search

syrte / ffb_predict

Feedback-Free Starbursts (FFB) at Cosmic Dawn: Observable Predictions for JWST
MIT License
0 stars 0 forks source link

Clarifications regarding ffb predict #1

Open mShuntov opened 5 months ago

mShuntov commented 5 months ago

Hi @syrte,

Thanks so much for sharing your code to generate FFB predictions, it is so useful to compare to observations. However, there are some results that are a bit puzzling to me, so I wonder if I am wrongly using the code.

  1. FFB predictions on the SMF exceed the HMF. When I generate the FFB SMF using

compute_dNdlgMs(z=zmean, lgMs=None, return_func=True)

with different eps_FFB_max and plot them together with the halo mass function generated from

compute_dNdlgMh(zmean, lgMh=Mh_arr, return_func=False),

I find that the FFB SMF for high eps_FFB_max exceeds the HMF, which is counterintuitive. I show this in the following figure, where I rescale the HMF by different epsilon values. The HMF scaled by epsilon=1 should in principle not be exceeded by the FFB model, or am I missing something? image

  1. SFRD vs z predictions don't match the observations When I make predictions for the SFRD with FFB using

compute_rho_SFR(z=zmean, MUV_lim=-17.)

And I compare with a literature compilation, including Madau&Dickinson 2014 and some JWST results, I get poor agreement. This is shown in the figure below. What I get is also different if I compare by eye with Fig.8 in your paper. Now, there is of course the issue of the different IMF assumed for the FFB model and in the literature. In the plot that I am showing, I do not rescale anything for different IMF assumptions, because I am not sure what is assumed in the IMF model, while in the literature it is usually Salpeter IMF. I know that the FFB is based off UniverseMachine, which assumes a Chabrier IMF. But even if I rescale the FFB SFRD to a Salpeter IMF, I don't get a good agreement. Am I getting something wrong?

image

Thanks for your help.

syrte commented 5 months ago

@mShuntov thanks for reaching out. sorry for my slow reply as I am in travel. i will get back to you asap.

syrte commented 1 month ago

@mShuntov I realize that I haven't replied to you yet. I thought I could reply to you sooner, but it slipped my mind during successive intensive travels from mid-May to recently. My sincere apologies for my late reply!

  1. The apparent discrepancy is because of the scatter in stellar-to-halo mass relation. Though an SMF higher than scaled HMF looks wired, I believe the results are physical. With such a scatter, the low-mass halos can contribute to the high stellar masses.

  2. I won't apply the FFB model for z<5 which is beyond the scope of our original intention. The SFR itself should be reliable (our model goes back to UniverseMachine at low z anyway). The discrepancy at z<5 in your plot comes from the Mstar-M_UV relation (via the integration for M_UV<-17). Our model computes the UV luminosity of galaxies using the Mstar-M_UV relation at high-z in Yung et al 2024 (see eq 12 and footnote 6 in our paper). However, this relation is redshift dependent (we adopted a constant relation as the difference is small at high z), so likely non-applicable for z<5. This behavior can be imporved in future.

What I get is also different if I compare by eye with Fig.8 in your paper.

This is weird if true (though it looks consistent with our paper by my eyes). Would you mind checking the the consistency between the curve in your plot and our prediction data table in this repo https://github.com/syrte/ffb_predict/tree/main/tables/rho_sfr ?

I hope the above clarify things a bit. Let me you if you have more questions!