A multitude of evidence has accumulated in support of the existence of dark matter in our Universe. There are already plenty of dark matter candidates. However, we do not know yet whether any of these candidates constitutes the whole or a part of the dark matter population despite the tremendous experimental efforts. In this thesis, we study several possible dark matter production mechanisms and the corresponding observational and theoretical constraints in the context of inflationary cosmology. Adopting a model-independent approach, we explore the parameter space for dark matter with a mass of order MeV and above showing that only small regions of the parameter space for the popular freeze-out mechanism are still viable. Nevertheless, the regions of the parameter space corresponding to the freeze-in and non-thermal dark matter scenarios are mostly unexplored. We, therefore, zoom into these regions and show that a connection to the inflationary observables can be established, which can help constrain these scenarios. We then consider the parameter space of a sub-eV dark matter candidate, the axion. We show that using the Cosmic Microwave Background radiation constraint on the effective number of relativistic species, an interesting constraint can be placed. This bound arises from the fact that the field whose angular excitations are the axions can be displaced from its minimum during inflation and later decays dominantly into ultra-relativistic, axions which contribute to the effective number of relativistic species. We finally consider the possible production of axion-like particle via non-perturbative effects due to their coupling to inflatons or moduli. We show that this mechanism is efficient only if the amplitude of inflaton/moduli oscillations is initially much larger than the mass scale associated with the axion-like particles. In this case, bounds can be placed on the corresponding parameter spaces.

We make use of cutting-edge simulations of galaxy formation in a Λ cold dark matter (ΛCDM) Universe to investigate the impact of the uncertainties inherent to certain observational techniques for estimating the masses of galaxies on the conclusions that are drawn from studies using such methods. By performing virtual 'observations' of simulated galaxies, we estimate their stellar and dynamical masses in the same way as in particular observational studies. The satellite galaxies of the Milky Way are highly attractive candidates for dynamical studies, due to their proximity; and in general, satellite galaxies dominate the clustering of galaxies on small scales. The total dynamical masses and internal mass distributions of individual galaxies, along with the clustering of galaxies as a function of intrinsic properties such as stellar mass, each reflect the structure and evolution history of the underlying invisible dark matter that forms the structural spine of the Universe and incubates the formation and evolution of galaxies over cosmic time. The observed stellar kinematics of dispersion-supported galaxies are often used to measure dynamical masses. Recently, several analytical relationships between the stellar line-of-sight velocity dispersion, the projected (2D) or deprojected (3D) half-light radius, and the total mass enclosed within the half-light radius, relying on the spherical Jeans equation, have been proposed. Here, we make use of the APOSTLE cosmological hydrodynamical simulations of the Local Group to test the validity and accuracy of such mass estimators for both dispersion and rotation-supported galaxies, for field and satellite galaxies, and for galaxies of varying masses, shapes, and velocity dispersion anisotropies. We find that the mass estimators of Walker et al. and Wolf et al. are able to recover the masses of dispersion-dominated systems with little systematic bias, but with a $1\sigma$ scatter of 25 and 23 percent, respectively. The error on the estimated mass is dominated by the impact of the 3D shape of the stellar mass distribution, which is difficult to constrain observationally. This intrinsic scatter becomes the dominant source of uncertainty in the masses estimated for galaxies like the dwarf spheroidal (dSph) satellites of the Milky Way, where the observational errors in their sizes and velocity dispersions are small. Such scatter may also affect the inner density profile slopes of dSphs as derived from multiple stellar populations, relaxing the significance with which Navarro-Frenk-White profiles may be excluded, depending on the degree to which the relevant properties of the different stellar populations are correlated. Additionally, we derive a new optimal mass estimator that removes the residual biases and achieves a statistically significant reduction in the scatter to 20 percent overall for dispersion-dominated galaxies, allowing more precise and accurate mass estimates. We present predictions for the two-point correlation function of galaxy clustering as a function of stellar mass, computed using two new versions of the GALFORM semi-analytic galaxy formation model. One model uses a universal stellar initial mass function (IMF), while the other assumes different IMFs for quiescent star formation and bursts. Particular consideration is given to how the assumptions required to estimate the stellar masses of observed galaxies (such as the choice of IMF, stellar population synthesis model, and dust extinction) influence the perceived dependence of galaxy clustering on stellar mass. Broad-band spectral energy distribution fitting is carried out to estimate stellar masses for the model galaxies in the same manner as in observational studies. We show clear differences between the clustering signals computed using the true and estimated model stellar masses. As such, we highlight the importance of applying our methodology to compare theoretical models to observations. We introduce an alternative scheme for the calculation of the merger time-scales for satellite galaxies in GALFORM, which takes into account the dark matter subhalo information from the underlying dark matter only simulation. This reduces the amplitude of small-scale clustering. The new merger scheme offers improved or similar agreement with observational clustering measurements, over the redshift range 0 < z < 0.7. We find reasonable agreement with clustering measurements from GAMA, but find larger discrepancies for some stellar mass ranges and separation scales with respect to measurements from SDSS and VIPERS, depending on the GALFORM model used.

Semi-analytical (SA) galaxy formation models have wide applications, and they are complementary to hydrodynamical simulations, which are more physically detailed but also much more computationally expensive. It is important to make semi-analytical models as physical as possible for the robustness of their applications. In this work we try to improve the modelling of two important processes, supernova (SN) feedback and gas cooling, in the SA model galform. We first improve the SN feedback recipe in a phenomenological way, using the constraints from four observations, including the Milky Way (MW) satellite galaxy luminosity function, the faint end of the field galaxy luminosity function, the redshift at which the universe was half reionized and the stellar metallicity of the MW satellites. We find that these observations favour a SN feedback model in which the feedback strength evolves with redshift. We further apply this improved model to investigate some details of reionization. We then develop a new, more physical model for gas cooling in halos in semi-analytical models. We compare this new cooling model with a cosmological hydrodynamical simulation with stripped-down galaxy formation physics running with the grid-based moving mesh code arepo, along with two previous models (GFC1 and GFC2) in galform and the models in l-galaxies and morgana. We find that generally all SA models predict cumulative cool masses close to the simulation, but the mass cooling rates in low redshift massive halos are overestimated. These SA models overpredict the specific angular momenta of the cool gas for low mass halos, while for low redshift massive halos, the predictions from the new cooling model generally agree better with the simulation than the earlier SA cooling models. We also use the simulation to investigate gas cooling in individual halos in more detail.

Primordial fluctuations are expected to be produced in the very early Universe, sourcing the anisotropies in the cosmic microwave background and seeding the formation of structures. In this thesis we study the effect of density perturbations produced during inflation on the large-scale galaxy bispectrum. We start by reviewing the basic concepts of modern cosmology and introducing the tools used in this research: Newtonian perturbation theory, statistics of random fields, the mass function of collapsed halos and the halo bias model. We then briefly describe how models of inflation source local-type non-Gaussian distributed primordial density perturbations. We apply these tools to justify the bivariate model for the halo density in the presence of primordial non-Gaussianity and derive some known results, like the scale-dependent halo bias. The aim is to show that the statistics of large-scale structure can be used to probe local-type non-Gaussianity of the primordial density field, complementary to existing constraints from the cosmic microwave background. Parametrising the amount of primordial non-Gaussianity with the leading-order non-linear parameter f_NL and the next-order one, g_NL, we will investigate how galaxy and matter bispectra can distinguish between them, despite their effects being nearly degenerate in the power spectra. We determine a connection between the sign of the halo bispectrum on large scales and the parameter g_NL and construct a combination of halo and matter bispectra that is sensitive to f_NL. After that, we will focus on local-type non-Gaussianity with f_NL only. It is known that the non-linear evolution of the matter density introduces a non-local tidal term in the halo bias model. Furthermore, we will show that the bivariate model in the Lagrangian frame leads to a novel non-local convective term in the Eulerian frame which can lead to non-negligible corrections in the halo bispectra, in particular on large scales or at high redshift. Finally, we address the problem of modelling redshift space distortions in the galaxy bispectrum, finding novel contributions with the characteristic large scale amplification induced by local-type non-Gaussianity. Therefore, redshift space distortions can potentially lead to a biased measurement of f_NL, if not properly accounted for. Moreover, we propose an analytic template for the monopole which can be used to t against data on large scales, extending models used in recent measurements. We conclude the thesis with some discussion of future developments. Observational constraints will also be discussed, based on idealised forecasts onf_NL { the accuracy of the determination of f_NL. Our findings suggest that the constraining power of the galaxy bispectrum in current surveys would provide f_NL measurements competitive with constraints from the cosmic microwave background and future surveys could improve this further.

Supernovae are important probes of cosmology. In 1999, Type Ia Supernovae (SNeIa) provided the first evidence for the accelerating expansion of the Universe (Riesset al., 1998, Perlmutter et al., 1999), and since then there have been many wide-field SN surveys with the scope of increasing the number of observed SNe, thus improving the constraints on cosmological parameters. Among these SN surveys,the Dark Energy Survey (DES) and the planned Large Synoptic Survey Telescope(LSST) will increase the number of available SNe Ia respectively to ' 3000 and ∼10⁵ (possibly ∼ 10⁶) in the coming decade. Weak gravitational lensing effects willthen become important for these new surveys. Weak gravitational lensing have different effects on the distance modulus measurements of SNe. Firstly, it introduces a non-Gaussian scatter on the distance moduli of SNe Ia, and this effect increases as a function of redshift. The non-Gaussian weak lensing distribution can also introduce a bias on the cosmological parameter values recovered by fitting the Hubble diagram. Secondly, it introduces spatial correlations on the magnitudes of close SN pairs, with angular separation of the order of arcminutes. Weak lensing of SNe can also be used to probe the growth of structures along the line-of-sight, giving further constraints on cosmological parameters like σ₈ and Ω_m. In Chapter 2, we present our results on the fit of the Hubble diagram from the Jointed Light-curve Analysis sample (JLA, Betoule et al. 2014) including weak lensing and peculiar velocities, the latter introducing an extra dispersion on the distance modulus measurements of low redshift SNe. We give constraints on the cosmological parameters when fitting for the the first four moments of the weak lensing distribution together with the variance induced by peculiar velocities. We test our method via numerical simulations and we find Ω_m=0.274±0.013 and σ₈=0.44^+0.63_-0.44 when fitting the JLA sample. We also apply the Kernel Density Estimation technique to reduce the problem of biased estimates of the moments measured on sparse data sample, and a boot-strap re-sampling method when computing the covariance between the moments. In Chapter 3 we propose to measure the two-point magnitude correlation function from SN data and compare such measurements to theoretical expectations. As available data sample appear to be insufficient to detect this weak correlation (we report a tentative detection with the JLA sample), we predict measurements with current (DES) and future (LSST) SN surveys, finding that the LSST should be able to detect such correlations at 6σ level of confidence (15,000 SNe over 70 deg² and assuming an intrinsic scatter of 0.15 magnitudes). DES (deep field) is expected to detect a cross-correlation between the Hubble residuals and the foreground galaxies at 12σ (integrated up to 9 arcminutes of separation and assuming an intrinsic scatter of 0.15 magnitudes), taking advantage of the higher galaxy density on the sky, while LSST should detect the same cross-correlation with signal-to-noise ⪆ 100. We also give forecasts on cosmological parameters when fitting Ωm and σ₈ from the twopoint magnitude auto-correlation function, i.e. we can achieve a 25% measurement of σ₈ from LSST (assuming 0.15 magnitudes of intrinsic scatter and applying a Gaussian prior on the matter density parameter). In Chapter 4, we investigate Type Ic Superluminous Supernovae (SLSNe Ic) as a new class of potential standard candles, which appear to be standardisable in their peak magnitudes with a scatter of only 0.2 – 0.3 magnitudes. Moreover, their exceptional peak magnitude (up to 100 times brighter than SNe Ia) allows them to be discovered to redshift ∼ 3, shedding new light on the deceleration epoch of the Universe. We give predictions for SLSN Ic redshift distribution within present (DES and SUDSS, which are expected to find 15 and 75 SLSNe respectively) and future surveys (LSST and Euclid, which should increase the available SLSNe to   10; 000 and 300 respectively, the latter up to redshift 4). We construct simulated Hubble diagrams for SLSNe Ic, spanning the likely values of intrinsic scatter for these sources ( 0:15 - 0:25 magnitudes), and fit the Hubble diagrams to infer cosmological constraints. We find that the addition of 75 SLSNe from SUDSS to the 3800 SNe Ia from DES can improve the constraints on w (the dark energy state parameter) and Ω_m by 20% (assuming a flat wCDM universe). Moreover, the combination of DES SNe Ia and 10,000 LSST SLSNe can measure Ω_m and w to 2% and 4% respectively. When considering temporal variations in w(a), we find possible uncertainties of 2%, 5% and 0.14 on Ω_m, w₀ and w_a respectively, from the combination of DES SNe Ia, LSST SLSNe and Planck Cosmic Microwave Background temperature power spectrum. We find that SLSNe from Euclid can constrain the matter density parameter to 10%, and can help constraining the equation-of-state parameters w₀ and w_a. All these surveys will also improve the knowledge about SLSN astrophysics, their progenitors and possible classification into sub-classes.

Galaxy SEDs contain a wealth of information about their stellar populations with the UV region tracing their hot component. In young populations this hot component comes from luminous O and B-type stars whereas, in old populations a hot component can be produced after sufficient mass loss, a phenomenon known as the UV upturn. The UV region of galaxies remains relatively unexplored and models lack calibration due to the paucity of observational data. However, by investigating features seen in the UV region it may be possible to explore both types of population with the potential to find indicators that can differentiate between the two. I use a large sample (∼ 275,000) of z ≥ 0.6 massive (log(M_∗/M_⊙) >∼ 11) galaxies taken from the Sloan Digital Sky Survey (SDSS) - III Baryon Oscillation Spectroscopic Survey (BOSS). I use both individual spectra and stacks and employ a suite of models including a UV contribution from old populations, spanning various effective temperatures, fuel consumptions, and metallicities. By investigating the effect of the UV upturn on the strength of mid-UV indices I find a subset that are able to differentiate between old and young UV ages; Mg I, Fe I, and BL3096. I find evidence for old stars contributing to the UV in massive galaxies, rather than star formation. The data favour models with low/medium upturn temperatures (10,000 - 25,000K) consistent with local galaxies, depending on the assumed metallicity, and with a larger fuel ( f ∼ 6.5 · 10⁻²M_⊙). Models with only one temperature are favoured over models with a temperature range, which would be typical of an extended horizontal branch. Old UV-bright populations are found in the whole working sample (92%), with a mass fraction peaking around 10 - 20%. Upturn galaxies are massive and have redder colours, in agreement with findings in the local Universe. I find that the upturn phenomenon appears at z ∼ 1 and its frequency increases towards lower redshift, as expected by the stellar evolution of low mass stars. These findings will help to constrain stellar evolution in the little explored UV upturn phase. The highest redshift galaxies are young and hence exhibit a pronounced UV spectrum due to their massive star components. The UV rest-frame is also what is actually sampled via optical and near-IR observations at these high redshifts. However, a comprehensive study of UV absorption lines, which may provide useful indicators for physical properties such as stellar age and metallicity, is still lacking. I exploit stellar population models of absorption line indices in the far-UV (1200 - 1900Å) to study the spectra of young high-z galaxies. Using high-z spectra from Sommariva et al. (2012), Erb et al. (2010), and VVDS, the central aim of this analysis is to assess the ability of the model indices to recover the stellar ages and metallicities found in the literature. Using a set of far-UV indices I fit both SSPs and CSPs to the strength of the absorption features found in the data as well as fitting the full far-UV spectral region. A simple test using mock galaxies shows the effect of dust to be negligible when fitting the indices in combination however, there may be more complicated effects that are not modelled well. The analysis shows that currently it is not possible to reliably derive the stellar ages or metallicities of high-z galaxies using the methods explored in this work. The large range of χ² values for the model fitting is likely due to the errors of the spectral indices being underestimated and the low quality of the UV data. However, issues may also lie on the model side. Emission lines are known to effect the far-UV region and could contaminate the absorption features investigated in this work. The spectra analysed have not been "cleaned" of emission lines and such features are not included in theoretical modelling.

Observations of the local Universe indicate that we are currently in an accelerated phase of cosmic expansion. This behaviour is compatible with our best theory of gravity, General Relativity, with the addition of a non-luminous form of energy that exerts a negative pressure known as ‘dark energy’. Alternatively, the observed acceleration could be an indication of the presence of new degrees of freedom active on cosmological scales. These degrees of freedom could be in the form of a dynamical field called ‘quintessence’, which is typically isolated from the rest of energy-momentum; although models which allow for interactions with ‘dark matter’ have recently been explored. This thesis explores a more radical idea; that the acceleration is a result of additional gravitational degrees of freedom acting on the largest scales. Any modification to our description of gravity on cosmological scales must take into account that we have very accurate tests of General Relativity within the solar system. Therefore the theories we explore rely on non-linearities that allow them to evade solar system tests via a ‘screening mechanism’. In particular, we focus on the use of special, non-linear, derivative structures which dominate on scales shorter than a characteristic length; suppressing the coupling to energy-momentum and thus screening the effect of the additional force. On the other hand, these interactions are negligible over cosmic scales and we recover a linear theory that communicates a fifth force. A typical theory of this type and the first to be discovered is the ‘Galileon’. We introduce this theory and discuss its behaviour around a static a spherically symmetric source which provides an example of ‘Vainshtein screening’. We introduce Ostrogradsky’s construction which proves that non-degenerate theories with higher order time derivatives always have instabilities. However, by being degenerate, the special structure of Galileons ensures that they evade this result. The same structure is used to construct vector theories with non-linear derivative self-interactions. These theories, named ‘vector Galileons’, break gauge symmetries and have been shown to have interesting cosmological applications. We introduce a way to spontaneously break the gauge symmetry and construct these theories via a Higgs mechanism. In addition to the purely gauge field interactions, our method generates new ghost-free scalar-vector interactions between the Higgs field and the gauge boson. We show how these additional terms are found to reduce, in a suitable decoupling limit, to scalar bi-Galileon interactions between the Higgs field and Goldstone bosons. Our formalism is first developed in the context of abelian symmetry, which allows us to connect with earlier work on the extension of the Proca action. We then show how this formalism is straightforwardly generalised to generate theories with non-abelian symmetry. Using an Arnowitt-Deser-Misner approach, we carefully reconsider the coupling with gravity of vector Galileons, with the aim of studying the necessary conditions to avoid the propagation of ghosts. We develop arguments that put on a more solid footing the results previously obtained in the literature. Moreover, working in analogy with the scalar counterpart, we find indications for the existence of a ‘beyond Horndeski’ theory involving vector degrees of freedom. After identifying the decoupled longitudinal mode of the vector Galileon with the scalar Galileon, we investigate the number of degrees of freedom present in the theory. We discuss how to construct the theory from the extrinsic curvature of the constant scalar field hypersurface, and find a simple expression for the action which guarantees the existence of the primary constraint necessary to avoid the Ostrogradsky instability. We then return to the ‘Galileonic Higgs mechanism’ and consider the effect of interactions between the higher order operators and a dynamical metric. We find a consistent covariantisation through the use of gravitational counter-terms that serve to also restrict the parameter space of the theory. After a brief introduction to cosmological perturbation theory, we explore the cosmological applications of the Galileonic Higgs. We find self-accelerating background solutions, associated with a non-trivial profile of the vector. We then expand the action to quadratic order in linear perturbations, diagonalise and discover that one of the modes is a ghost. This is in contrast with the positive results of related scenarios where an instability on Minkowski space is removed by gravitational interactions.

In this thesis I study how dark matter particles that interact through forces other than just gravity would affect the formation of structure in the Universe. This begins with a theoretical calculation of the location and rate at which these interactions take place throughout cosmic history. Giant galaxy clusters are expected to have the highest rates of dark matter interactions, at least for the simplest dark matter particle models. Predicting the formation of structure with non-standard dark matter requires the use of N-body simulations. I therefore introduce and test a set of modifications to the GADGET code that allow it to simulate a class of dark matter models known as self-interacting dark matter (SIDM). I focus particular attention on rarely discussed aspects of simulating SIDM; including how to handle particles scattering multiple times within a single time-step and how to implement scattering across processors. I also discuss how best to choose numerical parameters associated with the SIDM implementation and the range of numerical parameters that produce converged results. Because galaxy clusters should have particularly high rates of dark matter interactions, I use this code to perform simulations of a pair of merging galaxy clusters known as the 'Bullet Cluster'. At first these employ simple SIDM particle physics models for the dark matter. I demonstrate the importance of analysing simulations in an observationally motivated manner, finding that the way in which simulation outputs are compared with observations can have a significant impact on the derived constraints upon dark matter’s properties. I then look at what happens to these constraints for more complicated particle physics models of SIDM. In isolated systems, the effects of a complicated scattering cross-sections can be modelled using an appropriately-matched simple cross-section, while in systems like the Bullet Cluster, complicated cross-sections lead to phenomenology not seen with simpler particle models. Overall I find that SIDM remains a viable class of dark matter models, consistent with current observations.

Our understanding of the Universe is based on the ΛCDM model which, although the best cosmological model so far, relies on the presence of major unknown components – dark matter, dark energy, and an inflationary field – which in turn play a crucial role in the evolution of the Universe. These limitations of the model suggest that we may need to introduce modifications at cosmological scales. Indeed, a large variety of modified gravity theories have been proposed (see [1] for a review) and in order to better understand the behaviour of gravity in this regime, we must begin by constructing theoretical and observational tests of the ΛCDM model and the various alternative proposals. This thesis is concerned with testing gravity on cosmological scales, by analysing the viability of alternative gravitational theories, and scrutinising their theoretical consistency. In order to do this, we take two approaches. On the one hand, we explore the viability of a specific modified gravity theory, namely massive bigravity. The evolution of a perfectly homogeneous and isotropic Universe has been previously studied in detail in this model, and has been found to fit observational data. Hence, in this thesis we analyse the evolution of linear cosmological perturbations, where we find a number of interesting instabilities. On the other hand, we take a broader view and develop a method for parametrising linear cosmological perturbations that stays agnostic about the underlying theory of gravity. We apply this method to three classes of models: scalar-tensor, vector-tensor and bimetric theories, and as a result, in this case, we identify the complete forms of the quadratic actions for linear perturbations, and the number of free parameters that need to be defined, to cosmologically characterise these broad classes of gravity theories.

Many studies have shown that the majority of accretion onto supermassive black holes; i.e., active galactic nuclei (AGNs) are hidden from our view by obscuring "torus" of gas and dust with column densities of N_H ≥ 10²² cm⁻². Arguably, the most efficient method of identifying AGNs is in the X-ray waveband, where even heavily obscured AGNs have been detected. However, a significant fraction of the AGN population have remained hidden in X-rays due to their extreme torus column densities along our line of sight, N_H ≳ 1.5 x 10²⁴ cm⁻²; i.e., Compton-thick (CT). These CTAGNs are predicted to be abundant but their census is far from complete, even in our local universe, due the challenge in identifying them because of their faint fluxes. In this thesis, I present an updated census of the CTAGN population and the N_H distribution of AGN in our local universe, using a volume-limited mid-infrared selected sample of 20 AGNs within D = 15 Mpc. The volume-limited selection within a relatively small volume means that the AGN sample is less limited by flux (unlike most AGN samples), and the mid-infrared selection means that it is unbiased against both torus and host galaxy obscuration (unlike X-ray and optically selected AGN samples, respectively). The N_H values for each AGN are directly measured, by performing broadband X-ray spectroscopy (up to ~2 orders of magnitude in energy range) using data from multiple focusing X-ray observatories, primarily NuSTAR in combination with Chandra and XMM-Newton. For cases in which this is not possible, I use indirect multiwavelength techniques to identify potential CTAGNs. These techniques involve comparing the observed 2-10 keV fluxes from the AGNs to the [OIII]λ5007Å, 12μm, and [NeV]λ14.32μm fluxes, which act as indicators for the AGN intrinsic emission. The CTAGN fraction that I found; i.e., 30⁺²³₋₁₄%, is significantly higher than that observed in the hard X-ray AGN survey by Swift-BAT, but agrees very well with that inferred from the Swift-BAT survey, after taking into account sensitivity effects. I demonstrate that we can identify intrinsically lower luminosity CTAGNs that are missed by the Swift-BAT survey (i.e., down to L₂₋₁₀,_int ~ 10⁴⁰ erg s⁻¹). I provide case studies on two newly identified bona-fide CTAGNs in our local universe (NGC 1448 and NGC 5643), and demonstrate the challenges in characterising the properties of low luminosity AGN, primarily due to significant dilution and contamination by the X-ray emission from the host galaxy. I compare the AGN and host galaxy properties of my sample with that of the Swift-BAT AGN sample, and found the following: (1) the star formation rates between the two samples are consistent with each other; however, (2) my sample has a wider range of AGN Eddington ratio, extending down to a lower Eddington ratio than the Swift-BAT AGN sample; (3) my sample also shows a more diverse optical nuclear spectral type; and (4) dominates at lower black hole mass and galaxy stellar mass.