MaNGA Target Selection
The basic goal of the MaNGA target selection is to select a sample of galaxies:
- with a reproducible method based on only absolute magnitude (and color for a subsample) and redshift
- with a flat number density distribution in absolute i-band magnitude
- that can be covered by the MaNGA IFUs beyond a physical radius given in units of the effective radius (Re; the radius containing 50% of the light of the galaxy)
- maximizing the spatial resolution and S/N while satisfying the above requirements
With the need for known redshift and photometry, we restrict the selection to the SDSS DR7 spectroscopic footprint. We aim to observe approximately 10,000 galaxies during the 6 years of SDSS-IV.
We define absolute-magnitude-dependent redshift limits to select such a sample. The lower redshift limit is needed to ensure the angular sizes of galaxies in the sky can fit within our largest IFUs. The limit gets higher for more luminous galaxies because more luminous galaxies are larger in physical sizes. The upper redshift limit is then set accordingly so that we can have roughly equal number density of galaxies in the sky for different absolute magnitudes. The redshift bands get wider towards more luminous galaxies because luminous galaxies are rarer per unit volume. Note the y-axis in the figure below is logarithmic.
As described in Wake et al. 2017 (AJ, 154, 86), we undertook an extensive optimization exercise to simultaneously determine the selection criteria and the IFU size distribution given the details of the MaNGA instrumentation. Described below are the results of that optimization applied to the selection of the three main MaNGA samples. All the main target galaxies are selected using the reprocessed NSA photometry and are restricted to lie within the NYU-VAG LSS footprint as described here. Throughout we use photometric quantities defined using elliptical Petrosian photometry.
The main MaNGA sample consists of three components: Primary sample, Secondary sample, and the Color-Enhanced supplement.
The Primary sample is selected so that 80% of the galaxies can be covered by the 127 fiber IFU to a major axis radius of 1.5 Re and makes up 47% of the main MaNGA samples. The selection criteria are defined as upper and lower redshift cuts that depend on absolute i-band magnitude only and are described by the following function:
zlim = (A + B(Mi-5 log10 h) ) (1+exp[C(Mi-5 log10 h - D)])
with A = -0.056597, B = -0.0039264, C = -2.9119, and D = -22.8476 for the lower redshift limit and A = -0.011377, B = -0.0019220, C = -1.2924, and D = -22.1610 for the upper limit. Here h = H0/100 km s-1 Mpc-1. All absolute magnitudes and stellar mass in the NSA catalog are computed assuming h=1.
To remove the tail of low luminosity blue galaxies and produce a pseudo-stellar mass limit we also apply a magnitude dependent color cut of g-r > 0.4 (Mr-5 log10 h) + 7.4. Here g-r is the color in the rest frame.
We also apply a completeness cut corresponding to the magnitude limit of the input catalog:
z < -0.9335 -0.1839 (Mi-5 log10 h) -0.01222 (Mi-5 log10 h)2 -2.7668 × 10-4(Mi-5 log10 h)3
This completeness cut is also applied to the Secondary sample and the Color-enhanced supplement described below.
The Secondary sample is selected so that 80% of the galaxies can be covered by the 127 fiber IFU to a major axis radius of 2.5 Re and makes up 37% of the main MaNGA sample. As for the Primary sample the selection criteria are defined with upper and lower redshift cuts that depend on absolute i-band magnitude only and are described by the same functional form. For the Secondary sample the parameters are A = -0.056463, B = -0.0048895, C = -1.3773, and D = -22.3851 for the lower redshift limit and A = -0.048010, B = -0.0046639, C = -1.3719, and D = -22.3225 for the upper redshift limit.
As for the Primary, the same a magnitude dependent color cut is applied with g-r > 0.28 (Mr-5 log10 h) + 5.6.
The Secondary sample was designed to have a higher density of targets than is required to give the desired 2:1 ratio of Primary+ to Secondary galaxies. Therefore, before allocating IFUs (see below) we random sample the Secondary sample to the desired target density. Due to a recently discovered bug in the target selection code this random sampling is not truly random and in fact samples in such a way as to make the number density distribution flat as a function of stellar mass. This is a small change, since the density distribution was already quite close to being flat with stellar mass, but does have some consequences when calculating the appropriate weights to apply to the Secondary sample (see the weighting FAQ and Wake et al. 2017 (AJ, 154, 86) for more details). Secondary targets passing the random sampling have RANFLAG = 1.
Secondary targets that are not included after random sampling (RANFLAG = 0) can be allocated IFUs but only as filler targets if no other targets are available. As a result these filler Secondaries are likely to be a biased sampling of the full Secondary sample, for instance they are likely to be preferential selected for lower density regions.
Color-Enhanced & Primary+ Samples
The Color-Enhanced supplement fills in poorly sampled regions of the NUV-i vs Mi color-magnitude plane such as low luminosity red galaxies, high luminosity blue galaxies and "green valley" galaxies. It is constructed by expanding the Primary sample upper and lower redshift limits in underpopulated NUV-i vs (Mi-5 log10 h) bins. As for the Primary sample the goal is to cover the galaxies to 1.5 Re.
Combining the Primary sample and the Color-Enhanced supplement makes the Primary+ sample, which completely spans the full color-magnitude range targeted by MaNGA. As such the Primary+ (or the Primary or Secondary) sample can be used for population studies with appropriate weights applied (see Wake et al. 2017 (AJ, 154, 86)), but the Color-Enhanced supplement shall not be used by itself.
The Color-Enhanced supplmement makes up 16% of the main MaNGA sample. The Primary+ sample makes up 63% of the main MaNGA sample.
Tiling and IFU Allocation
One of our requirements is that the selection will be unbiased with respect to environment. This means that we should not choose to ignore a region where we have fewer galaxies than IFUs. It also means that we need to overlap tiles in denser regions on the sky to achieve similar completeness in dense environment as in voids. To achieve this we use an adaptive tiling routine that uses a 'gravitational' method to assign multiple overlapping plates to the densest regions. The method starts with an evenly distributed overlapping tiling. The target galaxies then 'pull' tiles as an $1/r^2$ law, with the mass of already assigned targets set to zero. The velocity with which the plates are allowed to move is damped to prevent oscillation of the tiles and a mild repulsive force between the plates is included. This has the effect of pulling plates towards the over dense regions and away from the voids. Currently only tiles with $>$ 7 targets are included in the final tiling.
Our IFU allocation method is designed to maximize the allocation of IFUs to galaxies of the appropriate size. It proceeds as follows:
- All galaxies within a given tile are selected.
- For each IFU size, say 19 fiber, galaxies that require a 19-fiber IFU to reach their target radius (e.g. 1.5 or 2.5 Re) are selected. Galaxies with a target radius smaller than the 19-fiber IFU are included in the 19-fiber IFU allocation (galaxies larger than the 127-fiber IFU are likewise assigned to the 127-fiber IFU).
- All available 19-fiber IFUs are assigned to these galaxies.
- If there are fewer galaxies that need 19-fiber IFUs than available 19-fiber IFUs, the remaining IFUs are put into a pool to be assigned later.
- This process is repeated for the remaining IFU sizes.
- The unassigned IFUs are then allocated to the remaining galaxies closest in target size, beginning with the largest galaxies and the largest remaining IFUs, and then working downwards.
This method produces a sample that has an angular size distribution that matches as closely as possible the IFU size distribution. Since we have optimized the IFU size distribution to match the actual galaxy target size distribution there is only a very small size bias introduced, which can be easily corrected for with weights if desired (see Wake et al. 2017 (AJ, 154, 86)).