The dynamic evolution of the baryonic intergalactic medium (IGM) caused by the underlying dark matter gravity is governed by the Navier-Stokes equations in which many cooling and heating processes are involved. However, it has long been recognized that the growth mode dynamics of cosmic matter clustering can be sketched by a random force driven Burgers' equation if cooling and heating are ignored. Just how well the dynamics of the IGM can be described as a Burgers fluid has not been fully investigated probably because cooling and heating are essential for a detailed understanding of the IGM. Using IGM samples produced by a cosmological hydrodynamic simulation in which heating and cooling processes are properly accounted for, we show that the IGM velocity field in the nonlinear regime shows the features of a Burgers fluid, that is, when the Reynolds number is high, the velocity field consists of an ensemble of shocks. Consequently, (1) the IGM velocity $v$ is generally smaller than that of dark matter; (2) for the smoothed field, the IGM velocity shows tight correlation with dark matter given by $v \simeq s v_{dm}$, with $s<1$, such that the lower the redshift, the smaller $s$; (3) the velocity PDFs are asymmetric between acceleration and deceleration events; (4) the PDF of velocity difference $\Delta v=v(x+r)-v(x)$ satisfies the scaling relation for a Burgers fluid, i.e., $P(\Delta v)=(1 r^y)F(\Delta v/r^y)$. We find the scaling function and parameters for the IGM which are applicable to the entire scale range of the samples (0.26 - 8 h$^{-1}$ Mpc). These properties show that the similarity mapping between the IGM and dark matter is violated on scales much larger than the Jeans length of the IGM.

We investigate the relationship between the mass and velocity fields of the intergalactic medium (IGM) and dark matter. Although the evolution of the IGM is dynamically governed by the gravity of the underlying dark matter field, some statistical properties of the IGM inevitably decouple from those of the dark matter once the nonlinearity of the dynamical equations and the stochastic nature of the field is considered. With simulation samples produced by a hybrid cosmological hydrodynamic/N-body code, which is effective in capturing shocks and complicated structures with high precision, we find that the one-point distributions of the IGM field are systematically different from that of dark matter as follows: 1.) the one-point distribution of the IGM peculiar velocity field is exponential at least at redshifts less than 2, while the dark matter velocity field is close to a Gaussian field; 2.) although the one-point distributions of the IGM and dark matter are similar, the point-by-point correlation between the IGM and dark matter density fields significantly differs on all scales and redshifts analyzed; 3.) the one-point density distributions of the difference between IGM and dark matter fields are highly non-Gaussian and long tailed. These discrepancies violate the similarity between the IGM and dark matter and cannot be explained simply as Jeans smoothing of the IGM. However, these statistical discrepancies are consistent with the fluids described by stochastic-force driven nonlinear dynamics.

It has been recently found that the local fluctuations of the QSO's Ly$\alpha$ absorption spectrum transmitted flux show spiky structures. This implies that the mass fields of the intergalactic medium (IGM) is intermittent. This feature cannot be explained by the clustering evolution of cosmic mass field in the linear regimes and is also difficult to incorporate into the hierarchical clustering scenario. We calculate the structure functions and intermittent exponent of the IGM and HI for full hydrodynamical simulation samples. The result shows the intermittent features of the Ly$\alpha$ transmitted flux fluctuations as well as the mass field of the IGM. We find that within the error bars of current data, all the intermittent behavior of the simulation samples are consistent with the observation. This result is different from our earlier result (Pando et al 2002), which shows that the intermittent behavior of samples generated by pseudo-hydro simulation cannot be fitted with observed data. One difference between the pseudo-hydro and full hydro simulations is in treating the dynamical relation between the IGM (or HI) and dark matter fields. The former assumes that the IGM density distribution traces the underlying dark matter point-by-point on scales larger than the Jeans length in either the linear or nonlinear regimes. However, hydrodynamic studies have found that a statistical discrepancy between the IGM field and underlying dark matter in nonlinear regime is possible. We find that the point-by-point correlation between the IGM density perturbations and dark matter become weaker on comoving scales less than 2 h$^{-1}$ Mpc (in LCDM model), which is larger than the IGM Jeans length.

We calculate the structure function and intermittent exponent of the 1.) Keck data, which consists of 29 high resolution, high signal to noise ratio (S/N) QSO Ly$\alpha$ absorption spectra, and 2.)the Ly$\alpha$ forest simulation samples produced via the pseudo hydro scheme for the low density cold dark matter (LCDM) model and warm dark matter (WDM) model with particle mass $m_W=300, 600, 800$ and 1000 eV. These two measures detect not only non-gaussianities, but also the type of non-gaussianty in the the field. We find that, 1.) the structure functions of the simulation samples are significantly larger than that of Keck data on scales less than about 100 h$^{-1}$ kpc, 2.) the intermittent exponent of the simulation samples is more negative than that of Keck data on all redshifts considered, 3.) the order-dependence of the structure functions of simulation samples are closer to the intermittency of hierarchical clustering on all scales, while the Keck data are closer to a lognormal field on small scales. These differences are independent of noise and show that the intermittent evolution modeled by the pseudo-hydro simulation is substantially different from observations, even though they are in good agreement in terms of second and lower order statistics. (Abridged)

The hierarchical clustering model of the cosmic mass field is examined in the context of intermittency. We show that the mass field satisfying the correlation hierarchy $\xi_n\simeq Q_n(\xi_2)^{n-1}$ is intermittent if $\kappa < d$, where $d$ is the dimension of the field, and $\kappa$ is the power-law index of the non-linear power spectrum in the discrete wavelet transform (DWT) representation. We also find that a field with singular clustering can be described by hierarchical clustering models with scale-dependent coefficients $Q_n$ and that this scale-dependence is completely determined by the intermittent exponent and $\kappa$. Moreover, the singular exponents of a field can be calculated by the asymptotic behavior of $Q_n$ when $n$ is large. Applying this result to the transmitted flux of HS1700 Ly$\alpha$ forests, we find that the underlying mass field of the Ly$\alpha$ forests is significantly intermittent. On physical scales less than about 2.0 h$^{-1}$ Mpc, the observed intermittent behavior is qualitatively different from the prediction of the hierarchical clustering with constant $Q_n$. The observations, however, do show the existence of an asymptotic value for the singular exponents. Therefore, the mass field can be described by the hierarchical clustering model with scale-dependent $Q_n$. The singular exponent indicates that the cosmic mass field at redshift $\sim 2$ is weakly singular at least on physical scales as small as 10 h$^{-1}$ kpc.

We investigate the weakly non-linear evolution of cosmic gravitational clustering in phase space by looking at the Zel'dovich solution in the discrete wavelet transform (DWT) representation. We show that if the initial perturbations are Gaussian, the relation between the evolved DWT mode and the initial perturbations in the weakly non-linear regime is quasi-local. That is, the evolved density perturbations are mainly determined by the initial perturbations localized in the same spatial range. Furthermore, we show that the evolved mode is monotonically related to the initial perturbed mode. Thus large (small) perturbed modes statistically correspond to the large (small) initial perturbed modes. We test this prediction by using QSO Ly$\alpha$ absorption samples. The results show that the weakly non-linear features for both the transmitted flux and identified forest lines are quasi-localized. The locality and monotonic properties provide a solid basis for a DWT scale-by-scale Gaussianization reconstruction algorithm proposed by Feng & Fang (Feng & Fang, 2000) for data in the weakly non-linear regime. With the Zel'dovich solution, we find also that the major non-Gaussianity caused by the weakly non-linear evolution is local scale-scale correlations. Therefore, to have a precise recovery of the initial Gaussian mass field, it is essential to remove the scale-scale correlations.

We show that scale-scale correlations are a generic feature of slow-roll inflation theories. These correlations result from the long-time tails characteristic of the time dependent correlations because the long wavelength density perturbation modes are diffusion-like. A relationship between the scale-scale correlations and time-correlations is established providing a way to reveal the time correlations of the perturbations during inflation. This mechanism provides for a testable prediction that the scale-scale correlations at two different spatial points will vanish.

We perform a discrete wavelet analysis of the COBE-DMR 4yr sky maps and find a significant scale-scale correlation on angular scales from about 11 to 22 degrees, only in the DMR face centered on the North Galactic Pole. This non-Gaussian signature does not arise either from the known foregrounds or the correlated noise maps, nor is it consistent with upper limits on the residual systematic errors in the DMR maps. Either the scale-scale correlations are caused by an unknown foreground contaminate or systematic errors on angular scales as large as 22 degrees, or the standard inflation plus cold dark matter paradigm is ruled out at the $> 99%$ confidence level.

We present an analysis of the non-Gaussianity in the distribution of Ly$\alpha$ forest lines in the QSO absorption spectra. Statistical tests performed on this data indicate that there may be large scale structure even though the power spectrum of the Ly$\alpha$ line distribution on large scales is found to be flat. It is apparent that higher (than two) order statistics are crucial in quantifying the clustering behavior of Ly$\alpha$ clouds. Using the discrete wavelet on three independent data sets of Ly$\alpha$ forests, we find that the distribution of Ly$\alpha$ forests does show non-Gaussian behavior on scales from 5 to 10 h$^{-1}$ Mpc with confidence level larger than 95%. Two data sets available on large scales are found to be non-Gaussian on even larger scales. These techniques are effective in discriminating among models of the Ly$\alpha$ forest formation, which are degenerate at second and lower order statistics (abridged).

It is of fundamental importance to determine if and how hierarchical clustering is involved in large-scale structure formation of the universe. Hierarchical evolution is characterized by rules which specify how dark matter halos are formed by the merging of halos at smaller scales. We show that scale-scale correlations of the matter density field are direct and sensitive measures to quantify this merging tree. Such correlations are most conveniently determined from discrete wavelet transforms. Analyzing two samples of Ly-alpha forests of QSO's absorption spectra, we find significant scale-scale correlations whose dependence is typical for a branching process. Therefore, models which predict a "history" independent evolution are ruled out and the halos hosting the Ly-alpha clouds must have gone through a "history" dependent merging process during their formation.

We present a detailed review of large-scale structure (LSS) study using the discrete wavelet transform (DWT). After describing how one constructs a wavelet decomposition we show how this bases can be used as a complete statistical discription of LSS. Among the topics studied are the the DWT estimation of the probability distribution function; the reconstruction of the power spectrum; the regularization of complex geometry in observational samples; cluster identification; extraction and identification of coherent structures; scale-decomposition of non-Gaussianity, such as spectra of skewnes and kurtosis and scale-scale correlations. These methods are applied to both observational and simulated samples of the QSO Lyman-alpha forests. It is clearly demonstrated that the statistical measures developed using the DWT are needed to distinguish between competing models of structure formation. The DWT also reveals physical features in these distributions not detected before. We conclude with a look towards the future of the use of the DWT in LSS.

Using a discrete wavelet based space-scale decomposition (SSD), the spectrum of the skewness and kurtosis is developed to describe the non-Gaussian signatures in cosmologically interesting samples. Because the basis of the discrete wavelet is compactly supported, the one-point distribution of the father function coefficients (FFCs) taken from one realization is a good estimate of the probability distribution function of the density if the ``fair sample hypothesis" holds. These FFC one-point distributions can also avoid the constraints of the central limit theorem on the detection of non-Gaussianity. Thus the FFC one-point distributions are effective in detecting non-Gaussian behavior in samples such as non-Gaussian clumps embedded in a Gaussain background, regardless of the number or density of the clumps. We demonstrate that the non-Gaussianity can reveal not only the magnitudes but also the scales of non-Gaussianity. Also calculated are the FFC one-point distributions, skewness and kurtosis spectra for real data and linearly simulated samples of QSO Ly$\alpha$ forests. When considering only second and lower order of statistics, such as the number density and two-point correlation functions, the simulated data show the same features as the real data. However, the the kurtosis spectra of samples given by different models are found to be different. On the other hand, the spectra of skewness and kurtosis for independent observational data sets are found to be the same. Moreover, the real data are significantly different from the non-Gaussianity spectrum of various posssible random samples. Therefore the non-Gaussain spectrum is necessary and valuable for model discrimination.

A method for measuring the spectrum of a density field by a discrete wavelet space-scale decomposition (SSD) has been studied. We show how the power spectrum can effectively be described by the father function coefficients (FFC) of the wavelet SSD. We demonstrate that the features of the spectrum, such as the magnitude, the index of a power law, and the typical scales, can be determined with high precision by the FFC reconstructed spectrum. This method does not require the mean density, which normally is poorly determined. The problem of the complex geometry of observed samples can also be easily solved because the basis are always orthogonal, regardless the geometry of the samples. Using this method, we examine the spectra inferred from Ly$\alpha$ forests of both simulated and real samples. We find that 1.) the magnitude of the 1-D spectra is significantly dfferent froma Poisson process; 2) the 1-d spectra are flat on scales less than 5 h$^{-1}$ Mpc, and increase slowly at larger ranges; 3.) the reconstructed 3-D spectra have about the same power as the COBE normalized linear spectrum of the SCDM model on scales less than 40 h$^{-1}$ Mpc, but is larger than the SCDM model on scales larger than 40 h$^{-1}$ Mpc; 4) the magnitudes of high redshift ($z>2.51$) spectra generally are larger than those of low redshift ($z<2.51$).

A wavelet space-scale decomposition (SSD) analysis of large scale structures in the universe has been developed. The SSD method of identifying and measuring structures in the spatial distribution of objects has been demonstrated. The position and strength (richness) of the identified clusters can be described by the corresponding coefficient of the wavelet transform. Using this technique, we systematically detected the clustering and its evolution of QSO's Ly$\alpha$ forest lines in real data and simulated samples. We showed that the clusters of Ly$\alpha$ absorbers do exist on scales as large as at least 20 h$^{-1}$ Mpc at significance levels of 2-4 $\sigma$. Independent data sets show about the same strength distribution of the decomposed clusters. The number densities of the clusters on scales of 10 - 20 h$^{-1}$ Mpc are found to evolve in an opposite sense as that of the lines themselves, i.e. they decrease with redshift. We also showed that the number density and the strength distribution of clusters can play an important role in testing or discriminating models, i.e. it can distinguish real data and simulated samples, which cannot be discriminated by traditional ways. We used Daubechies 4 and Mallat wavelets as the bases of the SSD. All above-mentioned conclusions do not depend on either wavelet basis.