List of Figures

	Figure 1: The thermal history of the Universe.
	Figure 2: Power spectrum for standard CDM. Parameters assumed are $Ω = 1$ , $n = 1$ , $H0$ = 50 km s^–1 Mpc^–1 and a baryon fraction of $Ωb = 0.04$ .
	Figure 3: Comparison of defect model predictions to current experimental data. All models were COBE normalised at $ℓ = 10$ . This figure was taken from Pen, Seljak & Turok [67].
	Figure 4: Matter power spectra computed from Boltzmann code summed over the eigenmodes. The upper curve shows the standard cold dark matter (sCDM) power spectrum. The defects generally have more power on small scales than large scales relative to the adiabatic sCDM model. The data points show the mass power spectrum as inferred from the galaxy distribution. This figure was taken from Pen, Seljak & Turok [67].
	Figure 5: Comparison between the observations and predictions for local strings. Shown are the results from simulations with $ΩΛ$ of 0.0 (short-long dashed), 0.3 (dotted-dashed), 0.5 (short dashed), 0.7 (long dashed), 0.9 (solid line) and standard CDM (dot-dashed).
	Figure 6: The expected level of the various anisotropic foregrounds for a typical CMB experiment (resolution $∼$ 1°) observing in the best low Galactic flux regions.
	Figure 7: A typical region of sky over a range of frequencies covered by CMB experiments. Synchrotron dominates at 1 GHz, free-free, synchrotron and CMB at 10 GHz, CMB at 100 GHz and dust at 1000 GHz. Flux is in $ΔT$ equivalent temperature in $μK$ . No atmospheric emission has been added. The CMB model shown here is for a standard cold dark matter model with $h = 0.5$ and $Ωb = 0.05$ .
	Figure 8: The frequency dependences of the thermal and kinetic SZ effects expressed as a brightness temperature change (top) and intensity change (bottom).
	Figure 9: Ryle telescope map of the sky towards quasar pair PC1643+4631 (the positions of which are shown as crosses). The central decrement is about $− 600 μK$ .
	Figure 10: SZ source counts with observational constraints, as a function of SZ flux density expressed at 400 GHz. The two hatched boxes show the 95% one-sided confidence limits from the VLA and the RT; due to the uncertain redshift of the clusters, there is a range of possible total SZ flux density, which has for a minimum the value observed in each beam and a maximum chosen here to correspond to $z > 1$ . From the SuZIE blank fields, one can deduce the 95% upper limit shown as the triangle pointing downwards (Church et al. [18]). We also plot the predictions of our fiducial open model ( $Ω = 0.2$ ) for all clusters (dashed line) and for those clusters with $z > 4$ .
	Figure 11: The COBE DMR 4 year data displayed as all-sky maps.
	Figure 12: Comparison between the maximum entropy reconstruction of the Tenerife Dec. 35° data at 15 GHz (solid line) and the COBE DMR predictions of Bunn et al. (1996) (dashed line) at 53 and 90 GHz.
	Figure 13: Saskatoon 3 year map showing region analysed as compared to the COBE full sky coverage.
	Figure 14: 16.5 GHz CAT image of 6° × 6° area centred on the CAT2 field, after discrete sources have been subtracted. Excess power can be seen in the central 2° × 2° primary beam (because the sensitivity drops sharply outside this area, the outer regions are a good indicator of the noise level on the map). The flux density range scale spans ±40 mJy per beam.
	Figure 15: Recent results from various CMB experiments. The solid line is the prediction (normalised to COBE) for standard CDM with $Ω∘ = 1.0$ , $Ωb = 0.1$ and $H ∘$ = 45 km s^–1 Mpc^–1. The Saskatoon points have a 14% calibration error.
	Figure 16: Artist’s impression of the completed VSA array.
	Figure 17: Artist’s impression of the MAP Satellite.
	Figure 18: Artist’s impression of the Planck Surveyor Satellite (formally COBRAS/SAMBA)
	Figure 19: Expected capability of a satellite experiment as a function of resolution. The percentage error in recovering cosmological parameters from the CMB power spectrum is shown versus the resolution available. This figure is taken from Bersanelli et al. 1996 [10].
	Figure 20: The left hand side shows the input maps used in the Planck simulations for a CDM simulation of the CMB and the thermal SZ effect. The right hand side shows the reconstructions obtained by MEM. It is easily seen that MEM does a very good job at reconstructed these two components. For comparison, the grey scales on the input maps are the same as on the reconstructed maps.
	Figure 21: Analytic fit to power spectrum versus experimental points. (From Hancock et al. [35], 1997.)
	Figure 22: The contours show 50, 5, 2 and 0.1 percentile likelihood contours for pairs of parameters determined from fits to the CMB power spectrum. The figures to the left show results for an experiment with resolution $∘ 𝜃FWHM = 1$ . Those to the right for a higher resolution experiment with $′ 𝜃FWHM = 10$ plotted on the same scale (central column) and with expanded scales (rightmost column). This figure is taken from Bersanelli et al. 1996 [10].