We present a version of the Fourier Bessel method first introduced by Fisher et al (1994) and Zaroubi et al (1994) with two extensions: (a) we amend the formalism to allow a generic galaxy weight which can be constant rather than the more conventional overweighting of galaxies at high distances, and (b) we correct for the masked zones by extrapolation of Fourier Bessel modes rather than by cloning from the galaxy distribution in neighbouring regions. We test the procedure extensively on N-body simulations and find that it gives generally unbiased results but that the reconstructed velocities tend to be overpredicted in high-density regions. Applying the formalism to the PSZz redshift catalog, we find that beta = 0.7 +/- 0.5 from a comparison of the reconstructed Local Group velocity to the CMB dipole. From an anisotropy test of the velocity field, we find that beta = 1 CDM models models normalized to the current cluster abundance can be excluded with 90% confidence. The density and velocity fields reconstructed agree with the fields found by Branchini et al (1998) in most points. We find a back-infall into the Great Attractor region (Hydra-Centaurus region) but tests suggest that this may be an artifact. We identify all the major clusters in our density field and confirm the existence of some previously identified possible ones.

We use rich clusters of galaxies in the Northern and Southern Galactic hemispheres up to a redshift z=0.12 to determine the cluster correlation function. We show that superclusters of galaxies and voids between them form a moderately regular network. As a result the correlation function determined for clusters located in rich superclusters oscillates: it has a series of regularly spaced secondary maxima and minima. The scale of the supercluster-void network, determined from the period of oscillations, is P=115 Mpc. The amplitude of oscillations is about 3 times larger than the estimated error. We compare the observed cluster correlation function with similar functions derived for popular models of structure formation. We find that the production of the observed cluster correlation function in any model with a smooth transition of the power spectrum from a Harrison-Zeldovich regime with positive spectral index on long wavelengths to a negative spectral index on short wavelengths is highly unlikely. The power spectrum must have an extra peak located at the wavelength equal to the period of oscillations of the correlation function.

We investigate properties of the correlation function of clusters of galaxies using geometrical models. On small scales the correlation function depends on the shape and the size of superclusters. On large scales it describes the geometry of the distribution of superclusters. If superclusters are distributed randomly then the correlation function on large scales is featureless. If superclusters and voids have a tendency to form a regular lattice then the correlation function on large scales has quasi-regularly spaced maxima and minima of decaying amplitude; i.e., it is oscillating. The period of oscillations is equal to the step size of the grid of the lattice. We calculate the power spectrum for our models and compare the geometrical information of the correlation function with other statistics. We find that geometric properties (the regularity of the distribution of clusters on large scales) are better quantified by the correlation function. We also analyse errors in the correlation function and the power spectrum by generating random realizations of models and finding the scatter of these realizations.

Using a new compilation of available data on galaxy clusters and superclusters we present evidence for a quasiregular three-dimensional network of rich superclusters and voids, with the regions of high density separated by about 120 Mpc. We calculate the power spectrum for clusters of galaxies; it has a peak on the wavelength equal to the step of the network; the excess in the amplitude of the spectrum over that of the cold dark matter model is by a factor of 1.4. The probability that the spectrum can be formed within the framework of the standard cosmogony is very small. If the cluster distribution reflects the distribution of all matter (luminous and dark), then there must exists some hithero unknown process that produces regular structure on large scales.

We have developed a parameter-independent method to detect local maxima of the two-point correlation function. By applying it to two samples of rich Abell clusters of galaxies with redshift limits z<0.08 and z<0.12 we detect three maxima centered at 150 Mpc, 300 Mpc and 430 Mpc with confidence levels 80% and higher. This sequence of fluctuations has an average interval of 140 Mpc, that can be explained by a power spectrum with a distinct peak at k=0.048+-0.005/Mpc.

Recently, the observed cellular nature of the large-scale structure of the Universe with its quasi-regular pattern of superclusters and voids has been pointed out by several authors. In this paper, we investigate properties of the initial power spectrum which lead to prediction of structure consistent with these observations.For this purpose, we analyze the evolution of structure within four sets of 2- and 3-dimensional cosmological models, which differ in their initial power spectrum. The models include HDM and CDM models as well as double power-law models. We discuss in detail the impact of model parameters such as the large scale and small scale power and the position and height of the maxima of the power spectra on the predicted structure. The best agreement with observations was observed in the model with the Harrison-Zeldovich spectrum on large scales, a power index $n\approx -1.5$ on small scales, and a maximum of the power spectrum at $\approx 130$~\Mpc. In this model the distribution of masses of clusters and superclusters, the correlation function of clusters, and the void distribution reproduce well the respective observed distributions. We also investigated the dynamical evolution of the supercluster-void structure. The results show that the basic supercluster-void network is formed very early and is essentially given by initial conditions.