Inconsistencies Between Halo Scaling Relations and Integral Scaling Relations

[daviesje]

The halo sampler introduced some new scaling relations, which change the way X-ray and stellar mass behave as a function of halo mass:

• The stellar-to-halo mass relation now includes an upper turnover at high M_halo
• L_X / SFR is now no longer constant, and scales with stellar mass and SFR via a metallicity

These new scaling relations were not applied to the previous model which integrates over the conditional mass function, which would be okay if we didn't use these integrals to determine a few things even when the halo sampler is used:

• The x-ray optical depth (TauX) still uses the global neutral fraction, obtained from integrals
• Halos below the minimum sampler mass are included based on these integrals, so these sources effectively have a cosntant L_x/SFR and no high mass stellar turnover (the latter will not matter but the former might)
• If FIXED_HALO_GRID=True all source fields are taken on the cell-scale using these integrals.

These factors lead to small inconsistencies within the model we would like to get rid of. Applying the high-mass stellar turnover in the integrals is simple enough, although it will slow down interpolation table construction due to 2 extra pow() calls in each integrand, which could matter for some minihalo runs (needs testing). On the other hand, using a non-constant L_X / SFR means that we can no longer combine the x-ray source grid with the lyman alpha and lyman werner source grids in SpinTemperature.c, this would at most double memory usage in runs without minihalos (including one more R X HII_TOT_NUM_PIXELS grid), and double the time spent on filtering the grids, so we should be careful here (about a third of the spintemp calculation and 3% of the whole run).