# Comments for the presentation


## Introduction

---

---


---


// COMMENTS:


---

// Explanation
- further modulation by _RSD_


    // From first principles
    // small scales to resolve sources + sinks + feedback
    // large scales to capture statistics

        // IF ASKED: difference with `21cmFAST`:
        // based on excursion formalistm -> only valid >= 1Mpc, which is ideal for large volumes + statistics => 21-cm forecasts

        // interesting to build emulators for instance


---


// From the xray emission
// primordial + heating term
// expansion + deposition by xrays
// => xrays are assumed to be the only source of heating


// $
//   x_("HII")(r bar M, z) = theta_"H" lr([R_b (M, z) - r], size: #150%)
// $

---

// COMMENTS:
// - contribution from the lyman lines
// - 1/r^2 decrease from spreading photons
// - more steep outwards + sharp drop due to redshifting out of line


---

== Revisiting the 21cm signal


---


### Procedure
Painting using all halos that match in a SINGLE step



---

OVERLAP EXPLICITLY ALLOWED


---
### Postprocessing
- ionization overlaps
- corrections due to RSD
- computation of derived quantities
- summary statistics

---
### Maps

---

### Signal

---

## Halo growth

### Motivation



### Effect on the flux profiles

// COMMENTS
// That will be directly affect the global signal as well
// shifting
//
// Yu-Siu already investigated the more nuanced effect of stochasticity but the approach we propose should supersede that


### Inferring growth from #smallcaps[Thesan] data
// ideal for rapid iterations

// in a parallelized fashion => want to stay fast
// fix the original mass for max. consistency
// fix the allowed dynamic range

// this sort of "breaks the degeneracy" between halos of the same mass but different growth histories


---

RESULT OF LOADING:
// COMMENTS:
// no clear trend between mass and growth rate

---

## Adaptations
---

### Central changes


// important since the bins are more now


// largely through vectorization -> still "native" python
// usage of HDF5
// solid caching mechanisms -> resume simulations, etc...


### Simplified usage

In a page or less

---

## Results
### Map outputs
  // the ones where the accretion rate is likely higher

  // in particular: no values where the coupling has become weaker
  // will become apparent in the signal as well


---

  // those are the ones where the diff vanishes: e.g. top right

---

  // more variation due to the different accretion rates

---
// Globally:
// more dynamic range while the mean systematically shifts towards the (biased) lower accretion rates

// Intermezzo - compare with lower alpha range - mostly similar but occasional contributions from higher alpha values
// => recommend keeping a wide range since it does not affect performance (if the bins are empty anyway)
// the more intersting discussion to be had is the effect of a more fine binning - thesan data already gives an indication which values will be most frequent
// => the implementation to test that is there


  // the halos themselves produce a stronger singal while the background is usually




---

### Signals

---

## Conclusion

---
### Summary

  // since it affects the SFR and thus the emissivity


  // change in profiles trivially

  // which could in theory be absorbed by shifting other paremeters

  // which we can hope to observe (although many are subtle)

// unique position of 21-cm cosmology -> cannot discuss observational constraints


  // invite you to check out


### Outlook
// finally ready for direct comparison with c2ray? now that parameters and loading have been properly implemented

// Assuming other relations related to production of photons is (hopefully by now well motivated) complex
// these cannot directly be inferred => expressed as a distribution as a function of another halo property

// the scale-up -> large volumes with usable merger trees
// comitting to reserving some 100s of node hours (which I would still quantify as fast)