We present synthetic dust polarization maps of simulated molecular clouds (MCs) with the goal to systematically explore the origin of the relative orientation of the magnetic field ($\bf{B}$) with respect to the MC sub-structures identified in density ($n$; 3D) and column density ($N$; 2D). The polarization maps are generated with the radiative transfer code POLARIS, including self-consistently calculated efficiencies for radiative torque alignment. The MCs are formed in two sets of 3D MHD simulations: in (i) colliding flows (CF), and (ii) the SILCC-Zoom simulations. In 3D, for the CF simulations with an initial field strength below $\sim$5 $\mu$G, $\bf{B}$ is oriented parallel or randomly with respect to the $n$-structures. For CF runs with stronger initial fields and all SILCC-Zoom simulations, which have an initial field strength of 3 $\mu$G, a flip from parallel to perpendicular orientation occurs at high densities of $n_\text{trans}$ $\simeq$ 10$^2$ - 10$^3$ cm$^{-3}$. We suggest that this flip happens if the MC's mass-to-flux ratio, $\mu$, is close to or below the critical value of 1. This corresponds to a field strength around 3 - 5 $\mu$G. In 2D, we use the Projected Rayleigh Statistics (PRS) to study the orientation of $\bf{B}$. If present, the flip in orientation occurs at $N_\text{trans}$ $\simeq$ 10$^{21 - 21.5}$ cm$^{-2}$, similar to the observed transition value from sub- to supercritical magnetic fields in the ISM. However, projection effects can reduce the power of the PRS method: Depending on the MC or LOS, the projected maps of the SILCC-Zoom simulations do not always show the flip, although expected from the 3D morphology. Such projection effects can explain the variety of recently observed field configurations, in particular within a single MC. Finally, we do not find a correlation between the observed orientation of $\bf{B}$ and the $N$-PDF.

The formation of molecular hydrogen (H$_2$) and carbon monoxide (CO) is sensitive to the volume and column density distribution of the turbulent interstellar medium. In this paper, we study H$_2$ and CO formation in a large set of hydrodynamical simulations of periodic boxes with driven supersonic turbulence, as well as in colliding flows with the \textsc{Flash} code. The simulations include a non-equilibrium chemistry network, gas self-gravity, and diffuse radiative transfer. We investigate the spatial resolution required to obtain a converged H$_2$ and CO mass fraction and formation history. From the numerical tests we find that H$_2$ converges at a spatial resolution of $\lesssim0.2$~pc, while the required resolution for CO convergence is $\lesssim 0.04$~pc in gas with solar metallicity which is subject to a solar neighbourhood interstellar radiation field. We derive two critical conditions from our numerical results: the simulation has to at least resolve the densities at which (1) the molecule formation time in each cell in the computational domain is equal to the dissociation time, and (2) the formation time is equal to the the typical cell crossing time. For both H$_2$ and CO, the second criterion is more restrictive. The formulae we derive can be used to check whether molecule formation is converged in any given simulation.

Precision photometric redshifts will be essential for extracting cosmological parameters from the next generation of wide-area imaging surveys. In this paper we introduce a photometric redshift algorithm, ArborZ, based on the machine-learning technique of Boosted Decision Trees. We study the algorithm using galaxies from the Sloan Digital Sky Survey and from mock catalogs intended to simulate both the SDSS and the upcoming Dark Energy Survey. We show that it improves upon the performance of existing algorithms. Moreover, the method naturally leads to the reconstruction of a full probability density function (PDF) for the photometric redshift of each galaxy, not merely a single "best estimate" and error, and also provides a photo-z quality figure-of-merit for each galaxy that can be used to reject outliers. We show that the stacked PDFs yield a more accurate reconstruction of the redshift distribution N(z). We discuss limitations of the current algorithm and ideas for future work.