISM; and discover,
characterize and time-resolve a large number of new polarized galactic and extragalactic
point sources.
COrE builds on the success of Planck/Herschel both in terms of hardware and software/science, reusing
many of the subsystems and methods developed by the mm/submm community. In its study of the cosmic
microwave background anisotropies, COrE ’s polarization maps will be of comparable or better quality than
the temperature maps that are being delivered by Planck in spite of the considerably weaker polarization
signal. COrE will achieve this with comparable angular resolution and frequency coverage, but twice the
frequency resolution and more than 30 times higher sensitivity than Planck . This improvement is made
possible by a massive increase in the number of detectors, from 63 to 6384, operating almost at the photon
shot noise limit, as well as from improved detector technology and a longer mission duration. One of the
lessons of Planck concerns the abundance of cosmic rays at L2, which would make it hazardous to bank on
building much more sensitive detectors behind a colder telescope. Of course, sensitivity to cosmic rays will
be part of the criteria we shall use in the detailed definition phase for selecting the final detector arrays
set-up. COrE ’s raw sensitivity will be matched by a system-wide approach to reject systematic errors. This
concern is actually the primary driver of the selected overall design. Rather than measuring polarization by
subtracting two linearly polarized intensities from different detectors as in Planck , which introduces errors by
aliasing temperature anisotropies into a spurious polarization signal, COrE will modulate the polarization
signal by using a rotating half-wave plate as the first element of the optical path from the sky to the
detector horns. This allows polarization to be measured directly on a short time scale and rejects spurious
polarization that is inevitably to some extent introduced inside the telescope/detector system.
In the area of Early Universe science, COrE will search for primordial gravitational waves, predicted from
cosmic inflation but not yet seen, by searching for a B mode of the CMB polarization of primordial origin.
Simulations demonstrate that COrE ’s sensitivity and foreground rejection capabilities will lower the limits
on the tensor-to-scalar ratio (T /S) by approximately two orders of magnitude, from the T /S ≈ 0.1 detection
threshold anticipated from Planck to 0.001! In addition to providing a confirmation of inflation by verifying
its most stunning and non-trivial prediction, a positive detection of B modes of primordial origin would
establish the energy scale of inflation, which cannot be determined from observing the scalar anisotropies
alone. There exist broad classes of models in this range. A negative result, likewise, would rule out large
classes of models and provide invaluable constrains for string theory model building by probing energies near
the Planck scale.
In a complementary way to probing primordial gravitational waves, COrE ’s temperature and polarization
maps offer the prospect of dramatic new insights into fundamental physics and possible signals from the
epoch of quantum gravity through the window of non-Gaussianity. The temperature and polarization maps
delivered by COrE will enable it to distinguish between non-Gaussian fingerprints from a range of processes
1
in acting during the Big Bang or in a pre-Big Bang phase. This fingerprinting capability sets it apart from
all other projected probes of non-Gaussianity. This exploration may reveal the action of multiple fields when
the seeds of cosmic structure were produced (“local” NG); non-canonical kinetic energy (“equilateral” NG);
remnants of a pre-inflationary phase (“flattened NG”); cosmic (super-)strings; a contracting phase with a
subsequent bounce; and even signatures of dark energy imprinted by cosmic structure in the late Universe.
Another key science objective is to reconstruct the gravitational lensing deflection field and measure its
power spectrum to cosmic-variance limits on all scales where linear perturbation theory is reliable. While
Planck will be able to make rudimentary measurements of the lensing spectrum using the temperature maps,
high-resolution polarization maps with low instrument noise can give a vastly superior reconstruction by
eliminating most of the intrinsic fluctuations as a source of statistical noise. Through lensing reconstruction,
COrE will be able to probe sub-eV neutrino masses and the dark energy sector in a way that is impossible
with the primary anisotropies alone. For example, the summed neutrino mass will be measured to 0.03 eV.
Measurements of neutrino oscillations establish lower bounds on neutrino masses and suggest two possible
hierarchies for the neutrino masses, a normal and an inverted hierarchy. In both of these, COrE data
alone will determine the lightest mass to 0.04 eV (95% confidence) — a sensitivity that cannot be reached
with laboratory experiments. Moreover, COrE could distinguish the inverted hierarchy from three massless
neutrinos at 3 σ.
In the area of galactic science, COrE will probe the galactic magnetic field in diffuse regions, allowing us
to establish the initial conditions for star formation and probe the physics of the interstellar medium. COrE
will measure both dust and synchrotron intensity and polarization. The accuracy of COrE information
will provide a unique opportunity to test magneto-hydrodynamical models and the impact of the magnetic
field on the interstellar medium energetics, and the interplay between turbulence and magnetic field. COrE
will be the first instrument with the combination of sensitivity and resolution needed to map the Galactic
magnetic field down to the sub-parsec scales where pre-stellar cores are formed. The inefficiency of starformation is still an open question: how and when does gravity overcome thermal, turbulence and magnetic
pressure, starting star formation, needs to be tested with high sensitivity and resolution data that only
COrE can provide. The survey from COrE at large scales will provide important new clues to the effects
of the Galactic magnetic field in supporting the diffuse medium in the Galactic gravitational potential and
in confining cosmic rays in the disk, providing stronger constrains on current dynamo scenarios. The manyfrequencies survey of COrE will open a new dimension to our understanding of interstellar dust nature and
evolution, and of the underlying physical mechanisms processing dust grains in the ISM. COrE data will
also be used to disentangle if the Galactic Center “hazes” detected by WMAP and Fermi are due to dark
matter annihilation or to astrophysical mechanisms at the Galactic Center: the two models, in fact, predict
different polarization properties for the haze, which can be measured very accurately with COrE .
COrE cosmology science will require accurate foreground removal and a very good assessment of the
relevant potential error. COrE is designed to generate itself all the data needed for that, in particular
by a large complement of relatively narrow spectral bands. Using all available information as of now, as
summarized in Planck Sky Model, simulations were done and analyzed using state-of-the-art component
separation algorithms. They provide confidence that COrE will acquire the multi-component data necessary
in order to achieve its science objectives with some margin for error and redundancy for verifying its results.
All this science and more will be produced by the exploitation of the main legacy product of COrE
full-sky maps of the three Stokes parameters I,Q,U, at 15 frequencies, i.e. 45 maps of incredible sensitivity,
providing a data mine for decades afterwards.
The mission has been endorsed by the Italian Space Agency ASI and by the French Space Agency CNES.
2
Contents
1 Science with COrE
3
2 From science requirements to mission design
15
3 Mission profile proposed to achieve these objectives
18
4 Model payload to achieve the science objectives
20
5 Technology development requirements
29
6 System requirements and spacecraft key factors
31
7 Science operations and archiving
34
8 Programmatic and cost analysis
36
9 Communication and Outreach
38
10 Selected references
39
1
Science with COrE
Observations of the microwave and far infrared sky have revolutionized our understanding of the origins
of the Universe and the processes at play within our galaxy and beyond. Before proceeding to a detailed
analysis of some of the highlights of the new science that will be made possible with COrE, it is appropriate to
review what is known, mainly from temperature maps but also from the presently available polarization data,
concerning the several components emitting in the microwave bands and then quickly review the physical
mechanisms causing this emission to be partially polarized, with a view toward understanding qualitatively
the new science that can be extracted.
The frequency coverage of COrE is broad, deployed over 15 frequency channels ranging from 45 GHz
to 795 GHz. For the temperature and polarization anisotropies, the central channels from about 70–220
GHz offer a relatively clean window in which the primordial CMB anisotropies dominate. At lower frequencies, a nonthermal galactic synchrotron component becomes an increasingly dominant contribution to the
anisotropic signal. The synchrotron component arises from the interaction of cosmic rays with the galactic
magnetic field and provides invaluable information to constrain the galactic magnetic field. This component
has a spectral index redder than a blackbody by approximately three inverse powers in frequency, allowing
it to be removed through this differing frequency dependence. The WMAP data has provided a wealth
of information concerning the degree of polarization of this synchrotron emission, providing one of the inputs for the COrE polarization cleaning forecasts. Nevertheless a more precise template will be needed,
hence the presence of 45 GHz channel. At low frequencies, there is also a free-free component, arising from
bremsstrahlung of the hot HII regions, tightly correlated with Hα emission, which serves as a useful tracer.
This emission however is not polarized. At the low end of the spectrum one also has evidence for spinning
dust emission, but the details and degree of polarization of this emission are not presently well understood.
Above the so-called CMB channels, above around 250 GHz, a component of cold (T ≈ 20K) dust starts
to become the dominant contribution to the anisotropies as one pushes into the exponentially falling Wien
regime of the CMB blackbody spectrum. The physics of the dust grains of the interstellar medium is complex and only partially understood. Our understanding is summarized by empirical models of the grain
populations drawing from a variety of observational inputs (frequency dependence of reddening including
the presence of absorption features, polarization of starlight, infrared emission,...) as well as more physical
arguments (bottom-up modelling informed by cosmic element abundances). The infrared emission can be
described approximately with a blackbody at the dust temperature and an emissivity index in the neighborhood of α = 1.7. Presently the most widely used dust model has two dust components with separate
temperatures and emissivity indices.
All these emission processes (with the exception of free-free) are partially polarized. The degree of
polarization of the synchrotron emission is around 10% and that of the dust around 5–7% , although large
uncertainties persist. For the primordial component, polarization can heuristically be understood as a
measure of the CMB quadrupole moment as seen by a typical electron of last scattering, roughly proportional
to the double derivative of the ‘temperature’ and the square of the mean distance between last and next-to-last
scattering. This proportionality explains why the electric-type (E) polarization from the scalar perturbations
receives an appreciable contribution from late-time reionization despite the fact that the reionization optical
depth is quite small (τ ≈ 0.087). This is the mechanism that allowed the WMAP polarization data to break
the degeneracy between the amplitude of the primordial scalar perturbations A and the reionization optical
depth τ. Current observations of the polarization of the primordial CMB component are in good agreement
3
`(` + 1)C` /2π [µK 2 ]
TT,scalar
TE,EE scalar
BB← EE, scalar lensed
BB, (T /S) = 10−1
BB, (T /S) = 10−2
BB, (T /S) = 10−3
BB, (T /S) = 10−4
Multipole number (`)
Figure 1:
Inflationary prediction for the CMB temperature and polarization anisotropies for the
scalar and tensor modes. The horizontal axis indicates the multipole number ` and the vertical axis indicates
`(` + 1)C`AB /(2π) in units of (µK)2 , which is roughly equivalent to the derivative of the power spectrum with respect
to ln `. The green curves indicate the TT, TE, and EE power spectra (from top to bottom) generated by the scalar
mode assuming the parameters from the best-fit model from WMAP seven-year data (H0 = 71.4 km s−1 Mpc−1 ,
Ωb = 0.045, Ωcdm = 0.220, ΩΛ = 0.73, τ = 0.086, and ns = 0.969). The BB scalar component (indicated by the heavy
red curve) results from the gravitational lensing of the EE polarized CMB anisotropy at the last scattering surface
z ≈ 1100 by structures situated mainly around redshift z ≈ 2. The top three blue curves (from top to bottom on the
left) indicate the TT, TE, BB, and EE spectra resulting from the tensor mode assuming a scale-invariant (nT = 0)
primordial spectrum and a tensor-to-scalar ratio (T /S) of 0.1, and the solid black curves indicate the BB spectra for
the descending values of (T /S) = r = 0.1, 0.01, 0.001 and 0.0001. For the TE cross-correlations we have plotted the log
of the absolute value, hence the downward spikes which correspond to sign changes.
with the scalar perturbations and cosmological parameters implicit from the temperature perturbations.
Polarization provides a powerful cross check because it probes the other of the two quadratures of the
cosmological perturbations.
The frontier of CMB polarization observations lies in searching for the B mode. The polarization field
on the celestial sphere may be divided into two components: an E mode, which may be expressed by means
of second derivatives acting on a potential, and a B mode where this pattern is rotated by 45◦ . For scalar
perturbations, which are the only ones that contribute to the matter power spectrum, only the E mode
polarization is possible within linearized cosmological perturbation theory. Nonlinear corrections, which
may be calculated reliably, contribute a gravitational lensing background (shown in red in the figure) having
a white noise spectrum at low multipole number `. This ‘white noise’ at low ` has a magnitude of around
5µK · arcmin, the precise value depending slightly on cosmological parameters. Any B mode with a black
body spectrum beyond the level expected from lensing is the tell-tale sign of primordial gravitational waves
from inflation, whose multipole spectral shape predictions are shown in Fig. 1 alongside the predictions for
the scalar anisotropies, shown in green. The expected B-mode anisotropies from inflation is parameterized
by the ratio of the primordial tensor perturbations relative to the scalar perturbations (T /S) or r.
The current COrE concept does not attempt to ‘clean’ the gravitational lensing B mode, but rather
accepts it as a background that can be well characterized and included in the analysis in the same way as
one typically deals with instrument noise in CMB experiments. Hence science requirement is to deliver a
foreground cleaned map with an accuracy in the neighborhood or slightly better than 5µK · arcmin after
foreground subtraction. This requires a superior raw sensitivity in order to clean out the foregrounds.
The so-called “foregrounds” may, on the one hand, be considered a nuisance for primordial cosmology.
They are the ‘dirt’ that must be removed to gain a glimpse at the pristine state of the primordial universe.
But on the other hand, these foregrounds, which will be characterized with exquisite precision, constitute a
gold mine for galactic and extragalactic science. In particular, the dust polarization maps produced, when
combined with 21cm maps to provide depth information, can be used to gain a better understanding of the
galactic magnetic field, which according to equipartition arguments plays a key role in the dynamics of the
interstellar medium. Numerous extragalactic polarized point sources will be discovered as well.
The following sections describe in more detail some of the highlights of the new science that one will be
able to do with COrE and formulate the science requirements for the COrE instrument. Then the following
sections describe the instrument proposed to satisfy these science goals. Additional information and details
as well as a bibliography are available on the website : wwww.core-mission.org.
4
COrE will be competing with suborbital experiments likewise aiming to detect non-zero (T /S) and carry
out other components of the science program presented here. Suborbital experiments have indeed played
and will continue to play an important role is developing and demonstrating new technology for space-based
CMB observation. Nevertheless, suborbital experiment are substantially handicapped in a number of ways
that are analyzed in detail in the section Why Space? below.
1.1
Detecting gravitational waves from inflation: the quest for B-modes
One of the striking developments of the past few decades is the theory of cosmic inflation. There is compelling
evidence that the universe underwent a period of very rapid expansion at very early times. As a result of
inflation, the universe ended up in a special, almost perfectly homogeneous state with a spatial geometry
which is almost exactly Euclidean. Quantum fluctuations of the space-time fabric during inflation will
have led to the formation of the cosmic web of matter structures and of gravitational waves. Inflation has
been extraordinarily successful when confronted with the new spate of high quality cosmological data, most
notably the WMAP data and we are clearly at the cusp of understanding the workings of the early Universe.
The temperature and polarization of the CMB are the observables of choice for achieving this goal.
The simplest model of inflation is based on a single scalar field φ having a potential V (φ). The energy
and pressure of the scalar field drive the expansion of the universe according to the Einstein equations.
For inflationary expansion we take the pressure and density such that p ≈ −ρ ≈ −V (φ) and find that
the expansion accelerates, with the scale factor a(t), evolving as a(t) ∝ eHt where the Hubble constant,
H ≡ ȧ/a evolves as H 2 = (8πG/3)V . Scalar perturbations arising from inflation imprint irregularities in
the energy density of the Universe with an amplitude AS that depends on the wave number k, characterized
by a scalar spectral index nS ≡ 1 + d ln A2S (k)/d ln k. Both AS and nS depend on V , V 0 and V 00 . The
spectrum of density perturbations is almost, but not exactly, scale invariant, with nS ≈ 1. Inflation also
generates tensor perturbations, transverse traceless perturbations in the metric of space time that are called
gravitational waves, with an amplitude AT (k). The relative contribution of gravity waves to curvature
perturbations is given by the tensor-to-scalar ratio r = (T /S) ≡ 16(AT /AS )2 ≈ 8MP2 l (V 0 /V )2 . Using
the COBE normalization AS = 1.91 × 10−5 , we find that V 1/4 = 3.3 × 1016 (T /S)1/4 GeV. Hence a
measurement of (T /S) directly probes the energy scale of inflation and is a window onto the
early Universe.
There is a vast range of models of inflation. Single-field inflation satisfies a simple relation ∆φ/MPl '
(r/0.01)1/2 where ∆φ is the scale of variation in the inflaton field. For such models, for example for chaotic
inflation, large variations are required if they are to be detected. Small-field inflation models may be
constrained more effectively by the measurements of nS . String theory contains an abundance of light scalar
fields which may be responsible for inflation. Examples are the hybrid inflation models which involve two or
more scalar fields, the so-called curvaton model of multi-field inflation which can boost the value of r and
other candidates with non-standard kinetic terms in the Lagrangian leading to Dirac-Born-Infeld inflation.
The panorama of inflationary models is clearly vast and intricate. Yet it is possible, given current
cosmological constraints, to forecast the expected amplitude of r for specific classes of models. We do so
in the left hand plot in Fig. 2, where we consider power law, chaotic (where the potential is of the form
V (φ) ' φp ) and spontaneous symmetry breaking inflationary models and find current constraints on r and
nS . A more ambitious, models-independent approach is taken in the right hand plot of Fig. 2, which shows
extracts from a two million point Monte-Carlo simulation of the inflationary flow equations. In essence
one constructs a hierarchy of flow equations that should encapsulate all models of inflation which can be
constructed as effective field theories. The results do not depend strongly on choices of prior ranges once
known observational constraints (on ns and dns /d ln k) are imposed, leading to somewhat model-independent
constraints on r. While the results from these simulations cannot be interpreted in a statistical way, they show
that models do not cover the observable parameter space uniformly and they show significant concentrations
of points with significant tensor/scalar ratio, as can be seen in the middle panel of Fig. 2. It is therefore
clear that r ∼ 10−3 is a natural target for a CMB polarization experiment.
A Fisher forecast of the ability to constrain r and nS was carried out using the power spectra determination errors obtained from two independent component separation simulations based on the Planck Sky
Model and the COrE instrument specifications, as described further below. The two approaches yield very
similar results, offering an encouraging cross check. We find that (T /S) = 1 × 10−3 can be distinguished
from 0 at the two sigma level (σr = 5 × 10−4 ) after marginalizing over all other cosmological parameters. In
the left panel of Fig. 2 the small ellipsoids illustrate what a detection for a fiducial model with r = 5 × 10−3
would look like.
One advantage of searching for B modes from space is that two windows of detection are available.
There is a bump about two orders of magnitude above the otherwise white noise spectrum at very low-`
(i.e., ` <
∼ 15) due to reionization. This is known as the “reionization bump” and cannot be probed from
the ground, because the measurement of the lowest multipoles from the ground is plagued by systematic
errors. The other window is known as the “recombination bump” and its statistical weight is centered around
` ≈ 70. According to the forecasts for COrE, for r = 0.005 the statistical weight is divided approximately
5
0
10
0
10
Spontaneous
Symmetry
Breaking
−1
10
−1
r
10
power law
chaotic p=8
chaotic p=1
chaotic p=0.1
−2
r
10
−2
10
WMAP constraints
−3
10
COrE constraints
−3
10
Allowed region if no tensor
modes detected with COrE
−4
10
−4
10
Allowed region if no tensor
modes detected with COrE
−5
0.95
1
1.05
ns
10
0.01
0.1
1
10
Δφ / mpl
Figure 2: Constraints on inflation from COrE. For a broad range of inflationary models COrE can be
expected to detect primordial gravitational waves from inflation. The large contours on the left panel show
the present constraints from WMAP7 in the r−ns plane. A few parameterized families of inflationary models
give an idea of representative model predictions. The small contours illustrate what a COrE detection would
look like if r = 0.005. The part of parameter space still allowed at 2σ in the case of a non-detection is shown in
grey. The right panel shows the ‘main sequence’ of inflationary models generated using a model-independent
approach. A Melchiorri & L Pagano, H Peiris & L Verde
equally between these two windows and the double measurement will allow the tensor spectral index nT to
be determining thus providing an important cross-check.
1.2
Probing primodrial non-Gaussianity
It has often been said that “Gaussianity,” along with spatial flatness and an approximately scale invariant
spectrum of adiabatic cosmological perturbations, is one of the inexorable consequences, or tests, of inflation.
Indeed it is remarkable that given the precise measures of the CMB that exist today, the primordial signal
turns out to be very nearly Gaussian, although the WMAP data does suggest some hints of non-Gaussianity
at low statistical significance. Following a series of seminal papers where the bispectral non-Gaussianity
was calculated for the first time in 2003, it was realized that large classes of models predicting measurable
non-Gaussianity exist and much of the current research in theoretical cosmology has shifted toward exploring
what patterns of non-Gaussianity are possible. These possibilities are closely linked to new physics near the
Planck scale and modifications of gravity.
COrE will improve non-Gaussianity constraints and will have a non-Gaussianity discovery potential
(defined more quantitatively below) a factor of ∼ 20 better than Planck , approaching the capabilities of an
ideal probe. COrE ’s competitive strength will be a vastly enhanced exploration of physically predicted NG
shapes compared to any other projected probe of NG.
Constraining primordial non-Gaussianity stringently, into the interesting ranges, requires high resolution
and full-sky coverage. For non-Gaussianity it is roughly the number of resolution elements that determines
the sensitivity. For ‘local’ non-Gaussianity full-sky coverage is essential because most of the statistical weight
derives from the coupling of the largest-scale modes to the modes at the limit of the resolution of the survey.
Polarization is able to probe to smaller scales because the ratio of the contaminant signal to the primordial
signal at large-` is larger for the polarization than for the temperature. Systematics control, the environment
at L2 and full sky-coverage make COrE the ideal CMB NG probe.
Measuring the shape of non-Gaussianity. An important advantage of COrE (compared to non-CMB
probes of NG) consists in the ability to recognize a wide range of shapes of possible non-Gaussianity. Fig. 3
show the different bispectral shapes predicted by different theories. In this way COrE will directly probe and
if present recognize the action of multiple fields when the seeds of cosmic structure were produced (“local”
NG); non-canonical kinetic energy (“equilateral” NG); remnants of a pre-inflationary phase (“flattened NG”);
cosmic (super-)strings; a contracting phase with a subsequent bounce, etc. Furthermore, these signatures
can be separated robustly from late time astrophysical NG from lensing, foregrounds, and spurious NG, e.g.
from residual instrumental systematics. A detection of any such primordial signal would rule out standard
slow-roll models of single-field inflation. A detection of “local” NG would be inconsistent with any single-field
inflationary model.
Guaranteed CMB NG as probe of dark energy: Beyond searches for primordial NG, COrE will detect
late-Universe NG imprinted on the CMB maps by the lensing/ISW correlation. This signal can be exploited
to yield the strongest dark energy constraints from the CMB alone. There would also be ancillary signatures,
6
Figure 3: Distinct shapes for the CMB bispectra bl1 l2 l3 plotted with positive (cyan) and negative (magenta)
isocontours.Shown are (from left to right) the ‘local’ model (e.g., multifield inflation), equilateral model (e.g.,
DBI inflation) and cosmic strings. Fergusson & Shellard
allowing tests of the standard cosmology and constraining modified gravity alternatives.
NG is typically characterised by the parameters fNL , the bispectrum amplitude for ‘local” NG arising
in models with multiple fields, and gNL , and τNL , the trispectrum amplitudes. Joint constraints on τNL and
fNL can probe consistency conditions between the bispectrum and trispectrum rule out standard multifield
inflation τNL ≥ (6fNL /5)2 . Beyond these special cases, general estimators have now been developed to
efficiently search for arbitrary shapes in the CMB data, allowing any model or mechanism to be directly
tested. The COrE constraint on local fNL improves by a factor of 2.5 compared to Planck to δfNL ≈ 2.
We define a conservative figure of merit to compare the predicted impact of COrE on NG constraints (or,
equivalently, its discovery potential) for three forms of NG: local, equilateral and flattened. While Planck
will reduce the constraint volume by a factor of 70 compared to WMAP, COrE would improve it by yet
another factor of 20. Considering the constraint volume based only on the polarization maps (which provides
independent information to the temperature maps and can hence provide an important consistency check),
from Planck to COrE there is a factor 110 improvement. This improvement will be even more dramatic
when considering combined constraints on a larger range of NG shapes.
Important further information can be obtained combining the bispectrum with trispectrum constraints.
The forecast precision with which local trispectrum parameters could be measured with COrE temperature
data alone are ∆gN L = 3×104 and ∆τN L = 1×104 , which will be significantly improved when also considering
polarization data. There are also models with a negligible bispectrum and a large trispectrum, including
inflation with a parity symmetry. Similarly, for cosmic strings the trispectrum will provide a significantly
stronger constraint on the string tension than the power spectrum or bispectrum.
What if there is no detection of primordial NG? In this case COrE would rule out all scenarios whose
natural parameter space result in high values of the NG parameters such as fNL , while rendering many
other early universe mechanisms cosmologically irrelevant. Examples include DBI inflation models typically
predicting |fNL | > 5 and competitors to inflation like ekpyrotic/cyclic models predicting measureable |fNL | ∼
10.
In a complementary way to probing primordial gravitational waves, COrE ’s probe of NG offers the
prospect of dramatic new insights into fundamental physics and possible signals from the epoch of quantum
gravity.
1.3
Fundamental physics with gravitational lensing
The post-recombination cosmic history, and in particular “dark parameters” such as those describing the
dark energy and light neutrino masses (much below 1 eV), mainly affects the primary CMB anisotropies
through the angular-diameter distance to recombination and are therefore largely degenerate. While these
degeneracies can be partly broken with external datasets, a very promising alternative route is to use the
effect of weak gravitational lensing of the CMB by large-scale structure. The lensing deflection is a sensitive
probe of the growth of structure over a range of redshifts peaking around z ≈ 2. Lensing has several effects
on the CMB; the most powerful in terms of providing a new window to the physics of the dark sector is the
(small) non-Gaussianity that it induces in the observed CMB.
While lensing studies using the CMB have much in common with present and planned surveys of weak
lensing of galaxies, there are important differences. The CMB provides the most distant source plane and
so the deflections are larger and are sourced generally at higher redshift than for galaxy lensing. At higher
redshift, the amplitude of the fluctuations is lower and so a wider range of scales can be probed in the linear
regime, removing many of the issues of non-linear evolution that complicate galaxy lensing. Moreover, there
are no astrophysical issues such as intrinsic alignments of galaxy shapes to worry about. In addition, the
7
Figure 4: Left: Reconstruction noise on the deflection power spectrum for an extended Planck mission (four
surveys; left) and COrE (right) using temperature alone (red) and temperature and polarization (blue). For
COrE , we also show the approximate noise level (green) for an improved iterative version of the reconstruction
estimator following Smith et al. The deflection power spectrum is also plotted based on the linear matter
power spectrum (black solid) and with non-linear corrections (black dashed). Right: The resulting errors
(including cosmic variance) on the deflection power spectrum from Planck (red) and COrE (blue) using lens
reconstruction with temperature and polarization (no iteration).
statistics of the CMB source plane are well understood so that shear and magnification are equally useful
observables. On the down-side, the CMB is a single source plane so there is no possibility for performing
tomographic studies. Note also that instrumental systematic effects are quite different for the two routes.
The different redshift ranges and systematic effects make CMB and galaxy lensing highly complementary,
and cross-correlations should be able to provide interesting new results.
1.3.1
Measurement and interpretation of the CMB lensing signal
One can construct estimators for the lensing deflections from off-diagonal correlations in the observed CMB.
Once reconstructed, the deflection field can be used as another cosmological observable. High resolution is
essential since, even for large-scale lenses, the majority of the information in the reconstruction comes from
small-scale CMB fluctuations. A highly significant detection of the deflection power spectrum, Cldd , should
be possible with Planck’s temperature maps (see Fig. 4 for forecasts) and detailed analysis is underway.
Detections are also expected soon from the ground (e.g. with SPT). However, the statistical noise in temperature reconstructions is such that Cldd will only ever be measured to the cosmic variance limit for multipoles
l < 100. To do better, and hence increase the lever-arm with which to constrain dark-sector physics, one
must use polarization data. Planck is too noisy for polarization to add much, but with COrE it improves
the reconstruction greatly; see Fig. 4. Most of the information comes from the E-B correlations and the
reconstructed deflection power should be cosmic-variance limited to l ≈ 500, roughly a 25-fold increase in
the number of reconstructed modes with S/N > 1. Significantly, COrE can mine all of the information in
the deflection power spectrum where linear theory is reliable.
1.3.2
Neutrino masses as a unique probe of physics beyond the Standard Model
While the Standard Model as first proposed implies that there are three exactly massless chiral neutrinos,
an abundance of evidence for flavor oscillations now exists that requires neutrinos to have mass. Exploration
of the neutrino sector therefore provides truly telling clues on how the Standard Model should be extended.
The most recent data compilations indicate squared mass differences between the three mass eigenstates of
∆m212 = (7.59 ± 0.20) × 10−5 eV2 and ∆m231 = ±(2.43 ± 0.13) × 10−3 eV2 (1σ errors). However, oscillation
experiments do not probe the absolute mass scale of the neutrinos. Taking the charged leptons and quarks
as a guide, we might expect the neutrino masses to be in one of two possible hierarchies: a normal hierarchy
with m1 , m2 m3 and a total mass close to 0.058 eV; or an inverted hierarchy with m3 m1 , m2 and total
mass 0.1 eV. However, one should be mindful that neutrinos could well be degenerate with m1 ∼ m2 ∼ m3 0.05 eV. While laboratory β-decay experiments can probe effective masses, the target values for hierarchical
masses are well below the detection limit of current and future experiments.
Cosmology, however, provides an alternative probe of absolute neutrino masses. Mass splittings from
oscillation data imply that at least two mass eigenstates are non-relativistic today. These impede the growth
of density perturbations on small scales and the resulting mass-dependent suppression of the matter power
spectrum can be measured with CMB lensing even for masses close to the hierarchical targets. We illustrate
8
Figure 5: Simulated deflection power spectrum from COrE assuming an inverted hierarchy of neutrino
masses with the minimum total mass allowed by oscillation data (m1 ≈ m2 = 0.05 eV and m3 = 0 eV). In
the upper panel, the solid lines are the theory power spectrum for this scenario (lower) and for three massless
neutrinos (upper). The difference between these spectra is plotted in the lower panel illustrating how COrE
can distinguish these scenarios from Cldd in the range l > 200.
the capabilities of COrE to distinguish the minimal-mass inverted hierarchy from a model with three massless
neutrinos via its reconstruction of the lensing deflection field in Fig. 5. If all other parameters were known,
lensing with COrE could constrain the summed neutrino mass to 0.012 eV (1σ). However this ignores the
issue of uncertainties in the other parameters. While future distance-indicator measurements may help,
to be conservative here we consider only constraints from COrE alone in a joint analysis of the unlensed
temperature and E-mode polarization and the reconstructed deflection field. In our MCMC simulations, we
vary the standard seven parameters of flat wCDM models plus neutrino-mass parameters for three cases.
First we consider a minimal-mass normal hierarchy and use an oscillation prior (ignoring the errors in the
squared-mass differences) to fix m2 and m3 in terms of the lightest mass m1 which we allow to vary. Second
we assume a minimal-mass inverted hierarchy and vary m3 . Finally we assume degenerate neutrinos and
vary the total mass about a model with massless neutrinos. We summarize our results as follows.
• If neutrinos have hierarchical masses, COrE will bound the lightest mass to m1 < 0.034 eV (normal)
and m3 < 0.045 eV (inverted) at 95% confidence.
• The minimal-mass inverted hierarchy could be distinguished from a scenario with three massless neutrinos at the 3σ level.
• If neutrino masses are degenerate, COrE will measure the total mass to 0.03 eV (1σ error). For
comparison, the error expected from the Planck nominal mission (including lensing) is 0.10 eV.
Current constraints on neutrino masses from cosmology are rather model dependent and the tightest
constraints come from combining multiple datasets (e.g. CMB and large-scale structure data). The 95%
limits on the total mass are in the range 0.3–2 eV. In models in which w 6= −1, CMB and large-scale
structure data indicate an upper limit at the 1 eV level. Our forecasted constraints from COrE are for
a single homogeneous dataset and are much less model dependent than with current data since the full
shape information in the deflection power spectrum is accessible. For example, Smith et al. show
Pthat for
a CMB lensing experiment comparable to COrE , there are no significant degeneracies between ν mν , w
and the spatial curvature. Our constraints from COrE are comparable to forecasts for tomographic studies
with future galaxy lensing surveys (assuming a Planck prior), for example LSST or Euclid, but with quite
different systematics.
9
1.4
Challenging the cosmology paradigm: ancillary cosmological science
1.4.1 Parameters
Even though the main design driver of COrE is to provide ultrasensitive measurements of the CMB polarization in order to detect tensor modes from inflation, the improvement in the E polarization power spectrum
and the T power spectrum at large multipole number ` will as a by-product improve the determination of the
cosmological parameters typically by a factor 2–3 and thus reduce the allowed volume in parameter space
even more significantly. COrE will tell us whether the simple models that today fit so well will continue to
agree with the data or be ruled out. COrE will also constrain more general models such as those with isocurvature modes, non-standard helium fraction, extra relativistic species, etc. In particular, the constraints on
nS and τ will improve by approximately a factor of two.
1.4.2 Reionization history of the universe
Reionization after the dark ages comes from hard radiation emitted by the very first structures formed
in the high-redshift Universe, either by very massive stars or by quasars. The current model-independent
uncertainty on the optical depth, τ , from WMAP is δτ ∼ 0.015. While one expects Planck to find a definitive
constraint on τ , a detailed characterization of the ionization history will be lacking. COrE will, through an
improved measurement of the E-mode in the ` ∼ 10 − 20 range, pin down the possible evolution histories at
high redshift, identifying whether reionization was abrupt, prolonged are underwent various eras. COrE’s
arcminute resolution will also allow a detailed characterization of the morphology of reionization, measuring
the spatial variations in the ionization fraction. The signal that arises from the interaction of the ionizing
regions is expect to peak at ` ∼ 2000.
1.4.3
Primordial magnetic fields
Synchrotron emission and Faraday rotation measurements provide increasing evidence that the large scale
structures in the Universe, such as galaxies and clusters, are pervaded by magnetic fields which are of order a
few micro-Gauss. An open question is whether these fields were seeded in the Early Universe (by fields of a few
nano-Gauss that arise in inflation or cosmological phase transitions) or at a later date through astrophysical
processes. Either way, they would have sourced CMB anisotropies leading to unique observational signatures.
Very large scale, quasi-homogeneous magnetic fields lead to very distinct cross correlations between E and
B modes. Smaller scale, stochastic magnetic fields would lead to a characteristic peak in the B-mode power
spectrum at ` ∼ 2000. COrE will detect magnetic fields at the sub nano-Gauss level permitting an accurate
characterization of their origin and evolution.
1.4.4 Topological defects
Topological defects arise generically as the result of spontaneous symmetry breaking, both in condensed
matter physics and in theories of physics at high energies relevant in the early universe. It is therefore
quite conceivable that defects could have been produced in the early universe, and in fact defects offered a
paradigm for structure formation competing with inflation before precision CMB observation showed that
defects could account for at most a small part of the matter power spectrum. Nevertheless topological
defects such as cosmic strings or textures can be formed at the end of hybrid inflation and have observable
consequences. They would contribute with similar amplitudes to all scalar, vector and tensor components of
E- and B-mode CMB polarization anisotropies. Their main distinguishing feature would be a reionization
bump at low ` and a white noise power spectrum rising to a peak somewhere between ` = 500 and 1000.
The amplitude of CMB anisotropies from defects is of order Gµ ∼ (η/MP )2 ∼ 10−6 , for a symmetry breaking
transition near the GUT scale, η ∼ 1016 GeV. Present bounds from WMAP-5yr require Gµ/c2 < 7 × 10−7 .
COrE, for both strings and textures, for a non-detection would lower the bound to Gµ/c2 < 2 × 10−8 . This
would allow us to explore and constrain a wide range of particle physics theories near the GUT scale.
1.5
1.5.1
The galactic interstellar medium
The Galactic magnetic field
The energy density in the magnetic field in the Milky Way is comparable to energy densities in cosmic rays
and in turbulent gas, indicating that these three components are in continuous interaction and providing
feedback to each other. The magnetic field is key to interstellar medium (ISM) energetics, star formation
and galaxy evolution, but its actual role remains quantitatively debated due to the paucity of data on the
field structure and its correlation with the gas dynamics.
Dust and synchrotron radiation from the Galaxy give the dominant contributions to the microwave sky
polarization. The two emissions provide complementary perspectives on the magnetic field structure. Synchrotron radiation traces the field over the whole volume of the Galaxy, while dust polarization traces the
field within interstellar clouds. COrE will produce the first all-sky maps of these two complementary tracers
of the field, and so constrain its orientation and strength with the resolution and sensitivity needed to bring
a complete picture of Galactic magnetism. This perspective holds promises of exciting steps forward in our
10
understanding of an elusive, but dynamically important, component of galaxies. We outline here three prime
questions we will be able to address with the COrE survey in a unique way.
What is the impact of the magnetic field on the interstellar medium energetics?
Synchrotron emission is a prime tracer of galactic magnetic fields. The synchrotron intensity allows
estimation of the total magnetic field strength, while its polarization fraction gives information about the ratio
of ordered to total field. The COrE data will be an eminent complement to upcoming Faraday tomography
measurements at lower frequencies with LOFAR, eVLA, ASKAP, and SKA. At the 45 GHz frequency of
COrE, Faraday rotation is negligible and the field orientation is directly related to the emission polarization.
Fluctuations in the synchrotron flux in intensity and polarization provide insight into magnetic turbulence. Several characteristic statistical properties are encoded in these data, for example the magnetic energy
and helicity spectra, as well as the spectrum of the magnetic tension force. The accuracy of the COrE information will thus provide a unique opportunity to study magneto-hydrodynamical turbulence and dynamo
action, which govern magnetic field amplification, in great detail. It will drastically increase the spectral
range of accurately probed magneto-hydrodynamical modes. Therefore, the detection potential for relevant
plasma physical processes and their characteristic scales such as turbulent energy injection and dissipation
wil be considerably increased. The interplay between turbulence and magnetic fields is also a key to star
formation.
Role of the magnetic field in structuring the interstellar matter and on star formation?
The COrE survey will provide the combination of sensitivity and angular resolution required to continuously map the Galactic magnetic field within interstellar clouds, over their full extent down to the sub-parsec
scales where pre-stellar cores are formed. No other experiment offers such a capability. Planck data will
provide much information on the Galactic magnetic field on large scales, but does not have the required sensitivity to map the field across the diffuse interstellar medium. Upcoming balloon-borne experiments such as
Polar-BLAST and PILOT do not either have the required sensitivity. Stellar polarization observations will
continue to develop, but they are intrinsically limited to a discrete set of sight lines. Ground based telescopes
at sub-mm and millimeter wavelengths including ALMA will image at sub-arcminute angular scales a variety of compact sources including pre-stellar condensations, but they cannot map polarization from diffuse
emission.
Star formation occurs as a result of the action of gravity, which is counteracted by thermal, magnetic and
turbulent pressures. In the diffuse interstellar medium, outside star-forming molecular clouds, the kinetic
energy from interstellar turbulence and magnetic energy are comparable. Both are much larger than the
clouds’ gravitational binding energy and the gas internal energy. For stars to form, gravity must become,
locally, the dominant force. This occurs where the turbulence energy has dissipated and matter has condensed
without increasing the magnetic field flux in comparable proportions. How and how frequently does this
occur? This question is key to our physical understanding of what regulates the efficiency of star formation.
Star formation is observed to be an inefficient process. The first scenario proposed to regulate star formation combines long cloud lifetimes and a low star formation efficiency. In this picture, molecular clouds are
prevented from collapsing on large scales by turbulent and magnetic pressures, while star formation is locally
controlled by the rate at which material can cross field lines through ambipolar diffusion. Observations and
numerical simulations are challenging this view. It is increasingly considered, but still debated, that the
bottle-neck of star formation is the formation of magnetically supercritical condensations where magnetic
pressure is too weak to balance gravity. Once they are formed these condensations rapidly collapse to form
stars. Star formation would be generally inefficient, because only a small fraction of the matter reaches this
stage within the cloud lifetimes. Within this second paradigm, molecular clouds are transient, dynamically
evolving gas concentrations produced by compressive motions out of diffuse interstellar matter. This stresses
the importance of the COrE survey to understand how the interplay between turbulence and magnetic field
sets the initial conditions of star formation within the diffuse interstellar medium.
Impact of the magnetic field on the large-scale structure and evolution of the Galaxy?
COrE is also expected to provide new important clues to the effects of the Galactic magnetic field at
larger scales. We know that the magnetic field helps to support the interstellar gas against its own weight
in the Galactic gravitational potential and that it confines cosmic rays to the Galactic disk. In this manner,
both the magnetic field and cosmic rays partake in the overall hydrostatic balance of the interstellar medium,
causing the gaseous disk to become much thicker. In turn, the thickening of the disk tends to make it unstable
to the Parker instability (a type of magnetic Rayleigh-Taylor instability). When this instability develops,
magnetic field lines ripple, and the interstellar gas slides down along them toward the magnetic troughs,
where it accumulates. The whole process, it has been suggested, could give birth to new molecular cloud
complexes and ultimately trigger star formation.
The magnetic field is also believed to play an important role in disk-halo interactions, although the exact
mechanisms involved are still poorly understood. A priori, one expects the magnetic field to impede mass
exchange between the disk and the halo. For instance, the magnetic field tends to contain super-bubbles and
prevent them from venting gas into the halo. On the other hand, it has been suggested that the magnetic
field might stabilize in-falling clouds and enable them to reach the disk without being disrupted by their
dynamical interaction with the hot halo gas. To test these different ideas, it is crucial to gain a better
observational knowledge of the Galactic magnetic field, particularly in the halo. Similarly, a better knowledge of the Galactic magnetic field, both in the disk and in the halo, will make it possible to place stronger
constraints on the existing dynamo scenarios.
11
Figure 6: The COrE data will allow us to map the turbulent component of the Galactic magnetic field, and thereby to
characterize the role it plays in shaping the diffuse Interstellar medium, and regulating star formation. The interstellar
medium filamentary structure is illustrated here with HI spectroscopic observations obtained with the Dominion Radio
Astrophysical Observatory. The three images correspond to three different velocity channels of one 4◦ × 4◦ field (data
from the Planck Deep Field HI survey, PI Peter Martin).
1.5.2
The polarization properties of interstellar dust
The unique combination of spectral and spatial information provided by the COrE survey will open a new
dimension to our understanding of dust, its nature and evolution within interstellar space. Understanding
interstellar dust evolution is a major challenge in astrophysics underlying key physical and chemical processes
in interstellar space.
The Galactic interstellar medium is dusty. Large dust grains (size > 10 nm) dominate the dust mass.
Within the diffuse interstellar medium, they are cold (∼ 10 − 20 K) and emit at far-IR to millimeter wavelengths. Dipolar emission from small carbon dust particles is a main emission component, often referred to as
the anomalous microwave emission. Both dust components are relevant to CORE observations. Interstellar
dust is not star dust. Behind this statement lies the idea that the composition of interstellar dust reflects the
action of interstellar processes, which contribute to break and re-build grains over timescales much shorter
than the injection timescale by stellar ejecta. While there is wide consensus on this conclusion, the processes
that drive dust evolution in space are still poorly understood.
The polarization of dust emission is related to grain properties (size, shape and magnetic susceptibility)
and the efficiency of grain alignment, both of which depend on local physical conditions (gas density, radiation
field). The physics of dust evolution raises several questions that we need to answer in order to characterize
the spectral dependence of dust polarization and achieve the required accuracy in component separation for
B-mode detection. Which dust components are polarized? How to account for grain alignment, and what is
its dependence on local physical conditions? This last question is also key to understand dust polarization
as a tracer of the magnetic field structure within interstellar matter. The wide spectral coverage and the
large number of spectral bands are unique assets of the CORE project to answer these questions.
1.5.3 The Galactic Center
Non thermal emission in the galactic center
The center of the Milky Way is the locus of high energy processes and a source of high energy cosmic
rays which interact with the local magnetic field as well as with local clouds of interstellar matter.
The polarization measurements of the plasma produced and diffusing from the Galactic Center (GC)
region can set crucial constraints on the nature of this non-thermal emission in the GC region, on the largescale structure of the magnetic field stretched along the out-flowing regions (thought to have an intensity
∼ 50 − 500 µG), on the sources of cosmic rays (likely SNRs), on the properties of the diffusive transport
(and on its effective time-scales, in comparison with the advective ones), and on the relative importance
of hadronic and leptonic processes for the production and acceleration of non-thermal particles in the GC
region.
Evidence for dark matter annihilation?
Dark Matter (DM) particle (χ) annihilation followed by secondary particle diffusion in the magnetic
field of the Galactic Center (GC) region could produce detectable γ-ray emission (by χχ → π 0 → γ + γ)
together with a diffuse halo (so called “haze”) of synchrotron and inverse Compton scattering of background
photons by energetic electrons (produced by χχ → X + π ± → X + e± ). The intensity, spatial extension
12
and spectrum of this emission would depend on the specific details of the DM particles composition and
annihilation cross-section, as well as on the Galactic magnetic field establishing the diffusion properties of
the secondary particles produced in DM annihilation.
The existence of an anomalous microwave “haze” in the WMAP data, coinciding with gamma-ray bubbles
observed by Fermi-LAT, centered around the GC has been largely debated. The interpretation is still
controversial: whereas emission from DM secondary particles has been originally invoked to explain the
anomalous WMAP microwave“haze”, this interpretation is at variance with an astrophysical scenario, where
the putative “haze” would be associated with star formation in the GC region, producing a fast wind with
substantial quantities of cosmic-ray ions that diffuse out to large angular scales on the sky.
A wide-frequency and high-sensitivity polarization survey of the GC region with COrE will provide an
unprecedented view of the GC region, and may lead to the discovery of an unambiguous microwave “haze”
associated with DM annihilation. Polarization is a key to disentangle among possible interpretations: while
strong polarization features (on both synchrotron and inverse Compton emission) are expected in a scenario of
SNe driven outflows and winds, a DM-produced microwave “haze” is expected to have quite low polarization,
because the secondary electrons are produced in-situ along the DM density profile of the Galaxy, and are
subject to a smooth and rather isotropic spatial diffusion in the GC magnetic field.
1.6
Characterizing extragalactic sources
Figure 7: Expected integral counts, as a function of the polarized flux, of radio sources and of different populations
of dusty star-forming galaxies, at 105 and 555 GHz.
Measurements of the polarization properties of radio sources at mm and sub-mm wavelengths provide
important insights into the physical properties (and in particular into the structure of magnetic fields) of the
innermost compact regions of relativistic jets in AGNs. These measurements will be also essential to assess
the point sources contamination of CMB polarization maps. In fact, at high Galactic latitudes, radio sources
◦
are expected to be the main contaminant on scales <
∼0.5 up to frequencies of a few hundred GHz.
The only blind polarization surveys presently available are those produced by WMAP. 22 objects have
detected in polarization in at least one WMAP channel, five of which, however, are doubtful. Planck can
double the number of detections, but the sample will still be too limited for a meaningful study of polarization
properties of the various radio source populations.
Ground based studies of high-frequency polarization properties of radio sources have been carried out by
following-up samples drawn from surveys in total intensity. In spite of recent efforts, polarimetric data at
mm wavelengths are limited, and essentially non-existent at sub-mm wavelengths.
The most extensive polarization information on complete samples selected at high radio frequencies was
obtained at ' 20 GHz. Radio sources are found to be significantly polarized (up to >
∼30%, although the
median polarization degree is approximately 3%).
Polarized emission in giant radio galaxies is an exceptional tool to understand the origin of relativistic
particles, the ultra high-E cosmic rays, and magnetic fields. In regions of the RG lobes where shocks are
strongest, the inflation of the lobe is likely accelerating particles to very high energies, producing also quite
flatter synchrotron spectra, and relatively high-polarization (20-40%) degrees that could be observed with
COrE in many nearby objects. In addition, Inverse Compton scattering of CMB photons in giant radiogalaxy
lobes produce an SZ effect. whose spectrum (both in intensity and polarization at a level of 10-20 %) is best
observed at high frequencies.
Very little is known about the polarization of dusty galaxies, that dominate the counts in total intensity
above ' 100–200 GHz, but there are indications that it is probably ≤ 1% .
The spectacular sensitivity of COrE will allow us to obtain the first large sample of sources blindly
selected in polarization. Source confusion is far less of a problem in polarization than in total intensity.
Adopting an average polarization degree of 3% for radio sources and of 1% for dusty star-forming galaxies,
we find that the confusion noise is always sub-dominant and becomes negligible above 100 GHz, at least in
13
Figure 8: H-ATLAS 350µm SDP field (14.4 deg2 ) smoothed to COrE resolution. Among the brightest sources there
are five strongly lensed high redshift (z ∼ 2–3) dusty galaxies, all potentially detectable by COrE . COrE will select
such extremely interesting sources over most of the whole sky.
high Galactic latitude regions where polarized diffuse emissions are low, and provided that the CMB can be
efficiently removed.
As illustrated by Fig. 7, at 105 GHz we expect to detect the polarized flux of ' 320 radio sources per
2
2 )1/2 ' 3 mJy. At higher frequencies also the polarized
sr above the 5σ detection limit Slim = 5(σnoise
+ σconf
dust emission from dusty star-forming galaxies should be detectable. Under the above assumptions for the
mean polarization degree, we have ' 12 radio sources and ' 5 normal spiral galaxies per sr with polarized
flux larger than the 5σ detection limit of ' 32.5 mJy at 555 GHz. At 795 GHz the 5σ detection limit is
' 48.5 mJy and the expected numbers of detections per sr are of ' 6 radio sources and of ' 16 dusty galaxies.
1.6.1
CORE and non-polarized point sources
In total intensity COrE will be so sensitive that that it will reach the source confusion limit. We can use
galaxy counts from large area Herschel surveys at COrE wavelengths to determine its confusion limits. For
a beam FWHM of 1.30 at 795 GHz and of 1.90 at 555 GHz, we find 5σ detection limits of ' 96 mJy and
of 89 mJy respectively, if we take into account only the Poisson fluctuations. If we allow for the effect of
clustering the 5σ confusion limits increase to 119 mJy and to 127 mJy at 795 and 555 GHz, respectively.
Surveys of small areas of the sky are already available at 350µm (857 GHz) with somewhat better angular
resolution and depth from Herschel . In Fig. 8 we show the 4◦ × 4◦ region surveyed by H-ATLAS, smoothed
to the 1.3 arcmin FHWM beam of COrE at 795 GHz. Amongst the brightest objects in this H-ATLAS
image are five strongly lensed background galaxies lying at z ∼ 2–3, which can easily be identified by their
red far-IR color and absence from radio surveys. COrE will be able to use its highest frequency channels
to provide a similar selection of high redshift lensed far-IR sources, but this will be possible over not the
limited areas covered with Herschel , but over most of the sky.
We estimate that COrE will be able to detect at 795 GHz ' 1.3 × 104 strongly lensed high-z dusty
galaxies above |b| ∼ 20◦ , where Galactic confusion should be of relatively minor importance. In addition,
it will detect, at the same frequency, ' 4 × 104 normal and starburst galaxies and ' 540 unlensed high-z
proto-spheroidal galaxies.
1.6.2
The Sunyaev Zel’dovich effect in galaxy clusters
Comptonization of CMB photons by electrons in the intra-cluster plasma (the Sunyaev-Zel’dovich effect SZE) is a powerful probe of the energetics, spectra and stratification of the electron distribution in clusters.
Due to its redshift-independent nature, the SZE is also a powerful cosmological probe. Precise observations
of the SZE over a wide ν range are crucial to unveil the structure of cluster atmospheres because the high-ν
(ν ≥ 350 GHz) part of the SZE spectrum is more sensitive to the relativistic effects of the electron distribution.
SZE observations in the millimeter range are becoming available with the advent of HERSCHEL-Spire (in the
∼ 600 − 1200 GHz range) and PLANCK-HFI (in the ∼ 100 − 850 GHz range). However, HERSCHEL lacks a
spectral coverage in the low-ν part of the SZE spectrum that is important for an unbiased determination of
the main cluster parameters (density, temperature and peculiar velocity), while PLANCK-HFI suffers from
14
a limited sensitivity and spatial resolution in order to determine and resolve the cluster parameters for the
majority of medium- and small-mass systems.
COrE will provide a substantial improvement in the study of the SZE in galaxy clusters: i) it will provide
a full coverage of the SZE spectrum with a sensitivity that allows a systematic study of the thermal, multitemperature and non-thermal properties of many galaxy clusters; ii) it will provide a spatially-resolved
spectral study of many nearby clusters especially at high-ν where the relativistic effects of single- and multitemperature, or supra-thermal, plasma are relevant; iii) it will provide relevant constraints on the polarization
of the thermal SZE in many nearby, hot clusters; iv) it will provide a more detailed description of the mm.
point-like source contamination for the determination of the spectral and spatial properties of the cluster
SZE (and of its polarization) especially in the frequency range probed by COrE.
2
2.1
From science requirements to mission design
Sensitivity
The primary science goal of COrE is to investigate the physics at work in the early universe, and to test
extensively the standard cosmological model through high precision mapping of the CMB temperature and
polarization anisotropies. Just as with previous CMB missions, this requires the sensitive measurement of
sky emission in the millimeter wave domain where CMB emission is the strongest. It also requires additional
channels to constrain and understand all astrophysical emissions which contribute to the total sky emission.
The level of polarization of CMB B modes from gravitational lensing, which is both an important scientific
objective and an ultimate limitation for measuring primordial tensor B modes, is the main driver for the
sensitivity of COrE. It sets the requirement that the COrE mission should measure sky polarization with a
sensitivity better than 5µK·arcmin in a set of frequency channels at the maximum level of CMB emission
relative to foregrounds.
The baseline COrE design includes six frequency channels between 75 and 225 GHz with each a sensitivity better than this requirement. Not only this allows for the target sensitivity, but it also provides the
redundancy to check that measured fluctuations indeed have the expected color of CMB fluctuations.
2.2
Frequency bands
COrE is also designed as a comprehensive explorer of all astrophysical emissions in a broad frequency range
from 40 to 800 GHz. For each pixel of the sky, it is essential that the number of independent measurements
be sufficient to constrain all parameters of all the relevant emissions. For significantly polarized emissions,
synchrotron is represented with 2 or 3 parameters (amplitude, spectral index, and running due to the aging
of the relativistic electrons), CMB with 1 parameter (temperature), and thermal dust is usually fit with 3
to 6 parameters (amplitude, spectral index, and temperature for one or two greybodies). The total of 6-10
parameters requires at least as many frequency channels for a meaningful fit.
When dealing with intensity maps, the spectral coverage must also allow for fitting unpolarized emissions:
free-free, spinning dust, thermal Sunyaev Zel’dovich effect, and molecular lines, with one or more parameter
for each. Contrarily to the polarization case however, for which external data sets are essentially non-existent
at the required sensitivity level, Hα maps, H2 maps, X-ray observations, and optical surveys are available as
tracers of total emission, and can help interpreting observations in total intensity.
Hence, we require a conservative four to five additional parameters for adequate frequency coverage in
temperature, for a total of ten to fifteen parameters to fit unpolarized emission.
In order to secure the capability to measure as many parameters as may be required to fit and interpret
the data, we require the COrE instrument to observe the sky in 15 different frequency bands. This is also
useful to avoid using the CMB sensitive channels as foreground monitors in polarization.
This broad spectral coverage is important not only for addressing all the science objectives of COrE, but
also, as discussed below, to guarantee the effectiveness of component separation for the prime CMB scientific
objectives. This coverage is a strength of the design of COrE as compared to previous CMB polarization
experiments, proposed or in operation.
2.3
Angular resolution
The observation of the primordial tensor CMB B-mode spectrum requires only modest resolution, of order
1◦ on the sky, significantly lower than already achieved by WMAP and Planck.
CMB lensing, however, generates deflections of order 2.7 arcminutes. An accurate measurement of such
deflections calls, for adequate sampling of the deflection field, for resolution of the same order of magnitude,
although the lensing can be measured at the power spectrum level even at a lower resolution.
For galactic science, the high resolution mapping of the galactic magnetic field in star-forming regions is
the key for understanding the interplay between turbulence and magnetic field, which leads to the formation
of pre-stellar condensations. A resolution of order 1 arcminute is adequate to complement what can be done
15
from the ground with ALMA and single dish sub-mm telescopes, which will provide polarization imaging at
sub-arcminute angular scales of a variety of compact sources including pre-stellar condensations.
As a last constraint, we note that galactic and extragalactic point sources can be a significant contaminant
for the measurement of CMB B modes, should the tensor-to-scalar ratio be low. As each point source has
its own emission SED, this contamination cannot be effectively removed using multi-frequency observations
unless the sources can be singled out in the maps in the best possible way. This calls for the best possible
angular resolution, which sets the lower limit on the flux limit of detectable point sources.
A limiting constraint for the COrE resolution, however, is set by the diffraction limit due to the size of
the telescope (set in turn by the size of the fairing for a space mission). We thus require the highest possible
resolution available from space, and demonstrate, as discussed in the science section, that it is adequate for
achieving the main science objectives of COrE.
2.4
Control of systematics
A rigorous control of systematic effects is an essential requirement for any experiment aiming to measure
CMB polarization. Systematic effects are essentially of two kinds. Additive errors (such as glitches due
to cosmic ray impacts stray signals due to temperature drifts of the payload, low frequency noise), and
multiplicative errors (errors on the instrumental response, such as inaccurate beam shapes, unknown cross
polarization leakage, sidelobes in the optical response, gain drifts).
This require both a very careful design of the instrument to minimize such systematics, but also the
capability to measure instrument imperfections in flight.
Extensive simulation work of systematic effects impact on the measurement of polarization has been done
in the context of Planck and for the EPIC project in the US, setting constraints on gain mismatch and beam
symmetry, which are taken into account in the COrE instrument design. In addition, we require the COrE
mission to provide four layers of redundancy: at the band level, with six different CMB–sensitive frequency
channels; at the detector level, with many different detectors re-observing the same sky within different time
periods; at the scanning strategy level, with a well connected coverage of the sky thanks to a well-designed
scanning strategy; finally, we require a polarization modulation which provides high coverage of polarization
angles per line of sight, while keeping the telescope attitude fixed w.r.t. the sky and therefore rejecting,
whatever their level systematics fixed w.r.t. the spacecraft.
2.5
Component separation
In the tour of the microwave sky in the bands covered by COrE, we described how the observed emission
in a particular frequency channel is the linear superposition of several components arising from distinct
physical processes. The various diffuse components identified—mainly emissions from the galactic ISM and
the primordial CMB—have differing frequency dependences. This feature allows for their separation, which
is a prerequisite for all the science described in the preceding sections.
Since the COBE detection, the development and application of component separation methods have
been a very active area of CMB research, and many methods have been successfully applied to temperature
component separation, both on real data and simulation for validation. In fact, until now foregrounds have
been much feared but have turned out to be less of a problem than anticipated, at least for the measurement
of the temperature power spectrum. In practice, many methods have been developed and optimized. At
their core, these methods derive from a number of simple ideas, including:
• masking the brightest point sources and modelling the residuals;
• cutting out the galaxy, less so for small wavenumbers and more aggressively for larger wavenumbers;
• extrapolating the frequency dependence of foreground components using physically motivated frequency
templates and templates obtained from correlation studies;
• using ‘blind’ methods to let the data express its natural decomposition into components based on the
hypothesis of statistical independence.
For polarization studies, a much higher degree of cleaning through component separation will be required
because for the B modes the polarized foregrounds outshine the targeted signal at all frequencies by about
two orders of magnitude or more (see Fig. 9). To study the COrE capabilities, several independent foreground
removal simulations were carried out. Fig. 9 shows the outcome of one of these simulations for r = T /S =
10−3 , based on the classical Internal Linear Combination technique, performed on simulated multi-frequency
maps with foregrounds modeled using the Planck Sky Model, with noise levels according to design COrE
sensitivities per frequency channel. As can be seen from the figure, a very simple blind component separation
method is adequate to measure the primordial B mode spectrum for r = 10−3 .
This simple exercise is just the outcome of one of many polarized component separation exercises made
for the preparation of a future space mission. Additional methodshave been developed among the proposers
of the COrE mission. One of such methods, also tested specifically for the COrE design, relies on a two step
analysis. In a “model learning” phase, a linear model is matched to the data using the CCA blind component separation method in harmonic space. Then, a “source reconstruction” phase extracts components
16
by inverting the linear system using a generalized Least Square solution in pixel space. The linearity of the
process allows for estimating foreground residual and noise errors by filtering separately a foreground-only
dataset and a noise-only dataset. This method results in errors forecasts in agreement with those obtained
in figure 9.
More technical details about the component separation for COrE can be found on the website : www.coremission.org/proposal/foregrounds.
Figure 9: Component separation exercise for B mode detection assuming (T /S) = 10−3 . The solid black curve shows
the predicted blackbody B mode power spectrum, which is a combination of the tensor B modes (black curve) and a
gravitational lensing background (not shown) making primordial E modes appear as B modes in part. The upper solid
blue curve shows the contribution of diffuse galactic emission in one of the “cleaner” channels (here 105 GHz). The
red curve indicates the instrument noise that would be obtained combining five CMB channels, an the light blue curve
indicates contamination by point sources after the brightest ones (S > 100 mJy at 20GHz and S > 500 mJy at 100
microns) have been cut out. The purple data points indicate the recovered raw primordial spectrum measurements, as
compared to the theoretical spectrum (purple line). The black points result after the gravitational lensing contribution
has been removed, leaving only the recovered tensor contribution. Here a galactic cut with a conservative fsky ' 0.50
has been used.
2.6
Why space?
Balloon experiments such as EBEX and SPIDER along with ground-based telescopes such as PolarBear and
SPT1 ), with thousands of detectors in one or two frequency bands and sufficiently long integration time, are
just now reaching levels of instantaneous sensitivity to achieve white-noise levels of 3µK per arcminute at
a single frequency, arguably sufficient to reach r ' 0.05 in the absence of contaminating systematics, with
angular resolutions varying from ten arcmin to one degree. These experiments have already shown us the
technical feasibility of measuring the CMB with thousands of detectors.
However, the full science goals that we have outlined above require a yet higher sensitivity to cosmological
and astrophysical polarization over a wider range of angular scales and frequencies. To achieve this in practice
further requires control of systematics at a level that cannot be achieved in suborbital experiments. A satellite
experiment like COrE is therefore a necessary next step. It will observe the full sky, have multiple bands over
more than a decade of frequency, and take data for very long integration times in the very quiet environment
of the L2 Lagrange point of the Sun-Earth system.
2.6.1
Full sky coverage
COrE ’s science goals are best achieved with high sensitivity over the full sky. This will enable measurement
of the reionization signal in polarization, crucial for confirmation of the last-scattering B-mode signal at
1
for an up-to-date list of suborbital experiments, see http://lambda.gsfc.nasa.gov/product/suborbit/su experiments.cfm
17
medium scale. Moreover, it will build up sufficient signal-to-noise on the non-Gaussian signal, constrained
by the number of independent triples (or quadruples, etc.) available. This will be crucial if we want to go
beyond detection of fNL to detailed non-Gaussian models of the higher-order moments as discussed above.
Situated between the strongly-radiating ground and Sun, suborbital experiments are highly constrained
in the amount of sky that they can observe as well as the way that they can observe it. They must scan at
constant elevation to avoid rapidly-changing ground pickup while simultaneously avoiding the Sun. Hence
these experiments can typically observe a maximum of a few thousand square degrees with only mild crosslinking of different scans, with details strongly dependent upon the location of the experiment.
2.6.2
Avoiding the atmosphere
The atmosphere allows observation in just a few atmospheric windows, incompatible with the wide frequency
coverage required to fulfill the scientific objectives of COrE. If only for this reason, COrE must be a space
mission.
Even for less ambitious science goals, the atmosphere turns out to be a severe problem for polarized CMB
observations from the ground.
The atmosphere indeed is a strong source of microwave emission. Although it is largely unpolarized, it
adds to the photon shot noise, reducing detector sensitivity and hence making necessary longer integration
times. Moreover, Zeeman splitting due to the Earth’s magnetic field leads to both linear and circular
polarization. The linear polarization may be small (≈ 10nK), but the circular polarization is much larger
(≈ 102 µK), and great care must be taken to avoid conversion into linear polarization within the instrument.
Another effect leading to significant polarization is the back-scattering of thermal radiation from the hot
ground by the atmosphere. This results primarily from large ice crystals (i.e., 100 – 1000 µm) in high-altitude
clouds, often optically thin (along with a smaller effect from molecular scattering). Simulations demonstrate
that at both the South Pole and in the Atacama desert such polarized emission will likely swamp the B mode
signal from primordial tensor modes.
2.6.3
Benign environment (especially L2)
Even with highly-constrained constant-elevation scans, suborbital experiments would need to reject far sidelobes from the 250 K × 2π sr ground at a level of roughly 10−10 to remain below the desired noise on large
angular scales. From a distant point in space such as L2, the Earth will instead subtend an angle of only 0.5◦ ,
reducing its emission by a factor 105 . Moreover, at L2 the Earth, like the Sun, can be more easily shielded,
limiting the emission from both to diffraction around (and temperature variations within) the shielding.
These same considerations apply to temporal variations in the ambient temperature, which can be up to
several degrees per day on Earth at mid-latitudes, compared to the nearly completely stable environment
in space. For COrE , any change in thermal loading is due to the behavior of the cooler itself and can be
strictly controlled and monitored, as well as kept to specific times during the mission.
One potential worry in deep space is the increased flux of cosmic rays. The rates at L2 for high-sensitivity
bolometers have been measured by Planck-HFI, and, while not negligible, have not prevented Planck from
reaching (and, indeed, surpassing) its sensitivity goal. Hence, the achieved Planck detector sensitivity gives
the proper guideline for COrE , multiplied by the vast increase in detector numbers. We are thus very
confident that any extra noise induced by cosmic rays will not challenge the COrE sensitivity goal.
The L2 environment also allows much longer integration times than can be achieved from balloons (with
an Ultra-Long Duration limit of roughly 40 days at present) and a higher duty cycle than ground-based
experiments.
Only from space can an experiment like COrE achieve its simultaneous goals of full-sky measurements of
astrophysical polarization over a decade in frequency and an angular resolution of 6 arcmin, and µK · arcmin
sensitivity coupled with thermal stability and 10−5 sidelobe rejection.
3
Mission profile proposed to achieve these objectives
To achieve the science objectives, a polarization surveyor with full sky coverage in 15 bands with an angular
resolution of the order of few arc minutes and an integrated sensitivity of 1 µK · arcmin is proposed. The
nominal mission duration is 4 years.
3.1
Launcher requirements.
The rejection of parasitic signals from the side lobes pick-up of the Moon, Earth and Sun emissions, together
with the need of a full sky coverage, lead to the choice of an L2 orbit for Planck . The same applies to
COrE . A Soyuz launcher that would take off from Kourou offers the best fare compatible with instrument
requirements. In fig. 10, we present the accommodation of COrE inside the fairing of a dual-payload type-S
bay.
18
Figure 10: Fairing accommodation of COrE .
3.2
Orbit requirements.
An 500, 000/800, 000 km orbit around L2, similar to Herschel , has been selected, to save payload mass,
taking advantage of the lower ∆v necessary with respect to the Planck orbit (100ms−1 instead of 400ms−1 ),
reducing the need to a single tank of 128kg of hydrazine instead of three. This is made possible thanks to
an improved “shadow cone” of the sunshield + V-grooves.
The mission will last at least 4 years, requiring periodic orbit maintenance. This long lifetime is achieved
with the use of a cryogenic system free of consumables (see § 4.3.7). The transfer from Earth to L2 would
last 2 to 4 months. The calibration and performance verification phase will last 2 months.
To achieve optimal sky coverage and redundancy (both in terms of hits per pixel and angular coverage),
COrE will scan the sky as described in § 6.1
3.3
Ground segment requirements
The payload will host a single, very homogeneous survey instrument. The mission concept is thus very simple
in terms of operations and Ground Segment. The most important operational constraint is the stability of
the system for as long as possible during the mission. The pointing strategy will be defined well in advance
and the Mission Operation Control will operate the spacecraft according to the pointing law, ensuring respect
of the constraints.
Every day the instrument will simply survey the sky for 24 hours according to the predefined program,
spinning at 0.5 rpm and producing a raw data-rate around 18 Mbit/s; we know from Planck that this can
be compressed to 4.5 Mbit/s for transmission. The data will be stored on-board and will be dumped to the
Earth via a high-gain antenna aligned to the spin axis for about 2 hours per day in the Ka band, during which
the spin axis will be directed to the Earth. The instrument will continue its survey during the downlink.
Ability to transmit more data or for longer duration will be limited to a short calibration and performance
verification phase before starting the survey, once cold and around the L2 orbit.
The high redundancy given by the nutation/precession scanning strategy will be the most powerful tool
to deal with any data discontinuities and downlink problems without the requirements of fast reaction to
events, as in the case of Planck . So the number of operations and procedures will be very limited after the
end of the calibration and performance verification phase.
3.4
Special requirements
This is a cold payload (∼ 30K for the telescope, ∼ 0.1K for the focal plane), which should be kept in the
shadow of the solar panels and V-grooves for the whole mission. The satellite is a re-orientable slow spinner,
< 20◦ from the anti-sun direction, basically pointing the earth to ease
whose spin axis will be at all times at ∼
downlink communications. We also require to be able to change the spin rate (nominally around 0.5 rpm)
by ±10% at 6 months time scales, which provides a powerful check for (spin-synchronous) systematic effects.
19
4
4.1
Model payload to achieve the science objectives
Overview of the payload
Based on the exceptional results from the Planck mission and on its actual performances we are now in a
better position to optimize a new satellite design. COrE is an evolutionary concept based heavily on Planck
and Herschel missions, plus a generation of ground based projects. The science breakthrough is via the great
increase of number of detectors through the use of large arrays of feedhorn coupled dual-polarized detectors
divided in 15 spectral bands. Moreover, in comparison to Planck, the incoming polarization is modulated
through the use a unique rotating reflective half-wave plate (RHWP) leading to a much greater accuracy in
the measurement of the polarization.
The instrument (Fig.10) is based on an off-axis reflective telescope able to accept a large focal plane area
(Fig.11 left) with limited aberrations and cross-polarization. The RHWP is located at the entrance aperture
of the telescope so that the instrumental polarization systematic effects are easier to remove. The incoming
radiation then goes through the telescope to be focused onto the horns that are coupled to Ortho-Mode
Transducers (OMT) through a circular waveguide. Each branch of the OMTs separating the two linear
polarizations is coupled to a detector. The cold optics (feedhorns - OMTs - detectors) is cooled at 100 mK
and surrounded by several successive temperature stages. The whole instrument is enclosed in a passively
cooled 35 K shield. The telescope mirrors and the RHWP could be at a lower temperature of 30 K.
Figure 11: Left: FPU: the highest frequency band (795 GHz) is at the center and the 45 GHz horns at the periphery.
Right: Ray-Tracing of the optical system showing the off-axis telescope and the RHWP. Fig.10 shows the instrument
inside the Soyuz fairing.
4.2
Description of the measurement technique
COrE is devoted to a full sky measurement of anisotropies and polarization of the CMB. The choice of a
far-Earth orbit, such as a halo or a quasi-periodic Lissajous orbit around the Sun-Earth Lagrangian point
L2, is a mandatory feature for COrE in order to minimize stray light contamination from sidelobe pick-up.
Requirements for far-sidelobe rejection are extremely tight for COrE . Because the broad features of the beam
pattern at large off-axis angles are essentially the same for all detectors at a given frequency, the required
rejection must be calculated against the total sensitivity per frequency channel. Using Planck as a guideline,
and assuming a factor of 10 better sensitivity for COrE in an L2 orbit, we need roughly -110dB rejection for
Sun contamination, -100dB for the Earth, and -80dB for the Moon. Stringent, highly frequency-dependent
rejection requirements will be imposed also by straylight from Galactic diffuse emission on intermediate
sidelobes. While challenging, stray light rejection limits for COrE can be achieved with very careful optical
design and can be tested with a moderate extrapolation of the state-of-the-art technology developed by
ESA for the Planck mission. Furthermore, the excellent thermal stability of the L2 environment is ideal to
minimize thermal systematic effects on the instrument. The experiences of both WMAP and Planck have
demonstrated exquisite thermal stability for small changes in the solar aspect angle, with very small and
slow residual thermal drifts.
From the data analysis point of view, we need full sky coverage, to optimize the redundancy per map
pixel and the variety of time scales involved in the re-visit of a given pixel in order to reject “naturally” the
statistical noise and most common systematics. Also, polarization determination requires a wide variety of
measurement angles per pixel. In the first B-Pol proposal, all the polarization modulation efforts were put
on the payload motion. This led to a sophisticated (but feasible) scanning strategy and set tight constraints
on sun shielding. For COrE , the addition of a HWP modulator relaxes these requirements and allows to
20
Parameter
β
µ̇
∆φ
α
ψEarth
φ̇
Table 1: Scan parameter definitions and allowed variation ranges.
Definition
Range
Potential criticalities
◦
telescope axis
65 − 85
> 85◦ : stray light, polarization angle redundancy
to spin axis angle
< 65◦ : excessive excursion α angle required
spin rate
0.1 − 2 rpm
< 0.1 rpm : 1/f noise, thermal fluctuations
> 2 rpm : bolometer time constant
Depointing angle
1−few arcmin < 1: limited redundancy in sky circles
> 2: insufficient sky sampling
Solar aspect angle
5 − 27.5◦
< 5: coupling with β and full-sky requirement
> 27.5: thermal effects, stray lights
◦
Maximum Earth
0 − 15
> 15: TM/TC requirements
aspect angle
Precession frequency
1-24 weeks
polarization angle redundancy, destriping
thermal effects
assume a scanning strategy à la Planck, with only a precession of the spin axis and possibly some level
of nutation. Assuming the COrE spacecraft to be a spinner, the spin rate µ̇ needs to be fast enough to
ensure redundancy for minimizing the effect of 1/f noise and thermal fluctuations, and slow enough to be
compatible with the bolometer time constants. Based on the Planck experience, we anticipate something of
the order of 0.5 rpm, and a step and point approach for the spin axis like in Planck, with similar de-pointing
parameters ∆φ of 1 to a few arcmin every few tens of circles and therefore maintain the nearly anti-solar
configuration (∆φ/∆t ∼ 1◦ /day). This would yield sufficient redundancy in sky circles and, simultaneously,
good sampling of the sky for each single beam up to 220 GHz, i.e. in all the main COrE cosmological
channels. At higher frequencies, where the angular resolution is θF W HM < 4 arcmin, adequate sampling
would be ensured by proper staggering of the feeds in the focal plane.
Tight constraints are imposed on the scanning parameters by thermal and stray light effects, and by
telecommunication requirements. The maximum excursion α of the spin axis from anti-Sun direction must
be limited within 27.5◦ in order to avoid thermal effects and stray light contamination from the Sun and Moon,
and to rule out power-modulation due to asymmetries in the spacecraft or shadowing effects on the solar
panels. Also, the angle of the spin-axis to Earth angle must be limited within 20◦ to ensure good visibility of
a fixed medium gain antenna for TM/TC activity. While steerable antenna designs are possible, we do not
favor their use since they would complicate the design and increase the risk. Depending on the dynamical
constraints and attitude control capabilities, modulation of the spin axis can be set either by precession
(with the advantage of maintaining a constant solar aspect angle) or with cross-ecliptic excursions at an
angle α = 90 − β. The optimal configuration will be determined via simulations that include polarization
angle redundancy, de-striping efficiency, and figures of merit for stray light and thermal effects. Table 1
summarizes a possible baseline based on the Planck experience.
The use of a three-axis stabilized satellite allows further detailed optimization around that basic concept.
For instance, one can imagine to combine the best of the Planck and WMAP scanning, by superimposing a
daily nutation, of 20◦ around the anti-Sun direction, bringing the spin axis back in the anti-earth position
after 24 hours. This has the effect of distributing the basic ring measurement2 to over a∼ 40◦ annulus,
therefore allowing to revisit the same sky pixels at various times over 40 days (constraining systematics of
various periods), at the expense of a somewhat uneven sky coverage. This detailed optimization is left to
the next stage of the study.
In addition to the classical compromise between survey size and sensitivity, COrE has to account for
the directivity of polarization. It is both an additional constraint on the system and a lever arm against
systematic effects. Polarization is usually described in terms of the Stokes parameters I, Q and U . In COrE
we propose to implement a polarization modulator made of a rotating Half-Wave Plate (hereafter RHWP)
which modulates simultaneously all the detectors in the focal plane, so that each detector measures
m = 0.5(I + Q cos 4α + U sin 4α) + n
(1)
where n is the noise and α = ωt is the orientation of the RHWP. Each measurement is the linear combination of three independent parameters. Three independent measurements, each at a different angle α, are
therefore required (at least) to determine the polarization state. In the absence of a polarization modulator,
these measurements would require to rotate the detector. However, that approach mixes beam-shape and
detector-related effects with real polarization signals. The introduction of the modulator has simplified very
significantly the measurement and scan strategy: the satellite does not need to be rotated, and the sky scan
can be very similar to the one used in Planck : the continuous rotation of the RHWP, at a frequency not com2
i.e. the measurements made with a stable pointing allowing revisiting the same pixel by the same detector but with several
RHWP orientations.
21
Figure 12: Angular power spectra of the 1/f noise timelines: Previous B-Pol proposal without HWP (left), Planck
and continuously rotating HWP at 1Hz (right).
mensurate to the spin frequency, provides the required change in α for each re-observation, after each scan,
of the same sky pixel. After at least 3 rotation, a ring of fully determined I, Q, U can be obtained, further
rotations allowing an improvement in signal-to-noise and a better discrimination against systematics3 . The
basic idea behind this modulated measurement, as equation 1 shows, is that polarization is modulated at 4α
when the modulator is rotated by α, whereas most systematic effects are fixed in the instrument reference
frame. Rotating the modulator with respect to the incoming polarization and providing sufficient angular
coverage is therefore analogous to a lock-in detection that selects polarization and cancels out systematic
effects. The rotation frequency of the modulator is around 0.1 ÷ 0.5 Hz, slow enough to avoid microphonic
effects.
Results of detailed simulations of this scan / modulation strategy are summarized in fig. 12. The improvement with respect to the Planck survey (which had not been optimized for polarization measurements)
is very significant. For simplicity and illustration purpose, we chose typical Planck parameters (α = 5◦ ,
β = 85◦ , µ̇ = 1 rpm) with a continuously rotating HWP (1Hz). For these simulations, we took standard
white and/or 1/f noise with fknee = 0.1Hz and a N ET = 10µK.sec1/2 .
A detailed optimization of scan parameters (angles and precession/nutation frequencies) could be done
in a second step. For this work, suffice is to say that the addition of a HWP enables COrE to relax previous
constraints on sun-shielding and scanning implementation. In light of these one detector simulations, a
continuous rotation of the HWP seems to better reject 1/f noise than a stepped HWP. However, we wish to
insist on the fact that (1) the 1/f rejection can be addressed via performant de-striping algorithms such as
those developed in the context of Planck; (2) stepping the HWP enables to fine tune the angular coverage
and can represent an important technical simplification.
4.3
4.3.1
Instrument conceptual design and key characteristics
Optical configuration
The specifications needed to achieve scientific goals of the mission and directly relevant to the optical design are the resolution, the edge taper/spillover, the beam ellipticity (requirement< 1%), and the crosspolarization (req.< −30 dB). In order to meet our requirements with high TRL technology and within the
dimension and weight budgets of a Soyuz launcher, we have chosen to base the payload on a unique reflective off-axis compact test range telescope configuration (such as for ground-based Clover and QUIET
projects) following the Dragone-Mitsugushi condition. Used in conjunction with feedhorn-coupled detectors
this design allows for a large focal plane unit (FPU) with limited aberrations necessary to fulfil the scientific
requirements. This technology is already rated TRL 9. However, because of the large number of detectors
that will be included further development is required (§ 5.1.1). To limit the dimensions of the FPU, the
horn apertures need to be small, resulting in a fast optical system. The resulting configuration has a 1.2 m
3
Decomposing the periodic sky signal in Fourier harmonics, one sees that the sky signal is contained in harmonics of the spin
frequency with a width all the more narrow than the series of rings is long, therefore offering a better rejection of all systematic
effect which are not at those precise harmonics.
22
Figure 13: GRASP beam simulations in the U-V plane for two extreme off-axis pixels: Right 45 GHz, Left 105 GHz.
Both give 5% beam ellipticity. Optimization/reduction of the beam ellipticity will happen during the Phase A.
projected diameter making use of F2 horns, each having a 3λ aperture diameter. Taking into account an
array filling factor of 0.8 and leaving enough spacing for the spectral filtering, the overall FPU diameter is
about 40 cm (Fig.11 right). The whole optical system is enclosed in a cavity formed by a shield at 35 K with
the RHWP acting as a cold stop. It is expected that the telescope could reach a temperature as low as 30 K
with passive cooling (compared to 40 K on Planck).
While this first iteration optical concept is already producing good performances, improvement will be
needed during a Phase A study. Attention will have to be paid to the beam ellipticity of the edge pixels
(about 5% at the moment) which could be improved with a lengthy optimization of the telescope / RHWP
designs as well as a re-shaping of the FPU surface (curved instead of the flat surface as assumed so far).
Also the possibility of having a colder RHWP will be investigated to reduce the spillover contribution and
to limit the effects of its emissivity (§ 4.3.4).
4.3.2
Polarization separation
Following the feed horn transmitting the full intensity of the incoming radiation through its circular or
square waveguide, the separation of the two linear polarizations is performed with an OMT. The assembly
Horn-OMT sets the minimum cross-polarization achievable between the two branches that conduct each
polarization to the detector. Waveguide OMTs are the best: return loss and isolation of the order of -20
dB and -50 dB respectively have been already achieved at 100 GHz on a 30% bandwidth, and -40 / -70
dB respectively at 30 GHz. Although further development is underway (§ 5.1.1), this TRL 9 technology is
potentially too heavy and, so far, allows similar performances only for frequencies below 150 GHz.
Baseline design: for this reason, the use of a lighter planar technology is preferred. For planar OMTs
the two polarizations are extracted by microstrip circuit probes and made easily available to the bolometers.
A completely symmetric planar OMT can be realized using four probes inside a circular waveguide which
are then recombined. Broadband four probe antenna OMTs have been realized and tested in the L-band:
within 30% bandwidth the measured return loss and cross-polarization were respectively -20 dB and -40 dB.
While this technology could, in principle, be scaled fairly easily for operation at higher frequencies, planar
OMTs have been developed for Clover (150/225 GHz) but with lower measured performances. Development
is thus required to reach higher performances for frequencies up to 800 GHz.
4.3.3 Detectors and readout electronics
In the frequency range of COrE , bolometers cooled to very low temperature (T ≤ 100mK) have currently
the best performances for large bandwidth detection. With a careful instrument design, their sensitivity
can be limited by the photon noise of the astrophysical source. The understanding of their behavior in
space environment has been dramatically improved thanks to the Planck satellite. In order to realize and
readout the large number of detectors required, transition edge superconducting bolometers (TESs) are used.
Various materials systems can be combined to form the thermal sensor: thin films such as NbSi or bilayer
like MoCu. The critical temperature of these sensors has to be matched to the power handling and cooling
requirements of the instrument. TESs have been developed extensively for astronomical observations from
millimeter through X-ray wavelengths, and have a long history of use.
Microstrip-coupled TES technology is the detector baseline for COrE , where thin-film waveguide
antennas in the OMT are used to couple radiation from a horn antenna onto the TES through a superconducting microstrip transmission line terminated with a resistor. Their realization will be done on the same
wafer as the OMTs using micro fabrication techniques used to develop superconducting bolometer arrays for
ground based experiments like QUBIC, SCUBA2 or Clover.
23
Detailed models4 indicate a photon NEP of about 4.10−18 W.Hz −1/2 in all wavebands. A sensitivity
requirement for the detectors has been chosen to 70% of the photon noise or 3.10−18 W.Hz −1/2 which increases
the total NEP by only 22%. Differential thermal conductances of typically 10pW/K are needed for the TESs
themselves to reach the required NEP, where an operating temperature of 100mK has been assumed. These
conductances are above the quantum conductance limit even for a structure supporting a number of acoustic
modes. Thermal conductance of 0.4pW/K are currently being designed and tested for the SAFARI satellite.
The time constant of the detectors should be lower than the time needed to scan a beam not to degrade
the angular resolution. The stringent constraint is therefore given by the high frequency channels where the
beam size is the smallest. With the proposed design, the required time constant ranges from 10ms to 0.6ms.
TESs show a strong electro-thermal feedback which speeds up their effective time response to the sub-ms
values required by COrE .
The readout system of bolometer arrays requires a multiplexing technique to decrease the thermal loading
on the last cryogenic stage but also to simplify the architecture. The COrE design is based on time domain
multiplexing using SQUIDs as first amplifier stage. A standard SiGe BiCMOS ASIC cooled to 80K will be
used to control the multiplexing sequence and to amplify the signal from the SQUIDs. This technology has
been tested in space by NASA during the MISSE-6 mission onboard of the ISS. A multiplexing factor of 24
has been demonstrated with a heat load of 20 mW down to 4 K. Assuming the same power dissipation, we
have finalized a new design to readout 128 detectors with one ASIC. The COrE architecture will require 50 of
these new ASICs, leading to a power load of about 1 W. Further improvement in terms of power dissipation
and multiplexing factor will be achieved during phase A.
We are also considering the use of superconductive Kinetic Inductance Detectors (KIDs), new devices
operating at 0.1K that promise high-sensitivity, large-format detector arrays. KIDs comprise a superconducting thin-film microwave resonator capacitively coupled to a probe transmission line. By exciting the
electrical resonance with a microwave probe signal, the transmission phase of the resonator can be monitored, allowing the deposition of energy or power to be detected. The KIDs read each pixel at a separate
frequency. Therefore, only one pair of coax leads is needed to connect several thousand pixels to the read-out
electronics simplifying the thermal architecture. These devices have been demonstrated in the laboratory
and at the IRAM telescope. Since space-compliant versions are still under development, these detectors are
at TRL 4.
4.3.4 Polarization modulator
A classic Half-Wave Plate operated in transmission cannot cover the full COrE spectral range. We have
therefore chosen to use a Reflecting Half-Wave Plate (RHWP). It consists of a free-standing wire grid polarizer
kept at a specific distance d from a flat mirror (Fig.14). It rotates around an axis orthogonal to its surface and
passing through its center. Assuming the RHWP oriented as in Fig.14, depending on the angle of incidence
φ (dictated by the optical setup) the RHWP will introduce a phase-shift between the two orthogonal s and
p polarizations, respectively orthogonal and parallel to the plane of incidence. When the phase-shift is equal
to 180◦ and the RHWP rotates with angular velocity ω the overall effect is a rotation of the reflected linear
polarization at a frequency 2ω. This happens only when the path difference between the two waves is equal
to π, i.e. at frequencies:
c
νnRHW P = (2n + 1)ν0RHW P
with
ν0RHW P =
(2)
4cos(φ)d
where n is an integer and c the speed of light. The modulation efficiency vs. frequency is sinusoidal with
maxima at frequencies νn . Without spectral filtering, the average modulation efficiency is about 0.5. Adding
spectral filtering (to be implemented - § 4.3.6), higher efficiencies can be achieved in narrower bands: a
reduction of 50% bandwidth around one of the peaks gives an efficiency of up to 80%. In our case, this
means that the bandwidth of most of the spectral channel is 15 GHz. For the last 3 high frequency bands
(555, 675 and 795GHz), the additional spectral filtering 5 allows to increase the bandwidth to 195 GHz.
The absorption is due to the power dissipated on the metal and on the wires. In a flat metal there
is always a slightly different absorption between the s and p polarizations, the latter being the lossier. The
losses increase with the frequency but they can also be reduced by decreasing the device temperature. For a
copper RHWP at 30 K used at φ = 45◦ incidence and assuming an ideal wire-grid, the absorption coefficients
for the s and p polarizations in the different channels are fixed and of order 10−4 .
A dielectric substrate RHWP based on photolithographic techniques could be used to solve the freestanding RHWP manufacture problems (§ 5.1.2). The air-gap between the wire-grid and the mirror can be
replaced by a dielectric substrate with the wire-grid directly realized on one of the substrate surfaces (Fig.14).
4
assuming 50% optical efficiency, 30K telescope with 1% emissivity, 15GHz bandwidth
The optimization of the partition of the available focal plane area among the different frequency channels should be studied
further. For the three highest frequency channels we have chosen broad bands containing several peaks (here 7) of the modulation
efficiency so that (∆ν)/ν ≈ 1/3 in order to increase sensitivity. An option for study is to further increase the sensitivity (in
µKarcmin) by approximately 40% through introducing periodic sub-band filters to block the frequencies of low modulation
efficiency. In both cases the polarization frequency bands have a comb-like structure with 7 teeth.
5
24
Figure 14: Left: Free-standing RHWP. Right: Dielectric substrate RHWP
The addition of a dielectric substrate introduces frequency-dependent standing waves (multiple reflections)
within it. Their amplitude will also depend on the type of polarization s or p travelling through it. Models
show that the resulting modulation efficiency deteriorates with increasing substrate refractive index. Using
a substrate with n=1.2 the efficiency drops down to 0.4 which can be improved through spectral filtering.
There are additional losses due to resonances that appear at frequencies corresponding to the device efficiency
peaks. However while the RHWP is rotating the s and p absorption coefficients fluctuate at a frequency 2ω
and the averaged fluctuation amplitudes are of order 6×10−5 and 1.5×10−5 , respectively.
While considered at TRL 4 only, we are choosing the dielectric substrate RHWP as the modulator
baseline for COrE as it offers the best potential performances and manufacturability. Indeed it is based on
the dielectric embedded interference filter technology that has been flown on Planck and Herschel.
4.3.5 Polarization modulator rotation mechanism
The RHWP will need to be rotated. While a step and integrate rotation could be considered, a continuous
rotation at 0.1 to a few rpm would be preferable both for the observation strategy (§ 4.2 but also to avoid
affecting the rotation of the satellite. Such a strategy has been adopted on the balloon borne experiments
Maxima and EBEX for which a superconductive magnetic bearing has been developed. It is able to rotate the
HWP at 2 Hz with low friction dissipation. This is a crucial parameter as the RHWP will have to operate at
the lowest temperature possible (30K for this instance) in order to reduce the systematic effects related to its
emissivity. The rotation will be performed by a fiberglass shaft linked to a step motor operated at a higher
temperature within the service module where a larger cooling power is available (shown in Fig.10). The
difficulty resides in having to rotate a HWP of 1.7 m diameter, weighting approximatively 25 kg including
its support structure while it is being kept at 30 K. The technology is rated TRL 8 (§ 5.1.3), the associated
risks being the mass and dissipated power.
4.3.6
Spectral filtering
Bolometric detectors being sensitive to a very wide spectral range, their spectral band definition requires
special attention. COrE requires a high in- band transmission in order to reach a 50% overall optical
efficiency whilst the rejection of optical/NIR power must be better than ≈ 10−12 while a rejection of ≈ 10−6
is enough in the FIR. Low frequency rejection can be easily achieved through the horn waveguide design.
The Planck-HFI-like high frequency band definition based on dielectric embedded interference filters, requires
several multilayer mesh filters in series located on the different temperature stages also limiting the thermal
power reaching the ultra-cold stages. This technology, selected as the baseline for COrE , allows for
use with mixed frequency array by arranging their geometry to suit the shape of the sub-array. It is also
conceivable to make a unique filter made of several regions. However, special attention will have to avoid
potential stray light in between regions/filters and cross-talk.
In order to reduce the number of interference filters mentioned above, superconducting planar bandpass filters could be lithographed onto the detector chips, which together with a planar OMT would lead to
intrinsic on-chip polarimetry. While parts of this technique has been demonstrated, it would require further
development.
Additional spectral filtering: As mentioned in § 4.3.4, the modulation efficiency of the high frequency
channels can be increased using broader bands in combination with sub-band selective filtering around the
peaks. This can be achieved for example using multilayer dielectrics that act as dielectric mirrors. However,
while the RHWP modulation efficiency peaks are centered at an odd multiple number of its basic frequency as
shown in Eq.2, the filter transmission bands are centered at an even multiple number of its basic frequency
following Eq.3. So a slightly different basic frequency ν0F ilter will be used for the filter so that we have
maximum coincidence between the two sets of peaks. The example in Fig.15 shows a possible filtering
around the 795 GHz peak using ν0F ilter = 15.29 and n = 26.
25
νnF ilter = 2n × ν0F ilter
(3)
Figure 15: Sub-band filtering: Filter transmission (blue) and RHWP efficiency (red).
4.3.7
Cryogenic chain
The mass and power requirements and the temperature stages being similar to those of Planck, a duplication
of the Planck cryogenic chain would bring the whole system to TRL 9. However several modification would
give the whole system superior performance. The main emphasis will be to develop a system based on
mechanical coolers and recyclable system in order to have a long duration mission. Below is a summary from
a CNES study.
Passive cooling:
The V-grooves that have been used in Planck to radiate some power into free space but laterally to the spin
axis have proven to be an efficient solution to a first passive cooling stage. They can also provide cooling to
intermediate stage electronics. An inner V-groove temperature of 46 K is currently achieved in orbit with an
available power of the order of 1 W for Planck. With further optimization and with an addition of a fourth
V-groove, we expect that a temperature of about 35 K could be achieved for COrE inner V-groove.
Active cooling - Mechanical coolers
The 35 K to 20K stage could be provided by a H2 JT with metal-sorption compressor from JPL (Planck).
However a Pulse Tube Cooler (PTC) providing 15 K is under study (§ 5.1.6).
The 15 K to 4 K stage is provided by a Joule-Thomson (JT) cooler similar to the Planck one.
Active cooling - Sub-K coolers
Because the OpenCycle dilution Refrigerator (OCDR) providing both the 1.7 K and 100 mK stages from
4 K on Planck has a limited lifetime, a recyclable system is preferred. The 0.1 K stage baseline could be
provided by an ADR cryocooler (e.g. from CEA SBT), requiring at pre-cooling (e.g. at 1.7K and 4K on
SAFARI/SPICA) or a continuous ADR 4-stage (NASA-GSFC) developed for MIRI (JWST). While ADRs
are already at TRL 9 (XRS on ASTRO-E2), the strong associated magnetic field might affect the operation
of the TES readout system (SQUIDs), although efforts are being dedicated to this issue.
For this reason a Closed Cycle Dilution Refrigerator (CCDR), evolution of the flight-proven OCDR
can offer a more convenient solution. The re-circulation of the helium isotopes (He-3 and He-4) of this system
allows theoretically to have no limitation on lifetime. Contrary to the OCDR, the CCDR requires a precooling stage around 1.7 K in order to evacuate the heat from thermalisation of He-3 and He-4 gas flows and
from condensation of He-3 (a fraction this heat is used for the evaporation of He-3 but not all of it). It is
not possible to produce a 1.7 K stage with a He-3 Joule-Thomson expansion within this dilution system. As
a consequence, the 1.7 K stage is produced by a dedicated He-4 sorption cooler based on the concept of the
Herschel He-3 sorption cooler. The CCDR system also provides extra cooling power between 200 mK and
600 mK if needed (e.g. for thermalisation of mechanical supports or harnesses).
4.4
Performance assessment with respect to science objectives
The sensitivity of a perfect instrument with a uniform coverage of the celestial sphere, expressed in terms of
noise angular amplitude spectrum cnoise , is given by Eq. 4 assuming a polarization modulation efficiency of
80%, where tmiss is the total mission duration, Ndet is the number of detectors and sdet is the sensitivity in
intensity of one detector.
c2noise
4π
= 2.4 ×
× s2det = (4.7µK.arcmin)2 ×
tmiss Ndet
4 years
tmiss
×
400
Ndet
×
!2
sdet
1
(4)
50µK.s 2
To reach such sensitivity, the proposed instrument architecture is designed in order to have background
limited performances. The detector number is furthermore optimized to reach the required sensitivity in
26
each channel and to fill the focal plane area. With the current design of the instrument 6 , this process leads
to the numbers given in table 2. The proposed configuration has a total of 6384 detectors in 15 frequency
bands. The core sensitivity is in the CMB channel (75GHz-225GHz) with 4950 detectors.
Table 2: COrE performances - assuming a 50% value for detection chain efficiency.
Central Freq.
∆ν
Ndetectors FWHM Unpol. sensitivity Q & U sensitivity
(GHz)
(GHz)
(arcmin)
(µK.arcmin)
(µK.arcmin)
45
15
64
23.3
5.2
9.0
75
15
300
14
2.7
4.7
105
15
400
10
2.7
4.6
15
550
7.8
2.6
4.5
135
165
15
750
6.4
2.6
4.6
195
15
1150
5.4
2.6
4.5
225
15
1800
4.7
2.6
4.5
255
15
575
4.1
6.0
10.4
285
15
375
3.7
10.0
17
15
100
3.3
26.6
46
315
375
15
64
2.8
67.8
117
435
15
64
2.4
147.6
255
555
195
64
1.9
218
589
675
195
64
1.6
1268
3420
795
195
64
1.3
7744
20881
4.5
Resources: mass, volume, power, on board data handling and telemetry
The resources have been established by the COrE team in collaboration with CNES and are based upon the
Planck mission. Instrument mass and power breakdowns are given in table 3. The total estimated mass of
the instrument (300 kg) and power (260 W) are higher than for the Planck instruments (LFI + HFI) (total
of 209 kg and 212 W) which is consistent with the fact that we have now only one instrument based on one
detector technology but with a much larger number of detectors.
The overall payload volume is similar to the Planck one. Stored under the Soyuz fairing the maximum
diameter is 3.8 m and about 4 m tall. With the solar panels fully deployed the diameter reaches 5.9 m.
Data storage capacity : a mass memory of 780 Gbits is needed (instead of 32 Gbits for Planck).
The data rate at instrument output is 4.5 Mbits/s, compression included (130 kbits/s on Planck).
Table 3: Table of mass and power for instrument sub-systems
Sub-system
Mass Power
Sub-system
Mass (kg)
30K Cold Box + telescope supports 150
Horns
50
Housekeeping electronics
60
Cryo Harness
30
OMTs+Detectors
30
FPU structure
20
Read out electronics
20
200
Total instrument
300
4.6
Power(W)
260
Pointing and alignment requirements
The beam of the instrument ranges from 23 arcmin to 1.3 arcmin at the highest frequency. The most
stringent requirement will come from the CMB channels (45 to 225 GHz) for which the ability to recover
the pointing information must be less than 1% of the beam FWHM. Applying the same criteria to the
high frequency channels, the pointing accuracy needs to be aboutf 1 arcsec. The telescope axis, horn axes
directions, and polarimeters main axes will be determined with respect to the satellite reference system
to within 0.5◦ during the calibration phase. This measurement will be repeated independently during the
verification / calibration phases at the beginning of the mission.
6
Instrument efficiency: 50%, telescope temperature: 30K with 1% emissivity, detector noise: 70% of the photon noise
27
4.7
Operating modes
After the verification phase the instrument will operate in a single survey mode.
4.8
Specific interface requirements: configuration needs, thermal needs
Similarly to Planck, COrE will be launched warm and will be cooled during its travel to the L2 orbit.
During the launch phase several sub-systems will be locked and released after launch. This has the principal
advantage of limiting the strength and dimension of the support which would be required to resist the launch
vibration level and which would undoubtedly increase the thermal conduction. These locking mechanisms
will be needed for the focal plane and the RHWP. Such a system has been adopted for Planck-HFI and is
then at TRL 9.
4.9
Calibration and other specific requirements
In order to keep the instrument simple and without single point failure, no on-board calibration system
will be implemented. COrE calibration will rely on ground calibration sequences and astrophysical sources
that will have been thoroughly studied by Planck and follow-up observations. Systematic effects affecting
the polarization purity of the incoming signal will be mitigated by the use of the RHWP in front of the
telescope. However, the RHWP will introduce systematic effects on the beam shape which will vary with its
orientation in need of characterization. The ground calibration will consist of complementary tests:
Sub-system level: (detectors, reflective optics, cold optics, RHWP, cooling system)
FPU: spectral response of each pixel, dark and illuminated I-V curves for various FPU temperatures.
Warm RF test in anechoic chamber: representative test of a few horns, OMTs and warm detectors
(heterodyne instead of TES) associated with the telescope, shields and the RHWP in order to measure the
main beam and far sidelobes. Similar test conducted during the Planck ground calibration campaigns have
been extremely beneficial. Emphasis will be on the measurement and modelling of the systematic effects
introduced by the RHWP.
Dedicated calibration facility reproducing the cryogenic and radiative environment expected during
the mission. A large vacuum chamber instrumented with a 2K cryogenic chain and with a full beam source
will be prepared. The chamber will allow the operation of the full instrument, observing a 2.7K load and
a polarized source modulated and oriented as needed. This setup will allow the precise determination of
the polarimetric parameters of the instrument (polarization efficiency and main axis angles) and of the
responsivity and noise of all the detectors for different radiative loading conditions. The proposers inherit
considerable experience from similar calibration facilities they built for Archeops and BOOMERanG, but
also through the calibration campaign performed on Planck at CSL-Liége.
4.10
Current heritage and Technology Readiness Level (TRL)
TRL levels associated to the instrument sub-systems as well as their heritage is given in table 4. It has to
be noted that by selecting the baseline for each sub-system, only 3 technologies out of 11 are not at TRL
9/8. The dielectric RWHP is based on known technology (same than filters), the TESs are being used on
ground based instruments and the planar OMT technology is showing encouraging results. The CCDF is
being developed in collaboration with industry.
Sub-system
Interference filters
Mechanical coolers
Horns
Wg OMTs
Telescope
Sorption cooler
4.11
Table
TRL
9
9
9
9
9
9
4: TRLs and Heritage of the various sub-systems
Heritage
Sub-system
TRL
Heritage
Herschel, Planck
ADR
9
XRS on ASTRO-E2
Planck
Rotation mechanism
8
Maxima, EBEX
WMAP, Planck
Planar OMTs
6
Clover, PAPPA
WMAP, Planck
TESs+Readout
6
GBT, CLOVER
WMAP, Planck
Closed cycle DF
4
Herschel
RHWP
4
Hertz
Proposed procurement approach
See § 6.8
28
4.12
Critical issues
Critical issues related to the payload sub-systems development are addressed in § 5.
In addition, care must be taken to reduce the RHWP emissivity (below ∼ 1%) to match the emissivities
for the two polarizations, and to keep the RHWP as cold as possible. The nominal temperature of the
RHWP is 30K, as for the rest of the telescope assembly. At this temperature, with a RHWP emissivity
of 0.1% at 100 GHz, and with ∼ 1% matching of the emissivities for the two polarizations, the RHWP
produces a spurious signal at a level of about 300 µK p-p at 2ωo . This frequency is far from the frequency
of modulation of a real polarized signal from the sky, 4ωo . However, non-linearities in the detectors might
produce a spurious low-level 4ωo signal. Bolometers being very linear, we expect this effect to be in the
sub-µK region. However, a lower temperature of the RHWP would certainly give some margin, would the
combination of detector non-linearity and emissivity mismatch produce a larger signal.
Because the power dissipated from the FPU, the telescope and the modulator (<10 mW) are reasonably
small, we do not exclude a RHWP and telescope colder than the baseline 30K even taking into account the
radiative heating from the last V-groove (about 10 mW, depending on the emissivity). The main problem
remains the conduction from the supports connecting the telescope/RHWP assembly to the SVM that need
to resist the launch phase giving a heat load in the hundreds of mW range. If, however, we can thermally
decouple from these supports after launch, we can easily reduce the heat load to a few mW. This can be done
using the so-called PODS (passive orbital disconnecting struts) or other active devices. The total 20 mW
thermal load might then be compensated by a JT cooler connected to the whole telescope and modulator
assembly. In these conditions, with a large emitting area (a few square meters) the assembly cools to a few
K in about one week, with a very important reduction of the radiative load on the focal plane box, on the
detectors, and a very significant reduction of the synchronous signal from the rotating wave plate.
5
5.1
Technology development requirements
Payload technology challenges and technology development strategy
5.1.1 Horns and OMTs
While planar OMT technology seems to be the most promising approach but in need of improvement, other
techniques will need to be investigated if it turns out that the planar technology shows limited performances.
Current areas of research are looking for the possibility to simplify the manufacture, to reduce the costs and
the mass of waveguide components while retaining their exceptional RF performances. Such developments
include drilled smooth walled horns (Fig.16) and platelet corrugated horns (Fig.16). Platelets technology
can also be applied to OMTs for which spark erosion manufacture tolerances reaches a few microns making
possible the extension to high frequencies. While the manufacture process is simplified and made cheaper,
the mass issue is still to be resolved. This can be tackled through stereolithography techniques that have
been applied by our US collaborators on a 100 GHz prototype horn with good RF characteristics. This
results in a horn made of metal coated plastic. While this reduces greatly the mass, further development
is needed in order to test the cryogenic performances. There is little doubt that within a few years these
techniques will be at a level of maturity allowing their inclusion in the instrument design.
Figure 16: Left: prototype of 100 GHz planar OMT (Fr). Center left: array of platelets horns (It). Center right:
array of drilled smooth walled horns (UK). Right: 97 GHz waveguide OMT (UK).
5.1.2 RHWP
The manufacture of the free standing RHWP would be certainly challenging considering that the required
diameter is greater than 1 meter. The mechanical structure of the free-standing wire-grid must be able to
guarantee constant spacing between the wires and constant distance from the mirror all across its surface.
To our knowledge, the biggest device of this type ever built so far has a diameter of 0.5 m (D.T. Chuss).
29
Commercially available dielectric laminates used in microwave engineering to build planar microstrip
circuits could be used to build a dielectric substrate RHWP. These laminates come in different sizes, thicknesses, refractive indices and they have copper depositions on both sides. The idea is to use one metallic side
as a mirror and photolithographically etch the other side of the substrate in order to realize metallic strips
that will behave like a wire-grid. The resolution required for the mask production in our frequency range is
achievable with commercial large size printers. The laminate can be joined to a thick metallic flat surface
to increase the structure stiffness. Note that the main problems of the free-standing RHWP are now solved:
the constant spacing between wires is guaranteed by the high accuracy of the photolithographic techniques
adopted whereas the constant distance mirror-wires comes directly from the accurate constant thickness of
the laminates that is required in microstrip circuits.
5.1.3 Rotation mechanism
The two main problems associated with this sub-system are thermal and mechanical: thermal dissipation of
the bearings, thermal conduction through the shaft and microphonic vibrations affecting the detectors.
Shaft: For the torque transmittance between the motor (located in the service module) and the RHWP,
a long fiberglass shaft is a good solution but is in need of further development. The interface between this
compound material and metallic part needs to be studied more carefully due to different thermal expansion
factors. Another critical point could be the dynamical aspect of the shaft. The shaft diameter will be
studied with the stress criteria (torque transmission) and the dynamic criteria (modes for launch and no
torsional modes to not interfere with the motor) potentially leading to higher thermal conduction between
the temperature stage of the motor and the 30 K of the RHWP. If so, magnetic contact-less coupling devices
could be investigated in order to reduce the thermo-mechanical aspects of the shaft.
Bearings Classic bearing mechanism for the RHWP will need to be studied for the lubrications problems
operated at 30 K. A dry lubrication solution might be adopted. However, in order to decrease the thermal
dissipation the High Temperature SuperConducting magnetic bearing, such as the one developed for EBEX,
would be a better solution to completely avoid the problem of lubrication but will need further development.
5.1.4 Detectors
Over the last few years many national funding agencies have invested heavily in establishing superconducting
detector fabrication facilities in Europe. Major facilities exist in the UK, France, the Netherlands, Sweden,
Germany and Italy. These facilities not only provide clean rooms with lithography, processing, and test
equipment, but also provide professional, traceable fabrication routes, producing high-quality high-yield
devices. A major part of the technology development, fabrication, and test programme will be to form a
single network that is coordinated by one lead organization, with all of the partner organizations working
towards the refinement of a single design across all of the chosen wavebands. Most of these laboratories are
working on the development of TESs and KIDs.
5.1.5
Closed cycle dilution fridge
The maturity level of CCDR is considered TRL 4, and should reach TRL 5 within 2 years with the ongoing
developments at Air Liquide / CNRS Neel institute.
5.1.6
Pulse Tube Cooler (PTC)
The 15/20 K stage could be provided by a PTC under development at Air Liquide/CEA (with a goal
to reach 15 K). Note that a 2-stage PTC with a 300 mW heat lift at 20 K (with intercept) has already
been demonstrated on the GSTP program 20-50 K Pulse Tube (ESA contract 20497/06/NL/PA). Ongoing
developments will allow to have more margin at the 20 K stage. This 20 K PTC stage requires a pre-cooling
at 80 K with a heat intercept around 6 W. Note that a pre-cooling at higher temperature (e.g. 120 K) is
possible, but with a slight decrease of the performance at 20 K. This heat intercept can be provided either
by a V-groove through thermal braids links for example (preferred solution), or by a dedicated 80 K PTC.
Maturity level of the 20 K PTC is considered TRL 4.
5.2
Mission and Spacecraft technology challenges
5.2.1 Data rate
COrE requires a data rate of 4.5 Mbits/s. Transmitting telemetry at such a high data rate requires a wide
frequency bandwidth, by far above the 10 MHz maximum allowed by the ECSS standard in the frequency
band allocated to L2 missions (8450-8500 MHz). The use of Ka band should therefore be considered as the
30
baseline for the science telemetry transmission of the COrE mission. In order to achieve the link budget, the
35 m ground station located at New Norcia is required as for Planck operations. From the on board side, a
0.5 m diameter high gain antenna pointing towards the Earth is required. Three configuration options are
identified for the antenna: 1-Antenna pointing mechanism (existing) 2-Active antenna pointing (R&D study
on-going) and 3-a fixed parabola antenna, pointed towards the Earth by the spacecraft. The conservative
option 3 is taken as the baseline.
6
System requirements and spacecraft key factors
The COrE mission is composed by a single instrument on-board a single spacecraft (which is quite similar to
Planck ), in orbit around the second Lagrangian point (L2) of the Sun-Earth system on a 500000/800000 km
orbit, like Herschel . It is a spinning survey instrument able to realise a new full sky survey about every 6
< 20◦ ) to anti-solar. In the
months7 , since the spin axis direction is constrained to remain relatively close ( ∼
following, we analyse the system implication of the changes with respect to Planck , like for instance a smaller
maximum diameter of the fairing on a Soyuz, the pointing strategy or the telemetry increase due to the large
increase in the number of detectors. This analysis shows the feasibility of the mission by capitalizing on the
technical heritage and demonstrated successes of the Planck /Herschel program.
6.1
Attitude and orbit control
The Lissajous orbit around L2 can be controlled by a 10 hours manoeuvre session per month. The propulsion
system is the same as that of Planck one, apart from the removal of 2 tanks since the new, Herschel like orbit does not need an insertion manoeuvre. Orbit maintenance is then similar to the Herschel one
(∆v = 13.5 m/s). The total needed (∆v = 100 + 30 m/s) for a total mass lower than 2000 kg can be obtained
with a single tank of 128 kg of Hydrazine, therefore removing two tanks as compared to Planck . The rest of
the propulsion system is identical from Planck .
The spacecraft attitude is similar to the Planck case. The spacecraft is spinning; the spin axis is maintained near anti-Sun, with some level of precession; and the telescope boresight is almost 90 degrees from
the spin axis. On top of this strategy, a large nutation of the spin axis is considered to increase redundancy.
The spinning rate is decreased to 0.5 rpm instead of the 1 rpm for Planck , which decreases constraints on the
needed speed of detectors, decreases the necessary data rate, and allow a standard solution for still making
it an easily steerable satellite.
In order to include the large nutation option, the baseline attitude control of the payload is based on an
assembly of (at least) 3 off-the-shelf wheels and a star tracker system. Assuming a comparable inertia to the
Planck one ∼ 2800 kg m2 , but with a 0.5 rpm rotation speed, the 150 N m s total angular momentum of the
satellite can then be steered with 68 N m s wheels (double with respect to the 30 N m s Herschel wheels). A
fourth redundant wheel would lead to a reaction wheel assembly very similar to that of Herschel . No specific
difficulty was identified since it is quite recurrent from Herschel /Planck .
6.2
On board data handling and telemetry
Data are partially treated on-board by a dedicated Data Processing Unit that compresses the data produced
by the readout electronics. Data are then transferred to the Service Module for storage and download. The
on-board data storage system should be able to save a few days of data ( i.e. 4.5×106 ×86400×2 ' 780 Gbit,
instead of the 32 Gbit memory on board Planck) in order to cope with downlink problems due to contingencies
in the telecommunication system. A strategy will be defined to optimize the data downlink after those events.
Detectors currently used for CMB observations are already close to photon noise limited, which means that
to increase the instrumental sensitivity, one has to increase the number of detectors. This directly impacts
on the telemetry requirements. Based on Planck experience, we request 16 bits per scientific data sample.
For an instrument equipped with ' 6400 detectors, and with a sampling rate matching the scan speed and
the FWHM of the beam for each frequency channel, this leads to 18 Mbps, which are reduced to 4.5 Mbps
by compression (as successfully demonstrated with Planck , which had an instrumental output of 130 kbit/s
). Assuming the telemetry is done with a 35 m antenna, e.g. the new Norcia antenna, and assuming a data
dump duration of two hours a day, we shall need a data dump rate of 4.5*12=54Mbit/s. This calls for a
high-gain antenna in the Ka band. In our current design, the antenna is fixed and aligned to the spin axis
of the satellite, with the satellite axis is constantly aligned to the Earth during the 2 hours data dump. This
duration could be extended if beneficial. Note that alternative designs could be even more advantageous,
like for instance an active antenna pointing.
Concerning ranging, commanding, and housekeeping data transmission functions, there is no requirement
exceeding the Planck TT&C performances. Hence the Planck RF architecture can be re-used from the shell
7
In the case of the specific scan-strategy of Planck , it takes about 9.5 months to achieve the first full coverage of the sky by
all detectors, and 6 more month to achieve the second complete survey (which starts 6 months after the beginning of survey 1,
even though the sky is not yet fully covered once by all detectors).
31
COrE mass budget
Element
Planck mass
(kg)
Payload
Structure, V-grooves
20K stage
4K
1.7 K stage
0.1 K stage
Telescope
RHWP movement
Instrument
95+104
SVM
767
Hydrazine
Other SVM
Total
384
78
1974
261
114
21
106
44
Planck → COrE evolution
+ 1 V-groove → + 20 kg
sorption cooler → redunded Pulse tube
similar JT
New, needed to start closed cycle 0.1K stage
open to close cycle → - helium tank + pump
+ RHWP; smaller Primary mirror Aperture
+ motor + shaft → + 20 kg
HFI + LFI → rough envelope from table 3
- 2 tanks → -30 kg
+ wheel assembly → +50 kg
+ Gyro (+50 kg) + Ka band & HGA (+10kg)
+Solar Array (+50 kg) + Sun-shield (+30kg)
- FOG gyro and environment Monitor
COrE mass
(kg)
281
26
21
10
30
60
20
300
887
128
69
1832
Table 5: COrE mass budget estimate by comparison with Planck . This appears to be compatible with Soyuz
lift off mass (2000kg at L2) with sufficient margins.
(X band transponder, TWT amplifier and antennas -low gain and medium gain) For the science telemetry,
as stated above, an additional emitting chain in Ka frequency band is required, including a TWT amplifier
from the shell. It can be coupled with the X band transponder such as TTC systems existing for deep space
missions (Bepi-Colombo, KaTE experiment) with the increase of data rate on the Ka link. As an alternative
solution, a Ka band emitter could be used dedicated to science data. The power increase for the Ka emitting
chain can be roughly estimated to 100 W, so 200 W in total for the TT&C subsystem while transmitting.
6.3
Mission operation concept (Ground Segment)
The Ground Segment operations are even simpler than those of Planck , since the payload will host a single
survey instrument, with high redundancy in sky coverage. As a baseline, after downlink the data will be
sent to the Instrument Team for analysis of the Housekeeping data and detection of anomalies. The Mission
Operation Control will take care of pointing reconstruction. The Instrument team will take care of the first
analysis of the quality of the data and of the data storage.
The pointing strategy will be defined well in advance and the Mission Operation Control will operate the
spacecraft according to the pointing law, ensuring respect of the constraints.
6.4
Estimated overall resources (mass and power)
We anticipate COrE to be very close in mass to Planck . Table 5 provides a ROM comparison of the Planck
versus COrE mass budget derived by the COrE mission Working Group with CNES support.
We also anticipate COrE to be close in power needed to Planck . Table 6 provides a ROM comparison of
the Planck versus COrE power budget derived by the COrE Mission Working Group with CNES support.
the COrE power is provided by a solar array in 12 sections: 6 sections of 1.3 m2 at the bottom of the satellite
(assuming an hexagonal shape), plus six deployable panels on each SVM panel (folded along the SVM for
launch). A total of 15 m2 is then available. This gives an end-of-life (∼ −10%) input power at maximum
sun aspect angle (∼ 10%) around 2000 W, i.e. identical to Planck (despite the smaller fairing), there again
providing comfortable margin as compared to the anticipated need.
6.5
Specific environmental constraints (EMC, Temperature, cleanliness)
The environmental constraints are very similar to the ones applied in the Planck mission. In addition to
that, the readout system based on SQUIDs generates specific problems of magnetic compatibility. This
issue is resolved internally in the instrument, by proper magnetic shielding already demonstrated in other
instruments.
32
Element
Payload
20K stage
4K
1.7 K stage
0.1 K stage
RHWP movement
Instrument
SVM
Transmission
ACS
Thermal control
PCDU
PCDU harness loss (6%)
Data Management
Propulsion (Orbit control mode)
Total worst case
COrE power budget
Planck power
Planck → COrE evolution
(W)
572
102
20
142+70
106
67
72
35
76
36
44
1342
COrE power
(W)
sorption cooler → redunded Pulse tube
similar JT
New, needed to start closed cycle 0.1K stage
open to close cycle → (+ pump)
+ motor + shaft → + 40 W
HFI + LFI → rough envelope from table 3
250
102
20
30
40
260
+Ka emitter & + 100W amplifier
+Wheels (4 × 30 W)
206
120
72
35
+ mass Memory
100
44
1279
Table 6: COrE power budget estimate by comparison with Planck .
6.6
Special requirements
The proposed scan and orbit allow for a fairly constant solar illumination of the satellite. The absence of
important dissipation in the cold instrument makes us confident that a stable operation temperature of 30K
can be reached using the solar panels as shields for the SVM, and Planck like V-grooves shielding the cold
instrument section. The passive cooling V-groove systems has therefore a TRL of 9.
6.7
Current heritage (assumed bus) and TRL
The service module is quite similar to the Planck one, with relatively minor modifications required to fit in
the Soyuz vector. Indeed, the spacecraft has to answer to similar general requirements corresponding to a
spinning cryogenic survey mission, and the mission was designed from start to take full advantage of the
Planck /Herschel successes.
The Soyuz fairing has a diameter of 3.8 m and a height of 5 m which is close to the Planck spacecraft
size (diameter 4.2m× height 4.2 m), but will nevertheless require a reduction of the diameter to 3.8 m, the
maximum size of Soyuz fairing. Also, the launcher Interface is the 1666mm Soyuz interface instead of
the 2624mm interface for the Planck /Ariane-5 interface. The central tube is then reduced, which is still
compatible with the presence of one hydrazine tank (128 kg of hydrazine). The height of the SVM is not
modified and remains equal to 840mm.
The 4th , upper V-groove shields the scientific payload from solar radiation and is itself shielded y the
rd
3 V-groove. The 3rd V-groove is in turn screened by the second and so on, while the first V-groove is
protected by the sun-shield, and in turn protects the upper stages from the re-radiation from the sun-shield.
The difference with Planck is the addition of a 4th V-groove to insure further passive cooling and take full
advantage of the removal of the high power load from the LFI, and the choice of a larger opening angle ' 28◦
for the shadow cone (whose basis is the solar-shield and whose size mark out the outer limit of the V-grooves),
to allow more flexibility in orbit/scanning choices by enabling larger solar aspect angles. We then propose
to add alongside the SVM deployable petals covered with solar arrays and linked by a skin or canvas, as in
WMAP . While folded at launch, this corona will create additional power and shield the 1st V-groove. Then
in flight, the diameter of the shield will be ' 3.8 + 2 × 0.893 × tan(26.5dg) ' 3.8 + 0.893 = 4.7m allowing
ample room to fulfil the power need of COrE , which according to table 6 should be smaller than that of
Planck (which had a smaller diameter and a larger central tube). This solution as flown on WMAP .
As recalled earlier, the other evolutions with respect to Planck concerns the transmission (calling for an
HGA in the Ka band), and the introduction of a rather standard wheel assembly for nulling the angular
momentum and pointing.
The spacecraft TRL thus appear to be 9 for all the component, with the possible exception of the
transmitter.
33
6.8
Proposed procurement approach
We propose a ”standard” procurement approach: the service module, the launch and the operations will be
provided by ESA, while the instrument will be provided by a consortium of national space agencies. The
question of which part of the optical and cryogenic chain is to be provided by ESA and by the consortium
is addressed in the programmatic section § 8. The consortium of participating institutes will be in charge
of the developments, in strict coordination with industrial partners selected for experience, reliability, and
geographical returns.
6.9
Critical issues.
Criticalities for the instrument have been detailed in §4.12. No further issues identified at spacecraft level.
7
Science operations and archiving
Successful COrE science operations, defined as the handling and archiving of the raw data, the characterization of the instrument performance and response functions, the production of polarized maps, derived
science products, and inferred science constraints will be an integral part of the mission. Achieving these
goals on the time scale defined in the proprietary data policy section below, will be ensured through the
combination of the following three factors:
Robustness of the science data through hierarchies of data redundancy
The data is very redundant in the time, frequency and detector domains. Specifically, the same sky location is
revisited at the spin frequency, at many times throughout the mission, and polarization observed in different
orientations at the rotation frequency of the half-wave plate. Nearly all sky locations will be seen by all
detectors, and in all frequency bands, at many times. Hence, COrE is not very vulnerable to occasional data
loss, and thus does not require very stringent checking of all the data on a daily basis.
An early and sustained commitment to a fully-fledged simulation capability
Starting from the heritage of the Planck mission we will be able to make informed decisions.
A comfortably feasible scaling of the data processing infrastructure compared to the Planck
With ∼6400 detectors sampling at 500 Hz over the mission life of 4 years, the COrE data set will be 1.6
Petabyte large, not counting ancillary information which would increase it by ∼ 25% to 2 Petabytes. This is
large by today’s standards, but this amount of data will be comfortably handled and archived with plausible
assumptions about the evolution of data storage and processing costs for a time scale of 10 years to launch.
COrE is certainly well within the envelope of anticipated data volumes from other astronomical projects
such as LOFAR, LSST, ALMA8 , which will be operational on a similar time-scale.
The experience, existing resources and lessons learned from the successful pre-launch and post-launch
science operations of the Planck mission, are major assets for COrE science operations. Specifically the
Planck heritage endows COrE operations with
• a tested and working library of simulation and data analysis codes of all major parts of the pipeline, implementing state-of-the-art models of astrophysical foreground signals both galactic and extragalactic,
instrument optics, thermal and bolometer systematics, pointing, etc.;
• an European (and international) network of CMB data analysis experts built through the Planck
experience (key members of the Planck science operations team are involved in COrE );
• significant experience in the design of data base infrastructure for CMB analysis including tracking of
data provenance, and integrating a geographically distributed analysis team.
This heritage allows the COrE team to reap the fruit of the intensive development effort for Planck ’s science
operations capabilities, starting operations with many building blocks already ins place. Combined with a
sustained and organized development effort through the pre-launch phase, we are confident we can reach the
COrE science goals in a timely and efficient way.
In the following we outline a blue-print for a plausible analysis of the COrE data, consisting of a TimeOrdered-Information (TOI) processing and polarized map-making, followed by science analyses,
such as component separation, and finally scientific exploitation.
TOI processing:COrE will scan the sky for at least four years with a sampling rate of ≈ 2 ms. A single
datum taken by on of the detectors can be modeled as the addition of sky signal, time-varying systematics
and instrumental noise. The sky signal arises as the convolution of the detector’s polarized beam response
with the sky in the orientation defined by the detector placement in the focal plane, the direction of the
spin axis and the angle of rotation of the space craft about the spin axis. Time-varying systematics can
be internal (e.g. as a consequence of variations in thermal state of the instrument) or external (e.g. from
8
http://www.lofar.org/, http://www.lsst.org/, http://www.alma.nrao.edu
34
cosmic ray hits). Using the several levels of redundancy of the data we can jointly estimate the instrumental
parameters, such as the polarized beam response, systematic effects, and noise properties, together with the
pointing and calibration information. Residual systematics will be estimated and removed combining the
TOI with on-ground and in-flight calibration data (arising, e.g. from planets, bright polarized sources etc.)
Polarized map-making: Using the information extracted at the TOI processing stage we can compress
the TOI into maps of the polarization Stokes parameters of the pixelized sky at the instrument frequencies.
This compression is done in an iterative fashion, using improved estimates of the time-invariant sky signal
to help separate out systematic effects in the time-domain which in turn compresses into more faithful maps
of the sky signal9 . The resulting set of single-frequency maps are the basis for further scientific analysis.
Component separation: At this stage frequency domain information is used to identify and extract
the different signal contributions in the set of single-frequency maps such as the CMB, galactic emission
(synchrotron, free-free and dust) and catalogues of galactic and extragalactic sources. The single frequency
maps and component separated estimates are the starting point for the further scientific exploitation of the
COrE data, the exploration of cosmic origins described in the science plan.
Scientific exploitation: While component separation already brings a level of scientific interpretation
to the data analysis, the component maps will be the basis of further specialized analysis of the correlations
in the CMB polarization anisotropy to extract the anisotropy power spectra, B-mode constraints, higher
order correlation functions for non-Gaussianity tests and the CMB lensing signal. Similarly the magnetic
field tomography and studies of the ISM based on polarized maps of the dust emission derived from higher
frequency channels. These tasks will be organized by developing a structure of science working groups
interfaced with the Data Processing and Exploitation Center (see below).
Extrapolating the performance of existing methods and software tools for the TOI processing, we estimate
that to produce all the single-frequency maps from four years of data from all core detectors would take
about a week on an average present-day supercomputer with ∼ 1, 000 processing units. The growing power
of parallel-computing platforms, combined with improvements in efficiency of massively-parallel algorithms
and codes, should bring the estimated time to make maps down by at least an order of magnitude, permitting
processing of the entire data set on the time scale of a day.
It is thus likely that human resources will remain the real bottleneck for the time-domain processing
efficiency. We thus envisage a significant need for automated data-processing tools and methods which we
will develop pre-launch. COrE will be able to benefit from, and contribute to on-going development of such
tools in other contexts.
7.1
Science Operations Architecture and Share of Responsibilities
Organizationally, the data analysis (DA) effort will be focused around the COrE Data Processing and Exploitation Center (hereafter DPEC). Though structurally centralized, the DPEC will need to be geographically distributed. It will be composed of two closely interacting and partially overlapping teams, referred
to as the Infrastructure Development (ID) and Algorithm Development (AD) teams. The ID team will be
primarily responsible for the development of the data analysis framework and infrastructure. It will consist
of dedicated programmers and a few supporting scientists, some of which will belong to the AD team as
well. The major tasks of this team will include work on databases for storing and retrieving diverse data
products as well as a high performance Input/Output layer. Closer to the launch it will be also responsible
for creating and streamlining the DA pipeline assembled from the building blocks provided by the AD team.
The AD team, will take a responsibility for the development of the DA algorithms and software tools relevant
to all stages of the DA. Primarily, it will be composed of dedicated scientists covering a broad range of skills
relevant to the DA task. A single, focused team working on all the aspects of the pipeline will ensure a
better exploitation of the team resources and coordination of work. The AT team will contribute to the
groundwork for the infrastructure design, and later on, will become a primary tester and then user of the
DPEC built pipeline. The overlap between the members of both teams will ensure the necessary level of
coordination and communication between the teams. Both teams will work together on the DA of the COrE
data set once those are available. The DPECs ID team will also manage the major computational resources
of the collaboration. We anticipate that at the peak of the DA those will amount to a dedicated, massively
parallel 1 PetaFlop/s platform supplemented by high performance and capacity (O(1) PB) storage.
7.2
Archive approach
The data will be stored as time ordered raw-data sets (one TORD per detector) during the survey. Cleaned,
calibrated timelines will be processed from the TORD, and stored for the subsequent polarized map-making
and spectral estimations. Maps and catalogues at each frequencies will then be produced.
These final data products as well as the necessary ancillary data (beams, calibration information, etc.)
should be archived as part of ESA’s facilities (as the Planck legacy archive), to serve in the long run the
scientific community and help it to exploit the extremely low noise and systematics free full sky data set
9
Time-varying sources are detected and masked in this process and then treated separately.
35
produced. The TORD themselves should also be retained in a long term repository, but would only be served
on special request after some refereeing process.
7.3
Proprietary data policy
The data will be pre-analyzed by the DPEC in the two years following the end of the survey. The resulting
data products will be calibrated maps of the Stokes parameters, one for each observed frequency. The DPEC
will be in charge of producing a public data release together with the first round of scientific publications on
the results of COrE .
In order to best serve the community, while remaining realistic concerning the duration needed to analyze
such a vast and complex data set to the exacting standards of precision cosmology, we propose an early release
of non-CMB data 2.5 years after the beginning of the survey (which itself should be 2 to 4 months after launch
) on the basis of the first two complete sky coverages. Indeed this would provide 15 months of analysis of the
first 15 months of survey. While very tight, this should be achievable with extensive pre-launch preparation,
and simulation activities assuming the mission goes well. We will analyze during the feasibility study the
exact specifications for these early products. One can imagine producing and releasing to the community
a COrE Early Compact Source Sample (CoCoSS), as well as some high-frequency maps at 350 GHz and
higher, at least the temperature part.
The release of the data obtained during the 4 years of nominal mission duration will be two years after
the end of the data taking, i.e. about 3 years later than the early release, together with the first round of
scientific publications on the cosmological results of COrE . The data products will be calibrated maps of the
Stokes parameters, one for each observed frequency and some relevant ancillary data to characterize them (
e.g. beams). The DPEC will be in charge of producing the public data releases.
8
Programmatic and cost analysis
8.1
Overall proposed mission management structure
Principal Investigator (PI)
Co-PIs (one per institute)
Overall
Project
Manager
Local
Project
Managers
(each
institute)
Co-Is
Instrument Scientist (IS)
Project Scientist (PS)
Calibration
Scientist
System manager
System engineers
• Thermal Arch.
• Mech. Arch.
• Optical Arch.
• Elec. Arch.
•
•
Steering Committee
PI, Co-PIs, IS, PS, PM
Scanning strategy
Scientist
Level coordinators
(for each area)
• DPC
• Level 1
•
• Level n
• Level S
Instr. Team
coordinators
(for each area)
• Optics
• Cryogeny
• Electronics
• Polarimetry
Ground Calib. Sc.
In-flight Calib. Sc.
Instrument Working Group
Data Processing Group
Figure 17: Proposed Management Structure for COrE .
The mission management structure is summarized in Fig. 17.
8.2
Mission schedule drivers (technology developments, etc.)
We do not have astronomical drivers: the measurement of the CMB can be performed any time, and
calibration sources are available about twice a year for any launch date.
In terms of technology, a successful implementation of the proposal for a launch in the 2010-2012 window
of the M3 AO calls for bringing the RHWP to a TRL of 5 by the end of 2014 at the time of the down
selection for implementation. This is the main driver.
8.3
Payload/Instrument Cost
This section provides a rough order of magnitude (ROM) cost for the satellite and for the instrument. Given
the very large commonalities with the Planck mission, of which COrE represents in many ways an ideal
36
COrE Consortium budget
Subsystem
ROM Cost [Me]
Management and Syst. Engineering
3
Rotating HWP assembly
12
0.1 K CC Dilution
11
Feed horns arrays
5
Quasi-optical filters
5
Detector arrays
12
Focal plane integration and test
10
Cold electronics/multiplexer
18
Warm readout electronics
12
Data Processing Unit
5
Instrument Level Integration
10
Instrument Level Tests
10
COrE Instrument sub-total
113
COrE Science Ground Segment
60
COrE Education & outreach
2
COrE Consortium overall cost
175
Potentially provided by
FR, IT
IT, UK, USA
FR, USA
IT, UK, USA
UK
FR, IT, UK, USA
FR, IT, UK, USA
FR, IT, UK, USA
FR, IT, SP
FR, IT, SP
FR, IT
FR, IT
Consortium
Consortium
Table 7: COrE instrumental consortium preliminary cost breakdown.
continuation, we have as often as possible taken Planck as a basis for our costing estimate. COrE spacecraft
is a direct heritor of the Planck one, reusing nearly all of its flight-proven developments (in particular
concerning the overall thermal architecture), with a less demanding orbit around L2, à la Herschel . On
the instrument side, COrE has many more detectors than Planck , but uses new technologies for massproduction of these devices. Most importantly, thanks to advances in bolometer technology, there is only
one instrument in COrE which can cover all the necessary frequency range. Apart from the obvious saving of
removing entirely an expensive instrument which accounted for maybe 40% of the overall instrumental cost,
this also results in an easier integration, calibration and management of the system. Overall, this should
lead to a significant cost reduction of COrE with respect to Planck .
The main subsystems of the instrument will be developed by a single consortium of the scientific institutions involved in COrE , and funded by the national agencies. In table 7 we report a preliminary breakdown
of the costs, and we mention countries who are potential providers of these specific subsystems; that list is
obviously not closed, but rather destined at showing that there are several interested parties to provide each
subsystem. As for Planck , we envisage a substantial involvement of our American colleagues.
One should note from Table 7 that we have assumed here that the first element of the optical system, i.e.
the RHWP, would be provided by the consortium and that all the (rest of the) telescope would be provided
by ESA. We have also assumed that the first part of the cryogenic chain is not part of the instrumental
deliveries, considering in particular that some of the possible coolers might very well be industrial deliveries.
Only the last part which is bound to be an integral part of the instrument design was kept in the instrument
part.
8.3.1
Assumed share of payload costs to ESA
Taken at first value, the split of responsibilities/deliveries described in the previous section between ESA
and the COrE instrument consortium would lead to an EU countries contribution of 175 Me to cover the
instrumental consortium related costs, ESA contributing the remaining 434 Me ( cf. § 8.4 and Table 8, which
includes 10% of that estimate of 434 Me as ESA’s internal cost).
But we think it might be advantageous to all parties to engage in a collaboration with NASA and our US
colleagues at a level up to 130 Me, which is quite compatible with the 200 M$ cap mentioned in the recently
released decadal report10 . The general idea of the proposed collaborative scheme is that the USA would
both participate at the similar percentage level to the overall cost of the COrE program and to the cost of
the instrumental realization and of the science ground segment. Even though it is too early to freeze any
detailed choice (before even a pre-phase A study took place), given the array of technologies developed in
the USA, it is anticipated that the US contribution, if funded, would naturally lead to provide parts of the
focal plane (for instance some of the frequency channels), and of the detectors electronics chain; the RHWP
10
Even after accounting for extra costs on the NASA side like participating in activities during the definition phases, or the
scientific exploitation of products during the proprietary phase, often at the 25% level of the DPC cost
37
COrE financial budget [Me]
1. Soyuz Launch
2. Spacecraft
80
374
2.1
2.2
2.3
2.4
Project Office
Engineering
AIT
SVM
2.4.1 Structure
8
2.4.2 Thermal
2
2.4.3 Power
21
2.4.4 AOCS
25
2.4.5 TTC TMCU 10
2.4.2 Avionics
26
2.5 Telescope (Structure+PM+SR)
2.6 Active Cryogenic chain (→ 1.7K)
2.7 COrE Instrument
3. ESA Ground Segment (MOC+SOC)
4. ESA internal cost (∼ 10% of ESA cost)
5. COrE Science ground Segment (DPC, not science)
6. COrE Education & outreach
7. COrE Science exploitation
Overall COrE program cost
20
50
10
92
30
59
113
50
43
60
2
619
Table 8: Overall COrE cost breakdown, excluding science exploitation of DPC products
COrE Program budget [Me]
ESA
349
NASA
130
EU countries
140
COrE program total
619
COrE Program budget [Me]
ESA
434
EU countries
175
COrE program total
619
Table 9: COrE program main contribution, excluding science exploitation, in the 2 cases considered.
might also be a transatlantic collaboration, as well as the Data Processing. One could also envisage, at the
frontier between ESA and instrument provision, the provision by the US of some of the cooling stages (the
3 stages of Planck active cooling chain were provided by 3 different countries, namely the US, UK, and FR),
as well as the wheel assembly or the telescope assembly.
8.3.2
Estimated non-ESA payload costs
See above.
8.4
Overall mission cost analysis
Table 8 summarizes our Rough Order of Magnitude estimate of the overall cost of the proposed COrE
program and Table 9 recall the expected breakdown overall between the ESA, EU and non EU countries,
considering the case, or not, of a US participation.
9
Communication and Outreach
The quest to understand the origins of our Universe has invariably provoked great interest among the
broad public. As part of the COrE mission, we propose hiring a full-time communications officer, whose
responsibilities would include (1) developing an extensive web-site with information concerning cosmology,
particle physics, and inflation suitable for students at a variety of levels, (2) liaising with the scientific and
38
general press regarding the COrE mission, and (3) organizing and publicizing a speakers’ bureau that would
bring COrE scientists in contact with school groups.
The outreach activities will focus on communicating the excitement of a single mission that has the
potential to answer questions ranging from the origin of the Universe, the origin of cosmic structure, neutrino
physics, and the origin of stars. To the extent possible we will focus on not just presenting information to the
public but to provide immersive and participatory experiences. The COrE website will include opportunities
to participate in distributed computing projects related to COrE science objectives, thereby involving the
general public in the scientific exploitation of COrE data. In addition, we will develop a presence of the
COrE mission on social media with opportunities for the public to interact with COrE scientists.
10
Selected references
We provide here an abridged set of essential references concerning the polarization, CMB, and sub-millimeter
science and instrumentation. A more complete bibliography as well as supplemental and expanded information relating to certain sections of this proposal may be found the COrE website:
www.core-mission.org/proposal/supplements
The COrE team has benefited from experience in a variety of suborbital and space CMB experiments, including most notably Boomerang, CLOVER, and PLANCK (http://www.esa.int/SPECIALS/Planck/index.html).
We have also benefited from several B-mode polarization satellite studies starting with the Frech CNESsponsored SAMPAN study and the Italian COFIS study of ASI, leading to European B-Pol proposal three
years ago. (www.b-pol.org) A more detailed American CMBPol study followed on this heritage, producing
a number of useful and detailed white papers, which broadly confirm the conclusions of the present study.
In particular, the following papers provide useful details and extensive references:
J Dunkley et al., “CMBPol Mission Concept Study: Prospects for polarized foreground removal,”
(arXiv:0811.3915[astro-ph]) (2008)
A Fraisse et al., “CMBPol Mission Concept Study: Foreground Science Knowledge and Prospects,”
(arXiv:0811.3920[astro-ph]) (2008)
D Baumann et al., “CMBPol Mission Concept Study: Probing Inflation with CMB Polarization,”
(arXiv:0811.3919[astro-ph]) (2008)
M Zaldarriaga et al., “CMBPol Mission Concept Study: Reionization Science with the Cosmic
Microwave Background,” (arXiv:0811.3918[astro-ph]) (2008)
K Smith et al., “CMBPol Mission Concept Study: Gravitational Lensing,” (arXiv:0811.3916[astroph]) (2008)
More information is at http://cmbpol.uchicago.edu/. We following publications are particularly relevant:
Reichborn-Kjennerud B. et al., “EBEX: A balloon-borne CMB polarization experiment,” Proceedings of the SPIE, Volume 7741, pp. 77411C-77411C-12 (2010) (arXiv:1007.3672)
D Chuss, “Quasioptical Reflective Polarization Modulation for the Beyond Einstein Inflation Probe”
in Polarization Modulators for CMBPol, J of Phys: Conf. Series 155 (2009).
L Verde, H Peiris & R Jimenez, “Optimizing CMB polarization experiments to constrain inflationary physics,” JCAP 0601, 019 (2006) (astro-ph/0506036)
de Zotti, G., et al., “Radio and millimeter continuum surveys and their astrophysical implications”,
(A&ARv 18, 1) (2010).
Lagache, G., et al., “Dusty Infrared Galaxies: Sources of the Cosmic Infrared Background”,
(ARA&A Annual 43, 727) (2005)
J Fergusson, M Liguori, P Shellard, “General CMB and Primordial Bispectrum Estimation I: Mode
Expansion, Map-Making and Measures of fN L ,” Phys. Rev.D82, 023502 (2010) (0912.5516/astroph.CO)
A Lewis & A Challinor, “Weak gravitational lensing of the CMB,” Phys. Rept. 429, 1 (2006)
(astro-ph/0601594)
C McKee & E Ostriker, “Theory of Star Formation,” ARAA 45, 565 (2007)
J Delabrouille & J Cardoso, “Diffuse Source Separation in CMB Observations,” (astro-ph/0702198)
S Leach et al., “Component separation methods for the PLANCK mission,” A& A 491, 597 (2008)
(arXiv:0805.0269)
A more detailed bibliography arranged section by section appears on the COrE website cited above.
39