CMB is radiation from the early universe, emitted when the universe was about 400 000 years old, and is coming from all directions. This radiation reflects the properties and structure of the early universe as well as the later history of the universe through which it has traveled.
CMB measurements are demanding because the anisotropy and polarization signals are so weak. They require extremely sensitive instruments, ideal observational conditions, and sophisticated data analysis to bring out their full potential. However, the weakness of the signal is a reflection of an important property of the CMB: We are observing perturbations in the early universe, which are still small enough, that their evolution has been linear. Thus their evolution can be calculated backwards to the original primordial perturbations generated in the very early universe. CMB is practically the only observational window to the extremely high energy physics of the very early universe, where these seeds for the structure of the universe were generated.
CMB measurements are contaminated by microwave radiation from our own galaxy and extragalactic objects. To be able to remove this "foreground", the observations are carried out at many different frequencies. While being a nuisance to cosmology, this foreground is of great interest to astronomers. The satellite CMB experiments measure the microwave sky at frequencies not accessible form earth-based observatories.
Currently the most important cosmological data set are the CMB observations of the WMAP satellite (NASA) [1]. The Planck satellite of the European Space Agency (ESA) [2] is the successor to WMAP. Planck will surpass WMAP in resolution, sensitivity and frequency coverage. While the WMAP data has already allowed a determination of several central cosmological parameters with good accuracy [3], these determinations are based on a number of simplicity assumptions, which need to be tested by more accurate and extensive measurements, which Planck will provide. Important questions remaining for Planck include the nature of primordial perturbations (adiabaticity, Gaussianity, scale-dependence, existence of tensor perturbations), which is the key to the origin of these perturbations. WMAP has also raised a number of questions, for whose answers Planck data will be needed. The WMAP observations at the largest angular scales show unexpected features challenging standard cosmological theory [4]. These results may have been compromised by foreground contamination in the data. The wider frequency coverage of Planck will allow a better separation of the foreground from the CMB.
The CMB observations can be expressed in terms of angular power spectra, which show the dependence of the anisotropies and polarization on the distance scale. These spectra show characteristic "acoustic peaks" related to oscillations in the primordial plasma [5]. The WMAP resolution reaches halfway to the third peak, whereas the Planck resolution is almost three times better reaching to the sixth or seventh peak. This will provide a long lever arm so that the scale dependence of the perturbations can be measured. There are four such spectra related to temperature anisotropy, two components of polarization, and the correlation of temperature with polarization. The higher sensitivity of Planck will be especially important for polarization measurements. WMAP has only been able to publish temperature and correlation spectra, whereas Planck will measure also the polarization spectra. The number of cosmological parameters that can be determined from the data is proportional to the number of spectra, and the number of acoustic peaks in these spectra. Thus Planck will be able to test many cosmological hypotheses beyond the reach of WMAP.
A cornerstone of Finnish participation to Planck is the set of 70 GHz detectors built in Finland by Ylinen Electronics under the supervision of Millilab (VTT). This 70 GHz channel is one of the most important of the 9 frequency channels in Planck, since at this frequency the foreground contamination is the smallest.
Finnish scientists are participating in Planck both in foreground studies (Metsahovi and Tuorla Observatories, and the Observatory of the University of Helsinki) and in cosmology (Department of Physical Sciences, University of Helsinki, and Helsinki Institute of Physics). This cosmology effort is described here. For the foreground (astronomy) effort see the Metsahovi and Tuorla Planck page and the Helsinki Observatory Planck page.
The cosmology effort consist of both a theoretical effort preparing for the utilization of the Planck data, and a data analysis effort within the Planck collaboration, which contributes to the success of the Planck mission, earns Finnish cosmologists the rights to the Planck data, and provides us with a detailed understanding of it necessary for a reliable use of the data.
Planck has detectors for 9 frequency channels. These detectors are divided into two instruments, the Low-Frequency Instrument (LFI) and the High-Frequency Instrument (HFI), which use different detection technologies. The Finnish 70 GHz detectors belong to the LFI. The Planck organization has separate data processing centers (DPC) for each instrument. (Part of the data analysis will be done separately, part jointly). The Planck Level-S center (Garching, Germany) is responsible for generating simulated data [6], for testing analysis methods.
The Planck collaboration is divided into working groups (WG), which have responsibilities at different parts of the data analysis. They work in collaboration with the DPCs, who are responsible for the final implementation of the data analysis.
The analysis begins by preprocessing and cleaning of the data from various artifacts and systematic effects (WG1). The outcome of this is the time-ordered data (TOD) from each detector.
Then frequency maps are constructed for each of the 9 frequency channels (WG3). Because of the extreme weakness of the signal this is a nontrivial task. A naïve mapping of the TOD on the sky would be dominated by noise, and the signal has to be extracted from the data using statistical methods. Correlated noise can be removed utilizing the fact that the same point on the sky is repeatedly observed over many different timescales. To do this as well as possible requires working simultaneously with all the data pertaining to the same frequency. In practice this means that this whole data need to be kept in memory simultaneously, requiring a large supercomputer. After the TODs have been compressed into frequency maps, the memory requirements of further processing are smaller.
In component separation (WG2) component maps (CMB and various foreground components) are extracted from the frequency maps. The foreground components are then the responsibility of WG5, WG6, and WG7.
From the CMB temperature and polarization maps, the angular power spectra are determined (WG3). Cosmological parameters are estimated from these spectra (WG3). There will be many different determinations corresponding to more restricted and more general cosmological models, with different numbers of relevant cosmological parameters. Many cosmological hypotheses will be tested this way.
If the primordial perturbations are Gaussian, all cosmological information is contained in the angular power spectra. Otherwise the CMB maps contain additional information (WG4). Different cosmological models predict Gaussianity to a different degree; and for some models the deviation will be observable by Planck.
Systematic effects affecting map-making to be studied include subpixel structure in the sky, detector beam shapes, gaps in the data, and features in the detector and cooler noise. To correct for the beam shapes there are two approaches: a deconvolution map-making method [10] and a MASTER [11] type correction in the power spectrum estimation stage.
In the future the work will shift more towards the later stage of data analysis: power spectrum estimation [12] and cosmological parameter determination. An important part of this is the study of error propagation. Errors in different map pixels are correlated as are errors in the different multipoles of the power spectra. These are properly carried by covariance matrices, but for high-resolution maps these matrices are prohibitively large, requiring an approximate solution. A hybrid scheme employing different approximations at large and small scales has been proposed [13]. Parameter estimation in a high-dimensional parameter space is also a challenging task. Markov Chain Monte Carlo [14] is the current baseline method.
Planck data will also be used to constrain models of dark energy [23], although here the role of other cosmological data [24,25] becomes relatively more important. An alternative for dark energy is modification of General Relativity to account for the accelerated expansion of the universe.